From c7205d2a73ea9033c3f4cc5b6f01ec3431fba348 Mon Sep 17 00:00:00 2001
From: Raoul Hodgson <raoul.hodgson@ed.ac.uk>
Date: Tue, 24 May 2022 14:30:26 +0100
Subject: [PATCH 001/146] Removed nvcc guards for json

---
 Grid/json/json.hpp                 | 4 +---
 Grid/serialisation/JSON_IO.cc      | 2 +-
 Grid/serialisation/Serialisation.h | 2 +-
 3 files changed, 3 insertions(+), 5 deletions(-)

diff --git a/Grid/json/json.hpp b/Grid/json/json.hpp
index 618aa7a1..5b5d2215 100644
--- a/Grid/json/json.hpp
+++ b/Grid/json/json.hpp
@@ -1,4 +1,3 @@
-#ifndef __NVCC__
 /*
     __ _____ _____ _____
  __|  |   __|     |   | |  JSON for Modern C++
@@ -18918,5 +18917,4 @@ inline nlohmann::json::json_pointer operator "" _json_pointer(const char* s, std
 #undef NLOHMANN_BASIC_JSON_TPL
 
 
-#endif
-#endif
+#endif
\ No newline at end of file
diff --git a/Grid/serialisation/JSON_IO.cc b/Grid/serialisation/JSON_IO.cc
index f2282099..e47a2969 100644
--- a/Grid/serialisation/JSON_IO.cc
+++ b/Grid/serialisation/JSON_IO.cc
@@ -26,7 +26,7 @@
     *************************************************************************************/
     /*  END LEGAL */
 #include <Grid/Grid.h>
-#if (!defined(GRID_CUDA)) && (!defined(GRID_HIP))
+#ifndef GRID_HIP
 
 NAMESPACE_BEGIN(Grid);
 
diff --git a/Grid/serialisation/Serialisation.h b/Grid/serialisation/Serialisation.h
index e14120af..4f395be2 100644
--- a/Grid/serialisation/Serialisation.h
+++ b/Grid/serialisation/Serialisation.h
@@ -36,7 +36,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 #include "BinaryIO.h"
 #include "TextIO.h"
 #include "XmlIO.h"
-#if (!defined(GRID_CUDA)) && (!defined(GRID_HIP))
+#ifndef GRID_HIP
 #include "JSON_IO.h"
 #endif
 

From 3ca0de1c409eb6c96ef95293d926915b8173346b Mon Sep 17 00:00:00 2001
From: Raoul Hodgson <raoul.hodgson@ed.ac.uk>
Date: Tue, 24 May 2022 14:37:33 +0100
Subject: [PATCH 002/146] Fix json write for vector<string>

---
 Grid/serialisation/JSON_IO.cc | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/Grid/serialisation/JSON_IO.cc b/Grid/serialisation/JSON_IO.cc
index e47a2969..c92203ba 100644
--- a/Grid/serialisation/JSON_IO.cc
+++ b/Grid/serialisation/JSON_IO.cc
@@ -82,7 +82,7 @@ void JSONWriter::writeDefault(const std::string &s,	const std::string &x)
   if (s.size())
     ss_ << "\""<< s << "\" : \"" << os.str() << "\" ," ;
   else
-    ss_ << os.str() << " ," ;
+    ss_ << "\""<< os.str() << "\" ," ;
 }
 
 // Reader implementation ///////////////////////////////////////////////////////

From e346154c5dfe8ba6ada0b11bcfc8cfba0cc57f6b Mon Sep 17 00:00:00 2001
From: Raoul Hodgson <raoul.hodgson@ed.ac.uk>
Date: Tue, 24 May 2022 15:47:01 +0100
Subject: [PATCH 003/146] Updated json CUDA compile guards

---
 Grid/serialisation/JSON_IO.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/Grid/serialisation/JSON_IO.h b/Grid/serialisation/JSON_IO.h
index b5255de3..090a36e1 100644
--- a/Grid/serialisation/JSON_IO.h
+++ b/Grid/serialisation/JSON_IO.h
@@ -54,7 +54,7 @@ namespace Grid
     void pop(void);
     template <typename U>
     void writeDefault(const std::string &s, const U &x);
-#ifdef __NVCC__
+#if defined(GRID_CUDA) || defined(GRID_HIP)
     void writeDefault(const std::string &s, const Grid::ComplexD &x) 
     { 
       std::complex<double> z(real(x),imag(x));
@@ -101,7 +101,7 @@ namespace Grid
     void readDefault(const std::string &s, std::vector<U> &output);
     template <typename U, typename P>
     void readDefault(const std::string &s, std::pair<U,P> &output);
-#ifdef __NVCC__
+#if defined(GRID_CUDA) || defined(GRID_HIP)
     void readDefault(const std::string &s, ComplexD &output)
     { 
       std::complex<double> z;

From da4daea57ab0afd2865f9282f13870e661e97be9 Mon Sep 17 00:00:00 2001
From: Raoul Hodgson <raoul.hodgson@ed.ac.uk>
Date: Tue, 24 May 2022 16:16:06 +0100
Subject: [PATCH 004/146] Updated json to latest release 3.10.5

---
 Grid/json/json.hpp | 24163 ++++++++++++++++++++++++-------------------
 1 file changed, 13667 insertions(+), 10496 deletions(-)

diff --git a/Grid/json/json.hpp b/Grid/json/json.hpp
index 5b5d2215..cb27e058 100644
--- a/Grid/json/json.hpp
+++ b/Grid/json/json.hpp
@@ -1,12 +1,12 @@
 /*
     __ _____ _____ _____
  __|  |   __|     |   | |  JSON for Modern C++
-|  |  |__   |  |  | | | |  version 3.2.0
+|  |  |__   |  |  | | | |  version 3.10.5
 |_____|_____|_____|_|___|  https://github.com/nlohmann/json
 
 Licensed under the MIT License <http://opensource.org/licenses/MIT>.
 SPDX-License-Identifier: MIT
-Copyright (c) 2013-2018 Niels Lohmann <http://nlohmann.me>.
+Copyright (c) 2013-2022 Niels Lohmann <http://nlohmann.me>.
 
 Permission is hereby  granted, free of charge, to any  person obtaining a copy
 of this software and associated  documentation files (the "Software"), to deal
@@ -27,588 +27,56 @@ OUT OF OR IN CONNECTION WITH THE SOFTWARE  OR THE USE OR OTHER DEALINGS IN THE
 SOFTWARE.
 */
 
-#ifndef NLOHMANN_JSON_HPP
-#define NLOHMANN_JSON_HPP
+/****************************************************************************\
+ * Note on documentation: The source files contain links to the online      *
+ * documentation of the public API at https://json.nlohmann.me. This URL    *
+ * contains the most recent documentation and should also be applicable to  *
+ * previous versions; documentation for deprecated functions is not         *
+ * removed, but marked deprecated. See "Generate documentation" section in  *
+ * file doc/README.md.                                                      *
+\****************************************************************************/
+
+#ifndef INCLUDE_NLOHMANN_JSON_HPP_
+#define INCLUDE_NLOHMANN_JSON_HPP_
 
 #define NLOHMANN_JSON_VERSION_MAJOR 3
-#define NLOHMANN_JSON_VERSION_MINOR 2
-#define NLOHMANN_JSON_VERSION_PATCH 0
+#define NLOHMANN_JSON_VERSION_MINOR 10
+#define NLOHMANN_JSON_VERSION_PATCH 5
 
 #include <algorithm> // all_of, find, for_each
-#include <cassert> // assert
-#include <ciso646> // and, not, or
 #include <cstddef> // nullptr_t, ptrdiff_t, size_t
 #include <functional> // hash, less
 #include <initializer_list> // initializer_list
-#include <iosfwd> // istream, ostream
-#include <iterator> // iterator_traits, random_access_iterator_tag
+#ifndef JSON_NO_IO
+    #include <iosfwd> // istream, ostream
+#endif  // JSON_NO_IO
+#include <iterator> // random_access_iterator_tag
+#include <memory> // unique_ptr
 #include <numeric> // accumulate
 #include <string> // string, stoi, to_string
 #include <utility> // declval, forward, move, pair, swap
-
-// #include <nlohmann/json_fwd.hpp>
-#ifndef NLOHMANN_JSON_FWD_HPP
-#define NLOHMANN_JSON_FWD_HPP
-
-#include <cstdint> // int64_t, uint64_t
-#include <map> // map
-#include <memory> // allocator
-#include <string> // string
 #include <vector> // vector
 
-/*!
-@brief namespace for Niels Lohmann
-@see https://github.com/nlohmann
-@since version 1.0.0
-*/
-namespace nlohmann
-{
-/*!
-@brief default JSONSerializer template argument
-
-This serializer ignores the template arguments and uses ADL
-([argument-dependent lookup](https://en.cppreference.com/w/cpp/language/adl))
-for serialization.
-*/
-template<typename T = void, typename SFINAE = void>
-struct adl_serializer;
-
-template<template<typename U, typename V, typename... Args> class ObjectType =
-         std::map,
-         template<typename U, typename... Args> class ArrayType = std::vector,
-         class StringType = std::string, class BooleanType = bool,
-         class NumberIntegerType = std::int64_t,
-         class NumberUnsignedType = std::uint64_t,
-         class NumberFloatType = double,
-         template<typename U> class AllocatorType = std::allocator,
-         template<typename T, typename SFINAE = void> class JSONSerializer =
-         adl_serializer>
-class basic_json;
-
-/*!
-@brief JSON Pointer
-
-A JSON pointer defines a string syntax for identifying a specific value
-within a JSON document. It can be used with functions `at` and
-`operator[]`. Furthermore, JSON pointers are the base for JSON patches.
-
-@sa [RFC 6901](https://tools.ietf.org/html/rfc6901)
-
-@since version 2.0.0
-*/
-template<typename BasicJsonType>
-class json_pointer;
-
-/*!
-@brief default JSON class
-
-This type is the default specialization of the @ref basic_json class which
-uses the standard template types.
-
-@since version 1.0.0
-*/
-using json = basic_json<>;
-}
-
-#endif
-
-// #include <nlohmann/detail/macro_scope.hpp>
-
-
-// This file contains all internal macro definitions
-// You MUST include macro_unscope.hpp at the end of json.hpp to undef all of them
-
-// exclude unsupported compilers
-#if !defined(JSON_SKIP_UNSUPPORTED_COMPILER_CHECK)
-    #if defined(__clang__)
-        #if (__clang_major__ * 10000 + __clang_minor__ * 100 + __clang_patchlevel__) < 30400
-            #error "unsupported Clang version - see https://github.com/nlohmann/json#supported-compilers"
-        #endif
-    #elif defined(__GNUC__) && !(defined(__ICC) || defined(__INTEL_COMPILER))
-        #if (__GNUC__ * 10000 + __GNUC_MINOR__ * 100 + __GNUC_PATCHLEVEL__) < 40800
-            #error "unsupported GCC version - see https://github.com/nlohmann/json#supported-compilers"
-        #endif
-    #endif
-#endif
-
-// disable float-equal warnings on GCC/clang
-#if defined(__clang__) || defined(__GNUC__) || defined(__GNUG__)
-    #pragma GCC diagnostic push
-    #pragma GCC diagnostic ignored "-Wfloat-equal"
-#endif
-
-// disable documentation warnings on clang
-#if defined(__clang__)
-    #pragma GCC diagnostic push
-    #pragma GCC diagnostic ignored "-Wdocumentation"
-#endif
-
-// allow for portable deprecation warnings
-#if defined(__clang__) || defined(__GNUC__) || defined(__GNUG__)
-    #define JSON_DEPRECATED __attribute__((deprecated))
-#elif defined(_MSC_VER)
-    #define JSON_DEPRECATED __declspec(deprecated)
-#else
-    #define JSON_DEPRECATED
-#endif
-
-// allow to disable exceptions
-#if (defined(__cpp_exceptions) || defined(__EXCEPTIONS) || defined(_CPPUNWIND)) && !defined(JSON_NOEXCEPTION)
-    #define JSON_THROW(exception) throw exception
-    #define JSON_TRY try
-    #define JSON_CATCH(exception) catch(exception)
-    #define JSON_INTERNAL_CATCH(exception) catch(exception)
-#else
-    #define JSON_THROW(exception) std::abort()
-    #define JSON_TRY if(true)
-    #define JSON_CATCH(exception) if(false)
-    #define JSON_INTERNAL_CATCH(exception) if(false)
-#endif
-
-// override exception macros
-#if defined(JSON_THROW_USER)
-    #undef JSON_THROW
-    #define JSON_THROW JSON_THROW_USER
-#endif
-#if defined(JSON_TRY_USER)
-    #undef JSON_TRY
-    #define JSON_TRY JSON_TRY_USER
-#endif
-#if defined(JSON_CATCH_USER)
-    #undef JSON_CATCH
-    #define JSON_CATCH JSON_CATCH_USER
-    #undef JSON_INTERNAL_CATCH
-    #define JSON_INTERNAL_CATCH JSON_CATCH_USER
-#endif
-#if defined(JSON_INTERNAL_CATCH_USER)
-    #undef JSON_INTERNAL_CATCH
-    #define JSON_INTERNAL_CATCH JSON_INTERNAL_CATCH_USER
-#endif
-
-// manual branch prediction
-#if defined(__clang__) || defined(__GNUC__) || defined(__GNUG__)
-    #define JSON_LIKELY(x)      __builtin_expect(!!(x), 1)
-    #define JSON_UNLIKELY(x)    __builtin_expect(!!(x), 0)
-#else
-    #define JSON_LIKELY(x)      x
-    #define JSON_UNLIKELY(x)    x
-#endif
-
-// C++ language standard detection
-#if (defined(__cplusplus) && __cplusplus >= 201703L) || (defined(_HAS_CXX17) && _HAS_CXX17 == 1) // fix for issue #464
-    #define JSON_HAS_CPP_17
-    #define JSON_HAS_CPP_14
-#elif (defined(__cplusplus) && __cplusplus >= 201402L) || (defined(_HAS_CXX14) && _HAS_CXX14 == 1)
-    #define JSON_HAS_CPP_14
-#endif
-
-// Ugly macros to avoid uglier copy-paste when specializing basic_json. They
-// may be removed in the future once the class is split.
-
-#define NLOHMANN_BASIC_JSON_TPL_DECLARATION                                \
-    template<template<typename, typename, typename...> class ObjectType,   \
-             template<typename, typename...> class ArrayType,              \
-             class StringType, class BooleanType, class NumberIntegerType, \
-             class NumberUnsignedType, class NumberFloatType,              \
-             template<typename> class AllocatorType,                       \
-             template<typename, typename = void> class JSONSerializer>
-
-#define NLOHMANN_BASIC_JSON_TPL                                            \
-    basic_json<ObjectType, ArrayType, StringType, BooleanType,             \
-    NumberIntegerType, NumberUnsignedType, NumberFloatType,                \
-    AllocatorType, JSONSerializer>
-
-// #include <nlohmann/detail/meta/cpp_future.hpp>
-
-
-#include <ciso646> // not
-#include <cstddef> // size_t
-#include <type_traits> // conditional, enable_if, false_type, integral_constant, is_constructible, is_integral, is_same, remove_cv, remove_reference, true_type
-
-namespace nlohmann
-{
-namespace detail
-{
-// alias templates to reduce boilerplate
-template<bool B, typename T = void>
-using enable_if_t = typename std::enable_if<B, T>::type;
-
-template<typename T>
-using uncvref_t = typename std::remove_cv<typename std::remove_reference<T>::type>::type;
-
-// implementation of C++14 index_sequence and affiliates
-// source: https://stackoverflow.com/a/32223343
-template<std::size_t... Ints>
-struct index_sequence
-{
-    using type = index_sequence;
-    using value_type = std::size_t;
-    static constexpr std::size_t size() noexcept
-    {
-        return sizeof...(Ints);
-    }
-};
-
-template<class Sequence1, class Sequence2>
-struct merge_and_renumber;
-
-template<std::size_t... I1, std::size_t... I2>
-struct merge_and_renumber<index_sequence<I1...>, index_sequence<I2...>>
-        : index_sequence < I1..., (sizeof...(I1) + I2)... > {};
-
-template<std::size_t N>
-struct make_index_sequence
-    : merge_and_renumber < typename make_index_sequence < N / 2 >::type,
-      typename make_index_sequence < N - N / 2 >::type > {};
-
-template<> struct make_index_sequence<0> : index_sequence<> {};
-template<> struct make_index_sequence<1> : index_sequence<0> {};
-
-template<typename... Ts>
-using index_sequence_for = make_index_sequence<sizeof...(Ts)>;
-
-// dispatch utility (taken from ranges-v3)
-template<unsigned N> struct priority_tag : priority_tag < N - 1 > {};
-template<> struct priority_tag<0> {};
-
-// taken from ranges-v3
-template<typename T>
-struct static_const
-{
-    static constexpr T value{};
-};
-
-template<typename T>
-constexpr T static_const<T>::value;
-}
-}
-
-// #include <nlohmann/detail/meta/type_traits.hpp>
-
-
-#include <ciso646> // not
-#include <limits> // numeric_limits
-#include <type_traits> // false_type, is_constructible, is_integral, is_same, true_type
-#include <utility> // declval
-
-// #include <nlohmann/json_fwd.hpp>
-
-// #include <nlohmann/detail/meta/cpp_future.hpp>
-
-// #include <nlohmann/detail/meta/detected.hpp>
+// #include <nlohmann/adl_serializer.hpp>
 
 
 #include <type_traits>
-
-// #include <nlohmann/detail/meta/void_t.hpp>
-
-
-namespace nlohmann
-{
-namespace detail
-{
-template <typename ...Ts> struct make_void
-{
-    using type = void;
-};
-template <typename ...Ts> using void_t = typename make_void<Ts...>::type;
-}
-}
-
-
-// http://en.cppreference.com/w/cpp/experimental/is_detected
-namespace nlohmann
-{
-namespace detail
-{
-struct nonesuch
-{
-    nonesuch() = delete;
-    ~nonesuch() = delete;
-    nonesuch(nonesuch const&) = delete;
-    void operator=(nonesuch const&) = delete;
-};
-
-template <class Default,
-          class AlwaysVoid,
-          template <class...> class Op,
-          class... Args>
-struct detector
-{
-    using value_t = std::false_type;
-    using type = Default;
-};
-
-template <class Default, template <class...> class Op, class... Args>
-struct detector<Default, void_t<Op<Args...>>, Op, Args...>
-{
-    using value_t = std::true_type;
-    using type = Op<Args...>;
-};
-
-template <template <class...> class Op, class... Args>
-using is_detected = typename detector<nonesuch, void, Op, Args...>::value_t;
-
-template <template <class...> class Op, class... Args>
-using detected_t = typename detector<nonesuch, void, Op, Args...>::type;
-
-template <class Default, template <class...> class Op, class... Args>
-using detected_or = detector<Default, void, Op, Args...>;
-
-template <class Default, template <class...> class Op, class... Args>
-using detected_or_t = typename detected_or<Default, Op, Args...>::type;
-
-template <class Expected, template <class...> class Op, class... Args>
-using is_detected_exact = std::is_same<Expected, detected_t<Op, Args...>>;
-
-template <class To, template <class...> class Op, class... Args>
-using is_detected_convertible =
-    std::is_convertible<detected_t<Op, Args...>, To>;
-}
-}
-
-// #include <nlohmann/detail/macro_scope.hpp>
-
-
-namespace nlohmann
-{
-/*!
-@brief detail namespace with internal helper functions
-
-This namespace collects functions that should not be exposed,
-implementations of some @ref basic_json methods, and meta-programming helpers.
-
-@since version 2.1.0
-*/
-namespace detail
-{
-/////////////
-// helpers //
-/////////////
-
-template<typename> struct is_basic_json : std::false_type {};
-
-NLOHMANN_BASIC_JSON_TPL_DECLARATION
-struct is_basic_json<NLOHMANN_BASIC_JSON_TPL> : std::true_type {};
-
-//////////////////////////
-// aliases for detected //
-//////////////////////////
-
-template <typename T>
-using mapped_type_t = typename T::mapped_type;
-
-template <typename T>
-using key_type_t = typename T::key_type;
-
-template <typename T>
-using value_type_t = typename T::value_type;
-
-template <typename T>
-using difference_type_t = typename T::difference_type;
-
-template <typename T>
-using pointer_t = typename T::pointer;
-
-template <typename T>
-using reference_t = typename T::reference;
-
-template <typename T>
-using iterator_category_t = typename T::iterator_category;
-
-template <typename T>
-using iterator_t = typename T::iterator;
-
-template <typename T, typename... Args>
-using to_json_function = decltype(T::to_json(std::declval<Args>()...));
-
-template <typename T, typename... Args>
-using from_json_function = decltype(T::from_json(std::declval<Args>()...));
-
-template <typename T, typename U>
-using get_template_function = decltype(std::declval<T>().template get<U>());
-
-///////////////////
-// is_ functions //
-///////////////////
-
-template <typename T, typename = void>
-struct is_iterator_traits : std::false_type {};
-
-template <typename T>
-struct is_iterator_traits<std::iterator_traits<T>>
-{
-  private:
-    using traits = std::iterator_traits<T>;
-
-  public:
-    static constexpr auto value =
-        is_detected<value_type_t, traits>::value &&
-        is_detected<difference_type_t, traits>::value &&
-        is_detected<pointer_t, traits>::value &&
-        is_detected<iterator_category_t, traits>::value &&
-        is_detected<reference_t, traits>::value;
-};
-
-// source: https://stackoverflow.com/a/37193089/4116453
-
-template <typename T, typename = void>
-struct is_complete_type : std::false_type {};
-
-template <typename T>
-struct is_complete_type<T, decltype(void(sizeof(T)))> : std::true_type {};
-
-template <typename BasicJsonType, typename CompatibleObjectType,
-          typename = void>
-struct is_compatible_object_type_impl : std::false_type {};
-
-template <typename BasicJsonType, typename CompatibleObjectType>
-struct is_compatible_object_type_impl <
-    BasicJsonType, CompatibleObjectType,
-    enable_if_t<is_detected<mapped_type_t, CompatibleObjectType>::value and
-    is_detected<key_type_t, CompatibleObjectType>::value >>
-{
-
-    using object_t = typename BasicJsonType::object_t;
-
-    // macOS's is_constructible does not play well with nonesuch...
-    static constexpr bool value =
-        std::is_constructible<typename object_t::key_type,
-        typename CompatibleObjectType::key_type>::value and
-        std::is_constructible<typename object_t::mapped_type,
-        typename CompatibleObjectType::mapped_type>::value;
-};
-
-template <typename BasicJsonType, typename CompatibleObjectType>
-struct is_compatible_object_type
-    : is_compatible_object_type_impl<BasicJsonType, CompatibleObjectType> {};
-
-template <typename BasicJsonType, typename CompatibleStringType,
-          typename = void>
-struct is_compatible_string_type_impl : std::false_type {};
-
-template <typename BasicJsonType, typename CompatibleStringType>
-struct is_compatible_string_type_impl <
-    BasicJsonType, CompatibleStringType,
-    enable_if_t<is_detected_exact<typename BasicJsonType::string_t::value_type,
-    value_type_t, CompatibleStringType>::value >>
-{
-    static constexpr auto value =
-        std::is_constructible<typename BasicJsonType::string_t, CompatibleStringType>::value;
-};
-
-template <typename BasicJsonType, typename CompatibleStringType>
-struct is_compatible_string_type
-    : is_compatible_string_type_impl<BasicJsonType, CompatibleStringType> {};
-
-template <typename BasicJsonType, typename CompatibleArrayType, typename = void>
-struct is_compatible_array_type_impl : std::false_type {};
-
-template <typename BasicJsonType, typename CompatibleArrayType>
-struct is_compatible_array_type_impl <
-    BasicJsonType, CompatibleArrayType,
-    enable_if_t<is_detected<value_type_t, CompatibleArrayType>::value and
-    is_detected<iterator_t, CompatibleArrayType>::value >>
-{
-    // This is needed because json_reverse_iterator has a ::iterator type...
-    // Therefore it is detected as a CompatibleArrayType.
-    // The real fix would be to have an Iterable concept.
-    static constexpr bool value = not is_iterator_traits<std::iterator_traits<CompatibleArrayType>>::value;
-};
-
-template <typename BasicJsonType, typename CompatibleArrayType>
-struct is_compatible_array_type
-    : is_compatible_array_type_impl<BasicJsonType, CompatibleArrayType> {};
-
-template <typename RealIntegerType, typename CompatibleNumberIntegerType,
-          typename = void>
-struct is_compatible_integer_type_impl : std::false_type {};
-
-template <typename RealIntegerType, typename CompatibleNumberIntegerType>
-struct is_compatible_integer_type_impl <
-    RealIntegerType, CompatibleNumberIntegerType,
-    enable_if_t<std::is_integral<RealIntegerType>::value and
-    std::is_integral<CompatibleNumberIntegerType>::value and
-    not std::is_same<bool, CompatibleNumberIntegerType>::value >>
-{
-    // is there an assert somewhere on overflows?
-    using RealLimits = std::numeric_limits<RealIntegerType>;
-    using CompatibleLimits = std::numeric_limits<CompatibleNumberIntegerType>;
-
-    static constexpr auto value =
-        std::is_constructible<RealIntegerType,
-        CompatibleNumberIntegerType>::value and
-        CompatibleLimits::is_integer and
-        RealLimits::is_signed == CompatibleLimits::is_signed;
-};
-
-template <typename RealIntegerType, typename CompatibleNumberIntegerType>
-struct is_compatible_integer_type
-    : is_compatible_integer_type_impl<RealIntegerType,
-      CompatibleNumberIntegerType> {};
-
-// trait checking if JSONSerializer<T>::from_json(json const&, udt&) exists
-template <typename BasicJsonType, typename T, typename = void>
-struct has_from_json : std::false_type {};
-
-template <typename BasicJsonType, typename T>
-struct has_from_json<BasicJsonType, T,
-           enable_if_t<not is_basic_json<T>::value>>
-{
-    using serializer = typename BasicJsonType::template json_serializer<T, void>;
-
-    static constexpr bool value =
-        is_detected_exact<void, from_json_function, serializer,
-        const BasicJsonType&, T&>::value;
-};
-
-// This trait checks if JSONSerializer<T>::from_json(json const&) exists
-// this overload is used for non-default-constructible user-defined-types
-template <typename BasicJsonType, typename T, typename = void>
-struct has_non_default_from_json : std::false_type {};
-
-template<typename BasicJsonType, typename T>
-struct has_non_default_from_json<BasicJsonType, T, enable_if_t<not is_basic_json<T>::value>>
-{
-    using serializer = typename BasicJsonType::template json_serializer<T, void>;
-
-    static constexpr bool value =
-        is_detected_exact<T, from_json_function, serializer,
-        const BasicJsonType&>::value;
-};
-
-// This trait checks if BasicJsonType::json_serializer<T>::to_json exists
-// Do not evaluate the trait when T is a basic_json type, to avoid template instantiation infinite recursion.
-template <typename BasicJsonType, typename T, typename = void>
-struct has_to_json : std::false_type {};
-
-template <typename BasicJsonType, typename T>
-struct has_to_json<BasicJsonType, T, enable_if_t<not is_basic_json<T>::value>>
-{
-    using serializer = typename BasicJsonType::template json_serializer<T, void>;
-
-    static constexpr bool value =
-        is_detected_exact<void, to_json_function, serializer, BasicJsonType&,
-        T>::value;
-};
-
-template <typename BasicJsonType, typename CompatibleType, typename = void>
-struct is_compatible_type_impl: std::false_type {};
-
-template <typename BasicJsonType, typename CompatibleType>
-struct is_compatible_type_impl <
-    BasicJsonType, CompatibleType,
-    enable_if_t<is_complete_type<CompatibleType>::value >>
-{
-    static constexpr bool value =
-        has_to_json<BasicJsonType, CompatibleType>::value;
-};
-
-template <typename BasicJsonType, typename CompatibleType>
-struct is_compatible_type
-    : is_compatible_type_impl<BasicJsonType, CompatibleType> {};
-}
-}
+#include <utility>
+
+// #include <nlohmann/detail/conversions/from_json.hpp>
+
+
+#include <algorithm> // transform
+#include <array> // array
+#include <forward_list> // forward_list
+#include <iterator> // inserter, front_inserter, end
+#include <map> // map
+#include <string> // string
+#include <tuple> // tuple, make_tuple
+#include <type_traits> // is_arithmetic, is_same, is_enum, underlying_type, is_convertible
+#include <unordered_map> // unordered_map
+#include <utility> // pair, declval
+#include <valarray> // valarray
 
 // #include <nlohmann/detail/exceptions.hpp>
 
@@ -616,339 +84,15 @@ struct is_compatible_type
 #include <exception> // exception
 #include <stdexcept> // runtime_error
 #include <string> // to_string
-
-namespace nlohmann
-{
-namespace detail
-{
-////////////////
-// exceptions //
-////////////////
-
-/*!
-@brief general exception of the @ref basic_json class
-
-This class is an extension of `std::exception` objects with a member @a id for
-exception ids. It is used as the base class for all exceptions thrown by the
-@ref basic_json class. This class can hence be used as "wildcard" to catch
-exceptions.
-
-Subclasses:
-- @ref parse_error for exceptions indicating a parse error
-- @ref invalid_iterator for exceptions indicating errors with iterators
-- @ref type_error for exceptions indicating executing a member function with
-                  a wrong type
-- @ref out_of_range for exceptions indicating access out of the defined range
-- @ref other_error for exceptions indicating other library errors
-
-@internal
-@note To have nothrow-copy-constructible exceptions, we internally use
-      `std::runtime_error` which can cope with arbitrary-length error messages.
-      Intermediate strings are built with static functions and then passed to
-      the actual constructor.
-@endinternal
-
-@liveexample{The following code shows how arbitrary library exceptions can be
-caught.,exception}
-
-@since version 3.0.0
-*/
-class exception : public std::exception
-{
-  public:
-    /// returns the explanatory string
-    const char* what() const noexcept override
-    {
-        return m.what();
-    }
-
-    /// the id of the exception
-    const int id;
-
-  protected:
-    exception(int id_, const char* what_arg) : id(id_), m(what_arg) {}
-
-    static std::string name(const std::string& ename, int id_)
-    {
-        return "[json.exception." + ename + "." + std::to_string(id_) + "] ";
-    }
-
-  private:
-    /// an exception object as storage for error messages
-    std::runtime_error m;
-};
-
-/*!
-@brief exception indicating a parse error
-
-This exception is thrown by the library when a parse error occurs. Parse errors
-can occur during the deserialization of JSON text, CBOR, MessagePack, as well
-as when using JSON Patch.
-
-Member @a byte holds the byte index of the last read character in the input
-file.
-
-Exceptions have ids 1xx.
-
-name / id                      | example message | description
------------------------------- | --------------- | -------------------------
-json.exception.parse_error.101 | parse error at 2: unexpected end of input; expected string literal | This error indicates a syntax error while deserializing a JSON text. The error message describes that an unexpected token (character) was encountered, and the member @a byte indicates the error position.
-json.exception.parse_error.102 | parse error at 14: missing or wrong low surrogate | JSON uses the `\uxxxx` format to describe Unicode characters. Code points above above 0xFFFF are split into two `\uxxxx` entries ("surrogate pairs"). This error indicates that the surrogate pair is incomplete or contains an invalid code point.
-json.exception.parse_error.103 | parse error: code points above 0x10FFFF are invalid | Unicode supports code points up to 0x10FFFF. Code points above 0x10FFFF are invalid.
-json.exception.parse_error.104 | parse error: JSON patch must be an array of objects | [RFC 6902](https://tools.ietf.org/html/rfc6902) requires a JSON Patch document to be a JSON document that represents an array of objects.
-json.exception.parse_error.105 | parse error: operation must have string member 'op' | An operation of a JSON Patch document must contain exactly one "op" member, whose value indicates the operation to perform. Its value must be one of "add", "remove", "replace", "move", "copy", or "test"; other values are errors.
-json.exception.parse_error.106 | parse error: array index '01' must not begin with '0' | An array index in a JSON Pointer ([RFC 6901](https://tools.ietf.org/html/rfc6901)) may be `0` or any number without a leading `0`.
-json.exception.parse_error.107 | parse error: JSON pointer must be empty or begin with '/' - was: 'foo' | A JSON Pointer must be a Unicode string containing a sequence of zero or more reference tokens, each prefixed by a `/` character.
-json.exception.parse_error.108 | parse error: escape character '~' must be followed with '0' or '1' | In a JSON Pointer, only `~0` and `~1` are valid escape sequences.
-json.exception.parse_error.109 | parse error: array index 'one' is not a number | A JSON Pointer array index must be a number.
-json.exception.parse_error.110 | parse error at 1: cannot read 2 bytes from vector | When parsing CBOR or MessagePack, the byte vector ends before the complete value has been read.
-json.exception.parse_error.112 | parse error at 1: error reading CBOR; last byte: 0xF8 | Not all types of CBOR or MessagePack are supported. This exception occurs if an unsupported byte was read.
-json.exception.parse_error.113 | parse error at 2: expected a CBOR string; last byte: 0x98 | While parsing a map key, a value that is not a string has been read.
-
-@note For an input with n bytes, 1 is the index of the first character and n+1
-      is the index of the terminating null byte or the end of file. This also
-      holds true when reading a byte vector (CBOR or MessagePack).
-
-@liveexample{The following code shows how a `parse_error` exception can be
-caught.,parse_error}
-
-@sa @ref exception for the base class of the library exceptions
-@sa @ref invalid_iterator for exceptions indicating errors with iterators
-@sa @ref type_error for exceptions indicating executing a member function with
-                    a wrong type
-@sa @ref out_of_range for exceptions indicating access out of the defined range
-@sa @ref other_error for exceptions indicating other library errors
-
-@since version 3.0.0
-*/
-class parse_error : public exception
-{
-  public:
-    /*!
-    @brief create a parse error exception
-    @param[in] id_       the id of the exception
-    @param[in] byte_     the byte index where the error occurred (or 0 if the
-                         position cannot be determined)
-    @param[in] what_arg  the explanatory string
-    @return parse_error object
-    */
-    static parse_error create(int id_, std::size_t byte_, const std::string& what_arg)
-    {
-        std::string w = exception::name("parse_error", id_) + "parse error" +
-                        (byte_ != 0 ? (" at " + std::to_string(byte_)) : "") +
-                        ": " + what_arg;
-        return parse_error(id_, byte_, w.c_str());
-    }
-
-    /*!
-    @brief byte index of the parse error
-
-    The byte index of the last read character in the input file.
-
-    @note For an input with n bytes, 1 is the index of the first character and
-          n+1 is the index of the terminating null byte or the end of file.
-          This also holds true when reading a byte vector (CBOR or MessagePack).
-    */
-    const std::size_t byte;
-
-  private:
-    parse_error(int id_, std::size_t byte_, const char* what_arg)
-        : exception(id_, what_arg), byte(byte_) {}
-};
-
-/*!
-@brief exception indicating errors with iterators
-
-This exception is thrown if iterators passed to a library function do not match
-the expected semantics.
-
-Exceptions have ids 2xx.
-
-name / id                           | example message | description
------------------------------------ | --------------- | -------------------------
-json.exception.invalid_iterator.201 | iterators are not compatible | The iterators passed to constructor @ref basic_json(InputIT first, InputIT last) are not compatible, meaning they do not belong to the same container. Therefore, the range (@a first, @a last) is invalid.
-json.exception.invalid_iterator.202 | iterator does not fit current value | In an erase or insert function, the passed iterator @a pos does not belong to the JSON value for which the function was called. It hence does not define a valid position for the deletion/insertion.
-json.exception.invalid_iterator.203 | iterators do not fit current value | Either iterator passed to function @ref erase(IteratorType first, IteratorType last) does not belong to the JSON value from which values shall be erased. It hence does not define a valid range to delete values from.
-json.exception.invalid_iterator.204 | iterators out of range | When an iterator range for a primitive type (number, boolean, or string) is passed to a constructor or an erase function, this range has to be exactly (@ref begin(), @ref end()), because this is the only way the single stored value is expressed. All other ranges are invalid.
-json.exception.invalid_iterator.205 | iterator out of range | When an iterator for a primitive type (number, boolean, or string) is passed to an erase function, the iterator has to be the @ref begin() iterator, because it is the only way to address the stored value. All other iterators are invalid.
-json.exception.invalid_iterator.206 | cannot construct with iterators from null | The iterators passed to constructor @ref basic_json(InputIT first, InputIT last) belong to a JSON null value and hence to not define a valid range.
-json.exception.invalid_iterator.207 | cannot use key() for non-object iterators | The key() member function can only be used on iterators belonging to a JSON object, because other types do not have a concept of a key.
-json.exception.invalid_iterator.208 | cannot use operator[] for object iterators | The operator[] to specify a concrete offset cannot be used on iterators belonging to a JSON object, because JSON objects are unordered.
-json.exception.invalid_iterator.209 | cannot use offsets with object iterators | The offset operators (+, -, +=, -=) cannot be used on iterators belonging to a JSON object, because JSON objects are unordered.
-json.exception.invalid_iterator.210 | iterators do not fit | The iterator range passed to the insert function are not compatible, meaning they do not belong to the same container. Therefore, the range (@a first, @a last) is invalid.
-json.exception.invalid_iterator.211 | passed iterators may not belong to container | The iterator range passed to the insert function must not be a subrange of the container to insert to.
-json.exception.invalid_iterator.212 | cannot compare iterators of different containers | When two iterators are compared, they must belong to the same container.
-json.exception.invalid_iterator.213 | cannot compare order of object iterators | The order of object iterators cannot be compared, because JSON objects are unordered.
-json.exception.invalid_iterator.214 | cannot get value | Cannot get value for iterator: Either the iterator belongs to a null value or it is an iterator to a primitive type (number, boolean, or string), but the iterator is different to @ref begin().
-
-@liveexample{The following code shows how an `invalid_iterator` exception can be
-caught.,invalid_iterator}
-
-@sa @ref exception for the base class of the library exceptions
-@sa @ref parse_error for exceptions indicating a parse error
-@sa @ref type_error for exceptions indicating executing a member function with
-                    a wrong type
-@sa @ref out_of_range for exceptions indicating access out of the defined range
-@sa @ref other_error for exceptions indicating other library errors
-
-@since version 3.0.0
-*/
-class invalid_iterator : public exception
-{
-  public:
-    static invalid_iterator create(int id_, const std::string& what_arg)
-    {
-        std::string w = exception::name("invalid_iterator", id_) + what_arg;
-        return invalid_iterator(id_, w.c_str());
-    }
-
-  private:
-    invalid_iterator(int id_, const char* what_arg)
-        : exception(id_, what_arg) {}
-};
-
-/*!
-@brief exception indicating executing a member function with a wrong type
-
-This exception is thrown in case of a type error; that is, a library function is
-executed on a JSON value whose type does not match the expected semantics.
-
-Exceptions have ids 3xx.
-
-name / id                     | example message | description
------------------------------ | --------------- | -------------------------
-json.exception.type_error.301 | cannot create object from initializer list | To create an object from an initializer list, the initializer list must consist only of a list of pairs whose first element is a string. When this constraint is violated, an array is created instead.
-json.exception.type_error.302 | type must be object, but is array | During implicit or explicit value conversion, the JSON type must be compatible to the target type. For instance, a JSON string can only be converted into string types, but not into numbers or boolean types.
-json.exception.type_error.303 | incompatible ReferenceType for get_ref, actual type is object | To retrieve a reference to a value stored in a @ref basic_json object with @ref get_ref, the type of the reference must match the value type. For instance, for a JSON array, the @a ReferenceType must be @ref array_t&.
-json.exception.type_error.304 | cannot use at() with string | The @ref at() member functions can only be executed for certain JSON types.
-json.exception.type_error.305 | cannot use operator[] with string | The @ref operator[] member functions can only be executed for certain JSON types.
-json.exception.type_error.306 | cannot use value() with string | The @ref value() member functions can only be executed for certain JSON types.
-json.exception.type_error.307 | cannot use erase() with string | The @ref erase() member functions can only be executed for certain JSON types.
-json.exception.type_error.308 | cannot use push_back() with string | The @ref push_back() and @ref operator+= member functions can only be executed for certain JSON types.
-json.exception.type_error.309 | cannot use insert() with | The @ref insert() member functions can only be executed for certain JSON types.
-json.exception.type_error.310 | cannot use swap() with number | The @ref swap() member functions can only be executed for certain JSON types.
-json.exception.type_error.311 | cannot use emplace_back() with string | The @ref emplace_back() member function can only be executed for certain JSON types.
-json.exception.type_error.312 | cannot use update() with string | The @ref update() member functions can only be executed for certain JSON types.
-json.exception.type_error.313 | invalid value to unflatten | The @ref unflatten function converts an object whose keys are JSON Pointers back into an arbitrary nested JSON value. The JSON Pointers must not overlap, because then the resulting value would not be well defined.
-json.exception.type_error.314 | only objects can be unflattened | The @ref unflatten function only works for an object whose keys are JSON Pointers.
-json.exception.type_error.315 | values in object must be primitive | The @ref unflatten function only works for an object whose keys are JSON Pointers and whose values are primitive.
-json.exception.type_error.316 | invalid UTF-8 byte at index 10: 0x7E | The @ref dump function only works with UTF-8 encoded strings; that is, if you assign a `std::string` to a JSON value, make sure it is UTF-8 encoded. |
-
-@liveexample{The following code shows how a `type_error` exception can be
-caught.,type_error}
-
-@sa @ref exception for the base class of the library exceptions
-@sa @ref parse_error for exceptions indicating a parse error
-@sa @ref invalid_iterator for exceptions indicating errors with iterators
-@sa @ref out_of_range for exceptions indicating access out of the defined range
-@sa @ref other_error for exceptions indicating other library errors
-
-@since version 3.0.0
-*/
-class type_error : public exception
-{
-  public:
-    static type_error create(int id_, const std::string& what_arg)
-    {
-        std::string w = exception::name("type_error", id_) + what_arg;
-        return type_error(id_, w.c_str());
-    }
-
-  private:
-    type_error(int id_, const char* what_arg) : exception(id_, what_arg) {}
-};
-
-/*!
-@brief exception indicating access out of the defined range
-
-This exception is thrown in case a library function is called on an input
-parameter that exceeds the expected range, for instance in case of array
-indices or nonexisting object keys.
-
-Exceptions have ids 4xx.
-
-name / id                       | example message | description
-------------------------------- | --------------- | -------------------------
-json.exception.out_of_range.401 | array index 3 is out of range | The provided array index @a i is larger than @a size-1.
-json.exception.out_of_range.402 | array index '-' (3) is out of range | The special array index `-` in a JSON Pointer never describes a valid element of the array, but the index past the end. That is, it can only be used to add elements at this position, but not to read it.
-json.exception.out_of_range.403 | key 'foo' not found | The provided key was not found in the JSON object.
-json.exception.out_of_range.404 | unresolved reference token 'foo' | A reference token in a JSON Pointer could not be resolved.
-json.exception.out_of_range.405 | JSON pointer has no parent | The JSON Patch operations 'remove' and 'add' can not be applied to the root element of the JSON value.
-json.exception.out_of_range.406 | number overflow parsing '10E1000' | A parsed number could not be stored as without changing it to NaN or INF.
-json.exception.out_of_range.407 | number overflow serializing '9223372036854775808' | UBJSON only supports integers numbers up to 9223372036854775807. |
-json.exception.out_of_range.408 | excessive array size: 8658170730974374167 | The size (following `#`) of an UBJSON array or object exceeds the maximal capacity. |
-
-@liveexample{The following code shows how an `out_of_range` exception can be
-caught.,out_of_range}
-
-@sa @ref exception for the base class of the library exceptions
-@sa @ref parse_error for exceptions indicating a parse error
-@sa @ref invalid_iterator for exceptions indicating errors with iterators
-@sa @ref type_error for exceptions indicating executing a member function with
-                    a wrong type
-@sa @ref other_error for exceptions indicating other library errors
-
-@since version 3.0.0
-*/
-class out_of_range : public exception
-{
-  public:
-    static out_of_range create(int id_, const std::string& what_arg)
-    {
-        std::string w = exception::name("out_of_range", id_) + what_arg;
-        return out_of_range(id_, w.c_str());
-    }
-
-  private:
-    out_of_range(int id_, const char* what_arg) : exception(id_, what_arg) {}
-};
-
-/*!
-@brief exception indicating other library errors
-
-This exception is thrown in case of errors that cannot be classified with the
-other exception types.
-
-Exceptions have ids 5xx.
-
-name / id                      | example message | description
------------------------------- | --------------- | -------------------------
-json.exception.other_error.501 | unsuccessful: {"op":"test","path":"/baz", "value":"bar"} | A JSON Patch operation 'test' failed. The unsuccessful operation is also printed.
-
-@sa @ref exception for the base class of the library exceptions
-@sa @ref parse_error for exceptions indicating a parse error
-@sa @ref invalid_iterator for exceptions indicating errors with iterators
-@sa @ref type_error for exceptions indicating executing a member function with
-                    a wrong type
-@sa @ref out_of_range for exceptions indicating access out of the defined range
-
-@liveexample{The following code shows how an `other_error` exception can be
-caught.,other_error}
-
-@since version 3.0.0
-*/
-class other_error : public exception
-{
-  public:
-    static other_error create(int id_, const std::string& what_arg)
-    {
-        std::string w = exception::name("other_error", id_) + what_arg;
-        return other_error(id_, w.c_str());
-    }
-
-  private:
-    other_error(int id_, const char* what_arg) : exception(id_, what_arg) {}
-};
-}
-}
+#include <vector> // vector
 
 // #include <nlohmann/detail/value_t.hpp>
 
 
 #include <array> // array
-#include <ciso646> // and
 #include <cstddef> // size_t
 #include <cstdint> // uint8_t
+#include <string> // string
 
 namespace nlohmann
 {
@@ -977,7 +121,7 @@ number_float), because the library distinguishes these three types for numbers:
 @ref basic_json::number_float_t is used for floating-point numbers or to
 approximate integers which do not fit in the limits of their respective type.
 
-@sa @ref basic_json::basic_json(const value_t value_type) -- create a JSON
+@sa see @ref basic_json::basic_json(const value_t value_type) -- create a JSON
 value with the default value for a given type
 
 @since version 1.0.0
@@ -992,61 +136,3696 @@ enum class value_t : std::uint8_t
     number_integer,   ///< number value (signed integer)
     number_unsigned,  ///< number value (unsigned integer)
     number_float,     ///< number value (floating-point)
-    discarded         ///< discarded by the the parser callback function
+    binary,           ///< binary array (ordered collection of bytes)
+    discarded         ///< discarded by the parser callback function
 };
 
 /*!
 @brief comparison operator for JSON types
 
 Returns an ordering that is similar to Python:
-- order: null < boolean < number < object < array < string
+- order: null < boolean < number < object < array < string < binary
 - furthermore, each type is not smaller than itself
 - discarded values are not comparable
+- binary is represented as a b"" string in python and directly comparable to a
+  string; however, making a binary array directly comparable with a string would
+  be surprising behavior in a JSON file.
 
 @since version 1.0.0
 */
 inline bool operator<(const value_t lhs, const value_t rhs) noexcept
 {
-    static constexpr std::array<std::uint8_t, 8> order = {{
+    static constexpr std::array<std::uint8_t, 9> order = {{
             0 /* null */, 3 /* object */, 4 /* array */, 5 /* string */,
-            1 /* boolean */, 2 /* integer */, 2 /* unsigned */, 2 /* float */
+            1 /* boolean */, 2 /* integer */, 2 /* unsigned */, 2 /* float */,
+            6 /* binary */
         }
     };
 
     const auto l_index = static_cast<std::size_t>(lhs);
     const auto r_index = static_cast<std::size_t>(rhs);
-    return l_index < order.size() and r_index < order.size() and order[l_index] < order[r_index];
-}
+    return l_index < order.size() && r_index < order.size() && order[l_index] < order[r_index];
 }
+}  // namespace detail
+}  // namespace nlohmann
+
+// #include <nlohmann/detail/string_escape.hpp>
+
+
+#include <string>
+// #include <nlohmann/detail/macro_scope.hpp>
+
+
+#include <utility> // declval, pair
+// #include <nlohmann/thirdparty/hedley/hedley.hpp>
+
+
+/* Hedley - https://nemequ.github.io/hedley
+ * Created by Evan Nemerson <evan@nemerson.com>
+ *
+ * To the extent possible under law, the author(s) have dedicated all
+ * copyright and related and neighboring rights to this software to
+ * the public domain worldwide. This software is distributed without
+ * any warranty.
+ *
+ * For details, see <http://creativecommons.org/publicdomain/zero/1.0/>.
+ * SPDX-License-Identifier: CC0-1.0
+ */
+
+#if !defined(JSON_HEDLEY_VERSION) || (JSON_HEDLEY_VERSION < 15)
+#if defined(JSON_HEDLEY_VERSION)
+    #undef JSON_HEDLEY_VERSION
+#endif
+#define JSON_HEDLEY_VERSION 15
+
+#if defined(JSON_HEDLEY_STRINGIFY_EX)
+    #undef JSON_HEDLEY_STRINGIFY_EX
+#endif
+#define JSON_HEDLEY_STRINGIFY_EX(x) #x
+
+#if defined(JSON_HEDLEY_STRINGIFY)
+    #undef JSON_HEDLEY_STRINGIFY
+#endif
+#define JSON_HEDLEY_STRINGIFY(x) JSON_HEDLEY_STRINGIFY_EX(x)
+
+#if defined(JSON_HEDLEY_CONCAT_EX)
+    #undef JSON_HEDLEY_CONCAT_EX
+#endif
+#define JSON_HEDLEY_CONCAT_EX(a,b) a##b
+
+#if defined(JSON_HEDLEY_CONCAT)
+    #undef JSON_HEDLEY_CONCAT
+#endif
+#define JSON_HEDLEY_CONCAT(a,b) JSON_HEDLEY_CONCAT_EX(a,b)
+
+#if defined(JSON_HEDLEY_CONCAT3_EX)
+    #undef JSON_HEDLEY_CONCAT3_EX
+#endif
+#define JSON_HEDLEY_CONCAT3_EX(a,b,c) a##b##c
+
+#if defined(JSON_HEDLEY_CONCAT3)
+    #undef JSON_HEDLEY_CONCAT3
+#endif
+#define JSON_HEDLEY_CONCAT3(a,b,c) JSON_HEDLEY_CONCAT3_EX(a,b,c)
+
+#if defined(JSON_HEDLEY_VERSION_ENCODE)
+    #undef JSON_HEDLEY_VERSION_ENCODE
+#endif
+#define JSON_HEDLEY_VERSION_ENCODE(major,minor,revision) (((major) * 1000000) + ((minor) * 1000) + (revision))
+
+#if defined(JSON_HEDLEY_VERSION_DECODE_MAJOR)
+    #undef JSON_HEDLEY_VERSION_DECODE_MAJOR
+#endif
+#define JSON_HEDLEY_VERSION_DECODE_MAJOR(version) ((version) / 1000000)
+
+#if defined(JSON_HEDLEY_VERSION_DECODE_MINOR)
+    #undef JSON_HEDLEY_VERSION_DECODE_MINOR
+#endif
+#define JSON_HEDLEY_VERSION_DECODE_MINOR(version) (((version) % 1000000) / 1000)
+
+#if defined(JSON_HEDLEY_VERSION_DECODE_REVISION)
+    #undef JSON_HEDLEY_VERSION_DECODE_REVISION
+#endif
+#define JSON_HEDLEY_VERSION_DECODE_REVISION(version) ((version) % 1000)
+
+#if defined(JSON_HEDLEY_GNUC_VERSION)
+    #undef JSON_HEDLEY_GNUC_VERSION
+#endif
+#if defined(__GNUC__) && defined(__GNUC_PATCHLEVEL__)
+    #define JSON_HEDLEY_GNUC_VERSION JSON_HEDLEY_VERSION_ENCODE(__GNUC__, __GNUC_MINOR__, __GNUC_PATCHLEVEL__)
+#elif defined(__GNUC__)
+    #define JSON_HEDLEY_GNUC_VERSION JSON_HEDLEY_VERSION_ENCODE(__GNUC__, __GNUC_MINOR__, 0)
+#endif
+
+#if defined(JSON_HEDLEY_GNUC_VERSION_CHECK)
+    #undef JSON_HEDLEY_GNUC_VERSION_CHECK
+#endif
+#if defined(JSON_HEDLEY_GNUC_VERSION)
+    #define JSON_HEDLEY_GNUC_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_GNUC_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch))
+#else
+    #define JSON_HEDLEY_GNUC_VERSION_CHECK(major,minor,patch) (0)
+#endif
+
+#if defined(JSON_HEDLEY_MSVC_VERSION)
+    #undef JSON_HEDLEY_MSVC_VERSION
+#endif
+#if defined(_MSC_FULL_VER) && (_MSC_FULL_VER >= 140000000) && !defined(__ICL)
+    #define JSON_HEDLEY_MSVC_VERSION JSON_HEDLEY_VERSION_ENCODE(_MSC_FULL_VER / 10000000, (_MSC_FULL_VER % 10000000) / 100000, (_MSC_FULL_VER % 100000) / 100)
+#elif defined(_MSC_FULL_VER) && !defined(__ICL)
+    #define JSON_HEDLEY_MSVC_VERSION JSON_HEDLEY_VERSION_ENCODE(_MSC_FULL_VER / 1000000, (_MSC_FULL_VER % 1000000) / 10000, (_MSC_FULL_VER % 10000) / 10)
+#elif defined(_MSC_VER) && !defined(__ICL)
+    #define JSON_HEDLEY_MSVC_VERSION JSON_HEDLEY_VERSION_ENCODE(_MSC_VER / 100, _MSC_VER % 100, 0)
+#endif
+
+#if defined(JSON_HEDLEY_MSVC_VERSION_CHECK)
+    #undef JSON_HEDLEY_MSVC_VERSION_CHECK
+#endif
+#if !defined(JSON_HEDLEY_MSVC_VERSION)
+    #define JSON_HEDLEY_MSVC_VERSION_CHECK(major,minor,patch) (0)
+#elif defined(_MSC_VER) && (_MSC_VER >= 1400)
+    #define JSON_HEDLEY_MSVC_VERSION_CHECK(major,minor,patch) (_MSC_FULL_VER >= ((major * 10000000) + (minor * 100000) + (patch)))
+#elif defined(_MSC_VER) && (_MSC_VER >= 1200)
+    #define JSON_HEDLEY_MSVC_VERSION_CHECK(major,minor,patch) (_MSC_FULL_VER >= ((major * 1000000) + (minor * 10000) + (patch)))
+#else
+    #define JSON_HEDLEY_MSVC_VERSION_CHECK(major,minor,patch) (_MSC_VER >= ((major * 100) + (minor)))
+#endif
+
+#if defined(JSON_HEDLEY_INTEL_VERSION)
+    #undef JSON_HEDLEY_INTEL_VERSION
+#endif
+#if defined(__INTEL_COMPILER) && defined(__INTEL_COMPILER_UPDATE) && !defined(__ICL)
+    #define JSON_HEDLEY_INTEL_VERSION JSON_HEDLEY_VERSION_ENCODE(__INTEL_COMPILER / 100, __INTEL_COMPILER % 100, __INTEL_COMPILER_UPDATE)
+#elif defined(__INTEL_COMPILER) && !defined(__ICL)
+    #define JSON_HEDLEY_INTEL_VERSION JSON_HEDLEY_VERSION_ENCODE(__INTEL_COMPILER / 100, __INTEL_COMPILER % 100, 0)
+#endif
+
+#if defined(JSON_HEDLEY_INTEL_VERSION_CHECK)
+    #undef JSON_HEDLEY_INTEL_VERSION_CHECK
+#endif
+#if defined(JSON_HEDLEY_INTEL_VERSION)
+    #define JSON_HEDLEY_INTEL_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_INTEL_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch))
+#else
+    #define JSON_HEDLEY_INTEL_VERSION_CHECK(major,minor,patch) (0)
+#endif
+
+#if defined(JSON_HEDLEY_INTEL_CL_VERSION)
+    #undef JSON_HEDLEY_INTEL_CL_VERSION
+#endif
+#if defined(__INTEL_COMPILER) && defined(__INTEL_COMPILER_UPDATE) && defined(__ICL)
+    #define JSON_HEDLEY_INTEL_CL_VERSION JSON_HEDLEY_VERSION_ENCODE(__INTEL_COMPILER, __INTEL_COMPILER_UPDATE, 0)
+#endif
+
+#if defined(JSON_HEDLEY_INTEL_CL_VERSION_CHECK)
+    #undef JSON_HEDLEY_INTEL_CL_VERSION_CHECK
+#endif
+#if defined(JSON_HEDLEY_INTEL_CL_VERSION)
+    #define JSON_HEDLEY_INTEL_CL_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_INTEL_CL_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch))
+#else
+    #define JSON_HEDLEY_INTEL_CL_VERSION_CHECK(major,minor,patch) (0)
+#endif
+
+#if defined(JSON_HEDLEY_PGI_VERSION)
+    #undef JSON_HEDLEY_PGI_VERSION
+#endif
+#if defined(__PGI) && defined(__PGIC__) && defined(__PGIC_MINOR__) && defined(__PGIC_PATCHLEVEL__)
+    #define JSON_HEDLEY_PGI_VERSION JSON_HEDLEY_VERSION_ENCODE(__PGIC__, __PGIC_MINOR__, __PGIC_PATCHLEVEL__)
+#endif
+
+#if defined(JSON_HEDLEY_PGI_VERSION_CHECK)
+    #undef JSON_HEDLEY_PGI_VERSION_CHECK
+#endif
+#if defined(JSON_HEDLEY_PGI_VERSION)
+    #define JSON_HEDLEY_PGI_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_PGI_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch))
+#else
+    #define JSON_HEDLEY_PGI_VERSION_CHECK(major,minor,patch) (0)
+#endif
+
+#if defined(JSON_HEDLEY_SUNPRO_VERSION)
+    #undef JSON_HEDLEY_SUNPRO_VERSION
+#endif
+#if defined(__SUNPRO_C) && (__SUNPRO_C > 0x1000)
+    #define JSON_HEDLEY_SUNPRO_VERSION JSON_HEDLEY_VERSION_ENCODE((((__SUNPRO_C >> 16) & 0xf) * 10) + ((__SUNPRO_C >> 12) & 0xf), (((__SUNPRO_C >> 8) & 0xf) * 10) + ((__SUNPRO_C >> 4) & 0xf), (__SUNPRO_C & 0xf) * 10)
+#elif defined(__SUNPRO_C)
+    #define JSON_HEDLEY_SUNPRO_VERSION JSON_HEDLEY_VERSION_ENCODE((__SUNPRO_C >> 8) & 0xf, (__SUNPRO_C >> 4) & 0xf, (__SUNPRO_C) & 0xf)
+#elif defined(__SUNPRO_CC) && (__SUNPRO_CC > 0x1000)
+    #define JSON_HEDLEY_SUNPRO_VERSION JSON_HEDLEY_VERSION_ENCODE((((__SUNPRO_CC >> 16) & 0xf) * 10) + ((__SUNPRO_CC >> 12) & 0xf), (((__SUNPRO_CC >> 8) & 0xf) * 10) + ((__SUNPRO_CC >> 4) & 0xf), (__SUNPRO_CC & 0xf) * 10)
+#elif defined(__SUNPRO_CC)
+    #define JSON_HEDLEY_SUNPRO_VERSION JSON_HEDLEY_VERSION_ENCODE((__SUNPRO_CC >> 8) & 0xf, (__SUNPRO_CC >> 4) & 0xf, (__SUNPRO_CC) & 0xf)
+#endif
+
+#if defined(JSON_HEDLEY_SUNPRO_VERSION_CHECK)
+    #undef JSON_HEDLEY_SUNPRO_VERSION_CHECK
+#endif
+#if defined(JSON_HEDLEY_SUNPRO_VERSION)
+    #define JSON_HEDLEY_SUNPRO_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_SUNPRO_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch))
+#else
+    #define JSON_HEDLEY_SUNPRO_VERSION_CHECK(major,minor,patch) (0)
+#endif
+
+#if defined(JSON_HEDLEY_EMSCRIPTEN_VERSION)
+    #undef JSON_HEDLEY_EMSCRIPTEN_VERSION
+#endif
+#if defined(__EMSCRIPTEN__)
+    #define JSON_HEDLEY_EMSCRIPTEN_VERSION JSON_HEDLEY_VERSION_ENCODE(__EMSCRIPTEN_major__, __EMSCRIPTEN_minor__, __EMSCRIPTEN_tiny__)
+#endif
+
+#if defined(JSON_HEDLEY_EMSCRIPTEN_VERSION_CHECK)
+    #undef JSON_HEDLEY_EMSCRIPTEN_VERSION_CHECK
+#endif
+#if defined(JSON_HEDLEY_EMSCRIPTEN_VERSION)
+    #define JSON_HEDLEY_EMSCRIPTEN_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_EMSCRIPTEN_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch))
+#else
+    #define JSON_HEDLEY_EMSCRIPTEN_VERSION_CHECK(major,minor,patch) (0)
+#endif
+
+#if defined(JSON_HEDLEY_ARM_VERSION)
+    #undef JSON_HEDLEY_ARM_VERSION
+#endif
+#if defined(__CC_ARM) && defined(__ARMCOMPILER_VERSION)
+    #define JSON_HEDLEY_ARM_VERSION JSON_HEDLEY_VERSION_ENCODE(__ARMCOMPILER_VERSION / 1000000, (__ARMCOMPILER_VERSION % 1000000) / 10000, (__ARMCOMPILER_VERSION % 10000) / 100)
+#elif defined(__CC_ARM) && defined(__ARMCC_VERSION)
+    #define JSON_HEDLEY_ARM_VERSION JSON_HEDLEY_VERSION_ENCODE(__ARMCC_VERSION / 1000000, (__ARMCC_VERSION % 1000000) / 10000, (__ARMCC_VERSION % 10000) / 100)
+#endif
+
+#if defined(JSON_HEDLEY_ARM_VERSION_CHECK)
+    #undef JSON_HEDLEY_ARM_VERSION_CHECK
+#endif
+#if defined(JSON_HEDLEY_ARM_VERSION)
+    #define JSON_HEDLEY_ARM_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_ARM_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch))
+#else
+    #define JSON_HEDLEY_ARM_VERSION_CHECK(major,minor,patch) (0)
+#endif
+
+#if defined(JSON_HEDLEY_IBM_VERSION)
+    #undef JSON_HEDLEY_IBM_VERSION
+#endif
+#if defined(__ibmxl__)
+    #define JSON_HEDLEY_IBM_VERSION JSON_HEDLEY_VERSION_ENCODE(__ibmxl_version__, __ibmxl_release__, __ibmxl_modification__)
+#elif defined(__xlC__) && defined(__xlC_ver__)
+    #define JSON_HEDLEY_IBM_VERSION JSON_HEDLEY_VERSION_ENCODE(__xlC__ >> 8, __xlC__ & 0xff, (__xlC_ver__ >> 8) & 0xff)
+#elif defined(__xlC__)
+    #define JSON_HEDLEY_IBM_VERSION JSON_HEDLEY_VERSION_ENCODE(__xlC__ >> 8, __xlC__ & 0xff, 0)
+#endif
+
+#if defined(JSON_HEDLEY_IBM_VERSION_CHECK)
+    #undef JSON_HEDLEY_IBM_VERSION_CHECK
+#endif
+#if defined(JSON_HEDLEY_IBM_VERSION)
+    #define JSON_HEDLEY_IBM_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_IBM_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch))
+#else
+    #define JSON_HEDLEY_IBM_VERSION_CHECK(major,minor,patch) (0)
+#endif
+
+#if defined(JSON_HEDLEY_TI_VERSION)
+    #undef JSON_HEDLEY_TI_VERSION
+#endif
+#if \
+    defined(__TI_COMPILER_VERSION__) && \
+    ( \
+      defined(__TMS470__) || defined(__TI_ARM__) || \
+      defined(__MSP430__) || \
+      defined(__TMS320C2000__) \
+    )
+#if (__TI_COMPILER_VERSION__ >= 16000000)
+    #define JSON_HEDLEY_TI_VERSION JSON_HEDLEY_VERSION_ENCODE(__TI_COMPILER_VERSION__ / 1000000, (__TI_COMPILER_VERSION__ % 1000000) / 1000, (__TI_COMPILER_VERSION__ % 1000))
+#endif
+#endif
+
+#if defined(JSON_HEDLEY_TI_VERSION_CHECK)
+    #undef JSON_HEDLEY_TI_VERSION_CHECK
+#endif
+#if defined(JSON_HEDLEY_TI_VERSION)
+    #define JSON_HEDLEY_TI_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_TI_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch))
+#else
+    #define JSON_HEDLEY_TI_VERSION_CHECK(major,minor,patch) (0)
+#endif
+
+#if defined(JSON_HEDLEY_TI_CL2000_VERSION)
+    #undef JSON_HEDLEY_TI_CL2000_VERSION
+#endif
+#if defined(__TI_COMPILER_VERSION__) && defined(__TMS320C2000__)
+    #define JSON_HEDLEY_TI_CL2000_VERSION JSON_HEDLEY_VERSION_ENCODE(__TI_COMPILER_VERSION__ / 1000000, (__TI_COMPILER_VERSION__ % 1000000) / 1000, (__TI_COMPILER_VERSION__ % 1000))
+#endif
+
+#if defined(JSON_HEDLEY_TI_CL2000_VERSION_CHECK)
+    #undef JSON_HEDLEY_TI_CL2000_VERSION_CHECK
+#endif
+#if defined(JSON_HEDLEY_TI_CL2000_VERSION)
+    #define JSON_HEDLEY_TI_CL2000_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_TI_CL2000_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch))
+#else
+    #define JSON_HEDLEY_TI_CL2000_VERSION_CHECK(major,minor,patch) (0)
+#endif
+
+#if defined(JSON_HEDLEY_TI_CL430_VERSION)
+    #undef JSON_HEDLEY_TI_CL430_VERSION
+#endif
+#if defined(__TI_COMPILER_VERSION__) && defined(__MSP430__)
+    #define JSON_HEDLEY_TI_CL430_VERSION JSON_HEDLEY_VERSION_ENCODE(__TI_COMPILER_VERSION__ / 1000000, (__TI_COMPILER_VERSION__ % 1000000) / 1000, (__TI_COMPILER_VERSION__ % 1000))
+#endif
+
+#if defined(JSON_HEDLEY_TI_CL430_VERSION_CHECK)
+    #undef JSON_HEDLEY_TI_CL430_VERSION_CHECK
+#endif
+#if defined(JSON_HEDLEY_TI_CL430_VERSION)
+    #define JSON_HEDLEY_TI_CL430_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_TI_CL430_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch))
+#else
+    #define JSON_HEDLEY_TI_CL430_VERSION_CHECK(major,minor,patch) (0)
+#endif
+
+#if defined(JSON_HEDLEY_TI_ARMCL_VERSION)
+    #undef JSON_HEDLEY_TI_ARMCL_VERSION
+#endif
+#if defined(__TI_COMPILER_VERSION__) && (defined(__TMS470__) || defined(__TI_ARM__))
+    #define JSON_HEDLEY_TI_ARMCL_VERSION JSON_HEDLEY_VERSION_ENCODE(__TI_COMPILER_VERSION__ / 1000000, (__TI_COMPILER_VERSION__ % 1000000) / 1000, (__TI_COMPILER_VERSION__ % 1000))
+#endif
+
+#if defined(JSON_HEDLEY_TI_ARMCL_VERSION_CHECK)
+    #undef JSON_HEDLEY_TI_ARMCL_VERSION_CHECK
+#endif
+#if defined(JSON_HEDLEY_TI_ARMCL_VERSION)
+    #define JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_TI_ARMCL_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch))
+#else
+    #define JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(major,minor,patch) (0)
+#endif
+
+#if defined(JSON_HEDLEY_TI_CL6X_VERSION)
+    #undef JSON_HEDLEY_TI_CL6X_VERSION
+#endif
+#if defined(__TI_COMPILER_VERSION__) && defined(__TMS320C6X__)
+    #define JSON_HEDLEY_TI_CL6X_VERSION JSON_HEDLEY_VERSION_ENCODE(__TI_COMPILER_VERSION__ / 1000000, (__TI_COMPILER_VERSION__ % 1000000) / 1000, (__TI_COMPILER_VERSION__ % 1000))
+#endif
+
+#if defined(JSON_HEDLEY_TI_CL6X_VERSION_CHECK)
+    #undef JSON_HEDLEY_TI_CL6X_VERSION_CHECK
+#endif
+#if defined(JSON_HEDLEY_TI_CL6X_VERSION)
+    #define JSON_HEDLEY_TI_CL6X_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_TI_CL6X_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch))
+#else
+    #define JSON_HEDLEY_TI_CL6X_VERSION_CHECK(major,minor,patch) (0)
+#endif
+
+#if defined(JSON_HEDLEY_TI_CL7X_VERSION)
+    #undef JSON_HEDLEY_TI_CL7X_VERSION
+#endif
+#if defined(__TI_COMPILER_VERSION__) && defined(__C7000__)
+    #define JSON_HEDLEY_TI_CL7X_VERSION JSON_HEDLEY_VERSION_ENCODE(__TI_COMPILER_VERSION__ / 1000000, (__TI_COMPILER_VERSION__ % 1000000) / 1000, (__TI_COMPILER_VERSION__ % 1000))
+#endif
+
+#if defined(JSON_HEDLEY_TI_CL7X_VERSION_CHECK)
+    #undef JSON_HEDLEY_TI_CL7X_VERSION_CHECK
+#endif
+#if defined(JSON_HEDLEY_TI_CL7X_VERSION)
+    #define JSON_HEDLEY_TI_CL7X_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_TI_CL7X_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch))
+#else
+    #define JSON_HEDLEY_TI_CL7X_VERSION_CHECK(major,minor,patch) (0)
+#endif
+
+#if defined(JSON_HEDLEY_TI_CLPRU_VERSION)
+    #undef JSON_HEDLEY_TI_CLPRU_VERSION
+#endif
+#if defined(__TI_COMPILER_VERSION__) && defined(__PRU__)
+    #define JSON_HEDLEY_TI_CLPRU_VERSION JSON_HEDLEY_VERSION_ENCODE(__TI_COMPILER_VERSION__ / 1000000, (__TI_COMPILER_VERSION__ % 1000000) / 1000, (__TI_COMPILER_VERSION__ % 1000))
+#endif
+
+#if defined(JSON_HEDLEY_TI_CLPRU_VERSION_CHECK)
+    #undef JSON_HEDLEY_TI_CLPRU_VERSION_CHECK
+#endif
+#if defined(JSON_HEDLEY_TI_CLPRU_VERSION)
+    #define JSON_HEDLEY_TI_CLPRU_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_TI_CLPRU_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch))
+#else
+    #define JSON_HEDLEY_TI_CLPRU_VERSION_CHECK(major,minor,patch) (0)
+#endif
+
+#if defined(JSON_HEDLEY_CRAY_VERSION)
+    #undef JSON_HEDLEY_CRAY_VERSION
+#endif
+#if defined(_CRAYC)
+    #if defined(_RELEASE_PATCHLEVEL)
+        #define JSON_HEDLEY_CRAY_VERSION JSON_HEDLEY_VERSION_ENCODE(_RELEASE_MAJOR, _RELEASE_MINOR, _RELEASE_PATCHLEVEL)
+    #else
+        #define JSON_HEDLEY_CRAY_VERSION JSON_HEDLEY_VERSION_ENCODE(_RELEASE_MAJOR, _RELEASE_MINOR, 0)
+    #endif
+#endif
+
+#if defined(JSON_HEDLEY_CRAY_VERSION_CHECK)
+    #undef JSON_HEDLEY_CRAY_VERSION_CHECK
+#endif
+#if defined(JSON_HEDLEY_CRAY_VERSION)
+    #define JSON_HEDLEY_CRAY_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_CRAY_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch))
+#else
+    #define JSON_HEDLEY_CRAY_VERSION_CHECK(major,minor,patch) (0)
+#endif
+
+#if defined(JSON_HEDLEY_IAR_VERSION)
+    #undef JSON_HEDLEY_IAR_VERSION
+#endif
+#if defined(__IAR_SYSTEMS_ICC__)
+    #if __VER__ > 1000
+        #define JSON_HEDLEY_IAR_VERSION JSON_HEDLEY_VERSION_ENCODE((__VER__ / 1000000), ((__VER__ / 1000) % 1000), (__VER__ % 1000))
+    #else
+        #define JSON_HEDLEY_IAR_VERSION JSON_HEDLEY_VERSION_ENCODE(__VER__ / 100, __VER__ % 100, 0)
+    #endif
+#endif
+
+#if defined(JSON_HEDLEY_IAR_VERSION_CHECK)
+    #undef JSON_HEDLEY_IAR_VERSION_CHECK
+#endif
+#if defined(JSON_HEDLEY_IAR_VERSION)
+    #define JSON_HEDLEY_IAR_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_IAR_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch))
+#else
+    #define JSON_HEDLEY_IAR_VERSION_CHECK(major,minor,patch) (0)
+#endif
+
+#if defined(JSON_HEDLEY_TINYC_VERSION)
+    #undef JSON_HEDLEY_TINYC_VERSION
+#endif
+#if defined(__TINYC__)
+    #define JSON_HEDLEY_TINYC_VERSION JSON_HEDLEY_VERSION_ENCODE(__TINYC__ / 1000, (__TINYC__ / 100) % 10, __TINYC__ % 100)
+#endif
+
+#if defined(JSON_HEDLEY_TINYC_VERSION_CHECK)
+    #undef JSON_HEDLEY_TINYC_VERSION_CHECK
+#endif
+#if defined(JSON_HEDLEY_TINYC_VERSION)
+    #define JSON_HEDLEY_TINYC_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_TINYC_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch))
+#else
+    #define JSON_HEDLEY_TINYC_VERSION_CHECK(major,minor,patch) (0)
+#endif
+
+#if defined(JSON_HEDLEY_DMC_VERSION)
+    #undef JSON_HEDLEY_DMC_VERSION
+#endif
+#if defined(__DMC__)
+    #define JSON_HEDLEY_DMC_VERSION JSON_HEDLEY_VERSION_ENCODE(__DMC__ >> 8, (__DMC__ >> 4) & 0xf, __DMC__ & 0xf)
+#endif
+
+#if defined(JSON_HEDLEY_DMC_VERSION_CHECK)
+    #undef JSON_HEDLEY_DMC_VERSION_CHECK
+#endif
+#if defined(JSON_HEDLEY_DMC_VERSION)
+    #define JSON_HEDLEY_DMC_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_DMC_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch))
+#else
+    #define JSON_HEDLEY_DMC_VERSION_CHECK(major,minor,patch) (0)
+#endif
+
+#if defined(JSON_HEDLEY_COMPCERT_VERSION)
+    #undef JSON_HEDLEY_COMPCERT_VERSION
+#endif
+#if defined(__COMPCERT_VERSION__)
+    #define JSON_HEDLEY_COMPCERT_VERSION JSON_HEDLEY_VERSION_ENCODE(__COMPCERT_VERSION__ / 10000, (__COMPCERT_VERSION__ / 100) % 100, __COMPCERT_VERSION__ % 100)
+#endif
+
+#if defined(JSON_HEDLEY_COMPCERT_VERSION_CHECK)
+    #undef JSON_HEDLEY_COMPCERT_VERSION_CHECK
+#endif
+#if defined(JSON_HEDLEY_COMPCERT_VERSION)
+    #define JSON_HEDLEY_COMPCERT_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_COMPCERT_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch))
+#else
+    #define JSON_HEDLEY_COMPCERT_VERSION_CHECK(major,minor,patch) (0)
+#endif
+
+#if defined(JSON_HEDLEY_PELLES_VERSION)
+    #undef JSON_HEDLEY_PELLES_VERSION
+#endif
+#if defined(__POCC__)
+    #define JSON_HEDLEY_PELLES_VERSION JSON_HEDLEY_VERSION_ENCODE(__POCC__ / 100, __POCC__ % 100, 0)
+#endif
+
+#if defined(JSON_HEDLEY_PELLES_VERSION_CHECK)
+    #undef JSON_HEDLEY_PELLES_VERSION_CHECK
+#endif
+#if defined(JSON_HEDLEY_PELLES_VERSION)
+    #define JSON_HEDLEY_PELLES_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_PELLES_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch))
+#else
+    #define JSON_HEDLEY_PELLES_VERSION_CHECK(major,minor,patch) (0)
+#endif
+
+#if defined(JSON_HEDLEY_MCST_LCC_VERSION)
+    #undef JSON_HEDLEY_MCST_LCC_VERSION
+#endif
+#if defined(__LCC__) && defined(__LCC_MINOR__)
+    #define JSON_HEDLEY_MCST_LCC_VERSION JSON_HEDLEY_VERSION_ENCODE(__LCC__ / 100, __LCC__ % 100, __LCC_MINOR__)
+#endif
+
+#if defined(JSON_HEDLEY_MCST_LCC_VERSION_CHECK)
+    #undef JSON_HEDLEY_MCST_LCC_VERSION_CHECK
+#endif
+#if defined(JSON_HEDLEY_MCST_LCC_VERSION)
+    #define JSON_HEDLEY_MCST_LCC_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_MCST_LCC_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch))
+#else
+    #define JSON_HEDLEY_MCST_LCC_VERSION_CHECK(major,minor,patch) (0)
+#endif
+
+#if defined(JSON_HEDLEY_GCC_VERSION)
+    #undef JSON_HEDLEY_GCC_VERSION
+#endif
+#if \
+    defined(JSON_HEDLEY_GNUC_VERSION) && \
+    !defined(__clang__) && \
+    !defined(JSON_HEDLEY_INTEL_VERSION) && \
+    !defined(JSON_HEDLEY_PGI_VERSION) && \
+    !defined(JSON_HEDLEY_ARM_VERSION) && \
+    !defined(JSON_HEDLEY_CRAY_VERSION) && \
+    !defined(JSON_HEDLEY_TI_VERSION) && \
+    !defined(JSON_HEDLEY_TI_ARMCL_VERSION) && \
+    !defined(JSON_HEDLEY_TI_CL430_VERSION) && \
+    !defined(JSON_HEDLEY_TI_CL2000_VERSION) && \
+    !defined(JSON_HEDLEY_TI_CL6X_VERSION) && \
+    !defined(JSON_HEDLEY_TI_CL7X_VERSION) && \
+    !defined(JSON_HEDLEY_TI_CLPRU_VERSION) && \
+    !defined(__COMPCERT__) && \
+    !defined(JSON_HEDLEY_MCST_LCC_VERSION)
+    #define JSON_HEDLEY_GCC_VERSION JSON_HEDLEY_GNUC_VERSION
+#endif
+
+#if defined(JSON_HEDLEY_GCC_VERSION_CHECK)
+    #undef JSON_HEDLEY_GCC_VERSION_CHECK
+#endif
+#if defined(JSON_HEDLEY_GCC_VERSION)
+    #define JSON_HEDLEY_GCC_VERSION_CHECK(major,minor,patch) (JSON_HEDLEY_GCC_VERSION >= JSON_HEDLEY_VERSION_ENCODE(major, minor, patch))
+#else
+    #define JSON_HEDLEY_GCC_VERSION_CHECK(major,minor,patch) (0)
+#endif
+
+#if defined(JSON_HEDLEY_HAS_ATTRIBUTE)
+    #undef JSON_HEDLEY_HAS_ATTRIBUTE
+#endif
+#if \
+  defined(__has_attribute) && \
+  ( \
+    (!defined(JSON_HEDLEY_IAR_VERSION) || JSON_HEDLEY_IAR_VERSION_CHECK(8,5,9)) \
+  )
+#  define JSON_HEDLEY_HAS_ATTRIBUTE(attribute) __has_attribute(attribute)
+#else
+#  define JSON_HEDLEY_HAS_ATTRIBUTE(attribute) (0)
+#endif
+
+#if defined(JSON_HEDLEY_GNUC_HAS_ATTRIBUTE)
+    #undef JSON_HEDLEY_GNUC_HAS_ATTRIBUTE
+#endif
+#if defined(__has_attribute)
+    #define JSON_HEDLEY_GNUC_HAS_ATTRIBUTE(attribute,major,minor,patch) JSON_HEDLEY_HAS_ATTRIBUTE(attribute)
+#else
+    #define JSON_HEDLEY_GNUC_HAS_ATTRIBUTE(attribute,major,minor,patch) JSON_HEDLEY_GNUC_VERSION_CHECK(major,minor,patch)
+#endif
+
+#if defined(JSON_HEDLEY_GCC_HAS_ATTRIBUTE)
+    #undef JSON_HEDLEY_GCC_HAS_ATTRIBUTE
+#endif
+#if defined(__has_attribute)
+    #define JSON_HEDLEY_GCC_HAS_ATTRIBUTE(attribute,major,minor,patch) JSON_HEDLEY_HAS_ATTRIBUTE(attribute)
+#else
+    #define JSON_HEDLEY_GCC_HAS_ATTRIBUTE(attribute,major,minor,patch) JSON_HEDLEY_GCC_VERSION_CHECK(major,minor,patch)
+#endif
+
+#if defined(JSON_HEDLEY_HAS_CPP_ATTRIBUTE)
+    #undef JSON_HEDLEY_HAS_CPP_ATTRIBUTE
+#endif
+#if \
+    defined(__has_cpp_attribute) && \
+    defined(__cplusplus) && \
+    (!defined(JSON_HEDLEY_SUNPRO_VERSION) || JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,15,0))
+    #define JSON_HEDLEY_HAS_CPP_ATTRIBUTE(attribute) __has_cpp_attribute(attribute)
+#else
+    #define JSON_HEDLEY_HAS_CPP_ATTRIBUTE(attribute) (0)
+#endif
+
+#if defined(JSON_HEDLEY_HAS_CPP_ATTRIBUTE_NS)
+    #undef JSON_HEDLEY_HAS_CPP_ATTRIBUTE_NS
+#endif
+#if !defined(__cplusplus) || !defined(__has_cpp_attribute)
+    #define JSON_HEDLEY_HAS_CPP_ATTRIBUTE_NS(ns,attribute) (0)
+#elif \
+    !defined(JSON_HEDLEY_PGI_VERSION) && \
+    !defined(JSON_HEDLEY_IAR_VERSION) && \
+    (!defined(JSON_HEDLEY_SUNPRO_VERSION) || JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,15,0)) && \
+    (!defined(JSON_HEDLEY_MSVC_VERSION) || JSON_HEDLEY_MSVC_VERSION_CHECK(19,20,0))
+    #define JSON_HEDLEY_HAS_CPP_ATTRIBUTE_NS(ns,attribute) JSON_HEDLEY_HAS_CPP_ATTRIBUTE(ns::attribute)
+#else
+    #define JSON_HEDLEY_HAS_CPP_ATTRIBUTE_NS(ns,attribute) (0)
+#endif
+
+#if defined(JSON_HEDLEY_GNUC_HAS_CPP_ATTRIBUTE)
+    #undef JSON_HEDLEY_GNUC_HAS_CPP_ATTRIBUTE
+#endif
+#if defined(__has_cpp_attribute) && defined(__cplusplus)
+    #define JSON_HEDLEY_GNUC_HAS_CPP_ATTRIBUTE(attribute,major,minor,patch) __has_cpp_attribute(attribute)
+#else
+    #define JSON_HEDLEY_GNUC_HAS_CPP_ATTRIBUTE(attribute,major,minor,patch) JSON_HEDLEY_GNUC_VERSION_CHECK(major,minor,patch)
+#endif
+
+#if defined(JSON_HEDLEY_GCC_HAS_CPP_ATTRIBUTE)
+    #undef JSON_HEDLEY_GCC_HAS_CPP_ATTRIBUTE
+#endif
+#if defined(__has_cpp_attribute) && defined(__cplusplus)
+    #define JSON_HEDLEY_GCC_HAS_CPP_ATTRIBUTE(attribute,major,minor,patch) __has_cpp_attribute(attribute)
+#else
+    #define JSON_HEDLEY_GCC_HAS_CPP_ATTRIBUTE(attribute,major,minor,patch) JSON_HEDLEY_GCC_VERSION_CHECK(major,minor,patch)
+#endif
+
+#if defined(JSON_HEDLEY_HAS_BUILTIN)
+    #undef JSON_HEDLEY_HAS_BUILTIN
+#endif
+#if defined(__has_builtin)
+    #define JSON_HEDLEY_HAS_BUILTIN(builtin) __has_builtin(builtin)
+#else
+    #define JSON_HEDLEY_HAS_BUILTIN(builtin) (0)
+#endif
+
+#if defined(JSON_HEDLEY_GNUC_HAS_BUILTIN)
+    #undef JSON_HEDLEY_GNUC_HAS_BUILTIN
+#endif
+#if defined(__has_builtin)
+    #define JSON_HEDLEY_GNUC_HAS_BUILTIN(builtin,major,minor,patch) __has_builtin(builtin)
+#else
+    #define JSON_HEDLEY_GNUC_HAS_BUILTIN(builtin,major,minor,patch) JSON_HEDLEY_GNUC_VERSION_CHECK(major,minor,patch)
+#endif
+
+#if defined(JSON_HEDLEY_GCC_HAS_BUILTIN)
+    #undef JSON_HEDLEY_GCC_HAS_BUILTIN
+#endif
+#if defined(__has_builtin)
+    #define JSON_HEDLEY_GCC_HAS_BUILTIN(builtin,major,minor,patch) __has_builtin(builtin)
+#else
+    #define JSON_HEDLEY_GCC_HAS_BUILTIN(builtin,major,minor,patch) JSON_HEDLEY_GCC_VERSION_CHECK(major,minor,patch)
+#endif
+
+#if defined(JSON_HEDLEY_HAS_FEATURE)
+    #undef JSON_HEDLEY_HAS_FEATURE
+#endif
+#if defined(__has_feature)
+    #define JSON_HEDLEY_HAS_FEATURE(feature) __has_feature(feature)
+#else
+    #define JSON_HEDLEY_HAS_FEATURE(feature) (0)
+#endif
+
+#if defined(JSON_HEDLEY_GNUC_HAS_FEATURE)
+    #undef JSON_HEDLEY_GNUC_HAS_FEATURE
+#endif
+#if defined(__has_feature)
+    #define JSON_HEDLEY_GNUC_HAS_FEATURE(feature,major,minor,patch) __has_feature(feature)
+#else
+    #define JSON_HEDLEY_GNUC_HAS_FEATURE(feature,major,minor,patch) JSON_HEDLEY_GNUC_VERSION_CHECK(major,minor,patch)
+#endif
+
+#if defined(JSON_HEDLEY_GCC_HAS_FEATURE)
+    #undef JSON_HEDLEY_GCC_HAS_FEATURE
+#endif
+#if defined(__has_feature)
+    #define JSON_HEDLEY_GCC_HAS_FEATURE(feature,major,minor,patch) __has_feature(feature)
+#else
+    #define JSON_HEDLEY_GCC_HAS_FEATURE(feature,major,minor,patch) JSON_HEDLEY_GCC_VERSION_CHECK(major,minor,patch)
+#endif
+
+#if defined(JSON_HEDLEY_HAS_EXTENSION)
+    #undef JSON_HEDLEY_HAS_EXTENSION
+#endif
+#if defined(__has_extension)
+    #define JSON_HEDLEY_HAS_EXTENSION(extension) __has_extension(extension)
+#else
+    #define JSON_HEDLEY_HAS_EXTENSION(extension) (0)
+#endif
+
+#if defined(JSON_HEDLEY_GNUC_HAS_EXTENSION)
+    #undef JSON_HEDLEY_GNUC_HAS_EXTENSION
+#endif
+#if defined(__has_extension)
+    #define JSON_HEDLEY_GNUC_HAS_EXTENSION(extension,major,minor,patch) __has_extension(extension)
+#else
+    #define JSON_HEDLEY_GNUC_HAS_EXTENSION(extension,major,minor,patch) JSON_HEDLEY_GNUC_VERSION_CHECK(major,minor,patch)
+#endif
+
+#if defined(JSON_HEDLEY_GCC_HAS_EXTENSION)
+    #undef JSON_HEDLEY_GCC_HAS_EXTENSION
+#endif
+#if defined(__has_extension)
+    #define JSON_HEDLEY_GCC_HAS_EXTENSION(extension,major,minor,patch) __has_extension(extension)
+#else
+    #define JSON_HEDLEY_GCC_HAS_EXTENSION(extension,major,minor,patch) JSON_HEDLEY_GCC_VERSION_CHECK(major,minor,patch)
+#endif
+
+#if defined(JSON_HEDLEY_HAS_DECLSPEC_ATTRIBUTE)
+    #undef JSON_HEDLEY_HAS_DECLSPEC_ATTRIBUTE
+#endif
+#if defined(__has_declspec_attribute)
+    #define JSON_HEDLEY_HAS_DECLSPEC_ATTRIBUTE(attribute) __has_declspec_attribute(attribute)
+#else
+    #define JSON_HEDLEY_HAS_DECLSPEC_ATTRIBUTE(attribute) (0)
+#endif
+
+#if defined(JSON_HEDLEY_GNUC_HAS_DECLSPEC_ATTRIBUTE)
+    #undef JSON_HEDLEY_GNUC_HAS_DECLSPEC_ATTRIBUTE
+#endif
+#if defined(__has_declspec_attribute)
+    #define JSON_HEDLEY_GNUC_HAS_DECLSPEC_ATTRIBUTE(attribute,major,minor,patch) __has_declspec_attribute(attribute)
+#else
+    #define JSON_HEDLEY_GNUC_HAS_DECLSPEC_ATTRIBUTE(attribute,major,minor,patch) JSON_HEDLEY_GNUC_VERSION_CHECK(major,minor,patch)
+#endif
+
+#if defined(JSON_HEDLEY_GCC_HAS_DECLSPEC_ATTRIBUTE)
+    #undef JSON_HEDLEY_GCC_HAS_DECLSPEC_ATTRIBUTE
+#endif
+#if defined(__has_declspec_attribute)
+    #define JSON_HEDLEY_GCC_HAS_DECLSPEC_ATTRIBUTE(attribute,major,minor,patch) __has_declspec_attribute(attribute)
+#else
+    #define JSON_HEDLEY_GCC_HAS_DECLSPEC_ATTRIBUTE(attribute,major,minor,patch) JSON_HEDLEY_GCC_VERSION_CHECK(major,minor,patch)
+#endif
+
+#if defined(JSON_HEDLEY_HAS_WARNING)
+    #undef JSON_HEDLEY_HAS_WARNING
+#endif
+#if defined(__has_warning)
+    #define JSON_HEDLEY_HAS_WARNING(warning) __has_warning(warning)
+#else
+    #define JSON_HEDLEY_HAS_WARNING(warning) (0)
+#endif
+
+#if defined(JSON_HEDLEY_GNUC_HAS_WARNING)
+    #undef JSON_HEDLEY_GNUC_HAS_WARNING
+#endif
+#if defined(__has_warning)
+    #define JSON_HEDLEY_GNUC_HAS_WARNING(warning,major,minor,patch) __has_warning(warning)
+#else
+    #define JSON_HEDLEY_GNUC_HAS_WARNING(warning,major,minor,patch) JSON_HEDLEY_GNUC_VERSION_CHECK(major,minor,patch)
+#endif
+
+#if defined(JSON_HEDLEY_GCC_HAS_WARNING)
+    #undef JSON_HEDLEY_GCC_HAS_WARNING
+#endif
+#if defined(__has_warning)
+    #define JSON_HEDLEY_GCC_HAS_WARNING(warning,major,minor,patch) __has_warning(warning)
+#else
+    #define JSON_HEDLEY_GCC_HAS_WARNING(warning,major,minor,patch) JSON_HEDLEY_GCC_VERSION_CHECK(major,minor,patch)
+#endif
+
+#if \
+    (defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L)) || \
+    defined(__clang__) || \
+    JSON_HEDLEY_GCC_VERSION_CHECK(3,0,0) || \
+    JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) || \
+    JSON_HEDLEY_IAR_VERSION_CHECK(8,0,0) || \
+    JSON_HEDLEY_PGI_VERSION_CHECK(18,4,0) || \
+    JSON_HEDLEY_ARM_VERSION_CHECK(4,1,0) || \
+    JSON_HEDLEY_TI_VERSION_CHECK(15,12,0) || \
+    JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(4,7,0) || \
+    JSON_HEDLEY_TI_CL430_VERSION_CHECK(2,0,1) || \
+    JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,1,0) || \
+    JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,0,0) || \
+    JSON_HEDLEY_TI_CL7X_VERSION_CHECK(1,2,0) || \
+    JSON_HEDLEY_TI_CLPRU_VERSION_CHECK(2,1,0) || \
+    JSON_HEDLEY_CRAY_VERSION_CHECK(5,0,0) || \
+    JSON_HEDLEY_TINYC_VERSION_CHECK(0,9,17) || \
+    JSON_HEDLEY_SUNPRO_VERSION_CHECK(8,0,0) || \
+    (JSON_HEDLEY_IBM_VERSION_CHECK(10,1,0) && defined(__C99_PRAGMA_OPERATOR))
+    #define JSON_HEDLEY_PRAGMA(value) _Pragma(#value)
+#elif JSON_HEDLEY_MSVC_VERSION_CHECK(15,0,0)
+    #define JSON_HEDLEY_PRAGMA(value) __pragma(value)
+#else
+    #define JSON_HEDLEY_PRAGMA(value)
+#endif
+
+#if defined(JSON_HEDLEY_DIAGNOSTIC_PUSH)
+    #undef JSON_HEDLEY_DIAGNOSTIC_PUSH
+#endif
+#if defined(JSON_HEDLEY_DIAGNOSTIC_POP)
+    #undef JSON_HEDLEY_DIAGNOSTIC_POP
+#endif
+#if defined(__clang__)
+    #define JSON_HEDLEY_DIAGNOSTIC_PUSH _Pragma("clang diagnostic push")
+    #define JSON_HEDLEY_DIAGNOSTIC_POP _Pragma("clang diagnostic pop")
+#elif JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0)
+    #define JSON_HEDLEY_DIAGNOSTIC_PUSH _Pragma("warning(push)")
+    #define JSON_HEDLEY_DIAGNOSTIC_POP _Pragma("warning(pop)")
+#elif JSON_HEDLEY_GCC_VERSION_CHECK(4,6,0)
+    #define JSON_HEDLEY_DIAGNOSTIC_PUSH _Pragma("GCC diagnostic push")
+    #define JSON_HEDLEY_DIAGNOSTIC_POP _Pragma("GCC diagnostic pop")
+#elif \
+    JSON_HEDLEY_MSVC_VERSION_CHECK(15,0,0) || \
+    JSON_HEDLEY_INTEL_CL_VERSION_CHECK(2021,1,0)
+    #define JSON_HEDLEY_DIAGNOSTIC_PUSH __pragma(warning(push))
+    #define JSON_HEDLEY_DIAGNOSTIC_POP __pragma(warning(pop))
+#elif JSON_HEDLEY_ARM_VERSION_CHECK(5,6,0)
+    #define JSON_HEDLEY_DIAGNOSTIC_PUSH _Pragma("push")
+    #define JSON_HEDLEY_DIAGNOSTIC_POP _Pragma("pop")
+#elif \
+    JSON_HEDLEY_TI_VERSION_CHECK(15,12,0) || \
+    JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(5,2,0) || \
+    JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,4,0) || \
+    JSON_HEDLEY_TI_CL6X_VERSION_CHECK(8,1,0) || \
+    JSON_HEDLEY_TI_CL7X_VERSION_CHECK(1,2,0) || \
+    JSON_HEDLEY_TI_CLPRU_VERSION_CHECK(2,1,0)
+    #define JSON_HEDLEY_DIAGNOSTIC_PUSH _Pragma("diag_push")
+    #define JSON_HEDLEY_DIAGNOSTIC_POP _Pragma("diag_pop")
+#elif JSON_HEDLEY_PELLES_VERSION_CHECK(2,90,0)
+    #define JSON_HEDLEY_DIAGNOSTIC_PUSH _Pragma("warning(push)")
+    #define JSON_HEDLEY_DIAGNOSTIC_POP _Pragma("warning(pop)")
+#else
+    #define JSON_HEDLEY_DIAGNOSTIC_PUSH
+    #define JSON_HEDLEY_DIAGNOSTIC_POP
+#endif
+
+/* JSON_HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_ is for
+   HEDLEY INTERNAL USE ONLY.  API subject to change without notice. */
+#if defined(JSON_HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_)
+    #undef JSON_HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_
+#endif
+#if defined(__cplusplus)
+#  if JSON_HEDLEY_HAS_WARNING("-Wc++98-compat")
+#    if JSON_HEDLEY_HAS_WARNING("-Wc++17-extensions")
+#      if JSON_HEDLEY_HAS_WARNING("-Wc++1z-extensions")
+#        define JSON_HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_(xpr) \
+    JSON_HEDLEY_DIAGNOSTIC_PUSH \
+    _Pragma("clang diagnostic ignored \"-Wc++98-compat\"") \
+    _Pragma("clang diagnostic ignored \"-Wc++17-extensions\"") \
+    _Pragma("clang diagnostic ignored \"-Wc++1z-extensions\"") \
+    xpr \
+    JSON_HEDLEY_DIAGNOSTIC_POP
+#      else
+#        define JSON_HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_(xpr) \
+    JSON_HEDLEY_DIAGNOSTIC_PUSH \
+    _Pragma("clang diagnostic ignored \"-Wc++98-compat\"") \
+    _Pragma("clang diagnostic ignored \"-Wc++17-extensions\"") \
+    xpr \
+    JSON_HEDLEY_DIAGNOSTIC_POP
+#      endif
+#    else
+#      define JSON_HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_(xpr) \
+    JSON_HEDLEY_DIAGNOSTIC_PUSH \
+    _Pragma("clang diagnostic ignored \"-Wc++98-compat\"") \
+    xpr \
+    JSON_HEDLEY_DIAGNOSTIC_POP
+#    endif
+#  endif
+#endif
+#if !defined(JSON_HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_)
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_(x) x
+#endif
+
+#if defined(JSON_HEDLEY_CONST_CAST)
+    #undef JSON_HEDLEY_CONST_CAST
+#endif
+#if defined(__cplusplus)
+#  define JSON_HEDLEY_CONST_CAST(T, expr) (const_cast<T>(expr))
+#elif \
+  JSON_HEDLEY_HAS_WARNING("-Wcast-qual") || \
+  JSON_HEDLEY_GCC_VERSION_CHECK(4,6,0) || \
+  JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0)
+#  define JSON_HEDLEY_CONST_CAST(T, expr) (__extension__ ({ \
+        JSON_HEDLEY_DIAGNOSTIC_PUSH \
+        JSON_HEDLEY_DIAGNOSTIC_DISABLE_CAST_QUAL \
+        ((T) (expr)); \
+        JSON_HEDLEY_DIAGNOSTIC_POP \
+    }))
+#else
+#  define JSON_HEDLEY_CONST_CAST(T, expr) ((T) (expr))
+#endif
+
+#if defined(JSON_HEDLEY_REINTERPRET_CAST)
+    #undef JSON_HEDLEY_REINTERPRET_CAST
+#endif
+#if defined(__cplusplus)
+    #define JSON_HEDLEY_REINTERPRET_CAST(T, expr) (reinterpret_cast<T>(expr))
+#else
+    #define JSON_HEDLEY_REINTERPRET_CAST(T, expr) ((T) (expr))
+#endif
+
+#if defined(JSON_HEDLEY_STATIC_CAST)
+    #undef JSON_HEDLEY_STATIC_CAST
+#endif
+#if defined(__cplusplus)
+    #define JSON_HEDLEY_STATIC_CAST(T, expr) (static_cast<T>(expr))
+#else
+    #define JSON_HEDLEY_STATIC_CAST(T, expr) ((T) (expr))
+#endif
+
+#if defined(JSON_HEDLEY_CPP_CAST)
+    #undef JSON_HEDLEY_CPP_CAST
+#endif
+#if defined(__cplusplus)
+#  if JSON_HEDLEY_HAS_WARNING("-Wold-style-cast")
+#    define JSON_HEDLEY_CPP_CAST(T, expr) \
+    JSON_HEDLEY_DIAGNOSTIC_PUSH \
+    _Pragma("clang diagnostic ignored \"-Wold-style-cast\"") \
+    ((T) (expr)) \
+    JSON_HEDLEY_DIAGNOSTIC_POP
+#  elif JSON_HEDLEY_IAR_VERSION_CHECK(8,3,0)
+#    define JSON_HEDLEY_CPP_CAST(T, expr) \
+    JSON_HEDLEY_DIAGNOSTIC_PUSH \
+    _Pragma("diag_suppress=Pe137") \
+    JSON_HEDLEY_DIAGNOSTIC_POP
+#  else
+#    define JSON_HEDLEY_CPP_CAST(T, expr) ((T) (expr))
+#  endif
+#else
+#  define JSON_HEDLEY_CPP_CAST(T, expr) (expr)
+#endif
+
+#if defined(JSON_HEDLEY_DIAGNOSTIC_DISABLE_DEPRECATED)
+    #undef JSON_HEDLEY_DIAGNOSTIC_DISABLE_DEPRECATED
+#endif
+#if JSON_HEDLEY_HAS_WARNING("-Wdeprecated-declarations")
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_DEPRECATED _Pragma("clang diagnostic ignored \"-Wdeprecated-declarations\"")
+#elif JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0)
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_DEPRECATED _Pragma("warning(disable:1478 1786)")
+#elif JSON_HEDLEY_INTEL_CL_VERSION_CHECK(2021,1,0)
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_DEPRECATED __pragma(warning(disable:1478 1786))
+#elif JSON_HEDLEY_PGI_VERSION_CHECK(20,7,0)
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_DEPRECATED _Pragma("diag_suppress 1215,1216,1444,1445")
+#elif JSON_HEDLEY_PGI_VERSION_CHECK(17,10,0)
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_DEPRECATED _Pragma("diag_suppress 1215,1444")
+#elif JSON_HEDLEY_GCC_VERSION_CHECK(4,3,0)
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_DEPRECATED _Pragma("GCC diagnostic ignored \"-Wdeprecated-declarations\"")
+#elif JSON_HEDLEY_MSVC_VERSION_CHECK(15,0,0)
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_DEPRECATED __pragma(warning(disable:4996))
+#elif JSON_HEDLEY_MCST_LCC_VERSION_CHECK(1,25,10)
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_DEPRECATED _Pragma("diag_suppress 1215,1444")
+#elif \
+    JSON_HEDLEY_TI_VERSION_CHECK(15,12,0) || \
+    (JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(4,8,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \
+    JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(5,2,0) || \
+    (JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,0,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \
+    JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,4,0) || \
+    (JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,0,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \
+    JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,3,0) || \
+    (JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,2,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \
+    JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,5,0) || \
+    JSON_HEDLEY_TI_CL7X_VERSION_CHECK(1,2,0) || \
+    JSON_HEDLEY_TI_CLPRU_VERSION_CHECK(2,1,0)
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_DEPRECATED _Pragma("diag_suppress 1291,1718")
+#elif JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,13,0) && !defined(__cplusplus)
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_DEPRECATED _Pragma("error_messages(off,E_DEPRECATED_ATT,E_DEPRECATED_ATT_MESS)")
+#elif JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,13,0) && defined(__cplusplus)
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_DEPRECATED _Pragma("error_messages(off,symdeprecated,symdeprecated2)")
+#elif JSON_HEDLEY_IAR_VERSION_CHECK(8,0,0)
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_DEPRECATED _Pragma("diag_suppress=Pe1444,Pe1215")
+#elif JSON_HEDLEY_PELLES_VERSION_CHECK(2,90,0)
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_DEPRECATED _Pragma("warn(disable:2241)")
+#else
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_DEPRECATED
+#endif
+
+#if defined(JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_PRAGMAS)
+    #undef JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_PRAGMAS
+#endif
+#if JSON_HEDLEY_HAS_WARNING("-Wunknown-pragmas")
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_PRAGMAS _Pragma("clang diagnostic ignored \"-Wunknown-pragmas\"")
+#elif JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0)
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_PRAGMAS _Pragma("warning(disable:161)")
+#elif JSON_HEDLEY_INTEL_CL_VERSION_CHECK(2021,1,0)
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_PRAGMAS __pragma(warning(disable:161))
+#elif JSON_HEDLEY_PGI_VERSION_CHECK(17,10,0)
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_PRAGMAS _Pragma("diag_suppress 1675")
+#elif JSON_HEDLEY_GCC_VERSION_CHECK(4,3,0)
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_PRAGMAS _Pragma("GCC diagnostic ignored \"-Wunknown-pragmas\"")
+#elif JSON_HEDLEY_MSVC_VERSION_CHECK(15,0,0)
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_PRAGMAS __pragma(warning(disable:4068))
+#elif \
+    JSON_HEDLEY_TI_VERSION_CHECK(16,9,0) || \
+    JSON_HEDLEY_TI_CL6X_VERSION_CHECK(8,0,0) || \
+    JSON_HEDLEY_TI_CL7X_VERSION_CHECK(1,2,0) || \
+    JSON_HEDLEY_TI_CLPRU_VERSION_CHECK(2,3,0)
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_PRAGMAS _Pragma("diag_suppress 163")
+#elif JSON_HEDLEY_TI_CL6X_VERSION_CHECK(8,0,0)
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_PRAGMAS _Pragma("diag_suppress 163")
+#elif JSON_HEDLEY_IAR_VERSION_CHECK(8,0,0)
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_PRAGMAS _Pragma("diag_suppress=Pe161")
+#elif JSON_HEDLEY_MCST_LCC_VERSION_CHECK(1,25,10)
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_PRAGMAS _Pragma("diag_suppress 161")
+#else
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_PRAGMAS
+#endif
+
+#if defined(JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_CPP_ATTRIBUTES)
+    #undef JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_CPP_ATTRIBUTES
+#endif
+#if JSON_HEDLEY_HAS_WARNING("-Wunknown-attributes")
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_CPP_ATTRIBUTES _Pragma("clang diagnostic ignored \"-Wunknown-attributes\"")
+#elif JSON_HEDLEY_GCC_VERSION_CHECK(4,6,0)
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_CPP_ATTRIBUTES _Pragma("GCC diagnostic ignored \"-Wdeprecated-declarations\"")
+#elif JSON_HEDLEY_INTEL_VERSION_CHECK(17,0,0)
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_CPP_ATTRIBUTES _Pragma("warning(disable:1292)")
+#elif JSON_HEDLEY_INTEL_CL_VERSION_CHECK(2021,1,0)
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_CPP_ATTRIBUTES __pragma(warning(disable:1292))
+#elif JSON_HEDLEY_MSVC_VERSION_CHECK(19,0,0)
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_CPP_ATTRIBUTES __pragma(warning(disable:5030))
+#elif JSON_HEDLEY_PGI_VERSION_CHECK(20,7,0)
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_CPP_ATTRIBUTES _Pragma("diag_suppress 1097,1098")
+#elif JSON_HEDLEY_PGI_VERSION_CHECK(17,10,0)
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_CPP_ATTRIBUTES _Pragma("diag_suppress 1097")
+#elif JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,14,0) && defined(__cplusplus)
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_CPP_ATTRIBUTES _Pragma("error_messages(off,attrskipunsup)")
+#elif \
+    JSON_HEDLEY_TI_VERSION_CHECK(18,1,0) || \
+    JSON_HEDLEY_TI_CL6X_VERSION_CHECK(8,3,0) || \
+    JSON_HEDLEY_TI_CL7X_VERSION_CHECK(1,2,0)
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_CPP_ATTRIBUTES _Pragma("diag_suppress 1173")
+#elif JSON_HEDLEY_IAR_VERSION_CHECK(8,0,0)
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_CPP_ATTRIBUTES _Pragma("diag_suppress=Pe1097")
+#elif JSON_HEDLEY_MCST_LCC_VERSION_CHECK(1,25,10)
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_CPP_ATTRIBUTES _Pragma("diag_suppress 1097")
+#else
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_CPP_ATTRIBUTES
+#endif
+
+#if defined(JSON_HEDLEY_DIAGNOSTIC_DISABLE_CAST_QUAL)
+    #undef JSON_HEDLEY_DIAGNOSTIC_DISABLE_CAST_QUAL
+#endif
+#if JSON_HEDLEY_HAS_WARNING("-Wcast-qual")
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_CAST_QUAL _Pragma("clang diagnostic ignored \"-Wcast-qual\"")
+#elif JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0)
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_CAST_QUAL _Pragma("warning(disable:2203 2331)")
+#elif JSON_HEDLEY_GCC_VERSION_CHECK(3,0,0)
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_CAST_QUAL _Pragma("GCC diagnostic ignored \"-Wcast-qual\"")
+#else
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_CAST_QUAL
+#endif
+
+#if defined(JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNUSED_FUNCTION)
+    #undef JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNUSED_FUNCTION
+#endif
+#if JSON_HEDLEY_HAS_WARNING("-Wunused-function")
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNUSED_FUNCTION _Pragma("clang diagnostic ignored \"-Wunused-function\"")
+#elif JSON_HEDLEY_GCC_VERSION_CHECK(3,4,0)
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNUSED_FUNCTION _Pragma("GCC diagnostic ignored \"-Wunused-function\"")
+#elif JSON_HEDLEY_MSVC_VERSION_CHECK(1,0,0)
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNUSED_FUNCTION __pragma(warning(disable:4505))
+#elif JSON_HEDLEY_MCST_LCC_VERSION_CHECK(1,25,10)
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNUSED_FUNCTION _Pragma("diag_suppress 3142")
+#else
+    #define JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNUSED_FUNCTION
+#endif
+
+#if defined(JSON_HEDLEY_DEPRECATED)
+    #undef JSON_HEDLEY_DEPRECATED
+#endif
+#if defined(JSON_HEDLEY_DEPRECATED_FOR)
+    #undef JSON_HEDLEY_DEPRECATED_FOR
+#endif
+#if \
+    JSON_HEDLEY_MSVC_VERSION_CHECK(14,0,0) || \
+    JSON_HEDLEY_INTEL_CL_VERSION_CHECK(2021,1,0)
+    #define JSON_HEDLEY_DEPRECATED(since) __declspec(deprecated("Since " # since))
+    #define JSON_HEDLEY_DEPRECATED_FOR(since, replacement) __declspec(deprecated("Since " #since "; use " #replacement))
+#elif \
+    (JSON_HEDLEY_HAS_EXTENSION(attribute_deprecated_with_message) && !defined(JSON_HEDLEY_IAR_VERSION)) || \
+    JSON_HEDLEY_GCC_VERSION_CHECK(4,5,0) || \
+    JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) || \
+    JSON_HEDLEY_ARM_VERSION_CHECK(5,6,0) || \
+    JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,13,0) || \
+    JSON_HEDLEY_PGI_VERSION_CHECK(17,10,0) || \
+    JSON_HEDLEY_TI_VERSION_CHECK(18,1,0) || \
+    JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(18,1,0) || \
+    JSON_HEDLEY_TI_CL6X_VERSION_CHECK(8,3,0) || \
+    JSON_HEDLEY_TI_CL7X_VERSION_CHECK(1,2,0) || \
+    JSON_HEDLEY_TI_CLPRU_VERSION_CHECK(2,3,0) || \
+    JSON_HEDLEY_MCST_LCC_VERSION_CHECK(1,25,10)
+    #define JSON_HEDLEY_DEPRECATED(since) __attribute__((__deprecated__("Since " #since)))
+    #define JSON_HEDLEY_DEPRECATED_FOR(since, replacement) __attribute__((__deprecated__("Since " #since "; use " #replacement)))
+#elif defined(__cplusplus) && (__cplusplus >= 201402L)
+    #define JSON_HEDLEY_DEPRECATED(since) JSON_HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_([[deprecated("Since " #since)]])
+    #define JSON_HEDLEY_DEPRECATED_FOR(since, replacement) JSON_HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_([[deprecated("Since " #since "; use " #replacement)]])
+#elif \
+    JSON_HEDLEY_HAS_ATTRIBUTE(deprecated) || \
+    JSON_HEDLEY_GCC_VERSION_CHECK(3,1,0) || \
+    JSON_HEDLEY_ARM_VERSION_CHECK(4,1,0) || \
+    JSON_HEDLEY_TI_VERSION_CHECK(15,12,0) || \
+    (JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(4,8,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \
+    JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(5,2,0) || \
+    (JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,0,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \
+    JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,4,0) || \
+    (JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,0,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \
+    JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,3,0) || \
+    (JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,2,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \
+    JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,5,0) || \
+    JSON_HEDLEY_TI_CL7X_VERSION_CHECK(1,2,0) || \
+    JSON_HEDLEY_TI_CLPRU_VERSION_CHECK(2,1,0) || \
+    JSON_HEDLEY_MCST_LCC_VERSION_CHECK(1,25,10) || \
+    JSON_HEDLEY_IAR_VERSION_CHECK(8,10,0)
+    #define JSON_HEDLEY_DEPRECATED(since) __attribute__((__deprecated__))
+    #define JSON_HEDLEY_DEPRECATED_FOR(since, replacement) __attribute__((__deprecated__))
+#elif \
+    JSON_HEDLEY_MSVC_VERSION_CHECK(13,10,0) || \
+    JSON_HEDLEY_PELLES_VERSION_CHECK(6,50,0) || \
+    JSON_HEDLEY_INTEL_CL_VERSION_CHECK(2021,1,0)
+    #define JSON_HEDLEY_DEPRECATED(since) __declspec(deprecated)
+    #define JSON_HEDLEY_DEPRECATED_FOR(since, replacement) __declspec(deprecated)
+#elif JSON_HEDLEY_IAR_VERSION_CHECK(8,0,0)
+    #define JSON_HEDLEY_DEPRECATED(since) _Pragma("deprecated")
+    #define JSON_HEDLEY_DEPRECATED_FOR(since, replacement) _Pragma("deprecated")
+#else
+    #define JSON_HEDLEY_DEPRECATED(since)
+    #define JSON_HEDLEY_DEPRECATED_FOR(since, replacement)
+#endif
+
+#if defined(JSON_HEDLEY_UNAVAILABLE)
+    #undef JSON_HEDLEY_UNAVAILABLE
+#endif
+#if \
+    JSON_HEDLEY_HAS_ATTRIBUTE(warning) || \
+    JSON_HEDLEY_GCC_VERSION_CHECK(4,3,0) || \
+    JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) || \
+    JSON_HEDLEY_MCST_LCC_VERSION_CHECK(1,25,10)
+    #define JSON_HEDLEY_UNAVAILABLE(available_since) __attribute__((__warning__("Not available until " #available_since)))
+#else
+    #define JSON_HEDLEY_UNAVAILABLE(available_since)
+#endif
+
+#if defined(JSON_HEDLEY_WARN_UNUSED_RESULT)
+    #undef JSON_HEDLEY_WARN_UNUSED_RESULT
+#endif
+#if defined(JSON_HEDLEY_WARN_UNUSED_RESULT_MSG)
+    #undef JSON_HEDLEY_WARN_UNUSED_RESULT_MSG
+#endif
+#if \
+    JSON_HEDLEY_HAS_ATTRIBUTE(warn_unused_result) || \
+    JSON_HEDLEY_GCC_VERSION_CHECK(3,4,0) || \
+    JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) || \
+    JSON_HEDLEY_TI_VERSION_CHECK(15,12,0) || \
+    (JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(4,8,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \
+    JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(5,2,0) || \
+    (JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,0,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \
+    JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,4,0) || \
+    (JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,0,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \
+    JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,3,0) || \
+    (JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,2,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \
+    JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,5,0) || \
+    JSON_HEDLEY_TI_CL7X_VERSION_CHECK(1,2,0) || \
+    JSON_HEDLEY_TI_CLPRU_VERSION_CHECK(2,1,0) || \
+    (JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,15,0) && defined(__cplusplus)) || \
+    JSON_HEDLEY_PGI_VERSION_CHECK(17,10,0) || \
+    JSON_HEDLEY_MCST_LCC_VERSION_CHECK(1,25,10)
+    #define JSON_HEDLEY_WARN_UNUSED_RESULT __attribute__((__warn_unused_result__))
+    #define JSON_HEDLEY_WARN_UNUSED_RESULT_MSG(msg) __attribute__((__warn_unused_result__))
+#elif (JSON_HEDLEY_HAS_CPP_ATTRIBUTE(nodiscard) >= 201907L)
+    #define JSON_HEDLEY_WARN_UNUSED_RESULT JSON_HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_([[nodiscard]])
+    #define JSON_HEDLEY_WARN_UNUSED_RESULT_MSG(msg) JSON_HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_([[nodiscard(msg)]])
+#elif JSON_HEDLEY_HAS_CPP_ATTRIBUTE(nodiscard)
+    #define JSON_HEDLEY_WARN_UNUSED_RESULT JSON_HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_([[nodiscard]])
+    #define JSON_HEDLEY_WARN_UNUSED_RESULT_MSG(msg) JSON_HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_([[nodiscard]])
+#elif defined(_Check_return_) /* SAL */
+    #define JSON_HEDLEY_WARN_UNUSED_RESULT _Check_return_
+    #define JSON_HEDLEY_WARN_UNUSED_RESULT_MSG(msg) _Check_return_
+#else
+    #define JSON_HEDLEY_WARN_UNUSED_RESULT
+    #define JSON_HEDLEY_WARN_UNUSED_RESULT_MSG(msg)
+#endif
+
+#if defined(JSON_HEDLEY_SENTINEL)
+    #undef JSON_HEDLEY_SENTINEL
+#endif
+#if \
+    JSON_HEDLEY_HAS_ATTRIBUTE(sentinel) || \
+    JSON_HEDLEY_GCC_VERSION_CHECK(4,0,0) || \
+    JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) || \
+    JSON_HEDLEY_ARM_VERSION_CHECK(5,4,0) || \
+    JSON_HEDLEY_MCST_LCC_VERSION_CHECK(1,25,10)
+    #define JSON_HEDLEY_SENTINEL(position) __attribute__((__sentinel__(position)))
+#else
+    #define JSON_HEDLEY_SENTINEL(position)
+#endif
+
+#if defined(JSON_HEDLEY_NO_RETURN)
+    #undef JSON_HEDLEY_NO_RETURN
+#endif
+#if JSON_HEDLEY_IAR_VERSION_CHECK(8,0,0)
+    #define JSON_HEDLEY_NO_RETURN __noreturn
+#elif \
+    JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) || \
+    JSON_HEDLEY_MCST_LCC_VERSION_CHECK(1,25,10)
+    #define JSON_HEDLEY_NO_RETURN __attribute__((__noreturn__))
+#elif defined(__STDC_VERSION__) && __STDC_VERSION__ >= 201112L
+    #define JSON_HEDLEY_NO_RETURN _Noreturn
+#elif defined(__cplusplus) && (__cplusplus >= 201103L)
+    #define JSON_HEDLEY_NO_RETURN JSON_HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_([[noreturn]])
+#elif \
+    JSON_HEDLEY_HAS_ATTRIBUTE(noreturn) || \
+    JSON_HEDLEY_GCC_VERSION_CHECK(3,2,0) || \
+    JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,11,0) || \
+    JSON_HEDLEY_ARM_VERSION_CHECK(4,1,0) || \
+    JSON_HEDLEY_IBM_VERSION_CHECK(10,1,0) || \
+    JSON_HEDLEY_TI_VERSION_CHECK(15,12,0) || \
+    (JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(4,8,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \
+    JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(5,2,0) || \
+    (JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,0,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \
+    JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,4,0) || \
+    (JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,0,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \
+    JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,3,0) || \
+    (JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,2,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \
+    JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,5,0) || \
+    JSON_HEDLEY_TI_CL7X_VERSION_CHECK(1,2,0) || \
+    JSON_HEDLEY_TI_CLPRU_VERSION_CHECK(2,1,0) || \
+    JSON_HEDLEY_IAR_VERSION_CHECK(8,10,0)
+    #define JSON_HEDLEY_NO_RETURN __attribute__((__noreturn__))
+#elif JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,10,0)
+    #define JSON_HEDLEY_NO_RETURN _Pragma("does_not_return")
+#elif \
+    JSON_HEDLEY_MSVC_VERSION_CHECK(13,10,0) || \
+    JSON_HEDLEY_INTEL_CL_VERSION_CHECK(2021,1,0)
+    #define JSON_HEDLEY_NO_RETURN __declspec(noreturn)
+#elif JSON_HEDLEY_TI_CL6X_VERSION_CHECK(6,0,0) && defined(__cplusplus)
+    #define JSON_HEDLEY_NO_RETURN _Pragma("FUNC_NEVER_RETURNS;")
+#elif JSON_HEDLEY_COMPCERT_VERSION_CHECK(3,2,0)
+    #define JSON_HEDLEY_NO_RETURN __attribute((noreturn))
+#elif JSON_HEDLEY_PELLES_VERSION_CHECK(9,0,0)
+    #define JSON_HEDLEY_NO_RETURN __declspec(noreturn)
+#else
+    #define JSON_HEDLEY_NO_RETURN
+#endif
+
+#if defined(JSON_HEDLEY_NO_ESCAPE)
+    #undef JSON_HEDLEY_NO_ESCAPE
+#endif
+#if JSON_HEDLEY_HAS_ATTRIBUTE(noescape)
+    #define JSON_HEDLEY_NO_ESCAPE __attribute__((__noescape__))
+#else
+    #define JSON_HEDLEY_NO_ESCAPE
+#endif
+
+#if defined(JSON_HEDLEY_UNREACHABLE)
+    #undef JSON_HEDLEY_UNREACHABLE
+#endif
+#if defined(JSON_HEDLEY_UNREACHABLE_RETURN)
+    #undef JSON_HEDLEY_UNREACHABLE_RETURN
+#endif
+#if defined(JSON_HEDLEY_ASSUME)
+    #undef JSON_HEDLEY_ASSUME
+#endif
+#if \
+    JSON_HEDLEY_MSVC_VERSION_CHECK(13,10,0) || \
+    JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) || \
+    JSON_HEDLEY_INTEL_CL_VERSION_CHECK(2021,1,0)
+    #define JSON_HEDLEY_ASSUME(expr) __assume(expr)
+#elif JSON_HEDLEY_HAS_BUILTIN(__builtin_assume)
+    #define JSON_HEDLEY_ASSUME(expr) __builtin_assume(expr)
+#elif \
+    JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,2,0) || \
+    JSON_HEDLEY_TI_CL6X_VERSION_CHECK(4,0,0)
+    #if defined(__cplusplus)
+        #define JSON_HEDLEY_ASSUME(expr) std::_nassert(expr)
+    #else
+        #define JSON_HEDLEY_ASSUME(expr) _nassert(expr)
+    #endif
+#endif
+#if \
+    (JSON_HEDLEY_HAS_BUILTIN(__builtin_unreachable) && (!defined(JSON_HEDLEY_ARM_VERSION))) || \
+    JSON_HEDLEY_GCC_VERSION_CHECK(4,5,0) || \
+    JSON_HEDLEY_PGI_VERSION_CHECK(18,10,0) || \
+    JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) || \
+    JSON_HEDLEY_IBM_VERSION_CHECK(13,1,5) || \
+    JSON_HEDLEY_CRAY_VERSION_CHECK(10,0,0) || \
+    JSON_HEDLEY_MCST_LCC_VERSION_CHECK(1,25,10)
+    #define JSON_HEDLEY_UNREACHABLE() __builtin_unreachable()
+#elif defined(JSON_HEDLEY_ASSUME)
+    #define JSON_HEDLEY_UNREACHABLE() JSON_HEDLEY_ASSUME(0)
+#endif
+#if !defined(JSON_HEDLEY_ASSUME)
+    #if defined(JSON_HEDLEY_UNREACHABLE)
+        #define JSON_HEDLEY_ASSUME(expr) JSON_HEDLEY_STATIC_CAST(void, ((expr) ? 1 : (JSON_HEDLEY_UNREACHABLE(), 1)))
+    #else
+        #define JSON_HEDLEY_ASSUME(expr) JSON_HEDLEY_STATIC_CAST(void, expr)
+    #endif
+#endif
+#if defined(JSON_HEDLEY_UNREACHABLE)
+    #if  \
+        JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,2,0) || \
+        JSON_HEDLEY_TI_CL6X_VERSION_CHECK(4,0,0)
+        #define JSON_HEDLEY_UNREACHABLE_RETURN(value) return (JSON_HEDLEY_STATIC_CAST(void, JSON_HEDLEY_ASSUME(0)), (value))
+    #else
+        #define JSON_HEDLEY_UNREACHABLE_RETURN(value) JSON_HEDLEY_UNREACHABLE()
+    #endif
+#else
+    #define JSON_HEDLEY_UNREACHABLE_RETURN(value) return (value)
+#endif
+#if !defined(JSON_HEDLEY_UNREACHABLE)
+    #define JSON_HEDLEY_UNREACHABLE() JSON_HEDLEY_ASSUME(0)
+#endif
+
+JSON_HEDLEY_DIAGNOSTIC_PUSH
+#if JSON_HEDLEY_HAS_WARNING("-Wpedantic")
+    #pragma clang diagnostic ignored "-Wpedantic"
+#endif
+#if JSON_HEDLEY_HAS_WARNING("-Wc++98-compat-pedantic") && defined(__cplusplus)
+    #pragma clang diagnostic ignored "-Wc++98-compat-pedantic"
+#endif
+#if JSON_HEDLEY_GCC_HAS_WARNING("-Wvariadic-macros",4,0,0)
+    #if defined(__clang__)
+        #pragma clang diagnostic ignored "-Wvariadic-macros"
+    #elif defined(JSON_HEDLEY_GCC_VERSION)
+        #pragma GCC diagnostic ignored "-Wvariadic-macros"
+    #endif
+#endif
+#if defined(JSON_HEDLEY_NON_NULL)
+    #undef JSON_HEDLEY_NON_NULL
+#endif
+#if \
+    JSON_HEDLEY_HAS_ATTRIBUTE(nonnull) || \
+    JSON_HEDLEY_GCC_VERSION_CHECK(3,3,0) || \
+    JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) || \
+    JSON_HEDLEY_ARM_VERSION_CHECK(4,1,0)
+    #define JSON_HEDLEY_NON_NULL(...) __attribute__((__nonnull__(__VA_ARGS__)))
+#else
+    #define JSON_HEDLEY_NON_NULL(...)
+#endif
+JSON_HEDLEY_DIAGNOSTIC_POP
+
+#if defined(JSON_HEDLEY_PRINTF_FORMAT)
+    #undef JSON_HEDLEY_PRINTF_FORMAT
+#endif
+#if defined(__MINGW32__) && JSON_HEDLEY_GCC_HAS_ATTRIBUTE(format,4,4,0) && !defined(__USE_MINGW_ANSI_STDIO)
+    #define JSON_HEDLEY_PRINTF_FORMAT(string_idx,first_to_check) __attribute__((__format__(ms_printf, string_idx, first_to_check)))
+#elif defined(__MINGW32__) && JSON_HEDLEY_GCC_HAS_ATTRIBUTE(format,4,4,0) && defined(__USE_MINGW_ANSI_STDIO)
+    #define JSON_HEDLEY_PRINTF_FORMAT(string_idx,first_to_check) __attribute__((__format__(gnu_printf, string_idx, first_to_check)))
+#elif \
+    JSON_HEDLEY_HAS_ATTRIBUTE(format) || \
+    JSON_HEDLEY_GCC_VERSION_CHECK(3,1,0) || \
+    JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) || \
+    JSON_HEDLEY_ARM_VERSION_CHECK(5,6,0) || \
+    JSON_HEDLEY_IBM_VERSION_CHECK(10,1,0) || \
+    JSON_HEDLEY_TI_VERSION_CHECK(15,12,0) || \
+    (JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(4,8,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \
+    JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(5,2,0) || \
+    (JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,0,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \
+    JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,4,0) || \
+    (JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,0,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \
+    JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,3,0) || \
+    (JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,2,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \
+    JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,5,0) || \
+    JSON_HEDLEY_TI_CL7X_VERSION_CHECK(1,2,0) || \
+    JSON_HEDLEY_TI_CLPRU_VERSION_CHECK(2,1,0) || \
+    JSON_HEDLEY_MCST_LCC_VERSION_CHECK(1,25,10)
+    #define JSON_HEDLEY_PRINTF_FORMAT(string_idx,first_to_check) __attribute__((__format__(__printf__, string_idx, first_to_check)))
+#elif JSON_HEDLEY_PELLES_VERSION_CHECK(6,0,0)
+    #define JSON_HEDLEY_PRINTF_FORMAT(string_idx,first_to_check) __declspec(vaformat(printf,string_idx,first_to_check))
+#else
+    #define JSON_HEDLEY_PRINTF_FORMAT(string_idx,first_to_check)
+#endif
+
+#if defined(JSON_HEDLEY_CONSTEXPR)
+    #undef JSON_HEDLEY_CONSTEXPR
+#endif
+#if defined(__cplusplus)
+    #if __cplusplus >= 201103L
+        #define JSON_HEDLEY_CONSTEXPR JSON_HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_(constexpr)
+    #endif
+#endif
+#if !defined(JSON_HEDLEY_CONSTEXPR)
+    #define JSON_HEDLEY_CONSTEXPR
+#endif
+
+#if defined(JSON_HEDLEY_PREDICT)
+    #undef JSON_HEDLEY_PREDICT
+#endif
+#if defined(JSON_HEDLEY_LIKELY)
+    #undef JSON_HEDLEY_LIKELY
+#endif
+#if defined(JSON_HEDLEY_UNLIKELY)
+    #undef JSON_HEDLEY_UNLIKELY
+#endif
+#if defined(JSON_HEDLEY_UNPREDICTABLE)
+    #undef JSON_HEDLEY_UNPREDICTABLE
+#endif
+#if JSON_HEDLEY_HAS_BUILTIN(__builtin_unpredictable)
+    #define JSON_HEDLEY_UNPREDICTABLE(expr) __builtin_unpredictable((expr))
+#endif
+#if \
+  (JSON_HEDLEY_HAS_BUILTIN(__builtin_expect_with_probability) && !defined(JSON_HEDLEY_PGI_VERSION)) || \
+  JSON_HEDLEY_GCC_VERSION_CHECK(9,0,0) || \
+  JSON_HEDLEY_MCST_LCC_VERSION_CHECK(1,25,10)
+#  define JSON_HEDLEY_PREDICT(expr, value, probability) __builtin_expect_with_probability(  (expr), (value), (probability))
+#  define JSON_HEDLEY_PREDICT_TRUE(expr, probability)   __builtin_expect_with_probability(!!(expr),    1   , (probability))
+#  define JSON_HEDLEY_PREDICT_FALSE(expr, probability)  __builtin_expect_with_probability(!!(expr),    0   , (probability))
+#  define JSON_HEDLEY_LIKELY(expr)                      __builtin_expect                 (!!(expr),    1                  )
+#  define JSON_HEDLEY_UNLIKELY(expr)                    __builtin_expect                 (!!(expr),    0                  )
+#elif \
+  (JSON_HEDLEY_HAS_BUILTIN(__builtin_expect) && !defined(JSON_HEDLEY_INTEL_CL_VERSION)) || \
+  JSON_HEDLEY_GCC_VERSION_CHECK(3,0,0) || \
+  JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) || \
+  (JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,15,0) && defined(__cplusplus)) || \
+  JSON_HEDLEY_ARM_VERSION_CHECK(4,1,0) || \
+  JSON_HEDLEY_IBM_VERSION_CHECK(10,1,0) || \
+  JSON_HEDLEY_TI_VERSION_CHECK(15,12,0) || \
+  JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(4,7,0) || \
+  JSON_HEDLEY_TI_CL430_VERSION_CHECK(3,1,0) || \
+  JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,1,0) || \
+  JSON_HEDLEY_TI_CL6X_VERSION_CHECK(6,1,0) || \
+  JSON_HEDLEY_TI_CL7X_VERSION_CHECK(1,2,0) || \
+  JSON_HEDLEY_TI_CLPRU_VERSION_CHECK(2,1,0) || \
+  JSON_HEDLEY_TINYC_VERSION_CHECK(0,9,27) || \
+  JSON_HEDLEY_CRAY_VERSION_CHECK(8,1,0) || \
+  JSON_HEDLEY_MCST_LCC_VERSION_CHECK(1,25,10)
+#  define JSON_HEDLEY_PREDICT(expr, expected, probability) \
+    (((probability) >= 0.9) ? __builtin_expect((expr), (expected)) : (JSON_HEDLEY_STATIC_CAST(void, expected), (expr)))
+#  define JSON_HEDLEY_PREDICT_TRUE(expr, probability) \
+    (__extension__ ({ \
+        double hedley_probability_ = (probability); \
+        ((hedley_probability_ >= 0.9) ? __builtin_expect(!!(expr), 1) : ((hedley_probability_ <= 0.1) ? __builtin_expect(!!(expr), 0) : !!(expr))); \
+    }))
+#  define JSON_HEDLEY_PREDICT_FALSE(expr, probability) \
+    (__extension__ ({ \
+        double hedley_probability_ = (probability); \
+        ((hedley_probability_ >= 0.9) ? __builtin_expect(!!(expr), 0) : ((hedley_probability_ <= 0.1) ? __builtin_expect(!!(expr), 1) : !!(expr))); \
+    }))
+#  define JSON_HEDLEY_LIKELY(expr)   __builtin_expect(!!(expr), 1)
+#  define JSON_HEDLEY_UNLIKELY(expr) __builtin_expect(!!(expr), 0)
+#else
+#  define JSON_HEDLEY_PREDICT(expr, expected, probability) (JSON_HEDLEY_STATIC_CAST(void, expected), (expr))
+#  define JSON_HEDLEY_PREDICT_TRUE(expr, probability) (!!(expr))
+#  define JSON_HEDLEY_PREDICT_FALSE(expr, probability) (!!(expr))
+#  define JSON_HEDLEY_LIKELY(expr) (!!(expr))
+#  define JSON_HEDLEY_UNLIKELY(expr) (!!(expr))
+#endif
+#if !defined(JSON_HEDLEY_UNPREDICTABLE)
+    #define JSON_HEDLEY_UNPREDICTABLE(expr) JSON_HEDLEY_PREDICT(expr, 1, 0.5)
+#endif
+
+#if defined(JSON_HEDLEY_MALLOC)
+    #undef JSON_HEDLEY_MALLOC
+#endif
+#if \
+    JSON_HEDLEY_HAS_ATTRIBUTE(malloc) || \
+    JSON_HEDLEY_GCC_VERSION_CHECK(3,1,0) || \
+    JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) || \
+    JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,11,0) || \
+    JSON_HEDLEY_ARM_VERSION_CHECK(4,1,0) || \
+    JSON_HEDLEY_IBM_VERSION_CHECK(12,1,0) || \
+    JSON_HEDLEY_TI_VERSION_CHECK(15,12,0) || \
+    (JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(4,8,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \
+    JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(5,2,0) || \
+    (JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,0,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \
+    JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,4,0) || \
+    (JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,0,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \
+    JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,3,0) || \
+    (JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,2,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \
+    JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,5,0) || \
+    JSON_HEDLEY_TI_CL7X_VERSION_CHECK(1,2,0) || \
+    JSON_HEDLEY_TI_CLPRU_VERSION_CHECK(2,1,0) || \
+    JSON_HEDLEY_MCST_LCC_VERSION_CHECK(1,25,10)
+    #define JSON_HEDLEY_MALLOC __attribute__((__malloc__))
+#elif JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,10,0)
+    #define JSON_HEDLEY_MALLOC _Pragma("returns_new_memory")
+#elif \
+    JSON_HEDLEY_MSVC_VERSION_CHECK(14,0,0) || \
+    JSON_HEDLEY_INTEL_CL_VERSION_CHECK(2021,1,0)
+    #define JSON_HEDLEY_MALLOC __declspec(restrict)
+#else
+    #define JSON_HEDLEY_MALLOC
+#endif
+
+#if defined(JSON_HEDLEY_PURE)
+    #undef JSON_HEDLEY_PURE
+#endif
+#if \
+  JSON_HEDLEY_HAS_ATTRIBUTE(pure) || \
+  JSON_HEDLEY_GCC_VERSION_CHECK(2,96,0) || \
+  JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) || \
+  JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,11,0) || \
+  JSON_HEDLEY_ARM_VERSION_CHECK(4,1,0) || \
+  JSON_HEDLEY_IBM_VERSION_CHECK(10,1,0) || \
+  JSON_HEDLEY_TI_VERSION_CHECK(15,12,0) || \
+  (JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(4,8,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \
+  JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(5,2,0) || \
+  (JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,0,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \
+  JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,4,0) || \
+  (JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,0,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \
+  JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,3,0) || \
+  (JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,2,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \
+  JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,5,0) || \
+  JSON_HEDLEY_TI_CL7X_VERSION_CHECK(1,2,0) || \
+  JSON_HEDLEY_TI_CLPRU_VERSION_CHECK(2,1,0) || \
+  JSON_HEDLEY_PGI_VERSION_CHECK(17,10,0) || \
+  JSON_HEDLEY_MCST_LCC_VERSION_CHECK(1,25,10)
+#  define JSON_HEDLEY_PURE __attribute__((__pure__))
+#elif JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,10,0)
+#  define JSON_HEDLEY_PURE _Pragma("does_not_write_global_data")
+#elif defined(__cplusplus) && \
+    ( \
+      JSON_HEDLEY_TI_CL430_VERSION_CHECK(2,0,1) || \
+      JSON_HEDLEY_TI_CL6X_VERSION_CHECK(4,0,0) || \
+      JSON_HEDLEY_TI_CL7X_VERSION_CHECK(1,2,0) \
+    )
+#  define JSON_HEDLEY_PURE _Pragma("FUNC_IS_PURE;")
+#else
+#  define JSON_HEDLEY_PURE
+#endif
+
+#if defined(JSON_HEDLEY_CONST)
+    #undef JSON_HEDLEY_CONST
+#endif
+#if \
+    JSON_HEDLEY_HAS_ATTRIBUTE(const) || \
+    JSON_HEDLEY_GCC_VERSION_CHECK(2,5,0) || \
+    JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) || \
+    JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,11,0) || \
+    JSON_HEDLEY_ARM_VERSION_CHECK(4,1,0) || \
+    JSON_HEDLEY_IBM_VERSION_CHECK(10,1,0) || \
+    JSON_HEDLEY_TI_VERSION_CHECK(15,12,0) || \
+    (JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(4,8,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \
+    JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(5,2,0) || \
+    (JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,0,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \
+    JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,4,0) || \
+    (JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,0,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \
+    JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,3,0) || \
+    (JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,2,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \
+    JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,5,0) || \
+    JSON_HEDLEY_TI_CL7X_VERSION_CHECK(1,2,0) || \
+    JSON_HEDLEY_TI_CLPRU_VERSION_CHECK(2,1,0) || \
+    JSON_HEDLEY_PGI_VERSION_CHECK(17,10,0) || \
+    JSON_HEDLEY_MCST_LCC_VERSION_CHECK(1,25,10)
+    #define JSON_HEDLEY_CONST __attribute__((__const__))
+#elif \
+    JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,10,0)
+    #define JSON_HEDLEY_CONST _Pragma("no_side_effect")
+#else
+    #define JSON_HEDLEY_CONST JSON_HEDLEY_PURE
+#endif
+
+#if defined(JSON_HEDLEY_RESTRICT)
+    #undef JSON_HEDLEY_RESTRICT
+#endif
+#if defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) && !defined(__cplusplus)
+    #define JSON_HEDLEY_RESTRICT restrict
+#elif \
+    JSON_HEDLEY_GCC_VERSION_CHECK(3,1,0) || \
+    JSON_HEDLEY_MSVC_VERSION_CHECK(14,0,0) || \
+    JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) || \
+    JSON_HEDLEY_INTEL_CL_VERSION_CHECK(2021,1,0) || \
+    JSON_HEDLEY_ARM_VERSION_CHECK(4,1,0) || \
+    JSON_HEDLEY_IBM_VERSION_CHECK(10,1,0) || \
+    JSON_HEDLEY_PGI_VERSION_CHECK(17,10,0) || \
+    JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,3,0) || \
+    JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,2,4) || \
+    JSON_HEDLEY_TI_CL6X_VERSION_CHECK(8,1,0) || \
+    JSON_HEDLEY_TI_CL7X_VERSION_CHECK(1,2,0) || \
+    (JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,14,0) && defined(__cplusplus)) || \
+    JSON_HEDLEY_IAR_VERSION_CHECK(8,0,0) || \
+    defined(__clang__) || \
+    JSON_HEDLEY_MCST_LCC_VERSION_CHECK(1,25,10)
+    #define JSON_HEDLEY_RESTRICT __restrict
+#elif JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,3,0) && !defined(__cplusplus)
+    #define JSON_HEDLEY_RESTRICT _Restrict
+#else
+    #define JSON_HEDLEY_RESTRICT
+#endif
+
+#if defined(JSON_HEDLEY_INLINE)
+    #undef JSON_HEDLEY_INLINE
+#endif
+#if \
+    (defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L)) || \
+    (defined(__cplusplus) && (__cplusplus >= 199711L))
+    #define JSON_HEDLEY_INLINE inline
+#elif \
+    defined(JSON_HEDLEY_GCC_VERSION) || \
+    JSON_HEDLEY_ARM_VERSION_CHECK(6,2,0)
+    #define JSON_HEDLEY_INLINE __inline__
+#elif \
+    JSON_HEDLEY_MSVC_VERSION_CHECK(12,0,0) || \
+    JSON_HEDLEY_INTEL_CL_VERSION_CHECK(2021,1,0) || \
+    JSON_HEDLEY_ARM_VERSION_CHECK(4,1,0) || \
+    JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(5,1,0) || \
+    JSON_HEDLEY_TI_CL430_VERSION_CHECK(3,1,0) || \
+    JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,2,0) || \
+    JSON_HEDLEY_TI_CL6X_VERSION_CHECK(8,0,0) || \
+    JSON_HEDLEY_TI_CL7X_VERSION_CHECK(1,2,0) || \
+    JSON_HEDLEY_TI_CLPRU_VERSION_CHECK(2,1,0) || \
+    JSON_HEDLEY_MCST_LCC_VERSION_CHECK(1,25,10)
+    #define JSON_HEDLEY_INLINE __inline
+#else
+    #define JSON_HEDLEY_INLINE
+#endif
+
+#if defined(JSON_HEDLEY_ALWAYS_INLINE)
+    #undef JSON_HEDLEY_ALWAYS_INLINE
+#endif
+#if \
+  JSON_HEDLEY_HAS_ATTRIBUTE(always_inline) || \
+  JSON_HEDLEY_GCC_VERSION_CHECK(4,0,0) || \
+  JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) || \
+  JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,11,0) || \
+  JSON_HEDLEY_ARM_VERSION_CHECK(4,1,0) || \
+  JSON_HEDLEY_IBM_VERSION_CHECK(10,1,0) || \
+  JSON_HEDLEY_TI_VERSION_CHECK(15,12,0) || \
+  (JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(4,8,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \
+  JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(5,2,0) || \
+  (JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,0,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \
+  JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,4,0) || \
+  (JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,0,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \
+  JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,3,0) || \
+  (JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,2,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \
+  JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,5,0) || \
+  JSON_HEDLEY_TI_CL7X_VERSION_CHECK(1,2,0) || \
+  JSON_HEDLEY_TI_CLPRU_VERSION_CHECK(2,1,0) || \
+  JSON_HEDLEY_MCST_LCC_VERSION_CHECK(1,25,10) || \
+  JSON_HEDLEY_IAR_VERSION_CHECK(8,10,0)
+#  define JSON_HEDLEY_ALWAYS_INLINE __attribute__((__always_inline__)) JSON_HEDLEY_INLINE
+#elif \
+  JSON_HEDLEY_MSVC_VERSION_CHECK(12,0,0) || \
+  JSON_HEDLEY_INTEL_CL_VERSION_CHECK(2021,1,0)
+#  define JSON_HEDLEY_ALWAYS_INLINE __forceinline
+#elif defined(__cplusplus) && \
+    ( \
+      JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(5,2,0) || \
+      JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,3,0) || \
+      JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,4,0) || \
+      JSON_HEDLEY_TI_CL6X_VERSION_CHECK(6,1,0) || \
+      JSON_HEDLEY_TI_CL7X_VERSION_CHECK(1,2,0) || \
+      JSON_HEDLEY_TI_CLPRU_VERSION_CHECK(2,1,0) \
+    )
+#  define JSON_HEDLEY_ALWAYS_INLINE _Pragma("FUNC_ALWAYS_INLINE;")
+#elif JSON_HEDLEY_IAR_VERSION_CHECK(8,0,0)
+#  define JSON_HEDLEY_ALWAYS_INLINE _Pragma("inline=forced")
+#else
+#  define JSON_HEDLEY_ALWAYS_INLINE JSON_HEDLEY_INLINE
+#endif
+
+#if defined(JSON_HEDLEY_NEVER_INLINE)
+    #undef JSON_HEDLEY_NEVER_INLINE
+#endif
+#if \
+    JSON_HEDLEY_HAS_ATTRIBUTE(noinline) || \
+    JSON_HEDLEY_GCC_VERSION_CHECK(4,0,0) || \
+    JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) || \
+    JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,11,0) || \
+    JSON_HEDLEY_ARM_VERSION_CHECK(4,1,0) || \
+    JSON_HEDLEY_IBM_VERSION_CHECK(10,1,0) || \
+    JSON_HEDLEY_TI_VERSION_CHECK(15,12,0) || \
+    (JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(4,8,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \
+    JSON_HEDLEY_TI_ARMCL_VERSION_CHECK(5,2,0) || \
+    (JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,0,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \
+    JSON_HEDLEY_TI_CL2000_VERSION_CHECK(6,4,0) || \
+    (JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,0,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \
+    JSON_HEDLEY_TI_CL430_VERSION_CHECK(4,3,0) || \
+    (JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,2,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \
+    JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,5,0) || \
+    JSON_HEDLEY_TI_CL7X_VERSION_CHECK(1,2,0) || \
+    JSON_HEDLEY_TI_CLPRU_VERSION_CHECK(2,1,0) || \
+    JSON_HEDLEY_MCST_LCC_VERSION_CHECK(1,25,10) || \
+    JSON_HEDLEY_IAR_VERSION_CHECK(8,10,0)
+    #define JSON_HEDLEY_NEVER_INLINE __attribute__((__noinline__))
+#elif \
+    JSON_HEDLEY_MSVC_VERSION_CHECK(13,10,0) || \
+    JSON_HEDLEY_INTEL_CL_VERSION_CHECK(2021,1,0)
+    #define JSON_HEDLEY_NEVER_INLINE __declspec(noinline)
+#elif JSON_HEDLEY_PGI_VERSION_CHECK(10,2,0)
+    #define JSON_HEDLEY_NEVER_INLINE _Pragma("noinline")
+#elif JSON_HEDLEY_TI_CL6X_VERSION_CHECK(6,0,0) && defined(__cplusplus)
+    #define JSON_HEDLEY_NEVER_INLINE _Pragma("FUNC_CANNOT_INLINE;")
+#elif JSON_HEDLEY_IAR_VERSION_CHECK(8,0,0)
+    #define JSON_HEDLEY_NEVER_INLINE _Pragma("inline=never")
+#elif JSON_HEDLEY_COMPCERT_VERSION_CHECK(3,2,0)
+    #define JSON_HEDLEY_NEVER_INLINE __attribute((noinline))
+#elif JSON_HEDLEY_PELLES_VERSION_CHECK(9,0,0)
+    #define JSON_HEDLEY_NEVER_INLINE __declspec(noinline)
+#else
+    #define JSON_HEDLEY_NEVER_INLINE
+#endif
+
+#if defined(JSON_HEDLEY_PRIVATE)
+    #undef JSON_HEDLEY_PRIVATE
+#endif
+#if defined(JSON_HEDLEY_PUBLIC)
+    #undef JSON_HEDLEY_PUBLIC
+#endif
+#if defined(JSON_HEDLEY_IMPORT)
+    #undef JSON_HEDLEY_IMPORT
+#endif
+#if defined(_WIN32) || defined(__CYGWIN__)
+#  define JSON_HEDLEY_PRIVATE
+#  define JSON_HEDLEY_PUBLIC   __declspec(dllexport)
+#  define JSON_HEDLEY_IMPORT   __declspec(dllimport)
+#else
+#  if \
+    JSON_HEDLEY_HAS_ATTRIBUTE(visibility) || \
+    JSON_HEDLEY_GCC_VERSION_CHECK(3,3,0) || \
+    JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,11,0) || \
+    JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) || \
+    JSON_HEDLEY_ARM_VERSION_CHECK(4,1,0) || \
+    JSON_HEDLEY_IBM_VERSION_CHECK(13,1,0) || \
+    ( \
+      defined(__TI_EABI__) && \
+      ( \
+        (JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,2,0) && defined(__TI_GNU_ATTRIBUTE_SUPPORT__)) || \
+        JSON_HEDLEY_TI_CL6X_VERSION_CHECK(7,5,0) \
+      ) \
+    ) || \
+    JSON_HEDLEY_MCST_LCC_VERSION_CHECK(1,25,10)
+#    define JSON_HEDLEY_PRIVATE __attribute__((__visibility__("hidden")))
+#    define JSON_HEDLEY_PUBLIC  __attribute__((__visibility__("default")))
+#  else
+#    define JSON_HEDLEY_PRIVATE
+#    define JSON_HEDLEY_PUBLIC
+#  endif
+#  define JSON_HEDLEY_IMPORT    extern
+#endif
+
+#if defined(JSON_HEDLEY_NO_THROW)
+    #undef JSON_HEDLEY_NO_THROW
+#endif
+#if \
+    JSON_HEDLEY_HAS_ATTRIBUTE(nothrow) || \
+    JSON_HEDLEY_GCC_VERSION_CHECK(3,3,0) || \
+    JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) || \
+    JSON_HEDLEY_MCST_LCC_VERSION_CHECK(1,25,10)
+    #define JSON_HEDLEY_NO_THROW __attribute__((__nothrow__))
+#elif \
+    JSON_HEDLEY_MSVC_VERSION_CHECK(13,1,0) || \
+    JSON_HEDLEY_INTEL_CL_VERSION_CHECK(2021,1,0) || \
+    JSON_HEDLEY_ARM_VERSION_CHECK(4,1,0)
+    #define JSON_HEDLEY_NO_THROW __declspec(nothrow)
+#else
+    #define JSON_HEDLEY_NO_THROW
+#endif
+
+#if defined(JSON_HEDLEY_FALL_THROUGH)
+    #undef JSON_HEDLEY_FALL_THROUGH
+#endif
+#if \
+    JSON_HEDLEY_HAS_ATTRIBUTE(fallthrough) || \
+    JSON_HEDLEY_GCC_VERSION_CHECK(7,0,0) || \
+    JSON_HEDLEY_MCST_LCC_VERSION_CHECK(1,25,10)
+    #define JSON_HEDLEY_FALL_THROUGH __attribute__((__fallthrough__))
+#elif JSON_HEDLEY_HAS_CPP_ATTRIBUTE_NS(clang,fallthrough)
+    #define JSON_HEDLEY_FALL_THROUGH JSON_HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_([[clang::fallthrough]])
+#elif JSON_HEDLEY_HAS_CPP_ATTRIBUTE(fallthrough)
+    #define JSON_HEDLEY_FALL_THROUGH JSON_HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_([[fallthrough]])
+#elif defined(__fallthrough) /* SAL */
+    #define JSON_HEDLEY_FALL_THROUGH __fallthrough
+#else
+    #define JSON_HEDLEY_FALL_THROUGH
+#endif
+
+#if defined(JSON_HEDLEY_RETURNS_NON_NULL)
+    #undef JSON_HEDLEY_RETURNS_NON_NULL
+#endif
+#if \
+    JSON_HEDLEY_HAS_ATTRIBUTE(returns_nonnull) || \
+    JSON_HEDLEY_GCC_VERSION_CHECK(4,9,0) || \
+    JSON_HEDLEY_MCST_LCC_VERSION_CHECK(1,25,10)
+    #define JSON_HEDLEY_RETURNS_NON_NULL __attribute__((__returns_nonnull__))
+#elif defined(_Ret_notnull_) /* SAL */
+    #define JSON_HEDLEY_RETURNS_NON_NULL _Ret_notnull_
+#else
+    #define JSON_HEDLEY_RETURNS_NON_NULL
+#endif
+
+#if defined(JSON_HEDLEY_ARRAY_PARAM)
+    #undef JSON_HEDLEY_ARRAY_PARAM
+#endif
+#if \
+    defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) && \
+    !defined(__STDC_NO_VLA__) && \
+    !defined(__cplusplus) && \
+    !defined(JSON_HEDLEY_PGI_VERSION) && \
+    !defined(JSON_HEDLEY_TINYC_VERSION)
+    #define JSON_HEDLEY_ARRAY_PARAM(name) (name)
+#else
+    #define JSON_HEDLEY_ARRAY_PARAM(name)
+#endif
+
+#if defined(JSON_HEDLEY_IS_CONSTANT)
+    #undef JSON_HEDLEY_IS_CONSTANT
+#endif
+#if defined(JSON_HEDLEY_REQUIRE_CONSTEXPR)
+    #undef JSON_HEDLEY_REQUIRE_CONSTEXPR
+#endif
+/* JSON_HEDLEY_IS_CONSTEXPR_ is for
+   HEDLEY INTERNAL USE ONLY.  API subject to change without notice. */
+#if defined(JSON_HEDLEY_IS_CONSTEXPR_)
+    #undef JSON_HEDLEY_IS_CONSTEXPR_
+#endif
+#if \
+    JSON_HEDLEY_HAS_BUILTIN(__builtin_constant_p) || \
+    JSON_HEDLEY_GCC_VERSION_CHECK(3,4,0) || \
+    JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) || \
+    JSON_HEDLEY_TINYC_VERSION_CHECK(0,9,19) || \
+    JSON_HEDLEY_ARM_VERSION_CHECK(4,1,0) || \
+    JSON_HEDLEY_IBM_VERSION_CHECK(13,1,0) || \
+    JSON_HEDLEY_TI_CL6X_VERSION_CHECK(6,1,0) || \
+    (JSON_HEDLEY_SUNPRO_VERSION_CHECK(5,10,0) && !defined(__cplusplus)) || \
+    JSON_HEDLEY_CRAY_VERSION_CHECK(8,1,0) || \
+    JSON_HEDLEY_MCST_LCC_VERSION_CHECK(1,25,10)
+    #define JSON_HEDLEY_IS_CONSTANT(expr) __builtin_constant_p(expr)
+#endif
+#if !defined(__cplusplus)
+#  if \
+       JSON_HEDLEY_HAS_BUILTIN(__builtin_types_compatible_p) || \
+       JSON_HEDLEY_GCC_VERSION_CHECK(3,4,0) || \
+       JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) || \
+       JSON_HEDLEY_IBM_VERSION_CHECK(13,1,0) || \
+       JSON_HEDLEY_CRAY_VERSION_CHECK(8,1,0) || \
+       JSON_HEDLEY_ARM_VERSION_CHECK(5,4,0) || \
+       JSON_HEDLEY_TINYC_VERSION_CHECK(0,9,24)
+#if defined(__INTPTR_TYPE__)
+    #define JSON_HEDLEY_IS_CONSTEXPR_(expr) __builtin_types_compatible_p(__typeof__((1 ? (void*) ((__INTPTR_TYPE__) ((expr) * 0)) : (int*) 0)), int*)
+#else
+    #include <stdint.h>
+    #define JSON_HEDLEY_IS_CONSTEXPR_(expr) __builtin_types_compatible_p(__typeof__((1 ? (void*) ((intptr_t) ((expr) * 0)) : (int*) 0)), int*)
+#endif
+#  elif \
+       ( \
+          defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 201112L) && \
+          !defined(JSON_HEDLEY_SUNPRO_VERSION) && \
+          !defined(JSON_HEDLEY_PGI_VERSION) && \
+          !defined(JSON_HEDLEY_IAR_VERSION)) || \
+       (JSON_HEDLEY_HAS_EXTENSION(c_generic_selections) && !defined(JSON_HEDLEY_IAR_VERSION)) || \
+       JSON_HEDLEY_GCC_VERSION_CHECK(4,9,0) || \
+       JSON_HEDLEY_INTEL_VERSION_CHECK(17,0,0) || \
+       JSON_HEDLEY_IBM_VERSION_CHECK(12,1,0) || \
+       JSON_HEDLEY_ARM_VERSION_CHECK(5,3,0)
+#if defined(__INTPTR_TYPE__)
+    #define JSON_HEDLEY_IS_CONSTEXPR_(expr) _Generic((1 ? (void*) ((__INTPTR_TYPE__) ((expr) * 0)) : (int*) 0), int*: 1, void*: 0)
+#else
+    #include <stdint.h>
+    #define JSON_HEDLEY_IS_CONSTEXPR_(expr) _Generic((1 ? (void*) ((intptr_t) * 0) : (int*) 0), int*: 1, void*: 0)
+#endif
+#  elif \
+       defined(JSON_HEDLEY_GCC_VERSION) || \
+       defined(JSON_HEDLEY_INTEL_VERSION) || \
+       defined(JSON_HEDLEY_TINYC_VERSION) || \
+       defined(JSON_HEDLEY_TI_ARMCL_VERSION) || \
+       JSON_HEDLEY_TI_CL430_VERSION_CHECK(18,12,0) || \
+       defined(JSON_HEDLEY_TI_CL2000_VERSION) || \
+       defined(JSON_HEDLEY_TI_CL6X_VERSION) || \
+       defined(JSON_HEDLEY_TI_CL7X_VERSION) || \
+       defined(JSON_HEDLEY_TI_CLPRU_VERSION) || \
+       defined(__clang__)
+#    define JSON_HEDLEY_IS_CONSTEXPR_(expr) ( \
+        sizeof(void) != \
+        sizeof(*( \
+                  1 ? \
+                  ((void*) ((expr) * 0L) ) : \
+((struct { char v[sizeof(void) * 2]; } *) 1) \
+                ) \
+              ) \
+                                            )
+#  endif
+#endif
+#if defined(JSON_HEDLEY_IS_CONSTEXPR_)
+    #if !defined(JSON_HEDLEY_IS_CONSTANT)
+        #define JSON_HEDLEY_IS_CONSTANT(expr) JSON_HEDLEY_IS_CONSTEXPR_(expr)
+    #endif
+    #define JSON_HEDLEY_REQUIRE_CONSTEXPR(expr) (JSON_HEDLEY_IS_CONSTEXPR_(expr) ? (expr) : (-1))
+#else
+    #if !defined(JSON_HEDLEY_IS_CONSTANT)
+        #define JSON_HEDLEY_IS_CONSTANT(expr) (0)
+    #endif
+    #define JSON_HEDLEY_REQUIRE_CONSTEXPR(expr) (expr)
+#endif
+
+#if defined(JSON_HEDLEY_BEGIN_C_DECLS)
+    #undef JSON_HEDLEY_BEGIN_C_DECLS
+#endif
+#if defined(JSON_HEDLEY_END_C_DECLS)
+    #undef JSON_HEDLEY_END_C_DECLS
+#endif
+#if defined(JSON_HEDLEY_C_DECL)
+    #undef JSON_HEDLEY_C_DECL
+#endif
+#if defined(__cplusplus)
+    #define JSON_HEDLEY_BEGIN_C_DECLS extern "C" {
+    #define JSON_HEDLEY_END_C_DECLS }
+    #define JSON_HEDLEY_C_DECL extern "C"
+#else
+    #define JSON_HEDLEY_BEGIN_C_DECLS
+    #define JSON_HEDLEY_END_C_DECLS
+    #define JSON_HEDLEY_C_DECL
+#endif
+
+#if defined(JSON_HEDLEY_STATIC_ASSERT)
+    #undef JSON_HEDLEY_STATIC_ASSERT
+#endif
+#if \
+  !defined(__cplusplus) && ( \
+      (defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 201112L)) || \
+      (JSON_HEDLEY_HAS_FEATURE(c_static_assert) && !defined(JSON_HEDLEY_INTEL_CL_VERSION)) || \
+      JSON_HEDLEY_GCC_VERSION_CHECK(6,0,0) || \
+      JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0) || \
+      defined(_Static_assert) \
+    )
+#  define JSON_HEDLEY_STATIC_ASSERT(expr, message) _Static_assert(expr, message)
+#elif \
+  (defined(__cplusplus) && (__cplusplus >= 201103L)) || \
+  JSON_HEDLEY_MSVC_VERSION_CHECK(16,0,0) || \
+  JSON_HEDLEY_INTEL_CL_VERSION_CHECK(2021,1,0)
+#  define JSON_HEDLEY_STATIC_ASSERT(expr, message) JSON_HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_(static_assert(expr, message))
+#else
+#  define JSON_HEDLEY_STATIC_ASSERT(expr, message)
+#endif
+
+#if defined(JSON_HEDLEY_NULL)
+    #undef JSON_HEDLEY_NULL
+#endif
+#if defined(__cplusplus)
+    #if __cplusplus >= 201103L
+        #define JSON_HEDLEY_NULL JSON_HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_(nullptr)
+    #elif defined(NULL)
+        #define JSON_HEDLEY_NULL NULL
+    #else
+        #define JSON_HEDLEY_NULL JSON_HEDLEY_STATIC_CAST(void*, 0)
+    #endif
+#elif defined(NULL)
+    #define JSON_HEDLEY_NULL NULL
+#else
+    #define JSON_HEDLEY_NULL ((void*) 0)
+#endif
+
+#if defined(JSON_HEDLEY_MESSAGE)
+    #undef JSON_HEDLEY_MESSAGE
+#endif
+#if JSON_HEDLEY_HAS_WARNING("-Wunknown-pragmas")
+#  define JSON_HEDLEY_MESSAGE(msg) \
+    JSON_HEDLEY_DIAGNOSTIC_PUSH \
+    JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_PRAGMAS \
+    JSON_HEDLEY_PRAGMA(message msg) \
+    JSON_HEDLEY_DIAGNOSTIC_POP
+#elif \
+  JSON_HEDLEY_GCC_VERSION_CHECK(4,4,0) || \
+  JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0)
+#  define JSON_HEDLEY_MESSAGE(msg) JSON_HEDLEY_PRAGMA(message msg)
+#elif JSON_HEDLEY_CRAY_VERSION_CHECK(5,0,0)
+#  define JSON_HEDLEY_MESSAGE(msg) JSON_HEDLEY_PRAGMA(_CRI message msg)
+#elif JSON_HEDLEY_IAR_VERSION_CHECK(8,0,0)
+#  define JSON_HEDLEY_MESSAGE(msg) JSON_HEDLEY_PRAGMA(message(msg))
+#elif JSON_HEDLEY_PELLES_VERSION_CHECK(2,0,0)
+#  define JSON_HEDLEY_MESSAGE(msg) JSON_HEDLEY_PRAGMA(message(msg))
+#else
+#  define JSON_HEDLEY_MESSAGE(msg)
+#endif
+
+#if defined(JSON_HEDLEY_WARNING)
+    #undef JSON_HEDLEY_WARNING
+#endif
+#if JSON_HEDLEY_HAS_WARNING("-Wunknown-pragmas")
+#  define JSON_HEDLEY_WARNING(msg) \
+    JSON_HEDLEY_DIAGNOSTIC_PUSH \
+    JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_PRAGMAS \
+    JSON_HEDLEY_PRAGMA(clang warning msg) \
+    JSON_HEDLEY_DIAGNOSTIC_POP
+#elif \
+  JSON_HEDLEY_GCC_VERSION_CHECK(4,8,0) || \
+  JSON_HEDLEY_PGI_VERSION_CHECK(18,4,0) || \
+  JSON_HEDLEY_INTEL_VERSION_CHECK(13,0,0)
+#  define JSON_HEDLEY_WARNING(msg) JSON_HEDLEY_PRAGMA(GCC warning msg)
+#elif \
+  JSON_HEDLEY_MSVC_VERSION_CHECK(15,0,0) || \
+  JSON_HEDLEY_INTEL_CL_VERSION_CHECK(2021,1,0)
+#  define JSON_HEDLEY_WARNING(msg) JSON_HEDLEY_PRAGMA(message(msg))
+#else
+#  define JSON_HEDLEY_WARNING(msg) JSON_HEDLEY_MESSAGE(msg)
+#endif
+
+#if defined(JSON_HEDLEY_REQUIRE)
+    #undef JSON_HEDLEY_REQUIRE
+#endif
+#if defined(JSON_HEDLEY_REQUIRE_MSG)
+    #undef JSON_HEDLEY_REQUIRE_MSG
+#endif
+#if JSON_HEDLEY_HAS_ATTRIBUTE(diagnose_if)
+#  if JSON_HEDLEY_HAS_WARNING("-Wgcc-compat")
+#    define JSON_HEDLEY_REQUIRE(expr) \
+    JSON_HEDLEY_DIAGNOSTIC_PUSH \
+    _Pragma("clang diagnostic ignored \"-Wgcc-compat\"") \
+    __attribute__((diagnose_if(!(expr), #expr, "error"))) \
+    JSON_HEDLEY_DIAGNOSTIC_POP
+#    define JSON_HEDLEY_REQUIRE_MSG(expr,msg) \
+    JSON_HEDLEY_DIAGNOSTIC_PUSH \
+    _Pragma("clang diagnostic ignored \"-Wgcc-compat\"") \
+    __attribute__((diagnose_if(!(expr), msg, "error"))) \
+    JSON_HEDLEY_DIAGNOSTIC_POP
+#  else
+#    define JSON_HEDLEY_REQUIRE(expr) __attribute__((diagnose_if(!(expr), #expr, "error")))
+#    define JSON_HEDLEY_REQUIRE_MSG(expr,msg) __attribute__((diagnose_if(!(expr), msg, "error")))
+#  endif
+#else
+#  define JSON_HEDLEY_REQUIRE(expr)
+#  define JSON_HEDLEY_REQUIRE_MSG(expr,msg)
+#endif
+
+#if defined(JSON_HEDLEY_FLAGS)
+    #undef JSON_HEDLEY_FLAGS
+#endif
+#if JSON_HEDLEY_HAS_ATTRIBUTE(flag_enum) && (!defined(__cplusplus) || JSON_HEDLEY_HAS_WARNING("-Wbitfield-enum-conversion"))
+    #define JSON_HEDLEY_FLAGS __attribute__((__flag_enum__))
+#else
+    #define JSON_HEDLEY_FLAGS
+#endif
+
+#if defined(JSON_HEDLEY_FLAGS_CAST)
+    #undef JSON_HEDLEY_FLAGS_CAST
+#endif
+#if JSON_HEDLEY_INTEL_VERSION_CHECK(19,0,0)
+#  define JSON_HEDLEY_FLAGS_CAST(T, expr) (__extension__ ({ \
+        JSON_HEDLEY_DIAGNOSTIC_PUSH \
+        _Pragma("warning(disable:188)") \
+        ((T) (expr)); \
+        JSON_HEDLEY_DIAGNOSTIC_POP \
+    }))
+#else
+#  define JSON_HEDLEY_FLAGS_CAST(T, expr) JSON_HEDLEY_STATIC_CAST(T, expr)
+#endif
+
+#if defined(JSON_HEDLEY_EMPTY_BASES)
+    #undef JSON_HEDLEY_EMPTY_BASES
+#endif
+#if \
+    (JSON_HEDLEY_MSVC_VERSION_CHECK(19,0,23918) && !JSON_HEDLEY_MSVC_VERSION_CHECK(20,0,0)) || \
+    JSON_HEDLEY_INTEL_CL_VERSION_CHECK(2021,1,0)
+    #define JSON_HEDLEY_EMPTY_BASES __declspec(empty_bases)
+#else
+    #define JSON_HEDLEY_EMPTY_BASES
+#endif
+
+/* Remaining macros are deprecated. */
+
+#if defined(JSON_HEDLEY_GCC_NOT_CLANG_VERSION_CHECK)
+    #undef JSON_HEDLEY_GCC_NOT_CLANG_VERSION_CHECK
+#endif
+#if defined(__clang__)
+    #define JSON_HEDLEY_GCC_NOT_CLANG_VERSION_CHECK(major,minor,patch) (0)
+#else
+    #define JSON_HEDLEY_GCC_NOT_CLANG_VERSION_CHECK(major,minor,patch) JSON_HEDLEY_GCC_VERSION_CHECK(major,minor,patch)
+#endif
+
+#if defined(JSON_HEDLEY_CLANG_HAS_ATTRIBUTE)
+    #undef JSON_HEDLEY_CLANG_HAS_ATTRIBUTE
+#endif
+#define JSON_HEDLEY_CLANG_HAS_ATTRIBUTE(attribute) JSON_HEDLEY_HAS_ATTRIBUTE(attribute)
+
+#if defined(JSON_HEDLEY_CLANG_HAS_CPP_ATTRIBUTE)
+    #undef JSON_HEDLEY_CLANG_HAS_CPP_ATTRIBUTE
+#endif
+#define JSON_HEDLEY_CLANG_HAS_CPP_ATTRIBUTE(attribute) JSON_HEDLEY_HAS_CPP_ATTRIBUTE(attribute)
+
+#if defined(JSON_HEDLEY_CLANG_HAS_BUILTIN)
+    #undef JSON_HEDLEY_CLANG_HAS_BUILTIN
+#endif
+#define JSON_HEDLEY_CLANG_HAS_BUILTIN(builtin) JSON_HEDLEY_HAS_BUILTIN(builtin)
+
+#if defined(JSON_HEDLEY_CLANG_HAS_FEATURE)
+    #undef JSON_HEDLEY_CLANG_HAS_FEATURE
+#endif
+#define JSON_HEDLEY_CLANG_HAS_FEATURE(feature) JSON_HEDLEY_HAS_FEATURE(feature)
+
+#if defined(JSON_HEDLEY_CLANG_HAS_EXTENSION)
+    #undef JSON_HEDLEY_CLANG_HAS_EXTENSION
+#endif
+#define JSON_HEDLEY_CLANG_HAS_EXTENSION(extension) JSON_HEDLEY_HAS_EXTENSION(extension)
+
+#if defined(JSON_HEDLEY_CLANG_HAS_DECLSPEC_DECLSPEC_ATTRIBUTE)
+    #undef JSON_HEDLEY_CLANG_HAS_DECLSPEC_DECLSPEC_ATTRIBUTE
+#endif
+#define JSON_HEDLEY_CLANG_HAS_DECLSPEC_ATTRIBUTE(attribute) JSON_HEDLEY_HAS_DECLSPEC_ATTRIBUTE(attribute)
+
+#if defined(JSON_HEDLEY_CLANG_HAS_WARNING)
+    #undef JSON_HEDLEY_CLANG_HAS_WARNING
+#endif
+#define JSON_HEDLEY_CLANG_HAS_WARNING(warning) JSON_HEDLEY_HAS_WARNING(warning)
+
+#endif /* !defined(JSON_HEDLEY_VERSION) || (JSON_HEDLEY_VERSION < X) */
+
+// #include <nlohmann/detail/meta/detected.hpp>
+
+
+#include <type_traits>
+
+// #include <nlohmann/detail/meta/void_t.hpp>
+
+
+namespace nlohmann
+{
+namespace detail
+{
+template<typename ...Ts> struct make_void
+{
+    using type = void;
+};
+template<typename ...Ts> using void_t = typename make_void<Ts...>::type;
+} // namespace detail
+}  // namespace nlohmann
+
+
+// https://en.cppreference.com/w/cpp/experimental/is_detected
+namespace nlohmann
+{
+namespace detail
+{
+struct nonesuch
+{
+    nonesuch() = delete;
+    ~nonesuch() = delete;
+    nonesuch(nonesuch const&) = delete;
+    nonesuch(nonesuch const&&) = delete;
+    void operator=(nonesuch const&) = delete;
+    void operator=(nonesuch&&) = delete;
+};
+
+template<class Default,
+         class AlwaysVoid,
+         template<class...> class Op,
+         class... Args>
+struct detector
+{
+    using value_t = std::false_type;
+    using type = Default;
+};
+
+template<class Default, template<class...> class Op, class... Args>
+struct detector<Default, void_t<Op<Args...>>, Op, Args...>
+{
+    using value_t = std::true_type;
+    using type = Op<Args...>;
+};
+
+template<template<class...> class Op, class... Args>
+using is_detected = typename detector<nonesuch, void, Op, Args...>::value_t;
+
+template<template<class...> class Op, class... Args>
+struct is_detected_lazy : is_detected<Op, Args...> { };
+
+template<template<class...> class Op, class... Args>
+using detected_t = typename detector<nonesuch, void, Op, Args...>::type;
+
+template<class Default, template<class...> class Op, class... Args>
+using detected_or = detector<Default, void, Op, Args...>;
+
+template<class Default, template<class...> class Op, class... Args>
+using detected_or_t = typename detected_or<Default, Op, Args...>::type;
+
+template<class Expected, template<class...> class Op, class... Args>
+using is_detected_exact = std::is_same<Expected, detected_t<Op, Args...>>;
+
+template<class To, template<class...> class Op, class... Args>
+using is_detected_convertible =
+    std::is_convertible<detected_t<Op, Args...>, To>;
+}  // namespace detail
+}  // namespace nlohmann
+
+
+// This file contains all internal macro definitions
+// You MUST include macro_unscope.hpp at the end of json.hpp to undef all of them
+
+// exclude unsupported compilers
+#if !defined(JSON_SKIP_UNSUPPORTED_COMPILER_CHECK)
+    #if defined(__clang__)
+        #if (__clang_major__ * 10000 + __clang_minor__ * 100 + __clang_patchlevel__) < 30400
+            #error "unsupported Clang version - see https://github.com/nlohmann/json#supported-compilers"
+        #endif
+    #elif defined(__GNUC__) && !(defined(__ICC) || defined(__INTEL_COMPILER))
+        #if (__GNUC__ * 10000 + __GNUC_MINOR__ * 100 + __GNUC_PATCHLEVEL__) < 40800
+            #error "unsupported GCC version - see https://github.com/nlohmann/json#supported-compilers"
+        #endif
+    #endif
+#endif
+
+// C++ language standard detection
+// if the user manually specified the used c++ version this is skipped
+#if !defined(JSON_HAS_CPP_20) && !defined(JSON_HAS_CPP_17) && !defined(JSON_HAS_CPP_14) && !defined(JSON_HAS_CPP_11)
+    #if (defined(__cplusplus) && __cplusplus >= 202002L) || (defined(_MSVC_LANG) && _MSVC_LANG >= 202002L)
+        #define JSON_HAS_CPP_20
+        #define JSON_HAS_CPP_17
+        #define JSON_HAS_CPP_14
+    #elif (defined(__cplusplus) && __cplusplus >= 201703L) || (defined(_HAS_CXX17) && _HAS_CXX17 == 1) // fix for issue #464
+        #define JSON_HAS_CPP_17
+        #define JSON_HAS_CPP_14
+    #elif (defined(__cplusplus) && __cplusplus >= 201402L) || (defined(_HAS_CXX14) && _HAS_CXX14 == 1)
+        #define JSON_HAS_CPP_14
+    #endif
+    // the cpp 11 flag is always specified because it is the minimal required version
+    #define JSON_HAS_CPP_11
+#endif
+
+#if !defined(JSON_HAS_FILESYSTEM) && !defined(JSON_HAS_EXPERIMENTAL_FILESYSTEM)
+    #ifdef JSON_HAS_CPP_17
+        #if defined(__cpp_lib_filesystem)
+            #define JSON_HAS_FILESYSTEM 1
+        #elif defined(__cpp_lib_experimental_filesystem)
+            #define JSON_HAS_EXPERIMENTAL_FILESYSTEM 1
+        #elif !defined(__has_include)
+            #define JSON_HAS_EXPERIMENTAL_FILESYSTEM 1
+        #elif __has_include(<filesystem>)
+            #define JSON_HAS_FILESYSTEM 1
+        #elif __has_include(<experimental/filesystem>)
+            #define JSON_HAS_EXPERIMENTAL_FILESYSTEM 1
+        #endif
+
+        // std::filesystem does not work on MinGW GCC 8: https://sourceforge.net/p/mingw-w64/bugs/737/
+        #if defined(__MINGW32__) && defined(__GNUC__) && __GNUC__ == 8
+            #undef JSON_HAS_FILESYSTEM
+            #undef JSON_HAS_EXPERIMENTAL_FILESYSTEM
+        #endif
+
+        // no filesystem support before GCC 8: https://en.cppreference.com/w/cpp/compiler_support
+        #if defined(__GNUC__) && !defined(__clang__) && __GNUC__ < 8
+            #undef JSON_HAS_FILESYSTEM
+            #undef JSON_HAS_EXPERIMENTAL_FILESYSTEM
+        #endif
+
+        // no filesystem support before Clang 7: https://en.cppreference.com/w/cpp/compiler_support
+        #if defined(__clang_major__) && __clang_major__ < 7
+            #undef JSON_HAS_FILESYSTEM
+            #undef JSON_HAS_EXPERIMENTAL_FILESYSTEM
+        #endif
+
+        // no filesystem support before MSVC 19.14: https://en.cppreference.com/w/cpp/compiler_support
+        #if defined(_MSC_VER) && _MSC_VER < 1940
+            #undef JSON_HAS_FILESYSTEM
+            #undef JSON_HAS_EXPERIMENTAL_FILESYSTEM
+        #endif
+
+        // no filesystem support before iOS 13
+        #if defined(__IPHONE_OS_VERSION_MIN_REQUIRED) && __IPHONE_OS_VERSION_MIN_REQUIRED < 130000
+            #undef JSON_HAS_FILESYSTEM
+            #undef JSON_HAS_EXPERIMENTAL_FILESYSTEM
+        #endif
+
+        // no filesystem support before macOS Catalina
+        #if defined(__MAC_OS_X_VERSION_MIN_REQUIRED) && __MAC_OS_X_VERSION_MIN_REQUIRED < 101500
+            #undef JSON_HAS_FILESYSTEM
+            #undef JSON_HAS_EXPERIMENTAL_FILESYSTEM
+        #endif
+    #endif
+#endif
+
+#ifndef JSON_HAS_EXPERIMENTAL_FILESYSTEM
+    #define JSON_HAS_EXPERIMENTAL_FILESYSTEM 0
+#endif
+
+#ifndef JSON_HAS_FILESYSTEM
+    #define JSON_HAS_FILESYSTEM 0
+#endif
+
+// disable documentation warnings on clang
+#if defined(__clang__)
+    #pragma clang diagnostic push
+    #pragma clang diagnostic ignored "-Wdocumentation"
+    #pragma clang diagnostic ignored "-Wdocumentation-unknown-command"
+#endif
+
+// allow disabling exceptions
+#if (defined(__cpp_exceptions) || defined(__EXCEPTIONS) || defined(_CPPUNWIND)) && !defined(JSON_NOEXCEPTION)
+    #define JSON_THROW(exception) throw exception
+    #define JSON_TRY try
+    #define JSON_CATCH(exception) catch(exception)
+    #define JSON_INTERNAL_CATCH(exception) catch(exception)
+#else
+    #include <cstdlib>
+    #define JSON_THROW(exception) std::abort()
+    #define JSON_TRY if(true)
+    #define JSON_CATCH(exception) if(false)
+    #define JSON_INTERNAL_CATCH(exception) if(false)
+#endif
+
+// override exception macros
+#if defined(JSON_THROW_USER)
+    #undef JSON_THROW
+    #define JSON_THROW JSON_THROW_USER
+#endif
+#if defined(JSON_TRY_USER)
+    #undef JSON_TRY
+    #define JSON_TRY JSON_TRY_USER
+#endif
+#if defined(JSON_CATCH_USER)
+    #undef JSON_CATCH
+    #define JSON_CATCH JSON_CATCH_USER
+    #undef JSON_INTERNAL_CATCH
+    #define JSON_INTERNAL_CATCH JSON_CATCH_USER
+#endif
+#if defined(JSON_INTERNAL_CATCH_USER)
+    #undef JSON_INTERNAL_CATCH
+    #define JSON_INTERNAL_CATCH JSON_INTERNAL_CATCH_USER
+#endif
+
+// allow overriding assert
+#if !defined(JSON_ASSERT)
+    #include <cassert> // assert
+    #define JSON_ASSERT(x) assert(x)
+#endif
+
+// allow to access some private functions (needed by the test suite)
+#if defined(JSON_TESTS_PRIVATE)
+    #define JSON_PRIVATE_UNLESS_TESTED public
+#else
+    #define JSON_PRIVATE_UNLESS_TESTED private
+#endif
+
+/*!
+@brief macro to briefly define a mapping between an enum and JSON
+@def NLOHMANN_JSON_SERIALIZE_ENUM
+@since version 3.4.0
+*/
+#define NLOHMANN_JSON_SERIALIZE_ENUM(ENUM_TYPE, ...)                                            \
+    template<typename BasicJsonType>                                                            \
+    inline void to_json(BasicJsonType& j, const ENUM_TYPE& e)                                   \
+    {                                                                                           \
+        static_assert(std::is_enum<ENUM_TYPE>::value, #ENUM_TYPE " must be an enum!");          \
+        static const std::pair<ENUM_TYPE, BasicJsonType> m[] = __VA_ARGS__;                     \
+        auto it = std::find_if(std::begin(m), std::end(m),                                      \
+                               [e](const std::pair<ENUM_TYPE, BasicJsonType>& ej_pair) -> bool  \
+        {                                                                                       \
+            return ej_pair.first == e;                                                          \
+        });                                                                                     \
+        j = ((it != std::end(m)) ? it : std::begin(m))->second;                                 \
+    }                                                                                           \
+    template<typename BasicJsonType>                                                            \
+    inline void from_json(const BasicJsonType& j, ENUM_TYPE& e)                                 \
+    {                                                                                           \
+        static_assert(std::is_enum<ENUM_TYPE>::value, #ENUM_TYPE " must be an enum!");          \
+        static const std::pair<ENUM_TYPE, BasicJsonType> m[] = __VA_ARGS__;                     \
+        auto it = std::find_if(std::begin(m), std::end(m),                                      \
+                               [&j](const std::pair<ENUM_TYPE, BasicJsonType>& ej_pair) -> bool \
+        {                                                                                       \
+            return ej_pair.second == j;                                                         \
+        });                                                                                     \
+        e = ((it != std::end(m)) ? it : std::begin(m))->first;                                  \
+    }
+
+// Ugly macros to avoid uglier copy-paste when specializing basic_json. They
+// may be removed in the future once the class is split.
+
+#define NLOHMANN_BASIC_JSON_TPL_DECLARATION                                \
+    template<template<typename, typename, typename...> class ObjectType,   \
+             template<typename, typename...> class ArrayType,              \
+             class StringType, class BooleanType, class NumberIntegerType, \
+             class NumberUnsignedType, class NumberFloatType,              \
+             template<typename> class AllocatorType,                       \
+             template<typename, typename = void> class JSONSerializer,     \
+             class BinaryType>
+
+#define NLOHMANN_BASIC_JSON_TPL                                            \
+    basic_json<ObjectType, ArrayType, StringType, BooleanType,             \
+    NumberIntegerType, NumberUnsignedType, NumberFloatType,                \
+    AllocatorType, JSONSerializer, BinaryType>
+
+// Macros to simplify conversion from/to types
+
+#define NLOHMANN_JSON_EXPAND( x ) x
+#define NLOHMANN_JSON_GET_MACRO(_1, _2, _3, _4, _5, _6, _7, _8, _9, _10, _11, _12, _13, _14, _15, _16, _17, _18, _19, _20, _21, _22, _23, _24, _25, _26, _27, _28, _29, _30, _31, _32, _33, _34, _35, _36, _37, _38, _39, _40, _41, _42, _43, _44, _45, _46, _47, _48, _49, _50, _51, _52, _53, _54, _55, _56, _57, _58, _59, _60, _61, _62, _63, _64, NAME,...) NAME
+#define NLOHMANN_JSON_PASTE(...) NLOHMANN_JSON_EXPAND(NLOHMANN_JSON_GET_MACRO(__VA_ARGS__, \
+        NLOHMANN_JSON_PASTE64, \
+        NLOHMANN_JSON_PASTE63, \
+        NLOHMANN_JSON_PASTE62, \
+        NLOHMANN_JSON_PASTE61, \
+        NLOHMANN_JSON_PASTE60, \
+        NLOHMANN_JSON_PASTE59, \
+        NLOHMANN_JSON_PASTE58, \
+        NLOHMANN_JSON_PASTE57, \
+        NLOHMANN_JSON_PASTE56, \
+        NLOHMANN_JSON_PASTE55, \
+        NLOHMANN_JSON_PASTE54, \
+        NLOHMANN_JSON_PASTE53, \
+        NLOHMANN_JSON_PASTE52, \
+        NLOHMANN_JSON_PASTE51, \
+        NLOHMANN_JSON_PASTE50, \
+        NLOHMANN_JSON_PASTE49, \
+        NLOHMANN_JSON_PASTE48, \
+        NLOHMANN_JSON_PASTE47, \
+        NLOHMANN_JSON_PASTE46, \
+        NLOHMANN_JSON_PASTE45, \
+        NLOHMANN_JSON_PASTE44, \
+        NLOHMANN_JSON_PASTE43, \
+        NLOHMANN_JSON_PASTE42, \
+        NLOHMANN_JSON_PASTE41, \
+        NLOHMANN_JSON_PASTE40, \
+        NLOHMANN_JSON_PASTE39, \
+        NLOHMANN_JSON_PASTE38, \
+        NLOHMANN_JSON_PASTE37, \
+        NLOHMANN_JSON_PASTE36, \
+        NLOHMANN_JSON_PASTE35, \
+        NLOHMANN_JSON_PASTE34, \
+        NLOHMANN_JSON_PASTE33, \
+        NLOHMANN_JSON_PASTE32, \
+        NLOHMANN_JSON_PASTE31, \
+        NLOHMANN_JSON_PASTE30, \
+        NLOHMANN_JSON_PASTE29, \
+        NLOHMANN_JSON_PASTE28, \
+        NLOHMANN_JSON_PASTE27, \
+        NLOHMANN_JSON_PASTE26, \
+        NLOHMANN_JSON_PASTE25, \
+        NLOHMANN_JSON_PASTE24, \
+        NLOHMANN_JSON_PASTE23, \
+        NLOHMANN_JSON_PASTE22, \
+        NLOHMANN_JSON_PASTE21, \
+        NLOHMANN_JSON_PASTE20, \
+        NLOHMANN_JSON_PASTE19, \
+        NLOHMANN_JSON_PASTE18, \
+        NLOHMANN_JSON_PASTE17, \
+        NLOHMANN_JSON_PASTE16, \
+        NLOHMANN_JSON_PASTE15, \
+        NLOHMANN_JSON_PASTE14, \
+        NLOHMANN_JSON_PASTE13, \
+        NLOHMANN_JSON_PASTE12, \
+        NLOHMANN_JSON_PASTE11, \
+        NLOHMANN_JSON_PASTE10, \
+        NLOHMANN_JSON_PASTE9, \
+        NLOHMANN_JSON_PASTE8, \
+        NLOHMANN_JSON_PASTE7, \
+        NLOHMANN_JSON_PASTE6, \
+        NLOHMANN_JSON_PASTE5, \
+        NLOHMANN_JSON_PASTE4, \
+        NLOHMANN_JSON_PASTE3, \
+        NLOHMANN_JSON_PASTE2, \
+        NLOHMANN_JSON_PASTE1)(__VA_ARGS__))
+#define NLOHMANN_JSON_PASTE2(func, v1) func(v1)
+#define NLOHMANN_JSON_PASTE3(func, v1, v2) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE2(func, v2)
+#define NLOHMANN_JSON_PASTE4(func, v1, v2, v3) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE3(func, v2, v3)
+#define NLOHMANN_JSON_PASTE5(func, v1, v2, v3, v4) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE4(func, v2, v3, v4)
+#define NLOHMANN_JSON_PASTE6(func, v1, v2, v3, v4, v5) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE5(func, v2, v3, v4, v5)
+#define NLOHMANN_JSON_PASTE7(func, v1, v2, v3, v4, v5, v6) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE6(func, v2, v3, v4, v5, v6)
+#define NLOHMANN_JSON_PASTE8(func, v1, v2, v3, v4, v5, v6, v7) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE7(func, v2, v3, v4, v5, v6, v7)
+#define NLOHMANN_JSON_PASTE9(func, v1, v2, v3, v4, v5, v6, v7, v8) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE8(func, v2, v3, v4, v5, v6, v7, v8)
+#define NLOHMANN_JSON_PASTE10(func, v1, v2, v3, v4, v5, v6, v7, v8, v9) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE9(func, v2, v3, v4, v5, v6, v7, v8, v9)
+#define NLOHMANN_JSON_PASTE11(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE10(func, v2, v3, v4, v5, v6, v7, v8, v9, v10)
+#define NLOHMANN_JSON_PASTE12(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE11(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11)
+#define NLOHMANN_JSON_PASTE13(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE12(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12)
+#define NLOHMANN_JSON_PASTE14(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE13(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13)
+#define NLOHMANN_JSON_PASTE15(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE14(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14)
+#define NLOHMANN_JSON_PASTE16(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE15(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15)
+#define NLOHMANN_JSON_PASTE17(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE16(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16)
+#define NLOHMANN_JSON_PASTE18(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE17(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17)
+#define NLOHMANN_JSON_PASTE19(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE18(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18)
+#define NLOHMANN_JSON_PASTE20(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE19(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19)
+#define NLOHMANN_JSON_PASTE21(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE20(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20)
+#define NLOHMANN_JSON_PASTE22(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE21(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21)
+#define NLOHMANN_JSON_PASTE23(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE22(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22)
+#define NLOHMANN_JSON_PASTE24(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE23(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23)
+#define NLOHMANN_JSON_PASTE25(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE24(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24)
+#define NLOHMANN_JSON_PASTE26(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE25(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25)
+#define NLOHMANN_JSON_PASTE27(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE26(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26)
+#define NLOHMANN_JSON_PASTE28(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE27(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27)
+#define NLOHMANN_JSON_PASTE29(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE28(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28)
+#define NLOHMANN_JSON_PASTE30(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE29(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29)
+#define NLOHMANN_JSON_PASTE31(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE30(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30)
+#define NLOHMANN_JSON_PASTE32(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE31(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31)
+#define NLOHMANN_JSON_PASTE33(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE32(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32)
+#define NLOHMANN_JSON_PASTE34(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE33(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33)
+#define NLOHMANN_JSON_PASTE35(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE34(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34)
+#define NLOHMANN_JSON_PASTE36(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE35(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35)
+#define NLOHMANN_JSON_PASTE37(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE36(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36)
+#define NLOHMANN_JSON_PASTE38(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE37(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37)
+#define NLOHMANN_JSON_PASTE39(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE38(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38)
+#define NLOHMANN_JSON_PASTE40(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE39(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39)
+#define NLOHMANN_JSON_PASTE41(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE40(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40)
+#define NLOHMANN_JSON_PASTE42(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE41(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41)
+#define NLOHMANN_JSON_PASTE43(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE42(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42)
+#define NLOHMANN_JSON_PASTE44(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE43(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43)
+#define NLOHMANN_JSON_PASTE45(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE44(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44)
+#define NLOHMANN_JSON_PASTE46(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE45(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45)
+#define NLOHMANN_JSON_PASTE47(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE46(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46)
+#define NLOHMANN_JSON_PASTE48(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE47(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47)
+#define NLOHMANN_JSON_PASTE49(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE48(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48)
+#define NLOHMANN_JSON_PASTE50(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE49(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49)
+#define NLOHMANN_JSON_PASTE51(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE50(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50)
+#define NLOHMANN_JSON_PASTE52(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE51(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51)
+#define NLOHMANN_JSON_PASTE53(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE52(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52)
+#define NLOHMANN_JSON_PASTE54(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE53(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53)
+#define NLOHMANN_JSON_PASTE55(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53, v54) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE54(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53, v54)
+#define NLOHMANN_JSON_PASTE56(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53, v54, v55) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE55(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53, v54, v55)
+#define NLOHMANN_JSON_PASTE57(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53, v54, v55, v56) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE56(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53, v54, v55, v56)
+#define NLOHMANN_JSON_PASTE58(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53, v54, v55, v56, v57) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE57(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53, v54, v55, v56, v57)
+#define NLOHMANN_JSON_PASTE59(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53, v54, v55, v56, v57, v58) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE58(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53, v54, v55, v56, v57, v58)
+#define NLOHMANN_JSON_PASTE60(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53, v54, v55, v56, v57, v58, v59) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE59(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53, v54, v55, v56, v57, v58, v59)
+#define NLOHMANN_JSON_PASTE61(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53, v54, v55, v56, v57, v58, v59, v60) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE60(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53, v54, v55, v56, v57, v58, v59, v60)
+#define NLOHMANN_JSON_PASTE62(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53, v54, v55, v56, v57, v58, v59, v60, v61) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE61(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53, v54, v55, v56, v57, v58, v59, v60, v61)
+#define NLOHMANN_JSON_PASTE63(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53, v54, v55, v56, v57, v58, v59, v60, v61, v62) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE62(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53, v54, v55, v56, v57, v58, v59, v60, v61, v62)
+#define NLOHMANN_JSON_PASTE64(func, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53, v54, v55, v56, v57, v58, v59, v60, v61, v62, v63) NLOHMANN_JSON_PASTE2(func, v1) NLOHMANN_JSON_PASTE63(func, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15, v16, v17, v18, v19, v20, v21, v22, v23, v24, v25, v26, v27, v28, v29, v30, v31, v32, v33, v34, v35, v36, v37, v38, v39, v40, v41, v42, v43, v44, v45, v46, v47, v48, v49, v50, v51, v52, v53, v54, v55, v56, v57, v58, v59, v60, v61, v62, v63)
+
+#define NLOHMANN_JSON_TO(v1) nlohmann_json_j[#v1] = nlohmann_json_t.v1;
+#define NLOHMANN_JSON_FROM(v1) nlohmann_json_j.at(#v1).get_to(nlohmann_json_t.v1);
+
+/*!
+@brief macro
+@def NLOHMANN_DEFINE_TYPE_INTRUSIVE
+@since version 3.9.0
+*/
+#define NLOHMANN_DEFINE_TYPE_INTRUSIVE(Type, ...)  \
+    friend void to_json(nlohmann::json& nlohmann_json_j, const Type& nlohmann_json_t) { NLOHMANN_JSON_EXPAND(NLOHMANN_JSON_PASTE(NLOHMANN_JSON_TO, __VA_ARGS__)) } \
+    friend void from_json(const nlohmann::json& nlohmann_json_j, Type& nlohmann_json_t) { NLOHMANN_JSON_EXPAND(NLOHMANN_JSON_PASTE(NLOHMANN_JSON_FROM, __VA_ARGS__)) }
+
+/*!
+@brief macro
+@def NLOHMANN_DEFINE_TYPE_NON_INTRUSIVE
+@since version 3.9.0
+*/
+#define NLOHMANN_DEFINE_TYPE_NON_INTRUSIVE(Type, ...)  \
+    inline void to_json(nlohmann::json& nlohmann_json_j, const Type& nlohmann_json_t) { NLOHMANN_JSON_EXPAND(NLOHMANN_JSON_PASTE(NLOHMANN_JSON_TO, __VA_ARGS__)) } \
+    inline void from_json(const nlohmann::json& nlohmann_json_j, Type& nlohmann_json_t) { NLOHMANN_JSON_EXPAND(NLOHMANN_JSON_PASTE(NLOHMANN_JSON_FROM, __VA_ARGS__)) }
+
+
+// inspired from https://stackoverflow.com/a/26745591
+// allows to call any std function as if (e.g. with begin):
+// using std::begin; begin(x);
+//
+// it allows using the detected idiom to retrieve the return type
+// of such an expression
+#define NLOHMANN_CAN_CALL_STD_FUNC_IMPL(std_name)                                 \
+    namespace detail {                                                            \
+    using std::std_name;                                                          \
+    \
+    template<typename... T>                                                       \
+    using result_of_##std_name = decltype(std_name(std::declval<T>()...));        \
+    }                                                                             \
+    \
+    namespace detail2 {                                                           \
+    struct std_name##_tag                                                         \
+    {                                                                             \
+    };                                                                            \
+    \
+    template<typename... T>                                                       \
+    std_name##_tag std_name(T&&...);                                              \
+    \
+    template<typename... T>                                                       \
+    using result_of_##std_name = decltype(std_name(std::declval<T>()...));        \
+    \
+    template<typename... T>                                                       \
+    struct would_call_std_##std_name                                              \
+    {                                                                             \
+        static constexpr auto const value = ::nlohmann::detail::                  \
+                                            is_detected_exact<std_name##_tag, result_of_##std_name, T...>::value; \
+    };                                                                            \
+    } /* namespace detail2 */ \
+    \
+    template<typename... T>                                                       \
+    struct would_call_std_##std_name : detail2::would_call_std_##std_name<T...>   \
+    {                                                                             \
+    }
+
+#ifndef JSON_USE_IMPLICIT_CONVERSIONS
+    #define JSON_USE_IMPLICIT_CONVERSIONS 1
+#endif
+
+#if JSON_USE_IMPLICIT_CONVERSIONS
+    #define JSON_EXPLICIT
+#else
+    #define JSON_EXPLICIT explicit
+#endif
+
+#ifndef JSON_DIAGNOSTICS
+    #define JSON_DIAGNOSTICS 0
+#endif
+
+
+namespace nlohmann
+{
+namespace detail
+{
+
+/*!
+@brief replace all occurrences of a substring by another string
+
+@param[in,out] s  the string to manipulate; changed so that all
+               occurrences of @a f are replaced with @a t
+@param[in]     f  the substring to replace with @a t
+@param[in]     t  the string to replace @a f
+
+@pre The search string @a f must not be empty. **This precondition is
+enforced with an assertion.**
+
+@since version 2.0.0
+*/
+inline void replace_substring(std::string& s, const std::string& f,
+                              const std::string& t)
+{
+    JSON_ASSERT(!f.empty());
+    for (auto pos = s.find(f);                // find first occurrence of f
+            pos != std::string::npos;         // make sure f was found
+            s.replace(pos, f.size(), t),      // replace with t, and
+            pos = s.find(f, pos + t.size()))  // find next occurrence of f
+    {}
 }
 
-// #include <nlohmann/detail/conversions/from_json.hpp>
+/*!
+ * @brief string escaping as described in RFC 6901 (Sect. 4)
+ * @param[in] s string to escape
+ * @return    escaped string
+ *
+ * Note the order of escaping "~" to "~0" and "/" to "~1" is important.
+ */
+inline std::string escape(std::string s)
+{
+    replace_substring(s, "~", "~0");
+    replace_substring(s, "/", "~1");
+    return s;
+}
+
+/*!
+ * @brief string unescaping as described in RFC 6901 (Sect. 4)
+ * @param[in] s string to unescape
+ * @return    unescaped string
+ *
+ * Note the order of escaping "~1" to "/" and "~0" to "~" is important.
+ */
+static void unescape(std::string& s)
+{
+    replace_substring(s, "~1", "/");
+    replace_substring(s, "~0", "~");
+}
+
+} // namespace detail
+} // namespace nlohmann
+
+// #include <nlohmann/detail/input/position_t.hpp>
 
 
-#include <algorithm> // transform
-#include <array> // array
-#include <ciso646> // and, not
-#include <forward_list> // forward_list
-#include <iterator> // inserter, front_inserter, end
-#include <map> // map
-#include <string> // string
-#include <tuple> // tuple, make_tuple
-#include <type_traits> // is_arithmetic, is_same, is_enum, underlying_type, is_convertible
-#include <unordered_map> // unordered_map
-#include <utility> // pair, declval
-#include <valarray> // valarray
+#include <cstddef> // size_t
 
-// #include <nlohmann/detail/exceptions.hpp>
+namespace nlohmann
+{
+namespace detail
+{
+/// struct to capture the start position of the current token
+struct position_t
+{
+    /// the total number of characters read
+    std::size_t chars_read_total = 0;
+    /// the number of characters read in the current line
+    std::size_t chars_read_current_line = 0;
+    /// the number of lines read
+    std::size_t lines_read = 0;
+
+    /// conversion to size_t to preserve SAX interface
+    constexpr operator size_t() const
+    {
+        return chars_read_total;
+    }
+};
+
+} // namespace detail
+} // namespace nlohmann
+
+// #include <nlohmann/detail/macro_scope.hpp>
+
+
+namespace nlohmann
+{
+namespace detail
+{
+////////////////
+// exceptions //
+////////////////
+
+/// @brief general exception of the @ref basic_json class
+/// @sa https://json.nlohmann.me/api/basic_json/exception/
+class exception : public std::exception
+{
+  public:
+    /// returns the explanatory string
+    const char* what() const noexcept override
+    {
+        return m.what();
+    }
+
+    /// the id of the exception
+    const int id; // NOLINT(cppcoreguidelines-non-private-member-variables-in-classes)
+
+  protected:
+    JSON_HEDLEY_NON_NULL(3)
+    exception(int id_, const char* what_arg) : id(id_), m(what_arg) {} // NOLINT(bugprone-throw-keyword-missing)
+
+    static std::string name(const std::string& ename, int id_)
+    {
+        return "[json.exception." + ename + "." + std::to_string(id_) + "] ";
+    }
+
+    template<typename BasicJsonType>
+    static std::string diagnostics(const BasicJsonType& leaf_element)
+    {
+#if JSON_DIAGNOSTICS
+        std::vector<std::string> tokens;
+        for (const auto* current = &leaf_element; current->m_parent != nullptr; current = current->m_parent)
+        {
+            switch (current->m_parent->type())
+            {
+                case value_t::array:
+                {
+                    for (std::size_t i = 0; i < current->m_parent->m_value.array->size(); ++i)
+                    {
+                        if (&current->m_parent->m_value.array->operator[](i) == current)
+                        {
+                            tokens.emplace_back(std::to_string(i));
+                            break;
+                        }
+                    }
+                    break;
+                }
+
+                case value_t::object:
+                {
+                    for (const auto& element : *current->m_parent->m_value.object)
+                    {
+                        if (&element.second == current)
+                        {
+                            tokens.emplace_back(element.first.c_str());
+                            break;
+                        }
+                    }
+                    break;
+                }
+
+                case value_t::null: // LCOV_EXCL_LINE
+                case value_t::string: // LCOV_EXCL_LINE
+                case value_t::boolean: // LCOV_EXCL_LINE
+                case value_t::number_integer: // LCOV_EXCL_LINE
+                case value_t::number_unsigned: // LCOV_EXCL_LINE
+                case value_t::number_float: // LCOV_EXCL_LINE
+                case value_t::binary: // LCOV_EXCL_LINE
+                case value_t::discarded: // LCOV_EXCL_LINE
+                default:   // LCOV_EXCL_LINE
+                    break; // LCOV_EXCL_LINE
+            }
+        }
+
+        if (tokens.empty())
+        {
+            return "";
+        }
+
+        return "(" + std::accumulate(tokens.rbegin(), tokens.rend(), std::string{},
+                                     [](const std::string & a, const std::string & b)
+        {
+            return a + "/" + detail::escape(b);
+        }) + ") ";
+#else
+        static_cast<void>(leaf_element);
+        return "";
+#endif
+    }
+
+  private:
+    /// an exception object as storage for error messages
+    std::runtime_error m;
+};
+
+/// @brief exception indicating a parse error
+/// @sa https://json.nlohmann.me/api/basic_json/parse_error/
+class parse_error : public exception
+{
+  public:
+    /*!
+    @brief create a parse error exception
+    @param[in] id_       the id of the exception
+    @param[in] pos       the position where the error occurred (or with
+                         chars_read_total=0 if the position cannot be
+                         determined)
+    @param[in] what_arg  the explanatory string
+    @return parse_error object
+    */
+    template<typename BasicJsonType>
+    static parse_error create(int id_, const position_t& pos, const std::string& what_arg, const BasicJsonType& context)
+    {
+        std::string w = exception::name("parse_error", id_) + "parse error" +
+                        position_string(pos) + ": " + exception::diagnostics(context) + what_arg;
+        return {id_, pos.chars_read_total, w.c_str()};
+    }
+
+    template<typename BasicJsonType>
+    static parse_error create(int id_, std::size_t byte_, const std::string& what_arg, const BasicJsonType& context)
+    {
+        std::string w = exception::name("parse_error", id_) + "parse error" +
+                        (byte_ != 0 ? (" at byte " + std::to_string(byte_)) : "") +
+                        ": " + exception::diagnostics(context) + what_arg;
+        return {id_, byte_, w.c_str()};
+    }
+
+    /*!
+    @brief byte index of the parse error
+
+    The byte index of the last read character in the input file.
+
+    @note For an input with n bytes, 1 is the index of the first character and
+          n+1 is the index of the terminating null byte or the end of file.
+          This also holds true when reading a byte vector (CBOR or MessagePack).
+    */
+    const std::size_t byte;
+
+  private:
+    parse_error(int id_, std::size_t byte_, const char* what_arg)
+        : exception(id_, what_arg), byte(byte_) {}
+
+    static std::string position_string(const position_t& pos)
+    {
+        return " at line " + std::to_string(pos.lines_read + 1) +
+               ", column " + std::to_string(pos.chars_read_current_line);
+    }
+};
+
+/// @brief exception indicating errors with iterators
+/// @sa https://json.nlohmann.me/api/basic_json/invalid_iterator/
+class invalid_iterator : public exception
+{
+  public:
+    template<typename BasicJsonType>
+    static invalid_iterator create(int id_, const std::string& what_arg, const BasicJsonType& context)
+    {
+        std::string w = exception::name("invalid_iterator", id_) + exception::diagnostics(context) + what_arg;
+        return {id_, w.c_str()};
+    }
+
+  private:
+    JSON_HEDLEY_NON_NULL(3)
+    invalid_iterator(int id_, const char* what_arg)
+        : exception(id_, what_arg) {}
+};
+
+/// @brief exception indicating executing a member function with a wrong type
+/// @sa https://json.nlohmann.me/api/basic_json/type_error/
+class type_error : public exception
+{
+  public:
+    template<typename BasicJsonType>
+    static type_error create(int id_, const std::string& what_arg, const BasicJsonType& context)
+    {
+        std::string w = exception::name("type_error", id_) + exception::diagnostics(context) + what_arg;
+        return {id_, w.c_str()};
+    }
+
+  private:
+    JSON_HEDLEY_NON_NULL(3)
+    type_error(int id_, const char* what_arg) : exception(id_, what_arg) {}
+};
+
+/// @brief exception indicating access out of the defined range
+/// @sa https://json.nlohmann.me/api/basic_json/out_of_range/
+class out_of_range : public exception
+{
+  public:
+    template<typename BasicJsonType>
+    static out_of_range create(int id_, const std::string& what_arg, const BasicJsonType& context)
+    {
+        std::string w = exception::name("out_of_range", id_) + exception::diagnostics(context) + what_arg;
+        return {id_, w.c_str()};
+    }
+
+  private:
+    JSON_HEDLEY_NON_NULL(3)
+    out_of_range(int id_, const char* what_arg) : exception(id_, what_arg) {}
+};
+
+/// @brief exception indicating other library errors
+/// @sa https://json.nlohmann.me/api/basic_json/other_error/
+class other_error : public exception
+{
+  public:
+    template<typename BasicJsonType>
+    static other_error create(int id_, const std::string& what_arg, const BasicJsonType& context)
+    {
+        std::string w = exception::name("other_error", id_) + exception::diagnostics(context) + what_arg;
+        return {id_, w.c_str()};
+    }
+
+  private:
+    JSON_HEDLEY_NON_NULL(3)
+    other_error(int id_, const char* what_arg) : exception(id_, what_arg) {}
+};
+
+}  // namespace detail
+}  // namespace nlohmann
 
 // #include <nlohmann/detail/macro_scope.hpp>
 
 // #include <nlohmann/detail/meta/cpp_future.hpp>
 
+
+#include <cstddef> // size_t
+#include <type_traits> // conditional, enable_if, false_type, integral_constant, is_constructible, is_integral, is_same, remove_cv, remove_reference, true_type
+#include <utility> // index_sequence, make_index_sequence, index_sequence_for
+
+// #include <nlohmann/detail/macro_scope.hpp>
+
+
+namespace nlohmann
+{
+namespace detail
+{
+
+template<typename T>
+using uncvref_t = typename std::remove_cv<typename std::remove_reference<T>::type>::type;
+
+#ifdef JSON_HAS_CPP_14
+
+// the following utilities are natively available in C++14
+using std::enable_if_t;
+using std::index_sequence;
+using std::make_index_sequence;
+using std::index_sequence_for;
+
+#else
+
+// alias templates to reduce boilerplate
+template<bool B, typename T = void>
+using enable_if_t = typename std::enable_if<B, T>::type;
+
+// The following code is taken from https://github.com/abseil/abseil-cpp/blob/10cb35e459f5ecca5b2ff107635da0bfa41011b4/absl/utility/utility.h
+// which is part of Google Abseil (https://github.com/abseil/abseil-cpp), licensed under the Apache License 2.0.
+
+//// START OF CODE FROM GOOGLE ABSEIL
+
+// integer_sequence
+//
+// Class template representing a compile-time integer sequence. An instantiation
+// of `integer_sequence<T, Ints...>` has a sequence of integers encoded in its
+// type through its template arguments (which is a common need when
+// working with C++11 variadic templates). `absl::integer_sequence` is designed
+// to be a drop-in replacement for C++14's `std::integer_sequence`.
+//
+// Example:
+//
+//   template< class T, T... Ints >
+//   void user_function(integer_sequence<T, Ints...>);
+//
+//   int main()
+//   {
+//     // user_function's `T` will be deduced to `int` and `Ints...`
+//     // will be deduced to `0, 1, 2, 3, 4`.
+//     user_function(make_integer_sequence<int, 5>());
+//   }
+template <typename T, T... Ints>
+struct integer_sequence
+{
+    using value_type = T;
+    static constexpr std::size_t size() noexcept
+    {
+        return sizeof...(Ints);
+    }
+};
+
+// index_sequence
+//
+// A helper template for an `integer_sequence` of `size_t`,
+// `absl::index_sequence` is designed to be a drop-in replacement for C++14's
+// `std::index_sequence`.
+template <size_t... Ints>
+using index_sequence = integer_sequence<size_t, Ints...>;
+
+namespace utility_internal
+{
+
+template <typename Seq, size_t SeqSize, size_t Rem>
+struct Extend;
+
+// Note that SeqSize == sizeof...(Ints). It's passed explicitly for efficiency.
+template <typename T, T... Ints, size_t SeqSize>
+struct Extend<integer_sequence<T, Ints...>, SeqSize, 0>
+{
+    using type = integer_sequence < T, Ints..., (Ints + SeqSize)... >;
+};
+
+template <typename T, T... Ints, size_t SeqSize>
+struct Extend<integer_sequence<T, Ints...>, SeqSize, 1>
+{
+    using type = integer_sequence < T, Ints..., (Ints + SeqSize)..., 2 * SeqSize >;
+};
+
+// Recursion helper for 'make_integer_sequence<T, N>'.
+// 'Gen<T, N>::type' is an alias for 'integer_sequence<T, 0, 1, ... N-1>'.
+template <typename T, size_t N>
+struct Gen
+{
+    using type =
+        typename Extend < typename Gen < T, N / 2 >::type, N / 2, N % 2 >::type;
+};
+
+template <typename T>
+struct Gen<T, 0>
+{
+    using type = integer_sequence<T>;
+};
+
+}  // namespace utility_internal
+
+// Compile-time sequences of integers
+
+// make_integer_sequence
+//
+// This template alias is equivalent to
+// `integer_sequence<int, 0, 1, ..., N-1>`, and is designed to be a drop-in
+// replacement for C++14's `std::make_integer_sequence`.
+template <typename T, T N>
+using make_integer_sequence = typename utility_internal::Gen<T, N>::type;
+
+// make_index_sequence
+//
+// This template alias is equivalent to `index_sequence<0, 1, ..., N-1>`,
+// and is designed to be a drop-in replacement for C++14's
+// `std::make_index_sequence`.
+template <size_t N>
+using make_index_sequence = make_integer_sequence<size_t, N>;
+
+// index_sequence_for
+//
+// Converts a typename pack into an index sequence of the same length, and
+// is designed to be a drop-in replacement for C++14's
+// `std::index_sequence_for()`
+template <typename... Ts>
+using index_sequence_for = make_index_sequence<sizeof...(Ts)>;
+
+//// END OF CODE FROM GOOGLE ABSEIL
+
+#endif
+
+// dispatch utility (taken from ranges-v3)
+template<unsigned N> struct priority_tag : priority_tag < N - 1 > {};
+template<> struct priority_tag<0> {};
+
+// taken from ranges-v3
+template<typename T>
+struct static_const
+{
+    static constexpr T value{};
+};
+
+template<typename T>
+constexpr T static_const<T>::value; // NOLINT(readability-redundant-declaration)
+
+}  // namespace detail
+}  // namespace nlohmann
+
+// #include <nlohmann/detail/meta/identity_tag.hpp>
+
+
+namespace nlohmann
+{
+namespace detail
+{
+// dispatching helper struct
+template <class T> struct identity_tag {};
+}  // namespace detail
+}  // namespace nlohmann
+
 // #include <nlohmann/detail/meta/type_traits.hpp>
 
+
+#include <limits> // numeric_limits
+#include <type_traits> // false_type, is_constructible, is_integral, is_same, true_type
+#include <utility> // declval
+#include <tuple> // tuple
+
+// #include <nlohmann/detail/macro_scope.hpp>
+
+
+// #include <nlohmann/detail/iterators/iterator_traits.hpp>
+
+
+#include <iterator> // random_access_iterator_tag
+
+// #include <nlohmann/detail/meta/void_t.hpp>
+
+// #include <nlohmann/detail/meta/cpp_future.hpp>
+
+
+namespace nlohmann
+{
+namespace detail
+{
+template<typename It, typename = void>
+struct iterator_types {};
+
+template<typename It>
+struct iterator_types <
+    It,
+    void_t<typename It::difference_type, typename It::value_type, typename It::pointer,
+    typename It::reference, typename It::iterator_category >>
+{
+    using difference_type = typename It::difference_type;
+    using value_type = typename It::value_type;
+    using pointer = typename It::pointer;
+    using reference = typename It::reference;
+    using iterator_category = typename It::iterator_category;
+};
+
+// This is required as some compilers implement std::iterator_traits in a way that
+// doesn't work with SFINAE. See https://github.com/nlohmann/json/issues/1341.
+template<typename T, typename = void>
+struct iterator_traits
+{
+};
+
+template<typename T>
+struct iterator_traits < T, enable_if_t < !std::is_pointer<T>::value >>
+            : iterator_types<T>
+{
+};
+
+template<typename T>
+struct iterator_traits<T*, enable_if_t<std::is_object<T>::value>>
+{
+    using iterator_category = std::random_access_iterator_tag;
+    using value_type = T;
+    using difference_type = ptrdiff_t;
+    using pointer = T*;
+    using reference = T&;
+};
+} // namespace detail
+} // namespace nlohmann
+
+// #include <nlohmann/detail/meta/call_std/begin.hpp>
+
+
+// #include <nlohmann/detail/macro_scope.hpp>
+
+
+namespace nlohmann
+{
+NLOHMANN_CAN_CALL_STD_FUNC_IMPL(begin);
+} // namespace nlohmann
+
+// #include <nlohmann/detail/meta/call_std/end.hpp>
+
+
+// #include <nlohmann/detail/macro_scope.hpp>
+
+
+namespace nlohmann
+{
+NLOHMANN_CAN_CALL_STD_FUNC_IMPL(end);
+}  // namespace nlohmann
+
+// #include <nlohmann/detail/meta/cpp_future.hpp>
+
+// #include <nlohmann/detail/meta/detected.hpp>
+
+// #include <nlohmann/json_fwd.hpp>
+#ifndef INCLUDE_NLOHMANN_JSON_FWD_HPP_
+#define INCLUDE_NLOHMANN_JSON_FWD_HPP_
+
+#include <cstdint> // int64_t, uint64_t
+#include <map> // map
+#include <memory> // allocator
+#include <string> // string
+#include <vector> // vector
+
+/*!
+@brief namespace for Niels Lohmann
+@see https://github.com/nlohmann
+@since version 1.0.0
+*/
+namespace nlohmann
+{
+/*!
+@brief default JSONSerializer template argument
+
+This serializer ignores the template arguments and uses ADL
+([argument-dependent lookup](https://en.cppreference.com/w/cpp/language/adl))
+for serialization.
+*/
+template<typename T = void, typename SFINAE = void>
+struct adl_serializer;
+
+/// a class to store JSON values
+/// @sa https://json.nlohmann.me/api/basic_json/
+template<template<typename U, typename V, typename... Args> class ObjectType =
+         std::map,
+         template<typename U, typename... Args> class ArrayType = std::vector,
+         class StringType = std::string, class BooleanType = bool,
+         class NumberIntegerType = std::int64_t,
+         class NumberUnsignedType = std::uint64_t,
+         class NumberFloatType = double,
+         template<typename U> class AllocatorType = std::allocator,
+         template<typename T, typename SFINAE = void> class JSONSerializer =
+         adl_serializer,
+         class BinaryType = std::vector<std::uint8_t>>
+class basic_json;
+
+/// @brief JSON Pointer defines a string syntax for identifying a specific value within a JSON document
+/// @sa https://json.nlohmann.me/api/json_pointer/
+template<typename BasicJsonType>
+class json_pointer;
+
+/*!
+@brief default specialization
+@sa https://json.nlohmann.me/api/json/
+*/
+using json = basic_json<>;
+
+/// @brief a minimal map-like container that preserves insertion order
+/// @sa https://json.nlohmann.me/api/ordered_map/
+template<class Key, class T, class IgnoredLess, class Allocator>
+struct ordered_map;
+
+/// @brief specialization that maintains the insertion order of object keys
+/// @sa https://json.nlohmann.me/api/ordered_json/
+using ordered_json = basic_json<nlohmann::ordered_map>;
+
+}  // namespace nlohmann
+
+#endif  // INCLUDE_NLOHMANN_JSON_FWD_HPP_
+
+
+namespace nlohmann
+{
+/*!
+@brief detail namespace with internal helper functions
+
+This namespace collects functions that should not be exposed,
+implementations of some @ref basic_json methods, and meta-programming helpers.
+
+@since version 2.1.0
+*/
+namespace detail
+{
+/////////////
+// helpers //
+/////////////
+
+// Note to maintainers:
+//
+// Every trait in this file expects a non CV-qualified type.
+// The only exceptions are in the 'aliases for detected' section
+// (i.e. those of the form: decltype(T::member_function(std::declval<T>())))
+//
+// In this case, T has to be properly CV-qualified to constraint the function arguments
+// (e.g. to_json(BasicJsonType&, const T&))
+
+template<typename> struct is_basic_json : std::false_type {};
+
+NLOHMANN_BASIC_JSON_TPL_DECLARATION
+struct is_basic_json<NLOHMANN_BASIC_JSON_TPL> : std::true_type {};
+
+//////////////////////
+// json_ref helpers //
+//////////////////////
+
+template<typename>
+class json_ref;
+
+template<typename>
+struct is_json_ref : std::false_type {};
+
+template<typename T>
+struct is_json_ref<json_ref<T>> : std::true_type {};
+
+//////////////////////////
+// aliases for detected //
+//////////////////////////
+
+template<typename T>
+using mapped_type_t = typename T::mapped_type;
+
+template<typename T>
+using key_type_t = typename T::key_type;
+
+template<typename T>
+using value_type_t = typename T::value_type;
+
+template<typename T>
+using difference_type_t = typename T::difference_type;
+
+template<typename T>
+using pointer_t = typename T::pointer;
+
+template<typename T>
+using reference_t = typename T::reference;
+
+template<typename T>
+using iterator_category_t = typename T::iterator_category;
+
+template<typename T, typename... Args>
+using to_json_function = decltype(T::to_json(std::declval<Args>()...));
+
+template<typename T, typename... Args>
+using from_json_function = decltype(T::from_json(std::declval<Args>()...));
+
+template<typename T, typename U>
+using get_template_function = decltype(std::declval<T>().template get<U>());
+
+// trait checking if JSONSerializer<T>::from_json(json const&, udt&) exists
+template<typename BasicJsonType, typename T, typename = void>
+struct has_from_json : std::false_type {};
+
+// trait checking if j.get<T> is valid
+// use this trait instead of std::is_constructible or std::is_convertible,
+// both rely on, or make use of implicit conversions, and thus fail when T
+// has several constructors/operator= (see https://github.com/nlohmann/json/issues/958)
+template <typename BasicJsonType, typename T>
+struct is_getable
+{
+    static constexpr bool value = is_detected<get_template_function, const BasicJsonType&, T>::value;
+};
+
+template<typename BasicJsonType, typename T>
+struct has_from_json < BasicJsonType, T, enable_if_t < !is_basic_json<T>::value >>
+{
+    using serializer = typename BasicJsonType::template json_serializer<T, void>;
+
+    static constexpr bool value =
+        is_detected_exact<void, from_json_function, serializer,
+        const BasicJsonType&, T&>::value;
+};
+
+// This trait checks if JSONSerializer<T>::from_json(json const&) exists
+// this overload is used for non-default-constructible user-defined-types
+template<typename BasicJsonType, typename T, typename = void>
+struct has_non_default_from_json : std::false_type {};
+
+template<typename BasicJsonType, typename T>
+struct has_non_default_from_json < BasicJsonType, T, enable_if_t < !is_basic_json<T>::value >>
+{
+    using serializer = typename BasicJsonType::template json_serializer<T, void>;
+
+    static constexpr bool value =
+        is_detected_exact<T, from_json_function, serializer,
+        const BasicJsonType&>::value;
+};
+
+// This trait checks if BasicJsonType::json_serializer<T>::to_json exists
+// Do not evaluate the trait when T is a basic_json type, to avoid template instantiation infinite recursion.
+template<typename BasicJsonType, typename T, typename = void>
+struct has_to_json : std::false_type {};
+
+template<typename BasicJsonType, typename T>
+struct has_to_json < BasicJsonType, T, enable_if_t < !is_basic_json<T>::value >>
+{
+    using serializer = typename BasicJsonType::template json_serializer<T, void>;
+
+    static constexpr bool value =
+        is_detected_exact<void, to_json_function, serializer, BasicJsonType&,
+        T>::value;
+};
+
+
+///////////////////
+// is_ functions //
+///////////////////
+
+// https://en.cppreference.com/w/cpp/types/conjunction
+template<class...> struct conjunction : std::true_type { };
+template<class B1> struct conjunction<B1> : B1 { };
+template<class B1, class... Bn>
+struct conjunction<B1, Bn...>
+: std::conditional<bool(B1::value), conjunction<Bn...>, B1>::type {};
+
+// https://en.cppreference.com/w/cpp/types/negation
+template<class B> struct negation : std::integral_constant < bool, !B::value > { };
+
+// Reimplementation of is_constructible and is_default_constructible, due to them being broken for
+// std::pair and std::tuple until LWG 2367 fix (see https://cplusplus.github.io/LWG/lwg-defects.html#2367).
+// This causes compile errors in e.g. clang 3.5 or gcc 4.9.
+template <typename T>
+struct is_default_constructible : std::is_default_constructible<T> {};
+
+template <typename T1, typename T2>
+struct is_default_constructible<std::pair<T1, T2>>
+            : conjunction<is_default_constructible<T1>, is_default_constructible<T2>> {};
+
+template <typename T1, typename T2>
+struct is_default_constructible<const std::pair<T1, T2>>
+            : conjunction<is_default_constructible<T1>, is_default_constructible<T2>> {};
+
+template <typename... Ts>
+struct is_default_constructible<std::tuple<Ts...>>
+            : conjunction<is_default_constructible<Ts>...> {};
+
+template <typename... Ts>
+struct is_default_constructible<const std::tuple<Ts...>>
+            : conjunction<is_default_constructible<Ts>...> {};
+
+
+template <typename T, typename... Args>
+struct is_constructible : std::is_constructible<T, Args...> {};
+
+template <typename T1, typename T2>
+struct is_constructible<std::pair<T1, T2>> : is_default_constructible<std::pair<T1, T2>> {};
+
+template <typename T1, typename T2>
+struct is_constructible<const std::pair<T1, T2>> : is_default_constructible<const std::pair<T1, T2>> {};
+
+template <typename... Ts>
+struct is_constructible<std::tuple<Ts...>> : is_default_constructible<std::tuple<Ts...>> {};
+
+template <typename... Ts>
+struct is_constructible<const std::tuple<Ts...>> : is_default_constructible<const std::tuple<Ts...>> {};
+
+
+template<typename T, typename = void>
+struct is_iterator_traits : std::false_type {};
+
+template<typename T>
+struct is_iterator_traits<iterator_traits<T>>
+{
+  private:
+    using traits = iterator_traits<T>;
+
+  public:
+    static constexpr auto value =
+        is_detected<value_type_t, traits>::value &&
+        is_detected<difference_type_t, traits>::value &&
+        is_detected<pointer_t, traits>::value &&
+        is_detected<iterator_category_t, traits>::value &&
+        is_detected<reference_t, traits>::value;
+};
+
+template<typename T>
+struct is_range
+{
+  private:
+    using t_ref = typename std::add_lvalue_reference<T>::type;
+
+    using iterator = detected_t<result_of_begin, t_ref>;
+    using sentinel = detected_t<result_of_end, t_ref>;
+
+    // to be 100% correct, it should use https://en.cppreference.com/w/cpp/iterator/input_or_output_iterator
+    // and https://en.cppreference.com/w/cpp/iterator/sentinel_for
+    // but reimplementing these would be too much work, as a lot of other concepts are used underneath
+    static constexpr auto is_iterator_begin =
+        is_iterator_traits<iterator_traits<iterator>>::value;
+
+  public:
+    static constexpr bool value = !std::is_same<iterator, nonesuch>::value && !std::is_same<sentinel, nonesuch>::value && is_iterator_begin;
+};
+
+template<typename R>
+using iterator_t = enable_if_t<is_range<R>::value, result_of_begin<decltype(std::declval<R&>())>>;
+
+template<typename T>
+using range_value_t = value_type_t<iterator_traits<iterator_t<T>>>;
+
+// The following implementation of is_complete_type is taken from
+// https://blogs.msdn.microsoft.com/vcblog/2015/12/02/partial-support-for-expression-sfinae-in-vs-2015-update-1/
+// and is written by Xiang Fan who agreed to using it in this library.
+
+template<typename T, typename = void>
+struct is_complete_type : std::false_type {};
+
+template<typename T>
+struct is_complete_type<T, decltype(void(sizeof(T)))> : std::true_type {};
+
+template<typename BasicJsonType, typename CompatibleObjectType,
+         typename = void>
+struct is_compatible_object_type_impl : std::false_type {};
+
+template<typename BasicJsonType, typename CompatibleObjectType>
+struct is_compatible_object_type_impl <
+    BasicJsonType, CompatibleObjectType,
+    enable_if_t < is_detected<mapped_type_t, CompatibleObjectType>::value&&
+    is_detected<key_type_t, CompatibleObjectType>::value >>
+{
+    using object_t = typename BasicJsonType::object_t;
+
+    // macOS's is_constructible does not play well with nonesuch...
+    static constexpr bool value =
+        is_constructible<typename object_t::key_type,
+        typename CompatibleObjectType::key_type>::value &&
+        is_constructible<typename object_t::mapped_type,
+        typename CompatibleObjectType::mapped_type>::value;
+};
+
+template<typename BasicJsonType, typename CompatibleObjectType>
+struct is_compatible_object_type
+    : is_compatible_object_type_impl<BasicJsonType, CompatibleObjectType> {};
+
+template<typename BasicJsonType, typename ConstructibleObjectType,
+         typename = void>
+struct is_constructible_object_type_impl : std::false_type {};
+
+template<typename BasicJsonType, typename ConstructibleObjectType>
+struct is_constructible_object_type_impl <
+    BasicJsonType, ConstructibleObjectType,
+    enable_if_t < is_detected<mapped_type_t, ConstructibleObjectType>::value&&
+    is_detected<key_type_t, ConstructibleObjectType>::value >>
+{
+    using object_t = typename BasicJsonType::object_t;
+
+    static constexpr bool value =
+        (is_default_constructible<ConstructibleObjectType>::value &&
+         (std::is_move_assignable<ConstructibleObjectType>::value ||
+          std::is_copy_assignable<ConstructibleObjectType>::value) &&
+         (is_constructible<typename ConstructibleObjectType::key_type,
+          typename object_t::key_type>::value &&
+          std::is_same <
+          typename object_t::mapped_type,
+          typename ConstructibleObjectType::mapped_type >::value)) ||
+        (has_from_json<BasicJsonType,
+         typename ConstructibleObjectType::mapped_type>::value ||
+         has_non_default_from_json <
+         BasicJsonType,
+         typename ConstructibleObjectType::mapped_type >::value);
+};
+
+template<typename BasicJsonType, typename ConstructibleObjectType>
+struct is_constructible_object_type
+    : is_constructible_object_type_impl<BasicJsonType,
+      ConstructibleObjectType> {};
+
+template<typename BasicJsonType, typename CompatibleStringType>
+struct is_compatible_string_type
+{
+    static constexpr auto value =
+        is_constructible<typename BasicJsonType::string_t, CompatibleStringType>::value;
+};
+
+template<typename BasicJsonType, typename ConstructibleStringType>
+struct is_constructible_string_type
+{
+    static constexpr auto value =
+        is_constructible<ConstructibleStringType,
+        typename BasicJsonType::string_t>::value;
+};
+
+template<typename BasicJsonType, typename CompatibleArrayType, typename = void>
+struct is_compatible_array_type_impl : std::false_type {};
+
+template<typename BasicJsonType, typename CompatibleArrayType>
+struct is_compatible_array_type_impl <
+    BasicJsonType, CompatibleArrayType,
+    enable_if_t <
+    is_detected<iterator_t, CompatibleArrayType>::value&&
+    is_iterator_traits<iterator_traits<detected_t<iterator_t, CompatibleArrayType>>>::value&&
+// special case for types like std::filesystem::path whose iterator's value_type are themselves
+// c.f. https://github.com/nlohmann/json/pull/3073
+    !std::is_same<CompatibleArrayType, detected_t<range_value_t, CompatibleArrayType>>::value >>
+{
+    static constexpr bool value =
+        is_constructible<BasicJsonType,
+        range_value_t<CompatibleArrayType>>::value;
+};
+
+template<typename BasicJsonType, typename CompatibleArrayType>
+struct is_compatible_array_type
+    : is_compatible_array_type_impl<BasicJsonType, CompatibleArrayType> {};
+
+template<typename BasicJsonType, typename ConstructibleArrayType, typename = void>
+struct is_constructible_array_type_impl : std::false_type {};
+
+template<typename BasicJsonType, typename ConstructibleArrayType>
+struct is_constructible_array_type_impl <
+    BasicJsonType, ConstructibleArrayType,
+    enable_if_t<std::is_same<ConstructibleArrayType,
+    typename BasicJsonType::value_type>::value >>
+            : std::true_type {};
+
+template<typename BasicJsonType, typename ConstructibleArrayType>
+struct is_constructible_array_type_impl <
+    BasicJsonType, ConstructibleArrayType,
+    enable_if_t < !std::is_same<ConstructibleArrayType,
+    typename BasicJsonType::value_type>::value&&
+    !is_compatible_string_type<BasicJsonType, ConstructibleArrayType>::value&&
+    is_default_constructible<ConstructibleArrayType>::value&&
+(std::is_move_assignable<ConstructibleArrayType>::value ||
+ std::is_copy_assignable<ConstructibleArrayType>::value)&&
+is_detected<iterator_t, ConstructibleArrayType>::value&&
+is_iterator_traits<iterator_traits<detected_t<iterator_t, ConstructibleArrayType>>>::value&&
+is_detected<range_value_t, ConstructibleArrayType>::value&&
+// special case for types like std::filesystem::path whose iterator's value_type are themselves
+// c.f. https://github.com/nlohmann/json/pull/3073
+!std::is_same<ConstructibleArrayType, detected_t<range_value_t, ConstructibleArrayType>>::value&&
+        is_complete_type <
+        detected_t<range_value_t, ConstructibleArrayType >>::value >>
+{
+    using value_type = range_value_t<ConstructibleArrayType>;
+
+    static constexpr bool value =
+        std::is_same<value_type,
+        typename BasicJsonType::array_t::value_type>::value ||
+        has_from_json<BasicJsonType,
+        value_type>::value ||
+        has_non_default_from_json <
+        BasicJsonType,
+        value_type >::value;
+};
+
+template<typename BasicJsonType, typename ConstructibleArrayType>
+struct is_constructible_array_type
+    : is_constructible_array_type_impl<BasicJsonType, ConstructibleArrayType> {};
+
+template<typename RealIntegerType, typename CompatibleNumberIntegerType,
+         typename = void>
+struct is_compatible_integer_type_impl : std::false_type {};
+
+template<typename RealIntegerType, typename CompatibleNumberIntegerType>
+struct is_compatible_integer_type_impl <
+    RealIntegerType, CompatibleNumberIntegerType,
+    enable_if_t < std::is_integral<RealIntegerType>::value&&
+    std::is_integral<CompatibleNumberIntegerType>::value&&
+    !std::is_same<bool, CompatibleNumberIntegerType>::value >>
+{
+    // is there an assert somewhere on overflows?
+    using RealLimits = std::numeric_limits<RealIntegerType>;
+    using CompatibleLimits = std::numeric_limits<CompatibleNumberIntegerType>;
+
+    static constexpr auto value =
+        is_constructible<RealIntegerType,
+        CompatibleNumberIntegerType>::value &&
+        CompatibleLimits::is_integer &&
+        RealLimits::is_signed == CompatibleLimits::is_signed;
+};
+
+template<typename RealIntegerType, typename CompatibleNumberIntegerType>
+struct is_compatible_integer_type
+    : is_compatible_integer_type_impl<RealIntegerType,
+      CompatibleNumberIntegerType> {};
+
+template<typename BasicJsonType, typename CompatibleType, typename = void>
+struct is_compatible_type_impl: std::false_type {};
+
+template<typename BasicJsonType, typename CompatibleType>
+struct is_compatible_type_impl <
+    BasicJsonType, CompatibleType,
+    enable_if_t<is_complete_type<CompatibleType>::value >>
+{
+    static constexpr bool value =
+        has_to_json<BasicJsonType, CompatibleType>::value;
+};
+
+template<typename BasicJsonType, typename CompatibleType>
+struct is_compatible_type
+    : is_compatible_type_impl<BasicJsonType, CompatibleType> {};
+
+template<typename T1, typename T2>
+struct is_constructible_tuple : std::false_type {};
+
+template<typename T1, typename... Args>
+struct is_constructible_tuple<T1, std::tuple<Args...>> : conjunction<is_constructible<T1, Args>...> {};
+
+// a naive helper to check if a type is an ordered_map (exploits the fact that
+// ordered_map inherits capacity() from std::vector)
+template <typename T>
+struct is_ordered_map
+{
+    using one = char;
+
+    struct two
+    {
+        char x[2]; // NOLINT(cppcoreguidelines-avoid-c-arrays,hicpp-avoid-c-arrays,modernize-avoid-c-arrays)
+    };
+
+    template <typename C> static one test( decltype(&C::capacity) ) ;
+    template <typename C> static two test(...);
+
+    enum { value = sizeof(test<T>(nullptr)) == sizeof(char) }; // NOLINT(cppcoreguidelines-pro-type-vararg,hicpp-vararg)
+};
+
+// to avoid useless casts (see https://github.com/nlohmann/json/issues/2893#issuecomment-889152324)
+template < typename T, typename U, enable_if_t < !std::is_same<T, U>::value, int > = 0 >
+T conditional_static_cast(U value)
+{
+    return static_cast<T>(value);
+}
+
+template<typename T, typename U, enable_if_t<std::is_same<T, U>::value, int> = 0>
+T conditional_static_cast(U value)
+{
+    return value;
+}
+
+}  // namespace detail
+}  // namespace nlohmann
+
 // #include <nlohmann/detail/value_t.hpp>
 
 
+#if JSON_HAS_EXPERIMENTAL_FILESYSTEM
+#include <experimental/filesystem>
+namespace nlohmann::detail
+{
+namespace std_fs = std::experimental::filesystem;
+} // namespace nlohmann::detail
+#elif JSON_HAS_FILESYSTEM
+#include <filesystem>
+namespace nlohmann::detail
+{
+namespace std_fs = std::filesystem;
+} // namespace nlohmann::detail
+#endif
+
 namespace nlohmann
 {
 namespace detail
@@ -1054,18 +3833,18 @@ namespace detail
 template<typename BasicJsonType>
 void from_json(const BasicJsonType& j, typename std::nullptr_t& n)
 {
-    if (JSON_UNLIKELY(not j.is_null()))
+    if (JSON_HEDLEY_UNLIKELY(!j.is_null()))
     {
-        JSON_THROW(type_error::create(302, "type must be null, but is " + std::string(j.type_name())));
+        JSON_THROW(type_error::create(302, "type must be null, but is " + std::string(j.type_name()), j));
     }
     n = nullptr;
 }
 
 // overloads for basic_json template parameters
-template<typename BasicJsonType, typename ArithmeticType,
-         enable_if_t<std::is_arithmetic<ArithmeticType>::value and
-                     not std::is_same<ArithmeticType, typename BasicJsonType::boolean_t>::value,
-                     int> = 0>
+template < typename BasicJsonType, typename ArithmeticType,
+           enable_if_t < std::is_arithmetic<ArithmeticType>::value&&
+                         !std::is_same<ArithmeticType, typename BasicJsonType::boolean_t>::value,
+                         int > = 0 >
 void get_arithmetic_value(const BasicJsonType& j, ArithmeticType& val)
 {
     switch (static_cast<value_t>(j))
@@ -1086,17 +3865,24 @@ void get_arithmetic_value(const BasicJsonType& j, ArithmeticType& val)
             break;
         }
 
+        case value_t::null:
+        case value_t::object:
+        case value_t::array:
+        case value_t::string:
+        case value_t::boolean:
+        case value_t::binary:
+        case value_t::discarded:
         default:
-            JSON_THROW(type_error::create(302, "type must be number, but is " + std::string(j.type_name())));
+            JSON_THROW(type_error::create(302, "type must be number, but is " + std::string(j.type_name()), j));
     }
 }
 
 template<typename BasicJsonType>
 void from_json(const BasicJsonType& j, typename BasicJsonType::boolean_t& b)
 {
-    if (JSON_UNLIKELY(not j.is_boolean()))
+    if (JSON_HEDLEY_UNLIKELY(!j.is_boolean()))
     {
-        JSON_THROW(type_error::create(302, "type must be boolean, but is " + std::string(j.type_name())));
+        JSON_THROW(type_error::create(302, "type must be boolean, but is " + std::string(j.type_name()), j));
     }
     b = *j.template get_ptr<const typename BasicJsonType::boolean_t*>();
 }
@@ -1104,25 +3890,25 @@ void from_json(const BasicJsonType& j, typename BasicJsonType::boolean_t& b)
 template<typename BasicJsonType>
 void from_json(const BasicJsonType& j, typename BasicJsonType::string_t& s)
 {
-    if (JSON_UNLIKELY(not j.is_string()))
+    if (JSON_HEDLEY_UNLIKELY(!j.is_string()))
     {
-        JSON_THROW(type_error::create(302, "type must be string, but is " + std::string(j.type_name())));
+        JSON_THROW(type_error::create(302, "type must be string, but is " + std::string(j.type_name()), j));
     }
     s = *j.template get_ptr<const typename BasicJsonType::string_t*>();
 }
 
 template <
-    typename BasicJsonType, typename CompatibleStringType,
+    typename BasicJsonType, typename ConstructibleStringType,
     enable_if_t <
-        is_compatible_string_type<BasicJsonType, CompatibleStringType>::value and
-        not std::is_same<typename BasicJsonType::string_t,
-                         CompatibleStringType>::value,
+        is_constructible_string_type<BasicJsonType, ConstructibleStringType>::value&&
+        !std::is_same<typename BasicJsonType::string_t,
+                      ConstructibleStringType>::value,
         int > = 0 >
-void from_json(const BasicJsonType& j, CompatibleStringType& s)
+void from_json(const BasicJsonType& j, ConstructibleStringType& s)
 {
-    if (JSON_UNLIKELY(not j.is_string()))
+    if (JSON_HEDLEY_UNLIKELY(!j.is_string()))
     {
-        JSON_THROW(type_error::create(302, "type must be string, but is " + std::string(j.type_name())));
+        JSON_THROW(type_error::create(302, "type must be string, but is " + std::string(j.type_name()), j));
     }
 
     s = *j.template get_ptr<const typename BasicJsonType::string_t*>();
@@ -1157,13 +3943,14 @@ void from_json(const BasicJsonType& j, EnumType& e)
 
 // forward_list doesn't have an insert method
 template<typename BasicJsonType, typename T, typename Allocator,
-         enable_if_t<std::is_convertible<BasicJsonType, T>::value, int> = 0>
+         enable_if_t<is_getable<BasicJsonType, T>::value, int> = 0>
 void from_json(const BasicJsonType& j, std::forward_list<T, Allocator>& l)
 {
-    if (JSON_UNLIKELY(not j.is_array()))
+    if (JSON_HEDLEY_UNLIKELY(!j.is_array()))
     {
-        JSON_THROW(type_error::create(302, "type must be array, but is " + std::string(j.type_name())));
+        JSON_THROW(type_error::create(302, "type must be array, but is " + std::string(j.type_name()), j));
     }
+    l.clear();
     std::transform(j.rbegin(), j.rend(),
                    std::front_inserter(l), [](const BasicJsonType & i)
     {
@@ -1173,15 +3960,29 @@ void from_json(const BasicJsonType& j, std::forward_list<T, Allocator>& l)
 
 // valarray doesn't have an insert method
 template<typename BasicJsonType, typename T,
-         enable_if_t<std::is_convertible<BasicJsonType, T>::value, int> = 0>
+         enable_if_t<is_getable<BasicJsonType, T>::value, int> = 0>
 void from_json(const BasicJsonType& j, std::valarray<T>& l)
 {
-    if (JSON_UNLIKELY(not j.is_array()))
+    if (JSON_HEDLEY_UNLIKELY(!j.is_array()))
     {
-        JSON_THROW(type_error::create(302, "type must be array, but is " + std::string(j.type_name())));
+        JSON_THROW(type_error::create(302, "type must be array, but is " + std::string(j.type_name()), j));
     }
     l.resize(j.size());
-    std::copy(j.m_value.array->begin(), j.m_value.array->end(), std::begin(l));
+    std::transform(j.begin(), j.end(), std::begin(l),
+                   [](const BasicJsonType & elem)
+    {
+        return elem.template get<T>();
+    });
+}
+
+template<typename BasicJsonType, typename T, std::size_t N>
+auto from_json(const BasicJsonType& j, T (&arr)[N])  // NOLINT(cppcoreguidelines-avoid-c-arrays,hicpp-avoid-c-arrays,modernize-avoid-c-arrays)
+-> decltype(j.template get<T>(), void())
+{
+    for (std::size_t i = 0; i < N; ++i)
+    {
+        arr[i] = j.at(i).template get<T>();
+    }
 }
 
 template<typename BasicJsonType>
@@ -1190,7 +3991,7 @@ void from_json_array_impl(const BasicJsonType& j, typename BasicJsonType::array_
     arr = *j.template get_ptr<const typename BasicJsonType::array_t*>();
 }
 
-template <typename BasicJsonType, typename T, std::size_t N>
+template<typename BasicJsonType, typename T, std::size_t N>
 auto from_json_array_impl(const BasicJsonType& j, std::array<T, N>& arr,
                           priority_tag<2> /*unused*/)
 -> decltype(j.template get<T>(), void())
@@ -1201,95 +4002,136 @@ auto from_json_array_impl(const BasicJsonType& j, std::array<T, N>& arr,
     }
 }
 
-template<typename BasicJsonType, typename CompatibleArrayType>
-auto from_json_array_impl(const BasicJsonType& j, CompatibleArrayType& arr, priority_tag<1> /*unused*/)
+template<typename BasicJsonType, typename ConstructibleArrayType,
+         enable_if_t<
+             std::is_assignable<ConstructibleArrayType&, ConstructibleArrayType>::value,
+             int> = 0>
+auto from_json_array_impl(const BasicJsonType& j, ConstructibleArrayType& arr, priority_tag<1> /*unused*/)
 -> decltype(
-    arr.reserve(std::declval<typename CompatibleArrayType::size_type>()),
-    j.template get<typename CompatibleArrayType::value_type>(),
+    arr.reserve(std::declval<typename ConstructibleArrayType::size_type>()),
+    j.template get<typename ConstructibleArrayType::value_type>(),
     void())
 {
     using std::end;
 
-    arr.reserve(j.size());
+    ConstructibleArrayType ret;
+    ret.reserve(j.size());
     std::transform(j.begin(), j.end(),
-                   std::inserter(arr, end(arr)), [](const BasicJsonType & i)
+                   std::inserter(ret, end(ret)), [](const BasicJsonType & i)
     {
         // get<BasicJsonType>() returns *this, this won't call a from_json
         // method when value_type is BasicJsonType
-        return i.template get<typename CompatibleArrayType::value_type>();
+        return i.template get<typename ConstructibleArrayType::value_type>();
     });
+    arr = std::move(ret);
 }
 
-template <typename BasicJsonType, typename CompatibleArrayType>
-void from_json_array_impl(const BasicJsonType& j, CompatibleArrayType& arr,
+template<typename BasicJsonType, typename ConstructibleArrayType,
+         enable_if_t<
+             std::is_assignable<ConstructibleArrayType&, ConstructibleArrayType>::value,
+             int> = 0>
+void from_json_array_impl(const BasicJsonType& j, ConstructibleArrayType& arr,
                           priority_tag<0> /*unused*/)
 {
     using std::end;
 
+    ConstructibleArrayType ret;
     std::transform(
-        j.begin(), j.end(), std::inserter(arr, end(arr)),
+        j.begin(), j.end(), std::inserter(ret, end(ret)),
         [](const BasicJsonType & i)
     {
         // get<BasicJsonType>() returns *this, this won't call a from_json
         // method when value_type is BasicJsonType
-        return i.template get<typename CompatibleArrayType::value_type>();
+        return i.template get<typename ConstructibleArrayType::value_type>();
     });
+    arr = std::move(ret);
 }
 
-template <typename BasicJsonType, typename CompatibleArrayType,
-          enable_if_t <
-              is_compatible_array_type<BasicJsonType, CompatibleArrayType>::value and
-              not is_compatible_object_type<BasicJsonType, CompatibleArrayType>::value and
-              not is_compatible_string_type<BasicJsonType, CompatibleArrayType>::value and
-              not is_basic_json<CompatibleArrayType>::value,
-              int > = 0 >
-
-auto from_json(const BasicJsonType& j, CompatibleArrayType& arr)
+template < typename BasicJsonType, typename ConstructibleArrayType,
+           enable_if_t <
+               is_constructible_array_type<BasicJsonType, ConstructibleArrayType>::value&&
+               !is_constructible_object_type<BasicJsonType, ConstructibleArrayType>::value&&
+               !is_constructible_string_type<BasicJsonType, ConstructibleArrayType>::value&&
+               !std::is_same<ConstructibleArrayType, typename BasicJsonType::binary_t>::value&&
+               !is_basic_json<ConstructibleArrayType>::value,
+               int > = 0 >
+auto from_json(const BasicJsonType& j, ConstructibleArrayType& arr)
 -> decltype(from_json_array_impl(j, arr, priority_tag<3> {}),
-j.template get<typename CompatibleArrayType::value_type>(),
+j.template get<typename ConstructibleArrayType::value_type>(),
 void())
 {
-    if (JSON_UNLIKELY(not j.is_array()))
+    if (JSON_HEDLEY_UNLIKELY(!j.is_array()))
     {
-        JSON_THROW(type_error::create(302, "type must be array, but is " +
-                                      std::string(j.type_name())));
+        JSON_THROW(type_error::create(302, "type must be array, but is " + std::string(j.type_name()), j));
     }
 
     from_json_array_impl(j, arr, priority_tag<3> {});
 }
 
-template<typename BasicJsonType, typename CompatibleObjectType,
-         enable_if_t<is_compatible_object_type<BasicJsonType, CompatibleObjectType>::value, int> = 0>
-void from_json(const BasicJsonType& j, CompatibleObjectType& obj)
+template < typename BasicJsonType, typename T, std::size_t... Idx >
+std::array<T, sizeof...(Idx)> from_json_inplace_array_impl(BasicJsonType&& j,
+        identity_tag<std::array<T, sizeof...(Idx)>> /*unused*/, index_sequence<Idx...> /*unused*/)
 {
-    if (JSON_UNLIKELY(not j.is_object()))
+    return { { std::forward<BasicJsonType>(j).at(Idx).template get<T>()... } };
+}
+
+template < typename BasicJsonType, typename T, std::size_t N >
+auto from_json(BasicJsonType&& j, identity_tag<std::array<T, N>> tag)
+-> decltype(from_json_inplace_array_impl(std::forward<BasicJsonType>(j), tag, make_index_sequence<N> {}))
+{
+    if (JSON_HEDLEY_UNLIKELY(!j.is_array()))
     {
-        JSON_THROW(type_error::create(302, "type must be object, but is " + std::string(j.type_name())));
+        JSON_THROW(type_error::create(302, "type must be array, but is " + std::string(j.type_name()), j));
     }
 
-    auto inner_object = j.template get_ptr<const typename BasicJsonType::object_t*>();
-    using value_type = typename CompatibleObjectType::value_type;
+    return from_json_inplace_array_impl(std::forward<BasicJsonType>(j), tag, make_index_sequence<N> {});
+}
+
+template<typename BasicJsonType>
+void from_json(const BasicJsonType& j, typename BasicJsonType::binary_t& bin)
+{
+    if (JSON_HEDLEY_UNLIKELY(!j.is_binary()))
+    {
+        JSON_THROW(type_error::create(302, "type must be binary, but is " + std::string(j.type_name()), j));
+    }
+
+    bin = *j.template get_ptr<const typename BasicJsonType::binary_t*>();
+}
+
+template<typename BasicJsonType, typename ConstructibleObjectType,
+         enable_if_t<is_constructible_object_type<BasicJsonType, ConstructibleObjectType>::value, int> = 0>
+void from_json(const BasicJsonType& j, ConstructibleObjectType& obj)
+{
+    if (JSON_HEDLEY_UNLIKELY(!j.is_object()))
+    {
+        JSON_THROW(type_error::create(302, "type must be object, but is " + std::string(j.type_name()), j));
+    }
+
+    ConstructibleObjectType ret;
+    const auto* inner_object = j.template get_ptr<const typename BasicJsonType::object_t*>();
+    using value_type = typename ConstructibleObjectType::value_type;
     std::transform(
         inner_object->begin(), inner_object->end(),
-        std::inserter(obj, obj.begin()),
+        std::inserter(ret, ret.begin()),
         [](typename BasicJsonType::object_t::value_type const & p)
     {
-        return value_type(p.first, p.second.template get<typename CompatibleObjectType::mapped_type>());
+        return value_type(p.first, p.second.template get<typename ConstructibleObjectType::mapped_type>());
     });
+    obj = std::move(ret);
 }
 
 // overload for arithmetic types, not chosen for basic_json template arguments
 // (BooleanType, etc..); note: Is it really necessary to provide explicit
 // overloads for boolean_t etc. in case of a custom BooleanType which is not
 // an arithmetic type?
-template<typename BasicJsonType, typename ArithmeticType,
-         enable_if_t <
-             std::is_arithmetic<ArithmeticType>::value and
-             not std::is_same<ArithmeticType, typename BasicJsonType::number_unsigned_t>::value and
-             not std::is_same<ArithmeticType, typename BasicJsonType::number_integer_t>::value and
-             not std::is_same<ArithmeticType, typename BasicJsonType::number_float_t>::value and
-             not std::is_same<ArithmeticType, typename BasicJsonType::boolean_t>::value,
-             int> = 0>
+template < typename BasicJsonType, typename ArithmeticType,
+           enable_if_t <
+               std::is_arithmetic<ArithmeticType>::value&&
+               !std::is_same<ArithmeticType, typename BasicJsonType::number_unsigned_t>::value&&
+               !std::is_same<ArithmeticType, typename BasicJsonType::number_integer_t>::value&&
+               !std::is_same<ArithmeticType, typename BasicJsonType::number_float_t>::value&&
+               !std::is_same<ArithmeticType, typename BasicJsonType::boolean_t>::value,
+               int > = 0 >
 void from_json(const BasicJsonType& j, ArithmeticType& val)
 {
     switch (static_cast<value_t>(j))
@@ -1315,111 +4157,157 @@ void from_json(const BasicJsonType& j, ArithmeticType& val)
             break;
         }
 
+        case value_t::null:
+        case value_t::object:
+        case value_t::array:
+        case value_t::string:
+        case value_t::binary:
+        case value_t::discarded:
         default:
-            JSON_THROW(type_error::create(302, "type must be number, but is " + std::string(j.type_name())));
+            JSON_THROW(type_error::create(302, "type must be number, but is " + std::string(j.type_name()), j));
     }
 }
 
+template<typename BasicJsonType, typename... Args, std::size_t... Idx>
+std::tuple<Args...> from_json_tuple_impl_base(BasicJsonType&& j, index_sequence<Idx...> /*unused*/)
+{
+    return std::make_tuple(std::forward<BasicJsonType>(j).at(Idx).template get<Args>()...);
+}
+
+template < typename BasicJsonType, class A1, class A2 >
+std::pair<A1, A2> from_json_tuple_impl(BasicJsonType&& j, identity_tag<std::pair<A1, A2>> /*unused*/, priority_tag<0> /*unused*/)
+{
+    return {std::forward<BasicJsonType>(j).at(0).template get<A1>(),
+            std::forward<BasicJsonType>(j).at(1).template get<A2>()};
+}
+
 template<typename BasicJsonType, typename A1, typename A2>
-void from_json(const BasicJsonType& j, std::pair<A1, A2>& p)
+void from_json_tuple_impl(BasicJsonType&& j, std::pair<A1, A2>& p, priority_tag<1> /*unused*/)
 {
-    p = {j.at(0).template get<A1>(), j.at(1).template get<A2>()};
-}
-
-template<typename BasicJsonType, typename Tuple, std::size_t... Idx>
-void from_json_tuple_impl(const BasicJsonType& j, Tuple& t, index_sequence<Idx...>)
-{
-    t = std::make_tuple(j.at(Idx).template get<typename std::tuple_element<Idx, Tuple>::type>()...);
+    p = from_json_tuple_impl(std::forward<BasicJsonType>(j), identity_tag<std::pair<A1, A2>> {}, priority_tag<0> {});
 }
 
 template<typename BasicJsonType, typename... Args>
-void from_json(const BasicJsonType& j, std::tuple<Args...>& t)
+std::tuple<Args...> from_json_tuple_impl(BasicJsonType&& j, identity_tag<std::tuple<Args...>> /*unused*/, priority_tag<2> /*unused*/)
 {
-    from_json_tuple_impl(j, t, index_sequence_for<Args...> {});
+    return from_json_tuple_impl_base<BasicJsonType, Args...>(std::forward<BasicJsonType>(j), index_sequence_for<Args...> {});
 }
 
-template <typename BasicJsonType, typename Key, typename Value, typename Compare, typename Allocator,
-          typename = enable_if_t<not std::is_constructible<
-                                     typename BasicJsonType::string_t, Key>::value>>
+template<typename BasicJsonType, typename... Args>
+void from_json_tuple_impl(BasicJsonType&& j, std::tuple<Args...>& t, priority_tag<3> /*unused*/)
+{
+    t = from_json_tuple_impl_base<BasicJsonType, Args...>(std::forward<BasicJsonType>(j), index_sequence_for<Args...> {});
+}
+
+template<typename BasicJsonType, typename TupleRelated>
+auto from_json(BasicJsonType&& j, TupleRelated&& t)
+-> decltype(from_json_tuple_impl(std::forward<BasicJsonType>(j), std::forward<TupleRelated>(t), priority_tag<3> {}))
+{
+    if (JSON_HEDLEY_UNLIKELY(!j.is_array()))
+    {
+        JSON_THROW(type_error::create(302, "type must be array, but is " + std::string(j.type_name()), j));
+    }
+
+    return from_json_tuple_impl(std::forward<BasicJsonType>(j), std::forward<TupleRelated>(t), priority_tag<3> {});
+}
+
+template < typename BasicJsonType, typename Key, typename Value, typename Compare, typename Allocator,
+           typename = enable_if_t < !std::is_constructible <
+                                        typename BasicJsonType::string_t, Key >::value >>
 void from_json(const BasicJsonType& j, std::map<Key, Value, Compare, Allocator>& m)
 {
-    if (JSON_UNLIKELY(not j.is_array()))
+    if (JSON_HEDLEY_UNLIKELY(!j.is_array()))
     {
-        JSON_THROW(type_error::create(302, "type must be array, but is " + std::string(j.type_name())));
+        JSON_THROW(type_error::create(302, "type must be array, but is " + std::string(j.type_name()), j));
     }
+    m.clear();
     for (const auto& p : j)
     {
-        if (JSON_UNLIKELY(not p.is_array()))
+        if (JSON_HEDLEY_UNLIKELY(!p.is_array()))
         {
-            JSON_THROW(type_error::create(302, "type must be array, but is " + std::string(p.type_name())));
+            JSON_THROW(type_error::create(302, "type must be array, but is " + std::string(p.type_name()), j));
         }
         m.emplace(p.at(0).template get<Key>(), p.at(1).template get<Value>());
     }
 }
 
-template <typename BasicJsonType, typename Key, typename Value, typename Hash, typename KeyEqual, typename Allocator,
-          typename = enable_if_t<not std::is_constructible<
-                                     typename BasicJsonType::string_t, Key>::value>>
+template < typename BasicJsonType, typename Key, typename Value, typename Hash, typename KeyEqual, typename Allocator,
+           typename = enable_if_t < !std::is_constructible <
+                                        typename BasicJsonType::string_t, Key >::value >>
 void from_json(const BasicJsonType& j, std::unordered_map<Key, Value, Hash, KeyEqual, Allocator>& m)
 {
-    if (JSON_UNLIKELY(not j.is_array()))
+    if (JSON_HEDLEY_UNLIKELY(!j.is_array()))
     {
-        JSON_THROW(type_error::create(302, "type must be array, but is " + std::string(j.type_name())));
+        JSON_THROW(type_error::create(302, "type must be array, but is " + std::string(j.type_name()), j));
     }
+    m.clear();
     for (const auto& p : j)
     {
-        if (JSON_UNLIKELY(not p.is_array()))
+        if (JSON_HEDLEY_UNLIKELY(!p.is_array()))
         {
-            JSON_THROW(type_error::create(302, "type must be array, but is " + std::string(p.type_name())));
+            JSON_THROW(type_error::create(302, "type must be array, but is " + std::string(p.type_name()), j));
         }
         m.emplace(p.at(0).template get<Key>(), p.at(1).template get<Value>());
     }
 }
 
+#if JSON_HAS_FILESYSTEM || JSON_HAS_EXPERIMENTAL_FILESYSTEM
+template<typename BasicJsonType>
+void from_json(const BasicJsonType& j, std_fs::path& p)
+{
+    if (JSON_HEDLEY_UNLIKELY(!j.is_string()))
+    {
+        JSON_THROW(type_error::create(302, "type must be string, but is " + std::string(j.type_name()), j));
+    }
+    p = *j.template get_ptr<const typename BasicJsonType::string_t*>();
+}
+#endif
+
 struct from_json_fn
 {
     template<typename BasicJsonType, typename T>
-    auto operator()(const BasicJsonType& j, T& val) const
-    noexcept(noexcept(from_json(j, val)))
-    -> decltype(from_json(j, val), void())
+    auto operator()(const BasicJsonType& j, T&& val) const
+    noexcept(noexcept(from_json(j, std::forward<T>(val))))
+    -> decltype(from_json(j, std::forward<T>(val)))
     {
-        return from_json(j, val);
+        return from_json(j, std::forward<T>(val));
     }
 };
-}
+}  // namespace detail
 
 /// namespace to hold default `from_json` function
 /// to see why this is required:
 /// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2015/n4381.html
-namespace
+namespace // NOLINT(cert-dcl59-cpp,fuchsia-header-anon-namespaces,google-build-namespaces)
 {
-constexpr const auto& from_json = detail::static_const<detail::from_json_fn>::value;
-}
-}
+constexpr const auto& from_json = detail::static_const<detail::from_json_fn>::value; // NOLINT(misc-definitions-in-headers)
+} // namespace
+} // namespace nlohmann
 
 // #include <nlohmann/detail/conversions/to_json.hpp>
 
 
-#include <ciso646> // or, and, not
+#include <algorithm> // copy
 #include <iterator> // begin, end
+#include <string> // string
 #include <tuple> // tuple, get
 #include <type_traits> // is_same, is_constructible, is_floating_point, is_enum, underlying_type
 #include <utility> // move, forward, declval, pair
 #include <valarray> // valarray
 #include <vector> // vector
 
-// #include <nlohmann/detail/meta/cpp_future.hpp>
-
-// #include <nlohmann/detail/meta/type_traits.hpp>
-
-// #include <nlohmann/detail/value_t.hpp>
+// #include <nlohmann/detail/macro_scope.hpp>
 
 // #include <nlohmann/detail/iterators/iteration_proxy.hpp>
 
 
 #include <cstddef> // size_t
-#include <string> // string, to_string
 #include <iterator> // input_iterator_tag
+#include <string> // string, to_string
+#include <tuple> // tuple_size, get, tuple_element
+#include <utility> // move
+
+// #include <nlohmann/detail/meta/type_traits.hpp>
 
 // #include <nlohmann/detail/value_t.hpp>
 
@@ -1428,100 +4316,114 @@ namespace nlohmann
 {
 namespace detail
 {
+template<typename string_type>
+void int_to_string( string_type& target, std::size_t value )
+{
+    // For ADL
+    using std::to_string;
+    target = to_string(value);
+}
+template<typename IteratorType> class iteration_proxy_value
+{
+  public:
+    using difference_type = std::ptrdiff_t;
+    using value_type = iteration_proxy_value;
+    using pointer = value_type * ;
+    using reference = value_type & ;
+    using iterator_category = std::input_iterator_tag;
+    using string_type = typename std::remove_cv< typename std::remove_reference<decltype( std::declval<IteratorType>().key() ) >::type >::type;
+
+  private:
+    /// the iterator
+    IteratorType anchor;
+    /// an index for arrays (used to create key names)
+    std::size_t array_index = 0;
+    /// last stringified array index
+    mutable std::size_t array_index_last = 0;
+    /// a string representation of the array index
+    mutable string_type array_index_str = "0";
+    /// an empty string (to return a reference for primitive values)
+    const string_type empty_str{};
+
+  public:
+    explicit iteration_proxy_value(IteratorType it) noexcept
+        : anchor(std::move(it))
+    {}
+
+    /// dereference operator (needed for range-based for)
+    iteration_proxy_value& operator*()
+    {
+        return *this;
+    }
+
+    /// increment operator (needed for range-based for)
+    iteration_proxy_value& operator++()
+    {
+        ++anchor;
+        ++array_index;
+
+        return *this;
+    }
+
+    /// equality operator (needed for InputIterator)
+    bool operator==(const iteration_proxy_value& o) const
+    {
+        return anchor == o.anchor;
+    }
+
+    /// inequality operator (needed for range-based for)
+    bool operator!=(const iteration_proxy_value& o) const
+    {
+        return anchor != o.anchor;
+    }
+
+    /// return key of the iterator
+    const string_type& key() const
+    {
+        JSON_ASSERT(anchor.m_object != nullptr);
+
+        switch (anchor.m_object->type())
+        {
+            // use integer array index as key
+            case value_t::array:
+            {
+                if (array_index != array_index_last)
+                {
+                    int_to_string( array_index_str, array_index );
+                    array_index_last = array_index;
+                }
+                return array_index_str;
+            }
+
+            // use key from the object
+            case value_t::object:
+                return anchor.key();
+
+            // use an empty key for all primitive types
+            case value_t::null:
+            case value_t::string:
+            case value_t::boolean:
+            case value_t::number_integer:
+            case value_t::number_unsigned:
+            case value_t::number_float:
+            case value_t::binary:
+            case value_t::discarded:
+            default:
+                return empty_str;
+        }
+    }
+
+    /// return value of the iterator
+    typename IteratorType::reference value() const
+    {
+        return anchor.value();
+    }
+};
+
 /// proxy class for the items() function
 template<typename IteratorType> class iteration_proxy
 {
   private:
-    /// helper class for iteration
-    class iteration_proxy_internal
-    {
-      public:
-        using difference_type = std::ptrdiff_t;
-        using value_type = iteration_proxy_internal;
-        using pointer = iteration_proxy_internal*;
-        using reference = iteration_proxy_internal&;
-        using iterator_category = std::input_iterator_tag;
-
-      private:
-        /// the iterator
-        IteratorType anchor;
-        /// an index for arrays (used to create key names)
-        std::size_t array_index = 0;
-        /// last stringified array index
-        mutable std::size_t array_index_last = 0;
-        /// a string representation of the array index
-        mutable std::string array_index_str = "0";
-        /// an empty string (to return a reference for primitive values)
-        const std::string empty_str = "";
-
-      public:
-        explicit iteration_proxy_internal(IteratorType it) noexcept : anchor(it) {}
-
-        iteration_proxy_internal(const iteration_proxy_internal&) = default;
-        iteration_proxy_internal& operator=(const iteration_proxy_internal&) = default;
-
-        /// dereference operator (needed for range-based for)
-        iteration_proxy_internal& operator*()
-        {
-            return *this;
-        }
-
-        /// increment operator (needed for range-based for)
-        iteration_proxy_internal& operator++()
-        {
-            ++anchor;
-            ++array_index;
-
-            return *this;
-        }
-
-        /// equality operator (needed for InputIterator)
-        bool operator==(const iteration_proxy_internal& o) const noexcept
-        {
-            return anchor == o.anchor;
-        }
-
-        /// inequality operator (needed for range-based for)
-        bool operator!=(const iteration_proxy_internal& o) const noexcept
-        {
-            return anchor != o.anchor;
-        }
-
-        /// return key of the iterator
-        const std::string& key() const
-        {
-            assert(anchor.m_object != nullptr);
-
-            switch (anchor.m_object->type())
-            {
-                // use integer array index as key
-                case value_t::array:
-                {
-                    if (array_index != array_index_last)
-                    {
-                        array_index_str = std::to_string(array_index);
-                        array_index_last = array_index;
-                    }
-                    return array_index_str;
-                }
-
-                // use key from the object
-                case value_t::object:
-                    return anchor.key();
-
-                // use an empty key for all primitive types
-                default:
-                    return empty_str;
-            }
-        }
-
-        /// return value of the iterator
-        typename IteratorType::reference value() const
-        {
-            return anchor.value();
-        }
-    };
-
     /// the container to iterate
     typename IteratorType::reference container;
 
@@ -1531,20 +4433,84 @@ template<typename IteratorType> class iteration_proxy
         : container(cont) {}
 
     /// return iterator begin (needed for range-based for)
-    iteration_proxy_internal begin() noexcept
+    iteration_proxy_value<IteratorType> begin() noexcept
     {
-        return iteration_proxy_internal(container.begin());
+        return iteration_proxy_value<IteratorType>(container.begin());
     }
 
     /// return iterator end (needed for range-based for)
-    iteration_proxy_internal end() noexcept
+    iteration_proxy_value<IteratorType> end() noexcept
     {
-        return iteration_proxy_internal(container.end());
+        return iteration_proxy_value<IteratorType>(container.end());
     }
 };
+// Structured Bindings Support
+// For further reference see https://blog.tartanllama.xyz/structured-bindings/
+// And see https://github.com/nlohmann/json/pull/1391
+template<std::size_t N, typename IteratorType, enable_if_t<N == 0, int> = 0>
+auto get(const nlohmann::detail::iteration_proxy_value<IteratorType>& i) -> decltype(i.key())
+{
+    return i.key();
 }
+// Structured Bindings Support
+// For further reference see https://blog.tartanllama.xyz/structured-bindings/
+// And see https://github.com/nlohmann/json/pull/1391
+template<std::size_t N, typename IteratorType, enable_if_t<N == 1, int> = 0>
+auto get(const nlohmann::detail::iteration_proxy_value<IteratorType>& i) -> decltype(i.value())
+{
+    return i.value();
 }
+}  // namespace detail
+}  // namespace nlohmann
 
+// The Addition to the STD Namespace is required to add
+// Structured Bindings Support to the iteration_proxy_value class
+// For further reference see https://blog.tartanllama.xyz/structured-bindings/
+// And see https://github.com/nlohmann/json/pull/1391
+namespace std
+{
+#if defined(__clang__)
+    // Fix: https://github.com/nlohmann/json/issues/1401
+    #pragma clang diagnostic push
+    #pragma clang diagnostic ignored "-Wmismatched-tags"
+#endif
+template<typename IteratorType>
+class tuple_size<::nlohmann::detail::iteration_proxy_value<IteratorType>>
+            : public std::integral_constant<std::size_t, 2> {};
+
+template<std::size_t N, typename IteratorType>
+class tuple_element<N, ::nlohmann::detail::iteration_proxy_value<IteratorType >>
+{
+  public:
+    using type = decltype(
+                     get<N>(std::declval <
+                            ::nlohmann::detail::iteration_proxy_value<IteratorType >> ()));
+};
+#if defined(__clang__)
+    #pragma clang diagnostic pop
+#endif
+} // namespace std
+
+// #include <nlohmann/detail/meta/cpp_future.hpp>
+
+// #include <nlohmann/detail/meta/type_traits.hpp>
+
+// #include <nlohmann/detail/value_t.hpp>
+
+
+#if JSON_HAS_EXPERIMENTAL_FILESYSTEM
+#include <experimental/filesystem>
+namespace nlohmann::detail
+{
+namespace std_fs = std::experimental::filesystem;
+} // namespace nlohmann::detail
+#elif JSON_HAS_FILESYSTEM
+#include <filesystem>
+namespace nlohmann::detail
+{
+namespace std_fs = std::filesystem;
+} // namespace nlohmann::detail
+#endif
 
 namespace nlohmann
 {
@@ -1554,6 +4520,13 @@ namespace detail
 // constructors //
 //////////////////
 
+/*
+ * Note all external_constructor<>::construct functions need to call
+ * j.m_value.destroy(j.m_type) to avoid a memory leak in case j contains an
+ * allocated value (e.g., a string). See bug issue
+ * https://github.com/nlohmann/json/issues/2865 for more information.
+ */
+
 template<value_t> struct external_constructor;
 
 template<>
@@ -1562,6 +4535,7 @@ struct external_constructor<value_t::boolean>
     template<typename BasicJsonType>
     static void construct(BasicJsonType& j, typename BasicJsonType::boolean_t b) noexcept
     {
+        j.m_value.destroy(j.m_type);
         j.m_type = value_t::boolean;
         j.m_value = b;
         j.assert_invariant();
@@ -1574,6 +4548,7 @@ struct external_constructor<value_t::string>
     template<typename BasicJsonType>
     static void construct(BasicJsonType& j, const typename BasicJsonType::string_t& s)
     {
+        j.m_value.destroy(j.m_type);
         j.m_type = value_t::string;
         j.m_value = s;
         j.assert_invariant();
@@ -1582,28 +4557,53 @@ struct external_constructor<value_t::string>
     template<typename BasicJsonType>
     static void construct(BasicJsonType& j, typename BasicJsonType::string_t&& s)
     {
+        j.m_value.destroy(j.m_type);
         j.m_type = value_t::string;
         j.m_value = std::move(s);
         j.assert_invariant();
     }
 
-    template<typename BasicJsonType, typename CompatibleStringType,
-             enable_if_t<not std::is_same<CompatibleStringType, typename BasicJsonType::string_t>::value,
-                         int> = 0>
+    template < typename BasicJsonType, typename CompatibleStringType,
+               enable_if_t < !std::is_same<CompatibleStringType, typename BasicJsonType::string_t>::value,
+                             int > = 0 >
     static void construct(BasicJsonType& j, const CompatibleStringType& str)
     {
+        j.m_value.destroy(j.m_type);
         j.m_type = value_t::string;
         j.m_value.string = j.template create<typename BasicJsonType::string_t>(str);
         j.assert_invariant();
     }
 };
 
+template<>
+struct external_constructor<value_t::binary>
+{
+    template<typename BasicJsonType>
+    static void construct(BasicJsonType& j, const typename BasicJsonType::binary_t& b)
+    {
+        j.m_value.destroy(j.m_type);
+        j.m_type = value_t::binary;
+        j.m_value = typename BasicJsonType::binary_t(b);
+        j.assert_invariant();
+    }
+
+    template<typename BasicJsonType>
+    static void construct(BasicJsonType& j, typename BasicJsonType::binary_t&& b)
+    {
+        j.m_value.destroy(j.m_type);
+        j.m_type = value_t::binary;
+        j.m_value = typename BasicJsonType::binary_t(std::move(b));
+        j.assert_invariant();
+    }
+};
+
 template<>
 struct external_constructor<value_t::number_float>
 {
     template<typename BasicJsonType>
     static void construct(BasicJsonType& j, typename BasicJsonType::number_float_t val) noexcept
     {
+        j.m_value.destroy(j.m_type);
         j.m_type = value_t::number_float;
         j.m_value = val;
         j.assert_invariant();
@@ -1616,6 +4616,7 @@ struct external_constructor<value_t::number_unsigned>
     template<typename BasicJsonType>
     static void construct(BasicJsonType& j, typename BasicJsonType::number_unsigned_t val) noexcept
     {
+        j.m_value.destroy(j.m_type);
         j.m_type = value_t::number_unsigned;
         j.m_value = val;
         j.assert_invariant();
@@ -1628,6 +4629,7 @@ struct external_constructor<value_t::number_integer>
     template<typename BasicJsonType>
     static void construct(BasicJsonType& j, typename BasicJsonType::number_integer_t val) noexcept
     {
+        j.m_value.destroy(j.m_type);
         j.m_type = value_t::number_integer;
         j.m_value = val;
         j.assert_invariant();
@@ -1640,40 +4642,49 @@ struct external_constructor<value_t::array>
     template<typename BasicJsonType>
     static void construct(BasicJsonType& j, const typename BasicJsonType::array_t& arr)
     {
+        j.m_value.destroy(j.m_type);
         j.m_type = value_t::array;
         j.m_value = arr;
+        j.set_parents();
         j.assert_invariant();
     }
 
     template<typename BasicJsonType>
     static void construct(BasicJsonType& j, typename BasicJsonType::array_t&& arr)
     {
+        j.m_value.destroy(j.m_type);
         j.m_type = value_t::array;
         j.m_value = std::move(arr);
+        j.set_parents();
         j.assert_invariant();
     }
 
-    template<typename BasicJsonType, typename CompatibleArrayType,
-             enable_if_t<not std::is_same<CompatibleArrayType, typename BasicJsonType::array_t>::value,
-                         int> = 0>
+    template < typename BasicJsonType, typename CompatibleArrayType,
+               enable_if_t < !std::is_same<CompatibleArrayType, typename BasicJsonType::array_t>::value,
+                             int > = 0 >
     static void construct(BasicJsonType& j, const CompatibleArrayType& arr)
     {
         using std::begin;
         using std::end;
+
+        j.m_value.destroy(j.m_type);
         j.m_type = value_t::array;
         j.m_value.array = j.template create<typename BasicJsonType::array_t>(begin(arr), end(arr));
+        j.set_parents();
         j.assert_invariant();
     }
 
     template<typename BasicJsonType>
     static void construct(BasicJsonType& j, const std::vector<bool>& arr)
     {
+        j.m_value.destroy(j.m_type);
         j.m_type = value_t::array;
         j.m_value = value_t::array;
         j.m_value.array->reserve(arr.size());
         for (const bool x : arr)
         {
             j.m_value.array->push_back(x);
+            j.set_parent(j.m_value.array->back());
         }
         j.assert_invariant();
     }
@@ -1682,10 +4693,15 @@ struct external_constructor<value_t::array>
              enable_if_t<std::is_convertible<T, BasicJsonType>::value, int> = 0>
     static void construct(BasicJsonType& j, const std::valarray<T>& arr)
     {
+        j.m_value.destroy(j.m_type);
         j.m_type = value_t::array;
         j.m_value = value_t::array;
         j.m_value.array->resize(arr.size());
-        std::copy(std::begin(arr), std::end(arr), j.m_value.array->begin());
+        if (arr.size() > 0)
+        {
+            std::copy(std::begin(arr), std::end(arr), j.m_value.array->begin());
+        }
+        j.set_parents();
         j.assert_invariant();
     }
 };
@@ -1696,28 +4712,34 @@ struct external_constructor<value_t::object>
     template<typename BasicJsonType>
     static void construct(BasicJsonType& j, const typename BasicJsonType::object_t& obj)
     {
+        j.m_value.destroy(j.m_type);
         j.m_type = value_t::object;
         j.m_value = obj;
+        j.set_parents();
         j.assert_invariant();
     }
 
     template<typename BasicJsonType>
     static void construct(BasicJsonType& j, typename BasicJsonType::object_t&& obj)
     {
+        j.m_value.destroy(j.m_type);
         j.m_type = value_t::object;
         j.m_value = std::move(obj);
+        j.set_parents();
         j.assert_invariant();
     }
 
-    template<typename BasicJsonType, typename CompatibleObjectType,
-             enable_if_t<not std::is_same<CompatibleObjectType, typename BasicJsonType::object_t>::value, int> = 0>
+    template < typename BasicJsonType, typename CompatibleObjectType,
+               enable_if_t < !std::is_same<CompatibleObjectType, typename BasicJsonType::object_t>::value, int > = 0 >
     static void construct(BasicJsonType& j, const CompatibleObjectType& obj)
     {
         using std::begin;
         using std::end;
 
+        j.m_value.destroy(j.m_type);
         j.m_type = value_t::object;
         j.m_value.object = j.template create<typename BasicJsonType::object_t>(begin(obj), end(obj));
+        j.set_parents();
         j.assert_invariant();
     }
 };
@@ -1781,19 +4803,25 @@ void to_json(BasicJsonType& j, const std::vector<bool>& e)
     external_constructor<value_t::array>::construct(j, e);
 }
 
-template <typename BasicJsonType, typename CompatibleArrayType,
-          enable_if_t<is_compatible_array_type<BasicJsonType,
-                      CompatibleArrayType>::value and
-                      not is_compatible_object_type<
-                          BasicJsonType, CompatibleArrayType>::value and
-                      not is_compatible_string_type<BasicJsonType, CompatibleArrayType>::value and
-                      not is_basic_json<CompatibleArrayType>::value,
-                      int> = 0>
+template < typename BasicJsonType, typename CompatibleArrayType,
+           enable_if_t < is_compatible_array_type<BasicJsonType,
+                         CompatibleArrayType>::value&&
+                         !is_compatible_object_type<BasicJsonType, CompatibleArrayType>::value&&
+                         !is_compatible_string_type<BasicJsonType, CompatibleArrayType>::value&&
+                         !std::is_same<typename BasicJsonType::binary_t, CompatibleArrayType>::value&&
+                         !is_basic_json<CompatibleArrayType>::value,
+                         int > = 0 >
 void to_json(BasicJsonType& j, const CompatibleArrayType& arr)
 {
     external_constructor<value_t::array>::construct(j, arr);
 }
 
+template<typename BasicJsonType>
+void to_json(BasicJsonType& j, const typename BasicJsonType::binary_t& bin)
+{
+    external_constructor<value_t::binary>::construct(j, bin);
+}
+
 template<typename BasicJsonType, typename T,
          enable_if_t<std::is_convertible<T, BasicJsonType>::value, int> = 0>
 void to_json(BasicJsonType& j, const std::valarray<T>& arr)
@@ -1807,8 +4835,8 @@ void to_json(BasicJsonType& j, typename BasicJsonType::array_t&& arr)
     external_constructor<value_t::array>::construct(j, std::move(arr));
 }
 
-template<typename BasicJsonType, typename CompatibleObjectType,
-         enable_if_t<is_compatible_object_type<BasicJsonType, CompatibleObjectType>::value and not is_basic_json<CompatibleObjectType>::value, int> = 0>
+template < typename BasicJsonType, typename CompatibleObjectType,
+           enable_if_t < is_compatible_object_type<BasicJsonType, CompatibleObjectType>::value&& !is_basic_json<CompatibleObjectType>::value, int > = 0 >
 void to_json(BasicJsonType& j, const CompatibleObjectType& obj)
 {
     external_constructor<value_t::object>::construct(j, obj);
@@ -1822,40 +4850,48 @@ void to_json(BasicJsonType& j, typename BasicJsonType::object_t&& obj)
 
 template <
     typename BasicJsonType, typename T, std::size_t N,
-    enable_if_t<not std::is_constructible<typename BasicJsonType::string_t,
-                const T (&)[N]>::value,
-                int> = 0 >
-void to_json(BasicJsonType& j, const T (&arr)[N])
+    enable_if_t < !std::is_constructible<typename BasicJsonType::string_t,
+                  const T(&)[N]>::value, // NOLINT(cppcoreguidelines-avoid-c-arrays,hicpp-avoid-c-arrays,modernize-avoid-c-arrays)
+                  int > = 0 >
+void to_json(BasicJsonType& j, const T(&arr)[N]) // NOLINT(cppcoreguidelines-avoid-c-arrays,hicpp-avoid-c-arrays,modernize-avoid-c-arrays)
 {
     external_constructor<value_t::array>::construct(j, arr);
 }
 
-template<typename BasicJsonType, typename... Args>
-void to_json(BasicJsonType& j, const std::pair<Args...>& p)
+template < typename BasicJsonType, typename T1, typename T2, enable_if_t < std::is_constructible<BasicJsonType, T1>::value&& std::is_constructible<BasicJsonType, T2>::value, int > = 0 >
+void to_json(BasicJsonType& j, const std::pair<T1, T2>& p)
 {
-    j = {p.first, p.second};
+    j = { p.first, p.second };
 }
 
 // for https://github.com/nlohmann/json/pull/1134
 template<typename BasicJsonType, typename T,
-         enable_if_t<std::is_same<T, typename iteration_proxy<typename BasicJsonType::iterator>::iteration_proxy_internal>::value, int> = 0>
-void to_json(BasicJsonType& j, T b) noexcept
+         enable_if_t<std::is_same<T, iteration_proxy_value<typename BasicJsonType::iterator>>::value, int> = 0>
+void to_json(BasicJsonType& j, const T& b)
 {
-    j = {{b.key(), b.value()}};
+    j = { {b.key(), b.value()} };
 }
 
 template<typename BasicJsonType, typename Tuple, std::size_t... Idx>
-void to_json_tuple_impl(BasicJsonType& j, const Tuple& t, index_sequence<Idx...>)
+void to_json_tuple_impl(BasicJsonType& j, const Tuple& t, index_sequence<Idx...> /*unused*/)
 {
-    j = {std::get<Idx>(t)...};
+    j = { std::get<Idx>(t)... };
 }
 
-template<typename BasicJsonType, typename... Args>
-void to_json(BasicJsonType& j, const std::tuple<Args...>& t)
+template<typename BasicJsonType, typename T, enable_if_t<is_constructible_tuple<BasicJsonType, T>::value, int > = 0>
+void to_json(BasicJsonType& j, const T& t)
 {
-    to_json_tuple_impl(j, t, index_sequence_for<Args...> {});
+    to_json_tuple_impl(j, t, make_index_sequence<std::tuple_size<T>::value> {});
 }
 
+#if JSON_HAS_FILESYSTEM || JSON_HAS_EXPERIMENTAL_FILESYSTEM
+template<typename BasicJsonType>
+void to_json(BasicJsonType& j, const std_fs::path& p)
+{
+    j = p.string();
+}
+#endif
+
 struct to_json_fn
 {
     template<typename BasicJsonType, typename T>
@@ -1865,22 +4901,313 @@ struct to_json_fn
         return to_json(j, std::forward<T>(val));
     }
 };
-}
+}  // namespace detail
 
 /// namespace to hold default `to_json` function
-namespace
+/// to see why this is required:
+/// http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2015/n4381.html
+namespace // NOLINT(cert-dcl59-cpp,fuchsia-header-anon-namespaces,google-build-namespaces)
 {
-constexpr const auto& to_json = detail::static_const<detail::to_json_fn>::value;
+constexpr const auto& to_json = detail::static_const<detail::to_json_fn>::value; // NOLINT(misc-definitions-in-headers)
+} // namespace
+} // namespace nlohmann
+
+// #include <nlohmann/detail/meta/identity_tag.hpp>
+
+// #include <nlohmann/detail/meta/type_traits.hpp>
+
+
+namespace nlohmann
+{
+
+/// @sa https://json.nlohmann.me/api/adl_serializer/
+template<typename ValueType, typename>
+struct adl_serializer
+{
+    /// @brief convert a JSON value to any value type
+    /// @sa https://json.nlohmann.me/api/adl_serializer/from_json/
+    template<typename BasicJsonType, typename TargetType = ValueType>
+    static auto from_json(BasicJsonType && j, TargetType& val) noexcept(
+        noexcept(::nlohmann::from_json(std::forward<BasicJsonType>(j), val)))
+    -> decltype(::nlohmann::from_json(std::forward<BasicJsonType>(j), val), void())
+    {
+        ::nlohmann::from_json(std::forward<BasicJsonType>(j), val);
+    }
+
+    /// @brief convert a JSON value to any value type
+    /// @sa https://json.nlohmann.me/api/adl_serializer/from_json/
+    template<typename BasicJsonType, typename TargetType = ValueType>
+    static auto from_json(BasicJsonType && j) noexcept(
+    noexcept(::nlohmann::from_json(std::forward<BasicJsonType>(j), detail::identity_tag<TargetType> {})))
+    -> decltype(::nlohmann::from_json(std::forward<BasicJsonType>(j), detail::identity_tag<TargetType> {}))
+    {
+        return ::nlohmann::from_json(std::forward<BasicJsonType>(j), detail::identity_tag<TargetType> {});
+    }
+
+    /// @brief convert any value type to a JSON value
+    /// @sa https://json.nlohmann.me/api/adl_serializer/to_json/
+    template<typename BasicJsonType, typename TargetType = ValueType>
+    static auto to_json(BasicJsonType& j, TargetType && val) noexcept(
+        noexcept(::nlohmann::to_json(j, std::forward<TargetType>(val))))
+    -> decltype(::nlohmann::to_json(j, std::forward<TargetType>(val)), void())
+    {
+        ::nlohmann::to_json(j, std::forward<TargetType>(val));
+    }
+};
+}  // namespace nlohmann
+
+// #include <nlohmann/byte_container_with_subtype.hpp>
+
+
+#include <cstdint> // uint8_t, uint64_t
+#include <tuple> // tie
+#include <utility> // move
+
+namespace nlohmann
+{
+
+/// @brief an internal type for a backed binary type
+/// @sa https://json.nlohmann.me/api/byte_container_with_subtype/
+template<typename BinaryType>
+class byte_container_with_subtype : public BinaryType
+{
+  public:
+    using container_type = BinaryType;
+    using subtype_type = std::uint64_t;
+
+    /// @sa https://json.nlohmann.me/api/byte_container_with_subtype/byte_container_with_subtype/
+    byte_container_with_subtype() noexcept(noexcept(container_type()))
+        : container_type()
+    {}
+
+    /// @sa https://json.nlohmann.me/api/byte_container_with_subtype/byte_container_with_subtype/
+    byte_container_with_subtype(const container_type& b) noexcept(noexcept(container_type(b)))
+        : container_type(b)
+    {}
+
+    /// @sa https://json.nlohmann.me/api/byte_container_with_subtype/byte_container_with_subtype/
+    byte_container_with_subtype(container_type&& b) noexcept(noexcept(container_type(std::move(b))))
+        : container_type(std::move(b))
+    {}
+
+    /// @sa https://json.nlohmann.me/api/byte_container_with_subtype/byte_container_with_subtype/
+    byte_container_with_subtype(const container_type& b, subtype_type subtype_) noexcept(noexcept(container_type(b)))
+        : container_type(b)
+        , m_subtype(subtype_)
+        , m_has_subtype(true)
+    {}
+
+    /// @sa https://json.nlohmann.me/api/byte_container_with_subtype/byte_container_with_subtype/
+    byte_container_with_subtype(container_type&& b, subtype_type subtype_) noexcept(noexcept(container_type(std::move(b))))
+        : container_type(std::move(b))
+        , m_subtype(subtype_)
+        , m_has_subtype(true)
+    {}
+
+    bool operator==(const byte_container_with_subtype& rhs) const
+    {
+        return std::tie(static_cast<const BinaryType&>(*this), m_subtype, m_has_subtype) ==
+               std::tie(static_cast<const BinaryType&>(rhs), rhs.m_subtype, rhs.m_has_subtype);
+    }
+
+    bool operator!=(const byte_container_with_subtype& rhs) const
+    {
+        return !(rhs == *this);
+    }
+
+    /// @brief sets the binary subtype
+    /// @sa https://json.nlohmann.me/api/byte_container_with_subtype/set_subtype/
+    void set_subtype(subtype_type subtype_) noexcept
+    {
+        m_subtype = subtype_;
+        m_has_subtype = true;
+    }
+
+    /// @brief return the binary subtype
+    /// @sa https://json.nlohmann.me/api/byte_container_with_subtype/subtype/
+    constexpr subtype_type subtype() const noexcept
+    {
+        return m_has_subtype ? m_subtype : static_cast<subtype_type>(-1);
+    }
+
+    /// @brief return whether the value has a subtype
+    /// @sa https://json.nlohmann.me/api/byte_container_with_subtype/has_subtype/
+    constexpr bool has_subtype() const noexcept
+    {
+        return m_has_subtype;
+    }
+
+    /// @brief clears the binary subtype
+    /// @sa https://json.nlohmann.me/api/byte_container_with_subtype/clear_subtype/
+    void clear_subtype() noexcept
+    {
+        m_subtype = 0;
+        m_has_subtype = false;
+    }
+
+  private:
+    subtype_type m_subtype = 0;
+    bool m_has_subtype = false;
+};
+
+}  // namespace nlohmann
+
+// #include <nlohmann/detail/conversions/from_json.hpp>
+
+// #include <nlohmann/detail/conversions/to_json.hpp>
+
+// #include <nlohmann/detail/exceptions.hpp>
+
+// #include <nlohmann/detail/hash.hpp>
+
+
+#include <cstdint> // uint8_t
+#include <cstddef> // size_t
+#include <functional> // hash
+
+// #include <nlohmann/detail/macro_scope.hpp>
+
+// #include <nlohmann/detail/value_t.hpp>
+
+
+namespace nlohmann
+{
+namespace detail
+{
+
+// boost::hash_combine
+inline std::size_t combine(std::size_t seed, std::size_t h) noexcept
+{
+    seed ^= h + 0x9e3779b9 + (seed << 6U) + (seed >> 2U);
+    return seed;
 }
+
+/*!
+@brief hash a JSON value
+
+The hash function tries to rely on std::hash where possible. Furthermore, the
+type of the JSON value is taken into account to have different hash values for
+null, 0, 0U, and false, etc.
+
+@tparam BasicJsonType basic_json specialization
+@param j JSON value to hash
+@return hash value of j
+*/
+template<typename BasicJsonType>
+std::size_t hash(const BasicJsonType& j)
+{
+    using string_t = typename BasicJsonType::string_t;
+    using number_integer_t = typename BasicJsonType::number_integer_t;
+    using number_unsigned_t = typename BasicJsonType::number_unsigned_t;
+    using number_float_t = typename BasicJsonType::number_float_t;
+
+    const auto type = static_cast<std::size_t>(j.type());
+    switch (j.type())
+    {
+        case BasicJsonType::value_t::null:
+        case BasicJsonType::value_t::discarded:
+        {
+            return combine(type, 0);
+        }
+
+        case BasicJsonType::value_t::object:
+        {
+            auto seed = combine(type, j.size());
+            for (const auto& element : j.items())
+            {
+                const auto h = std::hash<string_t> {}(element.key());
+                seed = combine(seed, h);
+                seed = combine(seed, hash(element.value()));
+            }
+            return seed;
+        }
+
+        case BasicJsonType::value_t::array:
+        {
+            auto seed = combine(type, j.size());
+            for (const auto& element : j)
+            {
+                seed = combine(seed, hash(element));
+            }
+            return seed;
+        }
+
+        case BasicJsonType::value_t::string:
+        {
+            const auto h = std::hash<string_t> {}(j.template get_ref<const string_t&>());
+            return combine(type, h);
+        }
+
+        case BasicJsonType::value_t::boolean:
+        {
+            const auto h = std::hash<bool> {}(j.template get<bool>());
+            return combine(type, h);
+        }
+
+        case BasicJsonType::value_t::number_integer:
+        {
+            const auto h = std::hash<number_integer_t> {}(j.template get<number_integer_t>());
+            return combine(type, h);
+        }
+
+        case BasicJsonType::value_t::number_unsigned:
+        {
+            const auto h = std::hash<number_unsigned_t> {}(j.template get<number_unsigned_t>());
+            return combine(type, h);
+        }
+
+        case BasicJsonType::value_t::number_float:
+        {
+            const auto h = std::hash<number_float_t> {}(j.template get<number_float_t>());
+            return combine(type, h);
+        }
+
+        case BasicJsonType::value_t::binary:
+        {
+            auto seed = combine(type, j.get_binary().size());
+            const auto h = std::hash<bool> {}(j.get_binary().has_subtype());
+            seed = combine(seed, h);
+            seed = combine(seed, static_cast<std::size_t>(j.get_binary().subtype()));
+            for (const auto byte : j.get_binary())
+            {
+                seed = combine(seed, std::hash<std::uint8_t> {}(byte));
+            }
+            return seed;
+        }
+
+        default:                   // LCOV_EXCL_LINE
+            JSON_ASSERT(false); // NOLINT(cert-dcl03-c,hicpp-static-assert,misc-static-assert) LCOV_EXCL_LINE
+            return 0;              // LCOV_EXCL_LINE
+    }
 }
 
+}  // namespace detail
+}  // namespace nlohmann
+
+// #include <nlohmann/detail/input/binary_reader.hpp>
+
+
+#include <algorithm> // generate_n
+#include <array> // array
+#include <cmath> // ldexp
+#include <cstddef> // size_t
+#include <cstdint> // uint8_t, uint16_t, uint32_t, uint64_t
+#include <cstdio> // snprintf
+#include <cstring> // memcpy
+#include <iterator> // back_inserter
+#include <limits> // numeric_limits
+#include <string> // char_traits, string
+#include <utility> // make_pair, move
+#include <vector> // vector
+
+// #include <nlohmann/detail/exceptions.hpp>
+
 // #include <nlohmann/detail/input/input_adapters.hpp>
 
 
-#include <cassert> // assert
+#include <array> // array
 #include <cstddef> // size_t
 #include <cstring> // strlen
-#include <istream> // istream
 #include <iterator> // begin, end, iterator_traits, random_access_iterator_tag, distance, next
 #include <memory> // shared_ptr, make_shared, addressof
 #include <numeric> // accumulate
@@ -1888,6 +5215,13 @@ constexpr const auto& to_json = detail::static_const<detail::to_json_fn>::value;
 #include <type_traits> // enable_if, is_base_of, is_pointer, is_integral, remove_pointer
 #include <utility> // pair, declval
 
+#ifndef JSON_NO_IO
+    #include <cstdio>   // FILE *
+    #include <istream>  // istream
+#endif                  // JSON_NO_IO
+
+// #include <nlohmann/detail/iterators/iterator_traits.hpp>
+
 // #include <nlohmann/detail/macro_scope.hpp>
 
 
@@ -1896,32 +5230,44 @@ namespace nlohmann
 namespace detail
 {
 /// the supported input formats
-enum class input_format_t { json, cbor, msgpack, ubjson };
+enum class input_format_t { json, cbor, msgpack, ubjson, bson };
 
 ////////////////////
 // input adapters //
 ////////////////////
 
+#ifndef JSON_NO_IO
 /*!
-@brief abstract input adapter interface
-
-Produces a stream of std::char_traits<char>::int_type characters from a
-std::istream, a buffer, or some other input type. Accepts the return of
-exactly one non-EOF character for future input. The int_type characters
-returned consist of all valid char values as positive values (typically
-unsigned char), plus an EOF value outside that range, specified by the value
-of the function std::char_traits<char>::eof(). This value is typically -1, but
-could be any arbitrary value which is not a valid char value.
+Input adapter for stdio file access. This adapter read only 1 byte and do not use any
+ buffer. This adapter is a very low level adapter.
 */
-struct input_adapter_protocol
+class file_input_adapter
 {
-    /// get a character [0,255] or std::char_traits<char>::eof().
-    virtual std::char_traits<char>::int_type get_character() = 0;
-    virtual ~input_adapter_protocol() = default;
+  public:
+    using char_type = char;
+
+    JSON_HEDLEY_NON_NULL(2)
+    explicit file_input_adapter(std::FILE* f) noexcept
+        : m_file(f)
+    {}
+
+    // make class move-only
+    file_input_adapter(const file_input_adapter&) = delete;
+    file_input_adapter(file_input_adapter&&) noexcept = default;
+    file_input_adapter& operator=(const file_input_adapter&) = delete;
+    file_input_adapter& operator=(file_input_adapter&&) = delete;
+    ~file_input_adapter() = default;
+
+    std::char_traits<char>::int_type get_character() noexcept
+    {
+        return std::fgetc(m_file);
+    }
+
+  private:
+    /// the file pointer to read from
+    std::FILE* m_file;
 };
 
-/// a type to simplify interfaces
-using input_adapter_t = std::shared_ptr<input_adapter_protocol>;
 
 /*!
 Input adapter for a (caching) istream. Ignores a UFT Byte Order Mark at
@@ -1932,103 +5278,111 @@ characters following those used in parsing the JSON input.  Clears the
 std::istream flags; any input errors (e.g., EOF) will be detected by the first
 subsequent call for input from the std::istream.
 */
-class input_stream_adapter : public input_adapter_protocol
+class input_stream_adapter
 {
   public:
-    ~input_stream_adapter() override
+    using char_type = char;
+
+    ~input_stream_adapter()
     {
         // clear stream flags; we use underlying streambuf I/O, do not
-        // maintain ifstream flags
-        is.clear();
+        // maintain ifstream flags, except eof
+        if (is != nullptr)
+        {
+            is->clear(is->rdstate() & std::ios::eofbit);
+        }
     }
 
     explicit input_stream_adapter(std::istream& i)
-        : is(i), sb(*i.rdbuf())
+        : is(&i), sb(i.rdbuf())
     {}
 
     // delete because of pointer members
     input_stream_adapter(const input_stream_adapter&) = delete;
     input_stream_adapter& operator=(input_stream_adapter&) = delete;
+    input_stream_adapter& operator=(input_stream_adapter&&) = delete;
+
+    input_stream_adapter(input_stream_adapter&& rhs) noexcept
+        : is(rhs.is), sb(rhs.sb)
+    {
+        rhs.is = nullptr;
+        rhs.sb = nullptr;
+    }
 
     // std::istream/std::streambuf use std::char_traits<char>::to_int_type, to
     // ensure that std::char_traits<char>::eof() and the character 0xFF do not
-    // end up as the same value, eg. 0xFFFFFFFF.
-    std::char_traits<char>::int_type get_character() override
+    // end up as the same value, e.g. 0xFFFFFFFF.
+    std::char_traits<char>::int_type get_character()
     {
-        return sb.sbumpc();
+        auto res = sb->sbumpc();
+        // set eof manually, as we don't use the istream interface.
+        if (JSON_HEDLEY_UNLIKELY(res == std::char_traits<char>::eof()))
+        {
+            is->clear(is->rdstate() | std::ios::eofbit);
+        }
+        return res;
     }
 
   private:
     /// the associated input stream
-    std::istream& is;
-    std::streambuf& sb;
+    std::istream* is = nullptr;
+    std::streambuf* sb = nullptr;
 };
+#endif  // JSON_NO_IO
 
-/// input adapter for buffer input
-class input_buffer_adapter : public input_adapter_protocol
+// General-purpose iterator-based adapter. It might not be as fast as
+// theoretically possible for some containers, but it is extremely versatile.
+template<typename IteratorType>
+class iterator_input_adapter
 {
   public:
-    input_buffer_adapter(const char* b, const std::size_t l)
-        : cursor(b), limit(b + l)
+    using char_type = typename std::iterator_traits<IteratorType>::value_type;
+
+    iterator_input_adapter(IteratorType first, IteratorType last)
+        : current(std::move(first)), end(std::move(last))
     {}
 
-    // delete because of pointer members
-    input_buffer_adapter(const input_buffer_adapter&) = delete;
-    input_buffer_adapter& operator=(input_buffer_adapter&) = delete;
-
-    std::char_traits<char>::int_type get_character() noexcept override
+    typename std::char_traits<char_type>::int_type get_character()
     {
-        if (JSON_LIKELY(cursor < limit))
+        if (JSON_HEDLEY_LIKELY(current != end))
         {
-            return std::char_traits<char>::to_int_type(*(cursor++));
+            auto result = std::char_traits<char_type>::to_int_type(*current);
+            std::advance(current, 1);
+            return result;
         }
 
-        return std::char_traits<char>::eof();
+        return std::char_traits<char_type>::eof();
     }
 
   private:
-    /// pointer to the current character
-    const char* cursor;
-    /// pointer past the last character
-    const char* const limit;
+    IteratorType current;
+    IteratorType end;
+
+    template<typename BaseInputAdapter, size_t T>
+    friend struct wide_string_input_helper;
+
+    bool empty() const
+    {
+        return current == end;
+    }
 };
 
-template<typename WideStringType>
-class wide_string_input_adapter : public input_adapter_protocol
+
+template<typename BaseInputAdapter, size_t T>
+struct wide_string_input_helper;
+
+template<typename BaseInputAdapter>
+struct wide_string_input_helper<BaseInputAdapter, 4>
 {
-  public:
-    explicit wide_string_input_adapter(const WideStringType& w) : str(w) {}
-
-    std::char_traits<char>::int_type get_character() noexcept override
-    {
-        // check if buffer needs to be filled
-        if (utf8_bytes_index == utf8_bytes_filled)
-        {
-            if (sizeof(typename WideStringType::value_type) == 2)
-            {
-                fill_buffer_utf16();
-            }
-            else
-            {
-                fill_buffer_utf32();
-            }
-
-            assert(utf8_bytes_filled > 0);
-            assert(utf8_bytes_index == 0);
-        }
-
-        // use buffer
-        assert(utf8_bytes_filled > 0);
-        assert(utf8_bytes_index < utf8_bytes_filled);
-        return utf8_bytes[utf8_bytes_index++];
-    }
-
-  private:
-    void fill_buffer_utf16()
+    // UTF-32
+    static void fill_buffer(BaseInputAdapter& input,
+                            std::array<std::char_traits<char>::int_type, 4>& utf8_bytes,
+                            size_t& utf8_bytes_index,
+                            size_t& utf8_bytes_filled)
     {
         utf8_bytes_index = 0;
 
-        if (current_wchar == str.size())
+        if (JSON_HEDLEY_UNLIKELY(input.empty()))
         {
             utf8_bytes[0] = std::char_traits<char>::eof();
             utf8_bytes_filled = 1;
@@ -2036,106 +5390,142 @@ class wide_string_input_adapter : public input_adapter_protocol
         else
         {
             // get the current character
-            const int wc = static_cast<int>(str[current_wchar++]);
-
-            // UTF-16 to UTF-8 encoding
-            if (wc < 0x80)
-            {
-                utf8_bytes[0] = wc;
-                utf8_bytes_filled = 1;
-            }
-            else if (wc <= 0x7FF)
-            {
-                utf8_bytes[0] = 0xC0 | ((wc >> 6));
-                utf8_bytes[1] = 0x80 | (wc & 0x3F);
-                utf8_bytes_filled = 2;
-            }
-            else if (0xD800 > wc or wc >= 0xE000)
-            {
-                utf8_bytes[0] = 0xE0 | ((wc >> 12));
-                utf8_bytes[1] = 0x80 | ((wc >> 6) & 0x3F);
-                utf8_bytes[2] = 0x80 | (wc & 0x3F);
-                utf8_bytes_filled = 3;
-            }
-            else
-            {
-                if (current_wchar < str.size())
-                {
-                    const int wc2 = static_cast<int>(str[current_wchar++]);
-                    const int charcode = 0x10000 + (((wc & 0x3FF) << 10) | (wc2 & 0x3FF));
-                    utf8_bytes[0] = 0xf0 | (charcode >> 18);
-                    utf8_bytes[1] = 0x80 | ((charcode >> 12) & 0x3F);
-                    utf8_bytes[2] = 0x80 | ((charcode >> 6) & 0x3F);
-                    utf8_bytes[3] = 0x80 | (charcode & 0x3F);
-                    utf8_bytes_filled = 4;
-                }
-                else
-                {
-                    // unknown character
-                    ++current_wchar;
-                    utf8_bytes[0] = wc;
-                    utf8_bytes_filled = 1;
-                }
-            }
-        }
-    }
-
-    void fill_buffer_utf32()
-    {
-        utf8_bytes_index = 0;
-
-        if (current_wchar == str.size())
-        {
-            utf8_bytes[0] = std::char_traits<char>::eof();
-            utf8_bytes_filled = 1;
-        }
-        else
-        {
-            // get the current character
-            const int wc = static_cast<int>(str[current_wchar++]);
+            const auto wc = input.get_character();
 
             // UTF-32 to UTF-8 encoding
             if (wc < 0x80)
             {
-                utf8_bytes[0] = wc;
+                utf8_bytes[0] = static_cast<std::char_traits<char>::int_type>(wc);
                 utf8_bytes_filled = 1;
             }
             else if (wc <= 0x7FF)
             {
-                utf8_bytes[0] = 0xC0 | ((wc >> 6) & 0x1F);
-                utf8_bytes[1] = 0x80 | (wc & 0x3F);
+                utf8_bytes[0] = static_cast<std::char_traits<char>::int_type>(0xC0u | ((static_cast<unsigned int>(wc) >> 6u) & 0x1Fu));
+                utf8_bytes[1] = static_cast<std::char_traits<char>::int_type>(0x80u | (static_cast<unsigned int>(wc) & 0x3Fu));
                 utf8_bytes_filled = 2;
             }
             else if (wc <= 0xFFFF)
             {
-                utf8_bytes[0] = 0xE0 | ((wc >> 12) & 0x0F);
-                utf8_bytes[1] = 0x80 | ((wc >> 6) & 0x3F);
-                utf8_bytes[2] = 0x80 | (wc & 0x3F);
+                utf8_bytes[0] = static_cast<std::char_traits<char>::int_type>(0xE0u | ((static_cast<unsigned int>(wc) >> 12u) & 0x0Fu));
+                utf8_bytes[1] = static_cast<std::char_traits<char>::int_type>(0x80u | ((static_cast<unsigned int>(wc) >> 6u) & 0x3Fu));
+                utf8_bytes[2] = static_cast<std::char_traits<char>::int_type>(0x80u | (static_cast<unsigned int>(wc) & 0x3Fu));
                 utf8_bytes_filled = 3;
             }
             else if (wc <= 0x10FFFF)
             {
-                utf8_bytes[0] = 0xF0 | ((wc >> 18 ) & 0x07);
-                utf8_bytes[1] = 0x80 | ((wc >> 12) & 0x3F);
-                utf8_bytes[2] = 0x80 | ((wc >> 6) & 0x3F);
-                utf8_bytes[3] = 0x80 | (wc & 0x3F);
+                utf8_bytes[0] = static_cast<std::char_traits<char>::int_type>(0xF0u | ((static_cast<unsigned int>(wc) >> 18u) & 0x07u));
+                utf8_bytes[1] = static_cast<std::char_traits<char>::int_type>(0x80u | ((static_cast<unsigned int>(wc) >> 12u) & 0x3Fu));
+                utf8_bytes[2] = static_cast<std::char_traits<char>::int_type>(0x80u | ((static_cast<unsigned int>(wc) >> 6u) & 0x3Fu));
+                utf8_bytes[3] = static_cast<std::char_traits<char>::int_type>(0x80u | (static_cast<unsigned int>(wc) & 0x3Fu));
                 utf8_bytes_filled = 4;
             }
             else
             {
                 // unknown character
-                utf8_bytes[0] = wc;
+                utf8_bytes[0] = static_cast<std::char_traits<char>::int_type>(wc);
                 utf8_bytes_filled = 1;
             }
         }
     }
+};
+
+template<typename BaseInputAdapter>
+struct wide_string_input_helper<BaseInputAdapter, 2>
+{
+    // UTF-16
+    static void fill_buffer(BaseInputAdapter& input,
+                            std::array<std::char_traits<char>::int_type, 4>& utf8_bytes,
+                            size_t& utf8_bytes_index,
+                            size_t& utf8_bytes_filled)
+    {
+        utf8_bytes_index = 0;
+
+        if (JSON_HEDLEY_UNLIKELY(input.empty()))
+        {
+            utf8_bytes[0] = std::char_traits<char>::eof();
+            utf8_bytes_filled = 1;
+        }
+        else
+        {
+            // get the current character
+            const auto wc = input.get_character();
+
+            // UTF-16 to UTF-8 encoding
+            if (wc < 0x80)
+            {
+                utf8_bytes[0] = static_cast<std::char_traits<char>::int_type>(wc);
+                utf8_bytes_filled = 1;
+            }
+            else if (wc <= 0x7FF)
+            {
+                utf8_bytes[0] = static_cast<std::char_traits<char>::int_type>(0xC0u | ((static_cast<unsigned int>(wc) >> 6u)));
+                utf8_bytes[1] = static_cast<std::char_traits<char>::int_type>(0x80u | (static_cast<unsigned int>(wc) & 0x3Fu));
+                utf8_bytes_filled = 2;
+            }
+            else if (0xD800 > wc || wc >= 0xE000)
+            {
+                utf8_bytes[0] = static_cast<std::char_traits<char>::int_type>(0xE0u | ((static_cast<unsigned int>(wc) >> 12u)));
+                utf8_bytes[1] = static_cast<std::char_traits<char>::int_type>(0x80u | ((static_cast<unsigned int>(wc) >> 6u) & 0x3Fu));
+                utf8_bytes[2] = static_cast<std::char_traits<char>::int_type>(0x80u | (static_cast<unsigned int>(wc) & 0x3Fu));
+                utf8_bytes_filled = 3;
+            }
+            else
+            {
+                if (JSON_HEDLEY_UNLIKELY(!input.empty()))
+                {
+                    const auto wc2 = static_cast<unsigned int>(input.get_character());
+                    const auto charcode = 0x10000u + (((static_cast<unsigned int>(wc) & 0x3FFu) << 10u) | (wc2 & 0x3FFu));
+                    utf8_bytes[0] = static_cast<std::char_traits<char>::int_type>(0xF0u | (charcode >> 18u));
+                    utf8_bytes[1] = static_cast<std::char_traits<char>::int_type>(0x80u | ((charcode >> 12u) & 0x3Fu));
+                    utf8_bytes[2] = static_cast<std::char_traits<char>::int_type>(0x80u | ((charcode >> 6u) & 0x3Fu));
+                    utf8_bytes[3] = static_cast<std::char_traits<char>::int_type>(0x80u | (charcode & 0x3Fu));
+                    utf8_bytes_filled = 4;
+                }
+                else
+                {
+                    utf8_bytes[0] = static_cast<std::char_traits<char>::int_type>(wc);
+                    utf8_bytes_filled = 1;
+                }
+            }
+        }
+    }
+};
+
+// Wraps another input apdater to convert wide character types into individual bytes.
+template<typename BaseInputAdapter, typename WideCharType>
+class wide_string_input_adapter
+{
+  public:
+    using char_type = char;
+
+    wide_string_input_adapter(BaseInputAdapter base)
+        : base_adapter(base) {}
+
+    typename std::char_traits<char>::int_type get_character() noexcept
+    {
+        // check if buffer needs to be filled
+        if (utf8_bytes_index == utf8_bytes_filled)
+        {
+            fill_buffer<sizeof(WideCharType)>();
+
+            JSON_ASSERT(utf8_bytes_filled > 0);
+            JSON_ASSERT(utf8_bytes_index == 0);
+        }
+
+        // use buffer
+        JSON_ASSERT(utf8_bytes_filled > 0);
+        JSON_ASSERT(utf8_bytes_index < utf8_bytes_filled);
+        return utf8_bytes[utf8_bytes_index++];
+    }
 
   private:
-    /// the wstring to process
-    const WideStringType& str;
+    BaseInputAdapter base_adapter;
 
-    /// index of the current wchar in str
-    std::size_t current_wchar = 0;
+    template<size_t T>
+    void fill_buffer()
+    {
+        wide_string_input_helper<BaseInputAdapter, T>::fill_buffer(base_adapter, utf8_bytes, utf8_bytes_index, utf8_bytes_filled);
+    }
 
     /// a buffer for UTF-8 bytes
     std::array<std::char_traits<char>::int_type, 4> utf8_bytes = {{0, 0, 0, 0}};
@@ -2146,129 +5536,893 @@ class wide_string_input_adapter : public input_adapter_protocol
     std::size_t utf8_bytes_filled = 0;
 };
 
-class input_adapter
+
+template<typename IteratorType, typename Enable = void>
+struct iterator_input_adapter_factory
+{
+    using iterator_type = IteratorType;
+    using char_type = typename std::iterator_traits<iterator_type>::value_type;
+    using adapter_type = iterator_input_adapter<iterator_type>;
+
+    static adapter_type create(IteratorType first, IteratorType last)
+    {
+        return adapter_type(std::move(first), std::move(last));
+    }
+};
+
+template<typename T>
+struct is_iterator_of_multibyte
+{
+    using value_type = typename std::iterator_traits<T>::value_type;
+    enum
+    {
+        value = sizeof(value_type) > 1
+    };
+};
+
+template<typename IteratorType>
+struct iterator_input_adapter_factory<IteratorType, enable_if_t<is_iterator_of_multibyte<IteratorType>::value>>
+{
+    using iterator_type = IteratorType;
+    using char_type = typename std::iterator_traits<iterator_type>::value_type;
+    using base_adapter_type = iterator_input_adapter<iterator_type>;
+    using adapter_type = wide_string_input_adapter<base_adapter_type, char_type>;
+
+    static adapter_type create(IteratorType first, IteratorType last)
+    {
+        return adapter_type(base_adapter_type(std::move(first), std::move(last)));
+    }
+};
+
+// General purpose iterator-based input
+template<typename IteratorType>
+typename iterator_input_adapter_factory<IteratorType>::adapter_type input_adapter(IteratorType first, IteratorType last)
+{
+    using factory_type = iterator_input_adapter_factory<IteratorType>;
+    return factory_type::create(first, last);
+}
+
+// Convenience shorthand from container to iterator
+// Enables ADL on begin(container) and end(container)
+// Encloses the using declarations in namespace for not to leak them to outside scope
+
+namespace container_input_adapter_factory_impl
+{
+
+using std::begin;
+using std::end;
+
+template<typename ContainerType, typename Enable = void>
+struct container_input_adapter_factory {};
+
+template<typename ContainerType>
+struct container_input_adapter_factory< ContainerType,
+       void_t<decltype(begin(std::declval<ContainerType>()), end(std::declval<ContainerType>()))>>
+       {
+           using adapter_type = decltype(input_adapter(begin(std::declval<ContainerType>()), end(std::declval<ContainerType>())));
+
+           static adapter_type create(const ContainerType& container)
+{
+    return input_adapter(begin(container), end(container));
+}
+       };
+
+} // namespace container_input_adapter_factory_impl
+
+template<typename ContainerType>
+typename container_input_adapter_factory_impl::container_input_adapter_factory<ContainerType>::adapter_type input_adapter(const ContainerType& container)
+{
+    return container_input_adapter_factory_impl::container_input_adapter_factory<ContainerType>::create(container);
+}
+
+#ifndef JSON_NO_IO
+// Special cases with fast paths
+inline file_input_adapter input_adapter(std::FILE* file)
+{
+    return file_input_adapter(file);
+}
+
+inline input_stream_adapter input_adapter(std::istream& stream)
+{
+    return input_stream_adapter(stream);
+}
+
+inline input_stream_adapter input_adapter(std::istream&& stream)
+{
+    return input_stream_adapter(stream);
+}
+#endif  // JSON_NO_IO
+
+using contiguous_bytes_input_adapter = decltype(input_adapter(std::declval<const char*>(), std::declval<const char*>()));
+
+// Null-delimited strings, and the like.
+template < typename CharT,
+           typename std::enable_if <
+               std::is_pointer<CharT>::value&&
+               !std::is_array<CharT>::value&&
+               std::is_integral<typename std::remove_pointer<CharT>::type>::value&&
+               sizeof(typename std::remove_pointer<CharT>::type) == 1,
+               int >::type = 0 >
+contiguous_bytes_input_adapter input_adapter(CharT b)
+{
+    auto length = std::strlen(reinterpret_cast<const char*>(b));
+    const auto* ptr = reinterpret_cast<const char*>(b);
+    return input_adapter(ptr, ptr + length);
+}
+
+template<typename T, std::size_t N>
+auto input_adapter(T (&array)[N]) -> decltype(input_adapter(array, array + N)) // NOLINT(cppcoreguidelines-avoid-c-arrays,hicpp-avoid-c-arrays,modernize-avoid-c-arrays)
+{
+    return input_adapter(array, array + N);
+}
+
+// This class only handles inputs of input_buffer_adapter type.
+// It's required so that expressions like {ptr, len} can be implicitly cast
+// to the correct adapter.
+class span_input_adapter
 {
   public:
-    // native support
+    template < typename CharT,
+               typename std::enable_if <
+                   std::is_pointer<CharT>::value&&
+                   std::is_integral<typename std::remove_pointer<CharT>::type>::value&&
+                   sizeof(typename std::remove_pointer<CharT>::type) == 1,
+                   int >::type = 0 >
+    span_input_adapter(CharT b, std::size_t l)
+        : ia(reinterpret_cast<const char*>(b), reinterpret_cast<const char*>(b) + l) {}
 
-    /// input adapter for input stream
-    input_adapter(std::istream& i)
-        : ia(std::make_shared<input_stream_adapter>(i)) {}
-
-    /// input adapter for input stream
-    input_adapter(std::istream&& i)
-        : ia(std::make_shared<input_stream_adapter>(i)) {}
-
-    input_adapter(const std::wstring& ws)
-        : ia(std::make_shared<wide_string_input_adapter<std::wstring>>(ws)) {}
-
-    input_adapter(const std::u16string& ws)
-        : ia(std::make_shared<wide_string_input_adapter<std::u16string>>(ws)) {}
-
-    input_adapter(const std::u32string& ws)
-        : ia(std::make_shared<wide_string_input_adapter<std::u32string>>(ws)) {}
-
-    /// input adapter for buffer
-    template<typename CharT,
-             typename std::enable_if<
-                 std::is_pointer<CharT>::value and
-                 std::is_integral<typename std::remove_pointer<CharT>::type>::value and
-                 sizeof(typename std::remove_pointer<CharT>::type) == 1,
-                 int>::type = 0>
-    input_adapter(CharT b, std::size_t l)
-        : ia(std::make_shared<input_buffer_adapter>(reinterpret_cast<const char*>(b), l)) {}
-
-    // derived support
-
-    /// input adapter for string literal
-    template<typename CharT,
-             typename std::enable_if<
-                 std::is_pointer<CharT>::value and
-                 std::is_integral<typename std::remove_pointer<CharT>::type>::value and
-                 sizeof(typename std::remove_pointer<CharT>::type) == 1,
-                 int>::type = 0>
-    input_adapter(CharT b)
-        : input_adapter(reinterpret_cast<const char*>(b),
-                        std::strlen(reinterpret_cast<const char*>(b))) {}
-
-    /// input adapter for iterator range with contiguous storage
     template<class IteratorType,
              typename std::enable_if<
-                 std::is_same<typename std::iterator_traits<IteratorType>::iterator_category, std::random_access_iterator_tag>::value,
+                 std::is_same<typename iterator_traits<IteratorType>::iterator_category, std::random_access_iterator_tag>::value,
                  int>::type = 0>
-    input_adapter(IteratorType first, IteratorType last)
+    span_input_adapter(IteratorType first, IteratorType last)
+        : ia(input_adapter(first, last)) {}
+
+    contiguous_bytes_input_adapter&& get()
     {
-#ifndef NDEBUG
-        // assertion to check that the iterator range is indeed contiguous,
-        // see http://stackoverflow.com/a/35008842/266378 for more discussion
-        const auto is_contiguous = std::accumulate(
-                                       first, last, std::pair<bool, int>(true, 0),
-                                       [&first](std::pair<bool, int> res, decltype(*first) val)
-        {
-            res.first &= (val == *(std::next(std::addressof(*first), res.second++)));
-            return res;
-        }).first;
-        assert(is_contiguous);
-#endif
-
-        // assertion to check that each element is 1 byte long
-        static_assert(
-            sizeof(typename std::iterator_traits<IteratorType>::value_type) == 1,
-            "each element in the iterator range must have the size of 1 byte");
-
-        const auto len = static_cast<size_t>(std::distance(first, last));
-        if (JSON_LIKELY(len > 0))
-        {
-            // there is at least one element: use the address of first
-            ia = std::make_shared<input_buffer_adapter>(reinterpret_cast<const char*>(&(*first)), len);
-        }
-        else
-        {
-            // the address of first cannot be used: use nullptr
-            ia = std::make_shared<input_buffer_adapter>(nullptr, len);
-        }
-    }
-
-    /// input adapter for array
-    template<class T, std::size_t N>
-    input_adapter(T (&array)[N])
-        : input_adapter(std::begin(array), std::end(array)) {}
-
-    /// input adapter for contiguous container
-    template<class ContiguousContainer, typename
-             std::enable_if<not std::is_pointer<ContiguousContainer>::value and
-                            std::is_base_of<std::random_access_iterator_tag, typename std::iterator_traits<decltype(std::begin(std::declval<ContiguousContainer const>()))>::iterator_category>::value,
-                            int>::type = 0>
-    input_adapter(const ContiguousContainer& c)
-        : input_adapter(std::begin(c), std::end(c)) {}
-
-    operator input_adapter_t()
-    {
-        return ia;
+        return std::move(ia); // NOLINT(hicpp-move-const-arg,performance-move-const-arg)
     }
 
   private:
-    /// the actual adapter
-    input_adapter_t ia = nullptr;
+    contiguous_bytes_input_adapter ia;
 };
-}
-}
+}  // namespace detail
+}  // namespace nlohmann
+
+// #include <nlohmann/detail/input/json_sax.hpp>
+
+
+#include <cstddef>
+#include <string> // string
+#include <utility> // move
+#include <vector> // vector
+
+// #include <nlohmann/detail/exceptions.hpp>
+
+// #include <nlohmann/detail/macro_scope.hpp>
+
+
+namespace nlohmann
+{
+
+/*!
+@brief SAX interface
+
+This class describes the SAX interface used by @ref nlohmann::json::sax_parse.
+Each function is called in different situations while the input is parsed. The
+boolean return value informs the parser whether to continue processing the
+input.
+*/
+template<typename BasicJsonType>
+struct json_sax
+{
+    using number_integer_t = typename BasicJsonType::number_integer_t;
+    using number_unsigned_t = typename BasicJsonType::number_unsigned_t;
+    using number_float_t = typename BasicJsonType::number_float_t;
+    using string_t = typename BasicJsonType::string_t;
+    using binary_t = typename BasicJsonType::binary_t;
+
+    /*!
+    @brief a null value was read
+    @return whether parsing should proceed
+    */
+    virtual bool null() = 0;
+
+    /*!
+    @brief a boolean value was read
+    @param[in] val  boolean value
+    @return whether parsing should proceed
+    */
+    virtual bool boolean(bool val) = 0;
+
+    /*!
+    @brief an integer number was read
+    @param[in] val  integer value
+    @return whether parsing should proceed
+    */
+    virtual bool number_integer(number_integer_t val) = 0;
+
+    /*!
+    @brief an unsigned integer number was read
+    @param[in] val  unsigned integer value
+    @return whether parsing should proceed
+    */
+    virtual bool number_unsigned(number_unsigned_t val) = 0;
+
+    /*!
+    @brief a floating-point number was read
+    @param[in] val  floating-point value
+    @param[in] s    raw token value
+    @return whether parsing should proceed
+    */
+    virtual bool number_float(number_float_t val, const string_t& s) = 0;
+
+    /*!
+    @brief a string value was read
+    @param[in] val  string value
+    @return whether parsing should proceed
+    @note It is safe to move the passed string value.
+    */
+    virtual bool string(string_t& val) = 0;
+
+    /*!
+    @brief a binary value was read
+    @param[in] val  binary value
+    @return whether parsing should proceed
+    @note It is safe to move the passed binary value.
+    */
+    virtual bool binary(binary_t& val) = 0;
+
+    /*!
+    @brief the beginning of an object was read
+    @param[in] elements  number of object elements or -1 if unknown
+    @return whether parsing should proceed
+    @note binary formats may report the number of elements
+    */
+    virtual bool start_object(std::size_t elements) = 0;
+
+    /*!
+    @brief an object key was read
+    @param[in] val  object key
+    @return whether parsing should proceed
+    @note It is safe to move the passed string.
+    */
+    virtual bool key(string_t& val) = 0;
+
+    /*!
+    @brief the end of an object was read
+    @return whether parsing should proceed
+    */
+    virtual bool end_object() = 0;
+
+    /*!
+    @brief the beginning of an array was read
+    @param[in] elements  number of array elements or -1 if unknown
+    @return whether parsing should proceed
+    @note binary formats may report the number of elements
+    */
+    virtual bool start_array(std::size_t elements) = 0;
+
+    /*!
+    @brief the end of an array was read
+    @return whether parsing should proceed
+    */
+    virtual bool end_array() = 0;
+
+    /*!
+    @brief a parse error occurred
+    @param[in] position    the position in the input where the error occurs
+    @param[in] last_token  the last read token
+    @param[in] ex          an exception object describing the error
+    @return whether parsing should proceed (must return false)
+    */
+    virtual bool parse_error(std::size_t position,
+                             const std::string& last_token,
+                             const detail::exception& ex) = 0;
+
+    json_sax() = default;
+    json_sax(const json_sax&) = default;
+    json_sax(json_sax&&) noexcept = default;
+    json_sax& operator=(const json_sax&) = default;
+    json_sax& operator=(json_sax&&) noexcept = default;
+    virtual ~json_sax() = default;
+};
+
+
+namespace detail
+{
+/*!
+@brief SAX implementation to create a JSON value from SAX events
+
+This class implements the @ref json_sax interface and processes the SAX events
+to create a JSON value which makes it basically a DOM parser. The structure or
+hierarchy of the JSON value is managed by the stack `ref_stack` which contains
+a pointer to the respective array or object for each recursion depth.
+
+After successful parsing, the value that is passed by reference to the
+constructor contains the parsed value.
+
+@tparam BasicJsonType  the JSON type
+*/
+template<typename BasicJsonType>
+class json_sax_dom_parser
+{
+  public:
+    using number_integer_t = typename BasicJsonType::number_integer_t;
+    using number_unsigned_t = typename BasicJsonType::number_unsigned_t;
+    using number_float_t = typename BasicJsonType::number_float_t;
+    using string_t = typename BasicJsonType::string_t;
+    using binary_t = typename BasicJsonType::binary_t;
+
+    /*!
+    @param[in,out] r  reference to a JSON value that is manipulated while
+                       parsing
+    @param[in] allow_exceptions_  whether parse errors yield exceptions
+    */
+    explicit json_sax_dom_parser(BasicJsonType& r, const bool allow_exceptions_ = true)
+        : root(r), allow_exceptions(allow_exceptions_)
+    {}
+
+    // make class move-only
+    json_sax_dom_parser(const json_sax_dom_parser&) = delete;
+    json_sax_dom_parser(json_sax_dom_parser&&) = default; // NOLINT(hicpp-noexcept-move,performance-noexcept-move-constructor)
+    json_sax_dom_parser& operator=(const json_sax_dom_parser&) = delete;
+    json_sax_dom_parser& operator=(json_sax_dom_parser&&) = default; // NOLINT(hicpp-noexcept-move,performance-noexcept-move-constructor)
+    ~json_sax_dom_parser() = default;
+
+    bool null()
+    {
+        handle_value(nullptr);
+        return true;
+    }
+
+    bool boolean(bool val)
+    {
+        handle_value(val);
+        return true;
+    }
+
+    bool number_integer(number_integer_t val)
+    {
+        handle_value(val);
+        return true;
+    }
+
+    bool number_unsigned(number_unsigned_t val)
+    {
+        handle_value(val);
+        return true;
+    }
+
+    bool number_float(number_float_t val, const string_t& /*unused*/)
+    {
+        handle_value(val);
+        return true;
+    }
+
+    bool string(string_t& val)
+    {
+        handle_value(val);
+        return true;
+    }
+
+    bool binary(binary_t& val)
+    {
+        handle_value(std::move(val));
+        return true;
+    }
+
+    bool start_object(std::size_t len)
+    {
+        ref_stack.push_back(handle_value(BasicJsonType::value_t::object));
+
+        if (JSON_HEDLEY_UNLIKELY(len != static_cast<std::size_t>(-1) && len > ref_stack.back()->max_size()))
+        {
+            JSON_THROW(out_of_range::create(408, "excessive object size: " + std::to_string(len), *ref_stack.back()));
+        }
+
+        return true;
+    }
+
+    bool key(string_t& val)
+    {
+        // add null at given key and store the reference for later
+        object_element = &(ref_stack.back()->m_value.object->operator[](val));
+        return true;
+    }
+
+    bool end_object()
+    {
+        ref_stack.back()->set_parents();
+        ref_stack.pop_back();
+        return true;
+    }
+
+    bool start_array(std::size_t len)
+    {
+        ref_stack.push_back(handle_value(BasicJsonType::value_t::array));
+
+        if (JSON_HEDLEY_UNLIKELY(len != static_cast<std::size_t>(-1) && len > ref_stack.back()->max_size()))
+        {
+            JSON_THROW(out_of_range::create(408, "excessive array size: " + std::to_string(len), *ref_stack.back()));
+        }
+
+        return true;
+    }
+
+    bool end_array()
+    {
+        ref_stack.back()->set_parents();
+        ref_stack.pop_back();
+        return true;
+    }
+
+    template<class Exception>
+    bool parse_error(std::size_t /*unused*/, const std::string& /*unused*/,
+                     const Exception& ex)
+    {
+        errored = true;
+        static_cast<void>(ex);
+        if (allow_exceptions)
+        {
+            JSON_THROW(ex);
+        }
+        return false;
+    }
+
+    constexpr bool is_errored() const
+    {
+        return errored;
+    }
+
+  private:
+    /*!
+    @invariant If the ref stack is empty, then the passed value will be the new
+               root.
+    @invariant If the ref stack contains a value, then it is an array or an
+               object to which we can add elements
+    */
+    template<typename Value>
+    JSON_HEDLEY_RETURNS_NON_NULL
+    BasicJsonType* handle_value(Value&& v)
+    {
+        if (ref_stack.empty())
+        {
+            root = BasicJsonType(std::forward<Value>(v));
+            return &root;
+        }
+
+        JSON_ASSERT(ref_stack.back()->is_array() || ref_stack.back()->is_object());
+
+        if (ref_stack.back()->is_array())
+        {
+            ref_stack.back()->m_value.array->emplace_back(std::forward<Value>(v));
+            return &(ref_stack.back()->m_value.array->back());
+        }
+
+        JSON_ASSERT(ref_stack.back()->is_object());
+        JSON_ASSERT(object_element);
+        *object_element = BasicJsonType(std::forward<Value>(v));
+        return object_element;
+    }
+
+    /// the parsed JSON value
+    BasicJsonType& root;
+    /// stack to model hierarchy of values
+    std::vector<BasicJsonType*> ref_stack {};
+    /// helper to hold the reference for the next object element
+    BasicJsonType* object_element = nullptr;
+    /// whether a syntax error occurred
+    bool errored = false;
+    /// whether to throw exceptions in case of errors
+    const bool allow_exceptions = true;
+};
+
+template<typename BasicJsonType>
+class json_sax_dom_callback_parser
+{
+  public:
+    using number_integer_t = typename BasicJsonType::number_integer_t;
+    using number_unsigned_t = typename BasicJsonType::number_unsigned_t;
+    using number_float_t = typename BasicJsonType::number_float_t;
+    using string_t = typename BasicJsonType::string_t;
+    using binary_t = typename BasicJsonType::binary_t;
+    using parser_callback_t = typename BasicJsonType::parser_callback_t;
+    using parse_event_t = typename BasicJsonType::parse_event_t;
+
+    json_sax_dom_callback_parser(BasicJsonType& r,
+                                 const parser_callback_t cb,
+                                 const bool allow_exceptions_ = true)
+        : root(r), callback(cb), allow_exceptions(allow_exceptions_)
+    {
+        keep_stack.push_back(true);
+    }
+
+    // make class move-only
+    json_sax_dom_callback_parser(const json_sax_dom_callback_parser&) = delete;
+    json_sax_dom_callback_parser(json_sax_dom_callback_parser&&) = default; // NOLINT(hicpp-noexcept-move,performance-noexcept-move-constructor)
+    json_sax_dom_callback_parser& operator=(const json_sax_dom_callback_parser&) = delete;
+    json_sax_dom_callback_parser& operator=(json_sax_dom_callback_parser&&) = default; // NOLINT(hicpp-noexcept-move,performance-noexcept-move-constructor)
+    ~json_sax_dom_callback_parser() = default;
+
+    bool null()
+    {
+        handle_value(nullptr);
+        return true;
+    }
+
+    bool boolean(bool val)
+    {
+        handle_value(val);
+        return true;
+    }
+
+    bool number_integer(number_integer_t val)
+    {
+        handle_value(val);
+        return true;
+    }
+
+    bool number_unsigned(number_unsigned_t val)
+    {
+        handle_value(val);
+        return true;
+    }
+
+    bool number_float(number_float_t val, const string_t& /*unused*/)
+    {
+        handle_value(val);
+        return true;
+    }
+
+    bool string(string_t& val)
+    {
+        handle_value(val);
+        return true;
+    }
+
+    bool binary(binary_t& val)
+    {
+        handle_value(std::move(val));
+        return true;
+    }
+
+    bool start_object(std::size_t len)
+    {
+        // check callback for object start
+        const bool keep = callback(static_cast<int>(ref_stack.size()), parse_event_t::object_start, discarded);
+        keep_stack.push_back(keep);
+
+        auto val = handle_value(BasicJsonType::value_t::object, true);
+        ref_stack.push_back(val.second);
+
+        // check object limit
+        if (ref_stack.back() && JSON_HEDLEY_UNLIKELY(len != static_cast<std::size_t>(-1) && len > ref_stack.back()->max_size()))
+        {
+            JSON_THROW(out_of_range::create(408, "excessive object size: " + std::to_string(len), *ref_stack.back()));
+        }
+
+        return true;
+    }
+
+    bool key(string_t& val)
+    {
+        BasicJsonType k = BasicJsonType(val);
+
+        // check callback for key
+        const bool keep = callback(static_cast<int>(ref_stack.size()), parse_event_t::key, k);
+        key_keep_stack.push_back(keep);
+
+        // add discarded value at given key and store the reference for later
+        if (keep && ref_stack.back())
+        {
+            object_element = &(ref_stack.back()->m_value.object->operator[](val) = discarded);
+        }
+
+        return true;
+    }
+
+    bool end_object()
+    {
+        if (ref_stack.back())
+        {
+            if (!callback(static_cast<int>(ref_stack.size()) - 1, parse_event_t::object_end, *ref_stack.back()))
+            {
+                // discard object
+                *ref_stack.back() = discarded;
+            }
+            else
+            {
+                ref_stack.back()->set_parents();
+            }
+        }
+
+        JSON_ASSERT(!ref_stack.empty());
+        JSON_ASSERT(!keep_stack.empty());
+        ref_stack.pop_back();
+        keep_stack.pop_back();
+
+        if (!ref_stack.empty() && ref_stack.back() && ref_stack.back()->is_structured())
+        {
+            // remove discarded value
+            for (auto it = ref_stack.back()->begin(); it != ref_stack.back()->end(); ++it)
+            {
+                if (it->is_discarded())
+                {
+                    ref_stack.back()->erase(it);
+                    break;
+                }
+            }
+        }
+
+        return true;
+    }
+
+    bool start_array(std::size_t len)
+    {
+        const bool keep = callback(static_cast<int>(ref_stack.size()), parse_event_t::array_start, discarded);
+        keep_stack.push_back(keep);
+
+        auto val = handle_value(BasicJsonType::value_t::array, true);
+        ref_stack.push_back(val.second);
+
+        // check array limit
+        if (ref_stack.back() && JSON_HEDLEY_UNLIKELY(len != static_cast<std::size_t>(-1) && len > ref_stack.back()->max_size()))
+        {
+            JSON_THROW(out_of_range::create(408, "excessive array size: " + std::to_string(len), *ref_stack.back()));
+        }
+
+        return true;
+    }
+
+    bool end_array()
+    {
+        bool keep = true;
+
+        if (ref_stack.back())
+        {
+            keep = callback(static_cast<int>(ref_stack.size()) - 1, parse_event_t::array_end, *ref_stack.back());
+            if (keep)
+            {
+                ref_stack.back()->set_parents();
+            }
+            else
+            {
+                // discard array
+                *ref_stack.back() = discarded;
+            }
+        }
+
+        JSON_ASSERT(!ref_stack.empty());
+        JSON_ASSERT(!keep_stack.empty());
+        ref_stack.pop_back();
+        keep_stack.pop_back();
+
+        // remove discarded value
+        if (!keep && !ref_stack.empty() && ref_stack.back()->is_array())
+        {
+            ref_stack.back()->m_value.array->pop_back();
+        }
+
+        return true;
+    }
+
+    template<class Exception>
+    bool parse_error(std::size_t /*unused*/, const std::string& /*unused*/,
+                     const Exception& ex)
+    {
+        errored = true;
+        static_cast<void>(ex);
+        if (allow_exceptions)
+        {
+            JSON_THROW(ex);
+        }
+        return false;
+    }
+
+    constexpr bool is_errored() const
+    {
+        return errored;
+    }
+
+  private:
+    /*!
+    @param[in] v  value to add to the JSON value we build during parsing
+    @param[in] skip_callback  whether we should skip calling the callback
+               function; this is required after start_array() and
+               start_object() SAX events, because otherwise we would call the
+               callback function with an empty array or object, respectively.
+
+    @invariant If the ref stack is empty, then the passed value will be the new
+               root.
+    @invariant If the ref stack contains a value, then it is an array or an
+               object to which we can add elements
+
+    @return pair of boolean (whether value should be kept) and pointer (to the
+            passed value in the ref_stack hierarchy; nullptr if not kept)
+    */
+    template<typename Value>
+    std::pair<bool, BasicJsonType*> handle_value(Value&& v, const bool skip_callback = false)
+    {
+        JSON_ASSERT(!keep_stack.empty());
+
+        // do not handle this value if we know it would be added to a discarded
+        // container
+        if (!keep_stack.back())
+        {
+            return {false, nullptr};
+        }
+
+        // create value
+        auto value = BasicJsonType(std::forward<Value>(v));
+
+        // check callback
+        const bool keep = skip_callback || callback(static_cast<int>(ref_stack.size()), parse_event_t::value, value);
+
+        // do not handle this value if we just learnt it shall be discarded
+        if (!keep)
+        {
+            return {false, nullptr};
+        }
+
+        if (ref_stack.empty())
+        {
+            root = std::move(value);
+            return {true, &root};
+        }
+
+        // skip this value if we already decided to skip the parent
+        // (https://github.com/nlohmann/json/issues/971#issuecomment-413678360)
+        if (!ref_stack.back())
+        {
+            return {false, nullptr};
+        }
+
+        // we now only expect arrays and objects
+        JSON_ASSERT(ref_stack.back()->is_array() || ref_stack.back()->is_object());
+
+        // array
+        if (ref_stack.back()->is_array())
+        {
+            ref_stack.back()->m_value.array->emplace_back(std::move(value));
+            return {true, &(ref_stack.back()->m_value.array->back())};
+        }
+
+        // object
+        JSON_ASSERT(ref_stack.back()->is_object());
+        // check if we should store an element for the current key
+        JSON_ASSERT(!key_keep_stack.empty());
+        const bool store_element = key_keep_stack.back();
+        key_keep_stack.pop_back();
+
+        if (!store_element)
+        {
+            return {false, nullptr};
+        }
+
+        JSON_ASSERT(object_element);
+        *object_element = std::move(value);
+        return {true, object_element};
+    }
+
+    /// the parsed JSON value
+    BasicJsonType& root;
+    /// stack to model hierarchy of values
+    std::vector<BasicJsonType*> ref_stack {};
+    /// stack to manage which values to keep
+    std::vector<bool> keep_stack {};
+    /// stack to manage which object keys to keep
+    std::vector<bool> key_keep_stack {};
+    /// helper to hold the reference for the next object element
+    BasicJsonType* object_element = nullptr;
+    /// whether a syntax error occurred
+    bool errored = false;
+    /// callback function
+    const parser_callback_t callback = nullptr;
+    /// whether to throw exceptions in case of errors
+    const bool allow_exceptions = true;
+    /// a discarded value for the callback
+    BasicJsonType discarded = BasicJsonType::value_t::discarded;
+};
+
+template<typename BasicJsonType>
+class json_sax_acceptor
+{
+  public:
+    using number_integer_t = typename BasicJsonType::number_integer_t;
+    using number_unsigned_t = typename BasicJsonType::number_unsigned_t;
+    using number_float_t = typename BasicJsonType::number_float_t;
+    using string_t = typename BasicJsonType::string_t;
+    using binary_t = typename BasicJsonType::binary_t;
+
+    bool null()
+    {
+        return true;
+    }
+
+    bool boolean(bool /*unused*/)
+    {
+        return true;
+    }
+
+    bool number_integer(number_integer_t /*unused*/)
+    {
+        return true;
+    }
+
+    bool number_unsigned(number_unsigned_t /*unused*/)
+    {
+        return true;
+    }
+
+    bool number_float(number_float_t /*unused*/, const string_t& /*unused*/)
+    {
+        return true;
+    }
+
+    bool string(string_t& /*unused*/)
+    {
+        return true;
+    }
+
+    bool binary(binary_t& /*unused*/)
+    {
+        return true;
+    }
+
+    bool start_object(std::size_t /*unused*/ = static_cast<std::size_t>(-1))
+    {
+        return true;
+    }
+
+    bool key(string_t& /*unused*/)
+    {
+        return true;
+    }
+
+    bool end_object()
+    {
+        return true;
+    }
+
+    bool start_array(std::size_t /*unused*/ = static_cast<std::size_t>(-1))
+    {
+        return true;
+    }
+
+    bool end_array()
+    {
+        return true;
+    }
+
+    bool parse_error(std::size_t /*unused*/, const std::string& /*unused*/, const detail::exception& /*unused*/)
+    {
+        return false;
+    }
+};
+}  // namespace detail
+
+}  // namespace nlohmann
 
 // #include <nlohmann/detail/input/lexer.hpp>
 
 
+#include <array> // array
 #include <clocale> // localeconv
 #include <cstddef> // size_t
-#include <cstdlib> // strtof, strtod, strtold, strtoll, strtoull
 #include <cstdio> // snprintf
+#include <cstdlib> // strtof, strtod, strtold, strtoll, strtoull
 #include <initializer_list> // initializer_list
 #include <string> // char_traits, string
+#include <utility> // move
 #include <vector> // vector
 
-// #include <nlohmann/detail/macro_scope.hpp>
-
 // #include <nlohmann/detail/input/input_adapters.hpp>
 
+// #include <nlohmann/detail/input/position_t.hpp>
+
+// #include <nlohmann/detail/macro_scope.hpp>
+
 
 namespace nlohmann
 {
@@ -2278,19 +6432,9 @@ namespace detail
 // lexer //
 ///////////
 
-/*!
-@brief lexical analysis
-
-This class organizes the lexical analysis during JSON deserialization.
-*/
 template<typename BasicJsonType>
-class lexer
+class lexer_base
 {
-    using number_integer_t = typename BasicJsonType::number_integer_t;
-    using number_unsigned_t = typename BasicJsonType::number_unsigned_t;
-    using number_float_t = typename BasicJsonType::number_float_t;
-    using string_t = typename BasicJsonType::string_t;
-
   public:
     /// token types for the parser
     enum class token_type
@@ -2315,6 +6459,8 @@ class lexer
     };
 
     /// return name of values of type token_type (only used for errors)
+    JSON_HEDLEY_RETURNS_NON_NULL
+    JSON_HEDLEY_CONST
     static const char* token_type_name(const token_type t) noexcept
     {
         switch (t)
@@ -2329,9 +6475,9 @@ class lexer
                 return "null literal";
             case token_type::value_string:
                 return "string literal";
-            case lexer::token_type::value_unsigned:
-            case lexer::token_type::value_integer:
-            case lexer::token_type::value_float:
+            case token_type::value_unsigned:
+            case token_type::value_integer:
+            case token_type::value_float:
                 return "number literal";
             case token_type::begin_array:
                 return "'['";
@@ -2357,13 +6503,37 @@ class lexer
                 // LCOV_EXCL_STOP
         }
     }
+};
+/*!
+@brief lexical analysis
 
-    explicit lexer(detail::input_adapter_t&& adapter)
-        : ia(std::move(adapter)), decimal_point_char(get_decimal_point()) {}
+This class organizes the lexical analysis during JSON deserialization.
+*/
+template<typename BasicJsonType, typename InputAdapterType>
+class lexer : public lexer_base<BasicJsonType>
+{
+    using number_integer_t = typename BasicJsonType::number_integer_t;
+    using number_unsigned_t = typename BasicJsonType::number_unsigned_t;
+    using number_float_t = typename BasicJsonType::number_float_t;
+    using string_t = typename BasicJsonType::string_t;
+    using char_type = typename InputAdapterType::char_type;
+    using char_int_type = typename std::char_traits<char_type>::int_type;
+
+  public:
+    using token_type = typename lexer_base<BasicJsonType>::token_type;
+
+    explicit lexer(InputAdapterType&& adapter, bool ignore_comments_ = false) noexcept
+        : ia(std::move(adapter))
+        , ignore_comments(ignore_comments_)
+        , decimal_point_char(static_cast<char_int_type>(get_decimal_point()))
+    {}
 
     // delete because of pointer members
     lexer(const lexer&) = delete;
+    lexer(lexer&&) = default; // NOLINT(hicpp-noexcept-move,performance-noexcept-move-constructor)
     lexer& operator=(lexer&) = delete;
+    lexer& operator=(lexer&&) = default; // NOLINT(hicpp-noexcept-move,performance-noexcept-move-constructor)
+    ~lexer() = default;
 
   private:
     /////////////////////
@@ -2371,10 +6541,11 @@ class lexer
     /////////////////////
 
     /// return the locale-dependent decimal point
+    JSON_HEDLEY_PURE
     static char get_decimal_point() noexcept
     {
-        const auto loc = localeconv();
-        assert(loc != nullptr);
+        const auto* loc = localeconv();
+        JSON_ASSERT(loc != nullptr);
         return (loc->decimal_point == nullptr) ? '.' : *(loc->decimal_point);
     }
 
@@ -2400,25 +6571,25 @@ class lexer
     int get_codepoint()
     {
         // this function only makes sense after reading `\u`
-        assert(current == 'u');
+        JSON_ASSERT(current == 'u');
         int codepoint = 0;
 
-        const auto factors = { 12, 8, 4, 0 };
+        const auto factors = { 12u, 8u, 4u, 0u };
         for (const auto factor : factors)
         {
             get();
 
-            if (current >= '0' and current <= '9')
+            if (current >= '0' && current <= '9')
             {
-                codepoint += ((current - 0x30) << factor);
+                codepoint += static_cast<int>((static_cast<unsigned int>(current) - 0x30u) << factor);
             }
-            else if (current >= 'A' and current <= 'F')
+            else if (current >= 'A' && current <= 'F')
             {
-                codepoint += ((current - 0x37) << factor);
+                codepoint += static_cast<int>((static_cast<unsigned int>(current) - 0x37u) << factor);
             }
-            else if (current >= 'a' and current <= 'f')
+            else if (current >= 'a' && current <= 'f')
             {
-                codepoint += ((current - 0x57) << factor);
+                codepoint += static_cast<int>((static_cast<unsigned int>(current) - 0x57u) << factor);
             }
             else
             {
@@ -2426,7 +6597,7 @@ class lexer
             }
         }
 
-        assert(0x0000 <= codepoint and codepoint <= 0xFFFF);
+        JSON_ASSERT(0x0000 <= codepoint && codepoint <= 0xFFFF);
         return codepoint;
     }
 
@@ -2445,15 +6616,15 @@ class lexer
 
     @return true if and only if no range violation was detected
     */
-    bool next_byte_in_range(std::initializer_list<int> ranges)
+    bool next_byte_in_range(std::initializer_list<char_int_type> ranges)
     {
-        assert(ranges.size() == 2 or ranges.size() == 4 or ranges.size() == 6);
+        JSON_ASSERT(ranges.size() == 2 || ranges.size() == 4 || ranges.size() == 6);
         add(current);
 
         for (auto range = ranges.begin(); range != ranges.end(); ++range)
         {
             get();
-            if (JSON_LIKELY(*range <= current and current <= *(++range)))
+            if (JSON_HEDLEY_LIKELY(*range <= current && current <= *(++range)))
             {
                 add(current);
             }
@@ -2470,7 +6641,7 @@ class lexer
     /*!
     @brief scan a string literal
 
-    This function scans a string according to Sect. 7 of RFC 7159. While
+    This function scans a string according to Sect. 7 of RFC 8259. While
     scanning, bytes are escaped and copied into buffer token_buffer. Then the
     function returns successfully, token_buffer is *not* null-terminated (as it
     may contain \0 bytes), and token_buffer.size() is the number of bytes in the
@@ -2488,7 +6659,7 @@ class lexer
         reset();
 
         // we entered the function by reading an open quote
-        assert(current == '\"');
+        JSON_ASSERT(current == '\"');
 
         while (true)
         {
@@ -2496,7 +6667,7 @@ class lexer
             switch (get())
             {
                 // end of file while parsing string
-                case std::char_traits<char>::eof():
+                case std::char_traits<char_type>::eof():
                 {
                     error_message = "invalid string: missing closing quote";
                     return token_type::parse_error;
@@ -2552,55 +6723,55 @@ class lexer
                             const int codepoint1 = get_codepoint();
                             int codepoint = codepoint1; // start with codepoint1
 
-                            if (JSON_UNLIKELY(codepoint1 == -1))
+                            if (JSON_HEDLEY_UNLIKELY(codepoint1 == -1))
                             {
                                 error_message = "invalid string: '\\u' must be followed by 4 hex digits";
                                 return token_type::parse_error;
                             }
 
                             // check if code point is a high surrogate
-                            if (0xD800 <= codepoint1 and codepoint1 <= 0xDBFF)
+                            if (0xD800 <= codepoint1 && codepoint1 <= 0xDBFF)
                             {
                                 // expect next \uxxxx entry
-                                if (JSON_LIKELY(get() == '\\' and get() == 'u'))
+                                if (JSON_HEDLEY_LIKELY(get() == '\\' && get() == 'u'))
                                 {
                                     const int codepoint2 = get_codepoint();
 
-                                    if (JSON_UNLIKELY(codepoint2 == -1))
+                                    if (JSON_HEDLEY_UNLIKELY(codepoint2 == -1))
                                     {
                                         error_message = "invalid string: '\\u' must be followed by 4 hex digits";
                                         return token_type::parse_error;
                                     }
 
                                     // check if codepoint2 is a low surrogate
-                                    if (JSON_LIKELY(0xDC00 <= codepoint2 and codepoint2 <= 0xDFFF))
+                                    if (JSON_HEDLEY_LIKELY(0xDC00 <= codepoint2 && codepoint2 <= 0xDFFF))
                                     {
                                         // overwrite codepoint
-                                        codepoint =
-                                            // high surrogate occupies the most significant 22 bits
-                                            (codepoint1 << 10)
-                                            // low surrogate occupies the least significant 15 bits
-                                            + codepoint2
-                                            // there is still the 0xD800, 0xDC00 and 0x10000 noise
-                                            // in the result so we have to subtract with:
-                                            // (0xD800 << 10) + DC00 - 0x10000 = 0x35FDC00
-                                            - 0x35FDC00;
+                                        codepoint = static_cast<int>(
+                                                        // high surrogate occupies the most significant 22 bits
+                                                        (static_cast<unsigned int>(codepoint1) << 10u)
+                                                        // low surrogate occupies the least significant 15 bits
+                                                        + static_cast<unsigned int>(codepoint2)
+                                                        // there is still the 0xD800, 0xDC00 and 0x10000 noise
+                                                        // in the result, so we have to subtract with:
+                                                        // (0xD800 << 10) + DC00 - 0x10000 = 0x35FDC00
+                                                        - 0x35FDC00u);
                                     }
                                     else
                                     {
-                                        error_message = "invalid string: surrogate U+DC00..U+DFFF must be followed by U+DC00..U+DFFF";
+                                        error_message = "invalid string: surrogate U+D800..U+DBFF must be followed by U+DC00..U+DFFF";
                                         return token_type::parse_error;
                                     }
                                 }
                                 else
                                 {
-                                    error_message = "invalid string: surrogate U+DC00..U+DFFF must be followed by U+DC00..U+DFFF";
+                                    error_message = "invalid string: surrogate U+D800..U+DBFF must be followed by U+DC00..U+DFFF";
                                     return token_type::parse_error;
                                 }
                             }
                             else
                             {
-                                if (JSON_UNLIKELY(0xDC00 <= codepoint1 and codepoint1 <= 0xDFFF))
+                                if (JSON_HEDLEY_UNLIKELY(0xDC00 <= codepoint1 && codepoint1 <= 0xDFFF))
                                 {
                                     error_message = "invalid string: surrogate U+DC00..U+DFFF must follow U+D800..U+DBFF";
                                     return token_type::parse_error;
@@ -2608,34 +6779,34 @@ class lexer
                             }
 
                             // result of the above calculation yields a proper codepoint
-                            assert(0x00 <= codepoint and codepoint <= 0x10FFFF);
+                            JSON_ASSERT(0x00 <= codepoint && codepoint <= 0x10FFFF);
 
                             // translate codepoint into bytes
                             if (codepoint < 0x80)
                             {
                                 // 1-byte characters: 0xxxxxxx (ASCII)
-                                add(codepoint);
+                                add(static_cast<char_int_type>(codepoint));
                             }
                             else if (codepoint <= 0x7FF)
                             {
                                 // 2-byte characters: 110xxxxx 10xxxxxx
-                                add(0xC0 | (codepoint >> 6));
-                                add(0x80 | (codepoint & 0x3F));
+                                add(static_cast<char_int_type>(0xC0u | (static_cast<unsigned int>(codepoint) >> 6u)));
+                                add(static_cast<char_int_type>(0x80u | (static_cast<unsigned int>(codepoint) & 0x3Fu)));
                             }
                             else if (codepoint <= 0xFFFF)
                             {
                                 // 3-byte characters: 1110xxxx 10xxxxxx 10xxxxxx
-                                add(0xE0 | (codepoint >> 12));
-                                add(0x80 | ((codepoint >> 6) & 0x3F));
-                                add(0x80 | (codepoint & 0x3F));
+                                add(static_cast<char_int_type>(0xE0u | (static_cast<unsigned int>(codepoint) >> 12u)));
+                                add(static_cast<char_int_type>(0x80u | ((static_cast<unsigned int>(codepoint) >> 6u) & 0x3Fu)));
+                                add(static_cast<char_int_type>(0x80u | (static_cast<unsigned int>(codepoint) & 0x3Fu)));
                             }
                             else
                             {
                                 // 4-byte characters: 11110xxx 10xxxxxx 10xxxxxx 10xxxxxx
-                                add(0xF0 | (codepoint >> 18));
-                                add(0x80 | ((codepoint >> 12) & 0x3F));
-                                add(0x80 | ((codepoint >> 6) & 0x3F));
-                                add(0x80 | (codepoint & 0x3F));
+                                add(static_cast<char_int_type>(0xF0u | (static_cast<unsigned int>(codepoint) >> 18u)));
+                                add(static_cast<char_int_type>(0x80u | ((static_cast<unsigned int>(codepoint) >> 12u) & 0x3Fu)));
+                                add(static_cast<char_int_type>(0x80u | ((static_cast<unsigned int>(codepoint) >> 6u) & 0x3Fu)));
+                                add(static_cast<char_int_type>(0x80u | (static_cast<unsigned int>(codepoint) & 0x3Fu)));
                             }
 
                             break;
@@ -2652,39 +6823,194 @@ class lexer
 
                 // invalid control characters
                 case 0x00:
+                {
+                    error_message = "invalid string: control character U+0000 (NUL) must be escaped to \\u0000";
+                    return token_type::parse_error;
+                }
+
                 case 0x01:
+                {
+                    error_message = "invalid string: control character U+0001 (SOH) must be escaped to \\u0001";
+                    return token_type::parse_error;
+                }
+
                 case 0x02:
+                {
+                    error_message = "invalid string: control character U+0002 (STX) must be escaped to \\u0002";
+                    return token_type::parse_error;
+                }
+
                 case 0x03:
+                {
+                    error_message = "invalid string: control character U+0003 (ETX) must be escaped to \\u0003";
+                    return token_type::parse_error;
+                }
+
                 case 0x04:
+                {
+                    error_message = "invalid string: control character U+0004 (EOT) must be escaped to \\u0004";
+                    return token_type::parse_error;
+                }
+
                 case 0x05:
+                {
+                    error_message = "invalid string: control character U+0005 (ENQ) must be escaped to \\u0005";
+                    return token_type::parse_error;
+                }
+
                 case 0x06:
+                {
+                    error_message = "invalid string: control character U+0006 (ACK) must be escaped to \\u0006";
+                    return token_type::parse_error;
+                }
+
                 case 0x07:
+                {
+                    error_message = "invalid string: control character U+0007 (BEL) must be escaped to \\u0007";
+                    return token_type::parse_error;
+                }
+
                 case 0x08:
+                {
+                    error_message = "invalid string: control character U+0008 (BS) must be escaped to \\u0008 or \\b";
+                    return token_type::parse_error;
+                }
+
                 case 0x09:
+                {
+                    error_message = "invalid string: control character U+0009 (HT) must be escaped to \\u0009 or \\t";
+                    return token_type::parse_error;
+                }
+
                 case 0x0A:
+                {
+                    error_message = "invalid string: control character U+000A (LF) must be escaped to \\u000A or \\n";
+                    return token_type::parse_error;
+                }
+
                 case 0x0B:
+                {
+                    error_message = "invalid string: control character U+000B (VT) must be escaped to \\u000B";
+                    return token_type::parse_error;
+                }
+
                 case 0x0C:
+                {
+                    error_message = "invalid string: control character U+000C (FF) must be escaped to \\u000C or \\f";
+                    return token_type::parse_error;
+                }
+
                 case 0x0D:
+                {
+                    error_message = "invalid string: control character U+000D (CR) must be escaped to \\u000D or \\r";
+                    return token_type::parse_error;
+                }
+
                 case 0x0E:
+                {
+                    error_message = "invalid string: control character U+000E (SO) must be escaped to \\u000E";
+                    return token_type::parse_error;
+                }
+
                 case 0x0F:
+                {
+                    error_message = "invalid string: control character U+000F (SI) must be escaped to \\u000F";
+                    return token_type::parse_error;
+                }
+
                 case 0x10:
+                {
+                    error_message = "invalid string: control character U+0010 (DLE) must be escaped to \\u0010";
+                    return token_type::parse_error;
+                }
+
                 case 0x11:
+                {
+                    error_message = "invalid string: control character U+0011 (DC1) must be escaped to \\u0011";
+                    return token_type::parse_error;
+                }
+
                 case 0x12:
+                {
+                    error_message = "invalid string: control character U+0012 (DC2) must be escaped to \\u0012";
+                    return token_type::parse_error;
+                }
+
                 case 0x13:
+                {
+                    error_message = "invalid string: control character U+0013 (DC3) must be escaped to \\u0013";
+                    return token_type::parse_error;
+                }
+
                 case 0x14:
+                {
+                    error_message = "invalid string: control character U+0014 (DC4) must be escaped to \\u0014";
+                    return token_type::parse_error;
+                }
+
                 case 0x15:
+                {
+                    error_message = "invalid string: control character U+0015 (NAK) must be escaped to \\u0015";
+                    return token_type::parse_error;
+                }
+
                 case 0x16:
+                {
+                    error_message = "invalid string: control character U+0016 (SYN) must be escaped to \\u0016";
+                    return token_type::parse_error;
+                }
+
                 case 0x17:
+                {
+                    error_message = "invalid string: control character U+0017 (ETB) must be escaped to \\u0017";
+                    return token_type::parse_error;
+                }
+
                 case 0x18:
+                {
+                    error_message = "invalid string: control character U+0018 (CAN) must be escaped to \\u0018";
+                    return token_type::parse_error;
+                }
+
                 case 0x19:
+                {
+                    error_message = "invalid string: control character U+0019 (EM) must be escaped to \\u0019";
+                    return token_type::parse_error;
+                }
+
                 case 0x1A:
+                {
+                    error_message = "invalid string: control character U+001A (SUB) must be escaped to \\u001A";
+                    return token_type::parse_error;
+                }
+
                 case 0x1B:
+                {
+                    error_message = "invalid string: control character U+001B (ESC) must be escaped to \\u001B";
+                    return token_type::parse_error;
+                }
+
                 case 0x1C:
+                {
+                    error_message = "invalid string: control character U+001C (FS) must be escaped to \\u001C";
+                    return token_type::parse_error;
+                }
+
                 case 0x1D:
+                {
+                    error_message = "invalid string: control character U+001D (GS) must be escaped to \\u001D";
+                    return token_type::parse_error;
+                }
+
                 case 0x1E:
+                {
+                    error_message = "invalid string: control character U+001E (RS) must be escaped to \\u001E";
+                    return token_type::parse_error;
+                }
+
                 case 0x1F:
                 {
-                    error_message = "invalid string: control character must be escaped";
+                    error_message = "invalid string: control character U+001F (US) must be escaped to \\u001F";
                     return token_type::parse_error;
                 }
 
@@ -2820,7 +7146,7 @@ class lexer
                 case 0xDE:
                 case 0xDF:
                 {
-                    if (JSON_UNLIKELY(not next_byte_in_range({0x80, 0xBF})))
+                    if (JSON_HEDLEY_UNLIKELY(!next_byte_in_range({0x80, 0xBF})))
                     {
                         return token_type::parse_error;
                     }
@@ -2830,7 +7156,7 @@ class lexer
                 // U+0800..U+0FFF: bytes E0 A0..BF 80..BF
                 case 0xE0:
                 {
-                    if (JSON_UNLIKELY(not (next_byte_in_range({0xA0, 0xBF, 0x80, 0xBF}))))
+                    if (JSON_HEDLEY_UNLIKELY(!(next_byte_in_range({0xA0, 0xBF, 0x80, 0xBF}))))
                     {
                         return token_type::parse_error;
                     }
@@ -2854,7 +7180,7 @@ class lexer
                 case 0xEE:
                 case 0xEF:
                 {
-                    if (JSON_UNLIKELY(not (next_byte_in_range({0x80, 0xBF, 0x80, 0xBF}))))
+                    if (JSON_HEDLEY_UNLIKELY(!(next_byte_in_range({0x80, 0xBF, 0x80, 0xBF}))))
                     {
                         return token_type::parse_error;
                     }
@@ -2864,7 +7190,7 @@ class lexer
                 // U+D000..U+D7FF: bytes ED 80..9F 80..BF
                 case 0xED:
                 {
-                    if (JSON_UNLIKELY(not (next_byte_in_range({0x80, 0x9F, 0x80, 0xBF}))))
+                    if (JSON_HEDLEY_UNLIKELY(!(next_byte_in_range({0x80, 0x9F, 0x80, 0xBF}))))
                     {
                         return token_type::parse_error;
                     }
@@ -2874,7 +7200,7 @@ class lexer
                 // U+10000..U+3FFFF F0 90..BF 80..BF 80..BF
                 case 0xF0:
                 {
-                    if (JSON_UNLIKELY(not (next_byte_in_range({0x90, 0xBF, 0x80, 0xBF, 0x80, 0xBF}))))
+                    if (JSON_HEDLEY_UNLIKELY(!(next_byte_in_range({0x90, 0xBF, 0x80, 0xBF, 0x80, 0xBF}))))
                     {
                         return token_type::parse_error;
                     }
@@ -2886,7 +7212,7 @@ class lexer
                 case 0xF2:
                 case 0xF3:
                 {
-                    if (JSON_UNLIKELY(not (next_byte_in_range({0x80, 0xBF, 0x80, 0xBF, 0x80, 0xBF}))))
+                    if (JSON_HEDLEY_UNLIKELY(!(next_byte_in_range({0x80, 0xBF, 0x80, 0xBF, 0x80, 0xBF}))))
                     {
                         return token_type::parse_error;
                     }
@@ -2896,7 +7222,7 @@ class lexer
                 // U+100000..U+10FFFF F4 80..8F 80..BF 80..BF
                 case 0xF4:
                 {
-                    if (JSON_UNLIKELY(not (next_byte_in_range({0x80, 0x8F, 0x80, 0xBF, 0x80, 0xBF}))))
+                    if (JSON_HEDLEY_UNLIKELY(!(next_byte_in_range({0x80, 0x8F, 0x80, 0xBF, 0x80, 0xBF}))))
                     {
                         return token_type::parse_error;
                     }
@@ -2913,16 +7239,90 @@ class lexer
         }
     }
 
+    /*!
+     * @brief scan a comment
+     * @return whether comment could be scanned successfully
+     */
+    bool scan_comment()
+    {
+        switch (get())
+        {
+            // single-line comments skip input until a newline or EOF is read
+            case '/':
+            {
+                while (true)
+                {
+                    switch (get())
+                    {
+                        case '\n':
+                        case '\r':
+                        case std::char_traits<char_type>::eof():
+                        case '\0':
+                            return true;
+
+                        default:
+                            break;
+                    }
+                }
+            }
+
+            // multi-line comments skip input until */ is read
+            case '*':
+            {
+                while (true)
+                {
+                    switch (get())
+                    {
+                        case std::char_traits<char_type>::eof():
+                        case '\0':
+                        {
+                            error_message = "invalid comment; missing closing '*/'";
+                            return false;
+                        }
+
+                        case '*':
+                        {
+                            switch (get())
+                            {
+                                case '/':
+                                    return true;
+
+                                default:
+                                {
+                                    unget();
+                                    continue;
+                                }
+                            }
+                        }
+
+                        default:
+                            continue;
+                    }
+                }
+            }
+
+            // unexpected character after reading '/'
+            default:
+            {
+                error_message = "invalid comment; expecting '/' or '*' after '/'";
+                return false;
+            }
+        }
+    }
+
+    JSON_HEDLEY_NON_NULL(2)
     static void strtof(float& f, const char* str, char** endptr) noexcept
     {
         f = std::strtof(str, endptr);
     }
 
+    JSON_HEDLEY_NON_NULL(2)
     static void strtof(double& f, const char* str, char** endptr) noexcept
     {
         f = std::strtod(str, endptr);
     }
 
+    JSON_HEDLEY_NON_NULL(2)
     static void strtof(long double& f, const char* str, char** endptr) noexcept
     {
         f = std::strtold(str, endptr);
@@ -2931,10 +7331,10 @@ class lexer
     /*!
     @brief scan a number literal
 
-    This function scans a string according to Sect. 6 of RFC 7159.
+    This function scans a string according to Sect. 6 of RFC 8259.
 
     The function is realized with a deterministic finite state machine derived
-    from the grammar described in RFC 7159. Starting in state "init", the
+    from the grammar described in RFC 8259. Starting in state "init", the
     input is read and used to determined the next state. Only state "done"
     accepts the number. State "error" is a trap state to model errors. In the
     table below, "anything" means any character but the ones listed before.
@@ -2945,7 +7345,7 @@ class lexer
     minus    | zero     | any1     | [error]  | [error] | [error] | [error]  | [error]
     zero     | done     | done     | exponent | done    | done    | decimal1 | done
     any1     | any1     | any1     | exponent | done    | done    | decimal1 | done
-    decimal1 | decimal2 | [error]  | [error]  | [error] | [error] | [error]  | [error]
+    decimal1 | decimal2 | decimal2 | [error]  | [error] | [error] | [error]  | [error]
     decimal2 | decimal2 | decimal2 | exponent | done    | done    | done     | done
     exponent | any2     | any2     | [error]  | sign    | sign    | [error]  | [error]
     sign     | any2     | any2     | [error]  | [error] | [error] | [error]  | [error]
@@ -2968,7 +7368,7 @@ class lexer
           locale's decimal point is used instead of `.` to work with the
           locale-dependent converters.
     */
-    token_type scan_number()
+    token_type scan_number()  // lgtm [cpp/use-of-goto]
     {
         // reset token_buffer to store the number's bytes
         reset();
@@ -3006,13 +7406,9 @@ class lexer
                 goto scan_number_any1;
             }
 
-            // LCOV_EXCL_START
-            default:
-            {
-                // all other characters are rejected outside scan_number()
-                assert(false);
-            }
-                // LCOV_EXCL_STOP
+            // all other characters are rejected outside scan_number()
+            default:            // LCOV_EXCL_LINE
+                JSON_ASSERT(false); // NOLINT(cert-dcl03-c,hicpp-static-assert,misc-static-assert) LCOV_EXCL_LINE
         }
 
 scan_number_minus:
@@ -3250,7 +7646,7 @@ scan_number_done:
         // we are done scanning a number)
         unget();
 
-        char* endptr = nullptr;
+        char* endptr = nullptr; // NOLINT(cppcoreguidelines-pro-type-vararg,hicpp-vararg)
         errno = 0;
 
         // try to parse integers first and fall back to floats
@@ -3259,7 +7655,7 @@ scan_number_done:
             const auto x = std::strtoull(token_buffer.data(), &endptr, 10);
 
             // we checked the number format before
-            assert(endptr == token_buffer.data() + token_buffer.size());
+            JSON_ASSERT(endptr == token_buffer.data() + token_buffer.size());
 
             if (errno == 0)
             {
@@ -3275,7 +7671,7 @@ scan_number_done:
             const auto x = std::strtoll(token_buffer.data(), &endptr, 10);
 
             // we checked the number format before
-            assert(endptr == token_buffer.data() + token_buffer.size());
+            JSON_ASSERT(endptr == token_buffer.data() + token_buffer.size());
 
             if (errno == 0)
             {
@@ -3292,7 +7688,7 @@ scan_number_done:
         strtof(value_float, token_buffer.data(), &endptr);
 
         // we checked the number format before
-        assert(endptr == token_buffer.data() + token_buffer.size());
+        JSON_ASSERT(endptr == token_buffer.data() + token_buffer.size());
 
         return token_type::value_float;
     }
@@ -3302,13 +7698,14 @@ scan_number_done:
     @param[in] length        the length of the passed literal text
     @param[in] return_type   the token type to return on success
     */
-    token_type scan_literal(const char* literal_text, const std::size_t length,
+    JSON_HEDLEY_NON_NULL(2)
+    token_type scan_literal(const char_type* literal_text, const std::size_t length,
                             token_type return_type)
     {
-        assert(current == literal_text[0]);
+        JSON_ASSERT(std::char_traits<char_type>::to_char_type(current) == literal_text[0]);
         for (std::size_t i = 1; i < length; ++i)
         {
-            if (JSON_UNLIKELY(get() != literal_text[i]))
+            if (JSON_HEDLEY_UNLIKELY(std::char_traits<char_type>::to_char_type(get()) != literal_text[i]))
             {
                 error_message = "invalid literal";
                 return token_type::parse_error;
@@ -3326,7 +7723,7 @@ scan_number_done:
     {
         token_buffer.clear();
         token_string.clear();
-        token_string.push_back(std::char_traits<char>::to_char_type(current));
+        token_string.push_back(std::char_traits<char_type>::to_char_type(current));
     }
 
     /*
@@ -3339,9 +7736,11 @@ scan_number_done:
 
     @return character read from the input
     */
-    std::char_traits<char>::int_type get()
+    char_int_type get()
     {
-        ++chars_read;
+        ++position.chars_read_total;
+        ++position.chars_read_current_line;
+
         if (next_unget)
         {
             // just reset the next_unget variable and work with current
@@ -3349,13 +7748,20 @@ scan_number_done:
         }
         else
         {
-            current = ia->get_character();
+            current = ia.get_character();
         }
 
-        if (JSON_LIKELY(current != std::char_traits<char>::eof()))
+        if (JSON_HEDLEY_LIKELY(current != std::char_traits<char_type>::eof()))
         {
-            token_string.push_back(std::char_traits<char>::to_char_type(current));
+            token_string.push_back(std::char_traits<char_type>::to_char_type(current));
         }
+
+        if (current == '\n')
+        {
+            ++position.lines_read;
+            position.chars_read_current_line = 0;
+        }
+
         return current;
     }
 
@@ -3363,25 +7769,40 @@ scan_number_done:
     @brief unget current character (read it again on next get)
 
     We implement unget by setting variable next_unget to true. The input is not
-    changed - we just simulate ungetting by modifying chars_read and
-    token_string. The next call to get() will behave as if the unget character
-    is read again.
+    changed - we just simulate ungetting by modifying chars_read_total,
+    chars_read_current_line, and token_string. The next call to get() will
+    behave as if the unget character is read again.
     */
     void unget()
     {
         next_unget = true;
-        --chars_read;
-        if (JSON_LIKELY(current != std::char_traits<char>::eof()))
+
+        --position.chars_read_total;
+
+        // in case we "unget" a newline, we have to also decrement the lines_read
+        if (position.chars_read_current_line == 0)
         {
-            assert(token_string.size() != 0);
+            if (position.lines_read > 0)
+            {
+                --position.lines_read;
+            }
+        }
+        else
+        {
+            --position.chars_read_current_line;
+        }
+
+        if (JSON_HEDLEY_LIKELY(current != std::char_traits<char_type>::eof()))
+        {
+            JSON_ASSERT(!token_string.empty());
             token_string.pop_back();
         }
     }
 
     /// add a character to token_buffer
-    void add(int c)
+    void add(char_int_type c)
     {
-        token_buffer.push_back(std::char_traits<char>::to_char_type(c));
+        token_buffer.push_back(static_cast<typename string_t::value_type>(c));
     }
 
   public:
@@ -3418,9 +7839,9 @@ scan_number_done:
     /////////////////////
 
     /// return position of last read token
-    constexpr std::size_t get_position() const noexcept
+    constexpr position_t get_position() const noexcept
     {
-        return chars_read;
+        return position;
     }
 
     /// return the last read token (for errors only).  Will never contain EOF
@@ -3432,17 +7853,17 @@ scan_number_done:
         std::string result;
         for (const auto c : token_string)
         {
-            if ('\x00' <= c and c <= '\x1F')
+            if (static_cast<unsigned char>(c) <= '\x1F')
             {
                 // escape control characters
-                char cs[9];
-                snprintf(cs, 9, "<U+%.4X>", static_cast<unsigned char>(c));
-                result += cs;
+                std::array<char, 9> cs{{}};
+                static_cast<void>((std::snprintf)(cs.data(), cs.size(), "<U+%.4X>", static_cast<unsigned char>(c))); // NOLINT(cppcoreguidelines-pro-type-vararg,hicpp-vararg)
+                result += cs.data();
             }
             else
             {
                 // add character as is
-                result.push_back(c);
+                result.push_back(static_cast<std::string::value_type>(c));
             }
         }
 
@@ -3450,6 +7871,7 @@ scan_number_done:
     }
 
     /// return syntax error message
+    JSON_HEDLEY_RETURNS_NON_NULL
     constexpr const char* get_error_message() const noexcept
     {
         return error_message;
@@ -3467,41 +7889,48 @@ scan_number_done:
     {
         if (get() == 0xEF)
         {
-            if (get() == 0xBB and get() == 0xBF)
-            {
-                // we completely parsed the BOM
-                return true;
-            }
-            else
-            {
-                // after reading 0xEF, an unexpected character followed
-                return false;
-            }
+            // check if we completely parse the BOM
+            return get() == 0xBB && get() == 0xBF;
         }
-        else
+
+        // the first character is not the beginning of the BOM; unget it to
+        // process is later
+        unget();
+        return true;
+    }
+
+    void skip_whitespace()
+    {
+        do
         {
-            // the first character is not the beginning of the BOM; unget it to
-            // process is later
-            unget();
-            return true;
+            get();
         }
+        while (current == ' ' || current == '\t' || current == '\n' || current == '\r');
     }
 
     token_type scan()
     {
         // initially, skip the BOM
-        if (chars_read == 0 and not skip_bom())
+        if (position.chars_read_total == 0 && !skip_bom())
         {
             error_message = "invalid BOM; must be 0xEF 0xBB 0xBF if given";
             return token_type::parse_error;
         }
 
         // read next character and ignore whitespace
-        do
+        skip_whitespace();
+
+        // ignore comments
+        while (ignore_comments && current == '/')
         {
-            get();
+            if (!scan_comment())
+            {
+                return token_type::parse_error;
+            }
+
+            // skip following whitespace
+            skip_whitespace();
         }
-        while (current == ' ' or current == '\t' or current == '\n' or current == '\r');
 
         switch (current)
         {
@@ -3521,11 +7950,20 @@ scan_number_done:
 
             // literals
             case 't':
-                return scan_literal("true", 4, token_type::literal_true);
+            {
+                std::array<char_type, 4> true_literal = {{static_cast<char_type>('t'), static_cast<char_type>('r'), static_cast<char_type>('u'), static_cast<char_type>('e')}};
+                return scan_literal(true_literal.data(), true_literal.size(), token_type::literal_true);
+            }
             case 'f':
-                return scan_literal("false", 5, token_type::literal_false);
+            {
+                std::array<char_type, 5> false_literal = {{static_cast<char_type>('f'), static_cast<char_type>('a'), static_cast<char_type>('l'), static_cast<char_type>('s'), static_cast<char_type>('e')}};
+                return scan_literal(false_literal.data(), false_literal.size(), token_type::literal_false);
+            }
             case 'n':
-                return scan_literal("null", 4, token_type::literal_null);
+            {
+                std::array<char_type, 4> null_literal = {{static_cast<char_type>('n'), static_cast<char_type>('u'), static_cast<char_type>('l'), static_cast<char_type>('l')}};
+                return scan_literal(null_literal.data(), null_literal.size(), token_type::literal_null);
+            }
 
             // string
             case '\"':
@@ -3548,7 +7986,7 @@ scan_number_done:
             // end of input (the null byte is needed when parsing from
             // string literals)
             case '\0':
-            case std::char_traits<char>::eof():
+            case std::char_traits<char_type>::eof():
                 return token_type::end_of_input;
 
             // error
@@ -3560,19 +7998,22 @@ scan_number_done:
 
   private:
     /// input adapter
-    detail::input_adapter_t ia = nullptr;
+    InputAdapterType ia;
+
+    /// whether comments should be ignored (true) or signaled as errors (false)
+    const bool ignore_comments = false;
 
     /// the current character
-    std::char_traits<char>::int_type current = std::char_traits<char>::eof();
+    char_int_type current = std::char_traits<char_type>::eof();
 
     /// whether the next get() call should just return current
     bool next_unget = false;
 
-    /// the number of characters read
-    std::size_t chars_read = 0;
+    /// the start position of the current token
+    position_t position {};
 
     /// raw input token string (for error messages)
-    std::vector<char> token_string {};
+    std::vector<char_type> token_string {};
 
     /// buffer for variable-length tokens (numbers, strings)
     string_t token_buffer {};
@@ -3586,22 +8027,10 @@ scan_number_done:
     number_float_t value_float = 0;
 
     /// the decimal point
-    const char decimal_point_char = '.';
+    const char_int_type decimal_point_char = '.';
 };
-}
-}
-
-// #include <nlohmann/detail/input/parser.hpp>
-
-
-#include <cassert> // assert
-#include <cmath> // isfinite
-#include <cstdint> // uint8_t
-#include <functional> // function
-#include <string> // string
-#include <utility> // move
-
-// #include <nlohmann/detail/exceptions.hpp>
+}  // namespace detail
+}  // namespace nlohmann
 
 // #include <nlohmann/detail/macro_scope.hpp>
 
@@ -3610,6 +8039,7 @@ scan_number_done:
 
 #include <cstdint> // size_t
 #include <utility> // declval
+#include <string> // string
 
 // #include <nlohmann/detail/meta/detected.hpp>
 
@@ -3620,53 +8050,57 @@ namespace nlohmann
 {
 namespace detail
 {
-template <typename T>
+template<typename T>
 using null_function_t = decltype(std::declval<T&>().null());
 
-template <typename T>
+template<typename T>
 using boolean_function_t =
     decltype(std::declval<T&>().boolean(std::declval<bool>()));
 
-template <typename T, typename Integer>
+template<typename T, typename Integer>
 using number_integer_function_t =
     decltype(std::declval<T&>().number_integer(std::declval<Integer>()));
 
-template <typename T, typename Unsigned>
+template<typename T, typename Unsigned>
 using number_unsigned_function_t =
     decltype(std::declval<T&>().number_unsigned(std::declval<Unsigned>()));
 
-template <typename T, typename Float, typename String>
+template<typename T, typename Float, typename String>
 using number_float_function_t = decltype(std::declval<T&>().number_float(
                                     std::declval<Float>(), std::declval<const String&>()));
 
-template <typename T, typename String>
+template<typename T, typename String>
 using string_function_t =
     decltype(std::declval<T&>().string(std::declval<String&>()));
 
-template <typename T>
+template<typename T, typename Binary>
+using binary_function_t =
+    decltype(std::declval<T&>().binary(std::declval<Binary&>()));
+
+template<typename T>
 using start_object_function_t =
     decltype(std::declval<T&>().start_object(std::declval<std::size_t>()));
 
-template <typename T, typename String>
+template<typename T, typename String>
 using key_function_t =
     decltype(std::declval<T&>().key(std::declval<String&>()));
 
-template <typename T>
+template<typename T>
 using end_object_function_t = decltype(std::declval<T&>().end_object());
 
-template <typename T>
+template<typename T>
 using start_array_function_t =
     decltype(std::declval<T&>().start_array(std::declval<std::size_t>()));
 
-template <typename T>
+template<typename T>
 using end_array_function_t = decltype(std::declval<T&>().end_array());
 
-template <typename T, typename Exception>
+template<typename T, typename Exception>
 using parse_error_function_t = decltype(std::declval<T&>().parse_error(
         std::declval<std::size_t>(), std::declval<const std::string&>(),
         std::declval<const Exception&>()));
 
-template <typename SAX, typename BasicJsonType>
+template<typename SAX, typename BasicJsonType>
 struct is_sax
 {
   private:
@@ -3677,19 +8111,18 @@ struct is_sax
     using number_unsigned_t = typename BasicJsonType::number_unsigned_t;
     using number_float_t = typename BasicJsonType::number_float_t;
     using string_t = typename BasicJsonType::string_t;
+    using binary_t = typename BasicJsonType::binary_t;
     using exception_t = typename BasicJsonType::exception;
 
   public:
     static constexpr bool value =
         is_detected_exact<bool, null_function_t, SAX>::value &&
         is_detected_exact<bool, boolean_function_t, SAX>::value &&
-        is_detected_exact<bool, number_integer_function_t, SAX,
-        number_integer_t>::value &&
-        is_detected_exact<bool, number_unsigned_function_t, SAX,
-        number_unsigned_t>::value &&
-        is_detected_exact<bool, number_float_function_t, SAX, number_float_t,
-        string_t>::value &&
+        is_detected_exact<bool, number_integer_function_t, SAX, number_integer_t>::value &&
+        is_detected_exact<bool, number_unsigned_function_t, SAX, number_unsigned_t>::value &&
+        is_detected_exact<bool, number_float_function_t, SAX, number_float_t, string_t>::value &&
         is_detected_exact<bool, string_function_t, SAX, string_t>::value &&
+        is_detected_exact<bool, binary_function_t, SAX, binary_t>::value &&
         is_detected_exact<bool, start_object_function_t, SAX>::value &&
         is_detected_exact<bool, key_function_t, SAX, string_t>::value &&
         is_detected_exact<bool, end_object_function_t, SAX>::value &&
@@ -3698,7 +8131,7 @@ struct is_sax
         is_detected_exact<bool, parse_error_function_t, SAX, exception_t>::value;
 };
 
-template <typename SAX, typename BasicJsonType>
+template<typename SAX, typename BasicJsonType>
 struct is_sax_static_asserts
 {
   private:
@@ -3709,6 +8142,7 @@ struct is_sax_static_asserts
     using number_unsigned_t = typename BasicJsonType::number_unsigned_t;
     using number_float_t = typename BasicJsonType::number_float_t;
     using string_t = typename BasicJsonType::string_t;
+    using binary_t = typename BasicJsonType::binary_t;
     using exception_t = typename BasicJsonType::exception;
 
   public:
@@ -3732,6 +8166,9 @@ struct is_sax_static_asserts
     static_assert(
         is_detected_exact<bool, string_function_t, SAX, string_t>::value,
         "Missing/invalid function: bool string(string_t&)");
+    static_assert(
+        is_detected_exact<bool, binary_function_t, SAX, binary_t>::value,
+        "Missing/invalid function: bool binary(binary_t&)");
     static_assert(is_detected_exact<bool, start_object_function_t, SAX>::value,
                   "Missing/invalid function: bool start_object(std::size_t)");
     static_assert(is_detected_exact<bool, key_function_t, SAX, string_t>::value,
@@ -3747,718 +8184,10 @@ struct is_sax_static_asserts
         "Missing/invalid function: bool parse_error(std::size_t, const "
         "std::string&, const exception&)");
 };
-}
-}
+}  // namespace detail
+}  // namespace nlohmann
 
-// #include <nlohmann/detail/input/input_adapters.hpp>
-
-// #include <nlohmann/detail/input/json_sax.hpp>
-
-
-#include <cstddef>
-#include <string>
-#include <vector>
-
-// #include <nlohmann/detail/input/parser.hpp>
-
-// #include <nlohmann/detail/exceptions.hpp>
-
-
-namespace nlohmann
-{
-
-/*!
-@brief SAX interface
-
-This class describes the SAX interface used by @ref nlohmann::json::sax_parse.
-Each function is called in different situations while the input is parsed. The
-boolean return value informs the parser whether to continue processing the
-input.
-*/
-template<typename BasicJsonType>
-struct json_sax
-{
-    /// type for (signed) integers
-    using number_integer_t = typename BasicJsonType::number_integer_t;
-    /// type for unsigned integers
-    using number_unsigned_t = typename BasicJsonType::number_unsigned_t;
-    /// type for floating-point numbers
-    using number_float_t = typename BasicJsonType::number_float_t;
-    /// type for strings
-    using string_t = typename BasicJsonType::string_t;
-
-    /*!
-    @brief a null value was read
-    @return whether parsing should proceed
-    */
-    virtual bool null() = 0;
-
-    /*!
-    @brief a boolean value was read
-    @param[in] val  boolean value
-    @return whether parsing should proceed
-    */
-    virtual bool boolean(bool val) = 0;
-
-    /*!
-    @brief an integer number was read
-    @param[in] val  integer value
-    @return whether parsing should proceed
-    */
-    virtual bool number_integer(number_integer_t val) = 0;
-
-    /*!
-    @brief an unsigned integer number was read
-    @param[in] val  unsigned integer value
-    @return whether parsing should proceed
-    */
-    virtual bool number_unsigned(number_unsigned_t val) = 0;
-
-    /*!
-    @brief an floating-point number was read
-    @param[in] val  floating-point value
-    @param[in] s    raw token value
-    @return whether parsing should proceed
-    */
-    virtual bool number_float(number_float_t val, const string_t& s) = 0;
-
-    /*!
-    @brief a string was read
-    @param[in] val  string value
-    @return whether parsing should proceed
-    @note It is safe to move the passed string.
-    */
-    virtual bool string(string_t& val) = 0;
-
-    /*!
-    @brief the beginning of an object was read
-    @param[in] elements  number of object elements or -1 if unknown
-    @return whether parsing should proceed
-    @note binary formats may report the number of elements
-    */
-    virtual bool start_object(std::size_t elements) = 0;
-
-    /*!
-    @brief an object key was read
-    @param[in] val  object key
-    @return whether parsing should proceed
-    @note It is safe to move the passed string.
-    */
-    virtual bool key(string_t& val) = 0;
-
-    /*!
-    @brief the end of an object was read
-    @return whether parsing should proceed
-    */
-    virtual bool end_object() = 0;
-
-    /*!
-    @brief the beginning of an array was read
-    @param[in] elements  number of array elements or -1 if unknown
-    @return whether parsing should proceed
-    @note binary formats may report the number of elements
-    */
-    virtual bool start_array(std::size_t elements) = 0;
-
-    /*!
-    @brief the end of an array was read
-    @return whether parsing should proceed
-    */
-    virtual bool end_array() = 0;
-
-    /*!
-    @brief a parse error occurred
-    @param[in] position    the position in the input where the error occurs
-    @param[in] last_token  the last read token
-    @param[in] error_msg   a detailed error message
-    @return whether parsing should proceed (must return false)
-    */
-    virtual bool parse_error(std::size_t position,
-                             const std::string& last_token,
-                             const detail::exception& ex) = 0;
-
-    virtual ~json_sax() = default;
-};
-
-
-namespace detail
-{
-/*!
-@brief SAX implementation to create a JSON value from SAX events
-
-This class implements the @ref json_sax interface and processes the SAX events
-to create a JSON value which makes it basically a DOM parser. The structure or
-hierarchy of the JSON value is managed by the stack `ref_stack` which contains
-a pointer to the respective array or object for each recursion depth.
-
-After successful parsing, the value that is passed by reference to the
-constructor contains the parsed value.
-
-@tparam BasicJsonType  the JSON type
-*/
-template<typename BasicJsonType>
-class json_sax_dom_parser
-{
-  public:
-    using number_integer_t = typename BasicJsonType::number_integer_t;
-    using number_unsigned_t = typename BasicJsonType::number_unsigned_t;
-    using number_float_t = typename BasicJsonType::number_float_t;
-    using string_t = typename BasicJsonType::string_t;
-
-    /*!
-    @param[in, out] r  reference to a JSON value that is manipulated while
-                       parsing
-    @param[in] allow_exceptions_  whether parse errors yield exceptions
-    */
-    explicit json_sax_dom_parser(BasicJsonType& r, const bool allow_exceptions_ = true)
-        : root(r), allow_exceptions(allow_exceptions_)
-    {}
-
-    bool null()
-    {
-        handle_value(nullptr);
-        return true;
-    }
-
-    bool boolean(bool val)
-    {
-        handle_value(val);
-        return true;
-    }
-
-    bool number_integer(number_integer_t val)
-    {
-        handle_value(val);
-        return true;
-    }
-
-    bool number_unsigned(number_unsigned_t val)
-    {
-        handle_value(val);
-        return true;
-    }
-
-    bool number_float(number_float_t val, const string_t&)
-    {
-        handle_value(val);
-        return true;
-    }
-
-    bool string(string_t& val)
-    {
-        handle_value(val);
-        return true;
-    }
-
-    bool start_object(std::size_t len)
-    {
-        ref_stack.push_back(handle_value(BasicJsonType::value_t::object));
-
-        if (JSON_UNLIKELY(len != std::size_t(-1) and len > ref_stack.back()->max_size()))
-        {
-            JSON_THROW(out_of_range::create(408,
-                                            "excessive object size: " + std::to_string(len)));
-        }
-
-        return true;
-    }
-
-    bool key(string_t& val)
-    {
-        // add null at given key and store the reference for later
-        object_element = &(ref_stack.back()->m_value.object->operator[](val));
-        return true;
-    }
-
-    bool end_object()
-    {
-        ref_stack.pop_back();
-        return true;
-    }
-
-    bool start_array(std::size_t len)
-    {
-        ref_stack.push_back(handle_value(BasicJsonType::value_t::array));
-
-        if (JSON_UNLIKELY(len != std::size_t(-1) and len > ref_stack.back()->max_size()))
-        {
-            JSON_THROW(out_of_range::create(408,
-                                            "excessive array size: " + std::to_string(len)));
-        }
-
-        return true;
-    }
-
-    bool end_array()
-    {
-        ref_stack.pop_back();
-        return true;
-    }
-
-    bool parse_error(std::size_t, const std::string&,
-                     const detail::exception& ex)
-    {
-        errored = true;
-        if (allow_exceptions)
-        {
-            // determine the proper exception type from the id
-            switch ((ex.id / 100) % 100)
-            {
-                case 1:
-                    JSON_THROW(*reinterpret_cast<const detail::parse_error*>(&ex));
-                case 4:
-                    JSON_THROW(*reinterpret_cast<const detail::out_of_range*>(&ex));
-                // LCOV_EXCL_START
-                case 2:
-                    JSON_THROW(*reinterpret_cast<const detail::invalid_iterator*>(&ex));
-                case 3:
-                    JSON_THROW(*reinterpret_cast<const detail::type_error*>(&ex));
-                case 5:
-                    JSON_THROW(*reinterpret_cast<const detail::other_error*>(&ex));
-                default:
-                    assert(false);
-                    // LCOV_EXCL_STOP
-            }
-        }
-        return false;
-    }
-
-    constexpr bool is_errored() const
-    {
-        return errored;
-    }
-
-  private:
-    /*!
-    @invariant If the ref stack is empty, then the passed value will be the new
-               root.
-    @invariant If the ref stack contains a value, then it is an array or an
-               object to which we can add elements
-    */
-    template<typename Value>
-    BasicJsonType* handle_value(Value&& v)
-    {
-        if (ref_stack.empty())
-        {
-            root = BasicJsonType(std::forward<Value>(v));
-            return &root;
-        }
-        else
-        {
-            assert(ref_stack.back()->is_array() or ref_stack.back()->is_object());
-            if (ref_stack.back()->is_array())
-            {
-                ref_stack.back()->m_value.array->emplace_back(std::forward<Value>(v));
-                return &(ref_stack.back()->m_value.array->back());
-            }
-            else
-            {
-                assert(object_element);
-                *object_element = BasicJsonType(std::forward<Value>(v));
-                return object_element;
-            }
-        }
-    }
-
-    /// the parsed JSON value
-    BasicJsonType& root;
-    /// stack to model hierarchy of values
-    std::vector<BasicJsonType*> ref_stack;
-    /// helper to hold the reference for the next object element
-    BasicJsonType* object_element = nullptr;
-    /// whether a syntax error occurred
-    bool errored = false;
-    /// whether to throw exceptions in case of errors
-    const bool allow_exceptions = true;
-};
-
-template<typename BasicJsonType>
-class json_sax_dom_callback_parser
-{
-  public:
-    using number_integer_t = typename BasicJsonType::number_integer_t;
-    using number_unsigned_t = typename BasicJsonType::number_unsigned_t;
-    using number_float_t = typename BasicJsonType::number_float_t;
-    using string_t = typename BasicJsonType::string_t;
-    using parser_callback_t = typename BasicJsonType::parser_callback_t;
-    using parse_event_t = typename BasicJsonType::parse_event_t;
-
-    json_sax_dom_callback_parser(BasicJsonType& r,
-                                 const parser_callback_t cb,
-                                 const bool allow_exceptions_ = true)
-        : root(r), callback(cb), allow_exceptions(allow_exceptions_)
-    {
-        keep_stack.push_back(true);
-    }
-
-    bool null()
-    {
-        handle_value(nullptr);
-        return true;
-    }
-
-    bool boolean(bool val)
-    {
-        handle_value(val);
-        return true;
-    }
-
-    bool number_integer(number_integer_t val)
-    {
-        handle_value(val);
-        return true;
-    }
-
-    bool number_unsigned(number_unsigned_t val)
-    {
-        handle_value(val);
-        return true;
-    }
-
-    bool number_float(number_float_t val, const string_t&)
-    {
-        handle_value(val);
-        return true;
-    }
-
-    bool string(string_t& val)
-    {
-        handle_value(val);
-        return true;
-    }
-
-    bool start_object(std::size_t len)
-    {
-        // check callback for object start
-        const bool keep = callback(static_cast<int>(ref_stack.size()), parse_event_t::object_start, discarded);
-        keep_stack.push_back(keep);
-
-        auto val = handle_value(BasicJsonType::value_t::object, true);
-        ref_stack.push_back(val.second);
-
-        // check object limit
-        if (ref_stack.back())
-        {
-            if (JSON_UNLIKELY(len != std::size_t(-1) and len > ref_stack.back()->max_size()))
-            {
-                JSON_THROW(out_of_range::create(408,
-                                                "excessive object size: " + std::to_string(len)));
-            }
-        }
-
-        return true;
-    }
-
-    bool key(string_t& val)
-    {
-        BasicJsonType k = BasicJsonType(val);
-
-        // check callback for key
-        const bool keep = callback(static_cast<int>(ref_stack.size()), parse_event_t::key, k);
-        key_keep_stack.push_back(keep);
-
-        // add discarded value at given key and store the reference for later
-        if (keep and ref_stack.back())
-        {
-            object_element = &(ref_stack.back()->m_value.object->operator[](val) = discarded);
-        }
-
-        return true;
-    }
-
-    bool end_object()
-    {
-        if (ref_stack.back())
-        {
-            if (not callback(static_cast<int>(ref_stack.size()) - 1, parse_event_t::object_end, *ref_stack.back()))
-            {
-                // discard object
-                *ref_stack.back() = discarded;
-            }
-        }
-
-        assert(not ref_stack.empty());
-        assert(not keep_stack.empty());
-        ref_stack.pop_back();
-        keep_stack.pop_back();
-
-        if (not ref_stack.empty() and ref_stack.back())
-        {
-            // remove discarded value
-            if (ref_stack.back()->is_object())
-            {
-                for (auto it = ref_stack.back()->begin(); it != ref_stack.back()->end(); ++it)
-                {
-                    if (it->is_discarded())
-                    {
-                        ref_stack.back()->erase(it);
-                        break;
-                    }
-                }
-            }
-        }
-
-        return true;
-    }
-
-    bool start_array(std::size_t len)
-    {
-        const bool keep = callback(static_cast<int>(ref_stack.size()), parse_event_t::array_start, discarded);
-        keep_stack.push_back(keep);
-
-        auto val = handle_value(BasicJsonType::value_t::array, true);
-        ref_stack.push_back(val.second);
-
-        // check array limit
-        if (ref_stack.back())
-        {
-            if (JSON_UNLIKELY(len != std::size_t(-1) and len > ref_stack.back()->max_size()))
-            {
-                JSON_THROW(out_of_range::create(408,
-                                                "excessive array size: " + std::to_string(len)));
-            }
-        }
-
-        return true;
-    }
-
-    bool end_array()
-    {
-        bool keep = true;
-
-        if (ref_stack.back())
-        {
-            keep = callback(static_cast<int>(ref_stack.size()) - 1, parse_event_t::array_end, *ref_stack.back());
-            if (not keep)
-            {
-                // discard array
-                *ref_stack.back() = discarded;
-            }
-        }
-
-        assert(not ref_stack.empty());
-        assert(not keep_stack.empty());
-        ref_stack.pop_back();
-        keep_stack.pop_back();
-
-        // remove discarded value
-        if (not keep and not ref_stack.empty())
-        {
-            if (ref_stack.back()->is_array())
-            {
-                ref_stack.back()->m_value.array->pop_back();
-            }
-        }
-
-        return true;
-    }
-
-    bool parse_error(std::size_t, const std::string&,
-                     const detail::exception& ex)
-    {
-        errored = true;
-        if (allow_exceptions)
-        {
-            // determine the proper exception type from the id
-            switch ((ex.id / 100) % 100)
-            {
-                case 1:
-                    JSON_THROW(*reinterpret_cast<const detail::parse_error*>(&ex));
-                case 4:
-                    JSON_THROW(*reinterpret_cast<const detail::out_of_range*>(&ex));
-                // LCOV_EXCL_START
-                case 2:
-                    JSON_THROW(*reinterpret_cast<const detail::invalid_iterator*>(&ex));
-                case 3:
-                    JSON_THROW(*reinterpret_cast<const detail::type_error*>(&ex));
-                case 5:
-                    JSON_THROW(*reinterpret_cast<const detail::other_error*>(&ex));
-                default:
-                    assert(false);
-                    // LCOV_EXCL_STOP
-            }
-        }
-        return false;
-    }
-
-    constexpr bool is_errored() const
-    {
-        return errored;
-    }
-
-  private:
-    /*!
-    @param[in] v  value to add to the JSON value we build during parsing
-    @param[in] skip_callback  whether we should skip calling the callback
-               function; this is required after start_array() and
-               start_object() SAX events, because otherwise we would call the
-               callback function with an empty array or object, respectively.
-
-    @invariant If the ref stack is empty, then the passed value will be the new
-               root.
-    @invariant If the ref stack contains a value, then it is an array or an
-               object to which we can add elements
-
-    @return pair of boolean (whether value should be kept) and pointer (to the
-            passed value in the ref_stack hierarchy; nullptr if not kept)
-    */
-    template<typename Value>
-    std::pair<bool, BasicJsonType*> handle_value(Value&& v, const bool skip_callback = false)
-    {
-        assert(not keep_stack.empty());
-
-        // do not handle this value if we know it would be added to a discarded
-        // container
-        if (not keep_stack.back())
-        {
-            return {false, nullptr};
-        }
-
-        // create value
-        auto value = BasicJsonType(std::forward<Value>(v));
-
-        // check callback
-        const bool keep = skip_callback or callback(static_cast<int>(ref_stack.size()), parse_event_t::value, value);
-
-        // do not handle this value if we just learnt it shall be discarded
-        if (not keep)
-        {
-            return {false, nullptr};
-        }
-
-        if (ref_stack.empty())
-        {
-            root = std::move(value);
-            return {true, &root};
-        }
-        else
-        {
-            // skip this value if we already decided to skip the parent
-            // (https://github.com/nlohmann/json/issues/971#issuecomment-413678360)
-            if (not ref_stack.back())
-            {
-                return {false, nullptr};
-            }
-
-            assert(ref_stack.back()->is_array() or ref_stack.back()->is_object());
-            if (ref_stack.back()->is_array())
-            {
-                ref_stack.back()->m_value.array->push_back(std::move(value));
-                return {true, &(ref_stack.back()->m_value.array->back())};
-            }
-            else
-            {
-                // check if we should store an element for the current key
-                assert(not key_keep_stack.empty());
-                const bool store_element = key_keep_stack.back();
-                key_keep_stack.pop_back();
-
-                if (not store_element)
-                {
-                    return {false, nullptr};
-                }
-
-                assert(object_element);
-                *object_element = std::move(value);
-                return {true, object_element};
-            }
-        }
-    }
-
-    /// the parsed JSON value
-    BasicJsonType& root;
-    /// stack to model hierarchy of values
-    std::vector<BasicJsonType*> ref_stack;
-    /// stack to manage which values to keep
-    std::vector<bool> keep_stack;
-    /// stack to manage which object keys to keep
-    std::vector<bool> key_keep_stack;
-    /// helper to hold the reference for the next object element
-    BasicJsonType* object_element = nullptr;
-    /// whether a syntax error occurred
-    bool errored = false;
-    /// callback function
-    const parser_callback_t callback = nullptr;
-    /// whether to throw exceptions in case of errors
-    const bool allow_exceptions = true;
-    /// a discarded value for the callback
-    BasicJsonType discarded = BasicJsonType::value_t::discarded;
-};
-
-template<typename BasicJsonType>
-class json_sax_acceptor
-{
-  public:
-    using number_integer_t = typename BasicJsonType::number_integer_t;
-    using number_unsigned_t = typename BasicJsonType::number_unsigned_t;
-    using number_float_t = typename BasicJsonType::number_float_t;
-    using string_t = typename BasicJsonType::string_t;
-
-    bool null()
-    {
-        return true;
-    }
-
-    bool boolean(bool)
-    {
-        return true;
-    }
-
-    bool number_integer(number_integer_t)
-    {
-        return true;
-    }
-
-    bool number_unsigned(number_unsigned_t)
-    {
-        return true;
-    }
-
-    bool number_float(number_float_t, const string_t&)
-    {
-        return true;
-    }
-
-    bool string(string_t&)
-    {
-        return true;
-    }
-
-    bool start_object(std::size_t = std::size_t(-1))
-    {
-        return true;
-    }
-
-    bool key(string_t&)
-    {
-        return true;
-    }
-
-    bool end_object()
-    {
-        return true;
-    }
-
-    bool start_array(std::size_t = std::size_t(-1))
-    {
-        return true;
-    }
-
-    bool end_array()
-    {
-        return true;
-    }
-
-    bool parse_error(std::size_t, const std::string&, const detail::exception&)
-    {
-        return false;
-    }
-};
-}
-
-}
-
-// #include <nlohmann/detail/input/lexer.hpp>
+// #include <nlohmann/detail/meta/type_traits.hpp>
 
 // #include <nlohmann/detail/value_t.hpp>
 
@@ -4467,1517 +8196,28 @@ namespace nlohmann
 {
 namespace detail
 {
-////////////
-// parser //
-////////////
+
+/// how to treat CBOR tags
+enum class cbor_tag_handler_t
+{
+    error,   ///< throw a parse_error exception in case of a tag
+    ignore,  ///< ignore tags
+    store    ///< store tags as binary type
+};
 
 /*!
-@brief syntax analysis
+@brief determine system byte order
 
-This class implements a recursive decent parser.
+@return true if and only if system's byte order is little endian
+
+@note from https://stackoverflow.com/a/1001328/266378
 */
-template<typename BasicJsonType>
-class parser
+static inline bool little_endianness(int num = 1) noexcept
 {
-    using number_integer_t = typename BasicJsonType::number_integer_t;
-    using number_unsigned_t = typename BasicJsonType::number_unsigned_t;
-    using number_float_t = typename BasicJsonType::number_float_t;
-    using string_t = typename BasicJsonType::string_t;
-    using lexer_t = lexer<BasicJsonType>;
-    using token_type = typename lexer_t::token_type;
-
-  public:
-    enum class parse_event_t : uint8_t
-    {
-        /// the parser read `{` and started to process a JSON object
-        object_start,
-        /// the parser read `}` and finished processing a JSON object
-        object_end,
-        /// the parser read `[` and started to process a JSON array
-        array_start,
-        /// the parser read `]` and finished processing a JSON array
-        array_end,
-        /// the parser read a key of a value in an object
-        key,
-        /// the parser finished reading a JSON value
-        value
-    };
-
-    using parser_callback_t =
-        std::function<bool(int depth, parse_event_t event, BasicJsonType& parsed)>;
-
-    /// a parser reading from an input adapter
-    explicit parser(detail::input_adapter_t&& adapter,
-                    const parser_callback_t cb = nullptr,
-                    const bool allow_exceptions_ = true)
-        : callback(cb), m_lexer(std::move(adapter)), allow_exceptions(allow_exceptions_)
-    {
-        // read first token
-        get_token();
-    }
-
-    /*!
-    @brief public parser interface
-
-    @param[in] strict      whether to expect the last token to be EOF
-    @param[in,out] result  parsed JSON value
-
-    @throw parse_error.101 in case of an unexpected token
-    @throw parse_error.102 if to_unicode fails or surrogate error
-    @throw parse_error.103 if to_unicode fails
-    */
-    void parse(const bool strict, BasicJsonType& result)
-    {
-        if (callback)
-        {
-            json_sax_dom_callback_parser<BasicJsonType> sdp(result, callback, allow_exceptions);
-            sax_parse_internal(&sdp);
-            result.assert_invariant();
-
-            // in strict mode, input must be completely read
-            if (strict and (get_token() != token_type::end_of_input))
-            {
-                sdp.parse_error(m_lexer.get_position(),
-                                m_lexer.get_token_string(),
-                                parse_error::create(101, m_lexer.get_position(), exception_message(token_type::end_of_input)));
-            }
-
-            // in case of an error, return discarded value
-            if (sdp.is_errored())
-            {
-                result = value_t::discarded;
-                return;
-            }
-
-            // set top-level value to null if it was discarded by the callback
-            // function
-            if (result.is_discarded())
-            {
-                result = nullptr;
-            }
-        }
-        else
-        {
-            json_sax_dom_parser<BasicJsonType> sdp(result, allow_exceptions);
-            sax_parse_internal(&sdp);
-            result.assert_invariant();
-
-            // in strict mode, input must be completely read
-            if (strict and (get_token() != token_type::end_of_input))
-            {
-                sdp.parse_error(m_lexer.get_position(),
-                                m_lexer.get_token_string(),
-                                parse_error::create(101, m_lexer.get_position(), exception_message(token_type::end_of_input)));
-            }
-
-            // in case of an error, return discarded value
-            if (sdp.is_errored())
-            {
-                result = value_t::discarded;
-                return;
-            }
-        }
-    }
-
-    /*!
-    @brief public accept interface
-
-    @param[in] strict  whether to expect the last token to be EOF
-    @return whether the input is a proper JSON text
-    */
-    bool accept(const bool strict = true)
-    {
-        json_sax_acceptor<BasicJsonType> sax_acceptor;
-        return sax_parse(&sax_acceptor, strict);
-    }
-
-    template <typename SAX>
-    bool sax_parse(SAX* sax, const bool strict = true)
-    {
-        (void)detail::is_sax_static_asserts<SAX, BasicJsonType> {};
-        const bool result = sax_parse_internal(sax);
-
-        // strict mode: next byte must be EOF
-        if (result and strict and (get_token() != token_type::end_of_input))
-        {
-            return sax->parse_error(m_lexer.get_position(),
-                                    m_lexer.get_token_string(),
-                                    parse_error::create(101, m_lexer.get_position(), exception_message(token_type::end_of_input)));
-        }
-
-        return result;
-    }
-
-  private:
-    template <typename SAX>
-    bool sax_parse_internal(SAX* sax)
-    {
-        // stack to remember the hieararchy of structured values we are parsing
-        // true = array; false = object
-        std::vector<bool> states;
-        // value to avoid a goto (see comment where set to true)
-        bool skip_to_state_evaluation = false;
-
-        while (true)
-        {
-            if (not skip_to_state_evaluation)
-            {
-                // invariant: get_token() was called before each iteration
-                switch (last_token)
-                {
-                    case token_type::begin_object:
-                    {
-                        if (JSON_UNLIKELY(not sax->start_object(std::size_t(-1))))
-                        {
-                            return false;
-                        }
-
-                        // closing } -> we are done
-                        if (get_token() == token_type::end_object)
-                        {
-                            if (JSON_UNLIKELY(not sax->end_object()))
-                            {
-                                return false;
-                            }
-                            break;
-                        }
-
-                        // parse key
-                        if (JSON_UNLIKELY(last_token != token_type::value_string))
-                        {
-                            return sax->parse_error(m_lexer.get_position(),
-                                                    m_lexer.get_token_string(),
-                                                    parse_error::create(101, m_lexer.get_position(), exception_message(token_type::value_string)));
-                        }
-                        else
-                        {
-                            if (JSON_UNLIKELY(not sax->key(m_lexer.get_string())))
-                            {
-                                return false;
-                            }
-                        }
-
-                        // parse separator (:)
-                        if (JSON_UNLIKELY(get_token() != token_type::name_separator))
-                        {
-                            return sax->parse_error(m_lexer.get_position(),
-                                                    m_lexer.get_token_string(),
-                                                    parse_error::create(101, m_lexer.get_position(), exception_message(token_type::name_separator)));
-                        }
-
-                        // remember we are now inside an object
-                        states.push_back(false);
-
-                        // parse values
-                        get_token();
-                        continue;
-                    }
-
-                    case token_type::begin_array:
-                    {
-                        if (JSON_UNLIKELY(not sax->start_array(std::size_t(-1))))
-                        {
-                            return false;
-                        }
-
-                        // closing ] -> we are done
-                        if (get_token() == token_type::end_array)
-                        {
-                            if (JSON_UNLIKELY(not sax->end_array()))
-                            {
-                                return false;
-                            }
-                            break;
-                        }
-
-                        // remember we are now inside an array
-                        states.push_back(true);
-
-                        // parse values (no need to call get_token)
-                        continue;
-                    }
-
-                    case token_type::value_float:
-                    {
-                        const auto res = m_lexer.get_number_float();
-
-                        if (JSON_UNLIKELY(not std::isfinite(res)))
-                        {
-                            return sax->parse_error(m_lexer.get_position(),
-                                                    m_lexer.get_token_string(),
-                                                    out_of_range::create(406, "number overflow parsing '" + m_lexer.get_token_string() + "'"));
-                        }
-                        else
-                        {
-                            if (JSON_UNLIKELY(not sax->number_float(res, m_lexer.get_string())))
-                            {
-                                return false;
-                            }
-                            break;
-                        }
-                    }
-
-                    case token_type::literal_false:
-                    {
-                        if (JSON_UNLIKELY(not sax->boolean(false)))
-                        {
-                            return false;
-                        }
-                        break;
-                    }
-
-                    case token_type::literal_null:
-                    {
-                        if (JSON_UNLIKELY(not sax->null()))
-                        {
-                            return false;
-                        }
-                        break;
-                    }
-
-                    case token_type::literal_true:
-                    {
-                        if (JSON_UNLIKELY(not sax->boolean(true)))
-                        {
-                            return false;
-                        }
-                        break;
-                    }
-
-                    case token_type::value_integer:
-                    {
-                        if (JSON_UNLIKELY(not sax->number_integer(m_lexer.get_number_integer())))
-                        {
-                            return false;
-                        }
-                        break;
-                    }
-
-                    case token_type::value_string:
-                    {
-                        if (JSON_UNLIKELY(not sax->string(m_lexer.get_string())))
-                        {
-                            return false;
-                        }
-                        break;
-                    }
-
-                    case token_type::value_unsigned:
-                    {
-                        if (JSON_UNLIKELY(not sax->number_unsigned(m_lexer.get_number_unsigned())))
-                        {
-                            return false;
-                        }
-                        break;
-                    }
-
-                    case token_type::parse_error:
-                    {
-                        // using "uninitialized" to avoid "expected" message
-                        return sax->parse_error(m_lexer.get_position(),
-                                                m_lexer.get_token_string(),
-                                                parse_error::create(101, m_lexer.get_position(), exception_message(token_type::uninitialized)));
-                    }
-
-                    default: // the last token was unexpected
-                    {
-                        return sax->parse_error(m_lexer.get_position(),
-                                                m_lexer.get_token_string(),
-                                                parse_error::create(101, m_lexer.get_position(), exception_message(token_type::literal_or_value)));
-                    }
-                }
-            }
-            else
-            {
-                skip_to_state_evaluation = false;
-            }
-
-            // we reached this line after we successfully parsed a value
-            if (states.empty())
-            {
-                // empty stack: we reached the end of the hieararchy: done
-                return true;
-            }
-            else
-            {
-                if (states.back())  // array
-                {
-                    // comma -> next value
-                    if (get_token() == token_type::value_separator)
-                    {
-                        // parse a new value
-                        get_token();
-                        continue;
-                    }
-
-                    // closing ]
-                    if (JSON_LIKELY(last_token == token_type::end_array))
-                    {
-                        if (JSON_UNLIKELY(not sax->end_array()))
-                        {
-                            return false;
-                        }
-
-                        // We are done with this array. Before we can parse a
-                        // new value, we need to evaluate the new state first.
-                        // By setting skip_to_state_evaluation to false, we
-                        // are effectively jumping to the beginning of this if.
-                        assert(not states.empty());
-                        states.pop_back();
-                        skip_to_state_evaluation = true;
-                        continue;
-                    }
-                    else
-                    {
-                        return sax->parse_error(m_lexer.get_position(),
-                                                m_lexer.get_token_string(),
-                                                parse_error::create(101, m_lexer.get_position(), exception_message(token_type::end_array)));
-                    }
-                }
-                else  // object
-                {
-                    // comma -> next value
-                    if (get_token() == token_type::value_separator)
-                    {
-                        // parse key
-                        if (JSON_UNLIKELY(get_token() != token_type::value_string))
-                        {
-                            return sax->parse_error(m_lexer.get_position(),
-                                                    m_lexer.get_token_string(),
-                                                    parse_error::create(101, m_lexer.get_position(), exception_message(token_type::value_string)));
-                        }
-                        else
-                        {
-                            if (JSON_UNLIKELY(not sax->key(m_lexer.get_string())))
-                            {
-                                return false;
-                            }
-                        }
-
-                        // parse separator (:)
-                        if (JSON_UNLIKELY(get_token() != token_type::name_separator))
-                        {
-                            return sax->parse_error(m_lexer.get_position(),
-                                                    m_lexer.get_token_string(),
-                                                    parse_error::create(101, m_lexer.get_position(), exception_message(token_type::name_separator)));
-                        }
-
-                        // parse values
-                        get_token();
-                        continue;
-                    }
-
-                    // closing }
-                    if (JSON_LIKELY(last_token == token_type::end_object))
-                    {
-                        if (JSON_UNLIKELY(not sax->end_object()))
-                        {
-                            return false;
-                        }
-
-                        // We are done with this object. Before we can parse a
-                        // new value, we need to evaluate the new state first.
-                        // By setting skip_to_state_evaluation to false, we
-                        // are effectively jumping to the beginning of this if.
-                        assert(not states.empty());
-                        states.pop_back();
-                        skip_to_state_evaluation = true;
-                        continue;
-                    }
-                    else
-                    {
-                        return sax->parse_error(m_lexer.get_position(),
-                                                m_lexer.get_token_string(),
-                                                parse_error::create(101, m_lexer.get_position(), exception_message(token_type::end_object)));
-                    }
-                }
-            }
-        }
-    }
-
-    /// get next token from lexer
-    token_type get_token()
-    {
-        return (last_token = m_lexer.scan());
-    }
-
-    std::string exception_message(const token_type expected)
-    {
-        std::string error_msg = "syntax error - ";
-        if (last_token == token_type::parse_error)
-        {
-            error_msg += std::string(m_lexer.get_error_message()) + "; last read: '" +
-                         m_lexer.get_token_string() + "'";
-        }
-        else
-        {
-            error_msg += "unexpected " + std::string(lexer_t::token_type_name(last_token));
-        }
-
-        if (expected != token_type::uninitialized)
-        {
-            error_msg += "; expected " + std::string(lexer_t::token_type_name(expected));
-        }
-
-        return error_msg;
-    }
-
-  private:
-    /// callback function
-    const parser_callback_t callback = nullptr;
-    /// the type of the last read token
-    token_type last_token = token_type::uninitialized;
-    /// the lexer
-    lexer_t m_lexer;
-    /// whether to throw exceptions in case of errors
-    const bool allow_exceptions = true;
-};
-}
+    return *reinterpret_cast<char*>(&num) == 1;
 }
 
-// #include <nlohmann/detail/iterators/primitive_iterator.hpp>
 
-
-#include <cstddef> // ptrdiff_t
-#include <limits>  // numeric_limits
-
-namespace nlohmann
-{
-namespace detail
-{
-/*
-@brief an iterator for primitive JSON types
-
-This class models an iterator for primitive JSON types (boolean, number,
-string). It's only purpose is to allow the iterator/const_iterator classes
-to "iterate" over primitive values. Internally, the iterator is modeled by
-a `difference_type` variable. Value begin_value (`0`) models the begin,
-end_value (`1`) models past the end.
-*/
-class primitive_iterator_t
-{
-  private:
-    using difference_type = std::ptrdiff_t;
-    static constexpr difference_type begin_value = 0;
-    static constexpr difference_type end_value = begin_value + 1;
-
-    /// iterator as signed integer type
-    difference_type m_it = (std::numeric_limits<std::ptrdiff_t>::min)();
-
-  public:
-    constexpr difference_type get_value() const noexcept
-    {
-        return m_it;
-    }
-
-    /// set iterator to a defined beginning
-    void set_begin() noexcept
-    {
-        m_it = begin_value;
-    }
-
-    /// set iterator to a defined past the end
-    void set_end() noexcept
-    {
-        m_it = end_value;
-    }
-
-    /// return whether the iterator can be dereferenced
-    constexpr bool is_begin() const noexcept
-    {
-        return m_it == begin_value;
-    }
-
-    /// return whether the iterator is at end
-    constexpr bool is_end() const noexcept
-    {
-        return m_it == end_value;
-    }
-
-    friend constexpr bool operator==(primitive_iterator_t lhs, primitive_iterator_t rhs) noexcept
-    {
-        return lhs.m_it == rhs.m_it;
-    }
-
-    friend constexpr bool operator<(primitive_iterator_t lhs, primitive_iterator_t rhs) noexcept
-    {
-        return lhs.m_it < rhs.m_it;
-    }
-
-    primitive_iterator_t operator+(difference_type n) noexcept
-    {
-        auto result = *this;
-        result += n;
-        return result;
-    }
-
-    friend constexpr difference_type operator-(primitive_iterator_t lhs, primitive_iterator_t rhs) noexcept
-    {
-        return lhs.m_it - rhs.m_it;
-    }
-
-    primitive_iterator_t& operator++() noexcept
-    {
-        ++m_it;
-        return *this;
-    }
-
-    primitive_iterator_t const operator++(int) noexcept
-    {
-        auto result = *this;
-        ++m_it;
-        return result;
-    }
-
-    primitive_iterator_t& operator--() noexcept
-    {
-        --m_it;
-        return *this;
-    }
-
-    primitive_iterator_t const operator--(int) noexcept
-    {
-        auto result = *this;
-        --m_it;
-        return result;
-    }
-
-    primitive_iterator_t& operator+=(difference_type n) noexcept
-    {
-        m_it += n;
-        return *this;
-    }
-
-    primitive_iterator_t& operator-=(difference_type n) noexcept
-    {
-        m_it -= n;
-        return *this;
-    }
-};
-}
-}
-
-// #include <nlohmann/detail/iterators/internal_iterator.hpp>
-
-
-// #include <nlohmann/detail/iterators/primitive_iterator.hpp>
-
-
-namespace nlohmann
-{
-namespace detail
-{
-/*!
-@brief an iterator value
-
-@note This structure could easily be a union, but MSVC currently does not allow
-unions members with complex constructors, see https://github.com/nlohmann/json/pull/105.
-*/
-template<typename BasicJsonType> struct internal_iterator
-{
-    /// iterator for JSON objects
-    typename BasicJsonType::object_t::iterator object_iterator {};
-    /// iterator for JSON arrays
-    typename BasicJsonType::array_t::iterator array_iterator {};
-    /// generic iterator for all other types
-    primitive_iterator_t primitive_iterator {};
-};
-}
-}
-
-// #include <nlohmann/detail/iterators/iter_impl.hpp>
-
-
-#include <ciso646> // not
-#include <iterator> // iterator, random_access_iterator_tag, bidirectional_iterator_tag, advance, next
-#include <type_traits> // conditional, is_const, remove_const
-
-// #include <nlohmann/detail/exceptions.hpp>
-
-// #include <nlohmann/detail/iterators/internal_iterator.hpp>
-
-// #include <nlohmann/detail/iterators/primitive_iterator.hpp>
-
-// #include <nlohmann/detail/macro_scope.hpp>
-
-// #include <nlohmann/detail/meta/cpp_future.hpp>
-
-// #include <nlohmann/detail/value_t.hpp>
-
-
-namespace nlohmann
-{
-namespace detail
-{
-// forward declare, to be able to friend it later on
-template<typename IteratorType> class iteration_proxy;
-
-/*!
-@brief a template for a bidirectional iterator for the @ref basic_json class
-
-This class implements a both iterators (iterator and const_iterator) for the
-@ref basic_json class.
-
-@note An iterator is called *initialized* when a pointer to a JSON value has
-      been set (e.g., by a constructor or a copy assignment). If the iterator is
-      default-constructed, it is *uninitialized* and most methods are undefined.
-      **The library uses assertions to detect calls on uninitialized iterators.**
-
-@requirement The class satisfies the following concept requirements:
--
-[BidirectionalIterator](https://en.cppreference.com/w/cpp/named_req/BidirectionalIterator):
-  The iterator that can be moved can be moved in both directions (i.e.
-  incremented and decremented).
-
-@since version 1.0.0, simplified in version 2.0.9, change to bidirectional
-       iterators in version 3.0.0 (see https://github.com/nlohmann/json/issues/593)
-*/
-template<typename BasicJsonType>
-class iter_impl
-{
-    /// allow basic_json to access private members
-    friend iter_impl<typename std::conditional<std::is_const<BasicJsonType>::value, typename std::remove_const<BasicJsonType>::type, const BasicJsonType>::type>;
-    friend BasicJsonType;
-    friend iteration_proxy<iter_impl>;
-
-    using object_t = typename BasicJsonType::object_t;
-    using array_t = typename BasicJsonType::array_t;
-    // make sure BasicJsonType is basic_json or const basic_json
-    static_assert(is_basic_json<typename std::remove_const<BasicJsonType>::type>::value,
-                  "iter_impl only accepts (const) basic_json");
-
-  public:
-
-    /// The std::iterator class template (used as a base class to provide typedefs) is deprecated in C++17.
-    /// The C++ Standard has never required user-defined iterators to derive from std::iterator.
-    /// A user-defined iterator should provide publicly accessible typedefs named
-    /// iterator_category, value_type, difference_type, pointer, and reference.
-    /// Note that value_type is required to be non-const, even for constant iterators.
-    using iterator_category = std::bidirectional_iterator_tag;
-
-    /// the type of the values when the iterator is dereferenced
-    using value_type = typename BasicJsonType::value_type;
-    /// a type to represent differences between iterators
-    using difference_type = typename BasicJsonType::difference_type;
-    /// defines a pointer to the type iterated over (value_type)
-    using pointer = typename std::conditional<std::is_const<BasicJsonType>::value,
-          typename BasicJsonType::const_pointer,
-          typename BasicJsonType::pointer>::type;
-    /// defines a reference to the type iterated over (value_type)
-    using reference =
-        typename std::conditional<std::is_const<BasicJsonType>::value,
-        typename BasicJsonType::const_reference,
-        typename BasicJsonType::reference>::type;
-
-    /// default constructor
-    iter_impl() = default;
-
-    /*!
-    @brief constructor for a given JSON instance
-    @param[in] object  pointer to a JSON object for this iterator
-    @pre object != nullptr
-    @post The iterator is initialized; i.e. `m_object != nullptr`.
-    */
-    explicit iter_impl(pointer object) noexcept : m_object(object)
-    {
-        assert(m_object != nullptr);
-
-        switch (m_object->m_type)
-        {
-            case value_t::object:
-            {
-                m_it.object_iterator = typename object_t::iterator();
-                break;
-            }
-
-            case value_t::array:
-            {
-                m_it.array_iterator = typename array_t::iterator();
-                break;
-            }
-
-            default:
-            {
-                m_it.primitive_iterator = primitive_iterator_t();
-                break;
-            }
-        }
-    }
-
-    /*!
-    @note The conventional copy constructor and copy assignment are implicitly
-          defined. Combined with the following converting constructor and
-          assignment, they support: (1) copy from iterator to iterator, (2)
-          copy from const iterator to const iterator, and (3) conversion from
-          iterator to const iterator. However conversion from const iterator
-          to iterator is not defined.
-    */
-
-    /*!
-    @brief converting constructor
-    @param[in] other  non-const iterator to copy from
-    @note It is not checked whether @a other is initialized.
-    */
-    iter_impl(const iter_impl<typename std::remove_const<BasicJsonType>::type>& other) noexcept
-        : m_object(other.m_object), m_it(other.m_it) {}
-
-    /*!
-    @brief converting assignment
-    @param[in,out] other  non-const iterator to copy from
-    @return const/non-const iterator
-    @note It is not checked whether @a other is initialized.
-    */
-    iter_impl& operator=(const iter_impl<typename std::remove_const<BasicJsonType>::type>& other) noexcept
-    {
-        m_object = other.m_object;
-        m_it = other.m_it;
-        return *this;
-    }
-
-  private:
-    /*!
-    @brief set the iterator to the first value
-    @pre The iterator is initialized; i.e. `m_object != nullptr`.
-    */
-    void set_begin() noexcept
-    {
-        assert(m_object != nullptr);
-
-        switch (m_object->m_type)
-        {
-            case value_t::object:
-            {
-                m_it.object_iterator = m_object->m_value.object->begin();
-                break;
-            }
-
-            case value_t::array:
-            {
-                m_it.array_iterator = m_object->m_value.array->begin();
-                break;
-            }
-
-            case value_t::null:
-            {
-                // set to end so begin()==end() is true: null is empty
-                m_it.primitive_iterator.set_end();
-                break;
-            }
-
-            default:
-            {
-                m_it.primitive_iterator.set_begin();
-                break;
-            }
-        }
-    }
-
-    /*!
-    @brief set the iterator past the last value
-    @pre The iterator is initialized; i.e. `m_object != nullptr`.
-    */
-    void set_end() noexcept
-    {
-        assert(m_object != nullptr);
-
-        switch (m_object->m_type)
-        {
-            case value_t::object:
-            {
-                m_it.object_iterator = m_object->m_value.object->end();
-                break;
-            }
-
-            case value_t::array:
-            {
-                m_it.array_iterator = m_object->m_value.array->end();
-                break;
-            }
-
-            default:
-            {
-                m_it.primitive_iterator.set_end();
-                break;
-            }
-        }
-    }
-
-  public:
-    /*!
-    @brief return a reference to the value pointed to by the iterator
-    @pre The iterator is initialized; i.e. `m_object != nullptr`.
-    */
-    reference operator*() const
-    {
-        assert(m_object != nullptr);
-
-        switch (m_object->m_type)
-        {
-            case value_t::object:
-            {
-                assert(m_it.object_iterator != m_object->m_value.object->end());
-                return m_it.object_iterator->second;
-            }
-
-            case value_t::array:
-            {
-                assert(m_it.array_iterator != m_object->m_value.array->end());
-                return *m_it.array_iterator;
-            }
-
-            case value_t::null:
-                JSON_THROW(invalid_iterator::create(214, "cannot get value"));
-
-            default:
-            {
-                if (JSON_LIKELY(m_it.primitive_iterator.is_begin()))
-                {
-                    return *m_object;
-                }
-
-                JSON_THROW(invalid_iterator::create(214, "cannot get value"));
-            }
-        }
-    }
-
-    /*!
-    @brief dereference the iterator
-    @pre The iterator is initialized; i.e. `m_object != nullptr`.
-    */
-    pointer operator->() const
-    {
-        assert(m_object != nullptr);
-
-        switch (m_object->m_type)
-        {
-            case value_t::object:
-            {
-                assert(m_it.object_iterator != m_object->m_value.object->end());
-                return &(m_it.object_iterator->second);
-            }
-
-            case value_t::array:
-            {
-                assert(m_it.array_iterator != m_object->m_value.array->end());
-                return &*m_it.array_iterator;
-            }
-
-            default:
-            {
-                if (JSON_LIKELY(m_it.primitive_iterator.is_begin()))
-                {
-                    return m_object;
-                }
-
-                JSON_THROW(invalid_iterator::create(214, "cannot get value"));
-            }
-        }
-    }
-
-    /*!
-    @brief post-increment (it++)
-    @pre The iterator is initialized; i.e. `m_object != nullptr`.
-    */
-    iter_impl const operator++(int)
-    {
-        auto result = *this;
-        ++(*this);
-        return result;
-    }
-
-    /*!
-    @brief pre-increment (++it)
-    @pre The iterator is initialized; i.e. `m_object != nullptr`.
-    */
-    iter_impl& operator++()
-    {
-        assert(m_object != nullptr);
-
-        switch (m_object->m_type)
-        {
-            case value_t::object:
-            {
-                std::advance(m_it.object_iterator, 1);
-                break;
-            }
-
-            case value_t::array:
-            {
-                std::advance(m_it.array_iterator, 1);
-                break;
-            }
-
-            default:
-            {
-                ++m_it.primitive_iterator;
-                break;
-            }
-        }
-
-        return *this;
-    }
-
-    /*!
-    @brief post-decrement (it--)
-    @pre The iterator is initialized; i.e. `m_object != nullptr`.
-    */
-    iter_impl const operator--(int)
-    {
-        auto result = *this;
-        --(*this);
-        return result;
-    }
-
-    /*!
-    @brief pre-decrement (--it)
-    @pre The iterator is initialized; i.e. `m_object != nullptr`.
-    */
-    iter_impl& operator--()
-    {
-        assert(m_object != nullptr);
-
-        switch (m_object->m_type)
-        {
-            case value_t::object:
-            {
-                std::advance(m_it.object_iterator, -1);
-                break;
-            }
-
-            case value_t::array:
-            {
-                std::advance(m_it.array_iterator, -1);
-                break;
-            }
-
-            default:
-            {
-                --m_it.primitive_iterator;
-                break;
-            }
-        }
-
-        return *this;
-    }
-
-    /*!
-    @brief  comparison: equal
-    @pre The iterator is initialized; i.e. `m_object != nullptr`.
-    */
-    bool operator==(const iter_impl& other) const
-    {
-        // if objects are not the same, the comparison is undefined
-        if (JSON_UNLIKELY(m_object != other.m_object))
-        {
-            JSON_THROW(invalid_iterator::create(212, "cannot compare iterators of different containers"));
-        }
-
-        assert(m_object != nullptr);
-
-        switch (m_object->m_type)
-        {
-            case value_t::object:
-                return (m_it.object_iterator == other.m_it.object_iterator);
-
-            case value_t::array:
-                return (m_it.array_iterator == other.m_it.array_iterator);
-
-            default:
-                return (m_it.primitive_iterator == other.m_it.primitive_iterator);
-        }
-    }
-
-    /*!
-    @brief  comparison: not equal
-    @pre The iterator is initialized; i.e. `m_object != nullptr`.
-    */
-    bool operator!=(const iter_impl& other) const
-    {
-        return not operator==(other);
-    }
-
-    /*!
-    @brief  comparison: smaller
-    @pre The iterator is initialized; i.e. `m_object != nullptr`.
-    */
-    bool operator<(const iter_impl& other) const
-    {
-        // if objects are not the same, the comparison is undefined
-        if (JSON_UNLIKELY(m_object != other.m_object))
-        {
-            JSON_THROW(invalid_iterator::create(212, "cannot compare iterators of different containers"));
-        }
-
-        assert(m_object != nullptr);
-
-        switch (m_object->m_type)
-        {
-            case value_t::object:
-                JSON_THROW(invalid_iterator::create(213, "cannot compare order of object iterators"));
-
-            case value_t::array:
-                return (m_it.array_iterator < other.m_it.array_iterator);
-
-            default:
-                return (m_it.primitive_iterator < other.m_it.primitive_iterator);
-        }
-    }
-
-    /*!
-    @brief  comparison: less than or equal
-    @pre The iterator is initialized; i.e. `m_object != nullptr`.
-    */
-    bool operator<=(const iter_impl& other) const
-    {
-        return not other.operator < (*this);
-    }
-
-    /*!
-    @brief  comparison: greater than
-    @pre The iterator is initialized; i.e. `m_object != nullptr`.
-    */
-    bool operator>(const iter_impl& other) const
-    {
-        return not operator<=(other);
-    }
-
-    /*!
-    @brief  comparison: greater than or equal
-    @pre The iterator is initialized; i.e. `m_object != nullptr`.
-    */
-    bool operator>=(const iter_impl& other) const
-    {
-        return not operator<(other);
-    }
-
-    /*!
-    @brief  add to iterator
-    @pre The iterator is initialized; i.e. `m_object != nullptr`.
-    */
-    iter_impl& operator+=(difference_type i)
-    {
-        assert(m_object != nullptr);
-
-        switch (m_object->m_type)
-        {
-            case value_t::object:
-                JSON_THROW(invalid_iterator::create(209, "cannot use offsets with object iterators"));
-
-            case value_t::array:
-            {
-                std::advance(m_it.array_iterator, i);
-                break;
-            }
-
-            default:
-            {
-                m_it.primitive_iterator += i;
-                break;
-            }
-        }
-
-        return *this;
-    }
-
-    /*!
-    @brief  subtract from iterator
-    @pre The iterator is initialized; i.e. `m_object != nullptr`.
-    */
-    iter_impl& operator-=(difference_type i)
-    {
-        return operator+=(-i);
-    }
-
-    /*!
-    @brief  add to iterator
-    @pre The iterator is initialized; i.e. `m_object != nullptr`.
-    */
-    iter_impl operator+(difference_type i) const
-    {
-        auto result = *this;
-        result += i;
-        return result;
-    }
-
-    /*!
-    @brief  addition of distance and iterator
-    @pre The iterator is initialized; i.e. `m_object != nullptr`.
-    */
-    friend iter_impl operator+(difference_type i, const iter_impl& it)
-    {
-        auto result = it;
-        result += i;
-        return result;
-    }
-
-    /*!
-    @brief  subtract from iterator
-    @pre The iterator is initialized; i.e. `m_object != nullptr`.
-    */
-    iter_impl operator-(difference_type i) const
-    {
-        auto result = *this;
-        result -= i;
-        return result;
-    }
-
-    /*!
-    @brief  return difference
-    @pre The iterator is initialized; i.e. `m_object != nullptr`.
-    */
-    difference_type operator-(const iter_impl& other) const
-    {
-        assert(m_object != nullptr);
-
-        switch (m_object->m_type)
-        {
-            case value_t::object:
-                JSON_THROW(invalid_iterator::create(209, "cannot use offsets with object iterators"));
-
-            case value_t::array:
-                return m_it.array_iterator - other.m_it.array_iterator;
-
-            default:
-                return m_it.primitive_iterator - other.m_it.primitive_iterator;
-        }
-    }
-
-    /*!
-    @brief  access to successor
-    @pre The iterator is initialized; i.e. `m_object != nullptr`.
-    */
-    reference operator[](difference_type n) const
-    {
-        assert(m_object != nullptr);
-
-        switch (m_object->m_type)
-        {
-            case value_t::object:
-                JSON_THROW(invalid_iterator::create(208, "cannot use operator[] for object iterators"));
-
-            case value_t::array:
-                return *std::next(m_it.array_iterator, n);
-
-            case value_t::null:
-                JSON_THROW(invalid_iterator::create(214, "cannot get value"));
-
-            default:
-            {
-                if (JSON_LIKELY(m_it.primitive_iterator.get_value() == -n))
-                {
-                    return *m_object;
-                }
-
-                JSON_THROW(invalid_iterator::create(214, "cannot get value"));
-            }
-        }
-    }
-
-    /*!
-    @brief  return the key of an object iterator
-    @pre The iterator is initialized; i.e. `m_object != nullptr`.
-    */
-    const typename object_t::key_type& key() const
-    {
-        assert(m_object != nullptr);
-
-        if (JSON_LIKELY(m_object->is_object()))
-        {
-            return m_it.object_iterator->first;
-        }
-
-        JSON_THROW(invalid_iterator::create(207, "cannot use key() for non-object iterators"));
-    }
-
-    /*!
-    @brief  return the value of an iterator
-    @pre The iterator is initialized; i.e. `m_object != nullptr`.
-    */
-    reference value() const
-    {
-        return operator*();
-    }
-
-  private:
-    /// associated JSON instance
-    pointer m_object = nullptr;
-    /// the actual iterator of the associated instance
-    internal_iterator<typename std::remove_const<BasicJsonType>::type> m_it;
-};
-}
-}
-
-// #include <nlohmann/detail/iterators/iteration_proxy.hpp>
-
-// #include <nlohmann/detail/iterators/json_reverse_iterator.hpp>
-
-
-#include <cstddef> // ptrdiff_t
-#include <iterator> // reverse_iterator
-#include <utility> // declval
-
-namespace nlohmann
-{
-namespace detail
-{
-//////////////////////
-// reverse_iterator //
-//////////////////////
-
-/*!
-@brief a template for a reverse iterator class
-
-@tparam Base the base iterator type to reverse. Valid types are @ref
-iterator (to create @ref reverse_iterator) and @ref const_iterator (to
-create @ref const_reverse_iterator).
-
-@requirement The class satisfies the following concept requirements:
--
-[BidirectionalIterator](https://en.cppreference.com/w/cpp/named_req/BidirectionalIterator):
-  The iterator that can be moved can be moved in both directions (i.e.
-  incremented and decremented).
-- [OutputIterator](https://en.cppreference.com/w/cpp/named_req/OutputIterator):
-  It is possible to write to the pointed-to element (only if @a Base is
-  @ref iterator).
-
-@since version 1.0.0
-*/
-template<typename Base>
-class json_reverse_iterator : public std::reverse_iterator<Base>
-{
-  public:
-    using difference_type = std::ptrdiff_t;
-    /// shortcut to the reverse iterator adapter
-    using base_iterator = std::reverse_iterator<Base>;
-    /// the reference type for the pointed-to element
-    using reference = typename Base::reference;
-
-    /// create reverse iterator from iterator
-    explicit json_reverse_iterator(const typename base_iterator::iterator_type& it) noexcept
-        : base_iterator(it) {}
-
-    /// create reverse iterator from base class
-    explicit json_reverse_iterator(const base_iterator& it) noexcept : base_iterator(it) {}
-
-    /// post-increment (it++)
-    json_reverse_iterator const operator++(int)
-    {
-        return static_cast<json_reverse_iterator>(base_iterator::operator++(1));
-    }
-
-    /// pre-increment (++it)
-    json_reverse_iterator& operator++()
-    {
-        return static_cast<json_reverse_iterator&>(base_iterator::operator++());
-    }
-
-    /// post-decrement (it--)
-    json_reverse_iterator const operator--(int)
-    {
-        return static_cast<json_reverse_iterator>(base_iterator::operator--(1));
-    }
-
-    /// pre-decrement (--it)
-    json_reverse_iterator& operator--()
-    {
-        return static_cast<json_reverse_iterator&>(base_iterator::operator--());
-    }
-
-    /// add to iterator
-    json_reverse_iterator& operator+=(difference_type i)
-    {
-        return static_cast<json_reverse_iterator&>(base_iterator::operator+=(i));
-    }
-
-    /// add to iterator
-    json_reverse_iterator operator+(difference_type i) const
-    {
-        return static_cast<json_reverse_iterator>(base_iterator::operator+(i));
-    }
-
-    /// subtract from iterator
-    json_reverse_iterator operator-(difference_type i) const
-    {
-        return static_cast<json_reverse_iterator>(base_iterator::operator-(i));
-    }
-
-    /// return difference
-    difference_type operator-(const json_reverse_iterator& other) const
-    {
-        return base_iterator(*this) - base_iterator(other);
-    }
-
-    /// access to successor
-    reference operator[](difference_type n) const
-    {
-        return *(this->operator+(n));
-    }
-
-    /// return the key of an object iterator
-    auto key() const -> decltype(std::declval<Base>().key())
-    {
-        auto it = --this->base();
-        return it.key();
-    }
-
-    /// return the value of an iterator
-    reference value() const
-    {
-        auto it = --this->base();
-        return it.operator * ();
-    }
-};
-}
-}
-
-// #include <nlohmann/detail/output/output_adapters.hpp>
-
-
-#include <algorithm> // copy
-#include <cstddef> // size_t
-#include <ios> // streamsize
-#include <iterator> // back_inserter
-#include <memory> // shared_ptr, make_shared
-#include <ostream> // basic_ostream
-#include <string> // basic_string
-#include <vector> // vector
-
-namespace nlohmann
-{
-namespace detail
-{
-/// abstract output adapter interface
-template<typename CharType> struct output_adapter_protocol
-{
-    virtual void write_character(CharType c) = 0;
-    virtual void write_characters(const CharType* s, std::size_t length) = 0;
-    virtual ~output_adapter_protocol() = default;
-};
-
-/// a type to simplify interfaces
-template<typename CharType>
-using output_adapter_t = std::shared_ptr<output_adapter_protocol<CharType>>;
-
-/// output adapter for byte vectors
-template<typename CharType>
-class output_vector_adapter : public output_adapter_protocol<CharType>
-{
-  public:
-    explicit output_vector_adapter(std::vector<CharType>& vec) : v(vec) {}
-
-    void write_character(CharType c) override
-    {
-        v.push_back(c);
-    }
-
-    void write_characters(const CharType* s, std::size_t length) override
-    {
-        std::copy(s, s + length, std::back_inserter(v));
-    }
-
-  private:
-    std::vector<CharType>& v;
-};
-
-/// output adapter for output streams
-template<typename CharType>
-class output_stream_adapter : public output_adapter_protocol<CharType>
-{
-  public:
-    explicit output_stream_adapter(std::basic_ostream<CharType>& s) : stream(s) {}
-
-    void write_character(CharType c) override
-    {
-        stream.put(c);
-    }
-
-    void write_characters(const CharType* s, std::size_t length) override
-    {
-        stream.write(s, static_cast<std::streamsize>(length));
-    }
-
-  private:
-    std::basic_ostream<CharType>& stream;
-};
-
-/// output adapter for basic_string
-template<typename CharType, typename StringType = std::basic_string<CharType>>
-class output_string_adapter : public output_adapter_protocol<CharType>
-{
-  public:
-    explicit output_string_adapter(StringType& s) : str(s) {}
-
-    void write_character(CharType c) override
-    {
-        str.push_back(c);
-    }
-
-    void write_characters(const CharType* s, std::size_t length) override
-    {
-        str.append(s, length);
-    }
-
-  private:
-    StringType& str;
-};
-
-template<typename CharType, typename StringType = std::basic_string<CharType>>
-class output_adapter
-{
-  public:
-    output_adapter(std::vector<CharType>& vec)
-        : oa(std::make_shared<output_vector_adapter<CharType>>(vec)) {}
-
-    output_adapter(std::basic_ostream<CharType>& s)
-        : oa(std::make_shared<output_stream_adapter<CharType>>(s)) {}
-
-    output_adapter(StringType& s)
-        : oa(std::make_shared<output_string_adapter<CharType, StringType>>(s)) {}
-
-    operator output_adapter_t<CharType>()
-    {
-        return oa;
-    }
-
-  private:
-    output_adapter_t<CharType> oa = nullptr;
-};
-}
-}
-
-// #include <nlohmann/detail/input/binary_reader.hpp>
-
-
-#include <algorithm> // generate_n
-#include <array> // array
-#include <cassert> // assert
-#include <cmath> // ldexp
-#include <cstddef> // size_t
-#include <cstdint> // uint8_t, uint16_t, uint32_t, uint64_t
-#include <cstdio> // snprintf
-#include <cstring> // memcpy
-#include <iterator> // back_inserter
-#include <limits> // numeric_limits
-#include <string> // char_traits, string
-#include <utility> // make_pair, move
-
-// #include <nlohmann/detail/input/input_adapters.hpp>
-
-// #include <nlohmann/detail/input/json_sax.hpp>
-
-// #include <nlohmann/detail/exceptions.hpp>
-
-// #include <nlohmann/detail/macro_scope.hpp>
-
-// #include <nlohmann/detail/meta/is_sax.hpp>
-
-// #include <nlohmann/detail/value_t.hpp>
-
-
-namespace nlohmann
-{
-namespace detail
-{
 ///////////////////
 // binary reader //
 ///////////////////
@@ -5985,14 +8225,17 @@ namespace detail
 /*!
 @brief deserialization of CBOR, MessagePack, and UBJSON values
 */
-template<typename BasicJsonType, typename SAX = json_sax_dom_parser<BasicJsonType>>
+template<typename BasicJsonType, typename InputAdapterType, typename SAX = json_sax_dom_parser<BasicJsonType>>
 class binary_reader
 {
     using number_integer_t = typename BasicJsonType::number_integer_t;
     using number_unsigned_t = typename BasicJsonType::number_unsigned_t;
     using number_float_t = typename BasicJsonType::number_float_t;
     using string_t = typename BasicJsonType::string_t;
+    using binary_t = typename BasicJsonType::binary_t;
     using json_sax_t = SAX;
+    using char_type = typename InputAdapterType::char_type;
+    using char_int_type = typename std::char_traits<char_type>::int_type;
 
   public:
     /*!
@@ -6000,30 +8243,43 @@ class binary_reader
 
     @param[in] adapter  input adapter to read from
     */
-    explicit binary_reader(input_adapter_t adapter) : ia(std::move(adapter))
+    explicit binary_reader(InputAdapterType&& adapter) noexcept : ia(std::move(adapter))
     {
         (void)detail::is_sax_static_asserts<SAX, BasicJsonType> {};
-        assert(ia);
     }
 
+    // make class move-only
+    binary_reader(const binary_reader&) = delete;
+    binary_reader(binary_reader&&) = default; // NOLINT(hicpp-noexcept-move,performance-noexcept-move-constructor)
+    binary_reader& operator=(const binary_reader&) = delete;
+    binary_reader& operator=(binary_reader&&) = default; // NOLINT(hicpp-noexcept-move,performance-noexcept-move-constructor)
+    ~binary_reader() = default;
+
     /*!
     @param[in] format  the binary format to parse
     @param[in] sax_    a SAX event processor
     @param[in] strict  whether to expect the input to be consumed completed
+    @param[in] tag_handler  how to treat CBOR tags
 
-    @return
+    @return whether parsing was successful
     */
+    JSON_HEDLEY_NON_NULL(3)
     bool sax_parse(const input_format_t format,
                    json_sax_t* sax_,
-                   const bool strict = true)
+                   const bool strict = true,
+                   const cbor_tag_handler_t tag_handler = cbor_tag_handler_t::error)
     {
         sax = sax_;
         bool result = false;
 
         switch (format)
         {
+            case input_format_t::bson:
+                result = parse_bson_internal();
+                break;
+
             case input_format_t::cbor:
-                result = parse_cbor_internal();
+                result = parse_cbor_internal(true, tag_handler);
                 break;
 
             case input_format_t::msgpack:
@@ -6034,14 +8290,13 @@ class binary_reader
                 result = parse_ubjson_internal();
                 break;
 
-            // LCOV_EXCL_START
-            default:
-                assert(false);
-                // LCOV_EXCL_STOP
+            case input_format_t::json: // LCOV_EXCL_LINE
+            default:            // LCOV_EXCL_LINE
+                JSON_ASSERT(false); // NOLINT(cert-dcl03-c,hicpp-static-assert,misc-static-assert) LCOV_EXCL_LINE
         }
 
         // strict mode: next byte must be EOF
-        if (result and strict)
+        if (result && strict)
         {
             if (format == input_format_t::ubjson)
             {
@@ -6052,42 +8307,281 @@ class binary_reader
                 get();
             }
 
-            if (JSON_UNLIKELY(current != std::char_traits<char>::eof()))
+            if (JSON_HEDLEY_UNLIKELY(current != std::char_traits<char_type>::eof()))
             {
-                return sax->parse_error(chars_read, get_token_string(), parse_error::create(110, chars_read, "expected end of input"));
+                return sax->parse_error(chars_read, get_token_string(),
+                                        parse_error::create(110, chars_read, exception_message(format, "expected end of input; last byte: 0x" + get_token_string(), "value"), BasicJsonType()));
             }
         }
 
         return result;
     }
 
+  private:
+    //////////
+    // BSON //
+    //////////
+
     /*!
-    @brief determine system byte order
-
-    @return true if and only if system's byte order is little endian
-
-    @note from http://stackoverflow.com/a/1001328/266378
+    @brief Reads in a BSON-object and passes it to the SAX-parser.
+    @return whether a valid BSON-value was passed to the SAX parser
     */
-    static constexpr bool little_endianess(int num = 1) noexcept
+    bool parse_bson_internal()
     {
-        return (*reinterpret_cast<char*>(&num) == 1);
+        std::int32_t document_size{};
+        get_number<std::int32_t, true>(input_format_t::bson, document_size);
+
+        if (JSON_HEDLEY_UNLIKELY(!sax->start_object(static_cast<std::size_t>(-1))))
+        {
+            return false;
+        }
+
+        if (JSON_HEDLEY_UNLIKELY(!parse_bson_element_list(/*is_array*/false)))
+        {
+            return false;
+        }
+
+        return sax->end_object();
     }
 
-  private:
+    /*!
+    @brief Parses a C-style string from the BSON input.
+    @param[in,out] result  A reference to the string variable where the read
+                            string is to be stored.
+    @return `true` if the \x00-byte indicating the end of the string was
+             encountered before the EOF; false` indicates an unexpected EOF.
+    */
+    bool get_bson_cstr(string_t& result)
+    {
+        auto out = std::back_inserter(result);
+        while (true)
+        {
+            get();
+            if (JSON_HEDLEY_UNLIKELY(!unexpect_eof(input_format_t::bson, "cstring")))
+            {
+                return false;
+            }
+            if (current == 0x00)
+            {
+                return true;
+            }
+            *out++ = static_cast<typename string_t::value_type>(current);
+        }
+    }
+
+    /*!
+    @brief Parses a zero-terminated string of length @a len from the BSON
+           input.
+    @param[in] len  The length (including the zero-byte at the end) of the
+                    string to be read.
+    @param[in,out] result  A reference to the string variable where the read
+                            string is to be stored.
+    @tparam NumberType The type of the length @a len
+    @pre len >= 1
+    @return `true` if the string was successfully parsed
+    */
+    template<typename NumberType>
+    bool get_bson_string(const NumberType len, string_t& result)
+    {
+        if (JSON_HEDLEY_UNLIKELY(len < 1))
+        {
+            auto last_token = get_token_string();
+            return sax->parse_error(chars_read, last_token, parse_error::create(112, chars_read, exception_message(input_format_t::bson, "string length must be at least 1, is " + std::to_string(len), "string"), BasicJsonType()));
+        }
+
+        return get_string(input_format_t::bson, len - static_cast<NumberType>(1), result) && get() != std::char_traits<char_type>::eof();
+    }
+
+    /*!
+    @brief Parses a byte array input of length @a len from the BSON input.
+    @param[in] len  The length of the byte array to be read.
+    @param[in,out] result  A reference to the binary variable where the read
+                            array is to be stored.
+    @tparam NumberType The type of the length @a len
+    @pre len >= 0
+    @return `true` if the byte array was successfully parsed
+    */
+    template<typename NumberType>
+    bool get_bson_binary(const NumberType len, binary_t& result)
+    {
+        if (JSON_HEDLEY_UNLIKELY(len < 0))
+        {
+            auto last_token = get_token_string();
+            return sax->parse_error(chars_read, last_token, parse_error::create(112, chars_read, exception_message(input_format_t::bson, "byte array length cannot be negative, is " + std::to_string(len), "binary"), BasicJsonType()));
+        }
+
+        // All BSON binary values have a subtype
+        std::uint8_t subtype{};
+        get_number<std::uint8_t>(input_format_t::bson, subtype);
+        result.set_subtype(subtype);
+
+        return get_binary(input_format_t::bson, len, result);
+    }
+
+    /*!
+    @brief Read a BSON document element of the given @a element_type.
+    @param[in] element_type The BSON element type, c.f. http://bsonspec.org/spec.html
+    @param[in] element_type_parse_position The position in the input stream,
+               where the `element_type` was read.
+    @warning Not all BSON element types are supported yet. An unsupported
+             @a element_type will give rise to a parse_error.114:
+             Unsupported BSON record type 0x...
+    @return whether a valid BSON-object/array was passed to the SAX parser
+    */
+    bool parse_bson_element_internal(const char_int_type element_type,
+                                     const std::size_t element_type_parse_position)
+    {
+        switch (element_type)
+        {
+            case 0x01: // double
+            {
+                double number{};
+                return get_number<double, true>(input_format_t::bson, number) && sax->number_float(static_cast<number_float_t>(number), "");
+            }
+
+            case 0x02: // string
+            {
+                std::int32_t len{};
+                string_t value;
+                return get_number<std::int32_t, true>(input_format_t::bson, len) && get_bson_string(len, value) && sax->string(value);
+            }
+
+            case 0x03: // object
+            {
+                return parse_bson_internal();
+            }
+
+            case 0x04: // array
+            {
+                return parse_bson_array();
+            }
+
+            case 0x05: // binary
+            {
+                std::int32_t len{};
+                binary_t value;
+                return get_number<std::int32_t, true>(input_format_t::bson, len) && get_bson_binary(len, value) && sax->binary(value);
+            }
+
+            case 0x08: // boolean
+            {
+                return sax->boolean(get() != 0);
+            }
+
+            case 0x0A: // null
+            {
+                return sax->null();
+            }
+
+            case 0x10: // int32
+            {
+                std::int32_t value{};
+                return get_number<std::int32_t, true>(input_format_t::bson, value) && sax->number_integer(value);
+            }
+
+            case 0x12: // int64
+            {
+                std::int64_t value{};
+                return get_number<std::int64_t, true>(input_format_t::bson, value) && sax->number_integer(value);
+            }
+
+            default: // anything else not supported (yet)
+            {
+                std::array<char, 3> cr{{}};
+                static_cast<void>((std::snprintf)(cr.data(), cr.size(), "%.2hhX", static_cast<unsigned char>(element_type))); // NOLINT(cppcoreguidelines-pro-type-vararg,hicpp-vararg)
+                return sax->parse_error(element_type_parse_position, std::string(cr.data()), parse_error::create(114, element_type_parse_position, "Unsupported BSON record type 0x" + std::string(cr.data()), BasicJsonType()));
+            }
+        }
+    }
+
+    /*!
+    @brief Read a BSON element list (as specified in the BSON-spec)
+
+    The same binary layout is used for objects and arrays, hence it must be
+    indicated with the argument @a is_array which one is expected
+    (true --> array, false --> object).
+
+    @param[in] is_array Determines if the element list being read is to be
+                        treated as an object (@a is_array == false), or as an
+                        array (@a is_array == true).
+    @return whether a valid BSON-object/array was passed to the SAX parser
+    */
+    bool parse_bson_element_list(const bool is_array)
+    {
+        string_t key;
+
+        while (auto element_type = get())
+        {
+            if (JSON_HEDLEY_UNLIKELY(!unexpect_eof(input_format_t::bson, "element list")))
+            {
+                return false;
+            }
+
+            const std::size_t element_type_parse_position = chars_read;
+            if (JSON_HEDLEY_UNLIKELY(!get_bson_cstr(key)))
+            {
+                return false;
+            }
+
+            if (!is_array && !sax->key(key))
+            {
+                return false;
+            }
+
+            if (JSON_HEDLEY_UNLIKELY(!parse_bson_element_internal(element_type, element_type_parse_position)))
+            {
+                return false;
+            }
+
+            // get_bson_cstr only appends
+            key.clear();
+        }
+
+        return true;
+    }
+
+    /*!
+    @brief Reads an array from the BSON input and passes it to the SAX-parser.
+    @return whether a valid BSON-array was passed to the SAX parser
+    */
+    bool parse_bson_array()
+    {
+        std::int32_t document_size{};
+        get_number<std::int32_t, true>(input_format_t::bson, document_size);
+
+        if (JSON_HEDLEY_UNLIKELY(!sax->start_array(static_cast<std::size_t>(-1))))
+        {
+            return false;
+        }
+
+        if (JSON_HEDLEY_UNLIKELY(!parse_bson_element_list(/*is_array*/true)))
+        {
+            return false;
+        }
+
+        return sax->end_array();
+    }
+
+    //////////
+    // CBOR //
+    //////////
+
     /*!
     @param[in] get_char  whether a new character should be retrieved from the
-                         input (true, default) or whether the last read
-                         character should be considered instead
+                         input (true) or whether the last read character should
+                         be considered instead (false)
+    @param[in] tag_handler how CBOR tags should be treated
 
     @return whether a valid CBOR value was passed to the SAX parser
     */
-    bool parse_cbor_internal(const bool get_char = true)
+    bool parse_cbor_internal(const bool get_char,
+                             const cbor_tag_handler_t tag_handler)
     {
         switch (get_char ? get() : current)
         {
             // EOF
-            case std::char_traits<char>::eof():
-                return unexpect_eof();
+            case std::char_traits<char_type>::eof():
+                return unexpect_eof(input_format_t::cbor, "value");
 
             // Integer 0x00..0x17 (0..23)
             case 0x00:
@@ -6118,26 +8612,26 @@ class binary_reader
 
             case 0x18: // Unsigned integer (one-byte uint8_t follows)
             {
-                uint8_t number;
-                return get_number(number) and sax->number_unsigned(number);
+                std::uint8_t number{};
+                return get_number(input_format_t::cbor, number) && sax->number_unsigned(number);
             }
 
             case 0x19: // Unsigned integer (two-byte uint16_t follows)
             {
-                uint16_t number;
-                return get_number(number) and sax->number_unsigned(number);
+                std::uint16_t number{};
+                return get_number(input_format_t::cbor, number) && sax->number_unsigned(number);
             }
 
             case 0x1A: // Unsigned integer (four-byte uint32_t follows)
             {
-                uint32_t number;
-                return get_number(number) and sax->number_unsigned(number);
+                std::uint32_t number{};
+                return get_number(input_format_t::cbor, number) && sax->number_unsigned(number);
             }
 
             case 0x1B: // Unsigned integer (eight-byte uint64_t follows)
             {
-                uint64_t number;
-                return get_number(number) and sax->number_unsigned(number);
+                std::uint64_t number{};
+                return get_number(input_format_t::cbor, number) && sax->number_unsigned(number);
             }
 
             // Negative integer -1-0x00..-1-0x17 (-1..-24)
@@ -6165,33 +8659,68 @@ class binary_reader
             case 0x35:
             case 0x36:
             case 0x37:
-                return sax->number_integer(static_cast<int8_t>(0x20 - 1 - current));
+                return sax->number_integer(static_cast<std::int8_t>(0x20 - 1 - current));
 
             case 0x38: // Negative integer (one-byte uint8_t follows)
             {
-                uint8_t number;
-                return get_number(number) and sax->number_integer(static_cast<number_integer_t>(-1) - number);
+                std::uint8_t number{};
+                return get_number(input_format_t::cbor, number) && sax->number_integer(static_cast<number_integer_t>(-1) - number);
             }
 
             case 0x39: // Negative integer -1-n (two-byte uint16_t follows)
             {
-                uint16_t number;
-                return get_number(number) and sax->number_integer(static_cast<number_integer_t>(-1) - number);
+                std::uint16_t number{};
+                return get_number(input_format_t::cbor, number) && sax->number_integer(static_cast<number_integer_t>(-1) - number);
             }
 
             case 0x3A: // Negative integer -1-n (four-byte uint32_t follows)
             {
-                uint32_t number;
-                return get_number(number) and sax->number_integer(static_cast<number_integer_t>(-1) - number);
+                std::uint32_t number{};
+                return get_number(input_format_t::cbor, number) && sax->number_integer(static_cast<number_integer_t>(-1) - number);
             }
 
             case 0x3B: // Negative integer -1-n (eight-byte uint64_t follows)
             {
-                uint64_t number;
-                return get_number(number) and sax->number_integer(static_cast<number_integer_t>(-1)
+                std::uint64_t number{};
+                return get_number(input_format_t::cbor, number) && sax->number_integer(static_cast<number_integer_t>(-1)
                         - static_cast<number_integer_t>(number));
             }
 
+            // Binary data (0x00..0x17 bytes follow)
+            case 0x40:
+            case 0x41:
+            case 0x42:
+            case 0x43:
+            case 0x44:
+            case 0x45:
+            case 0x46:
+            case 0x47:
+            case 0x48:
+            case 0x49:
+            case 0x4A:
+            case 0x4B:
+            case 0x4C:
+            case 0x4D:
+            case 0x4E:
+            case 0x4F:
+            case 0x50:
+            case 0x51:
+            case 0x52:
+            case 0x53:
+            case 0x54:
+            case 0x55:
+            case 0x56:
+            case 0x57:
+            case 0x58: // Binary data (one-byte uint8_t for n follows)
+            case 0x59: // Binary data (two-byte uint16_t for n follow)
+            case 0x5A: // Binary data (four-byte uint32_t for n follow)
+            case 0x5B: // Binary data (eight-byte uint64_t for n follow)
+            case 0x5F: // Binary data (indefinite length)
+            {
+                binary_t b;
+                return get_cbor_binary(b) && sax->binary(b);
+            }
+
             // UTF-8 string (0x00..0x17 bytes follow)
             case 0x60:
             case 0x61:
@@ -6224,7 +8753,7 @@ class binary_reader
             case 0x7F: // UTF-8 string (indefinite length)
             {
                 string_t s;
-                return get_cbor_string(s) and sax->string(s);
+                return get_cbor_string(s) && sax->string(s);
             }
 
             // array (0x00..0x17 data items follow)
@@ -6252,34 +8781,34 @@ class binary_reader
             case 0x95:
             case 0x96:
             case 0x97:
-                return get_cbor_array(static_cast<std::size_t>(current & 0x1F));
+                return get_cbor_array(static_cast<std::size_t>(static_cast<unsigned int>(current) & 0x1Fu), tag_handler);
 
             case 0x98: // array (one-byte uint8_t for n follows)
             {
-                uint8_t len;
-                return get_number(len) and get_cbor_array(static_cast<std::size_t>(len));
+                std::uint8_t len{};
+                return get_number(input_format_t::cbor, len) && get_cbor_array(static_cast<std::size_t>(len), tag_handler);
             }
 
             case 0x99: // array (two-byte uint16_t for n follow)
             {
-                uint16_t len;
-                return get_number(len) and get_cbor_array(static_cast<std::size_t>(len));
+                std::uint16_t len{};
+                return get_number(input_format_t::cbor, len) && get_cbor_array(static_cast<std::size_t>(len), tag_handler);
             }
 
             case 0x9A: // array (four-byte uint32_t for n follow)
             {
-                uint32_t len;
-                return get_number(len) and get_cbor_array(static_cast<std::size_t>(len));
+                std::uint32_t len{};
+                return get_number(input_format_t::cbor, len) && get_cbor_array(static_cast<std::size_t>(len), tag_handler);
             }
 
             case 0x9B: // array (eight-byte uint64_t for n follow)
             {
-                uint64_t len;
-                return get_number(len) and get_cbor_array(static_cast<std::size_t>(len));
+                std::uint64_t len{};
+                return get_number(input_format_t::cbor, len) && get_cbor_array(detail::conditional_static_cast<std::size_t>(len), tag_handler);
             }
 
             case 0x9F: // array (indefinite length)
-                return get_cbor_array(std::size_t(-1));
+                return get_cbor_array(static_cast<std::size_t>(-1), tag_handler);
 
             // map (0x00..0x17 pairs of data items follow)
             case 0xA0:
@@ -6306,34 +8835,144 @@ class binary_reader
             case 0xB5:
             case 0xB6:
             case 0xB7:
-                return get_cbor_object(static_cast<std::size_t>(current & 0x1F));
+                return get_cbor_object(static_cast<std::size_t>(static_cast<unsigned int>(current) & 0x1Fu), tag_handler);
 
             case 0xB8: // map (one-byte uint8_t for n follows)
             {
-                uint8_t len;
-                return get_number(len) and get_cbor_object(static_cast<std::size_t>(len));
+                std::uint8_t len{};
+                return get_number(input_format_t::cbor, len) && get_cbor_object(static_cast<std::size_t>(len), tag_handler);
             }
 
             case 0xB9: // map (two-byte uint16_t for n follow)
             {
-                uint16_t len;
-                return get_number(len) and get_cbor_object(static_cast<std::size_t>(len));
+                std::uint16_t len{};
+                return get_number(input_format_t::cbor, len) && get_cbor_object(static_cast<std::size_t>(len), tag_handler);
             }
 
             case 0xBA: // map (four-byte uint32_t for n follow)
             {
-                uint32_t len;
-                return get_number(len) and get_cbor_object(static_cast<std::size_t>(len));
+                std::uint32_t len{};
+                return get_number(input_format_t::cbor, len) && get_cbor_object(static_cast<std::size_t>(len), tag_handler);
             }
 
             case 0xBB: // map (eight-byte uint64_t for n follow)
             {
-                uint64_t len;
-                return get_number(len) and get_cbor_object(static_cast<std::size_t>(len));
+                std::uint64_t len{};
+                return get_number(input_format_t::cbor, len) && get_cbor_object(detail::conditional_static_cast<std::size_t>(len), tag_handler);
             }
 
             case 0xBF: // map (indefinite length)
-                return get_cbor_object(std::size_t(-1));
+                return get_cbor_object(static_cast<std::size_t>(-1), tag_handler);
+
+            case 0xC6: // tagged item
+            case 0xC7:
+            case 0xC8:
+            case 0xC9:
+            case 0xCA:
+            case 0xCB:
+            case 0xCC:
+            case 0xCD:
+            case 0xCE:
+            case 0xCF:
+            case 0xD0:
+            case 0xD1:
+            case 0xD2:
+            case 0xD3:
+            case 0xD4:
+            case 0xD8: // tagged item (1 bytes follow)
+            case 0xD9: // tagged item (2 bytes follow)
+            case 0xDA: // tagged item (4 bytes follow)
+            case 0xDB: // tagged item (8 bytes follow)
+            {
+                switch (tag_handler)
+                {
+                    case cbor_tag_handler_t::error:
+                    {
+                        auto last_token = get_token_string();
+                        return sax->parse_error(chars_read, last_token, parse_error::create(112, chars_read, exception_message(input_format_t::cbor, "invalid byte: 0x" + last_token, "value"), BasicJsonType()));
+                    }
+
+                    case cbor_tag_handler_t::ignore:
+                    {
+                        // ignore binary subtype
+                        switch (current)
+                        {
+                            case 0xD8:
+                            {
+                                std::uint8_t subtype_to_ignore{};
+                                get_number(input_format_t::cbor, subtype_to_ignore);
+                                break;
+                            }
+                            case 0xD9:
+                            {
+                                std::uint16_t subtype_to_ignore{};
+                                get_number(input_format_t::cbor, subtype_to_ignore);
+                                break;
+                            }
+                            case 0xDA:
+                            {
+                                std::uint32_t subtype_to_ignore{};
+                                get_number(input_format_t::cbor, subtype_to_ignore);
+                                break;
+                            }
+                            case 0xDB:
+                            {
+                                std::uint64_t subtype_to_ignore{};
+                                get_number(input_format_t::cbor, subtype_to_ignore);
+                                break;
+                            }
+                            default:
+                                break;
+                        }
+                        return parse_cbor_internal(true, tag_handler);
+                    }
+
+                    case cbor_tag_handler_t::store:
+                    {
+                        binary_t b;
+                        // use binary subtype and store in binary container
+                        switch (current)
+                        {
+                            case 0xD8:
+                            {
+                                std::uint8_t subtype{};
+                                get_number(input_format_t::cbor, subtype);
+                                b.set_subtype(detail::conditional_static_cast<typename binary_t::subtype_type>(subtype));
+                                break;
+                            }
+                            case 0xD9:
+                            {
+                                std::uint16_t subtype{};
+                                get_number(input_format_t::cbor, subtype);
+                                b.set_subtype(detail::conditional_static_cast<typename binary_t::subtype_type>(subtype));
+                                break;
+                            }
+                            case 0xDA:
+                            {
+                                std::uint32_t subtype{};
+                                get_number(input_format_t::cbor, subtype);
+                                b.set_subtype(detail::conditional_static_cast<typename binary_t::subtype_type>(subtype));
+                                break;
+                            }
+                            case 0xDB:
+                            {
+                                std::uint64_t subtype{};
+                                get_number(input_format_t::cbor, subtype);
+                                b.set_subtype(detail::conditional_static_cast<typename binary_t::subtype_type>(subtype));
+                                break;
+                            }
+                            default:
+                                return parse_cbor_internal(true, tag_handler);
+                        }
+                        get();
+                        return get_cbor_binary(b) && sax->binary(b);
+                    }
+
+                    default:                 // LCOV_EXCL_LINE
+                        JSON_ASSERT(false); // NOLINT(cert-dcl03-c,hicpp-static-assert,misc-static-assert) LCOV_EXCL_LINE
+                        return false;        // LCOV_EXCL_LINE
+                }
+            }
 
             case 0xF4: // false
                 return sax->boolean(false);
@@ -6346,17 +8985,20 @@ class binary_reader
 
             case 0xF9: // Half-Precision Float (two-byte IEEE 754)
             {
-                const int byte1 = get();
-                if (JSON_UNLIKELY(not unexpect_eof()))
+                const auto byte1_raw = get();
+                if (JSON_HEDLEY_UNLIKELY(!unexpect_eof(input_format_t::cbor, "number")))
                 {
                     return false;
                 }
-                const int byte2 = get();
-                if (JSON_UNLIKELY(not unexpect_eof()))
+                const auto byte2_raw = get();
+                if (JSON_HEDLEY_UNLIKELY(!unexpect_eof(input_format_t::cbor, "number")))
                 {
                     return false;
                 }
 
+                const auto byte1 = static_cast<unsigned char>(byte1_raw);
+                const auto byte2 = static_cast<unsigned char>(byte2_raw);
+
                 // code from RFC 7049, Appendix D, Figure 3:
                 // As half-precision floating-point numbers were only added
                 // to IEEE 754 in 2008, today's programming platforms often
@@ -6365,13 +9007,13 @@ class binary_reader
                 // without such support. An example of a small decoder for
                 // half-precision floating-point numbers in the C language
                 // is shown in Fig. 3.
-                const int half = (byte1 << 8) + byte2;
+                const auto half = static_cast<unsigned int>((byte1 << 8u) + byte2);
                 const double val = [&half]
                 {
-                    const int exp = (half >> 10) & 0x1F;
-                    const int mant = half & 0x3FF;
-                    assert(0 <= exp and exp <= 32);
-                    assert(0 <= mant and mant <= 1024);
+                    const int exp = (half >> 10u) & 0x1Fu;
+                    const unsigned int mant = half & 0x3FFu;
+                    JSON_ASSERT(0 <= exp&& exp <= 32);
+                    JSON_ASSERT(mant <= 1024);
                     switch (exp)
                     {
                         case 0:
@@ -6384,31 +9026,322 @@ class binary_reader
                             return std::ldexp(mant + 1024, exp - 25);
                     }
                 }();
-                return sax->number_float((half & 0x8000) != 0
+                return sax->number_float((half & 0x8000u) != 0
                                          ? static_cast<number_float_t>(-val)
                                          : static_cast<number_float_t>(val), "");
             }
 
             case 0xFA: // Single-Precision Float (four-byte IEEE 754)
             {
-                float number;
-                return get_number(number) and sax->number_float(static_cast<number_float_t>(number), "");
+                float number{};
+                return get_number(input_format_t::cbor, number) && sax->number_float(static_cast<number_float_t>(number), "");
             }
 
             case 0xFB: // Double-Precision Float (eight-byte IEEE 754)
             {
-                double number;
-                return get_number(number) and sax->number_float(static_cast<number_float_t>(number), "");
+                double number{};
+                return get_number(input_format_t::cbor, number) && sax->number_float(static_cast<number_float_t>(number), "");
             }
 
             default: // anything else (0xFF is handled inside the other types)
             {
                 auto last_token = get_token_string();
-                return sax->parse_error(chars_read, last_token, parse_error::create(112, chars_read, "error reading CBOR; last byte: 0x" + last_token));
+                return sax->parse_error(chars_read, last_token, parse_error::create(112, chars_read, exception_message(input_format_t::cbor, "invalid byte: 0x" + last_token, "value"), BasicJsonType()));
             }
         }
     }
 
+    /*!
+    @brief reads a CBOR string
+
+    This function first reads starting bytes to determine the expected
+    string length and then copies this number of bytes into a string.
+    Additionally, CBOR's strings with indefinite lengths are supported.
+
+    @param[out] result  created string
+
+    @return whether string creation completed
+    */
+    bool get_cbor_string(string_t& result)
+    {
+        if (JSON_HEDLEY_UNLIKELY(!unexpect_eof(input_format_t::cbor, "string")))
+        {
+            return false;
+        }
+
+        switch (current)
+        {
+            // UTF-8 string (0x00..0x17 bytes follow)
+            case 0x60:
+            case 0x61:
+            case 0x62:
+            case 0x63:
+            case 0x64:
+            case 0x65:
+            case 0x66:
+            case 0x67:
+            case 0x68:
+            case 0x69:
+            case 0x6A:
+            case 0x6B:
+            case 0x6C:
+            case 0x6D:
+            case 0x6E:
+            case 0x6F:
+            case 0x70:
+            case 0x71:
+            case 0x72:
+            case 0x73:
+            case 0x74:
+            case 0x75:
+            case 0x76:
+            case 0x77:
+            {
+                return get_string(input_format_t::cbor, static_cast<unsigned int>(current) & 0x1Fu, result);
+            }
+
+            case 0x78: // UTF-8 string (one-byte uint8_t for n follows)
+            {
+                std::uint8_t len{};
+                return get_number(input_format_t::cbor, len) && get_string(input_format_t::cbor, len, result);
+            }
+
+            case 0x79: // UTF-8 string (two-byte uint16_t for n follow)
+            {
+                std::uint16_t len{};
+                return get_number(input_format_t::cbor, len) && get_string(input_format_t::cbor, len, result);
+            }
+
+            case 0x7A: // UTF-8 string (four-byte uint32_t for n follow)
+            {
+                std::uint32_t len{};
+                return get_number(input_format_t::cbor, len) && get_string(input_format_t::cbor, len, result);
+            }
+
+            case 0x7B: // UTF-8 string (eight-byte uint64_t for n follow)
+            {
+                std::uint64_t len{};
+                return get_number(input_format_t::cbor, len) && get_string(input_format_t::cbor, len, result);
+            }
+
+            case 0x7F: // UTF-8 string (indefinite length)
+            {
+                while (get() != 0xFF)
+                {
+                    string_t chunk;
+                    if (!get_cbor_string(chunk))
+                    {
+                        return false;
+                    }
+                    result.append(chunk);
+                }
+                return true;
+            }
+
+            default:
+            {
+                auto last_token = get_token_string();
+                return sax->parse_error(chars_read, last_token, parse_error::create(113, chars_read, exception_message(input_format_t::cbor, "expected length specification (0x60-0x7B) or indefinite string type (0x7F); last byte: 0x" + last_token, "string"), BasicJsonType()));
+            }
+        }
+    }
+
+    /*!
+    @brief reads a CBOR byte array
+
+    This function first reads starting bytes to determine the expected
+    byte array length and then copies this number of bytes into the byte array.
+    Additionally, CBOR's byte arrays with indefinite lengths are supported.
+
+    @param[out] result  created byte array
+
+    @return whether byte array creation completed
+    */
+    bool get_cbor_binary(binary_t& result)
+    {
+        if (JSON_HEDLEY_UNLIKELY(!unexpect_eof(input_format_t::cbor, "binary")))
+        {
+            return false;
+        }
+
+        switch (current)
+        {
+            // Binary data (0x00..0x17 bytes follow)
+            case 0x40:
+            case 0x41:
+            case 0x42:
+            case 0x43:
+            case 0x44:
+            case 0x45:
+            case 0x46:
+            case 0x47:
+            case 0x48:
+            case 0x49:
+            case 0x4A:
+            case 0x4B:
+            case 0x4C:
+            case 0x4D:
+            case 0x4E:
+            case 0x4F:
+            case 0x50:
+            case 0x51:
+            case 0x52:
+            case 0x53:
+            case 0x54:
+            case 0x55:
+            case 0x56:
+            case 0x57:
+            {
+                return get_binary(input_format_t::cbor, static_cast<unsigned int>(current) & 0x1Fu, result);
+            }
+
+            case 0x58: // Binary data (one-byte uint8_t for n follows)
+            {
+                std::uint8_t len{};
+                return get_number(input_format_t::cbor, len) &&
+                       get_binary(input_format_t::cbor, len, result);
+            }
+
+            case 0x59: // Binary data (two-byte uint16_t for n follow)
+            {
+                std::uint16_t len{};
+                return get_number(input_format_t::cbor, len) &&
+                       get_binary(input_format_t::cbor, len, result);
+            }
+
+            case 0x5A: // Binary data (four-byte uint32_t for n follow)
+            {
+                std::uint32_t len{};
+                return get_number(input_format_t::cbor, len) &&
+                       get_binary(input_format_t::cbor, len, result);
+            }
+
+            case 0x5B: // Binary data (eight-byte uint64_t for n follow)
+            {
+                std::uint64_t len{};
+                return get_number(input_format_t::cbor, len) &&
+                       get_binary(input_format_t::cbor, len, result);
+            }
+
+            case 0x5F: // Binary data (indefinite length)
+            {
+                while (get() != 0xFF)
+                {
+                    binary_t chunk;
+                    if (!get_cbor_binary(chunk))
+                    {
+                        return false;
+                    }
+                    result.insert(result.end(), chunk.begin(), chunk.end());
+                }
+                return true;
+            }
+
+            default:
+            {
+                auto last_token = get_token_string();
+                return sax->parse_error(chars_read, last_token, parse_error::create(113, chars_read, exception_message(input_format_t::cbor, "expected length specification (0x40-0x5B) or indefinite binary array type (0x5F); last byte: 0x" + last_token, "binary"), BasicJsonType()));
+            }
+        }
+    }
+
+    /*!
+    @param[in] len  the length of the array or static_cast<std::size_t>(-1) for an
+                    array of indefinite size
+    @param[in] tag_handler how CBOR tags should be treated
+    @return whether array creation completed
+    */
+    bool get_cbor_array(const std::size_t len,
+                        const cbor_tag_handler_t tag_handler)
+    {
+        if (JSON_HEDLEY_UNLIKELY(!sax->start_array(len)))
+        {
+            return false;
+        }
+
+        if (len != static_cast<std::size_t>(-1))
+        {
+            for (std::size_t i = 0; i < len; ++i)
+            {
+                if (JSON_HEDLEY_UNLIKELY(!parse_cbor_internal(true, tag_handler)))
+                {
+                    return false;
+                }
+            }
+        }
+        else
+        {
+            while (get() != 0xFF)
+            {
+                if (JSON_HEDLEY_UNLIKELY(!parse_cbor_internal(false, tag_handler)))
+                {
+                    return false;
+                }
+            }
+        }
+
+        return sax->end_array();
+    }
+
+    /*!
+    @param[in] len  the length of the object or static_cast<std::size_t>(-1) for an
+                    object of indefinite size
+    @param[in] tag_handler how CBOR tags should be treated
+    @return whether object creation completed
+    */
+    bool get_cbor_object(const std::size_t len,
+                         const cbor_tag_handler_t tag_handler)
+    {
+        if (JSON_HEDLEY_UNLIKELY(!sax->start_object(len)))
+        {
+            return false;
+        }
+
+        if (len != 0)
+        {
+            string_t key;
+            if (len != static_cast<std::size_t>(-1))
+            {
+                for (std::size_t i = 0; i < len; ++i)
+                {
+                    get();
+                    if (JSON_HEDLEY_UNLIKELY(!get_cbor_string(key) || !sax->key(key)))
+                    {
+                        return false;
+                    }
+
+                    if (JSON_HEDLEY_UNLIKELY(!parse_cbor_internal(true, tag_handler)))
+                    {
+                        return false;
+                    }
+                    key.clear();
+                }
+            }
+            else
+            {
+                while (get() != 0xFF)
+                {
+                    if (JSON_HEDLEY_UNLIKELY(!get_cbor_string(key) || !sax->key(key)))
+                    {
+                        return false;
+                    }
+
+                    if (JSON_HEDLEY_UNLIKELY(!parse_cbor_internal(true, tag_handler)))
+                    {
+                        return false;
+                    }
+                    key.clear();
+                }
+            }
+        }
+
+        return sax->end_object();
+    }
+
+    /////////////
+    // MsgPack //
+    /////////////
+
     /*!
     @return whether a valid MessagePack value was passed to the SAX parser
     */
@@ -6417,8 +9350,8 @@ class binary_reader
         switch (get())
         {
             // EOF
-            case std::char_traits<char>::eof():
-                return unexpect_eof();
+            case std::char_traits<char_type>::eof():
+                return unexpect_eof(input_format_t::msgpack, "value");
 
             // positive fixint
             case 0x00:
@@ -6568,7 +9501,7 @@ class binary_reader
             case 0x8D:
             case 0x8E:
             case 0x8F:
-                return get_msgpack_object(static_cast<std::size_t>(current & 0x0F));
+                return get_msgpack_object(static_cast<std::size_t>(static_cast<unsigned int>(current) & 0x0Fu));
 
             // fixarray
             case 0x90:
@@ -6587,7 +9520,7 @@ class binary_reader
             case 0x9D:
             case 0x9E:
             case 0x9F:
-                return get_msgpack_array(static_cast<std::size_t>(current & 0x0F));
+                return get_msgpack_array(static_cast<std::size_t>(static_cast<unsigned int>(current) & 0x0Fu));
 
             // fixstr
             case 0xA0:
@@ -6622,9 +9555,12 @@ class binary_reader
             case 0xBD:
             case 0xBE:
             case 0xBF:
+            case 0xD9: // str 8
+            case 0xDA: // str 16
+            case 0xDB: // str 32
             {
                 string_t s;
-                return get_msgpack_string(s) and sax->string(s);
+                return get_msgpack_string(s) && sax->string(s);
             }
 
             case 0xC0: // nil
@@ -6636,96 +9572,104 @@ class binary_reader
             case 0xC3: // true
                 return sax->boolean(true);
 
+            case 0xC4: // bin 8
+            case 0xC5: // bin 16
+            case 0xC6: // bin 32
+            case 0xC7: // ext 8
+            case 0xC8: // ext 16
+            case 0xC9: // ext 32
+            case 0xD4: // fixext 1
+            case 0xD5: // fixext 2
+            case 0xD6: // fixext 4
+            case 0xD7: // fixext 8
+            case 0xD8: // fixext 16
+            {
+                binary_t b;
+                return get_msgpack_binary(b) && sax->binary(b);
+            }
+
             case 0xCA: // float 32
             {
-                float number;
-                return get_number(number) and sax->number_float(static_cast<number_float_t>(number), "");
+                float number{};
+                return get_number(input_format_t::msgpack, number) && sax->number_float(static_cast<number_float_t>(number), "");
             }
 
             case 0xCB: // float 64
             {
-                double number;
-                return get_number(number) and sax->number_float(static_cast<number_float_t>(number), "");
+                double number{};
+                return get_number(input_format_t::msgpack, number) && sax->number_float(static_cast<number_float_t>(number), "");
             }
 
             case 0xCC: // uint 8
             {
-                uint8_t number;
-                return get_number(number) and sax->number_unsigned(number);
+                std::uint8_t number{};
+                return get_number(input_format_t::msgpack, number) && sax->number_unsigned(number);
             }
 
             case 0xCD: // uint 16
             {
-                uint16_t number;
-                return get_number(number) and sax->number_unsigned(number);
+                std::uint16_t number{};
+                return get_number(input_format_t::msgpack, number) && sax->number_unsigned(number);
             }
 
             case 0xCE: // uint 32
             {
-                uint32_t number;
-                return get_number(number) and sax->number_unsigned(number);
+                std::uint32_t number{};
+                return get_number(input_format_t::msgpack, number) && sax->number_unsigned(number);
             }
 
             case 0xCF: // uint 64
             {
-                uint64_t number;
-                return get_number(number) and sax->number_unsigned(number);
+                std::uint64_t number{};
+                return get_number(input_format_t::msgpack, number) && sax->number_unsigned(number);
             }
 
             case 0xD0: // int 8
             {
-                int8_t number;
-                return get_number(number) and sax->number_integer(number);
+                std::int8_t number{};
+                return get_number(input_format_t::msgpack, number) && sax->number_integer(number);
             }
 
             case 0xD1: // int 16
             {
-                int16_t number;
-                return get_number(number) and sax->number_integer(number);
+                std::int16_t number{};
+                return get_number(input_format_t::msgpack, number) && sax->number_integer(number);
             }
 
             case 0xD2: // int 32
             {
-                int32_t number;
-                return get_number(number) and sax->number_integer(number);
+                std::int32_t number{};
+                return get_number(input_format_t::msgpack, number) && sax->number_integer(number);
             }
 
             case 0xD3: // int 64
             {
-                int64_t number;
-                return get_number(number) and sax->number_integer(number);
-            }
-
-            case 0xD9: // str 8
-            case 0xDA: // str 16
-            case 0xDB: // str 32
-            {
-                string_t s;
-                return get_msgpack_string(s) and sax->string(s);
+                std::int64_t number{};
+                return get_number(input_format_t::msgpack, number) && sax->number_integer(number);
             }
 
             case 0xDC: // array 16
             {
-                uint16_t len;
-                return get_number(len) and get_msgpack_array(static_cast<std::size_t>(len));
+                std::uint16_t len{};
+                return get_number(input_format_t::msgpack, len) && get_msgpack_array(static_cast<std::size_t>(len));
             }
 
             case 0xDD: // array 32
             {
-                uint32_t len;
-                return get_number(len) and get_msgpack_array(static_cast<std::size_t>(len));
+                std::uint32_t len{};
+                return get_number(input_format_t::msgpack, len) && get_msgpack_array(static_cast<std::size_t>(len));
             }
 
             case 0xDE: // map 16
             {
-                uint16_t len;
-                return get_number(len) and get_msgpack_object(static_cast<std::size_t>(len));
+                std::uint16_t len{};
+                return get_number(input_format_t::msgpack, len) && get_msgpack_object(static_cast<std::size_t>(len));
             }
 
             case 0xDF: // map 32
             {
-                uint32_t len;
-                return get_number(len) and get_msgpack_object(static_cast<std::size_t>(len));
+                std::uint32_t len{};
+                return get_number(input_format_t::msgpack, len) && get_msgpack_object(static_cast<std::size_t>(len));
             }
 
             // negative fixint
@@ -6761,306 +9705,16 @@ class binary_reader
             case 0xFD:
             case 0xFE:
             case 0xFF:
-                return sax->number_integer(static_cast<int8_t>(current));
+                return sax->number_integer(static_cast<std::int8_t>(current));
 
             default: // anything else
             {
                 auto last_token = get_token_string();
-                return sax->parse_error(chars_read, last_token, parse_error::create(112, chars_read, "error reading MessagePack; last byte: 0x" + last_token));
+                return sax->parse_error(chars_read, last_token, parse_error::create(112, chars_read, exception_message(input_format_t::msgpack, "invalid byte: 0x" + last_token, "value"), BasicJsonType()));
             }
         }
     }
 
-    /*!
-    @param[in] get_char  whether a new character should be retrieved from the
-                         input (true, default) or whether the last read
-                         character should be considered instead
-
-    @return whether a valid UBJSON value was passed to the SAX parser
-    */
-    bool parse_ubjson_internal(const bool get_char = true)
-    {
-        return get_ubjson_value(get_char ? get_ignore_noop() : current);
-    }
-
-    /*!
-    @brief get next character from the input
-
-    This function provides the interface to the used input adapter. It does
-    not throw in case the input reached EOF, but returns a -'ve valued
-    `std::char_traits<char>::eof()` in that case.
-
-    @return character read from the input
-    */
-    int get()
-    {
-        ++chars_read;
-        return (current = ia->get_character());
-    }
-
-    /*!
-    @return character read from the input after ignoring all 'N' entries
-    */
-    int get_ignore_noop()
-    {
-        do
-        {
-            get();
-        }
-        while (current == 'N');
-
-        return current;
-    }
-
-    /*
-    @brief read a number from the input
-
-    @tparam NumberType the type of the number
-    @param[out] result  number of type @a NumberType
-
-    @return whether conversion completed
-
-    @note This function needs to respect the system's endianess, because
-          bytes in CBOR, MessagePack, and UBJSON are stored in network order
-          (big endian) and therefore need reordering on little endian systems.
-    */
-    template<typename NumberType>
-    bool get_number(NumberType& result)
-    {
-        // step 1: read input into array with system's byte order
-        std::array<uint8_t, sizeof(NumberType)> vec;
-        for (std::size_t i = 0; i < sizeof(NumberType); ++i)
-        {
-            get();
-            if (JSON_UNLIKELY(not unexpect_eof()))
-            {
-                return false;
-            }
-
-            // reverse byte order prior to conversion if necessary
-            if (is_little_endian)
-            {
-                vec[sizeof(NumberType) - i - 1] = static_cast<uint8_t>(current);
-            }
-            else
-            {
-                vec[i] = static_cast<uint8_t>(current); // LCOV_EXCL_LINE
-            }
-        }
-
-        // step 2: convert array into number of type T and return
-        std::memcpy(&result, vec.data(), sizeof(NumberType));
-        return true;
-    }
-
-    /*!
-    @brief create a string by reading characters from the input
-
-    @tparam NumberType the type of the number
-    @param[in] len number of characters to read
-    @param[out] string created by reading @a len bytes
-
-    @return whether string creation completed
-
-    @note We can not reserve @a len bytes for the result, because @a len
-          may be too large. Usually, @ref unexpect_eof() detects the end of
-          the input before we run out of string memory.
-    */
-    template<typename NumberType>
-    bool get_string(const NumberType len, string_t& result)
-    {
-        bool success = true;
-        std::generate_n(std::back_inserter(result), len, [this, &success]()
-        {
-            get();
-            if (JSON_UNLIKELY(not unexpect_eof()))
-            {
-                success = false;
-            }
-            return static_cast<char>(current);
-        });
-        return success;
-    }
-
-    /*!
-    @brief reads a CBOR string
-
-    This function first reads starting bytes to determine the expected
-    string length and then copies this number of bytes into a string.
-    Additionally, CBOR's strings with indefinite lengths are supported.
-
-    @param[out] result  created string
-
-    @return whether string creation completed
-    */
-    bool get_cbor_string(string_t& result)
-    {
-        if (JSON_UNLIKELY(not unexpect_eof()))
-        {
-            return false;
-        }
-
-        switch (current)
-        {
-            // UTF-8 string (0x00..0x17 bytes follow)
-            case 0x60:
-            case 0x61:
-            case 0x62:
-            case 0x63:
-            case 0x64:
-            case 0x65:
-            case 0x66:
-            case 0x67:
-            case 0x68:
-            case 0x69:
-            case 0x6A:
-            case 0x6B:
-            case 0x6C:
-            case 0x6D:
-            case 0x6E:
-            case 0x6F:
-            case 0x70:
-            case 0x71:
-            case 0x72:
-            case 0x73:
-            case 0x74:
-            case 0x75:
-            case 0x76:
-            case 0x77:
-            {
-                return get_string(current & 0x1F, result);
-            }
-
-            case 0x78: // UTF-8 string (one-byte uint8_t for n follows)
-            {
-                uint8_t len;
-                return get_number(len) and get_string(len, result);
-            }
-
-            case 0x79: // UTF-8 string (two-byte uint16_t for n follow)
-            {
-                uint16_t len;
-                return get_number(len) and get_string(len, result);
-            }
-
-            case 0x7A: // UTF-8 string (four-byte uint32_t for n follow)
-            {
-                uint32_t len;
-                return get_number(len) and get_string(len, result);
-            }
-
-            case 0x7B: // UTF-8 string (eight-byte uint64_t for n follow)
-            {
-                uint64_t len;
-                return get_number(len) and get_string(len, result);
-            }
-
-            case 0x7F: // UTF-8 string (indefinite length)
-            {
-                while (get() != 0xFF)
-                {
-                    string_t chunk;
-                    if (not get_cbor_string(chunk))
-                    {
-                        return false;
-                    }
-                    result.append(chunk);
-                }
-                return true;
-            }
-
-            default:
-            {
-                auto last_token = get_token_string();
-                return sax->parse_error(chars_read, last_token, parse_error::create(113, chars_read, "expected a CBOR string; last byte: 0x" + last_token));
-            }
-        }
-    }
-
-    /*!
-    @param[in] len  the length of the array or std::size_t(-1) for an
-                    array of indefinite size
-    @return whether array creation completed
-    */
-    bool get_cbor_array(const std::size_t len)
-    {
-        if (JSON_UNLIKELY(not sax->start_array(len)))
-        {
-            return false;
-        }
-
-        if (len != std::size_t(-1))
-            for (std::size_t i = 0; i < len; ++i)
-            {
-                if (JSON_UNLIKELY(not parse_cbor_internal()))
-                {
-                    return false;
-                }
-            }
-        else
-        {
-            while (get() != 0xFF)
-            {
-                if (JSON_UNLIKELY(not parse_cbor_internal(false)))
-                {
-                    return false;
-                }
-            }
-        }
-
-        return sax->end_array();
-    }
-
-    /*!
-    @param[in] len  the length of the object or std::size_t(-1) for an
-                    object of indefinite size
-    @return whether object creation completed
-    */
-    bool get_cbor_object(const std::size_t len)
-    {
-        if (not JSON_UNLIKELY(sax->start_object(len)))
-        {
-            return false;
-        }
-
-        string_t key;
-        if (len != std::size_t(-1))
-        {
-            for (std::size_t i = 0; i < len; ++i)
-            {
-                get();
-                if (JSON_UNLIKELY(not get_cbor_string(key) or not sax->key(key)))
-                {
-                    return false;
-                }
-
-                if (JSON_UNLIKELY(not parse_cbor_internal()))
-                {
-                    return false;
-                }
-                key.clear();
-            }
-        }
-        else
-        {
-            while (get() != 0xFF)
-            {
-                if (JSON_UNLIKELY(not get_cbor_string(key) or not sax->key(key)))
-                {
-                    return false;
-                }
-
-                if (JSON_UNLIKELY(not parse_cbor_internal()))
-                {
-                    return false;
-                }
-                key.clear();
-            }
-        }
-
-        return sax->end_object();
-    }
-
     /*!
     @brief reads a MessagePack string
 
@@ -7073,7 +9727,7 @@ class binary_reader
     */
     bool get_msgpack_string(string_t& result)
     {
-        if (JSON_UNLIKELY(not unexpect_eof()))
+        if (JSON_HEDLEY_UNLIKELY(!unexpect_eof(input_format_t::msgpack, "string")))
         {
             return false;
         }
@@ -7114,49 +9768,166 @@ class binary_reader
             case 0xBE:
             case 0xBF:
             {
-                return get_string(current & 0x1F, result);
+                return get_string(input_format_t::msgpack, static_cast<unsigned int>(current) & 0x1Fu, result);
             }
 
             case 0xD9: // str 8
             {
-                uint8_t len;
-                return get_number(len) and get_string(len, result);
+                std::uint8_t len{};
+                return get_number(input_format_t::msgpack, len) && get_string(input_format_t::msgpack, len, result);
             }
 
             case 0xDA: // str 16
             {
-                uint16_t len;
-                return get_number(len) and get_string(len, result);
+                std::uint16_t len{};
+                return get_number(input_format_t::msgpack, len) && get_string(input_format_t::msgpack, len, result);
             }
 
             case 0xDB: // str 32
             {
-                uint32_t len;
-                return get_number(len) and get_string(len, result);
+                std::uint32_t len{};
+                return get_number(input_format_t::msgpack, len) && get_string(input_format_t::msgpack, len, result);
             }
 
             default:
             {
                 auto last_token = get_token_string();
-                return sax->parse_error(chars_read, last_token, parse_error::create(113, chars_read, "expected a MessagePack string; last byte: 0x" + last_token));
+                return sax->parse_error(chars_read, last_token, parse_error::create(113, chars_read, exception_message(input_format_t::msgpack, "expected length specification (0xA0-0xBF, 0xD9-0xDB); last byte: 0x" + last_token, "string"), BasicJsonType()));
             }
         }
     }
 
+    /*!
+    @brief reads a MessagePack byte array
+
+    This function first reads starting bytes to determine the expected
+    byte array length and then copies this number of bytes into a byte array.
+
+    @param[out] result  created byte array
+
+    @return whether byte array creation completed
+    */
+    bool get_msgpack_binary(binary_t& result)
+    {
+        // helper function to set the subtype
+        auto assign_and_return_true = [&result](std::int8_t subtype)
+        {
+            result.set_subtype(static_cast<std::uint8_t>(subtype));
+            return true;
+        };
+
+        switch (current)
+        {
+            case 0xC4: // bin 8
+            {
+                std::uint8_t len{};
+                return get_number(input_format_t::msgpack, len) &&
+                       get_binary(input_format_t::msgpack, len, result);
+            }
+
+            case 0xC5: // bin 16
+            {
+                std::uint16_t len{};
+                return get_number(input_format_t::msgpack, len) &&
+                       get_binary(input_format_t::msgpack, len, result);
+            }
+
+            case 0xC6: // bin 32
+            {
+                std::uint32_t len{};
+                return get_number(input_format_t::msgpack, len) &&
+                       get_binary(input_format_t::msgpack, len, result);
+            }
+
+            case 0xC7: // ext 8
+            {
+                std::uint8_t len{};
+                std::int8_t subtype{};
+                return get_number(input_format_t::msgpack, len) &&
+                       get_number(input_format_t::msgpack, subtype) &&
+                       get_binary(input_format_t::msgpack, len, result) &&
+                       assign_and_return_true(subtype);
+            }
+
+            case 0xC8: // ext 16
+            {
+                std::uint16_t len{};
+                std::int8_t subtype{};
+                return get_number(input_format_t::msgpack, len) &&
+                       get_number(input_format_t::msgpack, subtype) &&
+                       get_binary(input_format_t::msgpack, len, result) &&
+                       assign_and_return_true(subtype);
+            }
+
+            case 0xC9: // ext 32
+            {
+                std::uint32_t len{};
+                std::int8_t subtype{};
+                return get_number(input_format_t::msgpack, len) &&
+                       get_number(input_format_t::msgpack, subtype) &&
+                       get_binary(input_format_t::msgpack, len, result) &&
+                       assign_and_return_true(subtype);
+            }
+
+            case 0xD4: // fixext 1
+            {
+                std::int8_t subtype{};
+                return get_number(input_format_t::msgpack, subtype) &&
+                       get_binary(input_format_t::msgpack, 1, result) &&
+                       assign_and_return_true(subtype);
+            }
+
+            case 0xD5: // fixext 2
+            {
+                std::int8_t subtype{};
+                return get_number(input_format_t::msgpack, subtype) &&
+                       get_binary(input_format_t::msgpack, 2, result) &&
+                       assign_and_return_true(subtype);
+            }
+
+            case 0xD6: // fixext 4
+            {
+                std::int8_t subtype{};
+                return get_number(input_format_t::msgpack, subtype) &&
+                       get_binary(input_format_t::msgpack, 4, result) &&
+                       assign_and_return_true(subtype);
+            }
+
+            case 0xD7: // fixext 8
+            {
+                std::int8_t subtype{};
+                return get_number(input_format_t::msgpack, subtype) &&
+                       get_binary(input_format_t::msgpack, 8, result) &&
+                       assign_and_return_true(subtype);
+            }
+
+            case 0xD8: // fixext 16
+            {
+                std::int8_t subtype{};
+                return get_number(input_format_t::msgpack, subtype) &&
+                       get_binary(input_format_t::msgpack, 16, result) &&
+                       assign_and_return_true(subtype);
+            }
+
+            default:           // LCOV_EXCL_LINE
+                return false;  // LCOV_EXCL_LINE
+        }
+    }
+
     /*!
     @param[in] len  the length of the array
     @return whether array creation completed
     */
     bool get_msgpack_array(const std::size_t len)
     {
-        if (JSON_UNLIKELY(not sax->start_array(len)))
+        if (JSON_HEDLEY_UNLIKELY(!sax->start_array(len)))
         {
             return false;
         }
 
         for (std::size_t i = 0; i < len; ++i)
         {
-            if (JSON_UNLIKELY(not parse_msgpack_internal()))
+            if (JSON_HEDLEY_UNLIKELY(!parse_msgpack_internal()))
             {
                 return false;
             }
@@ -7171,7 +9942,7 @@ class binary_reader
     */
     bool get_msgpack_object(const std::size_t len)
     {
-        if (JSON_UNLIKELY(not sax->start_object(len)))
+        if (JSON_HEDLEY_UNLIKELY(!sax->start_object(len)))
         {
             return false;
         }
@@ -7180,12 +9951,12 @@ class binary_reader
         for (std::size_t i = 0; i < len; ++i)
         {
             get();
-            if (JSON_UNLIKELY(not get_msgpack_string(key) or not sax->key(key)))
+            if (JSON_HEDLEY_UNLIKELY(!get_msgpack_string(key) || !sax->key(key)))
             {
                 return false;
             }
 
-            if (JSON_UNLIKELY(not parse_msgpack_internal()))
+            if (JSON_HEDLEY_UNLIKELY(!parse_msgpack_internal()))
             {
                 return false;
             }
@@ -7195,6 +9966,22 @@ class binary_reader
         return sax->end_object();
     }
 
+    ////////////
+    // UBJSON //
+    ////////////
+
+    /*!
+    @param[in] get_char  whether a new character should be retrieved from the
+                         input (true, default) or whether the last read
+                         character should be considered instead
+
+    @return whether a valid UBJSON value was passed to the SAX parser
+    */
+    bool parse_ubjson_internal(const bool get_char = true)
+    {
+        return get_ubjson_value(get_char ? get_ignore_noop() : current);
+    }
+
     /*!
     @brief reads a UBJSON string
 
@@ -7213,10 +10000,10 @@ class binary_reader
     {
         if (get_char)
         {
-            get();  // TODO: may we ignore N here?
+            get();  // TODO(niels): may we ignore N here?
         }
 
-        if (JSON_UNLIKELY(not unexpect_eof()))
+        if (JSON_HEDLEY_UNLIKELY(!unexpect_eof(input_format_t::ubjson, "value")))
         {
             return false;
         }
@@ -7225,37 +10012,37 @@ class binary_reader
         {
             case 'U':
             {
-                uint8_t len;
-                return get_number(len) and get_string(len, result);
+                std::uint8_t len{};
+                return get_number(input_format_t::ubjson, len) && get_string(input_format_t::ubjson, len, result);
             }
 
             case 'i':
             {
-                int8_t len;
-                return get_number(len) and get_string(len, result);
+                std::int8_t len{};
+                return get_number(input_format_t::ubjson, len) && get_string(input_format_t::ubjson, len, result);
             }
 
             case 'I':
             {
-                int16_t len;
-                return get_number(len) and get_string(len, result);
+                std::int16_t len{};
+                return get_number(input_format_t::ubjson, len) && get_string(input_format_t::ubjson, len, result);
             }
 
             case 'l':
             {
-                int32_t len;
-                return get_number(len) and get_string(len, result);
+                std::int32_t len{};
+                return get_number(input_format_t::ubjson, len) && get_string(input_format_t::ubjson, len, result);
             }
 
             case 'L':
             {
-                int64_t len;
-                return get_number(len) and get_string(len, result);
+                std::int64_t len{};
+                return get_number(input_format_t::ubjson, len) && get_string(input_format_t::ubjson, len, result);
             }
 
             default:
                 auto last_token = get_token_string();
-                return sax->parse_error(chars_read, last_token, parse_error::create(113, chars_read, "expected a UBJSON string; last byte: 0x" + last_token));
+                return sax->parse_error(chars_read, last_token, parse_error::create(113, chars_read, exception_message(input_format_t::ubjson, "expected length type specification (U, i, I, l, L); last byte: 0x" + last_token, "string"), BasicJsonType()));
         }
     }
 
@@ -7269,8 +10056,8 @@ class binary_reader
         {
             case 'U':
             {
-                uint8_t number;
-                if (JSON_UNLIKELY(not get_number(number)))
+                std::uint8_t number{};
+                if (JSON_HEDLEY_UNLIKELY(!get_number(input_format_t::ubjson, number)))
                 {
                     return false;
                 }
@@ -7280,19 +10067,19 @@ class binary_reader
 
             case 'i':
             {
-                int8_t number;
-                if (JSON_UNLIKELY(not get_number(number)))
+                std::int8_t number{};
+                if (JSON_HEDLEY_UNLIKELY(!get_number(input_format_t::ubjson, number)))
                 {
                     return false;
                 }
-                result = static_cast<std::size_t>(number);
+                result = static_cast<std::size_t>(number); // NOLINT(bugprone-signed-char-misuse,cert-str34-c): number is not a char
                 return true;
             }
 
             case 'I':
             {
-                int16_t number;
-                if (JSON_UNLIKELY(not get_number(number)))
+                std::int16_t number{};
+                if (JSON_HEDLEY_UNLIKELY(!get_number(input_format_t::ubjson, number)))
                 {
                     return false;
                 }
@@ -7302,8 +10089,8 @@ class binary_reader
 
             case 'l':
             {
-                int32_t number;
-                if (JSON_UNLIKELY(not get_number(number)))
+                std::int32_t number{};
+                if (JSON_HEDLEY_UNLIKELY(!get_number(input_format_t::ubjson, number)))
                 {
                     return false;
                 }
@@ -7313,8 +10100,8 @@ class binary_reader
 
             case 'L':
             {
-                int64_t number;
-                if (JSON_UNLIKELY(not get_number(number)))
+                std::int64_t number{};
+                if (JSON_HEDLEY_UNLIKELY(!get_number(input_format_t::ubjson, number)))
                 {
                     return false;
                 }
@@ -7325,7 +10112,7 @@ class binary_reader
             default:
             {
                 auto last_token = get_token_string();
-                return sax->parse_error(chars_read, last_token, parse_error::create(113, chars_read, "byte after '#' must denote a number type; last byte: 0x" + last_token));
+                return sax->parse_error(chars_read, last_token, parse_error::create(113, chars_read, exception_message(input_format_t::ubjson, "expected length type specification (U, i, I, l, L) after '#'; last byte: 0x" + last_token, "size"), BasicJsonType()));
             }
         }
     }
@@ -7340,7 +10127,7 @@ class binary_reader
 
     @return whether pair creation completed
     */
-    bool get_ubjson_size_type(std::pair<std::size_t, int>& result)
+    bool get_ubjson_size_type(std::pair<std::size_t, char_int_type>& result)
     {
         result.first = string_t::npos; // size
         result.second = 0; // type
@@ -7350,28 +10137,30 @@ class binary_reader
         if (current == '$')
         {
             result.second = get();  // must not ignore 'N', because 'N' maybe the type
-            if (JSON_UNLIKELY(not unexpect_eof()))
+            if (JSON_HEDLEY_UNLIKELY(!unexpect_eof(input_format_t::ubjson, "type")))
             {
                 return false;
             }
 
             get_ignore_noop();
-            if (JSON_UNLIKELY(current != '#'))
+            if (JSON_HEDLEY_UNLIKELY(current != '#'))
             {
-                if (JSON_UNLIKELY(not unexpect_eof()))
+                if (JSON_HEDLEY_UNLIKELY(!unexpect_eof(input_format_t::ubjson, "value")))
                 {
                     return false;
                 }
                 auto last_token = get_token_string();
-                return sax->parse_error(chars_read, last_token, parse_error::create(112, chars_read, "expected '#' after UBJSON type information; last byte: 0x" + last_token));
+                return sax->parse_error(chars_read, last_token, parse_error::create(112, chars_read, exception_message(input_format_t::ubjson, "expected '#' after type information; last byte: 0x" + last_token, "size"), BasicJsonType()));
             }
 
             return get_ubjson_size_value(result.first);
         }
-        else if (current == '#')
+
+        if (current == '#')
         {
             return get_ubjson_size_value(result.first);
         }
+
         return true;
     }
 
@@ -7379,12 +10168,12 @@ class binary_reader
     @param prefix  the previously read or set type prefix
     @return whether value creation completed
     */
-    bool get_ubjson_value(const int prefix)
+    bool get_ubjson_value(const char_int_type prefix)
     {
         switch (prefix)
         {
-            case std::char_traits<char>::eof():  // EOF
-                return unexpect_eof();
+            case std::char_traits<char_type>::eof():  // EOF
+                return unexpect_eof(input_format_t::ubjson, "value");
 
             case 'T':  // true
                 return sax->boolean(true);
@@ -7396,66 +10185,71 @@ class binary_reader
 
             case 'U':
             {
-                uint8_t number;
-                return get_number(number) and sax->number_unsigned(number);
+                std::uint8_t number{};
+                return get_number(input_format_t::ubjson, number) && sax->number_unsigned(number);
             }
 
             case 'i':
             {
-                int8_t number;
-                return get_number(number) and sax->number_integer(number);
+                std::int8_t number{};
+                return get_number(input_format_t::ubjson, number) && sax->number_integer(number);
             }
 
             case 'I':
             {
-                int16_t number;
-                return get_number(number) and sax->number_integer(number);
+                std::int16_t number{};
+                return get_number(input_format_t::ubjson, number) && sax->number_integer(number);
             }
 
             case 'l':
             {
-                int32_t number;
-                return get_number(number) and sax->number_integer(number);
+                std::int32_t number{};
+                return get_number(input_format_t::ubjson, number) && sax->number_integer(number);
             }
 
             case 'L':
             {
-                int64_t number;
-                return get_number(number) and sax->number_integer(number);
+                std::int64_t number{};
+                return get_number(input_format_t::ubjson, number) && sax->number_integer(number);
             }
 
             case 'd':
             {
-                float number;
-                return get_number(number) and sax->number_float(static_cast<number_float_t>(number), "");
+                float number{};
+                return get_number(input_format_t::ubjson, number) && sax->number_float(static_cast<number_float_t>(number), "");
             }
 
             case 'D':
             {
-                double number;
-                return get_number(number) and sax->number_float(static_cast<number_float_t>(number), "");
+                double number{};
+                return get_number(input_format_t::ubjson, number) && sax->number_float(static_cast<number_float_t>(number), "");
+            }
+
+            case 'H':
+            {
+                return get_ubjson_high_precision_number();
             }
 
             case 'C':  // char
             {
                 get();
-                if (JSON_UNLIKELY(not unexpect_eof()))
+                if (JSON_HEDLEY_UNLIKELY(!unexpect_eof(input_format_t::ubjson, "char")))
                 {
                     return false;
                 }
-                if (JSON_UNLIKELY(current > 127))
+                if (JSON_HEDLEY_UNLIKELY(current > 127))
                 {
                     auto last_token = get_token_string();
-                    return sax->parse_error(chars_read, last_token, parse_error::create(113, chars_read, "byte after 'C' must be in range 0x00..0x7F; last byte: 0x" + last_token));
+                    return sax->parse_error(chars_read, last_token, parse_error::create(113, chars_read, exception_message(input_format_t::ubjson, "byte after 'C' must be in range 0x00..0x7F; last byte: 0x" + last_token, "char"), BasicJsonType()));
                 }
-                string_t s(1, static_cast<char>(current));
+                string_t s(1, static_cast<typename string_t::value_type>(current));
                 return sax->string(s);
             }
 
             case 'S':  // string
             {
                 string_t s;
-                return get_ubjson_string(s) and sax->string(s);
+                return get_ubjson_string(s) && sax->string(s);
             }
 
             case '[':  // array
@@ -7467,7 +10261,7 @@ class binary_reader
             default: // anything else
             {
                 auto last_token = get_token_string();
-                return sax->parse_error(chars_read, last_token, parse_error::create(112, chars_read, "error reading UBJSON; last byte: 0x" + last_token));
+                return sax->parse_error(chars_read, last_token, parse_error::create(112, chars_read, exception_message(input_format_t::ubjson, "invalid byte: 0x" + last_token, "value"), BasicJsonType()));
             }
         }
     }
@@ -7477,15 +10271,15 @@ class binary_reader
     */
     bool get_ubjson_array()
     {
-        std::pair<std::size_t, int> size_and_type;
-        if (JSON_UNLIKELY(not get_ubjson_size_type(size_and_type)))
+        std::pair<std::size_t, char_int_type> size_and_type;
+        if (JSON_HEDLEY_UNLIKELY(!get_ubjson_size_type(size_and_type)))
         {
             return false;
         }
 
         if (size_and_type.first != string_t::npos)
         {
-            if (JSON_UNLIKELY(not sax->start_array(size_and_type.first)))
+            if (JSON_HEDLEY_UNLIKELY(!sax->start_array(size_and_type.first)))
             {
                 return false;
             }
@@ -7496,7 +10290,7 @@ class binary_reader
                 {
                     for (std::size_t i = 0; i < size_and_type.first; ++i)
                     {
-                        if (JSON_UNLIKELY(not get_ubjson_value(size_and_type.second)))
+                        if (JSON_HEDLEY_UNLIKELY(!get_ubjson_value(size_and_type.second)))
                         {
                             return false;
                         }
@@ -7507,7 +10301,7 @@ class binary_reader
             {
                 for (std::size_t i = 0; i < size_and_type.first; ++i)
                 {
-                    if (JSON_UNLIKELY(not parse_ubjson_internal()))
+                    if (JSON_HEDLEY_UNLIKELY(!parse_ubjson_internal()))
                     {
                         return false;
                     }
@@ -7516,14 +10310,14 @@ class binary_reader
         }
         else
         {
-            if (JSON_UNLIKELY(not sax->start_array(std::size_t(-1))))
+            if (JSON_HEDLEY_UNLIKELY(!sax->start_array(static_cast<std::size_t>(-1))))
             {
                 return false;
             }
 
             while (current != ']')
             {
-                if (JSON_UNLIKELY(not parse_ubjson_internal(false)))
+                if (JSON_HEDLEY_UNLIKELY(!parse_ubjson_internal(false)))
                 {
                     return false;
                 }
@@ -7539,8 +10333,8 @@ class binary_reader
     */
     bool get_ubjson_object()
     {
-        std::pair<std::size_t, int> size_and_type;
-        if (JSON_UNLIKELY(not get_ubjson_size_type(size_and_type)))
+        std::pair<std::size_t, char_int_type> size_and_type;
+        if (JSON_HEDLEY_UNLIKELY(!get_ubjson_size_type(size_and_type)))
         {
             return false;
         }
@@ -7548,7 +10342,7 @@ class binary_reader
         string_t key;
         if (size_and_type.first != string_t::npos)
         {
-            if (JSON_UNLIKELY(not sax->start_object(size_and_type.first)))
+            if (JSON_HEDLEY_UNLIKELY(!sax->start_object(size_and_type.first)))
             {
                 return false;
             }
@@ -7557,11 +10351,11 @@ class binary_reader
             {
                 for (std::size_t i = 0; i < size_and_type.first; ++i)
                 {
-                    if (JSON_UNLIKELY(not get_ubjson_string(key) or not sax->key(key)))
+                    if (JSON_HEDLEY_UNLIKELY(!get_ubjson_string(key) || !sax->key(key)))
                     {
                         return false;
                     }
-                    if (JSON_UNLIKELY(not get_ubjson_value(size_and_type.second)))
+                    if (JSON_HEDLEY_UNLIKELY(!get_ubjson_value(size_and_type.second)))
                     {
                         return false;
                     }
@@ -7572,11 +10366,11 @@ class binary_reader
             {
                 for (std::size_t i = 0; i < size_and_type.first; ++i)
                 {
-                    if (JSON_UNLIKELY(not get_ubjson_string(key) or not sax->key(key)))
+                    if (JSON_HEDLEY_UNLIKELY(!get_ubjson_string(key) || !sax->key(key)))
                     {
                         return false;
                     }
-                    if (JSON_UNLIKELY(not parse_ubjson_internal()))
+                    if (JSON_HEDLEY_UNLIKELY(!parse_ubjson_internal()))
                     {
                         return false;
                     }
@@ -7586,18 +10380,18 @@ class binary_reader
         }
         else
         {
-            if (JSON_UNLIKELY(not sax->start_object(std::size_t(-1))))
+            if (JSON_HEDLEY_UNLIKELY(!sax->start_object(static_cast<std::size_t>(-1))))
             {
                 return false;
             }
 
             while (current != '}')
             {
-                if (JSON_UNLIKELY(not get_ubjson_string(key, false) or not sax->key(key)))
+                if (JSON_HEDLEY_UNLIKELY(!get_ubjson_string(key, false) || !sax->key(key)))
                 {
                     return false;
                 }
-                if (JSON_UNLIKELY(not parse_ubjson_internal()))
+                if (JSON_HEDLEY_UNLIKELY(!parse_ubjson_internal()))
                 {
                     return false;
                 }
@@ -7609,14 +10403,225 @@ class binary_reader
         return sax->end_object();
     }
 
+    // Note, no reader for UBJSON binary types is implemented because they do
+    // not exist
+
+    bool get_ubjson_high_precision_number()
+    {
+        // get size of following number string
+        std::size_t size{};
+        auto res = get_ubjson_size_value(size);
+        if (JSON_HEDLEY_UNLIKELY(!res))
+        {
+            return res;
+        }
+
+        // get number string
+        std::vector<char> number_vector;
+        for (std::size_t i = 0; i < size; ++i)
+        {
+            get();
+            if (JSON_HEDLEY_UNLIKELY(!unexpect_eof(input_format_t::ubjson, "number")))
+            {
+                return false;
+            }
+            number_vector.push_back(static_cast<char>(current));
+        }
+
+        // parse number string
+        using ia_type = decltype(detail::input_adapter(number_vector));
+        auto number_lexer = detail::lexer<BasicJsonType, ia_type>(detail::input_adapter(number_vector), false);
+        const auto result_number = number_lexer.scan();
+        const auto number_string = number_lexer.get_token_string();
+        const auto result_remainder = number_lexer.scan();
+
+        using token_type = typename detail::lexer_base<BasicJsonType>::token_type;
+
+        if (JSON_HEDLEY_UNLIKELY(result_remainder != token_type::end_of_input))
+        {
+            return sax->parse_error(chars_read, number_string, parse_error::create(115, chars_read, exception_message(input_format_t::ubjson, "invalid number text: " + number_lexer.get_token_string(), "high-precision number"), BasicJsonType()));
+        }
+
+        switch (result_number)
+        {
+            case token_type::value_integer:
+                return sax->number_integer(number_lexer.get_number_integer());
+            case token_type::value_unsigned:
+                return sax->number_unsigned(number_lexer.get_number_unsigned());
+            case token_type::value_float:
+                return sax->number_float(number_lexer.get_number_float(), std::move(number_string));
+            case token_type::uninitialized:
+            case token_type::literal_true:
+            case token_type::literal_false:
+            case token_type::literal_null:
+            case token_type::value_string:
+            case token_type::begin_array:
+            case token_type::begin_object:
+            case token_type::end_array:
+            case token_type::end_object:
+            case token_type::name_separator:
+            case token_type::value_separator:
+            case token_type::parse_error:
+            case token_type::end_of_input:
+            case token_type::literal_or_value:
+            default:
+                return sax->parse_error(chars_read, number_string, parse_error::create(115, chars_read, exception_message(input_format_t::ubjson, "invalid number text: " + number_lexer.get_token_string(), "high-precision number"), BasicJsonType()));
+        }
+    }
+
+    ///////////////////////
+    // Utility functions //
+    ///////////////////////
+
     /*!
+    @brief get next character from the input
+
+    This function provides the interface to the used input adapter. It does
+    not throw in case the input reached EOF, but returns a -'ve valued
+    `std::char_traits<char_type>::eof()` in that case.
+
+    @return character read from the input
+    */
+    char_int_type get()
+    {
+        ++chars_read;
+        return current = ia.get_character();
+    }
+
+    /*!
+    @return character read from the input after ignoring all 'N' entries
+    */
+    char_int_type get_ignore_noop()
+    {
+        do
+        {
+            get();
+        }
+        while (current == 'N');
+
+        return current;
+    }
+
+    /*
+    @brief read a number from the input
+
+    @tparam NumberType the type of the number
+    @param[in] format   the current format (for diagnostics)
+    @param[out] result  number of type @a NumberType
+
+    @return whether conversion completed
+
+    @note This function needs to respect the system's endianness, because
+          bytes in CBOR, MessagePack, and UBJSON are stored in network order
+          (big endian) and therefore need reordering on little endian systems.
+    */
+    template<typename NumberType, bool InputIsLittleEndian = false>
+    bool get_number(const input_format_t format, NumberType& result)
+    {
+        // step 1: read input into array with system's byte order
+        std::array<std::uint8_t, sizeof(NumberType)> vec{};
+        for (std::size_t i = 0; i < sizeof(NumberType); ++i)
+        {
+            get();
+            if (JSON_HEDLEY_UNLIKELY(!unexpect_eof(format, "number")))
+            {
+                return false;
+            }
+
+            // reverse byte order prior to conversion if necessary
+            if (is_little_endian != InputIsLittleEndian)
+            {
+                vec[sizeof(NumberType) - i - 1] = static_cast<std::uint8_t>(current);
+            }
+            else
+            {
+                vec[i] = static_cast<std::uint8_t>(current); // LCOV_EXCL_LINE
+            }
+        }
+
+        // step 2: convert array into number of type T and return
+        std::memcpy(&result, vec.data(), sizeof(NumberType));
+        return true;
+    }
+
+    /*!
+    @brief create a string by reading characters from the input
+
+    @tparam NumberType the type of the number
+    @param[in] format the current format (for diagnostics)
+    @param[in] len number of characters to read
+    @param[out] result string created by reading @a len bytes
+
+    @return whether string creation completed
+
+    @note We can not reserve @a len bytes for the result, because @a len
+          may be too large. Usually, @ref unexpect_eof() detects the end of
+          the input before we run out of string memory.
+    */
+    template<typename NumberType>
+    bool get_string(const input_format_t format,
+                    const NumberType len,
+                    string_t& result)
+    {
+        bool success = true;
+        for (NumberType i = 0; i < len; i++)
+        {
+            get();
+            if (JSON_HEDLEY_UNLIKELY(!unexpect_eof(format, "string")))
+            {
+                success = false;
+                break;
+            }
+            result.push_back(static_cast<typename string_t::value_type>(current));
+        }
+        return success;
+    }
+
+    /*!
+    @brief create a byte array by reading bytes from the input
+
+    @tparam NumberType the type of the number
+    @param[in] format the current format (for diagnostics)
+    @param[in] len number of bytes to read
+    @param[out] result byte array created by reading @a len bytes
+
+    @return whether byte array creation completed
+
+    @note We can not reserve @a len bytes for the result, because @a len
+          may be too large. Usually, @ref unexpect_eof() detects the end of
+          the input before we run out of memory.
+    */
+    template<typename NumberType>
+    bool get_binary(const input_format_t format,
+                    const NumberType len,
+                    binary_t& result)
+    {
+        bool success = true;
+        for (NumberType i = 0; i < len; i++)
+        {
+            get();
+            if (JSON_HEDLEY_UNLIKELY(!unexpect_eof(format, "binary")))
+            {
+                success = false;
+                break;
+            }
+            result.push_back(static_cast<std::uint8_t>(current));
+        }
+        return success;
+    }
+
+    /*!
+    @param[in] format   the current format (for diagnostics)
+    @param[in] context  further context information (for diagnostics)
     @return whether the last read character is not EOF
     */
-    bool unexpect_eof() const
+    JSON_HEDLEY_NON_NULL(3)
+    bool unexpect_eof(const input_format_t format, const char* context) const
     {
-        if (JSON_UNLIKELY(current == std::char_traits<char>::eof()))
+        if (JSON_HEDLEY_UNLIKELY(current == std::char_traits<char_type>::eof()))
         {
-            return sax->parse_error(chars_read, "<end of file>", parse_error::create(110, chars_read, "unexpected end of input"));
+            return sax->parse_error(chars_read, "<end of file>",
+                                    parse_error::create(110, chars_read, exception_message(format, "unexpected end of input", context), BasicJsonType()));
         }
         return true;
     }
@@ -7626,44 +10631,2674 @@ class binary_reader
     */
     std::string get_token_string() const
     {
-        char cr[3];
-        snprintf(cr, 3, "%.2hhX", static_cast<unsigned char>(current));
-        return std::string{cr};
+        std::array<char, 3> cr{{}};
+        static_cast<void>((std::snprintf)(cr.data(), cr.size(), "%.2hhX", static_cast<unsigned char>(current))); // NOLINT(cppcoreguidelines-pro-type-vararg,hicpp-vararg)
+        return std::string{cr.data()};
+    }
+
+    /*!
+    @param[in] format   the current format
+    @param[in] detail   a detailed error message
+    @param[in] context  further context information
+    @return a message string to use in the parse_error exceptions
+    */
+    std::string exception_message(const input_format_t format,
+                                  const std::string& detail,
+                                  const std::string& context) const
+    {
+        std::string error_msg = "syntax error while parsing ";
+
+        switch (format)
+        {
+            case input_format_t::cbor:
+                error_msg += "CBOR";
+                break;
+
+            case input_format_t::msgpack:
+                error_msg += "MessagePack";
+                break;
+
+            case input_format_t::ubjson:
+                error_msg += "UBJSON";
+                break;
+
+            case input_format_t::bson:
+                error_msg += "BSON";
+                break;
+
+            case input_format_t::json: // LCOV_EXCL_LINE
+            default:            // LCOV_EXCL_LINE
+                JSON_ASSERT(false); // NOLINT(cert-dcl03-c,hicpp-static-assert,misc-static-assert) LCOV_EXCL_LINE
+        }
+
+        return error_msg + " " + context + ": " + detail;
     }
 
   private:
     /// input adapter
-    input_adapter_t ia = nullptr;
+    InputAdapterType ia;
 
     /// the current character
-    int current = std::char_traits<char>::eof();
+    char_int_type current = std::char_traits<char_type>::eof();
 
     /// the number of characters read
     std::size_t chars_read = 0;
 
-    /// whether we can assume little endianess
-    const bool is_little_endian = little_endianess();
+    /// whether we can assume little endianness
+    const bool is_little_endian = little_endianness();
 
     /// the SAX parser
     json_sax_t* sax = nullptr;
 };
-}
-}
+}  // namespace detail
+}  // namespace nlohmann
+
+// #include <nlohmann/detail/input/input_adapters.hpp>
+
+// #include <nlohmann/detail/input/lexer.hpp>
+
+// #include <nlohmann/detail/input/parser.hpp>
+
+
+#include <cmath> // isfinite
+#include <cstdint> // uint8_t
+#include <functional> // function
+#include <string> // string
+#include <utility> // move
+#include <vector> // vector
+
+// #include <nlohmann/detail/exceptions.hpp>
+
+// #include <nlohmann/detail/input/input_adapters.hpp>
+
+// #include <nlohmann/detail/input/json_sax.hpp>
+
+// #include <nlohmann/detail/input/lexer.hpp>
+
+// #include <nlohmann/detail/macro_scope.hpp>
+
+// #include <nlohmann/detail/meta/is_sax.hpp>
+
+// #include <nlohmann/detail/value_t.hpp>
+
+
+namespace nlohmann
+{
+namespace detail
+{
+////////////
+// parser //
+////////////
+
+enum class parse_event_t : std::uint8_t
+{
+    /// the parser read `{` and started to process a JSON object
+    object_start,
+    /// the parser read `}` and finished processing a JSON object
+    object_end,
+    /// the parser read `[` and started to process a JSON array
+    array_start,
+    /// the parser read `]` and finished processing a JSON array
+    array_end,
+    /// the parser read a key of a value in an object
+    key,
+    /// the parser finished reading a JSON value
+    value
+};
+
+template<typename BasicJsonType>
+using parser_callback_t =
+    std::function<bool(int /*depth*/, parse_event_t /*event*/, BasicJsonType& /*parsed*/)>;
+
+/*!
+@brief syntax analysis
+
+This class implements a recursive descent parser.
+*/
+template<typename BasicJsonType, typename InputAdapterType>
+class parser
+{
+    using number_integer_t = typename BasicJsonType::number_integer_t;
+    using number_unsigned_t = typename BasicJsonType::number_unsigned_t;
+    using number_float_t = typename BasicJsonType::number_float_t;
+    using string_t = typename BasicJsonType::string_t;
+    using lexer_t = lexer<BasicJsonType, InputAdapterType>;
+    using token_type = typename lexer_t::token_type;
+
+  public:
+    /// a parser reading from an input adapter
+    explicit parser(InputAdapterType&& adapter,
+                    const parser_callback_t<BasicJsonType> cb = nullptr,
+                    const bool allow_exceptions_ = true,
+                    const bool skip_comments = false)
+        : callback(cb)
+        , m_lexer(std::move(adapter), skip_comments)
+        , allow_exceptions(allow_exceptions_)
+    {
+        // read first token
+        get_token();
+    }
+
+    /*!
+    @brief public parser interface
+
+    @param[in] strict      whether to expect the last token to be EOF
+    @param[in,out] result  parsed JSON value
+
+    @throw parse_error.101 in case of an unexpected token
+    @throw parse_error.102 if to_unicode fails or surrogate error
+    @throw parse_error.103 if to_unicode fails
+    */
+    void parse(const bool strict, BasicJsonType& result)
+    {
+        if (callback)
+        {
+            json_sax_dom_callback_parser<BasicJsonType> sdp(result, callback, allow_exceptions);
+            sax_parse_internal(&sdp);
+
+            // in strict mode, input must be completely read
+            if (strict && (get_token() != token_type::end_of_input))
+            {
+                sdp.parse_error(m_lexer.get_position(),
+                                m_lexer.get_token_string(),
+                                parse_error::create(101, m_lexer.get_position(),
+                                                    exception_message(token_type::end_of_input, "value"), BasicJsonType()));
+            }
+
+            // in case of an error, return discarded value
+            if (sdp.is_errored())
+            {
+                result = value_t::discarded;
+                return;
+            }
+
+            // set top-level value to null if it was discarded by the callback
+            // function
+            if (result.is_discarded())
+            {
+                result = nullptr;
+            }
+        }
+        else
+        {
+            json_sax_dom_parser<BasicJsonType> sdp(result, allow_exceptions);
+            sax_parse_internal(&sdp);
+
+            // in strict mode, input must be completely read
+            if (strict && (get_token() != token_type::end_of_input))
+            {
+                sdp.parse_error(m_lexer.get_position(),
+                                m_lexer.get_token_string(),
+                                parse_error::create(101, m_lexer.get_position(), exception_message(token_type::end_of_input, "value"), BasicJsonType()));
+            }
+
+            // in case of an error, return discarded value
+            if (sdp.is_errored())
+            {
+                result = value_t::discarded;
+                return;
+            }
+        }
+
+        result.assert_invariant();
+    }
+
+    /*!
+    @brief public accept interface
+
+    @param[in] strict  whether to expect the last token to be EOF
+    @return whether the input is a proper JSON text
+    */
+    bool accept(const bool strict = true)
+    {
+        json_sax_acceptor<BasicJsonType> sax_acceptor;
+        return sax_parse(&sax_acceptor, strict);
+    }
+
+    template<typename SAX>
+    JSON_HEDLEY_NON_NULL(2)
+    bool sax_parse(SAX* sax, const bool strict = true)
+    {
+        (void)detail::is_sax_static_asserts<SAX, BasicJsonType> {};
+        const bool result = sax_parse_internal(sax);
+
+        // strict mode: next byte must be EOF
+        if (result && strict && (get_token() != token_type::end_of_input))
+        {
+            return sax->parse_error(m_lexer.get_position(),
+                                    m_lexer.get_token_string(),
+                                    parse_error::create(101, m_lexer.get_position(), exception_message(token_type::end_of_input, "value"), BasicJsonType()));
+        }
+
+        return result;
+    }
+
+  private:
+    template<typename SAX>
+    JSON_HEDLEY_NON_NULL(2)
+    bool sax_parse_internal(SAX* sax)
+    {
+        // stack to remember the hierarchy of structured values we are parsing
+        // true = array; false = object
+        std::vector<bool> states;
+        // value to avoid a goto (see comment where set to true)
+        bool skip_to_state_evaluation = false;
+
+        while (true)
+        {
+            if (!skip_to_state_evaluation)
+            {
+                // invariant: get_token() was called before each iteration
+                switch (last_token)
+                {
+                    case token_type::begin_object:
+                    {
+                        if (JSON_HEDLEY_UNLIKELY(!sax->start_object(static_cast<std::size_t>(-1))))
+                        {
+                            return false;
+                        }
+
+                        // closing } -> we are done
+                        if (get_token() == token_type::end_object)
+                        {
+                            if (JSON_HEDLEY_UNLIKELY(!sax->end_object()))
+                            {
+                                return false;
+                            }
+                            break;
+                        }
+
+                        // parse key
+                        if (JSON_HEDLEY_UNLIKELY(last_token != token_type::value_string))
+                        {
+                            return sax->parse_error(m_lexer.get_position(),
+                                                    m_lexer.get_token_string(),
+                                                    parse_error::create(101, m_lexer.get_position(), exception_message(token_type::value_string, "object key"), BasicJsonType()));
+                        }
+                        if (JSON_HEDLEY_UNLIKELY(!sax->key(m_lexer.get_string())))
+                        {
+                            return false;
+                        }
+
+                        // parse separator (:)
+                        if (JSON_HEDLEY_UNLIKELY(get_token() != token_type::name_separator))
+                        {
+                            return sax->parse_error(m_lexer.get_position(),
+                                                    m_lexer.get_token_string(),
+                                                    parse_error::create(101, m_lexer.get_position(), exception_message(token_type::name_separator, "object separator"), BasicJsonType()));
+                        }
+
+                        // remember we are now inside an object
+                        states.push_back(false);
+
+                        // parse values
+                        get_token();
+                        continue;
+                    }
+
+                    case token_type::begin_array:
+                    {
+                        if (JSON_HEDLEY_UNLIKELY(!sax->start_array(static_cast<std::size_t>(-1))))
+                        {
+                            return false;
+                        }
+
+                        // closing ] -> we are done
+                        if (get_token() == token_type::end_array)
+                        {
+                            if (JSON_HEDLEY_UNLIKELY(!sax->end_array()))
+                            {
+                                return false;
+                            }
+                            break;
+                        }
+
+                        // remember we are now inside an array
+                        states.push_back(true);
+
+                        // parse values (no need to call get_token)
+                        continue;
+                    }
+
+                    case token_type::value_float:
+                    {
+                        const auto res = m_lexer.get_number_float();
+
+                        if (JSON_HEDLEY_UNLIKELY(!std::isfinite(res)))
+                        {
+                            return sax->parse_error(m_lexer.get_position(),
+                                                    m_lexer.get_token_string(),
+                                                    out_of_range::create(406, "number overflow parsing '" + m_lexer.get_token_string() + "'", BasicJsonType()));
+                        }
+
+                        if (JSON_HEDLEY_UNLIKELY(!sax->number_float(res, m_lexer.get_string())))
+                        {
+                            return false;
+                        }
+
+                        break;
+                    }
+
+                    case token_type::literal_false:
+                    {
+                        if (JSON_HEDLEY_UNLIKELY(!sax->boolean(false)))
+                        {
+                            return false;
+                        }
+                        break;
+                    }
+
+                    case token_type::literal_null:
+                    {
+                        if (JSON_HEDLEY_UNLIKELY(!sax->null()))
+                        {
+                            return false;
+                        }
+                        break;
+                    }
+
+                    case token_type::literal_true:
+                    {
+                        if (JSON_HEDLEY_UNLIKELY(!sax->boolean(true)))
+                        {
+                            return false;
+                        }
+                        break;
+                    }
+
+                    case token_type::value_integer:
+                    {
+                        if (JSON_HEDLEY_UNLIKELY(!sax->number_integer(m_lexer.get_number_integer())))
+                        {
+                            return false;
+                        }
+                        break;
+                    }
+
+                    case token_type::value_string:
+                    {
+                        if (JSON_HEDLEY_UNLIKELY(!sax->string(m_lexer.get_string())))
+                        {
+                            return false;
+                        }
+                        break;
+                    }
+
+                    case token_type::value_unsigned:
+                    {
+                        if (JSON_HEDLEY_UNLIKELY(!sax->number_unsigned(m_lexer.get_number_unsigned())))
+                        {
+                            return false;
+                        }
+                        break;
+                    }
+
+                    case token_type::parse_error:
+                    {
+                        // using "uninitialized" to avoid "expected" message
+                        return sax->parse_error(m_lexer.get_position(),
+                                                m_lexer.get_token_string(),
+                                                parse_error::create(101, m_lexer.get_position(), exception_message(token_type::uninitialized, "value"), BasicJsonType()));
+                    }
+
+                    case token_type::uninitialized:
+                    case token_type::end_array:
+                    case token_type::end_object:
+                    case token_type::name_separator:
+                    case token_type::value_separator:
+                    case token_type::end_of_input:
+                    case token_type::literal_or_value:
+                    default: // the last token was unexpected
+                    {
+                        return sax->parse_error(m_lexer.get_position(),
+                                                m_lexer.get_token_string(),
+                                                parse_error::create(101, m_lexer.get_position(), exception_message(token_type::literal_or_value, "value"), BasicJsonType()));
+                    }
+                }
+            }
+            else
+            {
+                skip_to_state_evaluation = false;
+            }
+
+            // we reached this line after we successfully parsed a value
+            if (states.empty())
+            {
+                // empty stack: we reached the end of the hierarchy: done
+                return true;
+            }
+
+            if (states.back())  // array
+            {
+                // comma -> next value
+                if (get_token() == token_type::value_separator)
+                {
+                    // parse a new value
+                    get_token();
+                    continue;
+                }
+
+                // closing ]
+                if (JSON_HEDLEY_LIKELY(last_token == token_type::end_array))
+                {
+                    if (JSON_HEDLEY_UNLIKELY(!sax->end_array()))
+                    {
+                        return false;
+                    }
+
+                    // We are done with this array. Before we can parse a
+                    // new value, we need to evaluate the new state first.
+                    // By setting skip_to_state_evaluation to false, we
+                    // are effectively jumping to the beginning of this if.
+                    JSON_ASSERT(!states.empty());
+                    states.pop_back();
+                    skip_to_state_evaluation = true;
+                    continue;
+                }
+
+                return sax->parse_error(m_lexer.get_position(),
+                                        m_lexer.get_token_string(),
+                                        parse_error::create(101, m_lexer.get_position(), exception_message(token_type::end_array, "array"), BasicJsonType()));
+            }
+
+            // states.back() is false -> object
+
+            // comma -> next value
+            if (get_token() == token_type::value_separator)
+            {
+                // parse key
+                if (JSON_HEDLEY_UNLIKELY(get_token() != token_type::value_string))
+                {
+                    return sax->parse_error(m_lexer.get_position(),
+                                            m_lexer.get_token_string(),
+                                            parse_error::create(101, m_lexer.get_position(), exception_message(token_type::value_string, "object key"), BasicJsonType()));
+                }
+
+                if (JSON_HEDLEY_UNLIKELY(!sax->key(m_lexer.get_string())))
+                {
+                    return false;
+                }
+
+                // parse separator (:)
+                if (JSON_HEDLEY_UNLIKELY(get_token() != token_type::name_separator))
+                {
+                    return sax->parse_error(m_lexer.get_position(),
+                                            m_lexer.get_token_string(),
+                                            parse_error::create(101, m_lexer.get_position(), exception_message(token_type::name_separator, "object separator"), BasicJsonType()));
+                }
+
+                // parse values
+                get_token();
+                continue;
+            }
+
+            // closing }
+            if (JSON_HEDLEY_LIKELY(last_token == token_type::end_object))
+            {
+                if (JSON_HEDLEY_UNLIKELY(!sax->end_object()))
+                {
+                    return false;
+                }
+
+                // We are done with this object. Before we can parse a
+                // new value, we need to evaluate the new state first.
+                // By setting skip_to_state_evaluation to false, we
+                // are effectively jumping to the beginning of this if.
+                JSON_ASSERT(!states.empty());
+                states.pop_back();
+                skip_to_state_evaluation = true;
+                continue;
+            }
+
+            return sax->parse_error(m_lexer.get_position(),
+                                    m_lexer.get_token_string(),
+                                    parse_error::create(101, m_lexer.get_position(), exception_message(token_type::end_object, "object"), BasicJsonType()));
+        }
+    }
+
+    /// get next token from lexer
+    token_type get_token()
+    {
+        return last_token = m_lexer.scan();
+    }
+
+    std::string exception_message(const token_type expected, const std::string& context)
+    {
+        std::string error_msg = "syntax error ";
+
+        if (!context.empty())
+        {
+            error_msg += "while parsing " + context + " ";
+        }
+
+        error_msg += "- ";
+
+        if (last_token == token_type::parse_error)
+        {
+            error_msg += std::string(m_lexer.get_error_message()) + "; last read: '" +
+                         m_lexer.get_token_string() + "'";
+        }
+        else
+        {
+            error_msg += "unexpected " + std::string(lexer_t::token_type_name(last_token));
+        }
+
+        if (expected != token_type::uninitialized)
+        {
+            error_msg += "; expected " + std::string(lexer_t::token_type_name(expected));
+        }
+
+        return error_msg;
+    }
+
+  private:
+    /// callback function
+    const parser_callback_t<BasicJsonType> callback = nullptr;
+    /// the type of the last read token
+    token_type last_token = token_type::uninitialized;
+    /// the lexer
+    lexer_t m_lexer;
+    /// whether to throw exceptions in case of errors
+    const bool allow_exceptions = true;
+};
+
+}  // namespace detail
+}  // namespace nlohmann
+
+// #include <nlohmann/detail/iterators/internal_iterator.hpp>
+
+
+// #include <nlohmann/detail/iterators/primitive_iterator.hpp>
+
+
+#include <cstddef> // ptrdiff_t
+#include <limits>  // numeric_limits
+
+// #include <nlohmann/detail/macro_scope.hpp>
+
+
+namespace nlohmann
+{
+namespace detail
+{
+/*
+@brief an iterator for primitive JSON types
+
+This class models an iterator for primitive JSON types (boolean, number,
+string). It's only purpose is to allow the iterator/const_iterator classes
+to "iterate" over primitive values. Internally, the iterator is modeled by
+a `difference_type` variable. Value begin_value (`0`) models the begin,
+end_value (`1`) models past the end.
+*/
+class primitive_iterator_t
+{
+  private:
+    using difference_type = std::ptrdiff_t;
+    static constexpr difference_type begin_value = 0;
+    static constexpr difference_type end_value = begin_value + 1;
+
+  JSON_PRIVATE_UNLESS_TESTED:
+    /// iterator as signed integer type
+    difference_type m_it = (std::numeric_limits<std::ptrdiff_t>::min)();
+
+  public:
+    constexpr difference_type get_value() const noexcept
+    {
+        return m_it;
+    }
+
+    /// set iterator to a defined beginning
+    void set_begin() noexcept
+    {
+        m_it = begin_value;
+    }
+
+    /// set iterator to a defined past the end
+    void set_end() noexcept
+    {
+        m_it = end_value;
+    }
+
+    /// return whether the iterator can be dereferenced
+    constexpr bool is_begin() const noexcept
+    {
+        return m_it == begin_value;
+    }
+
+    /// return whether the iterator is at end
+    constexpr bool is_end() const noexcept
+    {
+        return m_it == end_value;
+    }
+
+    friend constexpr bool operator==(primitive_iterator_t lhs, primitive_iterator_t rhs) noexcept
+    {
+        return lhs.m_it == rhs.m_it;
+    }
+
+    friend constexpr bool operator<(primitive_iterator_t lhs, primitive_iterator_t rhs) noexcept
+    {
+        return lhs.m_it < rhs.m_it;
+    }
+
+    primitive_iterator_t operator+(difference_type n) noexcept
+    {
+        auto result = *this;
+        result += n;
+        return result;
+    }
+
+    friend constexpr difference_type operator-(primitive_iterator_t lhs, primitive_iterator_t rhs) noexcept
+    {
+        return lhs.m_it - rhs.m_it;
+    }
+
+    primitive_iterator_t& operator++() noexcept
+    {
+        ++m_it;
+        return *this;
+    }
+
+    primitive_iterator_t const operator++(int) noexcept // NOLINT(readability-const-return-type)
+    {
+        auto result = *this;
+        ++m_it;
+        return result;
+    }
+
+    primitive_iterator_t& operator--() noexcept
+    {
+        --m_it;
+        return *this;
+    }
+
+    primitive_iterator_t const operator--(int) noexcept // NOLINT(readability-const-return-type)
+    {
+        auto result = *this;
+        --m_it;
+        return result;
+    }
+
+    primitive_iterator_t& operator+=(difference_type n) noexcept
+    {
+        m_it += n;
+        return *this;
+    }
+
+    primitive_iterator_t& operator-=(difference_type n) noexcept
+    {
+        m_it -= n;
+        return *this;
+    }
+};
+}  // namespace detail
+}  // namespace nlohmann
+
+
+namespace nlohmann
+{
+namespace detail
+{
+/*!
+@brief an iterator value
+
+@note This structure could easily be a union, but MSVC currently does not allow
+unions members with complex constructors, see https://github.com/nlohmann/json/pull/105.
+*/
+template<typename BasicJsonType> struct internal_iterator
+{
+    /// iterator for JSON objects
+    typename BasicJsonType::object_t::iterator object_iterator {};
+    /// iterator for JSON arrays
+    typename BasicJsonType::array_t::iterator array_iterator {};
+    /// generic iterator for all other types
+    primitive_iterator_t primitive_iterator {};
+};
+}  // namespace detail
+}  // namespace nlohmann
+
+// #include <nlohmann/detail/iterators/iter_impl.hpp>
+
+
+#include <iterator> // iterator, random_access_iterator_tag, bidirectional_iterator_tag, advance, next
+#include <type_traits> // conditional, is_const, remove_const
+
+// #include <nlohmann/detail/exceptions.hpp>
+
+// #include <nlohmann/detail/iterators/internal_iterator.hpp>
+
+// #include <nlohmann/detail/iterators/primitive_iterator.hpp>
+
+// #include <nlohmann/detail/macro_scope.hpp>
+
+// #include <nlohmann/detail/meta/cpp_future.hpp>
+
+// #include <nlohmann/detail/meta/type_traits.hpp>
+
+// #include <nlohmann/detail/value_t.hpp>
+
+
+namespace nlohmann
+{
+namespace detail
+{
+// forward declare, to be able to friend it later on
+template<typename IteratorType> class iteration_proxy;
+template<typename IteratorType> class iteration_proxy_value;
+
+/*!
+@brief a template for a bidirectional iterator for the @ref basic_json class
+This class implements a both iterators (iterator and const_iterator) for the
+@ref basic_json class.
+@note An iterator is called *initialized* when a pointer to a JSON value has
+      been set (e.g., by a constructor or a copy assignment). If the iterator is
+      default-constructed, it is *uninitialized* and most methods are undefined.
+      **The library uses assertions to detect calls on uninitialized iterators.**
+@requirement The class satisfies the following concept requirements:
+-
+[BidirectionalIterator](https://en.cppreference.com/w/cpp/named_req/BidirectionalIterator):
+  The iterator that can be moved can be moved in both directions (i.e.
+  incremented and decremented).
+@since version 1.0.0, simplified in version 2.0.9, change to bidirectional
+       iterators in version 3.0.0 (see https://github.com/nlohmann/json/issues/593)
+*/
+template<typename BasicJsonType>
+class iter_impl // NOLINT(cppcoreguidelines-special-member-functions,hicpp-special-member-functions)
+{
+    /// the iterator with BasicJsonType of different const-ness
+    using other_iter_impl = iter_impl<typename std::conditional<std::is_const<BasicJsonType>::value, typename std::remove_const<BasicJsonType>::type, const BasicJsonType>::type>;
+    /// allow basic_json to access private members
+    friend other_iter_impl;
+    friend BasicJsonType;
+    friend iteration_proxy<iter_impl>;
+    friend iteration_proxy_value<iter_impl>;
+
+    using object_t = typename BasicJsonType::object_t;
+    using array_t = typename BasicJsonType::array_t;
+    // make sure BasicJsonType is basic_json or const basic_json
+    static_assert(is_basic_json<typename std::remove_const<BasicJsonType>::type>::value,
+                  "iter_impl only accepts (const) basic_json");
+
+  public:
+
+    /// The std::iterator class template (used as a base class to provide typedefs) is deprecated in C++17.
+    /// The C++ Standard has never required user-defined iterators to derive from std::iterator.
+    /// A user-defined iterator should provide publicly accessible typedefs named
+    /// iterator_category, value_type, difference_type, pointer, and reference.
+    /// Note that value_type is required to be non-const, even for constant iterators.
+    using iterator_category = std::bidirectional_iterator_tag;
+
+    /// the type of the values when the iterator is dereferenced
+    using value_type = typename BasicJsonType::value_type;
+    /// a type to represent differences between iterators
+    using difference_type = typename BasicJsonType::difference_type;
+    /// defines a pointer to the type iterated over (value_type)
+    using pointer = typename std::conditional<std::is_const<BasicJsonType>::value,
+          typename BasicJsonType::const_pointer,
+          typename BasicJsonType::pointer>::type;
+    /// defines a reference to the type iterated over (value_type)
+    using reference =
+        typename std::conditional<std::is_const<BasicJsonType>::value,
+        typename BasicJsonType::const_reference,
+        typename BasicJsonType::reference>::type;
+
+    iter_impl() = default;
+    ~iter_impl() = default;
+    iter_impl(iter_impl&&) noexcept = default;
+    iter_impl& operator=(iter_impl&&) noexcept = default;
+
+    /*!
+    @brief constructor for a given JSON instance
+    @param[in] object  pointer to a JSON object for this iterator
+    @pre object != nullptr
+    @post The iterator is initialized; i.e. `m_object != nullptr`.
+    */
+    explicit iter_impl(pointer object) noexcept : m_object(object)
+    {
+        JSON_ASSERT(m_object != nullptr);
+
+        switch (m_object->m_type)
+        {
+            case value_t::object:
+            {
+                m_it.object_iterator = typename object_t::iterator();
+                break;
+            }
+
+            case value_t::array:
+            {
+                m_it.array_iterator = typename array_t::iterator();
+                break;
+            }
+
+            case value_t::null:
+            case value_t::string:
+            case value_t::boolean:
+            case value_t::number_integer:
+            case value_t::number_unsigned:
+            case value_t::number_float:
+            case value_t::binary:
+            case value_t::discarded:
+            default:
+            {
+                m_it.primitive_iterator = primitive_iterator_t();
+                break;
+            }
+        }
+    }
+
+    /*!
+    @note The conventional copy constructor and copy assignment are implicitly
+          defined. Combined with the following converting constructor and
+          assignment, they support: (1) copy from iterator to iterator, (2)
+          copy from const iterator to const iterator, and (3) conversion from
+          iterator to const iterator. However conversion from const iterator
+          to iterator is not defined.
+    */
+
+    /*!
+    @brief const copy constructor
+    @param[in] other const iterator to copy from
+    @note This copy constructor had to be defined explicitly to circumvent a bug
+          occurring on msvc v19.0 compiler (VS 2015) debug build. For more
+          information refer to: https://github.com/nlohmann/json/issues/1608
+    */
+    iter_impl(const iter_impl<const BasicJsonType>& other) noexcept
+        : m_object(other.m_object), m_it(other.m_it)
+    {}
+
+    /*!
+    @brief converting assignment
+    @param[in] other const iterator to copy from
+    @return const/non-const iterator
+    @note It is not checked whether @a other is initialized.
+    */
+    iter_impl& operator=(const iter_impl<const BasicJsonType>& other) noexcept
+    {
+        if (&other != this)
+        {
+            m_object = other.m_object;
+            m_it = other.m_it;
+        }
+        return *this;
+    }
+
+    /*!
+    @brief converting constructor
+    @param[in] other  non-const iterator to copy from
+    @note It is not checked whether @a other is initialized.
+    */
+    iter_impl(const iter_impl<typename std::remove_const<BasicJsonType>::type>& other) noexcept
+        : m_object(other.m_object), m_it(other.m_it)
+    {}
+
+    /*!
+    @brief converting assignment
+    @param[in] other  non-const iterator to copy from
+    @return const/non-const iterator
+    @note It is not checked whether @a other is initialized.
+    */
+    iter_impl& operator=(const iter_impl<typename std::remove_const<BasicJsonType>::type>& other) noexcept // NOLINT(cert-oop54-cpp)
+    {
+        m_object = other.m_object;
+        m_it = other.m_it;
+        return *this;
+    }
+
+  JSON_PRIVATE_UNLESS_TESTED:
+    /*!
+    @brief set the iterator to the first value
+    @pre The iterator is initialized; i.e. `m_object != nullptr`.
+    */
+    void set_begin() noexcept
+    {
+        JSON_ASSERT(m_object != nullptr);
+
+        switch (m_object->m_type)
+        {
+            case value_t::object:
+            {
+                m_it.object_iterator = m_object->m_value.object->begin();
+                break;
+            }
+
+            case value_t::array:
+            {
+                m_it.array_iterator = m_object->m_value.array->begin();
+                break;
+            }
+
+            case value_t::null:
+            {
+                // set to end so begin()==end() is true: null is empty
+                m_it.primitive_iterator.set_end();
+                break;
+            }
+
+            case value_t::string:
+            case value_t::boolean:
+            case value_t::number_integer:
+            case value_t::number_unsigned:
+            case value_t::number_float:
+            case value_t::binary:
+            case value_t::discarded:
+            default:
+            {
+                m_it.primitive_iterator.set_begin();
+                break;
+            }
+        }
+    }
+
+    /*!
+    @brief set the iterator past the last value
+    @pre The iterator is initialized; i.e. `m_object != nullptr`.
+    */
+    void set_end() noexcept
+    {
+        JSON_ASSERT(m_object != nullptr);
+
+        switch (m_object->m_type)
+        {
+            case value_t::object:
+            {
+                m_it.object_iterator = m_object->m_value.object->end();
+                break;
+            }
+
+            case value_t::array:
+            {
+                m_it.array_iterator = m_object->m_value.array->end();
+                break;
+            }
+
+            case value_t::null:
+            case value_t::string:
+            case value_t::boolean:
+            case value_t::number_integer:
+            case value_t::number_unsigned:
+            case value_t::number_float:
+            case value_t::binary:
+            case value_t::discarded:
+            default:
+            {
+                m_it.primitive_iterator.set_end();
+                break;
+            }
+        }
+    }
+
+  public:
+    /*!
+    @brief return a reference to the value pointed to by the iterator
+    @pre The iterator is initialized; i.e. `m_object != nullptr`.
+    */
+    reference operator*() const
+    {
+        JSON_ASSERT(m_object != nullptr);
+
+        switch (m_object->m_type)
+        {
+            case value_t::object:
+            {
+                JSON_ASSERT(m_it.object_iterator != m_object->m_value.object->end());
+                return m_it.object_iterator->second;
+            }
+
+            case value_t::array:
+            {
+                JSON_ASSERT(m_it.array_iterator != m_object->m_value.array->end());
+                return *m_it.array_iterator;
+            }
+
+            case value_t::null:
+                JSON_THROW(invalid_iterator::create(214, "cannot get value", *m_object));
+
+            case value_t::string:
+            case value_t::boolean:
+            case value_t::number_integer:
+            case value_t::number_unsigned:
+            case value_t::number_float:
+            case value_t::binary:
+            case value_t::discarded:
+            default:
+            {
+                if (JSON_HEDLEY_LIKELY(m_it.primitive_iterator.is_begin()))
+                {
+                    return *m_object;
+                }
+
+                JSON_THROW(invalid_iterator::create(214, "cannot get value", *m_object));
+            }
+        }
+    }
+
+    /*!
+    @brief dereference the iterator
+    @pre The iterator is initialized; i.e. `m_object != nullptr`.
+    */
+    pointer operator->() const
+    {
+        JSON_ASSERT(m_object != nullptr);
+
+        switch (m_object->m_type)
+        {
+            case value_t::object:
+            {
+                JSON_ASSERT(m_it.object_iterator != m_object->m_value.object->end());
+                return &(m_it.object_iterator->second);
+            }
+
+            case value_t::array:
+            {
+                JSON_ASSERT(m_it.array_iterator != m_object->m_value.array->end());
+                return &*m_it.array_iterator;
+            }
+
+            case value_t::null:
+            case value_t::string:
+            case value_t::boolean:
+            case value_t::number_integer:
+            case value_t::number_unsigned:
+            case value_t::number_float:
+            case value_t::binary:
+            case value_t::discarded:
+            default:
+            {
+                if (JSON_HEDLEY_LIKELY(m_it.primitive_iterator.is_begin()))
+                {
+                    return m_object;
+                }
+
+                JSON_THROW(invalid_iterator::create(214, "cannot get value", *m_object));
+            }
+        }
+    }
+
+    /*!
+    @brief post-increment (it++)
+    @pre The iterator is initialized; i.e. `m_object != nullptr`.
+    */
+    iter_impl const operator++(int) // NOLINT(readability-const-return-type)
+    {
+        auto result = *this;
+        ++(*this);
+        return result;
+    }
+
+    /*!
+    @brief pre-increment (++it)
+    @pre The iterator is initialized; i.e. `m_object != nullptr`.
+    */
+    iter_impl& operator++()
+    {
+        JSON_ASSERT(m_object != nullptr);
+
+        switch (m_object->m_type)
+        {
+            case value_t::object:
+            {
+                std::advance(m_it.object_iterator, 1);
+                break;
+            }
+
+            case value_t::array:
+            {
+                std::advance(m_it.array_iterator, 1);
+                break;
+            }
+
+            case value_t::null:
+            case value_t::string:
+            case value_t::boolean:
+            case value_t::number_integer:
+            case value_t::number_unsigned:
+            case value_t::number_float:
+            case value_t::binary:
+            case value_t::discarded:
+            default:
+            {
+                ++m_it.primitive_iterator;
+                break;
+            }
+        }
+
+        return *this;
+    }
+
+    /*!
+    @brief post-decrement (it--)
+    @pre The iterator is initialized; i.e. `m_object != nullptr`.
+    */
+    iter_impl const operator--(int) // NOLINT(readability-const-return-type)
+    {
+        auto result = *this;
+        --(*this);
+        return result;
+    }
+
+    /*!
+    @brief pre-decrement (--it)
+    @pre The iterator is initialized; i.e. `m_object != nullptr`.
+    */
+    iter_impl& operator--()
+    {
+        JSON_ASSERT(m_object != nullptr);
+
+        switch (m_object->m_type)
+        {
+            case value_t::object:
+            {
+                std::advance(m_it.object_iterator, -1);
+                break;
+            }
+
+            case value_t::array:
+            {
+                std::advance(m_it.array_iterator, -1);
+                break;
+            }
+
+            case value_t::null:
+            case value_t::string:
+            case value_t::boolean:
+            case value_t::number_integer:
+            case value_t::number_unsigned:
+            case value_t::number_float:
+            case value_t::binary:
+            case value_t::discarded:
+            default:
+            {
+                --m_it.primitive_iterator;
+                break;
+            }
+        }
+
+        return *this;
+    }
+
+    /*!
+    @brief comparison: equal
+    @pre The iterator is initialized; i.e. `m_object != nullptr`.
+    */
+    template < typename IterImpl, detail::enable_if_t < (std::is_same<IterImpl, iter_impl>::value || std::is_same<IterImpl, other_iter_impl>::value), std::nullptr_t > = nullptr >
+    bool operator==(const IterImpl& other) const
+    {
+        // if objects are not the same, the comparison is undefined
+        if (JSON_HEDLEY_UNLIKELY(m_object != other.m_object))
+        {
+            JSON_THROW(invalid_iterator::create(212, "cannot compare iterators of different containers", *m_object));
+        }
+
+        JSON_ASSERT(m_object != nullptr);
+
+        switch (m_object->m_type)
+        {
+            case value_t::object:
+                return (m_it.object_iterator == other.m_it.object_iterator);
+
+            case value_t::array:
+                return (m_it.array_iterator == other.m_it.array_iterator);
+
+            case value_t::null:
+            case value_t::string:
+            case value_t::boolean:
+            case value_t::number_integer:
+            case value_t::number_unsigned:
+            case value_t::number_float:
+            case value_t::binary:
+            case value_t::discarded:
+            default:
+                return (m_it.primitive_iterator == other.m_it.primitive_iterator);
+        }
+    }
+
+    /*!
+    @brief comparison: not equal
+    @pre The iterator is initialized; i.e. `m_object != nullptr`.
+    */
+    template < typename IterImpl, detail::enable_if_t < (std::is_same<IterImpl, iter_impl>::value || std::is_same<IterImpl, other_iter_impl>::value), std::nullptr_t > = nullptr >
+    bool operator!=(const IterImpl& other) const
+    {
+        return !operator==(other);
+    }
+
+    /*!
+    @brief comparison: smaller
+    @pre The iterator is initialized; i.e. `m_object != nullptr`.
+    */
+    bool operator<(const iter_impl& other) const
+    {
+        // if objects are not the same, the comparison is undefined
+        if (JSON_HEDLEY_UNLIKELY(m_object != other.m_object))
+        {
+            JSON_THROW(invalid_iterator::create(212, "cannot compare iterators of different containers", *m_object));
+        }
+
+        JSON_ASSERT(m_object != nullptr);
+
+        switch (m_object->m_type)
+        {
+            case value_t::object:
+                JSON_THROW(invalid_iterator::create(213, "cannot compare order of object iterators", *m_object));
+
+            case value_t::array:
+                return (m_it.array_iterator < other.m_it.array_iterator);
+
+            case value_t::null:
+            case value_t::string:
+            case value_t::boolean:
+            case value_t::number_integer:
+            case value_t::number_unsigned:
+            case value_t::number_float:
+            case value_t::binary:
+            case value_t::discarded:
+            default:
+                return (m_it.primitive_iterator < other.m_it.primitive_iterator);
+        }
+    }
+
+    /*!
+    @brief comparison: less than or equal
+    @pre The iterator is initialized; i.e. `m_object != nullptr`.
+    */
+    bool operator<=(const iter_impl& other) const
+    {
+        return !other.operator < (*this);
+    }
+
+    /*!
+    @brief comparison: greater than
+    @pre The iterator is initialized; i.e. `m_object != nullptr`.
+    */
+    bool operator>(const iter_impl& other) const
+    {
+        return !operator<=(other);
+    }
+
+    /*!
+    @brief comparison: greater than or equal
+    @pre The iterator is initialized; i.e. `m_object != nullptr`.
+    */
+    bool operator>=(const iter_impl& other) const
+    {
+        return !operator<(other);
+    }
+
+    /*!
+    @brief add to iterator
+    @pre The iterator is initialized; i.e. `m_object != nullptr`.
+    */
+    iter_impl& operator+=(difference_type i)
+    {
+        JSON_ASSERT(m_object != nullptr);
+
+        switch (m_object->m_type)
+        {
+            case value_t::object:
+                JSON_THROW(invalid_iterator::create(209, "cannot use offsets with object iterators", *m_object));
+
+            case value_t::array:
+            {
+                std::advance(m_it.array_iterator, i);
+                break;
+            }
+
+            case value_t::null:
+            case value_t::string:
+            case value_t::boolean:
+            case value_t::number_integer:
+            case value_t::number_unsigned:
+            case value_t::number_float:
+            case value_t::binary:
+            case value_t::discarded:
+            default:
+            {
+                m_it.primitive_iterator += i;
+                break;
+            }
+        }
+
+        return *this;
+    }
+
+    /*!
+    @brief subtract from iterator
+    @pre The iterator is initialized; i.e. `m_object != nullptr`.
+    */
+    iter_impl& operator-=(difference_type i)
+    {
+        return operator+=(-i);
+    }
+
+    /*!
+    @brief add to iterator
+    @pre The iterator is initialized; i.e. `m_object != nullptr`.
+    */
+    iter_impl operator+(difference_type i) const
+    {
+        auto result = *this;
+        result += i;
+        return result;
+    }
+
+    /*!
+    @brief addition of distance and iterator
+    @pre The iterator is initialized; i.e. `m_object != nullptr`.
+    */
+    friend iter_impl operator+(difference_type i, const iter_impl& it)
+    {
+        auto result = it;
+        result += i;
+        return result;
+    }
+
+    /*!
+    @brief subtract from iterator
+    @pre The iterator is initialized; i.e. `m_object != nullptr`.
+    */
+    iter_impl operator-(difference_type i) const
+    {
+        auto result = *this;
+        result -= i;
+        return result;
+    }
+
+    /*!
+    @brief return difference
+    @pre The iterator is initialized; i.e. `m_object != nullptr`.
+    */
+    difference_type operator-(const iter_impl& other) const
+    {
+        JSON_ASSERT(m_object != nullptr);
+
+        switch (m_object->m_type)
+        {
+            case value_t::object:
+                JSON_THROW(invalid_iterator::create(209, "cannot use offsets with object iterators", *m_object));
+
+            case value_t::array:
+                return m_it.array_iterator - other.m_it.array_iterator;
+
+            case value_t::null:
+            case value_t::string:
+            case value_t::boolean:
+            case value_t::number_integer:
+            case value_t::number_unsigned:
+            case value_t::number_float:
+            case value_t::binary:
+            case value_t::discarded:
+            default:
+                return m_it.primitive_iterator - other.m_it.primitive_iterator;
+        }
+    }
+
+    /*!
+    @brief access to successor
+    @pre The iterator is initialized; i.e. `m_object != nullptr`.
+    */
+    reference operator[](difference_type n) const
+    {
+        JSON_ASSERT(m_object != nullptr);
+
+        switch (m_object->m_type)
+        {
+            case value_t::object:
+                JSON_THROW(invalid_iterator::create(208, "cannot use operator[] for object iterators", *m_object));
+
+            case value_t::array:
+                return *std::next(m_it.array_iterator, n);
+
+            case value_t::null:
+                JSON_THROW(invalid_iterator::create(214, "cannot get value", *m_object));
+
+            case value_t::string:
+            case value_t::boolean:
+            case value_t::number_integer:
+            case value_t::number_unsigned:
+            case value_t::number_float:
+            case value_t::binary:
+            case value_t::discarded:
+            default:
+            {
+                if (JSON_HEDLEY_LIKELY(m_it.primitive_iterator.get_value() == -n))
+                {
+                    return *m_object;
+                }
+
+                JSON_THROW(invalid_iterator::create(214, "cannot get value", *m_object));
+            }
+        }
+    }
+
+    /*!
+    @brief return the key of an object iterator
+    @pre The iterator is initialized; i.e. `m_object != nullptr`.
+    */
+    const typename object_t::key_type& key() const
+    {
+        JSON_ASSERT(m_object != nullptr);
+
+        if (JSON_HEDLEY_LIKELY(m_object->is_object()))
+        {
+            return m_it.object_iterator->first;
+        }
+
+        JSON_THROW(invalid_iterator::create(207, "cannot use key() for non-object iterators", *m_object));
+    }
+
+    /*!
+    @brief return the value of an iterator
+    @pre The iterator is initialized; i.e. `m_object != nullptr`.
+    */
+    reference value() const
+    {
+        return operator*();
+    }
+
+  JSON_PRIVATE_UNLESS_TESTED:
+    /// associated JSON instance
+    pointer m_object = nullptr;
+    /// the actual iterator of the associated instance
+    internal_iterator<typename std::remove_const<BasicJsonType>::type> m_it {};
+};
+} // namespace detail
+} // namespace nlohmann
+
+// #include <nlohmann/detail/iterators/iteration_proxy.hpp>
+
+// #include <nlohmann/detail/iterators/json_reverse_iterator.hpp>
+
+
+#include <cstddef> // ptrdiff_t
+#include <iterator> // reverse_iterator
+#include <utility> // declval
+
+namespace nlohmann
+{
+namespace detail
+{
+//////////////////////
+// reverse_iterator //
+//////////////////////
+
+/*!
+@brief a template for a reverse iterator class
+
+@tparam Base the base iterator type to reverse. Valid types are @ref
+iterator (to create @ref reverse_iterator) and @ref const_iterator (to
+create @ref const_reverse_iterator).
+
+@requirement The class satisfies the following concept requirements:
+-
+[BidirectionalIterator](https://en.cppreference.com/w/cpp/named_req/BidirectionalIterator):
+  The iterator that can be moved can be moved in both directions (i.e.
+  incremented and decremented).
+- [OutputIterator](https://en.cppreference.com/w/cpp/named_req/OutputIterator):
+  It is possible to write to the pointed-to element (only if @a Base is
+  @ref iterator).
+
+@since version 1.0.0
+*/
+template<typename Base>
+class json_reverse_iterator : public std::reverse_iterator<Base>
+{
+  public:
+    using difference_type = std::ptrdiff_t;
+    /// shortcut to the reverse iterator adapter
+    using base_iterator = std::reverse_iterator<Base>;
+    /// the reference type for the pointed-to element
+    using reference = typename Base::reference;
+
+    /// create reverse iterator from iterator
+    explicit json_reverse_iterator(const typename base_iterator::iterator_type& it) noexcept
+        : base_iterator(it) {}
+
+    /// create reverse iterator from base class
+    explicit json_reverse_iterator(const base_iterator& it) noexcept : base_iterator(it) {}
+
+    /// post-increment (it++)
+    json_reverse_iterator const operator++(int) // NOLINT(readability-const-return-type)
+    {
+        return static_cast<json_reverse_iterator>(base_iterator::operator++(1));
+    }
+
+    /// pre-increment (++it)
+    json_reverse_iterator& operator++()
+    {
+        return static_cast<json_reverse_iterator&>(base_iterator::operator++());
+    }
+
+    /// post-decrement (it--)
+    json_reverse_iterator const operator--(int) // NOLINT(readability-const-return-type)
+    {
+        return static_cast<json_reverse_iterator>(base_iterator::operator--(1));
+    }
+
+    /// pre-decrement (--it)
+    json_reverse_iterator& operator--()
+    {
+        return static_cast<json_reverse_iterator&>(base_iterator::operator--());
+    }
+
+    /// add to iterator
+    json_reverse_iterator& operator+=(difference_type i)
+    {
+        return static_cast<json_reverse_iterator&>(base_iterator::operator+=(i));
+    }
+
+    /// add to iterator
+    json_reverse_iterator operator+(difference_type i) const
+    {
+        return static_cast<json_reverse_iterator>(base_iterator::operator+(i));
+    }
+
+    /// subtract from iterator
+    json_reverse_iterator operator-(difference_type i) const
+    {
+        return static_cast<json_reverse_iterator>(base_iterator::operator-(i));
+    }
+
+    /// return difference
+    difference_type operator-(const json_reverse_iterator& other) const
+    {
+        return base_iterator(*this) - base_iterator(other);
+    }
+
+    /// access to successor
+    reference operator[](difference_type n) const
+    {
+        return *(this->operator+(n));
+    }
+
+    /// return the key of an object iterator
+    auto key() const -> decltype(std::declval<Base>().key())
+    {
+        auto it = --this->base();
+        return it.key();
+    }
+
+    /// return the value of an iterator
+    reference value() const
+    {
+        auto it = --this->base();
+        return it.operator * ();
+    }
+};
+}  // namespace detail
+}  // namespace nlohmann
+
+// #include <nlohmann/detail/iterators/primitive_iterator.hpp>
+
+// #include <nlohmann/detail/json_pointer.hpp>
+
+
+#include <algorithm> // all_of
+#include <cctype> // isdigit
+#include <limits> // max
+#include <numeric> // accumulate
+#include <string> // string
+#include <utility> // move
+#include <vector> // vector
+
+// #include <nlohmann/detail/exceptions.hpp>
+
+// #include <nlohmann/detail/macro_scope.hpp>
+
+// #include <nlohmann/detail/string_escape.hpp>
+
+// #include <nlohmann/detail/value_t.hpp>
+
+
+namespace nlohmann
+{
+
+/// @brief JSON Pointer defines a string syntax for identifying a specific value within a JSON document
+/// @sa https://json.nlohmann.me/api/json_pointer/
+template<typename BasicJsonType>
+class json_pointer
+{
+    // allow basic_json to access private members
+    NLOHMANN_BASIC_JSON_TPL_DECLARATION
+    friend class basic_json;
+
+  public:
+    /// @brief create JSON pointer
+    /// @sa https://json.nlohmann.me/api/json_pointer/json_pointer/
+    explicit json_pointer(const std::string& s = "")
+        : reference_tokens(split(s))
+    {}
+
+    /// @brief return a string representation of the JSON pointer
+    /// @sa https://json.nlohmann.me/api/json_pointer/to_string/
+    std::string to_string() const
+    {
+        return std::accumulate(reference_tokens.begin(), reference_tokens.end(),
+                               std::string{},
+                               [](const std::string & a, const std::string & b)
+        {
+            return a + "/" + detail::escape(b);
+        });
+    }
+
+    /// @brief return a string representation of the JSON pointer
+    /// @sa https://json.nlohmann.me/api/json_pointer/operator_string/
+    operator std::string() const
+    {
+        return to_string();
+    }
+
+    /// @brief append another JSON pointer at the end of this JSON pointer
+    /// @sa https://json.nlohmann.me/api/json_pointer/operator_slasheq/
+    json_pointer& operator/=(const json_pointer& ptr)
+    {
+        reference_tokens.insert(reference_tokens.end(),
+                                ptr.reference_tokens.begin(),
+                                ptr.reference_tokens.end());
+        return *this;
+    }
+
+    /// @brief append an unescaped reference token at the end of this JSON pointer
+    /// @sa https://json.nlohmann.me/api/json_pointer/operator_slasheq/
+    json_pointer& operator/=(std::string token)
+    {
+        push_back(std::move(token));
+        return *this;
+    }
+
+    /// @brief append an array index at the end of this JSON pointer
+    /// @sa https://json.nlohmann.me/api/json_pointer/operator_slasheq/
+    json_pointer& operator/=(std::size_t array_idx)
+    {
+        return *this /= std::to_string(array_idx);
+    }
+
+    /// @brief create a new JSON pointer by appending the right JSON pointer at the end of the left JSON pointer
+    /// @sa https://json.nlohmann.me/api/json_pointer/operator_slash/
+    friend json_pointer operator/(const json_pointer& lhs,
+                                  const json_pointer& rhs)
+    {
+        return json_pointer(lhs) /= rhs;
+    }
+
+    /// @brief create a new JSON pointer by appending the unescaped token at the end of the JSON pointer
+    /// @sa https://json.nlohmann.me/api/json_pointer/operator_slash/
+    friend json_pointer operator/(const json_pointer& lhs, std::string token) // NOLINT(performance-unnecessary-value-param)
+    {
+        return json_pointer(lhs) /= std::move(token);
+    }
+
+    /// @brief create a new JSON pointer by appending the array-index-token at the end of the JSON pointer
+    /// @sa https://json.nlohmann.me/api/json_pointer/operator_slash/
+    friend json_pointer operator/(const json_pointer& lhs, std::size_t array_idx)
+    {
+        return json_pointer(lhs) /= array_idx;
+    }
+
+    /// @brief returns the parent of this JSON pointer
+    /// @sa https://json.nlohmann.me/api/json_pointer/parent_pointer/
+    json_pointer parent_pointer() const
+    {
+        if (empty())
+        {
+            return *this;
+        }
+
+        json_pointer res = *this;
+        res.pop_back();
+        return res;
+    }
+
+    /// @brief remove last reference token
+    /// @sa https://json.nlohmann.me/api/json_pointer/pop_back/
+    void pop_back()
+    {
+        if (JSON_HEDLEY_UNLIKELY(empty()))
+        {
+            JSON_THROW(detail::out_of_range::create(405, "JSON pointer has no parent", BasicJsonType()));
+        }
+
+        reference_tokens.pop_back();
+    }
+
+    /// @brief return last reference token
+    /// @sa https://json.nlohmann.me/api/json_pointer/back/
+    const std::string& back() const
+    {
+        if (JSON_HEDLEY_UNLIKELY(empty()))
+        {
+            JSON_THROW(detail::out_of_range::create(405, "JSON pointer has no parent", BasicJsonType()));
+        }
+
+        return reference_tokens.back();
+    }
+
+    /// @brief append an unescaped token at the end of the reference pointer
+    /// @sa https://json.nlohmann.me/api/json_pointer/push_back/
+    void push_back(const std::string& token)
+    {
+        reference_tokens.push_back(token);
+    }
+
+    /// @brief append an unescaped token at the end of the reference pointer
+    /// @sa https://json.nlohmann.me/api/json_pointer/push_back/
+    void push_back(std::string&& token)
+    {
+        reference_tokens.push_back(std::move(token));
+    }
+
+    /// @brief return whether pointer points to the root document
+    /// @sa https://json.nlohmann.me/api/json_pointer/empty/
+    bool empty() const noexcept
+    {
+        return reference_tokens.empty();
+    }
+
+  private:
+    /*!
+    @param[in] s  reference token to be converted into an array index
+
+    @return integer representation of @a s
+
+    @throw parse_error.106  if an array index begins with '0'
+    @throw parse_error.109  if an array index begins not with a digit
+    @throw out_of_range.404 if string @a s could not be converted to an integer
+    @throw out_of_range.410 if an array index exceeds size_type
+    */
+    static typename BasicJsonType::size_type array_index(const std::string& s)
+    {
+        using size_type = typename BasicJsonType::size_type;
+
+        // error condition (cf. RFC 6901, Sect. 4)
+        if (JSON_HEDLEY_UNLIKELY(s.size() > 1 && s[0] == '0'))
+        {
+            JSON_THROW(detail::parse_error::create(106, 0, "array index '" + s + "' must not begin with '0'", BasicJsonType()));
+        }
+
+        // error condition (cf. RFC 6901, Sect. 4)
+        if (JSON_HEDLEY_UNLIKELY(s.size() > 1 && !(s[0] >= '1' && s[0] <= '9')))
+        {
+            JSON_THROW(detail::parse_error::create(109, 0, "array index '" + s + "' is not a number", BasicJsonType()));
+        }
+
+        std::size_t processed_chars = 0;
+        unsigned long long res = 0;  // NOLINT(runtime/int)
+        JSON_TRY
+        {
+            res = std::stoull(s, &processed_chars);
+        }
+        JSON_CATCH(std::out_of_range&)
+        {
+            JSON_THROW(detail::out_of_range::create(404, "unresolved reference token '" + s + "'", BasicJsonType()));
+        }
+
+        // check if the string was completely read
+        if (JSON_HEDLEY_UNLIKELY(processed_chars != s.size()))
+        {
+            JSON_THROW(detail::out_of_range::create(404, "unresolved reference token '" + s + "'", BasicJsonType()));
+        }
+
+        // only triggered on special platforms (like 32bit), see also
+        // https://github.com/nlohmann/json/pull/2203
+        if (res >= static_cast<unsigned long long>((std::numeric_limits<size_type>::max)()))  // NOLINT(runtime/int)
+        {
+            JSON_THROW(detail::out_of_range::create(410, "array index " + s + " exceeds size_type", BasicJsonType())); // LCOV_EXCL_LINE
+        }
+
+        return static_cast<size_type>(res);
+    }
+
+  JSON_PRIVATE_UNLESS_TESTED:
+    json_pointer top() const
+    {
+        if (JSON_HEDLEY_UNLIKELY(empty()))
+        {
+            JSON_THROW(detail::out_of_range::create(405, "JSON pointer has no parent", BasicJsonType()));
+        }
+
+        json_pointer result = *this;
+        result.reference_tokens = {reference_tokens[0]};
+        return result;
+    }
+
+  private:
+    /*!
+    @brief create and return a reference to the pointed to value
+
+    @complexity Linear in the number of reference tokens.
+
+    @throw parse_error.109 if array index is not a number
+    @throw type_error.313 if value cannot be unflattened
+    */
+    BasicJsonType& get_and_create(BasicJsonType& j) const
+    {
+        auto* result = &j;
+
+        // in case no reference tokens exist, return a reference to the JSON value
+        // j which will be overwritten by a primitive value
+        for (const auto& reference_token : reference_tokens)
+        {
+            switch (result->type())
+            {
+                case detail::value_t::null:
+                {
+                    if (reference_token == "0")
+                    {
+                        // start a new array if reference token is 0
+                        result = &result->operator[](0);
+                    }
+                    else
+                    {
+                        // start a new object otherwise
+                        result = &result->operator[](reference_token);
+                    }
+                    break;
+                }
+
+                case detail::value_t::object:
+                {
+                    // create an entry in the object
+                    result = &result->operator[](reference_token);
+                    break;
+                }
+
+                case detail::value_t::array:
+                {
+                    // create an entry in the array
+                    result = &result->operator[](array_index(reference_token));
+                    break;
+                }
+
+                /*
+                The following code is only reached if there exists a reference
+                token _and_ the current value is primitive. In this case, we have
+                an error situation, because primitive values may only occur as
+                single value; that is, with an empty list of reference tokens.
+                */
+                case detail::value_t::string:
+                case detail::value_t::boolean:
+                case detail::value_t::number_integer:
+                case detail::value_t::number_unsigned:
+                case detail::value_t::number_float:
+                case detail::value_t::binary:
+                case detail::value_t::discarded:
+                default:
+                    JSON_THROW(detail::type_error::create(313, "invalid value to unflatten", j));
+            }
+        }
+
+        return *result;
+    }
+
+    /*!
+    @brief return a reference to the pointed to value
+
+    @note This version does not throw if a value is not present, but tries to
+          create nested values instead. For instance, calling this function
+          with pointer `"/this/that"` on a null value is equivalent to calling
+          `operator[]("this").operator[]("that")` on that value, effectively
+          changing the null value to an object.
+
+    @param[in] ptr  a JSON value
+
+    @return reference to the JSON value pointed to by the JSON pointer
+
+    @complexity Linear in the length of the JSON pointer.
+
+    @throw parse_error.106   if an array index begins with '0'
+    @throw parse_error.109   if an array index was not a number
+    @throw out_of_range.404  if the JSON pointer can not be resolved
+    */
+    BasicJsonType& get_unchecked(BasicJsonType* ptr) const
+    {
+        for (const auto& reference_token : reference_tokens)
+        {
+            // convert null values to arrays or objects before continuing
+            if (ptr->is_null())
+            {
+                // check if reference token is a number
+                const bool nums =
+                    std::all_of(reference_token.begin(), reference_token.end(),
+                                [](const unsigned char x)
+                {
+                    return std::isdigit(x);
+                });
+
+                // change value to array for numbers or "-" or to object otherwise
+                *ptr = (nums || reference_token == "-")
+                       ? detail::value_t::array
+                       : detail::value_t::object;
+            }
+
+            switch (ptr->type())
+            {
+                case detail::value_t::object:
+                {
+                    // use unchecked object access
+                    ptr = &ptr->operator[](reference_token);
+                    break;
+                }
+
+                case detail::value_t::array:
+                {
+                    if (reference_token == "-")
+                    {
+                        // explicitly treat "-" as index beyond the end
+                        ptr = &ptr->operator[](ptr->m_value.array->size());
+                    }
+                    else
+                    {
+                        // convert array index to number; unchecked access
+                        ptr = &ptr->operator[](array_index(reference_token));
+                    }
+                    break;
+                }
+
+                case detail::value_t::null:
+                case detail::value_t::string:
+                case detail::value_t::boolean:
+                case detail::value_t::number_integer:
+                case detail::value_t::number_unsigned:
+                case detail::value_t::number_float:
+                case detail::value_t::binary:
+                case detail::value_t::discarded:
+                default:
+                    JSON_THROW(detail::out_of_range::create(404, "unresolved reference token '" + reference_token + "'", *ptr));
+            }
+        }
+
+        return *ptr;
+    }
+
+    /*!
+    @throw parse_error.106   if an array index begins with '0'
+    @throw parse_error.109   if an array index was not a number
+    @throw out_of_range.402  if the array index '-' is used
+    @throw out_of_range.404  if the JSON pointer can not be resolved
+    */
+    BasicJsonType& get_checked(BasicJsonType* ptr) const
+    {
+        for (const auto& reference_token : reference_tokens)
+        {
+            switch (ptr->type())
+            {
+                case detail::value_t::object:
+                {
+                    // note: at performs range check
+                    ptr = &ptr->at(reference_token);
+                    break;
+                }
+
+                case detail::value_t::array:
+                {
+                    if (JSON_HEDLEY_UNLIKELY(reference_token == "-"))
+                    {
+                        // "-" always fails the range check
+                        JSON_THROW(detail::out_of_range::create(402,
+                                                                "array index '-' (" + std::to_string(ptr->m_value.array->size()) +
+                                                                ") is out of range", *ptr));
+                    }
+
+                    // note: at performs range check
+                    ptr = &ptr->at(array_index(reference_token));
+                    break;
+                }
+
+                case detail::value_t::null:
+                case detail::value_t::string:
+                case detail::value_t::boolean:
+                case detail::value_t::number_integer:
+                case detail::value_t::number_unsigned:
+                case detail::value_t::number_float:
+                case detail::value_t::binary:
+                case detail::value_t::discarded:
+                default:
+                    JSON_THROW(detail::out_of_range::create(404, "unresolved reference token '" + reference_token + "'", *ptr));
+            }
+        }
+
+        return *ptr;
+    }
+
+    /*!
+    @brief return a const reference to the pointed to value
+
+    @param[in] ptr  a JSON value
+
+    @return const reference to the JSON value pointed to by the JSON
+    pointer
+
+    @throw parse_error.106   if an array index begins with '0'
+    @throw parse_error.109   if an array index was not a number
+    @throw out_of_range.402  if the array index '-' is used
+    @throw out_of_range.404  if the JSON pointer can not be resolved
+    */
+    const BasicJsonType& get_unchecked(const BasicJsonType* ptr) const
+    {
+        for (const auto& reference_token : reference_tokens)
+        {
+            switch (ptr->type())
+            {
+                case detail::value_t::object:
+                {
+                    // use unchecked object access
+                    ptr = &ptr->operator[](reference_token);
+                    break;
+                }
+
+                case detail::value_t::array:
+                {
+                    if (JSON_HEDLEY_UNLIKELY(reference_token == "-"))
+                    {
+                        // "-" cannot be used for const access
+                        JSON_THROW(detail::out_of_range::create(402, "array index '-' (" + std::to_string(ptr->m_value.array->size()) + ") is out of range", *ptr));
+                    }
+
+                    // use unchecked array access
+                    ptr = &ptr->operator[](array_index(reference_token));
+                    break;
+                }
+
+                case detail::value_t::null:
+                case detail::value_t::string:
+                case detail::value_t::boolean:
+                case detail::value_t::number_integer:
+                case detail::value_t::number_unsigned:
+                case detail::value_t::number_float:
+                case detail::value_t::binary:
+                case detail::value_t::discarded:
+                default:
+                    JSON_THROW(detail::out_of_range::create(404, "unresolved reference token '" + reference_token + "'", *ptr));
+            }
+        }
+
+        return *ptr;
+    }
+
+    /*!
+    @throw parse_error.106   if an array index begins with '0'
+    @throw parse_error.109   if an array index was not a number
+    @throw out_of_range.402  if the array index '-' is used
+    @throw out_of_range.404  if the JSON pointer can not be resolved
+    */
+    const BasicJsonType& get_checked(const BasicJsonType* ptr) const
+    {
+        for (const auto& reference_token : reference_tokens)
+        {
+            switch (ptr->type())
+            {
+                case detail::value_t::object:
+                {
+                    // note: at performs range check
+                    ptr = &ptr->at(reference_token);
+                    break;
+                }
+
+                case detail::value_t::array:
+                {
+                    if (JSON_HEDLEY_UNLIKELY(reference_token == "-"))
+                    {
+                        // "-" always fails the range check
+                        JSON_THROW(detail::out_of_range::create(402,
+                                                                "array index '-' (" + std::to_string(ptr->m_value.array->size()) +
+                                                                ") is out of range", *ptr));
+                    }
+
+                    // note: at performs range check
+                    ptr = &ptr->at(array_index(reference_token));
+                    break;
+                }
+
+                case detail::value_t::null:
+                case detail::value_t::string:
+                case detail::value_t::boolean:
+                case detail::value_t::number_integer:
+                case detail::value_t::number_unsigned:
+                case detail::value_t::number_float:
+                case detail::value_t::binary:
+                case detail::value_t::discarded:
+                default:
+                    JSON_THROW(detail::out_of_range::create(404, "unresolved reference token '" + reference_token + "'", *ptr));
+            }
+        }
+
+        return *ptr;
+    }
+
+    /*!
+    @throw parse_error.106   if an array index begins with '0'
+    @throw parse_error.109   if an array index was not a number
+    */
+    bool contains(const BasicJsonType* ptr) const
+    {
+        for (const auto& reference_token : reference_tokens)
+        {
+            switch (ptr->type())
+            {
+                case detail::value_t::object:
+                {
+                    if (!ptr->contains(reference_token))
+                    {
+                        // we did not find the key in the object
+                        return false;
+                    }
+
+                    ptr = &ptr->operator[](reference_token);
+                    break;
+                }
+
+                case detail::value_t::array:
+                {
+                    if (JSON_HEDLEY_UNLIKELY(reference_token == "-"))
+                    {
+                        // "-" always fails the range check
+                        return false;
+                    }
+                    if (JSON_HEDLEY_UNLIKELY(reference_token.size() == 1 && !("0" <= reference_token && reference_token <= "9")))
+                    {
+                        // invalid char
+                        return false;
+                    }
+                    if (JSON_HEDLEY_UNLIKELY(reference_token.size() > 1))
+                    {
+                        if (JSON_HEDLEY_UNLIKELY(!('1' <= reference_token[0] && reference_token[0] <= '9')))
+                        {
+                            // first char should be between '1' and '9'
+                            return false;
+                        }
+                        for (std::size_t i = 1; i < reference_token.size(); i++)
+                        {
+                            if (JSON_HEDLEY_UNLIKELY(!('0' <= reference_token[i] && reference_token[i] <= '9')))
+                            {
+                                // other char should be between '0' and '9'
+                                return false;
+                            }
+                        }
+                    }
+
+                    const auto idx = array_index(reference_token);
+                    if (idx >= ptr->size())
+                    {
+                        // index out of range
+                        return false;
+                    }
+
+                    ptr = &ptr->operator[](idx);
+                    break;
+                }
+
+                case detail::value_t::null:
+                case detail::value_t::string:
+                case detail::value_t::boolean:
+                case detail::value_t::number_integer:
+                case detail::value_t::number_unsigned:
+                case detail::value_t::number_float:
+                case detail::value_t::binary:
+                case detail::value_t::discarded:
+                default:
+                {
+                    // we do not expect primitive values if there is still a
+                    // reference token to process
+                    return false;
+                }
+            }
+        }
+
+        // no reference token left means we found a primitive value
+        return true;
+    }
+
+    /*!
+    @brief split the string input to reference tokens
+
+    @note This function is only called by the json_pointer constructor.
+          All exceptions below are documented there.
+
+    @throw parse_error.107  if the pointer is not empty or begins with '/'
+    @throw parse_error.108  if character '~' is not followed by '0' or '1'
+    */
+    static std::vector<std::string> split(const std::string& reference_string)
+    {
+        std::vector<std::string> result;
+
+        // special case: empty reference string -> no reference tokens
+        if (reference_string.empty())
+        {
+            return result;
+        }
+
+        // check if nonempty reference string begins with slash
+        if (JSON_HEDLEY_UNLIKELY(reference_string[0] != '/'))
+        {
+            JSON_THROW(detail::parse_error::create(107, 1, "JSON pointer must be empty or begin with '/' - was: '" + reference_string + "'", BasicJsonType()));
+        }
+
+        // extract the reference tokens:
+        // - slash: position of the last read slash (or end of string)
+        // - start: position after the previous slash
+        for (
+            // search for the first slash after the first character
+            std::size_t slash = reference_string.find_first_of('/', 1),
+            // set the beginning of the first reference token
+            start = 1;
+            // we can stop if start == 0 (if slash == std::string::npos)
+            start != 0;
+            // set the beginning of the next reference token
+            // (will eventually be 0 if slash == std::string::npos)
+            start = (slash == std::string::npos) ? 0 : slash + 1,
+            // find next slash
+            slash = reference_string.find_first_of('/', start))
+        {
+            // use the text between the beginning of the reference token
+            // (start) and the last slash (slash).
+            auto reference_token = reference_string.substr(start, slash - start);
+
+            // check reference tokens are properly escaped
+            for (std::size_t pos = reference_token.find_first_of('~');
+                    pos != std::string::npos;
+                    pos = reference_token.find_first_of('~', pos + 1))
+            {
+                JSON_ASSERT(reference_token[pos] == '~');
+
+                // ~ must be followed by 0 or 1
+                if (JSON_HEDLEY_UNLIKELY(pos == reference_token.size() - 1 ||
+                                         (reference_token[pos + 1] != '0' &&
+                                          reference_token[pos + 1] != '1')))
+                {
+                    JSON_THROW(detail::parse_error::create(108, 0, "escape character '~' must be followed with '0' or '1'", BasicJsonType()));
+                }
+            }
+
+            // finally, store the reference token
+            detail::unescape(reference_token);
+            result.push_back(reference_token);
+        }
+
+        return result;
+    }
+
+  private:
+    /*!
+    @param[in] reference_string  the reference string to the current value
+    @param[in] value             the value to consider
+    @param[in,out] result        the result object to insert values to
+
+    @note Empty objects or arrays are flattened to `null`.
+    */
+    static void flatten(const std::string& reference_string,
+                        const BasicJsonType& value,
+                        BasicJsonType& result)
+    {
+        switch (value.type())
+        {
+            case detail::value_t::array:
+            {
+                if (value.m_value.array->empty())
+                {
+                    // flatten empty array as null
+                    result[reference_string] = nullptr;
+                }
+                else
+                {
+                    // iterate array and use index as reference string
+                    for (std::size_t i = 0; i < value.m_value.array->size(); ++i)
+                    {
+                        flatten(reference_string + "/" + std::to_string(i),
+                                value.m_value.array->operator[](i), result);
+                    }
+                }
+                break;
+            }
+
+            case detail::value_t::object:
+            {
+                if (value.m_value.object->empty())
+                {
+                    // flatten empty object as null
+                    result[reference_string] = nullptr;
+                }
+                else
+                {
+                    // iterate object and use keys as reference string
+                    for (const auto& element : *value.m_value.object)
+                    {
+                        flatten(reference_string + "/" + detail::escape(element.first), element.second, result);
+                    }
+                }
+                break;
+            }
+
+            case detail::value_t::null:
+            case detail::value_t::string:
+            case detail::value_t::boolean:
+            case detail::value_t::number_integer:
+            case detail::value_t::number_unsigned:
+            case detail::value_t::number_float:
+            case detail::value_t::binary:
+            case detail::value_t::discarded:
+            default:
+            {
+                // add primitive value with its reference string
+                result[reference_string] = value;
+                break;
+            }
+        }
+    }
+
+    /*!
+    @param[in] value  flattened JSON
+
+    @return unflattened JSON
+
+    @throw parse_error.109 if array index is not a number
+    @throw type_error.314  if value is not an object
+    @throw type_error.315  if object values are not primitive
+    @throw type_error.313  if value cannot be unflattened
+    */
+    static BasicJsonType
+    unflatten(const BasicJsonType& value)
+    {
+        if (JSON_HEDLEY_UNLIKELY(!value.is_object()))
+        {
+            JSON_THROW(detail::type_error::create(314, "only objects can be unflattened", value));
+        }
+
+        BasicJsonType result;
+
+        // iterate the JSON object values
+        for (const auto& element : *value.m_value.object)
+        {
+            if (JSON_HEDLEY_UNLIKELY(!element.second.is_primitive()))
+            {
+                JSON_THROW(detail::type_error::create(315, "values in object must be primitive", element.second));
+            }
+
+            // assign value to reference pointed to by JSON pointer; Note that if
+            // the JSON pointer is "" (i.e., points to the whole value), function
+            // get_and_create returns a reference to result itself. An assignment
+            // will then create a primitive value.
+            json_pointer(element.first).get_and_create(result) = element.second;
+        }
+
+        return result;
+    }
+
+    /*!
+    @brief compares two JSON pointers for equality
+
+    @param[in] lhs  JSON pointer to compare
+    @param[in] rhs  JSON pointer to compare
+    @return whether @a lhs is equal to @a rhs
+
+    @complexity Linear in the length of the JSON pointer
+
+    @exceptionsafety No-throw guarantee: this function never throws exceptions.
+    */
+    friend bool operator==(json_pointer const& lhs,
+                           json_pointer const& rhs) noexcept
+    {
+        return lhs.reference_tokens == rhs.reference_tokens;
+    }
+
+    /*!
+    @brief compares two JSON pointers for inequality
+
+    @param[in] lhs  JSON pointer to compare
+    @param[in] rhs  JSON pointer to compare
+    @return whether @a lhs is not equal @a rhs
+
+    @complexity Linear in the length of the JSON pointer
+
+    @exceptionsafety No-throw guarantee: this function never throws exceptions.
+    */
+    friend bool operator!=(json_pointer const& lhs,
+                           json_pointer const& rhs) noexcept
+    {
+        return !(lhs == rhs);
+    }
+
+    /// the reference tokens
+    std::vector<std::string> reference_tokens;
+};
+}  // namespace nlohmann
+
+// #include <nlohmann/detail/json_ref.hpp>
+
+
+#include <initializer_list>
+#include <utility>
+
+// #include <nlohmann/detail/meta/type_traits.hpp>
+
+
+namespace nlohmann
+{
+namespace detail
+{
+template<typename BasicJsonType>
+class json_ref
+{
+  public:
+    using value_type = BasicJsonType;
+
+    json_ref(value_type&& value)
+        : owned_value(std::move(value))
+    {}
+
+    json_ref(const value_type& value)
+        : value_ref(&value)
+    {}
+
+    json_ref(std::initializer_list<json_ref> init)
+        : owned_value(init)
+    {}
+
+    template <
+        class... Args,
+        enable_if_t<std::is_constructible<value_type, Args...>::value, int> = 0 >
+    json_ref(Args && ... args)
+        : owned_value(std::forward<Args>(args)...)
+    {}
+
+    // class should be movable only
+    json_ref(json_ref&&) noexcept = default;
+    json_ref(const json_ref&) = delete;
+    json_ref& operator=(const json_ref&) = delete;
+    json_ref& operator=(json_ref&&) = delete;
+    ~json_ref() = default;
+
+    value_type moved_or_copied() const
+    {
+        if (value_ref == nullptr)
+        {
+            return std::move(owned_value);
+        }
+        return *value_ref;
+    }
+
+    value_type const& operator*() const
+    {
+        return value_ref ? *value_ref : owned_value;
+    }
+
+    value_type const* operator->() const
+    {
+        return &** this;
+    }
+
+  private:
+    mutable value_type owned_value = nullptr;
+    value_type const* value_ref = nullptr;
+};
+}  // namespace detail
+}  // namespace nlohmann
+
+// #include <nlohmann/detail/macro_scope.hpp>
+
+// #include <nlohmann/detail/string_escape.hpp>
+
+// #include <nlohmann/detail/meta/cpp_future.hpp>
+
+// #include <nlohmann/detail/meta/type_traits.hpp>
 
 // #include <nlohmann/detail/output/binary_writer.hpp>
 
 
 #include <algorithm> // reverse
 #include <array> // array
+#include <cmath> // isnan, isinf
 #include <cstdint> // uint8_t, uint16_t, uint32_t, uint64_t
 #include <cstring> // memcpy
 #include <limits> // numeric_limits
+#include <string> // string
+#include <utility> // move
 
 // #include <nlohmann/detail/input/binary_reader.hpp>
 
+// #include <nlohmann/detail/macro_scope.hpp>
+
 // #include <nlohmann/detail/output/output_adapters.hpp>
 
 
+#include <algorithm> // copy
+#include <cstddef> // size_t
+#include <iterator> // back_inserter
+#include <memory> // shared_ptr, make_shared
+#include <string> // basic_string
+#include <vector> // vector
+
+#ifndef JSON_NO_IO
+    #include <ios>      // streamsize
+    #include <ostream>  // basic_ostream
+#endif  // JSON_NO_IO
+
+// #include <nlohmann/detail/macro_scope.hpp>
+
+
+namespace nlohmann
+{
+namespace detail
+{
+/// abstract output adapter interface
+template<typename CharType> struct output_adapter_protocol
+{
+    virtual void write_character(CharType c) = 0;
+    virtual void write_characters(const CharType* s, std::size_t length) = 0;
+    virtual ~output_adapter_protocol() = default;
+
+    output_adapter_protocol() = default;
+    output_adapter_protocol(const output_adapter_protocol&) = default;
+    output_adapter_protocol(output_adapter_protocol&&) noexcept = default;
+    output_adapter_protocol& operator=(const output_adapter_protocol&) = default;
+    output_adapter_protocol& operator=(output_adapter_protocol&&) noexcept = default;
+};
+
+/// a type to simplify interfaces
+template<typename CharType>
+using output_adapter_t = std::shared_ptr<output_adapter_protocol<CharType>>;
+
+/// output adapter for byte vectors
+template<typename CharType, typename AllocatorType = std::allocator<CharType>>
+class output_vector_adapter : public output_adapter_protocol<CharType>
+{
+  public:
+    explicit output_vector_adapter(std::vector<CharType, AllocatorType>& vec) noexcept
+        : v(vec)
+    {}
+
+    void write_character(CharType c) override
+    {
+        v.push_back(c);
+    }
+
+    JSON_HEDLEY_NON_NULL(2)
+    void write_characters(const CharType* s, std::size_t length) override
+    {
+        std::copy(s, s + length, std::back_inserter(v));
+    }
+
+  private:
+    std::vector<CharType, AllocatorType>& v;
+};
+
+#ifndef JSON_NO_IO
+/// output adapter for output streams
+template<typename CharType>
+class output_stream_adapter : public output_adapter_protocol<CharType>
+{
+  public:
+    explicit output_stream_adapter(std::basic_ostream<CharType>& s) noexcept
+        : stream(s)
+    {}
+
+    void write_character(CharType c) override
+    {
+        stream.put(c);
+    }
+
+    JSON_HEDLEY_NON_NULL(2)
+    void write_characters(const CharType* s, std::size_t length) override
+    {
+        stream.write(s, static_cast<std::streamsize>(length));
+    }
+
+  private:
+    std::basic_ostream<CharType>& stream;
+};
+#endif  // JSON_NO_IO
+
+/// output adapter for basic_string
+template<typename CharType, typename StringType = std::basic_string<CharType>>
+class output_string_adapter : public output_adapter_protocol<CharType>
+{
+  public:
+    explicit output_string_adapter(StringType& s) noexcept
+        : str(s)
+    {}
+
+    void write_character(CharType c) override
+    {
+        str.push_back(c);
+    }
+
+    JSON_HEDLEY_NON_NULL(2)
+    void write_characters(const CharType* s, std::size_t length) override
+    {
+        str.append(s, length);
+    }
+
+  private:
+    StringType& str;
+};
+
+template<typename CharType, typename StringType = std::basic_string<CharType>>
+class output_adapter
+{
+  public:
+    template<typename AllocatorType = std::allocator<CharType>>
+    output_adapter(std::vector<CharType, AllocatorType>& vec)
+        : oa(std::make_shared<output_vector_adapter<CharType, AllocatorType>>(vec)) {}
+
+#ifndef JSON_NO_IO
+    output_adapter(std::basic_ostream<CharType>& s)
+        : oa(std::make_shared<output_stream_adapter<CharType>>(s)) {}
+#endif  // JSON_NO_IO
+
+    output_adapter(StringType& s)
+        : oa(std::make_shared<output_string_adapter<CharType, StringType>>(s)) {}
+
+    operator output_adapter_t<CharType>()
+    {
+        return oa;
+    }
+
+  private:
+    output_adapter_t<CharType> oa = nullptr;
+};
+}  // namespace detail
+}  // namespace nlohmann
+
+
 namespace nlohmann
 {
 namespace detail
@@ -7678,19 +13313,53 @@ namespace detail
 template<typename BasicJsonType, typename CharType>
 class binary_writer
 {
+    using string_t = typename BasicJsonType::string_t;
+    using binary_t = typename BasicJsonType::binary_t;
+    using number_float_t = typename BasicJsonType::number_float_t;
+
   public:
     /*!
     @brief create a binary writer
 
     @param[in] adapter  output adapter to write to
     */
-    explicit binary_writer(output_adapter_t<CharType> adapter) : oa(adapter)
+    explicit binary_writer(output_adapter_t<CharType> adapter) : oa(std::move(adapter))
     {
-        assert(oa);
+        JSON_ASSERT(oa);
     }
 
     /*!
-    @brief[in] j  JSON value to serialize
+    @param[in] j  JSON value to serialize
+    @pre       j.type() == value_t::object
+    */
+    void write_bson(const BasicJsonType& j)
+    {
+        switch (j.type())
+        {
+            case value_t::object:
+            {
+                write_bson_object(*j.m_value.object);
+                break;
+            }
+
+            case value_t::null:
+            case value_t::array:
+            case value_t::string:
+            case value_t::boolean:
+            case value_t::number_integer:
+            case value_t::number_unsigned:
+            case value_t::number_float:
+            case value_t::binary:
+            case value_t::discarded:
+            default:
+            {
+                JSON_THROW(type_error::create(317, "to serialize to BSON, top-level type must be object, but is " + std::string(j.type_name()), j));
+            }
+        }
+    }
+
+    /*!
+    @param[in] j  JSON value to serialize
     */
     void write_cbor(const BasicJsonType& j)
     {
@@ -7698,15 +13367,15 @@ class binary_writer
         {
             case value_t::null:
             {
-                oa->write_character(static_cast<CharType>(0xF6));
+                oa->write_character(to_char_type(0xF6));
                 break;
             }
 
             case value_t::boolean:
             {
                 oa->write_character(j.m_value.boolean
-                                    ? static_cast<CharType>(0xF5)
-                                    : static_cast<CharType>(0xF4));
+                                    ? to_char_type(0xF5)
+                                    : to_char_type(0xF4));
                 break;
             }
 
@@ -7719,27 +13388,27 @@ class binary_writer
                     // code from the value_t::number_unsigned case here.
                     if (j.m_value.number_integer <= 0x17)
                     {
-                        write_number(static_cast<uint8_t>(j.m_value.number_integer));
+                        write_number(static_cast<std::uint8_t>(j.m_value.number_integer));
                     }
-                    else if (j.m_value.number_integer <= (std::numeric_limits<uint8_t>::max)())
+                    else if (j.m_value.number_integer <= (std::numeric_limits<std::uint8_t>::max)())
                     {
-                        oa->write_character(static_cast<CharType>(0x18));
-                        write_number(static_cast<uint8_t>(j.m_value.number_integer));
+                        oa->write_character(to_char_type(0x18));
+                        write_number(static_cast<std::uint8_t>(j.m_value.number_integer));
                     }
-                    else if (j.m_value.number_integer <= (std::numeric_limits<uint16_t>::max)())
+                    else if (j.m_value.number_integer <= (std::numeric_limits<std::uint16_t>::max)())
                     {
-                        oa->write_character(static_cast<CharType>(0x19));
-                        write_number(static_cast<uint16_t>(j.m_value.number_integer));
+                        oa->write_character(to_char_type(0x19));
+                        write_number(static_cast<std::uint16_t>(j.m_value.number_integer));
                     }
-                    else if (j.m_value.number_integer <= (std::numeric_limits<uint32_t>::max)())
+                    else if (j.m_value.number_integer <= (std::numeric_limits<std::uint32_t>::max)())
                     {
-                        oa->write_character(static_cast<CharType>(0x1A));
-                        write_number(static_cast<uint32_t>(j.m_value.number_integer));
+                        oa->write_character(to_char_type(0x1A));
+                        write_number(static_cast<std::uint32_t>(j.m_value.number_integer));
                     }
                     else
                     {
-                        oa->write_character(static_cast<CharType>(0x1B));
-                        write_number(static_cast<uint64_t>(j.m_value.number_integer));
+                        oa->write_character(to_char_type(0x1B));
+                        write_number(static_cast<std::uint64_t>(j.m_value.number_integer));
                     }
                 }
                 else
@@ -7749,27 +13418,27 @@ class binary_writer
                     const auto positive_number = -1 - j.m_value.number_integer;
                     if (j.m_value.number_integer >= -24)
                     {
-                        write_number(static_cast<uint8_t>(0x20 + positive_number));
+                        write_number(static_cast<std::uint8_t>(0x20 + positive_number));
                     }
-                    else if (positive_number <= (std::numeric_limits<uint8_t>::max)())
+                    else if (positive_number <= (std::numeric_limits<std::uint8_t>::max)())
                     {
-                        oa->write_character(static_cast<CharType>(0x38));
-                        write_number(static_cast<uint8_t>(positive_number));
+                        oa->write_character(to_char_type(0x38));
+                        write_number(static_cast<std::uint8_t>(positive_number));
                     }
-                    else if (positive_number <= (std::numeric_limits<uint16_t>::max)())
+                    else if (positive_number <= (std::numeric_limits<std::uint16_t>::max)())
                     {
-                        oa->write_character(static_cast<CharType>(0x39));
-                        write_number(static_cast<uint16_t>(positive_number));
+                        oa->write_character(to_char_type(0x39));
+                        write_number(static_cast<std::uint16_t>(positive_number));
                     }
-                    else if (positive_number <= (std::numeric_limits<uint32_t>::max)())
+                    else if (positive_number <= (std::numeric_limits<std::uint32_t>::max)())
                     {
-                        oa->write_character(static_cast<CharType>(0x3A));
-                        write_number(static_cast<uint32_t>(positive_number));
+                        oa->write_character(to_char_type(0x3A));
+                        write_number(static_cast<std::uint32_t>(positive_number));
                     }
                     else
                     {
-                        oa->write_character(static_cast<CharType>(0x3B));
-                        write_number(static_cast<uint64_t>(positive_number));
+                        oa->write_character(to_char_type(0x3B));
+                        write_number(static_cast<std::uint64_t>(positive_number));
                     }
                 }
                 break;
@@ -7779,35 +13448,51 @@ class binary_writer
             {
                 if (j.m_value.number_unsigned <= 0x17)
                 {
-                    write_number(static_cast<uint8_t>(j.m_value.number_unsigned));
+                    write_number(static_cast<std::uint8_t>(j.m_value.number_unsigned));
                 }
-                else if (j.m_value.number_unsigned <= (std::numeric_limits<uint8_t>::max)())
+                else if (j.m_value.number_unsigned <= (std::numeric_limits<std::uint8_t>::max)())
                 {
-                    oa->write_character(static_cast<CharType>(0x18));
-                    write_number(static_cast<uint8_t>(j.m_value.number_unsigned));
+                    oa->write_character(to_char_type(0x18));
+                    write_number(static_cast<std::uint8_t>(j.m_value.number_unsigned));
                 }
-                else if (j.m_value.number_unsigned <= (std::numeric_limits<uint16_t>::max)())
+                else if (j.m_value.number_unsigned <= (std::numeric_limits<std::uint16_t>::max)())
                 {
-                    oa->write_character(static_cast<CharType>(0x19));
-                    write_number(static_cast<uint16_t>(j.m_value.number_unsigned));
+                    oa->write_character(to_char_type(0x19));
+                    write_number(static_cast<std::uint16_t>(j.m_value.number_unsigned));
                 }
-                else if (j.m_value.number_unsigned <= (std::numeric_limits<uint32_t>::max)())
+                else if (j.m_value.number_unsigned <= (std::numeric_limits<std::uint32_t>::max)())
                 {
-                    oa->write_character(static_cast<CharType>(0x1A));
-                    write_number(static_cast<uint32_t>(j.m_value.number_unsigned));
+                    oa->write_character(to_char_type(0x1A));
+                    write_number(static_cast<std::uint32_t>(j.m_value.number_unsigned));
                 }
                 else
                 {
-                    oa->write_character(static_cast<CharType>(0x1B));
-                    write_number(static_cast<uint64_t>(j.m_value.number_unsigned));
+                    oa->write_character(to_char_type(0x1B));
+                    write_number(static_cast<std::uint64_t>(j.m_value.number_unsigned));
                 }
                 break;
             }
 
             case value_t::number_float:
             {
-                oa->write_character(get_cbor_float_prefix(j.m_value.number_float));
-                write_number(j.m_value.number_float);
+                if (std::isnan(j.m_value.number_float))
+                {
+                    // NaN is 0xf97e00 in CBOR
+                    oa->write_character(to_char_type(0xF9));
+                    oa->write_character(to_char_type(0x7E));
+                    oa->write_character(to_char_type(0x00));
+                }
+                else if (std::isinf(j.m_value.number_float))
+                {
+                    // Infinity is 0xf97c00, -Infinity is 0xf9fc00
+                    oa->write_character(to_char_type(0xf9));
+                    oa->write_character(j.m_value.number_float > 0 ? to_char_type(0x7C) : to_char_type(0xFC));
+                    oa->write_character(to_char_type(0x00));
+                }
+                else
+                {
+                    write_compact_float(j.m_value.number_float, detail::input_format_t::cbor);
+                }
                 break;
             }
 
@@ -7817,28 +13502,28 @@ class binary_writer
                 const auto N = j.m_value.string->size();
                 if (N <= 0x17)
                 {
-                    write_number(static_cast<uint8_t>(0x60 + N));
+                    write_number(static_cast<std::uint8_t>(0x60 + N));
                 }
-                else if (N <= (std::numeric_limits<uint8_t>::max)())
+                else if (N <= (std::numeric_limits<std::uint8_t>::max)())
                 {
-                    oa->write_character(static_cast<CharType>(0x78));
-                    write_number(static_cast<uint8_t>(N));
+                    oa->write_character(to_char_type(0x78));
+                    write_number(static_cast<std::uint8_t>(N));
                 }
-                else if (N <= (std::numeric_limits<uint16_t>::max)())
+                else if (N <= (std::numeric_limits<std::uint16_t>::max)())
                 {
-                    oa->write_character(static_cast<CharType>(0x79));
-                    write_number(static_cast<uint16_t>(N));
+                    oa->write_character(to_char_type(0x79));
+                    write_number(static_cast<std::uint16_t>(N));
                 }
-                else if (N <= (std::numeric_limits<uint32_t>::max)())
+                else if (N <= (std::numeric_limits<std::uint32_t>::max)())
                 {
-                    oa->write_character(static_cast<CharType>(0x7A));
-                    write_number(static_cast<uint32_t>(N));
+                    oa->write_character(to_char_type(0x7A));
+                    write_number(static_cast<std::uint32_t>(N));
                 }
                 // LCOV_EXCL_START
-                else if (N <= (std::numeric_limits<uint64_t>::max)())
+                else if (N <= (std::numeric_limits<std::uint64_t>::max)())
                 {
-                    oa->write_character(static_cast<CharType>(0x7B));
-                    write_number(static_cast<uint64_t>(N));
+                    oa->write_character(to_char_type(0x7B));
+                    write_number(static_cast<std::uint64_t>(N));
                 }
                 // LCOV_EXCL_STOP
 
@@ -7855,28 +13540,28 @@ class binary_writer
                 const auto N = j.m_value.array->size();
                 if (N <= 0x17)
                 {
-                    write_number(static_cast<uint8_t>(0x80 + N));
+                    write_number(static_cast<std::uint8_t>(0x80 + N));
                 }
-                else if (N <= (std::numeric_limits<uint8_t>::max)())
+                else if (N <= (std::numeric_limits<std::uint8_t>::max)())
                 {
-                    oa->write_character(static_cast<CharType>(0x98));
-                    write_number(static_cast<uint8_t>(N));
+                    oa->write_character(to_char_type(0x98));
+                    write_number(static_cast<std::uint8_t>(N));
                 }
-                else if (N <= (std::numeric_limits<uint16_t>::max)())
+                else if (N <= (std::numeric_limits<std::uint16_t>::max)())
                 {
-                    oa->write_character(static_cast<CharType>(0x99));
-                    write_number(static_cast<uint16_t>(N));
+                    oa->write_character(to_char_type(0x99));
+                    write_number(static_cast<std::uint16_t>(N));
                 }
-                else if (N <= (std::numeric_limits<uint32_t>::max)())
+                else if (N <= (std::numeric_limits<std::uint32_t>::max)())
                 {
-                    oa->write_character(static_cast<CharType>(0x9A));
-                    write_number(static_cast<uint32_t>(N));
+                    oa->write_character(to_char_type(0x9A));
+                    write_number(static_cast<std::uint32_t>(N));
                 }
                 // LCOV_EXCL_START
-                else if (N <= (std::numeric_limits<uint64_t>::max)())
+                else if (N <= (std::numeric_limits<std::uint64_t>::max)())
                 {
-                    oa->write_character(static_cast<CharType>(0x9B));
-                    write_number(static_cast<uint64_t>(N));
+                    oa->write_character(to_char_type(0x9B));
+                    write_number(static_cast<std::uint64_t>(N));
                 }
                 // LCOV_EXCL_STOP
 
@@ -7888,34 +13573,97 @@ class binary_writer
                 break;
             }
 
+            case value_t::binary:
+            {
+                if (j.m_value.binary->has_subtype())
+                {
+                    if (j.m_value.binary->subtype() <= (std::numeric_limits<std::uint8_t>::max)())
+                    {
+                        write_number(static_cast<std::uint8_t>(0xd8));
+                        write_number(static_cast<std::uint8_t>(j.m_value.binary->subtype()));
+                    }
+                    else if (j.m_value.binary->subtype() <= (std::numeric_limits<std::uint16_t>::max)())
+                    {
+                        write_number(static_cast<std::uint8_t>(0xd9));
+                        write_number(static_cast<std::uint16_t>(j.m_value.binary->subtype()));
+                    }
+                    else if (j.m_value.binary->subtype() <= (std::numeric_limits<std::uint32_t>::max)())
+                    {
+                        write_number(static_cast<std::uint8_t>(0xda));
+                        write_number(static_cast<std::uint32_t>(j.m_value.binary->subtype()));
+                    }
+                    else if (j.m_value.binary->subtype() <= (std::numeric_limits<std::uint64_t>::max)())
+                    {
+                        write_number(static_cast<std::uint8_t>(0xdb));
+                        write_number(static_cast<std::uint64_t>(j.m_value.binary->subtype()));
+                    }
+                }
+
+                // step 1: write control byte and the binary array size
+                const auto N = j.m_value.binary->size();
+                if (N <= 0x17)
+                {
+                    write_number(static_cast<std::uint8_t>(0x40 + N));
+                }
+                else if (N <= (std::numeric_limits<std::uint8_t>::max)())
+                {
+                    oa->write_character(to_char_type(0x58));
+                    write_number(static_cast<std::uint8_t>(N));
+                }
+                else if (N <= (std::numeric_limits<std::uint16_t>::max)())
+                {
+                    oa->write_character(to_char_type(0x59));
+                    write_number(static_cast<std::uint16_t>(N));
+                }
+                else if (N <= (std::numeric_limits<std::uint32_t>::max)())
+                {
+                    oa->write_character(to_char_type(0x5A));
+                    write_number(static_cast<std::uint32_t>(N));
+                }
+                // LCOV_EXCL_START
+                else if (N <= (std::numeric_limits<std::uint64_t>::max)())
+                {
+                    oa->write_character(to_char_type(0x5B));
+                    write_number(static_cast<std::uint64_t>(N));
+                }
+                // LCOV_EXCL_STOP
+
+                // step 2: write each element
+                oa->write_characters(
+                    reinterpret_cast<const CharType*>(j.m_value.binary->data()),
+                    N);
+
+                break;
+            }
+
             case value_t::object:
             {
                 // step 1: write control byte and the object size
                 const auto N = j.m_value.object->size();
                 if (N <= 0x17)
                 {
-                    write_number(static_cast<uint8_t>(0xA0 + N));
+                    write_number(static_cast<std::uint8_t>(0xA0 + N));
                 }
-                else if (N <= (std::numeric_limits<uint8_t>::max)())
+                else if (N <= (std::numeric_limits<std::uint8_t>::max)())
                 {
-                    oa->write_character(static_cast<CharType>(0xB8));
-                    write_number(static_cast<uint8_t>(N));
+                    oa->write_character(to_char_type(0xB8));
+                    write_number(static_cast<std::uint8_t>(N));
                 }
-                else if (N <= (std::numeric_limits<uint16_t>::max)())
+                else if (N <= (std::numeric_limits<std::uint16_t>::max)())
                 {
-                    oa->write_character(static_cast<CharType>(0xB9));
-                    write_number(static_cast<uint16_t>(N));
+                    oa->write_character(to_char_type(0xB9));
+                    write_number(static_cast<std::uint16_t>(N));
                 }
-                else if (N <= (std::numeric_limits<uint32_t>::max)())
+                else if (N <= (std::numeric_limits<std::uint32_t>::max)())
                 {
-                    oa->write_character(static_cast<CharType>(0xBA));
-                    write_number(static_cast<uint32_t>(N));
+                    oa->write_character(to_char_type(0xBA));
+                    write_number(static_cast<std::uint32_t>(N));
                 }
                 // LCOV_EXCL_START
-                else if (N <= (std::numeric_limits<uint64_t>::max)())
+                else if (N <= (std::numeric_limits<std::uint64_t>::max)())
                 {
-                    oa->write_character(static_cast<CharType>(0xBB));
-                    write_number(static_cast<uint64_t>(N));
+                    oa->write_character(to_char_type(0xBB));
+                    write_number(static_cast<std::uint64_t>(N));
                 }
                 // LCOV_EXCL_STOP
 
@@ -7928,13 +13676,14 @@ class binary_writer
                 break;
             }
 
+            case value_t::discarded:
             default:
                 break;
         }
     }
 
     /*!
-    @brief[in] j  JSON value to serialize
+    @param[in] j  JSON value to serialize
     */
     void write_msgpack(const BasicJsonType& j)
     {
@@ -7942,15 +13691,15 @@ class binary_writer
         {
             case value_t::null: // nil
             {
-                oa->write_character(static_cast<CharType>(0xC0));
+                oa->write_character(to_char_type(0xC0));
                 break;
             }
 
             case value_t::boolean: // true and false
             {
                 oa->write_character(j.m_value.boolean
-                                    ? static_cast<CharType>(0xC3)
-                                    : static_cast<CharType>(0xC2));
+                                    ? to_char_type(0xC3)
+                                    : to_char_type(0xC2));
                 break;
             }
 
@@ -7964,31 +13713,31 @@ class binary_writer
                     if (j.m_value.number_unsigned < 128)
                     {
                         // positive fixnum
-                        write_number(static_cast<uint8_t>(j.m_value.number_integer));
+                        write_number(static_cast<std::uint8_t>(j.m_value.number_integer));
                     }
-                    else if (j.m_value.number_unsigned <= (std::numeric_limits<uint8_t>::max)())
+                    else if (j.m_value.number_unsigned <= (std::numeric_limits<std::uint8_t>::max)())
                     {
                         // uint 8
-                        oa->write_character(static_cast<CharType>(0xCC));
-                        write_number(static_cast<uint8_t>(j.m_value.number_integer));
+                        oa->write_character(to_char_type(0xCC));
+                        write_number(static_cast<std::uint8_t>(j.m_value.number_integer));
                     }
-                    else if (j.m_value.number_unsigned <= (std::numeric_limits<uint16_t>::max)())
+                    else if (j.m_value.number_unsigned <= (std::numeric_limits<std::uint16_t>::max)())
                     {
                         // uint 16
-                        oa->write_character(static_cast<CharType>(0xCD));
-                        write_number(static_cast<uint16_t>(j.m_value.number_integer));
+                        oa->write_character(to_char_type(0xCD));
+                        write_number(static_cast<std::uint16_t>(j.m_value.number_integer));
                     }
-                    else if (j.m_value.number_unsigned <= (std::numeric_limits<uint32_t>::max)())
+                    else if (j.m_value.number_unsigned <= (std::numeric_limits<std::uint32_t>::max)())
                     {
                         // uint 32
-                        oa->write_character(static_cast<CharType>(0xCE));
-                        write_number(static_cast<uint32_t>(j.m_value.number_integer));
+                        oa->write_character(to_char_type(0xCE));
+                        write_number(static_cast<std::uint32_t>(j.m_value.number_integer));
                     }
-                    else if (j.m_value.number_unsigned <= (std::numeric_limits<uint64_t>::max)())
+                    else if (j.m_value.number_unsigned <= (std::numeric_limits<std::uint64_t>::max)())
                     {
                         // uint 64
-                        oa->write_character(static_cast<CharType>(0xCF));
-                        write_number(static_cast<uint64_t>(j.m_value.number_integer));
+                        oa->write_character(to_char_type(0xCF));
+                        write_number(static_cast<std::uint64_t>(j.m_value.number_integer));
                     }
                 }
                 else
@@ -7996,35 +13745,35 @@ class binary_writer
                     if (j.m_value.number_integer >= -32)
                     {
                         // negative fixnum
-                        write_number(static_cast<int8_t>(j.m_value.number_integer));
+                        write_number(static_cast<std::int8_t>(j.m_value.number_integer));
                     }
-                    else if (j.m_value.number_integer >= (std::numeric_limits<int8_t>::min)() and
-                             j.m_value.number_integer <= (std::numeric_limits<int8_t>::max)())
+                    else if (j.m_value.number_integer >= (std::numeric_limits<std::int8_t>::min)() &&
+                             j.m_value.number_integer <= (std::numeric_limits<std::int8_t>::max)())
                     {
                         // int 8
-                        oa->write_character(static_cast<CharType>(0xD0));
-                        write_number(static_cast<int8_t>(j.m_value.number_integer));
+                        oa->write_character(to_char_type(0xD0));
+                        write_number(static_cast<std::int8_t>(j.m_value.number_integer));
                     }
-                    else if (j.m_value.number_integer >= (std::numeric_limits<int16_t>::min)() and
-                             j.m_value.number_integer <= (std::numeric_limits<int16_t>::max)())
+                    else if (j.m_value.number_integer >= (std::numeric_limits<std::int16_t>::min)() &&
+                             j.m_value.number_integer <= (std::numeric_limits<std::int16_t>::max)())
                     {
                         // int 16
-                        oa->write_character(static_cast<CharType>(0xD1));
-                        write_number(static_cast<int16_t>(j.m_value.number_integer));
+                        oa->write_character(to_char_type(0xD1));
+                        write_number(static_cast<std::int16_t>(j.m_value.number_integer));
                     }
-                    else if (j.m_value.number_integer >= (std::numeric_limits<int32_t>::min)() and
-                             j.m_value.number_integer <= (std::numeric_limits<int32_t>::max)())
+                    else if (j.m_value.number_integer >= (std::numeric_limits<std::int32_t>::min)() &&
+                             j.m_value.number_integer <= (std::numeric_limits<std::int32_t>::max)())
                     {
                         // int 32
-                        oa->write_character(static_cast<CharType>(0xD2));
-                        write_number(static_cast<int32_t>(j.m_value.number_integer));
+                        oa->write_character(to_char_type(0xD2));
+                        write_number(static_cast<std::int32_t>(j.m_value.number_integer));
                     }
-                    else if (j.m_value.number_integer >= (std::numeric_limits<int64_t>::min)() and
-                             j.m_value.number_integer <= (std::numeric_limits<int64_t>::max)())
+                    else if (j.m_value.number_integer >= (std::numeric_limits<std::int64_t>::min)() &&
+                             j.m_value.number_integer <= (std::numeric_limits<std::int64_t>::max)())
                     {
                         // int 64
-                        oa->write_character(static_cast<CharType>(0xD3));
-                        write_number(static_cast<int64_t>(j.m_value.number_integer));
+                        oa->write_character(to_char_type(0xD3));
+                        write_number(static_cast<std::int64_t>(j.m_value.number_integer));
                     }
                 }
                 break;
@@ -8035,39 +13784,38 @@ class binary_writer
                 if (j.m_value.number_unsigned < 128)
                 {
                     // positive fixnum
-                    write_number(static_cast<uint8_t>(j.m_value.number_integer));
+                    write_number(static_cast<std::uint8_t>(j.m_value.number_integer));
                 }
-                else if (j.m_value.number_unsigned <= (std::numeric_limits<uint8_t>::max)())
+                else if (j.m_value.number_unsigned <= (std::numeric_limits<std::uint8_t>::max)())
                 {
                     // uint 8
-                    oa->write_character(static_cast<CharType>(0xCC));
-                    write_number(static_cast<uint8_t>(j.m_value.number_integer));
+                    oa->write_character(to_char_type(0xCC));
+                    write_number(static_cast<std::uint8_t>(j.m_value.number_integer));
                 }
-                else if (j.m_value.number_unsigned <= (std::numeric_limits<uint16_t>::max)())
+                else if (j.m_value.number_unsigned <= (std::numeric_limits<std::uint16_t>::max)())
                 {
                     // uint 16
-                    oa->write_character(static_cast<CharType>(0xCD));
-                    write_number(static_cast<uint16_t>(j.m_value.number_integer));
+                    oa->write_character(to_char_type(0xCD));
+                    write_number(static_cast<std::uint16_t>(j.m_value.number_integer));
                 }
-                else if (j.m_value.number_unsigned <= (std::numeric_limits<uint32_t>::max)())
+                else if (j.m_value.number_unsigned <= (std::numeric_limits<std::uint32_t>::max)())
                 {
                     // uint 32
-                    oa->write_character(static_cast<CharType>(0xCE));
-                    write_number(static_cast<uint32_t>(j.m_value.number_integer));
+                    oa->write_character(to_char_type(0xCE));
+                    write_number(static_cast<std::uint32_t>(j.m_value.number_integer));
                 }
-                else if (j.m_value.number_unsigned <= (std::numeric_limits<uint64_t>::max)())
+                else if (j.m_value.number_unsigned <= (std::numeric_limits<std::uint64_t>::max)())
                 {
                     // uint 64
-                    oa->write_character(static_cast<CharType>(0xCF));
-                    write_number(static_cast<uint64_t>(j.m_value.number_integer));
+                    oa->write_character(to_char_type(0xCF));
+                    write_number(static_cast<std::uint64_t>(j.m_value.number_integer));
                 }
                 break;
             }
 
             case value_t::number_float:
             {
-                oa->write_character(get_msgpack_float_prefix(j.m_value.number_float));
-                write_number(j.m_value.number_float);
+                write_compact_float(j.m_value.number_float, detail::input_format_t::msgpack);
                 break;
             }
 
@@ -8078,25 +13826,25 @@ class binary_writer
                 if (N <= 31)
                 {
                     // fixstr
-                    write_number(static_cast<uint8_t>(0xA0 | N));
+                    write_number(static_cast<std::uint8_t>(0xA0 | N));
                 }
-                else if (N <= (std::numeric_limits<uint8_t>::max)())
+                else if (N <= (std::numeric_limits<std::uint8_t>::max)())
                 {
                     // str 8
-                    oa->write_character(static_cast<CharType>(0xD9));
-                    write_number(static_cast<uint8_t>(N));
+                    oa->write_character(to_char_type(0xD9));
+                    write_number(static_cast<std::uint8_t>(N));
                 }
-                else if (N <= (std::numeric_limits<uint16_t>::max)())
+                else if (N <= (std::numeric_limits<std::uint16_t>::max)())
                 {
                     // str 16
-                    oa->write_character(static_cast<CharType>(0xDA));
-                    write_number(static_cast<uint16_t>(N));
+                    oa->write_character(to_char_type(0xDA));
+                    write_number(static_cast<std::uint16_t>(N));
                 }
-                else if (N <= (std::numeric_limits<uint32_t>::max)())
+                else if (N <= (std::numeric_limits<std::uint32_t>::max)())
                 {
                     // str 32
-                    oa->write_character(static_cast<CharType>(0xDB));
-                    write_number(static_cast<uint32_t>(N));
+                    oa->write_character(to_char_type(0xDB));
+                    write_number(static_cast<std::uint32_t>(N));
                 }
 
                 // step 2: write the string
@@ -8113,19 +13861,19 @@ class binary_writer
                 if (N <= 15)
                 {
                     // fixarray
-                    write_number(static_cast<uint8_t>(0x90 | N));
+                    write_number(static_cast<std::uint8_t>(0x90 | N));
                 }
-                else if (N <= (std::numeric_limits<uint16_t>::max)())
+                else if (N <= (std::numeric_limits<std::uint16_t>::max)())
                 {
                     // array 16
-                    oa->write_character(static_cast<CharType>(0xDC));
-                    write_number(static_cast<uint16_t>(N));
+                    oa->write_character(to_char_type(0xDC));
+                    write_number(static_cast<std::uint16_t>(N));
                 }
-                else if (N <= (std::numeric_limits<uint32_t>::max)())
+                else if (N <= (std::numeric_limits<std::uint32_t>::max)())
                 {
                     // array 32
-                    oa->write_character(static_cast<CharType>(0xDD));
-                    write_number(static_cast<uint32_t>(N));
+                    oa->write_character(to_char_type(0xDD));
+                    write_number(static_cast<std::uint32_t>(N));
                 }
 
                 // step 2: write each element
@@ -8136,6 +13884,89 @@ class binary_writer
                 break;
             }
 
+            case value_t::binary:
+            {
+                // step 0: determine if the binary type has a set subtype to
+                // determine whether or not to use the ext or fixext types
+                const bool use_ext = j.m_value.binary->has_subtype();
+
+                // step 1: write control byte and the byte string length
+                const auto N = j.m_value.binary->size();
+                if (N <= (std::numeric_limits<std::uint8_t>::max)())
+                {
+                    std::uint8_t output_type{};
+                    bool fixed = true;
+                    if (use_ext)
+                    {
+                        switch (N)
+                        {
+                            case 1:
+                                output_type = 0xD4; // fixext 1
+                                break;
+                            case 2:
+                                output_type = 0xD5; // fixext 2
+                                break;
+                            case 4:
+                                output_type = 0xD6; // fixext 4
+                                break;
+                            case 8:
+                                output_type = 0xD7; // fixext 8
+                                break;
+                            case 16:
+                                output_type = 0xD8; // fixext 16
+                                break;
+                            default:
+                                output_type = 0xC7; // ext 8
+                                fixed = false;
+                                break;
+                        }
+
+                    }
+                    else
+                    {
+                        output_type = 0xC4; // bin 8
+                        fixed = false;
+                    }
+
+                    oa->write_character(to_char_type(output_type));
+                    if (!fixed)
+                    {
+                        write_number(static_cast<std::uint8_t>(N));
+                    }
+                }
+                else if (N <= (std::numeric_limits<std::uint16_t>::max)())
+                {
+                    std::uint8_t output_type = use_ext
+                                               ? 0xC8 // ext 16
+                                               : 0xC5; // bin 16
+
+                    oa->write_character(to_char_type(output_type));
+                    write_number(static_cast<std::uint16_t>(N));
+                }
+                else if (N <= (std::numeric_limits<std::uint32_t>::max)())
+                {
+                    std::uint8_t output_type = use_ext
+                                               ? 0xC9 // ext 32
+                                               : 0xC6; // bin 32
+
+                    oa->write_character(to_char_type(output_type));
+                    write_number(static_cast<std::uint32_t>(N));
+                }
+
+                // step 1.5: if this is an ext type, write the subtype
+                if (use_ext)
+                {
+                    write_number(static_cast<std::int8_t>(j.m_value.binary->subtype()));
+                }
+
+                // step 2: write the byte string
+                oa->write_characters(
+                    reinterpret_cast<const CharType*>(j.m_value.binary->data()),
+                    N);
+
+                break;
+            }
+
             case value_t::object:
             {
                 // step 1: write control byte and the object size
@@ -8143,19 +13974,19 @@ class binary_writer
                 if (N <= 15)
                 {
                     // fixmap
-                    write_number(static_cast<uint8_t>(0x80 | (N & 0xF)));
+                    write_number(static_cast<std::uint8_t>(0x80 | (N & 0xF)));
                 }
-                else if (N <= (std::numeric_limits<uint16_t>::max)())
+                else if (N <= (std::numeric_limits<std::uint16_t>::max)())
                 {
                     // map 16
-                    oa->write_character(static_cast<CharType>(0xDE));
-                    write_number(static_cast<uint16_t>(N));
+                    oa->write_character(to_char_type(0xDE));
+                    write_number(static_cast<std::uint16_t>(N));
                 }
-                else if (N <= (std::numeric_limits<uint32_t>::max)())
+                else if (N <= (std::numeric_limits<std::uint32_t>::max)())
                 {
                     // map 32
-                    oa->write_character(static_cast<CharType>(0xDF));
-                    write_number(static_cast<uint32_t>(N));
+                    oa->write_character(to_char_type(0xDF));
+                    write_number(static_cast<std::uint32_t>(N));
                 }
 
                 // step 2: write each element
@@ -8167,6 +13998,7 @@ class binary_writer
                 break;
             }
 
+            case value_t::discarded:
             default:
                 break;
         }
@@ -8187,7 +14019,7 @@ class binary_writer
             {
                 if (add_prefix)
                 {
-                    oa->write_character(static_cast<CharType>('Z'));
+                    oa->write_character(to_char_type('Z'));
                 }
                 break;
             }
@@ -8195,9 +14027,11 @@ class binary_writer
             case value_t::boolean:
             {
                 if (add_prefix)
+                {
                     oa->write_character(j.m_value.boolean
-                                        ? static_cast<CharType>('T')
-                                        : static_cast<CharType>('F'));
+                                        ? to_char_type('T')
+                                        : to_char_type('F'));
+                }
                 break;
             }
 
@@ -8223,7 +14057,7 @@ class binary_writer
             {
                 if (add_prefix)
                 {
-                    oa->write_character(static_cast<CharType>('S'));
+                    oa->write_character(to_char_type('S'));
                 }
                 write_number_with_ubjson_prefix(j.m_value.string->size(), true);
                 oa->write_characters(
@@ -8236,13 +14070,13 @@ class binary_writer
             {
                 if (add_prefix)
                 {
-                    oa->write_character(static_cast<CharType>('['));
+                    oa->write_character(to_char_type('['));
                 }
 
                 bool prefix_required = true;
-                if (use_type and not j.m_value.array->empty())
+                if (use_type && !j.m_value.array->empty())
                 {
-                    assert(use_count);
+                    JSON_ASSERT(use_count);
                     const CharType first_prefix = ubjson_prefix(j.front());
                     const bool same_prefix = std::all_of(j.begin() + 1, j.end(),
                                                          [this, first_prefix](const BasicJsonType & v)
@@ -8253,14 +14087,14 @@ class binary_writer
                     if (same_prefix)
                     {
                         prefix_required = false;
-                        oa->write_character(static_cast<CharType>('$'));
+                        oa->write_character(to_char_type('$'));
                         oa->write_character(first_prefix);
                     }
                 }
 
                 if (use_count)
                 {
-                    oa->write_character(static_cast<CharType>('#'));
+                    oa->write_character(to_char_type('#'));
                     write_number_with_ubjson_prefix(j.m_value.array->size(), true);
                 }
 
@@ -8269,9 +14103,52 @@ class binary_writer
                     write_ubjson(el, use_count, use_type, prefix_required);
                 }
 
-                if (not use_count)
+                if (!use_count)
                 {
-                    oa->write_character(static_cast<CharType>(']'));
+                    oa->write_character(to_char_type(']'));
+                }
+
+                break;
+            }
+
+            case value_t::binary:
+            {
+                if (add_prefix)
+                {
+                    oa->write_character(to_char_type('['));
+                }
+
+                if (use_type && !j.m_value.binary->empty())
+                {
+                    JSON_ASSERT(use_count);
+                    oa->write_character(to_char_type('$'));
+                    oa->write_character('U');
+                }
+
+                if (use_count)
+                {
+                    oa->write_character(to_char_type('#'));
+                    write_number_with_ubjson_prefix(j.m_value.binary->size(), true);
+                }
+
+                if (use_type)
+                {
+                    oa->write_characters(
+                        reinterpret_cast<const CharType*>(j.m_value.binary->data()),
+                        j.m_value.binary->size());
+                }
+                else
+                {
+                    for (size_t i = 0; i < j.m_value.binary->size(); ++i)
+                    {
+                        oa->write_character(to_char_type('U'));
+                        oa->write_character(j.m_value.binary->data()[i]);
+                    }
+                }
+
+                if (!use_count)
+                {
+                    oa->write_character(to_char_type(']'));
                 }
 
                 break;
@@ -8281,13 +14158,13 @@ class binary_writer
             {
                 if (add_prefix)
                 {
-                    oa->write_character(static_cast<CharType>('{'));
+                    oa->write_character(to_char_type('{'));
                 }
 
                 bool prefix_required = true;
-                if (use_type and not j.m_value.object->empty())
+                if (use_type && !j.m_value.object->empty())
                 {
-                    assert(use_count);
+                    JSON_ASSERT(use_count);
                     const CharType first_prefix = ubjson_prefix(j.front());
                     const bool same_prefix = std::all_of(j.begin(), j.end(),
                                                          [this, first_prefix](const BasicJsonType & v)
@@ -8298,14 +14175,14 @@ class binary_writer
                     if (same_prefix)
                     {
                         prefix_required = false;
-                        oa->write_character(static_cast<CharType>('$'));
+                        oa->write_character(to_char_type('$'));
                         oa->write_character(first_prefix);
                     }
                 }
 
                 if (use_count)
                 {
-                    oa->write_character(static_cast<CharType>('#'));
+                    oa->write_character(to_char_type('#'));
                     write_number_with_ubjson_prefix(j.m_value.object->size(), true);
                 }
 
@@ -8318,47 +14195,387 @@ class binary_writer
                     write_ubjson(el.second, use_count, use_type, prefix_required);
                 }
 
-                if (not use_count)
+                if (!use_count)
                 {
-                    oa->write_character(static_cast<CharType>('}'));
+                    oa->write_character(to_char_type('}'));
                 }
 
                 break;
             }
 
+            case value_t::discarded:
             default:
                 break;
         }
     }
 
   private:
-    /*
-    @brief write a number to output input
+    //////////
+    // BSON //
+    //////////
 
-    @param[in] n number of type @a NumberType
-    @tparam NumberType the type of the number
-
-    @note This function needs to respect the system's endianess, because bytes
-          in CBOR, MessagePack, and UBJSON are stored in network order (big
-          endian) and therefore need reordering on little endian systems.
+    /*!
+    @return The size of a BSON document entry header, including the id marker
+            and the entry name size (and its null-terminator).
     */
-    template<typename NumberType>
-    void write_number(const NumberType n)
+    static std::size_t calc_bson_entry_header_size(const string_t& name, const BasicJsonType& j)
     {
-        // step 1: write number to array of length NumberType
-        std::array<CharType, sizeof(NumberType)> vec;
-        std::memcpy(vec.data(), &n, sizeof(NumberType));
-
-        // step 2: write array to output (with possible reordering)
-        if (is_little_endian)
+        const auto it = name.find(static_cast<typename string_t::value_type>(0));
+        if (JSON_HEDLEY_UNLIKELY(it != BasicJsonType::string_t::npos))
         {
-            // reverse byte order prior to conversion if necessary
-            std::reverse(vec.begin(), vec.end());
+            JSON_THROW(out_of_range::create(409, "BSON key cannot contain code point U+0000 (at byte " + std::to_string(it) + ")", j));
+            static_cast<void>(j);
         }
 
-        oa->write_characters(vec.data(), sizeof(NumberType));
+        return /*id*/ 1ul + name.size() + /*zero-terminator*/1u;
     }
 
+    /*!
+    @brief Writes the given @a element_type and @a name to the output adapter
+    */
+    void write_bson_entry_header(const string_t& name,
+                                 const std::uint8_t element_type)
+    {
+        oa->write_character(to_char_type(element_type)); // boolean
+        oa->write_characters(
+            reinterpret_cast<const CharType*>(name.c_str()),
+            name.size() + 1u);
+    }
+
+    /*!
+    @brief Writes a BSON element with key @a name and boolean value @a value
+    */
+    void write_bson_boolean(const string_t& name,
+                            const bool value)
+    {
+        write_bson_entry_header(name, 0x08);
+        oa->write_character(value ? to_char_type(0x01) : to_char_type(0x00));
+    }
+
+    /*!
+    @brief Writes a BSON element with key @a name and double value @a value
+    */
+    void write_bson_double(const string_t& name,
+                           const double value)
+    {
+        write_bson_entry_header(name, 0x01);
+        write_number<double, true>(value);
+    }
+
+    /*!
+    @return The size of the BSON-encoded string in @a value
+    */
+    static std::size_t calc_bson_string_size(const string_t& value)
+    {
+        return sizeof(std::int32_t) + value.size() + 1ul;
+    }
+
+    /*!
+    @brief Writes a BSON element with key @a name and string value @a value
+    */
+    void write_bson_string(const string_t& name,
+                           const string_t& value)
+    {
+        write_bson_entry_header(name, 0x02);
+
+        write_number<std::int32_t, true>(static_cast<std::int32_t>(value.size() + 1ul));
+        oa->write_characters(
+            reinterpret_cast<const CharType*>(value.c_str()),
+            value.size() + 1);
+    }
+
+    /*!
+    @brief Writes a BSON element with key @a name and null value
+    */
+    void write_bson_null(const string_t& name)
+    {
+        write_bson_entry_header(name, 0x0A);
+    }
+
+    /*!
+    @return The size of the BSON-encoded integer @a value
+    */
+    static std::size_t calc_bson_integer_size(const std::int64_t value)
+    {
+        return (std::numeric_limits<std::int32_t>::min)() <= value && value <= (std::numeric_limits<std::int32_t>::max)()
+               ? sizeof(std::int32_t)
+               : sizeof(std::int64_t);
+    }
+
+    /*!
+    @brief Writes a BSON element with key @a name and integer @a value
+    */
+    void write_bson_integer(const string_t& name,
+                            const std::int64_t value)
+    {
+        if ((std::numeric_limits<std::int32_t>::min)() <= value && value <= (std::numeric_limits<std::int32_t>::max)())
+        {
+            write_bson_entry_header(name, 0x10); // int32
+            write_number<std::int32_t, true>(static_cast<std::int32_t>(value));
+        }
+        else
+        {
+            write_bson_entry_header(name, 0x12); // int64
+            write_number<std::int64_t, true>(static_cast<std::int64_t>(value));
+        }
+    }
+
+    /*!
+    @return The size of the BSON-encoded unsigned integer in @a j
+    */
+    static constexpr std::size_t calc_bson_unsigned_size(const std::uint64_t value) noexcept
+    {
+        return (value <= static_cast<std::uint64_t>((std::numeric_limits<std::int32_t>::max)()))
+               ? sizeof(std::int32_t)
+               : sizeof(std::int64_t);
+    }
+
+    /*!
+    @brief Writes a BSON element with key @a name and unsigned @a value
+    */
+    void write_bson_unsigned(const string_t& name,
+                             const BasicJsonType& j)
+    {
+        if (j.m_value.number_unsigned <= static_cast<std::uint64_t>((std::numeric_limits<std::int32_t>::max)()))
+        {
+            write_bson_entry_header(name, 0x10 /* int32 */);
+            write_number<std::int32_t, true>(static_cast<std::int32_t>(j.m_value.number_unsigned));
+        }
+        else if (j.m_value.number_unsigned <= static_cast<std::uint64_t>((std::numeric_limits<std::int64_t>::max)()))
+        {
+            write_bson_entry_header(name, 0x12 /* int64 */);
+            write_number<std::int64_t, true>(static_cast<std::int64_t>(j.m_value.number_unsigned));
+        }
+        else
+        {
+            JSON_THROW(out_of_range::create(407, "integer number " + std::to_string(j.m_value.number_unsigned) + " cannot be represented by BSON as it does not fit int64", j));
+        }
+    }
+
+    /*!
+    @brief Writes a BSON element with key @a name and object @a value
+    */
+    void write_bson_object_entry(const string_t& name,
+                                 const typename BasicJsonType::object_t& value)
+    {
+        write_bson_entry_header(name, 0x03); // object
+        write_bson_object(value);
+    }
+
+    /*!
+    @return The size of the BSON-encoded array @a value
+    */
+    static std::size_t calc_bson_array_size(const typename BasicJsonType::array_t& value)
+    {
+        std::size_t array_index = 0ul;
+
+        const std::size_t embedded_document_size = std::accumulate(std::begin(value), std::end(value), static_cast<std::size_t>(0), [&array_index](std::size_t result, const typename BasicJsonType::array_t::value_type & el)
+        {
+            return result + calc_bson_element_size(std::to_string(array_index++), el);
+        });
+
+        return sizeof(std::int32_t) + embedded_document_size + 1ul;
+    }
+
+    /*!
+    @return The size of the BSON-encoded binary array @a value
+    */
+    static std::size_t calc_bson_binary_size(const typename BasicJsonType::binary_t& value)
+    {
+        return sizeof(std::int32_t) + value.size() + 1ul;
+    }
+
+    /*!
+    @brief Writes a BSON element with key @a name and array @a value
+    */
+    void write_bson_array(const string_t& name,
+                          const typename BasicJsonType::array_t& value)
+    {
+        write_bson_entry_header(name, 0x04); // array
+        write_number<std::int32_t, true>(static_cast<std::int32_t>(calc_bson_array_size(value)));
+
+        std::size_t array_index = 0ul;
+
+        for (const auto& el : value)
+        {
+            write_bson_element(std::to_string(array_index++), el);
+        }
+
+        oa->write_character(to_char_type(0x00));
+    }
+
+    /*!
+    @brief Writes a BSON element with key @a name and binary value @a value
+    */
+    void write_bson_binary(const string_t& name,
+                           const binary_t& value)
+    {
+        write_bson_entry_header(name, 0x05);
+
+        write_number<std::int32_t, true>(static_cast<std::int32_t>(value.size()));
+        write_number(value.has_subtype() ? static_cast<std::uint8_t>(value.subtype()) : static_cast<std::uint8_t>(0x00));
+
+        oa->write_characters(reinterpret_cast<const CharType*>(value.data()), value.size());
+    }
+
+    /*!
+    @brief Calculates the size necessary to serialize the JSON value @a j with its @a name
+    @return The calculated size for the BSON document entry for @a j with the given @a name.
+    */
+    static std::size_t calc_bson_element_size(const string_t& name,
+            const BasicJsonType& j)
+    {
+        const auto header_size = calc_bson_entry_header_size(name, j);
+        switch (j.type())
+        {
+            case value_t::object:
+                return header_size + calc_bson_object_size(*j.m_value.object);
+
+            case value_t::array:
+                return header_size + calc_bson_array_size(*j.m_value.array);
+
+            case value_t::binary:
+                return header_size + calc_bson_binary_size(*j.m_value.binary);
+
+            case value_t::boolean:
+                return header_size + 1ul;
+
+            case value_t::number_float:
+                return header_size + 8ul;
+
+            case value_t::number_integer:
+                return header_size + calc_bson_integer_size(j.m_value.number_integer);
+
+            case value_t::number_unsigned:
+                return header_size + calc_bson_unsigned_size(j.m_value.number_unsigned);
+
+            case value_t::string:
+                return header_size + calc_bson_string_size(*j.m_value.string);
+
+            case value_t::null:
+                return header_size + 0ul;
+
+            // LCOV_EXCL_START
+            case value_t::discarded:
+            default:
+                JSON_ASSERT(false); // NOLINT(cert-dcl03-c,hicpp-static-assert,misc-static-assert)
+                return 0ul;
+                // LCOV_EXCL_STOP
+        }
+    }
+
+    /*!
+    @brief Serializes the JSON value @a j to BSON and associates it with the
+           key @a name.
+    @param name The name to associate with the JSON entity @a j within the
+                current BSON document
+    */
+    void write_bson_element(const string_t& name,
+                            const BasicJsonType& j)
+    {
+        switch (j.type())
+        {
+            case value_t::object:
+                return write_bson_object_entry(name, *j.m_value.object);
+
+            case value_t::array:
+                return write_bson_array(name, *j.m_value.array);
+
+            case value_t::binary:
+                return write_bson_binary(name, *j.m_value.binary);
+
+            case value_t::boolean:
+                return write_bson_boolean(name, j.m_value.boolean);
+
+            case value_t::number_float:
+                return write_bson_double(name, j.m_value.number_float);
+
+            case value_t::number_integer:
+                return write_bson_integer(name, j.m_value.number_integer);
+
+            case value_t::number_unsigned:
+                return write_bson_unsigned(name, j);
+
+            case value_t::string:
+                return write_bson_string(name, *j.m_value.string);
+
+            case value_t::null:
+                return write_bson_null(name);
+
+            // LCOV_EXCL_START
+            case value_t::discarded:
+            default:
+                JSON_ASSERT(false); // NOLINT(cert-dcl03-c,hicpp-static-assert,misc-static-assert)
+                return;
+                // LCOV_EXCL_STOP
+        }
+    }
+
+    /*!
+    @brief Calculates the size of the BSON serialization of the given
+           JSON-object @a j.
+    @param[in] value  JSON value to serialize
+    @pre       value.type() == value_t::object
+    */
+    static std::size_t calc_bson_object_size(const typename BasicJsonType::object_t& value)
+    {
+        std::size_t document_size = std::accumulate(value.begin(), value.end(), static_cast<std::size_t>(0),
+                                    [](size_t result, const typename BasicJsonType::object_t::value_type & el)
+        {
+            return result += calc_bson_element_size(el.first, el.second);
+        });
+
+        return sizeof(std::int32_t) + document_size + 1ul;
+    }
+
+    /*!
+    @param[in] value  JSON value to serialize
+    @pre       value.type() == value_t::object
+    */
+    void write_bson_object(const typename BasicJsonType::object_t& value)
+    {
+        write_number<std::int32_t, true>(static_cast<std::int32_t>(calc_bson_object_size(value)));
+
+        for (const auto& el : value)
+        {
+            write_bson_element(el.first, el.second);
+        }
+
+        oa->write_character(to_char_type(0x00));
+    }
+
+    //////////
+    // CBOR //
+    //////////
+
+    static constexpr CharType get_cbor_float_prefix(float /*unused*/)
+    {
+        return to_char_type(0xFA);  // Single-Precision Float
+    }
+
+    static constexpr CharType get_cbor_float_prefix(double /*unused*/)
+    {
+        return to_char_type(0xFB);  // Double-Precision Float
+    }
+
+    /////////////
+    // MsgPack //
+    /////////////
+
+    static constexpr CharType get_msgpack_float_prefix(float /*unused*/)
+    {
+        return to_char_type(0xCA);  // float 32
+    }
+
+    static constexpr CharType get_msgpack_float_prefix(double /*unused*/)
+    {
+        return to_char_type(0xCB);  // float 64
+    }
+
+    ////////////
+    // UBJSON //
+    ////////////
+
     // UBJSON: write number (floating point)
     template<typename NumberType, typename std::enable_if<
                  std::is_floating_point<NumberType>::value, int>::type = 0>
@@ -8378,115 +14595,129 @@ class binary_writer
     void write_number_with_ubjson_prefix(const NumberType n,
                                          const bool add_prefix)
     {
-        if (n <= static_cast<uint64_t>((std::numeric_limits<int8_t>::max)()))
+        if (n <= static_cast<std::uint64_t>((std::numeric_limits<std::int8_t>::max)()))
         {
             if (add_prefix)
             {
-                oa->write_character(static_cast<CharType>('i'));  // int8
+                oa->write_character(to_char_type('i'));  // int8
             }
-            write_number(static_cast<uint8_t>(n));
+            write_number(static_cast<std::uint8_t>(n));
         }
-        else if (n <= (std::numeric_limits<uint8_t>::max)())
+        else if (n <= (std::numeric_limits<std::uint8_t>::max)())
         {
             if (add_prefix)
             {
-                oa->write_character(static_cast<CharType>('U'));  // uint8
+                oa->write_character(to_char_type('U'));  // uint8
             }
-            write_number(static_cast<uint8_t>(n));
+            write_number(static_cast<std::uint8_t>(n));
         }
-        else if (n <= static_cast<uint64_t>((std::numeric_limits<int16_t>::max)()))
+        else if (n <= static_cast<std::uint64_t>((std::numeric_limits<std::int16_t>::max)()))
         {
             if (add_prefix)
             {
-                oa->write_character(static_cast<CharType>('I'));  // int16
+                oa->write_character(to_char_type('I'));  // int16
             }
-            write_number(static_cast<int16_t>(n));
+            write_number(static_cast<std::int16_t>(n));
         }
-        else if (n <= static_cast<uint64_t>((std::numeric_limits<int32_t>::max)()))
+        else if (n <= static_cast<std::uint64_t>((std::numeric_limits<std::int32_t>::max)()))
         {
             if (add_prefix)
             {
-                oa->write_character(static_cast<CharType>('l'));  // int32
+                oa->write_character(to_char_type('l'));  // int32
             }
-            write_number(static_cast<int32_t>(n));
+            write_number(static_cast<std::int32_t>(n));
         }
-        else if (n <= static_cast<uint64_t>((std::numeric_limits<int64_t>::max)()))
+        else if (n <= static_cast<std::uint64_t>((std::numeric_limits<std::int64_t>::max)()))
         {
             if (add_prefix)
             {
-                oa->write_character(static_cast<CharType>('L'));  // int64
+                oa->write_character(to_char_type('L'));  // int64
             }
-            write_number(static_cast<int64_t>(n));
+            write_number(static_cast<std::int64_t>(n));
         }
         else
         {
-            JSON_THROW(out_of_range::create(407, "number overflow serializing " + std::to_string(n)));
+            if (add_prefix)
+            {
+                oa->write_character(to_char_type('H'));  // high-precision number
+            }
+
+            const auto number = BasicJsonType(n).dump();
+            write_number_with_ubjson_prefix(number.size(), true);
+            for (std::size_t i = 0; i < number.size(); ++i)
+            {
+                oa->write_character(to_char_type(static_cast<std::uint8_t>(number[i])));
+            }
         }
     }
 
     // UBJSON: write number (signed integer)
-    template<typename NumberType, typename std::enable_if<
-                 std::is_signed<NumberType>::value and
-                 not std::is_floating_point<NumberType>::value, int>::type = 0>
+    template < typename NumberType, typename std::enable_if <
+                   std::is_signed<NumberType>::value&&
+                   !std::is_floating_point<NumberType>::value, int >::type = 0 >
     void write_number_with_ubjson_prefix(const NumberType n,
                                          const bool add_prefix)
     {
-        if ((std::numeric_limits<int8_t>::min)() <= n and n <= (std::numeric_limits<int8_t>::max)())
+        if ((std::numeric_limits<std::int8_t>::min)() <= n && n <= (std::numeric_limits<std::int8_t>::max)())
         {
             if (add_prefix)
             {
-                oa->write_character(static_cast<CharType>('i'));  // int8
+                oa->write_character(to_char_type('i'));  // int8
             }
-            write_number(static_cast<int8_t>(n));
+            write_number(static_cast<std::int8_t>(n));
         }
-        else if (static_cast<int64_t>((std::numeric_limits<uint8_t>::min)()) <= n and n <= static_cast<int64_t>((std::numeric_limits<uint8_t>::max)()))
+        else if (static_cast<std::int64_t>((std::numeric_limits<std::uint8_t>::min)()) <= n && n <= static_cast<std::int64_t>((std::numeric_limits<std::uint8_t>::max)()))
         {
             if (add_prefix)
             {
-                oa->write_character(static_cast<CharType>('U'));  // uint8
+                oa->write_character(to_char_type('U'));  // uint8
             }
-            write_number(static_cast<uint8_t>(n));
+            write_number(static_cast<std::uint8_t>(n));
         }
-        else if ((std::numeric_limits<int16_t>::min)() <= n and n <= (std::numeric_limits<int16_t>::max)())
+        else if ((std::numeric_limits<std::int16_t>::min)() <= n && n <= (std::numeric_limits<std::int16_t>::max)())
         {
             if (add_prefix)
             {
-                oa->write_character(static_cast<CharType>('I'));  // int16
+                oa->write_character(to_char_type('I'));  // int16
             }
-            write_number(static_cast<int16_t>(n));
+            write_number(static_cast<std::int16_t>(n));
         }
-        else if ((std::numeric_limits<int32_t>::min)() <= n and n <= (std::numeric_limits<int32_t>::max)())
+        else if ((std::numeric_limits<std::int32_t>::min)() <= n && n <= (std::numeric_limits<std::int32_t>::max)())
         {
             if (add_prefix)
             {
-                oa->write_character(static_cast<CharType>('l'));  // int32
+                oa->write_character(to_char_type('l'));  // int32
             }
-            write_number(static_cast<int32_t>(n));
+            write_number(static_cast<std::int32_t>(n));
         }
-        else if ((std::numeric_limits<int64_t>::min)() <= n and n <= (std::numeric_limits<int64_t>::max)())
+        else if ((std::numeric_limits<std::int64_t>::min)() <= n && n <= (std::numeric_limits<std::int64_t>::max)())
         {
             if (add_prefix)
             {
-                oa->write_character(static_cast<CharType>('L'));  // int64
+                oa->write_character(to_char_type('L'));  // int64
             }
-            write_number(static_cast<int64_t>(n));
+            write_number(static_cast<std::int64_t>(n));
         }
         // LCOV_EXCL_START
         else
         {
-            JSON_THROW(out_of_range::create(407, "number overflow serializing " + std::to_string(n)));
+            if (add_prefix)
+            {
+                oa->write_character(to_char_type('H'));  // high-precision number
+            }
+
+            const auto number = BasicJsonType(n).dump();
+            write_number_with_ubjson_prefix(number.size(), true);
+            for (std::size_t i = 0; i < number.size(); ++i)
+            {
+                oa->write_character(to_char_type(static_cast<std::uint8_t>(number[i])));
+            }
         }
         // LCOV_EXCL_STOP
     }
 
     /*!
     @brief determine the type prefix of container values
-
-    @note This function does not need to be 100% accurate when it comes to
-          integer limits. In case a number exceeds the limits of int64_t,
-          this will be detected by a later call to function
-          write_number_with_ubjson_prefix. Therefore, we return 'L' for any
-          value that does not fit the previous limits.
     */
     CharType ubjson_prefix(const BasicJsonType& j) const noexcept
     {
@@ -8500,50 +14731,54 @@ class binary_writer
 
             case value_t::number_integer:
             {
-                if ((std::numeric_limits<int8_t>::min)() <= j.m_value.number_integer and j.m_value.number_integer <= (std::numeric_limits<int8_t>::max)())
+                if ((std::numeric_limits<std::int8_t>::min)() <= j.m_value.number_integer && j.m_value.number_integer <= (std::numeric_limits<std::int8_t>::max)())
                 {
                     return 'i';
                 }
-                else if ((std::numeric_limits<uint8_t>::min)() <= j.m_value.number_integer and j.m_value.number_integer <= (std::numeric_limits<uint8_t>::max)())
+                if ((std::numeric_limits<std::uint8_t>::min)() <= j.m_value.number_integer && j.m_value.number_integer <= (std::numeric_limits<std::uint8_t>::max)())
                 {
                     return 'U';
                 }
-                else if ((std::numeric_limits<int16_t>::min)() <= j.m_value.number_integer and j.m_value.number_integer <= (std::numeric_limits<int16_t>::max)())
+                if ((std::numeric_limits<std::int16_t>::min)() <= j.m_value.number_integer && j.m_value.number_integer <= (std::numeric_limits<std::int16_t>::max)())
                 {
                     return 'I';
                 }
-                else if ((std::numeric_limits<int32_t>::min)() <= j.m_value.number_integer and j.m_value.number_integer <= (std::numeric_limits<int32_t>::max)())
+                if ((std::numeric_limits<std::int32_t>::min)() <= j.m_value.number_integer && j.m_value.number_integer <= (std::numeric_limits<std::int32_t>::max)())
                 {
                     return 'l';
                 }
-                else  // no check and assume int64_t (see note above)
+                if ((std::numeric_limits<std::int64_t>::min)() <= j.m_value.number_integer && j.m_value.number_integer <= (std::numeric_limits<std::int64_t>::max)())
                 {
                     return 'L';
                 }
+                // anything else is treated as high-precision number
+                return 'H'; // LCOV_EXCL_LINE
             }
 
             case value_t::number_unsigned:
             {
-                if (j.m_value.number_unsigned <= (std::numeric_limits<int8_t>::max)())
+                if (j.m_value.number_unsigned <= static_cast<std::uint64_t>((std::numeric_limits<std::int8_t>::max)()))
                 {
                     return 'i';
                 }
-                else if (j.m_value.number_unsigned <= (std::numeric_limits<uint8_t>::max)())
+                if (j.m_value.number_unsigned <= static_cast<std::uint64_t>((std::numeric_limits<std::uint8_t>::max)()))
                 {
                     return 'U';
                 }
-                else if (j.m_value.number_unsigned <= (std::numeric_limits<int16_t>::max)())
+                if (j.m_value.number_unsigned <= static_cast<std::uint64_t>((std::numeric_limits<std::int16_t>::max)()))
                 {
                     return 'I';
                 }
-                else if (j.m_value.number_unsigned <= (std::numeric_limits<int32_t>::max)())
+                if (j.m_value.number_unsigned <= static_cast<std::uint64_t>((std::numeric_limits<std::int32_t>::max)()))
                 {
                     return 'l';
                 }
-                else  // no check and assume int64_t (see note above)
+                if (j.m_value.number_unsigned <= static_cast<std::uint64_t>((std::numeric_limits<std::int64_t>::max)()))
                 {
                     return 'L';
                 }
+                // anything else is treated as high-precision number
+                return 'H'; // LCOV_EXCL_LINE
             }
 
             case value_t::number_float:
@@ -8552,83 +14787,170 @@ class binary_writer
             case value_t::string:
                 return 'S';
 
-            case value_t::array:
+            case value_t::array: // fallthrough
+            case value_t::binary:
                 return '[';
 
             case value_t::object:
                 return '{';
 
+            case value_t::discarded:
             default:  // discarded values
                 return 'N';
         }
     }
 
-    static constexpr CharType get_cbor_float_prefix(float)
-    {
-        return static_cast<CharType>(0xFA);  // Single-Precision Float
-    }
-
-    static constexpr CharType get_cbor_float_prefix(double)
-    {
-        return static_cast<CharType>(0xFB);  // Double-Precision Float
-    }
-
-    static constexpr CharType get_msgpack_float_prefix(float)
-    {
-        return static_cast<CharType>(0xCA);  // float 32
-    }
-
-    static constexpr CharType get_msgpack_float_prefix(double)
-    {
-        return static_cast<CharType>(0xCB);  // float 64
-    }
-
-    static constexpr CharType get_ubjson_float_prefix(float)
+    static constexpr CharType get_ubjson_float_prefix(float /*unused*/)
     {
         return 'd';  // float 32
     }
 
-    static constexpr CharType get_ubjson_float_prefix(double)
+    static constexpr CharType get_ubjson_float_prefix(double /*unused*/)
     {
         return 'D';  // float 64
     }
 
+    ///////////////////////
+    // Utility functions //
+    ///////////////////////
+
+    /*
+    @brief write a number to output input
+    @param[in] n number of type @a NumberType
+    @tparam NumberType the type of the number
+    @tparam OutputIsLittleEndian Set to true if output data is
+                                 required to be little endian
+
+    @note This function needs to respect the system's endianness, because bytes
+          in CBOR, MessagePack, and UBJSON are stored in network order (big
+          endian) and therefore need reordering on little endian systems.
+    */
+    template<typename NumberType, bool OutputIsLittleEndian = false>
+    void write_number(const NumberType n)
+    {
+        // step 1: write number to array of length NumberType
+        std::array<CharType, sizeof(NumberType)> vec{};
+        std::memcpy(vec.data(), &n, sizeof(NumberType));
+
+        // step 2: write array to output (with possible reordering)
+        if (is_little_endian != OutputIsLittleEndian)
+        {
+            // reverse byte order prior to conversion if necessary
+            std::reverse(vec.begin(), vec.end());
+        }
+
+        oa->write_characters(vec.data(), sizeof(NumberType));
+    }
+
+    void write_compact_float(const number_float_t n, detail::input_format_t format)
+    {
+#ifdef __GNUC__
+#pragma GCC diagnostic push
+#pragma GCC diagnostic ignored "-Wfloat-equal"
+#endif
+        if (static_cast<double>(n) >= static_cast<double>(std::numeric_limits<float>::lowest()) &&
+                static_cast<double>(n) <= static_cast<double>((std::numeric_limits<float>::max)()) &&
+                static_cast<double>(static_cast<float>(n)) == static_cast<double>(n))
+        {
+            oa->write_character(format == detail::input_format_t::cbor
+                                ? get_cbor_float_prefix(static_cast<float>(n))
+                                : get_msgpack_float_prefix(static_cast<float>(n)));
+            write_number(static_cast<float>(n));
+        }
+        else
+        {
+            oa->write_character(format == detail::input_format_t::cbor
+                                ? get_cbor_float_prefix(n)
+                                : get_msgpack_float_prefix(n));
+            write_number(n);
+        }
+#ifdef __GNUC__
+#pragma GCC diagnostic pop
+#endif
+    }
+
+  public:
+    // The following to_char_type functions are implement the conversion
+    // between uint8_t and CharType. In case CharType is not unsigned,
+    // such a conversion is required to allow values greater than 128.
+    // See <https://github.com/nlohmann/json/issues/1286> for a discussion.
+    template < typename C = CharType,
+               enable_if_t < std::is_signed<C>::value && std::is_signed<char>::value > * = nullptr >
+    static constexpr CharType to_char_type(std::uint8_t x) noexcept
+    {
+        return *reinterpret_cast<char*>(&x);
+    }
+
+    template < typename C = CharType,
+               enable_if_t < std::is_signed<C>::value && std::is_unsigned<char>::value > * = nullptr >
+    static CharType to_char_type(std::uint8_t x) noexcept
+    {
+        static_assert(sizeof(std::uint8_t) == sizeof(CharType), "size of CharType must be equal to std::uint8_t");
+        static_assert(std::is_trivial<CharType>::value, "CharType must be trivial");
+        CharType result;
+        std::memcpy(&result, &x, sizeof(x));
+        return result;
+    }
+
+    template<typename C = CharType,
+             enable_if_t<std::is_unsigned<C>::value>* = nullptr>
+    static constexpr CharType to_char_type(std::uint8_t x) noexcept
+    {
+        return x;
+    }
+
+    template < typename InputCharType, typename C = CharType,
+               enable_if_t <
+                   std::is_signed<C>::value &&
+                   std::is_signed<char>::value &&
+                   std::is_same<char, typename std::remove_cv<InputCharType>::type>::value
+                   > * = nullptr >
+    static constexpr CharType to_char_type(InputCharType x) noexcept
+    {
+        return x;
+    }
+
   private:
-    /// whether we can assume little endianess
-    const bool is_little_endian = binary_reader<BasicJsonType>::little_endianess();
+    /// whether we can assume little endianness
+    const bool is_little_endian = little_endianness();
 
     /// the output
     output_adapter_t<CharType> oa = nullptr;
 };
-}
-}
+}  // namespace detail
+}  // namespace nlohmann
+
+// #include <nlohmann/detail/output/output_adapters.hpp>
 
 // #include <nlohmann/detail/output/serializer.hpp>
 
 
 #include <algorithm> // reverse, remove, fill, find, none_of
 #include <array> // array
-#include <cassert> // assert
-#include <ciso646> // and, or
 #include <clocale> // localeconv, lconv
 #include <cmath> // labs, isfinite, isnan, signbit
 #include <cstddef> // size_t, ptrdiff_t
 #include <cstdint> // uint8_t
 #include <cstdio> // snprintf
 #include <limits> // numeric_limits
-#include <string> // string
+#include <string> // string, char_traits
+#include <iomanip> // setfill, setw
+#include <sstream> // stringstream
 #include <type_traits> // is_same
-
-// #include <nlohmann/detail/exceptions.hpp>
+#include <utility> // move
 
 // #include <nlohmann/detail/conversions/to_chars.hpp>
 
 
-#include <cassert> // assert
-#include <ciso646> // or, and, not
+#include <array> // array
 #include <cmath>   // signbit, isfinite
 #include <cstdint> // intN_t, uintN_t
 #include <cstring> // memcpy, memmove
+#include <limits> // numeric_limits
+#include <type_traits> // conditional
+
+// #include <nlohmann/detail/macro_scope.hpp>
+
 
 namespace nlohmann
 {
@@ -8657,7 +14979,7 @@ For a detailed description of the algorithm see:
 namespace dtoa_impl
 {
 
-template <typename Target, typename Source>
+template<typename Target, typename Source>
 Target reinterpret_bits(const Source source)
 {
     static_assert(sizeof(Target) == sizeof(Source), "size mismatch");
@@ -8671,11 +14993,10 @@ struct diyfp // f * 2^e
 {
     static constexpr int kPrecision = 64; // = q
 
-    uint64_t f;
-    int e;
+    std::uint64_t f = 0;
+    int e = 0;
 
-    constexpr diyfp() noexcept : f(0), e(0) {}
-    constexpr diyfp(uint64_t f_, int e_) noexcept : f(f_), e(e_) {}
+    constexpr diyfp(std::uint64_t f_, int e_) noexcept : f(f_), e(e_) {}
 
     /*!
     @brief returns x - y
@@ -8683,10 +15004,10 @@ struct diyfp // f * 2^e
     */
     static diyfp sub(const diyfp& x, const diyfp& y) noexcept
     {
-        assert(x.e == y.e);
-        assert(x.f >= y.f);
+        JSON_ASSERT(x.e == y.e);
+        JSON_ASSERT(x.f >= y.f);
 
-        return diyfp(x.f - y.f, x.e);
+        return {x.f - y.f, x.e};
     }
 
     /*!
@@ -8720,23 +15041,23 @@ struct diyfp // f * 2^e
         //
         //   = p_lo + 2^64 p_hi
 
-        const uint64_t u_lo = x.f & 0xFFFFFFFF;
-        const uint64_t u_hi = x.f >> 32;
-        const uint64_t v_lo = y.f & 0xFFFFFFFF;
-        const uint64_t v_hi = y.f >> 32;
+        const std::uint64_t u_lo = x.f & 0xFFFFFFFFu;
+        const std::uint64_t u_hi = x.f >> 32u;
+        const std::uint64_t v_lo = y.f & 0xFFFFFFFFu;
+        const std::uint64_t v_hi = y.f >> 32u;
 
-        const uint64_t p0 = u_lo * v_lo;
-        const uint64_t p1 = u_lo * v_hi;
-        const uint64_t p2 = u_hi * v_lo;
-        const uint64_t p3 = u_hi * v_hi;
+        const std::uint64_t p0 = u_lo * v_lo;
+        const std::uint64_t p1 = u_lo * v_hi;
+        const std::uint64_t p2 = u_hi * v_lo;
+        const std::uint64_t p3 = u_hi * v_hi;
 
-        const uint64_t p0_hi = p0 >> 32;
-        const uint64_t p1_lo = p1 & 0xFFFFFFFF;
-        const uint64_t p1_hi = p1 >> 32;
-        const uint64_t p2_lo = p2 & 0xFFFFFFFF;
-        const uint64_t p2_hi = p2 >> 32;
+        const std::uint64_t p0_hi = p0 >> 32u;
+        const std::uint64_t p1_lo = p1 & 0xFFFFFFFFu;
+        const std::uint64_t p1_hi = p1 >> 32u;
+        const std::uint64_t p2_lo = p2 & 0xFFFFFFFFu;
+        const std::uint64_t p2_hi = p2 >> 32u;
 
-        uint64_t Q = p0_hi + p1_lo + p2_lo;
+        std::uint64_t Q = p0_hi + p1_lo + p2_lo;
 
         // The full product might now be computed as
         //
@@ -8747,11 +15068,11 @@ struct diyfp // f * 2^e
         // Effectively we only need to add the highest bit in p_lo to p_hi (and
         // Q_hi + 1 does not overflow).
 
-        Q += uint64_t{1} << (64 - 32 - 1); // round, ties up
+        Q += std::uint64_t{1} << (64u - 32u - 1u); // round, ties up
 
-        const uint64_t h = p3 + p2_hi + p1_hi + (Q >> 32);
+        const std::uint64_t h = p3 + p2_hi + p1_hi + (Q >> 32u);
 
-        return diyfp(h, x.e + y.e + 64);
+        return {h, x.e + y.e + 64};
     }
 
     /*!
@@ -8760,11 +15081,11 @@ struct diyfp // f * 2^e
     */
     static diyfp normalize(diyfp x) noexcept
     {
-        assert(x.f != 0);
+        JSON_ASSERT(x.f != 0);
 
-        while ((x.f >> 63) == 0)
+        while ((x.f >> 63u) == 0)
         {
-            x.f <<= 1;
+            x.f <<= 1u;
             x.e--;
         }
 
@@ -8779,10 +15100,10 @@ struct diyfp // f * 2^e
     {
         const int delta = x.e - target_exponent;
 
-        assert(delta >= 0);
-        assert(((x.f << delta) >> delta) == x.f);
+        JSON_ASSERT(delta >= 0);
+        JSON_ASSERT(((x.f << delta) >> delta) == x.f);
 
-        return diyfp(x.f << delta, target_exponent);
+        return {x.f << delta, target_exponent};
     }
 };
 
@@ -8799,11 +15120,11 @@ boundaries.
 
 @pre value must be finite and positive
 */
-template <typename FloatType>
+template<typename FloatType>
 boundaries compute_boundaries(FloatType value)
 {
-    assert(std::isfinite(value));
-    assert(value > 0);
+    JSON_ASSERT(std::isfinite(value));
+    JSON_ASSERT(value > 0);
 
     // Convert the IEEE representation into a diyfp.
     //
@@ -8818,15 +15139,15 @@ boundaries compute_boundaries(FloatType value)
     constexpr int      kPrecision = std::numeric_limits<FloatType>::digits; // = p (includes the hidden bit)
     constexpr int      kBias      = std::numeric_limits<FloatType>::max_exponent - 1 + (kPrecision - 1);
     constexpr int      kMinExp    = 1 - kBias;
-    constexpr uint64_t kHiddenBit = uint64_t{1} << (kPrecision - 1); // = 2^(p-1)
+    constexpr std::uint64_t kHiddenBit = std::uint64_t{1} << (kPrecision - 1); // = 2^(p-1)
 
-    using bits_type = typename std::conditional< kPrecision == 24, uint32_t, uint64_t >::type;
+    using bits_type = typename std::conditional<kPrecision == 24, std::uint32_t, std::uint64_t >::type;
 
-    const uint64_t bits = reinterpret_bits<bits_type>(value);
-    const uint64_t E = bits >> (kPrecision - 1);
-    const uint64_t F = bits & (kHiddenBit - 1);
+    const auto bits = static_cast<std::uint64_t>(reinterpret_bits<bits_type>(value));
+    const std::uint64_t E = bits >> (kPrecision - 1);
+    const std::uint64_t F = bits & (kHiddenBit - 1);
 
-    const bool is_denormal = (E == 0);
+    const bool is_denormal = E == 0;
     const diyfp v = is_denormal
                     ? diyfp(F, kMinExp)
                     : diyfp(F + kHiddenBit, static_cast<int>(E) - kBias);
@@ -8852,7 +15173,7 @@ boundaries compute_boundaries(FloatType value)
     //      -----------------+------+------+-------------+-------------+---  (B)
     //                       v-     m-     v             m+            v+
 
-    const bool lower_boundary_is_closer = (F == 0 and E > 1);
+    const bool lower_boundary_is_closer = F == 0 && E > 1;
     const diyfp m_plus = diyfp(2 * v.f + 1, v.e - 1);
     const diyfp m_minus = lower_boundary_is_closer
                           ? diyfp(4 * v.f - 1, v.e - 2)  // (B)
@@ -8927,7 +15248,7 @@ constexpr int kGamma = -32;
 
 struct cached_power // c = f * 2^e ~= 10^k
 {
-    uint64_t f;
+    std::uint64_t f;
     int e;
     int k;
 };
@@ -8951,7 +15272,7 @@ inline cached_power get_cached_power_for_binary_exponent(int e)
     //      ==> 2^(q - 1 + alpha) <= c * 2^(e + q)
     //      ==> 2^(alpha - e - 1) <= c
     //
-    // If c were an exakt power of ten, i.e. c = 10^k, one may determine k as
+    // If c were an exact power of ten, i.e. c = 10^k, one may determine k as
     //
     //      k = ceil( log_10( 2^(alpha - e - 1) ) )
     //        = ceil( (alpha - e - 1) * log_10(2) )
@@ -8991,110 +15312,110 @@ inline cached_power get_cached_power_for_binary_exponent(int e)
     // NB:
     // Actually this function returns c, such that -60 <= e_c + e + 64 <= -34.
 
-    constexpr int kCachedPowersSize = 79;
     constexpr int kCachedPowersMinDecExp = -300;
     constexpr int kCachedPowersDecStep = 8;
 
-    static constexpr cached_power kCachedPowers[] =
+    static constexpr std::array<cached_power, 79> kCachedPowers =
     {
-        { 0xAB70FE17C79AC6CA, -1060, -300 },
-        { 0xFF77B1FCBEBCDC4F, -1034, -292 },
-        { 0xBE5691EF416BD60C, -1007, -284 },
-        { 0x8DD01FAD907FFC3C,  -980, -276 },
-        { 0xD3515C2831559A83,  -954, -268 },
-        { 0x9D71AC8FADA6C9B5,  -927, -260 },
-        { 0xEA9C227723EE8BCB,  -901, -252 },
-        { 0xAECC49914078536D,  -874, -244 },
-        { 0x823C12795DB6CE57,  -847, -236 },
-        { 0xC21094364DFB5637,  -821, -228 },
-        { 0x9096EA6F3848984F,  -794, -220 },
-        { 0xD77485CB25823AC7,  -768, -212 },
-        { 0xA086CFCD97BF97F4,  -741, -204 },
-        { 0xEF340A98172AACE5,  -715, -196 },
-        { 0xB23867FB2A35B28E,  -688, -188 },
-        { 0x84C8D4DFD2C63F3B,  -661, -180 },
-        { 0xC5DD44271AD3CDBA,  -635, -172 },
-        { 0x936B9FCEBB25C996,  -608, -164 },
-        { 0xDBAC6C247D62A584,  -582, -156 },
-        { 0xA3AB66580D5FDAF6,  -555, -148 },
-        { 0xF3E2F893DEC3F126,  -529, -140 },
-        { 0xB5B5ADA8AAFF80B8,  -502, -132 },
-        { 0x87625F056C7C4A8B,  -475, -124 },
-        { 0xC9BCFF6034C13053,  -449, -116 },
-        { 0x964E858C91BA2655,  -422, -108 },
-        { 0xDFF9772470297EBD,  -396, -100 },
-        { 0xA6DFBD9FB8E5B88F,  -369,  -92 },
-        { 0xF8A95FCF88747D94,  -343,  -84 },
-        { 0xB94470938FA89BCF,  -316,  -76 },
-        { 0x8A08F0F8BF0F156B,  -289,  -68 },
-        { 0xCDB02555653131B6,  -263,  -60 },
-        { 0x993FE2C6D07B7FAC,  -236,  -52 },
-        { 0xE45C10C42A2B3B06,  -210,  -44 },
-        { 0xAA242499697392D3,  -183,  -36 },
-        { 0xFD87B5F28300CA0E,  -157,  -28 },
-        { 0xBCE5086492111AEB,  -130,  -20 },
-        { 0x8CBCCC096F5088CC,  -103,  -12 },
-        { 0xD1B71758E219652C,   -77,   -4 },
-        { 0x9C40000000000000,   -50,    4 },
-        { 0xE8D4A51000000000,   -24,   12 },
-        { 0xAD78EBC5AC620000,     3,   20 },
-        { 0x813F3978F8940984,    30,   28 },
-        { 0xC097CE7BC90715B3,    56,   36 },
-        { 0x8F7E32CE7BEA5C70,    83,   44 },
-        { 0xD5D238A4ABE98068,   109,   52 },
-        { 0x9F4F2726179A2245,   136,   60 },
-        { 0xED63A231D4C4FB27,   162,   68 },
-        { 0xB0DE65388CC8ADA8,   189,   76 },
-        { 0x83C7088E1AAB65DB,   216,   84 },
-        { 0xC45D1DF942711D9A,   242,   92 },
-        { 0x924D692CA61BE758,   269,  100 },
-        { 0xDA01EE641A708DEA,   295,  108 },
-        { 0xA26DA3999AEF774A,   322,  116 },
-        { 0xF209787BB47D6B85,   348,  124 },
-        { 0xB454E4A179DD1877,   375,  132 },
-        { 0x865B86925B9BC5C2,   402,  140 },
-        { 0xC83553C5C8965D3D,   428,  148 },
-        { 0x952AB45CFA97A0B3,   455,  156 },
-        { 0xDE469FBD99A05FE3,   481,  164 },
-        { 0xA59BC234DB398C25,   508,  172 },
-        { 0xF6C69A72A3989F5C,   534,  180 },
-        { 0xB7DCBF5354E9BECE,   561,  188 },
-        { 0x88FCF317F22241E2,   588,  196 },
-        { 0xCC20CE9BD35C78A5,   614,  204 },
-        { 0x98165AF37B2153DF,   641,  212 },
-        { 0xE2A0B5DC971F303A,   667,  220 },
-        { 0xA8D9D1535CE3B396,   694,  228 },
-        { 0xFB9B7CD9A4A7443C,   720,  236 },
-        { 0xBB764C4CA7A44410,   747,  244 },
-        { 0x8BAB8EEFB6409C1A,   774,  252 },
-        { 0xD01FEF10A657842C,   800,  260 },
-        { 0x9B10A4E5E9913129,   827,  268 },
-        { 0xE7109BFBA19C0C9D,   853,  276 },
-        { 0xAC2820D9623BF429,   880,  284 },
-        { 0x80444B5E7AA7CF85,   907,  292 },
-        { 0xBF21E44003ACDD2D,   933,  300 },
-        { 0x8E679C2F5E44FF8F,   960,  308 },
-        { 0xD433179D9C8CB841,   986,  316 },
-        { 0x9E19DB92B4E31BA9,  1013,  324 },
+        {
+            { 0xAB70FE17C79AC6CA, -1060, -300 },
+            { 0xFF77B1FCBEBCDC4F, -1034, -292 },
+            { 0xBE5691EF416BD60C, -1007, -284 },
+            { 0x8DD01FAD907FFC3C,  -980, -276 },
+            { 0xD3515C2831559A83,  -954, -268 },
+            { 0x9D71AC8FADA6C9B5,  -927, -260 },
+            { 0xEA9C227723EE8BCB,  -901, -252 },
+            { 0xAECC49914078536D,  -874, -244 },
+            { 0x823C12795DB6CE57,  -847, -236 },
+            { 0xC21094364DFB5637,  -821, -228 },
+            { 0x9096EA6F3848984F,  -794, -220 },
+            { 0xD77485CB25823AC7,  -768, -212 },
+            { 0xA086CFCD97BF97F4,  -741, -204 },
+            { 0xEF340A98172AACE5,  -715, -196 },
+            { 0xB23867FB2A35B28E,  -688, -188 },
+            { 0x84C8D4DFD2C63F3B,  -661, -180 },
+            { 0xC5DD44271AD3CDBA,  -635, -172 },
+            { 0x936B9FCEBB25C996,  -608, -164 },
+            { 0xDBAC6C247D62A584,  -582, -156 },
+            { 0xA3AB66580D5FDAF6,  -555, -148 },
+            { 0xF3E2F893DEC3F126,  -529, -140 },
+            { 0xB5B5ADA8AAFF80B8,  -502, -132 },
+            { 0x87625F056C7C4A8B,  -475, -124 },
+            { 0xC9BCFF6034C13053,  -449, -116 },
+            { 0x964E858C91BA2655,  -422, -108 },
+            { 0xDFF9772470297EBD,  -396, -100 },
+            { 0xA6DFBD9FB8E5B88F,  -369,  -92 },
+            { 0xF8A95FCF88747D94,  -343,  -84 },
+            { 0xB94470938FA89BCF,  -316,  -76 },
+            { 0x8A08F0F8BF0F156B,  -289,  -68 },
+            { 0xCDB02555653131B6,  -263,  -60 },
+            { 0x993FE2C6D07B7FAC,  -236,  -52 },
+            { 0xE45C10C42A2B3B06,  -210,  -44 },
+            { 0xAA242499697392D3,  -183,  -36 },
+            { 0xFD87B5F28300CA0E,  -157,  -28 },
+            { 0xBCE5086492111AEB,  -130,  -20 },
+            { 0x8CBCCC096F5088CC,  -103,  -12 },
+            { 0xD1B71758E219652C,   -77,   -4 },
+            { 0x9C40000000000000,   -50,    4 },
+            { 0xE8D4A51000000000,   -24,   12 },
+            { 0xAD78EBC5AC620000,     3,   20 },
+            { 0x813F3978F8940984,    30,   28 },
+            { 0xC097CE7BC90715B3,    56,   36 },
+            { 0x8F7E32CE7BEA5C70,    83,   44 },
+            { 0xD5D238A4ABE98068,   109,   52 },
+            { 0x9F4F2726179A2245,   136,   60 },
+            { 0xED63A231D4C4FB27,   162,   68 },
+            { 0xB0DE65388CC8ADA8,   189,   76 },
+            { 0x83C7088E1AAB65DB,   216,   84 },
+            { 0xC45D1DF942711D9A,   242,   92 },
+            { 0x924D692CA61BE758,   269,  100 },
+            { 0xDA01EE641A708DEA,   295,  108 },
+            { 0xA26DA3999AEF774A,   322,  116 },
+            { 0xF209787BB47D6B85,   348,  124 },
+            { 0xB454E4A179DD1877,   375,  132 },
+            { 0x865B86925B9BC5C2,   402,  140 },
+            { 0xC83553C5C8965D3D,   428,  148 },
+            { 0x952AB45CFA97A0B3,   455,  156 },
+            { 0xDE469FBD99A05FE3,   481,  164 },
+            { 0xA59BC234DB398C25,   508,  172 },
+            { 0xF6C69A72A3989F5C,   534,  180 },
+            { 0xB7DCBF5354E9BECE,   561,  188 },
+            { 0x88FCF317F22241E2,   588,  196 },
+            { 0xCC20CE9BD35C78A5,   614,  204 },
+            { 0x98165AF37B2153DF,   641,  212 },
+            { 0xE2A0B5DC971F303A,   667,  220 },
+            { 0xA8D9D1535CE3B396,   694,  228 },
+            { 0xFB9B7CD9A4A7443C,   720,  236 },
+            { 0xBB764C4CA7A44410,   747,  244 },
+            { 0x8BAB8EEFB6409C1A,   774,  252 },
+            { 0xD01FEF10A657842C,   800,  260 },
+            { 0x9B10A4E5E9913129,   827,  268 },
+            { 0xE7109BFBA19C0C9D,   853,  276 },
+            { 0xAC2820D9623BF429,   880,  284 },
+            { 0x80444B5E7AA7CF85,   907,  292 },
+            { 0xBF21E44003ACDD2D,   933,  300 },
+            { 0x8E679C2F5E44FF8F,   960,  308 },
+            { 0xD433179D9C8CB841,   986,  316 },
+            { 0x9E19DB92B4E31BA9,  1013,  324 },
+        }
     };
 
     // This computation gives exactly the same results for k as
     //      k = ceil((kAlpha - e - 1) * 0.30102999566398114)
     // for |e| <= 1500, but doesn't require floating-point operations.
     // NB: log_10(2) ~= 78913 / 2^18
-    assert(e >= -1500);
-    assert(e <=  1500);
+    JSON_ASSERT(e >= -1500);
+    JSON_ASSERT(e <=  1500);
     const int f = kAlpha - e - 1;
-    const int k = (f * 78913) / (1 << 18) + (f > 0);
+    const int k = (f * 78913) / (1 << 18) + static_cast<int>(f > 0);
 
     const int index = (-kCachedPowersMinDecExp + k + (kCachedPowersDecStep - 1)) / kCachedPowersDecStep;
-    assert(index >= 0);
-    assert(index < kCachedPowersSize);
-    static_cast<void>(kCachedPowersSize); // Fix warning.
+    JSON_ASSERT(index >= 0);
+    JSON_ASSERT(static_cast<std::size_t>(index) < kCachedPowers.size());
 
-    const cached_power cached = kCachedPowers[index];
-    assert(kAlpha <= cached.e + e + 64);
-    assert(kGamma >= cached.e + e + 64);
+    const cached_power cached = kCachedPowers[static_cast<std::size_t>(index)];
+    JSON_ASSERT(kAlpha <= cached.e + e + 64);
+    JSON_ASSERT(kGamma >= cached.e + e + 64);
 
     return cached;
 }
@@ -9103,7 +15424,7 @@ inline cached_power get_cached_power_for_binary_exponent(int e)
 For n != 0, returns k, such that pow10 := 10^(k-1) <= n < 10^k.
 For n == 0, returns 1 and sets pow10 := 1.
 */
-inline int find_largest_pow10(const uint32_t n, uint32_t& pow10)
+inline int find_largest_pow10(const std::uint32_t n, std::uint32_t& pow10)
 {
     // LCOV_EXCL_START
     if (n >= 1000000000)
@@ -9112,60 +15433,58 @@ inline int find_largest_pow10(const uint32_t n, uint32_t& pow10)
         return 10;
     }
     // LCOV_EXCL_STOP
-    else if (n >= 100000000)
+    if (n >= 100000000)
     {
         pow10 = 100000000;
         return  9;
     }
-    else if (n >= 10000000)
+    if (n >= 10000000)
     {
         pow10 = 10000000;
         return  8;
     }
-    else if (n >= 1000000)
+    if (n >= 1000000)
     {
         pow10 = 1000000;
         return  7;
     }
-    else if (n >= 100000)
+    if (n >= 100000)
     {
         pow10 = 100000;
         return  6;
     }
-    else if (n >= 10000)
+    if (n >= 10000)
     {
         pow10 = 10000;
         return  5;
     }
-    else if (n >= 1000)
+    if (n >= 1000)
     {
         pow10 = 1000;
         return  4;
     }
-    else if (n >= 100)
+    if (n >= 100)
     {
         pow10 = 100;
         return  3;
     }
-    else if (n >= 10)
+    if (n >= 10)
     {
         pow10 = 10;
         return  2;
     }
-    else
-    {
-        pow10 = 1;
-        return 1;
-    }
+
+    pow10 = 1;
+    return 1;
 }
 
-inline void grisu2_round(char* buf, int len, uint64_t dist, uint64_t delta,
-                         uint64_t rest, uint64_t ten_k)
+inline void grisu2_round(char* buf, int len, std::uint64_t dist, std::uint64_t delta,
+                         std::uint64_t rest, std::uint64_t ten_k)
 {
-    assert(len >= 1);
-    assert(dist <= delta);
-    assert(rest <= delta);
-    assert(ten_k > 0);
+    JSON_ASSERT(len >= 1);
+    JSON_ASSERT(dist <= delta);
+    JSON_ASSERT(rest <= delta);
+    JSON_ASSERT(ten_k > 0);
 
     //               <--------------------------- delta ---->
     //                                  <---- dist --------->
@@ -9187,10 +15506,10 @@ inline void grisu2_round(char* buf, int len, uint64_t dist, uint64_t delta,
     // integer arithmetic.
 
     while (rest < dist
-            and delta - rest >= ten_k
-            and (rest + ten_k < dist or dist - rest > rest + ten_k - dist))
+            && delta - rest >= ten_k
+            && (rest + ten_k < dist || dist - rest > rest + ten_k - dist))
     {
-        assert(buf[len - 1] != '0');
+        JSON_ASSERT(buf[len - 1] != '0');
         buf[len - 1]--;
         rest += ten_k;
     }
@@ -9218,11 +15537,11 @@ inline void grisu2_digit_gen(char* buffer, int& length, int& decimal_exponent,
     // Grisu2 generates the digits of M+ from left to right and stops as soon as
     // V is in [M-,M+].
 
-    assert(M_plus.e >= kAlpha);
-    assert(M_plus.e <= kGamma);
+    JSON_ASSERT(M_plus.e >= kAlpha);
+    JSON_ASSERT(M_plus.e <= kGamma);
 
-    uint64_t delta = diyfp::sub(M_plus, M_minus).f; // (significand of (M+ - M-), implicit exponent is e)
-    uint64_t dist  = diyfp::sub(M_plus, w      ).f; // (significand of (M+ - w ), implicit exponent is e)
+    std::uint64_t delta = diyfp::sub(M_plus, M_minus).f; // (significand of (M+ - M-), implicit exponent is e)
+    std::uint64_t dist  = diyfp::sub(M_plus, w      ).f; // (significand of (M+ - w ), implicit exponent is e)
 
     // Split M+ = f * 2^e into two parts p1 and p2 (note: e < 0):
     //
@@ -9231,18 +15550,18 @@ inline void grisu2_digit_gen(char* buffer, int& length, int& decimal_exponent,
     //         = ((p1        ) * 2^-e + (p2        )) * 2^e
     //         = p1 + p2 * 2^e
 
-    const diyfp one(uint64_t{1} << -M_plus.e, M_plus.e);
+    const diyfp one(std::uint64_t{1} << -M_plus.e, M_plus.e);
 
-    uint32_t p1 = static_cast<uint32_t>(M_plus.f >> -one.e); // p1 = f div 2^-e (Since -e >= 32, p1 fits into a 32-bit int.)
-    uint64_t p2 = M_plus.f & (one.f - 1);                    // p2 = f mod 2^-e
+    auto p1 = static_cast<std::uint32_t>(M_plus.f >> -one.e); // p1 = f div 2^-e (Since -e >= 32, p1 fits into a 32-bit int.)
+    std::uint64_t p2 = M_plus.f & (one.f - 1);                    // p2 = f mod 2^-e
 
     // 1)
     //
     // Generate the digits of the integral part p1 = d[n-1]...d[1]d[0]
 
-    assert(p1 > 0);
+    JSON_ASSERT(p1 > 0);
 
-    uint32_t pow10;
+    std::uint32_t pow10{};
     const int k = find_largest_pow10(p1, pow10);
 
     //      10^(k-1) <= p1 < 10^k, pow10 = 10^(k-1)
@@ -9270,13 +15589,13 @@ inline void grisu2_digit_gen(char* buffer, int& length, int& decimal_exponent,
         //      M+ = buffer * 10^n + (p1 + p2 * 2^e)    (buffer = 0 for n = k)
         //      pow10 = 10^(n-1) <= p1 < 10^n
         //
-        const uint32_t d = p1 / pow10;  // d = p1 div 10^(n-1)
-        const uint32_t r = p1 % pow10;  // r = p1 mod 10^(n-1)
+        const std::uint32_t d = p1 / pow10;  // d = p1 div 10^(n-1)
+        const std::uint32_t r = p1 % pow10;  // r = p1 mod 10^(n-1)
         //
         //      M+ = buffer * 10^n + (d * 10^(n-1) + r) + p2 * 2^e
         //         = (buffer * 10 + d) * 10^(n-1) + (r + p2 * 2^e)
         //
-        assert(d <= 9);
+        JSON_ASSERT(d <= 9);
         buffer[length++] = static_cast<char>('0' + d); // buffer := buffer * 10 + d
         //
         //      M+ = buffer * 10^(n-1) + (r + p2 * 2^e)
@@ -9296,7 +15615,7 @@ inline void grisu2_digit_gen(char* buffer, int& length, int& decimal_exponent,
         // Note:
         // Since rest and delta share the same exponent e, it suffices to
         // compare the significands.
-        const uint64_t rest = (uint64_t{p1} << -one.e) + p2;
+        const std::uint64_t rest = (std::uint64_t{p1} << -one.e) + p2;
         if (rest <= delta)
         {
             // V = buffer * 10^n, with M- <= V <= M+.
@@ -9312,7 +15631,7 @@ inline void grisu2_digit_gen(char* buffer, int& length, int& decimal_exponent,
             //
             //      10^n = (10^n * 2^-e) * 2^e = ulp * 2^e
             //
-            const uint64_t ten_n = uint64_t{pow10} << -one.e;
+            const std::uint64_t ten_n = std::uint64_t{pow10} << -one.e;
             grisu2_round(buffer, length, dist, delta, rest, ten_n);
 
             return;
@@ -9363,7 +15682,7 @@ inline void grisu2_digit_gen(char* buffer, int& length, int& decimal_exponent,
     //
     // and stop as soon as 10^-m * r * 2^e <= delta * 2^e
 
-    assert(p2 > delta);
+    JSON_ASSERT(p2 > delta);
 
     int m = 0;
     for (;;)
@@ -9374,16 +15693,16 @@ inline void grisu2_digit_gen(char* buffer, int& length, int& decimal_exponent,
         //         = buffer * 10^-m + 10^-m * (1/10 * (10 * p2)                   ) * 2^e
         //         = buffer * 10^-m + 10^-m * (1/10 * ((10*p2 div 2^-e) * 2^-e + (10*p2 mod 2^-e)) * 2^e
         //
-        assert(p2 <= UINT64_MAX / 10);
+        JSON_ASSERT(p2 <= (std::numeric_limits<std::uint64_t>::max)() / 10);
         p2 *= 10;
-        const uint64_t d = p2 >> -one.e;     // d = (10 * p2) div 2^-e
-        const uint64_t r = p2 & (one.f - 1); // r = (10 * p2) mod 2^-e
+        const std::uint64_t d = p2 >> -one.e;     // d = (10 * p2) div 2^-e
+        const std::uint64_t r = p2 & (one.f - 1); // r = (10 * p2) mod 2^-e
         //
         //      M+ = buffer * 10^-m + 10^-m * (1/10 * (d * 2^-e + r) * 2^e
         //         = buffer * 10^-m + 10^-m * (1/10 * (d + r * 2^e))
         //         = (buffer * 10 + d) * 10^(-m-1) + 10^(-m-1) * r * 2^e
         //
-        assert(d <= 9);
+        JSON_ASSERT(d <= 9);
         buffer[length++] = static_cast<char>('0' + d); // buffer := buffer * 10 + d
         //
         //      M+ = buffer * 10^(-m-1) + 10^(-m-1) * r * 2^e
@@ -9417,7 +15736,7 @@ inline void grisu2_digit_gen(char* buffer, int& length, int& decimal_exponent,
     //
     //      10^m * 10^-m = 1 = 2^-e * 2^e = ten_m * 2^e
     //
-    const uint64_t ten_m = one.f;
+    const std::uint64_t ten_m = one.f;
     grisu2_round(buffer, length, dist, delta, p2, ten_m);
 
     // By construction this algorithm generates the shortest possible decimal
@@ -9440,11 +15759,12 @@ v = buf * 10^decimal_exponent
 len is the length of the buffer (number of decimal digits)
 The buffer must be large enough, i.e. >= max_digits10.
 */
+JSON_HEDLEY_NON_NULL(1)
 inline void grisu2(char* buf, int& len, int& decimal_exponent,
                    diyfp m_minus, diyfp v, diyfp m_plus)
 {
-    assert(m_plus.e == m_minus.e);
-    assert(m_plus.e == v.e);
+    JSON_ASSERT(m_plus.e == m_minus.e);
+    JSON_ASSERT(m_plus.e == v.e);
 
     //  --------(-----------------------+-----------------------)--------    (A)
     //          m-                      v                       m+
@@ -9498,14 +15818,15 @@ v = buf * 10^decimal_exponent
 len is the length of the buffer (number of decimal digits)
 The buffer must be large enough, i.e. >= max_digits10.
 */
-template <typename FloatType>
+template<typename FloatType>
+JSON_HEDLEY_NON_NULL(1)
 void grisu2(char* buf, int& len, int& decimal_exponent, FloatType value)
 {
     static_assert(diyfp::kPrecision >= std::numeric_limits<FloatType>::digits + 3,
                   "internal error: not enough precision");
 
-    assert(std::isfinite(value));
-    assert(value > 0);
+    JSON_ASSERT(std::isfinite(value));
+    JSON_ASSERT(value > 0);
 
     // If the neighbors (and boundaries) of 'value' are always computed for double-precision
     // numbers, all float's can be recovered using strtod (and strtof). However, the resulting
@@ -9513,7 +15834,7 @@ void grisu2(char* buf, int& len, int& decimal_exponent, FloatType value)
     //
     // The documentation for 'std::to_chars' (https://en.cppreference.com/w/cpp/utility/to_chars)
     // says "value is converted to a string as if by std::sprintf in the default ("C") locale"
-    // and since sprintf promotes float's to double's, I think this is exactly what 'std::to_chars'
+    // and since sprintf promotes floats to doubles, I think this is exactly what 'std::to_chars'
     // does.
     // On the other hand, the documentation for 'std::to_chars' requires that "parsing the
     // representation using the corresponding std::from_chars function recovers value exactly". That
@@ -9537,10 +15858,12 @@ void grisu2(char* buf, int& len, int& decimal_exponent, FloatType value)
 @return a pointer to the element following the exponent.
 @pre -1000 < e < 1000
 */
+JSON_HEDLEY_NON_NULL(1)
+JSON_HEDLEY_RETURNS_NON_NULL
 inline char* append_exponent(char* buf, int e)
 {
-    assert(e > -1000);
-    assert(e <  1000);
+    JSON_ASSERT(e > -1000);
+    JSON_ASSERT(e <  1000);
 
     if (e < 0)
     {
@@ -9552,7 +15875,7 @@ inline char* append_exponent(char* buf, int e)
         *buf++ = '+';
     }
 
-    uint32_t k = static_cast<uint32_t>(e);
+    auto k = static_cast<std::uint32_t>(e);
     if (k < 10)
     {
         // Always print at least two digits in the exponent.
@@ -9587,11 +15910,13 @@ notation. Otherwise it will be printed in exponential notation.
 @pre min_exp < 0
 @pre max_exp > 0
 */
+JSON_HEDLEY_NON_NULL(1)
+JSON_HEDLEY_RETURNS_NON_NULL
 inline char* format_buffer(char* buf, int len, int decimal_exponent,
                            int min_exp, int max_exp)
 {
-    assert(min_exp < 0);
-    assert(max_exp > 0);
+    JSON_ASSERT(min_exp < 0);
+    JSON_ASSERT(max_exp > 0);
 
     const int k = len;
     const int n = len + decimal_exponent;
@@ -9600,40 +15925,40 @@ inline char* format_buffer(char* buf, int len, int decimal_exponent,
     // k is the length of the buffer (number of decimal digits)
     // n is the position of the decimal point relative to the start of the buffer.
 
-    if (k <= n and n <= max_exp)
+    if (k <= n && n <= max_exp)
     {
         // digits[000]
         // len <= max_exp + 2
 
-        std::memset(buf + k, '0', static_cast<size_t>(n - k));
+        std::memset(buf + k, '0', static_cast<size_t>(n) - static_cast<size_t>(k));
         // Make it look like a floating-point number (#362, #378)
         buf[n + 0] = '.';
         buf[n + 1] = '0';
-        return buf + (n + 2);
+        return buf + (static_cast<size_t>(n) + 2);
     }
 
-    if (0 < n and n <= max_exp)
+    if (0 < n && n <= max_exp)
     {
         // dig.its
         // len <= max_digits10 + 1
 
-        assert(k > n);
+        JSON_ASSERT(k > n);
 
-        std::memmove(buf + (n + 1), buf + n, static_cast<size_t>(k - n));
+        std::memmove(buf + (static_cast<size_t>(n) + 1), buf + n, static_cast<size_t>(k) - static_cast<size_t>(n));
         buf[n] = '.';
-        return buf + (k + 1);
+        return buf + (static_cast<size_t>(k) + 1U);
     }
 
-    if (min_exp < n and n <= 0)
+    if (min_exp < n && n <= 0)
     {
         // 0.[000]digits
         // len <= 2 + (-min_exp - 1) + max_digits10
 
-        std::memmove(buf + (2 + -n), buf, static_cast<size_t>(k));
+        std::memmove(buf + (2 + static_cast<size_t>(-n)), buf, static_cast<size_t>(k));
         buf[0] = '0';
         buf[1] = '.';
         std::memset(buf + 2, '0', static_cast<size_t>(-n));
-        return buf + (2 + (-n) + k);
+        return buf + (2U + static_cast<size_t>(-n) + static_cast<size_t>(k));
     }
 
     if (k == 1)
@@ -9648,9 +15973,9 @@ inline char* format_buffer(char* buf, int len, int decimal_exponent,
         // d.igitsE+123
         // len <= max_digits10 + 1 + 5
 
-        std::memmove(buf + 2, buf + 1, static_cast<size_t>(k - 1));
+        std::memmove(buf + 2, buf + 1, static_cast<size_t>(k) - 1);
         buf[1] = '.';
-        buf += 1 + k;
+        buf += 1 + static_cast<size_t>(k);
     }
 
     *buf++ = 'e';
@@ -9669,11 +15994,13 @@ format. Returns an iterator pointing past-the-end of the decimal representation.
 @note The buffer must be large enough.
 @note The result is NOT null-terminated.
 */
-template <typename FloatType>
-char* to_chars(char* first, char* last, FloatType value)
+template<typename FloatType>
+JSON_HEDLEY_NON_NULL(1, 2)
+JSON_HEDLEY_RETURNS_NON_NULL
+char* to_chars(char* first, const char* last, FloatType value)
 {
     static_cast<void>(last); // maybe unused - fix warning
-    assert(std::isfinite(value));
+    JSON_ASSERT(std::isfinite(value));
 
     // Use signbit(value) instead of (value < 0) since signbit works for -0.
     if (std::signbit(value))
@@ -9682,6 +16009,10 @@ char* to_chars(char* first, char* last, FloatType value)
         *first++ = '-';
     }
 
+#ifdef __GNUC__
+#pragma GCC diagnostic push
+#pragma GCC diagnostic ignored "-Wfloat-equal"
+#endif
     if (value == 0) // +-0
     {
         *first++ = '0';
@@ -9690,8 +16021,11 @@ char* to_chars(char* first, char* last, FloatType value)
         *first++ = '0';
         return first;
     }
+#ifdef __GNUC__
+#pragma GCC diagnostic pop
+#endif
 
-    assert(last - first >= std::numeric_limits<FloatType>::max_digits10);
+    JSON_ASSERT(last - first >= std::numeric_limits<FloatType>::max_digits10);
 
     // Compute v = buffer * 10^decimal_exponent.
     // The decimal digits are stored in the buffer, which needs to be interpreted
@@ -9701,16 +16035,16 @@ char* to_chars(char* first, char* last, FloatType value)
     int decimal_exponent = 0;
     dtoa_impl::grisu2(first, len, decimal_exponent, value);
 
-    assert(len <= std::numeric_limits<FloatType>::max_digits10);
+    JSON_ASSERT(len <= std::numeric_limits<FloatType>::max_digits10);
 
     // Format the buffer like printf("%.*g", prec, value)
     constexpr int kMinExp = -4;
     // Use digits10 here to increase compatibility with version 2.
     constexpr int kMaxExp = std::numeric_limits<FloatType>::digits10;
 
-    assert(last - first >= kMaxExp + 2);
-    assert(last - first >= 2 + (-kMinExp - 1) + std::numeric_limits<FloatType>::max_digits10);
-    assert(last - first >= std::numeric_limits<FloatType>::max_digits10 + 6);
+    JSON_ASSERT(last - first >= kMaxExp + 2);
+    JSON_ASSERT(last - first >= 2 + (-kMinExp - 1) + std::numeric_limits<FloatType>::max_digits10);
+    JSON_ASSERT(last - first >= std::numeric_limits<FloatType>::max_digits10 + 6);
 
     return dtoa_impl::format_buffer(first, len, decimal_exponent, kMinExp, kMaxExp);
 }
@@ -9718,10 +16052,14 @@ char* to_chars(char* first, char* last, FloatType value)
 } // namespace detail
 } // namespace nlohmann
 
+// #include <nlohmann/detail/exceptions.hpp>
+
 // #include <nlohmann/detail/macro_scope.hpp>
 
 // #include <nlohmann/detail/meta/cpp_future.hpp>
 
+// #include <nlohmann/detail/output/binary_writer.hpp>
+
 // #include <nlohmann/detail/output/output_adapters.hpp>
 
 // #include <nlohmann/detail/value_t.hpp>
@@ -9735,6 +16073,14 @@ namespace detail
 // serialization //
 ///////////////////
 
+/// how to treat decoding errors
+enum class error_handler_t
+{
+    strict,  ///< throw a type_error exception in case of invalid UTF-8
+    replace, ///< replace invalid UTF-8 sequences with U+FFFD
+    ignore   ///< ignore invalid UTF-8 sequences
+};
+
 template<typename BasicJsonType>
 class serializer
 {
@@ -9742,24 +16088,33 @@ class serializer
     using number_float_t = typename BasicJsonType::number_float_t;
     using number_integer_t = typename BasicJsonType::number_integer_t;
     using number_unsigned_t = typename BasicJsonType::number_unsigned_t;
-    static constexpr uint8_t UTF8_ACCEPT = 0;
-    static constexpr uint8_t UTF8_REJECT = 1;
+    using binary_char_t = typename BasicJsonType::binary_t::value_type;
+    static constexpr std::uint8_t UTF8_ACCEPT = 0;
+    static constexpr std::uint8_t UTF8_REJECT = 1;
 
   public:
     /*!
     @param[in] s  output stream to serialize to
     @param[in] ichar  indentation character to use
+    @param[in] error_handler_  how to react on decoding errors
     */
-    serializer(output_adapter_t<char> s, const char ichar)
-        : o(std::move(s)), loc(std::localeconv()),
-          thousands_sep(loc->thousands_sep == nullptr ? '\0' : * (loc->thousands_sep)),
-          decimal_point(loc->decimal_point == nullptr ? '\0' : * (loc->decimal_point)),
-          indent_char(ichar), indent_string(512, indent_char)
+    serializer(output_adapter_t<char> s, const char ichar,
+               error_handler_t error_handler_ = error_handler_t::strict)
+        : o(std::move(s))
+        , loc(std::localeconv())
+        , thousands_sep(loc->thousands_sep == nullptr ? '\0' : std::char_traits<char>::to_char_type(* (loc->thousands_sep)))
+        , decimal_point(loc->decimal_point == nullptr ? '\0' : std::char_traits<char>::to_char_type(* (loc->decimal_point)))
+        , indent_char(ichar)
+        , indent_string(512, indent_char)
+        , error_handler(error_handler_)
     {}
 
     // delete because of pointer members
     serializer(const serializer&) = delete;
     serializer& operator=(const serializer&) = delete;
+    serializer(serializer&&) = delete;
+    serializer& operator=(serializer&&) = delete;
+    ~serializer() = default;
 
     /*!
     @brief internal implementation of the serialization function
@@ -9772,13 +16127,19 @@ class serializer
     - strings and object keys are escaped using `escape_string()`
     - integer numbers are converted implicitly via `operator<<`
     - floating-point numbers are converted to a string using `"%g"` format
+    - binary values are serialized as objects containing the subtype and the
+      byte array
 
-    @param[in] val             value to serialize
-    @param[in] pretty_print    whether the output shall be pretty-printed
-    @param[in] indent_step     the indent level
-    @param[in] current_indent  the current indent level (only used internally)
+    @param[in] val               value to serialize
+    @param[in] pretty_print      whether the output shall be pretty-printed
+    @param[in] ensure_ascii If @a ensure_ascii is true, all non-ASCII characters
+    in the output are escaped with `\uXXXX` sequences, and the result consists
+    of ASCII characters only.
+    @param[in] indent_step       the indent level
+    @param[in] current_indent    the current indent level (only used internally)
     */
-    void dump(const BasicJsonType& val, const bool pretty_print,
+    void dump(const BasicJsonType& val,
+              const bool pretty_print,
               const bool ensure_ascii,
               const unsigned int indent_step,
               const unsigned int current_indent = 0)
@@ -9799,7 +16160,7 @@ class serializer
 
                     // variable to hold indentation for recursive calls
                     const auto new_indent = current_indent + indent_step;
-                    if (JSON_UNLIKELY(indent_string.size() < new_indent))
+                    if (JSON_HEDLEY_UNLIKELY(indent_string.size() < new_indent))
                     {
                         indent_string.resize(indent_string.size() * 2, ' ');
                     }
@@ -9817,8 +16178,8 @@ class serializer
                     }
 
                     // last element
-                    assert(i != val.m_value.object->cend());
-                    assert(std::next(i) == val.m_value.object->cend());
+                    JSON_ASSERT(i != val.m_value.object->cend());
+                    JSON_ASSERT(std::next(i) == val.m_value.object->cend());
                     o->write_characters(indent_string.c_str(), new_indent);
                     o->write_character('\"');
                     dump_escaped(i->first, ensure_ascii);
@@ -9845,8 +16206,8 @@ class serializer
                     }
 
                     // last element
-                    assert(i != val.m_value.object->cend());
-                    assert(std::next(i) == val.m_value.object->cend());
+                    JSON_ASSERT(i != val.m_value.object->cend());
+                    JSON_ASSERT(std::next(i) == val.m_value.object->cend());
                     o->write_character('\"');
                     dump_escaped(i->first, ensure_ascii);
                     o->write_characters("\":", 2);
@@ -9872,7 +16233,7 @@ class serializer
 
                     // variable to hold indentation for recursive calls
                     const auto new_indent = current_indent + indent_step;
-                    if (JSON_UNLIKELY(indent_string.size() < new_indent))
+                    if (JSON_HEDLEY_UNLIKELY(indent_string.size() < new_indent))
                     {
                         indent_string.resize(indent_string.size() * 2, ' ');
                     }
@@ -9887,7 +16248,7 @@ class serializer
                     }
 
                     // last element
-                    assert(not val.m_value.array->empty());
+                    JSON_ASSERT(!val.m_value.array->empty());
                     o->write_characters(indent_string.c_str(), new_indent);
                     dump(val.m_value.array->back(), true, ensure_ascii, indent_step, new_indent);
 
@@ -9908,7 +16269,7 @@ class serializer
                     }
 
                     // last element
-                    assert(not val.m_value.array->empty());
+                    JSON_ASSERT(!val.m_value.array->empty());
                     dump(val.m_value.array->back(), false, ensure_ascii, indent_step, current_indent);
 
                     o->write_character(']');
@@ -9925,6 +16286,79 @@ class serializer
                 return;
             }
 
+            case value_t::binary:
+            {
+                if (pretty_print)
+                {
+                    o->write_characters("{\n", 2);
+
+                    // variable to hold indentation for recursive calls
+                    const auto new_indent = current_indent + indent_step;
+                    if (JSON_HEDLEY_UNLIKELY(indent_string.size() < new_indent))
+                    {
+                        indent_string.resize(indent_string.size() * 2, ' ');
+                    }
+
+                    o->write_characters(indent_string.c_str(), new_indent);
+
+                    o->write_characters("\"bytes\": [", 10);
+
+                    if (!val.m_value.binary->empty())
+                    {
+                        for (auto i = val.m_value.binary->cbegin();
+                                i != val.m_value.binary->cend() - 1; ++i)
+                        {
+                            dump_integer(*i);
+                            o->write_characters(", ", 2);
+                        }
+                        dump_integer(val.m_value.binary->back());
+                    }
+
+                    o->write_characters("],\n", 3);
+                    o->write_characters(indent_string.c_str(), new_indent);
+
+                    o->write_characters("\"subtype\": ", 11);
+                    if (val.m_value.binary->has_subtype())
+                    {
+                        dump_integer(val.m_value.binary->subtype());
+                    }
+                    else
+                    {
+                        o->write_characters("null", 4);
+                    }
+                    o->write_character('\n');
+                    o->write_characters(indent_string.c_str(), current_indent);
+                    o->write_character('}');
+                }
+                else
+                {
+                    o->write_characters("{\"bytes\":[", 10);
+
+                    if (!val.m_value.binary->empty())
+                    {
+                        for (auto i = val.m_value.binary->cbegin();
+                                i != val.m_value.binary->cend() - 1; ++i)
+                        {
+                            dump_integer(*i);
+                            o->write_character(',');
+                        }
+                        dump_integer(val.m_value.binary->back());
+                    }
+
+                    o->write_characters("],\"subtype\":", 12);
+                    if (val.m_value.binary->has_subtype())
+                    {
+                        dump_integer(val.m_value.binary->subtype());
+                        o->write_character('}');
+                    }
+                    else
+                    {
+                        o->write_characters("null}", 5);
+                    }
+                }
+                return;
+            }
+
             case value_t::boolean:
             {
                 if (val.m_value.boolean)
@@ -9967,10 +16401,13 @@ class serializer
                 o->write_characters("null", 4);
                 return;
             }
+
+            default:            // LCOV_EXCL_LINE
+                JSON_ASSERT(false); // NOLINT(cert-dcl03-c,hicpp-static-assert,misc-static-assert) LCOV_EXCL_LINE
         }
     }
 
-  private:
+  JSON_PRIVATE_UNLESS_TESTED:
     /*!
     @brief dump escaped string
 
@@ -9987,13 +16424,17 @@ class serializer
     */
     void dump_escaped(const string_t& s, const bool ensure_ascii)
     {
-        uint32_t codepoint;
-        uint8_t state = UTF8_ACCEPT;
+        std::uint32_t codepoint{};
+        std::uint8_t state = UTF8_ACCEPT;
         std::size_t bytes = 0;  // number of bytes written to string_buffer
 
+        // number of bytes written at the point of the last valid byte
+        std::size_t bytes_after_last_accept = 0;
+        std::size_t undumped_chars = 0;
+
         for (std::size_t i = 0; i < s.size(); ++i)
         {
-            const auto byte = static_cast<uint8_t>(s[i]);
+            const auto byte = static_cast<std::uint8_t>(s[i]);
 
             switch (decode(state, codepoint, byte))
             {
@@ -10054,19 +16495,21 @@ class serializer
                         {
                             // escape control characters (0x00..0x1F) or, if
                             // ensure_ascii parameter is used, non-ASCII characters
-                            if ((codepoint <= 0x1F) or (ensure_ascii and (codepoint >= 0x7F)))
+                            if ((codepoint <= 0x1F) || (ensure_ascii && (codepoint >= 0x7F)))
                             {
                                 if (codepoint <= 0xFFFF)
                                 {
-                                    std::snprintf(string_buffer.data() + bytes, 7, "\\u%04x",
-                                                  static_cast<uint16_t>(codepoint));
+                                    // NOLINTNEXTLINE(cppcoreguidelines-pro-type-vararg,hicpp-vararg)
+                                    static_cast<void>((std::snprintf)(string_buffer.data() + bytes, 7, "\\u%04x",
+                                                                      static_cast<std::uint16_t>(codepoint)));
                                     bytes += 6;
                                 }
                                 else
                                 {
-                                    std::snprintf(string_buffer.data() + bytes, 13, "\\u%04x\\u%04x",
-                                                  static_cast<uint16_t>(0xD7C0 + (codepoint >> 10)),
-                                                  static_cast<uint16_t>(0xDC00 + (codepoint & 0x3FF)));
+                                    // NOLINTNEXTLINE(cppcoreguidelines-pro-type-vararg,hicpp-vararg)
+                                    static_cast<void>((std::snprintf)(string_buffer.data() + bytes, 13, "\\u%04x\\u%04x",
+                                                                      static_cast<std::uint16_t>(0xD7C0u + (codepoint >> 10u)),
+                                                                      static_cast<std::uint16_t>(0xDC00u + (codepoint & 0x3FFu))));
                                     bytes += 12;
                                 }
                             }
@@ -10088,29 +16531,99 @@ class serializer
                         o->write_characters(string_buffer.data(), bytes);
                         bytes = 0;
                     }
+
+                    // remember the byte position of this accept
+                    bytes_after_last_accept = bytes;
+                    undumped_chars = 0;
                     break;
                 }
 
                 case UTF8_REJECT:  // decode found invalid UTF-8 byte
                 {
-                    std::string sn(3, '\0');
-                    snprintf(&sn[0], sn.size(), "%.2X", byte);
-                    JSON_THROW(type_error::create(316, "invalid UTF-8 byte at index " + std::to_string(i) + ": 0x" + sn));
+                    switch (error_handler)
+                    {
+                        case error_handler_t::strict:
+                        {
+                            std::stringstream ss;
+                            ss << std::uppercase << std::setfill('0') << std::setw(2) << std::hex << (byte | 0);
+                            JSON_THROW(type_error::create(316, "invalid UTF-8 byte at index " + std::to_string(i) + ": 0x" + ss.str(), BasicJsonType()));
+                        }
+
+                        case error_handler_t::ignore:
+                        case error_handler_t::replace:
+                        {
+                            // in case we saw this character the first time, we
+                            // would like to read it again, because the byte
+                            // may be OK for itself, but just not OK for the
+                            // previous sequence
+                            if (undumped_chars > 0)
+                            {
+                                --i;
+                            }
+
+                            // reset length buffer to the last accepted index;
+                            // thus removing/ignoring the invalid characters
+                            bytes = bytes_after_last_accept;
+
+                            if (error_handler == error_handler_t::replace)
+                            {
+                                // add a replacement character
+                                if (ensure_ascii)
+                                {
+                                    string_buffer[bytes++] = '\\';
+                                    string_buffer[bytes++] = 'u';
+                                    string_buffer[bytes++] = 'f';
+                                    string_buffer[bytes++] = 'f';
+                                    string_buffer[bytes++] = 'f';
+                                    string_buffer[bytes++] = 'd';
+                                }
+                                else
+                                {
+                                    string_buffer[bytes++] = detail::binary_writer<BasicJsonType, char>::to_char_type('\xEF');
+                                    string_buffer[bytes++] = detail::binary_writer<BasicJsonType, char>::to_char_type('\xBF');
+                                    string_buffer[bytes++] = detail::binary_writer<BasicJsonType, char>::to_char_type('\xBD');
+                                }
+
+                                // write buffer and reset index; there must be 13 bytes
+                                // left, as this is the maximal number of bytes to be
+                                // written ("\uxxxx\uxxxx\0") for one code point
+                                if (string_buffer.size() - bytes < 13)
+                                {
+                                    o->write_characters(string_buffer.data(), bytes);
+                                    bytes = 0;
+                                }
+
+                                bytes_after_last_accept = bytes;
+                            }
+
+                            undumped_chars = 0;
+
+                            // continue processing the string
+                            state = UTF8_ACCEPT;
+                            break;
+                        }
+
+                        default:            // LCOV_EXCL_LINE
+                            JSON_ASSERT(false); // NOLINT(cert-dcl03-c,hicpp-static-assert,misc-static-assert) LCOV_EXCL_LINE
+                    }
+                    break;
                 }
 
                 default:  // decode found yet incomplete multi-byte code point
                 {
-                    if (not ensure_ascii)
+                    if (!ensure_ascii)
                     {
                         // code point will not be escaped - copy byte to buffer
                         string_buffer[bytes++] = s[i];
                     }
+                    ++undumped_chars;
                     break;
                 }
             }
         }
 
-        if (JSON_LIKELY(state == UTF8_ACCEPT))
+        // we finished processing the string
+        if (JSON_HEDLEY_LIKELY(state == UTF8_ACCEPT))
         {
             // write buffer
             if (bytes > 0)
@@ -10121,12 +16634,92 @@ class serializer
         else
         {
             // we finish reading, but do not accept: string was incomplete
-            std::string sn(3, '\0');
-            snprintf(&sn[0], sn.size(), "%.2X", static_cast<uint8_t>(s.back()));
-            JSON_THROW(type_error::create(316, "incomplete UTF-8 string; last byte: 0x" + sn));
+            switch (error_handler)
+            {
+                case error_handler_t::strict:
+                {
+                    std::stringstream ss;
+                    ss << std::uppercase << std::setfill('0') << std::setw(2) << std::hex << (static_cast<std::uint8_t>(s.back()) | 0);
+                    JSON_THROW(type_error::create(316, "incomplete UTF-8 string; last byte: 0x" + ss.str(), BasicJsonType()));
+                }
+
+                case error_handler_t::ignore:
+                {
+                    // write all accepted bytes
+                    o->write_characters(string_buffer.data(), bytes_after_last_accept);
+                    break;
+                }
+
+                case error_handler_t::replace:
+                {
+                    // write all accepted bytes
+                    o->write_characters(string_buffer.data(), bytes_after_last_accept);
+                    // add a replacement character
+                    if (ensure_ascii)
+                    {
+                        o->write_characters("\\ufffd", 6);
+                    }
+                    else
+                    {
+                        o->write_characters("\xEF\xBF\xBD", 3);
+                    }
+                    break;
+                }
+
+                default:            // LCOV_EXCL_LINE
+                    JSON_ASSERT(false); // NOLINT(cert-dcl03-c,hicpp-static-assert,misc-static-assert) LCOV_EXCL_LINE
+            }
         }
     }
 
+  private:
+    /*!
+    @brief count digits
+
+    Count the number of decimal (base 10) digits for an input unsigned integer.
+
+    @param[in] x  unsigned integer number to count its digits
+    @return    number of decimal digits
+    */
+    inline unsigned int count_digits(number_unsigned_t x) noexcept
+    {
+        unsigned int n_digits = 1;
+        for (;;)
+        {
+            if (x < 10)
+            {
+                return n_digits;
+            }
+            if (x < 100)
+            {
+                return n_digits + 1;
+            }
+            if (x < 1000)
+            {
+                return n_digits + 2;
+            }
+            if (x < 10000)
+            {
+                return n_digits + 3;
+            }
+            x = x / 10000u;
+            n_digits += 4;
+        }
+    }
+
+    // templates to avoid warnings about useless casts
+    template <typename NumberType, enable_if_t<std::is_signed<NumberType>::value, int> = 0>
+    bool is_negative_number(NumberType x)
+    {
+        return x < 0;
+    }
+
+    template < typename NumberType, enable_if_t <std::is_unsigned<NumberType>::value, int > = 0 >
+    bool is_negative_number(NumberType /*unused*/)
+    {
+        return false;
+    }
+
     /*!
     @brief dump an integer
 
@@ -10136,12 +16729,30 @@ class serializer
     @param[in] x  integer number (signed or unsigned) to dump
     @tparam NumberType either @a number_integer_t or @a number_unsigned_t
     */
-    template<typename NumberType, detail::enable_if_t<
-                 std::is_same<NumberType, number_unsigned_t>::value or
-                 std::is_same<NumberType, number_integer_t>::value,
-                 int> = 0>
+    template < typename NumberType, detail::enable_if_t <
+                   std::is_integral<NumberType>::value ||
+                   std::is_same<NumberType, number_unsigned_t>::value ||
+                   std::is_same<NumberType, number_integer_t>::value ||
+                   std::is_same<NumberType, binary_char_t>::value,
+                   int > = 0 >
     void dump_integer(NumberType x)
     {
+        static constexpr std::array<std::array<char, 2>, 100> digits_to_99
+        {
+            {
+                {{'0', '0'}}, {{'0', '1'}}, {{'0', '2'}}, {{'0', '3'}}, {{'0', '4'}}, {{'0', '5'}}, {{'0', '6'}}, {{'0', '7'}}, {{'0', '8'}}, {{'0', '9'}},
+                {{'1', '0'}}, {{'1', '1'}}, {{'1', '2'}}, {{'1', '3'}}, {{'1', '4'}}, {{'1', '5'}}, {{'1', '6'}}, {{'1', '7'}}, {{'1', '8'}}, {{'1', '9'}},
+                {{'2', '0'}}, {{'2', '1'}}, {{'2', '2'}}, {{'2', '3'}}, {{'2', '4'}}, {{'2', '5'}}, {{'2', '6'}}, {{'2', '7'}}, {{'2', '8'}}, {{'2', '9'}},
+                {{'3', '0'}}, {{'3', '1'}}, {{'3', '2'}}, {{'3', '3'}}, {{'3', '4'}}, {{'3', '5'}}, {{'3', '6'}}, {{'3', '7'}}, {{'3', '8'}}, {{'3', '9'}},
+                {{'4', '0'}}, {{'4', '1'}}, {{'4', '2'}}, {{'4', '3'}}, {{'4', '4'}}, {{'4', '5'}}, {{'4', '6'}}, {{'4', '7'}}, {{'4', '8'}}, {{'4', '9'}},
+                {{'5', '0'}}, {{'5', '1'}}, {{'5', '2'}}, {{'5', '3'}}, {{'5', '4'}}, {{'5', '5'}}, {{'5', '6'}}, {{'5', '7'}}, {{'5', '8'}}, {{'5', '9'}},
+                {{'6', '0'}}, {{'6', '1'}}, {{'6', '2'}}, {{'6', '3'}}, {{'6', '4'}}, {{'6', '5'}}, {{'6', '6'}}, {{'6', '7'}}, {{'6', '8'}}, {{'6', '9'}},
+                {{'7', '0'}}, {{'7', '1'}}, {{'7', '2'}}, {{'7', '3'}}, {{'7', '4'}}, {{'7', '5'}}, {{'7', '6'}}, {{'7', '7'}}, {{'7', '8'}}, {{'7', '9'}},
+                {{'8', '0'}}, {{'8', '1'}}, {{'8', '2'}}, {{'8', '3'}}, {{'8', '4'}}, {{'8', '5'}}, {{'8', '6'}}, {{'8', '7'}}, {{'8', '8'}}, {{'8', '9'}},
+                {{'9', '0'}}, {{'9', '1'}}, {{'9', '2'}}, {{'9', '3'}}, {{'9', '4'}}, {{'9', '5'}}, {{'9', '6'}}, {{'9', '7'}}, {{'9', '8'}}, {{'9', '9'}},
+            }
+        };
+
         // special case for "0"
         if (x == 0)
         {
@@ -10149,28 +16760,56 @@ class serializer
             return;
         }
 
-        const bool is_negative = not (x >= 0);  // see issue #755
-        std::size_t i = 0;
+        // use a pointer to fill the buffer
+        auto buffer_ptr = number_buffer.begin(); // NOLINT(llvm-qualified-auto,readability-qualified-auto,cppcoreguidelines-pro-type-vararg,hicpp-vararg)
 
-        while (x != 0)
+        number_unsigned_t abs_value;
+
+        unsigned int n_chars{};
+
+        if (is_negative_number(x))
         {
-            // spare 1 byte for '\0'
-            assert(i < number_buffer.size() - 1);
+            *buffer_ptr = '-';
+            abs_value = remove_sign(static_cast<number_integer_t>(x));
 
-            const auto digit = std::labs(static_cast<long>(x % 10));
-            number_buffer[i++] = static_cast<char>('0' + digit);
-            x /= 10;
+            // account one more byte for the minus sign
+            n_chars = 1 + count_digits(abs_value);
+        }
+        else
+        {
+            abs_value = static_cast<number_unsigned_t>(x);
+            n_chars = count_digits(abs_value);
         }
 
-        if (is_negative)
+        // spare 1 byte for '\0'
+        JSON_ASSERT(n_chars < number_buffer.size() - 1);
+
+        // jump to the end to generate the string from backward,
+        // so we later avoid reversing the result
+        buffer_ptr += n_chars;
+
+        // Fast int2ascii implementation inspired by "Fastware" talk by Andrei Alexandrescu
+        // See: https://www.youtube.com/watch?v=o4-CwDo2zpg
+        while (abs_value >= 100)
         {
-            // make sure there is capacity for the '-'
-            assert(i < number_buffer.size() - 2);
-            number_buffer[i++] = '-';
+            const auto digits_index = static_cast<unsigned>((abs_value % 100));
+            abs_value /= 100;
+            *(--buffer_ptr) = digits_to_99[digits_index][1];
+            *(--buffer_ptr) = digits_to_99[digits_index][0];
         }
 
-        std::reverse(number_buffer.begin(), number_buffer.begin() + i);
-        o->write_characters(number_buffer.data(), i);
+        if (abs_value >= 10)
+        {
+            const auto digits_index = static_cast<unsigned>(abs_value);
+            *(--buffer_ptr) = digits_to_99[digits_index][1];
+            *(--buffer_ptr) = digits_to_99[digits_index][0];
+        }
+        else
+        {
+            *(--buffer_ptr) = static_cast<char>('0' + abs_value);
+        }
+
+        o->write_characters(number_buffer.data(), n_chars);
     }
 
     /*!
@@ -10184,7 +16823,7 @@ class serializer
     void dump_float(number_float_t x)
     {
         // NaN / inf
-        if (not std::isfinite(x))
+        if (!std::isfinite(x))
         {
             o->write_characters("null", 4);
             return;
@@ -10196,16 +16835,16 @@ class serializer
         //
         // NB: The test below works if <long double> == <double>.
         static constexpr bool is_ieee_single_or_double
-            = (std::numeric_limits<number_float_t>::is_iec559 and std::numeric_limits<number_float_t>::digits == 24 and std::numeric_limits<number_float_t>::max_exponent == 128) or
-              (std::numeric_limits<number_float_t>::is_iec559 and std::numeric_limits<number_float_t>::digits == 53 and std::numeric_limits<number_float_t>::max_exponent == 1024);
+            = (std::numeric_limits<number_float_t>::is_iec559 && std::numeric_limits<number_float_t>::digits == 24 && std::numeric_limits<number_float_t>::max_exponent == 128) ||
+              (std::numeric_limits<number_float_t>::is_iec559 && std::numeric_limits<number_float_t>::digits == 53 && std::numeric_limits<number_float_t>::max_exponent == 1024);
 
         dump_float(x, std::integral_constant<bool, is_ieee_single_or_double>());
     }
 
     void dump_float(number_float_t x, std::true_type /*is_ieee_single_or_double*/)
     {
-        char* begin = number_buffer.data();
-        char* end = ::nlohmann::detail::to_chars(begin, begin + number_buffer.size(), x);
+        auto* begin = number_buffer.data();
+        auto* end = ::nlohmann::detail::to_chars(begin, begin + number_buffer.size(), x);
 
         o->write_characters(begin, static_cast<size_t>(end - begin));
     }
@@ -10216,26 +16855,28 @@ class serializer
         static constexpr auto d = std::numeric_limits<number_float_t>::max_digits10;
 
         // the actual conversion
-        std::ptrdiff_t len = snprintf(number_buffer.data(), number_buffer.size(), "%.*g", d, x);
+        // NOLINTNEXTLINE(cppcoreguidelines-pro-type-vararg,hicpp-vararg)
+        std::ptrdiff_t len = (std::snprintf)(number_buffer.data(), number_buffer.size(), "%.*g", d, x);
 
         // negative value indicates an error
-        assert(len > 0);
+        JSON_ASSERT(len > 0);
         // check if buffer was large enough
-        assert(static_cast<std::size_t>(len) < number_buffer.size());
+        JSON_ASSERT(static_cast<std::size_t>(len) < number_buffer.size());
 
         // erase thousands separator
         if (thousands_sep != '\0')
         {
-            const auto end = std::remove(number_buffer.begin(),
-                                         number_buffer.begin() + len, thousands_sep);
+            // NOLINTNEXTLINE(readability-qualified-auto,llvm-qualified-auto): std::remove returns an iterator, see https://github.com/nlohmann/json/issues/3081
+            const auto end = std::remove(number_buffer.begin(), number_buffer.begin() + len, thousands_sep);
             std::fill(end, number_buffer.end(), '\0');
-            assert((end - number_buffer.begin()) <= len);
+            JSON_ASSERT((end - number_buffer.begin()) <= len);
             len = (end - number_buffer.begin());
         }
 
         // convert decimal point to '.'
-        if (decimal_point != '\0' and decimal_point != '.')
+        if (decimal_point != '\0' && decimal_point != '.')
         {
+            // NOLINTNEXTLINE(readability-qualified-auto,llvm-qualified-auto): std::find returns an iterator, see https://github.com/nlohmann/json/issues/3081
             const auto dec_pos = std::find(number_buffer.begin(), number_buffer.end(), decimal_point);
             if (dec_pos != number_buffer.end())
             {
@@ -10245,12 +16886,12 @@ class serializer
 
         o->write_characters(number_buffer.data(), static_cast<std::size_t>(len));
 
-        // determine if need to append ".0"
+        // determine if we need to append ".0"
         const bool value_is_int_like =
             std::none_of(number_buffer.begin(), number_buffer.begin() + len + 1,
                          [](char c)
         {
-            return (c == '.' or c == 'e');
+            return c == '.' || c == 'e';
         });
 
         if (value_is_int_like)
@@ -10280,9 +16921,9 @@ class serializer
     @copyright Copyright (c) 2008-2009 Bjoern Hoehrmann <bjoern@hoehrmann.de>
     @sa http://bjoern.hoehrmann.de/utf-8/decoder/dfa/
     */
-    static uint8_t decode(uint8_t& state, uint32_t& codep, const uint8_t byte) noexcept
+    static std::uint8_t decode(std::uint8_t& state, std::uint32_t& codep, const std::uint8_t byte) noexcept
     {
-        static const std::array<uint8_t, 400> utf8d =
+        static const std::array<std::uint8_t, 400> utf8d =
         {
             {
                 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, // 00..1F
@@ -10302,16 +16943,45 @@ class serializer
             }
         };
 
-        const uint8_t type = utf8d[byte];
+        JSON_ASSERT(byte < utf8d.size());
+        const std::uint8_t type = utf8d[byte];
 
         codep = (state != UTF8_ACCEPT)
-                ? (byte & 0x3fu) | (codep << 6)
-                : static_cast<uint32_t>(0xff >> type) & (byte);
+                ? (byte & 0x3fu) | (codep << 6u)
+                : (0xFFu >> type) & (byte);
 
-        state = utf8d[256u + state * 16u + type];
+        std::size_t index = 256u + static_cast<size_t>(state) * 16u + static_cast<size_t>(type);
+        JSON_ASSERT(index < 400);
+        state = utf8d[index];
         return state;
     }
 
+    /*
+     * Overload to make the compiler happy while it is instantiating
+     * dump_integer for number_unsigned_t.
+     * Must never be called.
+     */
+    number_unsigned_t remove_sign(number_unsigned_t x)
+    {
+        JSON_ASSERT(false); // NOLINT(cert-dcl03-c,hicpp-static-assert,misc-static-assert) LCOV_EXCL_LINE
+        return x; // LCOV_EXCL_LINE
+    }
+
+    /*
+     * Helper function for dump_integer
+     *
+     * This function takes a negative signed integer and returns its absolute
+     * value as unsigned integer. The plus/minus shuffling is necessary as we can
+     * not directly remove the sign of an arbitrary signed integer as the
+     * absolute values of INT_MIN and INT_MAX are usually not the same. See
+     * #1708 for details.
+     */
+    inline number_unsigned_t remove_sign(number_integer_t x) noexcept
+    {
+        JSON_ASSERT(x < 0 && x < (std::numeric_limits<number_integer_t>::max)()); // NOLINT(misc-redundant-expression)
+        return static_cast<number_unsigned_t>(-(x + 1)) + 1;
+    }
+
   private:
     /// the output of the serializer
     output_adapter_t<char> o = nullptr;
@@ -10333,829 +17003,253 @@ class serializer
     const char indent_char;
     /// the indentation string
     string_t indent_string;
+
+    /// error_handler how to react on decoding errors
+    const error_handler_t error_handler;
 };
-}
-}
+}  // namespace detail
+}  // namespace nlohmann
 
-// #include <nlohmann/detail/json_ref.hpp>
+// #include <nlohmann/detail/value_t.hpp>
+
+// #include <nlohmann/json_fwd.hpp>
+
+// #include <nlohmann/ordered_map.hpp>
 
 
-#include <initializer_list>
-#include <utility>
-
-namespace nlohmann
-{
-namespace detail
-{
-template<typename BasicJsonType>
-class json_ref
-{
-  public:
-    using value_type = BasicJsonType;
-
-    json_ref(value_type&& value)
-        : owned_value(std::move(value)), value_ref(&owned_value), is_rvalue(true)
-    {}
-
-    json_ref(const value_type& value)
-        : value_ref(const_cast<value_type*>(&value)), is_rvalue(false)
-    {}
-
-    json_ref(std::initializer_list<json_ref> init)
-        : owned_value(init), value_ref(&owned_value), is_rvalue(true)
-    {}
-
-    template<class... Args>
-    json_ref(Args&& ... args)
-        : owned_value(std::forward<Args>(args)...), value_ref(&owned_value), is_rvalue(true)
-    {}
-
-    // class should be movable only
-    json_ref(json_ref&&) = default;
-    json_ref(const json_ref&) = delete;
-    json_ref& operator=(const json_ref&) = delete;
-
-    value_type moved_or_copied() const
-    {
-        if (is_rvalue)
-        {
-            return std::move(*value_ref);
-        }
-        return *value_ref;
-    }
-
-    value_type const& operator*() const
-    {
-        return *static_cast<value_type const*>(value_ref);
-    }
-
-    value_type const* operator->() const
-    {
-        return static_cast<value_type const*>(value_ref);
-    }
-
-  private:
-    mutable value_type owned_value = nullptr;
-    value_type* value_ref = nullptr;
-    const bool is_rvalue;
-};
-}
-}
-
-// #include <nlohmann/detail/json_pointer.hpp>
-
-
-#include <cassert> // assert
-#include <numeric> // accumulate
-#include <string> // string
+#include <functional> // less
+#include <initializer_list> // initializer_list
+#include <iterator> // input_iterator_tag, iterator_traits
+#include <memory> // allocator
+#include <stdexcept> // for out_of_range
+#include <type_traits> // enable_if, is_convertible
+#include <utility> // pair
 #include <vector> // vector
 
 // #include <nlohmann/detail/macro_scope.hpp>
 
-// #include <nlohmann/detail/exceptions.hpp>
-
-// #include <nlohmann/detail/value_t.hpp>
-
 
 namespace nlohmann
 {
-template<typename BasicJsonType>
-class json_pointer
+
+/// ordered_map: a minimal map-like container that preserves insertion order
+/// for use within nlohmann::basic_json<ordered_map>
+template <class Key, class T, class IgnoredLess = std::less<Key>,
+          class Allocator = std::allocator<std::pair<const Key, T>>>
+                  struct ordered_map : std::vector<std::pair<const Key, T>, Allocator>
 {
-    // allow basic_json to access private members
-    NLOHMANN_BASIC_JSON_TPL_DECLARATION
-    friend class basic_json;
+    using key_type = Key;
+    using mapped_type = T;
+    using Container = std::vector<std::pair<const Key, T>, Allocator>;
+    using iterator = typename Container::iterator;
+    using const_iterator = typename Container::const_iterator;
+    using size_type = typename Container::size_type;
+    using value_type = typename Container::value_type;
 
-  public:
-    /*!
-    @brief create JSON pointer
+    // Explicit constructors instead of `using Container::Container`
+    // otherwise older compilers choke on it (GCC <= 5.5, xcode <= 9.4)
+    ordered_map(const Allocator& alloc = Allocator()) : Container{alloc} {}
+    template <class It>
+    ordered_map(It first, It last, const Allocator& alloc = Allocator())
+        : Container{first, last, alloc} {}
+    ordered_map(std::initializer_list<T> init, const Allocator& alloc = Allocator() )
+        : Container{init, alloc} {}
 
-    Create a JSON pointer according to the syntax described in
-    [Section 3 of RFC6901](https://tools.ietf.org/html/rfc6901#section-3).
-
-    @param[in] s  string representing the JSON pointer; if omitted, the empty
-                  string is assumed which references the whole JSON value
-
-    @throw parse_error.107 if the given JSON pointer @a s is nonempty and does
-                           not begin with a slash (`/`); see example below
-
-    @throw parse_error.108 if a tilde (`~`) in the given JSON pointer @a s is
-    not followed by `0` (representing `~`) or `1` (representing `/`); see
-    example below
-
-    @liveexample{The example shows the construction several valid JSON pointers
-    as well as the exceptional behavior.,json_pointer}
-
-    @since version 2.0.0
-    */
-    explicit json_pointer(const std::string& s = "")
-        : reference_tokens(split(s))
-    {}
-
-    /*!
-    @brief return a string representation of the JSON pointer
-
-    @invariant For each JSON pointer `ptr`, it holds:
-    @code {.cpp}
-    ptr == json_pointer(ptr.to_string());
-    @endcode
-
-    @return a string representation of the JSON pointer
-
-    @liveexample{The example shows the result of `to_string`.,
-    json_pointer__to_string}
-
-    @since version 2.0.0
-    */
-    std::string to_string() const noexcept
+    std::pair<iterator, bool> emplace(const key_type& key, T&& t)
     {
-        return std::accumulate(reference_tokens.begin(), reference_tokens.end(),
-                               std::string{},
-                               [](const std::string & a, const std::string & b)
+        for (auto it = this->begin(); it != this->end(); ++it)
         {
-            return a + "/" + escape(b);
-        });
-    }
-
-    /// @copydoc to_string()
-    operator std::string() const
-    {
-        return to_string();
-    }
-
-    /*!
-    @param[in] s  reference token to be converted into an array index
-
-    @return integer representation of @a s
-
-    @throw out_of_range.404 if string @a s could not be converted to an integer
-    */
-    static int array_index(const std::string& s)
-    {
-        std::size_t processed_chars = 0;
-        const int res = std::stoi(s, &processed_chars);
-
-        // check if the string was completely read
-        if (JSON_UNLIKELY(processed_chars != s.size()))
-        {
-            JSON_THROW(detail::out_of_range::create(404, "unresolved reference token '" + s + "'"));
-        }
-
-        return res;
-    }
-
-  private:
-    /*!
-    @brief remove and return last reference pointer
-    @throw out_of_range.405 if JSON pointer has no parent
-    */
-    std::string pop_back()
-    {
-        if (JSON_UNLIKELY(is_root()))
-        {
-            JSON_THROW(detail::out_of_range::create(405, "JSON pointer has no parent"));
-        }
-
-        auto last = reference_tokens.back();
-        reference_tokens.pop_back();
-        return last;
-    }
-
-    /// return whether pointer points to the root document
-    bool is_root() const
-    {
-        return reference_tokens.empty();
-    }
-
-    json_pointer top() const
-    {
-        if (JSON_UNLIKELY(is_root()))
-        {
-            JSON_THROW(detail::out_of_range::create(405, "JSON pointer has no parent"));
-        }
-
-        json_pointer result = *this;
-        result.reference_tokens = {reference_tokens[0]};
-        return result;
-    }
-
-    /*!
-    @brief create and return a reference to the pointed to value
-
-    @complexity Linear in the number of reference tokens.
-
-    @throw parse_error.109 if array index is not a number
-    @throw type_error.313 if value cannot be unflattened
-    */
-    BasicJsonType& get_and_create(BasicJsonType& j) const
-    {
-        using size_type = typename BasicJsonType::size_type;
-        auto result = &j;
-
-        // in case no reference tokens exist, return a reference to the JSON value
-        // j which will be overwritten by a primitive value
-        for (const auto& reference_token : reference_tokens)
-        {
-            switch (result->m_type)
+            if (it->first == key)
             {
-                case detail::value_t::null:
-                {
-                    if (reference_token == "0")
-                    {
-                        // start a new array if reference token is 0
-                        result = &result->operator[](0);
-                    }
-                    else
-                    {
-                        // start a new object otherwise
-                        result = &result->operator[](reference_token);
-                    }
-                    break;
-                }
+                return {it, false};
+            }
+        }
+        Container::emplace_back(key, t);
+        return {--this->end(), true};
+    }
 
-                case detail::value_t::object:
-                {
-                    // create an entry in the object
-                    result = &result->operator[](reference_token);
-                    break;
-                }
+    T& operator[](const Key& key)
+    {
+        return emplace(key, T{}).first->second;
+    }
 
-                case detail::value_t::array:
-                {
-                    // create an entry in the array
-                    JSON_TRY
-                    {
-                        result = &result->operator[](static_cast<size_type>(array_index(reference_token)));
-                    }
-                    JSON_CATCH(std::invalid_argument&)
-                    {
-                        JSON_THROW(detail::parse_error::create(109, 0, "array index '" + reference_token + "' is not a number"));
-                    }
-                    break;
-                }
+    const T& operator[](const Key& key) const
+    {
+        return at(key);
+    }
 
-                /*
-                The following code is only reached if there exists a reference
-                token _and_ the current value is primitive. In this case, we have
-                an error situation, because primitive values may only occur as
-                single value; that is, with an empty list of reference tokens.
-                */
-                default:
-                    JSON_THROW(detail::type_error::create(313, "invalid value to unflatten"));
+    T& at(const Key& key)
+    {
+        for (auto it = this->begin(); it != this->end(); ++it)
+        {
+            if (it->first == key)
+            {
+                return it->second;
             }
         }
 
-        return *result;
+        JSON_THROW(std::out_of_range("key not found"));
     }
 
-    /*!
-    @brief return a reference to the pointed to value
-
-    @note This version does not throw if a value is not present, but tries to
-          create nested values instead. For instance, calling this function
-          with pointer `"/this/that"` on a null value is equivalent to calling
-          `operator[]("this").operator[]("that")` on that value, effectively
-          changing the null value to an object.
-
-    @param[in] ptr  a JSON value
-
-    @return reference to the JSON value pointed to by the JSON pointer
-
-    @complexity Linear in the length of the JSON pointer.
-
-    @throw parse_error.106   if an array index begins with '0'
-    @throw parse_error.109   if an array index was not a number
-    @throw out_of_range.404  if the JSON pointer can not be resolved
-    */
-    BasicJsonType& get_unchecked(BasicJsonType* ptr) const
+    const T& at(const Key& key) const
     {
-        using size_type = typename BasicJsonType::size_type;
-        for (const auto& reference_token : reference_tokens)
+        for (auto it = this->begin(); it != this->end(); ++it)
         {
-            // convert null values to arrays or objects before continuing
-            if (ptr->m_type == detail::value_t::null)
+            if (it->first == key)
             {
-                // check if reference token is a number
-                const bool nums =
-                    std::all_of(reference_token.begin(), reference_token.end(),
-                                [](const char x)
-                {
-                    return (x >= '0' and x <= '9');
-                });
-
-                // change value to array for numbers or "-" or to object otherwise
-                *ptr = (nums or reference_token == "-")
-                       ? detail::value_t::array
-                       : detail::value_t::object;
-            }
-
-            switch (ptr->m_type)
-            {
-                case detail::value_t::object:
-                {
-                    // use unchecked object access
-                    ptr = &ptr->operator[](reference_token);
-                    break;
-                }
-
-                case detail::value_t::array:
-                {
-                    // error condition (cf. RFC 6901, Sect. 4)
-                    if (JSON_UNLIKELY(reference_token.size() > 1 and reference_token[0] == '0'))
-                    {
-                        JSON_THROW(detail::parse_error::create(106, 0,
-                                                               "array index '" + reference_token +
-                                                               "' must not begin with '0'"));
-                    }
-
-                    if (reference_token == "-")
-                    {
-                        // explicitly treat "-" as index beyond the end
-                        ptr = &ptr->operator[](ptr->m_value.array->size());
-                    }
-                    else
-                    {
-                        // convert array index to number; unchecked access
-                        JSON_TRY
-                        {
-                            ptr = &ptr->operator[](
-                                static_cast<size_type>(array_index(reference_token)));
-                        }
-                        JSON_CATCH(std::invalid_argument&)
-                        {
-                            JSON_THROW(detail::parse_error::create(109, 0, "array index '" + reference_token + "' is not a number"));
-                        }
-                    }
-                    break;
-                }
-
-                default:
-                    JSON_THROW(detail::out_of_range::create(404, "unresolved reference token '" + reference_token + "'"));
+                return it->second;
             }
         }
 
-        return *ptr;
+        JSON_THROW(std::out_of_range("key not found"));
     }
 
-    /*!
-    @throw parse_error.106   if an array index begins with '0'
-    @throw parse_error.109   if an array index was not a number
-    @throw out_of_range.402  if the array index '-' is used
-    @throw out_of_range.404  if the JSON pointer can not be resolved
-    */
-    BasicJsonType& get_checked(BasicJsonType* ptr) const
+    size_type erase(const Key& key)
     {
-        using size_type = typename BasicJsonType::size_type;
-        for (const auto& reference_token : reference_tokens)
+        for (auto it = this->begin(); it != this->end(); ++it)
         {
-            switch (ptr->m_type)
+            if (it->first == key)
             {
-                case detail::value_t::object:
+                // Since we cannot move const Keys, re-construct them in place
+                for (auto next = it; ++next != this->end(); ++it)
                 {
-                    // note: at performs range check
-                    ptr = &ptr->at(reference_token);
-                    break;
+                    it->~value_type(); // Destroy but keep allocation
+                    new (&*it) value_type{std::move(*next)};
                 }
-
-                case detail::value_t::array:
-                {
-                    if (JSON_UNLIKELY(reference_token == "-"))
-                    {
-                        // "-" always fails the range check
-                        JSON_THROW(detail::out_of_range::create(402,
-                                                                "array index '-' (" + std::to_string(ptr->m_value.array->size()) +
-                                                                ") is out of range"));
-                    }
-
-                    // error condition (cf. RFC 6901, Sect. 4)
-                    if (JSON_UNLIKELY(reference_token.size() > 1 and reference_token[0] == '0'))
-                    {
-                        JSON_THROW(detail::parse_error::create(106, 0,
-                                                               "array index '" + reference_token +
-                                                               "' must not begin with '0'"));
-                    }
-
-                    // note: at performs range check
-                    JSON_TRY
-                    {
-                        ptr = &ptr->at(static_cast<size_type>(array_index(reference_token)));
-                    }
-                    JSON_CATCH(std::invalid_argument&)
-                    {
-                        JSON_THROW(detail::parse_error::create(109, 0, "array index '" + reference_token + "' is not a number"));
-                    }
-                    break;
-                }
-
-                default:
-                    JSON_THROW(detail::out_of_range::create(404, "unresolved reference token '" + reference_token + "'"));
+                Container::pop_back();
+                return 1;
             }
         }
-
-        return *ptr;
+        return 0;
     }
 
-    /*!
-    @brief return a const reference to the pointed to value
-
-    @param[in] ptr  a JSON value
-
-    @return const reference to the JSON value pointed to by the JSON
-    pointer
-
-    @throw parse_error.106   if an array index begins with '0'
-    @throw parse_error.109   if an array index was not a number
-    @throw out_of_range.402  if the array index '-' is used
-    @throw out_of_range.404  if the JSON pointer can not be resolved
-    */
-    const BasicJsonType& get_unchecked(const BasicJsonType* ptr) const
+    iterator erase(iterator pos)
     {
-        using size_type = typename BasicJsonType::size_type;
-        for (const auto& reference_token : reference_tokens)
+        return erase(pos, std::next(pos));
+    }
+
+    iterator erase(iterator first, iterator last)
+    {
+        const auto elements_affected = std::distance(first, last);
+        const auto offset = std::distance(Container::begin(), first);
+
+        // This is the start situation. We need to delete elements_affected
+        // elements (3 in this example: e, f, g), and need to return an
+        // iterator past the last deleted element (h in this example).
+        // Note that offset is the distance from the start of the vector
+        // to first. We will need this later.
+
+        // [ a, b, c, d, e, f, g, h, i, j ]
+        //               ^        ^
+        //             first    last
+
+        // Since we cannot move const Keys, we re-construct them in place.
+        // We start at first and re-construct (viz. copy) the elements from
+        // the back of the vector. Example for first iteration:
+
+        //               ,--------.
+        //               v        |   destroy e and re-construct with h
+        // [ a, b, c, d, e, f, g, h, i, j ]
+        //               ^        ^
+        //               it       it + elements_affected
+
+        for (auto it = first; std::next(it, elements_affected) != Container::end(); ++it)
         {
-            switch (ptr->m_type)
+            it->~value_type(); // destroy but keep allocation
+            new (&*it) value_type{std::move(*std::next(it, elements_affected))}; // "move" next element to it
+        }
+
+        // [ a, b, c, d, h, i, j, h, i, j ]
+        //               ^        ^
+        //             first    last
+
+        // remove the unneeded elements at the end of the vector
+        Container::resize(this->size() - static_cast<size_type>(elements_affected));
+
+        // [ a, b, c, d, h, i, j ]
+        //               ^        ^
+        //             first    last
+
+        // first is now pointing past the last deleted element, but we cannot
+        // use this iterator, because it may have been invalidated by the
+        // resize call. Instead, we can return begin() + offset.
+        return Container::begin() + offset;
+    }
+
+    size_type count(const Key& key) const
+    {
+        for (auto it = this->begin(); it != this->end(); ++it)
+        {
+            if (it->first == key)
             {
-                case detail::value_t::object:
-                {
-                    // use unchecked object access
-                    ptr = &ptr->operator[](reference_token);
-                    break;
-                }
-
-                case detail::value_t::array:
-                {
-                    if (JSON_UNLIKELY(reference_token == "-"))
-                    {
-                        // "-" cannot be used for const access
-                        JSON_THROW(detail::out_of_range::create(402,
-                                                                "array index '-' (" + std::to_string(ptr->m_value.array->size()) +
-                                                                ") is out of range"));
-                    }
-
-                    // error condition (cf. RFC 6901, Sect. 4)
-                    if (JSON_UNLIKELY(reference_token.size() > 1 and reference_token[0] == '0'))
-                    {
-                        JSON_THROW(detail::parse_error::create(106, 0,
-                                                               "array index '" + reference_token +
-                                                               "' must not begin with '0'"));
-                    }
-
-                    // use unchecked array access
-                    JSON_TRY
-                    {
-                        ptr = &ptr->operator[](
-                            static_cast<size_type>(array_index(reference_token)));
-                    }
-                    JSON_CATCH(std::invalid_argument&)
-                    {
-                        JSON_THROW(detail::parse_error::create(109, 0, "array index '" + reference_token + "' is not a number"));
-                    }
-                    break;
-                }
-
-                default:
-                    JSON_THROW(detail::out_of_range::create(404, "unresolved reference token '" + reference_token + "'"));
+                return 1;
             }
         }
-
-        return *ptr;
+        return 0;
     }
 
-    /*!
-    @throw parse_error.106   if an array index begins with '0'
-    @throw parse_error.109   if an array index was not a number
-    @throw out_of_range.402  if the array index '-' is used
-    @throw out_of_range.404  if the JSON pointer can not be resolved
-    */
-    const BasicJsonType& get_checked(const BasicJsonType* ptr) const
+    iterator find(const Key& key)
     {
-        using size_type = typename BasicJsonType::size_type;
-        for (const auto& reference_token : reference_tokens)
+        for (auto it = this->begin(); it != this->end(); ++it)
         {
-            switch (ptr->m_type)
+            if (it->first == key)
             {
-                case detail::value_t::object:
-                {
-                    // note: at performs range check
-                    ptr = &ptr->at(reference_token);
-                    break;
-                }
-
-                case detail::value_t::array:
-                {
-                    if (JSON_UNLIKELY(reference_token == "-"))
-                    {
-                        // "-" always fails the range check
-                        JSON_THROW(detail::out_of_range::create(402,
-                                                                "array index '-' (" + std::to_string(ptr->m_value.array->size()) +
-                                                                ") is out of range"));
-                    }
-
-                    // error condition (cf. RFC 6901, Sect. 4)
-                    if (JSON_UNLIKELY(reference_token.size() > 1 and reference_token[0] == '0'))
-                    {
-                        JSON_THROW(detail::parse_error::create(106, 0,
-                                                               "array index '" + reference_token +
-                                                               "' must not begin with '0'"));
-                    }
-
-                    // note: at performs range check
-                    JSON_TRY
-                    {
-                        ptr = &ptr->at(static_cast<size_type>(array_index(reference_token)));
-                    }
-                    JSON_CATCH(std::invalid_argument&)
-                    {
-                        JSON_THROW(detail::parse_error::create(109, 0, "array index '" + reference_token + "' is not a number"));
-                    }
-                    break;
-                }
-
-                default:
-                    JSON_THROW(detail::out_of_range::create(404, "unresolved reference token '" + reference_token + "'"));
+                return it;
             }
         }
-
-        return *ptr;
+        return Container::end();
     }
 
-    /*!
-    @brief split the string input to reference tokens
-
-    @note This function is only called by the json_pointer constructor.
-          All exceptions below are documented there.
-
-    @throw parse_error.107  if the pointer is not empty or begins with '/'
-    @throw parse_error.108  if character '~' is not followed by '0' or '1'
-    */
-    static std::vector<std::string> split(const std::string& reference_string)
+    const_iterator find(const Key& key) const
     {
-        std::vector<std::string> result;
-
-        // special case: empty reference string -> no reference tokens
-        if (reference_string.empty())
+        for (auto it = this->begin(); it != this->end(); ++it)
         {
-            return result;
-        }
-
-        // check if nonempty reference string begins with slash
-        if (JSON_UNLIKELY(reference_string[0] != '/'))
-        {
-            JSON_THROW(detail::parse_error::create(107, 1,
-                                                   "JSON pointer must be empty or begin with '/' - was: '" +
-                                                   reference_string + "'"));
-        }
-
-        // extract the reference tokens:
-        // - slash: position of the last read slash (or end of string)
-        // - start: position after the previous slash
-        for (
-            // search for the first slash after the first character
-            std::size_t slash = reference_string.find_first_of('/', 1),
-            // set the beginning of the first reference token
-            start = 1;
-            // we can stop if start == string::npos+1 = 0
-            start != 0;
-            // set the beginning of the next reference token
-            // (will eventually be 0 if slash == std::string::npos)
-            start = slash + 1,
-            // find next slash
-            slash = reference_string.find_first_of('/', start))
-        {
-            // use the text between the beginning of the reference token
-            // (start) and the last slash (slash).
-            auto reference_token = reference_string.substr(start, slash - start);
-
-            // check reference tokens are properly escaped
-            for (std::size_t pos = reference_token.find_first_of('~');
-                    pos != std::string::npos;
-                    pos = reference_token.find_first_of('~', pos + 1))
+            if (it->first == key)
             {
-                assert(reference_token[pos] == '~');
-
-                // ~ must be followed by 0 or 1
-                if (JSON_UNLIKELY(pos == reference_token.size() - 1 or
-                                  (reference_token[pos + 1] != '0' and
-                                   reference_token[pos + 1] != '1')))
-                {
-                    JSON_THROW(detail::parse_error::create(108, 0, "escape character '~' must be followed with '0' or '1'"));
-                }
-            }
-
-            // finally, store the reference token
-            unescape(reference_token);
-            result.push_back(reference_token);
-        }
-
-        return result;
-    }
-
-    /*!
-    @brief replace all occurrences of a substring by another string
-
-    @param[in,out] s  the string to manipulate; changed so that all
-                   occurrences of @a f are replaced with @a t
-    @param[in]     f  the substring to replace with @a t
-    @param[in]     t  the string to replace @a f
-
-    @pre The search string @a f must not be empty. **This precondition is
-    enforced with an assertion.**
-
-    @since version 2.0.0
-    */
-    static void replace_substring(std::string& s, const std::string& f,
-                                  const std::string& t)
-    {
-        assert(not f.empty());
-        for (auto pos = s.find(f);                // find first occurrence of f
-                pos != std::string::npos;         // make sure f was found
-                s.replace(pos, f.size(), t),      // replace with t, and
-                pos = s.find(f, pos + t.size()))  // find next occurrence of f
-        {}
-    }
-
-    /// escape "~"" to "~0" and "/" to "~1"
-    static std::string escape(std::string s)
-    {
-        replace_substring(s, "~", "~0");
-        replace_substring(s, "/", "~1");
-        return s;
-    }
-
-    /// unescape "~1" to tilde and "~0" to slash (order is important!)
-    static void unescape(std::string& s)
-    {
-        replace_substring(s, "~1", "/");
-        replace_substring(s, "~0", "~");
-    }
-
-    /*!
-    @param[in] reference_string  the reference string to the current value
-    @param[in] value             the value to consider
-    @param[in,out] result        the result object to insert values to
-
-    @note Empty objects or arrays are flattened to `null`.
-    */
-    static void flatten(const std::string& reference_string,
-                        const BasicJsonType& value,
-                        BasicJsonType& result)
-    {
-        switch (value.m_type)
-        {
-            case detail::value_t::array:
-            {
-                if (value.m_value.array->empty())
-                {
-                    // flatten empty array as null
-                    result[reference_string] = nullptr;
-                }
-                else
-                {
-                    // iterate array and use index as reference string
-                    for (std::size_t i = 0; i < value.m_value.array->size(); ++i)
-                    {
-                        flatten(reference_string + "/" + std::to_string(i),
-                                value.m_value.array->operator[](i), result);
-                    }
-                }
-                break;
-            }
-
-            case detail::value_t::object:
-            {
-                if (value.m_value.object->empty())
-                {
-                    // flatten empty object as null
-                    result[reference_string] = nullptr;
-                }
-                else
-                {
-                    // iterate object and use keys as reference string
-                    for (const auto& element : *value.m_value.object)
-                    {
-                        flatten(reference_string + "/" + escape(element.first), element.second, result);
-                    }
-                }
-                break;
-            }
-
-            default:
-            {
-                // add primitive value with its reference string
-                result[reference_string] = value;
-                break;
+                return it;
             }
         }
+        return Container::end();
     }
 
-    /*!
-    @param[in] value  flattened JSON
-
-    @return unflattened JSON
-
-    @throw parse_error.109 if array index is not a number
-    @throw type_error.314  if value is not an object
-    @throw type_error.315  if object values are not primitive
-    @throw type_error.313  if value cannot be unflattened
-    */
-    static BasicJsonType
-    unflatten(const BasicJsonType& value)
+    std::pair<iterator, bool> insert( value_type&& value )
     {
-        if (JSON_UNLIKELY(not value.is_object()))
-        {
-            JSON_THROW(detail::type_error::create(314, "only objects can be unflattened"));
-        }
+        return emplace(value.first, std::move(value.second));
+    }
 
-        BasicJsonType result;
-
-        // iterate the JSON object values
-        for (const auto& element : *value.m_value.object)
+    std::pair<iterator, bool> insert( const value_type& value )
+    {
+        for (auto it = this->begin(); it != this->end(); ++it)
         {
-            if (JSON_UNLIKELY(not element.second.is_primitive()))
+            if (it->first == value.first)
             {
-                JSON_THROW(detail::type_error::create(315, "values in object must be primitive"));
+                return {it, false};
             }
-
-            // assign value to reference pointed to by JSON pointer; Note that if
-            // the JSON pointer is "" (i.e., points to the whole value), function
-            // get_and_create returns a reference to result itself. An assignment
-            // will then create a primitive value.
-            json_pointer(element.first).get_and_create(result) = element.second;
         }
-
-        return result;
+        Container::push_back(value);
+        return {--this->end(), true};
     }
 
-    friend bool operator==(json_pointer const& lhs,
-                           json_pointer const& rhs) noexcept
+    template<typename InputIt>
+    using require_input_iter = typename std::enable_if<std::is_convertible<typename std::iterator_traits<InputIt>::iterator_category,
+            std::input_iterator_tag>::value>::type;
+
+    template<typename InputIt, typename = require_input_iter<InputIt>>
+    void insert(InputIt first, InputIt last)
     {
-        return (lhs.reference_tokens == rhs.reference_tokens);
-    }
-
-    friend bool operator!=(json_pointer const& lhs,
-                           json_pointer const& rhs) noexcept
-    {
-        return not (lhs == rhs);
-    }
-
-    /// the reference tokens
-    std::vector<std::string> reference_tokens;
-};
-}
-
-// #include <nlohmann/adl_serializer.hpp>
-
-
-#include <utility>
-
-// #include <nlohmann/detail/conversions/from_json.hpp>
-
-// #include <nlohmann/detail/conversions/to_json.hpp>
-
-
-namespace nlohmann
-{
-template<typename, typename>
-struct adl_serializer
-{
-    /*!
-    @brief convert a JSON value to any value type
-
-    This function is usually called by the `get()` function of the
-    @ref basic_json class (either explicit or via conversion operators).
-
-    @param[in] j         JSON value to read from
-    @param[in,out] val  value to write to
-    */
-    template<typename BasicJsonType, typename ValueType>
-    static auto from_json(BasicJsonType&& j, ValueType& val) noexcept(
-        noexcept(::nlohmann::from_json(std::forward<BasicJsonType>(j), val))) -> decltype(
-            ::nlohmann::from_json(std::forward<BasicJsonType>(j), val), void()
-        )
-    {
-        ::nlohmann::from_json(std::forward<BasicJsonType>(j), val);
-    }
-
-    /*!
-    @brief convert any value type to a JSON value
-
-    This function is usually called by the constructors of the @ref basic_json
-    class.
-
-    @param[in,out] j  JSON value to write to
-    @param[in] val     value to read from
-    */
-    template <typename BasicJsonType, typename ValueType>
-    static auto to_json(BasicJsonType& j, ValueType&& val) noexcept(
-        noexcept(::nlohmann::to_json(j, std::forward<ValueType>(val))))
-    -> decltype(::nlohmann::to_json(j, std::forward<ValueType>(val)),
-                void())
-    {
-        ::nlohmann::to_json(j, std::forward<ValueType>(val));
+        for (auto it = first; it != last; ++it)
+        {
+            insert(*it);
+        }
     }
 };
-}
 
+}  // namespace nlohmann
+
+
+#if defined(JSON_HAS_CPP_17)
+    #include <string_view>
+#endif
 
 /*!
 @brief namespace for Niels Lohmann
@@ -11168,66 +17262,7 @@ namespace nlohmann
 /*!
 @brief a class to store JSON values
 
-@tparam ObjectType type for JSON objects (`std::map` by default; will be used
-in @ref object_t)
-@tparam ArrayType type for JSON arrays (`std::vector` by default; will be used
-in @ref array_t)
-@tparam StringType type for JSON strings and object keys (`std::string` by
-default; will be used in @ref string_t)
-@tparam BooleanType type for JSON booleans (`bool` by default; will be used
-in @ref boolean_t)
-@tparam NumberIntegerType type for JSON integer numbers (`int64_t` by
-default; will be used in @ref number_integer_t)
-@tparam NumberUnsignedType type for JSON unsigned integer numbers (@c
-`uint64_t` by default; will be used in @ref number_unsigned_t)
-@tparam NumberFloatType type for JSON floating-point numbers (`double` by
-default; will be used in @ref number_float_t)
-@tparam AllocatorType type of the allocator to use (`std::allocator` by
-default)
-@tparam JSONSerializer the serializer to resolve internal calls to `to_json()`
-and `from_json()` (@ref adl_serializer by default)
-
-@requirement The class satisfies the following concept requirements:
-- Basic
- - [DefaultConstructible](https://en.cppreference.com/w/cpp/named_req/DefaultConstructible):
-   JSON values can be default constructed. The result will be a JSON null
-   value.
- - [MoveConstructible](https://en.cppreference.com/w/cpp/named_req/MoveConstructible):
-   A JSON value can be constructed from an rvalue argument.
- - [CopyConstructible](https://en.cppreference.com/w/cpp/named_req/CopyConstructible):
-   A JSON value can be copy-constructed from an lvalue expression.
- - [MoveAssignable](https://en.cppreference.com/w/cpp/named_req/MoveAssignable):
-   A JSON value van be assigned from an rvalue argument.
- - [CopyAssignable](https://en.cppreference.com/w/cpp/named_req/CopyAssignable):
-   A JSON value can be copy-assigned from an lvalue expression.
- - [Destructible](https://en.cppreference.com/w/cpp/named_req/Destructible):
-   JSON values can be destructed.
-- Layout
- - [StandardLayoutType](https://en.cppreference.com/w/cpp/named_req/StandardLayoutType):
-   JSON values have
-   [standard layout](https://en.cppreference.com/w/cpp/language/data_members#Standard_layout):
-   All non-static data members are private and standard layout types, the
-   class has no virtual functions or (virtual) base classes.
-- Library-wide
- - [EqualityComparable](https://en.cppreference.com/w/cpp/named_req/EqualityComparable):
-   JSON values can be compared with `==`, see @ref
-   operator==(const_reference,const_reference).
- - [LessThanComparable](https://en.cppreference.com/w/cpp/named_req/LessThanComparable):
-   JSON values can be compared with `<`, see @ref
-   operator<(const_reference,const_reference).
- - [Swappable](https://en.cppreference.com/w/cpp/named_req/Swappable):
-   Any JSON lvalue or rvalue of can be swapped with any lvalue or rvalue of
-   other compatible types, using unqualified function call @ref swap().
- - [NullablePointer](https://en.cppreference.com/w/cpp/named_req/NullablePointer):
-   JSON values can be compared against `std::nullptr_t` objects which are used
-   to model the `null` value.
-- Container
- - [Container](https://en.cppreference.com/w/cpp/named_req/Container):
-   JSON values can be used like STL containers and provide iterator access.
- - [ReversibleContainer](https://en.cppreference.com/w/cpp/named_req/ReversibleContainer);
-   JSON values can be used like STL containers and provide reverse iterator
-   access.
-
+@internal
 @invariant The member variables @a m_value and @a m_type have the following
 relationship:
 - If `m_type == value_t::object`, then `m_value.object != nullptr`.
@@ -11235,43 +17270,55 @@ relationship:
 - If `m_type == value_t::string`, then `m_value.string != nullptr`.
 The invariants are checked by member function assert_invariant().
 
-@internal
-@note ObjectType trick from http://stackoverflow.com/a/9860911
+@note ObjectType trick from https://stackoverflow.com/a/9860911
 @endinternal
 
-@see [RFC 7159: The JavaScript Object Notation (JSON) Data Interchange
-Format](http://rfc7159.net/rfc7159)
-
 @since version 1.0.0
 
 @nosubgrouping
 */
 NLOHMANN_BASIC_JSON_TPL_DECLARATION
-class basic_json
+class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-special-member-functions)
 {
   private:
     template<detail::value_t> friend struct detail::external_constructor;
     friend ::nlohmann::json_pointer<basic_json>;
-    friend ::nlohmann::detail::parser<basic_json>;
+
+    template<typename BasicJsonType, typename InputType>
+    friend class ::nlohmann::detail::parser;
     friend ::nlohmann::detail::serializer<basic_json>;
     template<typename BasicJsonType>
     friend class ::nlohmann::detail::iter_impl;
     template<typename BasicJsonType, typename CharType>
     friend class ::nlohmann::detail::binary_writer;
-    template<typename BasicJsonType, typename SAX>
+    template<typename BasicJsonType, typename InputType, typename SAX>
     friend class ::nlohmann::detail::binary_reader;
     template<typename BasicJsonType>
     friend class ::nlohmann::detail::json_sax_dom_parser;
     template<typename BasicJsonType>
     friend class ::nlohmann::detail::json_sax_dom_callback_parser;
+    friend class ::nlohmann::detail::exception;
 
     /// workaround type for MSVC
     using basic_json_t = NLOHMANN_BASIC_JSON_TPL;
 
+  JSON_PRIVATE_UNLESS_TESTED:
     // convenience aliases for types residing in namespace detail;
-    using lexer = ::nlohmann::detail::lexer<basic_json>;
-    using parser = ::nlohmann::detail::parser<basic_json>;
+    using lexer = ::nlohmann::detail::lexer_base<basic_json>;
 
+    template<typename InputAdapterType>
+    static ::nlohmann::detail::parser<basic_json, InputAdapterType> parser(
+        InputAdapterType adapter,
+        detail::parser_callback_t<basic_json>cb = nullptr,
+        const bool allow_exceptions = true,
+        const bool ignore_comments = false
+                                 )
+    {
+        return ::nlohmann::detail::parser<basic_json, InputAdapterType>(std::move(adapter),
+                std::move(cb), allow_exceptions, ignore_comments);
+    }
+
+  private:
     using primitive_iterator_t = ::nlohmann::detail::primitive_iterator_t;
     template<typename BasicJsonType>
     using internal_iterator = ::nlohmann::detail::internal_iterator<BasicJsonType>;
@@ -11284,9 +17331,11 @@ class basic_json
     template<typename CharType>
     using output_adapter_t = ::nlohmann::detail::output_adapter_t<CharType>;
 
-    using binary_reader = ::nlohmann::detail::binary_reader<basic_json>;
+    template<typename InputType>
+    using binary_reader = ::nlohmann::detail::binary_reader<basic_json, InputType>;
     template<typename CharType> using binary_writer = ::nlohmann::detail::binary_writer<basic_json, CharType>;
 
+  JSON_PRIVATE_UNLESS_TESTED:
     using serializer = ::nlohmann::detail::serializer<basic_json>;
 
   public:
@@ -11295,6 +17344,10 @@ class basic_json
     using json_pointer = ::nlohmann::json_pointer<basic_json>;
     template<typename T, typename SFINAE>
     using json_serializer = JSONSerializer<T, SFINAE>;
+    /// how to treat decoding errors
+    using error_handler_t = detail::error_handler_t;
+    /// how to treat CBOR tags
+    using cbor_tag_handler_t = detail::cbor_tag_handler_t;
     /// helper type for initializer lists of basic_json values
     using initializer_list_t = std::initializer_list<detail::json_ref<basic_json>>;
 
@@ -11310,17 +17363,11 @@ class basic_json
     /// Classes to implement user-defined exceptions.
     /// @{
 
-    /// @copydoc detail::exception
     using exception = detail::exception;
-    /// @copydoc detail::parse_error
     using parse_error = detail::parse_error;
-    /// @copydoc detail::invalid_iterator
     using invalid_iterator = detail::invalid_iterator;
-    /// @copydoc detail::type_error
     using type_error = detail::type_error;
-    /// @copydoc detail::out_of_range
     using out_of_range = detail::out_of_range;
-    /// @copydoc detail::other_error
     using other_error = detail::other_error;
 
     /// @}
@@ -11368,45 +17415,21 @@ class basic_json
     /// @}
 
 
-    /*!
-    @brief returns the allocator associated with the container
-    */
+    /// @brief returns the allocator associated with the container
+    /// @sa https://json.nlohmann.me/api/basic_json/get_allocator/
     static allocator_type get_allocator()
     {
         return allocator_type();
     }
 
-    /*!
-    @brief returns version information on the library
-
-    This function returns a JSON object with information about the library,
-    including the version number and information on the platform and compiler.
-
-    @return JSON object holding version information
-    key         | description
-    ----------- | ---------------
-    `compiler`  | Information on the used compiler. It is an object with the following keys: `c++` (the used C++ standard), `family` (the compiler family; possible values are `clang`, `icc`, `gcc`, `ilecpp`, `msvc`, `pgcpp`, `sunpro`, and `unknown`), and `version` (the compiler version).
-    `copyright` | The copyright line for the library as string.
-    `name`      | The name of the library as string.
-    `platform`  | The used platform as string. Possible values are `win32`, `linux`, `apple`, `unix`, and `unknown`.
-    `url`       | The URL of the project as string.
-    `version`   | The version of the library. It is an object with the following keys: `major`, `minor`, and `patch` as defined by [Semantic Versioning](http://semver.org), and `string` (the version string).
-
-    @liveexample{The following code shows an example output of the `meta()`
-    function.,meta}
-
-    @exceptionsafety Strong guarantee: if an exception is thrown, there are no
-    changes to any JSON value.
-
-    @complexity Constant.
-
-    @since 2.1.0
-    */
+    /// @brief returns version information on the library
+    /// @sa https://json.nlohmann.me/api/basic_json/meta/
+    JSON_HEDLEY_WARN_UNUSED_RESULT
     static basic_json meta()
     {
         basic_json result;
 
-        result["copyright"] = "(C) 2013-2017 Niels Lohmann";
+        result["copyright"] = "(C) 2013-2022 Niels Lohmann";
         result["name"] = "JSON for Modern C++";
         result["url"] = "https://github.com/nlohmann/json";
         result["version"]["string"] =
@@ -11467,6 +17490,8 @@ class basic_json
     /// the template arguments passed to class @ref basic_json.
     /// @{
 
+    /// @brief object key comparator type
+    /// @sa https://json.nlohmann.me/api/basic_json/object_comparator_t/
 #if defined(JSON_HAS_CPP_14)
     // Use transparent comparator if possible, combined with perfect forwarding
     // on find() and count() calls prevents unnecessary string construction.
@@ -11475,456 +17500,69 @@ class basic_json
     using object_comparator_t = std::less<StringType>;
 #endif
 
-    /*!
-    @brief a type for an object
-
-    [RFC 7159](http://rfc7159.net/rfc7159) describes JSON objects as follows:
-    > An object is an unordered collection of zero or more name/value pairs,
-    > where a name is a string and a value is a string, number, boolean, null,
-    > object, or array.
-
-    To store objects in C++, a type is defined by the template parameters
-    described below.
-
-    @tparam ObjectType  the container to store objects (e.g., `std::map` or
-    `std::unordered_map`)
-    @tparam StringType the type of the keys or names (e.g., `std::string`).
-    The comparison function `std::less<StringType>` is used to order elements
-    inside the container.
-    @tparam AllocatorType the allocator to use for objects (e.g.,
-    `std::allocator`)
-
-    #### Default type
-
-    With the default values for @a ObjectType (`std::map`), @a StringType
-    (`std::string`), and @a AllocatorType (`std::allocator`), the default
-    value for @a object_t is:
-
-    @code {.cpp}
-    std::map<
-      std::string, // key_type
-      basic_json, // value_type
-      std::less<std::string>, // key_compare
-      std::allocator<std::pair<const std::string, basic_json>> // allocator_type
-    >
-    @endcode
-
-    #### Behavior
-
-    The choice of @a object_t influences the behavior of the JSON class. With
-    the default type, objects have the following behavior:
-
-    - When all names are unique, objects will be interoperable in the sense
-      that all software implementations receiving that object will agree on
-      the name-value mappings.
-    - When the names within an object are not unique, it is unspecified which
-      one of the values for a given key will be chosen. For instance,
-      `{"key": 2, "key": 1}` could be equal to either `{"key": 1}` or
-      `{"key": 2}`.
-    - Internally, name/value pairs are stored in lexicographical order of the
-      names. Objects will also be serialized (see @ref dump) in this order.
-      For instance, `{"b": 1, "a": 2}` and `{"a": 2, "b": 1}` will be stored
-      and serialized as `{"a": 2, "b": 1}`.
-    - When comparing objects, the order of the name/value pairs is irrelevant.
-      This makes objects interoperable in the sense that they will not be
-      affected by these differences. For instance, `{"b": 1, "a": 2}` and
-      `{"a": 2, "b": 1}` will be treated as equal.
-
-    #### Limits
-
-    [RFC 7159](http://rfc7159.net/rfc7159) specifies:
-    > An implementation may set limits on the maximum depth of nesting.
-
-    In this class, the object's limit of nesting is not explicitly constrained.
-    However, a maximum depth of nesting may be introduced by the compiler or
-    runtime environment. A theoretical limit can be queried by calling the
-    @ref max_size function of a JSON object.
-
-    #### Storage
-
-    Objects are stored as pointers in a @ref basic_json type. That is, for any
-    access to object values, a pointer of type `object_t*` must be
-    dereferenced.
-
-    @sa @ref array_t -- type for an array value
-
-    @since version 1.0.0
-
-    @note The order name/value pairs are added to the object is *not*
-    preserved by the library. Therefore, iterating an object may return
-    name/value pairs in a different order than they were originally stored. In
-    fact, keys will be traversed in alphabetical order as `std::map` with
-    `std::less` is used by default. Please note this behavior conforms to [RFC
-    7159](http://rfc7159.net/rfc7159), because any order implements the
-    specified "unordered" nature of JSON objects.
-    */
+    /// @brief a type for an object
+    /// @sa https://json.nlohmann.me/api/basic_json/object_t/
     using object_t = ObjectType<StringType,
           basic_json,
           object_comparator_t,
           AllocatorType<std::pair<const StringType,
           basic_json>>>;
 
-    /*!
-    @brief a type for an array
-
-    [RFC 7159](http://rfc7159.net/rfc7159) describes JSON arrays as follows:
-    > An array is an ordered sequence of zero or more values.
-
-    To store objects in C++, a type is defined by the template parameters
-    explained below.
-
-    @tparam ArrayType  container type to store arrays (e.g., `std::vector` or
-    `std::list`)
-    @tparam AllocatorType allocator to use for arrays (e.g., `std::allocator`)
-
-    #### Default type
-
-    With the default values for @a ArrayType (`std::vector`) and @a
-    AllocatorType (`std::allocator`), the default value for @a array_t is:
-
-    @code {.cpp}
-    std::vector<
-      basic_json, // value_type
-      std::allocator<basic_json> // allocator_type
-    >
-    @endcode
-
-    #### Limits
-
-    [RFC 7159](http://rfc7159.net/rfc7159) specifies:
-    > An implementation may set limits on the maximum depth of nesting.
-
-    In this class, the array's limit of nesting is not explicitly constrained.
-    However, a maximum depth of nesting may be introduced by the compiler or
-    runtime environment. A theoretical limit can be queried by calling the
-    @ref max_size function of a JSON array.
-
-    #### Storage
-
-    Arrays are stored as pointers in a @ref basic_json type. That is, for any
-    access to array values, a pointer of type `array_t*` must be dereferenced.
-
-    @sa @ref object_t -- type for an object value
-
-    @since version 1.0.0
-    */
+    /// @brief a type for an array
+    /// @sa https://json.nlohmann.me/api/basic_json/array_t/
     using array_t = ArrayType<basic_json, AllocatorType<basic_json>>;
 
-    /*!
-    @brief a type for a string
-
-    [RFC 7159](http://rfc7159.net/rfc7159) describes JSON strings as follows:
-    > A string is a sequence of zero or more Unicode characters.
-
-    To store objects in C++, a type is defined by the template parameter
-    described below. Unicode values are split by the JSON class into
-    byte-sized characters during deserialization.
-
-    @tparam StringType  the container to store strings (e.g., `std::string`).
-    Note this container is used for keys/names in objects, see @ref object_t.
-
-    #### Default type
-
-    With the default values for @a StringType (`std::string`), the default
-    value for @a string_t is:
-
-    @code {.cpp}
-    std::string
-    @endcode
-
-    #### Encoding
-
-    Strings are stored in UTF-8 encoding. Therefore, functions like
-    `std::string::size()` or `std::string::length()` return the number of
-    bytes in the string rather than the number of characters or glyphs.
-
-    #### String comparison
-
-    [RFC 7159](http://rfc7159.net/rfc7159) states:
-    > Software implementations are typically required to test names of object
-    > members for equality. Implementations that transform the textual
-    > representation into sequences of Unicode code units and then perform the
-    > comparison numerically, code unit by code unit, are interoperable in the
-    > sense that implementations will agree in all cases on equality or
-    > inequality of two strings. For example, implementations that compare
-    > strings with escaped characters unconverted may incorrectly find that
-    > `"a\\b"` and `"a\u005Cb"` are not equal.
-
-    This implementation is interoperable as it does compare strings code unit
-    by code unit.
-
-    #### Storage
-
-    String values are stored as pointers in a @ref basic_json type. That is,
-    for any access to string values, a pointer of type `string_t*` must be
-    dereferenced.
-
-    @since version 1.0.0
-    */
+    /// @brief a type for a string
+    /// @sa https://json.nlohmann.me/api/basic_json/string_t/
     using string_t = StringType;
 
-    /*!
-    @brief a type for a boolean
-
-    [RFC 7159](http://rfc7159.net/rfc7159) implicitly describes a boolean as a
-    type which differentiates the two literals `true` and `false`.
-
-    To store objects in C++, a type is defined by the template parameter @a
-    BooleanType which chooses the type to use.
-
-    #### Default type
-
-    With the default values for @a BooleanType (`bool`), the default value for
-    @a boolean_t is:
-
-    @code {.cpp}
-    bool
-    @endcode
-
-    #### Storage
-
-    Boolean values are stored directly inside a @ref basic_json type.
-
-    @since version 1.0.0
-    */
+    /// @brief a type for a boolean
+    /// @sa https://json.nlohmann.me/api/basic_json/boolean_t/
     using boolean_t = BooleanType;
 
-    /*!
-    @brief a type for a number (integer)
-
-    [RFC 7159](http://rfc7159.net/rfc7159) describes numbers as follows:
-    > The representation of numbers is similar to that used in most
-    > programming languages. A number is represented in base 10 using decimal
-    > digits. It contains an integer component that may be prefixed with an
-    > optional minus sign, which may be followed by a fraction part and/or an
-    > exponent part. Leading zeros are not allowed. (...) Numeric values that
-    > cannot be represented in the grammar below (such as Infinity and NaN)
-    > are not permitted.
-
-    This description includes both integer and floating-point numbers.
-    However, C++ allows more precise storage if it is known whether the number
-    is a signed integer, an unsigned integer or a floating-point number.
-    Therefore, three different types, @ref number_integer_t, @ref
-    number_unsigned_t and @ref number_float_t are used.
-
-    To store integer numbers in C++, a type is defined by the template
-    parameter @a NumberIntegerType which chooses the type to use.
-
-    #### Default type
-
-    With the default values for @a NumberIntegerType (`int64_t`), the default
-    value for @a number_integer_t is:
-
-    @code {.cpp}
-    int64_t
-    @endcode
-
-    #### Default behavior
-
-    - The restrictions about leading zeros is not enforced in C++. Instead,
-      leading zeros in integer literals lead to an interpretation as octal
-      number. Internally, the value will be stored as decimal number. For
-      instance, the C++ integer literal `010` will be serialized to `8`.
-      During deserialization, leading zeros yield an error.
-    - Not-a-number (NaN) values will be serialized to `null`.
-
-    #### Limits
-
-    [RFC 7159](http://rfc7159.net/rfc7159) specifies:
-    > An implementation may set limits on the range and precision of numbers.
-
-    When the default type is used, the maximal integer number that can be
-    stored is `9223372036854775807` (INT64_MAX) and the minimal integer number
-    that can be stored is `-9223372036854775808` (INT64_MIN). Integer numbers
-    that are out of range will yield over/underflow when used in a
-    constructor. During deserialization, too large or small integer numbers
-    will be automatically be stored as @ref number_unsigned_t or @ref
-    number_float_t.
-
-    [RFC 7159](http://rfc7159.net/rfc7159) further states:
-    > Note that when such software is used, numbers that are integers and are
-    > in the range \f$[-2^{53}+1, 2^{53}-1]\f$ are interoperable in the sense
-    > that implementations will agree exactly on their numeric values.
-
-    As this range is a subrange of the exactly supported range [INT64_MIN,
-    INT64_MAX], this class's integer type is interoperable.
-
-    #### Storage
-
-    Integer number values are stored directly inside a @ref basic_json type.
-
-    @sa @ref number_float_t -- type for number values (floating-point)
-
-    @sa @ref number_unsigned_t -- type for number values (unsigned integer)
-
-    @since version 1.0.0
-    */
+    /// @brief a type for a number (integer)
+    /// @sa https://json.nlohmann.me/api/basic_json/number_integer_t/
     using number_integer_t = NumberIntegerType;
 
-    /*!
-    @brief a type for a number (unsigned)
-
-    [RFC 7159](http://rfc7159.net/rfc7159) describes numbers as follows:
-    > The representation of numbers is similar to that used in most
-    > programming languages. A number is represented in base 10 using decimal
-    > digits. It contains an integer component that may be prefixed with an
-    > optional minus sign, which may be followed by a fraction part and/or an
-    > exponent part. Leading zeros are not allowed. (...) Numeric values that
-    > cannot be represented in the grammar below (such as Infinity and NaN)
-    > are not permitted.
-
-    This description includes both integer and floating-point numbers.
-    However, C++ allows more precise storage if it is known whether the number
-    is a signed integer, an unsigned integer or a floating-point number.
-    Therefore, three different types, @ref number_integer_t, @ref
-    number_unsigned_t and @ref number_float_t are used.
-
-    To store unsigned integer numbers in C++, a type is defined by the
-    template parameter @a NumberUnsignedType which chooses the type to use.
-
-    #### Default type
-
-    With the default values for @a NumberUnsignedType (`uint64_t`), the
-    default value for @a number_unsigned_t is:
-
-    @code {.cpp}
-    uint64_t
-    @endcode
-
-    #### Default behavior
-
-    - The restrictions about leading zeros is not enforced in C++. Instead,
-      leading zeros in integer literals lead to an interpretation as octal
-      number. Internally, the value will be stored as decimal number. For
-      instance, the C++ integer literal `010` will be serialized to `8`.
-      During deserialization, leading zeros yield an error.
-    - Not-a-number (NaN) values will be serialized to `null`.
-
-    #### Limits
-
-    [RFC 7159](http://rfc7159.net/rfc7159) specifies:
-    > An implementation may set limits on the range and precision of numbers.
-
-    When the default type is used, the maximal integer number that can be
-    stored is `18446744073709551615` (UINT64_MAX) and the minimal integer
-    number that can be stored is `0`. Integer numbers that are out of range
-    will yield over/underflow when used in a constructor. During
-    deserialization, too large or small integer numbers will be automatically
-    be stored as @ref number_integer_t or @ref number_float_t.
-
-    [RFC 7159](http://rfc7159.net/rfc7159) further states:
-    > Note that when such software is used, numbers that are integers and are
-    > in the range \f$[-2^{53}+1, 2^{53}-1]\f$ are interoperable in the sense
-    > that implementations will agree exactly on their numeric values.
-
-    As this range is a subrange (when considered in conjunction with the
-    number_integer_t type) of the exactly supported range [0, UINT64_MAX],
-    this class's integer type is interoperable.
-
-    #### Storage
-
-    Integer number values are stored directly inside a @ref basic_json type.
-
-    @sa @ref number_float_t -- type for number values (floating-point)
-    @sa @ref number_integer_t -- type for number values (integer)
-
-    @since version 2.0.0
-    */
+    /// @brief a type for a number (unsigned)
+    /// @sa https://json.nlohmann.me/api/basic_json/number_unsigned_t/
     using number_unsigned_t = NumberUnsignedType;
 
-    /*!
-    @brief a type for a number (floating-point)
-
-    [RFC 7159](http://rfc7159.net/rfc7159) describes numbers as follows:
-    > The representation of numbers is similar to that used in most
-    > programming languages. A number is represented in base 10 using decimal
-    > digits. It contains an integer component that may be prefixed with an
-    > optional minus sign, which may be followed by a fraction part and/or an
-    > exponent part. Leading zeros are not allowed. (...) Numeric values that
-    > cannot be represented in the grammar below (such as Infinity and NaN)
-    > are not permitted.
-
-    This description includes both integer and floating-point numbers.
-    However, C++ allows more precise storage if it is known whether the number
-    is a signed integer, an unsigned integer or a floating-point number.
-    Therefore, three different types, @ref number_integer_t, @ref
-    number_unsigned_t and @ref number_float_t are used.
-
-    To store floating-point numbers in C++, a type is defined by the template
-    parameter @a NumberFloatType which chooses the type to use.
-
-    #### Default type
-
-    With the default values for @a NumberFloatType (`double`), the default
-    value for @a number_float_t is:
-
-    @code {.cpp}
-    double
-    @endcode
-
-    #### Default behavior
-
-    - The restrictions about leading zeros is not enforced in C++. Instead,
-      leading zeros in floating-point literals will be ignored. Internally,
-      the value will be stored as decimal number. For instance, the C++
-      floating-point literal `01.2` will be serialized to `1.2`. During
-      deserialization, leading zeros yield an error.
-    - Not-a-number (NaN) values will be serialized to `null`.
-
-    #### Limits
-
-    [RFC 7159](http://rfc7159.net/rfc7159) states:
-    > This specification allows implementations to set limits on the range and
-    > precision of numbers accepted. Since software that implements IEEE
-    > 754-2008 binary64 (double precision) numbers is generally available and
-    > widely used, good interoperability can be achieved by implementations
-    > that expect no more precision or range than these provide, in the sense
-    > that implementations will approximate JSON numbers within the expected
-    > precision.
-
-    This implementation does exactly follow this approach, as it uses double
-    precision floating-point numbers. Note values smaller than
-    `-1.79769313486232e+308` and values greater than `1.79769313486232e+308`
-    will be stored as NaN internally and be serialized to `null`.
-
-    #### Storage
-
-    Floating-point number values are stored directly inside a @ref basic_json
-    type.
-
-    @sa @ref number_integer_t -- type for number values (integer)
-
-    @sa @ref number_unsigned_t -- type for number values (unsigned integer)
-
-    @since version 1.0.0
-    */
+    /// @brief a type for a number (floating-point)
+    /// @sa https://json.nlohmann.me/api/basic_json/number_float_t/
     using number_float_t = NumberFloatType;
 
+    /// @brief a type for a packed binary type
+    /// @sa https://json.nlohmann.me/api/basic_json/binary_t/
+    using binary_t = nlohmann::byte_container_with_subtype<BinaryType>;
+
     /// @}
 
   private:
 
     /// helper for exception-safe object creation
     template<typename T, typename... Args>
+    JSON_HEDLEY_RETURNS_NON_NULL
     static T* create(Args&& ... args)
     {
         AllocatorType<T> alloc;
         using AllocatorTraits = std::allocator_traits<AllocatorType<T>>;
 
-        auto deleter = [&](T * object)
+        auto deleter = [&](T * obj)
         {
-            AllocatorTraits::deallocate(alloc, object, 1);
+            AllocatorTraits::deallocate(alloc, obj, 1);
         };
-        std::unique_ptr<T, decltype(deleter)> object(AllocatorTraits::allocate(alloc, 1), deleter);
-        AllocatorTraits::construct(alloc, object.get(), std::forward<Args>(args)...);
-        assert(object != nullptr);
-        return object.release();
+        std::unique_ptr<T, decltype(deleter)> obj(AllocatorTraits::allocate(alloc, 1), deleter);
+        AllocatorTraits::construct(alloc, obj.get(), std::forward<Args>(args)...);
+        JSON_ASSERT(obj != nullptr);
+        return obj.release();
     }
 
     ////////////////////////
     // JSON value storage //
     ////////////////////////
 
+  JSON_PRIVATE_UNLESS_TESTED:
     /*!
     @brief a JSON value
 
@@ -11941,6 +17579,7 @@ class basic_json
     number    | number_integer  | @ref number_integer_t
     number    | number_unsigned | @ref number_unsigned_t
     number    | number_float    | @ref number_float_t
+    binary    | binary          | pointer to @ref binary_t
     null      | null            | *no value is stored*
 
     @note Variable-length types (objects, arrays, and strings) are stored as
@@ -11957,6 +17596,8 @@ class basic_json
         array_t* array;
         /// string (stored with pointer to save storage)
         string_t* string;
+        /// binary (stored with pointer to save storage)
+        binary_t* binary;
         /// boolean
         boolean_t boolean;
         /// number (integer)
@@ -11999,27 +17640,33 @@ class basic_json
                     break;
                 }
 
+                case value_t::binary:
+                {
+                    binary = create<binary_t>();
+                    break;
+                }
+
                 case value_t::boolean:
                 {
-                    boolean = boolean_t(false);
+                    boolean = static_cast<boolean_t>(false);
                     break;
                 }
 
                 case value_t::number_integer:
                 {
-                    number_integer = number_integer_t(0);
+                    number_integer = static_cast<number_integer_t>(0);
                     break;
                 }
 
                 case value_t::number_unsigned:
                 {
-                    number_unsigned = number_unsigned_t(0);
+                    number_unsigned = static_cast<number_unsigned_t>(0);
                     break;
                 }
 
                 case value_t::number_float:
                 {
-                    number_float = number_float_t(0.0);
+                    number_float = static_cast<number_float_t>(0.0);
                     break;
                 }
 
@@ -12029,12 +17676,13 @@ class basic_json
                     break;
                 }
 
+                case value_t::discarded:
                 default:
                 {
                     object = nullptr;  // silence warning, see #821
-                    if (JSON_UNLIKELY(t == value_t::null))
+                    if (JSON_HEDLEY_UNLIKELY(t == value_t::null))
                     {
-                        JSON_THROW(other_error::create(500, "961c151d2e87f2686a955a9be24d316f1362bf21 3.2.0")); // LCOV_EXCL_LINE
+                        JSON_THROW(other_error::create(500, "961c151d2e87f2686a955a9be24d316f1362bf21 3.10.5", basic_json())); // LCOV_EXCL_LINE
                     }
                     break;
                 }
@@ -12042,43 +17690,86 @@ class basic_json
         }
 
         /// constructor for strings
-        json_value(const string_t& value)
-        {
-            string = create<string_t>(value);
-        }
+        json_value(const string_t& value) : string(create<string_t>(value)) {}
 
         /// constructor for rvalue strings
-        json_value(string_t&& value)
-        {
-            string = create<string_t>(std::move(value));
-        }
+        json_value(string_t&& value) : string(create<string_t>(std::move(value))) {}
 
         /// constructor for objects
-        json_value(const object_t& value)
-        {
-            object = create<object_t>(value);
-        }
+        json_value(const object_t& value) : object(create<object_t>(value)) {}
 
         /// constructor for rvalue objects
-        json_value(object_t&& value)
-        {
-            object = create<object_t>(std::move(value));
-        }
+        json_value(object_t&& value) : object(create<object_t>(std::move(value))) {}
 
         /// constructor for arrays
-        json_value(const array_t& value)
-        {
-            array = create<array_t>(value);
-        }
+        json_value(const array_t& value) : array(create<array_t>(value)) {}
 
         /// constructor for rvalue arrays
-        json_value(array_t&& value)
-        {
-            array = create<array_t>(std::move(value));
-        }
+        json_value(array_t&& value) : array(create<array_t>(std::move(value))) {}
 
-        void destroy(value_t t) noexcept
+        /// constructor for binary arrays
+        json_value(const typename binary_t::container_type& value) : binary(create<binary_t>(value)) {}
+
+        /// constructor for rvalue binary arrays
+        json_value(typename binary_t::container_type&& value) : binary(create<binary_t>(std::move(value))) {}
+
+        /// constructor for binary arrays (internal type)
+        json_value(const binary_t& value) : binary(create<binary_t>(value)) {}
+
+        /// constructor for rvalue binary arrays (internal type)
+        json_value(binary_t&& value) : binary(create<binary_t>(std::move(value))) {}
+
+        void destroy(value_t t)
         {
+            if (t == value_t::array || t == value_t::object)
+            {
+                // flatten the current json_value to a heap-allocated stack
+                std::vector<basic_json> stack;
+
+                // move the top-level items to stack
+                if (t == value_t::array)
+                {
+                    stack.reserve(array->size());
+                    std::move(array->begin(), array->end(), std::back_inserter(stack));
+                }
+                else
+                {
+                    stack.reserve(object->size());
+                    for (auto&& it : *object)
+                    {
+                        stack.push_back(std::move(it.second));
+                    }
+                }
+
+                while (!stack.empty())
+                {
+                    // move the last item to local variable to be processed
+                    basic_json current_item(std::move(stack.back()));
+                    stack.pop_back();
+
+                    // if current_item is array/object, move
+                    // its children to the stack to be processed later
+                    if (current_item.is_array())
+                    {
+                        std::move(current_item.m_value.array->begin(), current_item.m_value.array->end(), std::back_inserter(stack));
+
+                        current_item.m_value.array->clear();
+                    }
+                    else if (current_item.is_object())
+                    {
+                        for (auto&& it : *current_item.m_value.object)
+                        {
+                            stack.push_back(std::move(it.second));
+                        }
+
+                        current_item.m_value.object->clear();
+                    }
+
+                    // it's now safe that current_item get destructed
+                    // since it doesn't have any children
+                }
+            }
+
             switch (t)
             {
                 case value_t::object:
@@ -12105,6 +17796,20 @@ class basic_json
                     break;
                 }
 
+                case value_t::binary:
+                {
+                    AllocatorType<binary_t> alloc;
+                    std::allocator_traits<decltype(alloc)>::destroy(alloc, binary);
+                    std::allocator_traits<decltype(alloc)>::deallocate(alloc, binary, 1);
+                    break;
+                }
+
+                case value_t::null:
+                case value_t::boolean:
+                case value_t::number_integer:
+                case value_t::number_unsigned:
+                case value_t::number_float:
+                case value_t::discarded:
                 default:
                 {
                     break;
@@ -12113,6 +17818,7 @@ class basic_json
         }
     };
 
+  private:
     /*!
     @brief checks the class invariants
 
@@ -12121,12 +17827,123 @@ class basic_json
     invariant. Furthermore, it has to be called each time the type of a JSON
     value is changed, because the invariant expresses a relationship between
     @a m_type and @a m_value.
+
+    Furthermore, the parent relation is checked for arrays and objects: If
+    @a check_parents true and the value is an array or object, then the
+    container's elements must have the current value as parent.
+
+    @param[in] check_parents  whether the parent relation should be checked.
+               The value is true by default and should only be set to false
+               during destruction of objects when the invariant does not
+               need to hold.
     */
-    void assert_invariant() const noexcept
+    void assert_invariant(bool check_parents = true) const noexcept
     {
-        assert(m_type != value_t::object or m_value.object != nullptr);
-        assert(m_type != value_t::array or m_value.array != nullptr);
-        assert(m_type != value_t::string or m_value.string != nullptr);
+        JSON_ASSERT(m_type != value_t::object || m_value.object != nullptr);
+        JSON_ASSERT(m_type != value_t::array || m_value.array != nullptr);
+        JSON_ASSERT(m_type != value_t::string || m_value.string != nullptr);
+        JSON_ASSERT(m_type != value_t::binary || m_value.binary != nullptr);
+
+#if JSON_DIAGNOSTICS
+        JSON_TRY
+        {
+            // cppcheck-suppress assertWithSideEffect
+            JSON_ASSERT(!check_parents || !is_structured() || std::all_of(begin(), end(), [this](const basic_json & j)
+            {
+                return j.m_parent == this;
+            }));
+        }
+        JSON_CATCH(...) {} // LCOV_EXCL_LINE
+#endif
+        static_cast<void>(check_parents);
+    }
+
+    void set_parents()
+    {
+#if JSON_DIAGNOSTICS
+        switch (m_type)
+        {
+            case value_t::array:
+            {
+                for (auto& element : *m_value.array)
+                {
+                    element.m_parent = this;
+                }
+                break;
+            }
+
+            case value_t::object:
+            {
+                for (auto& element : *m_value.object)
+                {
+                    element.second.m_parent = this;
+                }
+                break;
+            }
+
+            case value_t::null:
+            case value_t::string:
+            case value_t::boolean:
+            case value_t::number_integer:
+            case value_t::number_unsigned:
+            case value_t::number_float:
+            case value_t::binary:
+            case value_t::discarded:
+            default:
+                break;
+        }
+#endif
+    }
+
+    iterator set_parents(iterator it, typename iterator::difference_type count_set_parents)
+    {
+#if JSON_DIAGNOSTICS
+        for (typename iterator::difference_type i = 0; i < count_set_parents; ++i)
+        {
+            (it + i)->m_parent = this;
+        }
+#else
+        static_cast<void>(count_set_parents);
+#endif
+        return it;
+    }
+
+    reference set_parent(reference j, std::size_t old_capacity = static_cast<std::size_t>(-1))
+    {
+#if JSON_DIAGNOSTICS
+        if (old_capacity != static_cast<std::size_t>(-1))
+        {
+            // see https://github.com/nlohmann/json/issues/2838
+            JSON_ASSERT(type() == value_t::array);
+            if (JSON_HEDLEY_UNLIKELY(m_value.array->capacity() != old_capacity))
+            {
+                // capacity has changed: update all parents
+                set_parents();
+                return j;
+            }
+        }
+
+        // ordered_json uses a vector internally, so pointers could have
+        // been invalidated; see https://github.com/nlohmann/json/issues/2962
+#ifdef JSON_HEDLEY_MSVC_VERSION
+#pragma warning(push )
+#pragma warning(disable : 4127) // ignore warning to replace if with if constexpr
+#endif
+        if (detail::is_ordered_map<object_t>::value)
+        {
+            set_parents();
+            return j;
+        }
+#ifdef JSON_HEDLEY_MSVC_VERSION
+#pragma warning( pop )
+#endif
+
+        j.m_parent = this;
+#else
+        static_cast<void>(j);
+        static_cast<void>(old_capacity);
+#endif
+        return j;
     }
 
   public:
@@ -12134,73 +17951,13 @@ class basic_json
     // JSON parser callback //
     //////////////////////////
 
-    /*!
-    @brief parser event types
+    /// @brief parser event types
+    /// @sa https://json.nlohmann.me/api/basic_json/parse_event_t/
+    using parse_event_t = detail::parse_event_t;
 
-    The parser callback distinguishes the following events:
-    - `object_start`: the parser read `{` and started to process a JSON object
-    - `key`: the parser read a key of a value in an object
-    - `object_end`: the parser read `}` and finished processing a JSON object
-    - `array_start`: the parser read `[` and started to process a JSON array
-    - `array_end`: the parser read `]` and finished processing a JSON array
-    - `value`: the parser finished reading a JSON value
-
-    @image html callback_events.png "Example when certain parse events are triggered"
-
-    @sa @ref parser_callback_t for more information and examples
-    */
-    using parse_event_t = typename parser::parse_event_t;
-
-    /*!
-    @brief per-element parser callback type
-
-    With a parser callback function, the result of parsing a JSON text can be
-    influenced. When passed to @ref parse, it is called on certain events
-    (passed as @ref parse_event_t via parameter @a event) with a set recursion
-    depth @a depth and context JSON value @a parsed. The return value of the
-    callback function is a boolean indicating whether the element that emitted
-    the callback shall be kept or not.
-
-    We distinguish six scenarios (determined by the event type) in which the
-    callback function can be called. The following table describes the values
-    of the parameters @a depth, @a event, and @a parsed.
-
-    parameter @a event | description | parameter @a depth | parameter @a parsed
-    ------------------ | ----------- | ------------------ | -------------------
-    parse_event_t::object_start | the parser read `{` and started to process a JSON object | depth of the parent of the JSON object | a JSON value with type discarded
-    parse_event_t::key | the parser read a key of a value in an object | depth of the currently parsed JSON object | a JSON string containing the key
-    parse_event_t::object_end | the parser read `}` and finished processing a JSON object | depth of the parent of the JSON object | the parsed JSON object
-    parse_event_t::array_start | the parser read `[` and started to process a JSON array | depth of the parent of the JSON array | a JSON value with type discarded
-    parse_event_t::array_end | the parser read `]` and finished processing a JSON array | depth of the parent of the JSON array | the parsed JSON array
-    parse_event_t::value | the parser finished reading a JSON value | depth of the value | the parsed JSON value
-
-    @image html callback_events.png "Example when certain parse events are triggered"
-
-    Discarding a value (i.e., returning `false`) has different effects
-    depending on the context in which function was called:
-
-    - Discarded values in structured types are skipped. That is, the parser
-      will behave as if the discarded value was never read.
-    - In case a value outside a structured type is skipped, it is replaced
-      with `null`. This case happens if the top-level element is skipped.
-
-    @param[in] depth  the depth of the recursion during parsing
-
-    @param[in] event  an event of type parse_event_t indicating the context in
-    the callback function has been called
-
-    @param[in,out] parsed  the current intermediate parse result; note that
-    writing to this value has no effect for parse_event_t::key events
-
-    @return Whether the JSON value which called the function during parsing
-    should be kept (`true`) or not (`false`). In the latter case, it is either
-    skipped completely or replaced by an empty discarded object.
-
-    @sa @ref parse for examples
-
-    @since version 1.0.0
-    */
-    using parser_callback_t = typename parser::parser_callback_t;
+    /// @brief per-element parser callback type
+    /// @sa https://json.nlohmann.me/api/basic_json/parser_callback_t/
+    using parser_callback_t = detail::parser_callback_t<basic_json>;
 
     //////////////////
     // constructors //
@@ -12211,163 +17968,42 @@ class basic_json
     /// assignment, static functions creating objects, and the destructor.
     /// @{
 
-    /*!
-    @brief create an empty value with a given type
-
-    Create an empty JSON value with a given type. The value will be default
-    initialized with an empty value which depends on the type:
-
-    Value type  | initial value
-    ----------- | -------------
-    null        | `null`
-    boolean     | `false`
-    string      | `""`
-    number      | `0`
-    object      | `{}`
-    array       | `[]`
-
-    @param[in] v  the type of the value to create
-
-    @complexity Constant.
-
-    @exceptionsafety Strong guarantee: if an exception is thrown, there are no
-    changes to any JSON value.
-
-    @liveexample{The following code shows the constructor for different @ref
-    value_t values,basic_json__value_t}
-
-    @sa @ref clear() -- restores the postcondition of this constructor
-
-    @since version 1.0.0
-    */
+    /// @brief create an empty value with a given type
+    /// @sa https://json.nlohmann.me/api/basic_json/basic_json/
     basic_json(const value_t v)
         : m_type(v), m_value(v)
     {
         assert_invariant();
     }
 
-    /*!
-    @brief create a null object
-
-    Create a `null` JSON value. It either takes a null pointer as parameter
-    (explicitly creating `null`) or no parameter (implicitly creating `null`).
-    The passed null pointer itself is not read -- it is only used to choose
-    the right constructor.
-
-    @complexity Constant.
-
-    @exceptionsafety No-throw guarantee: this constructor never throws
-    exceptions.
-
-    @liveexample{The following code shows the constructor with and without a
-    null pointer parameter.,basic_json__nullptr_t}
-
-    @since version 1.0.0
-    */
+    /// @brief create a null object
+    /// @sa https://json.nlohmann.me/api/basic_json/basic_json/
     basic_json(std::nullptr_t = nullptr) noexcept
         : basic_json(value_t::null)
     {
         assert_invariant();
     }
 
-    /*!
-    @brief create a JSON value
-
-    This is a "catch all" constructor for all compatible JSON types; that is,
-    types for which a `to_json()` method exists. The constructor forwards the
-    parameter @a val to that method (to `json_serializer<U>::to_json` method
-    with `U = uncvref_t<CompatibleType>`, to be exact).
-
-    Template type @a CompatibleType includes, but is not limited to, the
-    following types:
-    - **arrays**: @ref array_t and all kinds of compatible containers such as
-      `std::vector`, `std::deque`, `std::list`, `std::forward_list`,
-      `std::array`, `std::valarray`, `std::set`, `std::unordered_set`,
-      `std::multiset`, and `std::unordered_multiset` with a `value_type` from
-      which a @ref basic_json value can be constructed.
-    - **objects**: @ref object_t and all kinds of compatible associative
-      containers such as `std::map`, `std::unordered_map`, `std::multimap`,
-      and `std::unordered_multimap` with a `key_type` compatible to
-      @ref string_t and a `value_type` from which a @ref basic_json value can
-      be constructed.
-    - **strings**: @ref string_t, string literals, and all compatible string
-      containers can be used.
-    - **numbers**: @ref number_integer_t, @ref number_unsigned_t,
-      @ref number_float_t, and all convertible number types such as `int`,
-      `size_t`, `int64_t`, `float` or `double` can be used.
-    - **boolean**: @ref boolean_t / `bool` can be used.
-
-    See the examples below.
-
-    @tparam CompatibleType a type such that:
-    - @a CompatibleType is not derived from `std::istream`,
-    - @a CompatibleType is not @ref basic_json (to avoid hijacking copy/move
-         constructors),
-    - @a CompatibleType is not a different @ref basic_json type (i.e. with different template arguments)
-    - @a CompatibleType is not a @ref basic_json nested type (e.g.,
-         @ref json_pointer, @ref iterator, etc ...)
-    - @ref @ref json_serializer<U> has a
-         `to_json(basic_json_t&, CompatibleType&&)` method
-
-    @tparam U = `uncvref_t<CompatibleType>`
-
-    @param[in] val the value to be forwarded to the respective constructor
-
-    @complexity Usually linear in the size of the passed @a val, also
-                depending on the implementation of the called `to_json()`
-                method.
-
-    @exceptionsafety Depends on the called constructor. For types directly
-    supported by the library (i.e., all types for which no `to_json()` function
-    was provided), strong guarantee holds: if an exception is thrown, there are
-    no changes to any JSON value.
-
-    @liveexample{The following code shows the constructor with several
-    compatible types.,basic_json__CompatibleType}
-
-    @since version 2.1.0
-    */
-    template <typename CompatibleType,
-              typename U = detail::uncvref_t<CompatibleType>,
-              detail::enable_if_t<
-                  not detail::is_basic_json<U>::value and detail::is_compatible_type<basic_json_t, U>::value, int> = 0>
-    basic_json(CompatibleType && val) noexcept(noexcept(
+    /// @brief create a JSON value from compatible types
+    /// @sa https://json.nlohmann.me/api/basic_json/basic_json/
+    template < typename CompatibleType,
+               typename U = detail::uncvref_t<CompatibleType>,
+               detail::enable_if_t <
+                   !detail::is_basic_json<U>::value && detail::is_compatible_type<basic_json_t, U>::value, int > = 0 >
+    basic_json(CompatibleType && val) noexcept(noexcept( // NOLINT(bugprone-forwarding-reference-overload,bugprone-exception-escape)
                 JSONSerializer<U>::to_json(std::declval<basic_json_t&>(),
                                            std::forward<CompatibleType>(val))))
     {
         JSONSerializer<U>::to_json(*this, std::forward<CompatibleType>(val));
+        set_parents();
         assert_invariant();
     }
 
-    /*!
-    @brief create a JSON value from an existing one
-
-    This is a constructor for existing @ref basic_json types.
-    It does not hijack copy/move constructors, since the parameter has different
-    template arguments than the current ones.
-
-    The constructor tries to convert the internal @ref m_value of the parameter.
-
-    @tparam BasicJsonType a type such that:
-    - @a BasicJsonType is a @ref basic_json type.
-    - @a BasicJsonType has different template arguments than @ref basic_json_t.
-
-    @param[in] val the @ref basic_json value to be converted.
-
-    @complexity Usually linear in the size of the passed @a val, also
-                depending on the implementation of the called `to_json()`
-                method.
-
-    @exceptionsafety Depends on the called constructor. For types directly
-    supported by the library (i.e., all types for which no `to_json()` function
-    was provided), strong guarantee holds: if an exception is thrown, there are
-    no changes to any JSON value.
-
-    @since version 3.2.0
-    */
-    template <typename BasicJsonType,
-              detail::enable_if_t<
-                  detail::is_basic_json<BasicJsonType>::value and not std::is_same<basic_json, BasicJsonType>::value, int> = 0>
+    /// @brief create a JSON value from an existing one
+    /// @sa https://json.nlohmann.me/api/basic_json/basic_json/
+    template < typename BasicJsonType,
+               detail::enable_if_t <
+                   detail::is_basic_json<BasicJsonType>::value&& !std::is_same<basic_json, BasicJsonType>::value, int > = 0 >
     basic_json(const BasicJsonType& val)
     {
         using other_boolean_t = typename BasicJsonType::boolean_t;
@@ -12377,6 +18013,7 @@ class basic_json
         using other_string_t = typename BasicJsonType::string_t;
         using other_object_t = typename BasicJsonType::object_t;
         using other_array_t = typename BasicJsonType::array_t;
+        using other_binary_t = typename BasicJsonType::binary_t;
 
         switch (val.type())
         {
@@ -12401,90 +18038,24 @@ class basic_json
             case value_t::array:
                 JSONSerializer<other_array_t>::to_json(*this, val.template get_ref<const other_array_t&>());
                 break;
+            case value_t::binary:
+                JSONSerializer<other_binary_t>::to_json(*this, val.template get_ref<const other_binary_t&>());
+                break;
             case value_t::null:
                 *this = nullptr;
                 break;
             case value_t::discarded:
                 m_type = value_t::discarded;
                 break;
+            default:            // LCOV_EXCL_LINE
+                JSON_ASSERT(false); // NOLINT(cert-dcl03-c,hicpp-static-assert,misc-static-assert) LCOV_EXCL_LINE
         }
+        set_parents();
         assert_invariant();
     }
 
-    /*!
-    @brief create a container (array or object) from an initializer list
-
-    Creates a JSON value of type array or object from the passed initializer
-    list @a init. In case @a type_deduction is `true` (default), the type of
-    the JSON value to be created is deducted from the initializer list @a init
-    according to the following rules:
-
-    1. If the list is empty, an empty JSON object value `{}` is created.
-    2. If the list consists of pairs whose first element is a string, a JSON
-       object value is created where the first elements of the pairs are
-       treated as keys and the second elements are as values.
-    3. In all other cases, an array is created.
-
-    The rules aim to create the best fit between a C++ initializer list and
-    JSON values. The rationale is as follows:
-
-    1. The empty initializer list is written as `{}` which is exactly an empty
-       JSON object.
-    2. C++ has no way of describing mapped types other than to list a list of
-       pairs. As JSON requires that keys must be of type string, rule 2 is the
-       weakest constraint one can pose on initializer lists to interpret them
-       as an object.
-    3. In all other cases, the initializer list could not be interpreted as
-       JSON object type, so interpreting it as JSON array type is safe.
-
-    With the rules described above, the following JSON values cannot be
-    expressed by an initializer list:
-
-    - the empty array (`[]`): use @ref array(initializer_list_t)
-      with an empty initializer list in this case
-    - arrays whose elements satisfy rule 2: use @ref
-      array(initializer_list_t) with the same initializer list
-      in this case
-
-    @note When used without parentheses around an empty initializer list, @ref
-    basic_json() is called instead of this function, yielding the JSON null
-    value.
-
-    @param[in] init  initializer list with JSON values
-
-    @param[in] type_deduction internal parameter; when set to `true`, the type
-    of the JSON value is deducted from the initializer list @a init; when set
-    to `false`, the type provided via @a manual_type is forced. This mode is
-    used by the functions @ref array(initializer_list_t) and
-    @ref object(initializer_list_t).
-
-    @param[in] manual_type internal parameter; when @a type_deduction is set
-    to `false`, the created JSON value will use the provided type (only @ref
-    value_t::array and @ref value_t::object are valid); when @a type_deduction
-    is set to `true`, this parameter has no effect
-
-    @throw type_error.301 if @a type_deduction is `false`, @a manual_type is
-    `value_t::object`, but @a init contains an element which is not a pair
-    whose first element is a string. In this case, the constructor could not
-    create an object. If @a type_deduction would have be `true`, an array
-    would have been created. See @ref object(initializer_list_t)
-    for an example.
-
-    @complexity Linear in the size of the initializer list @a init.
-
-    @exceptionsafety Strong guarantee: if an exception is thrown, there are no
-    changes to any JSON value.
-
-    @liveexample{The example below shows how JSON values are created from
-    initializer lists.,basic_json__list_init_t}
-
-    @sa @ref array(initializer_list_t) -- create a JSON array
-    value from an initializer list
-    @sa @ref object(initializer_list_t) -- create a JSON object
-    value from an initializer list
-
-    @since version 1.0.0
-    */
+    /// @brief create a container (array or object) from an initializer list
+    /// @sa https://json.nlohmann.me/api/basic_json/basic_json/
     basic_json(initializer_list_t init,
                bool type_deduction = true,
                value_t manual_type = value_t::array)
@@ -12494,11 +18065,11 @@ class basic_json
         bool is_an_object = std::all_of(init.begin(), init.end(),
                                         [](const detail::json_ref<basic_json>& element_ref)
         {
-            return (element_ref->is_array() and element_ref->size() == 2 and (*element_ref)[0].is_string());
+            return element_ref->is_array() && element_ref->size() == 2 && (*element_ref)[0].is_string();
         });
 
         // adjust type if type deduction is not wanted
-        if (not type_deduction)
+        if (!type_deduction)
         {
             // if array is wanted, do not create an object though possible
             if (manual_type == value_t::array)
@@ -12507,9 +18078,9 @@ class basic_json
             }
 
             // if object is wanted but impossible, throw an exception
-            if (JSON_UNLIKELY(manual_type == value_t::object and not is_an_object))
+            if (JSON_HEDLEY_UNLIKELY(manual_type == value_t::object && !is_an_object))
             {
-                JSON_THROW(type_error::create(301, "cannot create object from initializer list"));
+                JSON_THROW(type_error::create(301, "cannot create object from initializer list", basic_json()));
             }
         }
 
@@ -12519,13 +18090,13 @@ class basic_json
             m_type = value_t::object;
             m_value = value_t::object;
 
-            std::for_each(init.begin(), init.end(), [this](const detail::json_ref<basic_json>& element_ref)
+            for (auto& element_ref : init)
             {
                 auto element = element_ref.moved_or_copied();
                 m_value.object->emplace(
                     std::move(*((*element.m_value.array)[0].m_value.string)),
                     std::move((*element.m_value.array)[1]));
-            });
+            }
         }
         else
         {
@@ -12534,190 +18105,94 @@ class basic_json
             m_value.array = create<array_t>(init.begin(), init.end());
         }
 
+        set_parents();
         assert_invariant();
     }
 
-    /*!
-    @brief explicitly create an array from an initializer list
+    /// @brief explicitly create a binary array (without subtype)
+    /// @sa https://json.nlohmann.me/api/basic_json/binary/
+    JSON_HEDLEY_WARN_UNUSED_RESULT
+    static basic_json binary(const typename binary_t::container_type& init)
+    {
+        auto res = basic_json();
+        res.m_type = value_t::binary;
+        res.m_value = init;
+        return res;
+    }
 
-    Creates a JSON array value from a given initializer list. That is, given a
-    list of values `a, b, c`, creates the JSON value `[a, b, c]`. If the
-    initializer list is empty, the empty array `[]` is created.
+    /// @brief explicitly create a binary array (with subtype)
+    /// @sa https://json.nlohmann.me/api/basic_json/binary/
+    JSON_HEDLEY_WARN_UNUSED_RESULT
+    static basic_json binary(const typename binary_t::container_type& init, typename binary_t::subtype_type subtype)
+    {
+        auto res = basic_json();
+        res.m_type = value_t::binary;
+        res.m_value = binary_t(init, subtype);
+        return res;
+    }
 
-    @note This function is only needed to express two edge cases that cannot
-    be realized with the initializer list constructor (@ref
-    basic_json(initializer_list_t, bool, value_t)). These cases
-    are:
-    1. creating an array whose elements are all pairs whose first element is a
-    string -- in this case, the initializer list constructor would create an
-    object, taking the first elements as keys
-    2. creating an empty array -- passing the empty initializer list to the
-    initializer list constructor yields an empty object
+    /// @brief explicitly create a binary array
+    /// @sa https://json.nlohmann.me/api/basic_json/binary/
+    JSON_HEDLEY_WARN_UNUSED_RESULT
+    static basic_json binary(typename binary_t::container_type&& init)
+    {
+        auto res = basic_json();
+        res.m_type = value_t::binary;
+        res.m_value = std::move(init);
+        return res;
+    }
 
-    @param[in] init  initializer list with JSON values to create an array from
-    (optional)
+    /// @brief explicitly create a binary array (with subtype)
+    /// @sa https://json.nlohmann.me/api/basic_json/binary/
+    JSON_HEDLEY_WARN_UNUSED_RESULT
+    static basic_json binary(typename binary_t::container_type&& init, typename binary_t::subtype_type subtype)
+    {
+        auto res = basic_json();
+        res.m_type = value_t::binary;
+        res.m_value = binary_t(std::move(init), subtype);
+        return res;
+    }
 
-    @return JSON array value
-
-    @complexity Linear in the size of @a init.
-
-    @exceptionsafety Strong guarantee: if an exception is thrown, there are no
-    changes to any JSON value.
-
-    @liveexample{The following code shows an example for the `array`
-    function.,array}
-
-    @sa @ref basic_json(initializer_list_t, bool, value_t) --
-    create a JSON value from an initializer list
-    @sa @ref object(initializer_list_t) -- create a JSON object
-    value from an initializer list
-
-    @since version 1.0.0
-    */
+    /// @brief explicitly create an array from an initializer list
+    /// @sa https://json.nlohmann.me/api/basic_json/array/
+    JSON_HEDLEY_WARN_UNUSED_RESULT
     static basic_json array(initializer_list_t init = {})
     {
         return basic_json(init, false, value_t::array);
     }
 
-    /*!
-    @brief explicitly create an object from an initializer list
-
-    Creates a JSON object value from a given initializer list. The initializer
-    lists elements must be pairs, and their first elements must be strings. If
-    the initializer list is empty, the empty object `{}` is created.
-
-    @note This function is only added for symmetry reasons. In contrast to the
-    related function @ref array(initializer_list_t), there are
-    no cases which can only be expressed by this function. That is, any
-    initializer list @a init can also be passed to the initializer list
-    constructor @ref basic_json(initializer_list_t, bool, value_t).
-
-    @param[in] init  initializer list to create an object from (optional)
-
-    @return JSON object value
-
-    @throw type_error.301 if @a init is not a list of pairs whose first
-    elements are strings. In this case, no object can be created. When such a
-    value is passed to @ref basic_json(initializer_list_t, bool, value_t),
-    an array would have been created from the passed initializer list @a init.
-    See example below.
-
-    @complexity Linear in the size of @a init.
-
-    @exceptionsafety Strong guarantee: if an exception is thrown, there are no
-    changes to any JSON value.
-
-    @liveexample{The following code shows an example for the `object`
-    function.,object}
-
-    @sa @ref basic_json(initializer_list_t, bool, value_t) --
-    create a JSON value from an initializer list
-    @sa @ref array(initializer_list_t) -- create a JSON array
-    value from an initializer list
-
-    @since version 1.0.0
-    */
+    /// @brief explicitly create an object from an initializer list
+    /// @sa https://json.nlohmann.me/api/basic_json/object/
+    JSON_HEDLEY_WARN_UNUSED_RESULT
     static basic_json object(initializer_list_t init = {})
     {
         return basic_json(init, false, value_t::object);
     }
 
-    /*!
-    @brief construct an array with count copies of given value
-
-    Constructs a JSON array value by creating @a cnt copies of a passed value.
-    In case @a cnt is `0`, an empty array is created.
-
-    @param[in] cnt  the number of JSON copies of @a val to create
-    @param[in] val  the JSON value to copy
-
-    @post `std::distance(begin(),end()) == cnt` holds.
-
-    @complexity Linear in @a cnt.
-
-    @exceptionsafety Strong guarantee: if an exception is thrown, there are no
-    changes to any JSON value.
-
-    @liveexample{The following code shows examples for the @ref
-    basic_json(size_type\, const basic_json&)
-    constructor.,basic_json__size_type_basic_json}
-
-    @since version 1.0.0
-    */
+    /// @brief construct an array with count copies of given value
+    /// @sa https://json.nlohmann.me/api/basic_json/basic_json/
     basic_json(size_type cnt, const basic_json& val)
         : m_type(value_t::array)
     {
         m_value.array = create<array_t>(cnt, val);
+        set_parents();
         assert_invariant();
     }
 
-    /*!
-    @brief construct a JSON container given an iterator range
-
-    Constructs the JSON value with the contents of the range `[first, last)`.
-    The semantics depends on the different types a JSON value can have:
-    - In case of a null type, invalid_iterator.206 is thrown.
-    - In case of other primitive types (number, boolean, or string), @a first
-      must be `begin()` and @a last must be `end()`. In this case, the value is
-      copied. Otherwise, invalid_iterator.204 is thrown.
-    - In case of structured types (array, object), the constructor behaves as
-      similar versions for `std::vector` or `std::map`; that is, a JSON array
-      or object is constructed from the values in the range.
-
-    @tparam InputIT an input iterator type (@ref iterator or @ref
-    const_iterator)
-
-    @param[in] first begin of the range to copy from (included)
-    @param[in] last end of the range to copy from (excluded)
-
-    @pre Iterators @a first and @a last must be initialized. **This
-         precondition is enforced with an assertion (see warning).** If
-         assertions are switched off, a violation of this precondition yields
-         undefined behavior.
-
-    @pre Range `[first, last)` is valid. Usually, this precondition cannot be
-         checked efficiently. Only certain edge cases are detected; see the
-         description of the exceptions below. A violation of this precondition
-         yields undefined behavior.
-
-    @warning A precondition is enforced with a runtime assertion that will
-             result in calling `std::abort` if this precondition is not met.
-             Assertions can be disabled by defining `NDEBUG` at compile time.
-             See https://en.cppreference.com/w/cpp/error/assert for more
-             information.
-
-    @throw invalid_iterator.201 if iterators @a first and @a last are not
-    compatible (i.e., do not belong to the same JSON value). In this case,
-    the range `[first, last)` is undefined.
-    @throw invalid_iterator.204 if iterators @a first and @a last belong to a
-    primitive type (number, boolean, or string), but @a first does not point
-    to the first element any more. In this case, the range `[first, last)` is
-    undefined. See example code below.
-    @throw invalid_iterator.206 if iterators @a first and @a last belong to a
-    null value. In this case, the range `[first, last)` is undefined.
-
-    @complexity Linear in distance between @a first and @a last.
-
-    @exceptionsafety Strong guarantee: if an exception is thrown, there are no
-    changes to any JSON value.
-
-    @liveexample{The example below shows several ways to create JSON values by
-    specifying a subrange with iterators.,basic_json__InputIt_InputIt}
-
-    @since version 1.0.0
-    */
-    template<class InputIT, typename std::enable_if<
-                 std::is_same<InputIT, typename basic_json_t::iterator>::value or
-                 std::is_same<InputIT, typename basic_json_t::const_iterator>::value, int>::type = 0>
+    /// @brief construct a JSON container given an iterator range
+    /// @sa https://json.nlohmann.me/api/basic_json/basic_json/
+    template < class InputIT, typename std::enable_if <
+                   std::is_same<InputIT, typename basic_json_t::iterator>::value ||
+                   std::is_same<InputIT, typename basic_json_t::const_iterator>::value, int >::type = 0 >
     basic_json(InputIT first, InputIT last)
     {
-        assert(first.m_object != nullptr);
-        assert(last.m_object != nullptr);
+        JSON_ASSERT(first.m_object != nullptr);
+        JSON_ASSERT(last.m_object != nullptr);
 
         // make sure iterator fits the current value
-        if (JSON_UNLIKELY(first.m_object != last.m_object))
+        if (JSON_HEDLEY_UNLIKELY(first.m_object != last.m_object))
         {
-            JSON_THROW(invalid_iterator::create(201, "iterators are not compatible"));
+            JSON_THROW(invalid_iterator::create(201, "iterators are not compatible", basic_json()));
         }
 
         // copy type from first iterator
@@ -12732,14 +18207,19 @@ class basic_json
             case value_t::number_unsigned:
             case value_t::string:
             {
-                if (JSON_UNLIKELY(not first.m_it.primitive_iterator.is_begin()
-                                  or not last.m_it.primitive_iterator.is_end()))
+                if (JSON_HEDLEY_UNLIKELY(!first.m_it.primitive_iterator.is_begin()
+                                         || !last.m_it.primitive_iterator.is_end()))
                 {
-                    JSON_THROW(invalid_iterator::create(204, "iterators out of range"));
+                    JSON_THROW(invalid_iterator::create(204, "iterators out of range", *first.m_object));
                 }
                 break;
             }
 
+            case value_t::null:
+            case value_t::object:
+            case value_t::array:
+            case value_t::binary:
+            case value_t::discarded:
             default:
                 break;
         }
@@ -12790,11 +18270,19 @@ class basic_json
                 break;
             }
 
+            case value_t::binary:
+            {
+                m_value = *first.m_object->m_value.binary;
+                break;
+            }
+
+            case value_t::null:
+            case value_t::discarded:
             default:
-                JSON_THROW(invalid_iterator::create(206, "cannot construct with iterators from " +
-                                                    std::string(first.m_object->type_name())));
+                JSON_THROW(invalid_iterator::create(206, "cannot construct with iterators from " + std::string(first.m_object->type_name()), *first.m_object));
         }
 
+        set_parents();
         assert_invariant();
     }
 
@@ -12803,36 +18291,13 @@ class basic_json
     // other constructors and destructor //
     ///////////////////////////////////////
 
-    /// @private
-    basic_json(const detail::json_ref<basic_json>& ref)
-        : basic_json(ref.moved_or_copied())
-    {}
+    template<typename JsonRef,
+             detail::enable_if_t<detail::conjunction<detail::is_json_ref<JsonRef>,
+                                 std::is_same<typename JsonRef::value_type, basic_json>>::value, int> = 0 >
+    basic_json(const JsonRef& ref) : basic_json(ref.moved_or_copied()) {}
 
-    /*!
-    @brief copy constructor
-
-    Creates a copy of a given JSON value.
-
-    @param[in] other  the JSON value to copy
-
-    @post `*this == other`
-
-    @complexity Linear in the size of @a other.
-
-    @exceptionsafety Strong guarantee: if an exception is thrown, there are no
-    changes to any JSON value.
-
-    @requirement This function helps `basic_json` satisfying the
-    [Container](https://en.cppreference.com/w/cpp/named_req/Container)
-    requirements:
-    - The complexity is linear.
-    - As postcondition, it holds: `other == basic_json(other)`.
-
-    @liveexample{The following code shows an example for the copy
-    constructor.,basic_json__basic_json}
-
-    @since version 1.0.0
-    */
+    /// @brief copy constructor
+    /// @sa https://json.nlohmann.me/api/basic_json/basic_json/
     basic_json(const basic_json& other)
         : m_type(other.m_type)
     {
@@ -12883,80 +18348,45 @@ class basic_json
                 break;
             }
 
+            case value_t::binary:
+            {
+                m_value = *other.m_value.binary;
+                break;
+            }
+
+            case value_t::null:
+            case value_t::discarded:
             default:
                 break;
         }
 
+        set_parents();
         assert_invariant();
     }
 
-    /*!
-    @brief move constructor
-
-    Move constructor. Constructs a JSON value with the contents of the given
-    value @a other using move semantics. It "steals" the resources from @a
-    other and leaves it as JSON null value.
-
-    @param[in,out] other  value to move to this object
-
-    @post `*this` has the same value as @a other before the call.
-    @post @a other is a JSON null value.
-
-    @complexity Constant.
-
-    @exceptionsafety No-throw guarantee: this constructor never throws
-    exceptions.
-
-    @requirement This function helps `basic_json` satisfying the
-    [MoveConstructible](https://en.cppreference.com/w/cpp/named_req/MoveConstructible)
-    requirements.
-
-    @liveexample{The code below shows the move constructor explicitly called
-    via std::move.,basic_json__moveconstructor}
-
-    @since version 1.0.0
-    */
+    /// @brief move constructor
+    /// @sa https://json.nlohmann.me/api/basic_json/basic_json/
     basic_json(basic_json&& other) noexcept
         : m_type(std::move(other.m_type)),
           m_value(std::move(other.m_value))
     {
         // check that passed value is valid
-        other.assert_invariant();
+        other.assert_invariant(false);
 
         // invalidate payload
         other.m_type = value_t::null;
         other.m_value = {};
 
+        set_parents();
         assert_invariant();
     }
 
-    /*!
-    @brief copy assignment
-
-    Copy assignment operator. Copies a JSON value via the "copy and swap"
-    strategy: It is expressed in terms of the copy constructor, destructor,
-    and the `swap()` member function.
-
-    @param[in] other  value to copy from
-
-    @complexity Linear.
-
-    @requirement This function helps `basic_json` satisfying the
-    [Container](https://en.cppreference.com/w/cpp/named_req/Container)
-    requirements:
-    - The complexity is linear.
-
-    @liveexample{The code below shows and example for the copy assignment. It
-    creates a copy of value `a` which is then swapped with `b`. Finally\, the
-    copy of `a` (which is the null value after the swap) is
-    destroyed.,basic_json__copyassignment}
-
-    @since version 1.0.0
-    */
-    reference& operator=(basic_json other) noexcept (
-        std::is_nothrow_move_constructible<value_t>::value and
-        std::is_nothrow_move_assignable<value_t>::value and
-        std::is_nothrow_move_constructible<json_value>::value and
+    /// @brief copy assignment
+    /// @sa https://json.nlohmann.me/api/basic_json/operator=/
+    basic_json& operator=(basic_json other) noexcept (
+        std::is_nothrow_move_constructible<value_t>::value&&
+        std::is_nothrow_move_assignable<value_t>::value&&
+        std::is_nothrow_move_constructible<json_value>::value&&
         std::is_nothrow_move_assignable<json_value>::value
     )
     {
@@ -12967,28 +18397,16 @@ class basic_json
         swap(m_type, other.m_type);
         swap(m_value, other.m_value);
 
+        set_parents();
         assert_invariant();
         return *this;
     }
 
-    /*!
-    @brief destructor
-
-    Destroys the JSON value and frees all allocated memory.
-
-    @complexity Linear.
-
-    @requirement This function helps `basic_json` satisfying the
-    [Container](https://en.cppreference.com/w/cpp/named_req/Container)
-    requirements:
-    - The complexity is linear.
-    - All stored elements are destroyed and all memory is freed.
-
-    @since version 1.0.0
-    */
+    /// @brief destructor
+    /// @sa https://json.nlohmann.me/api/basic_json/~basic_json/
     ~basic_json() noexcept
     {
-        assert_invariant();
+        assert_invariant(false);
         m_value.destroy(m_type);
     }
 
@@ -13003,47 +18421,15 @@ class basic_json
     /// Functions to inspect the type of a JSON value.
     /// @{
 
-    /*!
-    @brief serialization
-
-    Serialization function for JSON values. The function tries to mimic
-    Python's `json.dumps()` function, and currently supports its @a indent
-    and @a ensure_ascii parameters.
-
-    @param[in] indent If indent is nonnegative, then array elements and object
-    members will be pretty-printed with that indent level. An indent level of
-    `0` will only insert newlines. `-1` (the default) selects the most compact
-    representation.
-    @param[in] indent_char The character to use for indentation if @a indent is
-    greater than `0`. The default is ` ` (space).
-    @param[in] ensure_ascii If @a ensure_ascii is true, all non-ASCII characters
-    in the output are escaped with `\uXXXX` sequences, and the result consists
-    of ASCII characters only.
-
-    @return string containing the serialization of the JSON value
-
-    @throw type_error.316 if a string stored inside the JSON value is not
-                          UTF-8 encoded
-
-    @complexity Linear.
-
-    @exceptionsafety Strong guarantee: if an exception is thrown, there are no
-    changes in the JSON value.
-
-    @liveexample{The following example shows the effect of different @a indent\,
-    @a indent_char\, and @a ensure_ascii parameters to the result of the
-    serialization.,dump}
-
-    @see https://docs.python.org/2/library/json.html#json.dump
-
-    @since version 1.0.0; indentation character @a indent_char, option
-           @a ensure_ascii and exceptions added in version 3.0.0
-    */
-    string_t dump(const int indent = -1, const char indent_char = ' ',
-                  const bool ensure_ascii = false) const
+    /// @brief serialization
+    /// @sa https://json.nlohmann.me/api/basic_json/dump/
+    string_t dump(const int indent = -1,
+                  const char indent_char = ' ',
+                  const bool ensure_ascii = false,
+                  const error_handler_t error_handler = error_handler_t::strict) const
     {
         string_t result;
-        serializer s(detail::output_adapter<char, string_t>(result), indent_char);
+        serializer s(detail::output_adapter<char, string_t>(result), indent_char, error_handler);
 
         if (indent >= 0)
         {
@@ -13057,373 +18443,106 @@ class basic_json
         return result;
     }
 
-    /*!
-    @brief return the type of the JSON value (explicit)
-
-    Return the type of the JSON value as a value from the @ref value_t
-    enumeration.
-
-    @return the type of the JSON value
-            Value type                | return value
-            ------------------------- | -------------------------
-            null                      | value_t::null
-            boolean                   | value_t::boolean
-            string                    | value_t::string
-            number (integer)          | value_t::number_integer
-            number (unsigned integer) | value_t::number_unsigned
-            number (floating-point)   | value_t::number_float
-            object                    | value_t::object
-            array                     | value_t::array
-            discarded                 | value_t::discarded
-
-    @complexity Constant.
-
-    @exceptionsafety No-throw guarantee: this member function never throws
-    exceptions.
-
-    @liveexample{The following code exemplifies `type()` for all JSON
-    types.,type}
-
-    @sa @ref operator value_t() -- return the type of the JSON value (implicit)
-    @sa @ref type_name() -- return the type as string
-
-    @since version 1.0.0
-    */
+    /// @brief return the type of the JSON value (explicit)
+    /// @sa https://json.nlohmann.me/api/basic_json/type/
     constexpr value_t type() const noexcept
     {
         return m_type;
     }
 
-    /*!
-    @brief return whether type is primitive
-
-    This function returns true if and only if the JSON type is primitive
-    (string, number, boolean, or null).
-
-    @return `true` if type is primitive (string, number, boolean, or null),
-    `false` otherwise.
-
-    @complexity Constant.
-
-    @exceptionsafety No-throw guarantee: this member function never throws
-    exceptions.
-
-    @liveexample{The following code exemplifies `is_primitive()` for all JSON
-    types.,is_primitive}
-
-    @sa @ref is_structured() -- returns whether JSON value is structured
-    @sa @ref is_null() -- returns whether JSON value is `null`
-    @sa @ref is_string() -- returns whether JSON value is a string
-    @sa @ref is_boolean() -- returns whether JSON value is a boolean
-    @sa @ref is_number() -- returns whether JSON value is a number
-
-    @since version 1.0.0
-    */
+    /// @brief return whether type is primitive
+    /// @sa https://json.nlohmann.me/api/basic_json/is_primitive/
     constexpr bool is_primitive() const noexcept
     {
-        return is_null() or is_string() or is_boolean() or is_number();
+        return is_null() || is_string() || is_boolean() || is_number() || is_binary();
     }
 
-    /*!
-    @brief return whether type is structured
-
-    This function returns true if and only if the JSON type is structured
-    (array or object).
-
-    @return `true` if type is structured (array or object), `false` otherwise.
-
-    @complexity Constant.
-
-    @exceptionsafety No-throw guarantee: this member function never throws
-    exceptions.
-
-    @liveexample{The following code exemplifies `is_structured()` for all JSON
-    types.,is_structured}
-
-    @sa @ref is_primitive() -- returns whether value is primitive
-    @sa @ref is_array() -- returns whether value is an array
-    @sa @ref is_object() -- returns whether value is an object
-
-    @since version 1.0.0
-    */
+    /// @brief return whether type is structured
+    /// @sa https://json.nlohmann.me/api/basic_json/is_structured/
     constexpr bool is_structured() const noexcept
     {
-        return is_array() or is_object();
+        return is_array() || is_object();
     }
 
-    /*!
-    @brief return whether value is null
-
-    This function returns true if and only if the JSON value is null.
-
-    @return `true` if type is null, `false` otherwise.
-
-    @complexity Constant.
-
-    @exceptionsafety No-throw guarantee: this member function never throws
-    exceptions.
-
-    @liveexample{The following code exemplifies `is_null()` for all JSON
-    types.,is_null}
-
-    @since version 1.0.0
-    */
+    /// @brief return whether value is null
+    /// @sa https://json.nlohmann.me/api/basic_json/is_null/
     constexpr bool is_null() const noexcept
     {
-        return (m_type == value_t::null);
+        return m_type == value_t::null;
     }
 
-    /*!
-    @brief return whether value is a boolean
-
-    This function returns true if and only if the JSON value is a boolean.
-
-    @return `true` if type is boolean, `false` otherwise.
-
-    @complexity Constant.
-
-    @exceptionsafety No-throw guarantee: this member function never throws
-    exceptions.
-
-    @liveexample{The following code exemplifies `is_boolean()` for all JSON
-    types.,is_boolean}
-
-    @since version 1.0.0
-    */
+    /// @brief return whether value is a boolean
+    /// @sa https://json.nlohmann.me/api/basic_json/is_boolean/
     constexpr bool is_boolean() const noexcept
     {
-        return (m_type == value_t::boolean);
+        return m_type == value_t::boolean;
     }
 
-    /*!
-    @brief return whether value is a number
-
-    This function returns true if and only if the JSON value is a number. This
-    includes both integer (signed and unsigned) and floating-point values.
-
-    @return `true` if type is number (regardless whether integer, unsigned
-    integer or floating-type), `false` otherwise.
-
-    @complexity Constant.
-
-    @exceptionsafety No-throw guarantee: this member function never throws
-    exceptions.
-
-    @liveexample{The following code exemplifies `is_number()` for all JSON
-    types.,is_number}
-
-    @sa @ref is_number_integer() -- check if value is an integer or unsigned
-    integer number
-    @sa @ref is_number_unsigned() -- check if value is an unsigned integer
-    number
-    @sa @ref is_number_float() -- check if value is a floating-point number
-
-    @since version 1.0.0
-    */
+    /// @brief return whether value is a number
+    /// @sa https://json.nlohmann.me/api/basic_json/is_number/
     constexpr bool is_number() const noexcept
     {
-        return is_number_integer() or is_number_float();
+        return is_number_integer() || is_number_float();
     }
 
-    /*!
-    @brief return whether value is an integer number
-
-    This function returns true if and only if the JSON value is a signed or
-    unsigned integer number. This excludes floating-point values.
-
-    @return `true` if type is an integer or unsigned integer number, `false`
-    otherwise.
-
-    @complexity Constant.
-
-    @exceptionsafety No-throw guarantee: this member function never throws
-    exceptions.
-
-    @liveexample{The following code exemplifies `is_number_integer()` for all
-    JSON types.,is_number_integer}
-
-    @sa @ref is_number() -- check if value is a number
-    @sa @ref is_number_unsigned() -- check if value is an unsigned integer
-    number
-    @sa @ref is_number_float() -- check if value is a floating-point number
-
-    @since version 1.0.0
-    */
+    /// @brief return whether value is an integer number
+    /// @sa https://json.nlohmann.me/api/basic_json/is_number_integer/
     constexpr bool is_number_integer() const noexcept
     {
-        return (m_type == value_t::number_integer or m_type == value_t::number_unsigned);
+        return m_type == value_t::number_integer || m_type == value_t::number_unsigned;
     }
 
-    /*!
-    @brief return whether value is an unsigned integer number
-
-    This function returns true if and only if the JSON value is an unsigned
-    integer number. This excludes floating-point and signed integer values.
-
-    @return `true` if type is an unsigned integer number, `false` otherwise.
-
-    @complexity Constant.
-
-    @exceptionsafety No-throw guarantee: this member function never throws
-    exceptions.
-
-    @liveexample{The following code exemplifies `is_number_unsigned()` for all
-    JSON types.,is_number_unsigned}
-
-    @sa @ref is_number() -- check if value is a number
-    @sa @ref is_number_integer() -- check if value is an integer or unsigned
-    integer number
-    @sa @ref is_number_float() -- check if value is a floating-point number
-
-    @since version 2.0.0
-    */
+    /// @brief return whether value is an unsigned integer number
+    /// @sa https://json.nlohmann.me/api/basic_json/is_number_unsigned/
     constexpr bool is_number_unsigned() const noexcept
     {
-        return (m_type == value_t::number_unsigned);
+        return m_type == value_t::number_unsigned;
     }
 
-    /*!
-    @brief return whether value is a floating-point number
-
-    This function returns true if and only if the JSON value is a
-    floating-point number. This excludes signed and unsigned integer values.
-
-    @return `true` if type is a floating-point number, `false` otherwise.
-
-    @complexity Constant.
-
-    @exceptionsafety No-throw guarantee: this member function never throws
-    exceptions.
-
-    @liveexample{The following code exemplifies `is_number_float()` for all
-    JSON types.,is_number_float}
-
-    @sa @ref is_number() -- check if value is number
-    @sa @ref is_number_integer() -- check if value is an integer number
-    @sa @ref is_number_unsigned() -- check if value is an unsigned integer
-    number
-
-    @since version 1.0.0
-    */
+    /// @brief return whether value is a floating-point number
+    /// @sa https://json.nlohmann.me/api/basic_json/is_number_float/
     constexpr bool is_number_float() const noexcept
     {
-        return (m_type == value_t::number_float);
+        return m_type == value_t::number_float;
     }
 
-    /*!
-    @brief return whether value is an object
-
-    This function returns true if and only if the JSON value is an object.
-
-    @return `true` if type is object, `false` otherwise.
-
-    @complexity Constant.
-
-    @exceptionsafety No-throw guarantee: this member function never throws
-    exceptions.
-
-    @liveexample{The following code exemplifies `is_object()` for all JSON
-    types.,is_object}
-
-    @since version 1.0.0
-    */
+    /// @brief return whether value is an object
+    /// @sa https://json.nlohmann.me/api/basic_json/is_object/
     constexpr bool is_object() const noexcept
     {
-        return (m_type == value_t::object);
+        return m_type == value_t::object;
     }
 
-    /*!
-    @brief return whether value is an array
-
-    This function returns true if and only if the JSON value is an array.
-
-    @return `true` if type is array, `false` otherwise.
-
-    @complexity Constant.
-
-    @exceptionsafety No-throw guarantee: this member function never throws
-    exceptions.
-
-    @liveexample{The following code exemplifies `is_array()` for all JSON
-    types.,is_array}
-
-    @since version 1.0.0
-    */
+    /// @brief return whether value is an array
+    /// @sa https://json.nlohmann.me/api/basic_json/is_array/
     constexpr bool is_array() const noexcept
     {
-        return (m_type == value_t::array);
+        return m_type == value_t::array;
     }
 
-    /*!
-    @brief return whether value is a string
-
-    This function returns true if and only if the JSON value is a string.
-
-    @return `true` if type is string, `false` otherwise.
-
-    @complexity Constant.
-
-    @exceptionsafety No-throw guarantee: this member function never throws
-    exceptions.
-
-    @liveexample{The following code exemplifies `is_string()` for all JSON
-    types.,is_string}
-
-    @since version 1.0.0
-    */
+    /// @brief return whether value is a string
+    /// @sa https://json.nlohmann.me/api/basic_json/is_string/
     constexpr bool is_string() const noexcept
     {
-        return (m_type == value_t::string);
+        return m_type == value_t::string;
     }
 
-    /*!
-    @brief return whether value is discarded
+    /// @brief return whether value is a binary array
+    /// @sa https://json.nlohmann.me/api/basic_json/is_binary/
+    constexpr bool is_binary() const noexcept
+    {
+        return m_type == value_t::binary;
+    }
 
-    This function returns true if and only if the JSON value was discarded
-    during parsing with a callback function (see @ref parser_callback_t).
-
-    @note This function will always be `false` for JSON values after parsing.
-    That is, discarded values can only occur during parsing, but will be
-    removed when inside a structured value or replaced by null in other cases.
-
-    @return `true` if type is discarded, `false` otherwise.
-
-    @complexity Constant.
-
-    @exceptionsafety No-throw guarantee: this member function never throws
-    exceptions.
-
-    @liveexample{The following code exemplifies `is_discarded()` for all JSON
-    types.,is_discarded}
-
-    @since version 1.0.0
-    */
+    /// @brief return whether value is discarded
+    /// @sa https://json.nlohmann.me/api/basic_json/is_discarded/
     constexpr bool is_discarded() const noexcept
     {
-        return (m_type == value_t::discarded);
+        return m_type == value_t::discarded;
     }
 
-    /*!
-    @brief return the type of the JSON value (implicit)
-
-    Implicitly return the type of the JSON value as a value from the @ref
-    value_t enumeration.
-
-    @return the type of the JSON value
-
-    @complexity Constant.
-
-    @exceptionsafety No-throw guarantee: this member function never throws
-    exceptions.
-
-    @liveexample{The following code exemplifies the @ref value_t operator for
-    all JSON types.,operator__value_t}
-
-    @sa @ref type() -- return the type of the JSON value (explicit)
-    @sa @ref type_name() -- return the type as string
-
-    @since version 1.0.0
-    */
+    /// @brief return the type of the JSON value (implicit)
+    /// @sa https://json.nlohmann.me/api/basic_json/operator_value_t/
     constexpr operator value_t() const noexcept
     {
         return m_type;
@@ -13439,12 +18558,12 @@ class basic_json
     /// get a boolean (explicit)
     boolean_t get_impl(boolean_t* /*unused*/) const
     {
-        if (JSON_LIKELY(is_boolean()))
+        if (JSON_HEDLEY_LIKELY(is_boolean()))
         {
             return m_value.boolean;
         }
 
-        JSON_THROW(type_error::create(302, "type must be boolean, but is " + std::string(type_name())));
+        JSON_THROW(type_error::create(302, "type must be boolean, but is " + std::string(type_name()), *this));
     }
 
     /// get a pointer to the value (object)
@@ -13531,6 +18650,18 @@ class basic_json
         return is_number_float() ? &m_value.number_float : nullptr;
     }
 
+    /// get a pointer to the value (binary)
+    binary_t* get_impl_ptr(binary_t* /*unused*/) noexcept
+    {
+        return is_binary() ? m_value.binary : nullptr;
+    }
+
+    /// get a pointer to the value (binary)
+    constexpr const binary_t* get_impl_ptr(const binary_t* /*unused*/) const noexcept
+    {
+        return is_binary() ? m_value.binary : nullptr;
+    }
+
     /*!
     @brief helper function to implement get_ref()
 
@@ -13546,14 +18677,14 @@ class basic_json
     static ReferenceType get_ref_impl(ThisType& obj)
     {
         // delegate the call to get_ptr<>()
-        auto ptr = obj.template get_ptr<typename std::add_pointer<ReferenceType>::type>();
+        auto* ptr = obj.template get_ptr<typename std::add_pointer<ReferenceType>::type>();
 
-        if (JSON_LIKELY(ptr != nullptr))
+        if (JSON_HEDLEY_LIKELY(ptr != nullptr))
         {
             return *ptr;
         }
 
-        JSON_THROW(type_error::create(303, "incompatible ReferenceType for get_ref, actual type is " + std::string(obj.type_name())));
+        JSON_THROW(type_error::create(303, "incompatible ReferenceType for get_ref, actual type is " + std::string(obj.type_name()), obj));
     }
 
   public:
@@ -13561,51 +18692,28 @@ class basic_json
     /// Direct access to the stored value of a JSON value.
     /// @{
 
-    /*!
-    @brief get special-case overload
-
-    This overloads avoids a lot of template boilerplate, it can be seen as the
-    identity method
-
-    @tparam BasicJsonType == @ref basic_json
-
-    @return a copy of *this
-
-    @complexity Constant.
-
-    @since version 2.1.0
-    */
-    template<typename BasicJsonType, detail::enable_if_t<
-                 std::is_same<typename std::remove_const<BasicJsonType>::type, basic_json_t>::value,
-                 int> = 0>
-    basic_json get() const
+    /// @brief get a pointer value (implicit)
+    /// @sa https://json.nlohmann.me/api/basic_json/get_ptr/
+    template<typename PointerType, typename std::enable_if<
+                 std::is_pointer<PointerType>::value, int>::type = 0>
+    auto get_ptr() noexcept -> decltype(std::declval<basic_json_t&>().get_impl_ptr(std::declval<PointerType>()))
     {
-        return *this;
+        // delegate the call to get_impl_ptr<>()
+        return get_impl_ptr(static_cast<PointerType>(nullptr));
     }
 
-    /*!
-    @brief get special-case overload
-
-    This overloads converts the current @ref basic_json in a different
-    @ref basic_json type
-
-    @tparam BasicJsonType == @ref basic_json
-
-    @return a copy of *this, converted into @tparam BasicJsonType
-
-    @complexity Depending on the implementation of the called `from_json()`
-                method.
-
-    @since version 3.2.0
-    */
-    template<typename BasicJsonType, detail::enable_if_t<
-                 not std::is_same<BasicJsonType, basic_json>::value and
-                 detail::is_basic_json<BasicJsonType>::value, int> = 0>
-    BasicJsonType get() const
+    /// @brief get a pointer value (implicit)
+    /// @sa https://json.nlohmann.me/api/basic_json/get_ptr/
+    template < typename PointerType, typename std::enable_if <
+                   std::is_pointer<PointerType>::value&&
+                   std::is_const<typename std::remove_pointer<PointerType>::type>::value, int >::type = 0 >
+    constexpr auto get_ptr() const noexcept -> decltype(std::declval<const basic_json_t&>().get_impl_ptr(std::declval<PointerType>()))
     {
-        return *this;
+        // delegate the call to get_impl_ptr<>() const
+        return get_impl_ptr(static_cast<PointerType>(nullptr));
     }
 
+  private:
     /*!
     @brief get a value (explicit)
 
@@ -13629,7 +18737,6 @@ class basic_json
     - @ref json_serializer<ValueType> does not have a `from_json()` method of
       the form `ValueType from_json(const basic_json&)`
 
-    @tparam ValueTypeCV the provided value type
     @tparam ValueType the returned value type
 
     @return copy of the JSON value, converted to @a ValueType
@@ -13645,24 +18752,15 @@ class basic_json
 
     @since version 2.1.0
     */
-    template<typename ValueTypeCV, typename ValueType = detail::uncvref_t<ValueTypeCV>,
-             detail::enable_if_t <
-                 not detail::is_basic_json<ValueType>::value and
-                 detail::has_from_json<basic_json_t, ValueType>::value and
-                 not detail::has_non_default_from_json<basic_json_t, ValueType>::value,
-                 int> = 0>
-    ValueType get() const noexcept(noexcept(
-                                       JSONSerializer<ValueType>::from_json(std::declval<const basic_json_t&>(), std::declval<ValueType&>())))
+    template < typename ValueType,
+               detail::enable_if_t <
+                   detail::is_default_constructible<ValueType>::value&&
+                   detail::has_from_json<basic_json_t, ValueType>::value,
+                   int > = 0 >
+    ValueType get_impl(detail::priority_tag<0> /*unused*/) const noexcept(noexcept(
+                JSONSerializer<ValueType>::from_json(std::declval<const basic_json_t&>(), std::declval<ValueType&>())))
     {
-        // we cannot static_assert on ValueTypeCV being non-const, because
-        // there is support for get<const basic_json_t>(), which is why we
-        // still need the uncvref
-        static_assert(not std::is_reference<ValueTypeCV>::value,
-                      "get() cannot be used with reference types, you might want to use get_ref()");
-        static_assert(std::is_default_constructible<ValueType>::value,
-                      "types must be DefaultConstructible when used with get()");
-
-        ValueType ret;
+        auto ret = ValueType();
         JSONSerializer<ValueType>::from_json(*this, ret);
         return ret;
     }
@@ -13678,7 +18776,7 @@ class basic_json
 
     The function is equivalent to executing
     @code {.cpp}
-    return JSONSerializer<ValueTypeCV>::from_json(*this);
+    return JSONSerializer<ValueType>::from_json(*this);
     @endcode
 
     This overloads is chosen if:
@@ -13689,7 +18787,6 @@ class basic_json
     @note If @ref json_serializer<ValueType> has both overloads of
     `from_json()`, this one is chosen.
 
-    @tparam ValueTypeCV the provided value type
     @tparam ValueType the returned value type
 
     @return copy of the JSON value, converted to @a ValueType
@@ -13698,109 +18795,116 @@ class basic_json
 
     @since version 2.1.0
     */
-    template<typename ValueTypeCV, typename ValueType = detail::uncvref_t<ValueTypeCV>,
-             detail::enable_if_t<not std::is_same<basic_json_t, ValueType>::value and
-                                 detail::has_non_default_from_json<basic_json_t, ValueType>::value,
-                                 int> = 0>
-    ValueType get() const noexcept(noexcept(
-                                       JSONSerializer<ValueTypeCV>::from_json(std::declval<const basic_json_t&>())))
+    template < typename ValueType,
+               detail::enable_if_t <
+                   detail::has_non_default_from_json<basic_json_t, ValueType>::value,
+                   int > = 0 >
+    ValueType get_impl(detail::priority_tag<1> /*unused*/) const noexcept(noexcept(
+                JSONSerializer<ValueType>::from_json(std::declval<const basic_json_t&>())))
     {
-        static_assert(not std::is_reference<ValueTypeCV>::value,
-                      "get() cannot be used with reference types, you might want to use get_ref()");
-        return JSONSerializer<ValueTypeCV>::from_json(*this);
+        return JSONSerializer<ValueType>::from_json(*this);
     }
 
     /*!
-    @brief get a value (explicit)
+    @brief get special-case overload
 
-    Explicit type conversion between the JSON value and a compatible value.
-    The value is filled into the input parameter by calling the @ref json_serializer<ValueType>
-    `from_json()` method.
+    This overloads converts the current @ref basic_json in a different
+    @ref basic_json type
 
-    The function is equivalent to executing
-    @code {.cpp}
-    ValueType v;
-    JSONSerializer<ValueType>::from_json(*this, v);
-    @endcode
+    @tparam BasicJsonType == @ref basic_json
 
-    This overloads is chosen if:
-    - @a ValueType is not @ref basic_json,
-    - @ref json_serializer<ValueType> has a `from_json()` method of the form
-      `void from_json(const basic_json&, ValueType&)`, and
+    @return a copy of *this, converted into @a BasicJsonType
 
-    @tparam ValueType the input parameter type.
+    @complexity Depending on the implementation of the called `from_json()`
+                method.
 
-    @return the input parameter, allowing chaining calls.
-
-    @throw what @ref json_serializer<ValueType> `from_json()` method throws
-
-    @liveexample{The example below shows several conversions from JSON values
-    to other types. There a few things to note: (1) Floating-point numbers can
-    be converted to integers\, (2) A JSON array can be converted to a standard
-    `std::vector<short>`\, (3) A JSON object can be converted to C++
-    associative containers such as `std::unordered_map<std::string\,
-    json>`.,get_to}
-
-    @since version 3.3.0
+    @since version 3.2.0
     */
-    template<typename ValueType,
-             detail::enable_if_t <
-                 not detail::is_basic_json<ValueType>::value and
-                 detail::has_from_json<basic_json_t, ValueType>::value,
-                 int> = 0>
-    ValueType & get_to(ValueType& v) const noexcept(noexcept(
-                JSONSerializer<ValueType>::from_json(std::declval<const basic_json_t&>(), v)))
+    template < typename BasicJsonType,
+               detail::enable_if_t <
+                   detail::is_basic_json<BasicJsonType>::value,
+                   int > = 0 >
+    BasicJsonType get_impl(detail::priority_tag<2> /*unused*/) const
     {
-        JSONSerializer<ValueType>::from_json(*this, v);
-        return v;
+        return *this;
     }
 
-
     /*!
-    @brief get a pointer value (implicit)
+    @brief get special-case overload
 
-    Implicit pointer access to the internally stored JSON value. No copies are
-    made.
+    This overloads avoids a lot of template boilerplate, it can be seen as the
+    identity method
 
-    @warning Writing data to the pointee of the result yields an undefined
-    state.
+    @tparam BasicJsonType == @ref basic_json
 
-    @tparam PointerType pointer type; must be a pointer to @ref array_t, @ref
-    object_t, @ref string_t, @ref boolean_t, @ref number_integer_t,
-    @ref number_unsigned_t, or @ref number_float_t. Enforced by a static
-    assertion.
-
-    @return pointer to the internally stored JSON value if the requested
-    pointer type @a PointerType fits to the JSON value; `nullptr` otherwise
+    @return a copy of *this
 
     @complexity Constant.
 
-    @liveexample{The example below shows how pointers to internal values of a
-    JSON value can be requested. Note that no type conversions are made and a
-    `nullptr` is returned if the value and the requested pointer type does not
-    match.,get_ptr}
-
-    @since version 1.0.0
+    @since version 2.1.0
     */
-    template<typename PointerType, typename std::enable_if<
-                 std::is_pointer<PointerType>::value, int>::type = 0>
-    auto get_ptr() noexcept -> decltype(std::declval<basic_json_t&>().get_impl_ptr(std::declval<PointerType>()))
+    template<typename BasicJsonType,
+             detail::enable_if_t<
+                 std::is_same<BasicJsonType, basic_json_t>::value,
+                 int> = 0>
+    basic_json get_impl(detail::priority_tag<3> /*unused*/) const
     {
-        // delegate the call to get_impl_ptr<>()
-        return get_impl_ptr(static_cast<PointerType>(nullptr));
+        return *this;
     }
 
     /*!
-    @brief get a pointer value (implicit)
-    @copydoc get_ptr()
+    @brief get a pointer value (explicit)
+    @copydoc get()
     */
-    template<typename PointerType, typename std::enable_if<
-                 std::is_pointer<PointerType>::value and
-                 std::is_const<typename std::remove_pointer<PointerType>::type>::value, int>::type = 0>
-    constexpr auto get_ptr() const noexcept -> decltype(std::declval<const basic_json_t&>().get_impl_ptr(std::declval<PointerType>()))
+    template<typename PointerType,
+             detail::enable_if_t<
+                 std::is_pointer<PointerType>::value,
+                 int> = 0>
+    constexpr auto get_impl(detail::priority_tag<4> /*unused*/) const noexcept
+    -> decltype(std::declval<const basic_json_t&>().template get_ptr<PointerType>())
     {
-        // delegate the call to get_impl_ptr<>() const
-        return get_impl_ptr(static_cast<PointerType>(nullptr));
+        // delegate the call to get_ptr
+        return get_ptr<PointerType>();
+    }
+
+  public:
+    /*!
+    @brief get a (pointer) value (explicit)
+
+    Performs explicit type conversion between the JSON value and a compatible value if required.
+
+    - If the requested type is a pointer to the internally stored JSON value that pointer is returned.
+    No copies are made.
+
+    - If the requested type is the current @ref basic_json, or a different @ref basic_json convertible
+    from the current @ref basic_json.
+
+    - Otherwise the value is converted by calling the @ref json_serializer<ValueType> `from_json()`
+    method.
+
+    @tparam ValueTypeCV the provided value type
+    @tparam ValueType the returned value type
+
+    @return copy of the JSON value, converted to @tparam ValueType if necessary
+
+    @throw what @ref json_serializer<ValueType> `from_json()` method throws if conversion is required
+
+    @since version 2.1.0
+    */
+    template < typename ValueTypeCV, typename ValueType = detail::uncvref_t<ValueTypeCV>>
+#if defined(JSON_HAS_CPP_14)
+    constexpr
+#endif
+    auto get() const noexcept(
+    noexcept(std::declval<const basic_json_t&>().template get_impl<ValueType>(detail::priority_tag<4> {})))
+    -> decltype(std::declval<const basic_json_t&>().template get_impl<ValueType>(detail::priority_tag<4> {}))
+    {
+        // we cannot static_assert on ValueTypeCV being non-const, because
+        // there is support for get<const basic_json_t>(), which is why we
+        // still need the uncvref
+        static_assert(!std::is_reference<ValueTypeCV>::value,
+                      "get() cannot be used with reference types, you might want to use get_ref()");
+        return get_impl<ValueType>(detail::priority_tag<4> {});
     }
 
     /*!
@@ -13826,7 +18930,7 @@ class basic_json
     `nullptr` is returned if the value and the requested pointer type does not
     match.,get__PointerType}
 
-    @sa @ref get_ptr() for explicit pointer-member access
+    @sa see @ref get_ptr() for explicit pointer-member access
 
     @since version 1.0.0
     */
@@ -13838,44 +18942,47 @@ class basic_json
         return get_ptr<PointerType>();
     }
 
-    /*!
-    @brief get a pointer value (explicit)
-    @copydoc get()
-    */
-    template<typename PointerType, typename std::enable_if<
-                 std::is_pointer<PointerType>::value, int>::type = 0>
-    constexpr auto get() const noexcept -> decltype(std::declval<const basic_json_t&>().template get_ptr<PointerType>())
+    /// @brief get a value (explicit)
+    /// @sa https://json.nlohmann.me/api/basic_json/get_to/
+    template < typename ValueType,
+               detail::enable_if_t <
+                   !detail::is_basic_json<ValueType>::value&&
+                   detail::has_from_json<basic_json_t, ValueType>::value,
+                   int > = 0 >
+    ValueType & get_to(ValueType& v) const noexcept(noexcept(
+                JSONSerializer<ValueType>::from_json(std::declval<const basic_json_t&>(), v)))
     {
-        // delegate the call to get_ptr
-        return get_ptr<PointerType>();
+        JSONSerializer<ValueType>::from_json(*this, v);
+        return v;
     }
 
-    /*!
-    @brief get a reference value (implicit)
+    // specialization to allow calling get_to with a basic_json value
+    // see https://github.com/nlohmann/json/issues/2175
+    template<typename ValueType,
+             detail::enable_if_t <
+                 detail::is_basic_json<ValueType>::value,
+                 int> = 0>
+    ValueType & get_to(ValueType& v) const
+    {
+        v = *this;
+        return v;
+    }
 
-    Implicit reference access to the internally stored JSON value. No copies
-    are made.
+    template <
+        typename T, std::size_t N,
+        typename Array = T (&)[N], // NOLINT(cppcoreguidelines-avoid-c-arrays,hicpp-avoid-c-arrays,modernize-avoid-c-arrays)
+        detail::enable_if_t <
+            detail::has_from_json<basic_json_t, Array>::value, int > = 0 >
+    Array get_to(T (&v)[N]) const // NOLINT(cppcoreguidelines-avoid-c-arrays,hicpp-avoid-c-arrays,modernize-avoid-c-arrays)
+    noexcept(noexcept(JSONSerializer<Array>::from_json(
+                          std::declval<const basic_json_t&>(), v)))
+    {
+        JSONSerializer<Array>::from_json(*this, v);
+        return v;
+    }
 
-    @warning Writing data to the referee of the result yields an undefined
-    state.
-
-    @tparam ReferenceType reference type; must be a reference to @ref array_t,
-    @ref object_t, @ref string_t, @ref boolean_t, @ref number_integer_t, or
-    @ref number_float_t. Enforced by static assertion.
-
-    @return reference to the internally stored JSON value if the requested
-    reference type @a ReferenceType fits to the JSON value; throws
-    type_error.303 otherwise
-
-    @throw type_error.303 in case passed type @a ReferenceType is incompatible
-    with the stored JSON value; see example below
-
-    @complexity Constant.
-
-    @liveexample{The example shows several calls to `get_ref()`.,get_ref}
-
-    @since version 1.1.0
-    */
+    /// @brief get a reference value (implicit)
+    /// @sa https://json.nlohmann.me/api/basic_json/get_ref/
     template<typename ReferenceType, typename std::enable_if<
                  std::is_reference<ReferenceType>::value, int>::type = 0>
     ReferenceType get_ref()
@@ -13884,13 +18991,11 @@ class basic_json
         return get_ref_impl<ReferenceType>(*this);
     }
 
-    /*!
-    @brief get a reference value (implicit)
-    @copydoc get_ref()
-    */
-    template<typename ReferenceType, typename std::enable_if<
-                 std::is_reference<ReferenceType>::value and
-                 std::is_const<typename std::remove_reference<ReferenceType>::type>::value, int>::type = 0>
+    /// @brief get a reference value (implicit)
+    /// @sa https://json.nlohmann.me/api/basic_json/get_ref/
+    template < typename ReferenceType, typename std::enable_if <
+                   std::is_reference<ReferenceType>::value&&
+                   std::is_const<typename std::remove_reference<ReferenceType>::type>::value, int >::type = 0 >
     ReferenceType get_ref() const
     {
         // delegate call to get_ref_impl
@@ -13927,25 +19032,48 @@ class basic_json
     @since version 1.0.0
     */
     template < typename ValueType, typename std::enable_if <
-                   not std::is_pointer<ValueType>::value and
-                   not std::is_same<ValueType, detail::json_ref<basic_json>>::value and
-                   not std::is_same<ValueType, typename string_t::value_type>::value and
-                   not detail::is_basic_json<ValueType>::value
+                   detail::conjunction <
+                       detail::negation<std::is_pointer<ValueType>>,
+                       detail::negation<std::is_same<ValueType, detail::json_ref<basic_json>>>,
+                                        detail::negation<std::is_same<ValueType, typename string_t::value_type>>,
+                                        detail::negation<detail::is_basic_json<ValueType>>,
+                                        detail::negation<std::is_same<ValueType, std::initializer_list<typename string_t::value_type>>>,
 
-#ifndef _MSC_VER  // fix for issue #167 operator<< ambiguity under VS2015
-                   and not std::is_same<ValueType, std::initializer_list<typename string_t::value_type>>::value
-#if defined(JSON_HAS_CPP_17) && defined(_MSC_VER) and _MSC_VER <= 1914
-                   and not std::is_same<ValueType, typename std::string_view>::value
+#if defined(JSON_HAS_CPP_17) && (defined(__GNUC__) || (defined(_MSC_VER) && _MSC_VER >= 1910 && _MSC_VER <= 1914))
+                                                detail::negation<std::is_same<ValueType, std::string_view>>,
 #endif
-#endif
-                   and detail::is_detected<detail::get_template_function, const basic_json_t&, ValueType>::value
-                   , int >::type = 0 >
-    operator ValueType() const
+                                                detail::is_detected_lazy<detail::get_template_function, const basic_json_t&, ValueType>
+                                                >::value, int >::type = 0 >
+                                        JSON_EXPLICIT operator ValueType() const
     {
         // delegate the call to get<>() const
         return get<ValueType>();
     }
 
+    /// @brief get a binary value
+    /// @sa https://json.nlohmann.me/api/basic_json/get_binary/
+    binary_t& get_binary()
+    {
+        if (!is_binary())
+        {
+            JSON_THROW(type_error::create(302, "type must be binary, but is " + std::string(type_name()), *this));
+        }
+
+        return *get_ptr<binary_t*>();
+    }
+
+    /// @brief get a binary value
+    /// @sa https://json.nlohmann.me/api/basic_json/get_binary/
+    const binary_t& get_binary() const
+    {
+        if (!is_binary())
+        {
+            JSON_THROW(type_error::create(302, "type must be binary, but is " + std::string(type_name()), *this));
+        }
+
+        return *get_ptr<const binary_t*>();
+    }
+
     /// @}
 
 
@@ -13957,83 +19085,35 @@ class basic_json
     /// Access to the JSON value.
     /// @{
 
-    /*!
-    @brief access specified array element with bounds checking
-
-    Returns a reference to the element at specified location @a idx, with
-    bounds checking.
-
-    @param[in] idx  index of the element to access
-
-    @return reference to the element at index @a idx
-
-    @throw type_error.304 if the JSON value is not an array; in this case,
-    calling `at` with an index makes no sense. See example below.
-    @throw out_of_range.401 if the index @a idx is out of range of the array;
-    that is, `idx >= size()`. See example below.
-
-    @exceptionsafety Strong guarantee: if an exception is thrown, there are no
-    changes in the JSON value.
-
-    @complexity Constant.
-
-    @since version 1.0.0
-
-    @liveexample{The example below shows how array elements can be read and
-    written using `at()`. It also demonstrates the different exceptions that
-    can be thrown.,at__size_type}
-    */
+    /// @brief access specified array element with bounds checking
+    /// @sa https://json.nlohmann.me/api/basic_json/at/
     reference at(size_type idx)
     {
         // at only works for arrays
-        if (JSON_LIKELY(is_array()))
+        if (JSON_HEDLEY_LIKELY(is_array()))
         {
             JSON_TRY
             {
-                return m_value.array->at(idx);
+                return set_parent(m_value.array->at(idx));
             }
             JSON_CATCH (std::out_of_range&)
             {
                 // create better exception explanation
-                JSON_THROW(out_of_range::create(401, "array index " + std::to_string(idx) + " is out of range"));
+                JSON_THROW(out_of_range::create(401, "array index " + std::to_string(idx) + " is out of range", *this));
             }
         }
         else
         {
-            JSON_THROW(type_error::create(304, "cannot use at() with " + std::string(type_name())));
+            JSON_THROW(type_error::create(304, "cannot use at() with " + std::string(type_name()), *this));
         }
     }
 
-    /*!
-    @brief access specified array element with bounds checking
-
-    Returns a const reference to the element at specified location @a idx,
-    with bounds checking.
-
-    @param[in] idx  index of the element to access
-
-    @return const reference to the element at index @a idx
-
-    @throw type_error.304 if the JSON value is not an array; in this case,
-    calling `at` with an index makes no sense. See example below.
-    @throw out_of_range.401 if the index @a idx is out of range of the array;
-    that is, `idx >= size()`. See example below.
-
-    @exceptionsafety Strong guarantee: if an exception is thrown, there are no
-    changes in the JSON value.
-
-    @complexity Constant.
-
-    @since version 1.0.0
-
-    @liveexample{The example below shows how array elements can be read using
-    `at()`. It also demonstrates the different exceptions that can be thrown.,
-    at__size_type_const}
-    */
+    /// @brief access specified array element with bounds checking
+    /// @sa https://json.nlohmann.me/api/basic_json/at/
     const_reference at(size_type idx) const
     {
         // at only works for arrays
-        if (JSON_LIKELY(is_array()))
+        if (JSON_HEDLEY_LIKELY(is_array()))
         {
             JSON_TRY
             {
@@ -14042,100 +19122,44 @@ class basic_json
             JSON_CATCH (std::out_of_range&)
             {
                 // create better exception explanation
-                JSON_THROW(out_of_range::create(401, "array index " + std::to_string(idx) + " is out of range"));
+                JSON_THROW(out_of_range::create(401, "array index " + std::to_string(idx) + " is out of range", *this));
             }
         }
         else
         {
-            JSON_THROW(type_error::create(304, "cannot use at() with " + std::string(type_name())));
+            JSON_THROW(type_error::create(304, "cannot use at() with " + std::string(type_name()), *this));
         }
     }
 
-    /*!
-    @brief access specified object element with bounds checking
-
-    Returns a reference to the element at with specified key @a key, with
-    bounds checking.
-
-    @param[in] key  key of the element to access
-
-    @return reference to the element at key @a key
-
-    @throw type_error.304 if the JSON value is not an object; in this case,
-    calling `at` with a key makes no sense. See example below.
-    @throw out_of_range.403 if the key @a key is is not stored in the object;
-    that is, `find(key) == end()`. See example below.
-
-    @exceptionsafety Strong guarantee: if an exception is thrown, there are no
-    changes in the JSON value.
-
-    @complexity Logarithmic in the size of the container.
-
-    @sa @ref operator[](const typename object_t::key_type&) for unchecked
-    access by reference
-    @sa @ref value() for access by value with a default value
-
-    @since version 1.0.0
-
-    @liveexample{The example below shows how object elements can be read and
-    written using `at()`. It also demonstrates the different exceptions that
-    can be thrown.,at__object_t_key_type}
-    */
+    /// @brief access specified object element with bounds checking
+    /// @sa https://json.nlohmann.me/api/basic_json/at/
     reference at(const typename object_t::key_type& key)
     {
         // at only works for objects
-        if (JSON_LIKELY(is_object()))
+        if (JSON_HEDLEY_LIKELY(is_object()))
         {
             JSON_TRY
             {
-                return m_value.object->at(key);
+                return set_parent(m_value.object->at(key));
             }
             JSON_CATCH (std::out_of_range&)
             {
                 // create better exception explanation
-                JSON_THROW(out_of_range::create(403, "key '" + key + "' not found"));
+                JSON_THROW(out_of_range::create(403, "key '" + key + "' not found", *this));
             }
         }
         else
         {
-            JSON_THROW(type_error::create(304, "cannot use at() with " + std::string(type_name())));
+            JSON_THROW(type_error::create(304, "cannot use at() with " + std::string(type_name()), *this));
         }
     }
 
-    /*!
-    @brief access specified object element with bounds checking
-
-    Returns a const reference to the element at with specified key @a key,
-    with bounds checking.
-
-    @param[in] key  key of the element to access
-
-    @return const reference to the element at key @a key
-
-    @throw type_error.304 if the JSON value is not an object; in this case,
-    calling `at` with a key makes no sense. See example below.
-    @throw out_of_range.403 if the key @a key is is not stored in the object;
-    that is, `find(key) == end()`. See example below.
-
-    @exceptionsafety Strong guarantee: if an exception is thrown, there are no
-    changes in the JSON value.
-
-    @complexity Logarithmic in the size of the container.
-
-    @sa @ref operator[](const typename object_t::key_type&) for unchecked
-    access by reference
-    @sa @ref value() for access by value with a default value
-
-    @since version 1.0.0
-
-    @liveexample{The example below shows how object elements can be read using
-    `at()`. It also demonstrates the different exceptions that can be thrown.,
-    at__object_t_key_type_const}
-    */
+    /// @brief access specified object element with bounds checking
+    /// @sa https://json.nlohmann.me/api/basic_json/at/
     const_reference at(const typename object_t::key_type& key) const
     {
         // at only works for objects
-        if (JSON_LIKELY(is_object()))
+        if (JSON_HEDLEY_LIKELY(is_object()))
         {
             JSON_TRY
             {
@@ -14144,40 +19168,17 @@ class basic_json
             JSON_CATCH (std::out_of_range&)
             {
                 // create better exception explanation
-                JSON_THROW(out_of_range::create(403, "key '" + key + "' not found"));
+                JSON_THROW(out_of_range::create(403, "key '" + key + "' not found", *this));
             }
         }
         else
         {
-            JSON_THROW(type_error::create(304, "cannot use at() with " + std::string(type_name())));
+            JSON_THROW(type_error::create(304, "cannot use at() with " + std::string(type_name()), *this));
         }
     }
 
-    /*!
-    @brief access specified array element
-
-    Returns a reference to the element at specified location @a idx.
-
-    @note If @a idx is beyond the range of the array (i.e., `idx >= size()`),
-    then the array is silently filled up with `null` values to make `idx` a
-    valid reference to the last stored element.
-
-    @param[in] idx  index of the element to access
-
-    @return reference to the element at index @a idx
-
-    @throw type_error.305 if the JSON value is not an array or null; in that
-    cases, using the [] operator with an index makes no sense.
-
-    @complexity Constant if @a idx is in the range of the array. Otherwise
-    linear in `idx - size()`.
-
-    @liveexample{The example below shows how array elements can be read and
-    written using `[]` operator. Note the addition of `null`
-    values.,operatorarray__size_type}
-
-    @since version 1.0.0
-    */
+    /// @brief access specified array element
+    /// @sa https://json.nlohmann.me/api/basic_json/operator%5B%5D/
     reference operator[](size_type idx)
     {
         // implicitly convert null value to an empty array
@@ -14189,79 +19190,54 @@ class basic_json
         }
 
         // operator[] only works for arrays
-        if (JSON_LIKELY(is_array()))
+        if (JSON_HEDLEY_LIKELY(is_array()))
         {
             // fill up array with null values if given idx is outside range
             if (idx >= m_value.array->size())
             {
-                m_value.array->insert(m_value.array->end(),
-                                      idx - m_value.array->size() + 1,
-                                      basic_json());
+#if JSON_DIAGNOSTICS
+                // remember array size & capacity before resizing
+                const auto old_size = m_value.array->size();
+                const auto old_capacity = m_value.array->capacity();
+#endif
+                m_value.array->resize(idx + 1);
+
+#if JSON_DIAGNOSTICS
+                if (JSON_HEDLEY_UNLIKELY(m_value.array->capacity() != old_capacity))
+                {
+                    // capacity has changed: update all parents
+                    set_parents();
+                }
+                else
+                {
+                    // set parent for values added above
+                    set_parents(begin() + static_cast<typename iterator::difference_type>(old_size), static_cast<typename iterator::difference_type>(idx + 1 - old_size));
+                }
+#endif
+                assert_invariant();
             }
 
             return m_value.array->operator[](idx);
         }
 
-        JSON_THROW(type_error::create(305, "cannot use operator[] with a numeric argument with " + std::string(type_name())));
+        JSON_THROW(type_error::create(305, "cannot use operator[] with a numeric argument with " + std::string(type_name()), *this));
     }
 
-    /*!
-    @brief access specified array element
-
-    Returns a const reference to the element at specified location @a idx.
-
-    @param[in] idx  index of the element to access
-
-    @return const reference to the element at index @a idx
-
-    @throw type_error.305 if the JSON value is not an array; in that case,
-    using the [] operator with an index makes no sense.
-
-    @complexity Constant.
-
-    @liveexample{The example below shows how array elements can be read using
-    the `[]` operator.,operatorarray__size_type_const}
-
-    @since version 1.0.0
-    */
+    /// @brief access specified array element
+    /// @sa https://json.nlohmann.me/api/basic_json/operator%5B%5D/
     const_reference operator[](size_type idx) const
     {
         // const operator[] only works for arrays
-        if (JSON_LIKELY(is_array()))
+        if (JSON_HEDLEY_LIKELY(is_array()))
         {
             return m_value.array->operator[](idx);
         }
 
-        JSON_THROW(type_error::create(305, "cannot use operator[] with a numeric argument with " + std::string(type_name())));
+        JSON_THROW(type_error::create(305, "cannot use operator[] with a numeric argument with " + std::string(type_name()), *this));
     }
 
-    /*!
-    @brief access specified object element
-
-    Returns a reference to the element at with specified key @a key.
-
-    @note If @a key is not found in the object, then it is silently added to
-    the object and filled with a `null` value to make `key` a valid reference.
-    In case the value was `null` before, it is converted to an object.
-
-    @param[in] key  key of the element to access
-
-    @return reference to the element at key @a key
-
-    @throw type_error.305 if the JSON value is not an object or null; in that
-    cases, using the [] operator with a key makes no sense.
-
-    @complexity Logarithmic in the size of the container.
-
-    @liveexample{The example below shows how object elements can be read and
-    written using the `[]` operator.,operatorarray__key_type}
-
-    @sa @ref at(const typename object_t::key_type&) for access by reference
-    with range checking
-    @sa @ref value() for access by value with a default value
-
-    @since version 1.0.0
-    */
+    /// @brief access specified object element
+    /// @sa https://json.nlohmann.me/api/basic_json/operator%5B%5D/
     reference operator[](const typename object_t::key_type& key)
     {
         // implicitly convert null value to an empty object
@@ -14273,84 +19249,32 @@ class basic_json
         }
 
         // operator[] only works for objects
-        if (JSON_LIKELY(is_object()))
+        if (JSON_HEDLEY_LIKELY(is_object()))
         {
-            return m_value.object->operator[](key);
+            return set_parent(m_value.object->operator[](key));
         }
 
-        JSON_THROW(type_error::create(305, "cannot use operator[] with a string argument with " + std::string(type_name())));
+        JSON_THROW(type_error::create(305, "cannot use operator[] with a string argument with " + std::string(type_name()), *this));
     }
 
-    /*!
-    @brief read-only access specified object element
-
-    Returns a const reference to the element at with specified key @a key. No
-    bounds checking is performed.
-
-    @warning If the element with key @a key does not exist, the behavior is
-    undefined.
-
-    @param[in] key  key of the element to access
-
-    @return const reference to the element at key @a key
-
-    @pre The element with key @a key must exist. **This precondition is
-         enforced with an assertion.**
-
-    @throw type_error.305 if the JSON value is not an object; in that case,
-    using the [] operator with a key makes no sense.
-
-    @complexity Logarithmic in the size of the container.
-
-    @liveexample{The example below shows how object elements can be read using
-    the `[]` operator.,operatorarray__key_type_const}
-
-    @sa @ref at(const typename object_t::key_type&) for access by reference
-    with range checking
-    @sa @ref value() for access by value with a default value
-
-    @since version 1.0.0
-    */
+    /// @brief access specified object element
+    /// @sa https://json.nlohmann.me/api/basic_json/operator%5B%5D/
     const_reference operator[](const typename object_t::key_type& key) const
     {
         // const operator[] only works for objects
-        if (JSON_LIKELY(is_object()))
+        if (JSON_HEDLEY_LIKELY(is_object()))
         {
-            assert(m_value.object->find(key) != m_value.object->end());
+            JSON_ASSERT(m_value.object->find(key) != m_value.object->end());
             return m_value.object->find(key)->second;
         }
 
-        JSON_THROW(type_error::create(305, "cannot use operator[] with a string argument with " + std::string(type_name())));
+        JSON_THROW(type_error::create(305, "cannot use operator[] with a string argument with " + std::string(type_name()), *this));
     }
 
-    /*!
-    @brief access specified object element
-
-    Returns a reference to the element at with specified key @a key.
-
-    @note If @a key is not found in the object, then it is silently added to
-    the object and filled with a `null` value to make `key` a valid reference.
-    In case the value was `null` before, it is converted to an object.
-
-    @param[in] key  key of the element to access
-
-    @return reference to the element at key @a key
-
-    @throw type_error.305 if the JSON value is not an object or null; in that
-    cases, using the [] operator with a key makes no sense.
-
-    @complexity Logarithmic in the size of the container.
-
-    @liveexample{The example below shows how object elements can be read and
-    written using the `[]` operator.,operatorarray__key_type}
-
-    @sa @ref at(const typename object_t::key_type&) for access by reference
-    with range checking
-    @sa @ref value() for access by value with a default value
-
-    @since version 1.1.0
-    */
+    /// @brief access specified object element
+    /// @sa https://json.nlohmann.me/api/basic_json/operator%5B%5D/
     template<typename T>
+    JSON_HEDLEY_NON_NULL(2)
     reference operator[](T* key)
     {
         // implicitly convert null to object
@@ -14362,186 +19286,75 @@ class basic_json
         }
 
         // at only works for objects
-        if (JSON_LIKELY(is_object()))
+        if (JSON_HEDLEY_LIKELY(is_object()))
         {
-            return m_value.object->operator[](key);
+            return set_parent(m_value.object->operator[](key));
         }
 
-        JSON_THROW(type_error::create(305, "cannot use operator[] with a string argument with " + std::string(type_name())));
+        JSON_THROW(type_error::create(305, "cannot use operator[] with a string argument with " + std::string(type_name()), *this));
     }
 
-    /*!
-    @brief read-only access specified object element
-
-    Returns a const reference to the element at with specified key @a key. No
-    bounds checking is performed.
-
-    @warning If the element with key @a key does not exist, the behavior is
-    undefined.
-
-    @param[in] key  key of the element to access
-
-    @return const reference to the element at key @a key
-
-    @pre The element with key @a key must exist. **This precondition is
-         enforced with an assertion.**
-
-    @throw type_error.305 if the JSON value is not an object; in that case,
-    using the [] operator with a key makes no sense.
-
-    @complexity Logarithmic in the size of the container.
-
-    @liveexample{The example below shows how object elements can be read using
-    the `[]` operator.,operatorarray__key_type_const}
-
-    @sa @ref at(const typename object_t::key_type&) for access by reference
-    with range checking
-    @sa @ref value() for access by value with a default value
-
-    @since version 1.1.0
-    */
+    /// @brief access specified object element
+    /// @sa https://json.nlohmann.me/api/basic_json/operator%5B%5D/
     template<typename T>
+    JSON_HEDLEY_NON_NULL(2)
     const_reference operator[](T* key) const
     {
         // at only works for objects
-        if (JSON_LIKELY(is_object()))
+        if (JSON_HEDLEY_LIKELY(is_object()))
         {
-            assert(m_value.object->find(key) != m_value.object->end());
+            JSON_ASSERT(m_value.object->find(key) != m_value.object->end());
             return m_value.object->find(key)->second;
         }
 
-        JSON_THROW(type_error::create(305, "cannot use operator[] with a string argument with " + std::string(type_name())));
+        JSON_THROW(type_error::create(305, "cannot use operator[] with a string argument with " + std::string(type_name()), *this));
     }
 
-    /*!
-    @brief access specified object element with default value
-
-    Returns either a copy of an object's element at the specified key @a key
-    or a given default value if no element with key @a key exists.
-
-    The function is basically equivalent to executing
-    @code {.cpp}
-    try {
-        return at(key);
-    } catch(out_of_range) {
-        return default_value;
-    }
-    @endcode
-
-    @note Unlike @ref at(const typename object_t::key_type&), this function
-    does not throw if the given key @a key was not found.
-
-    @note Unlike @ref operator[](const typename object_t::key_type& key), this
-    function does not implicitly add an element to the position defined by @a
-    key. This function is furthermore also applicable to const objects.
-
-    @param[in] key  key of the element to access
-    @param[in] default_value  the value to return if @a key is not found
-
-    @tparam ValueType type compatible to JSON values, for instance `int` for
-    JSON integer numbers, `bool` for JSON booleans, or `std::vector` types for
-    JSON arrays. Note the type of the expected value at @a key and the default
-    value @a default_value must be compatible.
-
-    @return copy of the element at key @a key or @a default_value if @a key
-    is not found
-
-    @throw type_error.306 if the JSON value is not an object; in that case,
-    using `value()` with a key makes no sense.
-
-    @complexity Logarithmic in the size of the container.
-
-    @liveexample{The example below shows how object elements can be queried
-    with a default value.,basic_json__value}
-
-    @sa @ref at(const typename object_t::key_type&) for access by reference
-    with range checking
-    @sa @ref operator[](const typename object_t::key_type&) for unchecked
-    access by reference
-
-    @since version 1.0.0
-    */
-    template<class ValueType, typename std::enable_if<
-                 std::is_convertible<basic_json_t, ValueType>::value, int>::type = 0>
+    /// @brief access specified object element with default value
+    /// @sa https://json.nlohmann.me/api/basic_json/value/
+    /// using std::is_convertible in a std::enable_if will fail when using explicit conversions
+    template < class ValueType, typename std::enable_if <
+                   detail::is_getable<basic_json_t, ValueType>::value
+                   && !std::is_same<value_t, ValueType>::value, int >::type = 0 >
     ValueType value(const typename object_t::key_type& key, const ValueType& default_value) const
     {
         // at only works for objects
-        if (JSON_LIKELY(is_object()))
+        if (JSON_HEDLEY_LIKELY(is_object()))
         {
             // if key is found, return value and given default value otherwise
             const auto it = find(key);
             if (it != end())
             {
-                return *it;
+                return it->template get<ValueType>();
             }
 
             return default_value;
         }
 
-        JSON_THROW(type_error::create(306, "cannot use value() with " + std::string(type_name())));
+        JSON_THROW(type_error::create(306, "cannot use value() with " + std::string(type_name()), *this));
     }
 
-    /*!
-    @brief overload for a default value of type const char*
-    @copydoc basic_json::value(const typename object_t::key_type&, ValueType) const
-    */
+    /// @brief access specified object element with default value
+    /// @sa https://json.nlohmann.me/api/basic_json/value/
+    /// overload for a default value of type const char*
     string_t value(const typename object_t::key_type& key, const char* default_value) const
     {
         return value(key, string_t(default_value));
     }
 
-    /*!
-    @brief access specified object element via JSON Pointer with default value
-
-    Returns either a copy of an object's element at the specified key @a key
-    or a given default value if no element with key @a key exists.
-
-    The function is basically equivalent to executing
-    @code {.cpp}
-    try {
-        return at(ptr);
-    } catch(out_of_range) {
-        return default_value;
-    }
-    @endcode
-
-    @note Unlike @ref at(const json_pointer&), this function does not throw
-    if the given key @a key was not found.
-
-    @param[in] ptr  a JSON pointer to the element to access
-    @param[in] default_value  the value to return if @a ptr found no value
-
-    @tparam ValueType type compatible to JSON values, for instance `int` for
-    JSON integer numbers, `bool` for JSON booleans, or `std::vector` types for
-    JSON arrays. Note the type of the expected value at @a key and the default
-    value @a default_value must be compatible.
-
-    @return copy of the element at key @a key or @a default_value if @a key
-    is not found
-
-    @throw type_error.306 if the JSON value is not an object; in that case,
-    using `value()` with a key makes no sense.
-
-    @complexity Logarithmic in the size of the container.
-
-    @liveexample{The example below shows how object elements can be queried
-    with a default value.,basic_json__value_ptr}
-
-    @sa @ref operator[](const json_pointer&) for unchecked access by reference
-
-    @since version 2.0.2
-    */
+    /// @brief access specified object element via JSON Pointer with default value
+    /// @sa https://json.nlohmann.me/api/basic_json/value/
     template<class ValueType, typename std::enable_if<
-                 std::is_convertible<basic_json_t, ValueType>::value, int>::type = 0>
+                 detail::is_getable<basic_json_t, ValueType>::value, int>::type = 0>
     ValueType value(const json_pointer& ptr, const ValueType& default_value) const
     {
         // at only works for objects
-        if (JSON_LIKELY(is_object()))
+        if (JSON_HEDLEY_LIKELY(is_object()))
         {
             // if pointer resolves a value, return it or use default value
             JSON_TRY
             {
-                return ptr.get_checked(this);
+                return ptr.get_checked(this).template get<ValueType>();
             }
             JSON_INTERNAL_CATCH (out_of_range&)
             {
@@ -14549,87 +19362,34 @@ class basic_json
             }
         }
 
-        JSON_THROW(type_error::create(306, "cannot use value() with " + std::string(type_name())));
+        JSON_THROW(type_error::create(306, "cannot use value() with " + std::string(type_name()), *this));
     }
 
-    /*!
-    @brief overload for a default value of type const char*
-    @copydoc basic_json::value(const json_pointer&, ValueType) const
-    */
+    /// @brief access specified object element via JSON Pointer with default value
+    /// @sa https://json.nlohmann.me/api/basic_json/value/
+    /// overload for a default value of type const char*
+    JSON_HEDLEY_NON_NULL(3)
     string_t value(const json_pointer& ptr, const char* default_value) const
     {
         return value(ptr, string_t(default_value));
     }
 
-    /*!
-    @brief access the first element
-
-    Returns a reference to the first element in the container. For a JSON
-    container `c`, the expression `c.front()` is equivalent to `*c.begin()`.
-
-    @return In case of a structured type (array or object), a reference to the
-    first element is returned. In case of number, string, or boolean values, a
-    reference to the value is returned.
-
-    @complexity Constant.
-
-    @pre The JSON value must not be `null` (would throw `std::out_of_range`)
-    or an empty array or object (undefined behavior, **guarded by
-    assertions**).
-    @post The JSON value remains unchanged.
-
-    @throw invalid_iterator.214 when called on `null` value
-
-    @liveexample{The following code shows an example for `front()`.,front}
-
-    @sa @ref back() -- access the last element
-
-    @since version 1.0.0
-    */
+    /// @brief access the first element
+    /// @sa https://json.nlohmann.me/api/basic_json/front/
     reference front()
     {
         return *begin();
     }
 
-    /*!
-    @copydoc basic_json::front()
-    */
+    /// @brief access the first element
+    /// @sa https://json.nlohmann.me/api/basic_json/front/
     const_reference front() const
     {
         return *cbegin();
     }
 
-    /*!
-    @brief access the last element
-
-    Returns a reference to the last element in the container. For a JSON
-    container `c`, the expression `c.back()` is equivalent to
-    @code {.cpp}
-    auto tmp = c.end();
-    --tmp;
-    return *tmp;
-    @endcode
-
-    @return In case of a structured type (array or object), a reference to the
-    last element is returned. In case of number, string, or boolean values, a
-    reference to the value is returned.
-
-    @complexity Constant.
-
-    @pre The JSON value must not be `null` (would throw `std::out_of_range`)
-    or an empty array or object (undefined behavior, **guarded by
-    assertions**).
-    @post The JSON value remains unchanged.
-
-    @throw invalid_iterator.214 when called on a `null` value. See example
-    below.
-
-    @liveexample{The following code shows an example for `back()`.,back}
-
-    @sa @ref front() -- access the first element
-
-    @since version 1.0.0
-    */
+    /// @brief access the last element
+    /// @sa https://json.nlohmann.me/api/basic_json/back/
     reference back()
     {
         auto tmp = end();
@@ -14637,9 +19397,8 @@ class basic_json
         return *tmp;
     }
 
-    /*!
-    @copydoc basic_json::back()
-    */
+    /// @brief access the last element
+    /// @sa https://json.nlohmann.me/api/basic_json/back/
     const_reference back() const
     {
         auto tmp = cend();
@@ -14647,62 +19406,18 @@ class basic_json
         return *tmp;
     }
 
-    /*!
-    @brief remove element given an iterator
-
-    Removes the element specified by iterator @a pos. The iterator @a pos must
-    be valid and dereferenceable. Thus the `end()` iterator (which is valid,
-    but is not dereferenceable) cannot be used as a value for @a pos.
-
-    If called on a primitive type other than `null`, the resulting JSON value
-    will be `null`.
-
-    @param[in] pos iterator to the element to remove
-    @return Iterator following the last removed element. If the iterator @a
-    pos refers to the last element, the `end()` iterator is returned.
-
-    @tparam IteratorType an @ref iterator or @ref const_iterator
-
-    @post Invalidates iterators and references at or after the point of the
-    erase, including the `end()` iterator.
-
-    @throw type_error.307 if called on a `null` value; example: `"cannot use
-    erase() with null"`
-    @throw invalid_iterator.202 if called on an iterator which does not belong
-    to the current JSON value; example: `"iterator does not fit current
-    value"`
-    @throw invalid_iterator.205 if called on a primitive type with invalid
-    iterator (i.e., any iterator which is not `begin()`); example: `"iterator
-    out of range"`
-
-    @complexity The complexity depends on the type:
-    - objects: amortized constant
-    - arrays: linear in distance between @a pos and the end of the container
-    - strings: linear in the length of the string
-    - other types: constant
-
-    @liveexample{The example shows the result of `erase()` for different JSON
-    types.,erase__IteratorType}
-
-    @sa @ref erase(IteratorType, IteratorType) -- removes the elements in
-    the given range
-    @sa @ref erase(const typename object_t::key_type&) -- removes the element
-    from an object at the given key
-    @sa @ref erase(const size_type) -- removes the element from an array at
-    the given index
-
-    @since version 1.0.0
-    */
-    template<class IteratorType, typename std::enable_if<
-                 std::is_same<IteratorType, typename basic_json_t::iterator>::value or
-                 std::is_same<IteratorType, typename basic_json_t::const_iterator>::value, int>::type
-             = 0>
+    /// @brief remove element given an iterator
+    /// @sa https://json.nlohmann.me/api/basic_json/erase/
+    template < class IteratorType, typename std::enable_if <
+                   std::is_same<IteratorType, typename basic_json_t::iterator>::value ||
+                   std::is_same<IteratorType, typename basic_json_t::const_iterator>::value, int >::type
+               = 0 >
     IteratorType erase(IteratorType pos)
     {
         // make sure iterator fits the current value
-        if (JSON_UNLIKELY(this != pos.m_object))
+        if (JSON_HEDLEY_UNLIKELY(this != pos.m_object))
         {
-            JSON_THROW(invalid_iterator::create(202, "iterator does not fit current value"));
+            JSON_THROW(invalid_iterator::create(202, "iterator does not fit current value", *this));
         }
 
         IteratorType result = end();
@@ -14714,10 +19429,11 @@ class basic_json
             case value_t::number_integer:
             case value_t::number_unsigned:
             case value_t::string:
+            case value_t::binary:
             {
-                if (JSON_UNLIKELY(not pos.m_it.primitive_iterator.is_begin()))
+                if (JSON_HEDLEY_UNLIKELY(!pos.m_it.primitive_iterator.is_begin()))
                 {
-                    JSON_THROW(invalid_iterator::create(205, "iterator out of range"));
+                    JSON_THROW(invalid_iterator::create(205, "iterator out of range", *this));
                 }
 
                 if (is_string())
@@ -14727,6 +19443,13 @@ class basic_json
                     std::allocator_traits<decltype(alloc)>::deallocate(alloc, m_value.string, 1);
                     m_value.string = nullptr;
                 }
+                else if (is_binary())
+                {
+                    AllocatorType<binary_t> alloc;
+                    std::allocator_traits<decltype(alloc)>::destroy(alloc, m_value.binary);
+                    std::allocator_traits<decltype(alloc)>::deallocate(alloc, m_value.binary, 1);
+                    m_value.binary = nullptr;
+                }
 
                 m_type = value_t::null;
                 assert_invariant();
@@ -14745,69 +19468,27 @@ class basic_json
                 break;
             }
 
+            case value_t::null:
+            case value_t::discarded:
             default:
-                JSON_THROW(type_error::create(307, "cannot use erase() with " + std::string(type_name())));
+                JSON_THROW(type_error::create(307, "cannot use erase() with " + std::string(type_name()), *this));
         }
 
         return result;
     }
 
-    /*!
-    @brief remove elements given an iterator range
-
-    Removes the element specified by the range `[first; last)`. The iterator
-    @a first does not need to be dereferenceable if `first == last`: erasing
-    an empty range is a no-op.
-
-    If called on a primitive type other than `null`, the resulting JSON value
-    will be `null`.
-
-    @param[in] first iterator to the beginning of the range to remove
-    @param[in] last iterator past the end of the range to remove
-    @return Iterator following the last removed element. If the iterator @a
-    second refers to the last element, the `end()` iterator is returned.
-
-    @tparam IteratorType an @ref iterator or @ref const_iterator
-
-    @post Invalidates iterators and references at or after the point of the
-    erase, including the `end()` iterator.
-
-    @throw type_error.307 if called on a `null` value; example: `"cannot use
-    erase() with null"`
-    @throw invalid_iterator.203 if called on iterators which does not belong
-    to the current JSON value; example: `"iterators do not fit current value"`
-    @throw invalid_iterator.204 if called on a primitive type with invalid
-    iterators (i.e., if `first != begin()` and `last != end()`); example:
-    `"iterators out of range"`
-
-    @complexity The complexity depends on the type:
-    - objects: `log(size()) + std::distance(first, last)`
-    - arrays: linear in the distance between @a first and @a last, plus linear
-      in the distance between @a last and end of the container
-    - strings: linear in the length of the string
-    - other types: constant
-
-    @liveexample{The example shows the result of `erase()` for different JSON
-    types.,erase__IteratorType_IteratorType}
-
-    @sa @ref erase(IteratorType) -- removes the element at a given position
-    @sa @ref erase(const typename object_t::key_type&) -- removes the element
-    from an object at the given key
-    @sa @ref erase(const size_type) -- removes the element from an array at
-    the given index
-
-    @since version 1.0.0
-    */
-    template<class IteratorType, typename std::enable_if<
-                 std::is_same<IteratorType, typename basic_json_t::iterator>::value or
-                 std::is_same<IteratorType, typename basic_json_t::const_iterator>::value, int>::type
-             = 0>
+    /// @brief remove elements given an iterator range
+    /// @sa https://json.nlohmann.me/api/basic_json/erase/
+    template < class IteratorType, typename std::enable_if <
+                   std::is_same<IteratorType, typename basic_json_t::iterator>::value ||
+                   std::is_same<IteratorType, typename basic_json_t::const_iterator>::value, int >::type
+               = 0 >
     IteratorType erase(IteratorType first, IteratorType last)
     {
         // make sure iterator fits the current value
-        if (JSON_UNLIKELY(this != first.m_object or this != last.m_object))
+        if (JSON_HEDLEY_UNLIKELY(this != first.m_object || this != last.m_object))
         {
-            JSON_THROW(invalid_iterator::create(203, "iterators do not fit current value"));
+            JSON_THROW(invalid_iterator::create(203, "iterators do not fit current value", *this));
         }
 
         IteratorType result = end();
@@ -14819,11 +19500,12 @@ class basic_json
             case value_t::number_integer:
             case value_t::number_unsigned:
             case value_t::string:
+            case value_t::binary:
             {
-                if (JSON_LIKELY(not first.m_it.primitive_iterator.is_begin()
-                                or not last.m_it.primitive_iterator.is_end()))
+                if (JSON_HEDLEY_LIKELY(!first.m_it.primitive_iterator.is_begin()
+                                       || !last.m_it.primitive_iterator.is_end()))
                 {
-                    JSON_THROW(invalid_iterator::create(204, "iterators out of range"));
+                    JSON_THROW(invalid_iterator::create(204, "iterators out of range", *this));
                 }
 
                 if (is_string())
@@ -14833,6 +19515,13 @@ class basic_json
                     std::allocator_traits<decltype(alloc)>::deallocate(alloc, m_value.string, 1);
                     m_value.string = nullptr;
                 }
+                else if (is_binary())
+                {
+                    AllocatorType<binary_t> alloc;
+                    std::allocator_traits<decltype(alloc)>::destroy(alloc, m_value.binary);
+                    std::allocator_traits<decltype(alloc)>::deallocate(alloc, m_value.binary, 1);
+                    m_value.binary = nullptr;
+                }
 
                 m_type = value_t::null;
                 assert_invariant();
@@ -14853,92 +19542,45 @@ class basic_json
                 break;
             }
 
+            case value_t::null:
+            case value_t::discarded:
             default:
-                JSON_THROW(type_error::create(307, "cannot use erase() with " + std::string(type_name())));
+                JSON_THROW(type_error::create(307, "cannot use erase() with " + std::string(type_name()), *this));
         }
 
         return result;
     }
 
-    /*!
-    @brief remove element from a JSON object given a key
-
-    Removes elements from a JSON object with the key value @a key.
-
-    @param[in] key value of the elements to remove
-
-    @return Number of elements removed. If @a ObjectType is the default
-    `std::map` type, the return value will always be `0` (@a key was not
-    found) or `1` (@a key was found).
-
-    @post References and iterators to the erased elements are invalidated.
-    Other references and iterators are not affected.
-
-    @throw type_error.307 when called on a type other than JSON object;
-    example: `"cannot use erase() with null"`
-
-    @complexity `log(size()) + count(key)`
-
-    @liveexample{The example shows the effect of `erase()`.,erase__key_type}
-
-    @sa @ref erase(IteratorType) -- removes the element at a given position
-    @sa @ref erase(IteratorType, IteratorType) -- removes the elements in
-    the given range
-    @sa @ref erase(const size_type) -- removes the element from an array at
-    the given index
-
-    @since version 1.0.0
-    */
+    /// @brief remove element from a JSON object given a key
+    /// @sa https://json.nlohmann.me/api/basic_json/erase/
     size_type erase(const typename object_t::key_type& key)
     {
         // this erase only works for objects
-        if (JSON_LIKELY(is_object()))
+        if (JSON_HEDLEY_LIKELY(is_object()))
         {
             return m_value.object->erase(key);
         }
 
-        JSON_THROW(type_error::create(307, "cannot use erase() with " + std::string(type_name())));
+        JSON_THROW(type_error::create(307, "cannot use erase() with " + std::string(type_name()), *this));
     }
 
-    /*!
-    @brief remove element from a JSON array given an index
-
-    Removes element from a JSON array at the index @a idx.
-
-    @param[in] idx index of the element to remove
-
-    @throw type_error.307 when called on a type other than JSON object;
-    example: `"cannot use erase() with null"`
-    @throw out_of_range.401 when `idx >= size()`; example: `"array index 17
-    is out of range"`
-
-    @complexity Linear in distance between @a idx and the end of the container.
-
-    @liveexample{The example shows the effect of `erase()`.,erase__size_type}
-
-    @sa @ref erase(IteratorType) -- removes the element at a given position
-    @sa @ref erase(IteratorType, IteratorType) -- removes the elements in
-    the given range
-    @sa @ref erase(const typename object_t::key_type&) -- removes the element
-    from an object at the given key
-
-    @since version 1.0.0
-    */
+    /// @brief remove element from a JSON array given an index
+    /// @sa https://json.nlohmann.me/api/basic_json/erase/
     void erase(const size_type idx)
     {
         // this erase only works for arrays
-        if (JSON_LIKELY(is_array()))
+        if (JSON_HEDLEY_LIKELY(is_array()))
         {
-            if (JSON_UNLIKELY(idx >= size()))
+            if (JSON_HEDLEY_UNLIKELY(idx >= size()))
             {
-                JSON_THROW(out_of_range::create(401, "array index " + std::to_string(idx) + " is out of range"));
+                JSON_THROW(out_of_range::create(401, "array index " + std::to_string(idx) + " is out of range", *this));
             }
 
             m_value.array->erase(m_value.array->begin() + static_cast<difference_type>(idx));
         }
         else
         {
-            JSON_THROW(type_error::create(307, "cannot use erase() with " + std::string(type_name())));
+            JSON_THROW(type_error::create(307, "cannot use erase() with " + std::string(type_name()), *this));
         }
     }
 
@@ -14952,28 +19594,8 @@ class basic_json
     /// @name lookup
     /// @{
 
-    /*!
-    @brief find an element in a JSON object
-
-    Finds an element in a JSON object with key equivalent to @a key. If the
-    element is not found or the JSON value is not an object, end() is
-    returned.
-
-    @note This method always returns @ref end() when executed on a JSON type
-          that is not an object.
-
-    @param[in] key key value of the element to search for.
-
-    @return Iterator to an element with key equivalent to @a key. If no such
-    element is found or the JSON value is not an object, past-the-end (see
-    @ref end()) iterator is returned.
-
-    @complexity Logarithmic in the size of the JSON object.
-
-    @liveexample{The example shows how `find()` is used.,find__key_type}
-
-    @since version 1.0.0
-    */
+    /// @brief find an element in a JSON object
+    /// @sa https://json.nlohmann.me/api/basic_json/find/
     template<typename KeyT>
     iterator find(KeyT&& key)
     {
@@ -14987,10 +19609,8 @@ class basic_json
         return result;
     }
 
-    /*!
-    @brief find an element in a JSON object
-    @copydoc find(KeyT&&)
-    */
+    /// @brief find an element in a JSON object
+    /// @sa https://json.nlohmann.me/api/basic_json/find/
     template<typename KeyT>
     const_iterator find(KeyT&& key) const
     {
@@ -15004,27 +19624,8 @@ class basic_json
         return result;
     }
 
-    /*!
-    @brief returns the number of occurrences of a key in a JSON object
-
-    Returns the number of elements with key @a key. If ObjectType is the
-    default `std::map` type, the return value will always be `0` (@a key was
-    not found) or `1` (@a key was found).
-
-    @note This method always returns `0` when executed on a JSON type that is
-          not an object.
-
-    @param[in] key key value of the element to count
-
-    @return Number of elements with key @a key. If the JSON value is not an
-    object, the return value will be `0`.
-
-    @complexity Logarithmic in the size of the JSON object.
-
-    @liveexample{The example shows how `count()` is used.,count}
-
-    @since version 1.0.0
-    */
+    /// @brief returns the number of occurrences of a key in a JSON object
+    /// @sa https://json.nlohmann.me/api/basic_json/count/
     template<typename KeyT>
     size_type count(KeyT&& key) const
     {
@@ -15032,6 +19633,22 @@ class basic_json
         return is_object() ? m_value.object->count(std::forward<KeyT>(key)) : 0;
     }
 
+    /// @brief check the existence of an element in a JSON object
+    /// @sa https://json.nlohmann.me/api/basic_json/contains/
+    template < typename KeyT, typename std::enable_if <
+                   !std::is_same<typename std::decay<KeyT>::type, json_pointer>::value, int >::type = 0 >
+    bool contains(KeyT && key) const
+    {
+        return is_object() && m_value.object->find(std::forward<KeyT>(key)) != m_value.object->end();
+    }
+
+    /// @brief check the existence of an element in a JSON object given a JSON pointer
+    /// @sa https://json.nlohmann.me/api/basic_json/contains/
+    bool contains(const json_pointer& ptr) const
+    {
+        return ptr.contains(this);
+    }
+
     /// @}
 
 
@@ -15042,30 +19659,8 @@ class basic_json
     /// @name iterators
     /// @{
 
-    /*!
-    @brief returns an iterator to the first element
-
-    Returns an iterator to the first element.
-
-    @image html range-begin-end.svg "Illustration from cppreference.com"
-
-    @return iterator to the first element
-
-    @complexity Constant.
-
-    @requirement This function helps `basic_json` satisfying the
-    [Container](https://en.cppreference.com/w/cpp/named_req/Container)
-    requirements:
-    - The complexity is constant.
-
-    @liveexample{The following code shows an example for `begin()`.,begin}
-
-    @sa @ref cbegin() -- returns a const iterator to the beginning
-    @sa @ref end() -- returns an iterator to the end
-    @sa @ref cend() -- returns a const iterator to the end
-
-    @since version 1.0.0
-    */
+    /// @brief returns an iterator to the first element
+    /// @sa https://json.nlohmann.me/api/basic_json/begin/
     iterator begin() noexcept
     {
         iterator result(this);
@@ -15073,39 +19668,15 @@ class basic_json
         return result;
     }
 
-    /*!
-    @copydoc basic_json::cbegin()
-    */
+    /// @brief returns an iterator to the first element
+    /// @sa https://json.nlohmann.me/api/basic_json/begin/
     const_iterator begin() const noexcept
     {
         return cbegin();
     }
 
-    /*!
-    @brief returns a const iterator to the first element
-
-    Returns a const iterator to the first element.
-
-    @image html range-begin-end.svg "Illustration from cppreference.com"
-
-    @return const iterator to the first element
-
-    @complexity Constant.
-
-    @requirement This function helps `basic_json` satisfying the
-    [Container](https://en.cppreference.com/w/cpp/named_req/Container)
-    requirements:
-    - The complexity is constant.
-    - Has the semantics of `const_cast<const basic_json&>(*this).begin()`.
-
-    @liveexample{The following code shows an example for `cbegin()`.,cbegin}
-
-    @sa @ref begin() -- returns an iterator to the beginning
-    @sa @ref end() -- returns an iterator to the end
-    @sa @ref cend() -- returns a const iterator to the end
-
-    @since version 1.0.0
-    */
+    /// @brief returns a const iterator to the first element
+    /// @sa https://json.nlohmann.me/api/basic_json/cbegin/
     const_iterator cbegin() const noexcept
     {
         const_iterator result(this);
@@ -15113,30 +19684,8 @@ class basic_json
         return result;
     }
 
-    /*!
-    @brief returns an iterator to one past the last element
-
-    Returns an iterator to one past the last element.
-
-    @image html range-begin-end.svg "Illustration from cppreference.com"
-
-    @return iterator one past the last element
-
-    @complexity Constant.
-
-    @requirement This function helps `basic_json` satisfying the
-    [Container](https://en.cppreference.com/w/cpp/named_req/Container)
-    requirements:
-    - The complexity is constant.
-
-    @liveexample{The following code shows an example for `end()`.,end}
-
-    @sa @ref cend() -- returns a const iterator to the end
-    @sa @ref begin() -- returns an iterator to the beginning
-    @sa @ref cbegin() -- returns a const iterator to the beginning
-
-    @since version 1.0.0
-    */
+    /// @brief returns an iterator to one past the last element
+    /// @sa https://json.nlohmann.me/api/basic_json/end/
     iterator end() noexcept
     {
         iterator result(this);
@@ -15144,39 +19693,15 @@ class basic_json
         return result;
     }
 
-    /*!
-    @copydoc basic_json::cend()
-    */
+    /// @brief returns an iterator to one past the last element
+    /// @sa https://json.nlohmann.me/api/basic_json/end/
     const_iterator end() const noexcept
     {
         return cend();
     }
 
-    /*!
-    @brief returns a const iterator to one past the last element
-
-    Returns a const iterator to one past the last element.
-
-    @image html range-begin-end.svg "Illustration from cppreference.com"
-
-    @return const iterator one past the last element
-
-    @complexity Constant.
-
-    @requirement This function helps `basic_json` satisfying the
-    [Container](https://en.cppreference.com/w/cpp/named_req/Container)
-    requirements:
-    - The complexity is constant.
-    - Has the semantics of `const_cast<const basic_json&>(*this).end()`.
-
-    @liveexample{The following code shows an example for `cend()`.,cend}
-
-    @sa @ref end() -- returns an iterator to the end
-    @sa @ref begin() -- returns an iterator to the beginning
-    @sa @ref cbegin() -- returns a const iterator to the beginning
-
-    @since version 1.0.0
-    */
+    /// @brief returns an iterator to one past the last element
+    /// @sa https://json.nlohmann.me/api/basic_json/cend/
     const_iterator cend() const noexcept
     {
         const_iterator result(this);
@@ -15184,270 +19709,80 @@ class basic_json
         return result;
     }
 
-    /*!
-    @brief returns an iterator to the reverse-beginning
-
-    Returns an iterator to the reverse-beginning; that is, the last element.
-
-    @image html range-rbegin-rend.svg "Illustration from cppreference.com"
-
-    @complexity Constant.
-
-    @requirement This function helps `basic_json` satisfying the
-    [ReversibleContainer](https://en.cppreference.com/w/cpp/named_req/ReversibleContainer)
-    requirements:
-    - The complexity is constant.
-    - Has the semantics of `reverse_iterator(end())`.
-
-    @liveexample{The following code shows an example for `rbegin()`.,rbegin}
-
-    @sa @ref crbegin() -- returns a const reverse iterator to the beginning
-    @sa @ref rend() -- returns a reverse iterator to the end
-    @sa @ref crend() -- returns a const reverse iterator to the end
-
-    @since version 1.0.0
-    */
+    /// @brief returns an iterator to the reverse-beginning
+    /// @sa https://json.nlohmann.me/api/basic_json/rbegin/
     reverse_iterator rbegin() noexcept
     {
         return reverse_iterator(end());
     }
 
-    /*!
-    @copydoc basic_json::crbegin()
-    */
+    /// @brief returns an iterator to the reverse-beginning
+    /// @sa https://json.nlohmann.me/api/basic_json/rbegin/
     const_reverse_iterator rbegin() const noexcept
     {
         return crbegin();
     }
 
-    /*!
-    @brief returns an iterator to the reverse-end
-
-    Returns an iterator to the reverse-end; that is, one before the first
-    element.
-
-    @image html range-rbegin-rend.svg "Illustration from cppreference.com"
-
-    @complexity Constant.
-
-    @requirement This function helps `basic_json` satisfying the
-    [ReversibleContainer](https://en.cppreference.com/w/cpp/named_req/ReversibleContainer)
-    requirements:
-    - The complexity is constant.
-    - Has the semantics of `reverse_iterator(begin())`.
-
-    @liveexample{The following code shows an example for `rend()`.,rend}
-
-    @sa @ref crend() -- returns a const reverse iterator to the end
-    @sa @ref rbegin() -- returns a reverse iterator to the beginning
-    @sa @ref crbegin() -- returns a const reverse iterator to the beginning
-
-    @since version 1.0.0
-    */
+    /// @brief returns an iterator to the reverse-end
+    /// @sa https://json.nlohmann.me/api/basic_json/rend/
     reverse_iterator rend() noexcept
     {
         return reverse_iterator(begin());
     }
 
-    /*!
-    @copydoc basic_json::crend()
-    */
+    /// @brief returns an iterator to the reverse-end
+    /// @sa https://json.nlohmann.me/api/basic_json/rend/
     const_reverse_iterator rend() const noexcept
     {
         return crend();
     }
 
-    /*!
-    @brief returns a const reverse iterator to the last element
-
-    Returns a const iterator to the reverse-beginning; that is, the last
-    element.
-
-    @image html range-rbegin-rend.svg "Illustration from cppreference.com"
-
-    @complexity Constant.
-
-    @requirement This function helps `basic_json` satisfying the
-    [ReversibleContainer](https://en.cppreference.com/w/cpp/named_req/ReversibleContainer)
-    requirements:
-    - The complexity is constant.
-    - Has the semantics of `const_cast<const basic_json&>(*this).rbegin()`.
-
-    @liveexample{The following code shows an example for `crbegin()`.,crbegin}
-
-    @sa @ref rbegin() -- returns a reverse iterator to the beginning
-    @sa @ref rend() -- returns a reverse iterator to the end
-    @sa @ref crend() -- returns a const reverse iterator to the end
-
-    @since version 1.0.0
-    */
+    /// @brief returns a const reverse iterator to the last element
+    /// @sa https://json.nlohmann.me/api/basic_json/crbegin/
     const_reverse_iterator crbegin() const noexcept
     {
         return const_reverse_iterator(cend());
     }
 
-    /*!
-    @brief returns a const reverse iterator to one before the first
-
-    Returns a const reverse iterator to the reverse-end; that is, one before
-    the first element.
-
-    @image html range-rbegin-rend.svg "Illustration from cppreference.com"
-
-    @complexity Constant.
-
-    @requirement This function helps `basic_json` satisfying the
-    [ReversibleContainer](https://en.cppreference.com/w/cpp/named_req/ReversibleContainer)
-    requirements:
-    - The complexity is constant.
-    - Has the semantics of `const_cast<const basic_json&>(*this).rend()`.
-
-    @liveexample{The following code shows an example for `crend()`.,crend}
-
-    @sa @ref rend() -- returns a reverse iterator to the end
-    @sa @ref rbegin() -- returns a reverse iterator to the beginning
-    @sa @ref crbegin() -- returns a const reverse iterator to the beginning
-
-    @since version 1.0.0
-    */
+    /// @brief returns a const reverse iterator to one before the first
+    /// @sa https://json.nlohmann.me/api/basic_json/crend/
     const_reverse_iterator crend() const noexcept
     {
         return const_reverse_iterator(cbegin());
     }
 
   public:
-    /*!
-    @brief wrapper to access iterator member functions in range-based for
-
-    This function allows to access @ref iterator::key() and @ref
-    iterator::value() during range-based for loops. In these loops, a
-    reference to the JSON values is returned, so there is no access to the
-    underlying iterator.
-
-    For loop without iterator_wrapper:
-
-    @code{cpp}
-    for (auto it = j_object.begin(); it != j_object.end(); ++it)
-    {
-        std::cout << "key: " << it.key() << ", value:" << it.value() << '\n';
-    }
-    @endcode
-
-    Range-based for loop without iterator proxy:
-
-    @code{cpp}
-    for (auto it : j_object)
-    {
-        // "it" is of type json::reference and has no key() member
-        std::cout << "value: " << it << '\n';
-    }
-    @endcode
-
-    Range-based for loop with iterator proxy:
-
-    @code{cpp}
-    for (auto it : json::iterator_wrapper(j_object))
-    {
-        std::cout << "key: " << it.key() << ", value:" << it.value() << '\n';
-    }
-    @endcode
-
-    @note When iterating over an array, `key()` will return the index of the
-          element as string (see example).
-
-    @param[in] ref  reference to a JSON value
-    @return iteration proxy object wrapping @a ref with an interface to use in
-            range-based for loops
-
-    @liveexample{The following code shows how the wrapper is used,iterator_wrapper}
-
-    @exceptionsafety Strong guarantee: if an exception is thrown, there are no
-    changes in the JSON value.
-
-    @complexity Constant.
-
-    @note The name of this function is not yet final and may change in the
-    future.
-
-    @deprecated This stream operator is deprecated and will be removed in
-                future 4.0.0 of the library. Please use @ref items() instead;
-                that is, replace `json::iterator_wrapper(j)` with `j.items()`.
-    */
-    JSON_DEPRECATED
+    /// @brief wrapper to access iterator member functions in range-based for
+    /// @sa https://json.nlohmann.me/api/basic_json/items/
+    /// @deprecated This function is deprecated since 3.1.0 and will be removed in
+    ///             version 4.0.0 of the library. Please use @ref items() instead;
+    ///             that is, replace `json::iterator_wrapper(j)` with `j.items()`.
+    JSON_HEDLEY_DEPRECATED_FOR(3.1.0, items())
     static iteration_proxy<iterator> iterator_wrapper(reference ref) noexcept
     {
         return ref.items();
     }
 
-    /*!
-    @copydoc iterator_wrapper(reference)
-    */
-    JSON_DEPRECATED
+    /// @brief wrapper to access iterator member functions in range-based for
+    /// @sa https://json.nlohmann.me/api/basic_json/items/
+    /// @deprecated This function is deprecated since 3.1.0 and will be removed in
+    ///         version 4.0.0 of the library. Please use @ref items() instead;
+    ///         that is, replace `json::iterator_wrapper(j)` with `j.items()`.
+    JSON_HEDLEY_DEPRECATED_FOR(3.1.0, items())
     static iteration_proxy<const_iterator> iterator_wrapper(const_reference ref) noexcept
     {
         return ref.items();
     }
 
-    /*!
-    @brief helper to access iterator member functions in range-based for
-
-    This function allows to access @ref iterator::key() and @ref
-    iterator::value() during range-based for loops. In these loops, a
-    reference to the JSON values is returned, so there is no access to the
-    underlying iterator.
-
-    For loop without `items()` function:
-
-    @code{cpp}
-    for (auto it = j_object.begin(); it != j_object.end(); ++it)
-    {
-        std::cout << "key: " << it.key() << ", value:" << it.value() << '\n';
-    }
-    @endcode
-
-    Range-based for loop without `items()` function:
-
-    @code{cpp}
-    for (auto it : j_object)
-    {
-        // "it" is of type json::reference and has no key() member
-        std::cout << "value: " << it << '\n';
-    }
-    @endcode
-
-    Range-based for loop with `items()` function:
-
-    @code{cpp}
-    for (auto it : j_object.items())
-    {
-        std::cout << "key: " << it.key() << ", value:" << it.value() << '\n';
-    }
-    @endcode
-
-    @note When iterating over an array, `key()` will return the index of the
-          element as string (see example). For primitive types (e.g., numbers),
-          `key()` returns an empty string.
-
-    @return iteration proxy object wrapping @a ref with an interface to use in
-            range-based for loops
-
-    @liveexample{The following code shows how the function is used.,items}
-
-    @exceptionsafety Strong guarantee: if an exception is thrown, there are no
-    changes in the JSON value.
-
-    @complexity Constant.
-
-    @since version 3.1.0.
-    */
+    /// @brief helper to access iterator member functions in range-based for
+    /// @sa https://json.nlohmann.me/api/basic_json/items/
     iteration_proxy<iterator> items() noexcept
     {
         return iteration_proxy<iterator>(*this);
     }
 
-    /*!
-    @copydoc items()
-    */
+    /// @brief helper to access iterator member functions in range-based for
+    /// @sa https://json.nlohmann.me/api/basic_json/items/
     iteration_proxy<const_iterator> items() const noexcept
     {
         return iteration_proxy<const_iterator>(*this);
@@ -15463,47 +19798,8 @@ class basic_json
     /// @name capacity
     /// @{
 
-    /*!
-    @brief checks whether the container is empty.
-
-    Checks if a JSON value has no elements (i.e. whether its @ref size is `0`).
-
-    @return The return value depends on the different types and is
-            defined as follows:
-            Value type  | return value
-            ----------- | -------------
-            null        | `true`
-            boolean     | `false`
-            string      | `false`
-            number      | `false`
-            object      | result of function `object_t::empty()`
-            array       | result of function `array_t::empty()`
-
-    @liveexample{The following code uses `empty()` to check if a JSON
-    object contains any elements.,empty}
-
-    @complexity Constant, as long as @ref array_t and @ref object_t satisfy
-    the Container concept; that is, their `empty()` functions have constant
-    complexity.
-
-    @iterators No changes.
-
-    @exceptionsafety No-throw guarantee: this function never throws exceptions.
-
-    @note This function does not return whether a string stored as JSON value
-    is empty - it returns whether the JSON container itself is empty which is
-    false in the case of a string.
-
-    @requirement This function helps `basic_json` satisfying the
-    [Container](https://en.cppreference.com/w/cpp/named_req/Container)
-    requirements:
-    - The complexity is constant.
-    - Has the semantics of `begin() == end()`.
-
-    @sa @ref size() -- returns the number of elements
-
-    @since version 1.0.0
-    */
+    /// @brief checks whether the container is empty.
+    /// @sa https://json.nlohmann.me/api/basic_json/empty/
     bool empty() const noexcept
     {
         switch (m_type)
@@ -15526,6 +19822,13 @@ class basic_json
                 return m_value.object->empty();
             }
 
+            case value_t::string:
+            case value_t::boolean:
+            case value_t::number_integer:
+            case value_t::number_unsigned:
+            case value_t::number_float:
+            case value_t::binary:
+            case value_t::discarded:
             default:
             {
                 // all other types are nonempty
@@ -15534,48 +19837,8 @@ class basic_json
         }
     }
 
-    /*!
-    @brief returns the number of elements
-
-    Returns the number of elements in a JSON value.
-
-    @return The return value depends on the different types and is
-            defined as follows:
-            Value type  | return value
-            ----------- | -------------
-            null        | `0`
-            boolean     | `1`
-            string      | `1`
-            number      | `1`
-            object      | result of function object_t::size()
-            array       | result of function array_t::size()
-
-    @liveexample{The following code calls `size()` on the different value
-    types.,size}
-
-    @complexity Constant, as long as @ref array_t and @ref object_t satisfy
-    the Container concept; that is, their size() functions have constant
-    complexity.
-
-    @iterators No changes.
-
-    @exceptionsafety No-throw guarantee: this function never throws exceptions.
-
-    @note This function does not return the length of a string stored as JSON
-    value - it returns the number of elements in the JSON value which is 1 in
-    the case of a string.
-
-    @requirement This function helps `basic_json` satisfying the
-    [Container](https://en.cppreference.com/w/cpp/named_req/Container)
-    requirements:
-    - The complexity is constant.
-    - Has the semantics of `std::distance(begin(), end())`.
-
-    @sa @ref empty() -- checks whether the container is empty
-    @sa @ref max_size() -- returns the maximal number of elements
-
-    @since version 1.0.0
-    */
+    /// @brief returns the number of elements
+    /// @sa https://json.nlohmann.me/api/basic_json/size/
     size_type size() const noexcept
     {
         switch (m_type)
@@ -15598,6 +19861,13 @@ class basic_json
                 return m_value.object->size();
             }
 
+            case value_t::string:
+            case value_t::boolean:
+            case value_t::number_integer:
+            case value_t::number_unsigned:
+            case value_t::number_float:
+            case value_t::binary:
+            case value_t::discarded:
             default:
             {
                 // all other types have size 1
@@ -15606,46 +19876,8 @@ class basic_json
         }
     }
 
-    /*!
-    @brief returns the maximum possible number of elements
-
-    Returns the maximum number of elements a JSON value is able to hold due to
-    system or library implementation limitations, i.e. `std::distance(begin(),
-    end())` for the JSON value.
-
-    @return The return value depends on the different types and is
-            defined as follows:
-            Value type  | return value
-            ----------- | -------------
-            null        | `0` (same as `size()`)
-            boolean     | `1` (same as `size()`)
-            string      | `1` (same as `size()`)
-            number      | `1` (same as `size()`)
-            object      | result of function `object_t::max_size()`
-            array       | result of function `array_t::max_size()`
-
-    @liveexample{The following code calls `max_size()` on the different value
-    types. Note the output is implementation specific.,max_size}
-
-    @complexity Constant, as long as @ref array_t and @ref object_t satisfy
-    the Container concept; that is, their `max_size()` functions have constant
-    complexity.
-
-    @iterators No changes.
-
-    @exceptionsafety No-throw guarantee: this function never throws exceptions.
-
-    @requirement This function helps `basic_json` satisfying the
-    [Container](https://en.cppreference.com/w/cpp/named_req/Container)
-    requirements:
-    - The complexity is constant.
-    - Has the semantics of returning `b.size()` where `b` is the largest
-      possible JSON value.
-
-    @sa @ref size() -- returns the number of elements
-
-    @since version 1.0.0
-    */
+    /// @brief returns the maximum possible number of elements
+    /// @sa https://json.nlohmann.me/api/basic_json/max_size/
     size_type max_size() const noexcept
     {
         switch (m_type)
@@ -15662,6 +19894,14 @@ class basic_json
                 return m_value.object->max_size();
             }
 
+            case value_t::null:
+            case value_t::string:
+            case value_t::boolean:
+            case value_t::number_integer:
+            case value_t::number_unsigned:
+            case value_t::number_float:
+            case value_t::binary:
+            case value_t::discarded:
             default:
             {
                 // all other types have max_size() == size()
@@ -15680,42 +19920,8 @@ class basic_json
     /// @name modifiers
     /// @{
 
-    /*!
-    @brief clears the contents
-
-    Clears the content of a JSON value and resets it to the default value as
-    if @ref basic_json(value_t) would have been called with the current value
-    type from @ref type():
-
-    Value type  | initial value
-    ----------- | -------------
-    null        | `null`
-    boolean     | `false`
-    string      | `""`
-    number      | `0`
-    object      | `{}`
-    array       | `[]`
-
-    @post Has the same effect as calling
-    @code {.cpp}
-    *this = basic_json(type());
-    @endcode
-
-    @liveexample{The example below shows the effect of `clear()` to different
-    JSON types.,clear}
-
-    @complexity Linear in the size of the JSON value.
-
-    @iterators All iterators, pointers and references related to this container
-               are invalidated.
-
-    @exceptionsafety No-throw guarantee: this function never throws exceptions.
-
-    @sa @ref basic_json(value_t) -- constructor that creates an object with the
-        same value than calling `clear()`
-
-    @since version 1.0.0
-    */
+    /// @brief clears the contents
+    /// @sa https://json.nlohmann.me/api/basic_json/clear/
     void clear() noexcept
     {
         switch (m_type)
@@ -15750,6 +19956,12 @@ class basic_json
                 break;
             }
 
+            case value_t::binary:
+            {
+                m_value.binary->clear();
+                break;
+            }
+
             case value_t::array:
             {
                 m_value.array->clear();
@@ -15762,37 +19974,21 @@ class basic_json
                 break;
             }
 
+            case value_t::null:
+            case value_t::discarded:
             default:
                 break;
         }
     }
 
-    /*!
-    @brief add an object to an array
-
-    Appends the given element @a val to the end of the JSON value. If the
-    function is called on a JSON null value, an empty array is created before
-    appending @a val.
-
-    @param[in] val the value to add to the JSON array
-
-    @throw type_error.308 when called on a type other than JSON array or
-    null; example: `"cannot use push_back() with number"`
-
-    @complexity Amortized constant.
-
-    @liveexample{The example shows how `push_back()` and `+=` can be used to
-    add elements to a JSON array. Note how the `null` value was silently
-    converted to a JSON array.,push_back}
-
-    @since version 1.0.0
-    */
+    /// @brief add an object to an array
+    /// @sa https://json.nlohmann.me/api/basic_json/push_back/
     void push_back(basic_json&& val)
     {
         // push_back only works for null objects or arrays
-        if (JSON_UNLIKELY(not(is_null() or is_array())))
+        if (JSON_HEDLEY_UNLIKELY(!(is_null() || is_array())))
         {
-            JSON_THROW(type_error::create(308, "cannot use push_back() with " + std::string(type_name())));
+            JSON_THROW(type_error::create(308, "cannot use push_back() with " + std::string(type_name()), *this));
         }
 
         // transform null object into an array
@@ -15804,31 +20000,28 @@ class basic_json
         }
 
         // add element to array (move semantics)
+        const auto old_capacity = m_value.array->capacity();
         m_value.array->push_back(std::move(val));
-        // invalidate object
-        val.m_type = value_t::null;
+        set_parent(m_value.array->back(), old_capacity);
+        // if val is moved from, basic_json move constructor marks it null, so we do not call the destructor
     }
 
-    /*!
-    @brief add an object to an array
-    @copydoc push_back(basic_json&&)
-    */
+    /// @brief add an object to an array
+    /// @sa https://json.nlohmann.me/api/basic_json/operator+=/
     reference operator+=(basic_json&& val)
     {
         push_back(std::move(val));
         return *this;
     }
 
-    /*!
-    @brief add an object to an array
-    @copydoc push_back(basic_json&&)
-    */
+    /// @brief add an object to an array
+    /// @sa https://json.nlohmann.me/api/basic_json/push_back/
     void push_back(const basic_json& val)
     {
         // push_back only works for null objects or arrays
-        if (JSON_UNLIKELY(not(is_null() or is_array())))
+        if (JSON_HEDLEY_UNLIKELY(!(is_null() || is_array())))
         {
-            JSON_THROW(type_error::create(308, "cannot use push_back() with " + std::string(type_name())));
+            JSON_THROW(type_error::create(308, "cannot use push_back() with " + std::string(type_name()), *this));
         }
 
         // transform null object into an array
@@ -15840,45 +20033,27 @@ class basic_json
         }
 
         // add element to array
+        const auto old_capacity = m_value.array->capacity();
         m_value.array->push_back(val);
+        set_parent(m_value.array->back(), old_capacity);
     }
 
-    /*!
-    @brief add an object to an array
-    @copydoc push_back(basic_json&&)
-    */
+    /// @brief add an object to an array
+    /// @sa https://json.nlohmann.me/api/basic_json/operator+=/
     reference operator+=(const basic_json& val)
     {
         push_back(val);
         return *this;
     }
 
-    /*!
-    @brief add an object to an object
-
-    Inserts the given element @a val to the JSON object. If the function is
-    called on a JSON null value, an empty object is created before inserting
-    @a val.
-
-    @param[in] val the value to add to the JSON object
-
-    @throw type_error.308 when called on a type other than JSON object or
-    null; example: `"cannot use push_back() with number"`
-
-    @complexity Logarithmic in the size of the container, O(log(`size()`)).
-
-    @liveexample{The example shows how `push_back()` and `+=` can be used to
-    add elements to a JSON object. Note how the `null` value was silently
-    converted to a JSON object.,push_back__object_t__value}
-
-    @since version 1.0.0
-    */
+    /// @brief add an object to an object
+    /// @sa https://json.nlohmann.me/api/basic_json/push_back/
     void push_back(const typename object_t::value_type& val)
     {
         // push_back only works for null objects or objects
-        if (JSON_UNLIKELY(not(is_null() or is_object())))
+        if (JSON_HEDLEY_UNLIKELY(!(is_null() || is_object())))
         {
-            JSON_THROW(type_error::create(308, "cannot use push_back() with " + std::string(type_name())));
+            JSON_THROW(type_error::create(308, "cannot use push_back() with " + std::string(type_name()), *this));
         }
 
         // transform null object into an object
@@ -15889,48 +20064,24 @@ class basic_json
             assert_invariant();
         }
 
-        // add element to array
-        m_value.object->insert(val);
+        // add element to object
+        auto res = m_value.object->insert(val);
+        set_parent(res.first->second);
     }
 
-    /*!
-    @brief add an object to an object
-    @copydoc push_back(const typename object_t::value_type&)
-    */
+    /// @brief add an object to an object
+    /// @sa https://json.nlohmann.me/api/basic_json/operator+=/
     reference operator+=(const typename object_t::value_type& val)
     {
         push_back(val);
         return *this;
     }
 
-    /*!
-    @brief add an object to an object
-
-    This function allows to use `push_back` with an initializer list. In case
-
-    1. the current value is an object,
-    2. the initializer list @a init contains only two elements, and
-    3. the first element of @a init is a string,
-
-    @a init is converted into an object element and added using
-    @ref push_back(const typename object_t::value_type&). Otherwise, @a init
-    is converted to a JSON value and added using @ref push_back(basic_json&&).
-
-    @param[in] init  an initializer list
-
-    @complexity Linear in the size of the initializer list @a init.
-
-    @note This function is required to resolve an ambiguous overload error,
-          because pairs like `{"key", "value"}` can be both interpreted as
-          `object_t::value_type` or `std::initializer_list<basic_json>`, see
-          https://github.com/nlohmann/json/issues/235 for more information.
-
-    @liveexample{The example shows how initializer lists are treated as
-    objects when possible.,push_back__initializer_list}
-    */
+    /// @brief add an object to an object
+    /// @sa https://json.nlohmann.me/api/basic_json/push_back/
     void push_back(initializer_list_t init)
     {
-        if (is_object() and init.size() == 2 and (*init.begin())->is_string())
+        if (is_object() && init.size() == 2 && (*init.begin())->is_string())
         {
             basic_json&& key = init.begin()->moved_or_copied();
             push_back(typename object_t::value_type(
@@ -15942,44 +20093,23 @@ class basic_json
         }
     }
 
-    /*!
-    @brief add an object to an object
-    @copydoc push_back(initializer_list_t)
-    */
+    /// @brief add an object to an object
+    /// @sa https://json.nlohmann.me/api/basic_json/operator+=/
     reference operator+=(initializer_list_t init)
     {
         push_back(init);
         return *this;
     }
 
-    /*!
-    @brief add an object to an array
-
-    Creates a JSON value from the passed parameters @a args to the end of the
-    JSON value. If the function is called on a JSON null value, an empty array
-    is created before appending the value created from @a args.
-
-    @param[in] args arguments to forward to a constructor of @ref basic_json
-    @tparam Args compatible types to create a @ref basic_json object
-
-    @throw type_error.311 when called on a type other than JSON array or
-    null; example: `"cannot use emplace_back() with number"`
-
-    @complexity Amortized constant.
-
-    @liveexample{The example shows how `push_back()` can be used to add
-    elements to a JSON array. Note how the `null` value was silently converted
-    to a JSON array.,emplace_back}
-
-    @since version 2.0.8
-    */
+    /// @brief add an object to an array
+    /// @sa https://json.nlohmann.me/api/basic_json/emplace_back/
     template<class... Args>
-    void emplace_back(Args&& ... args)
+    reference emplace_back(Args&& ... args)
     {
         // emplace_back only works for null objects or arrays
-        if (JSON_UNLIKELY(not(is_null() or is_array())))
+        if (JSON_HEDLEY_UNLIKELY(!(is_null() || is_array())))
         {
-            JSON_THROW(type_error::create(311, "cannot use emplace_back() with " + std::string(type_name())));
+            JSON_THROW(type_error::create(311, "cannot use emplace_back() with " + std::string(type_name()), *this));
         }
 
         // transform null object into an array
@@ -15991,43 +20121,20 @@ class basic_json
         }
 
         // add element to array (perfect forwarding)
+        const auto old_capacity = m_value.array->capacity();
         m_value.array->emplace_back(std::forward<Args>(args)...);
+        return set_parent(m_value.array->back(), old_capacity);
     }
 
-    /*!
-    @brief add an object to an object if key does not exist
-
-    Inserts a new element into a JSON object constructed in-place with the
-    given @a args if there is no element with the key in the container. If the
-    function is called on a JSON null value, an empty object is created before
-    appending the value created from @a args.
-
-    @param[in] args arguments to forward to a constructor of @ref basic_json
-    @tparam Args compatible types to create a @ref basic_json object
-
-    @return a pair consisting of an iterator to the inserted element, or the
-            already-existing element if no insertion happened, and a bool
-            denoting whether the insertion took place.
-
-    @throw type_error.311 when called on a type other than JSON object or
-    null; example: `"cannot use emplace() with number"`
-
-    @complexity Logarithmic in the size of the container, O(log(`size()`)).
-
-    @liveexample{The example shows how `emplace()` can be used to add elements
-    to a JSON object. Note how the `null` value was silently converted to a
-    JSON object. Further note how no value is added if there was already one
-    value stored with the same key.,emplace}
-
-    @since version 2.0.8
-    */
+    /// @brief add an object to an object if key does not exist
+    /// @sa https://json.nlohmann.me/api/basic_json/emplace/
     template<class... Args>
     std::pair<iterator, bool> emplace(Args&& ... args)
     {
         // emplace only works for null objects or arrays
-        if (JSON_UNLIKELY(not(is_null() or is_object())))
+        if (JSON_HEDLEY_UNLIKELY(!(is_null() || is_object())))
         {
-            JSON_THROW(type_error::create(311, "cannot use emplace() with " + std::string(type_name())));
+            JSON_THROW(type_error::create(311, "cannot use emplace() with " + std::string(type_name()), *this));
         }
 
         // transform null object into an object
@@ -16040,6 +20147,8 @@ class basic_json
 
         // add element to array (perfect forwarding)
         auto res = m_value.object->emplace(std::forward<Args>(args)...);
+        set_parent(res.first->second);
+
         // create result iterator and set iterator to the result of emplace
         auto it = begin();
         it.m_it.object_iterator = res.first;
@@ -16055,7 +20164,7 @@ class basic_json
     iterator insert_iterator(const_iterator pos, Args&& ... args)
     {
         iterator result(this);
-        assert(m_value.array != nullptr);
+        JSON_ASSERT(m_value.array != nullptr);
 
         auto insert_pos = std::distance(m_value.array->begin(), pos.m_it.array_iterator);
         m_value.array->insert(pos.m_it.array_iterator, std::forward<Args>(args)...);
@@ -16065,318 +20174,143 @@ class basic_json
         // result.m_it.array_iterator = m_value.array->insert(pos.m_it.array_iterator, cnt, val);
         // but the return value of insert is missing in GCC 4.8, so it is written this way instead.
 
+        set_parents();
         return result;
     }
 
-    /*!
-    @brief inserts element
-
-    Inserts element @a val before iterator @a pos.
-
-    @param[in] pos iterator before which the content will be inserted; may be
-    the end() iterator
-    @param[in] val element to insert
-    @return iterator pointing to the inserted @a val.
-
-    @throw type_error.309 if called on JSON values other than arrays;
-    example: `"cannot use insert() with string"`
-    @throw invalid_iterator.202 if @a pos is not an iterator of *this;
-    example: `"iterator does not fit current value"`
-
-    @complexity Constant plus linear in the distance between @a pos and end of
-    the container.
-
-    @liveexample{The example shows how `insert()` is used.,insert}
-
-    @since version 1.0.0
-    */
+    /// @brief inserts element into array
+    /// @sa https://json.nlohmann.me/api/basic_json/insert/
     iterator insert(const_iterator pos, const basic_json& val)
     {
         // insert only works for arrays
-        if (JSON_LIKELY(is_array()))
+        if (JSON_HEDLEY_LIKELY(is_array()))
         {
             // check if iterator pos fits to this JSON value
-            if (JSON_UNLIKELY(pos.m_object != this))
+            if (JSON_HEDLEY_UNLIKELY(pos.m_object != this))
             {
-                JSON_THROW(invalid_iterator::create(202, "iterator does not fit current value"));
+                JSON_THROW(invalid_iterator::create(202, "iterator does not fit current value", *this));
             }
 
             // insert to array and return iterator
             return insert_iterator(pos, val);
         }
 
-        JSON_THROW(type_error::create(309, "cannot use insert() with " + std::string(type_name())));
+        JSON_THROW(type_error::create(309, "cannot use insert() with " + std::string(type_name()), *this));
     }
 
-    /*!
-    @brief inserts element
-    @copydoc insert(const_iterator, const basic_json&)
-    */
+    /// @brief inserts element into array
+    /// @sa https://json.nlohmann.me/api/basic_json/insert/
     iterator insert(const_iterator pos, basic_json&& val)
     {
         return insert(pos, val);
     }
 
-    /*!
-    @brief inserts elements
-
-    Inserts @a cnt copies of @a val before iterator @a pos.
-
-    @param[in] pos iterator before which the content will be inserted; may be
-    the end() iterator
-    @param[in] cnt number of copies of @a val to insert
-    @param[in] val element to insert
-    @return iterator pointing to the first element inserted, or @a pos if
-    `cnt==0`
-
-    @throw type_error.309 if called on JSON values other than arrays; example:
-    `"cannot use insert() with string"`
-    @throw invalid_iterator.202 if @a pos is not an iterator of *this;
-    example: `"iterator does not fit current value"`
-
-    @complexity Linear in @a cnt plus linear in the distance between @a pos
-    and end of the container.
-
-    @liveexample{The example shows how `insert()` is used.,insert__count}
-
-    @since version 1.0.0
-    */
+    /// @brief inserts copies of element into array
+    /// @sa https://json.nlohmann.me/api/basic_json/insert/
     iterator insert(const_iterator pos, size_type cnt, const basic_json& val)
     {
         // insert only works for arrays
-        if (JSON_LIKELY(is_array()))
+        if (JSON_HEDLEY_LIKELY(is_array()))
         {
             // check if iterator pos fits to this JSON value
-            if (JSON_UNLIKELY(pos.m_object != this))
+            if (JSON_HEDLEY_UNLIKELY(pos.m_object != this))
             {
-                JSON_THROW(invalid_iterator::create(202, "iterator does not fit current value"));
+                JSON_THROW(invalid_iterator::create(202, "iterator does not fit current value", *this));
             }
 
             // insert to array and return iterator
             return insert_iterator(pos, cnt, val);
         }
 
-        JSON_THROW(type_error::create(309, "cannot use insert() with " + std::string(type_name())));
+        JSON_THROW(type_error::create(309, "cannot use insert() with " + std::string(type_name()), *this));
     }
 
-    /*!
-    @brief inserts elements
-
-    Inserts elements from range `[first, last)` before iterator @a pos.
-
-    @param[in] pos iterator before which the content will be inserted; may be
-    the end() iterator
-    @param[in] first begin of the range of elements to insert
-    @param[in] last end of the range of elements to insert
-
-    @throw type_error.309 if called on JSON values other than arrays; example:
-    `"cannot use insert() with string"`
-    @throw invalid_iterator.202 if @a pos is not an iterator of *this;
-    example: `"iterator does not fit current value"`
-    @throw invalid_iterator.210 if @a first and @a last do not belong to the
-    same JSON value; example: `"iterators do not fit"`
-    @throw invalid_iterator.211 if @a first or @a last are iterators into
-    container for which insert is called; example: `"passed iterators may not
-    belong to container"`
-
-    @return iterator pointing to the first element inserted, or @a pos if
-    `first==last`
-
-    @complexity Linear in `std::distance(first, last)` plus linear in the
-    distance between @a pos and end of the container.
-
-    @liveexample{The example shows how `insert()` is used.,insert__range}
-
-    @since version 1.0.0
-    */
+    /// @brief inserts range of elements into array
+    /// @sa https://json.nlohmann.me/api/basic_json/insert/
     iterator insert(const_iterator pos, const_iterator first, const_iterator last)
     {
         // insert only works for arrays
-        if (JSON_UNLIKELY(not is_array()))
+        if (JSON_HEDLEY_UNLIKELY(!is_array()))
         {
-            JSON_THROW(type_error::create(309, "cannot use insert() with " + std::string(type_name())));
+            JSON_THROW(type_error::create(309, "cannot use insert() with " + std::string(type_name()), *this));
         }
 
         // check if iterator pos fits to this JSON value
-        if (JSON_UNLIKELY(pos.m_object != this))
+        if (JSON_HEDLEY_UNLIKELY(pos.m_object != this))
         {
-            JSON_THROW(invalid_iterator::create(202, "iterator does not fit current value"));
+            JSON_THROW(invalid_iterator::create(202, "iterator does not fit current value", *this));
         }
 
         // check if range iterators belong to the same JSON object
-        if (JSON_UNLIKELY(first.m_object != last.m_object))
+        if (JSON_HEDLEY_UNLIKELY(first.m_object != last.m_object))
         {
-            JSON_THROW(invalid_iterator::create(210, "iterators do not fit"));
+            JSON_THROW(invalid_iterator::create(210, "iterators do not fit", *this));
         }
 
-        if (JSON_UNLIKELY(first.m_object == this))
+        if (JSON_HEDLEY_UNLIKELY(first.m_object == this))
         {
-            JSON_THROW(invalid_iterator::create(211, "passed iterators may not belong to container"));
+            JSON_THROW(invalid_iterator::create(211, "passed iterators may not belong to container", *this));
         }
 
         // insert to array and return iterator
         return insert_iterator(pos, first.m_it.array_iterator, last.m_it.array_iterator);
     }
 
-    /*!
-    @brief inserts elements
-
-    Inserts elements from initializer list @a ilist before iterator @a pos.
-
-    @param[in] pos iterator before which the content will be inserted; may be
-    the end() iterator
-    @param[in] ilist initializer list to insert the values from
-
-    @throw type_error.309 if called on JSON values other than arrays; example:
-    `"cannot use insert() with string"`
-    @throw invalid_iterator.202 if @a pos is not an iterator of *this;
-    example: `"iterator does not fit current value"`
-
-    @return iterator pointing to the first element inserted, or @a pos if
-    `ilist` is empty
-
-    @complexity Linear in `ilist.size()` plus linear in the distance between
-    @a pos and end of the container.
-
-    @liveexample{The example shows how `insert()` is used.,insert__ilist}
-
-    @since version 1.0.0
-    */
+    /// @brief inserts elements from initializer list into array
+    /// @sa https://json.nlohmann.me/api/basic_json/insert/
     iterator insert(const_iterator pos, initializer_list_t ilist)
     {
         // insert only works for arrays
-        if (JSON_UNLIKELY(not is_array()))
+        if (JSON_HEDLEY_UNLIKELY(!is_array()))
         {
-            JSON_THROW(type_error::create(309, "cannot use insert() with " + std::string(type_name())));
+            JSON_THROW(type_error::create(309, "cannot use insert() with " + std::string(type_name()), *this));
         }
 
         // check if iterator pos fits to this JSON value
-        if (JSON_UNLIKELY(pos.m_object != this))
+        if (JSON_HEDLEY_UNLIKELY(pos.m_object != this))
         {
-            JSON_THROW(invalid_iterator::create(202, "iterator does not fit current value"));
+            JSON_THROW(invalid_iterator::create(202, "iterator does not fit current value", *this));
         }
 
         // insert to array and return iterator
         return insert_iterator(pos, ilist.begin(), ilist.end());
     }
 
-    /*!
-    @brief inserts elements
-
-    Inserts elements from range `[first, last)`.
-
-    @param[in] first begin of the range of elements to insert
-    @param[in] last end of the range of elements to insert
-
-    @throw type_error.309 if called on JSON values other than objects; example:
-    `"cannot use insert() with string"`
-    @throw invalid_iterator.202 if iterator @a first or @a last does does not
-    point to an object; example: `"iterators first and last must point to
-    objects"`
-    @throw invalid_iterator.210 if @a first and @a last do not belong to the
-    same JSON value; example: `"iterators do not fit"`
-
-    @complexity Logarithmic: `O(N*log(size() + N))`, where `N` is the number
-    of elements to insert.
-
-    @liveexample{The example shows how `insert()` is used.,insert__range_object}
-
-    @since version 3.0.0
-    */
+    /// @brief inserts range of elements into object
+    /// @sa https://json.nlohmann.me/api/basic_json/insert/
     void insert(const_iterator first, const_iterator last)
     {
         // insert only works for objects
-        if (JSON_UNLIKELY(not is_object()))
+        if (JSON_HEDLEY_UNLIKELY(!is_object()))
         {
-            JSON_THROW(type_error::create(309, "cannot use insert() with " + std::string(type_name())));
+            JSON_THROW(type_error::create(309, "cannot use insert() with " + std::string(type_name()), *this));
         }
 
         // check if range iterators belong to the same JSON object
-        if (JSON_UNLIKELY(first.m_object != last.m_object))
+        if (JSON_HEDLEY_UNLIKELY(first.m_object != last.m_object))
         {
-            JSON_THROW(invalid_iterator::create(210, "iterators do not fit"));
+            JSON_THROW(invalid_iterator::create(210, "iterators do not fit", *this));
         }
 
         // passed iterators must belong to objects
-        if (JSON_UNLIKELY(not first.m_object->is_object()))
+        if (JSON_HEDLEY_UNLIKELY(!first.m_object->is_object()))
         {
-            JSON_THROW(invalid_iterator::create(202, "iterators first and last must point to objects"));
+            JSON_THROW(invalid_iterator::create(202, "iterators first and last must point to objects", *this));
         }
 
         m_value.object->insert(first.m_it.object_iterator, last.m_it.object_iterator);
     }
 
-    /*!
-    @brief updates a JSON object from another object, overwriting existing keys
-
-    Inserts all values from JSON object @a j and overwrites existing keys.
-
-    @param[in] j  JSON object to read values from
-
-    @throw type_error.312 if called on JSON values other than objects; example:
-    `"cannot use update() with string"`
-
-    @complexity O(N*log(size() + N)), where N is the number of elements to
-                insert.
-
-    @liveexample{The example shows how `update()` is used.,update}
-
-    @sa https://docs.python.org/3.6/library/stdtypes.html#dict.update
-
-    @since version 3.0.0
-    */
-    void update(const_reference j)
+    /// @brief updates a JSON object from another object, overwriting existing keys
+    /// @sa https://json.nlohmann.me/api/basic_json/update/
+    void update(const_reference j, bool merge_objects = false)
     {
-        // implicitly convert null value to an empty object
-        if (is_null())
-        {
-            m_type = value_t::object;
-            m_value.object = create<object_t>();
-            assert_invariant();
-        }
-
-        if (JSON_UNLIKELY(not is_object()))
-        {
-            JSON_THROW(type_error::create(312, "cannot use update() with " + std::string(type_name())));
-        }
-        if (JSON_UNLIKELY(not j.is_object()))
-        {
-            JSON_THROW(type_error::create(312, "cannot use update() with " + std::string(j.type_name())));
-        }
-
-        for (auto it = j.cbegin(); it != j.cend(); ++it)
-        {
-            m_value.object->operator[](it.key()) = it.value();
-        }
+        update(j.begin(), j.end(), merge_objects);
     }
 
-    /*!
-    @brief updates a JSON object from another object, overwriting existing keys
-
-    Inserts all values from from range `[first, last)` and overwrites existing
-    keys.
-
-    @param[in] first begin of the range of elements to insert
-    @param[in] last end of the range of elements to insert
-
-    @throw type_error.312 if called on JSON values other than objects; example:
-    `"cannot use update() with string"`
-    @throw invalid_iterator.202 if iterator @a first or @a last does does not
-    point to an object; example: `"iterators first and last must point to
-    objects"`
-    @throw invalid_iterator.210 if @a first and @a last do not belong to the
-    same JSON value; example: `"iterators do not fit"`
-
-    @complexity O(N*log(size() + N)), where N is the number of elements to
-                insert.
-
-    @liveexample{The example shows how `update()` is used__range.,update}
-
-    @sa https://docs.python.org/3.6/library/stdtypes.html#dict.update
-
-    @since version 3.0.0
-    */
-    void update(const_iterator first, const_iterator last)
+    /// @brief updates a JSON object from another object, overwriting existing keys
+    /// @sa https://json.nlohmann.me/api/basic_json/update/
+    void update(const_iterator first, const_iterator last, bool merge_objects = false)
     {
         // implicitly convert null value to an empty object
         if (is_null())
@@ -16386,155 +20320,142 @@ class basic_json
             assert_invariant();
         }
 
-        if (JSON_UNLIKELY(not is_object()))
+        if (JSON_HEDLEY_UNLIKELY(!is_object()))
         {
-            JSON_THROW(type_error::create(312, "cannot use update() with " + std::string(type_name())));
+            JSON_THROW(type_error::create(312, "cannot use update() with " + std::string(type_name()), *this));
         }
 
         // check if range iterators belong to the same JSON object
-        if (JSON_UNLIKELY(first.m_object != last.m_object))
+        if (JSON_HEDLEY_UNLIKELY(first.m_object != last.m_object))
         {
-            JSON_THROW(invalid_iterator::create(210, "iterators do not fit"));
+            JSON_THROW(invalid_iterator::create(210, "iterators do not fit", *this));
         }
 
         // passed iterators must belong to objects
-        if (JSON_UNLIKELY(not first.m_object->is_object()
-                          or not last.m_object->is_object()))
+        if (JSON_HEDLEY_UNLIKELY(!first.m_object->is_object()))
         {
-            JSON_THROW(invalid_iterator::create(202, "iterators first and last must point to objects"));
+            JSON_THROW(type_error::create(312, "cannot use update() with " + std::string(first.m_object->type_name()), *first.m_object));
         }
 
         for (auto it = first; it != last; ++it)
         {
+            if (merge_objects && it.value().is_object())
+            {
+                auto it2 = m_value.object->find(it.key());
+                if (it2 != m_value.object->end())
+                {
+                    it2->second.update(it.value(), true);
+                    continue;
+                }
+            }
             m_value.object->operator[](it.key()) = it.value();
+#if JSON_DIAGNOSTICS
+            m_value.object->operator[](it.key()).m_parent = this;
+#endif
         }
     }
 
-    /*!
-    @brief exchanges the values
-
-    Exchanges the contents of the JSON value with those of @a other. Does not
-    invoke any move, copy, or swap operations on individual elements. All
-    iterators and references remain valid. The past-the-end iterator is
-    invalidated.
-
-    @param[in,out] other JSON value to exchange the contents with
-
-    @complexity Constant.
-
-    @liveexample{The example below shows how JSON values can be swapped with
-    `swap()`.,swap__reference}
-
-    @since version 1.0.0
-    */
+    /// @brief exchanges the values
+    /// @sa https://json.nlohmann.me/api/basic_json/swap/
     void swap(reference other) noexcept (
-        std::is_nothrow_move_constructible<value_t>::value and
-        std::is_nothrow_move_assignable<value_t>::value and
-        std::is_nothrow_move_constructible<json_value>::value and
+        std::is_nothrow_move_constructible<value_t>::value&&
+        std::is_nothrow_move_assignable<value_t>::value&&
+        std::is_nothrow_move_constructible<json_value>::value&&
         std::is_nothrow_move_assignable<json_value>::value
     )
     {
         std::swap(m_type, other.m_type);
         std::swap(m_value, other.m_value);
+
+        set_parents();
+        other.set_parents();
         assert_invariant();
     }
 
-    /*!
-    @brief exchanges the values
+    /// @brief exchanges the values
+    /// @sa https://json.nlohmann.me/api/basic_json/swap/
+    friend void swap(reference left, reference right) noexcept (
+        std::is_nothrow_move_constructible<value_t>::value&&
+        std::is_nothrow_move_assignable<value_t>::value&&
+        std::is_nothrow_move_constructible<json_value>::value&&
+        std::is_nothrow_move_assignable<json_value>::value
+    )
+    {
+        left.swap(right);
+    }
 
-    Exchanges the contents of a JSON array with those of @a other. Does not
-    invoke any move, copy, or swap operations on individual elements. All
-    iterators and references remain valid. The past-the-end iterator is
-    invalidated.
-
-    @param[in,out] other array to exchange the contents with
-
-    @throw type_error.310 when JSON value is not an array; example: `"cannot
-    use swap() with string"`
-
-    @complexity Constant.
-
-    @liveexample{The example below shows how arrays can be swapped with
-    `swap()`.,swap__array_t}
-
-    @since version 1.0.0
-    */
-    void swap(array_t& other)
+    /// @brief exchanges the values
+    /// @sa https://json.nlohmann.me/api/basic_json/swap/
+    void swap(array_t& other) // NOLINT(bugprone-exception-escape)
     {
         // swap only works for arrays
-        if (JSON_LIKELY(is_array()))
+        if (JSON_HEDLEY_LIKELY(is_array()))
         {
             std::swap(*(m_value.array), other);
         }
         else
         {
-            JSON_THROW(type_error::create(310, "cannot use swap() with " + std::string(type_name())));
+            JSON_THROW(type_error::create(310, "cannot use swap() with " + std::string(type_name()), *this));
         }
     }
 
-    /*!
-    @brief exchanges the values
-
-    Exchanges the contents of a JSON object with those of @a other. Does not
-    invoke any move, copy, or swap operations on individual elements. All
-    iterators and references remain valid. The past-the-end iterator is
-    invalidated.
-
-    @param[in,out] other object to exchange the contents with
-
-    @throw type_error.310 when JSON value is not an object; example:
-    `"cannot use swap() with string"`
-
-    @complexity Constant.
-
-    @liveexample{The example below shows how objects can be swapped with
-    `swap()`.,swap__object_t}
-
-    @since version 1.0.0
-    */
-    void swap(object_t& other)
+    /// @brief exchanges the values
+    /// @sa https://json.nlohmann.me/api/basic_json/swap/
+    void swap(object_t& other) // NOLINT(bugprone-exception-escape)
     {
         // swap only works for objects
-        if (JSON_LIKELY(is_object()))
+        if (JSON_HEDLEY_LIKELY(is_object()))
         {
             std::swap(*(m_value.object), other);
         }
         else
         {
-            JSON_THROW(type_error::create(310, "cannot use swap() with " + std::string(type_name())));
+            JSON_THROW(type_error::create(310, "cannot use swap() with " + std::string(type_name()), *this));
         }
     }
 
-    /*!
-    @brief exchanges the values
-
-    Exchanges the contents of a JSON string with those of @a other. Does not
-    invoke any move, copy, or swap operations on individual elements. All
-    iterators and references remain valid. The past-the-end iterator is
-    invalidated.
-
-    @param[in,out] other string to exchange the contents with
-
-    @throw type_error.310 when JSON value is not a string; example: `"cannot
-    use swap() with boolean"`
-
-    @complexity Constant.
-
-    @liveexample{The example below shows how strings can be swapped with
-    `swap()`.,swap__string_t}
-
-    @since version 1.0.0
-    */
-    void swap(string_t& other)
+    /// @brief exchanges the values
+    /// @sa https://json.nlohmann.me/api/basic_json/swap/
+    void swap(string_t& other) // NOLINT(bugprone-exception-escape)
     {
         // swap only works for strings
-        if (JSON_LIKELY(is_string()))
+        if (JSON_HEDLEY_LIKELY(is_string()))
         {
             std::swap(*(m_value.string), other);
         }
         else
         {
-            JSON_THROW(type_error::create(310, "cannot use swap() with " + std::string(type_name())));
+            JSON_THROW(type_error::create(310, "cannot use swap() with " + std::string(type_name()), *this));
+        }
+    }
+
+    /// @brief exchanges the values
+    /// @sa https://json.nlohmann.me/api/basic_json/swap/
+    void swap(binary_t& other) // NOLINT(bugprone-exception-escape)
+    {
+        // swap only works for strings
+        if (JSON_HEDLEY_LIKELY(is_binary()))
+        {
+            std::swap(*(m_value.binary), other);
+        }
+        else
+        {
+            JSON_THROW(type_error::create(310, "cannot use swap() with " + std::string(type_name()), *this));
+        }
+    }
+
+    /// @brief exchanges the values
+    /// @sa https://json.nlohmann.me/api/basic_json/swap/
+    void swap(typename binary_t::container_type& other) // NOLINT(bugprone-exception-escape)
+    {
+        // swap only works for strings
+        if (JSON_HEDLEY_LIKELY(is_binary()))
+        {
+            std::swap(*(m_value.binary), other);
+        }
+        else
+        {
+            JSON_THROW(type_error::create(310, "cannot use swap() with " + std::string(type_name()), *this));
         }
     }
 
@@ -16548,47 +20469,14 @@ class basic_json
     /// @name lexicographical comparison operators
     /// @{
 
-    /*!
-    @brief comparison: equal
-
-    Compares two JSON values for equality according to the following rules:
-    - Two JSON values are equal if (1) they are from the same type and (2)
-      their stored values are the same according to their respective
-      `operator==`.
-    - Integer and floating-point numbers are automatically converted before
-      comparison. Note than two NaN values are always treated as unequal.
-    - Two JSON null values are equal.
-
-    @note Floating-point inside JSON values numbers are compared with
-    `json::number_float_t::operator==` which is `double::operator==` by
-    default. To compare floating-point while respecting an epsilon, an alternative
-    [comparison function](https://github.com/mariokonrad/marnav/blob/master/src/marnav/math/floatingpoint.hpp#L34-#L39)
-    could be used, for instance
-    @code {.cpp}
-    template<typename T, typename = typename std::enable_if<std::is_floating_point<T>::value, T>::type>
-    inline bool is_same(T a, T b, T epsilon = std::numeric_limits<T>::epsilon()) noexcept
-    {
-        return std::abs(a - b) <= epsilon;
-    }
-    @endcode
-
-    @note NaN values never compare equal to themselves or to other NaN values.
-
-    @param[in] lhs  first JSON value to consider
-    @param[in] rhs  second JSON value to consider
-    @return whether the values @a lhs and @a rhs are equal
-
-    @exceptionsafety No-throw guarantee: this function never throws exceptions.
-
-    @complexity Linear.
-
-    @liveexample{The example demonstrates comparing several JSON
-    types.,operator__equal}
-
-    @since version 1.0.0
-    */
+    /// @brief comparison: equal
+    /// @sa https://json.nlohmann.me/api/basic_json/operator_eq/
     friend bool operator==(const_reference lhs, const_reference rhs) noexcept
     {
+#ifdef __GNUC__
+#pragma GCC diagnostic push
+#pragma GCC diagnostic ignored "-Wfloat-equal"
+#endif
         const auto lhs_type = lhs.type();
         const auto rhs_type = rhs.type();
 
@@ -16597,154 +20485,113 @@ class basic_json
             switch (lhs_type)
             {
                 case value_t::array:
-                    return (*lhs.m_value.array == *rhs.m_value.array);
+                    return *lhs.m_value.array == *rhs.m_value.array;
 
                 case value_t::object:
-                    return (*lhs.m_value.object == *rhs.m_value.object);
+                    return *lhs.m_value.object == *rhs.m_value.object;
 
                 case value_t::null:
                     return true;
 
                 case value_t::string:
-                    return (*lhs.m_value.string == *rhs.m_value.string);
+                    return *lhs.m_value.string == *rhs.m_value.string;
 
                 case value_t::boolean:
-                    return (lhs.m_value.boolean == rhs.m_value.boolean);
+                    return lhs.m_value.boolean == rhs.m_value.boolean;
 
                 case value_t::number_integer:
-                    return (lhs.m_value.number_integer == rhs.m_value.number_integer);
+                    return lhs.m_value.number_integer == rhs.m_value.number_integer;
 
                 case value_t::number_unsigned:
-                    return (lhs.m_value.number_unsigned == rhs.m_value.number_unsigned);
+                    return lhs.m_value.number_unsigned == rhs.m_value.number_unsigned;
 
                 case value_t::number_float:
-                    return (lhs.m_value.number_float == rhs.m_value.number_float);
+                    return lhs.m_value.number_float == rhs.m_value.number_float;
 
+                case value_t::binary:
+                    return *lhs.m_value.binary == *rhs.m_value.binary;
+
+                case value_t::discarded:
                 default:
                     return false;
             }
         }
-        else if (lhs_type == value_t::number_integer and rhs_type == value_t::number_float)
+        else if (lhs_type == value_t::number_integer && rhs_type == value_t::number_float)
         {
-            return (static_cast<number_float_t>(lhs.m_value.number_integer) == rhs.m_value.number_float);
+            return static_cast<number_float_t>(lhs.m_value.number_integer) == rhs.m_value.number_float;
         }
-        else if (lhs_type == value_t::number_float and rhs_type == value_t::number_integer)
+        else if (lhs_type == value_t::number_float && rhs_type == value_t::number_integer)
         {
-            return (lhs.m_value.number_float == static_cast<number_float_t>(rhs.m_value.number_integer));
+            return lhs.m_value.number_float == static_cast<number_float_t>(rhs.m_value.number_integer);
         }
-        else if (lhs_type == value_t::number_unsigned and rhs_type == value_t::number_float)
+        else if (lhs_type == value_t::number_unsigned && rhs_type == value_t::number_float)
         {
-            return (static_cast<number_float_t>(lhs.m_value.number_unsigned) == rhs.m_value.number_float);
+            return static_cast<number_float_t>(lhs.m_value.number_unsigned) == rhs.m_value.number_float;
         }
-        else if (lhs_type == value_t::number_float and rhs_type == value_t::number_unsigned)
+        else if (lhs_type == value_t::number_float && rhs_type == value_t::number_unsigned)
         {
-            return (lhs.m_value.number_float == static_cast<number_float_t>(rhs.m_value.number_unsigned));
+            return lhs.m_value.number_float == static_cast<number_float_t>(rhs.m_value.number_unsigned);
         }
-        else if (lhs_type == value_t::number_unsigned and rhs_type == value_t::number_integer)
+        else if (lhs_type == value_t::number_unsigned && rhs_type == value_t::number_integer)
         {
-            return (static_cast<number_integer_t>(lhs.m_value.number_unsigned) == rhs.m_value.number_integer);
+            return static_cast<number_integer_t>(lhs.m_value.number_unsigned) == rhs.m_value.number_integer;
         }
-        else if (lhs_type == value_t::number_integer and rhs_type == value_t::number_unsigned)
+        else if (lhs_type == value_t::number_integer && rhs_type == value_t::number_unsigned)
         {
-            return (lhs.m_value.number_integer == static_cast<number_integer_t>(rhs.m_value.number_unsigned));
+            return lhs.m_value.number_integer == static_cast<number_integer_t>(rhs.m_value.number_unsigned);
         }
 
         return false;
+#ifdef __GNUC__
+#pragma GCC diagnostic pop
+#endif
     }
 
-    /*!
-    @brief comparison: equal
-    @copydoc operator==(const_reference, const_reference)
-    */
+    /// @brief comparison: equal
+    /// @sa https://json.nlohmann.me/api/basic_json/operator_eq/
     template<typename ScalarType, typename std::enable_if<
                  std::is_scalar<ScalarType>::value, int>::type = 0>
-    friend bool operator==(const_reference lhs, const ScalarType rhs) noexcept
+    friend bool operator==(const_reference lhs, ScalarType rhs) noexcept
     {
-        return (lhs == basic_json(rhs));
+        return lhs == basic_json(rhs);
     }
 
-    /*!
-    @brief comparison: equal
-    @copydoc operator==(const_reference, const_reference)
-    */
+    /// @brief comparison: equal
+    /// @sa https://json.nlohmann.me/api/basic_json/operator_eq/
     template<typename ScalarType, typename std::enable_if<
                  std::is_scalar<ScalarType>::value, int>::type = 0>
-    friend bool operator==(const ScalarType lhs, const_reference rhs) noexcept
+    friend bool operator==(ScalarType lhs, const_reference rhs) noexcept
     {
-        return (basic_json(lhs) == rhs);
+        return basic_json(lhs) == rhs;
     }
 
-    /*!
-    @brief comparison: not equal
-
-    Compares two JSON values for inequality by calculating `not (lhs == rhs)`.
-
-    @param[in] lhs  first JSON value to consider
-    @param[in] rhs  second JSON value to consider
-    @return whether the values @a lhs and @a rhs are not equal
-
-    @complexity Linear.
-
-    @exceptionsafety No-throw guarantee: this function never throws exceptions.
-
-    @liveexample{The example demonstrates comparing several JSON
-    types.,operator__notequal}
-
-    @since version 1.0.0
-    */
+    /// @brief comparison: not equal
+    /// @sa https://json.nlohmann.me/api/basic_json/operator_ne/
     friend bool operator!=(const_reference lhs, const_reference rhs) noexcept
     {
-        return not (lhs == rhs);
+        return !(lhs == rhs);
     }
 
-    /*!
-    @brief comparison: not equal
-    @copydoc operator!=(const_reference, const_reference)
-    */
+    /// @brief comparison: not equal
+    /// @sa https://json.nlohmann.me/api/basic_json/operator_ne/
     template<typename ScalarType, typename std::enable_if<
                  std::is_scalar<ScalarType>::value, int>::type = 0>
-    friend bool operator!=(const_reference lhs, const ScalarType rhs) noexcept
+    friend bool operator!=(const_reference lhs, ScalarType rhs) noexcept
     {
-        return (lhs != basic_json(rhs));
+        return lhs != basic_json(rhs);
     }
 
-    /*!
-    @brief comparison: not equal
-    @copydoc operator!=(const_reference, const_reference)
-    */
+    /// @brief comparison: not equal
+    /// @sa https://json.nlohmann.me/api/basic_json/operator_ne/
     template<typename ScalarType, typename std::enable_if<
                  std::is_scalar<ScalarType>::value, int>::type = 0>
-    friend bool operator!=(const ScalarType lhs, const_reference rhs) noexcept
+    friend bool operator!=(ScalarType lhs, const_reference rhs) noexcept
     {
-        return (basic_json(lhs) != rhs);
+        return basic_json(lhs) != rhs;
     }
 
-    /*!
-    @brief comparison: less than
-
-    Compares whether one JSON value @a lhs is less than another JSON value @a
-    rhs according to the following rules:
-    - If @a lhs and @a rhs have the same type, the values are compared using
-      the default `<` operator.
-    - Integer and floating-point numbers are automatically converted before
-      comparison
-    - In case @a lhs and @a rhs have different types, the values are ignored
-      and the order of the types is considered, see
-      @ref operator<(const value_t, const value_t).
-
-    @param[in] lhs  first JSON value to consider
-    @param[in] rhs  second JSON value to consider
-    @return whether @a lhs is less than @a rhs
-
-    @complexity Linear.
-
-    @exceptionsafety No-throw guarantee: this function never throws exceptions.
-
-    @liveexample{The example demonstrates comparing several JSON
-    types.,operator__less}
-
-    @since version 1.0.0
-    */
+    /// @brief comparison: less than
+    /// @sa https://json.nlohmann.me/api/basic_json/operator_lt/
     friend bool operator<(const_reference lhs, const_reference rhs) noexcept
     {
         const auto lhs_type = lhs.type();
@@ -16755,54 +20602,60 @@ class basic_json
             switch (lhs_type)
             {
                 case value_t::array:
+                    // note parentheses are necessary, see
+                    // https://github.com/nlohmann/json/issues/1530
                     return (*lhs.m_value.array) < (*rhs.m_value.array);
 
                 case value_t::object:
-                    return *lhs.m_value.object < *rhs.m_value.object;
+                    return (*lhs.m_value.object) < (*rhs.m_value.object);
 
                 case value_t::null:
                     return false;
 
                 case value_t::string:
-                    return *lhs.m_value.string < *rhs.m_value.string;
+                    return (*lhs.m_value.string) < (*rhs.m_value.string);
 
                 case value_t::boolean:
-                    return lhs.m_value.boolean < rhs.m_value.boolean;
+                    return (lhs.m_value.boolean) < (rhs.m_value.boolean);
 
                 case value_t::number_integer:
-                    return lhs.m_value.number_integer < rhs.m_value.number_integer;
+                    return (lhs.m_value.number_integer) < (rhs.m_value.number_integer);
 
                 case value_t::number_unsigned:
-                    return lhs.m_value.number_unsigned < rhs.m_value.number_unsigned;
+                    return (lhs.m_value.number_unsigned) < (rhs.m_value.number_unsigned);
 
                 case value_t::number_float:
-                    return lhs.m_value.number_float < rhs.m_value.number_float;
+                    return (lhs.m_value.number_float) < (rhs.m_value.number_float);
 
+                case value_t::binary:
+                    return (*lhs.m_value.binary) < (*rhs.m_value.binary);
+
+                case value_t::discarded:
                 default:
                     return false;
             }
         }
-        else if (lhs_type == value_t::number_integer and rhs_type == value_t::number_float)
+        else if (lhs_type == value_t::number_integer && rhs_type == value_t::number_float)
         {
             return static_cast<number_float_t>(lhs.m_value.number_integer) < rhs.m_value.number_float;
         }
-        else if (lhs_type == value_t::number_float and rhs_type == value_t::number_integer)
+        else if (lhs_type == value_t::number_float && rhs_type == value_t::number_integer)
         {
             return lhs.m_value.number_float < static_cast<number_float_t>(rhs.m_value.number_integer);
         }
-        else if (lhs_type == value_t::number_unsigned and rhs_type == value_t::number_float)
+        else if (lhs_type == value_t::number_unsigned && rhs_type == value_t::number_float)
         {
             return static_cast<number_float_t>(lhs.m_value.number_unsigned) < rhs.m_value.number_float;
         }
-        else if (lhs_type == value_t::number_float and rhs_type == value_t::number_unsigned)
+        else if (lhs_type == value_t::number_float && rhs_type == value_t::number_unsigned)
         {
             return lhs.m_value.number_float < static_cast<number_float_t>(rhs.m_value.number_unsigned);
         }
-        else if (lhs_type == value_t::number_integer and rhs_type == value_t::number_unsigned)
+        else if (lhs_type == value_t::number_integer && rhs_type == value_t::number_unsigned)
         {
             return lhs.m_value.number_integer < static_cast<number_integer_t>(rhs.m_value.number_unsigned);
         }
-        else if (lhs_type == value_t::number_unsigned and rhs_type == value_t::number_integer)
+        else if (lhs_type == value_t::number_unsigned && rhs_type == value_t::number_integer)
         {
             return static_cast<number_integer_t>(lhs.m_value.number_unsigned) < rhs.m_value.number_integer;
         }
@@ -16813,164 +20666,97 @@ class basic_json
         return operator<(lhs_type, rhs_type);
     }
 
-    /*!
-    @brief comparison: less than
-    @copydoc operator<(const_reference, const_reference)
-    */
+    /// @brief comparison: less than
+    /// @sa https://json.nlohmann.me/api/basic_json/operator_lt/
     template<typename ScalarType, typename std::enable_if<
                  std::is_scalar<ScalarType>::value, int>::type = 0>
-    friend bool operator<(const_reference lhs, const ScalarType rhs) noexcept
+    friend bool operator<(const_reference lhs, ScalarType rhs) noexcept
     {
-        return (lhs < basic_json(rhs));
+        return lhs < basic_json(rhs);
     }
 
-    /*!
-    @brief comparison: less than
-    @copydoc operator<(const_reference, const_reference)
-    */
+    /// @brief comparison: less than
+    /// @sa https://json.nlohmann.me/api/basic_json/operator_lt/
     template<typename ScalarType, typename std::enable_if<
                  std::is_scalar<ScalarType>::value, int>::type = 0>
-    friend bool operator<(const ScalarType lhs, const_reference rhs) noexcept
+    friend bool operator<(ScalarType lhs, const_reference rhs) noexcept
     {
-        return (basic_json(lhs) < rhs);
+        return basic_json(lhs) < rhs;
     }
 
-    /*!
-    @brief comparison: less than or equal
-
-    Compares whether one JSON value @a lhs is less than or equal to another
-    JSON value by calculating `not (rhs < lhs)`.
-
-    @param[in] lhs  first JSON value to consider
-    @param[in] rhs  second JSON value to consider
-    @return whether @a lhs is less than or equal to @a rhs
-
-    @complexity Linear.
-
-    @exceptionsafety No-throw guarantee: this function never throws exceptions.
-
-    @liveexample{The example demonstrates comparing several JSON
-    types.,operator__greater}
-
-    @since version 1.0.0
-    */
+    /// @brief comparison: less than or equal
+    /// @sa https://json.nlohmann.me/api/basic_json/operator_le/
     friend bool operator<=(const_reference lhs, const_reference rhs) noexcept
     {
-        return not (rhs < lhs);
+        return !(rhs < lhs);
     }
 
-    /*!
-    @brief comparison: less than or equal
-    @copydoc operator<=(const_reference, const_reference)
-    */
+    /// @brief comparison: less than or equal
+    /// @sa https://json.nlohmann.me/api/basic_json/operator_le/
     template<typename ScalarType, typename std::enable_if<
                  std::is_scalar<ScalarType>::value, int>::type = 0>
-    friend bool operator<=(const_reference lhs, const ScalarType rhs) noexcept
+    friend bool operator<=(const_reference lhs, ScalarType rhs) noexcept
     {
-        return (lhs <= basic_json(rhs));
+        return lhs <= basic_json(rhs);
     }
 
-    /*!
-    @brief comparison: less than or equal
-    @copydoc operator<=(const_reference, const_reference)
-    */
+    /// @brief comparison: less than or equal
+    /// @sa https://json.nlohmann.me/api/basic_json/operator_le/
     template<typename ScalarType, typename std::enable_if<
                  std::is_scalar<ScalarType>::value, int>::type = 0>
-    friend bool operator<=(const ScalarType lhs, const_reference rhs) noexcept
+    friend bool operator<=(ScalarType lhs, const_reference rhs) noexcept
     {
-        return (basic_json(lhs) <= rhs);
+        return basic_json(lhs) <= rhs;
     }
 
-    /*!
-    @brief comparison: greater than
-
-    Compares whether one JSON value @a lhs is greater than another
-    JSON value by calculating `not (lhs <= rhs)`.
-
-    @param[in] lhs  first JSON value to consider
-    @param[in] rhs  second JSON value to consider
-    @return whether @a lhs is greater than to @a rhs
-
-    @complexity Linear.
-
-    @exceptionsafety No-throw guarantee: this function never throws exceptions.
-
-    @liveexample{The example demonstrates comparing several JSON
-    types.,operator__lessequal}
-
-    @since version 1.0.0
-    */
+    /// @brief comparison: greater than
+    /// @sa https://json.nlohmann.me/api/basic_json/operator_gt/
     friend bool operator>(const_reference lhs, const_reference rhs) noexcept
     {
-        return not (lhs <= rhs);
+        return !(lhs <= rhs);
     }
 
-    /*!
-    @brief comparison: greater than
-    @copydoc operator>(const_reference, const_reference)
-    */
+    /// @brief comparison: greater than
+    /// @sa https://json.nlohmann.me/api/basic_json/operator_gt/
     template<typename ScalarType, typename std::enable_if<
                  std::is_scalar<ScalarType>::value, int>::type = 0>
-    friend bool operator>(const_reference lhs, const ScalarType rhs) noexcept
+    friend bool operator>(const_reference lhs, ScalarType rhs) noexcept
     {
-        return (lhs > basic_json(rhs));
+        return lhs > basic_json(rhs);
     }
 
-    /*!
-    @brief comparison: greater than
-    @copydoc operator>(const_reference, const_reference)
-    */
+    /// @brief comparison: greater than
+    /// @sa https://json.nlohmann.me/api/basic_json/operator_gt/
     template<typename ScalarType, typename std::enable_if<
                  std::is_scalar<ScalarType>::value, int>::type = 0>
-    friend bool operator>(const ScalarType lhs, const_reference rhs) noexcept
+    friend bool operator>(ScalarType lhs, const_reference rhs) noexcept
     {
-        return (basic_json(lhs) > rhs);
+        return basic_json(lhs) > rhs;
     }
 
-    /*!
-    @brief comparison: greater than or equal
-
-    Compares whether one JSON value @a lhs is greater than or equal to another
-    JSON value by calculating `not (lhs < rhs)`.
-
-    @param[in] lhs  first JSON value to consider
-    @param[in] rhs  second JSON value to consider
-    @return whether @a lhs is greater than or equal to @a rhs
-
-    @complexity Linear.
-
-    @exceptionsafety No-throw guarantee: this function never throws exceptions.
-
-    @liveexample{The example demonstrates comparing several JSON
-    types.,operator__greaterequal}
-
-    @since version 1.0.0
-    */
+    /// @brief comparison: greater than or equal
+    /// @sa https://json.nlohmann.me/api/basic_json/operator_ge/
     friend bool operator>=(const_reference lhs, const_reference rhs) noexcept
     {
-        return not (lhs < rhs);
+        return !(lhs < rhs);
     }
 
-    /*!
-    @brief comparison: greater than or equal
-    @copydoc operator>=(const_reference, const_reference)
-    */
+    /// @brief comparison: greater than or equal
+    /// @sa https://json.nlohmann.me/api/basic_json/operator_ge/
     template<typename ScalarType, typename std::enable_if<
                  std::is_scalar<ScalarType>::value, int>::type = 0>
-    friend bool operator>=(const_reference lhs, const ScalarType rhs) noexcept
+    friend bool operator>=(const_reference lhs, ScalarType rhs) noexcept
     {
-        return (lhs >= basic_json(rhs));
+        return lhs >= basic_json(rhs);
     }
 
-    /*!
-    @brief comparison: greater than or equal
-    @copydoc operator>=(const_reference, const_reference)
-    */
+    /// @brief comparison: greater than or equal
+    /// @sa https://json.nlohmann.me/api/basic_json/operator_ge/
     template<typename ScalarType, typename std::enable_if<
                  std::is_scalar<ScalarType>::value, int>::type = 0>
-    friend bool operator>=(const ScalarType lhs, const_reference rhs) noexcept
+    friend bool operator>=(ScalarType lhs, const_reference rhs) noexcept
     {
-        return (basic_json(lhs) >= rhs);
+        return basic_json(lhs) >= rhs;
     }
 
     /// @}
@@ -16981,43 +20767,14 @@ class basic_json
 
     /// @name serialization
     /// @{
-
-    /*!
-    @brief serialize to stream
-
-    Serialize the given JSON value @a j to the output stream @a o. The JSON
-    value will be serialized using the @ref dump member function.
-
-    - The indentation of the output can be controlled with the member variable
-      `width` of the output stream @a o. For instance, using the manipulator
-      `std::setw(4)` on @a o sets the indentation level to `4` and the
-      serialization result is the same as calling `dump(4)`.
-
-    - The indentation character can be controlled with the member variable
-      `fill` of the output stream @a o. For instance, the manipulator
-      `std::setfill('\\t')` sets indentation to use a tab character rather than
-      the default space character.
-
-    @param[in,out] o  stream to serialize to
-    @param[in] j  JSON value to serialize
-
-    @return the stream @a o
-
-    @throw type_error.316 if a string stored inside the JSON value is not
-                          UTF-8 encoded
-
-    @complexity Linear.
-
-    @liveexample{The example below shows the serialization with different
-    parameters to `width` to adjust the indentation level.,operator_serialize}
-
-    @since version 1.0.0; indentation character added in version 3.0.0
-    */
+#ifndef JSON_NO_IO
+    /// @brief serialize to stream
+    /// @sa https://json.nlohmann.me/api/basic_json/operator_ltlt/
     friend std::ostream& operator<<(std::ostream& o, const basic_json& j)
     {
         // read width member and use it as indentation parameter if nonzero
-        const bool pretty_print = (o.width() > 0);
-        const auto indentation = (pretty_print ? o.width() : 0);
+        const bool pretty_print = o.width() > 0;
+        const auto indentation = pretty_print ? o.width() : 0;
 
         // reset width to 0 for subsequent calls to this stream
         o.width(0);
@@ -17028,20 +20785,18 @@ class basic_json
         return o;
     }
 
-    /*!
-    @brief serialize to stream
-    @deprecated This stream operator is deprecated and will be removed in
-                future 4.0.0 of the library. Please use
-                @ref operator<<(std::ostream&, const basic_json&)
-                instead; that is, replace calls like `j >> o;` with `o << j;`.
-    @since version 1.0.0; deprecated since version 3.0.0
-    */
-    JSON_DEPRECATED
+    /// @brief serialize to stream
+    /// @sa https://json.nlohmann.me/api/basic_json/operator_ltlt/
+    /// @deprecated This function is deprecated since 3.0.0 and will be removed in
+    ///             version 4.0.0 of the library. Please use
+    ///             operator<<(std::ostream&, const basic_json&) instead; that is,
+    ///             replace calls like `j >> o;` with `o << j;`.
+    JSON_HEDLEY_DEPRECATED_FOR(3.0.0, operator<<(std::ostream&, const basic_json&))
     friend std::ostream& operator>>(const basic_json& j, std::ostream& o)
     {
         return o << j;
     }
-
+#endif  // JSON_NO_IO
     /// @}
 
 
@@ -17052,337 +20807,181 @@ class basic_json
     /// @name deserialization
     /// @{
 
-    /*!
-    @brief deserialize from a compatible input
-
-    This function reads from a compatible input. Examples are:
-    - an array of 1-byte values
-    - strings with character/literal type with size of 1 byte
-    - input streams
-    - container with contiguous storage of 1-byte values. Compatible container
-      types include `std::vector`, `std::string`, `std::array`,
-      `std::valarray`, and `std::initializer_list`. Furthermore, C-style
-      arrays can be used with `std::begin()`/`std::end()`. User-defined
-      containers can be used as long as they implement random-access iterators
-      and a contiguous storage.
-
-    @pre Each element of the container has a size of 1 byte. Violating this
-    precondition yields undefined behavior. **This precondition is enforced
-    with a static assertion.**
-
-    @pre The container storage is contiguous. Violating this precondition
-    yields undefined behavior. **This precondition is enforced with an
-    assertion.**
-    @pre Each element of the container has a size of 1 byte. Violating this
-    precondition yields undefined behavior. **This precondition is enforced
-    with a static assertion.**
-
-    @warning There is no way to enforce all preconditions at compile-time. If
-             the function is called with a noncompliant container and with
-             assertions switched off, the behavior is undefined and will most
-             likely yield segmentation violation.
-
-    @param[in] i  input to read from
-    @param[in] cb  a parser callback function of type @ref parser_callback_t
-    which is used to control the deserialization by filtering unwanted values
-    (optional)
-
-    @return result of the deserialization
-
-    @throw parse_error.101 if a parse error occurs; example: `""unexpected end
-    of input; expected string literal""`
-    @throw parse_error.102 if to_unicode fails or surrogate error
-    @throw parse_error.103 if to_unicode fails
-
-    @complexity Linear in the length of the input. The parser is a predictive
-    LL(1) parser. The complexity can be higher if the parser callback function
-    @a cb has a super-linear complexity.
-
-    @note A UTF-8 byte order mark is silently ignored.
-
-    @liveexample{The example below demonstrates the `parse()` function reading
-    from an array.,parse__array__parser_callback_t}
-
-    @liveexample{The example below demonstrates the `parse()` function with
-    and without callback function.,parse__string__parser_callback_t}
-
-    @liveexample{The example below demonstrates the `parse()` function with
-    and without callback function.,parse__istream__parser_callback_t}
-
-    @liveexample{The example below demonstrates the `parse()` function reading
-    from a contiguous container.,parse__contiguouscontainer__parser_callback_t}
-
-    @since version 2.0.3 (contiguous containers)
-    */
-    static basic_json parse(detail::input_adapter&& i,
+    /// @brief deserialize from a compatible input
+    /// @sa https://json.nlohmann.me/api/basic_json/parse/
+    template<typename InputType>
+    JSON_HEDLEY_WARN_UNUSED_RESULT
+    static basic_json parse(InputType&& i,
                             const parser_callback_t cb = nullptr,
-                            const bool allow_exceptions = true)
+                            const bool allow_exceptions = true,
+                            const bool ignore_comments = false)
     {
         basic_json result;
-        parser(i, cb, allow_exceptions).parse(true, result);
+        parser(detail::input_adapter(std::forward<InputType>(i)), cb, allow_exceptions, ignore_comments).parse(true, result);
         return result;
     }
 
-    static bool accept(detail::input_adapter&& i)
+    /// @brief deserialize from a pair of character iterators
+    /// @sa https://json.nlohmann.me/api/basic_json/parse/
+    template<typename IteratorType>
+    JSON_HEDLEY_WARN_UNUSED_RESULT
+    static basic_json parse(IteratorType first,
+                            IteratorType last,
+                            const parser_callback_t cb = nullptr,
+                            const bool allow_exceptions = true,
+                            const bool ignore_comments = false)
     {
-        return parser(i).accept(true);
+        basic_json result;
+        parser(detail::input_adapter(std::move(first), std::move(last)), cb, allow_exceptions, ignore_comments).parse(true, result);
+        return result;
     }
 
-    /*!
-    @brief generate SAX events
+    JSON_HEDLEY_WARN_UNUSED_RESULT
+    JSON_HEDLEY_DEPRECATED_FOR(3.8.0, parse(ptr, ptr + len))
+    static basic_json parse(detail::span_input_adapter&& i,
+                            const parser_callback_t cb = nullptr,
+                            const bool allow_exceptions = true,
+                            const bool ignore_comments = false)
+    {
+        basic_json result;
+        parser(i.get(), cb, allow_exceptions, ignore_comments).parse(true, result);
+        return result;
+    }
 
-    The SAX event lister must follow the interface of @ref json_sax.
+    /// @brief check if the input is valid JSON
+    /// @sa https://json.nlohmann.me/api/basic_json/accept/
+    template<typename InputType>
+    static bool accept(InputType&& i,
+                       const bool ignore_comments = false)
+    {
+        return parser(detail::input_adapter(std::forward<InputType>(i)), nullptr, false, ignore_comments).accept(true);
+    }
 
-    This function reads from a compatible input. Examples are:
-    - an array of 1-byte values
-    - strings with character/literal type with size of 1 byte
-    - input streams
-    - container with contiguous storage of 1-byte values. Compatible container
-      types include `std::vector`, `std::string`, `std::array`,
-      `std::valarray`, and `std::initializer_list`. Furthermore, C-style
-      arrays can be used with `std::begin()`/`std::end()`. User-defined
-      containers can be used as long as they implement random-access iterators
-      and a contiguous storage.
+    /// @brief check if the input is valid JSON
+    /// @sa https://json.nlohmann.me/api/basic_json/accept/
+    template<typename IteratorType>
+    static bool accept(IteratorType first, IteratorType last,
+                       const bool ignore_comments = false)
+    {
+        return parser(detail::input_adapter(std::move(first), std::move(last)), nullptr, false, ignore_comments).accept(true);
+    }
 
-    @pre Each element of the container has a size of 1 byte. Violating this
-    precondition yields undefined behavior. **This precondition is enforced
-    with a static assertion.**
+    JSON_HEDLEY_WARN_UNUSED_RESULT
+    JSON_HEDLEY_DEPRECATED_FOR(3.8.0, accept(ptr, ptr + len))
+    static bool accept(detail::span_input_adapter&& i,
+                       const bool ignore_comments = false)
+    {
+        return parser(i.get(), nullptr, false, ignore_comments).accept(true);
+    }
 
-    @pre The container storage is contiguous. Violating this precondition
-    yields undefined behavior. **This precondition is enforced with an
-    assertion.**
-    @pre Each element of the container has a size of 1 byte. Violating this
-    precondition yields undefined behavior. **This precondition is enforced
-    with a static assertion.**
-
-    @warning There is no way to enforce all preconditions at compile-time. If
-             the function is called with a noncompliant container and with
-             assertions switched off, the behavior is undefined and will most
-             likely yield segmentation violation.
-
-    @param[in] i  input to read from
-    @param[in,out] sax  SAX event listener
-    @param[in] format  the format to parse (JSON, CBOR, MessagePack, or UBJSON)
-    @param[in] strict  whether the input has to be consumed completely
-
-    @return return value of the last processed SAX event
-
-    @throw parse_error.101 if a parse error occurs; example: `""unexpected end
-    of input; expected string literal""`
-    @throw parse_error.102 if to_unicode fails or surrogate error
-    @throw parse_error.103 if to_unicode fails
-
-    @complexity Linear in the length of the input. The parser is a predictive
-    LL(1) parser. The complexity can be higher if the SAX consumer @a sax has
-    a super-linear complexity.
-
-    @note A UTF-8 byte order mark is silently ignored.
-
-    @liveexample{The example below demonstrates the `sax_parse()` function
-    reading from string and processing the events with a user-defined SAX
-    event consumer.,sax_parse}
-
-    @since version 3.2.0
-    */
-    template <typename SAX>
-    static bool sax_parse(detail::input_adapter&& i, SAX* sax,
+    /// @brief generate SAX events
+    /// @sa https://json.nlohmann.me/api/basic_json/sax_parse/
+    template <typename InputType, typename SAX>
+    JSON_HEDLEY_NON_NULL(2)
+    static bool sax_parse(InputType&& i, SAX* sax,
                           input_format_t format = input_format_t::json,
-                          const bool strict = true)
+                          const bool strict = true,
+                          const bool ignore_comments = false)
     {
-        assert(sax);
-        switch (format)
-        {
-            case input_format_t::json:
-                return parser(std::move(i)).sax_parse(sax, strict);
-            default:
-                return detail::binary_reader<basic_json, SAX>(std::move(i)).sax_parse(format, sax, strict);
-        }
+        auto ia = detail::input_adapter(std::forward<InputType>(i));
+        return format == input_format_t::json
+               ? parser(std::move(ia), nullptr, true, ignore_comments).sax_parse(sax, strict)
+               : detail::binary_reader<basic_json, decltype(ia), SAX>(std::move(ia)).sax_parse(format, sax, strict);
     }
 
-    /*!
-    @brief deserialize from an iterator range with contiguous storage
-
-    This function reads from an iterator range of a container with contiguous
-    storage of 1-byte values. Compatible container types include
-    `std::vector`, `std::string`, `std::array`, `std::valarray`, and
-    `std::initializer_list`. Furthermore, C-style arrays can be used with
-    `std::begin()`/`std::end()`. User-defined containers can be used as long
-    as they implement random-access iterators and a contiguous storage.
-
-    @pre The iterator range is contiguous. Violating this precondition yields
-    undefined behavior. **This precondition is enforced with an assertion.**
-    @pre Each element in the range has a size of 1 byte. Violating this
-    precondition yields undefined behavior. **This precondition is enforced
-    with a static assertion.**
-
-    @warning There is no way to enforce all preconditions at compile-time. If
-             the function is called with noncompliant iterators and with
-             assertions switched off, the behavior is undefined and will most
-             likely yield segmentation violation.
-
-    @tparam IteratorType iterator of container with contiguous storage
-    @param[in] first  begin of the range to parse (included)
-    @param[in] last  end of the range to parse (excluded)
-    @param[in] cb  a parser callback function of type @ref parser_callback_t
-    which is used to control the deserialization by filtering unwanted values
-    (optional)
-    @param[in] allow_exceptions  whether to throw exceptions in case of a
-    parse error (optional, true by default)
-
-    @return result of the deserialization
-
-    @throw parse_error.101 in case of an unexpected token
-    @throw parse_error.102 if to_unicode fails or surrogate error
-    @throw parse_error.103 if to_unicode fails
-
-    @complexity Linear in the length of the input. The parser is a predictive
-    LL(1) parser. The complexity can be higher if the parser callback function
-    @a cb has a super-linear complexity.
-
-    @note A UTF-8 byte order mark is silently ignored.
-
-    @liveexample{The example below demonstrates the `parse()` function reading
-    from an iterator range.,parse__iteratortype__parser_callback_t}
-
-    @since version 2.0.3
-    */
-    template<class IteratorType, typename std::enable_if<
-                 std::is_base_of<
-                     std::random_access_iterator_tag,
-                     typename std::iterator_traits<IteratorType>::iterator_category>::value, int>::type = 0>
-    static basic_json parse(IteratorType first, IteratorType last,
-                            const parser_callback_t cb = nullptr,
-                            const bool allow_exceptions = true)
+    /// @brief generate SAX events
+    /// @sa https://json.nlohmann.me/api/basic_json/sax_parse/
+    template<class IteratorType, class SAX>
+    JSON_HEDLEY_NON_NULL(3)
+    static bool sax_parse(IteratorType first, IteratorType last, SAX* sax,
+                          input_format_t format = input_format_t::json,
+                          const bool strict = true,
+                          const bool ignore_comments = false)
     {
-        basic_json result;
-        parser(detail::input_adapter(first, last), cb, allow_exceptions).parse(true, result);
-        return result;
+        auto ia = detail::input_adapter(std::move(first), std::move(last));
+        return format == input_format_t::json
+               ? parser(std::move(ia), nullptr, true, ignore_comments).sax_parse(sax, strict)
+               : detail::binary_reader<basic_json, decltype(ia), SAX>(std::move(ia)).sax_parse(format, sax, strict);
     }
 
-    template<class IteratorType, typename std::enable_if<
-                 std::is_base_of<
-                     std::random_access_iterator_tag,
-                     typename std::iterator_traits<IteratorType>::iterator_category>::value, int>::type = 0>
-    static bool accept(IteratorType first, IteratorType last)
+    /// @brief generate SAX events
+    /// @sa https://json.nlohmann.me/api/basic_json/sax_parse/
+    /// @deprecated This function is deprecated since 3.8.0 and will be removed in
+    ///             version 4.0.0 of the library. Please use
+    ///             sax_parse(ptr, ptr + len) instead.
+    template <typename SAX>
+    JSON_HEDLEY_DEPRECATED_FOR(3.8.0, sax_parse(ptr, ptr + len, ...))
+    JSON_HEDLEY_NON_NULL(2)
+    static bool sax_parse(detail::span_input_adapter&& i, SAX* sax,
+                          input_format_t format = input_format_t::json,
+                          const bool strict = true,
+                          const bool ignore_comments = false)
     {
-        return parser(detail::input_adapter(first, last)).accept(true);
+        auto ia = i.get();
+        return format == input_format_t::json
+               // NOLINTNEXTLINE(hicpp-move-const-arg,performance-move-const-arg)
+               ? parser(std::move(ia), nullptr, true, ignore_comments).sax_parse(sax, strict)
+               // NOLINTNEXTLINE(hicpp-move-const-arg,performance-move-const-arg)
+               : detail::binary_reader<basic_json, decltype(ia), SAX>(std::move(ia)).sax_parse(format, sax, strict);
     }
-
-    template<class IteratorType, class SAX, typename std::enable_if<
-                 std::is_base_of<
-                     std::random_access_iterator_tag,
-                     typename std::iterator_traits<IteratorType>::iterator_category>::value, int>::type = 0>
-    static bool sax_parse(IteratorType first, IteratorType last, SAX* sax)
-    {
-        return parser(detail::input_adapter(first, last)).sax_parse(sax);
-    }
-
-    /*!
-    @brief deserialize from stream
-    @deprecated This stream operator is deprecated and will be removed in
-                version 4.0.0 of the library. Please use
-                @ref operator>>(std::istream&, basic_json&)
-                instead; that is, replace calls like `j << i;` with `i >> j;`.
-    @since version 1.0.0; deprecated since version 3.0.0
-    */
-    JSON_DEPRECATED
+#ifndef JSON_NO_IO
+    /// @brief deserialize from stream
+    /// @sa https://json.nlohmann.me/api/basic_json/operator_gtgt/
+    /// @deprecated This stream operator is deprecated since 3.0.0 and will be removed in
+    ///             version 4.0.0 of the library. Please use
+    ///             operator>>(std::istream&, basic_json&) instead; that is,
+    ///             replace calls like `j << i;` with `i >> j;`.
+    JSON_HEDLEY_DEPRECATED_FOR(3.0.0, operator>>(std::istream&, basic_json&))
     friend std::istream& operator<<(basic_json& j, std::istream& i)
     {
         return operator>>(i, j);
     }
 
-    /*!
-    @brief deserialize from stream
-
-    Deserializes an input stream to a JSON value.
-
-    @param[in,out] i  input stream to read a serialized JSON value from
-    @param[in,out] j  JSON value to write the deserialized input to
-
-    @throw parse_error.101 in case of an unexpected token
-    @throw parse_error.102 if to_unicode fails or surrogate error
-    @throw parse_error.103 if to_unicode fails
-
-    @complexity Linear in the length of the input. The parser is a predictive
-    LL(1) parser.
-
-    @note A UTF-8 byte order mark is silently ignored.
-
-    @liveexample{The example below shows how a JSON value is constructed by
-    reading a serialization from a stream.,operator_deserialize}
-
-    @sa parse(std::istream&, const parser_callback_t) for a variant with a
-    parser callback function to filter values while parsing
-
-    @since version 1.0.0
-    */
+    /// @brief deserialize from stream
+    /// @sa https://json.nlohmann.me/api/basic_json/operator_gtgt/
     friend std::istream& operator>>(std::istream& i, basic_json& j)
     {
         parser(detail::input_adapter(i)).parse(false, j);
         return i;
     }
-
+#endif  // JSON_NO_IO
     /// @}
 
     ///////////////////////////
     // convenience functions //
     ///////////////////////////
 
-    /*!
-    @brief return the type as string
-
-    Returns the type name as string to be used in error messages - usually to
-    indicate that a function was called on a wrong JSON type.
-
-    @return a string representation of a the @a m_type member:
-            Value type  | return value
-            ----------- | -------------
-            null        | `"null"`
-            boolean     | `"boolean"`
-            string      | `"string"`
-            number      | `"number"` (for all number types)
-            object      | `"object"`
-            array       | `"array"`
-            discarded   | `"discarded"`
-
-    @exceptionsafety No-throw guarantee: this function never throws exceptions.
-
-    @complexity Constant.
-
-    @liveexample{The following code exemplifies `type_name()` for all JSON
-    types.,type_name}
-
-    @sa @ref type() -- return the type of the JSON value
-    @sa @ref operator value_t() -- return the type of the JSON value (implicit)
-
-    @since version 1.0.0, public since 2.1.0, `const char*` and `noexcept`
-    since 3.0.0
-    */
+    /// @brief return the type as string
+    /// @sa https://json.nlohmann.me/api/basic_json/type_name/
+    JSON_HEDLEY_RETURNS_NON_NULL
     const char* type_name() const noexcept
     {
+        switch (m_type)
         {
-            switch (m_type)
-            {
-                case value_t::null:
-                    return "null";
-                case value_t::object:
-                    return "object";
-                case value_t::array:
-                    return "array";
-                case value_t::string:
-                    return "string";
-                case value_t::boolean:
-                    return "boolean";
-                case value_t::discarded:
-                    return "discarded";
-                default:
-                    return "number";
-            }
+            case value_t::null:
+                return "null";
+            case value_t::object:
+                return "object";
+            case value_t::array:
+                return "array";
+            case value_t::string:
+                return "string";
+            case value_t::boolean:
+                return "boolean";
+            case value_t::binary:
+                return "binary";
+            case value_t::discarded:
+                return "discarded";
+            case value_t::number_integer:
+            case value_t::number_unsigned:
+            case value_t::number_float:
+            default:
+                return "number";
         }
     }
 
 
-  private:
+  JSON_PRIVATE_UNLESS_TESTED:
     //////////////////////
     // member variables //
     //////////////////////
@@ -17393,6 +20992,11 @@ class basic_json
     /// the value of the current element
     json_value m_value = {};
 
+#if JSON_DIAGNOSTICS
+    /// a pointer to a parent value (for debugging purposes)
+    basic_json* m_parent = nullptr;
+#endif
+
     //////////////////////////////////////////
     // binary serialization/deserialization //
     //////////////////////////////////////////
@@ -17401,616 +21005,322 @@ class basic_json
     /// @{
 
   public:
-    /*!
-    @brief create a CBOR serialization of a given JSON value
-
-    Serializes a given JSON value @a j to a byte vector using the CBOR (Concise
-    Binary Object Representation) serialization format. CBOR is a binary
-    serialization format which aims to be more compact than JSON itself, yet
-    more efficient to parse.
-
-    The library uses the following mapping from JSON values types to
-    CBOR types according to the CBOR specification (RFC 7049):
-
-    JSON value type | value/range                                | CBOR type                          | first byte
-    --------------- | ------------------------------------------ | ---------------------------------- | ---------------
-    null            | `null`                                     | Null                               | 0xF6
-    boolean         | `true`                                     | True                               | 0xF5
-    boolean         | `false`                                    | False                              | 0xF4
-    number_integer  | -9223372036854775808..-2147483649          | Negative integer (8 bytes follow)  | 0x3B
-    number_integer  | -2147483648..-32769                        | Negative integer (4 bytes follow)  | 0x3A
-    number_integer  | -32768..-129                               | Negative integer (2 bytes follow)  | 0x39
-    number_integer  | -128..-25                                  | Negative integer (1 byte follow)   | 0x38
-    number_integer  | -24..-1                                    | Negative integer                   | 0x20..0x37
-    number_integer  | 0..23                                      | Integer                            | 0x00..0x17
-    number_integer  | 24..255                                    | Unsigned integer (1 byte follow)   | 0x18
-    number_integer  | 256..65535                                 | Unsigned integer (2 bytes follow)  | 0x19
-    number_integer  | 65536..4294967295                          | Unsigned integer (4 bytes follow)  | 0x1A
-    number_integer  | 4294967296..18446744073709551615           | Unsigned integer (8 bytes follow)  | 0x1B
-    number_unsigned | 0..23                                      | Integer                            | 0x00..0x17
-    number_unsigned | 24..255                                    | Unsigned integer (1 byte follow)   | 0x18
-    number_unsigned | 256..65535                                 | Unsigned integer (2 bytes follow)  | 0x19
-    number_unsigned | 65536..4294967295                          | Unsigned integer (4 bytes follow)  | 0x1A
-    number_unsigned | 4294967296..18446744073709551615           | Unsigned integer (8 bytes follow)  | 0x1B
-    number_float    | *any value*                                | Double-Precision Float             | 0xFB
-    string          | *length*: 0..23                            | UTF-8 string                       | 0x60..0x77
-    string          | *length*: 23..255                          | UTF-8 string (1 byte follow)       | 0x78
-    string          | *length*: 256..65535                       | UTF-8 string (2 bytes follow)      | 0x79
-    string          | *length*: 65536..4294967295                | UTF-8 string (4 bytes follow)      | 0x7A
-    string          | *length*: 4294967296..18446744073709551615 | UTF-8 string (8 bytes follow)      | 0x7B
-    array           | *size*: 0..23                              | array                              | 0x80..0x97
-    array           | *size*: 23..255                            | array (1 byte follow)              | 0x98
-    array           | *size*: 256..65535                         | array (2 bytes follow)             | 0x99
-    array           | *size*: 65536..4294967295                  | array (4 bytes follow)             | 0x9A
-    array           | *size*: 4294967296..18446744073709551615   | array (8 bytes follow)             | 0x9B
-    object          | *size*: 0..23                              | map                                | 0xA0..0xB7
-    object          | *size*: 23..255                            | map (1 byte follow)                | 0xB8
-    object          | *size*: 256..65535                         | map (2 bytes follow)               | 0xB9
-    object          | *size*: 65536..4294967295                  | map (4 bytes follow)               | 0xBA
-    object          | *size*: 4294967296..18446744073709551615   | map (8 bytes follow)               | 0xBB
-
-    @note The mapping is **complete** in the sense that any JSON value type
-          can be converted to a CBOR value.
-
-    @note If NaN or Infinity are stored inside a JSON number, they are
-          serialized properly. This behavior differs from the @ref dump()
-          function which serializes NaN or Infinity to `null`.
-
-    @note The following CBOR types are not used in the conversion:
-          - byte strings (0x40..0x5F)
-          - UTF-8 strings terminated by "break" (0x7F)
-          - arrays terminated by "break" (0x9F)
-          - maps terminated by "break" (0xBF)
-          - date/time (0xC0..0xC1)
-          - bignum (0xC2..0xC3)
-          - decimal fraction (0xC4)
-          - bigfloat (0xC5)
-          - tagged items (0xC6..0xD4, 0xD8..0xDB)
-          - expected conversions (0xD5..0xD7)
-          - simple values (0xE0..0xF3, 0xF8)
-          - undefined (0xF7)
-          - half and single-precision floats (0xF9-0xFA)
-          - break (0xFF)
-
-    @param[in] j  JSON value to serialize
-    @return MessagePack serialization as byte vector
-
-    @complexity Linear in the size of the JSON value @a j.
-
-    @liveexample{The example shows the serialization of a JSON value to a byte
-    vector in CBOR format.,to_cbor}
-
-    @sa http://cbor.io
-    @sa @ref from_cbor(detail::input_adapter, const bool strict) for the
-        analogous deserialization
-    @sa @ref to_msgpack(const basic_json&) for the related MessagePack format
-    @sa @ref to_ubjson(const basic_json&, const bool, const bool) for the
-             related UBJSON format
-
-    @since version 2.0.9
-    */
-    static std::vector<uint8_t> to_cbor(const basic_json& j)
+    /// @brief create a CBOR serialization of a given JSON value
+    /// @sa https://json.nlohmann.me/api/basic_json/to_cbor/
+    static std::vector<std::uint8_t> to_cbor(const basic_json& j)
     {
-        std::vector<uint8_t> result;
+        std::vector<std::uint8_t> result;
         to_cbor(j, result);
         return result;
     }
 
-    static void to_cbor(const basic_json& j, detail::output_adapter<uint8_t> o)
+    /// @brief create a CBOR serialization of a given JSON value
+    /// @sa https://json.nlohmann.me/api/basic_json/to_cbor/
+    static void to_cbor(const basic_json& j, detail::output_adapter<std::uint8_t> o)
     {
-        binary_writer<uint8_t>(o).write_cbor(j);
+        binary_writer<std::uint8_t>(o).write_cbor(j);
     }
 
+    /// @brief create a CBOR serialization of a given JSON value
+    /// @sa https://json.nlohmann.me/api/basic_json/to_cbor/
     static void to_cbor(const basic_json& j, detail::output_adapter<char> o)
     {
         binary_writer<char>(o).write_cbor(j);
     }
 
-    /*!
-    @brief create a MessagePack serialization of a given JSON value
-
-    Serializes a given JSON value @a j to a byte vector using the MessagePack
-    serialization format. MessagePack is a binary serialization format which
-    aims to be more compact than JSON itself, yet more efficient to parse.
-
-    The library uses the following mapping from JSON values types to
-    MessagePack types according to the MessagePack specification:
-
-    JSON value type | value/range                       | MessagePack type | first byte
-    --------------- | --------------------------------- | ---------------- | ----------
-    null            | `null`                            | nil              | 0xC0
-    boolean         | `true`                            | true             | 0xC3
-    boolean         | `false`                           | false            | 0xC2
-    number_integer  | -9223372036854775808..-2147483649 | int64            | 0xD3
-    number_integer  | -2147483648..-32769               | int32            | 0xD2
-    number_integer  | -32768..-129                      | int16            | 0xD1
-    number_integer  | -128..-33                         | int8             | 0xD0
-    number_integer  | -32..-1                           | negative fixint  | 0xE0..0xFF
-    number_integer  | 0..127                            | positive fixint  | 0x00..0x7F
-    number_integer  | 128..255                          | uint 8           | 0xCC
-    number_integer  | 256..65535                        | uint 16          | 0xCD
-    number_integer  | 65536..4294967295                 | uint 32          | 0xCE
-    number_integer  | 4294967296..18446744073709551615  | uint 64          | 0xCF
-    number_unsigned | 0..127                            | positive fixint  | 0x00..0x7F
-    number_unsigned | 128..255                          | uint 8           | 0xCC
-    number_unsigned | 256..65535                        | uint 16          | 0xCD
-    number_unsigned | 65536..4294967295                 | uint 32          | 0xCE
-    number_unsigned | 4294967296..18446744073709551615  | uint 64          | 0xCF
-    number_float    | *any value*                       | float 64         | 0xCB
-    string          | *length*: 0..31                   | fixstr           | 0xA0..0xBF
-    string          | *length*: 32..255                 | str 8            | 0xD9
-    string          | *length*: 256..65535              | str 16           | 0xDA
-    string          | *length*: 65536..4294967295       | str 32           | 0xDB
-    array           | *size*: 0..15                     | fixarray         | 0x90..0x9F
-    array           | *size*: 16..65535                 | array 16         | 0xDC
-    array           | *size*: 65536..4294967295         | array 32         | 0xDD
-    object          | *size*: 0..15                     | fix map          | 0x80..0x8F
-    object          | *size*: 16..65535                 | map 16           | 0xDE
-    object          | *size*: 65536..4294967295         | map 32           | 0xDF
-
-    @note The mapping is **complete** in the sense that any JSON value type
-          can be converted to a MessagePack value.
-
-    @note The following values can **not** be converted to a MessagePack value:
-          - strings with more than 4294967295 bytes
-          - arrays with more than 4294967295 elements
-          - objects with more than 4294967295 elements
-
-    @note The following MessagePack types are not used in the conversion:
-          - bin 8 - bin 32 (0xC4..0xC6)
-          - ext 8 - ext 32 (0xC7..0xC9)
-          - float 32 (0xCA)
-          - fixext 1 - fixext 16 (0xD4..0xD8)
-
-    @note Any MessagePack output created @ref to_msgpack can be successfully
-          parsed by @ref from_msgpack.
-
-    @note If NaN or Infinity are stored inside a JSON number, they are
-          serialized properly. This behavior differs from the @ref dump()
-          function which serializes NaN or Infinity to `null`.
-
-    @param[in] j  JSON value to serialize
-    @return MessagePack serialization as byte vector
-
-    @complexity Linear in the size of the JSON value @a j.
-
-    @liveexample{The example shows the serialization of a JSON value to a byte
-    vector in MessagePack format.,to_msgpack}
-
-    @sa http://msgpack.org
-    @sa @ref from_msgpack(const std::vector<uint8_t>&, const size_t) for the
-        analogous deserialization
-    @sa @ref to_cbor(const basic_json& for the related CBOR format
-    @sa @ref to_ubjson(const basic_json&, const bool, const bool) for the
-             related UBJSON format
-
-    @since version 2.0.9
-    */
-    static std::vector<uint8_t> to_msgpack(const basic_json& j)
+    /// @brief create a MessagePack serialization of a given JSON value
+    /// @sa https://json.nlohmann.me/api/basic_json/to_msgpack/
+    static std::vector<std::uint8_t> to_msgpack(const basic_json& j)
     {
-        std::vector<uint8_t> result;
+        std::vector<std::uint8_t> result;
         to_msgpack(j, result);
         return result;
     }
 
-    static void to_msgpack(const basic_json& j, detail::output_adapter<uint8_t> o)
+    /// @brief create a MessagePack serialization of a given JSON value
+    /// @sa https://json.nlohmann.me/api/basic_json/to_msgpack/
+    static void to_msgpack(const basic_json& j, detail::output_adapter<std::uint8_t> o)
     {
-        binary_writer<uint8_t>(o).write_msgpack(j);
+        binary_writer<std::uint8_t>(o).write_msgpack(j);
     }
 
+    /// @brief create a MessagePack serialization of a given JSON value
+    /// @sa https://json.nlohmann.me/api/basic_json/to_msgpack/
     static void to_msgpack(const basic_json& j, detail::output_adapter<char> o)
     {
         binary_writer<char>(o).write_msgpack(j);
     }
 
-    /*!
-    @brief create a UBJSON serialization of a given JSON value
-
-    Serializes a given JSON value @a j to a byte vector using the UBJSON
-    (Universal Binary JSON) serialization format. UBJSON aims to be more compact
-    than JSON itself, yet more efficient to parse.
-
-    The library uses the following mapping from JSON values types to
-    UBJSON types according to the UBJSON specification:
-
-    JSON value type | value/range                       | UBJSON type | marker
-    --------------- | --------------------------------- | ----------- | ------
-    null            | `null`                            | null        | `Z`
-    boolean         | `true`                            | true        | `T`
-    boolean         | `false`                           | false       | `F`
-    number_integer  | -9223372036854775808..-2147483649 | int64       | `L`
-    number_integer  | -2147483648..-32769               | int32       | `l`
-    number_integer  | -32768..-129                      | int16       | `I`
-    number_integer  | -128..127                         | int8        | `i`
-    number_integer  | 128..255                          | uint8       | `U`
-    number_integer  | 256..32767                        | int16       | `I`
-    number_integer  | 32768..2147483647                 | int32       | `l`
-    number_integer  | 2147483648..9223372036854775807   | int64       | `L`
-    number_unsigned | 0..127                            | int8        | `i`
-    number_unsigned | 128..255                          | uint8       | `U`
-    number_unsigned | 256..32767                        | int16       | `I`
-    number_unsigned | 32768..2147483647                 | int32       | `l`
-    number_unsigned | 2147483648..9223372036854775807   | int64       | `L`
-    number_float    | *any value*                       | float64     | `D`
-    string          | *with shortest length indicator*  | string      | `S`
-    array           | *see notes on optimized format*   | array       | `[`
-    object          | *see notes on optimized format*   | map         | `{`
-
-    @note The mapping is **complete** in the sense that any JSON value type
-          can be converted to a UBJSON value.
-
-    @note The following values can **not** be converted to a UBJSON value:
-          - strings with more than 9223372036854775807 bytes (theoretical)
-          - unsigned integer numbers above 9223372036854775807
-
-    @note The following markers are not used in the conversion:
-          - `Z`: no-op values are not created.
-          - `C`: single-byte strings are serialized with `S` markers.
-
-    @note Any UBJSON output created @ref to_ubjson can be successfully parsed
-          by @ref from_ubjson.
-
-    @note If NaN or Infinity are stored inside a JSON number, they are
-          serialized properly. This behavior differs from the @ref dump()
-          function which serializes NaN or Infinity to `null`.
-
-    @note The optimized formats for containers are supported: Parameter
-          @a use_size adds size information to the beginning of a container and
-          removes the closing marker. Parameter @a use_type further checks
-          whether all elements of a container have the same type and adds the
-          type marker to the beginning of the container. The @a use_type
-          parameter must only be used together with @a use_size = true. Note
-          that @a use_size = true alone may result in larger representations -
-          the benefit of this parameter is that the receiving side is
-          immediately informed on the number of elements of the container.
-
-    @param[in] j  JSON value to serialize
-    @param[in] use_size  whether to add size annotations to container types
-    @param[in] use_type  whether to add type annotations to container types
-                         (must be combined with @a use_size = true)
-    @return UBJSON serialization as byte vector
-
-    @complexity Linear in the size of the JSON value @a j.
-
-    @liveexample{The example shows the serialization of a JSON value to a byte
-    vector in UBJSON format.,to_ubjson}
-
-    @sa http://ubjson.org
-    @sa @ref from_ubjson(detail::input_adapter, const bool strict) for the
-        analogous deserialization
-    @sa @ref to_cbor(const basic_json& for the related CBOR format
-    @sa @ref to_msgpack(const basic_json&) for the related MessagePack format
-
-    @since version 3.1.0
-    */
-    static std::vector<uint8_t> to_ubjson(const basic_json& j,
-                                          const bool use_size = false,
-                                          const bool use_type = false)
+    /// @brief create a UBJSON serialization of a given JSON value
+    /// @sa https://json.nlohmann.me/api/basic_json/to_ubjson/
+    static std::vector<std::uint8_t> to_ubjson(const basic_json& j,
+            const bool use_size = false,
+            const bool use_type = false)
     {
-        std::vector<uint8_t> result;
+        std::vector<std::uint8_t> result;
         to_ubjson(j, result, use_size, use_type);
         return result;
     }
 
-    static void to_ubjson(const basic_json& j, detail::output_adapter<uint8_t> o,
+    /// @brief create a UBJSON serialization of a given JSON value
+    /// @sa https://json.nlohmann.me/api/basic_json/to_ubjson/
+    static void to_ubjson(const basic_json& j, detail::output_adapter<std::uint8_t> o,
                           const bool use_size = false, const bool use_type = false)
     {
-        binary_writer<uint8_t>(o).write_ubjson(j, use_size, use_type);
+        binary_writer<std::uint8_t>(o).write_ubjson(j, use_size, use_type);
     }
 
+    /// @brief create a UBJSON serialization of a given JSON value
+    /// @sa https://json.nlohmann.me/api/basic_json/to_ubjson/
     static void to_ubjson(const basic_json& j, detail::output_adapter<char> o,
                           const bool use_size = false, const bool use_type = false)
     {
         binary_writer<char>(o).write_ubjson(j, use_size, use_type);
     }
 
-    /*!
-    @brief create a JSON value from an input in CBOR format
+    /// @brief create a BSON serialization of a given JSON value
+    /// @sa https://json.nlohmann.me/api/basic_json/to_bson/
+    static std::vector<std::uint8_t> to_bson(const basic_json& j)
+    {
+        std::vector<std::uint8_t> result;
+        to_bson(j, result);
+        return result;
+    }
 
-    Deserializes a given input @a i to a JSON value using the CBOR (Concise
-    Binary Object Representation) serialization format.
+    /// @brief create a BSON serialization of a given JSON value
+    /// @sa https://json.nlohmann.me/api/basic_json/to_bson/
+    static void to_bson(const basic_json& j, detail::output_adapter<std::uint8_t> o)
+    {
+        binary_writer<std::uint8_t>(o).write_bson(j);
+    }
 
-    The library maps CBOR types to JSON value types as follows:
+    /// @brief create a BSON serialization of a given JSON value
+    /// @sa https://json.nlohmann.me/api/basic_json/to_bson/
+    static void to_bson(const basic_json& j, detail::output_adapter<char> o)
+    {
+        binary_writer<char>(o).write_bson(j);
+    }
 
-    CBOR type              | JSON value type | first byte
-    ---------------------- | --------------- | ----------
-    Integer                | number_unsigned | 0x00..0x17
-    Unsigned integer       | number_unsigned | 0x18
-    Unsigned integer       | number_unsigned | 0x19
-    Unsigned integer       | number_unsigned | 0x1A
-    Unsigned integer       | number_unsigned | 0x1B
-    Negative integer       | number_integer  | 0x20..0x37
-    Negative integer       | number_integer  | 0x38
-    Negative integer       | number_integer  | 0x39
-    Negative integer       | number_integer  | 0x3A
-    Negative integer       | number_integer  | 0x3B
-    Negative integer       | number_integer  | 0x40..0x57
-    UTF-8 string           | string          | 0x60..0x77
-    UTF-8 string           | string          | 0x78
-    UTF-8 string           | string          | 0x79
-    UTF-8 string           | string          | 0x7A
-    UTF-8 string           | string          | 0x7B
-    UTF-8 string           | string          | 0x7F
-    array                  | array           | 0x80..0x97
-    array                  | array           | 0x98
-    array                  | array           | 0x99
-    array                  | array           | 0x9A
-    array                  | array           | 0x9B
-    array                  | array           | 0x9F
-    map                    | object          | 0xA0..0xB7
-    map                    | object          | 0xB8
-    map                    | object          | 0xB9
-    map                    | object          | 0xBA
-    map                    | object          | 0xBB
-    map                    | object          | 0xBF
-    False                  | `false`         | 0xF4
-    True                   | `true`          | 0xF5
-    Nill                   | `null`          | 0xF6
-    Half-Precision Float   | number_float    | 0xF9
-    Single-Precision Float | number_float    | 0xFA
-    Double-Precision Float | number_float    | 0xFB
-
-    @warning The mapping is **incomplete** in the sense that not all CBOR
-             types can be converted to a JSON value. The following CBOR types
-             are not supported and will yield parse errors (parse_error.112):
-             - byte strings (0x40..0x5F)
-             - date/time (0xC0..0xC1)
-             - bignum (0xC2..0xC3)
-             - decimal fraction (0xC4)
-             - bigfloat (0xC5)
-             - tagged items (0xC6..0xD4, 0xD8..0xDB)
-             - expected conversions (0xD5..0xD7)
-             - simple values (0xE0..0xF3, 0xF8)
-             - undefined (0xF7)
-
-    @warning CBOR allows map keys of any type, whereas JSON only allows
-             strings as keys in object values. Therefore, CBOR maps with keys
-             other than UTF-8 strings are rejected (parse_error.113).
-
-    @note Any CBOR output created @ref to_cbor can be successfully parsed by
-          @ref from_cbor.
-
-    @param[in] i  an input in CBOR format convertible to an input adapter
-    @param[in] strict  whether to expect the input to be consumed until EOF
-                       (true by default)
-    @param[in] allow_exceptions  whether to throw exceptions in case of a
-    parse error (optional, true by default)
-
-    @return deserialized JSON value
-
-    @throw parse_error.110 if the given input ends prematurely or the end of
-    file was not reached when @a strict was set to true
-    @throw parse_error.112 if unsupported features from CBOR were
-    used in the given input @a v or if the input is not valid CBOR
-    @throw parse_error.113 if a string was expected as map key, but not found
-
-    @complexity Linear in the size of the input @a i.
-
-    @liveexample{The example shows the deserialization of a byte vector in CBOR
-    format to a JSON value.,from_cbor}
-
-    @sa http://cbor.io
-    @sa @ref to_cbor(const basic_json&) for the analogous serialization
-    @sa @ref from_msgpack(detail::input_adapter, const bool, const bool) for the
-        related MessagePack format
-    @sa @ref from_ubjson(detail::input_adapter, const bool, const bool) for the
-        related UBJSON format
-
-    @since version 2.0.9; parameter @a start_index since 2.1.1; changed to
-           consume input adapters, removed start_index parameter, and added
-           @a strict parameter since 3.0.0; added @allow_exceptions parameter
-           since 3.2.0
-    */
-    static basic_json from_cbor(detail::input_adapter&& i,
+    /// @brief create a JSON value from an input in CBOR format
+    /// @sa https://json.nlohmann.me/api/basic_json/from_cbor/
+    template<typename InputType>
+    JSON_HEDLEY_WARN_UNUSED_RESULT
+    static basic_json from_cbor(InputType&& i,
                                 const bool strict = true,
-                                const bool allow_exceptions = true)
+                                const bool allow_exceptions = true,
+                                const cbor_tag_handler_t tag_handler = cbor_tag_handler_t::error)
     {
         basic_json result;
         detail::json_sax_dom_parser<basic_json> sdp(result, allow_exceptions);
-        const bool res = binary_reader(detail::input_adapter(i)).sax_parse(input_format_t::cbor, &sdp, strict);
+        auto ia = detail::input_adapter(std::forward<InputType>(i));
+        const bool res = binary_reader<decltype(ia)>(std::move(ia)).sax_parse(input_format_t::cbor, &sdp, strict, tag_handler);
         return res ? result : basic_json(value_t::discarded);
     }
 
-    /*!
-    @copydoc from_cbor(detail::input_adapter, const bool, const bool)
-    */
-    template<typename A1, typename A2,
-             detail::enable_if_t<std::is_constructible<detail::input_adapter, A1, A2>::value, int> = 0>
-    static basic_json from_cbor(A1 && a1, A2 && a2,
+    /// @brief create a JSON value from an input in CBOR format
+    /// @sa https://json.nlohmann.me/api/basic_json/from_cbor/
+    template<typename IteratorType>
+    JSON_HEDLEY_WARN_UNUSED_RESULT
+    static basic_json from_cbor(IteratorType first, IteratorType last,
                                 const bool strict = true,
-                                const bool allow_exceptions = true)
+                                const bool allow_exceptions = true,
+                                const cbor_tag_handler_t tag_handler = cbor_tag_handler_t::error)
     {
         basic_json result;
         detail::json_sax_dom_parser<basic_json> sdp(result, allow_exceptions);
-        const bool res = binary_reader(detail::input_adapter(std::forward<A1>(a1), std::forward<A2>(a2))).sax_parse(input_format_t::cbor, &sdp, strict);
+        auto ia = detail::input_adapter(std::move(first), std::move(last));
+        const bool res = binary_reader<decltype(ia)>(std::move(ia)).sax_parse(input_format_t::cbor, &sdp, strict, tag_handler);
         return res ? result : basic_json(value_t::discarded);
     }
 
-    /*!
-    @brief create a JSON value from an input in MessagePack format
+    template<typename T>
+    JSON_HEDLEY_WARN_UNUSED_RESULT
+    JSON_HEDLEY_DEPRECATED_FOR(3.8.0, from_cbor(ptr, ptr + len))
+    static basic_json from_cbor(const T* ptr, std::size_t len,
+                                const bool strict = true,
+                                const bool allow_exceptions = true,
+                                const cbor_tag_handler_t tag_handler = cbor_tag_handler_t::error)
+    {
+        return from_cbor(ptr, ptr + len, strict, allow_exceptions, tag_handler);
+    }
 
-    Deserializes a given input @a i to a JSON value using the MessagePack
-    serialization format.
 
-    The library maps MessagePack types to JSON value types as follows:
+    JSON_HEDLEY_WARN_UNUSED_RESULT
+    JSON_HEDLEY_DEPRECATED_FOR(3.8.0, from_cbor(ptr, ptr + len))
+    static basic_json from_cbor(detail::span_input_adapter&& i,
+                                const bool strict = true,
+                                const bool allow_exceptions = true,
+                                const cbor_tag_handler_t tag_handler = cbor_tag_handler_t::error)
+    {
+        basic_json result;
+        detail::json_sax_dom_parser<basic_json> sdp(result, allow_exceptions);
+        auto ia = i.get();
+        // NOLINTNEXTLINE(hicpp-move-const-arg,performance-move-const-arg)
+        const bool res = binary_reader<decltype(ia)>(std::move(ia)).sax_parse(input_format_t::cbor, &sdp, strict, tag_handler);
+        return res ? result : basic_json(value_t::discarded);
+    }
 
-    MessagePack type | JSON value type | first byte
-    ---------------- | --------------- | ----------
-    positive fixint  | number_unsigned | 0x00..0x7F
-    fixmap           | object          | 0x80..0x8F
-    fixarray         | array           | 0x90..0x9F
-    fixstr           | string          | 0xA0..0xBF
-    nil              | `null`          | 0xC0
-    false            | `false`         | 0xC2
-    true             | `true`          | 0xC3
-    float 32         | number_float    | 0xCA
-    float 64         | number_float    | 0xCB
-    uint 8           | number_unsigned | 0xCC
-    uint 16          | number_unsigned | 0xCD
-    uint 32          | number_unsigned | 0xCE
-    uint 64          | number_unsigned | 0xCF
-    int 8            | number_integer  | 0xD0
-    int 16           | number_integer  | 0xD1
-    int 32           | number_integer  | 0xD2
-    int 64           | number_integer  | 0xD3
-    str 8            | string          | 0xD9
-    str 16           | string          | 0xDA
-    str 32           | string          | 0xDB
-    array 16         | array           | 0xDC
-    array 32         | array           | 0xDD
-    map 16           | object          | 0xDE
-    map 32           | object          | 0xDF
-    negative fixint  | number_integer  | 0xE0-0xFF
-
-    @warning The mapping is **incomplete** in the sense that not all
-             MessagePack types can be converted to a JSON value. The following
-             MessagePack types are not supported and will yield parse errors:
-              - bin 8 - bin 32 (0xC4..0xC6)
-              - ext 8 - ext 32 (0xC7..0xC9)
-              - fixext 1 - fixext 16 (0xD4..0xD8)
-
-    @note Any MessagePack output created @ref to_msgpack can be successfully
-          parsed by @ref from_msgpack.
-
-    @param[in] i  an input in MessagePack format convertible to an input
-                  adapter
-    @param[in] strict  whether to expect the input to be consumed until EOF
-                       (true by default)
-    @param[in] allow_exceptions  whether to throw exceptions in case of a
-    parse error (optional, true by default)
-
-    @return deserialized JSON value
-
-    @throw parse_error.110 if the given input ends prematurely or the end of
-    file was not reached when @a strict was set to true
-    @throw parse_error.112 if unsupported features from MessagePack were
-    used in the given input @a i or if the input is not valid MessagePack
-    @throw parse_error.113 if a string was expected as map key, but not found
-
-    @complexity Linear in the size of the input @a i.
-
-    @liveexample{The example shows the deserialization of a byte vector in
-    MessagePack format to a JSON value.,from_msgpack}
-
-    @sa http://msgpack.org
-    @sa @ref to_msgpack(const basic_json&) for the analogous serialization
-    @sa @ref from_cbor(detail::input_adapter, const bool, const bool) for the
-        related CBOR format
-    @sa @ref from_ubjson(detail::input_adapter, const bool, const bool) for
-        the related UBJSON format
-
-    @since version 2.0.9; parameter @a start_index since 2.1.1; changed to
-           consume input adapters, removed start_index parameter, and added
-           @a strict parameter since 3.0.0; added @allow_exceptions parameter
-           since 3.2.0
-    */
-    static basic_json from_msgpack(detail::input_adapter&& i,
+    /// @brief create a JSON value from an input in MessagePack format
+    /// @sa https://json.nlohmann.me/api/basic_json/from_msgpack/
+    template<typename InputType>
+    JSON_HEDLEY_WARN_UNUSED_RESULT
+    static basic_json from_msgpack(InputType&& i,
                                    const bool strict = true,
                                    const bool allow_exceptions = true)
     {
         basic_json result;
         detail::json_sax_dom_parser<basic_json> sdp(result, allow_exceptions);
-        const bool res = binary_reader(detail::input_adapter(i)).sax_parse(input_format_t::msgpack, &sdp, strict);
+        auto ia = detail::input_adapter(std::forward<InputType>(i));
+        const bool res = binary_reader<decltype(ia)>(std::move(ia)).sax_parse(input_format_t::msgpack, &sdp, strict);
         return res ? result : basic_json(value_t::discarded);
     }
 
-    /*!
-    @copydoc from_msgpack(detail::input_adapter, const bool, const bool)
-    */
-    template<typename A1, typename A2,
-             detail::enable_if_t<std::is_constructible<detail::input_adapter, A1, A2>::value, int> = 0>
-    static basic_json from_msgpack(A1 && a1, A2 && a2,
+    /// @brief create a JSON value from an input in MessagePack format
+    /// @sa https://json.nlohmann.me/api/basic_json/from_msgpack/
+    template<typename IteratorType>
+    JSON_HEDLEY_WARN_UNUSED_RESULT
+    static basic_json from_msgpack(IteratorType first, IteratorType last,
                                    const bool strict = true,
                                    const bool allow_exceptions = true)
     {
         basic_json result;
         detail::json_sax_dom_parser<basic_json> sdp(result, allow_exceptions);
-        const bool res = binary_reader(detail::input_adapter(std::forward<A1>(a1), std::forward<A2>(a2))).sax_parse(input_format_t::msgpack, &sdp, strict);
+        auto ia = detail::input_adapter(std::move(first), std::move(last));
+        const bool res = binary_reader<decltype(ia)>(std::move(ia)).sax_parse(input_format_t::msgpack, &sdp, strict);
         return res ? result : basic_json(value_t::discarded);
     }
 
-    /*!
-    @brief create a JSON value from an input in UBJSON format
+    template<typename T>
+    JSON_HEDLEY_WARN_UNUSED_RESULT
+    JSON_HEDLEY_DEPRECATED_FOR(3.8.0, from_msgpack(ptr, ptr + len))
+    static basic_json from_msgpack(const T* ptr, std::size_t len,
+                                   const bool strict = true,
+                                   const bool allow_exceptions = true)
+    {
+        return from_msgpack(ptr, ptr + len, strict, allow_exceptions);
+    }
 
-    Deserializes a given input @a i to a JSON value using the UBJSON (Universal
-    Binary JSON) serialization format.
+    JSON_HEDLEY_WARN_UNUSED_RESULT
+    JSON_HEDLEY_DEPRECATED_FOR(3.8.0, from_msgpack(ptr, ptr + len))
+    static basic_json from_msgpack(detail::span_input_adapter&& i,
+                                   const bool strict = true,
+                                   const bool allow_exceptions = true)
+    {
+        basic_json result;
+        detail::json_sax_dom_parser<basic_json> sdp(result, allow_exceptions);
+        auto ia = i.get();
+        // NOLINTNEXTLINE(hicpp-move-const-arg,performance-move-const-arg)
+        const bool res = binary_reader<decltype(ia)>(std::move(ia)).sax_parse(input_format_t::msgpack, &sdp, strict);
+        return res ? result : basic_json(value_t::discarded);
+    }
 
-    The library maps UBJSON types to JSON value types as follows:
-
-    UBJSON type | JSON value type                         | marker
-    ----------- | --------------------------------------- | ------
-    no-op       | *no value, next value is read*          | `N`
-    null        | `null`                                  | `Z`
-    false       | `false`                                 | `F`
-    true        | `true`                                  | `T`
-    float32     | number_float                            | `d`
-    float64     | number_float                            | `D`
-    uint8       | number_unsigned                         | `U`
-    int8        | number_integer                          | `i`
-    int16       | number_integer                          | `I`
-    int32       | number_integer                          | `l`
-    int64       | number_integer                          | `L`
-    string      | string                                  | `S`
-    char        | string                                  | `C`
-    array       | array (optimized values are supported)  | `[`
-    object      | object (optimized values are supported) | `{`
-
-    @note The mapping is **complete** in the sense that any UBJSON value can
-          be converted to a JSON value.
-
-    @param[in] i  an input in UBJSON format convertible to an input adapter
-    @param[in] strict  whether to expect the input to be consumed until EOF
-                       (true by default)
-    @param[in] allow_exceptions  whether to throw exceptions in case of a
-    parse error (optional, true by default)
-
-    @return deserialized JSON value
-
-    @throw parse_error.110 if the given input ends prematurely or the end of
-    file was not reached when @a strict was set to true
-    @throw parse_error.112 if a parse error occurs
-    @throw parse_error.113 if a string could not be parsed successfully
-
-    @complexity Linear in the size of the input @a i.
-
-    @liveexample{The example shows the deserialization of a byte vector in
-    UBJSON format to a JSON value.,from_ubjson}
-
-    @sa http://ubjson.org
-    @sa @ref to_ubjson(const basic_json&, const bool, const bool) for the
-             analogous serialization
-    @sa @ref from_cbor(detail::input_adapter, const bool, const bool) for the
-        related CBOR format
-    @sa @ref from_msgpack(detail::input_adapter, const bool, const bool) for
-        the related MessagePack format
-
-    @since version 3.1.0; added @allow_exceptions parameter since 3.2.0
-    */
-    static basic_json from_ubjson(detail::input_adapter&& i,
+    /// @brief create a JSON value from an input in UBJSON format
+    /// @sa https://json.nlohmann.me/api/basic_json/from_ubjson/
+    template<typename InputType>
+    JSON_HEDLEY_WARN_UNUSED_RESULT
+    static basic_json from_ubjson(InputType&& i,
                                   const bool strict = true,
                                   const bool allow_exceptions = true)
     {
         basic_json result;
         detail::json_sax_dom_parser<basic_json> sdp(result, allow_exceptions);
-        const bool res = binary_reader(detail::input_adapter(i)).sax_parse(input_format_t::ubjson, &sdp, strict);
+        auto ia = detail::input_adapter(std::forward<InputType>(i));
+        const bool res = binary_reader<decltype(ia)>(std::move(ia)).sax_parse(input_format_t::ubjson, &sdp, strict);
         return res ? result : basic_json(value_t::discarded);
     }
 
-    /*!
-    @copydoc from_ubjson(detail::input_adapter, const bool, const bool)
-    */
-    template<typename A1, typename A2,
-             detail::enable_if_t<std::is_constructible<detail::input_adapter, A1, A2>::value, int> = 0>
-    static basic_json from_ubjson(A1 && a1, A2 && a2,
+    /// @brief create a JSON value from an input in UBJSON format
+    /// @sa https://json.nlohmann.me/api/basic_json/from_ubjson/
+    template<typename IteratorType>
+    JSON_HEDLEY_WARN_UNUSED_RESULT
+    static basic_json from_ubjson(IteratorType first, IteratorType last,
                                   const bool strict = true,
                                   const bool allow_exceptions = true)
     {
         basic_json result;
         detail::json_sax_dom_parser<basic_json> sdp(result, allow_exceptions);
-        const bool res = binary_reader(detail::input_adapter(std::forward<A1>(a1), std::forward<A2>(a2))).sax_parse(input_format_t::ubjson, &sdp, strict);
+        auto ia = detail::input_adapter(std::move(first), std::move(last));
+        const bool res = binary_reader<decltype(ia)>(std::move(ia)).sax_parse(input_format_t::ubjson, &sdp, strict);
         return res ? result : basic_json(value_t::discarded);
     }
 
+    template<typename T>
+    JSON_HEDLEY_WARN_UNUSED_RESULT
+    JSON_HEDLEY_DEPRECATED_FOR(3.8.0, from_ubjson(ptr, ptr + len))
+    static basic_json from_ubjson(const T* ptr, std::size_t len,
+                                  const bool strict = true,
+                                  const bool allow_exceptions = true)
+    {
+        return from_ubjson(ptr, ptr + len, strict, allow_exceptions);
+    }
+
+    JSON_HEDLEY_WARN_UNUSED_RESULT
+    JSON_HEDLEY_DEPRECATED_FOR(3.8.0, from_ubjson(ptr, ptr + len))
+    static basic_json from_ubjson(detail::span_input_adapter&& i,
+                                  const bool strict = true,
+                                  const bool allow_exceptions = true)
+    {
+        basic_json result;
+        detail::json_sax_dom_parser<basic_json> sdp(result, allow_exceptions);
+        auto ia = i.get();
+        // NOLINTNEXTLINE(hicpp-move-const-arg,performance-move-const-arg)
+        const bool res = binary_reader<decltype(ia)>(std::move(ia)).sax_parse(input_format_t::ubjson, &sdp, strict);
+        return res ? result : basic_json(value_t::discarded);
+    }
+
+    /// @brief create a JSON value from an input in BSON format
+    /// @sa https://json.nlohmann.me/api/basic_json/from_bson/
+    template<typename InputType>
+    JSON_HEDLEY_WARN_UNUSED_RESULT
+    static basic_json from_bson(InputType&& i,
+                                const bool strict = true,
+                                const bool allow_exceptions = true)
+    {
+        basic_json result;
+        detail::json_sax_dom_parser<basic_json> sdp(result, allow_exceptions);
+        auto ia = detail::input_adapter(std::forward<InputType>(i));
+        const bool res = binary_reader<decltype(ia)>(std::move(ia)).sax_parse(input_format_t::bson, &sdp, strict);
+        return res ? result : basic_json(value_t::discarded);
+    }
+
+    /// @brief create a JSON value from an input in BSON format
+    /// @sa https://json.nlohmann.me/api/basic_json/from_bson/
+    template<typename IteratorType>
+    JSON_HEDLEY_WARN_UNUSED_RESULT
+    static basic_json from_bson(IteratorType first, IteratorType last,
+                                const bool strict = true,
+                                const bool allow_exceptions = true)
+    {
+        basic_json result;
+        detail::json_sax_dom_parser<basic_json> sdp(result, allow_exceptions);
+        auto ia = detail::input_adapter(std::move(first), std::move(last));
+        const bool res = binary_reader<decltype(ia)>(std::move(ia)).sax_parse(input_format_t::bson, &sdp, strict);
+        return res ? result : basic_json(value_t::discarded);
+    }
+
+    template<typename T>
+    JSON_HEDLEY_WARN_UNUSED_RESULT
+    JSON_HEDLEY_DEPRECATED_FOR(3.8.0, from_bson(ptr, ptr + len))
+    static basic_json from_bson(const T* ptr, std::size_t len,
+                                const bool strict = true,
+                                const bool allow_exceptions = true)
+    {
+        return from_bson(ptr, ptr + len, strict, allow_exceptions);
+    }
+
+    JSON_HEDLEY_WARN_UNUSED_RESULT
+    JSON_HEDLEY_DEPRECATED_FOR(3.8.0, from_bson(ptr, ptr + len))
+    static basic_json from_bson(detail::span_input_adapter&& i,
+                                const bool strict = true,
+                                const bool allow_exceptions = true)
+    {
+        basic_json result;
+        detail::json_sax_dom_parser<basic_json> sdp(result, allow_exceptions);
+        auto ia = i.get();
+        // NOLINTNEXTLINE(hicpp-move-const-arg,performance-move-const-arg)
+        const bool res = binary_reader<decltype(ia)>(std::move(ia)).sax_parse(input_format_t::bson, &sdp, strict);
+        return res ? result : basic_json(value_t::discarded);
+    }
     /// @}
 
     //////////////////////////
@@ -18020,180 +21330,36 @@ class basic_json
     /// @name JSON Pointer functions
     /// @{
 
-    /*!
-    @brief access specified element via JSON Pointer
-
-    Uses a JSON pointer to retrieve a reference to the respective JSON value.
-    No bound checking is performed. Similar to @ref operator[](const typename
-    object_t::key_type&), `null` values are created in arrays and objects if
-    necessary.
-
-    In particular:
-    - If the JSON pointer points to an object key that does not exist, it
-      is created an filled with a `null` value before a reference to it
-      is returned.
-    - If the JSON pointer points to an array index that does not exist, it
-      is created an filled with a `null` value before a reference to it
-      is returned. All indices between the current maximum and the given
-      index are also filled with `null`.
-    - The special value `-` is treated as a synonym for the index past the
-      end.
-
-    @param[in] ptr  a JSON pointer
-
-    @return reference to the element pointed to by @a ptr
-
-    @complexity Constant.
-
-    @throw parse_error.106   if an array index begins with '0'
-    @throw parse_error.109   if an array index was not a number
-    @throw out_of_range.404  if the JSON pointer can not be resolved
-
-    @liveexample{The behavior is shown in the example.,operatorjson_pointer}
-
-    @since version 2.0.0
-    */
+    /// @brief access specified element via JSON Pointer
+    /// @sa https://json.nlohmann.me/api/basic_json/operator%5B%5D/
     reference operator[](const json_pointer& ptr)
     {
         return ptr.get_unchecked(this);
     }
 
-    /*!
-    @brief access specified element via JSON Pointer
-
-    Uses a JSON pointer to retrieve a reference to the respective JSON value.
-    No bound checking is performed. The function does not change the JSON
-    value; no `null` values are created. In particular, the the special value
-    `-` yields an exception.
-
-    @param[in] ptr  JSON pointer to the desired element
-
-    @return const reference to the element pointed to by @a ptr
-
-    @complexity Constant.
-
-    @throw parse_error.106   if an array index begins with '0'
-    @throw parse_error.109   if an array index was not a number
-    @throw out_of_range.402  if the array index '-' is used
-    @throw out_of_range.404  if the JSON pointer can not be resolved
-
-    @liveexample{The behavior is shown in the example.,operatorjson_pointer_const}
-
-    @since version 2.0.0
-    */
+    /// @brief access specified element via JSON Pointer
+    /// @sa https://json.nlohmann.me/api/basic_json/operator%5B%5D/
     const_reference operator[](const json_pointer& ptr) const
     {
         return ptr.get_unchecked(this);
     }
 
-    /*!
-    @brief access specified element via JSON Pointer
-
-    Returns a reference to the element at with specified JSON pointer @a ptr,
-    with bounds checking.
-
-    @param[in] ptr  JSON pointer to the desired element
-
-    @return reference to the element pointed to by @a ptr
-
-    @throw parse_error.106 if an array index in the passed JSON pointer @a ptr
-    begins with '0'. See example below.
-
-    @throw parse_error.109 if an array index in the passed JSON pointer @a ptr
-    is not a number. See example below.
-
-    @throw out_of_range.401 if an array index in the passed JSON pointer @a ptr
-    is out of range. See example below.
-
-    @throw out_of_range.402 if the array index '-' is used in the passed JSON
-    pointer @a ptr. As `at` provides checked access (and no elements are
-    implicitly inserted), the index '-' is always invalid. See example below.
-
-    @throw out_of_range.403 if the JSON pointer describes a key of an object
-    which cannot be found. See example below.
-
-    @throw out_of_range.404 if the JSON pointer @a ptr can not be resolved.
-    See example below.
-
-    @exceptionsafety Strong guarantee: if an exception is thrown, there are no
-    changes in the JSON value.
-
-    @complexity Constant.
-
-    @since version 2.0.0
-
-    @liveexample{The behavior is shown in the example.,at_json_pointer}
-    */
+    /// @brief access specified element via JSON Pointer
+    /// @sa https://json.nlohmann.me/api/basic_json/at/
     reference at(const json_pointer& ptr)
     {
         return ptr.get_checked(this);
     }
 
-    /*!
-    @brief access specified element via JSON Pointer
-
-    Returns a const reference to the element at with specified JSON pointer @a
-    ptr, with bounds checking.
-
-    @param[in] ptr  JSON pointer to the desired element
-
-    @return reference to the element pointed to by @a ptr
-
-    @throw parse_error.106 if an array index in the passed JSON pointer @a ptr
-    begins with '0'. See example below.
-
-    @throw parse_error.109 if an array index in the passed JSON pointer @a ptr
-    is not a number. See example below.
-
-    @throw out_of_range.401 if an array index in the passed JSON pointer @a ptr
-    is out of range. See example below.
-
-    @throw out_of_range.402 if the array index '-' is used in the passed JSON
-    pointer @a ptr. As `at` provides checked access (and no elements are
-    implicitly inserted), the index '-' is always invalid. See example below.
-
-    @throw out_of_range.403 if the JSON pointer describes a key of an object
-    which cannot be found. See example below.
-
-    @throw out_of_range.404 if the JSON pointer @a ptr can not be resolved.
-    See example below.
-
-    @exceptionsafety Strong guarantee: if an exception is thrown, there are no
-    changes in the JSON value.
-
-    @complexity Constant.
-
-    @since version 2.0.0
-
-    @liveexample{The behavior is shown in the example.,at_json_pointer_const}
-    */
+    /// @brief access specified element via JSON Pointer
+    /// @sa https://json.nlohmann.me/api/basic_json/at/
     const_reference at(const json_pointer& ptr) const
     {
         return ptr.get_checked(this);
     }
 
-    /*!
-    @brief return flattened JSON value
-
-    The function creates a JSON object whose keys are JSON pointers (see [RFC
-    6901](https://tools.ietf.org/html/rfc6901)) and whose values are all
-    primitive. The original JSON value can be restored using the @ref
-    unflatten() function.
-
-    @return an object that maps JSON pointers to primitive values
-
-    @note Empty objects and arrays are flattened to `null` and will not be
-          reconstructed correctly by the @ref unflatten() function.
-
-    @complexity Linear in the size the JSON value.
-
-    @liveexample{The following code shows how a JSON object is flattened to an
-    object whose keys consist of JSON pointers.,flatten}
-
-    @sa @ref unflatten() for the reverse function
-
-    @since version 2.0.0
-    */
+    /// @brief return flattened JSON value
+    /// @sa https://json.nlohmann.me/api/basic_json/flatten/
     basic_json flatten() const
     {
         basic_json result(value_t::object);
@@ -18201,36 +21367,8 @@ class basic_json
         return result;
     }
 
-    /*!
-    @brief unflatten a previously flattened JSON value
-
-    The function restores the arbitrary nesting of a JSON value that has been
-    flattened before using the @ref flatten() function. The JSON value must
-    meet certain constraints:
-    1. The value must be an object.
-    2. The keys must be JSON pointers (see
-       [RFC 6901](https://tools.ietf.org/html/rfc6901))
-    3. The mapped values must be primitive JSON types.
-
-    @return the original JSON from a flattened version
-
-    @note Empty objects and arrays are flattened by @ref flatten() to `null`
-          values and can not unflattened to their original type. Apart from
-          this example, for a JSON value `j`, the following is always true:
-          `j == j.flatten().unflatten()`.
-
-    @complexity Linear in the size the JSON value.
-
-    @throw type_error.314  if value is not an object
-    @throw type_error.315  if object values are not primitive
-
-    @liveexample{The following code shows how a flattened JSON object is
-    unflattened into the original nested JSON object.,unflatten}
-
-    @sa @ref flatten() for the reverse function
-
-    @since version 2.0.0
-    */
+    /// @brief unflatten a previously flattened JSON value
+    /// @sa https://json.nlohmann.me/api/basic_json/unflatten/
     basic_json unflatten() const
     {
         return json_pointer::unflatten(*this);
@@ -18245,53 +21383,8 @@ class basic_json
     /// @name JSON Patch functions
     /// @{
 
-    /*!
-    @brief applies a JSON patch
-
-    [JSON Patch](http://jsonpatch.com) defines a JSON document structure for
-    expressing a sequence of operations to apply to a JSON) document. With
-    this function, a JSON Patch is applied to the current JSON value by
-    executing all operations from the patch.
-
-    @param[in] json_patch  JSON patch document
-    @return patched document
-
-    @note The application of a patch is atomic: Either all operations succeed
-          and the patched document is returned or an exception is thrown. In
-          any case, the original value is not changed: the patch is applied
-          to a copy of the value.
-
-    @throw parse_error.104 if the JSON patch does not consist of an array of
-    objects
-
-    @throw parse_error.105 if the JSON patch is malformed (e.g., mandatory
-    attributes are missing); example: `"operation add must have member path"`
-
-    @throw out_of_range.401 if an array index is out of range.
-
-    @throw out_of_range.403 if a JSON pointer inside the patch could not be
-    resolved successfully in the current JSON value; example: `"key baz not
-    found"`
-
-    @throw out_of_range.405 if JSON pointer has no parent ("add", "remove",
-    "move")
-
-    @throw other_error.501 if "test" operation was unsuccessful
-
-    @complexity Linear in the size of the JSON value and the length of the
-    JSON patch. As usually only a fraction of the JSON value is affected by
-    the patch, the complexity can usually be neglected.
-
-    @liveexample{The following code shows how a JSON patch is applied to a
-    value.,patch}
-
-    @sa @ref diff -- create a JSON patch by comparing two JSON values
-
-    @sa [RFC 6902 (JSON Patch)](https://tools.ietf.org/html/rfc6902)
-    @sa [RFC 6901 (JSON Pointer)](https://tools.ietf.org/html/rfc6901)
-
-    @since version 2.0.0
-    */
+    /// @brief applies a JSON patch
+    /// @sa https://json.nlohmann.me/api/basic_json/patch/
     basic_json patch(const basic_json& json_patch) const
     {
         // make a working copy to apply the patch to
@@ -18334,73 +21427,75 @@ class basic_json
         const auto operation_add = [&result](json_pointer & ptr, basic_json val)
         {
             // adding to the root of the target document means replacing it
-            if (ptr.is_root())
+            if (ptr.empty())
             {
                 result = val;
+                return;
             }
-            else
+
+            // make sure the top element of the pointer exists
+            json_pointer top_pointer = ptr.top();
+            if (top_pointer != ptr)
             {
-                // make sure the top element of the pointer exists
-                json_pointer top_pointer = ptr.top();
-                if (top_pointer != ptr)
+                result.at(top_pointer);
+            }
+
+            // get reference to parent of JSON pointer ptr
+            const auto last_path = ptr.back();
+            ptr.pop_back();
+            basic_json& parent = result[ptr];
+
+            switch (parent.m_type)
+            {
+                case value_t::null:
+                case value_t::object:
                 {
-                    result.at(top_pointer);
+                    // use operator[] to add value
+                    parent[last_path] = val;
+                    break;
                 }
 
-                // get reference to parent of JSON pointer ptr
-                const auto last_path = ptr.pop_back();
-                basic_json& parent = result[ptr];
-
-                switch (parent.m_type)
+                case value_t::array:
                 {
-                    case value_t::null:
-                    case value_t::object:
+                    if (last_path == "-")
                     {
-                        // use operator[] to add value
-                        parent[last_path] = val;
-                        break;
+                        // special case: append to back
+                        parent.push_back(val);
                     }
-
-                    case value_t::array:
+                    else
                     {
-                        if (last_path == "-")
+                        const auto idx = json_pointer::array_index(last_path);
+                        if (JSON_HEDLEY_UNLIKELY(idx > parent.size()))
                         {
-                            // special case: append to back
-                            parent.push_back(val);
+                            // avoid undefined behavior
+                            JSON_THROW(out_of_range::create(401, "array index " + std::to_string(idx) + " is out of range", parent));
                         }
-                        else
-                        {
-                            const auto idx = json_pointer::array_index(last_path);
-                            if (JSON_UNLIKELY(static_cast<size_type>(idx) > parent.size()))
-                            {
-                                // avoid undefined behavior
-                                JSON_THROW(out_of_range::create(401, "array index " + std::to_string(idx) + " is out of range"));
-                            }
-                            else
-                            {
-                                // default case: insert add offset
-                                parent.insert(parent.begin() + static_cast<difference_type>(idx), val);
-                            }
-                        }
-                        break;
-                    }
 
-                    // LCOV_EXCL_START
-                    default:
-                    {
-                        // if there exists a parent it cannot be primitive
-                        assert(false);
+                        // default case: insert add offset
+                        parent.insert(parent.begin() + static_cast<difference_type>(idx), val);
                     }
-                        // LCOV_EXCL_STOP
+                    break;
                 }
+
+                // if there exists a parent it cannot be primitive
+                case value_t::string: // LCOV_EXCL_LINE
+                case value_t::boolean: // LCOV_EXCL_LINE
+                case value_t::number_integer: // LCOV_EXCL_LINE
+                case value_t::number_unsigned: // LCOV_EXCL_LINE
+                case value_t::number_float: // LCOV_EXCL_LINE
+                case value_t::binary: // LCOV_EXCL_LINE
+                case value_t::discarded: // LCOV_EXCL_LINE
+                default:            // LCOV_EXCL_LINE
+                    JSON_ASSERT(false); // NOLINT(cert-dcl03-c,hicpp-static-assert,misc-static-assert) LCOV_EXCL_LINE
             }
         };
 
         // wrapper for "remove" operation; remove value at ptr
-        const auto operation_remove = [&result](json_pointer & ptr)
+        const auto operation_remove = [this, &result](json_pointer & ptr)
         {
             // get reference to parent of JSON pointer ptr
-            const auto last_path = ptr.pop_back();
+            const auto last_path = ptr.back();
+            ptr.pop_back();
             basic_json& parent = result.at(ptr);
 
             // remove child
@@ -18408,26 +21503,26 @@ class basic_json
             {
                 // perform range check
                 auto it = parent.find(last_path);
-                if (JSON_LIKELY(it != parent.end()))
+                if (JSON_HEDLEY_LIKELY(it != parent.end()))
                 {
                     parent.erase(it);
                 }
                 else
                 {
-                    JSON_THROW(out_of_range::create(403, "key '" + last_path + "' not found"));
+                    JSON_THROW(out_of_range::create(403, "key '" + last_path + "' not found", *this));
                 }
             }
             else if (parent.is_array())
             {
                 // note erase performs range check
-                parent.erase(static_cast<size_type>(json_pointer::array_index(last_path)));
+                parent.erase(json_pointer::array_index(last_path));
             }
         };
 
         // type check: top level value must be an array
-        if (JSON_UNLIKELY(not json_patch.is_array()))
+        if (JSON_HEDLEY_UNLIKELY(!json_patch.is_array()))
         {
-            JSON_THROW(parse_error::create(104, 0, "JSON patch must be an array of objects"));
+            JSON_THROW(parse_error::create(104, 0, "JSON patch must be an array of objects", json_patch));
         }
 
         // iterate and apply the operations
@@ -18445,15 +21540,17 @@ class basic_json
                 const auto error_msg = (op == "op") ? "operation" : "operation '" + op + "'";
 
                 // check if desired value is present
-                if (JSON_UNLIKELY(it == val.m_value.object->end()))
+                if (JSON_HEDLEY_UNLIKELY(it == val.m_value.object->end()))
                 {
-                    JSON_THROW(parse_error::create(105, 0, error_msg + " must have member '" + member + "'"));
+                    // NOLINTNEXTLINE(performance-inefficient-string-concatenation)
+                    JSON_THROW(parse_error::create(105, 0, error_msg + " must have member '" + member + "'", val));
                 }
 
                 // check if result is of type string
-                if (JSON_UNLIKELY(string_type and not it->second.is_string()))
+                if (JSON_HEDLEY_UNLIKELY(string_type && !it->second.is_string()))
                 {
-                    JSON_THROW(parse_error::create(105, 0, error_msg + " must have string member '" + member + "'"));
+                    // NOLINTNEXTLINE(performance-inefficient-string-concatenation)
+                    JSON_THROW(parse_error::create(105, 0, error_msg + " must have string member '" + member + "'", val));
                 }
 
                 // no error: return value
@@ -18461,14 +21558,14 @@ class basic_json
             };
 
             // type check: every element of the array must be an object
-            if (JSON_UNLIKELY(not val.is_object()))
+            if (JSON_HEDLEY_UNLIKELY(!val.is_object()))
             {
-                JSON_THROW(parse_error::create(104, 0, "JSON patch must be an array of objects"));
+                JSON_THROW(parse_error::create(104, 0, "JSON patch must be an array of objects", val));
             }
 
             // collect mandatory members
-            const std::string op = get_value("op", "op", true);
-            const std::string path = get_value(op, "path", true);
+            const auto op = get_value("op", "op", true).template get<std::string>();
+            const auto path = get_value(op, "path", true).template get<std::string>();
             json_pointer ptr(path);
 
             switch (get_op(op))
@@ -18494,7 +21591,7 @@ class basic_json
 
                 case patch_operations::move:
                 {
-                    const std::string from_path = get_value("move", "from", true);
+                    const auto from_path = get_value("move", "from", true).template get<std::string>();
                     json_pointer from_ptr(from_path);
 
                     // the "from" location must exist - use at()
@@ -18511,7 +21608,7 @@ class basic_json
 
                 case patch_operations::copy:
                 {
-                    const std::string from_path = get_value("copy", "from", true);
+                    const auto from_path = get_value("copy", "from", true).template get<std::string>();
                     const json_pointer from_ptr(from_path);
 
                     // the "from" location must exist - use at()
@@ -18539,19 +21636,20 @@ class basic_json
                     }
 
                     // throw an exception if test fails
-                    if (JSON_UNLIKELY(not success))
+                    if (JSON_HEDLEY_UNLIKELY(!success))
                     {
-                        JSON_THROW(other_error::create(501, "unsuccessful: " + val.dump()));
+                        JSON_THROW(other_error::create(501, "unsuccessful: " + val.dump(), val));
                     }
 
                     break;
                 }
 
                 case patch_operations::invalid:
+                default:
                 {
                     // op must be "add", "remove", "replace", "move", "copy", or
                     // "test"
-                    JSON_THROW(parse_error::create(105, 0, "operation value '" + op + "' is invalid"));
+                    JSON_THROW(parse_error::create(105, 0, "operation value '" + op + "' is invalid", val));
                 }
             }
         }
@@ -18559,39 +21657,9 @@ class basic_json
         return result;
     }
 
-    /*!
-    @brief creates a diff as a JSON patch
-
-    Creates a [JSON Patch](http://jsonpatch.com) so that value @a source can
-    be changed into the value @a target by calling @ref patch function.
-
-    @invariant For two JSON values @a source and @a target, the following code
-    yields always `true`:
-    @code {.cpp}
-    source.patch(diff(source, target)) == target;
-    @endcode
-
-    @note Currently, only `remove`, `add`, and `replace` operations are
-          generated.
-
-    @param[in] source  JSON value to compare from
-    @param[in] target  JSON value to compare against
-    @param[in] path    helper value to create JSON pointers
-
-    @return a JSON patch to convert the @a source to @a target
-
-    @complexity Linear in the lengths of @a source and @a target.
-
-    @liveexample{The following code shows how a JSON patch is created as a
-    diff for two JSON values.,diff}
-
-    @sa @ref patch -- apply a JSON patch
-    @sa @ref merge_patch -- apply a JSON Merge Patch
-
-    @sa [RFC 6902 (JSON Patch)](https://tools.ietf.org/html/rfc6902)
-
-    @since version 2.0.0
-    */
+    /// @brief creates a diff as a JSON patch
+    /// @sa https://json.nlohmann.me/api/basic_json/diff/
+    JSON_HEDLEY_WARN_UNUSED_RESULT
     static basic_json diff(const basic_json& source, const basic_json& target,
                            const std::string& path = "")
     {
@@ -18611,106 +21679,113 @@ class basic_json
             {
                 {"op", "replace"}, {"path", path}, {"value", target}
             });
+            return result;
         }
-        else
+
+        switch (source.type())
         {
-            switch (source.type())
+            case value_t::array:
             {
-                case value_t::array:
+                // first pass: traverse common elements
+                std::size_t i = 0;
+                while (i < source.size() && i < target.size())
                 {
-                    // first pass: traverse common elements
-                    std::size_t i = 0;
-                    while (i < source.size() and i < target.size())
-                    {
-                        // recursive call to compare array values at index i
-                        auto temp_diff = diff(source[i], target[i], path + "/" + std::to_string(i));
-                        result.insert(result.end(), temp_diff.begin(), temp_diff.end());
-                        ++i;
-                    }
-
-                    // i now reached the end of at least one array
-                    // in a second pass, traverse the remaining elements
-
-                    // remove my remaining elements
-                    const auto end_index = static_cast<difference_type>(result.size());
-                    while (i < source.size())
-                    {
-                        // add operations in reverse order to avoid invalid
-                        // indices
-                        result.insert(result.begin() + end_index, object(
-                        {
-                            {"op", "remove"},
-                            {"path", path + "/" + std::to_string(i)}
-                        }));
-                        ++i;
-                    }
-
-                    // add other remaining elements
-                    while (i < target.size())
-                    {
-                        result.push_back(
-                        {
-                            {"op", "add"},
-                            {"path", path + "/" + std::to_string(i)},
-                            {"value", target[i]}
-                        });
-                        ++i;
-                    }
-
-                    break;
+                    // recursive call to compare array values at index i
+                    auto temp_diff = diff(source[i], target[i], path + "/" + std::to_string(i));
+                    result.insert(result.end(), temp_diff.begin(), temp_diff.end());
+                    ++i;
                 }
 
-                case value_t::object:
+                // We now reached the end of at least one array
+                // in a second pass, traverse the remaining elements
+
+                // remove my remaining elements
+                const auto end_index = static_cast<difference_type>(result.size());
+                while (i < source.size())
                 {
-                    // first pass: traverse this object's elements
-                    for (auto it = source.cbegin(); it != source.cend(); ++it)
+                    // add operations in reverse order to avoid invalid
+                    // indices
+                    result.insert(result.begin() + end_index, object(
                     {
-                        // escape the key name to be used in a JSON patch
-                        const auto key = json_pointer::escape(it.key());
-
-                        if (target.find(it.key()) != target.end())
-                        {
-                            // recursive call to compare object values at key it
-                            auto temp_diff = diff(it.value(), target[it.key()], path + "/" + key);
-                            result.insert(result.end(), temp_diff.begin(), temp_diff.end());
-                        }
-                        else
-                        {
-                            // found a key that is not in o -> remove it
-                            result.push_back(object(
-                            {
-                                {"op", "remove"}, {"path", path + "/" + key}
-                            }));
-                        }
-                    }
-
-                    // second pass: traverse other object's elements
-                    for (auto it = target.cbegin(); it != target.cend(); ++it)
-                    {
-                        if (source.find(it.key()) == source.end())
-                        {
-                            // found a key that is not in this -> add it
-                            const auto key = json_pointer::escape(it.key());
-                            result.push_back(
-                            {
-                                {"op", "add"}, {"path", path + "/" + key},
-                                {"value", it.value()}
-                            });
-                        }
-                    }
-
-                    break;
+                        {"op", "remove"},
+                        {"path", path + "/" + std::to_string(i)}
+                    }));
+                    ++i;
                 }
 
-                default:
+                // add other remaining elements
+                while (i < target.size())
                 {
-                    // both primitive type: replace value
                     result.push_back(
                     {
-                        {"op", "replace"}, {"path", path}, {"value", target}
+                        {"op", "add"},
+                        {"path", path + "/-"},
+                        {"value", target[i]}
                     });
-                    break;
+                    ++i;
                 }
+
+                break;
+            }
+
+            case value_t::object:
+            {
+                // first pass: traverse this object's elements
+                for (auto it = source.cbegin(); it != source.cend(); ++it)
+                {
+                    // escape the key name to be used in a JSON patch
+                    const auto path_key = path + "/" + detail::escape(it.key());
+
+                    if (target.find(it.key()) != target.end())
+                    {
+                        // recursive call to compare object values at key it
+                        auto temp_diff = diff(it.value(), target[it.key()], path_key);
+                        result.insert(result.end(), temp_diff.begin(), temp_diff.end());
+                    }
+                    else
+                    {
+                        // found a key that is not in o -> remove it
+                        result.push_back(object(
+                        {
+                            {"op", "remove"}, {"path", path_key}
+                        }));
+                    }
+                }
+
+                // second pass: traverse other object's elements
+                for (auto it = target.cbegin(); it != target.cend(); ++it)
+                {
+                    if (source.find(it.key()) == source.end())
+                    {
+                        // found a key that is not in this -> add it
+                        const auto path_key = path + "/" + detail::escape(it.key());
+                        result.push_back(
+                        {
+                            {"op", "add"}, {"path", path_key},
+                            {"value", it.value()}
+                        });
+                    }
+                }
+
+                break;
+            }
+
+            case value_t::null:
+            case value_t::string:
+            case value_t::boolean:
+            case value_t::number_integer:
+            case value_t::number_unsigned:
+            case value_t::number_float:
+            case value_t::binary:
+            case value_t::discarded:
+            default:
+            {
+                // both primitive type: replace value
+                result.push_back(
+                {
+                    {"op", "replace"}, {"path", path}, {"value", target}
+                });
+                break;
             }
         }
 
@@ -18726,57 +21801,17 @@ class basic_json
     /// @name JSON Merge Patch functions
     /// @{
 
-    /*!
-    @brief applies a JSON Merge Patch
-
-    The merge patch format is primarily intended for use with the HTTP PATCH
-    method as a means of describing a set of modifications to a target
-    resource's content. This function applies a merge patch to the current
-    JSON value.
-
-    The function implements the following algorithm from Section 2 of
-    [RFC 7396 (JSON Merge Patch)](https://tools.ietf.org/html/rfc7396):
-
-    ```
-    define MergePatch(Target, Patch):
-      if Patch is an Object:
-        if Target is not an Object:
-          Target = {} // Ignore the contents and set it to an empty Object
-        for each Name/Value pair in Patch:
-          if Value is null:
-            if Name exists in Target:
-              remove the Name/Value pair from Target
-          else:
-            Target[Name] = MergePatch(Target[Name], Value)
-        return Target
-      else:
-        return Patch
-    ```
-
-    Thereby, `Target` is the current object; that is, the patch is applied to
-    the current value.
-
-    @param[in] patch  the patch to apply
-
-    @complexity Linear in the lengths of @a patch.
-
-    @liveexample{The following code shows how a JSON Merge Patch is applied to
-    a JSON document.,merge_patch}
-
-    @sa @ref patch -- apply a JSON patch
-    @sa [RFC 7396 (JSON Merge Patch)](https://tools.ietf.org/html/rfc7396)
-
-    @since version 3.0.0
-    */
-    void merge_patch(const basic_json& patch)
+    /// @brief applies a JSON Merge Patch
+    /// @sa https://json.nlohmann.me/api/basic_json/merge_patch/
+    void merge_patch(const basic_json& apply_patch)
     {
-        if (patch.is_object())
+        if (apply_patch.is_object())
         {
-            if (not is_object())
+            if (!is_object())
             {
                 *this = object();
             }
-            for (auto it = patch.begin(); it != patch.end(); ++it)
+            for (auto it = apply_patch.begin(); it != apply_patch.end(); ++it)
             {
                 if (it.value().is_null())
                 {
@@ -18790,44 +21825,44 @@ class basic_json
         }
         else
         {
-            *this = patch;
+            *this = apply_patch;
         }
     }
 
     /// @}
 };
+
+/// @brief user-defined to_string function for JSON values
+/// @sa https://json.nlohmann.me/api/basic_json/to_string/
+NLOHMANN_BASIC_JSON_TPL_DECLARATION
+std::string to_string(const NLOHMANN_BASIC_JSON_TPL& j)
+{
+    return j.dump();
+}
+
 } // namespace nlohmann
 
 ///////////////////////
 // nonmember support //
 ///////////////////////
 
-// specialization of std::swap, and std::hash
-namespace std
+namespace std // NOLINT(cert-dcl58-cpp)
 {
 
-/// hash value for JSON objects
-template<>
-struct hash<nlohmann::json>
+/// @brief hash value for JSON objects
+/// @sa https://json.nlohmann.me/api/basic_json/std_hash/
+NLOHMANN_BASIC_JSON_TPL_DECLARATION
+struct hash<nlohmann::NLOHMANN_BASIC_JSON_TPL>
 {
-    /*!
-    @brief return a hash value for a JSON object
-
-    @since version 1.0.0
-    */
-    std::size_t operator()(const nlohmann::json& j) const
+    std::size_t operator()(const nlohmann::NLOHMANN_BASIC_JSON_TPL& j) const
     {
-        // a naive hashing via the string representation
-        const auto& h = hash<nlohmann::json::string_t>();
-        return h(j.dump());
+        return nlohmann::detail::hash(j);
     }
 };
 
-/// specialization for std::less<value_t>
-/// @note: do not remove the space after '<',
-///        see https://github.com/nlohmann/json/pull/679
+// specialization for std::less<value_t>
 template<>
-struct less< ::nlohmann::detail::value_t>
+struct less< ::nlohmann::detail::value_t> // do not remove the space after '<', see https://github.com/nlohmann/json/pull/679
 {
     /*!
     @brief compare two value_t enum values
@@ -18840,53 +21875,34 @@ struct less< ::nlohmann::detail::value_t>
     }
 };
 
-/*!
-@brief exchanges the values of two JSON objects
+// C++20 prohibit function specialization in the std namespace.
+#ifndef JSON_HAS_CPP_20
 
-@since version 1.0.0
-*/
-template<>
-inline void swap<nlohmann::json>(nlohmann::json& j1, nlohmann::json& j2) noexcept(
-    is_nothrow_move_constructible<nlohmann::json>::value and
-    is_nothrow_move_assignable<nlohmann::json>::value
-)
+/// @brief exchanges the values of two JSON objects
+/// @sa https://json.nlohmann.me/api/basic_json/std_swap/
+NLOHMANN_BASIC_JSON_TPL_DECLARATION
+inline void swap(nlohmann::NLOHMANN_BASIC_JSON_TPL& j1, nlohmann::NLOHMANN_BASIC_JSON_TPL& j2) noexcept(  // NOLINT(readability-inconsistent-declaration-parameter-name)
+    is_nothrow_move_constructible<nlohmann::NLOHMANN_BASIC_JSON_TPL>::value&&                          // NOLINT(misc-redundant-expression)
+    is_nothrow_move_assignable<nlohmann::NLOHMANN_BASIC_JSON_TPL>::value)
 {
     j1.swap(j2);
 }
 
+#endif
+
 } // namespace std
 
-/*!
-@brief user-defined string literal for JSON values
-
-This operator implements a user-defined string literal for JSON objects. It
-can be used by adding `"_json"` to a string literal and returns a JSON object
-if no parse error occurred.
-
-@param[in] s  a string representation of a JSON object
-@param[in] n  the length of string @a s
-@return a JSON object
-
-@since version 1.0.0
-*/
+/// @brief user-defined string literal for JSON values
+/// @sa https://json.nlohmann.me/api/basic_json/operator_literal_json/
+JSON_HEDLEY_NON_NULL(1)
 inline nlohmann::json operator "" _json(const char* s, std::size_t n)
 {
     return nlohmann::json::parse(s, s + n);
 }
 
-/*!
-@brief user-defined string literal for JSON pointer
-
-This operator implements a user-defined string literal for JSON Pointers. It
-can be used by adding `"_json_pointer"` to a string literal and returns a JSON pointer
-object if no parse error occurred.
-
-@param[in] s  a string representation of a JSON Pointer
-@param[in] n  the length of string @a s
-@return a JSON pointer object
-
-@since version 2.0.0
-*/
+/// @brief user-defined string literal for JSON pointer
+/// @sa https://json.nlohmann.me/api/basic_json/operator_literal_json_pointer/
+JSON_HEDLEY_NON_NULL(1)
 inline nlohmann::json::json_pointer operator "" _json_pointer(const char* s, std::size_t n)
 {
     return nlohmann::json::json_pointer(std::string(s, n));
@@ -18895,26 +21911,181 @@ inline nlohmann::json::json_pointer operator "" _json_pointer(const char* s, std
 // #include <nlohmann/detail/macro_unscope.hpp>
 
 
-// restore GCC/clang diagnostic settings
-#if defined(__clang__) || defined(__GNUC__) || defined(__GNUG__)
-    #pragma GCC diagnostic pop
-#endif
+// restore clang diagnostic settings
 #if defined(__clang__)
-    #pragma GCC diagnostic pop
+    #pragma clang diagnostic pop
 #endif
 
 // clean up
+#undef JSON_ASSERT
 #undef JSON_INTERNAL_CATCH
 #undef JSON_CATCH
 #undef JSON_THROW
 #undef JSON_TRY
-#undef JSON_LIKELY
-#undef JSON_UNLIKELY
-#undef JSON_DEPRECATED
+#undef JSON_PRIVATE_UNLESS_TESTED
+#undef JSON_HAS_CPP_11
 #undef JSON_HAS_CPP_14
 #undef JSON_HAS_CPP_17
+#undef JSON_HAS_CPP_20
+#undef JSON_HAS_FILESYSTEM
+#undef JSON_HAS_EXPERIMENTAL_FILESYSTEM
 #undef NLOHMANN_BASIC_JSON_TPL_DECLARATION
 #undef NLOHMANN_BASIC_JSON_TPL
+#undef JSON_EXPLICIT
+#undef NLOHMANN_CAN_CALL_STD_FUNC_IMPL
+
+// #include <nlohmann/thirdparty/hedley/hedley_undef.hpp>
 
 
-#endif
\ No newline at end of file
+#undef JSON_HEDLEY_ALWAYS_INLINE
+#undef JSON_HEDLEY_ARM_VERSION
+#undef JSON_HEDLEY_ARM_VERSION_CHECK
+#undef JSON_HEDLEY_ARRAY_PARAM
+#undef JSON_HEDLEY_ASSUME
+#undef JSON_HEDLEY_BEGIN_C_DECLS
+#undef JSON_HEDLEY_CLANG_HAS_ATTRIBUTE
+#undef JSON_HEDLEY_CLANG_HAS_BUILTIN
+#undef JSON_HEDLEY_CLANG_HAS_CPP_ATTRIBUTE
+#undef JSON_HEDLEY_CLANG_HAS_DECLSPEC_DECLSPEC_ATTRIBUTE
+#undef JSON_HEDLEY_CLANG_HAS_EXTENSION
+#undef JSON_HEDLEY_CLANG_HAS_FEATURE
+#undef JSON_HEDLEY_CLANG_HAS_WARNING
+#undef JSON_HEDLEY_COMPCERT_VERSION
+#undef JSON_HEDLEY_COMPCERT_VERSION_CHECK
+#undef JSON_HEDLEY_CONCAT
+#undef JSON_HEDLEY_CONCAT3
+#undef JSON_HEDLEY_CONCAT3_EX
+#undef JSON_HEDLEY_CONCAT_EX
+#undef JSON_HEDLEY_CONST
+#undef JSON_HEDLEY_CONSTEXPR
+#undef JSON_HEDLEY_CONST_CAST
+#undef JSON_HEDLEY_CPP_CAST
+#undef JSON_HEDLEY_CRAY_VERSION
+#undef JSON_HEDLEY_CRAY_VERSION_CHECK
+#undef JSON_HEDLEY_C_DECL
+#undef JSON_HEDLEY_DEPRECATED
+#undef JSON_HEDLEY_DEPRECATED_FOR
+#undef JSON_HEDLEY_DIAGNOSTIC_DISABLE_CAST_QUAL
+#undef JSON_HEDLEY_DIAGNOSTIC_DISABLE_CPP98_COMPAT_WRAP_
+#undef JSON_HEDLEY_DIAGNOSTIC_DISABLE_DEPRECATED
+#undef JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_CPP_ATTRIBUTES
+#undef JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNKNOWN_PRAGMAS
+#undef JSON_HEDLEY_DIAGNOSTIC_DISABLE_UNUSED_FUNCTION
+#undef JSON_HEDLEY_DIAGNOSTIC_POP
+#undef JSON_HEDLEY_DIAGNOSTIC_PUSH
+#undef JSON_HEDLEY_DMC_VERSION
+#undef JSON_HEDLEY_DMC_VERSION_CHECK
+#undef JSON_HEDLEY_EMPTY_BASES
+#undef JSON_HEDLEY_EMSCRIPTEN_VERSION
+#undef JSON_HEDLEY_EMSCRIPTEN_VERSION_CHECK
+#undef JSON_HEDLEY_END_C_DECLS
+#undef JSON_HEDLEY_FLAGS
+#undef JSON_HEDLEY_FLAGS_CAST
+#undef JSON_HEDLEY_GCC_HAS_ATTRIBUTE
+#undef JSON_HEDLEY_GCC_HAS_BUILTIN
+#undef JSON_HEDLEY_GCC_HAS_CPP_ATTRIBUTE
+#undef JSON_HEDLEY_GCC_HAS_DECLSPEC_ATTRIBUTE
+#undef JSON_HEDLEY_GCC_HAS_EXTENSION
+#undef JSON_HEDLEY_GCC_HAS_FEATURE
+#undef JSON_HEDLEY_GCC_HAS_WARNING
+#undef JSON_HEDLEY_GCC_NOT_CLANG_VERSION_CHECK
+#undef JSON_HEDLEY_GCC_VERSION
+#undef JSON_HEDLEY_GCC_VERSION_CHECK
+#undef JSON_HEDLEY_GNUC_HAS_ATTRIBUTE
+#undef JSON_HEDLEY_GNUC_HAS_BUILTIN
+#undef JSON_HEDLEY_GNUC_HAS_CPP_ATTRIBUTE
+#undef JSON_HEDLEY_GNUC_HAS_DECLSPEC_ATTRIBUTE
+#undef JSON_HEDLEY_GNUC_HAS_EXTENSION
+#undef JSON_HEDLEY_GNUC_HAS_FEATURE
+#undef JSON_HEDLEY_GNUC_HAS_WARNING
+#undef JSON_HEDLEY_GNUC_VERSION
+#undef JSON_HEDLEY_GNUC_VERSION_CHECK
+#undef JSON_HEDLEY_HAS_ATTRIBUTE
+#undef JSON_HEDLEY_HAS_BUILTIN
+#undef JSON_HEDLEY_HAS_CPP_ATTRIBUTE
+#undef JSON_HEDLEY_HAS_CPP_ATTRIBUTE_NS
+#undef JSON_HEDLEY_HAS_DECLSPEC_ATTRIBUTE
+#undef JSON_HEDLEY_HAS_EXTENSION
+#undef JSON_HEDLEY_HAS_FEATURE
+#undef JSON_HEDLEY_HAS_WARNING
+#undef JSON_HEDLEY_IAR_VERSION
+#undef JSON_HEDLEY_IAR_VERSION_CHECK
+#undef JSON_HEDLEY_IBM_VERSION
+#undef JSON_HEDLEY_IBM_VERSION_CHECK
+#undef JSON_HEDLEY_IMPORT
+#undef JSON_HEDLEY_INLINE
+#undef JSON_HEDLEY_INTEL_CL_VERSION
+#undef JSON_HEDLEY_INTEL_CL_VERSION_CHECK
+#undef JSON_HEDLEY_INTEL_VERSION
+#undef JSON_HEDLEY_INTEL_VERSION_CHECK
+#undef JSON_HEDLEY_IS_CONSTANT
+#undef JSON_HEDLEY_IS_CONSTEXPR_
+#undef JSON_HEDLEY_LIKELY
+#undef JSON_HEDLEY_MALLOC
+#undef JSON_HEDLEY_MCST_LCC_VERSION
+#undef JSON_HEDLEY_MCST_LCC_VERSION_CHECK
+#undef JSON_HEDLEY_MESSAGE
+#undef JSON_HEDLEY_MSVC_VERSION
+#undef JSON_HEDLEY_MSVC_VERSION_CHECK
+#undef JSON_HEDLEY_NEVER_INLINE
+#undef JSON_HEDLEY_NON_NULL
+#undef JSON_HEDLEY_NO_ESCAPE
+#undef JSON_HEDLEY_NO_RETURN
+#undef JSON_HEDLEY_NO_THROW
+#undef JSON_HEDLEY_NULL
+#undef JSON_HEDLEY_PELLES_VERSION
+#undef JSON_HEDLEY_PELLES_VERSION_CHECK
+#undef JSON_HEDLEY_PGI_VERSION
+#undef JSON_HEDLEY_PGI_VERSION_CHECK
+#undef JSON_HEDLEY_PREDICT
+#undef JSON_HEDLEY_PRINTF_FORMAT
+#undef JSON_HEDLEY_PRIVATE
+#undef JSON_HEDLEY_PUBLIC
+#undef JSON_HEDLEY_PURE
+#undef JSON_HEDLEY_REINTERPRET_CAST
+#undef JSON_HEDLEY_REQUIRE
+#undef JSON_HEDLEY_REQUIRE_CONSTEXPR
+#undef JSON_HEDLEY_REQUIRE_MSG
+#undef JSON_HEDLEY_RESTRICT
+#undef JSON_HEDLEY_RETURNS_NON_NULL
+#undef JSON_HEDLEY_SENTINEL
+#undef JSON_HEDLEY_STATIC_ASSERT
+#undef JSON_HEDLEY_STATIC_CAST
+#undef JSON_HEDLEY_STRINGIFY
+#undef JSON_HEDLEY_STRINGIFY_EX
+#undef JSON_HEDLEY_SUNPRO_VERSION
+#undef JSON_HEDLEY_SUNPRO_VERSION_CHECK
+#undef JSON_HEDLEY_TINYC_VERSION
+#undef JSON_HEDLEY_TINYC_VERSION_CHECK
+#undef JSON_HEDLEY_TI_ARMCL_VERSION
+#undef JSON_HEDLEY_TI_ARMCL_VERSION_CHECK
+#undef JSON_HEDLEY_TI_CL2000_VERSION
+#undef JSON_HEDLEY_TI_CL2000_VERSION_CHECK
+#undef JSON_HEDLEY_TI_CL430_VERSION
+#undef JSON_HEDLEY_TI_CL430_VERSION_CHECK
+#undef JSON_HEDLEY_TI_CL6X_VERSION
+#undef JSON_HEDLEY_TI_CL6X_VERSION_CHECK
+#undef JSON_HEDLEY_TI_CL7X_VERSION
+#undef JSON_HEDLEY_TI_CL7X_VERSION_CHECK
+#undef JSON_HEDLEY_TI_CLPRU_VERSION
+#undef JSON_HEDLEY_TI_CLPRU_VERSION_CHECK
+#undef JSON_HEDLEY_TI_VERSION
+#undef JSON_HEDLEY_TI_VERSION_CHECK
+#undef JSON_HEDLEY_UNAVAILABLE
+#undef JSON_HEDLEY_UNLIKELY
+#undef JSON_HEDLEY_UNPREDICTABLE
+#undef JSON_HEDLEY_UNREACHABLE
+#undef JSON_HEDLEY_UNREACHABLE_RETURN
+#undef JSON_HEDLEY_VERSION
+#undef JSON_HEDLEY_VERSION_DECODE_MAJOR
+#undef JSON_HEDLEY_VERSION_DECODE_MINOR
+#undef JSON_HEDLEY_VERSION_DECODE_REVISION
+#undef JSON_HEDLEY_VERSION_ENCODE
+#undef JSON_HEDLEY_WARNING
+#undef JSON_HEDLEY_WARN_UNUSED_RESULT
+#undef JSON_HEDLEY_WARN_UNUSED_RESULT_MSG
+#undef JSON_HEDLEY_FALL_THROUGH
+
+
+
+#endif  // INCLUDE_NLOHMANN_JSON_HPP_

From 315ea18be2dd355d9b599a3580a2713e975c9a5f Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Mos=C3=A8=20Giordano?= <giordano@users.noreply.github.com>
Date: Tue, 14 Jun 2022 22:41:05 +0100
Subject: [PATCH 005/146] Update default value of gen-simd-width in README

---
 README.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/README.md b/README.md
index 88b922a5..4af52d78 100644
--- a/README.md
+++ b/README.md
@@ -148,7 +148,7 @@ If you want to build all the tests at once just use `make tests`.
 - `--enable-mkl[=<path>]`: use Intel MKL for FFT (and LAPACK if enabled) routines. A UNIX prefix containing the library can be specified (optional).
 - `--enable-numa`: enable NUMA first touch optimisation
 - `--enable-simd=<code>`: setup Grid for the SIMD target `<code>` (default: `GEN`). A list of possible SIMD targets is detailed in a section below.
-- `--enable-gen-simd-width=<size>`: select the size (in bytes) of the generic SIMD vector type (default: 32 bytes).
+- `--enable-gen-simd-width=<size>`: select the size (in bytes) of the generic SIMD vector type (default: 64 bytes).
 - `--enable-comms=<comm>`: Use `<comm>` for message passing (default: `none`). A list of possible SIMD targets is detailed in a section below.
 - `--enable-rng={sitmo|ranlux48|mt19937}`: choose the RNG (default: `sitmo `).
 - `--disable-timers`: disable system dependent high-resolution timers.

From 042ab1a052c9caadcfbeba427498b7719082c3db Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@users.noreply.github.com>
Date: Mon, 27 Jun 2022 13:21:39 -0400
Subject: [PATCH 006/146] Update GridStd.h

---
 Grid/GridStd.h | 1 +
 1 file changed, 1 insertion(+)

diff --git a/Grid/GridStd.h b/Grid/GridStd.h
index 28f6bc46..d0a8124a 100644
--- a/Grid/GridStd.h
+++ b/Grid/GridStd.h
@@ -16,6 +16,7 @@
 #include <functional>
 #include <stdio.h>
 #include <stdlib.h>
+#include <strings.h>
 #include <stdio.h>
 #include <signal.h>
 #include <ctime>

From f14e7e51e73abca5352ead21a1f25b8bdcd09be3 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 12 Jul 2022 10:56:22 -0700
Subject: [PATCH 007/146] Grid accelerator

---
 Grid/threads/Accelerator.cc |  2 ++
 Grid/threads/Accelerator.h  | 11 ++++++-----
 2 files changed, 8 insertions(+), 5 deletions(-)

diff --git a/Grid/threads/Accelerator.cc b/Grid/threads/Accelerator.cc
index 240758b3..1e07887b 100644
--- a/Grid/threads/Accelerator.cc
+++ b/Grid/threads/Accelerator.cc
@@ -195,12 +195,14 @@ void acceleratorInit(void)
 #ifdef GRID_SYCL
 
 cl::sycl::queue *theGridAccelerator;
+cl::sycl::queue *theCopyAccelerator;
 void acceleratorInit(void)
 {
   int nDevices = 1;
   cl::sycl::gpu_selector selector;
   cl::sycl::device selectedDevice { selector };
   theGridAccelerator = new sycl::queue (selectedDevice);
+  theCopyAccelerator = new sycl::queue (selectedDevice);
 
 #ifdef GRID_SYCL_LEVEL_ZERO_IPC
   zeInit(0);
diff --git a/Grid/threads/Accelerator.h b/Grid/threads/Accelerator.h
index 389f2cc4..132c28d2 100644
--- a/Grid/threads/Accelerator.h
+++ b/Grid/threads/Accelerator.h
@@ -262,6 +262,7 @@ NAMESPACE_END(Grid);
 NAMESPACE_BEGIN(Grid);
 
 extern cl::sycl::queue *theGridAccelerator;
+extern cl::sycl::queue *theCopyAccelerator;
 
 #ifdef __SYCL_DEVICE_ONLY__
 #define GRID_SIMT
@@ -298,7 +299,7 @@ accelerator_inline int acceleratorSIMTlane(int Nsimd) {
      });	   			              \
     });
 
-#define accelerator_barrier(dummy) theGridAccelerator->wait();
+#define accelerator_barrier(dummy) { printf(" theGridAccelerator::wait()\n"); ; theGridAccelerator->wait(); }
 
 inline void *acceleratorAllocShared(size_t bytes){ return malloc_shared(bytes,*theGridAccelerator);};
 inline void *acceleratorAllocDevice(size_t bytes){ return malloc_device(bytes,*theGridAccelerator);};
@@ -307,10 +308,10 @@ inline void acceleratorFreeDevice(void *ptr){free(ptr,*theGridAccelerator);};
 inline void acceleratorCopyDeviceToDeviceAsynch(void *from,void *to,size_t bytes)  {
   theGridAccelerator->memcpy(to,from,bytes);
 }
-inline void acceleratorCopySynchronise(void) {  theGridAccelerator->wait(); std::cout<<"acceleratorCopySynchronise() wait "<<std::endl; }
-inline void acceleratorCopyToDevice(void *from,void *to,size_t bytes)  { theGridAccelerator->memcpy(to,from,bytes); theGridAccelerator->wait();}
-inline void acceleratorCopyFromDevice(void *from,void *to,size_t bytes){ theGridAccelerator->memcpy(to,from,bytes); theGridAccelerator->wait();}
-inline void acceleratorMemSet(void *base,int value,size_t bytes) { theGridAccelerator->memset(base,value,bytes); theGridAccelerator->wait();}
+inline void acceleratorCopySynchronise(void) {  theCopyAccelerator->wait(); std::cout<<"acceleratorCopySynchronise() wait "<<std::endl; }
+inline void acceleratorCopyToDevice(void *from,void *to,size_t bytes)  { theCopyAccelerator->memcpy(to,from,bytes); theCopyAccelerator->wait();}
+inline void acceleratorCopyFromDevice(void *from,void *to,size_t bytes){ theCopyAccelerator->memcpy(to,from,bytes); theCopyAccelerator->wait();}
+inline void acceleratorMemSet(void *base,int value,size_t bytes) { theCopyAccelerator->memset(base,value,bytes); theCopyAccelerator->wait();}
 inline int  acceleratorIsCommunicable(void *ptr)
 {
 #if 0

From 5f8892bf03b3579b912e66e07917e4f936dddd7e Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 19 Jul 2022 09:31:51 -0700
Subject: [PATCH 008/146] Mistake pointed out by Camilo

---
 Grid/threads/Accelerator.h | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/Grid/threads/Accelerator.h b/Grid/threads/Accelerator.h
index 132c28d2..a7fd8db7 100644
--- a/Grid/threads/Accelerator.h
+++ b/Grid/threads/Accelerator.h
@@ -305,13 +305,13 @@ inline void *acceleratorAllocShared(size_t bytes){ return malloc_shared(bytes,*t
 inline void *acceleratorAllocDevice(size_t bytes){ return malloc_device(bytes,*theGridAccelerator);};
 inline void acceleratorFreeShared(void *ptr){free(ptr,*theGridAccelerator);};
 inline void acceleratorFreeDevice(void *ptr){free(ptr,*theGridAccelerator);};
-inline void acceleratorCopyDeviceToDeviceAsynch(void *from,void *to,size_t bytes)  {
-  theGridAccelerator->memcpy(to,from,bytes);
-}
-inline void acceleratorCopySynchronise(void) {  theCopyAccelerator->wait(); std::cout<<"acceleratorCopySynchronise() wait "<<std::endl; }
+
+inline void acceleratorCopyDeviceToDeviceAsynch(void *from,void *to,size_t bytes)  {  theCopyAccelerator->memcpy(to,from,bytes);}
+inline void acceleratorCopySynchronise(void) {  theCopyAccelerator->wait(); }
 inline void acceleratorCopyToDevice(void *from,void *to,size_t bytes)  { theCopyAccelerator->memcpy(to,from,bytes); theCopyAccelerator->wait();}
 inline void acceleratorCopyFromDevice(void *from,void *to,size_t bytes){ theCopyAccelerator->memcpy(to,from,bytes); theCopyAccelerator->wait();}
 inline void acceleratorMemSet(void *base,int value,size_t bytes) { theCopyAccelerator->memset(base,value,bytes); theCopyAccelerator->wait();}
+
 inline int  acceleratorIsCommunicable(void *ptr)
 {
 #if 0

From 2ab1af5754fb3007930ae7cee20395cf1363d61b Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 19 Jul 2022 09:51:06 -0700
Subject: [PATCH 009/146] Ensure no synchronize and not optoin dependent

---
 Grid/communicator/Communicator_mpi3.cc | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/Grid/communicator/Communicator_mpi3.cc b/Grid/communicator/Communicator_mpi3.cc
index 7b3e8847..b8ce7bca 100644
--- a/Grid/communicator/Communicator_mpi3.cc
+++ b/Grid/communicator/Communicator_mpi3.cc
@@ -392,9 +392,9 @@ double CartesianCommunicator::StencilSendToRecvFromBegin(std::vector<CommsReques
     acceleratorCopyDeviceToDeviceAsynch(xmit,shm,bytes);
   }
 
-  if ( CommunicatorPolicy == CommunicatorPolicySequential ) {
-    this->StencilSendToRecvFromComplete(list,dir);
-  }
+  //  if ( CommunicatorPolicy == CommunicatorPolicySequential ) {
+  //    this->StencilSendToRecvFromComplete(list,dir);
+  //  }
 
   return off_node_bytes;
 }

From d32b923b6c9724f57e8cb0fa27e0b7aa2c753cf0 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 2 Aug 2022 07:58:04 -0700
Subject: [PATCH 010/146] Fencing on a stream in SYCL is needed. Didn't know
 that ... gulp

---
 .../implementation/WilsonKernelsImplementation.h      |  3 +++
 Grid/threads/Accelerator.cc                           |  3 ++-
 Grid/threads/Accelerator.h                            | 11 ++++++++---
 3 files changed, 13 insertions(+), 4 deletions(-)

diff --git a/Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h b/Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h
index 9228b84c..9f6960af 100644
--- a/Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h
+++ b/Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h
@@ -498,6 +498,7 @@ void WilsonKernels<Impl>::DhopKernel(int Opt,StencilImpl &st,  DoubledGaugeField
 #ifndef GRID_CUDA
      if (Opt == WilsonKernelsStatic::OptInlineAsm  ) {  ASM_CALL(AsmDhopSiteDag);     return;}
 #endif
+     acceleratorFenceComputeStream();
    } else if( interior ) {
      if (Opt == WilsonKernelsStatic::OptGeneric    ) { KERNEL_CALL(GenericDhopSiteDagInt); return;}
      if (Opt == WilsonKernelsStatic::OptHandUnroll ) { KERNEL_CALL(HandDhopSiteDagInt);    return;}
@@ -505,11 +506,13 @@ void WilsonKernels<Impl>::DhopKernel(int Opt,StencilImpl &st,  DoubledGaugeField
      if (Opt == WilsonKernelsStatic::OptInlineAsm  ) {  ASM_CALL(AsmDhopSiteDagInt);     return;}
 #endif
    } else if( exterior ) {
+     acceleratorFenceComputeStream();
      if (Opt == WilsonKernelsStatic::OptGeneric    ) { KERNEL_CALL(GenericDhopSiteDagExt); return;}
      if (Opt == WilsonKernelsStatic::OptHandUnroll ) { KERNEL_CALL(HandDhopSiteDagExt);    return;}
 #ifndef GRID_CUDA
      if (Opt == WilsonKernelsStatic::OptInlineAsm  ) {  ASM_CALL(AsmDhopSiteDagExt);     return;}
 #endif
+     acceleratorFenceComputeStream();
    }
    assert(0 && " Kernel optimisation case not covered ");
   }
diff --git a/Grid/threads/Accelerator.cc b/Grid/threads/Accelerator.cc
index 1e07887b..163e4ac4 100644
--- a/Grid/threads/Accelerator.cc
+++ b/Grid/threads/Accelerator.cc
@@ -202,7 +202,8 @@ void acceleratorInit(void)
   cl::sycl::gpu_selector selector;
   cl::sycl::device selectedDevice { selector };
   theGridAccelerator = new sycl::queue (selectedDevice);
-  theCopyAccelerator = new sycl::queue (selectedDevice);
+  //  theCopyAccelerator = new sycl::queue (selectedDevice);
+  theCopyAccelerator = theGridAccelerator; // Should proceed concurrenlty anyway.
 
 #ifdef GRID_SYCL_LEVEL_ZERO_IPC
   zeInit(0);
diff --git a/Grid/threads/Accelerator.h b/Grid/threads/Accelerator.h
index a7fd8db7..c7366a1f 100644
--- a/Grid/threads/Accelerator.h
+++ b/Grid/threads/Accelerator.h
@@ -247,7 +247,12 @@ inline int  acceleratorIsCommunicable(void *ptr)
 //////////////////////////////////////////////
 // SyCL acceleration
 //////////////////////////////////////////////
-
+#ifdef GRID_SYCL
+inline void acceleratorFenceComputeStream(void){ accelerator_barrier();};
+#else
+// Ordering within a stream guaranteed on Nvidia & AMD
+inline void acceleratorFenceComputeStream(void){ };
+#endif
 #ifdef GRID_SYCL
 NAMESPACE_END(Grid);
 #include <CL/sycl.hpp>
@@ -299,15 +304,15 @@ accelerator_inline int acceleratorSIMTlane(int Nsimd) {
      });	   			              \
     });
 
-#define accelerator_barrier(dummy) { printf(" theGridAccelerator::wait()\n"); ; theGridAccelerator->wait(); }
+#define accelerator_barrier(dummy) { printf(" theGridAccelerator::wait()\n");  theGridAccelerator->wait(); }
 
 inline void *acceleratorAllocShared(size_t bytes){ return malloc_shared(bytes,*theGridAccelerator);};
 inline void *acceleratorAllocDevice(size_t bytes){ return malloc_device(bytes,*theGridAccelerator);};
 inline void acceleratorFreeShared(void *ptr){free(ptr,*theGridAccelerator);};
 inline void acceleratorFreeDevice(void *ptr){free(ptr,*theGridAccelerator);};
 
+inline void acceleratorCopySynchronise(void) {  printf(" theCopyAccelerator::wait()\n"); theCopyAccelerator->wait(); }
 inline void acceleratorCopyDeviceToDeviceAsynch(void *from,void *to,size_t bytes)  {  theCopyAccelerator->memcpy(to,from,bytes);}
-inline void acceleratorCopySynchronise(void) {  theCopyAccelerator->wait(); }
 inline void acceleratorCopyToDevice(void *from,void *to,size_t bytes)  { theCopyAccelerator->memcpy(to,from,bytes); theCopyAccelerator->wait();}
 inline void acceleratorCopyFromDevice(void *from,void *to,size_t bytes){ theCopyAccelerator->memcpy(to,from,bytes); theCopyAccelerator->wait();}
 inline void acceleratorMemSet(void *base,int value,size_t bytes) { theCopyAccelerator->memset(base,value,bytes); theCopyAccelerator->wait();}

From 84110166e420f2a5eddb4ee4bb38a386f087c299 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 2 Aug 2022 08:00:43 -0700
Subject: [PATCH 011/146] Fix the fence

---
 Grid/threads/Accelerator.h | 15 +++++++++------
 1 file changed, 9 insertions(+), 6 deletions(-)

diff --git a/Grid/threads/Accelerator.h b/Grid/threads/Accelerator.h
index c7366a1f..c97c91a3 100644
--- a/Grid/threads/Accelerator.h
+++ b/Grid/threads/Accelerator.h
@@ -248,12 +248,6 @@ inline int  acceleratorIsCommunicable(void *ptr)
 // SyCL acceleration
 //////////////////////////////////////////////
 #ifdef GRID_SYCL
-inline void acceleratorFenceComputeStream(void){ accelerator_barrier();};
-#else
-// Ordering within a stream guaranteed on Nvidia & AMD
-inline void acceleratorFenceComputeStream(void){ };
-#endif
-#ifdef GRID_SYCL
 NAMESPACE_END(Grid);
 #include <CL/sycl.hpp>
 #include <CL/sycl/usm.hpp>
@@ -517,7 +511,16 @@ inline void *acceleratorAllocCpu(size_t bytes){return memalign(GRID_ALLOC_ALIGN,
 inline void acceleratorFreeCpu  (void *ptr){free(ptr);};
 #endif
 
+//////////////////////////////////////////////
+// Fencing needed ONLY for SYCL
+//////////////////////////////////////////////
 
+#ifdef GRID_SYCL
+inline void acceleratorFenceComputeStream(void){ accelerator_barrier();};
+#else
+// Ordering within a stream guaranteed on Nvidia & AMD
+inline void acceleratorFenceComputeStream(void){ };
+#endif
 
 ///////////////////////////////////////////////////
 // Synchronise across local threads for divergence resynch

From bb0a0da47a4035ffc0fdbfbb61411979aefeae5f Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 2 Aug 2022 08:09:43 -0700
Subject: [PATCH 012/146] inon blocking caution due to SYCL

---
 Grid/algorithms/CoarsenedMatrix.h  | 2 +-
 Grid/algorithms/approx/Chebyshev.h | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/Grid/algorithms/CoarsenedMatrix.h b/Grid/algorithms/CoarsenedMatrix.h
index a9a82f34..ba4abecd 100644
--- a/Grid/algorithms/CoarsenedMatrix.h
+++ b/Grid/algorithms/CoarsenedMatrix.h
@@ -262,7 +262,7 @@ public:
 	autoView( Tnp_v , (*Tnp), AcceleratorWrite);
 	autoView( Tnm_v , (*Tnm), AcceleratorWrite);
 	const int Nsimd = CComplex::Nsimd();
-	accelerator_forNB(ss, FineGrid->oSites(), Nsimd, {
+	accelerator_for(ss, FineGrid->oSites(), Nsimd, {
 	  coalescedWrite(y_v[ss],xscale*y_v(ss)+mscale*Tn_v(ss));
 	  coalescedWrite(Tnp_v[ss],2.0*y_v(ss)-Tnm_v(ss));
         });
diff --git a/Grid/algorithms/approx/Chebyshev.h b/Grid/algorithms/approx/Chebyshev.h
index 584ed1d5..7c93f0b8 100644
--- a/Grid/algorithms/approx/Chebyshev.h
+++ b/Grid/algorithms/approx/Chebyshev.h
@@ -264,7 +264,7 @@ public:
       auto Tnp_v = Tnp->View();
       auto Tnm_v = Tnm->View();
       constexpr int Nsimd = vector_type::Nsimd();
-      accelerator_forNB(ss, in.Grid()->oSites(), Nsimd, {
+      accelerator_for(ss, in.Grid()->oSites(), Nsimd, {
 	  coalescedWrite(y_v[ss],xscale*y_v(ss)+mscale*Tn_v(ss));
 	  coalescedWrite(Tnp_v[ss],2.0*y_v(ss)-Tnm_v(ss));
       });

From 17d71771056264749b91be40d28fdb48a740f264 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 2 Aug 2022 08:33:39 -0700
Subject: [PATCH 013/146] Files for SYCL

---
 systems/PVC/benchmarks/run-1tile.sh     | 62 +++++++++++++++++++++++++
 systems/PVC/benchmarks/run-2tile-mpi.sh | 26 +++++++++++
 systems/PVC/benchmarks/wrap.sh          | 14 ++++++
 systems/PVC/config-command              | 15 ++++++
 systems/PVC/setup.sh                    | 11 +++++
 5 files changed, 128 insertions(+)
 create mode 100644 systems/PVC/benchmarks/run-1tile.sh
 create mode 100755 systems/PVC/benchmarks/run-2tile-mpi.sh
 create mode 100755 systems/PVC/benchmarks/wrap.sh
 create mode 100644 systems/PVC/config-command
 create mode 100644 systems/PVC/setup.sh

diff --git a/systems/PVC/benchmarks/run-1tile.sh b/systems/PVC/benchmarks/run-1tile.sh
new file mode 100644
index 00000000..923afd84
--- /dev/null
+++ b/systems/PVC/benchmarks/run-1tile.sh
@@ -0,0 +1,62 @@
+#!/bin/sh
+##SBATCH -p PVC-SPR-QZEH 
+##SBATCH -p PVC-ICX-QZNW
+#SBATCH -p QZ1J-ICX-PVC
+##SBATCH -p QZ1J-SPR-PVC-2C
+
+source /nfs/site/home/paboylex/ATS/GridNew/Grid/systems/PVC-nightly/setup.sh
+
+export NT=16
+
+export I_MPI_OFFLOAD=1
+export I_MPI_OFFLOAD_TOPOLIB=level_zero
+export I_MPI_OFFLOAD_DOMAIN_SIZE=-1
+
+# export IGC_EnableLSCFenceUGMBeforeEOT=0
+# export SYCL_PROGRAM_COMPILE_OPTIONS="-ze-opt-large-register-file=False"
+export SYCL_DEVICE_FILTER=gpu,level_zero
+#export IGC_ShaderDumpEnable=1 
+#export IGC_DumpToCurrentDir=1
+export I_MPI_OFFLOAD_CELL=tile
+export EnableImplicitScaling=0
+export EnableWalkerPartition=0
+export ZE_AFFINITY_MASK=0.0
+mpiexec -launcher ssh -n 1 -host localhost  ./Benchmark_dwf_fp32 --mpi 1.1.1.1 --grid 32.32.32.32 --accelerator-threads $NT --comms-sequential --shm-mpi 1 --cacheblocking 8.8.8.8
+
+export ZE_AFFINITY_MASK=0
+export I_MPI_OFFLOAD_CELL=device
+export EnableImplicitScaling=1
+export EnableWalkerPartition=1
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+#mpiexec -launcher ssh -n 2 -host localhost  vtune -collect gpu-hotspots -knob gpu-sampling-interval=1 -data-limit=0 -r ./vtune_run4 -- ./wrap.sh ./Benchmark_dwf_fp32 --mpi 2.1.1.1 --grid 64.32.32.32 --accelerator-threads $NT --comms-overlap --shm-mpi 1
+
+#mpiexec  -launcher ssh -n 1 -host localhost ./wrap.sh ./Benchmark_dwf_fp32 --mpi 1.1.1.1 --grid 64.32.32.32 --accelerator-threads $NT --comms-overlap --shm-mpi 1
+
+#mpiexec  -launcher ssh -n 2 -host localhost ./wrap.sh ./Benchmark_dwf_fp32 --mpi 2.1.1.1 --grid 64.32.32.32 --accelerator-threads $NT --comms-sequential --shm-mpi 1
+
+#mpiexec  -launcher ssh -n 2 -host localhost ./wrap.sh ./Benchmark_dwf_fp32 --mpi 2.1.1.1 --grid 64.32.32.32 --accelerator-threads $NT --comms-overlap --shm-mpi 1
+
+#mpiexec  -launcher ssh -n 2 -host localhost ./wrap.sh ./Benchmark_dwf_fp32 --mpi 2.1.1.1 --grid 64.32.32.32 --accelerator-threads $NT --comms-sequential --shm-mpi 0
+
+#mpirun -np 2 ./wrap.sh ./Benchmark_dwf_fp32 --mpi 1.1.1.2 --grid 16.32.32.64 --accelerator-threads $NT --comms-sequential --shm-mpi 0
+#mpirun -np 2 ./wrap.sh ./Benchmark_dwf_fp32 --mpi 1.1.1.2 --grid 32.32.32.64 --accelerator-threads $NT --comms-sequential --shm-mpi 1
+
diff --git a/systems/PVC/benchmarks/run-2tile-mpi.sh b/systems/PVC/benchmarks/run-2tile-mpi.sh
new file mode 100755
index 00000000..9db0b66b
--- /dev/null
+++ b/systems/PVC/benchmarks/run-2tile-mpi.sh
@@ -0,0 +1,26 @@
+#!/bin/bash
+##SBATCH -p PVC-SPR-QZEH 
+##SBATCH -p PVC-ICX-QZNW
+#SBATCH -p QZ1J-ICX-PVC
+
+source /nfs/site/home/paboylex/ATS/GridNew/Grid/systems/PVC-nightly/setup.sh
+
+export NT=16
+
+
+# export IGC_EnableLSCFenceUGMBeforeEOT=0
+# export SYCL_PROGRAM_COMPILE_OPTIONS="-ze-opt-large-register-file=False"
+#export IGC_ShaderDumpEnable=1 
+#export IGC_DumpToCurrentDir=1
+export I_MPI_OFFLOAD=1
+export I_MPI_OFFLOAD_TOPOLIB=level_zero
+export I_MPI_OFFLOAD_DOMAIN_SIZE=-1
+export SYCL_DEVICE_FILTER=gpu,level_zero
+export I_MPI_OFFLOAD_CELL=tile
+export EnableImplicitScaling=0
+export EnableWalkerPartition=0
+
+mpiexec -launcher ssh -n 1 -host localhost  ./wrap.sh ./Benchmark_dwf_fp32 --mpi 1.1.1.1 --grid 32.32.32.32 --accelerator-threads $NT --comms-sequential --shm-mpi 1 > 1tile.log
+
+mpiexec -launcher ssh -n 2 -host localhost  ./wrap.sh ./Benchmark_dwf_fp32 --mpi 2.1.1.1 --grid 64.32.32.32 --accelerator-threads $NT --comms-sequential --shm-mpi 1 > 2tile.log
+
diff --git a/systems/PVC/benchmarks/wrap.sh b/systems/PVC/benchmarks/wrap.sh
new file mode 100755
index 00000000..a352fff9
--- /dev/null
+++ b/systems/PVC/benchmarks/wrap.sh
@@ -0,0 +1,14 @@
+#!/bin/sh
+
+export ZE_AFFINITY_MASK=0.$MPI_LOCALRANKID
+
+echo Ranke $MPI_LOCALRANKID ZE_AFFINITY_MASK is $ZE_AFFINITY_MASK
+
+
+if [ $MPI_LOCALRANKID = "0" ] 
+then
+#  ~psteinbr/build_pti/ze_tracer -c $@
+  onetrace --chrome-kernel-timeline $@
+else
+  $@
+fi
diff --git a/systems/PVC/config-command b/systems/PVC/config-command
new file mode 100644
index 00000000..3f5b5993
--- /dev/null
+++ b/systems/PVC/config-command
@@ -0,0 +1,15 @@
+INSTALL=/nfs/site/home/azusayax/install
+../../configure \
+	--enable-simd=GPU \
+	--enable-gen-simd-width=64 \
+	--enable-comms=mpi \
+	--disable-accelerator-cshift \
+	--disable-gparity \
+	--disable-fermion-reps \
+	--enable-shm=nvlink \
+	--enable-accelerator=sycl \
+	--enable-unified=yes \
+	CXX=mpicxx \
+	LDFLAGS="-fsycl-device-code-split=per_kernel -fsycl-device-lib=all -lze_loader -L$INSTALL/lib" \
+	CXXFLAGS="-cxx=dpcpp -fsycl-unnamed-lambda -fsycl -no-fma -I$INSTALL/include -Wtautological-constant-compare"
+
diff --git a/systems/PVC/setup.sh b/systems/PVC/setup.sh
new file mode 100644
index 00000000..2a6f920b
--- /dev/null
+++ b/systems/PVC/setup.sh
@@ -0,0 +1,11 @@
+export https_proxy=http://proxy-chain.intel.com:911
+export LD_LIBRARY_PATH=/nfs/site/home/azusayax/install/lib:$LD_LIBRARY_PATH
+
+module load intel-release
+source /opt/intel/oneapi/PVC_setup.sh
+#source /opt/intel/oneapi/ATS_setup.sh
+module load intel/mpich/pvc45.3
+export PATH=~/ATS/pti-gpu/tools/onetrace/:$PATH
+
+#clsh embargo-ci-neo-022845
+#source /opt/intel/vtune_amplifier/amplxe-vars.sh

From 188d2c7a4dc77807b545f5f2813cdb589b9e44ca Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 2 Aug 2022 08:38:53 -0700
Subject: [PATCH 014/146] PVC default, ignore ATS

---
 Grid/threads/Accelerator.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/Grid/threads/Accelerator.h b/Grid/threads/Accelerator.h
index c97c91a3..d2cf8a01 100644
--- a/Grid/threads/Accelerator.h
+++ b/Grid/threads/Accelerator.h
@@ -289,7 +289,7 @@ accelerator_inline int acceleratorSIMTlane(int Nsimd) {
       cgh.parallel_for(					\
       cl::sycl::nd_range<3>(global,local), \
       [=] (cl::sycl::nd_item<3> item) /*mutable*/     \
-      [[intel::reqd_sub_group_size(8)]]	      \
+      [[intel::reqd_sub_group_size(16)]]	      \
       {						      \
       auto iter1    = item.get_global_id(0);	      \
       auto iter2    = item.get_global_id(1);	      \

From f922adf05e418312ddcb3e024c6bca179d7c8865 Mon Sep 17 00:00:00 2001
From: Gurtej Kanwar <gurtej@mit.edu>
Date: Sun, 21 Aug 2022 16:16:18 +0200
Subject: [PATCH 015/146] Fix Photon ComplexField type

---
 Grid/qcd/action/gauge/Photon.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/Grid/qcd/action/gauge/Photon.h b/Grid/qcd/action/gauge/Photon.h
index 465aa8bd..3d4baccd 100644
--- a/Grid/qcd/action/gauge/Photon.h
+++ b/Grid/qcd/action/gauge/Photon.h
@@ -49,7 +49,7 @@ NAMESPACE_BEGIN(Grid);
     
     typedef Lattice<SiteLink>  LinkField;
     typedef Lattice<SiteField> Field;
-    typedef Field              ComplexField;
+    typedef LinkField          ComplexField;
   };
   
   typedef QedGImpl<vComplex> QedGImplR;

From 554c2383593a85695f716eaebfcad539210ff689 Mon Sep 17 00:00:00 2001
From: Gurtej Kanwar <gurtej@mit.edu>
Date: Sun, 21 Aug 2022 17:28:57 +0200
Subject: [PATCH 016/146] Update OpenSSL digest to use high-level methods

This avoids deprecation warnings when compiling against OpenSSL 3.0
but should still be backwards compatible. It is the recommended way
to use the digest API going forward.
---
 Grid/util/Sha.h | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)

diff --git a/Grid/util/Sha.h b/Grid/util/Sha.h
index ee164c34..f3789ac4 100644
--- a/Grid/util/Sha.h
+++ b/Grid/util/Sha.h
@@ -27,6 +27,7 @@
     /*  END LEGAL */
 extern "C" {
 #include <openssl/sha.h>
+#include <openssl/evp.h>
 }
 #ifdef USE_IPP
 #include "ipp.h"
@@ -70,10 +71,8 @@ public:
   static inline std::vector<unsigned char> sha256(const void *data,size_t bytes)
   {
     std::vector<unsigned char> hash(SHA256_DIGEST_LENGTH);
-    SHA256_CTX sha256;
-    SHA256_Init  (&sha256);
-    SHA256_Update(&sha256, data,bytes);
-    SHA256_Final (&hash[0], &sha256);
+    auto digest = EVP_get_digestbyname("SHA256");
+    EVP_Digest(data, bytes, &hash[0], NULL, digest, NULL);
     return hash;
   }
   static inline std::vector<int> sha256_seeds(const std::string &s)

From 60dfb49afa48b75455fe076e25f67d09bef4a74e Mon Sep 17 00:00:00 2001
From: Gurtej Kanwar <gurtej@mit.edu>
Date: Sun, 21 Aug 2022 17:29:55 +0200
Subject: [PATCH 017/146] Remove FP16 tests when FP16 is disabled

---
 tests/Test_simd.cc | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/tests/Test_simd.cc b/tests/Test_simd.cc
index 468bc982..16205ee1 100644
--- a/tests/Test_simd.cc
+++ b/tests/Test_simd.cc
@@ -793,6 +793,7 @@ int main (int argc, char ** argv)
     }
     std::cout <<" OK ! "<<std::endl;
 
+#ifdef USE_FP16
     // Double to Half
     std::cout << GridLogMessage<< "Double to half" ;
     precisionChange(&H[0],&D[0],Ndp);
@@ -822,6 +823,7 @@ int main (int argc, char ** argv)
       assert( tmp < 1.0e-3 );
     }
     std::cout <<" OK ! "<<std::endl;
+#endif
 
   }
   Grid_finalize();

From 1713de35c0dc339564661dd7df8a72583f889e91 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Mon, 5 Sep 2022 21:50:02 -0400
Subject: [PATCH 018/146] Improved config flags

---
 systems/Crusher/config-command | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/systems/Crusher/config-command b/systems/Crusher/config-command
index c93ea9c8..dfd8d127 100644
--- a/systems/Crusher/config-command
+++ b/systems/Crusher/config-command
@@ -7,7 +7,7 @@
 --disable-fermion-reps \
 --with-gmp=$OLCF_GMP_ROOT \
 --with-mpfr=/opt/cray/pe/gcc/mpfr/3.1.4/ \
---disable-gparity \
+--enable-gparity \
 CXX=hipcc MPICXX=mpicxx \
 CXXFLAGS="-fPIC -I{$ROCM_PATH}/include/ -std=c++14 -I${MPICH_DIR}/include " \
  LDFLAGS=" -L${MPICH_DIR}/lib -lmpi -L${CRAY_MPICH_ROOTDIR}/gtl/lib -lmpi_gtl_hsa -lamdhip64 "

From 5b128a6f9f21c7f9309b9c7f019ff28d42b8df49 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Fri, 23 Sep 2022 16:18:47 -0400
Subject: [PATCH 019/146] MixedPrec Multishift with better precision scheme for
 GPU

---
 .../ConjugateGradientMultiShiftMixedPrec.h    | 52 +++++++------------
 1 file changed, 20 insertions(+), 32 deletions(-)

diff --git a/Grid/algorithms/iterative/ConjugateGradientMultiShiftMixedPrec.h b/Grid/algorithms/iterative/ConjugateGradientMultiShiftMixedPrec.h
index b1d90688..8a88cb07 100644
--- a/Grid/algorithms/iterative/ConjugateGradientMultiShiftMixedPrec.h
+++ b/Grid/algorithms/iterative/ConjugateGradientMultiShiftMixedPrec.h
@@ -164,12 +164,8 @@ public:
     RealD cp,bp,qq; //prev
   
     // Matrix mult fields
-    FieldF r_f(SinglePrecGrid);
     FieldF p_f(SinglePrecGrid);
-    FieldF tmp_f(SinglePrecGrid);
     FieldF mmp_f(SinglePrecGrid);
-    FieldF src_f(SinglePrecGrid);
-    precisionChange(src_f, src_d);
 
     // Check lightest mass
     for(int s=0;s<nshift;s++){
@@ -198,14 +194,13 @@ public:
       ps_d[s] = src_d;
     }
     // r and p for primary
-    r_f=src_f; //residual maintained in single
-    p_f=src_f;
     p_d = src_d; //primary copy --- make this a reference to ps_d to save axpys
-  
+    r_d = p_d;
+    
     //MdagM+m[0]
-    Linop_f.HermOpAndNorm(p_f,mmp_f,d,qq); // mmp = MdagM p        d=real(dot(p, mmp)),  qq=norm2(mmp)
-    axpy(mmp_f,mass[0],p_f,mmp_f);
-    RealD rn = norm2(p_f);
+    Linop_d.HermOpAndNorm(p_d,mmp_d,d,qq); // mmp = MdagM p        d=real(dot(p, mmp)),  qq=norm2(mmp)
+    axpy(mmp_d,mass[0],p_d,mmp_d);
+    RealD rn = norm2(p_d);
     d += rn*mass[0];
 
     b = -cp /d;
@@ -223,7 +218,7 @@ public:
   
     // r += b[0] A.p[0]
     // c= norm(r)
-    c=axpy_norm(r_f,b,mmp_f,r_f);
+    c=axpy_norm(r_d,b,mmp_d,r_d);
   
     for(int s=0;s<nshift;s++) {
       axpby(psi_d[s],0.,-bs[s]*alpha[s],src_d,src_d);
@@ -240,13 +235,8 @@ public:
     int k;
   
     for (k=1;k<=MaxIterations;k++){    
+
       a = c /cp;
-
-      //Update double precision search direction by residual
-      PrecChangeTimer.Start();
-      precisionChange(r_d, r_f);
-      PrecChangeTimer.Stop();
-
       AXPYTimer.Start();
       axpy(p_d,a,p_d,r_d); 
 
@@ -263,19 +253,23 @@ public:
       AXPYTimer.Stop();
 
       PrecChangeTimer.Start();
-      precisionChange(p_f, p_d); //get back single prec search direction for linop
+      precisionChangeFast(p_f, p_d); //get back single prec search direction for linop
       PrecChangeTimer.Stop();
 
       cp=c;
       MatrixTimer.Start();  
-      Linop_f.HermOp(p_f,mmp_f); 
-      d=real(innerProduct(p_f,mmp_f));    
+      Linop_f.HermOp(p_f,mmp_f);
       MatrixTimer.Stop();  
 
+      PrecChangeTimer.Start();
+      precisionChangeFast(mmp_d, mmp_f); // From Float to Double
+      PrecChangeTimer.Stop();
+
       AXPYTimer.Start();
-      axpy(mmp_f,mass[0],p_f,mmp_f);
+      d=real(innerProduct(p_d,mmp_d));    
+      axpy(mmp_d,mass[0],p_d,mmp_d);
       AXPYTimer.Stop();
-      RealD rn = norm2(p_f);
+      RealD rn = norm2(p_d);
       d += rn*mass[0];
     
       bp=b;
@@ -306,12 +300,11 @@ public:
       }
 
       //Perform reliable update if necessary; otherwise update residual from single-prec mmp
-      RealD c_f = axpy_norm(r_f,b,mmp_f,r_f);
+      c = axpy_norm(r_d,b,mmp_d,r_d);
       AXPYTimer.Stop();
 
-      c = c_f;
-
       if(k % ReliableUpdateFreq == 0){
+	RealD c_old = c;
 	//Replace r with true residual
 	MatrixTimer.Start();  
 	Linop_d.HermOp(psi_d[0],mmp_d); 
@@ -320,15 +313,10 @@ public:
 	AXPYTimer.Start();
 	axpy(mmp_d,mass[0],psi_d[0],mmp_d);
 
-	RealD c_d = axpy_norm(r_d, -1.0, mmp_d, src_d);
+	c = axpy_norm(r_d, -1.0, mmp_d, src_d);
 	AXPYTimer.Stop();
 
-	std::cout<<GridLogMessage<<"ConjugateGradientMultiShiftMixedPrec k="<<k<< ", replaced |r|^2 = "<<c_f <<" with |r|^2 = "<<c_d<<std::endl;
-	
-	PrecChangeTimer.Start();
-	precisionChange(r_f, r_d);
-	PrecChangeTimer.Stop();
-	c = c_d;
+	std::cout<<GridLogMessage<<"ConjugateGradientMultiShiftMixedPrec k="<<k<< ", replaced |r|^2 = "<<c_old <<" with |r|^2 = "<<c<<std::endl;
       }
     
       // Convergence checks

From e4c117aabf3690e3d9a8f50c0d3a3c1d32e6108a Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Fri, 23 Sep 2022 16:19:27 -0400
Subject: [PATCH 020/146] Compile fix, multishift mixed prec support

---
 Grid/lattice/Lattice_reduction.h |  4 +---
 Grid/lattice/Lattice_transfer.h  | 19 +++++++++++++++++++
 2 files changed, 20 insertions(+), 3 deletions(-)

diff --git a/Grid/lattice/Lattice_reduction.h b/Grid/lattice/Lattice_reduction.h
index bcd09c04..5f73e94b 100644
--- a/Grid/lattice/Lattice_reduction.h
+++ b/Grid/lattice/Lattice_reduction.h
@@ -92,9 +92,7 @@ inline typename vobj::scalar_objectD sumD_cpu(const vobj *arg, Integer osites)
     ssum = ssum+sumarray[i];
   } 
   
-  typedef typename vobj::scalar_object ssobj;
-  ssobj ret = ssum;
-  return ret;
+  return ssum;
 }
 /*
 Threaded max, don't use for now
diff --git a/Grid/lattice/Lattice_transfer.h b/Grid/lattice/Lattice_transfer.h
index aee55e93..de6ca886 100644
--- a/Grid/lattice/Lattice_transfer.h
+++ b/Grid/lattice/Lattice_transfer.h
@@ -1080,6 +1080,25 @@ vectorizeFromRevLexOrdArray( std::vector<sobj> &in, Lattice<vobj> &out)
   });
 }
 
+template<class VobjOut, class VobjIn>
+void precisionChangeFast(Lattice<VobjOut> &out, const Lattice<VobjIn> &in)
+{
+  typedef typename VobjOut::scalar_object SobjOut;
+  typedef typename VobjIn::scalar_object SobjIn;
+  conformable(out.Grid(),in.Grid());
+  out.Checkerboard() = in.Checkerboard();
+  int nsimd = out.Grid()->Nsimd();
+  autoView( out_v  , out, AcceleratorWrite);
+  autoView(  in_v ,   in, AcceleratorRead);
+  accelerator_for(idx,out.Grid()->oSites(),nsimd,{
+      auto itmp = coalescedRead(in_v[idx]);
+      auto otmp = coalescedRead(out_v[idx]);
+#ifdef GRID_SIMT
+      otmp=itmp;
+#endif
+      coalescedWrite(out_v[idx],otmp);
+  });
+}
 //Convert a Lattice from one precision to another
 template<class VobjOut, class VobjIn>
 void precisionChange(Lattice<VobjOut> &out, const Lattice<VobjIn> &in)

From 87ac7104f8017a5077fd4be310dbae273a94cbb1 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Fri, 23 Sep 2022 16:20:46 -0400
Subject: [PATCH 021/146] Prettier

---
 Grid/qcd/action/ActionBase.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/Grid/qcd/action/ActionBase.h b/Grid/qcd/action/ActionBase.h
index 353c4d3b..5027fcf4 100644
--- a/Grid/qcd/action/ActionBase.h
+++ b/Grid/qcd/action/ActionBase.h
@@ -50,9 +50,9 @@ public:
   RealD refresh_us;
   void  reset_timer(void)        {
     deriv_us = S_us = refresh_us = 0.0;
-    deriv_num=0;
     deriv_norm_sum = deriv_max_sum=0.0;
     Fdt_max_sum =  Fdt_norm_sum = 0.0;
+    deriv_num=0;
   }
   void  deriv_log(RealD nrm, RealD max,RealD Fdt_nrm,RealD Fdt_max) {
     deriv_max_sum+=max;

From 97cce103d78ce09211c871957c23c99897e40195 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Fri, 23 Sep 2022 16:21:49 -0400
Subject: [PATCH 022/146] Tolerances control

---
 .../GeneralEvenOddRationalRatio.h             | 53 +++++++++++++++++++
 1 file changed, 53 insertions(+)

diff --git a/Grid/qcd/action/pseudofermion/GeneralEvenOddRationalRatio.h b/Grid/qcd/action/pseudofermion/GeneralEvenOddRationalRatio.h
index 2b08cf49..027f37b3 100644
--- a/Grid/qcd/action/pseudofermion/GeneralEvenOddRationalRatio.h
+++ b/Grid/qcd/action/pseudofermion/GeneralEvenOddRationalRatio.h
@@ -115,6 +115,54 @@ NAMESPACE_BEGIN(Grid);
       
     public:
 
+      // allow non-uniform tolerances 
+      void SetTolerances(std::vector<RealD> action_tolerance,std::vector<RealD> md_tolerance)
+      {
+	assert(action_tolerance.size()==ApproxPowerAction.tolerances.size());
+	assert(    md_tolerance.size()==ApproxPowerMD.tolerances.size());
+	
+	// Fix up the tolerances
+	for(int i=0;i<ApproxPowerAction.tolerances.size();i++){
+	  ApproxPowerAction.tolerances[i]       = action_tolerance[i];
+	  ApproxNegPowerAction.tolerances[i]    = action_tolerance[i];
+	  ApproxHalfPowerAction.tolerances[i]   = action_tolerance[i];
+	  ApproxNegHalfPowerAction.tolerances[i]= action_tolerance[i];
+	  ApproxPowerMD.tolerances[i]       = md_tolerance[i];
+	  ApproxNegPowerMD.tolerances[i]    = md_tolerance[i];
+	  ApproxHalfPowerMD.tolerances[i]   = md_tolerance[i];
+	  ApproxNegHalfPowerMD.tolerances[i]= md_tolerance[i];
+	}
+	
+	// Print out - could deprecate
+	for(int i=0;i<ApproxPowerMD.tolerances.size();i++) {
+	  std::cout<<GridLogMessage << " ApproxPowerMD shift["<<i<<"] "
+		   <<" pole    "<<ApproxPowerMD.poles[i]
+		   <<" residue "<<ApproxPowerMD.residues[i]
+		   <<" tol     "<<ApproxPowerMD.tolerances[i]<<std::endl;
+	}
+	/*
+	  for(int i=0;i<ApproxNegPowerMD.tolerances.size();i++) {
+	  std::cout<<GridLogMessage << " ApproxNegPowerMD shift["<<i<<"] "
+		   <<" pole    "<<ApproxNegPowerMD.poles[i]
+		   <<" residue "<<ApproxNegPowerMD.residues[i]
+		   <<" tol     "<<ApproxNegPowerMD.tolerances[i]<<std::endl;
+	}
+	for(int i=0;i<ApproxHalfPowerMD.tolerances.size();i++) {
+	  std::cout<<GridLogMessage << " ApproxHalfPowerMD shift["<<i<<"] "
+		   <<" pole    "<<ApproxHalfPowerMD.poles[i]
+		   <<" residue "<<ApproxHalfPowerMD.residues[i]
+		   <<" tol     "<<ApproxHalfPowerMD.tolerances[i]<<std::endl;
+	}
+	for(int i=0;i<ApproxNegHalfPowerMD.tolerances.size();i++) {
+	  std::cout<<GridLogMessage << " ApproxNegHalfPowerMD shift["<<i<<"] "
+		   <<" pole    "<<ApproxNegHalfPowerMD.poles[i]
+		   <<" residue "<<ApproxNegHalfPowerMD.residues[i]
+		   <<" tol     "<<ApproxNegHalfPowerMD.tolerances[i]<<std::endl;
+	}
+	*/
+	
+      }
+      
       GeneralEvenOddRatioRationalPseudoFermionAction(FermionOperator<Impl>  &_NumOp, 
 						     FermionOperator<Impl>  &_DenOp, 
 						     const Params & p
@@ -149,6 +197,11 @@ NAMESPACE_BEGIN(Grid);
 	    ApproxNegHalfPowerMD.tolerances[i] = ApproxHalfPowerMD.tolerances[i] = param.md_tolerance;
 	}
 
+	std::vector<RealD> action_tolerance(ApproxHalfPowerAction.tolerances.size(),param.action_tolerance);
+	std::vector<RealD> md_tolerance    (ApproxHalfPowerMD.tolerances.size(),param.md_tolerance);
+
+	SetTolerances(action_tolerance, md_tolerance);
+	
 	std::cout<<GridLogMessage << action_name() << " initialize: complete" << std::endl;
       };
 

From eb552c3ecdbbe77e28d798e76ebcf28dbe44e80b Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Fri, 23 Sep 2022 16:22:28 -0400
Subject: [PATCH 023/146] dt info

---
 Grid/qcd/hmc/integrators/Integrator.h | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/Grid/qcd/hmc/integrators/Integrator.h b/Grid/qcd/hmc/integrators/Integrator.h
index 7f676ce7..48219c04 100644
--- a/Grid/qcd/hmc/integrators/Integrator.h
+++ b/Grid/qcd/hmc/integrators/Integrator.h
@@ -145,7 +145,7 @@ protected:
 
       //      DumpSliceNorm("force ",force,Nd-1);
       MomFilter->applyFilter(force);
-      std::cout << GridLogIntegrator << " update_P : Level [" << level <<"]["<<a <<"] "<<name<< std::endl;
+      std::cout << GridLogIntegrator << " update_P : Level [" << level <<"]["<<a <<"] "<<name<<" dt "<<ep<<  std::endl;
       DumpSliceNorm("force filtered ",force,Nd-1);
       
       Real force_abs   = std::sqrt(norm2(force)/U.Grid()->gSites()); //average per-site norm.  nb. norm2(latt) = \sum_x norm2(latt[x]) 
@@ -156,6 +156,7 @@ protected:
 
       as[level].actions.at(a)->deriv_log(force_abs,force_max,impulse_abs,impulse_max);
       
+      std::cout << GridLogIntegrator<< "["<<level<<"]["<<a<<"] dt           : " << ep <<" "<<name<<std::endl;
       std::cout << GridLogIntegrator<< "["<<level<<"]["<<a<<"] Force average: " << force_abs <<" "<<name<<std::endl;
       std::cout << GridLogIntegrator<< "["<<level<<"]["<<a<<"] Force max    : " << force_max <<" "<<name<<std::endl;
       std::cout << GridLogIntegrator<< "["<<level<<"]["<<a<<"] Fdt average  : " << impulse_abs <<" "<<name<<std::endl;

From a0d682687e373a5a15eaf9c853ee91ad1878c3a7 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Fri, 23 Sep 2022 16:22:53 -0400
Subject: [PATCH 024/146] Better logging of Fdt for force gradient

---
 Grid/qcd/hmc/integrators/Integrator_algorithm.h | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/Grid/qcd/hmc/integrators/Integrator_algorithm.h b/Grid/qcd/hmc/integrators/Integrator_algorithm.h
index b05c4ea8..ad556659 100644
--- a/Grid/qcd/hmc/integrators/Integrator_algorithm.h
+++ b/Grid/qcd/hmc/integrators/Integrator_algorithm.h
@@ -227,7 +227,8 @@ public:
     // Presently 4 force evals, and should have 3, so 1.33x too expensive.
     // could reduce this with sloppy CG to perhaps 1.15x too expensive
     // even without prediction.
-    this->update_P(Pfg, Ufg, level, 1.0);
+    this->update_P(Pfg, Ufg, level, fg_dt);
+    Pfg = Pfg*(1.0/fg_dt);
     this->update_U(Pfg, Ufg, fg_dt);
     this->update_P(Ufg, level, ep);
   }

From a2cefaa53afcd8a5f9e05f8e7639f17bb8bfc20c Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Fri, 23 Sep 2022 16:49:14 -0400
Subject: [PATCH 025/146] Faster

---
 systems/Crusher/config-command |  6 ++++--
 systems/Crusher/dwf.slurm      | 23 +++++++++++++----------
 systems/Crusher/sourceme.sh    |  8 ++++++--
 3 files changed, 23 insertions(+), 14 deletions(-)

diff --git a/systems/Crusher/config-command b/systems/Crusher/config-command
index dfd8d127..854c2c01 100644
--- a/systems/Crusher/config-command
+++ b/systems/Crusher/config-command
@@ -6,10 +6,12 @@
 --enable-simd=GPU \
 --disable-fermion-reps \
 --with-gmp=$OLCF_GMP_ROOT \
+--with-fftw=$FFTW_DIR/.. \
 --with-mpfr=/opt/cray/pe/gcc/mpfr/3.1.4/ \
 --enable-gparity \
 CXX=hipcc MPICXX=mpicxx \
-CXXFLAGS="-fPIC -I{$ROCM_PATH}/include/ -std=c++14 -I${MPICH_DIR}/include " \
- LDFLAGS=" -L${MPICH_DIR}/lib -lmpi -L${CRAY_MPICH_ROOTDIR}/gtl/lib -lmpi_gtl_hsa -lamdhip64 "
+CXXFLAGS="-fPIC -I{$ROCM_PATH}/include/ -std=c++14 -I${MPICH_DIR}/include -L/lib64 " \
+ LDFLAGS="-L/lib64 -L/opt/rocm-5.2.0/lib/ -L${MPICH_DIR}/lib -lmpi -L${CRAY_MPICH_ROOTDIR}/gtl/lib -lmpi_gtl_hsa -lamdhip64 "
+
 
 
diff --git a/systems/Crusher/dwf.slurm b/systems/Crusher/dwf.slurm
index 757ad4a2..20239e80 100644
--- a/systems/Crusher/dwf.slurm
+++ b/systems/Crusher/dwf.slurm
@@ -10,23 +10,26 @@
 #SBATCH -n 8
 #SBATCH --exclusive
 #SBATCH --gpu-bind=map_gpu:0,1,2,3,7,6,5,4
+#export BIND="--cpu-bind=verbose,map_ldom:3,3,1,1,2,2,0,0"
 
 DIR=.
 source sourceme.sh
-export MPIR_CVAR_GPU_EAGER_DEVICE_MEM=0
+export MPIR_CVAR_GPU_EAGER_DEVICE_MEM=16384
 export MPICH_GPU_SUPPORT_ENABLED=1
-export MPICH_SMP_SINGLE_COPY_MODE=XPMEM
-export OMP_NUM_THREADS=1
+export MPICH_SMP_SINGLE_COPY_MODE=CMA
+export OMP_NUM_THREADS=4
+export MPICH_OFI_NIC_POLICY=GPU
 
-echo MPICH_SMP_SINGLE_COPY_MODE $MPICH_SMP_SINGLE_COPY_MODE
+#PARAMS=" --accelerator-threads 8 --grid 64.64.32.16 --mpi 2.2.2.1 --comms-sequential --shm 2048 --shm-mpi 1"
+#srun --gpus-per-task 1 -n8 $BIND ./wrap.sh  ./benchmarks/Benchmark_dwf_fp32 $PARAMS
 
-echo working directory
-pwd
+PARAMS=" --accelerator-threads 8 --grid 16.16.32.32 --mpi 1.1.1.1 --comms-sequential --shm 2048 --shm-mpi 1"
+srun --gpus-per-task 1 -n1 $BIND ./wrap.sh  ./benchmarks/Benchmark_dwf_fp32 $PARAMS
 
-PARAMS=" --accelerator-threads 8 --grid 32.32.32.32 --mpi 1.1.1.1 --comms-sequential --shm 2048 --shm-mpi 0"
-srun --gpus-per-task 1 -n1  ./benchmarks/Benchmark_dwf_fp32 $PARAMS
+PARAMS=" --accelerator-threads 8 --grid 32.16.32.32 --mpi 1.1.1.2 --comms-sequential --shm 2048 --shm-mpi 1"
+srun --gpus-per-task 1 -n2 $BIND ./wrap.sh  ./benchmarks/Benchmark_dwf_fp32 $PARAMS
 
-PARAMS=" --accelerator-threads 8 --grid 64.64.64.32 --mpi 2.2.2.1 --comms-sequential --shm 2048 --shm-mpi 0"
-srun --gpus-per-task 1 -n8  ./benchmarks/Benchmark_dwf_fp32 $PARAMS
+PARAMS=" --accelerator-threads 8 --grid 32.32.32.64 --mpi 1.2.2.2 --comms-sequential --shm 2048 --shm-mpi 1"
+srun --gpus-per-task 1 -n8 $BIND ./wrap.sh  ./benchmarks/Benchmark_dwf_fp32 $PARAMS
 
 
diff --git a/systems/Crusher/sourceme.sh b/systems/Crusher/sourceme.sh
index ad0d6582..3cccb10a 100644
--- a/systems/Crusher/sourceme.sh
+++ b/systems/Crusher/sourceme.sh
@@ -1,8 +1,12 @@
+export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/sw/crusher/spack-envs/base/opt/cray-sles15-zen3/gcc-11.2.0/gperftools-2.9.1-72ubwtuc5wcz2meqltbfdb76epufgzo2/lib
+module load emacs 
+#module load gperftools
 module load PrgEnv-gnu
-module load rocm/5.1.0
+module load rocm/5.2.0
 module load cray-mpich/8.1.16
+#module load cray-mpich/8.1.17
 module load gmp
-#module load cray-fftw
+module load cray-fftw
 module load craype-accel-amd-gfx90a
 export LD_LIBRARY_PATH=/opt/gcc/mpfr/3.1.4/lib:$LD_LIBRARY_PATH
 #Hack for lib

From 68e4d833dd4f772f8b33ca12f45126138afea29a Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Fri, 23 Sep 2022 16:49:29 -0400
Subject: [PATCH 026/146] Run through wrapper script

---
 systems/Crusher/wrap.sh | 9 +++++++++
 1 file changed, 9 insertions(+)
 create mode 100755 systems/Crusher/wrap.sh

diff --git a/systems/Crusher/wrap.sh b/systems/Crusher/wrap.sh
new file mode 100755
index 00000000..eb58353c
--- /dev/null
+++ b/systems/Crusher/wrap.sh
@@ -0,0 +1,9 @@
+#!/bin/sh
+
+export HIP_VISIBLE_DEVICES=$ROCR_VISIBLE_DEVICES
+unset ROCR_VISIBLE_DEVICES
+
+#rank=$SLURM_PROCID
+#rocprof -d rocprof.$rank -o rocprof.$rank/results.rank$SLURM_PROCID.csv --sys-trace $@
+
+$@

From 7ffbc3e98e561a2f9696a1839649ac42e9b49450 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 27 Sep 2022 10:35:31 -0400
Subject: [PATCH 027/146] Double2 improved. REally don't like 'convertType' -
 localise to a GPT header

---
 Grid/lattice/Lattice_transfer.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/Grid/lattice/Lattice_transfer.h b/Grid/lattice/Lattice_transfer.h
index de6ca886..cf6e94c2 100644
--- a/Grid/lattice/Lattice_transfer.h
+++ b/Grid/lattice/Lattice_transfer.h
@@ -194,11 +194,11 @@ accelerator_inline void convertType(vComplexD2 & out, const ComplexD & in) {
 #endif
 
 accelerator_inline void convertType(vComplexF & out, const vComplexD2 & in) {
-  out.v = Optimization::PrecisionChange::DtoS(in._internal[0].v,in._internal[1].v);
+  precisionChange(out,in);
 }
 
 accelerator_inline void convertType(vComplexD2 & out, const vComplexF & in) {
-  Optimization::PrecisionChange::StoD(in.v,out._internal[0].v,out._internal[1].v);
+  precisionChange(out,in);
 }
 
 template<typename T1,typename T2>

From 6c9eef97264fb089489e5f26c79822b87b9cabea Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 27 Sep 2022 10:36:54 -0400
Subject: [PATCH 028/146] D2 fields

---
 Grid/qcd/QCD.h | 85 ++++++++++++++++++++++++++++++++++----------------
 1 file changed, 58 insertions(+), 27 deletions(-)

diff --git a/Grid/qcd/QCD.h b/Grid/qcd/QCD.h
index 81356a66..533c0e21 100644
--- a/Grid/qcd/QCD.h
+++ b/Grid/qcd/QCD.h
@@ -126,6 +126,7 @@ typedef iSpinMatrix<ComplexD >          SpinMatrixD;
 typedef iSpinMatrix<vComplex >          vSpinMatrix;
 typedef iSpinMatrix<vComplexF>          vSpinMatrixF;
 typedef iSpinMatrix<vComplexD>          vSpinMatrixD;
+typedef iSpinMatrix<vComplexD2>         vSpinMatrixD2;
 
 // Colour Matrix
 typedef iColourMatrix<Complex  >        ColourMatrix;
@@ -135,6 +136,7 @@ typedef iColourMatrix<ComplexD >        ColourMatrixD;
 typedef iColourMatrix<vComplex >        vColourMatrix;
 typedef iColourMatrix<vComplexF>        vColourMatrixF;
 typedef iColourMatrix<vComplexD>        vColourMatrixD;
+typedef iColourMatrix<vComplexD2>       vColourMatrixD2;
 
 // SpinColour matrix
 typedef iSpinColourMatrix<Complex  >    SpinColourMatrix;
@@ -144,6 +146,7 @@ typedef iSpinColourMatrix<ComplexD >    SpinColourMatrixD;
 typedef iSpinColourMatrix<vComplex >    vSpinColourMatrix;
 typedef iSpinColourMatrix<vComplexF>    vSpinColourMatrixF;
 typedef iSpinColourMatrix<vComplexD>    vSpinColourMatrixD;
+typedef iSpinColourMatrix<vComplexD2>   vSpinColourMatrixD2;
 
 // SpinColourSpinColour matrix
 typedef iSpinColourSpinColourMatrix<Complex  >    SpinColourSpinColourMatrix;
@@ -153,6 +156,7 @@ typedef iSpinColourSpinColourMatrix<ComplexD >    SpinColourSpinColourMatrixD;
 typedef iSpinColourSpinColourMatrix<vComplex >    vSpinColourSpinColourMatrix;
 typedef iSpinColourSpinColourMatrix<vComplexF>    vSpinColourSpinColourMatrixF;
 typedef iSpinColourSpinColourMatrix<vComplexD>    vSpinColourSpinColourMatrixD;
+typedef iSpinColourSpinColourMatrix<vComplexD2>   vSpinColourSpinColourMatrixD2;
 
 // SpinColourSpinColour matrix
 typedef iSpinColourSpinColourMatrix<Complex  >    SpinColourSpinColourMatrix;
@@ -162,33 +166,37 @@ typedef iSpinColourSpinColourMatrix<ComplexD >    SpinColourSpinColourMatrixD;
 typedef iSpinColourSpinColourMatrix<vComplex >    vSpinColourSpinColourMatrix;
 typedef iSpinColourSpinColourMatrix<vComplexF>    vSpinColourSpinColourMatrixF;
 typedef iSpinColourSpinColourMatrix<vComplexD>    vSpinColourSpinColourMatrixD;
+typedef iSpinColourSpinColourMatrix<vComplexD2>   vSpinColourSpinColourMatrixD2;
 
 // LorentzColour
 typedef iLorentzColourMatrix<Complex  > LorentzColourMatrix;
 typedef iLorentzColourMatrix<ComplexF > LorentzColourMatrixF;
 typedef iLorentzColourMatrix<ComplexD > LorentzColourMatrixD;
 
-typedef iLorentzColourMatrix<vComplex > vLorentzColourMatrix;
-typedef iLorentzColourMatrix<vComplexF> vLorentzColourMatrixF;
-typedef iLorentzColourMatrix<vComplexD> vLorentzColourMatrixD;
+typedef iLorentzColourMatrix<vComplex >  vLorentzColourMatrix;
+typedef iLorentzColourMatrix<vComplexF>  vLorentzColourMatrixF;
+typedef iLorentzColourMatrix<vComplexD>  vLorentzColourMatrixD;
+typedef iLorentzColourMatrix<vComplexD2> vLorentzColourMatrixD2;
 
 // DoubleStored gauge field
 typedef iDoubleStoredColourMatrix<Complex  > DoubleStoredColourMatrix;
 typedef iDoubleStoredColourMatrix<ComplexF > DoubleStoredColourMatrixF;
 typedef iDoubleStoredColourMatrix<ComplexD > DoubleStoredColourMatrixD;
 
-typedef iDoubleStoredColourMatrix<vComplex > vDoubleStoredColourMatrix;
-typedef iDoubleStoredColourMatrix<vComplexF> vDoubleStoredColourMatrixF;
-typedef iDoubleStoredColourMatrix<vComplexD> vDoubleStoredColourMatrixD;
+typedef iDoubleStoredColourMatrix<vComplex >  vDoubleStoredColourMatrix;
+typedef iDoubleStoredColourMatrix<vComplexF>  vDoubleStoredColourMatrixF;
+typedef iDoubleStoredColourMatrix<vComplexD>  vDoubleStoredColourMatrixD;
+typedef iDoubleStoredColourMatrix<vComplexD2> vDoubleStoredColourMatrixD2;
 
 //G-parity flavour matrix
 typedef iGparityFlavourMatrix<Complex> GparityFlavourMatrix;
 typedef iGparityFlavourMatrix<ComplexF> GparityFlavourMatrixF;
 typedef iGparityFlavourMatrix<ComplexD> GparityFlavourMatrixD;
 
-typedef iGparityFlavourMatrix<vComplex> vGparityFlavourMatrix;
-typedef iGparityFlavourMatrix<vComplexF> vGparityFlavourMatrixF;
-typedef iGparityFlavourMatrix<vComplexD> vGparityFlavourMatrixD;
+typedef iGparityFlavourMatrix<vComplex>   vGparityFlavourMatrix;
+typedef iGparityFlavourMatrix<vComplexF>  vGparityFlavourMatrixF;
+typedef iGparityFlavourMatrix<vComplexD>  vGparityFlavourMatrixD;
+typedef iGparityFlavourMatrix<vComplexD2> vGparityFlavourMatrixD2;
 
 
 // Spin vector
@@ -199,6 +207,7 @@ typedef iSpinVector<ComplexD>           SpinVectorD;
 typedef iSpinVector<vComplex >           vSpinVector;
 typedef iSpinVector<vComplexF>           vSpinVectorF;
 typedef iSpinVector<vComplexD>           vSpinVectorD;
+typedef iSpinVector<vComplexD2>          vSpinVectorD2;
 
 // Colour vector
 typedef iColourVector<Complex >         ColourVector;
@@ -208,6 +217,7 @@ typedef iColourVector<ComplexD>         ColourVectorD;
 typedef iColourVector<vComplex >         vColourVector;
 typedef iColourVector<vComplexF>         vColourVectorF;
 typedef iColourVector<vComplexD>         vColourVectorD;
+typedef iColourVector<vComplexD2>        vColourVectorD2;
 
 // SpinColourVector
 typedef iSpinColourVector<Complex >     SpinColourVector;
@@ -217,6 +227,7 @@ typedef iSpinColourVector<ComplexD>     SpinColourVectorD;
 typedef iSpinColourVector<vComplex >     vSpinColourVector;
 typedef iSpinColourVector<vComplexF>     vSpinColourVectorF;
 typedef iSpinColourVector<vComplexD>     vSpinColourVectorD;
+typedef iSpinColourVector<vComplexD2>    vSpinColourVectorD2;
 
 // HalfSpin vector
 typedef iHalfSpinVector<Complex >       HalfSpinVector;
@@ -226,15 +237,17 @@ typedef iHalfSpinVector<ComplexD>       HalfSpinVectorD;
 typedef iHalfSpinVector<vComplex >       vHalfSpinVector;
 typedef iHalfSpinVector<vComplexF>       vHalfSpinVectorF;
 typedef iHalfSpinVector<vComplexD>       vHalfSpinVectorD;
+typedef iHalfSpinVector<vComplexD2>      vHalfSpinVectorD2;
 
 // HalfSpinColour vector
 typedef iHalfSpinColourVector<Complex > HalfSpinColourVector;
 typedef iHalfSpinColourVector<ComplexF> HalfSpinColourVectorF;
 typedef iHalfSpinColourVector<ComplexD> HalfSpinColourVectorD;
     
-typedef iHalfSpinColourVector<vComplex > vHalfSpinColourVector;
-typedef iHalfSpinColourVector<vComplexF> vHalfSpinColourVectorF;
-typedef iHalfSpinColourVector<vComplexD> vHalfSpinColourVectorD;
+typedef iHalfSpinColourVector<vComplex >  vHalfSpinColourVector;
+typedef iHalfSpinColourVector<vComplexF>  vHalfSpinColourVectorF;
+typedef iHalfSpinColourVector<vComplexD>  vHalfSpinColourVectorD;
+typedef iHalfSpinColourVector<vComplexD2> vHalfSpinColourVectorD2;
 
 //G-parity flavour vector
 typedef iGparityFlavourVector<Complex >         GparityFlavourVector;
@@ -244,7 +257,7 @@ typedef iGparityFlavourVector<ComplexD>         GparityFlavourVectorD;
 typedef iGparityFlavourVector<vComplex >         vGparityFlavourVector;
 typedef iGparityFlavourVector<vComplexF>         vGparityFlavourVectorF;
 typedef iGparityFlavourVector<vComplexD>         vGparityFlavourVectorD;
-
+typedef iGparityFlavourVector<vComplexD2>        vGparityFlavourVectorD2;
     
 // singlets
 typedef iSinglet<Complex >         TComplex;     // FIXME This is painful. Tensor singlet complex type.
@@ -254,6 +267,7 @@ typedef iSinglet<ComplexD>         TComplexD;    // FIXME This is painful. Tenso
 typedef iSinglet<vComplex >        vTComplex ;   // what if we don't know the tensor structure
 typedef iSinglet<vComplexF>        vTComplexF;   // what if we don't know the tensor structure
 typedef iSinglet<vComplexD>        vTComplexD;   // what if we don't know the tensor structure
+typedef iSinglet<vComplexD2>       vTComplexD2;   // what if we don't know the tensor structure
 
 typedef iSinglet<Real >            TReal;        // Shouldn't need these; can I make it work without?
 typedef iSinglet<RealF>            TRealF;       // Shouldn't need these; can I make it work without?
@@ -271,47 +285,58 @@ typedef iSinglet<Integer >         TInteger;
 typedef Lattice<vColourMatrix>          LatticeColourMatrix;
 typedef Lattice<vColourMatrixF>         LatticeColourMatrixF;
 typedef Lattice<vColourMatrixD>         LatticeColourMatrixD;
+typedef Lattice<vColourMatrixD2>        LatticeColourMatrixD2;
 
 typedef Lattice<vSpinMatrix>            LatticeSpinMatrix;
 typedef Lattice<vSpinMatrixF>           LatticeSpinMatrixF;
 typedef Lattice<vSpinMatrixD>           LatticeSpinMatrixD;
+typedef Lattice<vSpinMatrixD2>          LatticeSpinMatrixD2;
 
 typedef Lattice<vSpinColourMatrix>      LatticeSpinColourMatrix;
 typedef Lattice<vSpinColourMatrixF>     LatticeSpinColourMatrixF;
 typedef Lattice<vSpinColourMatrixD>     LatticeSpinColourMatrixD;
+typedef Lattice<vSpinColourMatrixD2>    LatticeSpinColourMatrixD2;
 
 typedef Lattice<vSpinColourSpinColourMatrix>      LatticeSpinColourSpinColourMatrix;
 typedef Lattice<vSpinColourSpinColourMatrixF>     LatticeSpinColourSpinColourMatrixF;
 typedef Lattice<vSpinColourSpinColourMatrixD>     LatticeSpinColourSpinColourMatrixD;
+typedef Lattice<vSpinColourSpinColourMatrixD2>    LatticeSpinColourSpinColourMatrixD2;
 
-typedef Lattice<vLorentzColourMatrix>  LatticeLorentzColourMatrix;
-typedef Lattice<vLorentzColourMatrixF> LatticeLorentzColourMatrixF;
-typedef Lattice<vLorentzColourMatrixD> LatticeLorentzColourMatrixD;
+typedef Lattice<vLorentzColourMatrix>   LatticeLorentzColourMatrix;
+typedef Lattice<vLorentzColourMatrixF>  LatticeLorentzColourMatrixF;
+typedef Lattice<vLorentzColourMatrixD>  LatticeLorentzColourMatrixD;
+typedef Lattice<vLorentzColourMatrixD2> LatticeLorentzColourMatrixD2;
 
 // DoubleStored gauge field
-typedef Lattice<vDoubleStoredColourMatrix>  LatticeDoubleStoredColourMatrix;
-typedef Lattice<vDoubleStoredColourMatrixF> LatticeDoubleStoredColourMatrixF;
-typedef Lattice<vDoubleStoredColourMatrixD> LatticeDoubleStoredColourMatrixD;
+typedef Lattice<vDoubleStoredColourMatrix>   LatticeDoubleStoredColourMatrix;
+typedef Lattice<vDoubleStoredColourMatrixF>  LatticeDoubleStoredColourMatrixF;
+typedef Lattice<vDoubleStoredColourMatrixD>  LatticeDoubleStoredColourMatrixD;
+typedef Lattice<vDoubleStoredColourMatrixD2> LatticeDoubleStoredColourMatrixD2;
 
 typedef Lattice<vSpinVector>            LatticeSpinVector;
 typedef Lattice<vSpinVectorF>           LatticeSpinVectorF;
 typedef Lattice<vSpinVectorD>           LatticeSpinVectorD;
+typedef Lattice<vSpinVectorD2>          LatticeSpinVectorD2;
 
 typedef Lattice<vColourVector>          LatticeColourVector;
 typedef Lattice<vColourVectorF>         LatticeColourVectorF;
 typedef Lattice<vColourVectorD>         LatticeColourVectorD;
+typedef Lattice<vColourVectorD2>        LatticeColourVectorD2;
 
 typedef Lattice<vSpinColourVector>      LatticeSpinColourVector;
 typedef Lattice<vSpinColourVectorF>     LatticeSpinColourVectorF;
 typedef Lattice<vSpinColourVectorD>     LatticeSpinColourVectorD;
+typedef Lattice<vSpinColourVectorD2>    LatticeSpinColourVectorD2;
 
 typedef Lattice<vHalfSpinVector>        LatticeHalfSpinVector;
 typedef Lattice<vHalfSpinVectorF>       LatticeHalfSpinVectorF;
 typedef Lattice<vHalfSpinVectorD>       LatticeHalfSpinVectorD;
+typedef Lattice<vHalfSpinVectorD2>      LatticeHalfSpinVectorD2;
 
-typedef Lattice<vHalfSpinColourVector>  LatticeHalfSpinColourVector;
-typedef Lattice<vHalfSpinColourVectorF> LatticeHalfSpinColourVectorF;
-typedef Lattice<vHalfSpinColourVectorD> LatticeHalfSpinColourVectorD;
+typedef Lattice<vHalfSpinColourVector>   LatticeHalfSpinColourVector;
+typedef Lattice<vHalfSpinColourVectorF>  LatticeHalfSpinColourVectorF;
+typedef Lattice<vHalfSpinColourVectorD>  LatticeHalfSpinColourVectorD;
+typedef Lattice<vHalfSpinColourVectorD2> LatticeHalfSpinColourVectorD2;
 
 typedef Lattice<vTReal>            LatticeReal;
 typedef Lattice<vTRealF>           LatticeRealF;
@@ -320,6 +345,7 @@ typedef Lattice<vTRealD>           LatticeRealD;
 typedef Lattice<vTComplex>         LatticeComplex;
 typedef Lattice<vTComplexF>        LatticeComplexF;
 typedef Lattice<vTComplexD>        LatticeComplexD;
+typedef Lattice<vTComplexD2>       LatticeComplexD2;
 
 typedef Lattice<vTInteger>         LatticeInteger; // Predicates for "where"
 
@@ -327,37 +353,42 @@ typedef Lattice<vTInteger>         LatticeInteger; // Predicates for "where"
 ///////////////////////////////////////////
 // Physical names for things
 ///////////////////////////////////////////
-typedef LatticeHalfSpinColourVector  LatticeHalfFermion;
-typedef LatticeHalfSpinColourVectorF LatticeHalfFermionF;
-typedef LatticeHalfSpinColourVectorF LatticeHalfFermionD;
+typedef LatticeHalfSpinColourVector   LatticeHalfFermion;
+typedef LatticeHalfSpinColourVectorF  LatticeHalfFermionF;
+typedef LatticeHalfSpinColourVectorD  LatticeHalfFermionD;
+typedef LatticeHalfSpinColourVectorD2 LatticeHalfFermionD2;
 
 typedef LatticeSpinColourVector      LatticeFermion;
 typedef LatticeSpinColourVectorF     LatticeFermionF;
 typedef LatticeSpinColourVectorD     LatticeFermionD;
+typedef LatticeSpinColourVectorD2    LatticeFermionD2;
 
 typedef LatticeSpinColourMatrix                LatticePropagator;
 typedef LatticeSpinColourMatrixF               LatticePropagatorF;
 typedef LatticeSpinColourMatrixD               LatticePropagatorD;
+typedef LatticeSpinColourMatrixD2              LatticePropagatorD2;
 
 typedef LatticeLorentzColourMatrix             LatticeGaugeField;
 typedef LatticeLorentzColourMatrixF            LatticeGaugeFieldF;
 typedef LatticeLorentzColourMatrixD            LatticeGaugeFieldD;
+typedef LatticeLorentzColourMatrixD2           LatticeGaugeFieldD2;
 
 typedef LatticeDoubleStoredColourMatrix        LatticeDoubledGaugeField;
 typedef LatticeDoubleStoredColourMatrixF       LatticeDoubledGaugeFieldF;
 typedef LatticeDoubleStoredColourMatrixD       LatticeDoubledGaugeFieldD;
+typedef LatticeDoubleStoredColourMatrixD2      LatticeDoubledGaugeFieldD2;
 
 template<class GF> using LorentzScalar = Lattice<iScalar<typename GF::vector_object::element> >;
 
-// Uhgg... typing this hurt  ;)
-// (my keyboard got burning hot when I typed this, must be the anti-Fermion)
 typedef Lattice<vColourVector>          LatticeStaggeredFermion;    
 typedef Lattice<vColourVectorF>         LatticeStaggeredFermionF;    
 typedef Lattice<vColourVectorD>         LatticeStaggeredFermionD;    
+typedef Lattice<vColourVectorD2>        LatticeStaggeredFermionD2;    
 
 typedef Lattice<vColourMatrix>          LatticeStaggeredPropagator; 
 typedef Lattice<vColourMatrixF>         LatticeStaggeredPropagatorF; 
 typedef Lattice<vColourMatrixD>         LatticeStaggeredPropagatorD; 
+typedef Lattice<vColourMatrixD2>        LatticeStaggeredPropagatorD2; 
 
 //////////////////////////////////////////////////////////////////////////////
 // Peek and Poke named after physics attributes

From 9e81b42981b34484a75d1f2051b2ff3ffc01c14d Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 27 Sep 2022 10:37:19 -0400
Subject: [PATCH 029/146] D2 fields

---
 Grid/qcd/action/fermion/CloverHelpers.h | 1 +
 1 file changed, 1 insertion(+)

diff --git a/Grid/qcd/action/fermion/CloverHelpers.h b/Grid/qcd/action/fermion/CloverHelpers.h
index 57e71998..e93f3dd0 100644
--- a/Grid/qcd/action/fermion/CloverHelpers.h
+++ b/Grid/qcd/action/fermion/CloverHelpers.h
@@ -140,6 +140,7 @@ public:
     return NMAX;
   }
 
+  static int getNMAX(Lattice<iImplClover<vComplexD2>> &t, RealD R) {return getNMAX(1e-12,R);}
   static int getNMAX(Lattice<iImplClover<vComplexD>> &t, RealD R) {return getNMAX(1e-12,R);}
   static int getNMAX(Lattice<iImplClover<vComplexF>> &t, RealD R) {return getNMAX(1e-6,R);}
 

From e8bfbf2f7c14165cea962d73963281b657f99bd0 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 27 Sep 2022 10:37:45 -0400
Subject: [PATCH 030/146] D2 operators

---
 Grid/qcd/action/fermion/Fermion.h | 111 +++++++-----------------------
 1 file changed, 24 insertions(+), 87 deletions(-)

diff --git a/Grid/qcd/action/fermion/Fermion.h b/Grid/qcd/action/fermion/Fermion.h
index 223ff9dd..2f3fb9f2 100644
--- a/Grid/qcd/action/fermion/Fermion.h
+++ b/Grid/qcd/action/fermion/Fermion.h
@@ -112,28 +112,21 @@ NAMESPACE_CHECK(DWFutils);
 // Cayley 5d
 NAMESPACE_BEGIN(Grid);
 
-typedef WilsonFermion<WilsonImplR> WilsonFermionR;
+typedef WilsonFermion<WilsonImplD2> WilsonFermionD2;
 typedef WilsonFermion<WilsonImplF> WilsonFermionF;
 typedef WilsonFermion<WilsonImplD> WilsonFermionD;
 
-//typedef WilsonFermion<WilsonImplRL> WilsonFermionRL;
-//typedef WilsonFermion<WilsonImplFH> WilsonFermionFH;
-//typedef WilsonFermion<WilsonImplDF> WilsonFermionDF;
-
-typedef WilsonFermion<WilsonAdjImplR> WilsonAdjFermionR;
 typedef WilsonFermion<WilsonAdjImplF> WilsonAdjFermionF;
 typedef WilsonFermion<WilsonAdjImplD> WilsonAdjFermionD;
 
-typedef WilsonFermion<WilsonTwoIndexSymmetricImplR> WilsonTwoIndexSymmetricFermionR;
 typedef WilsonFermion<WilsonTwoIndexSymmetricImplF> WilsonTwoIndexSymmetricFermionF;
 typedef WilsonFermion<WilsonTwoIndexSymmetricImplD> WilsonTwoIndexSymmetricFermionD;
 
-typedef WilsonFermion<WilsonTwoIndexAntiSymmetricImplR> WilsonTwoIndexAntiSymmetricFermionR;
 typedef WilsonFermion<WilsonTwoIndexAntiSymmetricImplF> WilsonTwoIndexAntiSymmetricFermionF;
 typedef WilsonFermion<WilsonTwoIndexAntiSymmetricImplD> WilsonTwoIndexAntiSymmetricFermionD;
 
 // Twisted mass fermion
-typedef WilsonTMFermion<WilsonImplR> WilsonTMFermionR;
+typedef WilsonTMFermion<WilsonImplD2> WilsonTMFermionD2;
 typedef WilsonTMFermion<WilsonImplF> WilsonTMFermionF;
 typedef WilsonTMFermion<WilsonImplD> WilsonTMFermionD;
 
@@ -141,23 +134,20 @@ typedef WilsonTMFermion<WilsonImplD> WilsonTMFermionD;
 template <typename WImpl> using WilsonClover = WilsonCloverFermion<WImpl, CloverHelpers<WImpl>>;
 template <typename WImpl> using WilsonExpClover = WilsonCloverFermion<WImpl, ExpCloverHelpers<WImpl>>;
 
-typedef WilsonClover<WilsonImplR> WilsonCloverFermionR;
+typedef WilsonClover<WilsonImplD2> WilsonCloverFermionD2;
 typedef WilsonClover<WilsonImplF> WilsonCloverFermionF;
 typedef WilsonClover<WilsonImplD> WilsonCloverFermionD;
 
-typedef WilsonExpClover<WilsonImplR> WilsonExpCloverFermionR;
+typedef WilsonExpClover<WilsonImplD2> WilsonExpCloverFermionD2;
 typedef WilsonExpClover<WilsonImplF> WilsonExpCloverFermionF;
 typedef WilsonExpClover<WilsonImplD> WilsonExpCloverFermionD;
 
-typedef WilsonClover<WilsonAdjImplR> WilsonCloverAdjFermionR;
 typedef WilsonClover<WilsonAdjImplF> WilsonCloverAdjFermionF;
 typedef WilsonClover<WilsonAdjImplD> WilsonCloverAdjFermionD;
 
-typedef WilsonClover<WilsonTwoIndexSymmetricImplR> WilsonCloverTwoIndexSymmetricFermionR;
 typedef WilsonClover<WilsonTwoIndexSymmetricImplF> WilsonCloverTwoIndexSymmetricFermionF;
 typedef WilsonClover<WilsonTwoIndexSymmetricImplD> WilsonCloverTwoIndexSymmetricFermionD;
 
-typedef WilsonClover<WilsonTwoIndexAntiSymmetricImplR> WilsonCloverTwoIndexAntiSymmetricFermionR;
 typedef WilsonClover<WilsonTwoIndexAntiSymmetricImplF> WilsonCloverTwoIndexAntiSymmetricFermionF;
 typedef WilsonClover<WilsonTwoIndexAntiSymmetricImplD> WilsonCloverTwoIndexAntiSymmetricFermionD;
 
@@ -165,161 +155,108 @@ typedef WilsonClover<WilsonTwoIndexAntiSymmetricImplD> WilsonCloverTwoIndexAntiS
 template <typename WImpl> using CompactWilsonClover = CompactWilsonCloverFermion<WImpl, CompactCloverHelpers<WImpl>>;
 template <typename WImpl> using CompactWilsonExpClover = CompactWilsonCloverFermion<WImpl, CompactExpCloverHelpers<WImpl>>;
 
-typedef CompactWilsonClover<WilsonImplR> CompactWilsonCloverFermionR;
+typedef CompactWilsonClover<WilsonImplD2> CompactWilsonCloverFermionD2;
 typedef CompactWilsonClover<WilsonImplF> CompactWilsonCloverFermionF;
 typedef CompactWilsonClover<WilsonImplD> CompactWilsonCloverFermionD;
 
-typedef CompactWilsonExpClover<WilsonImplR> CompactWilsonExpCloverFermionR;
+typedef CompactWilsonExpClover<WilsonImplD2> CompactWilsonExpCloverFermionD2;
 typedef CompactWilsonExpClover<WilsonImplF> CompactWilsonExpCloverFermionF;
 typedef CompactWilsonExpClover<WilsonImplD> CompactWilsonExpCloverFermionD;
 
-typedef CompactWilsonClover<WilsonAdjImplR> CompactWilsonCloverAdjFermionR;
 typedef CompactWilsonClover<WilsonAdjImplF> CompactWilsonCloverAdjFermionF;
 typedef CompactWilsonClover<WilsonAdjImplD> CompactWilsonCloverAdjFermionD;
 
-typedef CompactWilsonClover<WilsonTwoIndexSymmetricImplR> CompactWilsonCloverTwoIndexSymmetricFermionR;
 typedef CompactWilsonClover<WilsonTwoIndexSymmetricImplF> CompactWilsonCloverTwoIndexSymmetricFermionF;
 typedef CompactWilsonClover<WilsonTwoIndexSymmetricImplD> CompactWilsonCloverTwoIndexSymmetricFermionD;
 
-typedef CompactWilsonClover<WilsonTwoIndexAntiSymmetricImplR> CompactWilsonCloverTwoIndexAntiSymmetricFermionR;
 typedef CompactWilsonClover<WilsonTwoIndexAntiSymmetricImplF> CompactWilsonCloverTwoIndexAntiSymmetricFermionF;
 typedef CompactWilsonClover<WilsonTwoIndexAntiSymmetricImplD> CompactWilsonCloverTwoIndexAntiSymmetricFermionD;
 
 // Domain Wall fermions
-typedef DomainWallFermion<WilsonImplR> DomainWallFermionR;
 typedef DomainWallFermion<WilsonImplF> DomainWallFermionF;
 typedef DomainWallFermion<WilsonImplD> DomainWallFermionD;
+typedef DomainWallFermion<WilsonImplD2> DomainWallFermionD2;
 
-//typedef DomainWallFermion<WilsonImplRL> DomainWallFermionRL;
-//typedef DomainWallFermion<WilsonImplFH> DomainWallFermionFH;
-//typedef DomainWallFermion<WilsonImplDF> DomainWallFermionDF;
-
-typedef DomainWallEOFAFermion<WilsonImplR> DomainWallEOFAFermionR;
+typedef DomainWallEOFAFermion<WilsonImplD2> DomainWallEOFAFermionD2;
 typedef DomainWallEOFAFermion<WilsonImplF> DomainWallEOFAFermionF;
 typedef DomainWallEOFAFermion<WilsonImplD> DomainWallEOFAFermionD;
 
-//typedef DomainWallEOFAFermion<WilsonImplRL> DomainWallEOFAFermionRL;
-//typedef DomainWallEOFAFermion<WilsonImplFH> DomainWallEOFAFermionFH;
-//typedef DomainWallEOFAFermion<WilsonImplDF> DomainWallEOFAFermionDF;
-
-typedef MobiusFermion<WilsonImplR> MobiusFermionR;
+typedef MobiusFermion<WilsonImplD2> MobiusFermionD2;
 typedef MobiusFermion<WilsonImplF> MobiusFermionF;
 typedef MobiusFermion<WilsonImplD> MobiusFermionD;
 
-//typedef MobiusFermion<WilsonImplRL> MobiusFermionRL;
-//typedef MobiusFermion<WilsonImplFH> MobiusFermionFH;
-//typedef MobiusFermion<WilsonImplDF> MobiusFermionDF;
-
-typedef MobiusEOFAFermion<WilsonImplR> MobiusEOFAFermionR;
+typedef MobiusEOFAFermion<WilsonImplD2> MobiusEOFAFermionD2;
 typedef MobiusEOFAFermion<WilsonImplF> MobiusEOFAFermionF;
 typedef MobiusEOFAFermion<WilsonImplD> MobiusEOFAFermionD;
 
-//typedef MobiusEOFAFermion<WilsonImplRL> MobiusEOFAFermionRL;
-//typedef MobiusEOFAFermion<WilsonImplFH> MobiusEOFAFermionFH;
-//typedef MobiusEOFAFermion<WilsonImplDF> MobiusEOFAFermionDF;
-
-typedef ZMobiusFermion<ZWilsonImplR> ZMobiusFermionR;
+typedef ZMobiusFermion<ZWilsonImplD2> ZMobiusFermionD2;
 typedef ZMobiusFermion<ZWilsonImplF> ZMobiusFermionF;
 typedef ZMobiusFermion<ZWilsonImplD> ZMobiusFermionD;
 
-//typedef ZMobiusFermion<ZWilsonImplRL> ZMobiusFermionRL;
-//typedef ZMobiusFermion<ZWilsonImplFH> ZMobiusFermionFH;
-//typedef ZMobiusFermion<ZWilsonImplDF> ZMobiusFermionDF;
-
-// Ls vectorised
-typedef ScaledShamirFermion<WilsonImplR> ScaledShamirFermionR;
+typedef ScaledShamirFermion<WilsonImplD2> ScaledShamirFermionD2;
 typedef ScaledShamirFermion<WilsonImplF> ScaledShamirFermionF;
 typedef ScaledShamirFermion<WilsonImplD> ScaledShamirFermionD;
 
-typedef MobiusZolotarevFermion<WilsonImplR> MobiusZolotarevFermionR;
+typedef MobiusZolotarevFermion<WilsonImplD2> MobiusZolotarevFermionD2;
 typedef MobiusZolotarevFermion<WilsonImplF> MobiusZolotarevFermionF;
 typedef MobiusZolotarevFermion<WilsonImplD> MobiusZolotarevFermionD;
-typedef ShamirZolotarevFermion<WilsonImplR> ShamirZolotarevFermionR;
+typedef ShamirZolotarevFermion<WilsonImplD2> ShamirZolotarevFermionD2;
 typedef ShamirZolotarevFermion<WilsonImplF> ShamirZolotarevFermionF;
 typedef ShamirZolotarevFermion<WilsonImplD> ShamirZolotarevFermionD;
 
-typedef OverlapWilsonCayleyTanhFermion<WilsonImplR> OverlapWilsonCayleyTanhFermionR;
+typedef OverlapWilsonCayleyTanhFermion<WilsonImplD2> OverlapWilsonCayleyTanhFermionD2;
 typedef OverlapWilsonCayleyTanhFermion<WilsonImplF> OverlapWilsonCayleyTanhFermionF;
 typedef OverlapWilsonCayleyTanhFermion<WilsonImplD> OverlapWilsonCayleyTanhFermionD;
-typedef OverlapWilsonCayleyZolotarevFermion<WilsonImplR> OverlapWilsonCayleyZolotarevFermionR;
+typedef OverlapWilsonCayleyZolotarevFermion<WilsonImplD2> OverlapWilsonCayleyZolotarevFermionD2;
 typedef OverlapWilsonCayleyZolotarevFermion<WilsonImplF> OverlapWilsonCayleyZolotarevFermionF;
 typedef OverlapWilsonCayleyZolotarevFermion<WilsonImplD> OverlapWilsonCayleyZolotarevFermionD;
 
 // Continued fraction
-typedef OverlapWilsonContFracTanhFermion<WilsonImplR> OverlapWilsonContFracTanhFermionR;
+typedef OverlapWilsonContFracTanhFermion<WilsonImplD2> OverlapWilsonContFracTanhFermionD2;
 typedef OverlapWilsonContFracTanhFermion<WilsonImplF> OverlapWilsonContFracTanhFermionF;
 typedef OverlapWilsonContFracTanhFermion<WilsonImplD> OverlapWilsonContFracTanhFermionD;
-typedef OverlapWilsonContFracZolotarevFermion<WilsonImplR> OverlapWilsonContFracZolotarevFermionR;
+typedef OverlapWilsonContFracZolotarevFermion<WilsonImplD2> OverlapWilsonContFracZolotarevFermionD2;
 typedef OverlapWilsonContFracZolotarevFermion<WilsonImplF> OverlapWilsonContFracZolotarevFermionF;
 typedef OverlapWilsonContFracZolotarevFermion<WilsonImplD> OverlapWilsonContFracZolotarevFermionD;
 
 // Partial fraction
-typedef OverlapWilsonPartialFractionTanhFermion<WilsonImplR> OverlapWilsonPartialFractionTanhFermionR;
+typedef OverlapWilsonPartialFractionTanhFermion<WilsonImplD2> OverlapWilsonPartialFractionTanhFermionD2;
 typedef OverlapWilsonPartialFractionTanhFermion<WilsonImplF> OverlapWilsonPartialFractionTanhFermionF;
 typedef OverlapWilsonPartialFractionTanhFermion<WilsonImplD> OverlapWilsonPartialFractionTanhFermionD;
 
-typedef OverlapWilsonPartialFractionZolotarevFermion<WilsonImplR> OverlapWilsonPartialFractionZolotarevFermionR;
+typedef OverlapWilsonPartialFractionZolotarevFermion<WilsonImplD2> OverlapWilsonPartialFractionZolotarevFermionD2;
 typedef OverlapWilsonPartialFractionZolotarevFermion<WilsonImplF> OverlapWilsonPartialFractionZolotarevFermionF;
 typedef OverlapWilsonPartialFractionZolotarevFermion<WilsonImplD> OverlapWilsonPartialFractionZolotarevFermionD;
 
 // Gparity cases; partial list until tested
-typedef WilsonFermion<GparityWilsonImplR>     GparityWilsonFermionR;
 typedef WilsonFermion<GparityWilsonImplF>     GparityWilsonFermionF;
 typedef WilsonFermion<GparityWilsonImplD>     GparityWilsonFermionD;
 
-//typedef WilsonFermion<GparityWilsonImplRL>     GparityWilsonFermionRL;
-//typedef WilsonFermion<GparityWilsonImplFH>     GparityWilsonFermionFH;
-//typedef WilsonFermion<GparityWilsonImplDF>     GparityWilsonFermionDF;
-
-typedef DomainWallFermion<GparityWilsonImplR> GparityDomainWallFermionR;
 typedef DomainWallFermion<GparityWilsonImplF> GparityDomainWallFermionF;
 typedef DomainWallFermion<GparityWilsonImplD> GparityDomainWallFermionD;
 
-//typedef DomainWallFermion<GparityWilsonImplRL> GparityDomainWallFermionRL;
-//typedef DomainWallFermion<GparityWilsonImplFH> GparityDomainWallFermionFH;
-//typedef DomainWallFermion<GparityWilsonImplDF> GparityDomainWallFermionDF;
-
-typedef DomainWallEOFAFermion<GparityWilsonImplR> GparityDomainWallEOFAFermionR;
+typedef DomainWallEOFAFermion<GparityWilsonImplR> GparityDomainWallEOFAFermionD2;
 typedef DomainWallEOFAFermion<GparityWilsonImplF> GparityDomainWallEOFAFermionF;
 typedef DomainWallEOFAFermion<GparityWilsonImplD> GparityDomainWallEOFAFermionD;
 
-//typedef DomainWallEOFAFermion<GparityWilsonImplRL> GparityDomainWallEOFAFermionRL;
-//typedef DomainWallEOFAFermion<GparityWilsonImplFH> GparityDomainWallEOFAFermionFH;
-//typedef DomainWallEOFAFermion<GparityWilsonImplDF> GparityDomainWallEOFAFermionDF;
-
-typedef WilsonTMFermion<GparityWilsonImplR> GparityWilsonTMFermionR;
+typedef WilsonTMFermion<GparityWilsonImplR> GparityWilsonTMFermionD2;
 typedef WilsonTMFermion<GparityWilsonImplF> GparityWilsonTMFermionF;
 typedef WilsonTMFermion<GparityWilsonImplD> GparityWilsonTMFermionD;
 
-//typedef WilsonTMFermion<GparityWilsonImplRL> GparityWilsonTMFermionRL;
-//typedef WilsonTMFermion<GparityWilsonImplFH> GparityWilsonTMFermionFH;
-//typedef WilsonTMFermion<GparityWilsonImplDF> GparityWilsonTMFermionDF;
-
-typedef MobiusFermion<GparityWilsonImplR> GparityMobiusFermionR;
+typedef MobiusFermion<GparityWilsonImplR> GparityMobiusFermionD2;
 typedef MobiusFermion<GparityWilsonImplF> GparityMobiusFermionF;
 typedef MobiusFermion<GparityWilsonImplD> GparityMobiusFermionD;
 
-//typedef MobiusFermion<GparityWilsonImplRL> GparityMobiusFermionRL;
-//typedef MobiusFermion<GparityWilsonImplFH> GparityMobiusFermionFH;
-//typedef MobiusFermion<GparityWilsonImplDF> GparityMobiusFermionDF;
-
-typedef MobiusEOFAFermion<GparityWilsonImplR> GparityMobiusEOFAFermionR;
+typedef MobiusEOFAFermion<GparityWilsonImplR> GparityMobiusEOFAFermionD2;
 typedef MobiusEOFAFermion<GparityWilsonImplF> GparityMobiusEOFAFermionF;
 typedef MobiusEOFAFermion<GparityWilsonImplD> GparityMobiusEOFAFermionD;
 
-//typedef MobiusEOFAFermion<GparityWilsonImplRL> GparityMobiusEOFAFermionRL;
-//typedef MobiusEOFAFermion<GparityWilsonImplFH> GparityMobiusEOFAFermionFH;
-//typedef MobiusEOFAFermion<GparityWilsonImplDF> GparityMobiusEOFAFermionDF;
-
-typedef ImprovedStaggeredFermion<StaggeredImplR> ImprovedStaggeredFermionR;
 typedef ImprovedStaggeredFermion<StaggeredImplF> ImprovedStaggeredFermionF;
 typedef ImprovedStaggeredFermion<StaggeredImplD> ImprovedStaggeredFermionD;
 
-typedef NaiveStaggeredFermion<StaggeredImplR> NaiveStaggeredFermionR;
 typedef NaiveStaggeredFermion<StaggeredImplF> NaiveStaggeredFermionF;
 typedef NaiveStaggeredFermion<StaggeredImplD> NaiveStaggeredFermionD;
 
-typedef ImprovedStaggeredFermion5D<StaggeredImplR> ImprovedStaggeredFermion5DR;
 typedef ImprovedStaggeredFermion5D<StaggeredImplF> ImprovedStaggeredFermion5DF;
 typedef ImprovedStaggeredFermion5D<StaggeredImplD> ImprovedStaggeredFermion5DD;
 

From 8f4e2ee545ac138152112f20d75b31cbc22cf77f Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 27 Sep 2022 10:53:46 -0400
Subject: [PATCH 031/146] Double2

---
 Grid/qcd/action/fermion/WilsonImpl.h | 12 +++---------
 1 file changed, 3 insertions(+), 9 deletions(-)

diff --git a/Grid/qcd/action/fermion/WilsonImpl.h b/Grid/qcd/action/fermion/WilsonImpl.h
index 2685796d..c7180115 100644
--- a/Grid/qcd/action/fermion/WilsonImpl.h
+++ b/Grid/qcd/action/fermion/WilsonImpl.h
@@ -37,7 +37,7 @@ NAMESPACE_BEGIN(Grid);
 template <class S, class Representation = FundamentalRepresentation,class Options = CoeffReal >
 class WilsonImpl : public PeriodicGaugeImpl<GaugeImplTypes<S, Representation::Dimension > > {
 public:
-
+  
   static const int Dimension = Representation::Dimension;
   static const bool isFundamental = Representation::isFundamental;
   static const bool LsVectorised=false;
@@ -242,19 +242,13 @@ public:
 typedef WilsonImpl<vComplex,  FundamentalRepresentation, CoeffReal > WilsonImplR;  // Real.. whichever prec
 typedef WilsonImpl<vComplexF, FundamentalRepresentation, CoeffReal > WilsonImplF;  // Float
 typedef WilsonImpl<vComplexD, FundamentalRepresentation, CoeffReal > WilsonImplD;  // Double
-
-//typedef WilsonImpl<vComplex,  FundamentalRepresentation, CoeffRealHalfComms > WilsonImplRL;  // Real.. whichever prec
-//typedef WilsonImpl<vComplexF, FundamentalRepresentation, CoeffRealHalfComms > WilsonImplFH;  // Float
-//typedef WilsonImpl<vComplexD, FundamentalRepresentation, CoeffRealHalfComms > WilsonImplDF;  // Double
+typedef WilsonImpl<vComplexD2, FundamentalRepresentation, CoeffReal > WilsonImplD2;  // Double
 
 typedef WilsonImpl<vComplex,  FundamentalRepresentation, CoeffComplex > ZWilsonImplR; // Real.. whichever prec
 typedef WilsonImpl<vComplexF, FundamentalRepresentation, CoeffComplex > ZWilsonImplF; // Float
 typedef WilsonImpl<vComplexD, FundamentalRepresentation, CoeffComplex > ZWilsonImplD; // Double
+typedef WilsonImpl<vComplexD2, FundamentalRepresentation, CoeffComplex > ZWilsonImplD2; // Double
 
-//typedef WilsonImpl<vComplex,  FundamentalRepresentation, CoeffComplexHalfComms > ZWilsonImplRL; // Real.. whichever prec
-//typedef WilsonImpl<vComplexF, FundamentalRepresentation, CoeffComplexHalfComms > ZWilsonImplFH; // Float
-//typedef WilsonImpl<vComplexD, FundamentalRepresentation, CoeffComplexHalfComms > ZWilsonImplDF; // Double
- 
 typedef WilsonImpl<vComplex,  AdjointRepresentation, CoeffReal > WilsonAdjImplR;   // Real.. whichever prec
 typedef WilsonImpl<vComplexF, AdjointRepresentation, CoeffReal > WilsonAdjImplF;  // Float
 typedef WilsonImpl<vComplexD, AdjointRepresentation, CoeffReal > WilsonAdjImplD;  // Double

From 70c83ec3bec087184308ccbe6d5006260e2c3f9a Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 27 Sep 2022 10:54:23 -0400
Subject: [PATCH 032/146] More instantiations

---
 .../action/fermion/instantiation/generate_instantiations.sh   | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/Grid/qcd/action/fermion/instantiation/generate_instantiations.sh b/Grid/qcd/action/fermion/instantiation/generate_instantiations.sh
index d6845e75..4ccc01e8 100755
--- a/Grid/qcd/action/fermion/instantiation/generate_instantiations.sh
+++ b/Grid/qcd/action/fermion/instantiation/generate_instantiations.sh
@@ -9,6 +9,7 @@ STAG5_IMPL_LIST=""
 WILSON_IMPL_LIST=" \
 	   WilsonImplF \
 	   WilsonImplD \
+	   WilsonImplD2 \
 	   WilsonAdjImplF \
 	   WilsonAdjImplD \
 	   WilsonTwoIndexSymmetricImplF \
@@ -25,8 +26,9 @@ COMPACT_WILSON_IMPL_LIST=" \
 DWF_IMPL_LIST=" \
 	   WilsonImplF \
 	   WilsonImplD \
+	   WilsonImplD2 \
 	   ZWilsonImplF \
-	   ZWilsonImplD "
+	   ZWilsonImplD2 "
 
 GDWF_IMPL_LIST=" \
 	   GparityWilsonImplF \

From 97448a93dc75313d076c85edf6de2fae1e5507bc Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 27 Sep 2022 10:55:25 -0400
Subject: [PATCH 033/146] Double2 compiles and dslash runs

---
 Grid/simd/Grid_vector_types.h | 24 ++++++------
 Grid/simd/Grid_vector_unops.h | 74 +++++++++++++++++++++++++++++++----
 Grid/simd/Simd.h              | 10 ++---
 3 files changed, 82 insertions(+), 26 deletions(-)

diff --git a/Grid/simd/Grid_vector_types.h b/Grid/simd/Grid_vector_types.h
index 4f952bb2..fd71a84a 100644
--- a/Grid/simd/Grid_vector_types.h
+++ b/Grid/simd/Grid_vector_types.h
@@ -1055,7 +1055,7 @@ accelerator_inline typename toComplexMapper<Rsimd>::Complexified toComplex(const
 }
 
 
-accelerator_inline void precisionChange(vRealF    *out,vRealD    *in,int nvec)
+accelerator_inline void precisionChange(vRealF    *out,const vRealD    *in,int nvec)
 {
   assert((nvec&0x1)==0);
   for(int m=0;m*2<nvec;m++){
@@ -1063,7 +1063,7 @@ accelerator_inline void precisionChange(vRealF    *out,vRealD    *in,int nvec)
     out[m].v=Optimization::PrecisionChange::DtoS(in[n].v,in[n+1].v);
   }
 }
-accelerator_inline void precisionChange(vRealH    *out,vRealD    *in,int nvec)
+accelerator_inline void precisionChange(vRealH    *out,const vRealD    *in,int nvec)
 {
   assert((nvec&0x3)==0);
   for(int m=0;m*4<nvec;m++){
@@ -1071,7 +1071,7 @@ accelerator_inline void precisionChange(vRealH    *out,vRealD    *in,int nvec)
     out[m].v=Optimization::PrecisionChange::DtoH(in[n].v,in[n+1].v,in[n+2].v,in[n+3].v);
   }
 }
-accelerator_inline void precisionChange(vRealH    *out,vRealF    *in,int nvec)
+accelerator_inline void precisionChange(vRealH    *out,const vRealF    *in,int nvec)
 {
   assert((nvec&0x1)==0);
   for(int m=0;m*2<nvec;m++){
@@ -1079,7 +1079,7 @@ accelerator_inline void precisionChange(vRealH    *out,vRealF    *in,int nvec)
     out[m].v=Optimization::PrecisionChange::StoH(in[n].v,in[n+1].v);
   }
 }
-accelerator_inline void precisionChange(vRealD    *out,vRealF    *in,int nvec)
+accelerator_inline void precisionChange(vRealD    *out,const vRealF    *in,int nvec)
 {
   assert((nvec&0x1)==0);
   for(int m=0;m*2<nvec;m++){
@@ -1095,7 +1095,7 @@ accelerator_inline void precisionChange(vRealD    *out,vRealF    *in,int nvec)
     //  |                                                 ~~~~~~~^
   }
 }
-accelerator_inline void precisionChange(vRealD    *out,vRealH    *in,int nvec)
+accelerator_inline void precisionChange(vRealD    *out,const vRealH    *in,int nvec)
 {
   assert((nvec&0x3)==0);
   for(int m=0;m*4<nvec;m++){
@@ -1103,7 +1103,7 @@ accelerator_inline void precisionChange(vRealD    *out,vRealH    *in,int nvec)
     Optimization::PrecisionChange::HtoD(in[m].v,out[n].v,out[n+1].v,out[n+2].v,out[n+3].v);
   }
 }
-accelerator_inline void precisionChange(vRealF    *out,vRealH    *in,int nvec)
+accelerator_inline void precisionChange(vRealF    *out,const vRealH    *in,int nvec)
 {
   assert((nvec&0x1)==0);
   for(int m=0;m*2<nvec;m++){
@@ -1111,12 +1111,12 @@ accelerator_inline void precisionChange(vRealF    *out,vRealH    *in,int nvec)
     Optimization::PrecisionChange::HtoS(in[m].v,out[n].v,out[n+1].v);
   }
 }
-accelerator_inline void precisionChange(vComplexF *out,vComplexD *in,int nvec){ precisionChange((vRealF *)out,(vRealD *)in,nvec);}
-accelerator_inline void precisionChange(vComplexH *out,vComplexD *in,int nvec){ precisionChange((vRealH *)out,(vRealD *)in,nvec);}
-accelerator_inline void precisionChange(vComplexH *out,vComplexF *in,int nvec){ precisionChange((vRealH *)out,(vRealF *)in,nvec);}
-accelerator_inline void precisionChange(vComplexD *out,vComplexF *in,int nvec){ precisionChange((vRealD *)out,(vRealF *)in,nvec);}
-accelerator_inline void precisionChange(vComplexD *out,vComplexH *in,int nvec){ precisionChange((vRealD *)out,(vRealH *)in,nvec);}
-accelerator_inline void precisionChange(vComplexF *out,vComplexH *in,int nvec){ precisionChange((vRealF *)out,(vRealH *)in,nvec);}
+accelerator_inline void precisionChange(vComplexF *out,const vComplexD *in,int nvec){ precisionChange((vRealF *)out,(vRealD *)in,nvec);}
+accelerator_inline void precisionChange(vComplexH *out,const vComplexD *in,int nvec){ precisionChange((vRealH *)out,(vRealD *)in,nvec);}
+accelerator_inline void precisionChange(vComplexH *out,const vComplexF *in,int nvec){ precisionChange((vRealH *)out,(vRealF *)in,nvec);}
+accelerator_inline void precisionChange(vComplexD *out,const vComplexF *in,int nvec){ precisionChange((vRealD *)out,(vRealF *)in,nvec);}
+accelerator_inline void precisionChange(vComplexD *out,const vComplexH *in,int nvec){ precisionChange((vRealD *)out,(vRealH *)in,nvec);}
+accelerator_inline void precisionChange(vComplexF *out,const vComplexH *in,int nvec){ precisionChange((vRealF *)out,(vRealH *)in,nvec);}
 
 // Check our vector types are of an appropriate size.
 
diff --git a/Grid/simd/Grid_vector_unops.h b/Grid/simd/Grid_vector_unops.h
index b89bb785..35f1721c 100644
--- a/Grid/simd/Grid_vector_unops.h
+++ b/Grid/simd/Grid_vector_unops.h
@@ -29,8 +29,7 @@ See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
 			   /*  END LEGAL */
-#ifndef GRID_VECTOR_UNOPS
-#define GRID_VECTOR_UNOPS
+#pragma once
 
 #include <cmath>
 
@@ -112,6 +111,9 @@ template <class scalar>
 struct ImagFunctor {
   accelerator scalar operator()(const scalar &a) const { return imag(a); }
 };
+/////////////
+// Unary operations
+/////////////
 template <class S, class V>
 accelerator_inline Grid_simd<S, V> real(const Grid_simd<S, V> &r) {
   return SimdApply(RealFunctor<S>(), r);
@@ -168,6 +170,65 @@ template <class S, class V>
 accelerator_inline Grid_simd<S, V> div(const Grid_simd<S, V> &r, Integer y) {
   return SimdApply(DivIntFunctor<S>(y), r);
 }
+/// Double 2 cases
+template <class S, class V>
+accelerator_inline Grid_simd2<S, V> real(const Grid_simd2<S, V> &r) {
+  return SimdApply(RealFunctor<S>(), r);
+}
+template <class S, class V>
+accelerator_inline Grid_simd2<S, V> imag(const Grid_simd2<S, V> &r) {
+  return SimdApply(ImagFunctor<S>(), r);
+}
+template <class S, class V>
+accelerator_inline Grid_simd2<S, V> sqrt(const Grid_simd2<S, V> &r) {
+  return SimdApply(SqrtRealFunctor<S>(), r);
+}
+template <class S, class V>
+accelerator_inline Grid_simd2<S, V> cos(const Grid_simd2<S, V> &r) {
+  return SimdApply(CosRealFunctor<S>(), r);
+}
+template <class S, class V>
+accelerator_inline Grid_simd2<S, V> sin(const Grid_simd2<S, V> &r) {
+  return SimdApply(SinRealFunctor<S>(), r);
+}
+template <class S, class V>
+accelerator_inline Grid_simd2<S, V> acos(const Grid_simd2<S, V> &r) {
+  return SimdApply(AcosRealFunctor<S>(), r);
+}
+template <class S, class V>
+accelerator_inline Grid_simd2<S, V> asin(const Grid_simd2<S, V> &r) {
+  return SimdApply(AsinRealFunctor<S>(), r);
+}
+template <class S, class V>
+accelerator_inline Grid_simd2<S, V> log(const Grid_simd2<S, V> &r) {
+  return SimdApply(LogRealFunctor<S>(), r);
+}
+template <class S, class V>
+accelerator_inline Grid_simd2<S, V> abs(const Grid_simd2<S, V> &r) {
+  return SimdApply(AbsRealFunctor<S>(), r);
+}
+template <class S, class V>
+accelerator_inline Grid_simd2<S, V> exp(const Grid_simd2<S, V> &r) {
+  return SimdApply(ExpFunctor<S>(), r);
+}
+template <class S, class V>
+accelerator_inline Grid_simd2<S, V> Not(const Grid_simd2<S, V> &r) {
+  return SimdApply(NotFunctor<S>(), r);
+}
+template <class S, class V>
+accelerator_inline Grid_simd2<S, V> pow(const Grid_simd2<S, V> &r, double y) {
+  return SimdApply(PowRealFunctor<S>(y), r);
+}
+template <class S, class V>
+accelerator_inline Grid_simd2<S, V> mod(const Grid_simd2<S, V> &r, Integer y) {
+  return SimdApply(ModIntFunctor<S>(y), r);
+}
+template <class S, class V>
+accelerator_inline Grid_simd2<S, V> div(const Grid_simd2<S, V> &r, Integer y) {
+  return SimdApply(DivIntFunctor<S>(y), r);
+}
+
+
 ////////////////////////////////////////////////////////////////////////////
 // Allows us to assign into **conformable** real vectors from complex
 ////////////////////////////////////////////////////////////////////////////
@@ -193,23 +254,22 @@ struct OrOrFunctor {
 ////////////////////////////////
 template <class S, class V>
 accelerator_inline Grid_simd<S, V> operator&(const Grid_simd<S, V> &x,
-				 const Grid_simd<S, V> &y) {
+					     const Grid_simd<S, V> &y) {
   return SimdApplyBinop(AndFunctor<S>(), x, y);
 }
 template <class S, class V>
 accelerator_inline Grid_simd<S, V> operator&&(const Grid_simd<S, V> &x,
-				  const Grid_simd<S, V> &y) {
+					      const Grid_simd<S, V> &y) {
   return SimdApplyBinop(AndAndFunctor<S>(), x, y);
 }
 template <class S, class V>
 accelerator_inline Grid_simd<S, V> operator|(const Grid_simd<S, V> &x,
-				 const Grid_simd<S, V> &y) {
+					     const Grid_simd<S, V> &y) {
   return SimdApplyBinop(OrFunctor<S>(), x, y);
 }
 template <class S, class V>
 accelerator_inline Grid_simd<S, V> operator||(const Grid_simd<S, V> &x,
-				  const Grid_simd<S, V> &y) {
+					      const Grid_simd<S, V> &y) {
   return SimdApplyBinop(OrOrFunctor<S>(), x, y);
 }
 NAMESPACE_END(Grid);
-#endif
diff --git a/Grid/simd/Simd.h b/Grid/simd/Simd.h
index 76ca3bef..ddee6a36 100644
--- a/Grid/simd/Simd.h
+++ b/Grid/simd/Simd.h
@@ -224,18 +224,14 @@ accelerator_inline void Gpermute(VectorSIMD &y,const VectorSIMD &b,int perm);
 NAMESPACE_END(Grid);
 
 #include <Grid/simd/Grid_vector_types.h>
+#include <Grid/simd/Grid_doubled_vector.h>
 #include <Grid/simd/Grid_vector_unops.h>
 
 NAMESPACE_BEGIN(Grid);
-// Default precision
-#ifdef GRID_DEFAULT_PRECISION_DOUBLE
+
+// Default precision is wired to double
 typedef vRealD vReal;
 typedef vComplexD vComplex;
-#else
-typedef vRealF vReal;
-typedef vComplexF vComplex;
-#endif
-
  
 inline std::ostream& operator<< (std::ostream& stream, const vComplexF &o){
   int nn=vComplexF::Nsimd();

From 234324599e407e9084665702391fc154c882e0b6 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 27 Sep 2022 10:56:10 -0400
Subject: [PATCH 034/146] Double2

---
 Grid/tensors/Tensor_inner.h  | 20 ++++-----
 Grid/tensors/Tensor_traits.h | 81 +++++-------------------------------
 2 files changed, 18 insertions(+), 83 deletions(-)

diff --git a/Grid/tensors/Tensor_inner.h b/Grid/tensors/Tensor_inner.h
index fd651cae..05f265d7 100644
--- a/Grid/tensors/Tensor_inner.h
+++ b/Grid/tensors/Tensor_inner.h
@@ -214,24 +214,20 @@ accelerator_inline vRealD    innerProductD2(const vRealD    &l,const vRealD    &
 
 accelerator_inline vComplexD2 innerProductD2(const vComplexF &l,const vComplexF &r)
 {  
-  vComplexD la,lb;
-  vComplexD ra,rb;
-  Optimization::PrecisionChange::StoD(l.v,la.v,lb.v);
-  Optimization::PrecisionChange::StoD(r.v,ra.v,rb.v);
+  vComplexD2 dl,dr;
   vComplexD2 ret;
-  ret._internal[0] = innerProduct(la,ra);
-  ret._internal[1] = innerProduct(lb,rb);
+  precisionChange(dl,l);
+  precisionChange(dr,r);
+  ret = innerProduct(dl,dr);
   return ret;
 }
 accelerator_inline vRealD2 innerProductD2(const vRealF &l,const vRealF &r)
 {  
-  vRealD la,lb;
-  vRealD ra,rb;
-  Optimization::PrecisionChange::StoD(l.v,la.v,lb.v);
-  Optimization::PrecisionChange::StoD(r.v,ra.v,rb.v);
+  vRealD2 dl,dr;
   vRealD2 ret;
-  ret._internal[0]=innerProduct(la,ra);
-  ret._internal[1]=innerProduct(lb,rb); 
+  precisionChange(dl,l);
+  precisionChange(dr,r);
+  ret=innerProduct(dl,dr);
   return ret;
 }
 
diff --git a/Grid/tensors/Tensor_traits.h b/Grid/tensors/Tensor_traits.h
index 89290214..99633cee 100644
--- a/Grid/tensors/Tensor_traits.h
+++ b/Grid/tensors/Tensor_traits.h
@@ -42,39 +42,6 @@ NAMESPACE_BEGIN(Grid);
   template<typename T>     struct isGridScalar                : public std::false_type { static constexpr bool notvalue = true; };
   template<class T>        struct isGridScalar<iScalar<T>>    : public std::true_type  { static constexpr bool notvalue = false; };
 
-  // Store double-precision data in single-precision grids for precision promoted localInnerProductD
-  template<typename T>
-  class TypePair {
-  public:
-    T _internal[2];
-    accelerator TypePair<T>& operator=(const Grid::Zero& o) {
-      _internal[0] = Zero();
-      _internal[1] = Zero();
-      return *this;
-    }
-
-    accelerator TypePair<T> operator+(const TypePair<T>& o) const {
-      TypePair<T> r;
-      r._internal[0] = _internal[0] + o._internal[0];
-      r._internal[1] = _internal[1] + o._internal[1];
-      return r;
-    }
-
-    accelerator TypePair<T>& operator+=(const TypePair<T>& o) {
-      _internal[0] += o._internal[0];
-      _internal[1] += o._internal[1];
-      return *this;
-    }
-
-    friend accelerator_inline void add(TypePair<T>* ret, const TypePair<T>* a, const TypePair<T>* b) {
-      add(&ret->_internal[0],&a->_internal[0],&b->_internal[0]);
-      add(&ret->_internal[1],&a->_internal[1],&b->_internal[1]);
-    }
-  };
-  typedef TypePair<ComplexD> ComplexD2;
-  typedef TypePair<RealD> RealD2;
-  typedef TypePair<vComplexD> vComplexD2;
-  typedef TypePair<vRealD> vRealD2;
 
   // Traits to identify fundamental data types
   template<typename T>     struct isGridFundamental                : public std::false_type { static constexpr bool notvalue = true; };
@@ -88,8 +55,6 @@ NAMESPACE_BEGIN(Grid);
   template<>               struct isGridFundamental<RealD>         : public std::true_type  { static constexpr bool notvalue = false; };
   template<>               struct isGridFundamental<vComplexD2>    : public std::true_type  { static constexpr bool notvalue = false; };
   template<>               struct isGridFundamental<vRealD2>       : public std::true_type  { static constexpr bool notvalue = false; };
-  template<>               struct isGridFundamental<ComplexD2>     : public std::true_type  { static constexpr bool notvalue = false; };
-  template<>               struct isGridFundamental<RealD2>        : public std::true_type  { static constexpr bool notvalue = false; };
 
 
 //////////////////////////////////////////////////////////////////////////////////
@@ -136,7 +101,7 @@ NAMESPACE_BEGIN(Grid);
     typedef ComplexF Complexified;
     typedef RealF Realified;
     typedef RealD DoublePrecision;
-    typedef RealD2 DoublePrecision2;
+    typedef RealD DoublePrecision2;
   };
   template<> struct GridTypeMapper<RealD> : public GridTypeMapper_Base {
     typedef RealD scalar_type;
@@ -151,19 +116,6 @@ NAMESPACE_BEGIN(Grid);
     typedef RealD DoublePrecision;
     typedef RealD DoublePrecision2;
   };
-  template<> struct GridTypeMapper<RealD2> : public GridTypeMapper_Base {
-    typedef RealD2 scalar_type;
-    typedef RealD2 scalar_typeD;
-    typedef RealD2 vector_type;
-    typedef RealD2 vector_typeD;
-    typedef RealD2 tensor_reduced;
-    typedef RealD2 scalar_object;
-    typedef RealD2 scalar_objectD;
-    typedef ComplexD2 Complexified;
-    typedef RealD2 Realified;
-    typedef RealD2 DoublePrecision;
-    typedef RealD2 DoublePrecision2;
-  };
   template<> struct GridTypeMapper<ComplexF> : public GridTypeMapper_Base {
     typedef ComplexF scalar_type;
     typedef ComplexD scalar_typeD;
@@ -175,7 +127,7 @@ NAMESPACE_BEGIN(Grid);
     typedef ComplexF Complexified;
     typedef RealF Realified;
     typedef ComplexD DoublePrecision;
-    typedef ComplexD2 DoublePrecision2;
+    typedef ComplexD DoublePrecision2;
   };
   template<> struct GridTypeMapper<ComplexD> : public GridTypeMapper_Base {
     typedef ComplexD scalar_type;
@@ -220,19 +172,6 @@ NAMESPACE_BEGIN(Grid);
   };
 #endif
 
-  template<> struct GridTypeMapper<ComplexD2> : public GridTypeMapper_Base {
-    typedef ComplexD2 scalar_type;
-    typedef ComplexD2 scalar_typeD;
-    typedef ComplexD2 vector_type;
-    typedef ComplexD2 vector_typeD;
-    typedef ComplexD2 tensor_reduced;
-    typedef ComplexD2 scalar_object;
-    typedef ComplexD2 scalar_objectD;
-    typedef ComplexD2 Complexified;
-    typedef RealD2 Realified;
-    typedef ComplexD2 DoublePrecision;
-    typedef ComplexD2 DoublePrecision2;
-  };
   template<> struct GridTypeMapper<Integer> : public GridTypeMapper_Base {
     typedef Integer scalar_type;
     typedef Integer scalar_typeD;
@@ -274,13 +213,13 @@ NAMESPACE_BEGIN(Grid);
     typedef vRealD DoublePrecision2;
   };
   template<> struct GridTypeMapper<vRealD2> : public GridTypeMapper_Base {
-    typedef RealD2  scalar_type;
-    typedef RealD2  scalar_typeD;
+    typedef RealD  scalar_type;
+    typedef RealD  scalar_typeD;
     typedef vRealD2 vector_type;
     typedef vRealD2 vector_typeD;
     typedef vRealD2 tensor_reduced;
-    typedef RealD2  scalar_object;
-    typedef RealD2  scalar_objectD;
+    typedef RealD  scalar_object;
+    typedef RealD  scalar_objectD;
     typedef vComplexD2 Complexified;
     typedef vRealD2 Realified;
     typedef vRealD2 DoublePrecision;
@@ -341,13 +280,13 @@ NAMESPACE_BEGIN(Grid);
     typedef vComplexD DoublePrecision2;
   };
   template<> struct GridTypeMapper<vComplexD2> : public GridTypeMapper_Base {
-    typedef ComplexD2  scalar_type;
-    typedef ComplexD2  scalar_typeD;
+    typedef ComplexD  scalar_type;
+    typedef ComplexD  scalar_typeD;
     typedef vComplexD2 vector_type;
     typedef vComplexD2 vector_typeD;
     typedef vComplexD2 tensor_reduced;
-    typedef ComplexD2  scalar_object;
-    typedef ComplexD2  scalar_objectD;
+    typedef ComplexD  scalar_object;
+    typedef ComplexD  scalar_objectD;
     typedef vComplexD2 Complexified;
     typedef vRealD2 Realified;
     typedef vComplexD2 DoublePrecision;

From af9ecb8b41fe48628b713be8dee191a13bf20b34 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 27 Sep 2022 10:56:55 -0400
Subject: [PATCH 035/146] Current tests compiling

---
 HMC/Mobius2p1f.cc                         |  6 ++--
 HMC/Mobius2p1fEOFA.cc                     |  8 +++---
 HMC/Mobius2p1fEOFA_F1.cc                  |  8 +++---
 HMC/Mobius2p1fRHMC.cc                     |  6 ++--
 HMC/Mobius2p1f_DD_RHMC.cc                 |  2 +-
 HMC/Mobius2p1f_DD_RHMC_96I.cc             |  8 +++---
 HMC/Mobius2p1f_DD_RHMC_96I_mixed.cc       | 10 +++----
 HMC/Mobius2p1f_DD_RHMC_96I_mixedmshift.cc | 34 +++++++++++++++++------
 8 files changed, 50 insertions(+), 32 deletions(-)

diff --git a/HMC/Mobius2p1f.cc b/HMC/Mobius2p1f.cc
index 5f82e0e7..4ab1f20f 100644
--- a/HMC/Mobius2p1f.cc
+++ b/HMC/Mobius2p1f.cc
@@ -39,7 +39,7 @@ int main(int argc, char **argv) {
 
    // Typedefs to simplify notation
   typedef WilsonImplR FermionImplPolicy;
-  typedef MobiusFermionR FermionAction;
+  typedef MobiusFermionD FermionAction;
   typedef typename FermionAction::FermionField FermionField;
 
   typedef Grid::XmlReader       Serialiser;
@@ -133,8 +133,8 @@ int main(int argc, char **argv) {
   ////////////////////////////////////
 
   //  FermionAction StrangeOp(U,*FGrid,*FrbGrid,*GridPtr,*GridRBPtr,light_mass,M5,b,c, Params);
-  //  DomainWallEOFAFermionR Strange_Op_L(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mf, mf, mb, shift_L, pm, M5);
-  //  DomainWallEOFAFermionR Strange_Op_R(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mb, mf, mb, shift_R, pm, M5);
+  //  DomainWallEOFAFermionD Strange_Op_L(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mf, mf, mb, shift_L, pm, M5);
+  //  DomainWallEOFAFermionD Strange_Op_R(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mb, mf, mb, shift_R, pm, M5);
   //  ExactOneFlavourRatioPseudoFermionAction EOFA(Strange_Op_L,Strange_Op_R,CG,ofp, false);
 
   FermionAction StrangeOp (U,*FGrid,*FrbGrid,*GridPtr,*GridRBPtr,strange_mass,M5,b,c, Params);
diff --git a/HMC/Mobius2p1fEOFA.cc b/HMC/Mobius2p1fEOFA.cc
index b1294da5..c961cbc9 100644
--- a/HMC/Mobius2p1fEOFA.cc
+++ b/HMC/Mobius2p1fEOFA.cc
@@ -175,9 +175,9 @@ int main(int argc, char **argv) {
 
    // Typedefs to simplify notation
   typedef WilsonImplR FermionImplPolicy;
-  typedef MobiusFermionR FermionAction;
+  typedef MobiusFermionD FermionAction;
   typedef MobiusFermionF FermionActionF;
-  typedef MobiusEOFAFermionR FermionEOFAAction;
+  typedef MobiusEOFAFermionD FermionEOFAAction;
   typedef MobiusEOFAFermionF FermionEOFAActionF;
   typedef typename FermionAction::FermionField FermionField;
   typedef typename FermionActionF::FermionField FermionFieldF;
@@ -293,9 +293,9 @@ int main(int argc, char **argv) {
   OFRp.precision= 50;
 
   
-  MobiusEOFAFermionR Strange_Op_L (U , *FGrid , *FrbGrid , *GridPtr , *GridRBPtr , strange_mass, strange_mass, pv_mass, 0.0, -1, M5, b, c);
+  MobiusEOFAFermionD Strange_Op_L (U , *FGrid , *FrbGrid , *GridPtr , *GridRBPtr , strange_mass, strange_mass, pv_mass, 0.0, -1, M5, b, c);
   MobiusEOFAFermionF Strange_Op_LF(UF, *FGridF, *FrbGridF, *GridPtrF, *GridRBPtrF, strange_mass, strange_mass, pv_mass, 0.0, -1, M5, b, c);
-  MobiusEOFAFermionR Strange_Op_R (U , *FGrid , *FrbGrid , *GridPtr , *GridRBPtr , pv_mass, strange_mass,      pv_mass, -1.0, 1, M5, b, c);
+  MobiusEOFAFermionD Strange_Op_R (U , *FGrid , *FrbGrid , *GridPtr , *GridRBPtr , pv_mass, strange_mass,      pv_mass, -1.0, 1, M5, b, c);
   MobiusEOFAFermionF Strange_Op_RF(UF, *FGridF, *FrbGridF, *GridPtrF, *GridRBPtrF, pv_mass, strange_mass,      pv_mass, -1.0, 1, M5, b, c);
 
   ConjugateGradient<FermionField>      ActionCG(ActionStoppingCondition,MaxCGIterations);
diff --git a/HMC/Mobius2p1fEOFA_F1.cc b/HMC/Mobius2p1fEOFA_F1.cc
index 3f0a7bf6..f910d69e 100644
--- a/HMC/Mobius2p1fEOFA_F1.cc
+++ b/HMC/Mobius2p1fEOFA_F1.cc
@@ -159,9 +159,9 @@ int main(int argc, char **argv) {
 
    // Typedefs to simplify notation
   typedef WilsonImplR FermionImplPolicy;
-  typedef MobiusFermionR FermionAction;
+  typedef MobiusFermionD FermionAction;
   typedef MobiusFermionF FermionActionF;
-  typedef MobiusEOFAFermionR FermionEOFAAction;
+  typedef MobiusEOFAFermionD FermionEOFAAction;
   typedef MobiusEOFAFermionF FermionEOFAActionF;
   typedef typename FermionAction::FermionField FermionField;
   typedef typename FermionActionF::FermionField FermionFieldF;
@@ -281,9 +281,9 @@ int main(int argc, char **argv) {
   OFRp.precision= 50;
 
   
-  MobiusEOFAFermionR Strange_Op_L (U , *FGrid , *FrbGrid , *GridPtr , *GridRBPtr , strange_mass, strange_mass, pv_mass, 0.0, -1, M5, b, c);
+  MobiusEOFAFermionD Strange_Op_L (U , *FGrid , *FrbGrid , *GridPtr , *GridRBPtr , strange_mass, strange_mass, pv_mass, 0.0, -1, M5, b, c);
   MobiusEOFAFermionF Strange_Op_LF(UF, *FGridF, *FrbGridF, *GridPtrF, *GridRBPtrF, strange_mass, strange_mass, pv_mass, 0.0, -1, M5, b, c);
-  MobiusEOFAFermionR Strange_Op_R (U , *FGrid , *FrbGrid , *GridPtr , *GridRBPtr , pv_mass, strange_mass,      pv_mass, -1.0, 1, M5, b, c);
+  MobiusEOFAFermionD Strange_Op_R (U , *FGrid , *FrbGrid , *GridPtr , *GridRBPtr , pv_mass, strange_mass,      pv_mass, -1.0, 1, M5, b, c);
   MobiusEOFAFermionF Strange_Op_RF(UF, *FGridF, *FrbGridF, *GridPtrF, *GridRBPtrF, pv_mass, strange_mass,      pv_mass, -1.0, 1, M5, b, c);
 
   ConjugateGradient<FermionField>      ActionCG(ActionStoppingCondition,MaxCGIterations);
diff --git a/HMC/Mobius2p1fRHMC.cc b/HMC/Mobius2p1fRHMC.cc
index b958d548..288a6c54 100644
--- a/HMC/Mobius2p1fRHMC.cc
+++ b/HMC/Mobius2p1fRHMC.cc
@@ -39,7 +39,7 @@ int main(int argc, char **argv) {
 
    // Typedefs to simplify notation
   typedef WilsonImplR FermionImplPolicy;
-  typedef MobiusFermionR FermionAction;
+  typedef MobiusFermionD FermionAction;
   typedef typename FermionAction::FermionField FermionField;
 
   typedef Grid::XmlReader       Serialiser;
@@ -137,8 +137,8 @@ int main(int argc, char **argv) {
   ////////////////////////////////////
 
   //  FermionAction StrangeOp(U,*FGrid,*FrbGrid,*GridPtr,*GridRBPtr,light_mass,M5,b,c, Params);
-  //  DomainWallEOFAFermionR Strange_Op_L(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mf, mf, mb, shift_L, pm, M5);
-  //  DomainWallEOFAFermionR Strange_Op_R(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mb, mf, mb, shift_R, pm, M5);
+  //  DomainWallEOFAFermionD Strange_Op_L(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mf, mf, mb, shift_L, pm, M5);
+  //  DomainWallEOFAFermionD Strange_Op_R(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mb, mf, mb, shift_R, pm, M5);
   //  ExactOneFlavourRatioPseudoFermionAction EOFA(Strange_Op_L,Strange_Op_R,CG,ofp, false);
 
   FermionAction StrangeOp (U,*FGrid,*FrbGrid,*GridPtr,*GridRBPtr,strange_mass,M5,b,c, Params);
diff --git a/HMC/Mobius2p1f_DD_RHMC.cc b/HMC/Mobius2p1f_DD_RHMC.cc
index aca408e7..39b4c1dd 100644
--- a/HMC/Mobius2p1f_DD_RHMC.cc
+++ b/HMC/Mobius2p1f_DD_RHMC.cc
@@ -37,7 +37,7 @@ int main(int argc, char **argv) {
 
    // Typedefs to simplify notation
   typedef WilsonImplR FermionImplPolicy;
-  typedef MobiusFermionR FermionAction;
+  typedef MobiusFermionD FermionAction;
   typedef typename FermionAction::FermionField FermionField;
 
   typedef Grid::XmlReader       Serialiser;
diff --git a/HMC/Mobius2p1f_DD_RHMC_96I.cc b/HMC/Mobius2p1f_DD_RHMC_96I.cc
index a6a2f26c..5158aed9 100644
--- a/HMC/Mobius2p1f_DD_RHMC_96I.cc
+++ b/HMC/Mobius2p1f_DD_RHMC_96I.cc
@@ -37,7 +37,7 @@ int main(int argc, char **argv) {
 
    // Typedefs to simplify notation
   typedef WilsonImplR FermionImplPolicy;
-  typedef MobiusFermionR FermionAction;
+  typedef MobiusFermionD FermionAction;
   typedef typename FermionAction::FermionField FermionField;
 
   typedef Grid::XmlReader       Serialiser;
@@ -286,8 +286,8 @@ int main(int argc, char **argv) {
     // ii) Break low bound, how rapidly?
     // iii) Run lanczos
     // iv)  Have CG return spectral range estimate
-    FermionField vec(StrangeOp.FermionRedBlackGrid());
-    FermionField res(StrangeOp.FermionRedBlackGrid());
+    FermionField vec(StrangeOp.FermionDedBlackGrid());
+    FermionField res(StrangeOp.FermionDedBlackGrid());
     vec = 1; // Fill with any old junk
 
     std::cout << "Bounds check on strange operator mass "<< StrangeOp.Mass()<<std::endl;
@@ -327,7 +327,7 @@ int main(int argc, char **argv) {
     
     auto grid4= GridPtr;
     auto rbgrid4= GridRBPtr;
-    auto rbgrid = StrangeOp.FermionRedBlackGrid();
+    auto rbgrid = StrangeOp.FermionDedBlackGrid();
     auto grid = StrangeOp.FermionGrid();
     if(1){
     const int Nstop = 5;
diff --git a/HMC/Mobius2p1f_DD_RHMC_96I_mixed.cc b/HMC/Mobius2p1f_DD_RHMC_96I_mixed.cc
index a01b68fc..36528c9f 100644
--- a/HMC/Mobius2p1f_DD_RHMC_96I_mixed.cc
+++ b/HMC/Mobius2p1f_DD_RHMC_96I_mixed.cc
@@ -156,7 +156,7 @@ int main(int argc, char **argv) {
    // Typedefs to simplify notation
   typedef WilsonImplR FermionImplPolicy;
   typedef WilsonImplF FermionImplPolicyF;
-  typedef MobiusFermionR FermionAction;
+  typedef MobiusFermionD FermionAction;
   typedef MobiusFermionF FermionActionF;
   typedef typename FermionAction::FermionField FermionField;
   typedef typename FermionActionF::FermionField FermionFieldF;
@@ -225,7 +225,7 @@ int main(int argc, char **argv) {
   SFRp.hi       = 90.0;
   SFRp.MaxIter  = 60000;
   SFRp.tolerance= 1.0e-8;
-  SFRp.mdtolerance= 1.0e-6;
+  SFRp.mdtolerance= 1.0e-3;
   SFRp.degree   = 12;
   SFRp.precision= 50;
   SFRp.BoundsCheckFreq=0;
@@ -235,7 +235,7 @@ int main(int argc, char **argv) {
   OFRp.hi       = 90.0;
   OFRp.MaxIter  = 60000;
   OFRp.tolerance= 1.0e-8;
-  OFRp.mdtolerance= 1.0e-6;
+  OFRp.mdtolerance= 1.0e-3;
   //  OFRp.degree   = 20; converges
   //  OFRp.degree   = 16;
   OFRp.degree   = 12;
@@ -274,7 +274,7 @@ int main(int argc, char **argv) {
   Block4[2] = Dirichlet[3];
   Block4[3] = Dirichlet[4];
 
-  int Width=3;
+  int Width=4;
   TheHMC.Resources.SetMomentumFilter(new DDHMCFilter<WilsonImplR::Field>(Block4,Width));
 
   //////////////////////////
@@ -311,7 +311,7 @@ int main(int argc, char **argv) {
   //  double StoppingCondition = 1e-14;
   //  double MDStoppingCondition = 1e-9;
   double StoppingCondition = 1e-10;
-  double MDStoppingCondition = 1e-7;
+  double MDStoppingCondition = 1e-6;
   double MDStoppingConditionLoose = 1e-6;
   double MaxCGIterations = 300000;
   ConjugateGradient<FermionField>  CG(StoppingCondition,MaxCGIterations);
diff --git a/HMC/Mobius2p1f_DD_RHMC_96I_mixedmshift.cc b/HMC/Mobius2p1f_DD_RHMC_96I_mixedmshift.cc
index 732c4666..63315ec3 100644
--- a/HMC/Mobius2p1f_DD_RHMC_96I_mixedmshift.cc
+++ b/HMC/Mobius2p1f_DD_RHMC_96I_mixedmshift.cc
@@ -128,14 +128,9 @@ template<class FermionOperatorD, class FermionOperatorF, class SchurOperatorD, c
       ////////////////////////////////////////////////////////////////////////////////////
       // Make a mixed precision conjugate gradient
       ////////////////////////////////////////////////////////////////////////////////////
-#if 1
       RealD delta=1.e-4;
       std::cout << GridLogMessage << "Calling reliable update Conjugate Gradient" <<std::endl;
       ConjugateGradientReliableUpdate<FieldD,FieldF> MPCG(Tolerance,MaxInnerIterations*MaxOuterIterations,delta,SinglePrecGrid5,LinOpF,LinOpD);
-#else      
-      std::cout << GridLogMessage << "Calling mixed precision Conjugate Gradient" <<std::endl;
-      MixedPrecisionConjugateGradient<FieldD,FieldF> MPCG(Tolerance,MaxInnerIterations,MaxOuterIterations,SinglePrecGrid5,LinOpF,LinOpD);
-#endif
       MPCG(src,psi);
     }
   };
@@ -153,7 +148,7 @@ int main(int argc, char **argv) {
   typedef WilsonImplR FermionImplPolicy;
   typedef WilsonImplF FermionImplPolicyF;
 
-  typedef MobiusFermionR FermionAction;
+  typedef MobiusFermionD FermionAction;
   typedef MobiusFermionF FermionActionF;
   typedef typename FermionAction::FermionField FermionField;
   typedef typename FermionActionF::FermionField FermionFieldF;
@@ -300,9 +295,13 @@ int main(int argc, char **argv) {
   // These lines are unecessary if BC are all periodic
   std::vector<Complex> boundary = {1,1,1,-1};
   FermionAction::ImplParams Params(boundary);
-  Params.dirichlet=NonDirichlet;
   FermionAction::ImplParams ParamsDir(boundary);
+  FermionActionF::ImplParams ParamsF(boundary);
+  FermionActionF::ImplParams ParamsDirF(boundary);
+  Params.dirichlet=NonDirichlet;
+  ParamsF.dirichlet=NonDirichlet;
   ParamsDir.dirichlet=Dirichlet;
+  ParamsDirF.dirichlet=Dirichlet;
 
   //  double StoppingCondition = 1e-14;
   //  double MDStoppingCondition = 1e-9;
@@ -323,15 +322,34 @@ int main(int argc, char **argv) {
   ////////////////////////////////////
   // Strange action
   ////////////////////////////////////
+
   FermionAction StrangeOp (U,*FGrid,*FrbGrid,*GridPtr,*GridRBPtr,strange_mass,M5,b,c, Params);
   FermionAction StrangePauliVillarsOp(U,*FGrid,*FrbGrid,*GridPtr,*GridRBPtr,pv_mass,  M5,b,c, Params);
 
   FermionAction StrangeOpDir (U,*FGrid,*FrbGrid,*GridPtr,*GridRBPtr,strange_mass,M5,b,c, ParamsDir);
   FermionAction StrangePauliVillarsOpDir(U,*FGrid,*FrbGrid,*GridPtr,*GridRBPtr,pv_mass,  M5,b,c, ParamsDir);
-  
+
+  FermionActionF StrangeOpF (UF,*FGridF,*FrbGridF,*GridPtrF,*GridRBPtrF,strange_mass,M5,b,c, ParamsF);
+  FermionActionF StrangePauliVillarsOpF(UF,*FGridF,*FrbGridF,*GridPtrF,*GridRBPtrF,pv_mass,  M5,b,c, ParamsF);
+
+  FermionActionF StrangeOpDirF (UF,*FGridF,*FrbGridF,*GridPtrF,*GridRBPtrF,strange_mass,M5,b,c, ParamsDirF);
+  FermionActionF StrangePauliVillarsOpDirF(UF,*FGridF,*FrbGridF,*GridPtrF,*GridRBPtrF,pv_mass,  M5,b,c, ParamsDirF);
+
+#if 1  
+  OneFlavourEvenOddRatioRationalMixedPrecPseudoFermionAction<FermionImplPolicy,FermionImplPolicyF> StrangePseudoFermionBdy(StrangeOpDir,StrangeOp,
+															   StrangeOpDirF,StrangeOpF,
+															   SFRp,500);
+  OneFlavourEvenOddRatioRationalMixedPrecPseudoFermionAction<FermionImplPolicy,FermionImplPolicyF> StrangePseudoFermionLocal(StrangePauliVillarsOpDir,StrangeOpDir,
+															     StrangePauliVillarsOpDirF,StrangeOpDirF,
+															     SFRp,500);
+  OneFlavourEvenOddRatioRationalMixedPrecPseudoFermionAction<FermionImplPolicy,FermionImplPolicyF> StrangePseudoFermionPVBdy(StrangePauliVillarsOp,StrangePauliVillarsOpDir,
+															     StrangePauliVillarsOpF,StrangePauliVillarsOpDirF,
+															     SFRp,500);
+#else
   OneFlavourEvenOddRatioRationalPseudoFermionAction<FermionImplPolicy> StrangePseudoFermionBdy(StrangeOpDir,StrangeOp,SFRp);
   OneFlavourEvenOddRatioRationalPseudoFermionAction<FermionImplPolicy> StrangePseudoFermionLocal(StrangePauliVillarsOpDir,StrangeOpDir,SFRp);
   OneFlavourEvenOddRatioRationalPseudoFermionAction<FermionImplPolicy> StrangePseudoFermionPVBdy(StrangePauliVillarsOp,StrangePauliVillarsOpDir,SFRp);
+#endif
   Level1.push_back(&StrangePseudoFermionBdy); // ok
   Level2.push_back(&StrangePseudoFermionLocal);
   Level1.push_back(&StrangePseudoFermionPVBdy); //ok

From 25df2d2c3ba7d3e5f3dd66cbc12d673851d95668 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 27 Sep 2022 10:57:12 -0400
Subject: [PATCH 036/146] Various precision options

---
 benchmarks/Benchmark_dwf_fp32.cc | 113 +++++++++++++++++--------------
 1 file changed, 62 insertions(+), 51 deletions(-)

diff --git a/benchmarks/Benchmark_dwf_fp32.cc b/benchmarks/Benchmark_dwf_fp32.cc
index 5ee764c4..029b2016 100644
--- a/benchmarks/Benchmark_dwf_fp32.cc
+++ b/benchmarks/Benchmark_dwf_fp32.cc
@@ -123,26 +123,44 @@ void Benchmark(int Ls, Coordinate Dirichlet)
 
   long unsigned int single_site_flops = 8*Nc*(7+16*Nc);
 
-  GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplexF::Nsimd()),GridDefaultMpi());
+  std::vector<int> seeds4({1,2,3,4});
+  std::vector<int> seeds5({5,6,7,8});
+#define DOUBLE
+#ifdef SINGLE
+  typedef vComplexF          Simd;
+  typedef LatticeFermionF    FermionField;
+  typedef LatticeGaugeFieldF GaugeField;
+  typedef LatticeColourMatrixF ColourMatrixField;
+  typedef DomainWallFermionF FermionAction;
+#endif
+#ifdef DOUBLE
+  typedef vComplexD          Simd;
+  typedef LatticeFermionD    FermionField;
+  typedef LatticeGaugeFieldD GaugeField;
+  typedef LatticeColourMatrixD ColourMatrixField;
+  typedef DomainWallFermionD FermionAction;
+#endif
+#ifdef DOUBLE2
+  typedef vComplexD2          Simd;
+  typedef LatticeFermionD2    FermionField;
+  typedef LatticeGaugeFieldD2 GaugeField;
+  typedef LatticeColourMatrixD2 ColourMatrixField;
+  typedef DomainWallFermionD2 FermionAction;
+#endif
+  
+  GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,Simd::Nsimd()),GridDefaultMpi());
   GridRedBlackCartesian * UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
   GridCartesian         * FGrid   = SpaceTimeGrid::makeFiveDimGrid(Ls,UGrid);
   GridRedBlackCartesian * FrbGrid = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGrid);
 
-  GridCartesian         * sUGrid   = SpaceTimeGrid::makeFourDimDWFGrid(GridDefaultLatt(),GridDefaultMpi());
-  GridRedBlackCartesian * sUrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(sUGrid);
-  GridCartesian         * sFGrid   = SpaceTimeGrid::makeFiveDimDWFGrid(Ls,UGrid);
-  GridRedBlackCartesian * sFrbGrid = SpaceTimeGrid::makeFiveDimDWFRedBlackGrid(Ls,UGrid);
-
-  std::vector<int> seeds4({1,2,3,4});
-  std::vector<int> seeds5({5,6,7,8});
-
   std::cout << GridLogMessage << "Initialising 4d RNG" << std::endl;
   GridParallelRNG          RNG4(UGrid);  RNG4.SeedUniqueString(std::string("The 4D RNG"));
 
   std::cout << GridLogMessage << "Initialising 5d RNG" << std::endl;
   GridParallelRNG          RNG5(FGrid);  RNG5.SeedUniqueString(std::string("The 5D RNG"));
 
-  LatticeFermionF src   (FGrid); random(RNG5,src);
+ 
+  FermionField src   (FGrid); random(RNG5,src);
 #if 0
   src = Zero();
   {
@@ -158,14 +176,14 @@ void Benchmark(int Ls, Coordinate Dirichlet)
   src = src*N2;
 #endif
 
-  LatticeFermionF result(FGrid); result=Zero();
-  LatticeFermionF    ref(FGrid);    ref=Zero();
-  LatticeFermionF    tmp(FGrid);
-  LatticeFermionF    err(FGrid);
+  FermionField result(FGrid); result=Zero();
+  FermionField    ref(FGrid);    ref=Zero();
+  FermionField    tmp(FGrid);
+  FermionField    err(FGrid);
 
   std::cout << GridLogMessage << "Drawing gauge field" << std::endl;
-  LatticeGaugeFieldF Umu(UGrid);
-  LatticeGaugeFieldF UmuCopy(UGrid);
+  GaugeField Umu(UGrid);
+  GaugeField UmuCopy(UGrid);
   SU<Nc>::HotConfiguration(RNG4,Umu);
   UmuCopy=Umu;
   std::cout << GridLogMessage << "Random gauge initialised " << std::endl;
@@ -179,13 +197,13 @@ void Benchmark(int Ls, Coordinate Dirichlet)
   std::cout << GridLogMessage << "Applying BCs for Dirichlet Block5 " << Dirichlet << std::endl;
   std::cout << GridLogMessage << "Applying BCs for Dirichlet Block4 " << Block << std::endl;
 
-  DirichletFilter<LatticeGaugeFieldF> Filter(Block);
+  DirichletFilter<GaugeField> Filter(Block);
   Filter.applyFilter(Umu);
   
   ////////////////////////////////////
   // Naive wilson implementation
   ////////////////////////////////////
-  std::vector<LatticeColourMatrixF> U(4,UGrid);
+  std::vector<ColourMatrixField> U(4,UGrid);
   for(int mu=0;mu<Nd;mu++){
     U[mu] = PeekIndex<LorentzIndex>(Umu,mu);
   }
@@ -236,10 +254,8 @@ void Benchmark(int Ls, Coordinate Dirichlet)
   std::cout << GridLogMessage<< "*****************************************************************" <<std::endl;
   std::cout << GridLogMessage<< "*****************************************************************" <<std::endl;
   std::cout << GridLogMessage<< "* Benchmarking DomainWallFermionR::Dhop                  "<<std::endl;
-  std::cout << GridLogMessage<< "* Vectorising space-time by "<<vComplexF::Nsimd()<<std::endl;
-  std::cout << GridLogMessage<< "* VComplexF size is "<<sizeof(vComplexF)<< " B"<<std::endl;
-  if ( sizeof(RealF)==4 )   std::cout << GridLogMessage<< "* SINGLE precision "<<std::endl;
-  if ( sizeof(RealF)==8 )   std::cout << GridLogMessage<< "* DOUBLE precision "<<std::endl;
+  std::cout << GridLogMessage<< "* Vectorising space-time by "<<Simd::Nsimd()<<std::endl;
+  std::cout << GridLogMessage<< "* VComplex size is "<<sizeof(Simd)<< " B"<<std::endl;
 #ifdef GRID_OMP
   if ( WilsonKernelsStatic::Comms == WilsonKernelsStatic::CommsAndCompute ) std::cout << GridLogMessage<< "* Using Overlapped Comms/Compute" <<std::endl;
   if ( WilsonKernelsStatic::Comms == WilsonKernelsStatic::CommsThenCompute) std::cout << GridLogMessage<< "* Using sequential comms compute" <<std::endl;
@@ -249,13 +265,14 @@ void Benchmark(int Ls, Coordinate Dirichlet)
   if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptInlineAsm ) std::cout << GridLogMessage<< "* Using Asm Nc=3   WilsonKernels" <<std::endl;
   std::cout << GridLogMessage<< "*****************************************************************" <<std::endl;
 
-  DomainWallFermionF::ImplParams p;
+  FermionAction::ImplParams p;
   p.dirichlet=Dirichlet;
-  DomainWallFermionF Dw(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,p);
+  FermionAction Dw(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,p);
   Dw.ImportGauge(Umu);
   
   int ncall =300;
-
+  RealD n2e;
+  
   if (1) {
     FGrid->Barrier();
     Dw.Dhop(src,result,0);
@@ -270,8 +287,8 @@ void Benchmark(int Ls, Coordinate Dirichlet)
     double volume=Ls;  for(int mu=0;mu<Nd;mu++) volume=volume*latt4[mu];
     double flops=single_site_flops*volume*ncall;
 
-    auto nsimd = vComplex::Nsimd();
-    auto simdwidth = sizeof(vComplex);
+    auto nsimd = Simd::Nsimd();
+    auto simdwidth = sizeof(Simd);
 
     // RF: Nd Wilson * Ls, Nd gauge * Ls, Nc colors
     double data_rf = volume * ((2*Nd+1)*Nd*Nc + 2*Nd*Nc*Nc) * simdwidth / nsimd * ncall / (1024.*1024.*1024.);
@@ -283,17 +300,16 @@ void Benchmark(int Ls, Coordinate Dirichlet)
     std::cout<<GridLogMessage << "mflop/s =   "<< flops/(t1-t0)<<std::endl;
     std::cout<<GridLogMessage << "mflop/s per rank =  "<< flops/(t1-t0)/NP<<std::endl;
     std::cout<<GridLogMessage << "mflop/s per node =  "<< flops/(t1-t0)/NN<<std::endl;
-    //    std::cout<<GridLogMessage << "RF  GiB/s (base 2) =   "<< 1000000. * data_rf/((t1-t0))<<std::endl;
-    //    std::cout<<GridLogMessage << "mem GiB/s (base 2) =   "<< 1000000. * data_mem/((t1-t0))<<std::endl;
     err = ref-result;
-    std::cout<<GridLogMessage << "norm diff   "<< norm2(err)<<std::endl;
+    n2e = norm2(err);
+    std::cout<<GridLogMessage << "norm diff   "<< n2e<< "  Line "<<__LINE__ <<std::endl;
 
-    if(( norm2(err)>1.0e-4) ) {
+    if(( n2e>1.0e-4) ) {
       std::cout<<GridLogMessage << "WRONG RESULT" << std::endl;
       FGrid->Barrier();
       exit(-1);
     }
-    assert (norm2(err)< 1.0e-4 );
+    assert (n2e< 1.0e-4 );
   }
 
   if (1)
@@ -348,14 +364,16 @@ void Benchmark(int Ls, Coordinate Dirichlet)
   std::cout<<GridLogMessage << "norm dag result "<< norm2(result)<<std::endl;
   std::cout<<GridLogMessage << "norm dag ref    "<< norm2(ref)<<std::endl;
   err = ref-result;
-  std::cout<<GridLogMessage << "norm dag diff   "<< norm2(err)<<std::endl;
-  assert((norm2(err)<1.0e-4));
+  n2e= norm2(err);
+  std::cout<<GridLogMessage << "norm dag diff   "<< n2e<< "  Line "<<__LINE__ <<std::endl;
+
+  assert((n2e)<1.0e-4);
   
-  LatticeFermionF src_e (FrbGrid);
-  LatticeFermionF src_o (FrbGrid);
-  LatticeFermionF r_e   (FrbGrid);
-  LatticeFermionF r_o   (FrbGrid);
-  LatticeFermionF r_eo  (FGrid);
+  FermionField src_e (FrbGrid);
+  FermionField src_o (FrbGrid);
+  FermionField r_e   (FrbGrid);
+  FermionField r_o   (FrbGrid);
+  FermionField r_eo  (FGrid);
 
   std::cout<<GridLogMessage << "Calling Deo and Doe and //assert Deo+Doe == Dunprec"<<std::endl;
   pickCheckerboard(Even,src_e,src);
@@ -367,10 +385,8 @@ void Benchmark(int Ls, Coordinate Dirichlet)
 
   // S-direction is INNERMOST and takes no part in the parity.
   std::cout << GridLogMessage<< "*********************************************************" <<std::endl;
-  std::cout << GridLogMessage<< "* Benchmarking DomainWallFermionF::DhopEO                "<<std::endl;
-  std::cout << GridLogMessage<< "* Vectorising space-time by "<<vComplexF::Nsimd()<<std::endl;
-  if ( sizeof(RealF)==4 )   std::cout << GridLogMessage<< "* SINGLE precision "<<std::endl;
-  if ( sizeof(RealF)==8 )   std::cout << GridLogMessage<< "* DOUBLE precision "<<std::endl;
+  std::cout << GridLogMessage<< "* Benchmarking DomainWallFermion::DhopEO                "<<std::endl;
+  std::cout << GridLogMessage<< "* Vectorising space-time by "<<Simd::Nsimd()<<std::endl;
 #ifdef GRID_OMP
   if ( WilsonKernelsStatic::Comms == WilsonKernelsStatic::CommsAndCompute ) std::cout << GridLogMessage<< "* Using Overlapped Comms/Compute" <<std::endl;
   if ( WilsonKernelsStatic::Comms == WilsonKernelsStatic::CommsThenCompute) std::cout << GridLogMessage<< "* Using sequential comms compute" <<std::endl;
@@ -384,13 +400,7 @@ void Benchmark(int Ls, Coordinate Dirichlet)
     Dw.DhopEO(src_o,r_e,DaggerNo);
     double t0=usecond();
     for(int i=0;i<ncall;i++){
-#ifdef CUDA_PROFILE
-      if(i==10) cudaProfilerStart();
-#endif
       Dw.DhopEO(src_o,r_e,DaggerNo);
-#ifdef CUDA_PROFILE
-      if(i==20) cudaProfilerStop();
-#endif
     }
     double t1=usecond();
     FGrid->Barrier();
@@ -414,8 +424,9 @@ void Benchmark(int Ls, Coordinate Dirichlet)
   setCheckerboard(r_eo,r_e);
 
   err = r_eo-result;
-  std::cout<<GridLogMessage << "norm diff   "<< norm2(err)<<std::endl;
-  assert(norm2(err)<1.0e-4);
+  n2e= norm2(err);
+  std::cout<<GridLogMessage << "norm diff   "<< n2e<< "  Line "<<__LINE__ <<std::endl;
+  assert(n2e<1.0e-4);
 
   pickCheckerboard(Even,src_e,err);
   pickCheckerboard(Odd,src_o,err);

From 48165c1dc12e1d630f5fd5c40e191ba108ed7cee Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 27 Sep 2022 10:58:00 -0400
Subject: [PATCH 037/146] Ticked off a few items

---
 TODO | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)

diff --git a/TODO b/TODO
index e23e040d..12c523a6 100644
--- a/TODO
+++ b/TODO
@@ -1,3 +1,11 @@
+======
+DDHMC
+======
+-- Reliable Update CG - DONE
+-- Multishift Mixed Precision - DONE
+-- Pole dependent residual  - DONE
+
+=======
 -- comms threads issue??
 -- Part done: Staggered kernel performance on GPU
 
@@ -8,7 +16,7 @@ General
 - Make representations code take Gimpl
 - Simplify the HMCand remove modules
 - Lattice_arith - are the mult, mac etc.. still needed after ET engine?
-- Lattice_rng - faster local only loop in init
+- Lattice_rng - faster local only loop in init -- DDHMC
 - Audit: accelerate A2Autils -- off critical path for HMC
 
 =========================================================

From fad2f969d9ec25ccaf028ee4aca65874a3f778b2 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 27 Sep 2022 10:58:43 -0400
Subject: [PATCH 038/146] Summit up to date

---
 systems/Summit/comms.4node    | 179 ----------------------------
 systems/Summit/config-command |   2 +-
 systems/Summit/dwf.24.4node   | 218 ++++++++++------------------------
 systems/Summit/dwf.32.4node   | 197 +++++++-----------------------
 systems/Summit/dwf4.lsf       |  13 +-
 5 files changed, 112 insertions(+), 497 deletions(-)
 delete mode 100644 systems/Summit/comms.4node

diff --git a/systems/Summit/comms.4node b/systems/Summit/comms.4node
deleted file mode 100644
index b0df0801..00000000
--- a/systems/Summit/comms.4node
+++ /dev/null
@@ -1,179 +0,0 @@
-OPENMPI detected
-AcceleratorCudaInit[0]: ========================
-AcceleratorCudaInit[0]: Device Number    : 0
-AcceleratorCudaInit[0]: ========================
-AcceleratorCudaInit[0]: Device identifier: Tesla V100-SXM2-16GB
-AcceleratorCudaInit[0]:   totalGlobalMem: 16911433728 
-AcceleratorCudaInit[0]:   managedMemory: 1 
-AcceleratorCudaInit[0]:   isMultiGpuBoard: 0 
-AcceleratorCudaInit[0]:   warpSize: 32 
-AcceleratorCudaInit[0]:   pciBusID: 4 
-AcceleratorCudaInit[0]:   pciDeviceID: 0 
-AcceleratorCudaInit[0]: maxGridSize (2147483647,65535,65535)
-AcceleratorCudaInit: rank 0 setting device to node rank 0
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-local rank 0 device 0 bus id: 0004:04:00.0
-AcceleratorCudaInit: ================================================
-SharedMemoryMpi:  World communicator of size 24
-SharedMemoryMpi:  Node  communicator of size 6
-0SharedMemoryMpi:  SharedMemoryMPI.cc acceleratorAllocDevice 1073741824bytes at 0x200060000000 for comms buffers 
-Setting up IPC
-
-__|__|__|__|__|__|__|__|__|__|__|__|__|__|__
-__|__|__|__|__|__|__|__|__|__|__|__|__|__|__
-__|_ |  |  |  |  |  |  |  |  |  |  |  | _|__
-__|_                                    _|__
-__|_   GGGG    RRRR    III    DDDD      _|__
-__|_  G        R   R    I     D   D     _|__
-__|_  G        R   R    I     D    D    _|__
-__|_  G  GG    RRRR     I     D    D    _|__
-__|_  G   G    R  R     I     D   D     _|__
-__|_   GGGG    R   R   III    DDDD      _|__
-__|_                                    _|__
-__|__|__|__|__|__|__|__|__|__|__|__|__|__|__
-__|__|__|__|__|__|__|__|__|__|__|__|__|__|__
-  |  |  |  |  |  |  |  |  |  |  |  |  |  |  
-
-
-Copyright (C) 2015 Peter Boyle, Azusa Yamaguchi, Guido Cossu, Antonin Portelli and other authors
-
-This program is free software; you can redistribute it and/or modify
-it under the terms of the GNU General Public License as published by
-the Free Software Foundation; either version 2 of the License, or
-(at your option) any later version.
-
-This program is distributed in the hope that it will be useful,
-but WITHOUT ANY WARRANTY; without even the implied warranty of
-MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-GNU General Public License for more details.
-Current Grid git commit hash=7cb1ff7395a5833ded6526c43891bd07a0436290: (HEAD -> develop, origin/develop, origin/HEAD) clean
-
-Grid : Message : ================================================ 
-Grid : Message : MPI is initialised and logging filters activated 
-Grid : Message : ================================================ 
-Grid : Message : Requested 1073741824 byte stencil comms buffers 
-AcceleratorCudaInit: rank 1 setting device to node rank 1
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-local rank 1 device 1 bus id: 0004:05:00.0
-AcceleratorCudaInit: rank 2 setting device to node rank 2
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-local rank 2 device 2 bus id: 0004:06:00.0
-AcceleratorCudaInit: rank 5 setting device to node rank 5
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-local rank 5 device 5 bus id: 0035:05:00.0
-AcceleratorCudaInit: rank 4 setting device to node rank 4
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-local rank 4 device 4 bus id: 0035:04:00.0
-AcceleratorCudaInit: rank 3 setting device to node rank 3
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-local rank 3 device 3 bus id: 0035:03:00.0
-Grid : Message : MemoryManager Cache 13529146982 bytes 
-Grid : Message : MemoryManager::Init() setting up
-Grid : Message : MemoryManager::Init() cache pool for recent allocations: SMALL 8 LARGE 2
-Grid : Message : MemoryManager::Init() Non unified: Caching accelerator data in dedicated memory
-Grid : Message : MemoryManager::Init() Using cudaMalloc
-Grid : Message : 2.137929 s : Grid is setup to use 6 threads
-Grid : Message : 2.137941 s : Number of iterations to average: 250
-Grid : Message : 2.137950 s : ====================================================================================================
-Grid : Message : 2.137958 s : = Benchmarking sequential halo exchange from host memory 
-Grid : Message : 2.137966 s : ====================================================================================================
-Grid : Message : 2.137974 s :  L  	 Ls  	    bytes		MB/s uni	MB/s bidi
-AcceleratorCudaInit: rank 22 setting device to node rank 4
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 10 setting device to node rank 4
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 15 setting device to node rank 3
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 21 setting device to node rank 3
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 20 setting device to node rank 2
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 7 setting device to node rank 1
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 9 setting device to node rank 3
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 11 setting device to node rank 5
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 8 setting device to node rank 2
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 6 setting device to node rank 0
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 19 setting device to node rank 1
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 23 setting device to node rank 5
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 18 setting device to node rank 0
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 12 setting device to node rank 0
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 16 setting device to node rank 4
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 13 setting device to node rank 1
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 14 setting device to node rank 2
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 17 setting device to node rank 5
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-Grid : Message : 2.604949 s :    8	8	     393216       89973.9  		179947.8
-Grid : Message : 2.668249 s :    8	8	     393216       18650.3  		37300.5
-Grid : Message : 2.732288 s :    8	8	     393216       18428.5  		36857.1
-Grid : Message : 2.753565 s :    8	8	     393216       55497.2  		110994.4
-Grid : Message : 2.808960 s :   12	8	    1327104       100181.5  		200363.0
-Grid : Message : 3.226900 s :   12	8	    1327104       20600.5  		41201.0
-Grid : Message : 3.167459 s :   12	8	    1327104       24104.6  		48209.2
-Grid : Message : 3.227660 s :   12	8	    1327104       66156.7  		132313.5
-Grid : Message : 3.413570 s :   16	8	    3145728       56174.4  		112348.8
-Grid : Message : 3.802697 s :   16	8	    3145728       24255.9  		48511.7
-Grid : Message : 4.190498 s :   16	8	    3145728       24336.7  		48673.4
-Grid : Message : 4.385171 s :   16	8	    3145728       48484.1  		96968.2
-Grid : Message : 4.805284 s :   20	8	    6144000       46380.5  		92761.1
-Grid : Message : 5.562975 s :   20	8	    6144000       24328.5  		48656.9
-Grid : Message : 6.322562 s :   20	8	    6144000       24266.7  		48533.4
-Grid : Message : 6.773598 s :   20	8	    6144000       40868.5  		81736.9
-Grid : Message : 7.600999 s :   24	8	   10616832       40198.3  		80396.6
-Grid : Message : 8.912917 s :   24	8	   10616832       24279.5  		48559.1
-Grid : Message : 10.220961 s :   24	8	   10616832       24350.2  		48700.4
-Grid : Message : 11.728250 s :   24	8	   10616832       37390.9  		74781.8
-Grid : Message : 12.497258 s :   28	8	   16859136       36792.2  		73584.5
-Grid : Message : 14.585387 s :   28	8	   16859136       24222.2  		48444.3
-Grid : Message : 16.664783 s :   28	8	   16859136       24323.4  		48646.8
-Grid : Message : 17.955238 s :   28	8	   16859136       39194.7  		78389.4
-Grid : Message : 20.136479 s :   32	8	   25165824       35718.3  		71436.5
-Grid : Message : 23.241958 s :   32	8	   25165824       24311.4  		48622.9
-Grid : Message : 26.344810 s :   32	8	   25165824       24331.9  		48663.7
-Grid : Message : 28.384420 s :   32	8	   25165824       37016.3  		74032.7
-Grid : Message : 28.388879 s : ====================================================================================================
-Grid : Message : 28.388894 s : = Benchmarking sequential halo exchange from GPU memory 
-Grid : Message : 28.388909 s : ====================================================================================================
-Grid : Message : 28.388924 s :  L  	 Ls  	    bytes		MB/s uni	MB/s bidi
-Grid : Message : 28.553993 s :    8	8	     393216       8272.4  		16544.7
-Grid : Message : 28.679592 s :    8	8	     393216       9395.4  		18790.8
-Grid : Message : 28.811112 s :    8	8	     393216       8971.0  		17942.0
-Grid : Message : 28.843770 s :    8	8	     393216       36145.6  		72291.2
-Grid : Message : 28.981754 s :   12	8	    1327104       49591.6  		99183.2
-Grid : Message : 29.299764 s :   12	8	    1327104       12520.8  		25041.7
-Grid : Message : 29.620288 s :   12	8	    1327104       12422.2  		24844.4
-Grid : Message : 29.657645 s :   12	8	    1327104       106637.5  		213275.1
-Grid : Message : 29.952933 s :   16	8	    3145728       43939.2  		87878.5
-Grid : Message : 30.585411 s :   16	8	    3145728       14922.1  		29844.2
-Grid : Message : 31.219781 s :   16	8	    3145728       14877.2  		29754.4
-Grid : Message : 31.285017 s :   16	8	    3145728       144724.3  		289448.7
-Grid : Message : 31.706443 s :   20	8	    6144000       54676.2  		109352.4
-Grid : Message : 32.739205 s :   20	8	    6144000       17848.0  		35696.1
-Grid : Message : 33.771852 s :   20	8	    6144000       17849.9  		35699.7
-Grid : Message : 33.871981 s :   20	8	    6144000       184141.4  		368282.8
-Grid : Message : 34.536808 s :   24	8	   10616832       55784.3  		111568.6
-Grid : Message : 36.275648 s :   24	8	   10616832       18317.6  		36635.3
-Grid : Message : 37.997181 s :   24	8	   10616832       18501.7  		37003.4
-Grid : Message : 38.140442 s :   24	8	   10616832       222383.9  		444767.9
-Grid : Message : 39.177222 s :   28	8	   16859136       56609.7  		113219.4
-Grid : Message : 41.874755 s :   28	8	   16859136       18749.9  		37499.8
-Grid : Message : 44.529381 s :   28	8	   16859136       19052.9  		38105.8
-Grid : Message : 44.742192 s :   28	8	   16859136       237717.1  		475434.2
-Grid : Message : 46.184000 s :   32	8	   25165824       57091.2  		114182.4
-Grid : Message : 50.734740 s :   32	8	   25165824       19411.0  		38821.9
-Grid : Message : 53.931228 s :   32	8	   25165824       19570.6  		39141.2
-Grid : Message : 54.238467 s :   32	8	   25165824       245765.6  		491531.2
-Grid : Message : 54.268664 s : ====================================================================================================
-Grid : Message : 54.268680 s : = All done; Bye Bye
-Grid : Message : 54.268691 s : ====================================================================================================
diff --git a/systems/Summit/config-command b/systems/Summit/config-command
index 4caf652e..2a856be0 100644
--- a/systems/Summit/config-command
+++ b/systems/Summit/config-command
@@ -3,7 +3,7 @@
 	      --enable-gen-simd-width=32 \
 	      --enable-unified=no \
 	       --enable-shm=no \
-	       --disable-gparity \
+	       --enable-gparity \
 	       --disable-setdevice \
 	       --disable-fermion-reps \
 	       --enable-accelerator=cuda \
diff --git a/systems/Summit/dwf.24.4node b/systems/Summit/dwf.24.4node
index 212e471c..d0ca9697 100644
--- a/systems/Summit/dwf.24.4node
+++ b/systems/Summit/dwf.24.4node
@@ -10,19 +10,16 @@ AcceleratorCudaInit[0]:   warpSize: 32
 AcceleratorCudaInit[0]:   pciBusID: 4 
 AcceleratorCudaInit[0]:   pciDeviceID: 0 
 AcceleratorCudaInit[0]: maxGridSize (2147483647,65535,65535)
-AcceleratorCudaInit: rank 0 setting device to node rank 0
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
+AcceleratorCudaInit: using default device 
+AcceleratorCudaInit: assume user either uses
+AcceleratorCudaInit: a) IBM jsrun, or 
+AcceleratorCudaInit: b) invokes through a wrapping script to set CUDA_VISIBLE_DEVICES, UCX_NET_DEVICES, and numa binding 
+AcceleratorCudaInit: Configure options --enable-setdevice=no 
 local rank 0 device 0 bus id: 0004:04:00.0
 AcceleratorCudaInit: ================================================
 SharedMemoryMpi:  World communicator of size 24
-SharedMemoryMpi:  Node  communicator of size 6
-0SharedMemoryMpi:  SharedMemoryMPI.cc acceleratorAllocDevice 2147483648bytes at 0x200080000000 for comms buffers 
-AcceleratorCudaInit: rank 3 setting device to node rank 3
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-local rank 3 device 3 bus id: 0035:03:00.0
-AcceleratorCudaInit: rank 5 setting device to node rank 5
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-local rank 5 device 5 bus id: 0035:05:00.0
+SharedMemoryMpi:  Node  communicator of size 1
+0SharedMemoryMpi:  SharedMemoryMPI.cc acceleratorAllocDevice 1073741824bytes at 0x200080000000 - 2000bfffffff for comms buffers 
 Setting up IPC
 
 __|__|__|__|__|__|__|__|__|__|__|__|__|__|__
@@ -36,6 +33,11 @@ __|_  G  GG    RRRR     I     D    D    _|__
 __|_  G   G    R  R     I     D   D     _|__
 __|_   GGGG    R   R   III    DDDD      _|__
 __|_                                    _|__
+local rank 5 device 0 bus id: 0035:05:00.0
+local rank 1 device 0 bus id: 0004:05:00.0
+local rank 2 device 0 bus id: 0004:06:00.0
+local rank 3 device 0 bus id: 0035:03:00.0
+local rank 4 device 0 bus id: 0035:04:00.0
 __|__|__|__|__|__|__|__|__|__|__|__|__|__|__
 __|__|__|__|__|__|__|__|__|__|__|__|__|__|__
   |  |  |  |  |  |  |  |  |  |  |  |  |  |  
@@ -45,15 +47,6 @@ Copyright (C) 2015 Peter Boyle, Azusa Yamaguchi, Guido Cossu, Antonin Portelli a
 
 This program is free software; you can redistribute it and/or modify
 it under the terms of the GNU General Public License as published by
-AcceleratorCudaInit: rank 4 setting device to node rank 4
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-local rank 4 device 4 bus id: 0035:04:00.0
-AcceleratorCudaInit: rank 1 setting device to node rank 1
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-local rank 1 device 1 bus id: 0004:05:00.0
-AcceleratorCudaInit: rank 2 setting device to node rank 2
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-local rank 2 device 2 bus id: 0004:06:00.0
 the Free Software Foundation; either version 2 of the License, or
 (at your option) any later version.
 
@@ -61,146 +54,63 @@ This program is distributed in the hope that it will be useful,
 but WITHOUT ANY WARRANTY; without even the implied warranty of
 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 GNU General Public License for more details.
-Current Grid git commit hash=7cb1ff7395a5833ded6526c43891bd07a0436290: (HEAD -> develop, origin/develop, origin/HEAD) clean
+Current Grid git commit hash=1713de35c0dc339564661dd7df8a72583f889e91: (HEAD -> feature/dirichlet) uncommited changes
 
 Grid : Message : ================================================ 
 Grid : Message : MPI is initialised and logging filters activated 
 Grid : Message : ================================================ 
-Grid : Message : Requested 2147483648 byte stencil comms buffers 
-Grid : Message : MemoryManager Cache 8388608000 bytes 
+Grid : Message : Requested 1073741824 byte stencil comms buffers 
+Grid : Message : MemoryManager Cache 4194304000 bytes 
 Grid : Message : MemoryManager::Init() setting up
 Grid : Message : MemoryManager::Init() cache pool for recent allocations: SMALL 8 LARGE 2
 Grid : Message : MemoryManager::Init() Non unified: Caching accelerator data in dedicated memory
 Grid : Message : MemoryManager::Init() Using cudaMalloc
-Grid : Message : 1.731905 s : Grid Layout
-Grid : Message : 1.731915 s : 	Global lattice size  : 48 48 48 72 
-Grid : Message : 1.731928 s : 	OpenMP threads       : 6
-Grid : Message : 1.731938 s : 	MPI tasks            : 2 2 2 3 
-AcceleratorCudaInit: rank 9 setting device to node rank 3
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 23 setting device to node rank 5
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 22 setting device to node rank 4
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 21 setting device to node rank 3
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 18 setting device to node rank 0
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 6 setting device to node rank 0
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 7 setting device to node rank 1
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 10 setting device to node rank 4
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 8 setting device to node rank 2
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 11 setting device to node rank 5
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 20 setting device to node rank 2
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 19 setting device to node rank 1
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 13 setting device to node rank 1
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 12 setting device to node rank 0
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 14 setting device to node rank 2
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 16 setting device to node rank 4
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 15 setting device to node rank 3
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 17 setting device to node rank 5
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-Grid : Message : 2.683494 s : Making s innermost grids
-Grid : Message : 2.780034 s : Initialising 4d RNG
-Grid : Message : 2.833099 s : Intialising parallel RNG with unique string 'The 4D RNG'
-Grid : Message : 2.833121 s : Seed SHA256: 49db4542db694e3b1a74bf2592a8c1b83bfebbe18401693c2609a4c3af1
-Grid : Message : 2.916841 s : Initialising 5d RNG
-Grid : Message : 3.762880 s : Intialising parallel RNG with unique string 'The 5D RNG'
-Grid : Message : 3.762902 s : Seed SHA256: b6316f2fac44ce14111f93e0296389330b077bfd0a7b359f781c58589f8a
-Grid : Message : 5.264345 s : Initialised RNGs
-Grid : Message : 6.489904 s : Drawing gauge field
-Grid : Message : 6.729262 s : Random gauge initialised 
-Grid : Message : 7.781273 s : Setting up Cshift based reference 
-Grid : Message : 8.725313 s : *****************************************************************
-Grid : Message : 8.725332 s : * Kernel options --dslash-generic, --dslash-unroll, --dslash-asm
-Grid : Message : 8.725342 s : *****************************************************************
-Grid : Message : 8.725352 s : *****************************************************************
-Grid : Message : 8.725362 s : * Benchmarking DomainWallFermionR::Dhop                  
-Grid : Message : 8.725372 s : * Vectorising space-time by 4
-Grid : Message : 8.725383 s : * VComplexF size is 32 B
-Grid : Message : 8.725395 s : * SINGLE precision 
-Grid : Message : 8.725405 s : * Using Overlapped Comms/Compute
-Grid : Message : 8.725415 s : * Using GENERIC Nc WilsonKernels
-Grid : Message : 8.725425 s : *****************************************************************
-Grid : Message : 9.465229 s : Called warmup
-Grid : Message : 58.646066 s : Called Dw 3000 times in 4.91764e+07 us
-Grid : Message : 58.646121 s : mflop/s =   1.02592e+07
-Grid : Message : 58.646134 s : mflop/s per rank =  427468
-Grid : Message : 58.646145 s : mflop/s per node =  2.56481e+06
-Grid : Message : 58.646156 s : RF  GiB/s (base 2) =   20846.5
-Grid : Message : 58.646166 s : mem GiB/s (base 2) =   13029.1
-Grid : Message : 58.648008 s : norm diff   1.04778e-13
-Grid : Message : 58.734885 s : #### Dhop calls report 
-Grid : Message : 58.734897 s : WilsonFermion5D Number of DhopEO Calls   : 6002
-Grid : Message : 58.734909 s : WilsonFermion5D TotalTime   /Calls        : 8217.71 us
-Grid : Message : 58.734922 s : WilsonFermion5D CommTime    /Calls        : 7109.5 us
-Grid : Message : 58.734933 s : WilsonFermion5D FaceTime    /Calls        : 446.623 us
-Grid : Message : 58.734943 s : WilsonFermion5D ComputeTime1/Calls        : 18.0558 us
-Grid : Message : 58.734953 s : WilsonFermion5D ComputeTime2/Calls        : 731.097 us
-Grid : Message : 58.734979 s : Average mflops/s per call                : 4.8157e+09
-Grid : Message : 58.734989 s : Average mflops/s per call per rank       : 2.00654e+08
-Grid : Message : 58.734999 s : Average mflops/s per call per node       : 1.20393e+09
-Grid : Message : 58.735008 s : Average mflops/s per call (full)         : 1.04183e+07
-Grid : Message : 58.735017 s : Average mflops/s per call per rank (full): 434094
-Grid : Message : 58.735026 s : Average mflops/s per call per node (full): 2.60456e+06
-Grid : Message : 58.735035 s : WilsonFermion5D Stencil
-Grid : Message : 58.735043 s : WilsonFermion5D StencilEven
-Grid : Message : 58.735051 s : WilsonFermion5D StencilOdd
-Grid : Message : 58.735059 s : WilsonFermion5D Stencil     Reporti()
-Grid : Message : 58.735067 s : WilsonFermion5D StencilEven Reporti()
-Grid : Message : 58.735075 s : WilsonFermion5D StencilOdd  Reporti()
-Grid : Message : 64.934380 s : Compare to naive wilson implementation Dag to verify correctness
-Grid : Message : 64.934740 s : Called DwDag
-Grid : Message : 64.934870 s : norm dag result 12.0422
-Grid : Message : 64.120756 s : norm dag ref    12.0422
-Grid : Message : 64.149389 s : norm dag diff   7.6644e-14
-Grid : Message : 64.317786 s : Calling Deo and Doe and //assert Deo+Doe == Dunprec
-Grid : Message : 64.465331 s : src_e0.499995
-Grid : Message : 64.524653 s : src_o0.500005
-Grid : Message : 64.558706 s : *********************************************************
-Grid : Message : 64.558717 s : * Benchmarking DomainWallFermionF::DhopEO                
-Grid : Message : 64.558727 s : * Vectorising space-time by 4
-Grid : Message : 64.558737 s : * SINGLE precision 
-Grid : Message : 64.558745 s : * Using Overlapped Comms/Compute
-Grid : Message : 64.558753 s : * Using GENERIC Nc WilsonKernels
-Grid : Message : 64.558761 s : *********************************************************
-Grid : Message : 92.702145 s : Deo mflop/s =   8.97692e+06
-Grid : Message : 92.702185 s : Deo mflop/s per rank   374038
-Grid : Message : 92.702198 s : Deo mflop/s per node   2.24423e+06
-Grid : Message : 92.702209 s : #### Dhop calls report 
-Grid : Message : 92.702223 s : WilsonFermion5D Number of DhopEO Calls   : 3001
-Grid : Message : 92.702240 s : WilsonFermion5D TotalTime   /Calls        : 9377.88 us
-Grid : Message : 92.702257 s : WilsonFermion5D CommTime    /Calls        : 8221.84 us
-Grid : Message : 92.702277 s : WilsonFermion5D FaceTime    /Calls        : 543.548 us
-Grid : Message : 92.702301 s : WilsonFermion5D ComputeTime1/Calls        : 20.936 us
-Grid : Message : 92.702322 s : WilsonFermion5D ComputeTime2/Calls        : 732.33 us
-Grid : Message : 92.702376 s : Average mflops/s per call                : 4.13001e+09
-Grid : Message : 92.702387 s : Average mflops/s per call per rank       : 1.72084e+08
-Grid : Message : 92.702397 s : Average mflops/s per call per node       : 1.0325e+09
-Grid : Message : 92.702407 s : Average mflops/s per call (full)         : 9.12937e+06
-Grid : Message : 92.702416 s : Average mflops/s per call per rank (full): 380391
-Grid : Message : 92.702426 s : Average mflops/s per call per node (full): 2.28234e+06
-Grid : Message : 92.702435 s : WilsonFermion5D Stencil
-Grid : Message : 92.702443 s : WilsonFermion5D StencilEven
-Grid : Message : 92.702451 s : WilsonFermion5D StencilOdd
-Grid : Message : 92.702459 s : WilsonFermion5D Stencil     Reporti()
-Grid : Message : 92.702467 s : WilsonFermion5D StencilEven Reporti()
-Grid : Message : 92.702475 s : WilsonFermion5D StencilOdd  Reporti()
-Grid : Message : 92.772983 s : r_e6.02121
-Grid : Message : 92.786384 s : r_o6.02102
-Grid : Message : 92.799622 s : res12.0422
-Grid : Message : 93.860500 s : norm diff   0
-Grid : Message : 93.162026 s : norm diff even  0
-Grid : Message : 93.197529 s : norm diff odd   0
+
+
+
+
+
+
+
+Grid : Message : 0.179000 s : ++++++++++++++++++++++++++++++++++++++++++++++++
+Grid : Message : 0.196000 s :  Testing with full communication 
+Grid : Message : 0.211000 s : ++++++++++++++++++++++++++++++++++++++++++++++++
+Grid : Message : 0.225000 s : Grid Layout
+Grid : Message : 0.233000 s : 	Global lattice size  : 48 48 48 72 
+Grid : Message : 0.246000 s : 	OpenMP threads       : 6
+Grid : Message : 0.255000 s : 	MPI tasks            : 2 2 2 3 
+Grid : Message : 0.182200 s : Initialising 4d RNG
+Grid : Message : 0.233863 s : Intialising parallel RNG with unique string 'The 4D RNG'
+Grid : Message : 0.233886 s : Seed SHA256: 49db4542db694e3b1a74bf2592a8c1b83bfebbe18401693c2609a4c3af1
+Grid : Message : 0.245805 s : Initialising 5d RNG
+Grid : Message : 1.710720 s : Intialising parallel RNG with unique string 'The 5D RNG'
+Grid : Message : 1.710950 s : Seed SHA256: b6316f2fac44ce14111f93e0296389330b077bfd0a7b359f781c58589f8a
+Grid : Message : 2.220272 s : Drawing gauge field
+Grid : Message : 2.418119 s : Random gauge initialised 
+Grid : Message : 2.418142 s : Applying BCs for Dirichlet Block5 [0 0 0 0 0]
+Grid : Message : 2.418156 s : Applying BCs for Dirichlet Block4 [0 0 0 0]
+Grid : Message : 2.489588 s : Setting up Cshift based reference 
+Grid : Message : 13.921239 s : *****************************************************************
+Grid : Message : 13.921261 s : * Kernel options --dslash-generic, --dslash-unroll, --dslash-asm
+Grid : Message : 13.921270 s : *****************************************************************
+Grid : Message : 13.921279 s : *****************************************************************
+Grid : Message : 13.921288 s : * Benchmarking DomainWallFermionR::Dhop                  
+Grid : Message : 13.921296 s : * Vectorising space-time by 4
+Grid : Message : 13.921305 s : * VComplexF size is 32 B
+Grid : Message : 13.921314 s : * SINGLE precision 
+Grid : Message : 13.921321 s : * Using Overlapped Comms/Compute
+Grid : Message : 13.921328 s : * Using GENERIC Nc WilsonKernels
+Grid : Message : 13.921335 s : *****************************************************************
+Grid : Message : 14.821339 s : Called warmup
+Grid : Message : 23.975467 s : Called Dw 300 times in 9.15155e+06 us
+Grid : Message : 23.975528 s : mflop/s =   5.51286e+06
+Grid : Message : 23.975543 s : mflop/s per rank =  229702
+Grid : Message : 23.975557 s : mflop/s per node =  229702
+Grid : Message : 23.989684 s : norm diff   5.09279e-313  Line 291
+Grid : Message : 39.450493 s : ----------------------------------------------------------------
+Grid : Message : 39.450517 s : Compare to naive wilson implementation Dag to verify correctness
+Grid : Message : 39.450526 s : ----------------------------------------------------------------
+Grid : Message : 39.450534 s : Called DwDag
+Grid : Message : 39.450542 s : norm dag result nan
+Grid : Message : 39.451564 s : norm dag ref    nan
+Grid : Message : 39.455714 s : norm dag diff   nan  Line 354
diff --git a/systems/Summit/dwf.32.4node b/systems/Summit/dwf.32.4node
index eed54f2d..fe21bad8 100644
--- a/systems/Summit/dwf.32.4node
+++ b/systems/Summit/dwf.32.4node
@@ -10,14 +10,21 @@ AcceleratorCudaInit[0]:   warpSize: 32
 AcceleratorCudaInit[0]:   pciBusID: 4 
 AcceleratorCudaInit[0]:   pciDeviceID: 0 
 AcceleratorCudaInit[0]: maxGridSize (2147483647,65535,65535)
-AcceleratorCudaInit: rank 0 setting device to node rank 0
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
+AcceleratorCudaInit: using default device 
+AcceleratorCudaInit: assume user either uses
+AcceleratorCudaInit: a) IBM jsrun, or 
+AcceleratorCudaInit: b) invokes through a wrapping script to set CUDA_VISIBLE_DEVICES, UCX_NET_DEVICES, and numa binding 
+AcceleratorCudaInit: Configure options --enable-setdevice=no 
 local rank 0 device 0 bus id: 0004:04:00.0
 AcceleratorCudaInit: ================================================
 SharedMemoryMpi:  World communicator of size 24
-SharedMemoryMpi:  Node  communicator of size 6
-0SharedMemoryMpi:  SharedMemoryMPI.cc acceleratorAllocDevice 2147483648bytes at 0x200080000000 for comms buffers 
+SharedMemoryMpi:  Node  communicator of size 1
+local rank 3 device 0 bus id: 0004:04:00.0
+local rank 2 device 0 bus id: 0004:04:00.0
+local rank 1 device 0 bus id: 0004:04:00.0
+0SharedMemoryMpi:  SharedMemoryMPI.cc acceleratorAllocDevice 1073741824bytes at 0x200080000000 - 2000bfffffff for comms buffers 
 Setting up IPC
+local rank 5 device 0 bus id: 0004:04:00.0
 
 __|__|__|__|__|__|__|__|__|__|__|__|__|__|__
 __|__|__|__|__|__|__|__|__|__|__|__|__|__|__
@@ -39,168 +46,46 @@ Copyright (C) 2015 Peter Boyle, Azusa Yamaguchi, Guido Cossu, Antonin Portelli a
 
 This program is free software; you can redistribute it and/or modify
 it under the terms of the GNU General Public License as published by
+local rank 4 device 0 bus id: 0004:04:00.0
 the Free Software Foundation; either version 2 of the License, or
 (at your option) any later version.
 
 This program is distributed in the hope that it will be useful,
 but WITHOUT ANY WARRANTY; without even the implied warranty of
 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-AcceleratorCudaInit: rank 2 setting device to node rank 2
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-local rank 2 device 2 bus id: 0004:06:00.0
-AcceleratorCudaInit: rank 1 setting device to node rank 1
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-local rank 1 device 1 bus id: 0004:05:00.0
-AcceleratorCudaInit: rank 4 setting device to node rank 4
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-local rank 4 device 4 bus id: 0035:04:00.0
-AcceleratorCudaInit: rank 3 setting device to node rank 3
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-local rank 3 device 3 bus id: 0035:03:00.0
-AcceleratorCudaInit: rank 5 setting device to node rank 5
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-local rank 5 device 5 bus id: 0035:05:00.0
 GNU General Public License for more details.
-Current Grid git commit hash=7cb1ff7395a5833ded6526c43891bd07a0436290: (HEAD -> develop, origin/develop, origin/HEAD) clean
+Current Grid git commit hash=1713de35c0dc339564661dd7df8a72583f889e91: (HEAD -> feature/dirichlet) uncommited changes
 
 Grid : Message : ================================================ 
 Grid : Message : MPI is initialised and logging filters activated 
 Grid : Message : ================================================ 
-Grid : Message : Requested 2147483648 byte stencil comms buffers 
-Grid : Message : MemoryManager Cache 8388608000 bytes 
+Grid : Message : Requested 1073741824 byte stencil comms buffers 
 Grid : Message : MemoryManager::Init() setting up
 Grid : Message : MemoryManager::Init() cache pool for recent allocations: SMALL 8 LARGE 2
-Grid : Message : MemoryManager::Init() Non unified: Caching accelerator data in dedicated memory
-Grid : Message : MemoryManager::Init() Using cudaMalloc
-Grid : Message : 1.544984 s : Grid Layout
-Grid : Message : 1.544992 s : 	Global lattice size  : 64 64 64 96 
-Grid : Message : 1.545003 s : 	OpenMP threads       : 6
-Grid : Message : 1.545011 s : 	MPI tasks            : 2 2 2 3 
-AcceleratorCudaInit: rank 8 setting device to node rank 2
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 6 setting device to node rank 0
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 11 setting device to node rank 5
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 16 setting device to node rank 4
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 17 setting device to node rank 5
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 13 setting device to node rank 1
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 12 setting device to node rank 0
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 21 setting device to node rank 3
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 23 setting device to node rank 5
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 22 setting device to node rank 4
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 19 setting device to node rank 1
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 18 setting device to node rank 0
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 7 setting device to node rank 1
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 10 setting device to node rank 4
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 9 setting device to node rank 3
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 14 setting device to node rank 2
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 15 setting device to node rank 3
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-AcceleratorCudaInit: rank 20 setting device to node rank 2
-AcceleratorCudaInit: Configure options --enable-setdevice=yes 
-Grid : Message : 2.994920 s : Making s innermost grids
-Grid : Message : 2.232502 s : Initialising 4d RNG
-Grid : Message : 2.397047 s : Intialising parallel RNG with unique string 'The 4D RNG'
-Grid : Message : 2.397069 s : Seed SHA256: 49db4542db694e3b1a74bf2592a8c1b83bfebbe18401693c2609a4c3af1
-Grid : Message : 2.653140 s : Initialising 5d RNG
-Grid : Message : 5.285347 s : Intialising parallel RNG with unique string 'The 5D RNG'
-Grid : Message : 5.285369 s : Seed SHA256: b6316f2fac44ce14111f93e0296389330b077bfd0a7b359f781c58589f8a
-Grid : Message : 9.994738 s : Initialised RNGs
-Grid : Message : 13.153426 s : Drawing gauge field
-Grid : Message : 13.825697 s : Random gauge initialised 
-Grid : Message : 18.537657 s : Setting up Cshift based reference 
-Grid : Message : 22.296755 s : *****************************************************************
-Grid : Message : 22.296781 s : * Kernel options --dslash-generic, --dslash-unroll, --dslash-asm
-Grid : Message : 22.296791 s : *****************************************************************
-Grid : Message : 22.296800 s : *****************************************************************
-Grid : Message : 22.296809 s : * Benchmarking DomainWallFermionR::Dhop                  
-Grid : Message : 22.296818 s : * Vectorising space-time by 4
-Grid : Message : 22.296828 s : * VComplexF size is 32 B
-Grid : Message : 22.296838 s : * SINGLE precision 
-Grid : Message : 22.296847 s : * Using Overlapped Comms/Compute
-Grid : Message : 22.296855 s : * Using GENERIC Nc WilsonKernels
-Grid : Message : 22.296863 s : *****************************************************************
-Grid : Message : 24.746452 s : Called warmup
-Grid : Message : 137.525756 s : Called Dw 3000 times in 1.12779e+08 us
-Grid : Message : 137.525818 s : mflop/s =   1.41383e+07
-Grid : Message : 137.525831 s : mflop/s per rank =  589097
-Grid : Message : 137.525843 s : mflop/s per node =  3.53458e+06
-Grid : Message : 137.525854 s : RF  GiB/s (base 2) =   28728.7
-Grid : Message : 137.525864 s : mem GiB/s (base 2) =   17955.5
-Grid : Message : 137.693645 s : norm diff   1.04885e-13
-Grid : Message : 137.965585 s : #### Dhop calls report 
-Grid : Message : 137.965598 s : WilsonFermion5D Number of DhopEO Calls   : 6002
-Grid : Message : 137.965612 s : WilsonFermion5D TotalTime   /Calls        : 18899.7 us
-Grid : Message : 137.965624 s : WilsonFermion5D CommTime    /Calls        : 16041.4 us
-Grid : Message : 137.965634 s : WilsonFermion5D FaceTime    /Calls        : 859.705 us
-Grid : Message : 137.965644 s : WilsonFermion5D ComputeTime1/Calls        : 70.5881 us
-Grid : Message : 137.965654 s : WilsonFermion5D ComputeTime2/Calls        : 2094.8 us
-Grid : Message : 137.965682 s : Average mflops/s per call                : 3.87638e+09
-Grid : Message : 137.965692 s : Average mflops/s per call per rank       : 1.61516e+08
-Grid : Message : 137.965702 s : Average mflops/s per call per node       : 9.69095e+08
-Grid : Message : 137.965712 s : Average mflops/s per call (full)         : 1.43168e+07
-Grid : Message : 137.965721 s : Average mflops/s per call per rank (full): 596533
-Grid : Message : 137.965730 s : Average mflops/s per call per node (full): 3.5792e+06
-Grid : Message : 137.965740 s : WilsonFermion5D Stencil
-Grid : Message : 137.965748 s : WilsonFermion5D StencilEven
-Grid : Message : 137.965756 s : WilsonFermion5D StencilOdd
-Grid : Message : 137.965764 s : WilsonFermion5D Stencil     Reporti()
-Grid : Message : 137.965772 s : WilsonFermion5D StencilEven Reporti()
-Grid : Message : 137.965780 s : WilsonFermion5D StencilOdd  Reporti()
-Grid : Message : 156.554605 s : Compare to naive wilson implementation Dag to verify correctness
-Grid : Message : 156.554632 s : Called DwDag
-Grid : Message : 156.554642 s : norm dag result 12.0421
-Grid : Message : 156.639265 s : norm dag ref    12.0421
-Grid : Message : 156.888281 s : norm dag diff   7.62057e-14
-Grid : Message : 157.609797 s : Calling Deo and Doe and //assert Deo+Doe == Dunprec
-Grid : Message : 158.208630 s : src_e0.499996
-Grid : Message : 158.162447 s : src_o0.500004
-Grid : Message : 158.267780 s : *********************************************************
-Grid : Message : 158.267791 s : * Benchmarking DomainWallFermionF::DhopEO                
-Grid : Message : 158.267801 s : * Vectorising space-time by 4
-Grid : Message : 158.267811 s : * SINGLE precision 
-Grid : Message : 158.267820 s : * Using Overlapped Comms/Compute
-Grid : Message : 158.267828 s : * Using GENERIC Nc WilsonKernels
-Grid : Message : 158.267836 s : *********************************************************
-Grid : Message : 216.487829 s : Deo mflop/s =   1.37283e+07
-Grid : Message : 216.487869 s : Deo mflop/s per rank   572011
-Grid : Message : 216.487881 s : Deo mflop/s per node   3.43206e+06
-Grid : Message : 216.487893 s : #### Dhop calls report 
-Grid : Message : 216.487903 s : WilsonFermion5D Number of DhopEO Calls   : 3001
-Grid : Message : 216.487913 s : WilsonFermion5D TotalTime   /Calls        : 19399.6 us
-Grid : Message : 216.487923 s : WilsonFermion5D CommTime    /Calls        : 16475.4 us
-Grid : Message : 216.487933 s : WilsonFermion5D FaceTime    /Calls        : 972.393 us
-Grid : Message : 216.487943 s : WilsonFermion5D ComputeTime1/Calls        : 49.8474 us
-Grid : Message : 216.487953 s : WilsonFermion5D ComputeTime2/Calls        : 2089.93 us
-Grid : Message : 216.488001 s : Average mflops/s per call                : 5.39682e+09
-Grid : Message : 216.488011 s : Average mflops/s per call per rank       : 2.24867e+08
-Grid : Message : 216.488020 s : Average mflops/s per call per node       : 1.3492e+09
-Grid : Message : 216.488030 s : Average mflops/s per call (full)         : 1.39479e+07
-Grid : Message : 216.488039 s : Average mflops/s per call per rank (full): 581162
-Grid : Message : 216.488048 s : Average mflops/s per call per node (full): 3.48697e+06
-Grid : Message : 216.488057 s : WilsonFermion5D Stencil
-Grid : Message : 216.488065 s : WilsonFermion5D StencilEven
-Grid : Message : 216.488073 s : WilsonFermion5D StencilOdd
-Grid : Message : 216.488081 s : WilsonFermion5D Stencil     Reporti()
-Grid : Message : 216.488089 s : WilsonFermion5D StencilEven Reporti()
-Grid : Message : 216.488097 s : WilsonFermion5D StencilOdd  Reporti()
-Grid : Message : 217.384495 s : r_e6.02113
-Grid : Message : 217.426121 s : r_o6.02096
-Grid : Message : 217.472636 s : res12.0421
-Grid : Message : 218.200068 s : norm diff   0
-Grid : Message : 218.645673 s : norm diff even  0
-Grid : Message : 218.816561 s : norm diff odd   0
+Grid : Message : MemoryManager::Init() Unified memory space
+Grid : Message : MemoryManager::Init() Using cudaMallocManaged
+
+
+
+
+
+
+
+Grid : Message : 0.139000 s : ++++++++++++++++++++++++++++++++++++++++++++++++
+Grid : Message : 0.151000 s :  Testing with full communication 
+Grid : Message : 0.158000 s : ++++++++++++++++++++++++++++++++++++++++++++++++
+Grid : Message : 0.165000 s : Grid Layout
+Grid : Message : 0.171000 s : 	Global lattice size  : 64 64 64 96 
+Grid : Message : 0.181000 s : 	OpenMP threads       : 6
+Grid : Message : 0.189000 s : 	MPI tasks            : 2 2 2 3 
+Grid : Message : 0.177717 s : Initialising 4d RNG
+Grid : Message : 0.342461 s : Intialising parallel RNG with unique string 'The 4D RNG'
+Grid : Message : 0.342483 s : Seed SHA256: 49db4542db694e3b1a74bf2592a8c1b83bfebbe18401693c2609a4c3af1
+Grid : Message : 0.370454 s : Initialising 5d RNG
+Grid : Message : 3.174160 s : Intialising parallel RNG with unique string 'The 5D RNG'
+Grid : Message : 3.174420 s : Seed SHA256: b6316f2fac44ce14111f93e0296389330b077bfd0a7b359f781c58589f8a
+Grid : Message : 22.119339 s : Drawing gauge field
+Grid : Message : 38.113060 s : Random gauge initialised 
+Grid : Message : 38.113320 s : Applying BCs for Dirichlet Block5 [0 0 0 0 0]
+Grid : Message : 38.113470 s : Applying BCs for Dirichlet Block4 [0 0 0 0]
+Grid : Message : 43.906786 s : Setting up Cshift based reference 
diff --git a/systems/Summit/dwf4.lsf b/systems/Summit/dwf4.lsf
index 7d940338..51560f71 100644
--- a/systems/Summit/dwf4.lsf
+++ b/systems/Summit/dwf4.lsf
@@ -7,16 +7,15 @@
 export OMP_NUM_THREADS=6
 export PAMI_IBV_ADAPTER_AFFINITY=1
 export PAMI_ENABLE_STRIPING=1
+export PAMI_DISABLE_IPC=1
 export OPT="--comms-concurrent --comms-overlap "
-#export GRID_ALLOC_NCACHE_LARGE=1
-export APP="./benchmarks/Benchmark_comms_host_device  --mpi 2.2.2.3 "
-jsrun --nrs 4 -a6 -g6 -c42 -dpacked -b packed:7 --latency_priority gpu-cpu --smpiargs=-gpu $APP > comms.4node
 
-APP="./benchmarks/Benchmark_dwf_fp32 --grid 48.48.48.72 --mpi 2.2.2.3 --shm 2048 --shm-force-mpi 1 --device-mem 8000 --shm-force-mpi 1 $OPT "
-jsrun --nrs 4 -a6 -g6 -c42 -dpacked -b packed:7 --latency_priority gpu-cpu --smpiargs=-gpu $APP > dwf.24.4node
 
-APP="./benchmarks/Benchmark_dwf_fp32 --grid 64.64.64.96 --mpi 2.2.2.3 --shm 2048 --shm-force-mpi 1 --device-mem 8000 --shm-force-mpi 1 $OPT "
-jsrun --nrs 4 -a6 -g6 -c42 -dpacked -b packed:7 --latency_priority gpu-cpu --smpiargs=-gpu $APP > dwf.32.4node
+APP="./wrap.sh ./benchmarks/Benchmark_dwf_fp32 --grid 48.48.48.72 --mpi 2.2.2.3 --shm 1024 --device-mem 4000 --shm-force-mpi 1 $OPT "
+jsrun --nrs 24 -a1 -g1 -c6 -dpacked -b packed:6 --latency_priority gpu-cpu --smpiargs="-gpu" $APP > dwf.24.4node
+
+APP="./wrap.sh ./benchmarks/Benchmark_comms_host_device --grid 48.48.48.72 --mpi 2.2.2.3 --shm 1024 --device-mem 4000 --shm-force-mpi 1 $OPT "
+jsrun  --smpiargs="-gpu"  --nrs 4 -a6 -g6 -c42 -dpacked -b packed:6  $APP > comms.24.4node
 
 
 

From 9296299b61508e8fb4a44fa32f656b7ab1857d9e Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 4 Oct 2022 11:10:34 -0700
Subject: [PATCH 039/146] Better commenting

---
 Grid/qcd/action/fermion/WilsonCompressor.h | 16 ++++++++--------
 1 file changed, 8 insertions(+), 8 deletions(-)

diff --git a/Grid/qcd/action/fermion/WilsonCompressor.h b/Grid/qcd/action/fermion/WilsonCompressor.h
index e0e08c1c..5c3351a6 100644
--- a/Grid/qcd/action/fermion/WilsonCompressor.h
+++ b/Grid/qcd/action/fermion/WilsonCompressor.h
@@ -117,19 +117,19 @@ public:
     typedef decltype(coalescedRead(*in))    sobj;
     typedef decltype(coalescedRead(*out0)) hsobj;
 
-    unsigned int Nsimd = vobj::Nsimd();
+    constexpr unsigned int Nsimd = vobj::Nsimd();
     unsigned int mask = Nsimd >> (type + 1);
     int lane = acceleratorSIMTlane(Nsimd);
     int j0 = lane &(~mask); // inner coor zero
     int j1 = lane |(mask) ; // inner coor one
-    const vobj *vp0 = &in[k];
-    const vobj *vp1 = &in[m];
-    const vobj *vp = (lane&mask) ? vp1:vp0;
-    auto sa = coalescedRead(*vp,j0);
-    auto sb = coalescedRead(*vp,j1);
+    const vobj *vp0 = &in[k];  // out0[j] = merge low bit of type from in[k] and in[m] 
+    const vobj *vp1 = &in[m];  // out1[j] = merge hi  bit of type from in[k] and in[m]
+    const vobj *vp = (lane&mask) ? vp1:vp0;// if my lane has high bit take vp1, low bit take vp0
+    auto sa = coalescedRead(*vp,j0); // lane to read for out 0, NB 50% read coalescing
+    auto sb = coalescedRead(*vp,j1); // lane to read for out 1
     hsobj psa, psb;
-    projector::Proj(psa,sa,mu,dag);
-    projector::Proj(psb,sb,mu,dag);
+    projector::Proj(psa,sa,mu,dag);  // spin project the result0
+    projector::Proj(psb,sb,mu,dag);  // spin project the result1
     coalescedWrite(out0[j],psa);
     coalescedWrite(out1[j],psb);
 #else

From e1e5c7502308e61b3649c87df56e1f3f31c7b90f Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 4 Oct 2022 11:11:10 -0700
Subject: [PATCH 040/146] Stencil gather improvements - SVM was running slow
 and used for a pointer array that wasn't needed to be in SVM

---
 Grid/stencil/Stencil.h | 32 +++++++++++++++++++-------------
 1 file changed, 19 insertions(+), 13 deletions(-)

diff --git a/Grid/stencil/Stencil.h b/Grid/stencil/Stencil.h
index c2bc8dab..65d878cb 100644
--- a/Grid/stencil/Stencil.h
+++ b/Grid/stencil/Stencil.h
@@ -80,11 +80,14 @@ void Gather_plane_simple_table (commVector<std::pair<int,int> >& table,const Lat
 ///////////////////////////////////////////////////////////////////
 template<class cobj,class vobj,class compressor>
 void Gather_plane_exchange_table(const Lattice<vobj> &rhs,
-				 commVector<cobj *> pointers,int dimension,int plane,int cbmask,compressor &compress,int type) __attribute__((noinline));
+				 commVector<cobj *> pointers,
+				 int dimension,int plane,
+				 int cbmask,compressor &compress,int type) __attribute__((noinline));
 
 template<class cobj,class vobj,class compressor>
-void Gather_plane_exchange_table(commVector<std::pair<int,int> >& table,const Lattice<vobj> &rhs,
-				 Vector<cobj *> pointers,int dimension,int plane,int cbmask,
+void Gather_plane_exchange_table(commVector<std::pair<int,int> >& table,
+				 const Lattice<vobj> &rhs,
+				 std::vector<cobj *> &pointers,int dimension,int plane,int cbmask,
 				 compressor &compress,int type)
 {
   assert( (table.size()&0x1)==0);
@@ -92,14 +95,15 @@ void Gather_plane_exchange_table(commVector<std::pair<int,int> >& table,const La
   int so  = plane*rhs.Grid()->_ostride[dimension]; // base offset for start of plane
 
   auto rhs_v = rhs.View(AcceleratorRead);
+  auto rhs_p = &rhs_v[0];
   auto p0=&pointers[0][0];
   auto p1=&pointers[1][0];
   auto tp=&table[0];
   accelerator_forNB(j, num, vobj::Nsimd(), {
-      compress.CompressExchange(p0,p1, &rhs_v[0], j,
-			      so+tp[2*j  ].second,
-			      so+tp[2*j+1].second,
-			      type);
+      compress.CompressExchange(p0,p1, rhs_p, j,
+				so+tp[2*j  ].second,
+				so+tp[2*j+1].second,
+				type);
   });
   rhs_v.ViewClose();
 }
@@ -230,8 +234,8 @@ public:
   };
   struct Merge {
     cobj * mpointer;
-    Vector<scalar_object *> rpointers;
-    Vector<cobj *> vpointers;
+    //    std::vector<scalar_object *> rpointers;
+    std::vector<cobj *> vpointers;
     Integer buffer_size;
     Integer type;
   };
@@ -406,6 +410,7 @@ public:
       comms_bytes+=bytes;
       shm_bytes  +=2*Packets[i].bytes-bytes;
     }
+    _grid->StencilBarrier();// Synch shared memory on a single nodes
   }
 
   void CommunicateComplete(std::vector<std::vector<CommsRequest_t> > &reqs)
@@ -420,7 +425,7 @@ public:
   ////////////////////////////////////////////////////////////////////////
   void Communicate(void)
   {
-    if ( CartesianCommunicator::CommunicatorPolicy == CartesianCommunicator::CommunicatorPolicySequential ){
+    if ( 0 ){
       thread_region {
 	// must be called in parallel region
 	int mythread  = thread_num();
@@ -569,7 +574,7 @@ public:
     d.buffer_size = buffer_size;
     dv.push_back(d);
   }
-  void AddMerge(cobj *merge_p,Vector<cobj *> &rpointers,Integer buffer_size,Integer type,std::vector<Merge> &mv) {
+  void AddMerge(cobj *merge_p,std::vector<cobj *> &rpointers,Integer buffer_size,Integer type,std::vector<Merge> &mv) {
     Merge m;
     m.type     = type;
     m.mpointer = merge_p;
@@ -582,6 +587,7 @@ public:
   }
   template<class decompressor>  void CommsMergeSHM(decompressor decompress) {
     mpi3synctime-=usecond();
+    accelerator_barrier();
     _grid->StencilBarrier();// Synch shared memory on a single nodes
     mpi3synctime+=usecond();
     shmmergetime-=usecond();
@@ -1114,8 +1120,8 @@ public:
     int bytes = (reduced_buffer_size*datum_bytes)/simd_layout;
     assert(bytes*simd_layout == reduced_buffer_size*datum_bytes);
 
-    Vector<cobj *> rpointers(maxl);
-    Vector<cobj *> spointers(maxl);
+    std::vector<cobj *> rpointers(maxl);
+    std::vector<cobj *> spointers(maxl);
 
     ///////////////////////////////////////////
     // Work out what to send where

From 03508448f88c08cfa146e5b0082b930d5de503d7 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 4 Oct 2022 11:12:15 -0700
Subject: [PATCH 041/146] Remove verbose

---
 Grid/threads/Accelerator.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/Grid/threads/Accelerator.h b/Grid/threads/Accelerator.h
index d2cf8a01..e17e85d1 100644
--- a/Grid/threads/Accelerator.h
+++ b/Grid/threads/Accelerator.h
@@ -298,14 +298,14 @@ accelerator_inline int acceleratorSIMTlane(int Nsimd) {
      });	   			              \
     });
 
-#define accelerator_barrier(dummy) { printf(" theGridAccelerator::wait()\n");  theGridAccelerator->wait(); }
+#define accelerator_barrier(dummy) { theGridAccelerator->wait(); }
 
 inline void *acceleratorAllocShared(size_t bytes){ return malloc_shared(bytes,*theGridAccelerator);};
 inline void *acceleratorAllocDevice(size_t bytes){ return malloc_device(bytes,*theGridAccelerator);};
 inline void acceleratorFreeShared(void *ptr){free(ptr,*theGridAccelerator);};
 inline void acceleratorFreeDevice(void *ptr){free(ptr,*theGridAccelerator);};
 
-inline void acceleratorCopySynchronise(void) {  printf(" theCopyAccelerator::wait()\n"); theCopyAccelerator->wait(); }
+inline void acceleratorCopySynchronise(void) {  theCopyAccelerator->wait(); }
 inline void acceleratorCopyDeviceToDeviceAsynch(void *from,void *to,size_t bytes)  {  theCopyAccelerator->memcpy(to,from,bytes);}
 inline void acceleratorCopyToDevice(void *from,void *to,size_t bytes)  { theCopyAccelerator->memcpy(to,from,bytes); theCopyAccelerator->wait();}
 inline void acceleratorCopyFromDevice(void *from,void *to,size_t bytes){ theCopyAccelerator->memcpy(to,from,bytes); theCopyAccelerator->wait();}

From 413312f9a9a98634363f5c5206be19e97ec55fca Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 4 Oct 2022 11:12:59 -0700
Subject: [PATCH 042/146] Benchmark the halo construction. THe bye counts are
 out and should be doubled for SIMD directions

---
 benchmarks/Benchmark_halo.cc | 131 +++++++++++++++++++++++++++++++++++
 1 file changed, 131 insertions(+)
 create mode 100644 benchmarks/Benchmark_halo.cc

diff --git a/benchmarks/Benchmark_halo.cc b/benchmarks/Benchmark_halo.cc
new file mode 100644
index 00000000..43138e67
--- /dev/null
+++ b/benchmarks/Benchmark_halo.cc
@@ -0,0 +1,131 @@
+ /*************************************************************************************
+    Grid physics library, www.github.com/paboyle/Grid
+    Source file: ./benchmarks/Benchmark_dwf.cc
+    Copyright (C) 2015
+
+    Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+    Author: paboyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+    See the full license in the file "LICENSE" in the top level distribution directory
+    *************************************************************************************/
+    /*  END LEGAL */
+#include <Grid/Grid.h>
+#ifdef GRID_CUDA
+#define CUDA_PROFILE
+#endif
+
+#ifdef CUDA_PROFILE
+#include <cuda_profiler_api.h>
+#endif
+
+using namespace std;
+using namespace Grid;
+
+template<class d>
+struct scal {
+  d internal;
+};
+
+  Gamma::Algebra Gmu [] = {
+    Gamma::Algebra::GammaX,
+    Gamma::Algebra::GammaY,
+    Gamma::Algebra::GammaZ,
+    Gamma::Algebra::GammaT
+  };
+
+
+int main (int argc, char ** argv)
+{
+  Grid_init(&argc,&argv);
+
+  Coordinate latt4= GridDefaultLatt();
+  Coordinate mpi  = GridDefaultMpi();
+  Coordinate simd = GridDefaultSimd(Nd,vComplexF::Nsimd());
+
+  GridLogLayout();
+
+  int Ls=16;
+  for(int i=0;i<argc;i++)
+    if(std::string(argv[i]) == "-Ls"){
+      std::stringstream ss(argv[i+1]); ss >> Ls;
+    }
+
+  
+  GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(latt4,simd ,mpi);
+  GridRedBlackCartesian * UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
+  GridCartesian         * FGrid   = SpaceTimeGrid::makeFiveDimGrid(Ls,UGrid);
+  GridRedBlackCartesian * FrbGrid = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGrid);
+
+  std::cout << GridLogMessage << "Making s innermost grids"<<std::endl;
+  GridCartesian         * sUGrid   = SpaceTimeGrid::makeFourDimDWFGrid(GridDefaultLatt(),GridDefaultMpi());
+  GridRedBlackCartesian * sUrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(sUGrid);
+  GridCartesian         * sFGrid   = SpaceTimeGrid::makeFiveDimDWFGrid(Ls,UGrid);
+  GridRedBlackCartesian * sFrbGrid = SpaceTimeGrid::makeFiveDimDWFRedBlackGrid(Ls,UGrid);
+
+  std::vector<int> seeds4({1,2,3,4});
+  std::vector<int> seeds5({5,6,7,8});
+
+  std::cout << GridLogMessage << "Initialising 4d RNG" << std::endl;
+  GridParallelRNG          RNG4(UGrid);  RNG4.SeedUniqueString(std::string("The 4D RNG"));
+  std::cout << GridLogMessage << "Initialising 5d RNG" << std::endl;
+  GridParallelRNG          RNG5(FGrid);  RNG5.SeedUniqueString(std::string("The 5D RNG"));
+  std::cout << GridLogMessage << "Initialised RNGs" << std::endl;
+
+  LatticeFermionF src   (FGrid); random(RNG5,src);
+  RealD N2 = 1.0/::sqrt(norm2(src));
+  src = src*N2;
+
+  std::cout << GridLogMessage << "Drawing gauge field" << std::endl;
+  LatticeGaugeFieldF Umu(UGrid);
+  SU<Nc>::HotConfiguration(RNG4,Umu);
+  std::cout << GridLogMessage << "Random gauge initialised " << std::endl;
+
+  RealD mass=0.1;
+  RealD M5  =1.8;
+
+  RealD NP = UGrid->_Nprocessors;
+  RealD NN = UGrid->NodeCount();
+
+  DomainWallFermionF Dw(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+
+  const int ncall = 500;
+  std::cout << GridLogMessage<< "*********************************************************" <<std::endl;
+  std::cout << GridLogMessage<< "* Benchmarking DomainWallFermionF::HaloGatherOpt         "<<std::endl;
+  std::cout << GridLogMessage<< "*********************************************************" <<std::endl;
+  {
+    typename DomainWallFermionF::Compressor compressor(0);
+    FGrid->Barrier();
+    Dw.Stencil.HaloExchangeOptGather(src,compressor);
+    double t0=usecond();
+    for(int i=0;i<ncall;i++){
+      Dw.Stencil.HaloExchangeOptGather(src,compressor);
+    }
+    double t1=usecond();
+    FGrid->Barrier();
+
+    double bytes=0.0;
+    if(mpi[0]) bytes+=latt4[1]*latt4[2]*latt4[3];
+    if(mpi[1]) bytes+=latt4[0]*latt4[2]*latt4[3];
+    if(mpi[2]) bytes+=latt4[0]*latt4[1]*latt4[3];
+    if(mpi[3]) bytes+=latt4[0]*latt4[1]*latt4[2];
+    bytes = bytes * Ls * 8.* (24.+12.)* 2.0;
+
+    std::cout<<GridLogMessage << "Gather us /call =   "<< (t1-t0)/ncall<<std::endl;
+    std::cout<<GridLogMessage << "Gather MBs /call =   "<< bytes*ncall/(t1-t0)<<std::endl;
+
+  }
+
+  Grid_finalize();
+  exit(0);
+}

From 0d5639f7075902182c59b9458a0e242f7ad3b7eb Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 4 Oct 2022 11:13:41 -0700
Subject: [PATCH 043/146] Run script update

---
 systems/PVC/benchmarks/run-2tile-mpi.sh | 11 +++++++++--
 systems/PVC/benchmarks/wrap.sh          |  4 ++--
 2 files changed, 11 insertions(+), 4 deletions(-)

diff --git a/systems/PVC/benchmarks/run-2tile-mpi.sh b/systems/PVC/benchmarks/run-2tile-mpi.sh
index 9db0b66b..34deac2c 100755
--- a/systems/PVC/benchmarks/run-2tile-mpi.sh
+++ b/systems/PVC/benchmarks/run-2tile-mpi.sh
@@ -1,6 +1,7 @@
 #!/bin/bash
 ##SBATCH -p PVC-SPR-QZEH 
 ##SBATCH -p PVC-ICX-QZNW
+
 #SBATCH -p QZ1J-ICX-PVC
 
 source /nfs/site/home/paboylex/ATS/GridNew/Grid/systems/PVC-nightly/setup.sh
@@ -19,8 +20,14 @@ export SYCL_DEVICE_FILTER=gpu,level_zero
 export I_MPI_OFFLOAD_CELL=tile
 export EnableImplicitScaling=0
 export EnableWalkerPartition=0
+#export SYCL_PI_LEVEL_ZERO_DEVICE_SCOPE_EVENTS=1
+export SYCL_PI_LEVEL_ZERO_USE_IMMEDIATE_COMMANDLISTS=1
+#export SYCL_PI_LEVEL_ZERO_USE_COPY_ENGINE=0
 
-mpiexec -launcher ssh -n 1 -host localhost  ./wrap.sh ./Benchmark_dwf_fp32 --mpi 1.1.1.1 --grid 32.32.32.32 --accelerator-threads $NT --comms-sequential --shm-mpi 1 > 1tile.log
+mpiexec -launcher ssh -n 2 -host localhost  ./wrap4gpu.sh ./Benchmark_dwf_fp32 --mpi 1.1.1.2 --grid 32.32.32.64 --accelerator-threads $NT  --shm-mpi 1 > dw.2tile.1x2.log
+mpiexec -launcher ssh -n 2 -host localhost  ./wrap4gpu.sh ./Benchmark_dwf_fp32 --mpi 2.1.1.1 --grid 64.32.32.32 --accelerator-threads $NT  --shm-mpi 1 > dw.2tile.2x1.log
+
+mpiexec -launcher ssh -n 2 -host localhost  ./wrap4gpu.sh ./Benchmark_halo --mpi 1.1.1.2 --grid 32.32.32.64 --accelerator-threads $NT  --shm-mpi 1 > halo.2tile.1x2.log
+mpiexec -launcher ssh -n 2 -host localhost  ./wrap4gpu.sh ./Benchmark_halo --mpi 2.1.1.1 --grid 64.32.32.32 --accelerator-threads $NT  --shm-mpi 1 > halo.2tile.2x1.log
 
-mpiexec -launcher ssh -n 2 -host localhost  ./wrap.sh ./Benchmark_dwf_fp32 --mpi 2.1.1.1 --grid 64.32.32.32 --accelerator-threads $NT --comms-sequential --shm-mpi 1 > 2tile.log
 
diff --git a/systems/PVC/benchmarks/wrap.sh b/systems/PVC/benchmarks/wrap.sh
index a352fff9..bb7b517d 100755
--- a/systems/PVC/benchmarks/wrap.sh
+++ b/systems/PVC/benchmarks/wrap.sh
@@ -7,8 +7,8 @@ echo Ranke $MPI_LOCALRANKID ZE_AFFINITY_MASK is $ZE_AFFINITY_MASK
 
 if [ $MPI_LOCALRANKID = "0" ] 
 then
-#  ~psteinbr/build_pti/ze_tracer -c $@
-  onetrace --chrome-kernel-timeline $@
+#  ~psteinbr/build_pti/ze_tracer -h $@
+  onetrace --chrome-device-timeline $@
 else
   $@
 fi

From 551a5f8dc8bb356b475f6f075902964429df93ba Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 11 Oct 2022 14:44:55 -0400
Subject: [PATCH 044/146] RRII gpu option

---
 Grid/communicator/Communicator_base.h         |   2 +-
 Grid/lattice/Lattice_ET.h                     |   2 +-
 Grid/lattice/Lattice_matrix_reduction.h       |   3 -
 Grid/lattice/Lattice_peekpoke.h               |  14 +-
 Grid/lattice/Lattice_reduction.h              |  12 +-
 Grid/lattice/Lattice_reduction_gpu.h          |   2 -
 Grid/lattice/Lattice_transfer.h               |   6 +-
 .../CayleyFermion5DImplementation.h           |  82 --
 Grid/qcd/utils/SUn.h                          |   1 -
 Grid/simd/Grid_a64fx-2.h                      |   4 +-
 Grid/simd/Grid_a64fx-fixedsize.h              |   8 +-
 Grid/simd/Grid_avx.h                          |   8 +-
 Grid/simd/Grid_avx512.h                       |   9 +-
 Grid/simd/Grid_gpu_rrii.h                     | 878 ++++++++++++++++++
 Grid/simd/Grid_gpu_vec.h                      |  10 +-
 Grid/simd/Grid_qpx.h                          |   4 +-
 Grid/simd/Grid_sse4.h                         |  10 +-
 Grid/simd/Grid_vector_types.h                 | 170 ++--
 Grid/simd/Simd.h                              |   2 +
 Grid/stencil/Stencil.h                        |   2 -
 Grid/tensors/Tensor_class.h                   |   3 +
 Grid/tensors/Tensor_extract_merge.h           |  49 +-
 configure.ac                                  |  84 +-
 systems/mac-arm/config-command-mpi            |   4 +-
 24 files changed, 1099 insertions(+), 270 deletions(-)
 create mode 100644 Grid/simd/Grid_gpu_rrii.h

diff --git a/Grid/communicator/Communicator_base.h b/Grid/communicator/Communicator_base.h
index cb3b9f0e..68cd36cc 100644
--- a/Grid/communicator/Communicator_base.h
+++ b/Grid/communicator/Communicator_base.h
@@ -131,7 +131,7 @@ public:
   template<class obj> void GlobalSum(obj &o){
     typedef typename obj::scalar_type scalar_type;
     int words = sizeof(obj)/sizeof(scalar_type);
-    scalar_type * ptr = (scalar_type *)& o;
+    scalar_type * ptr = (scalar_type *)& o; // Safe alias 
     GlobalSumVector(ptr,words);
   }
   
diff --git a/Grid/lattice/Lattice_ET.h b/Grid/lattice/Lattice_ET.h
index 4a8a7423..fdd31b28 100644
--- a/Grid/lattice/Lattice_ET.h
+++ b/Grid/lattice/Lattice_ET.h
@@ -63,7 +63,7 @@ accelerator_inline vobj predicatedWhere(const iobj &predicate,
   typename std::remove_const<vobj>::type ret;
 
   typedef typename vobj::scalar_object scalar_object;
-  typedef typename vobj::scalar_type scalar_type;
+  //  typedef typename vobj::scalar_type scalar_type;
   typedef typename vobj::vector_type vector_type;
 
   const int Nsimd = vobj::vector_type::Nsimd();
diff --git a/Grid/lattice/Lattice_matrix_reduction.h b/Grid/lattice/Lattice_matrix_reduction.h
index 7c470fef..abebbfd6 100644
--- a/Grid/lattice/Lattice_matrix_reduction.h
+++ b/Grid/lattice/Lattice_matrix_reduction.h
@@ -32,7 +32,6 @@ template<class vobj>
 static void sliceMaddMatrix (Lattice<vobj> &R,Eigen::MatrixXcd &aa,const Lattice<vobj> &X,const Lattice<vobj> &Y,int Orthog,RealD scale=1.0) 
 {    
   typedef typename vobj::scalar_object sobj;
-  typedef typename vobj::scalar_type scalar_type;
   typedef typename vobj::vector_type vector_type;
 
   int Nblock = X.Grid()->GlobalDimensions()[Orthog];
@@ -82,7 +81,6 @@ template<class vobj>
 static void sliceMulMatrix (Lattice<vobj> &R,Eigen::MatrixXcd &aa,const Lattice<vobj> &X,int Orthog,RealD scale=1.0) 
 {    
   typedef typename vobj::scalar_object sobj;
-  typedef typename vobj::scalar_type scalar_type;
   typedef typename vobj::vector_type vector_type;
 
   int Nblock = X.Grid()->GlobalDimensions()[Orthog];
@@ -130,7 +128,6 @@ template<class vobj>
 static void sliceInnerProductMatrix(  Eigen::MatrixXcd &mat, const Lattice<vobj> &lhs,const Lattice<vobj> &rhs,int Orthog) 
 {
   typedef typename vobj::scalar_object sobj;
-  typedef typename vobj::scalar_type scalar_type;
   typedef typename vobj::vector_type vector_type;
   
   GridBase *FullGrid  = lhs.Grid();
diff --git a/Grid/lattice/Lattice_peekpoke.h b/Grid/lattice/Lattice_peekpoke.h
index f3b485a4..b6a36b11 100644
--- a/Grid/lattice/Lattice_peekpoke.h
+++ b/Grid/lattice/Lattice_peekpoke.h
@@ -96,9 +96,6 @@ void pokeSite(const sobj &s,Lattice<vobj> &l,const Coordinate &site){
 
   GridBase *grid=l.Grid();
 
-  typedef typename vobj::scalar_type scalar_type;
-  typedef typename vobj::vector_type vector_type;
-
   int Nsimd = grid->Nsimd();
 
   assert( l.Checkerboard()== l.Grid()->CheckerBoard(site));
@@ -136,9 +133,6 @@ void peekSite(sobj &s,const Lattice<vobj> &l,const Coordinate &site){
         
   GridBase *grid=l.Grid();
 
-  typedef typename vobj::scalar_type scalar_type;
-  typedef typename vobj::vector_type vector_type;
-
   int Nsimd = grid->Nsimd();
 
   assert( l.Checkerboard() == l.Grid()->CheckerBoard(site));
@@ -179,11 +173,11 @@ inline void peekLocalSite(sobj &s,const LatticeView<vobj> &l,Coordinate &site)
   idx= grid->iIndex(site);
   odx= grid->oIndex(site);
   
-  scalar_type * vp = (scalar_type *)&l[odx];
+  const vector_type *vp = (const vector_type *) &l[odx];
   scalar_type * pt = (scalar_type *)&s;
       
   for(int w=0;w<words;w++){
-    pt[w] = vp[idx+w*Nsimd];
+    pt[w] = getlane(vp[w],idx);
   }
       
   return;
@@ -216,10 +210,10 @@ inline void pokeLocalSite(const sobj &s,LatticeView<vobj> &l,Coordinate &site)
   idx= grid->iIndex(site);
   odx= grid->oIndex(site);
 
-  scalar_type * vp = (scalar_type *)&l[odx];
+  vector_type * vp = (vector_type *)&l[odx];
   scalar_type * pt = (scalar_type *)&s;
   for(int w=0;w<words;w++){
-    vp[idx+w*Nsimd] = pt[w];
+    putlane(vp[w],pt[w],idx);
   }
   return;
 };
diff --git a/Grid/lattice/Lattice_reduction.h b/Grid/lattice/Lattice_reduction.h
index af6f8c00..87c3bf68 100644
--- a/Grid/lattice/Lattice_reduction.h
+++ b/Grid/lattice/Lattice_reduction.h
@@ -219,7 +219,6 @@ template<class vobj> inline RealD maxLocalNorm2(const Lattice<vobj> &arg)
 template<class vobj>
 inline ComplexD rankInnerProduct(const Lattice<vobj> &left,const Lattice<vobj> &right)
 {
-  typedef typename vobj::scalar_type scalar_type;
   typedef typename vobj::vector_typeD vector_type;
   ComplexD  nrm;
   
@@ -296,7 +295,6 @@ axpby_norm_fast(Lattice<vobj> &z,sobj a,sobj b,const Lattice<vobj> &x,const Latt
   conformable(z,x);
   conformable(x,y);
 
-  typedef typename vobj::scalar_type scalar_type;
   //  typedef typename vobj::vector_typeD vector_type;
   RealD  nrm;
   
@@ -341,7 +339,6 @@ innerProductNorm(ComplexD& ip, RealD &nrm, const Lattice<vobj> &left,const Latti
 {
   conformable(left,right);
 
-  typedef typename vobj::scalar_type scalar_type;
   typedef typename vobj::vector_typeD vector_type;
   Vector<ComplexD> tmp(2);
 
@@ -597,7 +594,8 @@ static void sliceNorm (std::vector<RealD> &sn,const Lattice<vobj> &rhs,int Ortho
 template<class vobj>
 static void sliceMaddVector(Lattice<vobj> &R,std::vector<RealD> &a,const Lattice<vobj> &X,const Lattice<vobj> &Y,
 			    int orthogdim,RealD scale=1.0) 
-{    
+{
+  // perhaps easier to just promote A to a field and use regular madd
   typedef typename vobj::scalar_object sobj;
   typedef typename vobj::scalar_type scalar_type;
   typedef typename vobj::vector_type vector_type;
@@ -628,8 +626,7 @@ static void sliceMaddVector(Lattice<vobj> &R,std::vector<RealD> &a,const Lattice
     for(int l=0;l<Nsimd;l++){
       grid->iCoorFromIindex(icoor,l);
       int ldx =r+icoor[orthogdim]*rd;
-      scalar_type *as =(scalar_type *)&av;
-      as[l] = scalar_type(a[ldx])*zscale;
+      av.putlane(scalar_type(a[ldx])*zscale,l);
     }
 
     tensor_reduced at; at=av;
@@ -669,7 +666,6 @@ template<class vobj>
 static void sliceMaddMatrix (Lattice<vobj> &R,Eigen::MatrixXcd &aa,const Lattice<vobj> &X,const Lattice<vobj> &Y,int Orthog,RealD scale=1.0) 
 {    
   typedef typename vobj::scalar_object sobj;
-  typedef typename vobj::scalar_type scalar_type;
   typedef typename vobj::vector_type vector_type;
 
   int Nblock = X.Grid()->GlobalDimensions()[Orthog];
@@ -723,7 +719,6 @@ template<class vobj>
 static void sliceMulMatrix (Lattice<vobj> &R,Eigen::MatrixXcd &aa,const Lattice<vobj> &X,int Orthog,RealD scale=1.0) 
 {    
   typedef typename vobj::scalar_object sobj;
-  typedef typename vobj::scalar_type scalar_type;
   typedef typename vobj::vector_type vector_type;
 
   int Nblock = X.Grid()->GlobalDimensions()[Orthog];
@@ -777,7 +772,6 @@ template<class vobj>
 static void sliceInnerProductMatrix(  Eigen::MatrixXcd &mat, const Lattice<vobj> &lhs,const Lattice<vobj> &rhs,int Orthog) 
 {
   typedef typename vobj::scalar_object sobj;
-  typedef typename vobj::scalar_type scalar_type;
   typedef typename vobj::vector_type vector_type;
   
   GridBase *FullGrid  = lhs.Grid();
diff --git a/Grid/lattice/Lattice_reduction_gpu.h b/Grid/lattice/Lattice_reduction_gpu.h
index bad86d2a..5f5c6cc0 100644
--- a/Grid/lattice/Lattice_reduction_gpu.h
+++ b/Grid/lattice/Lattice_reduction_gpu.h
@@ -250,8 +250,6 @@ inline typename vobj::scalar_objectD sumD_gpu_large(const vobj *lat, Integer osi
 template <class vobj>
 inline typename vobj::scalar_objectD sumD_gpu(const vobj *lat, Integer osites)
 {
-  typedef typename vobj::vector_type  vector;
-  typedef typename vobj::scalar_typeD scalarD;
   typedef typename vobj::scalar_objectD sobj;
   sobj ret;
   
diff --git a/Grid/lattice/Lattice_transfer.h b/Grid/lattice/Lattice_transfer.h
index aee55e93..d9b7a704 100644
--- a/Grid/lattice/Lattice_transfer.h
+++ b/Grid/lattice/Lattice_transfer.h
@@ -677,10 +677,10 @@ void localCopyRegion(const Lattice<vobj> &From,Lattice<vobj> & To,Coordinate Fro
       Integer idx_t = 0; for(int d=0;d<nd;d++) idx_t+=ist[d]*(Tcoor[d]/rdt[d]);
       Integer odx_f = 0; for(int d=0;d<nd;d++) odx_f+=osf[d]*(Fcoor[d]%rdf[d]);
       Integer odx_t = 0; for(int d=0;d<nd;d++) odx_t+=ost[d]*(Tcoor[d]%rdt[d]);
-      scalar_type * fp = (scalar_type *)&f_v[odx_f];
-      scalar_type * tp = (scalar_type *)&t_v[odx_t];
+      vector_type * fp = (vector_type *)&f_v[odx_f];
+      vector_type * tp = (vector_type *)&t_v[odx_t];
       for(int w=0;w<words;w++){
-	tp[idx_t+w*Nsimd] = fp[idx_f+w*Nsimd];  // FIXME IF RRII layout, type pun no worke
+	tp[w].putlane(fp[w].getlane(idx_f),idx_t);
       }
     }
   });
diff --git a/Grid/qcd/action/fermion/implementation/CayleyFermion5DImplementation.h b/Grid/qcd/action/fermion/implementation/CayleyFermion5DImplementation.h
index 208e5f81..2b8a3a18 100644
--- a/Grid/qcd/action/fermion/implementation/CayleyFermion5DImplementation.h
+++ b/Grid/qcd/action/fermion/implementation/CayleyFermion5DImplementation.h
@@ -905,88 +905,6 @@ void CayleyFermion5D<Impl>::SeqConservedCurrent(PropagatorField &q_in,
 #undef TopRowWithSource
 
 
-
-#if 0
-template<class Impl>
-void CayleyFermion5D<Impl>::MooeeInternalCompute(int dag, int inv,
-						 Vector<iSinglet<Simd> > & Matp,
-						 Vector<iSinglet<Simd> > & Matm)
-{
-  int Ls=this->Ls;
-
-  GridBase *grid = this->FermionRedBlackGrid();
-  int LLs = grid->_rdimensions[0];
-
-  if ( LLs == Ls ) {
-    return; // Not vectorised in 5th direction
-  }
-
-  Eigen::MatrixXcd Pplus  = Eigen::MatrixXcd::Zero(Ls,Ls);
-  Eigen::MatrixXcd Pminus = Eigen::MatrixXcd::Zero(Ls,Ls);
-  
-  for(int s=0;s<Ls;s++){
-    Pplus(s,s) = bee[s];
-    Pminus(s,s)= bee[s];
-  }
-  
-  for(int s=0;s<Ls-1;s++){
-    Pminus(s,s+1) = -cee[s];
-  }
-  
-  for(int s=0;s<Ls-1;s++){
-    Pplus(s+1,s) = -cee[s+1];
-  }
-  Pplus (0,Ls-1) = mass*cee[0];
-  Pminus(Ls-1,0) = mass*cee[Ls-1];
-  
-  Eigen::MatrixXcd PplusMat ;
-  Eigen::MatrixXcd PminusMat;
-  
-  if ( inv ) {
-    PplusMat =Pplus.inverse();
-    PminusMat=Pminus.inverse();
-  } else { 
-    PplusMat =Pplus;
-    PminusMat=Pminus;
-  }
-  
-  if(dag){
-    PplusMat.adjointInPlace();
-    PminusMat.adjointInPlace();
-  }
-  
-  typedef typename SiteHalfSpinor::scalar_type scalar_type;
-  const int Nsimd=Simd::Nsimd();
-  Matp.resize(Ls*LLs);
-  Matm.resize(Ls*LLs);
-
-  for(int s2=0;s2<Ls;s2++){
-    for(int s1=0;s1<LLs;s1++){
-      int istride = LLs;
-      int ostride = 1;
-      Simd Vp;
-      Simd Vm;
-      scalar_type *sp = (scalar_type *)&Vp;
-      scalar_type *sm = (scalar_type *)&Vm;
-      for(int l=0;l<Nsimd;l++){
-	if ( switcheroo<Coeff_t>::iscomplex() ) {
-	  sp[l] = PplusMat (l*istride+s1*ostride,s2);
-	  sm[l] = PminusMat(l*istride+s1*ostride,s2);
-	} else { 
-	  // if real
-	  scalar_type tmp;
-	  tmp = PplusMat (l*istride+s1*ostride,s2);
-	  sp[l] = scalar_type(tmp.real(),tmp.real());
-	  tmp = PminusMat(l*istride+s1*ostride,s2);
-	  sm[l] = scalar_type(tmp.real(),tmp.real());
-	}
-      }
-      Matp[LLs*s2+s1] = Vp;
-      Matm[LLs*s2+s1] = Vm;
-    }}
-}
-#endif
-
 NAMESPACE_END(Grid);
 
 
diff --git a/Grid/qcd/utils/SUn.h b/Grid/qcd/utils/SUn.h
index b9660c65..23eceea3 100644
--- a/Grid/qcd/utils/SUn.h
+++ b/Grid/qcd/utils/SUn.h
@@ -615,7 +615,6 @@ public:
     GridBase *grid = out.Grid();
 
     typedef typename LatticeMatrixType::vector_type vector_type;
-    typedef typename LatticeMatrixType::scalar_type scalar_type;
 
     typedef iSinglet<vector_type> vTComplexType;
 
diff --git a/Grid/simd/Grid_a64fx-2.h b/Grid/simd/Grid_a64fx-2.h
index 2ad8591c..a4ef70ae 100644
--- a/Grid/simd/Grid_a64fx-2.h
+++ b/Grid/simd/Grid_a64fx-2.h
@@ -501,7 +501,7 @@ struct Conj{
 struct TimesMinusI{
   // Complex
   template <typename T>
-  inline vec<T> operator()(vec<T> a, vec<T> b){
+  inline vec<T> operator()(vec<T> a){
     vec<T> out;
     const vec<typename acle<T>::uint> tbl_swap = acle<T>::tbl_swap();
     svbool_t pg1 = acle<T>::pg1();
@@ -520,7 +520,7 @@ struct TimesMinusI{
 struct TimesI{
   // Complex
   template <typename T>
-  inline vec<T> operator()(vec<T> a, vec<T> b){
+  inline vec<T> operator()(vec<T> a){
     vec<T> out;
     const vec<typename acle<T>::uint> tbl_swap = acle<T>::tbl_swap();
     svbool_t pg1 = acle<T>::pg1();
diff --git a/Grid/simd/Grid_a64fx-fixedsize.h b/Grid/simd/Grid_a64fx-fixedsize.h
index 6b450012..5bf1b0a3 100644
--- a/Grid/simd/Grid_a64fx-fixedsize.h
+++ b/Grid/simd/Grid_a64fx-fixedsize.h
@@ -418,7 +418,7 @@ struct Conj{
 
 struct TimesMinusI{
   // Complex float
-  inline vecf operator()(vecf a, vecf b){
+  inline vecf operator()(vecf a){
     lutf tbl_swap = acle<float>::tbl_swap();
     pred pg1 = acle<float>::pg1();
     pred pg_odd = acle<float>::pg_odd();
@@ -428,7 +428,7 @@ struct TimesMinusI{
     return svneg_m(a_v, pg_odd, a_v);
   }
   // Complex double
-  inline vecd operator()(vecd a, vecd b){
+  inline vecd operator()(vecd a){
     lutd tbl_swap = acle<double>::tbl_swap();
     pred pg1 = acle<double>::pg1();
     pred pg_odd = acle<double>::pg_odd();
@@ -441,7 +441,7 @@ struct TimesMinusI{
 
 struct TimesI{
   // Complex float
-  inline vecf operator()(vecf a, vecf b){
+  inline vecf operator()(vecf a){
     lutf tbl_swap = acle<float>::tbl_swap();
     pred pg1 = acle<float>::pg1();
     pred pg_even = acle<float>::pg_even();
@@ -451,7 +451,7 @@ struct TimesI{
     return svneg_m(a_v, pg_even, a_v);
   }
   // Complex double
-  inline vecd operator()(vecd a, vecd b){
+  inline vecd operator()(vecd a){
     lutd tbl_swap = acle<double>::tbl_swap();
     pred pg1 = acle<double>::pg1();
     pred pg_even = acle<double>::pg_even();
diff --git a/Grid/simd/Grid_avx.h b/Grid/simd/Grid_avx.h
index ad9800fb..f8962714 100644
--- a/Grid/simd/Grid_avx.h
+++ b/Grid/simd/Grid_avx.h
@@ -405,12 +405,12 @@ struct Conj{
 
 struct TimesMinusI{
   //Complex single
-  inline __m256 operator()(__m256 in, __m256 ret){
+  inline __m256 operator()(__m256 in){
     __m256 tmp =_mm256_addsub_ps(_mm256_setzero_ps(),in);   // r,-i
     return _mm256_shuffle_ps(tmp,tmp,_MM_SELECT_FOUR_FOUR(2,3,0,1)); //-i,r
   }
   //Complex double
-  inline __m256d operator()(__m256d in, __m256d ret){
+  inline __m256d operator()(__m256d in){
     __m256d tmp = _mm256_addsub_pd(_mm256_setzero_pd(),in); // r,-i
     return _mm256_shuffle_pd(tmp,tmp,0x5);
   }
@@ -418,12 +418,12 @@ struct TimesMinusI{
 
 struct TimesI{
   //Complex single
-  inline __m256 operator()(__m256 in, __m256 ret){
+  inline __m256 operator()(__m256 in){
     __m256 tmp =_mm256_shuffle_ps(in,in,_MM_SELECT_FOUR_FOUR(2,3,0,1)); // i,r
     return _mm256_addsub_ps(_mm256_setzero_ps(),tmp);          // i,-r
   }
   //Complex double
-  inline __m256d operator()(__m256d in, __m256d ret){
+  inline __m256d operator()(__m256d in){
     __m256d tmp = _mm256_shuffle_pd(in,in,0x5);
     return _mm256_addsub_pd(_mm256_setzero_pd(),tmp); // i,-r
   }
diff --git a/Grid/simd/Grid_avx512.h b/Grid/simd/Grid_avx512.h
index 839d4554..95b96143 100644
--- a/Grid/simd/Grid_avx512.h
+++ b/Grid/simd/Grid_avx512.h
@@ -271,14 +271,14 @@ struct Conj{
 
 struct TimesMinusI{
   //Complex single
-  inline __m512 operator()(__m512 in, __m512 ret){
+  inline __m512 operator()(__m512 in){
     //__m512 tmp = _mm512_mask_sub_ps(in,0xaaaa,_mm512_setzero_ps(),in); // real -imag 
     //return _mm512_shuffle_ps(tmp,tmp,_MM_SELECT_FOUR_FOUR(2,3,1,0));   // 0x4E??
     __m512 tmp = _mm512_shuffle_ps(in,in,_MM_SELECT_FOUR_FOUR(2,3,0,1));
     return _mm512_mask_sub_ps(tmp,0xaaaa,_mm512_setzero_ps(),tmp);
   }
   //Complex double
-  inline __m512d operator()(__m512d in, __m512d ret){
+  inline __m512d operator()(__m512d in){
     //__m512d tmp = _mm512_mask_sub_pd(in,0xaa,_mm512_setzero_pd(),in); // real -imag 
     //return _mm512_shuffle_pd(tmp,tmp,0x55);
     __m512d tmp = _mm512_shuffle_pd(in,in,0x55);
@@ -288,17 +288,16 @@ struct TimesMinusI{
 
 struct TimesI{
   //Complex single
-  inline __m512 operator()(__m512 in, __m512 ret){
+  inline __m512 operator()(__m512 in){
     __m512 tmp = _mm512_shuffle_ps(in,in,_MM_SELECT_FOUR_FOUR(2,3,0,1));
     return _mm512_mask_sub_ps(tmp,0x5555,_mm512_setzero_ps(),tmp); 
   }
   //Complex double
-  inline __m512d operator()(__m512d in, __m512d ret){
+  inline __m512d operator()(__m512d in){
     __m512d tmp = _mm512_shuffle_pd(in,in,0x55);
     return _mm512_mask_sub_pd(tmp,0x55,_mm512_setzero_pd(),tmp); 
   }
 
-
 };
   
 // Gpermute utilities consider coalescing into 1 Gpermute
diff --git a/Grid/simd/Grid_gpu_rrii.h b/Grid/simd/Grid_gpu_rrii.h
new file mode 100644
index 00000000..36f343e4
--- /dev/null
+++ b/Grid/simd/Grid_gpu_rrii.h
@@ -0,0 +1,878 @@
+    /*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/simd/Grid_gpu.h
+
+    Copyright (C) 2021
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+    *************************************************************************************/
+    /*  END LEGAL */
+//----------------------------------------------------------------------
+/*! @file Grid_gpu_rrii.h*/
+//----------------------------------------------------------------------
+
+//////////////////////////////
+// fp16
+//////////////////////////////
+#ifdef GRID_CUDA
+#include <cuda_fp16.h>
+#endif
+#ifdef GRID_HIP
+#include <hip/hip_fp16.h>
+#endif
+#if !defined(GRID_HIP) && !defined(GRID_CUDA) 
+namespace Grid {
+  typedef struct { uint16_t x;} half;
+}
+#endif
+namespace Grid {
+  accelerator_inline float half2float(half h)
+  {
+    float f;
+#if defined(GRID_CUDA) || defined(GRID_HIP)
+    f = __half2float(h);
+#else 
+    Grid_half hh; 
+    hh.x = h.x;
+    f=  sfw_half_to_float(hh);
+#endif
+    return f;
+  }
+  accelerator_inline half float2half(float f)
+  {
+    half h;
+#if defined(GRID_CUDA) || defined(GRID_HIP)
+    h = __float2half(f);
+#else
+    Grid_half hh = sfw_float_to_half(f);
+    h.x = hh.x;
+#endif
+    return h;
+  }
+}
+
+
+#define COALESCE_GRANULARITY ( GEN_SIMD_WIDTH )
+
+namespace Grid {
+
+////////////////////////////////////////////////////////////////////////
+// Real vector
+////////////////////////////////////////////////////////////////////////  
+template<int _N, class _datum>
+struct GpuVector {
+  _datum rrrr[_N];
+  static const int N = _N;
+  typedef _datum datum;
+};
+template<int N,class datum>
+inline accelerator GpuVector<N,datum> operator*(const GpuVector<N,datum> l,const GpuVector<N,datum> r) {
+  GpuVector<N,datum> ret;
+  for(int i=0;i<N;i++) { 
+    ret.rrrr[i] = l.rrrr[i]*r.rrrr[i];
+  }
+  return ret;
+}
+template<int N,class datum>
+inline accelerator GpuVector<N,datum> operator-(const GpuVector<N,datum> l,const GpuVector<N,datum> r) {
+  GpuVector<N,datum> ret;
+  for(int i=0;i<N;i++) { 
+    ret.rrrr[i] = l.rrrr[i]-r.rrrr[i];
+  }
+  return ret;
+}
+template<int N,class datum>
+inline accelerator GpuVector<N,datum> operator+(const GpuVector<N,datum> l,const GpuVector<N,datum> r) {
+  GpuVector<N,datum> ret;
+  for(int i=0;i<N;i++) { 
+    ret.rrrr[i] = l.rrrr[i]+r.rrrr[i];
+  }
+  return ret;
+}
+template<int N,class datum>
+inline accelerator GpuVector<N,datum> operator/(const GpuVector<N,datum> l,const GpuVector<N,datum> r) {
+  GpuVector<N,datum> ret;
+  for(int i=0;i<N;i++) { 
+    ret.rrrr[i] = l.rrrr[i]/r.rrrr[i];
+  }
+  return ret;
+}
+
+////////////////////////////////////////////////////////////////////////
+// Complex vector
+////////////////////////////////////////////////////////////////////////  
+template<int _N, class _datum>
+struct GpuComplexVector {
+  _datum rrrr[_N];
+  _datum iiii[_N];
+  static const int N = _N;
+  typedef _datum datum;
+};
+template<int N,class datum>
+inline accelerator GpuComplexVector<N,datum> operator*(const GpuComplexVector<N,datum> l,const GpuComplexVector<N,datum> r) {
+  GpuComplexVector<N,datum> ret;
+  for(int i=0;i<N;i++) { 
+    ret.rrrr[i] = l.rrrr[i]*r.rrrr[i] - l.iiii[i]*r.iiii[i];
+    ret.iiii[i] = l.rrrr[i]*r.iiii[i] + l.iiii[i]*r.rrrr[i];
+  }
+  return ret;
+}
+template<int N,class datum>
+inline accelerator GpuComplexVector<N,datum> operator-(const GpuComplexVector<N,datum> l,const GpuComplexVector<N,datum> r) {
+  GpuComplexVector<N,datum> ret;
+  for(int i=0;i<N;i++) { 
+    ret.rrrr[i] = l.rrrr[i]-r.rrrr[i];
+    ret.iiii[i] = l.iiii[i]-r.iiii[i];
+  }
+  return ret;
+}
+template<int N,class datum>
+inline accelerator GpuComplexVector<N,datum> operator+(const GpuComplexVector<N,datum> l,const GpuComplexVector<N,datum> r) {
+  GpuComplexVector<N,datum> ret;
+  for(int i=0;i<N;i++) { 
+    ret.rrrr[i] = l.rrrr[i]+r.rrrr[i];
+    ret.iiii[i] = l.iiii[i]+r.iiii[i];
+  }
+  return ret;
+}
+template<int N,class datum>
+inline accelerator GpuComplexVector<N,datum> operator/(const GpuComplexVector<N,datum> l,const GpuComplexVector<N,datum> r) {
+  GpuComplexVector<N,datum> ret;
+  for(int i=0;i<N;i++) { 
+    ret.rrrr[i] = l.rrrr[i]/r.rrrr[i];
+    ret.iiii[i] = l.iiii[i]/r.iiii[i];
+  }
+  return ret;
+}
+
+////////////////////////////////
+// SIMD counts
+////////////////////////////////
+
+constexpr int NSIMD_RealH    = COALESCE_GRANULARITY / sizeof(half);
+constexpr int NSIMD_ComplexH = COALESCE_GRANULARITY / sizeof(half);
+constexpr int NSIMD_RealF    = COALESCE_GRANULARITY / sizeof(float);
+constexpr int NSIMD_ComplexF = COALESCE_GRANULARITY / sizeof(float);
+constexpr int NSIMD_RealD    = COALESCE_GRANULARITY / sizeof(double);
+constexpr int NSIMD_ComplexD = COALESCE_GRANULARITY / sizeof(double);
+constexpr int NSIMD_Integer  = COALESCE_GRANULARITY / sizeof(Integer);
+
+typedef GpuVector<NSIMD_RealH   , half        > GpuVectorRH;
+typedef GpuComplexVector<NSIMD_ComplexH, half > GpuVectorCH;
+typedef GpuVector<NSIMD_RealF,    float       > GpuVectorRF;
+typedef GpuComplexVector<NSIMD_ComplexF, float> GpuVectorCF;
+typedef GpuVector<NSIMD_RealD,    double      > GpuVectorRD;
+typedef GpuComplexVector<NSIMD_ComplexD,double> GpuVectorCD;
+typedef GpuVector<NSIMD_Integer,  Integer     > GpuVectorI;
+
+namespace Optimization {
+
+  struct Vsplat{
+    //Complex float
+    accelerator_inline GpuVectorCF operator()(float a, float b){
+      GpuVectorCF ret;
+      for(int i=0;i<GpuVectorCF::N;i++){
+	ret.rrrr[i] = typename GpuVectorCF::datum(a);
+	ret.iiii[i] = typename GpuVectorCF::datum(b);
+      }
+      return ret;
+    }
+    // Real float
+    accelerator_inline GpuVectorRF operator()(float a){
+      GpuVectorRF ret;
+      for(int i=0;i<GpuVectorRF::N;i++){
+	ret.rrrr[i] = typename GpuVectorRF::datum(a);
+      }
+      return ret;
+    }
+    //Complex double
+    accelerator_inline GpuVectorCD operator()(double a, double b){
+      GpuVectorCD ret;
+      for(int i=0;i<GpuVectorCD::N;i++){
+	ret.rrrr[i] = typename GpuVectorCD::datum(a);
+	ret.iiii[i] = typename GpuVectorCD::datum(b);
+      }
+      return ret;
+    }
+    //Real double
+    accelerator_inline GpuVectorRD operator()(double a){
+      GpuVectorRD ret; 
+      for(int i=0;i<GpuVectorRD::N;i++){
+	ret.rrrr[i] = typename GpuVectorRD::datum(a);
+      }
+      return ret;
+    }
+    //Integer
+    accelerator_inline GpuVectorI operator()(Integer a){
+      GpuVectorI ret;
+      for(int i=0;i<GpuVectorI::N;i++){
+	ret.rrrr[i] = typename GpuVectorI::datum(a);
+      }
+      return ret;
+    }
+  };
+
+  struct Vstore{
+    template<int N,class datum,class P>
+    accelerator_inline void operator()(GpuVector<N,datum> a, P* Fp){
+      GpuVector<N,datum> *vF = (GpuVector<N,datum> *)Fp;
+      *vF = a;
+    }
+    template<int N,class datum,class P>
+    accelerator_inline void operator()(GpuComplexVector<N,datum> a, P* Fp){
+      GpuComplexVector<N,datum> *vF = (GpuComplexVector<N,datum> *)Fp;
+      *vF = a;
+    }
+  };
+
+  struct Vstream{
+    template<int N,class datum, class P>
+    accelerator_inline void operator()(P* F,GpuVector<N,datum> a){
+      GpuVector<N,datum> *vF = (GpuVector<N,datum> *)F;
+      *vF = a;
+    }
+    template<int N,class datum, class P>
+    accelerator_inline void operator()(P* F,GpuComplexVector<N,datum> a){
+      GpuComplexVector<N,datum> *vF = (GpuComplexVector<N,datum> *)F;
+      *vF = a;
+    }
+  };
+
+  struct Vset{
+    // Complex float 
+    accelerator_inline GpuVectorCF operator()(Grid::ComplexF *a){
+      typedef GpuVectorCF vec;
+      vec ret;
+      for(int i=0;i<vec::N;i++){
+	ret.rrrr[i] = vec::datum(a[i].real());
+	ret.iiii[i] = vec::datum(a[i].imag());
+      }
+      return ret;
+    }
+    // Complex double 
+    accelerator_inline GpuVectorCD operator()(Grid::ComplexD *a){
+      typedef GpuVectorCD vec;
+      vec ret;
+      for(int i=0;i<vec::N;i++){
+	ret.rrrr[i] = vec::datum(a[i].real());
+	ret.iiii[i] = vec::datum(a[i].imag());
+      }
+      return ret;
+    }
+    // Real float 
+    accelerator_inline GpuVectorRF operator()(float *a){
+      typedef GpuVectorRF vec;
+      vec ret;
+      for(int i=0;i<vec::N;i++){
+	ret.rrrr[i] = vec::datum(a[i]);
+      }
+      return ret;
+    }
+    // Real double
+    accelerator_inline GpuVectorRD operator()(double *a){
+      typedef GpuVectorRD vec;
+      vec ret;
+      for(int i=0;i<vec::N;i++){
+	ret.rrrr[i] = vec::datum(a[i]);
+      }
+      return ret;
+    }
+    // Integer
+    accelerator_inline GpuVectorI operator()(Integer *a){
+      typedef GpuVectorI vec;
+      vec ret;
+      for(int i=0;i<vec::N;i++){
+	ret.rrrr[i] = vec::datum(a[i]);
+      }
+      return ret;
+    }
+  };
+
+  template <typename Out_type, typename In_type>
+  struct Reduce{
+    //Need templated class to overload output type
+    //General form must generate error if compiled
+    accelerator_inline Out_type operator()(In_type in){
+      printf("Error, using wrong Reduce function\n");
+      exit(1);
+      return 0;
+    }
+  };
+
+  /////////////////////////////////////////////////////
+  // Arithmetic operations
+  /////////////////////////////////////////////////////
+  struct Sum{
+    //Real float
+    accelerator_inline GpuVectorRF operator()(GpuVectorRF a,GpuVectorRF b){
+      return a+b;
+    }
+    accelerator_inline GpuVectorRD operator()(GpuVectorRD a,GpuVectorRD b){
+      return a+b;
+    }
+    accelerator_inline GpuVectorCF operator()(GpuVectorCF a,GpuVectorCF b){
+      return a+b;
+    }
+    accelerator_inline GpuVectorCD operator()(GpuVectorCD a,GpuVectorCD b){
+      return a+b;
+    }
+    accelerator_inline GpuVectorI operator()(GpuVectorI a,GpuVectorI b){
+      return a+b;
+    }
+  };
+
+  struct Sub{
+    accelerator_inline GpuVectorRF operator()(GpuVectorRF a,GpuVectorRF b){
+      return a-b;
+    }
+    accelerator_inline GpuVectorRD operator()(GpuVectorRD a,GpuVectorRD b){
+      return a-b;
+    }
+    accelerator_inline GpuVectorCF operator()(GpuVectorCF a,GpuVectorCF b){
+      return a-b;
+    }
+    accelerator_inline GpuVectorCD operator()(GpuVectorCD a,GpuVectorCD b){
+      return a-b;
+    }
+    accelerator_inline GpuVectorI operator()(GpuVectorI a,GpuVectorI b){
+      return a-b;
+    }
+  };
+
+  struct MultRealPart{
+    accelerator_inline GpuVectorCF operator()(GpuVectorCF a,GpuVectorCF b){
+      typedef GpuVectorCF vec;
+      vec ret;
+      for(int i=0;i<vec::N;i++){
+	ret.rrrr[i] = a.rrrr[i]*b.rrrr[i];
+	ret.iiii[i] = a.rrrr[i]*b.iiii[i];
+      }
+      return ret;
+    }
+    accelerator_inline GpuVectorCD operator()(GpuVectorCD a,GpuVectorCD b){
+      typedef GpuVectorCD vec;
+      vec ret;
+      for(int i=0;i<vec::N;i++){
+	ret.rrrr[i] = a.rrrr[i]*b.rrrr[i];
+	ret.iiii[i] = a.rrrr[i]*b.iiii[i];
+      }
+      return ret;
+    }
+  };
+
+  struct MaddRealPart{
+    accelerator_inline GpuVectorCF operator()(GpuVectorCF a,GpuVectorCF b,GpuVectorCF c){
+      typedef GpuVectorCF vec;
+      vec ret;
+      for(int i=0;i<vec::N;i++){
+	ret.rrrr[i] = a.rrrr[i]*b.rrrr[i]+c.rrrr[i];
+	ret.iiii[i] = a.rrrr[i]*b.iiii[i]+c.iiii[i];
+      }
+      return ret;
+    }
+    accelerator_inline GpuVectorCD operator()(GpuVectorCD a,GpuVectorCD b,GpuVectorCD c){
+      typedef GpuVectorCD vec;
+      vec ret;
+      for(int i=0;i<vec::N;i++){
+	ret.rrrr[i] = a.rrrr[i]*b.rrrr[i]+c.rrrr[i];
+	ret.iiii[i] = a.rrrr[i]*b.iiii[i]+c.iiii[i];
+      }
+      return ret;
+    }
+  };
+
+  struct MultComplex{
+
+    accelerator_inline GpuVectorCF operator()(GpuVectorCF a,GpuVectorCF b){
+      return a*b;
+    }
+    accelerator_inline GpuVectorCD operator()(GpuVectorCD a,GpuVectorCD b){
+      return a*b;
+    }
+  };
+
+  struct Mult{
+    accelerator_inline void mac(GpuVectorRF &a, GpuVectorRF b, GpuVectorRF c){
+      a= a+b*c;
+    }
+    accelerator_inline void mac(GpuVectorRD &a, GpuVectorRD b, GpuVectorRD c){
+      a= a+b*c;
+    }
+    // Real float
+    accelerator_inline GpuVectorRF operator()(GpuVectorRF a, GpuVectorRF b){
+      return a*b;
+    }
+    // Real double
+    accelerator_inline GpuVectorRD operator()(GpuVectorRD a, GpuVectorRD b){
+      return a*b;
+    }
+    accelerator_inline GpuVectorI operator()(GpuVectorI a, GpuVectorI b){
+      return a*b;
+    }
+  };
+
+  struct Div{
+    // Real float
+    accelerator_inline GpuVectorRF operator()(GpuVectorRF a, GpuVectorRF b){
+      return a/b;
+    }
+    accelerator_inline GpuVectorRD operator()(GpuVectorRD a, GpuVectorRD b){
+      return a/b;
+    }
+    accelerator_inline GpuVectorI operator()(GpuVectorI a, GpuVectorI b){
+      return a/b;
+    }
+
+    // Danger -- element wise divide fro complex, not complex div. 
+    // See Grid_vector_types.h lines around 735, applied after "toReal"
+    accelerator_inline GpuVectorCF operator()(GpuVectorCF a, GpuVectorCF b){
+      return a/b;
+    }
+    accelerator_inline GpuVectorCD operator()(GpuVectorCD a, GpuVectorCD b){
+      return a/b;
+    }
+  };
+
+
+  struct Conj{
+    // Complex single
+    accelerator_inline GpuVectorCF operator()(GpuVectorCF in){
+      typedef GpuVectorCF vec;
+      vec ret;
+      for(int i=0;i<vec::N;i++){
+	ret.rrrr[i] = in.rrrr[i];
+	ret.iiii[i] =-in.iiii[i];
+      }
+      return ret;
+    }
+    accelerator_inline GpuVectorCD operator()(GpuVectorCD in){
+      typedef GpuVectorCD vec;
+      vec ret;
+      for(int i=0;i<vec::N;i++){
+	ret.rrrr[i] = in.rrrr[i];
+	ret.iiii[i] =-in.iiii[i];
+      }
+      return ret;
+    }
+  };
+
+  struct TimesMinusI{
+    //Complex single
+    accelerator_inline GpuVectorCF operator()(GpuVectorCF in){
+      typedef GpuVectorCF vec;
+      vec ret;
+      for(int i=0;i<vec::N;i++){
+	ret.rrrr[i] = in.iiii[i];
+	ret.iiii[i] =-in.rrrr[i];
+      }
+      return ret;
+    }
+    accelerator_inline GpuVectorCD operator()(GpuVectorCD in){
+      typedef GpuVectorCD vec;
+      vec ret;
+      for(int i=0;i<vec::N;i++){
+	ret.rrrr[i] = in.iiii[i];
+	ret.iiii[i] =-in.rrrr[i];
+      }
+      return ret;
+    }
+  };
+
+  struct TimesI{
+    //Complex single
+    accelerator_inline GpuVectorCF operator()(GpuVectorCF in){
+      typedef GpuVectorCF vec;
+      vec ret;
+      for(int i=0;i<vec::N;i++){
+	ret.rrrr[i] =-in.iiii[i];
+	ret.iiii[i] = in.rrrr[i];
+      }
+      return ret;
+    }
+    accelerator_inline GpuVectorCD operator()(GpuVectorCD in){
+      typedef GpuVectorCD vec;
+      vec ret;
+      for(int i=0;i<vec::N;i++){
+	ret.rrrr[i] =-in.iiii[i];
+	ret.iiii[i] = in.rrrr[i];
+      }
+      return ret;
+    }
+  };
+
+  struct Permute{
+
+    template <int n,int _N, class _datum >
+    static accelerator_inline GpuVector<_N,_datum> PermuteN(GpuVector<_N,_datum> &in) {   
+      typedef GpuVector<_N,_datum> vec;
+      vec out;					
+      unsigned int _mask = vec::N >> (n + 1);	
+      for(int i=0;i<vec::N;i++) {
+	out.rrrr[i] = in.rrrr[i^_mask];
+      }
+      return out;	
+    }
+    template <int n,int _N, class _datum >
+    static accelerator_inline GpuComplexVector<_N,_datum> PermuteN(GpuComplexVector<_N,_datum> &in) {   
+      typedef GpuComplexVector<_N,_datum> vec;
+      vec out;					
+      unsigned int _mask = vec::N >> (n + 1);	
+      for(int i=0;i<vec::N;i++) {
+	out.rrrr[i] = in.rrrr[i^_mask];
+	out.iiii[i] = in.iiii[i^_mask];
+      }
+      return out;	
+    }
+    
+    template <typename vec>  static accelerator_inline vec Permute0(vec in) { return PermuteN<0,vec::N,typename vec::datum>(in);  }
+    template <typename vec>  static accelerator_inline vec Permute1(vec in) { return PermuteN<1,vec::N,typename vec::datum>(in);  }
+    template <typename vec>  static accelerator_inline vec Permute2(vec in) { return PermuteN<2,vec::N,typename vec::datum>(in);  }
+    template <typename vec>  static accelerator_inline vec Permute3(vec in) { return PermuteN<3,vec::N,typename vec::datum>(in);  }
+    
+  };
+  
+  struct PrecisionChange {
+
+    ////////////////////////////////////////////////////////////////////////////////////
+    // Single / Half
+    ////////////////////////////////////////////////////////////////////////////////////
+     static accelerator_inline GpuVectorCH StoH (GpuVectorCF a,GpuVectorCF b) {
+      int N = GpuVectorCF::N;
+      GpuVectorCH h;
+      for(int i=0;i<N;i++) {
+        h.rrrr[i  ] = float2half(a.rrrr[i]);
+        h.iiii[i  ] = float2half(a.iiii[i]);
+	h.rrrr[i+N] = float2half(b.rrrr[i]);
+	h.iiii[i+N] = float2half(b.iiii[i]);
+      }
+      return h;
+    }
+    static accelerator_inline void  HtoS (GpuVectorCH h,GpuVectorCF &sa,GpuVectorCF &sb) {
+      int N = GpuVectorCF::N;
+      for(int i=0;i<N;i++) {
+	sa.rrrr[i] = half2float(h.rrrr[i  ]);
+	sa.iiii[i] = half2float(h.iiii[i  ]);
+	sb.rrrr[i] = half2float(h.rrrr[i+N]);
+	sb.iiii[i] = half2float(h.iiii[i+N]);
+      }
+    }
+    static accelerator_inline GpuVectorRH StoH (GpuVectorRF a,GpuVectorRF b) {
+      int N = GpuVectorRF::N;
+      GpuVectorRH h;
+      for(int i=0;i<N;i++) {
+        h.rrrr[i  ] = float2half(a.rrrr[i]);
+	h.rrrr[i+N] = float2half(b.rrrr[i]);
+      }
+      return h;
+    }
+    static accelerator_inline void  HtoS (GpuVectorRH h,GpuVectorRF &sa,GpuVectorRF &sb) {
+      int N = GpuVectorRF::N;
+      for(int i=0;i<N;i++) {
+	sa.rrrr[i] = half2float(h.rrrr[i  ]);
+	sb.rrrr[i] = half2float(h.rrrr[i+N]);
+      }
+    }
+
+    ////////////////////////////////////////////////////////////////////////////////////
+    // Double Single
+    ////////////////////////////////////////////////////////////////////////////////////
+    static accelerator_inline GpuVectorCF DtoS (GpuVectorCD a,GpuVectorCD b) {
+      int N = GpuVectorCD::N;
+      GpuVectorCF h;
+      for(int i=0;i<N;i++) {
+        h.rrrr[i  ] = a.rrrr[i];
+        h.iiii[i  ] = a.iiii[i];
+	h.rrrr[i+N] = b.rrrr[i];
+	h.iiii[i+N] = b.iiii[i];
+      }
+      return h;
+    }
+
+    static accelerator_inline void  StoD (GpuVectorCF h,GpuVectorCD &sa,GpuVectorCD &sb) {
+      int N = GpuVectorCD::N;
+      for(int i=0;i<N;i++) {
+	sa.rrrr[i] = h.rrrr[i  ];
+	sa.iiii[i] = h.iiii[i  ];
+	sb.rrrr[i] = h.rrrr[i+N];
+	sb.iiii[i] = h.iiii[i+N];
+      }
+    }
+
+    static accelerator_inline GpuVectorRF DtoS (GpuVectorRD a,GpuVectorRD b) {
+      int N = GpuVectorRD::N;
+      GpuVectorRF h;
+      for(int i=0;i<N;i++) {
+        h.rrrr[i  ] = a.rrrr[i];
+	h.rrrr[i+N] = b.rrrr[i];
+      }
+      return h;
+    }
+
+    static accelerator_inline void  StoD (GpuVectorRF h,GpuVectorRD &sa,GpuVectorRD &sb) {
+      int N = GpuVectorRD::N;
+      for(int i=0;i<N;i++) {
+	sa.rrrr[i] = h.rrrr[i  ];
+	sb.rrrr[i] = h.rrrr[i+N];
+      }
+    }
+
+    ////////////////////////////////////////////////////////////////////////////////////
+    // Double Half
+    ////////////////////////////////////////////////////////////////////////////////////
+    static accelerator_inline GpuVectorCH DtoH (GpuVectorCD a,GpuVectorCD b,GpuVectorCD c,GpuVectorCD d) {
+      GpuVectorCF sa,sb;
+      sa = DtoS(a,b);
+      sb = DtoS(c,d);
+      return StoH(sa,sb);
+    }
+    static accelerator_inline void HtoD (GpuVectorCH h,GpuVectorCD &a,GpuVectorCD &b,GpuVectorCD &c,GpuVectorCD &d) {
+      GpuVectorCF sa,sb;
+      HtoS(h,sa,sb);
+      StoD(sa,a,b);
+      StoD(sb,c,d);
+    }
+    static accelerator_inline GpuVectorRH DtoH (GpuVectorRD a,GpuVectorRD b,GpuVectorRD c,GpuVectorRD d) {
+      GpuVectorRF sa,sb;
+      sa = DtoS(a,b);
+      sb = DtoS(c,d);
+      return StoH(sa,sb);
+    }
+    static accelerator_inline void HtoD (GpuVectorRH h,GpuVectorRD &a,GpuVectorRD &b,GpuVectorRD &c,GpuVectorRD &d) {
+      GpuVectorRF sa,sb;
+      HtoS(h,sa,sb);
+      StoD(sa,a,b);
+      StoD(sb,c,d);
+    }
+  };
+
+struct Exchange{
+
+  template <int n,int _N, class _datum >
+  static accelerator_inline void ExchangeN(GpuVector<_N,_datum> &out1,
+					   GpuVector<_N,_datum> &out2,
+					   GpuVector<_N,_datum> &in1,
+					   GpuVector<_N,_datum> &in2 )
+  {   
+    typedef GpuVector<_N,_datum> vec;
+    unsigned int mask = vec::N >> (n + 1);
+    for(int i=0;i<vec::N;i++) {
+      int j1 = i&(~mask);
+      if  ( (i&mask) == 0 ) { out1.rrrr[i]=in1.rrrr[j1];}
+      else                  { out1.rrrr[i]=in2.rrrr[j1];}
+      int j2 = i|mask;
+      if  ( (i&mask) == 0 ) { out2.rrrr[i]=in1.rrrr[j2];}
+      else                  { out2.rrrr[i]=in2.rrrr[j2];}
+    }      
+  }
+  template <int n,int _N, class _datum >
+  static accelerator_inline void ExchangeN(GpuComplexVector<_N,_datum> &out1,
+					   GpuComplexVector<_N,_datum> &out2,
+					   GpuComplexVector<_N,_datum> &in1,
+					   GpuComplexVector<_N,_datum> &in2 )
+  {   
+    typedef GpuComplexVector<_N,_datum> vec;
+    unsigned int mask = vec::N >> (n + 1);
+    for(int i=0;i<vec::N;i++) {
+      int j1 = i&(~mask);
+      if  ( (i&mask) == 0 ) {
+	out1.rrrr[i]=in1.rrrr[j1];
+	out1.iiii[i]=in1.iiii[j1];
+      }
+      else                  {
+	out1.rrrr[i]=in2.rrrr[j1];
+	out1.iiii[i]=in2.iiii[j1];
+      }
+      int j2 = i|mask;
+      if  ( (i&mask) == 0 ) {
+	out2.rrrr[i]=in1.rrrr[j2];
+	out2.iiii[i]=in1.iiii[j2];
+      }
+      else                  {
+	out2.rrrr[i]=in2.rrrr[j2];
+	out2.iiii[i]=in2.iiii[j2];
+      }
+    }      
+  }
+  template <typename vec>
+  static accelerator_inline void Exchange0(vec &out1,vec &out2,vec &in1,vec &in2){
+    ExchangeN<0>(out1,out2,in1,in2);
+  };
+  template <typename vec>
+  static accelerator_inline void Exchange1(vec &out1,vec &out2,vec &in1,vec &in2){
+    ExchangeN<1>(out1,out2,in1,in2);
+  };
+  template <typename vec>
+  static accelerator_inline void Exchange2(vec &out1,vec &out2,vec &in1,vec &in2){
+    ExchangeN<2>(out1,out2,in1,in2);
+  };
+  template <typename vec>
+  static accelerator_inline void Exchange3(vec &out1,vec &out2,vec &in1,vec &in2){
+    ExchangeN<3>(out1,out2,in1,in2);
+  };
+
+};
+
+struct Rotate{
+
+  template <int n, typename vec> static accelerator_inline vec tRotate(vec in){
+    return rotate(in, n);
+  }
+    
+  template <int _N, class _datum >
+  static accelerator_inline GpuComplexVector<_N,_datum> rotate_template(GpuComplexVector<_N,_datum> &in, int n)
+  {
+    typedef GpuComplexVector<_N,_datum> vec;
+    vec out;
+    for(int i=0;i<vec::N;i++){
+      out.rrrr[i] = in.rrrr[(i + n)%vec::N];
+      out.iiii[i] = in.iiii[(i + n)%vec::N];
+    }
+    return out;
+  }
+
+  template <int _N, class _datum >
+  static accelerator_inline GpuVector<_N,_datum> rotate_template(GpuVector<_N,_datum> &in, int n)
+  {
+    typedef GpuVector<_N,_datum> vec;
+    vec out;
+    for(int i=0;i<vec::N;i++){
+      out.rrrr[i] = in.rrrr[(i + n)%vec::N];
+    }
+    return out;
+  }
+
+  typedef GpuVectorRH  SIMD_Htype; // Single precision type
+  typedef GpuVectorRF  SIMD_Ftype; // Single precision type
+  typedef GpuVectorRD  SIMD_Dtype; // Double precision type
+  typedef GpuVectorI   SIMD_Itype; // Integer type
+
+  typedef GpuVectorCH  SIMD_CHtype; // Single precision type
+  typedef GpuVectorCF  SIMD_CFtype; // Single precision type
+  typedef GpuVectorCD  SIMD_CDtype; // Double precision type
+
+  static accelerator_inline GpuVectorRH rotate(GpuVectorRH in, int n){ return rotate_template(in,n);}
+  static accelerator_inline GpuVectorRF rotate(GpuVectorRF in, int n){ return rotate_template(in,n);}
+  static accelerator_inline GpuVectorRD rotate(GpuVectorRD in, int n){ return rotate_template(in,n);}
+  static accelerator_inline GpuVectorI  rotate(GpuVectorI  in, int n){ return rotate_template(in,n);}
+  static accelerator_inline GpuVectorCH rotate(GpuVectorCH in, int n){ return rotate_template(in,n/2);} // Measure in complex not float
+  static accelerator_inline GpuVectorCF rotate(GpuVectorCF in, int n){ return rotate_template(in,n/2);}
+  static accelerator_inline GpuVectorCD rotate(GpuVectorCD in, int n){ return rotate_template(in,n/2);}
+
+};
+
+//////////////////////////////////////////////
+// Some Template specialization
+
+  //Complex float Reduce
+  template<>
+  accelerator_inline Grid::ComplexF 
+  Reduce<Grid::ComplexF, GpuVectorCF>::operator()(GpuVectorCF in)
+  {
+    Grid::ComplexF greduce(in.rrrr[0],in.iiii[0]);
+    for(int i=1;i<GpuVectorCF::N;i++) {
+      greduce = greduce+Grid::ComplexF(in.rrrr[i],in.iiii[i]);
+    }
+    return greduce;
+  }
+
+  template<>
+  accelerator_inline Grid::ComplexD
+  Reduce<Grid::ComplexD, GpuVectorCD>::operator()(GpuVectorCD in)
+  {
+    Grid::ComplexD greduce(in.rrrr[0],in.iiii[0]);
+    for(int i=1;i<GpuVectorCD::N;i++) {
+      greduce = greduce+ Grid::ComplexD(in.rrrr[i],in.iiii[i]);
+    }
+    return greduce;
+  }
+
+  // Real
+  template<>
+  accelerator_inline Grid::RealF 
+  Reduce<RealF, GpuVectorRF>::operator()(GpuVectorRF in)
+  {
+    RealF ret = in.rrrr[0];
+    for(int i=1;i<GpuVectorRF::N;i++) {
+      ret = ret+in.rrrr[i];
+    }
+    return ret;
+  }
+
+  template<>
+  accelerator_inline Grid::RealD 
+  Reduce<RealD, GpuVectorRD>::operator()(GpuVectorRD in)
+  {
+    RealD ret = in.rrrr[0];
+    for(int i=1;i<GpuVectorRD::N;i++) {
+      ret = ret+in.rrrr[i];
+    }
+    return ret;
+  }
+
+  template<>
+  accelerator_inline Integer
+  Reduce<Integer, GpuVectorI>::operator()(GpuVectorI in)
+  {
+    Integer ret = in.rrrr[0];
+    for(int i=1;i<GpuVectorI::N;i++) {
+      ret = ret+in.rrrr[i];
+    }
+    return ret;
+  }
+
+}// End optimizatoin
+
+//////////////////////////////////////////////////////////////////////////////////////
+// Here assign types 
+//////////////////////////////////////////////////////////////////////////////////////
+  typedef GpuVectorRH  SIMD_Htype; // Single precision type
+  typedef GpuVectorRF  SIMD_Ftype; // Single precision type
+  typedef GpuVectorRD  SIMD_Dtype; // Double precision type
+  typedef GpuVectorI   SIMD_Itype; // Integer type
+
+  typedef GpuVectorCH  SIMD_CHtype; // Single precision type
+  typedef GpuVectorCF  SIMD_CFtype; // Single precision type
+  typedef GpuVectorCD  SIMD_CDtype; // Double precision type
+
+  // prefetch utilities
+  accelerator_inline void v_prefetch0(int size, const char *ptr){};
+  accelerator_inline void prefetch_HINT_T0(const char *ptr){};
+
+  // Function name aliases
+  typedef Optimization::Vsplat   VsplatSIMD;
+  typedef Optimization::Vstore   VstoreSIMD;
+  typedef Optimization::Vset     VsetSIMD;
+  typedef Optimization::Vstream  VstreamSIMD;
+  template <typename S, typename T> using ReduceSIMD = Optimization::Reduce<S,T>;
+
+  // Arithmetic operations
+  typedef Optimization::Sum         SumSIMD;
+  typedef Optimization::Sub         SubSIMD;
+  typedef Optimization::Div         DivSIMD;
+  typedef Optimization::Mult        MultSIMD;
+  typedef Optimization::MultComplex MultComplexSIMD;
+  typedef Optimization::MultRealPart MultRealPartSIMD;
+  typedef Optimization::MaddRealPart MaddRealPartSIMD;
+  typedef Optimization::Conj        ConjSIMD;
+  typedef Optimization::TimesMinusI TimesMinusISIMD;
+  typedef Optimization::TimesI      TimesISIMD;
+
+}
diff --git a/Grid/simd/Grid_gpu_vec.h b/Grid/simd/Grid_gpu_vec.h
index b2c7588f..6f4528c7 100644
--- a/Grid/simd/Grid_gpu_vec.h
+++ b/Grid/simd/Grid_gpu_vec.h
@@ -38,7 +38,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 #ifdef GRID_HIP
 #include <hip/hip_fp16.h>
 #endif
-#ifdef GRID_SYCL
+#if !defined(GRID_CUDA) && !defined(GRID_HIP)
 namespace Grid {
   typedef struct { uint16_t x;} half;
   typedef struct { half   x; half   y;} half2;
@@ -486,7 +486,7 @@ namespace Optimization {
 
   struct TimesMinusI{
     //Complex single
-    accelerator_inline GpuVectorCF operator()(GpuVectorCF in,GpuVectorCF dummy){
+    accelerator_inline GpuVectorCF operator()(GpuVectorCF in){
       typedef GpuVectorCF vec;
       vec ret;
       for(int i=0;i<vec::N;i++){
@@ -495,7 +495,7 @@ namespace Optimization {
       }
       return ret;
     }
-    accelerator_inline GpuVectorCD operator()(GpuVectorCD in,GpuVectorCD dummy){
+    accelerator_inline GpuVectorCD operator()(GpuVectorCD in){
       typedef GpuVectorCD vec;
       vec ret;
       for(int i=0;i<vec::N;i++){
@@ -508,7 +508,7 @@ namespace Optimization {
 
   struct TimesI{
     //Complex single
-    accelerator_inline GpuVectorCF operator()(GpuVectorCF in,GpuVectorCF dummy){
+    accelerator_inline GpuVectorCF operator()(GpuVectorCF in){
       typedef GpuVectorCF vec;
       vec ret;
       for(int i=0;i<vec::N;i++){
@@ -517,7 +517,7 @@ namespace Optimization {
       }
       return ret;
     }
-    accelerator_inline GpuVectorCD operator()(GpuVectorCD in,GpuVectorCD dummy){
+    accelerator_inline GpuVectorCD operator()(GpuVectorCD in){
       typedef GpuVectorCD vec;
       vec ret;
       for(int i=0;i<vec::N;i++){
diff --git a/Grid/simd/Grid_qpx.h b/Grid/simd/Grid_qpx.h
index d93dbc5b..a1b08514 100644
--- a/Grid/simd/Grid_qpx.h
+++ b/Grid/simd/Grid_qpx.h
@@ -356,7 +356,7 @@ struct Conj{
   
 struct TimesMinusI{
   //Complex double
-  inline vector4double operator()(vector4double v, vector4double ret){
+  inline vector4double operator()(vector4double v){
     return vec_xxcpnmadd(v, (vector4double){1., 1., 1., 1.},
 			 (vector4double){0., 0., 0., 0.});
   }
@@ -367,7 +367,7 @@ struct TimesMinusI{
   
 struct TimesI{
   //Complex double
-  inline vector4double operator()(vector4double v, vector4double ret){
+  inline vector4double operator()(vector4double v){
     return vec_xxcpnmadd(v, (vector4double){-1., -1., -1., -1.},
 			 (vector4double){0., 0., 0., 0.});
   }
diff --git a/Grid/simd/Grid_sse4.h b/Grid/simd/Grid_sse4.h
index 2b2a80fd..cb93dd3e 100644
--- a/Grid/simd/Grid_sse4.h
+++ b/Grid/simd/Grid_sse4.h
@@ -273,27 +273,25 @@ struct Conj{
 
 struct TimesMinusI{
   //Complex single
-  inline __m128 operator()(__m128 in, __m128 ret){
+  inline __m128 operator()(__m128 in){
     __m128 tmp =_mm_addsub_ps(_mm_setzero_ps(),in); // r,-i
     return _mm_shuffle_ps(tmp,tmp,_MM_SELECT_FOUR_FOUR(2,3,0,1));
   }
   //Complex double
-  inline __m128d operator()(__m128d in, __m128d ret){
+  inline __m128d operator()(__m128d in){
     __m128d tmp =_mm_addsub_pd(_mm_setzero_pd(),in); // r,-i
     return _mm_shuffle_pd(tmp,tmp,0x1);
   }
-
-
 };
 
 struct TimesI{
   //Complex single
-  inline __m128 operator()(__m128 in, __m128 ret){
+  inline __m128 operator()(__m128 in){
     __m128 tmp =_mm_shuffle_ps(in,in,_MM_SELECT_FOUR_FOUR(2,3,0,1));
     return _mm_addsub_ps(_mm_setzero_ps(),tmp); // r,-i
   }
   //Complex double
-  inline __m128d operator()(__m128d in, __m128d ret){
+  inline __m128d operator()(__m128d in){
     __m128d tmp = _mm_shuffle_pd(in,in,0x1);
     return _mm_addsub_pd(_mm_setzero_pd(),tmp); // r,-i
   }
diff --git a/Grid/simd/Grid_vector_types.h b/Grid/simd/Grid_vector_types.h
index 4f952bb2..a8e24ae0 100644
--- a/Grid/simd/Grid_vector_types.h
+++ b/Grid/simd/Grid_vector_types.h
@@ -110,11 +110,10 @@ accelerator_inline Grid_half sfw_float_to_half(float ff) {
 #ifdef GPU_VEC
 #include "Grid_gpu_vec.h"
 #endif
-/*
-#ifdef GEN
-#include "Grid_generic.h"
+
+#ifdef GPU_RRII
+#include "Grid_gpu_rrii.h"
 #endif
-*/
 
 #ifdef GEN
   #if defined(A64FX) || defined(A64FXFIXEDSIZE) // breakout A64FX SVE ACLE here
@@ -131,7 +130,6 @@ accelerator_inline Grid_half sfw_float_to_half(float ff) {
       #include "Grid_a64fx-fixedsize.h"
     #endif
   #else
-    //#pragma message("building GEN") // generic
     #include "Grid_generic.h"
   #endif
 #endif
@@ -270,12 +268,14 @@ public:
   typedef Vector_type vector_type;
   typedef Scalar_type scalar_type;
 
+  /*
   typedef union conv_t_union {
     Vector_type v;
     Scalar_type s[sizeof(Vector_type) / sizeof(Scalar_type)];
     accelerator_inline conv_t_union(){};
   } conv_t;
-
+  */
+  
   Vector_type v;
 
   static accelerator_inline constexpr int Nsimd(void) {
@@ -555,15 +555,13 @@ public:
   template <class functor>
   friend accelerator_inline Grid_simd SimdApply(const functor &func, const Grid_simd &v) {
     Grid_simd ret;
-    Grid_simd::conv_t conv;
     Grid_simd::scalar_type s;
 
-    conv.v = v.v;
     for (int i = 0; i < Nsimd(); i++) {
-      s = conv.s[i];
-      conv.s[i] = func(s);
+      s = v.getlane(i);
+      s = func(s);
+      ret.putlane(s,i);
     }
-    ret.v = conv.v;
     return ret;
   }
   template <class functor>
@@ -571,18 +569,14 @@ public:
                                          const Grid_simd &x,
                                          const Grid_simd &y) {
     Grid_simd ret;
-    Grid_simd::conv_t cx;
-    Grid_simd::conv_t cy;
     Grid_simd::scalar_type sx,sy;
 
-    cx.v = x.v;
-    cy.v = y.v;
     for (int i = 0; i < Nsimd(); i++) {
-      sx = cx.s[i];
-      sy = cy.s[i];
-      cx.s[i] = func(sx,sy);
+      sx = x.getlane(i);
+      sy = y.getlane(i);
+      sx = func(sx,sy);
+      ret.putlane(sx,i);
     }
-    ret.v = cx.v;
     return ret;
   }
   ///////////////////////
@@ -645,15 +639,36 @@ public:
   ///////////////////////////////
   // Getting single lanes
   ///////////////////////////////
-  accelerator_inline Scalar_type getlane(int lane) {
+#ifdef GPU_RRII
+  template <class S = Scalar_type,IfComplex<S> = 0>
+  accelerator_inline Scalar_type getlane(int lane) const {
+    return Scalar_type(v.rrrr[lane],v.iiii[lane]);
+  }
+  template <class S = Scalar_type,IfComplex<S> = 0>
+  accelerator_inline void putlane(const Scalar_type &_S, int lane){
+    v.rrrr[lane] = real(_S);
+    v.iiii[lane] = imag(_S);
+  }
+  template <class S = Scalar_type,IfNotComplex<S> = 0>
+  accelerator_inline Scalar_type getlane(int lane) const {
+    return ((S*)&v)[lane];
+  }
+  template <class S = Scalar_type,IfNotComplex<S> = 0>
+  accelerator_inline void putlane(const S &_S, int lane){
+    ((Scalar_type*)&v)[lane] = _S;
+  }
+#else // Can pun to an array of complex
+  accelerator_inline Scalar_type getlane(int lane) const {
     return ((Scalar_type*)&v)[lane];
   }
-
   accelerator_inline void putlane(const Scalar_type &S, int lane){
     ((Scalar_type*)&v)[lane] = S;
   }
+#endif
+
 };  // end of Grid_simd class definition
 
+
 ///////////////////////////////
 // Define available types
 ///////////////////////////////
@@ -663,7 +678,7 @@ typedef Grid_simd<double , SIMD_Dtype> vRealD;
 typedef Grid_simd<Integer, SIMD_Itype> vInteger;
 typedef Grid_simd<uint16_t,SIMD_Htype> vRealH;
 
-#ifdef GPU_VEC
+#if defined(GPU_VEC) || defined(GPU_RRII)
 typedef Grid_simd<complex<uint16_t>, SIMD_CHtype> vComplexH;
 typedef Grid_simd<complex<float>   , SIMD_CFtype> vComplexF;
 typedef Grid_simd<complex<double>  , SIMD_CDtype> vComplexD;
@@ -763,6 +778,7 @@ accelerator_inline void vsplat(Grid_simd<S, V> &ret, NotEnableIf<is_complex<S>,
 }
 //////////////////////////
 
+
 ///////////////////////////////////////////////
 // Initialise to 1,0,i for the correct types
 ///////////////////////////////////////////////
@@ -907,34 +923,6 @@ accelerator_inline Grid_simd<S, V> fxmac(Grid_simd<S, V> a, Grid_simd<S, V> b, G
 // ----------------------------------------------
 
 
-// Distinguish between complex types and others
-template <class S, class V, IfComplex<S> = 0>
-accelerator_inline Grid_simd<S, V> operator/(Grid_simd<S, V> a, Grid_simd<S, V> b) {
-  typedef Grid_simd<S, V> simd;
-
-  simd ret;
-  simd den;
-  typename simd::conv_t conv;
-
-  ret = a * conjugate(b) ;
-  den = b * conjugate(b) ;
-
-  // duplicates real part
-  auto real_den  = toReal(den);
-  simd zden;
-  memcpy((void *)&zden.v,(void *)&real_den.v,sizeof(zden));
-  ret.v=binary<V>(ret.v, zden.v, DivSIMD());
-  return ret;
-};
-
-// Real/Integer types
-template <class S, class V, IfNotComplex<S> = 0>
-accelerator_inline Grid_simd<S, V> operator/(Grid_simd<S, V> a, Grid_simd<S, V> b) {
-  Grid_simd<S, V> ret;
-  ret.v = binary<V>(a.v, b.v, DivSIMD());
-  return ret;
-};
-
 ///////////////////////
 // Conjugate
 ///////////////////////
@@ -959,30 +947,29 @@ accelerator_inline Grid_simd<S, V> adj(const Grid_simd<S, V> &in) {
 ///////////////////////
 template <class S, class V, IfComplex<S> = 0>
 accelerator_inline void timesMinusI(Grid_simd<S, V> &ret, const Grid_simd<S, V> &in) {
-  ret.v = binary<V>(in.v, ret.v, TimesMinusISIMD());
+  ret.v = unary<V>(in.v, TimesMinusISIMD());
 }
 template <class S, class V, IfComplex<S> = 0>
 accelerator_inline Grid_simd<S, V> timesMinusI(const Grid_simd<S, V> &in) {
   Grid_simd<S, V> ret;
-  timesMinusI(ret, in);
+  ret.v=unary<V>(in.v, TimesMinusISIMD());
   return ret;
 }
 template <class S, class V, IfNotComplex<S> = 0>
 accelerator_inline Grid_simd<S, V> timesMinusI(const Grid_simd<S, V> &in) {
   return in;
 }
-
 ///////////////////////
 // timesI
 ///////////////////////
 template <class S, class V, IfComplex<S> = 0>
 accelerator_inline void timesI(Grid_simd<S, V> &ret, const Grid_simd<S, V> &in) {
-  ret.v = binary<V>(in.v, ret.v, TimesISIMD());
+  ret.v = unary<V>(in.v, TimesISIMD());
 }
 template <class S, class V, IfComplex<S> = 0>
 accelerator_inline Grid_simd<S, V> timesI(const Grid_simd<S, V> &in) {
   Grid_simd<S, V> ret;
-  timesI(ret, in);
+  ret.v= unary<V>(in.v, TimesISIMD());
   return ret;
 }
 template <class S, class V, IfNotComplex<S> = 0>
@@ -990,6 +977,35 @@ accelerator_inline Grid_simd<S, V> timesI(const Grid_simd<S, V> &in) {
   return in;
 }
 
+
+// Distinguish between complex types and others
+template <class S, class V, IfComplex<S> = 0>
+accelerator_inline Grid_simd<S, V> operator/(Grid_simd<S, V> a, Grid_simd<S, V> b) {
+  typedef Grid_simd<S, V> simd;
+
+  simd ret;
+  simd den;
+
+  ret = a * conjugate(b) ;
+  den = b * conjugate(b) ;
+
+  // duplicates real part
+  auto real_den  = toReal(den);
+  simd zden;
+  memcpy((void *)&zden.v,(void *)&real_den.v,sizeof(zden));
+  ret.v=binary<V>(ret.v, zden.v, DivSIMD());
+  return ret;
+};
+
+// Real/Integer types
+template <class S, class V, IfNotComplex<S> = 0>
+accelerator_inline Grid_simd<S, V> operator/(Grid_simd<S, V> a, Grid_simd<S, V> b) {
+  Grid_simd<S, V> ret;
+  ret.v = binary<V>(a.v, b.v, DivSIMD());
+  return ret;
+};
+
+
 /////////////////////
 // Inner, outer
 /////////////////////
@@ -1021,12 +1037,12 @@ template <class Csimd>  // must be a real arg
 accelerator_inline typename toRealMapper<Csimd>::Realified toReal(const Csimd &in) {
   typedef typename toRealMapper<Csimd>::Realified Rsimd;
   Rsimd ret;
-  typename Rsimd::conv_t conv;
-  memcpy((void *)&conv.v,(void *)&in.v,sizeof(conv.v));
+  int j=0;
   for (int i = 0; i < Rsimd::Nsimd(); i += 2) {
-    conv.s[i + 1] = conv.s[i];  // duplicate (r,r);(r,r);(r,r); etc...
+    auto s = real(in.getlane(j++));
+    ret.putlane(s,i);
+    ret.putlane(s,i+1);
   }
-  memcpy((void *)&ret.v,(void *)&conv.v,sizeof(ret.v));
   return ret;
 }
 
@@ -1039,18 +1055,19 @@ template <class Rsimd>  // must be a real arg
 accelerator_inline typename toComplexMapper<Rsimd>::Complexified toComplex(const Rsimd &in) {
 
   typedef typename toComplexMapper<Rsimd>::Complexified   Csimd;
-  typename Rsimd::conv_t conv;  // address as real
-
-  conv.v = in.v;
+  typedef typename Csimd::scalar_type scalar_type;
+  int j=0;
+  Csimd ret;
   for (int i = 0; i < Rsimd::Nsimd(); i += 2) {
-    assert(conv.s[i + 1] == conv.s[i]);
+    auto rr = in.getlane(i);
+    auto ri = in.getlane(i+1);
+    assert(rr==ri);
     // trap any cases where real was not duplicated
     // indicating the SIMD grids of real and imag assignment did not correctly
     // match
-    conv.s[i + 1] = 0.0;  // zero imaginary parts
+    scalar_type s(rr,0.0);
+    ret.putlane(s,j++);
   }
-  Csimd ret;
-  memcpy((void *)&ret.v,(void *)&conv.v,sizeof(ret.v));
   return ret;
 }
 
@@ -1146,6 +1163,27 @@ template <> struct is_simd<vInteger>   : public std::true_type {};
 template <typename T> using IfSimd    = Invoke<std::enable_if<is_simd<T>::value, int> >;
 template <typename T> using IfNotSimd = Invoke<std::enable_if<!is_simd<T>::value, unsigned> >;
 
+///////////////////////////////////////////////
+// Convenience insert / extract with complex support
+///////////////////////////////////////////////
+template <class S, class V>
+accelerator_inline S getlane(const Grid_simd<S, V> &in,int lane) {
+  return in.getlane(lane);
+}
+template <class S, class V>
+accelerator_inline void putlane(Grid_simd<S, V> &vec,const S &_S, int lane){
+  vec.putlane(_S,lane);
+}
+template <class S,IfNotSimd<S> = 0 >
+accelerator_inline S getlane(const S &in,int lane) {
+  return in;
+}
+template <class S,IfNotSimd<S> = 0 >
+accelerator_inline void putlane(S &vec,const S &_S, int lane){
+  vec = _S;
+}
+
+
 NAMESPACE_END(Grid);
 
 #endif
diff --git a/Grid/simd/Simd.h b/Grid/simd/Simd.h
index 76ca3bef..9de192bb 100644
--- a/Grid/simd/Simd.h
+++ b/Grid/simd/Simd.h
@@ -69,6 +69,7 @@ typedef RealF  Real;
 typedef thrust::complex<RealF> ComplexF;
 typedef thrust::complex<RealD> ComplexD;
 typedef thrust::complex<Real>  Complex;
+typedef thrust::complex<uint16_t>  ComplexH;
 template<class T> using complex = thrust::complex<T>;
 
 accelerator_inline ComplexD pow(const ComplexD& r,RealD y){ return(thrust::pow(r,(double)y)); }
@@ -77,6 +78,7 @@ accelerator_inline ComplexF pow(const ComplexF& r,RealF y){ return(thrust::pow(r
 typedef std::complex<RealF> ComplexF;
 typedef std::complex<RealD> ComplexD;
 typedef std::complex<Real>  Complex;
+typedef std::complex<uint16_t>  ComplexH; // Hack
 template<class T> using complex = std::complex<T>;
 
 accelerator_inline ComplexD pow(const ComplexD& r,RealD y){ return(std::pow(r,y)); }
diff --git a/Grid/stencil/Stencil.h b/Grid/stencil/Stencil.h
index 19b5e6ea..e23ff258 100644
--- a/Grid/stencil/Stencil.h
+++ b/Grid/stencil/Stencil.h
@@ -216,7 +216,6 @@ class CartesianStencil : public CartesianStencilAccelerator<vobj,cobj,Parameters
 public:
 
   typedef typename cobj::vector_type vector_type;
-  typedef typename cobj::scalar_type scalar_type;
   typedef typename cobj::scalar_object scalar_object;
   typedef const CartesianStencilView<vobj,cobj,Parameters> View_type;
   typedef typename View_type::StencilVector StencilVector;
@@ -1014,7 +1013,6 @@ public:
   int Gather(const Lattice<vobj> &rhs,int dimension,int shift,int cbmask,compressor & compress,int &face_idx, int point)
   {
     typedef typename cobj::vector_type vector_type;
-    typedef typename cobj::scalar_type scalar_type;
 
     int comms_send   = this->_comms_send[point] ;
     int comms_recv   = this->_comms_recv[point] ;
diff --git a/Grid/tensors/Tensor_class.h b/Grid/tensors/Tensor_class.h
index be045ede..f3114cb5 100644
--- a/Grid/tensors/Tensor_class.h
+++ b/Grid/tensors/Tensor_class.h
@@ -178,6 +178,7 @@ public:
     stream << "S {" << o._internal << "}";
     return stream;
   };
+  // FIXME These will break with change of data layout
   strong_inline const scalar_type * begin() const { return reinterpret_cast<const scalar_type *>(&_internal); }
   strong_inline       scalar_type * begin()       { return reinterpret_cast<      scalar_type *>(&_internal); }
   strong_inline const scalar_type * end()   const { return begin() + Traits::count; }
@@ -288,6 +289,7 @@ public:
   //      return _internal[i];
   //    }
 
+  // FIXME These will break with change of data layout
   strong_inline const scalar_type * begin() const { return reinterpret_cast<const scalar_type *>(_internal); }
   strong_inline       scalar_type * begin()       { return reinterpret_cast<      scalar_type *>(_internal); }
   strong_inline const scalar_type * end()   const { return begin() + Traits::count; }
@@ -430,6 +432,7 @@ public:
   //    return _internal[i][j];
   //  }
 
+  // FIXME These will break with change of data layout
   strong_inline const scalar_type * begin() const { return reinterpret_cast<const scalar_type *>(_internal[0]); }
   strong_inline       scalar_type * begin()       { return reinterpret_cast<      scalar_type *>(_internal[0]); }
   strong_inline const scalar_type * end()   const { return begin() + Traits::count; }
diff --git a/Grid/tensors/Tensor_extract_merge.h b/Grid/tensors/Tensor_extract_merge.h
index ab14f81f..c63a2439 100644
--- a/Grid/tensors/Tensor_extract_merge.h
+++ b/Grid/tensors/Tensor_extract_merge.h
@@ -1,5 +1,5 @@
 /*************************************************************************************
-n
+
     Grid physics library, www.github.com/paboyle/Grid 
 
     Source file: ./lib/tensors/Tensor_extract_merge.h
@@ -62,8 +62,18 @@ void extract(const vobj &vec,ExtractBuffer<sobj> &extracted)
   const int words=sizeof(vobj)/sizeof(vector_type);
   const int Nsimd=vector_type::Nsimd();
   const int Nextr=extracted.size();
+  vector_type * vp = (vector_type *)&vec;
   const int s=Nsimd/Nextr;
   sobj_scalar_type *sp = (sobj_scalar_type *) &extracted[0];
+  sobj_scalar_type stmp;
+  for(int w=0;w<words;w++){
+    for(int i=0;i<Nextr;i++){
+      stmp = vp[w].getlane(i*s);
+      sp[i*words+w] =stmp;
+      //      memcpy((char *)&sp[i*words+w],(char *)&stmp,sizeof(stmp));
+    }
+  }
+  /*
   scalar_type *vp = (scalar_type *)&vec;
   scalar_type      vtmp;
   sobj_scalar_type stmp;
@@ -74,6 +84,8 @@ void extract(const vobj &vec,ExtractBuffer<sobj> &extracted)
       memcpy((char *)&sp[i*words+w],(char *)&stmp,sizeof(stmp));
     }
   }
+  */
+  
   return;
 }
 
@@ -93,7 +105,7 @@ void   merge(vobj &vec,ExtractBuffer<sobj> &extracted)
   const int s=Nsimd/Nextr;
 
   sobj_scalar_type *sp = (sobj_scalar_type *)&extracted[0];
-  scalar_type *vp = (scalar_type *)&vec;
+  vector_type *vp = (vector_type *)&vec;
   scalar_type      vtmp;
   sobj_scalar_type stmp;
   for(int w=0;w<words;w++){
@@ -101,7 +113,8 @@ void   merge(vobj &vec,ExtractBuffer<sobj> &extracted)
       for(int ii=0;ii<s;ii++){
 	memcpy((char *)&stmp,(char *)&sp[i*words+w],sizeof(stmp));
 	vtmp = stmp;
-	memcpy((char *)&vp[w*Nsimd+i*s+ii],(char *)&vtmp,sizeof(vtmp));
+	vp[w].putlane(vtmp,i*s+ii);
+	//	memcpy((char *)&vp[w*Nsimd+i*s+ii],(char *)&vtmp,sizeof(vtmp));
       }
     }
   }
@@ -124,9 +137,9 @@ typename vobj::scalar_object extractLane(int lane, const vobj & __restrict__ vec
 
   scalar_object extracted;
   pointer __restrict__  sp = (pointer)&extracted; // Type pun
-  pointer __restrict__  vp = (pointer)&vec;
+  vector_type *vp = (vector_type *)&vec;
   for(int w=0;w<words;w++){
-    sp[w]=vp[w*Nsimd+lane];
+    sp[w]=vp[w].getlane(lane);
   }
   return extracted;
 }
@@ -143,9 +156,9 @@ void insertLane(int lane, vobj & __restrict__ vec,const typename vobj::scalar_ob
   constexpr int Nsimd=vector_type::Nsimd();
 
   pointer __restrict__ sp = (pointer)&extracted;
-  pointer __restrict__ vp = (pointer)&vec;
+  vector_type *vp = (vector_type *)&vec;
   for(int w=0;w<words;w++){
-    vp[w*Nsimd+lane]=sp[w];
+    vp[w].putlane(sp[w],lane);
   }
 }
 
@@ -164,15 +177,13 @@ void extract(const vobj &vec,const ExtractPointerArray<sobj> &extracted, int off
   const int Nextr=extracted.size();
   const int s = Nsimd/Nextr;
 
-  scalar_type * vp = (scalar_type *)&vec;
+  vector_type * vp = (vector_type *)&vec;
   scalar_type      vtmp;
   sobj_scalar_type stmp;
   for(int w=0;w<words;w++){
     for(int i=0;i<Nextr;i++){
       sobj_scalar_type * pointer = (sobj_scalar_type *)& extracted[i][offset];
-      memcpy((char *)&vtmp,(char *)&vp[w*Nsimd+i*s],sizeof(vtmp));
-      stmp = vtmp;
-      memcpy((char *)&pointer[w],(char *)&stmp,sizeof(stmp)); // may do a precision conversion
+      pointer[w] = vp[w].getlane(i*s);
     }
   }
 }
@@ -192,23 +203,21 @@ void merge(vobj &vec,const ExtractPointerArray<sobj> &extracted, int offset)
   const int Nextr=extracted.size();
   const int s = Nsimd/Nextr;
 
-  scalar_type * vp = (scalar_type *)&vec;
+  vector_type * vp = (vector_type *)&vec;
   scalar_type      vtmp;
   sobj_scalar_type stmp;
   for(int w=0;w<words;w++){
     for(int i=0;i<Nextr;i++){
       sobj_scalar_type * pointer = (sobj_scalar_type *)& extracted[i][offset];
       for(int ii=0;ii<s;ii++){
-	memcpy((char *)&stmp,(char *)&pointer[w],sizeof(stmp));
-	vtmp=stmp;
-	memcpy((char *)&vp[w*Nsimd+i*s+ii],(char *)&vtmp,sizeof(vtmp));
+	vtmp=pointer[w];
+	vp[w].putlane(vtmp,i*s+ii);
       }
     }
   }
 }
 
 
-
 //////////////////////////////////////////////////////////////////////////////////
 //Copy a single lane of a SIMD tensor type from one object to another
 //Output object must be of the same tensor type but may be of a different precision (i.e. it can have a different root data type)
@@ -239,12 +248,12 @@ void copyLane(vobjOut & __restrict__ vecOut, int lane_out, const vobjIn & __rest
   iscalar_type itmp;
   oscalar_type otmp;
 
-  opointer __restrict__  op = (opointer)&vecOut;
-  ipointer __restrict__  ip = (ipointer)&vecIn;
+  ovector_type * __restrict__ op = (ovector_type *)&vecOut;
+  ivector_type * __restrict__ ip = (ivector_type *)&vecIn;
   for(int w=0;w<owords;w++){
-    memcpy( (char*)&itmp, (char*)(ip + lane_in + iNsimd*w), sizeof(iscalar_type) );
+    itmp = ip[iNsimd*w].getlane(lane_in);
     otmp = itmp; //potential precision change
-    memcpy( (char*)(op + lane_out + oNsimd*w), (char*)&otmp, sizeof(oscalar_type) );
+    op[oNsimd*w].putlane(otmp,lane_out);
   }
 }
 
diff --git a/configure.ac b/configure.ac
index b26ccc9b..9e7b8b80 100644
--- a/configure.ac
+++ b/configure.ac
@@ -1,5 +1,5 @@
-AC_PREREQ([2.63])
-AC_INIT([Grid], [0.7.0], [https://github.com/paboyle/Grid], [Grid])
+AC_PREREQ([2.71])
+AC_INIT([Grid],[0.7.0],[https://github.com/paboyle/Grid],[Grid])
 AC_CANONICAL_BUILD
 AC_CANONICAL_HOST
 AC_CANONICAL_TARGET
@@ -20,7 +20,7 @@ m4_ifdef([AM_SILENT_RULES], [AM_SILENT_RULES([yes])])
 
 ################ Set flags
 # do not move!
-AC_ARG_ENABLE([debug],[AC_HELP_STRING([--enable-debug=yes|no], [enable debug compilation ])],
+AC_ARG_ENABLE([debug],[AS_HELP_STRING([--enable-debug=yes|no],[enable debug compilation ])],
     [ac_DEBUG=${enable_debug}], [ac_DEBUG=no])
 case ${ac_DEBUG} in
     yes)
@@ -114,7 +114,7 @@ AC_ARG_WITH([openssl],
 
 ############### lapack
 AC_ARG_ENABLE([lapack],
-    [AC_HELP_STRING([--enable-lapack=yes|no|prefix], [enable LAPACK])],
+    [AS_HELP_STRING([--enable-lapack=yes|no|prefix],[enable LAPACK])],
     [ac_LAPACK=${enable_lapack}], [ac_LAPACK=no])
 
 case ${ac_LAPACK} in
@@ -130,7 +130,7 @@ esac
 
 ############### tracing
 AC_ARG_ENABLE([tracing],
-    [AC_HELP_STRING([--enable-tracing=none|nvtx|roctx|timer], [enable tracing])],
+    [AS_HELP_STRING([--enable-tracing=none|nvtx|roctx|timer],[enable tracing])],
     [ac_TRACING=${enable_tracing}], [ac_TRACING=none])
 
 case ${ac_TRACING} in
@@ -150,19 +150,19 @@ esac
 
 ############### fermions
 AC_ARG_ENABLE([fermion-reps],
-     [AC_HELP_STRING([--enable-fermion-reps=yes|no], [enable extra fermion representation support])],
+     [AS_HELP_STRING([--enable-fermion-reps=yes|no],[enable extra fermion representation support])],
      [ac_FERMION_REPS=${enable_fermion_reps}], [ac_FERMION_REPS=yes])
 
 AM_CONDITIONAL(BUILD_FERMION_REPS, [ test "${ac_FERMION_REPS}X" == "yesX" ])
 
 AC_ARG_ENABLE([gparity],
-     [AC_HELP_STRING([--enable-gparity=yes|no], [enable G-parity support])],
+     [AS_HELP_STRING([--enable-gparity=yes|no],[enable G-parity support])],
      [ac_GPARITY=${enable_gparity}], [ac_GPARITY=yes])
 
 AM_CONDITIONAL(BUILD_GPARITY, [ test "${ac_GPARITY}X" == "yesX" ])
 
 AC_ARG_ENABLE([zmobius],
-     [AC_HELP_STRING([--enable-zmobius=yes|no], [enable Zmobius support])],
+     [AS_HELP_STRING([--enable-zmobius=yes|no],[enable Zmobius support])],
      [ac_ZMOBIUS=${enable_zmobius}], [ac_ZMOBIUS=yes])
 
 AM_CONDITIONAL(BUILD_ZMOBIUS, [ test "${ac_ZMOBIUS}X" == "yesX" ])
@@ -179,7 +179,7 @@ case ${ac_ZMOBIUS} in
 esac
 ############### Nc
 AC_ARG_ENABLE([Nc],
-    [AC_HELP_STRING([--enable-Nc=2|3|4|5], [enable number of colours])],
+    [AS_HELP_STRING([--enable-Nc=2|3|4|5],[enable number of colours])],
     [ac_Nc=${enable_Nc}], [ac_Nc=3])
 
 case ${ac_Nc} in
@@ -197,7 +197,7 @@ esac
 
 ############### FP16 conversions
 AC_ARG_ENABLE([sfw-fp16],
-    [AC_HELP_STRING([--enable-sfw-fp16=yes|no], [enable software fp16 comms])],
+    [AS_HELP_STRING([--enable-sfw-fp16=yes|no],[enable software fp16 comms])],
     [ac_SFW_FP16=${enable_sfw_fp16}], [ac_SFW_FP16=yes])
 case ${ac_SFW_FP16} in
     yes)
@@ -209,11 +209,11 @@ esac
 
 ############### Default to accelerator cshift, but revert to host if UCX is buggy or other reasons
 AC_ARG_ENABLE([accelerator-cshift],
-    [AC_HELP_STRING([--enable-accelerator-cshift=yes|no], [run cshift on the device])],
+    [AS_HELP_STRING([--enable-accelerator-cshift=yes|no],[run cshift on the device])],
     [ac_ACC_CSHIFT=${enable_accelerator_cshift}], [ac_ACC_CSHIFT=yes])
 
 AC_ARG_ENABLE([ucx-buggy],
-    [AC_HELP_STRING([--enable-ucx-buggy=yes|no], [enable workaround for UCX device buffer bugs])],
+    [AS_HELP_STRING([--enable-ucx-buggy=yes|no],[enable workaround for UCX device buffer bugs])],
     [ac_UCXBUGGY=${enable_ucx_buggy}], [ac_UCXBUGGY=no])
 
 case ${ac_UCXBUGGY} in
@@ -231,7 +231,7 @@ esac
 
 ############### SYCL/CUDA/HIP/none
 AC_ARG_ENABLE([accelerator],
-    [AC_HELP_STRING([--enable-accelerator=cuda|sycl|hip|none], [enable none,cuda,sycl,hip acceleration])],
+    [AS_HELP_STRING([--enable-accelerator=cuda|sycl|hip|none],[enable none,cuda,sycl,hip acceleration])],
     [ac_ACCELERATOR=${enable_accelerator}], [ac_ACCELERATOR=none])
 case ${ac_ACCELERATOR} in
     cuda)
@@ -254,7 +254,7 @@ esac
 
 ############### UNIFIED MEMORY
 AC_ARG_ENABLE([unified],
-    [AC_HELP_STRING([--enable-unified=yes|no], [enable unified address space for accelerator loops])],
+    [AS_HELP_STRING([--enable-unified=yes|no],[enable unified address space for accelerator loops])],
     [ac_UNIFIED=${enable_unified}], [ac_UNIFIED=yes])
 case ${ac_UNIFIED} in
     yes)
@@ -268,10 +268,10 @@ esac
 
 ############### Intel libraries
 AC_ARG_ENABLE([mkl],
-    [AC_HELP_STRING([--enable-mkl=yes|no|prefix], [enable Intel MKL for LAPACK & FFTW])],
+    [AS_HELP_STRING([--enable-mkl=yes|no|prefix],[enable Intel MKL for LAPACK & FFTW])],
     [ac_MKL=${enable_mkl}], [ac_MKL=no])
 AC_ARG_ENABLE([ipp],
-    [AC_HELP_STRING([--enable-ipp=yes|no|prefix], [enable Intel IPP for fast CRC32C])],
+    [AS_HELP_STRING([--enable-ipp=yes|no|prefix],[enable Intel IPP for fast CRC32C])],
     [ac_IPP=${enable_ipp}], [ac_IPP=no])
 
 case ${ac_MKL} in
@@ -369,8 +369,7 @@ CXXFLAGS=$CXXFLAGS_CPY
 LDFLAGS=$LDFLAGS_CPY
 
 ############### SIMD instruction selection
-AC_ARG_ENABLE([simd],[AC_HELP_STRING([--enable-simd=code],
-              [select SIMD target (cf. README.md)])], [ac_SIMD=${enable_simd}], [ac_SIMD=GEN])
+AC_ARG_ENABLE([simd],[AS_HELP_STRING([--enable-simd=code],[select SIMD target (cf. README.md)])], [ac_SIMD=${enable_simd}], [ac_SIMD=GEN])
 
 AC_ARG_ENABLE([gen-simd-width],
             [AS_HELP_STRING([--enable-gen-simd-width=size],
@@ -435,7 +434,13 @@ case ${ax_cv_cxx_compiler_vendor} in
   clang|gnu)
     case ${ac_SIMD} in
       GPU)
-        AC_DEFINE([GPU_VEC],[1],[GPU vectorised 512bit])
+        AC_DEFINE([GPU_VEC],[1],[GPU vectorised])
+        AC_DEFINE_UNQUOTED([GEN_SIMD_WIDTH],[$ac_gen_simd_width],
+                           [generic SIMD vector width (in bytes)])
+        SIMD_GEN_WIDTH_MSG=" (width= $ac_gen_simd_width)"
+	SIMD_FLAGS='';;
+      GPU-RRII)
+        AC_DEFINE([GPU_RRII],[1],[GPU vectorised with RRRR / IIII layout])
         AC_DEFINE_UNQUOTED([GEN_SIMD_WIDTH],[$ac_gen_simd_width],
                            [generic SIMD vector width (in bytes)])
         SIMD_GEN_WIDTH_MSG=" (width= $ac_gen_simd_width)"
@@ -504,6 +509,12 @@ case ${ax_cv_cxx_compiler_vendor} in
       GPU)
         AC_DEFINE([GPU_VEC],[1],[GPU vectorised ])
 	SIMD_FLAGS='';;
+      GPU-RRII)
+        AC_DEFINE([GPU_RRII],[1],[GPU vectorised with RRRR / IIII layout])
+        AC_DEFINE_UNQUOTED([GEN_SIMD_WIDTH],[$ac_gen_simd_width],
+                           [generic SIMD vector width (in bytes)])
+        SIMD_GEN_WIDTH_MSG=" (width= $ac_gen_simd_width)"
+	SIMD_FLAGS='';;
       SSE4)
         AC_DEFINE([SSE4],[1],[SSE4 intrinsics])
         SIMD_FLAGS='-msse4.2 -xsse4.2';;
@@ -551,8 +562,7 @@ AC_DEFINE([GRID_DEFAULT_PRECISION_DOUBLE],[1],[GRID_DEFAULT_PRECISION is DOUBLE]
 #########################################################
 ######################  GRID ALLOCATOR ALIGNMENT ##
 #########################################################
-AC_ARG_ENABLE([alloc-align],[AC_HELP_STRING([--enable-alloc-align=2MB|4k],
-              [Alignment in bytes of GRID Allocator ])],[ac_ALLOC_ALIGN=${enable_alloc_align}],[ac_ALLOC_ALIGN=2MB])
+AC_ARG_ENABLE([alloc-align],[AS_HELP_STRING([--enable-alloc-align=2MB|4k],[Alignment in bytes of GRID Allocator ])],[ac_ALLOC_ALIGN=${enable_alloc_align}],[ac_ALLOC_ALIGN=2MB])
 case ${ac_ALLOC_ALIGN} in
     4k)
      AC_DEFINE([GRID_ALLOC_ALIGN],[(4096)],[GRID_ALLOC_ALIGN]);;
@@ -561,8 +571,7 @@ case ${ac_ALLOC_ALIGN} in
     *);;
 esac
 
-AC_ARG_ENABLE([alloc-cache],[AC_HELP_STRING([--enable-alloc-cache ],
-              [Cache a pool of recent "frees" to reuse])],[ac_ALLOC_CACHE=${enable_alloc_cache}],[ac_ALLOC_CACHE=yes])
+AC_ARG_ENABLE([alloc-cache],[AS_HELP_STRING([--enable-alloc-cache ],[Cache a pool of recent "frees" to reuse])],[ac_ALLOC_CACHE=${enable_alloc_cache}],[ac_ALLOC_CACHE=yes])
 case ${ac_ALLOC_CACHE} in
     yes)
      AC_DEFINE([ALLOCATION_CACHE],[1],[ALLOCATION_CACHE]);;
@@ -573,8 +582,7 @@ esac
 #########################################################
 ######################  set GPU device to rank in node ##
 #########################################################
-AC_ARG_ENABLE([setdevice],[AC_HELP_STRING([--enable-setdevice | --disable-setdevice],
-              [Set GPU to rank in node with cudaSetDevice or similar])],[ac_SETDEVICE=${enable_SETDEVICE}],[ac_SETDEVICE=no])
+AC_ARG_ENABLE([setdevice],[AS_HELP_STRING([--enable-setdevice | --disable-setdevice],[Set GPU to rank in node with cudaSetDevice or similar])],[ac_SETDEVICE=${enable_SETDEVICE}],[ac_SETDEVICE=no])
 case ${ac_SETDEVICE} in
     yes)
 	echo ENABLE SET DEVICE
@@ -588,8 +596,7 @@ esac
 #########################################################
 ######################  Shared memory intranode #########
 #########################################################
-AC_ARG_ENABLE([shm],[AC_HELP_STRING([--enable-shm=shmopen|shmget|hugetlbfs|shmnone|nvlink|no|none],
-              [Select SHM allocation technique])],[ac_SHM=${enable_shm}],[ac_SHM=no])
+AC_ARG_ENABLE([shm],[AS_HELP_STRING([--enable-shm=shmopen|shmget|hugetlbfs|shmnone|nvlink|no|none],[Select SHM allocation technique])],[ac_SHM=${enable_shm}],[ac_SHM=no])
 
 case ${ac_SHM} in
 
@@ -626,15 +633,13 @@ case ${ac_SHM} in
 esac
 
 ######################  Shared base path for SHMMMAP
-AC_ARG_ENABLE([shmpath],[AC_HELP_STRING([--enable-shmpath=path],
-              [Select SHM mmap base path for hugetlbfs])],
+AC_ARG_ENABLE([shmpath],[AS_HELP_STRING([--enable-shmpath=path],[Select SHM mmap base path for hugetlbfs])],
 	      [ac_SHMPATH=${enable_shmpath}],
 	      [ac_SHMPATH=/var/lib/hugetlbfs/global/pagesize-2MB/])
 AC_DEFINE_UNQUOTED([GRID_SHM_PATH],["$ac_SHMPATH"],[Path to a hugetlbfs filesystem for MMAPing])
 
 ############### force MPI in SMP
-AC_ARG_ENABLE([shm-force-mpi],[AC_HELP_STRING([--enable-shm-force-mpi],
-              [Force MPI within shared memory])],[ac_SHM_FORCE_MPI=${enable_shm_force_mpi}],[ac_SHM_FORCE_MPI=no])
+AC_ARG_ENABLE([shm-force-mpi],[AS_HELP_STRING([--enable-shm-force-mpi],[Force MPI within shared memory])],[ac_SHM_FORCE_MPI=${enable_shm_force_mpi}],[ac_SHM_FORCE_MPI=no])
 case ${ac_SHM_FORCE_MPI} in
      yes)
         AC_DEFINE([GRID_SHM_FORCE_MPI],[1],[GRID_SHM_FORCE_MPI] )
@@ -643,8 +648,7 @@ case ${ac_SHM_FORCE_MPI} in
 esac
 
 ############### communication type selection
-AC_ARG_ENABLE([comms-threads],[AC_HELP_STRING([--enable-comms-threads | --disable-comms-threads],
-              [Use multiple threads in MPI calls])],[ac_COMMS_THREADS=${enable_comms_threads}],[ac_COMMS_THREADS=yes])
+AC_ARG_ENABLE([comms-threads],[AS_HELP_STRING([--enable-comms-threads | --disable-comms-threads],[Use multiple threads in MPI calls])],[ac_COMMS_THREADS=${enable_comms_threads}],[ac_COMMS_THREADS=yes])
 
 case ${ac_COMMS_THREADS} in
      yes)
@@ -654,8 +658,7 @@ case ${ac_COMMS_THREADS} in
 esac
 
 ############### communication type selection
-AC_ARG_ENABLE([comms],[AC_HELP_STRING([--enable-comms=none|mpi|mpi-auto],
-              [Select communications])],[ac_COMMS=${enable_comms}],[ac_COMMS=none])
+AC_ARG_ENABLE([comms],[AS_HELP_STRING([--enable-comms=none|mpi|mpi-auto],[Select communications])],[ac_COMMS=${enable_comms}],[ac_COMMS=none])
 
 
 case ${ac_COMMS} in
@@ -689,8 +692,8 @@ AM_CONDITIONAL(BUILD_COMMS_MPI3,  [ test "${comms_type}X" == "mpi3X" ] )
 AM_CONDITIONAL(BUILD_COMMS_NONE,  [ test "${comms_type}X" == "noneX" ])
 
 ############### RNG selection
-AC_ARG_ENABLE([rng],[AC_HELP_STRING([--enable-rng=ranlux48|mt19937|sitmo],\
-	            [Select Random Number Generator to be used])],\
+AC_ARG_ENABLE([rng],[AS_HELP_STRING([--enable-rng=ranlux48|mt19937|sitmo],[\
+	            Select Random Number Generator to be used])],\
 	            [ac_RNG=${enable_rng}],[ac_RNG=sitmo])
 
 case ${ac_RNG} in
@@ -709,8 +712,8 @@ case ${ac_RNG} in
 esac
 
 ############### Timer option
-AC_ARG_ENABLE([timers],[AC_HELP_STRING([--enable-timers],\
-	            [Enable system dependent high res timers])],\
+AC_ARG_ENABLE([timers],[AS_HELP_STRING([--enable-timers],[\
+	            Enable system dependent high res timers])],\
 	            [ac_TIMERS=${enable_timers}],[ac_TIMERS=yes])
 
 case ${ac_TIMERS} in
@@ -726,8 +729,7 @@ case ${ac_TIMERS} in
 esac
 
 ############### Chroma regression test
-AC_ARG_ENABLE([chroma],[AC_HELP_STRING([--enable-chroma],
-              [Expect chroma compiled under c++11 ])],ac_CHROMA=yes,ac_CHROMA=no)
+AC_ARG_ENABLE([chroma],[AS_HELP_STRING([--enable-chroma],[Expect chroma compiled under c++11 ])],ac_CHROMA=yes,ac_CHROMA=no)
 
 case ${ac_CHROMA} in
      yes|no)
diff --git a/systems/mac-arm/config-command-mpi b/systems/mac-arm/config-command-mpi
index d1e75c39..52506559 100644
--- a/systems/mac-arm/config-command-mpi
+++ b/systems/mac-arm/config-command-mpi
@@ -1 +1,3 @@
-CXX=mpicxx-openmpi-mp CXXFLAGS=-I/opt/local/include/ LDFLAGS=-L/opt/local/lib/ ../../configure --enable-simd=GEN --enable-debug --enable-comms=mpi
+#CXX=mpicxx-openmpi-mp CXXFLAGS=-I/opt/local/include/ LDFLAGS=-L/opt/local/lib/ ../../configure --enable-simd=GEN --enable-debug --enable-comms=mpi
+CXX=mpicxx-openmpi-mp CXXFLAGS=-I/opt/local/include/ LDFLAGS=-L/opt/local/lib/ ../../configure --enable-simd=GPU-RRII --enable-comms=mpi
+#CXX=mpicxx-openmpi-mp CXXFLAGS=-I/opt/local/include/ LDFLAGS=-L/opt/local/lib/ ../../configure --enable-simd=GPU --enable-debug --enable-comms=mpi

From d089739e2f1086c1f436dac1e7812dea395759e0 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Thu, 13 Oct 2022 17:55:50 -0400
Subject: [PATCH 045/146] Hack for lattice sites

---
 Grid/qcd/action/fermion/WilsonKernels.h       | 12 ++--
 .../WilsonKernelsImplementation.h             | 56 ++++++++++++-------
 2 files changed, 42 insertions(+), 26 deletions(-)

diff --git a/Grid/qcd/action/fermion/WilsonKernels.h b/Grid/qcd/action/fermion/WilsonKernels.h
index 68422f28..c5d48095 100644
--- a/Grid/qcd/action/fermion/WilsonKernels.h
+++ b/Grid/qcd/action/fermion/WilsonKernels.h
@@ -90,22 +90,22 @@ private:
       
   // Specialised variants
   static accelerator void GenericDhopSite(StencilView &st,  DoubledGaugeFieldView &U, SiteHalfSpinor * buf,
-					  int sF, int sU, const FermionFieldView &in, FermionFieldView &out);
+					  int Ls, int sF, int sU, const FermionFieldView &in, FermionFieldView &out);
       
   static accelerator void GenericDhopSiteDag(StencilView &st,  DoubledGaugeFieldView &U, SiteHalfSpinor * buf,
-						    int sF, int sU, const FermionFieldView &in, FermionFieldView &out);
+					     int Ls, int sF, int sU, const FermionFieldView &in, FermionFieldView &out);
   
   static accelerator void GenericDhopSiteInt(StencilView &st,  DoubledGaugeFieldView &U, SiteHalfSpinor * buf,
-						    int sF, int sU, const FermionFieldView &in, FermionFieldView &out);
+					     int Ls, int sF, int sU, const FermionFieldView &in, FermionFieldView &out);
       
   static accelerator void GenericDhopSiteDagInt(StencilView &st,  DoubledGaugeFieldView &U, SiteHalfSpinor * buf,
-						int sF, int sU, const FermionFieldView &in, FermionFieldView &out);
+						int Ls, int sF, int sU, const FermionFieldView &in, FermionFieldView &out);
   
   static accelerator void GenericDhopSiteExt(StencilView &st,  DoubledGaugeFieldView &U, SiteHalfSpinor * buf,
-					     int sF, int sU, const FermionFieldView &in, FermionFieldView &out);
+					     int Ls, int sF, int sU, const FermionFieldView &in, FermionFieldView &out);
       
   static accelerator void GenericDhopSiteDagExt(StencilView &st,  DoubledGaugeFieldView &U, SiteHalfSpinor * buf,
-						       int sF, int sU, const FermionFieldView &in, FermionFieldView &out);
+						int Ls, int sF, int sU, const FermionFieldView &in, FermionFieldView &out);
 
   static void AsmDhopSite(StencilView &st,  DoubledGaugeFieldView &U, SiteHalfSpinor * buf,
 			  int sF, int sU, int Ls, int Nsite, const FermionFieldView &in,FermionFieldView &out);
diff --git a/Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h b/Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h
index 939fda33..b5fe63cc 100644
--- a/Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h
+++ b/Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h
@@ -61,9 +61,13 @@ accelerator_inline void get_stencil(StencilEntry * mem, StencilEntry &chip)
     auto tmp = coalescedReadPermute(in[SE->_offset],ptype,perm,lane);	\
     spProj(chi,tmp);						\
   } else {							\
-    chi = coalescedRead(buf[SE->_offset],lane);			\
+    int s = sF %sU ;						\
+    chi = Zero();						\
+    if ( (s==0)||(s==Ls-1)) {					\
+      chi = coalescedRead(buf[SE->_offset],lane);		\
+    }								\
   }								\
-  acceleratorSynchronise();						\
+  acceleratorSynchronise();					\
   Impl::multLink(Uchi, U[sU], chi, Dir, SE, st);		\
   Recon(result, Uchi);
 
@@ -80,11 +84,14 @@ accelerator_inline void get_stencil(StencilEntry * mem, StencilEntry &chip)
 
 #define GENERIC_STENCIL_LEG_EXT(Dir,spProj,Recon)		\
   SE = st.GetEntry(ptype, Dir, sF);				\
-  if (!SE->_is_local ) {		\
-    auto chi = coalescedRead(buf[SE->_offset],lane);		\
-    Impl::multLink(Uchi, U[sU], chi, Dir, SE, st);		\
-    Recon(result, Uchi);					\
-    nmu++;							\
+  if (!SE->_is_local ) {					\
+    int s = sF %sU ;						\
+    if ( (s==0)||(s==Ls-1)) {					\
+      auto chi = coalescedRead(buf[SE->_offset],lane);		\
+      Impl::multLink(Uchi, U[sU], chi, Dir, SE, st);		\
+      Recon(result, Uchi);					\
+      nmu++;							\
+    }								\
   }								\
   acceleratorSynchronise();
 
@@ -111,7 +118,7 @@ accelerator_inline void get_stencil(StencilEntry * mem, StencilEntry &chip)
   ////////////////////////////////////////////////////////////////////
 template <class Impl> accelerator_inline
 void WilsonKernels<Impl>::GenericDhopSiteDag(StencilView &st, DoubledGaugeFieldView &U,
-					     SiteHalfSpinor *buf, int sF,
+					     SiteHalfSpinor *buf, int Ls, int sF,
 					     int sU, const FermionFieldView &in, FermionFieldView &out)
 {
   typedef decltype(coalescedRead(buf[0]))   calcHalfSpinor;
@@ -137,7 +144,7 @@ void WilsonKernels<Impl>::GenericDhopSiteDag(StencilView &st, DoubledGaugeFieldV
 
 template <class Impl> accelerator_inline
 void WilsonKernels<Impl>::GenericDhopSite(StencilView &st, DoubledGaugeFieldView &U,
-					  SiteHalfSpinor *buf, int sF,
+					  SiteHalfSpinor *buf, int Ls, int sF,
 					  int sU, const FermionFieldView &in, FermionFieldView &out)
 {
   typedef decltype(coalescedRead(buf[0])) calcHalfSpinor;
@@ -166,7 +173,7 @@ void WilsonKernels<Impl>::GenericDhopSite(StencilView &st, DoubledGaugeFieldView
   ////////////////////////////////////////////////////////////////////
 template <class Impl> accelerator_inline
 void WilsonKernels<Impl>::GenericDhopSiteDagInt(StencilView &st,  DoubledGaugeFieldView &U,
-						SiteHalfSpinor *buf, int sF,
+						SiteHalfSpinor *buf, int Ls, int sF,
 						int sU, const FermionFieldView &in, FermionFieldView &out)
 {
   typedef decltype(coalescedRead(buf[0])) calcHalfSpinor;
@@ -194,8 +201,8 @@ void WilsonKernels<Impl>::GenericDhopSiteDagInt(StencilView &st,  DoubledGaugeFi
 
 template <class Impl> accelerator_inline
 void WilsonKernels<Impl>::GenericDhopSiteInt(StencilView &st,  DoubledGaugeFieldView &U,
-							 SiteHalfSpinor *buf, int sF,
-							 int sU, const FermionFieldView &in, FermionFieldView &out)
+					     SiteHalfSpinor *buf, int Ls, int sF,
+					     int sU, const FermionFieldView &in, FermionFieldView &out)
 {
   typedef decltype(coalescedRead(buf[0])) calcHalfSpinor;
   typedef decltype(coalescedRead(in[0]))  calcSpinor;
@@ -224,7 +231,7 @@ void WilsonKernels<Impl>::GenericDhopSiteInt(StencilView &st,  DoubledGaugeField
 ////////////////////////////////////////////////////////////////////
 template <class Impl> accelerator_inline
 void WilsonKernels<Impl>::GenericDhopSiteDagExt(StencilView &st,  DoubledGaugeFieldView &U,
-						SiteHalfSpinor *buf, int sF,
+						SiteHalfSpinor *buf, int Ls, int sF,
 						int sU, const FermionFieldView &in, FermionFieldView &out)
 {
   typedef decltype(coalescedRead(buf[0])) calcHalfSpinor;
@@ -255,7 +262,7 @@ void WilsonKernels<Impl>::GenericDhopSiteDagExt(StencilView &st,  DoubledGaugeFi
 
 template <class Impl> accelerator_inline
 void WilsonKernels<Impl>::GenericDhopSiteExt(StencilView &st,  DoubledGaugeFieldView &U,
-					     SiteHalfSpinor *buf, int sF,
+					     SiteHalfSpinor *buf, int Ls, int sF,
 					     int sU, const FermionFieldView &in, FermionFieldView &out)
 {
   typedef decltype(coalescedRead(buf[0])) calcHalfSpinor;
@@ -420,6 +427,15 @@ void WilsonKernels<Impl>::DhopDirKernel( StencilImpl &st, DoubledGaugeField &U,S
       WilsonKernels<Impl>::A(st_v,U_v,buf,sF,sU,in_v,out_v);		\
   });
 
+#define KERNEL_CALLG(A)						\
+  const uint64_t    NN = Nsite*Ls;					\
+  accelerator_forNB( ss, NN, Simd::Nsimd(), {				\
+      int sF = ss;							\
+      int sU = ss/Ls;							\
+      WilsonKernels<Impl>::A(st_v,U_v,buf,Ls,sF,sU,in_v,out_v);		\
+    });									\
+ accelerator_barrier();
+
 #define KERNEL_CALL(A) KERNEL_CALLNB(A); accelerator_barrier();
 
 #define KERNEL_CALL_EXT(A)						\
@@ -450,7 +466,7 @@ void WilsonKernels<Impl>::DhopKernel(int Opt,StencilImpl &st,  DoubledGaugeField
     autoView(st_v , st,AcceleratorRead);
 
    if( interior && exterior ) {
-     if (Opt == WilsonKernelsStatic::OptGeneric    ) { KERNEL_CALL(GenericDhopSite); return;}
+     if (Opt == WilsonKernelsStatic::OptGeneric    ) { KERNEL_CALLG(GenericDhopSite); return;}
 #ifdef SYCL_HACK     
      if (Opt == WilsonKernelsStatic::OptHandUnroll ) { KERNEL_CALL(HandDhopSiteSycl);    return; }
 #else
@@ -460,13 +476,13 @@ void WilsonKernels<Impl>::DhopKernel(int Opt,StencilImpl &st,  DoubledGaugeField
      if (Opt == WilsonKernelsStatic::OptInlineAsm  ) {  ASM_CALL(AsmDhopSite);    return;}
 #endif
    } else if( interior ) {
-     if (Opt == WilsonKernelsStatic::OptGeneric    ) { KERNEL_CALLNB(GenericDhopSiteInt); return;}
+     if (Opt == WilsonKernelsStatic::OptGeneric    ) { KERNEL_CALLG(GenericDhopSiteInt); return;}
      if (Opt == WilsonKernelsStatic::OptHandUnroll ) { KERNEL_CALLNB(HandDhopSiteInt);    return;}
 #ifndef GRID_CUDA
      if (Opt == WilsonKernelsStatic::OptInlineAsm  ) {  ASM_CALL(AsmDhopSiteInt);    return;}
 #endif
    } else if( exterior ) {
-     if (Opt == WilsonKernelsStatic::OptGeneric    ) { KERNEL_CALL(GenericDhopSiteExt); return;}
+     if (Opt == WilsonKernelsStatic::OptGeneric    ) { KERNEL_CALLG(GenericDhopSiteExt); return;}
      if (Opt == WilsonKernelsStatic::OptHandUnroll ) { KERNEL_CALL(HandDhopSiteExt);    return;}
 #ifndef GRID_CUDA
      if (Opt == WilsonKernelsStatic::OptInlineAsm  ) {  ASM_CALL(AsmDhopSiteExt);    return;}
@@ -485,19 +501,19 @@ void WilsonKernels<Impl>::DhopKernel(int Opt,StencilImpl &st,  DoubledGaugeField
     autoView(st_v ,st,AcceleratorRead);
 
    if( interior && exterior ) {
-     if (Opt == WilsonKernelsStatic::OptGeneric    ) { KERNEL_CALL(GenericDhopSiteDag); return;}
+     if (Opt == WilsonKernelsStatic::OptGeneric    ) { KERNEL_CALLG(GenericDhopSiteDag); return;}
      if (Opt == WilsonKernelsStatic::OptHandUnroll ) { KERNEL_CALL(HandDhopSiteDag);    return;}
 #ifndef GRID_CUDA
      if (Opt == WilsonKernelsStatic::OptInlineAsm  ) {  ASM_CALL(AsmDhopSiteDag);     return;}
 #endif
    } else if( interior ) {
-     if (Opt == WilsonKernelsStatic::OptGeneric    ) { KERNEL_CALL(GenericDhopSiteDagInt); return;}
+     if (Opt == WilsonKernelsStatic::OptGeneric    ) { KERNEL_CALLG(GenericDhopSiteDagInt); return;}
      if (Opt == WilsonKernelsStatic::OptHandUnroll ) { KERNEL_CALL(HandDhopSiteDagInt);    return;}
 #ifndef GRID_CUDA
      if (Opt == WilsonKernelsStatic::OptInlineAsm  ) {  ASM_CALL(AsmDhopSiteDagInt);     return;}
 #endif
    } else if( exterior ) {
-     if (Opt == WilsonKernelsStatic::OptGeneric    ) { KERNEL_CALL(GenericDhopSiteDagExt); return;}
+     if (Opt == WilsonKernelsStatic::OptGeneric    ) { KERNEL_CALLG(GenericDhopSiteDagExt); return;}
      if (Opt == WilsonKernelsStatic::OptHandUnroll ) { KERNEL_CALL(HandDhopSiteDagExt);    return;}
 #ifndef GRID_CUDA
      if (Opt == WilsonKernelsStatic::OptInlineAsm  ) {  ASM_CALL(AsmDhopSiteDagExt);     return;}

From 8a07b5200960cabc24a3ceefe9c20c331b1c2954 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Thu, 13 Oct 2022 18:44:47 -0400
Subject: [PATCH 046/146] Dirichlet

---
 .../GeneralEvenOddRationalRatio.h             | 10 +++-
 .../pseudofermion/TwoFlavourEvenOddRatio.h    |  5 ++
 Grid/qcd/hmc/HMC.h                            |  2 +-
 Grid/qcd/hmc/integrators/Integrator.h         | 42 +++++++++++++++
 Grid/simd/Grid_vector_types.h                 | 53 -------------------
 configure.ac                                  |  2 +-
 systems/Crusher/config-command                |  2 +
 7 files changed, 59 insertions(+), 57 deletions(-)

diff --git a/Grid/qcd/action/pseudofermion/GeneralEvenOddRationalRatio.h b/Grid/qcd/action/pseudofermion/GeneralEvenOddRationalRatio.h
index 027f37b3..ff605362 100644
--- a/Grid/qcd/action/pseudofermion/GeneralEvenOddRationalRatio.h
+++ b/Grid/qcd/action/pseudofermion/GeneralEvenOddRationalRatio.h
@@ -59,7 +59,7 @@ NAMESPACE_BEGIN(Grid);
 
       typedef RationalActionParams Params;
       Params param;
-
+      RealD  RefreshAction;
       //For action evaluation
       MultiShiftFunction ApproxPowerAction   ;  //rational approx for X^{1/inv_pow}
       MultiShiftFunction ApproxNegPowerAction;  //rational approx for X^{-1/inv_pow}
@@ -270,12 +270,18 @@ NAMESPACE_BEGIN(Grid);
 	assert(NumOp.ConstEE() == 1);
 	assert(DenOp.ConstEE() == 1);
 	PhiEven = Zero();
-	std::cout<<GridLogMessage << action_name() << " refresh: starting" << std::endl;
+
+	RefreshAction = norm2( etaOdd );
+        std::cout<<GridLogMessage << action_name() << " refresh: action is " << RefreshAction << std::endl;
       };
 
       //////////////////////////////////////////////////////
       // S_f = chi^dag* P(V^dag*V)/Q(V^dag*V)* N(M^dag*M)/D(M^dag*M)* P(V^dag*V)/Q(V^dag*V)* chi       
       //////////////////////////////////////////////////////
+      virtual RealD Sinitial(const GaugeField &U) {
+	return RefreshAction;
+      }
+
       virtual RealD S(const GaugeField &U) {
 	std::cout<<GridLogMessage << action_name() << " compute action: starting" << std::endl;
 	ImportGauge(U);
diff --git a/Grid/qcd/action/pseudofermion/TwoFlavourEvenOddRatio.h b/Grid/qcd/action/pseudofermion/TwoFlavourEvenOddRatio.h
index 56dd6840..f348fde5 100644
--- a/Grid/qcd/action/pseudofermion/TwoFlavourEvenOddRatio.h
+++ b/Grid/qcd/action/pseudofermion/TwoFlavourEvenOddRatio.h
@@ -50,6 +50,8 @@ NAMESPACE_BEGIN(Grid);
       FermionField PhiOdd;   // the pseudo fermion field for this trajectory
       FermionField PhiEven;  // the pseudo fermion field for this trajectory
 
+      RealD RefreshAction;
+      
     public:
       TwoFlavourEvenOddRatioPseudoFermionAction(FermionOperator<Impl>  &_NumOp, 
                                                 FermionOperator<Impl>  &_DenOp, 
@@ -132,6 +134,9 @@ NAMESPACE_BEGIN(Grid);
         // Even det factors
         DenOp.MooeeDag(etaEven,tmp);
         NumOp.MooeeInvDag(tmp,PhiEven);
+
+	RefreshAction = norm2(etaEven)+norm2(etaOdd);
+	std::cout << " refresh " <<action_name()<< " action "<<RefreshAction<<std::endl;
       };
 
       //////////////////////////////////////////////////////
diff --git a/Grid/qcd/hmc/HMC.h b/Grid/qcd/hmc/HMC.h
index a9fe6f46..745e53c8 100644
--- a/Grid/qcd/hmc/HMC.h
+++ b/Grid/qcd/hmc/HMC.h
@@ -189,7 +189,7 @@ private:
     //////////////////////////////////////////////////////////////////////////////////////////////////////
     std::cout << GridLogMessage << "--------------------------------------------------\n";
     std::cout << GridLogMessage << "Compute initial action";
-    RealD H0 = TheIntegrator.S(U);  
+    RealD H0 = TheIntegrator.Sinitial(U);  
     std::cout << GridLogMessage << "--------------------------------------------------\n";
 
     std::streamsize current_precision = std::cout.precision();
diff --git a/Grid/qcd/hmc/integrators/Integrator.h b/Grid/qcd/hmc/integrators/Integrator.h
index 016b1a84..e1f9e3f0 100644
--- a/Grid/qcd/hmc/integrators/Integrator.h
+++ b/Grid/qcd/hmc/integrators/Integrator.h
@@ -419,6 +419,48 @@ public:
     return H;
   }
 
+  struct _Sinitial {
+    template <class FieldType, class Repr>
+    void operator()(std::vector<Action<FieldType>*> repr_set, Repr& Rep, int level, RealD& H) {
+      
+      for (int a = 0; a < repr_set.size(); ++a) {
+        RealD Hterm = repr_set.at(a)->Sinitial(Rep.U);
+        std::cout << GridLogMessage << "Sinitial Level " << level << " term " << a << " H Hirep = " << Hterm << std::endl;
+        H += Hterm;
+
+      }
+    }
+  } Sinitial_hireps{};
+
+  RealD Sinitial(Field& U) 
+  {  // here also U not used
+
+    std::cout << GridLogIntegrator << "Integrator initial action\n";
+
+    RealD H = - FieldImplementation::FieldSquareNorm(P)/HMC_MOMENTUM_DENOMINATOR; // - trace (P*P)/denom
+
+    RealD Hterm;
+
+    // Actions
+    for (int level = 0; level < as.size(); ++level) {
+      for (int actionID = 0; actionID < as[level].actions.size(); ++actionID) {
+        // get gauge field from the SmearingPolicy and
+        // based on the boolean is_smeared in actionID
+        Field& Us = Smearer.get_U(as[level].actions.at(actionID)->is_smeared);
+        std::cout << GridLogMessage << "S [" << level << "][" << actionID << "] action eval " << std::endl;
+	        as[level].actions.at(actionID)->S_timer_start();
+        Hterm = as[level].actions.at(actionID)->Sinitial(Us);
+   	        as[level].actions.at(actionID)->S_timer_stop();
+        std::cout << GridLogMessage << "S [" << level << "][" << actionID << "] H = " << Hterm << std::endl;
+        H += Hterm;
+      }
+      as[level].apply(Sinitial_hireps, Representations, level, H);
+    }
+
+    return H;
+  }
+
+  
   void integrate(Field& U) 
   {
     // reset the clocks
diff --git a/Grid/simd/Grid_vector_types.h b/Grid/simd/Grid_vector_types.h
index 42460c48..daf41cae 100644
--- a/Grid/simd/Grid_vector_types.h
+++ b/Grid/simd/Grid_vector_types.h
@@ -148,23 +148,6 @@ accelerator_inline Grid_half sfw_float_to_half(float ff) {
   #endif
 #endif
 
-/*
-#ifdef A64FXVLA
-#pragma message("building A64FX VLA")
-#if defined(ARMCLANGCOMPAT)
-  #pragma message("applying data types patch")
-#endif
-#include <arm_sve.h>
-#include "Grid_a64fx-2.h"
-#endif
-
-#ifdef A64FXVLS
-#pragma message("building A64FX VLS")
-#include <arm_sve.h>
-#include "Grid_a64fx-fixedsize.h"
-#endif
-*/
-
 #ifdef SSE4
 #include "Grid_sse4.h"
 #endif
@@ -1147,42 +1130,6 @@ static_assert(sizeof(SIMD_Ftype) == sizeof(SIMD_Itype), "SIMD vector lengths inc
 #endif
 #endif
 
-/////////////////////////////////////////
-// Some traits to recognise the types
-/////////////////////////////////////////
-template <typename T>
-struct is_simd : public std::false_type {};
-template <> struct is_simd<vRealF>     : public std::true_type {};
-template <> struct is_simd<vRealD>     : public std::true_type {};
-template <> struct is_simd<vRealH>     : public std::true_type {};
-template <> struct is_simd<vComplexF>  : public std::true_type {};
-template <> struct is_simd<vComplexD>  : public std::true_type {};
-template <> struct is_simd<vComplexH>  : public std::true_type {};
-template <> struct is_simd<vInteger>   : public std::true_type {};
-
-template <typename T> using IfSimd    = Invoke<std::enable_if<is_simd<T>::value, int> >;
-template <typename T> using IfNotSimd = Invoke<std::enable_if<!is_simd<T>::value, unsigned> >;
-
-///////////////////////////////////////////////
-// Convenience insert / extract with complex support
-///////////////////////////////////////////////
-template <class S, class V>
-accelerator_inline S getlane(const Grid_simd<S, V> &in,int lane) {
-  return in.getlane(lane);
-}
-template <class S, class V>
-accelerator_inline void putlane(Grid_simd<S, V> &vec,const S &_S, int lane){
-  vec.putlane(_S,lane);
-}
-template <class S,IfNotSimd<S> = 0 >
-accelerator_inline S getlane(const S &in,int lane) {
-  return in;
-}
-template <class S,IfNotSimd<S> = 0 >
-accelerator_inline void putlane(S &vec,const S &_S, int lane){
-  vec = _S;
-}
-
 
 NAMESPACE_END(Grid);
 
diff --git a/configure.ac b/configure.ac
index 9e7b8b80..2e6199c7 100644
--- a/configure.ac
+++ b/configure.ac
@@ -1,4 +1,4 @@
-AC_PREREQ([2.71])
+AC_PREREQ([2.69])
 AC_INIT([Grid],[0.7.0],[https://github.com/paboyle/Grid],[Grid])
 AC_CANONICAL_BUILD
 AC_CANONICAL_HOST
diff --git a/systems/Crusher/config-command b/systems/Crusher/config-command
index 854c2c01..57b93e03 100644
--- a/systems/Crusher/config-command
+++ b/systems/Crusher/config-command
@@ -14,4 +14,6 @@ CXXFLAGS="-fPIC -I{$ROCM_PATH}/include/ -std=c++14 -I${MPICH_DIR}/include -L/lib
  LDFLAGS="-L/lib64 -L/opt/rocm-5.2.0/lib/ -L${MPICH_DIR}/lib -lmpi -L${CRAY_MPICH_ROOTDIR}/gtl/lib -lmpi_gtl_hsa -lamdhip64 "
 
 
+#--enable-simd=GPU-RRII \
+
 

From 991667ba5e2a5dd972dd8f813e2a35be74b9adc9 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Thu, 13 Oct 2022 18:50:35 -0400
Subject: [PATCH 047/146] Revert

---
 Grid/qcd/action/fermion/WilsonKernels.h       | 12 ++--
 .../WilsonKernelsImplementation.h             | 56 +++++++------------
 2 files changed, 26 insertions(+), 42 deletions(-)

diff --git a/Grid/qcd/action/fermion/WilsonKernels.h b/Grid/qcd/action/fermion/WilsonKernels.h
index c5d48095..68422f28 100644
--- a/Grid/qcd/action/fermion/WilsonKernels.h
+++ b/Grid/qcd/action/fermion/WilsonKernels.h
@@ -90,22 +90,22 @@ private:
       
   // Specialised variants
   static accelerator void GenericDhopSite(StencilView &st,  DoubledGaugeFieldView &U, SiteHalfSpinor * buf,
-					  int Ls, int sF, int sU, const FermionFieldView &in, FermionFieldView &out);
+					  int sF, int sU, const FermionFieldView &in, FermionFieldView &out);
       
   static accelerator void GenericDhopSiteDag(StencilView &st,  DoubledGaugeFieldView &U, SiteHalfSpinor * buf,
-					     int Ls, int sF, int sU, const FermionFieldView &in, FermionFieldView &out);
+						    int sF, int sU, const FermionFieldView &in, FermionFieldView &out);
   
   static accelerator void GenericDhopSiteInt(StencilView &st,  DoubledGaugeFieldView &U, SiteHalfSpinor * buf,
-					     int Ls, int sF, int sU, const FermionFieldView &in, FermionFieldView &out);
+						    int sF, int sU, const FermionFieldView &in, FermionFieldView &out);
       
   static accelerator void GenericDhopSiteDagInt(StencilView &st,  DoubledGaugeFieldView &U, SiteHalfSpinor * buf,
-						int Ls, int sF, int sU, const FermionFieldView &in, FermionFieldView &out);
+						int sF, int sU, const FermionFieldView &in, FermionFieldView &out);
   
   static accelerator void GenericDhopSiteExt(StencilView &st,  DoubledGaugeFieldView &U, SiteHalfSpinor * buf,
-					     int Ls, int sF, int sU, const FermionFieldView &in, FermionFieldView &out);
+					     int sF, int sU, const FermionFieldView &in, FermionFieldView &out);
       
   static accelerator void GenericDhopSiteDagExt(StencilView &st,  DoubledGaugeFieldView &U, SiteHalfSpinor * buf,
-						int Ls, int sF, int sU, const FermionFieldView &in, FermionFieldView &out);
+						       int sF, int sU, const FermionFieldView &in, FermionFieldView &out);
 
   static void AsmDhopSite(StencilView &st,  DoubledGaugeFieldView &U, SiteHalfSpinor * buf,
 			  int sF, int sU, int Ls, int Nsite, const FermionFieldView &in,FermionFieldView &out);
diff --git a/Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h b/Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h
index 000b5445..bdba7cb2 100644
--- a/Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h
+++ b/Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h
@@ -61,13 +61,9 @@ accelerator_inline void get_stencil(StencilEntry * mem, StencilEntry &chip)
     auto tmp = coalescedReadPermute(in[SE->_offset],ptype,perm,lane);	\
     spProj(chi,tmp);						\
   } else {							\
-    int s = sF %sU ;						\
-    chi = Zero();						\
-    if ( (s==0)||(s==Ls-1)) {					\
-      chi = coalescedRead(buf[SE->_offset],lane);		\
-    }								\
+    chi = coalescedRead(buf[SE->_offset],lane);			\
   }								\
-  acceleratorSynchronise();					\
+  acceleratorSynchronise();						\
   Impl::multLink(Uchi, U[sU], chi, Dir, SE, st);		\
   Recon(result, Uchi);
 
@@ -84,14 +80,11 @@ accelerator_inline void get_stencil(StencilEntry * mem, StencilEntry &chip)
 
 #define GENERIC_STENCIL_LEG_EXT(Dir,spProj,Recon)		\
   SE = st.GetEntry(ptype, Dir, sF);				\
-  if (!SE->_is_local ) {					\
-    int s = sF %sU ;						\
-    if ( (s==0)||(s==Ls-1)) {					\
-      auto chi = coalescedRead(buf[SE->_offset],lane);		\
-      Impl::multLink(Uchi, U[sU], chi, Dir, SE, st);		\
-      Recon(result, Uchi);					\
-      nmu++;							\
-    }								\
+  if (!SE->_is_local ) {		\
+    auto chi = coalescedRead(buf[SE->_offset],lane);		\
+    Impl::multLink(Uchi, U[sU], chi, Dir, SE, st);		\
+    Recon(result, Uchi);					\
+    nmu++;							\
   }								\
   acceleratorSynchronise();
 
@@ -118,7 +111,7 @@ accelerator_inline void get_stencil(StencilEntry * mem, StencilEntry &chip)
   ////////////////////////////////////////////////////////////////////
 template <class Impl> accelerator_inline
 void WilsonKernels<Impl>::GenericDhopSiteDag(StencilView &st, DoubledGaugeFieldView &U,
-					     SiteHalfSpinor *buf, int Ls, int sF,
+					     SiteHalfSpinor *buf, int sF,
 					     int sU, const FermionFieldView &in, FermionFieldView &out)
 {
   typedef decltype(coalescedRead(buf[0]))   calcHalfSpinor;
@@ -144,7 +137,7 @@ void WilsonKernels<Impl>::GenericDhopSiteDag(StencilView &st, DoubledGaugeFieldV
 
 template <class Impl> accelerator_inline
 void WilsonKernels<Impl>::GenericDhopSite(StencilView &st, DoubledGaugeFieldView &U,
-					  SiteHalfSpinor *buf, int Ls, int sF,
+					  SiteHalfSpinor *buf, int sF,
 					  int sU, const FermionFieldView &in, FermionFieldView &out)
 {
   typedef decltype(coalescedRead(buf[0])) calcHalfSpinor;
@@ -173,7 +166,7 @@ void WilsonKernels<Impl>::GenericDhopSite(StencilView &st, DoubledGaugeFieldView
   ////////////////////////////////////////////////////////////////////
 template <class Impl> accelerator_inline
 void WilsonKernels<Impl>::GenericDhopSiteDagInt(StencilView &st,  DoubledGaugeFieldView &U,
-						SiteHalfSpinor *buf, int Ls, int sF,
+						SiteHalfSpinor *buf, int sF,
 						int sU, const FermionFieldView &in, FermionFieldView &out)
 {
   typedef decltype(coalescedRead(buf[0])) calcHalfSpinor;
@@ -201,8 +194,8 @@ void WilsonKernels<Impl>::GenericDhopSiteDagInt(StencilView &st,  DoubledGaugeFi
 
 template <class Impl> accelerator_inline
 void WilsonKernels<Impl>::GenericDhopSiteInt(StencilView &st,  DoubledGaugeFieldView &U,
-					     SiteHalfSpinor *buf, int Ls, int sF,
-					     int sU, const FermionFieldView &in, FermionFieldView &out)
+							 SiteHalfSpinor *buf, int sF,
+							 int sU, const FermionFieldView &in, FermionFieldView &out)
 {
   typedef decltype(coalescedRead(buf[0])) calcHalfSpinor;
   typedef decltype(coalescedRead(in[0]))  calcSpinor;
@@ -231,7 +224,7 @@ void WilsonKernels<Impl>::GenericDhopSiteInt(StencilView &st,  DoubledGaugeField
 ////////////////////////////////////////////////////////////////////
 template <class Impl> accelerator_inline
 void WilsonKernels<Impl>::GenericDhopSiteDagExt(StencilView &st,  DoubledGaugeFieldView &U,
-						SiteHalfSpinor *buf, int Ls, int sF,
+						SiteHalfSpinor *buf, int sF,
 						int sU, const FermionFieldView &in, FermionFieldView &out)
 {
   typedef decltype(coalescedRead(buf[0])) calcHalfSpinor;
@@ -262,7 +255,7 @@ void WilsonKernels<Impl>::GenericDhopSiteDagExt(StencilView &st,  DoubledGaugeFi
 
 template <class Impl> accelerator_inline
 void WilsonKernels<Impl>::GenericDhopSiteExt(StencilView &st,  DoubledGaugeFieldView &U,
-					     SiteHalfSpinor *buf, int Ls, int sF,
+					     SiteHalfSpinor *buf, int sF,
 					     int sU, const FermionFieldView &in, FermionFieldView &out)
 {
   typedef decltype(coalescedRead(buf[0])) calcHalfSpinor;
@@ -427,15 +420,6 @@ void WilsonKernels<Impl>::DhopDirKernel( StencilImpl &st, DoubledGaugeField &U,S
       WilsonKernels<Impl>::A(st_v,U_v,buf,sF,sU,in_v,out_v);		\
   });
 
-#define KERNEL_CALLG(A)						\
-  const uint64_t    NN = Nsite*Ls;					\
-  accelerator_forNB( ss, NN, Simd::Nsimd(), {				\
-      int sF = ss;							\
-      int sU = ss/Ls;							\
-      WilsonKernels<Impl>::A(st_v,U_v,buf,Ls,sF,sU,in_v,out_v);		\
-    });									\
- accelerator_barrier();
-
 #define KERNEL_CALL(A) KERNEL_CALLNB(A); accelerator_barrier();
 
 #define KERNEL_CALL_EXT(A)						\
@@ -466,7 +450,7 @@ void WilsonKernels<Impl>::DhopKernel(int Opt,StencilImpl &st,  DoubledGaugeField
     autoView(st_v , st,AcceleratorRead);
 
    if( interior && exterior ) {
-     if (Opt == WilsonKernelsStatic::OptGeneric    ) { KERNEL_CALLG(GenericDhopSite); return;}
+     if (Opt == WilsonKernelsStatic::OptGeneric    ) { KERNEL_CALL(GenericDhopSite); return;}
 #ifdef SYCL_HACK     
      if (Opt == WilsonKernelsStatic::OptHandUnroll ) { KERNEL_CALL(HandDhopSiteSycl);    return; }
 #else
@@ -476,13 +460,13 @@ void WilsonKernels<Impl>::DhopKernel(int Opt,StencilImpl &st,  DoubledGaugeField
      if (Opt == WilsonKernelsStatic::OptInlineAsm  ) {  ASM_CALL(AsmDhopSite);    return;}
 #endif
    } else if( interior ) {
-     if (Opt == WilsonKernelsStatic::OptGeneric    ) { KERNEL_CALLG(GenericDhopSiteInt); return;}
+     if (Opt == WilsonKernelsStatic::OptGeneric    ) { KERNEL_CALLNB(GenericDhopSiteInt); return;}
      if (Opt == WilsonKernelsStatic::OptHandUnroll ) { KERNEL_CALLNB(HandDhopSiteInt);    return;}
 #ifndef GRID_CUDA
      if (Opt == WilsonKernelsStatic::OptInlineAsm  ) {  ASM_CALL(AsmDhopSiteInt);    return;}
 #endif
    } else if( exterior ) {
-     if (Opt == WilsonKernelsStatic::OptGeneric    ) { KERNEL_CALLG(GenericDhopSiteExt); return;}
+     if (Opt == WilsonKernelsStatic::OptGeneric    ) { KERNEL_CALL(GenericDhopSiteExt); return;}
      if (Opt == WilsonKernelsStatic::OptHandUnroll ) { KERNEL_CALL(HandDhopSiteExt);    return;}
 #ifndef GRID_CUDA
      if (Opt == WilsonKernelsStatic::OptInlineAsm  ) {  ASM_CALL(AsmDhopSiteExt);    return;}
@@ -501,21 +485,21 @@ void WilsonKernels<Impl>::DhopKernel(int Opt,StencilImpl &st,  DoubledGaugeField
     autoView(st_v ,st,AcceleratorRead);
 
    if( interior && exterior ) {
-     if (Opt == WilsonKernelsStatic::OptGeneric    ) { KERNEL_CALLG(GenericDhopSiteDag); return;}
+     if (Opt == WilsonKernelsStatic::OptGeneric    ) { KERNEL_CALL(GenericDhopSiteDag); return;}
      if (Opt == WilsonKernelsStatic::OptHandUnroll ) { KERNEL_CALL(HandDhopSiteDag);    return;}
 #ifndef GRID_CUDA
      if (Opt == WilsonKernelsStatic::OptInlineAsm  ) {  ASM_CALL(AsmDhopSiteDag);     return;}
 #endif
      acceleratorFenceComputeStream();
    } else if( interior ) {
-     if (Opt == WilsonKernelsStatic::OptGeneric    ) { KERNEL_CALLG(GenericDhopSiteDagInt); return;}
+     if (Opt == WilsonKernelsStatic::OptGeneric    ) { KERNEL_CALL(GenericDhopSiteDagInt); return;}
      if (Opt == WilsonKernelsStatic::OptHandUnroll ) { KERNEL_CALL(HandDhopSiteDagInt);    return;}
 #ifndef GRID_CUDA
      if (Opt == WilsonKernelsStatic::OptInlineAsm  ) {  ASM_CALL(AsmDhopSiteDagInt);     return;}
 #endif
    } else if( exterior ) {
      acceleratorFenceComputeStream();
-     if (Opt == WilsonKernelsStatic::OptGeneric    ) { KERNEL_CALLG(GenericDhopSiteDagExt); return;}
+     if (Opt == WilsonKernelsStatic::OptGeneric    ) { KERNEL_CALL(GenericDhopSiteDagExt); return;}
      if (Opt == WilsonKernelsStatic::OptHandUnroll ) { KERNEL_CALL(HandDhopSiteDagExt);    return;}
 #ifndef GRID_CUDA
      if (Opt == WilsonKernelsStatic::OptInlineAsm  ) {  ASM_CALL(AsmDhopSiteDagExt);     return;}

From 2e8c3b0ddb64b294bc061741164b1a399de6416c Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 18 Oct 2022 18:10:01 -0400
Subject: [PATCH 048/146] Slow implementation of Shamir DWF

---
 Grid/qcd/action/fermion/DWFSlow.h | 291 ++++++++++++++++++++++++++++++
 1 file changed, 291 insertions(+)
 create mode 100644 Grid/qcd/action/fermion/DWFSlow.h

diff --git a/Grid/qcd/action/fermion/DWFSlow.h b/Grid/qcd/action/fermion/DWFSlow.h
new file mode 100644
index 00000000..61298504
--- /dev/null
+++ b/Grid/qcd/action/fermion/DWFSlow.h
@@ -0,0 +1,291 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/DWFSlow.h
+
+Copyright (C) 2022
+
+Author: Peter Boyle <pboyle@bnl.gov>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+			   /*  END LEGAL */
+#pragma once
+
+NAMESPACE_BEGIN(Grid);
+
+template <class Impl>
+class DWFSlowFermion : public FermionOperator<Impl>
+{
+public:
+  INHERIT_IMPL_TYPES(Impl);
+
+  ///////////////////////////////////////////////////////////////
+  // Implement the abstract base
+  ///////////////////////////////////////////////////////////////
+  GridBase *GaugeGrid(void) { return _grid4; }
+  GridBase *GaugeRedBlackGrid(void) { return _cbgrid4; }
+  GridBase *FermionGrid(void) { return _grid; }
+  GridBase *FermionRedBlackGrid(void) { return _cbgrid; }
+
+  FermionField _tmp;
+  FermionField &tmp(void) { return _tmp; }
+
+  //////////////////////////////////////////////////////////////////
+  // override multiply; cut number routines if pass dagger argument
+  // and also make interface more uniformly consistent
+  //////////////////////////////////////////////////////////////////
+  virtual void  M(const FermionField &in, FermionField &out)
+  {
+    FermionField tmp(_grid);
+    out = (5.0 - M5) * in;
+    Dhop(in,tmp,DaggerNo);
+    out = out + tmp;
+  }
+  virtual void  Mdag(const FermionField &in, FermionField &out)
+  {
+    FermionField tmp(_grid);
+    out = (5.0 - M5) * in;
+    Dhop(in,tmp,DaggerYes);
+    out = out + tmp;
+  };
+
+  /////////////////////////////////////////////////////////
+  // half checkerboard operations 5D redblack so just site identiy
+  /////////////////////////////////////////////////////////
+  void Meooe(const FermionField &in, FermionField &out)
+  {
+    if ( in.Checkerboard() == Odd ) {
+      this->DhopEO(in,out,DaggerNo);
+    } else {
+      this->DhopOE(in,out,DaggerNo);
+    }
+  }
+  void MeooeDag(const FermionField &in, FermionField &out)
+  {
+    if ( in.Checkerboard() == Odd ) {
+      this->DhopEO(in,out,DaggerYes);
+    } else {
+      this->DhopOE(in,out,DaggerYes);
+    }
+  };
+
+  // allow override for twisted mass and clover
+  virtual void Mooee(const FermionField &in, FermionField &out)
+  {
+    out = (5.0 - M5) * in;
+  }
+  virtual void MooeeDag(const FermionField &in, FermionField &out)
+  {
+    out = (5.0 - M5) * in;
+  }
+  virtual void MooeeInv(const FermionField &in, FermionField &out)
+  {
+    out = (1.0/(5.0 - M5)) * in;
+  };
+  virtual void MooeeInvDag(const FermionField &in, FermionField &out)
+  {
+    out = (1.0/(5.0 - M5)) * in;
+  };
+
+  virtual void  MomentumSpacePropagator(FermionField &out,const FermionField &in,RealD _mass,std::vector<double> twist) {} ;
+
+  ////////////////////////
+  // Derivative interface
+  ////////////////////////
+  // Interface calls an internal routine
+  void DhopDeriv(GaugeField &mat,const FermionField &U,const FermionField &V,int dag)  { assert(0);};
+  void DhopDerivOE(GaugeField &mat,const FermionField &U,const FermionField &V,int dag){ assert(0);};
+  void DhopDerivEO(GaugeField &mat,const FermionField &U,const FermionField &V,int dag){ assert(0);};
+
+  ///////////////////////////////////////////////////////////////
+  // non-hermitian hopping term; half cb or both
+  ///////////////////////////////////////////////////////////////
+  void Dhop(const FermionField &in, FermionField &out, int dag)
+  {
+    FermionField tmp(in.Grid());
+    Dhop5(in,out,MassField,MassField,dag );
+    for(int mu=0;mu<4;mu++){
+      DhopDirU(in,Umu[mu],Umu[mu],tmp,mu,dag );    out = out + tmp;
+    }
+  };
+  void DhopOE(const FermionField &in, FermionField &out, int dag)
+  {
+    FermionField tmp(in.Grid());
+    assert(in.Checkerboard()==Even);
+    Dhop5(in,out,MassFieldOdd,MassFieldEven,dag);
+    for(int mu=0;mu<4;mu++){
+      DhopDirU(in,UmuOdd[mu],UmuEven[mu],tmp,mu,dag );    out = out + tmp;
+    }
+  };
+  void DhopEO(const FermionField &in, FermionField &out, int dag)
+  {
+    FermionField tmp(in.Grid());
+    assert(in.Checkerboard()==Odd);
+    Dhop5(in,out, MassFieldEven,MassFieldOdd ,dag );  
+    for(int mu=0;mu<4;mu++){
+      DhopDirU(in,UmuEven[mu],UmuOdd[mu],tmp,mu,dag );    out = out + tmp;
+    }
+  };
+
+  ///////////////////////////////////////////////////////////////
+  // Multigrid assistance; force term uses too
+  ///////////////////////////////////////////////////////////////
+  void Mdir(const FermionField &in, FermionField &out, int dir, int disp){ assert(0);};
+  void MdirAll(const FermionField &in, std::vector<FermionField> &out)   { assert(0);};
+  void DhopDir(const FermionField &in, FermionField &out, int dir, int disp) { assert(0);};
+  void DhopDirAll(const FermionField &in, std::vector<FermionField> &out)    { assert(0);};
+  void DhopDirCalc(const FermionField &in, FermionField &out, int dirdisp,int gamma, int dag) { assert(0);};
+
+  void DhopDirU(const FermionField &in, const GaugeLinkField &U5e, const GaugeLinkField &U5o, FermionField &out, int mu, int dag)
+  {
+    RealD     sgn= 1.0;
+    if (dag ) sgn=-1.0;
+
+    Gamma::Algebra Gmu [] = {
+			 Gamma::Algebra::GammaX,
+			 Gamma::Algebra::GammaY,
+			 Gamma::Algebra::GammaZ,
+			 Gamma::Algebra::GammaT
+    };
+
+    //    mass is  1,1,1,1,-m has to multiply the round the world term
+    FermionField tmp (in.Grid());
+    tmp = U5e * Cshift(in,mu+1,1);
+    out = tmp - Gamma(Gmu[mu])*tmp*sgn;
+    
+    tmp = Cshift(adj(U5o)*in,mu+1,-1);
+    out = out + tmp + Gamma(Gmu[mu])*tmp*sgn;
+
+    out = -0.5*out;
+  };
+
+  void Dhop5(const FermionField &in, FermionField &out, ComplexField &massE, ComplexField &massO, int dag)
+  {
+    // Mass term.... must multiple the round world with mass = 1,1,1,1, -m
+    RealD     sgn= 1.0;
+    if (dag ) sgn=-1.0;
+
+    Gamma G5(Gamma::Algebra::Gamma5);
+
+    FermionField tmp (in.Grid());
+    tmp = massE*Cshift(in,0,1);
+    out = tmp - G5*tmp*sgn;
+    
+    tmp = Cshift(massO*in,0,-1);
+    out = out + tmp + G5*tmp*sgn;
+    out = -0.5*out;
+  };
+
+  // Constructor
+  DWFSlowFermion(GaugeField &_Umu, GridCartesian &Fgrid,
+		 GridRedBlackCartesian &Hgrid, RealD _mass, RealD _M5)
+    :
+    _grid(&Fgrid),
+    _cbgrid(&Hgrid),
+    _grid4(_Umu.Grid()),
+    Umu(Nd,&Fgrid),
+    UmuEven(Nd,&Hgrid),
+    UmuOdd(Nd,&Hgrid),
+    MassField(&Fgrid),
+    MassFieldEven(&Hgrid),
+    MassFieldOdd(&Hgrid),
+    M5(_M5),
+    mass(_mass),
+    _tmp(&Hgrid)
+    {
+      Ls=Fgrid._fdimensions[0];
+      ImportGauge(_Umu);
+
+      typedef typename FermionField::scalar_type scalar;
+
+      Lattice<iScalar<vInteger> > coor(&Fgrid);
+      LatticeCoordinate(coor, 0); // Scoor
+      ComplexField one(&Fgrid);
+      MassField =scalar(-mass);
+      one       =scalar(1.0);
+      MassField =where(coor==Integer(Ls-1),MassField,one);
+      for(int mu=0;mu<Nd;mu++){
+	pickCheckerboard(Even,UmuEven[mu],Umu[mu]);
+	pickCheckerboard(Odd ,UmuOdd[mu],Umu[mu]);
+      }
+      pickCheckerboard(Even,MassFieldEven,MassField);
+      pickCheckerboard(Odd ,MassFieldOdd,MassField);
+      
+    }
+  
+  // DoubleStore impl dependent
+  void ImportGauge(const GaugeField &_Umu4)
+  {
+    GaugeLinkField U4(_grid4);
+    for(int mu=0;mu<Nd;mu++){
+      U4 = PeekIndex<LorentzIndex>(_Umu4, mu);
+      for(int s=0;s<this->Ls;s++){
+	InsertSlice(U4,Umu[mu],s,0);
+      }
+    }
+  }
+
+  ///////////////////////////////////////////////////////////////
+  // Data members require to support the functionality
+  ///////////////////////////////////////////////////////////////
+
+public:
+  virtual RealD Mass(void) { return mass; }
+  virtual int   isTrivialEE(void) { return 1; };
+  RealD mass;
+  RealD M5;
+  int Ls;
+
+  GridBase *_grid4;
+  GridBase *_grid;
+  GridBase *_cbgrid4;
+  GridBase *_cbgrid;
+
+  // Copy of the gauge field , with even and odd subsets
+  std::vector<GaugeLinkField> Umu;
+  std::vector<GaugeLinkField> UmuEven;
+  std::vector<GaugeLinkField> UmuOdd;
+  ComplexField MassField;
+  ComplexField MassFieldEven;
+  ComplexField MassFieldOdd;
+
+  ///////////////////////////////////////////////////////////////
+  // Conserved current utilities
+  ///////////////////////////////////////////////////////////////
+  void ContractConservedCurrent(PropagatorField &q_in_1,
+                                PropagatorField &q_in_2,
+                                PropagatorField &q_out,
+                                PropagatorField &phys_src,
+                                Current curr_type,
+                                unsigned int mu){}
+  void SeqConservedCurrent(PropagatorField &q_in,
+                           PropagatorField &q_out,
+                           PropagatorField &phys_src,
+                           Current curr_type,
+                           unsigned int mu,
+                           unsigned int tmin,
+			   unsigned int tmax,
+			   ComplexField &lattice_cmplx){}
+};
+
+typedef DWFSlowFermion<WilsonImplF> DWFSlowFermionF;
+typedef DWFSlowFermion<WilsonImplD> DWFSlowFermionD;
+
+NAMESPACE_END(Grid);

From 9e4835a3e3769e3b310941b26a58e7a239e4fb43 Mon Sep 17 00:00:00 2001
From: Fabian Joswig <fabian.joswig@ed.ac.uk>
Date: Tue, 25 Oct 2022 15:19:43 +0100
Subject: [PATCH 049/146] feat: changed CompactWilsonExpClover exponentiation
 to Taylor expansion with Horner scheme.

---
 Grid/qcd/action/fermion/CloverHelpers.h       | 203 ++++--------------
 ...CompactWilsonCloverFermionImplementation.h |  11 +-
 2 files changed, 50 insertions(+), 164 deletions(-)

diff --git a/Grid/qcd/action/fermion/CloverHelpers.h b/Grid/qcd/action/fermion/CloverHelpers.h
index 57e71998..c1df81fb 100644
--- a/Grid/qcd/action/fermion/CloverHelpers.h
+++ b/Grid/qcd/action/fermion/CloverHelpers.h
@@ -242,6 +242,19 @@ public:
     // mass term is multiplied to exp(Clover) below
   }
 
+  // Can this be avoided?
+  static void IdentityTimesC(const CloverField& in, RealD c) {
+    int DimRep = Impl::Dimension;
+
+    autoView(in_v, in, AcceleratorWrite);
+
+    accelerator_for(ss, in.Grid()->oSites(), 1, {
+      for (int sa=0; sa<Ns; sa++)
+        for (int ca=0; ca<DimRep; ca++)
+          in_v[ss]()(sa,sa)(ca,ca) = c;
+    });
+  }
+
   static int getNMAX(RealD prec, RealD R) {
     /* compute stop condition for exponential */
     int NMAX=1;
@@ -254,176 +267,46 @@ public:
     return NMAX;
   }
 
-  static int getNMAX(Lattice<iImplCloverDiagonal<vComplexD>> &t, RealD R) {return getNMAX(1e-12,R);}
-  static int getNMAX(Lattice<iImplCloverDiagonal<vComplexF>> &t, RealD R) {return getNMAX(1e-6,R);}
+  static int getNMAX(Lattice<iImplClover<vComplexD>> &t, RealD R) {return getNMAX(1e-12,R);}
+  static int getNMAX(Lattice<iImplClover<vComplexF>> &t, RealD R) {return getNMAX(1e-6,R);}
 
-  static void ExponentiateHermitean6by6(const iMatrix<ComplexD,6> &arg, const RealD& alpha, const std::vector<RealD>& cN, const int Niter, iMatrix<ComplexD,6>& dest){
+  static void Exponentiate_Clover(CloverField& Clover, CloverField& CloverInv, RealD csw_t, RealD diag_mass) {
 
-  	  typedef iMatrix<ComplexD,6> mat;
+    GridBase* grid = Clover.Grid();
+    CloverField ExpClover(grid);
 
-  	  RealD qn[6];
-  	  RealD qnold[6];
-  	  RealD p[5];
-  	  RealD trA2, trA3, trA4;
+    int NMAX = getNMAX(Clover, 3.*csw_t/diag_mass);
 
-  	  mat A2, A3, A4, A5;
-  	  A2 = alpha * alpha * arg * arg;
-  	  A3 = alpha * arg * A2;
-  	  A4 = A2 * A2;
-  	  A5 = A2 * A3;
+    Clover *= (1.0/diag_mass);
 
-  	  trA2 = toReal( trace(A2) );
-  	  trA3 = toReal( trace(A3) );
-  	  trA4 = toReal( trace(A4));
-
-  	  p[0] = toReal( trace(A3 * A3)) / 6.0 - 0.125 * trA4 * trA2 - trA3 * trA3 / 18.0 + trA2 * trA2 * trA2/ 48.0;
-  	  p[1] = toReal( trace(A5)) / 5.0 - trA3 * trA2 / 6.0;
-  	  p[2] = toReal( trace(A4)) / 4.0 - 0.125 * trA2 * trA2;
-  	  p[3] = trA3 / 3.0;
-  	  p[4] = 0.5 * trA2;
-
-  	  qnold[0] = cN[Niter];
-  	  qnold[1] = 0.0;
-  	  qnold[2] = 0.0;
-  	  qnold[3] = 0.0;
-  	  qnold[4] = 0.0;
-  	  qnold[5] = 0.0;
-
-  	  for(int i = Niter-1; i >= 0; i--)
-  	  {
-  	   qn[0] = p[0] * qnold[5] + cN[i];
-  	   qn[1] = p[1] * qnold[5] + qnold[0];
-  	   qn[2] = p[2] * qnold[5] + qnold[1];
-  	   qn[3] = p[3] * qnold[5] + qnold[2];
-  	   qn[4] = p[4] * qnold[5] + qnold[3];
-  	   qn[5] = qnold[4];
-
-  	   qnold[0] = qn[0];
-  	   qnold[1] = qn[1];
-  	   qnold[2] = qn[2];
-  	   qnold[3] = qn[3];
-  	   qnold[4] = qn[4];
-  	   qnold[5] = qn[5];
-  	  }
-
-  	  mat unit(1.0);
-
-  	  dest = (qn[0] * unit + qn[1] * alpha * arg + qn[2] * A2 + qn[3] * A3 + qn[4] * A4 + qn[5] * A5);
-
-    }
-
-  static void Exponentiate_Clover(CloverDiagonalField& Diagonal, CloverTriangleField& Triangle, RealD csw_t, RealD diag_mass) {
-
-    GridBase* grid = Diagonal.Grid();
-    int NMAX = getNMAX(Diagonal, 3.*csw_t/diag_mass);
-
-    //
-    // Implementation completely in Daniel's layout
-    //
-
-    // Taylor expansion with Cayley-Hamilton recursion
-    // underlying Horner scheme as above
+    // Taylor expansion, slow but generic
+    // Horner scheme: a0 + a1 x + a2 x^2 + .. = a0 + x (a1 + x(...))
+    // qN = cN
+    // qn = cn + qn+1 X
     std::vector<RealD> cn(NMAX+1);
     cn[0] = 1.0;
-    for (int i=1; i<=NMAX; i++){
+    for (int i=1; i<=NMAX; i++)
       cn[i] = cn[i-1] / RealD(i);
+
+    ExpClover = Zero();
+    IdentityTimesC(ExpClover, cn[NMAX]);
+    for (int i=NMAX-1; i>=0; i--)
+      ExpClover = ExpClover * Clover + cn[i];
+
+    // prepare inverse
+    CloverInv = (-1.0)*Clover;
+
+    Clover = ExpClover * diag_mass;
+
+    ExpClover = Zero();
+    IdentityTimesC(ExpClover, cn[NMAX]);
+    for (int i=NMAX-1; i>=0; i--)
+      ExpClover = ExpClover * CloverInv + cn[i];
+
+    CloverInv = ExpClover * (1.0/diag_mass);
+
     }
 
-      // Taken over from Daniel's implementation
-      conformable(Diagonal, Triangle);
-
-      long lsites = grid->lSites();
-    {
-      typedef typename SiteCloverDiagonal::scalar_object scalar_object_diagonal;
-      typedef typename SiteCloverTriangle::scalar_object scalar_object_triangle;
-      typedef iMatrix<ComplexD,6> mat;
-
-      autoView(diagonal_v,  Diagonal,  CpuRead);
-      autoView(triangle_v,  Triangle,  CpuRead);
-      autoView(diagonalExp_v, Diagonal, CpuWrite);
-      autoView(triangleExp_v, Triangle, CpuWrite);
-
-      thread_for(site, lsites, { // NOTE: Not on GPU because of (peek/poke)LocalSite
-
-    	  mat srcCloverOpUL(0.0); // upper left block
-    	  mat srcCloverOpLR(0.0); // lower right block
-    	  mat ExpCloverOp;
-
-        scalar_object_diagonal diagonal_tmp     = Zero();
-        scalar_object_diagonal diagonal_exp_tmp = Zero();
-        scalar_object_triangle triangle_tmp     = Zero();
-        scalar_object_triangle triangle_exp_tmp = Zero();
-
-        Coordinate lcoor;
-        grid->LocalIndexToLocalCoor(site, lcoor);
-
-        peekLocalSite(diagonal_tmp, diagonal_v, lcoor);
-        peekLocalSite(triangle_tmp, triangle_v, lcoor);
-
-        int block;
-        block = 0;
-        for(int i = 0; i < 6; i++){
-        	for(int j = 0; j < 6; j++){
-        		if (i == j){
-        			srcCloverOpUL(i,j) = static_cast<ComplexD>(TensorRemove(diagonal_tmp()(block)(i)));
-        		}
-        		else{
-        			srcCloverOpUL(i,j) = static_cast<ComplexD>(TensorRemove(CompactHelpers::triangle_elem(triangle_tmp, block, i, j)));
-        		}
-        	}
-        }
-        block = 1;
-        for(int i = 0; i < 6; i++){
-          	for(int j = 0; j < 6; j++){
-           		if (i == j){
-           			srcCloverOpLR(i,j) = static_cast<ComplexD>(TensorRemove(diagonal_tmp()(block)(i)));
-           		}
-           		else{
-           			srcCloverOpLR(i,j) = static_cast<ComplexD>(TensorRemove(CompactHelpers::triangle_elem(triangle_tmp, block, i, j)));
-           		}
-            }
-        }
-
-        // exp(Clover)
-
-        ExponentiateHermitean6by6(srcCloverOpUL,1.0/diag_mass,cn,NMAX,ExpCloverOp);
-
-        block = 0;
-        for(int i = 0; i < 6; i++){
-        	for(int j = 0; j < 6; j++){
-            	if (i == j){
-            		diagonal_exp_tmp()(block)(i) = ExpCloverOp(i,j);
-            	}
-            	else if(i < j){
-            		triangle_exp_tmp()(block)(CompactHelpers::triangle_index(i, j)) = ExpCloverOp(i,j);
-            	}
-           	}
-        }
-
-        ExponentiateHermitean6by6(srcCloverOpLR,1.0/diag_mass,cn,NMAX,ExpCloverOp);
-
-        block = 1;
-        for(int i = 0; i < 6; i++){
-        	for(int j = 0; j < 6; j++){
-              	if (i == j){
-              		diagonal_exp_tmp()(block)(i) = ExpCloverOp(i,j);
-               	}
-               	else if(i < j){
-               		triangle_exp_tmp()(block)(CompactHelpers::triangle_index(i, j)) = ExpCloverOp(i,j);
-               	}
-            }
-        }
-
-        pokeLocalSite(diagonal_exp_tmp, diagonalExp_v, lcoor);
-        pokeLocalSite(triangle_exp_tmp, triangleExp_v, lcoor);
-      });
-    }
-
-    Diagonal *= diag_mass;
-    Triangle *= diag_mass;
-  }
-
-
   static GaugeLinkField Cmunu(std::vector<GaugeLinkField> &U, GaugeLinkField &lambda, int mu, int nu) {
     assert(0);
     return lambda;
diff --git a/Grid/qcd/action/fermion/implementation/CompactWilsonCloverFermionImplementation.h b/Grid/qcd/action/fermion/implementation/CompactWilsonCloverFermionImplementation.h
index e4864730..e042b985 100644
--- a/Grid/qcd/action/fermion/implementation/CompactWilsonCloverFermionImplementation.h
+++ b/Grid/qcd/action/fermion/implementation/CompactWilsonCloverFermionImplementation.h
@@ -305,6 +305,7 @@ void CompactWilsonCloverFermion<Impl, CloverHelpers>::ImportGauge(const GaugeFie
   GridBase* grid = _Umu.Grid();
   typename Impl::GaugeLinkField Bx(grid), By(grid), Bz(grid), Ex(grid), Ey(grid), Ez(grid);
   CloverField TmpOriginal(grid);
+  CloverField TmpInverse(grid);
 
   // Compute the field strength terms mu>nu
   double t2 = usecond();
@@ -326,14 +327,14 @@ void CompactWilsonCloverFermion<Impl, CloverHelpers>::ImportGauge(const GaugeFie
   TmpOriginal += Helpers::fillCloverZT(Ez) * csw_t;
   // Handle mass term based on clover policy
   CloverHelpers::MassTerm(TmpOriginal, this->diag_mass);
-  
+
   // Convert the data layout of the clover term
   double t4 = usecond();
-  CompactHelpers::ConvertLayout(TmpOriginal, Diagonal, Triangle);
+  CloverHelpers::Exponentiate_Clover(TmpOriginal, TmpInverse, csw_t, this->diag_mass);
 
   // Exponentiate the clover (nothing happens in case of the standard clover)
   double t5 = usecond();
-  CloverHelpers::Exponentiate_Clover(Diagonal, Triangle, csw_t, this->diag_mass);
+  CompactHelpers::ConvertLayout(TmpOriginal, Diagonal, Triangle);
 
   // Possible modify the boundary values
   double t6 = usecond();
@@ -341,7 +342,9 @@ void CompactWilsonCloverFermion<Impl, CloverHelpers>::ImportGauge(const GaugeFie
 
   // Invert the Clover term (explicit inversion needed for the improvement in case of open boundary conditions)
   double t7 = usecond();
-  CompactHelpers::Invert(Diagonal, Triangle, DiagonalInv, TriangleInv);
+  //CompactHelpers::Invert(Diagonal, Triangle, DiagonalInv, TriangleInv);
+  CompactHelpers::ConvertLayout(TmpInverse, DiagonalInv, TriangleInv);
+  //if(open_boundaries) handle differently!
 
   // Fill the remaining clover fields
   double t8 = usecond();

From 513d797ea63245fdffdafd5ea1ea66bb2e357168 Mon Sep 17 00:00:00 2001
From: Fabian Joswig <fabian.joswig@ed.ac.uk>
Date: Tue, 25 Oct 2022 16:17:22 +0100
Subject: [PATCH 050/146] fix: signature of
 CompactWilsonCloverHelpers::Exponentiate fixed.

---
 Grid/qcd/action/fermion/CloverHelpers.h                      | 5 +----
 .../CompactWilsonCloverFermionImplementation.h               | 4 ++--
 2 files changed, 3 insertions(+), 6 deletions(-)

diff --git a/Grid/qcd/action/fermion/CloverHelpers.h b/Grid/qcd/action/fermion/CloverHelpers.h
index c1df81fb..c2bb0d8f 100644
--- a/Grid/qcd/action/fermion/CloverHelpers.h
+++ b/Grid/qcd/action/fermion/CloverHelpers.h
@@ -208,10 +208,7 @@ public:
     Clover += diag_mass;
   }
 
-  static void Exponentiate_Clover(CloverDiagonalField& Diagonal,
-                          CloverTriangleField& Triangle,
-                          RealD csw_t, RealD diag_mass) {
-
+  static void Exponentiate_Clover(CloverField& Clover, CloverField& CloverInv, RealD csw_t, RealD diag_mass) {
     // Do nothing
   }
 
diff --git a/Grid/qcd/action/fermion/implementation/CompactWilsonCloverFermionImplementation.h b/Grid/qcd/action/fermion/implementation/CompactWilsonCloverFermionImplementation.h
index e042b985..e45a531f 100644
--- a/Grid/qcd/action/fermion/implementation/CompactWilsonCloverFermionImplementation.h
+++ b/Grid/qcd/action/fermion/implementation/CompactWilsonCloverFermionImplementation.h
@@ -365,8 +365,8 @@ void CompactWilsonCloverFermion<Impl, CloverHelpers>::ImportGauge(const GaugeFie
   std::cout << GridLogDebug << "allocations =                " << (t2 - t1) / 1e6 << std::endl;
   std::cout << GridLogDebug << "field strength =             " << (t3 - t2) / 1e6 << std::endl;
   std::cout << GridLogDebug << "fill clover =                " << (t4 - t3) / 1e6 << std::endl;
-  std::cout << GridLogDebug << "convert =                    " << (t5 - t4) / 1e6 << std::endl;
-  std::cout << GridLogDebug << "exponentiation =             " << (t6 - t5) / 1e6 << std::endl;
+  std::cout << GridLogDebug << "exponentiation =             " << (t5 - t4) / 1e6 << std::endl;
+  std::cout << GridLogDebug << "convert =                    " << (t6 - t5) / 1e6 << std::endl;
   std::cout << GridLogDebug << "boundaries =                 " << (t7 - t6) / 1e6 << std::endl;
   std::cout << GridLogDebug << "inversions =                 " << (t8 - t7) / 1e6 << std::endl;
   std::cout << GridLogDebug << "pick cbs =                   " << (t9 - t8) / 1e6 << std::endl;

From b36442e263e117ad771b4416dac469dcb6aba2fa Mon Sep 17 00:00:00 2001
From: Fabian Joswig <fabian.joswig@ed.ac.uk>
Date: Tue, 25 Oct 2022 16:57:01 +0100
Subject: [PATCH 051/146] feat: CloverHelpers::InvertClover implemented which
 handles the inversion of the Clover term depending on clover type and the
 boundary conditions.

---
 Grid/qcd/action/fermion/CloverHelpers.h       | 31 +++++++++++++++++--
 ...CompactWilsonCloverFermionImplementation.h |  6 ++--
 2 files changed, 31 insertions(+), 6 deletions(-)

diff --git a/Grid/qcd/action/fermion/CloverHelpers.h b/Grid/qcd/action/fermion/CloverHelpers.h
index c2bb0d8f..e8c45888 100644
--- a/Grid/qcd/action/fermion/CloverHelpers.h
+++ b/Grid/qcd/action/fermion/CloverHelpers.h
@@ -208,10 +208,20 @@ public:
     Clover += diag_mass;
   }
 
-  static void Exponentiate_Clover(CloverField& Clover, CloverField& CloverInv, RealD csw_t, RealD diag_mass) {
+  static void ExponentiateClover(CloverField& Clover, CloverField& CloverInv, RealD csw_t, RealD diag_mass) {
     // Do nothing
   }
 
+  static void InvertClover(CloverField& InvClover,
+                            const CloverDiagonalField& diagonal,
+                            const CloverTriangleField& triangle,
+                            CloverDiagonalField&       diagonalInv,
+                            CloverTriangleField&       triangleInv,
+                            bool open_boundaries) {
+
+    CompactHelpers::Invert(diagonal, triangle, diagonalInv, triangleInv);
+  }
+
   // TODO: implement Cmunu for better performances with compact layout, but don't do it
   // here, but rather in WilsonCloverHelpers.h -> CompactWilsonCloverHelpers
   static GaugeLinkField Cmunu(std::vector<GaugeLinkField> &U, GaugeLinkField &lambda, int mu, int nu) {
@@ -267,7 +277,7 @@ public:
   static int getNMAX(Lattice<iImplClover<vComplexD>> &t, RealD R) {return getNMAX(1e-12,R);}
   static int getNMAX(Lattice<iImplClover<vComplexF>> &t, RealD R) {return getNMAX(1e-6,R);}
 
-  static void Exponentiate_Clover(CloverField& Clover, CloverField& CloverInv, RealD csw_t, RealD diag_mass) {
+  static void ExponentiateClover(CloverField& Clover, CloverField& CloverInv, RealD csw_t, RealD diag_mass) {
 
     GridBase* grid = Clover.Grid();
     CloverField ExpClover(grid);
@@ -302,7 +312,24 @@ public:
 
     CloverInv = ExpClover * (1.0/diag_mass);
 
+  }
+
+  static void InvertClover(CloverField& InvClover,
+                            const CloverDiagonalField& diagonal,
+                            const CloverTriangleField& triangle,
+                            CloverDiagonalField&       diagonalInv,
+                            CloverTriangleField&       triangleInv,
+                            bool open_boundaries) {
+
+    if (open_boundaries)
+    {
+      CompactHelpers::Invert(diagonal, triangle, diagonalInv, triangleInv);
     }
+    else
+    {
+      CompactHelpers::ConvertLayout(InvClover, diagonalInv, triangleInv);
+    }
+  }
 
   static GaugeLinkField Cmunu(std::vector<GaugeLinkField> &U, GaugeLinkField &lambda, int mu, int nu) {
     assert(0);
diff --git a/Grid/qcd/action/fermion/implementation/CompactWilsonCloverFermionImplementation.h b/Grid/qcd/action/fermion/implementation/CompactWilsonCloverFermionImplementation.h
index e45a531f..6ee22113 100644
--- a/Grid/qcd/action/fermion/implementation/CompactWilsonCloverFermionImplementation.h
+++ b/Grid/qcd/action/fermion/implementation/CompactWilsonCloverFermionImplementation.h
@@ -330,7 +330,7 @@ void CompactWilsonCloverFermion<Impl, CloverHelpers>::ImportGauge(const GaugeFie
 
   // Convert the data layout of the clover term
   double t4 = usecond();
-  CloverHelpers::Exponentiate_Clover(TmpOriginal, TmpInverse, csw_t, this->diag_mass);
+  CloverHelpers::ExponentiateClover(TmpOriginal, TmpInverse, csw_t, this->diag_mass);
 
   // Exponentiate the clover (nothing happens in case of the standard clover)
   double t5 = usecond();
@@ -342,9 +342,7 @@ void CompactWilsonCloverFermion<Impl, CloverHelpers>::ImportGauge(const GaugeFie
 
   // Invert the Clover term (explicit inversion needed for the improvement in case of open boundary conditions)
   double t7 = usecond();
-  //CompactHelpers::Invert(Diagonal, Triangle, DiagonalInv, TriangleInv);
-  CompactHelpers::ConvertLayout(TmpInverse, DiagonalInv, TriangleInv);
-  //if(open_boundaries) handle differently!
+  CloverHelpers::InvertClover(TmpInverse, Diagonal, Triangle, DiagonalInv, TriangleInv, open_boundaries);
 
   // Fill the remaining clover fields
   double t8 = usecond();

From 86075fdd45a1f88108615a6dada2df2bee5af8ad Mon Sep 17 00:00:00 2001
From: Fabian Joswig <fabian.joswig@ed.ac.uk>
Date: Tue, 25 Oct 2022 17:05:34 +0100
Subject: [PATCH 052/146] feat: MassTerm and ExponentiateClover merged into
 InstantiateClover

---
 Grid/qcd/action/fermion/CloverHelpers.h             | 13 ++-----------
 .../CompactWilsonCloverFermionImplementation.h      | 12 ++++++------
 2 files changed, 8 insertions(+), 17 deletions(-)

diff --git a/Grid/qcd/action/fermion/CloverHelpers.h b/Grid/qcd/action/fermion/CloverHelpers.h
index e8c45888..b8b4fa59 100644
--- a/Grid/qcd/action/fermion/CloverHelpers.h
+++ b/Grid/qcd/action/fermion/CloverHelpers.h
@@ -204,14 +204,10 @@ public:
   typedef WilsonCloverHelpers<Impl> Helpers;
   typedef CompactWilsonCloverHelpers<Impl> CompactHelpers;
 
-  static void MassTerm(CloverField& Clover, RealD diag_mass) {
+  static void InstantiateClover(CloverField& Clover, CloverField& CloverInv, RealD csw_t, RealD diag_mass) {
     Clover += diag_mass;
   }
 
-  static void ExponentiateClover(CloverField& Clover, CloverField& CloverInv, RealD csw_t, RealD diag_mass) {
-    // Do nothing
-  }
-
   static void InvertClover(CloverField& InvClover,
                             const CloverDiagonalField& diagonal,
                             const CloverTriangleField& triangle,
@@ -244,11 +240,6 @@ public:
   template <typename vtype> using iImplClover = iScalar<iMatrix<iMatrix<vtype, Impl::Dimension>, Ns>>;
   typedef CompactWilsonCloverHelpers<Impl> CompactHelpers;
 
-  static void MassTerm(CloverField& Clover, RealD diag_mass) {
-    // do nothing!
-    // mass term is multiplied to exp(Clover) below
-  }
-
   // Can this be avoided?
   static void IdentityTimesC(const CloverField& in, RealD c) {
     int DimRep = Impl::Dimension;
@@ -277,7 +268,7 @@ public:
   static int getNMAX(Lattice<iImplClover<vComplexD>> &t, RealD R) {return getNMAX(1e-12,R);}
   static int getNMAX(Lattice<iImplClover<vComplexF>> &t, RealD R) {return getNMAX(1e-6,R);}
 
-  static void ExponentiateClover(CloverField& Clover, CloverField& CloverInv, RealD csw_t, RealD diag_mass) {
+  static void InstantiateClover(CloverField& Clover, CloverField& CloverInv, RealD csw_t, RealD diag_mass) {
 
     GridBase* grid = Clover.Grid();
     CloverField ExpClover(grid);
diff --git a/Grid/qcd/action/fermion/implementation/CompactWilsonCloverFermionImplementation.h b/Grid/qcd/action/fermion/implementation/CompactWilsonCloverFermionImplementation.h
index 6ee22113..0828f5fc 100644
--- a/Grid/qcd/action/fermion/implementation/CompactWilsonCloverFermionImplementation.h
+++ b/Grid/qcd/action/fermion/implementation/CompactWilsonCloverFermionImplementation.h
@@ -325,14 +325,14 @@ void CompactWilsonCloverFermion<Impl, CloverHelpers>::ImportGauge(const GaugeFie
   TmpOriginal += Helpers::fillCloverXT(Ex) * csw_t;
   TmpOriginal += Helpers::fillCloverYT(Ey) * csw_t;
   TmpOriginal += Helpers::fillCloverZT(Ez) * csw_t;
-  // Handle mass term based on clover policy
-  CloverHelpers::MassTerm(TmpOriginal, this->diag_mass);
+
+  // Instantiate the clover term
+  // - In case of the standard clover the mass term is added
+  // - In case of the exponential clover the clover term is exponentiated
+  double t4 = usecond();
+  CloverHelpers::InstantiateClover(TmpOriginal, TmpInverse, csw_t, this->diag_mass);
 
   // Convert the data layout of the clover term
-  double t4 = usecond();
-  CloverHelpers::ExponentiateClover(TmpOriginal, TmpInverse, csw_t, this->diag_mass);
-
-  // Exponentiate the clover (nothing happens in case of the standard clover)
   double t5 = usecond();
   CompactHelpers::ConvertLayout(TmpOriginal, Diagonal, Triangle);
 

From 9317d893b201a54c7dfa411fd0aaf6a32aa6f61b Mon Sep 17 00:00:00 2001
From: Fabian Joswig <fabian.joswig@ed.ac.uk>
Date: Tue, 25 Oct 2022 17:10:06 +0100
Subject: [PATCH 053/146] docs: details about inversion of CompactClover term
 added.

---
 .../CompactWilsonCloverFermionImplementation.h               | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/Grid/qcd/action/fermion/implementation/CompactWilsonCloverFermionImplementation.h b/Grid/qcd/action/fermion/implementation/CompactWilsonCloverFermionImplementation.h
index 0828f5fc..40bdfb43 100644
--- a/Grid/qcd/action/fermion/implementation/CompactWilsonCloverFermionImplementation.h
+++ b/Grid/qcd/action/fermion/implementation/CompactWilsonCloverFermionImplementation.h
@@ -340,7 +340,10 @@ void CompactWilsonCloverFermion<Impl, CloverHelpers>::ImportGauge(const GaugeFie
   double t6 = usecond();
   if(open_boundaries) CompactHelpers::ModifyBoundaries(Diagonal, Triangle, csw_t, cF, this->diag_mass);
 
-  // Invert the Clover term (explicit inversion needed for the improvement in case of open boundary conditions)
+  // Invert the Clover term
+  // In case of the exponential clover with (anti-)periodic boundary conditions exp(-Clover) saved
+  // in TmpInverse can be used. In all other cases the clover term has to be explictly inverted.
+  // TODO: For now this inversion is explictly done on the CPU
   double t7 = usecond();
   CloverHelpers::InvertClover(TmpInverse, Diagonal, Triangle, DiagonalInv, TriangleInv, open_boundaries);
 

From a2a879b668d008ad5f70f77e7160c8e855c444a0 Mon Sep 17 00:00:00 2001
From: Fabian Joswig <fabian.joswig@ed.ac.uk>
Date: Tue, 25 Oct 2022 17:20:42 +0100
Subject: [PATCH 054/146] docs: CompactClover Debug Info improved.

---
 .../CompactWilsonCloverFermionImplementation.h         | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/Grid/qcd/action/fermion/implementation/CompactWilsonCloverFermionImplementation.h b/Grid/qcd/action/fermion/implementation/CompactWilsonCloverFermionImplementation.h
index 40bdfb43..16f89725 100644
--- a/Grid/qcd/action/fermion/implementation/CompactWilsonCloverFermionImplementation.h
+++ b/Grid/qcd/action/fermion/implementation/CompactWilsonCloverFermionImplementation.h
@@ -336,7 +336,7 @@ void CompactWilsonCloverFermion<Impl, CloverHelpers>::ImportGauge(const GaugeFie
   double t5 = usecond();
   CompactHelpers::ConvertLayout(TmpOriginal, Diagonal, Triangle);
 
-  // Possible modify the boundary values
+  // Modify the clover term at the temporal boundaries in case of open boundary conditions
   double t6 = usecond();
   if(open_boundaries) CompactHelpers::ModifyBoundaries(Diagonal, Triangle, csw_t, cF, this->diag_mass);
 
@@ -366,10 +366,10 @@ void CompactWilsonCloverFermion<Impl, CloverHelpers>::ImportGauge(const GaugeFie
   std::cout << GridLogDebug << "allocations =                " << (t2 - t1) / 1e6 << std::endl;
   std::cout << GridLogDebug << "field strength =             " << (t3 - t2) / 1e6 << std::endl;
   std::cout << GridLogDebug << "fill clover =                " << (t4 - t3) / 1e6 << std::endl;
-  std::cout << GridLogDebug << "exponentiation =             " << (t5 - t4) / 1e6 << std::endl;
-  std::cout << GridLogDebug << "convert =                    " << (t6 - t5) / 1e6 << std::endl;
-  std::cout << GridLogDebug << "boundaries =                 " << (t7 - t6) / 1e6 << std::endl;
-  std::cout << GridLogDebug << "inversions =                 " << (t8 - t7) / 1e6 << std::endl;
+  std::cout << GridLogDebug << "exponentiate clover =        " << (t5 - t4) / 1e6 << std::endl;
+  std::cout << GridLogDebug << "convert layout =             " << (t6 - t5) / 1e6 << std::endl;
+  std::cout << GridLogDebug << "modify boundaries =          " << (t7 - t6) / 1e6 << std::endl;
+  std::cout << GridLogDebug << "invert clover =              " << (t8 - t7) / 1e6 << std::endl;
   std::cout << GridLogDebug << "pick cbs =                   " << (t9 - t8) / 1e6 << std::endl;
   std::cout << GridLogDebug << "total =                      " << (t9 - t0) / 1e6 << std::endl;
 }

From 5fa6a8b96d6d7d82dfb2ab2c0ebfb844871bcc6b Mon Sep 17 00:00:00 2001
From: Fabian Joswig <fabian.joswig@ed.ac.uk>
Date: Wed, 26 Oct 2022 12:40:28 +0100
Subject: [PATCH 055/146] docs: CompactClover debug info generalized.

---
 .../implementation/CompactWilsonCloverFermionImplementation.h   | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/Grid/qcd/action/fermion/implementation/CompactWilsonCloverFermionImplementation.h b/Grid/qcd/action/fermion/implementation/CompactWilsonCloverFermionImplementation.h
index 16f89725..c8d1ca30 100644
--- a/Grid/qcd/action/fermion/implementation/CompactWilsonCloverFermionImplementation.h
+++ b/Grid/qcd/action/fermion/implementation/CompactWilsonCloverFermionImplementation.h
@@ -366,7 +366,7 @@ void CompactWilsonCloverFermion<Impl, CloverHelpers>::ImportGauge(const GaugeFie
   std::cout << GridLogDebug << "allocations =                " << (t2 - t1) / 1e6 << std::endl;
   std::cout << GridLogDebug << "field strength =             " << (t3 - t2) / 1e6 << std::endl;
   std::cout << GridLogDebug << "fill clover =                " << (t4 - t3) / 1e6 << std::endl;
-  std::cout << GridLogDebug << "exponentiate clover =        " << (t5 - t4) / 1e6 << std::endl;
+  std::cout << GridLogDebug << "instantiate clover =         " << (t5 - t4) / 1e6 << std::endl;
   std::cout << GridLogDebug << "convert layout =             " << (t6 - t5) / 1e6 << std::endl;
   std::cout << GridLogDebug << "modify boundaries =          " << (t7 - t6) / 1e6 << std::endl;
   std::cout << GridLogDebug << "invert clover =              " << (t8 - t7) / 1e6 << std::endl;

From 184adeedb83c716b879d2026d1af902aa2457d02 Mon Sep 17 00:00:00 2001
From: Fabian Joswig <fabian.joswig@ed.ac.uk>
Date: Wed, 26 Oct 2022 12:53:46 +0100
Subject: [PATCH 056/146] feat: renamed open_boundaries to fixedBoundaries

---
 Grid/qcd/action/fermion/CloverHelpers.h       |  6 ++--
 .../fermion/CompactWilsonCloverFermion.h      |  2 +-
 ...CompactWilsonCloverFermionImplementation.h | 32 +++++++++----------
 3 files changed, 20 insertions(+), 20 deletions(-)

diff --git a/Grid/qcd/action/fermion/CloverHelpers.h b/Grid/qcd/action/fermion/CloverHelpers.h
index b8b4fa59..cd469ea7 100644
--- a/Grid/qcd/action/fermion/CloverHelpers.h
+++ b/Grid/qcd/action/fermion/CloverHelpers.h
@@ -213,7 +213,7 @@ public:
                             const CloverTriangleField& triangle,
                             CloverDiagonalField&       diagonalInv,
                             CloverTriangleField&       triangleInv,
-                            bool open_boundaries) {
+                            bool fixedBoundaries) {
 
     CompactHelpers::Invert(diagonal, triangle, diagonalInv, triangleInv);
   }
@@ -310,9 +310,9 @@ public:
                             const CloverTriangleField& triangle,
                             CloverDiagonalField&       diagonalInv,
                             CloverTriangleField&       triangleInv,
-                            bool open_boundaries) {
+                            bool fixedBoundaries) {
 
-    if (open_boundaries)
+    if (fixedBoundaries)
     {
       CompactHelpers::Invert(diagonal, triangle, diagonalInv, triangleInv);
     }
diff --git a/Grid/qcd/action/fermion/CompactWilsonCloverFermion.h b/Grid/qcd/action/fermion/CompactWilsonCloverFermion.h
index 249b20bd..d79b34d4 100644
--- a/Grid/qcd/action/fermion/CompactWilsonCloverFermion.h
+++ b/Grid/qcd/action/fermion/CompactWilsonCloverFermion.h
@@ -225,7 +225,7 @@ public:
   RealD csw_t;
   RealD cF;
 
-  bool open_boundaries;
+  bool fixedBoundaries;
 
   CloverDiagonalField Diagonal,    DiagonalEven,    DiagonalOdd;
   CloverDiagonalField DiagonalInv, DiagonalInvEven, DiagonalInvOdd;
diff --git a/Grid/qcd/action/fermion/implementation/CompactWilsonCloverFermionImplementation.h b/Grid/qcd/action/fermion/implementation/CompactWilsonCloverFermionImplementation.h
index c8d1ca30..7e3b7f00 100644
--- a/Grid/qcd/action/fermion/implementation/CompactWilsonCloverFermionImplementation.h
+++ b/Grid/qcd/action/fermion/implementation/CompactWilsonCloverFermionImplementation.h
@@ -48,7 +48,7 @@ CompactWilsonCloverFermion<Impl, CloverHelpers>::CompactWilsonCloverFermion(Gaug
   , csw_r(_csw_r)
   , csw_t(_csw_t)
   , cF(_cF)
-  , open_boundaries(impl_p.boundary_phases[Nd-1] == 0.0)
+  , fixedBoundaries(impl_p.boundary_phases[Nd-1] == 0.0)
   , Diagonal(&Fgrid),        Triangle(&Fgrid)
   , DiagonalEven(&Hgrid),    TriangleEven(&Hgrid)
   , DiagonalOdd(&Hgrid),     TriangleOdd(&Hgrid)
@@ -67,7 +67,7 @@ CompactWilsonCloverFermion<Impl, CloverHelpers>::CompactWilsonCloverFermion(Gaug
     csw_r /= clover_anisotropy.xi_0;
 
   ImportGauge(_Umu);
-  if (open_boundaries) {
+  if (fixedBoundaries) {
     this->BoundaryMaskEven.Checkerboard() = Even;
     this->BoundaryMaskOdd.Checkerboard() = Odd;
     CompactHelpers::SetupMasks(this->BoundaryMask, this->BoundaryMaskEven, this->BoundaryMaskOdd);
@@ -77,31 +77,31 @@ CompactWilsonCloverFermion<Impl, CloverHelpers>::CompactWilsonCloverFermion(Gaug
 template<class Impl, class CloverHelpers>
 void CompactWilsonCloverFermion<Impl, CloverHelpers>::Dhop(const FermionField& in, FermionField& out, int dag) {
   WilsonBase::Dhop(in, out, dag);
-  if(open_boundaries) ApplyBoundaryMask(out);
+  if(fixedBoundaries) ApplyBoundaryMask(out);
 }
 
 template<class Impl, class CloverHelpers>
 void CompactWilsonCloverFermion<Impl, CloverHelpers>::DhopOE(const FermionField& in, FermionField& out, int dag) {
   WilsonBase::DhopOE(in, out, dag);
-  if(open_boundaries) ApplyBoundaryMask(out);
+  if(fixedBoundaries) ApplyBoundaryMask(out);
 }
 
 template<class Impl, class CloverHelpers>
 void CompactWilsonCloverFermion<Impl, CloverHelpers>::DhopEO(const FermionField& in, FermionField& out, int dag) {
   WilsonBase::DhopEO(in, out, dag);
-  if(open_boundaries) ApplyBoundaryMask(out);
+  if(fixedBoundaries) ApplyBoundaryMask(out);
 }
 
 template<class Impl, class CloverHelpers>
 void CompactWilsonCloverFermion<Impl, CloverHelpers>::DhopDir(const FermionField& in, FermionField& out, int dir, int disp) {
   WilsonBase::DhopDir(in, out, dir, disp);
-  if(this->open_boundaries) ApplyBoundaryMask(out);
+  if(this->fixedBoundaries) ApplyBoundaryMask(out);
 }
 
 template<class Impl, class CloverHelpers>
 void CompactWilsonCloverFermion<Impl, CloverHelpers>::DhopDirAll(const FermionField& in, std::vector<FermionField>& out) {
   WilsonBase::DhopDirAll(in, out);
-  if(this->open_boundaries) {
+  if(this->fixedBoundaries) {
     for(auto& o : out) ApplyBoundaryMask(o);
   }
 }
@@ -112,7 +112,7 @@ void CompactWilsonCloverFermion<Impl, CloverHelpers>::M(const FermionField& in,
   WilsonBase::Dhop(in, out, DaggerNo); // call base to save applying bc
   Mooee(in, Tmp);
   axpy(out, 1.0, out, Tmp);
-  if(open_boundaries) ApplyBoundaryMask(out);
+  if(fixedBoundaries) ApplyBoundaryMask(out);
 }
 
 template<class Impl, class CloverHelpers>
@@ -121,19 +121,19 @@ void CompactWilsonCloverFermion<Impl, CloverHelpers>::Mdag(const FermionField& i
   WilsonBase::Dhop(in, out, DaggerYes);  // call base to save applying bc
   MooeeDag(in, Tmp);
   axpy(out, 1.0, out, Tmp);
-  if(open_boundaries) ApplyBoundaryMask(out);
+  if(fixedBoundaries) ApplyBoundaryMask(out);
 }
 
 template<class Impl, class CloverHelpers>
 void CompactWilsonCloverFermion<Impl, CloverHelpers>::Meooe(const FermionField& in, FermionField& out) {
   WilsonBase::Meooe(in, out);
-  if(open_boundaries) ApplyBoundaryMask(out);
+  if(fixedBoundaries) ApplyBoundaryMask(out);
 }
 
 template<class Impl, class CloverHelpers>
 void CompactWilsonCloverFermion<Impl, CloverHelpers>::MeooeDag(const FermionField& in, FermionField& out) {
   WilsonBase::MeooeDag(in, out);
-  if(open_boundaries) ApplyBoundaryMask(out);
+  if(fixedBoundaries) ApplyBoundaryMask(out);
 }
 
 template<class Impl, class CloverHelpers>
@@ -147,7 +147,7 @@ void CompactWilsonCloverFermion<Impl, CloverHelpers>::Mooee(const FermionField&
   } else {
     MooeeInternal(in, out, Diagonal, Triangle);
   }
-  if(open_boundaries) ApplyBoundaryMask(out);
+  if(fixedBoundaries) ApplyBoundaryMask(out);
 }
 
 template<class Impl, class CloverHelpers>
@@ -166,7 +166,7 @@ void CompactWilsonCloverFermion<Impl, CloverHelpers>::MooeeInv(const FermionFiel
   } else {
     MooeeInternal(in, out, DiagonalInv, TriangleInv);
   }
-  if(open_boundaries) ApplyBoundaryMask(out);
+  if(fixedBoundaries) ApplyBoundaryMask(out);
 }
 
 template<class Impl, class CloverHelpers>
@@ -186,7 +186,7 @@ void CompactWilsonCloverFermion<Impl, CloverHelpers>::MdirAll(const FermionField
 
 template<class Impl, class CloverHelpers>
 void CompactWilsonCloverFermion<Impl, CloverHelpers>::MDeriv(GaugeField& force, const FermionField& X, const FermionField& Y, int dag) {
-  assert(!open_boundaries); // TODO check for changes required for open bc
+  assert(!fixedBoundaries); // TODO check for changes required for open bc
 
   // NOTE: code copied from original clover term
   conformable(X.Grid(), Y.Grid());
@@ -338,14 +338,14 @@ void CompactWilsonCloverFermion<Impl, CloverHelpers>::ImportGauge(const GaugeFie
 
   // Modify the clover term at the temporal boundaries in case of open boundary conditions
   double t6 = usecond();
-  if(open_boundaries) CompactHelpers::ModifyBoundaries(Diagonal, Triangle, csw_t, cF, this->diag_mass);
+  if(fixedBoundaries) CompactHelpers::ModifyBoundaries(Diagonal, Triangle, csw_t, cF, this->diag_mass);
 
   // Invert the Clover term
   // In case of the exponential clover with (anti-)periodic boundary conditions exp(-Clover) saved
   // in TmpInverse can be used. In all other cases the clover term has to be explictly inverted.
   // TODO: For now this inversion is explictly done on the CPU
   double t7 = usecond();
-  CloverHelpers::InvertClover(TmpInverse, Diagonal, Triangle, DiagonalInv, TriangleInv, open_boundaries);
+  CloverHelpers::InvertClover(TmpInverse, Diagonal, Triangle, DiagonalInv, TriangleInv, fixedBoundaries);
 
   // Fill the remaining clover fields
   double t8 = usecond();

From 132d841b05217bf5e8efca1ef378c8505b700773 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Wed, 2 Nov 2022 19:33:22 -0400
Subject: [PATCH 057/146] Compile fix

---
 Grid/tensors/Tensor_extract_merge.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/Grid/tensors/Tensor_extract_merge.h b/Grid/tensors/Tensor_extract_merge.h
index c63a2439..87572faf 100644
--- a/Grid/tensors/Tensor_extract_merge.h
+++ b/Grid/tensors/Tensor_extract_merge.h
@@ -130,7 +130,7 @@ typename vobj::scalar_object extractLane(int lane, const vobj & __restrict__ vec
   typedef typename vobj::scalar_object scalar_object;
   typedef typename vobj::vector_type   vector_type;
   typedef typename ExtractTypeMap<scalar_type>::extract_type extract_type;
-  typedef extract_type * pointer;
+  typedef scalar_type * pointer;
 
   constexpr int words=sizeof(vobj)/sizeof(vector_type);
   constexpr int Nsimd=vector_type::Nsimd();
@@ -150,7 +150,7 @@ void insertLane(int lane, vobj & __restrict__ vec,const typename vobj::scalar_ob
   typedef typename vobj::vector_type vector_type;
   typedef typename vector_type::scalar_type scalar_type;
   typedef typename ExtractTypeMap<scalar_type>::extract_type extract_type;
-  typedef extract_type * pointer;
+  typedef scalar_type * pointer;
 
   constexpr int words=sizeof(vobj)/sizeof(vector_type);
   constexpr int Nsimd=vector_type::Nsimd();

From eae1c02111a03026dcfceb6ef8e8b6cf2a08b334 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Wed, 2 Nov 2022 19:50:32 -0400
Subject: [PATCH 058/146] Bounds check

---
 HMC/Mobius2p1f_DD_RHMC_96I.cc | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/HMC/Mobius2p1f_DD_RHMC_96I.cc b/HMC/Mobius2p1f_DD_RHMC_96I.cc
index 7ba5f5b5..967a3fc9 100644
--- a/HMC/Mobius2p1f_DD_RHMC_96I.cc
+++ b/HMC/Mobius2p1f_DD_RHMC_96I.cc
@@ -299,6 +299,7 @@ int main(int argc, char **argv) {
     //    MDCG(SdagS,vec,res);
 
 
+    vec = 1; // Fill with any old junk
     std::cout << "Bounds check on light quark operator mass "<< Denominators[0]->Mass() <<std::endl;
     SchurDifferentiableOperator<FermionImplPolicy> UdagU(*Denominators[0]);
     HighBoundCheck(UdagU,vec,OFRp.hi);
@@ -308,14 +309,16 @@ int main(int argc, char **argv) {
     //    MDCG(UdagU,vec,res);
 
 
+    vec = 1; // Fill with any old junk
     std::cout << "Bounds check on strange dirichlet operator mass "<< StrangeOpDir.Mass()<<std::endl;
     SchurDifferentiableOperator<FermionImplPolicy> SddagSd(StrangeOpDir);
-    HighBoundCheck(SddagSd,vec,OFRp.hi);
-    ChebyBoundsCheck(SddagSd,vec,OFRp.lo,OFRp.hi);
+    HighBoundCheck(SddagSd,vec,SFRp.hi);
+    ChebyBoundsCheck(SddagSd,vec,SFRp.lo,SFRp.hi);
     std::cout << "strange dirichlet inversion"<<std::endl;
     res=Zero();
     //    MDCG(SddagSd,vec,res);
 
+    vec = 1; // Fill with any old junk
     std::cout << "Bounds check on light dirichlet operator mass "<< Numerators[0]->Mass()<<std::endl;
     SchurDifferentiableOperator<FermionImplPolicy> UddagUd(*Numerators[0]);
     HighBoundCheck(UddagUd,vec,OFRp.hi);
@@ -337,7 +340,7 @@ int main(int argc, char **argv) {
     const int MaxIt= 10000;
     int Nconv;
     RealD resid = 1.0e-5;
-    if(0)
+    if(1)
     {
       int order = 501;
       RealD bound = 5.0e-4;

From d9dd9a5b5f3f7eb797acc4e9ea8f1e20c2e4b843 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Wed, 2 Nov 2022 19:51:50 -0400
Subject: [PATCH 059/146] LLVM update

---
 systems/Crusher/sourceme.sh | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/systems/Crusher/sourceme.sh b/systems/Crusher/sourceme.sh
index 3cccb10a..42e15fb1 100644
--- a/systems/Crusher/sourceme.sh
+++ b/systems/Crusher/sourceme.sh
@@ -2,7 +2,7 @@ export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/sw/crusher/spack-envs/base/opt/cray-sle
 module load emacs 
 #module load gperftools
 module load PrgEnv-gnu
-module load rocm/5.2.0
+module load rocm/5.3.0
 module load cray-mpich/8.1.16
 #module load cray-mpich/8.1.17
 module load gmp

From a3927a8a276ff499bd9d9a3b964e80a7eb16b646 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Wed, 2 Nov 2022 20:22:27 -0400
Subject: [PATCH 060/146] Dirichlet

---
 tests/Test_cayley_even_odd_vec.cc | 20 ++++++--------------
 1 file changed, 6 insertions(+), 14 deletions(-)

diff --git a/tests/Test_cayley_even_odd_vec.cc b/tests/Test_cayley_even_odd_vec.cc
index 9524fea0..e25d6fde 100644
--- a/tests/Test_cayley_even_odd_vec.cc
+++ b/tests/Test_cayley_even_odd_vec.cc
@@ -29,19 +29,6 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
- ;
-
-template<class d>
-struct scal {
-  d internal;
-};
-
-    Gamma::Algebra Gmu [] = {
-        Gamma::Algebra::GammaX,
-        Gamma::Algebra::GammaY,
-        Gamma::Algebra::GammaZ,
-        Gamma::Algebra::GammaT
-    };
 
 template<class What> 
 void  TestWhat(What & Ddwf,
@@ -88,7 +75,12 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage<<"**************************************************************"<<std::endl;
   std::cout<<GridLogMessage <<"DomainWallFermion vectorised test"<<std::endl;
   std::cout<<GridLogMessage<<"**************************************************************"<<std::endl;
-  DomainWallFermionD Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+  std::vector<Complex> boundary = {1,1,1,-1};
+  DomainWallFermionD::ImplParams Params(boundary);
+  Coordinate Dirichlet({0,8,8,16,32});
+  Params.dirichlet=Dirichlet;
+
+  DomainWallFermionD Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,Params);
   TestWhat<DomainWallFermionD>(Ddwf,FGrid,FrbGrid,UGrid,mass,M5,&RNG4,&RNG5);
 
   RealD b=1.5;// Scale factor b+c=2, b-c=1

From 78acae9b50c416fe26452cee818f3cda10752e0d Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Wed, 2 Nov 2022 20:24:17 -0400
Subject: [PATCH 061/146] Simple DWF for easy check

---
 Grid/qcd/action/fermion/DWFSlow.h | 291 ++++++++++++++++++++++++++++++
 1 file changed, 291 insertions(+)
 create mode 100644 Grid/qcd/action/fermion/DWFSlow.h

diff --git a/Grid/qcd/action/fermion/DWFSlow.h b/Grid/qcd/action/fermion/DWFSlow.h
new file mode 100644
index 00000000..61298504
--- /dev/null
+++ b/Grid/qcd/action/fermion/DWFSlow.h
@@ -0,0 +1,291 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/DWFSlow.h
+
+Copyright (C) 2022
+
+Author: Peter Boyle <pboyle@bnl.gov>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+			   /*  END LEGAL */
+#pragma once
+
+NAMESPACE_BEGIN(Grid);
+
+template <class Impl>
+class DWFSlowFermion : public FermionOperator<Impl>
+{
+public:
+  INHERIT_IMPL_TYPES(Impl);
+
+  ///////////////////////////////////////////////////////////////
+  // Implement the abstract base
+  ///////////////////////////////////////////////////////////////
+  GridBase *GaugeGrid(void) { return _grid4; }
+  GridBase *GaugeRedBlackGrid(void) { return _cbgrid4; }
+  GridBase *FermionGrid(void) { return _grid; }
+  GridBase *FermionRedBlackGrid(void) { return _cbgrid; }
+
+  FermionField _tmp;
+  FermionField &tmp(void) { return _tmp; }
+
+  //////////////////////////////////////////////////////////////////
+  // override multiply; cut number routines if pass dagger argument
+  // and also make interface more uniformly consistent
+  //////////////////////////////////////////////////////////////////
+  virtual void  M(const FermionField &in, FermionField &out)
+  {
+    FermionField tmp(_grid);
+    out = (5.0 - M5) * in;
+    Dhop(in,tmp,DaggerNo);
+    out = out + tmp;
+  }
+  virtual void  Mdag(const FermionField &in, FermionField &out)
+  {
+    FermionField tmp(_grid);
+    out = (5.0 - M5) * in;
+    Dhop(in,tmp,DaggerYes);
+    out = out + tmp;
+  };
+
+  /////////////////////////////////////////////////////////
+  // half checkerboard operations 5D redblack so just site identiy
+  /////////////////////////////////////////////////////////
+  void Meooe(const FermionField &in, FermionField &out)
+  {
+    if ( in.Checkerboard() == Odd ) {
+      this->DhopEO(in,out,DaggerNo);
+    } else {
+      this->DhopOE(in,out,DaggerNo);
+    }
+  }
+  void MeooeDag(const FermionField &in, FermionField &out)
+  {
+    if ( in.Checkerboard() == Odd ) {
+      this->DhopEO(in,out,DaggerYes);
+    } else {
+      this->DhopOE(in,out,DaggerYes);
+    }
+  };
+
+  // allow override for twisted mass and clover
+  virtual void Mooee(const FermionField &in, FermionField &out)
+  {
+    out = (5.0 - M5) * in;
+  }
+  virtual void MooeeDag(const FermionField &in, FermionField &out)
+  {
+    out = (5.0 - M5) * in;
+  }
+  virtual void MooeeInv(const FermionField &in, FermionField &out)
+  {
+    out = (1.0/(5.0 - M5)) * in;
+  };
+  virtual void MooeeInvDag(const FermionField &in, FermionField &out)
+  {
+    out = (1.0/(5.0 - M5)) * in;
+  };
+
+  virtual void  MomentumSpacePropagator(FermionField &out,const FermionField &in,RealD _mass,std::vector<double> twist) {} ;
+
+  ////////////////////////
+  // Derivative interface
+  ////////////////////////
+  // Interface calls an internal routine
+  void DhopDeriv(GaugeField &mat,const FermionField &U,const FermionField &V,int dag)  { assert(0);};
+  void DhopDerivOE(GaugeField &mat,const FermionField &U,const FermionField &V,int dag){ assert(0);};
+  void DhopDerivEO(GaugeField &mat,const FermionField &U,const FermionField &V,int dag){ assert(0);};
+
+  ///////////////////////////////////////////////////////////////
+  // non-hermitian hopping term; half cb or both
+  ///////////////////////////////////////////////////////////////
+  void Dhop(const FermionField &in, FermionField &out, int dag)
+  {
+    FermionField tmp(in.Grid());
+    Dhop5(in,out,MassField,MassField,dag );
+    for(int mu=0;mu<4;mu++){
+      DhopDirU(in,Umu[mu],Umu[mu],tmp,mu,dag );    out = out + tmp;
+    }
+  };
+  void DhopOE(const FermionField &in, FermionField &out, int dag)
+  {
+    FermionField tmp(in.Grid());
+    assert(in.Checkerboard()==Even);
+    Dhop5(in,out,MassFieldOdd,MassFieldEven,dag);
+    for(int mu=0;mu<4;mu++){
+      DhopDirU(in,UmuOdd[mu],UmuEven[mu],tmp,mu,dag );    out = out + tmp;
+    }
+  };
+  void DhopEO(const FermionField &in, FermionField &out, int dag)
+  {
+    FermionField tmp(in.Grid());
+    assert(in.Checkerboard()==Odd);
+    Dhop5(in,out, MassFieldEven,MassFieldOdd ,dag );  
+    for(int mu=0;mu<4;mu++){
+      DhopDirU(in,UmuEven[mu],UmuOdd[mu],tmp,mu,dag );    out = out + tmp;
+    }
+  };
+
+  ///////////////////////////////////////////////////////////////
+  // Multigrid assistance; force term uses too
+  ///////////////////////////////////////////////////////////////
+  void Mdir(const FermionField &in, FermionField &out, int dir, int disp){ assert(0);};
+  void MdirAll(const FermionField &in, std::vector<FermionField> &out)   { assert(0);};
+  void DhopDir(const FermionField &in, FermionField &out, int dir, int disp) { assert(0);};
+  void DhopDirAll(const FermionField &in, std::vector<FermionField> &out)    { assert(0);};
+  void DhopDirCalc(const FermionField &in, FermionField &out, int dirdisp,int gamma, int dag) { assert(0);};
+
+  void DhopDirU(const FermionField &in, const GaugeLinkField &U5e, const GaugeLinkField &U5o, FermionField &out, int mu, int dag)
+  {
+    RealD     sgn= 1.0;
+    if (dag ) sgn=-1.0;
+
+    Gamma::Algebra Gmu [] = {
+			 Gamma::Algebra::GammaX,
+			 Gamma::Algebra::GammaY,
+			 Gamma::Algebra::GammaZ,
+			 Gamma::Algebra::GammaT
+    };
+
+    //    mass is  1,1,1,1,-m has to multiply the round the world term
+    FermionField tmp (in.Grid());
+    tmp = U5e * Cshift(in,mu+1,1);
+    out = tmp - Gamma(Gmu[mu])*tmp*sgn;
+    
+    tmp = Cshift(adj(U5o)*in,mu+1,-1);
+    out = out + tmp + Gamma(Gmu[mu])*tmp*sgn;
+
+    out = -0.5*out;
+  };
+
+  void Dhop5(const FermionField &in, FermionField &out, ComplexField &massE, ComplexField &massO, int dag)
+  {
+    // Mass term.... must multiple the round world with mass = 1,1,1,1, -m
+    RealD     sgn= 1.0;
+    if (dag ) sgn=-1.0;
+
+    Gamma G5(Gamma::Algebra::Gamma5);
+
+    FermionField tmp (in.Grid());
+    tmp = massE*Cshift(in,0,1);
+    out = tmp - G5*tmp*sgn;
+    
+    tmp = Cshift(massO*in,0,-1);
+    out = out + tmp + G5*tmp*sgn;
+    out = -0.5*out;
+  };
+
+  // Constructor
+  DWFSlowFermion(GaugeField &_Umu, GridCartesian &Fgrid,
+		 GridRedBlackCartesian &Hgrid, RealD _mass, RealD _M5)
+    :
+    _grid(&Fgrid),
+    _cbgrid(&Hgrid),
+    _grid4(_Umu.Grid()),
+    Umu(Nd,&Fgrid),
+    UmuEven(Nd,&Hgrid),
+    UmuOdd(Nd,&Hgrid),
+    MassField(&Fgrid),
+    MassFieldEven(&Hgrid),
+    MassFieldOdd(&Hgrid),
+    M5(_M5),
+    mass(_mass),
+    _tmp(&Hgrid)
+    {
+      Ls=Fgrid._fdimensions[0];
+      ImportGauge(_Umu);
+
+      typedef typename FermionField::scalar_type scalar;
+
+      Lattice<iScalar<vInteger> > coor(&Fgrid);
+      LatticeCoordinate(coor, 0); // Scoor
+      ComplexField one(&Fgrid);
+      MassField =scalar(-mass);
+      one       =scalar(1.0);
+      MassField =where(coor==Integer(Ls-1),MassField,one);
+      for(int mu=0;mu<Nd;mu++){
+	pickCheckerboard(Even,UmuEven[mu],Umu[mu]);
+	pickCheckerboard(Odd ,UmuOdd[mu],Umu[mu]);
+      }
+      pickCheckerboard(Even,MassFieldEven,MassField);
+      pickCheckerboard(Odd ,MassFieldOdd,MassField);
+      
+    }
+  
+  // DoubleStore impl dependent
+  void ImportGauge(const GaugeField &_Umu4)
+  {
+    GaugeLinkField U4(_grid4);
+    for(int mu=0;mu<Nd;mu++){
+      U4 = PeekIndex<LorentzIndex>(_Umu4, mu);
+      for(int s=0;s<this->Ls;s++){
+	InsertSlice(U4,Umu[mu],s,0);
+      }
+    }
+  }
+
+  ///////////////////////////////////////////////////////////////
+  // Data members require to support the functionality
+  ///////////////////////////////////////////////////////////////
+
+public:
+  virtual RealD Mass(void) { return mass; }
+  virtual int   isTrivialEE(void) { return 1; };
+  RealD mass;
+  RealD M5;
+  int Ls;
+
+  GridBase *_grid4;
+  GridBase *_grid;
+  GridBase *_cbgrid4;
+  GridBase *_cbgrid;
+
+  // Copy of the gauge field , with even and odd subsets
+  std::vector<GaugeLinkField> Umu;
+  std::vector<GaugeLinkField> UmuEven;
+  std::vector<GaugeLinkField> UmuOdd;
+  ComplexField MassField;
+  ComplexField MassFieldEven;
+  ComplexField MassFieldOdd;
+
+  ///////////////////////////////////////////////////////////////
+  // Conserved current utilities
+  ///////////////////////////////////////////////////////////////
+  void ContractConservedCurrent(PropagatorField &q_in_1,
+                                PropagatorField &q_in_2,
+                                PropagatorField &q_out,
+                                PropagatorField &phys_src,
+                                Current curr_type,
+                                unsigned int mu){}
+  void SeqConservedCurrent(PropagatorField &q_in,
+                           PropagatorField &q_out,
+                           PropagatorField &phys_src,
+                           Current curr_type,
+                           unsigned int mu,
+                           unsigned int tmin,
+			   unsigned int tmax,
+			   ComplexField &lattice_cmplx){}
+};
+
+typedef DWFSlowFermion<WilsonImplF> DWFSlowFermionF;
+typedef DWFSlowFermion<WilsonImplD> DWFSlowFermionD;
+
+NAMESPACE_END(Grid);

From 006268f55624709aea0d353cbaf0d5875315b4b5 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Wed, 2 Nov 2022 20:24:51 -0400
Subject: [PATCH 062/146] DWF Slow version

---
 Grid/qcd/action/fermion/Fermion.h | 1 +
 1 file changed, 1 insertion(+)

diff --git a/Grid/qcd/action/fermion/Fermion.h b/Grid/qcd/action/fermion/Fermion.h
index 2f3fb9f2..bad736cf 100644
--- a/Grid/qcd/action/fermion/Fermion.h
+++ b/Grid/qcd/action/fermion/Fermion.h
@@ -47,6 +47,7 @@ Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
 ////////////////////////////////////////////
 // Fermion operators / actions
 ////////////////////////////////////////////
+#include <Grid/qcd/action/fermion/DWFSlow.h>       // Slow DWF
 
 #include <Grid/qcd/action/fermion/WilsonFermion.h>       // 4d wilson like
 NAMESPACE_CHECK(Wilson);

From 82e959f66cac74c3863eaa0b42f3c8b42ae874d8 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 8 Nov 2022 12:45:25 -0800
Subject: [PATCH 063/146] SYCL reduction

---
 Grid/lattice/Lattice_reduction.h | 22 ++++++++++++----------
 1 file changed, 12 insertions(+), 10 deletions(-)

diff --git a/Grid/lattice/Lattice_reduction.h b/Grid/lattice/Lattice_reduction.h
index 0ddac437..fb6a258c 100644
--- a/Grid/lattice/Lattice_reduction.h
+++ b/Grid/lattice/Lattice_reduction.h
@@ -28,6 +28,9 @@ Author: Christoph Lehner <christoph@lhnr.de>
 #if defined(GRID_CUDA)||defined(GRID_HIP)
 #include <Grid/lattice/Lattice_reduction_gpu.h>
 #endif
+#if defined(GRID_SYCL)
+#include <Grid/lattice/Lattice_reduction_sycl.h>
+#endif
 
 NAMESPACE_BEGIN(Grid);
 
@@ -127,7 +130,7 @@ inline Double max(const Double *arg, Integer osites)
 template<class vobj>
 inline typename vobj::scalar_object sum(const vobj *arg, Integer osites)
 {
-#if defined(GRID_CUDA)||defined(GRID_HIP)
+#if defined(GRID_CUDA)||defined(GRID_HIP)||defined(GRID_SYCL)
   return sum_gpu(arg,osites);
 #else
   return sum_cpu(arg,osites);
@@ -136,7 +139,7 @@ inline typename vobj::scalar_object sum(const vobj *arg, Integer osites)
 template<class vobj>
 inline typename vobj::scalar_objectD sumD(const vobj *arg, Integer osites)
 {
-#if defined(GRID_CUDA)||defined(GRID_HIP)
+#if defined(GRID_CUDA)||defined(GRID_HIP)||defined(GRID_SYCL)
   return sumD_gpu(arg,osites);
 #else
   return sumD_cpu(arg,osites);
@@ -145,7 +148,7 @@ inline typename vobj::scalar_objectD sumD(const vobj *arg, Integer osites)
 template<class vobj>
 inline typename vobj::scalar_objectD sumD_large(const vobj *arg, Integer osites)
 {
-#if defined(GRID_CUDA)||defined(GRID_HIP)
+#if defined(GRID_CUDA)||defined(GRID_HIP)||defined(GRID_SYCL)
   return sumD_gpu_large(arg,osites);
 #else
   return sumD_cpu(arg,osites);
@@ -155,13 +158,13 @@ inline typename vobj::scalar_objectD sumD_large(const vobj *arg, Integer osites)
 template<class vobj>
 inline typename vobj::scalar_object sum(const Lattice<vobj> &arg)
 {
-#if defined(GRID_CUDA)||defined(GRID_HIP)
-  autoView( arg_v, arg, AcceleratorRead);
   Integer osites = arg.Grid()->oSites();
-  auto ssum= sum_gpu(&arg_v[0],osites);
+#if defined(GRID_CUDA)||defined(GRID_HIP)||defined(GRID_SYCL)
+  typename vobj::scalar_object ssum;
+  autoView( arg_v, arg, AcceleratorRead);
+  ssum= sum_gpu(&arg_v[0],osites);
 #else
   autoView(arg_v, arg, CpuRead);
-  Integer osites = arg.Grid()->oSites();
   auto ssum= sum_cpu(&arg_v[0],osites);
 #endif  
   arg.Grid()->GlobalSum(ssum);
@@ -171,7 +174,7 @@ inline typename vobj::scalar_object sum(const Lattice<vobj> &arg)
 template<class vobj>
 inline typename vobj::scalar_object sum_large(const Lattice<vobj> &arg)
 {
-#if defined(GRID_CUDA)||defined(GRID_HIP)
+#if defined(GRID_CUDA)||defined(GRID_HIP)||defined(GRID_SYCL)
   autoView( arg_v, arg, AcceleratorRead);
   Integer osites = arg.Grid()->oSites();
   auto ssum= sum_gpu_large(&arg_v[0],osites);
@@ -235,11 +238,10 @@ inline ComplexD rankInnerProduct(const Lattice<vobj> &left,const Lattice<vobj> &
   typedef decltype(innerProductD(vobj(),vobj())) inner_t;
   Vector<inner_t> inner_tmp(sites);
   auto inner_tmp_v = &inner_tmp[0];
-    
   {
     autoView( left_v , left, AcceleratorRead);
     autoView( right_v,right, AcceleratorRead);
-
+    // This code could read coalesce
     // GPU - SIMT lane compliance...
     accelerator_for( ss, sites, 1,{
 	auto x_l = left_v[ss];

From bd68861b28b24ce85adcbcbe5a7c50b383446b3f Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 8 Nov 2022 12:49:26 -0800
Subject: [PATCH 064/146] SYCL sum

---
 Grid/lattice/Lattice_reduction_sycl.h | 125 ++++++++++++++++++++++++++
 1 file changed, 125 insertions(+)
 create mode 100644 Grid/lattice/Lattice_reduction_sycl.h

diff --git a/Grid/lattice/Lattice_reduction_sycl.h b/Grid/lattice/Lattice_reduction_sycl.h
new file mode 100644
index 00000000..90980c4c
--- /dev/null
+++ b/Grid/lattice/Lattice_reduction_sycl.h
@@ -0,0 +1,125 @@
+NAMESPACE_BEGIN(Grid);
+
+/////////////////////////////////////////////////////////////////////////////////////////////////////////
+// Possibly promote to double and sum
+/////////////////////////////////////////////////////////////////////////////////////////////////////////
+
+template <class vobj>
+inline typename vobj::scalar_objectD sumD_gpu_tensor(const vobj *lat, Integer osites) 
+{
+  typedef typename vobj::scalar_object sobj;
+  typedef typename vobj::scalar_objectD sobjD;
+  sobj *mysum =(sobj *) malloc_shared(sizeof(sobj),*theGridAccelerator);
+  sobj identity; zeroit(identity);
+  sobj ret ; 
+
+  Integer nsimd= vobj::Nsimd();
+  
+  theGridAccelerator->submit([&](cl::sycl::handler &cgh) {
+     auto Reduction = cl::sycl::reduction(mysum,identity,std::plus<>());
+     cgh.parallel_for(cl::sycl::range<1>{osites},
+		      Reduction,
+		      [=] (cl::sycl::id<1> item, auto &sum) {
+      auto osite   = item[0];
+      sum +=Reduce(lat[osite]);
+     });
+   });
+  theGridAccelerator->wait();
+  ret = mysum[0];
+  free(mysum,*theGridAccelerator);
+  sobjD dret; convertType(dret,ret);
+  return dret;
+}
+
+template <class vobj>
+inline typename vobj::scalar_objectD sumD_gpu_large(const vobj *lat, Integer osites)
+{
+  return sumD_gpu_tensor(lat,osites);
+}
+template <class vobj>
+inline typename vobj::scalar_objectD sumD_gpu_small(const vobj *lat, Integer osites)
+{
+  return sumD_gpu_large(lat,osites);
+}
+
+template <class vobj>
+inline typename vobj::scalar_objectD sumD_gpu(const vobj *lat, Integer osites)
+{
+  return sumD_gpu_large(lat,osites);
+}
+
+/////////////////////////////////////////////////////////////////////////////////////////////////////////
+// Return as same precision as input performing reduction in double precision though
+/////////////////////////////////////////////////////////////////////////////////////////////////////////
+template <class vobj>
+inline typename vobj::scalar_object sum_gpu(const vobj *lat, Integer osites) 
+{
+  typedef typename vobj::scalar_object sobj;
+  sobj result;
+  result = sumD_gpu(lat,osites);
+  return result;
+}
+
+template <class vobj>
+inline typename vobj::scalar_object sum_gpu_large(const vobj *lat, Integer osites)
+{
+  typedef typename vobj::scalar_object sobj;
+  sobj result;
+  result = sumD_gpu_large(lat,osites);
+  return result;
+}
+
+NAMESPACE_END(Grid);
+
+/*
+template<class Double> Double svm_reduce(Double *vec,uint64_t L)
+{
+  Double sumResult; zeroit(sumResult);
+  Double *d_sum =(Double *)cl::sycl::malloc_shared(sizeof(Double),*theGridAccelerator);
+  Double identity;  zeroit(identity);
+  theGridAccelerator->submit([&](cl::sycl::handler &cgh) {
+     auto Reduction = cl::sycl::reduction(d_sum,identity,std::plus<>());
+     cgh.parallel_for(cl::sycl::range<1>{L},
+		      Reduction,
+		      [=] (cl::sycl::id<1> index, auto &sum) {
+	 sum +=vec[index];
+     });
+   });
+  theGridAccelerator->wait();
+  Double ret = d_sum[0];
+  free(d_sum,*theGridAccelerator);
+  std::cout << " svm_reduce finished "<<L<<" sites sum = " << ret <<std::endl;
+  return ret;
+}
+
+template <class vobj>
+inline typename vobj::scalar_objectD sumD_gpu_repack(const vobj *lat, Integer osites)
+{
+  typedef typename vobj::vector_type  vector;
+  typedef typename vobj::scalar_type  scalar;
+
+  typedef typename vobj::scalar_typeD scalarD;
+  typedef typename vobj::scalar_objectD sobjD;
+
+  sobjD ret;
+  scalarD *ret_p = (scalarD *)&ret;
+  
+  const int nsimd = vobj::Nsimd();
+  const int words = sizeof(vobj)/sizeof(vector);
+
+  Vector<scalar> buffer(osites*nsimd);
+  scalar *buf = &buffer[0];
+  vector *dat = (vector *)lat;
+
+  for(int w=0;w<words;w++) {
+
+    accelerator_for(ss,osites,nsimd,{
+	int lane = acceleratorSIMTlane(nsimd);
+	buf[ss*nsimd+lane] = dat[ss*words+w].getlane(lane);
+    });
+    //Precision change at this point is to late to gain precision
+    ret_p[w] = svm_reduce(buf,nsimd*osites);
+  }
+  return ret;
+}
+*/

From 63fd1dfa62b86a2fae01f31ae6619c3ac6713119 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 8 Nov 2022 13:22:09 -0800
Subject: [PATCH 065/146] Config on PVC

---
 systems/PVC/config-command | 9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)

diff --git a/systems/PVC/config-command b/systems/PVC/config-command
index 3f5b5993..858edda7 100644
--- a/systems/PVC/config-command
+++ b/systems/PVC/config-command
@@ -2,14 +2,15 @@ INSTALL=/nfs/site/home/azusayax/install
 ../../configure \
 	--enable-simd=GPU \
 	--enable-gen-simd-width=64 \
-	--enable-comms=mpi \
+	--enable-comms=mpi-auto \
 	--disable-accelerator-cshift \
 	--disable-gparity \
 	--disable-fermion-reps \
 	--enable-shm=nvlink \
 	--enable-accelerator=sycl \
-	--enable-unified=yes \
-	CXX=mpicxx \
+	--enable-unified=no \
+	MPICXX=mpicxx \
+	CXX=dpcpp \
 	LDFLAGS="-fsycl-device-code-split=per_kernel -fsycl-device-lib=all -lze_loader -L$INSTALL/lib" \
-	CXXFLAGS="-cxx=dpcpp -fsycl-unnamed-lambda -fsycl -no-fma -I$INSTALL/include -Wtautological-constant-compare"
+	CXXFLAGS="-fsycl-unnamed-lambda -fsycl -no-fma -I$INSTALL/include -Wtautological-constant-compare"
 

From 187310136226348c0cd4cf507a79f76eab7fed0b Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 8 Nov 2022 13:22:45 -0800
Subject: [PATCH 066/146] PVC

---
 systems/PVC/benchmarks/run-1tile.sh | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
 mode change 100644 => 100755 systems/PVC/benchmarks/run-1tile.sh

diff --git a/systems/PVC/benchmarks/run-1tile.sh b/systems/PVC/benchmarks/run-1tile.sh
old mode 100644
new mode 100755
index 923afd84..0fe80247
--- a/systems/PVC/benchmarks/run-1tile.sh
+++ b/systems/PVC/benchmarks/run-1tile.sh
@@ -6,7 +6,7 @@
 
 source /nfs/site/home/paboylex/ATS/GridNew/Grid/systems/PVC-nightly/setup.sh
 
-export NT=16
+export NT=8
 
 export I_MPI_OFFLOAD=1
 export I_MPI_OFFLOAD_TOPOLIB=level_zero
@@ -21,7 +21,7 @@ export I_MPI_OFFLOAD_CELL=tile
 export EnableImplicitScaling=0
 export EnableWalkerPartition=0
 export ZE_AFFINITY_MASK=0.0
-mpiexec -launcher ssh -n 1 -host localhost  ./Benchmark_dwf_fp32 --mpi 1.1.1.1 --grid 32.32.32.32 --accelerator-threads $NT --comms-sequential --shm-mpi 1 --cacheblocking 8.8.8.8
+mpiexec -launcher ssh -n 1 -host localhost  ./Benchmark_dwf_fp32 --mpi 1.1.1.1 --grid 32.32.32.32 --accelerator-threads $NT --comms-sequential --shm-mpi 1 --device-mem 32768
 
 export ZE_AFFINITY_MASK=0
 export I_MPI_OFFLOAD_CELL=device

From 5c75aa50089bc53bc8a2a18dd16e8d6aab5041c4 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 8 Nov 2022 13:22:57 -0800
Subject: [PATCH 067/146] Device mem

---
 systems/PVC/benchmarks/run-2tile-mpi.sh | 17 +++++++++--------
 1 file changed, 9 insertions(+), 8 deletions(-)

diff --git a/systems/PVC/benchmarks/run-2tile-mpi.sh b/systems/PVC/benchmarks/run-2tile-mpi.sh
index 34deac2c..cefab776 100755
--- a/systems/PVC/benchmarks/run-2tile-mpi.sh
+++ b/systems/PVC/benchmarks/run-2tile-mpi.sh
@@ -8,7 +8,6 @@ source /nfs/site/home/paboylex/ATS/GridNew/Grid/systems/PVC-nightly/setup.sh
 
 export NT=16
 
-
 # export IGC_EnableLSCFenceUGMBeforeEOT=0
 # export SYCL_PROGRAM_COMPILE_OPTIONS="-ze-opt-large-register-file=False"
 #export IGC_ShaderDumpEnable=1 
@@ -20,14 +19,16 @@ export SYCL_DEVICE_FILTER=gpu,level_zero
 export I_MPI_OFFLOAD_CELL=tile
 export EnableImplicitScaling=0
 export EnableWalkerPartition=0
-#export SYCL_PI_LEVEL_ZERO_DEVICE_SCOPE_EVENTS=1
+export SYCL_PI_LEVEL_ZERO_DEVICE_SCOPE_EVENTS=1
 export SYCL_PI_LEVEL_ZERO_USE_IMMEDIATE_COMMANDLISTS=1
-#export SYCL_PI_LEVEL_ZERO_USE_COPY_ENGINE=0
+export SYCL_PI_LEVEL_ZERO_USE_COPY_ENGINE=0
 
-mpiexec -launcher ssh -n 2 -host localhost  ./wrap4gpu.sh ./Benchmark_dwf_fp32 --mpi 1.1.1.2 --grid 32.32.32.64 --accelerator-threads $NT  --shm-mpi 1 > dw.2tile.1x2.log
-mpiexec -launcher ssh -n 2 -host localhost  ./wrap4gpu.sh ./Benchmark_dwf_fp32 --mpi 2.1.1.1 --grid 64.32.32.32 --accelerator-threads $NT  --shm-mpi 1 > dw.2tile.2x1.log
-
-mpiexec -launcher ssh -n 2 -host localhost  ./wrap4gpu.sh ./Benchmark_halo --mpi 1.1.1.2 --grid 32.32.32.64 --accelerator-threads $NT  --shm-mpi 1 > halo.2tile.1x2.log
-mpiexec -launcher ssh -n 2 -host localhost  ./wrap4gpu.sh ./Benchmark_halo --mpi 2.1.1.1 --grid 64.32.32.32 --accelerator-threads $NT  --shm-mpi 1 > halo.2tile.2x1.log
+for i in 0 
+do
+mpiexec -launcher ssh -n 2 -host localhost  ./wrap4gpu.sh ./Benchmark_dwf_fp32 --mpi 1.1.1.2 --grid 32.32.32.64 --accelerator-threads $NT  --shm-mpi 1  --device-mem 32768
+mpiexec -launcher ssh -n 2 -host localhost  ./wrap4gpu.sh ./Benchmark_dwf_fp32 --mpi 2.1.1.1 --grid 64.32.32.32 --accelerator-threads $NT  --shm-mpi 1  --device-mem 32768
+done
+#mpiexec -launcher ssh -n 2 -host localhost  ./wrap4gpu.sh ./Benchmark_halo --mpi 1.1.1.2 --grid 32.32.32.64 --accelerator-threads $NT  --shm-mpi 1 > halo.2tile.1x2.log
+#mpiexec -launcher ssh -n 2 -host localhost  ./wrap4gpu.sh ./Benchmark_halo --mpi 2.1.1.1 --grid 64.32.32.32 --accelerator-threads $NT  --shm-mpi 1 > halo.2tile.2x1.log
 
 

From 0655dab46659e21dfb7eaa4a4c947296e315fa5e Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 8 Nov 2022 13:38:54 -0800
Subject: [PATCH 068/146] Open MP on host enabled

---
 systems/PVC/config-command | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/systems/PVC/config-command b/systems/PVC/config-command
index 858edda7..cd7bba1d 100644
--- a/systems/PVC/config-command
+++ b/systems/PVC/config-command
@@ -12,5 +12,5 @@ INSTALL=/nfs/site/home/azusayax/install
 	MPICXX=mpicxx \
 	CXX=dpcpp \
 	LDFLAGS="-fsycl-device-code-split=per_kernel -fsycl-device-lib=all -lze_loader -L$INSTALL/lib" \
-	CXXFLAGS="-fsycl-unnamed-lambda -fsycl -no-fma -I$INSTALL/include -Wtautological-constant-compare"
+	CXXFLAGS="-fsycl-unnamed-lambda -fsycl -no-fma -I$INSTALL/include -Wno-tautological-compare"
 

From a11c12e2e716cafb54b40fe45405e56144edea14 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 15 Nov 2022 16:20:01 -0500
Subject: [PATCH 069/146] Modifications for partial dirichlet BCs

---
 Grid/communicator/Communicator_base.h  |  2 +-
 Grid/communicator/Communicator_mpi3.cc | 18 +++++++-----------
 Grid/communicator/Communicator_none.cc |  2 +-
 3 files changed, 9 insertions(+), 13 deletions(-)

diff --git a/Grid/communicator/Communicator_base.h b/Grid/communicator/Communicator_base.h
index 68cd36cc..b98424a1 100644
--- a/Grid/communicator/Communicator_base.h
+++ b/Grid/communicator/Communicator_base.h
@@ -155,7 +155,7 @@ public:
 				    int xmit_to_rank,int do_xmit,
 				    void *recv,
 				    int recv_from_rank,int do_recv,
-				    int bytes,int dir);
+				    int xbytes,int rbytes,int dir);
   
   
   void StencilSendToRecvFromComplete(std::vector<CommsRequest_t> &waitall,int i);
diff --git a/Grid/communicator/Communicator_mpi3.cc b/Grid/communicator/Communicator_mpi3.cc
index cbdd224d..892e3dbe 100644
--- a/Grid/communicator/Communicator_mpi3.cc
+++ b/Grid/communicator/Communicator_mpi3.cc
@@ -343,7 +343,7 @@ double CartesianCommunicator::StencilSendToRecvFrom( void *xmit,
 						     int bytes,int dir)
 {
   std::vector<CommsRequest_t> list;
-  double offbytes = StencilSendToRecvFromBegin(list,xmit,dest,dox,recv,from,dor,bytes,dir);
+  double offbytes = StencilSendToRecvFromBegin(list,xmit,dest,dox,recv,from,dor,bytes,bytes,dir);
   StencilSendToRecvFromComplete(list,dir);
   return offbytes;
 }
@@ -353,7 +353,7 @@ double CartesianCommunicator::StencilSendToRecvFromBegin(std::vector<CommsReques
 							 int dest,int dox,
 							 void *recv,
 							 int from,int dor,
-							 int bytes,int dir)
+							 int xbytes,int rbytes,int dir)
 {
   int ncomm  =communicator_halo.size();
   int commdir=dir%ncomm;
@@ -375,31 +375,27 @@ double CartesianCommunicator::StencilSendToRecvFromBegin(std::vector<CommsReques
   if ( dor ) {
     if ( (gfrom ==MPI_UNDEFINED) || Stencil_force_mpi ) {
       tag= dir+from*32;
-      ierr=MPI_Irecv(recv, bytes, MPI_CHAR,from,tag,communicator_halo[commdir],&rrq);
+      ierr=MPI_Irecv(recv, rbytes, MPI_CHAR,from,tag,communicator_halo[commdir],&rrq);
       assert(ierr==0);
       list.push_back(rrq);
-      off_node_bytes+=bytes;
+      off_node_bytes+=rbytes;
     }
   }
   
   if (dox) {
     if ( (gdest == MPI_UNDEFINED) || Stencil_force_mpi ) {
       tag= dir+_processor*32;
-      ierr =MPI_Isend(xmit, bytes, MPI_CHAR,dest,tag,communicator_halo[commdir],&xrq);
+      ierr =MPI_Isend(xmit, xbytes, MPI_CHAR,dest,tag,communicator_halo[commdir],&xrq);
       assert(ierr==0);
       list.push_back(xrq);
-      off_node_bytes+=bytes;
+      off_node_bytes+=xbytes;
     } else {
       void *shm = (void *) this->ShmBufferTranslate(dest,recv);
       assert(shm!=NULL);
-      acceleratorCopyDeviceToDeviceAsynch(xmit,shm,bytes);
+      acceleratorCopyDeviceToDeviceAsynch(xmit,shm,xbytes);
     }
   }
 
-  //  if ( CommunicatorPolicy == CommunicatorPolicySequential ) {
-  //    this->StencilSendToRecvFromComplete(list,dir);
-  //  }
-
   return off_node_bytes;
 }
 void CartesianCommunicator::StencilSendToRecvFromComplete(std::vector<CommsRequest_t> &list,int dir)
diff --git a/Grid/communicator/Communicator_none.cc b/Grid/communicator/Communicator_none.cc
index 4b533c4b..5d917b79 100644
--- a/Grid/communicator/Communicator_none.cc
+++ b/Grid/communicator/Communicator_none.cc
@@ -126,7 +126,7 @@ double CartesianCommunicator::StencilSendToRecvFromBegin(std::vector<CommsReques
 							 int xmit_to_rank,int dox,
 							 void *recv,
 							 int recv_from_rank,int dor,
-							 int bytes, int dir)
+							 int xbytes,int rbytes, int dir)
 {
   return 2.0*bytes;
 }

From 0566fc6267ab60b97cc8cd22961171701717877c Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 15 Nov 2022 16:21:24 -0500
Subject: [PATCH 070/146] Partial Dirichlet

---
 Grid/qcd/action/ActionParams.h                |  11 +-
 Grid/qcd/action/fermion/WilsonCompressor.h    | 330 +++++++++++-------
 .../WilsonFermion5DImplementation.h           |  13 +-
 3 files changed, 223 insertions(+), 131 deletions(-)

diff --git a/Grid/qcd/action/ActionParams.h b/Grid/qcd/action/ActionParams.h
index 274ff318..b2a06280 100644
--- a/Grid/qcd/action/ActionParams.h
+++ b/Grid/qcd/action/ActionParams.h
@@ -39,29 +39,38 @@ struct GparityWilsonImplParams {
   Coordinate twists;
                      //mu=Nd-1 is assumed to be the time direction and a twist value of 1 indicates antiperiodic BCs
   Coordinate dirichlet; // Blocksize of dirichlet BCs
-  GparityWilsonImplParams() : twists(Nd, 0) { dirichlet.resize(0); };
+  int  partialDirichlet;
+  GparityWilsonImplParams() : twists(Nd, 0) {
+    dirichlet.resize(0);
+    partialDirichlet=0;
+  };
 };
   
 struct WilsonImplParams {
   bool overlapCommsCompute;
   Coordinate dirichlet; // Blocksize of dirichlet BCs
+  int  partialDirichlet;
   AcceleratorVector<Real,Nd> twist_n_2pi_L;
   AcceleratorVector<Complex,Nd> boundary_phases;
   WilsonImplParams()  {
     dirichlet.resize(0);
+    partialDirichlet=0;
     boundary_phases.resize(Nd, 1.0);
       twist_n_2pi_L.resize(Nd, 0.0);
   };
   WilsonImplParams(const AcceleratorVector<Complex,Nd> phi) : boundary_phases(phi), overlapCommsCompute(false) {
     twist_n_2pi_L.resize(Nd, 0.0);
+    partialDirichlet=0;
     dirichlet.resize(0);
   }
 };
 
 struct StaggeredImplParams {
   Coordinate dirichlet; // Blocksize of dirichlet BCs
+  int  partialDirichlet;
   StaggeredImplParams()
   {
+    partialDirichlet=0;
     dirichlet.resize(0);
   };
 };
diff --git a/Grid/qcd/action/fermion/WilsonCompressor.h b/Grid/qcd/action/fermion/WilsonCompressor.h
index de2f1979..e2ced552 100644
--- a/Grid/qcd/action/fermion/WilsonCompressor.h
+++ b/Grid/qcd/action/fermion/WilsonCompressor.h
@@ -32,17 +32,182 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 NAMESPACE_BEGIN(Grid);
 
+///////////////////////////////////////////////////////////////
+// Wilson compressor will need FaceGather policies for:
+// Periodic, Dirichlet, and partial Dirichlet for DWF
+///////////////////////////////////////////////////////////////
+class FaceGatherPartialDWF
+{
+public:
+  static int PartialCompressionFactor(GridBase *grid) {return grid->_fdimensions[0]/2;};
+  //  static int PartialCompressionFactor(GridBase *grid) { return 1;}
+  template<class vobj,class cobj,class compressor>
+  static void Gather_plane_simple (commVector<std::pair<int,int> >& table,
+				   const Lattice<vobj> &rhs,
+				   cobj *buffer,
+				   compressor &compress,
+				   int off,int so,int partial)
+  {
+    //DWF only hack: If a direction that is OFF node we use Partial Dirichlet
+    //  Shrinks local and remote comms buffers
+    GridBase *Grid = rhs.Grid();
+    int Ls = Grid->_rdimensions[0];
+    std::pair<int,int> *table_v = & table[0];
+    auto rhs_v = rhs.View(AcceleratorRead);
+    int vol=table.size()/Ls;
+    accelerator_forNB( idx,table.size(), vobj::Nsimd(), {
+	Integer i=idx/Ls;
+	Integer s=idx%Ls;
+	if(s==0)    compress.Compress(buffer[off+i    ],rhs_v[so+table_v[idx].second]);
+	if(s==Ls-1) compress.Compress(buffer[off+i+vol],rhs_v[so+table_v[idx].second]);
+    });
+  }
+  template<class decompressor,class Decompression>
+  static void DecompressFace(decompressor decompress,Decompression &dd)
+  {
+    auto Ls = dd.dims[0];
+    // Just pass in the Grid
+    auto kp = dd.kernel_p;
+    auto mp = dd.mpi_p;
+    int size= dd.buffer_size;
+    int vol= size/Ls;
+    accelerator_forNB(o,size,1,{
+	int idx=o/Ls;
+	int   s=o%Ls;
+	if ( s == 0 ) {
+	  int oo=idx;
+	  kp[o]=mp[oo];
+	} else if ( s == Ls-1 ) {
+	  int oo=vol+idx; 
+	  kp[o]=mp[oo];
+	} else {
+	  kp[o] = Zero();//fill rest with zero if partial dirichlet
+	}
+    });
+  }
+  ////////////////////////////////////////////////////////////////////////////////////////////
+  // Need to gather *interior portions* for ALL s-slices in simd directions
+  // Do the gather as need to treat SIMD lanes differently, and insert zeroes on receive side
+  // Reorder the fifth dim to be s=Ls-1 , s=0, s=1,...,Ls-2.
+  ////////////////////////////////////////////////////////////////////////////////////////////
+  template<class vobj,class cobj,class compressor>
+  static void Gather_plane_exchange(commVector<std::pair<int,int> >& table,const Lattice<vobj> &rhs,
+				    Vector<cobj *> pointers,int dimension,int plane,int cbmask,
+				    compressor &compress,int type,int partial)
+  {
+    GridBase *Grid = rhs.Grid();
+    int Ls = Grid->_rdimensions[0];
+
+    // insertion of zeroes...
+    assert( (table.size()&0x1)==0);
+    int num=table.size()/2;
+    int so  = plane*rhs.Grid()->_ostride[dimension]; // base offset for start of plane
+    
+    auto rhs_v = rhs.View(AcceleratorRead);
+    auto p0=&pointers[0][0];
+    auto p1=&pointers[1][0];
+    auto tp=&table[0];
+    int nnum=num/Ls;
+    accelerator_forNB(j, num, vobj::Nsimd(), {
+	//  Reorders both local and remote comms buffers
+	//  
+	int s  = j % Ls;
+	int sp1 = (s+1)%Ls;  // peri incremented s slice
+	
+	int hxyz= j/Ls;
+
+	int xyz0= hxyz*2; // xyzt part of coor
+	int xyz1= hxyz*2+1;
+	
+	int jj= hxyz + sp1*nnum ; // 0,1,2,3 -> Ls-1 slice , 0-slice, 1-slice ....
+	
+	int kk0= xyz0*Ls + s ; // s=0 goes to s=1
+	int kk1= xyz1*Ls + s ; // s=Ls-1 -> s=0
+	compress.CompressExchange(p0[jj],p1[jj],
+				  rhs_v[so+tp[kk0 ].second], // Same s, consecutive xyz sites
+				  rhs_v[so+tp[kk1 ].second], 
+				  type);
+    });
+    rhs_v.ViewClose();
+  }
+  // Merge routine is for SIMD faces
+  template<class decompressor,class Merger>
+  static void MergeFace(decompressor decompress,Merger &mm)
+  {
+    auto Ls = mm.dims[0];
+    int  num= mm.buffer_size/2; // relate vol and Ls to buffer size
+    auto mp = &mm.mpointer[0];
+    auto vp0= &mm.vpointers[0][0]; // First arg is exchange first
+    auto vp1= &mm.vpointers[1][0];
+    auto type= mm.type;
+    int nnum = num/Ls;
+    accelerator_forNB(o,num,vobj::Nsimd(),{
+
+	int  s=o%Ls;
+	int hxyz=o/Ls; // xyzt related component
+	int xyz0=hxyz*2;
+	int xyz1=hxyz*2+1;
+
+	int sp = (s+1)%Ls; 
+	int jj= hxyz + sp*nnum ; // 0,1,2,3 -> Ls-1 slice , 0-slice, 1-slice ....
+
+	int oo0= s+xyz0*Ls;
+	int oo1= s+xyz1*Ls;
+
+	// same ss0, ss1 pair goes to new layout
+	decompress.Exchange(mp[oo0],mp[oo1],vp0[jj],vp1[jj],type);
+      });
+  }
+};
+class FaceGatherDWFMixedBCs
+{
+public:
+  static int PartialCompressionFactor(GridBase *grid) {return grid->_fdimensions[0]/2;};
+  
+  template<class vobj,class cobj,class compressor>
+  static void Gather_plane_simple (commVector<std::pair<int,int> >& table,
+					 const Lattice<vobj> &rhs,
+					 cobj *buffer,
+					 compressor &compress,
+					 int off,int so,int partial)
+  {
+    if(partial) FaceGatherPartialDWF::Gather_plane_simple(table,rhs,buffer,compress,off,so,partial);
+    else        FaceGatherSimple::Gather_plane_simple(table,rhs,buffer,compress,off,so,partial);
+  }
+  template<class vobj,class cobj,class compressor>
+  static void Gather_plane_exchange(commVector<std::pair<int,int> >& table,const Lattice<vobj> &rhs,
+				    Vector<cobj *> pointers,int dimension,int plane,int cbmask,
+				    compressor &compress,int type,int partial)
+  {
+    if(partial) FaceGatherPartialDWF::Gather_plane_exchange(table,rhs,pointers,dimension, plane,cbmask,compress,type,partial);
+    else        FaceGatherSimple::Gather_plane_exchange    (table,rhs,pointers,dimension, plane,cbmask,compress,type,partial);
+  }
+  template<class decompressor,class Merger>
+  static void MergeFace(decompressor decompress,Merger &mm)
+  {
+    int partial = mm.partial;
+    if ( partial ) FaceGatherPartialDWF::MergeFace(decompress,mm);
+    else           FaceGatherSimple::MergeFace(decompress,mm);
+  }
+
+  template<class decompressor,class Decompression>
+  static void DecompressFace(decompressor decompress,Decompression &dd)
+  {
+    int partial = dd.partial;
+    if ( partial ) FaceGatherPartialDWF::DecompressFace(decompress,dd);
+    else           FaceGatherSimple::DecompressFace(decompress,dd);
+  }
+};
+
 /////////////////////////////////////////////////////////////////////////////////////////////
-// optimised versions supporting half precision too
+// optimised versions supporting half precision too??? Deprecate
 /////////////////////////////////////////////////////////////////////////////////////////////
 
-template<class _HCspinor,class _Hspinor,class _Spinor, class projector,typename SFINAE = void >
-class WilsonCompressorTemplate;
-
 
+//Could make FaceGather a template param, but then behaviour is runtime not compile time
 template<class _HCspinor,class _Hspinor,class _Spinor, class projector>
-class WilsonCompressorTemplate< _HCspinor, _Hspinor, _Spinor, projector,
-				typename std::enable_if<std::is_same<_HCspinor,_Hspinor>::value>::type >
+class WilsonCompressorTemplate  : public FaceGatherDWFMixedBCs
+//  : public FaceGatherSimple
 {
 public:
   
@@ -79,172 +244,81 @@ public:
   /*****************************************************/
   /* Exchange includes precision change if mpi data is not same */
   /*****************************************************/
-  accelerator_inline void Exchange(SiteHalfSpinor *mp,
-				   const SiteHalfSpinor * __restrict__ vp0,
-				   const SiteHalfSpinor * __restrict__ vp1,
-				   Integer type,Integer o) const {
+  accelerator_inline void Exchange(SiteHalfSpinor &mp0,
+				   SiteHalfSpinor &mp1,
+				   const SiteHalfSpinor & vp0,
+				   const SiteHalfSpinor & vp1,
+				   Integer type) const {
 #ifdef GRID_SIMT
-    exchangeSIMT(mp[2*o],mp[2*o+1],vp0[o],vp1[o],type);
+    exchangeSIMT(mp0,mp1,vp0,vp1,type);
 #else
     SiteHalfSpinor tmp1;
     SiteHalfSpinor tmp2;
-    exchange(tmp1,tmp2,vp0[o],vp1[o],type);
-    vstream(mp[2*o  ],tmp1);
-    vstream(mp[2*o+1],tmp2);
+    exchange(tmp1,tmp2,vp0,vp1,type);
+    vstream(mp0,tmp1);
+    vstream(mp1,tmp2);
 #endif
   }
-
+  
 
   /*****************************************************/
   /* Have a decompression step if mpi data is not same */
   /*****************************************************/
-  accelerator_inline void Decompress(SiteHalfSpinor * __restrict__ out,
-				     SiteHalfSpinor * __restrict__ in, Integer o) const {    
-    assert(0);
+  accelerator_inline void Decompress(SiteHalfSpinor &out,
+				     SiteHalfSpinor &in) const {    
+    out = in;
   }
 
   /*****************************************************/
   /* Compress Exchange                                 */
   /*****************************************************/
-  accelerator_inline void CompressExchange(SiteHalfSpinor * __restrict__ out0,
-					   SiteHalfSpinor * __restrict__ out1,
-					   const SiteSpinor * __restrict__ in,
-					   Integer j,Integer k, Integer m,Integer type) const
+  accelerator_inline void CompressExchange(SiteHalfSpinor &out0,
+					   SiteHalfSpinor &out1,
+					   const SiteSpinor &in0,
+					   const SiteSpinor &in1,
+					   Integer type) const
   {
 #ifdef GRID_SIMT
     typedef SiteSpinor vobj;
     typedef SiteHalfSpinor hvobj;
-    typedef decltype(coalescedRead(*in))    sobj;
-    typedef decltype(coalescedRead(*out0)) hsobj;
+    typedef decltype(coalescedRead(in0))    sobj;
+    typedef decltype(coalescedRead(out0)) hsobj;
 
     unsigned int Nsimd = vobj::Nsimd();
     unsigned int mask = Nsimd >> (type + 1);
     int lane = acceleratorSIMTlane(Nsimd);
     int j0 = lane &(~mask); // inner coor zero
     int j1 = lane |(mask) ; // inner coor one
-    const vobj *vp0 = &in[k];
-    const vobj *vp1 = &in[m];
+    const vobj *vp0 = &in0;
+    const vobj *vp1 = &in1;
     const vobj *vp = (lane&mask) ? vp1:vp0;
     auto sa = coalescedRead(*vp,j0);
     auto sb = coalescedRead(*vp,j1);
     hsobj psa, psb;
     projector::Proj(psa,sa,mu,dag);
     projector::Proj(psb,sb,mu,dag);
-    coalescedWrite(out0[j],psa);
-    coalescedWrite(out1[j],psb);
+    coalescedWrite(out0,psa);
+    coalescedWrite(out1,psb);
 #else
     SiteHalfSpinor temp1, temp2;
     SiteHalfSpinor temp3, temp4;
-    projector::Proj(temp1,in[k],mu,dag);
-    projector::Proj(temp2,in[m],mu,dag);
+    projector::Proj(temp1,in0,mu,dag);
+    projector::Proj(temp2,in1,mu,dag);
     exchange(temp3,temp4,temp1,temp2,type);
-    vstream(out0[j],temp3);
-    vstream(out1[j],temp4);
+    vstream(out0,temp3);
+    vstream(out1,temp4);
 #endif
   }
 
   /*****************************************************/
   /* Pass the info to the stencil */
   /*****************************************************/
-  accelerator_inline bool DecompressionStep(void) const { return false; }
+  accelerator_inline bool DecompressionStep(void) const {
+    return false;
+  }
 
 };
 
-#if 0
-template<class _HCspinor,class _Hspinor,class _Spinor, class projector>
-class WilsonCompressorTemplate< _HCspinor, _Hspinor, _Spinor, projector,
-				typename std::enable_if<!std::is_same<_HCspinor,_Hspinor>::value>::type >
-{
-public:
-  
-  int mu,dag;  
-
-  void Point(int p) { mu=p; };
-
-  WilsonCompressorTemplate(int _dag=0){
-    dag = _dag;
-  }
-
-  typedef _Spinor         SiteSpinor;
-  typedef _Hspinor     SiteHalfSpinor;
-  typedef _HCspinor SiteHalfCommSpinor;
-  typedef typename SiteHalfCommSpinor::vector_type vComplexLow;
-  typedef typename SiteHalfSpinor::vector_type     vComplexHigh;
-  constexpr static int Nw=sizeof(SiteHalfSpinor)/sizeof(vComplexHigh);
-
-  accelerator_inline int CommDatumSize(void) const {
-    return sizeof(SiteHalfCommSpinor);
-  }
-
-  /*****************************************************/
-  /* Compress includes precision change if mpi data is not same */
-  /*****************************************************/
-  accelerator_inline void Compress(SiteHalfSpinor &buf,const SiteSpinor &in) const {
-    SiteHalfSpinor hsp;
-    SiteHalfCommSpinor *hbuf = (SiteHalfCommSpinor *)buf;
-    projector::Proj(hsp,in,mu,dag);
-    precisionChange((vComplexLow *)&hbuf[o],(vComplexHigh *)&hsp,Nw);
-  }
-  accelerator_inline void Compress(SiteHalfSpinor &buf,const SiteSpinor &in) const {
-#ifdef GRID_SIMT
-    typedef decltype(coalescedRead(buf)) sobj;
-    sobj sp;
-    auto sin = coalescedRead(in);
-    projector::Proj(sp,sin,mu,dag);
-    coalescedWrite(buf,sp);
-#else
-    projector::Proj(buf,in,mu,dag);
-#endif
-  }
-
-  /*****************************************************/
-  /* Exchange includes precision change if mpi data is not same */
-  /*****************************************************/
-  accelerator_inline void Exchange(SiteHalfSpinor *mp,
-                       SiteHalfSpinor *vp0,
-                       SiteHalfSpinor *vp1,
-		       Integer type,Integer o) const {
-    SiteHalfSpinor vt0,vt1;
-    SiteHalfCommSpinor *vpp0 = (SiteHalfCommSpinor *)vp0;
-    SiteHalfCommSpinor *vpp1 = (SiteHalfCommSpinor *)vp1;
-    precisionChange((vComplexHigh *)&vt0,(vComplexLow *)&vpp0[o],Nw);
-    precisionChange((vComplexHigh *)&vt1,(vComplexLow *)&vpp1[o],Nw);
-    exchange(mp[2*o],mp[2*o+1],vt0,vt1,type);
-  }
-
-  /*****************************************************/
-  /* Have a decompression step if mpi data is not same */
-  /*****************************************************/
-  accelerator_inline void Decompress(SiteHalfSpinor *out, SiteHalfSpinor *in, Integer o) const {
-    SiteHalfCommSpinor *hin=(SiteHalfCommSpinor *)in;
-    precisionChange((vComplexHigh *)&out[o],(vComplexLow *)&hin[o],Nw);
-  }
-
-  /*****************************************************/
-  /* Compress Exchange                                 */
-  /*****************************************************/
-  accelerator_inline void CompressExchange(SiteHalfSpinor *out0,
-			       SiteHalfSpinor *out1,
-			       const SiteSpinor *in,
-			       Integer j,Integer k, Integer m,Integer type) const {
-    SiteHalfSpinor temp1, temp2,temp3,temp4;
-    SiteHalfCommSpinor *hout0 = (SiteHalfCommSpinor *)out0;
-    SiteHalfCommSpinor *hout1 = (SiteHalfCommSpinor *)out1;
-    projector::Proj(temp1,in[k],mu,dag);
-    projector::Proj(temp2,in[m],mu,dag);
-    exchange(temp3,temp4,temp1,temp2,type);
-    precisionChange((vComplexLow *)&hout0[j],(vComplexHigh *)&temp3,Nw);
-    precisionChange((vComplexLow *)&hout1[j],(vComplexHigh *)&temp4,Nw);
-  }
-
-  /*****************************************************/
-  /* Pass the info to the stencil */
-  /*****************************************************/
-  accelerator_inline bool DecompressionStep(void) const { return true; }
-
-};
-#endif
-
 #define DECLARE_PROJ(Projector,Compressor,spProj)			\
   class Projector {							\
   public:								\
diff --git a/Grid/qcd/action/fermion/implementation/WilsonFermion5DImplementation.h b/Grid/qcd/action/fermion/implementation/WilsonFermion5DImplementation.h
index 388094b2..992f4d20 100644
--- a/Grid/qcd/action/fermion/implementation/WilsonFermion5DImplementation.h
+++ b/Grid/qcd/action/fermion/implementation/WilsonFermion5DImplementation.h
@@ -148,12 +148,21 @@ void WilsonFermion5D<Impl>::ImportGauge(const GaugeField &_Umu)
   GaugeField HUmu(_Umu.Grid());
   HUmu = _Umu*(-0.5);
   if ( Dirichlet ) {
-    std::cout << GridLogDslash << " Dirichlet BCs 5d " <<Block<<std::endl;
+    std:: cout << GridLogMessage << "Checking block size multiple of rank boundaries for Dirichlet"<<std::endl;
+    for(int d=0;d<Nd;d++) {
+      int GaugeBlock = Block[d+1];
+      int ldim=GaugeGrid()->LocalDimensions()[d];
+      if (GaugeBlock) assert( (GaugeBlock%ldim)==0);
+    }
+  }
+  if ( Dirichlet && (!this->Params.partialDirichlet) ) {
+    std::cout << GridLogMessage << " Dirichlet filtering gauge field BCs block " <<Block<<std::endl;
     Coordinate GaugeBlock(Nd);
     for(int d=0;d<Nd;d++) GaugeBlock[d] = Block[d+1];
-    std::cout << GridLogDslash << " Dirichlet BCs 4d " <<GaugeBlock<<std::endl;
     DirichletFilter<GaugeField> Filter(GaugeBlock);
     Filter.applyFilter(HUmu);
+  } else {
+    std::cout << GridLogMessage << " Dirichlet "<< Dirichlet << " not filtered gauge field" <<std::endl;
   }
   Impl::DoubleStore(GaugeGrid(),Umu,HUmu);
   pickCheckerboard(Even,UmuEven,Umu);

From 5fe480d81c07a564a5615621c909428248f7e208 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 15 Nov 2022 16:21:45 -0500
Subject: [PATCH 071/146] Generic patch

---
 Grid/simd/Grid_generic.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/Grid/simd/Grid_generic.h b/Grid/simd/Grid_generic.h
index 35cb3bc3..4e527e33 100644
--- a/Grid/simd/Grid_generic.h
+++ b/Grid/simd/Grid_generic.h
@@ -244,7 +244,7 @@ struct Conj{
 struct TimesMinusI{
   // Complex
   template <typename T>
-  accelerator_inline vec<T> operator()(vec<T> a, vec<T> b){
+  accelerator_inline vec<T> operator()(vec<T> a){
     vec<T> out;
       
     VECTOR_FOR(i, W<T>::c, 1)
@@ -265,7 +265,7 @@ struct TimesMinusI{
 struct TimesI{
   // Complex
   template <typename T>
-  accelerator_inline vec<T> operator()(vec<T> a, vec<T> b){
+  accelerator_inline vec<T> operator()(vec<T> a){
     vec<T> out;
       
     VECTOR_FOR(i, W<T>::c, 1)

From 0db4f1803f520835f5e86eae71bc8d564191b266 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 15 Nov 2022 16:23:41 -0500
Subject: [PATCH 072/146] Partial dirichlet support

---
 Grid/stencil/SimpleCompressor.h | 152 +++++++++++++++++++++++++-------
 1 file changed, 120 insertions(+), 32 deletions(-)

diff --git a/Grid/stencil/SimpleCompressor.h b/Grid/stencil/SimpleCompressor.h
index b36d954f..eda6d9e7 100644
--- a/Grid/stencil/SimpleCompressor.h
+++ b/Grid/stencil/SimpleCompressor.h
@@ -3,26 +3,108 @@
 
 NAMESPACE_BEGIN(Grid);
 
-template<class vobj>
-accelerator_inline void exchangeSIMT(vobj &mp0,vobj &mp1,const vobj &vp0,const vobj &vp1,Integer type)
-{
-    typedef decltype(coalescedRead(mp0)) sobj;
-    unsigned int Nsimd = vobj::Nsimd();
-    unsigned int mask = Nsimd >> (type + 1);
-    int lane = acceleratorSIMTlane(Nsimd);
-    int j0 = lane &(~mask); // inner coor zero
-    int j1 = lane |(mask) ; // inner coor one
-    const vobj *vpa = &vp0;
-    const vobj *vpb = &vp1;
-    const vobj *vp = (lane&mask) ? (vpb) : (vpa);
-    auto sa = coalescedRead(vp[0],j0);
-    auto sb = coalescedRead(vp[0],j1);
-    coalescedWrite(mp0,sa);
-    coalescedWrite(mp1,sb);
-}
+class SimpleStencilParams{
+public:
+  Coordinate dirichlet;
+  int partialDirichlet;
+  SimpleStencilParams() { partialDirichlet = 0; };
+};
 
-template<class vobj>
-class SimpleCompressor {
+
+// Compressors will inherit buffer management policies
+// Standard comms buffer management
+class FaceGatherSimple
+{
+public:
+  static int PartialCompressionFactor(GridBase *grid) {return 1;};
+  // Decompress is after merge so ok
+  template<class vobj,class cobj,class compressor> 
+  static void Gather_plane_simple (commVector<std::pair<int,int> >& table,
+				   const Lattice<vobj> &rhs,
+				   cobj *buffer,
+				   compressor &compress,
+				   int off,int so,int partial)
+  {
+    int num=table.size();
+    std::pair<int,int> *table_v = & table[0];
+    
+    auto rhs_v = rhs.View(AcceleratorRead);
+    accelerator_forNB( i,num, vobj::Nsimd(), {
+	compress.Compress(buffer[off+table_v[i].first],rhs_v[so+table_v[i].second]);
+    });
+    rhs_v.ViewClose();
+  }
+  template<class vobj,class cobj,class compressor>
+  static void Gather_plane_exchange(commVector<std::pair<int,int> >& table,const Lattice<vobj> &rhs,
+				    Vector<cobj *> pointers,int dimension,int plane,int cbmask,
+				    compressor &compress,int type,int partial)
+  {
+    assert( (table.size()&0x1)==0);
+    int num=table.size()/2;
+    int so  = plane*rhs.Grid()->_ostride[dimension]; // base offset for start of plane
+    
+    auto rhs_v = rhs.View(AcceleratorRead);
+    auto p0=&pointers[0][0];
+    auto p1=&pointers[1][0];
+    auto tp=&table[0];
+    auto rhs_p = &rhs_v[0];
+    accelerator_forNB(j, num, vobj::Nsimd(), {
+	compress.CompressExchange(p0[j],p1[j],
+				  rhs_p[so+tp[2*j  ].second],
+				  rhs_p[so+tp[2*j+1].second],
+				  type);
+    });
+    rhs_v.ViewClose();
+  }
+
+  template<class decompressor,class Decompression>
+  static void DecompressFace(decompressor decompress,Decompression &dd)
+  {
+    auto kp = dd.kernel_p;
+    auto mp = dd.mpi_p;
+    accelerator_forNB(o,dd.buffer_size,1,{
+      decompress.Decompress(kp[o],mp[o]);
+    });
+  }
+  template<class decompressor,class Merger>
+  static void MergeFace(decompressor decompress,Merger &mm)
+  {
+    auto mp = &mm.mpointer[0];
+    auto vp0= &mm.vpointers[0][0];
+    auto vp1= &mm.vpointers[1][0];
+    auto type= mm.type;
+    accelerator_forNB(o,mm.buffer_size/2,vobj::Nsimd(),{
+	decompress.Exchange(mp[2*o],mp[2*o+1],vp0[o],vp1[o],type);
+    });
+  }
+};
+
+////////////////////////////////////
+// Wilson compressor will add alternate policies for Dirichlet
+// and possibly partial Dirichlet for DWF
+////////////////////////////////////
+/*
+class FaceGatherDirichlet
+{
+  // If it's dirichlet we don't assemble comms buffers
+  //
+  // Rely on zeroes in gauge field to drive the correct result
+  // NAN propgagation: field will locally wrap, so fermion should NOT contain NAN and just permute
+  template<class vobj,class cobj,class compressor>
+  static void Gather_plane_simple (commVector<std::pair<int,int> >& table,const Lattice<vobj> &rhs,cobj *buffer,compressor &compress, int off,int so){};
+  template<class vobj,class cobj,class compressor>
+  static void Gather_plane_exchange(commVector<std::pair<int,int> >& table,const Lattice<vobj> &rhs,
+				   Vector<cobj *> pointers,int dimension,int plane,int cbmask,
+				   compressor &compress,int type) {}
+  template<class decompressor,class Merger>
+  static void Merge(decompressor decompress,Merge &mm)  {  }
+  template<class decompressor,class Decompression>
+  static void Decompress(decompressor decompress,Decompression &dd) {}
+};
+*/
+
+template<class vobj,class FaceGather>
+class SimpleCompressorGather : public FaceGather {
 public:
   void Point(int) {};
   accelerator_inline int  CommDatumSize(void) const { return sizeof(vobj); }
@@ -30,20 +112,19 @@ public:
   accelerator_inline void Compress(vobj &buf,const vobj &in) const {
     coalescedWrite(buf,coalescedRead(in));
   }
-  accelerator_inline void Exchange(vobj *mp,vobj *vp0,vobj *vp1,Integer type,Integer o) const {
+  accelerator_inline void Exchange(vobj &mp0,vobj &mp1,vobj &vp0,vobj &vp1,Integer type) const {
 #ifdef GRID_SIMT
-    exchangeSIMT(mp[2*o],mp[2*o+1],vp0[o],vp1[o],type);
+    exchangeSIMT(mp0,mp1,vp0,vp1,type);
 #else
-    exchange(mp[2*o],mp[2*o+1],vp0[o],vp1[o],type);
+    exchange(mp0,mp1,vp0,vp1,type);
 #endif
   }
-  accelerator_inline void Decompress(vobj *out,vobj *in, int o) const { assert(0); }
-  accelerator_inline void CompressExchange(vobj *out0,vobj *out1,const vobj *in,
-					   int j,int k, int m,int type) const {
+  accelerator_inline void Decompress(vobj &out,vobj &in) const {  };
+  accelerator_inline void CompressExchange(vobj &out0,vobj &out1,const vobj &in0,const vobj &in1,int type) const {
 #ifdef GRID_SIMT
-    exchangeSIMT(out0[j],out1[j],in[k],in[m],type);
+    exchangeSIMT(out0,out1,in0,in1,type);
 #else
-    exchange(out0[j],out1[j],in[k],in[m],type);
+    exchange(out0,out1,in0,in1,type);
 #endif
   }
   // For cshift. Cshift should drop compressor coupling altogether 
@@ -52,11 +133,18 @@ public:
     return arg;
   }
 };
-class SimpleStencilParams{
-public:
-  Coordinate dirichlet;
-  SimpleStencilParams() {};
-};
+
+// Standard compressor never needs dirichlet.
+//
+// Get away with a local period wrap and rely on dirac operator to use a zero gauge link as it is faster
+//
+// Compressors that inherit Dirichlet and Non-dirichlet behaviour.
+//
+// Currently run-time behaviour through StencilParameters paramaters, p.dirichlet
+// combined with the FaceGatherSimple behaviour
+
+template <class vobj> using SimpleCompressor = SimpleCompressorGather<vobj,FaceGatherSimple>;
+//template <class vobj> using SimpleCompressorDirichlet = SimpleCompressorGather<vobj,FaceGatherDirichlet>;
 
 NAMESPACE_END(Grid);
 

From e2e269e03b785114d8989b292e6052d1b236d3cc Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 15 Nov 2022 16:24:26 -0500
Subject: [PATCH 073/146] Partial dirichlet BCs

---
 Grid/stencil/Stencil.h | 204 ++++++++++++++++++++++++++---------------
 1 file changed, 131 insertions(+), 73 deletions(-)

diff --git a/Grid/stencil/Stencil.h b/Grid/stencil/Stencil.h
index e23ff258..dd8b646e 100644
--- a/Grid/stencil/Stencil.h
+++ b/Grid/stencil/Stencil.h
@@ -59,6 +59,7 @@ NAMESPACE_BEGIN(Grid);
 void Gather_plane_table_compute (GridBase *grid,int dimension,int plane,int cbmask,
 				 int off,std::vector<std::pair<int,int> > & table);
 
+/*
 template<class vobj,class cobj,class compressor>
 void Gather_plane_simple_table (commVector<std::pair<int,int> >& table,const Lattice<vobj> &rhs,cobj *buffer,compressor &compress, int off,int so)   __attribute__((noinline));
 
@@ -103,6 +104,7 @@ void Gather_plane_exchange_table(commVector<std::pair<int,int> >& table,const La
   });
   rhs_v.ViewClose();
 }
+*/
 
 struct StencilEntry {
 #ifdef GRID_CUDA
@@ -133,8 +135,18 @@ class CartesianStencilAccelerator {
   int           _osites;
   StencilVector _directions;
   StencilVector _distances;
-  StencilVector _comms_send;
-  StencilVector _comms_recv;
+  ///////////////////////////////////////////////////
+  // If true, this is FULLY communicated per face
+  // Otherwise will either be full or partial dirichlet
+  ///////////////////////////////////////////////////
+  StencilVector _comms_send; 
+  StencilVector _comms_recv; // this is FULLY communicated per face
+  ///////////////////////////////////////////////////
+  // If true, this is partially communicated per face
+  ///////////////////////////////////////////////////
+  StencilVector _comms_partial_send; 
+  StencilVector _comms_partial_recv;
+  //
   StencilVector _comm_buf_size;
   StencilVector _permute_type;
   StencilVector same_node;
@@ -229,7 +241,8 @@ public:
     Integer from_rank;
     Integer do_send;
     Integer do_recv;
-    Integer bytes;
+    Integer xbytes;
+    Integer rbytes;
   };
   struct Merge {
     cobj * mpointer;
@@ -237,11 +250,15 @@ public:
     Vector<cobj *> vpointers;
     Integer buffer_size;
     Integer type;
+    Integer partial; // partial dirichlet BCs
+    Coordinate dims;
   };
   struct Decompress {
     cobj * kernel_p;
     cobj * mpi_p;
     Integer buffer_size;
+    Integer partial; // partial dirichlet BCs
+    Coordinate dims;
   };
   struct CopyReceiveBuffer {
     void * from_p;
@@ -252,7 +269,8 @@ public:
     Integer direction;
     Integer OrthogPlane;
     Integer DestProc;
-    Integer bytes;
+    Integer xbytes;
+    Integer rbytes;
     Integer lane;
     Integer cb;
     void *recv_buf;
@@ -277,6 +295,7 @@ public:
   }
 
   int face_table_computed;
+  int partialDirichlet;
   std::vector<commVector<std::pair<int,int> > > face_table ;
   Vector<int> surface_list;
 
@@ -365,7 +384,7 @@ public:
 					Packets[i].to_rank,Packets[i].do_send,
 					Packets[i].recv_buf,
 					Packets[i].from_rank,Packets[i].do_recv,
-					Packets[i].bytes,i);
+					Packets[i].xbytes,Packets[i].rbytes,i);
     }
   }
 
@@ -501,7 +520,9 @@ public:
     }
   }
   
-  Integer CheckForDuplicate(Integer direction, Integer OrthogPlane, Integer DestProc, void *recv_buf,Integer lane,Integer bytes,Integer cb)
+  Integer CheckForDuplicate(Integer direction, Integer OrthogPlane, Integer DestProc, void *recv_buf,Integer lane,
+			    Integer xbytes,Integer rbytes,
+			    Integer cb)
   {
     CachedTransfer obj;
     obj.direction   = direction;
@@ -509,19 +530,22 @@ public:
     obj.DestProc    = DestProc;
     obj.recv_buf    = recv_buf;
     obj.lane        = lane;
-    obj.bytes       = bytes;
+    obj.xbytes      = xbytes;
+    obj.rbytes      = rbytes;
     obj.cb          = cb;
 
     for(int i=0;i<CachedTransfers.size();i++){
       if (   (CachedTransfers[i].direction  ==direction)
 	   &&(CachedTransfers[i].OrthogPlane==OrthogPlane)
 	   &&(CachedTransfers[i].DestProc   ==DestProc)
-	   &&(CachedTransfers[i].bytes      ==bytes)
+	   &&(CachedTransfers[i].xbytes      ==xbytes)
+	   &&(CachedTransfers[i].rbytes      ==rbytes)
 	   &&(CachedTransfers[i].lane       ==lane)
 	   &&(CachedTransfers[i].cb         ==cb)
 	     ){
-
-	AddCopy(CachedTransfers[i].recv_buf,recv_buf,bytes);
+	// FIXME worry about duplicate with partial compression
+	// Wont happen as DWF has no duplicates, but...
+	AddCopy(CachedTransfers[i].recv_buf,recv_buf,rbytes);
 	return 1;
       }
     }
@@ -532,7 +556,7 @@ public:
   void AddPacket(void *xmit,void * rcv,
 		 Integer to, Integer do_send,
 		 Integer from, Integer do_recv,
-		 Integer bytes){
+		 Integer xbytes,Integer rbytes){
     Packet p;
     p.send_buf = xmit;
     p.recv_buf = rcv;
@@ -540,11 +564,16 @@ public:
     p.from_rank= from;
     p.do_send  = do_send;
     p.do_recv  = do_recv;
-    p.bytes    = bytes;
+    p.xbytes    = xbytes;
+    p.rbytes    = rbytes;
+    //    if (do_send) std::cout << GridLogMessage << " MPI packet to   "<<to<< " of size "<<xbytes<<std::endl;
+    //    if (do_recv) std::cout << GridLogMessage << " MPI packet from "<<from<< " of size "<<xbytes<<std::endl;
     Packets.push_back(p);
   }
   void AddDecompress(cobj *k_p,cobj *m_p,Integer buffer_size,std::vector<Decompress> &dv) {
     Decompress d;
+    d.partial  = this->partialDirichlet;
+    d.dims     = _grid->_fdimensions;
     d.kernel_p = k_p;
     d.mpi_p    = m_p;
     d.buffer_size = buffer_size;
@@ -552,6 +581,8 @@ public:
   }
   void AddMerge(cobj *merge_p,Vector<cobj *> &rpointers,Integer buffer_size,Integer type,std::vector<Merge> &mv) {
     Merge m;
+    m.partial  = this->partialDirichlet;
+    m.dims     = _grid->_fdimensions;
     m.type     = type;
     m.mpointer = merge_p;
     m.vpointers= rpointers;
@@ -571,21 +602,10 @@ public:
   void CommsMerge(decompressor decompress,std::vector<Merge> &mm,std::vector<Decompress> &dd)
   {
     for(int i=0;i<mm.size();i++){
-      auto mp = &mm[i].mpointer[0];
-      auto vp0= &mm[i].vpointers[0][0];
-      auto vp1= &mm[i].vpointers[1][0];
-      auto type= mm[i].type;
-      accelerator_forNB(o,mm[i].buffer_size/2,vobj::Nsimd(),{
-	  decompress.Exchange(mp,vp0,vp1,type,o);
-      });
+      decompressor::MergeFace(decompress,mm[i]);
     }
-
     for(int i=0;i<dd.size();i++){
-      auto kp = dd[i].kernel_p;
-      auto mp = dd[i].mpi_p;
-      accelerator_forNB(o,dd[i].buffer_size,1,{
-	decompress.Decompress(kp,mp,o);
-      });
+      decompressor::DecompressFace(decompress,dd[i]);
     }
   }
   ////////////////////////////////////////
@@ -664,6 +684,10 @@ public:
 	  if ( ( (ld*(pc+1) ) % block ) == 0 ) this->_comms_send[ii] = 0;
 	  if ( ( (ld*pc     ) % block ) == 0 ) this->_comms_recv[ii] = 0;
 	}
+	if ( partialDirichlet ) {
+	  this->_comms_partial_send[ii] = !this->_comms_send[ii];
+	  this->_comms_partial_recv[ii] = !this->_comms_recv[ii];
+	}
       }
     }
   }
@@ -691,7 +715,7 @@ public:
     this->same_node.resize(npoints);
 
     if ( p.dirichlet.size() ==0 ) p.dirichlet.resize(grid->Nd(),0);
-
+    partialDirichlet = p.partialDirichlet;
     DirichletBlock(p.dirichlet); // comms send/recv set up
 
     _unified_buffer_size=0;
@@ -827,7 +851,7 @@ public:
     GridBase *grid=_grid;
     const int Nsimd = grid->Nsimd();
 
-    int comms_recv      = this->_comms_recv[point];
+    int comms_recv      = this->_comms_recv[point] || this->_comms_partial_recv[point] ;
     int fd              = _grid->_fdimensions[dimension];
     int ld              = _grid->_ldimensions[dimension];
     int rd              = _grid->_rdimensions[dimension];
@@ -1014,8 +1038,10 @@ public:
   {
     typedef typename cobj::vector_type vector_type;
 
-    int comms_send   = this->_comms_send[point] ;
-    int comms_recv   = this->_comms_recv[point] ;
+    int comms_send   = this->_comms_send[point];
+    int comms_recv   = this->_comms_recv[point];
+    int comms_partial_send   = this->_comms_partial_send[point] ;
+    int comms_partial_recv   = this->_comms_partial_recv[point] ;
 
     assert(rhs.Grid()==_grid);
     //	  conformable(_grid,rhs.Grid());
@@ -1046,7 +1072,17 @@ public:
 	if (cbmask != 0x3) words=words>>1;
 
 	int bytes =  words * compress.CommDatumSize();
+	int xbytes;
+	int rbytes;
 
+	if ( comms_send ) xbytes = bytes; // Full send
+	else if ( comms_partial_send ) xbytes = bytes/compressor::PartialCompressionFactor(_grid);
+	else xbytes = 0; // full dirichlet
+
+	if ( comms_recv ) rbytes = bytes;
+	else if ( comms_partial_recv ) rbytes = bytes/compressor::PartialCompressionFactor(_grid);
+	else rbytes = 0;
+	
 	int so  = sx*rhs.Grid()->_ostride[dimension]; // base offset for start of plane
 	int comm_off = u_comm_offset;
 
@@ -1059,49 +1095,47 @@ public:
 	assert (xmit_to_rank   != _grid->ThisRank());
 	assert (recv_from_rank != _grid->ThisRank());
 
-	if( comms_send ) {
-
-	  if ( !face_table_computed ) {
-	    face_table.resize(face_idx+1);
-	    std::vector<std::pair<int,int> >  face_table_host ;
-	    Gather_plane_table_compute ((GridBase *)_grid,dimension,sx,cbmask,comm_off,face_table_host);
+	if ( !face_table_computed ) {
+	  face_table.resize(face_idx+1);
+	  std::vector<std::pair<int,int> >  face_table_host ;
+	  Gather_plane_table_compute ((GridBase *)_grid,dimension,sx,cbmask,comm_off,face_table_host);
+	  //	  std::cout << "bytes expect "<< bytes << " " << face_table_host.size()* compress.CommDatumSize()<<std::endl;
 	    face_table[face_idx].resize(face_table_host.size());
 	    acceleratorCopyToDevice(&face_table_host[0],
 				    &face_table[face_idx][0],
 				    face_table[face_idx].size()*sizeof(face_table_host[0]));
-	  }
-
-
-	  if ( compress.DecompressionStep() ) {
-	    recv_buf=u_simd_recv_buf[0];
-	  } else {
-	    recv_buf=this->u_recv_buf_p;
-	  }
-
-	  send_buf = this->u_send_buf_p; // Gather locally, must send
-	
-	  ////////////////////////////////////////////////////////
-	  // Gather locally
-	  ////////////////////////////////////////////////////////
-	  assert(send_buf!=NULL);
-	  Gather_plane_simple_table(face_table[face_idx],rhs,send_buf,compress,comm_off,so);
 	}
 
-	int duplicate = CheckForDuplicate(dimension,sx,comm_proc,(void *)&recv_buf[comm_off],0,bytes,cbmask);
-	if ( (!duplicate) ) { // Force comms for now
 
+	if ( (compress.DecompressionStep()&&comms_recv) || comms_partial_recv ) {
+	  recv_buf=u_simd_recv_buf[0];
+	} else {
+	  recv_buf=this->u_recv_buf_p;
+	}
+
+	////////////////////////////////////////////////////////
+	// Gather locally
+	////////////////////////////////////////////////////////
+	send_buf = this->u_send_buf_p; // Gather locally, must send
+	assert(send_buf!=NULL);
+
+	compressor::Gather_plane_simple(face_table[face_idx],rhs,send_buf,compress,comm_off,so,comms_partial_send);
+
+        int duplicate = CheckForDuplicate(dimension,sx,comm_proc,(void *)&recv_buf[comm_off],0,xbytes,rbytes,cbmask);
+	if ( !duplicate ) { // Force comms for now
+	  
 	  ///////////////////////////////////////////////////////////
 	  // Build a list of things to do after we synchronise GPUs
 	  // Start comms now???
 	  ///////////////////////////////////////////////////////////
 	  AddPacket((void *)&send_buf[comm_off],
 		    (void *)&recv_buf[comm_off],
-		    xmit_to_rank, comms_send,
-		    recv_from_rank, comms_recv,
-		    bytes);
+		    xmit_to_rank, comms_send|comms_partial_send,
+		    recv_from_rank, comms_recv|comms_partial_recv,
+		    xbytes,rbytes);
 	}
-	
-	if ( compress.DecompressionStep()  && comms_recv ) {
+
+	if ( (compress.DecompressionStep() && comms_recv) || comms_partial_recv ) {
 	  AddDecompress(&this->u_recv_buf_p[comm_off],
 			&recv_buf[comm_off],
 			words,Decompressions);
@@ -1109,7 +1143,6 @@ public:
 	
 	u_comm_offset+=words;
 	face_idx++;
-
       }
     }
     return 0;
@@ -1122,8 +1155,10 @@ public:
 
     const int maxl =2;// max layout in a direction
 
-    int comms_send   = this->_comms_send[point] ;
-    int comms_recv   = this->_comms_recv[point] ;
+    int comms_send   = this->_comms_send[point];
+    int comms_recv   = this->_comms_recv[point];
+    int comms_partial_send   = this->_comms_partial_send[point] ;
+    int comms_partial_recv   = this->_comms_partial_recv[point] ;
 
     int fd = _grid->_fdimensions[dimension];
     int rd = _grid->_rdimensions[dimension];
@@ -1153,6 +1188,11 @@ public:
 
     int datum_bytes = compress.CommDatumSize();
     int bytes = (reduced_buffer_size*datum_bytes)/simd_layout;
+
+    // how many bytes on wire : partial dirichlet or dirichlet may set to < bytes
+    int xbytes; 
+    int rbytes; 
+    
     assert(bytes*simd_layout == reduced_buffer_size*datum_bytes);
 
     Vector<cobj *> rpointers(maxl);
@@ -1182,22 +1222,37 @@ public:
 	if ( !face_table_computed ) {
 	  face_table.resize(face_idx+1);
 	  std::vector<std::pair<int,int> >  face_table_host ;
-				
+
 	  Gather_plane_table_compute ((GridBase *)_grid,dimension,sx,cbmask,comm_off,face_table_host);
 	  face_table[face_idx].resize(face_table_host.size());
 	  acceleratorCopyToDevice(&face_table_host[0],
 				  &face_table[face_idx][0],
 				  face_table[face_idx].size()*sizeof(face_table_host[0]));
+
 	}
 
-	if ( comms_send || comms_recv ) {
-	  Gather_plane_exchange_table(face_table[face_idx],rhs,spointers,dimension,sx,cbmask,compress,permute_type);
+	
+	if ( comms_send ) xbytes = bytes;
+	else if ( comms_partial_send ) xbytes = bytes/compressor::PartialCompressionFactor(_grid);
+	else xbytes = 0;
+
+	if ( comms_recv ) rbytes = bytes;
+	else if ( comms_partial_recv ) rbytes = bytes/compressor::PartialCompressionFactor(_grid);
+	else rbytes = 0;
+
+	// Gathers SIMD lanes for send and merge
+	// Different faces can be full comms or partial comms with  multiple ranks per node
+	if ( comms_send || comms_recv||comms_partial_send||comms_partial_recv ) {
+
+	  int partial = partialDirichlet;
+	  compressor::Gather_plane_exchange(face_table[face_idx],rhs,
+					    spointers,dimension,sx,cbmask,
+					    compress,permute_type,partial );
 	}
 	face_idx++;
 
-	//spointers[0] -- low
-	//spointers[1] -- high
-
+	//spointers[0] -- low simd coor
+	//spointers[1] -- high simd coor
 	for(int i=0;i<maxl;i++){
 
 	  int my_coor  = rd*i + x;            // self explanatory
@@ -1222,13 +1277,16 @@ public:
 	    _grid->ShiftedRanks(dimension,nbr_proc,xmit_to_rank,recv_from_rank);
 
 	    rpointers[i] = rp;
-
-	    int duplicate = CheckForDuplicate(dimension,sx,nbr_proc,(void *)rp,i,bytes,cbmask);
+	    
+	    int duplicate = CheckForDuplicate(dimension,sx,nbr_proc,(void *)rp,i,xbytes,rbytes,cbmask);
 	    if ( !duplicate  ) { 
+	      if ( (bytes != rbytes) && (rbytes!=0) ){
+		acceleratorMemSet(rp,0,bytes); // Zero prefill comms buffer to zero
+	      }
 	      AddPacket((void *)sp,(void *)rp,
-			xmit_to_rank,comms_send,
-			recv_from_rank,comms_recv,
-			bytes);
+			xmit_to_rank,comms_send|comms_partial_send,
+			recv_from_rank,comms_recv|comms_partial_recv,
+			xbytes,rbytes);
 	    }
 
 	  } else {
@@ -1238,7 +1296,7 @@ public:
 	  }
 	}
 
-	if ( comms_recv ) {
+	if ( comms_recv|comms_partial_recv ) {
 	  AddMerge(&this->u_recv_buf_p[comm_off],rpointers,reduced_buffer_size,permute_type,Mergers);
 	}
 

From 7d302a525def70d276a2c6e87c50a15a57ee001a Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 15 Nov 2022 16:24:55 -0500
Subject: [PATCH 074/146] Natural place for this routine is here

---
 Grid/tensors/Tensor_SIMT.h | 21 +++++++++++++++++++++
 1 file changed, 21 insertions(+)

diff --git a/Grid/tensors/Tensor_SIMT.h b/Grid/tensors/Tensor_SIMT.h
index 0a7d3382..8015d74c 100644
--- a/Grid/tensors/Tensor_SIMT.h
+++ b/Grid/tensors/Tensor_SIMT.h
@@ -31,6 +31,27 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 NAMESPACE_BEGIN(Grid);
 
+////////////////////////////////////////////////
+// Inside a GPU thread
+////////////////////////////////////////////////
+template<class vobj>
+accelerator_inline void exchangeSIMT(vobj &mp0,vobj &mp1,const vobj &vp0,const vobj &vp1,Integer type)
+{
+    typedef decltype(coalescedRead(mp0)) sobj;
+    unsigned int Nsimd = vobj::Nsimd();
+    unsigned int mask = Nsimd >> (type + 1);
+    int lane = acceleratorSIMTlane(Nsimd);
+    int j0 = lane &(~mask); // inner coor zero
+    int j1 = lane |(mask) ; // inner coor one
+    const vobj *vpa = &vp0;
+    const vobj *vpb = &vp1;
+    const vobj *vp = (lane&mask) ? (vpb) : (vpa);
+    auto sa = coalescedRead(vp[0],j0);
+    auto sb = coalescedRead(vp[0],j1);
+    coalescedWrite(mp0,sa);
+    coalescedWrite(mp1,sb);
+}
+
 
 #ifndef GRID_SIMT
 //////////////////////////////////////////

From 0352da34f019bea2fd5a7117600f3da68cf40179 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 15 Nov 2022 16:26:49 -0500
Subject: [PATCH 075/146] Several deleted files

---
 systems/Spock/comms.slurm    | 26 --------------------------
 systems/Spock/config-command | 14 --------------
 systems/Spock/dwf.slurm      | 26 --------------------------
 systems/Spock/dwf4.slurm     | 26 --------------------------
 systems/Spock/dwf8.slurm     | 24 ------------------------
 systems/Spock/mpiwrapper.sh  | 12 ------------
 systems/Spock/sourceme.sh    |  9 ---------
 7 files changed, 137 deletions(-)
 delete mode 100644 systems/Spock/comms.slurm
 delete mode 100644 systems/Spock/config-command
 delete mode 100644 systems/Spock/dwf.slurm
 delete mode 100644 systems/Spock/dwf4.slurm
 delete mode 100644 systems/Spock/dwf8.slurm
 delete mode 100755 systems/Spock/mpiwrapper.sh
 delete mode 100644 systems/Spock/sourceme.sh

diff --git a/systems/Spock/comms.slurm b/systems/Spock/comms.slurm
deleted file mode 100644
index 0841fda0..00000000
--- a/systems/Spock/comms.slurm
+++ /dev/null
@@ -1,26 +0,0 @@
-#!/bin/bash
-# Begin LSF Directives
-#SBATCH -A LGT104
-#SBATCH -t 01:00:00
-##SBATCH -U openmpThu
-#SBATCH -p ecp
-#SBATCH -J comms
-#SBATCH -o comms.%J
-#SBATCH -e comms.%J
-#SBATCH -N 1
-#SBATCH -n 2
-
-DIR=.
-module list
-export MPIR_CVAR_GPU_EAGER_DEVICE_MEM=0
-export MPICH_GPU_SUPPORT_ENABLED=1
-#export MPICH_SMP_SINGLE_COPY_MODE=XPMEM
-#export MPICH_SMP_SINGLE_COPY_MODE=CMA
-export MPICH_SMP_SINGLE_COPY_MODE=NONE
-export OMP_NUM_THREADS=8
-
-AT=8
-echo MPICH_SMP_SINGLE_COPY_MODE $MPICH_SMP_SINGLE_COPY_MODE
-PARAMS=" --accelerator-threads ${AT} --grid 64.64.32.32 --mpi 2.1.1.1 "
-srun -n2 --label -c$OMP_NUM_THREADS --gpus-per-task=1 ./mpiwrapper.sh ./benchmarks/Benchmark_comms_host_device $PARAMS
-
diff --git a/systems/Spock/config-command b/systems/Spock/config-command
deleted file mode 100644
index 3ffefe4f..00000000
--- a/systems/Spock/config-command
+++ /dev/null
@@ -1,14 +0,0 @@
-../../configure --enable-comms=mpi-auto \
---enable-unified=no \
---enable-shm=nvlink \
---enable-accelerator=hip \
---enable-gen-simd-width=64 \
---enable-simd=GPU \
---disable-fermion-reps \
---disable-gparity \
---with-gmp=$OLCF_GMP_ROOT \
---with-mpfr=/opt/cray/pe/gcc/mpfr/3.1.4/ \
-CXX=hipcc MPICXX=mpicxx \
-CXXFLAGS="-fPIC -I/opt/rocm-4.3.0/include/ -std=c++14 -I${MPICH_DIR}/include " \
---prefix=/ccs/home/chulwoo/Grid \
- LDFLAGS=" -L${MPICH_DIR}/lib -lmpi -L${CRAY_MPICH_ROOTDIR}/gtl/lib -lmpi_gtl_hsa "
diff --git a/systems/Spock/dwf.slurm b/systems/Spock/dwf.slurm
deleted file mode 100644
index 7144a270..00000000
--- a/systems/Spock/dwf.slurm
+++ /dev/null
@@ -1,26 +0,0 @@
-#!/bin/bash
-# Begin LSF Directives
-#SBATCH -A LGT104
-#SBATCH -t 01:00:00
-##SBATCH -U openmpThu
-#SBATCH -p ecp
-#SBATCH -J DWF
-#SBATCH -o DWF.%J
-#SBATCH -e DWF.%J
-#SBATCH -N 1
-#SBATCH -n 1
-
-DIR=.
-module list
-export MPIR_CVAR_GPU_EAGER_DEVICE_MEM=0
-export MPICH_GPU_SUPPORT_ENABLED=1
-#export MPICH_SMP_SINGLE_COPY_MODE=XPMEM
-#export MPICH_SMP_SINGLE_COPY_MODE=NONE
-export MPICH_SMP_SINGLE_COPY_MODE=CMA
-export OMP_NUM_THREADS=8
-
-AT=8
-echo MPICH_SMP_SINGLE_COPY_MODE $MPICH_SMP_SINGLE_COPY_MODE
-PARAMS=" --accelerator-threads ${AT} --grid 32.32.32.32 --mpi 1.1.1.1 --comms-overlap"
-srun -n1 --label -c$OMP_NUM_THREADS --gpus-per-task=1 ./mpiwrapper.sh ./benchmarks/Benchmark_dwf_fp32 $PARAMS
-
diff --git a/systems/Spock/dwf4.slurm b/systems/Spock/dwf4.slurm
deleted file mode 100644
index 261929ab..00000000
--- a/systems/Spock/dwf4.slurm
+++ /dev/null
@@ -1,26 +0,0 @@
-#!/bin/bash
-# Begin LSF Directives
-#SBATCH -A LGT104
-#SBATCH -t 01:00:00
-##SBATCH -U openmpThu
-#SBATCH -p ecp
-#SBATCH -J DWF
-#SBATCH -o DWF.%J
-#SBATCH -e DWF.%J
-#SBATCH -N 1
-#SBATCH -n 4
-
-DIR=.
-module list
-export MPIR_CVAR_GPU_EAGER_DEVICE_MEM=0
-export MPICH_GPU_SUPPORT_ENABLED=1
-#export MPICH_SMP_SINGLE_COPY_MODE=XPMEM
-export MPICH_SMP_SINGLE_COPY_MODE=NONE
-#export MPICH_SMP_SINGLE_COPY_MODE=CMA
-export OMP_NUM_THREADS=8
-
-AT=8
-echo MPICH_SMP_SINGLE_COPY_MODE $MPICH_SMP_SINGLE_COPY_MODE
-PARAMS=" --accelerator-threads ${AT} --grid 32.32.64.64 --mpi 1.1.2.2 --comms-overlap --shm 2048 --shm-mpi 0"
-srun -n4 --label -c$OMP_NUM_THREADS --gpus-per-task=1 ./mpiwrapper.sh ./benchmarks/Benchmark_dwf_fp32 $PARAMS
-
diff --git a/systems/Spock/dwf8.slurm b/systems/Spock/dwf8.slurm
deleted file mode 100644
index f2e12b97..00000000
--- a/systems/Spock/dwf8.slurm
+++ /dev/null
@@ -1,24 +0,0 @@
-#!/bin/bash
-# Begin LSF Directives
-#SBATCH -A LGT104
-#SBATCH -t 3:00:00
-#SBATCH -p ecp
-#SBATCH -J DWF
-#SBATCH -o DWF.%J
-#SBATCH -e DWF.%J
-#SBATCH -N 2
-#SBATCH -n 8
-
-DIR=.
-module list
-export MPIR_CVAR_GPU_EAGER_DEVICE_MEM=0
-export MPICH_GPU_SUPPORT_ENABLED=1
-export MPICH_SMP_SINGLE_COPY_MODE=CMA
-
-export OMP_NUM_THREADS=8
-
-AT=8
-echo MPICH_SMP_SINGLE_COPY_MODE $MPICH_SMP_SINGLE_COPY_MODE
-PARAMS=" --accelerator-threads ${AT} --grid 16.16.16.48 --mpi 1.2.2.2 --comms-overlap --shm 2048 --shm-mpi 0"
-srun -N2 -n8 --label -c$OMP_NUM_THREADS --gpus-per-task=1 ./mpiwrapper.sh ./HMC/Mobius2p1f_DD_RHMC $PARAMS
-
diff --git a/systems/Spock/mpiwrapper.sh b/systems/Spock/mpiwrapper.sh
deleted file mode 100755
index 76c4e364..00000000
--- a/systems/Spock/mpiwrapper.sh
+++ /dev/null
@@ -1,12 +0,0 @@
-#!/bin/bash
-
-lrank=$SLURM_LOCALID
-
-export ROCR_VISIBLE_DEVICES=$SLURM_LOCALID
-
-echo "`hostname` - $lrank device=$ROCR_VISIBLE_DEVICES binding=$BINDING"
-
-$*
-
-
-
diff --git a/systems/Spock/sourceme.sh b/systems/Spock/sourceme.sh
deleted file mode 100644
index 415341d0..00000000
--- a/systems/Spock/sourceme.sh
+++ /dev/null
@@ -1,9 +0,0 @@
-module load emacs
-module load PrgEnv-gnu
-module load rocm/4.5.0
-module load gmp
-module load cray-fftw
-module load craype-accel-amd-gfx908
-export MPIR_CVAR_GPU_EAGER_DEVICE_MEM=0
-export MPICH_GPU_SUPPORT_ENABLED=1
-export LD_LIBRARY_PATH=/opt/cray/pe/gcc/mpfr/3.1.4/lib/:$LD_LIBRARY_PATH

From 45a001e078c2a1a6bbefa82cdd5d24d6ac93579d Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 15 Nov 2022 16:27:20 -0500
Subject: [PATCH 076/146] Debug compile

---
 systems/mac-arm/config-command-mpi | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/systems/mac-arm/config-command-mpi b/systems/mac-arm/config-command-mpi
index 52506559..e91d74e6 100644
--- a/systems/mac-arm/config-command-mpi
+++ b/systems/mac-arm/config-command-mpi
@@ -1,3 +1,3 @@
-#CXX=mpicxx-openmpi-mp CXXFLAGS=-I/opt/local/include/ LDFLAGS=-L/opt/local/lib/ ../../configure --enable-simd=GEN --enable-debug --enable-comms=mpi
-CXX=mpicxx-openmpi-mp CXXFLAGS=-I/opt/local/include/ LDFLAGS=-L/opt/local/lib/ ../../configure --enable-simd=GPU-RRII --enable-comms=mpi
+CXX=mpicxx-openmpi-mp CXXFLAGS=-I/opt/local/include/ LDFLAGS=-L/opt/local/lib/ ../../configure --enable-simd=GEN --enable-debug --enable-comms=mpi
+#CXX=mpicxx-openmpi-mp CXXFLAGS=-I/opt/local/include/ LDFLAGS=-L/opt/local/lib/ ../../configure --enable-simd=GPU-RRII --enable-comms=mpi
 #CXX=mpicxx-openmpi-mp CXXFLAGS=-I/opt/local/include/ LDFLAGS=-L/opt/local/lib/ ../../configure --enable-simd=GPU --enable-debug --enable-comms=mpi

From e74666a09c7a160a31684c7a073778e9745cb97b Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 15 Nov 2022 16:34:21 -0500
Subject: [PATCH 077/146] Double length vector type for fast precision change

---
 Grid/simd/Grid_doubled_vector.h | 666 ++++++++++++++++++++++++++++++++
 1 file changed, 666 insertions(+)
 create mode 100644 Grid/simd/Grid_doubled_vector.h

diff --git a/Grid/simd/Grid_doubled_vector.h b/Grid/simd/Grid_doubled_vector.h
new file mode 100644
index 00000000..ee604750
--- /dev/null
+++ b/Grid/simd/Grid_doubled_vector.h
@@ -0,0 +1,666 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid
+
+    Source file: ./lib/simd/Grid_vector_types.h
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once
+
+NAMESPACE_BEGIN(Grid);
+
+template <class Scalar_type, class Vector_type>
+class Grid_simd2 {
+public:
+  typedef typename RealPart<Scalar_type>::type Real;
+  typedef Vector_type vector_type;
+  typedef Scalar_type scalar_type;
+
+  typedef union conv_t_union {
+    Vector_type v;
+    Scalar_type s[sizeof(Vector_type) / sizeof(Scalar_type)];
+    accelerator_inline conv_t_union(){};
+  } conv_t;
+
+  static constexpr int nvec=2;
+  Vector_type v[nvec];
+
+  static accelerator_inline constexpr int Nsimd(void) {
+    static_assert( (sizeof(Vector_type) / sizeof(Scalar_type) >= 1), " size mismatch " );
+    
+    return nvec*sizeof(Vector_type) / sizeof(Scalar_type);
+  }
+  
+  accelerator_inline Grid_simd2 &operator=(const Grid_simd2 &&rhs) {
+    for(int n=0;n<nvec;n++) v[n] = rhs.v[n];
+    return *this;
+  };
+  accelerator_inline Grid_simd2 &operator=(const Grid_simd2 &rhs) {
+    for(int n=0;n<nvec;n++) v[n] = rhs.v[n];
+    return *this;
+  };  // faster than not declaring it and leaving to the compiler
+
+  accelerator Grid_simd2() = default;
+  accelerator_inline Grid_simd2(const Grid_simd2 &rhs) {    for(int n=0;n<nvec;n++) v[n] = rhs.v[n];  };
+  accelerator_inline Grid_simd2(const Grid_simd2 &&rhs){    for(int n=0;n<nvec;n++) v[n] = rhs.v[n];  };
+  accelerator_inline Grid_simd2(const Real a) { vsplat(*this, Scalar_type(a)); };
+  // Enable if complex type
+  template <typename S = Scalar_type> accelerator_inline
+  Grid_simd2(const typename std::enable_if<is_complex<S>::value, S>::type a) {
+      vsplat(*this, a);
+  };
+
+  /////////////////////////////
+  // Constructors
+  /////////////////////////////
+  accelerator_inline Grid_simd2 &  operator=(const Zero &z) {
+    vzero(*this);
+    return (*this);
+  }
+
+  ///////////////////////////////////////////////
+  // mac, mult, sub, add, adj
+  ///////////////////////////////////////////////
+
+  friend accelerator_inline void mac(Grid_simd2 *__restrict__ y,
+				     const Grid_simd2 *__restrict__ a,
+				     const Grid_simd2 *__restrict__ x) {
+    *y = (*a) * (*x) + (*y);
+  };
+
+  friend accelerator_inline void mult(Grid_simd2 *__restrict__ y,
+				      const Grid_simd2 *__restrict__ l,
+				      const Grid_simd2 *__restrict__ r) {
+    *y = (*l) * (*r);
+  }
+
+  friend accelerator_inline void sub(Grid_simd2 *__restrict__ y,
+				     const Grid_simd2 *__restrict__ l,
+				     const Grid_simd2 *__restrict__ r) {
+    *y = (*l) - (*r);
+  }
+  friend accelerator_inline void add(Grid_simd2 *__restrict__ y,
+				     const Grid_simd2 *__restrict__ l,
+				     const Grid_simd2 *__restrict__ r) {
+    *y = (*l) + (*r);
+  }
+  friend accelerator_inline void mac(Grid_simd2 *__restrict__ y,
+				     const Scalar_type *__restrict__ a,
+				     const Grid_simd2 *__restrict__ x) {
+    *y = (*a) * (*x) + (*y);
+  };
+  friend accelerator_inline void mult(Grid_simd2 *__restrict__ y,
+				      const Scalar_type *__restrict__ l,
+				      const Grid_simd2 *__restrict__ r) {
+    *y = (*l) * (*r);
+  }
+  friend accelerator_inline void sub(Grid_simd2 *__restrict__ y,
+				     const Scalar_type *__restrict__ l,
+				     const Grid_simd2 *__restrict__ r) {
+    *y = (*l) - (*r);
+  }
+  friend accelerator_inline void add(Grid_simd2 *__restrict__ y,
+				     const Scalar_type *__restrict__ l,
+				     const Grid_simd2 *__restrict__ r) {
+    *y = (*l) + (*r);
+  }
+
+  friend accelerator_inline void mac(Grid_simd2 *__restrict__ y,
+				     const Grid_simd2 *__restrict__ a,
+				     const Scalar_type *__restrict__ x) {
+    *y = (*a) * (*x) + (*y);
+  };
+  friend accelerator_inline void mult(Grid_simd2 *__restrict__ y,
+				      const Grid_simd2 *__restrict__ l,
+				      const Scalar_type *__restrict__ r) {
+    *y = (*l) * (*r);
+  }
+  friend accelerator_inline void sub(Grid_simd2 *__restrict__ y,
+				     const Grid_simd2 *__restrict__ l,
+				     const Scalar_type *__restrict__ r) {
+    *y = (*l) - (*r);
+  }
+  friend accelerator_inline void add(Grid_simd2 *__restrict__ y,
+				     const Grid_simd2 *__restrict__ l,
+				     const Scalar_type *__restrict__ r) {
+    *y = (*l) + (*r);
+  }
+
+  ////////////////////////////////////////////////////////////////////////
+  // FIXME:  gonna remove these load/store, get, set, prefetch
+  ////////////////////////////////////////////////////////////////////////
+  friend accelerator_inline void vset(Grid_simd2 &ret, Scalar_type *a) {
+    for(int n=0;n<nvec;n++) vset(ret.v[n],a);
+  }
+
+  ///////////////////////
+  // Vstore
+  ///////////////////////
+  friend accelerator_inline void vstore(const Grid_simd2 &ret, Scalar_type *a) {
+    for(int n=0;n<nvec;n++) vstore(ret.v[n],a);
+  }
+
+  ///////////////////////
+  // Vprefetch
+  ///////////////////////
+  friend accelerator_inline void vprefetch(const Grid_simd2 &v) {
+    vprefetch(v.v[0]);
+  }
+
+  ///////////////////////
+  // Reduce
+  ///////////////////////
+  friend accelerator_inline Scalar_type Reduce(const Grid_simd2 &in) {
+    return Reduce(in.v[0])+ Reduce(in.v[1]);
+  }
+
+  ////////////////////////////
+  // operator scalar * simd
+  ////////////////////////////
+  friend accelerator_inline Grid_simd2 operator*(const Scalar_type &a, Grid_simd2 b) {
+    Grid_simd2 va;
+    vsplat(va, a);
+    return va * b;
+  }
+  friend accelerator_inline Grid_simd2 operator*(Grid_simd2 b, const Scalar_type &a) {
+    return a * b;
+  }
+
+  //////////////////////////////////
+  // Divides
+  //////////////////////////////////
+  friend accelerator_inline Grid_simd2 operator/(const Scalar_type &a, Grid_simd2 b) {
+    Grid_simd2 va;
+    vsplat(va, a);
+    return va / b;
+  }
+  friend accelerator_inline Grid_simd2 operator/(Grid_simd2 b, const Scalar_type &a) {
+    Grid_simd2 va;
+    vsplat(va, a);
+    return b / a;
+  }
+
+  ///////////////////////
+  // Unary negation
+  ///////////////////////
+  friend accelerator_inline Grid_simd2 operator-(const Grid_simd2 &r) {
+    Grid_simd2 ret;
+    vzero(ret);
+    ret = ret - r;
+    return ret;
+  }
+  // *=,+=,-= operators
+  accelerator_inline Grid_simd2 &operator*=(const Grid_simd2 &r) {
+    *this = (*this) * r;
+    return *this;
+  }
+  accelerator_inline Grid_simd2 &operator+=(const Grid_simd2 &r) {
+    *this = *this + r;
+    return *this;
+  }
+  accelerator_inline Grid_simd2 &operator-=(const Grid_simd2 &r) {
+    *this = *this - r;
+    return *this;
+  }
+
+  ///////////////////////////////////////
+  // Not all functions are supported
+  // through SIMD and must breakout to
+  // scalar type and back again. This
+  // provides support
+  ///////////////////////////////////////
+  template <class functor>
+  friend accelerator_inline Grid_simd2 SimdApply(const functor &func, const Grid_simd2 &v) {
+    Grid_simd2 ret;
+    for(int n=0;n<nvec;n++){
+      ret.v[n]=SimdApply(func,v.v[n]);
+    }
+    return ret;
+  }
+  template <class functor>
+  friend accelerator_inline Grid_simd2 SimdApplyBinop(const functor &func,
+                                         const Grid_simd2 &x,
+                                         const Grid_simd2 &y) {
+    Grid_simd2 ret;
+    for(int n=0;n<nvec;n++){
+      ret.v[n]=SimdApplyBinop(func,x.v[n],y.v[n]);
+    }
+    return ret;
+  }
+  ///////////////////////
+  // Exchange
+  // Al Ah , Bl Bh -> Al Bl Ah,Bh
+  ///////////////////////
+  friend accelerator_inline void exchange0(Grid_simd2 &out1,Grid_simd2 &out2,Grid_simd2 in1,Grid_simd2 in2){
+      out1.v[0] = in1.v[0];
+      out1.v[1] = in2.v[0];
+      out2.v[0] = in1.v[1];
+      out2.v[1] = in2.v[1];
+  }
+  friend accelerator_inline void exchange1(Grid_simd2 &out1,Grid_simd2 &out2,Grid_simd2 in1,Grid_simd2 in2){
+    exchange0(out1.v[0],out2.v[0],in1.v[0],in2.v[0]);
+    exchange0(out1.v[1],out2.v[1],in1.v[1],in2.v[1]);
+  }
+  friend accelerator_inline void exchange2(Grid_simd2 &out1,Grid_simd2 &out2,Grid_simd2 in1,Grid_simd2 in2){
+    exchange1(out1.v[0],out2.v[0],in1.v[0],in2.v[0]);
+    exchange1(out1.v[1],out2.v[1],in1.v[1],in2.v[1]);
+  }
+  friend accelerator_inline void exchange3(Grid_simd2 &out1,Grid_simd2 &out2,Grid_simd2 in1,Grid_simd2 in2){
+    exchange2(out1.v[0],out2.v[0],in1.v[0],in2.v[0]);
+    exchange2(out1.v[1],out2.v[1],in1.v[1],in2.v[1]);
+  }
+  friend accelerator_inline void exchange4(Grid_simd2 &out1,Grid_simd2 &out2,Grid_simd2 in1,Grid_simd2 in2){
+    exchange3(out1.v[0],out2.v[0],in1.v[0],in2.v[0]);
+    exchange3(out1.v[1],out2.v[1],in1.v[1],in2.v[1]);
+  }
+  friend accelerator_inline void exchange(Grid_simd2 &out1,Grid_simd2 &out2,Grid_simd2 in1,Grid_simd2 in2,int n)
+  {
+    if       (n==3) {
+      exchange3(out1,out2,in1,in2);
+    } else if(n==2) {
+      exchange2(out1,out2,in1,in2);
+    } else if(n==1) {
+      exchange1(out1,out2,in1,in2);
+    } else if(n==0) {
+      exchange0(out1,out2,in1,in2);
+    }
+  }
+  ////////////////////////////////////////////////////////////////////
+  // General permute; assumes vector length is same across
+  // all subtypes; may not be a good assumption, but could
+  // add the vector width as a template param for BG/Q for example
+  ////////////////////////////////////////////////////////////////////
+  friend accelerator_inline void permute0(Grid_simd2 &y, Grid_simd2 b) {
+    y.v[0]=b.v[1];
+    y.v[1]=b.v[0];
+  }
+  friend accelerator_inline void permute1(Grid_simd2 &y, Grid_simd2 b) {
+    permute0(y.v[0],b.v[0]);
+    permute0(y.v[1],b.v[1]);
+  }
+  friend accelerator_inline void permute2(Grid_simd2 &y, Grid_simd2 b) {
+    permute1(y.v[0],b.v[0]);
+    permute1(y.v[1],b.v[1]);
+  }
+  friend accelerator_inline void permute3(Grid_simd2 &y, Grid_simd2 b) {
+    permute2(y.v[0],b.v[0]);
+    permute2(y.v[1],b.v[1]);
+  }
+  friend accelerator_inline void permute4(Grid_simd2 &y, Grid_simd2 b) {
+    permute3(y.v[0],b.v[0]);
+    permute3(y.v[1],b.v[1]);
+  }
+  friend accelerator_inline void permute(Grid_simd2 &y, Grid_simd2 b, int perm) {
+    if(perm==3) permute3(y, b);
+    else if(perm==2) permute2(y, b);
+    else if(perm==1) permute1(y, b);
+    else if(perm==0) permute0(y, b);
+  }
+
+  ///////////////////////////////
+  // Getting single lanes
+  ///////////////////////////////
+  accelerator_inline Scalar_type getlane(int lane) const {
+    if(lane < vector_type::Nsimd() ) return v[0].getlane(lane);
+    else                             return v[1].getlane(lane%vector_type::Nsimd());
+  }
+
+  accelerator_inline void putlane(const Scalar_type &S, int lane){
+    if(lane < vector_type::Nsimd() ) v[0].putlane(S,lane);
+    else                             v[1].putlane(S,lane%vector_type::Nsimd());
+  }
+};  // end of Grid_simd2 class definition
+
+///////////////////////////////
+// Define available types
+///////////////////////////////
+
+typedef Grid_simd2<complex<double>  , vComplexD>  vComplexD2;
+typedef Grid_simd2<double           , vRealD>     vRealD2;
+
+
+
+/////////////////////////////////////////
+// Some traits to recognise the types
+/////////////////////////////////////////
+template <typename T>
+struct is_simd : public std::false_type {};
+template <> struct is_simd<vRealF>     : public std::true_type {};
+template <> struct is_simd<vRealD>     : public std::true_type {};
+template <> struct is_simd<vRealH>     : public std::true_type {};
+template <> struct is_simd<vComplexF>  : public std::true_type {};
+template <> struct is_simd<vComplexD>  : public std::true_type {};
+template <> struct is_simd<vComplexH>  : public std::true_type {};
+template <> struct is_simd<vInteger>   : public std::true_type {};
+template <> struct is_simd<vRealD2>    : public std::true_type {};
+template <> struct is_simd<vComplexD2> : public std::true_type {};
+
+template <typename T> using IfSimd    = Invoke<std::enable_if<is_simd<T>::value, int> >;
+template <typename T> using IfNotSimd = Invoke<std::enable_if<!is_simd<T>::value, unsigned> >;
+
+///////////////////////////////////////////////
+// insert / extract with complex support
+///////////////////////////////////////////////
+template <class S, class V>
+accelerator_inline S getlane(const Grid_simd<S, V> &in,int lane) {
+  return in.getlane(lane);
+}
+template <class S, class V>
+accelerator_inline void putlane(Grid_simd<S, V> &vec,const S &_S, int lane){
+  vec.putlane(_S,lane);
+}
+template <class S,IfNotSimd<S> = 0 >
+accelerator_inline S getlane(const S &in,int lane) {
+  return in;
+}
+template <class S,IfNotSimd<S> = 0 >
+accelerator_inline void putlane(S &vec,const S &_S, int lane){
+  vec = _S;
+}
+template <class S, class V>
+accelerator_inline S getlane(const Grid_simd2<S, V> &in,int lane) {
+  return in.getlane(lane);
+}
+template <class S, class V>
+accelerator_inline void putlane(Grid_simd2<S, V> &vec,const S &_S, int lane){
+  vec.putlane(_S,lane);
+}
+
+
+////////////////////////////////////////////////////////////////////
+// General rotate
+////////////////////////////////////////////////////////////////////
+
+template <class S, class V>
+accelerator_inline void vbroadcast(Grid_simd2<S,V> &ret,const Grid_simd2<S,V> &src,int lane){
+  S* typepun =(S*) &src;
+  vsplat(ret,typepun[lane]);
+}
+template <class S, class V, IfComplex<S> =0>
+accelerator_inline void rbroadcast(Grid_simd2<S,V> &ret,const Grid_simd2<S,V> &src,int lane){
+  typedef typename V::vector_type vector_type;
+  S* typepun =(S*) &src;
+  ret.v[0].v = unary<vector_type>(real(typepun[lane]), VsplatSIMD());
+  ret.v[1].v = unary<vector_type>(real(typepun[lane]), VsplatSIMD());
+}
+
+
+///////////////////////
+// Splat
+///////////////////////
+
+// this is only for the complex version
+template <class S, class V, IfComplex<S> = 0, class ABtype>
+accelerator_inline void vsplat(Grid_simd2<S, V> &ret, ABtype a, ABtype b) {
+  vsplat(ret.v[0],a,b);
+  vsplat(ret.v[1],a,b);
+}
+
+// overload if complex
+template <class S, class V>
+accelerator_inline void vsplat(Grid_simd2<S, V> &ret, EnableIf<is_complex<S>, S> c) {
+  vsplat(ret, real(c), imag(c));
+}
+template <class S, class V>
+accelerator_inline void rsplat(Grid_simd2<S, V> &ret, EnableIf<is_complex<S>, S> c) {
+  vsplat(ret, real(c), real(c));
+}
+
+// if real fill with a, if complex fill with a in the real part (first function
+// above)
+template <class S, class V>
+accelerator_inline void vsplat(Grid_simd2<S, V> &ret, NotEnableIf<is_complex<S>, S> a)
+{
+  vsplat(ret.v[0],a);
+  vsplat(ret.v[1],a);
+}
+//////////////////////////
+
+///////////////////////////////////////////////
+// Initialise to 1,0,i for the correct types
+///////////////////////////////////////////////
+// For complex types
+template <class S, class V, IfComplex<S> = 0>
+accelerator_inline void vone(Grid_simd2<S, V> &ret) {
+  vsplat(ret, S(1.0, 0.0));
+}
+template <class S, class V, IfComplex<S> = 0>
+accelerator_inline void vzero(Grid_simd2<S, V> &ret) {
+  vsplat(ret, S(0.0, 0.0));
+}  // use xor?
+template <class S, class V, IfComplex<S> = 0>
+accelerator_inline void vcomplex_i(Grid_simd2<S, V> &ret) {
+  vsplat(ret, S(0.0, 1.0));
+}
+
+template <class S, class V, IfComplex<S> = 0>
+accelerator_inline void visign(Grid_simd2<S, V> &ret) {
+  vsplat(ret, S(1.0, -1.0));
+}
+template <class S, class V, IfComplex<S> = 0>
+accelerator_inline void vrsign(Grid_simd2<S, V> &ret) {
+  vsplat(ret, S(-1.0, 1.0));
+}
+
+// if not complex overload here
+template <class S, class V, IfReal<S> = 0>
+accelerator_inline void vone(Grid_simd2<S, V> &ret) {
+  vsplat(ret, S(1.0));
+}
+template <class S, class V, IfReal<S> = 0>
+accelerator_inline void vzero(Grid_simd2<S, V> &ret) {
+  vsplat(ret, S(0.0));
+}
+
+// For integral types
+template <class S, class V, IfInteger<S> = 0>
+accelerator_inline void vone(Grid_simd2<S, V> &ret) {
+  vsplat(ret, 1);
+}
+template <class S, class V, IfInteger<S> = 0>
+accelerator_inline void vzero(Grid_simd2<S, V> &ret) {
+  vsplat(ret, 0);
+}
+template <class S, class V, IfInteger<S> = 0>
+accelerator_inline void vtrue(Grid_simd2<S, V> &ret) {
+  vsplat(ret, 0xFFFFFFFF);
+}
+template <class S, class V, IfInteger<S> = 0>
+accelerator_inline void vfalse(Grid_simd2<S, V> &ret) {
+  vsplat(ret, 0);
+}
+template <class S, class V>
+accelerator_inline void zeroit(Grid_simd2<S, V> &z) {
+  vzero(z);
+}
+
+///////////////////////
+// Vstream
+///////////////////////
+template <class S, class V, IfReal<S> = 0>
+accelerator_inline void vstream(Grid_simd2<S, V> &out, const Grid_simd2<S, V> &in) {
+  vstream(out.v[0],in.v[0]);
+  vstream(out.v[1],in.v[1]);
+}
+template <class S, class V, IfComplex<S> = 0>
+accelerator_inline void vstream(Grid_simd2<S, V> &out, const Grid_simd2<S, V> &in) {
+  vstream(out.v[0],in.v[0]);
+  vstream(out.v[1],in.v[1]);
+}
+template <class S, class V, IfInteger<S> = 0>
+accelerator_inline void vstream(Grid_simd2<S, V> &out, const Grid_simd2<S, V> &in) {
+  vstream(out.v[0],in.v[0]);
+  vstream(out.v[1],in.v[1]);
+}
+
+////////////////////////////////////
+// Arithmetic operator overloads +,-,*
+////////////////////////////////////
+template <class S, class V>
+accelerator_inline Grid_simd2<S, V> operator+(Grid_simd2<S, V> a, Grid_simd2<S, V> b) {
+  Grid_simd2<S, V> ret;
+  ret.v[0] = a.v[0]+b.v[0];
+  ret.v[1] = a.v[1]+b.v[1];
+  return ret;
+};
+
+template <class S, class V>
+accelerator_inline Grid_simd2<S, V> operator-(Grid_simd2<S, V> a, Grid_simd2<S, V> b) {
+  Grid_simd2<S, V> ret;
+  ret.v[0] = a.v[0]-b.v[0];
+  ret.v[1] = a.v[1]-b.v[1];
+  return ret;
+};
+
+// Distinguish between complex types and others
+template <class S, class V, IfComplex<S> = 0>
+accelerator_inline Grid_simd2<S, V> real_mult(Grid_simd2<S, V> a, Grid_simd2<S, V> b) {
+  Grid_simd2<S, V> ret;
+  ret.v[0] =real_mult(a.v[0],b.v[0]);
+  ret.v[1] =real_mult(a.v[1],b.v[1]);
+  return ret;
+};
+template <class S, class V, IfComplex<S> = 0>
+accelerator_inline Grid_simd2<S, V> real_madd(Grid_simd2<S, V> a, Grid_simd2<S, V> b, Grid_simd2<S,V> c) {
+  Grid_simd2<S, V> ret;
+  ret.v[0] =real_madd(a.v[0],b.v[0],c.v[0]);
+  ret.v[1] =real_madd(a.v[1],b.v[1],c.v[1]);
+  return ret;
+};
+
+
+// Distinguish between complex types and others
+template <class S, class V>
+accelerator_inline Grid_simd2<S, V> operator*(Grid_simd2<S, V> a, Grid_simd2<S, V> b) {
+  Grid_simd2<S, V> ret;
+  ret.v[0] = a.v[0]*b.v[0];
+  ret.v[1] = a.v[1]*b.v[1];
+  return ret;
+};
+
+// Distinguish between complex types and others
+template <class S, class V>
+accelerator_inline Grid_simd2<S, V> operator/(Grid_simd2<S, V> a, Grid_simd2<S, V> b) {
+  Grid_simd2<S, V> ret;
+  ret.v[0] = a.v[0]/b.v[0];
+  ret.v[1] = a.v[1]/b.v[1];
+  return ret;
+};
+
+///////////////////////
+// Conjugate
+///////////////////////
+template <class S, class V>
+accelerator_inline Grid_simd2<S, V> conjugate(const Grid_simd2<S, V> &in) {
+  Grid_simd2<S, V> ret;
+  ret.v[0] = conjugate(in.v[0]);
+  ret.v[1] = conjugate(in.v[1]);
+  return ret;
+}
+template <class S, class V, IfNotInteger<S> = 0>
+accelerator_inline Grid_simd2<S, V> adj(const Grid_simd2<S, V> &in) {
+  return conjugate(in);
+}
+
+///////////////////////
+// timesMinusI
+///////////////////////
+template <class S, class V>
+accelerator_inline void timesMinusI(Grid_simd2<S, V> &ret, const Grid_simd2<S, V> &in) {
+  timesMinusI(ret.v[0],in.v[0]);
+  timesMinusI(ret.v[1],in.v[1]);
+}
+template <class S, class V>
+accelerator_inline Grid_simd2<S, V> timesMinusI(const Grid_simd2<S, V> &in) {
+  Grid_simd2<S, V> ret;
+  timesMinusI(ret.v[0],in.v[0]);
+  timesMinusI(ret.v[1],in.v[1]);
+  return ret;
+}
+
+///////////////////////
+// timesI
+///////////////////////
+template <class S, class V>
+accelerator_inline void timesI(Grid_simd2<S, V> &ret, const Grid_simd2<S, V> &in) {
+  timesI(ret.v[0],in.v[0]);
+  timesI(ret.v[1],in.v[1]);
+}
+template <class S, class V>
+accelerator_inline Grid_simd2<S, V> timesI(const Grid_simd2<S, V> &in) {
+  Grid_simd2<S, V> ret;
+  timesI(ret.v[0],in.v[0]);
+  timesI(ret.v[1],in.v[1]);
+  return ret;
+}
+
+/////////////////////
+// Inner, outer
+/////////////////////
+template <class S, class V>
+accelerator_inline Grid_simd2<S, V> innerProduct(const Grid_simd2<S, V> &l,const Grid_simd2<S, V> &r) {
+  return conjugate(l) * r;
+}
+template <class S, class V>
+accelerator_inline Grid_simd2<S, V> outerProduct(const Grid_simd2<S, V> &l,const Grid_simd2<S, V> &r) {
+  return l * conjugate(r);
+}
+
+template <class S, class V>
+accelerator_inline Grid_simd2<S, V> trace(const Grid_simd2<S, V> &arg) {
+  return arg;
+}
+
+////////////////////////////////////////////////////////////
+// copy/splat complex real parts into real;
+// insert real into complex and zero imag;
+////////////////////////////////////////////////////////////
+accelerator_inline void precisionChange(vComplexD2 &out,const vComplexF  &in){
+  Optimization::PrecisionChange::StoD(in.v,out.v[0].v,out.v[1].v);
+}
+accelerator_inline void precisionChange(vComplexF  &out,const vComplexD2 &in){
+  out.v=Optimization::PrecisionChange::DtoS(in.v[0].v,in.v[1].v);
+}
+accelerator_inline void precisionChange(vComplexD2 *out,const vComplexF  *in,int nvec){
+  for(int m=0;m<nvec;m++){ precisionChange(out[m],in[m]); }
+}
+accelerator_inline void precisionChange(vComplexF  *out,const vComplexD2 *in,int nvec){
+  for(int m=0;m<nvec;m++){ precisionChange(out[m],in[m]); }
+}
+
+accelerator_inline void precisionChange(vRealD2 &out,const vRealF  &in){
+  Optimization::PrecisionChange::StoD(in.v,out.v[0].v,out.v[1].v);
+}
+accelerator_inline void precisionChange(vRealF  &out,const vRealD2 &in){
+  out.v=Optimization::PrecisionChange::DtoS(in.v[0].v,in.v[1].v);
+}
+accelerator_inline void precisionChange(vRealD2 *out,const vRealF  *in,int nvec){
+  for(int m=0;m<nvec;m++){ precisionChange(out[m],in[m]); }
+}
+accelerator_inline void precisionChange(vRealF  *out,const vRealD2 *in,int nvec){
+  for(int m=0;m<nvec;m++){ precisionChange(out[m],in[m]); }
+}
+
+NAMESPACE_END(Grid);
+
+

From ca62abd203006a40fdf5da2e9b5e3d74cfc76d02 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 15 Nov 2022 16:36:46 -0500
Subject: [PATCH 078/146] Record some perturbative free field calculation

---
 examples/Example_taku.cc  | 383 +++++++++++++++++++++++++++++++++
 examples/Example_taku2.cc | 433 ++++++++++++++++++++++++++++++++++++++
 2 files changed, 816 insertions(+)
 create mode 100644 examples/Example_taku.cc
 create mode 100644 examples/Example_taku2.cc

diff --git a/examples/Example_taku.cc b/examples/Example_taku.cc
new file mode 100644
index 00000000..b9ad272e
--- /dev/null
+++ b/examples/Example_taku.cc
@@ -0,0 +1,383 @@
+/*
+ * Warning: This code illustrative only: not well tested, and not meant for production use
+ * without regression / tests being applied
+ */
+
+#include <Grid/Grid.h>
+
+using namespace std;
+using namespace Grid;
+
+RealD LLscale =1.0;
+RealD LCscale =1.0;
+
+template<class Gimpl,class Field> class CovariantLaplacianCshift : public SparseMatrixBase<Field>
+{
+public:
+  INHERIT_GIMPL_TYPES(Gimpl);
+
+  GridBase *grid;
+  GaugeField U;
+  
+  CovariantLaplacianCshift(GaugeField &_U)    :
+    grid(_U.Grid()),
+    U(_U) {  };
+
+  virtual GridBase *Grid(void) { return grid; };
+
+  virtual void  M    (const Field &in, Field &out)
+  {
+    out=Zero();
+    for(int mu=0;mu<Nd-1;mu++) {
+      GaugeLinkField Umu = PeekIndex<LorentzIndex>(U, mu); // NB: Inefficent
+      out = out - Gimpl::CovShiftForward(Umu,mu,in);    
+      out = out - Gimpl::CovShiftBackward(Umu,mu,in);    
+      out = out + 2.0*in;
+    }
+  };
+  virtual void  Mdag (const Field &in, Field &out) { M(in,out);}; // Laplacian is hermitian
+  virtual  void Mdiag    (const Field &in, Field &out)                  {assert(0);}; // Unimplemented need only for multigrid
+  virtual  void Mdir     (const Field &in, Field &out,int dir, int disp){assert(0);}; // Unimplemented need only for multigrid
+  virtual  void MdirAll  (const Field &in, std::vector<Field> &out)     {assert(0);}; // Unimplemented need only for multigrid
+};
+
+void MakePhase(Coordinate mom,LatticeComplex &phase)
+{
+  GridBase *grid = phase.Grid();
+  auto latt_size = grid->GlobalDimensions();
+  ComplexD ci(0.0,1.0);
+  phase=Zero();
+
+  LatticeComplex coor(phase.Grid());
+  for(int mu=0;mu<Nd;mu++){
+    RealD TwoPiL =  M_PI * 2.0/ latt_size[mu];
+    LatticeCoordinate(coor,mu);
+    phase = phase + (TwoPiL * mom[mu]) * coor;
+  }
+  phase = exp(phase*ci);
+}
+void PointSource(Coordinate &coor,LatticePropagator &source)
+{
+  //  Coordinate coor({0,0,0,0});
+  source=Zero();
+  SpinColourMatrix kronecker; kronecker=1.0;
+  pokeSite(kronecker,source,coor);
+}
+void Z2WallSource(GridParallelRNG &RNG,int tslice,LatticePropagator &source)
+{
+  GridBase *grid = source.Grid();
+  LatticeComplex noise(grid);
+  LatticeComplex zz(grid); zz=Zero();
+  LatticeInteger t(grid);
+
+  RealD nrm=1.0/sqrt(2);
+  bernoulli(RNG, noise); // 0,1 50:50
+
+  noise = (2.*noise - Complex(1,1))*nrm;
+
+  LatticeCoordinate(t,Tdir);
+  noise = where(t==Integer(tslice), noise, zz);
+
+  source = 1.0;
+  source = source*noise;
+  std::cout << " Z2 wall " << norm2(source) << std::endl;
+}
+template<class Field>
+void GaussianSmear(LatticeGaugeField &U,Field &unsmeared,Field &smeared)
+{
+  typedef CovariantLaplacianCshift <PeriodicGimplR,Field> Laplacian_t;
+  Laplacian_t Laplacian(U);
+
+  Integer Iterations = 40;
+  Real width = 2.0;
+  Real coeff = (width*width) / Real(4*Iterations);
+
+  Field tmp(U.Grid());
+  smeared=unsmeared;
+  //  chi = (1-p^2/2N)^N kronecker
+  for(int n = 0; n < Iterations; ++n) {
+    Laplacian.M(smeared,tmp);
+    smeared = smeared - coeff*tmp;
+    std::cout << " smear iter " << n<<" " <<norm2(smeared)<<std::endl;
+  }
+}
+void GaussianSource(Coordinate &site,LatticeGaugeField &U,LatticePropagator &source)
+{
+  LatticePropagator tmp(source.Grid());
+  PointSource(site,source);
+  std::cout << " GaussianSource Kronecker "<< norm2(source)<<std::endl;
+  tmp = source;
+  GaussianSmear(U,tmp,source);
+  std::cout << " GaussianSource Smeared "<< norm2(source)<<std::endl;
+}
+void GaussianWallSource(GridParallelRNG &RNG,int tslice,LatticeGaugeField &U,LatticePropagator &source)
+{
+  Z2WallSource(RNG,tslice,source);
+  auto tmp = source;
+  GaussianSmear(U,tmp,source);
+}
+void SequentialSource(int tslice,Coordinate &mom,LatticePropagator &spectator,LatticePropagator &source)
+{
+  assert(mom.size()==Nd);
+  assert(mom[Tdir] == 0);
+
+  GridBase * grid = spectator.Grid();
+
+
+  LatticeInteger ts(grid);
+  LatticeCoordinate(ts,Tdir);
+  source = Zero();
+  source = where(ts==Integer(tslice),spectator,source); // Stick in a slice of the spectator, zero everywhere else
+
+  LatticeComplex phase(grid);
+  MakePhase(mom,phase);
+
+  source = source *phase;
+}
+template<class Action>
+void Solve(Action &D,LatticePropagator &source,LatticePropagator &propagator)
+{
+  GridBase *UGrid = D.GaugeGrid();
+  GridBase *FGrid = D.FermionGrid();
+
+  LatticeFermion src4  (UGrid); 
+  LatticeFermion src5  (FGrid); 
+  LatticeFermion result5(FGrid);
+  LatticeFermion result4(UGrid);
+  LatticePropagator prop5(FGrid);
+  
+  ConjugateGradient<LatticeFermion> CG(1.0e-8,100000);
+  SchurRedBlackDiagMooeeSolve<LatticeFermion> schur(CG);
+  ZeroGuesser<LatticeFermion> ZG; // Could be a DeflatedGuesser if have eigenvectors
+   for(int s=0;s<Nd;s++){
+    for(int c=0;c<Nc;c++){
+      PropToFerm<Action>(src4,source,s,c);
+
+      D.ImportPhysicalFermionSource(src4,src5);
+
+      result5=Zero();
+      schur(D,src5,result5,ZG);
+      std::cout<<GridLogMessage
+	       <<"spin "<<s<<" color "<<c
+	       <<" norm2(src5d) "   <<norm2(src5)
+               <<" norm2(result5d) "<<norm2(result5)<<std::endl;
+
+      D.ExportPhysicalFermionSolution(result5,result4);
+
+      FermToProp<Action>(prop5,result5,s,c);
+      FermToProp<Action>(propagator,result4,s,c);
+    }
+  }
+  LatticePropagator Axial_mu(UGrid); 
+  LatticePropagator Vector_mu(UGrid); 
+
+  LatticeComplex    PA (UGrid); 
+  LatticeComplex    VV (UGrid); 
+  LatticeComplex    PJ5q(UGrid);
+  LatticeComplex    PP (UGrid);
+
+  std::vector<TComplex> sumPA;
+  std::vector<TComplex> sumVV;
+  std::vector<TComplex> sumPP;
+  std::vector<TComplex> sumPJ5q;
+
+  Gamma g5(Gamma::Algebra::Gamma5);
+  D.ContractConservedCurrent(prop5,prop5,Axial_mu,source,Current::Axial,Tdir);
+  PA       = trace(g5*Axial_mu);      // Pseudoscalar-Axial conserved current
+  sliceSum(PA,sumPA,Tdir);
+
+  int Nt{static_cast<int>(sumPA.size())};
+
+  for(int t=0;t<Nt;t++) std::cout<<GridLogMessage <<"PAc["<<t<<"] "<<real(TensorRemove(sumPA[t]))*LCscale<<std::endl;
+
+  PP       = trace(adj(propagator)*propagator); // Pseudoscalar density
+  sliceSum(PP,sumPP,Tdir);
+  for(int t=0;t<Nt;t++) std::cout<<GridLogMessage <<"PP["<<t<<"] "<<real(TensorRemove(sumPP[t]))*LCscale<<std::endl;
+  
+  D.ContractJ5q(prop5,PJ5q);
+  sliceSum(PJ5q,sumPJ5q,Tdir);
+  for(int t=0;t<Nt;t++) std::cout<<GridLogMessage <<"PJ5q["<<t<<"] "<<real(TensorRemove(sumPJ5q[t]))<<std::endl;
+
+  Gamma::Algebra GammaV[3] = {
+    Gamma::Algebra::GammaX,
+    Gamma::Algebra::GammaY,
+    Gamma::Algebra::GammaZ
+  };
+  for( int mu=0;mu<3;mu++ ) {
+    Gamma gV(GammaV[mu]);
+    D.ContractConservedCurrent(prop5,prop5,Vector_mu,source,Current::Vector,mu);
+    //    auto ss=sliceSum(Vector_mu,Tdir);
+    //    for(int t=0;t<Nt;t++) std::cout<<GridLogMessage <<"ss["<<mu<<"]["<<t<<"] "<<ss[t]<<std::endl;
+    VV       = trace(gV*Vector_mu);     // (local) Vector-Vector conserved current
+    sliceSum(VV,sumVV,Tdir);
+    for(int t=0;t<Nt;t++){
+      RealD Ct = real(TensorRemove(sumVV[t]))*LCscale;
+      std::cout<<GridLogMessage <<"VVc["<<mu<<"]["<<t<<"] "<< Ct
+	       << " 2 pi^2 t^3 C(t) "<< 2 * M_PI *M_PI * t*t*t *Ct<<std::endl;
+    }
+  }
+
+}
+
+class MesonFile: Serializable {
+public:
+  GRID_SERIALIZABLE_CLASS_MEMBERS(MesonFile, std::vector<std::vector<Complex> >, data);
+};
+
+void MesonTrace(std::string file,LatticePropagator &q1,LatticePropagator &q2,LatticeComplex &phase)
+{
+  const int nchannel=3;
+  Gamma::Algebra Gammas[nchannel][2] = {
+    {Gamma::Algebra::GammaX,Gamma::Algebra::GammaX},
+    {Gamma::Algebra::GammaY,Gamma::Algebra::GammaY},
+    {Gamma::Algebra::GammaZ,Gamma::Algebra::GammaZ}
+  };
+
+  Gamma G5(Gamma::Algebra::Gamma5);
+
+  LatticeComplex meson_CF(q1.Grid());
+  MesonFile MF;
+
+  for(int ch=0;ch<nchannel;ch++){
+
+    Gamma Gsrc(Gammas[ch][0]);
+    Gamma Gsnk(Gammas[ch][1]);
+
+    meson_CF = trace(G5*adj(q1)*G5*Gsnk*q2*adj(Gsrc));
+
+    std::vector<TComplex> meson_T;
+    sliceSum(meson_CF,meson_T, Tdir);
+
+    int nt=meson_T.size();
+
+    std::vector<Complex> corr(nt);
+    for(int t=0;t<nt;t++){
+      corr[t] = TensorRemove(meson_T[t])*LLscale; // Yes this is ugly, not figured a work around
+      std::cout << " channel "<<ch<<" t "<<t<<" " <<real(corr[t])<< " 2 pi^2 t^3 C(t) "<< 2 * M_PI *M_PI * t*t*t *real(corr[t])<<std::endl;
+    }
+    MF.data.push_back(corr);
+  }
+
+  {
+    XmlWriter WR(file);
+    write(WR,"MesonFile",MF);
+  }
+}
+
+int main (int argc, char ** argv)
+{
+  const int Ls=32;
+
+  Grid_init(&argc,&argv);
+
+  // Double precision grids
+  GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), 
+								   GridDefaultSimd(Nd,vComplex::Nsimd()),
+								   GridDefaultMpi());
+  GridRedBlackCartesian * UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
+  GridCartesian         * FGrid   = SpaceTimeGrid::makeFiveDimGrid(Ls,UGrid);
+  GridRedBlackCartesian * FrbGrid = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGrid);
+
+  //////////////////////////////////////////////////////////////////////
+  // You can manage seeds however you like.
+  // Recommend SeedUniqueString.
+  //////////////////////////////////////////////////////////////////////
+  std::vector<int> seeds4({1,2,3,4}); 
+  GridParallelRNG          RNG4(UGrid);  RNG4.SeedFixedIntegers(seeds4);
+
+  LatticeGaugeField Umu(UGrid);
+  std::string config;
+  RealD M5=1.8;
+  if( argc > 1 && argv[1][0] != '-' )
+  {
+    std::cout<<GridLogMessage <<"Loading configuration from "<<argv[1]<<std::endl;
+    FieldMetaData header;
+    NerscIO::readConfiguration(Umu, header, argv[1]);
+    config=argv[1];
+    M5=1.8;
+  }
+  else
+  {
+    SU<Nc>::ColdConfiguration(Umu);
+    config="ColdConfig";
+    //    RealD P=1.0; // Don't scale
+    RealD P=0.5871119; // 48I
+    //    RealD P=0.6153342; // 64I
+    //    RealD P=0.6388238 // 32Ifine
+    RealD u0 = sqrt(sqrt(P));
+    RealD M5mf = M5 - 4.0*(1.0-u0);
+    RealD w0   = 1.0 - M5mf;
+#if 0
+    // M5=1.8 with U=u0
+    Umu = Umu * u0;
+    LLscale = 1.0;
+    LCscale = 1.0;
+    std::cout<<GridLogMessage <<"Gauge links are u=u0= "<<u0<<std::endl;
+    std::cout<<GridLogMessage <<"M5 =  "<<M5<<std::endl;
+#else
+    M5 = M5mf;
+    std::cout<<GridLogMessage <<"Gauge links are u=1  "<<std::endl;
+    std::cout<<GridLogMessage <<"u0="<<u0<<std::endl;
+    std::cout<<GridLogMessage <<"M5=M5mf =  "<<M5<<std::endl;
+    LLscale = 1.0/(1-w0*w0)/(1-w0*w0);
+    LCscale = 1.0/(1-w0*w0)/(1-w0*w0);
+#endif
+    std::cout<<GridLogMessage <<"LLscale =  "<<LLscale<<std::endl;
+    std::cout<<GridLogMessage <<"LCscale =  "<<LCscale<<std::endl;
+  }
+
+  std::vector<RealD> masses({ 0.00} ); // u/d, s, c ??
+
+  int nmass = masses.size();
+
+  std::vector<MobiusFermionD *> FermActs;
+  
+  std::cout<<GridLogMessage <<"======================"<<std::endl;
+  std::cout<<GridLogMessage <<"MobiusFermion action as Scaled Shamir kernel"<<std::endl;
+  std::cout<<GridLogMessage <<"======================"<<std::endl;
+
+  for(auto mass: masses) {
+
+    RealD b=1.5;// Scale factor b+c=2, b-c=1
+    RealD c=0.5;
+    
+    FermActs.push_back(new MobiusFermionD(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,b,c));
+   
+  }
+
+  LatticePropagator point_source(UGrid);
+  //  LatticePropagator wall_source(UGrid);
+
+  Coordinate Origin({0,0,0,0});
+  PointSource   (Origin,point_source);
+  //  Z2WallSource  (RNG4,0,wall_source);
+  
+  std::vector<LatticePropagator> PointProps(nmass,UGrid);
+  //  std::vector<LatticePropagator> GaussProps(nmass,UGrid);
+  //  std::vector<LatticePropagator> Z2Props   (nmass,UGrid);
+
+  for(int m=0;m<nmass;m++) {
+    
+    Solve(*FermActs[m],point_source   ,PointProps[m]);
+  }
+
+  LatticeComplex phase(UGrid);
+  Coordinate mom({0,0,0,0});
+  MakePhase(mom,phase);
+  
+  for(int m1=0 ;m1<nmass;m1++) {
+  for(int m2=m1;m2<nmass;m2++) {
+    std::stringstream ssp,ssg,ssz;
+
+    ssp<<config<< "_m" << m1 << "_m"<< m2 << "_point_meson.xml";
+    ssz<<config<< "_m" << m1 << "_m"<< m2 << "_wall_meson.xml";
+
+    MesonTrace(ssp.str(),PointProps[m1],PointProps[m2],phase);
+    //    MesonTrace(ssz.str(),Z2Props[m1],Z2Props[m2],phase);
+  }}
+
+  Grid_finalize();
+}
+
+
+
diff --git a/examples/Example_taku2.cc b/examples/Example_taku2.cc
new file mode 100644
index 00000000..75dea606
--- /dev/null
+++ b/examples/Example_taku2.cc
@@ -0,0 +1,433 @@
+/*
+ * Warning: This code illustrative only: not well tested, and not meant for production use
+ * without regression / tests being applied
+ */
+
+#include <Grid/Grid.h>
+
+using namespace std;
+using namespace Grid;
+
+RealD LLscale =1.0;
+RealD LCscale =1.0;
+
+template<class Gimpl,class Field> class CovariantLaplacianCshift : public SparseMatrixBase<Field>
+{
+public:
+  INHERIT_GIMPL_TYPES(Gimpl);
+
+  GridBase *grid;
+  GaugeField U;
+  
+  CovariantLaplacianCshift(GaugeField &_U)    :
+    grid(_U.Grid()),
+    U(_U) {  };
+
+  virtual GridBase *Grid(void) { return grid; };
+
+  virtual void  M    (const Field &in, Field &out)
+  {
+    out=Zero();
+    for(int mu=0;mu<Nd-1;mu++) {
+      GaugeLinkField Umu = PeekIndex<LorentzIndex>(U, mu); // NB: Inefficent
+      out = out - Gimpl::CovShiftForward(Umu,mu,in);    
+      out = out - Gimpl::CovShiftBackward(Umu,mu,in);    
+      out = out + 2.0*in;
+    }
+  };
+  virtual void  Mdag (const Field &in, Field &out) { M(in,out);}; // Laplacian is hermitian
+  virtual  void Mdiag    (const Field &in, Field &out)                  {assert(0);}; // Unimplemented need only for multigrid
+  virtual  void Mdir     (const Field &in, Field &out,int dir, int disp){assert(0);}; // Unimplemented need only for multigrid
+  virtual  void MdirAll  (const Field &in, std::vector<Field> &out)     {assert(0);}; // Unimplemented need only for multigrid
+};
+
+void MakePhase(Coordinate mom,LatticeComplex &phase)
+{
+  GridBase *grid = phase.Grid();
+  auto latt_size = grid->GlobalDimensions();
+  ComplexD ci(0.0,1.0);
+  phase=Zero();
+
+  LatticeComplex coor(phase.Grid());
+  for(int mu=0;mu<Nd;mu++){
+    RealD TwoPiL =  M_PI * 2.0/ latt_size[mu];
+    LatticeCoordinate(coor,mu);
+    phase = phase + (TwoPiL * mom[mu]) * coor;
+  }
+  phase = exp(phase*ci);
+}
+
+void PointSource(Coordinate &coor,LatticePropagator &source)
+{
+  //  Coordinate coor({0,0,0,0});
+  source=Zero();
+  SpinColourMatrix kronecker; kronecker=1.0;
+  pokeSite(kronecker,source,coor);
+}
+void Z2WallSource(GridParallelRNG &RNG,int tslice,LatticePropagator &source)
+{
+  GridBase *grid = source.Grid();
+  LatticeComplex noise(grid);
+  LatticeComplex zz(grid); zz=Zero();
+  LatticeInteger t(grid);
+
+  RealD nrm=1.0/sqrt(2);
+  bernoulli(RNG, noise); // 0,1 50:50
+
+  noise = (2.*noise - Complex(1,1))*nrm;
+
+  LatticeCoordinate(t,Tdir);
+  noise = where(t==Integer(tslice), noise, zz);
+
+  source = 1.0;
+  source = source*noise;
+  std::cout << " Z2 wall " << norm2(source) << std::endl;
+}
+template<class Field>
+void GaussianSmear(LatticeGaugeField &U,Field &unsmeared,Field &smeared)
+{
+  typedef CovariantLaplacianCshift <PeriodicGimplR,Field> Laplacian_t;
+  Laplacian_t Laplacian(U);
+
+  Integer Iterations = 40;
+  Real width = 2.0;
+  Real coeff = (width*width) / Real(4*Iterations);
+
+  Field tmp(U.Grid());
+  smeared=unsmeared;
+  //  chi = (1-p^2/2N)^N kronecker
+  for(int n = 0; n < Iterations; ++n) {
+    Laplacian.M(smeared,tmp);
+    smeared = smeared - coeff*tmp;
+    std::cout << " smear iter " << n<<" " <<norm2(smeared)<<std::endl;
+  }
+}
+void GaussianSource(Coordinate &site,LatticeGaugeField &U,LatticePropagator &source)
+{
+  LatticePropagator tmp(source.Grid());
+  PointSource(site,source);
+  std::cout << " GaussianSource Kronecker "<< norm2(source)<<std::endl;
+  tmp = source;
+  GaussianSmear(U,tmp,source);
+  std::cout << " GaussianSource Smeared "<< norm2(source)<<std::endl;
+}
+void GaussianWallSource(GridParallelRNG &RNG,int tslice,LatticeGaugeField &U,LatticePropagator &source)
+{
+  Z2WallSource(RNG,tslice,source);
+  auto tmp = source;
+  GaussianSmear(U,tmp,source);
+}
+void SequentialSource(int tslice,Coordinate &mom,LatticePropagator &spectator,LatticePropagator &source)
+{
+  assert(mom.size()==Nd);
+  assert(mom[Tdir] == 0);
+
+  GridBase * grid = spectator.Grid();
+
+
+  LatticeInteger ts(grid);
+  LatticeCoordinate(ts,Tdir);
+  source = Zero();
+  source = where(ts==Integer(tslice),spectator,source); // Stick in a slice of the spectator, zero everywhere else
+
+  LatticeComplex phase(grid);
+  MakePhase(mom,phase);
+
+  source = source *phase;
+}
+
+template<class Action>
+void MasslessFreePropagator(Action &D,LatticePropagator &source,LatticePropagator &propagator)
+{			   
+ GridBase *UGrid = source.Grid();
+  GridBase *FGrid = D.FermionGrid();
+  bool fiveD = true; //calculate 4d free propagator                                                                                                                 
+  RealD mass = D.Mass();
+  LatticeFermion src4  (UGrid);
+  LatticeFermion result4  (UGrid);
+  LatticeFermion result5(FGrid);
+  LatticeFermion src5(FGrid);
+  LatticePropagator prop5(FGrid);
+  for(int s=0;s<Nd;s++){
+    for(int c=0;c<Nc;c++){
+ 
+      PropToFerm<Action>(src4,source,s,c);
+
+      D.ImportPhysicalFermionSource(src4,src5);
+      D.FreePropagator(src5,result5,mass,true);
+      std::cout<<GridLogMessage
+               <<"spin "<<s<<" color "<<c
+               <<" norm2(src5d) "   <<norm2(src5)
+               <<" norm2(result5d) "<<norm2(result5)<<std::endl;
+
+      D.ExportPhysicalFermionSolution(result5,result4);
+
+      FermToProp<Action>(prop5,result5,s,c);
+      FermToProp<Action>(propagator,result4,s,c);
+    }
+  }
+
+  LatticePropagator Vector_mu(UGrid);
+  LatticeComplex    VV (UGrid);
+  std::vector<TComplex> sumVV;
+  Gamma::Algebra GammaV[3] = {
+    Gamma::Algebra::GammaX,
+    Gamma::Algebra::GammaY,
+    Gamma::Algebra::GammaZ
+  };
+  for( int mu=0;mu<3;mu++ ) {
+    Gamma gV(GammaV[mu]);
+    D.ContractConservedCurrent(prop5,prop5,Vector_mu,source,Current::Vector,mu);
+    VV       = trace(gV*Vector_mu);     // (local) Vector-Vector conserved current
+    sliceSum(VV,sumVV,Tdir);
+    int Nt = sumVV.size();
+    for(int t=0;t<Nt;t++){
+      RealD Ct = real(TensorRemove(sumVV[t]))*LCscale;
+      std::cout<<GridLogMessage <<"VVc["<<mu<<"]["<<t<<"] "<< Ct
+               << " 2 pi^2 t^3 C(t) "<< 2 * M_PI *M_PI * t*t*t *Ct<<std::endl;
+    }
+  }
+}
+
+template<class Action>
+void Solve(Action &D,LatticePropagator &source,LatticePropagator &propagator)
+{
+  GridBase *UGrid = D.GaugeGrid();
+  GridBase *FGrid = D.FermionGrid();
+
+  LatticeFermion src4  (UGrid); 
+  LatticeFermion src5  (FGrid); 
+  LatticeFermion result5(FGrid);
+  LatticeFermion result4(UGrid);
+  LatticePropagator prop5(FGrid);
+  
+  ConjugateGradient<LatticeFermion> CG(1.0e-6,100000);
+  SchurRedBlackDiagMooeeSolve<LatticeFermion> schur(CG);
+  ZeroGuesser<LatticeFermion> ZG; // Could be a DeflatedGuesser if have eigenvectors
+   for(int s=0;s<Nd;s++){
+    for(int c=0;c<Nc;c++){
+      PropToFerm<Action>(src4,source,s,c);
+
+      D.ImportPhysicalFermionSource(src4,src5);
+
+      result5=Zero();
+      schur(D,src5,result5,ZG);
+      std::cout<<GridLogMessage
+	       <<"spin "<<s<<" color "<<c
+	       <<" norm2(src5d) "   <<norm2(src5)
+               <<" norm2(result5d) "<<norm2(result5)<<std::endl;
+
+      D.ExportPhysicalFermionSolution(result5,result4);
+
+      FermToProp<Action>(prop5,result5,s,c);
+      FermToProp<Action>(propagator,result4,s,c);
+    }
+  }
+  LatticePropagator Axial_mu(UGrid); 
+  LatticePropagator Vector_mu(UGrid); 
+
+  LatticeComplex    PA (UGrid); 
+  LatticeComplex    VV (UGrid); 
+  LatticeComplex    PJ5q(UGrid);
+  LatticeComplex    PP (UGrid);
+
+  std::vector<TComplex> sumPA;
+  std::vector<TComplex> sumVV;
+  std::vector<TComplex> sumPP;
+  std::vector<TComplex> sumPJ5q;
+
+  Gamma g5(Gamma::Algebra::Gamma5);
+  D.ContractConservedCurrent(prop5,prop5,Axial_mu,source,Current::Axial,Tdir);
+  PA       = trace(g5*Axial_mu);      // Pseudoscalar-Axial conserved current
+  sliceSum(PA,sumPA,Tdir);
+
+  int Nt{static_cast<int>(sumPA.size())};
+
+  for(int t=0;t<Nt;t++) std::cout<<GridLogMessage <<"PAc["<<t<<"] "<<real(TensorRemove(sumPA[t]))*LCscale<<std::endl;
+
+  PP       = trace(adj(propagator)*propagator); // Pseudoscalar density
+  sliceSum(PP,sumPP,Tdir);
+  for(int t=0;t<Nt;t++) std::cout<<GridLogMessage <<"PP["<<t<<"] "<<real(TensorRemove(sumPP[t]))*LCscale<<std::endl;
+  
+  D.ContractJ5q(prop5,PJ5q);
+  sliceSum(PJ5q,sumPJ5q,Tdir);
+  for(int t=0;t<Nt;t++) std::cout<<GridLogMessage <<"PJ5q["<<t<<"] "<<real(TensorRemove(sumPJ5q[t]))<<std::endl;
+
+  Gamma::Algebra GammaV[3] = {
+    Gamma::Algebra::GammaX,
+    Gamma::Algebra::GammaY,
+    Gamma::Algebra::GammaZ
+  };
+  for( int mu=0;mu<3;mu++ ) {
+    Gamma gV(GammaV[mu]);
+    D.ContractConservedCurrent(prop5,prop5,Vector_mu,source,Current::Vector,mu);
+    //    auto ss=sliceSum(Vector_mu,Tdir);
+    //    for(int t=0;t<Nt;t++) std::cout<<GridLogMessage <<"ss["<<mu<<"]["<<t<<"] "<<ss[t]<<std::endl;
+    VV       = trace(gV*Vector_mu);     // (local) Vector-Vector conserved current
+    sliceSum(VV,sumVV,Tdir);
+    for(int t=0;t<Nt;t++){
+      RealD Ct = real(TensorRemove(sumVV[t]))*LCscale;
+      std::cout<<GridLogMessage <<"VVc["<<mu<<"]["<<t<<"] "<< Ct
+	       << " 2 pi^2 t^3 C(t) "<< 2 * M_PI *M_PI * t*t*t *Ct<<std::endl;
+    }
+  }
+
+}
+
+class MesonFile: Serializable {
+public:
+  GRID_SERIALIZABLE_CLASS_MEMBERS(MesonFile, std::vector<std::vector<Complex> >, data);
+};
+
+void MesonTrace(std::string file,LatticePropagator &q1,LatticePropagator &q2,LatticeComplex &phase)
+{
+  const int nchannel=3;
+  Gamma::Algebra Gammas[nchannel][2] = {
+    {Gamma::Algebra::GammaX,Gamma::Algebra::GammaX},
+    {Gamma::Algebra::GammaY,Gamma::Algebra::GammaY},
+    //    {Gamma::Algebra::GammaZ,Gamma::Algebra::GammaZ}
+    {Gamma::Algebra::Gamma5,Gamma::Algebra::Gamma5}
+  };
+
+  Gamma G5(Gamma::Algebra::Gamma5);
+
+  LatticeComplex meson_CF(q1.Grid());
+  MesonFile MF;
+
+  for(int ch=0;ch<nchannel;ch++){
+
+    Gamma Gsrc(Gammas[ch][0]);
+    Gamma Gsnk(Gammas[ch][1]);
+
+    meson_CF = trace(G5*adj(q1)*G5*Gsnk*q2*adj(Gsrc));
+
+    std::vector<TComplex> meson_T;
+    sliceSum(meson_CF,meson_T, Tdir);
+
+    int nt=meson_T.size();
+
+    std::vector<Complex> corr(nt);
+    for(int t=0;t<nt;t++){
+      corr[t] = TensorRemove(meson_T[t])*LLscale; // Yes this is ugly, not figured a work around
+      std::cout << " channel "<<ch<<" t "<<t<<" " <<real(corr[t])<< " 2 pi^2 t^3 C(t) "<< 2 * M_PI *M_PI * t*t*t *real(corr[t])<<std::endl;
+    }
+    MF.data.push_back(corr);
+  }
+
+  {
+    XmlWriter WR(file);
+    write(WR,"MesonFile",MF);
+  }
+}
+
+int main (int argc, char ** argv)
+{
+  const int Ls=8;
+
+  Grid_init(&argc,&argv);
+
+  // Double precision grids
+  GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), 
+								   GridDefaultSimd(Nd,vComplex::Nsimd()),
+								   GridDefaultMpi());
+  GridRedBlackCartesian * UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
+  GridCartesian         * FGrid   = SpaceTimeGrid::makeFiveDimGrid(Ls,UGrid);
+  GridRedBlackCartesian * FrbGrid = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGrid);
+
+  //////////////////////////////////////////////////////////////////////
+  // You can manage seeds however you like.
+  // Recommend SeedUniqueString.
+  //////////////////////////////////////////////////////////////////////
+  //  std::vector<int> seeds4({1,2,3,4}); 
+  //  GridParallelRNG          RNG4(UGrid);  RNG4.SeedFixedIntegers(seeds4);
+
+  LatticeGaugeField Umu(UGrid);
+  std::string config;
+  RealD M5=atof(getenv("M5"));
+  RealD mq = atof(getenv("mass"));
+  std::vector<RealD> masses({ mq} ); // u/d, s, c ??
+  if( argc > 1 && argv[1][0] != '-' )
+  {
+    std::cout<<GridLogMessage <<"Loading configuration from "<<argv[1]<<std::endl;
+    FieldMetaData header;
+    NerscIO::readConfiguration(Umu, header, argv[1]);
+    config=argv[1];
+    LLscale = 1.0;
+    LCscale = 1.0;
+  }
+  else
+  {
+    SU<Nc>::ColdConfiguration(Umu);
+    config="ColdConfig";
+    //    RealD P=1.0; // Don't scale
+    //    RealD P=0.6153342; // 64I
+    //    RealD P=0.6388238 // 32Ifine
+    //    RealD P=0.5871119; // 48I
+    //    RealD u0 = sqrt(sqrt(P));
+    //    Umu = Umu * u0;
+    RealD w0 = 1 - M5;
+    LLscale = 1.0/(1-w0*w0)/(1-w0*w0);
+    LCscale = 1.0/(1-w0*w0)/(1-w0*w0);
+    std::cout<<GridLogMessage <<"Gauge links are u=1 "<<std::endl;
+    std::cout<<GridLogMessage <<"M5 =  "<<M5<<std::endl;
+    std::cout<<GridLogMessage <<"mq =  "<<mq<<std::endl;
+    std::cout<<GridLogMessage <<"LLscale =  "<<LLscale<<std::endl;
+    std::cout<<GridLogMessage <<"LCscale =  "<<LCscale<<std::endl;
+  }
+
+  int nmass = masses.size();
+
+  std::vector<DomainWallFermionD *> FermActs;
+  
+  std::cout<<GridLogMessage <<"======================"<<std::endl;
+  std::cout<<GridLogMessage <<"DomainWallFermion action"<<std::endl;
+  std::cout<<GridLogMessage <<"======================"<<std::endl;
+
+  for(auto mass: masses) {
+
+    std::cout<<GridLogMessage <<"Making DomainWallFermion action"<<std::endl;
+    FermActs.push_back(new DomainWallFermionD(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5));
+    std::cout<<GridLogMessage <<"Made DomainWallFermion action"<<std::endl;
+   
+  }
+
+  LatticePropagator point_source(UGrid);
+
+  Coordinate Origin({0,0,0,0});
+  PointSource   (Origin,point_source);
+  
+  //  std::vector<LatticePropagator> PointProps(nmass,UGrid);
+  std::vector<LatticePropagator> FreeProps(nmass,UGrid);
+  LatticePropagator delta(UGrid);
+
+  for(int m=0;m<nmass;m++) {
+    //    Solve(*FermActs[m],point_source   ,PointProps[m]);
+    MasslessFreePropagator(*FermActs[m],point_source   ,FreeProps[m]);
+
+    //    delta = PointProps[m] - FreeProps[m];
+    //    std::cout << " delta "<<norm2(delta) << " FFT "<<norm2(FreeProps[m])<< " CG " <<norm2(PointProps[m])<<std::endl;
+  }
+
+  LatticeComplex phase(UGrid);
+  Coordinate mom({0,0,0,0});
+  MakePhase(mom,phase);
+  
+  for(int m1=0 ;m1<nmass;m1++) {
+  for(int m2=m1;m2<nmass;m2++) {
+    std::stringstream ssp,ssg,ssz;
+
+    ssp<<config<< "_m" << m1 << "_m"<< m2 << "_point_meson.xml";
+    ssz<<config<< "_m" << m1 << "_m"<< m2 << "_free_meson.xml";
+
+    //    std::cout << "CG determined VV correlation function"<<std::endl;
+    //    MesonTrace(ssp.str(),PointProps[m1],PointProps[m2],phase);
+    
+    std::cout << "FFT derived VV correlation function"<<std::endl;
+    MesonTrace(ssz.str(),FreeProps[m1],FreeProps[m2],phase);
+  }}
+
+  Grid_finalize();
+}
+
+
+

From 653039695bcb1f4f7ca1e96af659131563e55de3 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 15 Nov 2022 16:37:15 -0500
Subject: [PATCH 079/146] Partial dirichlet changes

---
 benchmarks/Benchmark_comms.cc | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/benchmarks/Benchmark_comms.cc b/benchmarks/Benchmark_comms.cc
index 21e048f4..5ffca2b6 100644
--- a/benchmarks/Benchmark_comms.cc
+++ b/benchmarks/Benchmark_comms.cc
@@ -220,7 +220,7 @@ int main (int argc, char ** argv)
 					      xmit_to_rank,1,
 					      (void *)&rbuf[mu][0],
 					      recv_from_rank,1,
-					      bytes,mu);
+					      bytes,bytes,mu);
 	
 	    comm_proc = mpi_layout[mu]-1;
 	  
@@ -231,7 +231,7 @@ int main (int argc, char ** argv)
 					      xmit_to_rank,1,
 					      (void *)&rbuf[mu+4][0],
 					      recv_from_rank,1,
-					      bytes,mu+4);
+					      bytes,bytes,mu+4);
 	  
 	  }
 	}
@@ -312,7 +312,7 @@ int main (int argc, char ** argv)
 					      xmit_to_rank,1,
 					      (void *)&rbuf[mu][0],
 					      recv_from_rank,1,
-					      bytes,mu);
+					      bytes,bytes,mu);
 	    Grid.StencilSendToRecvFromComplete(requests,mu);
 	    requests.resize(0);
 
@@ -325,7 +325,7 @@ int main (int argc, char ** argv)
 					      xmit_to_rank,1,
 					      (void *)&rbuf[mu+4][0],
 					      recv_from_rank,1,
-					      bytes,mu+4);
+					      bytes,bytes,mu+4);
 	    Grid.StencilSendToRecvFromComplete(requests,mu+4);
 	    requests.resize(0);
 	  
@@ -412,7 +412,7 @@ int main (int argc, char ** argv)
 	    }
             int tid = omp_get_thread_num();
 	    tbytes= Grid.StencilSendToRecvFrom((void *)&xbuf[dir][0], xmit_to_rank,1,
-					       (void *)&rbuf[dir][0], recv_from_rank,1, bytes,tid);
+					       (void *)&rbuf[dir][0], recv_from_rank,1, bytes,bytes,tid);
 
 	    thread_critical { dbytes+=tbytes; }
 	  }

From 1af7572c61754b70cf0e8cfee3ee6cfd64cffa04 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 15 Nov 2022 16:38:51 -0500
Subject: [PATCH 080/146] Some test HMCs for DDHMC

---
 HMC/Mobius2p1f_DD_EOFA_96I_mixed.cc  | 520 ++++++++++++++++++++++++
 HMC/Mobius2p1f_DD_EOFA_96I_mshift.cc | 567 +++++++++++++++++++++++++++
 2 files changed, 1087 insertions(+)
 create mode 100644 HMC/Mobius2p1f_DD_EOFA_96I_mixed.cc
 create mode 100644 HMC/Mobius2p1f_DD_EOFA_96I_mshift.cc

diff --git a/HMC/Mobius2p1f_DD_EOFA_96I_mixed.cc b/HMC/Mobius2p1f_DD_EOFA_96I_mixed.cc
new file mode 100644
index 00000000..24d8951e
--- /dev/null
+++ b/HMC/Mobius2p1f_DD_EOFA_96I_mixed.cc
@@ -0,0 +1,520 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./tests/Test_hmc_EODWFRatio.cc
+
+Copyright (C) 2015-2016
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Guido Cossu <guido.cossu@ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/Grid.h>
+
+NAMESPACE_BEGIN(Grid);
+
+template<class FermionOperatorD, class FermionOperatorF, class SchurOperatorD, class  SchurOperatorF> 
+  class MixedPrecisionConjugateGradientOperatorFunction : public OperatorFunction<typename FermionOperatorD::FermionField> {
+  public:
+    typedef typename FermionOperatorD::FermionField FieldD;
+    typedef typename FermionOperatorF::FermionField FieldF;
+
+    using OperatorFunction<FieldD>::operator();
+
+    RealD   Tolerance;
+    RealD   InnerTolerance; //Initial tolerance for inner CG. Defaults to Tolerance but can be changed
+    Integer MaxInnerIterations;
+    Integer MaxOuterIterations;
+    GridBase* SinglePrecGrid4; //Grid for single-precision fields
+    GridBase* SinglePrecGrid5; //Grid for single-precision fields
+    RealD OuterLoopNormMult; //Stop the outer loop and move to a final double prec solve when the residual is OuterLoopNormMult * Tolerance
+
+    FermionOperatorF &FermOpF;
+    FermionOperatorD &FermOpD;;
+    SchurOperatorF &LinOpF;
+    SchurOperatorD &LinOpD;
+
+    Integer TotalInnerIterations; //Number of inner CG iterations
+    Integer TotalOuterIterations; //Number of restarts
+    Integer TotalFinalStepIterations; //Number of CG iterations in final patch-up step
+
+    MixedPrecisionConjugateGradientOperatorFunction(RealD tol, 
+						    Integer maxinnerit, 
+						    Integer maxouterit, 
+						    GridBase* _sp_grid4, 
+						    GridBase* _sp_grid5, 
+						    FermionOperatorF &_FermOpF,
+						    FermionOperatorD &_FermOpD,
+						    SchurOperatorF   &_LinOpF,
+						    SchurOperatorD   &_LinOpD): 
+      LinOpF(_LinOpF),
+      LinOpD(_LinOpD),
+      FermOpF(_FermOpF),
+      FermOpD(_FermOpD),
+      Tolerance(tol), 
+      InnerTolerance(tol), 
+      MaxInnerIterations(maxinnerit), 
+      MaxOuterIterations(maxouterit), 
+      SinglePrecGrid4(_sp_grid4),
+      SinglePrecGrid5(_sp_grid5),
+      OuterLoopNormMult(100.) 
+    { 
+      /* Debugging instances of objects; references are stored
+      std::cout << GridLogMessage << " Mixed precision CG wrapper LinOpF " <<std::hex<< &LinOpF<<std::dec <<std::endl;
+      std::cout << GridLogMessage << " Mixed precision CG wrapper LinOpD " <<std::hex<< &LinOpD<<std::dec <<std::endl;
+      std::cout << GridLogMessage << " Mixed precision CG wrapper FermOpF " <<std::hex<< &FermOpF<<std::dec <<std::endl;
+      std::cout << GridLogMessage << " Mixed precision CG wrapper FermOpD " <<std::hex<< &FermOpD<<std::dec <<std::endl;
+      */
+    };
+
+    void operator()(LinearOperatorBase<FieldD> &LinOpU, const FieldD &src, FieldD &psi) {
+
+      std::cout << GridLogMessage << " Mixed precision CG wrapper operator() "<<std::endl;
+
+      SchurOperatorD * SchurOpU = static_cast<SchurOperatorD *>(&LinOpU);
+      
+      //      std::cout << GridLogMessage << " Mixed precision CG wrapper operator() FermOpU " <<std::hex<< &(SchurOpU->_Mat)<<std::dec <<std::endl;
+      //      std::cout << GridLogMessage << " Mixed precision CG wrapper operator() FermOpD " <<std::hex<< &(LinOpD._Mat) <<std::dec <<std::endl;
+      // Assumption made in code to extract gauge field
+      // We could avoid storing LinopD reference alltogether ?
+      assert(&(SchurOpU->_Mat)==&(LinOpD._Mat));
+
+      ////////////////////////////////////////////////////////////////////////////////////
+      // Must snarf a single precision copy of the gauge field in Linop_d argument
+      ////////////////////////////////////////////////////////////////////////////////////
+      typedef typename FermionOperatorF::GaugeField GaugeFieldF;
+      typedef typename FermionOperatorF::GaugeLinkField GaugeLinkFieldF;
+      typedef typename FermionOperatorD::GaugeField GaugeFieldD;
+      typedef typename FermionOperatorD::GaugeLinkField GaugeLinkFieldD;
+
+      GridBase * GridPtrF = SinglePrecGrid4;
+      GridBase * GridPtrD = FermOpD.Umu.Grid();
+      GaugeFieldF     U_f  (GridPtrF);
+      GaugeLinkFieldF Umu_f(GridPtrF);
+      //      std::cout << " Dim gauge field "<<GridPtrF->Nd()<<std::endl; // 4d
+      //      std::cout << " Dim gauge field "<<GridPtrD->Nd()<<std::endl; // 4d
+
+      ////////////////////////////////////////////////////////////////////////////////////
+      // Moving this to a Clone method of fermion operator would allow to duplicate the 
+      // physics parameters and decrease gauge field copies
+      ////////////////////////////////////////////////////////////////////////////////////
+      GaugeLinkFieldD Umu_d(GridPtrD);
+      for(int mu=0;mu<Nd*2;mu++){ 
+	Umu_d = PeekIndex<LorentzIndex>(FermOpD.Umu, mu);
+	precisionChange(Umu_f,Umu_d);
+	PokeIndex<LorentzIndex>(FermOpF.Umu, Umu_f, mu);
+      }
+      pickCheckerboard(Even,FermOpF.UmuEven,FermOpF.Umu);
+      pickCheckerboard(Odd ,FermOpF.UmuOdd ,FermOpF.Umu);
+
+      ////////////////////////////////////////////////////////////////////////////////////
+      // Make a mixed precision conjugate gradient
+      ////////////////////////////////////////////////////////////////////////////////////
+#if 1
+      RealD delta=1.e-4;
+      std::cout << GridLogMessage << "Calling reliable update Conjugate Gradient" <<std::endl;
+      ConjugateGradientReliableUpdate<FieldD,FieldF> MPCG(Tolerance,MaxInnerIterations*MaxOuterIterations,delta,SinglePrecGrid5,LinOpF,LinOpD);
+#else      
+      std::cout << GridLogMessage << "Calling mixed precision Conjugate Gradient" <<std::endl;
+      MixedPrecisionConjugateGradient<FieldD,FieldF> MPCG(Tolerance,MaxInnerIterations,MaxOuterIterations,SinglePrecGrid5,LinOpF,LinOpD);
+#endif
+      MPCG(src,psi);
+    }
+  };
+
+NAMESPACE_END(Grid);
+
+
+int main(int argc, char **argv) {
+  using namespace Grid;
+
+  Grid_init(&argc, &argv);
+
+  CartesianCommunicator::BarrierWorld();
+  std::cout << GridLogMessage << " Clock skew check" <<std::endl;
+  
+  int threads = GridThread::GetThreads();
+
+   // Typedefs to simplify notation
+  typedef WilsonImplR FermionImplPolicy;
+  typedef WilsonImplF FermionImplPolicyF;
+  typedef MobiusFermionD FermionAction;
+  typedef MobiusFermionF FermionActionF;
+  typedef MobiusEOFAFermionD FermionEOFAAction;
+  typedef MobiusEOFAFermionF FermionEOFAActionF;
+  typedef typename FermionAction::FermionField FermionField;
+  typedef typename FermionActionF::FermionField FermionFieldF;
+
+  typedef WilsonImplD2 FermionImplPolicyD2;
+  typedef MobiusFermionD2 FermionActionD2;
+  typedef MobiusEOFAFermionD2 FermionEOFAActionD2;
+  typedef typename FermionActionD2::FermionField FermionFieldD2;
+  
+  typedef Grid::XmlReader       Serialiser;
+
+  //::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
+  IntegratorParameters MD;
+  //  typedef GenericHMCRunner<LeapFrog> HMCWrapper;
+  //  MD.name    = std::string("Leap Frog");
+  typedef GenericHMCRunner<ForceGradient> HMCWrapper;
+  MD.name    = std::string("Force Gradient");
+  //typedef GenericHMCRunner<MinimumNorm2> HMCWrapper;
+  // MD.name    = std::string("MinimumNorm2");
+  //  MD.MDsteps =  4;
+  MD.MDsteps =  4;
+  MD.trajL   = 0.5;
+
+  HMCparameters HMCparams;
+  HMCparams.StartTrajectory  = 1077;
+  HMCparams.Trajectories     = 1;
+  HMCparams.NoMetropolisUntil=  0;
+  // "[HotStart, ColdStart, TepidStart, CheckpointStart]\n";
+  //  HMCparams.StartingType     =std::string("ColdStart");
+  HMCparams.StartingType     =std::string("CheckpointStart");
+  HMCparams.MD = MD;
+  HMCWrapper TheHMC(HMCparams);
+
+  // Grid from the command line arguments --grid and --mpi
+  TheHMC.Resources.AddFourDimGrid("gauge"); // use default simd lanes decomposition
+
+  CheckpointerParameters CPparams;
+  CPparams.config_prefix = "ckpoint_DDHMC_lat";
+  CPparams.rng_prefix    = "ckpoint_DDHMC_rng";
+  CPparams.saveInterval  = 1;
+  CPparams.format        = "IEEE64BIG";
+  TheHMC.Resources.LoadNerscCheckpointer(CPparams);
+  std::cout << "loaded NERSC checpointer"<<std::endl;
+  RNGModuleParameters RNGpar;
+  RNGpar.serial_seeds = "1 2 3 4 5";
+  RNGpar.parallel_seeds = "6 7 8 9 10";
+  TheHMC.Resources.SetRNGSeeds(RNGpar);
+
+  // Construct observables
+  // here there is too much indirection
+  typedef PlaquetteMod<HMCWrapper::ImplPolicy> PlaqObs;
+  TheHMC.Resources.AddObservable<PlaqObs>();
+  //////////////////////////////////////////////
+
+  const int Ls      = 12;
+  RealD M5  = 1.8;
+  RealD b   = 1.5;
+  RealD c   = 0.5;
+  Real beta         = 2.13;
+  //  Real light_mass   = 5.4e-4;
+  Real light_mass   = 7.8e-4;
+  Real strange_mass = 0.0362;
+  Real pv_mass      = 1.0;
+  //  std::vector<Real> hasenbusch({ light_mass, 3.8e-3, 0.0145, 0.045, 0.108, 0.25, 0.51 , pv_mass });
+  std::vector<Real> hasenbusch({ light_mass, 5e-3, 0.0145, 0.045, 0.108, 0.25, 0.51 , 0.75 , pv_mass });
+
+
+  OneFlavourRationalParams OFRp; // Up/down
+  OFRp.lo       = 2.0e-5;
+  OFRp.hi       = 90.0;
+  OFRp.MaxIter  = 60000;
+  OFRp.tolerance= 1.0e-5;
+  OFRp.mdtolerance= 1.0e-3;
+  //  OFRp.degree   = 20; converges
+  //  OFRp.degree   = 16;
+  OFRp.degree   = 12;
+  OFRp.precision= 80;
+  OFRp.BoundsCheckFreq=0;
+  std::vector<RealD> ActionTolByPole({
+      1.0e-8,1.0e-8,1.0e-8,1.0e-8,
+      1.0e-10,1.0e-10,1.0e-10,1.0e-10,
+      1.0e-10,1.0e-10,1.0e-10,1.0e-10
+    });
+  std::vector<RealD> MDTolByPole({
+      3.0e-5,1.0e-5,3.0e-6,1.0e-6,
+      1.0e-8,1.0e-8,1.0e-8,1.0e-8,
+      1.0e-10,1.0e-10,1.0e-10,1.0e-10
+    });
+
+  auto GridPtr   = TheHMC.Resources.GetCartesian();
+  auto GridRBPtr = TheHMC.Resources.GetRBCartesian();
+
+  typedef SchurDiagMooeeOperator<FermionActionF,FermionFieldF> LinearOperatorF;
+  typedef SchurDiagMooeeOperator<FermionAction ,FermionField > LinearOperatorD;
+  typedef SchurDiagMooeeOperator<FermionEOFAActionF,FermionFieldF> LinearOperatorEOFAF;
+  typedef SchurDiagMooeeOperator<FermionEOFAAction ,FermionField > LinearOperatorEOFAD;
+  typedef MixedPrecisionConjugateGradientOperatorFunction<MobiusFermionD,MobiusFermionF,LinearOperatorD,LinearOperatorF> MxPCG;
+  typedef MixedPrecisionConjugateGradientOperatorFunction<MobiusEOFAFermionD,MobiusEOFAFermionF,LinearOperatorEOFAD,LinearOperatorEOFAF> MxPCG_EOFA;
+
+  ////////////////////////////////////////////////////////////////
+  // Domain decomposed
+  ////////////////////////////////////////////////////////////////
+  Coordinate latt4  = GridPtr->GlobalDimensions();
+  Coordinate mpi    = GridPtr->ProcessorGrid();
+  Coordinate shm;
+
+  GlobalSharedMemory::GetShmDims(mpi,shm);
+  
+  Coordinate CommDim(Nd);
+  for(int d=0;d<Nd;d++) CommDim[d]= (mpi[d]/shm[d])>1 ? 1 : 0;
+
+  Coordinate NonDirichlet(Nd+1,0);
+  Coordinate Dirichlet(Nd+1,0);
+  Dirichlet[1] = CommDim[0]*latt4[0]/mpi[0] * shm[0];
+  Dirichlet[2] = CommDim[1]*latt4[1]/mpi[1] * shm[1];
+  Dirichlet[3] = CommDim[2]*latt4[2]/mpi[2] * shm[2];
+  Dirichlet[4] = CommDim[3]*latt4[3]/mpi[3] * shm[3];
+
+  Coordinate Block4(Nd);
+  Block4[0] = Dirichlet[1];
+  Block4[1] = Dirichlet[2];
+  Block4[2] = Dirichlet[3];
+  Block4[3] = Dirichlet[4];
+
+  int Width=3;
+  TheHMC.Resources.SetMomentumFilter(new DDHMCFilter<WilsonImplR::Field>(Block4,Width));
+
+  //////////////////////////
+  // Fermion Grids
+  //////////////////////////
+  auto FGrid     = SpaceTimeGrid::makeFiveDimGrid(Ls,GridPtr);
+  auto FrbGrid   = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,GridPtr);
+
+  Coordinate simdF = GridDefaultSimd(Nd,vComplexF::Nsimd());
+  auto GridPtrF   = SpaceTimeGrid::makeFourDimGrid(latt4,simdF,mpi);
+  auto GridRBPtrF = SpaceTimeGrid::makeFourDimRedBlackGrid(GridPtrF);
+  auto FGridF     = SpaceTimeGrid::makeFiveDimGrid(Ls,GridPtrF);
+  auto FrbGridF   = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,GridPtrF);
+
+  IwasakiGaugeActionR GaugeAction(beta);
+
+  // temporarily need a gauge field
+  LatticeGaugeField U(GridPtr);
+  LatticeGaugeFieldF UF(GridPtrF);
+
+  std::cout << GridLogMessage << " Running the HMC "<< std::endl;
+  TheHMC.ReadCommandLine(argc,argv);  // params on CML or from param file
+  TheHMC.initializeGaugeFieldAndRNGs(U);
+  std::cout << "loaded NERSC gauge field"<<std::endl;
+
+
+  // These lines are unecessary if BC are all periodic
+  std::vector<Complex> boundary = {1,1,1,-1};
+  FermionAction::ImplParams Params(boundary);
+  Params.dirichlet=NonDirichlet;
+  FermionAction::ImplParams ParamsDir(boundary);
+  ParamsDir.dirichlet=Dirichlet;
+
+  //  double StoppingCondition = 1e-14;
+  //  double MDStoppingCondition = 1e-9;
+  double StoppingCondition = 1e-8;
+  double MDStoppingCondition = 1e-7;
+  double MDStoppingConditionLoose = 1e-7;
+  double MDStoppingConditionStrange = 1e-7;
+  double MaxCGIterations = 300000;
+  ConjugateGradient<FermionField>  CG(StoppingCondition,MaxCGIterations);
+  ConjugateGradient<FermionField>  MDCG(MDStoppingCondition,MaxCGIterations);
+
+  ////////////////////////////////////
+  // Collect actions
+  ////////////////////////////////////
+  ActionLevel<HMCWrapper::Field> Level1(1);
+  ActionLevel<HMCWrapper::Field> Level2(2);
+  ActionLevel<HMCWrapper::Field> Level3(30);
+
+  ////////////////////////////////////
+  // Strange action
+  ////////////////////////////////////
+  FermionAction StrangeOp (U,*FGrid,*FrbGrid,*GridPtr,*GridRBPtr,strange_mass,M5,b,c, Params);
+  FermionAction StrangePauliVillarsOp(U,*FGrid,*FrbGrid,*GridPtr,*GridRBPtr,pv_mass,  M5,b,c, Params);
+
+  // Probably dominates the force - back to EOFA.
+  OneFlavourRationalParams SFRp;
+  SFRp.lo       = 0.1;
+  SFRp.hi       = 25.0;
+  SFRp.MaxIter  = 10000;
+  SFRp.tolerance= 1.0e-5;
+  SFRp.mdtolerance= 2.0e-4;
+  SFRp.degree   = 14;
+  SFRp.precision= 50;
+  
+  MobiusEOFAFermionD Strange_Op_L (U , *FGrid , *FrbGrid , *GridPtr , *GridRBPtr , strange_mass, strange_mass, pv_mass, 0.0, -1, M5, b, c);
+  MobiusEOFAFermionF Strange_Op_LF(UF, *FGridF, *FrbGridF, *GridPtrF, *GridRBPtrF, strange_mass, strange_mass, pv_mass, 0.0, -1, M5, b, c);
+  MobiusEOFAFermionD Strange_Op_R (U , *FGrid , *FrbGrid , *GridPtr , *GridRBPtr , pv_mass, strange_mass,      pv_mass, -1.0, 1, M5, b, c);
+  MobiusEOFAFermionF Strange_Op_RF(UF, *FGridF, *FrbGridF, *GridPtrF, *GridRBPtrF, pv_mass, strange_mass,      pv_mass, -1.0, 1, M5, b, c);
+  ConjugateGradient<FermionField>      ActionCG(StoppingCondition,MaxCGIterations);
+  ConjugateGradient<FermionField>  DerivativeCG(MDStoppingCondition,MaxCGIterations);
+  LinearOperatorEOFAD Strange_LinOp_L (Strange_Op_L);
+  LinearOperatorEOFAD Strange_LinOp_R (Strange_Op_R);
+  LinearOperatorEOFAF Strange_LinOp_LF(Strange_Op_LF);
+  LinearOperatorEOFAF Strange_LinOp_RF(Strange_Op_RF);
+
+  const int MX_inner = 1000;
+  MxPCG_EOFA ActionCGL(StoppingCondition,
+		       MX_inner,
+		       MaxCGIterations,
+		       GridPtrF,
+		       FrbGridF,
+		       Strange_Op_LF,Strange_Op_L,
+		       Strange_LinOp_LF,Strange_LinOp_L);
+
+  MxPCG_EOFA DerivativeCGL(MDStoppingConditionStrange,
+			   MX_inner,
+			   MaxCGIterations,
+			   GridPtrF,
+			   FrbGridF,
+			   Strange_Op_LF,Strange_Op_L,
+			   Strange_LinOp_LF,Strange_LinOp_L);
+  
+  MxPCG_EOFA ActionCGR(StoppingCondition,
+		       MX_inner,
+		       MaxCGIterations,
+		       GridPtrF,
+		       FrbGridF,
+		       Strange_Op_RF,Strange_Op_R,
+		       Strange_LinOp_RF,Strange_LinOp_R);
+  
+  MxPCG_EOFA DerivativeCGR(MDStoppingConditionStrange,
+			   MX_inner,
+			   MaxCGIterations,
+			   GridPtrF,
+			   FrbGridF,
+			   Strange_Op_RF,Strange_Op_R,
+			   Strange_LinOp_RF,Strange_LinOp_R);
+
+  ExactOneFlavourRatioPseudoFermionAction<FermionImplPolicy> 
+    EOFA(Strange_Op_L, Strange_Op_R, 
+	 ActionCG, 
+	 ActionCGL, ActionCGR,
+	 DerivativeCGL, DerivativeCGR,
+	 SFRp, true);
+  //  Level2.push_back(&EOFA);
+
+  ////////////////////////////////////
+  // up down action
+  ////////////////////////////////////
+  std::vector<Real> light_den;
+  std::vector<Real> light_num;
+  std::vector<int> dirichlet_den;
+  std::vector<int> dirichlet_num;
+
+  int n_hasenbusch = hasenbusch.size();
+  light_den.push_back(light_mass);  dirichlet_den.push_back(0);
+  for(int h=0;h<n_hasenbusch;h++){
+    light_den.push_back(hasenbusch[h]); dirichlet_den.push_back(1);
+  }
+
+  for(int h=0;h<n_hasenbusch;h++){
+    light_num.push_back(hasenbusch[h]); dirichlet_num.push_back(1);
+  }
+  light_num.push_back(pv_mass);  dirichlet_num.push_back(0);
+
+  std::vector<FermionAction *> Numerators;
+  std::vector<FermionAction *> Denominators;
+  std::vector<FermionActionF *> DenominatorsF;
+  std::vector<TwoFlavourEvenOddRatioPseudoFermionAction<FermionImplPolicy> *> Quotients;
+  std::vector<OneFlavourEvenOddRatioRationalPseudoFermionAction<FermionImplPolicy> *> Bdys;
+  std::vector<MxPCG *> ActionMPCG;
+  std::vector<MxPCG *> MPCG;
+
+  typedef SchurDiagMooeeOperator<FermionActionF,FermionFieldF> LinearOperatorF;
+  typedef SchurDiagMooeeOperator<FermionAction ,FermionField > LinearOperatorD;
+  std::vector<LinearOperatorD *> LinOpD;
+  std::vector<LinearOperatorF *> LinOpF; 
+  
+  for(int h=0;h<n_hasenbusch+1;h++){
+    std::cout << GridLogMessage
+	      << " 2f quotient Action ";
+    std::cout << "det D("<<light_den[h]<<")";
+    if ( dirichlet_den[h] ) std::cout << "^dirichlet    ";
+    std::cout << "/ det D("<<light_num[h]<<")";
+    if ( dirichlet_num[h] ) std::cout << "^dirichlet    ";
+    std::cout << std::endl;
+
+    FermionAction::ImplParams ParamsNum(boundary);
+    FermionAction::ImplParams ParamsDen(boundary);
+    FermionActionF::ImplParams ParamsDenF(boundary);
+    
+    if ( dirichlet_num[h]==1) ParamsNum.dirichlet = Dirichlet;
+    else                      ParamsNum.dirichlet = NonDirichlet;
+    Numerators.push_back  (new FermionAction(U,*FGrid,*FrbGrid,*GridPtr,*GridRBPtr,light_num[h],M5,b,c, ParamsNum));
+
+    if ( dirichlet_den[h]==1) ParamsDen.dirichlet = Dirichlet;
+    else                      ParamsDen.dirichlet = NonDirichlet;
+
+    Denominators.push_back(new FermionAction(U,*FGrid,*FrbGrid,*GridPtr,*GridRBPtr,light_den[h],M5,b,c, ParamsDen));
+
+    ParamsDenF.dirichlet = ParamsDen.dirichlet;
+    DenominatorsF.push_back(new FermionActionF(UF,*FGridF,*FrbGridF,*GridPtrF,*GridRBPtrF,light_den[h],M5,b,c, ParamsDenF));
+
+    LinOpD.push_back(new LinearOperatorD(*Denominators[h]));
+    LinOpF.push_back(new LinearOperatorF(*DenominatorsF[h]));
+
+    double conv  = MDStoppingCondition;
+    if (h<3) conv= MDStoppingConditionLoose; // Relax on first two hasenbusch factors
+    const int MX_inner = 5000;
+    MPCG.push_back(new MxPCG(conv,
+			     MX_inner,
+			     MaxCGIterations,
+			     GridPtrF,
+			     FrbGridF,
+			     *DenominatorsF[h],*Denominators[h],
+			     *LinOpF[h], *LinOpD[h]) );
+
+    ActionMPCG.push_back(new MxPCG(StoppingCondition,
+				   MX_inner,
+				   MaxCGIterations,
+				   GridPtrF,
+				   FrbGridF,
+				   *DenominatorsF[h],*Denominators[h],
+				   *LinOpF[h], *LinOpD[h]) );
+
+    
+    if(h!=0) {
+      //      Quotients.push_back (new TwoFlavourEvenOddRatioPseudoFermionAction<FermionImplPolicy>(*Numerators[h],*Denominators[h],MDCG,CG));
+      Quotients.push_back (new TwoFlavourEvenOddRatioPseudoFermionAction<FermionImplPolicy>(*Numerators[h],*Denominators[h],*MPCG[h],*ActionMPCG[h],CG));
+    } else {
+      Bdys.push_back( new OneFlavourEvenOddRatioRationalPseudoFermionAction<FermionImplPolicy>(*Numerators[h],*Denominators[h],OFRp));
+      Bdys.push_back( new OneFlavourEvenOddRatioRationalPseudoFermionAction<FermionImplPolicy>(*Numerators[h],*Denominators[h],OFRp));
+    }
+  }
+  for(int h=0;h<Bdys.size();h++){
+    Bdys[h]->SetTolerances(ActionTolByPole,MDTolByPole);
+  }
+  int nquo=Quotients.size();
+  Level2.push_back(Bdys[0]);
+  Level2.push_back(Bdys[1]);
+  for(int h=0;h<nquo-1;h++){
+    Level2.push_back(Quotients[h]);
+  }
+  Level2.push_back(Quotients[nquo-1]);
+
+  /////////////////////////////////////////////////////////////
+  // Gauge action
+  /////////////////////////////////////////////////////////////
+  Level3.push_back(&GaugeAction);
+  TheHMC.TheAction.push_back(Level1);
+  TheHMC.TheAction.push_back(Level2);
+  TheHMC.TheAction.push_back(Level3);
+  std::cout << GridLogMessage << " Action complete "<< std::endl;
+  /////////////////////////////////////////////////////////////
+
+  TheHMC.Run();  // no smearing
+
+  Grid_finalize();
+} // main
+
+
+
diff --git a/HMC/Mobius2p1f_DD_EOFA_96I_mshift.cc b/HMC/Mobius2p1f_DD_EOFA_96I_mshift.cc
new file mode 100644
index 00000000..f4ca8515
--- /dev/null
+++ b/HMC/Mobius2p1f_DD_EOFA_96I_mshift.cc
@@ -0,0 +1,567 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./tests/Test_hmc_EODWFRatio.cc
+
+Copyright (C) 2015-2016
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Guido Cossu <guido.cossu@ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/Grid.h>
+
+NAMESPACE_BEGIN(Grid);
+
+template<class FermionOperatorD, class FermionOperatorF, class SchurOperatorD, class  SchurOperatorF> 
+  class MixedPrecisionConjugateGradientOperatorFunction : public OperatorFunction<typename FermionOperatorD::FermionField> {
+  public:
+    typedef typename FermionOperatorD::FermionField FieldD;
+    typedef typename FermionOperatorF::FermionField FieldF;
+
+    using OperatorFunction<FieldD>::operator();
+
+    RealD   Tolerance;
+    RealD   InnerTolerance; //Initial tolerance for inner CG. Defaults to Tolerance but can be changed
+    Integer MaxInnerIterations;
+    Integer MaxOuterIterations;
+    GridBase* SinglePrecGrid4; //Grid for single-precision fields
+    GridBase* SinglePrecGrid5; //Grid for single-precision fields
+    RealD OuterLoopNormMult; //Stop the outer loop and move to a final double prec solve when the residual is OuterLoopNormMult * Tolerance
+
+    FermionOperatorF &FermOpF;
+    FermionOperatorD &FermOpD;;
+    SchurOperatorF &LinOpF;
+    SchurOperatorD &LinOpD;
+
+    Integer TotalInnerIterations; //Number of inner CG iterations
+    Integer TotalOuterIterations; //Number of restarts
+    Integer TotalFinalStepIterations; //Number of CG iterations in final patch-up step
+
+    MixedPrecisionConjugateGradientOperatorFunction(RealD tol, 
+						    Integer maxinnerit, 
+						    Integer maxouterit, 
+						    GridBase* _sp_grid4, 
+						    GridBase* _sp_grid5, 
+						    FermionOperatorF &_FermOpF,
+						    FermionOperatorD &_FermOpD,
+						    SchurOperatorF   &_LinOpF,
+						    SchurOperatorD   &_LinOpD): 
+      LinOpF(_LinOpF),
+      LinOpD(_LinOpD),
+      FermOpF(_FermOpF),
+      FermOpD(_FermOpD),
+      Tolerance(tol), 
+      InnerTolerance(tol), 
+      MaxInnerIterations(maxinnerit), 
+      MaxOuterIterations(maxouterit), 
+      SinglePrecGrid4(_sp_grid4),
+      SinglePrecGrid5(_sp_grid5),
+      OuterLoopNormMult(100.) 
+    { 
+      /* Debugging instances of objects; references are stored
+      std::cout << GridLogMessage << " Mixed precision CG wrapper LinOpF " <<std::hex<< &LinOpF<<std::dec <<std::endl;
+      std::cout << GridLogMessage << " Mixed precision CG wrapper LinOpD " <<std::hex<< &LinOpD<<std::dec <<std::endl;
+      std::cout << GridLogMessage << " Mixed precision CG wrapper FermOpF " <<std::hex<< &FermOpF<<std::dec <<std::endl;
+      std::cout << GridLogMessage << " Mixed precision CG wrapper FermOpD " <<std::hex<< &FermOpD<<std::dec <<std::endl;
+      */
+    };
+
+    void operator()(LinearOperatorBase<FieldD> &LinOpU, const FieldD &src, FieldD &psi) {
+
+      std::cout << GridLogMessage << " Mixed precision CG wrapper operator() "<<std::endl;
+
+      SchurOperatorD * SchurOpU = static_cast<SchurOperatorD *>(&LinOpU);
+      
+      //      std::cout << GridLogMessage << " Mixed precision CG wrapper operator() FermOpU " <<std::hex<< &(SchurOpU->_Mat)<<std::dec <<std::endl;
+      //      std::cout << GridLogMessage << " Mixed precision CG wrapper operator() FermOpD " <<std::hex<< &(LinOpD._Mat) <<std::dec <<std::endl;
+      // Assumption made in code to extract gauge field
+      // We could avoid storing LinopD reference alltogether ?
+      assert(&(SchurOpU->_Mat)==&(LinOpD._Mat));
+
+      ////////////////////////////////////////////////////////////////////////////////////
+      // Must snarf a single precision copy of the gauge field in Linop_d argument
+      ////////////////////////////////////////////////////////////////////////////////////
+      typedef typename FermionOperatorF::GaugeField GaugeFieldF;
+      typedef typename FermionOperatorF::GaugeLinkField GaugeLinkFieldF;
+      typedef typename FermionOperatorD::GaugeField GaugeFieldD;
+      typedef typename FermionOperatorD::GaugeLinkField GaugeLinkFieldD;
+
+      GridBase * GridPtrF = SinglePrecGrid4;
+      GridBase * GridPtrD = FermOpD.Umu.Grid();
+      GaugeFieldF     U_f  (GridPtrF);
+      GaugeLinkFieldF Umu_f(GridPtrF);
+      //      std::cout << " Dim gauge field "<<GridPtrF->Nd()<<std::endl; // 4d
+      //      std::cout << " Dim gauge field "<<GridPtrD->Nd()<<std::endl; // 4d
+
+      ////////////////////////////////////////////////////////////////////////////////////
+      // Moving this to a Clone method of fermion operator would allow to duplicate the 
+      // physics parameters and decrease gauge field copies
+      ////////////////////////////////////////////////////////////////////////////////////
+      GaugeLinkFieldD Umu_d(GridPtrD);
+      for(int mu=0;mu<Nd*2;mu++){ 
+	Umu_d = PeekIndex<LorentzIndex>(FermOpD.Umu, mu);
+	precisionChange(Umu_f,Umu_d);
+	PokeIndex<LorentzIndex>(FermOpF.Umu, Umu_f, mu);
+      }
+      pickCheckerboard(Even,FermOpF.UmuEven,FermOpF.Umu);
+      pickCheckerboard(Odd ,FermOpF.UmuOdd ,FermOpF.Umu);
+
+      ////////////////////////////////////////////////////////////////////////////////////
+      // Make a mixed precision conjugate gradient
+      ////////////////////////////////////////////////////////////////////////////////////
+#if 1
+      RealD delta=1.e-4;
+      std::cout << GridLogMessage << "Calling reliable update Conjugate Gradient" <<std::endl;
+      ConjugateGradientReliableUpdate<FieldD,FieldF> MPCG(Tolerance,MaxInnerIterations*MaxOuterIterations,delta,SinglePrecGrid5,LinOpF,LinOpD);
+#else      
+      std::cout << GridLogMessage << "Calling mixed precision Conjugate Gradient" <<std::endl;
+      MixedPrecisionConjugateGradient<FieldD,FieldF> MPCG(Tolerance,MaxInnerIterations,MaxOuterIterations,SinglePrecGrid5,LinOpF,LinOpD);
+#endif
+      MPCG(src,psi);
+    }
+  };
+
+NAMESPACE_END(Grid);
+
+
+int main(int argc, char **argv) {
+  using namespace Grid;
+
+  Grid_init(&argc, &argv);
+
+  CartesianCommunicator::BarrierWorld();
+  std::cout << GridLogMessage << " Clock skew check" <<std::endl;
+  
+  int threads = GridThread::GetThreads();
+
+   // Typedefs to simplify notation
+  typedef WilsonImplD FermionImplPolicy;
+  typedef MobiusFermionD FermionAction;
+  typedef MobiusEOFAFermionD FermionEOFAAction;
+  typedef typename FermionAction::FermionField FermionField;
+
+  typedef WilsonImplF FermionImplPolicyF;
+  typedef MobiusFermionF FermionActionF;
+  typedef MobiusEOFAFermionF FermionEOFAActionF;
+  typedef typename FermionActionF::FermionField FermionFieldF;
+
+  typedef WilsonImplD2 FermionImplPolicyD2;
+  typedef MobiusFermionD2 FermionActionD2;
+  typedef MobiusEOFAFermionD2 FermionEOFAActionD2;
+  typedef typename FermionActionD2::FermionField FermionFieldD2;
+
+  typedef Grid::XmlReader       Serialiser;
+
+  //::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
+  IntegratorParameters MD;
+  //  typedef GenericHMCRunner<LeapFrog> HMCWrapper;
+  //  MD.name    = std::string("Leap Frog");
+  typedef GenericHMCRunner<ForceGradient> HMCWrapper;
+  MD.name    = std::string("Force Gradient");
+  //typedef GenericHMCRunner<MinimumNorm2> HMCWrapper;
+  // MD.name    = std::string("MinimumNorm2");
+  //  MD.MDsteps =  4;
+  MD.MDsteps =  4;
+  MD.trajL   = 0.5;
+
+  HMCparameters HMCparams;
+  HMCparams.StartTrajectory  = 1077;
+  HMCparams.Trajectories     = 1;
+  HMCparams.NoMetropolisUntil=  0;
+  // "[HotStart, ColdStart, TepidStart, CheckpointStart]\n";
+  //  HMCparams.StartingType     =std::string("ColdStart");
+  HMCparams.StartingType     =std::string("CheckpointStart");
+  HMCparams.MD = MD;
+  HMCWrapper TheHMC(HMCparams);
+
+  // Grid from the command line arguments --grid and --mpi
+  TheHMC.Resources.AddFourDimGrid("gauge"); // use default simd lanes decomposition
+
+  CheckpointerParameters CPparams;
+  CPparams.config_prefix = "ckpoint_DDHMC_lat";
+  CPparams.rng_prefix    = "ckpoint_DDHMC_rng";
+  CPparams.saveInterval  = 1;
+  CPparams.format        = "IEEE64BIG";
+  TheHMC.Resources.LoadNerscCheckpointer(CPparams);
+  std::cout << "loaded NERSC checpointer"<<std::endl;
+  RNGModuleParameters RNGpar;
+  RNGpar.serial_seeds = "1 2 3 4 5";
+  RNGpar.parallel_seeds = "6 7 8 9 10";
+  TheHMC.Resources.SetRNGSeeds(RNGpar);
+
+  // Construct observables
+  // here there is too much indirection
+  typedef PlaquetteMod<HMCWrapper::ImplPolicy> PlaqObs;
+  TheHMC.Resources.AddObservable<PlaqObs>();
+  //////////////////////////////////////////////
+
+  const int Ls      = 12;
+  RealD M5  = 1.8;
+  RealD b   = 1.5;
+  RealD c   = 0.5;
+  Real beta         = 2.13;
+  //  Real light_mass   = 5.4e-4;
+  Real light_mass   = 7.8e-4;
+  Real strange_mass = 0.0362;
+  Real pv_mass      = 1.0;
+  std::vector<Real> hasenbusch({ light_mass, 3.8e-3, 0.0145, 0.045, 0.108, 0.25, 0.51 , pv_mass });
+  //  std::vector<Real> hasenbusch({ light_mass, 0.0145, 0.045, 0.108, 0.25, 0.51 , 0.75 , pv_mass });
+
+
+  OneFlavourRationalParams OFRp; // Up/down
+  OFRp.lo       = 4.0e-5;
+  OFRp.hi       = 90.0;
+  OFRp.MaxIter  = 60000;
+  OFRp.tolerance= 1.0e-5;
+  OFRp.mdtolerance= 1.0e-3;
+  //  OFRp.degree   = 20; converges
+  //  OFRp.degree   = 16;
+  OFRp.degree   = 12;
+  OFRp.precision= 80;
+  OFRp.BoundsCheckFreq=0;
+  std::vector<RealD> ActionTolByPole({
+      1.0e-8,1.0e-8,1.0e-8,1.0e-8,
+      1.0e-8,1.0e-8,1.0e-8,1.0e-8,
+      1.0e-8,1.0e-8,1.0e-8,1.0e-8
+    });
+  std::vector<RealD> MDTolByPole({
+      1.0e-6,3.0e-7,1.0e-7,1.0e-7,
+      1.0e-8,1.0e-8,1.0e-8,1.0e-8,
+      1.0e-8,1.0e-8,1.0e-8,1.0e-8
+    });
+
+  auto GridPtr   = TheHMC.Resources.GetCartesian();
+  auto GridRBPtr = TheHMC.Resources.GetRBCartesian();
+
+  typedef SchurDiagMooeeOperator<FermionActionF,FermionFieldF> LinearOperatorF;
+  typedef SchurDiagMooeeOperator<FermionAction ,FermionField > LinearOperatorD;
+  typedef SchurDiagMooeeOperator<FermionActionD2,FermionFieldD2 > LinearOperatorD2;
+  typedef SchurDiagMooeeOperator<FermionEOFAActionF,FermionFieldF> LinearOperatorEOFAF;
+  typedef SchurDiagMooeeOperator<FermionEOFAAction ,FermionField > LinearOperatorEOFAD;
+  typedef SchurDiagMooeeOperator<FermionEOFAActionD2,FermionFieldD2 > LinearOperatorEOFAD2;
+  typedef MixedPrecisionConjugateGradientOperatorFunction<MobiusFermionD,MobiusFermionF,LinearOperatorD,LinearOperatorF> MxPCG;
+  typedef MixedPrecisionConjugateGradientOperatorFunction<MobiusEOFAFermionD,MobiusEOFAFermionF,LinearOperatorEOFAD,LinearOperatorEOFAF> MxPCG_EOFA;
+
+  ////////////////////////////////////////////////////////////////
+  // Domain decomposed
+  ////////////////////////////////////////////////////////////////
+  Coordinate latt4  = GridPtr->GlobalDimensions();
+  Coordinate mpi    = GridPtr->ProcessorGrid();
+  Coordinate shm;
+
+  GlobalSharedMemory::GetShmDims(mpi,shm);
+  
+  Coordinate CommDim(Nd);
+  for(int d=0;d<Nd;d++) CommDim[d]= (mpi[d]/shm[d])>1 ? 1 : 0;
+
+  Coordinate NonDirichlet(Nd+1,0);
+  Coordinate Dirichlet(Nd+1,0);
+  Dirichlet[1] = CommDim[0]*latt4[0]/mpi[0] * shm[0];
+  Dirichlet[2] = CommDim[1]*latt4[1]/mpi[1] * shm[1];
+  Dirichlet[3] = CommDim[2]*latt4[2]/mpi[2] * shm[2];
+  Dirichlet[4] = CommDim[3]*latt4[3]/mpi[3] * shm[3];
+  //Dirichlet[1] = 0;
+  //Dirichlet[2] = 0;
+  //Dirichlet[3] = 0;
+
+  // 
+  Coordinate Block4(Nd);
+  Block4[0] = Dirichlet[1];
+  Block4[1] = Dirichlet[2];
+  Block4[2] = Dirichlet[3];
+  Block4[3] = Dirichlet[4];
+
+  int Width=4;
+  TheHMC.Resources.SetMomentumFilter(new DDHMCFilter<WilsonImplD::Field>(Block4,Width));
+
+  //////////////////////////
+  // Fermion Grids
+  //////////////////////////
+  auto FGrid     = SpaceTimeGrid::makeFiveDimGrid(Ls,GridPtr);
+  auto FrbGrid   = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,GridPtr);
+
+  Coordinate simdF = GridDefaultSimd(Nd,vComplexF::Nsimd());
+  auto GridPtrF   = SpaceTimeGrid::makeFourDimGrid(latt4,simdF,mpi);
+  auto GridRBPtrF = SpaceTimeGrid::makeFourDimRedBlackGrid(GridPtrF);
+  auto FGridF     = SpaceTimeGrid::makeFiveDimGrid(Ls,GridPtrF);
+  auto FrbGridF   = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,GridPtrF);
+
+  IwasakiGaugeActionR GaugeAction(beta);
+
+  // temporarily need a gauge field
+  LatticeGaugeFieldD  U(GridPtr); U=Zero();
+  LatticeGaugeFieldF  UF(GridPtrF); UF=Zero();
+  LatticeGaugeFieldD2 UD2(GridPtrF); UD2=Zero();
+
+  std::cout << GridLogMessage << " Running the HMC "<< std::endl;
+  TheHMC.ReadCommandLine(argc,argv);  // params on CML or from param file
+  TheHMC.initializeGaugeFieldAndRNGs(U);
+  std::cout << "loaded NERSC gauge field"<<std::endl;
+
+  // These lines are unecessary if BC are all periodic
+  std::vector<Complex> boundary = {1,1,1,-1};
+  FermionAction::ImplParams Params(boundary);
+  FermionAction::ImplParams ParamsDir(boundary);
+  FermionActionF::ImplParams ParamsF(boundary);
+  FermionActionF::ImplParams ParamsDirF(boundary);
+  Params.dirichlet=NonDirichlet;
+  ParamsF.dirichlet=NonDirichlet;
+  ParamsDir.dirichlet=Dirichlet;
+  ParamsDirF.dirichlet=Dirichlet;
+
+  //  double StoppingCondition = 1e-14;
+  //  double MDStoppingCondition = 1e-9;
+  double StoppingCondition = 1e-8;
+  double MDStoppingCondition = 1e-7;
+  double MDStoppingConditionLoose = 1e-7;
+  double MDStoppingConditionStrange = 1e-7;
+  double MaxCGIterations = 300000;
+  ConjugateGradient<FermionField>  CG(StoppingCondition,MaxCGIterations);
+  ConjugateGradient<FermionField>  MDCG(MDStoppingCondition,MaxCGIterations);
+
+  ////////////////////////////////////
+  // Collect actions
+  ////////////////////////////////////
+  ActionLevel<HMCWrapper::Field> Level1(1);
+  ActionLevel<HMCWrapper::Field> Level2(2);
+  ActionLevel<HMCWrapper::Field> Level3(30);
+
+  ////////////////////////////////////
+  // Strange action
+  ////////////////////////////////////
+  FermionAction StrangeOp (U,*FGrid,*FrbGrid,*GridPtr,*GridRBPtr,strange_mass,M5,b,c, Params);
+  FermionAction StrangePauliVillarsOp(U,*FGrid,*FrbGrid,*GridPtr,*GridRBPtr,pv_mass,  M5,b,c, Params);
+
+  // Probably dominates the force - back to EOFA.
+  OneFlavourRationalParams SFRp;
+  SFRp.lo       = 0.25;
+  SFRp.hi       = 25.0;
+  SFRp.MaxIter  = 10000;
+  SFRp.tolerance= 1.0e-5;
+  SFRp.mdtolerance= 2.0e-4;
+  SFRp.degree   = 8;
+  SFRp.precision= 50;
+  
+  MobiusEOFAFermionD Strange_Op_L (U , *FGrid , *FrbGrid , *GridPtr , *GridRBPtr , strange_mass, strange_mass, pv_mass, 0.0, -1, M5, b, c);
+  MobiusEOFAFermionF Strange_Op_LF(UF, *FGridF, *FrbGridF, *GridPtrF, *GridRBPtrF, strange_mass, strange_mass, pv_mass, 0.0, -1, M5, b, c);
+  MobiusEOFAFermionD Strange_Op_R (U , *FGrid , *FrbGrid , *GridPtr , *GridRBPtr , pv_mass, strange_mass,      pv_mass, -1.0, 1, M5, b, c);
+  MobiusEOFAFermionF Strange_Op_RF(UF, *FGridF, *FrbGridF, *GridPtrF, *GridRBPtrF, pv_mass, strange_mass,      pv_mass, -1.0, 1, M5, b, c);
+  ConjugateGradient<FermionField>      ActionCG(StoppingCondition,MaxCGIterations);
+  ConjugateGradient<FermionField>  DerivativeCG(MDStoppingCondition,MaxCGIterations);
+  LinearOperatorEOFAD Strange_LinOp_L (Strange_Op_L);
+  LinearOperatorEOFAD Strange_LinOp_R (Strange_Op_R);
+  LinearOperatorEOFAF Strange_LinOp_LF(Strange_Op_LF);
+  LinearOperatorEOFAF Strange_LinOp_RF(Strange_Op_RF);
+
+  const int MX_inner = 1000;
+  MxPCG_EOFA ActionCGL(StoppingCondition,
+		       MX_inner,
+		       MaxCGIterations,
+		       GridPtrF,
+		       FrbGridF,
+		       Strange_Op_LF,Strange_Op_L,
+		       Strange_LinOp_LF,Strange_LinOp_L);
+
+  MxPCG_EOFA DerivativeCGL(MDStoppingConditionStrange,
+			   MX_inner,
+			   MaxCGIterations,
+			   GridPtrF,
+			   FrbGridF,
+			   Strange_Op_LF,Strange_Op_L,
+			   Strange_LinOp_LF,Strange_LinOp_L);
+  
+  MxPCG_EOFA ActionCGR(StoppingCondition,
+		       MX_inner,
+		       MaxCGIterations,
+		       GridPtrF,
+		       FrbGridF,
+		       Strange_Op_RF,Strange_Op_R,
+		       Strange_LinOp_RF,Strange_LinOp_R);
+  
+  MxPCG_EOFA DerivativeCGR(MDStoppingConditionStrange,
+			   MX_inner,
+			   MaxCGIterations,
+			   GridPtrF,
+			   FrbGridF,
+			   Strange_Op_RF,Strange_Op_R,
+			   Strange_LinOp_RF,Strange_LinOp_R);
+
+  ExactOneFlavourRatioPseudoFermionAction<FermionImplPolicy> 
+    EOFA(Strange_Op_L, Strange_Op_R, 
+	 ActionCG, 
+	 ActionCGL, ActionCGR,
+	 DerivativeCGL, DerivativeCGR,
+	 SFRp, true);
+  //  Level2.push_back(&EOFA);
+
+  ////////////////////////////////////
+  // up down action
+  ////////////////////////////////////
+  std::vector<Real> light_den;
+  std::vector<Real> light_num;
+  std::vector<int> dirichlet_den;
+  std::vector<int> dirichlet_num;
+
+  int n_hasenbusch = hasenbusch.size();
+  light_den.push_back(light_mass);  dirichlet_den.push_back(0);
+  for(int h=0;h<n_hasenbusch;h++){
+    light_den.push_back(hasenbusch[h]); dirichlet_den.push_back(1);
+  }
+
+  for(int h=0;h<n_hasenbusch;h++){
+    light_num.push_back(hasenbusch[h]); dirichlet_num.push_back(1);
+  }
+  light_num.push_back(pv_mass);  dirichlet_num.push_back(0);
+
+  std::vector<FermionAction *> Numerators;
+  std::vector<FermionAction *> Denominators;
+  std::vector<FermionActionF *> NumeratorsF;
+  std::vector<FermionActionF *> DenominatorsF;
+  std::vector<FermionActionD2 *> NumeratorsD2;
+  std::vector<FermionActionD2 *> DenominatorsD2;
+  std::vector<TwoFlavourEvenOddRatioPseudoFermionAction<FermionImplPolicy> *> Quotients;
+  std::vector<MxPCG *> ActionMPCG;
+  std::vector<MxPCG *> MPCG;
+  
+#define MIXED_PRECISION
+#ifdef MIXED_PRECISION
+  std::vector<OneFlavourEvenOddRatioRationalMixedPrecPseudoFermionAction<FermionImplPolicy,FermionImplPolicyF,FermionImplPolicyD2> *> Bdys;
+#else
+  std::vector<OneFlavourEvenOddRatioRationalPseudoFermionAction<FermionImplPolicy> *> Bdys;
+#endif
+
+  typedef SchurDiagMooeeOperator<FermionActionF,FermionFieldF> LinearOperatorF;
+  typedef SchurDiagMooeeOperator<FermionAction ,FermionField > LinearOperatorD;
+  std::vector<LinearOperatorD *> LinOpD;
+  std::vector<LinearOperatorF *> LinOpF; 
+  
+  for(int h=0;h<n_hasenbusch+1;h++){
+    std::cout << GridLogMessage
+	      << " 2f quotient Action ";
+    std::cout << "det D("<<light_den[h]<<")";
+    if ( dirichlet_den[h] ) std::cout << "^dirichlet    ";
+    std::cout << "/ det D("<<light_num[h]<<")";
+    if ( dirichlet_num[h] ) std::cout << "^dirichlet    ";
+    std::cout << std::endl;
+
+    FermionAction::ImplParams ParamsNum(boundary);
+    FermionAction::ImplParams ParamsDen(boundary);
+    FermionActionF::ImplParams ParamsDenF(boundary);
+    FermionActionF::ImplParams ParamsNumF(boundary);
+    
+    if ( dirichlet_num[h]==1) ParamsNum.dirichlet = Dirichlet;
+    else                      ParamsNum.dirichlet = NonDirichlet;
+
+    if ( dirichlet_den[h]==1) ParamsDen.dirichlet = Dirichlet;
+    else                      ParamsDen.dirichlet = NonDirichlet;
+
+    Numerators.push_back  (new FermionAction(U,*FGrid,*FrbGrid,*GridPtr,*GridRBPtr,light_num[h],M5,b,c, ParamsNum));
+    Denominators.push_back(new FermionAction(U,*FGrid,*FrbGrid,*GridPtr,*GridRBPtr,light_den[h],M5,b,c, ParamsDen));
+
+    ParamsDenF.dirichlet = ParamsDen.dirichlet;
+    DenominatorsF.push_back(new FermionActionF(UF,*FGridF,*FrbGridF,*GridPtrF,*GridRBPtrF,light_den[h],M5,b,c, ParamsDenF));
+
+    ParamsNumF.dirichlet = ParamsNum.dirichlet;
+    NumeratorsF.push_back  (new FermionActionF(UF,*FGridF,*FrbGridF,*GridPtrF,*GridRBPtrF,light_num[h],M5,b,c, ParamsNumF));
+
+    LinOpD.push_back(new LinearOperatorD(*Denominators[h]));
+    LinOpF.push_back(new LinearOperatorF(*DenominatorsF[h]));
+
+    double conv  = MDStoppingCondition;
+    if (h<3) conv= MDStoppingConditionLoose; // Relax on first two hasenbusch factors
+    const int MX_inner = 5000;
+    MPCG.push_back(new MxPCG(conv,
+			     MX_inner,
+			     MaxCGIterations,
+			     GridPtrF,
+			     FrbGridF,
+			     *DenominatorsF[h],*Denominators[h],
+			     *LinOpF[h], *LinOpD[h]) );
+
+    ActionMPCG.push_back(new MxPCG(StoppingCondition,
+				   MX_inner,
+				   MaxCGIterations,
+				   GridPtrF,
+				   FrbGridF,
+				   *DenominatorsF[h],*Denominators[h],
+				   *LinOpF[h], *LinOpD[h]) );
+
+    
+    if(h!=0) {
+      //      Quotients.push_back (new TwoFlavourEvenOddRatioPseudoFermionAction<FermionImplPolicy>(*Numerators[h],*Denominators[h],MDCG,CG));
+      Quotients.push_back (new TwoFlavourEvenOddRatioPseudoFermionAction<FermionImplPolicy>(*Numerators[h],*Denominators[h],*MPCG[h],*ActionMPCG[h],CG));
+    } else {
+#ifdef MIXED_PRECISION
+      // Use the D2 data types and make them use same grid as single
+      FermionActionD2::ImplParams ParamsDenD2(boundary);
+      FermionActionD2::ImplParams ParamsNumD2(boundary);
+
+      ParamsDenD2.dirichlet = ParamsDen.dirichlet;
+      DenominatorsD2.push_back(new FermionActionD2(UD2,*FGridF,*FrbGridF,*GridPtrF,*GridRBPtrF,light_den[h],M5,b,c, ParamsDenD2));
+
+      ParamsNumD2.dirichlet = ParamsNum.dirichlet;
+      NumeratorsD2.push_back  (new FermionActionD2(UD2,*FGridF,*FrbGridF,*GridPtrF,*GridRBPtrF,light_num[h],M5,b,c, ParamsNumD2));
+    
+      Bdys.push_back( new OneFlavourEvenOddRatioRationalMixedPrecPseudoFermionAction<FermionImplPolicy,FermionImplPolicyF,FermionImplPolicyD2>(
+			   *Numerators[h],*Denominators[h],
+			   *NumeratorsF[h],*DenominatorsF[h],
+			   *NumeratorsD2[h],*DenominatorsD2[h],
+			   OFRp, 200) );
+      Bdys.push_back( new OneFlavourEvenOddRatioRationalMixedPrecPseudoFermionAction<FermionImplPolicy,FermionImplPolicyF,FermionImplPolicyD2>(
+			   *Numerators[h],*Denominators[h],
+			   *NumeratorsF[h],*DenominatorsF[h],
+			   *NumeratorsD2[h],*DenominatorsD2[h],
+			   OFRp, 200) );
+#else
+      Bdys.push_back( new OneFlavourEvenOddRatioRationalPseudoFermionAction<FermionImplPolicy>(*Numerators[h],*Denominators[h],OFRp));
+      Bdys.push_back( new OneFlavourEvenOddRatioRationalPseudoFermionAction<FermionImplPolicy>(*Numerators[h],*Denominators[h],OFRp));
+#endif
+    }
+  }
+  for(int h=0;h<Bdys.size();h++){
+    Bdys[h]->SetTolerances(ActionTolByPole,MDTolByPole);
+  }
+  int nquo=Quotients.size();
+  Level1.push_back(Bdys[0]);
+  Level1.push_back(Bdys[1]);
+  for(int h=0;h<nquo-1;h++){
+    Level2.push_back(Quotients[h]);
+  }
+  Level2.push_back(Quotients[nquo-1]);
+
+  /////////////////////////////////////////////////////////////
+  // Gauge action
+  /////////////////////////////////////////////////////////////
+  Level3.push_back(&GaugeAction);
+  TheHMC.TheAction.push_back(Level1);
+  TheHMC.TheAction.push_back(Level2);
+  TheHMC.TheAction.push_back(Level3);
+  std::cout << GridLogMessage << " Action complete "<< std::endl;
+  /////////////////////////////////////////////////////////////
+
+  TheHMC.Run();  // no smearing
+
+  Grid_finalize();
+} // main
+
+
+

From e047616571ac33d7d63ebe8865d150e77d850711 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 15 Nov 2022 16:39:39 -0500
Subject: [PATCH 081/146] Multilevel integrator test

---
 HMC/Mobius2p1f_DD_EOFA_96I_3level.cc | 516 +++++++++++++++++++++++++++
 1 file changed, 516 insertions(+)
 create mode 100644 HMC/Mobius2p1f_DD_EOFA_96I_3level.cc

diff --git a/HMC/Mobius2p1f_DD_EOFA_96I_3level.cc b/HMC/Mobius2p1f_DD_EOFA_96I_3level.cc
new file mode 100644
index 00000000..c305567c
--- /dev/null
+++ b/HMC/Mobius2p1f_DD_EOFA_96I_3level.cc
@@ -0,0 +1,516 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./tests/Test_hmc_EODWFRatio.cc
+
+Copyright (C) 2015-2016
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Guido Cossu <guido.cossu@ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/Grid.h>
+
+NAMESPACE_BEGIN(Grid);
+
+template<class FermionOperatorD, class FermionOperatorF, class SchurOperatorD, class  SchurOperatorF> 
+  class MixedPrecisionConjugateGradientOperatorFunction : public OperatorFunction<typename FermionOperatorD::FermionField> {
+  public:
+    typedef typename FermionOperatorD::FermionField FieldD;
+    typedef typename FermionOperatorF::FermionField FieldF;
+
+    using OperatorFunction<FieldD>::operator();
+
+    RealD   Tolerance;
+    RealD   InnerTolerance; //Initial tolerance for inner CG. Defaults to Tolerance but can be changed
+    Integer MaxInnerIterations;
+    Integer MaxOuterIterations;
+    GridBase* SinglePrecGrid4; //Grid for single-precision fields
+    GridBase* SinglePrecGrid5; //Grid for single-precision fields
+    RealD OuterLoopNormMult; //Stop the outer loop and move to a final double prec solve when the residual is OuterLoopNormMult * Tolerance
+
+    FermionOperatorF &FermOpF;
+    FermionOperatorD &FermOpD;;
+    SchurOperatorF &LinOpF;
+    SchurOperatorD &LinOpD;
+
+    Integer TotalInnerIterations; //Number of inner CG iterations
+    Integer TotalOuterIterations; //Number of restarts
+    Integer TotalFinalStepIterations; //Number of CG iterations in final patch-up step
+
+    MixedPrecisionConjugateGradientOperatorFunction(RealD tol, 
+						    Integer maxinnerit, 
+						    Integer maxouterit, 
+						    GridBase* _sp_grid4, 
+						    GridBase* _sp_grid5, 
+						    FermionOperatorF &_FermOpF,
+						    FermionOperatorD &_FermOpD,
+						    SchurOperatorF   &_LinOpF,
+						    SchurOperatorD   &_LinOpD): 
+      LinOpF(_LinOpF),
+      LinOpD(_LinOpD),
+      FermOpF(_FermOpF),
+      FermOpD(_FermOpD),
+      Tolerance(tol), 
+      InnerTolerance(tol), 
+      MaxInnerIterations(maxinnerit), 
+      MaxOuterIterations(maxouterit), 
+      SinglePrecGrid4(_sp_grid4),
+      SinglePrecGrid5(_sp_grid5),
+      OuterLoopNormMult(100.) 
+    { 
+      /* Debugging instances of objects; references are stored
+      std::cout << GridLogMessage << " Mixed precision CG wrapper LinOpF " <<std::hex<< &LinOpF<<std::dec <<std::endl;
+      std::cout << GridLogMessage << " Mixed precision CG wrapper LinOpD " <<std::hex<< &LinOpD<<std::dec <<std::endl;
+      std::cout << GridLogMessage << " Mixed precision CG wrapper FermOpF " <<std::hex<< &FermOpF<<std::dec <<std::endl;
+      std::cout << GridLogMessage << " Mixed precision CG wrapper FermOpD " <<std::hex<< &FermOpD<<std::dec <<std::endl;
+      */
+    };
+
+    void operator()(LinearOperatorBase<FieldD> &LinOpU, const FieldD &src, FieldD &psi) {
+
+      std::cout << GridLogMessage << " Mixed precision CG wrapper operator() "<<std::endl;
+
+      SchurOperatorD * SchurOpU = static_cast<SchurOperatorD *>(&LinOpU);
+      
+      //      std::cout << GridLogMessage << " Mixed precision CG wrapper operator() FermOpU " <<std::hex<< &(SchurOpU->_Mat)<<std::dec <<std::endl;
+      //      std::cout << GridLogMessage << " Mixed precision CG wrapper operator() FermOpD " <<std::hex<< &(LinOpD._Mat) <<std::dec <<std::endl;
+      // Assumption made in code to extract gauge field
+      // We could avoid storing LinopD reference alltogether ?
+      assert(&(SchurOpU->_Mat)==&(LinOpD._Mat));
+
+      ////////////////////////////////////////////////////////////////////////////////////
+      // Must snarf a single precision copy of the gauge field in Linop_d argument
+      ////////////////////////////////////////////////////////////////////////////////////
+      typedef typename FermionOperatorF::GaugeField GaugeFieldF;
+      typedef typename FermionOperatorF::GaugeLinkField GaugeLinkFieldF;
+      typedef typename FermionOperatorD::GaugeField GaugeFieldD;
+      typedef typename FermionOperatorD::GaugeLinkField GaugeLinkFieldD;
+
+      GridBase * GridPtrF = SinglePrecGrid4;
+      GridBase * GridPtrD = FermOpD.Umu.Grid();
+      GaugeFieldF     U_f  (GridPtrF);
+      GaugeLinkFieldF Umu_f(GridPtrF);
+      //      std::cout << " Dim gauge field "<<GridPtrF->Nd()<<std::endl; // 4d
+      //      std::cout << " Dim gauge field "<<GridPtrD->Nd()<<std::endl; // 4d
+
+      ////////////////////////////////////////////////////////////////////////////////////
+      // Moving this to a Clone method of fermion operator would allow to duplicate the 
+      // physics parameters and decrease gauge field copies
+      ////////////////////////////////////////////////////////////////////////////////////
+      GaugeLinkFieldD Umu_d(GridPtrD);
+      for(int mu=0;mu<Nd*2;mu++){ 
+	Umu_d = PeekIndex<LorentzIndex>(FermOpD.Umu, mu);
+	precisionChange(Umu_f,Umu_d);
+	PokeIndex<LorentzIndex>(FermOpF.Umu, Umu_f, mu);
+      }
+      pickCheckerboard(Even,FermOpF.UmuEven,FermOpF.Umu);
+      pickCheckerboard(Odd ,FermOpF.UmuOdd ,FermOpF.Umu);
+
+      ////////////////////////////////////////////////////////////////////////////////////
+      // Make a mixed precision conjugate gradient
+      ////////////////////////////////////////////////////////////////////////////////////
+#if 1
+      RealD delta=1.e-4;
+      std::cout << GridLogMessage << "Calling reliable update Conjugate Gradient" <<std::endl;
+      ConjugateGradientReliableUpdate<FieldD,FieldF> MPCG(Tolerance,MaxInnerIterations*MaxOuterIterations,delta,SinglePrecGrid5,LinOpF,LinOpD);
+#else      
+      std::cout << GridLogMessage << "Calling mixed precision Conjugate Gradient" <<std::endl;
+      MixedPrecisionConjugateGradient<FieldD,FieldF> MPCG(Tolerance,MaxInnerIterations,MaxOuterIterations,SinglePrecGrid5,LinOpF,LinOpD);
+#endif
+      MPCG(src,psi);
+    }
+  };
+
+NAMESPACE_END(Grid);
+
+
+int main(int argc, char **argv) {
+  using namespace Grid;
+
+  Grid_init(&argc, &argv);
+
+  CartesianCommunicator::BarrierWorld();
+  std::cout << GridLogMessage << " Clock skew check" <<std::endl;
+  
+  int threads = GridThread::GetThreads();
+
+   // Typedefs to simplify notation
+  typedef WilsonImplR FermionImplPolicy;
+  typedef WilsonImplF FermionImplPolicyF;
+  typedef MobiusFermionD FermionAction;
+  typedef MobiusFermionF FermionActionF;
+  typedef MobiusEOFAFermionD FermionEOFAAction;
+  typedef MobiusEOFAFermionF FermionEOFAActionF;
+  typedef typename FermionAction::FermionField FermionField;
+  typedef typename FermionActionF::FermionField FermionFieldF;
+
+  typedef Grid::XmlReader       Serialiser;
+
+  //::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
+  IntegratorParameters MD;
+  //  typedef GenericHMCRunner<LeapFrog> HMCWrapper;
+  //  MD.name    = std::string("Leap Frog");
+  //  typedef GenericHMCRunner<ForceGradient> HMCWrapper;
+  //  MD.name    = std::string("Force Gradient");
+  typedef GenericHMCRunner<MinimumNorm2> HMCWrapper;
+   MD.name    = std::string("MinimumNorm2");
+  MD.MDsteps =  4;
+  MD.trajL   = 0.5;
+
+  HMCparameters HMCparams;
+  HMCparams.StartTrajectory  = 1077;
+  HMCparams.Trajectories     = 1;
+  HMCparams.NoMetropolisUntil=  0;
+  // "[HotStart, ColdStart, TepidStart, CheckpointStart]\n";
+  //  HMCparams.StartingType     =std::string("ColdStart");
+  HMCparams.StartingType     =std::string("CheckpointStart");
+  HMCparams.MD = MD;
+  HMCWrapper TheHMC(HMCparams);
+
+  // Grid from the command line arguments --grid and --mpi
+  TheHMC.Resources.AddFourDimGrid("gauge"); // use default simd lanes decomposition
+
+  CheckpointerParameters CPparams;
+  CPparams.config_prefix = "ckpoint_DDHMC_lat";
+  CPparams.rng_prefix    = "ckpoint_DDHMC_rng";
+  CPparams.saveInterval  = 1;
+  CPparams.format        = "IEEE64BIG";
+  TheHMC.Resources.LoadNerscCheckpointer(CPparams);
+  std::cout << "loaded NERSC checpointer"<<std::endl;
+  RNGModuleParameters RNGpar;
+  RNGpar.serial_seeds = "1 2 3 4 5";
+  RNGpar.parallel_seeds = "6 7 8 9 10";
+  TheHMC.Resources.SetRNGSeeds(RNGpar);
+
+  // Construct observables
+  // here there is too much indirection
+  typedef PlaquetteMod<HMCWrapper::ImplPolicy> PlaqObs;
+  TheHMC.Resources.AddObservable<PlaqObs>();
+  //////////////////////////////////////////////
+
+  const int Ls      = 12;
+  RealD M5  = 1.8;
+  RealD b   = 1.5;
+  RealD c   = 0.5;
+  Real beta         = 2.13;
+  //  Real light_mass   = 5.4e-4;
+  Real light_mass   = 7.8e-4;
+  Real strange_mass = 0.0362;
+  Real pv_mass      = 1.0;
+  //  std::vector<Real> hasenbusch({ light_mass, 3.8e-3, 0.0145, 0.045, 0.108, 0.25, 0.51 , pv_mass });
+  std::vector<Real> hasenbusch({ light_mass, 5e-3, 0.0145, 0.045, 0.108, 0.25, 0.51 , 0.75 , pv_mass });
+
+
+  OneFlavourRationalParams OFRp; // Up/down
+  OFRp.lo       = 2.0e-5;
+  OFRp.hi       = 90.0;
+  OFRp.MaxIter  = 60000;
+  OFRp.tolerance= 1.0e-5;
+  OFRp.mdtolerance= 1.0e-3;
+  //  OFRp.degree   = 20; converges
+  //  OFRp.degree   = 16;
+  OFRp.degree   = 12;
+  OFRp.precision= 80;
+  OFRp.BoundsCheckFreq=0;
+  std::vector<RealD> ActionTolByPole({
+      1.0e-6,1.0e-6,1.0e-7,1.0e-8,
+      1.0e-8,1.0e-8,1.0e-8,1.0e-8,
+      1.0e-10,1.0e-10,1.0e-10,1.0e-10
+    });
+  std::vector<RealD> MDTolByPole({
+      3.0e-4,3.0e-4,3.0e-5,1.0e-5,
+      1.0e-5,1.0e-5,1.0e-5,1.0e-5,
+      1.0e-8,1.0e-10,1.0e-10,1.0e-10
+    });
+
+  auto GridPtr   = TheHMC.Resources.GetCartesian();
+  auto GridRBPtr = TheHMC.Resources.GetRBCartesian();
+
+  typedef SchurDiagMooeeOperator<FermionActionF,FermionFieldF> LinearOperatorF;
+  typedef SchurDiagMooeeOperator<FermionAction ,FermionField > LinearOperatorD;
+  typedef SchurDiagMooeeOperator<FermionEOFAActionF,FermionFieldF> LinearOperatorEOFAF;
+  typedef SchurDiagMooeeOperator<FermionEOFAAction ,FermionField > LinearOperatorEOFAD;
+  typedef MixedPrecisionConjugateGradientOperatorFunction<MobiusFermionD,MobiusFermionF,LinearOperatorD,LinearOperatorF> MxPCG;
+  typedef MixedPrecisionConjugateGradientOperatorFunction<MobiusEOFAFermionD,MobiusEOFAFermionF,LinearOperatorEOFAD,LinearOperatorEOFAF> MxPCG_EOFA;
+
+  ////////////////////////////////////////////////////////////////
+  // Domain decomposed
+  ////////////////////////////////////////////////////////////////
+  Coordinate latt4  = GridPtr->GlobalDimensions();
+  Coordinate mpi    = GridPtr->ProcessorGrid();
+  Coordinate shm;
+
+  GlobalSharedMemory::GetShmDims(mpi,shm);
+  
+  Coordinate CommDim(Nd);
+  for(int d=0;d<Nd;d++) CommDim[d]= (mpi[d]/shm[d])>1 ? 1 : 0;
+
+  Coordinate NonDirichlet(Nd+1,0);
+  Coordinate Dirichlet(Nd+1,0);
+  Dirichlet[1] = CommDim[0]*latt4[0]/mpi[0] * shm[0];
+  Dirichlet[2] = CommDim[1]*latt4[1]/mpi[1] * shm[1];
+  Dirichlet[3] = CommDim[2]*latt4[2]/mpi[2] * shm[2];
+  Dirichlet[4] = CommDim[3]*latt4[3]/mpi[3] * shm[3];
+
+  Coordinate Block4(Nd);
+  Block4[0] = Dirichlet[1];
+  Block4[1] = Dirichlet[2];
+  Block4[2] = Dirichlet[3];
+  Block4[3] = Dirichlet[4];
+
+  int Width=3;
+  TheHMC.Resources.SetMomentumFilter(new DDHMCFilter<WilsonImplR::Field>(Block4,Width));
+
+  //////////////////////////
+  // Fermion Grids
+  //////////////////////////
+  auto FGrid     = SpaceTimeGrid::makeFiveDimGrid(Ls,GridPtr);
+  auto FrbGrid   = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,GridPtr);
+
+  Coordinate simdF = GridDefaultSimd(Nd,vComplexF::Nsimd());
+  auto GridPtrF   = SpaceTimeGrid::makeFourDimGrid(latt4,simdF,mpi);
+  auto GridRBPtrF = SpaceTimeGrid::makeFourDimRedBlackGrid(GridPtrF);
+  auto FGridF     = SpaceTimeGrid::makeFiveDimGrid(Ls,GridPtrF);
+  auto FrbGridF   = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,GridPtrF);
+
+  IwasakiGaugeActionR GaugeAction(beta);
+
+  // temporarily need a gauge field
+  LatticeGaugeField U(GridPtr);
+  LatticeGaugeFieldF UF(GridPtrF);
+
+  std::cout << GridLogMessage << " Running the HMC "<< std::endl;
+  TheHMC.ReadCommandLine(argc,argv);  // params on CML or from param file
+  TheHMC.initializeGaugeFieldAndRNGs(U);
+  std::cout << "loaded NERSC gauge field"<<std::endl;
+
+
+  // These lines are unecessary if BC are all periodic
+  std::vector<Complex> boundary = {1,1,1,-1};
+  FermionAction::ImplParams Params(boundary);
+  Params.dirichlet=NonDirichlet;
+  FermionAction::ImplParams ParamsDir(boundary);
+  ParamsDir.dirichlet=Dirichlet;
+
+  //  double StoppingCondition = 1e-14;
+  //  double MDStoppingCondition = 1e-9;
+  double StoppingCondition = 1e-8;
+  double MDStoppingCondition = 1e-6;
+  double MDStoppingConditionLoose = 1e-6;
+  double MDStoppingConditionStrange = 1e-8;
+  double MaxCGIterations = 300000;
+  ConjugateGradient<FermionField>  CG(StoppingCondition,MaxCGIterations);
+  ConjugateGradient<FermionField>  MDCG(MDStoppingCondition,MaxCGIterations);
+
+  ////////////////////////////////////
+  // Collect actions
+  ////////////////////////////////////
+  ActionLevel<HMCWrapper::Field> Level1(1);
+  ActionLevel<HMCWrapper::Field> Level2(8); // 6 x 20 = 120 = 8 x 15 
+  ActionLevel<HMCWrapper::Field> Level3(15);
+
+  ////////////////////////////////////
+  // Strange action
+  ////////////////////////////////////
+  FermionAction StrangeOp (U,*FGrid,*FrbGrid,*GridPtr,*GridRBPtr,strange_mass,M5,b,c, Params);
+  FermionAction StrangePauliVillarsOp(U,*FGrid,*FrbGrid,*GridPtr,*GridRBPtr,pv_mass,  M5,b,c, Params);
+
+  // Probably dominates the force - back to EOFA.
+  OneFlavourRationalParams SFRp;
+  SFRp.lo       = 0.1;
+  SFRp.hi       = 25.0;
+  SFRp.MaxIter  = 10000;
+  SFRp.tolerance= 1.0e-5;
+  SFRp.mdtolerance= 1.0e-3;
+  SFRp.degree   = 14;
+  SFRp.precision= 50;
+  
+  MobiusEOFAFermionD Strange_Op_L (U , *FGrid , *FrbGrid , *GridPtr , *GridRBPtr , strange_mass, strange_mass, pv_mass, 0.0, -1, M5, b, c);
+  MobiusEOFAFermionF Strange_Op_LF(UF, *FGridF, *FrbGridF, *GridPtrF, *GridRBPtrF, strange_mass, strange_mass, pv_mass, 0.0, -1, M5, b, c);
+  MobiusEOFAFermionD Strange_Op_R (U , *FGrid , *FrbGrid , *GridPtr , *GridRBPtr , pv_mass, strange_mass,      pv_mass, -1.0, 1, M5, b, c);
+  MobiusEOFAFermionF Strange_Op_RF(UF, *FGridF, *FrbGridF, *GridPtrF, *GridRBPtrF, pv_mass, strange_mass,      pv_mass, -1.0, 1, M5, b, c);
+  ConjugateGradient<FermionField>      ActionCG(StoppingCondition,MaxCGIterations);
+  ConjugateGradient<FermionField>  DerivativeCG(MDStoppingCondition,MaxCGIterations);
+  LinearOperatorEOFAD Strange_LinOp_L (Strange_Op_L);
+  LinearOperatorEOFAD Strange_LinOp_R (Strange_Op_R);
+  LinearOperatorEOFAF Strange_LinOp_LF(Strange_Op_LF);
+  LinearOperatorEOFAF Strange_LinOp_RF(Strange_Op_RF);
+
+  const int MX_inner = 1000;
+  MxPCG_EOFA ActionCGL(StoppingCondition,
+		       MX_inner,
+		       MaxCGIterations,
+		       GridPtrF,
+		       FrbGridF,
+		       Strange_Op_LF,Strange_Op_L,
+		       Strange_LinOp_LF,Strange_LinOp_L);
+
+  MxPCG_EOFA DerivativeCGL(MDStoppingConditionStrange,
+			   MX_inner,
+			   MaxCGIterations,
+			   GridPtrF,
+			   FrbGridF,
+			   Strange_Op_LF,Strange_Op_L,
+			   Strange_LinOp_LF,Strange_LinOp_L);
+  
+  MxPCG_EOFA ActionCGR(StoppingCondition,
+		       MX_inner,
+		       MaxCGIterations,
+		       GridPtrF,
+		       FrbGridF,
+		       Strange_Op_RF,Strange_Op_R,
+		       Strange_LinOp_RF,Strange_LinOp_R);
+  
+  MxPCG_EOFA DerivativeCGR(MDStoppingConditionStrange,
+			   MX_inner,
+			   MaxCGIterations,
+			   GridPtrF,
+			   FrbGridF,
+			   Strange_Op_RF,Strange_Op_R,
+			   Strange_LinOp_RF,Strange_LinOp_R);
+
+  ExactOneFlavourRatioPseudoFermionAction<FermionImplPolicy> 
+    EOFA(Strange_Op_L, Strange_Op_R, 
+	 ActionCG, 
+	 ActionCGL, ActionCGR,
+	 DerivativeCGL, DerivativeCGR,
+	 SFRp, true);
+  Level2.push_back(&EOFA);
+
+  ////////////////////////////////////
+  // up down action
+  ////////////////////////////////////
+  std::vector<Real> light_den;
+  std::vector<Real> light_num;
+  std::vector<int> dirichlet_den;
+  std::vector<int> dirichlet_num;
+
+  int n_hasenbusch = hasenbusch.size();
+  light_den.push_back(light_mass);  dirichlet_den.push_back(0);
+  for(int h=0;h<n_hasenbusch;h++){
+    light_den.push_back(hasenbusch[h]); dirichlet_den.push_back(1);
+  }
+
+  for(int h=0;h<n_hasenbusch;h++){
+    light_num.push_back(hasenbusch[h]); dirichlet_num.push_back(1);
+  }
+  light_num.push_back(pv_mass);  dirichlet_num.push_back(0);
+
+  std::vector<FermionAction *> Numerators;
+  std::vector<FermionAction *> Denominators;
+  std::vector<FermionActionF *> DenominatorsF;
+  std::vector<TwoFlavourEvenOddRatioPseudoFermionAction<FermionImplPolicy> *> Quotients;
+  std::vector<OneFlavourEvenOddRatioRationalPseudoFermionAction<FermionImplPolicy> *> Bdys;
+  std::vector<MxPCG *> ActionMPCG;
+  std::vector<MxPCG *> MPCG;
+
+  typedef SchurDiagMooeeOperator<FermionActionF,FermionFieldF> LinearOperatorF;
+  typedef SchurDiagMooeeOperator<FermionAction ,FermionField > LinearOperatorD;
+  std::vector<LinearOperatorD *> LinOpD;
+  std::vector<LinearOperatorF *> LinOpF; 
+  
+  for(int h=0;h<n_hasenbusch+1;h++){
+    std::cout << GridLogMessage
+	      << " 2f quotient Action ";
+    std::cout << "det D("<<light_den[h]<<")";
+    if ( dirichlet_den[h] ) std::cout << "^dirichlet    ";
+    std::cout << "/ det D("<<light_num[h]<<")";
+    if ( dirichlet_num[h] ) std::cout << "^dirichlet    ";
+    std::cout << std::endl;
+
+    FermionAction::ImplParams ParamsNum(boundary);
+    FermionAction::ImplParams ParamsDen(boundary);
+    FermionActionF::ImplParams ParamsDenF(boundary);
+    
+    if ( dirichlet_num[h]==1) ParamsNum.dirichlet = Dirichlet;
+    else                      ParamsNum.dirichlet = NonDirichlet;
+    Numerators.push_back  (new FermionAction(U,*FGrid,*FrbGrid,*GridPtr,*GridRBPtr,light_num[h],M5,b,c, ParamsNum));
+
+    if ( dirichlet_den[h]==1) ParamsDen.dirichlet = Dirichlet;
+    else                      ParamsDen.dirichlet = NonDirichlet;
+
+    Denominators.push_back(new FermionAction(U,*FGrid,*FrbGrid,*GridPtr,*GridRBPtr,light_den[h],M5,b,c, ParamsDen));
+
+    ParamsDenF.dirichlet = ParamsDen.dirichlet;
+    DenominatorsF.push_back(new FermionActionF(UF,*FGridF,*FrbGridF,*GridPtrF,*GridRBPtrF,light_den[h],M5,b,c, ParamsDenF));
+
+    LinOpD.push_back(new LinearOperatorD(*Denominators[h]));
+    LinOpF.push_back(new LinearOperatorF(*DenominatorsF[h]));
+
+    double conv  = MDStoppingCondition;
+    if (h<3) conv= MDStoppingConditionLoose; // Relax on first two hasenbusch factors
+    const int MX_inner = 5000;
+    MPCG.push_back(new MxPCG(conv,
+			     MX_inner,
+			     MaxCGIterations,
+			     GridPtrF,
+			     FrbGridF,
+			     *DenominatorsF[h],*Denominators[h],
+			     *LinOpF[h], *LinOpD[h]) );
+
+    ActionMPCG.push_back(new MxPCG(StoppingCondition,
+				   MX_inner,
+				   MaxCGIterations,
+				   GridPtrF,
+				   FrbGridF,
+				   *DenominatorsF[h],*Denominators[h],
+				   *LinOpF[h], *LinOpD[h]) );
+
+    
+    if(h!=0) {
+      //      Quotients.push_back (new TwoFlavourEvenOddRatioPseudoFermionAction<FermionImplPolicy>(*Numerators[h],*Denominators[h],MDCG,CG));
+      Quotients.push_back (new TwoFlavourEvenOddRatioPseudoFermionAction<FermionImplPolicy>(*Numerators[h],*Denominators[h],*MPCG[h],*ActionMPCG[h],CG));
+    } else {
+      Bdys.push_back( new OneFlavourEvenOddRatioRationalPseudoFermionAction<FermionImplPolicy>(*Numerators[h],*Denominators[h],OFRp));
+      Bdys.push_back( new OneFlavourEvenOddRatioRationalPseudoFermionAction<FermionImplPolicy>(*Numerators[h],*Denominators[h],OFRp));
+    }
+  }
+  for(int h=0;h<Bdys.size();h++){
+    Bdys[h]->SetTolerances(ActionTolByPole,MDTolByPole);
+  }
+  int nquo=Quotients.size();
+  //  Level1.push_back(Bdys[0]);
+  //  Level1.push_back(Bdys[1]);
+  //  Level2.push_back(Bdys[0]);
+  //  Level2.push_back(Bdys[1]);
+  for(int h=0;h<nquo-1;h++){
+    //    Level2.push_back(Quotients[h]);
+  }
+  //  Level2.push_back(Quotients[nquo-1]);
+
+  /////////////////////////////////////////////////////////////
+  // Gauge action
+  /////////////////////////////////////////////////////////////
+  Level3.push_back(&GaugeAction);
+  TheHMC.TheAction.push_back(Level1);
+  TheHMC.TheAction.push_back(Level2);
+  TheHMC.TheAction.push_back(Level3);
+  std::cout << GridLogMessage << " Action complete "<< std::endl;
+  /////////////////////////////////////////////////////////////
+
+  TheHMC.Run();  // no smearing
+
+  Grid_finalize();
+} // main
+
+
+

From 0ae0e5f43678b69f94c23c0e1eb3ab15156fa5e7 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 15 Nov 2022 16:40:38 -0500
Subject: [PATCH 082/146] Partial Dirichlet test

---
 benchmarks/Benchmark_dwf_fp32_partial.cc | 463 +++++++++++++++++++++++
 1 file changed, 463 insertions(+)
 create mode 100644 benchmarks/Benchmark_dwf_fp32_partial.cc

diff --git a/benchmarks/Benchmark_dwf_fp32_partial.cc b/benchmarks/Benchmark_dwf_fp32_partial.cc
new file mode 100644
index 00000000..4db3022a
--- /dev/null
+++ b/benchmarks/Benchmark_dwf_fp32_partial.cc
@@ -0,0 +1,463 @@
+ /*************************************************************************************
+    Grid physics library, www.github.com/paboyle/Grid
+    Source file: ./benchmarks/Benchmark_dwf.cc
+    Copyright (C) 2015
+
+    Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+    Author: paboyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+    See the full license in the file "LICENSE" in the top level distribution directory
+    *************************************************************************************/
+    /*  END LEGAL */
+#include <Grid/Grid.h>
+#ifdef GRID_CUDA
+#define CUDA_PROFILE
+#endif
+
+#ifdef CUDA_PROFILE
+#include <cuda_profiler_api.h>
+#endif
+
+using namespace std;
+using namespace Grid;
+
+////////////////////////
+/// Move to domains ////
+////////////////////////
+
+Gamma::Algebra Gmu [] = {
+			 Gamma::Algebra::GammaX,
+			 Gamma::Algebra::GammaY,
+			 Gamma::Algebra::GammaZ,
+			 Gamma::Algebra::GammaT
+};
+
+void Benchmark(int Ls, Coordinate Dirichlet, int partial);
+
+int main (int argc, char ** argv)
+{
+  Grid_init(&argc,&argv);
+
+
+  int threads = GridThread::GetThreads();
+
+  int Ls=8;
+  for(int i=0;i<argc;i++) {
+    if(std::string(argv[i]) == "-Ls"){
+      std::stringstream ss(argv[i+1]); ss >> Ls;
+    }
+  }
+
+  //////////////////
+  // With comms
+  //////////////////
+  Coordinate Dirichlet(Nd+1,0);
+
+  for(auto partial : {0}) {
+  std::cout << "\n\n\n\n\n\n" <<std::endl;
+  std::cout << GridLogMessage<< "++++++++++++++++++++++++++++++++++++++++++++++++" <<std::endl;
+  std::cout << GridLogMessage<< " Testing with full communication " <<std::endl;
+  std::cout << GridLogMessage<< "++++++++++++++++++++++++++++++++++++++++++++++++" <<std::endl;
+    Benchmark(Ls,Dirichlet,partial);
+  }
+
+  //////////////////
+  // Domain decomposed
+  //////////////////
+  Coordinate latt4  = GridDefaultLatt();
+  Coordinate mpi    = GridDefaultMpi();
+  Coordinate CommDim(Nd);
+  //Coordinate shm({2,1,1,1});
+  Coordinate shm;
+  GlobalSharedMemory::GetShmDims(mpi,shm);
+
+  std::cout <<GridLogMessage << " Shared memory MPI decomp is " <<shm<<std::endl;
+
+  //////////////////////
+  // Node level
+  //////////////////////
+  for(int d=0;d<Nd;d++) CommDim[d]= (mpi[d]/shm[d])>1 ? 1 : 0;
+  Dirichlet[0] = 0;
+  Dirichlet[1] = CommDim[0]*latt4[0]/mpi[0] * shm[0];
+  Dirichlet[2] = CommDim[1]*latt4[1]/mpi[1] * shm[1];
+  Dirichlet[3] = CommDim[2]*latt4[2]/mpi[2] * shm[2];
+  Dirichlet[4] = CommDim[3]*latt4[3]/mpi[3] * shm[3];
+
+  for(auto partial : {0,1}) {
+    std::cout << "\n\n\n\n\n\n" <<std::endl;
+    std::cout << GridLogMessage<< "++++++++++++++++++++++++++++++++++++++++++++++++" <<std::endl;
+    std::cout << GridLogMessage<< " Testing without internode communication partial dirichlet="<<partial <<std::endl;
+    std::cout << GridLogMessage<< "++++++++++++++++++++++++++++++++++++++++++++++++" <<std::endl;
+    Benchmark(Ls,Dirichlet,partial);
+  }
+  
+  for(int d=0;d<Nd;d++) CommDim[d]= mpi[d]>1 ? 1 : 0;
+  Dirichlet[0] = 0;
+  Dirichlet[1] = CommDim[0]*latt4[0]/mpi[0];
+  Dirichlet[2] = CommDim[1]*latt4[1]/mpi[1];
+  Dirichlet[3] = CommDim[2]*latt4[2]/mpi[2];
+  Dirichlet[4] = CommDim[3]*latt4[3]/mpi[3];
+  
+  for(auto partial : {0,1}) {
+    std::cout << "\n\n\n\n\n\n" <<std::endl;
+    std::cout << GridLogMessage<< "++++++++++++++++++++++++++++++++++++++++++++++++" <<std::endl;
+    std::cout << GridLogMessage<< " Testing without intranode communication; partial dirichlet= "<<partial <<std::endl;
+    std::cout << GridLogMessage<< "++++++++++++++++++++++++++++++++++++++++++++++++" <<std::endl;
+    Benchmark(Ls,Dirichlet,partial);
+  }
+  Grid_finalize();
+  exit(0);
+}
+void Benchmark(int Ls, Coordinate Dirichlet, int partial)
+{
+  Coordinate latt4 = GridDefaultLatt();
+  GridLogLayout();
+
+  long unsigned int single_site_flops = 8*Nc*(7+16*Nc);
+
+  std::vector<int> seeds4({1,2,3,4});
+  std::vector<int> seeds5({5,6,7,8});
+#define SINGLE
+#ifdef SINGLE
+  typedef vComplexF          Simd;
+  typedef LatticeFermionF    FermionField;
+  typedef LatticeGaugeFieldF GaugeField;
+  typedef LatticeColourMatrixF ColourMatrixField;
+  typedef DomainWallFermionF FermionAction;
+#endif
+#ifdef DOUBLE
+  typedef vComplexD          Simd;
+  typedef LatticeFermionD    FermionField;
+  typedef LatticeGaugeFieldD GaugeField;
+  typedef LatticeColourMatrixD ColourMatrixField;
+  typedef DomainWallFermionD FermionAction;
+#endif
+#ifdef DOUBLE2
+  typedef vComplexD2          Simd;
+  typedef LatticeFermionD2    FermionField;
+  typedef LatticeGaugeFieldD2 GaugeField;
+  typedef LatticeColourMatrixD2 ColourMatrixField;
+  typedef DomainWallFermionD2 FermionAction;
+#endif
+  
+  GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,Simd::Nsimd()),GridDefaultMpi());
+  GridRedBlackCartesian * UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
+  GridCartesian         * FGrid   = SpaceTimeGrid::makeFiveDimGrid(Ls,UGrid);
+  GridRedBlackCartesian * FrbGrid = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGrid);
+
+  std::cout << GridLogMessage << "Initialising 4d RNG" << std::endl;
+  GridParallelRNG          RNG4(UGrid);  RNG4.SeedUniqueString(std::string("The 4D RNG"));
+
+  std::cout << GridLogMessage << "Initialising 5d RNG" << std::endl;
+  GridParallelRNG          RNG5(FGrid);  RNG5.SeedUniqueString(std::string("The 5D RNG"));
+
+ 
+  FermionField src   (FGrid); random(RNG5,src);
+#if 0
+  src = Zero();
+  {
+    Coordinate origin({0,0,0,latt4[2]-1,0});
+    SpinColourVectorF tmp;
+    tmp=Zero();
+    tmp()(0)(0)=Complex(-2.0,0.0);
+    std::cout << " source site 0 " << tmp<<std::endl;
+    pokeSite(tmp,src,origin);
+  }
+#else
+  RealD N2 = 1.0/::sqrt(norm2(src));
+  src = src*N2;
+#endif
+
+  FermionField result(FGrid); result=Zero();
+  FermionField    ref(FGrid);    ref=Zero();
+  FermionField    tmp(FGrid);
+  FermionField    err(FGrid);
+
+  std::cout << GridLogMessage << "Drawing gauge field" << std::endl;
+  GaugeField Umu(UGrid);
+  GaugeField UmuFull(UGrid);
+  GaugeField UmuCopy(UGrid);
+  SU<Nc>::HotConfiguration(RNG4,Umu);
+  UmuCopy=Umu;
+  UmuFull=Umu;
+  std::cout << GridLogMessage << "Random gauge initialised " << std::endl;
+
+  ////////////////////////////////////
+  // Apply BCs
+  ////////////////////////////////////
+  Coordinate Block(4);
+  for(int d=0;d<4;d++)  Block[d]= Dirichlet[d+1];
+
+  std::cout << GridLogMessage << "Applying BCs for Dirichlet Block5 " << Dirichlet << std::endl;
+  std::cout << GridLogMessage << "Applying BCs for Dirichlet Block4 " << Block << std::endl;
+
+  DirichletFilter<GaugeField> Filter(Block);
+  Filter.applyFilter(Umu);
+  if(!partial) Filter.applyFilter(UmuCopy);
+  
+  ////////////////////////////////////
+  // Naive wilson implementation
+  ////////////////////////////////////
+  std::vector<ColourMatrixField> U(4,UGrid);
+  std::vector<ColourMatrixField> Ucopy(4,UGrid);
+  for(int mu=0;mu<Nd;mu++){
+    U[mu] = PeekIndex<LorentzIndex>(Umu,mu);
+    Ucopy[mu] = PeekIndex<LorentzIndex>(UmuCopy,mu);
+  }
+
+  std::cout << GridLogMessage << "Setting up Cshift based reference " << std::endl;
+
+  if (1)
+  {
+    ref = Zero();
+    for(int mu=0;mu<Nd;mu++){
+
+      tmp = Cshift(src,mu+1,1);
+      {
+	autoView( tmp_v  , tmp  , CpuWrite);
+	autoView( U_v    , U[mu]  , CpuRead);
+	autoView( Ucopy_v, Ucopy[mu]  , CpuRead);
+	for(int ss=0;ss<U[mu].Grid()->oSites();ss++){
+	  for(int s=0;s<Ls;s++){
+	    if ( (s==0) || (s==Ls-1)){
+	      tmp_v[Ls*ss+s] = Ucopy_v[ss]*tmp_v[Ls*ss+s];
+	    } else {
+	      tmp_v[Ls*ss+s] = U_v[ss]*tmp_v[Ls*ss+s];
+	    }
+	  }
+	}
+      }
+      ref=ref + tmp - Gamma(Gmu[mu])*tmp;
+      {
+	autoView( tmp_v  , tmp  , CpuWrite);
+	autoView( U_v  , U[mu]  , CpuRead);
+	autoView( Ucopy_v, Ucopy[mu]  , CpuRead);
+	autoView( src_v, src    , CpuRead);
+	for(int ss=0;ss<U[mu].Grid()->oSites();ss++){
+	  for(int s=0;s<Ls;s++){
+	    if ( (s==0) || (s==Ls-1) ){
+	      tmp_v[Ls*ss+s] = adj(Ucopy_v[ss])*src_v[Ls*ss+s];
+	    } else {
+	      tmp_v[Ls*ss+s] = adj(U_v[ss])*src_v[Ls*ss+s];
+	    }
+	  }
+	}
+      }
+      tmp =Cshift(tmp,mu+1,-1);
+      ref=ref + tmp + Gamma(Gmu[mu])*tmp;
+    }
+    ref = -0.5*ref;
+  }
+
+  RealD mass=0.1;
+  RealD M5  =1.8;
+
+  RealD NP = UGrid->_Nprocessors;
+  RealD NN = UGrid->NodeCount();
+
+  std::cout << GridLogMessage<< "*****************************************************************" <<std::endl;
+  std::cout << GridLogMessage<< "* Kernel options --dslash-generic, --dslash-unroll, --dslash-asm" <<std::endl;
+  std::cout << GridLogMessage<< "*****************************************************************" <<std::endl;
+  std::cout << GridLogMessage<< "*****************************************************************" <<std::endl;
+  std::cout << GridLogMessage<< "* Benchmarking DomainWallFermionR::Dhop                  "<<std::endl;
+  std::cout << GridLogMessage<< "* Vectorising space-time by "<<Simd::Nsimd()<<std::endl;
+  std::cout << GridLogMessage <<"* BCs for Dirichlet Block4 " << Block << std::endl;
+  std::cout << GridLogMessage <<"* Partial Dirichlet BC = " << partial << std::endl;
+  std::cout << GridLogMessage<< "* VComplex size is "<<sizeof(Simd)<< " B"<<std::endl;
+#ifdef GRID_OMP
+  if ( WilsonKernelsStatic::Comms == WilsonKernelsStatic::CommsAndCompute ) std::cout << GridLogMessage<< "* Using Overlapped Comms/Compute" <<std::endl;
+  if ( WilsonKernelsStatic::Comms == WilsonKernelsStatic::CommsThenCompute) std::cout << GridLogMessage<< "* Using sequential comms compute" <<std::endl;
+#endif
+  if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptGeneric   ) std::cout << GridLogMessage<< "* Using GENERIC Nc WilsonKernels" <<std::endl;
+  if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptHandUnroll) std::cout << GridLogMessage<< "* Using Nc=3       WilsonKernels" <<std::endl;
+  if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptInlineAsm ) std::cout << GridLogMessage<< "* Using Asm Nc=3   WilsonKernels" <<std::endl;
+  std::cout << GridLogMessage<< "*****************************************************************" <<std::endl;
+
+  FermionAction::ImplParams p;
+  p.dirichlet=Dirichlet;
+  p.partialDirichlet=partial;
+  FermionAction Dw(UmuFull,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,p);
+  
+  int ncall =1;
+  RealD n2e;
+  
+  if (1) {
+    FGrid->Barrier();
+    Dw.Dhop(src,result,0);
+    std::cout<<GridLogMessage<<"Called warmup"<<std::endl;
+    double t0=usecond();
+    for(int i=0;i<ncall;i++){
+      Dw.Dhop(src,result,0);
+    }
+    double t1=usecond();
+    FGrid->Barrier();
+
+    double volume=Ls;  for(int mu=0;mu<Nd;mu++) volume=volume*latt4[mu];
+    double flops=single_site_flops*volume*ncall;
+
+    auto nsimd = Simd::Nsimd();
+    auto simdwidth = sizeof(Simd);
+
+    // RF: Nd Wilson * Ls, Nd gauge * Ls, Nc colors
+    double data_rf = volume * ((2*Nd+1)*Nd*Nc + 2*Nd*Nc*Nc) * simdwidth / nsimd * ncall / (1024.*1024.*1024.);
+
+    // mem: Nd Wilson * Ls, Nd gauge, Nc colors
+    double data_mem = (volume * (2*Nd+1)*Nd*Nc + (volume/Ls) *2*Nd*Nc*Nc) * simdwidth / nsimd * ncall / (1024.*1024.*1024.);
+
+    std::cout<<GridLogMessage << "Called Dw "<<ncall<<" times in "<<t1-t0<<" us"<<std::endl;
+    std::cout<<GridLogMessage << "mflop/s =   "<< flops/(t1-t0)<<std::endl;
+    std::cout<<GridLogMessage << "mflop/s per rank =  "<< flops/(t1-t0)/NP<<std::endl;
+    std::cout<<GridLogMessage << "mflop/s per node =  "<< flops/(t1-t0)/NN<<std::endl;
+    err = ref-result;
+    n2e = norm2(err);
+
+    std::cout<<GridLogMessage << "norm diff   "<< n2e<< "  Line "<<__LINE__ <<std::endl;
+
+    if(( n2e>1.0e-4) ) {
+      std::cout<<GridLogMessage << "WRONG RESULT" << std::endl;
+      FGrid->Barrier();
+
+      DumpSliceNorm("s-slice ref ",ref,1);
+      DumpSliceNorm("s-slice res ",result,1);
+      DumpSliceNorm("s-slice error ",err,1);
+      exit(-1);
+    }
+    assert (n2e< 1.0e-4 );
+  }
+
+  if (1)
+  { // Naive wilson dag implementation
+
+    ref = Zero();
+    for(int mu=0;mu<Nd;mu++){
+
+      tmp = Cshift(src,mu+1,1);
+      {
+	autoView( tmp_v  , tmp  , CpuWrite);
+	autoView( U_v    , U[mu]  , CpuRead);
+	autoView( Ucopy_v, Ucopy[mu]  , CpuRead);
+	for(int ss=0;ss<U[mu].Grid()->oSites();ss++){
+	  for(int s=0;s<Ls;s++){
+	    if ( (s==0) || (s==Ls-1)){
+	      tmp_v[Ls*ss+s] = Ucopy_v[ss]*tmp_v[Ls*ss+s];
+	    } else {
+	      tmp_v[Ls*ss+s] = U_v[ss]*tmp_v[Ls*ss+s];
+	    }
+	  }
+	}
+      }
+      ref=ref + tmp + Gamma(Gmu[mu])*tmp;
+      {
+	autoView( tmp_v  , tmp  , CpuWrite);
+	autoView( U_v  , U[mu]  , CpuRead);
+	autoView( Ucopy_v, Ucopy[mu]  , CpuRead);
+	autoView( src_v, src    , CpuRead);
+	for(int ss=0;ss<U[mu].Grid()->oSites();ss++){
+	  for(int s=0;s<Ls;s++){
+	    if ( (s==0) || (s==Ls-1) ){
+	      tmp_v[Ls*ss+s] = adj(Ucopy_v[ss])*src_v[Ls*ss+s];
+	    } else {
+	      tmp_v[Ls*ss+s] = adj(U_v[ss])*src_v[Ls*ss+s];
+	    }
+	  }
+	}
+      }
+      tmp =Cshift(tmp,mu+1,-1);
+      ref=ref + tmp - Gamma(Gmu[mu])*tmp;
+    }
+    ref = -0.5*ref;
+  }
+
+  Dw.Dhop(src,result,DaggerYes);
+
+  std::cout << GridLogMessage << "----------------------------------------------------------------" << std::endl;
+  std::cout << GridLogMessage << "Compare to naive wilson implementation Dag to verify correctness" << std::endl;
+  std::cout << GridLogMessage << "----------------------------------------------------------------" << std::endl;
+
+  std::cout<<GridLogMessage << "Called DwDag"<<std::endl;
+  std::cout<<GridLogMessage << "norm dag result "<< norm2(result)<<std::endl;
+  std::cout<<GridLogMessage << "norm dag ref    "<< norm2(ref)<<std::endl;
+  err = ref-result;
+  n2e= norm2(err);
+  std::cout<<GridLogMessage << "norm dag diff   "<< n2e<< "  Line "<<__LINE__ <<std::endl;
+
+  assert((n2e)<1.0e-4);
+  
+  FermionField src_e (FrbGrid);
+  FermionField src_o (FrbGrid);
+  FermionField r_e   (FrbGrid);
+  FermionField r_o   (FrbGrid);
+  FermionField r_eo  (FGrid);
+
+  std::cout<<GridLogMessage << "Calling Deo and Doe and //assert Deo+Doe == Dunprec"<<std::endl;
+  pickCheckerboard(Even,src_e,src);
+  pickCheckerboard(Odd,src_o,src);
+
+  std::cout<<GridLogMessage << "src_e"<<norm2(src_e)<<std::endl;
+  std::cout<<GridLogMessage << "src_o"<<norm2(src_o)<<std::endl;
+
+
+  // S-direction is INNERMOST and takes no part in the parity.
+  std::cout << GridLogMessage<< "*********************************************************" <<std::endl;
+  std::cout << GridLogMessage<< "* Benchmarking DomainWallFermion::DhopEO                "<<std::endl;
+  std::cout << GridLogMessage<< "* Vectorising space-time by "<<Simd::Nsimd()<<std::endl;
+#ifdef GRID_OMP
+  if ( WilsonKernelsStatic::Comms == WilsonKernelsStatic::CommsAndCompute ) std::cout << GridLogMessage<< "* Using Overlapped Comms/Compute" <<std::endl;
+  if ( WilsonKernelsStatic::Comms == WilsonKernelsStatic::CommsThenCompute) std::cout << GridLogMessage<< "* Using sequential comms compute" <<std::endl;
+#endif
+  if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptGeneric   ) std::cout << GridLogMessage<< "* Using GENERIC Nc WilsonKernels" <<std::endl;
+  if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptHandUnroll) std::cout << GridLogMessage<< "* Using Nc=3       WilsonKernels" <<std::endl;
+  if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptInlineAsm ) std::cout << GridLogMessage<< "* Using Asm Nc=3   WilsonKernels" <<std::endl;
+  std::cout << GridLogMessage<< "*********************************************************" <<std::endl;
+  {
+    FGrid->Barrier();
+    Dw.DhopEO(src_o,r_e,DaggerNo);
+    double t0=usecond();
+    for(int i=0;i<ncall;i++){
+      Dw.DhopEO(src_o,r_e,DaggerNo);
+    }
+    double t1=usecond();
+    FGrid->Barrier();
+
+    double volume=Ls;  for(int mu=0;mu<Nd;mu++) volume=volume*latt4[mu];
+    double flops=(single_site_flops*volume*ncall)/2.0;
+
+    std::cout<<GridLogMessage << "Deo mflop/s =   "<< flops/(t1-t0)<<std::endl;
+    std::cout<<GridLogMessage << "Deo mflop/s per rank   "<< flops/(t1-t0)/NP<<std::endl;
+    std::cout<<GridLogMessage << "Deo mflop/s per node   "<< flops/(t1-t0)/NN<<std::endl;
+  }
+  Dw.DhopEO(src_o,r_e,DaggerNo);
+  Dw.DhopOE(src_e,r_o,DaggerNo);
+  Dw.Dhop  (src  ,result,DaggerNo);
+
+  std::cout<<GridLogMessage << "r_e"<<norm2(r_e)<<std::endl;
+  std::cout<<GridLogMessage << "r_o"<<norm2(r_o)<<std::endl;
+  std::cout<<GridLogMessage << "res"<<norm2(result)<<std::endl;
+
+  setCheckerboard(r_eo,r_o);
+  setCheckerboard(r_eo,r_e);
+
+  err = r_eo-result;
+  n2e= norm2(err);
+  std::cout<<GridLogMessage << "norm diff   "<< n2e<< "  Line "<<__LINE__ <<std::endl;
+  assert(n2e<1.0e-4);
+
+  pickCheckerboard(Even,src_e,err);
+  pickCheckerboard(Odd,src_o,err);
+  std::cout<<GridLogMessage << "norm diff even  "<< norm2(src_e)<<std::endl;
+  std::cout<<GridLogMessage << "norm diff odd   "<< norm2(src_o)<<std::endl;
+
+  assert(norm2(src_e)<1.0e-4);
+  assert(norm2(src_o)<1.0e-4);
+}

From fe6e8f5ac6a4e4cd94e3e3593a4ea9e0189a37d6 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 15 Nov 2022 17:00:49 -0500
Subject: [PATCH 083/146] Benchmark_comms fix

---
 benchmarks/Benchmark_comms.cc | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/benchmarks/Benchmark_comms.cc b/benchmarks/Benchmark_comms.cc
index 5ffca2b6..00526893 100644
--- a/benchmarks/Benchmark_comms.cc
+++ b/benchmarks/Benchmark_comms.cc
@@ -412,7 +412,7 @@ int main (int argc, char ** argv)
 	    }
             int tid = omp_get_thread_num();
 	    tbytes= Grid.StencilSendToRecvFrom((void *)&xbuf[dir][0], xmit_to_rank,1,
-					       (void *)&rbuf[dir][0], recv_from_rank,1, bytes,bytes,tid);
+					       (void *)&rbuf[dir][0], recv_from_rank,1, bytes,tid);
 
 	    thread_critical { dbytes+=tbytes; }
 	  }

From 6209120de9d861576eb0ce0f1ef706c445fb2a2a Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 15 Nov 2022 17:25:58 -0500
Subject: [PATCH 084/146] Fix to GPU compile attempt

---
 Grid/stencil/SimpleCompressor.h | 2 +-
 Grid/stencil/Stencil.h          | 2 ++
 2 files changed, 3 insertions(+), 1 deletion(-)

diff --git a/Grid/stencil/SimpleCompressor.h b/Grid/stencil/SimpleCompressor.h
index eda6d9e7..ffbe752c 100644
--- a/Grid/stencil/SimpleCompressor.h
+++ b/Grid/stencil/SimpleCompressor.h
@@ -73,7 +73,7 @@ public:
     auto vp0= &mm.vpointers[0][0];
     auto vp1= &mm.vpointers[1][0];
     auto type= mm.type;
-    accelerator_forNB(o,mm.buffer_size/2,vobj::Nsimd(),{
+    accelerator_forNB(o,mm.buffer_size/2,Merger::Nsimd(),{
 	decompress.Exchange(mp[2*o],mp[2*o+1],vp0[o],vp1[o],type);
     });
   }
diff --git a/Grid/stencil/Stencil.h b/Grid/stencil/Stencil.h
index dd8b646e..1e0b5028 100644
--- a/Grid/stencil/Stencil.h
+++ b/Grid/stencil/Stencil.h
@@ -245,6 +245,7 @@ public:
     Integer rbytes;
   };
   struct Merge {
+    static constexpr int Nsimd = vobj::Nsimd();
     cobj * mpointer;
     Vector<scalar_object *> rpointers;
     Vector<cobj *> vpointers;
@@ -254,6 +255,7 @@ public:
     Coordinate dims;
   };
   struct Decompress {
+    static constexpr int Nsimd = vobj::Nsimd();
     cobj * kernel_p;
     cobj * mpi_p;
     Integer buffer_size;

From ddad25211b4c3db0b922fd8ff43a22302dd5a26b Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 15 Nov 2022 17:47:52 -0500
Subject: [PATCH 085/146] Extra instantiations

---
 .../WilsonImplD2/CayleyFermion5DInstantiationWilsonImplD2.cc     | 1 +
 .../ContinuedFractionFermion5DInstantiationWilsonImplD2.cc       | 1 +
 .../DomainWallEOFAFermionInstantiationWilsonImplD2.cc            | 1 +
 .../WilsonImplD2/MobiusEOFAFermionInstantiationWilsonImplD2.cc   | 1 +
 .../PartialFractionFermion5DInstantiationWilsonImplD2.cc         | 1 +
 .../WilsonImplD2/WilsonCloverFermionInstantiationWilsonImplD2.cc | 1 +
 .../WilsonImplD2/WilsonFermion5DInstantiationWilsonImplD2.cc     | 1 +
 .../WilsonImplD2/WilsonFermionInstantiationWilsonImplD2.cc       | 1 +
 .../WilsonImplD2/WilsonKernelsInstantiationWilsonImplD2.cc       | 1 +
 .../WilsonImplD2/WilsonTMFermionInstantiationWilsonImplD2.cc     | 1 +
 Grid/qcd/action/fermion/instantiation/WilsonImplD2/impl.h        | 1 +
 .../ZWilsonImplD2/CayleyFermion5DInstantiationZWilsonImplD2.cc   | 1 +
 .../ContinuedFractionFermion5DInstantiationZWilsonImplD2.cc      | 1 +
 .../DomainWallEOFAFermionInstantiationZWilsonImplD2.cc           | 1 +
 .../ZWilsonImplD2/MobiusEOFAFermionInstantiationZWilsonImplD2.cc | 1 +
 .../PartialFractionFermion5DInstantiationZWilsonImplD2.cc        | 1 +
 .../ZWilsonImplD2/WilsonFermion5DInstantiationZWilsonImplD2.cc   | 1 +
 .../ZWilsonImplD2/WilsonKernelsInstantiationZWilsonImplD2.cc     | 1 +
 Grid/qcd/action/fermion/instantiation/ZWilsonImplD2/impl.h       | 1 +
 19 files changed, 19 insertions(+)
 create mode 120000 Grid/qcd/action/fermion/instantiation/WilsonImplD2/CayleyFermion5DInstantiationWilsonImplD2.cc
 create mode 120000 Grid/qcd/action/fermion/instantiation/WilsonImplD2/ContinuedFractionFermion5DInstantiationWilsonImplD2.cc
 create mode 120000 Grid/qcd/action/fermion/instantiation/WilsonImplD2/DomainWallEOFAFermionInstantiationWilsonImplD2.cc
 create mode 120000 Grid/qcd/action/fermion/instantiation/WilsonImplD2/MobiusEOFAFermionInstantiationWilsonImplD2.cc
 create mode 120000 Grid/qcd/action/fermion/instantiation/WilsonImplD2/PartialFractionFermion5DInstantiationWilsonImplD2.cc
 create mode 120000 Grid/qcd/action/fermion/instantiation/WilsonImplD2/WilsonCloverFermionInstantiationWilsonImplD2.cc
 create mode 120000 Grid/qcd/action/fermion/instantiation/WilsonImplD2/WilsonFermion5DInstantiationWilsonImplD2.cc
 create mode 120000 Grid/qcd/action/fermion/instantiation/WilsonImplD2/WilsonFermionInstantiationWilsonImplD2.cc
 create mode 120000 Grid/qcd/action/fermion/instantiation/WilsonImplD2/WilsonKernelsInstantiationWilsonImplD2.cc
 create mode 120000 Grid/qcd/action/fermion/instantiation/WilsonImplD2/WilsonTMFermionInstantiationWilsonImplD2.cc
 create mode 100644 Grid/qcd/action/fermion/instantiation/WilsonImplD2/impl.h
 create mode 120000 Grid/qcd/action/fermion/instantiation/ZWilsonImplD2/CayleyFermion5DInstantiationZWilsonImplD2.cc
 create mode 120000 Grid/qcd/action/fermion/instantiation/ZWilsonImplD2/ContinuedFractionFermion5DInstantiationZWilsonImplD2.cc
 create mode 120000 Grid/qcd/action/fermion/instantiation/ZWilsonImplD2/DomainWallEOFAFermionInstantiationZWilsonImplD2.cc
 create mode 120000 Grid/qcd/action/fermion/instantiation/ZWilsonImplD2/MobiusEOFAFermionInstantiationZWilsonImplD2.cc
 create mode 120000 Grid/qcd/action/fermion/instantiation/ZWilsonImplD2/PartialFractionFermion5DInstantiationZWilsonImplD2.cc
 create mode 120000 Grid/qcd/action/fermion/instantiation/ZWilsonImplD2/WilsonFermion5DInstantiationZWilsonImplD2.cc
 create mode 120000 Grid/qcd/action/fermion/instantiation/ZWilsonImplD2/WilsonKernelsInstantiationZWilsonImplD2.cc
 create mode 100644 Grid/qcd/action/fermion/instantiation/ZWilsonImplD2/impl.h

diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplD2/CayleyFermion5DInstantiationWilsonImplD2.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplD2/CayleyFermion5DInstantiationWilsonImplD2.cc
new file mode 120000
index 00000000..cb1db625
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplD2/CayleyFermion5DInstantiationWilsonImplD2.cc
@@ -0,0 +1 @@
+../CayleyFermion5DInstantiation.cc.master
\ No newline at end of file
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplD2/ContinuedFractionFermion5DInstantiationWilsonImplD2.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplD2/ContinuedFractionFermion5DInstantiationWilsonImplD2.cc
new file mode 120000
index 00000000..c2d4b8fc
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplD2/ContinuedFractionFermion5DInstantiationWilsonImplD2.cc
@@ -0,0 +1 @@
+../ContinuedFractionFermion5DInstantiation.cc.master
\ No newline at end of file
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplD2/DomainWallEOFAFermionInstantiationWilsonImplD2.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplD2/DomainWallEOFAFermionInstantiationWilsonImplD2.cc
new file mode 120000
index 00000000..2f550a2b
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplD2/DomainWallEOFAFermionInstantiationWilsonImplD2.cc
@@ -0,0 +1 @@
+../DomainWallEOFAFermionInstantiation.cc.master
\ No newline at end of file
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplD2/MobiusEOFAFermionInstantiationWilsonImplD2.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplD2/MobiusEOFAFermionInstantiationWilsonImplD2.cc
new file mode 120000
index 00000000..7a8f1172
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplD2/MobiusEOFAFermionInstantiationWilsonImplD2.cc
@@ -0,0 +1 @@
+../MobiusEOFAFermionInstantiation.cc.master
\ No newline at end of file
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplD2/PartialFractionFermion5DInstantiationWilsonImplD2.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplD2/PartialFractionFermion5DInstantiationWilsonImplD2.cc
new file mode 120000
index 00000000..7f4cea71
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplD2/PartialFractionFermion5DInstantiationWilsonImplD2.cc
@@ -0,0 +1 @@
+../PartialFractionFermion5DInstantiation.cc.master
\ No newline at end of file
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplD2/WilsonCloverFermionInstantiationWilsonImplD2.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplD2/WilsonCloverFermionInstantiationWilsonImplD2.cc
new file mode 120000
index 00000000..9cc05107
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplD2/WilsonCloverFermionInstantiationWilsonImplD2.cc
@@ -0,0 +1 @@
+../WilsonCloverFermionInstantiation.cc.master
\ No newline at end of file
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplD2/WilsonFermion5DInstantiationWilsonImplD2.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplD2/WilsonFermion5DInstantiationWilsonImplD2.cc
new file mode 120000
index 00000000..804d0884
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplD2/WilsonFermion5DInstantiationWilsonImplD2.cc
@@ -0,0 +1 @@
+../WilsonFermion5DInstantiation.cc.master
\ No newline at end of file
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplD2/WilsonFermionInstantiationWilsonImplD2.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplD2/WilsonFermionInstantiationWilsonImplD2.cc
new file mode 120000
index 00000000..5f6ab65e
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplD2/WilsonFermionInstantiationWilsonImplD2.cc
@@ -0,0 +1 @@
+../WilsonFermionInstantiation.cc.master
\ No newline at end of file
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplD2/WilsonKernelsInstantiationWilsonImplD2.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplD2/WilsonKernelsInstantiationWilsonImplD2.cc
new file mode 120000
index 00000000..01c35e7b
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplD2/WilsonKernelsInstantiationWilsonImplD2.cc
@@ -0,0 +1 @@
+../WilsonKernelsInstantiation.cc.master
\ No newline at end of file
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplD2/WilsonTMFermionInstantiationWilsonImplD2.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplD2/WilsonTMFermionInstantiationWilsonImplD2.cc
new file mode 120000
index 00000000..d5789bcf
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplD2/WilsonTMFermionInstantiationWilsonImplD2.cc
@@ -0,0 +1 @@
+../WilsonTMFermionInstantiation.cc.master
\ No newline at end of file
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplD2/impl.h b/Grid/qcd/action/fermion/instantiation/WilsonImplD2/impl.h
new file mode 100644
index 00000000..a836ff03
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplD2/impl.h
@@ -0,0 +1 @@
+#define IMPLEMENTATION WilsonImplD2
diff --git a/Grid/qcd/action/fermion/instantiation/ZWilsonImplD2/CayleyFermion5DInstantiationZWilsonImplD2.cc b/Grid/qcd/action/fermion/instantiation/ZWilsonImplD2/CayleyFermion5DInstantiationZWilsonImplD2.cc
new file mode 120000
index 00000000..cb1db625
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplD2/CayleyFermion5DInstantiationZWilsonImplD2.cc
@@ -0,0 +1 @@
+../CayleyFermion5DInstantiation.cc.master
\ No newline at end of file
diff --git a/Grid/qcd/action/fermion/instantiation/ZWilsonImplD2/ContinuedFractionFermion5DInstantiationZWilsonImplD2.cc b/Grid/qcd/action/fermion/instantiation/ZWilsonImplD2/ContinuedFractionFermion5DInstantiationZWilsonImplD2.cc
new file mode 120000
index 00000000..c2d4b8fc
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplD2/ContinuedFractionFermion5DInstantiationZWilsonImplD2.cc
@@ -0,0 +1 @@
+../ContinuedFractionFermion5DInstantiation.cc.master
\ No newline at end of file
diff --git a/Grid/qcd/action/fermion/instantiation/ZWilsonImplD2/DomainWallEOFAFermionInstantiationZWilsonImplD2.cc b/Grid/qcd/action/fermion/instantiation/ZWilsonImplD2/DomainWallEOFAFermionInstantiationZWilsonImplD2.cc
new file mode 120000
index 00000000..2f550a2b
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplD2/DomainWallEOFAFermionInstantiationZWilsonImplD2.cc
@@ -0,0 +1 @@
+../DomainWallEOFAFermionInstantiation.cc.master
\ No newline at end of file
diff --git a/Grid/qcd/action/fermion/instantiation/ZWilsonImplD2/MobiusEOFAFermionInstantiationZWilsonImplD2.cc b/Grid/qcd/action/fermion/instantiation/ZWilsonImplD2/MobiusEOFAFermionInstantiationZWilsonImplD2.cc
new file mode 120000
index 00000000..7a8f1172
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplD2/MobiusEOFAFermionInstantiationZWilsonImplD2.cc
@@ -0,0 +1 @@
+../MobiusEOFAFermionInstantiation.cc.master
\ No newline at end of file
diff --git a/Grid/qcd/action/fermion/instantiation/ZWilsonImplD2/PartialFractionFermion5DInstantiationZWilsonImplD2.cc b/Grid/qcd/action/fermion/instantiation/ZWilsonImplD2/PartialFractionFermion5DInstantiationZWilsonImplD2.cc
new file mode 120000
index 00000000..7f4cea71
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplD2/PartialFractionFermion5DInstantiationZWilsonImplD2.cc
@@ -0,0 +1 @@
+../PartialFractionFermion5DInstantiation.cc.master
\ No newline at end of file
diff --git a/Grid/qcd/action/fermion/instantiation/ZWilsonImplD2/WilsonFermion5DInstantiationZWilsonImplD2.cc b/Grid/qcd/action/fermion/instantiation/ZWilsonImplD2/WilsonFermion5DInstantiationZWilsonImplD2.cc
new file mode 120000
index 00000000..804d0884
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplD2/WilsonFermion5DInstantiationZWilsonImplD2.cc
@@ -0,0 +1 @@
+../WilsonFermion5DInstantiation.cc.master
\ No newline at end of file
diff --git a/Grid/qcd/action/fermion/instantiation/ZWilsonImplD2/WilsonKernelsInstantiationZWilsonImplD2.cc b/Grid/qcd/action/fermion/instantiation/ZWilsonImplD2/WilsonKernelsInstantiationZWilsonImplD2.cc
new file mode 120000
index 00000000..01c35e7b
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplD2/WilsonKernelsInstantiationZWilsonImplD2.cc
@@ -0,0 +1 @@
+../WilsonKernelsInstantiation.cc.master
\ No newline at end of file
diff --git a/Grid/qcd/action/fermion/instantiation/ZWilsonImplD2/impl.h b/Grid/qcd/action/fermion/instantiation/ZWilsonImplD2/impl.h
new file mode 100644
index 00000000..067d6080
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplD2/impl.h
@@ -0,0 +1 @@
+#define IMPLEMENTATION ZWilsonImplD2

From e2a938e7f78b52fd606a2879d91ced6902179645 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 15 Nov 2022 17:48:18 -0500
Subject: [PATCH 086/146] GPU happy for compile...?

---
 Grid/qcd/action/fermion/WilsonCompressor.h | 2 +-
 Grid/stencil/SimpleCompressor.h            | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/Grid/qcd/action/fermion/WilsonCompressor.h b/Grid/qcd/action/fermion/WilsonCompressor.h
index e2ced552..9ecb1c49 100644
--- a/Grid/qcd/action/fermion/WilsonCompressor.h
+++ b/Grid/qcd/action/fermion/WilsonCompressor.h
@@ -141,7 +141,7 @@ public:
     auto vp1= &mm.vpointers[1][0];
     auto type= mm.type;
     int nnum = num/Ls;
-    accelerator_forNB(o,num,vobj::Nsimd(),{
+    accelerator_forNB(o,num,Merger::Nsimd,{
 
 	int  s=o%Ls;
 	int hxyz=o/Ls; // xyzt related component
diff --git a/Grid/stencil/SimpleCompressor.h b/Grid/stencil/SimpleCompressor.h
index ffbe752c..ccbfdb29 100644
--- a/Grid/stencil/SimpleCompressor.h
+++ b/Grid/stencil/SimpleCompressor.h
@@ -73,7 +73,7 @@ public:
     auto vp0= &mm.vpointers[0][0];
     auto vp1= &mm.vpointers[1][0];
     auto type= mm.type;
-    accelerator_forNB(o,mm.buffer_size/2,Merger::Nsimd(),{
+    accelerator_forNB(o,mm.buffer_size/2,Merger::Nsimd,{
 	decompress.Exchange(mp[2*o],mp[2*o+1],vp0[o],vp1[o],type);
     });
   }

From e51eaedc568fd678132e94bb3d70938048508fea Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 15 Nov 2022 22:58:30 -0500
Subject: [PATCH 087/146] Making tests compile

---
 Grid/qcd/action/ActionParams.h                |  9 ++++++++
 .../action/scalar/ScalarInteractionAction.h   |  6 ++---
 Grid/qcd/hmc/HMC.h                            |  7 +++---
 Grid/qcd/modules/Registration.h               | 12 +++++-----
 Grid/stencil/Stencil.h                        |  2 +-
 systems/Crusher/config-command                |  3 +--
 tests/core/Test_cf_coarsen_support.cc         |  8 +++----
 tests/core/Test_checker.cc                    |  4 ++--
 tests/core/Test_contfrac_even_odd.cc          | 16 +++++++-------
 tests/core/Test_dwf_eofa_even_odd.cc          |  4 ++--
 tests/core/Test_dwf_even_odd.cc               |  4 ++--
 tests/core/Test_gpwilson_even_odd.cc          |  8 +++----
 tests/core/Test_mobius_eofa_even_odd.cc       |  4 ++--
 tests/core/Test_mobius_even_odd.cc            |  6 ++---
 tests/core/Test_staggered.cc                  | 10 ++++-----
 tests/core/Test_staggered5D.cc                | 10 ++++-----
 tests/core/Test_staggered_naive.cc            | 10 ++++-----
 tests/core/Test_wilson_clover.cc              | 20 ++++++++---------
 tests/core/Test_wilson_conserved_current.cc   | 22 +++++++++----------
 tests/core/Test_wilson_even_odd.cc            |  4 ++--
 tests/core/Test_wilson_exp_clover.cc          | 20 ++++++++---------
 .../core/Test_wilson_twisted_mass_even_odd.cc |  4 ++--
 tests/core/Test_zmobius_even_odd.cc           |  4 ++--
 tests/forces/Test_contfrac_force.cc           |  2 +-
 tests/forces/Test_dwf_force.cc                |  2 +-
 tests/forces/Test_dwf_force_eofa.cc           |  4 ++--
 tests/forces/Test_dwf_gpforce.cc              |  6 ++---
 tests/forces/Test_dwf_gpforce_eofa.cc         |  2 +-
 tests/forces/Test_gpdwf_force.cc              |  6 ++---
 tests/forces/Test_gpdwf_force_1f_2f.cc        |  4 ++--
 tests/forces/Test_gpwilson_force.cc           |  6 ++---
 tests/forces/Test_mobius_force.cc             |  2 +-
 tests/forces/Test_mobius_force_eofa.cc        |  4 ++--
 tests/forces/Test_mobius_gpforce_eofa.cc      |  2 +-
 tests/forces/Test_partfrac_force.cc           |  2 +-
 tests/forces/Test_wilson_force.cc             |  2 +-
 tests/forces/Test_wilsonclover_force.cc       |  2 +-
 tests/forces/Test_zmobius_force.cc            |  2 +-
 tests/hmc/Test_action_dwf_gparity2fvs1f.cc    |  8 +++----
 tests/hmc/Test_hmc_EODWFRatio.cc              |  9 +-------
 tests/hmc/Test_hmc_EODWFRatio_Gparity.cc      |  8 +------
 tests/hmc/Test_hmc_EOMobiusRatio.cc           |  9 +-------
 .../hmc/Test_hmc_EOMobiusRatioManyFlavour.cc  |  8 +------
 .../Test_hmc_EOWilsonCloverFermionGauge.cc    |  2 +-
 tests/hmc/Test_hmc_EOWilsonFermionGauge.cc    |  2 +-
 tests/hmc/Test_hmc_EOWilsonRatio.cc           |  2 +-
 tests/hmc/Test_hmc_GparityIwasakiGauge.cc     |  2 +-
 tests/hmc/Test_hmc_GparityWilsonGauge.cc      |  2 +-
 tests/hmc/Test_hmc_Mobius2p1f.cc              | 10 ++++-----
 tests/hmc/Test_hmc_ScalarActionNxN.cc         |  3 +--
 tests/hmc/Test_hmc_WC2ASFG_Production.cc      |  2 +-
 tests/hmc/Test_hmc_WC2SFG_Production.cc       |  2 +-
 tests/hmc/Test_hmc_WCFG_Production.cc         |  2 +-
 tests/hmc/Test_hmc_WCMixedRepFG_Production.cc |  4 ++--
 tests/hmc/Test_hmc_WCadjFG_Production.cc      |  2 +-
 .../hmc/Test_hmc_WilsonAdjointFermionGauge.cc |  2 +-
 .../hmc/Test_hmc_WilsonCloverFermionGauge.cc  |  2 +-
 tests/hmc/Test_hmc_WilsonFermionGauge.cc      |  2 +-
 tests/hmc/Test_hmc_WilsonGauge.cc             |  6 +++--
 ..._WilsonMixedRepresentationsFermionGauge.cc |  4 ++--
 tests/hmc/Test_hmc_WilsonRatio.cc             |  2 +-
 tests/hmc/Test_hmc_WilsonTMFermionGauge.cc    |  2 +-
 ...hmc_WilsonTwoIndexSymmetricFermionGauge.cc |  2 +-
 tests/hmc/Test_rhmc_EOWilson1p1.cc            |  2 +-
 tests/hmc/Test_rhmc_EOWilsonRatio.cc          |  2 +-
 tests/hmc/Test_rhmc_Wilson1p1.cc              |  2 +-
 tests/hmc/Test_rhmc_WilsonRatio.cc            |  2 +-
 67 files changed, 168 insertions(+), 184 deletions(-)

diff --git a/Grid/qcd/action/ActionParams.h b/Grid/qcd/action/ActionParams.h
index b2a06280..b332a2c3 100644
--- a/Grid/qcd/action/ActionParams.h
+++ b/Grid/qcd/action/ActionParams.h
@@ -35,6 +35,15 @@ directory
 NAMESPACE_BEGIN(Grid);
 
 // These can move into a params header and be given MacroMagic serialisation
+struct DefaultImplParams {
+  Coordinate dirichlet; // Blocksize of dirichlet BCs
+  int  partialDirichlet;
+  DefaultImplParams()  {
+    dirichlet.resize(0);
+    partialDirichlet=0;
+  };
+};
+
 struct GparityWilsonImplParams {
   Coordinate twists;
                      //mu=Nd-1 is assumed to be the time direction and a twist value of 1 indicates antiperiodic BCs
diff --git a/Grid/qcd/action/scalar/ScalarInteractionAction.h b/Grid/qcd/action/scalar/ScalarInteractionAction.h
index e04dd486..7708a489 100644
--- a/Grid/qcd/action/scalar/ScalarInteractionAction.h
+++ b/Grid/qcd/action/scalar/ScalarInteractionAction.h
@@ -47,7 +47,7 @@ private:
   const unsigned int N = Impl::Group::Dimension;
 
   typedef typename Field::vector_object vobj;
-  typedef CartesianStencil<vobj, vobj,int> Stencil;
+  typedef CartesianStencil<vobj, vobj,DefaultImplParams> Stencil;
 
   SimpleCompressor<vobj> compressor;
   int npoint = 2 * Ndim;
@@ -82,7 +82,7 @@ public:
   virtual RealD S(const Field &p)
   {
     assert(p.Grid()->Nd() == Ndim);
-    static Stencil phiStencil(p.Grid(), npoint, 0, directions, displacements,0);
+    static Stencil phiStencil(p.Grid(), npoint, 0, directions, displacements);
     phiStencil.HaloExchange(p, compressor);
     Field action(p.Grid()), pshift(p.Grid()), phisquared(p.Grid());
     phisquared = p * p;
@@ -133,7 +133,7 @@ public:
     double interm_t = usecond();
 
     // move this outside
-    static Stencil phiStencil(p.Grid(), npoint, 0, directions, displacements,0);
+    static Stencil phiStencil(p.Grid(), npoint, 0, directions, displacements);
 
     phiStencil.HaloExchange(p, compressor);
     double halo_t = usecond();
diff --git a/Grid/qcd/hmc/HMC.h b/Grid/qcd/hmc/HMC.h
index 745e53c8..d4739fb0 100644
--- a/Grid/qcd/hmc/HMC.h
+++ b/Grid/qcd/hmc/HMC.h
@@ -143,6 +143,7 @@ private:
     GridBase *Grid = U.Grid();
 
     if(Params.PerformRandomShift){
+#if 0
       //////////////////////////////////////////////////////////////////////////////////////////////////////
       // Mainly for DDHMC perform a random translation of U modulo volume
       //////////////////////////////////////////////////////////////////////////////////////////////////////
@@ -167,11 +168,11 @@ private:
 	//shift all fields together in a way that respects the gauge BCs
 	for(int mu=0; mu < Grid->Nd(); mu++)
 	  Umu[mu] = FieldImplementation::CshiftLink(Umu[mu],d,shift);
-      }
 
-      for(int mu=0;mu<Grid->Nd();mu++) PokeIndex<LorentzIndex>(U,Umu[mu],mu);
-      
+	for(int mu=0;mu<Grid->Nd();mu++) PokeIndex<LorentzIndex>(U,Umu[mu],mu);
+      }
       std::cout << GridLogMessage << "--------------------------------------------------\n";
+#endif	
     }
 
     TheIntegrator.reset_timer();
diff --git a/Grid/qcd/modules/Registration.h b/Grid/qcd/modules/Registration.h
index 28a9fdae..db43abe1 100644
--- a/Grid/qcd/modules/Registration.h
+++ b/Grid/qcd/modules/Registration.h
@@ -78,13 +78,13 @@ static Registrar<OneFlavourRatioEOFModule<FermionImplementationPolicy>,
 // Now a specific registration with a fermion field
 // here must instantiate CG and CR for every new fermion field type (macro!!)
 
-static Registrar< ConjugateGradientModule<WilsonFermionR::FermionField>,   
-                  HMC_SolverModuleFactory<solver_string, WilsonFermionR::FermionField, Serialiser> > __CGWFmodXMLInit("ConjugateGradient"); 
+static Registrar< ConjugateGradientModule<WilsonFermionD::FermionField>,   
+                  HMC_SolverModuleFactory<solver_string, WilsonFermionD::FermionField, Serialiser> > __CGWFmodXMLInit("ConjugateGradient"); 
 
-static Registrar< BiCGSTABModule<WilsonFermionR::FermionField>,   
-                  HMC_SolverModuleFactory<solver_string, WilsonFermionR::FermionField, Serialiser> > __BiCGWFmodXMLInit("BiCGSTAB"); 
-static Registrar< ConjugateResidualModule<WilsonFermionR::FermionField>,   
-                  HMC_SolverModuleFactory<solver_string, WilsonFermionR::FermionField, Serialiser> > __CRWFmodXMLInit("ConjugateResidual"); 
+static Registrar< BiCGSTABModule<WilsonFermionD::FermionField>,   
+                  HMC_SolverModuleFactory<solver_string, WilsonFermionD::FermionField, Serialiser> > __BiCGWFmodXMLInit("BiCGSTAB"); 
+static Registrar< ConjugateResidualModule<WilsonFermionD::FermionField>,   
+                  HMC_SolverModuleFactory<solver_string, WilsonFermionD::FermionField, Serialiser> > __CRWFmodXMLInit("ConjugateResidual"); 
 
 // add the staggered, scalar versions here
 
diff --git a/Grid/stencil/Stencil.h b/Grid/stencil/Stencil.h
index 1e0b5028..ff8a6433 100644
--- a/Grid/stencil/Stencil.h
+++ b/Grid/stencil/Stencil.h
@@ -698,7 +698,7 @@ public:
 		   int checkerboard,
 		   const std::vector<int> &directions,
 		   const std::vector<int> &distances,
-		   Parameters p)
+		   Parameters p=Parameters())
   {
     face_table_computed=0;
     _grid    = grid;
diff --git a/systems/Crusher/config-command b/systems/Crusher/config-command
index 57b93e03..ca1d6348 100644
--- a/systems/Crusher/config-command
+++ b/systems/Crusher/config-command
@@ -4,11 +4,10 @@
 --enable-accelerator=hip \
 --enable-gen-simd-width=64 \
 --enable-simd=GPU \
---disable-fermion-reps \
 --with-gmp=$OLCF_GMP_ROOT \
 --with-fftw=$FFTW_DIR/.. \
 --with-mpfr=/opt/cray/pe/gcc/mpfr/3.1.4/ \
---enable-gparity \
+--disable-gparity \
 CXX=hipcc MPICXX=mpicxx \
 CXXFLAGS="-fPIC -I{$ROCM_PATH}/include/ -std=c++14 -I${MPICH_DIR}/include -L/lib64 " \
  LDFLAGS="-L/lib64 -L/opt/rocm-5.2.0/lib/ -L${MPICH_DIR}/lib -lmpi -L${CRAY_MPICH_ROOTDIR}/gtl/lib -lmpi_gtl_hsa -lamdhip64 "
diff --git a/tests/core/Test_cf_coarsen_support.cc b/tests/core/Test_cf_coarsen_support.cc
index ad0309b9..0812ab7e 100644
--- a/tests/core/Test_cf_coarsen_support.cc
+++ b/tests/core/Test_cf_coarsen_support.cc
@@ -75,8 +75,8 @@ int main (int argc, char ** argv)
   RealD M5=1.8;
 
   {
-    OverlapWilsonContFracTanhFermionR Dcf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,1.0);
-    HermitianLinearOperator<OverlapWilsonContFracTanhFermionR,LatticeFermion> HermIndefOp(Dcf);
+    OverlapWilsonContFracTanhFermionD Dcf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,1.0);
+    HermitianLinearOperator<OverlapWilsonContFracTanhFermionD,LatticeFermion> HermIndefOp(Dcf);
 
     HermIndefOp.Op(src,ref);
     HermIndefOp.OpDiag(src,result);
@@ -92,8 +92,8 @@ int main (int argc, char ** argv)
   }
 
   {
-    OverlapWilsonPartialFractionTanhFermionR Dpf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,1.0);
-    HermitianLinearOperator<OverlapWilsonPartialFractionTanhFermionR,LatticeFermion> HermIndefOp(Dpf);
+    OverlapWilsonPartialFractionTanhFermionD Dpf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,1.0);
+    HermitianLinearOperator<OverlapWilsonPartialFractionTanhFermionD,LatticeFermion> HermIndefOp(Dpf);
     
     HermIndefOp.Op(src,ref);
     HermIndefOp.OpDiag(src,result);
diff --git a/tests/core/Test_checker.cc b/tests/core/Test_checker.cc
index f87133e5..c2382e91 100644
--- a/tests/core/Test_checker.cc
+++ b/tests/core/Test_checker.cc
@@ -140,14 +140,14 @@ int main (int argc, char ** argv)
   
   // RealD mass=0.1;
   // RealD M5=1.8;
-  // DomainWallFermionR Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+  // DomainWallFermionD Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
 
   // LatticeFermion    src_o(FrbGrid);
   // LatticeFermion result_o(FrbGrid);
   // pickCheckerboard(Odd,src_o,src);
   // result_o=Zero();
 
-  // SchurDiagMooeeOperator<DomainWallFermionR,LatticeFermion> HermOpEO(Ddwf);
+  // SchurDiagMooeeOperator<DomainWallFermionD,LatticeFermion> HermOpEO(Ddwf);
   // ConjugateGradient<LatticeFermion> CG(1.0e-8,10000);
   // CG(HermOpEO,src_o,result_o);
 
diff --git a/tests/core/Test_contfrac_even_odd.cc b/tests/core/Test_contfrac_even_odd.cc
index 42bfe361..5731719a 100644
--- a/tests/core/Test_contfrac_even_odd.cc
+++ b/tests/core/Test_contfrac_even_odd.cc
@@ -76,20 +76,20 @@ int main (int argc, char ** argv)
   RealD M5  =1.8;
 
   std::cout<<GridLogMessage <<"OverlapWilsonContFracTanhFermion  test"<<std::endl;
-  OverlapWilsonContFracTanhFermionR Dcf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,1.0);
-  TestWhat<OverlapWilsonContFracTanhFermionR>(Dcf,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+  OverlapWilsonContFracTanhFermionD Dcf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,1.0);
+  TestWhat<OverlapWilsonContFracTanhFermionD>(Dcf,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
 
   std::cout<<GridLogMessage <<"OverlapWilsonContFracZolotarevFermion  test"<<std::endl;
-  OverlapWilsonContFracZolotarevFermionR Dcfz(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,0.1,6.0);
-  TestWhat<OverlapWilsonContFracZolotarevFermionR>(Dcfz,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+  OverlapWilsonContFracZolotarevFermionD Dcfz(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,0.1,6.0);
+  TestWhat<OverlapWilsonContFracZolotarevFermionD>(Dcfz,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
 
   std::cout<<GridLogMessage <<"OverlapWilsonPartialFractionTanhFermion  test"<<std::endl;
-  OverlapWilsonPartialFractionTanhFermionR Dpf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,1.0);
-  TestWhat<OverlapWilsonPartialFractionTanhFermionR>(Dpf,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+  OverlapWilsonPartialFractionTanhFermionD Dpf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,1.0);
+  TestWhat<OverlapWilsonPartialFractionTanhFermionD>(Dpf,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
 
   std::cout<<GridLogMessage <<"OverlapWilsonPartialFractionZolotarevFermion  test"<<std::endl;
-  OverlapWilsonPartialFractionZolotarevFermionR Dpfz(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,0.1,6.0);
-  TestWhat<OverlapWilsonPartialFractionZolotarevFermionR>(Dpfz,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+  OverlapWilsonPartialFractionZolotarevFermionD Dpfz(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,0.1,6.0);
+  TestWhat<OverlapWilsonPartialFractionZolotarevFermionD>(Dpfz,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
 
   Grid_finalize();
 }
diff --git a/tests/core/Test_dwf_eofa_even_odd.cc b/tests/core/Test_dwf_eofa_even_odd.cc
index 7812ebb8..532c740a 100644
--- a/tests/core/Test_dwf_eofa_even_odd.cc
+++ b/tests/core/Test_dwf_eofa_even_odd.cc
@@ -90,7 +90,7 @@ int main (int argc, char ** argv)
     RealD shift = 0.1234;
     RealD M5    = 1.8;
     int   pm    = 1;
-    DomainWallEOFAFermionR Ddwf(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mq1, mq2, mq3, shift, pm, M5);
+    DomainWallEOFAFermionD Ddwf(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mq1, mq2, mq3, shift, pm, M5);
 
     LatticeFermion src_e (FrbGrid);
     LatticeFermion src_o (FrbGrid);
@@ -216,7 +216,7 @@ int main (int argc, char ** argv)
     pickCheckerboard(Even, phi_e, phi);
     pickCheckerboard(Odd , phi_o, phi);
 
-    SchurDiagMooeeOperator<DomainWallEOFAFermionR,LatticeFermion> HermOpEO(Ddwf);
+    SchurDiagMooeeOperator<DomainWallEOFAFermionD,LatticeFermion> HermOpEO(Ddwf);
     HermOpEO.MpcDagMpc(chi_e, dchi_e);
     HermOpEO.MpcDagMpc(chi_o, dchi_o);
 
diff --git a/tests/core/Test_dwf_even_odd.cc b/tests/core/Test_dwf_even_odd.cc
index 924eb3b7..f915b439 100644
--- a/tests/core/Test_dwf_even_odd.cc
+++ b/tests/core/Test_dwf_even_odd.cc
@@ -86,7 +86,7 @@ int main (int argc, char ** argv)
 
   RealD mass=0.1;
   RealD M5  =1.8;
-  DomainWallFermionR Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+  DomainWallFermionD Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
 
   LatticeFermion src_e (FrbGrid);
   LatticeFermion src_o (FrbGrid);
@@ -213,7 +213,7 @@ int main (int argc, char ** argv)
   pickCheckerboard(Even,phi_e,phi);
   pickCheckerboard(Odd ,phi_o,phi);
 
-  SchurDiagMooeeOperator<DomainWallFermionR,LatticeFermion> HermOpEO(Ddwf);
+  SchurDiagMooeeOperator<DomainWallFermionD,LatticeFermion> HermOpEO(Ddwf);
   HermOpEO.MpcDagMpc(chi_e,dchi_e);
   HermOpEO.MpcDagMpc(chi_o,dchi_o);
 
diff --git a/tests/core/Test_gpwilson_even_odd.cc b/tests/core/Test_gpwilson_even_odd.cc
index d510657e..c8587435 100644
--- a/tests/core/Test_gpwilson_even_odd.cc
+++ b/tests/core/Test_gpwilson_even_odd.cc
@@ -52,7 +52,7 @@ int main (int argc, char ** argv)
   //  pRNG.SeedFixedIntegers(seeds);
   pRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9}));
 
-  typedef typename GparityWilsonFermionR::FermionField FermionField;
+  typedef typename GparityWilsonFermionD::FermionField FermionField;
 
   FermionField src   (&Grid); random(pRNG,src);
   FermionField phi   (&Grid); random(pRNG,phi);
@@ -80,10 +80,10 @@ int main (int argc, char ** argv)
 
   RealD mass=0.1;
 
-  GparityWilsonFermionR::ImplParams params;
+  GparityWilsonFermionD::ImplParams params;
   std::vector<int> twists(Nd,0);  twists[1] = 1;
   params.twists = twists;
-  GparityWilsonFermionR Dw(Umu,Grid,RBGrid,mass,params);
+  GparityWilsonFermionD Dw(Umu,Grid,RBGrid,mass,params);
 
   FermionField src_e   (&RBGrid);
   FermionField src_o   (&RBGrid);
@@ -199,7 +199,7 @@ int main (int argc, char ** argv)
   pickCheckerboard(Even,phi_e,phi);
   pickCheckerboard(Odd ,phi_o,phi);
 
-  SchurDiagMooeeOperator<GparityWilsonFermionR,FermionField> HermOpEO(Dw);
+  SchurDiagMooeeOperator<GparityWilsonFermionD,FermionField> HermOpEO(Dw);
   HermOpEO.MpcDagMpc(chi_e,dchi_e);
   HermOpEO.MpcDagMpc(chi_o,dchi_o);
 
diff --git a/tests/core/Test_mobius_eofa_even_odd.cc b/tests/core/Test_mobius_eofa_even_odd.cc
index 68ffe624..65d55896 100644
--- a/tests/core/Test_mobius_eofa_even_odd.cc
+++ b/tests/core/Test_mobius_eofa_even_odd.cc
@@ -92,7 +92,7 @@ int main (int argc, char ** argv)
     RealD shift = 0.1234;
     RealD M5    = 1.8;
     int   pm    = 1;
-    MobiusEOFAFermionR Ddwf(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mq1, mq2, mq3, shift, pm, M5, b, c);
+    MobiusEOFAFermionD Ddwf(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mq1, mq2, mq3, shift, pm, M5, b, c);
 
     LatticeFermion src_e (FrbGrid);
     LatticeFermion src_o (FrbGrid);
@@ -218,7 +218,7 @@ int main (int argc, char ** argv)
     pickCheckerboard(Even, phi_e, phi);
     pickCheckerboard(Odd , phi_o, phi);
 
-    SchurDiagMooeeOperator<MobiusEOFAFermionR,LatticeFermion> HermOpEO(Ddwf);
+    SchurDiagMooeeOperator<MobiusEOFAFermionD,LatticeFermion> HermOpEO(Ddwf);
     HermOpEO.MpcDagMpc(chi_e, dchi_e);
     HermOpEO.MpcDagMpc(chi_o, dchi_o);
 
diff --git a/tests/core/Test_mobius_even_odd.cc b/tests/core/Test_mobius_even_odd.cc
index e210f236..91125ac6 100644
--- a/tests/core/Test_mobius_even_odd.cc
+++ b/tests/core/Test_mobius_even_odd.cc
@@ -108,8 +108,8 @@ int main (int argc, char ** argv)
   omegas.push_back( std::complex<double>(0.0686324988446592,-0.0550658530827402) );
 #endif
 
-  MobiusFermionR Ddwf(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mass, M5, 0.5,0.5);
-  //  DomainWallFermionR Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+  MobiusFermionD Ddwf(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mass, M5, 0.5,0.5);
+  //  DomainWallFermionD Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
 
   LatticeFermion src_e (FrbGrid);
   LatticeFermion src_o (FrbGrid);
@@ -264,7 +264,7 @@ int main (int argc, char ** argv)
   pickCheckerboard(Odd ,phi_o,phi);
 
 
-  SchurDiagMooeeOperator<MobiusFermionR,LatticeFermion> HermOpEO(Ddwf);
+  SchurDiagMooeeOperator<MobiusFermionD,LatticeFermion> HermOpEO(Ddwf);
   HermOpEO.MpcDagMpc(chi_e,dchi_e);
   HermOpEO.MpcDagMpc(chi_o,dchi_o);
 
diff --git a/tests/core/Test_staggered.cc b/tests/core/Test_staggered.cc
index ba615ad2..f38815ae 100644
--- a/tests/core/Test_staggered.cc
+++ b/tests/core/Test_staggered.cc
@@ -53,9 +53,9 @@ int main (int argc, char ** argv)
   pRNG.SeedFixedIntegers(seeds);
   //  pRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9});
 
-  typedef typename ImprovedStaggeredFermionR::FermionField FermionField; 
-  typedef typename ImprovedStaggeredFermionR::ComplexField ComplexField; 
-  typename ImprovedStaggeredFermionR::ImplParams params; 
+  typedef typename ImprovedStaggeredFermionD::FermionField FermionField; 
+  typedef typename ImprovedStaggeredFermionD::ComplexField ComplexField; 
+  typename ImprovedStaggeredFermionD::ImplParams params; 
 
   FermionField src   (&Grid); random(pRNG,src);
   FermionField result(&Grid); result=Zero();
@@ -130,7 +130,7 @@ int main (int argc, char ** argv)
     //    ref = ref + mass * src;
   }
 
-  ImprovedStaggeredFermionR Ds(Umu,Umu,Grid,RBGrid,mass,c1,c2,u0,params);
+  ImprovedStaggeredFermionD Ds(Umu,Umu,Grid,RBGrid,mass,c1,c2,u0,params);
   
 
   std::cout<<GridLogMessage<<"=========================================================="<<std::endl;
@@ -269,7 +269,7 @@ int main (int argc, char ** argv)
   pickCheckerboard(Even,phi_e,phi);
   pickCheckerboard(Odd ,phi_o,phi);
 
-  SchurDiagMooeeOperator<ImprovedStaggeredFermionR,FermionField> HermOpEO(Ds);
+  SchurDiagMooeeOperator<ImprovedStaggeredFermionD,FermionField> HermOpEO(Ds);
   HermOpEO.MpcDagMpc(chi_e,dchi_e);
   HermOpEO.MpcDagMpc(chi_o,dchi_o);
 
diff --git a/tests/core/Test_staggered5D.cc b/tests/core/Test_staggered5D.cc
index b1b3be1d..32ad0d17 100644
--- a/tests/core/Test_staggered5D.cc
+++ b/tests/core/Test_staggered5D.cc
@@ -60,9 +60,9 @@ int main (int argc, char ** argv)
   pRNG4.SeedFixedIntegers(seeds);
   pRNG5.SeedFixedIntegers(seeds);
 
-  typedef typename ImprovedStaggeredFermion5DR::FermionField FermionField; 
-  typedef typename ImprovedStaggeredFermion5DR::ComplexField ComplexField; 
-  typename ImprovedStaggeredFermion5DR::ImplParams params; 
+  typedef typename ImprovedStaggeredFermion5DD::FermionField FermionField; 
+  typedef typename ImprovedStaggeredFermion5DD::ComplexField ComplexField; 
+  typename ImprovedStaggeredFermion5DD::ImplParams params; 
 
   FermionField src   (FGrid);
 
@@ -148,7 +148,7 @@ int main (int argc, char ** argv)
     }
   }
 
-  ImprovedStaggeredFermion5DR Ds(Umu,Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,c1,c2,u0,params);
+  ImprovedStaggeredFermion5DD Ds(Umu,Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,c1,c2,u0,params);
 
   std::cout<<GridLogMessage<<"=========================================================="<<std::endl;
   std::cout<<GridLogMessage<<"= Testing Dhop against cshift implementation         "<<std::endl;
@@ -288,7 +288,7 @@ int main (int argc, char ** argv)
   pickCheckerboard(Even,phi_e,phi);
   pickCheckerboard(Odd ,phi_o,phi);
 
-  SchurDiagMooeeOperator<ImprovedStaggeredFermion5DR,FermionField> HermOpEO(Ds);
+  SchurDiagMooeeOperator<ImprovedStaggeredFermion5DD,FermionField> HermOpEO(Ds);
   HermOpEO.MpcDagMpc(chi_e,dchi_e);
   HermOpEO.MpcDagMpc(chi_o,dchi_o);
 
diff --git a/tests/core/Test_staggered_naive.cc b/tests/core/Test_staggered_naive.cc
index d8ca9d5f..9d32ad46 100644
--- a/tests/core/Test_staggered_naive.cc
+++ b/tests/core/Test_staggered_naive.cc
@@ -52,9 +52,9 @@ int main (int argc, char ** argv)
   pRNG.SeedFixedIntegers(seeds);
   //  pRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9});
 
-  typedef typename NaiveStaggeredFermionR::FermionField FermionField; 
-  typedef typename NaiveStaggeredFermionR::ComplexField ComplexField; 
-  typename NaiveStaggeredFermionR::ImplParams params; 
+  typedef typename NaiveStaggeredFermionD::FermionField FermionField; 
+  typedef typename NaiveStaggeredFermionD::ComplexField ComplexField; 
+  typename NaiveStaggeredFermionD::ImplParams params; 
 
   FermionField src   (&Grid); random(pRNG,src);
   FermionField result(&Grid); result=Zero();
@@ -120,7 +120,7 @@ int main (int argc, char ** argv)
     //    ref = ref + mass * src;
   }
 
-  NaiveStaggeredFermionR Ds(Umu,Grid,RBGrid,mass,c1,u0,params);
+  NaiveStaggeredFermionD Ds(Umu,Grid,RBGrid,mass,c1,u0,params);
   
 
   std::cout<<GridLogMessage<<"=========================================================="<<std::endl;
@@ -258,7 +258,7 @@ int main (int argc, char ** argv)
   pickCheckerboard(Even,phi_e,phi);
   pickCheckerboard(Odd ,phi_o,phi);
 
-  SchurDiagMooeeOperator<NaiveStaggeredFermionR,FermionField> HermOpEO(Ds);
+  SchurDiagMooeeOperator<NaiveStaggeredFermionD,FermionField> HermOpEO(Ds);
   HermOpEO.MpcDagMpc(chi_e,dchi_e);
   HermOpEO.MpcDagMpc(chi_o,dchi_o);
 
diff --git a/tests/core/Test_wilson_clover.cc b/tests/core/Test_wilson_clover.cc
index 8f143070..0ce0513f 100644
--- a/tests/core/Test_wilson_clover.cc
+++ b/tests/core/Test_wilson_clover.cc
@@ -52,8 +52,8 @@ int main(int argc, char **argv)
   pRNG.SeedFixedIntegers(seeds);
   //  pRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9});
 
-  typedef typename WilsonCloverFermionR::FermionField FermionField;
-  typename WilsonCloverFermionR::ImplParams params;
+  typedef typename WilsonCloverFermionD::FermionField FermionField;
+  typename WilsonCloverFermionD::ImplParams params;
   WilsonAnisotropyCoefficients anis;
 
   FermionField src(&Grid);
@@ -88,8 +88,8 @@ int main(int argc, char **argv)
   RealD csw_r = 1.0;
   RealD csw_t = 1.0;
 
-  WilsonCloverFermionR Dwc(Umu, Grid, RBGrid, mass, csw_r, csw_t, anis, params);
-  CompactWilsonCloverFermionR Dwc_compact(Umu, Grid, RBGrid, mass, csw_r, csw_t, 1.0, anis, params);
+  WilsonCloverFermionD Dwc(Umu, Grid, RBGrid, mass, csw_r, csw_t, anis, params);
+  CompactWilsonCloverFermionD Dwc_compact(Umu, Grid, RBGrid, mass, csw_r, csw_t, 1.0, anis, params);
 
   std::cout << GridLogMessage << "==========================================================" << std::endl;
   std::cout << GridLogMessage << "= Testing that Deo + Doe = Dunprec " << std::endl;
@@ -324,8 +324,8 @@ int main(int argc, char **argv)
   }
   /////////////////
 
-  WilsonCloverFermionR Dwc_prime(U_prime, Grid, RBGrid, mass, csw_r, csw_t, anis, params);
-  CompactWilsonCloverFermionR Dwc_compact_prime(U_prime, Grid, RBGrid, mass, csw_r, csw_t, 1.0, anis, params);
+  WilsonCloverFermionD Dwc_prime(U_prime, Grid, RBGrid, mass, csw_r, csw_t, anis, params);
+  CompactWilsonCloverFermionD Dwc_compact_prime(U_prime, Grid, RBGrid, mass, csw_r, csw_t, 1.0, anis, params);
 
   tmp = Omega * src;
   pickCheckerboard(Even, src_e, tmp);
@@ -377,14 +377,14 @@ int main(int argc, char **argv)
   chi = Zero();
   phi = Zero();
 
-  WilsonFermionR Dw(Umu, Grid, RBGrid, mass, params);
+  WilsonFermionD Dw(Umu, Grid, RBGrid, mass, params);
 
   Dw.M(src, result);
   Dwc.M(src, chi);
 
   Dwc_prime.M(Omega * src, phi);
 
-  WilsonFermionR Dw_prime(U_prime, Grid, RBGrid, mass, params);
+  WilsonFermionD Dw_prime(U_prime, Grid, RBGrid, mass, params);
   Dw_prime.M(Omega * src, result2);
 
   err = result - adj(Omega) * result2;
@@ -411,7 +411,7 @@ int main(int argc, char **argv)
   chi = Zero();
   phi = Zero();
   err = Zero();
-  WilsonCloverFermionR Dwc_csw0(Umu, Grid, RBGrid, mass, 0.0, 0.0, anis, params); //  <-- Notice: csw=0
+  WilsonCloverFermionD Dwc_csw0(Umu, Grid, RBGrid, mass, 0.0, 0.0, anis, params); //  <-- Notice: csw=0
 
   pickCheckerboard(Even, phi_e, phi);
   pickCheckerboard(Odd, phi_o, phi);
@@ -437,7 +437,7 @@ int main(int argc, char **argv)
   chi = Zero();
   phi = Zero();
   err = Zero();
-  CompactWilsonCloverFermionR Dwc_compact_csw0(Umu, Grid, RBGrid, mass, 0.0, 0.0, 1.0, anis, params); //  <-- Notice: csw=0
+  CompactWilsonCloverFermionD Dwc_compact_csw0(Umu, Grid, RBGrid, mass, 0.0, 0.0, 1.0, anis, params); //  <-- Notice: csw=0
 
   pickCheckerboard(Even, phi_e, phi);
   pickCheckerboard(Odd, phi_o, phi);
diff --git a/tests/core/Test_wilson_conserved_current.cc b/tests/core/Test_wilson_conserved_current.cc
index 3ee1a271..c66bf940 100644
--- a/tests/core/Test_wilson_conserved_current.cc
+++ b/tests/core/Test_wilson_conserved_current.cc
@@ -74,7 +74,7 @@ int main (int argc, char ** argv)
     SU<Nc>::HotConfiguration(RNG4,Umu);
   }
 
-  typename WilsonCloverFermionR::ImplParams params;
+  typename WilsonCloverFermionD::ImplParams params;
   WilsonAnisotropyCoefficients anis;
   RealD mass = 0.1;
   RealD csw_r = 1.0;
@@ -83,32 +83,32 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage <<"=================================="<<std::endl;
   std::cout<<GridLogMessage <<"WilsonFermion test"<<std::endl;
   std::cout<<GridLogMessage <<"=================================="<<std::endl;
-  WilsonFermionR Dw(Umu,*UGrid,*UrbGrid,mass,params);
-  TestConserved<WilsonFermionR>(Dw,Umu,UGrid,UrbGrid,&RNG4);
+  WilsonFermionD Dw(Umu,*UGrid,*UrbGrid,mass,params);
+  TestConserved<WilsonFermionD>(Dw,Umu,UGrid,UrbGrid,&RNG4);
 
   std::cout<<GridLogMessage <<"=================================="<<std::endl;
   std::cout<<GridLogMessage <<"WilsonCloverFermion test"<<std::endl;
   std::cout<<GridLogMessage <<"=================================="<<std::endl;
-  WilsonCloverFermionR Dwc(Umu, *UGrid, *UrbGrid, mass, csw_r, csw_t, anis, params);
-  TestConserved<WilsonCloverFermionR>(Dwc,Umu,UGrid,UrbGrid,&RNG4);
+  WilsonCloverFermionD Dwc(Umu, *UGrid, *UrbGrid, mass, csw_r, csw_t, anis, params);
+  TestConserved<WilsonCloverFermionD>(Dwc,Umu,UGrid,UrbGrid,&RNG4);
 
   std::cout<<GridLogMessage <<"=================================="<<std::endl;
   std::cout<<GridLogMessage <<"CompactWilsonCloverFermion test"<<std::endl;
   std::cout<<GridLogMessage <<"=================================="<<std::endl;
-  CompactWilsonCloverFermionR Dwcc(Umu, *UGrid, *UrbGrid, mass, csw_r, csw_t, 1.0, anis, params);
-  TestConserved<CompactWilsonCloverFermionR>(Dwcc,Umu,UGrid,UrbGrid,&RNG4);
+  CompactWilsonCloverFermionD Dwcc(Umu, *UGrid, *UrbGrid, mass, csw_r, csw_t, 1.0, anis, params);
+  TestConserved<CompactWilsonCloverFermionD>(Dwcc,Umu,UGrid,UrbGrid,&RNG4);
 
   std::cout<<GridLogMessage <<"=================================="<<std::endl;
   std::cout<<GridLogMessage <<"WilsonExpCloverFermion test"<<std::endl;
   std::cout<<GridLogMessage <<"=================================="<<std::endl;
-  WilsonExpCloverFermionR Dewc(Umu, *UGrid, *UrbGrid, mass, csw_r, csw_t, anis, params);
-  TestConserved<WilsonExpCloverFermionR>(Dewc,Umu,UGrid,UrbGrid,&RNG4);
+  WilsonExpCloverFermionD Dewc(Umu, *UGrid, *UrbGrid, mass, csw_r, csw_t, anis, params);
+  TestConserved<WilsonExpCloverFermionD>(Dewc,Umu,UGrid,UrbGrid,&RNG4);
 
   std::cout<<GridLogMessage <<"=================================="<<std::endl;
   std::cout<<GridLogMessage <<"CompactWilsonExpCloverFermion test"<<std::endl;
   std::cout<<GridLogMessage <<"=================================="<<std::endl;
-  CompactWilsonExpCloverFermionR Dewcc(Umu, *UGrid, *UrbGrid, mass, csw_r, csw_t, 1.0, anis, params);
-  TestConserved<CompactWilsonExpCloverFermionR>(Dewcc,Umu,UGrid,UrbGrid,&RNG4);
+  CompactWilsonExpCloverFermionD Dewcc(Umu, *UGrid, *UrbGrid, mass, csw_r, csw_t, 1.0, anis, params);
+  TestConserved<CompactWilsonExpCloverFermionD>(Dewcc,Umu,UGrid,UrbGrid,&RNG4);
 
   Grid_finalize();
 }
diff --git a/tests/core/Test_wilson_even_odd.cc b/tests/core/Test_wilson_even_odd.cc
index 4d240b80..81081bd0 100644
--- a/tests/core/Test_wilson_even_odd.cc
+++ b/tests/core/Test_wilson_even_odd.cc
@@ -89,7 +89,7 @@ int main (int argc, char ** argv)
 
   RealD mass=0.1;
 
-  WilsonFermionR Dw(Umu,Grid,RBGrid,mass);
+  WilsonFermionD Dw(Umu,Grid,RBGrid,mass);
 
   LatticeFermion src_e   (&RBGrid);
   LatticeFermion src_o   (&RBGrid);
@@ -205,7 +205,7 @@ int main (int argc, char ** argv)
   pickCheckerboard(Even,phi_e,phi);
   pickCheckerboard(Odd ,phi_o,phi);
 
-  SchurDiagMooeeOperator<WilsonFermionR,LatticeFermion> HermOpEO(Dw);
+  SchurDiagMooeeOperator<WilsonFermionD,LatticeFermion> HermOpEO(Dw);
   HermOpEO.MpcDagMpc(chi_e,dchi_e);
   HermOpEO.MpcDagMpc(chi_o,dchi_o);
 
diff --git a/tests/core/Test_wilson_exp_clover.cc b/tests/core/Test_wilson_exp_clover.cc
index 8516d0dc..017d8823 100644
--- a/tests/core/Test_wilson_exp_clover.cc
+++ b/tests/core/Test_wilson_exp_clover.cc
@@ -52,8 +52,8 @@ int main(int argc, char **argv)
   pRNG.SeedFixedIntegers(seeds);
   //  pRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9});
 
-  typedef typename WilsonExpCloverFermionR::FermionField FermionField;
-  typename WilsonExpCloverFermionR::ImplParams params;
+  typedef typename WilsonExpCloverFermionD::FermionField FermionField;
+  typename WilsonExpCloverFermionD::ImplParams params;
   WilsonAnisotropyCoefficients anis;
 
   FermionField src(&Grid);
@@ -88,8 +88,8 @@ int main(int argc, char **argv)
   RealD csw_r = 1.0;
   RealD csw_t = 1.0;
 
-  WilsonExpCloverFermionR Dwc(Umu, Grid, RBGrid, mass, csw_r, csw_t, anis, params);
-  CompactWilsonExpCloverFermionR Dwc_compact(Umu, Grid, RBGrid, mass, csw_r, csw_t, 1.0, anis, params);
+  WilsonExpCloverFermionD Dwc(Umu, Grid, RBGrid, mass, csw_r, csw_t, anis, params);
+  CompactWilsonExpCloverFermionD Dwc_compact(Umu, Grid, RBGrid, mass, csw_r, csw_t, 1.0, anis, params);
 
   std::cout << GridLogMessage << "==========================================================" << std::endl;
   std::cout << GridLogMessage << "= Testing that Deo + Doe = Dunprec " << std::endl;
@@ -324,8 +324,8 @@ int main(int argc, char **argv)
   }
   /////////////////
 
-  WilsonExpCloverFermionR Dwc_prime(U_prime, Grid, RBGrid, mass, csw_r, csw_t, anis, params);
-  CompactWilsonExpCloverFermionR Dwc_compact_prime(U_prime, Grid, RBGrid, mass, csw_r, csw_t, 1.0, anis, params);
+  WilsonExpCloverFermionD Dwc_prime(U_prime, Grid, RBGrid, mass, csw_r, csw_t, anis, params);
+  CompactWilsonExpCloverFermionD Dwc_compact_prime(U_prime, Grid, RBGrid, mass, csw_r, csw_t, 1.0, anis, params);
 
   tmp = Omega * src;
   pickCheckerboard(Even, src_e, tmp);
@@ -377,14 +377,14 @@ int main(int argc, char **argv)
   chi = Zero();
   phi = Zero();
 
-  WilsonFermionR Dw(Umu, Grid, RBGrid, mass, params);
+  WilsonFermionD Dw(Umu, Grid, RBGrid, mass, params);
 
   Dw.M(src, result);
   Dwc.M(src, chi);
 
   Dwc_prime.M(Omega * src, phi);
 
-  WilsonFermionR Dw_prime(U_prime, Grid, RBGrid, mass, params);
+  WilsonFermionD Dw_prime(U_prime, Grid, RBGrid, mass, params);
   Dw_prime.M(Omega * src, result2);
 
   err = result - adj(Omega) * result2;
@@ -411,7 +411,7 @@ int main(int argc, char **argv)
   chi = Zero();
   phi = Zero();
   err = Zero();
-  WilsonExpCloverFermionR Dwc_csw0(Umu, Grid, RBGrid, mass, 0.0, 0.0, anis, params); //  <-- Notice: csw=0
+  WilsonExpCloverFermionD Dwc_csw0(Umu, Grid, RBGrid, mass, 0.0, 0.0, anis, params); //  <-- Notice: csw=0
 
   pickCheckerboard(Even, phi_e, phi);
   pickCheckerboard(Odd, phi_o, phi);
@@ -437,7 +437,7 @@ int main(int argc, char **argv)
   chi = Zero();
   phi = Zero();
   err = Zero();
-  CompactWilsonExpCloverFermionR Dwc_compact_csw0(Umu, Grid, RBGrid, mass, 0.0, 0.0, 1.0, anis, params); //  <-- Notice: csw=0
+  CompactWilsonExpCloverFermionD Dwc_compact_csw0(Umu, Grid, RBGrid, mass, 0.0, 0.0, 1.0, anis, params); //  <-- Notice: csw=0
 
   pickCheckerboard(Even, phi_e, phi);
   pickCheckerboard(Odd, phi_o, phi);
diff --git a/tests/core/Test_wilson_twisted_mass_even_odd.cc b/tests/core/Test_wilson_twisted_mass_even_odd.cc
index d9e798c3..0351f7cd 100644
--- a/tests/core/Test_wilson_twisted_mass_even_odd.cc
+++ b/tests/core/Test_wilson_twisted_mass_even_odd.cc
@@ -90,7 +90,7 @@ int main (int argc, char ** argv)
 
   RealD mass=0.1;
   RealD mu  = 0.1;
-  WilsonTMFermionR Dw(Umu,Grid,RBGrid,mass,mu);
+  WilsonTMFermionD Dw(Umu,Grid,RBGrid,mass,mu);
 
   LatticeFermion src_e   (&RBGrid);
   LatticeFermion src_o   (&RBGrid);
@@ -206,7 +206,7 @@ int main (int argc, char ** argv)
   pickCheckerboard(Even,phi_e,phi);
   pickCheckerboard(Odd ,phi_o,phi);
 
-  SchurDiagMooeeOperator<WilsonTMFermionR,LatticeFermion> HermOpEO(Dw);
+  SchurDiagMooeeOperator<WilsonTMFermionD,LatticeFermion> HermOpEO(Dw);
   HermOpEO.MpcDagMpc(chi_e,dchi_e);
   HermOpEO.MpcDagMpc(chi_o,dchi_o);
 
diff --git a/tests/core/Test_zmobius_even_odd.cc b/tests/core/Test_zmobius_even_odd.cc
index f6e18934..ee6fe860 100644
--- a/tests/core/Test_zmobius_even_odd.cc
+++ b/tests/core/Test_zmobius_even_odd.cc
@@ -123,7 +123,7 @@ int main (int argc, char ** argv)
 		 RealD _mass,RealD _M5,
 		 std::vector<ComplexD> &gamma, RealD b,RealD c,const ImplParams &p= ImplParams()) : 
   */
-  ZMobiusFermionR Ddwf(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mass, M5, omegas,RealD(1.),RealD(0.));
+  ZMobiusFermionD Ddwf(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mass, M5, omegas,RealD(1.),RealD(0.));
 
   LatticeFermion src_e (FrbGrid);
   LatticeFermion src_o (FrbGrid);
@@ -278,7 +278,7 @@ int main (int argc, char ** argv)
   pickCheckerboard(Odd ,phi_o,phi);
 
 
-  SchurDiagMooeeOperator<ZMobiusFermionR,LatticeFermion> HermOpEO(Ddwf);
+  SchurDiagMooeeOperator<ZMobiusFermionD,LatticeFermion> HermOpEO(Ddwf);
   HermOpEO.MpcDagMpc(chi_e,dchi_e);
   HermOpEO.MpcDagMpc(chi_o,dchi_o);
 
diff --git a/tests/forces/Test_contfrac_force.cc b/tests/forces/Test_contfrac_force.cc
index dc9eedce..526cde12 100644
--- a/tests/forces/Test_contfrac_force.cc
+++ b/tests/forces/Test_contfrac_force.cc
@@ -66,7 +66,7 @@ int main (int argc, char ** argv)
   ////////////////////////////////////
   RealD mass=0.01; 
   RealD M5=1.8; 
-  OverlapWilsonContFracTanhFermionR Dcf(U,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,1.0);
+  OverlapWilsonContFracTanhFermionD Dcf(U,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,1.0);
   Dcf.M   (phi,Mphi);
 
   ComplexD S    = innerProduct(Mphi,Mphi); // pdag MdagM p
diff --git a/tests/forces/Test_dwf_force.cc b/tests/forces/Test_dwf_force.cc
index e7d17347..1ae28bb2 100644
--- a/tests/forces/Test_dwf_force.cc
+++ b/tests/forces/Test_dwf_force.cc
@@ -67,7 +67,7 @@ int main (int argc, char ** argv)
   ////////////////////////////////////
   RealD mass=0.01; 
   RealD M5=1.8; 
-  DomainWallFermionR Ddwf(U,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+  DomainWallFermionD Ddwf(U,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
   Ddwf.M   (phi,Mphi);
 
   ComplexD S    = innerProduct(Mphi,Mphi); // pdag MdagM p
diff --git a/tests/forces/Test_dwf_force_eofa.cc b/tests/forces/Test_dwf_force_eofa.cc
index 525178d0..d820573b 100644
--- a/tests/forces/Test_dwf_force_eofa.cc
+++ b/tests/forces/Test_dwf_force_eofa.cc
@@ -80,8 +80,8 @@ int main (int argc, char** argv)
   RealD mf = 0.01;
   RealD mb = 1.0;
   RealD M5 = 1.8;
-  DomainWallEOFAFermionR Lop(U, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mf, mf, mb, 0.0, -1, M5);
-  DomainWallEOFAFermionR Rop(U, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mb, mf, mb, -1.0, 1, M5);
+  DomainWallEOFAFermionD Lop(U, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mf, mf, mb, 0.0, -1, M5);
+  DomainWallEOFAFermionD Rop(U, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mb, mf, mb, -1.0, 1, M5);
   OneFlavourRationalParams Params(0.95, 100.0, 5000, 1.0e-12, 12);
   ConjugateGradient<LatticeFermion> CG(1.0e-12, 5000);
   ExactOneFlavourRatioPseudoFermionAction<WilsonImplR> Meofa(Lop, Rop, CG, CG, CG, CG, CG, Params, true);
diff --git a/tests/forces/Test_dwf_gpforce.cc b/tests/forces/Test_dwf_gpforce.cc
index 9db2c563..72d30369 100644
--- a/tests/forces/Test_dwf_gpforce.cc
+++ b/tests/forces/Test_dwf_gpforce.cc
@@ -47,7 +47,7 @@ int main (int argc, char ** argv)
   GridCartesian         * FGrid   = SpaceTimeGrid::makeFiveDimGrid(Ls,UGrid);
   GridRedBlackCartesian * FrbGrid = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGrid);
 
-  typedef typename GparityDomainWallFermionR::FermionField FermionField;
+  typedef typename GparityDomainWallFermionD::FermionField FermionField;
 
   int threads = GridThread::GetThreads();
   std::cout<<GridLogMessage << "Grid is setup to use "<<threads<<" threads"<<std::endl;
@@ -76,10 +76,10 @@ int main (int argc, char ** argv)
   std::vector<int> twists(Nd,0);
   twists[nu] = 1;
   twists[Nd-1] = 1; //antiperiodic in time
-  GparityDomainWallFermionR::ImplParams params;
+  GparityDomainWallFermionD::ImplParams params;
   params.twists = twists;
  
-  GparityDomainWallFermionR Dw(U,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,params);
+  GparityDomainWallFermionD Dw(U,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,params);
 
   Dw.M   (phi,Mphi);
 
diff --git a/tests/forces/Test_dwf_gpforce_eofa.cc b/tests/forces/Test_dwf_gpforce_eofa.cc
index fd47d33c..08923faa 100644
--- a/tests/forces/Test_dwf_gpforce_eofa.cc
+++ b/tests/forces/Test_dwf_gpforce_eofa.cc
@@ -33,7 +33,7 @@ using namespace std;
 using namespace Grid;
 
 typedef GparityWilsonImplR FermionImplPolicy;
-typedef GparityDomainWallEOFAFermionR FermionAction;
+typedef GparityDomainWallEOFAFermionD FermionAction;
 typedef typename FermionAction::FermionField FermionField;
 
 int main (int argc, char** argv)
diff --git a/tests/forces/Test_gpdwf_force.cc b/tests/forces/Test_gpdwf_force.cc
index af1ce82b..58dbfc47 100644
--- a/tests/forces/Test_gpdwf_force.cc
+++ b/tests/forces/Test_gpdwf_force.cc
@@ -56,7 +56,7 @@ int main (int argc, char ** argv)
   int threads = GridThread::GetThreads();
   std::cout<<GridLogMessage << "Grid is setup to use "<<threads<<" threads"<<std::endl;
 
-  typedef typename GparityDomainWallFermionR::FermionField FermionField;
+  typedef typename GparityDomainWallFermionD::FermionField FermionField;
   FermionField phi        (FGrid); gaussian(RNG5,phi);
   FermionField Mphi       (FGrid); 
   FermionField MphiPrime  (FGrid); 
@@ -75,8 +75,8 @@ int main (int argc, char ** argv)
   std::vector<int> twists(Nd,0);
   twists[nu] = 1;
   twists[3] = 1;
-  GparityDomainWallFermionR::ImplParams params;  params.twists = twists;
-  GparityDomainWallFermionR Ddwf(U,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,params);
+  GparityDomainWallFermionD::ImplParams params;  params.twists = twists;
+  GparityDomainWallFermionD Ddwf(U,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,params);
   Ddwf.M   (phi,Mphi);
 
   ComplexD S    = innerProduct(Mphi,Mphi); // pdag MdagM p
diff --git a/tests/forces/Test_gpdwf_force_1f_2f.cc b/tests/forces/Test_gpdwf_force_1f_2f.cc
index 7e14eb08..c343b7ac 100644
--- a/tests/forces/Test_gpdwf_force_1f_2f.cc
+++ b/tests/forces/Test_gpdwf_force_1f_2f.cc
@@ -70,8 +70,8 @@ void convertFermion1f_from_2f(FermionField1f &out_1f, const FermionField2f &in_2
 
   int nuoff = is_4d ? 0 : 1;   //s in 0 direction
 
-  int L_2f = FGrid_2f->FullDimensions()[nu+nuoff];
-  int L_1f = FGrid_1f->FullDimensions()[nu+nuoff];
+  Integer L_2f = FGrid_2f->FullDimensions()[nu+nuoff];
+  Integer L_1f = FGrid_1f->FullDimensions()[nu+nuoff];
   assert(L_1f == 2 * L_2f);
   
   auto in_f0_2fgrid = PeekIndex<GparityFlavourIndex>(in_2f,0); //flavor 0 on 2f Grid
diff --git a/tests/forces/Test_gpwilson_force.cc b/tests/forces/Test_gpwilson_force.cc
index 7ab2ddeb..4c3380fe 100644
--- a/tests/forces/Test_gpwilson_force.cc
+++ b/tests/forces/Test_gpwilson_force.cc
@@ -50,7 +50,7 @@ int main (int argc, char ** argv)
   int threads = GridThread::GetThreads();
   std::cout<<GridLogMessage << "Grid is setup to use "<<threads<<" threads"<<std::endl;
 
-  typedef typename GparityDomainWallFermionR::FermionField FermionField;
+  typedef typename GparityDomainWallFermionD::FermionField FermionField;
   FermionField phi        (FGrid); gaussian(RNG4,phi);
   FermionField Mphi       (FGrid); 
   FermionField MphiPrime  (FGrid); 
@@ -70,8 +70,8 @@ int main (int argc, char ** argv)
   std::vector<int> twists(Nd,0);
   twists[nu] = 1;
   twists[3]=1;
-  GparityWilsonFermionR::ImplParams params;  params.twists = twists;
-  GparityWilsonFermionR Wil(U,*UGrid,*UrbGrid,mass,params);
+  GparityWilsonFermionD::ImplParams params;  params.twists = twists;
+  GparityWilsonFermionD Wil(U,*UGrid,*UrbGrid,mass,params);
   Wil.M   (phi,Mphi);
 
   ComplexD S    = innerProduct(Mphi,Mphi); // pdag MdagM p
diff --git a/tests/forces/Test_mobius_force.cc b/tests/forces/Test_mobius_force.cc
index d2326a81..3518007c 100644
--- a/tests/forces/Test_mobius_force.cc
+++ b/tests/forces/Test_mobius_force.cc
@@ -76,7 +76,7 @@ int main (int argc, char ** argv)
   p.boundary_phases[2] = 1.0;
   p.boundary_phases[3] =- 1.0;
   
-  MobiusFermionR Ddwf(U,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,b,c,p);
+  MobiusFermionD Ddwf(U,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,b,c,p);
   Ddwf.M   (phi,Mphi);
 
   ComplexD S    = innerProduct(Mphi,Mphi); // pdag MdagM p
diff --git a/tests/forces/Test_mobius_force_eofa.cc b/tests/forces/Test_mobius_force_eofa.cc
index 1d25771a..a8871faa 100644
--- a/tests/forces/Test_mobius_force_eofa.cc
+++ b/tests/forces/Test_mobius_force_eofa.cc
@@ -82,8 +82,8 @@ int main (int argc, char** argv)
   RealD mf = 0.01;
   RealD mb = 1.0;
   RealD M5 = 1.8;
-  MobiusEOFAFermionR Lop(U, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mf, mf, mb, 0.0, -1, M5, b, c);
-  MobiusEOFAFermionR Rop(U, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mb, mf, mb, -1.0, 1, M5, b, c);
+  MobiusEOFAFermionD Lop(U, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mf, mf, mb, 0.0, -1, M5, b, c);
+  MobiusEOFAFermionD Rop(U, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mb, mf, mb, -1.0, 1, M5, b, c);
   OneFlavourRationalParams Params(0.95, 100.0, 5000, 1.0e-12, 12);
   ConjugateGradient<LatticeFermion> CG(1.0e-12, 5000);
   ExactOneFlavourRatioPseudoFermionAction<WilsonImplR> Meofa(Lop, Rop, CG, CG, CG, CG, CG, Params, false);
diff --git a/tests/forces/Test_mobius_gpforce_eofa.cc b/tests/forces/Test_mobius_gpforce_eofa.cc
index 7f114615..dd71b565 100644
--- a/tests/forces/Test_mobius_gpforce_eofa.cc
+++ b/tests/forces/Test_mobius_gpforce_eofa.cc
@@ -34,7 +34,7 @@ using namespace Grid;
  ;
 
 typedef GparityWilsonImplR FermionImplPolicy;
-typedef GparityMobiusEOFAFermionR FermionAction;
+typedef GparityMobiusEOFAFermionD FermionAction;
 typedef typename FermionAction::FermionField FermionField;
 
 int main (int argc, char** argv)
diff --git a/tests/forces/Test_partfrac_force.cc b/tests/forces/Test_partfrac_force.cc
index 33f7b5fd..173f7626 100644
--- a/tests/forces/Test_partfrac_force.cc
+++ b/tests/forces/Test_partfrac_force.cc
@@ -69,7 +69,7 @@ int main (int argc, char ** argv)
   ////////////////////////////////////
   RealD mass=0.01; 
   RealD M5=1.8; 
-  OverlapWilsonPartialFractionTanhFermionR Dpf(U,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,1.0);
+  OverlapWilsonPartialFractionTanhFermionD Dpf(U,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,1.0);
   Dpf.M   (phi,Mphi);
 
   ComplexD S    = innerProduct(Mphi,Mphi); // pdag MdagM p
diff --git a/tests/forces/Test_wilson_force.cc b/tests/forces/Test_wilson_force.cc
index b7bf1268..f4bf8ed3 100644
--- a/tests/forces/Test_wilson_force.cc
+++ b/tests/forces/Test_wilson_force.cc
@@ -67,7 +67,7 @@ int main (int argc, char ** argv)
   // Unmodified matrix element
   ////////////////////////////////////
   RealD mass=-4.0; //kills the diagonal term
-  WilsonFermionR Dw     (U,     Grid,RBGrid,mass);
+  WilsonFermionD Dw     (U,     Grid,RBGrid,mass);
   Dw.M   (phi,Mphi);
 
   ComplexD S    = innerProduct(Mphi,Mphi); // pdag MdagM p
diff --git a/tests/forces/Test_wilsonclover_force.cc b/tests/forces/Test_wilsonclover_force.cc
index 6a28e4e2..8aa5eb9d 100644
--- a/tests/forces/Test_wilsonclover_force.cc
+++ b/tests/forces/Test_wilsonclover_force.cc
@@ -70,7 +70,7 @@ int main(int argc, char **argv)
   ////////////////////////////////////
   RealD mass = 0.1;
   Real csw = 1.0;
-  WilsonCloverFermionR Dw(U, Grid, RBGrid, mass, csw, csw);
+  WilsonCloverFermionD Dw(U, Grid, RBGrid, mass, csw, csw);
   Dw.ImportGauge(U);
   Dw.M(phi, Mphi);
   ComplexD S = innerProduct(Mphi, Mphi); // Action : pdag MdagM p
diff --git a/tests/forces/Test_zmobius_force.cc b/tests/forces/Test_zmobius_force.cc
index 89673bc7..5d3a86f4 100644
--- a/tests/forces/Test_zmobius_force.cc
+++ b/tests/forces/Test_zmobius_force.cc
@@ -81,7 +81,7 @@ int main (int argc, char ** argv)
   omegas.push_back( std::complex<double>(0.0686324988446592,0.0550658530827402) );
   omegas.push_back( std::complex<double>(0.0686324988446592,-0.0550658530827402) );
 
-  ZMobiusFermionR Ddwf(U, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mass, M5, omegas,b,c);
+  ZMobiusFermionD Ddwf(U, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mass, M5, omegas,b,c);
 
   Ddwf.M   (phi,Mphi);
 
diff --git a/tests/hmc/Test_action_dwf_gparity2fvs1f.cc b/tests/hmc/Test_action_dwf_gparity2fvs1f.cc
index 830bcead..46f87d93 100644
--- a/tests/hmc/Test_action_dwf_gparity2fvs1f.cc
+++ b/tests/hmc/Test_action_dwf_gparity2fvs1f.cc
@@ -59,7 +59,7 @@ void copy2fTo1fFermionField(FermionField1f &out, const FermionField2f &in, int g
   LatticeInteger xcoor_1f(out.Grid()); //5d lattice integer
   LatticeCoordinate(xcoor_1f,gpdir);
 
-  int L = dim_2f[gpdir];
+  Integer L = dim_2f[gpdir];
 
   out = where(xcoor_1f < L, f0_fullgrid_dbl, f1_fullgrid_dbl);
 }
@@ -76,7 +76,7 @@ void copy2fTo1fGaugeField(LatticeGaugeField &out, const LatticeGaugeField &in, i
   LatticeInteger xcoor_1f(out.Grid());
   LatticeCoordinate(xcoor_1f,gpdir);
 
-  int L = dim_2f[gpdir];
+  Integer L = dim_2f[gpdir];
   
   out = where(xcoor_1f < L, U_dbl, Uconj_dbl);
 }
@@ -140,11 +140,11 @@ int main(int argc, char **argv) {
   copy2fTo1fGaugeField(Umu_1f, Umu_2f, mu);  
 
   typedef GparityWilsonImplR FermionImplPolicy2f;
-  typedef GparityDomainWallFermionR FermionAction2f;
+  typedef GparityDomainWallFermionD FermionAction2f;
   typedef typename FermionAction2f::FermionField FermionField2f;
   
   typedef WilsonImplR FermionImplPolicy1f;
-  typedef DomainWallFermionR FermionAction1f;
+  typedef DomainWallFermionD FermionAction1f;
   typedef typename FermionAction1f::FermionField FermionField1f;
 
   std::cout << "Generating eta 2f" << std::endl;
diff --git a/tests/hmc/Test_hmc_EODWFRatio.cc b/tests/hmc/Test_hmc_EODWFRatio.cc
index 93469ffe..ff8521cb 100644
--- a/tests/hmc/Test_hmc_EODWFRatio.cc
+++ b/tests/hmc/Test_hmc_EODWFRatio.cc
@@ -43,7 +43,7 @@ int main(int argc, char **argv) {
    // Typedefs to simplify notation
   typedef GenericHMCRunner<MinimumNorm2> HMCWrapper;  // Uses the default minimum norm
   typedef WilsonImplR FermionImplPolicy;
-  typedef DomainWallFermionR FermionAction;
+  typedef DomainWallFermionD FermionAction;
   typedef typename FermionAction::FermionField FermionField;
 
 
@@ -136,16 +136,9 @@ int main(int argc, char **argv) {
   TheHMC.ReadCommandLine(argc, argv); // these can be parameters from file
 
   // Reset performance counters 
-  NumOp.ZeroCounters();
-  DenOp.ZeroCounters();
   TheHMC.Run();  // no smearing
   // TheHMC.Run(SmearingPolicy); // for smearing
 
-  std::cout << GridLogMessage << "Numerator report, Pauli-Villars term         : " << std::endl;
-  NumOp.Report();
-  std::cout << GridLogMessage << "Denominator report, Dw(m) term (includes CG) : " << std::endl;
-  DenOp.Report();
-
   Grid_finalize();
 
 } // main
diff --git a/tests/hmc/Test_hmc_EODWFRatio_Gparity.cc b/tests/hmc/Test_hmc_EODWFRatio_Gparity.cc
index 9ca0b0a0..f98d0edc 100644
--- a/tests/hmc/Test_hmc_EODWFRatio_Gparity.cc
+++ b/tests/hmc/Test_hmc_EODWFRatio_Gparity.cc
@@ -42,7 +42,7 @@ int main(int argc, char **argv) {
   typedef ConjugateHMCRunner<MinimumNorm2> HMCWrapper;  // Uses the default minimum norm
 
   typedef GparityWilsonImplR FermionImplPolicy;
-  typedef GparityDomainWallFermionR FermionAction;
+  typedef GparityDomainWallFermionD FermionAction;
   typedef typename FermionAction::FermionField FermionField;
 
 
@@ -132,15 +132,9 @@ int main(int argc, char **argv) {
   TheHMC.ReadCommandLine(argc, argv); // these can be parameters from file
 
   // Reset performance counters 
-  NumOp.ZeroCounters();
-  DenOp.ZeroCounters();
   TheHMC.Run();  // no smearing
   // TheHMC.Run(SmearingPolicy); // for smearing
 
-  std::cout << GridLogMessage << "Numerator report, Pauli-Villars term         : " << std::endl;
-  NumOp.Report();
-  std::cout << GridLogMessage << "Denominator report, Dw(m) term (includes CG) : " << std::endl;
-  DenOp.Report();
 
   Grid_finalize();
 
diff --git a/tests/hmc/Test_hmc_EOMobiusRatio.cc b/tests/hmc/Test_hmc_EOMobiusRatio.cc
index 63b4d4fa..0e0a6611 100644
--- a/tests/hmc/Test_hmc_EOMobiusRatio.cc
+++ b/tests/hmc/Test_hmc_EOMobiusRatio.cc
@@ -83,7 +83,7 @@ int main(int argc, char **argv) {
    // Typedefs to simplify notation
   typedef GenericHMCRunner<MinimumNorm2> HMCWrapper;  // Uses the default minimum norm
   typedef WilsonImplR FermionImplPolicy;
-  typedef MobiusFermionR FermionAction;
+  typedef MobiusFermionD FermionAction;
   typedef typename FermionAction::FermionField FermionField;
   // Serialiser
   typedef Grid::XmlReader       Serialiser;
@@ -211,8 +211,6 @@ int main(int argc, char **argv) {
   */
 
   // Reset performance counters 
-  NumOp.ZeroCounters();
-  DenOp.ZeroCounters();
 
   if (ApplySmearing){
     SmearingParameters SmPar(Reader);
@@ -225,11 +223,6 @@ int main(int argc, char **argv) {
     TheHMC.Run();  // no smearing
   }
 
-  std::cout << GridLogMessage << "Numerator report, Pauli-Villars term         : " << std::endl;
-  NumOp.Report();
-  std::cout << GridLogMessage << "Denominator report, Dw(m) term (includes CG) : " << std::endl;
-  DenOp.Report();
-
   Grid_finalize();
 } // main
 
diff --git a/tests/hmc/Test_hmc_EOMobiusRatioManyFlavour.cc b/tests/hmc/Test_hmc_EOMobiusRatioManyFlavour.cc
index 790433f2..3f29ae62 100644
--- a/tests/hmc/Test_hmc_EOMobiusRatioManyFlavour.cc
+++ b/tests/hmc/Test_hmc_EOMobiusRatioManyFlavour.cc
@@ -89,7 +89,7 @@ int main(int argc, char **argv) {
    // Typedefs to simplify notation
   typedef GenericHMCRunner<MinimumNorm2> HMCWrapper;  // Uses the default minimum norm
   typedef WilsonImplR FermionImplPolicy;
-  typedef MobiusFermionR FermionAction;
+  typedef MobiusFermionD FermionAction;
   typedef typename FermionAction::FermionField FermionField;
   // Serialiser
   typedef Grid::XmlReader       Serialiser;
@@ -226,8 +226,6 @@ int main(int argc, char **argv) {
   */
 
   // Reset performance counters 
-  NumOp.ZeroCounters();
-  DenOp.ZeroCounters();
 
   if (ApplySmearing){
     SmearingParameters SmPar(Reader);
@@ -240,10 +238,6 @@ int main(int argc, char **argv) {
     TheHMC.Run();  // no smearing
   }
 
-  std::cout << GridLogMessage << "Numerator report, Pauli-Villars term         : " << std::endl;
-  NumOp.Report();
-  std::cout << GridLogMessage << "Denominator report, Dw(m) term (includes CG) : " << std::endl;
-  DenOp.Report();
 
   Grid_finalize();
 } // main
diff --git a/tests/hmc/Test_hmc_EOWilsonCloverFermionGauge.cc b/tests/hmc/Test_hmc_EOWilsonCloverFermionGauge.cc
index 6b9b70b5..f6485f20 100644
--- a/tests/hmc/Test_hmc_EOWilsonCloverFermionGauge.cc
+++ b/tests/hmc/Test_hmc_EOWilsonCloverFermionGauge.cc
@@ -39,7 +39,7 @@ int main(int argc, char **argv) {
    // Typedefs to simplify notation
   typedef GenericHMCRunner<MinimumNorm2> HMCWrapper;  // Uses the default minimum norm
   typedef WilsonImplR FermionImplPolicy;
-  typedef WilsonCloverFermionR FermionAction;
+  typedef WilsonCloverFermionD FermionAction;
   typedef typename FermionAction::FermionField FermionField;
 
 
diff --git a/tests/hmc/Test_hmc_EOWilsonFermionGauge.cc b/tests/hmc/Test_hmc_EOWilsonFermionGauge.cc
index 3643c0ad..092e66d1 100644
--- a/tests/hmc/Test_hmc_EOWilsonFermionGauge.cc
+++ b/tests/hmc/Test_hmc_EOWilsonFermionGauge.cc
@@ -40,7 +40,7 @@ int main(int argc, char **argv) {
    // Typedefs to simplify notation
   typedef GenericHMCRunner<MinimumNorm2> HMCWrapper;  // Uses the default minimum norm
   typedef WilsonImplR FermionImplPolicy;
-  typedef WilsonFermionR FermionAction;
+  typedef WilsonFermionD FermionAction;
   typedef typename FermionAction::FermionField FermionField;
 
 
diff --git a/tests/hmc/Test_hmc_EOWilsonRatio.cc b/tests/hmc/Test_hmc_EOWilsonRatio.cc
index 675bc605..406aa34d 100644
--- a/tests/hmc/Test_hmc_EOWilsonRatio.cc
+++ b/tests/hmc/Test_hmc_EOWilsonRatio.cc
@@ -42,7 +42,7 @@ int main(int argc, char **argv) {
    // Typedefs to simplify notation
   typedef GenericHMCRunner<MinimumNorm2> HMCWrapper;  // Uses the default minimum norm
   typedef WilsonImplR FermionImplPolicy;
-  typedef WilsonFermionR FermionAction;
+  typedef WilsonFermionD FermionAction;
   typedef typename FermionAction::FermionField FermionField;
 
 
diff --git a/tests/hmc/Test_hmc_GparityIwasakiGauge.cc b/tests/hmc/Test_hmc_GparityIwasakiGauge.cc
index d4bfa0a5..d79404a0 100644
--- a/tests/hmc/Test_hmc_GparityIwasakiGauge.cc
+++ b/tests/hmc/Test_hmc_GparityIwasakiGauge.cc
@@ -41,7 +41,7 @@ int main(int argc, char **argv) {
   typedef ConjugateHMCRunner<MinimumNorm2> HMCWrapper;  // Uses the default minimum norm
 
   typedef GparityWilsonImplR FermionImplPolicy;
-  typedef GparityDomainWallFermionR FermionAction;
+  typedef GparityDomainWallFermionD FermionAction;
   typedef typename FermionAction::FermionField FermionField;
 
 
diff --git a/tests/hmc/Test_hmc_GparityWilsonGauge.cc b/tests/hmc/Test_hmc_GparityWilsonGauge.cc
index b8c078fe..76901235 100644
--- a/tests/hmc/Test_hmc_GparityWilsonGauge.cc
+++ b/tests/hmc/Test_hmc_GparityWilsonGauge.cc
@@ -42,7 +42,7 @@ int main(int argc, char **argv) {
   typedef ConjugateHMCRunner<MinimumNorm2> HMCWrapper;  // Uses the default minimum norm
 
   typedef GparityWilsonImplR FermionImplPolicy;
-  typedef GparityDomainWallFermionR FermionAction;
+  typedef GparityDomainWallFermionD FermionAction;
   typedef typename FermionAction::FermionField FermionField;
 
 
diff --git a/tests/hmc/Test_hmc_Mobius2p1f.cc b/tests/hmc/Test_hmc_Mobius2p1f.cc
index 508f5b5e..8c97fbb5 100644
--- a/tests/hmc/Test_hmc_Mobius2p1f.cc
+++ b/tests/hmc/Test_hmc_Mobius2p1f.cc
@@ -39,7 +39,7 @@ int main(int argc, char **argv) {
 
    // Typedefs to simplify notation
   typedef WilsonImplR FermionImplPolicy;
-  typedef MobiusFermionR FermionAction;
+  typedef MobiusFermionD FermionAction;
   typedef typename FermionAction::FermionField FermionField;
 
   typedef Grid::XmlReader       Serialiser;
@@ -148,14 +148,14 @@ int main(int argc, char **argv) {
   // Level1.push_back(&StrangePseudoFermion);
   
   // DJM: setup for EOFA ratio (Shamir)
-  // DomainWallEOFAFermionR Strange_Op_L(U, *FGrid, *FrbGrid, *GridPtr, *GridRBPtr, strange_mass, strange_mass, pv_mass, 0.0, -1, M5);
-  // DomainWallEOFAFermionR Strange_Op_R(U, *FGrid, *FrbGrid, *GridPtr, *GridRBPtr, pv_mass, strange_mass, pv_mass, -1.0, 1, M5);
+  // DomainWallEOFAFermionD Strange_Op_L(U, *FGrid, *FrbGrid, *GridPtr, *GridRBPtr, strange_mass, strange_mass, pv_mass, 0.0, -1, M5);
+  // DomainWallEOFAFermionD Strange_Op_R(U, *FGrid, *FrbGrid, *GridPtr, *GridRBPtr, pv_mass, strange_mass, pv_mass, -1.0, 1, M5);
   // ExactOneFlavourRatioPseudoFermionAction<FermionImplPolicy> EOFA(Strange_Op_L, Strange_Op_R, CG, OFRp, true);
   // Level1.push_back(&EOFA);
 
   // DJM: setup for EOFA ratio (Mobius)
-  MobiusEOFAFermionR Strange_Op_L(U, *FGrid, *FrbGrid, *GridPtr, *GridRBPtr, strange_mass, strange_mass, pv_mass, 0.0, -1, M5, b, c);
-  MobiusEOFAFermionR Strange_Op_R(U, *FGrid, *FrbGrid, *GridPtr, *GridRBPtr, pv_mass, strange_mass, pv_mass, -1.0, 1, M5, b, c);
+  MobiusEOFAFermionD Strange_Op_L(U, *FGrid, *FrbGrid, *GridPtr, *GridRBPtr, strange_mass, strange_mass, pv_mass, 0.0, -1, M5, b, c);
+  MobiusEOFAFermionD Strange_Op_R(U, *FGrid, *FrbGrid, *GridPtr, *GridRBPtr, pv_mass, strange_mass, pv_mass, -1.0, 1, M5, b, c);
   ExactOneFlavourRatioPseudoFermionAction<FermionImplPolicy> EOFA(Strange_Op_L, Strange_Op_R, CG, OFRp, true);
   Level1.push_back(&EOFA);
 
diff --git a/tests/hmc/Test_hmc_ScalarActionNxN.cc b/tests/hmc/Test_hmc_ScalarActionNxN.cc
index 726ecd4a..119a39dc 100644
--- a/tests/hmc/Test_hmc_ScalarActionNxN.cc
+++ b/tests/hmc/Test_hmc_ScalarActionNxN.cc
@@ -34,7 +34,7 @@ class ScalarActionParameters : Serializable {
     double, lambda,
     double, g);
 
-  ScalarActionParameters() = default;
+  ScalarActionParameters() {};
 
     template <class ReaderClass >
   ScalarActionParameters(Reader<ReaderClass>& Reader){
@@ -45,7 +45,6 @@ class ScalarActionParameters : Serializable {
 }
 
 using namespace Grid;
- ;
 
 template <class Impl>
 class MagMeas : public HmcObservable<typename Impl::Field> {
diff --git a/tests/hmc/Test_hmc_WC2ASFG_Production.cc b/tests/hmc/Test_hmc_WC2ASFG_Production.cc
index 0bbf4ece..55c84c0a 100644
--- a/tests/hmc/Test_hmc_WC2ASFG_Production.cc
+++ b/tests/hmc/Test_hmc_WC2ASFG_Production.cc
@@ -80,7 +80,7 @@ int main(int argc, char **argv)
   // Typedefs to simplify notation
   typedef GenericHMCRunnerHirep<TheRepresentations, MinimumNorm2> HMCWrapper; // Uses the default minimum norm
   typedef WilsonTwoIndexAntiSymmetricImplR FermionImplPolicy; // gauge field implemetation for the pseudofermions
-  typedef WilsonCloverTwoIndexAntiSymmetricFermionR FermionAction; // type of lattice fermions (Wilson, DW, ...)
+  typedef WilsonCloverTwoIndexAntiSymmetricFermionD FermionAction; // type of lattice fermions (Wilson, DW, ...)
   typedef typename FermionAction::FermionField FermionField;
   //typedef Grid::JSONReader Serialiser;
   typedef Grid::XmlReader Serialiser;
diff --git a/tests/hmc/Test_hmc_WC2SFG_Production.cc b/tests/hmc/Test_hmc_WC2SFG_Production.cc
index 64a3f1cb..80f7baaf 100644
--- a/tests/hmc/Test_hmc_WC2SFG_Production.cc
+++ b/tests/hmc/Test_hmc_WC2SFG_Production.cc
@@ -81,7 +81,7 @@ int main(int argc, char **argv)
   // Typedefs to simplify notation
   typedef GenericHMCRunnerHirep<TheRepresentations, MinimumNorm2> HMCWrapper; // Uses the default minimum norm
   typedef WilsonTwoIndexSymmetricImplR FermionImplPolicy; // gauge field implemetation for the pseudofermions
-  typedef WilsonCloverTwoIndexSymmetricFermionR FermionAction; // type of lattice fermions (Wilson, DW, ...)
+  typedef WilsonCloverTwoIndexSymmetricFermionD FermionAction; // type of lattice fermions (Wilson, DW, ...)
   typedef typename FermionAction::FermionField FermionField;
   //typedef Grid::JSONReader Serialiser;
   typedef Grid::XmlReader Serialiser;
diff --git a/tests/hmc/Test_hmc_WCFG_Production.cc b/tests/hmc/Test_hmc_WCFG_Production.cc
index cebe3791..bd02886d 100644
--- a/tests/hmc/Test_hmc_WCFG_Production.cc
+++ b/tests/hmc/Test_hmc_WCFG_Production.cc
@@ -79,7 +79,7 @@ int main(int argc, char **argv)
   // Typedefs to simplify notation
   typedef GenericHMCRunner<MinimumNorm2> HMCWrapper; // Uses the default minimum norm
   typedef WilsonImplR FermionImplPolicy;
-  typedef WilsonCloverFermionR FermionAction;
+  typedef WilsonCloverFermionD FermionAction;
   typedef typename FermionAction::FermionField FermionField;
   typedef Grid::XmlReader Serialiser;
 
diff --git a/tests/hmc/Test_hmc_WCMixedRepFG_Production.cc b/tests/hmc/Test_hmc_WCMixedRepFG_Production.cc
index 211900be..ac017860 100644
--- a/tests/hmc/Test_hmc_WCMixedRepFG_Production.cc
+++ b/tests/hmc/Test_hmc_WCMixedRepFG_Production.cc
@@ -84,11 +84,11 @@ int main(int argc, char **argv) {
   typedef GenericHMCRunnerHirep<TheRepresentations, MinimumNorm2> HMCWrapper;
 
   typedef WilsonImplR FundImplPolicy;
-  typedef WilsonCloverFermionR FundFermionAction; 
+  typedef WilsonCloverFermionD FundFermionAction; 
   typedef typename FundFermionAction::FermionField FundFermionField;
 
   typedef WilsonTwoIndexAntiSymmetricImplR ASymmImplPolicy; 
-  typedef WilsonCloverTwoIndexAntiSymmetricFermionR ASymmFermionAction; 
+  typedef WilsonCloverTwoIndexAntiSymmetricFermionD ASymmFermionAction; 
   typedef typename ASymmFermionAction::FermionField ASymmFermionField;
 
   typedef Grid::XmlReader Serialiser;
diff --git a/tests/hmc/Test_hmc_WCadjFG_Production.cc b/tests/hmc/Test_hmc_WCadjFG_Production.cc
index 5cf4bac1..a9673e6c 100644
--- a/tests/hmc/Test_hmc_WCadjFG_Production.cc
+++ b/tests/hmc/Test_hmc_WCadjFG_Production.cc
@@ -81,7 +81,7 @@ int main(int argc, char **argv)
   // Typedefs to simplify notation
   typedef GenericHMCRunnerHirep<TheRepresentations, MinimumNorm2> HMCWrapper; // Uses the default minimum norm
   typedef WilsonAdjImplR FermionImplPolicy; // gauge field implemetation for the pseudofermions
-  typedef WilsonCloverAdjFermionR FermionAction; // type of lattice fermions (Wilson, DW, ...)
+  typedef WilsonCloverAdjFermionD FermionAction; // type of lattice fermions (Wilson, DW, ...)
   typedef typename FermionAction::FermionField FermionField;
   typedef Grid::XmlReader Serialiser;
 
diff --git a/tests/hmc/Test_hmc_WilsonAdjointFermionGauge.cc b/tests/hmc/Test_hmc_WilsonAdjointFermionGauge.cc
index cc56cae3..774c9037 100644
--- a/tests/hmc/Test_hmc_WilsonAdjointFermionGauge.cc
+++ b/tests/hmc/Test_hmc_WilsonAdjointFermionGauge.cc
@@ -46,7 +46,7 @@ int main(int argc, char **argv) {
    // Typedefs to simplify notation
   typedef GenericHMCRunnerHirep<TheRepresentations, MinimumNorm2> HMCWrapper;
   typedef WilsonAdjImplR FermionImplPolicy; // gauge field implemetation for the pseudofermions
-  typedef WilsonAdjFermionR FermionAction; // type of lattice fermions (Wilson, DW, ...)
+  typedef WilsonAdjFermionD FermionAction; // type of lattice fermions (Wilson, DW, ...)
   typedef typename FermionAction::FermionField FermionField;
 
   //::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
diff --git a/tests/hmc/Test_hmc_WilsonCloverFermionGauge.cc b/tests/hmc/Test_hmc_WilsonCloverFermionGauge.cc
index 9e27e3ec..149e6c5c 100644
--- a/tests/hmc/Test_hmc_WilsonCloverFermionGauge.cc
+++ b/tests/hmc/Test_hmc_WilsonCloverFermionGauge.cc
@@ -41,7 +41,7 @@ int main(int argc, char **argv)
   // Typedefs to simplify notation
   typedef GenericHMCRunner<MinimumNorm2> HMCWrapper; // Uses the default minimum norm
   typedef WilsonImplR FermionImplPolicy;
-  typedef WilsonCloverFermionR FermionAction;
+  typedef WilsonCloverFermionD FermionAction;
   typedef typename FermionAction::FermionField FermionField;
 
   //::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
diff --git a/tests/hmc/Test_hmc_WilsonFermionGauge.cc b/tests/hmc/Test_hmc_WilsonFermionGauge.cc
index cc1f2474..a0c43c51 100644
--- a/tests/hmc/Test_hmc_WilsonFermionGauge.cc
+++ b/tests/hmc/Test_hmc_WilsonFermionGauge.cc
@@ -42,7 +42,7 @@ int main(int argc, char **argv) {
    // Typedefs to simplify notation
   typedef GenericHMCRunner<MinimumNorm2> HMCWrapper;  // Uses the default minimum norm
   typedef WilsonImplR FermionImplPolicy;
-  typedef WilsonFermionR FermionAction;
+  typedef WilsonFermionD FermionAction;
   typedef typename FermionAction::FermionField FermionField;
 
 
diff --git a/tests/hmc/Test_hmc_WilsonGauge.cc b/tests/hmc/Test_hmc_WilsonGauge.cc
index 4ef0e658..28feadf3 100644
--- a/tests/hmc/Test_hmc_WilsonGauge.cc
+++ b/tests/hmc/Test_hmc_WilsonGauge.cc
@@ -69,8 +69,10 @@ int main(int argc, char **argv)
   TopologyObsParameters TopParams;
   TopParams.interval = 5;
   TopParams.do_smearing = true;
-  TopParams.Smearing.steps = 200;
-  TopParams.Smearing.step_size = 0.01;
+  TopParams.Smearing.init_step_size = 0.01;
+  TopParams.Smearing.tolerance = 1e-5;
+  //  TopParams.Smearing.steps = 200;
+  //  TopParams.Smearing.step_size = 0.01;
   TopParams.Smearing.meas_interval = 50;
   TopParams.Smearing.maxTau = 2.0; 
   TheHMC.Resources.AddObservable<QObs>(TopParams);
diff --git a/tests/hmc/Test_hmc_WilsonMixedRepresentationsFermionGauge.cc b/tests/hmc/Test_hmc_WilsonMixedRepresentationsFermionGauge.cc
index 3b8cdda6..60d760a6 100644
--- a/tests/hmc/Test_hmc_WilsonMixedRepresentationsFermionGauge.cc
+++ b/tests/hmc/Test_hmc_WilsonMixedRepresentationsFermionGauge.cc
@@ -51,9 +51,9 @@ int main(int argc, char **argv) {
   typedef GenericHMCRunnerHirep<TheRepresentations, MinimumNorm2> HMCWrapper;
 
   typedef WilsonAdjImplR AdjImplPolicy; // gauge field implemetation for the pseudofermions
-  typedef WilsonAdjFermionR AdjFermionAction; // type of lattice fermions (Wilson, DW, ...)
+  typedef WilsonAdjFermionD AdjFermionAction; // type of lattice fermions (Wilson, DW, ...)
   typedef WilsonTwoIndexSymmetricImplR SymmImplPolicy; 
-  typedef WilsonTwoIndexSymmetricFermionR SymmFermionAction; 
+  typedef WilsonTwoIndexSymmetricFermionD SymmFermionAction; 
 
 
   typedef typename AdjFermionAction::FermionField AdjFermionField;
diff --git a/tests/hmc/Test_hmc_WilsonRatio.cc b/tests/hmc/Test_hmc_WilsonRatio.cc
index 3e3cac7e..e134dd83 100644
--- a/tests/hmc/Test_hmc_WilsonRatio.cc
+++ b/tests/hmc/Test_hmc_WilsonRatio.cc
@@ -41,7 +41,7 @@ int main(int argc, char **argv) {
    // Typedefs to simplify notation
   typedef GenericHMCRunner<MinimumNorm2> HMCWrapper;  // Uses the default minimum norm
   typedef WilsonImplR FermionImplPolicy;
-  typedef WilsonFermionR FermionAction;
+  typedef WilsonFermionD FermionAction;
   typedef typename FermionAction::FermionField FermionField;
 
 
diff --git a/tests/hmc/Test_hmc_WilsonTMFermionGauge.cc b/tests/hmc/Test_hmc_WilsonTMFermionGauge.cc
index 5928efbe..9916580e 100644
--- a/tests/hmc/Test_hmc_WilsonTMFermionGauge.cc
+++ b/tests/hmc/Test_hmc_WilsonTMFermionGauge.cc
@@ -42,7 +42,7 @@ int main(int argc, char **argv) {
    // Typedefs to simplify notation
   typedef GenericHMCRunner<MinimumNorm2> HMCWrapper;  // Uses the default minimum norm
   typedef WilsonImplR FermionImplPolicy;
-  typedef WilsonTMFermionR FermionAction;
+  typedef WilsonTMFermionD FermionAction;
   typedef typename FermionAction::FermionField FermionField;
 
 
diff --git a/tests/hmc/Test_hmc_WilsonTwoIndexSymmetricFermionGauge.cc b/tests/hmc/Test_hmc_WilsonTwoIndexSymmetricFermionGauge.cc
index 387842f7..bdcd27c5 100644
--- a/tests/hmc/Test_hmc_WilsonTwoIndexSymmetricFermionGauge.cc
+++ b/tests/hmc/Test_hmc_WilsonTwoIndexSymmetricFermionGauge.cc
@@ -45,7 +45,7 @@ int main(int argc, char **argv) {
   typedef GenericHMCRunnerHirep<TheRepresentations, MinimumNorm2> HMCWrapper;
 
   typedef WilsonTwoIndexSymmetricImplR FermionImplPolicy; // gauge field implemetation for the pseudofermions
-  typedef WilsonTwoIndexSymmetricFermionR FermionAction; // type of lattice fermions (Wilson, DW, ...)
+  typedef WilsonTwoIndexSymmetricFermionD FermionAction; // type of lattice fermions (Wilson, DW, ...)
   typedef typename FermionAction::FermionField FermionField;
 
   //::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
diff --git a/tests/hmc/Test_rhmc_EOWilson1p1.cc b/tests/hmc/Test_rhmc_EOWilson1p1.cc
index 51a966b1..1e0975ca 100644
--- a/tests/hmc/Test_rhmc_EOWilson1p1.cc
+++ b/tests/hmc/Test_rhmc_EOWilson1p1.cc
@@ -41,7 +41,7 @@ int main(int argc, char **argv) {
    // Typedefs to simplify notation
   typedef GenericHMCRunner<MinimumNorm2> HMCWrapper;  // Uses the default minimum norm
   typedef WilsonImplR FermionImplPolicy;
-  typedef WilsonFermionR FermionAction;
+  typedef WilsonFermionD FermionAction;
   typedef typename FermionAction::FermionField FermionField;
 
 
diff --git a/tests/hmc/Test_rhmc_EOWilsonRatio.cc b/tests/hmc/Test_rhmc_EOWilsonRatio.cc
index 44fb6d47..06d54215 100644
--- a/tests/hmc/Test_rhmc_EOWilsonRatio.cc
+++ b/tests/hmc/Test_rhmc_EOWilsonRatio.cc
@@ -42,7 +42,7 @@ int main(int argc, char **argv) {
    // Typedefs to simplify notation
   typedef GenericHMCRunner<MinimumNorm2> HMCWrapper;  // Uses the default minimum norm
   typedef WilsonImplR FermionImplPolicy;
-  typedef WilsonFermionR FermionAction;
+  typedef WilsonFermionD FermionAction;
   typedef typename FermionAction::FermionField FermionField;
 
 
diff --git a/tests/hmc/Test_rhmc_Wilson1p1.cc b/tests/hmc/Test_rhmc_Wilson1p1.cc
index 93b748d2..2935092c 100644
--- a/tests/hmc/Test_rhmc_Wilson1p1.cc
+++ b/tests/hmc/Test_rhmc_Wilson1p1.cc
@@ -42,7 +42,7 @@ int main(int argc, char **argv) {
    // Typedefs to simplify notation
   typedef GenericHMCRunner<MinimumNorm2> HMCWrapper;  // Uses the default minimum norm
   typedef WilsonImplR FermionImplPolicy;
-  typedef WilsonFermionR FermionAction;
+  typedef WilsonFermionD FermionAction;
   typedef typename FermionAction::FermionField FermionField;
 
 
diff --git a/tests/hmc/Test_rhmc_WilsonRatio.cc b/tests/hmc/Test_rhmc_WilsonRatio.cc
index 4896d329..60bb7641 100644
--- a/tests/hmc/Test_rhmc_WilsonRatio.cc
+++ b/tests/hmc/Test_rhmc_WilsonRatio.cc
@@ -41,7 +41,7 @@ int main(int argc, char **argv) {
    // Typedefs to simplify notation
   typedef GenericHMCRunner<MinimumNorm2> HMCWrapper;  // Uses the default minimum norm
   typedef WilsonImplR FermionImplPolicy;
-  typedef WilsonFermionR FermionAction;
+  typedef WilsonFermionD FermionAction;
   typedef typename FermionAction::FermionField FermionField;
 
 

From 3dbfce522320128e26fcc3fc7223e33fe221d964 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Wed, 16 Nov 2022 20:15:51 -0500
Subject: [PATCH 088/146] Tests clean build on HIP

---
 Grid/algorithms/CoarsenedMatrix.h             | 22 ++---
 Grid/algorithms/LinearOperator.h              |  1 +
 Grid/qcd/action/ActionParams.h                |  9 --
 Grid/stencil/Stencil.h                        | 10 ++
 Grid/tensors/Tensor_traits.h                  |  2 +-
 systems/Crusher/config-command                |  4 +-
 systems/Crusher/sourceme.sh                   |  2 +
 tests/debug/Test_cayley_cg.cc                 | 48 ++++-----
 tests/debug/Test_cayley_coarsen_support.cc    |  6 +-
 tests/debug/Test_cayley_even_odd.cc           | 32 +++---
 tests/debug/Test_cayley_ldop_cr.cc            |  6 +-
 tests/debug/Test_cayley_mres.cc               | 28 +++---
 tests/debug/Test_heatbath_dwf_eofa.cc         |  4 +-
 tests/debug/Test_heatbath_dwf_eofa_gparity.cc |  4 +-
 tests/debug/Test_heatbath_mobius_eofa.cc      |  4 +-
 .../Test_heatbath_mobius_eofa_gparity.cc      |  2 +-
 tests/debug/Test_reweight_dwf_eofa.cc         | 16 +--
 tests/debug/Test_reweight_dwf_eofa_gparity.cc | 20 ++--
 tests/debug/Test_reweight_mobius_eofa.cc      | 16 +--
 .../Test_reweight_mobius_eofa_gparity.cc      | 20 ++--
 tests/hmc/Test_hmc_WC2ASFG_Production.cc      |  6 +-
 tests/hmc/Test_hmc_WC2SFG_Production.cc       |  4 +
 tests/hmc/Test_hmc_WCMixedRepFG_Production.cc |  4 +
 tests/hmc/Test_hmc_WCadjFG_Production.cc      |  4 +
 .../hmc/Test_hmc_WilsonAdjointFermionGauge.cc |  6 +-
 ..._WilsonMixedRepresentationsFermionGauge.cc |  4 +
 ...hmc_WilsonTwoIndexSymmetricFermionGauge.cc |  4 +
 tests/lanczos/Test_WCMultiRep_lanczos.cc      | 13 ++-
 tests/lanczos/Test_compressed_lanczos.cc      |  4 +-
 .../Test_dwf_compressed_lanczos_reorg.cc      |  4 +-
 ..._dwf_compressed_lanczos_reorg_synthetic.cc |  4 +-
 tests/lanczos/Test_dwf_lanczos.cc             | 16 +--
 tests/lanczos/Test_wilson_lanczos.cc          |  4 +-
 tests/smearing/Test_WilsonFlow.cc             |  9 +-
 tests/solver/Test_cf_cr_unprec.cc             |  6 +-
 tests/solver/Test_coarse_even_odd.cc          |  4 +-
 tests/solver/Test_contfrac_cg.cc              | 16 +--
 tests/solver/Test_dwf_cg_prec.cc              |  7 +-
 tests/solver/Test_dwf_cg_schur.cc             |  2 +-
 tests/solver/Test_dwf_cg_unprec.cc            |  4 +-
 tests/solver/Test_dwf_cr_unprec.cc            |  6 +-
 tests/solver/Test_dwf_fpgcr.cc                |  6 +-
 tests/solver/Test_dwf_hdcr.cc                 | 14 +--
 tests/solver/Test_dwf_hdcr_16_rb.cc           | 14 +--
 tests/solver/Test_dwf_hdcr_24_regression.cc   | 18 ++--
 tests/solver/Test_dwf_hdcr_2level.cc          | 30 +++---
 tests/solver/Test_dwf_hdcr_48_rb.cc           | 14 +--
 tests/solver/Test_dwf_hdcr_48_regression.cc   | 18 ++--
 tests/solver/Test_dwf_mrhs_cg.cc              | 14 +--
 tests/solver/Test_dwf_mrhs_cg_mpi.cc          | 14 +--
 tests/solver/Test_dwf_mrhs_cg_mpieo.cc        | 14 +--
 tests/solver/Test_dwf_multigrid.cc            | 20 ++--
 tests/solver/Test_dwf_qmr_unprec.cc           |  8 +-
 tests/solver/Test_eofa_inv.cc                 |  4 +-
 tests/solver/Test_hw_multigrid.cc             | 10 +-
 tests/solver/Test_hw_multigrid_mixed_48.cc    | 98 +++++++++----------
 tests/solver/Test_hw_multigrid_mixed_48_rb.cc | 96 +++++++++---------
 tests/solver/Test_mobius_bcg.cc               | 14 +--
 tests/solver/Test_mobius_bcg_nosplit.cc       | 10 +-
 tests/solver/Test_mobius_bcg_phys_nosplit.cc  | 10 +-
 tests/solver/Test_mobius_bcg_prec_nosplit.cc  | 10 +-
 tests/solver/Test_split_grid.cc               | 14 +--
 tests/solver/Test_staggered_block_cg_prec.cc  | 18 ++--
 .../solver/Test_staggered_block_cg_unprec.cc  | 26 ++---
 tests/solver/Test_staggered_cagmres_unprec.cc | 10 +-
 tests/solver/Test_staggered_cg_prec.cc        |  8 +-
 tests/solver/Test_staggered_cg_schur.cc       |  6 +-
 tests/solver/Test_staggered_cg_unprec.cc      | 10 +-
 tests/solver/Test_staggered_fcagmres_prec.cc  | 10 +-
 tests/solver/Test_staggered_fgmres_prec.cc    | 10 +-
 tests/solver/Test_staggered_gmres_unprec.cc   | 10 +-
 tests/solver/Test_staggered_mr_unprec.cc      | 10 +-
 tests/solver/Test_staggered_multishift.cc     |  8 +-
 tests/solver/Test_wilson_cagmres_unprec.cc    |  4 +-
 tests/solver/Test_wilson_cg_prec.cc           |  4 +-
 tests/solver/Test_wilson_cg_schur.cc          |  2 +-
 tests/solver/Test_wilson_cg_unprec.cc         |  4 +-
 tests/solver/Test_wilson_cr_unprec.cc         |  4 +-
 tests/solver/Test_wilson_fcagmres_prec.cc     |  4 +-
 tests/solver/Test_wilson_fgmres_prec.cc       |  4 +-
 tests/solver/Test_wilson_gmres_unprec.cc      |  4 +-
 tests/solver/Test_wilson_mg.cc                |  6 +-
 tests/solver/Test_wilson_mr_unprec.cc         |  4 +-
 tests/solver/Test_wilson_qmr_unprec.cc        |  4 +-
 .../solver/Test_wilsonclover_bicgstab_prec.cc |  4 +-
 .../Test_wilsonclover_bicgstab_schur.cc       |  2 +-
 .../Test_wilsonclover_bicgstab_unprec.cc      |  4 +-
 .../Test_wilsonclover_cagmres_unprec.cc       |  8 +-
 tests/solver/Test_wilsonclover_cg_prec.cc     | 16 +--
 tests/solver/Test_wilsonclover_cg_schur.cc    |  8 +-
 tests/solver/Test_wilsonclover_cg_unprec.cc   | 16 +--
 .../solver/Test_wilsonclover_fcagmres_prec.cc |  8 +-
 tests/solver/Test_wilsonclover_fgmres_prec.cc |  8 +-
 .../solver/Test_wilsonclover_gmres_unprec.cc  |  8 +-
 tests/solver/Test_wilsonclover_mg.cc          |  6 +-
 tests/solver/Test_wilsonclover_mr_unprec.cc   |  8 +-
 tests/solver/Test_zmobius_cg_prec.cc          |  5 +-
 97 files changed, 564 insertions(+), 544 deletions(-)

diff --git a/Grid/algorithms/CoarsenedMatrix.h b/Grid/algorithms/CoarsenedMatrix.h
index ba4abecd..7008008c 100644
--- a/Grid/algorithms/CoarsenedMatrix.h
+++ b/Grid/algorithms/CoarsenedMatrix.h
@@ -324,9 +324,9 @@ public:
   GridBase*        _cbgrid;
   int hermitian;
 
-  CartesianStencil<siteVector,siteVector,int> Stencil; 
-  CartesianStencil<siteVector,siteVector,int> StencilEven;
-  CartesianStencil<siteVector,siteVector,int> StencilOdd;
+  CartesianStencil<siteVector,siteVector,DefaultImplParams> Stencil; 
+  CartesianStencil<siteVector,siteVector,DefaultImplParams> StencilEven;
+  CartesianStencil<siteVector,siteVector,DefaultImplParams> StencilOdd;
 
   std::vector<CoarseMatrix> A;
   std::vector<CoarseMatrix> Aeven;
@@ -631,7 +631,7 @@ public:
     assert(Aself != nullptr);
   }
 
-  void DselfInternal(CartesianStencil<siteVector,siteVector,int> &st, CoarseMatrix &a,
+  void DselfInternal(CartesianStencil<siteVector,siteVector,DefaultImplParams> &st, CoarseMatrix &a,
                        const CoarseVector &in, CoarseVector &out, int dag) {
     int point = geom.npoint-1;
     autoView( out_v, out, AcceleratorWrite);
@@ -694,7 +694,7 @@ public:
     }
   }
 
-  void DhopInternal(CartesianStencil<siteVector,siteVector,int> &st, std::vector<CoarseMatrix> &a,
+  void DhopInternal(CartesianStencil<siteVector,siteVector,DefaultImplParams> &st, std::vector<CoarseMatrix> &a,
                     const CoarseVector &in, CoarseVector &out, int dag) {
     SimpleCompressor<siteVector> compressor;
 
@@ -784,9 +784,9 @@ public:
     _cbgrid(new GridRedBlackCartesian(&CoarseGrid)),
     geom(CoarseGrid._ndimension),
     hermitian(hermitian_),
-    Stencil(&CoarseGrid,geom.npoint,Even,geom.directions,geom.displacements,0),
-    StencilEven(_cbgrid,geom.npoint,Even,geom.directions,geom.displacements,0),
-    StencilOdd(_cbgrid,geom.npoint,Odd,geom.directions,geom.displacements,0),
+    Stencil(&CoarseGrid,geom.npoint,Even,geom.directions,geom.displacements),
+    StencilEven(_cbgrid,geom.npoint,Even,geom.directions,geom.displacements),
+    StencilOdd(_cbgrid,geom.npoint,Odd,geom.directions,geom.displacements),
     A(geom.npoint,&CoarseGrid),
     Aeven(geom.npoint,_cbgrid),
     Aodd(geom.npoint,_cbgrid),
@@ -804,9 +804,9 @@ public:
     _cbgrid(&CoarseRBGrid),
     geom(CoarseGrid._ndimension),
     hermitian(hermitian_),
-    Stencil(&CoarseGrid,geom.npoint,Even,geom.directions,geom.displacements,0),
-    StencilEven(&CoarseRBGrid,geom.npoint,Even,geom.directions,geom.displacements,0),
-    StencilOdd(&CoarseRBGrid,geom.npoint,Odd,geom.directions,geom.displacements,0),
+    Stencil(&CoarseGrid,geom.npoint,Even,geom.directions,geom.displacements),
+    StencilEven(&CoarseRBGrid,geom.npoint,Even,geom.directions,geom.displacements),
+    StencilOdd(&CoarseRBGrid,geom.npoint,Odd,geom.directions,geom.displacements),
     A(geom.npoint,&CoarseGrid),
     Aeven(geom.npoint,&CoarseRBGrid),
     Aodd(geom.npoint,&CoarseRBGrid),
diff --git a/Grid/algorithms/LinearOperator.h b/Grid/algorithms/LinearOperator.h
index b1cf4d97..090fed46 100644
--- a/Grid/algorithms/LinearOperator.h
+++ b/Grid/algorithms/LinearOperator.h
@@ -526,6 +526,7 @@ public:
       (*this)(Linop,in[k],out[k]);
     }
   };
+  virtual ~OperatorFunction(){};
 };
 
 template<class Field> class LinearFunction {
diff --git a/Grid/qcd/action/ActionParams.h b/Grid/qcd/action/ActionParams.h
index b332a2c3..122dfb9c 100644
--- a/Grid/qcd/action/ActionParams.h
+++ b/Grid/qcd/action/ActionParams.h
@@ -34,15 +34,6 @@ directory
 
 NAMESPACE_BEGIN(Grid);
 
-// These can move into a params header and be given MacroMagic serialisation
-struct DefaultImplParams {
-  Coordinate dirichlet; // Blocksize of dirichlet BCs
-  int  partialDirichlet;
-  DefaultImplParams()  {
-    dirichlet.resize(0);
-    partialDirichlet=0;
-  };
-};
 
 struct GparityWilsonImplParams {
   Coordinate twists;
diff --git a/Grid/stencil/Stencil.h b/Grid/stencil/Stencil.h
index ff8a6433..da230e7e 100644
--- a/Grid/stencil/Stencil.h
+++ b/Grid/stencil/Stencil.h
@@ -52,6 +52,16 @@
 
 NAMESPACE_BEGIN(Grid);
 
+// These can move into a params header and be given MacroMagic serialisation
+struct DefaultImplParams {
+  Coordinate dirichlet; // Blocksize of dirichlet BCs
+  int  partialDirichlet;
+  DefaultImplParams()  {
+    dirichlet.resize(0);
+    partialDirichlet=0;
+  };
+};
+
 ///////////////////////////////////////////////////////////////////
 // Gather for when there *is* need to SIMD split with compression
 ///////////////////////////////////////////////////////////////////
diff --git a/Grid/tensors/Tensor_traits.h b/Grid/tensors/Tensor_traits.h
index 99633cee..58fdc6ce 100644
--- a/Grid/tensors/Tensor_traits.h
+++ b/Grid/tensors/Tensor_traits.h
@@ -143,7 +143,7 @@ NAMESPACE_BEGIN(Grid);
     typedef ComplexD DoublePrecision2;
   };
 
-#ifdef GRID_CUDA
+#if defined(GRID_CUDA) || defined(GRID_HIP)  
   template<> struct GridTypeMapper<std::complex<float> > : public GridTypeMapper_Base {
     typedef std::complex<float> scalar_type;
     typedef std::complex<double> scalar_typeD;
diff --git a/systems/Crusher/config-command b/systems/Crusher/config-command
index ca1d6348..bd825ab3 100644
--- a/systems/Crusher/config-command
+++ b/systems/Crusher/config-command
@@ -1,4 +1,6 @@
+CLIME=`spack find --paths c-lime@2-3-9 | grep c-lime| cut -c 15-`
 ../../configure --enable-comms=mpi-auto \
+--with-lime=$CLIME \
 --enable-unified=no \
 --enable-shm=nvlink \
 --enable-accelerator=hip \
@@ -7,7 +9,7 @@
 --with-gmp=$OLCF_GMP_ROOT \
 --with-fftw=$FFTW_DIR/.. \
 --with-mpfr=/opt/cray/pe/gcc/mpfr/3.1.4/ \
---disable-gparity \
+--disable-fermion-reps \
 CXX=hipcc MPICXX=mpicxx \
 CXXFLAGS="-fPIC -I{$ROCM_PATH}/include/ -std=c++14 -I${MPICH_DIR}/include -L/lib64 " \
  LDFLAGS="-L/lib64 -L/opt/rocm-5.2.0/lib/ -L${MPICH_DIR}/lib -lmpi -L${CRAY_MPICH_ROOTDIR}/gtl/lib -lmpi_gtl_hsa -lamdhip64 "
diff --git a/systems/Crusher/sourceme.sh b/systems/Crusher/sourceme.sh
index 42e15fb1..ac248b93 100644
--- a/systems/Crusher/sourceme.sh
+++ b/systems/Crusher/sourceme.sh
@@ -1,3 +1,5 @@
+. /autofs/nccs-svm1_home1/paboyle/Crusher/Grid/spack/share/spack/setup-env.sh
+spack load c-lime
 export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/sw/crusher/spack-envs/base/opt/cray-sles15-zen3/gcc-11.2.0/gperftools-2.9.1-72ubwtuc5wcz2meqltbfdb76epufgzo2/lib
 module load emacs 
 #module load gperftools
diff --git a/tests/debug/Test_cayley_cg.cc b/tests/debug/Test_cayley_cg.cc
index 5418a8af..74492fd9 100644
--- a/tests/debug/Test_cayley_cg.cc
+++ b/tests/debug/Test_cayley_cg.cc
@@ -125,10 +125,10 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage <<"======================"<<std::endl;
   std::cout<<GridLogMessage <<"DomainWallFermion test"<<std::endl;
   std::cout<<GridLogMessage <<"======================"<<std::endl;
-  DomainWallFermionR Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+  DomainWallFermionD Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
   DomainWallFermionF DdwfF(UmuF,*FGridF,*FrbGridF,*UGridF,*UrbGridF,mass,M5);
-  TestCGinversions<DomainWallFermionR>(Ddwf,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
-  TestReconstruct5DFA<DomainWallFermionR,DomainWallFermionF>(Ddwf,DdwfF,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+  TestCGinversions<DomainWallFermionD>(Ddwf,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+  TestReconstruct5DFA<DomainWallFermionD,DomainWallFermionF>(Ddwf,DdwfF,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
 
   RealD b=1.5;// Scale factor b+c=2, b-c=1
   RealD c=0.5;
@@ -137,54 +137,54 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage <<"======================"<<std::endl;
   std::cout<<GridLogMessage <<"MobiusFermion test"<<std::endl;
   std::cout<<GridLogMessage <<"======================"<<std::endl;
-  MobiusFermionR Dmob(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,b,c);
+  MobiusFermionD Dmob(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,b,c);
   MobiusFermionF DmobF(UmuF,*FGridF,*FrbGridF,*UGridF,*UrbGridF,mass,M5,b,c);
-  TestCGinversions<MobiusFermionR>(Dmob,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
-  TestReconstruct5DFA<MobiusFermionR,MobiusFermionF>(Dmob,DmobF,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+  TestCGinversions<MobiusFermionD>(Dmob,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+  TestReconstruct5DFA<MobiusFermionD,MobiusFermionF>(Dmob,DmobF,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
 
   std::cout<<GridLogMessage <<"======================"<<std::endl;
   std::cout<<GridLogMessage <<"ZMobiusFermion test"<<std::endl;
   std::cout<<GridLogMessage <<"======================"<<std::endl;
-  ZMobiusFermionR ZDmob(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,gamma,b,c);
-  TestCGinversions<ZMobiusFermionR>(ZDmob,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
-  TestReconstruct5D<ZMobiusFermionR>(ZDmob,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+  ZMobiusFermionD ZDmob(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,gamma,b,c);
+  TestCGinversions<ZMobiusFermionD>(ZDmob,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+  TestReconstruct5D<ZMobiusFermionD>(ZDmob,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
 
   std::cout<<GridLogMessage <<"======================"<<std::endl;
   std::cout<<GridLogMessage <<"MobiusZolotarevFermion test"<<std::endl;
   std::cout<<GridLogMessage <<"======================"<<std::endl;
-  MobiusZolotarevFermionR Dzolo(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,b,c,0.1,2.0);
-  TestCGinversions<MobiusZolotarevFermionR>(Dzolo,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
-  TestReconstruct5D<MobiusZolotarevFermionR>(Dzolo,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+  MobiusZolotarevFermionD Dzolo(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,b,c,0.1,2.0);
+  TestCGinversions<MobiusZolotarevFermionD>(Dzolo,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+  TestReconstruct5D<MobiusZolotarevFermionD>(Dzolo,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
 
   std::cout<<GridLogMessage <<"======================"<<std::endl;
   std::cout<<GridLogMessage <<"ScaledShamirFermion test"<<std::endl;
   std::cout<<GridLogMessage <<"======================"<<std::endl;
-  ScaledShamirFermionR Dsham(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,2.0);
+  ScaledShamirFermionD Dsham(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,2.0);
   ScaledShamirFermionF DshamF(UmuF,*FGridF,*FrbGridF,*UGridF,*UrbGridF,mass,M5,2.0);
-  TestCGinversions<ScaledShamirFermionR>(Dsham,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
-  TestReconstruct5DFA<ScaledShamirFermionR,ScaledShamirFermionF>(Dsham,DshamF,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+  TestCGinversions<ScaledShamirFermionD>(Dsham,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+  TestReconstruct5DFA<ScaledShamirFermionD,ScaledShamirFermionF>(Dsham,DshamF,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
 
   std::cout<<GridLogMessage <<"======================"<<std::endl;
   std::cout<<GridLogMessage <<"ShamirZolotarevFermion test"<<std::endl;
   std::cout<<GridLogMessage <<"======================"<<std::endl;
-  ShamirZolotarevFermionR Dshamz(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,0.1,2.0);
-  TestCGinversions<ShamirZolotarevFermionR>(Dshamz,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
-  TestReconstruct5D<ShamirZolotarevFermionR>(Dshamz,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+  ShamirZolotarevFermionD Dshamz(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,0.1,2.0);
+  TestCGinversions<ShamirZolotarevFermionD>(Dshamz,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+  TestReconstruct5D<ShamirZolotarevFermionD>(Dshamz,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
 
   std::cout<<GridLogMessage <<"======================"<<std::endl;
   std::cout<<GridLogMessage <<"OverlapWilsonCayleyTanhFermion test"<<std::endl;
   std::cout<<GridLogMessage <<"======================"<<std::endl;
-  OverlapWilsonCayleyTanhFermionR Dov(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,1.0);
+  OverlapWilsonCayleyTanhFermionD Dov(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,1.0);
   OverlapWilsonCayleyTanhFermionF DovF(UmuF,*FGridF,*FrbGridF,*UGridF,*UrbGridF,mass,M5,1.0);
-  TestCGinversions<OverlapWilsonCayleyTanhFermionR>(Dov,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
-  TestReconstruct5DFA<OverlapWilsonCayleyTanhFermionR,OverlapWilsonCayleyTanhFermionF>(Dov,DovF,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+  TestCGinversions<OverlapWilsonCayleyTanhFermionD>(Dov,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+  TestReconstruct5DFA<OverlapWilsonCayleyTanhFermionD,OverlapWilsonCayleyTanhFermionF>(Dov,DovF,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
 
   std::cout<<GridLogMessage <<"======================"<<std::endl;
   std::cout<<GridLogMessage <<"OverlapWilsonCayleyZolotarevFermion test"<<std::endl;
   std::cout<<GridLogMessage <<"======================"<<std::endl;
-  OverlapWilsonCayleyZolotarevFermionR Dovz(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,0.1,2.0);
-  TestCGinversions<OverlapWilsonCayleyZolotarevFermionR>(Dovz,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
-  TestReconstruct5D<OverlapWilsonCayleyZolotarevFermionR>(Dovz,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+  OverlapWilsonCayleyZolotarevFermionD Dovz(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,0.1,2.0);
+  TestCGinversions<OverlapWilsonCayleyZolotarevFermionD>(Dovz,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+  TestReconstruct5D<OverlapWilsonCayleyZolotarevFermionD>(Dovz,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
 
   Grid_finalize();
 }
diff --git a/tests/debug/Test_cayley_coarsen_support.cc b/tests/debug/Test_cayley_coarsen_support.cc
index b2f691d7..2190a9b0 100644
--- a/tests/debug/Test_cayley_coarsen_support.cc
+++ b/tests/debug/Test_cayley_coarsen_support.cc
@@ -95,8 +95,8 @@ int main (int argc, char ** argv)
   RealD mass=0.5;
   RealD M5=1.8;
 
-  DomainWallFermionR Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
-  Gamma5R5HermitianLinearOperator<DomainWallFermionR,LatticeFermion> HermIndefOp(Ddwf);
+  DomainWallFermionD Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+  Gamma5R5HermitianLinearOperator<DomainWallFermionD,LatticeFermion> HermIndefOp(Ddwf);
 
   HermIndefOp.Op(src,ref);
   HermIndefOp.OpDiag(src,result);
@@ -118,7 +118,7 @@ int main (int argc, char ** argv)
 
   std::cout<<GridLogMessage<<"Calling Aggregation class" <<std::endl;
 
-  MdagMLinearOperator<DomainWallFermionR,LatticeFermion> HermDefOp(Ddwf);
+  MdagMLinearOperator<DomainWallFermionD,LatticeFermion> HermDefOp(Ddwf);
   typedef Aggregation<vSpinColourVector,vTComplex,nbasis> Subspace;
 
   Subspace Aggregates(Coarse5d,FGrid,cb);
diff --git a/tests/debug/Test_cayley_even_odd.cc b/tests/debug/Test_cayley_even_odd.cc
index 5e800b26..b6eecc0f 100644
--- a/tests/debug/Test_cayley_even_odd.cc
+++ b/tests/debug/Test_cayley_even_odd.cc
@@ -76,41 +76,41 @@ int main (int argc, char ** argv)
   RealD mass=0.1;
   RealD M5  =1.8;
   std::cout<<GridLogMessage <<"DomainWallFermion test"<<std::endl;
-  DomainWallFermionR Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
-  TestWhat<DomainWallFermionR>(Ddwf,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+  DomainWallFermionD Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+  TestWhat<DomainWallFermionD>(Ddwf,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
 
   RealD b=1.5;// Scale factor b+c=2, b-c=1
   RealD c=0.5;
   std::vector<ComplexD> gamma(Ls,ComplexD(1.0,0.1));
   
   std::cout<<GridLogMessage <<"MobiusFermion test"<<std::endl;
-  MobiusFermionR Dmob(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,b,c);
-  TestWhat<MobiusFermionR>(Dmob,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+  MobiusFermionD Dmob(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,b,c);
+  TestWhat<MobiusFermionD>(Dmob,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
 
   std::cout<<GridLogMessage <<"ZMobiusFermion test"<<std::endl;
-  ZMobiusFermionR ZDmob(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,gamma,b,c);
-  TestWhat<ZMobiusFermionR>(ZDmob,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+  ZMobiusFermionD ZDmob(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,gamma,b,c);
+  TestWhat<ZMobiusFermionD>(ZDmob,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
 
   std::cout<<GridLogMessage <<"MobiusZolotarevFermion test"<<std::endl;
-  MobiusZolotarevFermionR Dzolo(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,b,c,0.1,2.0);
-  TestWhat<MobiusZolotarevFermionR>(Dzolo,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+  MobiusZolotarevFermionD Dzolo(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,b,c,0.1,2.0);
+  TestWhat<MobiusZolotarevFermionD>(Dzolo,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
 
   std::cout<<GridLogMessage <<"ScaledShamirFermion test"<<std::endl;
-  ScaledShamirFermionR Dsham(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,2.0);
-  TestWhat<ScaledShamirFermionR>(Dsham,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+  ScaledShamirFermionD Dsham(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,2.0);
+  TestWhat<ScaledShamirFermionD>(Dsham,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
 
 
   std::cout<<GridLogMessage <<"ShamirZolotarevFermion test"<<std::endl;
-  ShamirZolotarevFermionR Dshamz(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,0.1,2.0);
-  TestWhat<ShamirZolotarevFermionR>(Dshamz,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+  ShamirZolotarevFermionD Dshamz(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,0.1,2.0);
+  TestWhat<ShamirZolotarevFermionD>(Dshamz,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
 
   std::cout<<GridLogMessage <<"OverlapWilsonCayleyTanhFermion test"<<std::endl;
-  OverlapWilsonCayleyTanhFermionR Dov(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,1.0);
-  TestWhat<OverlapWilsonCayleyTanhFermionR>(Dov,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+  OverlapWilsonCayleyTanhFermionD Dov(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,1.0);
+  TestWhat<OverlapWilsonCayleyTanhFermionD>(Dov,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
 
   std::cout<<GridLogMessage <<"OverlapWilsonCayleyZolotarevFermion test"<<std::endl;
-  OverlapWilsonCayleyZolotarevFermionR Dovz(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,0.1,2.0);
-  TestWhat<OverlapWilsonCayleyZolotarevFermionR>(Dovz,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+  OverlapWilsonCayleyZolotarevFermionD Dovz(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,0.1,2.0);
+  TestWhat<OverlapWilsonCayleyZolotarevFermionD>(Dovz,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
 
   Grid_finalize();
 }
diff --git a/tests/debug/Test_cayley_ldop_cr.cc b/tests/debug/Test_cayley_ldop_cr.cc
index 416017e5..997b8df5 100644
--- a/tests/debug/Test_cayley_ldop_cr.cc
+++ b/tests/debug/Test_cayley_ldop_cr.cc
@@ -83,8 +83,8 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
   std::cout<<GridLogMessage << "Building g5R5 hermitian DWF operator" <<std::endl;
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
-  DomainWallFermionR Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
-  Gamma5R5HermitianLinearOperator<DomainWallFermionR,LatticeFermion> HermIndefOp(Ddwf);
+  DomainWallFermionD Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+  Gamma5R5HermitianLinearOperator<DomainWallFermionD,LatticeFermion> HermIndefOp(Ddwf);
 
   const int nbasis = 8;
 
@@ -95,7 +95,7 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
   std::cout<<GridLogMessage << "Calling Aggregation class to build subspace" <<std::endl;
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
-  MdagMLinearOperator<DomainWallFermionR,LatticeFermion> HermDefOp(Ddwf);
+  MdagMLinearOperator<DomainWallFermionD,LatticeFermion> HermDefOp(Ddwf);
   Subspace Aggregates(Coarse5d,FGrid,cb);
   Aggregates.CreateSubspace(RNG5,HermDefOp);
 
diff --git a/tests/debug/Test_cayley_mres.cc b/tests/debug/Test_cayley_mres.cc
index bfbc3cf7..26d3dc60 100644
--- a/tests/debug/Test_cayley_mres.cc
+++ b/tests/debug/Test_cayley_mres.cc
@@ -128,8 +128,8 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage <<"======================"<<std::endl;
   std::cout<<GridLogMessage <<"DomainWallFermion test"<<std::endl;
   std::cout<<GridLogMessage <<"======================"<<std::endl;
-  DomainWallFermionR Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
-  TestConserved<DomainWallFermionR>(Ddwf,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+  DomainWallFermionD Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+  TestConserved<DomainWallFermionD>(Ddwf,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
 
   RealD b=1.5;// Scale factor b+c=2, b-c=1
   RealD c=0.5;
@@ -138,23 +138,23 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage <<"======================"<<std::endl;
   std::cout<<GridLogMessage <<"MobiusFermion test"<<std::endl;
   std::cout<<GridLogMessage <<"======================"<<std::endl;
-  MobiusFermionR Dmob(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,b,c);
-  TestConserved<MobiusFermionR>(Dmob,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+  MobiusFermionD Dmob(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,b,c);
+  TestConserved<MobiusFermionD>(Dmob,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
 
   std::cout<<GridLogMessage <<"======================"<<std::endl;
   std::cout<<GridLogMessage <<"ScaledShamirFermion test"<<std::endl;
   std::cout<<GridLogMessage <<"======================"<<std::endl;
-  ScaledShamirFermionR Dsham(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,2.0);
-  TestConserved<ScaledShamirFermionR>(Dsham,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+  ScaledShamirFermionD Dsham(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,2.0);
+  TestConserved<ScaledShamirFermionD>(Dsham,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
 
   std::cout<<GridLogMessage <<"======================"<<std::endl;
   std::cout<<GridLogMessage <<"ZMobiusFermion test"<<std::endl;
   std::cout<<GridLogMessage <<"======================"<<std::endl;
   for(int s=0;s<Ls;s++) omegasrev[s]=conjugate(omegas[Ls-1-s]);
   //  for(int s=0;s<Ls;s++) omegasrev[s]=omegas[Ls-1-s];
-  ZMobiusFermionR ZDmob(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,omegas,b,c);
-  ZMobiusFermionR ZDmobrev(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,omegasrev,b,c);
-  TestConserved<ZMobiusFermionR>(ZDmob,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5,&ZDmobrev);
+  ZMobiusFermionD ZDmob(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,omegas,b,c);
+  ZMobiusFermionD ZDmobrev(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,omegasrev,b,c);
+  TestConserved<ZMobiusFermionD>(ZDmob,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5,&ZDmobrev);
 
   Grid_finalize();
 }
@@ -290,7 +290,7 @@ void  TestConserved(Action & Ddwf,
     const RealD DmuPAmu{real(TensorRemove(sumPA[t]-sumPA[(t-1+Nt)%Nt]))};
     std::cout<<GridLogMessage<<" t "<<t<<" DmuPAmu "<<DmuPAmu
              <<" PP "<<real(TensorRemove(sumPP[t]))<<" PJ5q "<<real(TensorRemove(sumPJ5q[t]))
-             <<" Ward Identity defect " <<(DmuPAmu - 2.*real(TensorRemove(Ddwf.mass*sumPP[t] + sumPJ5q[t])))<<std::endl;
+             <<" Ward Identity defect " <<(DmuPAmu - 2.*real(TensorRemove(Ddwf.Mass()*sumPP[t] + sumPJ5q[t])))<<std::endl;
   }
   
   ///////////////////////////////
@@ -539,7 +539,7 @@ void  TestConserved1(Action & Ddwf, Action & Ddwfrev,
   PA       = trace(g5*Axial_mu);
   PP       = trace(adj(prop4)*prop4);
 
-  Defect = Defect - 2.0*Ddwf.mass* PP;
+  Defect = Defect - 2.0*Ddwf.Mass()* PP;
   Defect = Defect - 2.0*PJ5q;
 
   std::vector<TComplex> sumPAref;
@@ -565,8 +565,8 @@ void  TestConserved1(Action & Ddwf, Action & Ddwfrev,
     std::cout    <<" PAc action    "<<real(TensorRemove(sumPA[t]));
     std::cout    <<" PJ5q ref      "<<real(TensorRemove(sumPJ5qref[t]));
     std::cout    <<" PJ5q action   "<<real(TensorRemove(sumPJ5q[t]));
-    std::cout <<"WTI defects "<<real(TensorRemove(sumPAref[t]-sumPAref[(t-1+Nt)%Nt] - 2.0*(Ddwf.mass*sumPP[t] + sumPJ5q[t]) ))<<",";
-    std::cout <<real(TensorRemove(sumPA[t]-sumPA[(t-1+Nt)%Nt] - 2.0*(Ddwf.mass*sumPP[t] + sumPJ5q[t]) ))<<"\n";
+    std::cout <<"WTI defects "<<real(TensorRemove(sumPAref[t]-sumPAref[(t-1+Nt)%Nt] - 2.0*(Ddwf.Mass()*sumPP[t] + sumPJ5q[t]) ))<<",";
+    std::cout <<real(TensorRemove(sumPA[t]-sumPA[(t-1+Nt)%Nt] - 2.0*(Ddwf.Mass()*sumPP[t] + sumPJ5q[t]) ))<<"\n";
   }
 }
 #endif
@@ -600,7 +600,7 @@ void  TestConserved1(Action & Ddwf, Action & Ddwfrev,
       // Dperp
       {
 	RealD diag = 5.0 - Ddwf.M5;
-	mass = Ddwf.mass;
+	mass = Ddwf.Mass();
 	autoView( psi,result5,CpuRead);
 	autoView( chi,tmp,   CpuWrite);
 	thread_for(sss,UGrid->oSites(),{
diff --git a/tests/debug/Test_heatbath_dwf_eofa.cc b/tests/debug/Test_heatbath_dwf_eofa.cc
index e1c18021..5920054d 100644
--- a/tests/debug/Test_heatbath_dwf_eofa.cc
+++ b/tests/debug/Test_heatbath_dwf_eofa.cc
@@ -77,8 +77,8 @@ int main(int argc, char** argv)
   LatticeGaugeField Umu(UGrid);
   SU<Nc>::HotConfiguration(RNG4, Umu);
 
-  DomainWallEOFAFermionR Lop(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mf,  mf, mpv,  0.0, -1, M5);
-  DomainWallEOFAFermionR Rop(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mpv, mf, mpv, -1.0,  1, M5);
+  DomainWallEOFAFermionD Lop(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mf,  mf, mpv,  0.0, -1, M5);
+  DomainWallEOFAFermionD Rop(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mpv, mf, mpv, -1.0,  1, M5);
 
   // Construct the action and test the heatbath (zero initial guess)
   {
diff --git a/tests/debug/Test_heatbath_dwf_eofa_gparity.cc b/tests/debug/Test_heatbath_dwf_eofa_gparity.cc
index 7eabfc65..982f35db 100644
--- a/tests/debug/Test_heatbath_dwf_eofa_gparity.cc
+++ b/tests/debug/Test_heatbath_dwf_eofa_gparity.cc
@@ -41,7 +41,7 @@ using namespace Grid;
  ;
 
 typedef GparityWilsonImplR FermionImplPolicy;
-typedef GparityDomainWallEOFAFermionR FermionAction;
+typedef GparityDomainWallEOFAFermionD FermionAction;
 typedef typename FermionAction::FermionField FermionField;
 
 // Parameters for test
@@ -82,7 +82,7 @@ int main(int argc, char** argv)
   LatticeGaugeField Umu(UGrid);
   SU<Nc>::HotConfiguration(RNG4, Umu);
 
-  // GparityDomainWallFermionR::ImplParams params;
+  // GparityDomainWallFermionD::ImplParams params;
   FermionAction::ImplParams params;
   FermionAction Lop(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mf,  mf, mpv,  0.0, -1, M5, params);
   FermionAction Rop(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mpv, mf, mpv, -1.0,  1, M5, params);
diff --git a/tests/debug/Test_heatbath_mobius_eofa.cc b/tests/debug/Test_heatbath_mobius_eofa.cc
index 48806642..3824daab 100644
--- a/tests/debug/Test_heatbath_mobius_eofa.cc
+++ b/tests/debug/Test_heatbath_mobius_eofa.cc
@@ -79,8 +79,8 @@ int main(int argc, char** argv)
   LatticeGaugeField Umu(UGrid);
   SU<Nc>::HotConfiguration(RNG4, Umu);
 
-  MobiusEOFAFermionR Lop(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mf,  mf, mpv,  0.0, -1, M5, b, c);
-  MobiusEOFAFermionR Rop(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mpv, mf, mpv, -1.0,  1, M5, b, c);
+  MobiusEOFAFermionD Lop(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mf,  mf, mpv,  0.0, -1, M5, b, c);
+  MobiusEOFAFermionD Rop(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mpv, mf, mpv, -1.0,  1, M5, b, c);
 
   // Construct the action and test the heatbath (zero initial guess)
   {
diff --git a/tests/debug/Test_heatbath_mobius_eofa_gparity.cc b/tests/debug/Test_heatbath_mobius_eofa_gparity.cc
index 52447e5e..fd3d96f8 100644
--- a/tests/debug/Test_heatbath_mobius_eofa_gparity.cc
+++ b/tests/debug/Test_heatbath_mobius_eofa_gparity.cc
@@ -41,7 +41,7 @@ using namespace Grid;
  ;
 
 typedef GparityWilsonImplR FermionImplPolicy;
-typedef GparityMobiusEOFAFermionR FermionAction;
+typedef GparityMobiusEOFAFermionD FermionAction;
 typedef typename FermionAction::FermionField FermionField;
 
 // Parameters for test
diff --git a/tests/debug/Test_reweight_dwf_eofa.cc b/tests/debug/Test_reweight_dwf_eofa.cc
index a150b18f..6a5452c7 100644
--- a/tests/debug/Test_reweight_dwf_eofa.cc
+++ b/tests/debug/Test_reweight_dwf_eofa.cc
@@ -105,10 +105,10 @@ int main(int argc, char **argv)
   SU<Nc>::HotConfiguration(RNG4, Umu);
 
   // Initialize RHMC fermion operators
-  DomainWallFermionR Ddwf_f(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mf, M5);
-  DomainWallFermionR Ddwf_b(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mb, M5);
-  SchurDiagMooeeOperator<DomainWallFermionR, LatticeFermion> MdagM(Ddwf_f);
-  SchurDiagMooeeOperator<DomainWallFermionR, LatticeFermion> VdagV(Ddwf_b);
+  DomainWallFermionD Ddwf_f(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mf, M5);
+  DomainWallFermionD Ddwf_b(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mb, M5);
+  SchurDiagMooeeOperator<DomainWallFermionD, LatticeFermion> MdagM(Ddwf_f);
+  SchurDiagMooeeOperator<DomainWallFermionD, LatticeFermion> VdagV(Ddwf_b);
 
   // Degree 12 rational approximations to x^(1/4) and x^(-1/4)
   double     lo = 0.0001;
@@ -153,10 +153,10 @@ int main(int argc, char **argv)
   RealD shift_L = 0.0;
   RealD shift_R = -1.0;
   int pm = 1;
-  DomainWallEOFAFermionR Deofa_L(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mf, mf, mb, shift_L, pm, M5);
-  DomainWallEOFAFermionR Deofa_R(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mb, mf, mb, shift_R, pm, M5);
-  MdagMLinearOperator<DomainWallEOFAFermionR, LatticeFermion> LdagL(Deofa_L);
-  MdagMLinearOperator<DomainWallEOFAFermionR, LatticeFermion> RdagR(Deofa_R);
+  DomainWallEOFAFermionD Deofa_L(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mf, mf, mb, shift_L, pm, M5);
+  DomainWallEOFAFermionD Deofa_R(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mb, mf, mb, shift_R, pm, M5);
+  MdagMLinearOperator<DomainWallEOFAFermionD, LatticeFermion> LdagL(Deofa_L);
+  MdagMLinearOperator<DomainWallEOFAFermionD, LatticeFermion> RdagR(Deofa_R);
 
   // Stochastically estimate reweighting factor via EOFA
   RealD k = Deofa_L.k;
diff --git a/tests/debug/Test_reweight_dwf_eofa_gparity.cc b/tests/debug/Test_reweight_dwf_eofa_gparity.cc
index df2d95a0..70ae94aa 100644
--- a/tests/debug/Test_reweight_dwf_eofa_gparity.cc
+++ b/tests/debug/Test_reweight_dwf_eofa_gparity.cc
@@ -33,7 +33,7 @@ using namespace std;
 using namespace Grid;
  ;
 
-typedef typename GparityDomainWallFermionR::FermionField FermionField;
+typedef typename GparityDomainWallFermionD::FermionField FermionField;
 
 // parameters for test
 const std::vector<int> grid_dim = { 8, 8, 8, 8 };
@@ -107,11 +107,11 @@ int main(int argc, char **argv)
   SU<Nc>::HotConfiguration(RNG4, Umu);
 
   // Initialize RHMC fermion operators
-  GparityDomainWallFermionR::ImplParams params;
-  GparityDomainWallFermionR Ddwf_f(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mf, M5, params);
-  GparityDomainWallFermionR Ddwf_b(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mb, M5, params);
-  SchurDiagMooeeOperator<GparityDomainWallFermionR, FermionField> MdagM(Ddwf_f);
-  SchurDiagMooeeOperator<GparityDomainWallFermionR, FermionField> VdagV(Ddwf_b);
+  GparityDomainWallFermionD::ImplParams params;
+  GparityDomainWallFermionD Ddwf_f(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mf, M5, params);
+  GparityDomainWallFermionD Ddwf_b(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mb, M5, params);
+  SchurDiagMooeeOperator<GparityDomainWallFermionD, FermionField> MdagM(Ddwf_f);
+  SchurDiagMooeeOperator<GparityDomainWallFermionD, FermionField> VdagV(Ddwf_b);
 
   // Degree 12 rational approximations to x^(1/4) and x^(-1/4)
   double     lo = 0.0001;
@@ -156,10 +156,10 @@ int main(int argc, char **argv)
   RealD shift_L = 0.0;
   RealD shift_R = -1.0;
   int pm = 1;
-  GparityDomainWallEOFAFermionR Deofa_L(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mf, mf, mb, shift_L, pm, M5, params);
-  GparityDomainWallEOFAFermionR Deofa_R(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mb, mf, mb, shift_R, pm, M5, params);
-  MdagMLinearOperator<GparityDomainWallEOFAFermionR, FermionField> LdagL(Deofa_L);
-  MdagMLinearOperator<GparityDomainWallEOFAFermionR, FermionField> RdagR(Deofa_R);
+  GparityDomainWallEOFAFermionD Deofa_L(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mf, mf, mb, shift_L, pm, M5, params);
+  GparityDomainWallEOFAFermionD Deofa_R(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mb, mf, mb, shift_R, pm, M5, params);
+  MdagMLinearOperator<GparityDomainWallEOFAFermionD, FermionField> LdagL(Deofa_L);
+  MdagMLinearOperator<GparityDomainWallEOFAFermionD, FermionField> RdagR(Deofa_R);
 
   // Stochastically estimate reweighting factor via EOFA
   RealD k = Deofa_L.k;
diff --git a/tests/debug/Test_reweight_mobius_eofa.cc b/tests/debug/Test_reweight_mobius_eofa.cc
index 88ecab7d..744dd302 100644
--- a/tests/debug/Test_reweight_mobius_eofa.cc
+++ b/tests/debug/Test_reweight_mobius_eofa.cc
@@ -107,10 +107,10 @@ int main(int argc, char **argv)
   SU<Nc>::HotConfiguration(RNG4, Umu);
 
   // Initialize RHMC fermion operators
-  MobiusFermionR Ddwf_f(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mf, M5, b, c);
-  MobiusFermionR Ddwf_b(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mb, M5, b, c);
-  SchurDiagMooeeOperator<MobiusFermionR, LatticeFermion> MdagM(Ddwf_f);
-  SchurDiagMooeeOperator<MobiusFermionR, LatticeFermion> VdagV(Ddwf_b);
+  MobiusFermionD Ddwf_f(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mf, M5, b, c);
+  MobiusFermionD Ddwf_b(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mb, M5, b, c);
+  SchurDiagMooeeOperator<MobiusFermionD, LatticeFermion> MdagM(Ddwf_f);
+  SchurDiagMooeeOperator<MobiusFermionD, LatticeFermion> VdagV(Ddwf_b);
 
   // Degree 12 rational approximations to x^(1/4) and x^(-1/4)
   double     lo = 0.0001;
@@ -155,10 +155,10 @@ int main(int argc, char **argv)
   RealD shift_L = 0.0;
   RealD shift_R = -1.0;
   int pm = 1;
-  MobiusEOFAFermionR Deofa_L(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mf, mf, mb, shift_L, pm, M5, b, c);
-  MobiusEOFAFermionR Deofa_R(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mb, mf, mb, shift_R, pm, M5, b, c);
-  MdagMLinearOperator<MobiusEOFAFermionR, LatticeFermion> LdagL(Deofa_L);
-  MdagMLinearOperator<MobiusEOFAFermionR, LatticeFermion> RdagR(Deofa_R);
+  MobiusEOFAFermionD Deofa_L(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mf, mf, mb, shift_L, pm, M5, b, c);
+  MobiusEOFAFermionD Deofa_R(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mb, mf, mb, shift_R, pm, M5, b, c);
+  MdagMLinearOperator<MobiusEOFAFermionD, LatticeFermion> LdagL(Deofa_L);
+  MdagMLinearOperator<MobiusEOFAFermionD, LatticeFermion> RdagR(Deofa_R);
 
   // Stochastically estimate reweighting factor via EOFA
   RealD k = Deofa_L.k;
diff --git a/tests/debug/Test_reweight_mobius_eofa_gparity.cc b/tests/debug/Test_reweight_mobius_eofa_gparity.cc
index 31708265..e2a4fb47 100644
--- a/tests/debug/Test_reweight_mobius_eofa_gparity.cc
+++ b/tests/debug/Test_reweight_mobius_eofa_gparity.cc
@@ -33,7 +33,7 @@ using namespace std;
 using namespace Grid;
  ;
 
-typedef typename GparityDomainWallFermionR::FermionField FermionField;
+typedef typename GparityDomainWallFermionD::FermionField FermionField;
 
 // parameters for test
 const std::vector<int> grid_dim = { 8, 8, 8, 8 };
@@ -109,11 +109,11 @@ int main(int argc, char **argv)
   SU<Nc>::HotConfiguration(RNG4, Umu);
 
   // Initialize RHMC fermion operators
-  GparityDomainWallFermionR::ImplParams params;
-  GparityMobiusFermionR Ddwf_f(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mf, M5, b, c, params);
-  GparityMobiusFermionR Ddwf_b(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mb, M5, b, c, params);
-  SchurDiagMooeeOperator<GparityMobiusFermionR, FermionField> MdagM(Ddwf_f);
-  SchurDiagMooeeOperator<GparityMobiusFermionR, FermionField> VdagV(Ddwf_b);
+  GparityDomainWallFermionD::ImplParams params;
+  GparityMobiusFermionD Ddwf_f(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mf, M5, b, c, params);
+  GparityMobiusFermionD Ddwf_b(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mb, M5, b, c, params);
+  SchurDiagMooeeOperator<GparityMobiusFermionD, FermionField> MdagM(Ddwf_f);
+  SchurDiagMooeeOperator<GparityMobiusFermionD, FermionField> VdagV(Ddwf_b);
 
   // Degree 12 rational approximations to x^(1/4) and x^(-1/4)
   double     lo = 0.0001;
@@ -158,10 +158,10 @@ int main(int argc, char **argv)
   RealD shift_L = 0.0;
   RealD shift_R = -1.0;
   int pm = 1;
-  GparityMobiusEOFAFermionR Deofa_L(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mf, mf, mb, shift_L, pm, M5, b, c, params);
-  GparityMobiusEOFAFermionR Deofa_R(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mb, mf, mb, shift_R, pm, M5, b, c, params);
-  MdagMLinearOperator<GparityMobiusEOFAFermionR, FermionField> LdagL(Deofa_L);
-  MdagMLinearOperator<GparityMobiusEOFAFermionR, FermionField> RdagR(Deofa_R);
+  GparityMobiusEOFAFermionD Deofa_L(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mf, mf, mb, shift_L, pm, M5, b, c, params);
+  GparityMobiusEOFAFermionD Deofa_R(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mb, mf, mb, shift_R, pm, M5, b, c, params);
+  MdagMLinearOperator<GparityMobiusEOFAFermionD, FermionField> LdagL(Deofa_L);
+  MdagMLinearOperator<GparityMobiusEOFAFermionD, FermionField> RdagR(Deofa_R);
 
   // Stochastically estimate reweighting factor via EOFA
   RealD k = Deofa_L.k;
diff --git a/tests/hmc/Test_hmc_WC2ASFG_Production.cc b/tests/hmc/Test_hmc_WC2ASFG_Production.cc
index 55c84c0a..90f43ede 100644
--- a/tests/hmc/Test_hmc_WC2ASFG_Production.cc
+++ b/tests/hmc/Test_hmc_WC2ASFG_Production.cc
@@ -28,7 +28,7 @@ directory
 /*  END LEGAL */
 #include <Grid/Grid.h>
 
-
+#ifdef ENABLE_FERMION_REPS
 namespace Grid{
   struct FermionParameters: Serializable {
     GRID_SERIALIZABLE_CLASS_MEMBERS(FermionParameters,
@@ -210,4 +210,6 @@ int main(int argc, char **argv)
   Grid_finalize();
 
 } // main
-
+#else
+int main(int argc, char **argv){}
+#endif
diff --git a/tests/hmc/Test_hmc_WC2SFG_Production.cc b/tests/hmc/Test_hmc_WC2SFG_Production.cc
index 80f7baaf..16ca05a0 100644
--- a/tests/hmc/Test_hmc_WC2SFG_Production.cc
+++ b/tests/hmc/Test_hmc_WC2SFG_Production.cc
@@ -29,6 +29,7 @@ directory
 #include <Grid/Grid.h>
 
 
+#ifdef ENABLE_FERMION_REPS
 namespace Grid{
   struct FermionParameters: Serializable {
     GRID_SERIALIZABLE_CLASS_MEMBERS(FermionParameters,
@@ -211,3 +212,6 @@ int main(int argc, char **argv)
   Grid_finalize();
 
 } // main
+#else
+int main(int argc, char **argv){}
+#endif
diff --git a/tests/hmc/Test_hmc_WCMixedRepFG_Production.cc b/tests/hmc/Test_hmc_WCMixedRepFG_Production.cc
index ac017860..3be63a15 100644
--- a/tests/hmc/Test_hmc_WCMixedRepFG_Production.cc
+++ b/tests/hmc/Test_hmc_WCMixedRepFG_Production.cc
@@ -32,6 +32,7 @@ directory
 #include "Grid/Grid.h"
 
 
+#ifdef ENABLE_FERMION_REPS
 namespace Grid{
   struct FermionParameters: Serializable {
     GRID_SERIALIZABLE_CLASS_MEMBERS(FermionParameters,
@@ -222,3 +223,6 @@ int main(int argc, char **argv) {
   Grid_finalize();
 
 } // main
+#else
+int main(int argc, char **argv){}
+#endif
diff --git a/tests/hmc/Test_hmc_WCadjFG_Production.cc b/tests/hmc/Test_hmc_WCadjFG_Production.cc
index a9673e6c..f0e2742d 100644
--- a/tests/hmc/Test_hmc_WCadjFG_Production.cc
+++ b/tests/hmc/Test_hmc_WCadjFG_Production.cc
@@ -29,6 +29,7 @@ directory
 #include <Grid/Grid.h>
 
 
+#ifdef ENABLE_FERMION_REPS
 namespace Grid{
   struct FermionParameters: Serializable {
     GRID_SERIALIZABLE_CLASS_MEMBERS(FermionParameters,
@@ -211,3 +212,6 @@ int main(int argc, char **argv)
 
 } // main
 
+#else
+int main(int argc, char **argv){}
+#endif
diff --git a/tests/hmc/Test_hmc_WilsonAdjointFermionGauge.cc b/tests/hmc/Test_hmc_WilsonAdjointFermionGauge.cc
index 774c9037..3d601d25 100644
--- a/tests/hmc/Test_hmc_WilsonAdjointFermionGauge.cc
+++ b/tests/hmc/Test_hmc_WilsonAdjointFermionGauge.cc
@@ -31,9 +31,10 @@ directory
 /*  END LEGAL */
 #include "Grid/Grid.h"
 
+#ifdef ENABLE_FERMION_REPS
+
 int main(int argc, char **argv) {
   using namespace Grid;
-   ;
 
   // Here change the allowed (higher) representations
   typedef Representations< FundamentalRepresentation, AdjointRepresentation > TheRepresentations;
@@ -127,3 +128,6 @@ int main(int argc, char **argv) {
 
 } // main
 
+#else
+int main(int argc, char **argv){}
+#endif
diff --git a/tests/hmc/Test_hmc_WilsonMixedRepresentationsFermionGauge.cc b/tests/hmc/Test_hmc_WilsonMixedRepresentationsFermionGauge.cc
index 60d760a6..66a325f2 100644
--- a/tests/hmc/Test_hmc_WilsonMixedRepresentationsFermionGauge.cc
+++ b/tests/hmc/Test_hmc_WilsonMixedRepresentationsFermionGauge.cc
@@ -33,6 +33,7 @@ directory
 
 
 
+#ifdef ENABLE_FERMION_REPS
 
 int main(int argc, char **argv) {
 #ifndef GRID_CUDA
@@ -138,3 +139,6 @@ int main(int argc, char **argv) {
 } // main
 
 
+#else
+int main(int argc, char **argv){}
+#endif
diff --git a/tests/hmc/Test_hmc_WilsonTwoIndexSymmetricFermionGauge.cc b/tests/hmc/Test_hmc_WilsonTwoIndexSymmetricFermionGauge.cc
index bdcd27c5..4769e396 100644
--- a/tests/hmc/Test_hmc_WilsonTwoIndexSymmetricFermionGauge.cc
+++ b/tests/hmc/Test_hmc_WilsonTwoIndexSymmetricFermionGauge.cc
@@ -29,6 +29,7 @@ directory
 /*  END LEGAL */
 #include "Grid/Grid.h"
 
+#ifdef ENABLE_FERMION_REPS
 int main(int argc, char **argv) {
   using namespace Grid;
    ;
@@ -127,3 +128,6 @@ int main(int argc, char **argv) {
 } // main
 
 
+#else
+int main(int argc, char **argv){}
+#endif
diff --git a/tests/lanczos/Test_WCMultiRep_lanczos.cc b/tests/lanczos/Test_WCMultiRep_lanczos.cc
index 58759c96..0bfc75be 100644
--- a/tests/lanczos/Test_WCMultiRep_lanczos.cc
+++ b/tests/lanczos/Test_WCMultiRep_lanczos.cc
@@ -28,19 +28,21 @@ directory
 /*  END LEGAL */
 #include <Grid/Grid.h>
 
+#ifdef ENABLE_FERMION_REPS
+
 using namespace std;
 using namespace Grid;
 
 
-//typedef WilsonCloverFermionR FermionOp;
-//typedef typename WilsonFermionR::FermionField FermionField;
+//typedef WilsonCloverFermionD FermionOp;
+//typedef typename WilsonFermionD::FermionField FermionField;
 
 typedef WilsonImplR FundImplPolicy;
-typedef WilsonCloverFermionR FundFermionAction; 
+typedef WilsonCloverFermionD FundFermionAction; 
 typedef typename FundFermionAction::FermionField FundFermionField;
 
 typedef WilsonTwoIndexAntiSymmetricImplR ASymmImplPolicy; 
-typedef WilsonCloverTwoIndexAntiSymmetricFermionR ASymmFermionAction; 
+typedef WilsonCloverTwoIndexAntiSymmetricFermionD ASymmFermionAction; 
 typedef typename ASymmFermionAction::FermionField ASymmFermionField;
 
 
@@ -175,3 +177,6 @@ NerscHmcCheckpointer<PeriodicGimplR> Checkpoint(CPparams);
 
   Grid_finalize();
 }
+#else
+int main(int argc,char **argv){ return 0;};
+#endif
diff --git a/tests/lanczos/Test_compressed_lanczos.cc b/tests/lanczos/Test_compressed_lanczos.cc
index d7d0d52d..28df3f99 100644
--- a/tests/lanczos/Test_compressed_lanczos.cc
+++ b/tests/lanczos/Test_compressed_lanczos.cc
@@ -188,8 +188,8 @@ int main (int argc, char ** argv) {
   std::cout << GridLogMessage << "Lattice dimensions: " << GridDefaultLatt() << "   Ls: " << Ls << std::endl;
 
   // ZMobius EO Operator
-  ZMobiusFermionR Ddwf(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mass, M5, Params.omega,1.,0.);
-  SchurDiagTwoOperator<ZMobiusFermionR,LatticeFermion> HermOp(Ddwf);
+  ZMobiusFermionD Ddwf(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mass, M5, Params.omega,1.,0.);
+  SchurDiagTwoOperator<ZMobiusFermionD,LatticeFermion> HermOp(Ddwf);
 
   // Eigenvector storage
   LanczosParams fine  =Params.FineParams;  
diff --git a/tests/lanczos/Test_dwf_compressed_lanczos_reorg.cc b/tests/lanczos/Test_dwf_compressed_lanczos_reorg.cc
index 669a7b6d..7a84a465 100644
--- a/tests/lanczos/Test_dwf_compressed_lanczos_reorg.cc
+++ b/tests/lanczos/Test_dwf_compressed_lanczos_reorg.cc
@@ -188,8 +188,8 @@ int main (int argc, char ** argv) {
   std::cout << GridLogMessage << "Lattice dimensions: " << GridDefaultLatt() << "   Ls: " << Ls << std::endl;
 
   // ZMobius EO Operator
-  ZMobiusFermionR Ddwf(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mass, M5, Params.omega,1.,0.);
-  SchurDiagTwoOperator<ZMobiusFermionR,LatticeFermion> HermOp(Ddwf);
+  ZMobiusFermionD Ddwf(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mass, M5, Params.omega,1.,0.);
+  SchurDiagTwoOperator<ZMobiusFermionD,LatticeFermion> HermOp(Ddwf);
 
   // Eigenvector storage
   LanczosParams fine  =Params.FineParams;  
diff --git a/tests/lanczos/Test_dwf_compressed_lanczos_reorg_synthetic.cc b/tests/lanczos/Test_dwf_compressed_lanczos_reorg_synthetic.cc
index f3cb567c..e82a9741 100644
--- a/tests/lanczos/Test_dwf_compressed_lanczos_reorg_synthetic.cc
+++ b/tests/lanczos/Test_dwf_compressed_lanczos_reorg_synthetic.cc
@@ -301,8 +301,8 @@ int main (int argc, char ** argv) {
   std::cout << GridLogMessage << "Lattice dimensions: " << GridDefaultLatt() << "   Ls: " << Ls << std::endl;
 
   // ZMobius EO Operator
-  ZMobiusFermionR Ddwf(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mass, M5, Params.omega,1.,0.);
-  SchurDiagTwoOperator<ZMobiusFermionR,LatticeFermion> HermOp(Ddwf);
+  ZMobiusFermionD Ddwf(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mass, M5, Params.omega,1.,0.);
+  SchurDiagTwoOperator<ZMobiusFermionD,LatticeFermion> HermOp(Ddwf);
 
   // Eigenvector storage
   LanczosParams fine  =Params.FineParams;  
diff --git a/tests/lanczos/Test_dwf_lanczos.cc b/tests/lanczos/Test_dwf_lanczos.cc
index 1fe29bb2..1723e756 100644
--- a/tests/lanczos/Test_dwf_lanczos.cc
+++ b/tests/lanczos/Test_dwf_lanczos.cc
@@ -35,8 +35,8 @@ template<typename Action>
 struct Setup{};
 
 template<>
-struct Setup<GparityMobiusFermionR>{
-  static GparityMobiusFermionR* getAction(LatticeGaugeField &Umu,
+struct Setup<GparityMobiusFermionD>{
+  static GparityMobiusFermionD* getAction(LatticeGaugeField &Umu,
 					  GridCartesian* FGrid, GridRedBlackCartesian* FrbGrid, GridCartesian* UGrid, GridRedBlackCartesian* UrbGrid){
     RealD mass=0.01;
     RealD M5=1.8;
@@ -44,17 +44,17 @@ struct Setup<GparityMobiusFermionR>{
     GparityMobiusFermionD ::ImplParams params;
     std::vector<int> twists({1,1,1,0});
     params.twists = twists;
-    return new GparityMobiusFermionR(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,mob_b,mob_b-1.,params);
+    return new GparityMobiusFermionD(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,mob_b,mob_b-1.,params);
   }
 };
 
 template<>
-struct Setup<DomainWallFermionR>{
-  static DomainWallFermionR* getAction(LatticeGaugeField &Umu,
+struct Setup<DomainWallFermionD>{
+  static DomainWallFermionD* getAction(LatticeGaugeField &Umu,
 					  GridCartesian* FGrid, GridRedBlackCartesian* FrbGrid, GridCartesian* UGrid, GridRedBlackCartesian* UrbGrid){
     RealD mass=0.01;
     RealD M5=1.8;
-    return new DomainWallFermionR(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+    return new DomainWallFermionD(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
   }
 };
 
@@ -127,9 +127,9 @@ int main (int argc, char ** argv)
   }
 
   if(action == "GparityMobius"){
-    run<GparityMobiusFermionR>();
+    run<GparityMobiusFermionD>();
   }else if(action == "DWF"){
-    run<DomainWallFermionR>();
+    run<DomainWallFermionD>();
   }else{
     std::cout << "Unknown action" << std::endl;
     exit(1);
diff --git a/tests/lanczos/Test_wilson_lanczos.cc b/tests/lanczos/Test_wilson_lanczos.cc
index af21fb1d..4814e8c6 100644
--- a/tests/lanczos/Test_wilson_lanczos.cc
+++ b/tests/lanczos/Test_wilson_lanczos.cc
@@ -32,8 +32,8 @@ using namespace std;
 using namespace Grid;
  ;
 
-typedef WilsonFermionR FermionOp;
-typedef typename WilsonFermionR::FermionField FermionField;
+typedef WilsonFermionD FermionOp;
+typedef typename WilsonFermionD::FermionField FermionField;
 
 
 RealD AllZero(RealD x) { return 0.; }
diff --git a/tests/smearing/Test_WilsonFlow.cc b/tests/smearing/Test_WilsonFlow.cc
index f339959a..e0726f87 100644
--- a/tests/smearing/Test_WilsonFlow.cc
+++ b/tests/smearing/Test_WilsonFlow.cc
@@ -96,13 +96,16 @@ int main(int argc, char **argv) {
   std::cout << GridLogMessage << "Initial plaquette: "
     << WilsonLoops<PeriodicGimplR>::avgPlaquette(Umu) << std::endl;
 
-  WilsonFlow<PeriodicGimplR> WF(WFPar.steps, WFPar.step_size, WFPar.meas_interval);
+  int t=WFPar.maxTau;
+  WilsonFlowAdaptive<PeriodicGimplR> WF(WFPar.step_size, WFPar.maxTau,
+					1.0e-4,
+					WFPar.meas_interval);
 
-  WF.smear_adaptive(Uflow, Umu, WFPar.maxTau);
+  WF.smear(Uflow, Umu);
 
   RealD WFlow_plaq = WilsonLoops<PeriodicGimplR>::avgPlaquette(Uflow);
   RealD WFlow_TC   = WilsonLoops<PeriodicGimplR>::TopologicalCharge(Uflow);
-  RealD WFlow_T0   = WF.energyDensityPlaquette(Uflow);
+  RealD WFlow_T0   = WF.energyDensityPlaquette(t,Uflow);
   std::cout << GridLogMessage << "Plaquette          "<< conf << "   " << WFlow_plaq << std::endl;
   std::cout << GridLogMessage << "T0                 "<< conf << "   " << WFlow_T0 << std::endl;
   std::cout << GridLogMessage << "TopologicalCharge  "<< conf << "   " << WFlow_TC   << std::endl;
diff --git a/tests/solver/Test_cf_cr_unprec.cc b/tests/solver/Test_cf_cr_unprec.cc
index aa750175..3addcb62 100644
--- a/tests/solver/Test_cf_cr_unprec.cc
+++ b/tests/solver/Test_cf_cr_unprec.cc
@@ -71,14 +71,14 @@ int main (int argc, char ** argv)
   RealD mass=0.1;
   RealD M5=1.8;
 
-  OverlapWilsonContFracTanhFermionR Dcf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,1.0);
+  OverlapWilsonContFracTanhFermionD Dcf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,1.0);
 
   ConjugateResidual<LatticeFermion> MCR(1.0e-8,10000);
 
-  MdagMLinearOperator<OverlapWilsonContFracTanhFermionR,LatticeFermion> HermPosDefOp(Dcf);
+  MdagMLinearOperator<OverlapWilsonContFracTanhFermionD,LatticeFermion> HermPosDefOp(Dcf);
   MCR(HermPosDefOp,src,result);
 
-  HermitianLinearOperator<OverlapWilsonContFracTanhFermionR,LatticeFermion> HermIndefOp(Dcf);
+  HermitianLinearOperator<OverlapWilsonContFracTanhFermionD,LatticeFermion> HermIndefOp(Dcf);
   MCR(HermIndefOp,src,result);
 
   Grid_finalize();
diff --git a/tests/solver/Test_coarse_even_odd.cc b/tests/solver/Test_coarse_even_odd.cc
index c7127121..9d2f8c22 100644
--- a/tests/solver/Test_coarse_even_odd.cc
+++ b/tests/solver/Test_coarse_even_odd.cc
@@ -108,8 +108,8 @@ int main(int argc, char** argv) {
   RealD mass = -0.30;
   RealD csw  = 1.9192;
 
-  WilsonCloverFermionR Dwc(Umu, *Grid_f, *RBGrid_f, mass, csw, csw);
-  MdagMLinearOperator<WilsonCloverFermionR, LatticeFermion> MdagMOp_Dwc(Dwc);
+  WilsonCloverFermionD Dwc(Umu, *Grid_f, *RBGrid_f, mass, csw, csw);
+  MdagMLinearOperator<WilsonCloverFermionD, LatticeFermion> MdagMOp_Dwc(Dwc);
 
   /////////////////////////////////////////////////////////////////////////////
   //                             Type definitions                            //
diff --git a/tests/solver/Test_contfrac_cg.cc b/tests/solver/Test_contfrac_cg.cc
index afabae4c..52599d07 100644
--- a/tests/solver/Test_contfrac_cg.cc
+++ b/tests/solver/Test_contfrac_cg.cc
@@ -102,21 +102,21 @@ int main (int argc, char ** argv)
 
 
   std::cout<<GridLogMessage <<"OverlapWilsonContFracTanhFermion  test"<<std::endl;
-  OverlapWilsonContFracTanhFermionR Dcf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,1.0);
-  TestCGinversions<OverlapWilsonContFracTanhFermionR>(Dcf,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+  OverlapWilsonContFracTanhFermionD Dcf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,1.0);
+  TestCGinversions<OverlapWilsonContFracTanhFermionD>(Dcf,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
 
   std::cout<<GridLogMessage <<"OverlapWilsonContFracZolotarevFermion  test"<<std::endl;
-  OverlapWilsonContFracZolotarevFermionR Dcfz(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,0.1,6.0);
-  TestCGinversions<OverlapWilsonContFracZolotarevFermionR>(Dcfz,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+  OverlapWilsonContFracZolotarevFermionD Dcfz(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,0.1,6.0);
+  TestCGinversions<OverlapWilsonContFracZolotarevFermionD>(Dcfz,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
 
 
   std::cout<<GridLogMessage <<"OverlapWilsonPartialFractionTanhFermion  test"<<std::endl;
-  OverlapWilsonPartialFractionTanhFermionR Dpf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,1.0);
-  TestCGinversions<OverlapWilsonPartialFractionTanhFermionR>(Dpf,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+  OverlapWilsonPartialFractionTanhFermionD Dpf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,1.0);
+  TestCGinversions<OverlapWilsonPartialFractionTanhFermionD>(Dpf,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
 
   std::cout<<GridLogMessage <<"OverlapWilsonPartialFractionZolotarevFermion  test"<<std::endl;
-  OverlapWilsonPartialFractionZolotarevFermionR Dpfz(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,0.1,6.0);
-  TestCGinversions<OverlapWilsonPartialFractionZolotarevFermionR>(Dpfz,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+  OverlapWilsonPartialFractionZolotarevFermionD Dpfz(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,0.1,6.0);
+  TestCGinversions<OverlapWilsonPartialFractionZolotarevFermionD>(Dpfz,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
 
 
   Grid_finalize();
diff --git a/tests/solver/Test_dwf_cg_prec.cc b/tests/solver/Test_dwf_cg_prec.cc
index debb736a..f4e346bf 100644
--- a/tests/solver/Test_dwf_cg_prec.cc
+++ b/tests/solver/Test_dwf_cg_prec.cc
@@ -79,7 +79,7 @@ int main(int argc, char** argv) {
 
   RealD mass = 0.01;
   RealD M5 = 1.8;
-  DomainWallFermionR Ddwf(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mass, M5);
+  DomainWallFermionD Ddwf(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mass, M5);
 
   LatticeFermion src_o(FrbGrid);
   LatticeFermion result_o(FrbGrid);
@@ -88,7 +88,7 @@ int main(int argc, char** argv) {
 
   GridStopWatch CGTimer;
 
-  SchurDiagMooeeOperator<DomainWallFermionR, LatticeFermion> HermOpEO(Ddwf);
+  SchurDiagMooeeOperator<DomainWallFermionD, LatticeFermion> HermOpEO(Ddwf);
   ConjugateGradient<LatticeFermion> CG(1.0e-5, 10000, 0);// switch off the assert
 
   CGTimer.Start();
@@ -98,8 +98,5 @@ int main(int argc, char** argv) {
   std::cout << GridLogMessage << "Total CG time : " << CGTimer.Elapsed()
             << std::endl;
 
-  std::cout << GridLogMessage << "######## Dhop calls summary" << std::endl;
-  Ddwf.Report();
-
   Grid_finalize();
 }
diff --git a/tests/solver/Test_dwf_cg_schur.cc b/tests/solver/Test_dwf_cg_schur.cc
index 6541e73d..bcc0cc40 100644
--- a/tests/solver/Test_dwf_cg_schur.cc
+++ b/tests/solver/Test_dwf_cg_schur.cc
@@ -70,7 +70,7 @@ int main (int argc, char ** argv)
   
   RealD mass=0.1;
   RealD M5=1.8;
-  DomainWallFermionR Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+  DomainWallFermionD Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
 
   ConjugateGradient<LatticeFermion> CG(1.0e-8,10000);
   SchurRedBlackDiagMooeeSolve<LatticeFermion> SchurSolver(CG);
diff --git a/tests/solver/Test_dwf_cg_unprec.cc b/tests/solver/Test_dwf_cg_unprec.cc
index c867ccf3..58614c49 100644
--- a/tests/solver/Test_dwf_cg_unprec.cc
+++ b/tests/solver/Test_dwf_cg_unprec.cc
@@ -70,9 +70,9 @@ int main (int argc, char ** argv)
   
   RealD mass=0.1;
   RealD M5=1.8;
-  DomainWallFermionR Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+  DomainWallFermionD Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
 
-  MdagMLinearOperator<DomainWallFermionR,LatticeFermion> HermOp(Ddwf);
+  MdagMLinearOperator<DomainWallFermionD,LatticeFermion> HermOp(Ddwf);
   ConjugateGradient<LatticeFermion> CG(1.0e-6,10000);
   CG(HermOp,src,result);
 
diff --git a/tests/solver/Test_dwf_cr_unprec.cc b/tests/solver/Test_dwf_cr_unprec.cc
index 8c8583ba..4d67231d 100644
--- a/tests/solver/Test_dwf_cr_unprec.cc
+++ b/tests/solver/Test_dwf_cr_unprec.cc
@@ -77,12 +77,12 @@ int main (int argc, char ** argv)
   
   RealD mass=0.5;
   RealD M5=1.8;
-  DomainWallFermionR Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+  DomainWallFermionD Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
 
-  MdagMLinearOperator<DomainWallFermionR,LatticeFermion> HermOp(Ddwf);
+  MdagMLinearOperator<DomainWallFermionD,LatticeFermion> HermOp(Ddwf);
   MCR(HermOp,src,result);
 
-  Gamma5R5HermitianLinearOperator<DomainWallFermionR,LatticeFermion> g5HermOp(Ddwf);
+  Gamma5R5HermitianLinearOperator<DomainWallFermionD,LatticeFermion> g5HermOp(Ddwf);
   MCR(g5HermOp,src,result);
 
 
diff --git a/tests/solver/Test_dwf_fpgcr.cc b/tests/solver/Test_dwf_fpgcr.cc
index 42cc8de1..3d779474 100644
--- a/tests/solver/Test_dwf_fpgcr.cc
+++ b/tests/solver/Test_dwf_fpgcr.cc
@@ -77,12 +77,12 @@ int main (int argc, char ** argv)
   
   RealD mass=0.5;
   RealD M5=1.8;
-  DomainWallFermionR Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+  DomainWallFermionD Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
 
   std::cout<<GridLogMessage<<"*********************************************************"<<std::endl;
   std::cout<<GridLogMessage<<"* Solving with MdagM VPGCR "<<std::endl;
   std::cout<<GridLogMessage<<"*********************************************************"<<std::endl;
-  MdagMLinearOperator<DomainWallFermionR,LatticeFermion> HermOp(Ddwf);
+  MdagMLinearOperator<DomainWallFermionD,LatticeFermion> HermOp(Ddwf);
   TrivialPrecon<LatticeFermion> simple;
   PrecGeneralisedConjugateResidual<LatticeFermion> PGCR(1.0e-6,10000,HermOp,simple,4,160);
 
@@ -92,7 +92,7 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage<<"*********************************************************"<<std::endl;
   std::cout<<GridLogMessage<<"* Solving with g5-VPGCR "<<std::endl;
   std::cout<<GridLogMessage<<"*********************************************************"<<std::endl;
-  Gamma5R5HermitianLinearOperator<DomainWallFermionR,LatticeFermion> g5HermOp(Ddwf);
+  Gamma5R5HermitianLinearOperator<DomainWallFermionD,LatticeFermion> g5HermOp(Ddwf);
   PrecGeneralisedConjugateResidual<LatticeFermion> PGCR5(1.0e-6,10000,g5HermOp,simple,4,160);
   result=Zero();
   PGCR5(src,result);
diff --git a/tests/solver/Test_dwf_hdcr.cc b/tests/solver/Test_dwf_hdcr.cc
index ba77dffa..31b58284 100644
--- a/tests/solver/Test_dwf_hdcr.cc
+++ b/tests/solver/Test_dwf_hdcr.cc
@@ -254,7 +254,7 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
   RealD mass=0.001;
   RealD M5=1.8;
-  DomainWallFermionR Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+  DomainWallFermionD Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
 
   typedef Aggregation<vSpinColourVector,vTComplex,nbasis>              Subspace;
   typedef CoarsenedMatrix<vSpinColourVector,vTComplex,nbasis>          CoarseOperator;
@@ -263,7 +263,7 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
   std::cout<<GridLogMessage << "Calling Aggregation class to build subspace" <<std::endl;
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
-  MdagMLinearOperator<DomainWallFermionR,LatticeFermion> HermDefOp(Ddwf);
+  MdagMLinearOperator<DomainWallFermionD,LatticeFermion> HermDefOp(Ddwf);
 
   Subspace Aggregates(Coarse5d,FGrid,0);
 
@@ -290,7 +290,7 @@ int main (int argc, char ** argv)
   typedef CoarsenedMatrix<vSpinColourVector,vTComplex,nbasis>    Level1Op;
   typedef CoarsenedMatrix<siteVector,iScalar<vTComplex>,nbasisc> Level2Op;
 
-  Gamma5R5HermitianLinearOperator<DomainWallFermionR,LatticeFermion> HermIndefOp(Ddwf);
+  Gamma5R5HermitianLinearOperator<DomainWallFermionD,LatticeFermion> HermIndefOp(Ddwf);
 
   Level1Op LDOp(*Coarse5d,*Coarse5dRB,1); LDOp.CoarsenOperator(FGrid,HermIndefOp,Aggregates);
 
@@ -351,16 +351,16 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
   std::cout<<GridLogMessage << "Building 3 level Multigrid            "<< std::endl;
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
-  typedef MultiGridPreconditioner<vSpinColourVector,  vTComplex,nbasis, DomainWallFermionR,DeflatedGuesser<CoarseVector> , NormalEquations<CoarseVector> >   TwoLevelMG;
+  typedef MultiGridPreconditioner<vSpinColourVector,  vTComplex,nbasis, DomainWallFermionD,DeflatedGuesser<CoarseVector> , NormalEquations<CoarseVector> >   TwoLevelMG;
   typedef MultiGridPreconditioner<siteVector,iScalar<vTComplex>,nbasisc,Level1Op, DeflatedGuesser<CoarseCoarseVector>, NormalEquations<CoarseCoarseVector> > CoarseMG;
-  typedef MultiGridPreconditioner<vSpinColourVector,  vTComplex,nbasis, DomainWallFermionR,ZeroGuesser<CoarseVector>, LinearFunction<CoarseVector> >     ThreeLevelMG;
+  typedef MultiGridPreconditioner<vSpinColourVector,  vTComplex,nbasis, DomainWallFermionD,ZeroGuesser<CoarseVector>, LinearFunction<CoarseVector> >     ThreeLevelMG;
 
   // MultiGrid preconditioner acting on the coarse space <-> coarsecoarse space
   ChebyshevSmoother<CoarseVector,  Level1Op >        CoarseSmoother(0.1,12.0,3,L1LinOp,LDOp);
-  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother(0.5,60.0,10,HermIndefOp,Ddwf);
+  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother(0.5,60.0,10,HermIndefOp,Ddwf);
 
   //  MirsSmoother<CoarseVector,  Level1Op >        CoarseCGSmoother(0.1,0.1,4,L1LinOp,LDOp);
-  //  MirsSmoother<LatticeFermion,DomainWallFermionR> FineCGSmoother(0.0,0.01,8,HermIndefOp,Ddwf);
+  //  MirsSmoother<LatticeFermion,DomainWallFermionD> FineCGSmoother(0.0,0.01,8,HermIndefOp,Ddwf);
 
   CoarseMG Level2Precon (CoarseAggregates, L2Op,
 			 L1LinOp,LDOp,
diff --git a/tests/solver/Test_dwf_hdcr_16_rb.cc b/tests/solver/Test_dwf_hdcr_16_rb.cc
index 4682272d..ae8e7ae5 100644
--- a/tests/solver/Test_dwf_hdcr_16_rb.cc
+++ b/tests/solver/Test_dwf_hdcr_16_rb.cc
@@ -268,7 +268,7 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
   RealD mass=0.001;
   RealD M5=1.8;
-  DomainWallFermionR Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+  DomainWallFermionD Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
 
   typedef Aggregation<vSpinColourVector,vTComplex,nbasis>              Subspace;
   typedef CoarsenedMatrix<vSpinColourVector,vTComplex,nbasis>          CoarseOperator;
@@ -277,7 +277,7 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
   std::cout<<GridLogMessage << "Calling Aggregation class to build subspace" <<std::endl;
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
-  MdagMLinearOperator<DomainWallFermionR,LatticeFermion> HermDefOp(Ddwf);
+  MdagMLinearOperator<DomainWallFermionD,LatticeFermion> HermDefOp(Ddwf);
 
   Subspace Aggregates(Coarse5d,FGrid,0);
 
@@ -311,7 +311,7 @@ int main (int argc, char ** argv)
   typedef CoarsenedMatrix<vSpinColourVector,vTComplex,nbasis>    Level1Op;
   typedef CoarsenedMatrix<siteVector,iScalar<vTComplex>,nbasisc> Level2Op;
 
-  Gamma5R5HermitianLinearOperator<DomainWallFermionR,LatticeFermion> HermIndefOp(Ddwf);
+  Gamma5R5HermitianLinearOperator<DomainWallFermionD,LatticeFermion> HermIndefOp(Ddwf);
 
   
   GridRedBlackCartesian * Coarse4dRB = SpaceTimeGrid::makeFourDimRedBlackGrid(Coarse4d);
@@ -338,11 +338,11 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "Building 3 level Multigrid            "<< std::endl;
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
 
-  typedef MultiGridPreconditioner<vSpinColourVector,  vTComplex,nbasis, DomainWallFermionR,ZeroGuesser<CoarseVector> , SolverWrapper<CoarseVector> >   TwoLevelMG;
+  typedef MultiGridPreconditioner<vSpinColourVector,  vTComplex,nbasis, DomainWallFermionD,ZeroGuesser<CoarseVector> , SolverWrapper<CoarseVector> >   TwoLevelMG;
   typedef MultiGridPreconditioner<siteVector,iScalar<vTComplex>,nbasisc,Level1Op, DeflatedGuesser<CoarseCoarseVector>, NormalEquations<CoarseCoarseVector> > CoarseMG;
-  typedef MultiGridPreconditioner<vSpinColourVector,  vTComplex,nbasis, DomainWallFermionR,ZeroGuesser<CoarseVector>, LinearFunction<CoarseVector> >     ThreeLevelMG;
+  typedef MultiGridPreconditioner<vSpinColourVector,  vTComplex,nbasis, DomainWallFermionD,ZeroGuesser<CoarseVector>, LinearFunction<CoarseVector> >     ThreeLevelMG;
 
-  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother(0.5,60.0,12,HermIndefOp,Ddwf);
+  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother(0.5,60.0,12,HermIndefOp,Ddwf);
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
   std::cout<<GridLogMessage << "Calling 2 level Multigrid            "<< std::endl;
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
@@ -382,7 +382,7 @@ int main (int argc, char ** argv)
     LatticeFermion result_o(FrbGrid);
     pickCheckerboard(Odd,src_o,src);
     result_o=Zero();
-    SchurDiagMooeeOperator<DomainWallFermionR,LatticeFermion> HermOpEO(Ddwf);
+    SchurDiagMooeeOperator<DomainWallFermionD,LatticeFermion> HermOpEO(Ddwf);
     //    pCG(HermOpEO,src_o,result_o);
   
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
diff --git a/tests/solver/Test_dwf_hdcr_24_regression.cc b/tests/solver/Test_dwf_hdcr_24_regression.cc
index e6363474..88ae7fd2 100644
--- a/tests/solver/Test_dwf_hdcr_24_regression.cc
+++ b/tests/solver/Test_dwf_hdcr_24_regression.cc
@@ -250,7 +250,7 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
   RealD mass=0.00078;
   RealD M5=1.8;
-  DomainWallFermionR Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+  DomainWallFermionD Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
 
   typedef Aggregation<vSpinColourVector,vTComplex,nbasis>              Subspace;
   typedef CoarsenedMatrix<vSpinColourVector,vTComplex,nbasis>          CoarseOperator;
@@ -259,7 +259,7 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
   std::cout<<GridLogMessage << "Calling Aggregation class to build subspace" <<std::endl;
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
-  MdagMLinearOperator<DomainWallFermionR,LatticeFermion> HermDefOp(Ddwf);
+  MdagMLinearOperator<DomainWallFermionD,LatticeFermion> HermDefOp(Ddwf);
 
   Subspace Aggregates(Coarse5d,FGrid,0);
 
@@ -292,7 +292,7 @@ int main (int argc, char ** argv)
   typedef CoarsenedMatrix<vSpinColourVector,vTComplex,nbasis>    Level1Op;
   typedef CoarsenedMatrix<siteVector,iScalar<vTComplex>,nbasisc> Level2Op;
 
-  Gamma5R5HermitianLinearOperator<DomainWallFermionR,LatticeFermion> HermIndefOp(Ddwf);
+  Gamma5R5HermitianLinearOperator<DomainWallFermionD,LatticeFermion> HermIndefOp(Ddwf);
 
   
   GridRedBlackCartesian * Coarse4dRB = SpaceTimeGrid::makeFourDimRedBlackGrid(Coarse4d);
@@ -391,18 +391,18 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
   std::cout<<GridLogMessage << "Building 3 level Multigrid            "<< std::endl;
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
-  //  typedef MultiGridPreconditioner<vSpinColourVector,  vTComplex,nbasis, DomainWallFermionR,DeflatedGuesser<CoarseVector> , NormalEquations<CoarseVector> >   TwoLevelMG;
-  typedef MultiGridPreconditioner<vSpinColourVector,  vTComplex,nbasis, DomainWallFermionR,ZeroGuesser<CoarseVector> , NormalEquations<CoarseVector> >   TwoLevelMG;
+  //  typedef MultiGridPreconditioner<vSpinColourVector,  vTComplex,nbasis, DomainWallFermionD,DeflatedGuesser<CoarseVector> , NormalEquations<CoarseVector> >   TwoLevelMG;
+  typedef MultiGridPreconditioner<vSpinColourVector,  vTComplex,nbasis, DomainWallFermionD,ZeroGuesser<CoarseVector> , NormalEquations<CoarseVector> >   TwoLevelMG;
   typedef MultiGridPreconditioner<siteVector,iScalar<vTComplex>,nbasisc,Level1Op, DeflatedGuesser<CoarseCoarseVector>, NormalEquations<CoarseCoarseVector> > CoarseMG;
-  typedef MultiGridPreconditioner<vSpinColourVector,  vTComplex,nbasis, DomainWallFermionR,ZeroGuesser<CoarseVector>, LinearFunction<CoarseVector> >     ThreeLevelMG;
+  typedef MultiGridPreconditioner<vSpinColourVector,  vTComplex,nbasis, DomainWallFermionD,ZeroGuesser<CoarseVector>, LinearFunction<CoarseVector> >     ThreeLevelMG;
 
-  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother(0.25,60.0,12,HermIndefOp,Ddwf);
+  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother(0.25,60.0,12,HermIndefOp,Ddwf);
   /*
   // MultiGrid preconditioner acting on the coarse space <-> coarsecoarse space
   ChebyshevSmoother<CoarseVector,  Level1Op >        CoarseSmoother(0.1,15.0,3,L1LinOp,LDOp);
 
   //  MirsSmoother<CoarseVector,  Level1Op >        CoarseCGSmoother(0.1,0.1,4,L1LinOp,LDOp);
-  //  MirsSmoother<LatticeFermion,DomainWallFermionR> FineCGSmoother(0.0,0.01,8,HermIndefOp,Ddwf);
+  //  MirsSmoother<LatticeFermion,DomainWallFermionD> FineCGSmoother(0.0,0.01,8,HermIndefOp,Ddwf);
 
   CoarseMG Level2Precon (CoarseAggregates, L2Op,
 			 L1LinOp,LDOp,
@@ -463,7 +463,7 @@ int main (int argc, char ** argv)
     LatticeFermion result_o(FrbGrid);
     pickCheckerboard(Odd,src_o,src);
     result_o=Zero();
-    SchurDiagMooeeOperator<DomainWallFermionR,LatticeFermion> HermOpEO(Ddwf);
+    SchurDiagMooeeOperator<DomainWallFermionD,LatticeFermion> HermOpEO(Ddwf);
     //    pCG(HermOpEO,src_o,result_o);
   
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
diff --git a/tests/solver/Test_dwf_hdcr_2level.cc b/tests/solver/Test_dwf_hdcr_2level.cc
index 819f3464..47e129f3 100644
--- a/tests/solver/Test_dwf_hdcr_2level.cc
+++ b/tests/solver/Test_dwf_hdcr_2level.cc
@@ -288,7 +288,7 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
   RealD mass=0.001;
   RealD M5=1.8;
-  DomainWallFermionR Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+  DomainWallFermionD Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
 
   typedef Aggregation<vSpinColourVector,vTComplex,nbasis>              Subspace;
   typedef CoarsenedMatrix<vSpinColourVector,vTComplex,nbasis>          CoarseOperator;
@@ -297,7 +297,7 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
   std::cout<<GridLogMessage << "Calling Aggregation class to build subspace" <<std::endl;
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
-  MdagMLinearOperator<DomainWallFermionR,LatticeFermion> HermDefOp(Ddwf);
+  MdagMLinearOperator<DomainWallFermionD,LatticeFermion> HermDefOp(Ddwf);
 
   Subspace Aggregates(Coarse5d,FGrid,0);
 
@@ -332,7 +332,7 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
   typedef CoarsenedMatrix<vSpinColourVector,vTComplex,nbasis>    Level1Op;
 
-  Gamma5R5HermitianLinearOperator<DomainWallFermionR,LatticeFermion> HermIndefOp(Ddwf);
+  Gamma5R5HermitianLinearOperator<DomainWallFermionD,LatticeFermion> HermIndefOp(Ddwf);
 
   Level1Op LDOp(*Coarse5d,*Coarse5dRB,1); LDOp.CoarsenOperator(FGrid,HermIndefOp,Aggregates);
 
@@ -375,21 +375,21 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
   std::cout<<GridLogMessage << "Building 2 level Multigrid            "<< std::endl;
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
-  typedef MultiGridPreconditioner<vSpinColourVector,  vTComplex,nbasis, DomainWallFermionR,DeflatedGuesser<CoarseVector> , NormalEquations<CoarseVector> >   TwoLevelMG;
+  typedef MultiGridPreconditioner<vSpinColourVector,  vTComplex,nbasis, DomainWallFermionD,DeflatedGuesser<CoarseVector> , NormalEquations<CoarseVector> >   TwoLevelMG;
 
   // MultiGrid preconditioner acting on the coarse space <-> coarsecoarse space
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother(0.5,60.0,14,HermIndefOp,Ddwf); // 72 iter 63s
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother(0.1,60.0,20,HermIndefOp,Ddwf); // 66 iter 69s
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother(0.5,60.0,20,HermIndefOp,Ddwf); // 63 iter 65  s
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother(1.0,60.0,20,HermIndefOp,Ddwf); // 69, 70
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother(1.0,60.0,14,HermIndefOp,Ddwf); // 77
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother(0.5,60.0,14,HermIndefOp,Ddwf); // 72 iter 63s
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother(0.1,60.0,20,HermIndefOp,Ddwf); // 66 iter 69s
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother(0.5,60.0,20,HermIndefOp,Ddwf); // 63 iter 65  s
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother(1.0,60.0,20,HermIndefOp,Ddwf); // 69, 70
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother(1.0,60.0,14,HermIndefOp,Ddwf); // 77
 
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother(0.5,60.0,10,HermIndefOp,Ddwf); // 23 iter 15.9s
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother(0.5,60.0,14,HermIndefOp,Ddwf); // 20, 16.9s
-  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother(0.5,60.0,12,HermIndefOp,Ddwf); // 21, 15.6s
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother(0.5,60.0,10,HermIndefOp,Ddwf); // 23 iter 15.9s
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother(0.5,60.0,14,HermIndefOp,Ddwf); // 20, 16.9s
+  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother(0.5,60.0,12,HermIndefOp,Ddwf); // 21, 15.6s
 
-  //  MirsSmoother<LatticeFermion,DomainWallFermionR> FineCGSmoother(0.05,0.01,20,HermIndefOp,Ddwf);
-  //  RedBlackSmoother<LatticeFermion,DomainWallFermionR> FineRBSmoother(0.00,0.001,100,Ddwf);
+  //  MirsSmoother<LatticeFermion,DomainWallFermionD> FineCGSmoother(0.05,0.01,20,HermIndefOp,Ddwf);
+  //  RedBlackSmoother<LatticeFermion,DomainWallFermionD> FineRBSmoother(0.00,0.001,100,Ddwf);
 
   // Wrap the 2nd level solver in a MultiGrid preconditioner acting on the fine space
   //  ZeroGuesser<CoarseVector> CoarseZeroGuesser;
@@ -416,7 +416,7 @@ int main (int argc, char ** argv)
 
 
   ConjugateGradient<LatticeFermion>          FineCG(1.0e-8,10000);
-  SchurDiagMooeeOperator<DomainWallFermionR,LatticeFermion> FineDiagMooee(Ddwf); //  M_ee - Meo Moo^-1 Moe 
+  SchurDiagMooeeOperator<DomainWallFermionD,LatticeFermion> FineDiagMooee(Ddwf); //  M_ee - Meo Moo^-1 Moe 
   LatticeFermion f_src_e(FrbGrid); f_src_e=1.0;
   LatticeFermion f_res_e(FrbGrid); f_res_e=Zero();
   FineCG(FineDiagMooee,f_src_e,f_res_e);
diff --git a/tests/solver/Test_dwf_hdcr_48_rb.cc b/tests/solver/Test_dwf_hdcr_48_rb.cc
index 2b76681e..25ac1dac 100644
--- a/tests/solver/Test_dwf_hdcr_48_rb.cc
+++ b/tests/solver/Test_dwf_hdcr_48_rb.cc
@@ -264,7 +264,7 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
   RealD mass=0.00078;
   RealD M5=1.8;
-  DomainWallFermionR Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+  DomainWallFermionD Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
 
   typedef Aggregation<vSpinColourVector,vTComplex,nbasis>              Subspace;
   typedef CoarsenedMatrix<vSpinColourVector,vTComplex,nbasis>          CoarseOperator;
@@ -273,7 +273,7 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
   std::cout<<GridLogMessage << "Calling Aggregation class to build subspace" <<std::endl;
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
-  MdagMLinearOperator<DomainWallFermionR,LatticeFermion> HermDefOp(Ddwf);
+  MdagMLinearOperator<DomainWallFermionD,LatticeFermion> HermDefOp(Ddwf);
 
   Subspace Aggregates(Coarse5d,FGrid,0);
 
@@ -306,7 +306,7 @@ int main (int argc, char ** argv)
   typedef CoarsenedMatrix<vSpinColourVector,vTComplex,nbasis>    Level1Op;
   typedef CoarsenedMatrix<siteVector,iScalar<vTComplex>,nbasisc> Level2Op;
 
-  Gamma5R5HermitianLinearOperator<DomainWallFermionR,LatticeFermion> HermIndefOp(Ddwf);
+  Gamma5R5HermitianLinearOperator<DomainWallFermionD,LatticeFermion> HermIndefOp(Ddwf);
 
   
   GridRedBlackCartesian * Coarse4dRB = SpaceTimeGrid::makeFourDimRedBlackGrid(Coarse4d);
@@ -332,9 +332,9 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "Building 3 level Multigrid            "<< std::endl;
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
 
-  typedef MultiGridPreconditioner<vSpinColourVector,  vTComplex,nbasis, DomainWallFermionR,ZeroGuesser<CoarseVector> , SolverWrapper<CoarseVector> >   TwoLevelMG;
+  typedef MultiGridPreconditioner<vSpinColourVector,  vTComplex,nbasis, DomainWallFermionD,ZeroGuesser<CoarseVector> , SolverWrapper<CoarseVector> >   TwoLevelMG;
   typedef MultiGridPreconditioner<siteVector,iScalar<vTComplex>,nbasisc,Level1Op, DeflatedGuesser<CoarseCoarseVector>, NormalEquations<CoarseCoarseVector> > CoarseMG;
-  typedef MultiGridPreconditioner<vSpinColourVector,  vTComplex,nbasis, DomainWallFermionR,ZeroGuesser<CoarseVector>, LinearFunction<CoarseVector> >     ThreeLevelMG;
+  typedef MultiGridPreconditioner<vSpinColourVector,  vTComplex,nbasis, DomainWallFermionD,ZeroGuesser<CoarseVector>, LinearFunction<CoarseVector> >     ThreeLevelMG;
 
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
   std::cout<<GridLogMessage << "Calling 2 level Multigrid            "<< std::endl;
@@ -349,7 +349,7 @@ int main (int argc, char ** argv)
     result=Zero();
 
     std::cout << GridLogMessage <<" tol  " << tols[t] << " cheby order " <<ords[o]<< " lo "<<los[l] <<std::endl;
-    ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother(los[l],60.0,ords[o],HermIndefOp,Ddwf);
+    ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother(los[l],60.0,ords[o],HermIndefOp,Ddwf);
     ZeroGuesser<CoarseVector> CoarseZeroGuesser;
     ConjugateGradient<CoarseVector>  CoarseCG(tols[t],10000);
     SchurRedBlackDiagMooeeSolve<CoarseVector> CoarseRBCG(CoarseCG);
@@ -376,7 +376,7 @@ int main (int argc, char ** argv)
     LatticeFermion result_o(FrbGrid);
     pickCheckerboard(Odd,src_o,src);
     result_o=Zero();
-    SchurDiagMooeeOperator<DomainWallFermionR,LatticeFermion> HermOpEO(Ddwf);
+    SchurDiagMooeeOperator<DomainWallFermionD,LatticeFermion> HermOpEO(Ddwf);
     pCG(HermOpEO,src_o,result_o);
 
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
diff --git a/tests/solver/Test_dwf_hdcr_48_regression.cc b/tests/solver/Test_dwf_hdcr_48_regression.cc
index 4dbb097b..8c56c8f9 100644
--- a/tests/solver/Test_dwf_hdcr_48_regression.cc
+++ b/tests/solver/Test_dwf_hdcr_48_regression.cc
@@ -248,7 +248,7 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
   RealD mass=0.00078;
   RealD M5=1.8;
-  DomainWallFermionR Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+  DomainWallFermionD Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
 
   typedef Aggregation<vSpinColourVector,vTComplex,nbasis>              Subspace;
   typedef CoarsenedMatrix<vSpinColourVector,vTComplex,nbasis>          CoarseOperator;
@@ -257,7 +257,7 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
   std::cout<<GridLogMessage << "Calling Aggregation class to build subspace" <<std::endl;
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
-  MdagMLinearOperator<DomainWallFermionR,LatticeFermion> HermDefOp(Ddwf);
+  MdagMLinearOperator<DomainWallFermionD,LatticeFermion> HermDefOp(Ddwf);
 
   Subspace Aggregates(Coarse5d,FGrid,0);
 
@@ -290,7 +290,7 @@ int main (int argc, char ** argv)
   typedef CoarsenedMatrix<vSpinColourVector,vTComplex,nbasis>    Level1Op;
   typedef CoarsenedMatrix<siteVector,iScalar<vTComplex>,nbasisc> Level2Op;
 
-  Gamma5R5HermitianLinearOperator<DomainWallFermionR,LatticeFermion> HermIndefOp(Ddwf);
+  Gamma5R5HermitianLinearOperator<DomainWallFermionD,LatticeFermion> HermIndefOp(Ddwf);
 
   
   GridRedBlackCartesian * Coarse4dRB = SpaceTimeGrid::makeFourDimRedBlackGrid(Coarse4d);
@@ -386,18 +386,18 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
   std::cout<<GridLogMessage << "Building 3 level Multigrid            "<< std::endl;
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
-  //  typedef MultiGridPreconditioner<vSpinColourVector,  vTComplex,nbasis, DomainWallFermionR,DeflatedGuesser<CoarseVector> , NormalEquations<CoarseVector> >   TwoLevelMG;
-  typedef MultiGridPreconditioner<vSpinColourVector,  vTComplex,nbasis, DomainWallFermionR,ZeroGuesser<CoarseVector> , NormalEquations<CoarseVector> >   TwoLevelMG;
+  //  typedef MultiGridPreconditioner<vSpinColourVector,  vTComplex,nbasis, DomainWallFermionD,DeflatedGuesser<CoarseVector> , NormalEquations<CoarseVector> >   TwoLevelMG;
+  typedef MultiGridPreconditioner<vSpinColourVector,  vTComplex,nbasis, DomainWallFermionD,ZeroGuesser<CoarseVector> , NormalEquations<CoarseVector> >   TwoLevelMG;
   typedef MultiGridPreconditioner<siteVector,iScalar<vTComplex>,nbasisc,Level1Op, DeflatedGuesser<CoarseCoarseVector>, NormalEquations<CoarseCoarseVector> > CoarseMG;
-  typedef MultiGridPreconditioner<vSpinColourVector,  vTComplex,nbasis, DomainWallFermionR,ZeroGuesser<CoarseVector>, LinearFunction<CoarseVector> >     ThreeLevelMG;
+  typedef MultiGridPreconditioner<vSpinColourVector,  vTComplex,nbasis, DomainWallFermionD,ZeroGuesser<CoarseVector>, LinearFunction<CoarseVector> >     ThreeLevelMG;
 
-  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother(0.25,60.0,12,HermIndefOp,Ddwf);
+  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother(0.25,60.0,12,HermIndefOp,Ddwf);
   /*
   // MultiGrid preconditioner acting on the coarse space <-> coarsecoarse space
   ChebyshevSmoother<CoarseVector,  Level1Op >        CoarseSmoother(0.1,15.0,3,L1LinOp,LDOp);
 
   //  MirsSmoother<CoarseVector,  Level1Op >        CoarseCGSmoother(0.1,0.1,4,L1LinOp,LDOp);
-  //  MirsSmoother<LatticeFermion,DomainWallFermionR> FineCGSmoother(0.0,0.01,8,HermIndefOp,Ddwf);
+  //  MirsSmoother<LatticeFermion,DomainWallFermionD> FineCGSmoother(0.0,0.01,8,HermIndefOp,Ddwf);
 
   CoarseMG Level2Precon (CoarseAggregates, L2Op,
 			 L1LinOp,LDOp,
@@ -458,7 +458,7 @@ int main (int argc, char ** argv)
     LatticeFermion result_o(FrbGrid);
     pickCheckerboard(Odd,src_o,src);
     result_o=Zero();
-    SchurDiagMooeeOperator<DomainWallFermionR,LatticeFermion> HermOpEO(Ddwf);
+    SchurDiagMooeeOperator<DomainWallFermionD,LatticeFermion> HermOpEO(Ddwf);
     pCG(HermOpEO,src_o,result_o);
   
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
diff --git a/tests/solver/Test_dwf_mrhs_cg.cc b/tests/solver/Test_dwf_mrhs_cg.cc
index 4242f64b..24ded46d 100644
--- a/tests/solver/Test_dwf_mrhs_cg.cc
+++ b/tests/solver/Test_dwf_mrhs_cg.cc
@@ -35,9 +35,9 @@ using namespace Grid;
 int main (int argc, char ** argv)
 {
 #ifdef HAVE_LIME 
-  typedef typename DomainWallFermionR::FermionField FermionField; 
-  typedef typename DomainWallFermionR::ComplexField ComplexField; 
-  typename DomainWallFermionR::ImplParams params; 
+  typedef typename DomainWallFermionD::FermionField FermionField; 
+  typedef typename DomainWallFermionD::ComplexField ComplexField; 
+  typename DomainWallFermionD::ImplParams params; 
 
   const int Ls=4;
 
@@ -194,15 +194,15 @@ int main (int argc, char ** argv)
 
   RealD mass=0.01;
   RealD M5=1.8;
-  DomainWallFermionR Dchk(Umu,*FGrid,*FrbGrid,*UGrid,*rbGrid,mass,M5);
-  DomainWallFermionR Ddwf(s_Umu,*SFGrid,*SFrbGrid,*SGrid,*SrbGrid,mass,M5);
+  DomainWallFermionD Dchk(Umu,*FGrid,*FrbGrid,*UGrid,*rbGrid,mass,M5);
+  DomainWallFermionD Ddwf(s_Umu,*SFGrid,*SFrbGrid,*SGrid,*SrbGrid,mass,M5);
 
   std::cout << GridLogMessage << "****************************************************************** "<<std::endl;
   std::cout << GridLogMessage << " Calling DWF CG "<<std::endl;
   std::cout << GridLogMessage << "****************************************************************** "<<std::endl;
 
-  MdagMLinearOperator<DomainWallFermionR,FermionField> HermOp(Ddwf);
-  MdagMLinearOperator<DomainWallFermionR,FermionField> HermOpCk(Dchk);
+  MdagMLinearOperator<DomainWallFermionD,FermionField> HermOp(Ddwf);
+  MdagMLinearOperator<DomainWallFermionD,FermionField> HermOpCk(Dchk);
   ConjugateGradient<FermionField> CG((1.0e-5/(me+1)),10000);
   s_res = Zero();
   CG(HermOp,s_src,s_res);
diff --git a/tests/solver/Test_dwf_mrhs_cg_mpi.cc b/tests/solver/Test_dwf_mrhs_cg_mpi.cc
index d0a32460..1a679f45 100644
--- a/tests/solver/Test_dwf_mrhs_cg_mpi.cc
+++ b/tests/solver/Test_dwf_mrhs_cg_mpi.cc
@@ -34,9 +34,9 @@ using namespace Grid;
 
 int main (int argc, char ** argv)
 {
-  typedef typename DomainWallFermionR::FermionField FermionField; 
-  typedef typename DomainWallFermionR::ComplexField ComplexField; 
-  typename DomainWallFermionR::ImplParams params; 
+  typedef typename DomainWallFermionD::FermionField FermionField; 
+  typedef typename DomainWallFermionD::ComplexField ComplexField; 
+  typename DomainWallFermionD::ImplParams params; 
 
   double stp=1.0e-5;
   const int Ls=4;
@@ -189,15 +189,15 @@ int main (int argc, char ** argv)
   std::cout << GridLogMessage << " Building the solvers"<<std::endl;
   RealD mass=0.01;
   RealD M5=1.8;
-  DomainWallFermionR Dchk(Umu,*FGrid,*FrbGrid,*UGrid,*rbGrid,mass,M5);
-  DomainWallFermionR Ddwf(s_Umu,*SFGrid,*SFrbGrid,*SGrid,*SrbGrid,mass,M5);
+  DomainWallFermionD Dchk(Umu,*FGrid,*FrbGrid,*UGrid,*rbGrid,mass,M5);
+  DomainWallFermionD Ddwf(s_Umu,*SFGrid,*SFrbGrid,*SGrid,*SrbGrid,mass,M5);
 
   std::cout << GridLogMessage << "****************************************************************** "<<std::endl;
   std::cout << GridLogMessage << " Calling DWF CG "<<std::endl;
   std::cout << GridLogMessage << "****************************************************************** "<<std::endl;
 
-  MdagMLinearOperator<DomainWallFermionR,FermionField> HermOp(Ddwf);
-  MdagMLinearOperator<DomainWallFermionR,FermionField> HermOpCk(Dchk);
+  MdagMLinearOperator<DomainWallFermionD,FermionField> HermOp(Ddwf);
+  MdagMLinearOperator<DomainWallFermionD,FermionField> HermOpCk(Dchk);
   ConjugateGradient<FermionField> CG((stp),10000);
   s_res = Zero();
   CG(HermOp,s_src,s_res);
diff --git a/tests/solver/Test_dwf_mrhs_cg_mpieo.cc b/tests/solver/Test_dwf_mrhs_cg_mpieo.cc
index 1906619f..0dcab577 100644
--- a/tests/solver/Test_dwf_mrhs_cg_mpieo.cc
+++ b/tests/solver/Test_dwf_mrhs_cg_mpieo.cc
@@ -34,9 +34,9 @@ using namespace Grid;
 
 int main (int argc, char ** argv)
 {
-  typedef typename DomainWallFermionR::FermionField FermionField; 
-  typedef typename DomainWallFermionR::ComplexField ComplexField; 
-  typename DomainWallFermionR::ImplParams params; 
+  typedef typename DomainWallFermionD::FermionField FermionField; 
+  typedef typename DomainWallFermionD::ComplexField ComplexField; 
+  typename DomainWallFermionD::ImplParams params; 
 
   const int Ls=4;
 
@@ -124,15 +124,15 @@ int main (int argc, char ** argv)
   ///////////////////////////////////////////////////////////////
   RealD mass=0.01;
   RealD M5=1.8;
-  DomainWallFermionR Dchk(Umu,*FGrid,*FrbGrid,*UGrid,*rbGrid,mass,M5);
-  DomainWallFermionR Ddwf(s_Umu,*SFGrid,*SFrbGrid,*SGrid,*SrbGrid,mass,M5);
+  DomainWallFermionD Dchk(Umu,*FGrid,*FrbGrid,*UGrid,*rbGrid,mass,M5);
+  DomainWallFermionD Ddwf(s_Umu,*SFGrid,*SFrbGrid,*SGrid,*SrbGrid,mass,M5);
 
   std::cout << GridLogMessage << "****************************************************************** "<<std::endl;
   std::cout << GridLogMessage << " Calling DWF CG "<<std::endl;
   std::cout << GridLogMessage << "****************************************************************** "<<std::endl;
 
-  MdagMLinearOperator<DomainWallFermionR,FermionField> HermOp(Ddwf);
-  MdagMLinearOperator<DomainWallFermionR,FermionField> HermOpCk(Dchk);
+  MdagMLinearOperator<DomainWallFermionD,FermionField> HermOp(Ddwf);
+  MdagMLinearOperator<DomainWallFermionD,FermionField> HermOpCk(Dchk);
   ConjugateGradient<FermionField> CG((1.0e-8/(me+1)),10000);
   s_res = Zero();
   CG(HermOp,s_src,s_res);
diff --git a/tests/solver/Test_dwf_multigrid.cc b/tests/solver/Test_dwf_multigrid.cc
index e670b358..1cd83375 100644
--- a/tests/solver/Test_dwf_multigrid.cc
+++ b/tests/solver/Test_dwf_multigrid.cc
@@ -397,8 +397,8 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
   RealD mass=0.001;
   RealD M5=1.8;
-  DomainWallFermionR Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
-  DomainWallFermionR Dpv (Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,1.0,M5);
+  DomainWallFermionD Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+  DomainWallFermionD Dpv (Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,1.0,M5);
 
   typedef Aggregation<vSpinColourVector,vTComplex,nbasis>              Subspace;
   typedef CoarsenedMatrix<vSpinColourVector,vTComplex,nbasis>          CoarseOperator;
@@ -407,7 +407,7 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
   std::cout<<GridLogMessage << "Calling Aggregation class to build subspace" <<std::endl;
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
-  MdagMLinearOperator<DomainWallFermionR,LatticeFermion> HermDefOp(Ddwf);
+  MdagMLinearOperator<DomainWallFermionD,LatticeFermion> HermDefOp(Ddwf);
 
   Subspace Aggregates(Coarse5d,FGrid,0);
 
@@ -435,8 +435,8 @@ int main (int argc, char ** argv)
   typedef CoarsenedMatrix<vSpinColourVector,vTComplex,nbasis>    Level1Op;
   typedef CoarsenedMatrix<siteVector,iScalar<vTComplex>,nbasisc> Level2Op;
 
-  Gamma5R5HermitianLinearOperator<DomainWallFermionR,LatticeFermion> HermIndefOp(Ddwf);
-  Gamma5R5HermitianLinearOperator<DomainWallFermionR,LatticeFermion> HermIndefOpPV(Dpv);
+  Gamma5R5HermitianLinearOperator<DomainWallFermionD,LatticeFermion> HermIndefOp(Ddwf);
+  Gamma5R5HermitianLinearOperator<DomainWallFermionD,LatticeFermion> HermIndefOpPV(Dpv);
 
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
   std::cout<<GridLogMessage << "Building coarse representation of Indef operator" <<std::endl;
@@ -470,10 +470,10 @@ int main (int argc, char ** argv)
   ConjugateGradient<LatticeFermion>          FineCG(tol,MaxIt);
   //  GeneralisedMinimalResidual<LatticeFermion> FineGMRES(tol,MaxIt,20);
   
-  MdagMLinearOperator<DomainWallFermionR,LatticeFermion>    FineMdagM(Ddwf);     //  M^\dag M
-  PVdagMLinearOperator<DomainWallFermionR,LatticeFermion>   FinePVdagM(Ddwf,Dpv);//  M_{pv}^\dag M
-  SchurDiagMooeeOperator<DomainWallFermionR,LatticeFermion> FineDiagMooee(Ddwf); //  M_ee - Meo Moo^-1 Moe 
-  SchurDiagOneOperator<DomainWallFermionR,LatticeFermion>   FineDiagOne(Ddwf);   //  1 - M_ee^{-1} Meo Moo^{-1} Moe e
+  MdagMLinearOperator<DomainWallFermionD,LatticeFermion>    FineMdagM(Ddwf);     //  M^\dag M
+  PVdagMLinearOperator<DomainWallFermionD,LatticeFermion>   FinePVdagM(Ddwf,Dpv);//  M_{pv}^\dag M
+  SchurDiagMooeeOperator<DomainWallFermionD,LatticeFermion> FineDiagMooee(Ddwf); //  M_ee - Meo Moo^-1 Moe 
+  SchurDiagOneOperator<DomainWallFermionD,LatticeFermion>   FineDiagOne(Ddwf);   //  1 - M_ee^{-1} Meo Moo^{-1} Moe e
 
    MdagMLinearOperator<Level1Op,CoarseVector> CoarseMdagM(LDOp);
   PVdagMLinearOperator<Level1Op,CoarseVector> CoarsePVdagM(LDOp,LDOpPV);
@@ -552,7 +552,7 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
 
   ConjugateGradient<CoarseVector>  CoarseMgridCG(0.001,1000);     
-  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother(0.5,60.0,10,HermIndefOp,Ddwf);
+  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother(0.5,60.0,10,HermIndefOp,Ddwf);
 
   typedef HDCRPreconditioner<vSpinColourVector,  vTComplex,nbasis, NormalEquations<CoarseVector> >   TwoLevelHDCR;
   TwoLevelHDCR TwoLevelPrecon(Aggregates,
diff --git a/tests/solver/Test_dwf_qmr_unprec.cc b/tests/solver/Test_dwf_qmr_unprec.cc
index 370e7409..eeb20c95 100644
--- a/tests/solver/Test_dwf_qmr_unprec.cc
+++ b/tests/solver/Test_dwf_qmr_unprec.cc
@@ -66,17 +66,17 @@ int main (int argc, char ** argv)
   
   RealD mass=0.0;
   RealD M5=-1.8;
-  DomainWallFermionR Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+  DomainWallFermionD Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
 
-  Gamma5R5HermitianLinearOperator<DomainWallFermionR,LatticeFermion> g5HermOp(Ddwf);
+  Gamma5R5HermitianLinearOperator<DomainWallFermionD,LatticeFermion> g5HermOp(Ddwf);
   QMR(g5HermOp,src,result);
   GMR(g5HermOp,src,result);
 
-  NonHermitianLinearOperator<DomainWallFermionR,LatticeFermion> NonHermOp(Ddwf);
+  NonHermitianLinearOperator<DomainWallFermionD,LatticeFermion> NonHermOp(Ddwf);
   QMR(NonHermOp,src,result);
   GMR(NonHermOp,src,result);
 
-  MdagMLinearOperator<DomainWallFermionR,LatticeFermion> HermOp(Ddwf);
+  MdagMLinearOperator<DomainWallFermionD,LatticeFermion> HermOp(Ddwf);
   ConjugateGradient<LatticeFermion> CG(1.0e-8,10000);
   CG(HermOp,src,result);
 
diff --git a/tests/solver/Test_eofa_inv.cc b/tests/solver/Test_eofa_inv.cc
index 564405c2..71952b97 100644
--- a/tests/solver/Test_eofa_inv.cc
+++ b/tests/solver/Test_eofa_inv.cc
@@ -83,8 +83,8 @@ int main (int argc, char** argv)
   RealD mf = 0.01;
   RealD mb = 1.0;
   RealD M5 = 1.8;
-  MobiusEOFAFermionR Lop(U, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mf, mf, mb, 0.0, -1, M5, b, c);
-  MobiusEOFAFermionR Rop(U, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mb, mf, mb, -1.0, 1, M5, b, c);
+  MobiusEOFAFermionD Lop(U, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mf, mf, mb, 0.0, -1, M5, b, c);
+  MobiusEOFAFermionD Rop(U, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mb, mf, mb, -1.0, 1, M5, b, c);
   OneFlavourRationalParams Params(0.95, 100.0, 5000, 1.0e-10, 12);
   ConjugateGradient<LatticeFermion> CG(1.0e-10, 5000);
   ExactOneFlavourRatioPseudoFermionAction<WilsonImplR> Meofa(Lop, Rop, CG, CG, CG, CG, CG, Params, false);
diff --git a/tests/solver/Test_hw_multigrid.cc b/tests/solver/Test_hw_multigrid.cc
index 66c88883..fd30bca7 100644
--- a/tests/solver/Test_hw_multigrid.cc
+++ b/tests/solver/Test_hw_multigrid.cc
@@ -292,9 +292,9 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
   RealD mass=0.001;
   RealD M5=1.8;
-  WilsonFermionR    Dw(Umu,*UGrid,*UrbGrid,-M5);
-  DomainWallFermionR Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
-  DomainWallFermionR Dpv (Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,1.0,M5);
+  WilsonFermionD    Dw(Umu,*UGrid,*UrbGrid,-M5);
+  DomainWallFermionD Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+  DomainWallFermionD Dpv (Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,1.0,M5);
 
   typedef Aggregation<vSpinColourVector,vTComplex,nbasis>              Subspace;
   typedef CoarsenedMatrix<vSpinColourVector,vTComplex,nbasis>          CoarseOperator;
@@ -304,7 +304,7 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
   std::cout<<GridLogMessage << "Calling Aggregation class to build subspace" <<std::endl;
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
-  MdagMLinearOperator<WilsonFermionR,LatticeFermion> SubspaceOp(Dw);
+  MdagMLinearOperator<WilsonFermionD,LatticeFermion> SubspaceOp(Dw);
 
   Subspace Aggregates4D(Coarse4d,UGrid,0);
   Subspace Aggregates5D(Coarse5d,FGrid,0);
@@ -335,7 +335,7 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
   typedef CoarsenedMatrix<vSpinColourVector,vTComplex,nbasis>    Level1Op;
 
-  NonHermitianLinearOperator<DomainWallFermionR,LatticeFermion>  LinOpDwf(Ddwf);
+  NonHermitianLinearOperator<DomainWallFermionD,LatticeFermion>  LinOpDwf(Ddwf);
 
   Level1Op LDOp  (*Coarse5d,*Coarse5dRB,0);   
   
diff --git a/tests/solver/Test_hw_multigrid_mixed_48.cc b/tests/solver/Test_hw_multigrid_mixed_48.cc
index 0e8d6a17..3a31ddbe 100644
--- a/tests/solver/Test_hw_multigrid_mixed_48.cc
+++ b/tests/solver/Test_hw_multigrid_mixed_48.cc
@@ -395,7 +395,7 @@ public:
   Geometry  geom;
   GridBase *Coarse5D;
   GridBase *Coarse4D;
-  CartesianStencil<siteVector,siteVector,int> Stencil; 
+  CartesianStencil<siteVector,siteVector,DefaultImplParams> Stencil; 
   CoarsenedMatrix<Fobj,CComplex,nbasis> &Dw;
 
   GridBase * Grid(void)         { return Coarse5D; };   // this is all the linalg routines need to know
@@ -409,7 +409,7 @@ public:
     Coarse5D(&CoarseGrid5),
     Dw(_Dw),
     geom(CoarseGrid5._ndimension),
-    Stencil( &CoarseGrid5,geom.npoint,Even,geom.directions,geom.displacements,0)
+    Stencil( &CoarseGrid5,geom.npoint,Even,geom.directions,geom.displacements,DefaultImplParams())
   { 
   };
 
@@ -981,9 +981,9 @@ int main (int argc, char ** argv)
 
   RealD mass=0.00078;
 
-  WilsonFermionR     Dw(Umu,*UGrid,*UrbGrid,-M5);
-  DomainWallFermionR Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
-  DomainWallFermionR Dpv (Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,1.0,M5);
+  WilsonFermionD     Dw(Umu,*UGrid,*UrbGrid,-M5);
+  DomainWallFermionD Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+  DomainWallFermionD Dpv (Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,1.0,M5);
 
   typedef Aggregation<vSpinColourVector,vTComplex,nbasis>              Subspace;
   typedef CoarsenedMatrix<vSpinColourVector,vTComplex,nbasis>          CoarseOperator;
@@ -994,21 +994,21 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "Calling Aggregation class to build subspace" <<std::endl;
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
   // How to find criticall mass?
-  // WilsonFermionR     Dw_null(Umu,*UGrid,*UrbGrid,-0.75); //   600 iters
-  // WilsonFermionR     Dw_null(Umu,*UGrid,*UrbGrid,-0.80); //   800 iters
-  //  WilsonFermionR     Dw_null(Umu,*UGrid,*UrbGrid,-0.82); // 1023 iters
-  //  WilsonFermionR     Dw_null(Umu,*UGrid,*UrbGrid,-0.85); // 1428 iters
-  //  WilsonFermionR     Dw_null(Umu,*UGrid,*UrbGrid,-0.87); //  1900 iters
-  //  WilsonFermionR     Dw_null(Umu,*UGrid,*UrbGrid,-0.90); // 3900   iters
-  //  WilsonFermionR     Dw_null(Umu,*UGrid,*UrbGrid,-0.92); // 6200   iters
-  //  WilsonFermionR     Dw_null(Umu,*UGrid,*UrbGrid,-0.94);  // 8882 iters
-  WilsonFermionR     Dw_null(Umu,*UGrid,*UrbGrid,-0.95);  // 9170  iters
-  //  WilsonFermionR     Dw_null(Umu,*UGrid,*UrbGrid,-0.96);  // 8882   iters
-  //  WilsonFermionR     Dw_null(Umu,*UGrid,*UrbGrid,-0.97);  // 8406  iters
-  //  WilsonFermionR     Dw_null(Umu,*UGrid,*UrbGrid,-0.99); // 6900   iters
-  //  WilsonFermionR     Dw_null(Umu,*UGrid,*UrbGrid,-1.01); // 6397   iters
-  //  WilsonFermionR     Dw_null(Umu,*UGrid,*UrbGrid,-1.00); // 5900   iters
-  MdagMLinearOperator<WilsonFermionR,LatticeFermion> MdagM_Dw(Dw_null);
+  // WilsonFermionD     Dw_null(Umu,*UGrid,*UrbGrid,-0.75); //   600 iters
+  // WilsonFermionD     Dw_null(Umu,*UGrid,*UrbGrid,-0.80); //   800 iters
+  //  WilsonFermionD     Dw_null(Umu,*UGrid,*UrbGrid,-0.82); // 1023 iters
+  //  WilsonFermionD     Dw_null(Umu,*UGrid,*UrbGrid,-0.85); // 1428 iters
+  //  WilsonFermionD     Dw_null(Umu,*UGrid,*UrbGrid,-0.87); //  1900 iters
+  //  WilsonFermionD     Dw_null(Umu,*UGrid,*UrbGrid,-0.90); // 3900   iters
+  //  WilsonFermionD     Dw_null(Umu,*UGrid,*UrbGrid,-0.92); // 6200   iters
+  //  WilsonFermionD     Dw_null(Umu,*UGrid,*UrbGrid,-0.94);  // 8882 iters
+  WilsonFermionD     Dw_null(Umu,*UGrid,*UrbGrid,-0.95);  // 9170  iters
+  //  WilsonFermionD     Dw_null(Umu,*UGrid,*UrbGrid,-0.96);  // 8882   iters
+  //  WilsonFermionD     Dw_null(Umu,*UGrid,*UrbGrid,-0.97);  // 8406  iters
+  //  WilsonFermionD     Dw_null(Umu,*UGrid,*UrbGrid,-0.99); // 6900   iters
+  //  WilsonFermionD     Dw_null(Umu,*UGrid,*UrbGrid,-1.01); // 6397   iters
+  //  WilsonFermionD     Dw_null(Umu,*UGrid,*UrbGrid,-1.00); // 5900   iters
+  MdagMLinearOperator<WilsonFermionD,LatticeFermion> MdagM_Dw(Dw_null);
 
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
   std::cout<<GridLogMessage << "Testing Wilson criticality " <<std::endl;
@@ -1057,7 +1057,7 @@ int main (int argc, char ** argv)
   typedef CoarsenedMatrix<vSpinColourVector,vTComplex,nbasis>    Level1Op4;
   typedef CoarseCayleyFermion<vSpinColourVector,vTComplex,nbasis> Level1Op5;
   Level1Op4 c_Dw    (*Coarse4d,0);
-  NonHermitianLinearOperator<WilsonFermionR,LatticeFermion>  LinOpDw(Dw);
+  NonHermitianLinearOperator<WilsonFermionD,LatticeFermion>  LinOpDw(Dw);
   c_Dw.CoarsenOperator(UGrid,LinOpDw,Aggregates4D); // contains the M5 from Dw(-M5)
   //  c_Dw.Test(Aggregates4D,UGrid,LinOpDw);
 
@@ -1127,8 +1127,8 @@ int main (int argc, char ** argv)
   ConjugateGradient<CoarseVector>            CoarseCG(tol,MaxIt);
   ConjugateGradient<LatticeFermion>          FineCG(tol,MaxIt);
   
-  NonHermitianLinearOperator<DomainWallFermionR,LatticeFermion> FineM(Ddwf);
-  MdagMLinearOperator<DomainWallFermionR,LatticeFermion>    FineMdagM(Ddwf);     //  M^\dag M
+  NonHermitianLinearOperator<DomainWallFermionD,LatticeFermion> FineM(Ddwf);
+  MdagMLinearOperator<DomainWallFermionD,LatticeFermion>    FineMdagM(Ddwf);     //  M^\dag M
 
   NonHermitianLinearOperator<Level1Op5,CoarseVector> CoarseM(c_Dwf);
   MdagMLinearOperator<Level1Op5,CoarseVector> CoarseMdagM(c_Dwf);
@@ -1233,39 +1233,39 @@ typedef HDCRPreconditioner<siteVector,iScalar<vTComplex>,nbasisc,NormalEquations
 
   // Wrap the 2nd level solver in a MultiGrid preconditioner acting on the fine space
 
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother1(0.5,60.0,14,FineM,Ddwf); // 26 iter, 39s
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother2(0.5,60.0,14,FineM,Ddwf);
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother1(0.5,60.0,14,FineM,Ddwf); // 26 iter, 39s
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother2(0.5,60.0,14,FineM,Ddwf);
 
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother1(0.5,60.0,12,FineM,Ddwf); // 25 iter, 38s
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother2(0.5,60.0,16,FineM,Ddwf);
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother1(0.5,60.0,12,FineM,Ddwf); // 25 iter, 38s
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother2(0.5,60.0,16,FineM,Ddwf);
 
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother1(0.5,60.0,12,FineM,Ddwf); // 23 iter, 39s
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother2(0.5,60.0,20,FineM,Ddwf);
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother1(0.5,60.0,12,FineM,Ddwf); // 23 iter, 39s
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother2(0.5,60.0,20,FineM,Ddwf);
 
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother1(0.5,60.0,10,FineM,Ddwf);24 iter, 44s
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother2(0.5,60.0,24,FineM,Ddwf);
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother1(0.5,60.0,10,FineM,Ddwf);24 iter, 44s
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother2(0.5,60.0,24,FineM,Ddwf);
 
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother1(0.5,60.0,12,FineM,Ddwf); // odd convergence tail at 10^-9 ish
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother2(0.1,60.0,24,FineM,Ddwf); // 33 iter, waas O(10-9 by 26)
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother1(0.5,60.0,12,FineM,Ddwf); // odd convergence tail at 10^-9 ish
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother2(0.1,60.0,24,FineM,Ddwf); // 33 iter, waas O(10-9 by 26)
 
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother1(0.5,60.0,12,FineM,Ddwf); // 25 iter, 39s
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother2(0.5,60.0,18,FineM,Ddwf); //
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother1(0.5,60.0,12,FineM,Ddwf); // 25 iter, 39s
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother2(0.5,60.0,18,FineM,Ddwf); //
 
-  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother1(0.5,60.0,16,FineM,Ddwf); 
-  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother2(0.5,60.0,16,FineM,Ddwf); //
+  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother1(0.5,60.0,16,FineM,Ddwf); 
+  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother2(0.5,60.0,16,FineM,Ddwf); //
 
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother1(0.5,60.0,11,FineM,Ddwf); // 33 iter, 49s
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother2(0.5,60.0,11,FineM,Ddwf);
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother1(0.5,60.0,12,FineM,Ddwf); // 26 iter, 37s
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother2(0.5,60.0,12,FineM,Ddwf);
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother1(0.4,60.0,12,FineM,Ddwf); //  iter 26 no change in final residual
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother2(0.4,60.0,12,FineM,Ddwf);
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother1(0.3,60.0,12,FineM,Ddwf); // 27 iter 39s.
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother2(0.3,60.0,12,FineM,Ddwf);
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother1(0.3,60.0,13,FineM,Ddwf); // 26 iter, but slower
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother2(0.3,60.0,13,FineM,Ddwf);
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother1(1.0,60.0,12,FineM,Ddwf); // 34 iter, slower
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother2(1.0,60.0,12,FineM,Ddwf);
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother1(0.5,60.0,11,FineM,Ddwf); // 33 iter, 49s
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother2(0.5,60.0,11,FineM,Ddwf);
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother1(0.5,60.0,12,FineM,Ddwf); // 26 iter, 37s
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother2(0.5,60.0,12,FineM,Ddwf);
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother1(0.4,60.0,12,FineM,Ddwf); //  iter 26 no change in final residual
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother2(0.4,60.0,12,FineM,Ddwf);
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother1(0.3,60.0,12,FineM,Ddwf); // 27 iter 39s.
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother2(0.3,60.0,12,FineM,Ddwf);
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother1(0.3,60.0,13,FineM,Ddwf); // 26 iter, but slower
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother2(0.3,60.0,13,FineM,Ddwf);
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother1(1.0,60.0,12,FineM,Ddwf); // 34 iter, slower
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother2(1.0,60.0,12,FineM,Ddwf);
 
   ThreeLevelMG ThreeLevelPrecon(Aggregates4D,
 				FineM,
diff --git a/tests/solver/Test_hw_multigrid_mixed_48_rb.cc b/tests/solver/Test_hw_multigrid_mixed_48_rb.cc
index e7ceb022..0f18893e 100644
--- a/tests/solver/Test_hw_multigrid_mixed_48_rb.cc
+++ b/tests/solver/Test_hw_multigrid_mixed_48_rb.cc
@@ -395,7 +395,7 @@ public:
   Geometry  geom;
   GridBase *Coarse5D;
   GridBase *Coarse4D;
-  CartesianStencil<siteVector,siteVector,int> Stencil; 
+  CartesianStencil<siteVector,siteVector,DefaultImplParams> Stencil; 
   CoarsenedMatrix<Fobj,CComplex,nbasis> &Dw;
 
   GridBase * Grid(void)         { return Coarse5D; };   // this is all the linalg routines need to know
@@ -409,7 +409,7 @@ public:
     Coarse5D(&CoarseGrid5),
     Dw(_Dw),
     geom(CoarseGrid5._ndimension),
-    Stencil( &CoarseGrid5,geom.npoint,Even,geom.directions,geom.displacements,0)
+    Stencil( &CoarseGrid5,geom.npoint,Even,geom.directions,geom.displacements,DefaultImplParams())
   { 
   };
 
@@ -1005,9 +1005,9 @@ int main (int argc, char ** argv)
 
   RealD mass=0.00078;
 
-  WilsonFermionR     Dw(Umu,*UGrid,*UrbGrid,-M5);
-  DomainWallFermionR Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
-  DomainWallFermionR Dpv (Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,1.0,M5);
+  WilsonFermionD     Dw(Umu,*UGrid,*UrbGrid,-M5);
+  DomainWallFermionD Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+  DomainWallFermionD Dpv (Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,1.0,M5);
 
   typedef Aggregation<vSpinColourVector,vTComplex,nbasis>              Subspace;
   typedef CoarsenedMatrix<vSpinColourVector,vTComplex,nbasis>          CoarseOperator;
@@ -1018,21 +1018,21 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "Calling Aggregation class to build subspace" <<std::endl;
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
   // How to find criticall mass?
-  // WilsonFermionR     Dw_null(Umu,*UGrid,*UrbGrid,-0.75); //   600 iters
-  // WilsonFermionR     Dw_null(Umu,*UGrid,*UrbGrid,-0.80); //   800 iters
-  //  WilsonFermionR     Dw_null(Umu,*UGrid,*UrbGrid,-0.82); // 1023 iters
-  //  WilsonFermionR     Dw_null(Umu,*UGrid,*UrbGrid,-0.85); // 1428 iters
-  //  WilsonFermionR     Dw_null(Umu,*UGrid,*UrbGrid,-0.87); //  1900 iters
-  //  WilsonFermionR     Dw_null(Umu,*UGrid,*UrbGrid,-0.90); // 3900   iters
-  //  WilsonFermionR     Dw_null(Umu,*UGrid,*UrbGrid,-0.92); // 6200   iters
-  //  WilsonFermionR     Dw_null(Umu,*UGrid,*UrbGrid,-0.94);  // 8882 iters
-  WilsonFermionR     Dw_null(Umu,*UGrid,*UrbGrid,-0.95);  // 9170  iters
-  //  WilsonFermionR     Dw_null(Umu,*UGrid,*UrbGrid,-0.96);  // 8882   iters
-  //  WilsonFermionR     Dw_null(Umu,*UGrid,*UrbGrid,-0.97);  // 8406  iters
-  //  WilsonFermionR     Dw_null(Umu,*UGrid,*UrbGrid,-0.99); // 6900   iters
-  //  WilsonFermionR     Dw_null(Umu,*UGrid,*UrbGrid,-1.01); // 6397   iters
-  //  WilsonFermionR     Dw_null(Umu,*UGrid,*UrbGrid,-1.00); // 5900   iters
-  MdagMLinearOperator<WilsonFermionR,LatticeFermion> MdagM_Dw(Dw_null);
+  // WilsonFermionD     Dw_null(Umu,*UGrid,*UrbGrid,-0.75); //   600 iters
+  // WilsonFermionD     Dw_null(Umu,*UGrid,*UrbGrid,-0.80); //   800 iters
+  //  WilsonFermionD     Dw_null(Umu,*UGrid,*UrbGrid,-0.82); // 1023 iters
+  //  WilsonFermionD     Dw_null(Umu,*UGrid,*UrbGrid,-0.85); // 1428 iters
+  //  WilsonFermionD     Dw_null(Umu,*UGrid,*UrbGrid,-0.87); //  1900 iters
+  //  WilsonFermionD     Dw_null(Umu,*UGrid,*UrbGrid,-0.90); // 3900   iters
+  //  WilsonFermionD     Dw_null(Umu,*UGrid,*UrbGrid,-0.92); // 6200   iters
+  //  WilsonFermionD     Dw_null(Umu,*UGrid,*UrbGrid,-0.94);  // 8882 iters
+  WilsonFermionD     Dw_null(Umu,*UGrid,*UrbGrid,-0.95);  // 9170  iters
+  //  WilsonFermionD     Dw_null(Umu,*UGrid,*UrbGrid,-0.96);  // 8882   iters
+  //  WilsonFermionD     Dw_null(Umu,*UGrid,*UrbGrid,-0.97);  // 8406  iters
+  //  WilsonFermionD     Dw_null(Umu,*UGrid,*UrbGrid,-0.99); // 6900   iters
+  //  WilsonFermionD     Dw_null(Umu,*UGrid,*UrbGrid,-1.01); // 6397   iters
+  //  WilsonFermionD     Dw_null(Umu,*UGrid,*UrbGrid,-1.00); // 5900   iters
+  MdagMLinearOperator<WilsonFermionD,LatticeFermion> MdagM_Dw(Dw_null);
 
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
   std::cout<<GridLogMessage << "Testing Wilson criticality " <<std::endl;
@@ -1081,7 +1081,7 @@ int main (int argc, char ** argv)
   typedef CoarsenedMatrix<vSpinColourVector,vTComplex,nbasis>    Level1Op4;
   typedef CoarseCayleyFermion<vSpinColourVector,vTComplex,nbasis> Level1Op5;
   Level1Op4 c_Dw    (*Coarse4d,0);
-  NonHermitianLinearOperator<WilsonFermionR,LatticeFermion>  LinOpDw(Dw);
+  NonHermitianLinearOperator<WilsonFermionD,LatticeFermion>  LinOpDw(Dw);
   c_Dw.CoarsenOperator(UGrid,LinOpDw,Aggregates4D); // contains the M5 from Dw(-M5)
   //  c_Dw.Test(Aggregates4D,UGrid,LinOpDw);
 
@@ -1148,8 +1148,8 @@ int main (int argc, char ** argv)
   ConjugateGradient<CoarseVector>            CoarseCG(tol,MaxIt);
   ConjugateGradient<LatticeFermion>          FineCG(tol,MaxIt);
   
-  NonHermitianLinearOperator<DomainWallFermionR,LatticeFermion> FineM(Ddwf);
-  MdagMLinearOperator<DomainWallFermionR,LatticeFermion>    FineMdagM(Ddwf);     //  M^\dag M
+  NonHermitianLinearOperator<DomainWallFermionD,LatticeFermion> FineM(Ddwf);
+  MdagMLinearOperator<DomainWallFermionD,LatticeFermion>    FineMdagM(Ddwf);     //  M^\dag M
 
   NonHermitianLinearOperator<Level1Op5,CoarseVector> CoarseM(c_Dwf);
   MdagMLinearOperator<Level1Op5,CoarseVector> CoarseMdagM(c_Dwf);
@@ -1272,38 +1272,38 @@ typedef HDCRPreconditioner<siteVector,iScalar<vTComplex>,nbasisc,LinearFunction<
 
   // Wrap the 2nd level solver in a MultiGrid preconditioner acting on the fine space
 
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother1(0.5,60.0,14,FineM,Ddwf); // 26 iter, 39s
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother2(0.5,60.0,14,FineM,Ddwf);
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother1(0.5,60.0,14,FineM,Ddwf); // 26 iter, 39s
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother2(0.5,60.0,14,FineM,Ddwf);
 
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother1(0.5,60.0,12,FineM,Ddwf); // 25 iter, 38s
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother2(0.5,60.0,16,FineM,Ddwf);
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother1(0.5,60.0,12,FineM,Ddwf); // 25 iter, 38s
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother2(0.5,60.0,16,FineM,Ddwf);
 
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother1(0.5,60.0,12,FineM,Ddwf); // 23 iter, 39s
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother2(0.5,60.0,20,FineM,Ddwf);
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother1(0.5,60.0,12,FineM,Ddwf); // 23 iter, 39s
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother2(0.5,60.0,20,FineM,Ddwf);
 
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother1(0.5,60.0,10,FineM,Ddwf);24 iter, 44s
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother2(0.5,60.0,24,FineM,Ddwf);
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother1(0.5,60.0,10,FineM,Ddwf);24 iter, 44s
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother2(0.5,60.0,24,FineM,Ddwf);
 
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother1(0.5,60.0,12,FineM,Ddwf); // odd convergence tail at 10^-9 ish
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother2(0.1,60.0,24,FineM,Ddwf); // 33 iter, waas O(10-9 by 26)
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother1(0.5,60.0,12,FineM,Ddwf); // odd convergence tail at 10^-9 ish
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother2(0.1,60.0,24,FineM,Ddwf); // 33 iter, waas O(10-9 by 26)
 
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother1(0.5,60.0,12,FineM,Ddwf); // 25 iter, 39s
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother2(0.5,60.0,18,FineM,Ddwf); //
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother1(0.5,60.0,12,FineM,Ddwf); // 25 iter, 39s
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother2(0.5,60.0,18,FineM,Ddwf); //
 
-  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother(f_lo,f_hi,f_ord,FineM,Ddwf); 
+  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother(f_lo,f_hi,f_ord,FineM,Ddwf); 
 
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother1(0.5,60.0,11,FineM,Ddwf); // 33 iter, 49s
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother2(0.5,60.0,11,FineM,Ddwf);
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother1(0.5,60.0,12,FineM,Ddwf); // 26 iter, 37s
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother2(0.5,60.0,12,FineM,Ddwf);
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother1(0.4,60.0,12,FineM,Ddwf); //  iter 26 no change in final residual
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother2(0.4,60.0,12,FineM,Ddwf);
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother1(0.3,60.0,12,FineM,Ddwf); // 27 iter 39s.
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother2(0.3,60.0,12,FineM,Ddwf);
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother1(0.3,60.0,13,FineM,Ddwf); // 26 iter, but slower
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother2(0.3,60.0,13,FineM,Ddwf);
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother1(1.0,60.0,12,FineM,Ddwf); // 34 iter, slower
-  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionR> FineSmoother2(1.0,60.0,12,FineM,Ddwf);
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother1(0.5,60.0,11,FineM,Ddwf); // 33 iter, 49s
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother2(0.5,60.0,11,FineM,Ddwf);
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother1(0.5,60.0,12,FineM,Ddwf); // 26 iter, 37s
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother2(0.5,60.0,12,FineM,Ddwf);
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother1(0.4,60.0,12,FineM,Ddwf); //  iter 26 no change in final residual
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother2(0.4,60.0,12,FineM,Ddwf);
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother1(0.3,60.0,12,FineM,Ddwf); // 27 iter 39s.
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother2(0.3,60.0,12,FineM,Ddwf);
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother1(0.3,60.0,13,FineM,Ddwf); // 26 iter, but slower
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother2(0.3,60.0,13,FineM,Ddwf);
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother1(1.0,60.0,12,FineM,Ddwf); // 34 iter, slower
+  //  ChebyshevSmoother<LatticeFermion,DomainWallFermionD> FineSmoother2(1.0,60.0,12,FineM,Ddwf);
 
   ThreeLevelMG ThreeLevelPrecon(Aggregates4D,
 				FineM,
diff --git a/tests/solver/Test_mobius_bcg.cc b/tests/solver/Test_mobius_bcg.cc
index 8092d61c..a54b4a05 100644
--- a/tests/solver/Test_mobius_bcg.cc
+++ b/tests/solver/Test_mobius_bcg.cc
@@ -33,9 +33,9 @@ using namespace Grid;
 
 int main (int argc, char ** argv)
 {
-  typedef typename MobiusFermionR::FermionField FermionField; 
-  typedef typename MobiusFermionR::ComplexField ComplexField; 
-  typename MobiusFermionR::ImplParams params; 
+  typedef typename MobiusFermionD::FermionField FermionField; 
+  typedef typename MobiusFermionD::ComplexField ComplexField; 
+  typename MobiusFermionD::ImplParams params; 
 
   const int Ls=12;
 
@@ -158,15 +158,15 @@ int main (int argc, char ** argv)
   RealD mobius_factor=32./12.;
   RealD mobius_b=0.5*(mobius_factor+1.);
   RealD mobius_c=0.5*(mobius_factor-1.);
-  MobiusFermionR Dchk(Umu,*FGrid,*FrbGrid,*UGrid,*rbGrid,mass,M5,mobius_b,mobius_c,params);
-  MobiusFermionR Ddwf(s_Umu,*SFGrid,*SFrbGrid,*SGrid,*SrbGrid,mass,M5,mobius_b,mobius_c,params);
+  MobiusFermionD Dchk(Umu,*FGrid,*FrbGrid,*UGrid,*rbGrid,mass,M5,mobius_b,mobius_c,params);
+  MobiusFermionD Ddwf(s_Umu,*SFGrid,*SFrbGrid,*SGrid,*SrbGrid,mass,M5,mobius_b,mobius_c,params);
 
   std::cout << GridLogMessage << "****************************************************************** "<<std::endl;
   std::cout << GridLogMessage << " Calling DWF CG "<<std::endl;
   std::cout << GridLogMessage << "****************************************************************** "<<std::endl;
 
-  MdagMLinearOperator<MobiusFermionR,FermionField> HermOp(Ddwf);
-  MdagMLinearOperator<MobiusFermionR,FermionField> HermOpCk(Dchk);
+  MdagMLinearOperator<MobiusFermionD,FermionField> HermOp(Ddwf);
+  MdagMLinearOperator<MobiusFermionD,FermionField> HermOpCk(Dchk);
   ConjugateGradient<FermionField> CG((stp),100000);
   s_res = Zero();
 
diff --git a/tests/solver/Test_mobius_bcg_nosplit.cc b/tests/solver/Test_mobius_bcg_nosplit.cc
index de02b1e3..f33a40ea 100644
--- a/tests/solver/Test_mobius_bcg_nosplit.cc
+++ b/tests/solver/Test_mobius_bcg_nosplit.cc
@@ -35,9 +35,9 @@ using namespace Grid;
 
 int main (int argc, char ** argv)
 {
-  typedef typename DomainWallFermionR::FermionField FermionField; 
-  typedef typename DomainWallFermionR::ComplexField ComplexField; 
-  typename DomainWallFermionR::ImplParams params; 
+  typedef typename DomainWallFermionD::FermionField FermionField; 
+  typedef typename DomainWallFermionD::ComplexField ComplexField; 
+  typename DomainWallFermionD::ImplParams params; 
 
   const int Ls=16;
 
@@ -106,13 +106,13 @@ int main (int argc, char ** argv)
   std::cout << GridLogMessage << " Building the solvers"<<std::endl;
   RealD mass=0.01;
   RealD M5=1.8;
-  DomainWallFermionR Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*rbGrid,mass,M5,params);
+  DomainWallFermionD Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*rbGrid,mass,M5,params);
 
   std::cout << GridLogMessage << "****************************************************************** "<<std::endl;
   std::cout << GridLogMessage << " Calling DWF CG "<<std::endl;
   std::cout << GridLogMessage << "****************************************************************** "<<std::endl;
 
-  MdagMLinearOperator<DomainWallFermionR,FermionField> HermOp(Ddwf);
+  MdagMLinearOperator<DomainWallFermionD,FermionField> HermOp(Ddwf);
   ConjugateGradient<FermionField> CG((stp),100000);
 
   for(int rhs=0;rhs<1;rhs++){
diff --git a/tests/solver/Test_mobius_bcg_phys_nosplit.cc b/tests/solver/Test_mobius_bcg_phys_nosplit.cc
index 2fe573ce..76a6f7e1 100644
--- a/tests/solver/Test_mobius_bcg_phys_nosplit.cc
+++ b/tests/solver/Test_mobius_bcg_phys_nosplit.cc
@@ -35,9 +35,9 @@ using namespace Grid;
 
 int main (int argc, char ** argv)
 {
-  typedef typename DomainWallFermionR::FermionField FermionField; 
-  typedef typename DomainWallFermionR::ComplexField ComplexField; 
-  typename DomainWallFermionR::ImplParams params; 
+  typedef typename DomainWallFermionD::FermionField FermionField; 
+  typedef typename DomainWallFermionD::ComplexField ComplexField; 
+  typename DomainWallFermionD::ImplParams params; 
 
   const int Ls=16;
 
@@ -107,7 +107,7 @@ int main (int argc, char ** argv)
   std::cout << GridLogMessage << " Building the solvers"<<std::endl;
   RealD mass=0.01;
   RealD M5=1.8;
-  DomainWallFermionR Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*rbGrid,mass,M5,params);
+  DomainWallFermionD Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*rbGrid,mass,M5,params);
   for(int s=0;s<nrhs;s++) {
     Ddwf.ImportPhysicalFermionSource(src4[s],src[s]);
   }
@@ -116,7 +116,7 @@ int main (int argc, char ** argv)
   std::cout << GridLogMessage << " Calling DWF CG "<<std::endl;
   std::cout << GridLogMessage << "****************************************************************** "<<std::endl;
 
-  MdagMLinearOperator<DomainWallFermionR,FermionField> HermOp(Ddwf);
+  MdagMLinearOperator<DomainWallFermionD,FermionField> HermOp(Ddwf);
   ConjugateGradient<FermionField> CG((stp),100000);
 
   for(int rhs=0;rhs<1;rhs++){
diff --git a/tests/solver/Test_mobius_bcg_prec_nosplit.cc b/tests/solver/Test_mobius_bcg_prec_nosplit.cc
index 3ac0d42b..f793893c 100644
--- a/tests/solver/Test_mobius_bcg_prec_nosplit.cc
+++ b/tests/solver/Test_mobius_bcg_prec_nosplit.cc
@@ -35,9 +35,9 @@ using namespace Grid;
 
 int main (int argc, char ** argv)
 {
-  typedef typename DomainWallFermionR::FermionField FermionField; 
-  typedef typename DomainWallFermionR::ComplexField ComplexField; 
-  typename DomainWallFermionR::ImplParams params; 
+  typedef typename DomainWallFermionD::FermionField FermionField; 
+  typedef typename DomainWallFermionD::ComplexField ComplexField; 
+  typename DomainWallFermionD::ImplParams params; 
 
   const int Ls=16;
 
@@ -106,13 +106,13 @@ int main (int argc, char ** argv)
   std::cout << GridLogMessage << " Building the solvers"<<std::endl;
   RealD mass=0.01;
   RealD M5=1.8;
-  DomainWallFermionR Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*rbGrid,mass,M5,params);
+  DomainWallFermionD Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*rbGrid,mass,M5,params);
 
   std::cout << GridLogMessage << "****************************************************************** "<<std::endl;
   std::cout << GridLogMessage << " Calling DWF CG "<<std::endl;
   std::cout << GridLogMessage << "****************************************************************** "<<std::endl;
 
-  MdagMLinearOperator<DomainWallFermionR,FermionField> HermOp(Ddwf);
+  MdagMLinearOperator<DomainWallFermionD,FermionField> HermOp(Ddwf);
   ConjugateGradient<FermionField> CG((stp),100000);
 
   for(int rhs=0;rhs<1;rhs++){
diff --git a/tests/solver/Test_split_grid.cc b/tests/solver/Test_split_grid.cc
index 85626c8e..39441c82 100644
--- a/tests/solver/Test_split_grid.cc
+++ b/tests/solver/Test_split_grid.cc
@@ -34,9 +34,9 @@ using namespace Grid;
 
 int main (int argc, char ** argv)
 {
-  typedef typename DomainWallFermionR::FermionField FermionField; 
-  typedef typename DomainWallFermionR::ComplexField ComplexField; 
-  typename DomainWallFermionR::ImplParams params; 
+  typedef typename DomainWallFermionD::FermionField FermionField; 
+  typedef typename DomainWallFermionD::ComplexField ComplexField; 
+  typename DomainWallFermionD::ImplParams params; 
 
   const int Ls=4;
 
@@ -117,15 +117,15 @@ int main (int argc, char ** argv)
   ///////////////////////////////////////////////////////////////
   RealD mass=0.01;
   RealD M5=1.8;
-  DomainWallFermionR Dchk(Umu,*FGrid,*FrbGrid,*UGrid,*rbGrid,mass,M5);
-  DomainWallFermionR Ddwf(s_Umu,*SFGrid,*SFrbGrid,*SGrid,*SrbGrid,mass,M5);
+  DomainWallFermionD Dchk(Umu,*FGrid,*FrbGrid,*UGrid,*rbGrid,mass,M5);
+  DomainWallFermionD Ddwf(s_Umu,*SFGrid,*SFrbGrid,*SGrid,*SrbGrid,mass,M5);
 
   std::cout << GridLogMessage << "****************************************************************** "<<std::endl;
   std::cout << GridLogMessage << " Calling DWF CG "<<std::endl;
   std::cout << GridLogMessage << "****************************************************************** "<<std::endl;
 
-  MdagMLinearOperator<DomainWallFermionR,FermionField> HermOp(Ddwf);
-  MdagMLinearOperator<DomainWallFermionR,FermionField> HermOpCk(Dchk);
+  MdagMLinearOperator<DomainWallFermionD,FermionField> HermOp(Ddwf);
+  MdagMLinearOperator<DomainWallFermionD,FermionField> HermOpCk(Dchk);
   ConjugateGradient<FermionField> CG((1.0e-8/(me+1)),10000);
   s_res = Zero();
   CG(HermOp,s_src,s_res);
diff --git a/tests/solver/Test_staggered_block_cg_prec.cc b/tests/solver/Test_staggered_block_cg_prec.cc
index c5306e85..4cb7801e 100644
--- a/tests/solver/Test_staggered_block_cg_prec.cc
+++ b/tests/solver/Test_staggered_block_cg_prec.cc
@@ -46,9 +46,9 @@ struct scal {
 
 int main (int argc, char ** argv)
 {
-  typedef typename ImprovedStaggeredFermion5DR::FermionField FermionField; 
-  typedef typename ImprovedStaggeredFermion5DR::ComplexField ComplexField; 
-  typename ImprovedStaggeredFermion5DR::ImplParams params; 
+  typedef typename ImprovedStaggeredFermion5DD::FermionField FermionField; 
+  typedef typename ImprovedStaggeredFermion5DD::ComplexField ComplexField; 
+  typename ImprovedStaggeredFermion5DD::ImplParams params; 
 
   const int Ls=8;
 
@@ -98,8 +98,8 @@ int main (int argc, char ** argv)
   RealD c1=9.0/8.0;
   RealD c2=-1.0/24.0;
   RealD u0=1.0;
-  ImprovedStaggeredFermion5DR Ds(Umu,Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,c1,c2,u0); 
-  SchurStaggeredOperator<ImprovedStaggeredFermion5DR,FermionField> HermOp(Ds);
+  ImprovedStaggeredFermion5DD Ds(Umu,Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,c1,c2,u0); 
+  SchurStaggeredOperator<ImprovedStaggeredFermion5DD,FermionField> HermOp(Ds);
 
   ConjugateGradient<FermionField> CG(1.0e-8,10000);
   int blockDim = 0;
@@ -111,8 +111,8 @@ int main (int argc, char ** argv)
   std::cout << GridLogMessage << "****************************************************************** "<<std::endl;
   std::cout << GridLogMessage << " Calling 4d CG "<<std::endl;
   std::cout << GridLogMessage << "****************************************************************** "<<std::endl;
-  ImprovedStaggeredFermionR Ds4d(Umu,Umu,*UGrid,*UrbGrid,mass,c1,c2,u0);
-  SchurStaggeredOperator<ImprovedStaggeredFermionR,FermionField> HermOp4d(Ds4d);
+  ImprovedStaggeredFermionD Ds4d(Umu,Umu,*UGrid,*UrbGrid,mass,c1,c2,u0);
+  SchurStaggeredOperator<ImprovedStaggeredFermionD,FermionField> HermOp4d(Ds4d);
   FermionField src4d(UGrid); random(pRNG,src4d);
   FermionField src4d_o(UrbGrid);   pickCheckerboard(Odd,src4d_o,src4d);
   FermionField result4d_o(UrbGrid); 
@@ -135,7 +135,6 @@ int main (int argc, char ** argv)
   std::cout << GridLogMessage << "************************************************************************ "<<std::endl;
   std::cout << GridLogMessage << " Calling 5d CG for "<<Ls <<" right hand sides" <<std::endl;
   std::cout << GridLogMessage << "************************************************************************ "<<std::endl;
-  Ds.ZeroCounters();
   result_o=Zero();
   {
     double t1=usecond();
@@ -152,7 +151,6 @@ int main (int argc, char ** argv)
   std::cout << GridLogMessage << "************************************************************************ "<<std::endl;
   std::cout << GridLogMessage << " Calling multiRHS CG for "<<Ls <<" right hand sides" <<std::endl;
   std::cout << GridLogMessage << "************************************************************************ "<<std::endl;
-  Ds.ZeroCounters();
   result_o=Zero();
   {
     double t1=usecond();
@@ -170,7 +168,6 @@ int main (int argc, char ** argv)
   std::cout << GridLogMessage << "************************************************************************ "<<std::endl;
   std::cout << GridLogMessage << " Calling Block CGrQ for "<<Ls <<" right hand sides" <<std::endl;
   std::cout << GridLogMessage << "************************************************************************ "<<std::endl;
-  Ds.ZeroCounters();
   result_o=Zero();
   {
     double t1=usecond();
@@ -187,7 +184,6 @@ int main (int argc, char ** argv)
   std::cout << GridLogMessage << "************************************************************************ "<<std::endl;
   std::cout << GridLogMessage << " Calling Block CG for "<<Ls <<" right hand sides" <<std::endl;
   std::cout << GridLogMessage << "************************************************************************ "<<std::endl;
-  Ds.ZeroCounters();
   result_o=Zero();
   {
     double t1=usecond();
diff --git a/tests/solver/Test_staggered_block_cg_unprec.cc b/tests/solver/Test_staggered_block_cg_unprec.cc
index a78b449a..75a00f23 100644
--- a/tests/solver/Test_staggered_block_cg_unprec.cc
+++ b/tests/solver/Test_staggered_block_cg_unprec.cc
@@ -30,12 +30,6 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
- ;
-
-template<class d>
-struct scal {
-  d internal;
-};
 
   Gamma::Algebra Gmu [] = {
     Gamma::Algebra::GammaX,
@@ -46,9 +40,9 @@ struct scal {
 
 int main (int argc, char ** argv)
 {
-  typedef typename ImprovedStaggeredFermion5DR::FermionField FermionField; 
-  typedef typename ImprovedStaggeredFermion5DR::ComplexField ComplexField; 
-  typename ImprovedStaggeredFermion5DR::ImplParams params; 
+  typedef typename ImprovedStaggeredFermion5DD::FermionField FermionField; 
+  typedef typename ImprovedStaggeredFermion5DD::ComplexField ComplexField; 
+  typename ImprovedStaggeredFermion5DD::ImplParams params; 
 
   const int Ls=8;
 
@@ -83,8 +77,8 @@ int main (int argc, char ** argv)
     volume=volume*latt_size[mu];
   }
 
-  ImprovedStaggeredFermion5DR Ds(Umu,Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,c1,c2,u0); 
-  MdagMLinearOperator<ImprovedStaggeredFermion5DR,FermionField> HermOp(Ds);
+  ImprovedStaggeredFermion5DD Ds(Umu,Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,c1,c2,u0); 
+  MdagMLinearOperator<ImprovedStaggeredFermion5DD,FermionField> HermOp(Ds);
 
   ConjugateGradient<FermionField> CG(1.0e-8,10000);
   int blockDim = 0;
@@ -95,8 +89,8 @@ int main (int argc, char ** argv)
   std::cout << GridLogMessage << "****************************************************************** "<<std::endl;
   std::cout << GridLogMessage << " Calling 4d CG "<<std::endl;
   std::cout << GridLogMessage << "****************************************************************** "<<std::endl;
-  ImprovedStaggeredFermionR Ds4d(Umu,Umu,*UGrid,*UrbGrid,mass,c1,c2,u0);
-  MdagMLinearOperator<ImprovedStaggeredFermionR,FermionField> HermOp4d(Ds4d);
+  ImprovedStaggeredFermionD Ds4d(Umu,Umu,*UGrid,*UrbGrid,mass,c1,c2,u0);
+  MdagMLinearOperator<ImprovedStaggeredFermionD,FermionField> HermOp4d(Ds4d);
   FermionField src4d(UGrid); random(pRNG,src4d);
   FermionField result4d(UGrid); result4d=Zero();
 
@@ -120,7 +114,6 @@ int main (int argc, char ** argv)
   std::cout << GridLogMessage << "************************************************************************ "<<std::endl;
   result=Zero();
 {
-  Ds.ZeroCounters();
   double t1=usecond();
   CG(HermOp,src,result);
   double t2=usecond();
@@ -129,7 +122,6 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "usec    =   "<< (t2-t1)<<std::endl;
     std::cout<<GridLogMessage << "flops   =   "<< flops<<std::endl;
     std::cout<<GridLogMessage << "mflop/s =   "<< flops/(t2-t1)<<std::endl;
-  Ds.Report();
 }
   std::cout << GridLogMessage << "************************************************************************ "<<std::endl;
 
@@ -137,7 +129,6 @@ int main (int argc, char ** argv)
   std::cout << GridLogMessage << " Calling multiRHS CG for "<<Ls <<" right hand sides" <<std::endl;
   std::cout << GridLogMessage << "************************************************************************ "<<std::endl;
   result=Zero();
-  Ds.ZeroCounters();
 {
   double t1=usecond();
   mCG(HermOp,src,result);
@@ -148,14 +139,12 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "flops   =   "<< flops<<std::endl;
   std::cout<<GridLogMessage << "mflop/s =   "<< flops/(t2-t1)<<std::endl;
 }
-  Ds.Report();
   std::cout << GridLogMessage << "************************************************************************ "<<std::endl;
 
   std::cout << GridLogMessage << "************************************************************************ "<<std::endl;
   std::cout << GridLogMessage << " Calling Block CG for "<<Ls <<" right hand sides" <<std::endl;
   std::cout << GridLogMessage << "************************************************************************ "<<std::endl;
   result=Zero();
-  Ds.ZeroCounters();
 {
   double t1=usecond();
   BCGrQ(HermOp,src,result);
@@ -166,7 +155,6 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "flops   =   "<< flops<<std::endl;
   std::cout<<GridLogMessage << "mflop/s =   "<< flops/(t2-t1)<<std::endl;
 }
-  Ds.Report();
   std::cout << GridLogMessage << "************************************************************************ "<<std::endl;
 
 
diff --git a/tests/solver/Test_staggered_cagmres_unprec.cc b/tests/solver/Test_staggered_cagmres_unprec.cc
index 1b7a2f56..dcdf086f 100644
--- a/tests/solver/Test_staggered_cagmres_unprec.cc
+++ b/tests/solver/Test_staggered_cagmres_unprec.cc
@@ -33,9 +33,9 @@ using namespace Grid;
 
 int main (int argc, char ** argv)
 {
-  typedef typename ImprovedStaggeredFermionR::FermionField FermionField;
-  typedef typename ImprovedStaggeredFermionR::ComplexField ComplexField;
-  typename ImprovedStaggeredFermionR::ImplParams params;
+  typedef typename ImprovedStaggeredFermionD::FermionField FermionField;
+  typedef typename ImprovedStaggeredFermionD::ComplexField ComplexField;
+  typename ImprovedStaggeredFermionD::ImplParams params;
 
   Grid_init(&argc,&argv);
 
@@ -62,9 +62,9 @@ int main (int argc, char ** argv)
   RealD c1=9.0/8.0;
   RealD c2=-1.0/24.0;
   RealD u0=1.0;
-  ImprovedStaggeredFermionR Ds(Umu,Umu,Grid,RBGrid,mass,c1,c2,u0);
+  ImprovedStaggeredFermionD Ds(Umu,Umu,Grid,RBGrid,mass,c1,c2,u0);
 
-  MdagMLinearOperator<ImprovedStaggeredFermionR,FermionField> HermOp(Ds);
+  MdagMLinearOperator<ImprovedStaggeredFermionD,FermionField> HermOp(Ds);
   CommunicationAvoidingGeneralisedMinimalResidual<FermionField> CAGMRES(1.0e-8, 10000, 25);
   CAGMRES(HermOp,src,result);
 
diff --git a/tests/solver/Test_staggered_cg_prec.cc b/tests/solver/Test_staggered_cg_prec.cc
index 854ef632..bc80da09 100644
--- a/tests/solver/Test_staggered_cg_prec.cc
+++ b/tests/solver/Test_staggered_cg_prec.cc
@@ -47,8 +47,8 @@ struct scal {
 
 int main (int argc, char ** argv)
 {
-  typedef typename ImprovedStaggeredFermionR::FermionField FermionField; 
-  typename ImprovedStaggeredFermionR::ImplParams params; 
+  typedef typename ImprovedStaggeredFermionD::FermionField FermionField; 
+  typename ImprovedStaggeredFermionD::ImplParams params; 
 
   Grid_init(&argc,&argv);
 
@@ -74,14 +74,14 @@ int main (int argc, char ** argv)
   RealD c1=9.0/8.0;
   RealD c2=-1.0/24.0;
   RealD u0=1.0;
-  ImprovedStaggeredFermionR Ds(Umu,Umu,Grid,RBGrid,mass,c1,c2,u0);
+  ImprovedStaggeredFermionD Ds(Umu,Umu,Grid,RBGrid,mass,c1,c2,u0);
 
   FermionField res_o(&RBGrid); 
   FermionField src_o(&RBGrid); 
   pickCheckerboard(Odd,src_o,src);
   res_o=Zero();
 
-  SchurStaggeredOperator<ImprovedStaggeredFermionR,FermionField> HermOpEO(Ds);
+  SchurStaggeredOperator<ImprovedStaggeredFermionD,FermionField> HermOpEO(Ds);
   ConjugateGradient<FermionField> CG(1.0e-8,10000);
   double t1=usecond();
   CG(HermOpEO,src_o,res_o);
diff --git a/tests/solver/Test_staggered_cg_schur.cc b/tests/solver/Test_staggered_cg_schur.cc
index d8e5bdd4..5d7d073e 100644
--- a/tests/solver/Test_staggered_cg_schur.cc
+++ b/tests/solver/Test_staggered_cg_schur.cc
@@ -45,8 +45,8 @@ struct scal {
 
 int main (int argc, char ** argv)
 {
-  typedef typename ImprovedStaggeredFermionR::FermionField FermionField; 
-  typename ImprovedStaggeredFermionR::ImplParams params; 
+  typedef typename ImprovedStaggeredFermionD::FermionField FermionField; 
+  typename ImprovedStaggeredFermionD::ImplParams params; 
   Grid_init(&argc,&argv);
 
   Coordinate latt_size   = GridDefaultLatt();
@@ -68,7 +68,7 @@ int main (int argc, char ** argv)
   RealD c1=9.0/8.0;
   RealD c2=-1.0/24.0;
   RealD u0=1.0;
-  ImprovedStaggeredFermionR Ds(Umu,Umu,Grid,RBGrid,mass,c1,c2,u0);
+  ImprovedStaggeredFermionD Ds(Umu,Umu,Grid,RBGrid,mass,c1,c2,u0);
 
   ConjugateGradient<FermionField> CG(1.0e-8,10000);
   SchurRedBlackStaggeredSolve<FermionField> SchurSolver(CG);
diff --git a/tests/solver/Test_staggered_cg_unprec.cc b/tests/solver/Test_staggered_cg_unprec.cc
index e023b910..466f1d04 100644
--- a/tests/solver/Test_staggered_cg_unprec.cc
+++ b/tests/solver/Test_staggered_cg_unprec.cc
@@ -47,9 +47,9 @@ struct scal {
 
 int main (int argc, char ** argv)
 {
-  typedef typename ImprovedStaggeredFermionR::FermionField FermionField; 
-  typedef typename ImprovedStaggeredFermionR::ComplexField ComplexField; 
-  typename ImprovedStaggeredFermionR::ImplParams params; 
+  typedef typename ImprovedStaggeredFermionD::FermionField FermionField; 
+  typedef typename ImprovedStaggeredFermionD::ComplexField ComplexField; 
+  typename ImprovedStaggeredFermionD::ImplParams params; 
 
   Grid_init(&argc,&argv);
 
@@ -76,9 +76,9 @@ int main (int argc, char ** argv)
   RealD c1=9.0/8.0;
   RealD c2=-1.0/24.0;
   RealD u0=1.0;
-  ImprovedStaggeredFermionR Ds(Umu,Umu,Grid,RBGrid,mass,c1,c2,u0);
+  ImprovedStaggeredFermionD Ds(Umu,Umu,Grid,RBGrid,mass,c1,c2,u0);
 
-  MdagMLinearOperator<ImprovedStaggeredFermionR,FermionField> HermOp(Ds);
+  MdagMLinearOperator<ImprovedStaggeredFermionD,FermionField> HermOp(Ds);
   ConjugateGradient<FermionField> CG(1.0e-6,10000);
   CG(HermOp,src,result);
 
diff --git a/tests/solver/Test_staggered_fcagmres_prec.cc b/tests/solver/Test_staggered_fcagmres_prec.cc
index 692d688e..7587748e 100644
--- a/tests/solver/Test_staggered_fcagmres_prec.cc
+++ b/tests/solver/Test_staggered_fcagmres_prec.cc
@@ -33,9 +33,9 @@ using namespace Grid;
 
 int main (int argc, char ** argv)
 {
-  typedef typename ImprovedStaggeredFermionR::FermionField FermionField;
-  typedef typename ImprovedStaggeredFermionR::ComplexField ComplexField;
-  typename ImprovedStaggeredFermionR::ImplParams params;
+  typedef typename ImprovedStaggeredFermionD::FermionField FermionField;
+  typedef typename ImprovedStaggeredFermionD::ComplexField ComplexField;
+  typename ImprovedStaggeredFermionD::ImplParams params;
 
   Grid_init(&argc,&argv);
 
@@ -62,9 +62,9 @@ int main (int argc, char ** argv)
   RealD c1=9.0/8.0;
   RealD c2=-1.0/24.0;
   RealD u0=1.0;
-  ImprovedStaggeredFermionR Ds(Umu,Umu,Grid,RBGrid,mass,c1,c2,u0);
+  ImprovedStaggeredFermionD Ds(Umu,Umu,Grid,RBGrid,mass,c1,c2,u0);
 
-  MdagMLinearOperator<ImprovedStaggeredFermionR,FermionField> HermOp(Ds);
+  MdagMLinearOperator<ImprovedStaggeredFermionD,FermionField> HermOp(Ds);
 
   TrivialPrecon<FermionField> simple;
 
diff --git a/tests/solver/Test_staggered_fgmres_prec.cc b/tests/solver/Test_staggered_fgmres_prec.cc
index fe6da67c..a3c65057 100644
--- a/tests/solver/Test_staggered_fgmres_prec.cc
+++ b/tests/solver/Test_staggered_fgmres_prec.cc
@@ -33,9 +33,9 @@ using namespace Grid;
 
 int main (int argc, char ** argv)
 {
-  typedef typename ImprovedStaggeredFermionR::FermionField FermionField;
-  typedef typename ImprovedStaggeredFermionR::ComplexField ComplexField;
-  typename ImprovedStaggeredFermionR::ImplParams params;
+  typedef typename ImprovedStaggeredFermionD::FermionField FermionField;
+  typedef typename ImprovedStaggeredFermionD::ComplexField ComplexField;
+  typename ImprovedStaggeredFermionD::ImplParams params;
 
   Grid_init(&argc,&argv);
 
@@ -62,9 +62,9 @@ int main (int argc, char ** argv)
   RealD c1=9.0/8.0;
   RealD c2=-1.0/24.0;
   RealD u0=1.0;
-  ImprovedStaggeredFermionR Ds(Umu,Umu,Grid,RBGrid,mass,c1,c2,u0);
+  ImprovedStaggeredFermionD Ds(Umu,Umu,Grid,RBGrid,mass,c1,c2,u0);
 
-  MdagMLinearOperator<ImprovedStaggeredFermionR,FermionField> HermOp(Ds);
+  MdagMLinearOperator<ImprovedStaggeredFermionD,FermionField> HermOp(Ds);
 
   TrivialPrecon<FermionField> simple;
 
diff --git a/tests/solver/Test_staggered_gmres_unprec.cc b/tests/solver/Test_staggered_gmres_unprec.cc
index ec9d4608..abfeab75 100644
--- a/tests/solver/Test_staggered_gmres_unprec.cc
+++ b/tests/solver/Test_staggered_gmres_unprec.cc
@@ -33,9 +33,9 @@ using namespace Grid;
 
 int main (int argc, char ** argv)
 {
-  typedef typename ImprovedStaggeredFermionR::FermionField FermionField;
-  typedef typename ImprovedStaggeredFermionR::ComplexField ComplexField;
-  typename ImprovedStaggeredFermionR::ImplParams params;
+  typedef typename ImprovedStaggeredFermionD::FermionField FermionField;
+  typedef typename ImprovedStaggeredFermionD::ComplexField ComplexField;
+  typename ImprovedStaggeredFermionD::ImplParams params;
 
   Grid_init(&argc,&argv);
 
@@ -62,9 +62,9 @@ int main (int argc, char ** argv)
   RealD c1=9.0/8.0;
   RealD c2=-1.0/24.0;
   RealD u0=1.0;
-  ImprovedStaggeredFermionR Ds(Umu,Umu,Grid,RBGrid,mass,c1,c2,u0);
+  ImprovedStaggeredFermionD Ds(Umu,Umu,Grid,RBGrid,mass,c1,c2,u0);
 
-  MdagMLinearOperator<ImprovedStaggeredFermionR,FermionField> HermOp(Ds);
+  MdagMLinearOperator<ImprovedStaggeredFermionD,FermionField> HermOp(Ds);
   GeneralisedMinimalResidual<FermionField> GMRES(1.0e-8, 10000, 25);
   GMRES(HermOp,src,result);
 
diff --git a/tests/solver/Test_staggered_mr_unprec.cc b/tests/solver/Test_staggered_mr_unprec.cc
index ddbb8de3..1cdd60f9 100644
--- a/tests/solver/Test_staggered_mr_unprec.cc
+++ b/tests/solver/Test_staggered_mr_unprec.cc
@@ -33,9 +33,9 @@ using namespace Grid;
 
 int main (int argc, char ** argv)
 {
-  typedef typename ImprovedStaggeredFermionR::FermionField FermionField;
-  typedef typename ImprovedStaggeredFermionR::ComplexField ComplexField;
-  typename ImprovedStaggeredFermionR::ImplParams params;
+  typedef typename ImprovedStaggeredFermionD::FermionField FermionField;
+  typedef typename ImprovedStaggeredFermionD::ComplexField ComplexField;
+  typename ImprovedStaggeredFermionD::ImplParams params;
 
   Grid_init(&argc,&argv);
 
@@ -62,9 +62,9 @@ int main (int argc, char ** argv)
   RealD c1=9.0/8.0;
   RealD c2=-1.0/24.0;
   RealD u0=1.0;
-  ImprovedStaggeredFermionR Ds(Umu,Umu,Grid,RBGrid,mass,c1,c2,u0);
+  ImprovedStaggeredFermionD Ds(Umu,Umu,Grid,RBGrid,mass,c1,c2,u0);
 
-  MdagMLinearOperator<ImprovedStaggeredFermionR,FermionField> HermOp(Ds);
+  MdagMLinearOperator<ImprovedStaggeredFermionD,FermionField> HermOp(Ds);
   MinimalResidual<FermionField> MR(1.0e-8,10000,0.8);
   MR(HermOp,src,result);
 
diff --git a/tests/solver/Test_staggered_multishift.cc b/tests/solver/Test_staggered_multishift.cc
index 856f0b87..9f6b37d6 100644
--- a/tests/solver/Test_staggered_multishift.cc
+++ b/tests/solver/Test_staggered_multishift.cc
@@ -46,8 +46,8 @@ struct scal {
 
 int main (int argc, char ** argv)
 {
-  typedef typename ImprovedStaggeredFermionR::FermionField FermionField; 
-  typename ImprovedStaggeredFermionR::ImplParams params; 
+  typedef typename ImprovedStaggeredFermionD::FermionField FermionField; 
+  typename ImprovedStaggeredFermionD::ImplParams params; 
 
   Grid_init(&argc,&argv);
 
@@ -90,8 +90,8 @@ int main (int argc, char ** argv)
   RealD c2=-1.0/24.0;
   RealD u0=1.0;
 
-  ImprovedStaggeredFermionR Ds(Umu,Umu,Grid,RBGrid,mass,c1,c2,u0);
-  SchurStaggeredOperator<ImprovedStaggeredFermionR,FermionField> HermOpEO(Ds);
+  ImprovedStaggeredFermionD Ds(Umu,Umu,Grid,RBGrid,mass,c1,c2,u0);
+  SchurStaggeredOperator<ImprovedStaggeredFermionD,FermionField> HermOpEO(Ds);
 
   FermionField src(&Grid); random(pRNG,src);
   FermionField src_o(&RBGrid); 
diff --git a/tests/solver/Test_wilson_cagmres_unprec.cc b/tests/solver/Test_wilson_cagmres_unprec.cc
index 226d0719..80381a27 100644
--- a/tests/solver/Test_wilson_cagmres_unprec.cc
+++ b/tests/solver/Test_wilson_cagmres_unprec.cc
@@ -55,9 +55,9 @@ int main (int argc, char ** argv)
   }
 
   RealD mass=0.5;
-  WilsonFermionR Dw(Umu,Grid,RBGrid,mass);
+  WilsonFermionD Dw(Umu,Grid,RBGrid,mass);
 
-  MdagMLinearOperator<WilsonFermionR,LatticeFermion> HermOp(Dw);
+  MdagMLinearOperator<WilsonFermionD,LatticeFermion> HermOp(Dw);
   CommunicationAvoidingGeneralisedMinimalResidual<LatticeFermion> CAGMRES(1.0e-8, 10000, 25);
   CAGMRES(HermOp,src,result);
 
diff --git a/tests/solver/Test_wilson_cg_prec.cc b/tests/solver/Test_wilson_cg_prec.cc
index a28e014e..cb480e8c 100644
--- a/tests/solver/Test_wilson_cg_prec.cc
+++ b/tests/solver/Test_wilson_cg_prec.cc
@@ -69,7 +69,7 @@ int main (int argc, char ** argv)
   }
   
   RealD mass=0.5;
-  WilsonFermionR Dw(Umu,Grid,RBGrid,mass);
+  WilsonFermionD Dw(Umu,Grid,RBGrid,mass);
 
   //  HermitianOperator<WilsonFermion,LatticeFermion> HermOp(Dw);
   //  ConjugateGradient<LatticeFermion> CG(1.0e-8,10000);
@@ -80,7 +80,7 @@ int main (int argc, char ** argv)
   pickCheckerboard(Odd,src_o,src);
   result_o=Zero();
 
-  SchurDiagMooeeOperator<WilsonFermionR,LatticeFermion> HermOpEO(Dw);
+  SchurDiagMooeeOperator<WilsonFermionD,LatticeFermion> HermOpEO(Dw);
   ConjugateGradient<LatticeFermion> CG(1.0e-8,10000);
   CG(HermOpEO,src_o,result_o);
   
diff --git a/tests/solver/Test_wilson_cg_schur.cc b/tests/solver/Test_wilson_cg_schur.cc
index 97482131..601eb6b2 100644
--- a/tests/solver/Test_wilson_cg_schur.cc
+++ b/tests/solver/Test_wilson_cg_schur.cc
@@ -64,7 +64,7 @@ int main (int argc, char ** argv)
   LatticeFermion resid(&Grid); 
 
   RealD mass=0.5;
-  WilsonFermionR Dw(Umu,Grid,RBGrid,mass);
+  WilsonFermionD Dw(Umu,Grid,RBGrid,mass);
 
   ConjugateGradient<LatticeFermion> CG(1.0e-8,10000);
   SchurRedBlackDiagMooeeSolve<LatticeFermion> SchurSolver(CG);
diff --git a/tests/solver/Test_wilson_cg_unprec.cc b/tests/solver/Test_wilson_cg_unprec.cc
index 07f6ba7b..f1ecebd3 100644
--- a/tests/solver/Test_wilson_cg_unprec.cc
+++ b/tests/solver/Test_wilson_cg_unprec.cc
@@ -68,9 +68,9 @@ int main (int argc, char ** argv)
   }  
   
   RealD mass=0.5;
-  WilsonFermionR Dw(Umu,Grid,RBGrid,mass);
+  WilsonFermionD Dw(Umu,Grid,RBGrid,mass);
 
-  MdagMLinearOperator<WilsonFermionR,LatticeFermion> HermOp(Dw);
+  MdagMLinearOperator<WilsonFermionD,LatticeFermion> HermOp(Dw);
   ConjugateGradient<LatticeFermion> CG(1.0e-8,10000);
   CG(HermOp,src,result);
 
diff --git a/tests/solver/Test_wilson_cr_unprec.cc b/tests/solver/Test_wilson_cr_unprec.cc
index 67510a23..a8b49afd 100644
--- a/tests/solver/Test_wilson_cr_unprec.cc
+++ b/tests/solver/Test_wilson_cr_unprec.cc
@@ -70,9 +70,9 @@ int main (int argc, char ** argv)
   }  
 
   RealD mass=0.5;
-  WilsonFermionR Dw(Umu,Grid,RBGrid,mass);
+  WilsonFermionD Dw(Umu,Grid,RBGrid,mass);
 
-  MdagMLinearOperator<WilsonFermionR,LatticeFermion> HermOp(Dw);
+  MdagMLinearOperator<WilsonFermionD,LatticeFermion> HermOp(Dw);
 
   ConjugateResidual<LatticeFermion> MCR(1.0e-8,10000);
 
diff --git a/tests/solver/Test_wilson_fcagmres_prec.cc b/tests/solver/Test_wilson_fcagmres_prec.cc
index d2a1acf4..66f9f518 100644
--- a/tests/solver/Test_wilson_fcagmres_prec.cc
+++ b/tests/solver/Test_wilson_fcagmres_prec.cc
@@ -55,9 +55,9 @@ int main (int argc, char ** argv)
   }
 
   RealD mass=0.5;
-  WilsonFermionR Dw(Umu,Grid,RBGrid,mass);
+  WilsonFermionD Dw(Umu,Grid,RBGrid,mass);
 
-  MdagMLinearOperator<WilsonFermionR,LatticeFermion> HermOp(Dw);
+  MdagMLinearOperator<WilsonFermionD,LatticeFermion> HermOp(Dw);
 
   TrivialPrecon<LatticeFermion> simple;
 
diff --git a/tests/solver/Test_wilson_fgmres_prec.cc b/tests/solver/Test_wilson_fgmres_prec.cc
index 02d8f9f2..61368636 100644
--- a/tests/solver/Test_wilson_fgmres_prec.cc
+++ b/tests/solver/Test_wilson_fgmres_prec.cc
@@ -55,9 +55,9 @@ int main (int argc, char ** argv)
   }
 
   RealD mass=0.5;
-  WilsonFermionR Dw(Umu,Grid,RBGrid,mass);
+  WilsonFermionD Dw(Umu,Grid,RBGrid,mass);
 
-  MdagMLinearOperator<WilsonFermionR,LatticeFermion> HermOp(Dw);
+  MdagMLinearOperator<WilsonFermionD,LatticeFermion> HermOp(Dw);
 
   TrivialPrecon<LatticeFermion> simple;
 
diff --git a/tests/solver/Test_wilson_gmres_unprec.cc b/tests/solver/Test_wilson_gmres_unprec.cc
index e52c047f..5f2728ce 100644
--- a/tests/solver/Test_wilson_gmres_unprec.cc
+++ b/tests/solver/Test_wilson_gmres_unprec.cc
@@ -55,9 +55,9 @@ int main (int argc, char ** argv)
   }
 
   RealD mass=0.5;
-  WilsonFermionR Dw(Umu,Grid,RBGrid,mass);
+  WilsonFermionD Dw(Umu,Grid,RBGrid,mass);
 
-  MdagMLinearOperator<WilsonFermionR,LatticeFermion> HermOp(Dw);
+  MdagMLinearOperator<WilsonFermionD,LatticeFermion> HermOp(Dw);
   GeneralisedMinimalResidual<LatticeFermion> GMRES(1.0e-8, 10000, 25);
   GMRES(HermOp,src,result);
 
diff --git a/tests/solver/Test_wilson_mg.cc b/tests/solver/Test_wilson_mg.cc
index 99a16e19..875bf32a 100644
--- a/tests/solver/Test_wilson_mg.cc
+++ b/tests/solver/Test_wilson_mg.cc
@@ -77,16 +77,16 @@ int main(int argc, char **argv) {
   // Note: We do chiral doubling, so actually only nbasis/2 full basis vectors are used
   const int nbasis = 40;
 
-  WilsonFermionR Dw(Umu, *FGrid, *FrbGrid, mass);
+  WilsonFermionD Dw(Umu, *FGrid, *FrbGrid, mass);
 
-  MdagMLinearOperator<WilsonFermionR, LatticeFermion> MdagMOpDw(Dw);
+  MdagMLinearOperator<WilsonFermionD, LatticeFermion> MdagMOpDw(Dw);
 
   std::cout << GridLogMessage << "**************************************************" << std::endl;
   std::cout << GridLogMessage << "Testing Multigrid for Wilson" << std::endl;
   std::cout << GridLogMessage << "**************************************************" << std::endl;
 
   TrivialPrecon<LatticeFermion> TrivialPrecon;
-  auto MGPreconDw = createMGInstance<vSpinColourVector, vTComplex, nbasis, WilsonFermionR>(mgParams, levelInfo, Dw, Dw);
+  auto MGPreconDw = createMGInstance<vSpinColourVector, vTComplex, nbasis, WilsonFermionD>(mgParams, levelInfo, Dw, Dw);
 
   MGPreconDw->setup();
 
diff --git a/tests/solver/Test_wilson_mr_unprec.cc b/tests/solver/Test_wilson_mr_unprec.cc
index fef83794..c71392e4 100644
--- a/tests/solver/Test_wilson_mr_unprec.cc
+++ b/tests/solver/Test_wilson_mr_unprec.cc
@@ -55,9 +55,9 @@ int main (int argc, char ** argv)
   }
 
   RealD mass=0.5;
-  WilsonFermionR Dw(Umu,Grid,RBGrid,mass);
+  WilsonFermionD Dw(Umu,Grid,RBGrid,mass);
 
-  MdagMLinearOperator<WilsonFermionR,LatticeFermion> HermOp(Dw);
+  MdagMLinearOperator<WilsonFermionD,LatticeFermion> HermOp(Dw);
   MinimalResidual<LatticeFermion> MR(1.0e-8,10000,0.8);
   MR(HermOp,src,result);
 
diff --git a/tests/solver/Test_wilson_qmr_unprec.cc b/tests/solver/Test_wilson_qmr_unprec.cc
index c0b42a28..0cd132e4 100644
--- a/tests/solver/Test_wilson_qmr_unprec.cc
+++ b/tests/solver/Test_wilson_qmr_unprec.cc
@@ -56,9 +56,9 @@ int main (int argc, char ** argv)
   QuasiMinimalResidual<LatticeFermion> QMR(1.0e-8,10000);
   
   RealD mass=0.0;
-  WilsonFermionR Dw(Umu,*Grid,*rbGrid,mass);
+  WilsonFermionD Dw(Umu,*Grid,*rbGrid,mass);
 
-  NonHermitianLinearOperator<WilsonFermionR,LatticeFermion> NonHermOp(Dw);
+  NonHermitianLinearOperator<WilsonFermionD,LatticeFermion> NonHermOp(Dw);
   QMR(NonHermOp,src,result);
 
   Grid_finalize();
diff --git a/tests/solver/Test_wilsonclover_bicgstab_prec.cc b/tests/solver/Test_wilsonclover_bicgstab_prec.cc
index b382b1bb..d265e687 100644
--- a/tests/solver/Test_wilsonclover_bicgstab_prec.cc
+++ b/tests/solver/Test_wilsonclover_bicgstab_prec.cc
@@ -70,14 +70,14 @@ int main (int argc, char ** argv)
   RealD mass = -0.1;
   RealD csw_r = 1.0;
   RealD csw_t = 1.0;
-  WilsonCloverFermionR Dw(Umu, Grid, RBGrid, mass, csw_r, csw_t);
+  WilsonCloverFermionD Dw(Umu, Grid, RBGrid, mass, csw_r, csw_t);
 
   LatticeFermion    src_o(&RBGrid);
   LatticeFermion result_o(&RBGrid);
   pickCheckerboard(Odd, src_o, src);
   result_o = Zero();
 
-  NonHermitianSchurDiagMooeeOperator<WilsonCloverFermionR,LatticeFermion> HermOp(Dw);
+  NonHermitianSchurDiagMooeeOperator<WilsonCloverFermionD,LatticeFermion> HermOp(Dw);
   BiCGSTAB<LatticeFermion> CG(1.0e-8,10000);
   CG(HermOp, src_o, result_o);
 
diff --git a/tests/solver/Test_wilsonclover_bicgstab_schur.cc b/tests/solver/Test_wilsonclover_bicgstab_schur.cc
index f09d7cd1..38bfdb72 100644
--- a/tests/solver/Test_wilsonclover_bicgstab_schur.cc
+++ b/tests/solver/Test_wilsonclover_bicgstab_schur.cc
@@ -70,7 +70,7 @@ int main (int argc, char ** argv)
   RealD mass = -0.1;
   RealD csw_r = 1.0;
   RealD csw_t = 1.0;
-  WilsonCloverFermionR Dw(Umu, Grid, RBGrid, mass, csw_r, csw_t);
+  WilsonCloverFermionD Dw(Umu, Grid, RBGrid, mass, csw_r, csw_t);
 
   BiCGSTAB<LatticeFermion> CG(1.0e-8,10000);
   NonHermitianSchurRedBlackDiagMooeeSolve<LatticeFermion> SchurSolver(CG);
diff --git a/tests/solver/Test_wilsonclover_bicgstab_unprec.cc b/tests/solver/Test_wilsonclover_bicgstab_unprec.cc
index f546a744..48f194b0 100644
--- a/tests/solver/Test_wilsonclover_bicgstab_unprec.cc
+++ b/tests/solver/Test_wilsonclover_bicgstab_unprec.cc
@@ -70,9 +70,9 @@ int main (int argc, char ** argv)
   RealD mass = -0.1;
   RealD csw_r = 1.0;
   RealD csw_t = 1.0;
-  WilsonCloverFermionR Dw(Umu, Grid, RBGrid, mass, csw_r, csw_t);
+  WilsonCloverFermionD Dw(Umu, Grid, RBGrid, mass, csw_r, csw_t);
 
-  NonHermitianLinearOperator<WilsonCloverFermionR,LatticeFermion> HermOp(Dw);
+  NonHermitianLinearOperator<WilsonCloverFermionD,LatticeFermion> HermOp(Dw);
   BiCGSTAB<LatticeFermion> CG(1.0e-8,10000);
   CG(HermOp,src,result);
 
diff --git a/tests/solver/Test_wilsonclover_cagmres_unprec.cc b/tests/solver/Test_wilsonclover_cagmres_unprec.cc
index a8818168..8b264139 100644
--- a/tests/solver/Test_wilsonclover_cagmres_unprec.cc
+++ b/tests/solver/Test_wilsonclover_cagmres_unprec.cc
@@ -44,8 +44,8 @@ int main (int argc, char ** argv)
   std::vector<int> seeds({1,2,3,4});
   GridParallelRNG          pRNG(&Grid);  pRNG.SeedFixedIntegers(seeds);
 
-  typedef typename WilsonCloverFermionR::FermionField FermionField;
-  typename WilsonCloverFermionR::ImplParams params;
+  typedef typename WilsonCloverFermionD::FermionField FermionField;
+  typename WilsonCloverFermionD::ImplParams params;
   WilsonAnisotropyCoefficients anis;
 
   FermionField src(&Grid); random(pRNG,src);
@@ -61,9 +61,9 @@ int main (int argc, char ** argv)
   RealD mass  = 0.5;
   RealD csw_r = 1.0;
   RealD csw_t = 1.0;
-  WilsonCloverFermionR Dwc(Umu,Grid,RBGrid,mass,csw_r,csw_t,anis,params);
+  WilsonCloverFermionD Dwc(Umu,Grid,RBGrid,mass,csw_r,csw_t,anis,params);
 
-  MdagMLinearOperator<WilsonCloverFermionR,FermionField> HermOp(Dwc);
+  MdagMLinearOperator<WilsonCloverFermionD,FermionField> HermOp(Dwc);
   CommunicationAvoidingGeneralisedMinimalResidual<FermionField> CAGMRES(1.0e-8, 10000, 25);
   CAGMRES(HermOp,src,result);
 
diff --git a/tests/solver/Test_wilsonclover_cg_prec.cc b/tests/solver/Test_wilsonclover_cg_prec.cc
index abf64a1f..a0b3a712 100644
--- a/tests/solver/Test_wilsonclover_cg_prec.cc
+++ b/tests/solver/Test_wilsonclover_cg_prec.cc
@@ -72,10 +72,10 @@ int main (int argc, char ** argv)
   RealD csw_r = 1.0;
   RealD csw_t = 1.0;
   RealD cF = 1.0;
-  WilsonCloverFermionR Dw(Umu, Grid, RBGrid, mass, csw_r, csw_t);
-  CompactWilsonCloverFermionR Dw_compact(Umu, Grid, RBGrid, mass, csw_r, csw_t, 0.0);
-  WilsonExpCloverFermionR Dwe(Umu, Grid, RBGrid, mass, csw_r, csw_t);
-  CompactWilsonExpCloverFermionR Dwe_compact(Umu, Grid, RBGrid, mass, csw_r, csw_t, 0.0);
+  WilsonCloverFermionD Dw(Umu, Grid, RBGrid, mass, csw_r, csw_t);
+  CompactWilsonCloverFermionD Dw_compact(Umu, Grid, RBGrid, mass, csw_r, csw_t, 0.0);
+  WilsonExpCloverFermionD Dwe(Umu, Grid, RBGrid, mass, csw_r, csw_t);
+  CompactWilsonExpCloverFermionD Dwe_compact(Umu, Grid, RBGrid, mass, csw_r, csw_t, 0.0);
 
 
   //  HermitianOperator<WilsonFermion,LatticeFermion> HermOp(Dw);
@@ -89,22 +89,22 @@ int main (int argc, char ** argv)
   ConjugateGradient<LatticeFermion> CG(1.0e-8,10000);
 
   std::cout << GridLogMessage << "Testing Wilson Clover" << std::endl;
-  SchurDiagMooeeOperator<WilsonCloverFermionR,LatticeFermion> HermOpEO(Dw);
+  SchurDiagMooeeOperator<WilsonCloverFermionD,LatticeFermion> HermOpEO(Dw);
   result_o=Zero();
   CG(HermOpEO,src_o,result_o);
 
   std::cout << GridLogMessage << "Testing Compact Wilson Clover" << std::endl;
-  SchurDiagMooeeOperator<CompactWilsonCloverFermionR,LatticeFermion> HermOpEO_compact(Dw_compact);
+  SchurDiagMooeeOperator<CompactWilsonCloverFermionD,LatticeFermion> HermOpEO_compact(Dw_compact);
   result_o=Zero();
   CG(HermOpEO_compact,src_o,result_o);
 
   std::cout << GridLogMessage << "Testing Wilson Exp Clover" << std::endl;
-  SchurDiagMooeeOperator<WilsonExpCloverFermionR,LatticeFermion> HermOpEO_exp(Dwe);
+  SchurDiagMooeeOperator<WilsonExpCloverFermionD,LatticeFermion> HermOpEO_exp(Dwe);
   result_o=Zero();
   CG(HermOpEO_exp,src_o,result_o);
 
   std::cout << GridLogMessage << "Testing Compact Wilson Exp Clover" << std::endl;
-  SchurDiagMooeeOperator<CompactWilsonExpCloverFermionR,LatticeFermion> HermOpEO_exp_compact(Dwe_compact);
+  SchurDiagMooeeOperator<CompactWilsonExpCloverFermionD,LatticeFermion> HermOpEO_exp_compact(Dwe_compact);
   result_o=Zero();
   CG(HermOpEO_exp_compact,src_o,result_o);
 
diff --git a/tests/solver/Test_wilsonclover_cg_schur.cc b/tests/solver/Test_wilsonclover_cg_schur.cc
index 50d06af7..50a1c4a6 100644
--- a/tests/solver/Test_wilsonclover_cg_schur.cc
+++ b/tests/solver/Test_wilsonclover_cg_schur.cc
@@ -72,22 +72,22 @@ int main (int argc, char ** argv)
   RealD cF = 1.0;
 
   std::cout << GridLogMessage << "Testing Wilson Clover" << std::endl;
-  WilsonCloverFermionR Dw(Umu, Grid, RBGrid, mass, csw_r, csw_t);
+  WilsonCloverFermionD Dw(Umu, Grid, RBGrid, mass, csw_r, csw_t);
   result=Zero();
   SchurSolver(Dw,src,result);
 
   std::cout << GridLogMessage << "Testing Compact Wilson Clover" << std::endl;
-  CompactWilsonCloverFermionR Dw_compact(Umu, Grid, RBGrid, mass, csw_r, csw_t, 0.0);
+  CompactWilsonCloverFermionD Dw_compact(Umu, Grid, RBGrid, mass, csw_r, csw_t, 0.0);
   result=Zero();
   SchurSolver(Dw_compact,src,result);
 
   std::cout << GridLogMessage << "Testing Wilson Exp Clover" << std::endl;
-  WilsonExpCloverFermionR Dwe(Umu, Grid, RBGrid, mass, csw_r, csw_t);
+  WilsonExpCloverFermionD Dwe(Umu, Grid, RBGrid, mass, csw_r, csw_t);
   result=Zero();
   SchurSolver(Dwe,src,result);
 
   std::cout << GridLogMessage << "Testing Compact Wilson Exp Clover" << std::endl;
-  CompactWilsonExpCloverFermionR Dwe_compact(Umu, Grid, RBGrid, mass, csw_r, csw_t, 0.0);
+  CompactWilsonExpCloverFermionD Dwe_compact(Umu, Grid, RBGrid, mass, csw_r, csw_t, 0.0);
   result=Zero();
   SchurSolver(Dwe_compact,src,result);
   
diff --git a/tests/solver/Test_wilsonclover_cg_unprec.cc b/tests/solver/Test_wilsonclover_cg_unprec.cc
index 2a859f11..25cf07ee 100644
--- a/tests/solver/Test_wilsonclover_cg_unprec.cc
+++ b/tests/solver/Test_wilsonclover_cg_unprec.cc
@@ -71,31 +71,31 @@ int main (int argc, char ** argv)
   RealD csw_r = 1.0;
   RealD csw_t = 1.0;
   RealD cF = 1.0;
-  WilsonCloverFermionR Dw(Umu, Grid, RBGrid, mass, csw_r, csw_t);
-  CompactWilsonCloverFermionR Dw_compact(Umu, Grid, RBGrid, mass, csw_r, csw_t, 0.0);
-  WilsonExpCloverFermionR Dwe(Umu, Grid, RBGrid, mass, csw_r, csw_t);
-  CompactWilsonExpCloverFermionR Dwe_compact(Umu, Grid, RBGrid, mass, csw_r, csw_t, 0.0);
+  WilsonCloverFermionD Dw(Umu, Grid, RBGrid, mass, csw_r, csw_t);
+  CompactWilsonCloverFermionD Dw_compact(Umu, Grid, RBGrid, mass, csw_r, csw_t, 0.0);
+  WilsonExpCloverFermionD Dwe(Umu, Grid, RBGrid, mass, csw_r, csw_t);
+  CompactWilsonExpCloverFermionD Dwe_compact(Umu, Grid, RBGrid, mass, csw_r, csw_t, 0.0);
 
   ConjugateGradient<LatticeFermion> CG(1.0e-8,10000);
 
   std::cout << GridLogMessage << "Testing Wilson Clover" << std::endl;
-  MdagMLinearOperator<WilsonCloverFermionR,LatticeFermion> HermOp(Dw);
+  MdagMLinearOperator<WilsonCloverFermionD,LatticeFermion> HermOp(Dw);
   result=Zero();
   CG(HermOp,src,result);
 
   std::cout << GridLogMessage << "Testing Compact Wilson Clover" << std::endl;
-  MdagMLinearOperator<CompactWilsonCloverFermionR,LatticeFermion> HermOp_compact(Dw_compact);
+  MdagMLinearOperator<CompactWilsonCloverFermionD,LatticeFermion> HermOp_compact(Dw_compact);
   result=Zero();
   CG(HermOp_compact,src,result);
 
 
   std::cout << GridLogMessage << "Testing Wilson Exp Clover" << std::endl;
-  MdagMLinearOperator<WilsonExpCloverFermionR,LatticeFermion> HermOp_exp(Dwe);
+  MdagMLinearOperator<WilsonExpCloverFermionD,LatticeFermion> HermOp_exp(Dwe);
   result=Zero();
   CG(HermOp_exp,src,result);
 
   std::cout << GridLogMessage << "Testing Compact Wilson Exp Clover" << std::endl;
-  MdagMLinearOperator<CompactWilsonExpCloverFermionR,LatticeFermion> HermOp_exp_compact(Dwe_compact);
+  MdagMLinearOperator<CompactWilsonExpCloverFermionD,LatticeFermion> HermOp_exp_compact(Dwe_compact);
   result=Zero();
   CG(HermOp_exp_compact,src,result);
 
diff --git a/tests/solver/Test_wilsonclover_fcagmres_prec.cc b/tests/solver/Test_wilsonclover_fcagmres_prec.cc
index 1a294821..77b2afff 100644
--- a/tests/solver/Test_wilsonclover_fcagmres_prec.cc
+++ b/tests/solver/Test_wilsonclover_fcagmres_prec.cc
@@ -44,8 +44,8 @@ int main (int argc, char ** argv)
   std::vector<int> seeds({1,2,3,4});
   GridParallelRNG          pRNG(&Grid);  pRNG.SeedFixedIntegers(seeds);
 
-  typedef typename WilsonCloverFermionR::FermionField FermionField;
-  typename WilsonCloverFermionR::ImplParams params;
+  typedef typename WilsonCloverFermionD::FermionField FermionField;
+  typename WilsonCloverFermionD::ImplParams params;
   WilsonAnisotropyCoefficients anis;
 
   FermionField src(&Grid); random(pRNG,src);
@@ -61,9 +61,9 @@ int main (int argc, char ** argv)
   RealD mass  = 0.5;
   RealD csw_r = 1.0;
   RealD csw_t = 1.0;
-  WilsonCloverFermionR Dwc(Umu,Grid,RBGrid,mass,csw_r,csw_t,anis,params);
+  WilsonCloverFermionD Dwc(Umu,Grid,RBGrid,mass,csw_r,csw_t,anis,params);
 
-  MdagMLinearOperator<WilsonCloverFermionR,FermionField> HermOp(Dwc);
+  MdagMLinearOperator<WilsonCloverFermionD,FermionField> HermOp(Dwc);
 
   TrivialPrecon<FermionField> simple;
 
diff --git a/tests/solver/Test_wilsonclover_fgmres_prec.cc b/tests/solver/Test_wilsonclover_fgmres_prec.cc
index 15bb4136..0f48871f 100644
--- a/tests/solver/Test_wilsonclover_fgmres_prec.cc
+++ b/tests/solver/Test_wilsonclover_fgmres_prec.cc
@@ -44,8 +44,8 @@ int main (int argc, char ** argv)
   std::vector<int> seeds({1,2,3,4});
   GridParallelRNG          pRNG(&Grid);  pRNG.SeedFixedIntegers(seeds);
 
-  typedef typename WilsonCloverFermionR::FermionField FermionField;
-  typename WilsonCloverFermionR::ImplParams params;
+  typedef typename WilsonCloverFermionD::FermionField FermionField;
+  typename WilsonCloverFermionD::ImplParams params;
   WilsonAnisotropyCoefficients anis;
 
   FermionField src(&Grid); random(pRNG,src);
@@ -61,9 +61,9 @@ int main (int argc, char ** argv)
   RealD mass  = 0.5;
   RealD csw_r = 1.0;
   RealD csw_t = 1.0;
-  WilsonCloverFermionR Dwc(Umu,Grid,RBGrid,mass,csw_r,csw_t,anis,params);
+  WilsonCloverFermionD Dwc(Umu,Grid,RBGrid,mass,csw_r,csw_t,anis,params);
 
-  MdagMLinearOperator<WilsonCloverFermionR,FermionField> HermOp(Dwc);
+  MdagMLinearOperator<WilsonCloverFermionD,FermionField> HermOp(Dwc);
 
   TrivialPrecon<FermionField> simple;
 
diff --git a/tests/solver/Test_wilsonclover_gmres_unprec.cc b/tests/solver/Test_wilsonclover_gmres_unprec.cc
index 00f33382..b660d716 100644
--- a/tests/solver/Test_wilsonclover_gmres_unprec.cc
+++ b/tests/solver/Test_wilsonclover_gmres_unprec.cc
@@ -44,8 +44,8 @@ int main (int argc, char ** argv)
   std::vector<int> seeds({1,2,3,4});
   GridParallelRNG          pRNG(&Grid);  pRNG.SeedFixedIntegers(seeds);
 
-  typedef typename WilsonCloverFermionR::FermionField FermionField;
-  typename WilsonCloverFermionR::ImplParams params;
+  typedef typename WilsonCloverFermionD::FermionField FermionField;
+  typename WilsonCloverFermionD::ImplParams params;
   WilsonAnisotropyCoefficients anis;
 
   FermionField src(&Grid); random(pRNG,src);
@@ -61,9 +61,9 @@ int main (int argc, char ** argv)
   RealD mass  = 0.5;
   RealD csw_r = 1.0;
   RealD csw_t = 1.0;
-  WilsonCloverFermionR Dwc(Umu,Grid,RBGrid,mass,csw_r,csw_t,anis,params);
+  WilsonCloverFermionD Dwc(Umu,Grid,RBGrid,mass,csw_r,csw_t,anis,params);
 
-  MdagMLinearOperator<WilsonCloverFermionR,FermionField> HermOp(Dwc);
+  MdagMLinearOperator<WilsonCloverFermionD,FermionField> HermOp(Dwc);
   GeneralisedMinimalResidual<FermionField> GMRES(1.0e-8, 10000, 25);
   GMRES(HermOp,src,result);
 
diff --git a/tests/solver/Test_wilsonclover_mg.cc b/tests/solver/Test_wilsonclover_mg.cc
index 605d225d..1b0e8bb7 100644
--- a/tests/solver/Test_wilsonclover_mg.cc
+++ b/tests/solver/Test_wilsonclover_mg.cc
@@ -80,16 +80,16 @@ int main(int argc, char **argv) {
   // Note: We do chiral doubling, so actually only nbasis/2 full basis vectors are used
   const int nbasis = 40;
 
-  WilsonCloverFermionR Dwc(Umu, *FGrid, *FrbGrid, mass, csw_r, csw_t);
+  WilsonCloverFermionD Dwc(Umu, *FGrid, *FrbGrid, mass, csw_r, csw_t);
 
-  MdagMLinearOperator<WilsonCloverFermionR, LatticeFermion> MdagMOpDwc(Dwc);
+  MdagMLinearOperator<WilsonCloverFermionD, LatticeFermion> MdagMOpDwc(Dwc);
 
   std::cout << GridLogMessage << "**************************************************" << std::endl;
   std::cout << GridLogMessage << "Testing Multigrid for Wilson Clover" << std::endl;
   std::cout << GridLogMessage << "**************************************************" << std::endl;
 
   TrivialPrecon<LatticeFermion> TrivialPrecon;
-  auto MGPreconDwc = createMGInstance<vSpinColourVector, vTComplex, nbasis, WilsonCloverFermionR>(mgParams, levelInfo, Dwc, Dwc);
+  auto MGPreconDwc = createMGInstance<vSpinColourVector, vTComplex, nbasis, WilsonCloverFermionD>(mgParams, levelInfo, Dwc, Dwc);
 
   MGPreconDwc->setup();
 
diff --git a/tests/solver/Test_wilsonclover_mr_unprec.cc b/tests/solver/Test_wilsonclover_mr_unprec.cc
index ab49ec1f..be721236 100644
--- a/tests/solver/Test_wilsonclover_mr_unprec.cc
+++ b/tests/solver/Test_wilsonclover_mr_unprec.cc
@@ -44,8 +44,8 @@ int main (int argc, char ** argv)
   std::vector<int> seeds({1,2,3,4});
   GridParallelRNG          pRNG(&Grid);  pRNG.SeedFixedIntegers(seeds);
 
-  typedef typename WilsonCloverFermionR::FermionField FermionField;
-  typename WilsonCloverFermionR::ImplParams params;
+  typedef typename WilsonCloverFermionD::FermionField FermionField;
+  typename WilsonCloverFermionD::ImplParams params;
   WilsonAnisotropyCoefficients anis;
 
   FermionField src(&Grid); random(pRNG,src);
@@ -61,9 +61,9 @@ int main (int argc, char ** argv)
   RealD mass  = 0.5;
   RealD csw_r = 1.0;
   RealD csw_t = 1.0;
-  WilsonCloverFermionR Dwc(Umu,Grid,RBGrid,mass,csw_r,csw_t,anis,params);
+  WilsonCloverFermionD Dwc(Umu,Grid,RBGrid,mass,csw_r,csw_t,anis,params);
 
-  MdagMLinearOperator<WilsonCloverFermionR,FermionField> HermOp(Dwc);
+  MdagMLinearOperator<WilsonCloverFermionD,FermionField> HermOp(Dwc);
   MinimalResidual<FermionField> MR(1.0e-8,10000,0.8);
   MR(HermOp,src,result);
 
diff --git a/tests/solver/Test_zmobius_cg_prec.cc b/tests/solver/Test_zmobius_cg_prec.cc
index 6b007afc..7f1f98b8 100644
--- a/tests/solver/Test_zmobius_cg_prec.cc
+++ b/tests/solver/Test_zmobius_cg_prec.cc
@@ -101,7 +101,7 @@ int main(int argc, char** argv) {
   omegas.push_back( std::complex<double>(0.0686324988446592,-0.0550658530827402) );
 #endif
 
-  ZMobiusFermionR Ddwf(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mass, M5, omegas,1.,0.);
+  ZMobiusFermionD Ddwf(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mass, M5, omegas,1.,0.);
 
   LatticeFermion src_o(FrbGrid);
   LatticeFermion result_o(FrbGrid);
@@ -110,7 +110,7 @@ int main(int argc, char** argv) {
 
   GridStopWatch CGTimer;
 
-  SchurDiagMooeeOperator<ZMobiusFermionR, LatticeFermion> HermOpEO(Ddwf);
+  SchurDiagMooeeOperator<ZMobiusFermionD, LatticeFermion> HermOpEO(Ddwf);
   ConjugateGradient<LatticeFermion> CG(1.0e-8, 10000, 0);// switch off the assert
 
   CGTimer.Start();
@@ -121,7 +121,6 @@ int main(int argc, char** argv) {
             << std::endl;
 
   std::cout << GridLogMessage << "######## Dhop calls summary" << std::endl;
-  Ddwf.Report();
 
   Grid_finalize();
 }

From a5c77f8b95c0759469c7abd32b6cf80dd3bf72bf Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Fri, 25 Nov 2022 00:40:27 -0500
Subject: [PATCH 089/146] Tracing moved in order

---
 Grid/GridCore.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/Grid/GridCore.h b/Grid/GridCore.h
index 8e04a859..41c64ef6 100644
--- a/Grid/GridCore.h
+++ b/Grid/GridCore.h
@@ -44,10 +44,10 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 #include <Grid/GridStd.h>
 #include <Grid/threads/Pragmas.h>
 #include <Grid/perfmon/Timer.h>
-#include <Grid/perfmon/Tracing.h>
 //#include <Grid/perfmon/PerfCount.h>
 #include <Grid/util/Util.h>
 #include <Grid/log/Log.h>
+#include <Grid/perfmon/Tracing.h>
 #include <Grid/allocator/Allocator.h>
 #include <Grid/simd/Simd.h>
 #include <Grid/threads/ThreadReduction.h>

From bc9579dac6cb5e2babfe7682abb95d5ae7aa5519 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Fri, 25 Nov 2022 00:40:45 -0500
Subject: [PATCH 090/146] Old code path removed

---
 Grid/algorithms/approx/Chebyshev.h | 16 +---------------
 1 file changed, 1 insertion(+), 15 deletions(-)

diff --git a/Grid/algorithms/approx/Chebyshev.h b/Grid/algorithms/approx/Chebyshev.h
index 7c93f0b8..1d6984f3 100644
--- a/Grid/algorithms/approx/Chebyshev.h
+++ b/Grid/algorithms/approx/Chebyshev.h
@@ -258,26 +258,12 @@ public:
     for(int n=2;n<order;n++){
 
       Linop.HermOp(*Tn,y);
-#if 0
-      auto y_v = y.View();
-      auto Tn_v = Tn->View();
-      auto Tnp_v = Tnp->View();
-      auto Tnm_v = Tnm->View();
-      constexpr int Nsimd = vector_type::Nsimd();
-      accelerator_for(ss, in.Grid()->oSites(), Nsimd, {
-	  coalescedWrite(y_v[ss],xscale*y_v(ss)+mscale*Tn_v(ss));
-	  coalescedWrite(Tnp_v[ss],2.0*y_v(ss)-Tnm_v(ss));
-      });
-      if ( Coeffs[n] != 0.0) {
-	axpy(out,Coeffs[n],*Tnp,out);
-      }
-#else
       axpby(y,xscale,mscale,y,(*Tn));
       axpby(*Tnp,2.0,-1.0,y,(*Tnm));
       if ( Coeffs[n] != 0.0) {
 	axpy(out,Coeffs[n],*Tnp,out);
       }
-#endif
+
       // Cycle pointers to avoid copies
       Field *swizzle = Tnm;
       Tnm    =Tn;

From c0fb20fc039870673e9ca3643fa07ca25ccb808d Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Fri, 25 Nov 2022 00:43:12 -0500
Subject: [PATCH 091/146] Audit check for wrongly locked data

---
 Grid/allocator/MemoryManagerCache.cc | 55 ++++++++++++++++++++++------
 1 file changed, 43 insertions(+), 12 deletions(-)

diff --git a/Grid/allocator/MemoryManagerCache.cc b/Grid/allocator/MemoryManagerCache.cc
index 3bb3db7e..f03ee79f 100644
--- a/Grid/allocator/MemoryManagerCache.cc
+++ b/Grid/allocator/MemoryManagerCache.cc
@@ -8,9 +8,8 @@ NAMESPACE_BEGIN(Grid);
 static char print_buffer [ MAXLINE ];
 
 #define mprintf(...) snprintf (print_buffer,MAXLINE, __VA_ARGS__ ); std::cout << GridLogMemory << print_buffer;
-//#define dprintf(...) printf (__VA_ARGS__ ); fflush(stdout);
-#define dprintf(...) 
-
+#define dprintf(...) snprintf (print_buffer,MAXLINE, __VA_ARGS__ ); std::cout << GridLogMemory << print_buffer;
+//#define dprintf(...) 
 
 
 ////////////////////////////////////////////////////////////
@@ -132,9 +131,11 @@ void MemoryManager::Evict(AcceleratorViewEntry &AccCache)
   ///////////////////////////////////////////////////////////////////////////
   assert(AccCache.state!=Empty);
   
-  mprintf("MemoryManager: Evict(%lx) %lx\n",(uint64_t)AccCache.CpuPtr,(uint64_t)AccCache.AccPtr); 
-  assert(AccCache.accLock==0);
-  assert(AccCache.cpuLock==0);
+  mprintf("MemoryManager: Evict cpu %lx acc %lx cpuLock %ld accLock %ld\n",
+	  (uint64_t)AccCache.CpuPtr,(uint64_t)AccCache.AccPtr,
+	  (uint64_t)AccCache.cpuLock,(uint64_t)AccCache.accLock); 
+  if (AccCache.accLock!=0) return;
+  if (AccCache.cpuLock!=0) return;
   if(AccCache.state==AccDirty) {
     Flush(AccCache);
   }
@@ -197,6 +198,7 @@ void MemoryManager::CpuDiscard(AcceleratorViewEntry &AccCache)
 void MemoryManager::ViewClose(void* Ptr,ViewMode mode)
 {
   if( (mode==AcceleratorRead)||(mode==AcceleratorWrite)||(mode==AcceleratorWriteDiscard) ){
+    dprintf("AcceleratorViewClose %lx\n",(uint64_t)Ptr);
     AcceleratorViewClose((uint64_t)Ptr);
   } else if( (mode==CpuRead)||(mode==CpuWrite)){
     CpuViewClose((uint64_t)Ptr);
@@ -208,6 +210,7 @@ void *MemoryManager::ViewOpen(void* _CpuPtr,size_t bytes,ViewMode mode,ViewAdvis
 {
   uint64_t CpuPtr = (uint64_t)_CpuPtr;
   if( (mode==AcceleratorRead)||(mode==AcceleratorWrite)||(mode==AcceleratorWriteDiscard) ){
+    dprintf("AcceleratorViewOpen %lx\n",(uint64_t)CpuPtr);
     return (void *) AcceleratorViewOpen(CpuPtr,bytes,mode,hint);
   } else if( (mode==CpuRead)||(mode==CpuWrite)){
     return (void *)CpuViewOpen(CpuPtr,bytes,mode,hint);
@@ -247,11 +250,12 @@ uint64_t MemoryManager::AcceleratorViewOpen(uint64_t CpuPtr,size_t bytes,ViewMod
   assert(AccCache.cpuLock==0);  // Programming error
 
   if(AccCache.state!=Empty) {
-    dprintf("ViewOpen found entry %lx %lx : %ld %ld\n",
+    dprintf("ViewOpen found entry %lx %lx : %ld %ld accLock %ld\n",
 		    (uint64_t)AccCache.CpuPtr,
 		    (uint64_t)CpuPtr,
 		    (uint64_t)AccCache.bytes,
-		    (uint64_t)bytes);
+	            (uint64_t)bytes,
+		    (uint64_t)AccCache.accLock);
     assert(AccCache.CpuPtr == CpuPtr);
     assert(AccCache.bytes  ==bytes);
   }
@@ -286,6 +290,7 @@ uint64_t MemoryManager::AcceleratorViewOpen(uint64_t CpuPtr,size_t bytes,ViewMod
       AccCache.state  = Consistent; // Empty + AccRead => Consistent
     }
     AccCache.accLock= 1;
+    dprintf("Copied Empty entry into device accLock= %d\n",AccCache.accLock);
   } else if(AccCache.state==CpuDirty ){
     if(mode==AcceleratorWriteDiscard) {
       CpuDiscard(AccCache);
@@ -298,21 +303,21 @@ uint64_t MemoryManager::AcceleratorViewOpen(uint64_t CpuPtr,size_t bytes,ViewMod
       AccCache.state  = Consistent; // CpuDirty + AccRead => Consistent
     }
     AccCache.accLock++;
-    dprintf("Copied CpuDirty entry into device accLock %d\n",AccCache.accLock);
+    dprintf("CpuDirty entry into device ++accLock= %d\n",AccCache.accLock);
   } else if(AccCache.state==Consistent) {
     if((mode==AcceleratorWrite)||(mode==AcceleratorWriteDiscard))
       AccCache.state  = AccDirty;   // Consistent + AcceleratorWrite=> AccDirty
     else
       AccCache.state  = Consistent; // Consistent + AccRead => Consistent
     AccCache.accLock++;
-    dprintf("Consistent entry into device accLock %d\n",AccCache.accLock);
+    dprintf("Consistent entry into device ++accLock= %d\n",AccCache.accLock);
   } else if(AccCache.state==AccDirty) {
     if((mode==AcceleratorWrite)||(mode==AcceleratorWriteDiscard))
       AccCache.state  = AccDirty; // AccDirty + AcceleratorWrite=> AccDirty
     else
       AccCache.state  = AccDirty; // AccDirty + AccRead => AccDirty
     AccCache.accLock++;
-    dprintf("AccDirty entry into device accLock %d\n",AccCache.accLock);
+    dprintf("AccDirty entry ++accLock= %d\n",AccCache.accLock);
   } else {
     assert(0);
   }
@@ -320,6 +325,7 @@ uint64_t MemoryManager::AcceleratorViewOpen(uint64_t CpuPtr,size_t bytes,ViewMod
   // If view is opened on device remove from LRU
   if(AccCache.LRU_valid==1){
     // must possibly remove from LRU as now locked on GPU
+    dprintf("AccCache entry removed from LRU \n");
     LRUremove(AccCache);
   }
 
@@ -340,10 +346,12 @@ void MemoryManager::AcceleratorViewClose(uint64_t CpuPtr)
   assert(AccCache.accLock>0);
 
   AccCache.accLock--;
-
   // Move to LRU queue if not locked and close on device
   if(AccCache.accLock==0) {
+    dprintf("AccleratorViewClose %lx AccLock decremented to %ld move to LRU queue\n",(uint64_t)CpuPtr,(uint64_t)AccCache.accLock);
     LRUinsert(AccCache);
+  } else {
+    dprintf("AccleratorViewClose %lx AccLock decremented to %ld\n",(uint64_t)CpuPtr,(uint64_t)AccCache.accLock);
   }
 }
 void MemoryManager::CpuViewClose(uint64_t CpuPtr)
@@ -479,6 +487,29 @@ int   MemoryManager::isOpen   (void* _CpuPtr)
     return 0;
   }
 }
+void MemoryManager::Audit(std::string s)
+{
+  for(auto it=AccViewTable.begin();it!=AccViewTable.end();it++){
+    auto &AccCache = it->second;
+    
+    std::string str;
+    if ( AccCache.state==Empty    ) str = std::string("Empty");
+    if ( AccCache.state==CpuDirty ) str = std::string("CpuDirty");
+    if ( AccCache.state==AccDirty ) str = std::string("AccDirty");
+    if ( AccCache.state==Consistent)str = std::string("Consistent");
+
+    if ( AccCache.cpuLock || AccCache.accLock ) { 
+      std::cout << GridLogError << s<< "\n\t 0x"<<std::hex<<AccCache.CpuPtr<<std::dec
+		<< "\t0x"<<std::hex<<AccCache.AccPtr<<std::dec<<"\t" <<str
+		<< "\t cpuLock  " << AccCache.cpuLock
+		<< "\t accLock  " << AccCache.accLock
+		<< "\t LRUvalid " << AccCache.LRU_valid<<std::endl;
+    }
+
+    assert( AccCache.cpuLock== 0 ) ;
+    assert( AccCache.accLock== 0 ) ;
+  }
+}
 
 void MemoryManager::PrintState(void* _CpuPtr)
 {

From b690b1cbe9ff8b851df1dae56e3325d3e467204f Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Fri, 25 Nov 2022 00:43:57 -0500
Subject: [PATCH 092/146] Audit

---
 Grid/allocator/MemoryManagerShared.cc | 1 +
 1 file changed, 1 insertion(+)

diff --git a/Grid/allocator/MemoryManagerShared.cc b/Grid/allocator/MemoryManagerShared.cc
index c072873b..2434ad47 100644
--- a/Grid/allocator/MemoryManagerShared.cc
+++ b/Grid/allocator/MemoryManagerShared.cc
@@ -13,6 +13,7 @@ uint64_t  MemoryManager::DeviceToHostBytes;
 uint64_t  MemoryManager::HostToDeviceXfer;
 uint64_t  MemoryManager::DeviceToHostXfer;
 
+void  MemoryManager::Audit(void){};
 void  MemoryManager::ViewClose(void* AccPtr,ViewMode mode){};
 void *MemoryManager::ViewOpen(void* CpuPtr,size_t bytes,ViewMode mode,ViewAdvise hint){ return CpuPtr; };
 int   MemoryManager::isOpen   (void* CpuPtr) { return 0;}

From 7d8231ba32ce754a3b72cb38af556073f0450e33 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Fri, 25 Nov 2022 00:44:57 -0500
Subject: [PATCH 093/146] Tracing

---
 Grid/log/Log.cc | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)

diff --git a/Grid/log/Log.cc b/Grid/log/Log.cc
index acccec0e..b54c368d 100644
--- a/Grid/log/Log.cc
+++ b/Grid/log/Log.cc
@@ -66,6 +66,7 @@ GridLogger GridLogError  (1, "Error" , GridLogColours, "RED");
 GridLogger GridLogWarning(1, "Warning", GridLogColours, "YELLOW");
 GridLogger GridLogMessage(1, "Message", GridLogColours, "NORMAL");
 GridLogger GridLogMemory (1, "Memory", GridLogColours, "NORMAL");
+GridLogger GridLogTracing(1, "Tracing", GridLogColours, "NORMAL");
 GridLogger GridLogDebug  (1, "Debug", GridLogColours, "PURPLE");
 GridLogger GridLogPerformance(1, "Performance", GridLogColours, "GREEN");
 GridLogger GridLogDslash     (1, "Dslash", GridLogColours, "BLUE");
@@ -77,7 +78,7 @@ void GridLogConfigure(std::vector<std::string> &logstreams) {
   GridLogError.Active(1);
   GridLogWarning.Active(0);
   GridLogMessage.Active(1); // at least the messages should be always on
-  GridLogMemory.Active(0); // at least the messages should be always on
+  GridLogMemory.Active(0); 
   GridLogIterative.Active(0);
   GridLogDebug.Active(0);
   GridLogPerformance.Active(0);
@@ -87,6 +88,7 @@ void GridLogConfigure(std::vector<std::string> &logstreams) {
   GridLogHMC.Active(1);
 
   for (int i = 0; i < logstreams.size(); i++) {
+    if (logstreams[i] == std::string("Tracing"))     GridLogTracing.Active(1);
     if (logstreams[i] == std::string("Memory"))      GridLogMemory.Active(1);
     if (logstreams[i] == std::string("Warning"))     GridLogWarning.Active(1);
     if (logstreams[i] == std::string("NoMessage"))   GridLogMessage.Active(0);
@@ -94,8 +96,8 @@ void GridLogConfigure(std::vector<std::string> &logstreams) {
     if (logstreams[i] == std::string("Debug"))       GridLogDebug.Active(1);
     if (logstreams[i] == std::string("Performance")) GridLogPerformance.Active(1);
     if (logstreams[i] == std::string("Dslash"))      GridLogDslash.Active(1);
-    if (logstreams[i] == std::string("NoIntegrator"))  GridLogIntegrator.Active(0);
-    if (logstreams[i] == std::string("NoHMC"))         GridLogHMC.Active(0);
+    if (logstreams[i] == std::string("NoIntegrator"))GridLogIntegrator.Active(0);
+    if (logstreams[i] == std::string("NoHMC"))       GridLogHMC.Active(0);
     if (logstreams[i] == std::string("Colours"))     GridLogColours.Active(1);
   }
 }

From 63a30ae34f9be125c84c887037855a2ccca1b72a Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Fri, 25 Nov 2022 00:45:05 -0500
Subject: [PATCH 094/146] Tracing

---
 Grid/log/Log.h | 1 +
 1 file changed, 1 insertion(+)

diff --git a/Grid/log/Log.h b/Grid/log/Log.h
index fd706771..2d663a3c 100644
--- a/Grid/log/Log.h
+++ b/Grid/log/Log.h
@@ -186,6 +186,7 @@ extern GridLogger GridLogIterative  ;
 extern GridLogger GridLogIntegrator  ;
 extern GridLogger GridLogHMC;
 extern GridLogger GridLogMemory;
+extern GridLogger GridLogTracing;
 extern Colours    GridLogColours;
 
 std::string demangle(const char* name) ;

From 121c9e2cebbd0b277fd5ca36e6238591a4779613 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Fri, 25 Nov 2022 00:45:21 -0500
Subject: [PATCH 095/146] Tracing

---
 Grid/perfmon/Tracing.h | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/Grid/perfmon/Tracing.h b/Grid/perfmon/Tracing.h
index d39350c9..5000cef4 100644
--- a/Grid/perfmon/Tracing.h
+++ b/Grid/perfmon/Tracing.h
@@ -1,4 +1,7 @@
 #pragma once
+
+NAMESPACE_BEGIN(Grid);
+
 #ifdef GRID_TRACING_NVTX
 #include <nvToolsExt.h>
 class GridTracer {
@@ -64,3 +67,4 @@ inline void traceStop(int ID) {  }
 #else
 #define GRID_TRACE(name) GridTracer uniq_name_using_macros##__COUNTER__(name);
 #endif
+NAMESPACE_END(Grid);

From d71672dca939ede2459ba4cc25d6a40c3ff9e513 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Fri, 25 Nov 2022 00:46:35 -0500
Subject: [PATCH 096/146] Bug fix

---
 Grid/qcd/action/fermion/WilsonCompressor.h | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/Grid/qcd/action/fermion/WilsonCompressor.h b/Grid/qcd/action/fermion/WilsonCompressor.h
index 9ecb1c49..89aeca61 100644
--- a/Grid/qcd/action/fermion/WilsonCompressor.h
+++ b/Grid/qcd/action/fermion/WilsonCompressor.h
@@ -61,6 +61,7 @@ public:
 	if(s==0)    compress.Compress(buffer[off+i    ],rhs_v[so+table_v[idx].second]);
 	if(s==Ls-1) compress.Compress(buffer[off+i+vol],rhs_v[so+table_v[idx].second]);
     });
+    rhs_v.ViewClose();
   }
   template<class decompressor,class Decompression>
   static void DecompressFace(decompressor decompress,Decompression &dd)
@@ -171,6 +172,7 @@ public:
 					 compressor &compress,
 					 int off,int so,int partial)
   {
+    //    std::cout << " face gather simple DWF partial "<<partial <<std::endl;
     if(partial) FaceGatherPartialDWF::Gather_plane_simple(table,rhs,buffer,compress,off,so,partial);
     else        FaceGatherSimple::Gather_plane_simple(table,rhs,buffer,compress,off,so,partial);
   }
@@ -179,6 +181,7 @@ public:
 				    Vector<cobj *> pointers,int dimension,int plane,int cbmask,
 				    compressor &compress,int type,int partial)
   {
+    //    std::cout << " face gather exch DWF partial "<<partial <<std::endl;
     if(partial) FaceGatherPartialDWF::Gather_plane_exchange(table,rhs,pointers,dimension, plane,cbmask,compress,type,partial);
     else        FaceGatherSimple::Gather_plane_exchange    (table,rhs,pointers,dimension, plane,cbmask,compress,type,partial);
   }
@@ -186,6 +189,7 @@ public:
   static void MergeFace(decompressor decompress,Merger &mm)
   {
     int partial = mm.partial;
+    //    std::cout << " merge DWF partial "<<partial <<std::endl;
     if ( partial ) FaceGatherPartialDWF::MergeFace(decompress,mm);
     else           FaceGatherSimple::MergeFace(decompress,mm);
   }
@@ -194,6 +198,7 @@ public:
   static void DecompressFace(decompressor decompress,Decompression &dd)
   {
     int partial = dd.partial;
+    //    std::cout << " decompress DWF partial "<<partial <<std::endl;
     if ( partial ) FaceGatherPartialDWF::DecompressFace(decompress,dd);
     else           FaceGatherSimple::DecompressFace(decompress,dd);
   }

From bae6c263dc6b201fb18b9e585f1594c03142083d Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Fri, 25 Nov 2022 00:47:01 -0500
Subject: [PATCH 097/146] Audit

---
 Grid/allocator/MemoryManager.h | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/Grid/allocator/MemoryManager.h b/Grid/allocator/MemoryManager.h
index 740d8d92..ad2d9ffc 100644
--- a/Grid/allocator/MemoryManager.h
+++ b/Grid/allocator/MemoryManager.h
@@ -36,6 +36,11 @@ NAMESPACE_BEGIN(Grid);
 
 #define GRID_ALLOC_SMALL_LIMIT (4096)
 
+#define STRINGIFY(x) #x
+#define TOSTRING(x) STRINGIFY(x)
+#define FILE_LINE __FILE__ ":" TOSTRING(__LINE__)
+#define AUDIT(a) MemoryManager::Audit(FILE_LINE)
+
 /*Pinning pages is costly*/
 ////////////////////////////////////////////////////////////////////////////
 // Advise the LatticeAccelerator class
@@ -94,6 +99,7 @@ private:
 
   static void PrintBytes(void);
  public:
+  static void Audit(std::string s);
   static void Init(void);
   static void InitMessage(void);
   static void *AcceleratorAllocate(size_t bytes);

From f6402cb6c4aa31b33ecfc661befe009efe089165 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Fri, 25 Nov 2022 00:50:33 -0500
Subject: [PATCH 098/146] AUDIT removal

---
 .../WilsonFermion5DImplementation.h           | 31 ++++++++++++-------
 1 file changed, 19 insertions(+), 12 deletions(-)

diff --git a/Grid/qcd/action/fermion/implementation/WilsonFermion5DImplementation.h b/Grid/qcd/action/fermion/implementation/WilsonFermion5DImplementation.h
index 992f4d20..4ca24789 100644
--- a/Grid/qcd/action/fermion/implementation/WilsonFermion5DImplementation.h
+++ b/Grid/qcd/action/fermion/implementation/WilsonFermion5DImplementation.h
@@ -96,6 +96,8 @@ WilsonFermion5D<Impl>::WilsonFermion5D(GaugeField &_Umu,
     Coordinate block = p.dirichlet;
     if ( block[0] || block[1] || block[2] || block[3] || block[4] ){
       Dirichlet = 1;
+      std::cout << GridLogMessage << " WilsonFermion: non-trivial Dirichlet condition "<< block << std::endl;
+      std::cout << GridLogMessage << " WilsonFermion: partial Dirichlet "<< p.partialDirichlet << std::endl;
       Block = block;
     }
   } else {
@@ -137,9 +139,6 @@ WilsonFermion5D<Impl>::WilsonFermion5D(GaugeField &_Umu,
   StencilEven.BuildSurfaceList(LLs,vol4);
    StencilOdd.BuildSurfaceList(LLs,vol4);
 
-   //  std::cout << GridLogMessage << " SurfaceLists "<< Stencil.surface_list.size()
-   //                       <<" " << StencilEven.surface_list.size()<<std::endl;
-
 }
 
 template<class Impl>
@@ -148,21 +147,29 @@ void WilsonFermion5D<Impl>::ImportGauge(const GaugeField &_Umu)
   GaugeField HUmu(_Umu.Grid());
   HUmu = _Umu*(-0.5);
   if ( Dirichlet ) {
+
+    if ( this->Params.partialDirichlet ) {
+      std::cout << GridLogMessage << " partialDirichlet BCs " <<Block<<std::endl;
+    } else {
+      std::cout << GridLogMessage << " FULL Dirichlet BCs " <<Block<<std::endl;
+    }
+    
     std:: cout << GridLogMessage << "Checking block size multiple of rank boundaries for Dirichlet"<<std::endl;
     for(int d=0;d<Nd;d++) {
       int GaugeBlock = Block[d+1];
       int ldim=GaugeGrid()->LocalDimensions()[d];
       if (GaugeBlock) assert( (GaugeBlock%ldim)==0);
     }
-  }
-  if ( Dirichlet && (!this->Params.partialDirichlet) ) {
-    std::cout << GridLogMessage << " Dirichlet filtering gauge field BCs block " <<Block<<std::endl;
-    Coordinate GaugeBlock(Nd);
-    for(int d=0;d<Nd;d++) GaugeBlock[d] = Block[d+1];
-    DirichletFilter<GaugeField> Filter(GaugeBlock);
-    Filter.applyFilter(HUmu);
-  } else {
-    std::cout << GridLogMessage << " Dirichlet "<< Dirichlet << " not filtered gauge field" <<std::endl;
+
+    if (!this->Params.partialDirichlet) {
+      std::cout << GridLogMessage << " Dirichlet filtering gauge field BCs block " <<Block<<std::endl;
+      Coordinate GaugeBlock(Nd);
+      for(int d=0;d<Nd;d++) GaugeBlock[d] = Block[d+1];
+      DirichletFilter<GaugeField> Filter(GaugeBlock);
+      Filter.applyFilter(HUmu);
+    } else {
+      std::cout << GridLogMessage << " Dirichlet "<< Dirichlet << " NOT filtered gauge field" <<std::endl;
+    }
   }
   Impl::DoubleStore(GaugeGrid(),Umu,HUmu);
   pickCheckerboard(Even,UmuEven,Umu);

From 5fa573dfd367e01c3b9bb56faf90e689e784959d Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Fri, 25 Nov 2022 00:51:04 -0500
Subject: [PATCH 099/146] partial send fix

---
 Grid/stencil/Stencil.h | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/Grid/stencil/Stencil.h b/Grid/stencil/Stencil.h
index da230e7e..cffede12 100644
--- a/Grid/stencil/Stencil.h
+++ b/Grid/stencil/Stencil.h
@@ -203,7 +203,7 @@ class CartesianStencilAccelerator {
 template<class vobj,class cobj,class Parameters>
 class CartesianStencilView : public CartesianStencilAccelerator<vobj,cobj,Parameters>
 {
- private:
+public:
   int *closed;
   StencilEntry *cpu_ptr;
   ViewMode      mode;
@@ -676,6 +676,8 @@ public:
       int block = dirichlet_block[dimension];
       this->_comms_send[ii] = comm_dim;
       this->_comms_recv[ii] = comm_dim;
+      this->_comms_partial_send[ii] = 0;
+      this->_comms_partial_recv[ii] = 0;
       if ( block && comm_dim ) {
 	assert(abs(displacement) < ld );
 	// Quiesce communication across block boundaries
@@ -1131,6 +1133,7 @@ public:
 	send_buf = this->u_send_buf_p; // Gather locally, must send
 	assert(send_buf!=NULL);
 
+	//	std::cout << " GatherPlaneSimple partial send "<< comms_partial_send<<std::endl;
 	compressor::Gather_plane_simple(face_table[face_idx],rhs,send_buf,compress,comm_off,so,comms_partial_send);
 
         int duplicate = CheckForDuplicate(dimension,sx,comm_proc,(void *)&recv_buf[comm_off],0,xbytes,rbytes,cbmask);

From e13930c8b228380b6556541c334a0ae40a97afae Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Wed, 30 Nov 2022 15:11:29 -0500
Subject: [PATCH 100/146] Faster fermtoprop case

---
 Grid/qcd/QCD.h | 23 +++++++++++++++++++++++
 1 file changed, 23 insertions(+)

diff --git a/Grid/qcd/QCD.h b/Grid/qcd/QCD.h
index 858aead7..b292e560 100644
--- a/Grid/qcd/QCD.h
+++ b/Grid/qcd/QCD.h
@@ -451,9 +451,20 @@ template<class vobj> void pokeLorentz(vobj &lhs,const decltype(peekIndex<Lorentz
 // Fermion <-> propagator assignements
 //////////////////////////////////////////////
 //template <class Prop, class Ferm>
+#define FAST_FERM_TO_PROP
 template <class Fimpl>
 void FermToProp(typename Fimpl::PropagatorField &p, const typename Fimpl::FermionField &f, const int s, const int c)
 {
+#ifdef FAST_FERM_TO_PROP
+  autoView(p_v,p,AcceleratorWrite);
+  autoView(f_v,f,AcceleratorRead);
+  accelerator_for(idx,p_v.oSites(),1,{
+      for(int ss = 0; ss < Ns; ++ss) {
+      for(int cc = 0; cc < Fimpl::Dimension; ++cc) {
+	p_v[idx]()(ss,s)(cc,c) = f_v[idx]()(ss)(cc); // Propagator sink index is LEFT, suitable for left mult by gauge link (e.g.)
+      }}
+  );
+#else
   for(int j = 0; j < Ns; ++j)
     {
       auto pjs = peekSpin(p, j, s);
@@ -465,12 +476,23 @@ void FermToProp(typename Fimpl::PropagatorField &p, const typename Fimpl::Fermio
 	}
       pokeSpin(p, pjs, j, s);
     }
+#endif
 }
     
 //template <class Prop, class Ferm>
 template <class Fimpl>
 void PropToFerm(typename Fimpl::FermionField &f, const typename Fimpl::PropagatorField &p, const int s, const int c)
 {
+#ifdef FAST_FERM_TO_PROP
+  autoView(p_v,p,AcceleratorWrite);
+  autoView(f_v,f,AcceleratorRead);
+  accelerator_for(idx,p_v.oSites(),1,{
+      for(int ss = 0; ss < Ns; ++ss) {
+      for(int cc = 0; cc < Fimpl::Dimension; ++cc) {
+	f_v[idx]()(ss)(cc) = p_v[idx]()(ss,s)(cc,c); // LEFT index is copied across for s,c right index
+      }}
+  );
+#else
   for(int j = 0; j < Ns; ++j)
     {
       auto pjs = peekSpin(p, j, s);
@@ -482,6 +504,7 @@ void PropToFerm(typename Fimpl::FermionField &f, const typename Fimpl::Propagato
 	}
       pokeSpin(f, fj, j);
     }
+#endif
 }
     
 //////////////////////////////////////////////

From 97a098636d816970631b120877b85e13691dd8ff Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Wed, 30 Nov 2022 15:36:35 -0500
Subject: [PATCH 101/146] FermToProp

---
 Grid/qcd/QCD.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/Grid/qcd/QCD.h b/Grid/qcd/QCD.h
index b292e560..9c56c12c 100644
--- a/Grid/qcd/QCD.h
+++ b/Grid/qcd/QCD.h
@@ -463,7 +463,7 @@ void FermToProp(typename Fimpl::PropagatorField &p, const typename Fimpl::Fermio
       for(int cc = 0; cc < Fimpl::Dimension; ++cc) {
 	p_v[idx]()(ss,s)(cc,c) = f_v[idx]()(ss)(cc); // Propagator sink index is LEFT, suitable for left mult by gauge link (e.g.)
       }}
-  );
+    });
 #else
   for(int j = 0; j < Ns; ++j)
     {
@@ -491,7 +491,7 @@ void PropToFerm(typename Fimpl::FermionField &f, const typename Fimpl::Propagato
       for(int cc = 0; cc < Fimpl::Dimension; ++cc) {
 	f_v[idx]()(ss)(cc) = p_v[idx]()(ss,s)(cc,c); // LEFT index is copied across for s,c right index
       }}
-  );
+    });
 #else
   for(int j = 0; j < Ns; ++j)
     {

From 67f569354e7a2cc6f3217857cbf633457f6755c6 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Wed, 30 Nov 2022 15:51:13 -0500
Subject: [PATCH 102/146] Partial dirichlet changes

---
 Grid/log/Log.cc                               |  1 +
 Grid/qcd/action/ActionBase.h                  | 12 ++++---
 Grid/qcd/action/fermion/WilsonCompressor.h    | 35 +++++++++++++------
 .../GeneralEvenOddRationalRatio.h             |  1 +
 .../pseudofermion/TwoFlavourEvenOddRatio.h    | 26 +++++++++++++-
 Grid/qcd/hmc/integrators/Integrator.h         | 23 +++++++++++-
 HMC/Mobius2p1f_DD_EOFA_96I_mshift.cc          | 28 +++++++++++----
 TODO                                          |  3 ++
 benchmarks/Benchmark_dwf_fp32_partial.cc      | 12 ++++---
 systems/Crusher/config-command                |  1 +
 10 files changed, 114 insertions(+), 28 deletions(-)

diff --git a/Grid/log/Log.cc b/Grid/log/Log.cc
index b54c368d..166aea0a 100644
--- a/Grid/log/Log.cc
+++ b/Grid/log/Log.cc
@@ -79,6 +79,7 @@ void GridLogConfigure(std::vector<std::string> &logstreams) {
   GridLogWarning.Active(0);
   GridLogMessage.Active(1); // at least the messages should be always on
   GridLogMemory.Active(0); 
+  GridLogTracing.Active(0); 
   GridLogIterative.Active(0);
   GridLogDebug.Active(0);
   GridLogPerformance.Active(0);
diff --git a/Grid/qcd/action/ActionBase.h b/Grid/qcd/action/ActionBase.h
index 5027fcf4..1e8d6d7a 100644
--- a/Grid/qcd/action/ActionBase.h
+++ b/Grid/qcd/action/ActionBase.h
@@ -55,14 +55,18 @@ public:
     deriv_num=0;
   }
   void  deriv_log(RealD nrm, RealD max,RealD Fdt_nrm,RealD Fdt_max) {
-    deriv_max_sum+=max;
+    if ( max > deriv_max_sum ) {
+      deriv_max_sum=max;
+    }
     deriv_norm_sum+=nrm;
-    Fdt_max_sum+=Fdt_max;
+    if ( Fdt_max > Fdt_max_sum ) {
+      Fdt_max_sum=Fdt_max;
+    }
     Fdt_norm_sum+=Fdt_nrm; deriv_num++;
   }
-  RealD deriv_max_average(void)       { return deriv_max_sum/deriv_num; };
+  RealD deriv_max_average(void)       { return deriv_max_sum; };
   RealD deriv_norm_average(void)      { return deriv_norm_sum/deriv_num; };
-  RealD Fdt_max_average(void)         { return Fdt_max_sum/deriv_num; };
+  RealD Fdt_max_average(void)         { return Fdt_max_sum; };
   RealD Fdt_norm_average(void)        { return Fdt_norm_sum/deriv_num; };
   RealD deriv_timer(void)        { return deriv_us; };
   RealD S_timer(void)            { return S_us; };
diff --git a/Grid/qcd/action/fermion/WilsonCompressor.h b/Grid/qcd/action/fermion/WilsonCompressor.h
index 89aeca61..f62b3aba 100644
--- a/Grid/qcd/action/fermion/WilsonCompressor.h
+++ b/Grid/qcd/action/fermion/WilsonCompressor.h
@@ -36,10 +36,12 @@ NAMESPACE_BEGIN(Grid);
 // Wilson compressor will need FaceGather policies for:
 // Periodic, Dirichlet, and partial Dirichlet for DWF
 ///////////////////////////////////////////////////////////////
+const int dwf_compressor_depth=2;
 class FaceGatherPartialDWF
 {
 public:
-  static int PartialCompressionFactor(GridBase *grid) {return grid->_fdimensions[0]/2;};
+  //  static int PartialCompressionFactor(GridBase *grid) {return grid->_fdimensions[0]/(2*dwf_compressor_depth);};
+  static int PartialCompressionFactor(GridBase *grid) {return 1;}
   //  static int PartialCompressionFactor(GridBase *grid) { return 1;}
   template<class vobj,class cobj,class compressor>
   static void Gather_plane_simple (commVector<std::pair<int,int> >& table,
@@ -52,14 +54,17 @@ public:
     //  Shrinks local and remote comms buffers
     GridBase *Grid = rhs.Grid();
     int Ls = Grid->_rdimensions[0];
+    //    int depth=dwf_compressor_depth;
+    int depth=Ls/2;
     std::pair<int,int> *table_v = & table[0];
     auto rhs_v = rhs.View(AcceleratorRead);
     int vol=table.size()/Ls;
     accelerator_forNB( idx,table.size(), vobj::Nsimd(), {
 	Integer i=idx/Ls;
 	Integer s=idx%Ls;
-	if(s==0)    compress.Compress(buffer[off+i    ],rhs_v[so+table_v[idx].second]);
-	if(s==Ls-1) compress.Compress(buffer[off+i+vol],rhs_v[so+table_v[idx].second]);
+	Integer sc=depth+s-(Ls-depth);
+	if(s<depth)     compress.Compress(buffer[off+i+s*vol],rhs_v[so+table_v[idx].second]);
+	if(s>=Ls-depth) compress.Compress(buffer[off+i+sc*vol],rhs_v[so+table_v[idx].second]);
     });
     rhs_v.ViewClose();
   }
@@ -67,6 +72,8 @@ public:
   static void DecompressFace(decompressor decompress,Decompression &dd)
   {
     auto Ls = dd.dims[0];
+    //    int depth=dwf_compressor_depth;
+    int depth=Ls/2;
     // Just pass in the Grid
     auto kp = dd.kernel_p;
     auto mp = dd.mpi_p;
@@ -75,11 +82,12 @@ public:
     accelerator_forNB(o,size,1,{
 	int idx=o/Ls;
 	int   s=o%Ls;
-	if ( s == 0 ) {
-	  int oo=idx;
+	if ( s < depth ) {
+	  int oo=s*vol+idx;
 	  kp[o]=mp[oo];
-	} else if ( s == Ls-1 ) {
-	  int oo=vol+idx; 
+	} else if ( s >= Ls-depth ) {
+	  int sc = depth + s - (Ls-depth);
+	  int oo=sc*vol+idx; 
 	  kp[o]=mp[oo];
 	} else {
 	  kp[o] = Zero();//fill rest with zero if partial dirichlet
@@ -98,7 +106,9 @@ public:
   {
     GridBase *Grid = rhs.Grid();
     int Ls = Grid->_rdimensions[0];
-
+    //    int depth=dwf_compressor_depth;
+    int depth = Ls/2;
+    
     // insertion of zeroes...
     assert( (table.size()&0x1)==0);
     int num=table.size()/2;
@@ -113,7 +123,7 @@ public:
 	//  Reorders both local and remote comms buffers
 	//  
 	int s  = j % Ls;
-	int sp1 = (s+1)%Ls;  // peri incremented s slice
+	int sp1 = (s+depth)%Ls;  // peri incremented s slice
 	
 	int hxyz= j/Ls;
 
@@ -136,6 +146,8 @@ public:
   static void MergeFace(decompressor decompress,Merger &mm)
   {
     auto Ls = mm.dims[0];
+    int depth = Ls/2;
+    //    int depth=dwf_compressor_depth;
     int  num= mm.buffer_size/2; // relate vol and Ls to buffer size
     auto mp = &mm.mpointer[0];
     auto vp0= &mm.vpointers[0][0]; // First arg is exchange first
@@ -149,7 +161,7 @@ public:
 	int xyz0=hxyz*2;
 	int xyz1=hxyz*2+1;
 
-	int sp = (s+1)%Ls; 
+	int sp = (s+depth)%Ls; 
 	int jj= hxyz + sp*nnum ; // 0,1,2,3 -> Ls-1 slice , 0-slice, 1-slice ....
 
 	int oo0= s+xyz0*Ls;
@@ -163,7 +175,8 @@ public:
 class FaceGatherDWFMixedBCs
 {
 public:
-  static int PartialCompressionFactor(GridBase *grid) {return grid->_fdimensions[0]/2;};
+  //  static int PartialCompressionFactor(GridBase *grid) {return grid->_fdimensions[0]/(2*dwf_compressor_depth);};
+  static int PartialCompressionFactor(GridBase *grid) {return 1;}
   
   template<class vobj,class cobj,class compressor>
   static void Gather_plane_simple (commVector<std::pair<int,int> >& table,
diff --git a/Grid/qcd/action/pseudofermion/GeneralEvenOddRationalRatio.h b/Grid/qcd/action/pseudofermion/GeneralEvenOddRationalRatio.h
index ff605362..cb680f2f 100644
--- a/Grid/qcd/action/pseudofermion/GeneralEvenOddRationalRatio.h
+++ b/Grid/qcd/action/pseudofermion/GeneralEvenOddRationalRatio.h
@@ -279,6 +279,7 @@ NAMESPACE_BEGIN(Grid);
       // S_f = chi^dag* P(V^dag*V)/Q(V^dag*V)* N(M^dag*M)/D(M^dag*M)* P(V^dag*V)/Q(V^dag*V)* chi       
       //////////////////////////////////////////////////////
       virtual RealD Sinitial(const GaugeField &U) {
+	std::cout << GridLogMessage << "Returning stored two flavour refresh action "<<RefreshAction<<std::endl;
 	return RefreshAction;
       }
 
diff --git a/Grid/qcd/action/pseudofermion/TwoFlavourEvenOddRatio.h b/Grid/qcd/action/pseudofermion/TwoFlavourEvenOddRatio.h
index f348fde5..476b1c53 100644
--- a/Grid/qcd/action/pseudofermion/TwoFlavourEvenOddRatio.h
+++ b/Grid/qcd/action/pseudofermion/TwoFlavourEvenOddRatio.h
@@ -38,7 +38,7 @@ NAMESPACE_BEGIN(Grid);
     class TwoFlavourEvenOddRatioPseudoFermionAction : public Action<typename Impl::GaugeField> {
     public:
       INHERIT_IMPL_TYPES(Impl);
-
+      
     private:
       FermionOperator<Impl> & NumOp;// the basic operator
       FermionOperator<Impl> & DenOp;// the basic operator
@@ -112,28 +112,48 @@ NAMESPACE_BEGIN(Grid);
         // NumOp == V
         // DenOp == M
         //
+    AUDIT();
         FermionField etaOdd (NumOp.FermionRedBlackGrid());
         FermionField etaEven(NumOp.FermionRedBlackGrid());
         FermionField tmp    (NumOp.FermionRedBlackGrid());
 
+    AUDIT();
         pickCheckerboard(Even,etaEven,eta);
+    AUDIT();
         pickCheckerboard(Odd,etaOdd,eta);
 
+    AUDIT();
         NumOp.ImportGauge(U);
+    AUDIT();
         DenOp.ImportGauge(U);
+	std::cout << " TwoFlavourRefresh:  Imported gauge "<<std::endl;
+    AUDIT();
 
         SchurDifferentiableOperator<Impl> Mpc(DenOp);
+    AUDIT();
         SchurDifferentiableOperator<Impl> Vpc(NumOp);
+    AUDIT();
 
+	std::cout << " TwoFlavourRefresh: Diff ops "<<std::endl;
+    AUDIT();
         // Odd det factors
         Mpc.MpcDag(etaOdd,PhiOdd);
+    AUDIT();
+	std::cout << " TwoFlavourRefresh: MpcDag "<<std::endl;
         tmp=Zero();
+    AUDIT();
+	std::cout << " TwoFlavourRefresh: Zero() guess "<<std::endl;
+    AUDIT();
         HeatbathSolver(Vpc,PhiOdd,tmp);
+    AUDIT();
+	std::cout << " TwoFlavourRefresh: Heatbath solver "<<std::endl;
         Vpc.Mpc(tmp,PhiOdd);            
+	std::cout << " TwoFlavourRefresh: Mpc "<<std::endl;
 
         // Even det factors
         DenOp.MooeeDag(etaEven,tmp);
         NumOp.MooeeInvDag(tmp,PhiEven);
+	std::cout << " TwoFlavourRefresh: Mee "<<std::endl;
 
 	RefreshAction = norm2(etaEven)+norm2(etaOdd);
 	std::cout << " refresh " <<action_name()<< " action "<<RefreshAction<<std::endl;
@@ -142,6 +162,10 @@ NAMESPACE_BEGIN(Grid);
       //////////////////////////////////////////////////////
       // S = phi^dag V (Mdag M)^-1 Vdag phi
       //////////////////////////////////////////////////////
+      virtual RealD Sinitial(const GaugeField &U) {
+	std::cout << GridLogMessage << "Returning stored two flavour refresh action "<<RefreshAction<<std::endl;
+	return RefreshAction;
+      }
       virtual RealD S(const GaugeField &U) {
 
         NumOp.ImportGauge(U);
diff --git a/Grid/qcd/hmc/integrators/Integrator.h b/Grid/qcd/hmc/integrators/Integrator.h
index e1f9e3f0..85a135ad 100644
--- a/Grid/qcd/hmc/integrators/Integrator.h
+++ b/Grid/qcd/hmc/integrators/Integrator.h
@@ -132,10 +132,17 @@ protected:
 
       Field& Us = Smearer.get_U(as[level].actions.at(a)->is_smeared);
       double start_force = usecond();
+
+      std::cout << GridLogMessage << "AuditForce["<<level<<"]["<<a<<"] before"<<std::endl;
+      AUDIT();
+      
       as[level].actions.at(a)->deriv_timer_start();
       as[level].actions.at(a)->deriv(Us, force);  // deriv should NOT include Ta
       as[level].actions.at(a)->deriv_timer_stop();
 
+      std::cout << GridLogMessage << "AuditForce["<<level<<"]["<<a<<"] after"<<std::endl;
+      AUDIT();
+
       std::cout << GridLogIntegrator << "Smearing (on/off): " << as[level].actions.at(a)->is_smeared << std::endl;
       auto name = as[level].actions.at(a)->action_name();
       if (as[level].actions.at(a)->is_smeared) Smearer.smeared_force(force);
@@ -284,7 +291,7 @@ public:
       for (int actionID = 0; actionID < as[level].actions.size(); ++actionID) {
 	std::cout << GridLogMessage 
 		  << as[level].actions.at(actionID)->action_name()
-		  <<"["<<level<<"]["<< actionID<<"] : "
+		  <<"["<<level<<"]["<< actionID<<"] :\n\t\t "
 		  <<" force max " << as[level].actions.at(actionID)->deriv_max_average()
 		  <<" norm "      << as[level].actions.at(actionID)->deriv_norm_average()
 		  <<" Fdt max  "  << as[level].actions.at(actionID)->Fdt_max_average()
@@ -364,9 +371,14 @@ public:
         std::cout << GridLogMessage << "refresh [" << level << "][" << actionID << "] "<<name << std::endl;
 
         Field& Us = Smearer.get_U(as[level].actions.at(actionID)->is_smeared);
+
+	std::cout << GridLogMessage << "AuditRefresh["<<level<<"]["<<actionID<<"] before"<<std::endl;
+	AUDIT();
 	as[level].actions.at(actionID)->refresh_timer_start();
         as[level].actions.at(actionID)->refresh(Us, sRNG, pRNG);
 	as[level].actions.at(actionID)->refresh_timer_stop();
+	std::cout << GridLogMessage << "AuditRefresh["<<level<<"]["<<actionID<<"] after"<<std::endl;
+	AUDIT();
       }
 
       // Refresh the higher representation actions
@@ -403,6 +415,7 @@ public:
     // Actions
     for (int level = 0; level < as.size(); ++level) {
       for (int actionID = 0; actionID < as[level].actions.size(); ++actionID) {
+	AUDIT();
         // get gauge field from the SmearingPolicy and
         // based on the boolean is_smeared in actionID
         Field& Us = Smearer.get_U(as[level].actions.at(actionID)->is_smeared);
@@ -412,6 +425,7 @@ public:
    	        as[level].actions.at(actionID)->S_timer_stop();
         std::cout << GridLogMessage << "S [" << level << "][" << actionID << "] H = " << Hterm << std::endl;
         H += Hterm;
+	AUDIT();
       }
       as[level].apply(S_hireps, Representations, level, H);
     }
@@ -424,7 +438,9 @@ public:
     void operator()(std::vector<Action<FieldType>*> repr_set, Repr& Rep, int level, RealD& H) {
       
       for (int a = 0; a < repr_set.size(); ++a) {
+	AUDIT();
         RealD Hterm = repr_set.at(a)->Sinitial(Rep.U);
+	AUDIT();
         std::cout << GridLogMessage << "Sinitial Level " << level << " term " << a << " H Hirep = " << Hterm << std::endl;
         H += Hterm;
 
@@ -449,8 +465,10 @@ public:
         Field& Us = Smearer.get_U(as[level].actions.at(actionID)->is_smeared);
         std::cout << GridLogMessage << "S [" << level << "][" << actionID << "] action eval " << std::endl;
 	        as[level].actions.at(actionID)->S_timer_start();
+	AUDIT();
         Hterm = as[level].actions.at(actionID)->Sinitial(Us);
    	        as[level].actions.at(actionID)->S_timer_stop();
+	AUDIT();
         std::cout << GridLogMessage << "S [" << level << "][" << actionID << "] H = " << Hterm << std::endl;
         H += Hterm;
       }
@@ -463,6 +481,7 @@ public:
   
   void integrate(Field& U) 
   {
+    AUDIT();
     // reset the clocks
     t_U = 0;
     for (int level = 0; level < as.size(); ++level) {
@@ -480,8 +499,10 @@ public:
       assert(fabs(t_U - t_P[level]) < 1.0e-6);  // must be the same
       std::cout << GridLogIntegrator << " times[" << level << "]= " << t_P[level] << " " << t_U << std::endl;
     }
+    AUDIT();
 
     FieldImplementation::Project(U);
+    AUDIT();
 
     // and that we indeed got to the end of the trajectory
     assert(fabs(t_U - Params.trajL) < 1.0e-6);
diff --git a/HMC/Mobius2p1f_DD_EOFA_96I_mshift.cc b/HMC/Mobius2p1f_DD_EOFA_96I_mshift.cc
index f4ca8515..e85bec7e 100644
--- a/HMC/Mobius2p1f_DD_EOFA_96I_mshift.cc
+++ b/HMC/Mobius2p1f_DD_EOFA_96I_mshift.cc
@@ -179,8 +179,11 @@ int main(int argc, char **argv) {
   MD.name    = std::string("Force Gradient");
   //typedef GenericHMCRunner<MinimumNorm2> HMCWrapper;
   // MD.name    = std::string("MinimumNorm2");
-  //  MD.MDsteps =  4;
-  MD.MDsteps =  4;
+  // TrajL = 2
+  // 4/2 => 0.6 dH
+  // 3/3 => ?? dH
+  //MD.MDsteps =  4;
+  MD.MDsteps =  3;
   MD.trajL   = 0.5;
 
   HMCparameters HMCparams;
@@ -223,7 +226,7 @@ int main(int argc, char **argv) {
   Real light_mass   = 7.8e-4;
   Real strange_mass = 0.0362;
   Real pv_mass      = 1.0;
-  std::vector<Real> hasenbusch({ light_mass, 3.8e-3, 0.0145, 0.045, 0.108, 0.25, 0.51 , pv_mass });
+  std::vector<Real> hasenbusch({ light_mass, 0.005, 0.0145, 0.045, 0.108, 0.25, 0.51 , pv_mass });
   //  std::vector<Real> hasenbusch({ light_mass, 0.0145, 0.045, 0.108, 0.25, 0.51 , 0.75 , pv_mass });
 
 
@@ -327,6 +330,8 @@ int main(int argc, char **argv) {
   ParamsF.dirichlet=NonDirichlet;
   ParamsDir.dirichlet=Dirichlet;
   ParamsDirF.dirichlet=Dirichlet;
+  ParamsDir.partialDirichlet=1;
+  ParamsDirF.partialDirichlet=1;
 
   //  double StoppingCondition = 1e-14;
   //  double MDStoppingCondition = 1e-9;
@@ -342,8 +347,8 @@ int main(int argc, char **argv) {
   // Collect actions
   ////////////////////////////////////
   ActionLevel<HMCWrapper::Field> Level1(1);
-  ActionLevel<HMCWrapper::Field> Level2(2);
-  ActionLevel<HMCWrapper::Field> Level3(30);
+  ActionLevel<HMCWrapper::Field> Level2(3);
+  ActionLevel<HMCWrapper::Field> Level3(15);
 
   ////////////////////////////////////
   // Strange action
@@ -474,13 +479,21 @@ int main(int argc, char **argv) {
     if ( dirichlet_den[h]==1) ParamsDen.dirichlet = Dirichlet;
     else                      ParamsDen.dirichlet = NonDirichlet;
 
+    if ( dirichlet_num[h]==1) ParamsNum.partialDirichlet = 1;
+    else                      ParamsNum.partialDirichlet = 0;
+
+    if ( dirichlet_den[h]==1) ParamsDen.partialDirichlet = 1;
+    else                      ParamsDen.partialDirichlet = 0;
+    
     Numerators.push_back  (new FermionAction(U,*FGrid,*FrbGrid,*GridPtr,*GridRBPtr,light_num[h],M5,b,c, ParamsNum));
     Denominators.push_back(new FermionAction(U,*FGrid,*FrbGrid,*GridPtr,*GridRBPtr,light_den[h],M5,b,c, ParamsDen));
 
     ParamsDenF.dirichlet = ParamsDen.dirichlet;
+    ParamsDenF.partialDirichlet = ParamsDen.partialDirichlet;
     DenominatorsF.push_back(new FermionActionF(UF,*FGridF,*FrbGridF,*GridPtrF,*GridRBPtrF,light_den[h],M5,b,c, ParamsDenF));
 
     ParamsNumF.dirichlet = ParamsNum.dirichlet;
+    ParamsNumF.partialDirichlet = ParamsNum.partialDirichlet;
     NumeratorsF.push_back  (new FermionActionF(UF,*FGridF,*FrbGridF,*GridPtrF,*GridRBPtrF,light_num[h],M5,b,c, ParamsNumF));
 
     LinOpD.push_back(new LinearOperatorD(*Denominators[h]));
@@ -516,9 +529,11 @@ int main(int argc, char **argv) {
       FermionActionD2::ImplParams ParamsNumD2(boundary);
 
       ParamsDenD2.dirichlet = ParamsDen.dirichlet;
+      ParamsDenD2.partialDirichlet = ParamsDen.partialDirichlet;
       DenominatorsD2.push_back(new FermionActionD2(UD2,*FGridF,*FrbGridF,*GridPtrF,*GridRBPtrF,light_den[h],M5,b,c, ParamsDenD2));
 
       ParamsNumD2.dirichlet = ParamsNum.dirichlet;
+      ParamsNumD2.partialDirichlet = ParamsNum.partialDirichlet;
       NumeratorsD2.push_back  (new FermionActionD2(UD2,*FGridF,*FrbGridF,*GridPtrF,*GridRBPtrF,light_num[h],M5,b,c, ParamsNumD2));
     
       Bdys.push_back( new OneFlavourEvenOddRatioRationalMixedPrecPseudoFermionAction<FermionImplPolicy,FermionImplPolicyF,FermionImplPolicyD2>(
@@ -543,7 +558,8 @@ int main(int argc, char **argv) {
   int nquo=Quotients.size();
   Level1.push_back(Bdys[0]);
   Level1.push_back(Bdys[1]);
-  for(int h=0;h<nquo-1;h++){
+  Level2.push_back(Quotients[0]);
+  for(int h=1;h<nquo-1;h++){
     Level2.push_back(Quotients[h]);
   }
   Level2.push_back(Quotients[nquo-1]);
diff --git a/TODO b/TODO
index 12c523a6..380af2b7 100644
--- a/TODO
+++ b/TODO
@@ -1,3 +1,6 @@
+Lattice_basis.h -- > HIP and SYCL GPU code
+
+
 ======
 DDHMC
 ======
diff --git a/benchmarks/Benchmark_dwf_fp32_partial.cc b/benchmarks/Benchmark_dwf_fp32_partial.cc
index 4db3022a..0cbd4a6c 100644
--- a/benchmarks/Benchmark_dwf_fp32_partial.cc
+++ b/benchmarks/Benchmark_dwf_fp32_partial.cc
@@ -88,6 +88,7 @@ int main (int argc, char ** argv)
   // Node level
   //////////////////////
   for(int d=0;d<Nd;d++) CommDim[d]= (mpi[d]/shm[d])>1 ? 1 : 0;
+  //  for(int d=0;d<Nd;d++) CommDim[d]= 1;
   Dirichlet[0] = 0;
   Dirichlet[1] = CommDim[0]*latt4[0]/mpi[0] * shm[0];
   Dirichlet[2] = CommDim[1]*latt4[1]/mpi[1] * shm[1];
@@ -222,7 +223,7 @@ void Benchmark(int Ls, Coordinate Dirichlet, int partial)
   {
     ref = Zero();
     for(int mu=0;mu<Nd;mu++){
-
+      int depth=dwf_compressor_depth;
       tmp = Cshift(src,mu+1,1);
       {
 	autoView( tmp_v  , tmp  , CpuWrite);
@@ -230,7 +231,7 @@ void Benchmark(int Ls, Coordinate Dirichlet, int partial)
 	autoView( Ucopy_v, Ucopy[mu]  , CpuRead);
 	for(int ss=0;ss<U[mu].Grid()->oSites();ss++){
 	  for(int s=0;s<Ls;s++){
-	    if ( (s==0) || (s==Ls-1)){
+	    if ( (s<depth) || (s>=Ls-depth)){
 	      tmp_v[Ls*ss+s] = Ucopy_v[ss]*tmp_v[Ls*ss+s];
 	    } else {
 	      tmp_v[Ls*ss+s] = U_v[ss]*tmp_v[Ls*ss+s];
@@ -246,7 +247,7 @@ void Benchmark(int Ls, Coordinate Dirichlet, int partial)
 	autoView( src_v, src    , CpuRead);
 	for(int ss=0;ss<U[mu].Grid()->oSites();ss++){
 	  for(int s=0;s<Ls;s++){
-	    if ( (s==0) || (s==Ls-1) ){
+	    if ( (s<depth) || (s>=Ls-depth)){
 	      tmp_v[Ls*ss+s] = adj(Ucopy_v[ss])*src_v[Ls*ss+s];
 	    } else {
 	      tmp_v[Ls*ss+s] = adj(U_v[ss])*src_v[Ls*ss+s];
@@ -342,6 +343,7 @@ void Benchmark(int Ls, Coordinate Dirichlet, int partial)
     ref = Zero();
     for(int mu=0;mu<Nd;mu++){
 
+      int depth=dwf_compressor_depth;
       tmp = Cshift(src,mu+1,1);
       {
 	autoView( tmp_v  , tmp  , CpuWrite);
@@ -349,7 +351,7 @@ void Benchmark(int Ls, Coordinate Dirichlet, int partial)
 	autoView( Ucopy_v, Ucopy[mu]  , CpuRead);
 	for(int ss=0;ss<U[mu].Grid()->oSites();ss++){
 	  for(int s=0;s<Ls;s++){
-	    if ( (s==0) || (s==Ls-1)){
+	    if ( (s<depth) || (s>=Ls-depth)){
 	      tmp_v[Ls*ss+s] = Ucopy_v[ss]*tmp_v[Ls*ss+s];
 	    } else {
 	      tmp_v[Ls*ss+s] = U_v[ss]*tmp_v[Ls*ss+s];
@@ -365,7 +367,7 @@ void Benchmark(int Ls, Coordinate Dirichlet, int partial)
 	autoView( src_v, src    , CpuRead);
 	for(int ss=0;ss<U[mu].Grid()->oSites();ss++){
 	  for(int s=0;s<Ls;s++){
-	    if ( (s==0) || (s==Ls-1) ){
+	    if ( (s<depth) || (s>=Ls-depth)){
 	      tmp_v[Ls*ss+s] = adj(Ucopy_v[ss])*src_v[Ls*ss+s];
 	    } else {
 	      tmp_v[Ls*ss+s] = adj(U_v[ss])*src_v[Ls*ss+s];
diff --git a/systems/Crusher/config-command b/systems/Crusher/config-command
index bd825ab3..3965767f 100644
--- a/systems/Crusher/config-command
+++ b/systems/Crusher/config-command
@@ -3,6 +3,7 @@ CLIME=`spack find --paths c-lime@2-3-9 | grep c-lime| cut -c 15-`
 --with-lime=$CLIME \
 --enable-unified=no \
 --enable-shm=nvlink \
+--enable-tracing=timer \
 --enable-accelerator=hip \
 --enable-gen-simd-width=64 \
 --enable-simd=GPU \

From d49694f38f1b07676c4200892922c9a89bca99ad Mon Sep 17 00:00:00 2001
From: Raoul Hodgson <raoul.hodgson@ed.ac.uk>
Date: Tue, 6 Dec 2022 15:48:54 +0000
Subject: [PATCH 103/146] PropToFerm fix

---
 Grid/qcd/QCD.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/Grid/qcd/QCD.h b/Grid/qcd/QCD.h
index 9c56c12c..39f45ada 100644
--- a/Grid/qcd/QCD.h
+++ b/Grid/qcd/QCD.h
@@ -484,8 +484,8 @@ template <class Fimpl>
 void PropToFerm(typename Fimpl::FermionField &f, const typename Fimpl::PropagatorField &p, const int s, const int c)
 {
 #ifdef FAST_FERM_TO_PROP
-  autoView(p_v,p,AcceleratorWrite);
-  autoView(f_v,f,AcceleratorRead);
+  autoView(p_v,p,AcceleratorRead);
+  autoView(f_v,f,AcceleratorWrite);
   accelerator_for(idx,p_v.oSites(),1,{
       for(int ss = 0; ss < Ns; ++ss) {
       for(int cc = 0; cc < Fimpl::Dimension; ++cc) {

From 40234f531ff83172cebbb7401916641d0df7ddd2 Mon Sep 17 00:00:00 2001
From: Raoul Hodgson <raoul.hodgson@ed.ac.uk>
Date: Tue, 6 Dec 2022 17:34:51 +0000
Subject: [PATCH 104/146] FermToProp accelerator_for -> thread_for

---
 Grid/qcd/QCD.h | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

diff --git a/Grid/qcd/QCD.h b/Grid/qcd/QCD.h
index 39f45ada..6dd6d5dc 100644
--- a/Grid/qcd/QCD.h
+++ b/Grid/qcd/QCD.h
@@ -456,9 +456,9 @@ template <class Fimpl>
 void FermToProp(typename Fimpl::PropagatorField &p, const typename Fimpl::FermionField &f, const int s, const int c)
 {
 #ifdef FAST_FERM_TO_PROP
-  autoView(p_v,p,AcceleratorWrite);
-  autoView(f_v,f,AcceleratorRead);
-  accelerator_for(idx,p_v.oSites(),1,{
+  autoView(p_v,p,CpuWrite);
+  autoView(f_v,f,CpuRead);
+  thread_for(idx,p_v.oSites(),{
       for(int ss = 0; ss < Ns; ++ss) {
       for(int cc = 0; cc < Fimpl::Dimension; ++cc) {
 	p_v[idx]()(ss,s)(cc,c) = f_v[idx]()(ss)(cc); // Propagator sink index is LEFT, suitable for left mult by gauge link (e.g.)
@@ -484,9 +484,9 @@ template <class Fimpl>
 void PropToFerm(typename Fimpl::FermionField &f, const typename Fimpl::PropagatorField &p, const int s, const int c)
 {
 #ifdef FAST_FERM_TO_PROP
-  autoView(p_v,p,AcceleratorRead);
-  autoView(f_v,f,AcceleratorWrite);
-  accelerator_for(idx,p_v.oSites(),1,{
+  autoView(p_v,p,CpuRead);
+  autoView(f_v,f,CpuWrite);
+  thread_for(idx,p_v.oSites(),{
       for(int ss = 0; ss < Ns; ++ss) {
       for(int cc = 0; cc < Fimpl::Dimension; ++cc) {
 	f_v[idx]()(ss)(cc) = p_v[idx]()(ss,s)(cc,c); // LEFT index is copied across for s,c right index

From ff6777a98daaa7fd433e15cf0256ca50d3dae9dd Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 13 Dec 2022 08:13:51 -0500
Subject: [PATCH 105/146] Variable depth experiments

---
 Grid/qcd/action/fermion/WilsonCompressor.h | 36 ++++++++++++++++------
 1 file changed, 27 insertions(+), 9 deletions(-)

diff --git a/Grid/qcd/action/fermion/WilsonCompressor.h b/Grid/qcd/action/fermion/WilsonCompressor.h
index f62b3aba..fd1bbe89 100644
--- a/Grid/qcd/action/fermion/WilsonCompressor.h
+++ b/Grid/qcd/action/fermion/WilsonCompressor.h
@@ -36,13 +36,16 @@ NAMESPACE_BEGIN(Grid);
 // Wilson compressor will need FaceGather policies for:
 // Periodic, Dirichlet, and partial Dirichlet for DWF
 ///////////////////////////////////////////////////////////////
-const int dwf_compressor_depth=2;
+const int dwf_compressor_depth=1;
+#define DWF_COMPRESS
 class FaceGatherPartialDWF
 {
 public:
-  //  static int PartialCompressionFactor(GridBase *grid) {return grid->_fdimensions[0]/(2*dwf_compressor_depth);};
-  static int PartialCompressionFactor(GridBase *grid) {return 1;}
-  //  static int PartialCompressionFactor(GridBase *grid) { return 1;}
+#ifdef DWF_COMPRESS
+  static int PartialCompressionFactor(GridBase *grid) {return grid->_fdimensions[0]/(2*dwf_compressor_depth);};
+#else
+  static int PartialCompressionFactor(GridBase *grid) { return 1;}
+#endif
   template<class vobj,class cobj,class compressor>
   static void Gather_plane_simple (commVector<std::pair<int,int> >& table,
 				   const Lattice<vobj> &rhs,
@@ -54,8 +57,11 @@ public:
     //  Shrinks local and remote comms buffers
     GridBase *Grid = rhs.Grid();
     int Ls = Grid->_rdimensions[0];
-    //    int depth=dwf_compressor_depth;
+#ifdef DWF_COMPRESS
+    int depth=dwf_compressor_depth;
+#else 
     int depth=Ls/2;
+#endif
     std::pair<int,int> *table_v = & table[0];
     auto rhs_v = rhs.View(AcceleratorRead);
     int vol=table.size()/Ls;
@@ -72,8 +78,11 @@ public:
   static void DecompressFace(decompressor decompress,Decompression &dd)
   {
     auto Ls = dd.dims[0];
-    //    int depth=dwf_compressor_depth;
+#ifdef DWF_COMPRESS
+    int depth=dwf_compressor_depth;
+#else
     int depth=Ls/2;
+#endif    
     // Just pass in the Grid
     auto kp = dd.kernel_p;
     auto mp = dd.mpi_p;
@@ -106,8 +115,11 @@ public:
   {
     GridBase *Grid = rhs.Grid();
     int Ls = Grid->_rdimensions[0];
-    //    int depth=dwf_compressor_depth;
+#ifdef DWF_COMPRESS
+    int depth=dwf_compressor_depth;
+#else
     int depth = Ls/2;
+#endif
     
     // insertion of zeroes...
     assert( (table.size()&0x1)==0);
@@ -146,8 +158,11 @@ public:
   static void MergeFace(decompressor decompress,Merger &mm)
   {
     auto Ls = mm.dims[0];
+#ifdef DWF_COMPRESS
+    int depth=dwf_compressor_depth;
+#else
     int depth = Ls/2;
-    //    int depth=dwf_compressor_depth;
+#endif
     int  num= mm.buffer_size/2; // relate vol and Ls to buffer size
     auto mp = &mm.mpointer[0];
     auto vp0= &mm.vpointers[0][0]; // First arg is exchange first
@@ -175,8 +190,11 @@ public:
 class FaceGatherDWFMixedBCs
 {
 public:
-  //  static int PartialCompressionFactor(GridBase *grid) {return grid->_fdimensions[0]/(2*dwf_compressor_depth);};
+#ifdef DWF_COMPRESS
+  static int PartialCompressionFactor(GridBase *grid) {return grid->_fdimensions[0]/(2*dwf_compressor_depth);};
+#else 
   static int PartialCompressionFactor(GridBase *grid) {return 1;}
+#endif
   
   template<class vobj,class cobj,class compressor>
   static void Gather_plane_simple (commVector<std::pair<int,int> >& table,

From b54d0f3c730009fb4992e7ff09d0417a23d56747 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 13 Dec 2022 08:14:27 -0500
Subject: [PATCH 106/146] Smaller deltaH down to 7000s on t=0.5 trajectory

---
 HMC/Mobius2p1f_DD_EOFA_96I_mshift.cc | 28 ++++++++++++++++++----------
 1 file changed, 18 insertions(+), 10 deletions(-)

diff --git a/HMC/Mobius2p1f_DD_EOFA_96I_mshift.cc b/HMC/Mobius2p1f_DD_EOFA_96I_mshift.cc
index e85bec7e..0a924486 100644
--- a/HMC/Mobius2p1f_DD_EOFA_96I_mshift.cc
+++ b/HMC/Mobius2p1f_DD_EOFA_96I_mshift.cc
@@ -181,7 +181,7 @@ int main(int argc, char **argv) {
   // MD.name    = std::string("MinimumNorm2");
   // TrajL = 2
   // 4/2 => 0.6 dH
-  // 3/3 => ?? dH
+  // 3/3 => 0.8 dH .. depth 3, slower
   //MD.MDsteps =  4;
   MD.MDsteps =  3;
   MD.trajL   = 0.5;
@@ -223,13 +223,15 @@ int main(int argc, char **argv) {
   RealD c   = 0.5;
   Real beta         = 2.13;
   //  Real light_mass   = 5.4e-4;
-  Real light_mass   = 7.8e-4;
+  Real light_mass     = 7.8e-4;
+  Real light_mass_dir = 0.01;
   Real strange_mass = 0.0362;
   Real pv_mass      = 1.0;
-  std::vector<Real> hasenbusch({ light_mass, 0.005, 0.0145, 0.045, 0.108, 0.25, 0.51 , pv_mass });
+  std::vector<Real> hasenbusch({ 0.01, 0.045, 0.108, 0.25, 0.51 , pv_mass });
+  //  std::vector<Real> hasenbusch({ light_mass, 0.01, 0.045, 0.108, 0.25, 0.51 , pv_mass });
+  //  std::vector<Real> hasenbusch({ light_mass, 0.005, 0.0145, 0.045, 0.108, 0.25, 0.51 , pv_mass }); // Updated
   //  std::vector<Real> hasenbusch({ light_mass, 0.0145, 0.045, 0.108, 0.25, 0.51 , 0.75 , pv_mass });
 
-
   OneFlavourRationalParams OFRp; // Up/down
   OFRp.lo       = 4.0e-5;
   OFRp.hi       = 90.0;
@@ -238,18 +240,24 @@ int main(int argc, char **argv) {
   OFRp.mdtolerance= 1.0e-3;
   //  OFRp.degree   = 20; converges
   //  OFRp.degree   = 16;
-  OFRp.degree   = 12;
+  OFRp.degree   = 18;
   OFRp.precision= 80;
   OFRp.BoundsCheckFreq=0;
   std::vector<RealD> ActionTolByPole({
       1.0e-8,1.0e-8,1.0e-8,1.0e-8,
       1.0e-8,1.0e-8,1.0e-8,1.0e-8,
-      1.0e-8,1.0e-8,1.0e-8,1.0e-8
+      1.0e-8,1.0e-8,1.0e-8,1.0e-8,
+      1.0e-8,1.0e-8,1.0e-8,1.0e-8,
+      1.0e-8,1.0e-8
     });
   std::vector<RealD> MDTolByPole({
-      1.0e-6,3.0e-7,1.0e-7,1.0e-7,
+      1.0e-5,5.0e-6,1.0e-6,1.0e-7, // soften convergence more more
+      //      1.0e-6,3.0e-7,1.0e-7,1.0e-7,
+      //      3.0e-6,1.0e-6,1.0e-7,1.0e-7, // soften convergence
       1.0e-8,1.0e-8,1.0e-8,1.0e-8,
-      1.0e-8,1.0e-8,1.0e-8,1.0e-8
+      1.0e-8,1.0e-8,1.0e-8,1.0e-8,
+      1.0e-8,1.0e-8,1.0e-8,1.0e-8,
+      1.0e-8,1.0e-8
     });
 
   auto GridPtr   = TheHMC.Resources.GetCartesian();
@@ -540,12 +548,12 @@ int main(int argc, char **argv) {
 			   *Numerators[h],*Denominators[h],
 			   *NumeratorsF[h],*DenominatorsF[h],
 			   *NumeratorsD2[h],*DenominatorsD2[h],
-			   OFRp, 200) );
+			   OFRp, 400) );
       Bdys.push_back( new OneFlavourEvenOddRatioRationalMixedPrecPseudoFermionAction<FermionImplPolicy,FermionImplPolicyF,FermionImplPolicyD2>(
 			   *Numerators[h],*Denominators[h],
 			   *NumeratorsF[h],*DenominatorsF[h],
 			   *NumeratorsD2[h],*DenominatorsD2[h],
-			   OFRp, 200) );
+			   OFRp, 400) );
 #else
       Bdys.push_back( new OneFlavourEvenOddRatioRationalPseudoFermionAction<FermionImplPolicy>(*Numerators[h],*Denominators[h],OFRp));
       Bdys.push_back( new OneFlavourEvenOddRatioRationalPseudoFermionAction<FermionImplPolicy>(*Numerators[h],*Denominators[h],OFRp));

From 5bb7ba92faa362aca13803a1f6a9e9a12dacd5ef Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 13 Dec 2022 08:15:11 -0500
Subject: [PATCH 107/146] Test for DDHMC force term

---
 tests/forces/Test_double_ratio.cc | 510 ++++++++++++++++++++++++++++++
 1 file changed, 510 insertions(+)
 create mode 100644 tests/forces/Test_double_ratio.cc

diff --git a/tests/forces/Test_double_ratio.cc b/tests/forces/Test_double_ratio.cc
new file mode 100644
index 00000000..0a350692
--- /dev/null
+++ b/tests/forces/Test_double_ratio.cc
@@ -0,0 +1,510 @@
+/*
+  2f Full det MdagM 10^6 force ~ 1.3e7
+rid : Message : 1767.283471 s : +++++++++++++++++++++++++++++++++++++++++++++++++++++++++
+Grid : Message : 1767.283476 s : S1 : 1.52885e+09
+Grid : Message : 1767.283480 s : S2 : 1.52886e+09
+Grid : Message : 1767.283482 s : dS : 8877.34
+Grid : Message : 1767.283483 s : dSpred : 8877.7
+Grid : Message : 1767.283484 s : diff : -0.360484
+Grid : Message : 1767.283485 s : *********************************************************
+
+  2f Full det MpcdagMpc 10^6 force ~ 1.8e6
+Grid : Message : 2399.576962 s : +++++++++++++++++++++++++++++++++++++++++++++++++++++++++
+Grid : Message : 2399.576968 s : S1 : 1.52885e+09
+Grid : Message : 2399.576972 s : S2 : 1.52886e+09
+Grid : Message : 2399.576974 s : dS : 9728.49
+Grid : Message : 2399.576975 s : dSpred : 9726.58
+Grid : Message : 2399.576976 s : diff : 1.90683
+Grid : Message : 2399.576977 s : *********************************************************
+
+  2f bdy MdagM 1500 force Force ~ 2800
+Grid : Message : 4622.385061 s : +++++++++++++++++++++++++++++++++++++++++++++++++++++++++
+Grid : Message : 4622.385067 s : S1 : 1.52885e+09
+Grid : Message : 4622.385071 s : S2 : 1.52885e+09
+Grid : Message : 4622.385072 s : dS : 25.4944
+Grid : Message : 4622.385073 s : dSpred : 25.4672
+Grid : Message : 4622.385074 s : diff : 0.0271414
+Grid : Message : 4622.385075 s : *********************************************************
+
+  2f bdy MpcdagMpc 10^6 force   Force ~ 2200
+Grid : Message : 4622.385061 s : +++++++++++++++++++++++++++++++++++++++++++++++++++++++++
+Grid : Message : 4622.385067 s : S1 : 1.52885e+09
+Grid : Message : 4622.385071 s : S2 : 1.52885e+09
+Grid : Message : 4622.385072 s : dS : 25.4944
+Grid : Message : 4622.385073 s : dSpred : 25.4672
+Grid : Message : 4622.385074 s : diff : 0.0271414
+Grid : Message : 4622.385075 s : *********************************************************
+  
+  1f Bdy Det
+//
+// These all had tol set by OFRp,  not through MDpoles
+// So assumptions it was Remez might be wrong.
+//
+Optimisation log:  looser rational AND MD tolerances sloppy
+MobiusForce.221179 -- same as HMC. dS is mispredicted Forece  ~2.8
+Grid : Message : 6582.258991 s : dS : 0.024478
+Grid : Message : 6582.258992 s : dSpred : 0.00791876
+Grid : Message : 6582.258994 s : diff : 0.0165592
+
+MobiusForce.221193 -- tight rational AND MD tolerances to 1e-8 ~ 2.8 same
+Grid : Message : 1964.939209 s : S1 : 7.64404e+08
+Grid : Message : 1964.939213 s : S2 : 7.64404e+08
+Grid : Message : 1964.939215 s : dS : -0.00775838 <--- too loose even on action
+Grid : Message : 1964.939216 s : dSpred : -0.00416793 
+Grid : Message : 1964.939217 s : diff : -0.00359045
+
+MobiusForce.221394 -- tight rational, MD tol sloppy Force ~ 2.8
+Grid : Message : 2376.921950 s : S1 : 764404436.44069
+Grid : Message : 2376.921954 s : S2 : 764404436.43299
+Grid : Message : 2376.921956 s : dS : -0.0076971054077148
+Grid : Message : 2376.921958 s : dSpred : -0.0041610472282526
+Grid : Message : 2376.921959 s : diff : -0.0035360581794623
+
+
+MobiusForce.221587 -- slightly sloppier action, coming from tol array
+                   -- much sloppier force
+		   -- degree 18
+      1.0e-8,1.0e-8,1.0e-8,1.0e-8,
+      1.0e-6,3.0e-7,1.0e-7,1.0e-7,  // Orig sloppy
+Grid : Message : 2438.875507 s : S1 : 764404436.42251
+Grid : Message : 2438.875512 s : S2 : 764404436.4148
+Grid : Message : 2438.875514 s : dS : -0.0077102184295654
+Grid : Message : 2438.875516 s : dSpred : -0.0075684496959103
+Grid : Message : 2438.875517 s : diff : -0.00014176873365508
+
+MobiusForce.221639       3.0e-6,1.0e-6,1.0e-7,1.0e-7, // soften convergence more
+
+Grid : Message : 2373.927550 s : +++++++++++++++++++++++++++++++++++++++++++++++++++++++++
+Grid : Message : 2373.927600 s : S1 : 764404436.42251
+Grid : Message : 2373.927640 s : S2 : 764404436.4148
+Grid : Message : 2373.927660 s : dS : -0.0077102184295654
+Grid : Message : 2373.927680 s : dSpred : -0.0075993463919849
+Grid : Message : 2373.927690 s : diff : -0.00011087203758051
+Grid : Message : 2373.927700 s : *********************************************************
+
+
+Grid : Message : 69.269319 s :  ApproxPowerMD shift[0]  pole    9.5166866092503e-06 residue -2.0047722631555e-08 tol     3e-06
+Grid : Message : 69.269321 s :  ApproxPowerMD shift[1]  pole    4.7123486192778e-05 residue -1.316766030683e-07 tol     1e-06
+Grid : Message : 69.269323 s :  ApproxPowerMD shift[2]  pole    0.00014860967743736 residue -6.109883117444e-07 tol     1e-07
+Grid : Message : 69.269325 s :  ApproxPowerMD shift[3]  pole    0.00041055696132763 residue -2.6088717433891e-06 tol     1e-07
+Grid : Message : 69.269327 s :  ApproxPowerMD shift[4]  pole    0.0010822555692906 residue -1.0853799412802e-05 tol     1e-08
+Grid : Message : 69.269329 s :  ApproxPowerMD shift[5]  pole    0.0028029613512087 residue -4.4741734470158e-05 tol     1e-08
+Grid : Message : 69.269331 s :  ApproxPowerMD shift[6]  pole    0.0072103567378527 residue -0.00018380499193253 tol     1e-08
+
+rusher 96I]$ more MobiusForce.221887
+      1.0e-5,3.0e-6,3.0e-7,1.0e-7, // soften convergence more more
+// <-- this is the dirichlet solve, why poorer conditioned???
+Grid : Message : 1627.226206 s : ConjugateGradientMultiShift k=3643 Shift 3 has converged
+Grid : Message : 1667.373045 s : ConjugateGradientMultiShift k=5381 Shift 2 has converged
+Grid : Message : 1705.236992 s : ConjugateGradientMultiShift k=7063 Shift 1 has converged
+Grid : Message : 1752.493182 s : ConjugateGradientMultiShift k=9220 Shift 0 has converged
+// 
+//Grid : Message : 1414.837250 s : OneFlavourEvenOddRatioRationalPseudoFermionAction deriv: doing (M^dag M)^{-1/2} ( (V^dag V)^{1/4} Phi)
+Grid : Message : 1523.416680 s : ConjugateGradientMultiShift k=3846 Shift 2 has converged
+Grid : Message : 1530.798503 s : ConjugateGradientMultiShift k=4143 Shift 1 has converged
+Grid : Message : 1536.153421 s : ConjugateGradientMultiShift k=4353 Shift 0 has converged <-- this is the non-dirichlet solve
+
+Grid : Message : 2339.927565 s : +++++++++++++++++++++++++++++++++++++++++++++++++++++++++
+Grid : Message : 2339.927571 s : S1 : 764404436.42251
+Grid : Message : 2339.927575 s : S2 : 764404436.4148
+Grid : Message : 2339.927577 s : dS : -0.0077102184295654
+Grid : Message : 2339.927579 s : dSpred : -0.0068752425267964
+Grid : Message : 2339.927580 s : diff : -0.00083497590276901
+Grid : Message : 2339.927581 s : *********************************************************
+Grid : Message : 2339.927582 s : Done
+Grid : Message : 2339.927582 s : *********************************************************
+
+Force 76 S {S {S {(9.0175185326468,-3.5764415623768e-36)}}}
+Force 77 S {S {S {(4.1289977678493,-4.3364721285803e-37)}}}
+Force 78 S {S {S {(3.2299269465841,6.0391022273495e-37)}}}
+Force 79 S {S {S {(3.0051199649288,-9.6243599973575e-37)}}}
+Force 80 S {S {S {(2.8924316727872,-1.3371248240604e-37)}}}
+Force 81 S {S {S {(2.8270868791781,1.792628885004e-37)}}}
+Force 82 S {S {S {(2.8676819960087,-1.3518185034456e-36)}}}
+Force 83 S {S {S {(2.7724152154523,1.4950818774521e-37)}}}
+Force 84 S {S {S {(3.0204624534964,-9.6475025423893e-36)}}}
+Force 85 S {S {S {(2.8631304063459,2.2426228161781e-37)}}}
+Force 86 S {S {S {(2.9025673908905,-1.3942465026706e-36)}}}
+Force 87 S {S {S {(2.8553405232646,-2.0938493124022e-38)}}}
+Force 88 S {S {S {(3.2820184381375,-1.422348164495e-36)}}}
+Force 89 S {S {S {(3.8974980085791,1.1682209795266e-35)}}}
+Force 90 S {S {S {(4.660053618223,-1.4399805797573e-37)}}}
+Force 91 S {S {S {(6.7993872372366,1.4524702072348e-36)}}}
+Full
+Grid : Message : 1523.416680 s : ConjugateGradientMultiShift k=3846 Shift 2 has converged
+Grid : Message : 1530.798503 s : ConjugateGradientMultiShift k=4143 Shift 1 has converged
+Grid : Message : 1536.153421 s : ConjugateGradientMultiShift k=4353 Shift 0 has converged
+PV solve depth 3
+Grid : Message : 1667.373045 s : ConjugateGradientMultiShift k=5381 Shift 2 has converged
+Grid : Message : 1705.236992 s : ConjugateGradientMultiShift k=7063 Shift 1 has converged
+Grid : Message : 1752.493182 s : ConjugateGradientMultiShift k=9220 Shift 0 has converged
+
+MobiusForce.222490 depth 1
+Grid : Message : 2155.595070 s : +++++++++++++++++++++++++++++++++++++++++++++++++++++++++
+Grid : Message : 2155.595076 s : S1 : 764404436.37475
+Grid : Message : 2155.595080 s : S2 : 764404436.21131
+Grid : Message : 2155.595082 s : dS : -0.16344606876373
+Grid : Message : 2155.595084 s : dSpred : -0.16235663327375
+Grid : Message : 2155.595085 s : diff : -0.0010894354899788
+
+Force 4 S {S {S {(24.512489110423,-7.4203080895657e-36)}}}
+Force 5 S {S {S {(14.442663101577,7.3909207307951e-37)}}}
+Force 6 S {S {S {(12.298567945213,2.1989091200069e-36)}}}
+Force 7 S {S {S {(11.582362859271,-2.2540104177017e-36)}}}
+Force 8 S {S {S {(11.465725500906,-2.9512255045332e-36)}}}
+Force 9 S {S {S {(10.869067954412,-2.8388188572358e-36)}}}
+Force 10 S {S {S {(10.937111429576,-3.3530976357206e-36)}}}
+Force 11 S {S {S {(11.23500117508,-1.4487967873885e-36)}}}
+Force 12 S {S {S {(10.900736551834,5.1427877848475e-36)}}} Force is bigger
+Force 13 S {S {S {(10.951921323651,-1.2098775605838e-35)}}}
+Force 14 S {S {S {(10.676529230575,-2.50527233519e-36)}}}
+Force 15 S {S {S {(10.98568474467,3.2193851533145e-36)}}}
+Force 16 S {S {S {(11.931707726568,-8.5223340434616e-37)}}}
+Force 17 S {S {S {(13.751904678482,7.6337337826369e-36)}}}
+Force 18 S {S {S {(17.518955473833,1.8073225643893e-36)}}}
+Force 19 S {S {S {(20.36519304598,-2.5184966466368e-36)}}}
+Full solve
+Grid : Message : 1441.297575 s : ConjugateGradientMultiShift k=3846 Shift 2 has converged
+Grid : Message : 1449.206520 s : ConjugateGradientMultiShift k=4143 Shift 1 has converged
+Grid : Message : 1454.352909 s : ConjugateGradientMultiShift k=4353 Shift 0 has converged
+
+Dirichlet solve -- why so expensive??
+Spectral radius worse?
+Grid : Message : 1571.887003 s : ConjugateGradientMultiShift k=5195 Shift 2 has converged
+Grid : Message : 1599.543760 s : ConjugateGradientMultiShift k=6508 Shift 1 has converged
+Grid : Message : 1625.368198 s : ConjugateGradientMultiShift k=7819 Shift 0 has converged
+
+
+dS is much bigger.
+
+
+MobiusForce.223606
+Grid : Message : 1123.276405 s : ConjugateGradientMultiShift k=3273 Shift 0 has converged
+Grid : Message : 1125.945359 s : ConjugateGradientMultiShift k=3407 Shift 1 has converged
+Grid : Message : 1127.896580 s : ConjugateGradientMultiShift k=3508 Shift 2 has converged <-- 2 takes longer
+first (bdy) hasenbusch mass raised to 0.005 -- reduces Dirchlet solve cost
+Force looks ok still
+Grid : Message : 1510.884960 s : OneFlavourEvenOddRatioRationalPseudoFermionAction compute action: complete
+Grid : Message : 1510.969380 s : +++++++++++++++++++++++++++++++++++++++++++++++++++++++++
+Grid : Message : 1510.969440 s : S1 : 764404436.37475
+Grid : Message : 1510.969480 s : S2 : 764404436.17379
+Grid : Message : 1510.969500 s : dS : -0.20095825195312
+Grid : Message : 1510.969520 s : dSpred : -0.20025674631954
+Grid : Message : 1510.969530 s : diff : -0.00070150563358654
+Force 76 S {S {S {(24.161229317675,2.0147973173094e-35)}}}
+Force 77 S {S {S {(15.841085162729,3.983456481349e-36)}}}
+Force 78 S {S {S {(11.031761776856,9.0394046210295e-35)}}}
+Force 79 S {S {S {(12.177830066719,1.583978637733e-36)}}}
+Force 80 S {S {S {(9.8372072482222,6.4284847310594e-37)}}}
+Force 81 S {S {S {(9.6588863493149,1.0501572656659e-35)}}}
+Force 82 S {S {S {(10.623076227724,-4.4161853392455e-35)}}}
+Force 83 S {S {S {(8.9477003784221,-7.067659784319e-37)}}}
+Force 84 S {S {S {(9.7663166497594,-2.1014900256825e-35)}}}
+Force 85 S {S {S {(8.9992648919057,-4.7107936109203e-36)}}}
+Force 86 S {S {S {(9.0399987268337,6.4652189295226e-37)}}}
+Force 87 S {S {S {(9.1319052497073,7.9566273871284e-37)}}}
+Force 88 S {S {S {(10.094569606113,-1.263656427134e-37)}}}
+Force 89 S {S {S {(11.563679905523,-1.2777623593438e-35)}}}
+Force 90 S {S {S {(13.653150474463,2.9093485182852e-37)}}}
+Force 91 S {S {S {(16.303719912019,2.9857556510886e-36)}}}
+
+MobiusForce.223749
+first (bdy) hasenbusch mass raised to 0.01 -- reduces Dirchlet solve cost
+Grid : Message : 1374.472462 s : OneFlavourEvenOddRatioRationalPseudoFermionAction compute action: complete
+Grid : Message : 1374.479206 s : +++++++++++++++++++++++++++++++++++++++++++++++++++++++++
+Grid : Message : 1374.479211 s : S1 : 764404436.37428
+Grid : Message : 1374.479215 s : S2 : 764404436.20009
+Grid : Message : 1374.479217 s : dS : -0.17418932914734
+Grid : Message : 1374.479219 s : dSpred : -0.17358090105485
+Grid : Message : 1374.479220 s : diff : -0.00060842809248995
+Force 76 S {S {S {(27.006858541753,4.2141472476979e-36)}}}
+Force 77 S {S {S {(19.388701462694,-5.1620365048422e-35)}}}
+Force 78 S {S {S {(13.502424539662,-2.4038859474316e-35)}}}
+Force 79 S {S {S {(15.555776987064,6.0567346426118e-36)}}}
+Force 80 S {S {S {(12.752116522904,-2.3720006631655e-35)}}}
+Force 81 S {S {S {(12.656857824233,1.6912424972456e-35)}}}
+Force 82 S {S {S {(15.159284452724,5.0898905390605e-36)}}}
+Force 83 S {S {S {(12.222695136014,-2.2061824913027e-35)}}}
+Force 84 S {S {S {(12.92077598466,9.6287681011731e-36)}}}
+Force 85 S {S {S {(11.884630495484,2.822655809912e-36)}}}
+Force 86 S {S {S {(11.896353116174,1.0926219990893e-35)}}}
+Force 87 S {S {S {(11.557019282287,2.1532117771187e-35)}}}
+Force 88 S {S {S {(11.945108384613,-3.0210204816133e-36)}}}
+Force 89 S {S {S {(13.295373801078,7.3115748621146e-36)}}}
+Force 90 S {S {S {(15.373728471417,-7.4923071185536e-36)}}}
+Force 91 S {S {S {(17.348173714234,1.0344350287236e-36)}}}
+
+MobiusForce.223829
+      1.0e-5,5.0e-6,1.0e-6,1.0e-7, // soften convergence more more
+Grid : Message : 1000.951387 s : ConjugateGradientMultiShift k=1881 Shift 0 has converged
+Grid : Message : 1002.619542 s : ConjugateGradientMultiShift k=1960 Shift 1 has converged
+Grid : Message : 1003.726982 s : ConjugateGradientMultiShift k=2014 Shift 4 has converged
+Grid : Message : 1005.698741 s : ConjugateGradientMultiShift k=2113 Shift 2 has converged
+Grid : Message : 1007.320875 s : ConjugateGradientMultiShift k=2197 Shift 3 has converged
+Grid : Message : 1351.171259 s : S1 : 764404436.37428
+Grid : Message : 1351.171263 s : S2 : 764404436.20009
+Grid : Message : 1351.171265 s : dS : -0.17418932914734
+Grid : Message : 1351.171266 s : dSpred : -0.1743248065338
+Grid : Message : 1351.171267 s : diff : 0.00013547738646566
+Force 76 S {S {S {(27.004288088317,6.035575744297e-35)}}}
+Force 77 S {S {S {(19.388023720604,-6.9736202362532e-36)}}}
+Force 78 S {S {S {(13.502663916173,6.4067380855692e-35)}}}
+Force 79 S {S {S {(15.55135748152,1.7219522871608e-35)}}}
+Force 80 S {S {S {(12.75135802213,-1.1303847551095e-35)}}}
+Force 81 S {S {S {(12.655732786276,1.689773129307e-36)}}}
+Force 82 S {S {S {(15.158469055699,-6.7205950772387e-35)}}}
+Force 83 S {S {S {(12.222907191126,-1.6775773754173e-35)}}}
+Force 84 S {S {S {(12.916025368247,-1.9641041234302e-35)}}}
+Force 85 S {S {S {(11.881879452577,-2.3054382955502e-36)}}}
+Force 86 S {S {S {(11.897253557199,-3.3617669065579e-35)}}}
+Force 87 S {S {S {(11.55717723524,-1.8690360178074e-36)}}}
+Force 88 S {S {S {(11.945590605851,-6.7208889508264e-36)}}}
+Force 89 S {S {S {(13.298173932749,-1.0322309768158e-35)}}}
+Force 90 S {S {S {(15.373845416836,7.4158999857501e-36)}}}
+Force 91 S {S {S {(17.348058307158,-1.8514036025451e-36)}}}
+-- could make the stopping condition mandatory if shift 0 is converged.
+-- Save 20% of iterations and single tunable
+*/
+
+//
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./tests/Test_double_ratio.cc
+
+    Copyright (C) 2022
+
+Author: Peter Boyle <pboyle@bnl.gov>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+    *************************************************************************************/
+    /*  END LEGAL */
+#include <Grid/Grid.h>
+
+using namespace std;
+using namespace Grid;
+
+typedef MobiusFermionD FermionAction;
+typedef WilsonImplD FimplD;
+typedef WilsonImplD FermionImplPolicy;
+
+template<class Gimpl>
+void ForceTest(Action<LatticeGaugeField> &action,LatticeGaugeField & U,MomentumFilterBase<LatticeGaugeField> &Filter)
+{
+  GridBase *UGrid = U.Grid();
+
+  std::vector<int> seeds({1,2,3,5});
+  GridSerialRNG            sRNG;         sRNG.SeedFixedIntegers(seeds);
+  GridParallelRNG          RNG4(UGrid);  RNG4.SeedFixedIntegers(seeds);
+
+  LatticeColourMatrix Pmu(UGrid); 
+  LatticeGaugeField P(UGrid); 
+  LatticeGaugeField UdSdU(UGrid); 
+
+  std::cout << GridLogMessage << "*********************************************************"<<std::endl;
+  std::cout << GridLogMessage << " Force test for "<<action.action_name()<<std::endl;
+  std::cout << GridLogMessage << "*********************************************************"<<std::endl;
+  
+  RealD eps=0.005;
+
+  std::cout << GridLogMessage << "+++++++++++++++++++++++++++++++++++++++++++++++++++++++++"<<std::endl;
+  std::cout << GridLogMessage << " Refresh "<<action.action_name()<<std::endl;
+  std::cout << GridLogMessage << "+++++++++++++++++++++++++++++++++++++++++++++++++++++++++"<<std::endl;
+  
+  Gimpl::generate_momenta(P,sRNG,RNG4);
+  Filter.applyFilter(P);
+
+  FieldMetaData header;
+  std::string file("./ckpoint_lat.2000");
+  NerscIO::readConfiguration(U,header,file);
+ 
+  action.refresh(U,sRNG,RNG4);
+
+  std::cout << GridLogMessage << "+++++++++++++++++++++++++++++++++++++++++++++++++++++++++"<<std::endl;
+  std::cout << GridLogMessage << " Action "<<action.action_name()<<std::endl;
+  std::cout << GridLogMessage << "+++++++++++++++++++++++++++++++++++++++++++++++++++++++++"<<std::endl;
+
+  RealD S1 = action.S(U);
+
+  Gimpl::update_field(P,U,eps);
+
+  std::cout << GridLogMessage << "+++++++++++++++++++++++++++++++++++++++++++++++++++++++++"<<std::endl;
+  std::cout << GridLogMessage << " Derivative "<<action.action_name()<<std::endl;
+  std::cout << GridLogMessage << "+++++++++++++++++++++++++++++++++++++++++++++++++++++++++"<<std::endl;
+  action.deriv(U,UdSdU);
+  UdSdU = Ta(UdSdU);
+  Filter.applyFilter(UdSdU);
+
+  DumpSliceNorm("Force",UdSdU,Nd-1);
+  
+  Gimpl::update_field(P,U,eps);
+  std::cout << GridLogMessage << "+++++++++++++++++++++++++++++++++++++++++++++++++++++++++"<<std::endl;
+  std::cout << GridLogMessage << " Action "<<action.action_name()<<std::endl;
+  std::cout << GridLogMessage << "+++++++++++++++++++++++++++++++++++++++++++++++++++++++++"<<std::endl;
+  
+  RealD S2 = action.S(U);
+
+  // Use the derivative
+  LatticeComplex dS(UGrid); dS = Zero();
+  for(int mu=0;mu<Nd;mu++){
+    auto UdSdUmu = PeekIndex<LorentzIndex>(UdSdU,mu);
+    Pmu= PeekIndex<LorentzIndex>(P,mu);
+    dS = dS - trace(Pmu*UdSdUmu)*eps*2.0*2.0;
+  }
+  ComplexD dSpred    = sum(dS);
+  RealD diff =  S2-S1-dSpred.real();
+
+  std::cout<< GridLogMessage << "+++++++++++++++++++++++++++++++++++++++++++++++++++++++++"<<std::endl;
+  std::cout<< GridLogMessage << "S1 : "<< S1    <<std::endl;
+  std::cout<< GridLogMessage << "S2 : "<< S2    <<std::endl;
+  std::cout<< GridLogMessage << "dS : "<< S2-S1 <<std::endl;
+  std::cout<< GridLogMessage << "dSpred : "<< dSpred.real() <<std::endl;
+  std::cout<< GridLogMessage << "diff : "<< diff<<std::endl;
+  std::cout<< GridLogMessage << "*********************************************************"<<std::endl;
+  //  assert(diff<1.0);
+  std::cout<< GridLogMessage << "Done" <<std::endl;
+  std::cout << GridLogMessage << "*********************************************************"<<std::endl;
+}
+
+int main (int argc, char ** argv)
+{
+  Grid_init(&argc,&argv);
+
+  std::cout << std::setprecision(14);
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate mpi_layout  = GridDefaultMpi();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate shm;
+  GlobalSharedMemory::GetShmDims(mpi_layout,shm);
+
+  const int Ls=12;
+  const int Nt = latt_size[3];
+  GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplex::Nsimd()),GridDefaultMpi());
+  GridRedBlackCartesian * UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
+  GridCartesian         * FGrid   = SpaceTimeGrid::makeFiveDimGrid(Ls,UGrid);
+  GridRedBlackCartesian * FrbGrid = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGrid);
+ 
+  ////////////////////////////////////////////////////////////////
+  // Domain decomposed operator
+  ////////////////////////////////////////////////////////////////
+  Coordinate CommDim(Nd);
+  for(int d=0;d<Nd;d++) CommDim[d]= (mpi_layout[d]/shm[d])>1 ? 1 : 0;
+
+  Coordinate NonDirichlet(Nd+1,0);
+  Coordinate Dirichlet(Nd+1,0);
+  Dirichlet[1] = CommDim[0]*latt_size[0]/mpi_layout[0] * shm[0];
+  Dirichlet[2] = CommDim[1]*latt_size[1]/mpi_layout[1] * shm[1];
+  Dirichlet[3] = CommDim[2]*latt_size[2]/mpi_layout[2] * shm[2];
+  Dirichlet[4] = CommDim[3]*latt_size[3]/mpi_layout[3] * shm[3];
+
+  Coordinate Block4(Nd);
+  Block4[0] = Dirichlet[1];
+  Block4[1] = Dirichlet[2];
+  Block4[2] = Dirichlet[3];
+  Block4[3] = Dirichlet[4];
+
+  std::vector<Complex> boundary = {1,1,1,-1};
+  FermionAction::ImplParams Params(boundary);
+  FermionAction::ImplParams ParamsDir(boundary);
+  Params.dirichlet=NonDirichlet;
+  ParamsDir.dirichlet=Dirichlet;
+  ParamsDir.partialDirichlet=1;
+
+  ///////////////////// Gauge Field and Gauge Forces ////////////////////////////
+  LatticeGaugeField U(UGrid);
+
+  RealD beta=6.0;
+  WilsonGaugeActionR PlaqAction(beta);
+  IwasakiGaugeActionR RectAction(beta);
+
+  MomentumFilterNone<LatticeGaugeField> FilterNone;
+  ForceTest<GimplTypesR>(PlaqAction,U,FilterNone);
+  ForceTest<GimplTypesR>(RectAction,U,FilterNone);
+
+  ////////////////////////////////////
+  // Action
+  ////////////////////////////////////
+  RealD mass=0.00078; 
+  RealD dmass=0.01;
+  RealD pvmass=1.0; 
+  RealD M5=1.8; 
+  RealD b=1.5;
+  RealD c=0.5;
+  
+  // Double versions
+  FermionAction DdwfPeriodic(U,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,b,c,Params);
+  FermionAction PVPeriodic  (U,*FGrid,*FrbGrid,*UGrid,*UrbGrid,pvmass,M5,b,c,Params);
+  FermionAction DdwfDirichlet(U,*FGrid,*FrbGrid,*UGrid,*UrbGrid,dmass,M5,b,c,ParamsDir);
+
+  double StoppingCondition = 1.0e-8;
+  double MaxCGIterations = 50000;
+  ConjugateGradient<LatticeFermion>  CG(StoppingCondition,MaxCGIterations);
+  
+  //////////////////// Two Flavour Determinant Ratio ///////////////////////////////
+  TwoFlavourRatioPseudoFermionAction<FimplD> Nf2(PVPeriodic, DdwfPeriodic,CG,CG);
+  //  ForceTest<GimplTypesR>(Nf2,U,FilterNone);
+
+  //////////////////// Two Flavour Determinant force test Even Odd ///////////////////////////////
+  TwoFlavourEvenOddRatioPseudoFermionAction<FimplD> Nf2eo(PVPeriodic, DdwfPeriodic,CG,CG);
+  //  ForceTest<GimplTypesR>(Nf2eo,U,FilterNone);
+
+  //////////////////// Domain forces ////////////////////
+  int Width=4;
+  DDHMCFilter<WilsonImplD::Field> DDHMCFilter(Block4,Width);
+  
+  //////////////////// Two flavour boundary det  ////////////////////
+  TwoFlavourRatioPseudoFermionAction<FimplD> BdyNf2(DdwfDirichlet, DdwfPeriodic,CG,CG);
+  //  ForceTest<GimplTypesR>(BdyNf2,U,DDHMCFilter);
+
+  //////////////////// Two flavour eo boundary det  ////////////////////
+  TwoFlavourEvenOddRatioPseudoFermionAction<FimplD> BdyNf2eo(DdwfDirichlet, DdwfPeriodic,CG,CG);
+  //  ForceTest<GimplTypesR>(BdyNf2eo,U,DDHMCFilter);
+
+  //////////////////// One flavour boundary det  ////////////////////
+  OneFlavourRationalParams OFRp; // Up/down
+  OFRp.lo       = 4.0e-5;
+  OFRp.hi       = 90.0;
+  OFRp.MaxIter  = 60000;
+  OFRp.tolerance= 1.0e-9;
+  OFRp.mdtolerance= 1.0e-8;
+  OFRp.degree   = 18;
+  OFRp.precision= 80;
+  OFRp.BoundsCheckFreq=0;
+  std::vector<RealD> ActionTolByPole({
+      1.0e-8,1.0e-8,1.0e-8,1.0e-8,
+      1.0e-8,1.0e-8,1.0e-8,1.0e-8,
+      1.0e-8,1.0e-8,1.0e-8,1.0e-8,
+      1.0e-8,1.0e-8,1.0e-8,1.0e-8,
+      1.0e-8,1.0e-8
+    });
+  std::vector<RealD> MDTolByPole({
+      1.0e-5,5.0e-6,1.0e-6,1.0e-7, // soften convergence more more
+      //      3.0e-6,1.0e-6,1.0e-7,1.0e-7, // soften convergence more
+      //      1.0e-6,3.0e-7,1.0e-7,1.0e-7,  // Orig sloppy
+      //      1.0e-8,1.0e-8,1.0e-8,1.0e-8,
+      1.0e-8,1.0e-8,1.0e-8,1.0e-8,
+      1.0e-8,1.0e-8,1.0e-8,1.0e-8,
+      1.0e-8,1.0e-8,1.0e-8,1.0e-8,
+      1.0e-8,1.0e-8
+    });
+  OneFlavourEvenOddRatioRationalPseudoFermionAction<FermionImplPolicy> BdySqrt(DdwfDirichlet,DdwfPeriodic,OFRp);
+  BdySqrt.SetTolerances(ActionTolByPole,MDTolByPole);
+  ForceTest<GimplTypesR>(BdySqrt,U,DDHMCFilter);
+
+  Grid_finalize();
+}

From 140684d706009df5525de9f5e00a4496b5986ec1 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 13 Dec 2022 08:15:38 -0500
Subject: [PATCH 108/146] Head to head vs HMC

---
 HMC/Mobius2p1f_EOFA_96I_hmc.cc | 492 +++++++++++++++++++++++++++++++++
 1 file changed, 492 insertions(+)
 create mode 100644 HMC/Mobius2p1f_EOFA_96I_hmc.cc

diff --git a/HMC/Mobius2p1f_EOFA_96I_hmc.cc b/HMC/Mobius2p1f_EOFA_96I_hmc.cc
new file mode 100644
index 00000000..d27d558e
--- /dev/null
+++ b/HMC/Mobius2p1f_EOFA_96I_hmc.cc
@@ -0,0 +1,492 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./tests/Test_hmc_EODWFRatio.cc
+
+Copyright (C) 2015-2016
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Guido Cossu <guido.cossu@ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/Grid.h>
+
+NAMESPACE_BEGIN(Grid);
+
+template<class FermionOperatorD, class FermionOperatorF, class SchurOperatorD, class  SchurOperatorF> 
+  class MixedPrecisionConjugateGradientOperatorFunction : public OperatorFunction<typename FermionOperatorD::FermionField> {
+  public:
+    typedef typename FermionOperatorD::FermionField FieldD;
+    typedef typename FermionOperatorF::FermionField FieldF;
+
+    using OperatorFunction<FieldD>::operator();
+
+    RealD   Tolerance;
+    RealD   InnerTolerance; //Initial tolerance for inner CG. Defaults to Tolerance but can be changed
+    Integer MaxInnerIterations;
+    Integer MaxOuterIterations;
+    GridBase* SinglePrecGrid4; //Grid for single-precision fields
+    GridBase* SinglePrecGrid5; //Grid for single-precision fields
+    RealD OuterLoopNormMult; //Stop the outer loop and move to a final double prec solve when the residual is OuterLoopNormMult * Tolerance
+
+    FermionOperatorF &FermOpF;
+    FermionOperatorD &FermOpD;;
+    SchurOperatorF &LinOpF;
+    SchurOperatorD &LinOpD;
+
+    Integer TotalInnerIterations; //Number of inner CG iterations
+    Integer TotalOuterIterations; //Number of restarts
+    Integer TotalFinalStepIterations; //Number of CG iterations in final patch-up step
+
+    MixedPrecisionConjugateGradientOperatorFunction(RealD tol, 
+						    Integer maxinnerit, 
+						    Integer maxouterit, 
+						    GridBase* _sp_grid4, 
+						    GridBase* _sp_grid5, 
+						    FermionOperatorF &_FermOpF,
+						    FermionOperatorD &_FermOpD,
+						    SchurOperatorF   &_LinOpF,
+						    SchurOperatorD   &_LinOpD): 
+      LinOpF(_LinOpF),
+      LinOpD(_LinOpD),
+      FermOpF(_FermOpF),
+      FermOpD(_FermOpD),
+      Tolerance(tol), 
+      InnerTolerance(tol), 
+      MaxInnerIterations(maxinnerit), 
+      MaxOuterIterations(maxouterit), 
+      SinglePrecGrid4(_sp_grid4),
+      SinglePrecGrid5(_sp_grid5),
+      OuterLoopNormMult(100.) 
+    { 
+      /* Debugging instances of objects; references are stored
+      std::cout << GridLogMessage << " Mixed precision CG wrapper LinOpF " <<std::hex<< &LinOpF<<std::dec <<std::endl;
+      std::cout << GridLogMessage << " Mixed precision CG wrapper LinOpD " <<std::hex<< &LinOpD<<std::dec <<std::endl;
+      std::cout << GridLogMessage << " Mixed precision CG wrapper FermOpF " <<std::hex<< &FermOpF<<std::dec <<std::endl;
+      std::cout << GridLogMessage << " Mixed precision CG wrapper FermOpD " <<std::hex<< &FermOpD<<std::dec <<std::endl;
+      */
+    };
+
+    void operator()(LinearOperatorBase<FieldD> &LinOpU, const FieldD &src, FieldD &psi) {
+
+      std::cout << GridLogMessage << " Mixed precision CG wrapper operator() "<<std::endl;
+
+      SchurOperatorD * SchurOpU = static_cast<SchurOperatorD *>(&LinOpU);
+      
+      //      std::cout << GridLogMessage << " Mixed precision CG wrapper operator() FermOpU " <<std::hex<< &(SchurOpU->_Mat)<<std::dec <<std::endl;
+      //      std::cout << GridLogMessage << " Mixed precision CG wrapper operator() FermOpD " <<std::hex<< &(LinOpD._Mat) <<std::dec <<std::endl;
+      // Assumption made in code to extract gauge field
+      // We could avoid storing LinopD reference alltogether ?
+      assert(&(SchurOpU->_Mat)==&(LinOpD._Mat));
+
+      ////////////////////////////////////////////////////////////////////////////////////
+      // Must snarf a single precision copy of the gauge field in Linop_d argument
+      ////////////////////////////////////////////////////////////////////////////////////
+      typedef typename FermionOperatorF::GaugeField GaugeFieldF;
+      typedef typename FermionOperatorF::GaugeLinkField GaugeLinkFieldF;
+      typedef typename FermionOperatorD::GaugeField GaugeFieldD;
+      typedef typename FermionOperatorD::GaugeLinkField GaugeLinkFieldD;
+
+      GridBase * GridPtrF = SinglePrecGrid4;
+      GridBase * GridPtrD = FermOpD.Umu.Grid();
+      GaugeFieldF     U_f  (GridPtrF);
+      GaugeLinkFieldF Umu_f(GridPtrF);
+      //      std::cout << " Dim gauge field "<<GridPtrF->Nd()<<std::endl; // 4d
+      //      std::cout << " Dim gauge field "<<GridPtrD->Nd()<<std::endl; // 4d
+
+      ////////////////////////////////////////////////////////////////////////////////////
+      // Moving this to a Clone method of fermion operator would allow to duplicate the 
+      // physics parameters and decrease gauge field copies
+      ////////////////////////////////////////////////////////////////////////////////////
+      GaugeLinkFieldD Umu_d(GridPtrD);
+      for(int mu=0;mu<Nd*2;mu++){ 
+	Umu_d = PeekIndex<LorentzIndex>(FermOpD.Umu, mu);
+	precisionChange(Umu_f,Umu_d);
+	PokeIndex<LorentzIndex>(FermOpF.Umu, Umu_f, mu);
+      }
+      pickCheckerboard(Even,FermOpF.UmuEven,FermOpF.Umu);
+      pickCheckerboard(Odd ,FermOpF.UmuOdd ,FermOpF.Umu);
+
+      ////////////////////////////////////////////////////////////////////////////////////
+      // Make a mixed precision conjugate gradient
+      ////////////////////////////////////////////////////////////////////////////////////
+#if 1
+      RealD delta=1.e-4;
+      std::cout << GridLogMessage << "Calling reliable update Conjugate Gradient" <<std::endl;
+      ConjugateGradientReliableUpdate<FieldD,FieldF> MPCG(Tolerance,MaxInnerIterations*MaxOuterIterations,delta,SinglePrecGrid5,LinOpF,LinOpD);
+#else      
+      std::cout << GridLogMessage << "Calling mixed precision Conjugate Gradient" <<std::endl;
+      MixedPrecisionConjugateGradient<FieldD,FieldF> MPCG(Tolerance,MaxInnerIterations,MaxOuterIterations,SinglePrecGrid5,LinOpF,LinOpD);
+#endif
+      MPCG(src,psi);
+    }
+  };
+
+NAMESPACE_END(Grid);
+
+
+int main(int argc, char **argv) {
+  using namespace Grid;
+
+  Grid_init(&argc, &argv);
+
+  CartesianCommunicator::BarrierWorld();
+  std::cout << GridLogMessage << " Clock skew check" <<std::endl;
+  
+  int threads = GridThread::GetThreads();
+
+   // Typedefs to simplify notation
+  typedef WilsonImplD FermionImplPolicy;
+  typedef MobiusFermionD FermionAction;
+  typedef MobiusEOFAFermionD FermionEOFAAction;
+  typedef typename FermionAction::FermionField FermionField;
+
+  typedef WilsonImplF FermionImplPolicyF;
+  typedef MobiusFermionF FermionActionF;
+  typedef MobiusEOFAFermionF FermionEOFAActionF;
+  typedef typename FermionActionF::FermionField FermionFieldF;
+
+  typedef WilsonImplD2 FermionImplPolicyD2;
+  typedef MobiusFermionD2 FermionActionD2;
+  typedef MobiusEOFAFermionD2 FermionEOFAActionD2;
+  typedef typename FermionActionD2::FermionField FermionFieldD2;
+
+  typedef Grid::XmlReader       Serialiser;
+
+  //::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
+  IntegratorParameters MD;
+  //  typedef GenericHMCRunner<LeapFrog> HMCWrapper;
+  //  MD.name    = std::string("Leap Frog");
+  typedef GenericHMCRunner<ForceGradient> HMCWrapper;
+  MD.name    = std::string("Force Gradient");
+  //typedef GenericHMCRunner<MinimumNorm2> HMCWrapper;
+  // MD.name    = std::string("MinimumNorm2");
+  // TrajL = 2
+  // 4/2 => 0.6 dH
+  // 3/3 => 0.8 dH .. depth 3, slower
+  //MD.MDsteps =  4;
+  MD.MDsteps =  3;
+  MD.trajL   = 0.5;
+
+  HMCparameters HMCparams;
+  HMCparams.StartTrajectory  = 1077;
+  HMCparams.Trajectories     = 1;
+  HMCparams.NoMetropolisUntil=  0;
+  // "[HotStart, ColdStart, TepidStart, CheckpointStart]\n";
+  //  HMCparams.StartingType     =std::string("ColdStart");
+  HMCparams.StartingType     =std::string("CheckpointStart");
+  HMCparams.MD = MD;
+  HMCWrapper TheHMC(HMCparams);
+
+  // Grid from the command line arguments --grid and --mpi
+  TheHMC.Resources.AddFourDimGrid("gauge"); // use default simd lanes decomposition
+
+  CheckpointerParameters CPparams;
+  CPparams.config_prefix = "ckpoint_DDHMC_lat";
+  CPparams.rng_prefix    = "ckpoint_DDHMC_rng";
+  CPparams.saveInterval  = 1;
+  CPparams.format        = "IEEE64BIG";
+  TheHMC.Resources.LoadNerscCheckpointer(CPparams);
+  std::cout << "loaded NERSC checpointer"<<std::endl;
+  RNGModuleParameters RNGpar;
+  RNGpar.serial_seeds = "1 2 3 4 5";
+  RNGpar.parallel_seeds = "6 7 8 9 10";
+  TheHMC.Resources.SetRNGSeeds(RNGpar);
+
+  // Construct observables
+  // here there is too much indirection
+  typedef PlaquetteMod<HMCWrapper::ImplPolicy> PlaqObs;
+  TheHMC.Resources.AddObservable<PlaqObs>();
+  //////////////////////////////////////////////
+
+  const int Ls      = 12;
+  RealD M5  = 1.8;
+  RealD b   = 1.5;
+  RealD c   = 0.5;
+  Real beta         = 2.13;
+  //  Real light_mass   = 5.4e-4;
+  Real light_mass     = 7.8e-4;
+  Real strange_mass = 0.0362;
+  Real pv_mass      = 1.0;
+  //  std::vector<Real> hasenbusch({ 0.01, 0.045, 0.108, 0.25, 0.51 , pv_mass });
+  //  std::vector<Real> hasenbusch({ light_mass, 0.01, 0.045, 0.108, 0.25, 0.51 , pv_mass });
+  std::vector<Real> hasenbusch({ light_mass, 0.005, 0.0145, 0.045, 0.108, 0.25, 0.51 , pv_mass }); // Updated
+  //  std::vector<Real> hasenbusch({ light_mass, 0.0145, 0.045, 0.108, 0.25, 0.51 , 0.75 , pv_mass });
+
+  auto GridPtr   = TheHMC.Resources.GetCartesian();
+  auto GridRBPtr = TheHMC.Resources.GetRBCartesian();
+
+  typedef SchurDiagMooeeOperator<FermionActionF,FermionFieldF> LinearOperatorF;
+  typedef SchurDiagMooeeOperator<FermionAction ,FermionField > LinearOperatorD;
+  typedef SchurDiagMooeeOperator<FermionEOFAActionF,FermionFieldF> LinearOperatorEOFAF;
+  typedef SchurDiagMooeeOperator<FermionEOFAAction ,FermionField > LinearOperatorEOFAD;
+  typedef MixedPrecisionConjugateGradientOperatorFunction<MobiusFermionD,MobiusFermionF,LinearOperatorD,LinearOperatorF> MxPCG;
+  typedef MixedPrecisionConjugateGradientOperatorFunction<MobiusEOFAFermionD,MobiusEOFAFermionF,LinearOperatorEOFAD,LinearOperatorEOFAF> MxPCG_EOFA;
+
+  ////////////////////////////////////////////////////////////////
+  // Domain decomposed
+  ////////////////////////////////////////////////////////////////
+  Coordinate latt4  = GridPtr->GlobalDimensions();
+  Coordinate mpi    = GridPtr->ProcessorGrid();
+  Coordinate shm;
+
+  GlobalSharedMemory::GetShmDims(mpi,shm);
+  
+  Coordinate CommDim(Nd);
+  for(int d=0;d<Nd;d++) CommDim[d]= (mpi[d]/shm[d])>1 ? 1 : 0;
+
+  Coordinate NonDirichlet(Nd+1,0);
+
+  //////////////////////////
+  // Fermion Grids
+  //////////////////////////
+  auto FGrid     = SpaceTimeGrid::makeFiveDimGrid(Ls,GridPtr);
+  auto FrbGrid   = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,GridPtr);
+
+  Coordinate simdF = GridDefaultSimd(Nd,vComplexF::Nsimd());
+  auto GridPtrF   = SpaceTimeGrid::makeFourDimGrid(latt4,simdF,mpi);
+  auto GridRBPtrF = SpaceTimeGrid::makeFourDimRedBlackGrid(GridPtrF);
+  auto FGridF     = SpaceTimeGrid::makeFiveDimGrid(Ls,GridPtrF);
+  auto FrbGridF   = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,GridPtrF);
+
+  IwasakiGaugeActionR GaugeAction(beta);
+
+  // temporarily need a gauge field
+  LatticeGaugeFieldD  U(GridPtr); U=Zero();
+  LatticeGaugeFieldF  UF(GridPtrF); UF=Zero();
+  LatticeGaugeFieldD2 UD2(GridPtrF); UD2=Zero();
+
+  std::cout << GridLogMessage << " Running the HMC "<< std::endl;
+  TheHMC.ReadCommandLine(argc,argv);  // params on CML or from param file
+  TheHMC.initializeGaugeFieldAndRNGs(U);
+  std::cout << "loaded NERSC gauge field"<<std::endl;
+
+  // These lines are unecessary if BC are all periodic
+  std::vector<Complex> boundary = {1,1,1,-1};
+  FermionAction::ImplParams Params(boundary);
+  FermionActionF::ImplParams ParamsF(boundary);
+  Params.dirichlet=NonDirichlet;
+  ParamsF.dirichlet=NonDirichlet;
+
+  //  double StoppingCondition = 1e-14;
+  //  double MDStoppingCondition = 1e-9;
+  double StoppingCondition = 1e-8;
+  double MDStoppingCondition = 1e-7;
+  double MDStoppingConditionLoose = 1e-7;
+  double MDStoppingConditionStrange = 1e-7;
+  double MaxCGIterations = 300000;
+  ConjugateGradient<FermionField>  CG(StoppingCondition,MaxCGIterations);
+  ConjugateGradient<FermionField>  MDCG(MDStoppingCondition,MaxCGIterations);
+
+  ////////////////////////////////////
+  // Collect actions
+  ////////////////////////////////////
+  ActionLevel<HMCWrapper::Field> Level1(1);
+  ActionLevel<HMCWrapper::Field> Level2(3);
+  ActionLevel<HMCWrapper::Field> Level3(15);
+
+  ////////////////////////////////////
+  // Strange action
+  ////////////////////////////////////
+  FermionAction StrangeOp (U,*FGrid,*FrbGrid,*GridPtr,*GridRBPtr,strange_mass,M5,b,c, Params);
+  FermionAction StrangePauliVillarsOp(U,*FGrid,*FrbGrid,*GridPtr,*GridRBPtr,pv_mass,  M5,b,c, Params);
+
+  // Probably dominates the force - back to EOFA.
+  OneFlavourRationalParams SFRp;
+  SFRp.lo       = 0.25;
+  SFRp.hi       = 25.0;
+  SFRp.MaxIter  = 10000;
+  SFRp.tolerance= 1.0e-5;
+  SFRp.mdtolerance= 2.0e-4;
+  SFRp.degree   = 8;
+  SFRp.precision= 50;
+  
+  MobiusEOFAFermionD Strange_Op_L (U , *FGrid , *FrbGrid , *GridPtr , *GridRBPtr , strange_mass, strange_mass, pv_mass, 0.0, -1, M5, b, c);
+  MobiusEOFAFermionF Strange_Op_LF(UF, *FGridF, *FrbGridF, *GridPtrF, *GridRBPtrF, strange_mass, strange_mass, pv_mass, 0.0, -1, M5, b, c);
+  MobiusEOFAFermionD Strange_Op_R (U , *FGrid , *FrbGrid , *GridPtr , *GridRBPtr , pv_mass, strange_mass,      pv_mass, -1.0, 1, M5, b, c);
+  MobiusEOFAFermionF Strange_Op_RF(UF, *FGridF, *FrbGridF, *GridPtrF, *GridRBPtrF, pv_mass, strange_mass,      pv_mass, -1.0, 1, M5, b, c);
+  ConjugateGradient<FermionField>      ActionCG(StoppingCondition,MaxCGIterations);
+  ConjugateGradient<FermionField>  DerivativeCG(MDStoppingCondition,MaxCGIterations);
+  LinearOperatorEOFAD Strange_LinOp_L (Strange_Op_L);
+  LinearOperatorEOFAD Strange_LinOp_R (Strange_Op_R);
+  LinearOperatorEOFAF Strange_LinOp_LF(Strange_Op_LF);
+  LinearOperatorEOFAF Strange_LinOp_RF(Strange_Op_RF);
+
+  const int MX_inner = 1000;
+  MxPCG_EOFA ActionCGL(StoppingCondition,
+		       MX_inner,
+		       MaxCGIterations,
+		       GridPtrF,
+		       FrbGridF,
+		       Strange_Op_LF,Strange_Op_L,
+		       Strange_LinOp_LF,Strange_LinOp_L);
+
+  MxPCG_EOFA DerivativeCGL(MDStoppingConditionStrange,
+			   MX_inner,
+			   MaxCGIterations,
+			   GridPtrF,
+			   FrbGridF,
+			   Strange_Op_LF,Strange_Op_L,
+			   Strange_LinOp_LF,Strange_LinOp_L);
+  
+  MxPCG_EOFA ActionCGR(StoppingCondition,
+		       MX_inner,
+		       MaxCGIterations,
+		       GridPtrF,
+		       FrbGridF,
+		       Strange_Op_RF,Strange_Op_R,
+		       Strange_LinOp_RF,Strange_LinOp_R);
+  
+  MxPCG_EOFA DerivativeCGR(MDStoppingConditionStrange,
+			   MX_inner,
+			   MaxCGIterations,
+			   GridPtrF,
+			   FrbGridF,
+			   Strange_Op_RF,Strange_Op_R,
+			   Strange_LinOp_RF,Strange_LinOp_R);
+
+  ExactOneFlavourRatioPseudoFermionAction<FermionImplPolicy> 
+    EOFA(Strange_Op_L, Strange_Op_R, 
+	 ActionCG, 
+	 ActionCGL, ActionCGR,
+	 DerivativeCGL, DerivativeCGR,
+	 SFRp, true);
+  //  Level2.push_back(&EOFA);
+
+  ////////////////////////////////////
+  // up down action
+  ////////////////////////////////////
+  std::vector<Real> light_den;
+  std::vector<Real> light_num;
+  std::vector<int> dirichlet_den;
+  std::vector<int> dirichlet_num;
+
+  int n_hasenbusch = hasenbusch.size();
+  light_den.push_back(light_mass);  dirichlet_den.push_back(0);
+  for(int h=0;h<n_hasenbusch;h++){
+    light_den.push_back(hasenbusch[h]); dirichlet_den.push_back(0);
+  }
+
+  for(int h=0;h<n_hasenbusch;h++){
+    light_num.push_back(hasenbusch[h]); dirichlet_num.push_back(0);
+  }
+  light_num.push_back(pv_mass);  dirichlet_num.push_back(0);
+
+  std::vector<FermionAction *> Numerators;
+  std::vector<FermionAction *> Denominators;
+  std::vector<FermionActionF *> NumeratorsF;
+  std::vector<FermionActionF *> DenominatorsF;
+  std::vector<FermionActionD2 *> NumeratorsD2;
+  std::vector<FermionActionD2 *> DenominatorsD2;
+  std::vector<TwoFlavourEvenOddRatioPseudoFermionAction<FermionImplPolicy> *> Quotients;
+  std::vector<MxPCG *> ActionMPCG;
+  std::vector<MxPCG *> MPCG;
+  
+#define MIXED_PRECISION
+#ifdef MIXED_PRECISION
+  std::vector<OneFlavourEvenOddRatioRationalMixedPrecPseudoFermionAction<FermionImplPolicy,FermionImplPolicyF,FermionImplPolicyD2> *> Bdys;
+#else
+  std::vector<OneFlavourEvenOddRatioRationalPseudoFermionAction<FermionImplPolicy> *> Bdys;
+#endif
+
+  typedef SchurDiagMooeeOperator<FermionActionF,FermionFieldF> LinearOperatorF;
+  typedef SchurDiagMooeeOperator<FermionAction ,FermionField > LinearOperatorD;
+  std::vector<LinearOperatorD *> LinOpD;
+  std::vector<LinearOperatorF *> LinOpF; 
+  
+  for(int h=0;h<n_hasenbusch+1;h++){
+    std::cout << GridLogMessage
+	      << " 2f quotient Action ";
+    std::cout << "det D("<<light_den[h]<<")";
+    if ( dirichlet_den[h] ) std::cout << "^dirichlet    ";
+    std::cout << "/ det D("<<light_num[h]<<")";
+    if ( dirichlet_num[h] ) std::cout << "^dirichlet    ";
+    std::cout << std::endl;
+
+    FermionAction::ImplParams ParamsNum(boundary);
+    FermionAction::ImplParams ParamsDen(boundary);
+    FermionActionF::ImplParams ParamsDenF(boundary);
+    FermionActionF::ImplParams ParamsNumF(boundary);
+    
+    ParamsNum.dirichlet = NonDirichlet;
+    ParamsDen.dirichlet = NonDirichlet;
+
+    ParamsNum.partialDirichlet = 0;
+    ParamsDen.partialDirichlet = 0;
+    
+    Numerators.push_back  (new FermionAction(U,*FGrid,*FrbGrid,*GridPtr,*GridRBPtr,light_num[h],M5,b,c, ParamsNum));
+    Denominators.push_back(new FermionAction(U,*FGrid,*FrbGrid,*GridPtr,*GridRBPtr,light_den[h],M5,b,c, ParamsDen));
+
+    ParamsDenF.dirichlet = ParamsDen.dirichlet;
+    ParamsDenF.partialDirichlet = ParamsDen.partialDirichlet;
+    DenominatorsF.push_back(new FermionActionF(UF,*FGridF,*FrbGridF,*GridPtrF,*GridRBPtrF,light_den[h],M5,b,c, ParamsDenF));
+
+    ParamsNumF.dirichlet = ParamsNum.dirichlet;
+    ParamsNumF.partialDirichlet = ParamsNum.partialDirichlet;
+    NumeratorsF.push_back  (new FermionActionF(UF,*FGridF,*FrbGridF,*GridPtrF,*GridRBPtrF,light_num[h],M5,b,c, ParamsNumF));
+
+    LinOpD.push_back(new LinearOperatorD(*Denominators[h]));
+    LinOpF.push_back(new LinearOperatorF(*DenominatorsF[h]));
+
+    double conv  = MDStoppingCondition;
+    if (h<3) conv= MDStoppingConditionLoose; // Relax on first two hasenbusch factors
+    const int MX_inner = 5000;
+    MPCG.push_back(new MxPCG(conv,
+			     MX_inner,
+			     MaxCGIterations,
+			     GridPtrF,
+			     FrbGridF,
+			     *DenominatorsF[h],*Denominators[h],
+			     *LinOpF[h], *LinOpD[h]) );
+
+    ActionMPCG.push_back(new MxPCG(StoppingCondition,
+				   MX_inner,
+				   MaxCGIterations,
+				   GridPtrF,
+				   FrbGridF,
+				   *DenominatorsF[h],*Denominators[h],
+				   *LinOpF[h], *LinOpD[h]) );
+
+    
+    Quotients.push_back (new TwoFlavourEvenOddRatioPseudoFermionAction<FermionImplPolicy>(*Numerators[h],*Denominators[h],*MPCG[h],*ActionMPCG[h],CG));
+  }
+  int nquo=Quotients.size();
+  for(int h=0;h<nquo;h++){
+    Level2.push_back(Quotients[h]);
+  }
+
+  /////////////////////////////////////////////////////////////
+  // Gauge action
+  /////////////////////////////////////////////////////////////
+  Level3.push_back(&GaugeAction);
+  TheHMC.TheAction.push_back(Level1);
+  TheHMC.TheAction.push_back(Level2);
+  TheHMC.TheAction.push_back(Level3);
+  std::cout << GridLogMessage << " Action complete "<< std::endl;
+  /////////////////////////////////////////////////////////////
+
+  TheHMC.Run();  // no smearing
+
+  Grid_finalize();
+} // main
+
+
+

From 4f85672674a018f1213af057e6e3b09ada600347 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Sat, 17 Dec 2022 20:16:11 -0500
Subject: [PATCH 109/146] Simpler test for PETSc

---
 tests/core/Test_fft_matt.cc | 160 ++++++++++++++++++++++++++++++++++++
 1 file changed, 160 insertions(+)
 create mode 100644 tests/core/Test_fft_matt.cc

diff --git a/tests/core/Test_fft_matt.cc b/tests/core/Test_fft_matt.cc
new file mode 100644
index 00000000..d4455a7e
--- /dev/null
+++ b/tests/core/Test_fft_matt.cc
@@ -0,0 +1,160 @@
+    /*************************************************************************************
+
+    grid` physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./tests/Test_cshift.cc
+
+    Copyright (C) 2015
+
+Author: Azusa Yamaguchi <ayamaguc@staffmail.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+    *************************************************************************************/
+    /*  END LEGAL */
+#include <Grid/Grid.h>
+
+using namespace Grid;
+ ;
+
+int main (int argc, char ** argv)
+{
+  Grid_init(&argc,&argv);
+
+  int threads = GridThread::GetThreads();
+  std::cout<<GridLogMessage << "Grid is setup to use "<<threads<<" threads"<<std::endl;
+
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplexD::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
+
+  int vol = 1;
+  for(int d=0;d<latt_size.size();d++){
+    vol = vol * latt_size[d];
+  }
+  GridCartesian         GRID(latt_size,simd_layout,mpi_layout);
+  GridRedBlackCartesian RBGRID(&GRID);
+
+  LatticeComplexD     one(&GRID);
+  LatticeComplexD      zz(&GRID);
+  LatticeComplexD       C(&GRID);
+  LatticeComplexD  Ctilde(&GRID);
+  LatticeComplexD  Cref  (&GRID);
+  LatticeComplexD  Csav  (&GRID);
+  LatticeComplexD    coor(&GRID);
+
+  LatticeSpinMatrixD    S(&GRID);
+  LatticeSpinMatrixD    Stilde(&GRID);
+  
+  Coordinate p({1,3,2,3});
+
+  one = ComplexD(1.0,0.0);
+  zz  = ComplexD(0.0,0.0);
+
+  ComplexD ci(0.0,1.0);
+
+  std::vector<int> seeds({1,2,3,4});
+  GridSerialRNG          sRNG;  sRNG.SeedFixedIntegers(seeds); // naughty seeding
+  GridParallelRNG          pRNG(&GRID);
+  pRNG.SeedFixedIntegers(seeds);
+
+  LatticeGaugeFieldD Umu(&GRID);
+
+  SU<Nc>::ColdConfiguration(pRNG,Umu); // Unit gauge
+
+  ////////////////////////////////////////////////////
+  // Wilson test
+  ////////////////////////////////////////////////////
+  {
+    LatticeFermionD    src(&GRID); gaussian(pRNG,src);
+    LatticeFermionD    tmp(&GRID);
+    LatticeFermionD    ref(&GRID);
+    
+    RealD mass=0.01;
+    WilsonFermionD Dw(Umu,GRID,RBGRID,mass);
+    
+    Dw.M(src,tmp);
+
+    std::cout << "Dw src = " <<norm2(src)<<std::endl;
+    std::cout << "Dw tmp = " <<norm2(tmp)<<std::endl;
+    
+    Dw.FreePropagator(tmp,ref,mass);
+
+    std::cout << "Dw ref = " <<norm2(ref)<<std::endl;
+    
+    ref = ref - src;
+    
+    std::cout << "Dw ref-src = " <<norm2(ref)<<std::endl;
+  }
+
+
+  ////////////////////////////////////////////////////
+  // Wilson prop
+  ////////////////////////////////////////////////////
+  {
+    std::cout<<"****************************************"<<std::endl;
+    std::cout << "Wilson Mom space 4d propagator \n";
+    std::cout<<"****************************************"<<std::endl;
+
+    LatticeFermionD    src(&GRID); gaussian(pRNG,src);
+    LatticeFermionD    tmp(&GRID);
+    LatticeFermionD    ref(&GRID);
+    LatticeFermionD    diff(&GRID);
+
+    src=Zero();
+    Coordinate point(4,0); // 0,0,0,0
+    SpinColourVectorD ferm;
+    ferm=Zero();
+    ferm()(0)(0) = ComplexD(1.0);
+    pokeSite(ferm,src,point);
+
+    RealD mass=0.01;
+    WilsonFermionD Dw(Umu,GRID,RBGRID,mass);
+
+    // Momentum space prop
+    std::cout << " Solving by FFT and Feynman rules" <<std::endl;
+    Dw.FreePropagator(src,ref,mass) ;
+
+    Gamma G5(Gamma::Algebra::Gamma5);
+
+    LatticeFermionD    result(&GRID); 
+    const int sdir=0;
+    
+    ////////////////////////////////////////////////////////////////////////
+    // Conjugate gradient on normal equations system
+    ////////////////////////////////////////////////////////////////////////
+    std::cout << " Solving by Conjugate Gradient (CGNE)" <<std::endl;
+    Dw.Mdag(src,tmp);
+    src=tmp;
+    MdagMLinearOperator<WilsonFermionD,LatticeFermionD> HermOp(Dw);
+    ConjugateGradient<LatticeFermionD> CG(1.0e-10,10000);
+    CG(HermOp,src,result);
+    
+    ////////////////////////////////////////////////////////////////////////
+    std::cout << " Taking difference" <<std::endl;
+    std::cout << "Dw result "<<norm2(result)<<std::endl;
+    std::cout << "Dw ref     "<<norm2(ref)<<std::endl;
+    
+    diff = ref - result;
+    std::cout << "result - ref     "<<norm2(diff)<<std::endl;
+
+    DumpSliceNorm("Slice Norm Solution ",result,Nd-1);
+  }
+
+  
+  Grid_finalize();
+}

From 281f8101fe8c679bfd52d149d53a669fdbd92c0d Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Sat, 17 Dec 2022 20:35:33 -0500
Subject: [PATCH 110/146] Matt FFT test

---
 tests/core/Test_fft_matt.cc | 160 ++++++++++++++++++++++++++++++++++++
 1 file changed, 160 insertions(+)
 create mode 100644 tests/core/Test_fft_matt.cc

diff --git a/tests/core/Test_fft_matt.cc b/tests/core/Test_fft_matt.cc
new file mode 100644
index 00000000..d4455a7e
--- /dev/null
+++ b/tests/core/Test_fft_matt.cc
@@ -0,0 +1,160 @@
+    /*************************************************************************************
+
+    grid` physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./tests/Test_cshift.cc
+
+    Copyright (C) 2015
+
+Author: Azusa Yamaguchi <ayamaguc@staffmail.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+    *************************************************************************************/
+    /*  END LEGAL */
+#include <Grid/Grid.h>
+
+using namespace Grid;
+ ;
+
+int main (int argc, char ** argv)
+{
+  Grid_init(&argc,&argv);
+
+  int threads = GridThread::GetThreads();
+  std::cout<<GridLogMessage << "Grid is setup to use "<<threads<<" threads"<<std::endl;
+
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplexD::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
+
+  int vol = 1;
+  for(int d=0;d<latt_size.size();d++){
+    vol = vol * latt_size[d];
+  }
+  GridCartesian         GRID(latt_size,simd_layout,mpi_layout);
+  GridRedBlackCartesian RBGRID(&GRID);
+
+  LatticeComplexD     one(&GRID);
+  LatticeComplexD      zz(&GRID);
+  LatticeComplexD       C(&GRID);
+  LatticeComplexD  Ctilde(&GRID);
+  LatticeComplexD  Cref  (&GRID);
+  LatticeComplexD  Csav  (&GRID);
+  LatticeComplexD    coor(&GRID);
+
+  LatticeSpinMatrixD    S(&GRID);
+  LatticeSpinMatrixD    Stilde(&GRID);
+  
+  Coordinate p({1,3,2,3});
+
+  one = ComplexD(1.0,0.0);
+  zz  = ComplexD(0.0,0.0);
+
+  ComplexD ci(0.0,1.0);
+
+  std::vector<int> seeds({1,2,3,4});
+  GridSerialRNG          sRNG;  sRNG.SeedFixedIntegers(seeds); // naughty seeding
+  GridParallelRNG          pRNG(&GRID);
+  pRNG.SeedFixedIntegers(seeds);
+
+  LatticeGaugeFieldD Umu(&GRID);
+
+  SU<Nc>::ColdConfiguration(pRNG,Umu); // Unit gauge
+
+  ////////////////////////////////////////////////////
+  // Wilson test
+  ////////////////////////////////////////////////////
+  {
+    LatticeFermionD    src(&GRID); gaussian(pRNG,src);
+    LatticeFermionD    tmp(&GRID);
+    LatticeFermionD    ref(&GRID);
+    
+    RealD mass=0.01;
+    WilsonFermionD Dw(Umu,GRID,RBGRID,mass);
+    
+    Dw.M(src,tmp);
+
+    std::cout << "Dw src = " <<norm2(src)<<std::endl;
+    std::cout << "Dw tmp = " <<norm2(tmp)<<std::endl;
+    
+    Dw.FreePropagator(tmp,ref,mass);
+
+    std::cout << "Dw ref = " <<norm2(ref)<<std::endl;
+    
+    ref = ref - src;
+    
+    std::cout << "Dw ref-src = " <<norm2(ref)<<std::endl;
+  }
+
+
+  ////////////////////////////////////////////////////
+  // Wilson prop
+  ////////////////////////////////////////////////////
+  {
+    std::cout<<"****************************************"<<std::endl;
+    std::cout << "Wilson Mom space 4d propagator \n";
+    std::cout<<"****************************************"<<std::endl;
+
+    LatticeFermionD    src(&GRID); gaussian(pRNG,src);
+    LatticeFermionD    tmp(&GRID);
+    LatticeFermionD    ref(&GRID);
+    LatticeFermionD    diff(&GRID);
+
+    src=Zero();
+    Coordinate point(4,0); // 0,0,0,0
+    SpinColourVectorD ferm;
+    ferm=Zero();
+    ferm()(0)(0) = ComplexD(1.0);
+    pokeSite(ferm,src,point);
+
+    RealD mass=0.01;
+    WilsonFermionD Dw(Umu,GRID,RBGRID,mass);
+
+    // Momentum space prop
+    std::cout << " Solving by FFT and Feynman rules" <<std::endl;
+    Dw.FreePropagator(src,ref,mass) ;
+
+    Gamma G5(Gamma::Algebra::Gamma5);
+
+    LatticeFermionD    result(&GRID); 
+    const int sdir=0;
+    
+    ////////////////////////////////////////////////////////////////////////
+    // Conjugate gradient on normal equations system
+    ////////////////////////////////////////////////////////////////////////
+    std::cout << " Solving by Conjugate Gradient (CGNE)" <<std::endl;
+    Dw.Mdag(src,tmp);
+    src=tmp;
+    MdagMLinearOperator<WilsonFermionD,LatticeFermionD> HermOp(Dw);
+    ConjugateGradient<LatticeFermionD> CG(1.0e-10,10000);
+    CG(HermOp,src,result);
+    
+    ////////////////////////////////////////////////////////////////////////
+    std::cout << " Taking difference" <<std::endl;
+    std::cout << "Dw result "<<norm2(result)<<std::endl;
+    std::cout << "Dw ref     "<<norm2(ref)<<std::endl;
+    
+    diff = ref - result;
+    std::cout << "result - ref     "<<norm2(diff)<<std::endl;
+
+    DumpSliceNorm("Slice Norm Solution ",result,Nd-1);
+  }
+
+  
+  Grid_finalize();
+}

From d8c29f5fcf321e4d1a4f0579cbdf698df481b2ce Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Sun, 18 Dec 2022 12:05:00 -0500
Subject: [PATCH 111/146] Updated FFT test for PETSc

---
 Grid/lattice/Lattice.h      |  1 +
 Grid/lattice/Lattice_crc.h  | 55 +++++++++++++++++++++
 tests/core/Test_fft_matt.cc | 95 ++++++++++++++++++++++++++++---------
 3 files changed, 129 insertions(+), 22 deletions(-)
 create mode 100644 Grid/lattice/Lattice_crc.h

diff --git a/Grid/lattice/Lattice.h b/Grid/lattice/Lattice.h
index 9f5f1da7..c4adf86a 100644
--- a/Grid/lattice/Lattice.h
+++ b/Grid/lattice/Lattice.h
@@ -46,3 +46,4 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 #include <Grid/lattice/Lattice_unary.h>
 #include <Grid/lattice/Lattice_transfer.h>
 #include <Grid/lattice/Lattice_basis.h>
+#include <Grid/lattice/Lattice_crc.h>
diff --git a/Grid/lattice/Lattice_crc.h b/Grid/lattice/Lattice_crc.h
new file mode 100644
index 00000000..142e2349
--- /dev/null
+++ b/Grid/lattice/Lattice_crc.h
@@ -0,0 +1,55 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/lattice/Lattice_crc.h
+
+    Copyright (C) 2021
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once
+
+NAMESPACE_BEGIN(Grid);
+
+template<class vobj> void DumpSliceNorm(std::string s,Lattice<vobj> &f,int mu=-1)
+{
+  auto ff = localNorm2(f);
+  if ( mu==-1 ) mu = f.Grid()->Nd()-1;
+  typedef typename vobj::tensor_reduced normtype;
+  typedef typename normtype::scalar_object scalar;
+  std::vector<scalar> sff;
+  sliceSum(ff,sff,mu);
+  for(int t=0;t<sff.size();t++){
+    std::cout << s<<" "<<t<<" "<<sff[t]<<std::endl;
+  }
+}
+
+template<class vobj> uint32_t crc(Lattice<vobj> & buf)
+{
+  autoView( buf_v , buf, CpuRead);
+  return ::crc32(0L,(unsigned char *)&buf_v[0],(size_t)sizeof(vobj)*buf.oSites());
+}
+
+#define CRC(U) std::cout << "FingerPrint "<<__FILE__ <<" "<< __LINE__ <<" "<< #U <<" "<<crc(U)<<std::endl;
+
+NAMESPACE_END(Grid);
+
+
diff --git a/tests/core/Test_fft_matt.cc b/tests/core/Test_fft_matt.cc
index d4455a7e..d92fa94e 100644
--- a/tests/core/Test_fft_matt.cc
+++ b/tests/core/Test_fft_matt.cc
@@ -1,5 +1,4 @@
     /*************************************************************************************
-
     grid` physics library, www.github.com/paboyle/Grid 
 
     Source file: ./tests/Test_cshift.cc
@@ -29,7 +28,14 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 #include <Grid/Grid.h>
 
 using namespace Grid;
- ;
+
+Gamma::Algebra Gmu [] = {
+  Gamma::Algebra::GammaX,
+  Gamma::Algebra::GammaY,
+  Gamma::Algebra::GammaZ,
+  Gamma::Algebra::GammaT,
+  Gamma::Algebra::Gamma5
+};
 
 int main (int argc, char ** argv)
 {
@@ -49,22 +55,7 @@ int main (int argc, char ** argv)
   GridCartesian         GRID(latt_size,simd_layout,mpi_layout);
   GridRedBlackCartesian RBGRID(&GRID);
 
-  LatticeComplexD     one(&GRID);
-  LatticeComplexD      zz(&GRID);
-  LatticeComplexD       C(&GRID);
-  LatticeComplexD  Ctilde(&GRID);
-  LatticeComplexD  Cref  (&GRID);
-  LatticeComplexD  Csav  (&GRID);
   LatticeComplexD    coor(&GRID);
-
-  LatticeSpinMatrixD    S(&GRID);
-  LatticeSpinMatrixD    Stilde(&GRID);
-  
-  Coordinate p({1,3,2,3});
-
-  one = ComplexD(1.0,0.0);
-  zz  = ComplexD(0.0,0.0);
-
   ComplexD ci(0.0,1.0);
 
   std::vector<int> seeds({1,2,3,4});
@@ -73,7 +64,6 @@ int main (int argc, char ** argv)
   pRNG.SeedFixedIntegers(seeds);
 
   LatticeGaugeFieldD Umu(&GRID);
-
   SU<Nc>::ColdConfiguration(pRNG,Umu); // Unit gauge
 
   ////////////////////////////////////////////////////
@@ -81,17 +71,78 @@ int main (int argc, char ** argv)
   ////////////////////////////////////////////////////
   {
     LatticeFermionD    src(&GRID); gaussian(pRNG,src);
+    LatticeFermionD    src_p(&GRID);
     LatticeFermionD    tmp(&GRID);
     LatticeFermionD    ref(&GRID);
+    LatticeFermionD    result(&GRID);
     
-    RealD mass=0.01;
+    RealD mass=0.1;
     WilsonFermionD Dw(Umu,GRID,RBGRID,mass);
     
-    Dw.M(src,tmp);
+    Dw.M(src,ref);
+    std::cout << "Norm src "<<norm2(src)<<std::endl;
+    std::cout << "Norm Dw x src "<<norm2(ref)<<std::endl;
+    {
+      FFT theFFT(&GRID);
 
+      ////////////////
+      // operator in Fourier space
+      ////////////////
+      tmp =ref;
+      theFFT.FFT_all_dim(result,tmp,FFT::forward);
+      std::cout<<"FFT[ Dw x src ]  "<< norm2(result)<<std::endl;    
+
+      tmp = src;
+      theFFT.FFT_all_dim(src_p,tmp,FFT::forward);
+      std::cout<<"FFT[ src      ]  "<< norm2(src_p)<<std::endl;
+      
+      /////////////////////////////////////////////////////////////////
+      // work out the predicted FT from Fourier
+      /////////////////////////////////////////////////////////////////
+      auto FGrid = &GRID;
+      LatticeFermionD    Kinetic(FGrid); Kinetic = Zero();
+      LatticeComplexD    kmu(FGrid); 
+      LatticeInteger     scoor(FGrid); 
+      LatticeComplexD    sk (FGrid); sk = Zero();
+      LatticeComplexD    sk2(FGrid); sk2= Zero();
+      LatticeComplexD    W(FGrid); W= Zero();
+      LatticeComplexD    one(FGrid); one =ComplexD(1.0,0.0);
+      ComplexD ci(0.0,1.0);
+    
+      for(int mu=0;mu<Nd;mu++) {
+	
+	RealD TwoPiL =  M_PI * 2.0/ latt_size[mu];
+
+	LatticeCoordinate(kmu,mu);
+
+	kmu = TwoPiL * kmu;
+      
+	sk2 = sk2 + 2.0*sin(kmu*0.5)*sin(kmu*0.5);
+	sk  = sk  +     sin(kmu)    *sin(kmu); 
+      
+	// -1/2 Dw ->  1/2 gmu (eip - emip) = i sinp gmu
+	Kinetic = Kinetic + sin(kmu)*ci*(Gamma(Gmu[mu])*src_p);
+	
+      }
+    
+      W = mass + sk2; 
+      Kinetic = Kinetic + W * src_p;
+    
+      std::cout<<"Momentum space src         "<< norm2(src_p)<<std::endl;
+      std::cout<<"Momentum space Dw x src    "<< norm2(Kinetic)<<std::endl;
+      std::cout<<"FT[Coordinate space Dw]    "<< norm2(result)<<std::endl;
+    
+      result = result - Kinetic;
+      std::cout<<"diff "<< norm2(result)<<std::endl;
+      
+    }
+
+    std::cout << " =======================================" <<std::endl;
+    std::cout << " Checking FourierFreePropagator x Dw = 1" <<std::endl;
+    std::cout << " =======================================" <<std::endl;
     std::cout << "Dw src = " <<norm2(src)<<std::endl;
     std::cout << "Dw tmp = " <<norm2(tmp)<<std::endl;
-    
+    Dw.M(src,tmp);
     Dw.FreePropagator(tmp,ref,mass);
 
     std::cout << "Dw ref = " <<norm2(ref)<<std::endl;
@@ -122,7 +173,7 @@ int main (int argc, char ** argv)
     ferm()(0)(0) = ComplexD(1.0);
     pokeSite(ferm,src,point);
 
-    RealD mass=0.01;
+    RealD mass=0.1;
     WilsonFermionD Dw(Umu,GRID,RBGRID,mass);
 
     // Momentum space prop

From 3791bc527b56c800c7e039b0e124691b76a1580c Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Wed, 30 Nov 2022 15:55:17 -0500
Subject: [PATCH 112/146] Logging pulled in from dirichlet branch

---
 Grid/allocator/MemoryManager.cc       |  2 +-
 Grid/allocator/MemoryManager.h        |  6 +++
 Grid/allocator/MemoryManagerCache.cc  | 69 ++++++++++++++++++++-------
 Grid/allocator/MemoryManagerShared.cc |  1 +
 Grid/log/Log.cc                       | 17 +++++--
 Grid/log/Log.h                        | 11 +++--
 6 files changed, 83 insertions(+), 23 deletions(-)

diff --git a/Grid/allocator/MemoryManager.cc b/Grid/allocator/MemoryManager.cc
index ef02f6aa..d055898f 100644
--- a/Grid/allocator/MemoryManager.cc
+++ b/Grid/allocator/MemoryManager.cc
@@ -40,7 +40,7 @@ void MemoryManager::PrintBytes(void)
 //////////////////////////////////////////////////////////////////////
 MemoryManager::AllocationCacheEntry MemoryManager::Entries[MemoryManager::NallocType][MemoryManager::NallocCacheMax];
 int MemoryManager::Victim[MemoryManager::NallocType];
-int MemoryManager::Ncache[MemoryManager::NallocType] = { 2, 8, 2, 8, 2, 8 };
+int MemoryManager::Ncache[MemoryManager::NallocType] = { 2, 8, 8, 16, 8, 16 };
 uint64_t MemoryManager::CacheBytes[MemoryManager::NallocType];
 //////////////////////////////////////////////////////////////////////
 // Actual allocation and deallocation utils
diff --git a/Grid/allocator/MemoryManager.h b/Grid/allocator/MemoryManager.h
index 740d8d92..ad2d9ffc 100644
--- a/Grid/allocator/MemoryManager.h
+++ b/Grid/allocator/MemoryManager.h
@@ -36,6 +36,11 @@ NAMESPACE_BEGIN(Grid);
 
 #define GRID_ALLOC_SMALL_LIMIT (4096)
 
+#define STRINGIFY(x) #x
+#define TOSTRING(x) STRINGIFY(x)
+#define FILE_LINE __FILE__ ":" TOSTRING(__LINE__)
+#define AUDIT(a) MemoryManager::Audit(FILE_LINE)
+
 /*Pinning pages is costly*/
 ////////////////////////////////////////////////////////////////////////////
 // Advise the LatticeAccelerator class
@@ -94,6 +99,7 @@ private:
 
   static void PrintBytes(void);
  public:
+  static void Audit(std::string s);
   static void Init(void);
   static void InitMessage(void);
   static void *AcceleratorAllocate(size_t bytes);
diff --git a/Grid/allocator/MemoryManagerCache.cc b/Grid/allocator/MemoryManagerCache.cc
index 04b3fe95..f03ee79f 100644
--- a/Grid/allocator/MemoryManagerCache.cc
+++ b/Grid/allocator/MemoryManagerCache.cc
@@ -3,8 +3,13 @@
 
 #warning "Using explicit device memory copies"
 NAMESPACE_BEGIN(Grid);
-//#define dprintf(...) printf ( __VA_ARGS__ ); fflush(stdout);
-#define dprintf(...)
+
+#define MAXLINE 512
+static char print_buffer [ MAXLINE ];
+
+#define mprintf(...) snprintf (print_buffer,MAXLINE, __VA_ARGS__ ); std::cout << GridLogMemory << print_buffer;
+#define dprintf(...) snprintf (print_buffer,MAXLINE, __VA_ARGS__ ); std::cout << GridLogMemory << print_buffer;
+//#define dprintf(...) 
 
 
 ////////////////////////////////////////////////////////////
@@ -104,7 +109,7 @@ void MemoryManager::AccDiscard(AcceleratorViewEntry &AccCache)
   ///////////////////////////////////////////////////////////
   assert(AccCache.state!=Empty);
   
-   dprintf("MemoryManager: Discard(%llx) %llx\n",(uint64_t)AccCache.CpuPtr,(uint64_t)AccCache.AccPtr); 
+  mprintf("MemoryManager: Discard(%lx) %lx\n",(uint64_t)AccCache.CpuPtr,(uint64_t)AccCache.AccPtr); 
   assert(AccCache.accLock==0);
   assert(AccCache.cpuLock==0);
   assert(AccCache.CpuPtr!=(uint64_t)NULL);
@@ -112,7 +117,7 @@ void MemoryManager::AccDiscard(AcceleratorViewEntry &AccCache)
     AcceleratorFree((void *)AccCache.AccPtr,AccCache.bytes);
     DeviceBytes   -=AccCache.bytes;
     LRUremove(AccCache);
-    dprintf("MemoryManager: Free(%llx) LRU %lld Total %lld\n",(uint64_t)AccCache.AccPtr,DeviceLRUBytes,DeviceBytes);  
+    dprintf("MemoryManager: Free(%lx) LRU %ld Total %ld\n",(uint64_t)AccCache.AccPtr,DeviceLRUBytes,DeviceBytes);  
   }
   uint64_t CpuPtr = AccCache.CpuPtr;
   EntryErase(CpuPtr);
@@ -126,9 +131,11 @@ void MemoryManager::Evict(AcceleratorViewEntry &AccCache)
   ///////////////////////////////////////////////////////////////////////////
   assert(AccCache.state!=Empty);
   
-  dprintf("MemoryManager: Evict(%llx) %llx\n",(uint64_t)AccCache.CpuPtr,(uint64_t)AccCache.AccPtr); 
-  assert(AccCache.accLock==0);
-  assert(AccCache.cpuLock==0);
+  mprintf("MemoryManager: Evict cpu %lx acc %lx cpuLock %ld accLock %ld\n",
+	  (uint64_t)AccCache.CpuPtr,(uint64_t)AccCache.AccPtr,
+	  (uint64_t)AccCache.cpuLock,(uint64_t)AccCache.accLock); 
+  if (AccCache.accLock!=0) return;
+  if (AccCache.cpuLock!=0) return;
   if(AccCache.state==AccDirty) {
     Flush(AccCache);
   }
@@ -137,7 +144,7 @@ void MemoryManager::Evict(AcceleratorViewEntry &AccCache)
     AcceleratorFree((void *)AccCache.AccPtr,AccCache.bytes);
     DeviceBytes   -=AccCache.bytes;
     LRUremove(AccCache);
-    dprintf("MemoryManager: Free(%llx) footprint now %lld \n",(uint64_t)AccCache.AccPtr,DeviceBytes);  
+    dprintf("MemoryManager: Free(%lx) footprint now %ld \n",(uint64_t)AccCache.AccPtr,DeviceBytes);  
   }
   uint64_t CpuPtr = AccCache.CpuPtr;
   EntryErase(CpuPtr);
@@ -150,7 +157,7 @@ void MemoryManager::Flush(AcceleratorViewEntry &AccCache)
   assert(AccCache.AccPtr!=(uint64_t)NULL);
   assert(AccCache.CpuPtr!=(uint64_t)NULL);
   acceleratorCopyFromDevice((void *)AccCache.AccPtr,(void *)AccCache.CpuPtr,AccCache.bytes);
-  dprintf("MemoryManager: Flush  %llx -> %llx\n",(uint64_t)AccCache.AccPtr,(uint64_t)AccCache.CpuPtr); fflush(stdout);
+  mprintf("MemoryManager: Flush  %lx -> %lx\n",(uint64_t)AccCache.AccPtr,(uint64_t)AccCache.CpuPtr); fflush(stdout);
   DeviceToHostBytes+=AccCache.bytes;
   DeviceToHostXfer++;
   AccCache.state=Consistent;
@@ -165,7 +172,7 @@ void MemoryManager::Clone(AcceleratorViewEntry &AccCache)
     AccCache.AccPtr=(uint64_t)AcceleratorAllocate(AccCache.bytes);
     DeviceBytes+=AccCache.bytes;
   }
-  dprintf("MemoryManager: Clone %llx <- %llx\n",(uint64_t)AccCache.AccPtr,(uint64_t)AccCache.CpuPtr); fflush(stdout);
+  mprintf("MemoryManager: Clone %lx <- %lx\n",(uint64_t)AccCache.AccPtr,(uint64_t)AccCache.CpuPtr); fflush(stdout);
   acceleratorCopyToDevice((void *)AccCache.CpuPtr,(void *)AccCache.AccPtr,AccCache.bytes);
   HostToDeviceBytes+=AccCache.bytes;
   HostToDeviceXfer++;
@@ -191,6 +198,7 @@ void MemoryManager::CpuDiscard(AcceleratorViewEntry &AccCache)
 void MemoryManager::ViewClose(void* Ptr,ViewMode mode)
 {
   if( (mode==AcceleratorRead)||(mode==AcceleratorWrite)||(mode==AcceleratorWriteDiscard) ){
+    dprintf("AcceleratorViewClose %lx\n",(uint64_t)Ptr);
     AcceleratorViewClose((uint64_t)Ptr);
   } else if( (mode==CpuRead)||(mode==CpuWrite)){
     CpuViewClose((uint64_t)Ptr);
@@ -202,6 +210,7 @@ void *MemoryManager::ViewOpen(void* _CpuPtr,size_t bytes,ViewMode mode,ViewAdvis
 {
   uint64_t CpuPtr = (uint64_t)_CpuPtr;
   if( (mode==AcceleratorRead)||(mode==AcceleratorWrite)||(mode==AcceleratorWriteDiscard) ){
+    dprintf("AcceleratorViewOpen %lx\n",(uint64_t)CpuPtr);
     return (void *) AcceleratorViewOpen(CpuPtr,bytes,mode,hint);
   } else if( (mode==CpuRead)||(mode==CpuWrite)){
     return (void *)CpuViewOpen(CpuPtr,bytes,mode,hint);
@@ -241,11 +250,12 @@ uint64_t MemoryManager::AcceleratorViewOpen(uint64_t CpuPtr,size_t bytes,ViewMod
   assert(AccCache.cpuLock==0);  // Programming error
 
   if(AccCache.state!=Empty) {
-    dprintf("ViewOpen found entry %llx %llx : %lld %lld\n",
+    dprintf("ViewOpen found entry %lx %lx : %ld %ld accLock %ld\n",
 		    (uint64_t)AccCache.CpuPtr,
 		    (uint64_t)CpuPtr,
 		    (uint64_t)AccCache.bytes,
-		    (uint64_t)bytes);
+	            (uint64_t)bytes,
+		    (uint64_t)AccCache.accLock);
     assert(AccCache.CpuPtr == CpuPtr);
     assert(AccCache.bytes  ==bytes);
   }
@@ -280,6 +290,7 @@ uint64_t MemoryManager::AcceleratorViewOpen(uint64_t CpuPtr,size_t bytes,ViewMod
       AccCache.state  = Consistent; // Empty + AccRead => Consistent
     }
     AccCache.accLock= 1;
+    dprintf("Copied Empty entry into device accLock= %d\n",AccCache.accLock);
   } else if(AccCache.state==CpuDirty ){
     if(mode==AcceleratorWriteDiscard) {
       CpuDiscard(AccCache);
@@ -292,21 +303,21 @@ uint64_t MemoryManager::AcceleratorViewOpen(uint64_t CpuPtr,size_t bytes,ViewMod
       AccCache.state  = Consistent; // CpuDirty + AccRead => Consistent
     }
     AccCache.accLock++;
-    dprintf("Copied CpuDirty entry into device accLock %d\n",AccCache.accLock);
+    dprintf("CpuDirty entry into device ++accLock= %d\n",AccCache.accLock);
   } else if(AccCache.state==Consistent) {
     if((mode==AcceleratorWrite)||(mode==AcceleratorWriteDiscard))
       AccCache.state  = AccDirty;   // Consistent + AcceleratorWrite=> AccDirty
     else
       AccCache.state  = Consistent; // Consistent + AccRead => Consistent
     AccCache.accLock++;
-    dprintf("Consistent entry into device accLock %d\n",AccCache.accLock);
+    dprintf("Consistent entry into device ++accLock= %d\n",AccCache.accLock);
   } else if(AccCache.state==AccDirty) {
     if((mode==AcceleratorWrite)||(mode==AcceleratorWriteDiscard))
       AccCache.state  = AccDirty; // AccDirty + AcceleratorWrite=> AccDirty
     else
       AccCache.state  = AccDirty; // AccDirty + AccRead => AccDirty
     AccCache.accLock++;
-    dprintf("AccDirty entry into device accLock %d\n",AccCache.accLock);
+    dprintf("AccDirty entry ++accLock= %d\n",AccCache.accLock);
   } else {
     assert(0);
   }
@@ -314,6 +325,7 @@ uint64_t MemoryManager::AcceleratorViewOpen(uint64_t CpuPtr,size_t bytes,ViewMod
   // If view is opened on device remove from LRU
   if(AccCache.LRU_valid==1){
     // must possibly remove from LRU as now locked on GPU
+    dprintf("AccCache entry removed from LRU \n");
     LRUremove(AccCache);
   }
 
@@ -334,10 +346,12 @@ void MemoryManager::AcceleratorViewClose(uint64_t CpuPtr)
   assert(AccCache.accLock>0);
 
   AccCache.accLock--;
-
   // Move to LRU queue if not locked and close on device
   if(AccCache.accLock==0) {
+    dprintf("AccleratorViewClose %lx AccLock decremented to %ld move to LRU queue\n",(uint64_t)CpuPtr,(uint64_t)AccCache.accLock);
     LRUinsert(AccCache);
+  } else {
+    dprintf("AccleratorViewClose %lx AccLock decremented to %ld\n",(uint64_t)CpuPtr,(uint64_t)AccCache.accLock);
   }
 }
 void MemoryManager::CpuViewClose(uint64_t CpuPtr)
@@ -473,6 +487,29 @@ int   MemoryManager::isOpen   (void* _CpuPtr)
     return 0;
   }
 }
+void MemoryManager::Audit(std::string s)
+{
+  for(auto it=AccViewTable.begin();it!=AccViewTable.end();it++){
+    auto &AccCache = it->second;
+    
+    std::string str;
+    if ( AccCache.state==Empty    ) str = std::string("Empty");
+    if ( AccCache.state==CpuDirty ) str = std::string("CpuDirty");
+    if ( AccCache.state==AccDirty ) str = std::string("AccDirty");
+    if ( AccCache.state==Consistent)str = std::string("Consistent");
+
+    if ( AccCache.cpuLock || AccCache.accLock ) { 
+      std::cout << GridLogError << s<< "\n\t 0x"<<std::hex<<AccCache.CpuPtr<<std::dec
+		<< "\t0x"<<std::hex<<AccCache.AccPtr<<std::dec<<"\t" <<str
+		<< "\t cpuLock  " << AccCache.cpuLock
+		<< "\t accLock  " << AccCache.accLock
+		<< "\t LRUvalid " << AccCache.LRU_valid<<std::endl;
+    }
+
+    assert( AccCache.cpuLock== 0 ) ;
+    assert( AccCache.accLock== 0 ) ;
+  }
+}
 
 void MemoryManager::PrintState(void* _CpuPtr)
 {
diff --git a/Grid/allocator/MemoryManagerShared.cc b/Grid/allocator/MemoryManagerShared.cc
index c072873b..2434ad47 100644
--- a/Grid/allocator/MemoryManagerShared.cc
+++ b/Grid/allocator/MemoryManagerShared.cc
@@ -13,6 +13,7 @@ uint64_t  MemoryManager::DeviceToHostBytes;
 uint64_t  MemoryManager::HostToDeviceXfer;
 uint64_t  MemoryManager::DeviceToHostXfer;
 
+void  MemoryManager::Audit(void){};
 void  MemoryManager::ViewClose(void* AccPtr,ViewMode mode){};
 void *MemoryManager::ViewOpen(void* CpuPtr,size_t bytes,ViewMode mode,ViewAdvise hint){ return CpuPtr; };
 int   MemoryManager::isOpen   (void* CpuPtr) { return 0;}
diff --git a/Grid/log/Log.cc b/Grid/log/Log.cc
index 9302b4cc..166aea0a 100644
--- a/Grid/log/Log.cc
+++ b/Grid/log/Log.cc
@@ -65,29 +65,40 @@ GridLogger GridLogSolver (1, "Solver", GridLogColours, "NORMAL");
 GridLogger GridLogError  (1, "Error" , GridLogColours, "RED");
 GridLogger GridLogWarning(1, "Warning", GridLogColours, "YELLOW");
 GridLogger GridLogMessage(1, "Message", GridLogColours, "NORMAL");
+GridLogger GridLogMemory (1, "Memory", GridLogColours, "NORMAL");
+GridLogger GridLogTracing(1, "Tracing", GridLogColours, "NORMAL");
 GridLogger GridLogDebug  (1, "Debug", GridLogColours, "PURPLE");
 GridLogger GridLogPerformance(1, "Performance", GridLogColours, "GREEN");
+GridLogger GridLogDslash     (1, "Dslash", GridLogColours, "BLUE");
 GridLogger GridLogIterative  (1, "Iterative", GridLogColours, "BLUE");
 GridLogger GridLogIntegrator (1, "Integrator", GridLogColours, "BLUE");
+GridLogger GridLogHMC (1, "HMC", GridLogColours, "BLUE");
 
 void GridLogConfigure(std::vector<std::string> &logstreams) {
-  GridLogError.Active(0);
+  GridLogError.Active(1);
   GridLogWarning.Active(0);
   GridLogMessage.Active(1); // at least the messages should be always on
+  GridLogMemory.Active(0); 
+  GridLogTracing.Active(0); 
   GridLogIterative.Active(0);
   GridLogDebug.Active(0);
   GridLogPerformance.Active(0);
+  GridLogDslash.Active(0);
   GridLogIntegrator.Active(1);
   GridLogColours.Active(0);
+  GridLogHMC.Active(1);
 
   for (int i = 0; i < logstreams.size(); i++) {
-    if (logstreams[i] == std::string("Error"))       GridLogError.Active(1);
+    if (logstreams[i] == std::string("Tracing"))     GridLogTracing.Active(1);
+    if (logstreams[i] == std::string("Memory"))      GridLogMemory.Active(1);
     if (logstreams[i] == std::string("Warning"))     GridLogWarning.Active(1);
     if (logstreams[i] == std::string("NoMessage"))   GridLogMessage.Active(0);
     if (logstreams[i] == std::string("Iterative"))   GridLogIterative.Active(1);
     if (logstreams[i] == std::string("Debug"))       GridLogDebug.Active(1);
     if (logstreams[i] == std::string("Performance")) GridLogPerformance.Active(1);
-    if (logstreams[i] == std::string("Integrator"))  GridLogIntegrator.Active(1);
+    if (logstreams[i] == std::string("Dslash"))      GridLogDslash.Active(1);
+    if (logstreams[i] == std::string("NoIntegrator"))GridLogIntegrator.Active(0);
+    if (logstreams[i] == std::string("NoHMC"))       GridLogHMC.Active(0);
     if (logstreams[i] == std::string("Colours"))     GridLogColours.Active(1);
   }
 }
diff --git a/Grid/log/Log.h b/Grid/log/Log.h
index 68693647..2d663a3c 100644
--- a/Grid/log/Log.h
+++ b/Grid/log/Log.h
@@ -138,7 +138,8 @@ public:
         stream << std::setw(log.topWidth);
       }
       stream << log.topName << log.background()<< " : ";
-      stream << log.colour() <<  std::left;
+      //      stream << log.colour() <<  std::left;
+      stream <<  std::left;
       if (log.chanWidth > 0)
       {
         stream << std::setw(log.chanWidth);
@@ -153,9 +154,9 @@ public:
 	stream << log.evidence()
 	       << now	       << log.background() << " : " ;
       }
-      stream << log.colour();
+      //      stream << log.colour();
+      stream <<  std::right;
       stream.flags(f);
-
       return stream;
     } else { 
       return devnull;
@@ -180,8 +181,12 @@ extern GridLogger GridLogWarning;
 extern GridLogger GridLogMessage;
 extern GridLogger GridLogDebug  ;
 extern GridLogger GridLogPerformance;
+extern GridLogger GridLogDslash;
 extern GridLogger GridLogIterative  ;
 extern GridLogger GridLogIntegrator  ;
+extern GridLogger GridLogHMC;
+extern GridLogger GridLogMemory;
+extern GridLogger GridLogTracing;
 extern Colours    GridLogColours;
 
 std::string demangle(const char* name) ;

From b00a4142e5f919847285caeac13187fe3f6e10b2 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Thu, 1 Dec 2022 00:18:11 -0500
Subject: [PATCH 113/146] A=A fix

---
 Grid/lattice/Lattice_base.h | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

diff --git a/Grid/lattice/Lattice_base.h b/Grid/lattice/Lattice_base.h
index 9c3d723f..34f13fa6 100644
--- a/Grid/lattice/Lattice_base.h
+++ b/Grid/lattice/Lattice_base.h
@@ -129,7 +129,7 @@ public:
     
     auto exprCopy = expr;
     ExpressionViewOpen(exprCopy);
-    auto me  = View(AcceleratorWriteDiscard);
+    auto me  = View(AcceleratorWrite);
     accelerator_for(ss,me.size(),vobj::Nsimd(),{
       auto tmp = eval(ss,exprCopy);
       coalescedWrite(me[ss],tmp);
@@ -152,7 +152,7 @@ public:
 
     auto exprCopy = expr;
     ExpressionViewOpen(exprCopy);
-    auto me  = View(AcceleratorWriteDiscard);
+    auto me  = View(AcceleratorWrite);
     accelerator_for(ss,me.size(),vobj::Nsimd(),{
       auto tmp = eval(ss,exprCopy);
       coalescedWrite(me[ss],tmp);
@@ -174,7 +174,7 @@ public:
     this->checkerboard=cb;
     auto exprCopy = expr;
     ExpressionViewOpen(exprCopy);
-    auto me  = View(AcceleratorWriteDiscard);
+    auto me  = View(AcceleratorWrite);
     accelerator_for(ss,me.size(),vobj::Nsimd(),{
       auto tmp = eval(ss,exprCopy);
       coalescedWrite(me[ss],tmp);
@@ -245,7 +245,7 @@ public:
   ///////////////////////////////////////////
   // user defined constructor
   ///////////////////////////////////////////
-  Lattice(GridBase *grid,ViewMode mode=AcceleratorWriteDiscard) { 
+  Lattice(GridBase *grid,ViewMode mode=AcceleratorWrite) { 
     this->_grid = grid;
     resize(this->_grid->oSites());
     assert((((uint64_t)&this->_odata[0])&0xF) ==0);
@@ -288,7 +288,7 @@ public:
     typename std::enable_if<!std::is_same<robj,vobj>::value,int>::type i=0;
     conformable(*this,r);
     this->checkerboard = r.Checkerboard();
-    auto me =   View(AcceleratorWriteDiscard);
+    auto me =   View(AcceleratorWrite);
     auto him= r.View(AcceleratorRead);
     accelerator_for(ss,me.size(),vobj::Nsimd(),{
       coalescedWrite(me[ss],him(ss));
@@ -303,7 +303,7 @@ public:
   inline Lattice<vobj> & operator = (const Lattice<vobj> & r){
     this->checkerboard = r.Checkerboard();
     conformable(*this,r);
-    auto me =   View(AcceleratorWriteDiscard);
+    auto me =   View(AcceleratorWrite);
     auto him= r.View(AcceleratorRead);
     accelerator_for(ss,me.size(),vobj::Nsimd(),{
       coalescedWrite(me[ss],him(ss));

From 43a45ec97bf0212844dc2bc8df9774f534bc29ba Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Thu, 1 Dec 2022 00:18:43 -0500
Subject: [PATCH 114/146] SSC_START

---
 Grid/perfmon/PerfCount.h | 14 ++++++--------
 1 file changed, 6 insertions(+), 8 deletions(-)

diff --git a/Grid/perfmon/PerfCount.h b/Grid/perfmon/PerfCount.h
index dd25b41e..62b2a740 100644
--- a/Grid/perfmon/PerfCount.h
+++ b/Grid/perfmon/PerfCount.h
@@ -30,6 +30,12 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 #ifndef GRID_PERFCOUNT_H
 #define GRID_PERFCOUNT_H
 
+
+#ifndef __SSC_START
+#define __SSC_START
+#define __SSC_STOP
+#endif
+
 #include <sys/time.h>
 #include <ctime>
 #include <chrono>
@@ -72,17 +78,9 @@ static long perf_event_open(struct perf_event_attr *hw_event, pid_t pid,
 inline uint64_t cyclecount(void){ 
   return 0;
 }
-#define __SSC_MARK(mark) __asm__ __volatile__ ("movl %0, %%ebx; .byte 0x64, 0x67, 0x90 " ::"i"(mark):"%ebx")
-#define __SSC_STOP  __SSC_MARK(0x110)
-#define __SSC_START __SSC_MARK(0x111)
-
 
 #else
 
-#define __SSC_MARK(mark) 
-#define __SSC_STOP  
-#define __SSC_START 
-
 /*
  * cycle counters arch dependent
  */

From 99b3697b031b201a2dbe11c1942e672bfc846f05 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Thu, 1 Dec 2022 00:19:33 -0500
Subject: [PATCH 115/146] More loggin

---
 Grid/allocator/MemoryManager.h | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/Grid/allocator/MemoryManager.h b/Grid/allocator/MemoryManager.h
index ad2d9ffc..c22a54f3 100644
--- a/Grid/allocator/MemoryManager.h
+++ b/Grid/allocator/MemoryManager.h
@@ -97,8 +97,8 @@ private:
   static void *Insert(void *ptr,size_t bytes,AllocationCacheEntry *entries,int ncache,int &victim,uint64_t &cbytes) ;
   static void *Lookup(size_t bytes,AllocationCacheEntry *entries,int ncache,uint64_t &cbytes) ;
 
-  static void PrintBytes(void);
  public:
+  static void PrintBytes(void);
   static void Audit(std::string s);
   static void Init(void);
   static void InitMessage(void);
@@ -119,6 +119,8 @@ private:
   static uint64_t     DeviceToHostBytes;
   static uint64_t     HostToDeviceXfer;
   static uint64_t     DeviceToHostXfer;
+  static uint64_t     DeviceEvictions;
+  static uint64_t     DeviceDestroy;
  
  private:
 #ifndef GRID_UVM
@@ -176,6 +178,7 @@ private:
 
  public:
   static void Print(void);
+  static void PrintAll(void);
   static void PrintState( void* CpuPtr);
   static int   isOpen   (void* CpuPtr);
   static void  ViewClose(void* CpuPtr,ViewMode mode);

From 37ba32776f410af38b43ddceff009be1c0cfcc62 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Thu, 1 Dec 2022 00:19:42 -0500
Subject: [PATCH 116/146] More logging

---
 Grid/allocator/MemoryManagerShared.cc | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/Grid/allocator/MemoryManagerShared.cc b/Grid/allocator/MemoryManagerShared.cc
index 2434ad47..ba95420e 100644
--- a/Grid/allocator/MemoryManagerShared.cc
+++ b/Grid/allocator/MemoryManagerShared.cc
@@ -12,8 +12,10 @@ uint64_t  MemoryManager::HostToDeviceBytes;
 uint64_t  MemoryManager::DeviceToHostBytes;
 uint64_t  MemoryManager::HostToDeviceXfer;
 uint64_t  MemoryManager::DeviceToHostXfer;
+uint64_t  MemoryManager::DeviceEvictions;
+uint64_t  MemoryManager::DeviceDestroy;
 
-void  MemoryManager::Audit(void){};
+void  MemoryManager::Audit(std::string s){};
 void  MemoryManager::ViewClose(void* AccPtr,ViewMode mode){};
 void *MemoryManager::ViewOpen(void* CpuPtr,size_t bytes,ViewMode mode,ViewAdvise hint){ return CpuPtr; };
 int   MemoryManager::isOpen   (void* CpuPtr) { return 0;}
@@ -22,6 +24,7 @@ void  MemoryManager::PrintState(void* CpuPtr)
 std::cout << GridLogMessage << "Host<->Device memory movement not currently managed by Grid." << std::endl;
 };
 void  MemoryManager::Print(void){};
+void  MemoryManager::PrintAll(void){};
 void  MemoryManager::NotifyDeletion(void *ptr){};
 
 NAMESPACE_END(Grid);

From 1822ced302f08435035921438cb414c364a98898 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Thu, 1 Dec 2022 00:24:08 -0500
Subject: [PATCH 117/146] Bug fix

---
 Grid/allocator/MemoryManagerCache.cc | 116 ++++++++++++++++++++-------
 1 file changed, 86 insertions(+), 30 deletions(-)

diff --git a/Grid/allocator/MemoryManagerCache.cc b/Grid/allocator/MemoryManagerCache.cc
index f03ee79f..3420c9cc 100644
--- a/Grid/allocator/MemoryManagerCache.cc
+++ b/Grid/allocator/MemoryManagerCache.cc
@@ -28,6 +28,8 @@ uint64_t  MemoryManager::HostToDeviceBytes;
 uint64_t  MemoryManager::DeviceToHostBytes;
 uint64_t  MemoryManager::HostToDeviceXfer;
 uint64_t  MemoryManager::DeviceToHostXfer;
+uint64_t  MemoryManager::DeviceEvictions;
+uint64_t  MemoryManager::DeviceDestroy;
 
 ////////////////////////////////////
 // Priority ordering for unlocked entries
@@ -115,8 +117,10 @@ void MemoryManager::AccDiscard(AcceleratorViewEntry &AccCache)
   assert(AccCache.CpuPtr!=(uint64_t)NULL);
   if(AccCache.AccPtr) {
     AcceleratorFree((void *)AccCache.AccPtr,AccCache.bytes);
+    DeviceDestroy++;
     DeviceBytes   -=AccCache.bytes;
     LRUremove(AccCache);
+    AccCache.AccPtr=(uint64_t) NULL;
     dprintf("MemoryManager: Free(%lx) LRU %ld Total %ld\n",(uint64_t)AccCache.AccPtr,DeviceLRUBytes,DeviceBytes);  
   }
   uint64_t CpuPtr = AccCache.CpuPtr;
@@ -126,28 +130,36 @@ void MemoryManager::AccDiscard(AcceleratorViewEntry &AccCache)
 void MemoryManager::Evict(AcceleratorViewEntry &AccCache)
 {
   ///////////////////////////////////////////////////////////////////////////
-  // Make CPU consistent, remove from Accelerator, remove entry
-  // Cannot be locked. If allocated must be in LRU pool.
+  // Make CPU consistent, remove from Accelerator, remove from LRU, LEAVE CPU only entry
+  // Cannot be acclocked. If allocated must be in LRU pool.
+  //
+  // Nov 2022... Felix issue: Allocating two CpuPtrs, can have an entry in LRU-q with CPUlock.
+  //                          and require to evict the AccPtr copy. Eviction was a mistake in CpuViewOpen
+  //                          but there is a weakness where CpuLock entries are attempted for erase
+  //                          Take these OUT LRU queue when CPU locked?
+  //                          Cannot take out the table as cpuLock data is important.
   ///////////////////////////////////////////////////////////////////////////
   assert(AccCache.state!=Empty);
   
   mprintf("MemoryManager: Evict cpu %lx acc %lx cpuLock %ld accLock %ld\n",
 	  (uint64_t)AccCache.CpuPtr,(uint64_t)AccCache.AccPtr,
 	  (uint64_t)AccCache.cpuLock,(uint64_t)AccCache.accLock); 
-  if (AccCache.accLock!=0) return;
-  if (AccCache.cpuLock!=0) return;
+  assert(AccCache.accLock==0); // Cannot evict so logic bomb
+  assert(AccCache.CpuPtr!=(uint64_t)NULL);
   if(AccCache.state==AccDirty) {
     Flush(AccCache);
   }
-  assert(AccCache.CpuPtr!=(uint64_t)NULL);
   if(AccCache.AccPtr) {
     AcceleratorFree((void *)AccCache.AccPtr,AccCache.bytes);
-    DeviceBytes   -=AccCache.bytes;
     LRUremove(AccCache);
+    AccCache.AccPtr=(uint64_t)NULL;
+    AccCache.state=CpuDirty; // CPU primary now
+    DeviceBytes   -=AccCache.bytes;
     dprintf("MemoryManager: Free(%lx) footprint now %ld \n",(uint64_t)AccCache.AccPtr,DeviceBytes);  
   }
-  uint64_t CpuPtr = AccCache.CpuPtr;
-  EntryErase(CpuPtr);
+  //  uint64_t CpuPtr = AccCache.CpuPtr;
+  DeviceEvictions++;
+  //  EntryErase(CpuPtr);
 }
 void MemoryManager::Flush(AcceleratorViewEntry &AccCache)
 {
@@ -221,13 +233,16 @@ void *MemoryManager::ViewOpen(void* _CpuPtr,size_t bytes,ViewMode mode,ViewAdvis
 }
 void  MemoryManager::EvictVictims(uint64_t bytes)
 {
+  assert(bytes<DeviceMaxBytes);
   while(bytes+DeviceLRUBytes > DeviceMaxBytes){
     if ( DeviceLRUBytes > 0){
       assert(LRU.size()>0);
-      uint64_t victim = LRU.back();
+      uint64_t victim = LRU.back(); // From the LRU
       auto AccCacheIterator = EntryLookup(victim);
       auto & AccCache = AccCacheIterator->second;
       Evict(AccCache);
+    } else {
+      return;
     }
   }
 }
@@ -322,7 +337,8 @@ uint64_t MemoryManager::AcceleratorViewOpen(uint64_t CpuPtr,size_t bytes,ViewMod
     assert(0);
   }
 
-  // If view is opened on device remove from LRU
+  assert(AccCache.accLock>0);
+  // If view is opened on device must remove from LRU
   if(AccCache.LRU_valid==1){
     // must possibly remove from LRU as now locked on GPU
     dprintf("AccCache entry removed from LRU \n");
@@ -389,7 +405,7 @@ uint64_t MemoryManager::CpuViewOpen(uint64_t CpuPtr,size_t bytes,ViewMode mode,V
   auto & AccCache = AccCacheIterator->second;
 
   if (!AccCache.AccPtr) {
-     EvictVictims(bytes);
+    EvictVictims(bytes);
   }
 
   assert((mode==CpuRead)||(mode==CpuWrite));
@@ -444,20 +460,28 @@ void  MemoryManager::NotifyDeletion(void *_ptr)
 void  MemoryManager::Print(void)
 {
   PrintBytes();
-  std::cout << GridLogDebug << "--------------------------------------------" << std::endl;
-  std::cout << GridLogDebug << "Memory Manager                             " << std::endl;
-  std::cout << GridLogDebug << "--------------------------------------------" << std::endl;
-  std::cout << GridLogDebug << DeviceBytes   << " bytes allocated on device " << std::endl;
-  std::cout << GridLogDebug << DeviceLRUBytes<< " bytes evictable on device " << std::endl;
-  std::cout << GridLogDebug << DeviceMaxBytes<< " bytes max on device       " << std::endl;
-  std::cout << GridLogDebug << HostToDeviceXfer << " transfers        to   device " << std::endl;
-  std::cout << GridLogDebug << DeviceToHostXfer << " transfers        from device " << std::endl;
-  std::cout << GridLogDebug << HostToDeviceBytes<< " bytes transfered to   device " << std::endl;
-  std::cout << GridLogDebug << DeviceToHostBytes<< " bytes transfered from device " << std::endl;
-  std::cout << GridLogDebug << AccViewTable.size()<< " vectors " << LRU.size()<<" evictable"<< std::endl;
-  std::cout << GridLogDebug << "--------------------------------------------" << std::endl;
-  std::cout << GridLogDebug << "CpuAddr\t\tAccAddr\t\tState\t\tcpuLock\taccLock\tLRU_valid "<<std::endl;
-  std::cout << GridLogDebug << "--------------------------------------------" << std::endl;
+  std::cout << GridLogMessage << "--------------------------------------------" << std::endl;
+  std::cout << GridLogMessage << "Memory Manager                             " << std::endl;
+  std::cout << GridLogMessage << "--------------------------------------------" << std::endl;
+  std::cout << GridLogMessage << DeviceBytes   << " bytes allocated on device " << std::endl;
+  std::cout << GridLogMessage << DeviceLRUBytes<< " bytes evictable on device " << std::endl;
+  std::cout << GridLogMessage << DeviceMaxBytes<< " bytes max on device       " << std::endl;
+  std::cout << GridLogMessage << HostToDeviceXfer << " transfers        to   device " << std::endl;
+  std::cout << GridLogMessage << DeviceToHostXfer << " transfers        from device " << std::endl;
+  std::cout << GridLogMessage << HostToDeviceBytes<< " bytes transfered to   device " << std::endl;
+  std::cout << GridLogMessage << DeviceToHostBytes<< " bytes transfered from device " << std::endl;
+  std::cout << GridLogMessage << DeviceEvictions  << " Evictions from device " << std::endl;
+  std::cout << GridLogMessage << DeviceDestroy    << " Destroyed vectors on device " << std::endl;
+  std::cout << GridLogMessage << AccViewTable.size()<< " vectors " << LRU.size()<<" evictable"<< std::endl;
+  std::cout << GridLogMessage << "--------------------------------------------" << std::endl;
+}
+void  MemoryManager::PrintAll(void)
+{
+  Print();
+  std::cout << GridLogMessage << std::endl;
+  std::cout << GridLogMessage << "--------------------------------------------" << std::endl;
+  std::cout << GridLogMessage << "CpuAddr\t\tAccAddr\t\tState\t\tcpuLock\taccLock\tLRU_valid "<<std::endl;
+  std::cout << GridLogMessage << "--------------------------------------------" << std::endl;
   for(auto it=AccViewTable.begin();it!=AccViewTable.end();it++){
     auto &AccCache = it->second;
     
@@ -467,13 +491,13 @@ void  MemoryManager::Print(void)
     if ( AccCache.state==AccDirty ) str = std::string("AccDirty");
     if ( AccCache.state==Consistent)str = std::string("Consistent");
 
-    std::cout << GridLogDebug << "0x"<<std::hex<<AccCache.CpuPtr<<std::dec
+    std::cout << GridLogMessage << "0x"<<std::hex<<AccCache.CpuPtr<<std::dec
 	      << "\t0x"<<std::hex<<AccCache.AccPtr<<std::dec<<"\t" <<str
 	      << "\t" << AccCache.cpuLock
 	      << "\t" << AccCache.accLock
 	      << "\t" << AccCache.LRU_valid<<std::endl;
   }
-  std::cout << GridLogDebug << "--------------------------------------------" << std::endl;
+  std::cout << GridLogMessage << "--------------------------------------------" << std::endl;
 
 };
 int   MemoryManager::isOpen   (void* _CpuPtr) 
@@ -489,6 +513,24 @@ int   MemoryManager::isOpen   (void* _CpuPtr)
 }
 void MemoryManager::Audit(std::string s)
 {
+  uint64_t CpuBytes=0;
+  uint64_t AccBytes=0;
+  uint64_t LruBytes1=0;
+  uint64_t LruBytes2=0;
+  uint64_t LruCnt=0;
+  uint64_t LockedBytes=0;
+  
+  std::cout << " Memory Manager::Audit() from "<<s<<std::endl;
+  for(auto it=LRU.begin();it!=LRU.end();it++){
+    uint64_t cpuPtr = *it;
+    assert(EntryPresent(cpuPtr));
+    auto AccCacheIterator = EntryLookup(cpuPtr);
+    auto & AccCache = AccCacheIterator->second;
+    LruBytes2+=AccCache.bytes;
+    assert(AccCache.LRU_valid==1);
+    assert(AccCache.LRU_entry==it);
+  }
+  std::cout << " Memory Manager::Audit() LRU queue matches table entries "<<std::endl;
   for(auto it=AccViewTable.begin();it!=AccViewTable.end();it++){
     auto &AccCache = it->second;
     
@@ -498,7 +540,13 @@ void MemoryManager::Audit(std::string s)
     if ( AccCache.state==AccDirty ) str = std::string("AccDirty");
     if ( AccCache.state==Consistent)str = std::string("Consistent");
 
-    if ( AccCache.cpuLock || AccCache.accLock ) { 
+    CpuBytes+=AccCache.bytes;
+    if( AccCache.AccPtr )    AccBytes+=AccCache.bytes;
+    if( AccCache.LRU_valid ) LruBytes1+=AccCache.bytes;
+    if( AccCache.LRU_valid ) LruCnt++;
+    
+    if ( AccCache.cpuLock || AccCache.accLock ) {
+      assert(AccCache.LRU_valid==0);
       std::cout << GridLogError << s<< "\n\t 0x"<<std::hex<<AccCache.CpuPtr<<std::dec
 		<< "\t0x"<<std::hex<<AccCache.AccPtr<<std::dec<<"\t" <<str
 		<< "\t cpuLock  " << AccCache.cpuLock
@@ -509,6 +557,14 @@ void MemoryManager::Audit(std::string s)
     assert( AccCache.cpuLock== 0 ) ;
     assert( AccCache.accLock== 0 ) ;
   }
+  std::cout << " Memory Manager::Audit() no locked table entries "<<std::endl;
+  assert(LruBytes1==LruBytes2);
+  assert(LruBytes1==DeviceLRUBytes);
+  std::cout << " Memory Manager::Audit() evictable bytes matches sum over table "<<std::endl;
+  assert(AccBytes==DeviceBytes);
+  std::cout << " Memory Manager::Audit() device bytes matches sum over table "<<std::endl;
+  assert(LruCnt == LRU.size());
+  std::cout << " Memory Manager::Audit() LRU entry count matches "<<std::endl;
 }
 
 void MemoryManager::PrintState(void* _CpuPtr)
@@ -526,8 +582,8 @@ void MemoryManager::PrintState(void* _CpuPtr)
     if ( AccCache.state==EvictNext) str = std::string("EvictNext");
 
     std::cout << GridLogMessage << "CpuAddr\t\tAccAddr\t\tState\t\tcpuLock\taccLock\tLRU_valid "<<std::endl;
-    std::cout << GridLogMessage << "0x"<<std::hex<<AccCache.CpuPtr<<std::dec
-    << "\t0x"<<std::hex<<AccCache.AccPtr<<std::dec<<"\t" <<str
+    std::cout << GridLogMessage << "\tx"<<std::hex<<AccCache.CpuPtr<<std::dec
+    << "\tx"<<std::hex<<AccCache.AccPtr<<std::dec<<"\t" <<str
     << "\t" << AccCache.cpuLock
     << "\t" << AccCache.accLock
     << "\t" << AccCache.LRU_valid<<std::endl;

From ede02b688392c2716c121e7b18f6b7d67954db25 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Thu, 1 Dec 2022 00:25:04 -0500
Subject: [PATCH 118/146] Memory manager debug Felix case

---
 Grid/perfmon/PerfCount.cc          |   5 +-
 Grid/perfmon/Timer.h               |  22 +++---
 Grid/perfmon/Tracing.h             |  70 ++++++++++++++++++
 systems/mac-arm/config-command-mpi |   2 +-
 tests/core/Test_memory_manager.cc  | 110 +++++++++++++++++++++++++++++
 5 files changed, 196 insertions(+), 13 deletions(-)
 create mode 100644 Grid/perfmon/Tracing.h
 create mode 100644 tests/core/Test_memory_manager.cc

diff --git a/Grid/perfmon/PerfCount.cc b/Grid/perfmon/PerfCount.cc
index 2062bb59..114c36a0 100644
--- a/Grid/perfmon/PerfCount.cc
+++ b/Grid/perfmon/PerfCount.cc
@@ -27,10 +27,13 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 /*  END LEGAL */
 
 #include <Grid/GridCore.h>
-#include <Grid/perfmon/PerfCount.h>
 
+#include <Grid/perfmon/Timer.h>
+#include <Grid/perfmon/PerfCount.h>
 NAMESPACE_BEGIN(Grid);
 
+GridTimePoint theProgramStart = GridClock::now();
+
 #define CacheControl(L,O,R) ((PERF_COUNT_HW_CACHE_##L)|(PERF_COUNT_HW_CACHE_OP_##O<<8)| (PERF_COUNT_HW_CACHE_RESULT_##R<<16))
 #define RawConfig(A,B) (A<<8|B)
 const PerformanceCounter::PerformanceCounterConfig PerformanceCounter::PerformanceCounterConfigs [] = {
diff --git a/Grid/perfmon/Timer.h b/Grid/perfmon/Timer.h
index 2a44faee..ba5df85a 100644
--- a/Grid/perfmon/Timer.h
+++ b/Grid/perfmon/Timer.h
@@ -35,17 +35,8 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 NAMESPACE_BEGIN(Grid)
 
-// Dress the output; use std::chrono
-// C++11 time facilities better?
-inline double usecond(void) {
-  struct timeval tv;
-#ifdef TIMERS_ON
-  gettimeofday(&tv,NULL);
-#endif
-  return 1.0*tv.tv_usec + 1.0e6*tv.tv_sec;
-}
-
-typedef  std::chrono::system_clock          GridClock;
+//typedef  std::chrono::system_clock          GridClock;
+typedef  std::chrono::high_resolution_clock   GridClock;
 typedef  std::chrono::time_point<GridClock> GridTimePoint;
 
 typedef  std::chrono::seconds               GridSecs;
@@ -53,6 +44,15 @@ typedef  std::chrono::milliseconds          GridMillisecs;
 typedef  std::chrono::microseconds          GridUsecs;
 typedef  std::chrono::microseconds          GridTime;
 
+extern GridTimePoint theProgramStart;
+// Dress the output; use std::chrono
+// C++11 time facilities better?
+inline double usecond(void) {
+  auto usecs = std::chrono::duration_cast<GridUsecs>(GridClock::now()-theProgramStart); 
+  return 1.0*usecs.count();
+}
+
+
 inline std::ostream& operator<< (std::ostream & stream, const GridSecs & time)
 {
   stream << time.count()<<" s";
diff --git a/Grid/perfmon/Tracing.h b/Grid/perfmon/Tracing.h
new file mode 100644
index 00000000..5000cef4
--- /dev/null
+++ b/Grid/perfmon/Tracing.h
@@ -0,0 +1,70 @@
+#pragma once
+
+NAMESPACE_BEGIN(Grid);
+
+#ifdef GRID_TRACING_NVTX
+#include <nvToolsExt.h>
+class GridTracer {
+public:
+  GridTracer(const char* name) {
+    nvtxRangePushA(name);
+  }
+  ~GridTracer() {
+    nvtxRangePop();
+  }
+};
+inline void tracePush(const char *name) { nvtxRangePushA(name); }
+inline void tracePop(const char *name) { nvtxRangePop(); }
+inline int  traceStart(const char *name) {  }
+inline void traceStop(int ID) {  }
+#endif
+
+#ifdef GRID_TRACING_ROCTX
+#include <roctracer/roctx.h>
+class GridTracer {
+ public:
+  GridTracer(const char* name) {
+    roctxRangePushA(name);
+    std::cout << "roctxRangePush "<<name<<std::endl;
+  }
+  ~GridTracer() {
+    roctxRangePop();
+    std::cout << "roctxRangePop "<<std::endl;
+  }
+};
+inline void tracePush(const char *name) { roctxRangePushA(name); }
+inline void tracePop(const char *name) { roctxRangePop(); }
+inline int  traceStart(const char *name) { roctxRangeStart(name); }
+inline void traceStop(int ID) { roctxRangeStop(ID); }
+#endif
+
+#ifdef GRID_TRACING_TIMER
+class GridTracer {
+ public:
+  const char *name;
+  double elapsed;
+  GridTracer(const char* _name) {
+    name = _name;
+    elapsed=-usecond();
+  }
+  ~GridTracer() {
+    elapsed+=usecond();
+    std::cout << GridLogTracing << name << " took " <<elapsed<< " us" <<std::endl;
+  }
+};
+inline void tracePush(const char *name) {  }
+inline void tracePop(const char *name) {  }
+inline int  traceStart(const char *name) { return 0; }
+inline void traceStop(int ID) {  }
+#endif
+
+#ifdef GRID_TRACING_NONE
+#define GRID_TRACE(name) 
+inline void tracePush(const char *name) {  }
+inline void tracePop(const char *name) {  }
+inline int  traceStart(const char *name) { return 0;  }
+inline void traceStop(int ID) {  }
+#else
+#define GRID_TRACE(name) GridTracer uniq_name_using_macros##__COUNTER__(name);
+#endif
+NAMESPACE_END(Grid);
diff --git a/systems/mac-arm/config-command-mpi b/systems/mac-arm/config-command-mpi
index d1e75c39..a900fc2d 100644
--- a/systems/mac-arm/config-command-mpi
+++ b/systems/mac-arm/config-command-mpi
@@ -1 +1 @@
-CXX=mpicxx-openmpi-mp CXXFLAGS=-I/opt/local/include/ LDFLAGS=-L/opt/local/lib/ ../../configure --enable-simd=GEN --enable-debug --enable-comms=mpi
+CXX=mpicxx-openmpi-mp CXXFLAGS=-I/opt/local/include/ LDFLAGS=-L/opt/local/lib/ ../../configure --enable-simd=GEN --enable-debug --enable-comms=mpi --enable-unified=no
diff --git a/tests/core/Test_memory_manager.cc b/tests/core/Test_memory_manager.cc
new file mode 100644
index 00000000..fb01dfe6
--- /dev/null
+++ b/tests/core/Test_memory_manager.cc
@@ -0,0 +1,110 @@
+    /*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./tests/Test_memory_manager.cc
+
+    Copyright (C) 2022
+
+Author: Peter Boyle <pboyle@bnl.gov>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+    *************************************************************************************/
+    /*  END LEGAL */
+#include <Grid/Grid.h>
+
+using namespace std;
+using namespace Grid;
+
+void  MemoryTest(GridCartesian         * FGrid,int N);
+
+int main (int argc, char ** argv)
+{
+  Grid_init(&argc,&argv);
+
+  GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplex::Nsimd()),GridDefaultMpi());
+
+  int N=100;
+  for(int i=0;i<N;i++){
+    std::cout << "============================"<<std::endl;
+    std::cout << "Epoch "<<i<<"/"<<N<<std::endl;
+    std::cout << "============================"<<std::endl;
+    MemoryTest(UGrid,256);
+    MemoryManager::Print();
+    AUDIT();
+  }
+  Grid_finalize();
+}
+
+void  MemoryTest(GridCartesian         * FGrid, int N)
+{
+  LatticeComplexD zero(FGrid); zero=Zero();
+  std::vector<LatticeComplexD> A(N,zero);//FGrid);
+
+  std::vector<ComplexD> B(N,ComplexD(0.0)); // Update sequentially on host
+
+  for(int v=0;v<N;v++) A[v] = Zero();
+
+  uint64_t counter = 0;
+  for(int epoch = 0;epoch<10000;epoch++){
+
+    int v  = random() %N; // Which vec
+    int w  = random() %2; // Write or read
+    int e  = random() %3; // expression or for loop
+    int dev= random() %2; // On device?
+    //    int e=1;
+    ComplexD zc = counter++;
+    
+    if ( w ) {
+      B[v] = B[v] + zc;
+      if ( e == 0 ) {
+	A[v] = A[v] + zc - A[v] + A[v];
+      } else {
+	if ( dev ) { 
+	  autoView(A_v,A[v],AcceleratorWrite);
+	  accelerator_for(ss,FGrid->oSites(),1,{
+	    A_v[ss] = A_v[ss] + zc;
+	    });
+	} else {
+	  autoView(A_v,A[v],CpuWrite);
+	  thread_for(ss,FGrid->oSites(),{
+	      A_v[ss] = A_v[ss] + zc;
+	    });
+	}
+      }
+    } else {
+      if ( e == 0 ) {
+	A[v] = A[v] + A[v] - A[v];
+      } else { 
+	if ( dev ) { 
+	  autoView(A_v,A[v],AcceleratorRead);
+	  accelerator_for(ss,FGrid->oSites(),1,{
+	      assert(B[v]==A_v[ss]()()().getlane(0));
+	    });
+	  //	std::cout << "["<<v<<"] checked on GPU"<<B[v]<<std::endl;
+	} else {
+	  autoView(A_v,A[v],CpuRead);
+	  thread_for(ss,FGrid->oSites(),{
+	      assert(B[v]==A_v[ss]()()().getlane(0));
+	    });
+	  //	std::cout << "["<<v<<"] checked on CPU"<<B[v]<<std::endl;
+	}
+      }    
+    }
+  }
+
+}

From 2ff868f7a5ce038e2c9614870edc48babe05793c Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Thu, 1 Dec 2022 00:35:05 -0500
Subject: [PATCH 119/146] CPU open doesn't need to free space

---
 Grid/allocator/MemoryManagerCache.cc | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/Grid/allocator/MemoryManagerCache.cc b/Grid/allocator/MemoryManagerCache.cc
index 3420c9cc..bae184ec 100644
--- a/Grid/allocator/MemoryManagerCache.cc
+++ b/Grid/allocator/MemoryManagerCache.cc
@@ -404,9 +404,10 @@ uint64_t MemoryManager::CpuViewOpen(uint64_t CpuPtr,size_t bytes,ViewMode mode,V
   auto AccCacheIterator = EntryLookup(CpuPtr);
   auto & AccCache = AccCacheIterator->second;
 
-  if (!AccCache.AccPtr) {
-    EvictVictims(bytes);
-  }
+  // CPU doesn't need to free space
+  //  if (!AccCache.AccPtr) {
+  //    EvictVictims(bytes);
+  //  }
 
   assert((mode==CpuRead)||(mode==CpuWrite));
   assert(AccCache.accLock==0);  // Programming error

From 4ca1bf7ccaaa471a51994564adb5e445bc62afd5 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Wed, 21 Dec 2022 07:23:16 -0500
Subject: [PATCH 120/146] Added gauge invariance test

---
 tests/core/Test_fft_matt.cc | 59 +++++++++++++++++++++++++++++++++++++
 1 file changed, 59 insertions(+)

diff --git a/tests/core/Test_fft_matt.cc b/tests/core/Test_fft_matt.cc
index d92fa94e..55234601 100644
--- a/tests/core/Test_fft_matt.cc
+++ b/tests/core/Test_fft_matt.cc
@@ -206,6 +206,65 @@ int main (int argc, char ** argv)
     DumpSliceNorm("Slice Norm Solution ",result,Nd-1);
   }
 
+  ////////////////////////////////////////////////////
+  //Gauge invariance test
+  ////////////////////////////////////////////////////
+  {
+    std::cout<<"****************************************"<<std::endl;
+    std::cout << "Gauge invariance test \n";
+    std::cout<<"****************************************"<<std::endl;
+    LatticeGaugeField     U_GT(&GRID); // Gauge transformed field
+    LatticeColourMatrix   g(&GRID);    // local Gauge xform matrix
+    U_GT = Umu;
+    // Make a random xform to teh gauge field
+    SU<Nc>::RandomGaugeTransform(pRNG,U_GT,g); // Unit gauge
+
+    LatticeFermionD    src(&GRID);
+    LatticeFermionD    tmp(&GRID);
+    LatticeFermionD    ref(&GRID);
+    LatticeFermionD    diff(&GRID);
+
+    // could loop over colors
+    src=Zero();
+    Coordinate point(4,0); // 0,0,0,0
+    SpinColourVectorD ferm;
+    ferm=Zero();
+    ferm()(0)(0) = ComplexD(1.0);
+    pokeSite(ferm,src,point);
+
+    RealD mass=0.1;
+    WilsonFermionD Dw(U_GT,GRID,RBGRID,mass);
+
+    // Momentum space prop
+    std::cout << " Solving by FFT and Feynman rules" <<std::endl;
+    Dw.FreePropagator(src,ref,mass) ;
+
+    Gamma G5(Gamma::Algebra::Gamma5);
+
+    LatticeFermionD    result(&GRID); 
+    const int sdir=0;
+    
+    ////////////////////////////////////////////////////////////////////////
+    // Conjugate gradient on normal equations system
+    ////////////////////////////////////////////////////////////////////////
+    std::cout << " Solving by Conjugate Gradient (CGNE)" <<std::endl;
+    Dw.Mdag(src,tmp);
+    src=tmp;
+    MdagMLinearOperator<WilsonFermionD,LatticeFermionD> HermOp(Dw);
+    ConjugateGradient<LatticeFermionD> CG(1.0e-10,10000);
+    CG(HermOp,src,result);
+    
+    ////////////////////////////////////////////////////////////////////////
+    std::cout << " Taking difference" <<std::endl;
+    std::cout << "Dw result "<<norm2(result)<<std::endl;
+    std::cout << "Dw ref     "<<norm2(ref)<<std::endl;
+    
+    diff = ref - result;
+    std::cout << "result - ref     "<<norm2(diff)<<std::endl;
+
+    DumpSliceNorm("Slice Norm Solution ",result,Nd-1);
+  }
+  
   
   Grid_finalize();
 }

From ad0270ac8c6f55f756ad3d3e70047178f6f1ec55 Mon Sep 17 00:00:00 2001
From: Fabian Joswig <fabian.joswig@ed.ac.uk>
Date: Thu, 12 Jan 2023 12:36:30 +0000
Subject: [PATCH 121/146] fix: diagnostic pragma warnings fixed for CUDA 12+

---
 Grid/DisableWarnings.h  | 2 +-
 Grid/Grid_Eigen_Dense.h | 2 +-
 Grid/pugixml/pugixml.cc | 2 +-
 3 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/Grid/DisableWarnings.h b/Grid/DisableWarnings.h
index 64e4faf4..015e19d1 100644
--- a/Grid/DisableWarnings.h
+++ b/Grid/DisableWarnings.h
@@ -45,7 +45,7 @@ directory
  //disables nvcc specific warning in json.hpp
 #pragma clang diagnostic ignored "-Wdeprecated-register"
 
-#if (__CUDACC_VER_MAJOR__ >= 11) && (__CUDACC_VER_MINOR__ >= 5)
+#ifdef __NVCC_DIAG_PRAGMA_SUPPORT__
  //disables nvcc specific warning in json.hpp
 #pragma nv_diag_suppress unsigned_compare_with_zero
 #pragma nv_diag_suppress cast_to_qualified_type
diff --git a/Grid/Grid_Eigen_Dense.h b/Grid/Grid_Eigen_Dense.h
index 5aee81de..bdd39a65 100644
--- a/Grid/Grid_Eigen_Dense.h
+++ b/Grid/Grid_Eigen_Dense.h
@@ -14,7 +14,7 @@
 /* NVCC save and restore compile environment*/
 #ifdef __NVCC__
 #pragma push
-#if (__CUDACC_VER_MAJOR__ >= 11) && (__CUDACC_VER_MINOR__ >= 5)
+#ifdef __NVCC_DIAG_PRAGMA_SUPPORT__
 #pragma nv_diag_suppress code_is_unreachable
 #else
 #pragma diag_suppress code_is_unreachable
diff --git a/Grid/pugixml/pugixml.cc b/Grid/pugixml/pugixml.cc
index 81973964..b2ba698b 100644
--- a/Grid/pugixml/pugixml.cc
+++ b/Grid/pugixml/pugixml.cc
@@ -16,7 +16,7 @@
 
 #ifdef __NVCC__
 #pragma push
-#if (__CUDACC_VER_MAJOR__ >= 11) && (__CUDACC_VER_MINOR__ >= 5)
+#ifdef __NVCC_DIAG_PRAGMA_SUPPORT__
 #pragma nv_diag_suppress declared_but_not_referenced // suppress "function was declared but never referenced warning"
 #else
 #pragma diag_suppress declared_but_not_referenced // suppress "function was declared but never referenced warning"

From 1db58a8acce1cddb20be116b5916e4b359ae259f Mon Sep 17 00:00:00 2001
From: Christopher Kelly <ckelly@bnl.gov>
Date: Tue, 21 Feb 2023 10:52:42 -0500
Subject: [PATCH 122/146] Precision change improvements

Added a new, much faster implementation of precision change that uses (optionally) a precomputed workspace containing pointer offsets that is device resident, such that all lattice copying occurs only on the device and no host<->device transfer is required, other than the pointer table. It also avoids the need to unpack and repack the fields using explicit lane copying. When this new precisionChange is called without a workspace, one will be computed on-the-fly; however it is still considerably faster than the original implementation.

In the special case of using double2 and when the Grids are the same, calls to the new precisionChange will automatically use precisionChangeFast, such that there is a single API call for all precision changes.

Reliable update and mixed-prec multishift have been modified to precompute precision change workspaces

Renamed the original precisionChange as precisionChangeOrig

Fixed incorrect pointer offset bug in copyLane

Added a test and a benchmark for precisionChange

Added a test for reliable update CG
---
 .../ConjugateGradientMultiShiftMixedPrec.h    |  11 +-
 .../ConjugateGradientReliableUpdate.h         |  32 ++-
 Grid/lattice/Lattice_transfer.h               | 127 +++++++++++-
 Grid/tensors/Tensor_extract_merge.h           |   6 +-
 benchmarks/Benchmark_prec_change.cc           | 189 ++++++++++++++++++
 tests/core/Test_prec_change.cc                | 124 ++++++++++++
 tests/solver/Test_dwf_relupcg_prec.cc         | 143 +++++++++++++
 7 files changed, 616 insertions(+), 16 deletions(-)
 create mode 100644 benchmarks/Benchmark_prec_change.cc
 create mode 100644 tests/core/Test_prec_change.cc
 create mode 100644 tests/solver/Test_dwf_relupcg_prec.cc

diff --git a/Grid/algorithms/iterative/ConjugateGradientMultiShiftMixedPrec.h b/Grid/algorithms/iterative/ConjugateGradientMultiShiftMixedPrec.h
index de1cfe01..a89a1e4a 100644
--- a/Grid/algorithms/iterative/ConjugateGradientMultiShiftMixedPrec.h
+++ b/Grid/algorithms/iterative/ConjugateGradientMultiShiftMixedPrec.h
@@ -130,6 +130,9 @@ public:
     GRID_TRACE("ConjugateGradientMultiShiftMixedPrec");
     GridBase *DoublePrecGrid = src_d.Grid();
 
+    precisionChangeWorkspace pc_wk_s_to_d(DoublePrecGrid,SinglePrecGrid);
+    precisionChangeWorkspace pc_wk_d_to_s(SinglePrecGrid,DoublePrecGrid);
+    
     ////////////////////////////////////////////////////////////////////////
     // Convenience references to the info stored in "MultiShiftFunction"
     ////////////////////////////////////////////////////////////////////////
@@ -200,10 +203,10 @@ public:
     r_d = p_d;
     
     //MdagM+m[0]
-    precisionChangeFast(p_f,p_d);
+    precisionChange(p_f, p_d, pc_wk_d_to_s);
 
     Linop_f.HermOpAndNorm(p_f,mmp_f,d,qq); // mmp = MdagM p        d=real(dot(p, mmp)),  qq=norm2(mmp)
-    precisionChangeFast(tmp_d,mmp_f);
+    precisionChange(tmp_d, mmp_f, pc_wk_s_to_d);
     Linop_d.HermOpAndNorm(p_d,mmp_d,d,qq); // mmp = MdagM p        d=real(dot(p, mmp)),  qq=norm2(mmp)
     tmp_d = tmp_d - mmp_d;
     std::cout << " Testing operators match "<<norm2(mmp_d)<<" f "<<norm2(mmp_f)<<" diff "<< norm2(tmp_d)<<std::endl;
@@ -263,7 +266,7 @@ public:
       AXPYTimer.Stop();
 
       PrecChangeTimer.Start();
-      precisionChangeFast(p_f, p_d); //get back single prec search direction for linop
+      precisionChange(p_f, p_d, pc_wk_d_to_s); //get back single prec search direction for linop
       PrecChangeTimer.Stop();
 
       cp=c;
@@ -272,7 +275,7 @@ public:
       MatrixTimer.Stop();  
 
       PrecChangeTimer.Start();
-      precisionChangeFast(mmp_d, mmp_f); // From Float to Double
+      precisionChange(mmp_d, mmp_f, pc_wk_s_to_d); // From Float to Double
       PrecChangeTimer.Stop();
 
       AXPYTimer.Start();
diff --git a/Grid/algorithms/iterative/ConjugateGradientReliableUpdate.h b/Grid/algorithms/iterative/ConjugateGradientReliableUpdate.h
index 094f847d..514800fc 100644
--- a/Grid/algorithms/iterative/ConjugateGradientReliableUpdate.h
+++ b/Grid/algorithms/iterative/ConjugateGradientReliableUpdate.h
@@ -48,7 +48,7 @@ public:
   LinearOperatorBase<FieldF> &Linop_f;
   LinearOperatorBase<FieldD> &Linop_d;
   GridBase* SinglePrecGrid;
-  RealD Delta; //reliable update parameter
+  RealD Delta; //reliable update parameter. A reliable update is performed when the residual drops by a factor of Delta relative to its value at the last update
 
   //Optional ability to switch to a different linear operator once the tolerance reaches a certain point. Useful for single/half -> single/single
   LinearOperatorBase<FieldF> *Linop_fallback;
@@ -65,7 +65,9 @@ public:
       ErrorOnNoConverge(err_on_no_conv),
       DoFinalCleanup(true),
       Linop_fallback(NULL)
-  {};
+  {
+    assert(Delta > 0. && Delta < 1. && "Expect  0 < Delta < 1");
+  };
 
   void setFallbackLinop(LinearOperatorBase<FieldF> &_Linop_fallback, const RealD _fallback_transition_tol){
     Linop_fallback = &_Linop_fallback;
@@ -116,9 +118,12 @@ public:
     }
 
     //Single prec initialization
+    precisionChangeWorkspace pc_wk_sp_to_dp(src.Grid(), SinglePrecGrid);
+    precisionChangeWorkspace pc_wk_dp_to_sp(SinglePrecGrid, src.Grid());
+    
     FieldF r_f(SinglePrecGrid);
     r_f.Checkerboard() = r.Checkerboard();
-    precisionChange(r_f, r);
+    precisionChange(r_f, r, pc_wk_dp_to_sp);
 
     FieldF psi_f(r_f);
     psi_f = Zero();
@@ -134,7 +139,8 @@ public:
     GridStopWatch LinalgTimer;
     GridStopWatch MatrixTimer;
     GridStopWatch SolverTimer;
-
+    GridStopWatch PrecChangeTimer;
+    
     SolverTimer.Start();
     int k = 0;
     int l = 0;
@@ -173,7 +179,9 @@ public:
       // Stopping condition
       if (cp <= rsq) {
 	//Although not written in the paper, I assume that I have to add on the final solution
-	precisionChange(mmp, psi_f);
+	PrecChangeTimer.Start();
+	precisionChange(mmp, psi_f, pc_wk_sp_to_dp);
+	PrecChangeTimer.Stop();
 	psi = psi + mmp;
 	
 	
@@ -194,7 +202,10 @@ public:
 	std::cout << GridLogMessage << "\tElapsed    " << SolverTimer.Elapsed() <<std::endl;
 	std::cout << GridLogMessage << "\tMatrix     " << MatrixTimer.Elapsed() <<std::endl;
 	std::cout << GridLogMessage << "\tLinalg     " << LinalgTimer.Elapsed() <<std::endl;
+	std::cout << GridLogMessage << "\tPrecChange " << PrecChangeTimer.Elapsed() <<std::endl;
+	std::cout << GridLogMessage << "\tPrecChange avg time " << PrecChangeTimer.Elapsed()/(2*l+1) <<std::endl;
 
+	
 	IterationsToComplete = k;	
 	ReliableUpdatesPerformed = l;
 	  
@@ -214,14 +225,21 @@ public:
       else if(cp < Delta * MaxResidSinceLastRelUp) { //reliable update
 	std::cout << GridLogMessage << "ConjugateGradientReliableUpdate "
 		  << cp << "(residual) < " << Delta << "(Delta) * " << MaxResidSinceLastRelUp << "(MaxResidSinceLastRelUp) on iteration " << k << " : performing reliable update\n";
-	precisionChange(mmp, psi_f);
+	PrecChangeTimer.Start();
+	precisionChange(mmp, psi_f, pc_wk_sp_to_dp);
+	PrecChangeTimer.Stop();
 	psi = psi + mmp;
 
+	MatrixTimer.Start();
 	Linop_d.HermOpAndNorm(psi, mmp, d, qq);
+	MatrixTimer.Stop();
+	
 	r = src - mmp;
 
 	psi_f = Zero();
-	precisionChange(r_f, r);
+	PrecChangeTimer.Start();
+	precisionChange(r_f, r, pc_wk_dp_to_sp);
+	PrecChangeTimer.Stop();
 	cp = norm2(r);
 	MaxResidSinceLastRelUp = cp;
 
diff --git a/Grid/lattice/Lattice_transfer.h b/Grid/lattice/Lattice_transfer.h
index 3cdb75d1..6f9fc480 100644
--- a/Grid/lattice/Lattice_transfer.h
+++ b/Grid/lattice/Lattice_transfer.h
@@ -1080,6 +1080,7 @@ vectorizeFromRevLexOrdArray( std::vector<sobj> &in, Lattice<vobj> &out)
   });
 }
 
+//Very fast precision change. Requires in/out objects to reside on same Grid (e.g. by using double2 for the double-precision field)
 template<class VobjOut, class VobjIn>
 void precisionChangeFast(Lattice<VobjOut> &out, const Lattice<VobjIn> &in)
 {
@@ -1097,9 +1098,9 @@ void precisionChangeFast(Lattice<VobjOut> &out, const Lattice<VobjIn> &in)
       precisionChange(vout,vin,N);
   });
 }
-//Convert a Lattice from one precision to another
+//Convert a Lattice from one precision to another (original, slow implementation)
 template<class VobjOut, class VobjIn>
-void precisionChange(Lattice<VobjOut> &out, const Lattice<VobjIn> &in)
+void precisionChangeOrig(Lattice<VobjOut> &out, const Lattice<VobjIn> &in)
 {
   assert(out.Grid()->Nd() == in.Grid()->Nd());
   for(int d=0;d<out.Grid()->Nd();d++){
@@ -1145,6 +1146,128 @@ void precisionChange(Lattice<VobjOut> &out, const Lattice<VobjIn> &in)
   });
 }
 
+//The workspace for a precision change operation allowing for the reuse of the mapping to save time on subsequent calls
+class precisionChangeWorkspace{
+  std::pair<Integer,Integer>* fmap_device; //device pointer
+  //maintain grids for checking
+  GridBase* _out_grid;
+  GridBase* _in_grid;
+public:
+  precisionChangeWorkspace(GridBase *out_grid, GridBase *in_grid): _out_grid(out_grid), _in_grid(in_grid){
+    //Build a map between the sites and lanes of the output field and the input field as we cannot use the Grids on the device
+    assert(out_grid->Nd() == in_grid->Nd());
+    for(int d=0;d<out_grid->Nd();d++){
+      assert(out_grid->FullDimensions()[d] == in_grid->FullDimensions()[d]);
+    }
+    int Nsimd_out = out_grid->Nsimd();
+
+    std::vector<Coordinate> out_icorrs(out_grid->Nsimd()); //reuse these
+    for(int lane=0; lane < out_grid->Nsimd(); lane++)
+      out_grid->iCoorFromIindex(out_icorrs[lane], lane);
+  
+    std::vector<std::pair<Integer,Integer> > fmap_host(out_grid->lSites()); //lsites = osites*Nsimd
+    thread_for(out_oidx,out_grid->oSites(),{
+	Coordinate out_ocorr; 
+	out_grid->oCoorFromOindex(out_ocorr, out_oidx);
+      
+	Coordinate lcorr; //the local coordinate (common to both in and out as full coordinate)
+	for(int out_lane=0; out_lane < Nsimd_out; out_lane++){
+	  out_grid->InOutCoorToLocalCoor(out_ocorr, out_icorrs[out_lane], lcorr);
+	
+	  //int in_oidx = in_grid->oIndex(lcorr), in_lane = in_grid->iIndex(lcorr);
+	  //Note oIndex and OcorrFromOindex (and same for iIndex) are not inverse for checkerboarded lattice, the former coordinates being defined on the full lattice and the latter on the reduced lattice
+	  //Until this is fixed we need to circumvent the problem locally. Here I will use the coordinates defined on the reduced lattice for simplicity
+	  int in_oidx = 0, in_lane = 0;
+	  for(int d=0;d<in_grid->_ndimension;d++){
+	    in_oidx += in_grid->_ostride[d] * ( lcorr[d] % in_grid->_rdimensions[d] );
+	    in_lane += in_grid->_istride[d] * ( lcorr[d] / in_grid->_rdimensions[d] );
+	  }
+	  fmap_host[out_lane + Nsimd_out*out_oidx] = std::pair<Integer,Integer>( in_oidx, in_lane );
+	}
+      });
+
+    //Copy the map to the device (if we had a way to tell if an accelerator is in use we could avoid this copy for CPU-only machines)
+    size_t fmap_bytes = out_grid->lSites() * sizeof(std::pair<Integer,Integer>);
+    fmap_device = (std::pair<Integer,Integer>*)acceleratorAllocDevice(fmap_bytes);
+    acceleratorCopyToDevice(fmap_host.data(), fmap_device, fmap_bytes); 
+  }
+
+  //Prevent moving or copying
+  precisionChangeWorkspace(const precisionChangeWorkspace &r) = delete;
+  precisionChangeWorkspace(precisionChangeWorkspace &&r) = delete;
+  precisionChangeWorkspace &operator=(const precisionChangeWorkspace &r) = delete;
+  precisionChangeWorkspace &operator=(precisionChangeWorkspace &&r) = delete;
+  
+  std::pair<Integer,Integer> const* getMap() const{ return fmap_device; }
+
+  void checkGrids(GridBase* out, GridBase* in) const{
+    conformable(out, _out_grid);
+    conformable(in, _in_grid);
+  }
+  
+  ~precisionChangeWorkspace(){
+    acceleratorFreeDevice(fmap_device);
+  }
+};
+
+
+//We would like to use precisionChangeFast when possible. However usage of this requires the Grids to be the same (runtime check)
+//*and* the precisionChange(VobjOut::vector_type, VobjIn, int) function to be defined for the types; this requires an extra compile-time check which we do using some SFINAE trickery
+template<class VobjOut, class VobjIn>
+auto _precisionChangeFastWrap(Lattice<VobjOut> &out, const Lattice<VobjIn> &in, int dummy)->decltype( precisionChange( ((typename VobjOut::vector_type*)0), ((typename VobjIn::vector_type*)0), 1), int()){
+  if(out.Grid() == in.Grid()){
+    precisionChangeFast(out,in);
+    return 1;
+  }else{
+    return 0;
+  }
+}
+template<class VobjOut, class VobjIn>
+int _precisionChangeFastWrap(Lattice<VobjOut> &out, const Lattice<VobjIn> &in, long dummy){ //note long here is intentional; it means the above is preferred if available
+  return 0;
+}
+
+
+//Convert a lattice of one precision to another. Much faster than original implementation but requires a pregenerated workspace
+//which contains the mapping data.
+template<class VobjOut, class VobjIn>
+void precisionChange(Lattice<VobjOut> &out, const Lattice<VobjIn> &in, const precisionChangeWorkspace &workspace){
+  if(_precisionChangeFastWrap(out,in,0)) return;
+  
+  static_assert( std::is_same<typename VobjOut::scalar_typeD, typename VobjIn::scalar_typeD>::value == 1, "precisionChange: tensor types must be the same" ); //if tensor types are same the DoublePrecision type must be the same
+
+  out.Checkerboard() = in.Checkerboard();
+  constexpr int Nsimd_out = VobjOut::Nsimd();
+
+  workspace.checkGrids(out.Grid(),in.Grid());
+  std::pair<Integer,Integer> const* fmap_device = workspace.getMap();
+
+  //Do the copy/precision change
+  autoView( out_v , out, AcceleratorWrite);
+  autoView( in_v , in, AcceleratorRead);
+
+  accelerator_for(out_oidx, out.Grid()->oSites(), 1,{
+      std::pair<Integer,Integer> const* fmap_osite = fmap_device + out_oidx*Nsimd_out;
+      for(int out_lane=0; out_lane < Nsimd_out; out_lane++){      
+	int in_oidx = fmap_osite[out_lane].first;
+	int in_lane = fmap_osite[out_lane].second;
+	copyLane(out_v[out_oidx], out_lane, in_v[in_oidx], in_lane);
+      }
+    });
+}
+
+//Convert a Lattice from one precision to another. Much faster than original implementation but slower than precisionChangeFast
+//or precisionChange called with pregenerated workspace, as it needs to internally generate the workspace on the host and copy to device
+template<class VobjOut, class VobjIn>
+void precisionChange(Lattice<VobjOut> &out, const Lattice<VobjIn> &in){
+  if(_precisionChangeFastWrap(out,in,0)) return;   
+  precisionChangeWorkspace workspace(out.Grid(), in.Grid());
+  precisionChange(out, in, workspace);
+}
+
+
+
+
 ////////////////////////////////////////////////////////////////////////////////
 // Communicate between grids
 ////////////////////////////////////////////////////////////////////////////////
diff --git a/Grid/tensors/Tensor_extract_merge.h b/Grid/tensors/Tensor_extract_merge.h
index 87572faf..f1407d1f 100644
--- a/Grid/tensors/Tensor_extract_merge.h
+++ b/Grid/tensors/Tensor_extract_merge.h
@@ -226,7 +226,7 @@ template<class vobjOut, class vobjIn>
 accelerator_inline 
 void copyLane(vobjOut & __restrict__ vecOut, int lane_out, const vobjIn & __restrict__ vecIn, int lane_in)
 {
-  static_assert( std::is_same<typename vobjOut::DoublePrecision, typename vobjIn::DoublePrecision>::value == 1, "copyLane: tensor types must be the same" ); //if tensor types are same the DoublePrecision type must be the same
+  static_assert( std::is_same<typename vobjOut::scalar_typeD, typename vobjIn::scalar_typeD>::value == 1, "copyLane: tensor types must be the same" ); //if tensor types are same the DoublePrecision type must be the same
 
   typedef typename vobjOut::vector_type ovector_type;  
   typedef typename vobjIn::vector_type ivector_type;  
@@ -251,9 +251,9 @@ void copyLane(vobjOut & __restrict__ vecOut, int lane_out, const vobjIn & __rest
   ovector_type * __restrict__ op = (ovector_type *)&vecOut;
   ivector_type * __restrict__ ip = (ivector_type *)&vecIn;
   for(int w=0;w<owords;w++){
-    itmp = ip[iNsimd*w].getlane(lane_in);
+    itmp = ip[w].getlane(lane_in);
     otmp = itmp; //potential precision change
-    op[oNsimd*w].putlane(otmp,lane_out);
+    op[w].putlane(otmp,lane_out);
   }
 }
 
diff --git a/benchmarks/Benchmark_prec_change.cc b/benchmarks/Benchmark_prec_change.cc
new file mode 100644
index 00000000..ba34f87e
--- /dev/null
+++ b/benchmarks/Benchmark_prec_change.cc
@@ -0,0 +1,189 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./benchmarks/Benchmark_prec_change.cc
+
+    Copyright (C) 2015
+
+Author: Christopher Kelly <ckelly@bnl.gov>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+    *************************************************************************************/
+    /*  END LEGAL */
+#include <Grid/Grid.h>
+
+using namespace std;
+using namespace Grid;
+
+int main (int argc, char ** argv)
+{
+  Grid_init(&argc,&argv);
+
+  int Ls = 12;
+  Coordinate latt4 = GridDefaultLatt();
+
+  GridCartesian         * UGridD   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplexD::Nsimd()),GridDefaultMpi());
+  GridRedBlackCartesian * UrbGridD = SpaceTimeGrid::makeFourDimRedBlackGrid(UGridD);
+  GridCartesian         * FGridD   = SpaceTimeGrid::makeFiveDimGrid(Ls,UGridD);
+  GridRedBlackCartesian * FrbGridD = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGridD);
+
+  GridCartesian         * UGridF   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplexF::Nsimd()),GridDefaultMpi());
+  GridRedBlackCartesian * UrbGridF = SpaceTimeGrid::makeFourDimRedBlackGrid(UGridF);
+  GridCartesian         * FGridF   = SpaceTimeGrid::makeFiveDimGrid(Ls,UGridF);
+  GridRedBlackCartesian * FrbGridF = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGridF);
+
+  
+  std::vector<int> seeds4({1,2,3,4});
+  std::vector<int> seeds5({5,6,7,8});
+  
+  std::cout << GridLogMessage << "Initialising 4d RNG" << std::endl;
+  GridParallelRNG          RNG4(UGridD);  RNG4.SeedFixedIntegers(seeds4);
+  std::cout << GridLogMessage << "Initialising 5d RNG" << std::endl;
+  GridParallelRNG          RNG5(FGridD);  RNG5.SeedFixedIntegers(seeds5);
+  std::cout << GridLogMessage << "Initialised RNGs" << std::endl;
+
+  LatticeFermionD field_d(FGridD), tmp_d(FGridD);
+  random(RNG5,field_d); tmp_d = field_d;
+
+  LatticeFermionD2 field_d2(FGridF), tmp_d2(FGridF);
+  precisionChange(field_d2, field_d); tmp_d2 = field_d2;
+
+  LatticeFermionF field_f(FGridF), tmp_f(FGridF);
+  precisionChange(field_f, field_d); tmp_f = field_f;
+
+  int N = 500;
+
+  double time_ds = 0, time_sd = 0;
+
+  std::cout<<GridLogMessage << "Benchmarking single<->double original implementation (fields initially device-resident)" << std::endl;
+  for(int i=0;i<N;i++){
+    //We want to benchmark the typical scenario of both fields being device resident
+    //To do this, invoke an operation that will open a device view and touch all sites
+    //with a write operation that invalidates the CPU copy
+    field_d = tmp_d;
+    field_f = tmp_f;
+
+    double start=usecond();
+    precisionChangeOrig(field_d,field_f);
+    double stop=usecond();
+    time_sd += stop - start;
+
+    field_d = tmp_d;
+    field_f = tmp_f;
+
+    start=usecond();
+    precisionChangeOrig(field_f,field_d);
+    stop=usecond();
+    time_ds += stop - start;   
+  }
+  std::cout << "d->s " << time_ds/N << "us" << " s->d " << time_sd/N << "us" << std::endl;
+
+
+  precisionChangeWorkspace wk_sp_to_dp(field_d.Grid(),field_f.Grid());
+  precisionChangeWorkspace wk_dp_to_sp(field_f.Grid(),field_d.Grid());
+  
+  std::cout<<GridLogMessage << "Benchmarking single<->double with pregenerated workspace(fields initially device-resident)" << std::endl;
+  time_sd = time_ds = 0;
+  for(int i=0;i<N;i++){
+    field_d = tmp_d;
+    field_f = tmp_f;
+
+    double start=usecond();
+    precisionChange(field_d,field_f, wk_sp_to_dp);
+    double stop=usecond();
+    time_sd += stop - start;
+
+    field_d = tmp_d;
+    field_f = tmp_f;
+
+    start=usecond();
+    precisionChange(field_f,field_d, wk_dp_to_sp);
+    stop=usecond();
+    time_ds += stop - start;   
+  }
+  std::cout << "d->s " << time_ds/N << "us" << " s->d " << time_sd/N << "us" << std::endl;
+  
+  std::cout<<GridLogMessage << "Benchmarking single<->double with workspace generated on-the-fly (fields initially device-resident)" << std::endl;
+  time_sd = time_ds = 0;
+  for(int i=0;i<N;i++){
+    field_d = tmp_d;
+    field_f = tmp_f;
+
+    double start=usecond();
+    precisionChange(field_d,field_f);
+    double stop=usecond();
+    time_sd += stop - start;
+
+    field_d = tmp_d;
+    field_f = tmp_f;
+
+    start=usecond();
+    precisionChange(field_f,field_d);
+    stop=usecond();
+    time_ds += stop - start;
+
+  }
+  std::cout << "d->s " << time_ds/N << "us" << " s->d " << time_sd/N << "us" << std::endl;
+
+
+  std::cout<<GridLogMessage << "Benchmarking single<->double2 (fields initially device-resident)" << std::endl;
+  time_sd = time_ds = 0;
+  for(int i=0;i<N;i++){
+    field_d2 = tmp_d2;
+    field_f = tmp_f;
+
+    double start=usecond();
+    precisionChangeFast(field_d2,field_f);
+    double stop=usecond();
+    time_sd += stop - start;
+
+    field_d2 = tmp_d2;
+    field_f = tmp_f;
+
+    start=usecond();
+    precisionChangeFast(field_f,field_d2);
+    stop=usecond();
+    time_ds += stop - start;
+  }
+  std::cout << "d->s " << time_ds/N << "us" << " s->d " << time_sd/N << "us" << std::endl;
+
+
+  std::cout<<GridLogMessage << "Benchmarking single<->double2 through standard precisionChange call(fields initially device-resident) [NB: perf should be the same as the previous test!]" << std::endl;
+  time_sd = time_ds = 0;
+  for(int i=0;i<N;i++){
+    field_d2 = tmp_d2;
+    field_f = tmp_f;
+
+    double start=usecond();
+    precisionChange(field_d2,field_f);
+    double stop=usecond();
+    time_sd += stop - start;
+
+    field_d2 = tmp_d2;
+    field_f = tmp_f;
+
+    start=usecond();
+    precisionChange(field_f,field_d2);
+    stop=usecond();
+    time_ds += stop - start;
+  }
+  std::cout << "d->s " << time_ds/N << "us" << " s->d " << time_sd/N << "us" << std::endl;
+
+  Grid_finalize();
+}
diff --git a/tests/core/Test_prec_change.cc b/tests/core/Test_prec_change.cc
new file mode 100644
index 00000000..06b9ae5c
--- /dev/null
+++ b/tests/core/Test_prec_change.cc
@@ -0,0 +1,124 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./tests/core/Test_prec_change.cc
+
+    Copyright (C) 2015
+
+Author: Christopher Kelly <ckelly@bnl.gov>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+    *************************************************************************************/
+    /*  END LEGAL */
+#include <Grid/Grid.h>
+
+using namespace std;
+using namespace Grid;
+
+int main (int argc, char ** argv)
+{
+  Grid_init(&argc,&argv);
+
+  int Ls = 12;
+  Coordinate latt4 = GridDefaultLatt();
+
+  GridCartesian         * UGridD   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplexD::Nsimd()),GridDefaultMpi());
+  GridRedBlackCartesian * UrbGridD = SpaceTimeGrid::makeFourDimRedBlackGrid(UGridD);
+  GridCartesian         * FGridD   = SpaceTimeGrid::makeFiveDimGrid(Ls,UGridD);
+  GridRedBlackCartesian * FrbGridD = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGridD);
+
+  GridCartesian         * UGridF   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplexF::Nsimd()),GridDefaultMpi());
+  GridRedBlackCartesian * UrbGridF = SpaceTimeGrid::makeFourDimRedBlackGrid(UGridF);
+  GridCartesian         * FGridF   = SpaceTimeGrid::makeFiveDimGrid(Ls,UGridF);
+  GridRedBlackCartesian * FrbGridF = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGridF);
+
+  
+  std::vector<int> seeds4({1,2,3,4});
+  std::vector<int> seeds5({5,6,7,8});
+  
+  std::cout << GridLogMessage << "Initialising 5d RNG" << std::endl;
+  GridParallelRNG          RNG5(FGridD);  RNG5.SeedFixedIntegers(seeds5);
+  GridParallelRNG          RNG5F(FGridF);  RNG5F.SeedFixedIntegers(seeds5);
+  std::cout << GridLogMessage << "Initialised RNGs" << std::endl;
+
+  LatticeFermionD field_d(FGridD), tmp_d(FGridD);
+  random(RNG5,field_d);
+  RealD norm2_field_d = norm2(field_d);
+  
+  LatticeFermionD2 field_d2(FGridF), tmp_d2(FGridF);
+  random(RNG5F,field_d2);
+  RealD norm2_field_d2 = norm2(field_d2);
+  
+  LatticeFermionF field_f(FGridF);
+  
+  //Test original implementation
+  {
+    std::cout << GridLogMessage << "Testing original implementation" << std::endl;
+    field_f = Zero();
+    precisionChangeOrig(field_f,field_d);
+    RealD Ndiff = (norm2_field_d - norm2(field_f))/norm2_field_d;
+    std::cout << GridLogMessage << (fabs(Ndiff) > 1e-05 ? "!!FAIL" : "Pass") << ": relative norm2 of single and double prec fields differs by " << Ndiff << std::endl;
+    tmp_d = Zero();
+    precisionChangeOrig(tmp_d, field_f);
+    Ndiff = norm2( LatticeFermionD(tmp_d-field_d) ) / norm2_field_d;
+    std::cout << GridLogMessage << (fabs(Ndiff) > 1e-05 ? "!!FAIL" : "Pass") << ": relative norm2 of back-converted and original double prec fields differs by " << Ndiff << std::endl;
+  }
+  //Test new implementation with pregenerated workspace
+  {
+    std::cout << GridLogMessage << "Testing new implementation with pregenerated workspace" << std::endl;
+    precisionChangeWorkspace wk_sp_to_dp(field_d.Grid(),field_f.Grid());
+    precisionChangeWorkspace wk_dp_to_sp(field_f.Grid(),field_d.Grid());
+    
+    field_f = Zero();
+    precisionChange(field_f,field_d,wk_dp_to_sp);
+    RealD Ndiff = (norm2_field_d - norm2(field_f))/norm2_field_d;
+    std::cout << GridLogMessage << (fabs(Ndiff) > 1e-05 ? "!!FAIL" : "Pass") << ": relative norm2 of single and double prec fields differs by " << Ndiff << std::endl;
+    tmp_d = Zero();
+    precisionChange(tmp_d, field_f,wk_sp_to_dp);
+    Ndiff = norm2( LatticeFermionD(tmp_d-field_d) ) / norm2_field_d;
+    std::cout << GridLogMessage << (fabs(Ndiff) > 1e-05 ? "!!FAIL" : "Pass") << ": relative norm2 of back-converted and original double prec fields differs by " << Ndiff << std::endl;
+  }
+  //Test new implementation without pregenerated workspace
+  {
+    std::cout << GridLogMessage << "Testing new implementation without pregenerated workspace" << std::endl;
+    field_f = Zero();
+    precisionChange(field_f,field_d);
+    RealD Ndiff = (norm2_field_d - norm2(field_f))/norm2_field_d;
+    std::cout << GridLogMessage << (fabs(Ndiff) > 1e-05 ? "!!FAIL" : "Pass") << ": relative norm2 of single and double prec fields differs by " << Ndiff << std::endl;
+    tmp_d = Zero();
+    precisionChange(tmp_d, field_f);
+    Ndiff = norm2( LatticeFermionD(tmp_d-field_d) ) / norm2_field_d;
+    std::cout << GridLogMessage << (fabs(Ndiff) > 1e-05 ? "!!FAIL" : "Pass") << ": relative norm2 of back-converted and original double prec fields differs by " << Ndiff << std::endl;
+  } 
+  //Test fast implementation
+  {
+    std::cout << GridLogMessage << "Testing fast (double2) implementation" << std::endl;
+    field_f = Zero();
+    precisionChangeFast(field_f,field_d2);
+    RealD Ndiff = (norm2_field_d2 - norm2(field_f))/norm2_field_d2;
+    std::cout << GridLogMessage << (fabs(Ndiff) > 1e-05 ? "!!FAIL" : "Pass") << ": relative norm2 of single and double prec fields differs by " << Ndiff << std::endl;
+    tmp_d2 = Zero();
+    precisionChangeFast(tmp_d2, field_f);
+    Ndiff = norm2( LatticeFermionD2(tmp_d2-field_d2) ) / norm2_field_d2;
+    std::cout << GridLogMessage << (fabs(Ndiff) > 1e-05 ? "!!FAIL" : "Pass") << ": relative norm2 of back-converted and original double prec fields differs by " << Ndiff << std::endl;
+  }
+  std::cout << "Done" << std::endl;
+  
+  Grid_finalize();
+}
diff --git a/tests/solver/Test_dwf_relupcg_prec.cc b/tests/solver/Test_dwf_relupcg_prec.cc
new file mode 100644
index 00000000..1d8c022a
--- /dev/null
+++ b/tests/solver/Test_dwf_relupcg_prec.cc
@@ -0,0 +1,143 @@
+    /*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./tests/solver/Test_dwf_relupcg_prec.cc
+
+    Copyright (C) 2015
+
+Author: Christopher Kelly <ckelly@bnl.gov>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+    *************************************************************************************/
+    /*  END LEGAL */
+#include <Grid/Grid.h>
+
+using namespace std;
+using namespace Grid;
+
+int main (int argc, char ** argv)
+{
+  Grid_init(&argc,&argv);
+
+  double relup_delta = 0.2;
+  for(int i=1;i<argc-1;i++){
+    std::string sarg = argv[i];
+    if(sarg == "--relup_delta"){
+      std::stringstream ss; ss << argv[i+1]; ss >> relup_delta;
+      std::cout << GridLogMessage << "Set reliable update Delta to " << relup_delta << std::endl;
+    }
+  }   
+  
+  const int Ls=12;
+
+  { 
+  GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplexD::Nsimd()),GridDefaultMpi());
+  GridRedBlackCartesian * UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
+  GridCartesian         * FGrid   = SpaceTimeGrid::makeFiveDimGrid(Ls,UGrid);
+  GridRedBlackCartesian * FrbGrid = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGrid);
+
+  GridCartesian         * UGrid_f   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplexF::Nsimd()),GridDefaultMpi());
+  GridRedBlackCartesian * UrbGrid_f = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid_f);
+  GridCartesian         * FGrid_f   = SpaceTimeGrid::makeFiveDimGrid(Ls,UGrid_f);
+  GridRedBlackCartesian * FrbGrid_f = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGrid_f);
+  
+  std::vector<int> seeds4({1,2,3,4});
+  std::vector<int> seeds5({5,6,7,8});
+  GridParallelRNG          RNG5(FGrid);  RNG5.SeedFixedIntegers(seeds5);
+  GridParallelRNG          RNG4(UGrid);  RNG4.SeedFixedIntegers(seeds4);
+
+  LatticeFermionD    src(FGrid); random(RNG5,src);
+  LatticeFermionD result(FGrid); result=Zero();
+  LatticeGaugeFieldD Umu(UGrid);
+  LatticeGaugeFieldF Umu_f(UGrid_f); 
+  
+  SU<Nc>::HotConfiguration(RNG4,Umu);
+
+  precisionChange(Umu_f,Umu);
+  
+  RealD mass=0.1;
+  RealD M5=1.8;
+  DomainWallFermionD Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+  DomainWallFermionF Ddwf_f(Umu_f,*FGrid_f,*FrbGrid_f,*UGrid_f,*UrbGrid_f,mass,M5);
+
+  LatticeFermionD    src_o(FrbGrid);
+  LatticeFermionD result_o(FrbGrid);
+  LatticeFermionD result_o_2(FrbGrid);
+  pickCheckerboard(Odd,src_o,src);
+  result_o.Checkerboard() = Odd;
+  result_o = Zero();
+  result_o_2.Checkerboard() = Odd;
+  result_o_2 = Zero();
+
+  SchurDiagMooeeOperator<DomainWallFermionD,LatticeFermionD> HermOpEO(Ddwf);
+  SchurDiagMooeeOperator<DomainWallFermionF,LatticeFermionF> HermOpEO_f(Ddwf_f);
+
+  std::cout << GridLogMessage << "::::::::::::: Starting mixed CG" << std::endl;
+  ConjugateGradientReliableUpdate<LatticeFermionD,LatticeFermionF> mCG(1e-8, 10000, relup_delta, FrbGrid_f, HermOpEO_f, HermOpEO);
+  double t1,t2,flops;
+  double MdagMsiteflops = 1452; // Mobius (real coeffs)
+  // CG overhead: 8 inner product, 4+8 axpy_norm, 4+4 linear comb (2 of)
+  double CGsiteflops = (8+4+8+4+4)*Nc*Ns ;
+  std:: cout << " MdagM site flops = "<< 4*MdagMsiteflops<<std::endl;
+  std:: cout << " CG    site flops = "<< CGsiteflops <<std::endl;
+  int iters, iters_cleanup, relups, tot_iters;
+  for(int i=0;i<10;i++){
+    result_o = Zero();
+    t1=usecond();
+    mCG(src_o,result_o);
+    t2=usecond();
+    iters = mCG.IterationsToComplete; //Number of single prec CG iterations
+    iters_cleanup = mCG.IterationsToCleanup;
+    relups = mCG.ReliableUpdatesPerformed;
+    tot_iters  = iters + iters_cleanup + relups; //relup cost MdagM application in double
+    
+    flops = MdagMsiteflops*4*FrbGrid->gSites()*tot_iters;
+    flops+= CGsiteflops*FrbGrid->gSites()*tot_iters;
+    std::cout << " SinglePrecision single prec iterations/sec "<< iters/(t2-t1)*1000.*1000.<<std::endl;
+    std::cout << " SinglePrecision double prec cleanup iterations/sec "<< iters_cleanup/(t2-t1)*1000.*1000.<<std::endl;
+    std::cout << " SinglePrecision reliable updates/sec "<< relups/(t2-t1)*1000.*1000.<<std::endl;
+    std::cout << " SinglePrecision GF/s "<< flops/(t2-t1)/1000.<<std::endl;
+  }
+  std::cout << GridLogMessage << "::::::::::::: Starting regular CG" << std::endl;
+  ConjugateGradient<LatticeFermionD> CG(1.0e-8,10000);
+  for(int i=0;i<1;i++){
+    result_o_2 = Zero();
+    t1=usecond();
+    CG(HermOpEO,src_o,result_o_2);
+    t2=usecond();
+    iters = CG.IterationsToComplete;
+    flops = MdagMsiteflops*4*FrbGrid->gSites()*iters; 
+    flops+= CGsiteflops*FrbGrid->gSites()*iters;
+    
+    std::cout << " DoublePrecision iterations/sec "<< iters/(t2-t1)*1000.*1000.<<std::endl;
+    std::cout << " DoublePrecision GF/s "<< flops/(t2-t1)/1000.<<std::endl;
+  }
+  
+  //  MemoryManager::Print();
+
+  LatticeFermionD diff_o(FrbGrid);
+  RealD diff = axpy_norm(diff_o, -1.0, result_o, result_o_2);
+
+  std::cout << GridLogMessage << "::::::::::::: Diff between mixed and regular CG: " << diff << std::endl;
+  }
+  
+  MemoryManager::Print();
+
+  Grid_finalize();
+}

From e82cf1d31159363c2d68908c935910ce203dadf0 Mon Sep 17 00:00:00 2001
From: Christopher Kelly <ckelly@bnl.gov>
Date: Thu, 23 Feb 2023 09:45:29 -0500
Subject: [PATCH 123/146] Further prec-change improvements

Mixed prec CG algorithm has been modified to precompute precision change workspaces

As the original Test_dwf_mixedcg_prec has been coopted to do a performance stability and reproducibility test, requiring the single-prec CG to be run 200 times, I have created a new version of Test_dwf_mixedcg_prec in the solver subdirectory that just does the mixed vs double CG test
---
 .../iterative/ConjugateGradientMixedPrec.h    |   9 +-
 tests/solver/Test_dwf_mixedcg_prec.cc         | 122 ++++++++++++++++++
 2 files changed, 128 insertions(+), 3 deletions(-)
 create mode 100644 tests/solver/Test_dwf_mixedcg_prec.cc

diff --git a/Grid/algorithms/iterative/ConjugateGradientMixedPrec.h b/Grid/algorithms/iterative/ConjugateGradientMixedPrec.h
index 31ac55e0..27fee791 100644
--- a/Grid/algorithms/iterative/ConjugateGradientMixedPrec.h
+++ b/Grid/algorithms/iterative/ConjugateGradientMixedPrec.h
@@ -108,7 +108,10 @@ NAMESPACE_BEGIN(Grid);
     GridStopWatch PrecChangeTimer;
     
     Integer &outer_iter = TotalOuterIterations; //so it will be equal to the final iteration count
-      
+
+    precisionChangeWorkspace pc_wk_sp_to_dp(DoublePrecGrid, SinglePrecGrid);
+    precisionChangeWorkspace pc_wk_dp_to_sp(SinglePrecGrid, DoublePrecGrid);
+    
     for(outer_iter = 0; outer_iter < MaxOuterIterations; outer_iter++){
       //Compute double precision rsd and also new RHS vector.
       Linop_d.HermOp(sol_d, tmp_d);
@@ -123,7 +126,7 @@ NAMESPACE_BEGIN(Grid);
       while(norm * inner_tol * inner_tol < stop) inner_tol *= 2;  // inner_tol = sqrt(stop/norm) ??
 
       PrecChangeTimer.Start();
-      precisionChange(src_f, src_d);
+      precisionChange(src_f, src_d, pc_wk_dp_to_sp);
       PrecChangeTimer.Stop();
       
       sol_f = Zero();
@@ -142,7 +145,7 @@ NAMESPACE_BEGIN(Grid);
       
       //Convert sol back to double and add to double prec solution
       PrecChangeTimer.Start();
-      precisionChange(tmp_d, sol_f);
+      precisionChange(tmp_d, sol_f, pc_wk_sp_to_dp);
       PrecChangeTimer.Stop();
       
       axpy(sol_d, 1.0, tmp_d, sol_d);
diff --git a/tests/solver/Test_dwf_mixedcg_prec.cc b/tests/solver/Test_dwf_mixedcg_prec.cc
new file mode 100644
index 00000000..dc88018e
--- /dev/null
+++ b/tests/solver/Test_dwf_mixedcg_prec.cc
@@ -0,0 +1,122 @@
+    /*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./tests/Test_dwf_cg_prec.cc
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+    *************************************************************************************/
+    /*  END LEGAL */
+#include <Grid/Grid.h>
+
+//using namespace std;
+using namespace Grid;
+
+int main (int argc, char ** argv)
+{
+  Grid_init(&argc,&argv);
+
+  const int Ls=12;
+
+  std::cout << GridLogMessage << "::::: NB: to enable a quick bit reproducibility check use the --checksums flag. " << std::endl;
+
+  GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplexD::Nsimd()),GridDefaultMpi());
+  GridRedBlackCartesian * UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
+  GridCartesian         * FGrid   = SpaceTimeGrid::makeFiveDimGrid(Ls,UGrid);
+  GridRedBlackCartesian * FrbGrid = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGrid);
+
+  GridCartesian         * UGrid_f   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplexF::Nsimd()),GridDefaultMpi());
+  GridRedBlackCartesian * UrbGrid_f = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid_f);
+  GridCartesian         * FGrid_f   = SpaceTimeGrid::makeFiveDimGrid(Ls,UGrid_f);
+  GridRedBlackCartesian * FrbGrid_f = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGrid_f);
+  
+  std::vector<int> seeds4({1,2,3,4});
+  std::vector<int> seeds5({5,6,7,8});
+  GridParallelRNG          RNG5(FGrid);  RNG5.SeedFixedIntegers(seeds5);
+  GridParallelRNG          RNG4(UGrid);  RNG4.SeedFixedIntegers(seeds4);
+
+  LatticeFermionD    src(FGrid); random(RNG5,src);
+  LatticeFermionD result(FGrid); result=Zero();
+  LatticeGaugeFieldD Umu(UGrid);
+  LatticeGaugeFieldF Umu_f(UGrid_f); 
+  
+  SU<Nc>::HotConfiguration(RNG4,Umu);
+
+  precisionChange(Umu_f,Umu);
+  
+  RealD mass=0.1;
+  RealD M5=1.8;
+  DomainWallFermionD Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+  DomainWallFermionF Ddwf_f(Umu_f,*FGrid_f,*FrbGrid_f,*UGrid_f,*UrbGrid_f,mass,M5);
+
+  LatticeFermionD    src_o(FrbGrid);
+  LatticeFermionD result_o(FrbGrid);
+  LatticeFermionD result_o_2(FrbGrid);
+  pickCheckerboard(Odd,src_o,src);
+  result_o.Checkerboard() = Odd;
+  result_o = Zero();
+  result_o_2.Checkerboard() = Odd;
+  result_o_2 = Zero();
+
+  SchurDiagMooeeOperator<DomainWallFermionD,LatticeFermionD> HermOpEO(Ddwf);
+  SchurDiagMooeeOperator<DomainWallFermionF,LatticeFermionF> HermOpEO_f(Ddwf_f);
+
+  std::cout << GridLogMessage << "::::::::::::: Starting mixed CG" << std::endl;
+  MixedPrecisionConjugateGradient<LatticeFermionD,LatticeFermionF> mCG(1.0e-8, 10000, 50, FrbGrid_f, HermOpEO_f, HermOpEO);
+  double t1,t2,flops;
+  double MdagMsiteflops = 1452; // Mobius (real coeffs)
+  // CG overhead: 8 inner product, 4+8 axpy_norm, 4+4 linear comb (2 of)
+  double CGsiteflops = (8+4+8+4+4)*Nc*Ns ;
+  std:: cout << " MdagM site flops = "<< 4*MdagMsiteflops<<std::endl;
+  std:: cout << " CG    site flops = "<< CGsiteflops <<std::endl;
+
+  result_o = Zero();
+  t1=usecond();
+  mCG(src_o,result_o);
+  t2=usecond();
+  int iters = mCG.TotalInnerIterations; //Number of inner CG iterations
+  flops = MdagMsiteflops*4*FrbGrid->gSites()*iters;
+  flops+= CGsiteflops*FrbGrid->gSites()*iters;
+  std::cout << " SinglePrecision iterations/sec "<< iters/(t2-t1)*1000.*1000.<<std::endl;
+  std::cout << " SinglePrecision GF/s "<< flops/(t2-t1)/1000.<<std::endl;
+
+  std::cout << GridLogMessage << "::::::::::::: Starting regular CG" << std::endl;
+  ConjugateGradient<LatticeFermionD> CG(1.0e-8,10000);
+  result_o_2 = Zero();
+  t1=usecond();
+  CG(HermOpEO,src_o,result_o_2);
+  t2=usecond();
+  iters = CG.IterationsToComplete;
+  flops = MdagMsiteflops*4*FrbGrid->gSites()*iters; 
+  flops+= CGsiteflops*FrbGrid->gSites()*iters;
+  
+  std::cout << " DoublePrecision iterations/sec "<< iters/(t2-t1)*1000.*1000.<<std::endl;
+  std::cout << " DoublePrecision GF/s "<< flops/(t2-t1)/1000.<<std::endl;
+
+  LatticeFermionD diff_o(FrbGrid);
+  RealD diff = axpy_norm(diff_o, -1.0, result_o, result_o_2);
+
+  std::cout << GridLogMessage << "::::::::::::: Diff between mixed and regular CG: " << diff << std::endl;
+
+  MemoryManager::Print();
+
+  Grid_finalize();
+}

From 83d86943db53e3b05f3ec7523b1afa8230531e40 Mon Sep 17 00:00:00 2001
From: Christopher Kelly <ckelly@phys.columbia.edu>
Date: Thu, 23 Feb 2023 13:09:45 -0500
Subject: [PATCH 124/146] Fixed compile bug in MemoryManagerShared caused by
 Audit function not being passed a string

---
 Grid/allocator/MemoryManagerShared.cc | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/Grid/allocator/MemoryManagerShared.cc b/Grid/allocator/MemoryManagerShared.cc
index 2434ad47..e291ef89 100644
--- a/Grid/allocator/MemoryManagerShared.cc
+++ b/Grid/allocator/MemoryManagerShared.cc
@@ -13,7 +13,7 @@ uint64_t  MemoryManager::DeviceToHostBytes;
 uint64_t  MemoryManager::HostToDeviceXfer;
 uint64_t  MemoryManager::DeviceToHostXfer;
 
-void  MemoryManager::Audit(void){};
+void  MemoryManager::Audit(std::string s){};
 void  MemoryManager::ViewClose(void* AccPtr,ViewMode mode){};
 void *MemoryManager::ViewOpen(void* CpuPtr,size_t bytes,ViewMode mode,ViewAdvise hint){ return CpuPtr; };
 int   MemoryManager::isOpen   (void* CpuPtr) { return 0;}

From 87697eb07ebaf510a776f79bc409efebc7860a23 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 14 Mar 2023 09:07:36 -0700
Subject: [PATCH 125/146] SHared compile

---
 Grid/allocator/MemoryManagerShared.cc | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/Grid/allocator/MemoryManagerShared.cc b/Grid/allocator/MemoryManagerShared.cc
index 2434ad47..e291ef89 100644
--- a/Grid/allocator/MemoryManagerShared.cc
+++ b/Grid/allocator/MemoryManagerShared.cc
@@ -13,7 +13,7 @@ uint64_t  MemoryManager::DeviceToHostBytes;
 uint64_t  MemoryManager::HostToDeviceXfer;
 uint64_t  MemoryManager::DeviceToHostXfer;
 
-void  MemoryManager::Audit(void){};
+void  MemoryManager::Audit(std::string s){};
 void  MemoryManager::ViewClose(void* AccPtr,ViewMode mode){};
 void *MemoryManager::ViewOpen(void* CpuPtr,size_t bytes,ViewMode mode,ViewAdvise hint){ return CpuPtr; };
 int   MemoryManager::isOpen   (void* CpuPtr) { return 0;}

From cad5b187ddcbce278bdca06b42dfb4ac6bbc59d0 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 14 Mar 2023 09:08:16 -0700
Subject: [PATCH 126/146] Cleanup

---
 Grid/qcd/action/fermion/WilsonKernels.h | 7 -------
 1 file changed, 7 deletions(-)

diff --git a/Grid/qcd/action/fermion/WilsonKernels.h b/Grid/qcd/action/fermion/WilsonKernels.h
index 68422f28..2d868c27 100644
--- a/Grid/qcd/action/fermion/WilsonKernels.h
+++ b/Grid/qcd/action/fermion/WilsonKernels.h
@@ -52,13 +52,6 @@ public:
   typedef AcceleratorVector<int,STENCIL_MAX> StencilVector;   
 public:
 
-#ifdef GRID_SYCL
-#define SYCL_HACK
-#endif  
-#ifdef SYCL_HACK
-  static void HandDhopSiteSycl(StencilVector st_perm,StencilEntry *st_p, SiteDoubledGaugeField *U,SiteHalfSpinor  *buf,
-			       int ss,int sU,const SiteSpinor *in, SiteSpinor *out);
-#endif
   
   static void DhopKernel(int Opt,StencilImpl &st,  DoubledGaugeField &U, SiteHalfSpinor * buf,
 			 int Ls, int Nsite, const FermionField &in, FermionField &out,

From eeb6e0a6e371407ec4d445f7d0459e6a951c2449 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 14 Mar 2023 09:10:27 -0700
Subject: [PATCH 127/146] Renable cache blocking and efficient UPI type SHM
 comms

---
 .../WilsonFermion5DImplementation.h           |   4 +
 .../WilsonFermionImplementation.h             |   3 +
 .../WilsonKernelsImplementation.h             |   3 +-
 Grid/stencil/Stencil.h                        | 104 ++++++++++++++++--
 configure.ac                                  |   8 ++
 5 files changed, 112 insertions(+), 10 deletions(-)

diff --git a/Grid/qcd/action/fermion/implementation/WilsonFermion5DImplementation.h b/Grid/qcd/action/fermion/implementation/WilsonFermion5DImplementation.h
index 4ca24789..1ddb30ba 100644
--- a/Grid/qcd/action/fermion/implementation/WilsonFermion5DImplementation.h
+++ b/Grid/qcd/action/fermion/implementation/WilsonFermion5DImplementation.h
@@ -63,6 +63,10 @@ WilsonFermion5D<Impl>::WilsonFermion5D(GaugeField &_Umu,
   _tmp(&FiveDimRedBlackGrid),
   Dirichlet(0)
 {
+  Stencil.lo     = &Lebesgue;
+  StencilEven.lo = &LebesgueEvenOdd;
+  StencilOdd.lo  = &LebesgueEvenOdd;
+  
   // some assertions
   assert(FiveDimGrid._ndimension==5);
   assert(FourDimGrid._ndimension==4);
diff --git a/Grid/qcd/action/fermion/implementation/WilsonFermionImplementation.h b/Grid/qcd/action/fermion/implementation/WilsonFermionImplementation.h
index 2833fdc4..1a262533 100644
--- a/Grid/qcd/action/fermion/implementation/WilsonFermionImplementation.h
+++ b/Grid/qcd/action/fermion/implementation/WilsonFermionImplementation.h
@@ -60,6 +60,9 @@ WilsonFermion<Impl>::WilsonFermion(GaugeField &_Umu, GridCartesian &Fgrid,
       _tmp(&Hgrid),
       anisotropyCoeff(anis)
 {
+  Stencil.lo     = &Lebesgue;
+  StencilEven.lo = &LebesgueEvenOdd;
+  StencilOdd.lo  = &LebesgueEvenOdd;
   // Allocate the required comms buffer
   ImportGauge(_Umu);
   if  (anisotropyCoeff.isAnisotropic){
diff --git a/Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h b/Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h
index bdba7cb2..b307fad4 100644
--- a/Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h
+++ b/Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h
@@ -433,7 +433,8 @@ void WilsonKernels<Impl>::DhopDirKernel( StencilImpl &st, DoubledGaugeField &U,S
     });									
 
 #define ASM_CALL(A)							\
-  thread_for( ss, Nsite, {						\
+  thread_for( sss, Nsite, {						\
+    int ss = st.lo->Reorder(sss);					\
     int sU = ss;							\
     int sF = ss*Ls;							\
     WilsonKernels<Impl>::A(st_v,U_v,buf,sF,sU,Ls,1,in_v,out_v);		\
diff --git a/Grid/stencil/Stencil.h b/Grid/stencil/Stencil.h
index cffede12..a74b720d 100644
--- a/Grid/stencil/Stencil.h
+++ b/Grid/stencil/Stencil.h
@@ -290,9 +290,9 @@ public:
 
 protected:
   GridBase *                        _grid;
-
 public:
   GridBase *Grid(void) const { return _grid; }
+  LebesgueOrder *lo;
 
   ////////////////////////////////////////////////////////////////////////
   // Needed to conveniently communicate gparity parameters into GPU memory
@@ -337,6 +337,7 @@ public:
   ////////////////////////////////////////
   // Stencil query
   ////////////////////////////////////////
+#ifdef SHM_FAST_PATH
   inline int SameNode(int point) {
 
     int dimension    = this->_directions[point];
@@ -356,7 +357,40 @@ public:
     if ( displacement == 0 ) return 1;
     return 0;
   }
+#else
+  // 
+  inline int SameNode(int point) {
 
+    int dimension    = this->_directions[point];
+    int displacement = this->_distances[point];
+
+    int pd              = _grid->_processors[dimension];
+    int fd              = _grid->_fdimensions[dimension];
+    int ld              = _grid->_ldimensions[dimension];
+    int rd              = _grid->_rdimensions[dimension];
+    int simd_layout     = _grid->_simd_layout[dimension];
+    int comm_dim        = _grid->_processors[dimension] >1 ;
+ 
+    int recv_from_rank;
+    int xmit_to_rank;
+
+    if ( ! comm_dim ) return 1;
+
+    int nbr_proc;
+    if (displacement>0) nbr_proc = 1;
+    else                 nbr_proc = pd-1;
+
+    // FIXME  this logic needs to be sorted for three link term
+    //    assert( (displacement==1) || (displacement==-1));
+    // Present hack only works for >= 4^4 subvol per node
+    _grid->ShiftedRanks(dimension,nbr_proc,xmit_to_rank,recv_from_rank);
+
+    void *shm = (void *) _grid->ShmBufferTranslate(recv_from_rank,this->u_recv_buf_p);
+
+    if ( (shm==NULL) ) return 0;
+    return 1;
+  }
+#endif
   //////////////////////////////////////////
   // Comms packet queue for asynch thread
   // Use OpenMP Tasks for cleaner ???
@@ -1056,7 +1090,7 @@ public:
     int comms_recv   = this->_comms_recv[point];
     int comms_partial_send   = this->_comms_partial_send[point] ;
     int comms_partial_recv   = this->_comms_partial_recv[point] ;
-
+    
     assert(rhs.Grid()==_grid);
     //	  conformable(_grid,rhs.Grid());
 
@@ -1127,11 +1161,32 @@ public:
 	  recv_buf=this->u_recv_buf_p;
 	}
 
+	// potential SHM fast path for intranode
+	int shm_send=0;
+	int shm_recv=0;
+#ifdef SHM_FAST_PATH
+	// Put directly in place if we can
+	send_buf = (cobj *)_grid->ShmBufferTranslate(xmit_to_rank,recv_buf);
+	if ( (send_buf==NULL) ) {
+	  shm_send=0;
+	  send_buf = this->u_send_buf_p;
+	} else {
+	  shm_send=1;
+	}
+	void *test_ptr = _grid->ShmBufferTranslate(recv_from_rank,recv_buf);
+	if ( test_ptr != NULL ) shm_recv = 1;
+	//	static int printed;
+	//	if (!printed){
+	  //	  std::cout << " GATHER FAST PATH SHM "<<shm_send<< " "<<shm_recv<<std::endl;
+	//	  printed = 1;
+	//	}
+#else
 	////////////////////////////////////////////////////////
 	// Gather locally
 	////////////////////////////////////////////////////////
 	send_buf = this->u_send_buf_p; // Gather locally, must send
 	assert(send_buf!=NULL);
+#endif
 
 	//	std::cout << " GatherPlaneSimple partial send "<< comms_partial_send<<std::endl;
 	compressor::Gather_plane_simple(face_table[face_idx],rhs,send_buf,compress,comm_off,so,comms_partial_send);
@@ -1143,10 +1198,13 @@ public:
 	  // Build a list of things to do after we synchronise GPUs
 	  // Start comms now???
 	  ///////////////////////////////////////////////////////////
+	  int do_send = (comms_send|comms_partial_send) && (!shm_send );
+	  int do_recv = (comms_send|comms_partial_send) && (!shm_recv );
+	  
 	  AddPacket((void *)&send_buf[comm_off],
 		    (void *)&recv_buf[comm_off],
-		    xmit_to_rank, comms_send|comms_partial_send,
-		    recv_from_rank, comms_recv|comms_partial_recv,
+		    xmit_to_rank, do_send,
+		    recv_from_rank, do_recv,
 		    xbytes,rbytes);
 	}
 
@@ -1288,19 +1346,47 @@ public:
 
 	    int recv_from_rank;
 	    int xmit_to_rank;
-
+	    int shm_send=0;
+	    int shm_recv=0;
 	    _grid->ShiftedRanks(dimension,nbr_proc,xmit_to_rank,recv_from_rank);
-
+#ifdef SHM_FAST_PATH
+  #warning STENCIL SHM FAST PATH SELECTED
+	    // shm == receive pointer         if offnode
+	    // shm == Translate[send pointer] if on node -- my view of his send pointer
+	    cobj *shm = (cobj *) _grid->ShmBufferTranslate(recv_from_rank,sp);
+	    if (shm==NULL) {
+	      shm = rp;
+	      // we found a packet that comes from MPI and contributes to this shift.
+	      // is_same_node is only used in the WilsonStencil, and gets set for this point in the stencil.
+	      // Kernel will add the exterior_terms except if is_same_node.
+	      // leg of stencil
+	      shm_recv=0;
+	    } else {
+	      shm_recv=1;
+	    }
+	    rpointers[i] = shm;
+	    // Test send side
+	    void *test_ptr = (void *) _grid->ShmBufferTranslate(xmit_to_rank,sp);
+	    if ( test_ptr != NULL ) shm_send = 1;
+	    //	    static int printed;
+	    //	    if (!printed){
+	    //	      std::cout << " GATHERSIMD FAST PATH SHM "<<shm_send<< " "<<shm_recv<<std::endl;
+	    //	      printed = 1;
+	    //	    }
+#else
 	    rpointers[i] = rp;
+#endif
 	    
 	    int duplicate = CheckForDuplicate(dimension,sx,nbr_proc,(void *)rp,i,xbytes,rbytes,cbmask);
 	    if ( !duplicate  ) { 
 	      if ( (bytes != rbytes) && (rbytes!=0) ){
 		acceleratorMemSet(rp,0,bytes); // Zero prefill comms buffer to zero
 	      }
+	      int do_send = (comms_send|comms_partial_send) && (!shm_send );
+	      int do_recv = (comms_send|comms_partial_send) && (!shm_recv );
 	      AddPacket((void *)sp,(void *)rp,
-			xmit_to_rank,comms_send|comms_partial_send,
-			recv_from_rank,comms_recv|comms_partial_recv,
+			xmit_to_rank,do_send,
+			recv_from_rank,do_send,
 			xbytes,rbytes);
 	    }
 
@@ -1310,7 +1396,7 @@ public:
 
 	  }
 	}
-
+	// rpointer may be doing a remote read in the gather over SHM
 	if ( comms_recv|comms_partial_recv ) {
 	  AddMerge(&this->u_recv_buf_p[comm_off],rpointers,reduced_buffer_size,permute_type,Mergers);
 	}
diff --git a/configure.ac b/configure.ac
index 2e6199c7..fedca3fe 100644
--- a/configure.ac
+++ b/configure.ac
@@ -646,6 +646,14 @@ case ${ac_SHM_FORCE_MPI} in
       ;;
      *) ;;
 esac
+############### force MPI in SMP
+AC_ARG_ENABLE([shm-fast-path],[AS_HELP_STRING([--enable-shm-fast-path],[Allow kernels to remote copy over intranode])],[ac_SHM_FAST_PATH=${enable_shm_fast_path}],[ac_SHM_FAST_PATH=no])
+case ${ac_SHM_FAST_PATH} in
+     yes)
+        AC_DEFINE([SHM_FAST_PATH],[1],[SHM_FAST_PATH] )
+      ;;
+     *) ;;
+esac
 
 ############### communication type selection
 AC_ARG_ENABLE([comms-threads],[AS_HELP_STRING([--enable-comms-threads | --disable-comms-threads],[Use multiple threads in MPI calls])],[ac_COMMS_THREADS=${enable_comms_threads}],[ac_COMMS_THREADS=yes])

From f36b87deb5451a3c1d47c4a25502319395478c3e Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 14 Mar 2023 12:09:00 -0700
Subject: [PATCH 128/146] syscall fix

---
 Grid/communicator/SharedMemoryMPI.cc | 1 +
 1 file changed, 1 insertion(+)

diff --git a/Grid/communicator/SharedMemoryMPI.cc b/Grid/communicator/SharedMemoryMPI.cc
index 4993a02e..9a273dc4 100644
--- a/Grid/communicator/SharedMemoryMPI.cc
+++ b/Grid/communicator/SharedMemoryMPI.cc
@@ -29,6 +29,7 @@ Author: Christoph Lehner <christoph@lhnr.de>
 
 #include <Grid/GridCore.h>
 #include <pwd.h>
+#include <syscall.h>
 
 #ifdef GRID_CUDA
 #include <cuda_runtime_api.h>

From 14cc142a1424d02883887572cef47d1677a91fb1 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 14 Mar 2023 12:09:26 -0700
Subject: [PATCH 129/146] Warning remove

---
 Grid/stencil/Stencil.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/Grid/stencil/Stencil.h b/Grid/stencil/Stencil.h
index a74b720d..29aa876f 100644
--- a/Grid/stencil/Stencil.h
+++ b/Grid/stencil/Stencil.h
@@ -387,7 +387,7 @@ public:
 
     void *shm = (void *) _grid->ShmBufferTranslate(recv_from_rank,this->u_recv_buf_p);
 
-    if ( (shm==NULL) ) return 0;
+    if ( shm==NULL ) return 0;
     return 1;
   }
 #endif

From 861e5d7f4c0bb56cbb5de550cb2a14e5c1ad15f7 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 14 Mar 2023 12:10:02 -0700
Subject: [PATCH 130/146] SYCL version update. Why do they keep making
 incompatible changes

---
 Grid/threads/Accelerator.h | 16 +++++++++++-----
 1 file changed, 11 insertions(+), 5 deletions(-)

diff --git a/Grid/threads/Accelerator.h b/Grid/threads/Accelerator.h
index 5ac36d15..04ae885b 100644
--- a/Grid/threads/Accelerator.h
+++ b/Grid/threads/Accelerator.h
@@ -248,17 +248,23 @@ inline int  acceleratorIsCommunicable(void *ptr)
 //////////////////////////////////////////////
 // SyCL acceleration
 //////////////////////////////////////////////
-#ifdef GRID_SYCL
-NAMESPACE_END(Grid);
-#include <CL/sycl.hpp>
-#include <CL/sycl/usm.hpp>
 
+#ifdef GRID_SYCL
 #define GRID_SYCL_LEVEL_ZERO_IPC
 
-#ifdef GRID_SYCL_LEVEL_ZERO_IPC
+NAMESPACE_END(Grid);
+#if 0
+#include <CL/sycl.hpp>
+#include <CL/sycl/usm.hpp>
 #include <level_zero/ze_api.h>
 #include <CL/sycl/backend/level_zero.hpp>
+#else
+#include <sycl/CL/sycl.hpp>
+#include <sycl/usm.hpp>
+#include <level_zero/ze_api.h>
+#include <sycl/ext/oneapi/backend/level_zero.hpp>
 #endif
+
 NAMESPACE_BEGIN(Grid);
 
 extern cl::sycl::queue *theGridAccelerator;

From a997d24743abb33c2ec1778bafd805793d1919f7 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 14 Mar 2023 12:10:31 -0700
Subject: [PATCH 131/146] Remove nofma

---
 systems/PVC/config-command | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/systems/PVC/config-command b/systems/PVC/config-command
index 3f5b5993..dc6b222c 100644
--- a/systems/PVC/config-command
+++ b/systems/PVC/config-command
@@ -11,5 +11,5 @@ INSTALL=/nfs/site/home/azusayax/install
 	--enable-unified=yes \
 	CXX=mpicxx \
 	LDFLAGS="-fsycl-device-code-split=per_kernel -fsycl-device-lib=all -lze_loader -L$INSTALL/lib" \
-	CXXFLAGS="-cxx=dpcpp -fsycl-unnamed-lambda -fsycl -no-fma -I$INSTALL/include -Wtautological-constant-compare"
+	CXXFLAGS="-cxx=icpx -fsycl-unnamed-lambda -fsycl -I$INSTALL/include -Wno-tautological-constant-compare"
 

From 39c0815d9e6a338cf294af6fa6b759401888a5cc Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 21 Mar 2023 08:57:29 -0400
Subject: [PATCH 132/146] WriteDiscard

---
 Grid/lattice/Lattice_base.h | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/Grid/lattice/Lattice_base.h b/Grid/lattice/Lattice_base.h
index 34f13fa6..d6289de2 100644
--- a/Grid/lattice/Lattice_base.h
+++ b/Grid/lattice/Lattice_base.h
@@ -129,7 +129,7 @@ public:
     
     auto exprCopy = expr;
     ExpressionViewOpen(exprCopy);
-    auto me  = View(AcceleratorWrite);
+    auto me  = View(AcceleratorWriteDiscard);
     accelerator_for(ss,me.size(),vobj::Nsimd(),{
       auto tmp = eval(ss,exprCopy);
       coalescedWrite(me[ss],tmp);
@@ -152,7 +152,7 @@ public:
 
     auto exprCopy = expr;
     ExpressionViewOpen(exprCopy);
-    auto me  = View(AcceleratorWrite);
+    auto me  = View(AcceleratorWriteDiscard);
     accelerator_for(ss,me.size(),vobj::Nsimd(),{
       auto tmp = eval(ss,exprCopy);
       coalescedWrite(me[ss],tmp);
@@ -174,7 +174,7 @@ public:
     this->checkerboard=cb;
     auto exprCopy = expr;
     ExpressionViewOpen(exprCopy);
-    auto me  = View(AcceleratorWrite);
+    auto me  = View(AcceleratorWriteDiscard);
     accelerator_for(ss,me.size(),vobj::Nsimd(),{
       auto tmp = eval(ss,exprCopy);
       coalescedWrite(me[ss],tmp);
@@ -288,8 +288,8 @@ public:
     typename std::enable_if<!std::is_same<robj,vobj>::value,int>::type i=0;
     conformable(*this,r);
     this->checkerboard = r.Checkerboard();
-    auto me =   View(AcceleratorWrite);
     auto him= r.View(AcceleratorRead);
+    auto me =   View(AcceleratorWriteDiscard);
     accelerator_for(ss,me.size(),vobj::Nsimd(),{
       coalescedWrite(me[ss],him(ss));
     });
@@ -303,8 +303,8 @@ public:
   inline Lattice<vobj> & operator = (const Lattice<vobj> & r){
     this->checkerboard = r.Checkerboard();
     conformable(*this,r);
-    auto me =   View(AcceleratorWrite);
     auto him= r.View(AcceleratorRead);
+    auto me =   View(AcceleratorWriteDiscard);
     accelerator_for(ss,me.size(),vobj::Nsimd(),{
       coalescedWrite(me[ss],him(ss));
     });

From e1c326558a54a38e138e9eeaead1aaabaab57440 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 21 Mar 2023 08:53:56 -0700
Subject: [PATCH 133/146] COmms improvements

---
 Grid/qcd/action/fermion/WilsonCompressor.h    | 34 ++++++++++---------
 .../WilsonKernelsImplementation.h             | 11 ++++++
 Grid/stencil/Stencil.h                        |  4 +--
 Grid/threads/Accelerator.h                    |  4 +--
 systems/PVC/benchmarks/run-1tile.sh           |  2 +-
 systems/PVC/benchmarks/run-2tile-mpi.sh       |  4 +--
 systems/PVC/benchmarks/wrap.sh                | 10 +++---
 systems/PVC/config-command                    |  6 ++--
 systems/PVC/setup.sh                          |  3 +-
 9 files changed, 46 insertions(+), 32 deletions(-)

diff --git a/Grid/qcd/action/fermion/WilsonCompressor.h b/Grid/qcd/action/fermion/WilsonCompressor.h
index fd1bbe89..5523ae8a 100644
--- a/Grid/qcd/action/fermion/WilsonCompressor.h
+++ b/Grid/qcd/action/fermion/WilsonCompressor.h
@@ -484,24 +484,26 @@ public:
 
     int dag = compress.dag;
     int face_idx=0;
+#define vet_same_node(a,b) \
+      { auto tmp = b;  }
     if ( dag ) { 
-      assert(this->same_node[Xp]==this->HaloGatherDir(source,XpCompress,Xp,face_idx));
-      assert(this->same_node[Yp]==this->HaloGatherDir(source,YpCompress,Yp,face_idx));
-      assert(this->same_node[Zp]==this->HaloGatherDir(source,ZpCompress,Zp,face_idx));
-      assert(this->same_node[Tp]==this->HaloGatherDir(source,TpCompress,Tp,face_idx));
-      assert(this->same_node[Xm]==this->HaloGatherDir(source,XmCompress,Xm,face_idx));
-      assert(this->same_node[Ym]==this->HaloGatherDir(source,YmCompress,Ym,face_idx));
-      assert(this->same_node[Zm]==this->HaloGatherDir(source,ZmCompress,Zm,face_idx));
-      assert(this->same_node[Tm]==this->HaloGatherDir(source,TmCompress,Tm,face_idx));
+      vet_same_node(this->same_node[Xp],this->HaloGatherDir(source,XpCompress,Xp,face_idx));
+      vet_same_node(this->same_node[Yp],this->HaloGatherDir(source,YpCompress,Yp,face_idx));
+      vet_same_node(this->same_node[Zp],this->HaloGatherDir(source,ZpCompress,Zp,face_idx));
+      vet_same_node(this->same_node[Tp],this->HaloGatherDir(source,TpCompress,Tp,face_idx));
+      vet_same_node(this->same_node[Xm],this->HaloGatherDir(source,XmCompress,Xm,face_idx));
+      vet_same_node(this->same_node[Ym],this->HaloGatherDir(source,YmCompress,Ym,face_idx));
+      vet_same_node(this->same_node[Zm],this->HaloGatherDir(source,ZmCompress,Zm,face_idx));
+      vet_same_node(this->same_node[Tm],this->HaloGatherDir(source,TmCompress,Tm,face_idx));
     } else {
-      assert(this->same_node[Xp]==this->HaloGatherDir(source,XmCompress,Xp,face_idx));
-      assert(this->same_node[Yp]==this->HaloGatherDir(source,YmCompress,Yp,face_idx));
-      assert(this->same_node[Zp]==this->HaloGatherDir(source,ZmCompress,Zp,face_idx));
-      assert(this->same_node[Tp]==this->HaloGatherDir(source,TmCompress,Tp,face_idx));
-      assert(this->same_node[Xm]==this->HaloGatherDir(source,XpCompress,Xm,face_idx));
-      assert(this->same_node[Ym]==this->HaloGatherDir(source,YpCompress,Ym,face_idx));
-      assert(this->same_node[Zm]==this->HaloGatherDir(source,ZpCompress,Zm,face_idx));
-      assert(this->same_node[Tm]==this->HaloGatherDir(source,TpCompress,Tm,face_idx));
+      vet_same_node(this->same_node[Xp],this->HaloGatherDir(source,XmCompress,Xp,face_idx));
+      vet_same_node(this->same_node[Yp],this->HaloGatherDir(source,YmCompress,Yp,face_idx));
+      vet_same_node(this->same_node[Zp],this->HaloGatherDir(source,ZmCompress,Zp,face_idx));
+      vet_same_node(this->same_node[Tp],this->HaloGatherDir(source,TmCompress,Tp,face_idx));
+      vet_same_node(this->same_node[Xm],this->HaloGatherDir(source,XpCompress,Xm,face_idx));
+      vet_same_node(this->same_node[Ym],this->HaloGatherDir(source,YpCompress,Ym,face_idx));
+      vet_same_node(this->same_node[Zm],this->HaloGatherDir(source,ZpCompress,Zm,face_idx));
+      vet_same_node(this->same_node[Tm],this->HaloGatherDir(source,TpCompress,Tm,face_idx));
     }
     this->face_table_computed=1;
     assert(this->u_comm_offset==this->_unified_buffer_size);
diff --git a/Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h b/Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h
index b307fad4..fcf1f1f3 100644
--- a/Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h
+++ b/Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h
@@ -439,6 +439,17 @@ void WilsonKernels<Impl>::DhopDirKernel( StencilImpl &st, DoubledGaugeField &U,S
     int sF = ss*Ls;							\
     WilsonKernels<Impl>::A(st_v,U_v,buf,sF,sU,Ls,1,in_v,out_v);		\
   });
+#define ASM_CALL_SLICE(A)						\
+  auto grid = in.Grid() ;						\
+  int nt = grid->LocalDimensions()[4];					\
+  int nxyz = Nsite/nt ;							\
+  for(int t=0;t<nt;t++){						\
+  thread_for( sss, nxyz, {						\
+    int ss = t*nxyz+sss;						\
+    int sU = ss;							\
+    int sF = ss*Ls;							\
+    WilsonKernels<Impl>::A(st_v,U_v,buf,sF,sU,Ls,1,in_v,out_v);		\
+    });}
 
 template <class Impl>
 void WilsonKernels<Impl>::DhopKernel(int Opt,StencilImpl &st,  DoubledGaugeField &U, SiteHalfSpinor * buf,
diff --git a/Grid/stencil/Stencil.h b/Grid/stencil/Stencil.h
index 29aa876f..c8703b9f 100644
--- a/Grid/stencil/Stencil.h
+++ b/Grid/stencil/Stencil.h
@@ -358,7 +358,7 @@ public:
     return 0;
   }
 #else
-  // 
+  // fancy calculation for shm code
   inline int SameNode(int point) {
 
     int dimension    = this->_directions[point];
@@ -378,7 +378,7 @@ public:
 
     int nbr_proc;
     if (displacement>0) nbr_proc = 1;
-    else                 nbr_proc = pd-1;
+    else                nbr_proc = pd-1;
 
     // FIXME  this logic needs to be sorted for three link term
     //    assert( (displacement==1) || (displacement==-1));
diff --git a/Grid/threads/Accelerator.h b/Grid/threads/Accelerator.h
index 04ae885b..2dde1433 100644
--- a/Grid/threads/Accelerator.h
+++ b/Grid/threads/Accelerator.h
@@ -305,14 +305,14 @@ accelerator_inline int acceleratorSIMTlane(int Nsimd) {
      });	   			              \
     });
 
-#define accelerator_barrier(dummy) { printf(" theGridAccelerator::wait()\n");  theGridAccelerator->wait(); }
+#define accelerator_barrier(dummy) { theGridAccelerator->wait(); }
 
 inline void *acceleratorAllocShared(size_t bytes){ return malloc_shared(bytes,*theGridAccelerator);};
 inline void *acceleratorAllocDevice(size_t bytes){ return malloc_device(bytes,*theGridAccelerator);};
 inline void acceleratorFreeShared(void *ptr){free(ptr,*theGridAccelerator);};
 inline void acceleratorFreeDevice(void *ptr){free(ptr,*theGridAccelerator);};
 
-inline void acceleratorCopySynchronise(void) {  printf(" theCopyAccelerator::wait()\n"); theCopyAccelerator->wait(); }
+inline void acceleratorCopySynchronise(void) {  theCopyAccelerator->wait(); }
 inline void acceleratorCopyDeviceToDeviceAsynch(void *from,void *to,size_t bytes)  {  theCopyAccelerator->memcpy(to,from,bytes);}
 inline void acceleratorCopyToDevice(void *from,void *to,size_t bytes)  { theCopyAccelerator->memcpy(to,from,bytes); theCopyAccelerator->wait();}
 inline void acceleratorCopyFromDevice(void *from,void *to,size_t bytes){ theCopyAccelerator->memcpy(to,from,bytes); theCopyAccelerator->wait();}
diff --git a/systems/PVC/benchmarks/run-1tile.sh b/systems/PVC/benchmarks/run-1tile.sh
index 923afd84..3c594ab6 100644
--- a/systems/PVC/benchmarks/run-1tile.sh
+++ b/systems/PVC/benchmarks/run-1tile.sh
@@ -21,7 +21,7 @@ export I_MPI_OFFLOAD_CELL=tile
 export EnableImplicitScaling=0
 export EnableWalkerPartition=0
 export ZE_AFFINITY_MASK=0.0
-mpiexec -launcher ssh -n 1 -host localhost  ./Benchmark_dwf_fp32 --mpi 1.1.1.1 --grid 32.32.32.32 --accelerator-threads $NT --comms-sequential --shm-mpi 1 --cacheblocking 8.8.8.8
+mpiexec -launcher ssh -n 1 -host localhost  ./Benchmark_dwf_fp32 --mpi 1.1.1.1 --grid 32.32.32.32 --accelerator-threads $NT --comms-sequential --shm-mpi 0
 
 export ZE_AFFINITY_MASK=0
 export I_MPI_OFFLOAD_CELL=device
diff --git a/systems/PVC/benchmarks/run-2tile-mpi.sh b/systems/PVC/benchmarks/run-2tile-mpi.sh
index 9db0b66b..fa56d5ec 100755
--- a/systems/PVC/benchmarks/run-2tile-mpi.sh
+++ b/systems/PVC/benchmarks/run-2tile-mpi.sh
@@ -20,7 +20,7 @@ export I_MPI_OFFLOAD_CELL=tile
 export EnableImplicitScaling=0
 export EnableWalkerPartition=0
 
-mpiexec -launcher ssh -n 1 -host localhost  ./wrap.sh ./Benchmark_dwf_fp32 --mpi 1.1.1.1 --grid 32.32.32.32 --accelerator-threads $NT --comms-sequential --shm-mpi 1 > 1tile.log
+#mpiexec -launcher ssh -n 1 -host localhost  ./wrap.sh ./Benchmark_dwf_fp32 --mpi 1.1.1.1 --grid 32.32.32.32 --accelerator-threads $NT --comms-sequential --shm-mpi 0 > 1tile.log
 
-mpiexec -launcher ssh -n 2 -host localhost  ./wrap.sh ./Benchmark_dwf_fp32 --mpi 2.1.1.1 --grid 64.32.32.32 --accelerator-threads $NT --comms-sequential --shm-mpi 1 > 2tile.log
+mpiexec -launcher ssh -n 2 -host localhost  ./wrap.sh ./Benchmark_dwf_fp32 --mpi 2.1.1.1 --grid 64.32.32.32 --accelerator-threads $NT --comms-sequential --shm-mpi 0
 
diff --git a/systems/PVC/benchmarks/wrap.sh b/systems/PVC/benchmarks/wrap.sh
index a352fff9..0e48625b 100755
--- a/systems/PVC/benchmarks/wrap.sh
+++ b/systems/PVC/benchmarks/wrap.sh
@@ -5,10 +5,10 @@ export ZE_AFFINITY_MASK=0.$MPI_LOCALRANKID
 echo Ranke $MPI_LOCALRANKID ZE_AFFINITY_MASK is $ZE_AFFINITY_MASK
 
 
-if [ $MPI_LOCALRANKID = "0" ] 
-then
+#if [ $MPI_LOCALRANKID = "0" ] 
+#then
 #  ~psteinbr/build_pti/ze_tracer -c $@
-  onetrace --chrome-kernel-timeline $@
-else
+#  onetrace --chrome-kernel-timeline $@
+#else
   $@
-fi
+#fi
diff --git a/systems/PVC/config-command b/systems/PVC/config-command
index dc6b222c..7549f2b4 100644
--- a/systems/PVC/config-command
+++ b/systems/PVC/config-command
@@ -1,4 +1,4 @@
-INSTALL=/nfs/site/home/azusayax/install
+INSTALL=/nfs/site/home/paboylx/prereqs/
 ../../configure \
 	--enable-simd=GPU \
 	--enable-gen-simd-width=64 \
@@ -8,8 +8,8 @@ INSTALL=/nfs/site/home/azusayax/install
 	--disable-fermion-reps \
 	--enable-shm=nvlink \
 	--enable-accelerator=sycl \
-	--enable-unified=yes \
+	--enable-unified=no \
 	CXX=mpicxx \
 	LDFLAGS="-fsycl-device-code-split=per_kernel -fsycl-device-lib=all -lze_loader -L$INSTALL/lib" \
-	CXXFLAGS="-cxx=icpx -fsycl-unnamed-lambda -fsycl -I$INSTALL/include -Wno-tautological-constant-compare"
+	CXXFLAGS="-cxx=icpx -fsycl-unnamed-lambda -fsycl -Wno-tautological-constant-compare -I$INSTALL/include"
 
diff --git a/systems/PVC/setup.sh b/systems/PVC/setup.sh
index 2a6f920b..9b515a62 100644
--- a/systems/PVC/setup.sh
+++ b/systems/PVC/setup.sh
@@ -1,5 +1,6 @@
 export https_proxy=http://proxy-chain.intel.com:911
-export LD_LIBRARY_PATH=/nfs/site/home/azusayax/install/lib:$LD_LIBRARY_PATH
+#export LD_LIBRARY_PATH=/nfs/site/home/azusayax/install/lib:$LD_LIBRARY_PATH
+export LD_LIBRARY_PATH=$HOME/prereqs/lib/:$LD_LIBRARY_PATH
 
 module load intel-release
 source /opt/intel/oneapi/PVC_setup.sh

From 0f2b7864362b92737fb011244c4f74d7cf4e3c62 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 21 Mar 2023 15:36:11 -0400
Subject: [PATCH 134/146] Vector -> vector

---
 Grid/stencil/SimpleCompressor.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/Grid/stencil/SimpleCompressor.h b/Grid/stencil/SimpleCompressor.h
index ccbfdb29..dabd70a6 100644
--- a/Grid/stencil/SimpleCompressor.h
+++ b/Grid/stencil/SimpleCompressor.h
@@ -36,7 +36,7 @@ public:
   }
   template<class vobj,class cobj,class compressor>
   static void Gather_plane_exchange(commVector<std::pair<int,int> >& table,const Lattice<vobj> &rhs,
-				    Vector<cobj *> pointers,int dimension,int plane,int cbmask,
+				    std::vector<cobj *> pointers,int dimension,int plane,int cbmask,
 				    compressor &compress,int type,int partial)
   {
     assert( (table.size()&0x1)==0);

From 88e218e8ee2bc1840aebcb5b7fcb9621add1f2ab Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 21 Mar 2023 15:37:58 -0400
Subject: [PATCH 135/146] Stencil updates

---
 Grid/stencil/Stencil.cc | 21 +++++++++++++++++++++
 Grid/stencil/Stencil.h  | 14 ++++++++++++++
 2 files changed, 35 insertions(+)

diff --git a/Grid/stencil/Stencil.cc b/Grid/stencil/Stencil.cc
index c1b33baa..27dc75ed 100644
--- a/Grid/stencil/Stencil.cc
+++ b/Grid/stencil/Stencil.cc
@@ -29,6 +29,27 @@
 
 NAMESPACE_BEGIN(Grid);
 
+uint64_t DslashFullCount;
+uint64_t DslashPartialCount;
+uint64_t DslashDirichletCount;
+
+void DslashResetCounts(void)
+{
+  DslashFullCount=0;
+  DslashPartialCount=0;
+  DslashDirichletCount=0;
+}
+void DslashGetCounts(uint64_t &dirichlet,uint64_t &partial,uint64_t &full)
+{
+  dirichlet = DslashDirichletCount;
+  partial   = DslashPartialCount;
+  full      = DslashFullCount;
+}
+void DslashLogFull(void)     { DslashFullCount++;}
+void DslashLogPartial(void)  { DslashPartialCount++;}
+void DslashLogDirichlet(void){ DslashDirichletCount++;}
+
+
 void Gather_plane_table_compute (GridBase *grid,int dimension,int plane,int cbmask,
 				 int off,std::vector<std::pair<int,int> > & table)
 {
diff --git a/Grid/stencil/Stencil.h b/Grid/stencil/Stencil.h
index 93a3cb68..356cf73f 100644
--- a/Grid/stencil/Stencil.h
+++ b/Grid/stencil/Stencil.h
@@ -120,6 +120,12 @@ void Gather_plane_exchange_table(commVector<std::pair<int,int> >& table,
 }
 */
 
+void DslashResetCounts(void);
+void DslashGetCounts(uint64_t &dirichlet,uint64_t &partial,uint64_t &full);
+void DslashLogFull(void);
+void DslashLogPartial(void);
+void DslashLogDirichlet(void);
+
 struct StencilEntry {
 #ifdef GRID_CUDA
   uint64_t _byte_offset;       // 8 bytes
@@ -312,6 +318,7 @@ public:
 
   int face_table_computed;
   int partialDirichlet;
+  int fullDirichlet;
   std::vector<commVector<std::pair<int,int> > > face_table ;
   Vector<int> surface_list;
 
@@ -408,6 +415,9 @@ public:
   void CommunicateComplete(std::vector<std::vector<CommsRequest_t> > &reqs)
   {
     _grid->StencilSendToRecvFromComplete(MpiReqs,0);
+    if   ( this->partialDirichlet ) DslashLogPartial();
+    else if ( this->fullDirichlet ) DslashLogDirichlet();
+    else DslashLogFull();
   }
   ////////////////////////////////////////////////////////////////////////
   // Blocking send and receive. Either sequential or parallel.
@@ -736,6 +746,10 @@ public:
     if ( p.dirichlet.size() ==0 ) p.dirichlet.resize(grid->Nd(),0);
     partialDirichlet = p.partialDirichlet;
     DirichletBlock(p.dirichlet); // comms send/recv set up
+    fullDirichlet=0;
+    for(int d=0;d<p.dirichlet.size();d++){
+      if (p.dirichlet[d]) fullDirichlet=1;
+    }
 
     _unified_buffer_size=0;
     surface_list.resize(0);

From c5bdf61215f7b399cf82710f768ccd1a024e0c1f Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 21 Mar 2023 15:38:39 -0400
Subject: [PATCH 136/146] AUdit fix

---
 Grid/allocator/MemoryManagerShared.cc | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/Grid/allocator/MemoryManagerShared.cc b/Grid/allocator/MemoryManagerShared.cc
index 2434ad47..e291ef89 100644
--- a/Grid/allocator/MemoryManagerShared.cc
+++ b/Grid/allocator/MemoryManagerShared.cc
@@ -13,7 +13,7 @@ uint64_t  MemoryManager::DeviceToHostBytes;
 uint64_t  MemoryManager::HostToDeviceXfer;
 uint64_t  MemoryManager::DeviceToHostXfer;
 
-void  MemoryManager::Audit(void){};
+void  MemoryManager::Audit(std::string s){};
 void  MemoryManager::ViewClose(void* AccPtr,ViewMode mode){};
 void *MemoryManager::ViewOpen(void* CpuPtr,size_t bytes,ViewMode mode,ViewAdvise hint){ return CpuPtr; };
 int   MemoryManager::isOpen   (void* CpuPtr) { return 0;}

From 4135f2dcd1875816ade5a9b0194a32e07576afcb Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 21 Mar 2023 15:41:41 -0400
Subject: [PATCH 137/146] Compressor

---
 Grid/qcd/action/fermion/WilsonCompressor.h | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/Grid/qcd/action/fermion/WilsonCompressor.h b/Grid/qcd/action/fermion/WilsonCompressor.h
index aaf612fc..a1234e3b 100644
--- a/Grid/qcd/action/fermion/WilsonCompressor.h
+++ b/Grid/qcd/action/fermion/WilsonCompressor.h
@@ -36,7 +36,7 @@ NAMESPACE_BEGIN(Grid);
 // Wilson compressor will need FaceGather policies for:
 // Periodic, Dirichlet, and partial Dirichlet for DWF
 ///////////////////////////////////////////////////////////////
-const int dwf_compressor_depth=1;
+const int dwf_compressor_depth=2;
 #define DWF_COMPRESS
 class FaceGatherPartialDWF
 {
@@ -110,7 +110,7 @@ public:
   ////////////////////////////////////////////////////////////////////////////////////////////
   template<class vobj,class cobj,class compressor>
   static void Gather_plane_exchange(commVector<std::pair<int,int> >& table,const Lattice<vobj> &rhs,
-				    Vector<cobj *> pointers,int dimension,int plane,int cbmask,
+				    std::vector<cobj *> pointers,int dimension,int plane,int cbmask,
 				    compressor &compress,int type,int partial)
   {
     GridBase *Grid = rhs.Grid();
@@ -209,7 +209,7 @@ public:
   }
   template<class vobj,class cobj,class compressor>
   static void Gather_plane_exchange(commVector<std::pair<int,int> >& table,const Lattice<vobj> &rhs,
-				    Vector<cobj *> pointers,int dimension,int plane,int cbmask,
+				    std::vector<cobj *> pointers,int dimension,int plane,int cbmask,
 				    compressor &compress,int type,int partial)
   {
     //    std::cout << " face gather exch DWF partial "<<partial <<std::endl;

From 68428fceab776303171b19917ab40e2582b9c57d Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 21 Mar 2023 15:58:49 -0400
Subject: [PATCH 138/146] Integrator update

---
 Grid/qcd/hmc/integrators/Integrator.h | 9 +++++++++
 1 file changed, 9 insertions(+)

diff --git a/Grid/qcd/hmc/integrators/Integrator.h b/Grid/qcd/hmc/integrators/Integrator.h
index 85a135ad..38d6e561 100644
--- a/Grid/qcd/hmc/integrators/Integrator.h
+++ b/Grid/qcd/hmc/integrators/Integrator.h
@@ -284,6 +284,15 @@ public:
 		  << as[level].actions.at(actionID)->deriv_us*1.0e-6<<" s"<< std::endl;
       }
     }
+    std::cout << GridLogMessage << "--------------------------- "<<std::endl;
+    std::cout << GridLogMessage << " Dslash counts "<<std::endl;
+    std::cout << GridLogMessage << "------------------------- "<<std::endl;
+    uint64_t full, partial, dirichlet;
+    DslashGetCounts(dirichlet,partial,full);
+    std::cout << GridLogMessage << " Full BCs               : "<<full<<std::endl;
+    std::cout << GridLogMessage << " Partial dirichlet BCs  : "<<partial<<std::endl;
+    std::cout << GridLogMessage << " Dirichlet BCs          : "<<dirichlet<<std::endl;
+
     std::cout << GridLogMessage << "--------------------------- "<<std::endl;
     std::cout << GridLogMessage << " Force average size "<<std::endl;
     std::cout << GridLogMessage << "------------------------- "<<std::endl;

From f17f879206197889cbbb456142fa42645a9ab113 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 21 Mar 2023 15:59:29 -0400
Subject: [PATCH 139/146] Test update

---
 tests/forces/Test_double_ratio.cc | 33 ++++++++++++++++++++++++++++++-
 1 file changed, 32 insertions(+), 1 deletion(-)

diff --git a/tests/forces/Test_double_ratio.cc b/tests/forces/Test_double_ratio.cc
index 0a350692..dc6713f9 100644
--- a/tests/forces/Test_double_ratio.cc
+++ b/tests/forces/Test_double_ratio.cc
@@ -476,6 +476,20 @@ int main (int argc, char ** argv)
   //  ForceTest<GimplTypesR>(BdyNf2eo,U,DDHMCFilter);
 
   //////////////////// One flavour boundary det  ////////////////////
+  RationalActionParams OFRp; // Up/down
+  OFRp.lo       = 6.0e-5;
+  OFRp.hi       = 90.0;
+  OFRp.inv_pow  = 2;
+  OFRp.MaxIter  = SP_iters; // get most shifts by 2000, stop sharing space
+  OFRp.action_tolerance= 1.0e-8;
+  OFRp.action_degree   = 18;
+  OFRp.md_tolerance= 1.0e-5;
+  OFRp.md_degree   = 14;
+  //  OFRp.degree   = 20; converges
+  //  OFRp.degree   = 16;
+  OFRp.precision= 80;
+  OFRp.BoundsCheckFreq=0;
+  /*
   OneFlavourRationalParams OFRp; // Up/down
   OFRp.lo       = 4.0e-5;
   OFRp.hi       = 90.0;
@@ -485,6 +499,23 @@ int main (int argc, char ** argv)
   OFRp.degree   = 18;
   OFRp.precision= 80;
   OFRp.BoundsCheckFreq=0;
+  */
+  std::vector<RealD> ActionTolByPole({
+      1.0e-7,1.0e-8,1.0e-8,1.0e-8,
+      1.0e-8,1.0e-8,1.0e-8,1.0e-8,
+      1.0e-8,1.0e-8,1.0e-8,1.0e-8,
+      1.0e-8,1.0e-8,1.0e-8,1.0e-8,
+      1.0e-8,1.0e-8
+    });
+  std::vector<RealD> MDTolByPole({
+      1.6e-5,5.0e-6,1.0e-6,3.0e-7, // soften convergence more more
+      //      1.0e-6,3.0e-7,1.0e-7,1.0e-7,
+      //      3.0e-6,1.0e-6,1.0e-7,1.0e-7, // soften convergence
+      1.0e-8,1.0e-8,1.0e-8,1.0e-8,
+      1.0e-8,1.0e-8,1.0e-8,1.0e-8,
+      1.0e-8,1.0e-8
+    });
+  /*
   std::vector<RealD> ActionTolByPole({
       1.0e-8,1.0e-8,1.0e-8,1.0e-8,
       1.0e-8,1.0e-8,1.0e-8,1.0e-8,
@@ -499,9 +530,9 @@ int main (int argc, char ** argv)
       //      1.0e-8,1.0e-8,1.0e-8,1.0e-8,
       1.0e-8,1.0e-8,1.0e-8,1.0e-8,
       1.0e-8,1.0e-8,1.0e-8,1.0e-8,
-      1.0e-8,1.0e-8,1.0e-8,1.0e-8,
       1.0e-8,1.0e-8
     });
+  */
   OneFlavourEvenOddRatioRationalPseudoFermionAction<FermionImplPolicy> BdySqrt(DdwfDirichlet,DdwfPeriodic,OFRp);
   BdySqrt.SetTolerances(ActionTolByPole,MDTolByPole);
   ForceTest<GimplTypesR>(BdySqrt,U,DDHMCFilter);

From 8b43be39c0fe73864e821ae0fa93863d97d5044e Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 21 Mar 2023 16:00:52 -0400
Subject: [PATCH 140/146] Config command

---
 systems/Crusher/config-command | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/systems/Crusher/config-command b/systems/Crusher/config-command
index 3965767f..d310ff55 100644
--- a/systems/Crusher/config-command
+++ b/systems/Crusher/config-command
@@ -1,12 +1,13 @@
 CLIME=`spack find --paths c-lime@2-3-9 | grep c-lime| cut -c 15-`
 ../../configure --enable-comms=mpi-auto \
 --with-lime=$CLIME \
---enable-unified=no \
+--enable-unified=yes \
 --enable-shm=nvlink \
 --enable-tracing=timer \
 --enable-accelerator=hip \
 --enable-gen-simd-width=64 \
 --enable-simd=GPU \
+--disable-accelerator-cshift \
 --with-gmp=$OLCF_GMP_ROOT \
 --with-fftw=$FFTW_DIR/.. \
 --with-mpfr=/opt/cray/pe/gcc/mpfr/3.1.4/ \

From c6621806ca73b9384b3bcae6cc8ef47157262be7 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 21 Mar 2023 17:27:09 -0400
Subject: [PATCH 141/146] Compiling on laptop and running

---
 Grid/qcd/action/fermion/WilsonCompressor.h | 4 ++--
 Grid/stencil/SimpleCompressor.h            | 2 +-
 tests/Test_dwf_mixedcg_prec.cc             | 2 +-
 3 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/Grid/qcd/action/fermion/WilsonCompressor.h b/Grid/qcd/action/fermion/WilsonCompressor.h
index 6e46e0b5..b2c07d18 100644
--- a/Grid/qcd/action/fermion/WilsonCompressor.h
+++ b/Grid/qcd/action/fermion/WilsonCompressor.h
@@ -110,7 +110,7 @@ public:
   ////////////////////////////////////////////////////////////////////////////////////////////
   template<class vobj,class cobj,class compressor>
   static void Gather_plane_exchange(commVector<std::pair<int,int> >& table,const Lattice<vobj> &rhs,
-				    Vector<cobj *> pointers,int dimension,int plane,int cbmask,
+				    std::vector<cobj *> pointers,int dimension,int plane,int cbmask,
 				    compressor &compress,int type,int partial)
   {
     GridBase *Grid = rhs.Grid();
@@ -209,7 +209,7 @@ public:
   }
   template<class vobj,class cobj,class compressor>
   static void Gather_plane_exchange(commVector<std::pair<int,int> >& table,const Lattice<vobj> &rhs,
-				    Vector<cobj *> pointers,int dimension,int plane,int cbmask,
+				    std::vector<cobj *> pointers,int dimension,int plane,int cbmask,
 				    compressor &compress,int type,int partial)
   {
     //    std::cout << " face gather exch DWF partial "<<partial <<std::endl;
diff --git a/Grid/stencil/SimpleCompressor.h b/Grid/stencil/SimpleCompressor.h
index ccbfdb29..dabd70a6 100644
--- a/Grid/stencil/SimpleCompressor.h
+++ b/Grid/stencil/SimpleCompressor.h
@@ -36,7 +36,7 @@ public:
   }
   template<class vobj,class cobj,class compressor>
   static void Gather_plane_exchange(commVector<std::pair<int,int> >& table,const Lattice<vobj> &rhs,
-				    Vector<cobj *> pointers,int dimension,int plane,int cbmask,
+				    std::vector<cobj *> pointers,int dimension,int plane,int cbmask,
 				    compressor &compress,int type,int partial)
   {
     assert( (table.size()&0x1)==0);
diff --git a/tests/Test_dwf_mixedcg_prec.cc b/tests/Test_dwf_mixedcg_prec.cc
index 1e6da515..cbc573d1 100644
--- a/tests/Test_dwf_mixedcg_prec.cc
+++ b/tests/Test_dwf_mixedcg_prec.cc
@@ -101,7 +101,7 @@ int main (int argc, char ** argv)
   std:: cout << " MdagM site flops = "<< 4*MdagMsiteflops<<std::endl;
   std:: cout << " CG    site flops = "<< CGsiteflops <<std::endl;
   int iters;
-  for(int i=0;i<200;i++){
+  for(int i=0;i<10;i++){
     result_o = Zero();
     t1=usecond();
     mCG(src_o,result_o);

From d0bb033ea2990065ef0b1afdf54dc3d5014b6110 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Wed, 22 Mar 2023 19:07:32 -0400
Subject: [PATCH 142/146] Device resident GPU block buffer instead of UVM as
 hit likely UVM bug. Code worked on CUDA 11.4 but fails on later drivers
 (certainly 530.30.02, but need to find the perlmutter driver version).

---
 Grid/lattice/Lattice_reduction_gpu.h | 23 +++++++++++++++++++----
 1 file changed, 19 insertions(+), 4 deletions(-)

diff --git a/Grid/lattice/Lattice_reduction_gpu.h b/Grid/lattice/Lattice_reduction_gpu.h
index 5f5c6cc0..4bdcce0b 100644
--- a/Grid/lattice/Lattice_reduction_gpu.h
+++ b/Grid/lattice/Lattice_reduction_gpu.h
@@ -211,13 +211,28 @@ inline typename vobj::scalar_objectD sumD_gpu_small(const vobj *lat, Integer osi
   assert(ok);
 
   Integer smemSize = numThreads * sizeof(sobj);
-
-  Vector<sobj> buffer(numBlocks);
+  // UVM seems to be buggy under later CUDA drivers
+  // This fails on A100 and driver 5.30.02 / CUDA 12.1
+  // Fails with multiple NVCC versions back to 11.4,
+  // which worked with earlier drivers.
+  // Not sure which driver had first fail and this bears checking
+  // Is awkward as must install multiple driver versions
+#undef UVM_BLOCK_BUFFER  
+#ifndef UVM_BLOCK_BUFFER  
+  commVector<sobj> buffer(numBlocks);
   sobj *buffer_v = &buffer[0];
-  
+  sobj result;
   reduceKernel<<< numBlocks, numThreads, smemSize >>>(lat, buffer_v, size);
   accelerator_barrier();
-  auto result = buffer_v[0];
+  acceleratorCopyFromDevice(buffer_v,&result,sizeof(result));
+#else
+  Vector<sobj> buffer(numBlocks);
+  sobj *buffer_v = &buffer[0];
+  sobj result;
+  reduceKernel<<< numBlocks, numThreads, smemSize >>>(lat, buffer_v, size);
+  accelerator_barrier();
+  result = *buffer_v;
+#endif
   return result;
 }
 

From 52384e34cf76df9129640869f8e7c75c055a0e31 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Wed, 22 Mar 2023 19:40:32 -0400
Subject: [PATCH 143/146] Discard on construct

---
 Grid/lattice/Lattice_base.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/Grid/lattice/Lattice_base.h b/Grid/lattice/Lattice_base.h
index dc860d8e..b0b759b5 100644
--- a/Grid/lattice/Lattice_base.h
+++ b/Grid/lattice/Lattice_base.h
@@ -248,7 +248,7 @@ public:
   ///////////////////////////////////////////
   // user defined constructor
   ///////////////////////////////////////////
-  Lattice(GridBase *grid,ViewMode mode=AcceleratorWrite) { 
+  Lattice(GridBase *grid,ViewMode mode=AcceleratorWriteDiscard) { 
     this->_grid = grid;
     resize(this->_grid->oSites());
     assert((((uint64_t)&this->_odata[0])&0xF) ==0);

From 90130e25e9309b1c89b6a47a60f895ca222e240c Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Thu, 23 Mar 2023 10:27:02 -0400
Subject: [PATCH 144/146] TODO list

---
 TODO | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/TODO b/TODO
index 380af2b7..a6d0f2ac 100644
--- a/TODO
+++ b/TODO
@@ -1,3 +1,10 @@
+- - Slice sum optimisation & A2A - atomic addition
+- - Also faster non-atomic reduction
+- - Remaining PRs
+- - DDHMC
+
+=================
+=================
 Lattice_basis.h -- > HIP and SYCL GPU code
 
 
@@ -8,6 +15,7 @@ DDHMC
 -- Multishift Mixed Precision - DONE
 -- Pole dependent residual  - DONE
 
+
 =======
 -- comms threads issue??
 -- Part done: Staggered kernel performance on GPU

From 1a7114d4b934088ec9c049e99bc7ee8d0daab6fa Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Thu, 23 Mar 2023 15:38:35 -0400
Subject: [PATCH 145/146] Temporary algorithm while sorting out mixed prec

---
 .../ConjugateGradientMultiShiftCleanup.h      | 373 ++++++++++++++++++
 1 file changed, 373 insertions(+)
 create mode 100644 Grid/algorithms/iterative/ConjugateGradientMultiShiftCleanup.h

diff --git a/Grid/algorithms/iterative/ConjugateGradientMultiShiftCleanup.h b/Grid/algorithms/iterative/ConjugateGradientMultiShiftCleanup.h
new file mode 100644
index 00000000..17714f09
--- /dev/null
+++ b/Grid/algorithms/iterative/ConjugateGradientMultiShiftCleanup.h
@@ -0,0 +1,373 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/algorithms/iterative/ConjugateGradientMultiShift.h
+
+    Copyright (C) 2015
+
+Author: Azusa Yamaguchi <ayamaguc@staffmail.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Christopher Kelly <ckelly@bnl.gov>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once
+
+NAMESPACE_BEGIN(Grid);
+
+//CK 2020: A variant of the multi-shift conjugate gradient with the matrix multiplication in single precision. 
+//The residual is stored in single precision, but the search directions and solution are stored in double precision. 
+//Every update_freq iterations the residual is corrected in double precision. 
+//For safety the a final regular CG is applied to clean up if necessary
+
+//PB Pure single, then double fixup
+
+template<class FieldD, class FieldF,
+	 typename std::enable_if< getPrecision<FieldD>::value == 2, int>::type = 0,
+	 typename std::enable_if< getPrecision<FieldF>::value == 1, int>::type = 0> 
+class ConjugateGradientMultiShiftMixedPrecCleanup : public OperatorMultiFunction<FieldD>,
+					     public OperatorFunction<FieldD>
+{
+public:                                                
+
+  using OperatorFunction<FieldD>::operator();
+
+  RealD   Tolerance;
+  Integer MaxIterationsMshift;
+  Integer MaxIterations;
+  Integer IterationsToComplete; //Number of iterations the CG took to finish. Filled in upon completion
+  std::vector<int> IterationsToCompleteShift;  // Iterations for this shift
+  int verbose;
+  MultiShiftFunction shifts;
+  std::vector<RealD> TrueResidualShift;
+
+  int ReliableUpdateFreq; //number of iterations between reliable updates
+
+  GridBase* SinglePrecGrid; //Grid for single-precision fields
+  LinearOperatorBase<FieldF> &Linop_f; //single precision
+
+  ConjugateGradientMultiShiftMixedPrecCleanup(Integer maxit, const MultiShiftFunction &_shifts,
+				       GridBase* _SinglePrecGrid, LinearOperatorBase<FieldF> &_Linop_f,
+				       int _ReliableUpdateFreq) : 
+    MaxIterationsMshift(maxit),  shifts(_shifts), SinglePrecGrid(_SinglePrecGrid), Linop_f(_Linop_f), ReliableUpdateFreq(_ReliableUpdateFreq),
+    MaxIterations(20000)
+  { 
+    verbose=1;
+    IterationsToCompleteShift.resize(_shifts.order);
+    TrueResidualShift.resize(_shifts.order);
+  }
+
+  void operator() (LinearOperatorBase<FieldD> &Linop, const FieldD &src, FieldD &psi)
+  {
+    GridBase *grid = src.Grid();
+    int nshift = shifts.order;
+    std::vector<FieldD> results(nshift,grid);
+    (*this)(Linop,src,results,psi);
+  }
+  void operator() (LinearOperatorBase<FieldD> &Linop, const FieldD &src, std::vector<FieldD> &results, FieldD &psi)
+  {
+    int nshift = shifts.order;
+
+    (*this)(Linop,src,results);
+  
+    psi = shifts.norm*src;
+    for(int i=0;i<nshift;i++){
+      psi = psi + shifts.residues[i]*results[i];
+    }
+
+    return;
+  }
+
+  void operator() (LinearOperatorBase<FieldD> &Linop_d, const FieldD &src_d, std::vector<FieldD> &psi_d)
+  { 
+    GRID_TRACE("ConjugateGradientMultiShiftMixedPrecCleanup");
+    GridBase *DoublePrecGrid = src_d.Grid();
+
+    ////////////////////////////////////////////////////////////////////////
+    // Convenience references to the info stored in "MultiShiftFunction"
+    ////////////////////////////////////////////////////////////////////////
+    int nshift = shifts.order;
+
+    std::vector<RealD> &mass(shifts.poles); // Make references to array in "shifts"
+    std::vector<RealD> &mresidual(shifts.tolerances);
+    std::vector<RealD> alpha(nshift,1.0);
+
+    //Double precision search directions
+    FieldD p_d(DoublePrecGrid);
+    std::vector<FieldF> ps_f (nshift, SinglePrecGrid);// Search directions (single precision)
+    std::vector<FieldF> psi_f(nshift, SinglePrecGrid);// solutions (single precision)
+
+    FieldD tmp_d(DoublePrecGrid);
+    FieldD r_d(DoublePrecGrid);
+    FieldF r_f(SinglePrecGrid);
+    FieldD mmp_d(DoublePrecGrid);
+
+    assert(psi_d.size()==nshift);
+    assert(mass.size()==nshift);
+    assert(mresidual.size()==nshift);
+  
+    // dynamic sized arrays on stack; 2d is a pain with vector
+    RealD  bs[nshift];
+    RealD  rsq[nshift];
+    RealD  rsqf[nshift];
+    RealD  z[nshift][2];
+    int     converged[nshift];
+  
+    const int       primary =0;
+  
+    //Primary shift fields CG iteration
+    RealD a,b,c,d;
+    RealD cp,bp,qq; //prev
+  
+    // Matrix mult fields
+    FieldF p_f(SinglePrecGrid);
+    FieldF mmp_f(SinglePrecGrid);
+
+    // Check lightest mass
+    for(int s=0;s<nshift;s++){
+      assert( mass[s]>= mass[primary] );
+      converged[s]=0;
+    }
+  
+    // Wire guess to zero
+    // Residuals "r" are src
+    // First search direction "p" is also src
+    cp = norm2(src_d);
+
+    // Handle trivial case of zero src.
+    if( cp == 0. ){
+      for(int s=0;s<nshift;s++){
+	psi_d[s] = Zero();
+	psi_f[s] = Zero();
+	IterationsToCompleteShift[s] = 1;
+	TrueResidualShift[s] = 0.;
+      }
+      return;
+    }
+
+    for(int s=0;s<nshift;s++){
+      rsq[s] = cp * mresidual[s] * mresidual[s];
+      rsqf[s] =rsq[s];
+      std::cout<<GridLogMessage<<"ConjugateGradientMultiShiftMixedPrecCleanup: shift "<< s <<" target resid "<<rsq[s]<<std::endl;
+      //      ps_d[s] = src_d;
+      precisionChangeFast(ps_f[s],src_d);
+    }
+    // r and p for primary
+    p_d = src_d; //primary copy --- make this a reference to ps_d to save axpys
+    r_d = p_d;
+    
+    //MdagM+m[0]
+    precisionChangeFast(p_f,p_d);
+    Linop_f.HermOpAndNorm(p_f,mmp_f,d,qq); // mmp = MdagM p        d=real(dot(p, mmp)),  qq=norm2(mmp)
+    precisionChangeFast(tmp_d,mmp_f);
+    Linop_d.HermOpAndNorm(p_d,mmp_d,d,qq); // mmp = MdagM p        d=real(dot(p, mmp)),  qq=norm2(mmp)
+    tmp_d = tmp_d - mmp_d;
+    std::cout << " Testing operators match "<<norm2(mmp_d)<<" f "<<norm2(mmp_f)<<" diff "<< norm2(tmp_d)<<std::endl;
+    //    assert(norm2(tmp_d)< 1.0e-4);
+
+    axpy(mmp_d,mass[0],p_d,mmp_d);
+    RealD rn = norm2(p_d);
+    d += rn*mass[0];
+
+    b = -cp /d;
+  
+    // Set up the various shift variables
+    int       iz=0;
+    z[0][1-iz] = 1.0;
+    z[0][iz]   = 1.0;
+    bs[0]      = b;
+    for(int s=1;s<nshift;s++){
+      z[s][1-iz] = 1.0;
+      z[s][iz]   = 1.0/( 1.0 - b*(mass[s]-mass[0]));
+      bs[s]      = b*z[s][iz]; 
+    }
+  
+    // r += b[0] A.p[0]
+    // c= norm(r)
+    c=axpy_norm(r_d,b,mmp_d,r_d);
+  
+    for(int s=0;s<nshift;s++) {
+      axpby(psi_d[s],0.,-bs[s]*alpha[s],src_d,src_d);
+      precisionChangeFast(psi_f[s],psi_d[s]);
+    }
+  
+    ///////////////////////////////////////
+    // Timers
+    ///////////////////////////////////////
+    GridStopWatch AXPYTimer, ShiftTimer, QRTimer, MatrixTimer, SolverTimer, PrecChangeTimer, CleanupTimer;
+
+    SolverTimer.Start();
+  
+    // Iteration loop
+    int k;
+  
+    for (k=1;k<=MaxIterationsMshift;k++){    
+
+      a = c /cp;
+      AXPYTimer.Start();
+      axpy(p_d,a,p_d,r_d); 
+      AXPYTimer.Stop();
+
+      PrecChangeTimer.Start();
+      precisionChangeFast(r_f, r_d);
+      PrecChangeTimer.Stop();
+
+      AXPYTimer.Start();
+      for(int s=0;s<nshift;s++){
+	if ( ! converged[s] ) { 
+	  if (s==0){
+	    axpy(ps_f[s],a,ps_f[s],r_f);
+	  } else{
+	    RealD as =a *z[s][iz]*bs[s] /(z[s][1-iz]*b);
+	    axpby(ps_f[s],z[s][iz],as,r_f,ps_f[s]);
+	  }
+	}
+      }
+      AXPYTimer.Stop();
+
+      cp=c;
+      PrecChangeTimer.Start();
+      precisionChangeFast(p_f, p_d); //get back single prec search direction for linop
+      PrecChangeTimer.Stop();
+      MatrixTimer.Start();  
+      Linop_f.HermOp(p_f,mmp_f);
+      MatrixTimer.Stop();  
+      PrecChangeTimer.Start();
+      precisionChangeFast(mmp_d, mmp_f); // From Float to Double
+      PrecChangeTimer.Stop();
+
+      d=real(innerProduct(p_d,mmp_d));    
+      axpy(mmp_d,mass[0],p_d,mmp_d);
+      RealD rn = norm2(p_d);
+      d += rn*mass[0];
+    
+      bp=b;
+      b=-cp/d;
+
+      // Toggle the recurrence history
+      bs[0] = b;
+      iz = 1-iz;
+      ShiftTimer.Start();
+      for(int s=1;s<nshift;s++){
+	if((!converged[s])){
+	  RealD z0 = z[s][1-iz];
+	  RealD z1 = z[s][iz];
+	  z[s][iz] = z0*z1*bp
+	    / (b*a*(z1-z0) + z1*bp*(1- (mass[s]-mass[0])*b)); 
+	  bs[s] = b*z[s][iz]/z0; // NB sign  rel to Mike
+	}
+      }
+      ShiftTimer.Stop();
+
+      //Update single precision solutions
+      AXPYTimer.Start();
+      for(int s=0;s<nshift;s++){
+	int ss = s;
+	if( (!converged[s]) ) { 
+	  axpy(psi_f[ss],-bs[s]*alpha[s],ps_f[s],psi_f[ss]);
+	}
+      }
+      c = axpy_norm(r_d,b,mmp_d,r_d);
+      AXPYTimer.Stop();
+    
+      // Convergence checks
+      int all_converged = 1;
+      for(int s=0;s<nshift;s++){
+      
+	if ( (!converged[s]) ){
+	  IterationsToCompleteShift[s] = k;
+	
+	  RealD css  = c * z[s][iz]* z[s][iz];
+	
+	  if(css<rsqf[s]){
+	    if ( ! converged[s] )
+	      std::cout<<GridLogMessage<<"ConjugateGradientMultiShiftMixedPrecCleanup k="<<k<<" Shift "<<s<<" has converged"<<std::endl;
+	    converged[s]=1;
+	  } else {
+	    all_converged=0;
+	  }
+
+	}
+      }
+
+      if ( all_converged || k == MaxIterationsMshift-1){
+
+	SolverTimer.Stop();
+
+	for(int s=0;s<nshift;s++){
+	  precisionChangeFast(psi_d[s],psi_f[s]);
+	}
+
+	
+	if ( all_converged ){
+	  std::cout<<GridLogMessage<< "ConjugateGradientMultiShiftMixedPrecCleanup: All shifts have converged iteration "<<k<<std::endl;
+	  std::cout<<GridLogMessage<< "ConjugateGradientMultiShiftMixedPrecCleanup: Checking solutions"<<std::endl;
+	} else {
+	  std::cout<<GridLogMessage<< "ConjugateGradientMultiShiftMixedPrecCleanup: Not all shifts have converged iteration "<<k<<std::endl;
+	}
+	
+	// Check answers 
+	for(int s=0; s < nshift; s++) { 
+	  Linop_d.HermOpAndNorm(psi_d[s],mmp_d,d,qq);
+	  axpy(tmp_d,mass[s],psi_d[s],mmp_d);
+	  axpy(r_d,-alpha[s],src_d,tmp_d);
+	  RealD rn = norm2(r_d);
+	  RealD cn = norm2(src_d);
+	  TrueResidualShift[s] = std::sqrt(rn/cn);
+	  std::cout<<GridLogMessage<<"ConjugateGradientMultiShiftMixedPrecCleanup: shift["<<s<<"] true residual "<< TrueResidualShift[s] << " target " << mresidual[s] << std::endl;
+
+	  //If we have not reached the desired tolerance, do a (mixed precision) CG cleanup
+	  if(rn >= rsq[s]){
+	    CleanupTimer.Start();
+	    std::cout<<GridLogMessage<<"ConjugateGradientMultiShiftMixedPrecCleanup: performing cleanup step for shift " << s << std::endl;
+
+	    //Setup linear operators for final cleanup
+	    ConjugateGradientMultiShiftMixedPrecSupport::ShiftedLinop<FieldD> Linop_shift_d(Linop_d, mass[s]);
+	    ConjugateGradientMultiShiftMixedPrecSupport::ShiftedLinop<FieldF> Linop_shift_f(Linop_f, mass[s]);
+					       
+	    MixedPrecisionConjugateGradient<FieldD,FieldF> cg(mresidual[s], MaxIterations, MaxIterations, SinglePrecGrid, Linop_shift_f, Linop_shift_d); 
+	    cg(src_d, psi_d[s]);
+	    
+	    TrueResidualShift[s] = cg.TrueResidual;
+	    CleanupTimer.Stop();
+	  }
+	}
+
+	std::cout << GridLogMessage << "ConjugateGradientMultiShiftMixedPrecCleanup: Time Breakdown for body"<<std::endl;
+	std::cout << GridLogMessage << "\tSolver    " << SolverTimer.Elapsed()     <<std::endl;
+	std::cout << GridLogMessage << "\t\tAXPY    " << AXPYTimer.Elapsed()     <<std::endl;
+	std::cout << GridLogMessage << "\t\tMatrix    " << MatrixTimer.Elapsed()     <<std::endl;
+	std::cout << GridLogMessage << "\t\tShift    " << ShiftTimer.Elapsed()     <<std::endl;
+	std::cout << GridLogMessage << "\t\tPrecision Change " << PrecChangeTimer.Elapsed()     <<std::endl;
+	std::cout << GridLogMessage << "\tFinal Cleanup " << CleanupTimer.Elapsed()     <<std::endl;
+	std::cout << GridLogMessage << "\tSolver+Cleanup " << SolverTimer.Elapsed() + CleanupTimer.Elapsed() << std::endl;
+
+	IterationsToComplete = k;	
+
+	return;
+      }
+   
+    }
+    std::cout<<GridLogMessage<<"CG multi shift did not converge"<<std::endl;
+    assert(0);
+  }
+
+};
+NAMESPACE_END(Grid);
+

From 8a1b9073f96391e9c281b24244c458cbaf9fcea4 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Thu, 23 Mar 2023 15:39:30 -0400
Subject: [PATCH 146/146] Mshift update

---
 HMC/Mobius2p1f_DD_EOFA_96I_mshift.cc |  34 +--
 TODO                                 |   2 +
 tests/forces/Test_bdy.cc             | 305 +++++++++++++++++++++++++++
 3 files changed, 313 insertions(+), 28 deletions(-)
 create mode 100644 tests/forces/Test_bdy.cc

diff --git a/HMC/Mobius2p1f_DD_EOFA_96I_mshift.cc b/HMC/Mobius2p1f_DD_EOFA_96I_mshift.cc
index 059a0f20..5572d11f 100644
--- a/HMC/Mobius2p1f_DD_EOFA_96I_mshift.cc
+++ b/HMC/Mobius2p1f_DD_EOFA_96I_mshift.cc
@@ -164,11 +164,6 @@ int main(int argc, char **argv) {
   typedef MobiusEOFAFermionF FermionEOFAActionF;
   typedef typename FermionActionF::FermionField FermionFieldF;
 
-  typedef WilsonImplD2 FermionImplPolicyD2;
-  typedef MobiusFermionD2 FermionActionD2;
-  typedef MobiusEOFAFermionD2 FermionEOFAActionD2;
-  typedef typename FermionActionD2::FermionField FermionFieldD2;
-
   typedef Grid::XmlReader       Serialiser;
 
   //::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
@@ -268,10 +263,8 @@ int main(int argc, char **argv) {
 
   typedef SchurDiagMooeeOperator<FermionActionF,FermionFieldF> LinearOperatorF;
   typedef SchurDiagMooeeOperator<FermionAction ,FermionField > LinearOperatorD;
-  typedef SchurDiagMooeeOperator<FermionActionD2,FermionFieldD2 > LinearOperatorD2;
   typedef SchurDiagMooeeOperator<FermionEOFAActionF,FermionFieldF> LinearOperatorEOFAF;
   typedef SchurDiagMooeeOperator<FermionEOFAAction ,FermionField > LinearOperatorEOFAD;
-  typedef SchurDiagMooeeOperator<FermionEOFAActionD2,FermionFieldD2 > LinearOperatorEOFAD2;
   typedef MixedPrecisionConjugateGradientOperatorFunction<MobiusFermionD,MobiusFermionF,LinearOperatorD,LinearOperatorF> MxPCG;
   typedef MixedPrecisionConjugateGradientOperatorFunction<MobiusEOFAFermionD,MobiusEOFAFermionF,LinearOperatorEOFAD,LinearOperatorEOFAF> MxPCG_EOFA;
 
@@ -324,7 +317,6 @@ int main(int argc, char **argv) {
   // temporarily need a gauge field
   LatticeGaugeFieldD  U(GridPtr); U=Zero();
   LatticeGaugeFieldF  UF(GridPtrF); UF=Zero();
-  LatticeGaugeFieldD2 UD2(GridPtrF); UD2=Zero();
 
   std::cout << GridLogMessage << " Running the HMC "<< std::endl;
   TheHMC.ReadCommandLine(argc,argv);  // params on CML or from param file
@@ -428,7 +420,7 @@ int main(int argc, char **argv) {
 	 ActionCGL, ActionCGR,
 	 DerivativeCGL, DerivativeCGR,
 	 SFRp, true);
-  //  Level2.push_back(&EOFA);
+  Level2.push_back(&EOFA);
 
   ////////////////////////////////////
   // up down action
@@ -453,15 +445,13 @@ int main(int argc, char **argv) {
   std::vector<FermionAction *> Denominators;
   std::vector<FermionActionF *> NumeratorsF;
   std::vector<FermionActionF *> DenominatorsF;
-  std::vector<FermionActionD2 *> NumeratorsD2;
-  std::vector<FermionActionD2 *> DenominatorsD2;
   std::vector<TwoFlavourEvenOddRatioPseudoFermionAction<FermionImplPolicy> *> Quotients;
   std::vector<MxPCG *> ActionMPCG;
   std::vector<MxPCG *> MPCG;
   
 #define MIXED_PRECISION
 #ifdef MIXED_PRECISION
-  std::vector<GeneralEvenOddRatioRationalMixedPrecPseudoFermionAction<FermionImplPolicy,FermionImplPolicyF,FermionImplPolicyD2> *> Bdys;
+  std::vector<GeneralEvenOddRatioRationalMixedPrecPseudoFermionAction<FermionImplPolicy,FermionImplPolicyF,FermionImplPolicy> *> Bdys;
 #else
   std::vector<GeneralEvenOddRatioRationalPseudoFermionAction<FermionImplPolicy> *> Bdys;
 #endif
@@ -536,27 +526,15 @@ int main(int argc, char **argv) {
       Quotients.push_back (new TwoFlavourEvenOddRatioPseudoFermionAction<FermionImplPolicy>(*Numerators[h],*Denominators[h],*MPCG[h],*ActionMPCG[h],CG));
     } else {
 #ifdef MIXED_PRECISION
-      // Use the D2 data types and make them use same grid as single
-      FermionActionD2::ImplParams ParamsDenD2(boundary);
-      FermionActionD2::ImplParams ParamsNumD2(boundary);
-
-      ParamsDenD2.dirichlet = ParamsDen.dirichlet;
-      ParamsDenD2.partialDirichlet = ParamsDen.partialDirichlet;
-      DenominatorsD2.push_back(new FermionActionD2(UD2,*FGridF,*FrbGridF,*GridPtrF,*GridRBPtrF,light_den[h],M5,b,c, ParamsDenD2));
-
-      ParamsNumD2.dirichlet = ParamsNum.dirichlet;
-      ParamsNumD2.partialDirichlet = ParamsNum.partialDirichlet;
-      NumeratorsD2.push_back  (new FermionActionD2(UD2,*FGridF,*FrbGridF,*GridPtrF,*GridRBPtrF,light_num[h],M5,b,c, ParamsNumD2));
-    
-      Bdys.push_back( new GeneralEvenOddRatioRationalMixedPrecPseudoFermionAction<FermionImplPolicy,FermionImplPolicyF,FermionImplPolicyD2>(
+      Bdys.push_back( new GeneralEvenOddRatioRationalMixedPrecPseudoFermionAction<FermionImplPolicy,FermionImplPolicyF,FermionImplPolicy>(
 			   *Numerators[h],*Denominators[h],
 			   *NumeratorsF[h],*DenominatorsF[h],
-			   *NumeratorsD2[h],*DenominatorsD2[h],
+			   *Numerators[h],*Denominators[h],
 			   OFRp, SP_iters) );
-      Bdys.push_back( new GeneralEvenOddRatioRationalMixedPrecPseudoFermionAction<FermionImplPolicy,FermionImplPolicyF,FermionImplPolicyD2>(
+      Bdys.push_back( new GeneralEvenOddRatioRationalMixedPrecPseudoFermionAction<FermionImplPolicy,FermionImplPolicyF,FermionImplPolicy>(
 			   *Numerators[h],*Denominators[h],
 			   *NumeratorsF[h],*DenominatorsF[h],
-			   *NumeratorsD2[h],*DenominatorsD2[h],
+			   *Numerators[h],*Denominators[h],
 			   OFRp, SP_iters) );
 #else
       Bdys.push_back( new GeneralEvenOddRatioRationalPseudoFermionAction<FermionImplPolicy>(*Numerators[h],*Denominators[h],OFRp));
diff --git a/TODO b/TODO
index a6d0f2ac..750deb55 100644
--- a/TODO
+++ b/TODO
@@ -2,6 +2,8 @@
 - - Also faster non-atomic reduction
 - - Remaining PRs
 - - DDHMC
+  - - MixedPrec is the action eval, high precision
+  - - MixedPrecCleanup is the force eval, low precision
 
 =================
 =================
diff --git a/tests/forces/Test_bdy.cc b/tests/forces/Test_bdy.cc
new file mode 100644
index 00000000..c2c97d0d
--- /dev/null
+++ b/tests/forces/Test_bdy.cc
@@ -0,0 +1,305 @@
+/*
+
+  2f Full det MdagM 10^6 force ~ 1.3e7
+rid : Message : 1767.283471 s : +++++++++++++++++++++++++++++++++++++++++++++++++++++++++
+Grid : Message : 1767.283476 s : S1 : 1.52885e+09
+Grid : Message : 1767.283480 s : S2 : 1.52886e+09
+Grid : Message : 1767.283482 s : dS : 8877.34
+Grid : Message : 1767.283483 s : dSpred : 8877.7
+Grid : Message : 1767.283484 s : diff : -0.360484
+Grid : Message : 1767.283485 s : *********************************************************
+
+  2f Full det MpcdagMpc 10^6 force ~ 1.8e6
+Grid : Message : 2399.576962 s : +++++++++++++++++++++++++++++++++++++++++++++++++++++++++
+Grid : Message : 2399.576968 s : S1 : 1.52885e+09
+Grid : Message : 2399.576972 s : S2 : 1.52886e+09
+Grid : Message : 2399.576974 s : dS : 9728.49
+Grid : Message : 2399.576975 s : dSpred : 9726.58
+Grid : Message : 2399.576976 s : diff : 1.90683
+Grid : Message : 2399.576977 s : *********************************************************
+
+  2f bdy MdagM 1500 force Force ~ 2800
+Grid : Message : 4622.385061 s : +++++++++++++++++++++++++++++++++++++++++++++++++++++++++
+Grid : Message : 4622.385067 s : S1 : 1.52885e+09
+Grid : Message : 4622.385071 s : S2 : 1.52885e+09
+Grid : Message : 4622.385072 s : dS : 25.4944
+Grid : Message : 4622.385073 s : dSpred : 25.4672
+Grid : Message : 4622.385074 s : diff : 0.0271414
+Grid : Message : 4622.385075 s : *********************************************************
+
+  2f bdy MpcdagMpc 10^6 force   Force ~ 2200
+Grid : Message : 4622.385061 s : +++++++++++++++++++++++++++++++++++++++++++++++++++++++++
+Grid : Message : 4622.385067 s : S1 : 1.52885e+09
+Grid : Message : 4622.385071 s : S2 : 1.52885e+09
+Grid : Message : 4622.385072 s : dS : 25.4944
+Grid : Message : 4622.385073 s : dSpred : 25.4672
+Grid : Message : 4622.385074 s : diff : 0.0271414
+Grid : Message : 4622.385075 s : *********************************************************
+  
+  1f Bdy Det
+  Optimisation log:  looser rational AND MD tolerances sloppy
+MobiusForce.221179 -- same as HMC. dS is mispredicted Forece  ~2.8
+Grid : Message : 6582.258991 s : dS : 0.024478
+Grid : Message : 6582.258992 s : dSpred : 0.00791876
+Grid : Message : 6582.258994 s : diff : 0.0165592
+
+MobiusForce.221193 -- tight rational AND MD tolerances to 1e-8 ~ 2.8 same
+Grid : Message : 1964.939209 s : S1 : 7.64404e+08
+Grid : Message : 1964.939213 s : S2 : 7.64404e+08
+Grid : Message : 1964.939215 s : dS : -0.00775838 <--- too loose even on action
+Grid : Message : 1964.939216 s : dSpred : -0.00416793 
+Grid : Message : 1964.939217 s : diff : -0.00359045
+
+MobiusForce.221394 -- looser rational, MD tol 1e-8 ~ 2.8 same
+Grid : Message : 1198.346720 s : S1 : 764404649.48886
+Grid : Message : 1198.346760 s : S2 : 764404649.5133
+Grid : Message : 1198.346780 s : dS : 0.024440884590149
+Grid : Message : 1198.346800 s : dSpred : 0.0079145154465184
+Grid : Message : 1198.346810 s : diff : 0.016526369143631
+
+MobiusForce.221394 -- tight rational, MD tol sloppy Force ~ 2.8
+Grid : Message : 2376.921950 s : S1 : 764404436.44069
+Grid : Message : 2376.921954 s : S2 : 764404436.43299
+Grid : Message : 2376.921956 s : dS : -0.0076971054077148
+Grid : Message : 2376.921958 s : dSpred : -0.0041610472282526
+Grid : Message : 2376.921959 s : diff : -0.0035360581794623
+
+*/
+
+//
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./tests/Test_double_ratio.cc
+
+    Copyright (C) 2022
+
+Author: Peter Boyle <pboyle@bnl.gov>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+    *************************************************************************************/
+    /*  END LEGAL */
+#include <Grid/Grid.h>
+
+using namespace std;
+using namespace Grid;
+
+typedef MobiusFermionD FermionAction;
+typedef WilsonImplD FimplD;
+typedef WilsonImplD FermionImplPolicy;
+
+template<class Gimpl>
+void ForceTest(Action<LatticeGaugeField> &action,LatticeGaugeField & U,MomentumFilterBase<LatticeGaugeField> &Filter)
+{
+  GridBase *UGrid = U.Grid();
+
+  std::vector<int> seeds({1,2,3,5});
+  GridSerialRNG            sRNG;         sRNG.SeedFixedIntegers(seeds);
+  GridParallelRNG          RNG4(UGrid);  RNG4.SeedFixedIntegers(seeds);
+
+  LatticeColourMatrix Pmu(UGrid); 
+  LatticeGaugeField P(UGrid); 
+  LatticeGaugeField UdSdU(UGrid); 
+
+  std::cout << GridLogMessage << "*********************************************************"<<std::endl;
+  std::cout << GridLogMessage << " Force test for "<<action.action_name()<<std::endl;
+  std::cout << GridLogMessage << "*********************************************************"<<std::endl;
+  
+  RealD eps=0.005;
+
+  std::cout << GridLogMessage << "+++++++++++++++++++++++++++++++++++++++++++++++++++++++++"<<std::endl;
+  std::cout << GridLogMessage << " Refresh "<<action.action_name()<<std::endl;
+  std::cout << GridLogMessage << "+++++++++++++++++++++++++++++++++++++++++++++++++++++++++"<<std::endl;
+  
+  Gimpl::generate_momenta(P,sRNG,RNG4);
+  Filter.applyFilter(P);
+
+#if  0
+  FieldMetaData header;
+  std::string file("./ckpoint_lat.2000");
+  NerscIO::readConfiguration(U,header,file);
+#else
+  U = 1.0;
+#endif
+  action.refresh(U,sRNG,RNG4);
+
+  std::cout << GridLogMessage << "+++++++++++++++++++++++++++++++++++++++++++++++++++++++++"<<std::endl;
+  std::cout << GridLogMessage << " Action "<<action.action_name()<<std::endl;
+  std::cout << GridLogMessage << "+++++++++++++++++++++++++++++++++++++++++++++++++++++++++"<<std::endl;
+
+  RealD S1 = action.S(U);
+
+  Gimpl::update_field(P,U,eps);
+
+  std::cout << GridLogMessage << "+++++++++++++++++++++++++++++++++++++++++++++++++++++++++"<<std::endl;
+  std::cout << GridLogMessage << " Derivative "<<action.action_name()<<std::endl;
+  std::cout << GridLogMessage << "+++++++++++++++++++++++++++++++++++++++++++++++++++++++++"<<std::endl;
+  action.deriv(U,UdSdU);
+  UdSdU = Ta(UdSdU);
+  Filter.applyFilter(UdSdU);
+
+  DumpSliceNorm("Force",UdSdU,Nd-1);
+  
+  Gimpl::update_field(P,U,eps);
+  std::cout << GridLogMessage << "+++++++++++++++++++++++++++++++++++++++++++++++++++++++++"<<std::endl;
+  std::cout << GridLogMessage << " Action "<<action.action_name()<<std::endl;
+  std::cout << GridLogMessage << "+++++++++++++++++++++++++++++++++++++++++++++++++++++++++"<<std::endl;
+  
+  RealD S2 = action.S(U);
+
+  // Use the derivative
+  LatticeComplex dS(UGrid); dS = Zero();
+  for(int mu=0;mu<Nd;mu++){
+    auto UdSdUmu = PeekIndex<LorentzIndex>(UdSdU,mu);
+    Pmu= PeekIndex<LorentzIndex>(P,mu);
+    dS = dS - trace(Pmu*UdSdUmu)*eps*2.0*2.0;
+  }
+  ComplexD dSpred    = sum(dS);
+  RealD diff =  S2-S1-dSpred.real();
+
+  std::cout<< GridLogMessage << "+++++++++++++++++++++++++++++++++++++++++++++++++++++++++"<<std::endl;
+  std::cout<< GridLogMessage << "S1 : "<< S1    <<std::endl;
+  std::cout<< GridLogMessage << "S2 : "<< S2    <<std::endl;
+  std::cout<< GridLogMessage << "dS : "<< S2-S1 <<std::endl;
+  std::cout<< GridLogMessage << "dSpred : "<< dSpred.real() <<std::endl;
+  std::cout<< GridLogMessage << "diff : "<< diff<<std::endl;
+  std::cout<< GridLogMessage << "*********************************************************"<<std::endl;
+  //  assert(diff<1.0);
+  std::cout<< GridLogMessage << "Done" <<std::endl;
+  std::cout << GridLogMessage << "*********************************************************"<<std::endl;
+}
+
+int main (int argc, char ** argv)
+{
+  Grid_init(&argc,&argv);
+
+  std::cout << std::setprecision(14);
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate mpi_layout  = GridDefaultMpi();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate shm;
+  GlobalSharedMemory::GetShmDims(mpi_layout,shm);
+
+  const int Ls=12;
+  const int Nt = latt_size[3];
+  GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplex::Nsimd()),GridDefaultMpi());
+  GridRedBlackCartesian * UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
+  GridCartesian         * FGrid   = SpaceTimeGrid::makeFiveDimGrid(Ls,UGrid);
+  GridRedBlackCartesian * FrbGrid = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGrid);
+ 
+  ////////////////////////////////////////////////////////////////
+  // Domain decomposed operator
+  ////////////////////////////////////////////////////////////////
+  Coordinate CommDim(Nd);
+  for(int d=0;d<Nd;d++) CommDim[d]= (mpi_layout[d]/shm[d])>1 ? 1 : 0;
+
+  Coordinate NonDirichlet(Nd+1,0);
+  Coordinate Dirichlet(Nd+1,0);
+  Dirichlet[1] = CommDim[0]*latt_size[0]/mpi_layout[0] * shm[0];
+  Dirichlet[2] = CommDim[1]*latt_size[1]/mpi_layout[1] * shm[1];
+  Dirichlet[3] = CommDim[2]*latt_size[2]/mpi_layout[2] * shm[2];
+  Dirichlet[4] = CommDim[3]*latt_size[3]/mpi_layout[3] * shm[3];
+
+  Coordinate Block4(Nd);
+  Block4[0] = Dirichlet[1];
+  Block4[1] = Dirichlet[2];
+  Block4[2] = Dirichlet[3];
+  Block4[3] = Dirichlet[4];
+
+  std::vector<Complex> boundary = {1,1,1,-1};
+  FermionAction::ImplParams Params(boundary);
+  FermionAction::ImplParams ParamsDir(boundary);
+  Params.dirichlet=NonDirichlet;
+  ParamsDir.dirichlet=Dirichlet;
+  ParamsDir.partialDirichlet=1;
+
+  ///////////////////// Gauge Field and Gauge Forces ////////////////////////////
+  LatticeGaugeField U(UGrid);
+
+  RealD beta=6.0;
+  WilsonGaugeActionR PlaqAction(beta);
+  IwasakiGaugeActionR RectAction(beta);
+
+  MomentumFilterNone<LatticeGaugeField> FilterNone;
+  ForceTest<GimplTypesR>(PlaqAction,U,FilterNone);
+  ForceTest<GimplTypesR>(RectAction,U,FilterNone);
+
+  ////////////////////////////////////
+  // Action
+  ////////////////////////////////////
+  RealD mass=0.00078; 
+  RealD pvmass=1.0; 
+  RealD M5=1.8; 
+  RealD b=1.5;
+  RealD c=0.5;
+  
+  // Double versions
+  FermionAction DdwfPeriodic(U,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,b,c,Params);
+  FermionAction PVPeriodic  (U,*FGrid,*FrbGrid,*UGrid,*UrbGrid,pvmass,M5,b,c,Params);
+  FermionAction DdwfDirichlet(U,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,b,c,ParamsDir);
+
+  double StoppingCondition = 1.0e-8;
+  double MaxCGIterations = 50000;
+  ConjugateGradient<LatticeFermion>  CG(StoppingCondition,MaxCGIterations);
+  
+  //////////////////// Two Flavour Determinant Ratio ///////////////////////////////
+  TwoFlavourRatioPseudoFermionAction<FimplD> Nf2(PVPeriodic, DdwfPeriodic,CG,CG);
+  //  ForceTest<GimplTypesR>(Nf2,U,FilterNone);
+
+  //////////////////// Two Flavour Determinant force test Even Odd ///////////////////////////////
+  TwoFlavourEvenOddRatioPseudoFermionAction<FimplD> Nf2eo(PVPeriodic, DdwfPeriodic,CG,CG);
+  //  ForceTest<GimplTypesR>(Nf2eo,U,FilterNone);
+
+  //////////////////// Domain forces ////////////////////
+  int Width=4;
+  DDHMCFilter<WilsonImplD::Field> DDHMCFilter(Block4,Width);
+  
+  //////////////////// Two flavour boundary det  ////////////////////
+  TwoFlavourRatioPseudoFermionAction<FimplD> BdyNf2(DdwfDirichlet, DdwfPeriodic,CG,CG);
+  //  ForceTest<GimplTypesR>(BdyNf2,U,DDHMCFilter);
+
+  //////////////////// Two flavour eo boundary det  ////////////////////
+  TwoFlavourEvenOddRatioPseudoFermionAction<FimplD> BdyNf2eo(DdwfDirichlet, DdwfPeriodic,CG,CG);
+  //  ForceTest<GimplTypesR>(BdyNf2eo,U,DDHMCFilter);
+
+  //////////////////// One flavour boundary det  ////////////////////
+  OneFlavourRationalParams OFRp; // Up/down
+  OFRp.lo       = 4.0e-5;
+  OFRp.hi       = 90.0;
+  OFRp.MaxIter  = 60000;
+  OFRp.tolerance= 1.0e-8;
+  OFRp.mdtolerance= 1.0e-6;
+  OFRp.degree   = 18;
+  OFRp.precision= 80;
+  OFRp.BoundsCheckFreq=0;
+  std::vector<RealD> ActionTolByPole({
+      1.0e-8,1.0e-8,1.0e-8,1.0e-8,
+      1.0e-8,1.0e-8,1.0e-8,1.0e-8,
+      1.0e-8,1.0e-8,1.0e-8,1.0e-8
+    });
+  std::vector<RealD> MDTolByPole({
+      1.0e-6,3.0e-7,1.0e-7,1.0e-7,  // Orig sloppy
+      //      1.0e-8,1.0e-8,1.0e-8,1.0e-8,
+      1.0e-8,1.0e-8,1.0e-8,1.0e-8,
+      1.0e-8,1.0e-8,1.0e-8,1.0e-8
+    });
+  OneFlavourEvenOddRatioRationalPseudoFermionAction<FermionImplPolicy> BdySqrt(DdwfDirichlet,DdwfPeriodic,OFRp);
+  ForceTest<GimplTypesR>(BdySqrt,U,DDHMCFilter);
+
+  Grid_finalize();
+}