Update with LIME library guard

Conflicts:
	benchmarks/Benchmark_su3.cc

.gitignore

@@ -92,6 +92,8 @@ build*/*
 #####################
 *.xcodeproj/*
 build.sh
+.vscode
+*.code-workspace
 
 # Eigen source #
 ################
@@ -106,6 +108,10 @@ lib/fftw/*
 m4/lt*
 m4/libtool.m4
 
+# github pages #
+################
+gh-pages/
+
 # Buck files #
 ##############
 .buck*
@@ -116,4 +122,11 @@ make-bin-BUCK.sh
 # generated sources #
 #####################
 lib/qcd/spin/gamma-gen/*.h
-lib/qcd/spin/gamma-gen/*.cc
+lib/qcd/spin/gamma-gen/*.cc
+lib/version.h
+
+# vs code editor files #
+########################
+.vscode/
+.vscode/settings.json
+settings.json

.travis.yml

@@ -7,64 +7,8 @@ cache:
 matrix:
   include:
     - os:        osx
-      osx_image: xcode7.2
+      osx_image: xcode8.3
       compiler: clang
-    - compiler: gcc
-      addons:
-        apt:
-          sources:
-            - ubuntu-toolchain-r-test
-          packages:
-            - g++-4.9
-            - libmpfr-dev
-            - libgmp-dev
-            - libmpc-dev
-            - libopenmpi-dev
-            - openmpi-bin
-            - binutils-dev
-      env: VERSION=-4.9
-    - compiler: gcc
-      addons:
-        apt:
-          sources:
-            - ubuntu-toolchain-r-test
-          packages:
-            - g++-5
-            - libmpfr-dev
-            - libgmp-dev
-            - libmpc-dev
-            - libopenmpi-dev
-            - openmpi-bin
-            - binutils-dev
-      env: VERSION=-5
-    - compiler: clang
-      addons:
-        apt:
-          sources:
-            - ubuntu-toolchain-r-test
-          packages:
-            - g++-4.8
-            - libmpfr-dev
-            - libgmp-dev
-            - libmpc-dev
-            - libopenmpi-dev
-            - openmpi-bin
-            - binutils-dev
-      env: CLANG_LINK=http://llvm.org/releases/3.8.0/clang+llvm-3.8.0-x86_64-linux-gnu-ubuntu-14.04.tar.xz
-    - compiler: clang
-      addons:
-        apt:
-          sources:
-            - ubuntu-toolchain-r-test
-          packages:
-            - g++-4.8
-            - libmpfr-dev
-            - libgmp-dev
-            - libmpc-dev
-            - libopenmpi-dev
-            - openmpi-bin
-            - binutils-dev
-      env: CLANG_LINK=http://llvm.org/releases/3.7.0/clang+llvm-3.7.0-x86_64-linux-gnu-ubuntu-14.04.tar.xz
       
 before_install:
     - export GRIDDIR=`pwd`
@@ -73,13 +17,15 @@ before_install:
     - if [[ "$TRAVIS_OS_NAME" == "linux" ]] && [[ "$CC" == "clang" ]]; then export LD_LIBRARY_PATH="${GRIDDIR}/clang/lib:${LD_LIBRARY_PATH}"; fi
     - if [[ "$TRAVIS_OS_NAME" == "osx" ]]; then brew update; fi
     - if [[ "$TRAVIS_OS_NAME" == "osx" ]]; then brew install libmpc; fi
-    - if [[ "$TRAVIS_OS_NAME" == "osx" ]]; then brew install openmpi; fi
-    - if [[ "$TRAVIS_OS_NAME" == "osx" ]] && [[ "$CC" == "gcc" ]]; then brew install gcc5; fi
     
 install:
     - export CC=$CC$VERSION
     - export CXX=$CXX$VERSION
     - echo $PATH
+    - which autoconf
+    - autoconf  --version
+    - which automake
+    - automake  --version
     - which $CC
     - $CC  --version
     - which $CXX
@@ -92,15 +38,10 @@ script:
     - cd build
     - ../configure --enable-precision=single --enable-simd=SSE4 --enable-comms=none
     - make -j4 
-    - ./benchmarks/Benchmark_dwf --threads 1
+    - ./benchmarks/Benchmark_dwf --threads 1 --debug-signals
     - echo make clean
     - ../configure --enable-precision=double --enable-simd=SSE4 --enable-comms=none
     - make -j4
-    - ./benchmarks/Benchmark_dwf --threads 1
-    - echo make clean
-    - if [[ "$TRAVIS_OS_NAME" == "linux" ]]; then export CXXFLAGS='-DMPI_UINT32_T=MPI_UNSIGNED -DMPI_UINT64_T=MPI_UNSIGNED_LONG'; fi
-    - ../configure --enable-precision=single --enable-simd=SSE4 --enable-comms=mpi-auto
-    - make -j4
-    - if [[ "$TRAVIS_OS_NAME" == "linux" ]]; then mpirun.openmpi -n 2 ./benchmarks/Benchmark_dwf --threads 1 --mpi 2.1.1.1; fi
-    - if [[ "$TRAVIS_OS_NAME" == "osx" ]]; then mpirun -n 2 ./benchmarks/Benchmark_dwf --threads 1 --mpi 2.1.1.1; fi
+    - ./benchmarks/Benchmark_dwf --threads 1 --debug-signals
+    - make check
 

Makefile.am

@@ -3,10 +3,19 @@ SUBDIRS = lib benchmarks tests extras
 
 include $(top_srcdir)/doxygen.inc
 
-tests: all
-	$(MAKE) -C tests tests
+bin_SCRIPTS=grid-config
 
-.PHONY: tests doxygen-run doxygen-doc $(DX_PS_GOAL) $(DX_PDF_GOAL)
+BUILT_SOURCES = version.h
+
+version.h:
+	echo "`git log -n 1 --format=format:"#define GITHASH \\"%H:%d\\"%n" HEAD`" > $(srcdir)/lib/version.h
+
+.PHONY: bench check tests doxygen-run doxygen-doc $(DX_PS_GOAL) $(DX_PDF_GOAL)
+
+tests-local: all
+bench-local: all
+check-local: all
 
 AM_CXXFLAGS += -I$(top_builddir)/include
+
 ACLOCAL_AMFLAGS = -I m4

README.md

@@ -1,41 +1,13 @@
-# Grid
-<table>
-<tr>
-    <td>Last stable release</td>
-    <td><a href="https://travis-ci.org/paboyle/Grid">
-    <img src="https://travis-ci.org/paboyle/Grid.svg?branch=master"></a>
-    </td>
-</tr>
-<tr>
-    <td>Development branch</td>
-    <td><a href="https://travis-ci.org/paboyle/Grid">
-    <img src="https://travis-ci.org/paboyle/Grid.svg?branch=develop"></a>
-    </td>
-</tr>
-</table>
+# Grid [![Teamcity status](http://ci.cliath.ph.ed.ac.uk/app/rest/builds/aggregated/strob:(buildType:(affectedProject(id:Grid)),branch:name:develop)/statusIcon.svg)](http://ci.cliath.ph.ed.ac.uk/project.html?projectId=Grid&tab=projectOverview) [![Travis status](https://travis-ci.org/paboyle/Grid.svg?branch=develop)](https://travis-ci.org/paboyle/Grid)
 
 **Data parallel C++ mathematical object library.**
 
 License: GPL v2.
 
-Last update Nov 2016.
+Last update June 2017.
 
 _Please do not send pull requests to the `master` branch which is reserved for releases._
 
-### Bug report
-
-_To help us tracking and solving more efficiently issues with Grid, please report problems using the issue system of GitHub rather than sending emails to Grid developers._
-
-When you file an issue, please go though the following checklist:
-
-1. Check that the code is pointing to the `HEAD` of `develop` or any commit in `master` which is tagged with a version number. 
-2. Give a description of the target platform (CPU, network, compiler). Please give the full CPU part description, using for example `cat /proc/cpuinfo | grep 'model name' | uniq` (Linux) or `sysctl machdep.cpu.brand_string` (macOS) and the full output the `--version` option of your compiler.
-3. Give the exact `configure` command used.
-4. Attach `config.log`.
-5. Attach `config.summary`.
-6. Attach the output of `make V=1`.
-7. Describe the issue and any previous attempt to solve it. If relevant, show how to reproduce the issue using a minimal working example.
-
 
 
 ### Description
@@ -58,13 +30,68 @@ optimally use MPI, OpenMP and SIMD parallelism under the hood. This is a signifi
 for most programmers.
 
 The layout transformations are parametrised by the SIMD vector length. This adapts according to the architecture.
-Presently SSE4 (128 bit) AVX, AVX2, QPX (256 bit), IMCI, and AVX512 (512 bit) targets are supported (ARM NEON on the way).
+Presently SSE4, ARM NEON (128 bits) AVX, AVX2, QPX (256 bits), IMCI and AVX512 (512 bits) targets are supported.
 
-These are presented as `vRealF`, `vRealD`, `vComplexF`, and `vComplexD` internal vector data types. These may be useful in themselves for other programmers.
+These are presented as `vRealF`, `vRealD`, `vComplexF`, and `vComplexD` internal vector data types. 
 The corresponding scalar types are named `RealF`, `RealD`, `ComplexF` and `ComplexD`.
 
 MPI, OpenMP, and SIMD parallelism are present in the library.
-Please see https://arxiv.org/abs/1512.03487 for more detail.
+Please see [this paper](https://arxiv.org/abs/1512.03487) for more detail.
+
+
+### Compilers
+
+Intel ICPC v16.0.3 and later
+
+Clang v3.5 and later (need 3.8 and later for OpenMP)
+
+GCC   v4.9.x (recommended)
+
+GCC   v6.3 and later
+
+### Important: 
+
+Some versions of GCC appear to have a bug under high optimisation (-O2, -O3).
+
+The safety of these compiler versions cannot be guaranteed at this time. Follow Issue 100 for details and updates.
+
+GCC   v5.x
+
+GCC   v6.1, v6.2
+
+### Bug report
+
+_To help us tracking and solving more efficiently issues with Grid, please report problems using the issue system of GitHub rather than sending emails to Grid developers._
+
+When you file an issue, please go though the following checklist:
+
+1. Check that the code is pointing to the `HEAD` of `develop` or any commit in `master` which is tagged with a version number. 
+2. Give a description of the target platform (CPU, network, compiler). Please give the full CPU part description, using for example `cat /proc/cpuinfo | grep 'model name' | uniq` (Linux) or `sysctl machdep.cpu.brand_string` (macOS) and the full output the `--version` option of your compiler.
+3. Give the exact `configure` command used.
+4. Attach `config.log`.
+5. Attach `grid.config.summary`.
+6. Attach the output of `make V=1`.
+7. Describe the issue and any previous attempt to solve it. If relevant, show how to reproduce the issue using a minimal working example.
+
+### Required libraries
+Grid requires:
+
+[GMP](https://gmplib.org/), 
+
+[MPFR](http://www.mpfr.org/) 
+
+Bootstrapping grid downloads and uses for internal dense matrix (non-QCD operations) the Eigen library.
+
+Grid optionally uses:
+
+[HDF5](https://support.hdfgroup.org/HDF5/)  
+
+[LIME](http://usqcd-software.github.io/c-lime/) for ILDG and SciDAC file format support. 
+
+[FFTW](http://www.fftw.org) either generic version or via the Intel MKL library.
+
+LAPACK either generic version or Intel MKL library.
+
 
 ### Quick start
 First, start by cloning the repository:
@@ -95,10 +122,10 @@ install Grid. Other options are detailed in the next section, you can also use `
 `CXX`, `CXXFLAGS`, `LDFLAGS`, ... environment variables can be modified to
 customise the build.
 
-Finally, you can build and install Grid:
+Finally, you can build, check, and install Grid:
 
 ``` bash
-make; make install
+make; make check; make install
 ```
 
 To minimise the build time, only the tests at the root of the `tests` directory are built by default. If you want to build tests in the sub-directory `<subdir>` you can execute:
@@ -121,7 +148,7 @@ If you want to build all the tests at once just use `make tests`.
 - `--enable-gen-simd-width=<size>`: select the size (in bytes) of the generic SIMD vector type (default: 32 bytes).
 - `--enable-precision={single|double}`: set the default precision (default: `double`).
 - `--enable-precision=<comm>`: Use `<comm>` for message passing (default: `none`). A list of possible SIMD targets is detailed in a section below.
-- `--enable-rng={ranlux48|mt19937}`: choose the RNG (default: `ranlux48 `).
+- `--enable-rng={sitmo|ranlux48|mt19937}`: choose the RNG (default: `sitmo `).
 - `--disable-timers`: disable system dependent high-resolution timers.
 - `--enable-chroma`: enable Chroma regression tests.
 - `--enable-doxygen-doc`: enable the Doxygen documentation generation (build with `make doxygen-doc`)
@@ -135,7 +162,6 @@ The following options can be use with the `--enable-comms=` option to target dif
 | `none`         | no communications                                             |
 | `mpi[-auto]`   | MPI communications                                            |
 | `mpi3[-auto]`  | MPI communications using MPI 3 shared memory                  |
-| `mpi3l[-auto]` | MPI communications using MPI 3 shared memory and leader model |
 | `shmem `       | Cray SHMEM communications                                     |
 
 For the MPI interfaces the optional `-auto` suffix instructs the `configure` scripts to determine all the necessary compilation and linking flags. This is done by extracting the informations from the MPI wrapper specified in the environment variable `MPICXX` (if not specified `configure` will scan though a list of default names). The `-auto` suffix is not supported by the Cray environment wrapper scripts. Use the standard versions instead.  
@@ -153,18 +179,19 @@ The following options can be use with the `--enable-simd=` option to target diff
 | `AVXFMA4`   | AVX (256 bit) + FMA4                   |
 | `AVX2`      | AVX 2 (256 bit)                        |
 | `AVX512`    | AVX 512 bit                            |
-| `QPX`       | QPX (256 bit)                          |
+| `NEONv8`    | [ARM NEON](http://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc.den0024a/ch07s03.html) (128 bit)                     |
+| `QPX`       | IBM QPX (256 bit)                      |
 
 Alternatively, some CPU codenames can be directly used:
 
 | `<code>`    | Description                            |
 | ----------- | -------------------------------------- |
-| `KNC`       | [Intel Xeon Phi codename Knights Corner](http://ark.intel.com/products/codename/57721/Knights-Corner) |
 | `KNL`       | [Intel Xeon Phi codename Knights Landing](http://ark.intel.com/products/codename/48999/Knights-Landing) |
+| `SKL`       | [Intel Skylake with AVX512 extensions](https://ark.intel.com/products/codename/37572/Skylake#@server) |
 | `BGQ`       | Blue Gene/Q                            |
 
 #### Notes:
-- We currently support AVX512 only for the Intel compiler. Support for GCC and clang will appear in future versions of Grid when the AVX512 support within GCC and clang will be more advanced.
+- We currently support AVX512 for the Intel compiler and GCC (KNL and SKL target). Support for clang will appear in future versions of Grid when the AVX512 support in the compiler will be more advanced.
 - For BG/Q only [bgclang](http://trac.alcf.anl.gov/projects/llvm-bgq) is supported. We do not presently plan to support more compilers for this platform.
 - BG/Q performances are currently rather poor. This is being investigated for future versions.
 - The vector size for the `GEN` target can be specified with the `configure` script option `--enable-gen-simd-width`.
@@ -176,21 +203,205 @@ The following configuration is recommended for the Intel Knights Landing platfor
 ``` bash
 ../configure --enable-precision=double\
              --enable-simd=KNL        \
-             --enable-comms=mpi-auto \
-             --with-gmp=<path>        \
-             --with-mpfr=<path>       \
+             --enable-comms=mpi-auto  \
              --enable-mkl             \
              CXX=icpc MPICXX=mpiicpc
 ```
+The MKL flag enables use of BLAS and FFTW from the Intel Math Kernels Library.
 
-where `<path>` is the UNIX prefix where GMP and MPFR are installed. If you are working on a Cray machine that does not use the `mpiicpc` wrapper, please use:
+If you are working on a Cray machine that does not use the `mpiicpc` wrapper, please use:
 
 ``` bash
 ../configure --enable-precision=double\
              --enable-simd=KNL        \
              --enable-comms=mpi       \
-             --with-gmp=<path>        \
-             --with-mpfr=<path>       \
              --enable-mkl             \
              CXX=CC CC=cc
-```
+```
+
+If gmp and mpfr are NOT in standard places (/usr/) these flags may be needed:
+``` bash
+               --with-gmp=<path>        \
+               --with-mpfr=<path>       \
+```
+where `<path>` is the UNIX prefix where GMP and MPFR are installed. 
+
+Knight's Landing with Intel Omnipath adapters with two adapters per node 
+presently performs better with use of more than one rank per node, using shared memory 
+for interior communication. This is the mpi3 communications implementation. 
+We recommend four ranks per node for best performance, but optimum is local volume dependent.
+
+``` bash
+../configure --enable-precision=double\
+             --enable-simd=KNL        \
+             --enable-comms=mpi3-auto \
+             --enable-mkl             \
+             CC=icpc MPICXX=mpiicpc 
+```
+
+### Build setup for Intel Haswell Xeon platform
+
+The following configuration is recommended for the Intel Haswell platform:
+
+``` bash
+../configure --enable-precision=double\
+             --enable-simd=AVX2       \
+             --enable-comms=mpi3-auto \
+             --enable-mkl             \
+             CXX=icpc MPICXX=mpiicpc
+```
+The MKL flag enables use of BLAS and FFTW from the Intel Math Kernels Library.
+
+If gmp and mpfr are NOT in standard places (/usr/) these flags may be needed:
+``` bash
+               --with-gmp=<path>        \
+               --with-mpfr=<path>       \
+```
+where `<path>` is the UNIX prefix where GMP and MPFR are installed. 
+
+If you are working on a Cray machine that does not use the `mpiicpc` wrapper, please use:
+
+``` bash
+../configure --enable-precision=double\
+             --enable-simd=AVX2       \
+             --enable-comms=mpi3      \
+             --enable-mkl             \
+             CXX=CC CC=cc
+```
+Since Dual socket nodes are commonplace, we recommend MPI-3 as the default with the use of 
+one rank per socket. If using the Intel MPI library, threads should be pinned to NUMA domains using
+```
+        export I_MPI_PIN=1
+```
+This is the default.
+
+### Build setup for Intel Skylake Xeon platform
+
+The following configuration is recommended for the Intel Skylake platform:
+
+``` bash
+../configure --enable-precision=double\
+             --enable-simd=AVX512     \
+             --enable-comms=mpi3      \
+             --enable-mkl             \
+             CXX=mpiicpc
+```
+The MKL flag enables use of BLAS and FFTW from the Intel Math Kernels Library.
+
+If gmp and mpfr are NOT in standard places (/usr/) these flags may be needed:
+``` bash
+               --with-gmp=<path>        \
+               --with-mpfr=<path>       \
+```
+where `<path>` is the UNIX prefix where GMP and MPFR are installed. 
+
+If you are working on a Cray machine that does not use the `mpiicpc` wrapper, please use:
+
+``` bash
+../configure --enable-precision=double\
+             --enable-simd=AVX512     \
+             --enable-comms=mpi3      \
+             --enable-mkl             \
+             CXX=CC CC=cc
+```
+Since Dual socket nodes are commonplace, we recommend MPI-3 as the default with the use of 
+one rank per socket. If using the Intel MPI library, threads should be pinned to NUMA domains using
+``` 
+        export I_MPI_PIN=1
+```
+This is the default. 
+
+#### Expected Skylake Gold 6148 dual socket (single prec, single node 20+20 cores) performance using NUMA MPI mapping): 
+
+mpirun -n 2 benchmarks/Benchmark_dwf --grid 16.16.16.16 --mpi 2.1.1.1 --cacheblocking 2.2.2.2 --dslash-asm --shm 1024 --threads 18 
+
+TBA
+
+
+### Build setup for AMD EPYC / RYZEN
+
+The AMD EPYC is a multichip module comprising 32 cores spread over four distinct chips each with 8 cores.
+So, even with a single socket node there is a quad-chip module. Dual socket nodes with 64 cores total
+are common. Each chip within the module exposes a separate NUMA domain.
+There are four NUMA domains per socket and we recommend one MPI rank per NUMA domain.
+MPI-3 is recommended with the use of four ranks per socket,
+and 8 threads per rank. 
+
+The following configuration is recommended for the AMD EPYC platform.
+
+``` bash
+../configure --enable-precision=double\
+             --enable-simd=AVX2       \
+             --enable-comms=mpi3 \
+             CXX=mpicxx 
+```
+
+If gmp and mpfr are NOT in standard places (/usr/) these flags may be needed:
+``` bash
+               --with-gmp=<path>        \
+               --with-mpfr=<path>       \
+```
+where `<path>` is the UNIX prefix where GMP and MPFR are installed. 
+
+Using MPICH and g++ v4.9.2, best performance can be obtained using explicit GOMP_CPU_AFFINITY flags for each MPI rank.
+This can be done by invoking MPI on a wrapper script omp_bind.sh to handle this. 
+
+It is recommended to run 8 MPI ranks on a single dual socket AMD EPYC, with 8 threads per rank using MPI3 and
+shared memory to communicate within this node:
+
+mpirun -np 8 ./omp_bind.sh ./Benchmark_dwf --mpi 2.2.2.1 --dslash-unroll --threads 8 --grid 16.16.16.16 --cacheblocking 4.4.4.4 
+
+Where omp_bind.sh does the following:
+```
+#!/bin/bash
+
+numanode=` expr $PMI_RANK % 8 `
+basecore=`expr $numanode \* 16`
+core0=`expr $basecore + 0 `
+core1=`expr $basecore + 2 `
+core2=`expr $basecore + 4 `
+core3=`expr $basecore + 6 `
+core4=`expr $basecore + 8 `
+core5=`expr $basecore + 10 `
+core6=`expr $basecore + 12 `
+core7=`expr $basecore + 14 `
+
+export GOMP_CPU_AFFINITY="$core0 $core1 $core2 $core3 $core4 $core5 $core6 $core7"
+echo GOMP_CUP_AFFINITY $GOMP_CPU_AFFINITY
+
+$@
+```
+
+Performance:
+
+#### Expected AMD EPYC 7601 dual socket (single prec, single node 32+32 cores) performance using NUMA MPI mapping): 
+
+mpirun  -np 8 ./omp_bind.sh ./Benchmark_dwf --threads 8 --mpi 2.2.2.1 --dslash-unroll --grid 16.16.16.16 --cacheblocking 4.4.4.4
+
+TBA
+
+### Build setup for BlueGene/Q
+
+To be written...
+
+### Build setup for ARM Neon
+
+To be written...
+
+### Build setup for laptops, other compilers, non-cluster builds
+
+Many versions of g++ and clang++ work with Grid, and involve merely replacing CXX (and MPICXX),
+and omit the enable-mkl flag. 
+
+Single node builds are enabled with 
+```
+            --enable-comms=none
+```
+
+FFTW support that is not in the default search path may then enabled with
+```
+    --with-fftw=<installpath>
+```
+
+BLAS will not be compiled in by default, and Lanczos will default to Eigen diagonalisation.
+

TODO

@@ -1,6 +1,51 @@
 TODO:
 ---------------
 
+Code item work list
+
+a) namespaces & indentation
+ GRID_BEGIN_NAMESPACE();
+ GRID_END_NAMESPACE();
+-- delete QCD namespace
+
+b) GPU branch
+- start branch
+- Increase Macro use in core library support; prepare for change
+- Audit volume of "device" code
+- Virtual function audit
+- Start port once Nvidia box is up
+- Cut down volume of code for first port? How?
+
+Physics item work list:
+
+1)- BG/Q port and check ; Andrew says ok.
+2)- Consistent linear solver flop count/rate -- PARTIAL, time but no flop/s yet
+3)- Physical propagator interface
+4)- Multigrid Wilson and DWF, compare to other Multigrid implementations
+5)- HDCR resume
+
+----------------------------
+Recent DONE 
+-- RNG I/O in ILDG/SciDAC (minor) 
+-- Precision conversion and sort out localConvert      <-- partial/easy
+-- Conserved currents (Andrew)
+-- Split grid
+-- Christoph's local basis expansion Lanczos
+-- MultiRHS with spread out extra dim -- Go through filesystem with SciDAC I/O ; <-- DONE ; bmark cori
+-- Lanczos Remove DenseVector, DenseMatrix; Use Eigen instead. <-- DONE
+-- GaugeFix into central location                      <-- DONE
+-- Scidac and Ildg metadata handling                   <-- DONE
+-- Binary I/O MPI2 IO                                  <-- DONE
+-- Binary I/O speed up & x-strips                      <-- DONE
+-- Cut down the exterior overhead                      <-- DONE
+-- Interior legs from SHM comms                        <-- DONE
+-- Half-precision comms                                <-- DONE
+-- Merge high precision reduction into develop         <-- DONE
+-- BlockCG, BCGrQ                                      <-- DONE
+-- multiRHS DWF; benchmark on Cori/BNL for comms elimination <-- DONE
+   -- slice* linalg routines for multiRHS, BlockCG    
+
+-----
 * Forces; the UdSdU  term in gauge force term is half of what I think it should
   be. This is a consequence of taking ONLY the first term in:
 
@@ -21,16 +66,8 @@ TODO:
   This means we must double the force in the Test_xxx_force routines, and is the origin of the factor of two.
   This 2x is applied by hand in the fermion routines and in the Test_rect_force routine.
 
-
-Policies:
-
-* Link smearing/boundary conds; Policy class based implementation ; framework more in place
-
 * Support different boundary conditions (finite temp, chem. potential ... )
 
-* Support different fermion representations? 
-  - contained entirely within the integrator presently
-
 - Sign of force term.
 
 - Reversibility test.
@@ -41,11 +78,6 @@ Policies:
 
 - Audit oIndex usage for cb behaviour
 
-- Rectangle gauge actions.
-  Iwasaki,
-  Symanzik,
-  ... etc...
-
 - Prepare multigrid for HMC. - Alternate setup schemes.
 
 - Support for ILDG --- ugly, not done
@@ -55,9 +87,11 @@ Policies:
 - FFTnD ?
 
 - Gparity; hand opt use template specialisation elegance to enable the optimised paths ?
+
 - Gparity force term; Gparity (R)HMC.
-- Random number state save restore
+
 - Mobius implementation clean up to rmove #if 0 stale code sequences
+
 - CG -- profile carefully, kernel fusion, whole CG performance measurements.
 
 ================================================================
@@ -90,6 +124,7 @@ Insert/Extract
 Not sure of status of this -- reverify. Things are working nicely now though.
 
 * Make the Tensor types and Complex etc... play more nicely.
+
   - TensorRemove is a hack, come up with a long term rationalised approach to Complex vs. Scalar<Scalar<Scalar<Complex > > >
     QDP forces use of "toDouble" to get back to non tensor scalar. This role is presently taken TensorRemove, but I
     want to introduce a syntax that does not require this.
@@ -112,6 +147,8 @@ Not sure of status of this -- reverify. Things are working nicely now though.
 RECENT
 ---------------
 
+  - Support different fermion representations? -- DONE
+  - contained entirely within the integrator presently
   - Clean up HMC                                                             -- DONE
   - LorentzScalar<GaugeField> gets Gauge link type (cleaner).                -- DONE
   - Simplified the integrators a bit.                                        -- DONE
@@ -123,6 +160,26 @@ RECENT
   - Parallel io improvements                                  -- DONE
   - Plaquette and link trace checks into nersc reader from the Grid_nersc_io.cc test. -- DONE
 
+
+DONE:
+- MultiArray -- MultiRHS done
+- ConjugateGradientMultiShift -- DONE
+- MCR                         -- DONE
+- Remez -- Mike or Boost?     -- DONE
+- Proto (ET)                  -- DONE
+- uBlas                       -- DONE ; Eigen
+- Potentially Useful Boost libraries -- DONE ; Eigen
+- Aligned allocator; memory pool -- DONE
+- Multiprecision              -- DONE
+- Serialization               -- DONE
+- Regex -- Not needed
+- Tokenize -- Why?
+
+- Random number state save restore -- DONE
+- Rectangle gauge actions. -- DONE
+  Iwasaki,
+  Symanzik,
+  ... etc...
 Done: Cayley, Partial , ContFrac force terms.
 
 DONE
@@ -207,6 +264,7 @@ Done
 FUNCTIONALITY: it pleases me to keep track of things I have done (keeps me arguably sane)
 ======================================================================================================
 
+* Link smearing/boundary conds; Policy class based implementation ; framework more in place -- DONE
 * Command line args for geometry, simd, etc. layout. Is it necessary to have -- DONE
   user pass these? Is this a QCD specific?
 

VERSION

@@ -1,6 +1,5 @@
-Version : 0.6.0
+Version : 0.8.0
 
-- AVX512, AVX2, AVX, SSE good
-- Clang 3.5 and above, ICPC v16 and above, GCC 4.9 and above
-- MPI and MPI3
-- HiRep, Smearing, Generic gauge group
+- Clang 3.5 and above, ICPC v16 and above, GCC 6.3 and above recommended
+- MPI and MPI3 comms optimisations for KNL and OPA finished
+- Half precision comms

benchmarks/Benchmark_IO.cc

@@ -0,0 +1,108 @@
+#include <Grid/Grid.h>
+#ifdef HAVE_LIME
+
+using namespace std;
+using namespace Grid;
+using namespace Grid::QCD;
+
+#define MSG cout << GridLogMessage
+#define SEP \
+"============================================================================="
+#ifndef BENCH_IO_LMAX
+#define BENCH_IO_LMAX 40
+#endif
+
+typedef function<void(const string, LatticeFermion &)> WriterFn;
+typedef function<void(LatticeFermion &, const string)> ReaderFn;
+
+string filestem(const int l)
+{
+  return "iobench_l" + to_string(l);
+}
+
+void limeWrite(const string filestem, LatticeFermion &vec)
+{
+  emptyUserRecord record;
+  ScidacWriter    binWriter(vec._grid->IsBoss());
+
+  binWriter.open(filestem + ".bin");
+  binWriter.writeScidacFieldRecord(vec, record);
+  binWriter.close();
+}
+
+void limeRead(LatticeFermion &vec, const string filestem)
+{
+  emptyUserRecord record;
+  ScidacReader    binReader;
+
+  binReader.open(filestem + ".bin");
+  binReader.readScidacFieldRecord(vec, record);
+  binReader.close();
+}
+
+void writeBenchmark(const int l, const WriterFn &write)
+{
+  auto                      mpi  = GridDefaultMpi();
+  auto                      simd = GridDefaultSimd(Nd, vComplex::Nsimd());
+  vector<int>               latt = {l*mpi[0], l*mpi[1], l*mpi[2], l*mpi[3]};
+  unique_ptr<GridCartesian> gPt(SpaceTimeGrid::makeFourDimGrid(latt, simd, mpi));
+  GridCartesian             *g = gPt.get();
+  GridParallelRNG           rng(g);
+  LatticeFermion            vec(g);
+  emptyUserRecord           record;
+  ScidacWriter              binWriter(g->IsBoss());
+
+  cout << "-- Local volume " << l << "^4" << endl;
+  random(rng, vec);
+  write(filestem(l), vec);
+}
+
+void readBenchmark(const int l, const ReaderFn &read)
+{
+  auto                      mpi  = GridDefaultMpi();
+  auto                      simd = GridDefaultSimd(Nd, vComplex::Nsimd());
+  vector<int>               latt = {l*mpi[0], l*mpi[1], l*mpi[2], l*mpi[3]};
+  unique_ptr<GridCartesian> gPt(SpaceTimeGrid::makeFourDimGrid(latt, simd, mpi));
+  GridCartesian             *g = gPt.get();
+  LatticeFermion            vec(g);
+  emptyUserRecord           record;
+  ScidacReader              binReader;
+
+  cout << "-- Local volume " << l << "^4" << endl;
+  read(vec, filestem(l));
+}
+
+int main (int argc, char ** argv)
+{
+  Grid_init(&argc,&argv);
+
+  auto simd = GridDefaultSimd(Nd,vComplex::Nsimd());
+  auto mpi  = GridDefaultMpi();
+
+  int64_t threads = GridThread::GetThreads();
+  MSG << "Grid is setup to use " << threads << " threads" << endl;
+  MSG << SEP << endl;
+  MSG << "Benchmark Lime write" << endl;
+  MSG << SEP << endl;
+  for (int l = 4; l <= BENCH_IO_LMAX; l += 2)
+  {
+    writeBenchmark(l, limeWrite);
+  }
+
+  MSG << "Benchmark Lime read" << endl;
+  MSG << SEP << endl;
+  for (int l = 4; l <= BENCH_IO_LMAX; l += 2)
+  {
+    readBenchmark(l, limeRead);
+  }
+
+  Grid_finalize();
+
+  return EXIT_SUCCESS;
+}
+#else
+int main (int argc, char ** argv)
+{
+  return EXIT_SUCCESS;
+}
+#endif

benchmarks/Benchmark_ITT.cc

@@ -0,0 +1,800 @@
+    /*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./benchmarks/Benchmark_memory_bandwidth.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>
+
+using namespace std;
+using namespace Grid;
+using namespace Grid::QCD;
+
+typedef WilsonFermion5D<DomainWallVec5dImplR> WilsonFermion5DR;
+typedef WilsonFermion5D<DomainWallVec5dImplF> WilsonFermion5DF;
+typedef WilsonFermion5D<DomainWallVec5dImplD> WilsonFermion5DD;
+
+
+std::vector<int> L_list;
+std::vector<int> Ls_list;
+std::vector<double> mflop_list;
+
+double mflop_ref;
+double mflop_ref_err;
+
+int NN_global;
+
+struct time_statistics{
+  double mean;
+  double err;
+  double min;
+  double max;
+
+  void statistics(std::vector<double> v){
+      double sum = std::accumulate(v.begin(), v.end(), 0.0);
+      mean = sum / v.size();
+
+      std::vector<double> diff(v.size());
+      std::transform(v.begin(), v.end(), diff.begin(), [=](double x) { return x - mean; });
+      double sq_sum = std::inner_product(diff.begin(), diff.end(), diff.begin(), 0.0);
+      err = std::sqrt(sq_sum / (v.size()*(v.size() - 1)));
+
+      auto result = std::minmax_element(v.begin(), v.end());
+      min = *result.first;
+      max = *result.second;
+}
+};
+
+void comms_header(){
+  std::cout <<GridLogMessage << " L  "<<"\t"<<" Ls  "<<"\t"
+            <<std::setw(11)<<"bytes"<<"MB/s uni (err/min/max)"<<"\t\t"<<"MB/s bidi (err/min/max)"<<std::endl;
+};
+
+Gamma::Algebra Gmu [] = {
+  Gamma::Algebra::GammaX,
+  Gamma::Algebra::GammaY,
+  Gamma::Algebra::GammaZ,
+  Gamma::Algebra::GammaT
+};
+struct controls {
+  int Opt;
+  int CommsOverlap;
+  Grid::CartesianCommunicator::CommunicatorPolicy_t CommsAsynch;
+  //  int HugePages;
+};
+
+class Benchmark {
+public:
+  static void Decomposition (void ) {
+
+    int threads = GridThread::GetThreads();
+    std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+    std::cout<<GridLogMessage << "= Grid is setup to use "<<threads<<" threads"<<std::endl;
+    std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+    std::cout<<GridLogMessage<<"Grid Default Decomposition patterns\n";
+    std::cout<<GridLogMessage<<"\tOpenMP threads : "<<GridThread::GetThreads()<<std::endl;
+    std::cout<<GridLogMessage<<"\tMPI tasks      : "<<GridCmdVectorIntToString(GridDefaultMpi())<<std::endl;
+    std::cout<<GridLogMessage<<"\tvReal          : "<<sizeof(vReal )*8    <<"bits ; " <<GridCmdVectorIntToString(GridDefaultSimd(4,vReal::Nsimd()))<<std::endl;
+    std::cout<<GridLogMessage<<"\tvRealF         : "<<sizeof(vRealF)*8    <<"bits ; " <<GridCmdVectorIntToString(GridDefaultSimd(4,vRealF::Nsimd()))<<std::endl;
+    std::cout<<GridLogMessage<<"\tvRealD         : "<<sizeof(vRealD)*8    <<"bits ; " <<GridCmdVectorIntToString(GridDefaultSimd(4,vRealD::Nsimd()))<<std::endl;
+    std::cout<<GridLogMessage<<"\tvComplex       : "<<sizeof(vComplex )*8 <<"bits ; " <<GridCmdVectorIntToString(GridDefaultSimd(4,vComplex::Nsimd()))<<std::endl;
+    std::cout<<GridLogMessage<<"\tvComplexF      : "<<sizeof(vComplexF)*8 <<"bits ; " <<GridCmdVectorIntToString(GridDefaultSimd(4,vComplexF::Nsimd()))<<std::endl;
+    std::cout<<GridLogMessage<<"\tvComplexD      : "<<sizeof(vComplexD)*8 <<"bits ; " <<GridCmdVectorIntToString(GridDefaultSimd(4,vComplexD::Nsimd()))<<std::endl;
+    std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+
+  }
+
+  static void Comms(void)
+  {
+    int Nloop=200;
+    int nmu=0;
+    int maxlat=32;
+
+    std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplexD::Nsimd());
+    std::vector<int> mpi_layout  = GridDefaultMpi();
+
+    for(int mu=0;mu<Nd;mu++) if (mpi_layout[mu]>1) nmu++;
+
+    std::vector<double> t_time(Nloop);
+    time_statistics timestat;
+
+    std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
+    std::cout<<GridLogMessage << "= Benchmarking threaded STENCIL halo exchange in "<<nmu<<" dimensions"<<std::endl;
+    std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
+    comms_header();
+
+    for(int lat=4;lat<=maxlat;lat+=4){
+      for(int Ls=8;Ls<=8;Ls*=2){
+
+	std::vector<int> latt_size  ({lat*mpi_layout[0],
+	      lat*mpi_layout[1],
+	      lat*mpi_layout[2],
+	      lat*mpi_layout[3]});
+
+	GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
+	RealD Nrank = Grid._Nprocessors;
+	RealD Nnode = Grid.NodeCount();
+	RealD ppn = Nrank/Nnode;
+
+	std::vector<HalfSpinColourVectorD *> xbuf(8);
+	std::vector<HalfSpinColourVectorD *> rbuf(8);
+	Grid.ShmBufferFreeAll();
+	for(int d=0;d<8;d++){
+	  xbuf[d] = (HalfSpinColourVectorD *)Grid.ShmBufferMalloc(lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
+	  rbuf[d] = (HalfSpinColourVectorD *)Grid.ShmBufferMalloc(lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
+	  bzero((void *)xbuf[d],lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
+	  bzero((void *)rbuf[d],lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
+	}
+
+	int bytes=lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD);
+	int ncomm;
+	double dbytes;
+	std::vector<double> times(Nloop);
+	for(int i=0;i<Nloop;i++){
+
+	  double start=usecond();
+
+	  dbytes=0;
+	  ncomm=0;
+
+	  parallel_for(int dir=0;dir<8;dir++){
+
+	    double tbytes;
+	    int mu =dir % 4;
+
+	    if (mpi_layout[mu]>1 ) {
+	        
+	      int xmit_to_rank;
+	      int recv_from_rank;
+	      if ( dir == mu ) { 
+		int comm_proc=1;
+		Grid.ShiftedRanks(mu,comm_proc,xmit_to_rank,recv_from_rank);
+	      } else { 
+		int comm_proc = mpi_layout[mu]-1;
+		Grid.ShiftedRanks(mu,comm_proc,xmit_to_rank,recv_from_rank);
+	      }
+	      tbytes= Grid.StencilSendToRecvFrom((void *)&xbuf[dir][0], xmit_to_rank,
+						 (void *)&rbuf[dir][0], recv_from_rank,
+						 bytes,dir);
+	  
+#ifdef GRID_OMP
+#pragma omp atomic
+#endif
+	      ncomm++;
+
+#ifdef GRID_OMP
+#pragma omp atomic
+#endif
+	      dbytes+=tbytes;
+	    }
+	  }
+	  Grid.Barrier();
+	  double stop=usecond();
+	  t_time[i] = stop-start; // microseconds
+	}
+
+	timestat.statistics(t_time);
+	//	for(int i=0;i<t_time.size();i++){
+	//	  std::cout << i<<" "<<t_time[i]<<std::endl;
+	//	}
+
+	dbytes=dbytes*ppn;
+	double xbytes    = dbytes*0.5;
+	double rbytes    = dbytes*0.5;
+	double bidibytes = dbytes;
+
+	std::cout<<GridLogMessage << std::setw(4) << lat<<"\t"<<Ls<<"\t"
+		 <<std::setw(11) << bytes<< std::fixed << std::setprecision(1) << std::setw(7)
+		 <<std::right<< xbytes/timestat.mean<<"  "<< xbytes*timestat.err/(timestat.mean*timestat.mean)<< " "
+		 <<xbytes/timestat.max <<" "<< xbytes/timestat.min  
+		 << "\t\t"<<std::setw(7)<< bidibytes/timestat.mean<< "  " << bidibytes*timestat.err/(timestat.mean*timestat.mean) << " "
+		 << bidibytes/timestat.max << " " << bidibytes/timestat.min << std::endl;
+
+ 
+	
+	    }
+    }    
+
+    return;
+  }
+
+  static void Memory(void)
+  {
+    const int Nvec=8;
+    typedef Lattice< iVector< vReal,Nvec> > LatticeVec;
+    typedef iVector<vReal,Nvec> Vec;
+
+    std::vector<int> simd_layout = GridDefaultSimd(Nd,vReal::Nsimd());
+    std::vector<int> mpi_layout  = GridDefaultMpi();
+
+    std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+    std::cout<<GridLogMessage << "= Benchmarking a*x + y bandwidth"<<std::endl;
+    std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+    std::cout<<GridLogMessage << "  L  "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s"<<"\t\t"<<"Gflop/s"<<"\t\t seconds"<< "\t\tGB/s / node"<<std::endl;
+    std::cout<<GridLogMessage << "----------------------------------------------------------"<<std::endl;
+  
+    uint64_t NP;
+    uint64_t NN;
+
+
+  uint64_t lmax=48;
+#define NLOOP (100*lmax*lmax*lmax*lmax/lat/lat/lat/lat)
+
+    GridSerialRNG          sRNG;      sRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9}));
+    for(int lat=8;lat<=lmax;lat+=4){
+
+      std::vector<int> latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
+      int64_t vol= latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
+      GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
+
+      NP= Grid.RankCount();
+      NN =Grid.NodeCount();
+
+      Vec rn ; random(sRNG,rn);
+
+      LatticeVec z(&Grid); z=rn;
+      LatticeVec x(&Grid); x=rn;
+      LatticeVec y(&Grid); y=rn;
+      double a=2.0;
+
+      uint64_t Nloop=NLOOP;
+
+      double start=usecond();
+      for(int i=0;i<Nloop;i++){
+	z=a*x-y;
+        x._odata[0]=z._odata[0]; // force serial dependency to prevent optimise away
+        y._odata[4]=z._odata[4];
+      }
+      double stop=usecond();
+      double time = (stop-start)/Nloop*1000;
+     
+      double flops=vol*Nvec*2;// mul,add
+      double bytes=3.0*vol*Nvec*sizeof(Real);
+      std::cout<<GridLogMessage<<std::setprecision(3) 
+	       << lat<<"\t\t"<<bytes<<"   \t\t"<<bytes/time<<"\t\t"<<flops/time<<"\t\t"<<(stop-start)/1000./1000.
+	       << "\t\t"<< bytes/time/NN <<std::endl;
+
+    }
+  };
+
+  static double DWF5(int Ls,int L)
+  {
+    RealD mass=0.1;
+    RealD M5  =1.8;
+
+    double mflops;
+    double mflops_best = 0;
+    double mflops_worst= 0;
+    std::vector<double> mflops_all;
+
+    ///////////////////////////////////////////////////////
+    // Set/Get the layout & grid size
+    ///////////////////////////////////////////////////////
+    int threads = GridThread::GetThreads();
+    std::vector<int> mpi = GridDefaultMpi(); assert(mpi.size()==4);
+    std::vector<int> local({L,L,L,L});
+
+    GridCartesian         * TmpGrid   = SpaceTimeGrid::makeFourDimGrid(std::vector<int>({64,64,64,64}), 
+								       GridDefaultSimd(Nd,vComplex::Nsimd()),GridDefaultMpi());
+    uint64_t NP = TmpGrid->RankCount();
+    uint64_t NN = TmpGrid->NodeCount();
+    NN_global=NN;
+    uint64_t SHM=NP/NN;
+
+    std::vector<int> internal;
+    if      ( SHM == 1 )   internal = std::vector<int>({1,1,1,1});
+    else if ( SHM == 2 )   internal = std::vector<int>({2,1,1,1});
+    else if ( SHM == 4 )   internal = std::vector<int>({2,2,1,1});
+    else if ( SHM == 8 )   internal = std::vector<int>({2,2,2,1});
+    else assert(0);
+
+    std::vector<int> nodes({mpi[0]/internal[0],mpi[1]/internal[1],mpi[2]/internal[2],mpi[3]/internal[3]});
+    std::vector<int> latt4({local[0]*nodes[0],local[1]*nodes[1],local[2]*nodes[2],local[3]*nodes[3]});
+
+    ///////// Welcome message ////////////
+    std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+    std::cout<<GridLogMessage << "Benchmark DWF Ls vec on "<<L<<"^4 local volume "<<std::endl;
+    std::cout<<GridLogMessage << "* Global volume  : "<<GridCmdVectorIntToString(latt4)<<std::endl;
+    std::cout<<GridLogMessage << "* Ls             : "<<Ls<<std::endl;
+    std::cout<<GridLogMessage << "* MPI ranks      : "<<GridCmdVectorIntToString(mpi)<<std::endl;
+    std::cout<<GridLogMessage << "* Intranode      : "<<GridCmdVectorIntToString(internal)<<std::endl;
+    std::cout<<GridLogMessage << "* nodes          : "<<GridCmdVectorIntToString(nodes)<<std::endl;
+    std::cout<<GridLogMessage << "* Using "<<threads<<" threads"<<std::endl;
+    std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+
+    ///////// Lattice Init ////////////
+    GridCartesian         * UGrid    = SpaceTimeGrid::makeFourDimGrid(latt4, GridDefaultSimd(Nd,vComplex::Nsimd()),GridDefaultMpi());
+    GridRedBlackCartesian * UrbGrid  = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
+    GridCartesian         * sUGrid   = SpaceTimeGrid::makeFourDimDWFGrid(latt4,GridDefaultMpi());
+    GridRedBlackCartesian * sUrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(sUGrid);
+    GridCartesian         * sFGrid   = SpaceTimeGrid::makeFiveDimDWFGrid(Ls,UGrid);
+    GridRedBlackCartesian * sFrbGrid = SpaceTimeGrid::makeFiveDimDWFRedBlackGrid(Ls,UGrid);
+
+    ///////// RNG Init ////////////
+    std::vector<int> seeds4({1,2,3,4});
+    std::vector<int> seeds5({5,6,7,8});
+    GridParallelRNG          RNG4(UGrid);  RNG4.SeedFixedIntegers(seeds4);
+    GridParallelRNG          RNG5(sFGrid);  RNG5.SeedFixedIntegers(seeds5);
+    std::cout << GridLogMessage << "Initialised RNGs" << std::endl;
+
+    ///////// Source preparation ////////////
+    LatticeFermion src   (sFGrid); random(RNG5,src);
+    LatticeFermion tmp   (sFGrid);
+
+    RealD N2 = 1.0/::sqrt(norm2(src));
+    src = src*N2;
+    
+    LatticeGaugeField Umu(UGrid);  SU3::HotConfiguration(RNG4,Umu); 
+
+    WilsonFermion5DR sDw(Umu,*sFGrid,*sFrbGrid,*sUGrid,*sUrbGrid,M5);
+    LatticeFermion src_e (sFrbGrid);
+    LatticeFermion src_o (sFrbGrid);
+    LatticeFermion r_e   (sFrbGrid);
+    LatticeFermion r_o   (sFrbGrid);
+    LatticeFermion r_eo  (sFGrid);
+    LatticeFermion err   (sFGrid);
+    {
+
+      pickCheckerboard(Even,src_e,src);
+      pickCheckerboard(Odd,src_o,src);
+
+#if defined(AVX512) 
+      const int num_cases = 6;
+      std::string fmt("A/S ; A/O ; U/S ; U/O ; G/S ; G/O ");
+#else
+      const int num_cases = 4;
+      std::string fmt("U/S ; U/O ; G/S ; G/O ");
+#endif
+      controls Cases [] = {
+#ifdef AVX512
+	{ QCD::WilsonKernelsStatic::OptInlineAsm , QCD::WilsonKernelsStatic::CommsThenCompute ,CartesianCommunicator::CommunicatorPolicySequential  },
+	{ QCD::WilsonKernelsStatic::OptInlineAsm , QCD::WilsonKernelsStatic::CommsAndCompute  ,CartesianCommunicator::CommunicatorPolicySequential  },
+#endif
+	{ QCD::WilsonKernelsStatic::OptHandUnroll, QCD::WilsonKernelsStatic::CommsThenCompute ,CartesianCommunicator::CommunicatorPolicySequential  },
+	{ QCD::WilsonKernelsStatic::OptHandUnroll, QCD::WilsonKernelsStatic::CommsAndCompute  ,CartesianCommunicator::CommunicatorPolicySequential  },
+	{ QCD::WilsonKernelsStatic::OptGeneric   , QCD::WilsonKernelsStatic::CommsThenCompute ,CartesianCommunicator::CommunicatorPolicySequential  },
+	{ QCD::WilsonKernelsStatic::OptGeneric   , QCD::WilsonKernelsStatic::CommsAndCompute  ,CartesianCommunicator::CommunicatorPolicySequential  }
+      }; 
+
+      for(int c=0;c<num_cases;c++) {
+
+	QCD::WilsonKernelsStatic::Comms = Cases[c].CommsOverlap;
+	QCD::WilsonKernelsStatic::Opt   = Cases[c].Opt;
+	CartesianCommunicator::SetCommunicatorPolicy(Cases[c].CommsAsynch);
+
+	std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+	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;
+	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;
+	if ( sizeof(Real)==4 )   std::cout << GridLogMessage<< "* SINGLE precision "<<std::endl;
+	if ( sizeof(Real)==8 )   std::cout << GridLogMessage<< "* DOUBLE precision "<<std::endl;
+	std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+
+	int nwarm = 100;
+	uint64_t ncall = 1000;
+
+	double t0=usecond();
+	sFGrid->Barrier();
+	for(int i=0;i<nwarm;i++){
+	  sDw.DhopEO(src_o,r_e,DaggerNo);
+	}
+	sFGrid->Barrier();
+	double t1=usecond();
+
+	sDw.ZeroCounters();
+	time_statistics timestat;
+	std::vector<double> t_time(ncall);
+	for(uint64_t i=0;i<ncall;i++){
+	  t0=usecond();
+	  sDw.DhopEO(src_o,r_e,DaggerNo);
+	  t1=usecond();
+	  t_time[i] = t1-t0;
+	}
+	sFGrid->Barrier();
+	
+	double volume=Ls;  for(int mu=0;mu<Nd;mu++) volume=volume*latt4[mu];
+	double flops=(1344.0*volume)/2;
+	double mf_hi, mf_lo, mf_err;
+
+	timestat.statistics(t_time);
+	mf_hi = flops/timestat.min;
+	mf_lo = flops/timestat.max;
+	mf_err= flops/timestat.min * timestat.err/timestat.mean;
+
+	mflops = flops/timestat.mean;
+	mflops_all.push_back(mflops);
+	if ( mflops_best == 0   ) mflops_best = mflops;
+	if ( mflops_worst== 0   ) mflops_worst= mflops;
+	if ( mflops>mflops_best ) mflops_best = mflops;
+	if ( mflops<mflops_worst) mflops_worst= mflops;
+
+	std::cout<<GridLogMessage << std::fixed << std::setprecision(1)<<"sDeo mflop/s =   "<< mflops << " ("<<mf_err<<") " << mf_lo<<"-"<<mf_hi <<std::endl;
+	std::cout<<GridLogMessage << std::fixed << std::setprecision(1)<<"sDeo mflop/s per rank   "<< mflops/NP<<std::endl;
+	std::cout<<GridLogMessage << std::fixed << std::setprecision(1)<<"sDeo mflop/s per node   "<< mflops/NN<<std::endl;
+
+	sDw.Report();
+
+      }
+      double robust = mflops_worst/mflops_best;;
+      std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+      std::cout<<GridLogMessage << L<<"^4 x "<<Ls<< " sDeo Best  mflop/s        =   "<< mflops_best << " ; " << mflops_best/NN<<" per node " <<std::endl;
+      std::cout<<GridLogMessage << L<<"^4 x "<<Ls<< " sDeo Worst mflop/s        =   "<< mflops_worst<< " ; " << mflops_worst/NN<<" per node " <<std::endl;
+
+      std::cout<<GridLogMessage <<std::setprecision(3)<< L<<"^4 x "<<Ls<< " Performance Robustness   =   "<< robust <<std::endl;
+      std::cout<<GridLogMessage <<fmt << std::endl;
+      std::cout<<GridLogMessage;
+
+      for(int i=0;i<mflops_all.size();i++){
+	std::cout<<mflops_all[i]/NN<<" ; " ;
+      }
+      std::cout<<std::endl;
+      std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+
+    }
+    return mflops_best;
+  }
+
+  static double DWF(int Ls,int L, double & robust)
+  {
+    RealD mass=0.1;
+    RealD M5  =1.8;
+
+    double mflops;
+    double mflops_best = 0;
+    double mflops_worst= 0;
+    std::vector<double> mflops_all;
+
+    ///////////////////////////////////////////////////////
+    // Set/Get the layout & grid size
+    ///////////////////////////////////////////////////////
+    int threads = GridThread::GetThreads();
+    std::vector<int> mpi = GridDefaultMpi(); assert(mpi.size()==4);
+    std::vector<int> local({L,L,L,L});
+
+    GridCartesian         * TmpGrid   = SpaceTimeGrid::makeFourDimGrid(std::vector<int>({64,64,64,64}), 
+								       GridDefaultSimd(Nd,vComplex::Nsimd()),GridDefaultMpi());
+    uint64_t NP = TmpGrid->RankCount();
+    uint64_t NN = TmpGrid->NodeCount();
+    NN_global=NN;
+    uint64_t SHM=NP/NN;
+
+    std::vector<int> internal;
+    if      ( SHM == 1 )   internal = std::vector<int>({1,1,1,1});
+    else if ( SHM == 2 )   internal = std::vector<int>({2,1,1,1});
+    else if ( SHM == 4 )   internal = std::vector<int>({2,2,1,1});
+    else if ( SHM == 8 )   internal = std::vector<int>({2,2,2,1});
+    else assert(0);
+
+    std::vector<int> nodes({mpi[0]/internal[0],mpi[1]/internal[1],mpi[2]/internal[2],mpi[3]/internal[3]});
+    std::vector<int> latt4({local[0]*nodes[0],local[1]*nodes[1],local[2]*nodes[2],local[3]*nodes[3]});
+
+    ///////// Welcome message ////////////
+    std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+    std::cout<<GridLogMessage << "Benchmark DWF on "<<L<<"^4 local volume "<<std::endl;
+    std::cout<<GridLogMessage << "* Global volume  : "<<GridCmdVectorIntToString(latt4)<<std::endl;
+    std::cout<<GridLogMessage << "* Ls             : "<<Ls<<std::endl;
+    std::cout<<GridLogMessage << "* MPI ranks      : "<<GridCmdVectorIntToString(mpi)<<std::endl;
+    std::cout<<GridLogMessage << "* Intranode      : "<<GridCmdVectorIntToString(internal)<<std::endl;
+    std::cout<<GridLogMessage << "* nodes          : "<<GridCmdVectorIntToString(nodes)<<std::endl;
+    std::cout<<GridLogMessage << "* Using "<<threads<<" threads"<<std::endl;
+    std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+
+
+    ///////// Lattice Init ////////////
+    GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(latt4, GridDefaultSimd(Nd,vComplex::Nsimd()),GridDefaultMpi());
+    GridRedBlackCartesian * UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
+    GridCartesian         * FGrid   = SpaceTimeGrid::makeFiveDimGrid(Ls,UGrid);
+    GridRedBlackCartesian * FrbGrid = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGrid);
+
+    
+    ///////// RNG Init ////////////
+    std::vector<int> seeds4({1,2,3,4});
+    std::vector<int> seeds5({5,6,7,8});
+    GridParallelRNG          RNG4(UGrid);  RNG4.SeedFixedIntegers(seeds4);
+    GridParallelRNG          RNG5(FGrid);  RNG5.SeedFixedIntegers(seeds5);
+    std::cout << GridLogMessage << "Initialised RNGs" << std::endl;
+
+    ///////// Source preparation ////////////
+    LatticeFermion src   (FGrid); random(RNG5,src);
+    LatticeFermion ref   (FGrid);
+    LatticeFermion tmp   (FGrid);
+
+    RealD N2 = 1.0/::sqrt(norm2(src));
+    src = src*N2;
+    
+    LatticeGaugeField Umu(UGrid);  SU3::HotConfiguration(RNG4,Umu); 
+
+    DomainWallFermionR Dw(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+
+    ////////////////////////////////////
+    // Naive wilson implementation
+    ////////////////////////////////////
+    {
+      LatticeGaugeField Umu5d(FGrid); 
+      std::vector<LatticeColourMatrix> U(4,FGrid);
+      for(int ss=0;ss<Umu._grid->oSites();ss++){
+	for(int s=0;s<Ls;s++){
+	  Umu5d._odata[Ls*ss+s] = Umu._odata[ss];
+	}
+      }
+      ref = zero;
+      for(int mu=0;mu<Nd;mu++){
+	U[mu] = PeekIndex<LorentzIndex>(Umu5d,mu);
+      }
+      for(int mu=0;mu<Nd;mu++){
+	
+	tmp = U[mu]*Cshift(src,mu+1,1);
+	ref=ref + tmp - Gamma(Gmu[mu])*tmp;
+	
+	tmp =adj(U[mu])*src;
+	tmp =Cshift(tmp,mu+1,-1);
+	ref=ref + tmp + Gamma(Gmu[mu])*tmp;
+      }
+      ref = -0.5*ref;
+    }
+
+    LatticeFermion src_e (FrbGrid);
+    LatticeFermion src_o (FrbGrid);
+    LatticeFermion r_e   (FrbGrid);
+    LatticeFermion r_o   (FrbGrid);
+    LatticeFermion r_eo  (FGrid);
+    LatticeFermion err   (FGrid);
+    {
+
+      pickCheckerboard(Even,src_e,src);
+      pickCheckerboard(Odd,src_o,src);
+
+#if defined(AVX512) 
+      const int num_cases = 6;
+      std::string fmt("A/S ; A/O ; U/S ; U/O ; G/S ; G/O ");
+#else
+      const int num_cases = 4;
+      std::string fmt("U/S ; U/O ; G/S ; G/O ");
+#endif
+      controls Cases [] = {
+#ifdef AVX512
+	{ QCD::WilsonKernelsStatic::OptInlineAsm , QCD::WilsonKernelsStatic::CommsThenCompute ,CartesianCommunicator::CommunicatorPolicySequential  },
+	{ QCD::WilsonKernelsStatic::OptInlineAsm , QCD::WilsonKernelsStatic::CommsAndCompute  ,CartesianCommunicator::CommunicatorPolicySequential  },
+#endif
+	{ QCD::WilsonKernelsStatic::OptHandUnroll, QCD::WilsonKernelsStatic::CommsThenCompute ,CartesianCommunicator::CommunicatorPolicySequential  },
+	{ QCD::WilsonKernelsStatic::OptHandUnroll, QCD::WilsonKernelsStatic::CommsAndCompute  ,CartesianCommunicator::CommunicatorPolicySequential  },
+	{ QCD::WilsonKernelsStatic::OptGeneric   , QCD::WilsonKernelsStatic::CommsThenCompute ,CartesianCommunicator::CommunicatorPolicySequential  },
+	{ QCD::WilsonKernelsStatic::OptGeneric   , QCD::WilsonKernelsStatic::CommsAndCompute  ,CartesianCommunicator::CommunicatorPolicySequential  }
+      }; 
+
+      for(int c=0;c<num_cases;c++) {
+
+	QCD::WilsonKernelsStatic::Comms = Cases[c].CommsOverlap;
+	QCD::WilsonKernelsStatic::Opt   = Cases[c].Opt;
+	CartesianCommunicator::SetCommunicatorPolicy(Cases[c].CommsAsynch);
+
+	std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+	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;
+	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;
+	if ( sizeof(Real)==4 )   std::cout << GridLogMessage<< "* SINGLE precision "<<std::endl;
+	if ( sizeof(Real)==8 )   std::cout << GridLogMessage<< "* DOUBLE precision "<<std::endl;
+	std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+
+	int nwarm = 200;
+	double t0=usecond();
+	FGrid->Barrier();
+	for(int i=0;i<nwarm;i++){
+	  Dw.DhopEO(src_o,r_e,DaggerNo);
+	}
+	FGrid->Barrier();
+	double t1=usecond();
+	//	uint64_t ncall = (uint64_t) 2.5*1000.0*1000.0*nwarm/(t1-t0);
+	//	if (ncall < 500) ncall = 500;
+	uint64_t ncall = 1000;
+
+	FGrid->Broadcast(0,&ncall,sizeof(ncall));
+
+	//	std::cout << GridLogMessage << " Estimate " << ncall << " calls per second"<<std::endl;
+	Dw.ZeroCounters();
+
+	time_statistics timestat;
+	std::vector<double> t_time(ncall);
+	for(uint64_t i=0;i<ncall;i++){
+	  t0=usecond();
+	  Dw.DhopEO(src_o,r_e,DaggerNo);
+	  t1=usecond();
+	  t_time[i] = t1-t0;
+	}
+	FGrid->Barrier();
+	
+	double volume=Ls;  for(int mu=0;mu<Nd;mu++) volume=volume*latt4[mu];
+	double flops=(1344.0*volume)/2;
+	double mf_hi, mf_lo, mf_err;
+
+	timestat.statistics(t_time);
+	mf_hi = flops/timestat.min;
+	mf_lo = flops/timestat.max;
+	mf_err= flops/timestat.min * timestat.err/timestat.mean;
+
+	mflops = flops/timestat.mean;
+	mflops_all.push_back(mflops);
+	if ( mflops_best == 0   ) mflops_best = mflops;
+	if ( mflops_worst== 0   ) mflops_worst= mflops;
+	if ( mflops>mflops_best ) mflops_best = mflops;
+	if ( mflops<mflops_worst) mflops_worst= mflops;
+
+	std::cout<<GridLogMessage << std::fixed << std::setprecision(1)<<"Deo mflop/s =   "<< mflops << " ("<<mf_err<<") " << mf_lo<<"-"<<mf_hi <<std::endl;
+	std::cout<<GridLogMessage << std::fixed << std::setprecision(1)<<"Deo mflop/s per rank   "<< mflops/NP<<std::endl;
+	std::cout<<GridLogMessage << std::fixed << std::setprecision(1)<<"Deo mflop/s per node   "<< mflops/NN<<std::endl;
+
+	Dw.Report();
+
+	Dw.DhopEO(src_o,r_e,DaggerNo);
+	Dw.DhopOE(src_e,r_o,DaggerNo);
+	setCheckerboard(r_eo,r_o);
+	setCheckerboard(r_eo,r_e);
+	err = r_eo-ref; 
+	std::cout<<GridLogMessage << "norm diff   "<< norm2(err)<<std::endl;
+	assert((norm2(err)<1.0e-4));
+
+      }
+      robust = mflops_worst/mflops_best;
+      std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+      std::cout<<GridLogMessage << L<<"^4 x "<<Ls<< " Deo Best  mflop/s        =   "<< mflops_best << " ; " << mflops_best/NN<<" per node " <<std::endl;
+      std::cout<<GridLogMessage << L<<"^4 x "<<Ls<< " Deo Worst mflop/s        =   "<< mflops_worst<< " ; " << mflops_worst/NN<<" per node " <<std::endl;
+      std::cout<<GridLogMessage << std::fixed<<std::setprecision(3)<< L<<"^4 x "<<Ls<< " Performance Robustness   =   "<< robust  <<std::endl;
+      std::cout<<GridLogMessage <<fmt << std::endl;
+      std::cout<<GridLogMessage ;
+
+      for(int i=0;i<mflops_all.size();i++){
+	std::cout<<mflops_all[i]/NN<<" ; " ;
+      }
+      std::cout<<std::endl;
+      std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+
+    }
+    return mflops_best;
+  }
+
+};
+
+int main (int argc, char ** argv)
+{
+  Grid_init(&argc,&argv);
+
+  CartesianCommunicator::SetCommunicatorPolicy(CartesianCommunicator::CommunicatorPolicySequential);
+#ifdef KNL
+  LebesgueOrder::Block = std::vector<int>({8,2,2,2});
+#else
+  LebesgueOrder::Block = std::vector<int>({2,2,2,2});
+#endif
+  Benchmark::Decomposition();
+
+  int do_memory=1;
+  int do_comms =1;
+  int do_su3   =0;
+  int do_wilson=1;
+  int do_dwf   =1;
+
+  if ( do_su3 ) {
+    // empty for now
+  }
+#if 1
+  int sel=2;
+  std::vector<int> L_list({8,12,16,24});
+#else
+  int sel=1;
+  std::vector<int> L_list({8,12});
+#endif
+  int selm1=sel-1;
+  std::vector<double> robust_list;
+
+  std::vector<double> wilson;
+  std::vector<double> dwf4;
+  std::vector<double> dwf5;
+
+  if ( do_wilson ) {
+    int Ls=1;
+    std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+    std::cout<<GridLogMessage << " Wilson dslash 4D vectorised" <<std::endl;
+    std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+    for(int l=0;l<L_list.size();l++){
+      double robust;
+      wilson.push_back(Benchmark::DWF(1,L_list[l],robust));
+    }
+  }
+
+  int Ls=16;
+  if ( do_dwf ) {
+    std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+    std::cout<<GridLogMessage << " Domain wall dslash 4D vectorised" <<std::endl;
+    std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+    for(int l=0;l<L_list.size();l++){
+      double robust;
+      double result = Benchmark::DWF(Ls,L_list[l],robust) ;
+      dwf4.push_back(result);
+      robust_list.push_back(robust);
+    }
+  }
+
+  if ( do_dwf ) {
+    std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+    std::cout<<GridLogMessage << " Domain wall dslash 4D vectorised" <<std::endl;
+    std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+    for(int l=0;l<L_list.size();l++){
+      dwf5.push_back(Benchmark::DWF5(Ls,L_list[l]));
+    }
+
+  }
+
+  if ( do_dwf ) {
+
+  std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+  std::cout<<GridLogMessage << " Summary table Ls="<<Ls <<std::endl;
+  std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+  std::cout<<GridLogMessage << "L \t\t Wilson \t DWF4 \t DWF5 " <<std::endl;
+  for(int l=0;l<L_list.size();l++){
+    std::cout<<GridLogMessage << L_list[l] <<" \t\t "<< wilson[l]<<" \t "<<dwf4[l]<<" \t "<<dwf5[l] <<std::endl;
+  }
+  std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+  }
+
+  int NN=NN_global;
+  if ( do_memory ) {
+    std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+    std::cout<<GridLogMessage << " Memory benchmark " <<std::endl;
+    std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+    Benchmark::Memory();
+  }
+
+  if ( do_comms && (NN>1) ) {
+    std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+    std::cout<<GridLogMessage << " Communications benchmark " <<std::endl;
+    std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+    Benchmark::Comms();
+  }
+
+  if ( do_dwf ) {
+  std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+  std::cout<<GridLogMessage << " Per Node Summary table Ls="<<Ls <<std::endl;
+  std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+  std::cout<<GridLogMessage << " L \t\t Wilson\t\t DWF4  \t\t DWF5 " <<std::endl;
+  for(int l=0;l<L_list.size();l++){
+    std::cout<<GridLogMessage << L_list[l] <<" \t\t "<< wilson[l]/NN<<" \t "<<dwf4[l]/NN<<" \t "<<dwf5[l] /NN<<std::endl;
+  }
+  std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+
+  std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+  std::cout<<GridLogMessage << " Comparison point     result: "  << 0.5*(dwf4[sel]+dwf4[selm1])/NN << " Mflop/s per node"<<std::endl;
+  std::cout<<GridLogMessage << " Comparison point is 0.5*("<<dwf4[sel]/NN<<"+"<<dwf4[selm1]/NN << ") "<<std::endl;
+  std::cout<<std::setprecision(3);
+  std::cout<<GridLogMessage << " Comparison point robustness: "  << robust_list[sel] <<std::endl;
+  std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
+
+  }
+
+
+  Grid_finalize();
+}

benchmarks/Benchmark_comms.cc

@@ -31,6 +31,32 @@ using namespace std;
 using namespace Grid;
 using namespace Grid::QCD;
 
+struct time_statistics{
+  double mean;
+  double err;
+  double min;
+  double max;
+
+  void statistics(std::vector<double> v){
+      double sum = std::accumulate(v.begin(), v.end(), 0.0);
+      mean = sum / v.size();
+
+      std::vector<double> diff(v.size());
+      std::transform(v.begin(), v.end(), diff.begin(), [=](double x) { return x - mean; });
+      double sq_sum = std::inner_product(diff.begin(), diff.end(), diff.begin(), 0.0);
+      err = std::sqrt(sq_sum / (v.size()*(v.size() - 1)));
+
+      auto result = std::minmax_element(v.begin(), v.end());
+      min = *result.first;
+      max = *result.second;
+}
+};
+
+void header(){
+  std::cout <<GridLogMessage << " L  "<<"\t"<<" Ls  "<<"\t"
+            <<std::setw(11)<<"bytes"<<"MB/s uni (err/min/max)"<<"\t\t"<<"MB/s bidi (err/min/max)"<<std::endl;
+};
+
 int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
@@ -40,17 +66,21 @@ int main (int argc, char ** argv)
   int threads = GridThread::GetThreads();
   std::cout<<GridLogMessage << "Grid is setup to use "<<threads<<" threads"<<std::endl;
 
-  int Nloop=10;
+  int Nloop=100;
   int nmu=0;
+  int maxlat=32;
   for(int mu=0;mu<Nd;mu++) if (mpi_layout[mu]>1) nmu++;
 
+  std::cout << GridLogMessage << "Number of iterations to average: "<< Nloop << std::endl;
+  std::vector<double> t_time(Nloop);
+  time_statistics timestat;
+
   std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
   std::cout<<GridLogMessage << "= Benchmarking concurrent halo exchange in "<<nmu<<" dimensions"<<std::endl;
   std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
-  std::cout<<GridLogMessage << "  L  "<<"\t\t"<<" Ls  "<<"\t\t"<<"bytes"<<"\t\t"<<"MB/s uni"<<"\t\t"<<"MB/s bidi"<<std::endl;
-  int maxlat=16;
-  for(int lat=4;lat<=maxlat;lat+=2){
-    for(int Ls=1;Ls<=16;Ls*=2){
+  header();
+  for(int lat=4;lat<=maxlat;lat+=4){
+    for(int Ls=8;Ls<=8;Ls*=2){
 
       std::vector<int> latt_size  ({lat*mpi_layout[0],
       				    lat*mpi_layout[1],
@@ -58,17 +88,25 @@ int main (int argc, char ** argv)
       				    lat*mpi_layout[3]});
 
       GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
+      RealD Nrank = Grid._Nprocessors;
+      RealD Nnode = Grid.NodeCount();
+      RealD ppn = Nrank/Nnode;
 
-      std::vector<std::vector<HalfSpinColourVectorD> > xbuf(8,std::vector<HalfSpinColourVectorD>(lat*lat*lat*Ls));
-      std::vector<std::vector<HalfSpinColourVectorD> > rbuf(8,std::vector<HalfSpinColourVectorD>(lat*lat*lat*Ls));
+      std::vector<Vector<HalfSpinColourVectorD> > xbuf(8);	
+      std::vector<Vector<HalfSpinColourVectorD> > rbuf(8);
 
       int ncomm;
       int bytes=lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD);
+      for(int mu=0;mu<8;mu++){
+	xbuf[mu].resize(lat*lat*lat*Ls);
+	rbuf[mu].resize(lat*lat*lat*Ls);
+	//	std::cout << " buffers " << std::hex << (uint64_t)&xbuf[mu][0] <<" " << (uint64_t)&rbuf[mu][0] <<std::endl;
+      }
 
-      double start=usecond();
       for(int i=0;i<Nloop;i++){
+      double start=usecond();
 
-	std::vector<CartesianCommunicator::CommsRequest_t> requests;
+	std::vector<CommsRequest_t> requests;
 
 	ncomm=0;
 	for(int mu=0;mu<4;mu++){
@@ -79,7 +117,6 @@ int main (int argc, char ** argv)
 	    int comm_proc=1;
 	    int xmit_to_rank;
 	    int recv_from_rank;
-	    
 	    Grid.ShiftedRanks(mu,comm_proc,xmit_to_rank,recv_from_rank);
 	    Grid.SendToRecvFromBegin(requests,
 				   (void *)&xbuf[mu][0],
@@ -102,18 +139,24 @@ int main (int argc, char ** argv)
 	}
 	Grid.SendToRecvFromComplete(requests);
 	Grid.Barrier();
-
+	double stop=usecond();
+	t_time[i] = stop-start; // microseconds
       }
-      double stop=usecond();
 
-      double dbytes    = bytes;
-      double xbytes    = Nloop*dbytes*2.0*ncomm;
+      timestat.statistics(t_time);
+
+      double dbytes    = bytes*ppn;
+      double xbytes    = dbytes*2.0*ncomm;
       double rbytes    = xbytes;
       double bidibytes = xbytes+rbytes;
 
-      double time = stop-start; // microseconds
+      std::cout<<GridLogMessage << std::setw(4) << lat<<"\t"<<Ls<<"\t"
+               <<std::setw(11) << bytes<< std::fixed << std::setprecision(1) << std::setw(7)
+               <<std::right<< xbytes/timestat.mean<<"  "<< xbytes*timestat.err/(timestat.mean*timestat.mean)<< " "
+               <<xbytes/timestat.max <<" "<< xbytes/timestat.min  
+               << "\t\t"<<std::setw(7)<< bidibytes/timestat.mean<< "  " << bidibytes*timestat.err/(timestat.mean*timestat.mean) << " "
+               << bidibytes/timestat.max << " " << bidibytes/timestat.min << std::endl;
 
-      std::cout<<GridLogMessage << lat<<"\t\t"<<Ls<<"\t\t"<<bytes<<"\t\t"<<xbytes/time<<"\t\t"<<bidibytes/time<<std::endl;
     }
   }    
 
@@ -121,25 +164,32 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
   std::cout<<GridLogMessage << "= Benchmarking sequential halo exchange in "<<nmu<<" dimensions"<<std::endl;
   std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
-  std::cout<<GridLogMessage << "  L  "<<"\t\t"<<" Ls  "<<"\t\t"<<"bytes"<<"\t\t"<<"MB/s uni"<<"\t\t"<<"MB/s bidi"<<std::endl;
+  header();
 
-
-  for(int lat=4;lat<=maxlat;lat+=2){
-    for(int Ls=1;Ls<=16;Ls*=2){
+  for(int lat=4;lat<=maxlat;lat+=4){
+    for(int Ls=8;Ls<=8;Ls*=2){
 
       std::vector<int> latt_size  ({lat,lat,lat,lat});
 
       GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
+      RealD Nrank = Grid._Nprocessors;
+      RealD Nnode = Grid.NodeCount();
+      RealD ppn = Nrank/Nnode;
 
-      std::vector<std::vector<HalfSpinColourVectorD> > xbuf(8,std::vector<HalfSpinColourVectorD>(lat*lat*lat*Ls));
-      std::vector<std::vector<HalfSpinColourVectorD> > rbuf(8,std::vector<HalfSpinColourVectorD>(lat*lat*lat*Ls));
+      std::vector<Vector<HalfSpinColourVectorD> > xbuf(8);
+      std::vector<Vector<HalfSpinColourVectorD> > rbuf(8);
 
+      for(int mu=0;mu<8;mu++){
+	xbuf[mu].resize(lat*lat*lat*Ls);
+	rbuf[mu].resize(lat*lat*lat*Ls);
+	//	std::cout << " buffers " << std::hex << (uint64_t)&xbuf[mu][0] <<" " << (uint64_t)&rbuf[mu][0] <<std::endl;
+      }
 
       int ncomm;
       int bytes=lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD);
 
-      double start=usecond();
       for(int i=0;i<Nloop;i++){
+      double start=usecond();
     
 	ncomm=0;
 	for(int mu=0;mu<4;mu++){
@@ -152,7 +202,7 @@ int main (int argc, char ** argv)
 	    int recv_from_rank;
 	    
 	    {
-	      std::vector<CartesianCommunicator::CommsRequest_t> requests;
+	      std::vector<CommsRequest_t> requests;
 	      Grid.ShiftedRanks(mu,comm_proc,xmit_to_rank,recv_from_rank);
 	      Grid.SendToRecvFromBegin(requests,
 				       (void *)&xbuf[mu][0],
@@ -165,7 +215,7 @@ int main (int argc, char ** argv)
 
 	    comm_proc = mpi_layout[mu]-1;
 	    {
-	      std::vector<CartesianCommunicator::CommsRequest_t> requests;
+	      std::vector<CommsRequest_t> requests;
 	      Grid.ShiftedRanks(mu,comm_proc,xmit_to_rank,recv_from_rank);
 	      Grid.SendToRecvFromBegin(requests,
 				       (void *)&xbuf[mu+4][0],
@@ -178,30 +228,37 @@ int main (int argc, char ** argv)
 	  }
 	}
 	Grid.Barrier();
+	double stop=usecond();
+	t_time[i] = stop-start; // microseconds
+
       }
 
-      double stop=usecond();
+      timestat.statistics(t_time);
       
-      double dbytes    = bytes;
-      double xbytes    = Nloop*dbytes*2.0*ncomm;
+      double dbytes    = bytes*ppn;
+      double xbytes    = dbytes*2.0*ncomm;
       double rbytes    = xbytes;
       double bidibytes = xbytes+rbytes;
 
-      double time = stop-start;
+    std::cout<<GridLogMessage << std::setw(4) << lat<<"\t"<<Ls<<"\t"
+               <<std::setw(11) << bytes<< std::fixed << std::setprecision(1) << std::setw(7)
+               <<std::right<< xbytes/timestat.mean<<"  "<< xbytes*timestat.err/(timestat.mean*timestat.mean)<< " "
+               <<xbytes/timestat.max <<" "<< xbytes/timestat.min  
+               << "\t\t"<<std::setw(7)<< bidibytes/timestat.mean<< "  " << bidibytes*timestat.err/(timestat.mean*timestat.mean) << " "
+               << bidibytes/timestat.max << " " << bidibytes/timestat.min << std::endl;
 
-      std::cout<<GridLogMessage << lat<<"\t\t"<<Ls<<"\t\t"<<bytes<<"\t\t"<<xbytes/time<<"\t\t"<<bidibytes/time<<std::endl;
+      
     }
   }  
 
 
-  Nloop=100;
   std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
   std::cout<<GridLogMessage << "= Benchmarking concurrent STENCIL halo exchange in "<<nmu<<" dimensions"<<std::endl;
   std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
-  std::cout<<GridLogMessage << "  L  "<<"\t\t"<<" Ls  "<<"\t\t"<<"bytes"<<"\t\t"<<"MB/s uni"<<"\t\t"<<"MB/s bidi"<<std::endl;
+  header();
 
-  for(int lat=4;lat<=maxlat;lat+=2){
-    for(int Ls=1;Ls<=16;Ls*=2){
+  for(int lat=4;lat<=maxlat;lat+=4){
+    for(int Ls=8;Ls<=8;Ls*=2){
 
       std::vector<int> latt_size  ({lat*mpi_layout[0],
       				    lat*mpi_layout[1],
@@ -209,6 +266,9 @@ int main (int argc, char ** argv)
       				    lat*mpi_layout[3]});
 
       GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
+      RealD Nrank = Grid._Nprocessors;
+      RealD Nnode = Grid.NodeCount();
+      RealD ppn = Nrank/Nnode;
 
       std::vector<HalfSpinColourVectorD *> xbuf(8);
       std::vector<HalfSpinColourVectorD *> rbuf(8);
@@ -216,73 +276,86 @@ int main (int argc, char ** argv)
       for(int d=0;d<8;d++){
 	xbuf[d] = (HalfSpinColourVectorD *)Grid.ShmBufferMalloc(lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
 	rbuf[d] = (HalfSpinColourVectorD *)Grid.ShmBufferMalloc(lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
+	bzero((void *)xbuf[d],lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
+	bzero((void *)rbuf[d],lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
       }
 
       int ncomm;
       int bytes=lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD);
 
-      double start=usecond();
+      double dbytes;
       for(int i=0;i<Nloop;i++){
+	double start=usecond();
 
-	std::vector<CartesianCommunicator::CommsRequest_t> requests;
-
+	dbytes=0;
 	ncomm=0;
+
+	std::vector<CommsRequest_t> requests;
+
 	for(int mu=0;mu<4;mu++){
 	
+
 	  if (mpi_layout[mu]>1 ) {
 	  
 	    ncomm++;
 	    int comm_proc=1;
 	    int xmit_to_rank;
 	    int recv_from_rank;
-	    
 	    Grid.ShiftedRanks(mu,comm_proc,xmit_to_rank,recv_from_rank);
-	    Grid.StencilSendToRecvFromBegin(requests,
-					    (void *)&xbuf[mu][0],
-					    xmit_to_rank,
-					    (void *)&rbuf[mu][0],
-					    recv_from_rank,
-					    bytes);
+	    dbytes+=
+	      Grid.StencilSendToRecvFromBegin(requests,
+					      (void *)&xbuf[mu][0],
+					      xmit_to_rank,
+					      (void *)&rbuf[mu][0],
+					      recv_from_rank,
+					      bytes,mu);
 	
 	    comm_proc = mpi_layout[mu]-1;
 	  
 	    Grid.ShiftedRanks(mu,comm_proc,xmit_to_rank,recv_from_rank);
-	    Grid.StencilSendToRecvFromBegin(requests,
-					    (void *)&xbuf[mu+4][0],
-					    xmit_to_rank,
-					    (void *)&rbuf[mu+4][0],
-					    recv_from_rank,
-					    bytes);
+	    dbytes+=
+	      Grid.StencilSendToRecvFromBegin(requests,
+					      (void *)&xbuf[mu+4][0],
+					      xmit_to_rank,
+					      (void *)&rbuf[mu+4][0],
+					      recv_from_rank,
+					      bytes,mu+4);
 	  
 	  }
 	}
-	Grid.StencilSendToRecvFromComplete(requests);
+	Grid.StencilSendToRecvFromComplete(requests,0);
 	Grid.Barrier();
-
+	double stop=usecond();
+	t_time[i] = stop-start; // microseconds
+	
       }
-      double stop=usecond();
 
-      double dbytes    = bytes;
-      double xbytes    = Nloop*dbytes*2.0*ncomm;
-      double rbytes    = xbytes;
-      double bidibytes = xbytes+rbytes;
+      timestat.statistics(t_time);
+
+      dbytes=dbytes*ppn;
+      double xbytes    = dbytes*0.5;
+      double rbytes    = dbytes*0.5;
+      double bidibytes = dbytes;
+
+      std::cout<<GridLogMessage << std::setw(4) << lat<<"\t"<<Ls<<"\t"
+               <<std::setw(11) << bytes<< std::fixed << std::setprecision(1) << std::setw(7)
+               <<std::right<< xbytes/timestat.mean<<"  "<< xbytes*timestat.err/(timestat.mean*timestat.mean)<< " "
+               <<xbytes/timestat.max <<" "<< xbytes/timestat.min  
+               << "\t\t"<<std::setw(7)<< bidibytes/timestat.mean<< "  " << bidibytes*timestat.err/(timestat.mean*timestat.mean) << " "
+               << bidibytes/timestat.max << " " << bidibytes/timestat.min << std::endl;
 
-      double time = stop-start; // microseconds
 
-      std::cout<<GridLogMessage << lat<<"\t\t"<<Ls<<"\t\t"<<bytes<<"\t\t"<<xbytes/time<<"\t\t"<<bidibytes/time<<std::endl;
     }
   }    
 
 
-
-  Nloop=100;
   std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
   std::cout<<GridLogMessage << "= Benchmarking sequential STENCIL halo exchange in "<<nmu<<" dimensions"<<std::endl;
   std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
-  std::cout<<GridLogMessage << "  L  "<<"\t\t"<<" Ls  "<<"\t\t"<<"bytes"<<"\t\t"<<"MB/s uni"<<"\t\t"<<"MB/s bidi"<<std::endl;
+  header();
 
-  for(int lat=4;lat<=maxlat;lat+=2){
-    for(int Ls=1;Ls<=16;Ls*=2){
+  for(int lat=4;lat<=maxlat;lat+=4){
+    for(int Ls=8;Ls<=8;Ls*=2){
 
       std::vector<int> latt_size  ({lat*mpi_layout[0],
       				    lat*mpi_layout[1],
@@ -290,6 +363,9 @@ int main (int argc, char ** argv)
       				    lat*mpi_layout[3]});
 
       GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
+      RealD Nrank = Grid._Nprocessors;
+      RealD Nnode = Grid.NodeCount();
+      RealD ppn = Nrank/Nnode;
 
       std::vector<HalfSpinColourVectorD *> xbuf(8);
       std::vector<HalfSpinColourVectorD *> rbuf(8);
@@ -297,16 +373,18 @@ int main (int argc, char ** argv)
       for(int d=0;d<8;d++){
 	xbuf[d] = (HalfSpinColourVectorD *)Grid.ShmBufferMalloc(lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
 	rbuf[d] = (HalfSpinColourVectorD *)Grid.ShmBufferMalloc(lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
+	bzero((void *)xbuf[d],lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
+	bzero((void *)rbuf[d],lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
       }
 
       int ncomm;
       int bytes=lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD);
-
-      double start=usecond();
+      double dbytes;
       for(int i=0;i<Nloop;i++){
+	double start=usecond();
 
-	std::vector<CartesianCommunicator::CommsRequest_t> requests;
-
+	std::vector<CommsRequest_t> requests;
+	dbytes=0;
 	ncomm=0;
 	for(int mu=0;mu<4;mu++){
 	
@@ -318,44 +396,146 @@ int main (int argc, char ** argv)
 	    int recv_from_rank;
 	    
 	    Grid.ShiftedRanks(mu,comm_proc,xmit_to_rank,recv_from_rank);
-	    Grid.StencilSendToRecvFromBegin(requests,
-					    (void *)&xbuf[mu][0],
-					    xmit_to_rank,
-					    (void *)&rbuf[mu][0],
-					    recv_from_rank,
-					    bytes);
-	    //	    Grid.StencilSendToRecvFromComplete(requests);
-	    //	    requests.resize(0);
+	    dbytes+=
+	      Grid.StencilSendToRecvFromBegin(requests,
+					      (void *)&xbuf[mu][0],
+					      xmit_to_rank,
+					      (void *)&rbuf[mu][0],
+					      recv_from_rank,
+					      bytes,mu);
+	    Grid.StencilSendToRecvFromComplete(requests,mu);
+	    requests.resize(0);
 
 	    comm_proc = mpi_layout[mu]-1;
 	  
 	    Grid.ShiftedRanks(mu,comm_proc,xmit_to_rank,recv_from_rank);
-	    Grid.StencilSendToRecvFromBegin(requests,
-					    (void *)&xbuf[mu+4][0],
-					    xmit_to_rank,
-					    (void *)&rbuf[mu+4][0],
-					    recv_from_rank,
-					    bytes);
-	    Grid.StencilSendToRecvFromComplete(requests);
+	    dbytes+=
+	      Grid.StencilSendToRecvFromBegin(requests,
+					      (void *)&xbuf[mu+4][0],
+					      xmit_to_rank,
+					      (void *)&rbuf[mu+4][0],
+					      recv_from_rank,
+					      bytes,mu+4);
+	    Grid.StencilSendToRecvFromComplete(requests,mu+4);
 	    requests.resize(0);
 	  
 	  }
 	}
 	Grid.Barrier();
-
+	double stop=usecond();
+	t_time[i] = stop-start; // microseconds
+	
       }
-      double stop=usecond();
 
-      double dbytes    = bytes;
-      double xbytes    = Nloop*dbytes*2.0*ncomm;
-      double rbytes    = xbytes;
-      double bidibytes = xbytes+rbytes;
+      timestat.statistics(t_time);
 
-      double time = stop-start; // microseconds
+      dbytes=dbytes*ppn;
+      double xbytes    = dbytes*0.5;
+      double rbytes    = dbytes*0.5;
+      double bidibytes = dbytes;
 
-      std::cout<<GridLogMessage << lat<<"\t\t"<<Ls<<"\t\t"<<bytes<<"\t\t"<<xbytes/time<<"\t\t"<<bidibytes/time<<std::endl;
+
+      std::cout<<GridLogMessage << std::setw(4) << lat<<"\t"<<Ls<<"\t"
+               <<std::setw(11) << bytes<< std::fixed << std::setprecision(1) << std::setw(7)
+               <<std::right<< xbytes/timestat.mean<<"  "<< xbytes*timestat.err/(timestat.mean*timestat.mean)<< " "
+               <<xbytes/timestat.max <<" "<< xbytes/timestat.min  
+               << "\t\t"<<std::setw(7)<< bidibytes/timestat.mean<< "  " << bidibytes*timestat.err/(timestat.mean*timestat.mean) << " "
+               << bidibytes/timestat.max << " " << bidibytes/timestat.min << std::endl;
+ 
     }
   }    
 
+
+
+  std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
+  std::cout<<GridLogMessage << "= Benchmarking threaded STENCIL halo exchange in "<<nmu<<" dimensions"<<std::endl;
+  std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
+  header();
+
+  for(int lat=4;lat<=maxlat;lat+=4){
+    for(int Ls=8;Ls<=8;Ls*=2){
+
+      std::vector<int> latt_size  ({lat*mpi_layout[0],
+      				    lat*mpi_layout[1],
+      				    lat*mpi_layout[2],
+      				    lat*mpi_layout[3]});
+
+      GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
+      RealD Nrank = Grid._Nprocessors;
+      RealD Nnode = Grid.NodeCount();
+      RealD ppn = Nrank/Nnode;
+
+      std::vector<HalfSpinColourVectorD *> xbuf(8);
+      std::vector<HalfSpinColourVectorD *> rbuf(8);
+      Grid.ShmBufferFreeAll();
+      for(int d=0;d<8;d++){
+	xbuf[d] = (HalfSpinColourVectorD *)Grid.ShmBufferMalloc(lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
+	rbuf[d] = (HalfSpinColourVectorD *)Grid.ShmBufferMalloc(lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
+	bzero((void *)xbuf[d],lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
+	bzero((void *)rbuf[d],lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD));
+      }
+
+      int ncomm;
+      int bytes=lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD);
+      double dbytes;
+      for(int i=0;i<Nloop;i++){
+	double start=usecond();
+
+	std::vector<CommsRequest_t> requests;
+	dbytes=0;
+	ncomm=0;
+
+	parallel_for(int dir=0;dir<8;dir++){
+
+	  double tbytes;
+	  int mu =dir % 4;
+
+	  if (mpi_layout[mu]>1 ) {
+	  
+	    ncomm++;
+	    int xmit_to_rank;
+	    int recv_from_rank;
+	    if ( dir == mu ) { 
+	      int comm_proc=1;
+	      Grid.ShiftedRanks(mu,comm_proc,xmit_to_rank,recv_from_rank);
+	    } else { 
+	      int comm_proc = mpi_layout[mu]-1;
+	      Grid.ShiftedRanks(mu,comm_proc,xmit_to_rank,recv_from_rank);
+	    }
+
+	    tbytes= Grid.StencilSendToRecvFrom((void *)&xbuf[dir][0], xmit_to_rank,
+					       (void *)&rbuf[dir][0], recv_from_rank, bytes,dir);
+
+#pragma omp atomic
+	    dbytes+=tbytes;
+	  }
+	}
+	Grid.Barrier();
+	double stop=usecond();
+	t_time[i] = stop-start; // microseconds
+      }
+
+      timestat.statistics(t_time);
+
+      dbytes=dbytes*ppn;
+      double xbytes    = dbytes*0.5;
+      double rbytes    = dbytes*0.5;
+      double bidibytes = dbytes;
+
+
+      std::cout<<GridLogMessage << std::setw(4) << lat<<"\t"<<Ls<<"\t"
+               <<std::setw(11) << bytes<< std::fixed << std::setprecision(1) << std::setw(7)
+               <<std::right<< xbytes/timestat.mean<<"  "<< xbytes*timestat.err/(timestat.mean*timestat.mean)<< " "
+               <<xbytes/timestat.max <<" "<< xbytes/timestat.min  
+               << "\t\t"<<std::setw(7)<< bidibytes/timestat.mean<< "  " << bidibytes*timestat.err/(timestat.mean*timestat.mean) << " "
+               << bidibytes/timestat.max << " " << bidibytes/timestat.min << std::endl;
+ 
+    }
+  }    
+
+  std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
+  std::cout<<GridLogMessage << "= All done; Bye Bye"<<std::endl;
+  std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
+
   Grid_finalize();
 }

benchmarks/Benchmark_dwf.cc

@@ -1,28 +1,22 @@
-    /*************************************************************************************
-
+ /*************************************************************************************
     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>
+    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 */
@@ -48,16 +42,25 @@ typedef WilsonFermion5D<DomainWallVec5dImplR> WilsonFermion5DR;
 typedef WilsonFermion5D<DomainWallVec5dImplF> WilsonFermion5DF;
 typedef WilsonFermion5D<DomainWallVec5dImplD> WilsonFermion5DD;
 
-
 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;
 
   std::vector<int> latt4 = GridDefaultLatt();
-  const int Ls=8;
+  int Ls=16;
+  for(int i=0;i<argc;i++)
+    if(std::string(argv[i]) == "-Ls"){
+      std::stringstream ss(argv[i+1]); ss >> Ls;
+    }
+
+  GridLogLayout();
+
+  long unsigned int single_site_flops = 8*QCD::Nc*(7+16*QCD::Nc);
+
+
   GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplex::Nsimd()),GridDefaultMpi());
   GridRedBlackCartesian * UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
   GridCartesian         * FGrid   = SpaceTimeGrid::makeFiveDimGrid(Ls,UGrid);
@@ -71,35 +74,66 @@ int main (int argc, char ** argv)
 
   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.SeedFixedIntegers(seeds4);
+  std::cout << GridLogMessage << "Initialising 5d RNG" << std::endl;
   GridParallelRNG          RNG5(FGrid);  RNG5.SeedFixedIntegers(seeds5);
+  std::cout << GridLogMessage << "Initialised RNGs" << std::endl;
 
   LatticeFermion src   (FGrid); random(RNG5,src);
+#if 0
+  src = zero;
+  {
+    std::vector<int> 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
+
+
   LatticeFermion result(FGrid); result=zero;
   LatticeFermion    ref(FGrid);    ref=zero;
   LatticeFermion    tmp(FGrid);
   LatticeFermion    err(FGrid);
 
+  std::cout << GridLogMessage << "Drawing gauge field" << std::endl;
   LatticeGaugeField Umu(UGrid); 
-  random(RNG4,Umu);
-
-  LatticeGaugeField Umu5d(FGrid); 
+  SU3::HotConfiguration(RNG4,Umu); 
+  std::cout << GridLogMessage << "Random gauge initialised " << std::endl;
+#if 0
+  Umu=1.0;
+  for(int mu=0;mu<Nd;mu++){
+    LatticeColourMatrix ttmp(UGrid);
+    ttmp = PeekIndex<LorentzIndex>(Umu,mu);
+    //    if (mu !=2 ) ttmp = 0;
+    //    ttmp = ttmp* pow(10.0,mu);
+    PokeIndex<LorentzIndex>(Umu,ttmp,mu);
+  }
+  std::cout << GridLogMessage << "Forced to diagonal " << std::endl;
+#endif
 
+  ////////////////////////////////////
+  // Naive wilson implementation
+  ////////////////////////////////////
   // replicate across fifth dimension
+  LatticeGaugeField Umu5d(FGrid); 
+  std::vector<LatticeColourMatrix> U(4,FGrid);
   for(int ss=0;ss<Umu._grid->oSites();ss++){
     for(int s=0;s<Ls;s++){
       Umu5d._odata[Ls*ss+s] = Umu._odata[ss];
     }
   }
-
-  ////////////////////////////////////
-  // Naive wilson implementation
-  ////////////////////////////////////
-  std::vector<LatticeColourMatrix> U(4,FGrid);
   for(int mu=0;mu<Nd;mu++){
     U[mu] = PeekIndex<LorentzIndex>(Umu5d,mu);
   }
+  std::cout << GridLogMessage << "Setting up Cshift based reference " << std::endl;
 
   if (1)
   {
@@ -120,8 +154,7 @@ int main (int argc, char ** argv)
   RealD M5  =1.8;
 
   RealD NP = UGrid->_Nprocessors;
-
-  DomainWallFermionR Dw(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+  RealD NN = UGrid->NodeCount();
 
   std::cout << GridLogMessage<< "*****************************************************************" <<std::endl;
   std::cout << GridLogMessage<< "* Kernel options --dslash-generic, --dslash-unroll, --dslash-asm" <<std::endl;
@@ -131,15 +164,22 @@ int main (int argc, char ** argv)
   std::cout << GridLogMessage<< "* Vectorising space-time by "<<vComplex::Nsimd()<<std::endl;
   if ( sizeof(Real)==4 )   std::cout << GridLogMessage<< "* SINGLE precision "<<std::endl;
   if ( sizeof(Real)==8 )   std::cout << GridLogMessage<< "* DOUBLE precision "<<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;
 
-  int ncall =100;
+  DomainWallFermionR Dw(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+  int ncall =500;
   if (1) {
     FGrid->Barrier();
     Dw.ZeroCounters();
+    Dw.Dhop(src,result,0);
+    std::cout<<GridLogMessage<<"Called warmup"<<std::endl;
     double t0=usecond();
     for(int i=0;i<ncall;i++){
       __SSC_START;
@@ -150,19 +190,58 @@ int main (int argc, char ** argv)
     FGrid->Barrier();
     
     double volume=Ls;  for(int mu=0;mu<Nd;mu++) volume=volume*latt4[mu];
-    double flops=1344*volume*ncall;
+    double flops=single_site_flops*volume*ncall;
 
     std::cout<<GridLogMessage << "Called Dw "<<ncall<<" times in "<<t1-t0<<" us"<<std::endl;
-    std::cout<<GridLogMessage << "norm result "<< norm2(result)<<std::endl;
-    std::cout<<GridLogMessage << "norm ref    "<< norm2(ref)<<std::endl;
+    //    std::cout<<GridLogMessage << "norm result "<< norm2(result)<<std::endl;
+    //    std::cout<<GridLogMessage << "norm ref    "<< norm2(ref)<<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; 
     std::cout<<GridLogMessage << "norm diff   "<< norm2(err)<<std::endl;
+
+    /*
+    if(( norm2(err)>1.0e-4) ) { 
+      std::cout << "RESULT\n " << result<<std::endl;
+      std::cout << "REF   \n " << ref   <<std::endl;
+      std::cout << "ERR   \n " << err   <<std::endl;
+      FGrid->Barrier();
+      exit(-1);
+    }
+    */
     assert (norm2(err)< 1.0e-4 );
     Dw.Report();
   }
 
+  DomainWallFermionRL DwH(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+  if (1) {
+    FGrid->Barrier();
+    DwH.ZeroCounters();
+    DwH.Dhop(src,result,0);
+    double t0=usecond();
+    for(int i=0;i<ncall;i++){
+      __SSC_START;
+      DwH.Dhop(src,result,0);
+      __SSC_STOP;
+    }
+    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;
+
+    std::cout<<GridLogMessage << "Called half prec comms 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; 
+    std::cout<<GridLogMessage << "norm diff   "<< norm2(err)<<std::endl;
+
+    assert (norm2(err)< 1.0e-3 );
+    DwH.Report();
+  }
+
   if (1)
   {
 
@@ -171,6 +250,10 @@ int main (int argc, char ** argv)
     std::cout << GridLogMessage<< "* Vectorising fifth dimension by "<<vComplex::Nsimd()<<std::endl;
     if ( sizeof(Real)==4 )   std::cout << GridLogMessage<< "* SINGLE precision "<<std::endl;
     if ( sizeof(Real)==8 )   std::cout << GridLogMessage<< "* DOUBLE precision "<<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;
@@ -182,21 +265,13 @@ int main (int argc, char ** argv)
     LatticeFermion sresult(sFGrid);
 
     WilsonFermion5DR sDw(Umu,*sFGrid,*sFrbGrid,*sUGrid,*sUrbGrid,M5);
-  
-    for(int x=0;x<latt4[0];x++){
-    for(int y=0;y<latt4[1];y++){
-    for(int z=0;z<latt4[2];z++){
-    for(int t=0;t<latt4[3];t++){
-    for(int s=0;s<Ls;s++){
-      std::vector<int> site({s,x,y,z,t});
-      SpinColourVector tmp;
-      peekSite(tmp,src,site);
-      pokeSite(tmp,ssrc,site);
-    }}}}}
+
+    localConvert(src,ssrc);
     std::cout<<GridLogMessage<< "src norms "<< norm2(src)<<" " <<norm2(ssrc)<<std::endl;
     FGrid->Barrier();
-    double t0=usecond();
+    sDw.Dhop(ssrc,sresult,0);
     sDw.ZeroCounters();
+    double t0=usecond();
     for(int i=0;i<ncall;i++){
       __SSC_START;
       sDw.Dhop(ssrc,sresult,0);
@@ -205,51 +280,58 @@ int main (int argc, char ** argv)
     double t1=usecond();
     FGrid->Barrier();
     double volume=Ls;  for(int mu=0;mu<Nd;mu++) volume=volume*latt4[mu];
-    double flops=1344*volume*ncall;
+    double flops=single_site_flops*volume*ncall;
 
     std::cout<<GridLogMessage << "Called Dw s_inner "<<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;
+    //    std::cout<<GridLogMessage<< "res norms "<< norm2(result)<<" " <<norm2(sresult)<<std::endl;
     sDw.Report();
-  
-    if(0){
-      for(int i=0;i< PerformanceCounter::NumTypes(); i++ ){
-	sDw.Dhop(ssrc,sresult,0);
-	PerformanceCounter Counter(i);
-	Counter.Start();
-	sDw.Dhop(ssrc,sresult,0);
-	Counter.Stop();
-	Counter.Report();
-      }
-    }
-
-    std::cout<<GridLogMessage<< "res norms "<< norm2(result)<<" " <<norm2(sresult)<<std::endl;
-
     RealD sum=0;
-    for(int x=0;x<latt4[0];x++){
-    for(int y=0;y<latt4[1];y++){
-    for(int z=0;z<latt4[2];z++){
-    for(int t=0;t<latt4[3];t++){
-    for(int s=0;s<Ls;s++){
-      std::vector<int> site({s,x,y,z,t});
-      SpinColourVector normal, simd;
-      peekSite(normal,result,site);
-      peekSite(simd,sresult,site);
-      sum=sum+norm2(normal-simd);
-      if (norm2(normal-simd) > 1.0e-6 ) {
-	std::cout << "site "<<x<<","<<y<<","<<z<<","<<t<<","<<s<<" "<<norm2(normal-simd)<<std::endl;
-	std::cout << "site "<<x<<","<<y<<","<<z<<","<<t<<","<<s<<" normal "<<normal<<std::endl;
-	std::cout << "site "<<x<<","<<y<<","<<z<<","<<t<<","<<s<<" simd   "<<simd<<std::endl;
-      }
-    }}}}}
-    std::cout<<GridLogMessage<<" difference between normal and simd is "<<sum<<std::endl;
-    assert (sum< 1.0e-4 );
 
+    err=zero;
+    localConvert(sresult,err);
+    err = err - ref;
+    sum = norm2(err);
+    std::cout<<GridLogMessage<<" difference between normal ref and simd is "<<sum<<std::endl;
+    if(sum > 1.0e-4 ){
+      std::cout<< "sD REF\n " <<ref << std::endl;
+      std::cout<< "sD ERR   \n " <<err  <<std::endl;
+    }
+    //    assert(sum < 1.0e-4);
 
-    if (1) {
+    err=zero;
+    localConvert(sresult,err);
+    err = err - result;
+    sum = norm2(err);
+    std::cout<<GridLogMessage<<" difference between normal result and simd is "<<sum<<std::endl;
+    if(sum > 1.0e-4 ){
+      std::cout<< "sD REF\n " <<result << std::endl;
+      std::cout<< "sD ERR   \n " << err  <<std::endl;
+    }
+    assert(sum < 1.0e-4);
+
+    
+    if(1){
+      std::cout << GridLogMessage<< "*********************************************************" <<std::endl;
+      std::cout << GridLogMessage<< "* Benchmarking WilsonFermion5D<DomainWallVec5dImplR>::DhopEO "<<std::endl;
+      std::cout << GridLogMessage<< "* Vectorising fifth dimension by "<<vComplex::Nsimd()<<std::endl;
+      if ( sizeof(Real)==4 )   std::cout << GridLogMessage<< "* SINGLE precision "<<std::endl;
+      if ( sizeof(Real)==8 )   std::cout << GridLogMessage<< "* DOUBLE precision "<<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;
 
       LatticeFermion sr_eo(sFGrid);
-
       LatticeFermion ssrc_e (sFrbGrid);
       LatticeFermion ssrc_o (sFrbGrid);
       LatticeFermion sr_e   (sFrbGrid);
@@ -257,39 +339,30 @@ int main (int argc, char ** argv)
 
       pickCheckerboard(Even,ssrc_e,ssrc);
       pickCheckerboard(Odd,ssrc_o,ssrc);
-
-      setCheckerboard(sr_eo,ssrc_o);
-      setCheckerboard(sr_eo,ssrc_e);
+      //      setCheckerboard(sr_eo,ssrc_o);
+      //      setCheckerboard(sr_eo,ssrc_e);
 
       sr_e = zero;
       sr_o = zero;
 
-      std::cout << GridLogMessage<< "*********************************************************" <<std::endl;
-      std::cout << GridLogMessage<< "* Benchmarking WilsonFermion5D<DomainWallVec5dImplR>::DhopEO "<<std::endl;
-      std::cout << GridLogMessage<< "* Vectorising fifth dimension by "<<vComplex::Nsimd()<<std::endl;
-      if ( sizeof(Real)==4 )   std::cout << GridLogMessage<< "* SINGLE precision "<<std::endl;
-      if ( sizeof(Real)==8 )   std::cout << GridLogMessage<< "* DOUBLE precision "<<std::endl;
-      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();
+      sDw.DhopEO(ssrc_o, sr_e, DaggerNo);
       sDw.ZeroCounters();
-      sDw.stat.init("DhopEO");
+      //      sDw.stat.init("DhopEO");
       double t0=usecond();
       for (int i = 0; i < ncall; i++) {
         sDw.DhopEO(ssrc_o, sr_e, DaggerNo);
       }
       double t1=usecond();
       FGrid->Barrier();
-      sDw.stat.print();
+      //      sDw.stat.print();
 
       double volume=Ls;  for(int mu=0;mu<Nd;mu++) volume=volume*latt4[mu];
-      double flops=(1344.0*volume*ncall)/2;
+      double flops=(single_site_flops*volume*ncall)/2.0;
 
       std::cout<<GridLogMessage << "sDeo mflop/s =   "<< flops/(t1-t0)<<std::endl;
       std::cout<<GridLogMessage << "sDeo mflop/s per rank   "<< flops/(t1-t0)/NP<<std::endl;
+      std::cout<<GridLogMessage << "sDeo mflop/s per node   "<< flops/(t1-t0)/NN<<std::endl;
       sDw.Report();
 
       sDw.DhopEO(ssrc_o,sr_e,DaggerNo);
@@ -298,24 +371,43 @@ int main (int argc, char ** argv)
 
       pickCheckerboard(Even,ssrc_e,sresult);
       pickCheckerboard(Odd ,ssrc_o,sresult);
+
       ssrc_e = ssrc_e - sr_e;
       RealD error = norm2(ssrc_e);
-
       std::cout<<GridLogMessage << "sE norm diff   "<< norm2(ssrc_e)<< "  vec nrm"<<norm2(sr_e) <<std::endl;
-      ssrc_o = ssrc_o - sr_o;
 
+      ssrc_o = ssrc_o - sr_o;
       error+= norm2(ssrc_o);
       std::cout<<GridLogMessage << "sO norm diff   "<< norm2(ssrc_o)<< "  vec nrm"<<norm2(sr_o) <<std::endl;
-      if(error>1.0e-4) { 
+
+      if(( error>1.0e-4) ) { 
 	setCheckerboard(ssrc,ssrc_o);
 	setCheckerboard(ssrc,ssrc_e);
-	std::cout<< ssrc << std::endl;
+	std::cout<< "DIFF\n " <<ssrc << std::endl;
+	setCheckerboard(ssrc,sr_o);
+	setCheckerboard(ssrc,sr_e);
+	std::cout<< "CBRESULT\n " <<ssrc << std::endl;
+	std::cout<< "RESULT\n " <<sresult<< std::endl;
       }
+      assert(error<1.0e-4);
     }
 
+  if(0){
+    std::cout << "Single cache warm call to sDw.Dhop " <<std::endl;
+    for(int i=0;i< PerformanceCounter::NumTypes(); i++ ){
+      sDw.Dhop(ssrc,sresult,0);
+      PerformanceCounter Counter(i);
+      Counter.Start();
+      sDw.Dhop(ssrc,sresult,0);
+      Counter.Stop();
+      Counter.Report();
+    }
+  }
 
   }
 
+
+
   if (1)
   { // Naive wilson dag implementation
     ref = zero;
@@ -324,25 +416,30 @@ int main (int argc, char ** argv)
       //    ref =  src - Gamma(Gamma::Algebra::GammaX)* src ; // 1+gamma_x
       tmp = U[mu]*Cshift(src,mu+1,1);
       for(int i=0;i<ref._odata.size();i++){
-  ref._odata[i]+= tmp._odata[i] + Gamma(Gmu[mu])*tmp._odata[i]; ;
+	ref._odata[i]+= tmp._odata[i] + Gamma(Gmu[mu])*tmp._odata[i]; ;
       }
 
       tmp =adj(U[mu])*src;
       tmp =Cshift(tmp,mu+1,-1);
       for(int i=0;i<ref._odata.size();i++){
-  ref._odata[i]+= tmp._odata[i] - Gamma(Gmu[mu])*tmp._odata[i]; ;
+	ref._odata[i]+= tmp._odata[i] - Gamma(Gmu[mu])*tmp._odata[i]; ;
       }
     }
     ref = -0.5*ref;
   }
+  //  dump=1;
   Dw.Dhop(src,result,1);
   std::cout << GridLogMessage << "Compare to naive wilson implementation Dag to verify correctness" << std::endl;
   std::cout<<GridLogMessage << "Called DwDag"<<std::endl;
-  std::cout<<GridLogMessage << "norm result "<< norm2(result)<<std::endl;
-  std::cout<<GridLogMessage << "norm ref    "<< norm2(ref)<<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; 
-  std::cout<<GridLogMessage << "norm diff   "<< norm2(err)<<std::endl;
-  assert(norm2(err)<1.0e-4);
+  std::cout<<GridLogMessage << "norm dag diff   "<< norm2(err)<<std::endl;
+  if((norm2(err)>1.0e-4)){
+	std::cout<< "DAG RESULT\n "  <<ref     << std::endl;
+	std::cout<< "DAG sRESULT\n " <<result  << std::endl;
+	std::cout<< "DAG ERR   \n "  << err    <<std::endl;
+  }
   LatticeFermion src_e (FrbGrid);
   LatticeFermion src_o (FrbGrid);
   LatticeFermion r_e   (FrbGrid);
@@ -350,18 +447,24 @@ int main (int argc, char ** argv)
   LatticeFermion r_eo  (FGrid);
 
 
-  std::cout<<GridLogMessage << "Calling Deo and Doe and assert Deo+Doe == Dunprec"<<std::endl;
+  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 DomainWallFermionR::DhopEO                "<<std::endl;
   std::cout << GridLogMessage<< "* Vectorising space-time by "<<vComplex::Nsimd()<<std::endl;
   if ( sizeof(Real)==4 )   std::cout << GridLogMessage<< "* SINGLE precision "<<std::endl;
   if ( sizeof(Real)==8 )   std::cout << GridLogMessage<< "* DOUBLE precision "<<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;
@@ -369,6 +472,7 @@ int main (int argc, char ** argv)
   {
     Dw.ZeroCounters();
     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);
@@ -377,10 +481,11 @@ int main (int argc, char ** argv)
     FGrid->Barrier();
     
     double volume=Ls;  for(int mu=0;mu<Nd;mu++) volume=volume*latt4[mu];
-    double flops=(1344.0*volume*ncall)/2;
+    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.Report();
   }
   Dw.DhopEO(src_o,r_e,DaggerNo);
@@ -396,14 +501,20 @@ int main (int argc, char ** argv)
 
   err = r_eo-result; 
   std::cout<<GridLogMessage << "norm diff   "<< norm2(err)<<std::endl;
-  assert(norm2(err)<1.0e-4);
+  if((norm2(err)>1.0e-4)){
+	std::cout<< "Deo RESULT\n " <<r_eo << std::endl;
+	std::cout<< "Deo REF\n " <<result  << std::endl;
+	std::cout<< "Deo ERR   \n " << err <<std::endl;
+  }
 
   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);
-
   Grid_finalize();
+  exit(0);
 }
+

benchmarks/Benchmark_dwf_sweep.cc

@@ -51,6 +51,7 @@ int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
+
   std::cout << GridLogMessage<< "*****************************************************************" <<std::endl;
   std::cout << GridLogMessage<< "* Kernel options --dslash-generic, --dslash-unroll, --dslash-asm" <<std::endl;
   std::cout << GridLogMessage<< "*****************************************************************" <<std::endl;
@@ -107,6 +108,7 @@ void benchDw(std::vector<int> & latt4, int Ls, int threads,int report )
   GridRedBlackCartesian * UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
   GridCartesian         * FGrid   = SpaceTimeGrid::makeFiveDimGrid(Ls,UGrid);
   GridRedBlackCartesian * FrbGrid = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGrid);
+  long unsigned int single_site_flops = 8*QCD::Nc*(7+16*QCD::Nc);
 
   std::vector<int> seeds4({1,2,3,4});
   std::vector<int> seeds5({5,6,7,8});
@@ -196,7 +198,7 @@ void benchDw(std::vector<int> & latt4, int Ls, int threads,int report )
   
   if ( ! report ) {
     double volume=Ls;  for(int mu=0;mu<Nd;mu++) volume=volume*latt4[mu];
-    double flops=1344*volume*ncall;
+    double flops=single_site_flops*volume*ncall;
     std::cout <<"\t"<<NP<< "\t"<<flops/(t1-t0)<< "\t";
   }
   
@@ -228,7 +230,7 @@ void benchDw(std::vector<int> & latt4, int Ls, int threads,int report )
     
     if(!report){
       double volume=Ls;  for(int mu=0;mu<Nd;mu++) volume=volume*latt4[mu];
-      double flops=(1344.0*volume*ncall)/2;
+      double flops=(single_site_flops*volume*ncall)/2.0;
       std::cout<< flops/(t1-t0);
     }
   }
@@ -237,6 +239,7 @@ void benchDw(std::vector<int> & latt4, int Ls, int threads,int report )
 #define CHECK_SDW
 void benchsDw(std::vector<int> & latt4, int Ls, int threads, int report )
 {
+  long unsigned int single_site_flops = 8*QCD::Nc*(7+16*QCD::Nc);
 
   GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(latt4, GridDefaultSimd(Nd,vComplex::Nsimd()),GridDefaultMpi());
   GridRedBlackCartesian * UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
@@ -321,7 +324,7 @@ void benchsDw(std::vector<int> & latt4, int Ls, int threads, int report )
     Counter.Report();
   } else { 
     double volume=Ls;  for(int mu=0;mu<Nd;mu++) volume=volume*latt4[mu];
-    double flops=1344*volume*ncall;
+    double flops=single_site_flops*volume*ncall;
     std::cout<<"\t"<< flops/(t1-t0);
   }
 
@@ -358,7 +361,7 @@ void benchsDw(std::vector<int> & latt4, int Ls, int threads, int report )
     CounterSdw.Report();
   } else {
     double volume=Ls;  for(int mu=0;mu<Nd;mu++) volume=volume*latt4[mu];
-    double flops=(1344.0*volume*ncall)/2;
+    double flops=(single_site_flops*volume*ncall)/2.0;
     std::cout<<"\t"<< flops/(t1-t0);
   }
 }

benchmarks/Benchmark_gparity.cc

@@ -0,0 +1,190 @@
+#include <Grid/Grid.h>
+#include <sstream>
+using namespace std;
+using namespace Grid;
+using namespace Grid::QCD;
+
+template<class d>
+struct scal {
+  d internal;
+};
+
+  Gamma::Algebra Gmu [] = {
+    Gamma::Algebra::GammaX,
+    Gamma::Algebra::GammaY,
+    Gamma::Algebra::GammaZ,
+    Gamma::Algebra::GammaT
+  };
+
+typedef typename GparityDomainWallFermionF::FermionField GparityLatticeFermionF;
+typedef typename GparityDomainWallFermionD::FermionField GparityLatticeFermionD;
+
+
+
+int main (int argc, char ** argv)
+{
+  Grid_init(&argc,&argv);
+
+  int Ls=16;
+  for(int i=0;i<argc;i++)
+    if(std::string(argv[i]) == "-Ls"){
+      std::stringstream ss(argv[i+1]); ss >> Ls;
+    }
+
+
+  int threads = GridThread::GetThreads();
+  std::cout<<GridLogMessage << "Grid is setup to use "<<threads<<" threads"<<std::endl;
+  std::cout<<GridLogMessage << "Ls = " << Ls << std::endl;
+
+  std::vector<int> latt4 = GridDefaultLatt();
+
+  GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplexF::Nsimd()),GridDefaultMpi());
+  GridRedBlackCartesian * UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
+  GridCartesian         * FGrid   = SpaceTimeGrid::makeFiveDimGrid(Ls,UGrid);
+  GridRedBlackCartesian * FrbGrid = SpaceTimeGrid::makeFiveDimRedBlackGrid(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.SeedFixedIntegers(seeds4);
+  std::cout << GridLogMessage << "Initialising 5d RNG" << std::endl;
+  GridParallelRNG          RNG5(FGrid);  RNG5.SeedFixedIntegers(seeds5);
+  std::cout << GridLogMessage << "Initialised RNGs" << std::endl;
+
+  GparityLatticeFermionF src   (FGrid); random(RNG5,src);
+  RealD N2 = 1.0/::sqrt(norm2(src));
+  src = src*N2;
+
+  GparityLatticeFermionF result(FGrid); result=zero;
+  GparityLatticeFermionF    ref(FGrid);    ref=zero;
+  GparityLatticeFermionF    tmp(FGrid);
+  GparityLatticeFermionF    err(FGrid);
+
+  std::cout << GridLogMessage << "Drawing gauge field" << std::endl;
+  LatticeGaugeFieldF Umu(UGrid); 
+  SU3::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();
+
+  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 DomainWallFermion::Dhop                  "<<std::endl;
+  std::cout << GridLogMessage<< "* Vectorising space-time by "<<vComplexF::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;
+
+
+
+  std::cout << GridLogMessage<< "* SINGLE/SINGLE"<<std::endl;
+  GparityDomainWallFermionF Dw(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+  int ncall =1000;
+  if (1) {
+    FGrid->Barrier();
+    Dw.ZeroCounters();
+    Dw.Dhop(src,result,0);
+    std::cout<<GridLogMessage<<"Called warmup"<<std::endl;
+    double t0=usecond();
+    for(int i=0;i<ncall;i++){
+      __SSC_START;
+      Dw.Dhop(src,result,0);
+      __SSC_STOP;
+    }
+    double t1=usecond();
+    FGrid->Barrier();
+    
+    double volume=Ls;  for(int mu=0;mu<Nd;mu++) volume=volume*latt4[mu];
+    double flops=2*1320*volume*ncall;
+
+    std::cout<<GridLogMessage << "Called Dw "<<ncall<<" times in "<<t1-t0<<" us"<<std::endl;
+    //    std::cout<<GridLogMessage << "norm result "<< norm2(result)<<std::endl;
+    //    std::cout<<GridLogMessage << "norm ref    "<< norm2(ref)<<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;
+    Dw.Report();
+  }
+
+  std::cout << GridLogMessage<< "* SINGLE/HALF"<<std::endl;
+  GparityDomainWallFermionFH DwH(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+  if (1) {
+    FGrid->Barrier();
+    DwH.ZeroCounters();
+    DwH.Dhop(src,result,0);
+    double t0=usecond();
+    for(int i=0;i<ncall;i++){
+      __SSC_START;
+      DwH.Dhop(src,result,0);
+      __SSC_STOP;
+    }
+    double t1=usecond();
+    FGrid->Barrier();
+    
+    double volume=Ls;  for(int mu=0;mu<Nd;mu++) volume=volume*latt4[mu];
+    double flops=2*1320*volume*ncall;
+
+    std::cout<<GridLogMessage << "Called half prec comms 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;
+    DwH.Report();
+  }
+
+  GridCartesian         * UGrid_d   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplexD::Nsimd()),GridDefaultMpi());
+  GridRedBlackCartesian * UrbGrid_d = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid_d);
+  GridCartesian         * FGrid_d   = SpaceTimeGrid::makeFiveDimGrid(Ls,UGrid_d);
+  GridRedBlackCartesian * FrbGrid_d = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGrid_d);
+
+  
+  std::cout << GridLogMessage<< "* DOUBLE/DOUBLE"<<std::endl;
+  GparityLatticeFermionD src_d(FGrid_d);
+  precisionChange(src_d,src);
+
+  LatticeGaugeFieldD Umu_d(UGrid_d); 
+  precisionChange(Umu_d,Umu);
+
+  GparityLatticeFermionD result_d(FGrid_d);
+
+  GparityDomainWallFermionD DwD(Umu_d,*FGrid_d,*FrbGrid_d,*UGrid_d,*UrbGrid_d,mass,M5);
+  if (1) {
+    FGrid_d->Barrier();
+    DwD.ZeroCounters();
+    DwD.Dhop(src_d,result_d,0);
+    std::cout<<GridLogMessage<<"Called warmup"<<std::endl;
+    double t0=usecond();
+    for(int i=0;i<ncall;i++){
+      __SSC_START;
+      DwD.Dhop(src_d,result_d,0);
+      __SSC_STOP;
+    }
+    double t1=usecond();
+    FGrid_d->Barrier();
+    
+    double volume=Ls;  for(int mu=0;mu<Nd;mu++) volume=volume*latt4[mu];
+    double flops=2*1320*volume*ncall;
+
+    std::cout<<GridLogMessage << "Called Dw "<<ncall<<" times in "<<t1-t0<<" us"<<std::endl;
+    //    std::cout<<GridLogMessage << "norm result "<< norm2(result)<<std::endl;
+    //    std::cout<<GridLogMessage << "norm ref    "<< norm2(ref)<<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;
+    DwD.Report();
+  }
+
+  Grid_finalize();
+}
+

benchmarks/Benchmark_memory_asynch.cc

@@ -66,7 +66,8 @@ int main (int argc, char ** argv)
 
     Vec tsum; tsum = zero;
 
-    GridParallelRNG          pRNG(&Grid);      pRNG.SeedRandomDevice();
+    GridParallelRNG          pRNG(&Grid);      
+    pRNG.SeedFixedIntegers(std::vector<int>({56,17,89,101}));
 
     std::vector<double> stop(threads);
     Vector<Vec> sum(threads);
@@ -77,8 +78,7 @@ int main (int argc, char ** argv)
     }
 
     double start=usecond();
-PARALLEL_FOR_LOOP
-    for(int t=0;t<threads;t++){
+    parallel_for(int t=0;t<threads;t++){
 
       sum[t] = x[t]._odata[0];
       for(int i=0;i<Nloop;i++){

benchmarks/Benchmark_memory_bandwidth.cc

@@ -55,21 +55,21 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
   std::cout<<GridLogMessage << "  L  "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s"<<"\t\t"<<"Gflop/s"<<"\t\t seconds"<<std::endl;
   std::cout<<GridLogMessage << "----------------------------------------------------------"<<std::endl;
-  uint64_t lmax=44;
-#define NLOOP (1*lmax*lmax*lmax*lmax/vol)
-  for(int lat=4;lat<=lmax;lat+=4){
+  uint64_t lmax=64;
+#define NLOOP (10*lmax*lmax*lmax*lmax/vol)
+  for(int lat=8;lat<=lmax;lat+=8){
 
       std::vector<int> latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
-      int vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
+      int64_t vol= latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
       GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
 
       uint64_t Nloop=NLOOP;
 
-      //      GridParallelRNG          pRNG(&Grid);      pRNG.SeedRandomDevice();
+      //      GridParallelRNG          pRNG(&Grid);      pRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9}));
 
-      LatticeVec z(&Grid); //random(pRNG,z);
-      LatticeVec x(&Grid); //random(pRNG,x);
-      LatticeVec y(&Grid); //random(pRNG,y);
+      LatticeVec z(&Grid);// random(pRNG,z);
+      LatticeVec x(&Grid);// random(pRNG,x);
+      LatticeVec y(&Grid);// random(pRNG,y);
       double a=2.0;
 
 
@@ -83,7 +83,7 @@ int main (int argc, char ** argv)
       double time = (stop-start)/Nloop*1000;
       
       double flops=vol*Nvec*2;// mul,add
-      double bytes=3*vol*Nvec*sizeof(Real);
+      double bytes=3.0*vol*Nvec*sizeof(Real);
       std::cout<<GridLogMessage<<std::setprecision(3) << lat<<"\t\t"<<bytes<<"   \t\t"<<bytes/time<<"\t\t"<<flops/time<<"\t\t"<<(stop-start)/1000./1000.<<std::endl;
 
     }
@@ -94,17 +94,17 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "  L  "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s"<<"\t\t"<<"Gflop/s"<<"\t\t seconds"<<std::endl;
   std::cout<<GridLogMessage << "----------------------------------------------------------"<<std::endl;
   
-  for(int lat=4;lat<=lmax;lat+=4){
+  for(int lat=8;lat<=lmax;lat+=8){
 
       std::vector<int> latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
-      int vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
+      int64_t vol= latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
       GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
 
-      //      GridParallelRNG          pRNG(&Grid);      pRNG.SeedRandomDevice();
+      //      GridParallelRNG          pRNG(&Grid);      pRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9}));
 
-      LatticeVec z(&Grid); //random(pRNG,z);
-      LatticeVec x(&Grid); //random(pRNG,x);
-      LatticeVec y(&Grid); //random(pRNG,y);
+      LatticeVec z(&Grid);// random(pRNG,z);
+      LatticeVec x(&Grid);// random(pRNG,x);
+      LatticeVec y(&Grid);// random(pRNG,y);
       double a=2.0;
 
       uint64_t Nloop=NLOOP;
@@ -119,7 +119,7 @@ int main (int argc, char ** argv)
       double time = (stop-start)/Nloop*1000;
      
       double flops=vol*Nvec*2;// mul,add
-      double bytes=3*vol*Nvec*sizeof(Real);
+      double bytes=3.0*vol*Nvec*sizeof(Real);
       std::cout<<GridLogMessage<<std::setprecision(3) << lat<<"\t\t"<<bytes<<"   \t\t"<<bytes/time<<"\t\t"<<flops/time<<"\t\t"<<(stop-start)/1000./1000.<<std::endl;
 
     }
@@ -129,20 +129,20 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
   std::cout<<GridLogMessage << "  L  "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s"<<"\t\t"<<"Gflop/s"<<"\t\t seconds"<<std::endl;
 
-  for(int lat=4;lat<=lmax;lat+=4){
+  for(int lat=8;lat<=lmax;lat+=8){
 
 
       std::vector<int> latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
-      int vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
+      int64_t vol= latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
       uint64_t Nloop=NLOOP;
 
       GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
 
-      //      GridParallelRNG          pRNG(&Grid);      pRNG.SeedRandomDevice();
+      //      GridParallelRNG          pRNG(&Grid);      pRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9}));
 
-      LatticeVec z(&Grid); //random(pRNG,z);
-      LatticeVec x(&Grid); //random(pRNG,x);
-      LatticeVec y(&Grid); //random(pRNG,y);
+      LatticeVec z(&Grid);// random(pRNG,z);
+      LatticeVec x(&Grid);// random(pRNG,x);
+      LatticeVec y(&Grid);// random(pRNG,y);
       RealD a=2.0;
 
 
@@ -154,7 +154,7 @@ int main (int argc, char ** argv)
       double stop=usecond();
       double time = (stop-start)/Nloop*1000;
       
-      double bytes=2*vol*Nvec*sizeof(Real);
+      double bytes=2.0*vol*Nvec*sizeof(Real);
       double flops=vol*Nvec*1;// mul
       std::cout<<GridLogMessage <<std::setprecision(3) << lat<<"\t\t"<<bytes<<"   \t\t"<<bytes/time<<"\t\t"<<flops/time<<"\t\t"<<(stop-start)/1000./1000.<<std::endl;
 
@@ -166,17 +166,17 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "  L  "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s"<<"\t\t"<<"Gflop/s"<<"\t\t seconds"<<std::endl;
   std::cout<<GridLogMessage << "----------------------------------------------------------"<<std::endl;
 
-  for(int lat=4;lat<=lmax;lat+=4){
+  for(int lat=8;lat<=lmax;lat+=8){
 
       std::vector<int> latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
-      int vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
+      int64_t vol= latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
       uint64_t Nloop=NLOOP;
       GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
 
-      //      GridParallelRNG          pRNG(&Grid);      pRNG.SeedRandomDevice();
-      LatticeVec z(&Grid); //random(pRNG,z);
-      LatticeVec x(&Grid); //random(pRNG,x);
-      LatticeVec y(&Grid); //random(pRNG,y);
+      //      GridParallelRNG          pRNG(&Grid);      pRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9}));
+      LatticeVec z(&Grid);// random(pRNG,z);
+      LatticeVec x(&Grid);// random(pRNG,x);
+      LatticeVec y(&Grid);// random(pRNG,y);
       RealD a=2.0;
       Real nn;      
       double start=usecond();
@@ -187,7 +187,7 @@ int main (int argc, char ** argv)
       double stop=usecond();
       double time = (stop-start)/Nloop*1000;
       
-      double bytes=vol*Nvec*sizeof(Real);
+      double bytes=1.0*vol*Nvec*sizeof(Real);
       double flops=vol*Nvec*2;// mul,add
       std::cout<<GridLogMessage<<std::setprecision(3) << lat<<"\t\t"<<bytes<<"  \t\t"<<bytes/time<<"\t\t"<<flops/time<< "\t\t"<<(stop-start)/1000./1000.<< "\t\t " <<std::endl;
 

benchmarks/Benchmark_staggered.cc

@@ -0,0 +1,134 @@
+    /*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./benchmarks/Benchmark_staggered.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>
+
+using namespace std;
+using namespace Grid;
+using namespace Grid::QCD;
+
+int main (int argc, char ** argv)
+{
+  Grid_init(&argc,&argv);
+
+  std::vector<int> latt_size   = GridDefaultLatt();
+  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  std::vector<int> mpi_layout  = GridDefaultMpi();
+  GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
+  GridRedBlackCartesian     RBGrid(&Grid);
+
+  int threads = GridThread::GetThreads();
+  std::cout<<GridLogMessage << "Grid is setup to use "<<threads<<" threads"<<std::endl;
+  std::cout<<GridLogMessage << "Grid floating point word size is REALF"<< sizeof(RealF)<<std::endl;
+  std::cout<<GridLogMessage << "Grid floating point word size is REALD"<< sizeof(RealD)<<std::endl;
+  std::cout<<GridLogMessage << "Grid floating point word size is REAL"<< sizeof(Real)<<std::endl;
+
+  std::vector<int> seeds({1,2,3,4});
+  GridParallelRNG          pRNG(&Grid);
+  pRNG.SeedFixedIntegers(seeds);
+  //  pRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9});
+
+  typedef typename ImprovedStaggeredFermionR::FermionField FermionField; 
+  typename ImprovedStaggeredFermionR::ImplParams params; 
+
+  FermionField src   (&Grid); random(pRNG,src);
+  FermionField result(&Grid); result=zero;
+  FermionField    ref(&Grid);    ref=zero;
+  FermionField    tmp(&Grid);    tmp=zero;
+  FermionField    err(&Grid);    tmp=zero;
+  LatticeGaugeField Umu(&Grid); random(pRNG,Umu);
+  std::vector<LatticeColourMatrix> U(4,&Grid);
+
+  double volume=1;
+  for(int mu=0;mu<Nd;mu++){
+    volume=volume*latt_size[mu];
+  }  
+
+  // Only one non-zero (y)
+#if 0
+  Umu=zero;
+  Complex cone(1.0,0.0);
+  for(int nn=0;nn<Nd;nn++){
+    random(pRNG,U[nn]);
+    if(1) {
+      if (nn!=2) { U[nn]=zero; std::cout<<GridLogMessage << "zeroing gauge field in dir "<<nn<<std::endl; }
+      //      else       { U[nn]= cone;std::cout<<GridLogMessage << "unit gauge field in dir "<<nn<<std::endl; }
+      else       { std::cout<<GridLogMessage << "random gauge field in dir "<<nn<<std::endl; }
+    }
+    PokeIndex<LorentzIndex>(Umu,U[nn],nn);
+  }
+#endif
+
+  for(int mu=0;mu<Nd;mu++){
+    U[mu] = PeekIndex<LorentzIndex>(Umu,mu);
+  }
+  ref = zero;
+  /*  
+  { // Naive wilson implementation
+    ref = zero;
+    for(int mu=0;mu<Nd;mu++){
+      //    ref =  src + Gamma(Gamma::GammaX)* src ; // 1-gamma_x
+      tmp = U[mu]*Cshift(src,mu,1);
+      for(int i=0;i<ref._odata.size();i++){
+	ref._odata[i]+= tmp._odata[i] - Gamma(Gmu[mu])*tmp._odata[i]; ;
+      }
+
+      tmp =adj(U[mu])*src;
+      tmp =Cshift(tmp,mu,-1);
+      for(int i=0;i<ref._odata.size();i++){
+	ref._odata[i]+= tmp._odata[i] + Gamma(Gmu[mu])*tmp._odata[i]; ;
+      }
+    }
+  }
+  ref = -0.5*ref;
+  */
+
+  RealD mass=0.1;
+  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,params);
+  
+  std::cout<<GridLogMessage << "Calling Ds"<<std::endl;
+  int ncall=1000;
+  double t0=usecond();
+  for(int i=0;i<ncall;i++){
+    Ds.Dhop(src,result,0);
+  }
+  double t1=usecond();
+  double flops=(16*(3*(6+8+8)) + 15*3*2)*volume*ncall; // == 66*16 +  == 1146
+  
+  std::cout<<GridLogMessage << "Called Ds"<<std::endl;
+  std::cout<<GridLogMessage << "norm result "<< norm2(result)<<std::endl;
+  std::cout<<GridLogMessage << "norm ref    "<< norm2(ref)<<std::endl;
+  std::cout<<GridLogMessage << "mflop/s =   "<< flops/(t1-t0)<<std::endl;
+  err = ref-result; 
+  std::cout<<GridLogMessage << "norm diff   "<< norm2(err)<<std::endl;
+
+  Grid_finalize();
+}

benchmarks/Benchmark_su3.cc

@@ -35,13 +35,16 @@ using namespace Grid::QCD;
 int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
+#define LMAX (32)
+#define LMIN (16)
+#define LINC (4)
 
-  int Nloop=1000;
+  int64_t Nloop=2000;
 
   std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
   std::vector<int> mpi_layout  = GridDefaultMpi();
 
-  int threads = GridThread::GetThreads();
+  int64_t threads = GridThread::GetThreads();
   std::cout<<GridLogMessage << "Grid is setup to use "<<threads<<" threads"<<std::endl;
 
   std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
@@ -50,19 +53,19 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "  L  "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s\t\t GFlop/s"<<std::endl;
   std::cout<<GridLogMessage << "----------------------------------------------------------"<<std::endl;
 
-  for(int lat=2;lat<=32;lat+=2){
+  for(int lat=LMIN;lat<=LMAX;lat+=LINC){
 
       std::vector<int> latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
-      int vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
+      int64_t vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
       GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
-      //      GridParallelRNG          pRNG(&Grid);      pRNG.SeedRandomDevice();
+      GridParallelRNG          pRNG(&Grid);      pRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9}));
 
-      LatticeColourMatrix z(&Grid);// random(pRNG,z);
-      LatticeColourMatrix x(&Grid);// random(pRNG,x);
-      LatticeColourMatrix y(&Grid);// random(pRNG,y);
+      LatticeColourMatrix z(&Grid); random(pRNG,z);
+      LatticeColourMatrix x(&Grid); random(pRNG,x);
+      LatticeColourMatrix y(&Grid); random(pRNG,y);
 
       double start=usecond();
-      for(int i=0;i<Nloop;i++){
+      for(int64_t i=0;i<Nloop;i++){
 	x=x*y;
       }
       double stop=usecond();
@@ -82,20 +85,20 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "  L  "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s\t\t GFlop/s"<<std::endl;
   std::cout<<GridLogMessage << "----------------------------------------------------------"<<std::endl;
 
-  for(int lat=2;lat<=32;lat+=2){
+  for(int lat=LMIN;lat<=LMAX;lat+=LINC){
 
       std::vector<int> latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
-      int vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
+      int64_t vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
 
       GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
-      //      GridParallelRNG          pRNG(&Grid);      pRNG.SeedRandomDevice();
+      GridParallelRNG          pRNG(&Grid);      pRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9}));
 
-      LatticeColourMatrix z(&Grid); //random(pRNG,z);
-      LatticeColourMatrix x(&Grid); //random(pRNG,x);
-      LatticeColourMatrix y(&Grid); //random(pRNG,y);
+      LatticeColourMatrix z(&Grid); random(pRNG,z);
+      LatticeColourMatrix x(&Grid); random(pRNG,x);
+      LatticeColourMatrix y(&Grid); random(pRNG,y);
 
       double start=usecond();
-      for(int i=0;i<Nloop;i++){
+      for(int64_t i=0;i<Nloop;i++){
 	z=x*y;
       }
       double stop=usecond();
@@ -113,20 +116,20 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "  L  "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s\t\t GFlop/s"<<std::endl;
   std::cout<<GridLogMessage << "----------------------------------------------------------"<<std::endl;
 
-  for(int lat=2;lat<=32;lat+=2){
+  for(int lat=LMIN;lat<=LMAX;lat+=LINC){
 
       std::vector<int> latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
-      int vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
+      int64_t vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
 
       GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
-      //      GridParallelRNG          pRNG(&Grid);      pRNG.SeedRandomDevice();
+      GridParallelRNG          pRNG(&Grid);      pRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9}));
 
-      LatticeColourMatrix z(&Grid); //random(pRNG,z);
-      LatticeColourMatrix x(&Grid); //random(pRNG,x);
-      LatticeColourMatrix y(&Grid); //random(pRNG,y);
+      LatticeColourMatrix z(&Grid); random(pRNG,z);
+      LatticeColourMatrix x(&Grid); random(pRNG,x);
+      LatticeColourMatrix y(&Grid); random(pRNG,y);
 
       double start=usecond();
-      for(int i=0;i<Nloop;i++){
+      for(int64_t i=0;i<Nloop;i++){
 	mult(z,x,y);
       }
       double stop=usecond();
@@ -144,30 +147,107 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "  L  "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s\t\t GFlop/s"<<std::endl;
   std::cout<<GridLogMessage << "----------------------------------------------------------"<<std::endl;
 
-  for(int lat=2;lat<=32;lat+=2){
+  for(int lat=LMIN;lat<=LMAX;lat+=LINC){
+    
+    std::vector<int> latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
+    int64_t vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
+    
+    GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
+    GridParallelRNG          pRNG(&Grid);      pRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9}));
+    
+    LatticeColourMatrix z(&Grid); random(pRNG,z);
+    LatticeColourMatrix x(&Grid); random(pRNG,x);
+    LatticeColourMatrix y(&Grid); random(pRNG,y);
+    
+    double start=usecond();
+    for(int64_t i=0;i<Nloop;i++){
+      mac(z,x,y);
+    }
+    double stop=usecond();
+    double time = (stop-start)/Nloop*1000.0;
+    
+    double bytes=3*vol*Nc*Nc*sizeof(Complex);
+    double flops=Nc*Nc*(6+8+8)*vol;
+    std::cout<<GridLogMessage<<std::setprecision(3) << lat<<"\t\t"<<bytes<<"   \t\t"<<bytes/time<<"\t\t" << flops/time<<std::endl;
+    
+  }
+
+  std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
+  std::cout<<GridLogMessage << "= Benchmarking SU3xSU3  CovShiftForward(z,x,y)"<<std::endl;
+  std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
+  std::cout<<GridLogMessage << "  L  "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s\t\t GFlop/s"<<std::endl;
+  std::cout<<GridLogMessage << "----------------------------------------------------------"<<std::endl;
+
+  for(int lat=LMIN;lat<=LMAX;lat+=LINC){
 
       std::vector<int> latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
-      int vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
+      int64_t vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
 
       GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
-      //      GridParallelRNG          pRNG(&Grid);      pRNG.SeedRandomDevice();
+      GridParallelRNG          pRNG(&Grid);      pRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9}));
 
-      LatticeColourMatrix z(&Grid); //random(pRNG,z);
-      LatticeColourMatrix x(&Grid); //random(pRNG,x);
-      LatticeColourMatrix y(&Grid); //random(pRNG,y);
+      LatticeColourMatrix z(&Grid); random(pRNG,z);
+      LatticeColourMatrix x(&Grid); random(pRNG,x);
+      LatticeColourMatrix y(&Grid); random(pRNG,y);
 
-      double start=usecond();
-      for(int i=0;i<Nloop;i++){
-	mac(z,x,y);
+      for(int mu=0;mu<4;mu++){
+	      double start=usecond();
+	      for(int64_t i=0;i<Nloop;i++){
+	        z = PeriodicBC::CovShiftForward(x,mu,y);
+	    }
+	    double stop=usecond();
+	    double time = (stop-start)/Nloop*1000.0;
+	
+	
+	    double bytes=3*vol*Nc*Nc*sizeof(Complex);
+	    double flops=Nc*Nc*(6+8+8)*vol;
+	    std::cout<<GridLogMessage<<std::setprecision(3) << lat<<"\t\t"<<bytes<<"   \t\t"<<bytes/time<<"\t\t" << flops/time<<std::endl;
       }
-      double stop=usecond();
-      double time = (stop-start)/Nloop*1000.0;
-      
-      double bytes=3*vol*Nc*Nc*sizeof(Complex);
-      double flops=Nc*Nc*(8+8+8)*vol;
-      std::cout<<GridLogMessage<<std::setprecision(3) << lat<<"\t\t"<<bytes<<"   \t\t"<<bytes/time<<"\t\t" << flops/time<<std::endl;
+  }
+#if 1
+  std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
+  std::cout<<GridLogMessage << "= Benchmarking SU3xSU3  z= x * Cshift(y)"<<std::endl;
+  std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
+  std::cout<<GridLogMessage << "  L  "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s\t\t GFlop/s"<<std::endl;
+  std::cout<<GridLogMessage << "----------------------------------------------------------"<<std::endl;
 
+  for(int lat=LMIN;lat<=LMAX;lat+=LINC){
+      std::vector<int> latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
+      int64_t vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
+
+      GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
+      GridParallelRNG          pRNG(&Grid);      pRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9}));
+
+      LatticeColourMatrix z(&Grid); random(pRNG,z);
+      LatticeColourMatrix x(&Grid); random(pRNG,x);
+      LatticeColourMatrix y(&Grid); random(pRNG,y);
+      LatticeColourMatrix tmp(&Grid);
+
+      for(int mu=0;mu<4;mu++){
+	double tshift=0;
+	double tmult =0;
+
+	double start=usecond();
+	for(int64_t i=0;i<Nloop;i++){
+	  tshift-=usecond();
+	  tmp = Cshift(y,mu,-1);
+	  tshift+=usecond();
+	  tmult-=usecond();
+	  z   = x*tmp;
+	  tmult+=usecond();
+	}
+	double stop=usecond();
+	double time = (stop-start)/Nloop;
+	tshift = tshift/Nloop;
+	tmult  = tmult /Nloop;
+	
+	double bytes=3*vol*Nc*Nc*sizeof(Complex);
+	double flops=Nc*Nc*(6+8+8)*vol;
+	std::cout<<GridLogMessage<<std::setprecision(3) << "total us "<<time<<" shift "<<tshift <<" mult "<<tmult<<std::endl;
+	time = time * 1000; // convert to NS for GB/s
+	std::cout<<GridLogMessage<<std::setprecision(3) << lat<<"\t\t"<<bytes<<"   \t\t"<<bytes/time<<"\t\t" << flops/time<<std::endl;
+      }
     }
-
+#endif
   Grid_finalize();
 }

benchmarks/Benchmark_wilson.cc

@@ -4,7 +4,7 @@
 
     Source file: ./benchmarks/Benchmark_wilson.cc
 
-    Copyright (C) 2015
+    Copyright (C) 2018
 
 Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 Author: paboyle <paboyle@ph.ed.ac.uk>
@@ -32,6 +32,9 @@ using namespace std;
 using namespace Grid;
 using namespace Grid::QCD;
 
+
+#include "Grid/util/Profiling.h"
+
 template<class d>
 struct scal {
   d internal;
@@ -45,6 +48,7 @@ struct scal {
   };
 
 bool overlapComms = false;
+bool perfProfiling = false;
 
 int main (int argc, char ** argv)
 {
@@ -53,23 +57,34 @@ int main (int argc, char ** argv)
   if( GridCmdOptionExists(argv,argv+argc,"--asynch") ){
     overlapComms = true;
   }
+  if( GridCmdOptionExists(argv,argv+argc,"--perf") ){
+    perfProfiling = true;
+  }
+
+  long unsigned int single_site_flops = 8*QCD::Nc*(7+16*QCD::Nc);
+
 
   std::vector<int> latt_size   = GridDefaultLatt();
   std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
   std::vector<int> mpi_layout  = GridDefaultMpi();
   GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
-  GridRedBlackCartesian     RBGrid(latt_size,simd_layout,mpi_layout);
+  GridRedBlackCartesian     RBGrid(&Grid);
 
   int threads = GridThread::GetThreads();
-  std::cout<<GridLogMessage << "Grid is setup to use "<<threads<<" threads"<<std::endl;
+
+  GridLogLayout();
+
   std::cout<<GridLogMessage << "Grid floating point word size is REALF"<< sizeof(RealF)<<std::endl;
   std::cout<<GridLogMessage << "Grid floating point word size is REALD"<< sizeof(RealD)<<std::endl;
   std::cout<<GridLogMessage << "Grid floating point word size is REAL"<< sizeof(Real)<<std::endl;
+  std::cout<<GridLogMessage << "Grid number of colours : "<< QCD::Nc <<std::endl;
+  std::cout<<GridLogMessage << "Benchmarking Wilson operator in the fundamental representation" << std::endl;
+
 
   std::vector<int> seeds({1,2,3,4});
   GridParallelRNG          pRNG(&Grid);
   pRNG.SeedFixedIntegers(seeds);
-  //  pRNG.SeedRandomDevice();
+  //  pRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9});
 
   LatticeFermion src   (&Grid); random(pRNG,src);
   LatticeFermion result(&Grid); result=zero;
@@ -134,9 +149,25 @@ int main (int argc, char ** argv)
     Dw.Dhop(src,result,0);
   }
   double t1=usecond();
-  double flops=1344*volume*ncall;
+  double flops=single_site_flops*volume*ncall;
   
+  if (perfProfiling){
+  std::cout<<GridLogMessage << "Profiling Dw with perf"<<std::endl;
+    
+  System::profile("kernel", [&]() {
+    for(int i=0;i<ncall;i++){
+      Dw.Dhop(src,result,0);
+    }
+  });
+
+  std::cout<<GridLogMessage << "Generated kernel.data"<<std::endl;
+  std::cout<<GridLogMessage << "Use with: perf report -i kernel.data"<<std::endl;
+
+  }
+
+
   std::cout<<GridLogMessage << "Called Dw"<<std::endl;
+  std::cout<<GridLogMessage << "flops per site " << single_site_flops << std::endl;
   std::cout<<GridLogMessage << "norm result "<< norm2(result)<<std::endl;
   std::cout<<GridLogMessage << "norm ref    "<< norm2(ref)<<std::endl;
   std::cout<<GridLogMessage << "mflop/s =   "<< flops/(t1-t0)<<std::endl;

benchmarks/Benchmark_wilson_sweep.cc

@@ -62,6 +62,7 @@ int main (int argc, char ** argv)
   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<< "* Number of colours "<< QCD::Nc <<std::endl;
   std::cout << GridLogMessage<< "* Benchmarking WilsonFermionR::Dhop                  "<<std::endl;
   std::cout << GridLogMessage<< "* Vectorising space-time by "<<vComplex::Nsimd()<<std::endl;
   if ( sizeof(Real)==4 )   std::cout << GridLogMessage<< "* SINGLE precision "<<std::endl;
@@ -69,13 +70,15 @@ int main (int argc, char ** argv)
   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 << "* OpenMP threads       : "<< GridThread::GetThreads() <<std::endl;
+  std::cout << GridLogMessage << "* MPI tasks            : "<< GridCmdVectorIntToString(mpi_layout) << std::endl;
   std::cout << GridLogMessage<< "*****************************************************************" <<std::endl;
 
-  std::cout<<GridLogMessage << "============================================================================="<< std::endl;
-  std::cout<<GridLogMessage << "= Benchmarking Wilson" << std::endl;
-  std::cout<<GridLogMessage << "============================================================================="<< std::endl;
-  std::cout<<GridLogMessage << "Volume\t\t\tWilson/MFLOPs\tWilsonDag/MFLOPs" << std::endl;
-  std::cout<<GridLogMessage << "============================================================================="<< std::endl;
+  std::cout<<GridLogMessage << "================================================================================================="<< std::endl;
+  std::cout<<GridLogMessage << "= Benchmarking Wilson operator in the fundamental representation" << std::endl;
+  std::cout<<GridLogMessage << "================================================================================================="<< std::endl;
+  std::cout<<GridLogMessage << "Volume\t\t\tWilson/MFLOPs\tWilsonDag/MFLOPs\tWilsonEO/MFLOPs\tWilsonDagEO/MFLOPs" << std::endl;
+  std::cout<<GridLogMessage << "================================================================================================="<< std::endl;
 
   int Lmax = 32;
   int dmin = 0;
@@ -93,17 +96,24 @@ int main (int argc, char ** argv)
 	  std::cout << latt_size.back() << "\t\t";
 
 	  GridCartesian           Grid(latt_size,simd_layout,mpi_layout);
-	  GridRedBlackCartesian RBGrid(latt_size,simd_layout,mpi_layout);
+	  GridRedBlackCartesian RBGrid(&Grid);
 
 	  GridParallelRNG  pRNG(&Grid); pRNG.SeedFixedIntegers(seeds);
 	  LatticeGaugeField Umu(&Grid); random(pRNG,Umu);
-	  LatticeFermion    src(&Grid); random(pRNG,src);
-	  LatticeFermion result(&Grid); result=zero;
+	  LatticeFermion        src(&Grid); random(pRNG,src);
+	  LatticeFermion    src_o(&RBGrid); pickCheckerboard(Odd,src_o,src);
+	  LatticeFermion     result(&Grid); result=zero;
+	  LatticeFermion result_e(&RBGrid); result_e=zero;
 
 	  double volume = std::accumulate(latt_size.begin(),latt_size.end(),1,std::multiplies<int>());
 
 	  WilsonFermionR Dw(Umu,Grid,RBGrid,mass,params);
-      
+
+    // Full operator      
+	  bench_wilson(src,result,Dw,volume,DaggerNo);
+	  bench_wilson(src,result,Dw,volume,DaggerYes);
+    std::cout << "\t";
+    // EO
 	  bench_wilson(src,result,Dw,volume,DaggerNo);
 	  bench_wilson(src,result,Dw,volume,DaggerYes);
 	  std::cout << std::endl;
@@ -122,9 +132,26 @@ void bench_wilson (
 		   int const           dag )
 {
   int ncall    = 1000;
+  long unsigned int single_site_flops = 8*QCD::Nc*(7+16*QCD::Nc);
   double t0    = usecond();
   for(int i=0; i<ncall; i++) { Dw.Dhop(src,result,dag); }
   double t1    = usecond();
-  double flops = 1344 * volume * ncall;
+  double flops = single_site_flops * volume * ncall;
+  std::cout << flops/(t1-t0) << "\t\t";
+}
+
+void bench_wilson_eo (
+		   LatticeFermion &    src,
+		   LatticeFermion & result,
+		   WilsonFermionR &     Dw,
+		   double const     volume,
+		   int const           dag )
+{
+  int ncall    = 1000;
+  long unsigned int single_site_flops = 8*QCD::Nc*(7+16*QCD::Nc);
+  double t0    = usecond();
+  for(int i=0; i<ncall; i++) { Dw.DhopEO(src,result,dag); }
+  double t1    = usecond();
+  double flops = (single_site_flops * volume * ncall)/2.0;
   std::cout << flops/(t1-t0) << "\t\t";
 }

benchmarks/Makefile.am

@@ -1,11 +1,7 @@
 include Make.inc
 
-simple: simple_su3_test.o simple_su3_expr.o simple_simd_test.o
-
-EXTRA_LIBRARIES = libsimple_su3_test.a libsimple_su3_expr.a libsimple_simd_test.a
-
-libsimple_su3_test_a_SOURCES = simple_su3_test.cc
-
-libsimple_su3_expr_a_SOURCES = simple_su3_expr.cc
-
-libsimple_simd_test_a_SOURCES = simple_simd_test.cc
+bench-local: all
+	./Benchmark_su3
+	./Benchmark_memory_bandwidth
+	./Benchmark_wilson
+	./Benchmark_dwf --dslash-unroll

bootstrap.sh

@@ -1,11 +1,9 @@
 #!/usr/bin/env bash
 
-EIGEN_URL='http://bitbucket.org/eigen/eigen/get/3.2.9.tar.bz2'
+EIGEN_URL='http://bitbucket.org/eigen/eigen/get/3.3.3.tar.bz2'
 
 echo "-- deploying Eigen source..."
-wget ${EIGEN_URL} --no-check-certificate
-./scripts/update_eigen.sh `basename ${EIGEN_URL}`
-rm `basename ${EIGEN_URL}`
+wget ${EIGEN_URL} --no-check-certificate && ./scripts/update_eigen.sh `basename ${EIGEN_URL}` && rm `basename ${EIGEN_URL}`
 
 echo '-- generating Make.inc files...'
 ./scripts/filelist

configure.ac

@@ -1,16 +1,23 @@
 AC_PREREQ([2.63])
-AC_INIT([Grid], [0.6.0], [https://github.com/paboyle/Grid], [Grid])
+AC_INIT([Grid], [0.7.0], [https://github.com/paboyle/Grid], [Grid])
 AC_CANONICAL_BUILD
 AC_CANONICAL_HOST
 AC_CANONICAL_TARGET
-AM_INIT_AUTOMAKE(subdir-objects)
+AM_INIT_AUTOMAKE([subdir-objects 1.13])
+AM_EXTRA_RECURSIVE_TARGETS([tests bench])
 AC_CONFIG_MACRO_DIR([m4])
 AC_CONFIG_SRCDIR([lib/Grid.h])
 AC_CONFIG_HEADERS([lib/Config.h],[sed -i 's|PACKAGE_|GRID_|' lib/Config.h])
 m4_ifdef([AM_SILENT_RULES], [AM_SILENT_RULES([yes])])
 
-############### Checks for programs
+################ Get git info
+#AC_REVISION([m4_esyscmd_s([./scripts/configure.commit])])
+
+################ Set flags
+# do not move!
 CXXFLAGS="-O3 $CXXFLAGS"
+
+############### Checks for programs
 AC_PROG_CXX
 AC_PROG_RANLIB
 
@@ -24,12 +31,14 @@ AX_GXX_VERSION
 AC_DEFINE_UNQUOTED([GXX_VERSION],["$GXX_VERSION"],
       [version of g++ that will compile the code])
 
+
+
 ############### Checks for typedefs, structures, and compiler characteristics
 AC_TYPE_SIZE_T
 AC_TYPE_UINT32_T
 AC_TYPE_UINT64_T
 
-############### OpenMP 
+############### OpenMP
 AC_OPENMP
 ac_openmp=no
 if test "${OPENMP_CXXFLAGS}X" != "X"; then
@@ -45,9 +54,14 @@ AC_CHECK_HEADERS(malloc/malloc.h)
 AC_CHECK_HEADERS(malloc.h)
 AC_CHECK_HEADERS(endian.h)
 AC_CHECK_HEADERS(execinfo.h)
+AC_CHECK_HEADERS(numaif.h)
 AC_CHECK_DECLS([ntohll],[], [], [[#include <arpa/inet.h>]])
 AC_CHECK_DECLS([be64toh],[], [], [[#include <arpa/inet.h>]])
 
+############## Standard libraries
+AC_CHECK_LIB([m],[cos])
+AC_CHECK_LIB([stdc++],[abort])
+
 ############### GMP and MPFR
 AC_ARG_WITH([gmp],
     [AS_HELP_STRING([--with-gmp=prefix],
@@ -60,16 +74,23 @@ AC_ARG_WITH([mpfr],
     [AM_CXXFLAGS="-I$with_mpfr/include $AM_CXXFLAGS"]
     [AM_LDFLAGS="-L$with_mpfr/lib $AM_LDFLAGS"])
 
-############### FFTW3 
-AC_ARG_WITH([fftw],    
+############### FFTW3
+AC_ARG_WITH([fftw],
             [AS_HELP_STRING([--with-fftw=prefix],
             [try this for a non-standard install prefix of the FFTW3 library])],
             [AM_CXXFLAGS="-I$with_fftw/include $AM_CXXFLAGS"]
             [AM_LDFLAGS="-L$with_fftw/lib $AM_LDFLAGS"])
 
-############### lapack 
+############### LIME
+AC_ARG_WITH([lime],
+            [AS_HELP_STRING([--with-lime=prefix],
+            [try this for a non-standard install prefix of the LIME library])],
+            [AM_CXXFLAGS="-I$with_lime/include $AM_CXXFLAGS"]
+            [AM_LDFLAGS="-L$with_lime/lib $AM_LDFLAGS"])
+
+############### lapack
 AC_ARG_ENABLE([lapack],
-    [AC_HELP_STRING([--enable-lapack=yes|no|prefix], [enable LAPACK])], 
+    [AC_HELP_STRING([--enable-lapack=yes|no|prefix], [enable LAPACK])],
     [ac_LAPACK=${enable_lapack}], [ac_LAPACK=no])
 
 case ${ac_LAPACK} in
@@ -83,6 +104,18 @@ case ${ac_LAPACK} in
         AC_DEFINE([USE_LAPACK],[1],[use LAPACK]);;
 esac
 
+############### FP16 conversions
+AC_ARG_ENABLE([sfw-fp16],
+    [AC_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)
+      AC_DEFINE([SFW_FP16],[1],[software conversion to fp16]);;
+    no);;
+    *)
+      AC_MSG_ERROR(["SFW FP16 option not supported ${ac_SFW_FP16}"]);;
+esac
+
 ############### MKL
 AC_ARG_ENABLE([mkl],
     [AC_HELP_STRING([--enable-mkl=yes|no|prefix], [enable Intel MKL for LAPACK & FFTW])],
@@ -108,7 +141,7 @@ AC_ARG_WITH([hdf5],
 
 ############### first-touch
 AC_ARG_ENABLE([numa],
-    [AC_HELP_STRING([--enable-numa=yes|no|prefix], [enable first touch numa opt])], 
+    [AC_HELP_STRING([--enable-numa=yes|no|prefix], [enable first touch numa opt])],
     [ac_NUMA=${enable_NUMA}],[ac_NUMA=no])
 
 case ${ac_NUMA} in
@@ -134,8 +167,8 @@ if test "${ac_MKL}x" != "nox"; then
 fi
 
 AC_SEARCH_LIBS([__gmpf_init], [gmp],
-               [AC_SEARCH_LIBS([mpfr_init], [mpfr], 
-                               [AC_DEFINE([HAVE_LIBMPFR], [1], 
+               [AC_SEARCH_LIBS([mpfr_init], [mpfr],
+                               [AC_DEFINE([HAVE_LIBMPFR], [1],
                                           [Define to 1 if you have the `MPFR' library])]
                                [have_mpfr=true], [AC_MSG_ERROR([MPFR library not found])])]
                [AC_DEFINE([HAVE_LIBGMP], [1], [Define to 1 if you have the `GMP' library])]
@@ -144,7 +177,7 @@ AC_SEARCH_LIBS([__gmpf_init], [gmp],
 if test "${ac_LAPACK}x" != "nox"; then
     AC_SEARCH_LIBS([LAPACKE_sbdsdc], [lapack], [],
                    [AC_MSG_ERROR("LAPACK enabled but library not found")])
-fi   
+fi
 
 AC_SEARCH_LIBS([fftw_execute], [fftw3],
                [AC_SEARCH_LIBS([fftwf_execute], [fftw3f], [],
@@ -152,6 +185,23 @@ AC_SEARCH_LIBS([fftw_execute], [fftw3],
                [AC_DEFINE([HAVE_FFTW], [1], [Define to 1 if you have the `FFTW' library])]
                [have_fftw=true])
 
+AC_SEARCH_LIBS([limeCreateReader], [lime],
+               [AC_DEFINE([HAVE_LIME], [1], [Define to 1 if you have the `LIME' library])]
+               [have_lime=true],
+	       [AC_MSG_WARN(C-LIME library was not found in your system.
+In order to use ILGG file format please install or provide the correct path to your installation
+Info at: http://usqcd.jlab.org/usqcd-docs/c-lime/)])
+
+AC_SEARCH_LIBS([crc32], [z],
+               [AC_DEFINE([HAVE_ZLIB], [1], [Define to 1 if you have the `LIBZ' library])]
+               [have_zlib=true] [LIBS="${LIBS} -lz"],
+	       [AC_MSG_ERROR(zlib library was not found in your system.)])
+
+AC_SEARCH_LIBS([move_pages], [numa],
+               [AC_DEFINE([HAVE_LIBNUMA], [1], [Define to 1 if you have the `LIBNUMA' library])]
+               [have_libnuma=true] [LIBS="${LIBS} -lnuma"],
+	       [AC_MSG_WARN(libnuma library was not found in your system. Some optimisations will not apply)])
+
 AC_SEARCH_LIBS([H5Fopen], [hdf5_cpp],
                [AC_DEFINE([HAVE_HDF5], [1], [Define to 1 if you have the `HDF5' library])]
                [have_hdf5=true]
@@ -176,27 +226,38 @@ case ${ax_cv_cxx_compiler_vendor} in
     case ${ac_SIMD} in
       SSE4)
         AC_DEFINE([SSE4],[1],[SSE4 intrinsics])
-        SIMD_FLAGS='-msse4.2';;
+	case ${ac_SFW_FP16} in
+	  yes)
+	  SIMD_FLAGS='-msse4.2';;
+	  no)
+	  SIMD_FLAGS='-msse4.2 -mf16c';;
+	  *)
+          AC_MSG_ERROR(["SFW_FP16 must be either yes or no value ${ac_SFW_FP16} "]);;
+	esac;;
       AVX)
         AC_DEFINE([AVX1],[1],[AVX intrinsics])
-        SIMD_FLAGS='-mavx';;
+        SIMD_FLAGS='-mavx -mf16c';;
       AVXFMA4)
         AC_DEFINE([AVXFMA4],[1],[AVX intrinsics with FMA4])
-        SIMD_FLAGS='-mavx -mfma4';;
+        SIMD_FLAGS='-mavx -mfma4 -mf16c';;
       AVXFMA)
         AC_DEFINE([AVXFMA],[1],[AVX intrinsics with FMA3])
-        SIMD_FLAGS='-mavx -mfma';;
+        SIMD_FLAGS='-mavx -mfma -mf16c';;
       AVX2)
         AC_DEFINE([AVX2],[1],[AVX2 intrinsics])
-        SIMD_FLAGS='-mavx2 -mfma';;
+        SIMD_FLAGS='-mavx2 -mfma -mf16c';;
       AVX512)
         AC_DEFINE([AVX512],[1],[AVX512 intrinsics])
         SIMD_FLAGS='-mavx512f -mavx512pf -mavx512er -mavx512cd';;
+      SKL)
+        AC_DEFINE([AVX512],[1],[AVX512 intrinsics for SkyLake Xeon])
+        SIMD_FLAGS='-march=skylake-avx512';;
       KNC)
         AC_DEFINE([IMCI],[1],[IMCI intrinsics for Knights Corner])
         SIMD_FLAGS='';;
       KNL)
         AC_DEFINE([AVX512],[1],[AVX512 intrinsics])
+        AC_DEFINE([KNL],[1],[Knights landing processor])
         SIMD_FLAGS='-march=knl';;
       GEN)
         AC_DEFINE([GEN],[1],[generic vector code])
@@ -204,6 +265,9 @@ case ${ax_cv_cxx_compiler_vendor} in
                            [generic SIMD vector width (in bytes)])
         SIMD_GEN_WIDTH_MSG=" (width= $ac_gen_simd_width)"
         SIMD_FLAGS='';;
+      NEONv8)
+        AC_DEFINE([NEONV8],[1],[ARMv8 NEON])
+        SIMD_FLAGS='-march=armv8-a';;
       QPX|BGQ)
         AC_DEFINE([QPX],[1],[QPX intrinsics for BG/Q])
         SIMD_FLAGS='';;
@@ -232,6 +296,7 @@ case ${ax_cv_cxx_compiler_vendor} in
         SIMD_FLAGS='';;
       KNL)
         AC_DEFINE([AVX512],[1],[AVX512 intrinsics for Knights Landing])
+        AC_DEFINE([KNL],[1],[Knights landing processor])
         SIMD_FLAGS='-xmic-avx512';;
       GEN)
         AC_DEFINE([GEN],[1],[generic vector code])
@@ -269,10 +334,43 @@ case ${ac_PRECISION} in
      double)
        AC_DEFINE([GRID_DEFAULT_PRECISION_DOUBLE],[1],[GRID_DEFAULT_PRECISION is DOUBLE] )
      ;;
+     *)
+     AC_MSG_ERROR([${ac_PRECISION} unsupported --enable-precision option]);
+     ;;
 esac
 
+######################  Shared memory allocation technique under MPI3
+AC_ARG_ENABLE([shm],[AC_HELP_STRING([--enable-shm=shmopen|hugetlbfs|shmnone],
+              [Select SHM allocation technique])],[ac_SHM=${enable_shm}],[ac_SHM=shmopen])
+
+case ${ac_SHM} in
+
+     shmopen)
+     AC_DEFINE([GRID_MPI3_SHMOPEN],[1],[GRID_MPI3_SHMOPEN] )
+     ;;
+
+     shmnone)
+     AC_DEFINE([GRID_MPI3_SHM_NONE],[1],[GRID_MPI3_SHM_NONE] )
+     ;;
+
+     hugetlbfs)
+     AC_DEFINE([GRID_MPI3_SHMMMAP],[1],[GRID_MPI3_SHMMMAP] )
+     ;;
+
+     *)
+     AC_MSG_ERROR([${ac_SHM} unsupported --enable-shm option]);
+     ;;
+esac
+
+######################  Shared base path for SHMMMAP
+AC_ARG_ENABLE([shmpath],[AC_HELP_STRING([--enable-shmpath=path],
+              [Select SHM mmap base path for hugetlbfs])],
+	      [ac_SHMPATH=${enable_shmpath}],
+	      [ac_SHMPATH=/var/lib/hugetlbfs/pagesize-2MB/])
+AC_DEFINE_UNQUOTED([GRID_SHM_PATH],["$ac_SHMPATH"],[Path to a hugetlbfs filesystem for MMAPing])
+
 ############### communication type selection
-AC_ARG_ENABLE([comms],[AC_HELP_STRING([--enable-comms=none|mpi|mpi-auto|mpi3|mpi3-auto|shmem],
+AC_ARG_ENABLE([comms],[AC_HELP_STRING([--enable-comms=none|mpi|mpi-auto],
               [Select communications])],[ac_COMMS=${enable_comms}],[ac_COMMS=none])
 
 case ${ac_COMMS} in
@@ -280,24 +378,12 @@ case ${ac_COMMS} in
         AC_DEFINE([GRID_COMMS_NONE],[1],[GRID_COMMS_NONE] )
         comms_type='none'
      ;;
-     mpi3l*)
-       AC_DEFINE([GRID_COMMS_MPI3L],[1],[GRID_COMMS_MPI3L] )
-       comms_type='mpi3l'
-     ;;
-     mpi3*)
+     mpi*)
         AC_DEFINE([GRID_COMMS_MPI3],[1],[GRID_COMMS_MPI3] )
         comms_type='mpi3'
      ;;
-     mpi*)
-        AC_DEFINE([GRID_COMMS_MPI],[1],[GRID_COMMS_MPI] )
-        comms_type='mpi'
-     ;;
-     shmem)
-        AC_DEFINE([GRID_COMMS_SHMEM],[1],[GRID_COMMS_SHMEM] )
-        comms_type='shmem'
-     ;;
      *)
-        AC_MSG_ERROR([${ac_COMMS} unsupported --enable-comms option]); 
+        AC_MSG_ERROR([${ac_COMMS} unsupported --enable-comms option]);
      ;;
 esac
 case ${ac_COMMS} in
@@ -315,13 +401,13 @@ esac
 AM_CONDITIONAL(BUILD_COMMS_SHMEM, [ test "${comms_type}X" == "shmemX" ])
 AM_CONDITIONAL(BUILD_COMMS_MPI,   [ test "${comms_type}X" == "mpiX" ])
 AM_CONDITIONAL(BUILD_COMMS_MPI3,  [ test "${comms_type}X" == "mpi3X" ] )
-AM_CONDITIONAL(BUILD_COMMS_MPI3L, [ test "${comms_type}X" == "mpi3lX" ] )
+AM_CONDITIONAL(BUILD_COMMS_MPIT,  [ test "${comms_type}X" == "mpitX" ] )
 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_RNG=${enable_rng}],[ac_RNG=ranlux48])
+	            [ac_RNG=${enable_rng}],[ac_RNG=sitmo])
 
 case ${ac_RNG} in
      ranlux48)
@@ -334,7 +420,7 @@ case ${ac_RNG} in
       AC_DEFINE([RNG_SITMO],[1],[RNG_SITMO] )
      ;;
      *)
-      AC_MSG_ERROR([${ac_RNG} unsupported --enable-rng option]); 
+      AC_MSG_ERROR([${ac_RNG} unsupported --enable-rng option]);
      ;;
 esac
 
@@ -351,7 +437,7 @@ case ${ac_TIMERS} in
       AC_DEFINE([TIMERS_OFF],[1],[TIMERS_OFF] )
      ;;
      *)
-      AC_MSG_ERROR([${ac_TIMERS} unsupported --enable-timers option]); 
+      AC_MSG_ERROR([${ac_TIMERS} unsupported --enable-timers option]);
      ;;
 esac
 
@@ -363,7 +449,7 @@ case ${ac_CHROMA} in
      yes|no)
      ;;
      *)
-       AC_MSG_ERROR([${ac_CHROMA} unsupported --enable-chroma option]); 
+       AC_MSG_ERROR([${ac_CHROMA} unsupported --enable-chroma option]);
      ;;
 esac
 
@@ -384,12 +470,67 @@ DX_INIT_DOXYGEN([$PACKAGE_NAME], [doxygen.cfg])
 
 ############### Ouput
 cwd=`pwd -P`; cd ${srcdir}; abs_srcdir=`pwd -P`; cd ${cwd}
+GRID_CXXFLAGS="$AM_CXXFLAGS $CXXFLAGS"
+GRID_LDFLAGS="$AM_LDFLAGS $LDFLAGS"
+GRID_LIBS=$LIBS
+GRID_SHORT_SHA=`git rev-parse --short HEAD`
+GRID_SHA=`git rev-parse HEAD`
+GRID_BRANCH=`git rev-parse --abbrev-ref HEAD`
 AM_CXXFLAGS="-I${abs_srcdir}/include $AM_CXXFLAGS"
 AM_CFLAGS="-I${abs_srcdir}/include $AM_CFLAGS"
 AM_LDFLAGS="-L${cwd}/lib $AM_LDFLAGS"
 AC_SUBST([AM_CFLAGS])
 AC_SUBST([AM_CXXFLAGS])
 AC_SUBST([AM_LDFLAGS])
+AC_SUBST([GRID_CXXFLAGS])
+AC_SUBST([GRID_LDFLAGS])
+AC_SUBST([GRID_LIBS])
+AC_SUBST([GRID_SHA])
+AC_SUBST([GRID_BRANCH])
+
+git_commit=`cd $srcdir && ./scripts/configure.commit`
+
+echo "~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+Summary of configuration for $PACKAGE v$VERSION
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+----- GIT VERSION -------------------------------------
+$git_commit
+----- PLATFORM ----------------------------------------
+architecture (build)        : $build_cpu
+os (build)                  : $build_os
+architecture (target)       : $target_cpu
+os (target)                 : $target_os
+compiler vendor             : ${ax_cv_cxx_compiler_vendor}
+compiler version            : ${ax_cv_gxx_version}
+----- BUILD OPTIONS -----------------------------------
+SIMD                        : ${ac_SIMD}${SIMD_GEN_WIDTH_MSG}
+Threading                   : ${ac_openmp}
+Communications type         : ${comms_type}
+Shared memory allocator     : ${ac_SHM}
+Shared memory mmap path     : ${ac_SHMPATH}
+Default precision           : ${ac_PRECISION}
+Software FP16 conversion    : ${ac_SFW_FP16}
+RNG choice                  : ${ac_RNG}
+GMP                         : `if test "x$have_gmp" = xtrue; then echo yes; else echo no; fi`
+LAPACK                      : ${ac_LAPACK}
+FFTW                        : `if test "x$have_fftw" = xtrue; then echo yes; else echo no; fi`
+LIME (ILDG support)         : `if test "x$have_lime" = xtrue; then echo yes; else echo no; fi`
+HDF5                        : `if test "x$have_hdf5" = xtrue; then echo yes; else echo no; fi`
+build DOXYGEN documentation : `if test "$DX_FLAG_doc" = '1'; then echo yes; else echo no; fi`
+----- BUILD FLAGS -------------------------------------
+CXXFLAGS:
+`echo ${AM_CXXFLAGS} ${CXXFLAGS} | tr ' ' '\n' | sed 's/^-/    -/g'`
+LDFLAGS:
+`echo ${AM_LDFLAGS} ${LDFLAGS} | tr ' ' '\n' | sed 's/^-/    -/g'`
+LIBS:
+`echo ${LIBS} | tr ' ' '\n' | sed 's/^-/    -/g'`
+-------------------------------------------------------" > grid.configure.summary
+
+GRID_SUMMARY="`cat grid.configure.summary`"
+AM_SUBST_NOTMAKE([GRID_SUMMARY])
+AC_SUBST([GRID_SUMMARY])
+
+AC_CONFIG_FILES([grid-config], [chmod +x grid-config])
 AC_CONFIG_FILES(Makefile)
 AC_CONFIG_FILES(lib/Makefile)
 AC_CONFIG_FILES(tests/Makefile)
@@ -400,42 +541,16 @@ AC_CONFIG_FILES(tests/forces/Makefile)
 AC_CONFIG_FILES(tests/hadrons/Makefile)
 AC_CONFIG_FILES(tests/hmc/Makefile)
 AC_CONFIG_FILES(tests/solver/Makefile)
+AC_CONFIG_FILES(tests/lanczos/Makefile)
+AC_CONFIG_FILES(tests/smearing/Makefile)
 AC_CONFIG_FILES(tests/qdpxx/Makefile)
+AC_CONFIG_FILES(tests/testu01/Makefile)
 AC_CONFIG_FILES(benchmarks/Makefile)
 AC_CONFIG_FILES(extras/Makefile)
 AC_CONFIG_FILES(extras/Hadrons/Makefile)
 AC_OUTPUT
 
-echo "~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-Summary of configuration for $PACKAGE v$VERSION
-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+echo ""
+cat grid.configure.summary
+echo ""
 
------ PLATFORM ----------------------------------------
-architecture (build)        : $build_cpu
-os (build)                  : $build_os
-architecture (target)       : $target_cpu
-os (target)                 : $target_os
-compiler vendor             : ${ax_cv_cxx_compiler_vendor}
-compiler version            : ${ax_cv_gxx_version}
------ BUILD OPTIONS -----------------------------------
-SIMD                        : ${ac_SIMD}${SIMD_GEN_WIDTH_MSG}
-Threading                   : ${ac_openmp} 
-Communications type         : ${comms_type}
-Default precision           : ${ac_PRECISION}
-RNG choice                  : ${ac_RNG} 
-GMP                         : `if test "x$have_gmp" = xtrue; then echo yes; else echo no; fi`
-LAPACK                      : ${ac_LAPACK}
-FFTW                        : `if test "x$have_fftw" = xtrue; then echo yes; else echo no; fi`
-HDF5                        : `if test "x$have_hdf5" = xtrue; then echo yes; else echo no; fi`
-build DOXYGEN documentation : `if test "$DX_FLAG_doc" = '1'; then echo yes; else echo no; fi`
------ BUILD FLAGS -------------------------------------
-CXXFLAGS:
-`echo ${AM_CXXFLAGS} ${CXXFLAGS} | tr ' ' '\n' | sed 's/^-/    -/g'`
-LDFLAGS:
-`echo ${AM_LDFLAGS} ${LDFLAGS} | tr ' ' '\n' | sed 's/^-/    -/g'`
-LIBS:
-`echo ${LIBS} | tr ' ' '\n' | sed 's/^-/    -/g'`
--------------------------------------------------------" > config.summary
-echo ""
-cat config.summary
-echo ""

extras/Hadrons/Application.cc

@@ -4,8 +4,7 @@ Grid physics library, www.github.com/paboyle/Grid
 
 Source file: extras/Hadrons/Application.cc
 
-Copyright (C) 2015
-Copyright (C) 2016
+Copyright (C) 2015-2018
 
 Author: Antonin Portelli <antonin.portelli@me.com>
 
@@ -43,12 +42,7 @@ using namespace Hadrons;
 // constructors ////////////////////////////////////////////////////////////////
 Application::Application(void)
 {
-    LOG(Message) << "Modules available:" << std::endl;
-    auto list = ModuleFactory::getInstance().getBuilderList();
-    for (auto &m: list)
-    {
-        LOG(Message) << "  " << m << std::endl;
-    }
+    initLogger();
     auto dim = GridDefaultLatt(), mpi = GridDefaultMpi(), loc(dim);
     locVol_ = 1;
     for (unsigned int d = 0; d < dim.size(); ++d)
@@ -73,12 +67,6 @@ Application::Application(const std::string parameterFileName)
     parameterFileName_ = parameterFileName;
 }
 
-// environment shortcut ////////////////////////////////////////////////////////
-Environment & Application::env(void) const
-{
-    return Environment::getInstance();
-}
-
 // access //////////////////////////////////////////////////////////////////////
 void Application::setPar(const Application::GlobalPar &par)
 {
@@ -94,14 +82,13 @@ const Application::GlobalPar & Application::getPar(void)
 // execute /////////////////////////////////////////////////////////////////////
 void Application::run(void)
 {
-    if (!parameterFileName_.empty() and (env().getNModule() == 0))
+    if (!parameterFileName_.empty() and (vm().getNModule() == 0))
     {
         parseParameterFile(parameterFileName_);
     }
-    if (!scheduled_)
-    {
-        schedule();
-    }
+    vm().printContent();
+    env().printContent();
+    schedule();
     printSchedule();
     configLoop();
 }
@@ -124,12 +111,20 @@ void Application::parseParameterFile(const std::string parameterFileName)
     LOG(Message) << "Building application from '" << parameterFileName << "'..." << std::endl;
     read(reader, "parameters", par);
     setPar(par);
-    push(reader, "modules");
-    push(reader, "module");
+    if (!push(reader, "modules"))
+    {
+        HADRON_ERROR(Parsing, "Cannot open node 'modules' in parameter file '" 
+                              + parameterFileName + "'");
+    }
+    if (!push(reader, "module"))
+    {
+        HADRON_ERROR(Parsing, "Cannot open node 'modules/module' in parameter file '" 
+                              + parameterFileName + "'");
+    }
     do
     {
         read(reader, "id", id);
-        env().createModule(id.name, id.type, reader);
+        vm().createModule(id.name, id.type, reader);
     } while (reader.nextElement("module"));
     pop(reader);
     pop(reader);
@@ -139,7 +134,7 @@ void Application::saveParameterFile(const std::string parameterFileName)
 {
     XmlWriter          writer(parameterFileName);
     ObjectId           id;
-    const unsigned int nMod = env().getNModule();
+    const unsigned int nMod = vm().getNModule();
     
     LOG(Message) << "Saving application to '" << parameterFileName << "'..." << std::endl;
     write(writer, "parameters", getPar());
@@ -147,10 +142,10 @@ void Application::saveParameterFile(const std::string parameterFileName)
     for (unsigned int i = 0; i < nMod; ++i)
     {
         push(writer, "module");
-        id.name = env().getModuleName(i);
-        id.type = env().getModule(i)->getRegisteredName();
+        id.name = vm().getModuleName(i);
+        id.type = vm().getModule(i)->getRegisteredName();
         write(writer, "id", id);
-        env().getModule(i)->saveParameters(writer, "options");
+        vm().getModule(i)->saveParameters(writer, "options");
         pop(writer);
     }
     pop(writer);
@@ -158,94 +153,13 @@ void Application::saveParameterFile(const std::string parameterFileName)
 }
 
 // schedule computation ////////////////////////////////////////////////////////
-#define MEM_MSG(size)\
-sizeString((size)*locVol_) << " (" << sizeString(size)  << "/site)"
-
-#define DEFINE_MEMPEAK \
-auto memPeak = [this](const std::vector<unsigned int> &program)\
-{\
-    unsigned int memPeak;\
-    bool         msg;\
-    \
-    msg = HadronsLogMessage.isActive();\
-    HadronsLogMessage.Active(false);\
-    env().dryRun(true);\
-    memPeak = env().executeProgram(program);\
-    env().dryRun(false);\
-    env().freeAll();\
-    HadronsLogMessage.Active(true);\
-    \
-    return memPeak;\
-}
-
 void Application::schedule(void)
 {
-    DEFINE_MEMPEAK;
-    
-    // build module dependency graph
-    LOG(Message) << "Building module graph..." << std::endl;
-    auto graph = env().makeModuleGraph();
-    auto con = graph.getConnectedComponents();
-    
-    // constrained topological sort using a genetic algorithm
-    LOG(Message) << "Scheduling computation..." << std::endl;
-    LOG(Message) << "               #module= " << graph.size() << std::endl;
-    LOG(Message) << "       population size= " << par_.genetic.popSize << std::endl;
-    LOG(Message) << "       max. generation= " << par_.genetic.maxGen << std::endl;
-    LOG(Message) << "  max. cst. generation= " << par_.genetic.maxCstGen << std::endl;
-    LOG(Message) << "         mutation rate= " << par_.genetic.mutationRate << std::endl;
-    
-    unsigned int                               k = 0, gen, prevPeak, nCstPeak = 0;
-    std::random_device                         rd;
-    GeneticScheduler<unsigned int>::Parameters par;
-    
-    par.popSize      = par_.genetic.popSize;
-    par.mutationRate = par_.genetic.mutationRate;
-    par.seed         = rd();
-    memPeak_         = 0;
-    CartesianCommunicator::BroadcastWorld(0, &(par.seed), sizeof(par.seed));
-    for (unsigned int i = 0; i < con.size(); ++i)
+    if (!scheduled_ and !loadedSchedule_)
     {
-        GeneticScheduler<unsigned int> scheduler(con[i], memPeak, par);
-        
-        gen = 0;
-        do
-        {
-            LOG(Debug) << "Generation " << gen << ":" << std::endl;
-            scheduler.nextGeneration();
-            if (gen != 0)
-            {
-                if (prevPeak == scheduler.getMinValue())
-                {
-                    nCstPeak++;
-                }
-                else
-                {
-                    nCstPeak = 0;
-                }
-            }
-            
-            prevPeak = scheduler.getMinValue();
-            if (gen % 10 == 0)
-            {
-                LOG(Iterative) << "Generation " << gen << ": "
-                               << MEM_MSG(scheduler.getMinValue()) << std::endl;
-            }
-            
-            gen++;
-        } while ((gen < par_.genetic.maxGen)
-                 and (nCstPeak < par_.genetic.maxCstGen));
-        auto &t = scheduler.getMinSchedule();
-        if (scheduler.getMinValue() > memPeak_)
-        {
-            memPeak_ = scheduler.getMinValue();
-        }
-        for (unsigned int j = 0; j < t.size(); ++j)
-        {
-            program_.push_back(t[j]);
-        }
+        program_   = vm().schedule(par_.genetic);
+        scheduled_ = true;
     }
-    scheduled_ = true;
 }
 
 void Application::saveSchedule(const std::string filename)
@@ -255,21 +169,19 @@ void Application::saveSchedule(const std::string filename)
     
     if (!scheduled_)
     {
-        HADRON_ERROR("Computation not scheduled");
+        HADRON_ERROR(Definition, "Computation not scheduled");
     }
     LOG(Message) << "Saving current schedule to '" << filename << "'..."
                  << std::endl;
     for (auto address: program_)
     {
-        program.push_back(env().getModuleName(address));
+        program.push_back(vm().getModuleName(address));
     }
     write(writer, "schedule", program);
 }
 
 void Application::loadSchedule(const std::string filename)
 {
-    DEFINE_MEMPEAK;
-    
     TextReader               reader(filename);
     std::vector<std::string> program;
     
@@ -279,24 +191,24 @@ void Application::loadSchedule(const std::string filename)
     program_.clear();
     for (auto &name: program)
     {
-        program_.push_back(env().getModuleAddress(name));
+        program_.push_back(vm().getModuleAddress(name));
     }
-    scheduled_ = true;
-    memPeak_   = memPeak(program_);
+    loadedSchedule_ = true;
 }
 
 void Application::printSchedule(void)
 {
     if (!scheduled_)
     {
-        HADRON_ERROR("Computation not scheduled");
+        HADRON_ERROR(Definition, "Computation not scheduled");
     }
-    LOG(Message) << "Schedule (memory peak: " << MEM_MSG(memPeak_) << "):"
+    auto peak = vm().memoryNeeded(program_);
+    LOG(Message) << "Schedule (memory needed: " << sizeString(peak) << "):"
                  << std::endl;
     for (unsigned int i = 0; i < program_.size(); ++i)
     {
         LOG(Message) << std::setw(4) << i + 1 << ": "
-                     << env().getModuleName(program_[i]) << std::endl;
+                     << vm().getModuleName(program_[i]) << std::endl;
     }
 }
 
@@ -309,8 +221,8 @@ void Application::configLoop(void)
     {
         LOG(Message) << BIG_SEP << " Starting measurement for trajectory " << t
                      << " " << BIG_SEP << std::endl;
-        env().setTrajectory(t);
-        env().executeProgram(program_);
+        vm().setTrajectory(t);
+        vm().executeProgram(program_);
     }
     LOG(Message) << BIG_SEP << " End of measurement " << BIG_SEP << std::endl;
     env().freeAll();

extras/Hadrons/Application.hpp

@@ -4,8 +4,7 @@ Grid physics library, www.github.com/paboyle/Grid
 
 Source file: extras/Hadrons/Application.hpp
 
-Copyright (C) 2015
-Copyright (C) 2016
+Copyright (C) 2015-2018
 
 Author: Antonin Portelli <antonin.portelli@me.com>
 
@@ -31,8 +30,7 @@ See the full license in the file "LICENSE" in the top level distribution directo
 #define Hadrons_Application_hpp_
 
 #include <Grid/Hadrons/Global.hpp>
-#include <Grid/Hadrons/Environment.hpp>
-#include <Grid/Hadrons/ModuleFactory.hpp>
+#include <Grid/Hadrons/VirtualMachine.hpp>
 #include <Grid/Hadrons/Modules.hpp>
 
 BEGIN_HADRONS_NAMESPACE
@@ -51,25 +49,13 @@ public:
                                         unsigned int, end,
                                         unsigned int, step);
     };
-    class GeneticPar: Serializable
-    {
-    public:
-        GeneticPar(void):
-            popSize{20}, maxGen{1000}, maxCstGen{100}, mutationRate{.1} {};
-    public:
-        GRID_SERIALIZABLE_CLASS_MEMBERS(GeneticPar,
-                                        unsigned int, popSize,
-                                        unsigned int, maxGen,
-                                        unsigned int, maxCstGen,
-                                        double      , mutationRate);
-    };
     class GlobalPar: Serializable
     {
     public:
         GRID_SERIALIZABLE_CLASS_MEMBERS(GlobalPar,
-                                        TrajRange,   trajCounter,
-                                        GeneticPar,  genetic,
-                                        std::string, seed);
+                                        TrajRange,                  trajCounter,
+                                        VirtualMachine::GeneticPar, genetic,
+                                        std::string,                seed);
     };
 public:
     // constructors
@@ -100,14 +86,15 @@ public:
     void configLoop(void);
 private:
     // environment shortcut
-    Environment & env(void) const;
+    DEFINE_ENV_ALIAS;
+    // virtual machine shortcut
+    DEFINE_VM_ALIAS;
 private:
-    long unsigned int         locVol_;
-    std::string               parameterFileName_{""};
-    GlobalPar                 par_;
-    std::vector<unsigned int> program_;
-    Environment::Size         memPeak_;
-    bool                      scheduled_{false};
+    long unsigned int       locVol_;
+    std::string             parameterFileName_{""};
+    GlobalPar               par_;
+    VirtualMachine::Program program_;
+    bool                    scheduled_{false}, loadedSchedule_{false};
 };
 
 /******************************************************************************
@@ -117,14 +104,16 @@ private:
 template <typename M>
 void Application::createModule(const std::string name)
 {
-    env().createModule<M>(name);
+    vm().createModule<M>(name);
+    scheduled_ = false;
 }
 
 template <typename M>
 void Application::createModule(const std::string name,
                                const typename M::Par &par)
 {
-    env().createModule<M>(name, par);
+    vm().createModule<M>(name, par);
+    scheduled_ = false;
 }
 
 END_HADRONS_NAMESPACE

extras/Hadrons/Environment.cc

@@ -4,8 +4,7 @@ Grid physics library, www.github.com/paboyle/Grid
 
 Source file: extras/Hadrons/Environment.cc
 
-Copyright (C) 2015
-Copyright (C) 2016
+Copyright (C) 2015-2018
 
 Author: Antonin Portelli <antonin.portelli@me.com>
 
@@ -35,15 +34,19 @@ using namespace Grid;
 using namespace QCD;
 using namespace Hadrons;
 
+#define ERROR_NO_ADDRESS(address)\
+HADRON_ERROR(Definition, "no object with address " + std::to_string(address));
+
 /******************************************************************************
  *                       Environment implementation                           *
  ******************************************************************************/
 // constructor /////////////////////////////////////////////////////////////////
 Environment::Environment(void)
 {
-    nd_ = GridDefaultLatt().size();
+    dim_ = GridDefaultLatt();
+    nd_  = dim_.size();
     grid4d_.reset(SpaceTimeGrid::makeFourDimGrid(
-        GridDefaultLatt(), GridDefaultSimd(nd_, vComplex::Nsimd()),
+        dim_, GridDefaultSimd(nd_, vComplex::Nsimd()),
         GridDefaultMpi()));
     gridRb4d_.reset(SpaceTimeGrid::makeFourDimRedBlackGrid(grid4d_.get()));
     auto loc = getGrid()->LocalDimensions();
@@ -55,28 +58,6 @@ Environment::Environment(void)
     rng4d_.reset(new GridParallelRNG(grid4d_.get()));
 }
 
-// dry run /////////////////////////////////////////////////////////////////////
-void Environment::dryRun(const bool isDry)
-{
-    dryRun_ = isDry;
-}
-
-bool Environment::isDryRun(void) const
-{
-    return dryRun_;
-}
-
-// trajectory number ///////////////////////////////////////////////////////////
-void Environment::setTrajectory(const unsigned int traj)
-{
-    traj_ = traj;
-}
-
-unsigned int Environment::getTrajectory(void) const
-{
-    return traj_;
-}
-
 // grids ///////////////////////////////////////////////////////////////////////
 void Environment::createGrid(const unsigned int Ls)
 {
@@ -104,7 +85,7 @@ GridCartesian * Environment::getGrid(const unsigned int Ls) const
     }
     catch(std::out_of_range &)
     {
-        HADRON_ERROR("no grid with Ls= " << Ls);
+        HADRON_ERROR(Definition, "no grid with Ls= " + std::to_string(Ls));
     }
 }
 
@@ -123,7 +104,7 @@ GridRedBlackCartesian * Environment::getRbGrid(const unsigned int Ls) const
     }
     catch(std::out_of_range &)
     {
-        HADRON_ERROR("no red-black 5D grid with Ls= " << Ls);
+        HADRON_ERROR(Definition, "no red-black 5D grid with Ls= " + std::to_string(Ls));
     }
 }
 
@@ -132,6 +113,21 @@ unsigned int Environment::getNd(void) const
     return nd_;
 }
 
+std::vector<int> Environment::getDim(void) const
+{
+    return dim_;
+}
+
+int Environment::getDim(const unsigned int mu) const
+{
+    return dim_[mu];
+}
+
+unsigned long int Environment::getLocalVolume(void) const
+{
+    return locVol_;
+}
+
 // random number generator /////////////////////////////////////////////////////
 void Environment::setSeed(const std::vector<int> &seed)
 {
@@ -143,276 +139,6 @@ GridParallelRNG * Environment::get4dRng(void) const
     return rng4d_.get();
 }
 
-// module management ///////////////////////////////////////////////////////////
-void Environment::pushModule(Environment::ModPt &pt)
-{
-    std::string name = pt->getName();
-    
-    if (!hasModule(name))
-    {
-        std::vector<unsigned int> inputAddress;
-        unsigned int              address;
-        ModuleInfo                m;
-        
-        m.data = std::move(pt);
-        m.type = typeIdPt(*m.data.get());
-        m.name = name;
-        auto input  = m.data->getInput();
-        for (auto &in: input)
-        {
-            if (!hasObject(in))
-            {
-                addObject(in , -1);
-            }
-            m.input.push_back(objectAddress_[in]);
-        }
-        auto output = m.data->getOutput();
-        module_.push_back(std::move(m));
-        address              = static_cast<unsigned int>(module_.size() - 1);
-        moduleAddress_[name] = address;
-        for (auto &out: output)
-        {
-            if (!hasObject(out))
-            {
-                addObject(out, address);
-            }
-            else
-            {
-                if (object_[objectAddress_[out]].module < 0)
-                {
-                    object_[objectAddress_[out]].module = address;
-                }
-                else
-                {
-                    HADRON_ERROR("object '" + out
-                                 + "' is already produced by module '"
-                                 + module_[object_[getObjectAddress(out)].module].name
-                                 + "' (while pushing module '" + name + "')");
-                }
-            }
-        }
-    }
-    else
-    {
-        HADRON_ERROR("module '" + name + "' already exists");
-    }
-}
-
-unsigned int Environment::getNModule(void) const
-{
-    return module_.size();
-}
-
-void Environment::createModule(const std::string name, const std::string type,
-                               XmlReader &reader)
-{
-    auto &factory = ModuleFactory::getInstance();
-    auto pt       = factory.create(type, name);
-    
-    pt->parseParameters(reader, "options");
-    pushModule(pt);
-}
-
-ModuleBase * Environment::getModule(const unsigned int address) const
-{
-    if (hasModule(address))
-    {
-        return module_[address].data.get();
-    }
-    else
-    {
-        HADRON_ERROR("no module with address " + std::to_string(address));
-    }
-}
-
-ModuleBase * Environment::getModule(const std::string name) const
-{
-    return getModule(getModuleAddress(name));
-}
-
-unsigned int Environment::getModuleAddress(const std::string name) const
-{
-    if (hasModule(name))
-    {
-        return moduleAddress_.at(name);
-    }
-    else
-    {
-        HADRON_ERROR("no module with name '" + name + "'");
-    }
-}
-
-std::string Environment::getModuleName(const unsigned int address) const
-{
-    if (hasModule(address))
-    {
-        return module_[address].name;
-    }
-    else
-    {
-        HADRON_ERROR("no module with address " + std::to_string(address));
-    }
-}
-
-std::string Environment::getModuleType(const unsigned int address) const
-{
-    if (hasModule(address))
-    {
-        return typeName(module_[address].type);
-    }
-    else
-    {
-        HADRON_ERROR("no module with address " + std::to_string(address));
-    }
-}
-
-std::string Environment::getModuleType(const std::string name) const
-{
-    return getModuleType(getModuleAddress(name));
-}
-
-bool Environment::hasModule(const unsigned int address) const
-{
-    return (address < module_.size());
-}
-
-bool Environment::hasModule(const std::string name) const
-{
-    return (moduleAddress_.find(name) != moduleAddress_.end());
-}
-
-Graph<unsigned int> Environment::makeModuleGraph(void) const
-{
-    Graph<unsigned int> moduleGraph;
-    
-    for (unsigned int i = 0; i < module_.size(); ++i)
-    {
-        moduleGraph.addVertex(i);
-        for (auto &j: module_[i].input)
-        {
-            moduleGraph.addEdge(object_[j].module, i);
-        }
-    }
-    
-    return moduleGraph;
-}
-
-#define BIG_SEP "==============="
-#define SEP     "---------------"
-#define MEM_MSG(size)\
-sizeString((size)*locVol_) << " (" << sizeString(size)  << "/site)"
-
-Environment::Size
-Environment::executeProgram(const std::vector<unsigned int> &p)
-{
-    Size                                memPeak = 0, sizeBefore, sizeAfter;
-    std::vector<std::set<unsigned int>> freeProg;
-    bool                                continueCollect, nothingFreed;
-    
-    // build garbage collection schedule
-    freeProg.resize(p.size());
-    for (unsigned int i = 0; i < object_.size(); ++i)
-    {
-        auto pred = [i, this](const unsigned int j)
-        {
-            auto &in = module_[j].input;
-            auto it  = std::find(in.begin(), in.end(), i);
-            
-            return (it != in.end()) or (j == object_[i].module);
-        };
-        auto it = std::find_if(p.rbegin(), p.rend(), pred);
-        if (it != p.rend())
-        {
-            freeProg[p.rend() - it - 1].insert(i);
-        }
-    }
-    
-    // program execution
-    for (unsigned int i = 0; i < p.size(); ++i)
-    {
-        // execute module
-        if (!isDryRun())
-        {
-            LOG(Message) << SEP << " Measurement step " << i+1 << "/"
-                         << p.size() << " (module '" << module_[p[i]].name
-                         << "') " << SEP << std::endl;
-        }
-        (*module_[p[i]].data)();
-        sizeBefore = getTotalSize();
-        // print used memory after execution
-        if (!isDryRun())
-        {
-            LOG(Message) << "Allocated objects: " << MEM_MSG(sizeBefore)
-                         << std::endl;
-        }
-        if (sizeBefore > memPeak)
-        {
-            memPeak = sizeBefore;
-        }
-        // garbage collection for step i
-        if (!isDryRun())
-        {
-            LOG(Message) << "Garbage collection..." << std::endl;
-        }
-        nothingFreed = true;
-        do
-        {
-            continueCollect = false;
-            auto toFree = freeProg[i];
-            for (auto &j: toFree)
-            {
-                // continue garbage collection while there are still
-                // objects without owners
-                continueCollect = continueCollect or !hasOwners(j);
-                if(freeObject(j))
-                {
-                    // if an object has been freed, remove it from
-                    // the garbage collection schedule
-                    freeProg[i].erase(j);
-                    nothingFreed = false;
-                }
-            }
-        } while (continueCollect);
-        // any remaining objects in step i garbage collection schedule
-        // is scheduled for step i + 1
-        if (i + 1 < p.size())
-        {
-            for (auto &j: freeProg[i])
-            {
-                freeProg[i + 1].insert(j);
-            }
-        }
-        // print used memory after garbage collection if necessary
-        if (!isDryRun())
-        {
-            sizeAfter = getTotalSize();
-            if (sizeBefore != sizeAfter)
-            {
-                LOG(Message) << "Allocated objects: " << MEM_MSG(sizeAfter)
-                             << std::endl;
-            }
-            else
-            {
-                LOG(Message) << "Nothing to free" << std::endl;
-            }
-        }
-    }
-    
-    return memPeak;
-}
-
-Environment::Size Environment::executeProgram(const std::vector<std::string> &p)
-{
-    std::vector<unsigned int> pAddress;
-    
-    for (auto &n: p)
-    {
-        pAddress.push_back(getModuleAddress(n));
-    }
-    
-    return executeProgram(pAddress);
-}
-
 // general memory management ///////////////////////////////////////////////////
 void Environment::addObject(const std::string name, const int moduleAddress)
 {
@@ -422,46 +148,25 @@ void Environment::addObject(const std::string name, const int moduleAddress)
         
         info.name   = name;
         info.module = moduleAddress;
+        info.data   = nullptr;
         object_.push_back(std::move(info));
         objectAddress_[name] = static_cast<unsigned int>(object_.size() - 1);
     }
     else
     {
-        HADRON_ERROR("object '" + name + "' already exists");
+        HADRON_ERROR(Definition, "object '" + name + "' already exists");
     }
 }
 
-void Environment::registerObject(const unsigned int address,
-                                 const unsigned int size, const unsigned int Ls)
+void Environment::setObjectModule(const unsigned int objAddress,
+                                  const int modAddress)
 {
-    if (!hasRegisteredObject(address))
-    {
-        if (hasObject(address))
-        {
-            object_[address].size         = size;
-            object_[address].Ls           = Ls;
-            object_[address].isRegistered = true;
-        }
-        else
-        {
-            HADRON_ERROR("no object with address " + std::to_string(address));
-        }
-    }
-    else
-    {
-        HADRON_ERROR("object with address " + std::to_string(address)
-                     + " already registered");
-    }
+    object_[objAddress].module = modAddress;
 }
 
-void Environment::registerObject(const std::string name,
-                                 const unsigned int size, const unsigned int Ls)
+unsigned int Environment::getMaxAddress(void) const
 {
-    if (!hasObject(name))
-    {
-        addObject(name);
-    }
-    registerObject(getObjectAddress(name), size, Ls);
+    return object_.size();
 }
 
 unsigned int Environment::getObjectAddress(const std::string name) const
@@ -472,7 +177,7 @@ unsigned int Environment::getObjectAddress(const std::string name) const
     }
     else
     {
-        HADRON_ERROR("no object with name '" + name + "'");
+        HADRON_ERROR(Definition, "no object with name '" + name + "'");
     }
 }
 
@@ -484,24 +189,26 @@ std::string Environment::getObjectName(const unsigned int address) const
     }
     else
     {
-        HADRON_ERROR("no object with address " + std::to_string(address));
+        ERROR_NO_ADDRESS(address);
     }
 }
 
 std::string Environment::getObjectType(const unsigned int address) const
 {
-    if (hasRegisteredObject(address))
+    if (hasObject(address))
     {
-        return typeName(object_[address].type);
-    }
-    else if (hasObject(address))
-    {
-        HADRON_ERROR("object with address " + std::to_string(address)
-                     + " exists but is not registered");
+        if (object_[address].type)
+        {
+            return typeName(object_[address].type);
+        }
+        else
+        {
+            return "<no type>";
+        }
     }
     else
     {
-        HADRON_ERROR("no object with address " + std::to_string(address));
+        ERROR_NO_ADDRESS(address);
     }
 }
 
@@ -512,18 +219,13 @@ std::string Environment::getObjectType(const std::string name) const
 
 Environment::Size Environment::getObjectSize(const unsigned int address) const
 {
-    if (hasRegisteredObject(address))
+    if (hasObject(address))
     {
         return object_[address].size;
     }
-    else if (hasObject(address))
-    {
-        HADRON_ERROR("object with address " + std::to_string(address)
-                     + " exists but is not registered");
-    }
     else
     {
-        HADRON_ERROR("no object with address " + std::to_string(address));
+        ERROR_NO_ADDRESS(address);
     }
 }
 
@@ -532,20 +234,49 @@ Environment::Size Environment::getObjectSize(const std::string name) const
     return getObjectSize(getObjectAddress(name));
 }
 
-unsigned int Environment::getObjectLs(const unsigned int address) const
+Environment::Storage Environment::getObjectStorage(const unsigned int address) const
 {
-    if (hasRegisteredObject(address))
+    if (hasObject(address))
     {
-        return object_[address].Ls;
-    }
-    else if (hasObject(address))
-    {
-        HADRON_ERROR("object with address " + std::to_string(address)
-                     + " exists but is not registered");
+        return object_[address].storage;
     }
     else
     {
-        HADRON_ERROR("no object with address " + std::to_string(address));
+        ERROR_NO_ADDRESS(address);
+    }
+}
+
+Environment::Storage Environment::getObjectStorage(const std::string name) const
+{
+    return getObjectStorage(getObjectAddress(name));
+}
+
+int Environment::getObjectModule(const unsigned int address) const
+{
+    if (hasObject(address))
+    {
+        return object_[address].module;
+    }
+    else
+    {
+        ERROR_NO_ADDRESS(address);
+    }
+}
+
+int Environment::getObjectModule(const std::string name) const
+{
+    return getObjectModule(getObjectAddress(name));
+}
+
+unsigned int Environment::getObjectLs(const unsigned int address) const
+{
+    if (hasObject(address))
+    {
+        return object_[address].Ls;
+    }
+    else
+    {
+        ERROR_NO_ADDRESS(address);
     }
 }
 
@@ -566,30 +297,6 @@ bool Environment::hasObject(const std::string name) const
     return ((it != objectAddress_.end()) and hasObject(it->second));
 }
 
-bool Environment::hasRegisteredObject(const unsigned int address) const
-{
-    if (hasObject(address))
-    {
-        return object_[address].isRegistered;
-    }
-    else
-    {
-        return false;
-    }
-}
-
-bool Environment::hasRegisteredObject(const std::string name) const
-{
-    if (hasObject(name))
-    {
-        return hasRegisteredObject(getObjectAddress(name));
-    }
-    else
-    {
-        return false;
-    }
-}
-
 bool Environment::hasCreatedObject(const unsigned int address) const
 {
     if (hasObject(address))
@@ -630,92 +337,27 @@ Environment::Size Environment::getTotalSize(void) const
     
     for (auto &o: object_)
     {
-        if (o.isRegistered)
-        {
-            size += o.size;
-        }
+        size += o.size;
     }
     
     return size;
 }
 
-void Environment::addOwnership(const unsigned int owner,
-                               const unsigned int property)
+void Environment::freeObject(const unsigned int address)
 {
-    if (hasObject(property))
+    if (hasCreatedObject(address))
     {
-        object_[property].owners.insert(owner);
-    }
-    else
-    {
-        HADRON_ERROR("no object with address " + std::to_string(property));
-    }
-    if (hasObject(owner))
-    {
-        object_[owner].properties.insert(property);
-    }
-    else
-    {
-        HADRON_ERROR("no object with address " + std::to_string(owner));
+        LOG(Message) << "Destroying object '" << object_[address].name
+                     << "'" << std::endl;
     }
+    object_[address].size = 0;
+    object_[address].type = nullptr;
+    object_[address].data.reset(nullptr);
 }
 
-void Environment::addOwnership(const std::string owner,
-                               const std::string property)
+void Environment::freeObject(const std::string name)
 {
-    addOwnership(getObjectAddress(owner), getObjectAddress(property));
-}
-
-bool Environment::hasOwners(const unsigned int address) const
-{
-    
-    if (hasObject(address))
-    {
-        return (!object_[address].owners.empty());
-    }
-    else
-    {
-        HADRON_ERROR("no object with address " + std::to_string(address));
-    }
-}
-
-bool Environment::hasOwners(const std::string name) const
-{
-    return hasOwners(getObjectAddress(name));
-}
-
-bool Environment::freeObject(const unsigned int address)
-{
-    if (!hasOwners(address))
-    {
-        if (!isDryRun() and object_[address].isRegistered)
-        {
-            LOG(Message) << "Destroying object '" << object_[address].name
-                         << "'" << std::endl;
-        }
-        for (auto &p: object_[address].properties)
-        {
-            object_[p].owners.erase(address);
-        }
-        object_[address].size         = 0;
-        object_[address].Ls           = 0;
-        object_[address].isRegistered = false;
-        object_[address].type         = nullptr;
-        object_[address].owners.clear();
-        object_[address].properties.clear();
-        object_[address].data.reset(nullptr);
-        
-        return true;
-    }
-    else
-    {
-        return false;
-    }
-}
-
-bool Environment::freeObject(const std::string name)
-{
-    return freeObject(getObjectAddress(name));
+    freeObject(getObjectAddress(name));
 }
 
 void Environment::freeAll(void)
@@ -726,18 +368,24 @@ void Environment::freeAll(void)
     }
 }
 
-void Environment::printContent(void)
+void Environment::protectObjects(const bool protect)
 {
-    LOG(Message) << "Modules: " << std::endl;
-    for (unsigned int i = 0; i < module_.size(); ++i)
-    {
-        LOG(Message) << std::setw(4) << i << ": "
-                     << getModuleName(i) << std::endl;
-    }
-    LOG(Message) << "Objects: " << std::endl;
+    protect_ = protect;
+}
+
+bool Environment::objectsProtected(void) const
+{
+    return protect_;
+}
+
+// print environment content ///////////////////////////////////////////////////
+void Environment::printContent(void) const
+{
+    LOG(Debug) << "Objects: " << std::endl;
     for (unsigned int i = 0; i < object_.size(); ++i)
     {
-        LOG(Message) << std::setw(4) << i << ": "
-                     << getObjectName(i) << std::endl;
+        LOG(Debug) << std::setw(4) << i << ": "
+                   << getObjectName(i) << " ("
+                   << sizeString(getObjectSize(i)) << ")" << std::endl;
     }
 }

extras/Hadrons/Environment.hpp

@@ -4,8 +4,7 @@ Grid physics library, www.github.com/paboyle/Grid
 
 Source file: extras/Hadrons/Environment.hpp
 
-Copyright (C) 2015
-Copyright (C) 2016
+Copyright (C) 2015-2018
 
 Author: Antonin Portelli <antonin.portelli@me.com>
 
@@ -31,20 +30,12 @@ See the full license in the file "LICENSE" in the top level distribution directo
 #define Hadrons_Environment_hpp_
 
 #include <Grid/Hadrons/Global.hpp>
-#include <Grid/Hadrons/Graph.hpp>
-
-#ifndef SITE_SIZE_TYPE
-#define SITE_SIZE_TYPE unsigned int
-#endif
 
 BEGIN_HADRONS_NAMESPACE
 
 /******************************************************************************
  *                         Global environment                                 *
  ******************************************************************************/
-// forward declaration of Module
-class ModuleBase;
-
 class Object
 {
 public:
@@ -66,137 +57,100 @@ private:
     std::unique_ptr<T> objPt_{nullptr};
 };
 
+#define DEFINE_ENV_ALIAS \
+inline Environment & env(void) const\
+{\
+    return Environment::getInstance();\
+}
+
 class Environment
 {
     SINGLETON(Environment);
 public:
     typedef SITE_SIZE_TYPE                         Size;
-    typedef std::unique_ptr<ModuleBase>            ModPt;
     typedef std::unique_ptr<GridCartesian>         GridPt;
     typedef std::unique_ptr<GridRedBlackCartesian> GridRbPt;
     typedef std::unique_ptr<GridParallelRNG>       RngPt;
-    typedef std::unique_ptr<LatticeBase>           LatticePt;
+    enum class Storage {object, cache, temporary};
 private:
-    struct ModuleInfo
-    {
-        const std::type_info      *type{nullptr};
-        std::string               name;
-        ModPt                     data{nullptr};
-        std::vector<unsigned int> input;
-    };
     struct ObjInfo
     {
         Size                    size{0};
+        Storage                 storage{Storage::object};
         unsigned int            Ls{0};
-        bool                    isRegistered{false};
         const std::type_info    *type{nullptr};
         std::string             name;
         int                     module{-1};
-        std::set<unsigned int>  owners, properties;
         std::unique_ptr<Object> data{nullptr};
     };
 public:
-    // dry run
-    void                    dryRun(const bool isDry);
-    bool                    isDryRun(void) const;
-    // trajectory number
-    void                    setTrajectory(const unsigned int traj);
-    unsigned int            getTrajectory(void) const;
     // grids
     void                    createGrid(const unsigned int Ls);
     GridCartesian *         getGrid(const unsigned int Ls = 1) const;
     GridRedBlackCartesian * getRbGrid(const unsigned int Ls = 1) const;
+    std::vector<int>        getDim(void) const;
+    int                     getDim(const unsigned int mu) const;
+    unsigned long int       getLocalVolume(void) const;
     unsigned int            getNd(void) const;
     // random number generator
     void                    setSeed(const std::vector<int> &seed);
     GridParallelRNG *       get4dRng(void) const;
-    // module management
-    void                    pushModule(ModPt &pt);
-    template <typename M>
-    void                    createModule(const std::string name);
-    template <typename M>
-    void                    createModule(const std::string name,
-                                         const typename M::Par &par);
-    void                    createModule(const std::string name,
-                                         const std::string type,
-                                         XmlReader &reader);
-    unsigned int            getNModule(void) const;
-    ModuleBase *            getModule(const unsigned int address) const;
-    ModuleBase *            getModule(const std::string name) const;
-    template <typename M>
-    M *                     getModule(const unsigned int address) const;
-    template <typename M>
-    M *                     getModule(const std::string name) const;
-    unsigned int            getModuleAddress(const std::string name) const;
-    std::string             getModuleName(const unsigned int address) const;
-    std::string             getModuleType(const unsigned int address) const;
-    std::string             getModuleType(const std::string name) const;
-    bool                    hasModule(const unsigned int address) const;
-    bool                    hasModule(const std::string name) const;
-    Graph<unsigned int>     makeModuleGraph(void) const;
-    Size                    executeProgram(const std::vector<unsigned int> &p);
-    Size                    executeProgram(const std::vector<std::string> &p);
     // general memory management
     void                    addObject(const std::string name,
                                       const int moduleAddress = -1);
-    void                    registerObject(const unsigned int address,
-                                           const unsigned int size,
-                                           const unsigned int Ls = 1);
-    void                    registerObject(const std::string name,
-                                           const unsigned int size,
-                                           const unsigned int Ls = 1);
-    template <typename T>
-    unsigned int            lattice4dSize(void) const;
-    template <typename T>
-    void                    registerLattice(const unsigned int address,
-                                            const unsigned int Ls = 1);
-    template <typename T>
-    void                    registerLattice(const std::string name,
-                                            const unsigned int Ls = 1);
-    template <typename T>
-    void                    setObject(const unsigned int address, T *object);
-    template <typename T>
-    void                    setObject(const std::string name, T *object);
+    template <typename B, typename T, typename ... Ts>
+    void                    createDerivedObject(const std::string name,
+                                                const Environment::Storage storage,
+                                                const unsigned int Ls,
+                                                Ts && ... args);
+    template <typename T, typename ... Ts>
+    void                    createObject(const std::string name,
+                                         const Environment::Storage storage,
+                                         const unsigned int Ls,
+                                         Ts && ... args);
+    void                    setObjectModule(const unsigned int objAddress,
+                                            const int modAddress);
     template <typename T>
     T *                     getObject(const unsigned int address) const;
     template <typename T>
     T *                     getObject(const std::string name) const;
-    template <typename T>
-    T *                     createLattice(const unsigned int address);
-    template <typename T>
-    T *                     createLattice(const std::string name);
+    unsigned int            getMaxAddress(void) const;
     unsigned int            getObjectAddress(const std::string name) const;
     std::string             getObjectName(const unsigned int address) const;
     std::string             getObjectType(const unsigned int address) const;
     std::string             getObjectType(const std::string name) const;
     Size                    getObjectSize(const unsigned int address) const;
     Size                    getObjectSize(const std::string name) const;
+    Storage                 getObjectStorage(const unsigned int address) const;
+    Storage                 getObjectStorage(const std::string name) const;
+    int                     getObjectModule(const unsigned int address) const;
+    int                     getObjectModule(const std::string name) const;
     unsigned int            getObjectLs(const unsigned int address) const;
     unsigned int            getObjectLs(const std::string name) const;
     bool                    hasObject(const unsigned int address) const;
     bool                    hasObject(const std::string name) const;
-    bool                    hasRegisteredObject(const unsigned int address) const;
-    bool                    hasRegisteredObject(const std::string name) const;
     bool                    hasCreatedObject(const unsigned int address) const;
     bool                    hasCreatedObject(const std::string name) const;
     bool                    isObject5d(const unsigned int address) const;
     bool                    isObject5d(const std::string name) const;
+    template <typename T>
+    bool                    isObjectOfType(const unsigned int address) const;
+    template <typename T>
+    bool                    isObjectOfType(const std::string name) const;
     Environment::Size       getTotalSize(void) const;
-    void                    addOwnership(const unsigned int owner,
-                                         const unsigned int property);
-    void                    addOwnership(const std::string owner,
-                                         const std::string property);
-    bool                    hasOwners(const unsigned int address) const;
-    bool                    hasOwners(const std::string name) const;
-    bool                    freeObject(const unsigned int address);
-    bool                    freeObject(const std::string name);
+    void                    freeObject(const unsigned int address);
+    void                    freeObject(const std::string name);
     void                    freeAll(void);
-    void                    printContent(void);
+    void                    protectObjects(const bool protect);
+    bool                    objectsProtected(void) const;
+    // print environment content
+    void                    printContent(void) const;
 private:
     // general
-    bool                                   dryRun_{false};
-    unsigned int                           traj_, locVol_;
+    unsigned long int                      locVol_;
+    bool                                   protect_{true};
     // grids
+    std::vector<int>                       dim_;
     GridPt                                 grid4d_;
     std::map<unsigned int, GridPt>         grid5d_;
     GridRbPt                               gridRb4d_;
@@ -204,11 +158,6 @@ private:
     unsigned int                           nd_;
     // random number generator
     RngPt                                  rng4d_;
-    // module and related maps
-    std::vector<ModuleInfo>                module_;
-    std::map<std::string, unsigned int>    moduleAddress_;
-    // lattice store
-    std::map<unsigned int, LatticePt>      lattice_;
     // object store
     std::vector<ObjInfo>                   object_;
     std::map<std::string, unsigned int>    objectAddress_;
@@ -245,116 +194,85 @@ void Holder<T>::reset(T *pt)
 /******************************************************************************
  *                     Environment template implementation                    *
  ******************************************************************************/
-// module management ///////////////////////////////////////////////////////////
-template <typename M>
-void Environment::createModule(const std::string name)
+// general memory management ///////////////////////////////////////////////////
+template <typename B, typename T, typename ... Ts>
+void Environment::createDerivedObject(const std::string name,
+                                      const Environment::Storage storage,
+                                      const unsigned int Ls,
+                                      Ts && ... args)
 {
-    ModPt pt(new M(name));
+    if (!hasObject(name))
+    {
+        addObject(name);
+    }
     
-    pushModule(pt);
-}
-
-template <typename M>
-void Environment::createModule(const std::string name,
-                               const typename M::Par &par)
-{
-    ModPt pt(new M(name));
+    unsigned int address = getObjectAddress(name);
     
-    static_cast<M *>(pt.get())->setPar(par);
-    pushModule(pt);
-}
-
-template <typename M>
-M * Environment::getModule(const unsigned int address) const
-{
-    if (auto *pt = dynamic_cast<M *>(getModule(address)))
+    if (!object_[address].data or !objectsProtected())
     {
-        return pt;
+        MemoryStats memStats;
+    
+        if (!MemoryProfiler::stats)
+        {
+            MemoryProfiler::stats = &memStats;
+        }
+        size_t initMem           = MemoryProfiler::stats->currentlyAllocated;
+        object_[address].storage = storage;
+        object_[address].Ls      = Ls;
+        object_[address].data.reset(new Holder<B>(new T(std::forward<Ts>(args)...)));
+        object_[address].size    = MemoryProfiler::stats->maxAllocated - initMem;
+        object_[address].type    = &typeid(T);
+        if (MemoryProfiler::stats == &memStats)
+        {
+            MemoryProfiler::stats = nullptr;
+        }
     }
-    else
+    // object already exists, no error if it is a cache, error otherwise
+    else if ((object_[address].storage != Storage::cache) or 
+             (object_[address].storage != storage)        or
+             (object_[address].name    != name)           or
+             (object_[address].type    != &typeid(T)))
     {
-        HADRON_ERROR("module '" + module_[address].name
-                     + "' does not have type " + typeid(M).name()
-                     + "(object type: " + getModuleType(address) + ")");
+        HADRON_ERROR(Definition, "object '" + name + "' already allocated");
     }
 }
 
-template <typename M>
-M * Environment::getModule(const std::string name) const
+template <typename T, typename ... Ts>
+void Environment::createObject(const std::string name, 
+                               const Environment::Storage storage,
+                               const unsigned int Ls,
+                               Ts && ... args)
 {
-    return getModule<M>(getModuleAddress(name));
-}
-
-template <typename T>
-unsigned int Environment::lattice4dSize(void) const
-{
-    return sizeof(typename T::vector_object)/getGrid()->Nsimd();
-}
-
-template <typename T>
-void Environment::registerLattice(const unsigned int address,
-                                  const unsigned int Ls)
-{
-    createGrid(Ls);
-    registerObject(address, Ls*lattice4dSize<T>(), Ls);
-}
-
-template <typename T>
-void Environment::registerLattice(const std::string name, const unsigned int Ls)
-{
-    createGrid(Ls);
-    registerObject(name, Ls*lattice4dSize<T>(), Ls);
-}
-
-template <typename T>
-void Environment::setObject(const unsigned int address, T *object)
-{
-    if (hasRegisteredObject(address))
-    {
-        object_[address].data.reset(new Holder<T>(object));
-        object_[address].type = &typeid(T);
-    }
-    else if (hasObject(address))
-    {
-        HADRON_ERROR("object with address " + std::to_string(address) +
-                     " exists but is not registered");
-    }
-    else
-    {
-        HADRON_ERROR("no object with address " + std::to_string(address));
-    }
-}
-
-template <typename T>
-void Environment::setObject(const std::string name, T *object)
-{
-    setObject(getObjectAddress(name), object);
+    createDerivedObject<T, T>(name, storage, Ls, std::forward<Ts>(args)...);
 }
 
 template <typename T>
 T * Environment::getObject(const unsigned int address) const
 {
-    if (hasRegisteredObject(address))
+    if (hasObject(address))
     {
-        if (auto h = dynamic_cast<Holder<T> *>(object_[address].data.get()))
+        if (hasCreatedObject(address))
         {
-            return h->getPt();
+            if (auto h = dynamic_cast<Holder<T> *>(object_[address].data.get()))
+            {
+                return h->getPt();
+            }
+            else
+            {
+                HADRON_ERROR(Definition, "object with address " + std::to_string(address) +
+                            " does not have type '" + typeName(&typeid(T)) +
+                            "' (has type '" + getObjectType(address) + "')");
+            }
         }
         else
         {
-            HADRON_ERROR("object with address " + std::to_string(address) +
-                         " does not have type '" + typeid(T).name() +
-                         "' (has type '" + getObjectType(address) + "')");
+            HADRON_ERROR(Definition, "object with address " + std::to_string(address) +
+                         " is empty");
         }
     }
-    else if (hasObject(address))
-    {
-        HADRON_ERROR("object with address " + std::to_string(address) +
-                     " exists but is not registered");
-    }
     else
     {
-        HADRON_ERROR("no object with address " + std::to_string(address));
+        HADRON_ERROR(Definition, "no object with address " + std::to_string(address));
     }
 }
 
@@ -365,19 +283,29 @@ T * Environment::getObject(const std::string name) const
 }
 
 template <typename T>
-T * Environment::createLattice(const unsigned int address)
+bool Environment::isObjectOfType(const unsigned int address) const
 {
-    GridCartesian *g = getGrid(getObjectLs(address));
-    
-    setObject(address, new T(g));
-    
-    return getObject<T>(address);
+    if (hasObject(address))
+    {
+        if (auto h = dynamic_cast<Holder<T> *>(object_[address].data.get()))
+        {
+            return true;
+        }
+        else
+        {
+            return false;
+        }
+    }
+    else
+    {
+        HADRON_ERROR(Definition, "no object with address " + std::to_string(address));
+    }
 }
 
 template <typename T>
-T * Environment::createLattice(const std::string name)
+bool Environment::isObjectOfType(const std::string name) const
 {
-    return createLattice<T>(getObjectAddress(name));
+    return isObjectOfType<T>(getObjectAddress(name));
 }
 
 END_HADRONS_NAMESPACE

extras/Hadrons/Exceptions.cc

@@ -0,0 +1,57 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Exceptions.cc
+
+Copyright (C) 2015-2018
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+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/Hadrons/Exceptions.hpp>
+
+#ifndef ERR_SUFF
+#define ERR_SUFF " (" + loc + ")"
+#endif
+
+#define CONST_EXC(name, init) \
+name::name(std::string msg, std::string loc)\
+:init\
+{}
+
+using namespace Grid;
+using namespace Hadrons;
+using namespace Exceptions;
+
+// logic errors
+CONST_EXC(Logic, logic_error(msg + ERR_SUFF))
+CONST_EXC(Definition, Logic("definition error: " + msg, loc))
+CONST_EXC(Implementation, Logic("implementation error: " + msg, loc))
+CONST_EXC(Range, Logic("range error: " + msg, loc))
+CONST_EXC(Size, Logic("size error: " + msg, loc))
+// runtime errors
+CONST_EXC(Runtime, runtime_error(msg + ERR_SUFF))
+CONST_EXC(Argument, Runtime("argument error: " + msg, loc))
+CONST_EXC(Io, Runtime("IO error: " + msg, loc))
+CONST_EXC(Memory, Runtime("memory error: " + msg, loc))
+CONST_EXC(Parsing, Runtime("parsing error: " + msg, loc))
+CONST_EXC(Program, Runtime("program error: " + msg, loc))
+CONST_EXC(System, Runtime("system error: " + msg, loc))

extras/Hadrons/Exceptions.hpp

@@ -0,0 +1,72 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Exceptions.hpp
+
+Copyright (C) 2015-2018
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+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 */
+
+#ifndef Hadrons_Exceptions_hpp_
+#define Hadrons_Exceptions_hpp_
+
+#include <stdexcept>
+#ifndef Hadrons_Global_hpp_
+#include <Grid/Hadrons/Global.hpp>
+#endif
+
+#define SRC_LOC std::string(__FUNCTION__) + " at " + std::string(__FILE__) + ":"\
+                + std::to_string(__LINE__)
+#define HADRON_ERROR(exc, msg)\
+LOG(Error) << msg << std::endl;\
+throw(Exceptions::exc(msg, SRC_LOC));
+
+#define DECL_EXC(name, base) \
+class name: public base\
+{\
+public:\
+    name(std::string msg, std::string loc);\
+}
+
+BEGIN_HADRONS_NAMESPACE
+
+namespace Exceptions
+{
+    // logic errors
+    DECL_EXC(Logic, std::logic_error);
+    DECL_EXC(Definition, Logic);
+    DECL_EXC(Implementation, Logic);
+    DECL_EXC(Range, Logic);
+    DECL_EXC(Size, Logic);
+    // runtime errors
+    DECL_EXC(Runtime, std::runtime_error);
+    DECL_EXC(Argument, Runtime);
+    DECL_EXC(Io, Runtime);
+    DECL_EXC(Memory, Runtime);
+    DECL_EXC(Parsing, Runtime);
+    DECL_EXC(Program, Runtime);
+    DECL_EXC(System, Runtime);
+}
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_Exceptions_hpp_

extras/Hadrons/Factory.hpp

@@ -4,8 +4,7 @@ Grid physics library, www.github.com/paboyle/Grid
 
 Source file: extras/Hadrons/Factory.hpp
 
-Copyright (C) 2015
-Copyright (C) 2016
+Copyright (C) 2015-2018
 
 Author: Antonin Portelli <antonin.portelli@me.com>
 
@@ -95,7 +94,7 @@ std::unique_ptr<T> Factory<T>::create(const std::string type,
     }
     catch (std::out_of_range &)
     {
-        HADRON_ERROR("object of type '" + type + "' unknown");
+        HADRON_ERROR(Argument, "object of type '" + type + "' unknown");
     }
     
     return func(name);

extras/Hadrons/GeneticScheduler.hpp

@@ -4,8 +4,7 @@ Grid physics library, www.github.com/paboyle/Grid
 
 Source file: extras/Hadrons/GeneticScheduler.hpp
 
-Copyright (C) 2015
-Copyright (C) 2016
+Copyright (C) 2015-2018
 
 Author: Antonin Portelli <antonin.portelli@me.com>
 
@@ -38,13 +37,13 @@ BEGIN_HADRONS_NAMESPACE
 /******************************************************************************
  *                   Scheduler based on a genetic algorithm                   *
  ******************************************************************************/
-template <typename T>
+template <typename V, typename T>
 class GeneticScheduler
 {
 public:
-    typedef std::vector<T>                   Gene;
-    typedef std::pair<Gene *, Gene *>        GenePair;
-    typedef std::function<int(const Gene &)> ObjFunc;
+    typedef std::vector<T>                 Gene;
+    typedef std::pair<Gene *, Gene *>      GenePair;
+    typedef std::function<V(const Gene &)> ObjFunc;
     struct Parameters
     {
         double       mutationRate;
@@ -65,7 +64,7 @@ public:
     void benchmarkCrossover(const unsigned int nIt);
     // print population
     friend std::ostream & operator<<(std::ostream &out,
-                                     const GeneticScheduler<T> &s)
+                                     const GeneticScheduler<V, T> &s)
     {
         out << "[";
         for (auto &p: s.population_)
@@ -87,19 +86,19 @@ private:
     void     mutation(Gene &m, const Gene &c);
     
 private:
-    Graph<T>                 &graph_;
-    const ObjFunc            &func_;
-    const Parameters         par_;
-    std::multimap<int, Gene> population_;
-    std::mt19937             gen_;
+    Graph<T>               &graph_;
+    const ObjFunc          &func_;
+    const Parameters       par_;
+    std::multimap<V, Gene> population_;
+    std::mt19937           gen_;
 };
 
 /******************************************************************************
  *                       template implementation                              *
  ******************************************************************************/
 // constructor /////////////////////////////////////////////////////////////////
-template <typename T>
-GeneticScheduler<T>::GeneticScheduler(Graph<T> &graph, const ObjFunc &func,
+template <typename V, typename T>
+GeneticScheduler<V, T>::GeneticScheduler(Graph<T> &graph, const ObjFunc &func,
                                       const Parameters &par)
 : graph_(graph)
 , func_(func)
@@ -109,22 +108,22 @@ GeneticScheduler<T>::GeneticScheduler(Graph<T> &graph, const ObjFunc &func,
 }
 
 // access //////////////////////////////////////////////////////////////////////
-template <typename T>
-const typename GeneticScheduler<T>::Gene &
-GeneticScheduler<T>::getMinSchedule(void)
+template <typename V, typename T>
+const typename GeneticScheduler<V, T>::Gene &
+GeneticScheduler<V, T>::getMinSchedule(void)
 {
     return population_.begin()->second;
 }
 
-template <typename T>
-int GeneticScheduler<T>::getMinValue(void)
+template <typename V, typename T>
+int GeneticScheduler<V, T>::getMinValue(void)
 {
     return population_.begin()->first;
 }
 
 // breed a new generation //////////////////////////////////////////////////////
-template <typename T>
-void GeneticScheduler<T>::nextGeneration(void)
+template <typename V, typename T>
+void GeneticScheduler<V, T>::nextGeneration(void)
 {
     // random initialization of the population if necessary
     if (population_.size() != par_.popSize)
@@ -158,8 +157,8 @@ void GeneticScheduler<T>::nextGeneration(void)
 }
 
 // evolution steps /////////////////////////////////////////////////////////////
-template <typename T>
-void GeneticScheduler<T>::initPopulation(void)
+template <typename V, typename T>
+void GeneticScheduler<V, T>::initPopulation(void)
 {
     population_.clear();
     for (unsigned int i = 0; i < par_.popSize; ++i)
@@ -170,8 +169,8 @@ void GeneticScheduler<T>::initPopulation(void)
     }
 }
 
-template <typename T>
-void GeneticScheduler<T>::doCrossover(void)
+template <typename V, typename T>
+void GeneticScheduler<V, T>::doCrossover(void)
 {
     auto p = selectPair();
     Gene &p1 = *(p.first), &p2 = *(p.second);
@@ -185,8 +184,8 @@ void GeneticScheduler<T>::doCrossover(void)
     }
 }
 
-template <typename T>
-void GeneticScheduler<T>::doMutation(void)
+template <typename V, typename T>
+void GeneticScheduler<V, T>::doMutation(void)
 {
     std::uniform_real_distribution<double>      mdis(0., 1.);
     std::uniform_int_distribution<unsigned int> pdis(0, population_.size() - 1);
@@ -206,40 +205,35 @@ void GeneticScheduler<T>::doMutation(void)
 }
 
 // genetic operators ///////////////////////////////////////////////////////////
-template <typename T>
-typename GeneticScheduler<T>::GenePair GeneticScheduler<T>::selectPair(void)
+template <typename V, typename T>
+typename GeneticScheduler<V, T>::GenePair GeneticScheduler<V, T>::selectPair(void)
 {
     std::vector<double> prob;
     unsigned int        ind;
     Gene                *p1, *p2;
+    const double        max = population_.rbegin()->first;
     
+
     for (auto &c: population_)
     {
-        prob.push_back(1./c.first);
-    }
-    do
-    {
-        double probCpy;
-        
-        std::discrete_distribution<unsigned int> dis1(prob.begin(), prob.end());
-        auto rIt = population_.begin();
-        ind = dis1(gen_);
-        std::advance(rIt, ind);
-        p1 = &(rIt->second);
-        probCpy   = prob[ind];
-        prob[ind] = 0.;
-        std::discrete_distribution<unsigned int> dis2(prob.begin(), prob.end());
-        rIt = population_.begin();
-        std::advance(rIt, dis2(gen_));
-        p2 = &(rIt->second);
-        prob[ind] = probCpy;
-    } while (p1 == p2);
+        prob.push_back(std::exp((c.first-1.)/max));
+    }        
+    std::discrete_distribution<unsigned int> dis1(prob.begin(), prob.end());
+    auto rIt = population_.begin();
+    ind = dis1(gen_);
+    std::advance(rIt, ind);
+    p1 = &(rIt->second);
+    prob[ind] = 0.;
+    std::discrete_distribution<unsigned int> dis2(prob.begin(), prob.end());
+    rIt = population_.begin();
+    std::advance(rIt, dis2(gen_));
+    p2 = &(rIt->second);
     
     return std::make_pair(p1, p2);
 }
 
-template <typename T>
-void GeneticScheduler<T>::crossover(Gene &c1, Gene &c2, const Gene &p1,
+template <typename V, typename T>
+void GeneticScheduler<V, T>::crossover(Gene &c1, Gene &c2, const Gene &p1,
                                     const Gene &p2)
 {
     Gene                                        buf;
@@ -273,8 +267,8 @@ void GeneticScheduler<T>::crossover(Gene &c1, Gene &c2, const Gene &p1,
     }
 }
 
-template <typename T>
-void GeneticScheduler<T>::mutation(Gene &m, const Gene &c)
+template <typename V, typename T>
+void GeneticScheduler<V, T>::mutation(Gene &m, const Gene &c)
 {
     Gene                                        buf;
     std::uniform_int_distribution<unsigned int> dis(0, c.size() - 1);
@@ -303,8 +297,8 @@ void GeneticScheduler<T>::mutation(Gene &m, const Gene &c)
     }
 }
 
-template <typename T>
-void GeneticScheduler<T>::benchmarkCrossover(const unsigned int nIt)
+template <typename V, typename T>
+void GeneticScheduler<V, T>::benchmarkCrossover(const unsigned int nIt)
 {
     Gene   p1, p2, c1, c2;
     double neg = 0., eq = 0., pos = 0., total;

extras/Hadrons/Global.cc

@@ -4,8 +4,7 @@ Grid physics library, www.github.com/paboyle/Grid
 
 Source file: extras/Hadrons/Global.cc
 
-Copyright (C) 2015
-Copyright (C) 2016
+Copyright (C) 2015-2018
 
 Author: Antonin Portelli <antonin.portelli@me.com>
 
@@ -39,31 +38,19 @@ HadronsLogger Hadrons::HadronsLogMessage(1,"Message");
 HadronsLogger Hadrons::HadronsLogIterative(1,"Iterative");
 HadronsLogger Hadrons::HadronsLogDebug(1,"Debug");
 
-// pretty size formatting //////////////////////////////////////////////////////
-std::string Hadrons::sizeString(long unsigned int bytes)
-
+void Hadrons::initLogger(void)
 {
-    constexpr unsigned int bufSize = 256;
-    const char             *suffixes[7] = {"", "K", "M", "G", "T", "P", "E"};
-    char                   buf[256];
-    long unsigned int      s     = 0;
-    double                 count = bytes;
-    
-    while (count >= 1024 && s < 7)
-    {
-        s++;
-        count /= 1024;
-    }
-    if (count - floor(count) == 0.0)
-    {
-        snprintf(buf, bufSize, "%d %sB", (int)count, suffixes[s]);
-    }
-    else
-    {
-        snprintf(buf, bufSize, "%.1f %sB", count, suffixes[s]);
-    }
-    
-    return std::string(buf);
+    auto w = std::string("Hadrons").length();
+    GridLogError.setTopWidth(w);
+    GridLogWarning.setTopWidth(w);
+    GridLogMessage.setTopWidth(w);
+    GridLogIterative.setTopWidth(w);
+    GridLogDebug.setTopWidth(w);
+    HadronsLogError.Active(GridLogError.isActive());
+    HadronsLogWarning.Active(GridLogWarning.isActive());
+    HadronsLogMessage.Active(GridLogMessage.isActive());
+    HadronsLogIterative.Active(GridLogIterative.isActive());
+    HadronsLogDebug.Active(GridLogDebug.isActive());
 }
 
 // type utilities //////////////////////////////////////////////////////////////
@@ -80,3 +67,10 @@ std::string Hadrons::typeName(const std::type_info *info)
     
     return name;
 }
+
+// default writers/readers /////////////////////////////////////////////////////
+#ifdef HAVE_HDF5
+const std::string Hadrons::resultFileExt = "h5";
+#else
+const std::string Hadrons::resultFileExt = "xml";
+#endif

extras/Hadrons/Global.hpp

@@ -4,10 +4,10 @@ Grid physics library, www.github.com/paboyle/Grid
 
 Source file: extras/Hadrons/Global.hpp
 
-Copyright (C) 2015
-Copyright (C) 2016
+Copyright (C) 2015-2018
 
 Author: Antonin Portelli <antonin.portelli@me.com>
+Author: Lanny91 <andrew.lawson@gmail.com>
 
 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
@@ -35,6 +35,10 @@ See the full license in the file "LICENSE" in the top level distribution directo
 #include <Grid/Grid.h>
 #include <cxxabi.h>
 
+#ifndef SITE_SIZE_TYPE
+#define SITE_SIZE_TYPE size_t
+#endif
+
 #define BEGIN_HADRONS_NAMESPACE \
 namespace Grid {\
 using namespace QCD;\
@@ -51,23 +55,46 @@ using Grid::operator<<;
  * error with GCC 5 (clang & GCC 6 compile fine without it).
  */
 
-// FIXME: find a way to do that in a more general fashion
 #ifndef FIMPL
 #define FIMPL WilsonImplR
 #endif
+#ifndef SIMPL
+#define SIMPL ScalarImplCR
+#endif
+#ifndef GIMPL
+#define GIMPL GimplTypesR
+#endif
 
 BEGIN_HADRONS_NAMESPACE
 
 // type aliases
-#define TYPE_ALIASES(FImpl, suffix)\
-typedef FermionOperator<FImpl>                       FMat##suffix;             \
-typedef typename FImpl::FermionField                 FermionField##suffix;     \
-typedef typename FImpl::PropagatorField              PropagatorField##suffix;  \
-typedef typename FImpl::SitePropagator               SitePropagator##suffix;   \
-typedef typename FImpl::DoubledGaugeField            DoubledGaugeField##suffix;\
-typedef std::function<void(FermionField##suffix &,                             \
+#define FERM_TYPE_ALIASES(FImpl, suffix)\
+typedef FermionOperator<FImpl>                        FMat##suffix;            \
+typedef typename FImpl::FermionField                  FermionField##suffix;    \
+typedef typename FImpl::PropagatorField               PropagatorField##suffix; \
+typedef typename FImpl::SitePropagator::scalar_object SitePropagator##suffix;  \
+typedef std::vector<SitePropagator##suffix>           SlicedPropagator##suffix;
+
+#define GAUGE_TYPE_ALIASES(FImpl, suffix)\
+typedef typename FImpl::DoubledGaugeField DoubledGaugeField##suffix;
+
+#define SCALAR_TYPE_ALIASES(SImpl, suffix)\
+typedef typename SImpl::Field ScalarField##suffix;\
+typedef typename SImpl::Field PropagatorField##suffix;
+
+#define SOLVER_TYPE_ALIASES(FImpl, suffix)\
+typedef std::function<void(FermionField##suffix &,\
                       const FermionField##suffix &)> SolverFn##suffix;
 
+#define SINK_TYPE_ALIASES(suffix)\
+typedef std::function<SlicedPropagator##suffix\
+                      (const PropagatorField##suffix &)> SinkFn##suffix;
+
+#define FGS_TYPE_ALIASES(FImpl, suffix)\
+FERM_TYPE_ALIASES(FImpl, suffix)\
+GAUGE_TYPE_ALIASES(FImpl, suffix)\
+SOLVER_TYPE_ALIASES(FImpl, suffix)
+
 // logger
 class HadronsLogger: public Logger
 {
@@ -77,11 +104,6 @@ public:
 };
 
 #define LOG(channel) std::cout << HadronsLog##channel
-#define HADRON_ERROR(msg)\
-LOG(Error) << msg << " (" << __FUNCTION__ << " at " << __FILE__ << ":"\
-           << __LINE__ << ")" << std::endl;\
-abort();
-
 #define DEBUG_VAR(var) LOG(Debug) << #var << "= " << (var) << std::endl;
 
 extern HadronsLogger HadronsLogError;
@@ -90,6 +112,8 @@ extern HadronsLogger HadronsLogMessage;
 extern HadronsLogger HadronsLogIterative;
 extern HadronsLogger HadronsLogDebug;
 
+void initLogger(void);
+
 // singleton pattern
 #define SINGLETON(name)\
 public:\
@@ -115,9 +139,6 @@ public:\
 private:\
     name(void) = default;
 
-// pretty size formating
-std::string sizeString(long unsigned int bytes);
-
 // type utilities
 template <typename T>
 const std::type_info * typeIdPt(const T &x)
@@ -145,6 +166,22 @@ std::string typeName(void)
     return typeName(typeIdPt<T>());
 }
 
+// default writers/readers
+extern const std::string resultFileExt;
+
+#ifdef HAVE_HDF5
+typedef Hdf5Reader ResultReader;
+typedef Hdf5Writer ResultWriter;
+#else
+typedef XmlReader ResultReader;
+typedef XmlWriter ResultWriter;
+#endif
+
+#define RESULT_FILE_NAME(name) \
+name + "." + std::to_string(vm().getTrajectory()) + "." + resultFileExt
+
 END_HADRONS_NAMESPACE
 
+#include <Grid/Hadrons/Exceptions.hpp>
+
 #endif // Hadrons_Global_hpp_

extras/Hadrons/Graph.hpp

@@ -4,8 +4,7 @@ Grid physics library, www.github.com/paboyle/Grid
 
 Source file: extras/Hadrons/Graph.hpp
 
-Copyright (C) 2015
-Copyright (C) 2016
+Copyright (C) 2015-2018
 
 Author: Antonin Portelli <antonin.portelli@me.com>
 
@@ -185,7 +184,7 @@ void Graph<T>::removeVertex(const T &value)
     }
     else
     {
-        HADRON_ERROR("vertex " << value << " does not exists");
+        HADRON_ERROR(Range, "vertex does not exists");
     }
 
     // remove all edges containing the vertex
@@ -214,7 +213,7 @@ void Graph<T>::removeEdge(const Edge &e)
     }
     else
     {
-        HADRON_ERROR("edge "  << e << " does not exists");
+        HADRON_ERROR(Range, "edge does not exists");
     }
 }
 
@@ -260,7 +259,7 @@ void Graph<T>::mark(const T &value, const bool doMark)
     }
     else
     {
-        HADRON_ERROR("vertex " << value << " does not exists");
+        HADRON_ERROR(Range, "vertex does not exists");
     }
 }
 
@@ -298,7 +297,7 @@ bool Graph<T>::isMarked(const T &value) const
     }
     else
     {
-        HADRON_ERROR("vertex " << value << " does not exists");
+        HADRON_ERROR(Range, "vertex does not exists");
         
         return false;
     }
@@ -430,7 +429,7 @@ std::vector<T> Graph<T>::getAdjacentVertices(const T &value) const
     {
         return ((e.first == value) or (e.second == value));
     };
-    auto eIt = find_if(edgeSet_.begin(), edgeSet_.end(), pred);
+    auto eIt = std::find_if(edgeSet_.begin(), edgeSet_.end(), pred);
     
     while (eIt != edgeSet_.end())
     {
@@ -442,7 +441,7 @@ std::vector<T> Graph<T>::getAdjacentVertices(const T &value) const
         {
             adjacentVertex.push_back((*eIt).first);
         }
-        eIt = find_if(++eIt, edgeSet_.end(), pred);
+        eIt = std::find_if(++eIt, edgeSet_.end(), pred);
     }
     
     return adjacentVertex;
@@ -458,12 +457,12 @@ std::vector<T> Graph<T>::getChildren(const T &value) const
     {
         return (e.first == value);
     };
-    auto eIt = find_if(edgeSet_.begin(), edgeSet_.end(), pred);
+    auto eIt = std::find_if(edgeSet_.begin(), edgeSet_.end(), pred);
     
     while (eIt != edgeSet_.end())
     {
         child.push_back((*eIt).second);
-        eIt = find_if(++eIt, edgeSet_.end(), pred);
+        eIt = std::find_if(++eIt, edgeSet_.end(), pred);
     }
     
     return child;
@@ -479,12 +478,12 @@ std::vector<T> Graph<T>::getParents(const T &value) const
     {
         return (e.second == value);
     };
-    auto eIt = find_if(edgeSet_.begin(), edgeSet_.end(), pred);
+    auto eIt = std::find_if(edgeSet_.begin(), edgeSet_.end(), pred);
     
     while (eIt != edgeSet_.end())
     {
         parent.push_back((*eIt).first);
-        eIt = find_if(++eIt, edgeSet_.end(), pred);
+        eIt = std::find_if(++eIt, edgeSet_.end(), pred);
     }
     
     return parent;
@@ -544,7 +543,7 @@ std::vector<T> Graph<T>::topoSort(void)
     {
         if (tmpMarked.at(v))
         {
-            HADRON_ERROR("cannot topologically sort a cyclic graph");
+            HADRON_ERROR(Range, "cannot topologically sort a cyclic graph");
         }
         if (!isMarked(v))
         {
@@ -603,7 +602,7 @@ std::vector<T> Graph<T>::topoSort(Gen &gen)
     {
         if (tmpMarked.at(v))
         {
-            HADRON_ERROR("cannot topologically sort a cyclic graph");
+            HADRON_ERROR(Range, "cannot topologically sort a cyclic graph");
         }
         if (!isMarked(v))
         {

extras/Hadrons/HadronsXmlRun.cc

@@ -4,8 +4,7 @@ Grid physics library, www.github.com/paboyle/Grid
 
 Source file: extras/Hadrons/HadronsXmlRun.cc
 
-Copyright (C) 2015
-Copyright (C) 2016
+Copyright (C) 2015-2018
 
 Author: Antonin Portelli <antonin.portelli@me.com>
 
@@ -55,12 +54,6 @@ int main(int argc, char *argv[])
     
     // initialization
     Grid_init(&argc, &argv);
-    HadronsLogError.Active(GridLogError.isActive());
-    HadronsLogWarning.Active(GridLogWarning.isActive());
-    HadronsLogMessage.Active(GridLogMessage.isActive());
-    HadronsLogIterative.Active(GridLogIterative.isActive());
-    HadronsLogDebug.Active(GridLogDebug.isActive());
-    LOG(Message) << "Grid initialized" << std::endl;
     
     // execution
     Application application(parameterFileName);

extras/Hadrons/HadronsXmlSchedule.cc

@@ -4,8 +4,7 @@ Grid physics library, www.github.com/paboyle/Grid
 
 Source file: extras/Hadrons/HadronsXmlSchedule.cc
 
-Copyright (C) 2015
-Copyright (C) 2016
+Copyright (C) 2015-2018
 
 Author: Antonin Portelli <antonin.portelli@me.com>
 
@@ -49,12 +48,6 @@ int main(int argc, char *argv[])
     
     // initialization
     Grid_init(&argc, &argv);
-    HadronsLogError.Active(GridLogError.isActive());
-    HadronsLogWarning.Active(GridLogWarning.isActive());
-    HadronsLogMessage.Active(GridLogMessage.isActive());
-    HadronsLogIterative.Active(GridLogIterative.isActive());
-    HadronsLogDebug.Active(GridLogDebug.isActive());
-    LOG(Message) << "Grid initialized" << std::endl;
     
     // execution
     Application application;

extras/Hadrons/Makefile.am

@@ -7,20 +7,24 @@ libHadrons_a_SOURCES = \
     $(modules_cc)      \
     Application.cc     \
     Environment.cc     \
+	Exceptions.cc      \
     Global.cc          \
-    Module.cc
+    Module.cc		   \
+	VirtualMachine.cc
 libHadrons_adir = $(pkgincludedir)/Hadrons
 nobase_libHadrons_a_HEADERS = \
 	$(modules_hpp)            \
 	Application.hpp           \
 	Environment.hpp           \
+	Exceptions.hpp            \
 	Factory.hpp               \
 	GeneticScheduler.hpp      \
 	Global.hpp                \
 	Graph.hpp                 \
 	Module.hpp                \
 	Modules.hpp               \
-	ModuleFactory.hpp
+	ModuleFactory.hpp         \
+	VirtualMachine.hpp
 
 HadronsXmlRun_SOURCES = HadronsXmlRun.cc
 HadronsXmlRun_LDADD   = libHadrons.a -lGrid

extras/Hadrons/Module.cc

@@ -4,8 +4,7 @@ Grid physics library, www.github.com/paboyle/Grid
 
 Source file: extras/Hadrons/Module.cc
 
-Copyright (C) 2015
-Copyright (C) 2016
+Copyright (C) 2015-2018
 
 Author: Antonin Portelli <antonin.portelli@me.com>
 
@@ -39,7 +38,6 @@ using namespace Hadrons;
 // constructor /////////////////////////////////////////////////////////////////
 ModuleBase::ModuleBase(const std::string name)
 : name_(name)
-, env_(Environment::getInstance())
 {}
 
 // access //////////////////////////////////////////////////////////////////////
@@ -48,15 +46,10 @@ std::string ModuleBase::getName(void) const
     return name_;
 }
 
-Environment & ModuleBase::env(void) const
-{
-    return env_;
-}
-
 // get factory registration name if available
 std::string ModuleBase::getRegisteredName(void)
 {
-    HADRON_ERROR("module '" + getName() + "' has a type not registered"
+    HADRON_ERROR(Definition, "module '" + getName() + "' has no registered type"
                  + " in the factory");
 }
 
@@ -64,8 +57,5 @@ std::string ModuleBase::getRegisteredName(void)
 void ModuleBase::operator()(void)
 {
     setup();
-    if (!env().isDryRun())
-    {
-        execute();
-    }
+    execute();
 }

extras/Hadrons/Module.hpp

@@ -4,8 +4,7 @@ Grid physics library, www.github.com/paboyle/Grid
 
 Source file: extras/Hadrons/Module.hpp
 
-Copyright (C) 2015
-Copyright (C) 2016
+Copyright (C) 2015-2018
 
 Author: Antonin Portelli <antonin.portelli@me.com>
 
@@ -31,7 +30,7 @@ See the full license in the file "LICENSE" in the top level distribution directo
 #define Hadrons_Module_hpp_
 
 #include <Grid/Hadrons/Global.hpp>
-#include <Grid/Hadrons/Environment.hpp>
+#include <Grid/Hadrons/VirtualMachine.hpp>
 
 BEGIN_HADRONS_NAMESPACE
 
@@ -87,6 +86,56 @@ public:\
 static ns##mod##ModuleRegistrar ns##mod##ModuleRegistrarInstance;
 
 #define ARG(...) __VA_ARGS__
+#define MACRO_REDIRECT(arg1, arg2, arg3, macro, ...) macro
+
+#define envGet(type, name)\
+*env().template getObject<type>(name)
+
+#define envGetTmp(type, var)\
+type &var = *env().template getObject<type>(getName() + "_tmp_" + #var)
+
+#define envHasType(type, name)\
+env().template isObjectOfType<type>(name)
+
+#define envCreate(type, name, Ls, ...)\
+env().template createObject<type>(name, Environment::Storage::object, Ls, __VA_ARGS__)
+
+#define envCreateDerived(base, type, name, Ls, ...)\
+env().template createDerivedObject<base, type>(name, Environment::Storage::object, Ls, __VA_ARGS__)
+
+#define envCreateLat4(type, name)\
+envCreate(type, name, 1, env().getGrid())
+
+#define envCreateLat5(type, name, Ls)\
+envCreate(type, name, Ls, env().getGrid(Ls))
+
+#define envCreateLat(...)\
+MACRO_REDIRECT(__VA_ARGS__, envCreateLat5, envCreateLat4)(__VA_ARGS__)
+
+#define envCache(type, name, Ls, ...)\
+env().template createObject<type>(name, Environment::Storage::cache, Ls, __VA_ARGS__)
+
+#define envCacheLat4(type, name)\
+envCache(type, name, 1, env().getGrid())
+
+#define envCacheLat5(type, name, Ls)\
+envCache(type, name, Ls, env().getGrid(Ls))
+
+#define envCacheLat(...)\
+MACRO_REDIRECT(__VA_ARGS__, envCacheLat5, envCacheLat4)(__VA_ARGS__)
+
+#define envTmp(type, name, Ls, ...)\
+env().template createObject<type>(getName() + "_tmp_" + name,         \
+                                  Environment::Storage::temporary, Ls, __VA_ARGS__)
+
+#define envTmpLat4(type, name)\
+envTmp(type, name, 1, env().getGrid())
+
+#define envTmpLat5(type, name, Ls)\
+envTmp(type, name, Ls, env().getGrid(Ls))
+
+#define envTmpLat(...)\
+MACRO_REDIRECT(__VA_ARGS__, envTmpLat5, envTmpLat4)(__VA_ARGS__)
 
 /******************************************************************************
  *                            Module class                                    *
@@ -101,23 +150,30 @@ public:
     virtual ~ModuleBase(void) = default;
     // access
     std::string getName(void) const;
-    Environment &env(void) const;
     // get factory registration name if available
     virtual std::string getRegisteredName(void);
     // dependencies/products
     virtual std::vector<std::string> getInput(void) = 0;
+    virtual std::vector<std::string> getReference(void)
+    {
+        return std::vector<std::string>(0);
+    };
     virtual std::vector<std::string> getOutput(void) = 0;
     // parse parameters
     virtual void parseParameters(XmlReader &reader, const std::string name) = 0;
     virtual void saveParameters(XmlWriter &writer, const std::string name) = 0;
     // setup
     virtual void setup(void) {};
+    virtual void execute(void) = 0;
     // execution
     void operator()(void);
-    virtual void execute(void) = 0;
+protected:
+    // environment shortcut
+    DEFINE_ENV_ALIAS;
+    // virtual machine shortcut
+    DEFINE_VM_ALIAS;
 private:
     std::string name_;
-    Environment &env_;
 };
 
 // derived class, templating the parameter class

extras/Hadrons/ModuleFactory.hpp

@@ -4,8 +4,7 @@ Grid physics library, www.github.com/paboyle/Grid
 
 Source file: extras/Hadrons/ModuleFactory.hpp
 
-Copyright (C) 2015
-Copyright (C) 2016
+Copyright (C) 2015-2018
 
 Author: Antonin Portelli <antonin.portelli@me.com>
 

extras/Hadrons/Modules.hpp

@@ -4,10 +4,10 @@ Grid physics library, www.github.com/paboyle/Grid
 
 Source file: extras/Hadrons/Modules.hpp
 
-Copyright (C) 2015
-Copyright (C) 2016
+Copyright (C) 2015-2018
 
 Author: Antonin Portelli <antonin.portelli@me.com>
+Author: Lanny91 <andrew.lawson@gmail.com>
 
 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
@@ -26,15 +26,40 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution directory
 *************************************************************************************/
 /*  END LEGAL */
-#include <Grid/Hadrons/Modules/MAction/DWF.hpp>
-#include <Grid/Hadrons/Modules/MAction/Wilson.hpp>
 #include <Grid/Hadrons/Modules/MContraction/Baryon.hpp>
 #include <Grid/Hadrons/Modules/MContraction/Meson.hpp>
-#include <Grid/Hadrons/Modules/MGauge/Load.hpp>
-#include <Grid/Hadrons/Modules/MGauge/Random.hpp>
-#include <Grid/Hadrons/Modules/MGauge/Unit.hpp>
-#include <Grid/Hadrons/Modules/MSolver/RBPrecCG.hpp>
-#include <Grid/Hadrons/Modules/MSource/Point.hpp>
+#include <Grid/Hadrons/Modules/MContraction/WeakHamiltonian.hpp>
+#include <Grid/Hadrons/Modules/MContraction/WeakHamiltonianNonEye.hpp>
+#include <Grid/Hadrons/Modules/MContraction/DiscLoop.hpp>
+#include <Grid/Hadrons/Modules/MContraction/WeakNeutral4ptDisc.hpp>
+#include <Grid/Hadrons/Modules/MContraction/Gamma3pt.hpp>
+#include <Grid/Hadrons/Modules/MContraction/WardIdentity.hpp>
+#include <Grid/Hadrons/Modules/MContraction/WeakHamiltonianEye.hpp>
+#include <Grid/Hadrons/Modules/MFermion/GaugeProp.hpp>
 #include <Grid/Hadrons/Modules/MSource/SeqGamma.hpp>
+#include <Grid/Hadrons/Modules/MSource/Point.hpp>
+#include <Grid/Hadrons/Modules/MSource/Wall.hpp>
 #include <Grid/Hadrons/Modules/MSource/Z2.hpp>
-#include <Grid/Hadrons/Modules/Quark.hpp>
+#include <Grid/Hadrons/Modules/MSource/SeqConserved.hpp>
+#include <Grid/Hadrons/Modules/MSink/Smear.hpp>
+#include <Grid/Hadrons/Modules/MSink/Point.hpp>
+#include <Grid/Hadrons/Modules/MSolver/RBPrecCG.hpp>
+#include <Grid/Hadrons/Modules/MGauge/Unit.hpp>
+#include <Grid/Hadrons/Modules/MGauge/Random.hpp>
+#include <Grid/Hadrons/Modules/MGauge/StochEm.hpp>
+#include <Grid/Hadrons/Modules/MGauge/FundtoHirep.hpp>
+#include <Grid/Hadrons/Modules/MUtilities/TestSeqGamma.hpp>
+#include <Grid/Hadrons/Modules/MUtilities/TestSeqConserved.hpp>
+#include <Grid/Hadrons/Modules/MLoop/NoiseLoop.hpp>
+#include <Grid/Hadrons/Modules/MScalar/FreeProp.hpp>
+#include <Grid/Hadrons/Modules/MScalar/Scalar.hpp>
+#include <Grid/Hadrons/Modules/MScalar/ChargedProp.hpp>
+#include <Grid/Hadrons/Modules/MAction/DWF.hpp>
+#include <Grid/Hadrons/Modules/MAction/Wilson.hpp>
+#include <Grid/Hadrons/Modules/MAction/WilsonClover.hpp>
+#include <Grid/Hadrons/Modules/MScalarSUN/Div.hpp>
+#include <Grid/Hadrons/Modules/MScalarSUN/TrMag.hpp>
+#include <Grid/Hadrons/Modules/MScalarSUN/TwoPoint.hpp>
+#include <Grid/Hadrons/Modules/MScalarSUN/TrPhi.hpp>
+#include <Grid/Hadrons/Modules/MIO/LoadNersc.hpp>
+#include <Grid/Hadrons/Modules/MIO/LoadBinary.hpp>

extras/Hadrons/Modules/MAction/DWF.hpp

@@ -4,10 +4,10 @@ Grid physics library, www.github.com/paboyle/Grid
 
 Source file: extras/Hadrons/Modules/MAction/DWF.hpp
 
-Copyright (C) 2015
-Copyright (C) 2016
+Copyright (C) 2015-2018
 
 Author: Antonin Portelli <antonin.portelli@me.com>
+Author: Lanny91 <andrew.lawson@gmail.com>
 
 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
@@ -27,8 +27,8 @@ See the full license in the file "LICENSE" in the top level distribution directo
 *************************************************************************************/
 /*  END LEGAL */
 
-#ifndef Hadrons_DWF_hpp_
-#define Hadrons_DWF_hpp_
+#ifndef Hadrons_MAction_DWF_hpp_
+#define Hadrons_MAction_DWF_hpp_
 
 #include <Grid/Hadrons/Global.hpp>
 #include <Grid/Hadrons/Module.hpp>
@@ -48,14 +48,15 @@ public:
                                     std::string, gauge,
                                     unsigned int, Ls,
                                     double      , mass,
-                                    double      , M5);
+                                    double      , M5,
+                                    std::string , boundary);
 };
 
 template <typename FImpl>
 class TDWF: public Module<DWFPar>
 {
 public:
-    TYPE_ALIASES(FImpl,);
+    FGS_TYPE_ALIASES(FImpl,);
 public:
     // constructor
     TDWF(const std::string name);
@@ -64,6 +65,7 @@ public:
     // dependency relation
     virtual std::vector<std::string> getInput(void);
     virtual std::vector<std::string> getOutput(void);
+protected:
     // setup
     virtual void setup(void);
     // execution
@@ -102,33 +104,32 @@ std::vector<std::string> TDWF<FImpl>::getOutput(void)
 template <typename FImpl>
 void TDWF<FImpl>::setup(void)
 {
-    unsigned int size;
-    
-    size = 2*env().template lattice4dSize<typename FImpl::DoubledGaugeField>();
-    env().registerObject(getName(), size, par().Ls);
+    LOG(Message) << "Setting up domain wall fermion matrix with m= "
+                 << par().mass << ", M5= " << par().M5 << " and Ls= "
+                 << par().Ls << " using gauge field '" << par().gauge << "'"
+                 << std::endl;
+    LOG(Message) << "Fermion boundary conditions: " << par().boundary
+                 << std::endl;
+                 
+    env().createGrid(par().Ls);
+    auto &U    = envGet(LatticeGaugeField, par().gauge);
+    auto &g4   = *env().getGrid();
+    auto &grb4 = *env().getRbGrid();
+    auto &g5   = *env().getGrid(par().Ls);
+    auto &grb5 = *env().getRbGrid(par().Ls);
+    std::vector<Complex> boundary = strToVec<Complex>(par().boundary);
+    typename DomainWallFermion<FImpl>::ImplParams implParams(boundary);
+    envCreateDerived(FMat, DomainWallFermion<FImpl>, getName(), par().Ls, U, g5,
+                     grb5, g4, grb4, par().mass, par().M5, implParams);
 }
 
 // execution ///////////////////////////////////////////////////////////////////
 template <typename FImpl>
 void TDWF<FImpl>::execute(void)
-{
-    LOG(Message) << "Setting up domain wall fermion matrix with m= "
-                 << par().mass << ", M5= " << par().M5 << " and Ls= "
-                 << par().Ls << " using gauge field '" << par().gauge << "'"
-                 << std::endl;
-    env().createGrid(par().Ls);
-    auto &U      = *env().template getObject<LatticeGaugeField>(par().gauge);
-    auto &g4     = *env().getGrid();
-    auto &grb4   = *env().getRbGrid();
-    auto &g5     = *env().getGrid(par().Ls);
-    auto &grb5   = *env().getRbGrid(par().Ls);
-    FMat *fMatPt = new DomainWallFermion<FImpl>(U, g5, grb5, g4, grb4,
-                                                par().mass, par().M5);
-    env().setObject(getName(), fMatPt);
-}
+{}
 
 END_MODULE_NAMESPACE
 
 END_HADRONS_NAMESPACE
 
-#endif // Hadrons_DWF_hpp_
+#endif // Hadrons_MAction_DWF_hpp_

extras/Hadrons/Modules/MAction/Wilson.hpp

@@ -4,10 +4,10 @@ Grid physics library, www.github.com/paboyle/Grid
 
 Source file: extras/Hadrons/Modules/MAction/Wilson.hpp
 
-Copyright (C) 2015
-Copyright (C) 2016
+Copyright (C) 2015-2018
 
 Author: Antonin Portelli <antonin.portelli@me.com>
+Author: Lanny91 <andrew.lawson@gmail.com>
 
 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
@@ -27,8 +27,8 @@ See the full license in the file "LICENSE" in the top level distribution directo
 *************************************************************************************/
 /*  END LEGAL */
 
-#ifndef Hadrons_Wilson_hpp_
-#define Hadrons_Wilson_hpp_
+#ifndef Hadrons_MAction_Wilson_hpp_
+#define Hadrons_MAction_Wilson_hpp_
 
 #include <Grid/Hadrons/Global.hpp>
 #include <Grid/Hadrons/Module.hpp>
@@ -46,14 +46,15 @@ class WilsonPar: Serializable
 public:
     GRID_SERIALIZABLE_CLASS_MEMBERS(WilsonPar,
                                     std::string, gauge,
-                                    double     , mass);
+                                    double     , mass,
+                                    std::string, boundary);
 };
 
 template <typename FImpl>
 class TWilson: public Module<WilsonPar>
 {
 public:
-    TYPE_ALIASES(FImpl,);
+    FGS_TYPE_ALIASES(FImpl,);
 public:
     // constructor
     TWilson(const std::string name);
@@ -62,6 +63,7 @@ public:
     // dependencies/products
     virtual std::vector<std::string> getInput(void);
     virtual std::vector<std::string> getOutput(void);
+protected:
     // setup
     virtual void setup(void);
     // execution
@@ -100,24 +102,24 @@ std::vector<std::string> TWilson<FImpl>::getOutput(void)
 template <typename FImpl>
 void TWilson<FImpl>::setup(void)
 {
-    unsigned int size;
-    
-    size = 2*env().template lattice4dSize<typename FImpl::DoubledGaugeField>();
-    env().registerObject(getName(), size);
+    LOG(Message) << "Setting up TWilson fermion matrix with m= " << par().mass
+                 << " using gauge field '" << par().gauge << "'" << std::endl;
+    LOG(Message) << "Fermion boundary conditions: " << par().boundary
+                 << std::endl;
+                 
+    auto &U      = envGet(LatticeGaugeField, par().gauge);
+    auto &grid   = *env().getGrid();
+    auto &gridRb = *env().getRbGrid();
+    std::vector<Complex> boundary = strToVec<Complex>(par().boundary);
+    typename WilsonFermion<FImpl>::ImplParams implParams(boundary);
+    envCreateDerived(FMat, WilsonFermion<FImpl>, getName(), 1, U, grid, gridRb,
+                     par().mass, implParams);
 }
 
 // execution ///////////////////////////////////////////////////////////////////
 template <typename FImpl>
 void TWilson<FImpl>::execute()
-{
-    LOG(Message) << "Setting up TWilson fermion matrix with m= " << par().mass
-                 << " using gauge field '" << par().gauge << "'" << std::endl;
-    auto &U      = *env().template getObject<LatticeGaugeField>(par().gauge);
-    auto &grid   = *env().getGrid();
-    auto &gridRb = *env().getRbGrid();
-    FMat *fMatPt = new WilsonFermion<FImpl>(U, grid, gridRb, par().mass);
-    env().setObject(getName(), fMatPt);
-}
+{}
 
 END_MODULE_NAMESPACE
 

extras/Hadrons/Modules/MAction/WilsonClover.hpp

@@ -0,0 +1,153 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MAction/Wilson.hpp
+
+Copyright (C) 2015
+Copyright (C) 2016
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+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 */
+
+#ifndef Hadrons_MAction_WilsonClover_hpp_
+#define Hadrons_MAction_WilsonClover_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                            TWilson quark action                            *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MAction)
+
+class WilsonCloverPar: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(WilsonCloverPar,
+                                    std::string, gauge,
+                                    double     , mass,
+				                    double     , csw_r,
+				                    double     , csw_t,
+				                    WilsonAnisotropyCoefficients ,clover_anisotropy,
+                                    std::string, boundary
+				    );
+};
+
+template <typename FImpl>
+class TWilsonClover: public Module<WilsonCloverPar>
+{
+public:
+    FGS_TYPE_ALIASES(FImpl,);
+public:
+    // constructor
+    TWilsonClover(const std::string name);
+    // destructor
+    virtual ~TWilsonClover(void) = default;
+    // dependencies/products
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+    // setup
+    virtual void setup(void);
+    // execution
+    virtual void execute(void);
+};
+
+MODULE_REGISTER_NS(WilsonClover, TWilsonClover<FIMPL>, MAction);
+
+/******************************************************************************
+ *                     TWilsonClover template implementation                        *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+template <typename FImpl>
+TWilsonClover<FImpl>::TWilsonClover(const std::string name)
+: Module<WilsonCloverPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+template <typename FImpl>
+std::vector<std::string> TWilsonClover<FImpl>::getInput(void)
+{
+    std::vector<std::string> in = {par().gauge};
+
+    return in;
+}
+
+template <typename FImpl>
+std::vector<std::string> TWilsonClover<FImpl>::getOutput(void)
+{
+    std::vector<std::string> out = {getName()};
+
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TWilsonClover<FImpl>::setup(void)
+{
+    //unsigned int size;
+
+    // size = 2*env().template lattice4dSize<typename FImpl::DoubledGaugeField>();
+    // env().registerObject(getName(), size);
+
+
+    LOG(Message) << "Setting up TWilsonClover fermion matrix with m= " << par().mass
+                 << " using gauge field '" << par().gauge << "'" << std::endl;
+    LOG(Message) << "Fermion boundary conditions: " << par().boundary 
+                 << std::endl;
+    LOG(Message) << "Clover term csw_r: " << par().csw_r
+                 << " csw_t: " << par().csw_t
+                 << std::endl;
+    auto &U      = envGet(LatticeGaugeField, par().gauge);
+    auto &grid   = *env().getGrid();
+    auto &gridRb = *env().getRbGrid();
+    std::vector<Complex> boundary = strToVec<Complex>(par().boundary);
+    typename WilsonCloverFermion<FImpl>::ImplParams implParams(boundary);
+    envCreateDerived(FMat, WilsonCloverFermion<FImpl>, getName(), 1, U, grid, gridRb, par().mass,
+						  par().csw_r,
+						  par().csw_t,
+					      par().clover_anisotropy,
+						  implParams); 
+
+
+    //FMat *fMatPt = new WilsonCloverFermion<FImpl>(U, grid, gridRb, par().mass,
+	//					  par().csw_r,
+	//					  par().csw_t,
+	//				      par().clover_anisotropy,
+	//					  implParams);
+    //env().setObject(getName(), fMatPt);
+
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TWilsonClover<FImpl>::execute()
+{
+
+}
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_WilsonClover_hpp_

extras/Hadrons/Modules/MContraction/Baryon.hpp

@@ -4,10 +4,10 @@ Grid physics library, www.github.com/paboyle/Grid
 
 Source file: extras/Hadrons/Modules/MContraction/Baryon.hpp
 
-Copyright (C) 2015
-Copyright (C) 2016
+Copyright (C) 2015-2018
 
 Author: Antonin Portelli <antonin.portelli@me.com>
+Author: Lanny91 <andrew.lawson@gmail.com>
 
 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
@@ -27,8 +27,8 @@ See the full license in the file "LICENSE" in the top level distribution directo
 *************************************************************************************/
 /*  END LEGAL */
 
-#ifndef Hadrons_Baryon_hpp_
-#define Hadrons_Baryon_hpp_
+#ifndef Hadrons_MContraction_Baryon_hpp_
+#define Hadrons_MContraction_Baryon_hpp_
 
 #include <Grid/Hadrons/Global.hpp>
 #include <Grid/Hadrons/Module.hpp>
@@ -55,9 +55,9 @@ template <typename FImpl1, typename FImpl2, typename FImpl3>
 class TBaryon: public Module<BaryonPar>
 {
 public:
-    TYPE_ALIASES(FImpl1, 1);
-    TYPE_ALIASES(FImpl2, 2);
-    TYPE_ALIASES(FImpl3, 3);
+    FERM_TYPE_ALIASES(FImpl1, 1);
+    FERM_TYPE_ALIASES(FImpl2, 2);
+    FERM_TYPE_ALIASES(FImpl3, 3);
     class Result: Serializable
     {
     public:
@@ -72,6 +72,9 @@ public:
     // dependency relation
     virtual std::vector<std::string> getInput(void);
     virtual std::vector<std::string> getOutput(void);
+protected:
+    // setup
+    virtual void setup(void);
     // execution
     virtual void execute(void);
 };
@@ -99,11 +102,18 @@ std::vector<std::string> TBaryon<FImpl1, FImpl2, FImpl3>::getInput(void)
 template <typename FImpl1, typename FImpl2, typename FImpl3>
 std::vector<std::string> TBaryon<FImpl1, FImpl2, FImpl3>::getOutput(void)
 {
-    std::vector<std::string> out = {getName()};
+    std::vector<std::string> out = {};
     
     return out;
 }
 
+// setup ///////////////////////////////////////////////////////////////////////
+template <typename FImpl1, typename FImpl2, typename FImpl3>
+void TBaryon<FImpl1, FImpl2, FImpl3>::setup(void)
+{
+    envTmpLat(LatticeComplex, "c");
+}
+
 // execution ///////////////////////////////////////////////////////////////////
 template <typename FImpl1, typename FImpl2, typename FImpl3>
 void TBaryon<FImpl1, FImpl2, FImpl3>::execute(void)
@@ -112,20 +122,20 @@ void TBaryon<FImpl1, FImpl2, FImpl3>::execute(void)
                  << " quarks '" << par().q1 << "', '" << par().q2 << "', and '"
                  << par().q3 << "'" << std::endl;
     
-    XmlWriter             writer(par().output);
-    PropagatorField1      &q1 = *env().template getObject<PropagatorField1>(par().q1);
-    PropagatorField2      &q2 = *env().template getObject<PropagatorField2>(par().q2);
-    PropagatorField3      &q3 = *env().template getObject<PropagatorField3>(par().q2);
-    LatticeComplex        c(env().getGrid());
-    Result                result;
+    ResultWriter writer(RESULT_FILE_NAME(par().output));
+    auto       &q1 = envGet(PropagatorField1, par().q1);
+    auto       &q2 = envGet(PropagatorField2, par().q2);
+    auto       &q3 = envGet(PropagatorField3, par().q2);
+    envGetTmp(LatticeComplex, c);
+    Result     result;
     
     // FIXME: do contractions
     
-    write(writer, "meson", result);
+    // write(writer, "meson", result);
 }
 
 END_MODULE_NAMESPACE
 
 END_HADRONS_NAMESPACE
 
-#endif // Hadrons_Baryon_hpp_
+#endif // Hadrons_MContraction_Baryon_hpp_

extras/Hadrons/Modules/MContraction/DiscLoop.hpp

@@ -0,0 +1,146 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MContraction/DiscLoop.hpp
+
+Copyright (C) 2015-2018
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+Author: Lanny91 <andrew.lawson@gmail.com>
+
+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 */
+
+#ifndef Hadrons_MContraction_DiscLoop_hpp_
+#define Hadrons_MContraction_DiscLoop_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                                DiscLoop                                    *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MContraction)
+
+class DiscLoopPar: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(DiscLoopPar,
+                                    std::string,    q_loop,
+                                    Gamma::Algebra, gamma,
+                                    std::string,    output);
+};
+
+template <typename FImpl>
+class TDiscLoop: public Module<DiscLoopPar>
+{
+    FERM_TYPE_ALIASES(FImpl,);
+    class Result: Serializable
+    {
+    public:
+        GRID_SERIALIZABLE_CLASS_MEMBERS(Result,
+                                        Gamma::Algebra, gamma,
+                                        std::vector<Complex>, corr);
+    };
+public:
+    // constructor
+    TDiscLoop(const std::string name);
+    // destructor
+    virtual ~TDiscLoop(void) = default;
+    // dependency relation
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+protected:
+    // setup
+    virtual void setup(void);
+    // execution
+    virtual void execute(void);
+};
+
+MODULE_REGISTER_NS(DiscLoop, TDiscLoop<FIMPL>, MContraction);
+
+/******************************************************************************
+ *                       TDiscLoop implementation                             *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+template <typename FImpl>
+TDiscLoop<FImpl>::TDiscLoop(const std::string name)
+: Module<DiscLoopPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+template <typename FImpl>
+std::vector<std::string> TDiscLoop<FImpl>::getInput(void)
+{
+    std::vector<std::string> in = {par().q_loop};
+    
+    return in;
+}
+
+template <typename FImpl>
+std::vector<std::string> TDiscLoop<FImpl>::getOutput(void)
+{
+    std::vector<std::string> out = {};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TDiscLoop<FImpl>::setup(void)
+{
+    envTmpLat(LatticeComplex, "c");
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TDiscLoop<FImpl>::execute(void)
+{
+    LOG(Message) << "Computing disconnected loop contraction '" << getName() 
+                 << "' using '" << par().q_loop << "' with " << par().gamma 
+                 << " insertion." << std::endl;
+
+    ResultWriter          writer(RESULT_FILE_NAME(par().output));
+    auto                  &q_loop = envGet(PropagatorField, par().q_loop);
+    Gamma                 gamma(par().gamma);
+    std::vector<TComplex> buf;
+    Result                result;
+
+    envGetTmp(LatticeComplex, c);
+    c = trace(gamma*q_loop);
+    sliceSum(c, buf, Tp);
+
+    result.gamma = par().gamma;
+    result.corr.resize(buf.size());
+    for (unsigned int t = 0; t < buf.size(); ++t)
+    {
+        result.corr[t] = TensorRemove(buf[t]);
+    }
+
+    write(writer, "disc", result);
+}
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MContraction_DiscLoop_hpp_

extras/Hadrons/Modules/MContraction/Gamma3pt.hpp

@@ -0,0 +1,186 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MContraction/Gamma3pt.hpp
+
+Copyright (C) 2015-2018
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+Author: Lanny91 <andrew.lawson@gmail.com>
+
+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 */
+
+#ifndef Hadrons_MContraction_Gamma3pt_hpp_
+#define Hadrons_MContraction_Gamma3pt_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/*
+ * 3pt contraction with gamma matrix insertion.
+ *
+ * Schematic:
+ *
+ *             q2           q3
+ *        /----<------*------<----¬
+ *       /          gamma          \
+ *      /                           \
+ *   i *                            * f
+ *      \                          /
+ *       \                        /
+ *        \----------->----------/
+ *                   q1
+ *
+ *      trace(g5*q1*adj(q2)*g5*gamma*q3)
+ * 
+ *  options:
+ *   - q1: sink smeared propagator, source at i
+ *   - q2: propagator, source at i
+ *   - q3: propagator, source at f
+ *   - gamma: gamma matrix to insert
+ *   - tSnk: sink position for propagator q1.
+ *
+ */
+
+/******************************************************************************
+ *                               Gamma3pt                                     *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MContraction)
+
+class Gamma3ptPar: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(Gamma3ptPar,
+                                    std::string,    q1,
+                                    std::string,    q2,
+                                    std::string,    q3,
+                                    Gamma::Algebra, gamma,
+                                    unsigned int,   tSnk,
+                                    std::string,    output);
+};
+
+template <typename FImpl1, typename FImpl2, typename FImpl3>
+class TGamma3pt: public Module<Gamma3ptPar>
+{
+    FERM_TYPE_ALIASES(FImpl1, 1);
+    FERM_TYPE_ALIASES(FImpl2, 2);
+    FERM_TYPE_ALIASES(FImpl3, 3);
+    class Result: Serializable
+    {
+    public:
+        GRID_SERIALIZABLE_CLASS_MEMBERS(Result,
+                                        Gamma::Algebra, gamma,
+                                        std::vector<Complex>, corr);
+    };
+public:
+    // constructor
+    TGamma3pt(const std::string name);
+    // destructor
+    virtual ~TGamma3pt(void) = default;
+    // dependency relation
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+protected:
+    // setup
+    virtual void setup(void);
+    // execution
+    virtual void execute(void);
+};
+
+MODULE_REGISTER_NS(Gamma3pt, ARG(TGamma3pt<FIMPL, FIMPL, FIMPL>), MContraction);
+
+/******************************************************************************
+ *                       TGamma3pt implementation                             *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+template <typename FImpl1, typename FImpl2, typename FImpl3>
+TGamma3pt<FImpl1, FImpl2, FImpl3>::TGamma3pt(const std::string name)
+: Module<Gamma3ptPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+template <typename FImpl1, typename FImpl2, typename FImpl3>
+std::vector<std::string> TGamma3pt<FImpl1, FImpl2, FImpl3>::getInput(void)
+{
+    std::vector<std::string> in = {par().q1, par().q2, par().q3};
+    
+    return in;
+}
+
+template <typename FImpl1, typename FImpl2, typename FImpl3>
+std::vector<std::string> TGamma3pt<FImpl1, FImpl2, FImpl3>::getOutput(void)
+{
+    std::vector<std::string> out = {};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+template <typename FImpl1, typename FImpl2, typename FImpl3>
+void TGamma3pt<FImpl1, FImpl2, FImpl3>::setup(void)
+{
+    envTmpLat(LatticeComplex, "c");
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+template <typename FImpl1, typename FImpl2, typename FImpl3>
+void TGamma3pt<FImpl1, FImpl2, FImpl3>::execute(void)
+{
+    LOG(Message) << "Computing 3pt contractions '" << getName() << "' using"
+                 << " quarks '" << par().q1 << "', '" << par().q2 << "' and '"
+                 << par().q3 << "', with " << par().gamma << " insertion." 
+                 << std::endl;
+
+    // Initialise variables. q2 and q3 are normal propagators, q1 may be 
+    // sink smeared.
+    ResultWriter          writer(RESULT_FILE_NAME(par().output));
+    auto                  &q1 = envGet(SlicedPropagator1, par().q1);
+    auto                  &q2 = envGet(PropagatorField2, par().q2);
+    auto                  &q3 = envGet(PropagatorField2, par().q3);
+    Gamma                 g5(Gamma::Algebra::Gamma5);
+    Gamma                 gamma(par().gamma);
+    std::vector<TComplex> buf;
+    Result                result;
+    
+    // Extract relevant timeslice of sinked propagator q1, then contract &
+    // sum over all spacial positions of gamma insertion.
+    SitePropagator1 q1Snk = q1[par().tSnk];
+    envGetTmp(LatticeComplex, c);
+    c = trace(g5*q1Snk*adj(q2)*(g5*gamma)*q3);
+    sliceSum(c, buf, Tp);
+
+    result.gamma = par().gamma;
+    result.corr.resize(buf.size());
+    for (unsigned int t = 0; t < buf.size(); ++t)
+    {
+        result.corr[t] = TensorRemove(buf[t]);
+    }
+
+    write(writer, "gamma3pt", result);
+}
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MContraction_Gamma3pt_hpp_

extras/Hadrons/Modules/MContraction/Meson.hpp

@@ -4,10 +4,10 @@ Grid physics library, www.github.com/paboyle/Grid
 
 Source file: extras/Hadrons/Modules/MContraction/Meson.hpp
 
-Copyright (C) 2015
-Copyright (C) 2016
+Copyright (C) 2015-2018
 
 Author: Antonin Portelli <antonin.portelli@me.com>
+Author: Lanny91 <andrew.lawson@gmail.com>
 
 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
@@ -27,8 +27,8 @@ See the full license in the file "LICENSE" in the top level distribution directo
 *************************************************************************************/
 /*  END LEGAL */
 
-#ifndef Hadrons_Meson_hpp_
-#define Hadrons_Meson_hpp_
+#ifndef Hadrons_MContraction_Meson_hpp_
+#define Hadrons_MContraction_Meson_hpp_
 
 #include <Grid/Hadrons/Global.hpp>
 #include <Grid/Hadrons/Module.hpp>
@@ -36,32 +36,55 @@ See the full license in the file "LICENSE" in the top level distribution directo
 
 BEGIN_HADRONS_NAMESPACE
 
+/*
+ 
+ Meson contractions
+ -----------------------------
+ 
+ * options:
+ - q1: input propagator 1 (string)
+ - q2: input propagator 2 (string)
+ - gammas: gamma products to insert at sink & source, pairs of gamma matrices 
+           (space-separated strings) in angled brackets (i.e. <g_sink g_src>),
+           in a sequence (e.g. "<Gamma5 Gamma5><Gamma5 GammaT>").
+
+           Special values: "all" - perform all possible contractions.
+ - sink: module to compute the sink to use in contraction (string).
+*/
+
 /******************************************************************************
  *                                TMeson                                       *
  ******************************************************************************/
 BEGIN_MODULE_NAMESPACE(MContraction)
 
+typedef std::pair<Gamma::Algebra, Gamma::Algebra> GammaPair;
+
 class MesonPar: Serializable
 {
 public:
     GRID_SERIALIZABLE_CLASS_MEMBERS(MesonPar,
-                                    std::string,    q1,
-                                    std::string,    q2,
-                                    std::string,    output,
-                                    Gamma::Algebra, gammaSource,
-                                    Gamma::Algebra, gammaSink);
+                                    std::string, q1,
+                                    std::string, q2,
+                                    std::string, gammas,
+                                    std::string, sink,
+                                    std::string, output);
 };
 
 template <typename FImpl1, typename FImpl2>
 class TMeson: public Module<MesonPar>
 {
 public:
-    TYPE_ALIASES(FImpl1, 1);
-    TYPE_ALIASES(FImpl2, 2);
+    FERM_TYPE_ALIASES(FImpl1, 1);
+    FERM_TYPE_ALIASES(FImpl2, 2);
+    FERM_TYPE_ALIASES(ScalarImplCR, Scalar);
+    SINK_TYPE_ALIASES(Scalar);
     class Result: Serializable
     {
     public:
-        GRID_SERIALIZABLE_CLASS_MEMBERS(Result, std::vector<Complex>, corr);
+        GRID_SERIALIZABLE_CLASS_MEMBERS(Result,
+                                        Gamma::Algebra, gamma_snk,
+                                        Gamma::Algebra, gamma_src,
+                                        std::vector<Complex>, corr);
     };
 public:
     // constructor
@@ -71,6 +94,10 @@ public:
     // dependencies/products
     virtual std::vector<std::string> getInput(void);
     virtual std::vector<std::string> getOutput(void);
+    virtual void parseGammaString(std::vector<GammaPair> &gammaList);
+protected:
+    // execution
+    virtual void setup(void);
     // execution
     virtual void execute(void);
 };
@@ -90,7 +117,7 @@ TMeson<FImpl1, FImpl2>::TMeson(const std::string name)
 template <typename FImpl1, typename FImpl2>
 std::vector<std::string> TMeson<FImpl1, FImpl2>::getInput(void)
 {
-    std::vector<std::string> input = {par().q1, par().q2};
+    std::vector<std::string> input = {par().q1, par().q2, par().sink};
     
     return input;
 }
@@ -98,12 +125,46 @@ std::vector<std::string> TMeson<FImpl1, FImpl2>::getInput(void)
 template <typename FImpl1, typename FImpl2>
 std::vector<std::string> TMeson<FImpl1, FImpl2>::getOutput(void)
 {
-    std::vector<std::string> output = {getName()};
+    std::vector<std::string> output = {};
     
     return output;
 }
 
+template <typename FImpl1, typename FImpl2>
+void TMeson<FImpl1, FImpl2>::parseGammaString(std::vector<GammaPair> &gammaList)
+{
+    gammaList.clear();
+    // Determine gamma matrices to insert at source/sink.
+    if (par().gammas.compare("all") == 0)
+    {
+        // Do all contractions.
+        for (unsigned int i = 1; i < Gamma::nGamma; i += 2)
+        {
+            for (unsigned int j = 1; j < Gamma::nGamma; j += 2)
+            {
+                gammaList.push_back(std::make_pair((Gamma::Algebra)i, 
+                                                   (Gamma::Algebra)j));
+            }
+        }
+    }
+    else
+    {
+        // Parse individual contractions from input string.
+        gammaList = strToVec<GammaPair>(par().gammas);
+    } 
+}
+
 // execution ///////////////////////////////////////////////////////////////////
+template <typename FImpl1, typename FImpl2>
+void TMeson<FImpl1, FImpl2>::setup(void)
+{
+    envTmpLat(LatticeComplex, "c");
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+#define mesonConnected(q1, q2, gSnk, gSrc) \
+(g5*(gSnk))*(q1)*(adj(gSrc)*g5)*adj(q2)
+
 template <typename FImpl1, typename FImpl2>
 void TMeson<FImpl1, FImpl2>::execute(void)
 {
@@ -111,21 +172,72 @@ void TMeson<FImpl1, FImpl2>::execute(void)
                  << " quarks '" << par().q1 << "' and '" << par().q2 << "'"
                  << std::endl;
     
-    XmlWriter             writer(par().output);
-    PropagatorField1      &q1 = *env().template getObject<PropagatorField1>(par().q1);
-    PropagatorField2      &q2 = *env().template getObject<PropagatorField2>(par().q2);
-    LatticeComplex        c(env().getGrid());
-    Gamma                 gSrc(par().gammaSource), gSnk(par().gammaSink);
-    Gamma                 g5(Gamma::Algebra::Gamma5);
-    std::vector<TComplex> buf;
-    Result                result;
+    ResultWriter           writer(RESULT_FILE_NAME(par().output));
+    std::vector<TComplex>  buf;
+    std::vector<Result>    result;
+    Gamma                  g5(Gamma::Algebra::Gamma5);
+    std::vector<GammaPair> gammaList;
+    int                    nt = env().getDim(Tp);
     
-    c = trace(gSnk*q1*adj(gSrc)*g5*adj(q2)*g5);
-    sliceSum(c, buf, Tp);
-    result.corr.resize(buf.size());
-    for (unsigned int t = 0; t < buf.size(); ++t)
+    parseGammaString(gammaList);
+    result.resize(gammaList.size());
+    for (unsigned int i = 0; i < result.size(); ++i)
     {
-        result.corr[t] = TensorRemove(buf[t]);
+        result[i].gamma_snk = gammaList[i].first;
+        result[i].gamma_src = gammaList[i].second;
+        result[i].corr.resize(nt);
+    }
+    if (envHasType(SlicedPropagator1, par().q1) and
+        envHasType(SlicedPropagator2, par().q2))
+    {
+        auto &q1 = envGet(SlicedPropagator1, par().q1);
+        auto &q2 = envGet(SlicedPropagator2, par().q2);
+        
+        LOG(Message) << "(propagator already sinked)" << std::endl;
+        for (unsigned int i = 0; i < result.size(); ++i)
+        {
+            Gamma gSnk(gammaList[i].first);
+            Gamma gSrc(gammaList[i].second);
+            
+            for (unsigned int t = 0; t < buf.size(); ++t)
+            {
+                result[i].corr[t] = TensorRemove(trace(mesonConnected(q1[t], q2[t], gSnk, gSrc)));
+            }
+        }
+    }
+    else
+    {
+        auto &q1 = envGet(PropagatorField1, par().q1);
+        auto &q2 = envGet(PropagatorField2, par().q2);
+        
+        envGetTmp(LatticeComplex, c);
+        LOG(Message) << "(using sink '" << par().sink << "')" << std::endl;
+        for (unsigned int i = 0; i < result.size(); ++i)
+        {
+            Gamma       gSnk(gammaList[i].first);
+            Gamma       gSrc(gammaList[i].second);
+            std::string ns;
+                
+            ns = vm().getModuleNamespace(env().getObjectModule(par().sink));
+            if (ns == "MSource")
+            {
+                PropagatorField1 &sink = envGet(PropagatorField1, par().sink);
+                
+                c = trace(mesonConnected(q1, q2, gSnk, gSrc)*sink);
+                sliceSum(c, buf, Tp);
+            }
+            else if (ns == "MSink")
+            {
+                SinkFnScalar &sink = envGet(SinkFnScalar, par().sink);
+                
+                c   = trace(mesonConnected(q1, q2, gSnk, gSrc));
+                buf = sink(c);
+            }
+            for (unsigned int t = 0; t < buf.size(); ++t)
+            {
+                result[i].corr[t] = TensorRemove(buf[t]);
+            }
+        }
     }
     write(writer, "meson", result);
 }
@@ -134,4 +246,4 @@ END_MODULE_NAMESPACE
 
 END_HADRONS_NAMESPACE
 
-#endif // Hadrons_Meson_hpp_
+#endif // Hadrons_MContraction_Meson_hpp_

extras/Hadrons/Modules/MContraction/WardIdentity.hpp

@@ -0,0 +1,224 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MContraction/WardIdentity.hpp
+
+Copyright (C) 2015-2018
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+Author: Lanny91 <andrew.lawson@gmail.com>
+
+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 */
+
+#ifndef Hadrons_MContraction_WardIdentity_hpp_
+#define Hadrons_MContraction_WardIdentity_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/*
+  Ward Identity contractions
+ -----------------------------
+ 
+ * options:
+ - q:          propagator, 5D if available (string)
+ - action:     action module used for propagator solution (string)
+ - mass:       mass of quark (double)
+ - test_axial: whether or not to test PCAC relation.
+*/
+
+/******************************************************************************
+ *                              WardIdentity                                  *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MContraction)
+
+class WardIdentityPar: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(WardIdentityPar,
+                                    std::string, q,
+                                    std::string, action,
+                                    double,      mass,
+                                    bool,        test_axial);
+};
+
+template <typename FImpl>
+class TWardIdentity: public Module<WardIdentityPar>
+{
+public:
+    FERM_TYPE_ALIASES(FImpl,);
+public:
+    // constructor
+    TWardIdentity(const std::string name);
+    // destructor
+    virtual ~TWardIdentity(void) = default;
+    // dependency relation
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+protected:
+    // setup
+    virtual void setup(void);
+    // execution
+    virtual void execute(void);
+private:
+    unsigned int Ls_;
+};
+
+MODULE_REGISTER_NS(WardIdentity, TWardIdentity<FIMPL>, MContraction);
+
+/******************************************************************************
+ *                     TWardIdentity implementation                           *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+template <typename FImpl>
+TWardIdentity<FImpl>::TWardIdentity(const std::string name)
+: Module<WardIdentityPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+template <typename FImpl>
+std::vector<std::string> TWardIdentity<FImpl>::getInput(void)
+{
+    std::vector<std::string> in = {par().q, par().action};
+    
+    return in;
+}
+
+template <typename FImpl>
+std::vector<std::string> TWardIdentity<FImpl>::getOutput(void)
+{
+    std::vector<std::string> out = {};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TWardIdentity<FImpl>::setup(void)
+{
+    Ls_ = env().getObjectLs(par().q);
+    if (Ls_ != env().getObjectLs(par().action))
+    {
+        HADRON_ERROR(Size, "Ls mismatch between quark action and propagator");
+    }
+    envTmpLat(PropagatorField, "tmp");
+    envTmpLat(PropagatorField, "vector_WI");
+    if (par().test_axial)
+    {
+        envTmpLat(PropagatorField, "psi");
+        envTmpLat(LatticeComplex,  "PP");
+        envTmpLat(LatticeComplex,  "axial_defect");
+        envTmpLat(LatticeComplex,  "PJ5q");
+    }
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TWardIdentity<FImpl>::execute(void)
+{
+    LOG(Message) << "Performing Ward Identity checks for quark '" << par().q
+                 << "'." << std::endl;
+
+    auto  &q   = envGet(PropagatorField, par().q);
+    auto  &act = envGet(FMat, par().action);
+    Gamma g5(Gamma::Algebra::Gamma5);
+
+    // Compute D_mu V_mu, D here is backward derivative.
+    envGetTmp(PropagatorField, tmp);
+    envGetTmp(PropagatorField, vector_WI);
+    vector_WI    = zero;
+    for (unsigned int mu = 0; mu < Nd; ++mu)
+    {
+        act.ContractConservedCurrent(q, q, tmp, Current::Vector, mu);
+        tmp -= Cshift(tmp, mu, -1);
+        vector_WI += tmp;
+    }
+
+    // Test ward identity D_mu V_mu = 0;
+    LOG(Message) << "Vector Ward Identity check Delta_mu V_mu = " 
+                 << norm2(vector_WI) << std::endl;
+
+    if (par().test_axial)
+    {
+        envGetTmp(PropagatorField, psi);
+        envGetTmp(LatticeComplex, PP);
+        envGetTmp(LatticeComplex, axial_defect);
+        envGetTmp(LatticeComplex, PJ5q);
+        std::vector<TComplex> axial_buf;
+
+        // Compute <P|D_mu A_mu>, D is backwards derivative.
+        axial_defect = zero;
+        for (unsigned int mu = 0; mu < Nd; ++mu)
+        {
+            act.ContractConservedCurrent(q, q, tmp, Current::Axial, mu);
+            tmp -= Cshift(tmp, mu, -1);
+            axial_defect += trace(g5*tmp);
+        }
+
+        // Get <P|J5q> for 5D (zero for 4D) and <P|P>.
+        PJ5q = zero;
+        if (Ls_ > 1)
+        {
+            // <P|P>
+            ExtractSlice(tmp, q, 0, 0);
+            psi  = 0.5 * (tmp - g5*tmp);
+            ExtractSlice(tmp, q, Ls_ - 1, 0);
+            psi += 0.5 * (tmp + g5*tmp);
+            PP = trace(adj(psi)*psi);
+
+            // <P|5Jq>
+            ExtractSlice(tmp, q, Ls_/2 - 1, 0);
+            psi  = 0.5 * (tmp + g5*tmp);
+            ExtractSlice(tmp, q, Ls_/2, 0);
+            psi += 0.5 * (tmp - g5*tmp);
+            PJ5q = trace(adj(psi)*psi);
+        }
+        else
+        {
+            PP = trace(adj(q)*q);
+        }
+
+        // Test ward identity <P|D_mu A_mu> = 2m<P|P> + 2<P|J5q>
+        LOG(Message) << "|D_mu A_mu|^2 = " << norm2(axial_defect) << std::endl;
+        LOG(Message) << "|PP|^2        = " << norm2(PP) << std::endl;
+        LOG(Message) << "|PJ5q|^2      = " << norm2(PJ5q) << std::endl;
+        LOG(Message) << "Axial Ward Identity defect Delta_mu A_mu = "
+                     << norm2(axial_defect) << std::endl;
+    
+        // Axial defect by timeslice.
+        axial_defect -= 2.*(par().mass*PP + PJ5q);
+        LOG(Message) << "Check Axial defect by timeslice" << std::endl;
+        sliceSum(axial_defect, axial_buf, Tp);
+        for (int t = 0; t < axial_buf.size(); ++t)
+        {
+            LOG(Message) << "t = " << t << ": " 
+                         << TensorRemove(axial_buf[t]) << std::endl;
+        }
+    }
+}
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_WardIdentity_hpp_

extras/Hadrons/Modules/MContraction/WeakHamiltonian.hpp

@@ -0,0 +1,118 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MContraction/WeakHamiltonian.hpp
+
+Copyright (C) 2015-2018
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+Author: Lanny91 <andrew.lawson@gmail.com>
+
+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 */
+
+#ifndef Hadrons_MContraction_WeakHamiltonian_hpp_
+#define Hadrons_MContraction_WeakHamiltonian_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                         WeakHamiltonian                                    *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MContraction)
+
+/*******************************************************************************
+ * Utilities for contractions involving the Weak Hamiltonian.
+ ******************************************************************************/
+//// Sum and store correlator.
+#define MAKE_DIAG(exp, buf, res, n)\
+sliceSum(exp, buf, Tp);\
+res.name = (n);\
+res.corr.resize(buf.size());\
+for (unsigned int t = 0; t < buf.size(); ++t)\
+{\
+    res.corr[t] = TensorRemove(buf[t]);\
+}
+
+//// Contraction of mu index: use 'mu' variable in exp.
+#define SUM_MU(buf,exp)\
+buf = zero;\
+for (unsigned int mu = 0; mu < ndim; ++mu)\
+{\
+    buf += exp;\
+}
+
+enum 
+{
+  i_V = 0,
+  i_A = 1,
+  n_i = 2
+};
+
+class WeakHamiltonianPar: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(WeakHamiltonianPar,
+                                    std::string, q1,
+                                    std::string, q2,
+                                    std::string, q3,
+                                    std::string, q4,
+                                    unsigned int, tSnk,
+                                    std::string, output);
+};
+
+#define MAKE_WEAK_MODULE(modname)\
+class T##modname: public Module<WeakHamiltonianPar>\
+{\
+public:\
+    FERM_TYPE_ALIASES(FIMPL,)\
+    class Result: Serializable\
+    {\
+    public:\
+        GRID_SERIALIZABLE_CLASS_MEMBERS(Result,\
+                                        std::string, name,\
+                                        std::vector<Complex>, corr);\
+    };\
+public:\
+    /* constructor */ \
+    T##modname(const std::string name);\
+    /* destructor */ \
+    virtual ~T##modname(void) = default;\
+    /* dependency relation */ \
+    virtual std::vector<std::string> getInput(void);\
+    virtual std::vector<std::string> getOutput(void);\
+public:\
+    std::vector<std::string> VA_label = {"V", "A"};\
+protected:\
+    /* setup */ \
+    virtual void setup(void);\
+    /* execution */ \
+    virtual void execute(void);\
+};\
+MODULE_REGISTER_NS(modname, T##modname, MContraction);
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MContraction_WeakHamiltonian_hpp_

extras/Hadrons/Modules/MContraction/WeakHamiltonianEye.cc

@@ -0,0 +1,151 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MContraction/WeakHamiltonianEye.cc
+
+Copyright (C) 2015-2018
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+Author: Lanny91 <andrew.lawson@gmail.com>
+
+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/Hadrons/Modules/MContraction/WeakHamiltonianEye.hpp>
+
+using namespace Grid;
+using namespace Hadrons;
+using namespace MContraction;
+
+/*
+ * Weak Hamiltonian current-current contractions, Eye-type.
+ * 
+ * These contractions are generated by the Q1 and Q2 operators in the physical
+ * basis (see e.g. Fig 3 of arXiv:1507.03094).
+ * 
+ * Schematics:        q4                 |                  
+ *                  /-<-¬                |                             
+ *                 /     \               |             q2           q3
+ *                 \     /               |        /----<------*------<----¬                        
+ *            q2    \   /    q3          |       /          /-*-¬          \
+ *       /-----<-----* *-----<----¬      |      /          /     \          \
+ *    i *            H_W           * f   |   i *           \     /  q4      * f
+ *       \                        /      |      \           \->-/          /   
+ *        \                      /       |       \                        /       
+ *         \---------->---------/        |        \----------->----------/        
+ *                   q1                  |                   q1                  
+ *                                       |
+ *                Saucer (S)             |                  Eye (E)
+ * 
+ * S: trace(q3*g5*q1*adj(q2)*g5*gL[mu][p_1]*q4*gL[mu][p_2])
+ * E: trace(q3*g5*q1*adj(q2)*g5*gL[mu][p_1])*trace(q4*gL[mu][p_2])
+ * 
+ * Note q1 must be sink smeared.
+ */
+
+/******************************************************************************
+ *                  TWeakHamiltonianEye implementation                        *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+TWeakHamiltonianEye::TWeakHamiltonianEye(const std::string name)
+: Module<WeakHamiltonianPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+std::vector<std::string> TWeakHamiltonianEye::getInput(void)
+{
+    std::vector<std::string> in = {par().q1, par().q2, par().q3, par().q4};
+    
+    return in;
+}
+
+std::vector<std::string> TWeakHamiltonianEye::getOutput(void)
+{
+    std::vector<std::string> out = {};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+void TWeakHamiltonianEye::setup(void)
+{
+    unsigned int ndim = env().getNd();
+
+    envTmpLat(LatticeComplex,  "expbuf");
+    envTmpLat(PropagatorField, "tmp1");
+    envTmpLat(LatticeComplex,  "tmp2");
+    envTmp(std::vector<PropagatorField>, "S_body", 1, ndim, PropagatorField(env().getGrid()));
+    envTmp(std::vector<PropagatorField>, "S_loop", 1, ndim, PropagatorField(env().getGrid()));
+    envTmp(std::vector<LatticeComplex>,  "E_body", 1, ndim, LatticeComplex(env().getGrid()));
+    envTmp(std::vector<LatticeComplex>,  "E_loop", 1, ndim, LatticeComplex(env().getGrid()));
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+void TWeakHamiltonianEye::execute(void)
+{
+    LOG(Message) << "Computing Weak Hamiltonian (Eye type) contractions '" 
+                 << getName() << "' using quarks '" << par().q1 << "', '" 
+                 << par().q2 << ", '" << par().q3 << "' and '" << par().q4 
+                 << "'." << std::endl;
+
+    ResultWriter           writer(RESULT_FILE_NAME(par().output));
+    auto                   &q1 = envGet(SlicedPropagator, par().q1);
+    auto                   &q2 = envGet(PropagatorField, par().q2);
+    auto                   &q3 = envGet(PropagatorField, par().q3);
+    auto                   &q4 = envGet(PropagatorField, par().q4);
+    Gamma                  g5  = Gamma(Gamma::Algebra::Gamma5);
+    std::vector<TComplex>  corrbuf;
+    std::vector<Result>    result(n_eye_diag);
+    unsigned int ndim    = env().getNd();
+
+    envGetTmp(LatticeComplex,               expbuf); 
+    envGetTmp(PropagatorField,              tmp1);
+    envGetTmp(LatticeComplex,               tmp2);
+    envGetTmp(std::vector<PropagatorField>, S_body);
+    envGetTmp(std::vector<PropagatorField>, S_loop);
+    envGetTmp(std::vector<LatticeComplex>,  E_body);
+    envGetTmp(std::vector<LatticeComplex>,  E_loop);
+
+    // Get sink timeslice of q1.
+    SitePropagator q1Snk = q1[par().tSnk];
+
+    // Setup for S-type contractions.
+    for (int mu = 0; mu < ndim; ++mu)
+    {
+        S_body[mu] = MAKE_SE_BODY(q1Snk, q2, q3, GammaL(Gamma::gmu[mu]));
+        S_loop[mu] = MAKE_SE_LOOP(q4, GammaL(Gamma::gmu[mu]));
+    }
+
+    // Perform S-type contractions.    
+    SUM_MU(expbuf, trace(S_body[mu]*S_loop[mu]))
+    MAKE_DIAG(expbuf, corrbuf, result[S_diag], "HW_S")
+
+    // Recycle sub-expressions for E-type contractions.
+    for (unsigned int mu = 0; mu < ndim; ++mu)
+    {
+        E_body[mu] = trace(S_body[mu]);
+        E_loop[mu] = trace(S_loop[mu]);
+    }
+
+    // Perform E-type contractions.
+    SUM_MU(expbuf, E_body[mu]*E_loop[mu])
+    MAKE_DIAG(expbuf, corrbuf, result[E_diag], "HW_E")
+
+    write(writer, "HW_Eye", result);
+}

extras/Hadrons/Modules/MContraction/WeakHamiltonianEye.hpp

@@ -0,0 +1,59 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MContraction/WeakHamiltonianEye.hpp
+
+Copyright (C) 2015-2018
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+Author: Lanny91 <andrew.lawson@gmail.com>
+
+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 */
+
+#ifndef Hadrons_MContraction_WeakHamiltonianEye_hpp_
+#define Hadrons_MContraction_WeakHamiltonianEye_hpp_
+
+#include <Grid/Hadrons/Modules/MContraction/WeakHamiltonian.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                         WeakHamiltonianEye                                 *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MContraction)
+
+enum
+{
+    S_diag = 0,
+    E_diag = 1,
+    n_eye_diag = 2
+};
+
+// Saucer and Eye subdiagram contractions.
+#define MAKE_SE_BODY(Q_1, Q_2, Q_3, gamma) (Q_3*g5*Q_1*adj(Q_2)*g5*gamma)
+#define MAKE_SE_LOOP(Q_loop, gamma) (Q_loop*gamma)
+
+MAKE_WEAK_MODULE(WeakHamiltonianEye)
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MContraction_WeakHamiltonianEye_hpp_

extras/Hadrons/Modules/MContraction/WeakHamiltonianNonEye.cc

@@ -0,0 +1,148 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MContraction/WeakHamiltonianNonEye.cc
+
+Copyright (C) 2015-2018
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+Author: Lanny91 <andrew.lawson@gmail.com>
+
+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/Hadrons/Modules/MContraction/WeakHamiltonianNonEye.hpp>
+
+using namespace Grid;
+using namespace Hadrons;
+using namespace MContraction;
+
+/*
+ * Weak Hamiltonian current-current contractions, Non-Eye-type.
+ * 
+ * These contractions are generated by the Q1 and Q2 operators in the physical
+ * basis (see e.g. Fig 3 of arXiv:1507.03094).
+ * 
+ * Schematic:     
+ *            q2             q3          |           q2              q3
+ *          /--<--¬       /--<--¬        |        /--<--¬         /--<--¬       
+ *         /       \     /       \       |       /       \       /       \      
+ *        /         \   /         \      |      /         \     /         \     
+ *       /           \ /           \     |     /           \   /           \    
+ *    i *             * H_W         *  f |  i *             * * H_W         * f 
+ *      \             *             |    |     \           /   \           /
+ *       \           / \           /     |      \         /     \         /    
+ *        \         /   \         /      |       \       /       \       /  
+ *         \       /     \       /       |        \-->--/         \-->--/      
+ *          \-->--/       \-->--/        |          q1               q4 
+ *            q1             q4          |
+ *                Connected (C)          |                 Wing (W)
+ *
+ * C: trace(q1*adj(q2)*g5*gL[mu]*q3*adj(q4)*g5*gL[mu])
+ * W: trace(q1*adj(q2)*g5*gL[mu])*trace(q3*adj(q4)*g5*gL[mu])
+ * 
+ */
+
+/******************************************************************************
+ *                  TWeakHamiltonianNonEye implementation                     *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+TWeakHamiltonianNonEye::TWeakHamiltonianNonEye(const std::string name)
+: Module<WeakHamiltonianPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+std::vector<std::string> TWeakHamiltonianNonEye::getInput(void)
+{
+    std::vector<std::string> in = {par().q1, par().q2, par().q3, par().q4};
+    
+    return in;
+}
+
+std::vector<std::string> TWeakHamiltonianNonEye::getOutput(void)
+{
+    std::vector<std::string> out = {};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+void TWeakHamiltonianNonEye::setup(void)
+{
+    unsigned int ndim = env().getNd();
+
+    envTmpLat(LatticeComplex,  "expbuf");
+    envTmpLat(PropagatorField, "tmp1");
+    envTmpLat(LatticeComplex,  "tmp2");
+    envTmp(std::vector<PropagatorField>, "C_i_side_loop", 1, ndim, PropagatorField(env().getGrid()));
+    envTmp(std::vector<PropagatorField>, "C_f_side_loop", 1, ndim, PropagatorField(env().getGrid()));
+    envTmp(std::vector<LatticeComplex>,  "W_i_side_loop", 1, ndim, LatticeComplex(env().getGrid()));
+    envTmp(std::vector<LatticeComplex>,  "W_f_side_loop", 1, ndim, LatticeComplex(env().getGrid()));
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+void TWeakHamiltonianNonEye::execute(void)
+{
+    LOG(Message) << "Computing Weak Hamiltonian (Non-Eye type) contractions '" 
+                 << getName() << "' using quarks '" << par().q1 << "', '" 
+                 << par().q2 << ", '" << par().q3 << "' and '" << par().q4 
+                 << "'." << std::endl;
+    
+    ResultWriter          writer(RESULT_FILE_NAME(par().output));
+    auto                  &q1 = envGet(PropagatorField, par().q1);
+    auto                  &q2 = envGet(PropagatorField, par().q2);
+    auto                  &q3 = envGet(PropagatorField, par().q3);
+    auto                  &q4 = envGet(PropagatorField, par().q4);
+    Gamma                 g5  = Gamma(Gamma::Algebra::Gamma5);
+    std::vector<TComplex> corrbuf;
+    std::vector<Result>   result(n_noneye_diag); 
+    unsigned int          ndim = env().getNd();
+
+    envGetTmp(LatticeComplex,               expbuf); 
+    envGetTmp(PropagatorField,              tmp1);
+    envGetTmp(LatticeComplex,               tmp2);
+    envGetTmp(std::vector<PropagatorField>, C_i_side_loop);
+    envGetTmp(std::vector<PropagatorField>, C_f_side_loop);
+    envGetTmp(std::vector<LatticeComplex>,  W_i_side_loop);
+    envGetTmp(std::vector<LatticeComplex>,  W_f_side_loop);
+
+    // Setup for C-type contractions.
+    for (int mu = 0; mu < ndim; ++mu)
+    {
+        C_i_side_loop[mu] = MAKE_CW_SUBDIAG(q1, q2, GammaL(Gamma::gmu[mu]));
+        C_f_side_loop[mu] = MAKE_CW_SUBDIAG(q3, q4, GammaL(Gamma::gmu[mu]));
+    }
+
+    // Perform C-type contractions.    
+    SUM_MU(expbuf, trace(C_i_side_loop[mu]*C_f_side_loop[mu]))
+    MAKE_DIAG(expbuf, corrbuf, result[C_diag], "HW_C")
+
+    // Recycle sub-expressions for W-type contractions.
+    for (unsigned int mu = 0; mu < ndim; ++mu)
+    {
+        W_i_side_loop[mu] = trace(C_i_side_loop[mu]);
+        W_f_side_loop[mu] = trace(C_f_side_loop[mu]);
+    }
+
+    // Perform W-type contractions.
+    SUM_MU(expbuf, W_i_side_loop[mu]*W_f_side_loop[mu])
+    MAKE_DIAG(expbuf, corrbuf, result[W_diag], "HW_W")
+
+    write(writer, "HW_NonEye", result);
+}

extras/Hadrons/Modules/MContraction/WeakHamiltonianNonEye.hpp

@@ -0,0 +1,58 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MContraction/WeakHamiltonianNonEye.hpp
+
+Copyright (C) 2015-2018
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+Author: Lanny91 <andrew.lawson@gmail.com>
+
+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 */
+
+#ifndef Hadrons_MContraction_WeakHamiltonianNonEye_hpp_
+#define Hadrons_MContraction_WeakHamiltonianNonEye_hpp_
+
+#include <Grid/Hadrons/Modules/MContraction/WeakHamiltonian.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                         WeakHamiltonianNonEye                              *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MContraction)
+
+enum
+{
+    W_diag = 0,
+    C_diag = 1,
+    n_noneye_diag = 2
+};
+
+// Wing and Connected subdiagram contractions
+#define MAKE_CW_SUBDIAG(Q_1, Q_2, gamma) (Q_1*adj(Q_2)*g5*gamma)
+
+MAKE_WEAK_MODULE(WeakHamiltonianNonEye)
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MContraction_WeakHamiltonianNonEye_hpp_

extras/Hadrons/Modules/MContraction/WeakNeutral4ptDisc.cc

@@ -0,0 +1,142 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MContraction/WeakNeutral4ptDisc.cc
+
+Copyright (C) 2015-2018
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+Author: Lanny91 <andrew.lawson@gmail.com>
+
+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/Hadrons/Modules/MContraction/WeakNeutral4ptDisc.hpp>
+
+using namespace Grid;
+using namespace Hadrons;
+using namespace MContraction;
+
+/*
+ * Weak Hamiltonian + current contractions, disconnected topology for neutral 
+ * mesons.
+ * 
+ * These contractions are generated by operators Q_1,...,10 of the dS=1 Weak
+ * Hamiltonian in the physical basis and an additional current J (see e.g. 
+ * Fig 11 of arXiv:1507.03094).
+ * 
+ * Schematic:
+ *                        
+ *           q2          q4             q3
+ *       /--<--¬     /---<--¬       /---<--¬
+ *     /         \ /         \     /        \
+ *  i *           * H_W      |  J *          * f
+ *     \         / \         /     \        /
+ *      \--->---/   \-------/       \------/
+ *          q1 
+ * 
+ * options
+ * - q1: input propagator 1 (string)
+ * - q2: input propagator 2 (string)
+ * - q3: input propagator 3 (string), assumed to be sequential propagator 
+ * - q4: input propagator 4 (string), assumed to be a loop
+ * 
+ * type 1: trace(q1*adj(q2)*g5*gL[mu])*trace(loop*gL[mu])*trace(q3*g5)
+ * type 2: trace(q1*adj(q2)*g5*gL[mu]*loop*gL[mu])*trace(q3*g5)
+ */
+
+/*******************************************************************************
+ *                  TWeakNeutral4ptDisc implementation                         *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+TWeakNeutral4ptDisc::TWeakNeutral4ptDisc(const std::string name)
+: Module<WeakHamiltonianPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+std::vector<std::string> TWeakNeutral4ptDisc::getInput(void)
+{
+    std::vector<std::string> in = {par().q1, par().q2, par().q3, par().q4};
+    
+    return in;
+}
+
+std::vector<std::string> TWeakNeutral4ptDisc::getOutput(void)
+{
+    std::vector<std::string> out = {};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+void TWeakNeutral4ptDisc::setup(void)
+{
+    unsigned int ndim = env().getNd();
+
+    envTmpLat(LatticeComplex,  "expbuf");
+    envTmpLat(PropagatorField, "tmp");
+    envTmpLat(LatticeComplex,  "curr");
+    envTmp(std::vector<PropagatorField>, "meson", 1, ndim, PropagatorField(env().getGrid()));
+    envTmp(std::vector<PropagatorField>, "loop", 1, ndim,  PropagatorField(env().getGrid()));
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+void TWeakNeutral4ptDisc::execute(void)
+{
+    LOG(Message) << "Computing Weak Hamiltonian neutral disconnected contractions '" 
+                 << getName() << "' using quarks '" << par().q1 << "', '" 
+                 << par().q2 << ", '" << par().q3 << "' and '" << par().q4 
+                 << "'." << std::endl;
+
+    ResultWriter          writer(RESULT_FILE_NAME(par().output));
+    auto                  &q1 = envGet(PropagatorField, par().q1);
+    auto                  &q2 = envGet(PropagatorField, par().q2);
+    auto                  &q3 = envGet(PropagatorField, par().q3);
+    auto                  &q4 = envGet(PropagatorField, par().q4);
+    Gamma                 g5  = Gamma(Gamma::Algebra::Gamma5);
+    std::vector<TComplex> corrbuf;
+    std::vector<Result>   result(n_neut_disc_diag);
+    unsigned int          ndim = env().getNd();
+
+    envGetTmp(LatticeComplex,               expbuf); 
+    envGetTmp(PropagatorField,              tmp);
+    envGetTmp(LatticeComplex,               curr);
+    envGetTmp(std::vector<PropagatorField>, meson);
+    envGetTmp(std::vector<PropagatorField>, loop);
+
+    // Setup for type 1 contractions.
+    for (int mu = 0; mu < ndim; ++mu)
+    {
+        meson[mu] = MAKE_DISC_MESON(q1, q2, GammaL(Gamma::gmu[mu]));
+        loop[mu] = MAKE_DISC_LOOP(q4, GammaL(Gamma::gmu[mu]));
+    }
+    curr = MAKE_DISC_CURR(q3, GammaL(Gamma::Algebra::Gamma5));
+
+    // Perform type 1 contractions.    
+    SUM_MU(expbuf, trace(meson[mu]*loop[mu]))
+    expbuf *= curr;
+    MAKE_DIAG(expbuf, corrbuf, result[neut_disc_1_diag], "HW_disc0_1")
+
+    // Perform type 2 contractions.
+    SUM_MU(expbuf, trace(meson[mu])*trace(loop[mu]))
+    expbuf *= curr;
+    MAKE_DIAG(expbuf, corrbuf, result[neut_disc_2_diag], "HW_disc0_2")
+
+    write(writer, "HW_disc0", result);
+}

extras/Hadrons/Modules/MContraction/WeakNeutral4ptDisc.hpp

@@ -0,0 +1,60 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MContraction/WeakNeutral4ptDisc.hpp
+
+Copyright (C) 2015-2018
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+Author: Lanny91 <andrew.lawson@gmail.com>
+
+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 */
+
+#ifndef Hadrons_MContraction_WeakNeutral4ptDisc_hpp_
+#define Hadrons_MContraction_WeakNeutral4ptDisc_hpp_
+
+#include <Grid/Hadrons/Modules/MContraction/WeakHamiltonian.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                         WeakNeutral4ptDisc                                 *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MContraction)
+
+enum
+{
+    neut_disc_1_diag = 0,
+    neut_disc_2_diag = 1,
+    n_neut_disc_diag = 2
+};
+
+// Neutral 4pt disconnected subdiagram contractions.
+#define MAKE_DISC_MESON(Q_1, Q_2, gamma) (Q_1*adj(Q_2)*g5*gamma)
+#define MAKE_DISC_LOOP(Q_LOOP, gamma) (Q_LOOP*gamma)
+#define MAKE_DISC_CURR(Q_c, gamma) (trace(Q_c*gamma))
+
+MAKE_WEAK_MODULE(WeakNeutral4ptDisc)
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MContraction_WeakNeutral4ptDisc_hpp_

extras/Hadrons/Modules/MFermion/GaugeProp.hpp

@@ -2,12 +2,12 @@
 
 Grid physics library, www.github.com/paboyle/Grid 
 
-Source file: extras/Hadrons/Modules/Quark.hpp
+Source file: extras/Hadrons/Modules/MFermion/GaugeProp.hpp
 
-Copyright (C) 2015
-Copyright (C) 2016
+Copyright (C) 2015-2018
 
 Author: Antonin Portelli <antonin.portelli@me.com>
+Author: Lanny91 <andrew.lawson@gmail.com>
 
 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
@@ -27,8 +27,8 @@ See the full license in the file "LICENSE" in the top level distribution directo
 *************************************************************************************/
 /*  END LEGAL */
 
-#ifndef Hadrons_Quark_hpp_
-#define Hadrons_Quark_hpp_
+#ifndef Hadrons_MFermion_GaugeProp_hpp_
+#define Hadrons_MFermion_GaugeProp_hpp_
 
 #include <Grid/Hadrons/Global.hpp>
 #include <Grid/Hadrons/Module.hpp>
@@ -37,29 +37,53 @@ See the full license in the file "LICENSE" in the top level distribution directo
 BEGIN_HADRONS_NAMESPACE
 
 /******************************************************************************
- *                               TQuark                                       *
+ * 5D -> 4D and 4D -> 5D conversions.                                         *
  ******************************************************************************/
-class QuarkPar: Serializable
+template<class vobj> // Note that 5D object is modified.
+inline void make_4D(Lattice<vobj> &in_5d, Lattice<vobj> &out_4d, int Ls)
+{
+    axpby_ssp_pminus(in_5d, 0., in_5d, 1., in_5d, 0, 0);
+    axpby_ssp_pplus(in_5d, 1., in_5d, 1., in_5d, 0, Ls-1);
+    ExtractSlice(out_4d, in_5d, 0, 0);
+}
+
+template<class vobj>
+inline void make_5D(Lattice<vobj> &in_4d, Lattice<vobj> &out_5d, int Ls)
+{
+    out_5d = zero;
+    InsertSlice(in_4d, out_5d, 0, 0);
+    InsertSlice(in_4d, out_5d, Ls-1, 0);
+    axpby_ssp_pplus(out_5d, 0., out_5d, 1., out_5d, 0, 0);
+    axpby_ssp_pminus(out_5d, 0., out_5d, 1., out_5d, Ls-1, Ls-1);
+}
+
+/******************************************************************************
+ *                                GaugeProp                                   *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MFermion)
+
+class GaugePropPar: Serializable
 {
 public:
-    GRID_SERIALIZABLE_CLASS_MEMBERS(QuarkPar,
+    GRID_SERIALIZABLE_CLASS_MEMBERS(GaugePropPar,
                                     std::string, source,
                                     std::string, solver);
 };
 
 template <typename FImpl>
-class TQuark: public Module<QuarkPar>
+class TGaugeProp: public Module<GaugePropPar>
 {
 public:
-    TYPE_ALIASES(FImpl,);
+    FGS_TYPE_ALIASES(FImpl,);
 public:
     // constructor
-    TQuark(const std::string name);
+    TGaugeProp(const std::string name);
     // destructor
-    virtual ~TQuark(void) = default;
-    // dependencies/products
+    virtual ~TGaugeProp(void) = default;
+    // dependency relation
     virtual std::vector<std::string> getInput(void);
     virtual std::vector<std::string> getOutput(void);
+protected:
     // setup
     virtual void setup(void);
     // execution
@@ -69,20 +93,19 @@ private:
     SolverFn     *solver_{nullptr};
 };
 
-MODULE_REGISTER(Quark, TQuark<FIMPL>);
-
+MODULE_REGISTER_NS(GaugeProp, TGaugeProp<FIMPL>, MFermion);
 /******************************************************************************
- *                          TQuark implementation                             *
+ *                      TGaugeProp implementation                             *
  ******************************************************************************/
 // constructor /////////////////////////////////////////////////////////////////
 template <typename FImpl>
-TQuark<FImpl>::TQuark(const std::string name)
-: Module(name)
+TGaugeProp<FImpl>::TGaugeProp(const std::string name)
+: Module<GaugePropPar>(name)
 {}
 
 // dependencies/products ///////////////////////////////////////////////////////
 template <typename FImpl>
-std::vector<std::string> TQuark<FImpl>::getInput(void)
+std::vector<std::string> TGaugeProp<FImpl>::getInput(void)
 {
     std::vector<std::string> in = {par().source, par().solver};
     
@@ -90,7 +113,7 @@ std::vector<std::string> TQuark<FImpl>::getInput(void)
 }
 
 template <typename FImpl>
-std::vector<std::string> TQuark<FImpl>::getOutput(void)
+std::vector<std::string> TGaugeProp<FImpl>::getOutput(void)
 {
     std::vector<std::string> out = {getName(), getName() + "_5d"};
     
@@ -99,56 +122,52 @@ std::vector<std::string> TQuark<FImpl>::getOutput(void)
 
 // setup ///////////////////////////////////////////////////////////////////////
 template <typename FImpl>
-void TQuark<FImpl>::setup(void)
+void TGaugeProp<FImpl>::setup(void)
 {
     Ls_ = env().getObjectLs(par().solver);
-    env().template registerLattice<PropagatorField>(getName());
+    envCreateLat(PropagatorField, getName());
+    envTmpLat(FermionField, "source", Ls_);
+    envTmpLat(FermionField, "sol", Ls_);
+    envTmpLat(FermionField, "tmp");
     if (Ls_ > 1)
     {
-        env().template registerLattice<PropagatorField>(getName() + "_5d", Ls_);
+        envCreateLat(PropagatorField, getName() + "_5d", Ls_);
     }
 }
 
 // execution ///////////////////////////////////////////////////////////////////
 template <typename FImpl>
-void TQuark<FImpl>::execute(void)
+void TGaugeProp<FImpl>::execute(void)
 {
     LOG(Message) << "Computing quark propagator '" << getName() << "'"
                  << std::endl;
     
-    FermionField    source(env().getGrid(Ls_)), sol(env().getGrid(Ls_)),
-                    tmp(env().getGrid());
-    std::string     propName = (Ls_ == 1) ? getName() : (getName() + "_5d");
-    PropagatorField &prop    = *env().template createLattice<PropagatorField>(propName);
-    PropagatorField &fullSrc = *env().template getObject<PropagatorField>(par().source);
-    SolverFn        &solver  = *env().template getObject<SolverFn>(par().solver);
-    if (Ls_ > 1)
-    {
-        env().template createLattice<PropagatorField>(getName());
-    }
+    std::string propName = (Ls_ == 1) ? getName() : (getName() + "_5d");
+    auto        &prop    = envGet(PropagatorField, propName);
+    auto        &fullSrc = envGet(PropagatorField, par().source);
+    auto        &solver  = envGet(SolverFn, par().solver);
     
+    envGetTmp(FermionField, source);
+    envGetTmp(FermionField, sol);
+    envGetTmp(FermionField, tmp);
     LOG(Message) << "Inverting using solver '" << par().solver
                  << "' on source '" << par().source << "'" << std::endl;
     for (unsigned int s = 0; s < Ns; ++s)
-    for (unsigned int c = 0; c < Nc; ++c)
+      for (unsigned int c = 0; c < FImpl::Dimension; ++c)
     {
         LOG(Message) << "Inversion for spin= " << s << ", color= " << c
-        << std::endl;
+                     << std::endl;
         // source conversion for 4D sources
         if (!env().isObject5d(par().source))
         {
             if (Ls_ == 1)
             {
-                PropToFerm(source, fullSrc, s, c);
+               PropToFerm<FImpl>(source, fullSrc, s, c);
             }
             else
             {
-                source = zero;
-                PropToFerm(tmp, fullSrc, s, c);
-                InsertSlice(tmp, source, 0, 0);
-                InsertSlice(tmp, source, Ls_-1, 0);
-                axpby_ssp_pplus(source, 0., source, 1., source, 0, 0);
-                axpby_ssp_pminus(source, 0., source, 1., source, Ls_-1, Ls_-1);
+                PropToFerm<FImpl>(tmp, fullSrc, s, c);
+                make_5D(tmp, source, Ls_);
             }
         }
         // source conversion for 5D sources
@@ -156,30 +175,28 @@ void TQuark<FImpl>::execute(void)
         {
             if (Ls_ != env().getObjectLs(par().source))
             {
-                HADRON_ERROR("Ls mismatch between quark action and source");
+                HADRON_ERROR(Size, "Ls mismatch between quark action and source");
             }
             else
             {
-                PropToFerm(source, fullSrc, s, c);
+                PropToFerm<FImpl>(source, fullSrc, s, c);
             }
         }
         sol = zero;
         solver(sol, source);
-        FermToProp(prop, sol, s, c);
+        FermToProp<FImpl>(prop, sol, s, c);
         // create 4D propagators from 5D one if necessary
         if (Ls_ > 1)
         {
-            PropagatorField &p4d =
-                *env().template getObject<PropagatorField>(getName());
-            
-            axpby_ssp_pminus(sol, 0., sol, 1., sol, 0, 0);
-            axpby_ssp_pplus(sol, 0., sol, 1., sol, 0, Ls_-1);
-            ExtractSlice(tmp, sol, 0, 0);
-            FermToProp(p4d, tmp, s, c);
+            PropagatorField &p4d = envGet(PropagatorField, getName());
+            make_4D(sol, tmp, Ls_);
+            FermToProp<FImpl>(p4d, tmp, s, c);
         }
     }
 }
 
+END_MODULE_NAMESPACE
+
 END_HADRONS_NAMESPACE
 
-#endif // Hadrons_Quark_hpp_
+#endif // Hadrons_MFermion_GaugeProp_hpp_

extras/Hadrons/Modules/MGauge/FundtoHirep.cc

@@ -0,0 +1,75 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MGauge/FundtoHirep.cc
+
+Copyright (C) 2015
+Copyright (C) 2016
+
+
+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/Hadrons/Modules/MGauge/FundtoHirep.hpp>
+
+using namespace Grid;
+using namespace Hadrons;
+using namespace MGauge;
+
+// constructor /////////////////////////////////////////////////////////////////
+template <class Rep>
+TFundtoHirep<Rep>::TFundtoHirep(const std::string name)
+: Module<FundtoHirepPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+template <class Rep>
+std::vector<std::string> TFundtoHirep<Rep>::getInput(void)
+{
+    std::vector<std::string> in;
+    return in;
+}
+
+template <class Rep>
+std::vector<std::string> TFundtoHirep<Rep>::getOutput(void)
+{
+    std::vector<std::string> out = {getName()};
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+template <typename Rep>
+void TFundtoHirep<Rep>::setup(void)
+{
+    envCreateLat(typename Rep::LatticeField, getName());
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+template <class Rep>
+void TFundtoHirep<Rep>::execute(void)
+{
+    auto &U      = *env().template getObject<LatticeGaugeField>(par().gaugeconf);
+    LOG(Message) << "Transforming Representation" << std::endl;
+
+    Rep TargetRepresentation(U._grid);
+    TargetRepresentation.update_representation(U);
+
+    auto &URep = envGet(typename Rep::LatticeField, getName());
+    URep = TargetRepresentation.U;
+}

extras/Hadrons/Modules/MGauge/FundtoHirep.hpp

@@ -0,0 +1,77 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MGauge/FundtoHirep.hpp
+
+Copyright (C) 2015
+Copyright (C) 2016
+
+Author: David Preti <david.preti@to.infn.it>
+	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 */
+
+#ifndef Hadrons_MGauge_FundtoHirep_hpp_
+#define Hadrons_MGauge_FundtoHirep_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                         Load a NERSC configuration                         *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MGauge)
+
+class FundtoHirepPar: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(FundtoHirepPar,
+                                    std::string, gaugeconf);
+};
+
+template <class Rep>
+class TFundtoHirep: public Module<FundtoHirepPar>
+{
+public:
+    // constructor
+    TFundtoHirep(const std::string name);
+    // destructor
+    virtual ~TFundtoHirep(void) = default;
+    // dependency relation
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+    // setup
+    void setup(void);
+    // execution
+    void execute(void);
+};
+
+//MODULE_REGISTER_NS(FundtoAdjoint,   TFundtoHirep<AdjointRepresentation>, MGauge);
+//MODULE_REGISTER_NS(FundtoTwoIndexSym, TFundtoHirep<TwoIndexSymmetricRepresentation>, MGauge);
+//MODULE_REGISTER_NS(FundtoTwoIndexAsym, TFundtoHirep<TwoIndexAntiSymmetricRepresentation>, MGauge);
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MGauge_FundtoHirep_hpp_

extras/Hadrons/Modules/MGauge/Random.cc

@@ -4,8 +4,7 @@ Grid physics library, www.github.com/paboyle/Grid
 
 Source file: extras/Hadrons/Modules/MGauge/Random.cc
 
-Copyright (C) 2015
-Copyright (C) 2016
+Copyright (C) 2015-2018
 
 Author: Antonin Portelli <antonin.portelli@me.com>
 
@@ -44,7 +43,9 @@ TRandom::TRandom(const std::string name)
 // dependencies/products ///////////////////////////////////////////////////////
 std::vector<std::string> TRandom::getInput(void)
 {
-    return std::vector<std::string>();
+    std::vector<std::string> in;
+    
+    return in;
 }
 
 std::vector<std::string> TRandom::getOutput(void)
@@ -57,13 +58,14 @@ std::vector<std::string> TRandom::getOutput(void)
 // setup ///////////////////////////////////////////////////////////////////////
 void TRandom::setup(void)
 {
-    env().registerLattice<LatticeGaugeField>(getName());
+    envCreateLat(LatticeGaugeField, getName());
 }
 
 // execution ///////////////////////////////////////////////////////////////////
 void TRandom::execute(void)
 {
     LOG(Message) << "Generating random gauge configuration" << std::endl;
-    LatticeGaugeField &U = *env().createLattice<LatticeGaugeField>(getName());
+    
+    auto &U = envGet(LatticeGaugeField, getName());
     SU3::HotConfiguration(*env().get4dRng(), U);
 }

extras/Hadrons/Modules/MGauge/Random.hpp

@@ -4,8 +4,7 @@ Grid physics library, www.github.com/paboyle/Grid
 
 Source file: extras/Hadrons/Modules/MGauge/Random.hpp
 
-Copyright (C) 2015
-Copyright (C) 2016
+Copyright (C) 2015-2018
 
 Author: Antonin Portelli <antonin.portelli@me.com>
 
@@ -27,8 +26,8 @@ See the full license in the file "LICENSE" in the top level distribution directo
 *************************************************************************************/
 /*  END LEGAL */
 
-#ifndef Hadrons_Random_hpp_
-#define Hadrons_Random_hpp_
+#ifndef Hadrons_MGauge_Random_hpp_
+#define Hadrons_MGauge_Random_hpp_
 
 #include <Grid/Hadrons/Global.hpp>
 #include <Grid/Hadrons/Module.hpp>
@@ -51,6 +50,7 @@ public:
     // dependency relation
     virtual std::vector<std::string> getInput(void);
     virtual std::vector<std::string> getOutput(void);
+protected:
     // setup
     virtual void setup(void);
     // execution
@@ -63,4 +63,4 @@ END_MODULE_NAMESPACE
 
 END_HADRONS_NAMESPACE
 
-#endif // Hadrons_Random_hpp_
+#endif // Hadrons_MGauge_Random_hpp_

extras/Hadrons/Modules/MGauge/StochEm.cc

@@ -0,0 +1,84 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MGauge/StochEm.cc
+
+Copyright (C) 2015-2018
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+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/Hadrons/Modules/MGauge/StochEm.hpp>
+
+using namespace Grid;
+using namespace Hadrons;
+using namespace MGauge;
+
+/******************************************************************************
+*                  TStochEm implementation                             *
+******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+TStochEm::TStochEm(const std::string name)
+: Module<StochEmPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+std::vector<std::string> TStochEm::getInput(void)
+{
+    std::vector<std::string> in;
+    
+    return in;
+}
+
+std::vector<std::string> TStochEm::getOutput(void)
+{
+    std::vector<std::string> out = {getName()};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+void TStochEm::setup(void)
+{
+    if (!env().hasCreatedObject("_" + getName() + "_weight"))
+    {
+        envCacheLat(EmComp, "_" + getName() + "_weight");
+    }
+    envCreateLat(EmField, getName());
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+void TStochEm::execute(void)
+{
+    LOG(Message) << "Generating stochatic EM potential..." << std::endl;
+
+    PhotonR photon(par().gauge, par().zmScheme);
+    auto    &a = envGet(EmField, getName());
+    auto    &w = envGet(EmComp, "_" + getName() + "_weight");
+    
+    if (!env().hasCreatedObject("_" + getName() + "_weight"))
+    {
+        LOG(Message) << "Caching stochatic EM potential weight (gauge: "
+                     << par().gauge << ", zero-mode scheme: "
+                     << par().zmScheme << ")..." << std::endl;
+        photon.StochasticWeight(w);
+    }
+    photon.StochasticField(a, *env().get4dRng(), w);
+}

extras/Hadrons/Modules/MGauge/StochEm.hpp

@@ -0,0 +1,76 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MGauge/StochEm.hpp
+
+Copyright (C) 2015-2018
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+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 */
+#ifndef Hadrons_MGauge_StochEm_hpp_
+#define Hadrons_MGauge_StochEm_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                         StochEm                                 *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MGauge)
+
+class StochEmPar: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(StochEmPar,
+                                    PhotonR::Gauge,    gauge,
+                                    PhotonR::ZmScheme, zmScheme);
+};
+
+class TStochEm: public Module<StochEmPar>
+{
+public:
+    typedef PhotonR::GaugeField     EmField;
+    typedef PhotonR::GaugeLinkField EmComp;
+public:
+    // constructor
+    TStochEm(const std::string name);
+    // destructor
+    virtual ~TStochEm(void) = default;
+    // dependency relation
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+protected:
+    // setup
+    virtual void setup(void);
+    // execution
+    virtual void execute(void);
+};
+
+MODULE_REGISTER_NS(StochEm, TStochEm, MGauge);
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MGauge_StochEm_hpp_

extras/Hadrons/Modules/MGauge/Unit.cc

@@ -4,8 +4,7 @@ Grid physics library, www.github.com/paboyle/Grid
 
 Source file: extras/Hadrons/Modules/MGauge/Unit.cc
 
-Copyright (C) 2015
-Copyright (C) 2016
+Copyright (C) 2015-2018
 
 Author: Antonin Portelli <antonin.portelli@me.com>
 
@@ -57,13 +56,14 @@ std::vector<std::string> TUnit::getOutput(void)
 // setup ///////////////////////////////////////////////////////////////////////
 void TUnit::setup(void)
 {
-    env().registerLattice<LatticeGaugeField>(getName());
+    envCreateLat(LatticeGaugeField, getName());
 }
 
 // execution ///////////////////////////////////////////////////////////////////
 void TUnit::execute(void)
 {
     LOG(Message) << "Creating unit gauge configuration" << std::endl;
-    LatticeGaugeField &U = *env().createLattice<LatticeGaugeField>(getName());
+    
+    auto &U = envGet(LatticeGaugeField, getName());
     SU3::ColdConfiguration(*env().get4dRng(), U);
 }

extras/Hadrons/Modules/MGauge/Unit.hpp

@@ -4,8 +4,7 @@ Grid physics library, www.github.com/paboyle/Grid
 
 Source file: extras/Hadrons/Modules/MGauge/Unit.hpp
 
-Copyright (C) 2015
-Copyright (C) 2016
+Copyright (C) 2015-2018
 
 Author: Antonin Portelli <antonin.portelli@me.com>
 
@@ -27,8 +26,8 @@ See the full license in the file "LICENSE" in the top level distribution directo
 *************************************************************************************/
 /*  END LEGAL */
 
-#ifndef Hadrons_Unit_hpp_
-#define Hadrons_Unit_hpp_
+#ifndef Hadrons_MGauge_Unit_hpp_
+#define Hadrons_MGauge_Unit_hpp_
 
 #include <Grid/Hadrons/Global.hpp>
 #include <Grid/Hadrons/Module.hpp>
@@ -51,6 +50,7 @@ public:
     // dependencies/products
     virtual std::vector<std::string> getInput(void);
     virtual std::vector<std::string> getOutput(void);
+protected:
     // setup
     virtual void setup(void);
     // execution
@@ -63,4 +63,4 @@ END_MODULE_NAMESPACE
 
 END_HADRONS_NAMESPACE
 
-#endif // Hadrons_Unit_hpp_
+#endif // Hadrons_MGauge_Unit_hpp_

extras/Hadrons/Modules/MIO/LoadBinary.hpp

@@ -0,0 +1,140 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MIO/LoadBinary.hpp
+
+Copyright (C) 2015-2018
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+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 */
+#ifndef Hadrons_MIO_LoadBinary_hpp_
+#define Hadrons_MIO_LoadBinary_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                       Load a binary configurations                         *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MIO)
+
+class LoadBinaryPar: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(LoadBinaryPar,
+                                    std::string, file,
+                                    std::string, format);
+};
+
+template <typename Impl>
+class TLoadBinary: public Module<LoadBinaryPar>
+{
+public:
+    typedef typename Impl::Field                  Field;
+    typedef typename Impl::Simd                   Simd;
+    typedef typename Field::vector_object         vobj;
+    typedef typename vobj::scalar_object          sobj;
+    typedef typename sobj::DoublePrecision        sobj_double;
+    typedef BinarySimpleMunger<sobj_double, sobj> Munger;
+public:
+    // constructor
+    TLoadBinary(const std::string name);
+    // destructor
+    virtual ~TLoadBinary(void) = default;
+    // dependency relation
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+    // setup
+    virtual void setup(void);
+    // execution
+    virtual void execute(void);
+};
+
+MODULE_REGISTER_NS(LoadBinary, TLoadBinary<GIMPL>, MIO);
+MODULE_REGISTER_NS(LoadBinaryScalarSU2, TLoadBinary<ScalarNxNAdjImplR<2>>, MIO);
+MODULE_REGISTER_NS(LoadBinaryScalarSU3, TLoadBinary<ScalarNxNAdjImplR<3>>, MIO);
+MODULE_REGISTER_NS(LoadBinaryScalarSU4, TLoadBinary<ScalarNxNAdjImplR<4>>, MIO);
+MODULE_REGISTER_NS(LoadBinaryScalarSU5, TLoadBinary<ScalarNxNAdjImplR<5>>, MIO);
+MODULE_REGISTER_NS(LoadBinaryScalarSU6, TLoadBinary<ScalarNxNAdjImplR<6>>, MIO);
+
+/******************************************************************************
+ *                         TLoadBinary implementation                         *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+template <typename Impl>
+TLoadBinary<Impl>::TLoadBinary(const std::string name)
+: Module<LoadBinaryPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+template <typename Impl>
+std::vector<std::string> TLoadBinary<Impl>::getInput(void)
+{
+    std::vector<std::string> in;
+    
+    return in;
+}
+
+template <typename Impl>
+std::vector<std::string> TLoadBinary<Impl>::getOutput(void)
+{
+    std::vector<std::string> out = {getName()};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+template <typename Impl>
+void TLoadBinary<Impl>::setup(void)
+{
+    envCreateLat(Field, getName());
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+template <typename Impl>
+void TLoadBinary<Impl>::execute(void)
+{
+    Munger      munge;
+    uint32_t    nersc_csum, scidac_csuma, scidac_csumb;
+    auto        &U = envGet(Field, getName());
+    std::string filename = par().file + "."
+                           + std::to_string(vm().getTrajectory());
+
+    LOG(Message) << "Loading " << par().format 
+                 << " binary configuration from file '" << filename
+                 << "'" << std::endl;
+    BinaryIO::readLatticeObject<vobj, sobj_double>(U, filename, munge, 0, 
+                                                   par().format, nersc_csum,
+                                                   scidac_csuma, scidac_csumb);
+    LOG(Message) << "Checksums:" << std::endl;
+    LOG(Message) << "  NERSC    " << nersc_csum << std::endl;
+    LOG(Message) << "  SciDAC A " << scidac_csuma << std::endl;
+    LOG(Message) << "  SciDAC B " << scidac_csumb << std::endl;
+}
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MIO_LoadBinary_hpp_

extras/Hadrons/Modules/MIO/LoadNersc.cc

@@ -2,10 +2,9 @@
 
 Grid physics library, www.github.com/paboyle/Grid 
 
-Source file: extras/Hadrons/Modules/MGauge/Load.cc
+Source file: extras/Hadrons/Modules/MIO/LoadNersc.cc
 
-Copyright (C) 2015
-Copyright (C) 2016
+Copyright (C) 2015-2018
 
 Author: Antonin Portelli <antonin.portelli@me.com>
 
@@ -26,30 +25,29 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution directory
 *************************************************************************************/
 /*  END LEGAL */
-
-#include <Grid/Hadrons/Modules/MGauge/Load.hpp>
+#include <Grid/Hadrons/Modules/MIO/LoadNersc.hpp>
 
 using namespace Grid;
 using namespace Hadrons;
-using namespace MGauge;
+using namespace MIO;
 
 /******************************************************************************
-*                           TLoad implementation                               *
+*                       TLoadNersc implementation                             *
 ******************************************************************************/
 // constructor /////////////////////////////////////////////////////////////////
-TLoad::TLoad(const std::string name)
-: Module<LoadPar>(name)
+TLoadNersc::TLoadNersc(const std::string name)
+: Module<LoadNerscPar>(name)
 {}
 
 // dependencies/products ///////////////////////////////////////////////////////
-std::vector<std::string> TLoad::getInput(void)
+std::vector<std::string> TLoadNersc::getInput(void)
 {
     std::vector<std::string> in;
     
     return in;
 }
 
-std::vector<std::string> TLoad::getOutput(void)
+std::vector<std::string> TLoadNersc::getOutput(void)
 {
     std::vector<std::string> out = {getName()};
     
@@ -57,22 +55,22 @@ std::vector<std::string> TLoad::getOutput(void)
 }
 
 // setup ///////////////////////////////////////////////////////////////////////
-void TLoad::setup(void)
+void TLoadNersc::setup(void)
 {
-    env().registerLattice<LatticeGaugeField>(getName());
+    envCreateLat(LatticeGaugeField, getName());
 }
 
 // execution ///////////////////////////////////////////////////////////////////
-void TLoad::execute(void)
+void TLoadNersc::execute(void)
 {
-    NerscField  header;
-    std::string fileName = par().file + "."
-                           + std::to_string(env().getTrajectory());
-    
+    FieldMetaData header;
+    std::string   fileName = par().file + "."
+                             + std::to_string(vm().getTrajectory());
     LOG(Message) << "Loading NERSC configuration from file '" << fileName
                  << "'" << std::endl;
-    LatticeGaugeField &U = *env().createLattice<LatticeGaugeField>(getName());
+
+    auto &U = envGet(LatticeGaugeField, getName());
     NerscIO::readConfiguration(U, header, fileName);
     LOG(Message) << "NERSC header:" << std::endl;
-    dump_nersc_header(header, LOG(Message));
+    dump_meta_data(header, LOG(Message));
 }

extras/Hadrons/Modules/MIO/LoadNersc.hpp

@@ -2,10 +2,9 @@
 
 Grid physics library, www.github.com/paboyle/Grid 
 
-Source file: extras/Hadrons/Modules/MGauge/Load.hpp
+Source file: extras/Hadrons/Modules/MIO/LoadNersc.hpp
 
-Copyright (C) 2015
-Copyright (C) 2016
+Copyright (C) 2015-2018
 
 Author: Antonin Portelli <antonin.portelli@me.com>
 
@@ -26,9 +25,8 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution directory
 *************************************************************************************/
 /*  END LEGAL */
-
-#ifndef Hadrons_Load_hpp_
-#define Hadrons_Load_hpp_
+#ifndef Hadrons_MIO_LoadNersc_hpp_
+#define Hadrons_MIO_LoadNersc_hpp_
 
 #include <Grid/Hadrons/Global.hpp>
 #include <Grid/Hadrons/Module.hpp>
@@ -37,24 +35,24 @@ See the full license in the file "LICENSE" in the top level distribution directo
 BEGIN_HADRONS_NAMESPACE
 
 /******************************************************************************
- *                         Load a NERSC configuration                         *
+ *                       Load a NERSC configuration                           *
  ******************************************************************************/
-BEGIN_MODULE_NAMESPACE(MGauge)
+BEGIN_MODULE_NAMESPACE(MIO)
 
-class LoadPar: Serializable
+class LoadNerscPar: Serializable
 {
 public:
-    GRID_SERIALIZABLE_CLASS_MEMBERS(LoadPar,
+    GRID_SERIALIZABLE_CLASS_MEMBERS(LoadNerscPar,
                                     std::string, file);
 };
 
-class TLoad: public Module<LoadPar>
+class TLoadNersc: public Module<LoadNerscPar>
 {
 public:
     // constructor
-    TLoad(const std::string name);
+    TLoadNersc(const std::string name);
     // destructor
-    virtual ~TLoad(void) = default;
+    virtual ~TLoadNersc(void) = default;
     // dependency relation
     virtual std::vector<std::string> getInput(void);
     virtual std::vector<std::string> getOutput(void);
@@ -64,10 +62,10 @@ public:
     virtual void execute(void);
 };
 
-MODULE_REGISTER_NS(Load, TLoad, MGauge);
+MODULE_REGISTER_NS(LoadNersc, TLoadNersc, MIO);
 
 END_MODULE_NAMESPACE
 
 END_HADRONS_NAMESPACE
 
-#endif // Hadrons_Load_hpp_
+#endif // Hadrons_MIO_LoadNersc_hpp_

extras/Hadrons/Modules/MLoop/NoiseLoop.hpp

@@ -0,0 +1,134 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MLoop/NoiseLoop.hpp
+
+Copyright (C) 2015-2018
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+Author: Lanny91 <andrew.lawson@gmail.com>
+
+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 */
+
+#ifndef Hadrons_MLoop_NoiseLoop_hpp_
+#define Hadrons_MLoop_NoiseLoop_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/*
+ 
+ Noise loop propagator
+ -----------------------------
+ * loop_x = q_x * adj(eta_x)
+ 
+ * options:
+ - q = Result of inversion on noise source.
+ - eta = noise source.
+
+ */
+
+
+/******************************************************************************
+ *                         NoiseLoop                                          *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MLoop)
+
+class NoiseLoopPar: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(NoiseLoopPar,
+                                    std::string, q,
+                                    std::string, eta);
+};
+
+template <typename FImpl>
+class TNoiseLoop: public Module<NoiseLoopPar>
+{
+public:
+    FERM_TYPE_ALIASES(FImpl,);
+public:
+    // constructor
+    TNoiseLoop(const std::string name);
+    // destructor
+    virtual ~TNoiseLoop(void) = default;
+    // dependency relation
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+protected:
+    // setup
+    virtual void setup(void);
+    // execution
+    virtual void execute(void);
+};
+
+MODULE_REGISTER_NS(NoiseLoop, TNoiseLoop<FIMPL>, MLoop);
+
+/******************************************************************************
+ *                 TNoiseLoop implementation                                  *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+template <typename FImpl>
+TNoiseLoop<FImpl>::TNoiseLoop(const std::string name)
+: Module<NoiseLoopPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+template <typename FImpl>
+std::vector<std::string> TNoiseLoop<FImpl>::getInput(void)
+{
+    std::vector<std::string> in = {par().q, par().eta};
+    
+    return in;
+}
+
+template <typename FImpl>
+std::vector<std::string> TNoiseLoop<FImpl>::getOutput(void)
+{
+    std::vector<std::string> out = {getName()};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TNoiseLoop<FImpl>::setup(void)
+{
+    envCreateLat(PropagatorField, getName());
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TNoiseLoop<FImpl>::execute(void)
+{
+    auto &loop = envGet(PropagatorField, getName());
+    auto &q    = envGet(PropagatorField, par().q);
+    auto &eta  = envGet(PropagatorField, par().eta);
+    loop = q*adj(eta);
+}
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MLoop_NoiseLoop_hpp_

extras/Hadrons/Modules/MScalar/ChargedProp.cc

@@ -0,0 +1,249 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MScalar/ChargedProp.cc
+
+Copyright (C) 2015-2018
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+Author: James Harrison <jch1g10@soton.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/Hadrons/Modules/MScalar/ChargedProp.hpp>
+#include <Grid/Hadrons/Modules/MScalar/Scalar.hpp>
+
+using namespace Grid;
+using namespace Hadrons;
+using namespace MScalar;
+
+/******************************************************************************
+*                     TChargedProp implementation                             *
+******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+TChargedProp::TChargedProp(const std::string name)
+: Module<ChargedPropPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+std::vector<std::string> TChargedProp::getInput(void)
+{
+    std::vector<std::string> in = {par().source, par().emField};
+    
+    return in;
+}
+
+std::vector<std::string> TChargedProp::getOutput(void)
+{
+    std::vector<std::string> out = {getName()};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+void TChargedProp::setup(void)
+{
+    freeMomPropName_ = FREEMOMPROP(par().mass);
+    phaseName_.clear();
+    for (unsigned int mu = 0; mu < env().getNd(); ++mu)
+    {
+        phaseName_.push_back("_shiftphase_" + std::to_string(mu));
+    }
+    GFSrcName_ = getName() + "_DinvSrc";
+    fftName_   = getName() + "_fft";
+
+    freeMomPropDone_ = env().hasCreatedObject(freeMomPropName_);
+    GFSrcDone_       = env().hasCreatedObject(GFSrcName_);
+    phasesDone_      = env().hasCreatedObject(phaseName_[0]);
+    envCacheLat(ScalarField, freeMomPropName_);
+    for (unsigned int mu = 0; mu < env().getNd(); ++mu)
+    {
+        envCacheLat(ScalarField, phaseName_[mu]);
+    }
+    envCacheLat(ScalarField, GFSrcName_);
+    envCreateLat(ScalarField, getName());
+    envTmpLat(ScalarField, "buf");
+    envTmpLat(ScalarField, "result");
+    envTmpLat(ScalarField, "Amu");
+    envCache(FFT, fftName_, 1, env().getGrid());
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+void TChargedProp::execute(void)
+{
+    // CACHING ANALYTIC EXPRESSIONS
+    makeCaches();
+
+    // PROPAGATOR CALCULATION
+    LOG(Message) << "Computing charged scalar propagator"
+                 << " (mass= " << par().mass
+                 << ", charge= " << par().charge << ")..." << std::endl;
+    
+    auto   &prop  = envGet(ScalarField, getName());
+    auto   &GFSrc = envGet(ScalarField, GFSrcName_);
+    auto   &G     = envGet(ScalarField, freeMomPropName_);
+    auto   &fft   = envGet(FFT, fftName_);
+    double q      = par().charge;
+    envGetTmp(ScalarField, result); 
+    envGetTmp(ScalarField, buf); 
+
+    // G*F*Src
+    prop = GFSrc;
+
+    // - q*G*momD1*G*F*Src (momD1 = F*D1*Finv)
+    buf = GFSrc;
+    momD1(buf, fft);
+    buf = G*buf;
+    prop = prop - q*buf;
+
+    // + q^2*G*momD1*G*momD1*G*F*Src (here buf = G*momD1*G*F*Src)
+    momD1(buf, fft);
+    prop = prop + q*q*G*buf;
+
+    // - q^2*G*momD2*G*F*Src (momD2 = F*D2*Finv)
+    buf = GFSrc;
+    momD2(buf, fft);
+    prop = prop - q*q*G*buf;
+
+    // final FT
+    fft.FFT_all_dim(prop, prop, FFT::backward);
+    
+    // OUTPUT IF NECESSARY
+    if (!par().output.empty())
+    {
+        std::string           filename = par().output + "." +
+                                         std::to_string(vm().getTrajectory());
+        
+        LOG(Message) << "Saving zero-momentum projection to '"
+                     << filename << "'..." << std::endl;
+        
+        ResultWriter          writer(RESULT_FILE_NAME(par().output));
+        std::vector<TComplex> vecBuf;
+        std::vector<Complex>  result;
+        
+        sliceSum(prop, vecBuf, Tp);
+        result.resize(vecBuf.size());
+        for (unsigned int t = 0; t < vecBuf.size(); ++t)
+        {
+            result[t] = TensorRemove(vecBuf[t]);
+        }
+        write(writer, "charge", q);
+        write(writer, "prop", result);
+    }
+}
+
+void TChargedProp::makeCaches(void)
+{
+    auto &freeMomProp = envGet(ScalarField, freeMomPropName_);
+    auto &GFSrc       = envGet(ScalarField, GFSrcName_);
+    auto &fft         = envGet(FFT, fftName_);
+
+    if (!freeMomPropDone_)
+    {
+        LOG(Message) << "Caching momentum space free scalar propagator"
+                     << " (mass= " << par().mass << ")..." << std::endl;
+        SIMPL::MomentumSpacePropagator(freeMomProp, par().mass);
+    }
+    if (!GFSrcDone_)
+    {   
+        FFT  fft(env().getGrid());
+        auto &source = envGet(ScalarField, par().source);
+
+        LOG(Message) << "Caching G*F*src..." << std::endl;
+        fft.FFT_all_dim(GFSrc, source, FFT::forward);
+        GFSrc = freeMomProp*GFSrc;
+    }
+    if (!phasesDone_)
+    {
+        std::vector<int> &l = env().getGrid()->_fdimensions;
+        Complex          ci(0.0,1.0);
+        
+        LOG(Message) << "Caching shift phases..." << std::endl;
+        for (unsigned int mu = 0; mu < env().getNd(); ++mu)
+        {
+            Real twoPiL = M_PI*2./l[mu];
+            auto &phmu  = envGet(ScalarField, phaseName_[mu]);
+            
+            LatticeCoordinate(phmu, mu);
+            phmu = exp(ci*twoPiL*phmu);
+            phase_.push_back(&phmu);
+        }
+    }
+}
+
+void TChargedProp::momD1(ScalarField &s, FFT &fft)
+{
+    auto        &A = envGet(EmField, par().emField);
+    Complex     ci(0.0,1.0);
+
+    envGetTmp(ScalarField, buf);
+    envGetTmp(ScalarField, result);
+    envGetTmp(ScalarField, Amu);
+
+    result = zero;
+    for (unsigned int mu = 0; mu < env().getNd(); ++mu)
+    {
+        Amu = peekLorentz(A, mu);
+        buf = (*phase_[mu])*s;
+        fft.FFT_all_dim(buf, buf, FFT::backward);
+        buf = Amu*buf;
+        fft.FFT_all_dim(buf, buf, FFT::forward);
+        result = result - ci*buf;
+    }
+    fft.FFT_all_dim(s, s, FFT::backward);
+    for (unsigned int mu = 0; mu < env().getNd(); ++mu)
+    {
+        Amu = peekLorentz(A, mu);
+        buf = Amu*s;
+        fft.FFT_all_dim(buf, buf, FFT::forward);
+        result = result + ci*adj(*phase_[mu])*buf;
+    }
+
+    s = result;
+}
+
+void TChargedProp::momD2(ScalarField &s, FFT &fft)
+{
+    auto &A = envGet(EmField, par().emField);
+
+    envGetTmp(ScalarField, buf);
+    envGetTmp(ScalarField, result);
+    envGetTmp(ScalarField, Amu);
+
+    result = zero;
+    for (unsigned int mu = 0; mu < env().getNd(); ++mu)
+    {
+        Amu = peekLorentz(A, mu);
+        buf = (*phase_[mu])*s;
+        fft.FFT_all_dim(buf, buf, FFT::backward);
+        buf = Amu*Amu*buf;
+        fft.FFT_all_dim(buf, buf, FFT::forward);
+        result = result + .5*buf;
+    }
+    fft.FFT_all_dim(s, s, FFT::backward);
+    for (unsigned int mu = 0; mu < env().getNd(); ++mu)
+    {
+        Amu = peekLorentz(A, mu);        
+        buf = Amu*Amu*s;
+        fft.FFT_all_dim(buf, buf, FFT::forward);
+        result = result + .5*adj(*phase_[mu])*buf;
+    }
+
+    s = result;
+}

extras/Hadrons/Modules/MScalar/ChargedProp.hpp

@@ -0,0 +1,89 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MScalar/ChargedProp.hpp
+
+Copyright (C) 2015-2018
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+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 */
+#ifndef Hadrons_MScalar_ChargedProp_hpp_
+#define Hadrons_MScalar_ChargedProp_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                       Charged scalar propagator                            *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MScalar)
+
+class ChargedPropPar: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(ChargedPropPar,
+                                    std::string, emField,
+                                    std::string, source,
+                                    double,      mass,
+                                    double,      charge,
+                                    std::string, output);
+};
+
+class TChargedProp: public Module<ChargedPropPar>
+{
+public:
+    SCALAR_TYPE_ALIASES(SIMPL,);
+    typedef PhotonR::GaugeField     EmField;
+    typedef PhotonR::GaugeLinkField EmComp;
+public:
+    // constructor
+    TChargedProp(const std::string name);
+    // destructor
+    virtual ~TChargedProp(void) = default;
+    // dependency relation
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+protected:
+    // setup
+    virtual void setup(void);
+    // execution
+    virtual void execute(void);
+private:
+    void makeCaches(void);
+    void momD1(ScalarField &s, FFT &fft);
+    void momD2(ScalarField &s, FFT &fft);
+private:
+    bool                       freeMomPropDone_, GFSrcDone_, phasesDone_;
+    std::string                freeMomPropName_, GFSrcName_, fftName_;
+    std::vector<std::string>   phaseName_;
+    std::vector<ScalarField *> phase_;
+};
+
+MODULE_REGISTER_NS(ChargedProp, TChargedProp, MScalar);
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MScalar_ChargedProp_hpp_

extras/Hadrons/Modules/MScalar/FreeProp.cc

@@ -0,0 +1,99 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MScalar/FreeProp.cc
+
+Copyright (C) 2015-2018
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+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/Hadrons/Modules/MScalar/FreeProp.hpp>
+#include <Grid/Hadrons/Modules/MScalar/Scalar.hpp>
+
+using namespace Grid;
+using namespace Hadrons;
+using namespace MScalar;
+
+/******************************************************************************
+*                        TFreeProp implementation                             *
+******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+TFreeProp::TFreeProp(const std::string name)
+: Module<FreePropPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+std::vector<std::string> TFreeProp::getInput(void)
+{
+    std::vector<std::string> in = {par().source};
+    
+    return in;
+}
+
+std::vector<std::string> TFreeProp::getOutput(void)
+{
+    std::vector<std::string> out = {getName()};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+void TFreeProp::setup(void)
+{
+    freeMomPropName_ = FREEMOMPROP(par().mass);
+    
+    freePropDone_ = env().hasCreatedObject(freeMomPropName_);
+    envCacheLat(ScalarField, freeMomPropName_);
+    envCreateLat(ScalarField, getName());
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+void TFreeProp::execute(void)
+{
+    auto &freeMomProp = envGet(ScalarField, freeMomPropName_);
+    auto &prop        = envGet(ScalarField, getName());
+    auto &source      = envGet(ScalarField, par().source);
+
+    if (!freePropDone_)
+    {
+        LOG(Message) << "Caching momentum space free scalar propagator"
+                     << " (mass= " << par().mass << ")..." << std::endl;
+        SIMPL::MomentumSpacePropagator(freeMomProp, par().mass);
+    }
+    LOG(Message) << "Computing free scalar propagator..." << std::endl;
+    SIMPL::FreePropagator(source, prop, freeMomProp);
+    
+    if (!par().output.empty())
+    {
+        TextWriter            writer(par().output + "." +
+                                     std::to_string(vm().getTrajectory()));
+        std::vector<TComplex> buf;
+        std::vector<Complex>  result;
+        
+        sliceSum(prop, buf, Tp);
+        result.resize(buf.size());
+        for (unsigned int t = 0; t < buf.size(); ++t)
+        {
+            result[t] = TensorRemove(buf[t]);
+        }
+        write(writer, "prop", result);
+    }
+}

extras/Hadrons/Modules/MScalar/FreeProp.hpp

@@ -0,0 +1,79 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MScalar/FreeProp.hpp
+
+Copyright (C) 2015-2018
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+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 */
+#ifndef Hadrons_MScalar_FreeProp_hpp_
+#define Hadrons_MScalar_FreeProp_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                               FreeProp                                     *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MScalar)
+
+class FreePropPar: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(FreePropPar,
+                                    std::string, source,
+                                    double,      mass,
+                                    std::string, output);
+};
+
+class TFreeProp: public Module<FreePropPar>
+{
+public:
+    SCALAR_TYPE_ALIASES(SIMPL,);
+public:
+    // constructor
+    TFreeProp(const std::string name);
+    // destructor
+    virtual ~TFreeProp(void) = default;
+    // dependency relation
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+protected:
+    // setup
+    virtual void setup(void);
+    // execution
+    virtual void execute(void);
+private:
+    std::string freeMomPropName_;
+    bool        freePropDone_;
+};
+
+MODULE_REGISTER_NS(FreeProp, TFreeProp, MScalar);
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MScalar_FreeProp_hpp_

extras/Hadrons/Modules/MScalar/Scalar.hpp

@@ -0,0 +1,33 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MScalar/Scalar.hpp
+
+Copyright (C) 2015-2018
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+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 */
+#ifndef Hadrons_Scalar_hpp_
+#define Hadrons_Scalar_hpp_
+
+#define FREEMOMPROP(m) "_scalar_mom_prop_" + std::to_string(m)
+
+#endif // Hadrons_Scalar_hpp_

extras/Hadrons/Modules/MScalarSUN/Div.hpp

@@ -0,0 +1,166 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MScalarSUN/Div.hpp
+
+Copyright (C) 2015-2018
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+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 */
+#ifndef Hadrons_MScalarSUN_Div_hpp_
+#define Hadrons_MScalarSUN_Div_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                         Div                                 *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MScalarSUN)
+
+class DivPar: Serializable
+{
+public:
+    GRID_SERIALIZABLE_ENUM(DiffType, undef, forward, 1, backward, 2, central, 3);
+    GRID_SERIALIZABLE_CLASS_MEMBERS(DivPar,
+                                    std::vector<std::string>, op,
+                                    DiffType,                 type,
+                                    std::string,              output);
+};
+
+template <typename SImpl>
+class TDiv: public Module<DivPar>
+{
+public:
+    typedef typename SImpl::Field        Field;
+    typedef typename SImpl::ComplexField ComplexField;
+    class Result: Serializable
+    {
+    public:
+        GRID_SERIALIZABLE_CLASS_MEMBERS(Result,
+                                        DivPar::DiffType, type,
+                                        Complex,          value);
+    };
+public:
+    // constructor
+    TDiv(const std::string name);
+    // destructor
+    virtual ~TDiv(void) = default;
+    // dependency relation
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+    // setup
+    virtual void setup(void);
+    // execution
+    virtual void execute(void);
+};
+
+MODULE_REGISTER_NS(DivSU2, TDiv<ScalarNxNAdjImplR<2>>, MScalarSUN);
+MODULE_REGISTER_NS(DivSU3, TDiv<ScalarNxNAdjImplR<3>>, MScalarSUN);
+MODULE_REGISTER_NS(DivSU4, TDiv<ScalarNxNAdjImplR<4>>, MScalarSUN);
+MODULE_REGISTER_NS(DivSU5, TDiv<ScalarNxNAdjImplR<5>>, MScalarSUN);
+MODULE_REGISTER_NS(DivSU6, TDiv<ScalarNxNAdjImplR<6>>, MScalarSUN);
+
+/******************************************************************************
+ *                 TDiv implementation                             *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+template <typename SImpl>
+TDiv<SImpl>::TDiv(const std::string name)
+: Module<DivPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+template <typename SImpl>
+std::vector<std::string> TDiv<SImpl>::getInput(void)
+{
+    return par().op;
+}
+
+template <typename SImpl>
+std::vector<std::string> TDiv<SImpl>::getOutput(void)
+{
+    std::vector<std::string> out = {getName()};
+
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+template <typename SImpl>
+void TDiv<SImpl>::setup(void)
+{
+    if (par().op.size() != env().getNd())
+    {
+        HADRON_ERROR(Size, "the number of components differs from number of dimensions");
+    }
+    envCreateLat(ComplexField, getName());
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+template <typename SImpl>
+void TDiv<SImpl>::execute(void)
+{
+    const auto nd = env().getNd();
+
+    LOG(Message) << "Computing the " << par().type << " divergence of [";
+    for (unsigned int mu = 0; mu < nd; ++mu)
+    {
+        std::cout << par().op[mu] << ((mu == nd - 1) ? "]" : ", ");
+    }
+    std::cout << std::endl;
+
+    auto &div = envGet(ComplexField, getName());
+    div = zero;
+    for (unsigned int mu = 0; mu < nd; ++mu)
+    {
+        auto &op = envGet(ComplexField, par().op[mu]);
+        switch(par().type)
+        {
+            case DivPar::DiffType::backward:
+                div += op - Cshift(op, mu, -1);
+                break;
+            case DivPar::DiffType::forward:
+                div += Cshift(op, mu, 1) - op;
+                break;
+            case DivPar::DiffType::central:
+                div += 0.5*(Cshift(op, mu, 1) - Cshift(op, mu, -1));
+                break;
+        }
+    }
+    if (!par().output.empty())
+    {
+        Result       r;
+        ResultWriter writer(RESULT_FILE_NAME(par().output));
+
+        r.type  = par().type;
+        r.value = TensorRemove(sum(div));
+        write(writer, "div", r);
+    }
+}
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MScalarSUN_Div_hpp_

extras/Hadrons/Modules/MScalarSUN/TrMag.hpp

@@ -0,0 +1,146 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MScalarSUN/TrMag.hpp
+
+Copyright (C) 2015-2018
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+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 */
+#ifndef Hadrons_MScalarSUN_TrMag_hpp_
+#define Hadrons_MScalarSUN_TrMag_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                       Module to compute tr(mag^n)                          *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MScalarSUN)
+
+class TrMagPar: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(TrMagPar,
+                                    std::string,  field,
+                                    unsigned int, maxPow,
+                                    std::string,  output);
+};
+
+template <typename SImpl>
+class TTrMag: public Module<TrMagPar>
+{
+public:
+    typedef typename SImpl::Field        Field;
+    typedef typename SImpl::ComplexField ComplexField;
+    class Result: Serializable
+    {
+    public:
+        GRID_SERIALIZABLE_CLASS_MEMBERS(Result,
+                                        std::string, op,
+                                        Real,        value);
+    };
+public:
+    // constructor
+    TTrMag(const std::string name);
+    // destructor
+    virtual ~TTrMag(void) = default;
+    // dependency relation
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+    // setup
+    virtual void setup(void);
+    // execution
+    virtual void execute(void);
+};
+
+MODULE_REGISTER_NS(TrMagSU2, TTrMag<ScalarNxNAdjImplR<2>>, MScalarSUN);
+MODULE_REGISTER_NS(TrMagSU3, TTrMag<ScalarNxNAdjImplR<3>>, MScalarSUN);
+MODULE_REGISTER_NS(TrMagSU4, TTrMag<ScalarNxNAdjImplR<4>>, MScalarSUN);
+MODULE_REGISTER_NS(TrMagSU5, TTrMag<ScalarNxNAdjImplR<5>>, MScalarSUN);
+MODULE_REGISTER_NS(TrMagSU6, TTrMag<ScalarNxNAdjImplR<6>>, MScalarSUN);
+
+/******************************************************************************
+ *                         TTrMag implementation                              *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+template <typename SImpl>
+TTrMag<SImpl>::TTrMag(const std::string name)
+: Module<TrMagPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+template <typename SImpl>
+std::vector<std::string> TTrMag<SImpl>::getInput(void)
+{
+    std::vector<std::string> in = {par().field};
+    
+    return in;
+}
+
+template <typename SImpl>
+std::vector<std::string> TTrMag<SImpl>::getOutput(void)
+{
+    std::vector<std::string> out = {};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+template <typename SImpl>
+void TTrMag<SImpl>::setup(void)
+{}
+
+// execution ///////////////////////////////////////////////////////////////////
+template <typename SImpl>
+void TTrMag<SImpl>::execute(void)
+{
+    LOG(Message) << "Computing tr(mag^n) for n even up to " << par().maxPow
+                 << "..." << std::endl;
+
+    std::vector<Result> result;
+    ResultWriter        writer(RESULT_FILE_NAME(par().output));
+    auto                &phi = envGet(Field, par().field);
+
+    auto m2 = sum(phi), mn = m2;
+
+    m2 = -m2*m2;
+    mn = 1.;
+    for (unsigned int n = 2; n <= par().maxPow; n += 2)
+    {
+        Result r;
+
+        mn = mn*m2;
+        r.op    = "tr(mag^" + std::to_string(n) + ")";
+        r.value = TensorRemove(trace(mn)).real();
+        result.push_back(r);
+    }
+    write(writer, "trmag", result);
+}
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MScalarSUN_TrMag_hpp_

extras/Hadrons/Modules/MScalarSUN/TrPhi.hpp

@@ -0,0 +1,182 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MScalarSUN/TrPhi.hpp
+
+Copyright (C) 2015-2018
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+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 */
+#ifndef Hadrons_MScalarSUN_TrPhi_hpp_
+#define Hadrons_MScalarSUN_TrPhi_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                         Module to compute tr(phi^n)                        *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MScalarSUN)
+
+class TrPhiPar: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(TrPhiPar,
+                                    std::string,  field,
+                                    unsigned int, maxPow,
+                                    std::string,  output);
+};
+
+template <typename SImpl>
+class TTrPhi: public Module<TrPhiPar>
+{
+public:
+    typedef typename SImpl::Field        Field;
+    typedef typename SImpl::ComplexField ComplexField;
+    class Result: Serializable
+    {
+    public:
+        GRID_SERIALIZABLE_CLASS_MEMBERS(Result,
+                                        std::string, op,
+                                        Real,        value);
+    };
+public:
+    // constructor
+    TTrPhi(const std::string name);
+    // destructor
+    virtual ~TTrPhi(void) = default;
+    // dependency relation
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+    // setup
+    virtual void setup(void);
+    // execution
+    virtual void execute(void);
+private:
+    // output name generator
+    std::string outName(const unsigned int n);
+};
+
+MODULE_REGISTER_NS(TrPhiSU2, TTrPhi<ScalarNxNAdjImplR<2>>, MScalarSUN);
+MODULE_REGISTER_NS(TrPhiSU3, TTrPhi<ScalarNxNAdjImplR<3>>, MScalarSUN);
+MODULE_REGISTER_NS(TrPhiSU4, TTrPhi<ScalarNxNAdjImplR<4>>, MScalarSUN);
+MODULE_REGISTER_NS(TrPhiSU5, TTrPhi<ScalarNxNAdjImplR<5>>, MScalarSUN);
+MODULE_REGISTER_NS(TrPhiSU6, TTrPhi<ScalarNxNAdjImplR<6>>, MScalarSUN);
+
+/******************************************************************************
+ *                          TTrPhi implementation                             *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+template <typename SImpl>
+TTrPhi<SImpl>::TTrPhi(const std::string name)
+: Module<TrPhiPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+template <typename SImpl>
+std::vector<std::string> TTrPhi<SImpl>::getInput(void)
+{
+    std::vector<std::string> in = {par().field};
+    
+    return in;
+}
+
+template <typename SImpl>
+std::vector<std::string> TTrPhi<SImpl>::getOutput(void)
+{
+    std::vector<std::string> out;
+
+    for (unsigned int n = 2; n <= par().maxPow; n += 2)
+    {
+        out.push_back(outName(n));
+    }
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+template <typename SImpl>
+void TTrPhi<SImpl>::setup(void)
+{
+    if (par().maxPow < 2)
+    {
+        HADRON_ERROR(Size, "'maxPow' should be at least equal to 2");
+    }
+    envTmpLat(Field, "phi2");
+    envTmpLat(Field, "buf");
+    for (unsigned int n = 2; n <= par().maxPow; n += 2)
+    {
+        envCreateLat(ComplexField, outName(n));
+    }
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+template <typename SImpl>
+void TTrPhi<SImpl>::execute(void)
+{
+    LOG(Message) << "Computing tr(phi^n) for n even up to " << par().maxPow
+                 << "..." << std::endl; 
+
+    std::vector<Result> result;
+    auto                &phi = envGet(Field, par().field);
+
+    envGetTmp(Field, phi2);
+    envGetTmp(Field, buf);
+    buf  = 1.;
+    phi2 = -phi*phi; 
+    for (unsigned int n = 2; n <= par().maxPow; n += 2)
+    {
+        auto &phin = envGet(ComplexField, outName(n));
+
+        buf  = buf*phi2;
+        phin = trace(buf);
+        if (!par().output.empty())
+        {
+            Result r;
+
+            r.op    = "tr(phi^" + std::to_string(n) + ")";
+            r.value = TensorRemove(sum(phin)).real();
+            result.push_back(r);
+        }
+    }
+    if (result.size() > 0)
+    {
+        ResultWriter writer(RESULT_FILE_NAME(par().output));
+
+        write(writer, "trphi", result);
+    }
+}
+
+// output name generator ///////////////////////////////////////////////////////
+template <typename SImpl>
+std::string TTrPhi<SImpl>::outName(const unsigned int n)
+{
+    return getName() + "_" + std::to_string(n);
+}
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MScalarSUN_TrPhi_hpp_

extras/Hadrons/Modules/MScalarSUN/TwoPoint.hpp

@@ -0,0 +1,184 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MScalarSUN/TwoPoint.hpp
+
+Copyright (C) 2015-2018
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+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 */
+#ifndef Hadrons_MScalarSUN_TwoPoint_hpp_
+#define Hadrons_MScalarSUN_TwoPoint_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                 2-pt functions for a given set of operators                *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MScalarSUN)
+
+class TwoPointPar: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(TwoPointPar,
+                                    std::vector<std::string>, op,
+                                    std::string,              output);
+};
+
+template <typename SImpl>
+class TTwoPoint: public Module<TwoPointPar>
+{
+public:
+    typedef typename SImpl::Field        Field;
+    typedef typename SImpl::ComplexField ComplexField;
+    class Result: Serializable
+    {
+    public:
+        GRID_SERIALIZABLE_CLASS_MEMBERS(Result,
+                                        std::string, sink,
+                                        std::string, source,
+                                        std::vector<Complex>, data);
+    };
+public:
+    // constructor
+    TTwoPoint(const std::string name);
+    // destructor
+    virtual ~TTwoPoint(void) = default;
+    // dependency relation
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+    // setup
+    virtual void setup(void);
+    // execution
+    virtual void execute(void);
+private:
+    // make 2-pt function
+    template <class SinkSite, class SourceSite>
+    std::vector<Complex> makeTwoPoint(const std::vector<SinkSite>   &sink,
+                                      const std::vector<SourceSite> &source);
+};
+
+MODULE_REGISTER_NS(TwoPointSU2, TTwoPoint<ScalarNxNAdjImplR<2>>, MScalarSUN);
+MODULE_REGISTER_NS(TwoPointSU3, TTwoPoint<ScalarNxNAdjImplR<3>>, MScalarSUN);
+MODULE_REGISTER_NS(TwoPointSU4, TTwoPoint<ScalarNxNAdjImplR<4>>, MScalarSUN);
+MODULE_REGISTER_NS(TwoPointSU5, TTwoPoint<ScalarNxNAdjImplR<5>>, MScalarSUN);
+MODULE_REGISTER_NS(TwoPointSU6, TTwoPoint<ScalarNxNAdjImplR<6>>, MScalarSUN);
+
+/******************************************************************************
+ *                 TTwoPoint implementation                             *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+template <typename SImpl>
+TTwoPoint<SImpl>::TTwoPoint(const std::string name)
+: Module<TwoPointPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+template <typename SImpl>
+std::vector<std::string> TTwoPoint<SImpl>::getInput(void)
+{   
+    return par().op;
+}
+
+template <typename SImpl>
+std::vector<std::string> TTwoPoint<SImpl>::getOutput(void)
+{
+    std::vector<std::string> out = {};
+
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+template <typename SImpl>
+void TTwoPoint<SImpl>::setup(void)
+{
+    const unsigned int nt = env().getDim().back();
+    envTmp(std::vector<std::vector<TComplex>>, "slicedOp", 1, par().op.size(), 
+           std::vector<TComplex>(nt));
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+template <typename SImpl>
+void TTwoPoint<SImpl>::execute(void)
+{
+    LOG(Message) << "Computing 2-point functions for operators:" << std::endl;
+    for (auto &o: par().op)
+    {
+        LOG(Message) << "  '" << o << "'" << std::endl;
+    }
+
+    ResultWriter        writer(RESULT_FILE_NAME(par().output));
+    const unsigned int  nd = env().getDim().size();
+    std::vector<Result> result;
+    
+    envGetTmp(std::vector<std::vector<TComplex>>, slicedOp);
+    for (unsigned int i = 0; i < par().op.size(); ++i)
+    {
+        auto &op = envGet(ComplexField, par().op[i]);
+
+        sliceSum(op, slicedOp[i], nd - 1);
+    }
+    for (unsigned int i = 0; i < par().op.size(); ++i)
+    for (unsigned int j = 0; j < par().op.size(); ++j)
+    {
+        Result r;
+
+        r.sink   = par().op[i];
+        r.source = par().op[j];
+        r.data   = makeTwoPoint(slicedOp[i], slicedOp[j]);
+        result.push_back(r);
+    }
+    write(writer, "twopt", result);
+}
+
+// make 2-pt function //////////////////////////////////////////////////////////
+template <class SImpl>
+template <class SinkSite, class SourceSite>
+std::vector<Complex> TTwoPoint<SImpl>::makeTwoPoint(
+                                  const std::vector<SinkSite>   &sink,
+                                  const std::vector<SourceSite> &source)
+{
+    assert(sink.size() == source.size());
+    
+    unsigned int         nt = sink.size();
+    std::vector<Complex> res(nt, 0.);
+    
+    for (unsigned int dt = 0; dt < nt; ++dt)
+    {
+        for (unsigned int t  = 0; t < nt; ++t)
+        {
+            res[dt] += TensorRemove(trace(sink[(t+dt)%nt]*source[t]));
+        }
+        res[dt] *= 1./static_cast<double>(nt);
+    }
+    
+    return res;
+}
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MScalarSUN_TwoPoint_hpp_

extras/Hadrons/Modules/MSink/Point.hpp

@@ -0,0 +1,155 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MSink/Point.hpp
+
+Copyright (C) 2015-2018
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+Author: Lanny91 <andrew.lawson@gmail.com>
+
+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 */
+
+#ifndef Hadrons_MSink_Point_hpp_
+#define Hadrons_MSink_Point_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                                   Point                                    *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MSink)
+
+class PointPar: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(PointPar,
+                                    std::string, mom);
+};
+
+template <typename FImpl>
+class TPoint: public Module<PointPar>
+{
+public:
+    FERM_TYPE_ALIASES(FImpl,);
+    SINK_TYPE_ALIASES();
+public:
+    // constructor
+    TPoint(const std::string name);
+    // destructor
+    virtual ~TPoint(void) = default;
+    // dependency relation
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+protected:
+    // setup
+    virtual void setup(void);
+    // execution
+    virtual void execute(void);
+private:
+    bool        hasPhase_{false}; 
+    std::string momphName_;
+};
+
+MODULE_REGISTER_NS(Point,       TPoint<FIMPL>,        MSink);
+MODULE_REGISTER_NS(ScalarPoint, TPoint<ScalarImplCR>, MSink);
+
+/******************************************************************************
+ *                          TPoint implementation                             *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+template <typename FImpl>
+TPoint<FImpl>::TPoint(const std::string name)
+: Module<PointPar>(name)
+, momphName_ (name + "_momph")
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+template <typename FImpl>
+std::vector<std::string> TPoint<FImpl>::getInput(void)
+{
+    std::vector<std::string> in;
+    
+    return in;
+}
+
+template <typename FImpl>
+std::vector<std::string> TPoint<FImpl>::getOutput(void)
+{
+    std::vector<std::string> out = {getName()};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TPoint<FImpl>::setup(void)
+{
+    envTmpLat(LatticeComplex, "coor");
+    envCacheLat(LatticeComplex, momphName_);
+    envCreate(SinkFn, getName(), 1, nullptr);
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TPoint<FImpl>::execute(void)
+{   
+    LOG(Message) << "Setting up point sink function for momentum ["
+                 << par().mom << "]" << std::endl;
+
+    auto &ph = envGet(LatticeComplex, momphName_);
+    
+    if (!hasPhase_)
+    {
+        Complex           i(0.0,1.0);
+        std::vector<Real> p;
+
+        envGetTmp(LatticeComplex, coor);
+        p  = strToVec<Real>(par().mom);
+        ph = zero;
+        for(unsigned int mu = 0; mu < env().getNd(); mu++)
+        {
+            LatticeCoordinate(coor, mu);
+            ph = ph + (p[mu]/env().getGrid()->_fdimensions[mu])*coor;
+        }
+        ph = exp((Real)(2*M_PI)*i*ph);
+        hasPhase_ = true;
+    }
+    auto sink = [&ph](const PropagatorField &field)
+    {
+        SlicedPropagator res;
+        PropagatorField  tmp = ph*field;
+        
+        sliceSum(tmp, res, Tp);
+        
+        return res;
+    };
+    envGet(SinkFn, getName()) = sink;
+}
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MSink_Point_hpp_

extras/Hadrons/Modules/MSink/Smear.hpp

@@ -0,0 +1,127 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MSink/Smear.hpp
+
+Copyright (C) 2015-2018
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+Author: Lanny91 <andrew.lawson@gmail.com>
+
+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 */
+
+#ifndef Hadrons_MSink_Smear_hpp_
+#define Hadrons_MSink_Smear_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                                 Smear                                      *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MSink)
+
+class SmearPar: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(SmearPar,
+                                    std::string, q,
+                                    std::string, sink);
+};
+
+template <typename FImpl>
+class TSmear: public Module<SmearPar>
+{
+public:
+    FERM_TYPE_ALIASES(FImpl,);
+    SINK_TYPE_ALIASES();
+public:
+    // constructor
+    TSmear(const std::string name);
+    // destructor
+    virtual ~TSmear(void) = default;
+    // dependency relation
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+protected:
+    // setup
+    virtual void setup(void);
+    // execution
+    virtual void execute(void);
+};
+
+MODULE_REGISTER_NS(Smear, TSmear<FIMPL>, MSink);
+
+/******************************************************************************
+ *                          TSmear implementation                             *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+template <typename FImpl>
+TSmear<FImpl>::TSmear(const std::string name)
+: Module<SmearPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+template <typename FImpl>
+std::vector<std::string> TSmear<FImpl>::getInput(void)
+{
+    std::vector<std::string> in = {par().q, par().sink};
+    
+    return in;
+}
+
+template <typename FImpl>
+std::vector<std::string> TSmear<FImpl>::getOutput(void)
+{
+    std::vector<std::string> out = {getName()};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TSmear<FImpl>::setup(void)
+{
+    envCreate(SlicedPropagator, getName(), 1, env().getDim(Tp));
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TSmear<FImpl>::execute(void)
+{
+    LOG(Message) << "Sink smearing propagator '" << par().q
+                 << "' using sink function '" << par().sink << "'."
+                 << std::endl;
+
+    auto &sink = envGet(SinkFn, par().sink);
+    auto &q    = envGet(PropagatorField, par().q);
+    auto &out  = envGet(SlicedPropagator, getName());
+    
+    out = sink(q);
+}
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MSink_Smear_hpp_

extras/Hadrons/Modules/MSolver/RBPrecCG.hpp

@@ -4,8 +4,7 @@ Grid physics library, www.github.com/paboyle/Grid
 
 Source file: extras/Hadrons/Modules/MSolver/RBPrecCG.hpp
 
-Copyright (C) 2015
-Copyright (C) 2016
+Copyright (C) 2015-2018
 
 Author: Antonin Portelli <antonin.portelli@me.com>
 
@@ -27,8 +26,8 @@ See the full license in the file "LICENSE" in the top level distribution directo
 *************************************************************************************/
 /*  END LEGAL */
 
-#ifndef Hadrons_RBPrecCG_hpp_
-#define Hadrons_RBPrecCG_hpp_
+#ifndef Hadrons_MSolver_RBPrecCG_hpp_
+#define Hadrons_MSolver_RBPrecCG_hpp_
 
 #include <Grid/Hadrons/Global.hpp>
 #include <Grid/Hadrons/Module.hpp>
@@ -53,7 +52,7 @@ template <typename FImpl>
 class TRBPrecCG: public Module<RBPrecCGPar>
 {
 public:
-    TYPE_ALIASES(FImpl,);
+    FGS_TYPE_ALIASES(FImpl,);
 public:
     // constructor
     TRBPrecCG(const std::string name);
@@ -61,7 +60,9 @@ public:
     virtual ~TRBPrecCG(void) = default;
     // dependencies/products
     virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getReference(void);
     virtual std::vector<std::string> getOutput(void);
+protected:
     // setup
     virtual void setup(void);
     // execution
@@ -83,11 +84,19 @@ TRBPrecCG<FImpl>::TRBPrecCG(const std::string name)
 template <typename FImpl>
 std::vector<std::string> TRBPrecCG<FImpl>::getInput(void)
 {
-    std::vector<std::string> in = {par().action};
+    std::vector<std::string> in = {};
     
     return in;
 }
 
+template <typename FImpl>
+std::vector<std::string> TRBPrecCG<FImpl>::getReference(void)
+{
+    std::vector<std::string> ref = {par().action};
+    
+    return ref;
+}
+
 template <typename FImpl>
 std::vector<std::string> TRBPrecCG<FImpl>::getOutput(void)
 {
@@ -100,17 +109,12 @@ std::vector<std::string> TRBPrecCG<FImpl>::getOutput(void)
 template <typename FImpl>
 void TRBPrecCG<FImpl>::setup(void)
 {
-    auto Ls = env().getObjectLs(par().action);
-    
-    env().registerObject(getName(), 0, Ls);
-    env().addOwnership(getName(), par().action);
-}
+    LOG(Message) << "setting up Schur red-black preconditioned CG for"
+                 << " action '" << par().action << "' with residual "
+                 << par().residual << std::endl;
 
-// execution ///////////////////////////////////////////////////////////////////
-template <typename FImpl>
-void TRBPrecCG<FImpl>::execute(void)
-{
-    auto &mat   = *(env().template getObject<FMat>(par().action));
+    auto Ls     = env().getObjectLs(par().action);
+    auto &mat   = envGet(FMat, par().action);
     auto solver = [&mat, this](FermionField &sol, const FermionField &source)
     {
         ConjugateGradient<FermionField>           cg(par().residual, 10000);
@@ -118,15 +122,16 @@ void TRBPrecCG<FImpl>::execute(void)
         
         schurSolver(mat, source, sol);
     };
-    
-    LOG(Message) << "setting up Schur red-black preconditioned CG for"
-                 << " action '" << par().action << "' with residual "
-                 << par().residual << std::endl;
-    env().setObject(getName(), new SolverFn(solver));
+    envCreate(SolverFn, getName(), Ls, solver);
 }
 
+// execution ///////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TRBPrecCG<FImpl>::execute(void)
+{}
+
 END_MODULE_NAMESPACE
 
 END_HADRONS_NAMESPACE
 
-#endif // Hadrons_RBPrecCG_hpp_
+#endif // Hadrons_MSolver_RBPrecCG_hpp_

extras/Hadrons/Modules/MSource/Point.hpp

@@ -4,10 +4,10 @@ Grid physics library, www.github.com/paboyle/Grid
 
 Source file: extras/Hadrons/Modules/MSource/Point.hpp
 
-Copyright (C) 2015
-Copyright (C) 2016
+Copyright (C) 2015-2018
 
 Author: Antonin Portelli <antonin.portelli@me.com>
+Author: Lanny91 <andrew.lawson@gmail.com>
 
 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
@@ -27,8 +27,8 @@ See the full license in the file "LICENSE" in the top level distribution directo
 *************************************************************************************/
 /*  END LEGAL */
 
-#ifndef Hadrons_Point_hpp_
-#define Hadrons_Point_hpp_
+#ifndef Hadrons_MSource_Point_hpp_
+#define Hadrons_MSource_Point_hpp_
 
 #include <Grid/Hadrons/Global.hpp>
 #include <Grid/Hadrons/Module.hpp>
@@ -63,7 +63,7 @@ template <typename FImpl>
 class TPoint: public Module<PointPar>
 {
 public:
-    TYPE_ALIASES(FImpl,);
+    FERM_TYPE_ALIASES(FImpl,);
 public:
     // constructor
     TPoint(const std::string name);
@@ -72,13 +72,15 @@ public:
     // dependency relation
     virtual std::vector<std::string> getInput(void);
     virtual std::vector<std::string> getOutput(void);
+protected:
     // setup
     virtual void setup(void);
     // execution
     virtual void execute(void);
 };
 
-MODULE_REGISTER_NS(Point, TPoint<FIMPL>, MSource);
+MODULE_REGISTER_NS(Point,       TPoint<FIMPL>,        MSource);
+MODULE_REGISTER_NS(ScalarPoint, TPoint<ScalarImplCR>, MSource);
 
 /******************************************************************************
  *                       TPoint template implementation                       *
@@ -110,19 +112,20 @@ std::vector<std::string> TPoint<FImpl>::getOutput(void)
 template <typename FImpl>
 void TPoint<FImpl>::setup(void)
 {
-    env().template registerLattice<PropagatorField>(getName());
+    envCreateLat(PropagatorField, getName());
 }
 
 // execution ///////////////////////////////////////////////////////////////////
 template <typename FImpl>
 void TPoint<FImpl>::execute(void)
 {
-    std::vector<int> position = strToVec<int>(par().position);
-    typename SitePropagator::scalar_object id;
-    
     LOG(Message) << "Creating point source at position [" << par().position
-                 << "]" << std::endl;
-    PropagatorField &src = *env().template createLattice<PropagatorField>(getName());
+                << "]" << std::endl;
+
+    std::vector<int> position = strToVec<int>(par().position);
+    auto             &src     = envGet(PropagatorField, getName());
+    SitePropagator   id;
+    
     id  = 1.;
     src = zero;
     pokeSite(id, src, position);
@@ -132,4 +135,4 @@ END_MODULE_NAMESPACE
 
 END_HADRONS_NAMESPACE
 
-#endif // Hadrons_Point_hpp_
+#endif // Hadrons_MSource_Point_hpp_

extras/Hadrons/Modules/MSource/SeqConserved.hpp

@@ -0,0 +1,160 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MSource/SeqConserved.hpp
+
+Copyright (C) 2015-2018
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+Author: Lanny91 <andrew.lawson@gmail.com>
+
+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 */
+
+#ifndef Hadrons_MSource_SeqConserved_hpp_
+#define Hadrons_MSource_SeqConserved_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/*
+ 
+ Sequential source
+ -----------------------------
+ * src_x = q_x * theta(x_3 - tA) * theta(tB - x_3) * J_mu * exp(i x.mom)
+ 
+ * options:
+ - q: input propagator (string)
+ - action: fermion action used for propagator q (string)
+ - tA: begin timeslice (integer)
+ - tB: end timesilce (integer)
+ - curr_type: type of conserved current to insert (Current)
+ - mu: Lorentz index of current to insert (integer)
+ - mom: momentum insertion, space-separated float sequence (e.g ".1 .2 1. 0.")
+ 
+ */
+
+/******************************************************************************
+ *                              SeqConserved                                  *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MSource)
+
+class SeqConservedPar: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(SeqConservedPar,
+                                    std::string,  q,
+                                    std::string,  action,
+                                    unsigned int, tA,
+                                    unsigned int, tB,
+                                    Current,      curr_type,
+                                    unsigned int, mu,
+                                    std::string,  mom);
+};
+
+template <typename FImpl>
+class TSeqConserved: public Module<SeqConservedPar>
+{
+public:
+    FERM_TYPE_ALIASES(FImpl,);
+public:
+    // constructor
+    TSeqConserved(const std::string name);
+    // destructor
+    virtual ~TSeqConserved(void) = default;
+    // dependency relation
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+protected:
+    // setup
+    virtual void setup(void);
+    // execution
+    virtual void execute(void);
+};
+
+MODULE_REGISTER_NS(SeqConserved, TSeqConserved<FIMPL>, MSource);
+
+/******************************************************************************
+ *                      TSeqConserved implementation                          *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+template <typename FImpl>
+TSeqConserved<FImpl>::TSeqConserved(const std::string name)
+: Module<SeqConservedPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+template <typename FImpl>
+std::vector<std::string> TSeqConserved<FImpl>::getInput(void)
+{
+    std::vector<std::string> in = {par().q, par().action};
+    
+    return in;
+}
+
+template <typename FImpl>
+std::vector<std::string> TSeqConserved<FImpl>::getOutput(void)
+{
+    std::vector<std::string> out = {getName()};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TSeqConserved<FImpl>::setup(void)
+{
+    auto Ls_ = env().getObjectLs(par().action);
+    envCreateLat(PropagatorField, getName(), Ls_);
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TSeqConserved<FImpl>::execute(void)
+{
+    if (par().tA == par().tB)
+    {
+        LOG(Message) << "Generating sequential source with conserved "
+                     << par().curr_type << " current insertion (mu = " 
+                     << par().mu << ") at " << "t = " << par().tA << std::endl;
+    }
+    else
+    {
+        LOG(Message) << "Generating sequential source with conserved "
+                     << par().curr_type << " current insertion (mu = " 
+                     << par().mu << ") for " << par().tA << " <= t <= " 
+                     << par().tB << std::endl;
+    }
+    auto &src = envGet(PropagatorField, getName());
+    auto &q   = envGet(PropagatorField, par().q);
+    auto &mat = envGet(FMat, par().action);
+
+    std::vector<Real> mom = strToVec<Real>(par().mom);
+    mat.SeqConservedCurrent(q, src, par().curr_type, par().mu, 
+                            mom, par().tA, par().tB);
+}
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_SeqConserved_hpp_

extras/Hadrons/Modules/MSource/SeqGamma.hpp

@@ -4,10 +4,10 @@ Grid physics library, www.github.com/paboyle/Grid
 
 Source file: extras/Hadrons/Modules/MSource/SeqGamma.hpp
 
-Copyright (C) 2015
-Copyright (C) 2016
+Copyright (C) 2015-2018
 
 Author: Antonin Portelli <antonin.portelli@me.com>
+Author: Lanny91 <andrew.lawson@gmail.com>
 
 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
@@ -27,8 +27,8 @@ See the full license in the file "LICENSE" in the top level distribution directo
 *************************************************************************************/
 /*  END LEGAL */
 
-#ifndef Hadrons_SeqGamma_hpp_
-#define Hadrons_SeqGamma_hpp_
+#ifndef Hadrons_MSource_SeqGamma_hpp_
+#define Hadrons_MSource_SeqGamma_hpp_
 
 #include <Grid/Hadrons/Global.hpp>
 #include <Grid/Hadrons/Module.hpp>
@@ -71,7 +71,7 @@ template <typename FImpl>
 class TSeqGamma: public Module<SeqGammaPar>
 {
 public:
-    TYPE_ALIASES(FImpl,);
+    FGS_TYPE_ALIASES(FImpl,);
 public:
     // constructor
     TSeqGamma(const std::string name);
@@ -80,10 +80,14 @@ public:
     // dependency relation
     virtual std::vector<std::string> getInput(void);
     virtual std::vector<std::string> getOutput(void);
+protected:
     // setup
     virtual void setup(void);
     // execution
     virtual void execute(void);
+private:
+    bool        hasPhase_{false};
+    std::string momphName_, tName_;
 };
 
 MODULE_REGISTER_NS(SeqGamma, TSeqGamma<FIMPL>, MSource);
@@ -95,6 +99,8 @@ MODULE_REGISTER_NS(SeqGamma, TSeqGamma<FIMPL>, MSource);
 template <typename FImpl>
 TSeqGamma<FImpl>::TSeqGamma(const std::string name)
 : Module<SeqGammaPar>(name)
+, momphName_ (name + "_momph")
+, tName_ (name + "_t")
 {}
 
 // dependencies/products ///////////////////////////////////////////////////////
@@ -118,7 +124,10 @@ std::vector<std::string> TSeqGamma<FImpl>::getOutput(void)
 template <typename FImpl>
 void TSeqGamma<FImpl>::setup(void)
 {
-    env().template registerLattice<PropagatorField>(getName());
+    envCreateLat(PropagatorField, getName());
+    envCacheLat(Lattice<iScalar<vInteger>>, tName_);
+    envCacheLat(LatticeComplex, momphName_);
+    envTmpLat(LatticeComplex, "coor");
 }
 
 // execution ///////////////////////////////////////////////////////////////////
@@ -136,23 +145,29 @@ void TSeqGamma<FImpl>::execute(void)
                      << " sequential source for "
                      << par().tA << " <= t <= " << par().tB << std::endl;
     }
-    PropagatorField &src = *env().template createLattice<PropagatorField>(getName());
-    PropagatorField &q   = *env().template getObject<PropagatorField>(par().q);
-    Lattice<iScalar<vInteger>> t(env().getGrid());
-    LatticeComplex             ph(env().getGrid()), coor(env().getGrid());
-    Gamma                      g(par().gamma);
-    std::vector<Real>          p;
-    Complex                    i(0.0,1.0);
+    auto  &src = envGet(PropagatorField, getName());
+    auto  &q   = envGet(PropagatorField, par().q);
+    auto  &ph  = envGet(LatticeComplex, momphName_);
+    auto  &t   = envGet(Lattice<iScalar<vInteger>>, tName_);
+    Gamma g(par().gamma);
     
-    p  = strToVec<Real>(par().mom);
-    ph = zero;
-    for(unsigned int mu = 0; mu < env().getNd(); mu++)
+    if (!hasPhase_)
     {
-        LatticeCoordinate(coor, mu);
-        ph = ph + p[mu]*coor;
+        Complex           i(0.0,1.0);
+        std::vector<Real> p;
+
+        envGetTmp(LatticeComplex, coor);
+        p  = strToVec<Real>(par().mom);
+        ph = zero;
+        for(unsigned int mu = 0; mu < env().getNd(); mu++)
+        {
+            LatticeCoordinate(coor, mu);
+            ph = ph + (p[mu]/env().getGrid()->_fdimensions[mu])*coor;
+        }
+        ph = exp((Real)(2*M_PI)*i*ph);
+        LatticeCoordinate(t, Tp);
+        hasPhase_ = true;
     }
-    ph = exp(i*ph);
-    LatticeCoordinate(t, Tp);
     src = where((t >= par().tA) and (t <= par().tB), ph*(g*q), 0.*q);
 }
 
@@ -160,4 +175,4 @@ END_MODULE_NAMESPACE
 
 END_HADRONS_NAMESPACE
 
-#endif // Hadrons_SeqGamma_hpp_
+#endif // Hadrons_MSource_SeqGamma_hpp_

extras/Hadrons/Modules/MSource/Wall.hpp

@@ -0,0 +1,161 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MSource/Wall.hpp
+
+Copyright (C) 2015-2018
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+Author: Lanny91 <andrew.lawson@gmail.com>
+
+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 */
+
+#ifndef Hadrons_MSource_WallSource_hpp_
+#define Hadrons_MSource_WallSource_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/*
+ 
+ Wall source
+ -----------------------------
+ * src_x = delta(x_3 - tW) * exp(i x.mom)
+ 
+ * options:
+ - tW: source timeslice (integer)
+ - mom: momentum insertion, space-separated float sequence (e.g ".1 .2 1. 0.")
+ 
+ */
+
+/******************************************************************************
+ *                         Wall                                               *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MSource)
+
+class WallPar: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(WallPar,
+                                    unsigned int, tW,
+                                    std::string, mom);
+};
+
+template <typename FImpl>
+class TWall: public Module<WallPar>
+{
+public:
+    FERM_TYPE_ALIASES(FImpl,);
+public:
+    // constructor
+    TWall(const std::string name);
+    // destructor
+    virtual ~TWall(void) = default;
+    // dependency relation
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+protected:
+    // setup
+    virtual void setup(void);
+    // execution
+    virtual void execute(void);
+private:
+    bool        hasPhase_{false};
+    std::string momphName_, tName_;
+};
+
+MODULE_REGISTER_NS(Wall, TWall<FIMPL>, MSource);
+
+/******************************************************************************
+ *                 TWall implementation                                       *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+template <typename FImpl>
+TWall<FImpl>::TWall(const std::string name)
+: Module<WallPar>(name)
+, momphName_ (name + "_momph")
+, tName_ (name + "_t")
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+template <typename FImpl>
+std::vector<std::string> TWall<FImpl>::getInput(void)
+{
+    std::vector<std::string> in = {};
+    
+    return in;
+}
+
+template <typename FImpl>
+std::vector<std::string> TWall<FImpl>::getOutput(void)
+{
+    std::vector<std::string> out = {getName()};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TWall<FImpl>::setup(void)
+{
+    envCreateLat(PropagatorField, getName());
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TWall<FImpl>::execute(void)
+{    
+    LOG(Message) << "Generating wall source at t = " << par().tW 
+                 << " with momentum " << par().mom << std::endl;
+    
+    auto  &src = envGet(PropagatorField, getName());
+    auto  &ph  = envGet(LatticeComplex, momphName_);
+    auto  &t   = envGet(Lattice<iScalar<vInteger>>, tName_);
+    
+    if (!hasPhase_)
+    {
+        Complex           i(0.0,1.0);
+        std::vector<Real> p;
+
+        envGetTmp(LatticeComplex, coor);
+        p  = strToVec<Real>(par().mom);
+        ph = zero;
+        for(unsigned int mu = 0; mu < env().getNd(); mu++)
+        {
+            LatticeCoordinate(coor, mu);
+            ph = ph + (p[mu]/env().getGrid()->_fdimensions[mu])*coor;
+        }
+        ph = exp((Real)(2*M_PI)*i*ph);
+        LatticeCoordinate(t, Tp);
+        hasPhase_ = true;
+    }
+
+    src = 1.;
+    src = where((t == par().tW), src*ph, 0.*src);
+}
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_MSource_WallSource_hpp_

extras/Hadrons/Modules/MSource/Z2.hpp

@@ -4,8 +4,7 @@ Grid physics library, www.github.com/paboyle/Grid
 
 Source file: extras/Hadrons/Modules/MSource/Z2.hpp
 
-Copyright (C) 2015
-Copyright (C) 2016
+Copyright (C) 2015-2018
 
 Author: Antonin Portelli <antonin.portelli@me.com>
 
@@ -27,8 +26,8 @@ See the full license in the file "LICENSE" in the top level distribution directo
 *************************************************************************************/
 /*  END LEGAL */
 
-#ifndef Hadrons_Z2_hpp_
-#define Hadrons_Z2_hpp_
+#ifndef Hadrons_MSource_Z2_hpp_
+#define Hadrons_MSource_Z2_hpp_
 
 #include <Grid/Hadrons/Global.hpp>
 #include <Grid/Hadrons/Module.hpp>
@@ -67,7 +66,7 @@ template <typename FImpl>
 class TZ2: public Module<Z2Par>
 {
 public:
-    TYPE_ALIASES(FImpl,);
+    FERM_TYPE_ALIASES(FImpl,);
 public:
     // constructor
     TZ2(const std::string name);
@@ -76,13 +75,18 @@ public:
     // dependency relation
     virtual std::vector<std::string> getInput(void);
     virtual std::vector<std::string> getOutput(void);
+protected:
     // setup
     virtual void setup(void);
     // execution
     virtual void execute(void);
+private:
+    bool        hasT_{false};
+    std::string tName_;
 };
 
-MODULE_REGISTER_NS(Z2, TZ2<FIMPL>, MSource);
+MODULE_REGISTER_NS(Z2,       TZ2<FIMPL>,        MSource);
+MODULE_REGISTER_NS(ScalarZ2, TZ2<ScalarImplCR>, MSource);
 
 /******************************************************************************
  *                       TZ2 template implementation                          *
@@ -91,6 +95,7 @@ MODULE_REGISTER_NS(Z2, TZ2<FIMPL>, MSource);
 template <typename FImpl>
 TZ2<FImpl>::TZ2(const std::string name)
 : Module<Z2Par>(name)
+, tName_ (name + "_t")
 {}
 
 // dependencies/products ///////////////////////////////////////////////////////
@@ -114,29 +119,36 @@ std::vector<std::string> TZ2<FImpl>::getOutput(void)
 template <typename FImpl>
 void TZ2<FImpl>::setup(void)
 {
-    env().template registerLattice<PropagatorField>(getName());
+    envCreateLat(PropagatorField, getName());
+    envCacheLat(Lattice<iScalar<vInteger>>, tName_);
+    envTmpLat(LatticeComplex, "eta");
 }
 
 // execution ///////////////////////////////////////////////////////////////////
 template <typename FImpl>
 void TZ2<FImpl>::execute(void)
 {
-    Lattice<iScalar<vInteger>> t(env().getGrid());
-    LatticeComplex             eta(env().getGrid());
-    Complex                    shift(1., 1.);
-    
     if (par().tA == par().tB)
     {
         LOG(Message) << "Generating Z_2 wall source at t= " << par().tA
-        << std::endl;
+                     << std::endl;
     }
     else
     {
         LOG(Message) << "Generating Z_2 band for " << par().tA << " <= t <= "
-        << par().tB << std::endl;
+                     << par().tB << std::endl;
     }
-    PropagatorField &src = *env().template createLattice<PropagatorField>(getName());
-    LatticeCoordinate(t, Tp);
+    
+    auto    &src = envGet(PropagatorField, getName());
+    auto    &t   = envGet(Lattice<iScalar<vInteger>>, tName_);
+    Complex shift(1., 1.);
+
+    if (!hasT_)
+    {
+        LatticeCoordinate(t, Tp);
+        hasT_ = true;
+    }
+    envGetTmp(LatticeComplex, eta);
     bernoulli(*env().get4dRng(), eta);
     eta = (2.*eta - shift)*(1./::sqrt(2.));
     eta = where((t >= par().tA) and (t <= par().tB), eta, 0.*eta);
@@ -148,4 +160,4 @@ END_MODULE_NAMESPACE
 
 END_HADRONS_NAMESPACE
 
-#endif // Hadrons_Z2_hpp_
+#endif // Hadrons_MSource_Z2_hpp_

extras/Hadrons/Modules/MUtilities/TestSeqConserved.hpp

@@ -0,0 +1,186 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MUtilities/TestSeqConserved.hpp
+
+Copyright (C) 2015-2018
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+Author: Lanny91 <andrew.lawson@gmail.com>
+
+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 */
+
+#ifndef Hadrons_MUtilities_TestSeqConserved_hpp_
+#define Hadrons_MUtilities_TestSeqConserved_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/*
+  Ward Identity contractions using sequential propagators.
+ -----------------------------
+ 
+ * options:
+ - q:      point source propagator, 5D if available (string)
+ - qSeq:   result of sequential insertion of conserved current using q (string)
+ - action: action used for computation of q (string)
+ - origin: string giving point source origin of q (string)
+ - t_J:    time at which sequential current is inserted (int)
+ - mu:     Lorentz index of current inserted (int)
+ - curr:   current type, e.g. vector/axial (Current)
+*/
+
+/******************************************************************************
+ *                            TestSeqConserved                                *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MUtilities)
+
+class TestSeqConservedPar: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(TestSeqConservedPar,
+                                    std::string,  q,
+                                    std::string,  qSeq,
+                                    std::string,  action,
+                                    std::string,  origin,
+                                    unsigned int, t_J,
+                                    unsigned int, mu,
+                                    Current,      curr);
+};
+
+template <typename FImpl>
+class TTestSeqConserved: public Module<TestSeqConservedPar>
+{
+public:
+    FERM_TYPE_ALIASES(FImpl,);
+public:
+    // constructor
+    TTestSeqConserved(const std::string name);
+    // destructor
+    virtual ~TTestSeqConserved(void) = default;
+    // dependency relation
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+protected:
+    // setup
+    virtual void setup(void);
+    // execution
+    virtual void execute(void);
+};
+
+MODULE_REGISTER_NS(TestSeqConserved, TTestSeqConserved<FIMPL>, MUtilities);
+
+/******************************************************************************
+ *                     TTestSeqConserved implementation                       *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+template <typename FImpl>
+TTestSeqConserved<FImpl>::TTestSeqConserved(const std::string name)
+: Module<TestSeqConservedPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+template <typename FImpl>
+std::vector<std::string> TTestSeqConserved<FImpl>::getInput(void)
+{
+    std::vector<std::string> in = {par().q, par().qSeq, par().action};
+    
+    return in;
+}
+
+template <typename FImpl>
+std::vector<std::string> TTestSeqConserved<FImpl>::getOutput(void)
+{
+    std::vector<std::string> out = {getName()};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TTestSeqConserved<FImpl>::setup(void)
+{
+    auto Ls = env().getObjectLs(par().q);
+    if (Ls != env().getObjectLs(par().action))
+    {
+        HADRON_ERROR(Size, "Ls mismatch between quark action and propagator");
+    }
+    envTmpLat(PropagatorField, "tmp");
+    envTmpLat(LatticeComplex, "c");
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TTestSeqConserved<FImpl>::execute(void)
+{
+    // Check sequential insertion of current gives same result as conserved 
+    // current sink upon contraction. Assume q uses a point source.
+
+    auto                  &q    = envGet(PropagatorField, par().q);
+    auto                  &qSeq = envGet(PropagatorField, par().qSeq);
+    auto                  &act  = envGet(FMat, par().action);
+    Gamma                 g5(Gamma::Algebra::Gamma5);
+    Gamma::Algebra        gA = (par().curr == Current::Axial) ?
+                                  Gamma::Algebra::Gamma5 :
+                                  Gamma::Algebra::Identity;
+    Gamma                 g(gA);
+    SitePropagator        qSite;
+    Complex               test_S, test_V, check_S, check_V;
+    std::vector<TComplex> check_buf;
+    std::vector<int>      siteCoord;
+
+    envGetTmp(PropagatorField, tmp);
+    envGetTmp(LatticeComplex, c);
+    siteCoord = strToVec<int>(par().origin);
+    peekSite(qSite, qSeq, siteCoord);
+    test_S = trace(qSite*g);
+    test_V = trace(qSite*g*Gamma::gmu[par().mu]);
+    act.ContractConservedCurrent(q, q, tmp, par().curr, par().mu);
+    c = trace(tmp*g);
+    sliceSum(c, check_buf, Tp);
+    check_S = TensorRemove(check_buf[par().t_J]);
+
+    c = trace(tmp*g*Gamma::gmu[par().mu]);
+    sliceSum(c, check_buf, Tp);
+    check_V = TensorRemove(check_buf[par().t_J]);
+
+    LOG(Message) << "Test S  = " << abs(test_S)   << std::endl;
+    LOG(Message) << "Test V  = " << abs(test_V) << std::endl;
+    LOG(Message) << "Check S = " << abs(check_S) << std::endl;
+    LOG(Message) << "Check V = " << abs(check_V) << std::endl;
+
+    // Check difference = 0
+    check_S -= test_S;
+    check_V -= test_V;
+
+    LOG(Message) << "Consistency check for sequential conserved " 
+                 << par().curr << " current insertion: " << std::endl; 
+    LOG(Message) << "Diff S  = " << abs(check_S) << std::endl;
+    LOG(Message) << "Diff V  = " << abs(check_V) << std::endl;
+}
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_TestSeqConserved_hpp_

extras/Hadrons/Modules/MUtilities/TestSeqGamma.hpp

@@ -0,0 +1,150 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/Modules/MUtilities/TestSeqGamma.hpp
+
+Copyright (C) 2015-2018
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+Author: Lanny91 <andrew.lawson@gmail.com>
+
+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 */
+
+#ifndef Hadrons_MUtilities_TestSeqGamma_hpp_
+#define Hadrons_MUtilities_TestSeqGamma_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Module.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+/******************************************************************************
+ *                              TestSeqGamma                                  *
+ ******************************************************************************/
+BEGIN_MODULE_NAMESPACE(MUtilities)
+
+class TestSeqGammaPar: Serializable
+{
+public:
+    GRID_SERIALIZABLE_CLASS_MEMBERS(TestSeqGammaPar,
+                                    std::string,    q,
+                                    std::string,    qSeq,
+                                    std::string,    origin,
+                                    Gamma::Algebra, gamma,
+                                    unsigned int,   t_g);
+};
+
+template <typename FImpl>
+class TTestSeqGamma: public Module<TestSeqGammaPar>
+{
+public:
+    FERM_TYPE_ALIASES(FImpl,);
+public:
+    // constructor
+    TTestSeqGamma(const std::string name);
+    // destructor
+    virtual ~TTestSeqGamma(void) = default;
+    // dependency relation
+    virtual std::vector<std::string> getInput(void);
+    virtual std::vector<std::string> getOutput(void);
+protected:
+    // setup
+    virtual void setup(void);
+    // execution
+    virtual void execute(void);
+};
+
+MODULE_REGISTER_NS(TestSeqGamma, TTestSeqGamma<FIMPL>, MUtilities);
+
+/******************************************************************************
+ *                      TTestSeqGamma implementation                          *
+ ******************************************************************************/
+// constructor /////////////////////////////////////////////////////////////////
+template <typename FImpl>
+TTestSeqGamma<FImpl>::TTestSeqGamma(const std::string name)
+: Module<TestSeqGammaPar>(name)
+{}
+
+// dependencies/products ///////////////////////////////////////////////////////
+template <typename FImpl>
+std::vector<std::string> TTestSeqGamma<FImpl>::getInput(void)
+{
+    std::vector<std::string> in = {par().q, par().qSeq};
+    
+    return in;
+}
+
+template <typename FImpl>
+std::vector<std::string> TTestSeqGamma<FImpl>::getOutput(void)
+{
+    std::vector<std::string> out = {getName()};
+    
+    return out;
+}
+
+// setup ///////////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TTestSeqGamma<FImpl>::setup(void)
+{
+    envTmpLat(LatticeComplex, "c");
+}
+
+// execution ///////////////////////////////////////////////////////////////////
+template <typename FImpl>
+void TTestSeqGamma<FImpl>::execute(void)
+{
+    auto                  &q    = envGet(PropagatorField, par().q);
+    auto                  &qSeq = envGet(PropagatorField, par().qSeq);
+    Gamma                 g5(Gamma::Algebra::Gamma5);
+    Gamma                 g(par().gamma);
+    SitePropagator        qSite;
+    Complex               test, check;
+    std::vector<TComplex> check_buf;
+    std::vector<int>      siteCoord;
+
+    // Check sequential insertion of gamma matrix gives same result as 
+    // insertion of gamma at sink upon contraction. Assume q uses a point 
+    // source.
+    
+    envGetTmp(LatticeComplex, c);
+    siteCoord = strToVec<int>(par().origin);
+    peekSite(qSite, qSeq, siteCoord);
+    test = trace(g*qSite);
+
+    c = trace(adj(g)*g5*adj(q)*g5*g*q);
+    sliceSum(c, check_buf, Tp);
+    check = TensorRemove(check_buf[par().t_g]);
+
+    LOG(Message) << "Seq Result = " << abs(test)  << std::endl;
+    LOG(Message) << "Reference  = " << abs(check) << std::endl;
+
+    // Check difference = 0
+    check -= test;
+
+    LOG(Message) << "Consistency check for sequential " << par().gamma  
+                 << " insertion = " << abs(check) << std::endl;
+}
+
+END_MODULE_NAMESPACE
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_TestSeqGamma_hpp_

extras/Hadrons/Modules/templates/Module_in_NS.hpp.template

@@ -1,5 +1,5 @@
-#ifndef Hadrons____FILEBASENAME____hpp_
-#define Hadrons____FILEBASENAME____hpp_
+#ifndef Hadrons____NAMESPACE_______FILEBASENAME____hpp_
+#define Hadrons____NAMESPACE_______FILEBASENAME____hpp_
 
 #include <Grid/Hadrons/Global.hpp>
 #include <Grid/Hadrons/Module.hpp>
@@ -41,4 +41,4 @@ END_MODULE_NAMESPACE
 
 END_HADRONS_NAMESPACE
 
-#endif // Hadrons____FILEBASENAME____hpp_
+#endif // Hadrons____NAMESPACE_______FILEBASENAME____hpp_

extras/Hadrons/Modules/templates/Module_tmp_in_NS.hpp.template

@@ -1,5 +1,5 @@
-#ifndef Hadrons____FILEBASENAME____hpp_
-#define Hadrons____FILEBASENAME____hpp_
+#ifndef Hadrons____NAMESPACE_______FILEBASENAME____hpp_
+#define Hadrons____NAMESPACE_______FILEBASENAME____hpp_
 
 #include <Grid/Hadrons/Global.hpp>
 #include <Grid/Hadrons/Module.hpp>
@@ -82,4 +82,4 @@ END_MODULE_NAMESPACE
 
 END_HADRONS_NAMESPACE
 
-#endif // Hadrons____FILEBASENAME____hpp_
+#endif // Hadrons____NAMESPACE_______FILEBASENAME____hpp_

extras/Hadrons/VirtualMachine.cc

@@ -0,0 +1,622 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/VirtualMachine.cc
+
+Copyright (C) 2015-2018
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+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/Hadrons/VirtualMachine.hpp>
+#include <Grid/Hadrons/GeneticScheduler.hpp>
+#include <Grid/Hadrons/ModuleFactory.hpp>
+
+using namespace Grid;
+using namespace QCD;
+using namespace Hadrons;
+
+/******************************************************************************
+ *                      VirtualMachine implementation                         *
+ ******************************************************************************/
+// trajectory counter //////////////////////////////////////////////////////////
+void VirtualMachine::setTrajectory(const unsigned int traj)
+{
+    traj_ = traj;
+}
+
+unsigned int VirtualMachine::getTrajectory(void) const
+{
+    return traj_;
+}
+
+// module management ///////////////////////////////////////////////////////////
+void VirtualMachine::pushModule(VirtualMachine::ModPt &pt)
+{
+    std::string name = pt->getName();
+    
+    if (!hasModule(name))
+    {
+        std::vector<unsigned int> inputAddress;
+        unsigned int              address;
+        ModuleInfo                m;
+        
+        // module registration -------------------------------------------------
+        m.data = std::move(pt);
+        m.type = typeIdPt(*m.data.get());
+        m.name = name;
+        // input dependencies
+        for (auto &in: m.data->getInput())
+        {
+            if (!env().hasObject(in))
+            {
+                // if object does not exist, add it with no creator module
+                env().addObject(in , -1);
+            }
+            m.input.push_back(env().getObjectAddress(in));
+        }
+        // reference dependencies
+        for (auto &ref: m.data->getReference())
+        {
+            if (!env().hasObject(ref))
+            {
+                // if object does not exist, add it with no creator module
+                env().addObject(ref , -1);
+            }
+            m.input.push_back(env().getObjectAddress(ref));
+        }
+        auto inCopy = m.input;
+        // if module has inputs with references, they need to be added as
+        // an input
+        for (auto &in: inCopy)
+        {
+            int inm = env().getObjectModule(in);
+
+            if (inm > 0)
+            {
+                if (getModule(inm)->getReference().size() > 0)
+                {
+                    for (auto &rin: getModule(inm)->getReference())
+                    {
+                        m.input.push_back(env().getObjectAddress(rin));
+                    }
+                }
+            }
+        }
+        module_.push_back(std::move(m));
+        address              = static_cast<unsigned int>(module_.size() - 1);
+        moduleAddress_[name] = address;
+        // connecting outputs to potential inputs ------------------------------
+        for (auto &out: getModule(address)->getOutput())
+        {
+            if (!env().hasObject(out))
+            {
+                // output does not exists, add it
+                env().addObject(out, address);
+            }
+            else
+            {
+                if (env().getObjectModule(env().getObjectAddress(out)) < 0)
+                {
+                    // output exists but without creator, correct it
+                    env().setObjectModule(env().getObjectAddress(out), address);
+                }
+                else
+                {
+                    // output already fully registered, error
+                    HADRON_ERROR(Definition, "object '" + out
+                                 + "' is already produced by module '"
+                                 + module_[env().getObjectModule(out)].name
+                                 + "' (while pushing module '" + name + "')");
+                }
+                if (getModule(address)->getReference().size() > 0)
+                {
+                    // module has references, dependency should be propagated
+                    // to children modules; find module with `out` as an input
+                    // and add references to their input
+                    auto pred = [this, out](const ModuleInfo &n)
+                    {
+                        auto &in = n.input;
+                        auto it  = std::find(in.begin(), in.end(), 
+                                             env().getObjectAddress(out));
+                        
+                        return (it != in.end());
+                    };
+                    auto it = std::find_if(module_.begin(), module_.end(), pred);
+                    while (it != module_.end())
+                    {
+                        for (auto &ref: getModule(address)->getReference())
+                        {
+                            it->input.push_back(env().getObjectAddress(ref));
+                        }
+                        it = std::find_if(++it, module_.end(), pred);
+                    }   
+                }
+            }
+        }
+        graphOutdated_         = true;
+        memoryProfileOutdated_ = true;
+    }
+    else
+    {
+        HADRON_ERROR(Definition, "module '" + name + "' already exists");
+    }
+}
+
+unsigned int VirtualMachine::getNModule(void) const
+{
+    return module_.size();
+}
+
+void VirtualMachine::createModule(const std::string name, const std::string type,
+                                  XmlReader &reader)
+{
+    auto &factory = ModuleFactory::getInstance();
+    auto pt       = factory.create(type, name);
+    
+    pt->parseParameters(reader, "options");
+    pushModule(pt);
+}
+
+ModuleBase * VirtualMachine::getModule(const unsigned int address) const
+{
+    if (hasModule(address))
+    {
+        return module_[address].data.get();
+    }
+    else
+    {
+        HADRON_ERROR(Definition, "no module with address " + std::to_string(address));
+    }
+}
+
+ModuleBase * VirtualMachine::getModule(const std::string name) const
+{
+    return getModule(getModuleAddress(name));
+}
+
+unsigned int VirtualMachine::getModuleAddress(const std::string name) const
+{
+    if (hasModule(name))
+    {
+        return moduleAddress_.at(name);
+    }
+    else
+    {
+        HADRON_ERROR(Definition, "no module with name '" + name + "'");
+    }
+}
+
+std::string VirtualMachine::getModuleName(const unsigned int address) const
+{
+    if (hasModule(address))
+    {
+        return module_[address].name;
+    }
+    else
+    {
+        HADRON_ERROR(Definition, "no module with address " + std::to_string(address));
+    }
+}
+
+std::string VirtualMachine::getModuleType(const unsigned int address) const
+{
+    if (hasModule(address))
+    {
+        return typeName(module_[address].type);
+    }
+    else
+    {
+        HADRON_ERROR(Definition, "no module with address " + std::to_string(address));
+    }
+}
+
+std::string VirtualMachine::getModuleType(const std::string name) const
+{
+    return getModuleType(getModuleAddress(name));
+}
+
+std::string VirtualMachine::getModuleNamespace(const unsigned int address) const
+{
+    std::string type = getModuleType(address), ns;
+    
+    auto pos2 = type.rfind("::");
+    auto pos1 = type.rfind("::", pos2 - 2);
+    
+    return type.substr(pos1 + 2, pos2 - pos1 - 2);
+}
+
+std::string VirtualMachine::getModuleNamespace(const std::string name) const
+{
+    return getModuleNamespace(getModuleAddress(name));
+}
+
+bool VirtualMachine::hasModule(const unsigned int address) const
+{
+    return (address < module_.size());
+}
+
+bool VirtualMachine::hasModule(const std::string name) const
+{
+    return (moduleAddress_.find(name) != moduleAddress_.end());
+}
+
+// print VM content ////////////////////////////////////////////////////////////
+void VirtualMachine::printContent(void) const
+{
+    LOG(Debug) << "Modules: " << std::endl;
+    for (unsigned int i = 0; i < module_.size(); ++i)
+    {
+        LOG(Debug) << std::setw(4) << i << ": "
+                   << getModuleName(i) << std::endl;
+    }
+}
+
+// module graph ////////////////////////////////////////////////////////////////
+Graph<unsigned int> VirtualMachine::getModuleGraph(void)
+{
+    if (graphOutdated_)
+    {
+        makeModuleGraph();
+        graphOutdated_ = false;
+    }
+
+    return graph_;
+}
+
+void VirtualMachine::makeModuleGraph(void)
+{
+    Graph<unsigned int> graph;
+    
+    // create vertices
+    for (unsigned int m = 0; m < module_.size(); ++m)
+    {
+        graph.addVertex(m);
+    }
+    // create edges
+    for (unsigned int m = 0; m < module_.size(); ++m)
+    {
+        for (auto &in: module_[m].input)
+        {
+            graph.addEdge(env().getObjectModule(in), m);
+        }
+    }
+    graph_ = graph;
+}
+
+// memory profile //////////////////////////////////////////////////////////////
+const VirtualMachine::MemoryProfile & VirtualMachine::getMemoryProfile(void)
+{
+    if (memoryProfileOutdated_)
+    {
+        makeMemoryProfile();
+        memoryProfileOutdated_ = false;
+    }
+
+    return profile_;
+}
+
+void VirtualMachine::makeMemoryProfile(void)
+{
+    bool protect = env().objectsProtected();
+    bool hmsg    = HadronsLogMessage.isActive();
+    bool gmsg    = GridLogMessage.isActive();
+    bool err     = HadronsLogError.isActive();
+    auto program = getModuleGraph().topoSort();
+
+    resetProfile();
+    profile_.module.resize(getNModule());
+    env().protectObjects(false);
+    GridLogMessage.Active(false);
+    HadronsLogMessage.Active(false);
+    HadronsLogError.Active(false);
+    for (auto it = program.rbegin(); it != program.rend(); ++it) 
+    {
+        auto a = *it;
+
+        if (profile_.module[a].empty())
+        {
+            LOG(Debug) << "Profiling memory for module '" << module_[a].name
+                       << "' (" << a << ")..." << std::endl;
+            memoryProfile(a);
+            env().freeAll();
+        }
+    }
+    env().protectObjects(protect);
+    GridLogMessage.Active(gmsg);
+    HadronsLogMessage.Active(hmsg);
+    HadronsLogError.Active(err);
+    LOG(Debug) << "Memory profile:" << std::endl;
+    LOG(Debug) << "----------------" << std::endl;
+    for (unsigned int a = 0; a < profile_.module.size(); ++a)
+    {
+        LOG(Debug) << getModuleName(a) << " (" << a << ")" << std::endl;
+        for (auto &o: profile_.module[a])
+        {
+            LOG(Debug) << "|__ " << env().getObjectName(o.first) << " ("
+                       << sizeString(o.second) << ")" << std::endl;
+        }
+        LOG(Debug) << std::endl;
+    }
+    LOG(Debug) << "----------------" << std::endl;
+}
+
+void VirtualMachine::resetProfile(void)
+{
+    profile_.module.clear();
+    profile_.object.clear();
+}
+
+void VirtualMachine::resizeProfile(void)
+{
+    if (env().getMaxAddress() > profile_.object.size())
+    {
+        MemoryPrint empty;
+
+        empty.size   = 0;
+        empty.module = -1;
+        profile_.object.resize(env().getMaxAddress(), empty);
+    }
+}
+
+void VirtualMachine::updateProfile(const unsigned int address)
+{
+    resizeProfile();
+    for (unsigned int a = 0; a < env().getMaxAddress(); ++a)
+    {
+        if (env().hasCreatedObject(a) and (profile_.object[a].module == -1))
+        {
+            profile_.object[a].size     = env().getObjectSize(a);
+            profile_.object[a].storage  = env().getObjectStorage(a);
+            profile_.object[a].module   = address;
+            profile_.module[address][a] = profile_.object[a].size;
+            if (env().getObjectModule(a) < 0)
+            {
+                env().setObjectModule(a, address);
+            }
+        }
+    }
+}
+
+void VirtualMachine::cleanEnvironment(void)
+{
+    resizeProfile();
+    for (unsigned int a = 0; a < env().getMaxAddress(); ++a)
+    {
+        if (env().hasCreatedObject(a) and (profile_.object[a].module == -1))
+        {
+            env().freeObject(a);
+        }
+    }
+}
+
+void VirtualMachine::memoryProfile(const unsigned int address)
+{
+    auto m = getModule(address);
+
+    LOG(Debug) << "Setting up module '" << m->getName() 
+               << "' (" << address << ")..." << std::endl;
+    try
+    {
+        m->setup();
+        updateProfile(address);
+    }
+    catch (Exceptions::Definition &)
+    {
+        cleanEnvironment();
+        for (auto &in: m->getInput())
+        {
+            memoryProfile(env().getObjectModule(in));
+        }
+        for (auto &ref: m->getReference())
+        {
+            memoryProfile(env().getObjectModule(ref));
+        }
+        m->setup();
+        updateProfile(address);
+    }
+}
+
+void VirtualMachine::memoryProfile(const std::string name)
+{
+    memoryProfile(getModuleAddress(name));
+}
+
+// garbage collector ///////////////////////////////////////////////////////////
+VirtualMachine::GarbageSchedule 
+VirtualMachine::makeGarbageSchedule(const Program &p) const
+{
+    GarbageSchedule freeProg;
+    
+    freeProg.resize(p.size());
+    for (unsigned int a = 0; a < env().getMaxAddress(); ++a)
+    {
+        if (env().getObjectStorage(a) == Environment::Storage::temporary)
+        {
+            auto it = std::find(p.begin(), p.end(), env().getObjectModule(a));
+
+            if (it != p.end())
+            {
+                freeProg[std::distance(p.begin(), it)].insert(a);
+            }
+        }
+        else if (env().getObjectStorage(a) == Environment::Storage::object)
+        {
+            auto pred = [a, this](const unsigned int b)
+            {
+                auto &in = module_[b].input;
+                auto it  = std::find(in.begin(), in.end(), a);
+                
+                return (it != in.end()) or (b == env().getObjectModule(a));
+            };
+            auto it = std::find_if(p.rbegin(), p.rend(), pred);
+            if (it != p.rend())
+            {
+                freeProg[std::distance(it, p.rend()) - 1].insert(a);
+            }
+        }
+    }
+
+    return freeProg;
+}
+
+// high-water memory function //////////////////////////////////////////////////
+VirtualMachine::Size VirtualMachine::memoryNeeded(const Program &p)
+{
+    const MemoryProfile &profile = getMemoryProfile();
+    GarbageSchedule     freep    = makeGarbageSchedule(p);
+    Size                current = 0, max = 0;
+
+    for (unsigned int i = 0; i < p.size(); ++i)
+    {
+        for (auto &o: profile.module[p[i]])
+        {
+            current += o.second;
+        }
+        max = std::max(current, max);
+        for (auto &o: freep[i])
+        {
+            current -= profile.object[o].size;
+        }
+    }
+
+    return max;
+}
+
+// genetic scheduler ///////////////////////////////////////////////////////////
+VirtualMachine::Program VirtualMachine::schedule(const GeneticPar &par)
+{
+    typedef GeneticScheduler<Size, unsigned int> Scheduler;
+
+    auto graph = getModuleGraph();
+
+    //constrained topological sort using a genetic algorithm
+    LOG(Message) << "Scheduling computation..." << std::endl;
+    LOG(Message) << "               #module= " << graph.size() << std::endl;
+    LOG(Message) << "       population size= " << par.popSize << std::endl;
+    LOG(Message) << "       max. generation= " << par.maxGen << std::endl;
+    LOG(Message) << "  max. cst. generation= " << par.maxCstGen << std::endl;
+    LOG(Message) << "         mutation rate= " << par.mutationRate << std::endl;
+    
+    unsigned int          k = 0, gen, prevPeak, nCstPeak = 0;
+    std::random_device    rd;
+    Scheduler::Parameters gpar;
+    
+    gpar.popSize      = par.popSize;
+    gpar.mutationRate = par.mutationRate;
+    gpar.seed         = rd();
+    CartesianCommunicator::BroadcastWorld(0, &(gpar.seed), sizeof(gpar.seed));
+    Scheduler::ObjFunc memPeak = [this](const Program &p)->Size
+    {
+        return memoryNeeded(p);
+    };
+    Scheduler scheduler(graph, memPeak, gpar);
+    gen = 0;
+    do
+    {
+        LOG(Debug) << "Generation " << gen << ":" << std::endl;
+        scheduler.nextGeneration();
+        if (gen != 0)
+        {
+            if (prevPeak == scheduler.getMinValue())
+            {
+                nCstPeak++;
+            }
+            else
+            {
+                nCstPeak = 0;
+            }
+        }
+        
+        prevPeak = scheduler.getMinValue();
+        if (gen % 10 == 0)
+        {
+            LOG(Iterative) << "Generation " << gen << ": "
+                           << sizeString(scheduler.getMinValue()) << std::endl;
+        }
+        
+        gen++;
+    } while ((gen < par.maxGen) and (nCstPeak < par.maxCstGen));
+    
+    return scheduler.getMinSchedule();
+}
+
+// general execution ///////////////////////////////////////////////////////////
+#define BIG_SEP "==============="
+#define SEP     "---------------"
+#define MEM_MSG(size) sizeString(size)
+
+void VirtualMachine::executeProgram(const Program &p) const
+{
+    Size            memPeak = 0, sizeBefore, sizeAfter;
+    GarbageSchedule freeProg;
+    
+    // build garbage collection schedule
+    LOG(Debug) << "Building garbage collection schedule..." << std::endl;
+    freeProg = makeGarbageSchedule(p);
+
+    // program execution
+    LOG(Debug) << "Executing program..." << std::endl;
+    for (unsigned int i = 0; i < p.size(); ++i)
+    {
+        // execute module
+        LOG(Message) << SEP << " Measurement step " << i + 1 << "/"
+                     << p.size() << " (module '" << module_[p[i]].name
+                     << "') " << SEP << std::endl;
+        (*module_[p[i]].data)();
+        sizeBefore = env().getTotalSize();
+        // print used memory after execution
+        LOG(Message) << "Allocated objects: " << MEM_MSG(sizeBefore)
+                     << std::endl;
+        if (sizeBefore > memPeak)
+        {
+            memPeak = sizeBefore;
+        }
+        // garbage collection for step i
+        LOG(Message) << "Garbage collection..." << std::endl;
+        for (auto &j: freeProg[i])
+        {
+            env().freeObject(j);
+        }
+        // print used memory after garbage collection if necessary
+        sizeAfter = env().getTotalSize();
+        if (sizeBefore != sizeAfter)
+        {
+            LOG(Message) << "Allocated objects: " << MEM_MSG(sizeAfter)
+                            << std::endl;
+        }
+        else
+        {
+            LOG(Message) << "Nothing to free" << std::endl;
+        }
+    }
+}
+
+void VirtualMachine::executeProgram(const std::vector<std::string> &p) const
+{
+    Program pAddress;
+    
+    for (auto &n: p)
+    {
+        pAddress.push_back(getModuleAddress(n));
+    }
+    executeProgram(pAddress);
+}

extras/Hadrons/VirtualMachine.hpp

@@ -0,0 +1,207 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid 
+
+Source file: extras/Hadrons/VirtualMachine.hpp
+
+Copyright (C) 2015-2018
+
+Author: Antonin Portelli <antonin.portelli@me.com>
+
+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 */
+
+#ifndef Hadrons_VirtualMachine_hpp_
+#define Hadrons_VirtualMachine_hpp_
+
+#include <Grid/Hadrons/Global.hpp>
+#include <Grid/Hadrons/Graph.hpp>
+#include <Grid/Hadrons/Environment.hpp>
+
+BEGIN_HADRONS_NAMESPACE
+
+#define DEFINE_VM_ALIAS \
+inline VirtualMachine & vm(void) const\
+{\
+    return VirtualMachine::getInstance();\
+}
+
+/******************************************************************************
+ *                   Virtual machine for module execution                     *
+ ******************************************************************************/
+// forward declaration of Module
+class ModuleBase;
+
+class VirtualMachine
+{
+    SINGLETON_DEFCTOR(VirtualMachine);
+public:
+    typedef SITE_SIZE_TYPE                      Size;
+    typedef std::unique_ptr<ModuleBase>         ModPt;
+    typedef std::vector<std::set<unsigned int>> GarbageSchedule;
+    typedef std::vector<unsigned int>           Program;
+    struct MemoryPrint
+    {
+        Size                 size;
+        Environment::Storage storage;
+        int                  module;
+    };
+    struct MemoryProfile
+    {
+        std::vector<std::map<unsigned int, Size>> module;
+        std::vector<MemoryPrint>                  object;
+    };
+    class GeneticPar: Serializable
+    {
+    public:
+        GeneticPar(void):
+            popSize{20}, maxGen{1000}, maxCstGen{100}, mutationRate{.1} {};
+    public:
+        GRID_SERIALIZABLE_CLASS_MEMBERS(GeneticPar,
+                                        unsigned int, popSize,
+                                        unsigned int, maxGen,
+                                        unsigned int, maxCstGen,
+                                        double      , mutationRate);
+    };
+private:
+    struct ModuleInfo
+    {
+        const std::type_info      *type{nullptr};
+        std::string               name;
+        ModPt                     data{nullptr};
+        std::vector<unsigned int> input;
+        size_t                    maxAllocated;
+    };
+public:
+    // trajectory counter
+    void                setTrajectory(const unsigned int traj);
+    unsigned int        getTrajectory(void) const;
+    // module management
+    void                pushModule(ModPt &pt);
+    template <typename M>
+    void                createModule(const std::string name);
+    template <typename M>
+    void                createModule(const std::string name,
+                                         const typename M::Par &par);
+    void                createModule(const std::string name,
+                                         const std::string type,
+                                         XmlReader &reader);
+    unsigned int        getNModule(void) const;
+    ModuleBase *        getModule(const unsigned int address) const;
+    ModuleBase *        getModule(const std::string name) const;
+    template <typename M>
+    M *                 getModule(const unsigned int address) const;
+    template <typename M>
+    M *                 getModule(const std::string name) const;
+    unsigned int        getModuleAddress(const std::string name) const;
+    std::string         getModuleName(const unsigned int address) const;
+    std::string         getModuleType(const unsigned int address) const;
+    std::string         getModuleType(const std::string name) const;
+    std::string         getModuleNamespace(const unsigned int address) const;
+    std::string         getModuleNamespace(const std::string name) const;
+    bool                hasModule(const unsigned int address) const;
+    bool                hasModule(const std::string name) const;
+    // print VM content
+    void                printContent(void) const;
+    // module graph (could be a const reference if topoSort was const)
+    Graph<unsigned int> getModuleGraph(void);
+    // memory profile
+    const MemoryProfile &getMemoryProfile(void);
+    // garbage collector
+    GarbageSchedule     makeGarbageSchedule(const Program &p) const;
+    // high-water memory function
+    Size                memoryNeeded(const Program &p);
+    // genetic scheduler
+    Program             schedule(const GeneticPar &par);
+    // general execution
+    void                executeProgram(const Program &p) const;
+    void                executeProgram(const std::vector<std::string> &p) const;
+private:
+    // environment shortcut
+    DEFINE_ENV_ALIAS;
+    // module graph
+    void makeModuleGraph(void);
+    // memory profile
+    void makeMemoryProfile(void);
+    void resetProfile(void);
+    void resizeProfile(void);
+    void updateProfile(const unsigned int address);
+    void cleanEnvironment(void);
+    void memoryProfile(const std::string name);
+    void memoryProfile(const unsigned int address);
+private:
+    // general
+    unsigned int                        traj_;
+    // module and related maps
+    std::vector<ModuleInfo>             module_;
+    std::map<std::string, unsigned int> moduleAddress_;
+    std::string                         currentModule_{""};
+    // module graph
+    bool                                graphOutdated_{true};
+    Graph<unsigned int>                 graph_;
+    // memory profile
+    bool                                memoryProfileOutdated_{true};
+    MemoryProfile                       profile_;
+};
+
+/******************************************************************************
+ *                   VirtualMachine template implementation                   *
+ ******************************************************************************/
+// module management ///////////////////////////////////////////////////////////
+template <typename M>
+void VirtualMachine::createModule(const std::string name)
+{
+    ModPt pt(new M(name));
+    
+    pushModule(pt);
+}
+
+template <typename M>
+void VirtualMachine::createModule(const std::string name,
+                               const typename M::Par &par)
+{
+    ModPt pt(new M(name));
+    
+    static_cast<M *>(pt.get())->setPar(par);
+    pushModule(pt);
+}
+
+template <typename M>
+M * VirtualMachine::getModule(const unsigned int address) const
+{
+    if (auto *pt = dynamic_cast<M *>(getModule(address)))
+    {
+        return pt;
+    }
+    else
+    {
+        HADRON_ERROR(Definition, "module '" + module_[address].name
+                     + "' does not have type " + typeid(M).name()
+                     + "(has type: " + getModuleType(address) + ")");
+    }
+}
+
+template <typename M>
+M * VirtualMachine::getModule(const std::string name) const
+{
+    return getModule<M>(getModuleAddress(name));
+}
+
+END_HADRONS_NAMESPACE
+
+#endif // Hadrons_VirtualMachine_hpp_

extras/Hadrons/modules.inc

@@ -1,19 +1,53 @@
 modules_cc =\
-  Modules/MGauge/Load.cc \
+  Modules/MScalar/ChargedProp.cc \
+  Modules/MScalar/FreeProp.cc \
+  Modules/MContraction/WeakHamiltonianEye.cc \
+  Modules/MContraction/WeakNeutral4ptDisc.cc \
+  Modules/MContraction/WeakHamiltonianNonEye.cc \
+  Modules/MGauge/Unit.cc \
+  Modules/MGauge/StochEm.cc \
   Modules/MGauge/Random.cc \
-  Modules/MGauge/Unit.cc
+  Modules/MGauge/FundtoHirep.cc \
+  Modules/MScalar/FreeProp.cc \
+  Modules/MScalar/ChargedProp.cc \
+  Modules/MIO/LoadNersc.cc
 
 modules_hpp =\
-  Modules/MAction/DWF.hpp \
-  Modules/MAction/Wilson.hpp \
   Modules/MContraction/Baryon.hpp \
   Modules/MContraction/Meson.hpp \
-  Modules/MGauge/Load.hpp \
-  Modules/MGauge/Random.hpp \
-  Modules/MGauge/Unit.hpp \
-  Modules/MSolver/RBPrecCG.hpp \
-  Modules/MSource/Point.hpp \
+  Modules/MContraction/WeakHamiltonian.hpp \
+  Modules/MContraction/WeakHamiltonianNonEye.hpp \
+  Modules/MContraction/DiscLoop.hpp \
+  Modules/MContraction/WeakNeutral4ptDisc.hpp \
+  Modules/MContraction/Gamma3pt.hpp \
+  Modules/MContraction/WardIdentity.hpp \
+  Modules/MContraction/WeakHamiltonianEye.hpp \
+  Modules/MFermion/GaugeProp.hpp \
   Modules/MSource/SeqGamma.hpp \
+  Modules/MSource/Point.hpp \
+  Modules/MSource/Wall.hpp \
   Modules/MSource/Z2.hpp \
-  Modules/Quark.hpp
+  Modules/MSource/SeqConserved.hpp \
+  Modules/MSink/Smear.hpp \
+  Modules/MSink/Point.hpp \
+  Modules/MSolver/RBPrecCG.hpp \
+  Modules/MGauge/Unit.hpp \
+  Modules/MGauge/Random.hpp \
+  Modules/MGauge/StochEm.hpp \
+  Modules/MGauge/FundtoHirep.hpp \
+  Modules/MUtilities/TestSeqGamma.hpp \
+  Modules/MUtilities/TestSeqConserved.hpp \
+  Modules/MLoop/NoiseLoop.hpp \
+  Modules/MScalar/FreeProp.hpp \
+  Modules/MScalar/Scalar.hpp \
+  Modules/MScalar/ChargedProp.hpp \
+  Modules/MAction/DWF.hpp \
+  Modules/MAction/Wilson.hpp \
+  Modules/MAction/WilsonClover.hpp \
+  Modules/MScalarSUN/Div.hpp \
+  Modules/MScalarSUN/TrMag.hpp \
+  Modules/MScalarSUN/TwoPoint.hpp \
+  Modules/MScalarSUN/TrPhi.hpp \
+  Modules/MIO/LoadNersc.hpp \
+  Modules/MIO/LoadBinary.hpp
 

extras/qed-fvol/Global.cc

@@ -0,0 +1,11 @@
+#include <qed-fvol/Global.hpp>
+
+using namespace Grid;
+using namespace QCD;
+using namespace QedFVol;
+
+QedFVolLogger QedFVol::QedFVolLogError(1,"Error");
+QedFVolLogger QedFVol::QedFVolLogWarning(1,"Warning");
+QedFVolLogger QedFVol::QedFVolLogMessage(1,"Message");
+QedFVolLogger QedFVol::QedFVolLogIterative(1,"Iterative");
+QedFVolLogger QedFVol::QedFVolLogDebug(1,"Debug");

extras/qed-fvol/Global.hpp

@@ -0,0 +1,42 @@
+#ifndef QedFVol_Global_hpp_
+#define QedFVol_Global_hpp_
+
+#include <Grid/Grid.h>
+
+#define BEGIN_QEDFVOL_NAMESPACE \
+namespace Grid {\
+using namespace QCD;\
+namespace QedFVol {\
+using Grid::operator<<;
+#define END_QEDFVOL_NAMESPACE }}
+
+/* the 'using Grid::operator<<;' statement prevents a very nasty compilation
+ * error with GCC (clang compiles fine without it).
+ */
+
+BEGIN_QEDFVOL_NAMESPACE
+
+class QedFVolLogger: public Logger
+{
+public:
+    QedFVolLogger(int on, std::string nm): Logger("QedFVol", on, nm,
+                                                  GridLogColours, "BLACK"){};
+};
+
+#define LOG(channel) std::cout << QedFVolLog##channel
+#define QEDFVOL_ERROR(msg)\
+LOG(Error) << msg << " (" << __FUNCTION__ << " at " << __FILE__ << ":"\
+           << __LINE__ << ")" << std::endl;\
+abort();
+
+#define DEBUG_VAR(var) LOG(Debug) << #var << "= " << (var) << std::endl;
+
+extern QedFVolLogger QedFVolLogError;
+extern QedFVolLogger QedFVolLogWarning;
+extern QedFVolLogger QedFVolLogMessage;
+extern QedFVolLogger QedFVolLogIterative;
+extern QedFVolLogger QedFVolLogDebug;
+
+END_QEDFVOL_NAMESPACE
+
+#endif // QedFVol_Global_hpp_

extras/qed-fvol/Makefile.am

@@ -0,0 +1,9 @@
+AM_CXXFLAGS += -I$(top_srcdir)/extras
+
+bin_PROGRAMS = qed-fvol
+
+qed_fvol_SOURCES =   \
+    qed-fvol.cc      \
+    Global.cc
+
+qed_fvol_LDADD   = -lGrid

extras/qed-fvol/WilsonLoops.h

@@ -0,0 +1,265 @@
+#ifndef QEDFVOL_WILSONLOOPS_H
+#define QEDFVOL_WILSONLOOPS_H
+
+#include <Global.hpp>
+
+BEGIN_QEDFVOL_NAMESPACE
+
+template <class Gimpl> class NewWilsonLoops : public Gimpl {
+public:
+  INHERIT_GIMPL_TYPES(Gimpl);
+
+  typedef typename Gimpl::GaugeLinkField GaugeMat;
+  typedef typename Gimpl::GaugeField GaugeLorentz;
+
+  //////////////////////////////////////////////////
+  // directed plaquette oriented in mu,nu plane
+  //////////////////////////////////////////////////
+  static void dirPlaquette(GaugeMat &plaq, const std::vector<GaugeMat> &U,
+                           const int mu, const int nu) {
+    // Annoyingly, must use either scope resolution to find dependent base
+    // class,
+    // or this-> ; there is no "this" in a static method. This forces explicit
+    // Gimpl scope
+    // resolution throughout the usage in this file, and rather defeats the
+    // purpose of deriving
+    // from Gimpl.
+    plaq = Gimpl::CovShiftBackward(
+        U[mu], mu, Gimpl::CovShiftBackward(
+                       U[nu], nu, Gimpl::CovShiftForward(U[mu], mu, U[nu])));
+  }
+  //////////////////////////////////////////////////
+  // trace of directed plaquette oriented in mu,nu plane
+  //////////////////////////////////////////////////
+  static void traceDirPlaquette(LatticeComplex &plaq,
+                                const std::vector<GaugeMat> &U, const int mu,
+                                const int nu) {
+    GaugeMat sp(U[0]._grid);
+    dirPlaquette(sp, U, mu, nu);
+    plaq = trace(sp);
+  }
+  //////////////////////////////////////////////////
+  // sum over all planes of plaquette
+  //////////////////////////////////////////////////
+  static void sitePlaquette(LatticeComplex &Plaq,
+                            const std::vector<GaugeMat> &U) {
+    LatticeComplex sitePlaq(U[0]._grid);
+    Plaq = zero;
+    for (int mu = 1; mu < U[0]._grid->_ndimension; mu++) {
+      for (int nu = 0; nu < mu; nu++) {
+        traceDirPlaquette(sitePlaq, U, mu, nu);
+        Plaq = Plaq + sitePlaq;
+      }
+    }
+  }
+  //////////////////////////////////////////////////
+  // sum over all x,y,z,t and over all planes of plaquette
+  //////////////////////////////////////////////////
+  static Real sumPlaquette(const GaugeLorentz &Umu) {
+    std::vector<GaugeMat> U(4, Umu._grid);
+
+    for (int mu = 0; mu < Umu._grid->_ndimension; mu++) {
+      U[mu] = PeekIndex<LorentzIndex>(Umu, mu);
+    }
+
+    LatticeComplex Plaq(Umu._grid);
+
+    sitePlaquette(Plaq, U);
+
+    TComplex Tp = sum(Plaq);
+    Complex p = TensorRemove(Tp);
+    return p.real();
+  }
+  //////////////////////////////////////////////////
+  // average over all x,y,z,t and over all planes of plaquette
+  //////////////////////////////////////////////////
+  static Real avgPlaquette(const GaugeLorentz &Umu) {
+    int ndim = Umu._grid->_ndimension;
+    Real sumplaq = sumPlaquette(Umu);
+    Real vol = Umu._grid->gSites();
+    Real faces = (1.0 * ndim * (ndim - 1)) / 2.0;
+    return sumplaq / vol / faces / Nc; // Nc dependent... FIXME
+  }
+
+  //////////////////////////////////////////////////
+  // Wilson loop of size (R1, R2), oriented in mu,nu plane
+  //////////////////////////////////////////////////
+  static void wilsonLoop(GaugeMat &wl, const std::vector<GaugeMat> &U,
+                           const int Rmu, const int Rnu,
+                           const int mu, const int nu) {
+    wl = U[nu];
+
+    for(int i = 0; i < Rnu-1; i++){
+      wl = Gimpl::CovShiftForward(U[nu], nu, wl);
+    }
+
+    for(int i = 0; i < Rmu; i++){
+      wl = Gimpl::CovShiftForward(U[mu], mu, wl);
+    }
+
+    for(int i = 0; i < Rnu; i++){
+      wl = Gimpl::CovShiftBackward(U[nu], nu, wl);
+    }
+
+    for(int i = 0; i < Rmu; i++){
+      wl = Gimpl::CovShiftBackward(U[mu], mu, wl);
+    }
+  }
+  //////////////////////////////////////////////////
+  // trace of Wilson Loop oriented in mu,nu plane
+  //////////////////////////////////////////////////
+  static void traceWilsonLoop(LatticeComplex &wl,
+                                const std::vector<GaugeMat> &U,
+                                const int Rmu, const int Rnu,
+                                const int mu, const int nu) {
+    GaugeMat sp(U[0]._grid);
+    wilsonLoop(sp, U, Rmu, Rnu, mu, nu);
+    wl = trace(sp);
+  }
+  //////////////////////////////////////////////////
+  // sum over all planes of Wilson loop
+  //////////////////////////////////////////////////
+  static void siteWilsonLoop(LatticeComplex &Wl,
+                            const std::vector<GaugeMat> &U,
+                            const int R1, const int R2) {
+    LatticeComplex siteWl(U[0]._grid);
+    Wl = zero;
+    for (int mu = 1; mu < U[0]._grid->_ndimension; mu++) {
+      for (int nu = 0; nu < mu; nu++) {
+        traceWilsonLoop(siteWl, U, R1, R2, mu, nu);
+        Wl = Wl + siteWl;
+        traceWilsonLoop(siteWl, U, R2, R1, mu, nu);
+        Wl = Wl + siteWl;
+      }
+    }
+  }
+  //////////////////////////////////////////////////
+  // sum over planes of Wilson loop with length R1
+  // in the time direction
+  //////////////////////////////////////////////////
+  static void siteTimelikeWilsonLoop(LatticeComplex &Wl,
+                            const std::vector<GaugeMat> &U,
+                            const int R1, const int R2) {
+    LatticeComplex siteWl(U[0]._grid);
+
+    int ndim = U[0]._grid->_ndimension;
+
+    Wl = zero;
+    for (int nu = 0; nu < ndim - 1; nu++) {
+      traceWilsonLoop(siteWl, U, R1, R2, ndim-1, nu);
+      Wl = Wl + siteWl;
+    }
+  }
+  //////////////////////////////////////////////////
+  // sum Wilson loop over all planes orthogonal to the time direction
+  //////////////////////////////////////////////////
+  static void siteSpatialWilsonLoop(LatticeComplex &Wl,
+                            const std::vector<GaugeMat> &U,
+                            const int R1, const int R2) {
+    LatticeComplex siteWl(U[0]._grid);
+
+    Wl = zero;
+    for (int mu = 1; mu < U[0]._grid->_ndimension - 1; mu++) {
+      for (int nu = 0; nu < mu; nu++) {
+        traceWilsonLoop(siteWl, U, R1, R2, mu, nu);
+        Wl = Wl + siteWl;
+        traceWilsonLoop(siteWl, U, R2, R1, mu, nu);
+        Wl = Wl + siteWl;
+      }
+    }
+  }
+  //////////////////////////////////////////////////
+  // sum over all x,y,z,t and over all planes of Wilson loop
+  //////////////////////////////////////////////////
+  static Real sumWilsonLoop(const GaugeLorentz &Umu,
+                            const int R1, const int R2) {
+    std::vector<GaugeMat> U(4, Umu._grid);
+
+    for (int mu = 0; mu < Umu._grid->_ndimension; mu++) {
+      U[mu] = PeekIndex<LorentzIndex>(Umu, mu);
+    }
+
+    LatticeComplex Wl(Umu._grid);
+
+    siteWilsonLoop(Wl, U, R1, R2);
+
+    TComplex Tp = sum(Wl);
+    Complex p = TensorRemove(Tp);
+    return p.real();
+  }
+  //////////////////////////////////////////////////
+  // sum over all x,y,z,t and over all planes of timelike Wilson loop
+  //////////////////////////////////////////////////
+  static Real sumTimelikeWilsonLoop(const GaugeLorentz &Umu,
+                            const int R1, const int R2) {
+    std::vector<GaugeMat> U(4, Umu._grid);
+
+    for (int mu = 0; mu < Umu._grid->_ndimension; mu++) {
+      U[mu] = PeekIndex<LorentzIndex>(Umu, mu);
+    }
+
+    LatticeComplex Wl(Umu._grid);
+
+    siteTimelikeWilsonLoop(Wl, U, R1, R2);
+
+    TComplex Tp = sum(Wl);
+    Complex p = TensorRemove(Tp);
+    return p.real();
+  }
+  //////////////////////////////////////////////////
+  // sum over all x,y,z,t and over all planes of spatial Wilson loop
+  //////////////////////////////////////////////////
+  static Real sumSpatialWilsonLoop(const GaugeLorentz &Umu,
+                            const int R1, const int R2) {
+    std::vector<GaugeMat> U(4, Umu._grid);
+
+    for (int mu = 0; mu < Umu._grid->_ndimension; mu++) {
+      U[mu] = PeekIndex<LorentzIndex>(Umu, mu);
+    }
+
+    LatticeComplex Wl(Umu._grid);
+
+    siteSpatialWilsonLoop(Wl, U, R1, R2);
+
+    TComplex Tp = sum(Wl);
+    Complex p = TensorRemove(Tp);
+    return p.real();
+  }
+  //////////////////////////////////////////////////
+  // average over all x,y,z,t and over all planes of Wilson loop
+  //////////////////////////////////////////////////
+  static Real avgWilsonLoop(const GaugeLorentz &Umu,
+                            const int R1, const int R2) {
+    int ndim = Umu._grid->_ndimension;
+    Real sumWl = sumWilsonLoop(Umu, R1, R2);
+    Real vol = Umu._grid->gSites();
+    Real faces = 1.0 * ndim * (ndim - 1);
+    return sumWl / vol / faces / Nc; // Nc dependent... FIXME
+  }
+  //////////////////////////////////////////////////
+  // average over all x,y,z,t and over all planes of timelike Wilson loop
+  //////////////////////////////////////////////////
+  static Real avgTimelikeWilsonLoop(const GaugeLorentz &Umu,
+                            const int R1, const int R2) {
+    int ndim = Umu._grid->_ndimension;
+    Real sumWl = sumTimelikeWilsonLoop(Umu, R1, R2);
+    Real vol = Umu._grid->gSites();
+    Real faces = 1.0 * (ndim - 1);
+    return sumWl / vol / faces / Nc; // Nc dependent... FIXME
+  }
+  //////////////////////////////////////////////////
+  // average over all x,y,z,t and over all planes of spatial Wilson loop
+  //////////////////////////////////////////////////
+  static Real avgSpatialWilsonLoop(const GaugeLorentz &Umu,
+                            const int R1, const int R2) {
+    int ndim = Umu._grid->_ndimension;
+    Real sumWl = sumSpatialWilsonLoop(Umu, R1, R2);
+    Real vol = Umu._grid->gSites();
+    Real faces = 1.0 * (ndim - 1) * (ndim - 2);
+    return sumWl / vol / faces / Nc; // Nc dependent... FIXME
+  }
+};
+
+END_QEDFVOL_NAMESPACE
+
+#endif // QEDFVOL_WILSONLOOPS_H

extras/qed-fvol/qed-fvol.cc

@@ -0,0 +1,88 @@
+#include <Global.hpp>
+#include <WilsonLoops.h>
+
+using namespace Grid;
+using namespace QCD;
+using namespace QedFVol;
+
+typedef PeriodicGaugeImpl<QedGimplR>    QedPeriodicGimplR;
+typedef PhotonR::GaugeField             EmField;
+typedef PhotonR::GaugeLinkField         EmComp;
+
+const int NCONFIGS = 10;
+const int NWILSON = 10;
+
+int main(int argc, char *argv[])
+{
+    // parse command line
+    std::string parameterFileName;
+    
+    if (argc < 2)
+    {
+        std::cerr << "usage: " << argv[0] << " <parameter file> [Grid options]";
+        std::cerr << std::endl;
+        std::exit(EXIT_FAILURE);
+    }
+    parameterFileName = argv[1];
+    
+    // initialization
+    Grid_init(&argc, &argv);
+    QedFVolLogError.Active(GridLogError.isActive());
+    QedFVolLogWarning.Active(GridLogWarning.isActive());
+    QedFVolLogMessage.Active(GridLogMessage.isActive());
+    QedFVolLogIterative.Active(GridLogIterative.isActive());
+    QedFVolLogDebug.Active(GridLogDebug.isActive());
+    LOG(Message) << "Grid initialized" << std::endl;
+    
+    // QED stuff
+    std::vector<int> latt_size   = GridDefaultLatt();
+    std::vector<int> simd_layout = GridDefaultSimd(4, vComplex::Nsimd());
+    std::vector<int> mpi_layout  = GridDefaultMpi();
+    GridCartesian    grid(latt_size,simd_layout,mpi_layout);
+    GridParallelRNG  pRNG(&grid);
+    PhotonR          photon(PhotonR::Gauge::feynman,
+                            PhotonR::ZmScheme::qedL);
+    EmField          a(&grid);
+    EmField          expA(&grid);
+
+    Complex imag_unit(0, 1);
+
+    Real wlA;
+    std::vector<Real> logWlAvg(NWILSON, 0.0), logWlTime(NWILSON, 0.0), logWlSpace(NWILSON, 0.0);
+
+    pRNG.SeedRandomDevice();
+
+    LOG(Message) << "Wilson loop calculation beginning" << std::endl;
+    for(int ic = 0; ic < NCONFIGS; ic++){
+        LOG(Message) << "Configuration " << ic <<std::endl;
+        photon.StochasticField(a, pRNG);
+
+        // Exponentiate photon field
+        expA = exp(imag_unit*a);
+
+        // Calculate Wilson loops
+        for(int iw=1; iw<=NWILSON; iw++){
+            wlA = NewWilsonLoops<QedPeriodicGimplR>::avgWilsonLoop(expA, iw, iw) * 3;
+            logWlAvg[iw-1] -= 2*log(wlA);
+            wlA = NewWilsonLoops<QedPeriodicGimplR>::avgTimelikeWilsonLoop(expA, iw, iw) * 3;
+            logWlTime[iw-1] -= 2*log(wlA);
+            wlA = NewWilsonLoops<QedPeriodicGimplR>::avgSpatialWilsonLoop(expA, iw, iw) * 3;
+            logWlSpace[iw-1] -= 2*log(wlA);
+        }
+    }
+    LOG(Message) << "Wilson loop calculation completed" << std::endl;
+    
+    // Calculate Wilson loops
+    for(int iw=1; iw<=10; iw++){
+        LOG(Message) << iw << 'x' << iw << " Wilson loop" << std::endl;
+        LOG(Message) << "-2log(W) average: " << logWlAvg[iw-1]/NCONFIGS << std::endl;
+        LOG(Message) << "-2log(W) timelike: " << logWlTime[iw-1]/NCONFIGS << std::endl;
+        LOG(Message) << "-2log(W) spatial: " << logWlSpace[iw-1]/NCONFIGS << std::endl;
+    }
+
+    // epilogue
+    LOG(Message) << "Grid is finalizing now" << std::endl;
+    Grid_finalize();
+    
+    return EXIT_SUCCESS;
+}

gcc-bug-report/README

@@ -20,4 +20,17 @@ The simple testcase in this directory is the submitted bug report that encapsula
 problem. The test case works with icpc and with clang++, but fails consistently on g++
 current variants.
 
-Peter
+Peter
+
+
+************
+
+Second GCC bug reported, see Issue 100.
+
+https://wandbox.org/permlink/tzssJza6R9XnqANw
+https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80652
+
+Getting Travis fails under gcc-5 for Test_simd, now that I added more comprehensive testing to the
+CI test suite. The limitations of Travis runtime limits & weak cores are being shown.
+
+Travis uses 5.4.1 for g++-5.

grid-config.in

@@ -0,0 +1,86 @@
+#! /bin/sh
+
+prefix=@prefix@
+exec_prefix=@exec_prefix@
+includedir=@includedir@
+
+usage()
+{
+  cat <<EOF
+Usage: grid-config [OPTION]
+
+Known values for OPTION are:
+
+  --prefix     show Grid installation prefix
+  --cxxflags   print pre-processor and compiler flags
+  --ldflags    print library linking flags
+  --libs       print library linking information
+  --summary    print full build summary
+  --help       display this help and exit
+  --version    output version information
+  --git        print git revision
+
+EOF
+  
+  exit $1
+}
+
+if test $# -eq 0; then
+  usage 1
+fi
+
+cflags=false
+libs=false
+
+while test $# -gt 0; do
+  case "$1" in
+    -*=*) optarg=`echo "$1" | sed 's/[-_a-zA-Z0-9]*=//'` ;;
+    *) optarg= ;;
+  esac
+  
+  case "$1" in
+    --prefix)
+      echo $prefix
+    ;;
+    
+    --version)
+      echo @VERSION@
+      exit 0
+    ;;
+    
+    --git)
+      echo "@GRID_BRANCH@ @GRID_SHA@"
+      exit 0
+    ;;
+    
+    --help)
+      usage 0
+    ;;
+    
+    --cxxflags)
+      echo @GRID_CXXFLAGS@
+    ;;
+    
+    --ldflags)
+      echo @GRID_LDFLAGS@
+    ;;
+    
+    --libs)
+      echo @GRID_LIBS@
+    ;;
+    
+    --summary)
+      echo ""
+      echo "@GRID_SUMMARY@"
+      echo ""
+    ;;
+    
+    *)
+      usage
+      exit 1
+    ;;
+  esac
+  shift
+done
+
+exit 0

lib/AlignedAllocator.cc

@@ -1,65 +0,0 @@
-
-
-
-#include <Grid/Grid.h>
-
-namespace Grid {
-
-int PointerCache::victim;
-
-  PointerCache::PointerCacheEntry PointerCache::Entries[PointerCache::Ncache];
-
-void *PointerCache::Insert(void *ptr,size_t bytes) {
-
-  if (bytes < 4096 ) return NULL;
-
-#ifdef _OPENMP
-  assert(omp_in_parallel()==0);
-#endif 
-  void * ret = NULL;
-  int v = -1;
-
-  for(int e=0;e<Ncache;e++) {
-    if ( Entries[e].valid==0 ) {
-      v=e; 
-      break;
-    }
-  }
-
-  if ( v==-1 ) {
-    v=victim;
-    victim = (victim+1)%Ncache;
-  }
-
-  if ( Entries[v].valid ) {
-    ret = Entries[v].address;
-    Entries[v].valid = 0;
-    Entries[v].address = NULL;
-    Entries[v].bytes = 0;
-  }
-
-  Entries[v].address=ptr;
-  Entries[v].bytes  =bytes;
-  Entries[v].valid  =1;
-
-  return ret;
-}
-
-void *PointerCache::Lookup(size_t bytes) {
-
- if (bytes < 4096 ) return NULL;
-
-#ifdef _OPENMP
-  assert(omp_in_parallel()==0);
-#endif 
-
-  for(int e=0;e<Ncache;e++){
-    if ( Entries[e].valid && ( Entries[e].bytes == bytes ) ) {
-      Entries[e].valid = 0;
-      return Entries[e].address;
-    }
-  }
-  return NULL;
-}
-
-}

lib/DisableWarnings.h

@@ -0,0 +1,37 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/DisableWarnings.h
+
+Copyright (C) 2016
+
+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 */
+
+#ifndef DISABLE_WARNINGS_H
+#define DISABLE_WARNINGS_H
+
+ //disables and intel compiler specific warning (in json.hpp)
+#pragma warning disable 488  
+
+
+#endif