From c79ea0dcef2c73802fa7902dbc92e7e5da8ea03d Mon Sep 17 00:00:00 2001
From: paboyle <paboyle@ph.ed.ac.uk>
Date: Fri, 22 Apr 2016 21:52:54 -0700
Subject: [PATCH] Fixingn IMCI

---
 configure.ac                               | 11 +----
 lib/algorithms/approx/Remez.h              |  6 ++-
 lib/qcd/action/fermion/WilsonKernelsAsm.cc |  2 +-
 lib/simd/Grid_imci.h                       | 52 +++++++++++++++++++++-
 4 files changed, 58 insertions(+), 13 deletions(-)

diff --git a/configure.ac b/configure.ac
index 4da6e3d9..88d85c99 100644
--- a/configure.ac
+++ b/configure.ac
@@ -55,15 +55,6 @@ echo :::::::::::::::::::::::::::::::::::::::::::
 
 AC_CHECK_FUNCS([gettimeofday])
 
-#AC_CHECK_LIB([gmp],[__gmpf_init],,
-#        [AC_MSG_ERROR(GNU Multiple Precision GMP library was not found in your system.
-#Please install or provide the correct path to your installation
-#Info at: http://www.gmplib.org)])
-
-#AC_CHECK_LIB([mpfr],[mpfr_init],,
-#        [AC_MSG_ERROR(GNU Multiple Precision MPFR library was not found in your system.
-#Please install or provide the correct path to your installation
-#Info at: http://www.mpfr.org/)])
 
 #
 # SIMD instructions selection
@@ -124,7 +115,7 @@ case ${ac_SIMD} in
        echo Configuring for IMCI
        AC_DEFINE([IMCI],[1],[IMCI Intrinsics for Knights Corner] )
        supported="cross compilation"
-       ac_ZMM=yes;
+       ac_ZMM=no;
      ;;
      NEONv8)
        echo Configuring for experimental ARMv8a support 
diff --git a/lib/algorithms/approx/Remez.h b/lib/algorithms/approx/Remez.h
index 6e3cf05b..4a56d5d2 100644
--- a/lib/algorithms/approx/Remez.h
+++ b/lib/algorithms/approx/Remez.h
@@ -16,9 +16,13 @@
 #define INCLUDED_ALG_REMEZ_H
 
 #include <stddef.h>
+#include <Config.h>
 
-//#include <algorithms/approx/bigfloat.h>
+#ifdef HAVE_GMP_H
+#include <algorithms/approx/bigfloat.h>
+#else
 #include <algorithms/approx/bigfloat_double.h>
+#endif
 
 #define JMAX 10000 //Maximum number of iterations of Newton's approximation
 #define SUM_MAX 10 // Maximum number of terms in exponential
diff --git a/lib/qcd/action/fermion/WilsonKernelsAsm.cc b/lib/qcd/action/fermion/WilsonKernelsAsm.cc
index 859d1a20..980a2b17 100644
--- a/lib/qcd/action/fermion/WilsonKernelsAsm.cc
+++ b/lib/qcd/action/fermion/WilsonKernelsAsm.cc
@@ -27,7 +27,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     *************************************************************************************/
     /*  END LEGAL */
 #include <Grid.h>
-#if defined(AVX512) || defined (IMCI)
+#if defined(AVX512) 
 //#if defined (IMCI)
 
 #include <simd/Intel512wilson.h>
diff --git a/lib/simd/Grid_imci.h b/lib/simd/Grid_imci.h
index 044c5cb4..119a31a3 100644
--- a/lib/simd/Grid_imci.h
+++ b/lib/simd/Grid_imci.h
@@ -36,7 +36,9 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 //----------------------------------------------------------------------
 
 #include <immintrin.h>
+#include <zmmintrin.h>
 
+namespace Grid{
 namespace Optimization {
   
   struct Vsplat{
@@ -316,6 +318,54 @@ namespace Optimization {
 
   };
  
+  struct Rotate{
+
+    static inline __m512 rotate(__m512 in,int n){ 
+      switch(n){
+      case 0: return tRotate<0>(in);break;
+      case 1: return tRotate<1>(in);break;
+      case 2: return tRotate<2>(in);break;
+      case 3: return tRotate<3>(in);break;
+      case 4: return tRotate<4>(in);break;
+      case 5: return tRotate<5>(in);break;
+      case 6: return tRotate<6>(in);break;
+      case 7: return tRotate<7>(in);break;
+
+      case 8 : return tRotate<8>(in);break;
+      case 9 : return tRotate<9>(in);break;
+      case 10: return tRotate<10>(in);break;
+      case 11: return tRotate<11>(in);break;
+      case 12: return tRotate<12>(in);break;
+      case 13: return tRotate<13>(in);break;
+      case 14: return tRotate<14>(in);break;
+      case 15: return tRotate<15>(in);break;
+      default: assert(0);
+      }
+    }
+    static inline __m512d rotate(__m512d in,int n){ 
+      switch(n){
+      case 0: return tRotate<0>(in);break;
+      case 1: return tRotate<1>(in);break;
+      case 2: return tRotate<2>(in);break;
+      case 3: return tRotate<3>(in);break;
+      case 4: return tRotate<4>(in);break;
+      case 5: return tRotate<5>(in);break;
+      case 6: return tRotate<6>(in);break;
+      case 7: return tRotate<7>(in);break;
+      default: assert(0);
+      }
+    }
+
+    template<int n> static inline __m512 tRotate(__m512 in){ 
+      return (__m512)_mm512_alignr_epi32((__m512i)in,(__m512i)in,n);          
+    };
+
+    template<int n> static inline __m512d tRotate(__m512d in){ 
+      return (__m512d)_mm512_alignr_epi32((__m512i)in,(__m512i)in,2*n);          
+    };
+
+  };
+
 
 
   //////////////////////////////////////////////
@@ -358,7 +408,7 @@ namespace Optimization {
 
 //////////////////////////////////////////////////////////////////////////////////////
 // Here assign types 
-namespace Grid {
+
   typedef __m512 SIMD_Ftype;  // Single precision type
   typedef __m512d SIMD_Dtype; // Double precision type
   typedef __m512i SIMD_Itype; // Integer type