From a3c0fb79b6fab365aca3cce8c7d736b804dbca55 Mon Sep 17 00:00:00 2001
From: Christopher Kelly <ckelly@phys.columbia.edu>
Date: Mon, 11 Jul 2016 17:15:22 -0400
Subject: [PATCH 1/2] Fix to iVector and iMatrix pokeIndex and checkerboard
 local site indexing.

---
 lib/cartesian/Cartesian_base.h      |  2 +-
 lib/cartesian/Cartesian_red_black.h | 15 +++++++++++++--
 lib/tensors/Tensor_index.h          |  8 ++++----
 3 files changed, 18 insertions(+), 7 deletions(-)

diff --git a/lib/cartesian/Cartesian_base.h b/lib/cartesian/Cartesian_base.h
index 8272ac71..6fac30e9 100644
--- a/lib/cartesian/Cartesian_base.h
+++ b/lib/cartesian/Cartesian_base.h
@@ -123,7 +123,7 @@ public:
     //////////////////////////////////////////////////////////
     // SIMD lane addressing
     //////////////////////////////////////////////////////////
-    inline int iIndex(std::vector<int> &lcoor)
+    virtual int iIndex(std::vector<int> &lcoor)
     {
         int idx=0;
         for(int d=0;d<_ndimension;d++) idx+=_istride[d]*(lcoor[d]/_rdimensions[d]);
diff --git a/lib/cartesian/Cartesian_red_black.h b/lib/cartesian/Cartesian_red_black.h
index 2424d8dc..d6b6afe6 100644
--- a/lib/cartesian/Cartesian_red_black.h
+++ b/lib/cartesian/Cartesian_red_black.h
@@ -224,9 +224,20 @@ protected:
 	  idx+=_ostride[d]*(coor[d]%_rdimensions[d]);
 	}
       }
-        return idx;
+      return idx;
     };
-        
+    virtual int iIndex(std::vector<int> &lcoor)
+    {
+      int idx=0;
+      for(int d=0;d<_ndimension;d++) {
+	if( d==_checker_dim ) {
+	  idx+=_istride[d]*(lcoor[d]/(2*_rdimensions[d]));
+	} else { 
+	  idx+=_istride[d]*(lcoor[d]/_rdimensions[d]);
+	}
+      }
+      return idx;
+    }
 };
 
 }
diff --git a/lib/tensors/Tensor_index.h b/lib/tensors/Tensor_index.h
index bad9b1f1..7f34f3ac 100644
--- a/lib/tensors/Tensor_index.h
+++ b/lib/tensors/Tensor_index.h
@@ -146,14 +146,14 @@ class TensorIndexRecursion {
     }
 
   template<class vtype,int N> inline static 
-    void pokeIndex(iVector<vtype,N> &ret, const iVector<decltype(TensorIndexRecursion<Level-1>::peekIndex(ret._internal,0)),N> &arg, int i)
+    void pokeIndex(iVector<vtype,N> &ret, const iVector<decltype(TensorIndexRecursion<Level-1>::peekIndex(ret._internal[0],0)),N> &arg, int i)
     {
       for(int ii=0;ii<N;ii++){
 	TensorIndexRecursion<Level-1>::pokeIndex(ret._internal[ii],arg._internal[ii],i);
       }
     }
   template<class vtype,int N> inline static 
-    void pokeIndex(iVector<vtype,N> &ret, const iVector<decltype(TensorIndexRecursion<Level-1>::peekIndex(ret._internal,0)),N> &arg, int i,int j)
+    void pokeIndex(iVector<vtype,N> &ret, const iVector<decltype(TensorIndexRecursion<Level-1>::peekIndex(ret._internal[0],0)),N> &arg, int i,int j)
     {
       for(int ii=0;ii<N;ii++){
 	TensorIndexRecursion<Level-1>::pokeIndex(ret._internal[ii],arg._internal[ii],i,j);
@@ -161,7 +161,7 @@ class TensorIndexRecursion {
     }
 
   template<class vtype,int N> inline static 
-    void pokeIndex(iMatrix<vtype,N> &ret, const iMatrix<decltype(TensorIndexRecursion<Level-1>::peekIndex(ret._internal,0)),N> &arg, int i)
+    void pokeIndex(iMatrix<vtype,N> &ret, const iMatrix<decltype(TensorIndexRecursion<Level-1>::peekIndex(ret._internal[0][0],0)),N> &arg, int i)
     {
       for(int ii=0;ii<N;ii++){
       for(int jj=0;jj<N;jj++){
@@ -169,7 +169,7 @@ class TensorIndexRecursion {
       }}
     }
   template<class vtype,int N> inline static 
-    void pokeIndex(iMatrix<vtype,N> &ret, const iMatrix<decltype(TensorIndexRecursion<Level-1>::peekIndex(ret._internal,0)),N> &arg, int i,int j)
+    void pokeIndex(iMatrix<vtype,N> &ret, const iMatrix<decltype(TensorIndexRecursion<Level-1>::peekIndex(ret._internal[0][0],0)),N> &arg, int i,int j)
     {
       for(int ii=0;ii<N;ii++){
       for(int jj=0;jj<N;jj++){

From 6f47fbb1e2ec67650580baef917cab1efb57fb6c Mon Sep 17 00:00:00 2001
From: Christopher Kelly <ckelly@phys.columbia.edu>
Date: Wed, 13 Jul 2016 10:49:18 -0400
Subject: [PATCH 2/2] Disabled parallel for loops in ExtractSlice and
 InsertSlice due to race conditions. Likely will need to do so for
 localConvert too.

---
 lib/lattice/Lattice_transfer.h | 14 +++++++-------
 1 file changed, 7 insertions(+), 7 deletions(-)

diff --git a/lib/lattice/Lattice_transfer.h b/lib/lattice/Lattice_transfer.h
index bd53cde9..2fa72014 100644
--- a/lib/lattice/Lattice_transfer.h
+++ b/lib/lattice/Lattice_transfer.h
@@ -386,7 +386,7 @@ void InsertSlice(Lattice<vobj> &lowDim,Lattice<vobj> & higherDim,int slice, int
   }
 
   // the above should guarantee that the operations are local
-PARALLEL_FOR_LOOP
+  //PARALLEL_FOR_LOOP
   for(int idx=0;idx<lg->lSites();idx++){
     std::vector<int> lcoor(nl);
     std::vector<int> hcoor(nh);
@@ -420,15 +420,15 @@ void ExtractSlice(Lattice<vobj> &lowDim, Lattice<vobj> & higherDim,int slice, in
   assert(hg->_processors[orthog]==1);
 
   int dl; dl = 0;
-  for(int d=0;d<nh;d++){
-    if ( d != orthog) {
-      assert(lg->_processors[dl]  == hg->_processors[d]);
-      assert(lg->_ldimensions[dl] == hg->_ldimensions[d]);
-      dl++;
+    for(int d=0;d<nh;d++){
+      if ( d != orthog) {
+	assert(lg->_processors[dl]  == hg->_processors[d]);
+	assert(lg->_ldimensions[dl] == hg->_ldimensions[d]);
+	dl++;
     }
   }
   // the above should guarantee that the operations are local
-PARALLEL_FOR_LOOP
+  //PARALLEL_FOR_LOOP
   for(int idx=0;idx<lg->lSites();idx++){
     std::vector<int> lcoor(nl);
     std::vector<int> hcoor(nh);