diff --git a/TODO b/TODO
index a8fa5ce9..256f2609 100644
--- a/TODO
+++ b/TODO
@@ -10,7 +10,7 @@ FUNCTIONALITY:
 * How to do U[mu] ... lorentz part of type structure or not. more like chroma if not. -- DONE
 
 * subdirs lib, tests ??                              ----- DONE
-  - lib/math
+  - lib/math        
   - lib/cartesian
   - lib/cshift
   - lib/stencil
@@ -26,7 +26,7 @@ FUNCTIONALITY:
 Not done, or just incomplete
 * random number generation
 
-* Consider switch std::vector to boost arrays.
+* Consider switch std::vector to boost arrays or something lighter weight
   boost::multi_array<type, 3> A()...    to replace multi1d, multi2d etc..
 
 * How to define simple matrix operations, such as flavour matrices?
diff --git a/configure b/configure
index 48fafcf1..9e554925 100755
--- a/configure
+++ b/configure
@@ -1365,8 +1365,9 @@ Optional Features:
   --disable-dependency-tracking
                           speeds up one-time build
   --disable-openmp        do not use OpenMP
-  --enable-simd=SSE|AVX|AVX2|AVX512
-                          Select instructions
+  --enable-simd=SSE|AVX|AVX2|AVX512|MIC
+                          Select instructions to be SSE4.0, AVX 1.0, AVX
+                          2.0+FMA, AVX 512, MIC
   --enable-comms=none|mpi Select communications
 
 Some influential environment variables:
@@ -4945,6 +4946,18 @@ fi
 
 done
 
+for ac_header in endian.h
+do :
+  ac_fn_c_check_header_mongrel "$LINENO" "endian.h" "ac_cv_header_endian_h" "$ac_includes_default"
+if test "x$ac_cv_header_endian_h" = xyes; then :
+  cat >>confdefs.h <<_ACEOF
+#define HAVE_ENDIAN_H 1
+_ACEOF
+
+fi
+
+done
+
 
 # Checks for typedefs, structures, and compiler characteristics.
 ac_fn_c_check_type "$LINENO" "size_t" "ac_cv_type_size_t" "$ac_includes_default"
@@ -4999,32 +5012,11 @@ _ACEOF
 fi
 done
 
-for ac_func in ntohll
-do :
-  ac_fn_c_check_func "$LINENO" "ntohll" "ac_cv_func_ntohll"
-if test "x$ac_cv_func_ntohll" = xyes; then :
-  cat >>confdefs.h <<_ACEOF
-#define HAVE_NTOHLL 1
-_ACEOF
-
-fi
-done
-
-for ac_func in be64toh
-do :
-  ac_fn_c_check_func "$LINENO" "be64toh" "ac_cv_func_be64toh"
-if test "x$ac_cv_func_be64toh" = xyes; then :
-  cat >>confdefs.h <<_ACEOF
-#define HAVE_BE64TOH 1
-_ACEOF
-
-fi
-done
-
 
 # Check whether --enable-simd was given.
 if test "${enable_simd+set}" = set; then :
-  enableval=$enable_simd; ac_SIMD=${enable_simd}
+  enableval=$enable_simd; \
+	ac_SIMD=${enable_simd}
 else
   ac_SIMD=AVX2
 fi
@@ -5049,8 +5041,8 @@ $as_echo "#define AVX1 1" >>confdefs.h
 $as_echo "#define AVX2 1" >>confdefs.h
 
      ;;
-     AVX512)
-       echo Configuring for AVX512
+     AVX512|MIC)
+       echo Configuring for AVX512 and MIC
 
 $as_echo "#define AVX512 1" >>confdefs.h
 
diff --git a/lib/Grid_config.h b/lib/Grid_config.h
index 2171ea6b..4152540e 100644
--- a/lib/Grid_config.h
+++ b/lib/Grid_config.h
@@ -16,8 +16,8 @@
 /* GRID_COMMS_NONE */
 /* #undef GRID_COMMS_NONE */
 
-/* Define to 1 if you have the `be64toh' function. */
-/* #undef HAVE_BE64TOH */
+/* Define to 1 if you have the <endian.h> header file. */
+/* #undef HAVE_ENDIAN_H */
 
 /* Define to 1 if you have the `gettimeofday' function. */
 #define HAVE_GETTIMEOFDAY 1
@@ -34,9 +34,6 @@
 /* Define to 1 if you have the <memory.h> header file. */
 #define HAVE_MEMORY_H 1
 
-/* Define to 1 if you have the `ntohll' function. */
-/* #undef HAVE_NTOHLL */
-
 /* Define to 1 if you have the <stdint.h> header file. */
 #define HAVE_STDINT_H 1
 
diff --git a/lib/Grid_config.h.in b/lib/Grid_config.h.in
index 2381525f..2dc0bda4 100644
--- a/lib/Grid_config.h.in
+++ b/lib/Grid_config.h.in
@@ -15,8 +15,8 @@
 /* GRID_COMMS_NONE */
 #undef GRID_COMMS_NONE
 
-/* Define to 1 if you have the `be64toh' function. */
-#undef HAVE_BE64TOH
+/* Define to 1 if you have the <endian.h> header file. */
+#undef HAVE_ENDIAN_H
 
 /* Define to 1 if you have the `gettimeofday' function. */
 #undef HAVE_GETTIMEOFDAY
@@ -33,9 +33,6 @@
 /* Define to 1 if you have the <memory.h> header file. */
 #undef HAVE_MEMORY_H
 
-/* Define to 1 if you have the `ntohll' function. */
-#undef HAVE_NTOHLL
-
 /* Define to 1 if you have the <stdint.h> header file. */
 #undef HAVE_STDINT_H
 
diff --git a/lib/Grid_stencil.h b/lib/Grid_stencil.h
index d8debac0..7bafb879 100644
--- a/lib/Grid_stencil.h
+++ b/lib/Grid_stencil.h
@@ -283,7 +283,7 @@ namespace Grid {
 	    if ( comm_any ) {
 
 	      for(int i=0;i<Nsimd;i++){
-		pointers[i] = (scalar_type *)&send_buf_extract[i][0];
+		pointers[Nsimd-1-i] = (scalar_type *)&send_buf_extract[i][0];
 	      }
 	      Gather_plane_extract(rhs,pointers,dimension,sx,cbmask);
 	      
@@ -332,9 +332,9 @@ namespace Grid {
 	      for(int i=0;i<Nsimd;i++){
 		if ( permute_slice ) {
 		  PermuteMap=i^toggle_bit;
-		  pointers[i] = rpointers[PermuteMap];
+		  pointers[Nsimd-1-i] = rpointers[PermuteMap];
 		} else {
-		  pointers[i] = rpointers[i];
+		  pointers[Nsimd-1-i] = rpointers[i];
 		}
 	      }
 
diff --git a/lib/Grid_summation.h b/lib/Grid_summation.h
index dff2df61..366cef06 100644
--- a/lib/Grid_summation.h
+++ b/lib/Grid_summation.h
@@ -1,12 +1,71 @@
 #ifndef GRID_SUMMATION_H
 #define GRID_SUMMATION_H
 
+void subdivides(GridBase *coarse,GridBase *fine)
+{
+  assert(coarse->_ndimension == fine->_ndimension);
+
+  int _ndimension = coarse->_ndimension;
+
+  // local and global volumes subdivide cleanly after SIMDization
+  for(int d=0;d<_ndimension;d++){
+    assert((fine->_rdimensions[d] / coarse->_rdimensions[d])* coarse->_rdimensions[d]==fine->_rdimensions[d]); 
+    assert((fine->_ldimensions[d] / coarse->_ldimensions[d])* coarse->_ldimensions[d]==fine->_ldimensions[d]); 
+    assert((fine->_gdimensions[d] / coarse->_gdimensions[d])* coarse->_gdimensions[d]==fine->_gdimensions[d]); 
+    assert((fine->_fdimensions[d] / coarse->_fdimensions[d])* coarse->_fdimensions[d]==fine->_fdimensions[d]); 
+  }
+}
+// Generic name : Coarsen?
+//              : SubMeshSum?
+//
 template<class vobj>
-inline void sumBlocks(Lattice<vobj> &coarseData,const Lattice<vobj &fineData)
+inline void sumBlocks(Lattice<vobj> &coarseData,const Lattice<vobj> &fineData)
 {
   GridBase * fine  = findData._grid;
   GridBase * coarse= findData._grid;
+
+  subdivides(coars,fine); // require they map
+
+  int _ndimension = coarse->_ndimension;
   
+  std::vector<bool> replicated(_ndimension,false);
+  std::vector<int>  block_r   (_dimension);
+  std::vector<int>  block_f   (_dimension);
+
+  ///////////////////////////////////////////////////////////
+  // Detect whether the result is replicated in dimension d
+  ///////////////////////////////////////////////////////////
+  for(int d=0 ; d<_ndimension;d++){
+    if ( (_fdimensions[d] == 1) && (coarse->_processors[d]>1) ) {
+      replicated[d]=true;
+    }
+    block_r[d] = fine->_rdimensions[d] / coarse->_rdimensions[d];
+    block_l[d] = fine->_ldimensions[d] / coarse->_ldimensions[d];
+    block_f[d] = fine->_fdimensions[d] / coarse->_fdimensions[d];
+  }
+
+  coaseData=zero;
+
+  //FIXME Bagel's strategy: loop over fine sites
+  // identify corresponding coarse site, but coarse sites are 
+  // divided across threads. Not so easy to do in openmp but
+  // there must be a way
+  for(int sf=0;sf<fine->oSites();sf++){
+    
+    int sc;
+    vobj sum=zero;
+    std::vector<int> coor_c(_ndimension);
+    std::vector<int> coor_f(_ndimension);
+
+    GridBase::CoorFromIndex(coor_f,sf,fine->_rdimensions);
+
+    for(int d=0;d<_ndimension;d++) coor_c[d]=coor_f[d]/fine->_rdimensions;
+
+    GridBase::IndexFromCoor(coor_c,sc,coarse->_rdimensions);
+
+    coarseData._odata[sc]=coarseData._odata[sc]+fineData._odata[sf];
+
+  }
   return;
 }
 #endif