From 36ea5f6b7772a2295db0e4578f069706c393ea47 Mon Sep 17 00:00:00 2001
From: paboyle <paboyle@ph.ed.ac.uk>
Date: Sat, 24 Feb 2018 22:20:14 +0000
Subject: [PATCH] gpu friendly coordinates ; no std::vector on GPU

---
 benchmarks/Benchmark_ITT.cc              | 56 ++++++++++++------------
 benchmarks/Benchmark_comms.cc            | 20 ++++-----
 benchmarks/Benchmark_dwf.cc              |  4 +-
 benchmarks/Benchmark_gparity.cc          |  2 +-
 benchmarks/Benchmark_memory_asynch.cc    |  6 +--
 benchmarks/Benchmark_memory_bandwidth.cc | 12 ++---
 benchmarks/Benchmark_mooee.cc            |  2 +-
 benchmarks/Benchmark_staggered.cc        |  6 +--
 benchmarks/Benchmark_su3.cc              | 12 ++---
 benchmarks/Benchmark_wilson.cc           |  6 +--
 benchmarks/Benchmark_wilson_sweep.cc     |  9 ++--
 11 files changed, 67 insertions(+), 68 deletions(-)

diff --git a/benchmarks/Benchmark_ITT.cc b/benchmarks/Benchmark_ITT.cc
index b689866d..266251c5 100644
--- a/benchmarks/Benchmark_ITT.cc
+++ b/benchmarks/Benchmark_ITT.cc
@@ -109,8 +109,8 @@ public:
     int nmu=0;
     int maxlat=32;
 
-    std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplexD::Nsimd());
-    std::vector<int> mpi_layout  = GridDefaultMpi();
+    Coordinate simd_layout = GridDefaultSimd(Nd,vComplexD::Nsimd());
+    Coordinate mpi_layout  = GridDefaultMpi();
 
     for(int mu=0;mu<Nd;mu++) if (mpi_layout[mu]>1) nmu++;
 
@@ -125,7 +125,7 @@ public:
     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],
+	Coordinate latt_size  ({lat*mpi_layout[0],
 	      lat*mpi_layout[1],
 	      lat*mpi_layout[2],
 	      lat*mpi_layout[3]});
@@ -217,8 +217,8 @@ public:
     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();
+    Coordinate simd_layout = GridDefaultSimd(Nd,vReal::Nsimd());
+    Coordinate mpi_layout  = GridDefaultMpi();
 
     std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
     std::cout<<GridLogMessage << "= Benchmarking a*x + y bandwidth"<<std::endl;
@@ -236,7 +236,7 @@ public:
     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]});
+      Coordinate 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);
 
@@ -284,25 +284,25 @@ public:
     // 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});
+    Coordinate mpi = GridDefaultMpi(); assert(mpi.size()==4);
+    Coordinate local({L,L,L,L});
 
-    GridCartesian         * TmpGrid   = SpaceTimeGrid::makeFourDimGrid(std::vector<int>({64,64,64,64}), 
+    GridCartesian         * TmpGrid   = SpaceTimeGrid::makeFourDimGrid(Coordinate({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});
+    Coordinate internal;
+    if      ( SHM == 1 )   internal = Coordinate({1,1,1,1});
+    else if ( SHM == 2 )   internal = Coordinate({2,1,1,1});
+    else if ( SHM == 4 )   internal = Coordinate({2,2,1,1});
+    else if ( SHM == 8 )   internal = Coordinate({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]});
+    Coordinate nodes({mpi[0]/internal[0],mpi[1]/internal[1],mpi[2]/internal[2],mpi[3]/internal[3]});
+    Coordinate latt4({local[0]*nodes[0],local[1]*nodes[1],local[2]*nodes[2],local[3]*nodes[3]});
 
     ///////// Welcome message ////////////
     std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
@@ -466,25 +466,25 @@ public:
     // 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});
+    Coordinate mpi = GridDefaultMpi(); assert(mpi.size()==4);
+    Coordinate local({L,L,L,L});
 
-    GridCartesian         * TmpGrid   = SpaceTimeGrid::makeFourDimGrid(std::vector<int>({64,64,64,64}), 
+    GridCartesian         * TmpGrid   = SpaceTimeGrid::makeFourDimGrid(Coordinate({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});
+    Coordinate internal;
+    if      ( SHM == 1 )   internal = Coordinate({1,1,1,1});
+    else if ( SHM == 2 )   internal = Coordinate({2,1,1,1});
+    else if ( SHM == 4 )   internal = Coordinate({2,2,1,1});
+    else if ( SHM == 8 )   internal = Coordinate({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]});
+    Coordinate nodes({mpi[0]/internal[0],mpi[1]/internal[1],mpi[2]/internal[2],mpi[3]/internal[3]});
+    Coordinate latt4({local[0]*nodes[0],local[1]*nodes[1],local[2]*nodes[2],local[3]*nodes[3]});
 
     ///////// Welcome message ////////////
     std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
@@ -701,10 +701,10 @@ int main (int argc, char ** argv)
   }
 #if 1
   int sel=2;
-  std::vector<int> L_list({8,12,16,24});
+  Coordinate L_list({8,12,16,24});
 #else
   int sel=1;
-  std::vector<int> L_list({8,12});
+  Coordinate L_list({8,12});
 #endif
   int selm1=sel-1;
   std::vector<double> robust_list;
diff --git a/benchmarks/Benchmark_comms.cc b/benchmarks/Benchmark_comms.cc
index 8ece8580..0f95b115 100644
--- a/benchmarks/Benchmark_comms.cc
+++ b/benchmarks/Benchmark_comms.cc
@@ -61,8 +61,8 @@ int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplexD::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplexD::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
   int threads = GridThread::GetThreads();
   std::cout<<GridLogMessage << "Grid is setup to use "<<threads<<" threads"<<std::endl;
 
@@ -82,7 +82,7 @@ int main (int argc, char ** argv)
   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],
+      Coordinate latt_size  ({lat*mpi_layout[0],
       				    lat*mpi_layout[1],
       				    lat*mpi_layout[2],
       				    lat*mpi_layout[3]});
@@ -169,7 +169,7 @@ int main (int argc, char ** argv)
   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});
+      Coordinate latt_size  ({lat,lat,lat,lat});
 
       GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
       RealD Nrank = Grid._Nprocessors;
@@ -260,7 +260,7 @@ int main (int argc, char ** argv)
   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],
+      Coordinate latt_size  ({lat*mpi_layout[0],
       				    lat*mpi_layout[1],
       				    lat*mpi_layout[2],
       				    lat*mpi_layout[3]});
@@ -357,7 +357,7 @@ int main (int argc, char ** argv)
   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],
+      Coordinate latt_size  ({lat*mpi_layout[0],
       				    lat*mpi_layout[1],
       				    lat*mpi_layout[2],
       				    lat*mpi_layout[3]});
@@ -455,10 +455,10 @@ int main (int argc, char ** argv)
   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]});
+      Coordinate 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;
diff --git a/benchmarks/Benchmark_dwf.cc b/benchmarks/Benchmark_dwf.cc
index 6dc3799b..8972a496 100644
--- a/benchmarks/Benchmark_dwf.cc
+++ b/benchmarks/Benchmark_dwf.cc
@@ -50,7 +50,7 @@ int main (int argc, char ** argv)
   int threads = GridThread::GetThreads();
   std::cout<<GridLogMessage << "Grid is setup to use "<<threads<<" threads"<<std::endl;
 
-  std::vector<int> latt4 = GridDefaultLatt();
+  Coordinate latt4 = GridDefaultLatt();
   int Ls=16;
   for(int i=0;i<argc;i++)
     if(std::string(argv[i]) == "-Ls"){
@@ -82,7 +82,7 @@ int main (int argc, char ** argv)
 #if 0
   src = Zero();
   {
-    std::vector<int> origin({0,0,0,latt4[2]-1,0});
+    Coordinate origin({0,0,0,latt4[2]-1,0});
     SpinColourVectorF tmp;
     tmp=Zero();
     tmp()(0)(0)=Complex(-2.0,0.0);
diff --git a/benchmarks/Benchmark_gparity.cc b/benchmarks/Benchmark_gparity.cc
index 640c6081..cb89ec25 100644
--- a/benchmarks/Benchmark_gparity.cc
+++ b/benchmarks/Benchmark_gparity.cc
@@ -36,7 +36,7 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "Grid is setup to use "<<threads<<" threads"<<std::endl;
   std::cout<<GridLogMessage << "Ls = " << Ls << std::endl;
 
-  std::vector<int> latt4 = GridDefaultLatt();
+  Coordinate latt4 = GridDefaultLatt();
 
   GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplexF::Nsimd()),GridDefaultMpi());
   GridRedBlackCartesian * UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
diff --git a/benchmarks/Benchmark_memory_asynch.cc b/benchmarks/Benchmark_memory_asynch.cc
index e833aab5..c13eec16 100644
--- a/benchmarks/Benchmark_memory_asynch.cc
+++ b/benchmarks/Benchmark_memory_asynch.cc
@@ -41,8 +41,8 @@ int main (int argc, char ** argv)
   typedef iVector<vReal,Nvec> Vec;
 
 
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vReal::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vReal::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
 
   int threads = GridThread::GetThreads();
   std::cout<<GridLogMessage << "Grid is setup to use "<<threads<<" threads"<<std::endl;
@@ -58,7 +58,7 @@ int main (int argc, char ** argv)
 
     int Nloop=lmax*4/lat;
 
-    std::vector<int> latt_size  ({2*mpi_layout[0],2*mpi_layout[1],4*mpi_layout[2],lat*mpi_layout[3]});
+    Coordinate latt_size  ({2*mpi_layout[0],2*mpi_layout[1],4*mpi_layout[2],lat*mpi_layout[3]});
 
     GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
 
diff --git a/benchmarks/Benchmark_memory_bandwidth.cc b/benchmarks/Benchmark_memory_bandwidth.cc
index 12b4b4e3..7312ed67 100644
--- a/benchmarks/Benchmark_memory_bandwidth.cc
+++ b/benchmarks/Benchmark_memory_bandwidth.cc
@@ -43,8 +43,8 @@ int main (int argc, char ** argv)
 
   Vec rn = Zero();
 
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vReal::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vReal::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
 
   int threads = GridThread::GetThreads();
   std::cout<<GridLogMessage << "Grid is setup to use "<<threads<<" threads"<<std::endl;
@@ -59,7 +59,7 @@ int main (int argc, char ** argv)
 #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]});
+      Coordinate 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);
 
@@ -96,7 +96,7 @@ int main (int argc, char ** argv)
   
   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]});
+      Coordinate 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);
 
@@ -132,7 +132,7 @@ int main (int argc, char ** argv)
   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]});
+      Coordinate 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];
       uint64_t Nloop=NLOOP;
 
@@ -168,7 +168,7 @@ int main (int argc, char ** argv)
 
   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]});
+      Coordinate 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];
       uint64_t Nloop=NLOOP;
       GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
diff --git a/benchmarks/Benchmark_mooee.cc b/benchmarks/Benchmark_mooee.cc
index 788dc7d5..dfaaae30 100644
--- a/benchmarks/Benchmark_mooee.cc
+++ b/benchmarks/Benchmark_mooee.cc
@@ -40,7 +40,7 @@ int main (int argc, char ** argv)
   int threads = GridThread::GetThreads();
   std::cout<<GridLogMessage << "Grid is setup to use "<<threads<<" threads"<<std::endl;
 
-  std::vector<int> latt4 = GridDefaultLatt();
+  Coordinate latt4 = GridDefaultLatt();
   const int Ls=16;
   GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplex::Nsimd()),GridDefaultMpi());
   GridRedBlackCartesian * UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
diff --git a/benchmarks/Benchmark_staggered.cc b/benchmarks/Benchmark_staggered.cc
index ba99ce38..93086927 100644
--- a/benchmarks/Benchmark_staggered.cc
+++ b/benchmarks/Benchmark_staggered.cc
@@ -36,9 +36,9 @@ 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();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
   GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
   GridRedBlackCartesian     RBGrid(&Grid);
 
diff --git a/benchmarks/Benchmark_su3.cc b/benchmarks/Benchmark_su3.cc
index 49e34136..4d7c51c3 100644
--- a/benchmarks/Benchmark_su3.cc
+++ b/benchmarks/Benchmark_su3.cc
@@ -39,8 +39,8 @@ int main (int argc, char ** argv)
 
   int64_t Nloop=20;
 
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
 
   int64_t threads = GridThread::GetThreads();
   std::cout<<GridLogMessage << "Grid is setup to use "<<threads<<" threads"<<std::endl;
@@ -53,7 +53,7 @@ int main (int argc, char ** argv)
 
   for(int lat=2;lat<=LMAX;lat+=2){
 
-      std::vector<int> latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
+      Coordinate 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}));
@@ -85,7 +85,7 @@ int main (int argc, char ** argv)
 
   for(int lat=2;lat<=LMAX;lat+=2){
 
-      std::vector<int> latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
+      Coordinate 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);
@@ -116,7 +116,7 @@ int main (int argc, char ** argv)
 
   for(int lat=2;lat<=LMAX;lat+=2){
 
-      std::vector<int> latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
+      Coordinate 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);
@@ -147,7 +147,7 @@ int main (int argc, char ** argv)
 
   for(int lat=2;lat<=LMAX;lat+=2){
 
-      std::vector<int> latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
+      Coordinate 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);
diff --git a/benchmarks/Benchmark_wilson.cc b/benchmarks/Benchmark_wilson.cc
index fe8728fa..ad550bb8 100644
--- a/benchmarks/Benchmark_wilson.cc
+++ b/benchmarks/Benchmark_wilson.cc
@@ -54,9 +54,9 @@ int main (int argc, char ** argv)
     overlapComms = true;
   }
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
   GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
   GridRedBlackCartesian     RBGrid(&Grid);
 
diff --git a/benchmarks/Benchmark_wilson_sweep.cc b/benchmarks/Benchmark_wilson_sweep.cc
index ea773f7b..ce2780f9 100644
--- a/benchmarks/Benchmark_wilson_sweep.cc
+++ b/benchmarks/Benchmark_wilson_sweep.cc
@@ -53,8 +53,8 @@ int main (int argc, char ** argv)
   typename WilsonFermionR::ImplParams params;
   params.overlapCommsCompute = overlapComms;
 
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
   std::vector<int> seeds({1,2,3,4});
   RealD mass = 0.1;
 
@@ -83,14 +83,13 @@ int main (int argc, char ** argv)
   if ( getenv("DMIN") ) dmin=atoi(getenv("DMIN"));
   for (int L=8; L<=Lmax; L*=2)
     {
-      std::vector<int> latt_size = std::vector<int>(4,L);
+      Coordinate latt_size = Coordinate(4,L);
       for(int d=4; d>dmin; d--)
 	{
 	  if ( d<=3 ) { latt_size[d] *= 2; }
 
 	  std::cout << GridLogMessage;
-	  std::copy( latt_size.begin(), --latt_size.end(), std::ostream_iterator<int>( std::cout, std::string("x").c_str() ) );
-	  std::cout << latt_size.back() << "\t\t";
+	  std::cout << latt_size;
 
 	  GridCartesian           Grid(latt_size,simd_layout,mpi_layout);
 	  GridRedBlackCartesian RBGrid(&Grid);