portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2024-11-09 23:45:36 +00:00
Code Releases Activity

Populate the Cshift_table in the GPU

Browse Source 
Cshift is allocated in Unified memory and used
in the LambdaApply kernels but also populated
from the host. This creates a lot of Unified HtoD
and DtoH mem operations and has a negative effect
in performance. With this commit we populate the
Cshift table in the device with the
populate_Cshift_table() kernel.
This commit is contained in:
Makis Kappas 2023-01-11 21:26:25 +00:00
parent 4ca1bf7cca
commit 7d62f1d6d2
1 changed files with 40 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

40
Grid/cshift/Cshift_common.h
View File

@ -297,6 +297,30 @@ template<class vobj> void Scatter_plane_merge(Lattice<vobj> &rhs,ExtractPointerA
} }
} }
#if (defined(GRID_CUDA) || defined(GRID_HIP)) && defined(ACCELERATOR_CSHIFT)
template <typename T>
T iDivUp(T a, T b) // Round a / b to nearest higher integer value
{ return (a % b != 0) ? (a / b + 1) : (a / b); }
template <typename T>
__global__ void populate_Cshift_table(T* vector, T lo, T ro, T e1, T e2, T stride)
{
int idx = blockIdx.x*blockDim.x + threadIdx.x;
if (idx >= e1*e2) return;
int n, b, o;
n = idx / e2;
b = idx % e2;
o = n*stride + b;
vector[2*idx + 0] = lo + o;
vector[2*idx + 1] = ro + o;
}
#endif
////////////////////////////////////////////////////// //////////////////////////////////////////////////////
// local to node block strided copies // local to node block strided copies
////////////////////////////////////////////////////// //////////////////////////////////////////////////////
@ -321,12 +345,20 @@ template<class vobj> void Copy_plane(Lattice<vobj>& lhs,const Lattice<vobj> &rhs
int ent=0; int ent=0;
if(cbmask == 0x3 ){ if(cbmask == 0x3 ){
#if (defined(GRID_CUDA) || defined(GRID_HIP)) && defined(ACCELERATOR_CSHIFT)
ent = e1*e2;
dim3 blockSize(acceleratorThreads());
dim3 gridSize(iDivUp((unsigned int)ent, blockSize.x));
populate_Cshift_table<<<gridSize, blockSize>>>(&Cshift_table[0].first, lo, ro, e1, e2, stride);
accelerator_barrier();
#else
for(int n=0;n<e1;n++){ for(int n=0;n<e1;n++){
for(int b=0;b<e2;b++){ for(int b=0;b<e2;b++){
int o =n*stride+b; int o =n*stride+b;
Cshift_table[ent++] = std::pair<int,int>(lo+o,ro+o); Cshift_table[ent++] = std::pair<int,int>(lo+o,ro+o);
} }
} }
#endif
} else { } else {
for(int n=0;n<e1;n++){ for(int n=0;n<e1;n++){
for(int b=0;b<e2;b++){ for(int b=0;b<e2;b++){
@ -377,11 +409,19 @@ template<class vobj> void Copy_plane_permute(Lattice<vobj>& lhs,const Lattice<vo
int ent=0; int ent=0;
if ( cbmask == 0x3 ) { if ( cbmask == 0x3 ) {
#if (defined(GRID_CUDA) || defined(GRID_HIP)) && defined(ACCELERATOR_CSHIFT)
ent = e1*e2;
dim3 blockSize(acceleratorThreads());
dim3 gridSize(iDivUp((unsigned int)ent, blockSize.x));
populate_Cshift_table<<<gridSize, blockSize>>>(&Cshift_table[0].first, lo, ro, e1, e2, stride);
accelerator_barrier();
#else
for(int n=0;n<e1;n++){ for(int n=0;n<e1;n++){
for(int b=0;b<e2;b++){ for(int b=0;b<e2;b++){
int o =n*stride; int o =n*stride;
Cshift_table[ent++] = std::pair<int,int>(lo+o+b,ro+o+b); Cshift_table[ent++] = std::pair<int,int>(lo+o+b,ro+o+b);
}} }}
#endif
} else { } else {
for(int n=0;n<e1;n++){ for(int n=0;n<e1;n++){
for(int b=0;b<e2;b++){ for(int b=0;b<e2;b++){