portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2025-06-19 08:17:05 +01:00
Code Releases Activity

Imported coalescedReadGeneralPermute GPU implementation from Christoph

Browse Source 
Fixed bug in padded staple code where extract was being called on the result before the GPU view was closed
Fixed compile issue with pointer cast in padded staple code
Added timing summaries of padded staple code and timing breakdown of staple implementation to Test_padded_cell_staple
This commit is contained in:
Christopher Kelly
2023-06-21 16:01:01 -04:00
parent 7b11075102
commit 4241c7d4a3
2 changed files with 185 additions and 131 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
Grid/tensors/Tensor_SIMT.h
View File

@ -93,7 +93,7 @@ void coalescedWriteNonTemporal(vobj & __restrict__ vec,const vobj & __restrict__
{
vstream(vec, extracted);
}
#else
#else //==GRID_SIMT
//#ifndef GRID_SYCL
@ -176,6 +176,14 @@ typename vobj::scalar_object coalescedReadPermute(const vobj & __restrict__ vec,
return extractLane(plane,vec);
}
template<class vobj> accelerator_inline
typename vobj::scalar_object coalescedReadGeneralPermute(const vobj & __restrict__ vec,int perm_mask,int nd,int lane=acceleratorSIMTlane(vobj::Nsimd()))
{
int plane = lane;
for (int d=0;d<nd;d++)
plane = (perm_mask & (0x1 << d)) ? plane ^ (vobj::Nsimd() >> (d + 1)) : plane;
return extractLane(plane,vec);
}
template<class vobj> accelerator_inline
void coalescedWrite(vobj & __restrict__ vec,const typename vobj::scalar_object & __restrict__ extracted,int lane=acceleratorSIMTlane(vobj::Nsimd()))
{
insertLane(lane,vec,extracted);