portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2026-08-25 20:09:36 +01:00
Code Releases Activity

Added coalescedReadGeneralPermute CPU implementation based on Christoph's GPT code

Browse Source 
In a test code, implemented a padded-cell version of the staple and rectangular-staple calculation
This commit is contained in:
Christopher Kelly
2023-06-20 16:14:25 -04:00
parent d72e914cf0
commit abc658dca5
3 changed files with 546 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files
Grid/tensors/Tensor_SIMT.h
+10
View File
@@ -73,6 +73,16 @@ vobj coalescedReadPermute(const vobj & __restrict__ vec,int ptype,int doperm,int
return vec;
}
}
//'perm_mask' acts as a bitmask
template<class vobj> accelerator_inline
vobj coalescedReadGeneralPermute(const vobj & __restrict__ vec,int perm_mask,int nd,int lane=0)
{
auto obj = vec, tmp = vec;
for (int d=0;d<nd;d++)
if (perm_mask & (0x1 << d)) { permute(obj,tmp,d); tmp=obj;}
return obj;
}
template<class vobj> accelerator_inline
void coalescedWrite(vobj & __restrict__ vec,const vobj & __restrict__ extracted,int lane=0)
{