portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2025-04-09 21:50:45 +01:00
Code Releases Activity

Minor GPU vector width changeÂ

Browse Source
This commit is contained in:
Peter Boyle 2018-07-05 07:02:04 -04:00
parent 4b04ae3611
commit 1bb456c0c5
1 changed files with 1 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
lib/simd/Grid_gpu_vec.h
View File

@ -36,6 +36,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
namespace Grid { namespace Grid {
#define COALESCE_GRANULARITY (32) // bytes for coalesce granularity of target: Pascal, Volta
template<class pair> template<class pair>
class GpuComplex { class GpuComplex {
@ -118,8 +119,6 @@ inline accelerator GpuVector<N,datum> operator/(const GpuVector<N,datum> l,const
return ret; return ret;
} }
#define COALESCE_GRANULARITY (64) // 512 bit coalesce size target
constexpr int NSIMD_RealH = COALESCE_GRANULARITY / sizeof(half); constexpr int NSIMD_RealH = COALESCE_GRANULARITY / sizeof(half);
constexpr int NSIMD_ComplexH = COALESCE_GRANULARITY / sizeof(half2); constexpr int NSIMD_ComplexH = COALESCE_GRANULARITY / sizeof(half2);
constexpr int NSIMD_RealF = COALESCE_GRANULARITY / sizeof(float); constexpr int NSIMD_RealF = COALESCE_GRANULARITY / sizeof(float);