portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2025-04-25 21:25:56 +01:00
Code Releases Activity

Fix gpu MultRealPart and MaddRealPart bug

Browse Source
This commit is contained in:
gfilaci 2019-03-29 16:43:31 +00:00
parent 6da9aa9971
commit 8b6541fb60
1 changed files with 12 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

24
Grid/simd/Grid_gpu.h
View File

@ -245,18 +245,18 @@ namespace Optimization {
struct MultRealPart{ struct MultRealPart{
accelerator_inline float4 operator()(float4 a, float4 b){ accelerator_inline float4 operator()(float4 a, float4 b){
float4 ymm0; float4 ymm0;
ymm0.x = a.y; ymm0.x = a.x;
ymm0.y = a.y; ymm0.y = a.x;
ymm0.z = a.w; ymm0.z = a.z;
ymm0.w = a.w; ymm0.w = a.z;
return ymm0*b; return ymm0*b;
// ymm0 = _mm_shuffle_ps(a,a,_MM_SELECT_FOUR_FOUR(2,2,0,0)); // ymm0 <- ar ar, // ymm0 = _mm_shuffle_ps(a,a,_MM_SELECT_FOUR_FOUR(2,2,0,0)); // ymm0 <- ar ar,
// return _mm_mul_ps(ymm0,b); // ymm0 <- ar bi, ar br // return _mm_mul_ps(ymm0,b); // ymm0 <- ar bi, ar br
} }
accelerator_inline double2 operator()(double2 a, double2 b){ accelerator_inline double2 operator()(double2 a, double2 b){
double2 ymm0; double2 ymm0;
ymm0.x = a.y; ymm0.x = a.x;
ymm0.y = a.y; ymm0.y = a.x;
return ymm0*b; return ymm0*b;
// ymm0 = _mm_shuffle_pd(a,a,0x0); // ymm0 <- ar ar, ar,ar b'00,00 // ymm0 = _mm_shuffle_pd(a,a,0x0); // ymm0 <- ar ar, ar,ar b'00,00
// return _mm_mul_pd(ymm0,b); // ymm0 <- ar bi, ar br // return _mm_mul_pd(ymm0,b); // ymm0 <- ar bi, ar br
@ -265,17 +265,17 @@ namespace Optimization {
struct MaddRealPart{ struct MaddRealPart{
accelerator_inline float4 operator()(float4 a, float4 b, float4 c){ accelerator_inline float4 operator()(float4 a, float4 b, float4 c){
float4 ymm0; // = _mm_shuffle_ps(a,a,_MM_SELECT_FOUR_FOUR(2,2,0,0)); // ymm0 <- ar ar, float4 ymm0; // = _mm_shuffle_ps(a,a,_MM_SELECT_FOUR_FOUR(2,2,0,0)); // ymm0 <- ar ar,
ymm0.x = a.y; ymm0.x = a.x;
ymm0.y = a.y; ymm0.y = a.x;
ymm0.z = a.w; ymm0.z = a.z;
ymm0.w = a.w; ymm0.w = a.z;
return c+ymm0*b; return c+ymm0*b;
} }
accelerator_inline double2 operator()(double2 a, double2 b, double2 c){ accelerator_inline double2 operator()(double2 a, double2 b, double2 c){
// ymm0 = _mm_shuffle_pd( a, a, 0x0 ); // ymm0 = _mm_shuffle_pd( a, a, 0x0 );
double2 ymm0; double2 ymm0;
ymm0.x = a.y; ymm0.x = a.x;
ymm0.y = a.y; ymm0.y = a.x;
return c+ymm0*b; return c+ymm0*b;
} }
}; };