portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2024-11-10 07:55:35 +00:00
Code Releases Activity

Intel vs. Clang intrinsics differences absorbed

Browse Source
This commit is contained in:
azusayamaguchi 2016-10-21 09:08:36 +01:00
parent 202078eb1b
commit 20a091c3ed
1 changed files with 19 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

21
lib/simd/Grid_avx512.h
View File

@ -42,6 +42,22 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
namespace Grid{ namespace Grid{
namespace Optimization { namespace Optimization {
template<class vtype>
union uconv {
__m512 f;
vtype v;
};
union u512f {
__m512 v;
float f[8];
};
union u512d {
__m512 v;
double f[4];
};
struct Vsplat{ struct Vsplat{
//Complex float //Complex float
inline __m512 operator()(float a, float b){ inline __m512 operator()(float a, float b){
@ -361,8 +377,9 @@ namespace Optimization {
// Some Template specialization // Some Template specialization
// Hack for CLANG until mm512_reduce_add_ps etc... are implemented in GCC and Clang releases // Hack for CLANG until mm512_reduce_add_ps etc... are implemented in GCC and Clang releases
#undef GNU_CLANG_COMPILER
#ifdef GNU_CLANG_COMPILER #ifndef __INTEL_COMPILER
#warning "Slow reduction due to incomplete reduce intrinsics"
//Complex float Reduce //Complex float Reduce
template<> template<>
inline Grid::ComplexF Reduce<Grid::ComplexF, __m512>::operator()(__m512 in){ inline Grid::ComplexF Reduce<Grid::ComplexF, __m512>::operator()(__m512 in){