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

Half precision conversion working on AVX512 now too

Browse Source
This commit is contained in:
Peter Boyle 2017-04-13 17:35:11 +01:00
parent b9113ed310
commit 951be75292
1 changed files with 4 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
lib/simd/Grid_avx512.h
View File

@ -347,8 +347,8 @@ namespace Optimization {
#ifdef USE_FP16 #ifdef USE_FP16
__m256i ha = _mm512_cvtps_ph(a,0); __m256i ha = _mm512_cvtps_ph(a,0);
__m256i hb = _mm512_cvtps_ph(b,0); __m256i hb = _mm512_cvtps_ph(b,0);
h =(__m512i) _mm512_castps256_ps512(ha); h =(__m512i) _mm512_castps256_ps512((__m256)ha);
h =(__m512i) _mm512_insertf64x4((__m512d)h,(__m512d)hb,1); h =(__m512i) _mm512_insertf64x4((__m512d)h,(__m256d)hb,1);
#else #else
assert(0); assert(0);
#endif #endif
@ -356,8 +356,8 @@ namespace Optimization {
} }
static inline void HtoS (__m512i h,__m512 &sa,__m512 &sb) { static inline void HtoS (__m512i h,__m512 &sa,__m512 &sb) {
#ifdef USE_FP16 #ifdef USE_FP16
sa = _mm512_cvtph_ps(_mm512_extractf256_ps(h,0)); sa = _mm512_cvtph_ps((__m256i)_mm512_extractf64x4_pd((__m512d)h,0));
sb = _mm512_cvtph_ps(_mm512_extractf256_ps(h,1)); sb = _mm512_cvtph_ps((__m256i)_mm512_extractf64x4_pd((__m512d)h,1));
#else #else
assert(0); assert(0);
#endif #endif