portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2025-04-04 19:25:56 +01:00
Code Releases Activity

USE_FP16 macro

Browse Source
This commit is contained in:
paboyle 2017-04-14 14:17:14 +01:00
parent 42fb49d3fd
commit 9e2d29c644
2 changed files with 6 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
lib/simd/Grid_avx.h
View File

@ -470,13 +470,14 @@ namespace Optimization {
return in;
};
};
#define USE_FP16
struct PrecisionChange {
static inline __m256i StoH (__m256 a,__m256 b) {
__m256 h ;
#ifdef USE_FP16
__m128i ha = _mm256_cvtps_ph(a,0);
__m128i hb = _mm256_cvtps_ph(b,0);
__m256 h = _mm256_castps128_ps256(ha);
h = _mm256_castps128_ps256(ha);
h = _mm256_insertf128_ps(h,hb,1);
#else
assert(0);

5
lib/simd/Grid_avx512.h
View File

@ -340,13 +340,14 @@ namespace Optimization {
};
};
#define USE_FP16
struct PrecisionChange {
static inline __m512i StoH (__m512 a,__m512 b) {
__m512 h ;
#ifdef USE_FP16
__m256i ha = _mm512_cvtps_ph(a,0);
__m256i hb = _mm512_cvtps_ph(b,0);
__m512 h = _mm512_castps256_ps512(ha);
h = _mm512_castps256_ps512(ha);
h = _mm512_insertf256_ps(h,hb,1);
#else
assert(0);