moebiusband/MD-Bench
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Add working version of Simd4xn kernel with half neighbor lists

Browse Source 
Signed-off-by: Rafael Ravedutti <rafaelravedutti@gmail.com>
This commit is contained in:
Rafael Ravedutti
2022-03-23 15:54:18 +01:00
parent e72323ab6a
commit 0e742766b7
3 changed files with 32 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
gromacs/includes/simd/avx512_double.h
View File

@@ -40,6 +40,7 @@ static inline MD_SIMD_MASK simd_mask_cond_lt(MD_SIMD_FLOAT a, MD_SIMD_FLOAT b) {
static inline MD_SIMD_MASK simd_mask_from_u32(unsigned int a) { return _cvtu32_mask8(a); }
static inline unsigned int simd_mask_to_u32(MD_SIMD_MASK a) { return _cvtmask8_u32(a); }
static inline MD_SIMD_FLOAT simd_load(MD_FLOAT *p) { return _mm512_load_pd(p); }
static inline void simd_store(MD_FLOAT *p, MD_SIMD_FLOAT a) { _mm512_store_pd(p, a); }
static inline MD_SIMD_FLOAT select_by_mask(MD_SIMD_FLOAT a, MD_SIMD_MASK m) { return _mm512_mask_mov_pd(_mm512_setzero_pd(), m, a); }
static inline MD_FLOAT simd_h_reduce_sum(MD_SIMD_FLOAT a) {
MD_SIMD_FLOAT x = _mm512_add_pd(a, _mm512_shuffle_f64x2(a, a, 0xee));

1
gromacs/includes/simd/avx512_float.h
View File

@@ -42,6 +42,7 @@ static inline MD_SIMD_MASK simd_mask_cond_lt(MD_SIMD_FLOAT a, MD_SIMD_FLOAT b) {
static inline MD_SIMD_MASK simd_mask_from_u32(unsigned int a) { return _cvtu32_mask16(a); }
static inline unsigned int simd_mask_to_u32(MD_SIMD_MASK a) { return _cvtmask16_u32(a); }
static inline MD_SIMD_FLOAT simd_load(MD_FLOAT *p) { return _mm512_load_ps(p); }
static inline void simd_store(MD_FLOAT *p, MD_SIMD_FLOAT a) { _mm512_store_ps(p, a); }
static inline MD_SIMD_FLOAT select_by_mask(MD_SIMD_FLOAT a, MD_SIMD_MASK m) { return _mm512_mask_mov_ps(_mm512_setzero_ps(), m, a); }
static inline MD_FLOAT simd_h_reduce_sum(MD_SIMD_FLOAT a) {
// This would only be called in a Mx16 configuration, which is not valid in GROMACS