ForkMaster/netgen
1
0
Fork 0
mirror of https://github.com/NGSolve/netgen.git synced 2025-01-23 11:20:34 +05:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'fix_windows_avx2' into 'master'

Browse Source 
fix AVX2 for Windows build

See merge request ngsolve/netgen!514
This commit is contained in:
Hochsteger, Matthias 2022-08-23 09:33:38 +02:00
commit cee2ca18fc
1 changed files with 5 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
libsrc/core/simd_avx512.hpp
View File

@ -143,7 +143,7 @@ namespace ngcore
}
};
NETGEN_INLINE SIMD<double,8> operator- (SIMD<double,8> a) { return -a.Data(); }
NETGEN_INLINE SIMD<double,8> operator- (SIMD<double,8> a) { return _mm512_xor_pd(a.Data(), _mm512_set1_pd(-0.0)); } //{ return -a.Data(); }
NETGEN_INLINE SIMD<double,8> operator+ (SIMD<double,8> a, SIMD<double,8> b) { return _mm512_add_pd(a.Data(),b.Data()); }
NETGEN_INLINE SIMD<double,8> operator- (SIMD<double,8> a, SIMD<double,8> b) { return _mm512_sub_pd(a.Data(),b.Data()); }
NETGEN_INLINE SIMD<double,8> operator* (SIMD<double,8> a, SIMD<double,8> b) { return _mm512_mul_pd(a.Data(),b.Data()); }
@ -154,7 +154,7 @@ namespace ngcore
NETGEN_INLINE SIMD<double,8> sqrt (SIMD<double,8> a) { return _mm512_sqrt_pd(a.Data()); }
NETGEN_INLINE SIMD<double,8> floor (SIMD<double,8> a) { return _mm512_floor_pd(a.Data()); }
NETGEN_INLINE SIMD<double,8> ceil (SIMD<double,8> a) { return _mm512_ceil_pd(a.Data()); }
NETGEN_INLINE SIMD<double,8> fabs (SIMD<double,8> a) { return _mm512_max_pd(a.Data(), -a.Data()); }
NETGEN_INLINE SIMD<double,8> fabs (SIMD<double,8> a) { return _mm512_max_pd(a.Data(), ( - a).Data()); }
NETGEN_INLINE SIMD<mask64,8> operator<= (SIMD<double,8> a , SIMD<double,8> b)
{ return _mm512_cmp_pd_mask (a.Data(), b.Data(), _CMP_LE_OQ); }
@ -233,9 +233,9 @@ namespace ngcore
// sum01 b a b a b a b a
// sum23 d c d c d c d c
// __m512 perm = _mm512_permutex2var_pd (sum01.Data(), _mm512_set_epi64(1,2,3,4,5,6,7,8), sum23.Data());
__m256d ab = _mm512_extractf64x4_pd(sum01.Data(),0) + _mm512_extractf64x4_pd(sum01.Data(),1);
__m256d cd = _mm512_extractf64x4_pd(sum23.Data(),0) + _mm512_extractf64x4_pd(sum23.Data(),1);
return _mm256_add_pd (_mm256_permute2f128_pd (ab, cd, 1+2*16), _mm256_blend_pd (ab, cd, 12));
SIMD<double,4> ab = _mm512_extractf64x4_pd(sum01.Data(),0) + _mm512_extractf64x4_pd(sum01.Data(),1);
SIMD<double,4> cd = _mm512_extractf64x4_pd(sum23.Data(),0) + _mm512_extractf64x4_pd(sum23.Data(),1);
return _mm256_add_pd (_mm256_permute2f128_pd (ab.Data(), cd.Data(), 1 + 2 * 16), _mm256_blend_pd(ab.Data(), cd.Data(), 12));
}
NETGEN_INLINE SIMD<double,8> FMA (SIMD<double,8> a, SIMD<double,8> b, SIMD<double,8> c)