mirror/opencv
1
0
Fork 0
mirror of https://github.com/opencv/opencv.git synced 2025-06-07 01:13:28 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

Merge pull request #21011 from vrabaud:3.4

Browse Source
This commit is contained in:
Alexander Alekhin 2021-11-05 09:25:50 +00:00
commit 2ce47fda88
2 changed files with 26 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
modules/core/include/opencv2/core/hal/intrin_avx.hpp
View File

@ -1390,11 +1390,21 @@ OPENCV_HAL_IMPL_AVX_CHECK_SHORT(v_int16x16)
////////// Other math /////////
/** Some frequent operations **/
#if CV_FMA3
#define OPENCV_HAL_IMPL_AVX_MULADD(_Tpvec, suffix) \
inline _Tpvec v_fma(const _Tpvec& a, const _Tpvec& b, const _Tpvec& c) \
{ return _Tpvec(_mm256_fmadd_##suffix(a.val, b.val, c.val)); } \
inline _Tpvec v_muladd(const _Tpvec& a, const _Tpvec& b, const _Tpvec& c) \
{ return _Tpvec(_mm256_fmadd_##suffix(a.val, b.val, c.val)); } \
{ return _Tpvec(_mm256_fmadd_##suffix(a.val, b.val, c.val)); }
#else
#define OPENCV_HAL_IMPL_AVX_MULADD(_Tpvec, suffix) \
inline _Tpvec v_fma(const _Tpvec& a, const _Tpvec& b, const _Tpvec& c) \
{ return _Tpvec(_mm256_add_##suffix(_mm256_mul_##suffix(a.val, b.val), c.val)); } \
inline _Tpvec v_muladd(const _Tpvec& a, const _Tpvec& b, const _Tpvec& c) \
{ return _Tpvec(_mm256_add_##suffix(_mm256_mul_##suffix(a.val, b.val), c.val)); }
#endif
#define OPENCV_HAL_IMPL_AVX_MISC(_Tpvec, suffix) \
inline _Tpvec v_sqrt(const _Tpvec& x) \
{ return _Tpvec(_mm256_sqrt_##suffix(x.val)); } \
inline _Tpvec v_sqr_magnitude(const _Tpvec& a, const _Tpvec& b) \
@ -1404,6 +1414,8 @@ OPENCV_HAL_IMPL_AVX_CHECK_SHORT(v_int16x16)
OPENCV_HAL_IMPL_AVX_MULADD(v_float32x8, ps)
OPENCV_HAL_IMPL_AVX_MULADD(v_float64x4, pd)
OPENCV_HAL_IMPL_AVX_MISC(v_float32x8, ps)
OPENCV_HAL_IMPL_AVX_MISC(v_float64x4, pd)
inline v_int32x8 v_fma(const v_int32x8& a, const v_int32x8& b, const v_int32x8& c)
{

14
modules/core/include/opencv2/core/hal/intrin_avx512.hpp
View File

@ -1385,11 +1385,21 @@ inline v_uint64x8 v_popcount(const v_uint64x8& a) { return v_popcount(v_reinte
////////// Other math /////////
/** Some frequent operations **/
#if CV_FMA3
#define OPENCV_HAL_IMPL_AVX512_MULADD(_Tpvec, suffix) \
inline _Tpvec v_fma(const _Tpvec& a, const _Tpvec& b, const _Tpvec& c) \
{ return _Tpvec(_mm512_fmadd_##suffix(a.val, b.val, c.val)); } \
inline _Tpvec v_muladd(const _Tpvec& a, const _Tpvec& b, const _Tpvec& c) \
{ return _Tpvec(_mm512_fmadd_##suffix(a.val, b.val, c.val)); } \
{ return _Tpvec(_mm512_fmadd_##suffix(a.val, b.val, c.val)); }
#else
#define OPENCV_HAL_IMPL_AVX512_MULADD(_Tpvec, suffix) \
inline _Tpvec v_fma(const _Tpvec& a, const _Tpvec& b, const _Tpvec& c) \
{ return _Tpvec(_mm512_add_##suffix(_mm512_mul_##suffix(a.val, b.val), c.val)); } \
inline _Tpvec v_muladd(const _Tpvec& a, const _Tpvec& b, const _Tpvec& c) \
{ return _Tpvec(_mm512_add_##suffix(_mm512_mul_##suffix(a.val, b.val), c.val)); }
#endif
#define OPENCV_HAL_IMPL_AVX512_MISC(_Tpvec, suffix) \
inline _Tpvec v_sqrt(const _Tpvec& x) \
{ return _Tpvec(_mm512_sqrt_##suffix(x.val)); } \
inline _Tpvec v_sqr_magnitude(const _Tpvec& a, const _Tpvec& b) \
@ -1399,6 +1409,8 @@ inline v_uint64x8 v_popcount(const v_uint64x8& a) { return v_popcount(v_reinte
OPENCV_HAL_IMPL_AVX512_MULADD(v_float32x16, ps)
OPENCV_HAL_IMPL_AVX512_MULADD(v_float64x8, pd)
OPENCV_HAL_IMPL_AVX512_MISC(v_float32x16, ps)
OPENCV_HAL_IMPL_AVX512_MISC(v_float64x8, pd)
inline v_int32x16 v_fma(const v_int32x16& a, const v_int32x16& b, const v_int32x16& c)
{ return a * b + c; }