devtools/eigen
1
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2026-04-10 11:34:33 +08:00
Code Issues Packages Projects Releases Wiki Activity

Remove inline assembly for FMA (AVX) and add remaining extensions as packet ops: pmsub, pnmadd, and pnmsub.

Browse Source
This commit is contained in:
Rasmus Munk Larsen
2022-01-26 04:25:41 +00:00
parent 4e629b3c1b
commit 51311ec651
5 changed files with 152 additions and 31 deletions
Download Patch File Download Diff File Expand all files Collapse all files

24
Eigen/src/Core/arch/SSE/PacketMath.h
View File

@@ -364,6 +364,12 @@ template<> EIGEN_STRONG_INLINE Packet4i pmadd(const Packet4i& a, const Packet4i&
#ifdef EIGEN_VECTORIZE_FMA
template<> EIGEN_STRONG_INLINE Packet4f pmadd(const Packet4f& a, const Packet4f& b, const Packet4f& c) { return _mm_fmadd_ps(a,b,c); }
template<> EIGEN_STRONG_INLINE Packet2d pmadd(const Packet2d& a, const Packet2d& b, const Packet2d& c) { return _mm_fmadd_pd(a,b,c); }
template<> EIGEN_STRONG_INLINE Packet4f pmsub(const Packet4f& a, const Packet4f& b, const Packet4f& c) { return _mm_fmsub_ps(a,b,c); }
template<> EIGEN_STRONG_INLINE Packet2d pmsub(const Packet2d& a, const Packet2d& b, const Packet2d& c) { return _mm_fmsub_pd(a,b,c); }
template<> EIGEN_STRONG_INLINE Packet4f pnmadd(const Packet4f& a, const Packet4f& b, const Packet4f& c) { return _mm_fnmadd_ps(a,b,c); }
template<> EIGEN_STRONG_INLINE Packet2d pnmadd(const Packet2d& a, const Packet2d& b, const Packet2d& c) { return _mm_fnmadd_pd(a,b,c); }
template<> EIGEN_STRONG_INLINE Packet4f pnmsub(const Packet4f& a, const Packet4f& b, const Packet4f& c) { return _mm_fnmsub_ps(a,b,c); }
template<> EIGEN_STRONG_INLINE Packet2d pnmsub(const Packet2d& a, const Packet2d& b, const Packet2d& c) { return _mm_fnmsub_pd(a,b,c); }
#endif
#ifdef EIGEN_VECTORIZE_SSE4_1
@@ -1263,6 +1269,24 @@ template<> EIGEN_STRONG_INLINE float pmadd(const float& a, const float& b, const
template<> EIGEN_STRONG_INLINE double pmadd(const double& a, const double& b, const double& c) {
return ::fma(a,b,c);
}
template<> EIGEN_STRONG_INLINE float pmsub(const float& a, const float& b, const float& c) {
return ::fmaf(a,b,-c);
}
template<> EIGEN_STRONG_INLINE double pmsub(const double& a, const double& b, const double& c) {
return ::fma(a,b,-c);
}
template<> EIGEN_STRONG_INLINE float pnmadd(const float& a, const float& b, const float& c) {
return ::fmaf(-a,b,c);
}
template<> EIGEN_STRONG_INLINE double pnmadd(const double& a, const double& b, const double& c) {
return ::fma(-a,b,c);
}
template<> EIGEN_STRONG_INLINE float pnmsub(const float& a, const float& b, const float& c) {
return ::fmaf(-a,b,-c);
}
template<> EIGEN_STRONG_INLINE double pnmsub(const double& a, const double& b, const double& c) {
return ::fma(-a,b,-c);
}
#endif
#ifdef EIGEN_VECTORIZE_SSE4_1