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 packet ops pinsertfirst and pinsertlast that are only used in a single place, and can be replaced by other ops when constructing the first/final packet in linspaced_op_impl::packetOp.

Browse Source 
I cannot measure any performance changes for SSE, AVX, or AVX512.

name                                 old time/op             new time/op             delta
BM_LinSpace<float>/1                 1.63ns ± 0%             1.63ns ± 0%   ~             (p=0.762 n=5+5)
BM_LinSpace<float>/8                 4.92ns ± 3%             4.89ns ± 3%   ~             (p=0.421 n=5+5)
BM_LinSpace<float>/64                34.6ns ± 0%             34.6ns ± 0%   ~             (p=0.841 n=5+5)
BM_LinSpace<float>/512                217ns ± 0%              217ns ± 0%   ~             (p=0.421 n=5+5)
BM_LinSpace<float>/4k                1.68µs ± 0%             1.68µs ± 0%   ~             (p=1.000 n=5+5)
BM_LinSpace<float>/32k               13.3µs ± 0%             13.3µs ± 0%   ~             (p=0.905 n=5+4)
BM_LinSpace<float>/256k               107µs ± 0%              107µs ± 0%   ~             (p=0.841 n=5+5)
BM_LinSpace<float>/1M                 427µs ± 0%              427µs ± 0%   ~             (p=0.690 n=5+5)
This commit is contained in:
Rasmus Munk Larsen
2020-05-08 15:41:50 -07:00
parent 5c4e19fbe7
commit c1d944dd91
11 changed files with 6 additions and 270 deletions
Download Patch File Download Diff File Expand all files Collapse all files

38
Eigen/src/Core/arch/SSE/PacketMath.h
View File

@@ -936,44 +936,6 @@ template<> EIGEN_STRONG_INLINE Packet2d pblend(const Selector<2>& ifPacket, cons
#endif
}
template<> EIGEN_STRONG_INLINE Packet4f pinsertfirst(const Packet4f& a, float b)
{
#ifdef EIGEN_VECTORIZE_SSE4_1
return _mm_blend_ps(a,pset1<Packet4f>(b),1);
#else
return _mm_move_ss(a, _mm_load_ss(&b));
#endif
}
template<> EIGEN_STRONG_INLINE Packet2d pinsertfirst(const Packet2d& a, double b)
{
#ifdef EIGEN_VECTORIZE_SSE4_1
return _mm_blend_pd(a,pset1<Packet2d>(b),1);
#else
return _mm_move_sd(a, _mm_load_sd(&b));
#endif
}
template<> EIGEN_STRONG_INLINE Packet4f pinsertlast(const Packet4f& a, float b)
{
#ifdef EIGEN_VECTORIZE_SSE4_1
return _mm_blend_ps(a,pset1<Packet4f>(b),(1<<3));
#else
const Packet4f mask = _mm_castsi128_ps(_mm_setr_epi32(0x0,0x0,0x0,0xFFFFFFFF));
return _mm_or_ps(_mm_andnot_ps(mask, a), _mm_and_ps(mask, pset1<Packet4f>(b)));
#endif
}
template<> EIGEN_STRONG_INLINE Packet2d pinsertlast(const Packet2d& a, double b)
{
#ifdef EIGEN_VECTORIZE_SSE4_1
return _mm_blend_pd(a,pset1<Packet2d>(b),(1<<1));
#else
const Packet2d mask = _mm_castsi128_pd(_mm_setr_epi32(0x0,0x0,0xFFFFFFFF,0xFFFFFFFF));
return _mm_or_pd(_mm_andnot_pd(mask, a), _mm_and_pd(mask, pset1<Packet2d>(b)));
#endif
}
// Scalar path for pmadd with FMA to ensure consistency with vectorized path.
#ifdef EIGEN_VECTORIZE_FMA
template<> EIGEN_STRONG_INLINE float pmadd(const float& a, const float& b, const float& c) {