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

1)provide a better generic paddsub op implementation

Browse Source 
2)make paddsub op support the Packet2cf/Packet4f/Packet2f in NEON
3)make paddsub op support the Packet2cf/Packet4f in SSE
This commit is contained in:
Guoqiang QI
2021-01-13 22:54:03 +00:00
committed by Rasmus Munk Larsen
parent 352f1422d3
commit 38ae5353ab
10 changed files with 115 additions and 40 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
Eigen/src/Core/arch/SSE/Complex.h
View File

@@ -66,6 +66,13 @@ template<> struct unpacket_traits<Packet2cf> {
template<> EIGEN_STRONG_INLINE Packet2cf padd<Packet2cf>(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(_mm_add_ps(a.v,b.v)); }
template<> EIGEN_STRONG_INLINE Packet2cf psub<Packet2cf>(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(_mm_sub_ps(a.v,b.v)); }
template<> EIGEN_STRONG_INLINE Packet2cf pxor<Packet2cf>(const Packet2cf& a, const Packet2cf& b);
template<> EIGEN_STRONG_INLINE Packet2cf paddsub<Packet2cf>(const Packet2cf& a, const Packet2cf& b)
{
const Packet4f mask = _mm_castsi128_ps(_mm_setr_epi32(0x80000000,0x80000000,0x0,0x0));
return Packet2cf(padd(a.v, pxor(mask, b.v)));
}
template<> EIGEN_STRONG_INLINE Packet2cf pnegate(const Packet2cf& a)
{
const __m128 mask = _mm_castsi128_ps(_mm_setr_epi32(0x80000000,0x80000000,0x80000000,0x80000000));

22
Eigen/src/Core/arch/SSE/PacketMath.h
View File

@@ -301,6 +301,28 @@ template<> EIGEN_STRONG_INLINE Packet2d psub<Packet2d>(const Packet2d& a, const
template<> EIGEN_STRONG_INLINE Packet4i psub<Packet4i>(const Packet4i& a, const Packet4i& b) { return _mm_sub_epi32(a,b); }
template<> EIGEN_STRONG_INLINE Packet16b psub<Packet16b>(const Packet16b& a, const Packet16b& b) { return _mm_xor_si128(a,b); }
template<> EIGEN_STRONG_INLINE Packet4f pxor<Packet4f>(const Packet4f& a, const Packet4f& b);
template<> EIGEN_STRONG_INLINE Packet4f paddsub<Packet4f>(const Packet4f& a, const Packet4f& b)
{
#ifdef EIGEN_VECTORIZE_SSE3
return _mm_addsub_ps(a,b);
#else
const Packet4f mask = _mm_castsi128_ps(_mm_setr_epi32(0x80000000,0x0,0x80000000,0x0));
return padd(a, pxor(mask, b));
#endif
}
template<> EIGEN_STRONG_INLINE Packet2d pxor<Packet2d>(const Packet2d& , const Packet2d& );
template<> EIGEN_STRONG_INLINE Packet2d paddsub<Packet2d>(const Packet2d& a, const Packet2d& b)
{
#ifdef EIGEN_VECTORIZE_SSE3
return _mm_addsub_pd(a,b);
#else
const Packet2d mask = _mm_castsi128_pd(_mm_setr_epi32(0x0,0x80000000,0x0,0x0));
return padd(a, pxor(mask, b));
#endif
}
template<> EIGEN_STRONG_INLINE Packet4f pnegate(const Packet4f& a)
{
const Packet4f mask = _mm_castsi128_ps(_mm_setr_epi32(0x80000000,0x80000000,0x80000000,0x80000000));