GitHub-Proxy/eigen
1
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2025-08-12 11:49:02 +08:00
Code Issues Packages Projects Releases Wiki Activity

Makes gcc to generate a pshufd instruction for pset1

Browse Source
This commit is contained in:
Gael Guennebaud 2014-03-20 10:14:26 +01:00
parent 2a564695f0
commit c39a3fa7a1
1 changed files with 14 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

15
Eigen/src/Core/arch/SSE/PacketMath.h
View File

@ -110,7 +110,20 @@ template<> EIGEN_STRONG_INLINE Packet4f pset1<Packet4f>(const float& from) { re
template<> EIGEN_STRONG_INLINE Packet2d pset1<Packet2d>(const double& from) { return _mm_set_pd(from,from); } template<> EIGEN_STRONG_INLINE Packet2d pset1<Packet2d>(const double& from) { return _mm_set_pd(from,from); }
template<> EIGEN_STRONG_INLINE Packet4i pset1<Packet4i>(const int& from) { return _mm_set_epi32(from,from,from,from); } template<> EIGEN_STRONG_INLINE Packet4i pset1<Packet4i>(const int& from) { return _mm_set_epi32(from,from,from,from); }
#else #else
template<> EIGEN_STRONG_INLINE Packet4f pset1<Packet4f>(const float& from) { return _mm_set1_ps(from); }
// GCC generates a shufps instruction for set1_ps instead of the more efficient pshufd instruction.
// However, with AVX, we want it to generate a vbroadcastss.
// Moreover, we cannot use intrinsics here because then gcc generates crappy code in some cases (see bug 203)
#if (defined __GNUC__) && (!defined __INTEL_COMPILER) && (!defined __clang__) && (!defined __AVX__)
template<> EIGEN_STRONG_INLINE Packet4f pset1<Packet4f>(const float& from) {
Packet4f res;
asm("pshufd $0, %[a], %[b]" : [b] "=x" (res) : [a] "x" (from));
return res;
}
#else
template<> EIGEN_STRONG_INLINE Packet4f pset1<Packet4f>(const float& from) { return _mm_set_ps1(from); }
#endif
template<> EIGEN_STRONG_INLINE Packet2d pset1<Packet2d>(const double& from) { return _mm_set1_pd(from); } template<> EIGEN_STRONG_INLINE Packet2d pset1<Packet2d>(const double& from) { return _mm_set1_pd(from); }
template<> EIGEN_STRONG_INLINE Packet4i pset1<Packet4i>(const int& from) { return _mm_set1_epi32(from); } template<> EIGEN_STRONG_INLINE Packet4i pset1<Packet4i>(const int& from) { return _mm_set1_epi32(from); }
#endif #endif