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

New gebp kernel handling up to 3 packets x 4 register-level blocks. Huge speeup on Haswell.

Browse Source 
This changeset also introduce new vector functions: ploadquad and predux4.
This commit is contained in:
Gael Guennebaud
2014-04-16 17:05:11 +02:00
parent feaf7c7e6d
commit d5a795f673
12 changed files with 1064 additions and 633 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
Eigen/src/Core/arch/NEON/PacketMath.h
View File

@@ -101,8 +101,8 @@ EIGEN_STRONG_INLINE void vst1q_f32(float* to, float32x4_t from) { ::vst1q
EIGEN_STRONG_INLINE void vst1_f32 (float* to, float32x2_t from) { ::vst1_f32 ((float32_t*)to,from); }
#endif
template<> struct unpacket_traits<Packet4f> { typedef float type; enum {size=4}; };
template<> struct unpacket_traits<Packet4i> { typedef int type; enum {size=4}; };
template<> struct unpacket_traits<Packet4f> { typedef float type; enum {size=4}; typedef Packet4f half; };
template<> struct unpacket_traits<Packet4i> { typedef int type; enum {size=4}; typedef Packet4i half; };
template<> EIGEN_STRONG_INLINE Packet4f pset1<Packet4f>(const float& from) { return vdupq_n_f32(from); }
template<> EIGEN_STRONG_INLINE Packet4i pset1<Packet4i>(const int& from) { return vdupq_n_s32(from); }