diff options
author | Gael Guennebaud <g.gael@free.fr> | 2014-04-16 17:05:11 +0200 |
---|---|---|
committer | Gael Guennebaud <g.gael@free.fr> | 2014-04-16 17:05:11 +0200 |
commit | d5a795f67366db20a132cc70e4f0217f42372357 (patch) | |
tree | 74df7a911811e64a4fa0baff940abe9c97abd5b6 /Eigen/src/Core/arch/NEON | |
parent | feaf7c7e6d01a4804cee5949a01ece1f8a46866f (diff) |
New gebp kernel handling up to 3 packets x 4 register-level blocks. Huge speeup on Haswell.
This changeset also introduce new vector functions: ploadquad and predux4.
Diffstat (limited to 'Eigen/src/Core/arch/NEON')
-rw-r--r-- | Eigen/src/Core/arch/NEON/Complex.h | 2 | ||||
-rw-r--r-- | Eigen/src/Core/arch/NEON/PacketMath.h | 4 |
2 files changed, 3 insertions, 3 deletions
diff --git a/Eigen/src/Core/arch/NEON/Complex.h b/Eigen/src/Core/arch/NEON/Complex.h index e49c1a873..7ca76714f 100644 --- a/Eigen/src/Core/arch/NEON/Complex.h +++ b/Eigen/src/Core/arch/NEON/Complex.h @@ -47,7 +47,7 @@ template<> struct packet_traits<std::complex<float> > : default_packet_traits }; }; -template<> struct unpacket_traits<Packet2cf> { typedef std::complex<float> type; enum {size=2}; }; +template<> struct unpacket_traits<Packet2cf> { typedef std::complex<float> type; enum {size=2}; typedef Packet2cf half; }; template<> EIGEN_STRONG_INLINE Packet2cf pset1<Packet2cf>(const std::complex<float>& from) { diff --git a/Eigen/src/Core/arch/NEON/PacketMath.h b/Eigen/src/Core/arch/NEON/PacketMath.h index fae7b55fc..83150507a 100644 --- a/Eigen/src/Core/arch/NEON/PacketMath.h +++ b/Eigen/src/Core/arch/NEON/PacketMath.h @@ -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); } |