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.

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); }