diff options
author | Gael Guennebaud <g.gael@free.fr> | 2015-08-07 10:44:01 +0200 |
---|---|---|
committer | Gael Guennebaud <g.gael@free.fr> | 2015-08-07 10:44:01 +0200 |
commit | ce57dbd937ca250863d4d52f3d0940d57551a25c (patch) | |
tree | 39f15bd3db545a9e52d88ba48c37ddb6083f7ade /Eigen/src/Core/arch | |
parent | 2afdef6a54e7fd09a4a6d5e933cf15ffa129beb6 (diff) |
Let unpacket_traits<> exposes the required alignment and make use of it everywhere
Diffstat (limited to 'Eigen/src/Core/arch')
-rw-r--r-- | Eigen/src/Core/arch/AVX/Complex.h | 4 | ||||
-rw-r--r-- | Eigen/src/Core/arch/AVX/PacketMath.h | 6 | ||||
-rw-r--r-- | Eigen/src/Core/arch/AltiVec/Complex.h | 4 | ||||
-rwxr-xr-x | Eigen/src/Core/arch/AltiVec/PacketMath.h | 6 | ||||
-rw-r--r-- | Eigen/src/Core/arch/CUDA/PacketMath.h | 4 | ||||
-rw-r--r-- | Eigen/src/Core/arch/NEON/Complex.h | 4 | ||||
-rw-r--r-- | Eigen/src/Core/arch/NEON/PacketMath.h | 6 | ||||
-rw-r--r-- | Eigen/src/Core/arch/SSE/Complex.h | 4 | ||||
-rwxr-xr-x | Eigen/src/Core/arch/SSE/PacketMath.h | 6 |
9 files changed, 22 insertions, 22 deletions
diff --git a/Eigen/src/Core/arch/AVX/Complex.h b/Eigen/src/Core/arch/AVX/Complex.h index 003a1fc3c..b16e0ddd4 100644 --- a/Eigen/src/Core/arch/AVX/Complex.h +++ b/Eigen/src/Core/arch/AVX/Complex.h @@ -45,7 +45,7 @@ template<> struct packet_traits<std::complex<float> > : default_packet_traits }; }; -template<> struct unpacket_traits<Packet4cf> { typedef std::complex<float> type; enum {size=4}; typedef Packet2cf half; }; +template<> struct unpacket_traits<Packet4cf> { typedef std::complex<float> type; enum {size=4, alignment=Aligned32}; typedef Packet2cf half; }; template<> EIGEN_STRONG_INLINE Packet4cf padd<Packet4cf>(const Packet4cf& a, const Packet4cf& b) { return Packet4cf(_mm256_add_ps(a.v,b.v)); } template<> EIGEN_STRONG_INLINE Packet4cf psub<Packet4cf>(const Packet4cf& a, const Packet4cf& b) { return Packet4cf(_mm256_sub_ps(a.v,b.v)); } @@ -267,7 +267,7 @@ template<> struct packet_traits<std::complex<double> > : default_packet_traits }; }; -template<> struct unpacket_traits<Packet2cd> { typedef std::complex<double> type; enum {size=2}; typedef Packet1cd half; }; +template<> struct unpacket_traits<Packet2cd> { typedef std::complex<double> type; enum {size=2, alignment=Aligned32}; typedef Packet1cd half; }; template<> EIGEN_STRONG_INLINE Packet2cd padd<Packet2cd>(const Packet2cd& a, const Packet2cd& b) { return Packet2cd(_mm256_add_pd(a.v,b.v)); } template<> EIGEN_STRONG_INLINE Packet2cd psub<Packet2cd>(const Packet2cd& a, const Packet2cd& b) { return Packet2cd(_mm256_sub_pd(a.v,b.v)); } diff --git a/Eigen/src/Core/arch/AVX/PacketMath.h b/Eigen/src/Core/arch/AVX/PacketMath.h index a2306fd1a..4dda35b41 100644 --- a/Eigen/src/Core/arch/AVX/PacketMath.h +++ b/Eigen/src/Core/arch/AVX/PacketMath.h @@ -100,9 +100,9 @@ template<> struct packet_traits<int> : default_packet_traits }; */ -template<> struct unpacket_traits<Packet8f> { typedef float type; typedef Packet4f half; enum {size=8}; }; -template<> struct unpacket_traits<Packet4d> { typedef double type; typedef Packet2d half; enum {size=4}; }; -template<> struct unpacket_traits<Packet8i> { typedef int type; typedef Packet4i half; enum {size=8}; }; +template<> struct unpacket_traits<Packet8f> { typedef float type; typedef Packet4f half; enum {size=8, alignment=Aligned32}; }; +template<> struct unpacket_traits<Packet4d> { typedef double type; typedef Packet2d half; enum {size=4, alignment=Aligned32}; }; +template<> struct unpacket_traits<Packet8i> { typedef int type; typedef Packet4i half; enum {size=8, alignment=Aligned32}; }; template<> EIGEN_STRONG_INLINE Packet8f pset1<Packet8f>(const float& from) { return _mm256_set1_ps(from); } template<> EIGEN_STRONG_INLINE Packet4d pset1<Packet4d>(const double& from) { return _mm256_set1_pd(from); } diff --git a/Eigen/src/Core/arch/AltiVec/Complex.h b/Eigen/src/Core/arch/AltiVec/Complex.h index 1f00e27d8..58c296171 100644 --- a/Eigen/src/Core/arch/AltiVec/Complex.h +++ b/Eigen/src/Core/arch/AltiVec/Complex.h @@ -53,7 +53,7 @@ template<> struct packet_traits<std::complex<float> > : default_packet_traits }; }; -template<> struct unpacket_traits<Packet2cf> { typedef std::complex<float> type; enum {size=2}; typedef Packet2cf half; }; +template<> struct unpacket_traits<Packet2cf> { typedef std::complex<float> type; enum {size=2, alignment=Aligned16}; typedef Packet2cf half; }; template<> EIGEN_STRONG_INLINE Packet2cf pset1<Packet2cf>(const std::complex<float>& from) { @@ -275,7 +275,7 @@ template<> struct packet_traits<std::complex<double> > : default_packet_traits }; }; -template<> struct unpacket_traits<Packet1cd> { typedef std::complex<double> type; enum {size=1}; typedef Packet1cd half; }; +template<> struct unpacket_traits<Packet1cd> { typedef std::complex<double> type; enum {size=1, alignment=Aligned16}; typedef Packet1cd half; }; template<> EIGEN_STRONG_INLINE Packet1cd pload <Packet1cd>(const std::complex<double>* from) { EIGEN_DEBUG_ALIGNED_LOAD return Packet1cd(pload<Packet2d>((const double*)from)); } template<> EIGEN_STRONG_INLINE Packet1cd ploadu<Packet1cd>(const std::complex<double>* from) { EIGEN_DEBUG_UNALIGNED_LOAD return Packet1cd(ploadu<Packet2d>((const double*)from)); } diff --git a/Eigen/src/Core/arch/AltiVec/PacketMath.h b/Eigen/src/Core/arch/AltiVec/PacketMath.h index d647427ce..7ab777009 100755 --- a/Eigen/src/Core/arch/AltiVec/PacketMath.h +++ b/Eigen/src/Core/arch/AltiVec/PacketMath.h @@ -148,8 +148,8 @@ template<> struct packet_traits<int> : default_packet_traits }; -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<> struct unpacket_traits<Packet4f> { typedef float type; enum {size=4, alignment=Aligned16}; typedef Packet4f half; }; +template<> struct unpacket_traits<Packet4i> { typedef int type; enum {size=4, alignment=Aligned16}; typedef Packet4i half; }; inline std::ostream & operator <<(std::ostream & s, const Packet16uc & v) { @@ -756,7 +756,7 @@ template<> struct packet_traits<double> : default_packet_traits }; }; -template<> struct unpacket_traits<Packet2d> { typedef double type; enum {size=2}; typedef Packet2d half; }; +template<> struct unpacket_traits<Packet2d> { typedef double type; enum {size=2, alignment=Aligned16}; typedef Packet2d half; }; inline std::ostream & operator <<(std::ostream & s, const Packet2d & v) diff --git a/Eigen/src/Core/arch/CUDA/PacketMath.h b/Eigen/src/Core/arch/CUDA/PacketMath.h index ceed1d1ef..8b767a4e7 100644 --- a/Eigen/src/Core/arch/CUDA/PacketMath.h +++ b/Eigen/src/Core/arch/CUDA/PacketMath.h @@ -65,8 +65,8 @@ template<> struct packet_traits<double> : default_packet_traits }; -template<> struct unpacket_traits<float4> { typedef float type; enum {size=4}; typedef float4 half; }; -template<> struct unpacket_traits<double2> { typedef double type; enum {size=2}; typedef double2 half; }; +template<> struct unpacket_traits<float4> { typedef float type; enum {size=4, alignment=Aligned16}; typedef float4 half; }; +template<> struct unpacket_traits<double2> { typedef double type; enum {size=2, alignment=Aligned16}; typedef double2 half; }; template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE float4 pset1<float4>(const float& from) { return make_float4(from, from, from, from); diff --git a/Eigen/src/Core/arch/NEON/Complex.h b/Eigen/src/Core/arch/NEON/Complex.h index 60ab56a47..d2322b307 100644 --- a/Eigen/src/Core/arch/NEON/Complex.h +++ b/Eigen/src/Core/arch/NEON/Complex.h @@ -48,7 +48,7 @@ template<> struct packet_traits<std::complex<float> > : default_packet_traits }; }; -template<> struct unpacket_traits<Packet2cf> { typedef std::complex<float> type; enum {size=2}; typedef Packet2cf half; }; +template<> struct unpacket_traits<Packet2cf> { typedef std::complex<float> type; enum {size=2, alignment=Aligned16}; typedef Packet2cf half; }; template<> EIGEN_STRONG_INLINE Packet2cf pset1<Packet2cf>(const std::complex<float>& from) { @@ -306,7 +306,7 @@ template<> struct packet_traits<std::complex<double> > : default_packet_traits }; }; -template<> struct unpacket_traits<Packet1cd> { typedef std::complex<double> type; enum {size=1}; typedef Packet1cd half; }; +template<> struct unpacket_traits<Packet1cd> { typedef std::complex<double> type; enum {size=1, alignment=Aligned16}; typedef Packet1cd half; }; template<> EIGEN_STRONG_INLINE Packet1cd pload<Packet1cd>(const std::complex<double>* from) { EIGEN_DEBUG_ALIGNED_LOAD return Packet1cd(pload<Packet2d>((const double*)from)); } template<> EIGEN_STRONG_INLINE Packet1cd ploadu<Packet1cd>(const std::complex<double>* from) { EIGEN_DEBUG_UNALIGNED_LOAD return Packet1cd(ploadu<Packet2d>((const double*)from)); } diff --git a/Eigen/src/Core/arch/NEON/PacketMath.h b/Eigen/src/Core/arch/NEON/PacketMath.h index cda9056a4..a72f91f21 100644 --- a/Eigen/src/Core/arch/NEON/PacketMath.h +++ b/Eigen/src/Core/arch/NEON/PacketMath.h @@ -114,8 +114,8 @@ 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}; typedef Packet4f half; }; -template<> struct unpacket_traits<Packet4i> { typedef int type; enum {size=4}; typedef Packet4i half; }; +template<> struct unpacket_traits<Packet4f> { typedef float type; enum {size=4, alignment=Aligned16}; typedef Packet4f half; }; +template<> struct unpacket_traits<Packet4i> { typedef int type; enum {size=4, alignment=Aligned16}; 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); } @@ -570,7 +570,7 @@ template<> struct packet_traits<double> : default_packet_traits }; }; -template<> struct unpacket_traits<Packet2d> { typedef double type; enum {size=2}; typedef Packet2d half; }; +template<> struct unpacket_traits<Packet2d> { typedef double type; enum {size=2, alignment=Aligned16}; typedef Packet2d half; }; template<> EIGEN_STRONG_INLINE Packet2d pset1<Packet2d>(const double& from) { return vdupq_n_f64(from); } diff --git a/Eigen/src/Core/arch/SSE/Complex.h b/Eigen/src/Core/arch/SSE/Complex.h index 3713b73da..2a44b6272 100644 --- a/Eigen/src/Core/arch/SSE/Complex.h +++ b/Eigen/src/Core/arch/SSE/Complex.h @@ -50,7 +50,7 @@ template<> struct packet_traits<std::complex<float> > : default_packet_traits }; #endif -template<> struct unpacket_traits<Packet2cf> { typedef std::complex<float> type; enum {size=2}; typedef Packet2cf half; }; +template<> struct unpacket_traits<Packet2cf> { typedef std::complex<float> type; enum {size=2, alignment=Aligned16}; typedef Packet2cf half; }; 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)); } @@ -297,7 +297,7 @@ template<> struct packet_traits<std::complex<double> > : default_packet_traits }; #endif -template<> struct unpacket_traits<Packet1cd> { typedef std::complex<double> type; enum {size=1}; typedef Packet1cd half; }; +template<> struct unpacket_traits<Packet1cd> { typedef std::complex<double> type; enum {size=1, alignment=Aligned16}; typedef Packet1cd half; }; template<> EIGEN_STRONG_INLINE Packet1cd padd<Packet1cd>(const Packet1cd& a, const Packet1cd& b) { return Packet1cd(_mm_add_pd(a.v,b.v)); } template<> EIGEN_STRONG_INLINE Packet1cd psub<Packet1cd>(const Packet1cd& a, const Packet1cd& b) { return Packet1cd(_mm_sub_pd(a.v,b.v)); } diff --git a/Eigen/src/Core/arch/SSE/PacketMath.h b/Eigen/src/Core/arch/SSE/PacketMath.h index 38a84273d..d5e852799 100755 --- a/Eigen/src/Core/arch/SSE/PacketMath.h +++ b/Eigen/src/Core/arch/SSE/PacketMath.h @@ -144,9 +144,9 @@ template<> struct packet_traits<int> : default_packet_traits }; }; -template<> struct unpacket_traits<Packet4f> { typedef float type; enum {size=4}; typedef Packet4f half; }; -template<> struct unpacket_traits<Packet2d> { typedef double type; enum {size=2}; typedef Packet2d half; }; -template<> struct unpacket_traits<Packet4i> { typedef int type; enum {size=4}; typedef Packet4i half; }; +template<> struct unpacket_traits<Packet4f> { typedef float type; enum {size=4, alignment=Aligned16}; typedef Packet4f half; }; +template<> struct unpacket_traits<Packet2d> { typedef double type; enum {size=2, alignment=Aligned16}; typedef Packet2d half; }; +template<> struct unpacket_traits<Packet4i> { typedef int type; enum {size=4, alignment=Aligned16}; typedef Packet4i half; }; #if EIGEN_COMP_MSVC==1500 // Workaround MSVC 9 internal compiler error. |