diff options
author | 2010-04-24 00:58:44 +0300 | |
---|---|---|
committer | 2010-04-24 00:58:44 +0300 | |
commit | 9337f371d2b08e83a56df394c838305b8c8642d8 (patch) | |
tree | 3863bc8ca18553ea0d7ac34e8fab3aa5eda682f4 /Eigen/src/Core/arch | |
parent | 5acf46bd12edb79e9a83c42810654dec88227726 (diff) |
(proper commit this time)
replaced _mm_prefetch in GeneralBlockPanelKernel.h, with ei_prefetch() inline function.
Implemented NEON and AltiVec versions, copied SSE version over from GeneralBlockPanelKernel.h.
Also in GCC case (or rather !_MSC_VER) it's implemented using __builtin_prefetch().
NEON managed to give a small but welcome boost, 0.88GFLOPS -> 0.91GFLOPS.
Diffstat (limited to 'Eigen/src/Core/arch')
-rw-r--r-- | Eigen/src/Core/arch/AltiVec/PacketMath.h | 9 | ||||
-rw-r--r-- | Eigen/src/Core/arch/NEON/PacketMath.h | 7 | ||||
-rw-r--r-- | Eigen/src/Core/arch/SSE/PacketMath.h | 4 |
3 files changed, 18 insertions, 2 deletions
diff --git a/Eigen/src/Core/arch/AltiVec/PacketMath.h b/Eigen/src/Core/arch/AltiVec/PacketMath.h index c6fc670d8..0a7b07645 100644 --- a/Eigen/src/Core/arch/AltiVec/PacketMath.h +++ b/Eigen/src/Core/arch/AltiVec/PacketMath.h @@ -67,6 +67,8 @@ typedef __vector unsigned char Packet16uc; #define _EIGEN_DECLARE_CONST_Packet4i(NAME,X) \ Packet4i ei_p4i_##NAME = ei_pset1<int>(X) +#define DST_CHAN 1 +#define DST_CTRL(size, count, stride) (((size) << 24) | ((count) << 16) | (stride)) // Define global static constants: static Packet4f ei_p4f_COUNTDOWN = { 3.0, 2.0, 1.0, 0.0 }; @@ -291,8 +293,8 @@ template<> EIGEN_STRONG_INLINE void ei_pstoreu<float>(float* to, const Packet4f edgeAlign = vec_lvsl(0, to); // permute map to extract edges edges=vec_perm(LSQ,MSQ,edgeAlign); // extract the edges align = vec_lvsr( 0, to ); // permute map to misalign data - MSQ = vec_perm(edges,(Packet16uc)from,align); // misalign the data (MSQ) - LSQ = vec_perm((Packet16uc)from,edges,align); // misalign the data (LSQ) + MSQ = vec_perm(edges,(Packet16uc)from,align); // misalign the data (MSQ) + LSQ = vec_perm((Packet16uc)from,edges,align); // misalign the data (LSQ) vec_st( LSQ, 15, (unsigned char *)to ); // Store the LSQ part first vec_st( MSQ, 0, (unsigned char *)to ); // Store the MSQ part } @@ -315,6 +317,9 @@ template<> EIGEN_STRONG_INLINE void ei_pstoreu<int>(int* to, const Packet4i vec_st( MSQ, 0, (unsigned char *)to ); // Store the MSQ part } +template<> EIGEN_STRONG_INLINE void ei_prefetch<float>(const float* addr) { vec_dstt(addr, DST_CTRL(2,2,32), DST_CHAN); } +template<> EIGEN_STRONG_INLINE void ei_prefetch<int>(const int* addr) { vec_dstt(addr, DST_CTRL(2,2,32), DST_CHAN); } + template<> EIGEN_STRONG_INLINE float ei_pfirst<Packet4f>(const Packet4f& a) { float EIGEN_ALIGN16 x[4]; vec_st(a, 0, x); return x[0]; } template<> EIGEN_STRONG_INLINE int ei_pfirst<Packet4i>(const Packet4i& a) { int EIGEN_ALIGN16 x[4]; vec_st(a, 0, x); return x[0]; } diff --git a/Eigen/src/Core/arch/NEON/PacketMath.h b/Eigen/src/Core/arch/NEON/PacketMath.h index 2acb3633a..96c75101c 100644 --- a/Eigen/src/Core/arch/NEON/PacketMath.h +++ b/Eigen/src/Core/arch/NEON/PacketMath.h @@ -53,6 +53,10 @@ typedef int32x4_t Packet4i; #define _EIGEN_DECLARE_CONST_Packet4i(NAME,X) \ const Packet4i ei_p4i_##NAME = ei_pset1<int>(X) +#ifndef __pld +#define __pld(x) asm volatile ( " pld [%[addr]]\n" :: [addr] "r" (x) : "cc" ); +#endif + template<> struct ei_packet_traits<float> : ei_default_packet_traits { typedef Packet4f type; enum {size=4}; @@ -168,6 +172,9 @@ template<> EIGEN_STRONG_INLINE void ei_pstore<int>(int* to, const Packet4i template<> EIGEN_STRONG_INLINE void ei_pstoreu<float>(float* to, const Packet4f& from) { EIGEN_DEBUG_UNALIGNED_STORE vst1q_f32(to, from); } template<> EIGEN_STRONG_INLINE void ei_pstoreu<int>(int* to, const Packet4i& from) { EIGEN_DEBUG_UNALIGNED_STORE vst1q_s32(to, from); } +template<> EIGEN_STRONG_INLINE void ei_prefetch<float>(const float* addr) { __pld(addr); } +template<> EIGEN_STRONG_INLINE void ei_prefetch<int>(const int* addr) { __pld(addr); } + // FIXME only store the 2 first elements ? template<> EIGEN_STRONG_INLINE float ei_pfirst<Packet4f>(const Packet4f& a) { float EIGEN_ALIGN16 x[4]; vst1q_f32(x, a); return x[0]; } template<> EIGEN_STRONG_INLINE int ei_pfirst<Packet4i>(const Packet4i& a) { int EIGEN_ALIGN16 x[4]; vst1q_s32(x, a); return x[0]; } diff --git a/Eigen/src/Core/arch/SSE/PacketMath.h b/Eigen/src/Core/arch/SSE/PacketMath.h index 77f15d982..d360081cd 100644 --- a/Eigen/src/Core/arch/SSE/PacketMath.h +++ b/Eigen/src/Core/arch/SSE/PacketMath.h @@ -233,6 +233,10 @@ template<> EIGEN_STRONG_INLINE void ei_pstoreu<double>(double* to, const Packet2 template<> EIGEN_STRONG_INLINE void ei_pstoreu<float>(float* to, const Packet4f& from) { EIGEN_DEBUG_UNALIGNED_STORE ei_pstoreu((double*)to, _mm_castps_pd(from)); } template<> EIGEN_STRONG_INLINE void ei_pstoreu<int>(int* to, const Packet4i& from) { EIGEN_DEBUG_UNALIGNED_STORE ei_pstoreu((double*)to, _mm_castsi128_pd(from)); } +template<> EIGEN_STRONG_INLINE void ei_prefetch<float>(const float* addr) { _mm_prefetch((const char*)(addr), _MM_HINT_T0); } +template<> EIGEN_STRONG_INLINE void ei_prefetch<double>(const double* addr) { _mm_prefetch((const char*)(addr), _MM_HINT_T0); } +template<> EIGEN_STRONG_INLINE void ei_prefetch<int>(const int* addr) { _mm_prefetch((const char*)(addr), _MM_HINT_T0); } + #if defined(_MSC_VER) && (_MSC_VER <= 1500) && defined(_WIN64) // The temporary variable fixes an internal compilation error. // Direct of the struct members fixed bug #62. |