diff options
| author |  Benoit Steiner <benoit.steiner.goog@gmail.com> | 2014-04-14 10:44:17 -0700 |
|---|---|---|
| committer | Benoit Steiner <benoit.steiner.goog@gmail.com> | 2014-04-14 10:44:17 -0700 |
| commit | feaf7c7e6d01a4804cee5949a01ece1f8a46866f (patch) | |
| tree | a9a0d13174d4e1cf27333e2aef520d42a084bc26 /Eigen | |
| parent | 3b2321e3ab157b314286227e08d738d74bfbf5d1 (diff) | |
Optimized SSE unaligned loads and stores when compiling a 64bit target with a recent version of gcc (ie gcc 4.8).
Diffstat (limited to 'Eigen')
| -rwxr-xr-x | Eigen/src/Core/arch/SSE/PacketMath.h | 9 |
1 files changed, 8 insertions, 1 deletions
diff --git a/Eigen/src/Core/arch/SSE/PacketMath.h b/Eigen/src/Core/arch/SSE/PacketMath.h index ea05a3415..bc17726b4 100755 --- a/Eigen/src/Core/arch/SSE/PacketMath.h +++ b/Eigen/src/Core/arch/SSE/PacketMath.h @@ -271,14 +271,17 @@ template<> EIGEN_STRONG_INLINE Packet4i pload<Packet4i>(const int* from) { E // TODO: do the same for MSVC (ICC is compatible) // NOTE: with the code below, MSVC's compiler crashes! -#if defined(__GNUC__) && defined(__i386__) +#if defined(__GNUC__) && (defined(__i386__) || (defined(__x86_64) && EIGEN_GNUC_AT_LEAST(4, 8))) // bug 195: gcc/i386 emits weird x87 fldl/fstpl instructions for _mm_load_sd #define EIGEN_AVOID_CUSTOM_UNALIGNED_LOADS 1 + #define EIGEN_AVOID_CUSTOM_UNALIGNED_STORES 1 #elif defined(__clang__) // bug 201: Segfaults in __mm_loadh_pd with clang 2.8 #define EIGEN_AVOID_CUSTOM_UNALIGNED_LOADS 1 + #define EIGEN_AVOID_CUSTOM_UNALIGNED_STORES 0 #else #define EIGEN_AVOID_CUSTOM_UNALIGNED_LOADS 0 + #define EIGEN_AVOID_CUSTOM_UNALIGNED_STORES 0 #endif template<> EIGEN_STRONG_INLINE Packet4f ploadu<Packet4f>(const float* from) @@ -338,8 +341,12 @@ template<> EIGEN_STRONG_INLINE void pstore<int>(int* to, const Packet4i& f template<> EIGEN_STRONG_INLINE void pstoreu<double>(double* to, const Packet2d& from) { EIGEN_DEBUG_UNALIGNED_STORE +#if EIGEN_AVOID_CUSTOM_UNALIGNED_STORES + _mm_storeu_pd(to, from); +#else _mm_storel_pd((to), from); _mm_storeh_pd((to+1), from); +#endif } template<> EIGEN_STRONG_INLINE void pstoreu<float>(float* to, const Packet4f& from) { EIGEN_DEBUG_UNALIGNED_STORE pstoreu(reinterpret_cast<double*>(to), _mm_castps_pd(from)); } template<> EIGEN_STRONG_INLINE void pstoreu<int>(int* to, const Packet4i& from) { EIGEN_DEBUG_UNALIGNED_STORE pstoreu(reinterpret_cast<double*>(to), _mm_castsi128_pd(from)); } |