diff options
Diffstat (limited to 'Eigen/src/Core')
| -rw-r--r-- | Eigen/src/Core/arch/AVX512/PacketMath.h | 4 | ||||
| -rw-r--r-- | Eigen/src/Core/util/Macros.h | 18 |
2 files changed, 21 insertions, 1 deletions
diff --git a/Eigen/src/Core/arch/AVX512/PacketMath.h b/Eigen/src/Core/arch/AVX512/PacketMath.h index 9fbb256a1..e3f8c1c87 100644 --- a/Eigen/src/Core/arch/AVX512/PacketMath.h +++ b/Eigen/src/Core/arch/AVX512/PacketMath.h @@ -466,7 +466,9 @@ EIGEN_STRONG_INLINE Packet16i ploadu<Packet16i>(const int* from) { // {a0, a0 a1, a1, a2, a2, a3, a3, a4, a4, a5, a5, a6, a6, a7, a7} template <> EIGEN_STRONG_INLINE Packet16f ploaddup<Packet16f>(const float* from) { - __m256i low_half = _mm256_load_si256(reinterpret_cast<const __m256i*>(from)); + // an unaligned load is required here as there is no requirement + // on the alignment of input pointer 'from' + __m256i low_half = _mm256_loadu_si256(reinterpret_cast<const __m256i*>(from)); __m512 even_elements = _mm512_castsi512_ps(_mm512_cvtepu32_epi64(low_half)); __m512 pairs = _mm512_permute_ps(even_elements, _MM_SHUFFLE(2, 2, 0, 0)); return pairs; diff --git a/Eigen/src/Core/util/Macros.h b/Eigen/src/Core/util/Macros.h index 3af6c4e37..dd4866724 100644 --- a/Eigen/src/Core/util/Macros.h +++ b/Eigen/src/Core/util/Macros.h @@ -638,6 +638,15 @@ #endif #endif +#ifndef EIGEN_HAS_CXX11_OVERRIDE_FINAL + #if EIGEN_MAX_CPP_VER>=11 && \ + (__cplusplus >= 201103L || EIGEN_COMP_MSVC >= 1700) + #define EIGEN_HAS_CXX11_OVERRIDE_FINAL 1 + #else + #define EIGEN_HAS_CXX11_OVERRIDE_FINAL 0 + #endif +#endif + #if defined(EIGEN_CUDACC) && EIGEN_HAS_CONSTEXPR // While available already with c++11, this is useful mostly starting with c++14 and relaxed constexpr rules #if defined(__NVCC__) @@ -1095,6 +1104,15 @@ bool all(T t, Ts ... ts){ return t && all(ts...); } } #endif +#if EIGEN_HAS_CXX11_OVERRIDE_FINAL +// provide override and final specifiers if they are available: +# define EIGEN_OVERRIDE override +# define EIGEN_FINAL final +#else +# define EIGEN_OVERRIDE +# define EIGEN_FINAL +#endif + // Wrapping #pragma unroll in a macro since it is required for SYCL #if defined(__SYCL_DEVICE_ONLY__) #if defined(_MSC_VER) |