diff options
author | 2018-09-28 16:57:32 +0200 | |
---|---|---|
committer | 2018-09-28 16:57:32 +0200 | |
commit | 626942d9ddcc17c21c2d79a690537e54237275bc (patch) | |
tree | 82c20171f9b3a2f1bdaf2e822bf0a67d90e9c67c /Eigen | |
parent | e95696acb313a84b33a18cc300de418b05dc58e5 (diff) |
fix alignment issue in ploaddup for AVX512
Diffstat (limited to 'Eigen')
-rw-r--r-- | Eigen/src/Core/arch/AVX512/PacketMath.h | 4 |
1 files changed, 3 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; |