diff options
| author |  Eugene Zhulenev <ezhulenev@google.com> | 2019-04-23 11:41:02 -0700 |
|---|---|---|
| committer | Eugene Zhulenev <ezhulenev@google.com> | 2019-04-23 11:41:02 -0700 |
| commit | 68a2a8c4454cfa653d1ad0ec9de729c43e6a151d (patch) | |
| tree | b3f7e97612143e19d552cd8e64221f0af3643cbe /Eigen/src/Core/arch/AVX | |
| parent | 8c7a6feb8e3b3328995ea033ccbcf7f817b9b41c (diff) | |
Use packet ops instead of AVX2 intrinsics
Diffstat (limited to 'Eigen/src/Core/arch/AVX')
| -rw-r--r-- | Eigen/src/Core/arch/AVX/PacketMath.h | 8 |
1 files changed, 4 insertions, 4 deletions
diff --git a/Eigen/src/Core/arch/AVX/PacketMath.h b/Eigen/src/Core/arch/AVX/PacketMath.h index 3f94f85eb..9d13895e3 100644 --- a/Eigen/src/Core/arch/AVX/PacketMath.h +++ b/Eigen/src/Core/arch/AVX/PacketMath.h @@ -387,10 +387,10 @@ template<> EIGEN_STRONG_INLINE Packet4d ploadu<Packet4d>(const double* from) { E template<> EIGEN_STRONG_INLINE Packet8i ploadu<Packet8i>(const int* from) { EIGEN_DEBUG_UNALIGNED_LOAD return _mm256_loadu_si256(reinterpret_cast<const __m256i*>(from)); } template<> EIGEN_STRONG_INLINE Packet8f ploadu<Packet8f>(const float* from, uint8_t umask) { - __m256i mask = _mm256_set1_epi8(static_cast<char>(umask)); - const __m256i bit_mask = _mm256_set_epi32(0xffffff7f, 0xffffffbf, 0xffffffdf, 0xffffffef, 0xfffffff7, 0xfffffffb, 0xfffffffd, 0xfffffffe); - mask = _mm256_or_si256(mask, bit_mask); - mask = _mm256_cmpeq_epi32(mask, _mm256_set1_epi32(0xffffffff)); + Packet8i mask = _mm256_set1_epi8(static_cast<char>(umask)); + const Packet8i bit_mask = _mm256_set_epi32(0xffffff7f, 0xffffffbf, 0xffffffdf, 0xffffffef, 0xfffffff7, 0xfffffffb, 0xfffffffd, 0xfffffffe); + mask = por<Packet8i>(mask, bit_mask); + mask = pcmp_eq<Packet8i>(mask, _mm256_set1_epi32(0xffffffff)); EIGEN_DEBUG_UNALIGNED_LOAD return _mm256_maskload_ps(from, mask); } |