diff options
Diffstat (limited to 'Eigen/src/Core/arch/AVX/PacketMath.h')
-rw-r--r-- | Eigen/src/Core/arch/AVX/PacketMath.h | 8 |
1 files changed, 8 insertions, 0 deletions
diff --git a/Eigen/src/Core/arch/AVX/PacketMath.h b/Eigen/src/Core/arch/AVX/PacketMath.h index 9d13895e3..5011b98ea 100644 --- a/Eigen/src/Core/arch/AVX/PacketMath.h +++ b/Eigen/src/Core/arch/AVX/PacketMath.h @@ -431,6 +431,14 @@ template<> EIGEN_STRONG_INLINE void pstoreu<float>(float* to, const Packet8f& template<> EIGEN_STRONG_INLINE void pstoreu<double>(double* to, const Packet4d& from) { EIGEN_DEBUG_UNALIGNED_STORE _mm256_storeu_pd(to, from); } template<> EIGEN_STRONG_INLINE void pstoreu<int>(int* to, const Packet8i& from) { EIGEN_DEBUG_UNALIGNED_STORE _mm256_storeu_si256(reinterpret_cast<__m256i*>(to), from); } +template<> EIGEN_STRONG_INLINE void pstoreu<float>(float* to, const Packet8f& from, uint8_t umask) { + 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_STORE return _mm256_maskstore_ps(to, mask, from); +} + // NOTE: leverage _mm256_i32gather_ps and _mm256_i32gather_pd if AVX2 instructions are available // NOTE: for the record the following seems to be slower: return _mm256_i32gather_ps(from, _mm256_set1_epi32(stride), 4); template<> EIGEN_DEVICE_FUNC inline Packet8f pgather<float, Packet8f>(const float* from, Index stride) |