GitHub-Proxy/eigen
1
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2025-08-13 20:26:03 +08:00
Code Issues Packages Projects Releases Wiki Activity

Removing unused API to fix compile error in TensorFlow due to

Browse Source 
AVX512VL, AVX512BW usage
This commit is contained in:
Anuj Rawat 2019-05-12 14:43:10 +00:00
parent 4ccd1ece92
commit ad372084f5
1 changed files with 1 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
Eigen/src/Core/arch/GPU/PacketMathHalf.h
View File

@ -567,7 +567,7 @@ struct packet_traits<half> : default_packet_traits {
}; };
template<> struct unpacket_traits<Packet16h> { typedef Eigen::half type; typedef uint16_t mask_t; enum {size=16, alignment=Aligned32, vectorizable=true, masked_load_available=true, masked_store_available=false}; typedef Packet16h half; }; template<> struct unpacket_traits<Packet16h> { typedef Eigen::half type; enum {size=16, alignment=Aligned32, vectorizable=true, masked_load_available=false, masked_store_available=false}; typedef Packet16h half; };
template<> EIGEN_STRONG_INLINE Packet16h pset1<Packet16h>(const Eigen::half& from) { template<> EIGEN_STRONG_INLINE Packet16h pset1<Packet16h>(const Eigen::half& from) {
Packet16h result; Packet16h result;
@ -591,14 +591,6 @@ template<> EIGEN_STRONG_INLINE Packet16h ploadu<Packet16h>(const Eigen::half* fr
return result; return result;
} }
template<> EIGEN_STRONG_INLINE Packet16h ploadu<Packet16h>(const Eigen::half* from,
uint16_t umask) {
__mmask16 mask = static_cast<__mmask16>(umask);
Packet16h result;
result.x = _mm256_maskz_loadu_epi16(mask, reinterpret_cast<const __m256i*>(from));
return result;
}
template<> EIGEN_STRONG_INLINE void pstore<half>(Eigen::half* to, const Packet16h& from) { template<> EIGEN_STRONG_INLINE void pstore<half>(Eigen::half* to, const Packet16h& from) {
// (void*) -> workaround clang warning: // (void*) -> workaround clang warning:
// cast from 'Eigen::half *' to '__m256i *' increases required alignment from 2 to 32 // cast from 'Eigen::half *' to '__m256i *' increases required alignment from 2 to 32