devtools/eigen
1
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2026-04-10 11:34:33 +08:00
Code Issues Packages Projects Releases Wiki Activity

Introducing "vectorized" byte on unpacket_traits structs

Browse Source 
This is a preparation to a change on gebp_traits, where a new template
argument will be introduced to dictate the packet size, so it won't be
bound to the current/max packet size only anymore.

By having packet types defined early on gebp_traits, one has now to
act on packet types, not scalars anymore, for the enum values defined
on that class. One approach for reaching the vectorizable/size
properties one needs there could be getting the packet's scalar again
with unpacket_traits<>, then the size/Vectorizable enum entries from
packet_traits<>. It turns out guards like "#ifndef
EIGEN_VECTORIZE_AVX512" at AVX/PacketMath.h will hide smaller packet
variations of packet_traits<> for some types (and it makes sense to
keep that). In other words, one can't go back to the scalar and create
a new PacketType, as this will always lead to the maximum packet type
for the architecture.

The less costly/invasive solution for that, thus, is to add the
vectorizable info on every unpacket_traits struct as well.
This commit is contained in:
Gustavo Lima Chaves
2018-12-19 14:24:44 -08:00
parent efa4c9c40f
commit e763fcd09e
18 changed files with 46 additions and 43 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
Eigen/src/Core/arch/SSE/Complex.h
View File

@@ -50,7 +50,7 @@ template<> struct packet_traits<std::complex<float> > : default_packet_traits
};
#endif
template<> struct unpacket_traits<Packet2cf> { typedef std::complex<float> type; enum {size=2, alignment=Aligned16}; typedef Packet2cf half; };
template<> struct unpacket_traits<Packet2cf> { typedef std::complex<float> type; enum {size=2, alignment=Aligned16, vectorizable=true}; typedef Packet2cf half; };
template<> EIGEN_STRONG_INLINE Packet2cf padd<Packet2cf>(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(_mm_add_ps(a.v,b.v)); }
template<> EIGEN_STRONG_INLINE Packet2cf psub<Packet2cf>(const Packet2cf& a, const Packet2cf& b) { return Packet2cf(_mm_sub_ps(a.v,b.v)); }
@@ -280,7 +280,7 @@ template<> struct packet_traits<std::complex<double> > : default_packet_traits
};
#endif
template<> struct unpacket_traits<Packet1cd> { typedef std::complex<double> type; enum {size=1, alignment=Aligned16}; typedef Packet1cd half; };
template<> struct unpacket_traits<Packet1cd> { typedef std::complex<double> type; enum {size=1, alignment=Aligned16, vectorizable=true}; typedef Packet1cd half; };
template<> EIGEN_STRONG_INLINE Packet1cd padd<Packet1cd>(const Packet1cd& a, const Packet1cd& b) { return Packet1cd(_mm_add_pd(a.v,b.v)); }
template<> EIGEN_STRONG_INLINE Packet1cd psub<Packet1cd>(const Packet1cd& a, const Packet1cd& b) { return Packet1cd(_mm_sub_pd(a.v,b.v)); }

6
Eigen/src/Core/arch/SSE/PacketMath.h
View File

@@ -166,17 +166,17 @@ template<> struct unpacket_traits<Packet4f> {
typedef float type;
typedef Packet4f half;
typedef Packet4i integer_packet;
enum {size=4, alignment=Aligned16};
enum {size=4, alignment=Aligned16, vectorizable=true};
};
template<> struct unpacket_traits<Packet2d> {
typedef double type;
typedef Packet2d half;
enum {size=2, alignment=Aligned16};
enum {size=2, alignment=Aligned16, vectorizable=true};
};
template<> struct unpacket_traits<Packet4i> {
typedef int type;
typedef Packet4i half;
enum {size=4, alignment=Aligned16};
enum {size=4, alignment=Aligned16, vectorizable=false};
};
#ifndef EIGEN_VECTORIZE_AVX