GitHub-Proxy/eigen
1
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2025-08-12 11:49:02 +08:00
Code Issues Packages Projects Releases Wiki Activity

Enable the use of half-packet in coeff-based product.

Browse Source 
For instance, Matrix4f*Vector4f is now vectorized again when using AVX.
This commit is contained in:
Gael Guennebaud 2016-04-12 23:03:03 +02:00
parent e3a184785c
commit b67c983291
1 changed files with 21 additions and 15 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

24
Eigen/src/Core/ProductEvaluators.h
View File

@ -410,8 +410,6 @@ struct product_evaluator<Product<Lhs, Rhs, LazyProduct>, ProductTag, DenseShape,
typedef Product<Lhs, Rhs, LazyProduct> XprType; typedef Product<Lhs, Rhs, LazyProduct> XprType;
typedef typename XprType::Scalar Scalar; typedef typename XprType::Scalar Scalar;
typedef typename XprType::CoeffReturnType CoeffReturnType; typedef typename XprType::CoeffReturnType CoeffReturnType;
typedef typename XprType::PacketScalar PacketScalar;
typedef typename XprType::PacketReturnType PacketReturnType;
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
explicit product_evaluator(const XprType& xpr) explicit product_evaluator(const XprType& xpr)
@ -443,9 +441,13 @@ struct product_evaluator<Product<Lhs, Rhs, LazyProduct>, ProductTag, DenseShape,
ColsAtCompileTime = RhsNestedCleaned::ColsAtCompileTime, ColsAtCompileTime = RhsNestedCleaned::ColsAtCompileTime,
InnerSize = EIGEN_SIZE_MIN_PREFER_FIXED(LhsNestedCleaned::ColsAtCompileTime, RhsNestedCleaned::RowsAtCompileTime), InnerSize = EIGEN_SIZE_MIN_PREFER_FIXED(LhsNestedCleaned::ColsAtCompileTime, RhsNestedCleaned::RowsAtCompileTime),
MaxRowsAtCompileTime = LhsNestedCleaned::MaxRowsAtCompileTime, MaxRowsAtCompileTime = LhsNestedCleaned::MaxRowsAtCompileTime,
MaxColsAtCompileTime = RhsNestedCleaned::MaxColsAtCompileTime, MaxColsAtCompileTime = RhsNestedCleaned::MaxColsAtCompileTime
};
PacketSize = packet_traits<Scalar>::size, typedef typename find_best_packet<Scalar,RowsAtCompileTime>::type LhsVecPacketType;
typedef typename find_best_packet<Scalar,ColsAtCompileTime>::type RhsVecPacketType;
enum {
LhsCoeffReadCost = LhsEtorType::CoeffReadCost, LhsCoeffReadCost = LhsEtorType::CoeffReadCost,
RhsCoeffReadCost = RhsEtorType::CoeffReadCost, RhsCoeffReadCost = RhsEtorType::CoeffReadCost,
@ -459,19 +461,23 @@ struct product_evaluator<Product<Lhs, Rhs, LazyProduct>, ProductTag, DenseShape,
LhsFlags = LhsEtorType::Flags, LhsFlags = LhsEtorType::Flags,
RhsFlags = RhsEtorType::Flags, RhsFlags = RhsEtorType::Flags,
LhsAlignment = LhsEtorType::Alignment,
RhsAlignment = RhsEtorType::Alignment,
LhsRowMajor = LhsFlags & RowMajorBit, LhsRowMajor = LhsFlags & RowMajorBit,
RhsRowMajor = RhsFlags & RowMajorBit, RhsRowMajor = RhsFlags & RowMajorBit,
LhsVecPacketSize = unpacket_traits<LhsVecPacketType>::size,
RhsVecPacketSize = unpacket_traits<RhsVecPacketType>::size,
//
LhsAlignment = EIGEN_PLAIN_ENUM_MIN(LhsEtorType::Alignment,LhsVecPacketSize*int(sizeof(typename LhsNestedCleaned::Scalar))),
RhsAlignment = EIGEN_PLAIN_ENUM_MIN(RhsEtorType::Alignment,RhsVecPacketSize*int(sizeof(typename RhsNestedCleaned::Scalar))),
SameType = is_same<typename LhsNestedCleaned::Scalar,typename RhsNestedCleaned::Scalar>::value, SameType = is_same<typename LhsNestedCleaned::Scalar,typename RhsNestedCleaned::Scalar>::value,
CanVectorizeRhs = RhsRowMajor && (RhsFlags & PacketAccessBit) CanVectorizeRhs = RhsRowMajor && (RhsFlags & PacketAccessBit)
&& (ColsAtCompileTime == Dynamic || ((ColsAtCompileTime % PacketSize) == 0) ), && (ColsAtCompileTime == Dynamic || ((ColsAtCompileTime % RhsVecPacketSize) == 0) ),
CanVectorizeLhs = (!LhsRowMajor) && (LhsFlags & PacketAccessBit) CanVectorizeLhs = (!LhsRowMajor) && (LhsFlags & PacketAccessBit)
&& (RowsAtCompileTime == Dynamic || ((RowsAtCompileTime % PacketSize) == 0) ), && (RowsAtCompileTime == Dynamic || ((RowsAtCompileTime % LhsVecPacketSize) == 0) ),
EvalToRowMajor = (MaxRowsAtCompileTime==1&&MaxColsAtCompileTime!=1) ? 1 EvalToRowMajor = (MaxRowsAtCompileTime==1&&MaxColsAtCompileTime!=1) ? 1
: (MaxColsAtCompileTime==1&&MaxRowsAtCompileTime!=1) ? 0 : (MaxColsAtCompileTime==1&&MaxRowsAtCompileTime!=1) ? 0