GitHub-Proxy/eigen
1
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2025-09-12 09:23:12 +08:00
Code Issues Packages Projects Releases Wiki Activity

Backport changes from old to new expression engines

Browse Source
This commit is contained in:
Gael Guennebaud 2014-06-20 16:17:57 +02:00
parent b29b81a1f4
commit 7fa87a8b12
4 changed files with 41 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
Eigen/src/Core/CoreEvaluators.h
View File

@ -658,7 +658,7 @@ struct evaluator<Map<PlainObjectType, MapOptions, StrideType> >
&& ( bool(IsDynamicSize)
|| HasNoOuterStride
|| ( OuterStrideAtCompileTime!=Dynamic
&& ((static_cast<int>(sizeof(Scalar))*OuterStrideAtCompileTime)%16)==0 ) ),
&& ((static_cast<int>(sizeof(Scalar))*OuterStrideAtCompileTime)%EIGEN_ALIGN_BYTES)==0 ) ),
Flags0 = evaluator<PlainObjectType>::Flags,
Flags1 = IsAligned ? (int(Flags0) | AlignedBit) : (int(Flags0) & ~AlignedBit),
Flags2 = (bool(HasNoStride) || bool(PlainObjectType::IsVectorAtCompileTime))
@ -723,8 +723,9 @@ struct evaluator<Block<ArgType, BlockRows, BlockCols, InnerPanel> >
MaskPacketAccessBit = (InnerSize == Dynamic || (InnerSize % packet_traits<Scalar>::size) == 0)
&& (InnerStrideAtCompileTime == 1)
? PacketAccessBit : 0,
MaskAlignedBit = (InnerPanel && (OuterStrideAtCompileTime!=Dynamic) && (((OuterStrideAtCompileTime * int(sizeof(Scalar))) % 16) == 0)) ? AlignedBit : 0,
FlagsLinearAccessBit = (RowsAtCompileTime == 1 || ColsAtCompileTime == 1) ? LinearAccessBit : 0,
MaskAlignedBit = (InnerPanel && (OuterStrideAtCompileTime!=Dynamic) && (((OuterStrideAtCompileTime * int(sizeof(Scalar))) % EIGEN_ALIGN_BYTES) == 0)) ? AlignedBit : 0,
FlagsLinearAccessBit = (RowsAtCompileTime == 1 || ColsAtCompileTime == 1 || (InnerPanel && (traits<XprType>::Flags&LinearAccessBit))) ? LinearAccessBit : 0,
FlagsRowMajorBit = XprType::Flags&RowMajorBit,
Flags0 = evaluator<ArgType>::Flags & ( (HereditaryBits & ~RowMajorBit) |
DirectAccessBit |
@ -825,7 +826,7 @@ struct block_evaluator<ArgType, BlockRows, BlockCols, InnerPanel, /* HasDirectAc
: mapbase_evaluator<XprType, typename XprType::PlainObject>(block)
{
// FIXME this should be an internal assertion
eigen_assert(EIGEN_IMPLIES(evaluator<XprType>::Flags&AlignedBit, (size_t(block.data()) % 16) == 0) && "data is not aligned");
eigen_assert(EIGEN_IMPLIES(evaluator<XprType>::Flags&AlignedBit, (size_t(block.data()) % EIGEN_ALIGN_BYTES) == 0) && "data is not aligned");
}
};

34
Eigen/src/Core/products/GeneralMatrixMatrix.h
View File

@ -471,6 +471,38 @@ struct generic_product_impl<Lhs,Rhs,DenseShape,DenseShape,GemmProduct>
MaxDepthAtCompileTime = EIGEN_SIZE_MIN_PREFER_FIXED(Lhs::MaxColsAtCompileTime,Rhs::MaxRowsAtCompileTime)
};
typedef generic_product_impl<Lhs,Rhs,DenseShape,DenseShape,CoeffBasedProductMode> lazyproduct;
template<typename Dst>
static void evalTo(Dst& dst, const Lhs& lhs, const Rhs& rhs)
{
if((rhs.rows()+dst.rows()+dst.cols())<20 && rhs.rows()>0)
lazyproduct::evalTo(dst, lhs, rhs);
else
{
dst.setZero();
scaleAndAddTo(dst, lhs, rhs, Scalar(1));
}
}
template<typename Dst>
static void addTo(Dst& dst, const Lhs& lhs, const Rhs& rhs)
{
if((rhs.rows()+dst.rows()+dst.cols())<20 && rhs.rows()>0)
lazyproduct::addTo(dst, lhs, rhs);
else
scaleAndAddTo(dst,lhs, rhs, Scalar(1));
}
template<typename Dst>
static void subTo(Dst& dst, const Lhs& lhs, const Rhs& rhs)
{
if((rhs.rows()+dst.rows()+dst.cols())<20 && rhs.rows()>0)
lazyproduct::subTo(dst, lhs, rhs);
else
scaleAndAddTo(dst, lhs, rhs, Scalar(-1));
}
template<typename Dest>
static void scaleAndAddTo(Dest& dst, const Lhs& a_lhs, const Rhs& a_rhs, const Scalar& alpha)
{
@ -494,7 +526,7 @@ struct generic_product_impl<Lhs,Rhs,DenseShape,DenseShape,GemmProduct>
(Dest::Flags&RowMajorBit) ? RowMajor : ColMajor>,
ActualLhsTypeCleaned, ActualRhsTypeCleaned, Dest, BlockingType> GemmFunctor;
BlockingType blocking(dst.rows(), dst.cols(), lhs.cols());
BlockingType blocking(dst.rows(), dst.cols(), lhs.cols(), true);
internal::parallelize_gemm<(Dest::MaxRowsAtCompileTime>32 || Dest::MaxRowsAtCompileTime==Dynamic)>
(GemmFunctor(lhs, rhs, dst, actualAlpha, blocking), a_lhs.rows(), a_rhs.cols(), Dest::Flags&RowMajorBit);

2
Eigen/src/Core/products/GeneralMatrixMatrixTriangular.h
View File

@ -266,6 +266,8 @@ template<typename MatrixType, unsigned int UpLo>
template<typename ProductType>
TriangularView<MatrixType,UpLo>& TriangularView<MatrixType,UpLo>::_assignProduct(const ProductType& prod, const Scalar& alpha)
{
eigen_assert(m_matrix.rows() == prod.rows() && m_matrix.cols() == prod.cols());
general_product_to_triangular_selector<MatrixType, ProductType, UpLo, internal::traits<ProductType>::InnerSize==1>::run(m_matrix.const_cast_derived(), prod, alpha);
return *this;

2
Eigen/src/Core/util/XprHelper.h
View File

@ -187,7 +187,7 @@ class compute_matrix_evaluator_flags
((Options&DontAlign)==0)
&& (
#if EIGEN_ALIGN_STATICALLY
((!is_dynamic_size_storage) && (((MaxCols*MaxRows*int(sizeof(Scalar))) % 16) == 0))
((!is_dynamic_size_storage) && (((MaxCols*MaxRows*int(sizeof(Scalar))) % EIGEN_ALIGN_BYTES) == 0))
#else
0
#endif