mirror/eigen
2
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2025-03-31 19:00:35 +08:00
Code Issues Packages Projects Releases Wiki Activity

bug #1158: PartialReduxExpr is a vector expression, and it thus must expose the LinearAccessBit flag

Browse Source
This commit is contained in:
Gael Guennebaud 2016-01-28 13:16:30 +01:00
parent 9bcadb7fd1
commit df15fbc452
2 changed files with 4 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
Eigen/src/Core/CoreEvaluators.h
View File

@ -994,7 +994,7 @@ struct evaluator<PartialReduxExpr<ArgType, MemberOp, Direction> >
CoeffReadCost = TraversalSize==Dynamic ? HugeCost
: TraversalSize * evaluator<ArgType>::CoeffReadCost + int(CostOpType::value),
Flags = (traits<XprType>::Flags&RowMajorBit) | (evaluator<ArgType>::Flags&(HereditaryBits&(~RowMajorBit))),
Flags = (traits<XprType>::Flags&RowMajorBit) | (evaluator<ArgType>::Flags&(HereditaryBits&(~RowMajorBit))) | LinearAccessBit,
Alignment = 0 // FIXME this will need to be improved once PartialReduxExpr is vectorized
};

3
test/vectorwiseop.cpp
View File

@ -210,6 +210,9 @@ template<typename MatrixType> void vectorwiseop_matrix(const MatrixType& m)
VERIFY_IS_APPROX(m1.cwiseAbs().colwise().maxCoeff(), m1.colwise().template lpNorm<Infinity>());
VERIFY_IS_APPROX(m1.cwiseAbs().rowwise().maxCoeff(), m1.rowwise().template lpNorm<Infinity>());
// regression for bug 1158
VERIFY_IS_APPROX(m1.cwiseAbs().colwise().sum().x(), m1.col(0).cwiseAbs().sum());
// test normalized
m2 = m1.colwise().normalized();
VERIFY_IS_APPROX(m2.col(c), m1.col(c).normalized());