diff options
| author |  Gael Guennebaud <g.gael@free.fr> | 2015-10-21 20:58:33 +0200 |
|---|---|---|
| committer | Gael Guennebaud <g.gael@free.fr> | 2015-10-21 20:58:33 +0200 |
| commit | e78bc111f1dc0a7af2360b836c94c33d67e55fc5 (patch) | |
| tree | 0c17ed598ca7faf949227f2dd92f189f6b0f0dcb | |
| parent | b178cc347968675bdae942dbdcb7de9ed9daa564 (diff) | |
bug #1090: fix a shortcoming in redux logic for which slice-vectorization plus unrolling might happen.
| -rw-r--r-- | Eigen/src/Core/Redux.h | 5 | ||||
| -rw-r--r-- | test/redux.cpp | 8 |
2 files changed, 11 insertions, 2 deletions
diff --git a/Eigen/src/Core/Redux.h b/Eigen/src/Core/Redux.h index 87b4a9c46..309898b36 100644 --- a/Eigen/src/Core/Redux.h +++ b/Eigen/src/Core/Redux.h @@ -269,8 +269,9 @@ struct redux_impl<Func, Derived, LinearVectorizedTraversal, NoUnrolling> } }; -template<typename Func, typename Derived> -struct redux_impl<Func, Derived, SliceVectorizedTraversal, NoUnrolling> +// NOTE: for SliceVectorizedTraversal we simply bypass unrolling +template<typename Func, typename Derived, int Unrolling> +struct redux_impl<Func, Derived, SliceVectorizedTraversal, Unrolling> { typedef typename Derived::Scalar Scalar; typedef typename packet_traits<Scalar>::type PacketType; diff --git a/test/redux.cpp b/test/redux.cpp index 9b0767c73..849faf55e 100644 --- a/test/redux.cpp +++ b/test/redux.cpp @@ -56,6 +56,14 @@ template<typename MatrixType> void matrixRedux(const MatrixType& m) VERIFY_IS_APPROX(m1_for_prod.block(r0,c0,r1,c1).prod(), m1_for_prod.block(r0,c0,r1,c1).eval().prod()); VERIFY_IS_APPROX(m1.block(r0,c0,r1,c1).real().minCoeff(), m1.block(r0,c0,r1,c1).real().eval().minCoeff()); VERIFY_IS_APPROX(m1.block(r0,c0,r1,c1).real().maxCoeff(), m1.block(r0,c0,r1,c1).real().eval().maxCoeff()); + + // regression for bug 1090 + const int R1 = MatrixType::RowsAtCompileTime>=2 ? MatrixType::RowsAtCompileTime/2 : 6; + const int C1 = MatrixType::ColsAtCompileTime>=2 ? MatrixType::ColsAtCompileTime/2 : 6; + if(R1<=rows-r0 && C1<=cols-c0) + { + VERIFY_IS_APPROX( (m1.template block<R1,C1>(r0,c0).sum()), m1.block(r0,c0,R1,C1).sum() ); + } // test empty objects VERIFY_IS_APPROX(m1.block(r0,c0,0,0).sum(), Scalar(0)); |