diff --git a/Eigen/src/Cholesky/LDLT.h b/Eigen/src/Cholesky/LDLT.h index e01ae82..abd30bd 100644 --- a/Eigen/src/Cholesky/LDLT.h +++ b/Eigen/src/Cholesky/LDLT.h @@ -464,7 +464,7 @@ LDLT& LDLT::compute(const MatrixType& a) */ template template -LDLT& LDLT::rankUpdate(const MatrixBase& w, const typename NumTraits::Real& sigma) +LDLT& LDLT::rankUpdate(const MatrixBase& w, const typename LDLT::RealScalar& sigma) { const Index size = w.rows(); if (m_isInitialized) diff --git a/Eigen/src/Core/Assign.h b/Eigen/src/Core/Assign.h index bcfc261..f481731 100644 --- a/Eigen/src/Core/Assign.h +++ b/Eigen/src/Core/Assign.h @@ -449,7 +449,7 @@ struct assign_impl::JointAlignment }; const Scalar *dst_ptr = &dst.coeffRef(0,0); - if((!bool(dstIsAligned)) && (Index(dst_ptr) % sizeof(Scalar))>0) + if((!bool(dstIsAligned)) && (size_t(dst_ptr) % sizeof(Scalar))>0) { // the pointer is not aligend-on scalar, so alignment is not possible return assign_impl::run(dst, src); diff --git a/Eigen/src/Core/DenseBase.h b/Eigen/src/Core/DenseBase.h index 32f60d8..dc20e54 100644 --- a/Eigen/src/Core/DenseBase.h +++ b/Eigen/src/Core/DenseBase.h @@ -183,10 +183,6 @@ template class DenseBase /** \returns the number of nonzero coefficients which is in practice the number * of stored coefficients. */ inline Index nonZeros() const { return size(); } - /** \returns true if either the number of rows or the number of columns is equal to 1. - * In other words, this function returns - * \code rows()==1 || cols()==1 \endcode - * \sa rows(), cols(), IsVectorAtCompileTime. */ /** \returns the outer size. * diff --git a/Eigen/src/Core/GeneralProduct.h b/Eigen/src/Core/GeneralProduct.h index 9e805a8..0eae529 100644 --- a/Eigen/src/Core/GeneralProduct.h +++ b/Eigen/src/Core/GeneralProduct.h @@ -257,7 +257,7 @@ template class GeneralProduct : public ProductBase, Lhs, Rhs> { - template struct IsRowMajor : internal::conditional<(int(T::Flags)&RowMajorBit), internal::true_type, internal::false_type>::type {}; + template struct is_row_major : internal::conditional<(int(T::Flags)&RowMajorBit), internal::true_type, internal::false_type>::type {}; public: EIGEN_PRODUCT_PUBLIC_INTERFACE(GeneralProduct) @@ -281,22 +281,22 @@ class GeneralProduct template inline void evalTo(Dest& dest) const { - internal::outer_product_selector_run(*this, dest, set(), IsRowMajor()); + internal::outer_product_selector_run(*this, dest, set(), is_row_major()); } template inline void addTo(Dest& dest) const { - internal::outer_product_selector_run(*this, dest, add(), IsRowMajor()); + internal::outer_product_selector_run(*this, dest, add(), is_row_major()); } template inline void subTo(Dest& dest) const { - internal::outer_product_selector_run(*this, dest, sub(), IsRowMajor()); + internal::outer_product_selector_run(*this, dest, sub(), is_row_major()); } template void scaleAndAddTo(Dest& dest, const Scalar& alpha) const { - internal::outer_product_selector_run(*this, dest, adds(alpha), IsRowMajor()); + internal::outer_product_selector_run(*this, dest, adds(alpha), is_row_major()); } }; diff --git a/Eigen/src/Core/MapBase.h b/Eigen/src/Core/MapBase.h index 92e1141..a9828f7 100644 --- a/Eigen/src/Core/MapBase.h +++ b/Eigen/src/Core/MapBase.h @@ -123,7 +123,7 @@ template class MapBase return internal::ploadt(m_data + index * innerStride()); } - inline MapBase(PointerType dataPtr) : m_data(dataPtr), m_rows(RowsAtCompileTime), m_cols(ColsAtCompileTime) + explicit inline MapBase(PointerType dataPtr) : m_data(dataPtr), m_rows(RowsAtCompileTime), m_cols(ColsAtCompileTime) { EIGEN_STATIC_ASSERT_FIXED_SIZE(Derived) checkSanity(); diff --git a/Eigen/src/Core/MathFunctions.h b/Eigen/src/Core/MathFunctions.h index 2bfc5eb..adf2f9c 100644 --- a/Eigen/src/Core/MathFunctions.h +++ b/Eigen/src/Core/MathFunctions.h @@ -294,7 +294,7 @@ struct hypot_impl RealScalar _x = abs(x); RealScalar _y = abs(y); RealScalar p = (max)(_x, _y); - if(p==RealScalar(0)) return 0; + if(p==RealScalar(0)) return RealScalar(0); RealScalar q = (min)(_x, _y); RealScalar qp = q/p; return p * sqrt(RealScalar(1) + qp*qp); diff --git a/Eigen/src/Core/PlainObjectBase.h b/Eigen/src/Core/PlainObjectBase.h index 8a3e454..ffd3a06 100644 --- a/Eigen/src/Core/PlainObjectBase.h +++ b/Eigen/src/Core/PlainObjectBase.h @@ -437,6 +437,22 @@ class PlainObjectBase : public internal::dense_xpr_base::type } #endif + /** Copy constructor */ + EIGEN_STRONG_INLINE PlainObjectBase(const PlainObjectBase& other) + : m_storage() + { + _check_template_params(); + lazyAssign(other); + } + + template + EIGEN_STRONG_INLINE PlainObjectBase(const DenseBase &other) + : m_storage() + { + _check_template_params(); + lazyAssign(other); + } + EIGEN_STRONG_INLINE PlainObjectBase(Index a_size, Index nbRows, Index nbCols) : m_storage(a_size, nbRows, nbCols) { diff --git a/Eigen/src/Core/Ref.h b/Eigen/src/Core/Ref.h index f53674c..7a3beca 100644 --- a/Eigen/src/Core/Ref.h +++ b/Eigen/src/Core/Ref.h @@ -244,6 +244,15 @@ template class Ref< // std::cout << int(StrideType::InnerStrideAtCompileTime) << " - " << int(Derived::InnerStrideAtCompileTime) << "\n"; construct(expr.derived(), typename Traits::template match::type()); } + + inline Ref(const Ref& other) : Base(other) { + // copy constructor shall not copy the m_object, to avoid unnecessary malloc and copy + } + + template + inline Ref(const RefBase& other) { + construct(other.derived(), typename Traits::template match::type()); + } protected: diff --git a/Eigen/src/Core/arch/NEON/PacketMath.h b/Eigen/src/Core/arch/NEON/PacketMath.h index 94dfab3..d49670e 100644 --- a/Eigen/src/Core/arch/NEON/PacketMath.h +++ b/Eigen/src/Core/arch/NEON/PacketMath.h @@ -384,6 +384,7 @@ template<> EIGEN_STRONG_INLINE int predux_max(const Packet4i& a) a_lo = vget_low_s32(a); a_hi = vget_high_s32(a); max = vpmax_s32(a_lo, a_hi); + max = vpmax_s32(max, max); return vget_lane_s32(max, 0); } diff --git a/Eigen/src/Core/products/TriangularMatrixMatrix_MKL.h b/Eigen/src/Core/products/TriangularMatrixMatrix_MKL.h index ba41a1c..4cc56a4 100644 --- a/Eigen/src/Core/products/TriangularMatrixMatrix_MKL.h +++ b/Eigen/src/Core/products/TriangularMatrixMatrix_MKL.h @@ -109,7 +109,7 @@ struct product_triangular_matrix_matrix_trmm #define EIGEN_MKL_VML_THRESHOLD 128 +/* MKL_DOMAIN_BLAS, etc are defined only in 10.3 update 7 */ +/* MKL_BLAS, etc are not defined in 11.2 */ +#ifdef MKL_DOMAIN_ALL +#define EIGEN_MKL_DOMAIN_ALL MKL_DOMAIN_ALL +#else +#define EIGEN_MKL_DOMAIN_ALL MKL_ALL +#endif + +#ifdef MKL_DOMAIN_BLAS +#define EIGEN_MKL_DOMAIN_BLAS MKL_DOMAIN_BLAS +#else +#define EIGEN_MKL_DOMAIN_BLAS MKL_BLAS +#endif + +#ifdef MKL_DOMAIN_FFT +#define EIGEN_MKL_DOMAIN_FFT MKL_DOMAIN_FFT +#else +#define EIGEN_MKL_DOMAIN_FFT MKL_FFT +#endif + +#ifdef MKL_DOMAIN_VML +#define EIGEN_MKL_DOMAIN_VML MKL_DOMAIN_VML +#else +#define EIGEN_MKL_DOMAIN_VML MKL_VML +#endif + +#ifdef MKL_DOMAIN_PARDISO +#define EIGEN_MKL_DOMAIN_PARDISO MKL_DOMAIN_PARDISO +#else +#define EIGEN_MKL_DOMAIN_PARDISO MKL_PARDISO +#endif + namespace Eigen { typedef std::complex dcomplex; diff --git a/Eigen/src/Core/util/Macros.h b/Eigen/src/Core/util/Macros.h index f69970f..42671e8 100644 --- a/Eigen/src/Core/util/Macros.h +++ b/Eigen/src/Core/util/Macros.h @@ -13,7 +13,7 @@ #define EIGEN_WORLD_VERSION 3 #define EIGEN_MAJOR_VERSION 2 -#define EIGEN_MINOR_VERSION 5 +#define EIGEN_MINOR_VERSION 6 #define EIGEN_VERSION_AT_LEAST(x,y,z) (EIGEN_WORLD_VERSION>x || (EIGEN_WORLD_VERSION>=x && \ (EIGEN_MAJOR_VERSION>y || (EIGEN_MAJOR_VERSION>=y && \ @@ -314,7 +314,7 @@ namespace Eigen { // just an empty macro ! #define EIGEN_EMPTY -#if defined(_MSC_VER) && (_MSC_VER < 1800) && (!defined(__INTEL_COMPILER)) +#if defined(_MSC_VER) && (_MSC_VER < 1900) && (!defined(__INTEL_COMPILER)) #define EIGEN_INHERIT_ASSIGNMENT_EQUAL_OPERATOR(Derived) \ using Base::operator =; #elif defined(__clang__) // workaround clang bug (see http://forum.kde.org/viewtopic.php?f=74&t=102653) diff --git a/Eigen/src/Eigenvalues/SelfAdjointEigenSolver.h b/Eigen/src/Eigenvalues/SelfAdjointEigenSolver.h index c2e76c8..1131c8a 100644 --- a/Eigen/src/Eigenvalues/SelfAdjointEigenSolver.h +++ b/Eigen/src/Eigenvalues/SelfAdjointEigenSolver.h @@ -80,6 +80,8 @@ template class SelfAdjointEigenSolver /** \brief Scalar type for matrices of type \p _MatrixType. */ typedef typename MatrixType::Scalar Scalar; typedef typename MatrixType::Index Index; + + typedef Matrix EigenvectorsType; /** \brief Real scalar type for \p _MatrixType. * @@ -225,7 +227,7 @@ template class SelfAdjointEigenSolver * * \sa eigenvalues() */ - const MatrixType& eigenvectors() const + const EigenvectorsType& eigenvectors() const { eigen_assert(m_isInitialized && "SelfAdjointEigenSolver is not initialized."); eigen_assert(m_eigenvectorsOk && "The eigenvectors have not been computed together with the eigenvalues."); @@ -356,7 +358,7 @@ template class SelfAdjointEigenSolver EIGEN_STATIC_ASSERT_NON_INTEGER(Scalar); } - MatrixType m_eivec; + EigenvectorsType m_eivec; RealVectorType m_eivalues; typename TridiagonalizationType::SubDiagonalType m_subdiag; ComputationInfo m_info; @@ -381,7 +383,7 @@ template class SelfAdjointEigenSolver * "implicit symmetric QR step with Wilkinson shift" */ namespace internal { -template +template static void tridiagonal_qr_step(RealScalar* diag, RealScalar* subdiag, Index start, Index end, Scalar* matrixQ, Index n); } @@ -413,7 +415,7 @@ SelfAdjointEigenSolver& SelfAdjointEigenSolver // declare some aliases RealVectorType& diag = m_eivalues; - MatrixType& mat = m_eivec; + EigenvectorsType& mat = m_eivec; // map the matrix coefficients to [-1:1] to avoid over- and underflow. mat = matrix.template triangularView(); @@ -449,7 +451,7 @@ SelfAdjointEigenSolver& SelfAdjointEigenSolver while (start>0 && m_subdiag[start-1]!=0) start--; - internal::tridiagonal_qr_step(diag.data(), m_subdiag.data(), start, end, computeEigenvectors ? m_eivec.data() : (Scalar*)0, n); + internal::tridiagonal_qr_step(diag.data(), m_subdiag.data(), start, end, computeEigenvectors ? m_eivec.data() : (Scalar*)0, n); } if (iter <= m_maxIterations * n) @@ -498,6 +500,7 @@ template struct direct_selfadjoint_eigenvalues struct direct_selfadjoint_eigenvalues struct direct_selfadjoint_eigenvalues struct direct_selfadjoint_eigenvalues& SelfAdjointEigenSolver } namespace internal { -template +template static void tridiagonal_qr_step(RealScalar* diag, RealScalar* subdiag, Index start, Index end, Scalar* matrixQ, Index n) { using std::abs; @@ -784,8 +788,7 @@ static void tridiagonal_qr_step(RealScalar* diag, RealScalar* subdiag, Index sta // apply the givens rotation to the unit matrix Q = Q * G if (matrixQ) { - // FIXME if StorageOrder == RowMajor this operation is not very efficient - Map > q(matrixQ,n,n); + Map > q(matrixQ,n,n); q.applyOnTheRight(k,k+1,rot); } } diff --git a/Eigen/src/Geometry/AlignedBox.h b/Eigen/src/Geometry/AlignedBox.h index b226336..7e1cd9e 100644 --- a/Eigen/src/Geometry/AlignedBox.h +++ b/Eigen/src/Geometry/AlignedBox.h @@ -163,7 +163,7 @@ EIGEN_MAKE_ALIGNED_OPERATOR_NEW_IF_VECTORIZABLE_FIXED_SIZE(_Scalar,_AmbientDim) * a uniform distribution */ inline VectorType sample() const { - VectorType r; + VectorType r(dim()); for(Index d=0; d Identity() { return Quaternion(1, 0, 0, 0); } + static inline Quaternion Identity() { return Quaternion(Scalar(1), Scalar(0), Scalar(0), Scalar(0)); } /** \sa QuaternionBase::Identity(), MatrixBase::setIdentity() */ - inline QuaternionBase& setIdentity() { coeffs() << 0, 0, 0, 1; return *this; } + inline QuaternionBase& setIdentity() { coeffs() << Scalar(0), Scalar(0), Scalar(0), Scalar(1); return *this; } /** \returns the squared norm of the quaternion's coefficients * \sa QuaternionBase::norm(), MatrixBase::squaredNorm() diff --git a/Eigen/src/IterativeLinearSolvers/BasicPreconditioners.h b/Eigen/src/IterativeLinearSolvers/BasicPreconditioners.h index 73ca9bf..1f3c060 100644 --- a/Eigen/src/IterativeLinearSolvers/BasicPreconditioners.h +++ b/Eigen/src/IterativeLinearSolvers/BasicPreconditioners.h @@ -65,10 +65,10 @@ class DiagonalPreconditioner { typename MatType::InnerIterator it(mat,j); while(it && it.index()!=j) ++it; - if(it && it.index()==j) + if(it && it.index()==j && it.value()!=Scalar(0)) m_invdiag(j) = Scalar(1)/it.value(); else - m_invdiag(j) = 0; + m_invdiag(j) = Scalar(1); } m_isInitialized = true; return *this; diff --git a/Eigen/src/OrderingMethods/Amd.h b/Eigen/src/OrderingMethods/Amd.h index 1c28bdf..70550b8 100644 --- a/Eigen/src/OrderingMethods/Amd.h +++ b/Eigen/src/OrderingMethods/Amd.h @@ -137,9 +137,6 @@ void minimum_degree_ordering(SparseMatrix& C, Permutation degree[i] = len[i]; // degree of node i } mark = internal::cs_wclear(0, 0, w, n); /* clear w */ - elen[n] = -2; /* n is a dead element */ - Cp[n] = -1; /* n is a root of assembly tree */ - w[n] = 0; /* n is a dead element */ /* --- Initialize degree lists ------------------------------------------ */ for(i = 0; i < n; i++) @@ -153,7 +150,7 @@ void minimum_degree_ordering(SparseMatrix& C, Permutation } d = degree[i]; - if(d == 1) /* node i is empty */ + if(d == 1 && has_diag) /* node i is empty */ { elen[i] = -2; /* element i is dead */ nel++; diff --git a/Eigen/src/SparseCore/SparseVector.h b/Eigen/src/SparseCore/SparseVector.h index 7e15c81..49865d0 100644 --- a/Eigen/src/SparseCore/SparseVector.h +++ b/Eigen/src/SparseCore/SparseVector.h @@ -158,6 +158,7 @@ class SparseVector Index inner = IsColVector ? row : col; Index outer = IsColVector ? col : row; + EIGEN_ONLY_USED_FOR_DEBUG(outer); eigen_assert(outer==0); return insert(inner); } diff --git a/Eigen/src/SparseLU/SparseLU.h b/Eigen/src/SparseLU/SparseLU.h index 4514cfd..bdc4f19 100644 --- a/Eigen/src/SparseLU/SparseLU.h +++ b/Eigen/src/SparseLU/SparseLU.h @@ -749,8 +749,8 @@ struct SparseLUMatrixUReturnType : internal::no_assignment_operator } else { - Map, 0, OuterStride<> > A( &(m_mapL.valuePtr()[luptr]), nsupc, nsupc, OuterStride<>(lda) ); - Map< Matrix, 0, OuterStride<> > U (&(X(fsupc,0)), nsupc, nrhs, OuterStride<>(n) ); + Map, 0, OuterStride<> > A( &(m_mapL.valuePtr()[luptr]), nsupc, nsupc, OuterStride<>(lda) ); + Map< Matrix, 0, OuterStride<> > U (&(X(fsupc,0)), nsupc, nrhs, OuterStride<>(n) ); U = A.template triangularView().solve(U); } diff --git a/Eigen/src/SparseLU/SparseLUImpl.h b/Eigen/src/SparseLU/SparseLUImpl.h index 14d7089..99d651e4 100644 --- a/Eigen/src/SparseLU/SparseLUImpl.h +++ b/Eigen/src/SparseLU/SparseLUImpl.h @@ -21,6 +21,8 @@ class SparseLUImpl { public: typedef Matrix ScalarVector; + typedef Matrix ScalarMatrix; + typedef Map > MappedMatrixBlock; typedef Matrix IndexVector; typedef typename ScalarVector::RealScalar RealScalar; typedef Ref > BlockScalarVector; diff --git a/Eigen/src/SparseLU/SparseLU_Memory.h b/Eigen/src/SparseLU/SparseLU_Memory.h index 1ffa7d5..45f96d1 100644 --- a/Eigen/src/SparseLU/SparseLU_Memory.h +++ b/Eigen/src/SparseLU/SparseLU_Memory.h @@ -153,8 +153,8 @@ Index SparseLUImpl::memInit(Index m, Index n, Index annz, Index lw { Index& num_expansions = glu.num_expansions; //No memory expansions so far num_expansions = 0; - glu.nzumax = glu.nzlumax = (std::min)(fillratio * annz / n, m) * n; // estimated number of nonzeros in U - glu.nzlmax = (std::max)(Index(4), fillratio) * annz / 4; // estimated nnz in L factor + glu.nzumax = glu.nzlumax = (std::min)(fillratio * (annz+1) / n, m) * n; // estimated number of nonzeros in U + glu.nzlmax = (std::max)(Index(4), fillratio) * (annz+1) / 4; // estimated nnz in L factor // Return the estimated size to the user if necessary Index tempSpace; tempSpace = (2*panel_size + 4 + LUNoMarker) * m * sizeof(Index) + (panel_size + 1) * m * sizeof(Scalar); diff --git a/Eigen/src/SparseLU/SparseLU_SupernodalMatrix.h b/Eigen/src/SparseLU/SparseLU_SupernodalMatrix.h index b16afd6..54a5694 100644 --- a/Eigen/src/SparseLU/SparseLU_SupernodalMatrix.h +++ b/Eigen/src/SparseLU/SparseLU_SupernodalMatrix.h @@ -236,7 +236,7 @@ void MappedSuperNodalMatrix::solveInPlace( MatrixBase&X) con Index n = X.rows(); Index nrhs = X.cols(); const Scalar * Lval = valuePtr(); // Nonzero values - Matrix work(n, nrhs); // working vector + Matrix work(n, nrhs); // working vector work.setZero(); for (Index k = 0; k <= nsuper(); k ++) { @@ -267,12 +267,12 @@ void MappedSuperNodalMatrix::solveInPlace( MatrixBase&X) con Index lda = colIndexPtr()[fsupc+1] - luptr; // Triangular solve - Map, 0, OuterStride<> > A( &(Lval[luptr]), nsupc, nsupc, OuterStride<>(lda) ); - Map< Matrix, 0, OuterStride<> > U (&(X(fsupc,0)), nsupc, nrhs, OuterStride<>(n) ); + Map, 0, OuterStride<> > A( &(Lval[luptr]), nsupc, nsupc, OuterStride<>(lda) ); + Map< Matrix, 0, OuterStride<> > U (&(X(fsupc,0)), nsupc, nrhs, OuterStride<>(n) ); U = A.template triangularView().solve(U); // Matrix-vector product - new (&A) Map, 0, OuterStride<> > ( &(Lval[luptr+nsupc]), nrow, nsupc, OuterStride<>(lda) ); + new (&A) Map, 0, OuterStride<> > ( &(Lval[luptr+nsupc]), nrow, nsupc, OuterStride<>(lda) ); work.block(0, 0, nrow, nrhs) = A * U; //Begin Scatter diff --git a/Eigen/src/SparseLU/SparseLU_column_bmod.h b/Eigen/src/SparseLU/SparseLU_column_bmod.h index f24bd87..cacc7e9 100644 --- a/Eigen/src/SparseLU/SparseLU_column_bmod.h +++ b/Eigen/src/SparseLU/SparseLU_column_bmod.h @@ -162,11 +162,11 @@ Index SparseLUImpl::column_bmod(const Index jcol, const Index nseg // points to the beginning of jcol in snode L\U(jsupno) ufirst = glu.xlusup(jcol) + d_fsupc; Index lda = glu.xlusup(jcol+1) - glu.xlusup(jcol); - Map, 0, OuterStride<> > A( &(glu.lusup.data()[luptr]), nsupc, nsupc, OuterStride<>(lda) ); + MappedMatrixBlock A( &(glu.lusup.data()[luptr]), nsupc, nsupc, OuterStride<>(lda) ); VectorBlock u(glu.lusup, ufirst, nsupc); u = A.template triangularView().solve(u); - new (&A) Map, 0, OuterStride<> > ( &(glu.lusup.data()[luptr+nsupc]), nrow, nsupc, OuterStride<>(lda) ); + new (&A) MappedMatrixBlock ( &(glu.lusup.data()[luptr+nsupc]), nrow, nsupc, OuterStride<>(lda) ); VectorBlock l(glu.lusup, ufirst+nsupc, nrow); l.noalias() -= A * u; diff --git a/Eigen/src/SparseLU/SparseLU_kernel_bmod.h b/Eigen/src/SparseLU/SparseLU_kernel_bmod.h index 0d0283b..6af0267 100644 --- a/Eigen/src/SparseLU/SparseLU_kernel_bmod.h +++ b/Eigen/src/SparseLU/SparseLU_kernel_bmod.h @@ -56,7 +56,7 @@ EIGEN_DONT_INLINE void LU_kernel_bmod::run(const int segsi // Dense triangular solve -- start effective triangle luptr += lda * no_zeros + no_zeros; // Form Eigen matrix and vector - Map, 0, OuterStride<> > A( &(lusup.data()[luptr]), segsize, segsize, OuterStride<>(lda) ); + Map, 0, OuterStride<> > A( &(lusup.data()[luptr]), segsize, segsize, OuterStride<>(lda) ); Map > u(tempv.data(), segsize); u = A.template triangularView().solve(u); @@ -65,7 +65,7 @@ EIGEN_DONT_INLINE void LU_kernel_bmod::run(const int segsi luptr += segsize; const Index PacketSize = internal::packet_traits::size; Index ldl = internal::first_multiple(nrow, PacketSize); - Map, 0, OuterStride<> > B( &(lusup.data()[luptr]), nrow, segsize, OuterStride<>(lda) ); + Map, 0, OuterStride<> > B( &(lusup.data()[luptr]), nrow, segsize, OuterStride<>(lda) ); Index aligned_offset = internal::first_aligned(tempv.data()+segsize, PacketSize); Index aligned_with_B_offset = (PacketSize-internal::first_aligned(B.data(), PacketSize))%PacketSize; Map, 0, OuterStride<> > l(tempv.data()+segsize+aligned_offset+aligned_with_B_offset, nrow, OuterStride<>(ldl) ); diff --git a/Eigen/src/SparseLU/SparseLU_panel_bmod.h b/Eigen/src/SparseLU/SparseLU_panel_bmod.h index da0e0fc..9d2ff29 100644 --- a/Eigen/src/SparseLU/SparseLU_panel_bmod.h +++ b/Eigen/src/SparseLU/SparseLU_panel_bmod.h @@ -102,7 +102,7 @@ void SparseLUImpl::panel_bmod(const Index m, const Index w, const if(nsupc >= 2) { Index ldu = internal::first_multiple(u_rows, PacketSize); - Map, Aligned, OuterStride<> > U(tempv.data(), u_rows, u_cols, OuterStride<>(ldu)); + Map > U(tempv.data(), u_rows, u_cols, OuterStride<>(ldu)); // gather U Index u_col = 0; @@ -136,17 +136,17 @@ void SparseLUImpl::panel_bmod(const Index m, const Index w, const Index lda = glu.xlusup(fsupc+1) - glu.xlusup(fsupc); no_zeros = (krep - u_rows + 1) - fsupc; luptr += lda * no_zeros + no_zeros; - Map, 0, OuterStride<> > A(glu.lusup.data()+luptr, u_rows, u_rows, OuterStride<>(lda) ); + MappedMatrixBlock A(glu.lusup.data()+luptr, u_rows, u_rows, OuterStride<>(lda) ); U = A.template triangularView().solve(U); // update luptr += u_rows; - Map, 0, OuterStride<> > B(glu.lusup.data()+luptr, nrow, u_rows, OuterStride<>(lda) ); + MappedMatrixBlock B(glu.lusup.data()+luptr, nrow, u_rows, OuterStride<>(lda) ); eigen_assert(tempv.size()>w*ldu + nrow*w + 1); Index ldl = internal::first_multiple(nrow, PacketSize); Index offset = (PacketSize-internal::first_aligned(B.data(), PacketSize)) % PacketSize; - Map, 0, OuterStride<> > L(tempv.data()+w*ldu+offset, nrow, u_cols, OuterStride<>(ldl)); + MappedMatrixBlock L(tempv.data()+w*ldu+offset, nrow, u_cols, OuterStride<>(ldl)); L.setZero(); internal::sparselu_gemm(L.rows(), L.cols(), B.cols(), B.data(), B.outerStride(), U.data(), U.outerStride(), L.data(), L.outerStride()); diff --git a/Eigen/src/SparseLU/SparseLU_pivotL.h b/Eigen/src/SparseLU/SparseLU_pivotL.h index 457789c..2e49ef6 100644 --- a/Eigen/src/SparseLU/SparseLU_pivotL.h +++ b/Eigen/src/SparseLU/SparseLU_pivotL.h @@ -71,7 +71,7 @@ Index SparseLUImpl::pivotL(const Index jcol, const RealScalar& dia // Determine the largest abs numerical value for partial pivoting Index diagind = iperm_c(jcol); // diagonal index - RealScalar pivmax = 0.0; + RealScalar pivmax(-1.0); Index pivptr = nsupc; Index diag = emptyIdxLU; RealScalar rtemp; @@ -87,8 +87,9 @@ Index SparseLUImpl::pivotL(const Index jcol, const RealScalar& dia } // Test for singularity - if ( pivmax == 0.0 ) { - pivrow = lsub_ptr[pivptr]; + if ( pivmax <= RealScalar(0.0) ) { + // if pivmax == -1, the column is structurally empty, otherwise it is only numerically zero + pivrow = pivmax < RealScalar(0.0) ? diagind : lsub_ptr[pivptr]; perm_r(pivrow) = jcol; return (jcol+1); }