Skip to content

GitLab

Unverified Commit 4c4419e0 authored by Lucas Beyer's avatar Lucas Beyer Committed by GitHub
Browse files

Merge pull request #108 from STulling/master 

Fix Windows MSVC install by updating Eigen Library
parents 4d5343c3 13b115ab
Expand all Hide whitespace changes
Inline Side-by-side
Showing with 2086 additions and 1293 deletions
pydensecrf/densecrf/include/Eigen/SparseCholesky
View file @ 4c4419e0
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// Copyright (C) 2008-2013 Gael Guennebaud <gael.guennebaud@inria.fr>
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#ifndef EIGEN_SPARSECHOLESKY_MODULE_H
#define EIGEN_SPARSECHOLESKY_MODULE_H
#include "SparseCore"
#include "OrderingMethods"
#include "src/Core/util/DisableStupidWarnings.h"
/** \ingroup Sparse_modules
/**
* \defgroup SparseCholesky_Module SparseCholesky module
*
* This module currently provides two variants of the direct sparse Cholesky decomposition for selfadjoint (hermitian) matrices.
......@@ -20,11 +30,16 @@
* \endcode
*/
#include "src/misc/Solve.h"
#include "src/misc/SparseSolve.h"
#ifdef EIGEN_MPL2_ONLY
#error The SparseCholesky module has nothing to offer in MPL2 only mode
#endif
#include "src/SparseCholesky/SimplicialCholesky.h"
#ifndef EIGEN_MPL2_ONLY
#include "src/SparseCholesky/SimplicialCholesky_impl.h"
#endif
#include "src/Core/util/ReenableStupidWarnings.h"
#endif // EIGEN_SPARSECHOLESKY_MODULE_H
pydensecrf/densecrf/include/Eigen/SparseCore
View file @ 4c4419e0
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#ifndef EIGEN_SPARSECORE_MODULE_H
#define EIGEN_SPARSECORE_MODULE_H
......@@ -11,10 +18,10 @@
#include <cstring>
#include <algorithm>
/** \ingroup Sparse_modules
/**
* \defgroup SparseCore_Module SparseCore module
*
* This module provides a sparse matrix representation, and basic associatd matrix manipulations
* This module provides a sparse matrix representation, and basic associated matrix manipulations
* and operations.
*
* See the \ref TutorialSparse "Sparse tutorial"
......@@ -26,39 +33,35 @@
* This module depends on: Core.
*/
namespace Eigen {
/** The type used to identify a general sparse storage. */
struct Sparse {};
}
#include "src/SparseCore/SparseUtil.h"
#include "src/SparseCore/SparseMatrixBase.h"
#include "src/SparseCore/SparseAssign.h"
#include "src/SparseCore/CompressedStorage.h"
#include "src/SparseCore/AmbiVector.h"
#include "src/SparseCore/SparseCompressedBase.h"
#include "src/SparseCore/SparseMatrix.h"
#include "src/SparseCore/SparseMap.h"
#include "src/SparseCore/MappedSparseMatrix.h"
#include "src/SparseCore/SparseVector.h"
#include "src/SparseCore/CoreIterators.h"
#include "src/SparseCore/SparseBlock.h"
#include "src/SparseCore/SparseTranspose.h"
#include "src/SparseCore/SparseRef.h"
#include "src/SparseCore/SparseCwiseUnaryOp.h"
#include "src/SparseCore/SparseCwiseBinaryOp.h"
#include "src/SparseCore/SparseTranspose.h"
#include "src/SparseCore/SparseBlock.h"
#include "src/SparseCore/SparseDot.h"
#include "src/SparseCore/SparsePermutation.h"
#include "src/SparseCore/SparseAssign.h"
#include "src/SparseCore/SparseRedux.h"
#include "src/SparseCore/SparseFuzzy.h"
#include "src/SparseCore/SparseView.h"
#include "src/SparseCore/SparseDiagonalProduct.h"
#include "src/SparseCore/ConservativeSparseSparseProduct.h"
#include "src/SparseCore/SparseSparseProductWithPruning.h"
#include "src/SparseCore/SparseProduct.h"
#include "src/SparseCore/SparseDenseProduct.h"
#include "src/SparseCore/SparseDiagonalProduct.h"
#include "src/SparseCore/SparseTriangularView.h"
#include "src/SparseCore/SparseSelfAdjointView.h"
#include "src/SparseCore/SparseTriangularView.h"
#include "src/SparseCore/TriangularSolver.h"
#include "src/SparseCore/SparseView.h"
#include "src/SparseCore/SparsePermutation.h"
#include "src/SparseCore/SparseFuzzy.h"
#include "src/SparseCore/SparseSolverBase.h"
#include "src/Core/util/ReenableStupidWarnings.h"
......
pydensecrf/densecrf/include/Eigen/SparseLU 0 → 100644
View file @ 4c4419e0
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// Copyright (C) 2012 Désiré Nuentsa-Wakam <desire.nuentsa_wakam@inria.fr>
// Copyright (C) 2012 Gael Guennebaud <gael.guennebaud@inria.fr>
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#ifndef EIGEN_SPARSELU_MODULE_H
#define EIGEN_SPARSELU_MODULE_H
#include "SparseCore"
/**
* \defgroup SparseLU_Module SparseLU module
* This module defines a supernodal factorization of general sparse matrices.
* The code is fully optimized for supernode-panel updates with specialized kernels.
* Please, see the documentation of the SparseLU class for more details.
*/
// Ordering interface
#include "OrderingMethods"
#include "src/SparseLU/SparseLU_gemm_kernel.h"
#include "src/SparseLU/SparseLU_Structs.h"
#include "src/SparseLU/SparseLU_SupernodalMatrix.h"
#include "src/SparseLU/SparseLUImpl.h"
#include "src/SparseCore/SparseColEtree.h"
#include "src/SparseLU/SparseLU_Memory.h"
#include "src/SparseLU/SparseLU_heap_relax_snode.h"
#include "src/SparseLU/SparseLU_relax_snode.h"
#include "src/SparseLU/SparseLU_pivotL.h"
#include "src/SparseLU/SparseLU_panel_dfs.h"
#include "src/SparseLU/SparseLU_kernel_bmod.h"
#include "src/SparseLU/SparseLU_panel_bmod.h"
#include "src/SparseLU/SparseLU_column_dfs.h"
#include "src/SparseLU/SparseLU_column_bmod.h"
#include "src/SparseLU/SparseLU_copy_to_ucol.h"
#include "src/SparseLU/SparseLU_pruneL.h"
#include "src/SparseLU/SparseLU_Utils.h"
#include "src/SparseLU/SparseLU.h"
#endif // EIGEN_SPARSELU_MODULE_H
pydensecrf/densecrf/include/Eigen/SparseQR 0 → 100644
View file @ 4c4419e0
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#ifndef EIGEN_SPARSEQR_MODULE_H
#define EIGEN_SPARSEQR_MODULE_H
#include "SparseCore"
#include "OrderingMethods"
#include "src/Core/util/DisableStupidWarnings.h"
/** \defgroup SparseQR_Module SparseQR module
* \brief Provides QR decomposition for sparse matrices
*
* This module provides a simplicial version of the left-looking Sparse QR decomposition.
* The columns of the input matrix should be reordered to limit the fill-in during the
* decomposition. Built-in methods (COLAMD, AMD) or external methods (METIS) can be used to this end.
* See the \link OrderingMethods_Module OrderingMethods\endlink module for the list
* of built-in and external ordering methods.
*
* \code
* #include <Eigen/SparseQR>
* \endcode
*
*
*/
#include "src/SparseCore/SparseColEtree.h"
#include "src/SparseQR/SparseQR.h"
#include "src/Core/util/ReenableStupidWarnings.h"
#endif
pydensecrf/densecrf/include/Eigen/StdDeque
View file @ 4c4419e0
......@@ -14,7 +14,7 @@
#include "Core"
#include <deque>
#if (defined(_MSC_VER) && defined(_WIN64)) /* MSVC auto aligns in 64 bit builds */
#if EIGEN_COMP_MSVC && EIGEN_OS_WIN64 && (EIGEN_MAX_STATIC_ALIGN_BYTES<=16) /* MSVC auto aligns up to 16 bytes in 64 bit builds */
#define EIGEN_DEFINE_STL_DEQUE_SPECIALIZATION(...)
......
pydensecrf/densecrf/include/Eigen/StdList
View file @ 4c4419e0
......@@ -13,7 +13,7 @@
#include "Core"
#include <list>
#if (defined(_MSC_VER) && defined(_WIN64)) /* MSVC auto aligns in 64 bit builds */
#if EIGEN_COMP_MSVC && EIGEN_OS_WIN64 && (EIGEN_MAX_STATIC_ALIGN_BYTES<=16) /* MSVC auto aligns up to 16 bytes in 64 bit builds */
#define EIGEN_DEFINE_STL_LIST_SPECIALIZATION(...)
......
pydensecrf/densecrf/include/Eigen/StdVector
View file @ 4c4419e0
......@@ -14,7 +14,7 @@
#include "Core"
#include <vector>
#if (defined(_MSC_VER) && defined(_WIN64)) /* MSVC auto aligns in 64 bit builds */
#if EIGEN_COMP_MSVC && EIGEN_OS_WIN64 && (EIGEN_MAX_STATIC_ALIGN_BYTES<=16) /* MSVC auto aligns up to 16 bytes in 64 bit builds */
#define EIGEN_DEFINE_STL_VECTOR_SPECIALIZATION(...)
......
pydensecrf/densecrf/include/Eigen/SuperLUSupport
View file @ 4c4419e0
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#ifndef EIGEN_SUPERLUSUPPORT_MODULE_H
#define EIGEN_SUPERLUSUPPORT_MODULE_H
......@@ -36,6 +43,8 @@ namespace Eigen { struct SluMatrix; }
* - class SuperLU: a supernodal sequential LU factorization.
* - class SuperILU: a supernodal sequential incomplete LU factorization (to be used as a preconditioner for iterative methods).
*
* \warning This wrapper requires at least versions 4.0 of SuperLU. The 3.x versions are not supported.
*
* \warning When including this module, you have to use SUPERLU_EMPTY instead of EMPTY which is no longer defined because it is too polluting.
*
* \code
......@@ -48,12 +57,8 @@ namespace Eigen { struct SluMatrix; }
*
*/
#include "src/misc/Solve.h"
#include "src/misc/SparseSolve.h"
#include "src/SuperLUSupport/SuperLUSupport.h"
#include "src/Core/util/ReenableStupidWarnings.h"
#endif // EIGEN_SUPERLUSUPPORT_MODULE_H
pydensecrf/densecrf/include/Eigen/UmfPackSupport
View file @ 4c4419e0
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#ifndef EIGEN_UMFPACKSUPPORT_MODULE_H
#define EIGEN_UMFPACKSUPPORT_MODULE_H
......@@ -12,7 +19,7 @@ extern "C" {
/** \ingroup Support_modules
* \defgroup UmfPackSupport_Module UmfPackSupport module
*
* This module provides an interface to the UmfPack library which is part of the <a href="http://www.cise.ufl.edu/research/sparse/SuiteSparse/">suitesparse</a> package.
* This module provides an interface to the UmfPack library which is part of the <a href="http://www.suitesparse.com">suitesparse</a> package.
* It provides the following factorization class:
* - class UmfPackLU: a multifrontal sequential LU factorization.
*
......@@ -26,9 +33,6 @@ extern "C" {
*
*/
#include "src/misc/Solve.h"
#include "src/misc/SparseSolve.h"
#include "src/UmfPackSupport/UmfPackSupport.h"
#include "src/Core/util/ReenableStupidWarnings.h"
......
pydensecrf/densecrf/include/Eigen/src/Cholesky/LLT.h
View file @ 4c4419e0
......@@ -10,7 +10,7 @@
#ifndef EIGEN_LLT_H
#define EIGEN_LLT_H
namespace Eigen {
namespace Eigen {
namespace internal{
template<typename MatrixType, int UpLo> struct LLT_Traits;
......@@ -22,9 +22,9 @@ template<typename MatrixType, int UpLo> struct LLT_Traits;
*
* \brief Standard Cholesky decomposition (LL^T) of a matrix and associated features
*
* \param MatrixType the type of the matrix of which we are computing the LL^T Cholesky decomposition
* \param UpLo the triangular part that will be used for the decompositon: Lower (default) or Upper.
* The other triangular part won't be read.
* \tparam _MatrixType the type of the matrix of which we are computing the LL^T Cholesky decomposition
* \tparam _UpLo the triangular part that will be used for the decompositon: Lower (default) or Upper.
* The other triangular part won't be read.
*
* This class performs a LL^T Cholesky decomposition of a symmetric, positive definite
* matrix A such that A = LL^* = U^*U, where L is lower triangular.
......@@ -40,12 +40,18 @@ template<typename MatrixType, int UpLo> struct LLT_Traits;
*
* Example: \include LLT_example.cpp
* Output: \verbinclude LLT_example.out
*
* \sa MatrixBase::llt(), class LDLT
*/
/* HEY THIS DOX IS DISABLED BECAUSE THERE's A BUG EITHER HERE OR IN LDLT ABOUT THAT (OR BOTH)
* Note that during the decomposition, only the upper triangular part of A is considered. Therefore,
* the strict lower part does not have to store correct values.
*
* \b Performance: for best performance, it is recommended to use a column-major storage format
* with the Lower triangular part (the default), or, equivalently, a row-major storage format
* with the Upper triangular part. Otherwise, you might get a 20% slowdown for the full factorization
* step, and rank-updates can be up to 3 times slower.
*
* This class supports the \link InplaceDecomposition inplace decomposition \endlink mechanism.
*
* Note that during the decomposition, only the lower (or upper, as defined by _UpLo) triangular part of A is considered.
* Therefore, the strict lower part does not have to store correct values.
*
* \sa MatrixBase::llt(), SelfAdjointView::llt(), class LDLT
*/
template<typename _MatrixType, int _UpLo> class LLT
{
......@@ -54,12 +60,12 @@ template<typename _MatrixType, int _UpLo> class LLT
enum {
RowsAtCompileTime = MatrixType::RowsAtCompileTime,
ColsAtCompileTime = MatrixType::ColsAtCompileTime,
Options = MatrixType::Options,
MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime
};
typedef typename MatrixType::Scalar Scalar;
typedef typename NumTraits<typename MatrixType::Scalar>::Real RealScalar;
typedef typename MatrixType::Index Index;
typedef Eigen::Index Index; ///< \deprecated since Eigen 3.3
typedef typename MatrixType::StorageIndex StorageIndex;
enum {
PacketSize = internal::packet_traits<Scalar>::size,
......@@ -83,14 +89,30 @@ template<typename _MatrixType, int _UpLo> class LLT
* according to the specified problem \a size.
* \sa LLT()
*/
LLT(Index size) : m_matrix(size, size),
explicit LLT(Index size) : m_matrix(size, size),
m_isInitialized(false) {}
LLT(const MatrixType& matrix)
template<typename InputType>
explicit LLT(const EigenBase<InputType>& matrix)
: m_matrix(matrix.rows(), matrix.cols()),
m_isInitialized(false)
{
compute(matrix);
compute(matrix.derived());
}
/** \brief Constructs a LDLT factorization from a given matrix
*
* This overloaded constructor is provided for \link InplaceDecomposition inplace decomposition \endlink when
* \c MatrixType is a Eigen::Ref.
*
* \sa LLT(const EigenBase&)
*/
template<typename InputType>
explicit LLT(EigenBase<InputType>& matrix)
: m_matrix(matrix.derived()),
m_isInitialized(false)
{
compute(matrix.derived());
}
/** \returns a view of the upper triangular matrix U */
......@@ -115,33 +137,33 @@ template<typename _MatrixType, int _UpLo> class LLT
* Example: \include LLT_solve.cpp
* Output: \verbinclude LLT_solve.out
*
* \sa solveInPlace(), MatrixBase::llt()
* \sa solveInPlace(), MatrixBase::llt(), SelfAdjointView::llt()
*/
template<typename Rhs>
inline const internal::solve_retval<LLT, Rhs>
inline const Solve<LLT, Rhs>
solve(const MatrixBase<Rhs>& b) const
{
eigen_assert(m_isInitialized && "LLT is not initialized.");
eigen_assert(m_matrix.rows()==b.rows()
&& "LLT::solve(): invalid number of rows of the right hand side matrix b");
return internal::solve_retval<LLT, Rhs>(*this, b.derived());
return Solve<LLT, Rhs>(*this, b.derived());
}
#ifdef EIGEN2_SUPPORT
template<typename OtherDerived, typename ResultType>
bool solve(const MatrixBase<OtherDerived>& b, ResultType *result) const
{
*result = this->solve(b);
return true;
}
bool isPositiveDefinite() const { return true; }
#endif
template<typename Derived>
void solveInPlace(MatrixBase<Derived> &bAndX) const;
void solveInPlace(const MatrixBase<Derived> &bAndX) const;
LLT& compute(const MatrixType& matrix);
template<typename InputType>
LLT& compute(const EigenBase<InputType>& matrix);
/** \returns an estimate of the reciprocal condition number of the matrix of
* which \c *this is the Cholesky decomposition.
*/
RealScalar rcond() const
{
eigen_assert(m_isInitialized && "LLT is not initialized.");
eigen_assert(m_info == Success && "LLT failed because matrix appears to be negative");
return internal::rcond_estimate_helper(m_l1_norm, *this);
}
/** \returns the LLT decomposition matrix
*
......@@ -159,7 +181,7 @@ template<typename _MatrixType, int _UpLo> class LLT
/** \brief Reports whether previous computation was successful.
*
* \returns \c Success if computation was succesful,
* \c NumericalIssue if the matrix.appears to be negative.
* \c NumericalIssue if the matrix.appears not to be positive definite.
*/
ComputationInfo info() const
{
......@@ -167,18 +189,38 @@ template<typename _MatrixType, int _UpLo> class LLT
return m_info;
}
/** \returns the adjoint of \c *this, that is, a const reference to the decomposition itself as the underlying matrix is self-adjoint.
*
* This method is provided for compatibility with other matrix decompositions, thus enabling generic code such as:
* \code x = decomposition.adjoint().solve(b) \endcode
*/
const LLT& adjoint() const { return *this; };
inline Index rows() const { return m_matrix.rows(); }
inline Index cols() const { return m_matrix.cols(); }
template<typename VectorType>
LLT rankUpdate(const VectorType& vec, const RealScalar& sigma = 1);
#ifndef EIGEN_PARSED_BY_DOXYGEN
template<typename RhsType, typename DstType>
EIGEN_DEVICE_FUNC
void _solve_impl(const RhsType &rhs, DstType &dst) const;
#endif
protected:
static void check_template_parameters()
{
EIGEN_STATIC_ASSERT_NON_INTEGER(Scalar);
}
/** \internal
* Used to compute and store L
* The strict upper part is not used and even not initialized.
*/
MatrixType m_matrix;
RealScalar m_l1_norm;
bool m_isInitialized;
ComputationInfo m_info;
};
......@@ -188,18 +230,18 @@ namespace internal {
template<typename Scalar, int UpLo> struct llt_inplace;
template<typename MatrixType, typename VectorType>
static typename MatrixType::Index llt_rank_update_lower(MatrixType& mat, const VectorType& vec, const typename MatrixType::RealScalar& sigma)
static Index llt_rank_update_lower(MatrixType& mat, const VectorType& vec, const typename MatrixType::RealScalar& sigma)
{
using std::sqrt;
typedef typename MatrixType::Scalar Scalar;
typedef typename MatrixType::RealScalar RealScalar;
typedef typename MatrixType::Index Index;
typedef typename MatrixType::ColXpr ColXpr;
typedef typename internal::remove_all<ColXpr>::type ColXprCleaned;
typedef typename ColXprCleaned::SegmentReturnType ColXprSegment;
typedef Matrix<Scalar,Dynamic,1> TempVectorType;
typedef typename TempVectorType::SegmentReturnType TempVecSegment;
int n = mat.cols();
Index n = mat.cols();
eigen_assert(mat.rows()==n && vec.size()==n);
TempVectorType temp;
......@@ -211,12 +253,12 @@ static typename MatrixType::Index llt_rank_update_lower(MatrixType& mat, const V
// i.e., for sigma > 0
temp = sqrt(sigma) * vec;
for(int i=0; i<n; ++i)
for(Index i=0; i<n; ++i)
{
JacobiRotation<Scalar> g;
g.makeGivens(mat(i,i), -temp(i), &mat(i,i));
int rs = n-i-1;
Index rs = n-i-1;
if(rs>0)
{
ColXprSegment x(mat.col(i).tail(rs));
......@@ -229,12 +271,12 @@ static typename MatrixType::Index llt_rank_update_lower(MatrixType& mat, const V
{
temp = vec;
RealScalar beta = 1;
for(int j=0; j<n; ++j)
for(Index j=0; j<n; ++j)
{
RealScalar Ljj = real(mat.coeff(j,j));
RealScalar dj = abs2(Ljj);
RealScalar Ljj = numext::real(mat.coeff(j,j));
RealScalar dj = numext::abs2(Ljj);
Scalar wj = temp.coeff(j);
RealScalar swj2 = sigma*abs2(wj);
RealScalar swj2 = sigma*numext::abs2(wj);
RealScalar gamma = dj*beta + swj2;
RealScalar x = dj + swj2/beta;
......@@ -250,7 +292,7 @@ static typename MatrixType::Index llt_rank_update_lower(MatrixType& mat, const V
{
temp.tail(rs) -= (wj/Ljj) * mat.col(j).tail(rs);
if(gamma != 0)
mat.col(j).tail(rs) = (nLjj/Ljj) * mat.col(j).tail(rs) + (nLjj * sigma*conj(wj)/gamma)*temp.tail(rs);
mat.col(j).tail(rs) = (nLjj/Ljj) * mat.col(j).tail(rs) + (nLjj * sigma*numext::conj(wj)/gamma)*temp.tail(rs);
}
}
}
......@@ -261,10 +303,10 @@ template<typename Scalar> struct llt_inplace<Scalar, Lower>
{
typedef typename NumTraits<Scalar>::Real RealScalar;
template<typename MatrixType>
static typename MatrixType::Index unblocked(MatrixType& mat)
static Index unblocked(MatrixType& mat)
{
typedef typename MatrixType::Index Index;
using std::sqrt;
eigen_assert(mat.rows()==mat.cols());
const Index size = mat.rows();
for(Index k = 0; k < size; ++k)
......@@ -275,21 +317,20 @@ template<typename Scalar> struct llt_inplace<Scalar, Lower>
Block<MatrixType,1,Dynamic> A10(mat,k,0,1,k);
Block<MatrixType,Dynamic,Dynamic> A20(mat,k+1,0,rs,k);
RealScalar x = real(mat.coeff(k,k));
RealScalar x = numext::real(mat.coeff(k,k));
if (k>0) x -= A10.squaredNorm();
if (x<=RealScalar(0))
return k;
mat.coeffRef(k,k) = x = sqrt(x);
if (k>0 && rs>0) A21.noalias() -= A20 * A10.adjoint();
if (rs>0) A21 *= RealScalar(1)/x;
if (rs>0) A21 /= x;
}
return -1;
}
template<typename MatrixType>
static typename MatrixType::Index blocked(MatrixType& m)
static Index blocked(MatrixType& m)
{
typedef typename MatrixType::Index Index;
eigen_assert(m.rows()==m.cols());
Index size = m.rows();
if(size<32)
......@@ -314,36 +355,36 @@ template<typename Scalar> struct llt_inplace<Scalar, Lower>
Index ret;
if((ret=unblocked(A11))>=0) return k+ret;
if(rs>0) A11.adjoint().template triangularView<Upper>().template solveInPlace<OnTheRight>(A21);
if(rs>0) A22.template selfadjointView<Lower>().rankUpdate(A21,-1); // bottleneck
if(rs>0) A22.template selfadjointView<Lower>().rankUpdate(A21,typename NumTraits<RealScalar>::Literal(-1)); // bottleneck
}
return -1;
}
template<typename MatrixType, typename VectorType>
static typename MatrixType::Index rankUpdate(MatrixType& mat, const VectorType& vec, const RealScalar& sigma)
static Index rankUpdate(MatrixType& mat, const VectorType& vec, const RealScalar& sigma)
{
return Eigen::internal::llt_rank_update_lower(mat, vec, sigma);
}
};
template<typename Scalar> struct llt_inplace<Scalar, Upper>
{
typedef typename NumTraits<Scalar>::Real RealScalar;
template<typename MatrixType>
static EIGEN_STRONG_INLINE typename MatrixType::Index unblocked(MatrixType& mat)
static EIGEN_STRONG_INLINE Index unblocked(MatrixType& mat)
{
Transpose<MatrixType> matt(mat);
return llt_inplace<Scalar, Lower>::unblocked(matt);
}
template<typename MatrixType>
static EIGEN_STRONG_INLINE typename MatrixType::Index blocked(MatrixType& mat)
static EIGEN_STRONG_INLINE Index blocked(MatrixType& mat)
{
Transpose<MatrixType> matt(mat);
return llt_inplace<Scalar, Lower>::blocked(matt);
}
template<typename MatrixType, typename VectorType>
static typename MatrixType::Index rankUpdate(MatrixType& mat, const VectorType& vec, const RealScalar& sigma)
static Index rankUpdate(MatrixType& mat, const VectorType& vec, const RealScalar& sigma)
{
Transpose<MatrixType> matt(mat);
return llt_inplace<Scalar, Lower>::rankUpdate(matt, vec.conjugate(), sigma);
......@@ -354,8 +395,8 @@ template<typename MatrixType> struct LLT_Traits<MatrixType,Lower>
{
typedef const TriangularView<const MatrixType, Lower> MatrixL;
typedef const TriangularView<const typename MatrixType::AdjointReturnType, Upper> MatrixU;
static inline MatrixL getL(const MatrixType& m) { return m; }
static inline MatrixU getU(const MatrixType& m) { return m.adjoint(); }
static inline MatrixL getL(const MatrixType& m) { return MatrixL(m); }
static inline MatrixU getU(const MatrixType& m) { return MatrixU(m.adjoint()); }
static bool inplace_decomposition(MatrixType& m)
{ return llt_inplace<typename MatrixType::Scalar, Lower>::blocked(m)==-1; }
};
......@@ -364,8 +405,8 @@ template<typename MatrixType> struct LLT_Traits<MatrixType,Upper>
{
typedef const TriangularView<const typename MatrixType::AdjointReturnType, Lower> MatrixL;
typedef const TriangularView<const MatrixType, Upper> MatrixU;
static inline MatrixL getL(const MatrixType& m) { return m.adjoint(); }
static inline MatrixU getU(const MatrixType& m) { return m; }
static inline MatrixL getL(const MatrixType& m) { return MatrixL(m.adjoint()); }
static inline MatrixU getU(const MatrixType& m) { return MatrixU(m); }
static bool inplace_decomposition(MatrixType& m)
{ return llt_inplace<typename MatrixType::Scalar, Upper>::blocked(m)==-1; }
};
......@@ -380,12 +421,29 @@ template<typename MatrixType> struct LLT_Traits<MatrixType,Upper>
* Output: \verbinclude TutorialLinAlgComputeTwice.out
*/
template<typename MatrixType, int _UpLo>
LLT<MatrixType,_UpLo>& LLT<MatrixType,_UpLo>::compute(const MatrixType& a)
template<typename InputType>
LLT<MatrixType,_UpLo>& LLT<MatrixType,_UpLo>::compute(const EigenBase<InputType>& a)
{
check_template_parameters();
eigen_assert(a.rows()==a.cols());
const Index size = a.rows();
m_matrix.resize(size, size);
m_matrix = a;
if (!internal::is_same_dense(m_matrix, a.derived()))
m_matrix = a.derived();
// Compute matrix L1 norm = max abs column sum.
m_l1_norm = RealScalar(0);
// TODO move this code to SelfAdjointView
for (Index col = 0; col < size; ++col) {
RealScalar abs_col_sum;
if (_UpLo == Lower)
abs_col_sum = m_matrix.col(col).tail(size - col).template lpNorm<1>() + m_matrix.row(col).head(col).template lpNorm<1>();
else
abs_col_sum = m_matrix.col(col).head(col).template lpNorm<1>() + m_matrix.row(col).tail(size - col).template lpNorm<1>();
if (abs_col_sum > m_l1_norm)
m_l1_norm = abs_col_sum;
}
m_isInitialized = true;
bool ok = Traits::inplace_decomposition(m_matrix);
......@@ -413,39 +471,33 @@ LLT<_MatrixType,_UpLo> LLT<_MatrixType,_UpLo>::rankUpdate(const VectorType& v, c
return *this;
}
namespace internal {
template<typename _MatrixType, int UpLo, typename Rhs>
struct solve_retval<LLT<_MatrixType, UpLo>, Rhs>
: solve_retval_base<LLT<_MatrixType, UpLo>, Rhs>
{
typedef LLT<_MatrixType,UpLo> LLTType;
EIGEN_MAKE_SOLVE_HELPERS(LLTType,Rhs)
template<typename Dest> void evalTo(Dest& dst) const
{
dst = rhs();
dec().solveInPlace(dst);
}
};
#ifndef EIGEN_PARSED_BY_DOXYGEN
template<typename _MatrixType,int _UpLo>
template<typename RhsType, typename DstType>
void LLT<_MatrixType,_UpLo>::_solve_impl(const RhsType &rhs, DstType &dst) const
{
dst = rhs;
solveInPlace(dst);
}
#endif
/** \internal use x = llt_object.solve(x);
*
*
* This is the \em in-place version of solve().
*
* \param bAndX represents both the right-hand side matrix b and result x.
*
* \returns true always! If you need to check for existence of solutions, use another decomposition like LU, QR, or SVD.
* This version avoids a copy when the right hand side matrix b is not needed anymore.
*
* This version avoids a copy when the right hand side matrix b is not
* needed anymore.
* \warning The parameter is only marked 'const' to make the C++ compiler accept a temporary expression here.
* This function will const_cast it, so constness isn't honored here.
*
* \sa LLT::solve(), MatrixBase::llt()
*/
template<typename MatrixType, int _UpLo>
template<typename Derived>
void LLT<MatrixType,_UpLo>::solveInPlace(MatrixBase<Derived> &bAndX) const
void LLT<MatrixType,_UpLo>::solveInPlace(const MatrixBase<Derived> &bAndX) const
{
eigen_assert(m_isInitialized && "LLT is not initialized.");
eigen_assert(m_matrix.rows()==bAndX.rows());
......@@ -465,6 +517,7 @@ MatrixType LLT<MatrixType,_UpLo>::reconstructedMatrix() const
/** \cholesky_module
* \returns the LLT decomposition of \c *this
* \sa SelfAdjointView::llt()
*/
template<typename Derived>
inline const LLT<typename MatrixBase<Derived>::PlainObject>
......@@ -475,6 +528,7 @@ MatrixBase<Derived>::llt() const
/** \cholesky_module
* \returns the LLT decomposition of \c *this
* \sa SelfAdjointView::llt()
*/
template<typename MatrixType, unsigned int UpLo>
inline const LLT<typename SelfAdjointView<MatrixType, UpLo>::PlainObject, UpLo>
......
pydensecrf/densecrf/include/Eigen/src/Cholesky/LLT_LAPACKE.h 0 → 100644
View file @ 4c4419e0
/*
Copyright (c) 2011, Intel Corporation. All rights reserved.
Redistribution and use in source and binary forms, with or without modification,
are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright notice, this
list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright notice,
this list of conditions and the following disclaimer in the documentation
and/or other materials provided with the distribution.
* Neither the name of Intel Corporation nor the names of its contributors may
be used to endorse or promote products derived from this software without
specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR
ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON
ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
********************************************************************************
* Content : Eigen bindings to LAPACKe
* LLt decomposition based on LAPACKE_?potrf function.
********************************************************************************
*/
#ifndef EIGEN_LLT_LAPACKE_H
#define EIGEN_LLT_LAPACKE_H
namespace Eigen {
namespace internal {
template<typename Scalar> struct lapacke_llt;
#define EIGEN_LAPACKE_LLT(EIGTYPE, BLASTYPE, LAPACKE_PREFIX) \
template<> struct lapacke_llt<EIGTYPE> \
{ \
template<typename MatrixType> \
static inline Index potrf(MatrixType& m, char uplo) \
{ \
lapack_int matrix_order; \
lapack_int size, lda, info, StorageOrder; \
EIGTYPE* a; \
eigen_assert(m.rows()==m.cols()); \
/* Set up parameters for ?potrf */ \
size = convert_index<lapack_int>(m.rows()); \
StorageOrder = MatrixType::Flags&RowMajorBit?RowMajor:ColMajor; \
matrix_order = StorageOrder==RowMajor ? LAPACK_ROW_MAJOR : LAPACK_COL_MAJOR; \
a = &(m.coeffRef(0,0)); \
lda = convert_index<lapack_int>(m.outerStride()); \
\
info = LAPACKE_##LAPACKE_PREFIX##potrf( matrix_order, uplo, size, (BLASTYPE*)a, lda ); \
info = (info==0) ? -1 : info>0 ? info-1 : size; \
return info; \
} \
}; \
template<> struct llt_inplace<EIGTYPE, Lower> \
{ \
template<typename MatrixType> \
static Index blocked(MatrixType& m) \
{ \
return lapacke_llt<EIGTYPE>::potrf(m, 'L'); \
} \
template<typename MatrixType, typename VectorType> \
static Index rankUpdate(MatrixType& mat, const VectorType& vec, const typename MatrixType::RealScalar& sigma) \
{ return Eigen::internal::llt_rank_update_lower(mat, vec, sigma); } \
}; \
template<> struct llt_inplace<EIGTYPE, Upper> \
{ \
template<typename MatrixType> \
static Index blocked(MatrixType& m) \
{ \
return lapacke_llt<EIGTYPE>::potrf(m, 'U'); \
} \
template<typename MatrixType, typename VectorType> \
static Index rankUpdate(MatrixType& mat, const VectorType& vec, const typename MatrixType::RealScalar& sigma) \
{ \
Transpose<MatrixType> matt(mat); \
return llt_inplace<EIGTYPE, Lower>::rankUpdate(matt, vec.conjugate(), sigma); \
} \
};
EIGEN_LAPACKE_LLT(double, double, d)
EIGEN_LAPACKE_LLT(float, float, s)
EIGEN_LAPACKE_LLT(dcomplex, lapack_complex_double, z)
EIGEN_LAPACKE_LLT(scomplex, lapack_complex_float, c)
} // end namespace internal
} // end namespace Eigen
#endif // EIGEN_LLT_LAPACKE_H
pydensecrf/densecrf/include/Eigen/src/Core/Array.h
View file @ 4c4419e0
......@@ -12,7 +12,16 @@
namespace Eigen {
/** \class Array
namespace internal {
template<typename _Scalar, int _Rows, int _Cols, int _Options, int _MaxRows, int _MaxCols>
struct traits<Array<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols> > : traits<Matrix<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols> >
{
typedef ArrayXpr XprKind;
typedef ArrayBase<Array<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols> > XprBase;
};
}
/** \class Array
* \ingroup Core_Module
*
* \brief General-purpose arrays with easy API for coefficient-wise operations
......@@ -24,20 +33,14 @@ namespace Eigen {
* API for the %Matrix class provides easy access to linear-algebra
* operations.
*
* See documentation of class Matrix for detailed information on the template parameters
* storage layout.
*
* This class can be extended with the help of the plugin mechanism described on the page
* \ref TopicCustomizingEigen by defining the preprocessor symbol \c EIGEN_ARRAY_PLUGIN.
* \ref TopicCustomizing_Plugins by defining the preprocessor symbol \c EIGEN_ARRAY_PLUGIN.
*
* \sa \ref TutorialArrayClass, \ref TopicClassHierarchy
* \sa \blank \ref TutorialArrayClass, \ref TopicClassHierarchy
*/
namespace internal {
template<typename _Scalar, int _Rows, int _Cols, int _Options, int _MaxRows, int _MaxCols>
struct traits<Array<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols> > : traits<Matrix<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols> >
{
typedef ArrayXpr XprKind;
typedef ArrayBase<Array<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols> > XprBase;
};
}
template<typename _Scalar, int _Rows, int _Cols, int _Options, int _MaxRows, int _MaxCols>
class Array
: public PlainObjectBase<Array<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols> >
......@@ -69,11 +72,27 @@ class Array
* the usage of 'using'. This should be done only for operator=.
*/
template<typename OtherDerived>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Array& operator=(const EigenBase<OtherDerived> &other)
{
return Base::operator=(other);
}
/** Set all the entries to \a value.
* \sa DenseBase::setConstant(), DenseBase::fill()
*/
/* This overload is needed because the usage of
* using Base::operator=;
* fails on MSVC. Since the code below is working with GCC and MSVC, we skipped
* the usage of 'using'. This should be done only for operator=.
*/
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Array& operator=(const Scalar &value)
{
Base::setConstant(value);
return *this;
}
/** Copies the value of the expression \a other into \c *this with automatic resizing.
*
* *this might be resized to match the dimensions of \a other. If *this was a null matrix (not already initialized),
......@@ -84,7 +103,8 @@ class Array
* remain row-vectors and vectors remain vectors.
*/
template<typename OtherDerived>
EIGEN_STRONG_INLINE Array& operator=(const ArrayBase<OtherDerived>& other)
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Array& operator=(const DenseBase<OtherDerived>& other)
{
return Base::_set(other);
}
......@@ -92,11 +112,12 @@ class Array
/** This is a special case of the templated operator=. Its purpose is to
* prevent a default operator= from hiding the templated operator=.
*/
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Array& operator=(const Array& other)
{
return Base::_set(other);
}
/** Default constructor.
*
* For fixed-size matrices, does nothing.
......@@ -107,122 +128,122 @@ class Array
*
* \sa resize(Index,Index)
*/
EIGEN_STRONG_INLINE explicit Array() : Base()
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Array() : Base()
{
Base::_check_template_params();
EIGEN_INITIALIZE_BY_ZERO_IF_THAT_OPTION_IS_ENABLED
EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED
}
#ifndef EIGEN_PARSED_BY_DOXYGEN
// FIXME is it still needed ??
/** \internal */
EIGEN_DEVICE_FUNC
Array(internal::constructor_without_unaligned_array_assert)
: Base(internal::constructor_without_unaligned_array_assert())
{
Base::_check_template_params();
EIGEN_INITIALIZE_BY_ZERO_IF_THAT_OPTION_IS_ENABLED
EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED
}
#endif
/** Constructs a vector or row-vector with given dimension. \only_for_vectors
*
* Note that this is only useful for dynamic-size vectors. For fixed-size vectors,
* it is redundant to pass the dimension here, so it makes more sense to use the default
* constructor Matrix() instead.
*/
EIGEN_STRONG_INLINE explicit Array(Index dim)
: Base(dim, RowsAtCompileTime == 1 ? 1 : dim, ColsAtCompileTime == 1 ? 1 : dim)
#if EIGEN_HAS_RVALUE_REFERENCES
EIGEN_DEVICE_FUNC
Array(Array&& other) EIGEN_NOEXCEPT_IF(std::is_nothrow_move_constructible<Scalar>::value)
: Base(std::move(other))
{
Base::_check_template_params();
EIGEN_STATIC_ASSERT_VECTOR_ONLY(Array)
eigen_assert(dim >= 0);
eigen_assert(SizeAtCompileTime == Dynamic || SizeAtCompileTime == dim);
EIGEN_INITIALIZE_BY_ZERO_IF_THAT_OPTION_IS_ENABLED
}
EIGEN_DEVICE_FUNC
Array& operator=(Array&& other) EIGEN_NOEXCEPT_IF(std::is_nothrow_move_assignable<Scalar>::value)
{
other.swap(*this);
return *this;
}
#endif
#ifndef EIGEN_PARSED_BY_DOXYGEN
template<typename T>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE explicit Array(const T& x)
{
Base::_check_template_params();
Base::template _init1<T>(x);
}
template<typename T0, typename T1>
EIGEN_STRONG_INLINE Array(const T0& x, const T1& y)
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Array(const T0& val0, const T1& val1)
{
Base::_check_template_params();
this->template _init2<T0,T1>(x, y);
this->template _init2<T0,T1>(val0, val1);
}
#else
/** constructs an uninitialized matrix with \a rows rows and \a cols columns.
/** \brief Constructs a fixed-sized array initialized with coefficients starting at \a data */
EIGEN_DEVICE_FUNC explicit Array(const Scalar *data);
/** Constructs a vector or row-vector with given dimension. \only_for_vectors
*
* This is useful for dynamic-size matrices. For fixed-size matrices,
* Note that this is only useful for dynamic-size vectors. For fixed-size vectors,
* it is redundant to pass the dimension here, so it makes more sense to use the default
* constructor Array() instead.
*/
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE explicit Array(Index dim);
/** constructs an initialized 1x1 Array with the given coefficient */
Array(const Scalar& value);
/** constructs an uninitialized array with \a rows rows and \a cols columns.
*
* This is useful for dynamic-size arrays. For fixed-size arrays,
* it is redundant to pass these parameters, so one should use the default constructor
* Matrix() instead. */
* Array() instead. */
Array(Index rows, Index cols);
/** constructs an initialized 2D vector with given coefficients */
Array(const Scalar& x, const Scalar& y);
Array(const Scalar& val0, const Scalar& val1);
#endif
/** constructs an initialized 3D vector with given coefficients */
EIGEN_STRONG_INLINE Array(const Scalar& x, const Scalar& y, const Scalar& z)
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Array(const Scalar& val0, const Scalar& val1, const Scalar& val2)
{
Base::_check_template_params();
EIGEN_STATIC_ASSERT_VECTOR_SPECIFIC_SIZE(Array, 3)
m_storage.data()[0] = x;
m_storage.data()[1] = y;
m_storage.data()[2] = z;
m_storage.data()[0] = val0;
m_storage.data()[1] = val1;
m_storage.data()[2] = val2;
}
/** constructs an initialized 4D vector with given coefficients */
EIGEN_STRONG_INLINE Array(const Scalar& x, const Scalar& y, const Scalar& z, const Scalar& w)
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Array(const Scalar& val0, const Scalar& val1, const Scalar& val2, const Scalar& val3)
{
Base::_check_template_params();
EIGEN_STATIC_ASSERT_VECTOR_SPECIFIC_SIZE(Array, 4)
m_storage.data()[0] = x;
m_storage.data()[1] = y;
m_storage.data()[2] = z;
m_storage.data()[3] = w;
m_storage.data()[0] = val0;
m_storage.data()[1] = val1;
m_storage.data()[2] = val2;
m_storage.data()[3] = val3;
}
explicit Array(const Scalar *data);
/** Constructor copying the value of the expression \a other */
template<typename OtherDerived>
EIGEN_STRONG_INLINE Array(const ArrayBase<OtherDerived>& other)
: Base(other.rows() * other.cols(), other.rows(), other.cols())
{
Base::_check_template_params();
Base::_set_noalias(other);
}
/** Copy constructor */
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Array(const Array& other)
: Base(other.rows() * other.cols(), other.rows(), other.cols())
{
Base::_check_template_params();
Base::_set_noalias(other);
}
/** Copy constructor with in-place evaluation */
template<typename OtherDerived>
EIGEN_STRONG_INLINE Array(const ReturnByValue<OtherDerived>& other)
{
Base::_check_template_params();
Base::resize(other.rows(), other.cols());
other.evalTo(*this);
}
: Base(other)
{ }
/** \sa MatrixBase::operator=(const EigenBase<OtherDerived>&) */
template<typename OtherDerived>
EIGEN_STRONG_INLINE Array(const EigenBase<OtherDerived> &other)
: Base(other.derived().rows() * other.derived().cols(), other.derived().rows(), other.derived().cols())
{
Base::_check_template_params();
Base::resize(other.rows(), other.cols());
*this = other;
}
private:
struct PrivateType {};
public:
/** Override MatrixBase::swap() since for dynamic-sized matrices of same type it is enough to swap the
* data pointers.
*/
/** \sa MatrixBase::operator=(const EigenBase<OtherDerived>&) */
template<typename OtherDerived>
void swap(ArrayBase<OtherDerived> const & other)
{ this->_swap(other.derived()); }
inline Index innerStride() const { return 1; }
inline Index outerStride() const { return this->innerSize(); }
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Array(const EigenBase<OtherDerived> &other,
typename internal::enable_if<internal::is_convertible<typename OtherDerived::Scalar,Scalar>::value,
PrivateType>::type = PrivateType())
: Base(other.derived())
{ }
EIGEN_DEVICE_FUNC inline Index innerStride() const { return 1; }
EIGEN_DEVICE_FUNC inline Index outerStride() const { return this->innerSize(); }
#ifdef EIGEN_ARRAY_PLUGIN
#include EIGEN_ARRAY_PLUGIN
......
pydensecrf/densecrf/include/Eigen/src/Core/ArrayBase.h
View file @ 4c4419e0
......@@ -32,7 +32,7 @@ template<typename ExpressionType> class MatrixWrapper;
* \tparam Derived is the derived type, e.g., an array or an expression type.
*
* This class can be extended with the help of the plugin mechanism described on the page
* \ref TopicCustomizingEigen by defining the preprocessor symbol \c EIGEN_ARRAYBASE_PLUGIN.
* \ref TopicCustomizing_Plugins by defining the preprocessor symbol \c EIGEN_ARRAYBASE_PLUGIN.
*
* \sa class MatrixBase, \ref TopicClassHierarchy
*/
......@@ -46,11 +46,7 @@ template<typename Derived> class ArrayBase
typedef ArrayBase Eigen_BaseClassForSpecializationOfGlobalMathFuncImpl;
using internal::special_scalar_op_base<Derived,typename internal::traits<Derived>::Scalar,
typename NumTraits<typename internal::traits<Derived>::Scalar>::Real>::operator*;
typedef typename internal::traits<Derived>::StorageKind StorageKind;
typedef typename internal::traits<Derived>::Index Index;
typedef typename internal::traits<Derived>::Scalar Scalar;
typedef typename internal::packet_traits<Scalar>::type PacketScalar;
typedef typename NumTraits<Scalar>::Real RealScalar;
......@@ -64,8 +60,7 @@ template<typename Derived> class ArrayBase
using Base::MaxSizeAtCompileTime;
using Base::IsVectorAtCompileTime;
using Base::Flags;
using Base::CoeffReadCost;
using Base::derived;
using Base::const_cast_derived;
using Base::rows;
......@@ -85,25 +80,14 @@ template<typename Derived> class ArrayBase
#endif // not EIGEN_PARSED_BY_DOXYGEN
#ifndef EIGEN_PARSED_BY_DOXYGEN
/** \internal the plain matrix type corresponding to this expression. Note that is not necessarily
* exactly the return type of eval(): in the case of plain matrices, the return type of eval() is a const
* reference to a matrix, not a matrix! It is however guaranteed that the return type of eval() is either
* PlainObject or const PlainObject&.
*/
typedef Array<typename internal::traits<Derived>::Scalar,
internal::traits<Derived>::RowsAtCompileTime,
internal::traits<Derived>::ColsAtCompileTime,
AutoAlign | (internal::traits<Derived>::Flags&RowMajorBit ? RowMajor : ColMajor),
internal::traits<Derived>::MaxRowsAtCompileTime,
internal::traits<Derived>::MaxColsAtCompileTime
> PlainObject;
typedef typename Base::PlainObject PlainObject;
/** \internal Represents a matrix with all coefficients equal to one another*/
typedef CwiseNullaryOp<internal::scalar_constant_op<Scalar>,Derived> ConstantReturnType;
typedef CwiseNullaryOp<internal::scalar_constant_op<Scalar>,PlainObject> ConstantReturnType;
#endif // not EIGEN_PARSED_BY_DOXYGEN
#define EIGEN_CURRENT_STORAGE_BASE_CLASS Eigen::ArrayBase
#define EIGEN_DOC_UNARY_ADDONS(X,Y)
# include "../plugins/CommonCwiseUnaryOps.h"
# include "../plugins/MatrixCwiseUnaryOps.h"
# include "../plugins/ArrayCwiseUnaryOps.h"
......@@ -114,45 +98,63 @@ template<typename Derived> class ArrayBase
# include EIGEN_ARRAYBASE_PLUGIN
# endif
#undef EIGEN_CURRENT_STORAGE_BASE_CLASS
#undef EIGEN_DOC_UNARY_ADDONS
/** Special case of the template operator=, in order to prevent the compiler
* from generating a default operator= (issue hit with g++ 4.1)
*/
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
Derived& operator=(const ArrayBase& other)
{
return internal::assign_selector<Derived,Derived>::run(derived(), other.derived());
internal::call_assignment(derived(), other.derived());
return derived();
}
Derived& operator+=(const Scalar& scalar)
{ return *this = derived() + scalar; }
Derived& operator-=(const Scalar& scalar)
{ return *this = derived() - scalar; }
/** Set all the entries to \a value.
* \sa DenseBase::setConstant(), DenseBase::fill() */
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
Derived& operator=(const Scalar &value)
{ Base::setConstant(value); return derived(); }
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
Derived& operator+=(const Scalar& scalar);
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
Derived& operator-=(const Scalar& scalar);
template<typename OtherDerived>
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
Derived& operator+=(const ArrayBase<OtherDerived>& other);
template<typename OtherDerived>
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
Derived& operator-=(const ArrayBase<OtherDerived>& other);
template<typename OtherDerived>
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
Derived& operator*=(const ArrayBase<OtherDerived>& other);
template<typename OtherDerived>
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
Derived& operator/=(const ArrayBase<OtherDerived>& other);
public:
EIGEN_DEVICE_FUNC
ArrayBase<Derived>& array() { return *this; }
EIGEN_DEVICE_FUNC
const ArrayBase<Derived>& array() const { return *this; }
/** \returns an \link MatrixBase Matrix \endlink expression of this array
/** \returns an \link Eigen::MatrixBase Matrix \endlink expression of this array
* \sa MatrixBase::array() */
MatrixWrapper<Derived> matrix() { return derived(); }
const MatrixWrapper<const Derived> matrix() const { return derived(); }
EIGEN_DEVICE_FUNC
MatrixWrapper<Derived> matrix() { return MatrixWrapper<Derived>(derived()); }
EIGEN_DEVICE_FUNC
const MatrixWrapper<const Derived> matrix() const { return MatrixWrapper<const Derived>(derived()); }
// template<typename Dest>
// inline void evalTo(Dest& dst) const { dst = matrix(); }
protected:
ArrayBase() : Base() {}
EIGEN_DEFAULT_COPY_CONSTRUCTOR(ArrayBase)
EIGEN_DEFAULT_EMPTY_CONSTRUCTOR_AND_DESTRUCTOR(ArrayBase)
private:
explicit ArrayBase(Index);
......@@ -173,11 +175,10 @@ template<typename Derived> class ArrayBase
*/
template<typename Derived>
template<typename OtherDerived>
EIGEN_STRONG_INLINE Derived &
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived &
ArrayBase<Derived>::operator-=(const ArrayBase<OtherDerived> &other)
{
SelfCwiseBinaryOp<internal::scalar_difference_op<Scalar>, Derived, OtherDerived> tmp(derived());
tmp = other.derived();
call_assignment(derived(), other.derived(), internal::sub_assign_op<Scalar,typename OtherDerived::Scalar>());
return derived();
}
......@@ -187,11 +188,10 @@ ArrayBase<Derived>::operator-=(const ArrayBase<OtherDerived> &other)
*/
template<typename Derived>
template<typename OtherDerived>
EIGEN_STRONG_INLINE Derived &
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived &
ArrayBase<Derived>::operator+=(const ArrayBase<OtherDerived>& other)
{
SelfCwiseBinaryOp<internal::scalar_sum_op<Scalar>, Derived, OtherDerived> tmp(derived());
tmp = other.derived();
call_assignment(derived(), other.derived(), internal::add_assign_op<Scalar,typename OtherDerived::Scalar>());
return derived();
}
......@@ -201,11 +201,10 @@ ArrayBase<Derived>::operator+=(const ArrayBase<OtherDerived>& other)
*/
template<typename Derived>
template<typename OtherDerived>
EIGEN_STRONG_INLINE Derived &
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived &
ArrayBase<Derived>::operator*=(const ArrayBase<OtherDerived>& other)
{
SelfCwiseBinaryOp<internal::scalar_product_op<Scalar>, Derived, OtherDerived> tmp(derived());
tmp = other.derived();
call_assignment(derived(), other.derived(), internal::mul_assign_op<Scalar,typename OtherDerived::Scalar>());
return derived();
}
......@@ -215,11 +214,10 @@ ArrayBase<Derived>::operator*=(const ArrayBase<OtherDerived>& other)
*/
template<typename Derived>
template<typename OtherDerived>
EIGEN_STRONG_INLINE Derived &
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived &
ArrayBase<Derived>::operator/=(const ArrayBase<OtherDerived>& other)
{
SelfCwiseBinaryOp<internal::scalar_quotient_op<Scalar>, Derived, OtherDerived> tmp(derived());
tmp = other.derived();
call_assignment(derived(), other.derived(), internal::div_assign_op<Scalar,typename OtherDerived::Scalar>());
return derived();
}
......
pydensecrf/densecrf/include/Eigen/src/Core/ArrayWrapper.h
View file @ 4c4419e0
......@@ -29,6 +29,12 @@ struct traits<ArrayWrapper<ExpressionType> >
: public traits<typename remove_all<typename ExpressionType::Nested>::type >
{
typedef ArrayXpr XprKind;
// Let's remove NestByRefBit
enum {
Flags0 = traits<typename remove_all<typename ExpressionType::Nested>::type >::Flags,
LvalueBitFlag = is_lvalue<ExpressionType>::value ? LvalueBit : 0,
Flags = (Flags0 & ~(NestByRefBit | LvalueBit)) | LvalueBitFlag
};
};
}
......@@ -39,6 +45,7 @@ class ArrayWrapper : public ArrayBase<ArrayWrapper<ExpressionType> >
typedef ArrayBase<ArrayWrapper> Base;
EIGEN_DENSE_PUBLIC_INTERFACE(ArrayWrapper)
EIGEN_INHERIT_ASSIGNMENT_OPERATORS(ArrayWrapper)
typedef typename internal::remove_all<ExpressionType>::type NestedExpression;
typedef typename internal::conditional<
internal::is_lvalue<ExpressionType>::value,
......@@ -46,76 +53,45 @@ class ArrayWrapper : public ArrayBase<ArrayWrapper<ExpressionType> >
const Scalar
>::type ScalarWithConstIfNotLvalue;
typedef typename internal::nested<ExpressionType>::type NestedExpressionType;
typedef typename internal::ref_selector<ExpressionType>::non_const_type NestedExpressionType;
inline ArrayWrapper(ExpressionType& matrix) : m_expression(matrix) {}
using Base::coeffRef;
EIGEN_DEVICE_FUNC
explicit EIGEN_STRONG_INLINE ArrayWrapper(ExpressionType& matrix) : m_expression(matrix) {}
EIGEN_DEVICE_FUNC
inline Index rows() const { return m_expression.rows(); }
EIGEN_DEVICE_FUNC
inline Index cols() const { return m_expression.cols(); }
EIGEN_DEVICE_FUNC
inline Index outerStride() const { return m_expression.outerStride(); }
EIGEN_DEVICE_FUNC
inline Index innerStride() const { return m_expression.innerStride(); }
EIGEN_DEVICE_FUNC
inline ScalarWithConstIfNotLvalue* data() { return m_expression.data(); }
EIGEN_DEVICE_FUNC
inline const Scalar* data() const { return m_expression.data(); }
inline CoeffReturnType coeff(Index row, Index col) const
{
return m_expression.coeff(row, col);
}
inline Scalar& coeffRef(Index row, Index col)
{
return m_expression.const_cast_derived().coeffRef(row, col);
}
inline const Scalar& coeffRef(Index row, Index col) const
{
return m_expression.const_cast_derived().coeffRef(row, col);
}
inline CoeffReturnType coeff(Index index) const
{
return m_expression.coeff(index);
}
inline Scalar& coeffRef(Index index)
EIGEN_DEVICE_FUNC
inline const Scalar& coeffRef(Index rowId, Index colId) const
{
return m_expression.const_cast_derived().coeffRef(index);
return m_expression.coeffRef(rowId, colId);
}
EIGEN_DEVICE_FUNC
inline const Scalar& coeffRef(Index index) const
{
return m_expression.const_cast_derived().coeffRef(index);
}
template<int LoadMode>
inline const PacketScalar packet(Index row, Index col) const
{
return m_expression.template packet<LoadMode>(row, col);
}
template<int LoadMode>
inline void writePacket(Index row, Index col, const PacketScalar& x)
{
m_expression.const_cast_derived().template writePacket<LoadMode>(row, col, x);
}
template<int LoadMode>
inline const PacketScalar packet(Index index) const
{
return m_expression.template packet<LoadMode>(index);
}
template<int LoadMode>
inline void writePacket(Index index, const PacketScalar& x)
{
m_expression.const_cast_derived().template writePacket<LoadMode>(index, x);
return m_expression.coeffRef(index);
}
template<typename Dest>
EIGEN_DEVICE_FUNC
inline void evalTo(Dest& dst) const { dst = m_expression; }
const typename internal::remove_all<NestedExpressionType>::type&
EIGEN_DEVICE_FUNC
nestedExpression() const
{
return m_expression;
......@@ -123,10 +99,12 @@ class ArrayWrapper : public ArrayBase<ArrayWrapper<ExpressionType> >
/** Forwards the resizing request to the nested expression
* \sa DenseBase::resize(Index) */
void resize(Index newSize) { m_expression.const_cast_derived().resize(newSize); }
EIGEN_DEVICE_FUNC
void resize(Index newSize) { m_expression.resize(newSize); }
/** Forwards the resizing request to the nested expression
* \sa DenseBase::resize(Index,Index)*/
void resize(Index nbRows, Index nbCols) { m_expression.const_cast_derived().resize(nbRows,nbCols); }
EIGEN_DEVICE_FUNC
void resize(Index rows, Index cols) { m_expression.resize(rows,cols); }
protected:
NestedExpressionType m_expression;
......@@ -149,6 +127,12 @@ struct traits<MatrixWrapper<ExpressionType> >
: public traits<typename remove_all<typename ExpressionType::Nested>::type >
{
typedef MatrixXpr XprKind;
// Let's remove NestByRefBit
enum {
Flags0 = traits<typename remove_all<typename ExpressionType::Nested>::type >::Flags,
LvalueBitFlag = is_lvalue<ExpressionType>::value ? LvalueBit : 0,
Flags = (Flags0 & ~(NestByRefBit | LvalueBit)) | LvalueBitFlag
};
};
}
......@@ -159,6 +143,7 @@ class MatrixWrapper : public MatrixBase<MatrixWrapper<ExpressionType> >
typedef MatrixBase<MatrixWrapper<ExpressionType> > Base;
EIGEN_DENSE_PUBLIC_INTERFACE(MatrixWrapper)
EIGEN_INHERIT_ASSIGNMENT_OPERATORS(MatrixWrapper)
typedef typename internal::remove_all<ExpressionType>::type NestedExpression;
typedef typename internal::conditional<
internal::is_lvalue<ExpressionType>::value,
......@@ -166,72 +151,40 @@ class MatrixWrapper : public MatrixBase<MatrixWrapper<ExpressionType> >
const Scalar
>::type ScalarWithConstIfNotLvalue;
typedef typename internal::nested<ExpressionType>::type NestedExpressionType;
typedef typename internal::ref_selector<ExpressionType>::non_const_type NestedExpressionType;
inline MatrixWrapper(ExpressionType& matrix) : m_expression(matrix) {}
using Base::coeffRef;
EIGEN_DEVICE_FUNC
explicit inline MatrixWrapper(ExpressionType& matrix) : m_expression(matrix) {}
EIGEN_DEVICE_FUNC
inline Index rows() const { return m_expression.rows(); }
EIGEN_DEVICE_FUNC
inline Index cols() const { return m_expression.cols(); }
EIGEN_DEVICE_FUNC
inline Index outerStride() const { return m_expression.outerStride(); }
EIGEN_DEVICE_FUNC
inline Index innerStride() const { return m_expression.innerStride(); }
EIGEN_DEVICE_FUNC
inline ScalarWithConstIfNotLvalue* data() { return m_expression.data(); }
EIGEN_DEVICE_FUNC
inline const Scalar* data() const { return m_expression.data(); }
inline CoeffReturnType coeff(Index row, Index col) const
{
return m_expression.coeff(row, col);
}
inline Scalar& coeffRef(Index row, Index col)
{
return m_expression.const_cast_derived().coeffRef(row, col);
}
inline const Scalar& coeffRef(Index row, Index col) const
{
return m_expression.derived().coeffRef(row, col);
}
inline CoeffReturnType coeff(Index index) const
{
return m_expression.coeff(index);
}
inline Scalar& coeffRef(Index index)
EIGEN_DEVICE_FUNC
inline const Scalar& coeffRef(Index rowId, Index colId) const
{
return m_expression.const_cast_derived().coeffRef(index);
return m_expression.derived().coeffRef(rowId, colId);
}
EIGEN_DEVICE_FUNC
inline const Scalar& coeffRef(Index index) const
{
return m_expression.const_cast_derived().coeffRef(index);
}
template<int LoadMode>
inline const PacketScalar packet(Index row, Index col) const
{
return m_expression.template packet<LoadMode>(row, col);
}
template<int LoadMode>
inline void writePacket(Index row, Index col, const PacketScalar& x)
{
m_expression.const_cast_derived().template writePacket<LoadMode>(row, col, x);
}
template<int LoadMode>
inline const PacketScalar packet(Index index) const
{
return m_expression.template packet<LoadMode>(index);
}
template<int LoadMode>
inline void writePacket(Index index, const PacketScalar& x)
{
m_expression.const_cast_derived().template writePacket<LoadMode>(index, x);
return m_expression.coeffRef(index);
}
EIGEN_DEVICE_FUNC
const typename internal::remove_all<NestedExpressionType>::type&
nestedExpression() const
{
......@@ -240,10 +193,12 @@ class MatrixWrapper : public MatrixBase<MatrixWrapper<ExpressionType> >
/** Forwards the resizing request to the nested expression
* \sa DenseBase::resize(Index) */
void resize(Index newSize) { m_expression.const_cast_derived().resize(newSize); }
EIGEN_DEVICE_FUNC
void resize(Index newSize) { m_expression.resize(newSize); }
/** Forwards the resizing request to the nested expression
* \sa DenseBase::resize(Index,Index)*/
void resize(Index nbRows, Index nbCols) { m_expression.const_cast_derived().resize(nbRows,nbCols); }
EIGEN_DEVICE_FUNC
void resize(Index rows, Index cols) { m_expression.resize(rows,cols); }
protected:
NestedExpressionType m_expression;
......
pydensecrf/densecrf/include/Eigen/src/Core/AssignEvaluator.h 0 → 100644
View file @ 4c4419e0
This diff is collapsed.
pydensecrf/densecrf/include/Eigen/src/Core/Assign_MKL.h
View file @ 4c4419e0
/*
Copyright (c) 2011, Intel Corporation. All rights reserved.
Copyright (C) 2015 Gael Guennebaud <gael.guennebaud@inria.fr>
Redistribution and use in source and binary forms, with or without modification,
are permitted provided that the following conditions are met:
......@@ -37,17 +38,13 @@ namespace Eigen {
namespace internal {
template<typename Op> struct vml_call
{ enum { IsSupported = 0 }; };
template<typename Dst, typename Src, typename UnaryOp>
template<typename Dst, typename Src>
class vml_assign_traits
{
private:
enum {
DstHasDirectAccess = Dst::Flags & DirectAccessBit,
SrcHasDirectAccess = Src::Flags & DirectAccessBit,
StorageOrdersAgree = (int(Dst::IsRowMajor) == int(Src::IsRowMajor)),
InnerSize = int(Dst::IsVectorAtCompileTime) ? int(Dst::SizeAtCompileTime)
: int(Dst::Flags)&RowMajorBit ? int(Dst::ColsAtCompileTime)
......@@ -57,165 +54,122 @@ class vml_assign_traits
: int(Dst::MaxRowsAtCompileTime),
MaxSizeAtCompileTime = Dst::SizeAtCompileTime,
MightEnableVml = vml_call<UnaryOp>::IsSupported && StorageOrdersAgree && DstHasDirectAccess && SrcHasDirectAccess
&& Src::InnerStrideAtCompileTime==1 && Dst::InnerStrideAtCompileTime==1,
MightEnableVml = StorageOrdersAgree && DstHasDirectAccess && SrcHasDirectAccess && Src::InnerStrideAtCompileTime==1 && Dst::InnerStrideAtCompileTime==1,
MightLinearize = MightEnableVml && (int(Dst::Flags) & int(Src::Flags) & LinearAccessBit),
VmlSize = MightLinearize ? MaxSizeAtCompileTime : InnerMaxSize,
LargeEnough = VmlSize==Dynamic || VmlSize>=EIGEN_MKL_VML_THRESHOLD,
MayEnableVml = MightEnableVml && LargeEnough,
MayLinearize = MayEnableVml && MightLinearize
LargeEnough = VmlSize==Dynamic || VmlSize>=EIGEN_MKL_VML_THRESHOLD
};
public:
enum {
Traversal = MayLinearize ? LinearVectorizedTraversal
: MayEnableVml ? InnerVectorizedTraversal
: DefaultTraversal
EnableVml = MightEnableVml && LargeEnough,
Traversal = MightLinearize ? LinearTraversal : DefaultTraversal
};
};
template<typename Derived1, typename Derived2, typename UnaryOp, int Traversal, int Unrolling,
int VmlTraversal = vml_assign_traits<Derived1, Derived2, UnaryOp>::Traversal >
struct vml_assign_impl
: assign_impl<Derived1, Eigen::CwiseUnaryOp<UnaryOp, Derived2>,Traversal,Unrolling,BuiltIn>
{
};
template<typename Derived1, typename Derived2, typename UnaryOp, int Traversal, int Unrolling>
struct vml_assign_impl<Derived1, Derived2, UnaryOp, Traversal, Unrolling, InnerVectorizedTraversal>
{
typedef typename Derived1::Scalar Scalar;
typedef typename Derived1::Index Index;
static inline void run(Derived1& dst, const CwiseUnaryOp<UnaryOp, Derived2>& src)
{
// in case we want to (or have to) skip VML at runtime we can call:
// assign_impl<Derived1,Eigen::CwiseUnaryOp<UnaryOp, Derived2>,Traversal,Unrolling,BuiltIn>::run(dst,src);
const Index innerSize = dst.innerSize();
const Index outerSize = dst.outerSize();
for(Index outer = 0; outer < outerSize; ++outer) {
const Scalar *src_ptr = src.IsRowMajor ? &(src.nestedExpression().coeffRef(outer,0)) :
&(src.nestedExpression().coeffRef(0, outer));
Scalar *dst_ptr = dst.IsRowMajor ? &(dst.coeffRef(outer,0)) : &(dst.coeffRef(0, outer));
vml_call<UnaryOp>::run(src.functor(), innerSize, src_ptr, dst_ptr );
}
}
};
template<typename Derived1, typename Derived2, typename UnaryOp, int Traversal, int Unrolling>
struct vml_assign_impl<Derived1, Derived2, UnaryOp, Traversal, Unrolling, LinearVectorizedTraversal>
{
static inline void run(Derived1& dst, const CwiseUnaryOp<UnaryOp, Derived2>& src)
{
// in case we want to (or have to) skip VML at runtime we can call:
// assign_impl<Derived1,Eigen::CwiseUnaryOp<UnaryOp, Derived2>,Traversal,Unrolling,BuiltIn>::run(dst,src);
vml_call<UnaryOp>::run(src.functor(), dst.size(), src.nestedExpression().data(), dst.data() );
}
};
// Macroses
#define EIGEN_MKL_VML_SPECIALIZE_ASSIGN(TRAVERSAL,UNROLLING) \
template<typename Derived1, typename Derived2, typename UnaryOp> \
struct assign_impl<Derived1, Eigen::CwiseUnaryOp<UnaryOp, Derived2>, TRAVERSAL, UNROLLING, Specialized> { \
static inline void run(Derived1 &dst, const Eigen::CwiseUnaryOp<UnaryOp, Derived2> &src) { \
vml_assign_impl<Derived1,Derived2,UnaryOp,TRAVERSAL,UNROLLING>::run(dst, src); \
} \
};
EIGEN_MKL_VML_SPECIALIZE_ASSIGN(DefaultTraversal,NoUnrolling)
EIGEN_MKL_VML_SPECIALIZE_ASSIGN(DefaultTraversal,CompleteUnrolling)
EIGEN_MKL_VML_SPECIALIZE_ASSIGN(DefaultTraversal,InnerUnrolling)
EIGEN_MKL_VML_SPECIALIZE_ASSIGN(LinearTraversal,NoUnrolling)
EIGEN_MKL_VML_SPECIALIZE_ASSIGN(LinearTraversal,CompleteUnrolling)
EIGEN_MKL_VML_SPECIALIZE_ASSIGN(InnerVectorizedTraversal,NoUnrolling)
EIGEN_MKL_VML_SPECIALIZE_ASSIGN(InnerVectorizedTraversal,CompleteUnrolling)
EIGEN_MKL_VML_SPECIALIZE_ASSIGN(InnerVectorizedTraversal,InnerUnrolling)
EIGEN_MKL_VML_SPECIALIZE_ASSIGN(LinearVectorizedTraversal,CompleteUnrolling)
EIGEN_MKL_VML_SPECIALIZE_ASSIGN(LinearVectorizedTraversal,NoUnrolling)
EIGEN_MKL_VML_SPECIALIZE_ASSIGN(SliceVectorizedTraversal,NoUnrolling)
#define EIGEN_PP_EXPAND(ARG) ARG
#if !defined (EIGEN_FAST_MATH) || (EIGEN_FAST_MATH != 1)
#define EIGEN_MKL_VML_MODE VML_HA
#define EIGEN_VMLMODE_EXPAND_LA , VML_HA
#else
#define EIGEN_MKL_VML_MODE VML_LA
#define EIGEN_VMLMODE_EXPAND_LA , VML_LA
#endif
#define EIGEN_MKL_VML_DECLARE_UNARY_CALL(EIGENOP, VMLOP, EIGENTYPE, VMLTYPE) \
template<> struct vml_call< scalar_##EIGENOP##_op<EIGENTYPE> > { \
enum { IsSupported = 1 }; \
static inline void run( const scalar_##EIGENOP##_op<EIGENTYPE>& /*func*/, \
int size, const EIGENTYPE* src, EIGENTYPE* dst) { \
VMLOP(size, (const VMLTYPE*)src, (VMLTYPE*)dst); \
} \
#define EIGEN_VMLMODE_EXPAND__
#define EIGEN_VMLMODE_PREFIX_LA vm
#define EIGEN_VMLMODE_PREFIX__ v
#define EIGEN_VMLMODE_PREFIX(VMLMODE) EIGEN_CAT(EIGEN_VMLMODE_PREFIX_,VMLMODE)
#define EIGEN_MKL_VML_DECLARE_UNARY_CALL(EIGENOP, VMLOP, EIGENTYPE, VMLTYPE, VMLMODE) \
template< typename DstXprType, typename SrcXprNested> \
struct Assignment<DstXprType, CwiseUnaryOp<scalar_##EIGENOP##_op<EIGENTYPE>, SrcXprNested>, assign_op<EIGENTYPE,EIGENTYPE>, \
Dense2Dense, typename enable_if<vml_assign_traits<DstXprType,SrcXprNested>::EnableVml>::type> { \
typedef CwiseUnaryOp<scalar_##EIGENOP##_op<EIGENTYPE>, SrcXprNested> SrcXprType; \
static void run(DstXprType &dst, const SrcXprType &src, const assign_op<EIGENTYPE,EIGENTYPE> &func) { \
resize_if_allowed(dst, src, func); \
eigen_assert(dst.rows() == src.rows() && dst.cols() == src.cols()); \
if(vml_assign_traits<DstXprType,SrcXprNested>::Traversal==LinearTraversal) { \
VMLOP(dst.size(), (const VMLTYPE*)src.nestedExpression().data(), \
(VMLTYPE*)dst.data() EIGEN_PP_EXPAND(EIGEN_VMLMODE_EXPAND_##VMLMODE) ); \
} else { \
const Index outerSize = dst.outerSize(); \
for(Index outer = 0; outer < outerSize; ++outer) { \
const EIGENTYPE *src_ptr = src.IsRowMajor ? &(src.nestedExpression().coeffRef(outer,0)) : \
&(src.nestedExpression().coeffRef(0, outer)); \
EIGENTYPE *dst_ptr = dst.IsRowMajor ? &(dst.coeffRef(outer,0)) : &(dst.coeffRef(0, outer)); \
VMLOP( dst.innerSize(), (const VMLTYPE*)src_ptr, \
(VMLTYPE*)dst_ptr EIGEN_PP_EXPAND(EIGEN_VMLMODE_EXPAND_##VMLMODE)); \
} \
} \
} \
}; \
#define EIGEN_MKL_VML_DECLARE_UNARY_CALLS_REAL(EIGENOP, VMLOP, VMLMODE) \
EIGEN_MKL_VML_DECLARE_UNARY_CALL(EIGENOP, EIGEN_CAT(EIGEN_VMLMODE_PREFIX(VMLMODE),s##VMLOP), float, float, VMLMODE) \
EIGEN_MKL_VML_DECLARE_UNARY_CALL(EIGENOP, EIGEN_CAT(EIGEN_VMLMODE_PREFIX(VMLMODE),d##VMLOP), double, double, VMLMODE)
#define EIGEN_MKL_VML_DECLARE_UNARY_CALLS_CPLX(EIGENOP, VMLOP, VMLMODE) \
EIGEN_MKL_VML_DECLARE_UNARY_CALL(EIGENOP, EIGEN_CAT(EIGEN_VMLMODE_PREFIX(VMLMODE),c##VMLOP), scomplex, MKL_Complex8, VMLMODE) \
EIGEN_MKL_VML_DECLARE_UNARY_CALL(EIGENOP, EIGEN_CAT(EIGEN_VMLMODE_PREFIX(VMLMODE),z##VMLOP), dcomplex, MKL_Complex16, VMLMODE)
#define EIGEN_MKL_VML_DECLARE_UNARY_CALLS(EIGENOP, VMLOP, VMLMODE) \
EIGEN_MKL_VML_DECLARE_UNARY_CALLS_REAL(EIGENOP, VMLOP, VMLMODE) \
EIGEN_MKL_VML_DECLARE_UNARY_CALLS_CPLX(EIGENOP, VMLOP, VMLMODE)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS(sin, Sin, LA)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS(asin, Asin, LA)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS(sinh, Sinh, LA)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS(cos, Cos, LA)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS(acos, Acos, LA)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS(cosh, Cosh, LA)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS(tan, Tan, LA)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS(atan, Atan, LA)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS(tanh, Tanh, LA)
// EIGEN_MKL_VML_DECLARE_UNARY_CALLS(abs, Abs, _)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS(exp, Exp, LA)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS(log, Ln, LA)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS(log10, Log10, LA)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS(sqrt, Sqrt, _)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS_REAL(square, Sqr, _)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS_CPLX(arg, Arg, _)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS_REAL(round, Round, _)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS_REAL(floor, Floor, _)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS_REAL(ceil, Ceil, _)
#define EIGEN_MKL_VML_DECLARE_POW_CALL(EIGENOP, VMLOP, EIGENTYPE, VMLTYPE, VMLMODE) \
template< typename DstXprType, typename SrcXprNested, typename Plain> \
struct Assignment<DstXprType, CwiseBinaryOp<scalar_##EIGENOP##_op<EIGENTYPE,EIGENTYPE>, SrcXprNested, \
const CwiseNullaryOp<internal::scalar_constant_op<EIGENTYPE>,Plain> >, assign_op<EIGENTYPE,EIGENTYPE>, \
Dense2Dense, typename enable_if<vml_assign_traits<DstXprType,SrcXprNested>::EnableVml>::type> { \
typedef CwiseBinaryOp<scalar_##EIGENOP##_op<EIGENTYPE,EIGENTYPE>, SrcXprNested, \
const CwiseNullaryOp<internal::scalar_constant_op<EIGENTYPE>,Plain> > SrcXprType; \
static void run(DstXprType &dst, const SrcXprType &src, const assign_op<EIGENTYPE,EIGENTYPE> &func) { \
resize_if_allowed(dst, src, func); \
eigen_assert(dst.rows() == src.rows() && dst.cols() == src.cols()); \
VMLTYPE exponent = reinterpret_cast<const VMLTYPE&>(src.rhs().functor().m_other); \
if(vml_assign_traits<DstXprType,SrcXprNested>::Traversal==LinearTraversal) \
{ \
VMLOP( dst.size(), (const VMLTYPE*)src.lhs().data(), exponent, \
(VMLTYPE*)dst.data() EIGEN_PP_EXPAND(EIGEN_VMLMODE_EXPAND_##VMLMODE) ); \
} else { \
const Index outerSize = dst.outerSize(); \
for(Index outer = 0; outer < outerSize; ++outer) { \
const EIGENTYPE *src_ptr = src.IsRowMajor ? &(src.lhs().coeffRef(outer,0)) : \
&(src.lhs().coeffRef(0, outer)); \
EIGENTYPE *dst_ptr = dst.IsRowMajor ? &(dst.coeffRef(outer,0)) : &(dst.coeffRef(0, outer)); \
VMLOP( dst.innerSize(), (const VMLTYPE*)src_ptr, exponent, \
(VMLTYPE*)dst_ptr EIGEN_PP_EXPAND(EIGEN_VMLMODE_EXPAND_##VMLMODE)); \
} \
} \
} \
};
#define EIGEN_MKL_VML_DECLARE_UNARY_CALL_LA(EIGENOP, VMLOP, EIGENTYPE, VMLTYPE) \
template<> struct vml_call< scalar_##EIGENOP##_op<EIGENTYPE> > { \
enum { IsSupported = 1 }; \
static inline void run( const scalar_##EIGENOP##_op<EIGENTYPE>& /*func*/, \
int size, const EIGENTYPE* src, EIGENTYPE* dst) { \
MKL_INT64 vmlMode = EIGEN_MKL_VML_MODE; \
VMLOP(size, (const VMLTYPE*)src, (VMLTYPE*)dst, vmlMode); \
} \
};
#define EIGEN_MKL_VML_DECLARE_POW_CALL(EIGENOP, VMLOP, EIGENTYPE, VMLTYPE) \
template<> struct vml_call< scalar_##EIGENOP##_op<EIGENTYPE> > { \
enum { IsSupported = 1 }; \
static inline void run( const scalar_##EIGENOP##_op<EIGENTYPE>& func, \
int size, const EIGENTYPE* src, EIGENTYPE* dst) { \
EIGENTYPE exponent = func.m_exponent; \
MKL_INT64 vmlMode = EIGEN_MKL_VML_MODE; \
VMLOP(&size, (const VMLTYPE*)src, (const VMLTYPE*)&exponent, \
(VMLTYPE*)dst, &vmlMode); \
} \
};
#define EIGEN_MKL_VML_DECLARE_UNARY_CALLS_REAL(EIGENOP, VMLOP) \
EIGEN_MKL_VML_DECLARE_UNARY_CALL(EIGENOP, vs##VMLOP, float, float) \
EIGEN_MKL_VML_DECLARE_UNARY_CALL(EIGENOP, vd##VMLOP, double, double)
#define EIGEN_MKL_VML_DECLARE_UNARY_CALLS_COMPLEX(EIGENOP, VMLOP) \
EIGEN_MKL_VML_DECLARE_UNARY_CALL(EIGENOP, vc##VMLOP, scomplex, MKL_Complex8) \
EIGEN_MKL_VML_DECLARE_UNARY_CALL(EIGENOP, vz##VMLOP, dcomplex, MKL_Complex16)
#define EIGEN_MKL_VML_DECLARE_UNARY_CALLS(EIGENOP, VMLOP) \
EIGEN_MKL_VML_DECLARE_UNARY_CALLS_REAL(EIGENOP, VMLOP) \
EIGEN_MKL_VML_DECLARE_UNARY_CALLS_COMPLEX(EIGENOP, VMLOP)
#define EIGEN_MKL_VML_DECLARE_UNARY_CALLS_REAL_LA(EIGENOP, VMLOP) \
EIGEN_MKL_VML_DECLARE_UNARY_CALL_LA(EIGENOP, vms##VMLOP, float, float) \
EIGEN_MKL_VML_DECLARE_UNARY_CALL_LA(EIGENOP, vmd##VMLOP, double, double)
#define EIGEN_MKL_VML_DECLARE_UNARY_CALLS_COMPLEX_LA(EIGENOP, VMLOP) \
EIGEN_MKL_VML_DECLARE_UNARY_CALL_LA(EIGENOP, vmc##VMLOP, scomplex, MKL_Complex8) \
EIGEN_MKL_VML_DECLARE_UNARY_CALL_LA(EIGENOP, vmz##VMLOP, dcomplex, MKL_Complex16)
#define EIGEN_MKL_VML_DECLARE_UNARY_CALLS_LA(EIGENOP, VMLOP) \
EIGEN_MKL_VML_DECLARE_UNARY_CALLS_REAL_LA(EIGENOP, VMLOP) \
EIGEN_MKL_VML_DECLARE_UNARY_CALLS_COMPLEX_LA(EIGENOP, VMLOP)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS_LA(sin, Sin)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS_LA(asin, Asin)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS_LA(cos, Cos)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS_LA(acos, Acos)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS_LA(tan, Tan)
//EIGEN_MKL_VML_DECLARE_UNARY_CALLS(abs, Abs)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS_LA(exp, Exp)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS_LA(log, Ln)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS_LA(sqrt, Sqrt)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS_REAL(square, Sqr)
// The vm*powx functions are not avaibale in the windows version of MKL.
#ifdef _WIN32
EIGEN_MKL_VML_DECLARE_POW_CALL(pow, vmspowx_, float, float)
EIGEN_MKL_VML_DECLARE_POW_CALL(pow, vmdpowx_, double, double)
EIGEN_MKL_VML_DECLARE_POW_CALL(pow, vmcpowx_, scomplex, MKL_Complex8)
EIGEN_MKL_VML_DECLARE_POW_CALL(pow, vmzpowx_, dcomplex, MKL_Complex16)
#endif
EIGEN_MKL_VML_DECLARE_POW_CALL(pow, vmsPowx, float, float, LA)
EIGEN_MKL_VML_DECLARE_POW_CALL(pow, vmdPowx, double, double, LA)
EIGEN_MKL_VML_DECLARE_POW_CALL(pow, vmcPowx, scomplex, MKL_Complex8, LA)
EIGEN_MKL_VML_DECLARE_POW_CALL(pow, vmzPowx, dcomplex, MKL_Complex16, LA)
} // end namespace internal
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment