GitHub-Proxy/eigen
1
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2025-08-14 12:46:00 +08:00
Code Issues Packages Projects Releases Wiki Activity

Fix cuda device warnings

Browse Source
This commit is contained in:
Antonio Sanchez 2025-02-28 15:21:23 -08:00
parent c53002f5fb
commit 23b1682723
53 changed files with 655 additions and 600 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

25
CMakeLists.txt
View File

@ -176,7 +176,7 @@ if(NOT MSVC)
ei_add_cxx_compiler_flag("-Wall")
ei_add_cxx_compiler_flag("-Wextra")
#ei_add_cxx_compiler_flag("-Weverything") # clang
ei_add_cxx_compiler_flag("-Wundef")
ei_add_cxx_compiler_flag("-Wcast-align")
ei_add_cxx_compiler_flag("-Wchar-subscripts")
@ -191,29 +191,29 @@ if(NOT MSVC)
ei_add_cxx_compiler_flag("-Wc++11-extensions")
ei_add_cxx_compiler_flag("-Wdouble-promotion")
# ei_add_cxx_compiler_flag("-Wconversion")
# -Wshadow is insanely too strict with gcc, hopefully it will become usable with gcc 6
# if(NOT CMAKE_COMPILER_IS_GNUCXX OR (CMAKE_CXX_COMPILER_VERSION VERSION_GREATER "5.0.0"))
if(NOT CMAKE_COMPILER_IS_GNUCXX)
ei_add_cxx_compiler_flag("-Wshadow")
endif()
ei_add_cxx_compiler_flag("-Wno-psabi")
ei_add_cxx_compiler_flag("-Wno-variadic-macros")
ei_add_cxx_compiler_flag("-Wno-long-long")
ei_add_cxx_compiler_flag("-fno-check-new")
ei_add_cxx_compiler_flag("-fno-common")
ei_add_cxx_compiler_flag("-fstrict-aliasing")
ei_add_cxx_compiler_flag("-wd981") # disable ICC's "operands are evaluated in unspecified order" remark
ei_add_cxx_compiler_flag("-wd2304") # disable ICC's "warning #2304: non-explicit constructor with single argument may cause implicit type conversion" produced by -Wnon-virtual-dtor
# The -ansi flag must be added last, otherwise it is also used as a linker flag by check_cxx_compiler_flag making it fails
# Moreover we should not set both -strict-ansi and -ansi
check_cxx_compiler_flag("-strict-ansi" COMPILER_SUPPORT_STRICTANSI)
ei_add_cxx_compiler_flag("-Qunused-arguments") # disable clang warning: argument unused during compilation: '-ansi'
if(COMPILER_SUPPORT_STRICTANSI)
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -strict-ansi")
else()
@ -224,7 +224,7 @@ if(NOT MSVC)
ei_add_cxx_compiler_flag("-pie")
ei_add_cxx_compiler_flag("-fPIE")
endif()
set(CMAKE_REQUIRED_FLAGS "")
option(EIGEN_TEST_SSE2 "Enable/Disable SSE2 in tests/examples" OFF)
@ -398,6 +398,7 @@ if(EIGEN_TEST_NO_EXCEPTIONS)
message(STATUS "Disabling exceptions in tests/examples")
endif()
set(EIGEN_CUDA_CXX_FLAGS "" CACHE STRING "Additional flags to pass to the cuda compiler.")
set(EIGEN_CUDA_COMPUTE_ARCH 30 CACHE STRING "The CUDA compute architecture level to target when compiling CUDA code")
include_directories(${CMAKE_CURRENT_SOURCE_DIR} ${CMAKE_CURRENT_BINARY_DIR})
@ -600,11 +601,11 @@ if (NOT CMAKE_VERSION VERSION_LESS 3.0)
else (NOT CMAKE_VERSION VERSION_LESS 3.0)
# Fallback to legacy Eigen3Config.cmake without the imported target
# If CMakePackageConfigHelpers module is available (CMake >= 2.8.8)
# create a relocatable Config file, otherwise leave the hardcoded paths
# create a relocatable Config file, otherwise leave the hardcoded paths
include(CMakePackageConfigHelpers OPTIONAL RESULT_VARIABLE CPCH_PATH)
if(CPCH_PATH)
configure_package_config_file (
${CMAKE_CURRENT_SOURCE_DIR}/cmake/Eigen3ConfigLegacy.cmake.in
@ -613,7 +614,7 @@ else (NOT CMAKE_VERSION VERSION_LESS 3.0)
INSTALL_DESTINATION ${CMAKEPACKAGE_INSTALL_DIR}
NO_CHECK_REQUIRED_COMPONENTS_MACRO # Eigen does not provide components
)
else()
else()
# The PACKAGE_* variables are defined by the configure_package_config_file
# but without it we define them manually to the hardcoded paths
set(PACKAGE_INIT "")

2
Eigen/Core
View File

@ -123,7 +123,7 @@
#endif
#endif
#ifndef EIGEN_DONT_VECTORIZE
#if !defined(EIGEN_DONT_VECTORIZE) && !defined(EIGEN_CUDACC)
#if defined (EIGEN_SSE2_ON_NON_MSVC_BUT_NOT_OLD_GCC) || defined(EIGEN_SSE2_ON_MSVC_2008_OR_LATER)

4
Eigen/src/Cholesky/LDLT.h
View File

@ -44,7 +44,7 @@ namespace internal {
* decomposition to determine whether a system of equations has a solution.
*
* This class supports the \link InplaceDecomposition inplace decomposition \endlink mechanism.
*
*
* \sa MatrixBase::ldlt(), SelfAdjointView::ldlt(), class LLT
*/
template<typename _MatrixType, int _UpLo> class LDLT
@ -558,7 +558,7 @@ LDLT<MatrixType,_UpLo>& LDLT<MatrixType,_UpLo>::rankUpdate(const MatrixBase<Deri
#ifndef EIGEN_PARSED_BY_DOXYGEN
template<typename _MatrixType, int _UpLo>
template<typename RhsType, typename DstType>
void LDLT<_MatrixType,_UpLo>::_solve_impl(const RhsType &rhs, DstType &dst) const
EIGEN_DEVICE_FUNC void LDLT<_MatrixType,_UpLo>::_solve_impl(const RhsType &rhs, DstType &dst) const
{
eigen_assert(rhs.rows() == rows());
// dst = P b

2
Eigen/src/Cholesky/LLT.h
View File

@ -475,7 +475,7 @@ LLT<_MatrixType,_UpLo> LLT<_MatrixType,_UpLo>::rankUpdate(const VectorType& v, c
#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
EIGEN_DEVICE_FUNC void LLT<_MatrixType,_UpLo>::_solve_impl(const RhsType &rhs, DstType &dst) const
{
dst = rhs;
solveInPlace(dst);

4
Eigen/src/Core/Assign.h
View File

@ -16,7 +16,7 @@ namespace Eigen {
template<typename Derived>
template<typename OtherDerived>
EIGEN_STRONG_INLINE Derived& DenseBase<Derived>
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& DenseBase<Derived>
::lazyAssign(const DenseBase<OtherDerived>& other)
{
enum{
@ -29,7 +29,7 @@ EIGEN_STRONG_INLINE Derived& DenseBase<Derived>
eigen_assert(rows() == other.rows() && cols() == other.cols());
internal::call_assignment_no_alias(derived(),other.derived());
return derived();
}

66
Eigen/src/Core/AssignEvaluator.h
View File

@ -17,7 +17,7 @@ namespace Eigen {
// This implementation is based on Assign.h
namespace internal {
/***************************************************************************
* Part 1 : the logic deciding a strategy for traversal and unrolling *
***************************************************************************/
@ -29,12 +29,12 @@ struct copy_using_evaluator_traits
{
typedef typename DstEvaluator::XprType Dst;
typedef typename Dst::Scalar DstScalar;
enum {
DstFlags = DstEvaluator::Flags,
SrcFlags = SrcEvaluator::Flags
};
public:
enum {
DstAlignment = DstEvaluator::Alignment,
@ -135,7 +135,7 @@ public:
? int(CompleteUnrolling)
: int(NoUnrolling) )
: int(Traversal) == int(LinearTraversal)
? ( bool(MayUnrollCompletely) ? int(CompleteUnrolling)
? ( bool(MayUnrollCompletely) ? int(CompleteUnrolling)
: int(NoUnrolling) )
#if EIGEN_UNALIGNED_VECTORIZE
: int(Traversal) == int(SliceVectorizedTraversal)
@ -195,7 +195,7 @@ struct copy_using_evaluator_DefaultTraversal_CompleteUnrolling
// FIXME: this is not very clean, perhaps this information should be provided by the kernel?
typedef typename Kernel::DstEvaluatorType DstEvaluatorType;
typedef typename DstEvaluatorType::XprType DstXprType;
enum {
outer = Index / DstXprType::InnerSizeAtCompileTime,
inner = Index % DstXprType::InnerSizeAtCompileTime
@ -261,7 +261,7 @@ struct copy_using_evaluator_innervec_CompleteUnrolling
typedef typename Kernel::DstEvaluatorType DstEvaluatorType;
typedef typename DstEvaluatorType::XprType DstXprType;
typedef typename Kernel::PacketType PacketType;
enum {
outer = Index / DstXprType::InnerSizeAtCompileTime,
inner = Index % DstXprType::InnerSizeAtCompileTime,
@ -426,7 +426,7 @@ struct dense_assignment_loop<Kernel, LinearVectorizedTraversal, CompleteUnrollin
{
typedef typename Kernel::DstEvaluatorType::XprType DstXprType;
typedef typename Kernel::PacketType PacketType;
enum { size = DstXprType::SizeAtCompileTime,
packetSize =unpacket_traits<PacketType>::size,
alignedSize = (size/packetSize)*packetSize };
@ -599,14 +599,14 @@ protected:
typedef typename DstEvaluatorTypeT::XprType DstXprType;
typedef typename SrcEvaluatorTypeT::XprType SrcXprType;
public:
typedef DstEvaluatorTypeT DstEvaluatorType;
typedef SrcEvaluatorTypeT SrcEvaluatorType;
typedef typename DstEvaluatorType::Scalar Scalar;
typedef copy_using_evaluator_traits<DstEvaluatorTypeT, SrcEvaluatorTypeT, Functor> AssignmentTraits;
typedef typename AssignmentTraits::PacketType PacketType;
EIGEN_DEVICE_FUNC generic_dense_assignment_kernel(DstEvaluatorType &dst, const SrcEvaluatorType &src, const Functor &func, DstXprType& dstExpr)
: m_dst(dst), m_src(src), m_functor(func), m_dstExpr(dstExpr)
{
@ -614,58 +614,58 @@ public:
AssignmentTraits::debug();
#endif
}
EIGEN_DEVICE_FUNC Index size() const { return m_dstExpr.size(); }
EIGEN_DEVICE_FUNC Index innerSize() const { return m_dstExpr.innerSize(); }
EIGEN_DEVICE_FUNC Index outerSize() const { return m_dstExpr.outerSize(); }
EIGEN_DEVICE_FUNC Index rows() const { return m_dstExpr.rows(); }
EIGEN_DEVICE_FUNC Index cols() const { return m_dstExpr.cols(); }
EIGEN_DEVICE_FUNC Index outerStride() const { return m_dstExpr.outerStride(); }
EIGEN_DEVICE_FUNC DstEvaluatorType& dstEvaluator() { return m_dst; }
EIGEN_DEVICE_FUNC const SrcEvaluatorType& srcEvaluator() const { return m_src; }
/// Assign src(row,col) to dst(row,col) through the assignment functor.
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void assignCoeff(Index row, Index col)
{
m_functor.assignCoeff(m_dst.coeffRef(row,col), m_src.coeff(row,col));
}
/// \sa assignCoeff(Index,Index)
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void assignCoeff(Index index)
{
m_functor.assignCoeff(m_dst.coeffRef(index), m_src.coeff(index));
}
/// \sa assignCoeff(Index,Index)
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void assignCoeffByOuterInner(Index outer, Index inner)
{
Index row = rowIndexByOuterInner(outer, inner);
Index col = colIndexByOuterInner(outer, inner);
Index row = rowIndexByOuterInner(outer, inner);
Index col = colIndexByOuterInner(outer, inner);
assignCoeff(row, col);
}
template<int StoreMode, int LoadMode, typename PacketType>
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void assignPacket(Index row, Index col)
{
m_functor.template assignPacket<StoreMode>(&m_dst.coeffRef(row,col), m_src.template packet<LoadMode,PacketType>(row,col));
}
template<int StoreMode, int LoadMode, typename PacketType>
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void assignPacket(Index index)
{
m_functor.template assignPacket<StoreMode>(&m_dst.coeffRef(index), m_src.template packet<LoadMode,PacketType>(index));
}
template<int StoreMode, int LoadMode, typename PacketType>
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void assignPacketByOuterInner(Index outer, Index inner)
{
Index row = rowIndexByOuterInner(outer, inner);
Index row = rowIndexByOuterInner(outer, inner);
Index col = colIndexByOuterInner(outer, inner);
assignPacket<StoreMode,LoadMode,PacketType>(row, col);
}
EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE Index rowIndexByOuterInner(Index outer, Index inner)
{
typedef typename DstEvaluatorType::ExpressionTraits Traits;
@ -688,7 +688,7 @@ public:
{
return m_dstExpr.data();
}
protected:
DstEvaluatorType& m_dst;
const SrcEvaluatorType& m_src;
@ -734,7 +734,7 @@ EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void call_dense_assignment_loop(DstXprType
resize_if_allowed(dst, src, func);
DstEvaluatorType dstEvaluator(dst);
typedef generic_dense_assignment_kernel<DstEvaluatorType,SrcEvaluatorType,Functor> Kernel;
Kernel kernel(dstEvaluator, srcEvaluator, func, dst.const_cast_derived());
@ -762,7 +762,7 @@ struct EigenBase2EigenBase {};
template<typename,typename> struct AssignmentKind { typedef EigenBase2EigenBase Kind; };
template<> struct AssignmentKind<DenseShape,DenseShape> { typedef Dense2Dense Kind; };
// This is the main assignment class
template< typename DstXprType, typename SrcXprType, typename Functor,
typename Kind = typename AssignmentKind< typename evaluator_traits<DstXprType>::Shape , typename evaluator_traits<SrcXprType>::Shape >::Kind,
@ -787,7 +787,7 @@ void call_assignment(const Dst& dst, const Src& src)
{
call_assignment(dst, src, internal::assign_op<typename Dst::Scalar,typename Src::Scalar>());
}
// Deal with "assume-aliasing"
template<typename Dst, typename Src, typename Func>
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
@ -827,12 +827,12 @@ void call_assignment_no_alias(Dst& dst, const Src& src, const Func& func)
typedef typename internal::conditional<NeedToTranspose, Transpose<Dst>, Dst>::type ActualDstTypeCleaned;
typedef typename internal::conditional<NeedToTranspose, Transpose<Dst>, Dst&>::type ActualDstType;
ActualDstType actualDst(dst);
// TODO check whether this is the right place to perform these checks:
EIGEN_STATIC_ASSERT_LVALUE(Dst)
EIGEN_STATIC_ASSERT_SAME_MATRIX_SIZE(ActualDstTypeCleaned,Src)
EIGEN_CHECK_BINARY_COMPATIBILIY(Func,typename ActualDstTypeCleaned::Scalar,typename Src::Scalar);
Assignment<ActualDstTypeCleaned,Src,Func>::run(actualDst, src, func);
}
template<typename Dst, typename Src>
@ -869,13 +869,12 @@ template<typename Dst, typename Src> void check_for_aliasing(const Dst &dst, con
template< typename DstXprType, typename SrcXprType, typename Functor, typename Weak>
struct Assignment<DstXprType, SrcXprType, Functor, Dense2Dense, Weak>
{
EIGEN_DEVICE_FUNC
static EIGEN_STRONG_INLINE void run(DstXprType &dst, const SrcXprType &src, const Functor &func)
static EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void run(DstXprType &dst, const SrcXprType &src, const Functor &func)
{
#ifndef EIGEN_NO_DEBUG
internal::check_for_aliasing(dst, src);
#endif
call_dense_assignment_loop(dst, src, func);
}
};
@ -887,8 +886,7 @@ struct Assignment<DstXprType, SrcXprType, Functor, Dense2Dense, Weak>
template< typename DstXprType, typename SrcXprType, typename Functor, typename Weak>
struct Assignment<DstXprType, SrcXprType, Functor, EigenBase2EigenBase, Weak>
{
EIGEN_DEVICE_FUNC
static EIGEN_STRONG_INLINE void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op<typename DstXprType::Scalar,typename SrcXprType::Scalar> &/*func*/)
static EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op<typename DstXprType::Scalar,typename SrcXprType::Scalar> &/*func*/)
{
Index dstRows = src.rows();
Index dstCols = src.cols();

10
Eigen/src/Core/CommaInitializer.h
View File

@ -11,7 +11,7 @@
#ifndef EIGEN_COMMAINITIALIZER_H
#define EIGEN_COMMAINITIALIZER_H
namespace Eigen {
namespace Eigen {
/** \class CommaInitializer
* \ingroup Core_Module
@ -44,7 +44,7 @@ struct CommaInitializer
m_xpr.block(0, 0, other.rows(), other.cols()) = other;
}
/* Copy/Move constructor which transfers ownership. This is crucial in
/* Copy/Move constructor which transfers ownership. This is crucial in
* absence of return value optimization to avoid assertions during destruction. */
// FIXME in C++11 mode this could be replaced by a proper RValue constructor
EIGEN_DEVICE_FUNC
@ -135,13 +135,13 @@ struct CommaInitializer
*
* Example: \include MatrixBase_set.cpp
* Output: \verbinclude MatrixBase_set.out
*
*
* \note According the c++ standard, the argument expressions of this comma initializer are evaluated in arbitrary order.
*
* \sa CommaInitializer::finished(), class CommaInitializer
*/
template<typename Derived>
inline CommaInitializer<Derived> DenseBase<Derived>::operator<< (const Scalar& s)
EIGEN_DEVICE_FUNC inline CommaInitializer<Derived> DenseBase<Derived>::operator<< (const Scalar& s)
{
return CommaInitializer<Derived>(*static_cast<Derived*>(this), s);
}
@ -149,7 +149,7 @@ inline CommaInitializer<Derived> DenseBase<Derived>::operator<< (const Scalar& s
/** \sa operator<<(const Scalar&) */
template<typename Derived>
template<typename OtherDerived>
inline CommaInitializer<Derived>
EIGEN_DEVICE_FUNC inline CommaInitializer<Derived>
DenseBase<Derived>::operator<<(const DenseBase<OtherDerived>& other)
{
return CommaInitializer<Derived>(*static_cast<Derived *>(this), other);

8
Eigen/src/Core/CwiseBinaryOp.h
View File

@ -74,7 +74,7 @@ class CwiseBinaryOpImpl;
* \sa MatrixBase::binaryExpr(const MatrixBase<OtherDerived> &,const CustomBinaryOp &) const, class CwiseUnaryOp, class CwiseNullaryOp
*/
template<typename BinaryOp, typename LhsType, typename RhsType>
class CwiseBinaryOp :
class CwiseBinaryOp :
public CwiseBinaryOpImpl<
BinaryOp, LhsType, RhsType,
typename internal::cwise_promote_storage_type<typename internal::traits<LhsType>::StorageKind,
@ -83,7 +83,7 @@ class CwiseBinaryOp :
internal::no_assignment_operator
{
public:
typedef typename internal::remove_all<BinaryOp>::type Functor;
typedef typename internal::remove_all<LhsType>::type Lhs;
typedef typename internal::remove_all<RhsType>::type Rhs;
@ -158,7 +158,7 @@ public:
*/
template<typename Derived>
template<typename OtherDerived>
EIGEN_STRONG_INLINE Derived &
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived &
MatrixBase<Derived>::operator-=(const MatrixBase<OtherDerived> &other)
{
call_assignment(derived(), other.derived(), internal::sub_assign_op<Scalar,typename OtherDerived::Scalar>());
@ -171,7 +171,7 @@ MatrixBase<Derived>::operator-=(const MatrixBase<OtherDerived> &other)
*/
template<typename Derived>
template<typename OtherDerived>
EIGEN_STRONG_INLINE Derived &
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived &
MatrixBase<Derived>::operator+=(const MatrixBase<OtherDerived>& other)
{
call_assignment(derived(), other.derived(), internal::add_assign_op<Scalar,typename OtherDerived::Scalar>());

6
Eigen/src/Core/CwiseNullaryOp.h
View File

@ -126,12 +126,12 @@ DenseBase<Derived>::NullaryExpr(Index rows, Index cols, const CustomNullaryOp& f
*
* Here is an example with C++11 random generators: \include random_cpp11.cpp
* Output: \verbinclude random_cpp11.out
*
*
* \sa class CwiseNullaryOp
*/
template<typename Derived>
template<typename CustomNullaryOp>
EIGEN_STRONG_INLINE const CwiseNullaryOp<CustomNullaryOp, typename DenseBase<Derived>::PlainObject>
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const CwiseNullaryOp<CustomNullaryOp, typename DenseBase<Derived>::PlainObject>
DenseBase<Derived>::NullaryExpr(Index size, const CustomNullaryOp& func)
{
EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived)
@ -170,7 +170,7 @@ DenseBase<Derived>::NullaryExpr(const CustomNullaryOp& func)
* \sa class CwiseNullaryOp
*/
template<typename Derived>
EIGEN_STRONG_INLINE const typename DenseBase<Derived>::ConstantReturnType
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename DenseBase<Derived>::ConstantReturnType
DenseBase<Derived>::Constant(Index rows, Index cols, const Scalar& value)
{
return DenseBase<Derived>::NullaryExpr(rows, cols, internal::scalar_constant_op<Scalar>(value));

18
Eigen/src/Core/Diagonal.h
View File

@ -11,7 +11,7 @@
#ifndef EIGEN_DIAGONAL_H
#define EIGEN_DIAGONAL_H
namespace Eigen {
namespace Eigen {
/** \class Diagonal
* \ingroup Core_Module
@ -149,8 +149,8 @@ template<typename MatrixType, int _DiagIndex> class Diagonal
}
EIGEN_DEVICE_FUNC
inline const typename internal::remove_all<typename MatrixType::Nested>::type&
nestedExpression() const
inline const typename internal::remove_all<typename MatrixType::Nested>::type&
nestedExpression() const
{
return m_matrix;
}
@ -187,7 +187,7 @@ template<typename MatrixType, int _DiagIndex> class Diagonal
*
* \sa class Diagonal */
template<typename Derived>
inline typename MatrixBase<Derived>::DiagonalReturnType
EIGEN_DEVICE_FUNC inline typename MatrixBase<Derived>::DiagonalReturnType
MatrixBase<Derived>::diagonal()
{
return DiagonalReturnType(derived());
@ -195,7 +195,7 @@ MatrixBase<Derived>::diagonal()
/** This is the const version of diagonal(). */
template<typename Derived>
inline typename MatrixBase<Derived>::ConstDiagonalReturnType
EIGEN_DEVICE_FUNC inline typename MatrixBase<Derived>::ConstDiagonalReturnType
MatrixBase<Derived>::diagonal() const
{
return ConstDiagonalReturnType(derived());
@ -213,7 +213,7 @@ MatrixBase<Derived>::diagonal() const
*
* \sa MatrixBase::diagonal(), class Diagonal */
template<typename Derived>
inline typename MatrixBase<Derived>::DiagonalDynamicIndexReturnType
EIGEN_DEVICE_FUNC inline typename MatrixBase<Derived>::DiagonalDynamicIndexReturnType
MatrixBase<Derived>::diagonal(Index index)
{
return DiagonalDynamicIndexReturnType(derived(), index);
@ -221,7 +221,7 @@ MatrixBase<Derived>::diagonal(Index index)
/** This is the const version of diagonal(Index). */
template<typename Derived>
inline typename MatrixBase<Derived>::ConstDiagonalDynamicIndexReturnType
EIGEN_DEVICE_FUNC inline typename MatrixBase<Derived>::ConstDiagonalDynamicIndexReturnType
MatrixBase<Derived>::diagonal(Index index) const
{
return ConstDiagonalDynamicIndexReturnType(derived(), index);
@ -240,7 +240,7 @@ MatrixBase<Derived>::diagonal(Index index) const
* \sa MatrixBase::diagonal(), class Diagonal */
template<typename Derived>
template<int Index_>
inline typename MatrixBase<Derived>::template DiagonalIndexReturnType<Index_>::Type
EIGEN_DEVICE_FUNC inline typename MatrixBase<Derived>::template DiagonalIndexReturnType<Index_>::Type
MatrixBase<Derived>::diagonal()
{
return typename DiagonalIndexReturnType<Index_>::Type(derived());
@ -249,7 +249,7 @@ MatrixBase<Derived>::diagonal()
/** This is the const version of diagonal<int>(). */
template<typename Derived>
template<int Index_>
inline typename MatrixBase<Derived>::template ConstDiagonalIndexReturnType<Index_>::Type
EIGEN_DEVICE_FUNC inline typename MatrixBase<Derived>::template ConstDiagonalIndexReturnType<Index_>::Type
MatrixBase<Derived>::diagonal() const
{
return typename ConstDiagonalIndexReturnType<Index_>::Type(derived());

16
Eigen/src/Core/DiagonalMatrix.h
View File

@ -11,7 +11,7 @@
#ifndef EIGEN_DIAGONALMATRIX_H
#define EIGEN_DIAGONALMATRIX_H
namespace Eigen {
namespace Eigen {
#ifndef EIGEN_PARSED_BY_DOXYGEN
template<typename Derived>
@ -44,7 +44,7 @@ class DiagonalBase : public EigenBase<Derived>
EIGEN_DEVICE_FUNC
DenseMatrixType toDenseMatrix() const { return derived(); }
EIGEN_DEVICE_FUNC
inline const DiagonalVectorType& diagonal() const { return derived().diagonal(); }
EIGEN_DEVICE_FUNC
@ -70,7 +70,7 @@ class DiagonalBase : public EigenBase<Derived>
{
return InverseReturnType(diagonal().cwiseInverse());
}
EIGEN_DEVICE_FUNC
inline const DiagonalWrapper<const EIGEN_EXPR_BINARYOP_SCALAR_RETURN_TYPE(DiagonalVectorType,Scalar,product) >
operator*(const Scalar& scalar) const
@ -273,7 +273,7 @@ class DiagonalWrapper
* \sa class DiagonalWrapper, class DiagonalMatrix, diagonal(), isDiagonal()
**/
template<typename Derived>
inline const DiagonalWrapper<const Derived>
EIGEN_DEVICE_FUNC inline const DiagonalWrapper<const Derived>
MatrixBase<Derived>::asDiagonal() const
{
return DiagonalWrapper<const Derived>(derived());
@ -318,20 +318,20 @@ template<> struct AssignmentKind<DenseShape,DiagonalShape> { typedef Diagonal2De
template< typename DstXprType, typename SrcXprType, typename Functor>
struct Assignment<DstXprType, SrcXprType, Functor, Diagonal2Dense>
{
static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op<typename DstXprType::Scalar,typename SrcXprType::Scalar> &/*func*/)
static EIGEN_DEVICE_FUNC void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op<typename DstXprType::Scalar,typename SrcXprType::Scalar> &/*func*/)
{
Index dstRows = src.rows();
Index dstCols = src.cols();
if((dst.rows()!=dstRows) || (dst.cols()!=dstCols))
dst.resize(dstRows, dstCols);
dst.setZero();
dst.diagonal() = src.diagonal();
}
static void run(DstXprType &dst, const SrcXprType &src, const internal::add_assign_op<typename DstXprType::Scalar,typename SrcXprType::Scalar> &/*func*/)
{ dst.diagonal() += src.diagonal(); }
static void run(DstXprType &dst, const SrcXprType &src, const internal::sub_assign_op<typename DstXprType::Scalar,typename SrcXprType::Scalar> &/*func*/)
{ dst.diagonal() -= src.diagonal(); }
};

20
Eigen/src/Core/Dot.h
View File

@ -10,7 +10,7 @@
#ifndef EIGEN_DOT_H
#define EIGEN_DOT_H
namespace Eigen {
namespace Eigen {
namespace internal {
@ -78,7 +78,7 @@ MatrixBase<Derived>::dot(const MatrixBase<OtherDerived>& other) const
typedef internal::scalar_conj_product_op<Scalar,typename OtherDerived::Scalar> func;
EIGEN_CHECK_BINARY_COMPATIBILIY(func,Scalar,typename OtherDerived::Scalar);
#endif
eigen_assert(size() == other.size());
return internal::dot_nocheck<Derived,OtherDerived>::run(*this, other);
@ -93,7 +93,7 @@ MatrixBase<Derived>::dot(const MatrixBase<OtherDerived>& other) const
* \sa dot(), norm(), lpNorm()
*/
template<typename Derived>
EIGEN_STRONG_INLINE typename NumTraits<typename internal::traits<Derived>::Scalar>::Real MatrixBase<Derived>::squaredNorm() const
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE typename NumTraits<typename internal::traits<Derived>::Scalar>::Real MatrixBase<Derived>::squaredNorm() const
{
return numext::real((*this).cwiseAbs2().sum());
}
@ -105,7 +105,7 @@ EIGEN_STRONG_INLINE typename NumTraits<typename internal::traits<Derived>::Scala
* \sa lpNorm(), dot(), squaredNorm()
*/
template<typename Derived>
EIGEN_STRONG_INLINE typename NumTraits<typename internal::traits<Derived>::Scalar>::Real MatrixBase<Derived>::norm() const
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE typename NumTraits<typename internal::traits<Derived>::Scalar>::Real MatrixBase<Derived>::norm() const
{
return numext::sqrt(squaredNorm());
}
@ -120,7 +120,7 @@ EIGEN_STRONG_INLINE typename NumTraits<typename internal::traits<Derived>::Scala
* \sa norm(), normalize()
*/
template<typename Derived>
EIGEN_STRONG_INLINE const typename MatrixBase<Derived>::PlainObject
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename MatrixBase<Derived>::PlainObject
MatrixBase<Derived>::normalized() const
{
typedef typename internal::nested_eval<Derived,2>::type _Nested;
@ -142,7 +142,7 @@ MatrixBase<Derived>::normalized() const
* \sa norm(), normalized()
*/
template<typename Derived>
EIGEN_STRONG_INLINE void MatrixBase<Derived>::normalize()
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void MatrixBase<Derived>::normalize()
{
RealScalar z = squaredNorm();
// NOTE: after extensive benchmarking, this conditional does not impact performance, at least on recent x86 CPU
@ -163,7 +163,7 @@ EIGEN_STRONG_INLINE void MatrixBase<Derived>::normalize()
* \sa stableNorm(), stableNormalize(), normalized()
*/
template<typename Derived>
EIGEN_STRONG_INLINE const typename MatrixBase<Derived>::PlainObject
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename MatrixBase<Derived>::PlainObject
MatrixBase<Derived>::stableNormalized() const
{
typedef typename internal::nested_eval<Derived,3>::type _Nested;
@ -188,7 +188,7 @@ MatrixBase<Derived>::stableNormalized() const
* \sa stableNorm(), stableNormalized(), normalize()
*/
template<typename Derived>
EIGEN_STRONG_INLINE void MatrixBase<Derived>::stableNormalize()
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void MatrixBase<Derived>::stableNormalize()
{
RealScalar w = cwiseAbs().maxCoeff();
RealScalar z = (derived()/w).squaredNorm();
@ -260,9 +260,9 @@ struct lpNorm_selector<Derived, Infinity>
template<typename Derived>
template<int p>
#ifndef EIGEN_PARSED_BY_DOXYGEN
inline typename NumTraits<typename internal::traits<Derived>::Scalar>::Real
EIGEN_DEVICE_FUNC inline typename NumTraits<typename internal::traits<Derived>::Scalar>::Real
#else
MatrixBase<Derived>::RealScalar
EIGEN_DEVICE_FUNC MatrixBase<Derived>::RealScalar
#endif
MatrixBase<Derived>::lpNorm() const
{

8
Eigen/src/Core/Fuzzy.h
View File

@ -11,7 +11,7 @@
#ifndef EIGEN_FUZZY_H
#define EIGEN_FUZZY_H
namespace Eigen {
namespace Eigen {
namespace internal
{
@ -100,7 +100,7 @@ struct isMuchSmallerThan_scalar_selector<Derived, true>
*/
template<typename Derived>
template<typename OtherDerived>
bool DenseBase<Derived>::isApprox(
EIGEN_DEVICE_FUNC bool DenseBase<Derived>::isApprox(
const DenseBase<OtherDerived>& other,
const RealScalar& prec
) const
@ -122,7 +122,7 @@ bool DenseBase<Derived>::isApprox(
* \sa isApprox(), isMuchSmallerThan(const DenseBase<OtherDerived>&, RealScalar) const
*/
template<typename Derived>
bool DenseBase<Derived>::isMuchSmallerThan(
EIGEN_DEVICE_FUNC bool DenseBase<Derived>::isMuchSmallerThan(
const typename NumTraits<Scalar>::Real& other,
const RealScalar& prec
) const
@ -142,7 +142,7 @@ bool DenseBase<Derived>::isMuchSmallerThan(
*/
template<typename Derived>
template<typename OtherDerived>
bool DenseBase<Derived>::isMuchSmallerThan(
EIGEN_DEVICE_FUNC bool DenseBase<Derived>::isMuchSmallerThan(
const DenseBase<OtherDerived>& other,
const RealScalar& prec
) const

10
Eigen/src/Core/GeneralProduct.h
View File

@ -207,12 +207,12 @@ template<> struct gemv_dense_selector<OnTheRight,ColMajor,true>
typedef typename Rhs::Scalar RhsScalar;
typedef typename Dest::Scalar ResScalar;
typedef typename Dest::RealScalar RealScalar;
typedef internal::blas_traits<Lhs> LhsBlasTraits;
typedef typename LhsBlasTraits::DirectLinearAccessType ActualLhsType;
typedef internal::blas_traits<Rhs> RhsBlasTraits;
typedef typename RhsBlasTraits::DirectLinearAccessType ActualRhsType;
typedef Map<Matrix<ResScalar,Dynamic,1>, EIGEN_PLAIN_ENUM_MIN(AlignedMax,internal::packet_traits<ResScalar>::size)> MappedDest;
ActualLhsType actualLhs = LhsBlasTraits::extract(lhs);
@ -300,7 +300,7 @@ template<> struct gemv_dense_selector<OnTheRight,RowMajor,true>
typedef typename Lhs::Scalar LhsScalar;
typedef typename Rhs::Scalar RhsScalar;
typedef typename Dest::Scalar ResScalar;
typedef internal::blas_traits<Lhs> LhsBlasTraits;
typedef typename LhsBlasTraits::DirectLinearAccessType ActualLhsType;
typedef internal::blas_traits<Rhs> RhsBlasTraits;
@ -386,7 +386,7 @@ template<> struct gemv_dense_selector<OnTheRight,RowMajor,false>
*/
template<typename Derived>
template<typename OtherDerived>
inline const Product<Derived, OtherDerived>
EIGEN_DEVICE_FUNC inline const Product<Derived, OtherDerived>
MatrixBase<Derived>::operator*(const MatrixBase<OtherDerived> &other) const
{
// A note regarding the function declaration: In MSVC, this function will sometimes
@ -428,7 +428,7 @@ MatrixBase<Derived>::operator*(const MatrixBase<OtherDerived> &other) const
*/
template<typename Derived>
template<typename OtherDerived>
const Product<Derived,OtherDerived,LazyProduct>
EIGEN_DEVICE_FUNC const Product<Derived,OtherDerived,LazyProduct>
MatrixBase<Derived>::lazyProduct(const MatrixBase<OtherDerived> &other) const
{
enum {

42
Eigen/src/Core/GenericPacketMath.h
View File

@ -237,7 +237,7 @@ ploaddup(const typename unpacket_traits<Packet>::type* from) { return *from; }
* For instance, for a packet of 8 elements, 2 scalars will be read from \a *from and
* replicated to form: {from[0],from[0],from[0],from[0],from[1],from[1],from[1],from[1]}
* Currently, this function is only used in matrix products.
* For packet-size smaller or equal to 4, this function is equivalent to pload1
* For packet-size smaller or equal to 4, this function is equivalent to pload1
*/
template<typename Packet> EIGEN_DEVICE_FUNC inline Packet
ploadquad(const typename unpacket_traits<Packet>::type* from)
@ -359,77 +359,77 @@ template<typename Packet> EIGEN_DEVICE_FUNC inline Packet pcplxflip(const Packet
***************************/
/** \internal \returns the sine of \a a (coeff-wise) */
template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
template<typename Packet> EIGEN_DEVICE_FUNC EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
Packet psin(const Packet& a) { using std::sin; return sin(a); }
/** \internal \returns the cosine of \a a (coeff-wise) */
template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
template<typename Packet> EIGEN_DEVICE_FUNC EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
Packet pcos(const Packet& a) { using std::cos; return cos(a); }
/** \internal \returns the tan of \a a (coeff-wise) */
template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
template<typename Packet> EIGEN_DEVICE_FUNC EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
Packet ptan(const Packet& a) { using std::tan; return tan(a); }
/** \internal \returns the arc sine of \a a (coeff-wise) */
template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
template<typename Packet> EIGEN_DEVICE_FUNC EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
Packet pasin(const Packet& a) { using std::asin; return asin(a); }
/** \internal \returns the arc cosine of \a a (coeff-wise) */
template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
template<typename Packet> EIGEN_DEVICE_FUNC EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
Packet pacos(const Packet& a) { using std::acos; return acos(a); }
/** \internal \returns the arc tangent of \a a (coeff-wise) */
template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
template<typename Packet> EIGEN_DEVICE_FUNC EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
Packet patan(const Packet& a) { using std::atan; return atan(a); }
/** \internal \returns the hyperbolic sine of \a a (coeff-wise) */
template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
template<typename Packet> EIGEN_DEVICE_FUNC EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
Packet psinh(const Packet& a) { using std::sinh; return sinh(a); }
/** \internal \returns the hyperbolic cosine of \a a (coeff-wise) */
template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
template<typename Packet> EIGEN_DEVICE_FUNC EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
Packet pcosh(const Packet& a) { using std::cosh; return cosh(a); }
/** \internal \returns the hyperbolic tan of \a a (coeff-wise) */
template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
template<typename Packet> EIGEN_DEVICE_FUNC EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
Packet ptanh(const Packet& a) { using std::tanh; return tanh(a); }
/** \internal \returns the exp of \a a (coeff-wise) */
template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
template<typename Packet> EIGEN_DEVICE_FUNC EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
Packet pexp(const Packet& a) { using std::exp; return exp(a); }
/** \internal \returns the log of \a a (coeff-wise) */
template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
template<typename Packet> EIGEN_DEVICE_FUNC EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
Packet plog(const Packet& a) { using std::log; return log(a); }
/** \internal \returns the log1p of \a a (coeff-wise) */
template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
template<typename Packet> EIGEN_DEVICE_FUNC EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
Packet plog1p(const Packet& a) { return numext::log1p(a); }
/** \internal \returns the log10 of \a a (coeff-wise) */
template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
template<typename Packet> EIGEN_DEVICE_FUNC EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
Packet plog10(const Packet& a) { using std::log10; return log10(a); }
/** \internal \returns the square-root of \a a (coeff-wise) */
template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
template<typename Packet> EIGEN_DEVICE_FUNC EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
Packet psqrt(const Packet& a) { using std::sqrt; return sqrt(a); }
/** \internal \returns the reciprocal square-root of \a a (coeff-wise) */
template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
template<typename Packet> EIGEN_DEVICE_FUNC EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
Packet prsqrt(const Packet& a) {
return pdiv(pset1<Packet>(1), psqrt(a));
}
/** \internal \returns the rounded value of \a a (coeff-wise) */
template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
template<typename Packet> EIGEN_DEVICE_FUNC EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
Packet pround(const Packet& a) { using numext::round; return round(a); }
/** \internal \returns the floor of \a a (coeff-wise) */
template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
template<typename Packet> EIGEN_DEVICE_FUNC EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
Packet pfloor(const Packet& a) { using numext::floor; return floor(a); }
/** \internal \returns the ceil of \a a (coeff-wise) */
template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
template<typename Packet> EIGEN_DEVICE_FUNC EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
Packet pceil(const Packet& a) { using numext::ceil; return ceil(a); }
/***************************************************************************
@ -494,14 +494,14 @@ struct palign_impl
/** \internal update \a first using the concatenation of the packet_size minus \a Offset last elements
* of \a first and \a Offset first elements of \a second.
*
*
* This function is currently only used to optimize matrix-vector products on unligned matrices.
* It takes 2 packets that represent a contiguous memory array, and returns a packet starting
* at the position \a Offset. For instance, for packets of 4 elements, we have:
* Input:
* - first = {f0,f1,f2,f3}
* - second = {s0,s1,s2,s3}
* Output:
* Output:
* - if Offset==0 then {f0,f1,f2,f3}
* - if Offset==1 then {f1,f2,f3,s0}
* - if Offset==2 then {f2,f3,s0,s1}

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

@ -99,7 +99,7 @@ template<typename ExpressionType> class NestByValue
/** \returns an expression of the temporary version of *this.
*/
template<typename Derived>
inline const NestByValue<Derived>
EIGEN_DEVICE_FUNC inline const NestByValue<Derived>
DenseBase<Derived>::nestByValue() const
{
return NestByValue<Derived>(derived());

140
Eigen/src/Core/ProductEvaluators.h
View File

@ -14,7 +14,7 @@
#define EIGEN_PRODUCTEVALUATORS_H
namespace Eigen {
namespace internal {
/** \internal
@ -22,19 +22,19 @@ namespace internal {
* Since products require special treatments to handle all possible cases,
* we simply deffer the evaluation logic to a product_evaluator class
* which offers more partial specialization possibilities.
*
*
* \sa class product_evaluator
*/
template<typename Lhs, typename Rhs, int Options>
struct evaluator<Product<Lhs, Rhs, Options> >
struct evaluator<Product<Lhs, Rhs, Options> >
: public product_evaluator<Product<Lhs, Rhs, Options> >
{
typedef Product<Lhs, Rhs, Options> XprType;
typedef product_evaluator<XprType> Base;
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE explicit evaluator(const XprType& xpr) : Base(xpr) {}
};
// Catch "scalar * ( A * B )" and transform it to "(A*scalar) * B"
// TODO we should apply that rule only if that's really helpful
template<typename Lhs, typename Rhs, typename Scalar1, typename Scalar2, typename Plain1>
@ -62,12 +62,12 @@ struct evaluator<CwiseBinaryOp<internal::scalar_product_op<Scalar1,Scalar2>,
template<typename Lhs, typename Rhs, int DiagIndex>
struct evaluator<Diagonal<const Product<Lhs, Rhs, DefaultProduct>, DiagIndex> >
struct evaluator<Diagonal<const Product<Lhs, Rhs, DefaultProduct>, DiagIndex> >
: public evaluator<Diagonal<const Product<Lhs, Rhs, LazyProduct>, DiagIndex> >
{
typedef Diagonal<const Product<Lhs, Rhs, DefaultProduct>, DiagIndex> XprType;
typedef evaluator<Diagonal<const Product<Lhs, Rhs, LazyProduct>, DiagIndex> > Base;
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE explicit evaluator(const XprType& xpr)
: Base(Diagonal<const Product<Lhs, Rhs, LazyProduct>, DiagIndex>(
Product<Lhs, Rhs, LazyProduct>(xpr.nestedExpression().lhs(), xpr.nestedExpression().rhs()),
@ -108,23 +108,23 @@ struct product_evaluator<Product<Lhs, Rhs, Options>, ProductTag, LhsShape, RhsSh
: m_result(xpr.rows(), xpr.cols())
{
::new (static_cast<Base*>(this)) Base(m_result);
// FIXME shall we handle nested_eval here?,
// if so, then we must take care at removing the call to nested_eval in the specializations (e.g., in permutation_matrix_product, transposition_matrix_product, etc.)
// typedef typename internal::nested_eval<Lhs,Rhs::ColsAtCompileTime>::type LhsNested;
// typedef typename internal::nested_eval<Rhs,Lhs::RowsAtCompileTime>::type RhsNested;
// typedef typename internal::remove_all<LhsNested>::type LhsNestedCleaned;
// typedef typename internal::remove_all<RhsNested>::type RhsNestedCleaned;
//
//
// const LhsNested lhs(xpr.lhs());
// const RhsNested rhs(xpr.rhs());
//
//
// generic_product_impl<LhsNestedCleaned, RhsNestedCleaned>::evalTo(m_result, lhs, rhs);
generic_product_impl<Lhs, Rhs, LhsShape, RhsShape, ProductTag>::evalTo(m_result, xpr.lhs(), xpr.rhs());
}
protected:
protected:
PlainObject m_result;
};
@ -137,7 +137,7 @@ struct Assignment<DstXprType, Product<Lhs,Rhs,Options>, internal::assign_op<Scal
typename enable_if<(Options==DefaultProduct || Options==AliasFreeProduct)>::type>
{
typedef Product<Lhs,Rhs,Options> SrcXprType;
static EIGEN_STRONG_INLINE
static EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC
void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op<Scalar,Scalar> &)
{
Index dstRows = src.rows();
@ -155,7 +155,7 @@ struct Assignment<DstXprType, Product<Lhs,Rhs,Options>, internal::add_assign_op<
typename enable_if<(Options==DefaultProduct || Options==AliasFreeProduct)>::type>
{
typedef Product<Lhs,Rhs,Options> SrcXprType;
static EIGEN_STRONG_INLINE
static EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC
void run(DstXprType &dst, const SrcXprType &src, const internal::add_assign_op<Scalar,Scalar> &)
{
eigen_assert(dst.rows() == src.rows() && dst.cols() == src.cols());
@ -170,7 +170,7 @@ struct Assignment<DstXprType, Product<Lhs,Rhs,Options>, internal::sub_assign_op<
typename enable_if<(Options==DefaultProduct || Options==AliasFreeProduct)>::type>
{
typedef Product<Lhs,Rhs,Options> SrcXprType;
static EIGEN_STRONG_INLINE
static EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC
void run(DstXprType &dst, const SrcXprType &src, const internal::sub_assign_op<Scalar,Scalar> &)
{
eigen_assert(dst.rows() == src.rows() && dst.cols() == src.cols());
@ -190,7 +190,7 @@ struct Assignment<DstXprType, CwiseBinaryOp<internal::scalar_product_op<ScalarBi
typedef CwiseBinaryOp<internal::scalar_product_op<ScalarBis,Scalar>,
const CwiseNullaryOp<internal::scalar_constant_op<ScalarBis>,Plain>,
const Product<Lhs,Rhs,DefaultProduct> > SrcXprType;
static EIGEN_STRONG_INLINE
static EIGEN_STRONG_INLINE EIGEN_DEVICE_FUNC
void run(DstXprType &dst, const SrcXprType &src, const AssignFunc& func)
{
call_assignment_no_alias(dst, (src.lhs().functor().m_other * src.rhs().lhs())*src.rhs().rhs(), func);
@ -250,13 +250,13 @@ struct generic_product_impl<Lhs,Rhs,DenseShape,DenseShape,InnerProduct>
{
dst.coeffRef(0,0) = (lhs.transpose().cwiseProduct(rhs)).sum();
}
template<typename Dst>
static EIGEN_STRONG_INLINE void addTo(Dst& dst, const Lhs& lhs, const Rhs& rhs)
{
dst.coeffRef(0,0) += (lhs.transpose().cwiseProduct(rhs)).sum();
}
template<typename Dst>
static EIGEN_STRONG_INLINE void subTo(Dst& dst, const Lhs& lhs, const Rhs& rhs)
{ dst.coeffRef(0,0) -= (lhs.transpose().cwiseProduct(rhs)).sum(); }
@ -298,7 +298,7 @@ struct generic_product_impl<Lhs,Rhs,DenseShape,DenseShape,OuterProduct>
{
template<typename T> struct is_row_major : internal::conditional<(int(T::Flags)&RowMajorBit), internal::true_type, internal::false_type>::type {};
typedef typename Product<Lhs,Rhs>::Scalar Scalar;
// TODO it would be nice to be able to exploit our *_assign_op functors for that purpose
struct set { template<typename Dst, typename Src> void operator()(const Dst& dst, const Src& src) const { dst.const_cast_derived() = src; } };
struct add { template<typename Dst, typename Src> void operator()(const Dst& dst, const Src& src) const { dst.const_cast_derived() += src; } };
@ -310,31 +310,31 @@ struct generic_product_impl<Lhs,Rhs,DenseShape,DenseShape,OuterProduct>
dst.const_cast_derived() += m_scale * src;
}
};
template<typename Dst>
static EIGEN_STRONG_INLINE void evalTo(Dst& dst, const Lhs& lhs, const Rhs& rhs)
{
internal::outer_product_selector_run(dst, lhs, rhs, set(), is_row_major<Dst>());
}
template<typename Dst>
static EIGEN_STRONG_INLINE void addTo(Dst& dst, const Lhs& lhs, const Rhs& rhs)
{
internal::outer_product_selector_run(dst, lhs, rhs, add(), is_row_major<Dst>());
}
template<typename Dst>
static EIGEN_STRONG_INLINE void subTo(Dst& dst, const Lhs& lhs, const Rhs& rhs)
{
internal::outer_product_selector_run(dst, lhs, rhs, sub(), is_row_major<Dst>());
}
template<typename Dst>
static EIGEN_STRONG_INLINE void scaleAndAddTo(Dst& dst, const Lhs& lhs, const Rhs& rhs, const Scalar& alpha)
{
internal::outer_product_selector_run(dst, lhs, rhs, adds(alpha), is_row_major<Dst>());
}
};
@ -343,7 +343,7 @@ template<typename Lhs, typename Rhs, typename Derived>
struct generic_product_impl_base
{
typedef typename Product<Lhs,Rhs>::Scalar Scalar;
template<typename Dst>
static EIGEN_STRONG_INLINE void evalTo(Dst& dst, const Lhs& lhs, const Rhs& rhs)
{ dst.setZero(); scaleAndAddTo(dst, lhs, rhs, Scalar(1)); }
@ -355,7 +355,7 @@ struct generic_product_impl_base
template<typename Dst>
static EIGEN_STRONG_INLINE void subTo(Dst& dst, const Lhs& lhs, const Rhs& rhs)
{ scaleAndAddTo(dst, lhs, rhs, Scalar(-1)); }
template<typename Dst>
static EIGEN_STRONG_INLINE void scaleAndAddTo(Dst& dst, const Lhs& lhs, const Rhs& rhs, const Scalar& alpha)
{ Derived::scaleAndAddTo(dst,lhs,rhs,alpha); }
@ -385,12 +385,12 @@ struct generic_product_impl<Lhs,Rhs,DenseShape,DenseShape,GemvProduct>
};
template<typename Lhs, typename Rhs>
struct generic_product_impl<Lhs,Rhs,DenseShape,DenseShape,CoeffBasedProductMode>
struct generic_product_impl<Lhs,Rhs,DenseShape,DenseShape,CoeffBasedProductMode>
{
typedef typename Product<Lhs,Rhs>::Scalar Scalar;
template<typename Dst>
static EIGEN_STRONG_INLINE void evalTo(Dst& dst, const Lhs& lhs, const Rhs& rhs)
static EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void evalTo(Dst& dst, const Lhs& lhs, const Rhs& rhs)
{
// Same as: dst.noalias() = lhs.lazyProduct(rhs);
// but easier on the compiler side
@ -403,7 +403,7 @@ struct generic_product_impl<Lhs,Rhs,DenseShape,DenseShape,CoeffBasedProductMode>
// dst.noalias() += lhs.lazyProduct(rhs);
call_assignment_no_alias(dst, lhs.lazyProduct(rhs), internal::add_assign_op<typename Dst::Scalar,Scalar>());
}
template<typename Dst>
static EIGEN_STRONG_INLINE void subTo(Dst& dst, const Lhs& lhs, const Rhs& rhs)
{
@ -435,8 +435,8 @@ struct generic_product_impl<Lhs,Rhs,DenseShape,DenseShape,CoeffBasedProductMode>
{
call_assignment_no_alias(dst, lhs.lazyProduct(rhs), func);
}
// template<typename Dst>
// static inline void scaleAndAddTo(Dst& dst, const Lhs& lhs, const Rhs& rhs, const Scalar& alpha)
// { dst.noalias() += alpha * lhs.lazyProduct(rhs); }
@ -497,7 +497,7 @@ struct product_evaluator<Product<Lhs, Rhs, LazyProduct>, ProductTag, DenseShape,
typedef typename internal::nested_eval<Lhs,Rhs::ColsAtCompileTime>::type LhsNested;
typedef typename internal::nested_eval<Rhs,Lhs::RowsAtCompileTime>::type RhsNested;
typedef typename internal::remove_all<LhsNested>::type LhsNestedCleaned;
typedef typename internal::remove_all<RhsNested>::type RhsNestedCleaned;
@ -516,7 +516,7 @@ struct product_evaluator<Product<Lhs, Rhs, LazyProduct>, ProductTag, DenseShape,
typedef typename find_best_packet<Scalar,ColsAtCompileTime>::type RhsVecPacketType;
enum {
LhsCoeffReadCost = LhsEtorType::CoeffReadCost,
RhsCoeffReadCost = RhsEtorType::CoeffReadCost,
CoeffReadCost = InnerSize==0 ? NumTraits<Scalar>::ReadCost
@ -525,10 +525,10 @@ struct product_evaluator<Product<Lhs, Rhs, LazyProduct>, ProductTag, DenseShape,
+ (InnerSize - 1) * NumTraits<Scalar>::AddCost,
Unroll = CoeffReadCost <= EIGEN_UNROLLING_LIMIT,
LhsFlags = LhsEtorType::Flags,
RhsFlags = RhsEtorType::Flags,
LhsRowMajor = LhsFlags & RowMajorBit,
RhsRowMajor = RhsFlags & RowMajorBit,
@ -538,7 +538,7 @@ struct product_evaluator<Product<Lhs, Rhs, LazyProduct>, ProductTag, DenseShape,
// Here, we don't care about alignment larger than the usable packet size.
LhsAlignment = EIGEN_PLAIN_ENUM_MIN(LhsEtorType::Alignment,LhsVecPacketSize*int(sizeof(typename LhsNestedCleaned::Scalar))),
RhsAlignment = EIGEN_PLAIN_ENUM_MIN(RhsEtorType::Alignment,RhsVecPacketSize*int(sizeof(typename RhsNestedCleaned::Scalar))),
SameType = is_same<typename LhsNestedCleaned::Scalar,typename RhsNestedCleaned::Scalar>::value,
CanVectorizeRhs = bool(RhsRowMajor) && (RhsFlags & PacketAccessBit) && (ColsAtCompileTime!=1),
@ -553,7 +553,7 @@ struct product_evaluator<Product<Lhs, Rhs, LazyProduct>, ProductTag, DenseShape,
// TODO enable vectorization for mixed types
| (SameType && (CanVectorizeLhs || CanVectorizeRhs) ? PacketAccessBit : 0)
| (XprType::IsVectorAtCompileTime ? LinearAccessBit : 0),
LhsOuterStrideBytes = int(LhsNestedCleaned::OuterStrideAtCompileTime) * int(sizeof(typename LhsNestedCleaned::Scalar)),
RhsOuterStrideBytes = int(RhsNestedCleaned::OuterStrideAtCompileTime) * int(sizeof(typename RhsNestedCleaned::Scalar)),
@ -572,7 +572,7 @@ struct product_evaluator<Product<Lhs, Rhs, LazyProduct>, ProductTag, DenseShape,
&& (LhsFlags & RhsFlags & ActualPacketAccessBit)
&& (InnerSize % packet_traits<Scalar>::size == 0)
};
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const CoeffReturnType coeff(Index row, Index col) const
{
return (m_lhs.row(row).transpose().cwiseProduct( m_rhs.col(col) )).sum();
@ -611,7 +611,7 @@ struct product_evaluator<Product<Lhs, Rhs, LazyProduct>, ProductTag, DenseShape,
protected:
typename internal::add_const_on_value_type<LhsNested>::type m_lhs;
typename internal::add_const_on_value_type<RhsNested>::type m_rhs;
LhsEtorType m_lhsImpl;
RhsEtorType m_rhsImpl;
@ -730,7 +730,7 @@ struct generic_product_impl<Lhs,Rhs,TriangularShape,DenseShape,ProductTag>
: generic_product_impl_base<Lhs,Rhs,generic_product_impl<Lhs,Rhs,TriangularShape,DenseShape,ProductTag> >
{
typedef typename Product<Lhs,Rhs>::Scalar Scalar;
template<typename Dest>
static void scaleAndAddTo(Dest& dst, const Lhs& lhs, const Rhs& rhs, const Scalar& alpha)
{
@ -744,7 +744,7 @@ struct generic_product_impl<Lhs,Rhs,DenseShape,TriangularShape,ProductTag>
: generic_product_impl_base<Lhs,Rhs,generic_product_impl<Lhs,Rhs,DenseShape,TriangularShape,ProductTag> >
{
typedef typename Product<Lhs,Rhs>::Scalar Scalar;
template<typename Dest>
static void scaleAndAddTo(Dest& dst, const Lhs& lhs, const Rhs& rhs, const Scalar& alpha)
{
@ -765,7 +765,7 @@ struct generic_product_impl<Lhs,Rhs,SelfAdjointShape,DenseShape,ProductTag>
: generic_product_impl_base<Lhs,Rhs,generic_product_impl<Lhs,Rhs,SelfAdjointShape,DenseShape,ProductTag> >
{
typedef typename Product<Lhs,Rhs>::Scalar Scalar;
template<typename Dest>
static void scaleAndAddTo(Dest& dst, const Lhs& lhs, const Rhs& rhs, const Scalar& alpha)
{
@ -778,7 +778,7 @@ struct generic_product_impl<Lhs,Rhs,DenseShape,SelfAdjointShape,ProductTag>
: generic_product_impl_base<Lhs,Rhs,generic_product_impl<Lhs,Rhs,DenseShape,SelfAdjointShape,ProductTag> >
{
typedef typename Product<Lhs,Rhs>::Scalar Scalar;
template<typename Dest>
static void scaleAndAddTo(Dest& dst, const Lhs& lhs, const Rhs& rhs, const Scalar& alpha)
{
@ -790,7 +790,7 @@ struct generic_product_impl<Lhs,Rhs,DenseShape,SelfAdjointShape,ProductTag>
/***************************************************************************
* Diagonal products
***************************************************************************/
template<typename MatrixType, typename DiagonalType, typename Derived, int ProductOrder>
struct diagonal_product_evaluator_base
: evaluator_base<Derived>
@ -799,7 +799,7 @@ struct diagonal_product_evaluator_base
public:
enum {
CoeffReadCost = NumTraits<Scalar>::MulCost + evaluator<MatrixType>::CoeffReadCost + evaluator<DiagonalType>::CoeffReadCost,
MatrixFlags = evaluator<MatrixType>::Flags,
DiagFlags = evaluator<DiagonalType>::Flags,
_StorageOrder = MatrixFlags & RowMajorBit ? RowMajor : ColMajor,
@ -817,14 +817,14 @@ public:
|| (DiagonalType::SizeAtCompileTime==Dynamic && MatrixType::RowsAtCompileTime==1 && ProductOrder==OnTheLeft)
|| (DiagonalType::SizeAtCompileTime==Dynamic && MatrixType::ColsAtCompileTime==1 && ProductOrder==OnTheRight)
};
diagonal_product_evaluator_base(const MatrixType &mat, const DiagonalType &diag)
: m_diagImpl(diag), m_matImpl(mat)
{
EIGEN_INTERNAL_CHECK_COST_VALUE(NumTraits<Scalar>::MulCost);
EIGEN_INTERNAL_CHECK_COST_VALUE(CoeffReadCost);
}
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Scalar coeff(Index idx) const
{
if(AsScalarProduct)
@ -832,7 +832,7 @@ public:
else
return m_diagImpl.coeff(idx) * m_matImpl.coeff(idx);
}
protected:
template<int LoadMode,typename PacketType>
EIGEN_STRONG_INLINE PacketType packet_impl(Index row, Index col, Index id, internal::true_type) const
@ -840,7 +840,7 @@ protected:
return internal::pmul(m_matImpl.template packet<LoadMode,PacketType>(row, col),
internal::pset1<PacketType>(m_diagImpl.coeff(id)));
}
template<int LoadMode,typename PacketType>
EIGEN_STRONG_INLINE PacketType packet_impl(Index row, Index col, Index id, internal::false_type) const
{
@ -851,7 +851,7 @@ protected:
return internal::pmul(m_matImpl.template packet<LoadMode,PacketType>(row, col),
m_diagImpl.template packet<DiagonalPacketLoadMode,PacketType>(id));
}
evaluator<DiagonalType> m_diagImpl;
evaluator<MatrixType> m_matImpl;
};
@ -866,10 +866,10 @@ struct product_evaluator<Product<Lhs, Rhs, ProductKind>, ProductTag, DiagonalSha
using Base::m_matImpl;
using Base::coeff;
typedef typename Base::Scalar Scalar;
typedef Product<Lhs, Rhs, ProductKind> XprType;
typedef typename XprType::PlainObject PlainObject;
enum {
StorageOrder = int(Rhs::Flags) & RowMajorBit ? RowMajor : ColMajor
};
@ -878,12 +878,12 @@ struct product_evaluator<Product<Lhs, Rhs, ProductKind>, ProductTag, DiagonalSha
: Base(xpr.rhs(), xpr.lhs().diagonal())
{
}
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Scalar coeff(Index row, Index col) const
{
return m_diagImpl.coeff(row) * m_matImpl.coeff(row, col);
}
#ifndef __CUDACC__
template<int LoadMode,typename PacketType>
EIGEN_STRONG_INLINE PacketType packet(Index row, Index col) const
@ -893,7 +893,7 @@ struct product_evaluator<Product<Lhs, Rhs, ProductKind>, ProductTag, DiagonalSha
return this->template packet_impl<LoadMode,PacketType>(row,col, row,
typename internal::conditional<int(StorageOrder)==RowMajor, internal::true_type, internal::false_type>::type());
}
template<int LoadMode,typename PacketType>
EIGEN_STRONG_INLINE PacketType packet(Index idx) const
{
@ -912,22 +912,22 @@ struct product_evaluator<Product<Lhs, Rhs, ProductKind>, ProductTag, DenseShape,
using Base::m_matImpl;
using Base::coeff;
typedef typename Base::Scalar Scalar;
typedef Product<Lhs, Rhs, ProductKind> XprType;
typedef typename XprType::PlainObject PlainObject;
enum { StorageOrder = int(Lhs::Flags) & RowMajorBit ? RowMajor : ColMajor };
EIGEN_DEVICE_FUNC explicit product_evaluator(const XprType& xpr)
: Base(xpr.lhs(), xpr.rhs().diagonal())
{
}
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Scalar coeff(Index row, Index col) const
{
return m_matImpl.coeff(row, col) * m_diagImpl.coeff(col);
}
#ifndef __CUDACC__
template<int LoadMode,typename PacketType>
EIGEN_STRONG_INLINE PacketType packet(Index row, Index col) const
@ -935,7 +935,7 @@ struct product_evaluator<Product<Lhs, Rhs, ProductKind>, ProductTag, DenseShape,
return this->template packet_impl<LoadMode,PacketType>(row,col, col,
typename internal::conditional<int(StorageOrder)==ColMajor, internal::true_type, internal::false_type>::type());
}
template<int LoadMode,typename PacketType>
EIGEN_STRONG_INLINE PacketType packet(Index idx) const
{
@ -1017,7 +1017,7 @@ template<typename Lhs, typename Rhs, int ProductTag, typename MatrixShape>
struct generic_product_impl<Lhs, Rhs, PermutationShape, MatrixShape, ProductTag>
{
template<typename Dest>
static void evalTo(Dest& dst, const Lhs& lhs, const Rhs& rhs)
static EIGEN_DEVICE_FUNC void evalTo(Dest& dst, const Lhs& lhs, const Rhs& rhs)
{
permutation_matrix_product<Rhs, OnTheLeft, false, MatrixShape>::run(dst, lhs, rhs);
}
@ -1027,7 +1027,7 @@ template<typename Lhs, typename Rhs, int ProductTag, typename MatrixShape>
struct generic_product_impl<Lhs, Rhs, MatrixShape, PermutationShape, ProductTag>
{
template<typename Dest>
static void evalTo(Dest& dst, const Lhs& lhs, const Rhs& rhs)
static EIGEN_DEVICE_FUNC void evalTo(Dest& dst, const Lhs& lhs, const Rhs& rhs)
{
permutation_matrix_product<Lhs, OnTheRight, false, MatrixShape>::run(dst, rhs, lhs);
}
@ -1037,7 +1037,7 @@ template<typename Lhs, typename Rhs, int ProductTag, typename MatrixShape>
struct generic_product_impl<Inverse<Lhs>, Rhs, PermutationShape, MatrixShape, ProductTag>
{
template<typename Dest>
static void evalTo(Dest& dst, const Inverse<Lhs>& lhs, const Rhs& rhs)
static EIGEN_DEVICE_FUNC void evalTo(Dest& dst, const Inverse<Lhs>& lhs, const Rhs& rhs)
{
permutation_matrix_product<Rhs, OnTheLeft, true, MatrixShape>::run(dst, lhs.nestedExpression(), rhs);
}
@ -1047,7 +1047,7 @@ template<typename Lhs, typename Rhs, int ProductTag, typename MatrixShape>
struct generic_product_impl<Lhs, Inverse<Rhs>, MatrixShape, PermutationShape, ProductTag>
{
template<typename Dest>
static void evalTo(Dest& dst, const Lhs& lhs, const Inverse<Rhs>& rhs)
static EIGEN_DEVICE_FUNC void evalTo(Dest& dst, const Lhs& lhs, const Inverse<Rhs>& rhs)
{
permutation_matrix_product<Lhs, OnTheRight, true, MatrixShape>::run(dst, rhs.nestedExpression(), lhs);
}
@ -1069,7 +1069,7 @@ struct transposition_matrix_product
{
typedef typename nested_eval<ExpressionType, 1>::type MatrixType;
typedef typename remove_all<MatrixType>::type MatrixTypeCleaned;
template<typename Dest, typename TranspositionType>
static inline void run(Dest& dst, const TranspositionType& tr, const ExpressionType& xpr)
{
@ -1094,7 +1094,7 @@ template<typename Lhs, typename Rhs, int ProductTag, typename MatrixShape>
struct generic_product_impl<Lhs, Rhs, TranspositionsShape, MatrixShape, ProductTag>
{
template<typename Dest>
static void evalTo(Dest& dst, const Lhs& lhs, const Rhs& rhs)
static EIGEN_DEVICE_FUNC void evalTo(Dest& dst, const Lhs& lhs, const Rhs& rhs)
{
transposition_matrix_product<Rhs, OnTheLeft, false, MatrixShape>::run(dst, lhs, rhs);
}
@ -1104,7 +1104,7 @@ template<typename Lhs, typename Rhs, int ProductTag, typename MatrixShape>
struct generic_product_impl<Lhs, Rhs, MatrixShape, TranspositionsShape, ProductTag>
{
template<typename Dest>
static void evalTo(Dest& dst, const Lhs& lhs, const Rhs& rhs)
static EIGEN_DEVICE_FUNC void evalTo(Dest& dst, const Lhs& lhs, const Rhs& rhs)
{
transposition_matrix_product<Lhs, OnTheRight, false, MatrixShape>::run(dst, rhs, lhs);
}
@ -1115,7 +1115,7 @@ template<typename Lhs, typename Rhs, int ProductTag, typename MatrixShape>
struct generic_product_impl<Transpose<Lhs>, Rhs, TranspositionsShape, MatrixShape, ProductTag>
{
template<typename Dest>
static void evalTo(Dest& dst, const Transpose<Lhs>& lhs, const Rhs& rhs)
static EIGEN_DEVICE_FUNC void evalTo(Dest& dst, const Transpose<Lhs>& lhs, const Rhs& rhs)
{
transposition_matrix_product<Rhs, OnTheLeft, true, MatrixShape>::run(dst, lhs.nestedExpression(), rhs);
}
@ -1125,7 +1125,7 @@ template<typename Lhs, typename Rhs, int ProductTag, typename MatrixShape>
struct generic_product_impl<Lhs, Transpose<Rhs>, MatrixShape, TranspositionsShape, ProductTag>
{
template<typename Dest>
static void evalTo(Dest& dst, const Lhs& lhs, const Transpose<Rhs>& rhs)
static EIGEN_DEVICE_FUNC void evalTo(Dest& dst, const Lhs& lhs, const Transpose<Rhs>& rhs)
{
transposition_matrix_product<Lhs, OnTheRight, true, MatrixShape>::run(dst, rhs.nestedExpression(), lhs);
}

24
Eigen/src/Core/Random.h
View File

@ -10,7 +10,7 @@
#ifndef EIGEN_RANDOM_H
#define EIGEN_RANDOM_H
namespace Eigen {
namespace Eigen {
namespace internal {
@ -29,16 +29,16 @@ struct functor_traits<scalar_random_op<Scalar> >
*
* Numbers are uniformly spread through their whole definition range for integer types,
* and in the [-1:1] range for floating point scalar types.
*
*
* The parameters \a rows and \a cols are the number of rows and of columns of
* the returned matrix. Must be compatible with this MatrixBase type.
*
* \not_reentrant
*
*
* This variant is meant to be used for dynamic-size matrix types. For fixed-size types,
* it is redundant to pass \a rows and \a cols as arguments, so Random() should be used
* instead.
*
*
*
* Example: \include MatrixBase_random_int_int.cpp
* Output: \verbinclude MatrixBase_random_int_int.out
@ -46,7 +46,7 @@ struct functor_traits<scalar_random_op<Scalar> >
* This expression has the "evaluate before nesting" flag so that it will be evaluated into
* a temporary matrix whenever it is nested in a larger expression. This prevents unexpected
* behavior with expressions involving random matrices.
*
*
* See DenseBase::NullaryExpr(Index, const CustomNullaryOp&) for an example using C++11 random generators.
*
* \sa DenseBase::setRandom(), DenseBase::Random(Index), DenseBase::Random()
@ -93,7 +93,7 @@ DenseBase<Derived>::Random(Index size)
*
* Numbers are uniformly spread through their whole definition range for integer types,
* and in the [-1:1] range for floating point scalar types.
*
*
* This variant is only for fixed-size MatrixBase types. For dynamic-size types, you
* need to use the variants taking size arguments.
*
@ -103,7 +103,7 @@ DenseBase<Derived>::Random(Index size)
* This expression has the "evaluate before nesting" flag so that it will be evaluated into
* a temporary matrix whenever it is nested in a larger expression. This prevents unexpected
* behavior with expressions involving random matrices.
*
*
* \not_reentrant
*
* \sa DenseBase::setRandom(), DenseBase::Random(Index,Index), DenseBase::Random(Index)
@ -119,16 +119,16 @@ DenseBase<Derived>::Random()
*
* Numbers are uniformly spread through their whole definition range for integer types,
* and in the [-1:1] range for floating point scalar types.
*
*
* \not_reentrant
*
*
* Example: \include MatrixBase_setRandom.cpp
* Output: \verbinclude MatrixBase_setRandom.out
*
* \sa class CwiseNullaryOp, setRandom(Index), setRandom(Index,Index)
*/
template<typename Derived>
inline Derived& DenseBase<Derived>::setRandom()
EIGEN_DEVICE_FUNC inline Derived& DenseBase<Derived>::setRandom()
{
return *this = Random(rows(), cols());
}
@ -137,7 +137,7 @@ inline Derived& DenseBase<Derived>::setRandom()
*
* Numbers are uniformly spread through their whole definition range for integer types,
* and in the [-1:1] range for floating point scalar types.
*
*
* \only_for_vectors
* \not_reentrant
*
@ -160,7 +160,7 @@ PlainObjectBase<Derived>::setRandom(Index newSize)
* and in the [-1:1] range for floating point scalar types.
*
* \not_reentrant
*
*
* \param rows the new number of rows
* \param cols the new number of columns
*

38
Eigen/src/Core/Redux.h
View File

@ -11,7 +11,7 @@
#ifndef EIGEN_REDUX_H
#define EIGEN_REDUX_H
namespace Eigen {
namespace Eigen {
namespace internal {
@ -60,7 +60,7 @@ public:
enum {
Unrolling = Cost <= UnrollingLimit ? CompleteUnrolling : NoUnrolling
};
#ifdef EIGEN_DEBUG_ASSIGN
static void debug()
{
@ -128,7 +128,7 @@ template<typename Func, typename Derived, int Start>
struct redux_novec_unroller<Func, Derived, Start, 0>
{
typedef typename Derived::Scalar Scalar;
EIGEN_DEVICE_FUNC
EIGEN_DEVICE_FUNC
static EIGEN_STRONG_INLINE Scalar run(const Derived&, const Func&) { return Scalar(); }
};
@ -215,7 +215,7 @@ struct redux_impl<Func, Derived, LinearVectorizedTraversal, NoUnrolling>
static Scalar run(const Derived &mat, const Func& func)
{
const Index size = mat.size();
const Index packetSize = redux_traits<Func, Derived>::PacketSize;
const int packetAlignment = unpacket_traits<PacketScalar>::alignment;
enum {
@ -336,12 +336,12 @@ class redux_evaluator
public:
typedef _XprType XprType;
EIGEN_DEVICE_FUNC explicit redux_evaluator(const XprType &xpr) : m_evaluator(xpr), m_xpr(xpr) {}
typedef typename XprType::Scalar Scalar;
typedef typename XprType::CoeffReturnType CoeffReturnType;
typedef typename XprType::PacketScalar PacketScalar;
typedef typename XprType::PacketReturnType PacketReturnType;
enum {
MaxRowsAtCompileTime = XprType::MaxRowsAtCompileTime,
MaxColsAtCompileTime = XprType::MaxColsAtCompileTime,
@ -353,7 +353,7 @@ public:
CoeffReadCost = evaluator<XprType>::CoeffReadCost,
Alignment = evaluator<XprType>::Alignment
};
EIGEN_DEVICE_FUNC Index rows() const { return m_xpr.rows(); }
EIGEN_DEVICE_FUNC Index cols() const { return m_xpr.cols(); }
EIGEN_DEVICE_FUNC Index size() const { return m_xpr.size(); }
@ -375,17 +375,17 @@ public:
template<int LoadMode, typename PacketType>
PacketType packet(Index index) const
{ return m_evaluator.template packet<LoadMode,PacketType>(index); }
EIGEN_DEVICE_FUNC
CoeffReturnType coeffByOuterInner(Index outer, Index inner) const
{ return m_evaluator.coeff(IsRowMajor ? outer : inner, IsRowMajor ? inner : outer); }
template<int LoadMode, typename PacketType>
PacketType packetByOuterInner(Index outer, Index inner) const
{ return m_evaluator.template packet<LoadMode,PacketType>(IsRowMajor ? outer : inner, IsRowMajor ? inner : outer); }
const XprType & nestedExpression() const { return m_xpr; }
protected:
internal::evaluator<XprType> m_evaluator;
const XprType &m_xpr;
@ -407,14 +407,14 @@ protected:
*/
template<typename Derived>
template<typename Func>
EIGEN_STRONG_INLINE typename internal::traits<Derived>::Scalar
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE typename internal::traits<Derived>::Scalar
DenseBase<Derived>::redux(const Func& func) const
{
eigen_assert(this->rows()>0 && this->cols()>0 && "you are using an empty matrix");
typedef typename internal::redux_evaluator<Derived> ThisEvaluator;
ThisEvaluator thisEval(derived());
return internal::redux_impl<Func, ThisEvaluator>::run(thisEval, func);
}
@ -422,7 +422,7 @@ DenseBase<Derived>::redux(const Func& func) const
* \warning the result is undefined if \c *this contains NaN.
*/
template<typename Derived>
EIGEN_STRONG_INLINE typename internal::traits<Derived>::Scalar
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE typename internal::traits<Derived>::Scalar
DenseBase<Derived>::minCoeff() const
{
return derived().redux(Eigen::internal::scalar_min_op<Scalar,Scalar>());
@ -432,7 +432,7 @@ DenseBase<Derived>::minCoeff() const
* \warning the result is undefined if \c *this contains NaN.
*/
template<typename Derived>
EIGEN_STRONG_INLINE typename internal::traits<Derived>::Scalar
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE typename internal::traits<Derived>::Scalar
DenseBase<Derived>::maxCoeff() const
{
return derived().redux(Eigen::internal::scalar_max_op<Scalar,Scalar>());
@ -445,7 +445,7 @@ DenseBase<Derived>::maxCoeff() const
* \sa trace(), prod(), mean()
*/
template<typename Derived>
EIGEN_STRONG_INLINE typename internal::traits<Derived>::Scalar
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE typename internal::traits<Derived>::Scalar
DenseBase<Derived>::sum() const
{
if(SizeAtCompileTime==0 || (SizeAtCompileTime==Dynamic && size()==0))
@ -458,7 +458,7 @@ DenseBase<Derived>::sum() const
* \sa trace(), prod(), sum()
*/
template<typename Derived>
EIGEN_STRONG_INLINE typename internal::traits<Derived>::Scalar
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE typename internal::traits<Derived>::Scalar
DenseBase<Derived>::mean() const
{
#ifdef __INTEL_COMPILER
@ -479,7 +479,7 @@ DenseBase<Derived>::mean() const
* \sa sum(), mean(), trace()
*/
template<typename Derived>
EIGEN_STRONG_INLINE typename internal::traits<Derived>::Scalar
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE typename internal::traits<Derived>::Scalar
DenseBase<Derived>::prod() const
{
if(SizeAtCompileTime==0 || (SizeAtCompileTime==Dynamic && size()==0))
@ -494,7 +494,7 @@ DenseBase<Derived>::prod() const
* \sa diagonal(), sum()
*/
template<typename Derived>
EIGEN_STRONG_INLINE typename internal::traits<Derived>::Scalar
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE typename internal::traits<Derived>::Scalar
MatrixBase<Derived>::trace() const
{
return derived().diagonal().sum();

12
Eigen/src/Core/Replicate.h
View File

@ -10,7 +10,7 @@
#ifndef EIGEN_REPLICATE_H
#define EIGEN_REPLICATE_H
namespace Eigen {
namespace Eigen {
namespace internal {
template<typename MatrixType,int RowFactor,int ColFactor>
@ -35,7 +35,7 @@ struct traits<Replicate<MatrixType,RowFactor,ColFactor> >
IsRowMajor = MaxRowsAtCompileTime==1 && MaxColsAtCompileTime!=1 ? 1
: MaxColsAtCompileTime==1 && MaxRowsAtCompileTime!=1 ? 0
: (MatrixType::Flags & RowMajorBit) ? 1 : 0,
// FIXME enable DirectAccess with negative strides?
Flags = IsRowMajor ? RowMajorBit : 0
};
@ -95,8 +95,8 @@ template<typename MatrixType,int RowFactor,int ColFactor> class Replicate
EIGEN_DEVICE_FUNC
const _MatrixTypeNested& nestedExpression() const
{
return m_matrix;
{
return m_matrix;
}
protected:
@ -115,7 +115,7 @@ template<typename MatrixType,int RowFactor,int ColFactor> class Replicate
*/
template<typename Derived>
template<int RowFactor, int ColFactor>
const Replicate<Derived,RowFactor,ColFactor>
EIGEN_DEVICE_FUNC const Replicate<Derived,RowFactor,ColFactor>
DenseBase<Derived>::replicate() const
{
return Replicate<Derived,RowFactor,ColFactor>(derived());
@ -130,7 +130,7 @@ DenseBase<Derived>::replicate() const
* \sa VectorwiseOp::replicate(), DenseBase::replicate(), class Replicate
*/
template<typename ExpressionType, int Direction>
const typename VectorwiseOp<ExpressionType,Direction>::ReplicateReturnType
EIGEN_DEVICE_FUNC const typename VectorwiseOp<ExpressionType,Direction>::ReplicateReturnType
VectorwiseOp<ExpressionType,Direction>::replicate(Index factor) const
{
return typename VectorwiseOp<ExpressionType,Direction>::ReplicateReturnType

6
Eigen/src/Core/ReturnByValue.h
View File

@ -79,7 +79,7 @@ template<typename Derived> class ReturnByValue
template<typename Derived>
template<typename OtherDerived>
Derived& DenseBase<Derived>::operator=(const ReturnByValue<OtherDerived>& other)
EIGEN_DEVICE_FUNC Derived& DenseBase<Derived>::operator=(const ReturnByValue<OtherDerived>& other)
{
other.evalTo(derived());
return derived();
@ -90,7 +90,7 @@ namespace internal {
// Expression is evaluated in a temporary; default implementation of Assignment is bypassed so that
// when a ReturnByValue expression is assigned, the evaluator is not constructed.
// TODO: Finalize port to new regime; ReturnByValue should not exist in the expression world
template<typename Derived>
struct evaluator<ReturnByValue<Derived> >
: public evaluator<typename internal::traits<Derived>::ReturnType>
@ -98,7 +98,7 @@ struct evaluator<ReturnByValue<Derived> >
typedef ReturnByValue<Derived> XprType;
typedef typename internal::traits<Derived>::ReturnType PlainObject;
typedef evaluator<PlainObject> Base;
EIGEN_DEVICE_FUNC explicit evaluator(const XprType& xpr)
: m_result(xpr.rows(), xpr.cols())
{

14
Eigen/src/Core/Reverse.h
View File

@ -12,7 +12,7 @@
#ifndef EIGEN_REVERSE_H
#define EIGEN_REVERSE_H
namespace Eigen {
namespace Eigen {
namespace internal {
@ -44,7 +44,7 @@ template<typename PacketType> struct reverse_packet_cond<PacketType,false>
static inline PacketType run(const PacketType& x) { return x; }
};
} // end namespace internal
} // end namespace internal
/** \class Reverse
* \ingroup Core_Module
@ -98,7 +98,7 @@ template<typename MatrixType, int Direction> class Reverse
}
EIGEN_DEVICE_FUNC const typename internal::remove_all<typename MatrixType::Nested>::type&
nestedExpression() const
nestedExpression() const
{
return m_matrix;
}
@ -114,7 +114,7 @@ template<typename MatrixType, int Direction> class Reverse
*
*/
template<typename Derived>
inline typename DenseBase<Derived>::ReverseReturnType
EIGEN_DEVICE_FUNC inline typename DenseBase<Derived>::ReverseReturnType
DenseBase<Derived>::reverse()
{
return ReverseReturnType(derived());
@ -136,7 +136,7 @@ DenseBase<Derived>::reverse()
*
* \sa VectorwiseOp::reverseInPlace(), reverse() */
template<typename Derived>
inline void DenseBase<Derived>::reverseInPlace()
EIGEN_DEVICE_FUNC inline void DenseBase<Derived>::reverseInPlace()
{
if(cols()>rows())
{
@ -161,7 +161,7 @@ inline void DenseBase<Derived>::reverseInPlace()
}
namespace internal {
template<int Direction>
struct vectorwise_reverse_inplace_impl;
@ -201,7 +201,7 @@ struct vectorwise_reverse_inplace_impl<Horizontal>
*
* \sa DenseBase::reverseInPlace(), reverse() */
template<typename ExpressionType, int Direction>
void VectorwiseOp<ExpressionType,Direction>::reverseInPlace()
EIGEN_DEVICE_FUNC void VectorwiseOp<ExpressionType,Direction>::reverseInPlace()
{
internal::vectorwise_reverse_inplace_impl<Direction>::run(_expression().const_cast_derived());
}

22
Eigen/src/Core/SelfAdjointView.h
View File

@ -10,7 +10,7 @@
#ifndef EIGEN_SELFADJOINTMATRIX_H
#define EIGEN_SELFADJOINTMATRIX_H
namespace Eigen {
namespace Eigen {
/** \class SelfAdjointView
* \ingroup Core_Module
@ -58,7 +58,7 @@ template<typename _MatrixType, unsigned int UpLo> class SelfAdjointView
typedef MatrixTypeNestedCleaned NestedExpression;
/** \brief The type of coefficients in this matrix */
typedef typename internal::traits<SelfAdjointView>::Scalar Scalar;
typedef typename internal::traits<SelfAdjointView>::Scalar Scalar;
typedef typename MatrixType::StorageIndex StorageIndex;
typedef typename internal::remove_all<typename MatrixType::ConjugateReturnType>::type MatrixConjugateReturnType;
@ -131,7 +131,7 @@ template<typename _MatrixType, unsigned int UpLo> class SelfAdjointView
{
return Product<OtherDerived,SelfAdjointView>(lhs.derived(),rhs);
}
friend EIGEN_DEVICE_FUNC
const SelfAdjointView<const EIGEN_SCALAR_BINARYOP_EXPR_RETURN_TYPE(Scalar,MatrixType,product),UpLo>
operator*(const Scalar& s, const SelfAdjointView& mat)
@ -287,17 +287,17 @@ protected:
using Base::m_src;
using Base::m_functor;
public:
typedef typename Base::DstEvaluatorType DstEvaluatorType;
typedef typename Base::SrcEvaluatorType SrcEvaluatorType;
typedef typename Base::Scalar Scalar;
typedef typename Base::AssignmentTraits AssignmentTraits;
EIGEN_DEVICE_FUNC triangular_dense_assignment_kernel(DstEvaluatorType &dst, const SrcEvaluatorType &src, const Functor &func, DstXprType& dstExpr)
: Base(dst, src, func, dstExpr)
{}
EIGEN_DEVICE_FUNC void assignCoeff(Index row, Index col)
{
eigen_internal_assert(row!=col);
@ -305,12 +305,12 @@ public:
m_functor.assignCoeff(m_dst.coeffRef(row,col), tmp);
m_functor.assignCoeff(m_dst.coeffRef(col,row), numext::conj(tmp));
}
EIGEN_DEVICE_FUNC void assignDiagonalCoeff(Index id)
{
Base::assignCoeff(id,id);
}
EIGEN_DEVICE_FUNC void assignOppositeCoeff(Index, Index)
{ eigen_internal_assert(false && "should never be called"); }
};
@ -324,7 +324,7 @@ public:
/** This is the const version of MatrixBase::selfadjointView() */
template<typename Derived>
template<unsigned int UpLo>
typename MatrixBase<Derived>::template ConstSelfAdjointViewReturnType<UpLo>::Type
EIGEN_DEVICE_FUNC typename MatrixBase<Derived>::template ConstSelfAdjointViewReturnType<UpLo>::Type
MatrixBase<Derived>::selfadjointView() const
{
return typename ConstSelfAdjointViewReturnType<UpLo>::Type(derived());
@ -341,7 +341,7 @@ MatrixBase<Derived>::selfadjointView() const
*/
template<typename Derived>
template<unsigned int UpLo>
typename MatrixBase<Derived>::template SelfAdjointViewReturnType<UpLo>::Type
EIGEN_DEVICE_FUNC typename MatrixBase<Derived>::template SelfAdjointViewReturnType<UpLo>::Type
MatrixBase<Derived>::selfadjointView()
{
return typename SelfAdjointViewReturnType<UpLo>::Type(derived());

26
Eigen/src/Core/Solve.h
View File

@ -13,7 +13,7 @@
namespace Eigen {
template<typename Decomposition, typename RhsType, typename StorageKind> class SolveImpl;
/** \class Solve
* \ingroup Core_Module
*
@ -64,11 +64,11 @@ class Solve : public SolveImpl<Decomposition,RhsType,typename internal::traits<R
public:
typedef typename internal::traits<Solve>::PlainObject PlainObject;
typedef typename internal::traits<Solve>::StorageIndex StorageIndex;
Solve(const Decomposition &dec, const RhsType &rhs)
: m_dec(dec), m_rhs(rhs)
{}
EIGEN_DEVICE_FUNC Index rows() const { return m_dec.cols(); }
EIGEN_DEVICE_FUNC Index cols() const { return m_rhs.cols(); }
@ -87,14 +87,14 @@ class SolveImpl<Decomposition,RhsType,Dense>
: public MatrixBase<Solve<Decomposition,RhsType> >
{
typedef Solve<Decomposition,RhsType> Derived;
public:
typedef MatrixBase<Solve<Decomposition,RhsType> > Base;
EIGEN_DENSE_PUBLIC_INTERFACE(Derived)
private:
Scalar coeff(Index row, Index col) const;
Scalar coeff(Index i) const;
};
@ -119,15 +119,15 @@ struct evaluator<Solve<Decomposition,RhsType> >
typedef evaluator<PlainObject> Base;
enum { Flags = Base::Flags | EvalBeforeNestingBit };
EIGEN_DEVICE_FUNC explicit evaluator(const SolveType& solve)
: m_result(solve.rows(), solve.cols())
{
::new (static_cast<Base*>(this)) Base(m_result);
solve.dec()._solve_impl(solve.rhs(), m_result);
}
protected:
protected:
PlainObject m_result;
};
@ -137,7 +137,7 @@ template<typename DstXprType, typename DecType, typename RhsType, typename Scala
struct Assignment<DstXprType, Solve<DecType,RhsType>, internal::assign_op<Scalar,Scalar>, Dense2Dense>
{
typedef Solve<DecType,RhsType> SrcXprType;
static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op<Scalar,Scalar> &)
static EIGEN_DEVICE_FUNC void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op<Scalar,Scalar> &)
{
Index dstRows = src.rows();
Index dstCols = src.cols();
@ -153,7 +153,7 @@ template<typename DstXprType, typename DecType, typename RhsType, typename Scala
struct Assignment<DstXprType, Solve<Transpose<const DecType>,RhsType>, internal::assign_op<Scalar,Scalar>, Dense2Dense>
{
typedef Solve<Transpose<const DecType>,RhsType> SrcXprType;
static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op<Scalar,Scalar> &)
static EIGEN_DEVICE_FUNC void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op<Scalar,Scalar> &)
{
Index dstRows = src.rows();
Index dstCols = src.cols();
@ -170,13 +170,13 @@ struct Assignment<DstXprType, Solve<CwiseUnaryOp<internal::scalar_conjugate_op<t
internal::assign_op<Scalar,Scalar>, Dense2Dense>
{
typedef Solve<CwiseUnaryOp<internal::scalar_conjugate_op<typename DecType::Scalar>, const Transpose<const DecType> >,RhsType> SrcXprType;
static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op<Scalar,Scalar> &)
static EIGEN_DEVICE_FUNC void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op<Scalar,Scalar> &)
{
Index dstRows = src.rows();
Index dstCols = src.cols();
if((dst.rows()!=dstRows) || (dst.cols()!=dstCols))
dst.resize(dstRows, dstCols);
src.dec().nestedExpression().nestedExpression().template _solve_impl_transposed<true>(src.rhs(), dst);
}
};

10
Eigen/src/Core/SolveTriangular.h
View File

@ -10,7 +10,7 @@
#ifndef EIGEN_SOLVETRIANGULAR_H
#define EIGEN_SOLVETRIANGULAR_H
namespace Eigen {
namespace Eigen {
namespace internal {
@ -64,7 +64,7 @@ struct triangular_solver_selector<Lhs,Rhs,Side,Mode,NoUnrolling,1>
ei_declare_aligned_stack_constructed_variable(RhsScalar,actualRhs,rhs.size(),
(useRhsDirectly ? rhs.data() : 0));
if(!useRhsDirectly)
MappedRhs(actualRhs,rhs.size()) = rhs;
@ -148,7 +148,7 @@ struct triangular_solver_selector<Lhs,Rhs,OnTheRight,Mode,CompleteUnrolling,1> {
{
Transpose<const Lhs> trLhs(lhs);
Transpose<Rhs> trRhs(rhs);
triangular_solver_unroller<Transpose<const Lhs>,Transpose<Rhs>,
((Mode&Upper)==Upper ? Lower : Upper) | (Mode&UnitDiag),
0,Rhs::SizeAtCompileTime>::run(trLhs,trRhs);
@ -164,7 +164,7 @@ struct triangular_solver_selector<Lhs,Rhs,OnTheRight,Mode,CompleteUnrolling,1> {
#ifndef EIGEN_PARSED_BY_DOXYGEN
template<typename MatrixType, unsigned int Mode>
template<int Side, typename OtherDerived>
void TriangularViewImpl<MatrixType,Mode,Dense>::solveInPlace(const MatrixBase<OtherDerived>& _other) const
EIGEN_DEVICE_FUNC void TriangularViewImpl<MatrixType,Mode,Dense>::solveInPlace(const MatrixBase<OtherDerived>& _other) const
{
OtherDerived& other = _other.const_cast_derived();
eigen_assert( derived().cols() == derived().rows() && ((Side==OnTheLeft && derived().cols() == other.rows()) || (Side==OnTheRight && derived().cols() == other.cols())) );
@ -187,7 +187,7 @@ void TriangularViewImpl<MatrixType,Mode,Dense>::solveInPlace(const MatrixBase<Ot
template<typename Derived, unsigned int Mode>
template<int Side, typename Other>
const internal::triangular_solve_retval<Side,TriangularView<Derived,Mode>,Other>
EIGEN_DEVICE_FUNC const internal::triangular_solve_retval<Side,TriangularView<Derived,Mode>,Other>
TriangularViewImpl<Derived,Mode,Dense>::solve(const MatrixBase<Other>& other) const
{
return internal::triangular_solve_retval<Side,TriangularViewType,Other>(derived(), other.derived());

14
Eigen/src/Core/Transpose.h
View File

@ -11,7 +11,7 @@
#ifndef EIGEN_TRANSPOSE_H
#define EIGEN_TRANSPOSE_H
namespace Eigen {
namespace Eigen {
namespace internal {
template<typename MatrixType>
@ -170,7 +170,7 @@ template<typename MatrixType> class TransposeImpl<MatrixType,Dense>
*
* \sa transposeInPlace(), adjoint() */
template<typename Derived>
inline Transpose<Derived>
EIGEN_DEVICE_FUNC inline Transpose<Derived>
DenseBase<Derived>::transpose()
{
return TransposeReturnType(derived());
@ -182,7 +182,7 @@ DenseBase<Derived>::transpose()
*
* \sa transposeInPlace(), adjoint() */
template<typename Derived>
inline typename DenseBase<Derived>::ConstTransposeReturnType
EIGEN_DEVICE_FUNC inline typename DenseBase<Derived>::ConstTransposeReturnType
DenseBase<Derived>::transpose() const
{
return ConstTransposeReturnType(derived());
@ -208,7 +208,7 @@ DenseBase<Derived>::transpose() const
*
* \sa adjointInPlace(), transpose(), conjugate(), class Transpose, class internal::scalar_conjugate_op */
template<typename Derived>
inline const typename MatrixBase<Derived>::AdjointReturnType
EIGEN_DEVICE_FUNC inline const typename MatrixBase<Derived>::AdjointReturnType
MatrixBase<Derived>::adjoint() const
{
return AdjointReturnType(this->transpose());
@ -278,12 +278,12 @@ struct inplace_transpose_selector<MatrixType,false,MatchPacketSize> { // non squ
* Notice however that this method is only useful if you want to replace a matrix by its own transpose.
* If you just need the transpose of a matrix, use transpose().
*
* \note if the matrix is not square, then \c *this must be a resizable matrix.
* \note if the matrix is not square, then \c *this must be a resizable matrix.
* This excludes (non-square) fixed-size matrices, block-expressions and maps.
*
* \sa transpose(), adjoint(), adjointInPlace() */
template<typename Derived>
inline void DenseBase<Derived>::transposeInPlace()
EIGEN_DEVICE_FUNC inline void DenseBase<Derived>::transposeInPlace()
{
eigen_assert((rows() == cols() || (RowsAtCompileTime == Dynamic && ColsAtCompileTime == Dynamic))
&& "transposeInPlace() called on a non-square non-resizable matrix");
@ -314,7 +314,7 @@ inline void DenseBase<Derived>::transposeInPlace()
*
* \sa transpose(), adjoint(), transposeInPlace() */
template<typename Derived>
inline void MatrixBase<Derived>::adjointInPlace()
EIGEN_DEVICE_FUNC inline void MatrixBase<Derived>::adjointInPlace()
{
derived() = adjoint().eval();
}

98
Eigen/src/Core/TriangularMatrix.h
View File

@ -11,12 +11,12 @@
#ifndef EIGEN_TRIANGULARMATRIX_H
#define EIGEN_TRIANGULARMATRIX_H
namespace Eigen {
namespace Eigen {
namespace internal {
template<int Side, typename TriangularType, typename Rhs> struct triangular_solve_retval;
}
/** \class TriangularBase
@ -34,16 +34,16 @@ template<typename Derived> class TriangularBase : public EigenBase<Derived>
ColsAtCompileTime = internal::traits<Derived>::ColsAtCompileTime,
MaxRowsAtCompileTime = internal::traits<Derived>::MaxRowsAtCompileTime,
MaxColsAtCompileTime = internal::traits<Derived>::MaxColsAtCompileTime,
SizeAtCompileTime = (internal::size_at_compile_time<internal::traits<Derived>::RowsAtCompileTime,
internal::traits<Derived>::ColsAtCompileTime>::ret),
/**< This is equal to the number of coefficients, i.e. the number of
* rows times the number of columns, or to \a Dynamic if this is not
* known at compile-time. \sa RowsAtCompileTime, ColsAtCompileTime */
MaxSizeAtCompileTime = (internal::size_at_compile_time<internal::traits<Derived>::MaxRowsAtCompileTime,
internal::traits<Derived>::MaxColsAtCompileTime>::ret)
};
typedef typename internal::traits<Derived>::Scalar Scalar;
typedef typename internal::traits<Derived>::StorageKind StorageKind;
@ -63,7 +63,7 @@ template<typename Derived> class TriangularBase : public EigenBase<Derived>
inline Index outerStride() const { return derived().outerStride(); }
EIGEN_DEVICE_FUNC
inline Index innerStride() const { return derived().innerStride(); }
// dummy resize function
void resize(Index rows, Index cols)
{
@ -155,7 +155,7 @@ template<typename Derived> class TriangularBase : public EigenBase<Derived>
* \param MatrixType the type of the object in which we are taking the triangular part
* \param Mode the kind of triangular matrix expression to construct. Can be #Upper,
* #Lower, #UnitUpper, #UnitLower, #StrictlyUpper, or #StrictlyLower.
* This is in fact a bit field; it must have either #Upper or #Lower,
* This is in fact a bit field; it must have either #Upper or #Lower,
* and additionally it may have #UnitDiag or #ZeroDiag or neither.
*
* This class represents a triangular part of a matrix, not necessarily square. Strictly speaking, for rectangular
@ -197,7 +197,7 @@ template<typename _MatrixType, unsigned int _Mode> class TriangularView
typedef typename internal::traits<TriangularView>::MatrixTypeNestedNonRef MatrixTypeNestedNonRef;
typedef typename internal::remove_all<typename MatrixType::ConjugateReturnType>::type MatrixConjugateReturnType;
public:
typedef typename internal::traits<TriangularView>::StorageKind StorageKind;
@ -216,7 +216,7 @@ template<typename _MatrixType, unsigned int _Mode> class TriangularView
EIGEN_DEVICE_FUNC
explicit inline TriangularView(MatrixType& matrix) : m_matrix(matrix)
{}
EIGEN_INHERIT_ASSIGNMENT_OPERATORS(TriangularView)
/** \copydoc EigenBase::rows() */
@ -233,7 +233,7 @@ template<typename _MatrixType, unsigned int _Mode> class TriangularView
/** \returns a reference to the nested expression */
EIGEN_DEVICE_FUNC
NestedExpression& nestedExpression() { return m_matrix; }
typedef TriangularView<const MatrixConjugateReturnType,Mode> ConjugateReturnType;
/** \sa MatrixBase::conjugate() const */
EIGEN_DEVICE_FUNC
@ -255,7 +255,7 @@ template<typename _MatrixType, unsigned int _Mode> class TriangularView
typename MatrixType::TransposeReturnType tmp(m_matrix);
return TransposeReturnType(tmp);
}
typedef TriangularView<const typename MatrixType::ConstTransposeReturnType,TransposeMode> ConstTransposeReturnType;
/** \sa MatrixBase::transpose() const */
EIGEN_DEVICE_FUNC
@ -266,10 +266,10 @@ template<typename _MatrixType, unsigned int _Mode> class TriangularView
template<typename Other>
EIGEN_DEVICE_FUNC
inline const Solve<TriangularView, Other>
inline const Solve<TriangularView, Other>
solve(const MatrixBase<Other>& other) const
{ return Solve<TriangularView, Other>(*this, other.derived()); }
// workaround MSVC ICE
#if EIGEN_COMP_MSVC
template<int Side, typename Other>
@ -313,7 +313,7 @@ template<typename _MatrixType, unsigned int _Mode> class TriangularView
else
return m_matrix.diagonal().prod();
}
protected:
MatrixTypeNested m_matrix;
@ -375,7 +375,7 @@ template<typename _MatrixType, unsigned int _Mode> class TriangularViewImpl<_Mat
internal::call_assignment_no_alias(derived(), other.derived(), internal::sub_assign_op<Scalar,typename Other::Scalar>());
return derived();
}
/** \sa MatrixBase::operator*=() */
EIGEN_DEVICE_FUNC
TriangularViewType& operator*=(const typename internal::traits<MatrixType>::Scalar& other) { return *this = derived().nestedExpression() * other; }
@ -556,7 +556,7 @@ template<typename _MatrixType, unsigned int _Mode> class TriangularViewImpl<_Mat
// FIXME should we keep that possibility
template<typename MatrixType, unsigned int Mode>
template<typename OtherDerived>
inline TriangularView<MatrixType, Mode>&
EIGEN_DEVICE_FUNC inline TriangularView<MatrixType, Mode>&
TriangularViewImpl<MatrixType, Mode, Dense>::operator=(const MatrixBase<OtherDerived>& other)
{
internal::call_assignment_no_alias(derived(), other.derived(), internal::assign_op<Scalar,typename OtherDerived::Scalar>());
@ -566,7 +566,7 @@ TriangularViewImpl<MatrixType, Mode, Dense>::operator=(const MatrixBase<OtherDer
// FIXME should we keep that possibility
template<typename MatrixType, unsigned int Mode>
template<typename OtherDerived>
void TriangularViewImpl<MatrixType, Mode, Dense>::lazyAssign(const MatrixBase<OtherDerived>& other)
EIGEN_DEVICE_FUNC void TriangularViewImpl<MatrixType, Mode, Dense>::lazyAssign(const MatrixBase<OtherDerived>& other)
{
internal::call_assignment_no_alias(derived(), other.template triangularView<Mode>());
}
@ -575,7 +575,7 @@ void TriangularViewImpl<MatrixType, Mode, Dense>::lazyAssign(const MatrixBase<Ot
template<typename MatrixType, unsigned int Mode>
template<typename OtherDerived>
inline TriangularView<MatrixType, Mode>&
EIGEN_DEVICE_FUNC inline TriangularView<MatrixType, Mode>&
TriangularViewImpl<MatrixType, Mode, Dense>::operator=(const TriangularBase<OtherDerived>& other)
{
eigen_assert(Mode == int(OtherDerived::Mode));
@ -585,7 +585,7 @@ TriangularViewImpl<MatrixType, Mode, Dense>::operator=(const TriangularBase<Othe
template<typename MatrixType, unsigned int Mode>
template<typename OtherDerived>
void TriangularViewImpl<MatrixType, Mode, Dense>::lazyAssign(const TriangularBase<OtherDerived>& other)
EIGEN_DEVICE_FUNC void TriangularViewImpl<MatrixType, Mode, Dense>::lazyAssign(const TriangularBase<OtherDerived>& other)
{
eigen_assert(Mode == int(OtherDerived::Mode));
internal::call_assignment_no_alias(derived(), other.derived());
@ -600,7 +600,7 @@ void TriangularViewImpl<MatrixType, Mode, Dense>::lazyAssign(const TriangularBas
* If the matrix is triangular, the opposite part is set to zero. */
template<typename Derived>
template<typename DenseDerived>
void TriangularBase<Derived>::evalTo(MatrixBase<DenseDerived> &other) const
EIGEN_DEVICE_FUNC void TriangularBase<Derived>::evalTo(MatrixBase<DenseDerived> &other) const
{
evalToLazy(other.derived());
}
@ -626,7 +626,7 @@ void TriangularBase<Derived>::evalTo(MatrixBase<DenseDerived> &other) const
*/
template<typename Derived>
template<unsigned int Mode>
typename MatrixBase<Derived>::template TriangularViewReturnType<Mode>::Type
EIGEN_DEVICE_FUNC typename MatrixBase<Derived>::template TriangularViewReturnType<Mode>::Type
MatrixBase<Derived>::triangularView()
{
return typename TriangularViewReturnType<Mode>::Type(derived());
@ -635,7 +635,7 @@ MatrixBase<Derived>::triangularView()
/** This is the const version of MatrixBase::triangularView() */
template<typename Derived>
template<unsigned int Mode>
typename MatrixBase<Derived>::template ConstTriangularViewReturnType<Mode>::Type
EIGEN_DEVICE_FUNC typename MatrixBase<Derived>::template ConstTriangularViewReturnType<Mode>::Type
MatrixBase<Derived>::triangularView() const
{
return typename ConstTriangularViewReturnType<Mode>::Type(derived());
@ -700,7 +700,7 @@ bool MatrixBase<Derived>::isLowerTriangular(const RealScalar& prec) const
namespace internal {
// TODO currently a triangular expression has the form TriangularView<.,.>
// in the future triangular-ness should be defined by the expression traits
// such that Transpose<TriangularView<.,.> > is valid. (currently TriangularBase::transpose() is overloaded to make it work)
@ -728,7 +728,7 @@ struct Dense2Triangular {};
template<typename Kernel, unsigned int Mode, int UnrollCount, bool ClearOpposite> struct triangular_assignment_loop;
/** \internal Specialization of the dense assignment kernel for triangular matrices.
* The main difference is that the triangular, diagonal, and opposite parts are processed through three different functions.
* \tparam UpLo must be either Lower or Upper
@ -745,17 +745,17 @@ protected:
using Base::m_src;
using Base::m_functor;
public:
typedef typename Base::DstEvaluatorType DstEvaluatorType;
typedef typename Base::SrcEvaluatorType SrcEvaluatorType;
typedef typename Base::Scalar Scalar;
typedef typename Base::AssignmentTraits AssignmentTraits;
EIGEN_DEVICE_FUNC triangular_dense_assignment_kernel(DstEvaluatorType &dst, const SrcEvaluatorType &src, const Functor &func, DstXprType& dstExpr)
: Base(dst, src, func, dstExpr)
{}
#ifdef EIGEN_INTERNAL_DEBUGGING
EIGEN_DEVICE_FUNC void assignCoeff(Index row, Index col)
{
@ -765,16 +765,16 @@ public:
#else
using Base::assignCoeff;
#endif
EIGEN_DEVICE_FUNC void assignDiagonalCoeff(Index id)
{
if(Mode==UnitDiag && SetOpposite) m_functor.assignCoeff(m_dst.coeffRef(id,id), Scalar(1));
else if(Mode==ZeroDiag && SetOpposite) m_functor.assignCoeff(m_dst.coeffRef(id,id), Scalar(0));
else if(Mode==0) Base::assignCoeff(id,id);
}
EIGEN_DEVICE_FUNC void assignOppositeCoeff(Index row, Index col)
{
{
eigen_internal_assert(row!=col);
if(SetOpposite)
m_functor.assignCoeff(m_dst.coeffRef(row,col), Scalar(0));
@ -795,17 +795,17 @@ void call_triangular_assignment_loop(DstXprType& dst, const SrcXprType& src, con
if((dst.rows()!=dstRows) || (dst.cols()!=dstCols))
dst.resize(dstRows, dstCols);
DstEvaluatorType dstEvaluator(dst);
typedef triangular_dense_assignment_kernel< Mode&(Lower|Upper),Mode&(UnitDiag|ZeroDiag|SelfAdjoint),SetOpposite,
DstEvaluatorType,SrcEvaluatorType,Functor> Kernel;
Kernel kernel(dstEvaluator, srcEvaluator, func, dst.const_cast_derived());
enum {
unroll = DstXprType::SizeAtCompileTime != Dynamic
&& SrcEvaluatorType::CoeffReadCost < HugeCost
&& DstXprType::SizeAtCompileTime * (DstEvaluatorType::CoeffReadCost+SrcEvaluatorType::CoeffReadCost) / 2 <= EIGEN_UNROLLING_LIMIT
};
triangular_assignment_loop<Kernel, Mode, unroll ? int(DstXprType::SizeAtCompileTime) : Dynamic, SetOpposite>::run(kernel);
}
@ -827,8 +827,8 @@ struct Assignment<DstXprType, SrcXprType, Functor, Triangular2Triangular>
EIGEN_DEVICE_FUNC static void run(DstXprType &dst, const SrcXprType &src, const Functor &func)
{
eigen_assert(int(DstXprType::Mode) == int(SrcXprType::Mode));
call_triangular_assignment_loop<DstXprType::Mode, false>(dst, src, func);
call_triangular_assignment_loop<DstXprType::Mode, false>(dst, src, func);
}
};
@ -837,7 +837,7 @@ struct Assignment<DstXprType, SrcXprType, Functor, Triangular2Dense>
{
EIGEN_DEVICE_FUNC static void run(DstXprType &dst, const SrcXprType &src, const Functor &func)
{
call_triangular_assignment_loop<SrcXprType::Mode, (SrcXprType::Mode&SelfAdjoint)==0>(dst, src, func);
call_triangular_assignment_loop<SrcXprType::Mode, (SrcXprType::Mode&SelfAdjoint)==0>(dst, src, func);
}
};
@ -846,7 +846,7 @@ struct Assignment<DstXprType, SrcXprType, Functor, Dense2Triangular>
{
EIGEN_DEVICE_FUNC static void run(DstXprType &dst, const SrcXprType &src, const Functor &func)
{
call_triangular_assignment_loop<DstXprType::Mode, false>(dst, src, func);
call_triangular_assignment_loop<DstXprType::Mode, false>(dst, src, func);
}
};
@ -857,19 +857,19 @@ struct triangular_assignment_loop
// FIXME: this is not very clean, perhaps this information should be provided by the kernel?
typedef typename Kernel::DstEvaluatorType DstEvaluatorType;
typedef typename DstEvaluatorType::XprType DstXprType;
enum {
col = (UnrollCount-1) / DstXprType::RowsAtCompileTime,
row = (UnrollCount-1) % DstXprType::RowsAtCompileTime
};
typedef typename Kernel::Scalar Scalar;
EIGEN_DEVICE_FUNC
static inline void run(Kernel &kernel)
{
triangular_assignment_loop<Kernel, Mode, UnrollCount-1, SetOpposite>::run(kernel);
if(row==col)
kernel.assignDiagonalCoeff(row);
else if( ((Mode&Lower) && row>col) || ((Mode&Upper) && row<col) )
@ -912,10 +912,10 @@ struct triangular_assignment_loop<Kernel, Mode, Dynamic, SetOpposite>
}
else
i = maxi;
if(i<kernel.rows()) // then i==j
kernel.assignDiagonalCoeff(i++);
if (((Mode&Upper) && SetOpposite) || (Mode&Lower))
{
for(; i < kernel.rows(); ++i)
@ -932,20 +932,20 @@ struct triangular_assignment_loop<Kernel, Mode, Dynamic, SetOpposite>
* If the matrix is triangular, the opposite part is set to zero. */
template<typename Derived>
template<typename DenseDerived>
void TriangularBase<Derived>::evalToLazy(MatrixBase<DenseDerived> &other) const
EIGEN_DEVICE_FUNC void TriangularBase<Derived>::evalToLazy(MatrixBase<DenseDerived> &other) const
{
other.derived().resize(this->rows(), this->cols());
internal::call_triangular_assignment_loop<Derived::Mode,(Derived::Mode&SelfAdjoint)==0 /* SetOpposite */>(other.derived(), derived().nestedExpression());
}
namespace internal {
// Triangular = Product
template< typename DstXprType, typename Lhs, typename Rhs, typename Scalar>
struct Assignment<DstXprType, Product<Lhs,Rhs,DefaultProduct>, internal::assign_op<Scalar,typename Product<Lhs,Rhs,DefaultProduct>::Scalar>, Dense2Triangular>
{
typedef Product<Lhs,Rhs,DefaultProduct> SrcXprType;
static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op<Scalar,typename SrcXprType::Scalar> &)
static EIGEN_DEVICE_FUNC void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op<Scalar,typename SrcXprType::Scalar> &)
{
Index dstRows = src.rows();
Index dstCols = src.cols();
@ -961,7 +961,7 @@ template< typename DstXprType, typename Lhs, typename Rhs, typename Scalar>
struct Assignment<DstXprType, Product<Lhs,Rhs,DefaultProduct>, internal::add_assign_op<Scalar,typename Product<Lhs,Rhs,DefaultProduct>::Scalar>, Dense2Triangular>
{
typedef Product<Lhs,Rhs,DefaultProduct> SrcXprType;
static void run(DstXprType &dst, const SrcXprType &src, const internal::add_assign_op<Scalar,typename SrcXprType::Scalar> &)
static EIGEN_DEVICE_FUNC void run(DstXprType &dst, const SrcXprType &src, const internal::add_assign_op<Scalar,typename SrcXprType::Scalar> &)
{
dst._assignProduct(src, 1, 1);
}
@ -972,7 +972,7 @@ template< typename DstXprType, typename Lhs, typename Rhs, typename Scalar>
struct Assignment<DstXprType, Product<Lhs,Rhs,DefaultProduct>, internal::sub_assign_op<Scalar,typename Product<Lhs,Rhs,DefaultProduct>::Scalar>, Dense2Triangular>
{
typedef Product<Lhs,Rhs,DefaultProduct> SrcXprType;
static void run(DstXprType &dst, const SrcXprType &src, const internal::sub_assign_op<Scalar,typename SrcXprType::Scalar> &)
static EIGEN_DEVICE_FUNC void run(DstXprType &dst, const SrcXprType &src, const internal::sub_assign_op<Scalar,typename SrcXprType::Scalar> &)
{
dst._assignProduct(src, -1, 1);
}

4
Eigen/src/Core/VectorwiseOp.h
View File

@ -670,7 +670,7 @@ template<typename ExpressionType, int Direction> class VectorwiseOp
* \sa rowwise(), class VectorwiseOp, \ref TutorialReductionsVisitorsBroadcasting
*/
template<typename Derived>
inline typename DenseBase<Derived>::ColwiseReturnType
EIGEN_DEVICE_FUNC inline typename DenseBase<Derived>::ColwiseReturnType
DenseBase<Derived>::colwise()
{
return ColwiseReturnType(derived());
@ -684,7 +684,7 @@ DenseBase<Derived>::colwise()
* \sa colwise(), class VectorwiseOp, \ref TutorialReductionsVisitorsBroadcasting
*/
template<typename Derived>
inline typename DenseBase<Derived>::RowwiseReturnType
EIGEN_DEVICE_FUNC inline typename DenseBase<Derived>::RowwiseReturnType
DenseBase<Derived>::rowwise()
{
return RowwiseReturnType(derived());

4
Eigen/src/Core/arch/CUDA/PacketMath.h
View File

@ -167,10 +167,10 @@ template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE double2 ploadu<double2>(const d
return make_double2(from[0], from[1]);
}
template<> EIGEN_STRONG_INLINE float4 ploaddup<float4>(const float* from) {
template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE float4 ploaddup<float4>(const float* from) {
return make_float4(from[0], from[0], from[1], from[1]);
}
template<> EIGEN_STRONG_INLINE double2 ploaddup<double2>(const double* from) {
template<> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE double2 ploaddup<double2>(const double* from) {
return make_double2(from[0], from[0]);
}

28
Eigen/src/Core/products/GeneralMatrixMatrixTriangular.h
View File

@ -10,7 +10,7 @@
#ifndef EIGEN_GENERAL_MATRIX_MATRIX_TRIANGULAR_H
#define EIGEN_GENERAL_MATRIX_MATRIX_TRIANGULAR_H
namespace Eigen {
namespace Eigen {
template<typename Scalar, typename Index, int StorageOrder, int UpLo, bool ConjLhs, bool ConjRhs>
struct selfadjoint_rank1_update;
@ -27,7 +27,7 @@ namespace internal {
// forward declarations (defined at the end of this file)
template<typename LhsScalar, typename RhsScalar, typename Index, int mr, int nr, bool ConjLhs, bool ConjRhs, int ResInnerStride, int UpLo>
struct tribb_kernel;
/* Optimized matrix-matrix product evaluating only one triangular half */
template <typename Index,
typename LhsScalar, int LhsStorageOrder, bool ConjugateLhs,
@ -164,7 +164,7 @@ struct tribb_kernel
if(UpLo==Upper)
gebp_kernel1(res.getSubMapper(0, j), blockA, actual_b, j, depth, actualBlockSize, alpha,
-1, -1, 0, 0);
// selfadjoint micro block
{
Index i = j;
@ -186,7 +186,7 @@ struct tribb_kernel
if(UpLo==Lower)
{
Index i = j+actualBlockSize;
gebp_kernel1(res.getSubMapper(i, j), blockA+depth*i, actual_b, size-i,
gebp_kernel1(res.getSubMapper(i, j), blockA+depth*i, actual_b, size-i,
depth, actualBlockSize, alpha, -1, -1, 0, 0);
}
}
@ -207,13 +207,13 @@ struct general_product_to_triangular_selector<MatrixType,ProductType,UpLo,true>
static void run(MatrixType& mat, const ProductType& prod, const typename MatrixType::Scalar& alpha, bool beta)
{
typedef typename MatrixType::Scalar Scalar;
typedef typename internal::remove_all<typename ProductType::LhsNested>::type Lhs;
typedef internal::blas_traits<Lhs> LhsBlasTraits;
typedef typename LhsBlasTraits::DirectLinearAccessType ActualLhs;
typedef typename internal::remove_all<ActualLhs>::type _ActualLhs;
typename internal::add_const_on_value_type<ActualLhs>::type actualLhs = LhsBlasTraits::extract(prod.lhs());
typedef typename internal::remove_all<typename ProductType::RhsNested>::type Rhs;
typedef internal::blas_traits<Rhs> RhsBlasTraits;
typedef typename RhsBlasTraits::DirectLinearAccessType ActualRhs;
@ -230,18 +230,18 @@ struct general_product_to_triangular_selector<MatrixType,ProductType,UpLo,true>
UseLhsDirectly = _ActualLhs::InnerStrideAtCompileTime==1,
UseRhsDirectly = _ActualRhs::InnerStrideAtCompileTime==1
};
internal::gemv_static_vector_if<Scalar,Lhs::SizeAtCompileTime,Lhs::MaxSizeAtCompileTime,!UseLhsDirectly> static_lhs;
ei_declare_aligned_stack_constructed_variable(Scalar, actualLhsPtr, actualLhs.size(),
(UseLhsDirectly ? const_cast<Scalar*>(actualLhs.data()) : static_lhs.data()));
if(!UseLhsDirectly) Map<typename _ActualLhs::PlainObject>(actualLhsPtr, actualLhs.size()) = actualLhs;
internal::gemv_static_vector_if<Scalar,Rhs::SizeAtCompileTime,Rhs::MaxSizeAtCompileTime,!UseRhsDirectly> static_rhs;
ei_declare_aligned_stack_constructed_variable(Scalar, actualRhsPtr, actualRhs.size(),
(UseRhsDirectly ? const_cast<Scalar*>(actualRhs.data()) : static_rhs.data()));
if(!UseRhsDirectly) Map<typename _ActualRhs::PlainObject>(actualRhsPtr, actualRhs.size()) = actualRhs;
selfadjoint_rank1_update<Scalar,Index,StorageOrder,UpLo,
LhsBlasTraits::NeedToConjugate && NumTraits<Scalar>::IsComplex,
RhsBlasTraits::NeedToConjugate && NumTraits<Scalar>::IsComplex>
@ -259,7 +259,7 @@ struct general_product_to_triangular_selector<MatrixType,ProductType,UpLo,false>
typedef typename LhsBlasTraits::DirectLinearAccessType ActualLhs;
typedef typename internal::remove_all<ActualLhs>::type _ActualLhs;
typename internal::add_const_on_value_type<ActualLhs>::type actualLhs = LhsBlasTraits::extract(prod.lhs());
typedef typename internal::remove_all<typename ProductType::RhsNested>::type Rhs;
typedef internal::blas_traits<Rhs> RhsBlasTraits;
typedef typename RhsBlasTraits::DirectLinearAccessType ActualRhs;
@ -302,13 +302,13 @@ struct general_product_to_triangular_selector<MatrixType,ProductType,UpLo,false>
template<typename _MatrixType, unsigned int _Mode>
template<typename ProductType>
TriangularView<_MatrixType,_Mode>& TriangularViewImpl<_MatrixType,_Mode,Dense>::_assignProduct(const ProductType& prod, const Scalar& alpha, bool beta)
EIGEN_DEVICE_FUNC TriangularView<_MatrixType,_Mode>& TriangularViewImpl<_MatrixType,_Mode,Dense>::_assignProduct(const ProductType& prod, const Scalar& alpha, bool beta)
{
EIGEN_STATIC_ASSERT((_Mode&UnitDiag)==0, WRITING_TO_TRIANGULAR_PART_WITH_UNIT_DIAGONAL_IS_NOT_SUPPORTED);
eigen_assert(derived().nestedExpression().rows() == prod.rows() && derived().cols() == prod.cols());
general_product_to_triangular_selector<_MatrixType, ProductType, _Mode, internal::traits<ProductType>::InnerSize==1>::run(derived().nestedExpression().const_cast_derived(), prod, alpha, beta);
return derived();
}

8
Eigen/src/Core/products/SelfadjointProduct.h
View File

@ -16,7 +16,7 @@
* It corresponds to the level 3 SYRK and level 2 SYR Blas routines.
**********************************************************************/
namespace Eigen {
namespace Eigen {
template<typename Scalar, typename Index, int UpLo, bool ConjLhs, bool ConjRhs>
@ -68,10 +68,10 @@ struct selfadjoint_product_selector<MatrixType,OtherType,UpLo,true>
ei_declare_aligned_stack_constructed_variable(Scalar, actualOtherPtr, other.size(),
(UseOtherDirectly ? const_cast<Scalar*>(actualOther.data()) : static_other.data()));
if(!UseOtherDirectly)
Map<typename _ActualOtherType::PlainObject>(actualOtherPtr, actualOther.size()) = actualOther;
selfadjoint_rank1_update<Scalar,Index,StorageOrder,UpLo,
OtherBlasTraits::NeedToConjugate && NumTraits<Scalar>::IsComplex,
(!OtherBlasTraits::NeedToConjugate) && NumTraits<Scalar>::IsComplex>
@ -120,7 +120,7 @@ struct selfadjoint_product_selector<MatrixType,OtherType,UpLo,false>
template<typename MatrixType, unsigned int UpLo>
template<typename DerivedU>
SelfAdjointView<MatrixType,UpLo>& SelfAdjointView<MatrixType,UpLo>
EIGEN_DEVICE_FUNC SelfAdjointView<MatrixType,UpLo>& SelfAdjointView<MatrixType,UpLo>
::rankUpdate(const MatrixBase<DerivedU>& u, const Scalar& alpha)
{
selfadjoint_product_selector<MatrixType,DerivedU,UpLo>::run(_expression().const_cast_derived(), u.derived(), alpha);

4
Eigen/src/Core/products/SelfadjointRank2Update.h
View File

@ -10,7 +10,7 @@
#ifndef EIGEN_SELFADJOINTRANK2UPTADE_H
#define EIGEN_SELFADJOINTRANK2UPTADE_H
namespace Eigen {
namespace Eigen {
namespace internal {
@ -57,7 +57,7 @@ template<bool Cond, typename T> struct conj_expr_if
template<typename MatrixType, unsigned int UpLo>
template<typename DerivedU, typename DerivedV>
SelfAdjointView<MatrixType,UpLo>& SelfAdjointView<MatrixType,UpLo>
EIGEN_DEVICE_FUNC SelfAdjointView<MatrixType,UpLo>& SelfAdjointView<MatrixType,UpLo>
::rankUpdate(const MatrixBase<DerivedU>& u, const MatrixBase<DerivedV>& v, const Scalar& alpha)
{
typedef internal::blas_traits<DerivedU> UBlasTraits;

12
Eigen/src/Eigenvalues/MatrixBaseEigenvalues.h
View File

@ -11,7 +11,7 @@
#ifndef EIGEN_MATRIXBASEEIGENVALUES_H
#define EIGEN_MATRIXBASEEIGENVALUES_H
namespace Eigen {
namespace Eigen {
namespace internal {
@ -42,13 +42,13 @@ struct eigenvalues_selector<Derived, false>
} // end namespace internal
/** \brief Computes the eigenvalues of a matrix
/** \brief Computes the eigenvalues of a matrix
* \returns Column vector containing the eigenvalues.
*
* \eigenvalues_module
* This function computes the eigenvalues with the help of the EigenSolver
* class (for real matrices) or the ComplexEigenSolver class (for complex
* matrices).
* matrices).
*
* The eigenvalues are repeated according to their algebraic multiplicity,
* so there are as many eigenvalues as rows in the matrix.
@ -83,8 +83,8 @@ MatrixBase<Derived>::eigenvalues() const
*
* \sa SelfAdjointEigenSolver::eigenvalues(), MatrixBase::eigenvalues()
*/
template<typename MatrixType, unsigned int UpLo>
inline typename SelfAdjointView<MatrixType, UpLo>::EigenvaluesReturnType
template<typename MatrixType, unsigned int UpLo>
EIGEN_DEVICE_FUNC inline typename SelfAdjointView<MatrixType, UpLo>::EigenvaluesReturnType
SelfAdjointView<MatrixType, UpLo>::eigenvalues() const
{
PlainObject thisAsMatrix(*this);
@ -147,7 +147,7 @@ MatrixBase<Derived>::operatorNorm() const
* \sa eigenvalues(), MatrixBase::operatorNorm()
*/
template<typename MatrixType, unsigned int UpLo>
inline typename SelfAdjointView<MatrixType, UpLo>::RealScalar
EIGEN_DEVICE_FUNC inline typename SelfAdjointView<MatrixType, UpLo>::RealScalar
SelfAdjointView<MatrixType, UpLo>::operatorNorm() const
{
return eigenvalues().cwiseAbs().maxCoeff();

8
Eigen/src/LU/FullPivLU.h
View File

@ -53,7 +53,7 @@ template<typename _MatrixType> struct traits<FullPivLU<_MatrixType> >
* Output: \verbinclude class_FullPivLU.out
*
* This class supports the \link InplaceDecomposition inplace decomposition \endlink mechanism.
*
*
* \sa MatrixBase::fullPivLu(), MatrixBase::determinant(), MatrixBase::inverse()
*/
template<typename _MatrixType> class FullPivLU
@ -744,7 +744,7 @@ struct image_retval<FullPivLU<_MatrixType> >
#ifndef EIGEN_PARSED_BY_DOXYGEN
template<typename _MatrixType>
template<typename RhsType, typename DstType>
void FullPivLU<_MatrixType>::_solve_impl(const RhsType &rhs, DstType &dst) const
EIGEN_DEVICE_FUNC void FullPivLU<_MatrixType>::_solve_impl(const RhsType &rhs, DstType &dst) const
{
/* The decomposition PAQ = LU can be rewritten as A = P^{-1} L U Q^{-1}.
* So we proceed as follows:
@ -792,7 +792,7 @@ void FullPivLU<_MatrixType>::_solve_impl(const RhsType &rhs, DstType &dst) const
template<typename _MatrixType>
template<bool Conjugate, typename RhsType, typename DstType>
void FullPivLU<_MatrixType>::_solve_impl_transposed(const RhsType &rhs, DstType &dst) const
EIGEN_DEVICE_FUNC void FullPivLU<_MatrixType>::_solve_impl_transposed(const RhsType &rhs, DstType &dst) const
{
/* The decomposition PAQ = LU can be rewritten as A = P^{-1} L U Q^{-1},
* and since permutations are real and unitary, we can write this
@ -864,7 +864,7 @@ struct Assignment<DstXprType, Inverse<FullPivLU<MatrixType> >, internal::assign_
{
typedef FullPivLU<MatrixType> LuType;
typedef Inverse<LuType> SrcXprType;
static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op<typename DstXprType::Scalar,typename MatrixType::Scalar> &)
static EIGEN_DEVICE_FUNC void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op<typename DstXprType::Scalar,typename MatrixType::Scalar> &)
{
dst = src.nestedExpression().solve(MatrixType::Identity(src.rows(), src.cols()));
}

20
Eigen/src/LU/InverseImpl.h
View File

@ -11,7 +11,7 @@
#ifndef EIGEN_INVERSE_IMPL_H
#define EIGEN_INVERSE_IMPL_H
namespace Eigen {
namespace Eigen {
namespace internal {
@ -72,7 +72,7 @@ struct compute_inverse_and_det_with_check<MatrixType, ResultType, 1>
****************************/
template<typename MatrixType, typename ResultType>
EIGEN_DEVICE_FUNC
EIGEN_DEVICE_FUNC
inline void compute_inverse_size2_helper(
const MatrixType& matrix, const typename ResultType::Scalar& invdet,
ResultType& result)
@ -122,7 +122,7 @@ struct compute_inverse_and_det_with_check<MatrixType, ResultType, 2>
****************************/
template<typename MatrixType, int i, int j>
EIGEN_DEVICE_FUNC
EIGEN_DEVICE_FUNC
inline typename MatrixType::Scalar cofactor_3x3(const MatrixType& m)
{
enum {
@ -200,7 +200,7 @@ struct compute_inverse_and_det_with_check<MatrixType, ResultType, 3>
****************************/
template<typename Derived>
EIGEN_DEVICE_FUNC
EIGEN_DEVICE_FUNC
inline const typename Derived::Scalar general_det3_helper
(const MatrixBase<Derived>& matrix, int i1, int i2, int i3, int j1, int j2, int j3)
{
@ -209,7 +209,7 @@ inline const typename Derived::Scalar general_det3_helper
}
template<typename MatrixType, int i, int j>
EIGEN_DEVICE_FUNC
EIGEN_DEVICE_FUNC
inline typename MatrixType::Scalar cofactor_4x4(const MatrixType& matrix)
{
enum {
@ -290,13 +290,13 @@ template<typename DstXprType, typename XprType>
struct Assignment<DstXprType, Inverse<XprType>, internal::assign_op<typename DstXprType::Scalar,typename XprType::Scalar>, Dense2Dense>
{
typedef Inverse<XprType> SrcXprType;
static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op<typename DstXprType::Scalar,typename XprType::Scalar> &)
static EIGEN_DEVICE_FUNC void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op<typename DstXprType::Scalar,typename XprType::Scalar> &)
{
Index dstRows = src.rows();
Index dstCols = src.cols();
if((dst.rows()!=dstRows) || (dst.cols()!=dstCols))
dst.resize(dstRows, dstCols);
const int Size = EIGEN_PLAIN_ENUM_MIN(XprType::ColsAtCompileTime,DstXprType::ColsAtCompileTime);
EIGEN_ONLY_USED_FOR_DEBUG(Size);
eigen_assert(( (Size<=1) || (Size>4) || (extract_data(src.nestedExpression())!=extract_data(dst)))
@ -304,14 +304,14 @@ struct Assignment<DstXprType, Inverse<XprType>, internal::assign_op<typename Dst
typedef typename internal::nested_eval<XprType,XprType::ColsAtCompileTime>::type ActualXprType;
typedef typename internal::remove_all<ActualXprType>::type ActualXprTypeCleanded;
ActualXprType actual_xpr(src.nestedExpression());
compute_inverse<ActualXprTypeCleanded, DstXprType>::run(actual_xpr, dst);
}
};
} // end namespace internal
/** \lu_module

4
Eigen/src/LU/PartialPivLU.h
View File

@ -69,7 +69,7 @@ struct enable_if_ref<Ref<T>,Derived> {
* The data of the LU decomposition can be directly accessed through the methods matrixLU(), permutationP().
*
* This class supports the \link InplaceDecomposition inplace decomposition \endlink mechanism.
*
*
* \sa MatrixBase::partialPivLu(), MatrixBase::determinant(), MatrixBase::inverse(), MatrixBase::computeInverse(), class FullPivLU
*/
template<typename _MatrixType> class PartialPivLU
@ -572,7 +572,7 @@ struct Assignment<DstXprType, Inverse<PartialPivLU<MatrixType> >, internal::assi
{
typedef PartialPivLU<MatrixType> LuType;
typedef Inverse<LuType> SrcXprType;
static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op<typename DstXprType::Scalar,typename LuType::Scalar> &)
static EIGEN_DEVICE_FUNC void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op<typename DstXprType::Scalar,typename LuType::Scalar> &)
{
dst = src.nestedExpression().solve(MatrixType::Identity(src.rows(), src.cols()));
}

6
Eigen/src/QR/ColPivHouseholderQR.h
View File

@ -42,7 +42,7 @@ template<typename _MatrixType> struct traits<ColPivHouseholderQR<_MatrixType> >
* numerical stability. It is slower than HouseholderQR, and faster than FullPivHouseholderQR.
*
* This class supports the \link InplaceDecomposition inplace decomposition \endlink mechanism.
*
*
* \sa MatrixBase::colPivHouseholderQr()
*/
template<typename _MatrixType> class ColPivHouseholderQR
@ -582,7 +582,7 @@ void ColPivHouseholderQR<MatrixType>::computeInPlace()
#ifndef EIGEN_PARSED_BY_DOXYGEN
template<typename _MatrixType>
template<typename RhsType, typename DstType>
void ColPivHouseholderQR<_MatrixType>::_solve_impl(const RhsType &rhs, DstType &dst) const
EIGEN_DEVICE_FUNC void ColPivHouseholderQR<_MatrixType>::_solve_impl(const RhsType &rhs, DstType &dst) const
{
eigen_assert(rhs.rows() == rows());
@ -618,7 +618,7 @@ struct Assignment<DstXprType, Inverse<ColPivHouseholderQR<MatrixType> >, interna
{
typedef ColPivHouseholderQR<MatrixType> QrType;
typedef Inverse<QrType> SrcXprType;
static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op<typename DstXprType::Scalar,typename QrType::Scalar> &)
static EIGEN_DEVICE_FUNC void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op<typename DstXprType::Scalar,typename QrType::Scalar> &)
{
dst = src.nestedExpression().solve(MatrixType::Identity(src.rows(), src.cols()));
}

6
Eigen/src/QR/CompleteOrthogonalDecomposition.h
View File

@ -41,7 +41,7 @@ struct traits<CompleteOrthogonalDecomposition<_MatrixType> >
* size rank-by-rank. \b A may be rank deficient.
*
* This class supports the \link InplaceDecomposition inplace decomposition \endlink mechanism.
*
*
* \sa MatrixBase::completeOrthogonalDecomposition()
*/
template <typename _MatrixType>
@ -489,7 +489,7 @@ void CompleteOrthogonalDecomposition<MatrixType>::applyZAdjointOnTheLeftInPlace(
#ifndef EIGEN_PARSED_BY_DOXYGEN
template <typename _MatrixType>
template <typename RhsType, typename DstType>
void CompleteOrthogonalDecomposition<_MatrixType>::_solve_impl(
EIGEN_DEVICE_FUNC void CompleteOrthogonalDecomposition<_MatrixType>::_solve_impl(
const RhsType& rhs, DstType& dst) const {
eigen_assert(rhs.rows() == this->rows());
@ -532,7 +532,7 @@ struct Assignment<DstXprType, Inverse<CompleteOrthogonalDecomposition<MatrixType
{
typedef CompleteOrthogonalDecomposition<MatrixType> CodType;
typedef Inverse<CodType> SrcXprType;
static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op<typename DstXprType::Scalar,typename CodType::Scalar> &)
static EIGEN_DEVICE_FUNC void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op<typename DstXprType::Scalar,typename CodType::Scalar> &)
{
dst = src.nestedExpression().solve(MatrixType::Identity(src.rows(), src.rows()));
}

34
Eigen/src/QR/FullPivHouseholderQR.h
View File

@ -11,7 +11,7 @@
#ifndef EIGEN_FULLPIVOTINGHOUSEHOLDERQR_H
#define EIGEN_FULLPIVOTINGHOUSEHOLDERQR_H
namespace Eigen {
namespace Eigen {
namespace internal {
@ -40,18 +40,18 @@ struct traits<FullPivHouseholderQRMatrixQReturnType<MatrixType> >
* \tparam _MatrixType the type of the matrix of which we are computing the QR decomposition
*
* This class performs a rank-revealing QR decomposition of a matrix \b A into matrices \b P, \b P', \b Q and \b R
* such that
* such that
* \f[
* \mathbf{P} \, \mathbf{A} \, \mathbf{P}' = \mathbf{Q} \, \mathbf{R}
* \f]
* by using Householder transformations. Here, \b P and \b P' are permutation matrices, \b Q a unitary matrix
* by using Householder transformations. Here, \b P and \b P' are permutation matrices, \b Q a unitary matrix
* and \b R an upper triangular matrix.
*
* This decomposition performs a very prudent full pivoting in order to be rank-revealing and achieve optimal
* numerical stability. The trade-off is that it is slower than HouseholderQR and ColPivHouseholderQR.
*
* This class supports the \link InplaceDecomposition inplace decomposition \endlink mechanism.
*
*
* \sa MatrixBase::fullPivHouseholderQr()
*/
template<typename _MatrixType> class FullPivHouseholderQR
@ -114,12 +114,12 @@ template<typename _MatrixType> class FullPivHouseholderQR
*
* This constructor computes the QR factorization of the matrix \a matrix by calling
* the method compute(). It is a short cut for:
*
*
* \code
* FullPivHouseholderQR<MatrixType> qr(matrix.rows(), matrix.cols());
* qr.compute(matrix);
* \endcode
*
*
* \sa compute()
*/
template<typename InputType>
@ -317,9 +317,9 @@ template<typename _MatrixType> class FullPivHouseholderQR
inline Index rows() const { return m_qr.rows(); }
inline Index cols() const { return m_qr.cols(); }
/** \returns a const reference to the vector of Householder coefficients used to represent the factor \c Q.
*
*
* For advanced uses only.
*/
const HCoeffsType& hCoeffs() const { return m_hCoeffs; }
@ -392,7 +392,7 @@ template<typename _MatrixType> class FullPivHouseholderQR
* diagonal coefficient of U.
*/
RealScalar maxPivot() const { return m_maxpivot; }
#ifndef EIGEN_PARSED_BY_DOXYGEN
template<typename RhsType, typename DstType>
EIGEN_DEVICE_FUNC
@ -400,14 +400,14 @@ template<typename _MatrixType> class FullPivHouseholderQR
#endif
protected:
static void check_template_parameters()
{
EIGEN_STATIC_ASSERT_NON_INTEGER(Scalar);
}
void computeInPlace();
MatrixType m_qr;
HCoeffsType m_hCoeffs;
IntDiagSizeVectorType m_rows_transpositions;
@ -463,7 +463,7 @@ void FullPivHouseholderQR<MatrixType>::computeInPlace()
Index cols = m_qr.cols();
Index size = (std::min)(rows,cols);
m_hCoeffs.resize(size);
m_temp.resize(cols);
@ -539,7 +539,7 @@ void FullPivHouseholderQR<MatrixType>::computeInPlace()
#ifndef EIGEN_PARSED_BY_DOXYGEN
template<typename _MatrixType>
template<typename RhsType, typename DstType>
void FullPivHouseholderQR<_MatrixType>::_solve_impl(const RhsType &rhs, DstType &dst) const
EIGEN_DEVICE_FUNC void FullPivHouseholderQR<_MatrixType>::_solve_impl(const RhsType &rhs, DstType &dst) const
{
eigen_assert(rhs.rows() == rows());
const Index l_rank = rank();
@ -574,14 +574,14 @@ void FullPivHouseholderQR<_MatrixType>::_solve_impl(const RhsType &rhs, DstType
#endif
namespace internal {
template<typename DstXprType, typename MatrixType>
struct Assignment<DstXprType, Inverse<FullPivHouseholderQR<MatrixType> >, internal::assign_op<typename DstXprType::Scalar,typename FullPivHouseholderQR<MatrixType>::Scalar>, Dense2Dense>
{
typedef FullPivHouseholderQR<MatrixType> QrType;
typedef Inverse<QrType> SrcXprType;
static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op<typename DstXprType::Scalar,typename QrType::Scalar> &)
{
static EIGEN_DEVICE_FUNC void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op<typename DstXprType::Scalar,typename QrType::Scalar> &)
{
dst = src.nestedExpression().solve(MatrixType::Identity(src.rows(), src.cols()));
}
};

22
Eigen/src/QR/HouseholderQR.h
View File

@ -12,7 +12,7 @@
#ifndef EIGEN_QR_H
#define EIGEN_QR_H
namespace Eigen {
namespace Eigen {
/** \ingroup QR_Module
*
@ -24,7 +24,7 @@ namespace Eigen {
* \tparam _MatrixType the type of the matrix of which we are computing the QR decomposition
*
* This class performs a QR decomposition of a matrix \b A into matrices \b Q and \b R
* such that
* such that
* \f[
* \mathbf{A} = \mathbf{Q} \, \mathbf{R}
* \f]
@ -85,12 +85,12 @@ template<typename _MatrixType> class HouseholderQR
*
* This constructor computes the QR factorization of the matrix \a matrix by calling
* the method compute(). It is a short cut for:
*
*
* \code
* HouseholderQR<MatrixType> qr(matrix.rows(), matrix.cols());
* qr.compute(matrix);
* \endcode
*
*
* \sa compute()
*/
template<typename InputType>
@ -204,13 +204,13 @@ template<typename _MatrixType> class HouseholderQR
inline Index rows() const { return m_qr.rows(); }
inline Index cols() const { return m_qr.cols(); }
/** \returns a const reference to the vector of Householder coefficients used to represent the factor \c Q.
*
*
* For advanced uses only.
*/
const HCoeffsType& hCoeffs() const { return m_hCoeffs; }
#ifndef EIGEN_PARSED_BY_DOXYGEN
template<typename RhsType, typename DstType>
EIGEN_DEVICE_FUNC
@ -218,14 +218,14 @@ template<typename _MatrixType> class HouseholderQR
#endif
protected:
static void check_template_parameters()
{
EIGEN_STATIC_ASSERT_NON_INTEGER(Scalar);
}
void computeInPlace();
MatrixType m_qr;
HCoeffsType m_hCoeffs;
RowVectorType m_temp;
@ -347,7 +347,7 @@ struct householder_qr_inplace_blocked
#ifndef EIGEN_PARSED_BY_DOXYGEN
template<typename _MatrixType>
template<typename RhsType, typename DstType>
void HouseholderQR<_MatrixType>::_solve_impl(const RhsType &rhs, DstType &dst) const
EIGEN_DEVICE_FUNC void HouseholderQR<_MatrixType>::_solve_impl(const RhsType &rhs, DstType &dst) const
{
const Index rank = (std::min)(rows(), cols());
eigen_assert(rhs.rows() == rows());
@ -379,7 +379,7 @@ template<typename MatrixType>
void HouseholderQR<MatrixType>::computeInPlace()
{
check_template_parameters();
Index rows = m_qr.rows();
Index cols = m_qr.cols();
Index size = (std::min)(rows,cols);

20
Eigen/src/SVD/SVDBase.h
View File

@ -34,12 +34,12 @@ namespace Eigen {
*
* Singular values are always sorted in decreasing order.
*
*
*
* You can ask for only \em thin \a U or \a V to be computed, meaning the following. In case of a rectangular n-by-p matrix, letting \a m be the
* smaller value among \a n and \a p, there are only \a m singular vectors; the remaining columns of \a U and \a V do not correspond to actual
* singular vectors. Asking for \em thin \a U or \a V means asking for only their \a m first columns to be formed. So \a U is then a n-by-m matrix,
* and \a V is then a p-by-m matrix. Notice that thin \a U and \a V are all you need for (least squares) solving.
*
*
* If the input matrix has inf or nan coefficients, the result of the computation is undefined, but the computation is guaranteed to
* terminate in finite (and reasonable) time.
* \sa class BDCSVD, class JacobiSVD
@ -67,7 +67,7 @@ public:
typedef Matrix<Scalar, RowsAtCompileTime, RowsAtCompileTime, MatrixOptions, MaxRowsAtCompileTime, MaxRowsAtCompileTime> MatrixUType;
typedef Matrix<Scalar, ColsAtCompileTime, ColsAtCompileTime, MatrixOptions, MaxColsAtCompileTime, MaxColsAtCompileTime> MatrixVType;
typedef typename internal::plain_diag_type<MatrixType, RealScalar>::type SingularValuesType;
Derived& derived() { return *static_cast<Derived*>(this); }
const Derived& derived() const { return *static_cast<const Derived*>(this); }
@ -120,7 +120,7 @@ public:
eigen_assert(m_isInitialized && "SVD is not initialized.");
return m_nonzeroSingularValues;
}
/** \returns the rank of the matrix of which \c *this is the SVD.
*
* \note This method has to determine which singular values should be considered nonzero.
@ -137,7 +137,7 @@ public:
while(i>=0 && m_singularValues.coeff(i) < premultiplied_threshold) --i;
return i+1;
}
/** Allows to prescribe a threshold to be used by certain methods, such as rank() and solve(),
* which need to determine when singular values are to be considered nonzero.
* This is not used for the SVD decomposition itself.
@ -193,7 +193,7 @@ public:
inline Index rows() const { return m_rows; }
inline Index cols() const { return m_cols; }
/** \returns a (least squares) solution of \f$ A x = b \f$ using the current SVD decomposition of A.
*
* \param b the right-hand-side of the equation to solve.
@ -211,7 +211,7 @@ public:
eigen_assert(computeU() && computeV() && "SVD::solve() requires both unitaries U and V to be computed (thin unitaries suffice).");
return Solve<Derived, Rhs>(derived(), b.derived());
}
#ifndef EIGEN_PARSED_BY_DOXYGEN
template<typename RhsType, typename DstType>
EIGEN_DEVICE_FUNC
@ -219,12 +219,12 @@ public:
#endif
protected:
static void check_template_parameters()
{
EIGEN_STATIC_ASSERT_NON_INTEGER(Scalar);
}
// return true if already allocated
bool allocate(Index rows, Index cols, unsigned int computationOptions) ;
@ -258,7 +258,7 @@ protected:
#ifndef EIGEN_PARSED_BY_DOXYGEN
template<typename Derived>
template<typename RhsType, typename DstType>
void SVDBase<Derived>::_solve_impl(const RhsType &rhs, DstType &dst) const
EIGEN_DEVICE_FUNC void SVDBase<Derived>::_solve_impl(const RhsType &rhs, DstType &dst) const
{
eigen_assert(rhs.rows() == rows());

24
Eigen/src/SparseCore/SparseAssign.h
View File

@ -10,9 +10,9 @@
#ifndef EIGEN_SPARSEASSIGN_H
#define EIGEN_SPARSEASSIGN_H
namespace Eigen {
namespace Eigen {
template<typename Derived>
template<typename Derived>
template<typename OtherDerived>
Derived& SparseMatrixBase<Derived>::operator=(const EigenBase<OtherDerived> &other)
{
@ -104,7 +104,7 @@ void assign_sparse_to_sparse(DstXprType &dst, const SrcXprType &src)
enum { Flip = (DstEvaluatorType::Flags & RowMajorBit) != (SrcEvaluatorType::Flags & RowMajorBit) };
DstXprType temp(src.rows(), src.cols());
temp.reserve((std::max)(src.rows(),src.cols())*2);
@ -127,7 +127,7 @@ void assign_sparse_to_sparse(DstXprType &dst, const SrcXprType &src)
template< typename DstXprType, typename SrcXprType, typename Functor>
struct Assignment<DstXprType, SrcXprType, Functor, Sparse2Sparse>
{
static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op<typename DstXprType::Scalar,typename SrcXprType::Scalar> &/*func*/)
static EIGEN_DEVICE_FUNC void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op<typename DstXprType::Scalar,typename SrcXprType::Scalar> &/*func*/)
{
assign_sparse_to_sparse(dst.derived(), src.derived());
}
@ -137,15 +137,15 @@ struct Assignment<DstXprType, SrcXprType, Functor, Sparse2Sparse>
template< typename DstXprType, typename SrcXprType, typename Functor>
struct Assignment<DstXprType, SrcXprType, Functor, Sparse2Dense>
{
static void run(DstXprType &dst, const SrcXprType &src, const Functor &func)
static EIGEN_DEVICE_FUNC void run(DstXprType &dst, const SrcXprType &src, const Functor &func)
{
if(internal::is_same<Functor,internal::assign_op<typename DstXprType::Scalar,typename SrcXprType::Scalar> >::value)
dst.setZero();
internal::evaluator<SrcXprType> srcEval(src);
resize_if_allowed(dst, src, func);
internal::evaluator<DstXprType> dstEval(dst);
const Index outerEvaluationSize = (internal::evaluator<SrcXprType>::Flags&RowMajorBit) ? src.rows() : src.cols();
for (Index j=0; j<outerEvaluationSize; ++j)
for (typename internal::evaluator<SrcXprType>::InnerIterator i(srcEval,j); i; ++i)
@ -159,7 +159,7 @@ template<typename DstXprType, typename DecType, typename RhsType, typename Scala
struct Assignment<DstXprType, Solve<DecType,RhsType>, internal::assign_op<Scalar,Scalar>, Sparse2Sparse>
{
typedef Solve<DecType,RhsType> SrcXprType;
static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op<Scalar,Scalar> &)
static EIGEN_DEVICE_FUNC void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op<Scalar,Scalar> &)
{
Index dstRows = src.rows();
Index dstCols = src.cols();
@ -182,7 +182,7 @@ struct Assignment<DstXprType, SrcXprType, Functor, Diagonal2Sparse>
typedef Array<StorageIndex,Dynamic,1> ArrayXI;
typedef Array<Scalar,Dynamic,1> ArrayXS;
template<int Options>
static void run(SparseMatrix<Scalar,Options,StorageIndex> &dst, const SrcXprType &src, const internal::assign_op<typename DstXprType::Scalar,typename SrcXprType::Scalar> &/*func*/)
static EIGEN_DEVICE_FUNC void run(SparseMatrix<Scalar,Options,StorageIndex> &dst, const SrcXprType &src, const internal::assign_op<typename DstXprType::Scalar,typename SrcXprType::Scalar> &/*func*/)
{
Index dstRows = src.rows();
Index dstCols = src.cols();
@ -196,16 +196,16 @@ struct Assignment<DstXprType, SrcXprType, Functor, Diagonal2Sparse>
Map<ArrayXI>(dst.outerIndexPtr(), size+1).setLinSpaced(0,StorageIndex(size));
Map<ArrayXS>(dst.valuePtr(), size) = src.diagonal();
}
template<typename DstDerived>
static void run(SparseMatrixBase<DstDerived> &dst, const SrcXprType &src, const internal::assign_op<typename DstXprType::Scalar,typename SrcXprType::Scalar> &/*func*/)
{
dst.diagonal() = src.diagonal();
}
static void run(DstXprType &dst, const SrcXprType &src, const internal::add_assign_op<typename DstXprType::Scalar,typename SrcXprType::Scalar> &/*func*/)
{ dst.diagonal() += src.diagonal(); }
static void run(DstXprType &dst, const SrcXprType &src, const internal::sub_assign_op<typename DstXprType::Scalar,typename SrcXprType::Scalar> &/*func*/)
{ dst.diagonal() -= src.diagonal(); }
};

10
Eigen/src/SparseCore/SparseProduct.h
View File

@ -10,7 +10,7 @@
#ifndef EIGEN_SPARSEPRODUCT_H
#define EIGEN_SPARSEPRODUCT_H
namespace Eigen {
namespace Eigen {
/** \returns an expression of the product of two sparse matrices.
* By default a conservative product preserving the symbolic non zeros is performed.
@ -102,13 +102,13 @@ template< typename DstXprType, typename Lhs, typename Rhs>
struct Assignment<DstXprType, Product<Lhs,Rhs,AliasFreeProduct>, internal::assign_op<typename DstXprType::Scalar,typename Product<Lhs,Rhs,AliasFreeProduct>::Scalar>, Sparse2Dense>
{
typedef Product<Lhs,Rhs,AliasFreeProduct> SrcXprType;
static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op<typename DstXprType::Scalar,typename SrcXprType::Scalar> &)
static EIGEN_DEVICE_FUNC void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op<typename DstXprType::Scalar,typename SrcXprType::Scalar> &)
{
Index dstRows = src.rows();
Index dstCols = src.cols();
if((dst.rows()!=dstRows) || (dst.cols()!=dstCols))
dst.resize(dstRows, dstCols);
generic_product_impl<Lhs, Rhs>::evalTo(dst,src.lhs(),src.rhs());
}
};
@ -118,7 +118,7 @@ template< typename DstXprType, typename Lhs, typename Rhs>
struct Assignment<DstXprType, Product<Lhs,Rhs,AliasFreeProduct>, internal::add_assign_op<typename DstXprType::Scalar,typename Product<Lhs,Rhs,AliasFreeProduct>::Scalar>, Sparse2Dense>
{
typedef Product<Lhs,Rhs,AliasFreeProduct> SrcXprType;
static void run(DstXprType &dst, const SrcXprType &src, const internal::add_assign_op<typename DstXprType::Scalar,typename SrcXprType::Scalar> &)
static EIGEN_DEVICE_FUNC void run(DstXprType &dst, const SrcXprType &src, const internal::add_assign_op<typename DstXprType::Scalar,typename SrcXprType::Scalar> &)
{
generic_product_impl<Lhs, Rhs>::addTo(dst,src.lhs(),src.rhs());
}
@ -129,7 +129,7 @@ template< typename DstXprType, typename Lhs, typename Rhs>
struct Assignment<DstXprType, Product<Lhs,Rhs,AliasFreeProduct>, internal::sub_assign_op<typename DstXprType::Scalar,typename Product<Lhs,Rhs,AliasFreeProduct>::Scalar>, Sparse2Dense>
{
typedef Product<Lhs,Rhs,AliasFreeProduct> SrcXprType;
static void run(DstXprType &dst, const SrcXprType &src, const internal::sub_assign_op<typename DstXprType::Scalar,typename SrcXprType::Scalar> &)
static EIGEN_DEVICE_FUNC void run(DstXprType &dst, const SrcXprType &src, const internal::sub_assign_op<typename DstXprType::Scalar,typename SrcXprType::Scalar> &)
{
generic_product_impl<Lhs, Rhs>::subTo(dst,src.lhs(),src.rhs());
}

78
Eigen/src/SparseCore/SparseSelfAdjointView.h
View File

@ -10,8 +10,8 @@
#ifndef EIGEN_SPARSE_SELFADJOINTVIEW_H
#define EIGEN_SPARSE_SELFADJOINTVIEW_H
namespace Eigen {
namespace Eigen {
/** \ingroup SparseCore_Module
* \class SparseSelfAdjointView
*
@ -27,7 +27,7 @@ namespace Eigen {
* \sa SparseMatrixBase::selfadjointView()
*/
namespace internal {
template<typename MatrixType, unsigned int Mode>
struct traits<SparseSelfAdjointView<MatrixType,Mode> > : traits<MatrixType> {
};
@ -44,7 +44,7 @@ template<typename MatrixType, unsigned int _Mode> class SparseSelfAdjointView
: public EigenBase<SparseSelfAdjointView<MatrixType,_Mode> >
{
public:
enum {
Mode = _Mode,
TransposeMode = ((Mode & Upper) ? Lower : 0) | ((Mode & Lower) ? Upper : 0),
@ -58,7 +58,7 @@ template<typename MatrixType, unsigned int _Mode> class SparseSelfAdjointView
typedef Matrix<StorageIndex,Dynamic,1> VectorI;
typedef typename internal::ref_selector<MatrixType>::non_const_type MatrixTypeNested;
typedef typename internal::remove_all<MatrixTypeNested>::type _MatrixTypeNested;
explicit inline SparseSelfAdjointView(MatrixType& matrix) : m_matrix(matrix)
{
eigen_assert(rows()==cols() && "SelfAdjointView is only for squared matrices");
@ -94,7 +94,7 @@ template<typename MatrixType, unsigned int _Mode> class SparseSelfAdjointView
{
return Product<OtherDerived, SparseSelfAdjointView>(lhs.derived(), rhs);
}
/** Efficient sparse self-adjoint matrix times dense vector/matrix product */
template<typename OtherDerived>
Product<SparseSelfAdjointView,OtherDerived>
@ -121,7 +121,7 @@ template<typename MatrixType, unsigned int _Mode> class SparseSelfAdjointView
*/
template<typename DerivedU>
SparseSelfAdjointView& rankUpdate(const SparseMatrixBase<DerivedU>& u, const Scalar& alpha = Scalar(1));
/** \returns an expression of P H P^-1 */
// TODO implement twists in a more evaluator friendly fashion
SparseSymmetricPermutationProduct<_MatrixTypeNested,Mode> twistedBy(const PermutationMatrix<Dynamic,Dynamic,StorageIndex>& perm) const
@ -148,7 +148,7 @@ template<typename MatrixType, unsigned int _Mode> class SparseSelfAdjointView
PermutationMatrix<Dynamic,Dynamic,StorageIndex> pnull;
return *this = src.twistedBy(pnull);
}
void resize(Index rows, Index cols)
{
EIGEN_ONLY_USED_FOR_DEBUG(rows);
@ -156,7 +156,7 @@ template<typename MatrixType, unsigned int _Mode> class SparseSelfAdjointView
eigen_assert(rows == this->rows() && cols == this->cols()
&& "SparseSelfadjointView::resize() does not actually allow to resize.");
}
protected:
MatrixTypeNested m_matrix;
@ -203,7 +203,7 @@ SparseSelfAdjointView<MatrixType,Mode>::rankUpdate(const SparseMatrixBase<Derive
}
namespace internal {
// TODO currently a selfadjoint expression has the form SelfAdjointView<.,.>
// in the future selfadjoint-ness should be defined by the expression traits
// such that Transpose<SelfAdjointView<.,.> > is valid. (currently TriangularBase::transpose() is overloaded to make it work)
@ -226,7 +226,7 @@ struct Assignment<DstXprType, SrcXprType, Functor, SparseSelfAdjoint2Sparse>
typedef internal::assign_op<typename DstXprType::Scalar,typename SrcXprType::Scalar> AssignOpType;
template<typename DestScalar,int StorageOrder>
static void run(SparseMatrix<DestScalar,StorageOrder,StorageIndex> &dst, const SrcXprType &src, const AssignOpType&/*func*/)
static EIGEN_DEVICE_FUNC void run(SparseMatrix<DestScalar,StorageOrder,StorageIndex> &dst, const SrcXprType &src, const AssignOpType&/*func*/)
{
internal::permute_symm_to_fullsymm<SrcXprType::Mode>(src.matrix(), dst);
}
@ -257,7 +257,7 @@ struct Assignment<DstXprType, SrcXprType, Functor, SparseSelfAdjoint2Sparse>
run(tmp, src, AssignOpType());
dst -= tmp;
}
template<typename DestScalar>
static void run(DynamicSparseMatrix<DestScalar,ColMajor,StorageIndex>& dst, const SrcXprType &src, const AssignOpType&/*func*/)
{
@ -280,13 +280,13 @@ template<int Mode, typename SparseLhsType, typename DenseRhsType, typename Dense
inline void sparse_selfadjoint_time_dense_product(const SparseLhsType& lhs, const DenseRhsType& rhs, DenseResType& res, const AlphaType& alpha)
{
EIGEN_ONLY_USED_FOR_DEBUG(alpha);
typedef typename internal::nested_eval<SparseLhsType,DenseRhsType::MaxColsAtCompileTime>::type SparseLhsTypeNested;
typedef typename internal::remove_all<SparseLhsTypeNested>::type SparseLhsTypeNestedCleaned;
typedef evaluator<SparseLhsTypeNestedCleaned> LhsEval;
typedef typename LhsEval::InnerIterator LhsIterator;
typedef typename SparseLhsType::Scalar LhsScalar;
enum {
LhsIsRowMajor = (LhsEval::Flags&RowMajorBit)==RowMajorBit,
ProcessFirstHalf =
@ -295,7 +295,7 @@ inline void sparse_selfadjoint_time_dense_product(const SparseLhsType& lhs, cons
|| ( (Mode&Lower) && LhsIsRowMajor),
ProcessSecondHalf = !ProcessFirstHalf
};
SparseLhsTypeNested lhs_nested(lhs);
LhsEval lhsEval(lhs_nested);
@ -349,7 +349,7 @@ struct generic_product_impl<LhsView, Rhs, SparseSelfAdjointShape, DenseShape, Pr
typedef typename nested_eval<Rhs,Dynamic>::type RhsNested;
LhsNested lhsNested(lhsView.matrix());
RhsNested rhsNested(rhs);
internal::sparse_selfadjoint_time_dense_product<LhsView::Mode>(lhsNested, rhsNested, dst, alpha);
}
};
@ -366,7 +366,7 @@ struct generic_product_impl<Lhs, RhsView, DenseShape, SparseSelfAdjointShape, Pr
typedef typename nested_eval<Rhs,Dynamic>::type RhsNested;
LhsNested lhsNested(lhs);
RhsNested rhsNested(rhsView.matrix());
// transpose everything
Transpose<Dest> dstT(dst);
internal::sparse_selfadjoint_time_dense_product<RhsView::TransposeMode>(rhsNested.transpose(), lhsNested.transpose(), dstT, alpha);
@ -390,7 +390,7 @@ struct product_evaluator<Product<LhsView, Rhs, DefaultProduct>, ProductTag, Spar
::new (static_cast<Base*>(this)) Base(m_result);
generic_product_impl<typename Rhs::PlainObject, Rhs, SparseShape, SparseShape, ProductTag>::evalTo(m_result, m_lhs, xpr.rhs());
}
protected:
typename Rhs::PlainObject m_lhs;
PlainObject m_result;
@ -410,7 +410,7 @@ struct product_evaluator<Product<Lhs, RhsView, DefaultProduct>, ProductTag, Spar
::new (static_cast<Base*>(this)) Base(m_result);
generic_product_impl<Lhs, typename Lhs::PlainObject, SparseShape, SparseShape, ProductTag>::evalTo(m_result, xpr.lhs(), m_rhs);
}
protected:
typename Lhs::PlainObject m_rhs;
PlainObject m_result;
@ -432,13 +432,13 @@ void permute_symm_to_fullsymm(const MatrixType& mat, SparseMatrix<typename Matri
typedef Matrix<StorageIndex,Dynamic,1> VectorI;
typedef evaluator<MatrixType> MatEval;
typedef typename evaluator<MatrixType>::InnerIterator MatIterator;
MatEval matEval(mat);
Dest& dest(_dest.derived());
enum {
StorageOrderMatch = int(Dest::IsRowMajor) == int(MatrixType::IsRowMajor)
};
Index size = mat.rows();
VectorI count;
count.resize(size);
@ -465,7 +465,7 @@ void permute_symm_to_fullsymm(const MatrixType& mat, SparseMatrix<typename Matri
}
}
Index nnz = count.sum();
// reserve space
dest.resizeNonZeros(nnz);
dest.outerIndexPtr()[0] = 0;
@ -473,7 +473,7 @@ void permute_symm_to_fullsymm(const MatrixType& mat, SparseMatrix<typename Matri
dest.outerIndexPtr()[j+1] = dest.outerIndexPtr()[j] + count[j];
for(Index j=0; j<size; ++j)
count[j] = dest.outerIndexPtr()[j];
// copy data
for(StorageIndex j = 0; j<size; ++j)
{
@ -482,10 +482,10 @@ void permute_symm_to_fullsymm(const MatrixType& mat, SparseMatrix<typename Matri
StorageIndex i = internal::convert_index<StorageIndex>(it.index());
Index r = it.row();
Index c = it.col();
StorageIndex jp = perm ? perm[j] : j;
StorageIndex ip = perm ? perm[i] : i;
if(Mode==int(Upper|Lower))
{
Index k = count[StorageOrderMatch ? jp : ip]++;
@ -531,7 +531,7 @@ void permute_symm_to_symm(const MatrixType& mat, SparseMatrix<typename MatrixTyp
};
MatEval matEval(mat);
Index size = mat.rows();
VectorI count(size);
count.setZero();
@ -544,7 +544,7 @@ void permute_symm_to_symm(const MatrixType& mat, SparseMatrix<typename MatrixTyp
StorageIndex i = it.index();
if((int(SrcMode)==int(Lower) && i<j) || (int(SrcMode)==int(Upper) && i>j))
continue;
StorageIndex ip = perm ? perm[i] : i;
count[int(DstMode)==int(Lower) ? (std::min)(ip,jp) : (std::max)(ip,jp)]++;
}
@ -555,22 +555,22 @@ void permute_symm_to_symm(const MatrixType& mat, SparseMatrix<typename MatrixTyp
dest.resizeNonZeros(dest.outerIndexPtr()[size]);
for(Index j=0; j<size; ++j)
count[j] = dest.outerIndexPtr()[j];
for(StorageIndex j = 0; j<size; ++j)
{
for(MatIterator it(matEval,j); it; ++it)
{
StorageIndex i = it.index();
if((int(SrcMode)==int(Lower) && i<j) || (int(SrcMode)==int(Upper) && i>j))
continue;
StorageIndex jp = perm ? perm[j] : j;
StorageIndex ip = perm? perm[i] : i;
Index k = count[int(DstMode)==int(Lower) ? (std::min)(ip,jp) : (std::max)(ip,jp)]++;
dest.innerIndexPtr()[k] = int(DstMode)==int(Lower) ? (std::max)(ip,jp) : (std::min)(ip,jp);
if(!StorageOrderMatch) std::swap(ip,jp);
if( ((int(DstMode)==int(Lower) && ip<jp) || (int(DstMode)==int(Upper) && ip>jp)))
dest.valuePtr()[k] = numext::conj(it.value());
@ -609,17 +609,17 @@ class SparseSymmetricPermutationProduct
typedef Matrix<StorageIndex,Dynamic,1> VectorI;
typedef typename MatrixType::Nested MatrixTypeNested;
typedef typename internal::remove_all<MatrixTypeNested>::type NestedExpression;
SparseSymmetricPermutationProduct(const MatrixType& mat, const Perm& perm)
: m_matrix(mat), m_perm(perm)
{}
inline Index rows() const { return m_matrix.rows(); }
inline Index cols() const { return m_matrix.cols(); }
const NestedExpression& matrix() const { return m_matrix; }
const Perm& perm() const { return m_perm; }
protected:
MatrixTypeNested m_matrix;
const Perm& m_perm;
@ -627,21 +627,21 @@ class SparseSymmetricPermutationProduct
};
namespace internal {
template<typename DstXprType, typename MatrixType, int Mode, typename Scalar>
struct Assignment<DstXprType, SparseSymmetricPermutationProduct<MatrixType,Mode>, internal::assign_op<Scalar,typename MatrixType::Scalar>, Sparse2Sparse>
{
typedef SparseSymmetricPermutationProduct<MatrixType,Mode> SrcXprType;
typedef typename DstXprType::StorageIndex DstIndex;
template<int Options>
static void run(SparseMatrix<Scalar,Options,DstIndex> &dst, const SrcXprType &src, const internal::assign_op<Scalar,typename MatrixType::Scalar> &)
static EIGEN_DEVICE_FUNC void run(SparseMatrix<Scalar,Options,DstIndex> &dst, const SrcXprType &src, const internal::assign_op<Scalar,typename MatrixType::Scalar> &)
{
// internal::permute_symm_to_fullsymm<Mode>(m_matrix,_dest,m_perm.indices().data());
SparseMatrix<Scalar,(Options&RowMajor)==RowMajor ? ColMajor : RowMajor, DstIndex> tmp;
internal::permute_symm_to_fullsymm<Mode>(src.matrix(),tmp,src.perm().indices().data());
dst = tmp;
}
template<typename DestType,unsigned int DestMode>
static void run(SparseSelfAdjointView<DestType,DestMode>& dst, const SrcXprType &src, const internal::assign_op<Scalar,typename MatrixType::Scalar> &)
{

196
Eigen/src/SparseQR/SparseQR.h
View File

@ -42,31 +42,31 @@ namespace internal {
* \ingroup SparseQR_Module
* \class SparseQR
* \brief Sparse left-looking rank-revealing QR factorization
*
* This class implements a left-looking rank-revealing QR decomposition
*
* This class implements a left-looking rank-revealing QR decomposition
* of sparse matrices. When a column has a norm less than a given tolerance
* it is implicitly permuted to the end. The QR factorization thus obtained is
* given by A*P = Q*R where R is upper triangular or trapezoidal.
*
* it is implicitly permuted to the end. The QR factorization thus obtained is
* given by A*P = Q*R where R is upper triangular or trapezoidal.
*
* P is the column permutation which is the product of the fill-reducing and the
* rank-revealing permutations. Use colsPermutation() to get it.
*
* Q is the orthogonal matrix represented as products of Householder reflectors.
*
* Q is the orthogonal matrix represented as products of Householder reflectors.
* Use matrixQ() to get an expression and matrixQ().adjoint() to get the adjoint.
* You can then apply it to a vector.
*
*
* R is the sparse triangular or trapezoidal matrix. The later occurs when A is rank-deficient.
* matrixR().topLeftCorner(rank(), rank()) always returns a triangular factor of full rank.
*
*
* \tparam _MatrixType The type of the sparse matrix A, must be a column-major SparseMatrix<>
* \tparam _OrderingType The fill-reducing ordering method. See the \link OrderingMethods_Module
* \tparam _OrderingType The fill-reducing ordering method. See the \link OrderingMethods_Module
* OrderingMethods \endlink module for the list of built-in and external ordering methods.
*
*
* \implsparsesolverconcept
*
* \warning The input sparse matrix A must be in compressed mode (see SparseMatrix::makeCompressed()).
* \warning For complex matrices matrixQ().transpose() will actually return the adjoint matrix.
*
*
*/
template<typename _MatrixType, typename _OrderingType>
class SparseQR : public SparseSolverBase<SparseQR<_MatrixType,_OrderingType> >
@ -90,26 +90,26 @@ class SparseQR : public SparseSolverBase<SparseQR<_MatrixType,_OrderingType> >
ColsAtCompileTime = MatrixType::ColsAtCompileTime,
MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime
};
public:
SparseQR () : m_analysisIsok(false), m_lastError(""), m_useDefaultThreshold(true),m_isQSorted(false),m_isEtreeOk(false)
{ }
/** Construct a QR factorization of the matrix \a mat.
*
*
* \warning The matrix \a mat must be in compressed mode (see SparseMatrix::makeCompressed()).
*
*
* \sa compute()
*/
explicit SparseQR(const MatrixType& mat) : m_analysisIsok(false), m_lastError(""), m_useDefaultThreshold(true),m_isQSorted(false),m_isEtreeOk(false)
{
compute(mat);
}
/** Computes the QR factorization of the sparse matrix \a mat.
*
*
* \warning The matrix \a mat must be in compressed mode (see SparseMatrix::makeCompressed()).
*
*
* \sa analyzePattern(), factorize()
*/
void compute(const MatrixType& mat)
@ -119,15 +119,15 @@ class SparseQR : public SparseSolverBase<SparseQR<_MatrixType,_OrderingType> >
}
void analyzePattern(const MatrixType& mat);
void factorize(const MatrixType& mat);
/** \returns the number of rows of the represented matrix.
/** \returns the number of rows of the represented matrix.
*/
inline Index rows() const { return m_pmat.rows(); }
/** \returns the number of columns of the represented matrix.
/** \returns the number of columns of the represented matrix.
*/
inline Index cols() const { return m_pmat.cols();}
/** \returns a const reference to the \b sparse upper triangular matrix R of the QR factorization.
* \warning The entries of the returned matrix are not sorted. This means that using it in algorithms
* expecting sorted entries will fail. This include random coefficient accesses (SpaseMatrix::coeff()),
@ -142,7 +142,7 @@ class SparseQR : public SparseSolverBase<SparseQR<_MatrixType,_OrderingType> >
* \endcode
*/
const QRMatrixType& matrixR() const { return m_R; }
/** \returns the number of non linearly dependent columns as determined by the pivoting threshold.
*
* \sa setPivotThreshold()
@ -150,9 +150,9 @@ class SparseQR : public SparseSolverBase<SparseQR<_MatrixType,_OrderingType> >
Index rank() const
{
eigen_assert(m_isInitialized && "The factorization should be called first, use compute()");
return m_nonzeropivots;
return m_nonzeropivots;
}
/** \returns an expression of the matrix Q as products of sparse Householder reflectors.
* The common usage of this function is to apply it to a dense matrix or vector
* \code
@ -171,23 +171,23 @@ class SparseQR : public SparseSolverBase<SparseQR<_MatrixType,_OrderingType> >
* reflectors are stored unsorted, two transpositions are needed to sort
* them before performing the product.
*/
SparseQRMatrixQReturnType<SparseQR> matrixQ() const
SparseQRMatrixQReturnType<SparseQR> matrixQ() const
{ return SparseQRMatrixQReturnType<SparseQR>(*this); }
/** \returns a const reference to the column permutation P that was applied to A such that A*P = Q*R
* It is the combination of the fill-in reducing permutation and numerical column pivoting.
*/
const PermutationType& colsPermutation() const
{
{
eigen_assert(m_isInitialized && "Decomposition is not initialized.");
return m_outputPerm_c;
}
/** \returns A string describing the type of error.
* This method is provided to ease debugging, not to handle errors.
*/
std::string lastErrorMessage() const { return m_lastError; }
/** \internal */
template<typename Rhs, typename Dest>
bool _solve_impl(const MatrixBase<Rhs> &B, MatrixBase<Dest> &dest) const
@ -196,21 +196,21 @@ class SparseQR : public SparseSolverBase<SparseQR<_MatrixType,_OrderingType> >
eigen_assert(this->rows() == B.rows() && "SparseQR::solve() : invalid number of rows in the right hand side matrix");
Index rank = this->rank();
// Compute Q^* * b;
typename Dest::PlainObject y, b;
y = this->matrixQ().adjoint() * B;
b = y;
// Solve with the triangular matrix R
y.resize((std::max<Index>)(cols(),y.rows()),y.cols());
y.topRows(rank) = this->matrixR().topLeftCorner(rank, rank).template triangularView<Upper>().solve(b.topRows(rank));
y.bottomRows(y.rows()-rank).setZero();
// Apply the column permutation
if (m_perm_c.size()) dest = colsPermutation() * y.topRows(cols());
else dest = y.topRows(cols());
m_info = Success;
return true;
}
@ -225,13 +225,13 @@ class SparseQR : public SparseSolverBase<SparseQR<_MatrixType,_OrderingType> >
m_useDefaultThreshold = false;
m_threshold = threshold;
}
/** \returns the solution X of \f$ A X = B \f$ using the current decomposition of A.
*
* \sa compute()
*/
template<typename Rhs>
inline const Solve<SparseQR, Rhs> solve(const MatrixBase<Rhs>& B) const
inline const Solve<SparseQR, Rhs> solve(const MatrixBase<Rhs>& B) const
{
eigen_assert(m_isInitialized && "The factorization should be called first, use compute()");
eigen_assert(this->rows() == B.rows() && "SparseQR::solve() : invalid number of rows in the right hand side matrix");
@ -244,14 +244,14 @@ class SparseQR : public SparseSolverBase<SparseQR<_MatrixType,_OrderingType> >
eigen_assert(this->rows() == B.rows() && "SparseQR::solve() : invalid number of rows in the right hand side matrix");
return Solve<SparseQR, Rhs>(*this, B.derived());
}
/** \brief Reports whether previous computation was successful.
*
* \returns \c Success if computation was successful,
* \c NumericalIssue if the QR factorization reports a numerical problem
* \c InvalidInput if the input matrix is invalid
*
* \sa iparm()
* \sa iparm()
*/
ComputationInfo info() const
{
@ -270,7 +270,7 @@ class SparseQR : public SparseSolverBase<SparseQR<_MatrixType,_OrderingType> >
this->m_isQSorted = true;
}
protected:
bool m_analysisIsok;
bool m_factorizationIsok;
@ -290,18 +290,18 @@ class SparseQR : public SparseSolverBase<SparseQR<_MatrixType,_OrderingType> >
IndexVector m_firstRowElt; // First element in each row
bool m_isQSorted; // whether Q is sorted or not
bool m_isEtreeOk; // whether the elimination tree match the initial input matrix
template <typename, typename > friend struct SparseQR_QProduct;
};
/** \brief Preprocessing step of a QR factorization
*
/** \brief Preprocessing step of a QR factorization
*
* \warning The matrix \a mat must be in compressed mode (see SparseMatrix::makeCompressed()).
*
*
* In this step, the fill-reducing permutation is computed and applied to the columns of A
* and the column elimination tree is computed as well. Only the sparsity pattern of \a mat is exploited.
*
*
* \note In this step it is assumed that there is no empty row in the matrix \a mat.
*/
template <typename MatrixType, typename OrderingType>
@ -311,26 +311,26 @@ void SparseQR<MatrixType,OrderingType>::analyzePattern(const MatrixType& mat)
// Copy to a column major matrix if the input is rowmajor
typename internal::conditional<MatrixType::IsRowMajor,QRMatrixType,const MatrixType&>::type matCpy(mat);
// Compute the column fill reducing ordering
OrderingType ord;
ord(matCpy, m_perm_c);
OrderingType ord;
ord(matCpy, m_perm_c);
Index n = mat.cols();
Index m = mat.rows();
Index diagSize = (std::min)(m,n);
if (!m_perm_c.size())
{
m_perm_c.resize(n);
m_perm_c.indices().setLinSpaced(n, 0,StorageIndex(n-1));
}
// Compute the column elimination tree of the permuted matrix
m_outputPerm_c = m_perm_c.inverse();
internal::coletree(matCpy, m_etree, m_firstRowElt, m_outputPerm_c.indices().data());
m_isEtreeOk = true;
m_R.resize(m, n);
m_Q.resize(m, diagSize);
// Allocate space for nonzero elements : rough estimation
m_R.reserve(2*mat.nonZeros()); //FIXME Get a more accurate estimation through symbolic factorization with the etree
m_Q.reserve(2*mat.nonZeros());
@ -339,17 +339,17 @@ void SparseQR<MatrixType,OrderingType>::analyzePattern(const MatrixType& mat)
}
/** \brief Performs the numerical QR factorization of the input matrix
*
*
* The function SparseQR::analyzePattern(const MatrixType&) must have been called beforehand with
* a matrix having the same sparsity pattern than \a mat.
*
*
* \param mat The sparse column-major matrix
*/
template <typename MatrixType, typename OrderingType>
void SparseQR<MatrixType,OrderingType>::factorize(const MatrixType& mat)
{
using std::abs;
eigen_assert(m_analysisIsok && "analyzePattern() should be called before this step");
StorageIndex m = StorageIndex(mat.rows());
StorageIndex n = StorageIndex(mat.cols());
@ -359,7 +359,7 @@ void SparseQR<MatrixType,OrderingType>::factorize(const MatrixType& mat)
Index nzcolR, nzcolQ; // Number of nonzero for the current column of R and Q
ScalarVector tval(m); // The dense vector used to compute the current column
RealScalar pivotThreshold = m_threshold;
m_R.setZero();
m_Q.setZero();
m_pmat = mat;
@ -371,12 +371,12 @@ void SparseQR<MatrixType,OrderingType>::factorize(const MatrixType& mat)
}
m_pmat.uncompress(); // To have the innerNonZeroPtr allocated
// Apply the fill-in reducing permutation lazily:
{
// If the input is row major, copy the original column indices,
// otherwise directly use the input matrix
//
//
IndexVector originalOuterIndicesCpy;
const StorageIndex *originalOuterIndices = mat.outerIndexPtr();
if(MatrixType::IsRowMajor)
@ -384,20 +384,20 @@ void SparseQR<MatrixType,OrderingType>::factorize(const MatrixType& mat)
originalOuterIndicesCpy = IndexVector::Map(m_pmat.outerIndexPtr(),n+1);
originalOuterIndices = originalOuterIndicesCpy.data();
}
for (int i = 0; i < n; i++)
{
Index p = m_perm_c.size() ? m_perm_c.indices()(i) : i;
m_pmat.outerIndexPtr()[p] = originalOuterIndices[i];
m_pmat.innerNonZeroPtr()[p] = originalOuterIndices[i+1] - originalOuterIndices[i];
m_pmat.outerIndexPtr()[p] = originalOuterIndices[i];
m_pmat.innerNonZeroPtr()[p] = originalOuterIndices[i+1] - originalOuterIndices[i];
}
}
/* Compute the default threshold as in MatLab, see:
* Tim Davis, "Algorithm 915, SuiteSparseQR: Multifrontal Multithreaded Rank-Revealing
* Sparse QR Factorization, ACM Trans. on Math. Soft. 38(1), 2011, Page 8:3
* Sparse QR Factorization, ACM Trans. on Math. Soft. 38(1), 2011, Page 8:3
*/
if(m_useDefaultThreshold)
if(m_useDefaultThreshold)
{
RealScalar max2Norm = 0.0;
for (int j = 0; j < n; j++) max2Norm = numext::maxi(max2Norm, m_pmat.col(j).norm());
@ -405,10 +405,10 @@ void SparseQR<MatrixType,OrderingType>::factorize(const MatrixType& mat)
max2Norm = RealScalar(1);
pivotThreshold = 20 * (m + n) * max2Norm * NumTraits<RealScalar>::epsilon();
}
// Initialize the numerical permutation
m_pivotperm.setIdentity(n);
StorageIndex nonzeroCol = 0; // Record the number of valid pivots
m_Q.startVec(0);
@ -421,8 +421,8 @@ void SparseQR<MatrixType,OrderingType>::factorize(const MatrixType& mat)
Qidx(0) = nonzeroCol;
nzcolR = 0; nzcolQ = 1;
bool found_diag = nonzeroCol>=m;
tval.setZero();
tval.setZero();
// Symbolic factorization: find the nonzero locations of the column k of the factors R and Q, i.e.,
// all the nodes (with indexes lower than rank) reachable through the column elimination tree (etree) rooted at node k.
// Note: if the diagonal entry does not exist, then its contribution must be explicitly added,
@ -432,7 +432,7 @@ void SparseQR<MatrixType,OrderingType>::factorize(const MatrixType& mat)
StorageIndex curIdx = nonzeroCol;
if(itp) curIdx = StorageIndex(itp.row());
if(curIdx == nonzeroCol) found_diag = true;
// Get the nonzeros indexes of the current column of R
StorageIndex st = m_firstRowElt(curIdx); // The traversal of the etree starts here
if (st < 0 )
@ -442,7 +442,7 @@ void SparseQR<MatrixType,OrderingType>::factorize(const MatrixType& mat)
return;
}
// Traverse the etree
// Traverse the etree
Index bi = nzcolR;
for (; mark(st) != col; st = m_etree(st))
{
@ -454,13 +454,13 @@ void SparseQR<MatrixType,OrderingType>::factorize(const MatrixType& mat)
// Reverse the list to get the topological ordering
Index nt = nzcolR-bi;
for(Index i = 0; i < nt/2; i++) std::swap(Ridx(bi+i), Ridx(nzcolR-i-1));
// Copy the current (curIdx,pcol) value of the input matrix
if(itp) tval(curIdx) = itp.value();
else tval(curIdx) = Scalar(0);
// Compute the pattern of Q(:,k)
if(curIdx > nonzeroCol && mark(curIdx) != col )
if(curIdx > nonzeroCol && mark(curIdx) != col )
{
Qidx(nzcolQ) = curIdx; // Add this row to the pattern of Q,
mark(curIdx) = col; // and mark it as visited
@ -472,15 +472,15 @@ void SparseQR<MatrixType,OrderingType>::factorize(const MatrixType& mat)
for (Index i = nzcolR-1; i >= 0; i--)
{
Index curIdx = Ridx(i);
// Apply the curIdx-th householder vector to the current column (temporarily stored into tval)
Scalar tdot(0);
// First compute q' * tval
tdot = m_Q.col(curIdx).dot(tval);
tdot *= m_hcoeffs(curIdx);
// Then update tval = tval - q * tau
// FIXME: tval -= tdot * m_Q.col(curIdx) should amount to the same (need to check/add support for efficient "dense ?= sparse")
for (typename QRMatrixType::InnerIterator itq(m_Q, curIdx); itq; ++itq)
@ -500,16 +500,16 @@ void SparseQR<MatrixType,OrderingType>::factorize(const MatrixType& mat)
}
}
} // End update current column
Scalar tau = RealScalar(0);
RealScalar beta = 0;
if(nonzeroCol < diagSize)
{
// Compute the Householder reflection that eliminate the current column
// FIXME this step should call the Householder module.
Scalar c0 = nzcolQ ? tval(Qidx(0)) : Scalar(0);
// First, the squared norm of Q((col+1):m, col)
RealScalar sqrNorm = 0.;
for (Index itq = 1; itq < nzcolQ; ++itq) sqrNorm += numext::abs2(tval(Qidx(itq)));
@ -528,7 +528,7 @@ void SparseQR<MatrixType,OrderingType>::factorize(const MatrixType& mat)
for (Index itq = 1; itq < nzcolQ; ++itq)
tval(Qidx(itq)) /= (c0 - beta);
tau = numext::conj((beta-c0) / beta);
}
}
@ -536,7 +536,7 @@ void SparseQR<MatrixType,OrderingType>::factorize(const MatrixType& mat)
for (Index i = nzcolR-1; i >= 0; i--)
{
Index curIdx = Ridx(i);
if(curIdx < nonzeroCol)
if(curIdx < nonzeroCol)
{
m_R.insertBackByOuterInnerUnordered(col, curIdx) = tval(curIdx);
tval(curIdx) = Scalar(0.);
@ -562,17 +562,17 @@ void SparseQR<MatrixType,OrderingType>::factorize(const MatrixType& mat)
else
{
// Zero pivot found: move implicitly this column to the end
for (Index j = nonzeroCol; j < n-1; j++)
for (Index j = nonzeroCol; j < n-1; j++)
std::swap(m_pivotperm.indices()(j), m_pivotperm.indices()[j+1]);
// Recompute the column elimination tree
internal::coletree(m_pmat, m_etree, m_firstRowElt, m_pivotperm.indices().data());
m_isEtreeOk = false;
}
}
m_hcoeffs.tail(diagSize-nonzeroCol).setZero();
// Finalize the column pointers of the sparse matrices R and Q
m_Q.finalize();
m_Q.makeCompressed();
@ -581,18 +581,18 @@ void SparseQR<MatrixType,OrderingType>::factorize(const MatrixType& mat)
m_isQSorted = false;
m_nonzeropivots = nonzeroCol;
if(nonzeroCol<n)
{
// Permute the triangular factor to put the 'dead' columns to the end
QRMatrixType tempR(m_R);
m_R = tempR * m_pivotperm;
// Update the column permutation
m_outputPerm_c = m_outputPerm_c * m_pivotperm;
}
m_isInitialized = true;
m_isInitialized = true;
m_factorizationIsok = true;
m_info = Success;
}
@ -602,12 +602,12 @@ struct SparseQR_QProduct : ReturnByValue<SparseQR_QProduct<SparseQRType, Derived
{
typedef typename SparseQRType::QRMatrixType MatrixType;
typedef typename SparseQRType::Scalar Scalar;
// Get the references
SparseQR_QProduct(const SparseQRType& qr, const Derived& other, bool transpose) :
// Get the references
SparseQR_QProduct(const SparseQRType& qr, const Derived& other, bool transpose) :
m_qr(qr),m_other(other),m_transpose(transpose) {}
inline Index rows() const { return m_qr.matrixQ().rows(); }
inline Index cols() const { return m_other.cols(); }
// Assign to a vector
template<typename DesType>
void evalTo(DesType& res) const
@ -651,7 +651,7 @@ struct SparseQR_QProduct : ReturnByValue<SparseQR_QProduct<SparseQRType, Derived
}
}
}
const SparseQRType& m_qr;
const Derived& m_other;
bool m_transpose; // TODO this actually means adjoint
@ -659,7 +659,7 @@ struct SparseQR_QProduct : ReturnByValue<SparseQR_QProduct<SparseQRType, Derived
template<typename SparseQRType>
struct SparseQRMatrixQReturnType : public EigenBase<SparseQRMatrixQReturnType<SparseQRType> >
{
{
typedef typename SparseQRType::Scalar Scalar;
typedef Matrix<Scalar,Dynamic,Dynamic> DenseMatrix;
enum {
@ -701,7 +701,7 @@ struct SparseQRMatrixQTransposeReturnType
};
namespace internal {
template<typename SparseQRType>
struct evaluator_traits<SparseQRMatrixQReturnType<SparseQRType> >
{
@ -716,7 +716,7 @@ struct Assignment<DstXprType, SparseQRMatrixQReturnType<SparseQRType>, internal:
typedef SparseQRMatrixQReturnType<SparseQRType> SrcXprType;
typedef typename DstXprType::Scalar Scalar;
typedef typename DstXprType::StorageIndex StorageIndex;
static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op<Scalar,Scalar> &/*func*/)
static EIGEN_DEVICE_FUNC void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op<Scalar,Scalar> &/*func*/)
{
typename DstXprType::PlainObject idMat(src.rows(), src.cols());
idMat.setIdentity();
@ -732,7 +732,7 @@ struct Assignment<DstXprType, SparseQRMatrixQReturnType<SparseQRType>, internal:
typedef SparseQRMatrixQReturnType<SparseQRType> SrcXprType;
typedef typename DstXprType::Scalar Scalar;
typedef typename DstXprType::StorageIndex StorageIndex;
static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op<Scalar,Scalar> &/*func*/)
static EIGEN_DEVICE_FUNC void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op<Scalar,Scalar> &/*func*/)
{
dst = src.m_qr.matrixQ() * DstXprType::Identity(src.m_qr.rows(), src.m_qr.rows());
}

13
cmake/EigenConfigureTesting.cmake
View File

@ -1,7 +1,7 @@
include(EigenTesting)
include(CheckCXXSourceCompiles)
# configure the "site" and "buildname"
# configure the "site" and "buildname"
ei_set_sitename()
# retrieve and store the build string
@ -11,6 +11,15 @@ add_custom_target(buildtests)
add_custom_target(check COMMAND "ctest")
add_dependencies(check buildtests)
# Convenience target for only building GPU tests.
add_custom_target(buildtests_gpu)
add_custom_target(check_gpu COMMAND "ctest" "--output-on-failure"
"--no-compress-output"
"--build-no-clean"
"-T" "test"
"-L" "gpu")
add_dependencies(check_gpu buildtests_gpu)
# check whether /bin/bash exists (disabled as not used anymore)
# find_file(EIGEN_BIN_BASH_EXISTS "/bin/bash" PATHS "/" NO_DEFAULT_PATH)
@ -50,7 +59,7 @@ if(CMAKE_COMPILER_IS_GNUCXX)
set(CTEST_CUSTOM_COVERAGE_EXCLUDE "/test/")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${COVERAGE_FLAGS}")
endif(EIGEN_COVERAGE_TESTING)
elseif(MSVC)
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /D_CRT_SECURE_NO_WARNINGS /D_SCL_SECURE_NO_WARNINGS")
endif(CMAKE_COMPILER_IS_GNUCXX)

9
cmake/EigenTesting.cmake
View File

@ -18,7 +18,9 @@ macro(ei_add_test_internal testname testname_with_suffix)
set(filename ${testname}.cpp)
endif()
set(is_gpu_test OFF)
if(EIGEN_ADD_TEST_FILENAME_EXTENSION STREQUAL cu)
set(is_gpu_test ON)
if(EIGEN_TEST_CUDA_CLANG)
set_source_files_properties(${filename} PROPERTIES LANGUAGE CXX)
if(CUDA_64_BIT_DEVICE_CODE)
@ -48,6 +50,9 @@ macro(ei_add_test_internal testname testname_with_suffix)
else()
add_dependencies(buildtests ${targetname})
endif()
if (is_gpu_test)
add_dependencies(buildtests_gpu ${targetname})
endif()
if(EIGEN_NO_ASSERTION_CHECKING)
ei_add_target_property(${targetname} COMPILE_FLAGS "-DEIGEN_NO_ASSERTION_CHECKING=1")
@ -98,6 +103,10 @@ macro(ei_add_test_internal testname testname_with_suffix)
endif()
add_test(${testname_with_suffix} "${targetname}")
if (is_gpu_test)
# Add gpu tag for testing only GPU tests.
set_property(TEST ${testname_with_suffix} APPEND PROPERTY LABELS "gpu")
endif()
# Specify target and test labels accoirding to EIGEN_CURRENT_SUBPROJECT
get_property(current_subproject GLOBAL PROPERTY EIGEN_CURRENT_SUBPROJECT)

31
test/CMakeLists.txt
View File

@ -362,14 +362,35 @@ if(EIGEN_TEST_CUDA)
find_package(CUDA 5.0)
if(CUDA_FOUND)
set(CUDA_PROPAGATE_HOST_FLAGS OFF)
if("${CMAKE_CXX_COMPILER_ID}" STREQUAL "Clang")
set(CUDA_NVCC_FLAGS "-ccbin ${CMAKE_C_COMPILER}" CACHE STRING "nvcc flags" FORCE)
if( (NOT EIGEN_TEST_CXX11) OR (CMAKE_VERSION VERSION_LESS 3.3))
string(APPEND EIGEN_CUDA_CXX11_FLAGS " -std=c++11")
endif()
if(EIGEN_TEST_CUDA_CLANG)
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11 --cuda-gpu-arch=sm_30")
string(APPEND CMAKE_CXX_FLAGS " --cuda-path=${CUDA_TOOLKIT_ROOT_DIR}")
foreach(GPU IN LISTS EIGEN_CUDA_COMPUTE_ARCH)
string(APPEND CMAKE_CXX_FLAGS " --cuda-gpu-arch=sm_${GPU}")
endforeach()
string(APPEND CMAKE_CXX_FLAGS " ${EIGEN_CUDA_CXX_FLAGS}")
else()
set(CUDA_PROPAGATE_HOST_FLAGS OFF)
set(NVCC_ARCH_FLAGS)
# Define an -arch=sm_<arch>, otherwise if GPU does not exactly match one of
# those in the arch list for -gencode, the kernels will fail to run with
# cudaErrorNoKernelImageForDevice
# This can happen with newer cards (e.g. sm_75) and compiling with older
# versions of nvcc (e.g. 9.2) that do not support their specific arch.
list(LENGTH EIGEN_CUDA_COMPUTE_ARCH EIGEN_CUDA_COMPUTE_ARCH_SIZE)
if(EIGEN_CUDA_COMPUTE_ARCH_SIZE)
list(GET EIGEN_CUDA_COMPUTE_ARCH 0 EIGEN_CUDA_COMPUTE_DEFAULT)
set(NVCC_ARCH_FLAGS " -arch=sm_${EIGEN_CUDA_COMPUTE_DEFAULT}")
endif()
foreach(ARCH IN LISTS EIGEN_CUDA_COMPUTE_ARCH)
string(APPEND NVCC_ARCH_FLAGS " -gencode arch=compute_${ARCH},code=sm_${ARCH}")
endforeach()
set(CUDA_NVCC_FLAGS "--expt-relaxed-constexpr -Xcudafe \"--display_error_number\" ${NVCC_ARCH_FLAGS} ${CUDA_NVCC_FLAGS} ${EIGEN_CUDA_CXX_FLAGS}")
cuda_include_directories("${CMAKE_CURRENT_BINARY_DIR}" "${CUDA_TOOLKIT_ROOT_DIR}/include")
endif()
cuda_include_directories(${CMAKE_CURRENT_BINARY_DIR})
set(EIGEN_ADD_TEST_FILENAME_EXTENSION "cu")
ei_add_test(cuda_basic)

14
test/cuda_common.h
View File

@ -37,26 +37,26 @@ void run_on_cuda(const Kernel& ker, int n, const Input& in, Output& out)
typename Output::Scalar* d_out;
std::ptrdiff_t in_bytes = in.size() * sizeof(typename Input::Scalar);
std::ptrdiff_t out_bytes = out.size() * sizeof(typename Output::Scalar);
cudaMalloc((void**)(&d_in), in_bytes);
cudaMalloc((void**)(&d_out), out_bytes);
cudaMemcpy(d_in, in.data(), in_bytes, cudaMemcpyHostToDevice);
cudaMemcpy(d_out, out.data(), out_bytes, cudaMemcpyHostToDevice);
// Simple and non-optimal 1D mapping assuming n is not too large
// That's only for unit testing!
dim3 Blocks(128);
dim3 Grids( (n+int(Blocks.x)-1)/int(Blocks.x) );
cudaThreadSynchronize();
cudaDeviceSynchronize();
run_on_cuda_meta_kernel<<<Grids,Blocks>>>(ker, n, d_in, d_out);
cudaThreadSynchronize();
cudaDeviceSynchronize();
// check inputs have not been modified
cudaMemcpy(const_cast<typename Input::Scalar*>(in.data()), d_in, in_bytes, cudaMemcpyDeviceToHost);
cudaMemcpy(out.data(), d_out, out_bytes, cudaMemcpyDeviceToHost);
cudaFree(d_in);
cudaFree(d_out);
}

12
unsupported/Eigen/CXX11/src/Tensor/TensorReduction.h
View File

@ -678,15 +678,15 @@ struct TensorEvaluator<const TensorReductionOp<Op, Dims, ArgType, MakePointer_>,
template <typename S, typename O, bool V> friend struct internal::FullReducerShard;
#endif
#if defined(EIGEN_USE_GPU) && defined(__CUDACC__)
template <int B, int N, typename S, typename R, typename I> friend void internal::FullReductionKernel(R, const S, I, typename S::CoeffReturnType*, unsigned int*);
template <int B, int N, typename S, typename R, typename I> friend __global__ void internal::FullReductionKernel(R, const S, I, typename S::CoeffReturnType*, unsigned int*);
#ifdef EIGEN_HAS_CUDA_FP16
template <typename S, typename R, typename I> friend void internal::ReductionInitFullReduxKernelHalfFloat(R, const S, I, half2*);
template <int B, int N, typename S, typename R, typename I> friend void internal::FullReductionKernelHalfFloat(R, const S, I, half*, half2*);
template <int NPT, typename S, typename R, typename I> friend void internal::InnerReductionKernelHalfFloat(R, const S, I, I, half*);
template <typename S, typename R, typename I> friend __global__ void internal::ReductionInitFullReduxKernelHalfFloat(R, const S, I, half2*);
template <int B, int N, typename S, typename R, typename I> friend __global__ void internal::FullReductionKernelHalfFloat(R, const S, I, half*, half2*);
template <int NPT, typename S, typename R, typename I> friend __global__ void internal::InnerReductionKernelHalfFloat(R, const S, I, I, half*);
#endif
template <int NPT, typename S, typename R, typename I> friend void internal::InnerReductionKernel(R, const S, I, I, typename S::CoeffReturnType*);
template <int NPT, typename S, typename R, typename I> friend __global__ void internal::InnerReductionKernel(R, const S, I, I, typename S::CoeffReturnType*);
template <int NPT, typename S, typename R, typename I> friend void internal::OuterReductionKernel(R, const S, I, I, typename S::CoeffReturnType*);
template <int NPT, typename S, typename R, typename I> friend __global__ void internal::OuterReductionKernel(R, const S, I, I, typename S::CoeffReturnType*);
#endif
template <typename S, typename O, typename D> friend struct internal::InnerReducer;

20
unsupported/Eigen/CXX11/src/Tensor/TensorReductionCuda.h
View File

@ -168,7 +168,12 @@ __global__ void FullReductionKernel(Reducer reducer, const Self input, Index num
#pragma unroll
for (int offset = warpSize/2; offset > 0; offset /= 2) {
#if defined(EIGEN_CUDA_SDK_VER) && EIGEN_CUDA_SDK_VER < 90000
reducer.reduce(__shfl_down(accum, offset, warpSize), &accum);
#else
reducer.reduce(__shfl_down_sync(0xFFFFFFFF, accum, offset, warpSize), &accum);
#endif
}
if ((threadIdx.x & (warpSize - 1)) == 0) {
@ -244,7 +249,11 @@ __global__ void FullReductionKernelHalfFloat(Reducer reducer, const Self input,
#pragma unroll
for (int offset = warpSize/2; offset > 0; offset /= 2) {
#if defined(EIGEN_CUDA_SDK_VER) && EIGEN_CUDA_SDK_VER < 90000
reducer.reducePacket(__shfl_down(accum, offset, warpSize), &accum);
#else
reducer.reducePacket(__shfl_down_sync(0xFFFFFFFF, accum, offset, warpSize), &accum);
#endif
}
if ((threadIdx.x & (warpSize - 1)) == 0) {
@ -426,7 +435,11 @@ __global__ void InnerReductionKernel(Reducer reducer, const Self input, Index nu
#pragma unroll
for (int offset = warpSize/2; offset > 0; offset /= 2) {
#if defined(EIGEN_CUDA_SDK_VER) && EIGEN_CUDA_SDK_VER < 90000
reducer.reduce(__shfl_down(reduced_val, offset), &reduced_val);
#else
reducer.reduce(__shfl_down_sync(0xFFFFFFFF, reduced_val, offset), &reduced_val);
#endif
}
if ((threadIdx.x & (warpSize - 1)) == 0) {
@ -516,8 +529,15 @@ __global__ void InnerReductionKernelHalfFloat(Reducer reducer, const Self input,
#pragma unroll
for (int offset = warpSize/2; offset > 0; offset /= 2) {
#if defined(EIGEN_CUDA_SDK_VER) && EIGEN_CUDA_SDK_VER < 90000
reducer.reducePacket(__shfl_down(reduced_val1, offset, warpSize), &reduced_val1);
reducer.reducePacket(__shfl_down(reduced_val2, offset, warpSize), &reduced_val2);
#else
reducer.reducePacket(__shfl_down_sync(0xFFFFFFFF, reduced_val1, offset, warpSize), &reduced_val1);
reducer.reducePacket(__shfl_down_sync(0xFFFFFFFF, reduced_val2, offset, warpSize), &reduced_val2);
#endif
}
half val1 = __low2half(reduced_val1);

12
unsupported/Eigen/src/SpecialFunctions/SpecialFunctionsPacketMath.h
View File

@ -15,27 +15,27 @@ namespace Eigen {
namespace internal {
/** \internal \returns the ln(|gamma(\a a)|) (coeff-wise) */
template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
template<typename Packet> EIGEN_DEVICE_FUNC EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
Packet plgamma(const Packet& a) { using numext::lgamma; return lgamma(a); }
/** \internal \returns the derivative of lgamma, psi(\a a) (coeff-wise) */
template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
template<typename Packet> EIGEN_DEVICE_FUNC EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
Packet pdigamma(const Packet& a) { using numext::digamma; return digamma(a); }
/** \internal \returns the zeta function of two arguments (coeff-wise) */
template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
template<typename Packet> EIGEN_DEVICE_FUNC EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
Packet pzeta(const Packet& x, const Packet& q) { using numext::zeta; return zeta(x, q); }
/** \internal \returns the polygamma function (coeff-wise) */
template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
template<typename Packet> EIGEN_DEVICE_FUNC EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
Packet ppolygamma(const Packet& n, const Packet& x) { using numext::polygamma; return polygamma(n, x); }
/** \internal \returns the erf(\a a) (coeff-wise) */
template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
template<typename Packet> EIGEN_DEVICE_FUNC EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
Packet perf(const Packet& a) { using numext::erf; return erf(a); }
/** \internal \returns the erfc(\a a) (coeff-wise) */
template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
template<typename Packet> EIGEN_DEVICE_FUNC EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
Packet perfc(const Packet& a) { using numext::erfc; return erfc(a); }
/** \internal \returns the incomplete gamma function igamma(\a a, \a x) */

9
unsupported/test/CMakeLists.txt
View File

@ -216,17 +216,14 @@ if(CUDA_FOUND AND EIGEN_TEST_CUDA)
message(STATUS "Flags used to compile cuda code: " ${CMAKE_CXX_FLAGS})
if( (NOT EIGEN_TEST_CXX11) OR (CMAKE_VERSION VERSION_LESS 3.3))
set(EIGEN_CUDA_CXX11_FLAG "-std=c++11")
else()
# otherwise the flag has already been added because of the above set(CMAKE_CXX_STANDARD 11)
set(EIGEN_CUDA_CXX11_FLAG "")
string(APPEND EIGEN_CUDA_CXX11_FLAGS " -std=c++11")
endif()
if("${CMAKE_CXX_COMPILER_ID}" STREQUAL "Clang")
set(CUDA_NVCC_FLAGS "-ccbin ${CMAKE_C_COMPILER}" CACHE STRING "nvcc flags" FORCE)
endif()
if(EIGEN_TEST_CUDA_CLANG)
string(APPEND CMAKE_CXX_FLAGS " --cuda-path=${CUDA_TOOLKIT_ROOT_DIR} ${EIGEN_CUDA_CXX11_FLAG}")
string(APPEND CMAKE_CXX_FLAGS " --cuda-path=${CUDA_TOOLKIT_ROOT_DIR} ${EIGEN_CUDA_CXX11_FLAGS}")
foreach(ARCH IN LISTS EIGEN_CUDA_COMPUTE_ARCH)
string(APPEND CMAKE_CXX_FLAGS " --cuda-gpu-arch=sm_${ARCH}")
endforeach()
@ -246,7 +243,7 @@ if(CUDA_FOUND AND EIGEN_TEST_CUDA)
foreach(ARCH IN LISTS EIGEN_CUDA_COMPUTE_ARCH)
string(APPEND NVCC_ARCH_FLAGS " -gencode arch=compute_${ARCH},code=sm_${ARCH}")
endforeach()
set(CUDA_NVCC_FLAGS "--expt-relaxed-constexpr -Xcudafe \"--display_error_number\" ${NVCC_ARCH_FLAGS} ${CUDA_NVCC_FLAGS} ${EIGEN_CUDA_CXX11_FLAG}")
set(CUDA_NVCC_FLAGS "--expt-relaxed-constexpr -Xcudafe \"--display_error_number\" ${NVCC_ARCH_FLAGS} ${CUDA_NVCC_FLAGS} ${EIGEN_CUDA_CXX11_FLAGS}")
cuda_include_directories("${CMAKE_CURRENT_BINARY_DIR}" "${CUDA_TOOLKIT_ROOT_DIR}/include")
endif()