| 1 | // This file is part of Eigen, a lightweight C++ template library |
| 2 | // for linear algebra. |
| 3 | // |
| 4 | // Copyright (C) 2009-2014 Gael Guennebaud <gael.guennebaud@inria.fr> |
| 5 | // |
| 6 | // This Source Code Form is subject to the terms of the Mozilla |
| 7 | // Public License v. 2.0. If a copy of the MPL was not distributed |
| 8 | // with this file, You can obtain one at http://mozilla.org/MPL/2.0/. |
| 9 | |
| 10 | #ifndef EIGEN_SPARSE_SELFADJOINTVIEW_H |
| 11 | #define EIGEN_SPARSE_SELFADJOINTVIEW_H |
| 12 | |
| 13 | namespace Eigen { |
| 14 | |
| 15 | /** \ingroup SparseCore_Module |
| 16 | * \class SparseSelfAdjointView |
| 17 | * |
| 18 | * \brief Pseudo expression to manipulate a triangular sparse matrix as a selfadjoint matrix. |
| 19 | * |
| 20 | * \param MatrixType the type of the dense matrix storing the coefficients |
| 21 | * \param Mode can be either \c #Lower or \c #Upper |
| 22 | * |
| 23 | * This class is an expression of a sefladjoint matrix from a triangular part of a matrix |
| 24 | * with given dense storage of the coefficients. It is the return type of MatrixBase::selfadjointView() |
| 25 | * and most of the time this is the only way that it is used. |
| 26 | * |
| 27 | * \sa SparseMatrixBase::selfadjointView() |
| 28 | */ |
| 29 | namespace internal { |
| 30 | |
| 31 | template<typename MatrixType, unsigned int Mode> |
| 32 | struct traits<SparseSelfAdjointView<MatrixType,Mode> > : traits<MatrixType> { |
| 33 | }; |
| 34 | |
| 35 | template<int SrcMode,int DstMode,typename MatrixType,int DestOrder> |
| 36 | void permute_symm_to_symm(const MatrixType& mat, SparseMatrix<typename MatrixType::Scalar,DestOrder,typename MatrixType::StorageIndex>& _dest, const typename MatrixType::StorageIndex* perm = 0); |
| 37 | |
| 38 | template<int Mode,typename MatrixType,int DestOrder> |
| 39 | void permute_symm_to_fullsymm(const MatrixType& mat, SparseMatrix<typename MatrixType::Scalar,DestOrder,typename MatrixType::StorageIndex>& _dest, const typename MatrixType::StorageIndex* perm = 0); |
| 40 | |
| 41 | } |
| 42 | |
| 43 | template<typename MatrixType, unsigned int _Mode> class SparseSelfAdjointView |
| 44 | : public EigenBase<SparseSelfAdjointView<MatrixType,_Mode> > |
| 45 | { |
| 46 | public: |
| 47 | |
| 48 | enum { |
| 49 | Mode = _Mode, |
| 50 | TransposeMode = ((Mode & Upper) ? Lower : 0) | ((Mode & Lower) ? Upper : 0), |
| 51 | RowsAtCompileTime = internal::traits<SparseSelfAdjointView>::RowsAtCompileTime, |
| 52 | ColsAtCompileTime = internal::traits<SparseSelfAdjointView>::ColsAtCompileTime |
| 53 | }; |
| 54 | |
| 55 | typedef EigenBase<SparseSelfAdjointView> Base; |
| 56 | typedef typename MatrixType::Scalar Scalar; |
| 57 | typedef typename MatrixType::StorageIndex StorageIndex; |
| 58 | typedef Matrix<StorageIndex,Dynamic,1> VectorI; |
| 59 | typedef typename internal::ref_selector<MatrixType>::non_const_type MatrixTypeNested; |
| 60 | typedef typename internal::remove_all<MatrixTypeNested>::type _MatrixTypeNested; |
| 61 | |
| 62 | explicit inline SparseSelfAdjointView(MatrixType& matrix) : m_matrix(matrix) |
| 63 | { |
| 64 | eigen_assert(rows()==cols() && "SelfAdjointView is only for squared matrices" ); |
| 65 | } |
| 66 | |
| 67 | inline Index rows() const { return m_matrix.rows(); } |
| 68 | inline Index cols() const { return m_matrix.cols(); } |
| 69 | |
| 70 | /** \internal \returns a reference to the nested matrix */ |
| 71 | const _MatrixTypeNested& matrix() const { return m_matrix; } |
| 72 | typename internal::remove_reference<MatrixTypeNested>::type& matrix() { return m_matrix; } |
| 73 | |
| 74 | /** \returns an expression of the matrix product between a sparse self-adjoint matrix \c *this and a sparse matrix \a rhs. |
| 75 | * |
| 76 | * Note that there is no algorithmic advantage of performing such a product compared to a general sparse-sparse matrix product. |
| 77 | * Indeed, the SparseSelfadjointView operand is first copied into a temporary SparseMatrix before computing the product. |
| 78 | */ |
| 79 | template<typename OtherDerived> |
| 80 | Product<SparseSelfAdjointView, OtherDerived> |
| 81 | operator*(const SparseMatrixBase<OtherDerived>& rhs) const |
| 82 | { |
| 83 | return Product<SparseSelfAdjointView, OtherDerived>(*this, rhs.derived()); |
| 84 | } |
| 85 | |
| 86 | /** \returns an expression of the matrix product between a sparse matrix \a lhs and a sparse self-adjoint matrix \a rhs. |
| 87 | * |
| 88 | * Note that there is no algorithmic advantage of performing such a product compared to a general sparse-sparse matrix product. |
| 89 | * Indeed, the SparseSelfadjointView operand is first copied into a temporary SparseMatrix before computing the product. |
| 90 | */ |
| 91 | template<typename OtherDerived> friend |
| 92 | Product<OtherDerived, SparseSelfAdjointView> |
| 93 | operator*(const SparseMatrixBase<OtherDerived>& lhs, const SparseSelfAdjointView& rhs) |
| 94 | { |
| 95 | return Product<OtherDerived, SparseSelfAdjointView>(lhs.derived(), rhs); |
| 96 | } |
| 97 | |
| 98 | /** Efficient sparse self-adjoint matrix times dense vector/matrix product */ |
| 99 | template<typename OtherDerived> |
| 100 | Product<SparseSelfAdjointView,OtherDerived> |
| 101 | operator*(const MatrixBase<OtherDerived>& rhs) const |
| 102 | { |
| 103 | return Product<SparseSelfAdjointView,OtherDerived>(*this, rhs.derived()); |
| 104 | } |
| 105 | |
| 106 | /** Efficient dense vector/matrix times sparse self-adjoint matrix product */ |
| 107 | template<typename OtherDerived> friend |
| 108 | Product<OtherDerived,SparseSelfAdjointView> |
| 109 | operator*(const MatrixBase<OtherDerived>& lhs, const SparseSelfAdjointView& rhs) |
| 110 | { |
| 111 | return Product<OtherDerived,SparseSelfAdjointView>(lhs.derived(), rhs); |
| 112 | } |
| 113 | |
| 114 | /** Perform a symmetric rank K update of the selfadjoint matrix \c *this: |
| 115 | * \f$ this = this + \alpha ( u u^* ) \f$ where \a u is a vector or matrix. |
| 116 | * |
| 117 | * \returns a reference to \c *this |
| 118 | * |
| 119 | * To perform \f$ this = this + \alpha ( u^* u ) \f$ you can simply |
| 120 | * call this function with u.adjoint(). |
| 121 | */ |
| 122 | template<typename DerivedU> |
| 123 | SparseSelfAdjointView& rankUpdate(const SparseMatrixBase<DerivedU>& u, const Scalar& alpha = Scalar(1)); |
| 124 | |
| 125 | /** \returns an expression of P H P^-1 */ |
| 126 | // TODO implement twists in a more evaluator friendly fashion |
| 127 | SparseSymmetricPermutationProduct<_MatrixTypeNested,Mode> twistedBy(const PermutationMatrix<Dynamic,Dynamic,StorageIndex>& perm) const |
| 128 | { |
| 129 | return SparseSymmetricPermutationProduct<_MatrixTypeNested,Mode>(m_matrix, perm); |
| 130 | } |
| 131 | |
| 132 | template<typename SrcMatrixType,int SrcMode> |
| 133 | SparseSelfAdjointView& operator=(const SparseSymmetricPermutationProduct<SrcMatrixType,SrcMode>& permutedMatrix) |
| 134 | { |
| 135 | internal::call_assignment_no_alias_no_transpose(*this, permutedMatrix); |
| 136 | return *this; |
| 137 | } |
| 138 | |
| 139 | SparseSelfAdjointView& operator=(const SparseSelfAdjointView& src) |
| 140 | { |
| 141 | PermutationMatrix<Dynamic,Dynamic,StorageIndex> pnull; |
| 142 | return *this = src.twistedBy(pnull); |
| 143 | } |
| 144 | |
| 145 | template<typename SrcMatrixType,unsigned int SrcMode> |
| 146 | SparseSelfAdjointView& operator=(const SparseSelfAdjointView<SrcMatrixType,SrcMode>& src) |
| 147 | { |
| 148 | PermutationMatrix<Dynamic,Dynamic,StorageIndex> pnull; |
| 149 | return *this = src.twistedBy(pnull); |
| 150 | } |
| 151 | |
| 152 | void resize(Index rows, Index cols) |
| 153 | { |
| 154 | EIGEN_ONLY_USED_FOR_DEBUG(rows); |
| 155 | EIGEN_ONLY_USED_FOR_DEBUG(cols); |
| 156 | eigen_assert(rows == this->rows() && cols == this->cols() |
| 157 | && "SparseSelfadjointView::resize() does not actually allow to resize." ); |
| 158 | } |
| 159 | |
| 160 | protected: |
| 161 | |
| 162 | MatrixTypeNested m_matrix; |
| 163 | //mutable VectorI m_countPerRow; |
| 164 | //mutable VectorI m_countPerCol; |
| 165 | private: |
| 166 | template<typename Dest> void evalTo(Dest &) const; |
| 167 | }; |
| 168 | |
| 169 | /*************************************************************************** |
| 170 | * Implementation of SparseMatrixBase methods |
| 171 | ***************************************************************************/ |
| 172 | |
| 173 | template<typename Derived> |
| 174 | template<unsigned int UpLo> |
| 175 | typename SparseMatrixBase<Derived>::template ConstSelfAdjointViewReturnType<UpLo>::Type SparseMatrixBase<Derived>::selfadjointView() const |
| 176 | { |
| 177 | return SparseSelfAdjointView<const Derived, UpLo>(derived()); |
| 178 | } |
| 179 | |
| 180 | template<typename Derived> |
| 181 | template<unsigned int UpLo> |
| 182 | typename SparseMatrixBase<Derived>::template SelfAdjointViewReturnType<UpLo>::Type SparseMatrixBase<Derived>::selfadjointView() |
| 183 | { |
| 184 | return SparseSelfAdjointView<Derived, UpLo>(derived()); |
| 185 | } |
| 186 | |
| 187 | /*************************************************************************** |
| 188 | * Implementation of SparseSelfAdjointView methods |
| 189 | ***************************************************************************/ |
| 190 | |
| 191 | template<typename MatrixType, unsigned int Mode> |
| 192 | template<typename DerivedU> |
| 193 | SparseSelfAdjointView<MatrixType,Mode>& |
| 194 | SparseSelfAdjointView<MatrixType,Mode>::rankUpdate(const SparseMatrixBase<DerivedU>& u, const Scalar& alpha) |
| 195 | { |
| 196 | SparseMatrix<Scalar,(MatrixType::Flags&RowMajorBit)?RowMajor:ColMajor> tmp = u * u.adjoint(); |
| 197 | if(alpha==Scalar(0)) |
| 198 | m_matrix = tmp.template triangularView<Mode>(); |
| 199 | else |
| 200 | m_matrix += alpha * tmp.template triangularView<Mode>(); |
| 201 | |
| 202 | return *this; |
| 203 | } |
| 204 | |
| 205 | namespace internal { |
| 206 | |
| 207 | // TODO currently a selfadjoint expression has the form SelfAdjointView<.,.> |
| 208 | // in the future selfadjoint-ness should be defined by the expression traits |
| 209 | // such that Transpose<SelfAdjointView<.,.> > is valid. (currently TriangularBase::transpose() is overloaded to make it work) |
| 210 | template<typename MatrixType, unsigned int Mode> |
| 211 | struct evaluator_traits<SparseSelfAdjointView<MatrixType,Mode> > |
| 212 | { |
| 213 | typedef typename storage_kind_to_evaluator_kind<typename MatrixType::StorageKind>::Kind Kind; |
| 214 | typedef SparseSelfAdjointShape Shape; |
| 215 | }; |
| 216 | |
| 217 | struct SparseSelfAdjoint2Sparse {}; |
| 218 | |
| 219 | template<> struct AssignmentKind<SparseShape,SparseSelfAdjointShape> { typedef SparseSelfAdjoint2Sparse Kind; }; |
| 220 | template<> struct AssignmentKind<SparseSelfAdjointShape,SparseShape> { typedef Sparse2Sparse Kind; }; |
| 221 | |
| 222 | template< typename DstXprType, typename SrcXprType, typename Functor> |
| 223 | struct Assignment<DstXprType, SrcXprType, Functor, SparseSelfAdjoint2Sparse> |
| 224 | { |
| 225 | typedef typename DstXprType::StorageIndex StorageIndex; |
| 226 | typedef internal::assign_op<typename DstXprType::Scalar,typename SrcXprType::Scalar> AssignOpType; |
| 227 | |
| 228 | template<typename DestScalar,int StorageOrder> |
| 229 | static void run(SparseMatrix<DestScalar,StorageOrder,StorageIndex> &dst, const SrcXprType &src, const AssignOpType&/*func*/) |
| 230 | { |
| 231 | internal::permute_symm_to_fullsymm<SrcXprType::Mode>(src.matrix(), dst); |
| 232 | } |
| 233 | |
| 234 | // FIXME: the handling of += and -= in sparse matrices should be cleanup so that next two overloads could be reduced to: |
| 235 | template<typename DestScalar,int StorageOrder,typename AssignFunc> |
| 236 | static void run(SparseMatrix<DestScalar,StorageOrder,StorageIndex> &dst, const SrcXprType &src, const AssignFunc& func) |
| 237 | { |
| 238 | SparseMatrix<DestScalar,StorageOrder,StorageIndex> tmp(src.rows(),src.cols()); |
| 239 | run(tmp, src, AssignOpType()); |
| 240 | call_assignment_no_alias_no_transpose(dst, tmp, func); |
| 241 | } |
| 242 | |
| 243 | template<typename DestScalar,int StorageOrder> |
| 244 | static void run(SparseMatrix<DestScalar,StorageOrder,StorageIndex> &dst, const SrcXprType &src, |
| 245 | const internal::add_assign_op<typename DstXprType::Scalar,typename SrcXprType::Scalar>& /* func */) |
| 246 | { |
| 247 | SparseMatrix<DestScalar,StorageOrder,StorageIndex> tmp(src.rows(),src.cols()); |
| 248 | run(tmp, src, AssignOpType()); |
| 249 | dst += tmp; |
| 250 | } |
| 251 | |
| 252 | template<typename DestScalar,int StorageOrder> |
| 253 | static void run(SparseMatrix<DestScalar,StorageOrder,StorageIndex> &dst, const SrcXprType &src, |
| 254 | const internal::sub_assign_op<typename DstXprType::Scalar,typename SrcXprType::Scalar>& /* func */) |
| 255 | { |
| 256 | SparseMatrix<DestScalar,StorageOrder,StorageIndex> tmp(src.rows(),src.cols()); |
| 257 | run(tmp, src, AssignOpType()); |
| 258 | dst -= tmp; |
| 259 | } |
| 260 | |
| 261 | template<typename DestScalar> |
| 262 | static void run(DynamicSparseMatrix<DestScalar,ColMajor,StorageIndex>& dst, const SrcXprType &src, const AssignOpType&/*func*/) |
| 263 | { |
| 264 | // TODO directly evaluate into dst; |
| 265 | SparseMatrix<DestScalar,ColMajor,StorageIndex> tmp(dst.rows(),dst.cols()); |
| 266 | internal::permute_symm_to_fullsymm<SrcXprType::Mode>(src.matrix(), tmp); |
| 267 | dst = tmp; |
| 268 | } |
| 269 | }; |
| 270 | |
| 271 | } // end namespace internal |
| 272 | |
| 273 | /*************************************************************************** |
| 274 | * Implementation of sparse self-adjoint time dense matrix |
| 275 | ***************************************************************************/ |
| 276 | |
| 277 | namespace internal { |
| 278 | |
| 279 | template<int Mode, typename SparseLhsType, typename DenseRhsType, typename DenseResType, typename AlphaType> |
| 280 | inline void sparse_selfadjoint_time_dense_product(const SparseLhsType& lhs, const DenseRhsType& rhs, DenseResType& res, const AlphaType& alpha) |
| 281 | { |
| 282 | EIGEN_ONLY_USED_FOR_DEBUG(alpha); |
| 283 | |
| 284 | typedef typename internal::nested_eval<SparseLhsType,DenseRhsType::MaxColsAtCompileTime>::type SparseLhsTypeNested; |
| 285 | typedef typename internal::remove_all<SparseLhsTypeNested>::type SparseLhsTypeNestedCleaned; |
| 286 | typedef evaluator<SparseLhsTypeNestedCleaned> LhsEval; |
| 287 | typedef typename LhsEval::InnerIterator LhsIterator; |
| 288 | typedef typename SparseLhsType::Scalar LhsScalar; |
| 289 | |
| 290 | enum { |
| 291 | LhsIsRowMajor = (LhsEval::Flags&RowMajorBit)==RowMajorBit, |
| 292 | ProcessFirstHalf = |
| 293 | ((Mode&(Upper|Lower))==(Upper|Lower)) |
| 294 | || ( (Mode&Upper) && !LhsIsRowMajor) |
| 295 | || ( (Mode&Lower) && LhsIsRowMajor), |
| 296 | ProcessSecondHalf = !ProcessFirstHalf |
| 297 | }; |
| 298 | |
| 299 | SparseLhsTypeNested lhs_nested(lhs); |
| 300 | LhsEval lhsEval(lhs_nested); |
| 301 | |
| 302 | // work on one column at once |
| 303 | for (Index k=0; k<rhs.cols(); ++k) |
| 304 | { |
| 305 | for (Index j=0; j<lhs.outerSize(); ++j) |
| 306 | { |
| 307 | LhsIterator i(lhsEval,j); |
| 308 | // handle diagonal coeff |
| 309 | if (ProcessSecondHalf) |
| 310 | { |
| 311 | while (i && i.index()<j) ++i; |
| 312 | if(i && i.index()==j) |
| 313 | { |
| 314 | res.coeffRef(j,k) += alpha * i.value() * rhs.coeff(j,k); |
| 315 | ++i; |
| 316 | } |
| 317 | } |
| 318 | |
| 319 | // premultiplied rhs for scatters |
| 320 | typename ScalarBinaryOpTraits<AlphaType, typename DenseRhsType::Scalar>::ReturnType rhs_j(alpha*rhs(j,k)); |
| 321 | // accumulator for partial scalar product |
| 322 | typename DenseResType::Scalar res_j(0); |
| 323 | for(; (ProcessFirstHalf ? i && i.index() < j : i) ; ++i) |
| 324 | { |
| 325 | LhsScalar lhs_ij = i.value(); |
| 326 | if(!LhsIsRowMajor) lhs_ij = numext::conj(lhs_ij); |
| 327 | res_j += lhs_ij * rhs.coeff(i.index(),k); |
| 328 | res(i.index(),k) += numext::conj(lhs_ij) * rhs_j; |
| 329 | } |
| 330 | res.coeffRef(j,k) += alpha * res_j; |
| 331 | |
| 332 | // handle diagonal coeff |
| 333 | if (ProcessFirstHalf && i && (i.index()==j)) |
| 334 | res.coeffRef(j,k) += alpha * i.value() * rhs.coeff(j,k); |
| 335 | } |
| 336 | } |
| 337 | } |
| 338 | |
| 339 | |
| 340 | template<typename LhsView, typename Rhs, int ProductType> |
| 341 | struct generic_product_impl<LhsView, Rhs, SparseSelfAdjointShape, DenseShape, ProductType> |
| 342 | : generic_product_impl_base<LhsView, Rhs, generic_product_impl<LhsView, Rhs, SparseSelfAdjointShape, DenseShape, ProductType> > |
| 343 | { |
| 344 | template<typename Dest> |
| 345 | static void scaleAndAddTo(Dest& dst, const LhsView& lhsView, const Rhs& rhs, const typename Dest::Scalar& alpha) |
| 346 | { |
| 347 | typedef typename LhsView::_MatrixTypeNested Lhs; |
| 348 | typedef typename nested_eval<Lhs,Dynamic>::type LhsNested; |
| 349 | typedef typename nested_eval<Rhs,Dynamic>::type RhsNested; |
| 350 | LhsNested lhsNested(lhsView.matrix()); |
| 351 | RhsNested rhsNested(rhs); |
| 352 | |
| 353 | internal::sparse_selfadjoint_time_dense_product<LhsView::Mode>(lhsNested, rhsNested, dst, alpha); |
| 354 | } |
| 355 | }; |
| 356 | |
| 357 | template<typename Lhs, typename RhsView, int ProductType> |
| 358 | struct generic_product_impl<Lhs, RhsView, DenseShape, SparseSelfAdjointShape, ProductType> |
| 359 | : generic_product_impl_base<Lhs, RhsView, generic_product_impl<Lhs, RhsView, DenseShape, SparseSelfAdjointShape, ProductType> > |
| 360 | { |
| 361 | template<typename Dest> |
| 362 | static void scaleAndAddTo(Dest& dst, const Lhs& lhs, const RhsView& rhsView, const typename Dest::Scalar& alpha) |
| 363 | { |
| 364 | typedef typename RhsView::_MatrixTypeNested Rhs; |
| 365 | typedef typename nested_eval<Lhs,Dynamic>::type LhsNested; |
| 366 | typedef typename nested_eval<Rhs,Dynamic>::type RhsNested; |
| 367 | LhsNested lhsNested(lhs); |
| 368 | RhsNested rhsNested(rhsView.matrix()); |
| 369 | |
| 370 | // transpose everything |
| 371 | Transpose<Dest> dstT(dst); |
| 372 | internal::sparse_selfadjoint_time_dense_product<RhsView::TransposeMode>(rhsNested.transpose(), lhsNested.transpose(), dstT, alpha); |
| 373 | } |
| 374 | }; |
| 375 | |
| 376 | // NOTE: these two overloads are needed to evaluate the sparse selfadjoint view into a full sparse matrix |
| 377 | // TODO: maybe the copy could be handled by generic_product_impl so that these overloads would not be needed anymore |
| 378 | |
| 379 | template<typename LhsView, typename Rhs, int ProductTag> |
| 380 | struct product_evaluator<Product<LhsView, Rhs, DefaultProduct>, ProductTag, SparseSelfAdjointShape, SparseShape> |
| 381 | : public evaluator<typename Product<typename Rhs::PlainObject, Rhs, DefaultProduct>::PlainObject> |
| 382 | { |
| 383 | typedef Product<LhsView, Rhs, DefaultProduct> XprType; |
| 384 | typedef typename XprType::PlainObject PlainObject; |
| 385 | typedef evaluator<PlainObject> Base; |
| 386 | |
| 387 | product_evaluator(const XprType& xpr) |
| 388 | : m_lhs(xpr.lhs()), m_result(xpr.rows(), xpr.cols()) |
| 389 | { |
| 390 | ::new (static_cast<Base*>(this)) Base(m_result); |
| 391 | generic_product_impl<typename Rhs::PlainObject, Rhs, SparseShape, SparseShape, ProductTag>::evalTo(m_result, m_lhs, xpr.rhs()); |
| 392 | } |
| 393 | |
| 394 | protected: |
| 395 | typename Rhs::PlainObject m_lhs; |
| 396 | PlainObject m_result; |
| 397 | }; |
| 398 | |
| 399 | template<typename Lhs, typename RhsView, int ProductTag> |
| 400 | struct product_evaluator<Product<Lhs, RhsView, DefaultProduct>, ProductTag, SparseShape, SparseSelfAdjointShape> |
| 401 | : public evaluator<typename Product<Lhs, typename Lhs::PlainObject, DefaultProduct>::PlainObject> |
| 402 | { |
| 403 | typedef Product<Lhs, RhsView, DefaultProduct> XprType; |
| 404 | typedef typename XprType::PlainObject PlainObject; |
| 405 | typedef evaluator<PlainObject> Base; |
| 406 | |
| 407 | product_evaluator(const XprType& xpr) |
| 408 | : m_rhs(xpr.rhs()), m_result(xpr.rows(), xpr.cols()) |
| 409 | { |
| 410 | ::new (static_cast<Base*>(this)) Base(m_result); |
| 411 | generic_product_impl<Lhs, typename Lhs::PlainObject, SparseShape, SparseShape, ProductTag>::evalTo(m_result, xpr.lhs(), m_rhs); |
| 412 | } |
| 413 | |
| 414 | protected: |
| 415 | typename Lhs::PlainObject m_rhs; |
| 416 | PlainObject m_result; |
| 417 | }; |
| 418 | |
| 419 | } // namespace internal |
| 420 | |
| 421 | /*************************************************************************** |
| 422 | * Implementation of symmetric copies and permutations |
| 423 | ***************************************************************************/ |
| 424 | namespace internal { |
| 425 | |
| 426 | template<int Mode,typename MatrixType,int DestOrder> |
| 427 | void permute_symm_to_fullsymm(const MatrixType& mat, SparseMatrix<typename MatrixType::Scalar,DestOrder,typename MatrixType::StorageIndex>& _dest, const typename MatrixType::StorageIndex* perm) |
| 428 | { |
| 429 | typedef typename MatrixType::StorageIndex StorageIndex; |
| 430 | typedef typename MatrixType::Scalar Scalar; |
| 431 | typedef SparseMatrix<Scalar,DestOrder,StorageIndex> Dest; |
| 432 | typedef Matrix<StorageIndex,Dynamic,1> VectorI; |
| 433 | typedef evaluator<MatrixType> MatEval; |
| 434 | typedef typename evaluator<MatrixType>::InnerIterator MatIterator; |
| 435 | |
| 436 | MatEval matEval(mat); |
| 437 | Dest& dest(_dest.derived()); |
| 438 | enum { |
| 439 | StorageOrderMatch = int(Dest::IsRowMajor) == int(MatrixType::IsRowMajor) |
| 440 | }; |
| 441 | |
| 442 | Index size = mat.rows(); |
| 443 | VectorI count; |
| 444 | count.resize(size); |
| 445 | count.setZero(); |
| 446 | dest.resize(size,size); |
| 447 | for(Index j = 0; j<size; ++j) |
| 448 | { |
| 449 | Index jp = perm ? perm[j] : j; |
| 450 | for(MatIterator it(matEval,j); it; ++it) |
| 451 | { |
| 452 | Index i = it.index(); |
| 453 | Index r = it.row(); |
| 454 | Index c = it.col(); |
| 455 | Index ip = perm ? perm[i] : i; |
| 456 | if(Mode==(Upper|Lower)) |
| 457 | count[StorageOrderMatch ? jp : ip]++; |
| 458 | else if(r==c) |
| 459 | count[ip]++; |
| 460 | else if(( Mode==Lower && r>c) || ( Mode==Upper && r<c)) |
| 461 | { |
| 462 | count[ip]++; |
| 463 | count[jp]++; |
| 464 | } |
| 465 | } |
| 466 | } |
| 467 | Index nnz = count.sum(); |
| 468 | |
| 469 | // reserve space |
| 470 | dest.resizeNonZeros(nnz); |
| 471 | dest.outerIndexPtr()[0] = 0; |
| 472 | for(Index j=0; j<size; ++j) |
| 473 | dest.outerIndexPtr()[j+1] = dest.outerIndexPtr()[j] + count[j]; |
| 474 | for(Index j=0; j<size; ++j) |
| 475 | count[j] = dest.outerIndexPtr()[j]; |
| 476 | |
| 477 | // copy data |
| 478 | for(StorageIndex j = 0; j<size; ++j) |
| 479 | { |
| 480 | for(MatIterator it(matEval,j); it; ++it) |
| 481 | { |
| 482 | StorageIndex i = internal::convert_index<StorageIndex>(it.index()); |
| 483 | Index r = it.row(); |
| 484 | Index c = it.col(); |
| 485 | |
| 486 | StorageIndex jp = perm ? perm[j] : j; |
| 487 | StorageIndex ip = perm ? perm[i] : i; |
| 488 | |
| 489 | if(Mode==(Upper|Lower)) |
| 490 | { |
| 491 | Index k = count[StorageOrderMatch ? jp : ip]++; |
| 492 | dest.innerIndexPtr()[k] = StorageOrderMatch ? ip : jp; |
| 493 | dest.valuePtr()[k] = it.value(); |
| 494 | } |
| 495 | else if(r==c) |
| 496 | { |
| 497 | Index k = count[ip]++; |
| 498 | dest.innerIndexPtr()[k] = ip; |
| 499 | dest.valuePtr()[k] = it.value(); |
| 500 | } |
| 501 | else if(( (Mode&Lower)==Lower && r>c) || ( (Mode&Upper)==Upper && r<c)) |
| 502 | { |
| 503 | if(!StorageOrderMatch) |
| 504 | std::swap(ip,jp); |
| 505 | Index k = count[jp]++; |
| 506 | dest.innerIndexPtr()[k] = ip; |
| 507 | dest.valuePtr()[k] = it.value(); |
| 508 | k = count[ip]++; |
| 509 | dest.innerIndexPtr()[k] = jp; |
| 510 | dest.valuePtr()[k] = numext::conj(it.value()); |
| 511 | } |
| 512 | } |
| 513 | } |
| 514 | } |
| 515 | |
| 516 | template<int _SrcMode,int _DstMode,typename MatrixType,int DstOrder> |
| 517 | void permute_symm_to_symm(const MatrixType& mat, SparseMatrix<typename MatrixType::Scalar,DstOrder,typename MatrixType::StorageIndex>& _dest, const typename MatrixType::StorageIndex* perm) |
| 518 | { |
| 519 | typedef typename MatrixType::StorageIndex StorageIndex; |
| 520 | typedef typename MatrixType::Scalar Scalar; |
| 521 | SparseMatrix<Scalar,DstOrder,StorageIndex>& dest(_dest.derived()); |
| 522 | typedef Matrix<StorageIndex,Dynamic,1> VectorI; |
| 523 | typedef evaluator<MatrixType> MatEval; |
| 524 | typedef typename evaluator<MatrixType>::InnerIterator MatIterator; |
| 525 | |
| 526 | enum { |
| 527 | SrcOrder = MatrixType::IsRowMajor ? RowMajor : ColMajor, |
| 528 | StorageOrderMatch = int(SrcOrder) == int(DstOrder), |
| 529 | DstMode = DstOrder==RowMajor ? (_DstMode==Upper ? Lower : Upper) : _DstMode, |
| 530 | SrcMode = SrcOrder==RowMajor ? (_SrcMode==Upper ? Lower : Upper) : _SrcMode |
| 531 | }; |
| 532 | |
| 533 | MatEval matEval(mat); |
| 534 | |
| 535 | Index size = mat.rows(); |
| 536 | VectorI count(size); |
| 537 | count.setZero(); |
| 538 | dest.resize(size,size); |
| 539 | for(StorageIndex j = 0; j<size; ++j) |
| 540 | { |
| 541 | StorageIndex jp = perm ? perm[j] : j; |
| 542 | for(MatIterator it(matEval,j); it; ++it) |
| 543 | { |
| 544 | StorageIndex i = it.index(); |
| 545 | if((int(SrcMode)==int(Lower) && i<j) || (int(SrcMode)==int(Upper) && i>j)) |
| 546 | continue; |
| 547 | |
| 548 | StorageIndex ip = perm ? perm[i] : i; |
| 549 | count[int(DstMode)==int(Lower) ? (std::min)(ip,jp) : (std::max)(ip,jp)]++; |
| 550 | } |
| 551 | } |
| 552 | dest.outerIndexPtr()[0] = 0; |
| 553 | for(Index j=0; j<size; ++j) |
| 554 | dest.outerIndexPtr()[j+1] = dest.outerIndexPtr()[j] + count[j]; |
| 555 | dest.resizeNonZeros(dest.outerIndexPtr()[size]); |
| 556 | for(Index j=0; j<size; ++j) |
| 557 | count[j] = dest.outerIndexPtr()[j]; |
| 558 | |
| 559 | for(StorageIndex j = 0; j<size; ++j) |
| 560 | { |
| 561 | |
| 562 | for(MatIterator it(matEval,j); it; ++it) |
| 563 | { |
| 564 | StorageIndex i = it.index(); |
| 565 | if((int(SrcMode)==int(Lower) && i<j) || (int(SrcMode)==int(Upper) && i>j)) |
| 566 | continue; |
| 567 | |
| 568 | StorageIndex jp = perm ? perm[j] : j; |
| 569 | StorageIndex ip = perm? perm[i] : i; |
| 570 | |
| 571 | Index k = count[int(DstMode)==int(Lower) ? (std::min)(ip,jp) : (std::max)(ip,jp)]++; |
| 572 | dest.innerIndexPtr()[k] = int(DstMode)==int(Lower) ? (std::max)(ip,jp) : (std::min)(ip,jp); |
| 573 | |
| 574 | if(!StorageOrderMatch) std::swap(ip,jp); |
| 575 | if( ((int(DstMode)==int(Lower) && ip<jp) || (int(DstMode)==int(Upper) && ip>jp))) |
| 576 | dest.valuePtr()[k] = numext::conj(it.value()); |
| 577 | else |
| 578 | dest.valuePtr()[k] = it.value(); |
| 579 | } |
| 580 | } |
| 581 | } |
| 582 | |
| 583 | } |
| 584 | |
| 585 | // TODO implement twists in a more evaluator friendly fashion |
| 586 | |
| 587 | namespace internal { |
| 588 | |
| 589 | template<typename MatrixType, int Mode> |
| 590 | struct traits<SparseSymmetricPermutationProduct<MatrixType,Mode> > : traits<MatrixType> { |
| 591 | }; |
| 592 | |
| 593 | } |
| 594 | |
| 595 | template<typename MatrixType,int Mode> |
| 596 | class SparseSymmetricPermutationProduct |
| 597 | : public EigenBase<SparseSymmetricPermutationProduct<MatrixType,Mode> > |
| 598 | { |
| 599 | public: |
| 600 | typedef typename MatrixType::Scalar Scalar; |
| 601 | typedef typename MatrixType::StorageIndex StorageIndex; |
| 602 | enum { |
| 603 | RowsAtCompileTime = internal::traits<SparseSymmetricPermutationProduct>::RowsAtCompileTime, |
| 604 | ColsAtCompileTime = internal::traits<SparseSymmetricPermutationProduct>::ColsAtCompileTime |
| 605 | }; |
| 606 | protected: |
| 607 | typedef PermutationMatrix<Dynamic,Dynamic,StorageIndex> Perm; |
| 608 | public: |
| 609 | typedef Matrix<StorageIndex,Dynamic,1> VectorI; |
| 610 | typedef typename MatrixType::Nested MatrixTypeNested; |
| 611 | typedef typename internal::remove_all<MatrixTypeNested>::type NestedExpression; |
| 612 | |
| 613 | SparseSymmetricPermutationProduct(const MatrixType& mat, const Perm& perm) |
| 614 | : m_matrix(mat), m_perm(perm) |
| 615 | {} |
| 616 | |
| 617 | inline Index rows() const { return m_matrix.rows(); } |
| 618 | inline Index cols() const { return m_matrix.cols(); } |
| 619 | |
| 620 | const NestedExpression& matrix() const { return m_matrix; } |
| 621 | const Perm& perm() const { return m_perm; } |
| 622 | |
| 623 | protected: |
| 624 | MatrixTypeNested m_matrix; |
| 625 | const Perm& m_perm; |
| 626 | |
| 627 | }; |
| 628 | |
| 629 | namespace internal { |
| 630 | |
| 631 | template<typename DstXprType, typename MatrixType, int Mode, typename Scalar> |
| 632 | struct Assignment<DstXprType, SparseSymmetricPermutationProduct<MatrixType,Mode>, internal::assign_op<Scalar,typename MatrixType::Scalar>, Sparse2Sparse> |
| 633 | { |
| 634 | typedef SparseSymmetricPermutationProduct<MatrixType,Mode> SrcXprType; |
| 635 | typedef typename DstXprType::StorageIndex DstIndex; |
| 636 | template<int Options> |
| 637 | static void run(SparseMatrix<Scalar,Options,DstIndex> &dst, const SrcXprType &src, const internal::assign_op<Scalar,typename MatrixType::Scalar> &) |
| 638 | { |
| 639 | // internal::permute_symm_to_fullsymm<Mode>(m_matrix,_dest,m_perm.indices().data()); |
| 640 | SparseMatrix<Scalar,(Options&RowMajor)==RowMajor ? ColMajor : RowMajor, DstIndex> tmp; |
| 641 | internal::permute_symm_to_fullsymm<Mode>(src.matrix(),tmp,src.perm().indices().data()); |
| 642 | dst = tmp; |
| 643 | } |
| 644 | |
| 645 | template<typename DestType,unsigned int DestMode> |
| 646 | static void run(SparseSelfAdjointView<DestType,DestMode>& dst, const SrcXprType &src, const internal::assign_op<Scalar,typename MatrixType::Scalar> &) |
| 647 | { |
| 648 | internal::permute_symm_to_symm<Mode,DestMode>(src.matrix(),dst.matrix(),src.perm().indices().data()); |
| 649 | } |
| 650 | }; |
| 651 | |
| 652 | } // end namespace internal |
| 653 | |
| 654 | } // end namespace Eigen |
| 655 | |
| 656 | #endif // EIGEN_SPARSE_SELFADJOINTVIEW_H |
| 657 | |