00001
00002
00003
00004
00005
00006
00007
00008
00009
00010
00011 #ifndef EIGEN_GMRES_H
00012 #define EIGEN_GMRES_H
00013
00014 namespace Eigen {
00015
00016 namespace internal {
00017
00055 template<typename MatrixType, typename Rhs, typename Dest, typename Preconditioner>
00056 bool gmres(const MatrixType & mat, const Rhs & rhs, Dest & x, const Preconditioner & precond,
00057 int &iters, const int &restart, typename Dest::RealScalar & tol_error) {
00058
00059 using std::sqrt;
00060 using std::abs;
00061
00062 typedef typename Dest::RealScalar RealScalar;
00063 typedef typename Dest::Scalar Scalar;
00064 typedef Matrix < RealScalar, Dynamic, 1 > RealVectorType;
00065 typedef Matrix < Scalar, Dynamic, 1 > VectorType;
00066 typedef Matrix < Scalar, Dynamic, Dynamic > FMatrixType;
00067
00068 RealScalar tol = tol_error;
00069 const int maxIters = iters;
00070 iters = 0;
00071
00072 const int m = mat.rows();
00073
00074 VectorType p0 = rhs - mat*x;
00075 VectorType r0 = precond.solve(p0);
00076
00077
00078 VectorType w = VectorType::Zero(restart + 1);
00079
00080 FMatrixType H = FMatrixType::Zero(m, restart + 1);
00081 VectorType tau = VectorType::Zero(restart + 1);
00082 std::vector < JacobiRotation < Scalar > > G(restart);
00083
00084
00085 VectorType e;
00086 RealScalar beta;
00087 r0.makeHouseholder(e, tau.coeffRef(0), beta);
00088 w(0)=(Scalar) beta;
00089 H.bottomLeftCorner(m - 1, 1) = e;
00090
00091 for (int k = 1; k <= restart; ++k) {
00092
00093 ++iters;
00094
00095 VectorType v = VectorType::Unit(m, k - 1), workspace(m);
00096
00097
00098 for (int i = k - 1; i >= 0; --i) {
00099 v.tail(m - i).applyHouseholderOnTheLeft(H.col(i).tail(m - i - 1), tau.coeffRef(i), workspace.data());
00100 }
00101
00102
00103 VectorType t=mat*v;
00104 v=precond.solve(t);
00105
00106
00107 for (int i = 0; i < k; ++i) {
00108 v.tail(m - i).applyHouseholderOnTheLeft(H.col(i).tail(m - i - 1), tau.coeffRef(i), workspace.data());
00109 }
00110
00111 if (v.tail(m - k).norm() != 0.0) {
00112
00113 if (k <= restart) {
00114
00115
00116 VectorType e(m - k - 1);
00117 RealScalar beta;
00118 v.tail(m - k).makeHouseholder(e, tau.coeffRef(k), beta);
00119 H.col(k).tail(m - k - 1) = e;
00120
00121
00122 v.tail(m - k).applyHouseholderOnTheLeft(H.col(k).tail(m - k - 1), tau.coeffRef(k), workspace.data());
00123
00124 }
00125 }
00126
00127 if (k > 1) {
00128 for (int i = 0; i < k - 1; ++i) {
00129
00130 v.applyOnTheLeft(i, i + 1, G[i].adjoint());
00131 }
00132 }
00133
00134 if (k<m && v(k) != (Scalar) 0) {
00135
00136 G[k - 1].makeGivens(v(k - 1), v(k));
00137
00138
00139 v.applyOnTheLeft(k - 1, k, G[k - 1].adjoint());
00140 w.applyOnTheLeft(k - 1, k, G[k - 1].adjoint());
00141
00142 }
00143
00144
00145 H.col(k - 1).head(k) = v.head(k);
00146
00147 bool stop=(k==m || abs(w(k)) < tol || iters == maxIters);
00148
00149 if (stop || k == restart) {
00150
00151
00152 VectorType y = w.head(k);
00153 H.topLeftCorner(k, k).template triangularView < Eigen::Upper > ().solveInPlace(y);
00154
00155
00156 VectorType x_new = y(k - 1) * VectorType::Unit(m, k - 1);
00157
00158
00159 x_new.tail(m - k + 1).applyHouseholderOnTheLeft(H.col(k - 1).tail(m - k), tau.coeffRef(k - 1), workspace.data());
00160
00161 for (int i = k - 2; i >= 0; --i) {
00162 x_new += y(i) * VectorType::Unit(m, i);
00163
00164 x_new.tail(m - i).applyHouseholderOnTheLeft(H.col(i).tail(m - i - 1), tau.coeffRef(i), workspace.data());
00165 }
00166
00167 x += x_new;
00168
00169 if (stop) {
00170 return true;
00171 } else {
00172 k=0;
00173
00174
00175 VectorType p0=mat*x;
00176 VectorType p1=precond.solve(p0);
00177 r0 = rhs - p1;
00178
00179 w = VectorType::Zero(restart + 1);
00180 H = FMatrixType::Zero(m, restart + 1);
00181 tau = VectorType::Zero(restart + 1);
00182
00183
00184 RealScalar beta;
00185 r0.makeHouseholder(e, tau.coeffRef(0), beta);
00186 w(0)=(Scalar) beta;
00187 H.bottomLeftCorner(m - 1, 1) = e;
00188
00189 }
00190
00191 }
00192
00193
00194
00195 }
00196
00197 return false;
00198
00199 }
00200
00201 }
00202
00203 template< typename _MatrixType,
00204 typename _Preconditioner = DiagonalPreconditioner<typename _MatrixType::Scalar> >
00205 class GMRES;
00206
00207 namespace internal {
00208
00209 template< typename _MatrixType, typename _Preconditioner>
00210 struct traits<GMRES<_MatrixType,_Preconditioner> >
00211 {
00212 typedef _MatrixType MatrixType;
00213 typedef _Preconditioner Preconditioner;
00214 };
00215
00216 }
00217
00263 template< typename _MatrixType, typename _Preconditioner>
00264 class GMRES : public IterativeSolverBase<GMRES<_MatrixType,_Preconditioner> >
00265 {
00266 typedef IterativeSolverBase<GMRES> Base;
00267 using Base::mp_matrix;
00268 using Base::m_error;
00269 using Base::m_iterations;
00270 using Base::m_info;
00271 using Base::m_isInitialized;
00272
00273 private:
00274 int m_restart;
00275
00276 public:
00277 typedef _MatrixType MatrixType;
00278 typedef typename MatrixType::Scalar Scalar;
00279 typedef typename MatrixType::Index Index;
00280 typedef typename MatrixType::RealScalar RealScalar;
00281 typedef _Preconditioner Preconditioner;
00282
00283 public:
00284
00286 GMRES() : Base(), m_restart(30) {}
00287
00298 GMRES(const MatrixType& A) : Base(A), m_restart(30) {}
00299
00300 ~GMRES() {}
00301
00304 int get_restart() { return m_restart; }
00305
00309 void set_restart(const int restart) { m_restart=restart; }
00310
00316 template<typename Rhs,typename Guess>
00317 inline const internal::solve_retval_with_guess<GMRES, Rhs, Guess>
00318 solveWithGuess(const MatrixBase<Rhs>& b, const Guess& x0) const
00319 {
00320 eigen_assert(m_isInitialized && "GMRES is not initialized.");
00321 eigen_assert(Base::rows()==b.rows()
00322 && "GMRES::solve(): invalid number of rows of the right hand side matrix b");
00323 return internal::solve_retval_with_guess
00324 <GMRES, Rhs, Guess>(*this, b.derived(), x0);
00325 }
00326
00328 template<typename Rhs,typename Dest>
00329 void _solveWithGuess(const Rhs& b, Dest& x) const
00330 {
00331 bool failed = false;
00332 for(int j=0; j<b.cols(); ++j)
00333 {
00334 m_iterations = Base::maxIterations();
00335 m_error = Base::m_tolerance;
00336
00337 typename Dest::ColXpr xj(x,j);
00338 if(!internal::gmres(*mp_matrix, b.col(j), xj, Base::m_preconditioner, m_iterations, m_restart, m_error))
00339 failed = true;
00340 }
00341 m_info = failed ? NumericalIssue
00342 : m_error <= Base::m_tolerance ? Success
00343 : NoConvergence;
00344 m_isInitialized = true;
00345 }
00346
00348 template<typename Rhs,typename Dest>
00349 void _solve(const Rhs& b, Dest& x) const
00350 {
00351 x.setZero();
00352 _solveWithGuess(b,x);
00353 }
00354
00355 protected:
00356
00357 };
00358
00359
00360 namespace internal {
00361
00362 template<typename _MatrixType, typename _Preconditioner, typename Rhs>
00363 struct solve_retval<GMRES<_MatrixType, _Preconditioner>, Rhs>
00364 : solve_retval_base<GMRES<_MatrixType, _Preconditioner>, Rhs>
00365 {
00366 typedef GMRES<_MatrixType, _Preconditioner> Dec;
00367 EIGEN_MAKE_SOLVE_HELPERS(Dec,Rhs)
00368
00369 template<typename Dest> void evalTo(Dest& dst) const
00370 {
00371 dec()._solve(rhs(),dst);
00372 }
00373 };
00374
00375 }
00376
00377 }
00378
00379 #endif // EIGEN_GMRES_H