template<class ScalarType, class MV, class OP, const bool supportsScalarType = Belos::Details::LapackSupportsScalar<ScalarType>::value && ! Teuchos::ScalarTraits<ScalarType>::isComplex>
class Belos::PCPGSolMgr< ScalarType, MV, OP, supportsScalarType >
PCPG iterative linear solver.
- Author
- David Day
PCPG is a CG-based "seed solver." This means that it does preconditioned CG to build up a matrix polynomial, then can reuse that polynomial to compute solutions of successive linear systems, possibly with different right-hand sides. Belos also implements a Block GMRES - based seed solver, Belos::GmresPolySolMgr.
Users must ensure that each linear system has the same coefficient matrix. The seed space is invariant during an individual linear system solve. Finally, due to finite precision arithmetic, the off-diaognal "P'AP" terms grow.
One often sees PCPG in context with the FETI domain decomposition method.
Definition at line 155 of file BelosPCPGSolMgr.hpp.