Assigns the block to the current local index. At the same time the local defect is calculated for the index and stored in the rhs. Afterwards the is incremented for the next block.
Assigns the block to the current local index. At the same time the local defect is calculated for the index and stored in the rhs. Afterwards the is incremented for the next block.
Assigns the block to the current local index. At the same time the local defect is calculated for the index and stored in the rhs. Afterwards the is incremented for the next block.
All of our Krylow solvers are preconditioned versions. There are sequential preconditioners (e,g. SeqJacobi, SeqSOR, SeqSSOR) as well as parallel preconditioners (e.g. AMG, BlockPreconditioner) available for plugging them into the solvers together with matching ScalarProducts.
Some of the available preconditioners (e.g. SeqJacobi, SeqSOR, SeqSSOR)) may be given an additional int as a template parameter, the block recursion level. These preconditioners can be used on block-recursive matrices with an arbitrary hierarchy depth (eg. BCRSMatrix<BCRSMatrix<FieldMatrix,n,m> > >. Given a block recursion level those preconditioners work as normal on the offdiagonal blocks, treating them as traditional matrix entries. For the diagonal values a special procedure applies: If the diagonal is treated as a matrix itself and the preconditioner is applied recursively on the matrix representing the diagonal value with block level . For the case that the diagonal is treated as a matrix entry resulting in a linear solve or an identity operation depending on the algorithm.
Apply one step of the preconditioner to the system A(v)=d.
On entry v=0 and d=b-A(x) (although this might not be computed in that way. On exit v contains the update, i.e one step computes where is the approximate inverse of the operator characterizing the preconditioner.
Apply one step of the preconditioner to the system A(v)=d.
On entry v=0 and d=b-A(x) (although this might not be computed in that way. On exit v contains the update, i.e one step computes where is the approximate inverse of the operator characterizing the preconditioner.
Assigns the block to the current local index. At the same time the local defect is calculated for the index and stored in the rhs. Afterwards the is incremented for the next block.
Assigns the block to the current local index. At the same time the local defect is calculated for the index and stored in the rhs. Afterwards the is incremented for the next block.
Assigns the block to the current local index. At the same time the local defect is calculated for the index and stored in the rhs. Afterwards the is incremented for the next block.
If true the decomposition of the exact local solvers is computed on the fly for each subdomain and iteration step. If false all decompositions are computed in pre and only forward and backward substitution takes place in the iteration steps.
Array of sets of rowindices belonging to an overlapping subdomain
relaxationFactor
relaxation factor
onTheFly_
If true the decomposition of the exact local solvers is computed on the fly for each subdomain and iteration step. If false all decompositions are computed in pre and only forward and backward substitution takes place in the iteration steps.
Warning
Each rowindex should be part of at least one subdomain!