smith::BlockSchurPreconditioner Class Reference

Simple 2x2 block Schur complement preconditioner for block systems. More...

#include <block_preconditioner.hpp>

Collaboration diagram for smith::BlockSchurPreconditioner:

Public Member Functions
	BlockSchurPreconditioner (mfem::Array< int > &offsets, std::vector< std::unique_ptr< mfem::Solver >> solvers, BlockSchurType type=BlockSchurType::Diagonal, SchurApproxType approxType=SchurApproxType::DiagInv, std::vector< BlockOverride > overrides={})
	Construct a new 2x2 block Schur complement preconditioner. More...
virtual void	Mult (const mfem::Vector &in, mfem::Vector &out) const
	The action of the precondition on the block vector (b_1, b_2) More...
virtual void	SetOperator (const mfem::Operator &jacobian)
	Set the preconditioner to use the supplied linearized block Jacobian. More...

Detailed Description

Simple 2x2 block Schur complement preconditioner for block systems.

Uses two solvers, one for $ A_{11} $ and one for an approximate Schur complement $ S $. Call SetOperator() with an mfem::BlockOperator, then use Mult() to apply the selected Schur preconditioner type.

Definition at line 200 of file block_preconditioner.hpp.

Constructor & Destructor Documentation

BlockSchurPreconditioner()

smith::BlockSchurPreconditioner::BlockSchurPreconditioner	(	mfem::Array< int > &	offsets,
		std::vector< std::unique_ptr< mfem::Solver >>	solvers,
		BlockSchurType	type = BlockSchurType::Diagonal,
		SchurApproxType	approxType = SchurApproxType::DiagInv,
		std::vector< BlockOverride >	overrides = {}
	)

Construct a new 2x2 block Schur complement preconditioner.

Parameters

offsets	Offsets describing the 2-block layout.
solvers	Two solvers, for $ A_{11} $ and the Schur complement approximation.
type	Preconditioner variant (diagonal, lower, upper, or full).
approxType	Schur complement approximation strategy for the (1,1) block.
overrides	Optional list of (block index, operator) pairs used in place of the corresponding Jacobian diagonal block. For Schur systems, index 0 overrides $A_{11}$ and index 1 provides a custom Schur operator when approxType is SchurApproxType::Custom.

Definition at line 233 of file block_preconditioner.cpp.

Member Function Documentation

Mult()

void smith::BlockSchurPreconditioner::Mult ( const mfem::Vector &  in, mfem::Vector &  out  ) const

virtual

The action of the precondition on the block vector (b_1, b_2)

Parameters

in	The block input vector (b_1, b_2)
out	The block output vector P^-1(b_1, b_2)

Definition at line 304 of file block_preconditioner.cpp.

SetOperator()

void smith::BlockSchurPreconditioner::SetOperator ( const mfem::Operator &  jacobian )

virtual

Set the preconditioner to use the supplied linearized block Jacobian.

The Schur complement approximation is given by S_approx = A22 - A21 * diag(A11)^{-1} * A12

Parameters

jacobian

The supplied linearized Jacobian. Note that it is always a block operator

Definition at line 393 of file block_preconditioner.cpp.

---

The documentation for this class was generated from the following files:

• src/smith/numerics/block_preconditioner.hpp
• src/smith/numerics/block_preconditioner.cpp