preconditionneur_petsc_deriv

Preconditioners available with petsc solvers

preconditionneur_petsc_block_jacobi_icc

Synonyms: block_jacobi_icc

Inherits from: preconditionneur_petsc_deriv

Incomplete Cholesky factorization for symmetric matrix with the PETSc implementation.

Parameters:

[level] (type: int) factorization level (default value, 1). In parallel, the factorization is done by block (one per processor by default).
[ordering] (type: string into ['natural', 'rcm']) The ordering of the local matrix is natural by default, but rcm ordering, which reduces the bandwith of the local matrix, may interestingly improves the quality of the decomposition and reduces the number of iterations.

preconditionneur_petsc_block_jacobi_ilu

Synonyms: block_jacobi_ilu

Inherits from: preconditionneur_petsc_deriv

preconditionner

Parameters:

[level] (type: int) not_set

preconditionneur_petsc_boomeramg

Synonyms: boomeramg

Inherits from: preconditionneur_petsc_deriv

Multigrid preconditioner (no option is available yet, look at CLI command and Petsc documentation to try other options).

preconditionneur_petsc_c_amg

Synonyms: c-amg

Inherits from: preconditionneur_petsc_deriv

preconditionner

preconditionneur_petsc_diag

Synonyms: diag

Inherits from: preconditionneur_petsc_deriv

Diagonal (Jacobi) preconditioner.

preconditionneur_petsc_eisentat

Synonyms: eisentat

Inherits from: preconditionneur_petsc_deriv

SSOR version with Eisenstat trick which reduces the number of computations and thus CPU cost...

Parameters:

[omega] (type: float) relaxation factor

preconditionneur_petsc_ilu_mumps

Synonyms: ilu_mumps

Inherits from: preconditionneur_petsc_deriv

Incomplete LU factorization with Block Low Ranking from the MUMPS library. Mapped at runtime onto Petsc's cholesky pc with mat_mumps_icntl_35=1 (BLR enabled).

Parameters: