Preconditioners for MLFMM

A few preconditioners are available for the multilevel fast multipole method (MLFMM).

Note: It is recommended to use the default settings. Changing the preconditioner is recommended only for advanced users.
On the Solve/Run tab, in the Solution Settings group, click the  Solver Settings icon. On the Solver Settings dialog, click the Advanced tab.


Figure 1. The Solver Settings dialog (Advanced tab).
Incomplete LU decomposition (ILU) (128)
Preconditioner for an MLFMM solution that uses an incomplete LU-decomposition of the matrix. Note, not available for parallel.
Multilevel FEM-MLFMM LU/diagonal decomposition (2010)
Preconditioner for a hybrid MLFMM / FEM solution that uses a multilevel sparse LU decomposition of the combined and partitioned system.

When using the parallel Solver, the factorisation type, which slightly impacts runtime and memory, can be specified.

Block-Jacobi preconditioner of MLFMM one-level-up (4096)
Preconditioner for an MLFMM solution that uses LU-decomposition on entries in the boxes, one-level-up from the finest boxes. Note, not available for parallel.
Sparse approximate inverse (SPAI) (8192)
Preconditioner for an MLFMM solution. This preconditioner uses less memory than the default in most cases, but runtime could be longer.
Tip: Switch to this preconditioner if the default one fails.
Sparse LU (8193)
Preconditioner for an MLFMM solution that uses a sparse LU decomposition of the MLFMM near field matrix.

When using the parallel Solver, the factorisation type, which slightly impacts runtime and memory, can be specified.

Note: The number in brackets corresponds to the value of the I3 field at the CG card in the model .pre file.