KSPFCG

Implements the Flexible Conjugate Gradient method (FCG) [Not00], [AV91]. Unlike most KSP methods this allows the preconditioner to be nonlinear. Flexible Krylov Methods

Options Database Keys

• -ksp_fcg_mmax - maximum number of search directions, similar to the restart in KSPGMRES and KSPFGMRES

• -ksp_fcg_nprealloc - number of directions to preallocate

• -ksp_fcg_truncation_type <standard,notay> - truncation approach for directions

Notes

KSPFCG requires the matrix to be symmetric positive-definite (SPD); for non-SPD problems use KSPFGMRES or KSPGCR.

Supports left preconditioning only.

Contributed by

Patrick Sanan

References

[AV91]

O. Axelsson and P. S. Vassilevski. A black box generalized Conjugate Gradient solver with inner iterations and variable-step preconditioning. SIAM Journal on Matrix Analysis and Applications, 12(4):625–644, 1991.

[Not00]

Yvan Notay. Flexible conjugate gradients. SIAM Journal on Scientific Computing, 22(4):1444–1460, 2000.

See Also

KSP: Linear System Solvers, Flexible Krylov Methods, KSPGCR, KSPPIPEGCR, KSPPIPEFCG, KSPFGMRES, KSPCG, KSPFCGSetMmax(), KSPFCGGetMmax(), KSPFCGSetNprealloc(), KSPFCGGetNprealloc(), KSPFCGSetTruncationType(), KSPFCGGetTruncationType(), KSPFCDTruncationType

Level

beginner

Location

src/ksp/ksp/impls/fcg/fcg.c

Examples

src/ksp/ksp/tutorials/ex7.c

---

Index of all KSP routines
Table of Contents for all manual pages
Index of all manual pages