-ksp_gmres_restart <restart, default=30> | - see KSPGMRES | |
-ksp_hpddm_type <type, default=gmres> | - any of gmres, bgmres, cg, bcg, gcrodr, bgcrodr, bfbcg, or preonly, see KSPHPDDMType | |
-ksp_hpddm_deflation_tol <eps, default=-1.0> | - tolerance when deflating right-hand sides inside block methods (no deflation by default, only relevant with block methods) | |
-ksp_hpddm_enlarge_krylov_subspace <p, default=1> | - split the initial right-hand side into multiple vectors (only relevant with nonblock methods) | |
-ksp_hpddm_orthogonalization <type, default=cgs> | - any of cgs or mgs, see KSPGMRES | |
-ksp_hpddm_qr <type, default=cholqr> | - distributed QR factorizations with any of cholqr, cgs, or mgs (only relevant with block methods) | |
-ksp_hpddm_variant <type, default=left> | - any of left, right, or flexible (this option is superseded by KSPSetPCSide()) | |
-ksp_hpddm_recycle <n, default=0> | - number of harmonic Ritz vectors to compute (only relevant with GCRODR or BGCRODR) | |
-ksp_hpddm_recycle_target <type, default=SM> | - criterion to select harmonic Ritz vectors using either SM, LM, SR, LR, SI, or LI (only relevant with GCRODR or BGCRODR). For BGCRODR, if PETSc is compiled with SLEPc, this option is not relevant, since SLEPc is used instead. Options are set with the prefix -ksp_hpddm_recycle_eps_ | |
-ksp_hpddm_recycle_strategy <type, default=A> | - generalized eigenvalue problem A or B to solve for recycling (only relevant with flexible GCRODR or BGCRODR) | |
-ksp_hpddm_recycle_symmetric <true, default=false> | - symmetric generalized eigenproblems in BGCRODR, useful to switch to distributed solvers like EPSELEMENTAL (only relevant when PETSc is compiled with SLEPc) |
1980 | - The Block Conjugate Gradient Algorithm and Related Methods. O'Leary. Linear Algebra and its Applications. | |
2006 | - Recycling Krylov Subspaces for Sequences of Linear Systems. Parks, de Sturler, Mackey, Johnson, and Maiti. SIAM Journal on Scientific Computing | |
2013 | - A Modified Block Flexible GMRES Method with Deflation at Each Iteration for the Solution of Non-Hermitian Linear Systems with Multiple Right-Hand Sides. Calandra, Gratton, Lago, Vasseur, and Carvalho. SIAM Journal on Scientific Computing. | |
2016 | - Block Iterative Methods and Recycling for Improved Scalability of Linear Solvers. Jolivet and Tournier. SC16. | |
2017 | - A breakdown-free block conjugate gradient method. Ji and Li. BIT Numerical Mathematics. |