Use the -snes_force_iteration option for VINEWTONRSLS methodThis option has not be systematically added for all the various convergence tests fordifferent SNES solversCommit-type: bug-fix, featu
Use the -snes_force_iteration option for VINEWTONRSLS methodThis option has not be systematically added for all the various convergence tests fordifferent SNES solversCommit-type: bug-fix, featureReported-by: Henrik Buesing <hbuesing@eonerc.rwth-aachen.de>
show more ...
SNES+Adaptor: Now -snes_adapt_initial can take a number of iterates- You can use this, along with a PetscDS exact solution, to do refinement with an analytic guide
Merge remote-tracking branch 'origin/Fande-Kong/remove_SNESHasDM'
Merge remote-tracking branch 'origin/knepley/feature-adaptor-plex' into next-tmp
SNES: SNESDestroy() can be called with snes->dm == NULL
Have SNES display if -snes_fd or -snes_fd_color are used to compute the JacobianCommit-type: featureReported-by: Ed Bueler <elbueler@alaska.edu>
Remove SNESHasDMThis may confuse users because PETSC always attach DM to SNES even thoughusers do not call SNESSetDM
Merge remote-tracking branch 'origin/barry/use-mpi-datatype-F90Array'
Merge branch 'maint'
Merge remote-tracking branch 'origin/jed/fix-dmcoarsenhookadd-identical' into maint
KSPSetDM, SNESSetDM, and TSSetDM: documentationA DM can only be used to solve one problem at a time because informationabout the problem is stored on the DM.Reported-by: Blaise A Bourdin <bourdi
KSPSetDM, SNESSetDM, and TSSetDM: documentationA DM can only be used to solve one problem at a time because informationabout the problem is stored on the DM.Reported-by: Blaise A Bourdin <bourdin@lsu.edu>
No need to distract users with this detail since there are colorings in parallelCommit-type: documentation
doc clarificationCommit-type: documentationReported-by: Mark Adams <mfadams@lbl.gov>
MATMFFD: added MatSetUp_MFFDmove layouts setup to MatSetUp_MFFDAdded missing checks for public API callsSNES: added missing checks for collective logical booleans in SNESSetUseMatrixFree()
F90Array routines now take a MPI_Datatype argument instead of PetscDataTypeThis is part of the refactorization to limit and maybe eventually eliminate PetscDataType sinceit duplicates MPI_Datatype
F90Array routines now take a MPI_Datatype argument instead of PetscDataTypeThis is part of the refactorization to limit and maybe eventually eliminate PetscDataType sinceit duplicates MPI_Datatype functionality (poorly).Commit-type: style-fixThanks-to: Jed Brown <jed@jedbrown.org>
DMCoarsenHookRemove: new interface to clean up stale hooksThe hooks are not automatically cleaned out when a DM outlives the SNESand is used for another solve, leading to a hook referencing anone
DMCoarsenHookRemove: new interface to clean up stale hooksThe hooks are not automatically cleaned out when a DM outlives the SNESand is used for another solve, leading to a hook referencing anonexistent SNES.Reported-by: Lawrence Mitchell <wencel@gmail.com>
SNES+ex12: Added support for ascii_info_detail in -snes_converged_reason- It prints the L_2 error using the DS information- SNES ex12 output updated
SNES: Added -snes_adapt_sequence- Eventually remove snes_grid_sequence
SNES: Add DMAdaptor call to SNES
fix: redundant CHKERRQ() call after SETERRQ()Craycc complained about a stray CHKERRQ after SETERRQ in plextree.cReported-by: Patrick Sanan <patrick.sanan@gmail.com>
-snes_fd did not work for some cases with TS and -pc_type mgThis was due to SNESComputeJacobianDefault() getting its work vectors from SNES instead ofthe DM (historical design).Commit-type: bug-
-snes_fd did not work for some cases with TS and -pc_type mgThis was due to SNESComputeJacobianDefault() getting its work vectors from SNES instead ofthe DM (historical design).Commit-type: bug-fixFunded-by: ECPTime: 1.3 hoursReported-by: Ed Bueler <elbueler@alaska.edu>
SNESDefaultMatrixFreeCreate2: add MatSetUp
Merge remote-tracking branch 'origin/barry/feature-snes-force-iteration'
1...<<11121314151617181920>>...74