DM+Plex: Now classify cells by polytope type - Defined the polytope type of a cell. added names and special shortcutpointer like depth - Fixed DMLabelMakeValid() to optimize for contiguousruns of
DM+Plex: Now classify cells by polytope type - Defined the polytope type of a cell. added names and special shortcutpointer like depth - Fixed DMLabelMakeValid() to optimize for contiguousruns of points - Added cell polytope type label to DM - This needsspecial processing, like depth - Added DMPlexGetPointDepth(),DMPlexGetCellType(), DMPlexGetCellTypeLabel() - DMPlexComputeCellTypes()infers the cell type for all points using dim/depth/coneSize/ghostCells- DMPlexGetRawFace(Hybrid)_Internal() now takes the cell type. Thismeans we can unify the functions later. - Replaced switch(dim/coneSize)constructs with cell type
show more ...
Merge remote-tracking branch 'origin/maint'
Documentation fixes
Double CHKERRQ
Merge branch 'stefanozampini/update-suitesparse' into 'master'Update SuiteSparse to use github releases + fixes to SBAIJ hermitian complex caseSee merge request petsc/petsc!2305
PCBDDC: adjust conversions and fix solver type for sub_schursneed to get rid of these conversions somedayupdate outputs
PCBDDC: do not use MKL for SVD based constraints contractionReported-by: Pierre Jolivet <Pierre.Jolivet@enseeiht.fr>
Replace PetscInt with PetscMPIInt when returning a rank
PCBDDC: Fix valgrind issuehttps://gitlab.com/petsc/petsc/-/jobs/333447997not ok snes_tutorials-ex56_attach_mat_nearnullspace-1_bddc_approx_gamg # Error code: 10# ==22390== Invalid read of size 8
PCBDDC: Fix valgrind issuehttps://gitlab.com/petsc/petsc/-/jobs/333447997not ok snes_tutorials-ex56_attach_mat_nearnullspace-1_bddc_approx_gamg # Error code: 10# ==22390== Invalid read of size 8# ==22390== at 0x5F30EAD: PCSetData_AGG (agg.c:623)# ==22390== by 0x5F246DF: PCSetUp_GAMG (gamg.c:513)# ==22390== by 0x5FFF1D0: PCSetUp (precon.c:894)# ==22390== by 0x6116E23: KSPSetUp (itfunc.c:377)# ==22390== by 0x5E27EB5: PCBDDCSetUpLocalSolvers (bddcprivate.c:5547)# ==22390== by 0x5DBED6E: PCSetUp_BDDC (bddc.c:1758)# ==22390== by 0x5FFF1D0: PCSetUp (precon.c:894)# ==22390== by 0x6116E23: KSPSetUp (itfunc.c:377)# ==22390== by 0x6117B0A: KSPSolve (itfunc.c:707)# ==22390== by 0x623ACAF: SNESSolve_NEWTONLS (ls.c:225)# ==22390== by 0x61E01CB: SNESSolve (snes.c:4461)# ==22390== by 0x411A04: main (ex56.c:474)# ==22390== Address 0x11a8ad60 is 0 bytes inside a block of size 144 free'd# ==22390== at 0x4C2BDEC: free (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)# ==22390== by 0x4FD9792: PetscFreeAlign (mal.c:105)# ==22390== by 0x55B13A9: MatDestroy_SeqDense (dense.c:1469)# ==22390== by 0x540D881: MatDestroy (matrix.c:1271)# ==22390== by 0x5E2663A: MatNullSpacePropagateAny_Private (bddcprivate.c:5395)# ==22390== by 0x5E279D5: PCBDDCSetUpLocalSolvers (bddcprivate.c:5514)# ==22390== by 0x5DBED6E: PCSetUp_BDDC (bddc.c:1758)# ==22390== by 0x5FFF1D0: PCSetUp (precon.c:894)# ==22390== by 0x6116E23: KSPSetUp (itfunc.c:377)# ==22390== by 0x6117B0A: KSPSolve (itfunc.c:707)# ==22390== by 0x623ACAF: SNESSolve_NEWTONLS (ls.c:225)# ==22390== by 0x61E01CB: SNESSolve (snes.c:4461)# ==22390== by 0x411A04: main (ex56.c:474)
Merge branch 'jed/pcbddc-lumped' into 'master'Jed/pcbddc lumpedSee merge request petsc/petsc!1390
Bug fixes, code factorization, better coverage
PCBDDC: add lumped variant, aka M_1This variant, also known as M_1 in the literature, skips extension ofbalanced interface values to the interior by solving a Dirichletproblem. It gives up a fac
PCBDDC: add lumped variant, aka M_1This variant, also known as M_1 in the literature, skips extension ofbalanced interface values to the interior by solving a Dirichletproblem. It gives up a factor of H/h in condition number relative toDirichlet (standard) BDDC, but may be appropriate for smallsubdomains, inexact subdomain solvers, and/or as an ingredient incomposite preconditioners.Based-on-patch-by: Stefano Zampini <stefano.zampini@gmail.com>
KPS => KSP
PCBDDC: explicitly setup local KSPsso that we get a less distorted view in log_view
PCBDDC: directly copy values into deluxe matricesthis speeds up a bit the setup phase
Merge branch 'master' into 'master'Add --download-hpddmSee merge request petsc/petsc!1994
Fixes and improvements for PCHPDDM and KSPHPDDM
PCBDDC: speed up matrix assembly for sub_schurs by using VecScatter
PCBDDC: allow factoring and inverting local schur complements on GPUsRemove MatConvert -> I prefer to error if some matrix type is not supportedFew minor improvements to prevent unneeded data tran
PCBDDC: allow factoring and inverting local schur complements on GPUsRemove MatConvert -> I prefer to error if some matrix type is not supportedFew minor improvements to prevent unneeded data transfersFix old MKL dsygvx query routine
PCBDDCOrthonormalizeVecs: make number of vectors input-output
PCBDDC: added option to ignore nearnullspace for constraintsadded code to propagate nearnullspace to local matrices
PCBDDC: added a couple of events for approximate solvers
Some typos.
DMGetSection: replace all internal usage with DMGetLocalSectionSame for DMSetSection -> DMSetLocalSection. We retain the interfacefor DMGetSection/DMSetSection without deprecation warnings (for n
DMGetSection: replace all internal usage with DMGetLocalSectionSame for DMSetSection -> DMSetLocalSection. We retain the interfacefor DMGetSection/DMSetSection without deprecation warnings (for now).Suggested-by: Barry Smith <bsmith@mcs.anl.gov>
12345678910>>...53