Merge branch 'barry/feature-options-associated-with-object' [PR #1194]* barry/feature-options-associated-with-object:Complete support for individual PETSc objects to use their own options database
Merge branch 'barry/feature-options-associated-with-object' [PR #1194]* barry/feature-options-associated-with-object:Complete support for individual PETSc objects to use their own options database with PetscObjectSetOptions()
show more ...
Merge branch 'pr1197/lm/fix-pcpatch-options/master' [PR #1197]* pr1197/lm/fix-pcpatch-options/master:PCPatch: fix setFromOptions for !save_operatorsIf we do not save the patch operators, their cr
Merge branch 'pr1197/lm/fix-pcpatch-options/master' [PR #1197]* pr1197/lm/fix-pcpatch-options/master:PCPatch: fix setFromOptions for !save_operatorsIf we do not save the patch operators, their creation is deferred until PCPatch_Apply (rather than PCPath_Setup).As a result, we must also defer KSPSetFromOptions on each patch KSP until the operator has been attached.I think this won’t result in KSPSetFromOptions being called all the time, but we need to delay this call until the KSP has an operator, otherwise the defaults system produces nonsense.
PCPatch: correct logic for setFromOptions calls on subkspsAdditionally fix PCSetUpOnBlocks_PATCH for the !save_operators case.
PCPatch: fix setFromOptions for !save_operatorsIf we do not save the patch operators, their creation is deferreduntil PCPatch_Apply (rather than PCPath_Setup). As a result, we mustalso defer KSPS
PCPatch: fix setFromOptions for !save_operatorsIf we do not save the patch operators, their creation is deferreduntil PCPatch_Apply (rather than PCPath_Setup). As a result, we mustalso defer KSPSetFromOptions on each patch KSP until the operator hasbeen attached.
PCPatch: fix getGlobalDofs for !combinedThe offset is the sum of the offsets of all the fields I think. Thiscode path is never exercised.
PCPatch: fix patchSection setup for case !combinedIf the discretisation info comes in as one section per field, thelogic for computing a combined patchSection is different.
Complete support for individual PETSc objects to use their own options database with PetscObjectSetOptions()Commit-type: feature, exampleReported-by: "Blondel, Sophie" <sblondel@utk.edu>
Merge remote-tracking branch 'origin/stefano_zampini/weakly-symmetric-elasticity-rebased'
Fix for -pc_patch_exclude_subspaces code when the user hasn't yet setthe number of fields of the problem
Do some housekeeping and add missing CHKERRQ calls
C90 fixes
Generalise the exclude_subspace option of Vanka construction in PCPATCH to exclude_subspaces.This is important if you have more than one constraint, e.g. MHD problems withone multiplier enforcing
Generalise the exclude_subspace option of Vanka construction in PCPATCH to exclude_subspaces.This is important if you have more than one constraint, e.g. MHD problems withone multiplier enforcing solenoidality of the velocity and another enforcingsolenoidality of the magnetic field.
fix memleaks flagged by valgrind# > ==29645== 312 (40 direct, 272 indirect) bytes in 1 blocks are definitely lost in loss record 3 of 3# > ==29645== at 0x4C2CC70: calloc (in /usr/lib/valgrind/v
fix memleaks flagged by valgrind# > ==29645== 312 (40 direct, 272 indirect) bytes in 1 blocks are definitely lost in loss record 3 of 3# > ==29645== at 0x4C2CC70: calloc (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)# > ==29645== by 0x63009A0: kh_init_HSetI (hashseti.h:6)# > ==29645== by 0x63015BA: PetscHSetICreate (hashseti.h:6)# > ==29645== by 0x630F621: PCPatchCreateCellPatchDiscretisationInfo (pcpatch.c:857)# > ==29645== by 0x6317E86: PCSetUp_PATCH (pcpatch.c:1426)# > ==29645== by 0x632A345: PCSetUp (precon.c:932)# > ==29645== by 0x6475DDA: KSPSetUp (itfunc.c:391)# > ==29645== by 0x647985E: KSPSolve (itfunc.c:723)# > ==29645== by 0x661E5AA: SNESSolve_NEWTONLS (ls.c:224)# > ==29645== by 0x6596F9A: SNESSolve (snes.c:4394)# > ==29645== by 0x4139A7: main (ex62.c:676)
PCPatch: User patches and subspace exclusion are incompatible
PCPatch: Small fix
PCPatch: C89 fix
PC+Doc: Patch documentation fixes
PCPatch: Updated to new hash table interface
Small fix: if the construct type is set to something not USER or PYTHON,set patch->user_patches = PETSC_FALSE.
Some compatibility fixes for PCPATCH
PCPatch: Two people should not try to be clever at the same time- You cannot reuse a patch IS because the geometry info is being cached
Plex+PCPatch: Converted to IS input and Toby's geometry- This has broken all tests. Very depressing
PC: reinstate viewPatches code
PC: fix bc condensation (need to ignore condensed local dofs)
123456789