Backed out changeset 5a9ebf885615. Now PCFIELSPLIT Schur attaches A11's nullspace to S.Hg-commit: 1723d4624521b3378f92ed1e7cc618e9fed1f549
Check for fieldsplit blocks' constant nullspace directly and avoid attaching A11's nullspace to S.Hg-commit: 5a9ebf885615c6be8b56b4ed082d32dcd538285e
Correct split set up for fieldsplit type schur, including tabbing, prefixes and DM forwarding.Hg-commit: 943dde820f7f44163d29c1e93687cb3ebdbe8d8d
new fortran interfaceHg-commit: 4a6638305aacc052919d0dcea60bf9ed7e619628
added PCFieldSplitGetSubKSP() for FortranHg-commit: 40f9515d8b016d37466aebf307f39996fe7288c9
Now the inner A solve in S = C^T A^{-1} B by default is the same KSP as the (0,0) block. If you give any options with -fieldsplit_1_inner_* (or <split name>_inner_*) then it creates a new KSP with th
Now the inner A solve in S = C^T A^{-1} B by default is the same KSP as the (0,0) block. If you give any options with -fieldsplit_1_inner_* (or <split name>_inner_*) then it creates a new KSP with that prefix- Added MatSchurComplementSetKSP()- Added PetscOptionsFindPairPrefix_Private()Hg-commit: 780f6f61e3b8a31069a311f5f8e3c99c14add8a8
show more ...
Use KSPIncrementTabLevel() to set tabbing on the PC as well.Hg-commit: 39dae4a883c70c8621ff35a63f76df22c2f556f3
backout all chmod stuff in bf971ba58125[but keep diffs to src/mat/impls/aij/mpi/mpimatmatmult.c ]Hg-commit: 4a812fc1f35496ce2717c7da9ef12ffabbd13258
removed a spurious call to MatSetBlockSize.Hg-commit: bf971ba58125bd8cf7ceb2a9d0f5cdf27c27cc6a
added -valgrind -debugger option to petscmpiexecHg-commit: 398f06830ac9d0920f8d4e4d0bac972857d520d4
DMComplex: Fix for field constraints- Now field constraints are numbered relative to the field, not the whole point- Commented out debugging printHg-commit: 05c248dec7e38814ede96e40affc7d714dd080
DMComplex: Fix for field constraints- Now field constraints are numbered relative to the field, not the whole point- Commented out debugging printHg-commit: 05c248dec7e38814ede96e40affc7d714dd080e4
Now FieldSplit can be configure for nested splits from the command line: -pc_type fieldsplit -pc_fieldsplit_0_fields 0,1 -pc_fieldsplit_1_fields 2makes splits for Stokes+Temperature in SNES ex31
Now FieldSplit can be configure for nested splits from the command line: -pc_type fieldsplit -pc_fieldsplit_0_fields 0,1 -pc_fieldsplit_1_fields 2makes splits for Stokes+Temperature in SNES ex31Hg-commit: 0d4ccb990bb888c029826f39adc2fe0dcd82d4de
Split DMCreateDecomposition() into Field and Domain decompositions. Reenabled use of DM in ASM.Hg-commit: 2e2ee6a01d1a543670ea1a565999371852551596
Splitting up DM decomposition into field and subdomain decompositions.Hg-commit: 25ae1465cd37db36ff0f764d65ffd3f1bf1fc0cc
FieldSPlit Change: Now preconditioning matrices can be attached to a FS IS- Better FS viewing- Fixed bug in FS MatSchurComplementUpdate()- Allow preconditioning Mats to be passed to DMMesh createF
FieldSPlit Change: Now preconditioning matrices can be attached to a FS IS- Better FS viewing- Fixed bug in FS MatSchurComplementUpdate()- Allow preconditioning Mats to be passed to DMMesh createFieldSplitHg-commit: e65437940315da9d5a1a76d44d1731b9cf8f9749
Added PCFieldSplitGetType()Hg-commit: 626a1fef5c74312e7f2b9637639c1ce41736ed88
Better doc for FieldSplitHg-commit: 7af37839b074ff072a1adb1116c72fb2a58f334b
fix fieldsplit indexing bug on resetHg-commit: 769fda0d533de7c6e032237e398671fef0a09e90
FieldSplit should forward DMs and prefixes to Schur solvers, use correct tab levels on their PCs, and handle reset correctly.Hg-commit: a36eb42b26eeec4cdd64c76a892badd1985fede2
Doc fix for -pc_fieldsplit_schur_precondition <self,user,diag>, repair some formattingHg-commit: 29b351f18c2eb014490c1350a75b2d55b7f22a4c
FIELDSPLIT now looks for near null spaces in the IS defining the splitDMMesh constructFieldSplit() now takes near null spaces and sets then in the ISHg-commit: 2ad289ac99e0db0a970c4d51fcd91b9580e7
FIELDSPLIT now looks for near null spaces in the IS defining the splitDMMesh constructFieldSplit() now takes near null spaces and sets then in the ISHg-commit: 2ad289ac99e0db0a970c4d51fcd91b9580e7b934
norm is a real number not a complexHg-commit: 1d36d068558494b2cba8ae03cc58e8ed8299fa27
Fixed FieldSplit to detect a constant null space in the blocks, now ex62 works fully- Normalized null space vector- Pushed down S^{-1} tolerance for ex62Hg-commit: 7d9b5a448dfeae0624c1566357e15f6
Fixed FieldSplit to detect a constant null space in the blocks, now ex62 works fully- Normalized null space vector- Pushed down S^{-1} tolerance for ex62Hg-commit: 7d9b5a448dfeae0624c1566357e15f674f896fd6
PetscTypeCompare -> PetscObjectTypeCompareHg-commit: 3fc8f665421ddb862bf4d1ba95505378246340be
more ex50 snes non-DMMG test runs matching old ex19 test runs including one for BAIJ matricesHg-commit: f83b3b6c6fe6387792b93845fc602177b9a17b93
1...<<11121314151617181920>>...23