Add DMDA SNES local Jacobian support, does not work for rediscretized coarse levels; removed old junk from example

Hg-commit: c5e2e9b14c81fedb8c1dae1e0e3b686f5ee813ba

merge SNES/DM refactor

Hg-commit: 31d66a4917111e3b9e87bf19170c2c9421b4e8be

Refactor SNES DM infrastructure

The residual and Jacobian function pointers are now stored in a managed
context on the DM. Coarsening of this context is handled by a
DMCoarsenHook, so the coarse lev

Refactor SNES DM infrastructure

The residual and Jacobian function pointers are now stored in a managed
context on the DM. Coarsening of this context is handled by a
DMCoarsenHook, so the coarse level DM automatically has function
pointers for residual evaluation. The SNESDM context (attached to a DM)
has copy-on-write semantics for all DMs other than the one it was
originally created on. That means that changes to the function pointers
from the fine level are automatically propagated unless a coarse level
is changed manually.

Hg-commit: ce0066853e1ef4402603a33070cb3bfd75c87f86

show more ...

Changed the FormLocalJacobian() signature for DMComplex
- Now it passes the preconditioning matrix too
- I know we talked about using a different DM for that, but it is not ready yet
- Added code for

Changed the FormLocalJacobian() signature for DMComplex
- Now it passes the preconditioning matrix too
- I know we talked about using a different DM for that, but it is not ready yet
- Added code for computing the Jacobian MF, not tested

Hg-commit: dea97b5339e02d2153ed14e5fb4f4fbd3b23fc63

show more ...

Remove DMMGSetInitialGuessLocal()

There was no way to set a context for this function, no way to call it
without strange casts, and no way to set the function pointer without
reaching into private d

Remove DMMGSetInitialGuessLocal()

There was no way to set a context for this function, no way to call it
without strange casts, and no way to set the function pointer without
reaching into private data structures.

Hg-commit: 1d2de3df26a7ae8b42b8b5bd57018ff4d764b4cc

show more ...

src/snes/utils/sieve is removed - so update makefile

Hg-commit: 68e7001c0ef2f7baa13b5c468882a859d273208f

Lots of work on DMComplex
- Added PetscSectionCreateGlobalSection()
- This makes a Section which has global offsets, negative for ghosts
- Now negative sizes are not counted for PetscSectionGetSt

Lots of work on DMComplex
- Added PetscSectionCreateGlobalSection()
- This makes a Section which has global offsets, negative for ghosts
- Now negative sizes are not counted for PetscSectionGetStorageSize()
- Added DMPrintCellVector() and DMPrintCellMatrix()
- Added PetscQuadrature structure
- Added SNESDMComplexComputeFunction/Jacobian()

Hg-commit: 93e35c5fb2848c5fd712d4bd760214daf5e64247

show more ...

Removed dead code

Hg-commit: 73367a9d7a7bfa65a1015cae713985aaf385454b

merge release-3.2

Hg-commit: 92ecc738af2e54343bf34f9c473146d1479f768d

Fix a few format strings

Hg-commit: 71f429175a71a799d01a1cefdb2806fbcdd7ae81

better documentation for using AMS with Java and options
fixed two errors in using PetscOptionsXXX() when working with interactive options gui

Hg-commit: 89bc248f9523eb90041707320a4018e521612d7e

added -da_refine allowing mimicing -dmmg_nlevels without DMMG
PCML removes the DM object so that PCMG won't try to use it to generate restriction/interpolation
PCML marks itself as using Galerkin coa

added -da_refine allowing mimicing -dmmg_nlevels without DMMG
PCML removes the DM object so that PCMG won't try to use it to generate restriction/interpolation
PCML marks itself as using Galerkin coarse grid corrections so that -ksp_view shows it correctly.

Hg-commit: 1a4947988e50cdb8d6e314af4e8a4e969913512d

show more ...

Better error message for SNESDAFormFunction without calling SNESSetDM()

Hg-commit: 22555ec12e083037ca3a5c757e9b8622e1383725

basic -snes_grid_sequence <levels> is working for ex57.c needs work before will work with VI solvers and not yet determined how to get it working cleanly with multigrid for inner solver

Hg-commit: 9

basic -snes_grid_sequence <levels> is working for ex57.c needs work before will work with VI solvers and not yet determined how to get it working cleanly with multigrid for inner solver

Hg-commit: 963cdbcd2b6c84fb51974286fdf549257ab56035

show more ...

fixed PCReset_FieldSplit() to work in certain cases
changed SNESVI solver to again use PCReset() instead of making a new KSP each time
added ex58.c to allow start of testing of geometric multigrid fo

fixed PCReset_FieldSplit() to work in certain cases
changed SNESVI solver to again use PCReset() instead of making a new KSP each time
added ex58.c to allow start of testing of geometric multigrid for SNESVI solver

Hg-commit: 65bdb5e3e7bc2fd1d7345aee42689e28d0077897

show more ...

Removed last vestiges of C++ from ex12
- Remove some Mesh stuff from SNES

Hg-commit: 46bd5e36ae04339f8caac15e1128066c061bb6ba

ex12 now works in one case, but relies on a hack
- ex12 will not work in general because global Vecs do not store boundary values
- Initialize PetscSection correctly
- Small fix to L2G mapping

Hg-co

ex12 now works in one case, but relies on a hack
- ex12 will not work in general because global Vecs do not store boundary values
- Initialize PetscSection correctly
- Small fix to L2G mapping

Hg-commit: 6d085f306946b98152a6f2ab4ebae937918a0bfd

show more ...

sieve specific code got put in without #ifdef protection

Hg-commit: 79c83b802a52f5e9992e50ba9f24852222af72ed

Added SNES ex12 to solve Bratu on an unstructured mesh using FEM
- Not completely working yet (and some cruft)
- Added SNESMeshFormFunction()/Jacobian()
- Added PetscSection to encode unstructured da

Added SNES ex12 to solve Bratu on an unstructured mesh using FEM
- Not completely working yet (and some cruft)
- Added SNESMeshFormFunction()/Jacobian()
- Added PetscSection to encode unstructured data distributions
- Added new support functions for Mesh
- Added InsertModes for BC

Hg-commit: ca869cc893afde5164c99fa2051bdd35be4e6133

show more ...

Mass revert file modes that were made executable in f6e61ee9c329

Hg-commit: 116d90d0b66125cee0db3e7351047057e756baf5