introduced another scalable symbolic matmatmult that does not use bitarray
don't know how all these other files got changed

Hg-commit: 9cf1fa173221b7baf45abd24f9c6760aa244faf0

Rename DMGetMatrix() to DMCreateMatrix()

Hg-commit: 045e13fbce4683d1c6e7208d70318a9e4dde5c0b

Remove stored slice->globalvector, always create it in DMCreateGlobalVector_Sliced

Hg-commit: 049f44cc37c1d42cd3e753f4e2fa30275145309b

DMSliced needed to compose a DM for DMVecGetArray() to function correctly

Hg-commit: a2500e2364ebaf07eea564424ab2d21610a6f27f

commit after merge, some manual merges

Hg-commit: a60db41ebcbbb90ab0d02727666a69b6d3ee88a0

fix 0fb0bab0f904 to use new SUBMANSEC
[MANSEC for manulapages SUBMANSEC -if it exists - for generatefortranstubs]

Hg-commit: 30b6565f23703ed83e6a6b5642c6c0823387fe48

Fix MANSEC in DM

Hg-commit: 0fb0bab0f9048f45fa9db15d944449946b222ec4

changed examples to pass pointer to destroy function
most work

Hg-commit: bbc5d1464c4cbe6ea83fdf46e15c0f41ea396c06

converted rest of source code to take pointers into destructor reoutines. make test workds

Hg-commit: 78c641225c0516099423eeea3cc66e68fd2f62f3

commit after terrible merge

Hg-commit: 6ee623422368621da3dca005325b05a2fcd6d6bb

partcially completed making PETSc object destructors take & for the argument

Hg-commit: e5e776c3749e6dc0b2c108f7f84f10a7aaba55f5

SNES ex10 now runs in serial
- Added MatSetValuesTopology()
- DMGetMatrix_Mesh() now sets the local to global mapping
- We compose the Mesh with global/local vectors
- Added DMCreateLocalToGlobalMapp

SNES ex10 now runs in serial
- Added MatSetValuesTopology()
- DMGetMatrix_Mesh() now sets the local to global mapping
- We compose the Mesh with global/local vectors
- Added DMCreateLocalToGlobalMapping() to Mesh
- Added matrix filling to DMMeshCreateMatrix()
- Fixed DMDestroy() for all subclasses
- Moved DMDASetMatrPreallocateOnly() to DMSetMatrixPreallocateOnly()
and changed option to -dm_preallocate_only
- Fixed test output

Hg-commit: 75eb7f4d3ab7b8fbfedb6b44954aff1021a116c1

show more ...