add MatMultTransColoringApplyDenToSp()

Hg-commit: 55f04facce91feb6114badef2b92d0e50ded7ae3

Fix MatInvertBlockDiagonal_SeqAIJ, the old code produced the transpose so it was worthless for nonsymmetric problems

Hg-commit: 0e6ed694016804d0acbdbcfb01574005ffa20378

add MatMultTransposeColoringApply()

Hg-commit: d828d4c5d7928c859bac4fe88e9190babd4c122b

add MatMultTransposeColoringCreate() and MatMultTransposeColoringDestroy()

Hg-commit: 07822c8102acbf3d860363a2e0292b9f4bdf6afc

Move an=A->cmap->n into USE_ARRAY block to silence unused variable warning (since USE_ARRAY is off by default)

Hg-commit: 46a78ae510706d58ac21475a4c8d0c87f2817d2b

add array implementation for sparse dot product - slower :-(

Hg-commit: 4808c53e08bf424152591d45922351d2b3278e73

implement MatMatMultTransposeNumeric_SeqAIJ_SeqAIJ()

Hg-commit: 5f7c3284a881788d8b85f388a59ed05989359b47

fixed a few PetscOptionsGetInt(PETSC_NULL, to use object prefix

Hg-commit: 18077e155b608bf1c6379b862b95c69ac5c98077

implement MatMatMultTransposeSymbolic_SeqAIJ_SeqAIJ()

Hg-commit: 764b5e7101e34400970de155641432cfe097af59

rename MatMatTransposeMult() for A^T * B; MatMatMultTranspose() for A * B^T (not implemented yet)

Hg-commit: a142087cf282822ab67182371e8b850cbc565e08

changed MatDuplicate_SeqAIJ() to handle CUSP case

Hg-commit: 21d153a26151c2657ffec0ae380feacb96bf10cf

minor cleanup MatPtAPNumeric_SeqAIJ_SeqAIJ()

Hg-commit: cd73944d83a5093aa65010455630fa59740ac818

better fix for MatAXPY() updating gpu flag for changed matrix (also MatShift())

Hg-commit: 2ddd8efc3f35f1b73a00ba4ade4ddde9b3697fc4

updated AIJCUSP matrix so that MatAXPY() results in reloading the matrix to the GPU because the GPU matrix is no longer valid

Hg-commit: 4a78d299751875f3374d33efba517927757445e3

commit after merge of patch for 3.2 for pbjacobi preconditioning

Hg-commit: ac19ef62f9a4f3b94290ce733a2eecc95134aec2

fixes for txgpupetsc so it does not turn off MatMult_SeqAIJCUSP without inodes

Hg-commit: 317794c7c85893baf63fec41155938ffc7c802e0

fixed a giant array to be malloc-ed in my triple function

Hg-commit: 61119cb1a21c1d2a094ef9fe3a2d1150b8f102fc

removed an extra var from CPU triple construction

Hg-commit: ac95e47cf2df44ed9c78458fe72c72570b828ee5

Turn off Fortran for the new matrix constructor

Hg-commit: bcce49f070eb0449ec5b49aca7407856ed240029

finished initial implementation of MatCreateSeqAIJFromTriple.

Hg-commit: f0020c3688658a43c266938cf7f99a35cd163649

added base for MatCreateSeqAIJFromTriple

Hg-commit: 83c792bacc2ee63fd4e1d48fffc5b81ee9dfb32a

merge from 3.2

Hg-commit: 136b1a81df3a1c609acad95fbce77ebeead8502d

commit after merge

Hg-commit: f3ef751a6e049b1f35fb61c484c44e4d842f3a24

added KSPDefaultConvergedDestroy() for Fortran

Hg-commit: f55cb2ae1785272de91118d0c4cad25a7adf3942

Pthreadclasses run on Apple now.

Hg-commit: ce806d3ce874ef1a7dbe61011710bda3cb40d10e