Added disclaimer to Prometheus

Hg-commit: 061814d339d7cf06290600f9046ebb0358a2cc34

fixed mpiuni MPI_Comm_create
removed unneeded PetscCommDuplicate() from gamg.c
removed use numdiff in error checking

Hg-commit: 0a2aff779677c1dd6eba619a17823ebbfaec4d79

update mumps.py with scotch -> ptscotch change

Hg-commit: fceaf6822b5055283b586d434f47d8de736ae0c1

Fixed another bug Barry had put in because of a copy+paste error in metis.py (too many popLanguage() calls)

Hg-commit: b841d6d3f548b8e0c2f88747ffd8b34297a1d26d

Revert Barry's metis/parmetis changes in 6f723f11a7bd.

Both metis and parmetis have now been updated to not use the cxx variables. George had set this by default error in his makefile (neither packa

Revert Barry's metis/parmetis changes in 6f723f11a7bd.

Both metis and parmetis have now been updated to not use the cxx variables. George had set this by default error in his makefile (neither package needs C++ at all).

Hg-commit: c95339f32df1688f9ed2ba673f08d339a4afe48e

show more ...

commit after merge

Hg-commit: 02e72cd24f43c7653870c99ff8bf13bb9d393bdc

metis and parmetis need C++ compiler? WHY WHY WHY

Hg-commit: 6f723f11a7bd3bd95af5b5d1a56c55225d91f5c1

Remove cxx from parmetis.py. It doesn't need it and breaks builds that are C only

Hg-commit: fe7028a65da349e44622005dde8db9d47826833b

pass -D__thread="" to metis make on Apple so it compiles

Hg-commit: 52e18d98a52c392d0379b97e059010747de5fc34

Revert Barry's change to parmetis.py because it doesn't work on Cray (of course). Instead, use the patched version of metis/parmetis that accepts a 'cflags=' variable

Hg-commit: 111e7969d83b7f3bbadf

Revert Barry's change to parmetis.py because it doesn't work on Cray (of course). Instead, use the patched version of metis/parmetis that accepts a 'cflags=' variable

Hg-commit: 111e7969d83b7f3bbadfbbf5707256cdcccde13f

show more ...

need to pass cc and cxx in parametis.py to the parmetis make otherwise it uses the global mpicc mpicxx

Hg-commit: 49f58997defa5e3b3481a8e41b2b8fb959ba832e

Updated parmetis to 4.0.2. This is a big change for externalpackages:

Renamed ParMetis.py -> parmetis.py for consistency with tarball name
Renamed ParMetis -> parmetis in other package files
Use cus

Updated parmetis to 4.0.2. This is a big change for externalpackages:

Renamed ParMetis.py -> parmetis.py for consistency with tarball name
Renamed ParMetis -> parmetis in other package files
Use custom patched SuperLU_DIST (no PETSc interface change)
Use custom patched MUMPS (no PETSc interface change)

Hg-commit: 0c0bf07c7a32c9ed97da8f2f3a62077ec4fe9299

show more ...

Added metis.py 5.0.2 to packages and then added 'metis' to __init__.py. What did that do? NOBODY KNOWS.

Hg-commit: 6111c3453be120f357d1edc981b6eff0bb3a5b10

Implement self.found for CMake.py (still doesn't work as a 'required' packages in self.deps ... lack of documentation on what needs to be defined)

Hg-commit: 2aa3d143e009dfc433ebe38e15b0a7bfd37bdd54

merge from 32

Hg-commit: 869f6f0297004459ce96fa4c0c1d83c0634e6724

commit after pull from 3.2

Hg-commit: 12bf7f190e2a9b0eafa5d8c78eb160b07da3d705

pass VALGRIND_INCLUDE only in directory needed

Hg-commit: 0910059a918421cca01476a2c545fa60235f3b12

Fixed bug in basis tabulation for numComp > 2, Added 3D Stokes test
- 3D is still broken since the P2 solution does not have 0 residual

Hg-commit: 9d52bf95105e5c871f951af0e3540009af384753

TEMPORARY fix to RestrictVecVisitor, ex56 gets correct error for quadratics
- Must fix visitor correctly by putting components in PetscSection fields
- Must fix ex56 DMMeshComputeVertexFunction() for

TEMPORARY fix to RestrictVecVisitor, ex56 gets correct error for quadratics
- Must fix visitor correctly by putting components in PetscSection fields
- Must fix ex56 DMMeshComputeVertexFunction() for more general functions
- Fix allocation in ex12
- Added BC type to ex56
- Now generating numDof per dimension
- Fixed ex56 ComputeError()

Hg-commit: f05d8b07d707dc00d0ba40a55d26ad7685be8db1

show more ...

get that damn debugging crap out of Configure

Hg-commit: 0ccbc8202526251b9524d4e99b39f7a50dba29d3

Remove TSDefaultComputeJacobian{,Color}() and MatFDColoringApplyTS(); update all the examples that were still using the old style

Hg-commit: 6677905c965471eda292ff1997e1a953222a27c5

revert with-64-bit-indices change in 22bd3b9565e3. Also fix superlu_dist

Hg-commit: d9170cec19a9adfc8cbfc6720aeb8d583b8c08ec

[petsc-maint #94852] PaStiX integration update

Here is a new patch to fix some stuff with PaStiX in PetsC :

- In configuration scripts :
- Added self.requires32bitint = 0 to enable int64_t

[petsc-maint #94852] PaStiX integration update

Here is a new patch to fix some stuff with PaStiX in PetsC :

- In configuration scripts :
- Added self.requires32bitint = 0 to enable int64_t in PaStiX.py and
PTScotch.py;
- Added -DWITH_SCOTCH in PaStiX.py because new PaStiX version needs
-DWITH_SCOTCH to use scotch;
- Added -DDISTRIBUTED in PaStiX.py to make distributed interface
available in PaStiX library when self.mpi.found is true (wrapper
not written yet...);
- Use of "self.framework.argDB['with-64-bit-indices']" instead of
"self.libraryOptions.integerSize == 64" in PTScotch.py and PaStiX.py to
check the integer type (better ?);
- make scotch instead of PT-Scotch when self.mpi.found is false (Do I
also have to remove self.mpi from deps ?);
- Deactivate MPI in PaStiX library if self.mpi.found is false.
- In pastix wrapper :
- Use of all arithmetics, pastix function is choosen through PASTIX_CALL
macro, using the macro PASTIX_USE_COMPLEX to check if the type of scalar is
complex and sizeof to check if we PetscScalar are floats or doubles;
- Call pastix_checkMatric only if -mat_pastix_check is set to 1 (If
there is any problem, you may set default to 1).

Maybe we could add the fact that scotch can be compiled without PT-Scotch
when self.mpi.found is false (make scotch instead of make ptscotch).

Hg-commit: 22bd3b9565e3bd4de1fbd5a97649c229f85375fc

show more ...

update superlu tarball [to match updated superlu_dist]

Hg-commit: 4ab421c10b4b88300c1a33ce7e322ca34d74261d

commit after merge

Hg-commit: 6fb8b404bff6cd1873f79cfbe803d4926d1b8414