Removed BlockSolve95
- It is no longer supported
- ILU functionality is subsumed by PILUT in Hypre

Hg-commit: d675b710d2423988e415bd4d531a95abb123ca4c

started hypre HYPRE_StructMatrix based PETSc matrix type
hypre contains now outdated struct stencil and matrix stuff (will be removed next)

Hg-commit: 09c845b0ad214d20d8d422125e3595c17ebee30a

fixes and more error checking in many XXXGetType() calls

Hg-commit: 50b8601b7453642d7b3c91015c3af122e271a9ef

completed first generation of PCEXOTIC with support for a face and a wirebasket based
coarse grid space; ready for testing with pflotran

Hg-commit: 429f53fb595cd45cf53abcc3b513c1f769ef5a7f

From: mstefano@milan.westerngeco.slb.com
Subject: [Fwd: Re: [PETSC #18599] Valgrind problem with MatGetSubMatrix]
Date: November 14, 2008 2:30:37 AM CST
To: petsc-maint@mcs.anl.gov
Cc: pets

From: mstefano@milan.westerngeco.slb.com
Subject: [Fwd: Re: [PETSC #18599] Valgrind problem with MatGetSubMatrix]
Date: November 14, 2008 2:30:37 AM CST
To: petsc-maint@mcs.anl.gov
Cc: petsc-maint@mcs.anl.gov

I think I've found the bug.

Look at src/mat/impls/aij/seq/aij.c, line 1699.

There, variable "i" is declared as a PetscInt.

Then, look at line 1710 or 1712.

The variable "i" is used to get the truth answer from the ISSorted function.

As long as you don't build PETSc with --with-64-bit-indices, all is OK.

But (and this is my case) if you built PETSc with 64 bit indices, the PetscTruth data type is smaller (as byte size) than PetscInt. The result is that only part of the bytes of the "i" variable are initialized by ISSorted.

I strongly suggest to add a new variable to be used exclusively as a PetscTruth flag for the ISSorted calls.

Fixed this bug all over the place

Hg-commit: 79369c761f62f38e25800abc8725233db2a5486a

show more ...

PCApply_MG does not need to check if mg[levels-1]->r exists, it is now ALWAYS set in PCSetUp_MG

Hg-commit: 3664f08236c7bf595f3fb207cfd4da1c766b822a

Registered BFBt PC

Hg-commit: 11a138bc23ced53264aeee2301eec9f95dcae63c

moved assignment of LUFactorNumeric and CholeskyFactorNumeric into appropriate symbolic factorization (previously it was set
in the function table or in the getfactor
moved assignment of solve() i

moved assignment of LUFactorNumeric and CholeskyFactorNumeric into appropriate symbolic factorization (previously it was set
in the function table or in the getfactor
moved assignment of solve() into appropriate numeric factorization, previously it was in function table or get factor or symbolic factor

Hg-commit: 83ba8e21d43889fe945455ea7c1fb95368f1a281

show more ...

added petscapp.bib pub

Hg-commit: 98227e0e771a8e736c3f38b40d1fba4a54925a11

When KSPRICHARDSON is used without a KSP monitor and the PC has a PCApplyRichardson() then
ksp->its and converged reason are properly set to indicate type of convergence.

Also with HYPRE BoomerAMG

When KSPRICHARDSON is used without a KSP monitor and the PC has a PCApplyRichardson() then
ksp->its and converged reason are properly set to indicate type of convergence.

Also with HYPRE BoomerAMG when used as a solver (with KSPRichardson) the KSP convergence
criteria is passed down to BoomAMG and replaces the default (usually 0.0) criteria we normally
pass to BoomerAMG (for when we use BoomerAMG as a preconditioner.

Keita,

Thanks for pointing this out. This is a "feature" of some of the preconditioners when used with
Richardson. We call the preconditioner with a number of iterations directly and the preconditioner
decides when to exit (rather than all other cases where KSP decides), hence we did not have
information about the number of iterations. So in previous versions we did not set the ksp->its;
petsc-dev was changed to set ksp->its to the maximum number of iterations requested in the
preconditioner.

This too is not really correct because some preconditioners when used with Richardson
(specifically PCMG and hypre's BoomerAMG) will do the convergence test internally and
may not use the maximum number of iterations requested. I will modify petsc-dev today to
collect this information back from the preconditioner and pass it back correctly to ksp->its.

Hg-commit: d8d04f2052ab2294999b137bb681449db81389c7

show more ...

commit after merge

Hg-commit: 40fbcb1ffa2ac14ad5b1cd85db9cdd19834e6819

tentative working version of ibcgs

Hg-commit: de0dbdc36195e8aaaafee5ef5ba612be2c7c55d9

commit after most of merge of petsc-dev with petsc-dev-new-solvers with new way of handling external solvers

Hg-commit: 7c86fa28bc8468b955cc6a62edd1de55d6c3d700

basic transistion to new model for external direct solvers all tests pass

Hg-commit: 709a5887e4f17b074188a2f60d4f00cb8895dd52

moved the const out of XXType definitions and into the Set and Get Methods

Hg-commit: d4d9a237045a151847ef3121b00df73d952d0999