removed cheb eig in GAMG, added CG eig est to tests

added PREONLY default for coarse grid solver in GAMG

Merge branch 'barry/propagate-pcsetup-failures'

Infrastructure that allows failures in PCSetUp(), PCApply(), MatMult() etc due to, for example a zero pivot or function evaluation outside its domain, to propagate up and become a KSP_DIVERGED inste

Infrastructure that allows failures in PCSetUp(), PCApply(), MatMult() etc due to, for example a zero pivot or function evaluation outside its domain, to propagate up and become a KSP_DIVERGED instead of generating an error that stops the program. In response to Issue 96. This includes failures in MatCreateSNESMF() applications due to domain errors. The mechanism to propagate some errors is by setting Info or Nan into the output vector and using the norm or inner product reductions in SNES or KSP to propagate the error condition to all processes and then handling them immediately after the norm or inner product.

This allows, for example, ODE integrators to try again with a smaller time-step if the PCSetUp failed instead of requiring a complete restart of the run with other options.

Currently some error conditions, such as function domain error in a line search may not get progated up using the correct SNESConvergedReason

See src/snes/examples/tests/ex69.c for the handling of several conditions

show more ...

changed GAMG square graph parameter to int from bool

use %D for print N in GAMG

fixed missing CHKERRQ() detected by clang's --analyze

updated PETSc directory layout to match standard packaging strategies

include/petsc finclude,private,mpiuni
lib/petsc conf
bin/petsc*

MPI_SUM/MAX/MIN should be used with MPIU_INT not MPIU_SUM/MAX/MIN

Reported-by: Jed Brown <jbrown@mcs.anl.gov>

fixed up PetscInfo in GAMG

Merge branch 'barry/clean-gamg'

Conflicts:
src/ksp/pc/impls/gamg/gamg.c

bad compiler indicates that flid may be used without initialization while it is actually initialized in both branches of the code

merged mallocs/frees in GAMG routines

PCGAMG did not respect MPIAIJ matrix type when passed in on one process
It used AIJ on one process in some cases causing crashes.
Now PCGAMG uses consistently the Seq or MPI nature of the input matri

PCGAMG did not respect MPIAIJ matrix type when passed in on one process
It used AIJ on one process in some cases causing crashes.
Now PCGAMG uses consistently the Seq or MPI nature of the input matrix to
create its new matrices and no longer crashes with input MPIAIJ matrix on
one process.

Reported-by: Randall Mackie <rlmackie862@gmail.com>

show more ...

removed eigen test in gamg driver

fixed non-PETSc standard usage of const in GAMG code

Merge branch 'master' into barry/clean-gamg

Need updates to KSPCHEBYSHEV since I will be adding to them

There is no verbose in PETSc!

Removed the verbose mechanism from GAMG code and used PetscInfo() for reports instead.

missing error check on MPI call

start moving GAMGHash to use static inline functions to reduce overhead

very minor cleaning of some GAMG source code

MatDuplicate() for sequential AIJ did not copy over required field
MatDuplicate() does not copy over DM hence things like MatView() fail
Changed PCGAMGFilterGraph() to diagonally scale and use absolu

MatDuplicate() for sequential AIJ did not copy over required field
MatDuplicate() does not copy over DM hence things like MatView() fail
Changed PCGAMGFilterGraph() to diagonally scale and use absolute values of graph matrix even for negative threshold; don't know if this is needed in the algorithm or not

show more ...

added an option of negative thresold for generating the graph from the matrix to mean keep even the zero entries in the matrix
This saves all the calls to MatSetValues() for generating the graph matr

added an option of negative thresold for generating the graph from the matrix to mean keep even the zero entries in the matrix
This saves all the calls to MatSetValues() for generating the graph matrix. (MatSetValues calls are known to be slow).

Surprisingly to me it save 20% of the total time on my laptop for one process for
./ex45 -da_refine 4 -pc_type gamg -ksp_monitor -pc_gamg_threshold -1.0 -ksp_type fgmres -ksp_view -log_summary
compared to
./ex45 -da_refine 4 -pc_type gamg -ksp_monitor -pc_gamg_threshold 0.0 -ksp_type fgmres -ksp_view -log_summary
it generated smaller coarse grid matrices and had very similar convergence rate so not only did it essentially eliminate the PCGAMGGraph_AGG time but the other times decreased as well

show more ...

added infrastructure for PCView() to provide info on GAMG options and sub GAMG options, such as AGG

removed some unneeded rank and size from gamg code