PCBDDC: remove unneeded code for disconnected subdomains detection in PCBDDCAnalyzeInterface

corrected code for PCBDDCBenignCheck
NullSpace propogation disabled: all this nullspace thing is fragile

PCBDDC: remove unneeded code for disconnected subdomains detection in PCBDDCAnalyzeInterface

corrected code for PCBDDCBenignCheck
NullSpace propogation disabled: all this nullspace thing is fragile at the moment, need to be rethought

show more ...

Merge branch 'master' into barry-allow-vecset-nan-3.6

Merge branch 'stefano_zampini/feature-pardiso-schur' of https://bitbucket.org/petsc/petsc into stefano_zampini/feature-pardiso-schur

PCBDDC: remove explicit references to mumps in the source code

PCBDDC: complete optimization of benign subspace trick

There's no longer need of PtAP operations on local matrices

Begin renaming the PetscOptions objects for the future where PetscOptions will not be a singleton

Merge branch 'master' into knepley/fix-examples-plex

* master: (112 commits)
correct expression for implict ODE/DAE in table 11 of users manual
remove warning about deprecated MPI call in exampl

Merge branch 'master' into knepley/fix-examples-plex

* master: (112 commits)
correct expression for implict ODE/DAE in table 11 of users manual
remove warning about deprecated MPI call in examples
rm variables declared in middle of subroutine
tolerance for equality in PETSc test should depend on machine epsilon
tolerance for acccuracy in test example should depend on machine epsilon
increase convergence tolerance to get same result in single precision as double for convergence history
update no change output for gasm test which had not been properly updated
remove C++ // comment
example mistakenly used MPI_COMM_WORLD as if it was rank in print statement thus producing output only with MPI Uni
grep out name of DM object since on some systems the address printed has extra zeros thus producing a false difference in output
increase tolerance for test example
build: revert el6-4 build to cg - as el6 compilers give link errors
fix error type
insure similar convergence history for single, double, quad
fix ex104.c based on Barry's suggestion
increase example convergence tolerance to get matching convergence history in single and double precision
increase convergence tolerance so that test example has similar convergence history in single and double precisions
changed from cg to gmres for eigenestimator since in single precision cg produced a possible indefinite matrix report
change test example tolerance to depend on machine epsilon so as to give correct results in various precisions
loosen convergence tolerance to have similar convergence history with single, double, quad precisions
...

show more ...

Merge branch 'master' into shri/add-tsevent-to-powergrid-examples

Merge branch 'master' into stefano_zampini/feature-pardiso-schur

Branch got out of sync with master

PCBDDC: corrected logic around change of basis and benign trick

in case the benign trick has been requested, the local matrices are always swapped, instead of swapping the global matrix

We always start with petsc/master. Trying to fix a bug (submatrix reuse) reported by
Dmitry.
Merge remote-tracking branch 'petsc/master' into fix-pcgasm-reuse

Merge branch 'master' into pr263/PierreBdR/ts-domain-error-in-stage/master

Merge branch 'master' into barry/improve-git-download

fix violations of PETSc style guide: Usage of SETERRQ and NULL

Merge branch 'master' into stefano_zampini/feature-pcbddc-saddlepoint

Conflicts:
src/ksp/pc/impls/bddc/bddcprivate.c

Merge branch 'master' into tisaac/dmforest

Get 3.6.2 patches

* master: (1360 commits)
remove nasty memory leak with CUDA code when using block Jacobi with multiple blocks per process
print abs

Merge branch 'master' into tisaac/dmforest

Get 3.6.2 patches

* master: (1360 commits)
remove nasty memory leak with CUDA code when using block Jacobi with multiple blocks per process
print abs in Fortran test example in order to have same result with real and complex builds
changed outdate mpirun and -np to mpiexec and -n
force sqrt to be taken in the precision of PetscReal in fortran examples
continuation line too soon for some compilers
stop configure if user selected C++ clanguage, single precision, complex combination since we do not support it
fixes for Fortran examples to compile cleanly in single precision that do not involve complex numbers
fixes to Fortran examples to compile cleanly in single precision complex
fix previous mistake
must cast 1. to PetscScalar before dividing by a PetscScalar (C++ complex class is terrible)
must cast 0.0 to PetscScalar before comparison with a PetscScalar
changed order of suggested locations of downloadable PETSc to Barry's preference
Increase patchlevel to 3.6.2
remove debug print statement from two commits before
remove debug print statements I accidently left in the last commit
PetscDraw: Fix dummy definition of PetscDrawCollective{Begin|End}()
remove cleanupMumps flag from MatDestroy_Mumps() it should always free the memory it allocated Bug detected while adding features to libMesh
improved viewing of fractional counts in xml nested log viewing fixed bug in output of doubles to xml nested log files
chombo: build only libs - not examples
fix typo. 'gcc-4.8 --with-cxx-dialect=C++11' gives a warning with this line - even though its in a #ifdef thats never enabled.
...

show more ...

PCBDDC: no need to change the iteration matrix when the user specified the change of basis with the benign subspace trick

PCBDDC: avoid PtAP during presolve using the benign subspace trick

With the trick, there's no reason to use pmat != amat, so avoid projecting amat local since pmat local has been already projected

PCBDDC: allow adaptive selection of constraints using the exact Schur complement instead of the (possibly) incorrect one

In order to implement this efficiently and to reuse the solvers, some modific

PCBDDC: allow adaptive selection of constraints using the exact Schur complement instead of the (possibly) incorrect one

In order to implement this efficiently and to reuse the solvers, some modifications has been made to the code

show more ...

To add a space to all comments I have made before. Barry pointed out
that a comment without a space right after '/*' may cause bfort to crash

Merge remote-tracking branch 'petsc/master' into fixed-c

To add a space to all comments I have made before. Barry pointed out
that a comment without a space right after '/*' may cause bfort to crash

Merge remote-tracking branch 'petsc/master' into fixed-comment-space

Conflicts:
include/petsc/finclude/petscmat.h
include/petscmat.h
src/mat/examples/tests/makefile
src/mat/impls/aij/mpi/mpiaij.h

show more ...

PCBDDC: added API for the specification of primal vertices in global ordering

propagate info on primal vertices to coarser levels when using the benign subspace trick (still need to be improved)

Merge branch 'pr329/master/Fande-Kong/matpartitioning-hierarch' into pr361/master/Fande-Kong/pcgasm-increaseoverlap: need hierarch to build GASM subdomains.

Conflicts:
src/mat/examples/tests/makefi

Merge branch 'pr329/master/Fande-Kong/matpartitioning-hierarch' into pr361/master/Fande-Kong/pcgasm-increaseoverlap: need hierarch to build GASM subdomains.

Conflicts:
src/mat/examples/tests/makefile
src/mat/impls/aij/mpi/mpiaij.h

show more ...

PCBDDC: fixes for multiple PCSetUp calls using the benign trick

PCBDDC: added MatSeqAIJCompress to remove zero entries from SeqAIJ matrices

this is very very useful when changing the local basis with a PtAP operation, which does not care about possible zero entr

PCBDDC: added MatSeqAIJCompress to remove zero entries from SeqAIJ matrices

this is very very useful when changing the local basis with a PtAP operation, which does not care about possible zero entries in the result
MatDetectSubdomains now implements a threshold check for square dense matrices

show more ...

PCBDDC: major improvement in benign code

Allows for multiple local benign dofs (in case any of the local matrices represents disconnected subdomains)
Extended accordingly BDDC class and BDDCGraph cl

PCBDDC: major improvement in benign code

Allows for multiple local benign dofs (in case any of the local matrices represents disconnected subdomains)
Extended accordingly BDDC class and BDDCGraph class
Added MatDetectDisconnectedComponent to identify them

show more ...