PC+SNES: Attach a level id to subsolvers in a multilevel solve
- We should have done this years ago

merged PetscViewerGetSingleton() and PetscViewerGetSubcomm() into PetscViewerGetSubViewer()

Does not currently work, needs fixes to work correctly recursively

Merge branch 'master' into knepley/feature-snes-deflation

* master: (206 commits)
Bib: Added ref
fix the output file for the ts example ex20adj
Bugfix: setting post-step function in ztsf.c
f

Merge branch 'master' into knepley/feature-snes-deflation

* master: (206 commits)
Bib: Added ref
fix the output file for the ts example ex20adj
Bugfix: setting post-step function in ztsf.c
fix makefile format error with 'all-legacy'
doc: handle missing dev.html in a release tarball
doc: add in petscistypes.h to doc sourcefile list
Increase patchlevel to 3.6.1
update some example output files
Doc+Vec: Fixed VecSetValuesBlocked() Reported by: Shao-Ching Huang <huangsc@gmail.com>
clarify that PETSc's "thread safe" code only works with OpenMP
allow adjoint integration over split time intervals
DT: Goddman it. Why are people stepping on my commits
Bib: Added ref
fix the bug in the endpoint version of discrete adjoint theta and correct the affected example outputs
SNES: Added SNESQN types to Fortran Reported-by: Chris Bradley <c.bradley@auckland.ac.nz>
fix: buggy change in 33da457b864643cfcac2768e47197c286471a5a4
updated petscthreads.html with information on thread per object model
fix for ex56
removed the dead links to threads.html
send petsc repository messages to Karl no longer from his mcs address
...

Conflicts:
config/builder.py
src/dm/impls/plex/examples/tests/ex3.c
src/dm/impls/plex/examples/tests/makefile
src/dm/impls/plex/makefile

show more ...

Clarified the meaning of the output in SNESTEST and SNESUpdateCheckJacobian().

NGMRES: option to restart on F_M increase

Merge branch 'maint' of bitbucket.org:petsc/petsc into maint

SNESSolve_KSPOnly() did not compute function at new solution unless monitoring is turned on
but SNESSolve_FAS() internals assume that each subsnes solve it calls does compute the new function value s

SNESSolve_KSPOnly() did not compute function at new solution unless monitoring is turned on
but SNESSolve_FAS() internals assume that each subsnes solve it calls does compute the new function value so broke
when -fas_levels_snes_monitor was not turned on but -fas_levels_snes_type kspmonitor was used

Reported-by: Lawrence Mitchell <lawrence.mitchell@imperial.ac.uk>

show more ...

Have SNESVIGetInactiveSet return the *actual* current inactive set.

Previously it returned the previous inactive set.

improve manual page slightly for clarity

Don't call SNESSetFromOptions on subsneses in SetUp

Merged in anriseth/petsc/anriseth/set-vec-sol-if-null (pull request #339)

Make sure vec_sol is not NULL before running VecDuplicateVecs

Make sure vec_sol is not NULL before running VecDuplicateVecs

Set KSPOperators after updating Jacobian

Merge branch 'master' into knepley/feature-snes-deflation

* master: (239 commits)
Bib: Moved entry
revert sources to dev - reset PETSC_VERSION_RELEASE - reset changes/dev.html - enable cuda in c

Merge branch 'master' into knepley/feature-snes-deflation

* master: (239 commits)
Bib: Moved entry
revert sources to dev - reset PETSC_VERSION_RELEASE - reset changes/dev.html - enable cuda in configure
release: set v3.6 strings
doc: add PCGAMGSetReuseInterpolation() to changes file
fix: workarround static initialization issue (perhaps compiler bug) on opensolaris c/complex build
fix: ex40,ex41 errors in src/ts/examples/tutorials/ on bsd with c++/complex Perhaps this code nees a separate section when 'enorm == 0) - as pow(0,-0.33) = inf.
manual: fix format
more changes..
PCGASM/MatGetSubMatricesMPI() changes in dev.html.
changes: more updates
Add size checks for static arrays of assumed max size [-Warray-bounds]
Doc: Update to Changes
configure: default to useThreads= 0as the useThreads=1 code currently prints output in jumbled order [producing hard to read configure.log]
Delete PETSc viewer attribute if none are left on the linked list instead of putting an empty attribute
MatGetSubMatricesMPI_MPIAIJ(): guard for an empty B-matrix when delogging memory in MatSetSeqMats().
PCBDDC: fix typos in the specification of MPI operation
MPIU_SUM was used with MPIU_INT when it should only be used with MPIU_REAL or MPIU_SCALAR
fixed bad PetscPrintf() formatting in two examples
Plex: In pointwise callbacks, use embedding dimension
updates for generating Apple installations due to changes in locations of include files
...

show more ...

fixed definition of active constraints to only include f > 0 or f < 0 to match citation
removed dead vi examples
removed dead vi routine
updated vi examples output
added command line option to zero e

fixed definition of active constraints to only include f > 0 or f < 0 to match citation
removed dead vi examples
removed dead vi routine
updated vi examples output
added command line option to zero equality of x[] value with its constraint instead of hardwired 1.e-8

Reported-by: Asbjørn Nilsen Riseth <riseth@maths.ox.ac.uk>

show more ...

Merge branch 'master' into knepley/feature-snes-deflation

* master: (32 commits)
MatGetSubmatrix() could hang when some processes thought the entire matrix as being selected and others did not.

Merge branch 'master' into knepley/feature-snes-deflation

* master: (32 commits)
MatGetSubmatrix() could hang when some processes thought the entire matrix as being selected and others did not.
PetscOptionsStringToScalar() and hence PetscOptionsScalarArray() did not work correctly for complex numbers with negative imaginary part
Have SNESCOMPOSITE honour -snes_max_fail for the number of inner solver failures it will tolerate.
fiat: add fallback url
added Stefano and Lisandro to PETSc authors list updated PETSc version number to 3.6 and year of publication to 2015 in the users manuals
deeper cleanup of matrix when the type is reset
need to call KSPSetFromOptions() after KSPSetOperators() so that command line null space is attached to correct matrix updated example output for null space not attached to KSP
Doc: Acknowledge Patrick
bugs in example that should never have gotten into next bug in code calling VecGetArray() on read only vector
doc: seperate -> separate
configure: help - prevent from printing duplcate sections - all PACAKGENAMEs should be CAPS
remove C99 code from example update output from example
updated list of PETSc enum objects that bfort will treat as native for generating Fortran stubs
-help would print the various -xxx_type values on multiple lines making a | grep -ksp_type (for example) not list all the possible options
fix memory leak as reported by the nightly tests
print help message for PetscOptionsGetViewer() the message is rather long
removed dead code about PETSc web server
removed unneeded custom XXXViewFromOptions() functions and updated to use PetscObject as second argument calling sequence
changed second argument of PetscObjectViewFromOptions() to PetscObject and all routines that use this routine
bugfix as reported by PETSc blame digest (next) 2015-05-25
...

show more ...

Merged in pefarrell/petsc/pefarrell/snes-composite-max-fail (pull request #320)

Have SNESCOMPOSITE honour -snes_max_fail

Have SNESCOMPOSITE honour -snes_max_fail for the number of
inner solver failures it will tolerate.

SNES+Doc: Fix option name

changed second argument of PetscObjectViewFromOptions() to PetscObject and all routines that use this routine

SNESFAS+Doc: More comments

NCG: Added NCG type to viewer

initial size of the trust region is set as a percent of the norm of the initial guess, it should not be related to the initial norm of the function (which is kind of nuts).

Reported-by: Mark Lohry <

initial size of the trust region is set as a percent of the norm of the initial guess, it should not be related to the initial norm of the function (which is kind of nuts).

Reported-by: Mark Lohry <mlohry@princeton.edu>

show more ...

The default bounds for SNESVISetVariableBounds() in the manual page were reversed from correct values

Reported-by: Mark Lohry <mlohry@gmail.com>

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 ...