Merged petsc/petsc into rmills/petsc master.

ABI compatibility for debug and optimized builds

Merge branch 'barry/citations'

Sys Citations: initialize citations from Fortran

Merge branch 'jed/comm-attr-delete' into maint

Fix problem with MS-MPI where keyval is invalid when Petsc_DelComm is
called. Now we split into Petsc_DelComm_Inner and Petsc_DelComm_Outer
and no lon

Merge branch 'jed/comm-attr-delete' into maint

Fix problem with MS-MPI where keyval is invalid when Petsc_DelComm is
called. Now we split into Petsc_DelComm_Inner and Petsc_DelComm_Outer
and no longer need to use MPI_Attr_get.

* jed/comm-attr-delete:
sys comm: remove redundant self-call to MPI_Attr_get in DelComm_Outer
sys comm: separate destructors for inner and outer comms
sys: use union for Bcast of an MPI_Comm in PetscObjectName
sys comm: use union instead of void* and memcpy for MPI_Comm as an attribute
sys: cannot do input validation when !PetscInitialized

show more ...

Merge branch 'jed/comm-attr-delete'

Fix problem with MS-MPI where keyval is invalid when Petsc_DelComm is
called. Now we split into Petsc_DelComm_Inner and Petsc_DelComm_Outer
and no longer need to

Merge branch 'jed/comm-attr-delete'

Fix problem with MS-MPI where keyval is invalid when Petsc_DelComm is
called. Now we split into Petsc_DelComm_Inner and Petsc_DelComm_Outer
and no longer need to use MPI_Attr_get.

* jed/comm-attr-delete:
sys comm: remove redundant self-call to MPI_Attr_get in DelComm_Outer
sys comm: separate destructors for inner and outer comms
sys: use union for Bcast of an MPI_Comm in PetscObjectName
sys comm: use union instead of void* and memcpy for MPI_Comm as an attribute
sys: cannot do input validation when !PetscInitialized

show more ...

sys comm: remove redundant self-call to MPI_Attr_get in DelComm_Outer

MS-MPI (2008 and 2012) invalidates the communicator before calling the
attribute destructor. This is considered a bug as eviden

sys comm: remove redundant self-call to MPI_Attr_get in DelComm_Outer

MS-MPI (2008 and 2012) invalidates the communicator before calling the
attribute destructor. This is considered a bug as evidenced by test
attrdeleteget added to MPICH in:

commit c532aacc14f77a04ffca0c212731700a5d3d9a01
Author: Jed Brown <jedbrown@mcs.anl.gov>
Date: Sat May 18 12:55:38 2013 -0500

test: add attrdeleteget, MPI_Attr_get called from delete_fn

With MS-MPI 64-bit from HPC Pack 2008 and 2012, MPI_Attr_get returns
error code 773 when called from delete_fn on a communicator obtained
from MPI_Comm_split.

The standard is not explicit that the 'comm' argument of delete_fn must
be valid, so this test is only in effect when !USE_STRICT_MPI.

Signed-off-by: Jed Brown <jedbrown@mcs.anl.gov>
Signed-off-by: Satish Balay <balay@mcs.anl.gov>

However, in PETSc, the attribute value is already passed as an argument,
so we need not call MPI_Attr_get to obtain it.

Note that although we still call MPI_Attr_delete and MPI_Attr_get on the
inner comm, that communicator has not been invalidated yet.

show more ...

sys comm: separate destructors for inner and outer comms

This simplifies control flow relative to the old version where the same
destructor was used for both inner and outer comm. The outer comm
de

sys comm: separate destructors for inner and outer comms

This simplifies control flow relative to the old version where the same
destructor was used for both inner and outer comm. The outer comm
destructor still redundantly calls MPI_Attr_get, but demonstrates that
this call is unnecessary. The redundancy will be removed in the sequel.

show more ...

sys comm: use union instead of void* and memcpy for MPI_Comm as an attribute

We pack an MPI_Comm into the attribute pointer so that there is no PETSc
memory allocation needed to hold the reference.

sys comm: use union instead of void* and memcpy for MPI_Comm as an attribute

We pack an MPI_Comm into the attribute pointer so that there is no PETSc
memory allocation needed to hold the reference. The degree of the
pointer can be confusing when using a loose void* pointer and memcpy, so
use a union instead.

show more ...

sys: cannot do input validation when !PetscInitialized

PetscInitialized() and PetscFinalized() may be called before
PetscInitialize() or after PetscFinalize(), thus they cannot use
validation or sta

sys: cannot do input validation when !PetscInitialized

PetscInitialized() and PetscFinalized() may be called before
PetscInitialize() or after PetscFinalize(), thus they cannot use
validation or stack macros.

show more ...

added PetscPOpenSetMachine() and command line form

Sys: destroy threadcomm after processing -options_table and -options_left

Creation of a PetscObject (like a viewer) implies threadcomm setup, so
we must ensure that the threadcomm is destroyed after

Sys: destroy threadcomm after processing -options_table and -options_left

Creation of a PetscObject (like a viewer) implies threadcomm setup, so
we must ensure that the threadcomm is destroyed after the last
PetscObject.

show more ...

completed initial implementation of PetscCitationsRegister() and added proper cites for hypre, ml, mumps

-citations option causes the print out of approriate bibtex citation indices for all packages USED in the run
Most of the appropriate citations still need to be added

Merge bug fix/patch releases up to 3.2-p7 into maint-3.2

* commit 'a9157d915ed96804280c3dc077186a2a633fd603': (157 commits)
Fix PyPI download URL to help pip figure out the release version
#1109

Merge bug fix/patch releases up to 3.2-p7 into maint-3.2

* commit 'a9157d915ed96804280c3dc077186a2a633fd603': (157 commits)
Fix PyPI download URL to help pip figure out the release version
#110990 fix timestamp order for .so vs .a during 'make install'
Update FFTW download
Increase patchlevel to 3.2-p7
...

show more ...

Merge branch 'barry/rm-xxxregisteralldestroy'

* barry/rm-xxxregisteralldestroy:
TSSSP: Fix to use proper TSSSPInitializePackage() system
flag should have been false
removed unnneeded and unuse

Merge branch 'barry/rm-xxxregisteralldestroy'

* barry/rm-xxxregisteralldestroy:
TSSSP: Fix to use proper TSSSPInitializePackage() system
flag should have been false
removed unnneeded and unused XXXRegisterDestroy() routines
removed PetscFunctionListDestroyAll() and put PetscFunctionListDestroy(&XXXList) into the package finalize

show more ...

Merged petsc/petsc into rmills/petsc master

removed unnneeded and unused XXXRegisterDestroy() routines

removed PetscFunctionListDestroyAll() and put PetscFunctionListDestroy(&XXXList) into the package finalize

Merge branch 'barry/rm-xxxregisterdynamic'

* barry/rm-xxxregisterdynamic:
Registration: remove stale 'XXRegisterDynamic)' entries in man pages
TS examples: fix use of PetscFunctionList and add t

Merge branch 'barry/rm-xxxregisterdynamic'

* barry/rm-xxxregisterdynamic:
Registration: remove stale 'XXRegisterDynamic)' entries in man pages
TS examples: fix use of PetscFunctionList and add to nightlies
Changes: Note updates to XRegisterDynamic/PetscObjectComposeFunctionDynamic
PetscObjectComposeFunctionDynamic: remove stale docs and usage
developers.tex: remove complications from function composition with dlls
removed string version of function name for XXXRegister(), PetscFunctionListAdd() and PetscObjectComposeFunction()
changes: document PetscFunctionListAdd() API change
developers.tex: update documentation of PetscObjectComposeFunction
removed path and MPI_Comm arguments from PetscFunctionListFind/Add()
removed path argument to XXXInitializePackage() and XXXRegister()
removed XXXRegisterDynamic() but kept the APIs for everything else underneath the same phase I of the update to handling registering function pointers

show more ...

PetscInitialize() and PetscFinalize() should only process -server argument (with PetscPOpen) if PETSc is built with server support

Merged petsc/petsc into rmills/petsc master

Merge branch 'barry/rm-xxxregisterdynamic' into jed/ts-eimex

PetscObjectComposeFunctionDynamic() and TSRegisterDynamic() were
replaced by PetscObjectComposeFunction() and TSRegister(), both of which

Merge branch 'barry/rm-xxxregisterdynamic' into jed/ts-eimex

PetscObjectComposeFunctionDynamic() and TSRegisterDynamic() were
replaced by PetscObjectComposeFunction() and TSRegister(), both of which
drop the string name argument.

* barry/rm-xxxregisterdynamic: (82 commits)
...

Conflicts:
src/ts/interface/tsregall.c

show more ...

removed path argument to XXXInitializePackage() and XXXRegister()

Merge branch 'barry/ams' into master

* commit 'eb4227cfdd169cc03749ba3ef12d3932f2f322bf': (24 commits)
forgotten files
basic snooping on PetscStack is now working
split AMSJavascript.py into A

Merge branch 'barry/ams' into master

* commit 'eb4227cfdd169cc03749ba3ef12d3932f2f322bf': (24 commits)
forgotten files
basic snooping on PetscStack is now working
split AMSJavascript.py into AMS.py for AMS javascript and AMSSnoopObjects.py for example showing published objects
fix bug in previous commit
added error returns for JSON-RPC AMS calls
don't let web accessor change length of array setting to publisher
Pass arrays of values back from webserver to AMS publisher
fix for last commit, missing errorcode and wrong variable name
KSPLogResidualHistory and SNESLogConvergenceHistory need to lock for AMS and have error checking when called
shutdown webserver if you started it on PetscFinalize(), what about when crash, should still shut it down
Added support for starting server automatically by application code messed around finding a bug in AMS
got value back from webclient to application
goddman script Jed sent me doesn;t work, it loses changes
split webserver into standalone problem that is not part of running PETSc application had to do this since impossible for AMS publisher and accessor to be in same application better to just totally redo AMS! But that is to large a project
moved webserver code to its own directory and file
added callback for changable fields in pyjs representation of AMS memories
add AMS viewer for PC, server can only run without any memory debugging
changed AMS publishing to use the Viewer model added simple AMS viewer for mat,ksp,snes,ts added --ksp/snes/ts_view_pre for starting the viewer before the solve (ugly but useful for AMS) added PetscStackCalls around AMS function calls when appropriate (note do not want these in webserver because do not webserver stopping just because AMS server is down
more work on moving AMS model to viewer
Fix for SNESSetConvergenceHistory() when input a is null
...

show more ...