Move definition of PetscSplitReduction to petscimpl.h

PetscSection: destroy perm in PetscSectionReset()

Merge branch 'maint'

Fixes bug in VecMTDot_Seq/MPI where it did not work correctly for more than four vectors

Reported-by: Patrick Sanan <patrick.sanan@gmail.com>

Merge branch 'jed/vecview-native' into maint

* jed/vecview-native:
Vec: restore support for PETSC_VIEWER_NATIVE
Vec: add missing Fortran enum

Merge branch 'jed/fix-checkpointer-char' into maint

* jed/fix-checkpointer-char:
Sys: PetscCheckPointer PETSC_CHAR checks char*, not char**
Sys: PetscCheckPointer for PETSC_CHAR checks char*, no

Merge branch 'jed/fix-checkpointer-char' into maint

* jed/fix-checkpointer-char:
Sys: PetscCheckPointer PETSC_CHAR checks char*, not char**
Sys: PetscCheckPointer for PETSC_CHAR checks char*, not char**

show more ...

Merge branch 'jed/vecview-native'

* jed/vecview-native:
Vec: restore support for PETSC_VIEWER_NATIVE
Vec: add missing Fortran enum

Merge branch 'knepley/fix-l2g-checks'

* knepley/fix-l2g-checks:
IS: Fixed checks for l2g

IS: Fixed checks for l2g

PCHYPRE: added interface to auxiliary space Maxwell solver AMS

Interface is under testing.
Some specific functions could be namespaced to AMS

Merge branch 'barry/fix-islocaltoglobalmappingcreateis-with-blocksize'

fix for ISLocalToGlobalCreateIS() for bs > 0. Code copied a size for each index when it should be for each block index
this happended because the documentation for ISLocalToGlobalCreate() was wrong s

fix for ISLocalToGlobalCreateIS() for bs > 0. Code copied a size for each index when it should be for each block index
this happended because the documentation for ISLocalToGlobalCreate() was wrong since it said the size was number of indices, not number of block indices

show more ...

Merge branch 'barry/always-include-complex'

Merge branch 'barry/fix-islocaltoglobalmappingcreateis-with-blocksize'

fix ISLocalToGlobalMappingCreateIS() to respect the block size of the IS
Added test case for it

Reported-by: Glenn Hammond <gehammo@sandia.gov>

Merge branch 'knepley/fix-vec-init'

* knepley/fix-vec-init:
Vec: Fixed storage initialization to also zero ghost portion

Merge branch 'jed/fix-checkpointer-char'

* jed/fix-checkpointer-char:
Sys: PetscCheckPointer PETSC_CHAR checks char*, not char**
Sys: PetscCheckPointer for PETSC_CHAR checks char*, not char**

Sys: PetscCheckPointer PETSC_CHAR checks char*, not char**

Fix VecScatterCreateCommon_PtoS.

For 32bit int builds this function was trying to insert a method into
the scatter struct directly rather than into the vtable.

Vec: restore support for PETSC_VIEWER_NATIVE

It is useful for debugging to be able to write a DMDA Vec in PETSc
ordering instead of in native ordering. This functionality was removed
in

commit 7

Vec: restore support for PETSC_VIEWER_NATIVE

It is useful for debugging to be able to write a DMDA Vec in PETSc
ordering instead of in native ordering. This functionality was removed
in

commit 742cafdd48f845c927348e65bdee73999c739268
Author: Shri Abhyankar <abhyshr@mcs.anl.gov>
Date: Thu Jul 22 14:57:46 2010 -0500

Removed PETSC_VIEWER_NATIVE related code

and discussed on petsc-dev in

http://mid.mail-archive.com/8738qk1d3u.fsf@mcs.anl.gov

This commit partially reverts the above and adds a test.
PETSC_VIEWER_NATIVE is the documented way to write the Vec contents
without converting to natural ordering. This interface should be
preserved until there is a cleaner alternative.

Reported-by: Gautam Bisht <gbisht@lbl.gov>

show more ...

CUDA: Added PETSC_SKIP_COMPLEX to remaining .cu files

CUSP: Define PETSC_SKIP_COMPLEX in seqcusp code to fix compilation.

Complex numbers aren't properly supported with CUSP anyway,
so this change does not restrict anything.

remove PETSC_DESIRE_COMPLEX and instead include complex with petscmath.h UNLESS PETSC_SKIP_COMPLEX is set (for SSL includes currently)

Merge branch 'knepley/fix-plex-coords-parallel'

* knepley/fix-plex-coords-parallel:
DMPlex: Incorrect periodicity information
KSP+SNES: Converged output now includes prefix
DMPlex ex4: Fixed t

Merge branch 'knepley/fix-plex-coords-parallel'

* knepley/fix-plex-coords-parallel:
DMPlex: Incorrect periodicity information
KSP+SNES: Converged output now includes prefix
DMPlex ex4: Fixed test output
Mat+Coloring: Greedy coloring fix - Use ocolors[] for off-diagonal piece
PetscSection: Added check for mismatch between SF and Section
DMPlex: DMPlexReplace_Static() fixes - Should share coordinateDM, not just the section, so we get correct SF - Pass on periodicity information
DMPlex: Refinement should create SF before coordinates, so coordinateDM gets correct SF - Also pass on periodicity information
DMPlex: Distribution should create SF before coordinates, so coordinateDM gets correct SF - Also pass on periodicity information

show more ...

Merged in dmeiser/petsc/fix-vecscattercusp_stos (pull request #180)

Conflicts:
src/vec/vec/examples/tests/makefile

* fix-vecscattercusp_stos: Fixes an incorrect use of VecCUSPGetArrayWrite() inste

Merged in dmeiser/petsc/fix-vecscattercusp_stos (pull request #180)

Conflicts:
src/vec/vec/examples/tests/makefile

* fix-vecscattercusp_stos: Fixes an incorrect use of VecCUSPGetArrayWrite() instead
of VecCUSPGetArrayReadWrite() inside VecScatterCUSP_StoS.

show more ...