CUSP, ViennaCL: VecAYPX no longer counts flops if alpha == 0.0

This is now more consistent with the CPU case, where copies are not counted as FLOPs.

Merge branch 'balay/remove-sdir'

__SDIR__ is gone. This is an ABI change (PetscError) so dependent
packages must be recompiled.

* balay/remove-sdir:
__SDIR__: remove remaining traces and documen

Merge branch 'balay/remove-sdir'

__SDIR__ is gone. This is an ABI change (PetscError) so dependent
packages must be recompiled.

* balay/remove-sdir:
__SDIR__: remove remaining traces and document in changes.html
__SDIR__: remove usage of __SDIR__ from PETSc sources.

This merge fixes one instance of __SDIR__ that was added after
'balay/remove-sdir' branched off.

show more ...

Merge branch 'jed/doc-restore-zeros-pointer'

* jed/doc-restore-zeros-pointer:
DMPlex: fix my mistake using NULL in place of (int) 0
doc: restore functions in PETSc zero their arguments

Conflict

Merge branch 'jed/doc-restore-zeros-pointer'

* jed/doc-restore-zeros-pointer:
DMPlex: fix my mistake using NULL in place of (int) 0
doc: restore functions in PETSc zero their arguments

Conflicts:
src/dm/impls/plex/plex.c

show more ...

Merge branch 'barry/add-vecgetarray-fortran-derived-type'

* barry/add-vecgetarray-fortran-derived-type:
added example showing how to use VecGetArray() from Fortran but treating the array as an arr

Merge branch 'barry/add-vecgetarray-fortran-derived-type'

* barry/add-vecgetarray-fortran-derived-type:
added example showing how to use VecGetArray() from Fortran but treating the array as an array of Fortran derived types

show more ...

ViennaCL: Fixed unused variable and function warnings.

This was mostly due to a poor handling of option parsing.
The two old functions
ViennaCLSetFromOptions
VecSetFromOptions_SeqViennaCL
were ref

ViennaCL: Fixed unused variable and function warnings.

This was mostly due to a poor handling of option parsing.
The two old functions
ViennaCLSetFromOptions
VecSetFromOptions_SeqViennaCL
were refactored:
- VecSetFromOptions_SeqViennaCL was completely removed, as it only forwarded.
- ViennaCLSetFromOptions was renamed to PetscObjectSetFromOptions_ViennaCL
for better accordance with the PETSc naming scheme (thanks to Jed for pointing this out)
The options are now checked right before the respective objects are instantiated,
thus delaying the OpenCL setup as long as possible.

show more ...

Banished all mention of the old MatGetColoring() and use in examples

__SDIR__: remove usage of __SDIR__ from PETSc sources.

Compiler: Fix warnings from MPI impls which do not initialize outputs

(cherry picked from commit d892089bd4a83d5450f25ac15dd2b0a661537089)

Merge branch 'knepley/fix-plex-examples'

* knepley/fix-plex-examples:
SNES ex62: Remove code generation
DMPlex ex7: Added missing test output
SNES ex52: Removed old Jacobian stuff, and fixed c

Merge branch 'knepley/fix-plex-examples'

* knepley/fix-plex-examples:
SNES ex62: Remove code generation
DMPlex ex7: Added missing test output
SNES ex52: Removed old Jacobian stuff, and fixed call to DMPlexProjectFunction() - Fixed calls for new element handling
SNES ex12: Added a performance profiling mode
DMPlex ex8: Fixed leak
PetscSection: Must reset the section when changing the number of fields
DMPlex ex7: Fixed test output - Corrected orientations in interpolation
DMPlex ex1: Fixed test output - Uniform refinement for quads changed
DMPlex: Fix overagressive checks

show more ...

CUSP, ViennaCL: Fixed wrong behavior of VecAYPX() for CUSP and ViennaCL vectors.

The case alpha == 0.0 was not handled correctly, where a copy y <- x is needed,
but instead a no-op was implemented.

CUSP, ViennaCL: Fixed wrong behavior of VecAYPX() for CUSP and ViennaCL vectors.

The case alpha == 0.0 was not handled correctly, where a copy y <- x is needed,
but instead a no-op was implemented.

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

show more ...

Compiler: Fix warnings from MPI impls which do not initialize outputs

PetscSection: Must reset the section when changing the number of fields

Added several additional MPI_Ops to PetscSFBasic.

Added MPI_LAND,MPI_BAND,MPI_LOR,MPI_BOR,MPI_LXOR,MPI_BXOR and MPI_PROD
implementations for FetchAndOp and Unpack instances. Changed ex1.c
so that t

Added several additional MPI_Ops to PetscSFBasic.

Added MPI_LAND,MPI_BAND,MPI_LOR,MPI_BOR,MPI_LXOR,MPI_BXOR and MPI_PROD
implementations for FetchAndOp and Unpack instances. Changed ex1.c
so that the user can specify the op by command line.

show more ...

Merge branch 'paulmullowney/cusp-vector-scatter-with-fix'

* paulmullowney/cusp-vector-scatter-with-fix:
CUDA/CUSP: Implementation of Sequential to Sequential Vector Scatters

Merge branch 'knepley/fix-plex-preallocation'

* knepley/fix-plex-preallocation:
PetscSF: Fixed PetscSFCreateEmbeddedSF() - We were using sf->nleaves for the leaf buffer, which is completely wrong.

Merge branch 'knepley/fix-plex-preallocation'

* knepley/fix-plex-preallocation:
PetscSF: Fixed PetscSFCreateEmbeddedSF() - We were using sf->nleaves for the leaf buffer, which is completely wrong. We have to use the largest thing in sf->mine[]
DMPlex: Fix completely broken code in PetscSFCreateRemoteOffsets() - Non-broken example was in PetscSFDistributeSection()
DMPlex: Fix leaks in preallocation debugging
DMPlex: Put in guard for preallocation sections which are smaller than the SF

show more ...

Merge branch 'knepley/feature-closure-cache'

* knepley/feature-closure-cache:
DMPlex: Added declaration of DMPlexCreateClosureIndex()
SNES ex62: Use closure index
DMPlex: Use closure index if

Merge branch 'knepley/feature-closure-cache'

* knepley/feature-closure-cache:
DMPlex: Added declaration of DMPlexCreateClosureIndex()
SNES ex62: Use closure index
DMPlex: Use closure index if available for DMPlexVecGetClosure()
DMPlex: Added DMPlexCreateClosureIndex() - Uses a PetscSection+IS and stashes in the original Section
PetscSection: Added an index for optimization of closure queries - Added PetscSectionGet/SetClosureIndex() - The index is keyed to the DM which built it (use PetscObject) - Only use a single key for now since a lot of checking would be slow

Conflicts:
src/snes/examples/tutorials/ex62.c

show more ...

PetscSection: Added PetscSectionPermute()

PetscSF: Fixed PetscSFCreateEmbeddedSF()
- We were using sf->nleaves for the leaf buffer, which is completely wrong. We have to use the largest thing in sf->mine[]

Merge branch 'jed/remove-dynamic-loading-option'

Option --with-dynamic-loading has been removed. Similar functionality
is available via the run-time option -dynamic_library_preload.

* jed/remove-d

Merge branch 'jed/remove-dynamic-loading-option'

Option --with-dynamic-loading has been removed. Similar functionality
is available via the run-time option -dynamic_library_preload.

* jed/remove-dynamic-loading-option:
Sys: PetscContainerCreate must call PetscSysInitializePackage
Remove --with-dynamic-loading / PETSC_USE_DYNAMIC_LIBRARIES

Conflicts:
src/dm/impls/mesh/mesh.c
src/dm/interface/dlregisdmdm.c
src/docs/website/documentation/changes/dev.html

show more ...

added example showing how to use VecGetArray() from Fortran but treating the array as an array of Fortran derived types

DMPlex: Fix completely broken code in PetscSFCreateRemoteOffsets()
- Non-broken example was in PetscSFDistributeSection()

Adding a test for GPU Sequential to Sequential Vector Scatters

All vector Sequential to Sequential vector scatters tested in
this new example using SCATTER_FORWARD and ADD_VALUES for scatter
paramet

Adding a test for GPU Sequential to Sequential Vector Scatters

All vector Sequential to Sequential vector scatters tested in
this new example using SCATTER_FORWARD and ADD_VALUES for scatter
parameters. This includes all four combinations of sequential
general and sequential strided. Switching -vec_type between
cusp and seq yields the same results.

show more ...

Original commit by Paul Mullowney (augmented by PETSC_HAVE_CUSP include
guards by Karl Rupp):
Implementation of Sequential to Sequential VecScatters on the GPU.

In this commit, I've built a working

Original commit by Paul Mullowney (augmented by PETSC_HAVE_CUSP include
guards by Karl Rupp):
Implementation of Sequential to Sequential VecScatters on the GPU.

In this commit, I've built a working prototype for sequential
to sequential vector scatters for CUSP vectors. I've also reorganized
the parallel to parallel vector scatters in a new infrastructure.

The design of the code is as follows. Currently, I distinguish
between PtoP (parallel to parallel) VecScatters and StoS (sequential
to sequential) VecScatters. In cuspvecimpl.h, a high level struct
called PetscCUSPIndices stores a void * pointer and enumerated type
describing whether the scatter is PtoP or StoS (and later PtoS and
perhaps StoP). The actual type of scatter information built
and stored in the void * pointer depends on the calling code.

For instance, in vscat.c, there are the methods VecScatterBegin_SGToSG,
VecScatterBegin_SGToSS, ... These methods build the indices for
the sequential scatter on demand via the function
VecScatterCUSPIndicesCreate_StoS. Those indices are stored in
a struct of type _p_VecScatterCUSPIndices_StoS. the appropriate
data in that struct is filled depending on whether the input
and output are general or strided. Some additional meta data is
also stored.

A similar routine is called from vpscat.c for building the
PtoP vector scatter. The structure of this hasn't changed although
the constructing API method is now VecScatterCUSPIndicesCreate_PtoP
(it used to be PetscCUSPIndicesCreate).

A single destructor function is given for the PtoP and StoS scatters:
VecScatterCUSPIndicesDestroy. It used to be PetscCUSPIndicesDestroy.

I've made all of these methods PETSC_INTERN. I can't see any
reason to expose these methods to the user.

All of the source code for doing the sequential scatter computation
on the GPU is moved into a file vecscattercusp.cu (src/vec/vec/impls/seq/seqcusp).
I haven't moved the VecCUSPCopyToGPUSome and similar methods into
this file, although perhaps they belong there.

All scatter types implemented including FORWARD and REVERSE as well
as INSERT, ADD, MAX.

show more ...

doc: restore functions in PETSc zero their arguments

This is done to prevent accidental misuse. In general, one can pass
NULL for the pointer argument to keep the pointer around for longer, but
use

doc: restore functions in PETSc zero their arguments

This is done to prevent accidental misuse. In general, one can pass
NULL for the pointer argument to keep the pointer around for longer, but
use of arrays after restoration is invalid.

Suggested-by: Wolfgang Bangerth <bangerth@math.tamu.edu>

show more ...

Merge branch 'jed/object-id'

* jed/object-id:
Sys: mark PetscObjectState modification as logically collective
Sys: convert object ID to 64-bit and add PetscObjectGetId()
Sys: typedef PetscObje

Merge branch 'jed/object-id'

* jed/object-id:
Sys: mark PetscObjectState modification as logically collective
Sys: convert object ID to 64-bit and add PetscObjectGetId()
Sys: typedef PetscObjectState and change all accessors
unified PetscObjectStateXXX() commands. Fixed PetscObjectStateQuery() to PetscObjectStateGet()

show more ...