Add needed petscao.h

(previously example including petsc.h which I removed but I did not realize that ao was needed)

Reported-by: nightly tests

Vec example should only include petscvec.h not petsc.h

a variety of corrections suggested by the clang static analyzer

Time: 1 hour

Fixes found by clang static analyzer
mostly set but unused variables plus a couple of true bugs

Remove #include <petscmat.h> in src/vec.

This violates the layering and leads to undefined symbols when building
with sub-libraries.

Fix complex conjugation in VecDot_SeqCUDA and VecMDot_SeqCUDA

PetscSFDistributeSection(): handle case when nrooots < end of chart

PetscSF: fix PetscSFGetMultiSF() for sparse leaves

PetscSF: make regression test that fails

The multiSF doesn't use the sf->mine array.

Unify Max function for real and complex scalars

Fix VECCUDA compilation with complex and single precision

Enable complex scalars with VECCUDA

few more clang static analyzer fixes

many issues detected by clang static analyzer

Time: .5 hours

Merge branch 'pr421/alex/feature-veccuda'

* pr421/alex/feature-veccuda
The rationale of this pull request is to have GPU-enabled vectors purely based on CUDA,
and with the possibility of placing a u

Merge branch 'pr421/alex/feature-veccuda'

* pr421/alex/feature-veccuda
The rationale of this pull request is to have GPU-enabled vectors purely based on CUDA,
and with the possibility of placing a user-provided array on the GPU side.

Conflicts:
src/vec/vec/impls/seq/seqcusp/cuspvecimpl.h
src/vec/vec/impls/seq/seqviennacl/viennaclvecimpl.h

show more ...

Minor changes and fix to bugs introduced in 82f73ecaa

Add missing code for the case PETSC_HAVE_VECCUDA

Fix bugs introduced in 82f73ecaa

Use VECCUDA with MATAIJCUSPARSE

Removing annoying warning in nightly tests

Reported-by: nightly tests

Merge branch 'tisaac/dmforest'

* tisaac/dmforest: (36 commits)
DMP4est: fix transfer of labels to parents
DMPlexCreateInterpolatorTree: fix handling of global dofs (can be < 0)
DMPlexComputeIn

Merge branch 'tisaac/dmforest'

* tisaac/dmforest: (36 commits)
DMP4est: fix transfer of labels to parents
DMPlexCreateInterpolatorTree: fix handling of global dofs (can be < 0)
DMPlexComputeInterpolatorTree(): fix workspace size
DM: warn when no injection implemented
DMPlexComputeInterpolatorTree(): use row layout to preallocate
DMP4est: all processes agree on root type
DMPlexComputeInterpolatorTree(): valgrind fixes
DMP4est: reset childID if anctecedent is not parent
DMP4est: send labels from children to parents and bcast
DMForest: added DMForestGetAdaptivityPurpose()
DMP4est: enable caching of results of DMPforestGetTransferSF()
Plex: fix empty partition errors
DMPforestGetTransferSF_Internal(): handle different type roots on different processes
DMP4est: broadcast labels when transfering for agreement
PetscSFBasic: fix usage of MPI_Reduce_local
DMPforestLabelsInitialize(): search through parents when finding a point in the closure
DMP4est: try to connect high strata points when transfering
DMP4est: can't use PetscSFCreateSectionSF. It compresses when we don't want to
DMP4est: expand transfer sfs by the CellSF
DMP4est: add DMForestCreateCellSF_p4est
...

show more ...

CUDA: Fixed visibility and disabled Fortran bindings for VecScatters on GPU.

Replace kernels with thrust

Fix visibility issues in CUSP and VECCUDA

Fix fortran stubs for the case --with-fortran-interfaces

Merged duplicated functions into a single one.
VecScatterInitializeForGPU
VecScatterFinalizeForGPU