Kokkos: lazy initialization
Pass memtypes from VecScatter to SF
SFKokkos: only supports the default stream
Merge remote-tracking branch 'origin/maint' into master
Correct "-use_cuda_aware_mpi" to "-use_gpu_aware_mpi" in PetscSFSetFromOptions() man page.
Minor
Make Kokkos as the default backend of SF if available
Implement sfcuda and sfkokkos as runtime options
Add SF implemented in Kokkos
Put more cuda related stuff to sfcuda.cu
Rename sfpackcuda.cu to sfcuda.cu
PetscTimSort[WithArray]() implements Tim Peters hybrid sort algorithm(https://bugs.python.org/file4451/timsort.txt) which takes advantage of already orderedconsecutive elements in data to sort "rea
PetscTimSort[WithArray]() implements Tim Peters hybrid sort algorithm(https://bugs.python.org/file4451/timsort.txt) which takes advantage of already orderedconsecutive elements in data to sort "real world data" faster.
show more ...
Fix ISRestoreIndicesF90() for zero length array.https://gitlab.com/petsc/petsc/-/issues/607Commit-type: bug-fix/spend 30mReported-by: Fabian Jakub
Turn on checkbadSource test to generate an error when found; fix all source code that causes errorsCommit-type: portability-fix, testing-fix, style-fix, feature, maintainability/spend 1.5h
Vec: Replace 0 -> NULL for pointers
Merge remote-tracking branch 'origin/maint'
Change IS_COLORING_MAX and MPIU_COLORING_VALUE
PetscSection: refactor closure permutations to enable multiple depthsPetscSectionGetClosurePermutation(), PetscSectionSetClosurePermutation(), andPetscSectionGetClosureInversePermutation() all req
PetscSection: refactor closure permutations to enable multiple depthsPetscSectionGetClosurePermutation(), PetscSectionSetClosurePermutation(), andPetscSectionGetClosureInversePermutation() all require a new argument depth andthe getters require closure size to be specified by the caller. This allowsmultiple closure permutations to be specified, e.g., for mixed topology meshesand boundary faces.
SF: Add partial AtomicInsert support for PetscComplexUser should not rely on conflicting inserts.
Updated the use of sizeof() when possible instead of using hardwired sizes for functions whose string argumentsneed sizes. In particular done for PetscOptionsGetString(), PetscOptionsString(), and a
Updated the use of sizeof() when possible instead of using hardwired sizes for functions whose string argumentsneed sizes. In particular done for PetscOptionsGetString(), PetscOptionsString(), and a few other utilitiesCommit-type: style-fixDevelopment Tools: Vim, Emacs, Eclipse
Merge branch 'stefanozampini/feature-add-hpackages-rebased' into 'master'Fixes to MatProduct + many bug fixes + new hierarchical matrix supportSee merge request petsc/petsc!2800
Plex+SNES: Large update to support multiple DSes- Now allow simple divisions of the domain with DMSetField()- DMPlexComputeResidual_Internal() supports multiple DSes- DMPlexComputeJacobian_Interna
Plex+SNES: Large update to support multiple DSes- Now allow simple divisions of the domain with DMSetField()- DMPlexComputeResidual_Internal() supports multiple DSes- DMPlexComputeJacobian_Internal() supports multiple DSes- DMSNESCheckDisretization() supports multiple DSes- DMSNESCheckResidual() supports multiple DSes- DMSNESCheckJacoiban() supports multiple DSes- DMComputeL2FieldDiff() supports multiple DSes- PetscConveEst supports multiple DSes- Fixed bug refering to cStart from a cellIS- Added PetscDSGetQuadrature()- Added DMSetField_Internal()- Added DMSetRegionNumDS()- Protect PetscSectionView() from empty bcIndices- ISDifference now allows a NULL second argument
PetscLayoutsCreateSF: convenient routine to map two layouts via SF
1...<<21222324252627282930>>...96