| c1314260 | 10-Jan-2023 |
Junchao Zhang <jczhang@mcs.anl.gov> |
Mat: copy the MatDenseGetArrayAndMemType() cuda code to HIP |
| cd3f9d89 | 05-Jan-2023 |
Junchao Zhang <jczhang@mcs.anl.gov> |
Mat: add MatDenseGetArrayAndMemType and friends
Fix #1299 |
| e944a159 | 23-Jan-2023 |
Junchao Zhang <jczhang@mcs.anl.gov> |
MatAIJKokkos: update code to sync with Kokkos-kernels 4.0 |
| f3fa974c | 13-Jan-2023 |
Jacob Faibussowitsch <jacob.fai@gmail.com> |
Modernize and clean up petscsys.h
- Deprecate PETSC_NULL.
- Switch from preprocessor defines to anonymous enums for constant expressions.
- Surround macro arguments in ().
- Remove redundant casts.
Modernize and clean up petscsys.h
- Deprecate PETSC_NULL.
- Switch from preprocessor defines to anonymous enums for constant expressions.
- Surround macro arguments in ().
- Remove redundant casts.
- Clean up overly complex preprocessor defines in various PetscMem*() functions.
show more ...
|
| 18290094 | 15-Jan-2023 |
Satish Balay <balay@mcs.anl.gov> |
Merge branch 'barry/2022-12-27/rm-vecgetarray-fortran-examples' into 'main'
Replace use of VecGetArray() and friends from Fortran examples and docs
See merge request petsc/petsc!5950 |
| 0ab4885d | 28-Dec-2022 |
Barry Smith <bsmith@mcs.anl.gov> |
Deprecate VecGetArray() and similar routines in favor of VecGetArrayF90() and similar routines, and update Docs
Commit-type: housekeeping
/spend 1h |
| 0b98dbb4 | 30-Dec-2022 |
Barry Smith <bsmith@mcs.anl.gov> |
Add MatMPIAIJGetSeqAIJF90(), MatMPIAIJRestoreSeqAIJF90() |
| 99d14cb2 | 12-Jan-2023 |
Satish Balay <balay@mcs.anl.gov> |
Merge branch 'jolivet/feature-removed-errorchecking' into 'main'
Remove remaining traces of --with-errorchecking
See merge request petsc/petsc!5962 |
| 8eb1d50f | 09-Jan-2023 |
Pierre Jolivet <pierre@joliv.et> |
Fix -Wunused-parameter for CUDA files |
| 2f181314 | 08-Jan-2023 |
Pierre Jolivet <pierre@joliv.et> |
Remove remaining traces of --with-errorchecking |
| cedd07ca | 06-Jan-2023 |
Pierre Jolivet <pierre@joliv.et> |
Fix -Wunused-parameter and -Wimplicit-fallthrough |
| 1ea65430 | 04-Jan-2023 |
Satish Balay <balay@mcs.anl.gov> |
Merge remote-tracking branch 'origin/release' |
| 5249878f | 04-Jan-2023 |
Satish Balay <balay@mcs.anl.gov> |
Merge branch 'hzhang-bantingl/superlu-matsolvetranspose' into 'main'
Superlu_dist: add tests of MatSolveTranspose() MatMatSolveTranspose()
See merge request petsc/petsc!5942 |
| 78bc9606 | 16-Dec-2022 |
Hong Zhang <hzhang@mcs.anl.gov> |
followup MR5723; add tests of MatSolveTranspose() MatMatSolveTranspose() to src/mat/tests/ex125.c |
| bac8e595 | 03-Jan-2023 |
Satish Balay <balay@mcs.anl.gov> |
Merge branch 'adams/gamg-mpiaij-fixes' into 'main'
Adams/gamg mpiaij fixes
See merge request petsc/petsc!5929 |
| 35cb6cd3 | 30-Dec-2022 |
Pierre Jolivet <pierre@joliv.et> |
Fix some typos |
| 337345a6 | 21-Dec-2022 |
Satish Balay <balay@mcs.anl.gov> |
Merge remote-tracking branch 'origin/release' |
| cdd8bf47 | 21-Dec-2022 |
Junchao Zhang <jczhang@mcs.anl.gov> |
Fix a typo (bug) in MatInvertBlockDiagonal_SeqAIJ and also add tests for it |
| 2cf69117 | 18-Dec-2022 |
markadams4 <mfadams@lbl.gov> |
remove hardwired array in mpiaij |
| 851da9e1 | 17-Dec-2022 |
Hong Zhang <hongzhang@anl.gov> |
Improves safety in AVX512 kernels
- Use masked loads when AVX512VL is available
- Disable manual remainder optimization for platforms without AVX512VL support such as KNL. This prevents reading unal
Improves safety in AVX512 kernels
- Use masked loads when AVX512VL is available
- Disable manual remainder optimization for platforms without AVX512VL support such as KNL. This prevents reading unallocated memory but may lead to downgraded performance of MatMult() for MatAIJ on KNL.
Reported-by: "Matthew Kolbe"
show more ...
|
| 7815ee07 | 14-Dec-2022 |
Satish Balay <balay@mcs.anl.gov> |
Merge remote-tracking branch 'origin/release' |
| 4e208921 | 11-Dec-2022 |
Jed Brown <jed@jedbrown.org> |
MatGetValues_SeqAIJ: avoid mutable access to avoid incrementing state
MatInvertVariableBlockDiagonal_SeqAIJ: add flop counting
This fixes a -pc_type vpbjacobi performance bug in which the
precondit
MatGetValues_SeqAIJ: avoid mutable access to avoid incrementing state
MatInvertVariableBlockDiagonal_SeqAIJ: add flop counting
This fixes a -pc_type vpbjacobi performance bug in which the
preconditioner is rebuilt every iteration because its setup mutates
matrix state.
show more ...
|
| 185d4773 | 09-Dec-2022 |
Satish Balay <balay@mcs.anl.gov> |
Merge remote-tracking branch 'origin/release' |
| a7269746 | 09-Dec-2022 |
Jose E. Roman <jroman@dsic.upv.es> |
Fix compiler warning caused by having two variables with the same name
mpiov.c(57): warning #823: reference is to variable "i" (declared at line 32) -- under old for-init scoping rules it would have
Fix compiler warning caused by having two variables with the same name
mpiov.c(57): warning #823: reference is to variable "i" (declared at line 32) -- under old for-init scoping rules it would have been variable "i" (declared at line 47)
PetscCall(ISRestoreIndices(is[i], &idx));
^
show more ...
|
| 9a88ca10 | 29-Nov-2022 |
Barry Smith <bsmith@mcs.anl.gov> |
Make MatIncreaseOverlap_AIJ respect blocking |