Suggested-by: Jacob Faibussowitsch <jacob.fai@gmail.com>

Barry Smith authored 1 month ago and Satish Balay committed 1 month ago

Fix bug in PetscShmgetDeallocateArray needs to call PetscCheck(!shmdt(next->addr), since that is not called in the unmap on rank 0

Reported-by: Marsha Berger
Reported-by: Randy Leveque

- Add PetscDrawHGAddWeightedValue()

James Wright authored 4 months ago and Satish Balay committed 1 month ago

Co-authored-by: Barry Smith <bsmith@mcs.anl.gov>

Co-authored-by: Jed Brown <jed@jedbrown.org>

Rather than erroring out.

Accessing the FlowSolutionPointers arrays requires a different index
than the solution id

See discussion: !7993 (comment 2201318837)

Equivalent to Petsc{Int,Real,Scalar}View, but user selects the number of
columns-per-row printed to screen

Francesco Ballarin authored 4 months ago and Satish Balay committed 4 months ago

patch from https://salsa.debian.org/science-team/petsc/-/blob/fecc0768a4860dd3516c454fb3ff3613b61ff482/debian/patches/soname_extension

Co-authored-by: Jed Brown <jed@jedbrown.org>

Barry Smith authored 4 months ago and Satish Balay committed 4 months ago

Check for env variable TORCHELASTIC_RUN_ID to "ensure" LOCAL_RANK env variable is truly associted with PyTorch.

Reported-by: Stefano Zampini

Changes the previous subset of destructor APIs that used PetscErrorCode (*)(void *) (mostly those that
used PetscContainer

- Now allows any context to be a PetscObject
- Will provide a cleaner mapping to bindings in other languages
- Simplifies the maintenance of PETSc source code; improves clarity

Not backward compatible, compiler warnings will tell users what functions need to be updated