configure: MUMPS 5.6.1

Fix bug with SuperLU_DIST if single precision version is not available but PETSc tries to use it

Merge branch 'barry/2023-06-15/add-fortran-file-viewer' into 'main'

Allow one to attach a PETSCVIEWERASCII to a Fortran file unit

See merge request petsc/petsc!6607

HAVE_FORTRAN should be USE_FORTRAN_BINDINGS since it is about generating the Fortran bindings, not about if the Fortran compiler exists

MATSEQAIJHIPSPARSE: fix for bug in ROCM

Fix found linter errors

Merge branch 'jacobf/2023-06-13/mat-vec-more-cupm-functions' into 'main'

Add Some Missing Vec/Mat CUPM Functions

Closes #1399

See merge request petsc/petsc!6596

Fix hash state reset bug in MPIAIJCUSPARSE

PC: offload PCSetUp_VPBJacobi_Kokkos to device

Merge branch 'release'

MatSetPreallocationCOO_MPIAIJ: set the nonzerostate as the matrix is assmbled

Mat: mark the mat as was_assembled after MatSetPreallocationCOO_MPIAIJ because we have built Mvctx

Mat: save the cost of MatStash in MatAssemblyBegin/End_MPI_Hash when user ignores off-proc entries

Mat: fix a bug that MatSetValues_MPI_Hash() did not honor MAT_IGNORE_OFF_PROC_ENTRIES

Fix make checkbadSource

triggered with a merge of !6507 (main) and !6612 (release) changes

Merge remote-tracking branch 'origin/release'

Fix errors introduced in !6090 reported in https://gitlab.com/petsc/petsc/-/issues/1389

If matrices were destroyed (possibly through a type change) before their hash state was destroyed they would

Fix errors introduced in !6090 reported in https://gitlab.com/petsc/petsc/-/issues/1389

If matrices were destroyed (possibly through a type change) before their hash state was destroyed they would result in incorrect results or memory leaks

Commit-type: bug-fix
Reported-by: Pierre Jolivet

show more ...

Allow MatResetPreallocation to be called before the matrix has ever been assembled

Fix some incorrect error checking and remove some unneeded paths

Reported-by:Karsten Lettmann <karsten.lettmann@un

Allow MatResetPreallocation to be called before the matrix has ever been assembled

Fix some incorrect error checking and remove some unneeded paths

Reported-by:Karsten Lettmann <karsten.lettmann@uni-oldenburg.de>

show more ...

Mat tests: add HYPRE COO tests

support COO with hacking permutation array of AIJ

minor: Fix typo nozero -> nonzero

Merge remote-tracking branch 'origin/release'

MATMPIBAIJ: fix non const operation in conversion to MPIAIJ

Make MatDisAssemble_MPIAIJ and MatDisAssemble_MPIBAIJ support null offdiagonal block

Merge remote-tracking branch 'origin/release'

MATSOLVERMUMPS: fix wrong FLOP counts and nnz

Merge remote-tracking branch 'origin/release'