Merge branch 'barry/2025-10-18/optimize-aij-ops' into 'main'

Refactor handling of diagonal marking in *AIJ and SELL matrices

See merge request petsc/petsc!8797

- Replace MatMissingDiagonal() and MatMarkDiagonals_SeqXXX() with MatGetDiagonalMarkers_SeqXXX()

- Mat_SeqXXX->diag is not automatically created during MatAssemblyEnd() saving memory and time

- Replace MatMissingDiagonal() and MatMarkDiagonals_SeqXXX() with MatGetDiagonalMarkers_SeqXXX()

- Mat_SeqXXX->diag is not automatically created during MatAssemblyEnd() saving memory and time

- Accessing Mat_SeqXXX->diag now requires the use of MatGetDiagonalMarkers_SeqXXX() except when the current values are known to be correct; for example during numerical factorizations and solves

- Mat_SeqXXX->diag is now never shared among matrices; hence the free_diag flag is gone. That was always a risky proposition since any of the owning matrices could chang the values thus making them incorrect for other owners.

show more ...

Merge branch 'barry/2025-10-24/fix-dmshellsetdestroyctx' into 'main'

Finish converting the function prototypes of destroy for contexts to use PetscCtxDestroyFn

See merge request petsc/petsc!8810

Change names of Mat_XXX product contexts to MatProductCtx_XXX for code maintainability

Update destroy callback of all MatProductCtx and MatShellSetMatProductOperation() to use PetscCtxDestroyFn

Merge remote-tracking branch 'origin/release'

Merge branch 'jczhang/2025-08-01/fix-MatCreateMPIAIJWithSeqAIJ' into 'release'

Change the prototype of a ctor of Mat_SeqAIJKokkos

See merge request petsc/petsc!8626

Mat: change the prototype of a ctor of Mat_SeqAIJKokkos

To correctly mark the a,i,j on host are allocated by Mat_SeqAIJKokkos, so don't use them when Mat_SeqAIJKokkos is deleted.

close #1798

Merge remote-tracking branch 'origin/release'

Merge branch 'zach/fixes-gpu-mi300a' into 'release'

MATHYPRE and Kokkos Fixes

See merge request petsc/petsc!8510

Merge remote-tracking branch 'origin/release'

Merge branch 'jczhang/2025-07-01/fix-MatCreateMPIAIJWithSeqAIJ-device' into 'release'

Fix bugs related to MatCreateMPIAIJWithSeqAIJ

See merge request petsc/petsc!8524

aijkokkos: fix a bug, in MatAssemblyEnd_SeqAIJKokkos, before deleting aijkok, need to preserve a,i,j on host

Split KokkosDualViewSync into Host and Device versions

Merge branch 'jczhang/2025-03-18/revise-aijkokkos-matsolve' into 'main'

Add options to do factorization and solve on host for matseqaijkokkos

See merge request petsc/petsc!8209

aijkokkos: support factorization on host but solve on device

Merge branch 'stevendargaville/mat-kokkoscreatedevice' into 'main'

Added new API calls to allow kokkos matrices to be built with no host preallocation

See merge request petsc/petsc!8206

Changes to allow building gpu matrices on the device.

Changed MatCreateMPIAIJWithSeqAIJ() so global sizes must be given to prevent reduction.

Changed MatCreateMPIAIJWithSeqAIJ() so B can be passed

Changes to allow building gpu matrices on the device.

Changed MatCreateMPIAIJWithSeqAIJ() so global sizes must be given to prevent reduction.

Changed MatCreateMPIAIJWithSeqAIJ() so B can be passed in with local indices, hence no compactification occurs.

Added MatCreateSeqAIJKokkosWithKokkosViews() which creates a Mat with no preallocation on the host.

Added test that checks building kokkos matrices without any host preallocation

show more ...

Merge branch 'jczhang/2025-03-05/revise-PetscGetKokkosExecutionSpace' into 'main'

Return execution space instead of reference to simplify the code

See merge request petsc/petsc!8182

Kokkos: return execution space instead of reference to simplify the code

A Kokkos execution space is a shared pointer. We don't need to reference it

Merge branch 'jczhang/2025-01-30/feature-support-AMD-MI300A' into 'main'

Add support of AMD MI300A

Closes #1703

See merge request petsc/petsc!8110

Kokkos: add support of AMD MI300A

* Use HostMirrorMemorySpace instead of HostSpace to fix compile errors on MI300A

* Replace Kokkos::HostSpace with HostMirrorMemorySpace to fix compile errors on MI

Kokkos: add support of AMD MI300A

* Use HostMirrorMemorySpace instead of HostSpace to fix compile errors on MI300A

* Replace Kokkos::HostSpace with HostMirrorMemorySpace to fix compile errors on MI300A, since the latter is what Kokkos::DualView use for its host view

* Fix a subtle bug in KokkosDualViewSync() w.r.t MI300A. Suppose we want to sync a petsc VecKokkos v on host. On MI300A, the host copy v_h and the device copy v_d share the memory. So in the old code, we used if (v_dual.need_sync_host()) to skip the device to host memory copy. But I should not skip the exec.fence(). As the device might still have kernels writing v_d, we still need to sync the device/stream to make v_d ready for use on CPU (via v_h).

show more ...

Merge branch 'jczhang/2024-06-03/misc-gpu-stream-improve' into 'main'

Misc. GPU stream improvement

See merge request petsc/petsc!7614

Kokkos: try to always have an execution space argument

Merge branch 'barry/2023-06-07/optimize-multivecs-zhang' into 'main'

Optimize VecMDot_Seq as suggested by Junchao Zhang using BLAS 2 gemv

See merge request petsc/petsc!6580

Vec: add GEMV optimizations for VecMDot and friends for VecKokkos