Skip to content

GPU CUDA variant of ISSM#35

Draft
justinh2002 wants to merge 12 commits into
access-developmentfrom
justinh2002/GPU_cuda
Draft

GPU CUDA variant of ISSM#35
justinh2002 wants to merge 12 commits into
access-developmentfrom
justinh2002/GPU_cuda

Conversation

@justinh2002

Copy link
Copy Markdown
Collaborator

No description provided.

@justinh2002 justinh2002 changed the title change dims CUDA variant of ISSM May 29, 2026
@justinh2002 justinh2002 changed the title CUDA variant of ISSM GPU CUDA variant of ISSM May 29, 2026
In +ad builds PVec resolves to adjoint_petsc::ADVec (ADVecImpl*) which
is not compatible with VecScatterCreate*/VecGetArrayRead — those APIs
only accept plain Vec.

+cuda and +ad are mutually exclusive (Spack conflict), so:
  _HAVE_PETSC_CUDA_ defined  -> scatter / VecGetArrayRead (Vec only)
  _HAVE_PETSC_CUDA_ not set  -> original VecGetValues path (handles ADVec)

Fixes CI failure in gcc-ad build (PR #35).
@justinh2002 justinh2002 marked this pull request as draft June 2, 2026 03:19
The strncmp commit accidentally dropped the preallocation call, closing
braces and return statement, leaving only the fold marker. Restore them.
@justinh2002

Copy link
Copy Markdown
Collaborator Author

!redeploy

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant