Skip to content

avoid Xcompiler flags if building with llvm+enzyme+cuda #240

Description

@chapman39

when using enzyme + cuda, you need to set your cuda compiler from nvcc to llvm. therefore, we shouldn't be setting nvcc related flags anymore.

error: error: unknown argument: '-Xcompiler=-fopenmp=libomp'unknown argument: '-Xcompiler=-fopenmp=libomp'

see MFEM as reference for the kind of logic we will need: https://github.com/chapman39/spack-packages/blob/4d3d0f8ee5ffd6f099727f46ec704f3f4ee98e5c/repos/spack_repo/builtin/packages/mfem/package.py#L609

punting this for now since Smith doesn't build tribol via Spack anymore

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions