E3SM-Project / E3SM-Project/HOMMEXX
Meaningful platform/architecture-dependent MPI run lines for tests
Open
- Dominant language
- Fortran
- Stars
- 11
- Forks
- 0
- PR merge metrics
- No merged PRs in 30d
Description
On GPU, prtcB-flat-r3_cxx_vs_f90 is still not BFB with the Fortran on CPU. Differences are strongly suggestive of error from finite precision. Could be a gfortran CPU vs nvcc GPU thing, or could be deeper. A relevant observation is that on CPU with OpenMP, the GPU code path is BFB with the Fortran one. We can't rule out needing to make GPU-specific tests.
Contributor guide
No contributing guide indexed for this repository
Assessment
This issue has not been assessed yet.