Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Updates the libraries for GNU-GPU on Frontier #6352

Merged
merged 1 commit into from
Apr 23, 2024

Conversation

bishtgautam
Copy link
Contributor

  • Uses -L/opt/cray/pe/gcc//11.2.0/snos/lib64/ -lgfortran instead of
    -L/opt/cray/pe/gcc-libs -lgfortran as the lib in /opt/cray/pe/gcc-libs
    corresponds to GCC 12.2.0, while we are using GCC 11.2.0.
  • Adds missing -L/opt/rocm-5.4.0/lib -lhsa-runtime64

[BFB]

- Uses `-L/opt/cray/pe/gcc//11.2.0/snos/lib64/ -lgfortran` instead of
  `-L/opt/cray/pe/gcc-libs -lgfortran` as the lib in `/opt/cray/pe/gcc-libs`
  corresponds to GCC 12.2.0, while we are using GCC 11.2.0.
- Adds missing `-L/opt/rocm-5.4.0/lib -lhsa-runtime64`

[BFB]
Copy link

PR Preview Action v1.4.7
🚀 Deployed preview to https://E3SM-Project.github.io/E3SM/pr-preview/pr-6352/
on branch gh-pages at 2024-04-17 20:35 UTC

@rljacob rljacob requested review from jgfouca and removed request for grnydawn April 17, 2024 20:38
@rljacob
Copy link
Member

rljacob commented Apr 18, 2024

is this BFB ?

@bishtgautam
Copy link
Contributor Author

It should be BFB because the compiler hasn't been updated.

@grnydawn
Copy link
Contributor

I will try a couple of cases to see if they show BFB results and then try to merge it. If you have any specific compset and resolution for the BFB test, please let me know.

@bishtgautam
Copy link
Contributor Author

I don't have any compset/resolution recommendations. I didn't know we had any baselines on Frontier for --compiler gnugpu. Maybe SCREAM has tests that use the gnugpu, but I'm not sure if those tests are run from E3SM repo or the SCREAM repo.

@grnydawn
Copy link
Contributor

@bishtgautam, you're right that there is no baseline on Frontier. Sorry for the confusion. I ran the e3sm_developer test suite with this PR. Except for input data download failures (due to my incorrect wget settings), I encountered three build errors similar to the following. However, I believe this issue should be handled separately from this PR, and I think I can merge it to next and master.

/autofs/nccs-svm1_home1/grnydawn/repos/github/E3SM/components/homme/src/share/compose/compose_slmm_islmpi.hpp:548:15: error: unknown type name 'omp_lock_t'; did you mean '_IO_lock_t'?
  ListOfLists<omp_lock_t, HDT> ri_lidi_locks;
              ^~~~~~~~~~
              _IO_lock_t

/lustre/orion/cli115/proj-shared/grnydawn/e3sm_scratch/ERP_Ld3.ne4pg2_oQU480.F2010.frontier_gnugpu.20240418_145407_gjzl94/bld/gnugpu/mpich/nodebug/threads/mct/include/impl/Kokkos_ViewMapping.hpp:2695:50: error: cannot form a reference to 'void'
  using return_type = typename Traits::value_type&;

/lustre/orion/cli115/proj-shared/grnydawn/e3sm_scratch/ERP_Ld3.ne4pg2_oQU480.F2010.frontier_gnugpu.20240418_145407_gjzl94/bld/gnugpu/mpich/nodebug/threads/mct/include/impl/Kokkos_ViewMapping.hpp:3358:27: error: invalid application of 'sizeof' to an incomplete type 'typename ViewTraits<void *, Device<Serial, HostSpace>>::value_type' (aka 'void')
  enum { MemorySpanSize = sizeof(typename Traits::value_type) };
                          ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

grnydawn added a commit that referenced this pull request Apr 23, 2024
Updates the libraries for GNU-GPU on Frontier
@grnydawn grnydawn merged commit 35b65ce into master Apr 23, 2024
11 checks passed
@grnydawn grnydawn deleted the bishtgautam/machines/frontier-gnugpu branch April 23, 2024 20:03
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants