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

Unit tests fail on Frontier after recent module updates #104

Open
philipwjones opened this issue Jul 22, 2024 · 11 comments
Open

Unit tests fail on Frontier after recent module updates #104

philipwjones opened this issue Jul 22, 2024 · 11 comments
Assignees
Labels
bug Something isn't working Omega

Comments

@philipwjones
Copy link

Recent updates of modules on Frontier are preventing successful builds, so the E3SM config_machine must be updated.

As a workaround, the default compiler configurations appear to work for the crayclang (CPU) and crayomegagpu (GPU) options:
PrgEnv-amd/8.5.0
PrgEnv-cray-amd/8.5.0
PrgEnv-cray/8.5.0
PrgEnv-gnu-amd/8.5.0
PrgEnv-gnu/8.5.0
amd/5.7.1
craype/2.7.31
cce/17.0.0
cpe/23.12
gcc/12.2.0
rocm/5.7.1

cray-netcdf-hdf5parallel/4.9.0.9
cray-hdf5-parallel/1.12.2.9
cray-parallel-netcdf/1.12.3.9

I don't know if these options will work for the E3SM builds, so the eventual solution may be different.

@philipwjones philipwjones added bug Something isn't working Omega labels Jul 22, 2024
@philipwjones
Copy link
Author

Workaround posted here as promised on today's omega call @sbrus89 @xylar

@xylar
Copy link

xylar commented Jul 22, 2024

Thanks @philipwjones!

1 similar comment
@sbrus89
Copy link

sbrus89 commented Jul 22, 2024

Thanks @philipwjones!

@philipwjones
Copy link
Author

I recently inadvertantly built omega with the old modules to test a PR and they appear to all be working again - not sure if OLCF reinstalled old mods for back-compatibility? We should still update to more recent versions, but workaround may no longer be necessary.

@grnydawn
Copy link

@philipwjones, thanks for the info. I was also able to successfully build and run Omega without updating the modules. In any case, I will keep working on updating the e3sm machine config and then update the Omega config.

@sbrus89
Copy link

sbrus89 commented Jul 31, 2024

@philipwjones and @grnydawn are either of you still able to build/run Omega without updating modules on Frontier? I tried this today and was able to build successfully, but am seeing a bunch of /opt/cray/pe/gcc-libs/libstdc++.so.6: version 'GLIBCXX_3.4.32' not found-related errors when running the ctests.

@philipwjones
Copy link
Author

@sbrus89 Yup, still works for me. I usually build with crayclang for cpu and crayomegagpu for gpu. Both successfully built and ran the develop branch with both old and new modules.

@grnydawn
Copy link

@sbrus89 , I could not reproduce it because I could not get allocation on Frontier today. If you can share the location of the Omega build directory with me, I will look into it.

@mwarusz
Copy link
Member

mwarusz commented Jul 31, 2024

@sbrus89
I got the same error on gpu when I tried testing #105 on Frontier yesterday. I was using crayclanggpu. Cpu build with gnu worked fine.

@grnydawn
Copy link

@sbrus89, @mwarusz, try to use "crayomegagpu" instead of "crayclanggpu." Until Omega is fully compatible with E3SM configurations, we want to use Omega-specific configurations including "crayomegagpu," "gnuomegagpu," and "amdomegagpu."

@sbrus89
Copy link

sbrus89 commented Aug 1, 2024

Thanks @philipwjones and @grnydawn!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
bug Something isn't working Omega
Projects
None yet
Development

No branches or pull requests

5 participants