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

Add VASP recipe for GH200 #83

Merged
merged 28 commits into from
Jun 4, 2024
Merged

Add VASP recipe for GH200 #83

merged 28 commits into from
Jun 4, 2024

Conversation

AdhocMan
Copy link
Collaborator

@AdhocMan AdhocMan commented Apr 18, 2024

Initial recipe for VASP on the A100 and GH200 architecture.
The VASP source code and binaries are restricted to licensed users only.
JFrog has been configured such that only users in the vasp6 group and the CI pipeline should have access rights.

@AdhocMan AdhocMan changed the title Add VASP recipe for A100 Add VASP recipe for A100 and GH200 Apr 18, 2024
@RMeli
Copy link
Member

RMeli commented Apr 19, 2024

I think this needs a documentation page too (docs/uenv-vasp). In particular, it needs to have the "Licensing Terms and Conditions" that we have in the user portal. In #85 I added it as a danger box.

@AdhocMan AdhocMan changed the title Add VASP recipe for A100 and GH200 Add VASP recipe for GH200 Apr 29, 2024
@AdhocMan
Copy link
Collaborator Author

I think this needs a documentation page too (docs/uenv-vasp). In particular, it needs to have the "Licensing Terms and Conditions" that we have in the user portal. In #85 I added it as a danger box.

Thanks for the suggestion, I've added some documentation about licensing.

The A100 recipe has now been removed for now, since it seems to be a while until we can test it again. The only difference is the cuda architecture spec, so it will be easy to add again.

@AdhocMan
Copy link
Collaborator Author

cscs-ci run alps;system=santis;uarch=gh200;uenv=vasp:v6.4.2

@AdhocMan
Copy link
Collaborator Author

cscs-ci run alps;system=santis;uarch=gh200;uenv=vasp:v6.4.2

1 similar comment
@AdhocMan
Copy link
Collaborator Author

cscs-ci run alps;system=santis;uarch=gh200;uenv=vasp:v6.4.2

@AdhocMan
Copy link
Collaborator Author

cscs-ci run alps;system=santis;uarch=gh200;uenv=vasp:v6.4.2

@AdhocMan
Copy link
Collaborator Author

cscs-ci run alps;system=santis;uarch=gh200;uenv=vasp:v6.4.2

@AdhocMan
Copy link
Collaborator Author

cscs-ci run alps;system=santis;uarch=gh200;uenv=vasp:v6.4.2

recipes/vasp/v6.4.2/gh200/config.yaml Outdated Show resolved Hide resolved
recipes/vasp/v6.4.2/gh200/compilers.yaml Show resolved Hide resolved
recipes/vasp/v6.4.2/gh200/environments.yaml Show resolved Hide resolved
recipes/vasp/v6.4.2/gh200/environments.yaml Outdated Show resolved Hide resolved
recipes/vasp/v6.4.2/gh200/environments.yaml Show resolved Hide resolved
docs/uenv-vasp.md Show resolved Hide resolved
@AdhocMan
Copy link
Collaborator Author

cscs-ci run alps;system=santis;uarch=gh200;uenv=vasp:v6.4.2

@AdhocMan
Copy link
Collaborator Author

cscs-ci run alps;system=santis;uarch=gh200;uenv=vasp:v6.4.2

@AdhocMan
Copy link
Collaborator Author

AdhocMan commented May 8, 2024

cscs-ci run alps;system=santis;uarch=gh200;uenv=vasp:v6.4.2

@AdhocMan
Copy link
Collaborator Author

AdhocMan commented May 8, 2024

I've tested upgrading compilers and discovered some issues:

  • OpenBLAS does not seem to compile with NVHPC 23.9 and later using spack. One of the tests always segfaults. Also happens with the latest version of OpenBLAS
  • The recipe does not compile with NVHPC 24.1 and 24.3. An error in an OpenACC section of VASP occurs.

So I've updated the recipe to NVHPC 23.9 and NVPL libraries for BLAS and LAPACK.

docs/uenv-vasp.md Outdated Show resolved Hide resolved
@AdhocMan
Copy link
Collaborator Author

cscs-ci run alps;system=santis;uarch=gh200;uenv=vasp:v6.4.2

@AdhocMan
Copy link
Collaborator Author

Some of the tests in the included testsuite of VASP fail when compiled using the architecture flag -tp neoverse-v2. Using -tp host instead fixed the issues.
So the configuration has been changed to use a modified spack based on v0.21.0, where the architecture flag for NVHPC is is changed.

@AdhocMan AdhocMan requested a review from bcumming May 16, 2024 15:48
@bcumming
Copy link
Member

bcumming commented Jun 1, 2024

cscs-ci run alps;system=todi;uarch=gh200;uenv=vasp:v6.4.2

@AdhocMan
Copy link
Collaborator Author

AdhocMan commented Jun 3, 2024

cscs-ci run alps;system=todi;uarch=gh200;uenv=vasp:v6.4.2

@bcumming bcumming merged commit 6b30119 into eth-cscs:main Jun 4, 2024
2 checks passed
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.

None yet

4 participants