build: add extra cuda13, rocm7; remove cuda11, rocm4,5#2561
build: add extra cuda13, rocm7; remove cuda11, rocm4,5#2561havogt merged 6 commits intoGridTools:mainfrom
Conversation
| image: ${CSCS_REGISTRY_PATH}/public/${ARCH}/base/gt4py-ci-${PY_VERSION}:${DOCKER_TAG} | ||
| variables: | ||
| TEST_VARIANTS: 'cpu' # Extended jobs should redefine which variants (cpu, cuda12, rocm6_0) to test | ||
| TEST_VARIANTS: 'cpu' # Extended jobs should redefine which variants (cpu, cuda12, rocm6) to test |
There was a problem hiding this comment.
Question: is it clear for you what is this variable used for? I'm not longer sure if I understand why it is needed. Also, do you explicitly keep the cuda12 variant instead of cuda13 here?
There was a problem hiding this comment.
Trying to reverse engineer, I think we define all SUBVARIANTs in test_common, then TEST_VARIANTS should be used in the concrete test_cscs_gh200, test_cscs_amd_rocm to pick which to actually run.
We are still using cuda12 on santis by default, therefore I didn't bump the version (assuming you are not talking about the comment, but about what we run).
There was a problem hiding this comment.
We are still using cuda12 on santis by default, therefore I didn't bump the version (assuming you are not talking about the comment, but about what we run).
Understood, thanks.
Trying to reverse engineer, I think we define all SUBVARIANTs in test_common, then TEST_VARIANTS should be used in the concrete test_cscs_gh200, test_cscs_amd_rocm to pick which to actually run.
I see now, thanks. It's kind of weird how it's done: TEST_VARIANTS is used to exclude the variants which do not match its content in the exclude_variants_rules ...
2 participants
No description provided.