-
Notifications
You must be signed in to change notification settings - Fork 101
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
GitLab CI testing expansion #1644
base: develop
Are you sure you want to change the base?
Conversation
… run when develop was changed (Issue 1611.3)
Had to update some of the tests to not build with openMPTarget as they aren't supported.
…tel_compiler_build_instructions
@long58 I started reading your PR. First, thank you for fixing the multi-project! I was actually looking at that today, and couldn’t see what was wrong. I think GitLab changed the syntax at some point to remove those quotes. So, thanks for the help. We can see that the status reports a failure named "skipped-draft-pr". I think this will go away if you merge "develop" in you branch. I’ll complete the review tomorrow. |
@long58 good work on this so far. I'll review in more detail tomorrow. For now, I suggest that we remove the OpenMP target testing from this branch. Then, we can work on the OpenMP target stuff in a separate PR. It will take some effort to try various compilers to figure out what works. We've always had difficulty with compiler support for that back-end. @adrienbernede I think it would be best to test RAJAPerf against the PR branch to be merged. However, it would be sufficient I think to see whether RAJA Perf passes or fails (so we can fix breakages in RAJA Perf) without requiring that check to pass before merging the RAJA branch. Is that what you have in mind? |
@rhornung67 the new data point I find in your comment is that you don’t want to bind the RAJA PR status to the RAJAPerf status. I am find with that, and it can be achieve removing the |
@rhornung67 The intel 19.1.2 + gcc 10.3.1 job is timing out on poodle.
Given that poodle is quite small and allocations limited, I would argue that our CI on it should be more targeted. Currently any ruby job is also run on poodle, is that really necessary ? |
@adrienbernede I think we can remove that job. No users are exercising that compiler anymore, as far as I know. I suggest adding a job for intel/2023.2.1, which is the latest oneapi compiler. |
@@ -31,7 +31,7 @@ variables: | |||
# Project specific variants for poodle | |||
PROJECT_POODLE_VARIANTS: "~shared +openmp +vectorization +tests" | |||
# Project specific deps for poodle | |||
PROJECT_POODLE_DEPS: "" | |||
PROJECT_POODLE_DEPS: "^blt@develop " |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@long58 @rhornung67 let’s not deal with this in this PR, but blt@develop is heavily used in the CI. We should check whether that is still necessary, it’s not good to point at a moving target in CI.
- Uncommented test cases for omptarget - Deactivated lassen omptarget testing - made removed strategy: depend from rajaperf multiproject testing so that the status of rajaperf will not influence raja's testing status
Fixed this in my most recent commit. |
I went ahead and deactivated the OpenMP Target job, so that we can spend more time on it in a separate PR, as you suggested. |
Yes, let's tackle OpenMP target testing in a separate PR. @long58 I owe you some guidance on compiler choice. I hope to get back to you on this sometime this week. |
@long58 while trying to fix the conflicts in radiuss-spack-configs for you, I stumbled on something: |
@long58 apparently Spack does not like your new compiler. I’m trying to find why... |
The Spack error message is pretty cryptic: ==> Error: concretization failed for the following reasons:
I double-checked that service user |
@adrienbernede It looks like the update from spack version 02-18 to 05-26 broke the concretization of the new compiler. Any thoughts as to why this might be the case? Everything works when I change the spack version back to the old one. |
Summary
In order to accomplish this, I needed to make changes to the radiuss_spack_configs github project to add new compilers, and modify the raja and camp package.py files. I plan to make another pull request there integrating those changes, but for now, my branch of radiuss spack configs (used here) is up to date with the latest version of radiuss_spack_configs/develop.
Also, some (~9) of the OpenMPTarget tests on lassen are not consistent in whether they pass or fail. They will pass sometimes, and fail other times, and I'm really not entirely sure why.
This feature completes all objectives of Github issue #1611