You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The proposed config would run the tests once with the unmodified sycl branch and once with the tag update and compare the results, ignoring any fails that happen in both runs. This could help to identify unrelated fails, which we do have a high rate of due to how fast intel-llvm moves contribution wise. Ideally this would extend to running all the e2e jobs like this but in practice our hardware resources likely wouldn't support that much extra load, we'd probably need to pick one or two jobs to double up instead.
The text was updated successfully, but these errors were encountered:
The big issue where e2e failures are a problem is L0 on PVC. It should get better hopefully now that intel/llvm is also running tests on PVC. Nvm, that intel/llvm patch no longer includes adding a PVC job... :(
What we could maybe do is have a nightly job that runs against latest intel/llvm (maybe even runs twice, to detect flaky tests), which would then create a list of tests to xfail for the normal runs.
The proposed config would run the tests once with the unmodified sycl branch and once with the tag update and compare the results, ignoring any fails that happen in both runs. This could help to identify unrelated fails, which we do have a high rate of due to how fast intel-llvm moves contribution wise. Ideally this would extend to running all the e2e jobs like this but in practice our hardware resources likely wouldn't support that much extra load, we'd probably need to pick one or two jobs to double up instead.
The text was updated successfully, but these errors were encountered: