Repository navigation
test(miopen)[ALMIOPEN-2867]: add find/optimizer/misc gtest families to test_categories.yaml (1/9) - #13403
Conversation
…es.yaml (1/9) Chunk 1/9: find/allocator/misc suites, Adam/AmpAdam, pure unprefixed suites. Jenkins runs these via its *FP32*/*FP16*/... filter; no pattern selected them. Placement follows measured per-test time (Jenkins build #1034 and TheRock gfx942 logs, median ms): Smoke/unprefixed cases also run in standard; Full/ cases run in comprehensive/full only. Newly selected: 1094 cases (1094 that Jenkins runs, 0 beyond Jenkins). Added per tier (cases, est. serial time; divide by 4 shards for per-shard): standard +74 cases, +0.6 min comprehensive +1094 cases, +1.3 min full +1094 cases, +1.3 min - Full/ cases are excluded from standard (it is already at its time budget). Checked against the TheRock 2026-09-22 miopen_gtest (real --gtest_filter): nothing removed in any tier or arch variant, quick/ffm-quick unchanged, and the added cases equal an independently derived expectation. Part of a 9-way split of one change; chunks touch disjoint cases and may merge in any order, one at a time (they edit the same YAML anchors, so rebase between merges).
⏳ Waiting on Required Checks⏳ Required checks have not all concluded yet. This bot run is exiting without failing so those checks can report independently.
📖 Need help? See the Policy FAQ for details on every check and how to fix failures. |
|
⏳ Waiting for required checks to start or finish:
|
aalabsi-amd
changed the title
JonathanLichtnerAMD
left a comment
JonathanLichtnerAMD
left a comment
There was a problem hiding this comment.
LGTM. One minor comment
| @@ -64,10 +64,26 @@ positive_patterns: &positive_patterns | |||
| - "*/GPU_Dropout*" | |||
| - "*/GPU_MIOpenDriver*" | |||
| - "*/GPU_HipGraphExist*" | |||
| # Jenkins-parity: Smoke/ and unprefixed cases also run in standard; Full/ cases run in comprehensive/full only. | |||
| # Chunk: find/allocator/misc suites, Adam/AmpAdam, pure unprefixed suites. | |||
There was a problem hiding this comment.
This comment seems to imply that all the Full/* tests that were just added only run in comprehensive/full, but that is only true for a subset of these. Are these primarily Smoke/* tests? If so, we could add that explicitly here (or add one generic Smoke/* line, maybe immediately before Standard/*, below)
Maybe moving to Smoke/, Standard/, Full/* could help clean this up, so I am okay if you want to defer this
There was a problem hiding this comment.
One more note: you mention testing on standard, comprehensive and full in the TestPlan. Can we make this a checkbox list and also add doing sweeps on other architectures, such as gfx90a, navi, etc, or we might get surprised by nightly failures. The rock tests some architectures that our Jenkins CI didn't cover, and we have different timeout constraints for the rock's GHA vs Jenkins. Thanks!
|
Closing as this will be part of the full PR #13107 |
JIRA ID : ALMIOPEN-2867 Chunk 1/9: find/allocator/misc suites, Adam/AmpAdam, pure unprefixed suites.
Motivation
Chunk 1/9 of jenkins tests parity: find/allocator/misc suites, Adam/AmpAdam, pure unprefixed suites.
Technical Details
Jenkins runs these via its FP32/FP16/... filter; no pattern selected them. Placement follows measured per-test time (Jenkins old run and TheRock gfx942 logs, median ms): Smoke/unprefixed cases also run in standard; Full/ cases run in comprehensive/full only.
Newly selected: 1094 cases (1094 that Jenkins runs, 0 beyond Jenkins). Added per tier (cases, est. serial time; divide by 4 shards for per-shard):
standard +74 cases, +0.6 min
comprehensive +1094 cases, +1.3 min
full +1094 cases, +1.3 min
Checked against the TheRock 2026-09-22 miopen_gtest (real --gtest_filter): nothing removed in any tier or arch variant, quick/ffm-quick unchanged, and the added cases equal an independently derived expectation.
Part of a 9-way split of one change; chunks touch disjoint cases and may merge in any order, one at a time (they edit the same YAML anchors, so rebase between merges).
Test Plan
Test it with three cats standard, comprehensive and full and make sure tests pass fully and time added is reasonable and within budget.
Test Result
Submission Checklist