Repository navigation
Run the known-failing GPU-buffer CI lanes nightly, not on every PR - #187
Merged
Merged
Conversation
The GPU Buffer lanes of the two emulated-hardware AIS workflows fail on every run. Their client step is continue-on-error, so the jobs show green, but on every pull request and develop push they cost ~18 runner minutes and reported nothing, and the two-VM GPU job was the slowest job in CI. Split each workflow's matrix into two jobs: - The host-buffer job (gpu-direct, two-vm-gpu-direct) runs on pull requests, pushes, and manual runs as before, and skips the new scheduled run, which would only repeat it. - The GPU-buffer job (gpu-direct-gpu-buffer, two-vm-gpu-direct-gpu-buffer) runs on a daily schedule (06:41 UTC) and on manual runs. It's still non-fatal, and still reports the known failure in the job summary. The GPU-buffer job shares the host-buffer job's container, env, and steps through YAML anchors, so the two can't drift apart. Neither job takes its name from the matrix any more, since a skipped job shows its name unexpanded. When the GPU-direct transfer works under emulation, move the GPU Buffer entries back into the host-buffer jobs as fatal, and delete the GPU-buffer jobs and the schedules. Co-Authored-By: Claude <noreply@anthropic.com>
derobins
requested review from
akolliasAMD,
gaoikawa,
jordan-turbofish,
kurtmcmillan,
riley-dixon,
sbates130272 and
zbyrne
as code owners
October 13, 2026 11:23
Contributor
There was a problem hiding this comment.
🟢 Approval recommended
The event gating and shared job configuration correctly implement the described CI split without affecting host-buffer PR coverage.
0 open findings
What changed in this PR
Moves known-failing GPU-buffer CI lanes from per-PR execution to nightly/manual runs while preserving host-buffer coverage.
Changes:
- Adds daily 06:41 UTC schedules.
- Splits host- and GPU-buffer tests into separately gated jobs.
- Reuses container, environment, and steps through YAML anchors.
| File | Description |
|---|---|
.github/workflows/hipobject-hardware-test-two-vm-gpu-direct.yml |
Schedules the two-VM GPU-buffer lane nightly. |
.github/workflows/hipobject-hardware-test-gpu-direct.yml |
Schedules the single-VM GPU-buffer lane nightly. |
🧠 Review effort: Balanced
💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The GPU Buffer lanes of the two emulated-hardware AIS workflows fail on every run. Their client step is continue-on-error, so the jobs show green, but on every pull request and develop push they cost ~18 runner minutes and reported nothing, and the two-VM GPU job was the slowest job in CI.
Split each workflow's matrix into two jobs:
The GPU-buffer job shares the host-buffer job's container, env, and steps through YAML anchors, so the two can't drift apart. Neither job takes its name from the matrix any more, since a skipped job shows its name unexpanded.
When the GPU-direct transfer works under emulation, move the GPU Buffer entries back into the host-buffer jobs as fatal, and delete the GPU-buffer jobs and the schedules.