ci: fix the self-hosted gpu job and make it opt-in - #20
Merged
Conversation
The gpu job built via 'make -C cuda' and ran ./cuda/test / ./cuda/bench — none of which exist (there is no cuda/Makefile; the CUDA targets are CMake's amr_cuda_tests / amr_cuda_bench). It also ran on every push, so when the self-hosted A2000 is offline the job sits queued indefinitely and the whole run never completes (the prior dev run was auto-cancelled after ~24h). Build via CMake to the real targets, and gate the job to workflow_dispatch so GPU runtime validation is on-demand (run it when the runner is online) instead of stalling every merge. Drop continue-on-error so an opt-in run now reports failures honestly. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Problem
The
gpujob was broken two ways:make -C cuda ARCH=sm_86and ran./cuda/test/./cuda/bench— none exist (nocuda/Makefile; the CUDA targets are CMake'samr_cuda_tests/amr_cuda_bench). It only ever "passed" becausecontinue-on-error: truemasked it while the runner was offline.pushto dev/main. With the self-hosted A2000 offline, a self-hosted job sitsqueuedindefinitely, so the whole run never completes — the last dev run was auto-cancelled after 23h53m, meaning trunk CI never reports a clean green.Fix
amr_cuda_tests,amr_cuda_bench) and run those binaries.workflow_dispatch— GPU runtime validation is now on-demand (dispatch it when the A2000 is online) instead of stalling every merge. All other jobs (lint, cpu ×4, cuda-build compile-check) still run on push/PR and now complete cleanly.continue-on-errorso an opt-in GPU run reports failures honestly.The CUDA backend is still compile-checked on every push/PR by the existing
cuda-buildjob; only the runtime execution moved to opt-in.🤖 Generated with Claude Code