Add opt-in GPU spreading for the parallel test suite#588
Draft
michel2323 wants to merge 1 commit into
Draft
Conversation
ONEAPI_TEST_SPREAD_GPUS=1 pins each test worker process to a distinct GPU via ZE_AFFINITY_MASK (claimed round-robin through an atomic mkdir counter, set before `using oneAPI` so the Level Zero driver picks it up at init). This spreads the suite across all tiles instead of oversubscribing device 0. device() is task-local and Malt runs each test in a fresh task, so a device! in init_worker_code would not stick — process-level pinning is the robust approach. Default (unset) keeps every worker on the first device, preserving single-tile oversubscription which is useful for surfacing contention bugs. Verified: 6 concurrent claimers -> 6 distinct device UUIDs; real harness with --jobs=4 spreads cleanly (SUCCESS). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #588 +/- ##
==========================================
- Coverage 80.92% 80.89% -0.04%
==========================================
Files 48 48
Lines 3234 3234
==========================================
- Hits 2617 2616 -1
- Misses 617 618 +1 ☔ View full report in Codecov by Harness. 🚀 New features to boost your workflow:
|
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
Adds an opt-in mechanism to spread the parallel test suite across multiple GPU tiles instead of oversubscribing device 0.
With
ONEAPI_TEST_SPREAD_GPUS=1, each test worker process is pinned to a distinct GPU viaZE_AFFINITY_MASK(claimed round-robin through an atomicmkdircounter, set beforeusing oneAPIso the Level Zero driver picks it up at init), passed through the ParallelTestRunnerenvkwarg.device()is task-local and Malt runs each test in a fresh task, so adevice!ininit_worker_codewould not stick — process-level pinning is the robust approach.Notes
Fully opt-in: when
ONEAPI_TEST_SPREAD_GPUSis unset, behavior is identical to currentmain(every worker stays on the first device).ZE_AFFINITY_MASKis a standard Level Zero variable, stack-agnostic.🤖 Generated with Claude Code