CI: run CUDA.jl's tests with 3 CPU threads - #955
Merged
Merged
Conversation
The CUDA agents budget 2 threads per job, which limits CUDA.jl's test suite to 2 workers even though it is compile-bound and the CI hosts have CPU to spare. CUDA.jl's runtests.jl still caps the worker count by the available GPU memory.
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #955 +/- ##
=======================================
Coverage 86.54% 86.54%
=======================================
Files 29 29
Lines 5729 5729
=======================================
Hits 4958 4958
Misses 771 771 ☔ 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
The CUDA agents set
JULIA_CPU_THREADS=2, so ParallelTestRunner runs CUDA.jl's test suite with 2 workers. The suite is compile-bound: on gpuci a typical run keeps both workers at 100% CPU, peaks at ~1 GB of GPU memory per worker (of 5 GB on an A100 MIG slice), and takes 50-80 minutes. Meanwhile the host has ~15 of its 48 threads and ~300 GB of RAM idle, and the cuda queue is backlogged (median wait ~2 hours).This raises the budget to 3 threads for the downstream CUDA.jl step. CUDA.jl's
runtests.jlstill caps the worker count by free GPU memory (1 GiB per worker), so small devices are unaffected. Same change as JuliaGPU/CUDA.jl#3298.