search

nv-legate / cunumeric

An Aspiring Drop-In Replacement for NumPy at Scale
https://docs.nvidia.com/cunumeric/24.06/
Apache License 2.0
613 stars 69 forks source link

Add Cmake flag to ensure we're propagating thread_count #1102

Closed reazulhoque closed 9 months ago

reazulhoque commented 10 months ago

Thread count is not being propagated to Cmake. This fixes that.

copy-pr-bot[bot] commented 10 months ago

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

magnatelee commented 10 months ago

@reazulhoque please add the sign key to your GitHub account. the policy is that we don't accept PRs with unverified or unsigned commits.

magnatelee commented 10 months ago

also I don't recall we had any issue with the parallel build... can you describe what you're trying to fix with this PR?

reazulhoque commented 10 months ago

@magnatelee We are propagating to cmake what users have provided with -j flag to install.py.

Without this fix user provided option will not be respected by cmake.

manopapad commented 10 months ago

To be clear, install.py does pass -j directly to ninja/make, which is the last stage of the build, and covers all C++ compilation. We would have to pass this to cmake if we were using cmake --build rather than ninja/make directly. The problem is that compilation of the rust profiler expects to read -j from cmake.