patchpkg: patch python to use devbox CUDA libs #2296
Merged
Conversation
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
gcurtis force-pushed the gcurtis/cuda branch 5 times, most recently from 7507bdf to 04f319e Compare Automatically patch python to use any `cudaPackages.*` packages that are in devbox.json. This will only work if the CUDA drivers are already installed on the host system. The patching process is: 1. When generating the patch flake, look for the system's `libcuda.so` library (installed by the driver) and copy it into the flake's directory. 2. Nix copies the flake's source directory (and therefore libcuda.so) into the nix store when building it. 3. The flake calls `devbox patch` which adds a `DT_NEEDED` entry to the python binary for `libcuda.so`. It also adds the lib directories of any other `cudaPackages.*` packages that it finds in the environment.
gcurtis force-pushed the gcurtis/cuda branch from 04f319e to 9c758fd Compare 
savil reviewed Sep 25, 2024
savil approved these changes Sep 25, 2024
Collaborator
savil left a comment
savil left a comment There was a problem hiding this comment.
LGTM, modulo one question below
mikeland73 reviewed Sep 26, 2024
| return refs, nil | ||
| } | ||
| | ||
| func (d *DerivationBuilder) findCUDA(ctx context.Context, out *packageFS) error { |
Collaborator
There was a problem hiding this comment.
Food for future thought: Is there a way to do this in a more plugin-like way? Maybe we allow built in plugins to use go code (or in the future compile plugins as needed).
Collaborator Author
There was a problem hiding this comment.
I like the idea of aligning plugins and patching. This code runs as part of the flake builder (in the nix build sandbox), so plugins would just need something similar.
gcurtis force-pushed the gcurtis/cuda branch from 470e552 to 490bfe5 Compare 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.
Automatically patch python to use any
cudaPackages.*packages that are in devbox.json. This will only work if the CUDA drivers are already installed on the host system.The patching process is:
libcuda.solibrary (installed by the driver) and copy it into the flake’s directory.devbox patchwhich adds aDT_NEEDEDentry to the python binary forlibcuda.so. It also adds the lib directories of any othercudaPackages.*packages that it finds in the environment.