fix arm sgl-kernel link issue #7066
Conversation
There was a problem hiding this comment.
Summary of Changes
Hello @zhyncs, I'm Gemini Code Assist1! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!
This pull request aims to resolve a build-time linking problem affecting the sgl-kernel compilation process on ARM systems. It achieves this by making specific adjustments to the build environment setup within the docker container used for cross-compilation or native ARM builds.
Highlights
- Build Fix: Addresses a linking issue specifically encountered when building the sgl-kernel on ARM (aarch64) architecture.
- Build Environment: Modifies the build script to adjust environment variables and build settings within the docker container used for ARM builds.
Using Gemini Code Assist
The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.
Invoking Gemini
You can request assistance from Gemini at any point in your pull request via creating an issue comment (i.e. comment on the pull request page) using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands.
| Feature | Command | Description |
|---|---|---|
| Code Review | /gemini review |
Performs a code review for the current pull request in its current state. |
| Pull Request Summary | /gemini summary |
Provides a summary of the current pull request in its current state. |
| Comment | @gemini-code-assist | Responds in comments when explicitly tagged, both in issue comments and review comments. |
| Help | /gemini help |
Displays a list of available commands. |
Customization
To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.
Limitations & Feedback
Gemini Code Assist is currently in preview and may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments to provide feedback.
You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.
Footnotes
-
Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configureGemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩
There was a problem hiding this comment.
Code Review
This PR aims to fix a linking issue for ARM sgl-kernel builds by modifying the build.sh script. It introduces a setting for CMAKE_BUILD_PARALLEL_LEVEL and an update to LD_LIBRARY_PATH.
The LD_LIBRARY_PATH modification seems like a direct and plausible fix. However, the CMAKE_BUILD_PARALLEL_LEVEL setting, as currently placed in the host environment, is unlikely to affect the build parallelism inside the Docker container where it would be needed. For it to be effective, it should be set within the container's script or explicitly passed to Docker.
Providing details in the PR description about the specific link issue and the rationale for these changes would be beneficial.
| @@ -15,6 +15,7 @@ echo "ARCH: $ARCH" | |||
| if [ ${ARCH} = "aarch64" ]; then | |||
| LIBCUDA_ARCH="sbsa" | |||
| BUILDER_NAME="pytorch/manylinuxaarch64-builder" | |||
| CMAKE_BUILD_PARALLEL_LEVEL=16 | |||
There was a problem hiding this comment.
The variable CMAKE_BUILD_PARALLEL_LEVEL is set here in the host shell's environment. However, the main build process (using uv build, which in turn uses scikit-build-core) executes inside a Docker container. Environment variables from the host shell are not automatically inherited by the Docker container's environment unless they are explicitly passed using an option like -e with the docker run command.
Consequently, this setting for CMAKE_BUILD_PARALLEL_LEVEL is unlikely to affect the parallelism of the build process running inside the container. If the intention is to limit build parallelism for aarch64 architectures within the Docker environment (e.g., to mitigate high memory usage during linking), this variable needs to be defined and exported inside the script that bash -c "..." executes within the docker run command, or it must be explicitly passed to the docker run command using the -e CMAKE_BUILD_PARALLEL_LEVEL flag.
| @@ -40,6 +41,7 @@ docker run --rm \ | |||
| tar -xzf cmake-\${CMAKE_VERSION_MAJOR}.\${CMAKE_VERSION_MINOR}-linux-${ARCH}.tar.gz | |||
| mv cmake-\${CMAKE_VERSION_MAJOR}.\${CMAKE_VERSION_MINOR}-linux-${ARCH} /opt/cmake | |||
| export PATH=/opt/cmake/bin:\$PATH | |||
| export LD_LIBRARY_PATH=/lib64:\$LD_LIBRARY_PATH | |||
There was a problem hiding this comment.
Adding /lib64 to the LD_LIBRARY_PATH within the Docker container's build script is a common strategy to resolve issues where shared libraries required by the linker or at runtime are not found in standard search paths. This change appears to be a direct measure to address the "arm sgl-kernel link issue" mentioned in the PR title.
It's advisable to confirm that /lib64 is the correct directory containing the necessary libraries for the aarch64 builder environment and that this modification doesn't inadvertently cause incorrect library versions to be prioritized if more specific versions are expected from other locations in LD_LIBRARY_PATH (though prepending /lib64 while preserving the existing $LD_LIBRARY_PATH is a standard practice).
Motivation
Modifications
Checklist