-
-
Notifications
You must be signed in to change notification settings - Fork 33.2k
GH-136895: Update JIT builds to use LLVM 20 #140329
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
base: main
Are you sure you want to change the base?
Conversation
With LLVM 20, individual files are greater than the 100MiB single file limit for items checked into git. Therefore, this PR pulls down binaries from GitHub releases, as `.tar.xz` files to additionally maximize compression ratio. Currently this is somewhat of a first draft, as there are things like hash checking needed to be done.
This reverts commit e6450de.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
(CI changes 👍)
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM, thanks for addressing the comments.
Alright, this took me longer than expected for two reasons:
Bumping to LLVM 20 required some changes to the infrastructure we use to grab LLVM as a dependency on Windows. As of 20, LLVM now contains files that exceed GitHub's allowable size, so checking these directly into
cpython-bin-deps
' tree was not an option. Instead, this PR adds the ability to pull from release artifacts (see https://github.com/python/cpython-bin-deps/releases/tag/llvm-20.1.8.0). This also makes the process of bumping LLVM for Windows a little less hairy. Thank you, @emmatyping, for much of this code, and thank you, @zware, for publishing the release artifact for me 🙏 !For macOS x86_64 debug builds, external symbol references generated by LLVM 20 can exceed the ±2GB PC-relative addressing range that
patch_32r
requires. This manifested as assertion failures in free-threading tests (see https://github.com/savannahostrowski/cpython/actions/runs/18438327725/job/52537027963?pr=10 as an example). After going down a rabbit hole, I believe the correct fix here is to implement x86_64 trampolines (similar to our existing aarch64 implementation) to handle out-of-range symbols...and so I've done that. AFAICT, trampolines are only needed on macOS right now, as the other target platforms have different relocation mechanisms.