Avoid inlining FnPtr::new in StructGenerator bindings - #566
Open
jamienicol wants to merge 2 commits into
Open
Conversation
This avoids pathologically slow compilation for ARMv7 targets with `-Cprofile-generate`. Bindings emitted by StructGenerator call FnPtr::new hundreds of times. Under `-Cprofile-generate`, each inlined call adds profiling instrumentation to the same loader function. On ARMv7 targets, LLVM can then spend an unreasonable amount of time in the instruction scheduling stage of code generation for this function. In local testing, compilation remained in this stage after more than 90 minutes. Preventing FnPtr::new from being inlined avoids this, restoring normal compilation times.
Author
|
See servo/gleam#235 for the motivation behind this. Adding bindings for an additional extension increases |
Author
|
The clippy warning seems unrelated to my change, but I pushed a second commit to resolve it anyway. Does this look okay to merge @MarijnS95 ? I realise it's probably a rustc/llvm bug, but this repo does at least seem to have history of avoiding inlining for the sake of compilation times (e.g. #495) |
jamienicol
added a commit
to jamienicol/firefox
that referenced
this pull request
Aug 6, 2026
…. r?#gfx-reviewers Update gleam to include servo/gleam#235, adding bindings for ARB_base_instance/EXT_base_instance extensions. Unfortunately this gleam PR cannot be merged until gl_generator merges rust-windowing/gl-rs#566 and releases a new version. Without the gl_generator fix, arm32 profile-generate builds take hours due to what is presumably a rustc/LLVM bug. In the meantime, use [patch.crates-io] to vendor specific git revisions of gl_generator and gleam including the required fixes. Once both the gl_generator and gleam pull requests have been merged and new versions of the respective crates have been published, we should switch back to using the crates.io versions.
jamienicol
added a commit
to jamienicol/firefox
that referenced
this pull request
Aug 6, 2026
…. r?#gfx-reviewers Update gleam to include servo/gleam#235, adding bindings for ARB_base_instance/EXT_base_instance extensions. Unfortunately this gleam PR cannot be merged until gl_generator merges rust-windowing/gl-rs#566 and releases a new version. Without the gl_generator fix, arm32 profile-generate builds take hours due to what is presumably a rustc/LLVM bug. In the meantime, use [patch.crates-io] to vendor specific git revisions of gl_generator and gleam including the required fixes. Once both the gl_generator and gleam pull requests have been merged and new versions of the respective crates have been published, we should switch back to using the crates.io versions.
jamienicol
added a commit
to jamienicol/firefox
that referenced
this pull request
Aug 6, 2026
…. r?#gfx-reviewers Update gleam to include servo/gleam#235, adding bindings for ARB_base_instance/EXT_base_instance extensions. Unfortunately this gleam PR cannot be merged until gl_generator merges rust-windowing/gl-rs#566 and releases a new version. Without the gl_generator fix, arm32 profile-generate builds take hours due to what is presumably a rustc/LLVM bug. In the meantime, use [patch.crates-io] to vendor specific git revisions of gl_generator and gleam including the required fixes. Once both the gl_generator and gleam pull requests have been merged and new versions of the respective crates have been published, we should switch back to using the crates.io versions. Differential Revision: https://phabricator.services.mozilla.com/D316861
lando-worker Bot
pushed a commit
to mozilla-firefox/firefox
that referenced
this pull request
Aug 7, 2026
…. r=gfx-reviewers,supply-chain-reviewers,ahale Update gleam to include servo/gleam#235, adding bindings for ARB_base_instance/EXT_base_instance extensions. Unfortunately this gleam PR cannot be merged until gl_generator merges rust-windowing/gl-rs#566 and releases a new version. Without the gl_generator fix, arm32 profile-generate builds take hours due to what is presumably a rustc/LLVM bug. In the meantime, use [patch.crates-io] to vendor specific git revisions of gl_generator and gleam including the required fixes. Once both the gl_generator and gleam pull requests have been merged and new versions of the respective crates have been published, we should switch back to using the crates.io versions. Differential Revision: https://phabricator.services.mozilla.com/D316861
github-actions Bot
pushed a commit
to servo/webrender
that referenced
this pull request
Aug 9, 2026
…. r=gfx-reviewers,supply-chain-reviewers,ahale Update gleam to include servo/gleam#235, adding bindings for ARB_base_instance/EXT_base_instance extensions. Unfortunately this gleam PR cannot be merged until gl_generator merges rust-windowing/gl-rs#566 and releases a new version. Without the gl_generator fix, arm32 profile-generate builds take hours due to what is presumably a rustc/LLVM bug. In the meantime, use [patch.crates-io] to vendor specific git revisions of gl_generator and gleam including the required fixes. Once both the gl_generator and gleam pull requests have been merged and new versions of the respective crates have been published, we should switch back to using the crates.io versions. Differential Revision: https://phabricator.services.mozilla.com/D316861
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.
This avoids pathologically slow compilation for ARMv7 targets with
-Cprofile-generate.Bindings emitted by StructGenerator call FnPtr::new hundreds of times. Under
-Cprofile-generate, each inlined call adds profiling instrumentation to the same loader function. On ARMv7 targets, LLVM can then spend an unreasonable amount of time in the instruction scheduling stage of code generation for this function. In local testing, compilation remained in this stage after more than 90 minutes.Preventing FnPtr::new from being inlined avoids this, restoring normal compilation times.