perf(rust/tla): skip compute_tla if there is no module use TLA#8487
Merged
hyf0 merged 4 commits intorolldown:mainfrom Feb 27, 2026
Merged
perf(rust/tla): skip compute_tla if there is no module use TLA#8487hyf0 merged 4 commits intorolldown:mainfrom
hyf0 merged 4 commits intorolldown:mainfrom
Conversation
Member
|
Is count a bit of unnecessary? I would go for having a big flag under module scanner and return a |
Contributor
Author
Hi, thanks for the feedback. I think the counter is necessary because it makes it easier to check if any other TLA modules remain in the |
Member
|
Well done! |
Contributor
There was a problem hiding this comment.
Pull request overview
This PR implements a performance optimization for the compute_tla function by tracking the number of modules using top-level await (TLA) during the scan stage. The implementation adds an early return when no TLA modules exist, avoiding unnecessary computation for the common case where projects don't use TLA.
Changes:
- Added
tla_module_countfield to track TLA-using modules throughout the pipeline - Implemented early return in
compute_tlawhen count is zero - Updated merge logic in
ScanStageCacheto maintain accurate counts during partial scans
Reviewed changes
Copilot reviewed 5 out of 5 changed files in this pull request and generated 1 comment.
Show a summary per file
| File | Description |
|---|---|
| crates/rolldown/src/types/scan_stage_cache.rs | Added module_has_tla helper and logic to track TLA count during cache merge operations |
| crates/rolldown/src/stages/scan_stage.rs | Added tla_module_count field to scan stage output structures |
| crates/rolldown/src/stages/link_stage/mod.rs | Added tla_module_count field to LinkStage struct |
| crates/rolldown/src/stages/link_stage/compute_tla.rs | Added early return optimization when no TLA modules exist |
| crates/rolldown/src/module_loader/module_loader.rs | Added TLA counting during module loading |
hyf0
approved these changes
Feb 27, 2026
2b3172f to
b4a6959
Compare
✅ Deploy Preview for rolldown-rs canceled.
|
shulaoda
added a commit
that referenced
this pull request
Mar 5, 2026
## [1.0.0-rc.7] - 2026-03-05 ⚡ Smarter Code Generation Defaults - DCE-only minification and smart constant inlining are now enabled by default - Produces cleaner, smaller output bundles without requiring explicit configuration 💡 LLM-Friendly Bundle Analyzer Reports - New markdown output format for the bundle analyzer plugin with bundle summaries, module graphs, dependency chains, and optimization suggestions - Optimization suggestions now also recommend using the entriesAware option when common chunks contain modules only reachable from specific entries ### 💥 BREAKING CHANGES - enable minify: 'dce-only' by default (#8465) by @IWANABETHATGUY - settings `inlineConst: { mode: 'smart', pass: 1}` by default (#8444) by @IWANABETHATGUY ### 🚀 Features - binding: add original getter to BindingMagicString (#8533) by @IWANABETHATGUY - native-magic-string: add `offset` property support (#8531) by @IWANABETHATGUY - add `output.strict` option to control `"use strict"` directive emission (#8489) by @Copilot - watch: expose `watcher.compareContentsForPolling` (#8526) by @hyf0 - watch: use new watcher to support watch mode (#8475) by @hyf0 - rust/watch: handle bulk-change (#8466) by @hyf0 - add LLM-friendly markdown output format to bundle analyzer plugin (#8242) by @IWANABETHATGUY ### 🐛 Bug Fixes - expose `plugins` on `NormalizedInputOptions` for `buildStart` hook (#8521) by @Copilot - only uppercase facade symbols in JSX preserve mode (#8519) by @IWANABETHATGUY - binding: export BindingResult in generated dts header (#8537) by @minsoo-web - pre-resolve paths option to avoid `invoke_sync` deadlock (#8518) by @IWANABETHATGUY - remove debug-only jsx_preset and UntranspiledSyntaxError (#8511) by @IWANABETHATGUY - apply `topLevelVar` to exported `const`/`let` declarations (#8507) by @IWANABETHATGUY - rolldown_plugin_vite_web_worker_post: avoid replacing `new.target` (#8488) by @sapphi-red - update copyright year to 2026 (#8486) by @maciekzygmunt ### 🚜 Refactor - rust: use Oxc's SymbolFlags::ConstVariable instead of custom IsConst flag (#8543) by @Dunqing - rust: remove FacadeScoping, use Scoping::create_symbol for facade symbols (#8540) by @Dunqing - rust/watch: remove hacky `reset_closed_for_watch_mode` (#8530) by @hyf0 - binding: return &str instead of String in filename() getter (#8534) by @IWANABETHATGUY - rust: remove old watch mode implementation (#8525) by @hyf0 - rust/watch: simply watch logic in the binding layer (#8516) by @hyf0 - rust/watch: tweak struct/function names (#8464) by @hyf0 ### 📚 Documentation - explain how external modules work in rolldown (#8457) by @sapphi-red - add some diagrams using graphviz (#8499) by @sapphi-red - use `vitepress-plugin-graphviz` (#8498) by @sapphi-red - list s390x/ppc64le prebuilt binaries (#8495) by @crusty-voidzero - fix error type for `RolldownBuild.generate` and others (#8490) by @sapphi-red ### ⚡ Performance - string_wizard: reduce allocations and add ASCII fast paths (#8541) by @IWANABETHATGUY - use IndexBitSet to replace IndexVec<XXXIdx, bool> for module/stmt inclusion tracking (#8503) by @IWANABETHATGUY - plugin: use IndexBitSet to optimize skipped plugins checking (#8497) by @ShroXd - rust/tla: skip compute_tla if there is no module use TLA (#8487) by @ShroXd ### 🧪 Testing - node/watch: make watch tests run in concurrent and retry-able (#8512) by @hyf0 - add test case for static flag tree-shaking (#8476) by @IWANABETHATGUY - migrate post-banner sourcemap-with-shebang to Rust (#8477) by @Copilot ### ⚙️ Miscellaneous Tasks - vscode: `formatOnSave` for markdown files using oxc formatter (#8536) by @minsoo-web - deps: update test262 submodule for tests (#8528) by @sapphi-red - remove `retry` workaround from output paths test fixtures (#8520) by @Copilot - docs: add Shuyuan Wang (h-a-n-a) and remove from acknowledgements (#8509) by @Copilot - consolidate top_level_var test cases using configVariants (#8508) by @IWANABETHATGUY - add s390x and ppc64le linux gnu targets (#8493) by @Brooooooklyn ###◀️ Revert - fix(rolldown): increase tokio blocking threads size for watch mode (#8517) by @hyf0 ### ❤️ New Contributors * @minsoo-web made their first contribution in [#8536](#8536) * @crusty-voidzero made their first contribution in [#8495](#8495) * @maciekzygmunt made their first contribution in [#8486](#8486) Co-authored-by: shulaoda <165626830+shulaoda@users.noreply.github.com>
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
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.
A todo in
compute_tlaindicates that this process could be skipped when there is no module that uses TLA — which is also the common case for many real-world projects.This PR tracks the count of TLA-using modules during the scan stage and stores it in
tla_module_count. Link stage will check this counter and return early if the count is zero.