Skip to content

Comments

perf(rolldown): take advantage of tokio blocking threads#6270

Merged
graphite-app[bot] merged 1 commit intomainfrom
09-18-perf_rolldown_take_advantage_of_tokio_blocking_threads
Sep 18, 2025
Merged

perf(rolldown): take advantage of tokio blocking threads#6270
graphite-app[bot] merged 1 commit intomainfrom
09-18-perf_rolldown_take_advantage_of_tokio_blocking_threads

Conversation

@Brooooooklyn
Copy link
Member

@Brooooooklyn Brooooooklyn commented Sep 18, 2025

closes #3754

Benchmark in apps/10000:

1.0.0-beta.38

❯ hyperfine --warmup 3 --runs 10 --command-name "rolldown-latest" "../../node_modules/.bin/rolldown build --config rolldown.config.mjs"
Benchmark 1: rolldown-latest
  Time (mean ± σ):     553.6 ms ±  21.4 ms    [User: 1107.1 ms, System: 2948.2 ms]
  Range (min … max):   518.7 ms … 579.5 ms    10 runs

current

❯ hyperfine --warmup 3 --runs 10 --command-name "rolldown-latest" "../../node_modules/.bin/rolldown build --config rolldown.config.mjs"
Benchmark 1: rolldown-latest
  Time (mean ± σ):     458.1 ms ±   6.5 ms    [User: 1052.2 ms, System: 1092.4 ms]
  Range (min … max):   447.5 ms … 468.8 ms    10 runs

I don't know why codspeed crashes, I added a feature for codspeed so that under codspeed it doesn't use spawn_blocking.

Copy link
Member Author

Brooooooklyn commented Sep 18, 2025


How to use the Graphite Merge Queue

Add the label graphite: merge to this PR to add it to the merge queue.

You must have a Graphite account in order to use the merge queue. Sign up using this link.

An organization admin has enabled the Graphite Merge Queue in this repository.

Please do not merge from GitHub as this will restart CI on PRs being processed by the merge queue.

This stack of pull requests is managed by Graphite. Learn more about stacking.

@netlify
Copy link

netlify bot commented Sep 18, 2025

Deploy Preview for rolldown-rs canceled.

Name Link
🔨 Latest commit 2c840d5
🔍 Latest deploy log https://app.netlify.com/projects/rolldown-rs/deploys/68cc3d787952500008941236

@Brooooooklyn Brooooooklyn force-pushed the 09-18-perf_rolldown_take_advantage_of_tokio_blocking_threads branch from dcd6f8d to a21c8d4 Compare September 18, 2025 09:32
Copy link
Contributor

Copilot AI left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull Request Overview

This PR optimizes the performance of rolldown by leveraging tokio's blocking thread pool for file system operations, resulting in approximately 17% performance improvement (553ms to 458ms in benchmarks). The changes implement platform-specific optimizations for macOS and refactor file loading to use async-friendly blocking operations.

  • Platform-specific thread pool configuration with reduced thread count and increased stack size for macOS
  • Asynchronous file loading using tokio's spawn_blocking for better thread utilization
  • Function signature updates to support owned filesystem instances for thread-safe operations

Reviewed Changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

File Description
crates/rolldown_binding/src/lib.rs Configures platform-specific tokio runtime with macOS optimizations
crates/rolldown/src/utils/load_source.rs Refactors file loading to use tokio blocking threads with owned filesystem
crates/rolldown/src/module_loader/module_task.rs Updates function call to pass cloned filesystem instance

Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.

@github-actions
Copy link
Contributor

github-actions bot commented Sep 18, 2025

Benchmarks Rust

  • target: main(9ecf32c)
  • pr: 09-18-perf_rolldown_take_advantage_of_tokio_blocking_threads(2c840d5)
group                                                        pr                                     target
-----                                                        --                                     ------
bundle/bundle@multi-duplicated-top-level-symbol              1.01    144.8±1.46ms        ? ?/sec    1.00    144.0±1.58ms        ? ?/sec
bundle/bundle@multi-duplicated-top-level-symbol-sourcemap    1.00    157.2±1.04ms        ? ?/sec    1.00    156.6±4.94ms        ? ?/sec
bundle/bundle@rome_ts                                        1.01    230.9±3.28ms        ? ?/sec    1.00    229.0±2.92ms        ? ?/sec
bundle/bundle@rome_ts-sourcemap                              1.01    255.9±2.13ms        ? ?/sec    1.00    254.6±1.74ms        ? ?/sec
bundle/bundle@threejs                                        1.00     82.5±1.40ms        ? ?/sec    1.00     82.8±1.12ms        ? ?/sec
bundle/bundle@threejs-sourcemap                              1.00     93.2±0.96ms        ? ?/sec    1.00     93.0±1.03ms        ? ?/sec
bundle/bundle@threejs10x                                     1.00    932.1±6.74ms        ? ?/sec    1.00    929.5±6.17ms        ? ?/sec
bundle/bundle@threejs10x-sourcemap                           1.01   1027.6±6.10ms        ? ?/sec    1.00   1014.9±4.05ms        ? ?/sec
scan/scan@rome_ts                                            1.01    209.2±2.71ms        ? ?/sec    1.00    207.9±3.58ms        ? ?/sec
scan/scan@threejs                                            1.01     75.0±0.74ms        ? ?/sec    1.00     74.3±0.67ms        ? ?/sec
scan/scan@threejs10x                                         1.02   824.6±21.27ms        ? ?/sec    1.00   810.6±15.86ms        ? ?/sec

@Brooooooklyn Brooooooklyn force-pushed the 09-18-perf_rolldown_take_advantage_of_tokio_blocking_threads branch from a21c8d4 to b95b2be Compare September 18, 2025 09:53
@Brooooooklyn Brooooooklyn marked this pull request as draft September 18, 2025 11:59
@Brooooooklyn Brooooooklyn force-pushed the 09-18-perf_rolldown_take_advantage_of_tokio_blocking_threads branch 2 times, most recently from 29735d9 to 61a2698 Compare September 18, 2025 12:55
@Brooooooklyn Brooooooklyn marked this pull request as ready for review September 18, 2025 13:04
@CPunisher
Copy link

codspeed crashes because it only support up to 500 threads. So you also need to limit the max block thread count in benchmark code.

Copy link
Member

IWANABETHATGUY commented Sep 18, 2025

Merge activity

closes #3754

Benchmark in apps/10000:

### `1.0.0-beta.38`

```
❯ hyperfine --warmup 3 --runs 10 --command-name "rolldown-latest" "../../node_modules/.bin/rolldown build --config rolldown.config.mjs"
Benchmark 1: rolldown-latest
  Time (mean ± σ):     553.6 ms ±  21.4 ms    [User: 1107.1 ms, System: 2948.2 ms]
  Range (min … max):   518.7 ms … 579.5 ms    10 runs
```

### `current`

```
❯ hyperfine --warmup 3 --runs 10 --command-name "rolldown-latest" "../../node_modules/.bin/rolldown build --config rolldown.config.mjs"
Benchmark 1: rolldown-latest
  Time (mean ± σ):     458.1 ms ±   6.5 ms    [User: 1052.2 ms, System: 1092.4 ms]
  Range (min … max):   447.5 ms … 468.8 ms    10 runs
```

I don't know why `codspeed` crashes, I added a feature for `codspeed` so that under `codspeed` it doesn't use `spawn_blocking`.
@graphite-app graphite-app bot force-pushed the 09-18-perf_rolldown_take_advantage_of_tokio_blocking_threads branch from 38cfd74 to 2c840d5 Compare September 18, 2025 17:12
@graphite-app graphite-app bot merged commit 2c840d5 into main Sep 18, 2025
27 checks passed
@graphite-app graphite-app bot deleted the 09-18-perf_rolldown_take_advantage_of_tokio_blocking_threads branch September 18, 2025 17:20
shulaoda pushed a commit that referenced this pull request Sep 22, 2025
## [1.0.0-beta.39] - 2025-09-22

⚡ MacOS Performance Boost

- 10%-30% faster bundling
- Up to 45% in extreme cases

### 🚀 Features

- rolldown: oxc v0.90.0 + other oxc crate updates (#6274) by @Boshen
- support ignoring `/* @vite-ignore */` comment for dynamic imports (#6264) by @IWANABETHATGUY
- dev: merge mergeable queued tasks into one (#6253) by @hyf0
- dev: ignore self imports when computing update for invalidate (#6252) by @sapphi-red
- support `#__NO_SIDE_EFFECTS__` annotation for export default function declaration (#6239) by @IWANABETHATGUY
- rolldown_plugin_dynamic_import_vars: respect `@vite-ignore` comment (#6240) by @shulaoda
- rolldown_plugin_import_glob: support `exhaustive` option (#6236) by @shulaoda
- cross chunk /* @__NO_SIDE_EFFECTS__ */ support (#6207) by @IWANABETHATGUY
- rolldown_plugin_react_refresh_wrapper: make the plugin callable (#6229) by @sapphi-red

### 🐛 Bug Fixes

- handle optional chaining in scope hoisting member expression rewriting (#6278) by @shulaoda
- release: Fix semver commit release format to use plus instead of dash (#6273) by @Copilot
- dev: `ensure_latest_build_output` should work correctly when `eagerBuild: false` (#6262) by @hyf0
- browser: does not work in stackblitz (#6257) by @shulaoda
- only inline constants that are not reassigned (#6247) by @IWANABETHATGUY
- handle lexical declarations in root scope (#6230) by @IWANABETHATGUY

### 🚜 Refactor

- move oxc runtime helper generation from `build.rs` to task generator (#6284) by @shulaoda
- remove unused `SpanExt` (#6266) by @IWANABETHATGUY
- remove `@rolldown-ignore` ignore comment support (#6265) by @IWANABETHATGUY
- enhance CommonJsAstType to track property names (#6250) by @IWANABETHATGUY
- simplify usage for `Address::from` (#6245) by @IWANABETHATGUY
- remove `stmt_info_idx` in StmtInfo (#6244) by @IWANABETHATGUY
- extract `CrossModuleOptimizationRunnerContext` in cross_module_optimization (#6243) by @IWANABETHATGUY
- dev: task-queue based deisgn (#6233) by @hyf0
- extract immutable context from AstScanner (#6238) by @IWANABETHATGUY
- dev: rename `WatcherEventService` to `BuildDriverService` (#6232) by @hyf0
- move FlatOptions to rolldown_common (#6235) by @IWANABETHATGUY
- extract mutable state from ScopeHoistingFinalizerContext (#6234) by @IWANABETHATGUY
- wait `this.load` using futures instead of callbacks (#6222) by @sapphi-red
- rolldown_plugin_transform: remove unused `runtime_resolve_base` option (#6226) by @sapphi-red

### 📚 Documentation

- add "edit this page on GitHub" button (#6241) by @siaeyy

### ⚡ Performance

- rolldown: minor performance optimization (#6271) by @Brooooooklyn
- rolldown: take advantage of tokio blocking threads (#6270) by @Brooooooklyn

### 🧪 Testing

- dev: add rebuild after HMR case (#6260) by @sapphi-red
- dev: support full reload in tests (#6259) by @sapphi-red
- dev: extract waitFor from waitForPathExists (#6261) by @sapphi-red
- add class like cases for empty function optimizations (#6248) by @sapphi-red

### ⚙️ Miscellaneous Tasks

- deps: update github-actions (#6298) by @renovate[bot]
- deps: update dependency dprint/dprint-plugin-typescript to v0.95.11 (#6296) by @renovate[bot]
- deps: update dependency dprint/dprint-plugin-markdown to v0.19.0 (#6295) by @renovate[bot]
- deps: update dependency g-plane/pretty_yaml to v0.5.1 (#6294) by @renovate[bot]
- use a forked version of openharmony-rs/setup-ohos-sdk (#6290) by @Boshen
- deps: update dependency tsdown to v0.15.4 (#6286) by @renovate[bot]
- deps: update dependency rolldown-plugin-dts to v0.16.7 (#6285) by @renovate[bot]
- rust: continue to make `cargo publish --workspace` work (#6281) by @Boshen
- deps: update dependency tsdown to v0.15.3 (#6280) by @renovate[bot]
- rust: fixes for `cargo publish --dry-run --workspace` (#6279) by @Boshen
- deps: update dependency rust to v1.90.0 (#6277) by @renovate[bot]
- deps: update dependency rolldown-plugin-dts to v0.16.6 (#6276) by @renovate[bot]
- adding tracing for all sub stage in LinkStage:link (#6268) by @IWANABETHATGUY
- deps: update oxc-project/setup-node action to v1.0.4 (#6258) by @renovate[bot]
- deps: update github-actions (#6249) by @renovate[bot]
- deps: update dependency tsdown to v0.15.2 (#6242) by @renovate[bot]
- dev: output error when the initial bundle failed (#6228) by @sapphi-red

### ❤️ New Contributors

* @siaeyy made their first contribution in [#6241](#6241)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

[Perf]: Optimize multi thread IO in macOS

3 participants