Switch RcStr from std::sync::Arc to triomphe::Arc #8715
Merged
Conversation
This file contains 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
|
The latest updates on your projects. Learn more about Vercel for Git ↗︎
8 Skipped Deployments
|
|
This stack of pull requests is managed by Graphite. Learn more about stacking. |
🟢 Turbopack Benchmark CI successful 🟢Thanks |
|
✅ This change can build |
|
This saves a weakref count on each Arc
kdy1
approved these changes
Jul 11, 2024
kdy1
added a commit
to vercel/next.js
that referenced
this pull request
Jul 12, 2024
# Turbopack * vercel/turborepo#8686 <!-- Tobias Koppers - improve performance of the graph aggregation --> * vercel/turborepo#8694 <!-- Benjamin Woodruff - Replace `unreachable` with `bail` in `finalize_css` --> * vercel/turborepo#8661 <!-- Benjamin Woodruff - Create a minimum viable `ResolvedVc` type --> * vercel/turborepo#8662 <!-- Benjamin Woodruff - Add a minimum viable ResolvedValue marker trait --> * vercel/turborepo#8678 <!-- Benjamin Woodruff - Add derive macro for ResolvedValue --> * vercel/turborepo#8715 <!-- Benjamin Woodruff - Switch RcStr from std::sync::Arc to triomphe::Arc --> * vercel/turborepo#8699 <!-- Donny/강동윤 - perf: Merge multiple `EsmBinding` --> * vercel/turborepo#8720 <!-- Benjamin Woodruff - Add support for `#[turbo_tasks::value(resolved)]` and `#[turbo_tasks::value_trait(resolved)]` --> * vercel/turborepo#8706 <!-- Donny/강동윤 - build: Update `swc_core` to `v0.96.9` --> ### What? Update SWC crates to swc-project/swc@226617e and swc-project/plugins@b7658c3 ### Why? To keep in sync ### How? - Closes #64890 - Closes #63104
ForsakenHarmony
pushed a commit
to vercel/next.js
that referenced
this pull request
Jul 25, 2024
) # Description This removes the weakref from each `Arc`, which we weren't using, saving 64 bits per unique `RcStr`. https://docs.rs/triomphe/latest/triomphe/struct.Arc.html The difference in memory is too small to easily measure with heaptrack, but it should be a small win regardless, and it's a drop-in replacement that doesn't make the code any more complicated. # Testing Instructions Built using ``` cargo build --release -p next-build-test ``` Ran heaptrack on a single page (`/sink`) in `shadcn/ui` with: ``` cd ~/ui/apps/www RUST_LOG=next_build_test=info heaptrack ~/nextpack/target/release/next-build-test run sequential 1 1 '/sink' ``` And analyzed the results with ``` heaptrack --analyze /home/bgw.linux/ui/apps/www/heaptrack.next-build-test.3066837.zst | tail -100 ``` ## Heaptrack Before PR ``` total runtime: 135.14s. calls to allocation functions: 49700807 (367770/s) temporary memory allocations: 4126130 (30532/s) peak heap memory consumption: 1.18G peak RSS (including heaptrack overhead): 2.76G total memory leaked: 11.93M suppressed leaks: 7.24K ``` ``` total runtime: 138.50s. calls to allocation functions: 49771465 (359355/s) temporary memory allocations: 4410274 (31842/s) peak heap memory consumption: 1.18G peak RSS (including heaptrack overhead): 2.75G total memory leaked: 12.09M suppressed leaks: 7.24K ``` ## Heaptrack After PR ``` total runtime: 145.38s. calls to allocation functions: 49714711 (341966/s) temporary memory allocations: 4164393 (28645/s) peak heap memory consumption: 1.18G peak RSS (including heaptrack overhead): 2.75G total memory leaked: 11.67M suppressed leaks: 7.24K ``` ``` total runtime: 128.03s. calls to allocation functions: 49739679 (388515/s) temporary memory allocations: 4111893 (32117/s) peak heap memory consumption: 1.18G peak RSS (including heaptrack overhead): 2.76G total memory leaked: 11.88M suppressed leaks: 7.24K ```
ForsakenHarmony
pushed a commit
to vercel/next.js
that referenced
this pull request
Jul 29, 2024
) # Description This removes the weakref from each `Arc`, which we weren't using, saving 64 bits per unique `RcStr`. https://docs.rs/triomphe/latest/triomphe/struct.Arc.html The difference in memory is too small to easily measure with heaptrack, but it should be a small win regardless, and it's a drop-in replacement that doesn't make the code any more complicated. # Testing Instructions Built using ``` cargo build --release -p next-build-test ``` Ran heaptrack on a single page (`/sink`) in `shadcn/ui` with: ``` cd ~/ui/apps/www RUST_LOG=next_build_test=info heaptrack ~/nextpack/target/release/next-build-test run sequential 1 1 '/sink' ``` And analyzed the results with ``` heaptrack --analyze /home/bgw.linux/ui/apps/www/heaptrack.next-build-test.3066837.zst | tail -100 ``` ## Heaptrack Before PR ``` total runtime: 135.14s. calls to allocation functions: 49700807 (367770/s) temporary memory allocations: 4126130 (30532/s) peak heap memory consumption: 1.18G peak RSS (including heaptrack overhead): 2.76G total memory leaked: 11.93M suppressed leaks: 7.24K ``` ``` total runtime: 138.50s. calls to allocation functions: 49771465 (359355/s) temporary memory allocations: 4410274 (31842/s) peak heap memory consumption: 1.18G peak RSS (including heaptrack overhead): 2.75G total memory leaked: 12.09M suppressed leaks: 7.24K ``` ## Heaptrack After PR ``` total runtime: 145.38s. calls to allocation functions: 49714711 (341966/s) temporary memory allocations: 4164393 (28645/s) peak heap memory consumption: 1.18G peak RSS (including heaptrack overhead): 2.75G total memory leaked: 11.67M suppressed leaks: 7.24K ``` ``` total runtime: 128.03s. calls to allocation functions: 49739679 (388515/s) temporary memory allocations: 4111893 (32117/s) peak heap memory consumption: 1.18G peak RSS (including heaptrack overhead): 2.76G total memory leaked: 11.88M suppressed leaks: 7.24K ```
ForsakenHarmony
pushed a commit
to vercel/next.js
that referenced
this pull request
Jul 29, 2024
) # Description This removes the weakref from each `Arc`, which we weren't using, saving 64 bits per unique `RcStr`. https://docs.rs/triomphe/latest/triomphe/struct.Arc.html The difference in memory is too small to easily measure with heaptrack, but it should be a small win regardless, and it's a drop-in replacement that doesn't make the code any more complicated. # Testing Instructions Built using ``` cargo build --release -p next-build-test ``` Ran heaptrack on a single page (`/sink`) in `shadcn/ui` with: ``` cd ~/ui/apps/www RUST_LOG=next_build_test=info heaptrack ~/nextpack/target/release/next-build-test run sequential 1 1 '/sink' ``` And analyzed the results with ``` heaptrack --analyze /home/bgw.linux/ui/apps/www/heaptrack.next-build-test.3066837.zst | tail -100 ``` ## Heaptrack Before PR ``` total runtime: 135.14s. calls to allocation functions: 49700807 (367770/s) temporary memory allocations: 4126130 (30532/s) peak heap memory consumption: 1.18G peak RSS (including heaptrack overhead): 2.76G total memory leaked: 11.93M suppressed leaks: 7.24K ``` ``` total runtime: 138.50s. calls to allocation functions: 49771465 (359355/s) temporary memory allocations: 4410274 (31842/s) peak heap memory consumption: 1.18G peak RSS (including heaptrack overhead): 2.75G total memory leaked: 12.09M suppressed leaks: 7.24K ``` ## Heaptrack After PR ``` total runtime: 145.38s. calls to allocation functions: 49714711 (341966/s) temporary memory allocations: 4164393 (28645/s) peak heap memory consumption: 1.18G peak RSS (including heaptrack overhead): 2.75G total memory leaked: 11.67M suppressed leaks: 7.24K ``` ``` total runtime: 128.03s. calls to allocation functions: 49739679 (388515/s) temporary memory allocations: 4111893 (32117/s) peak heap memory consumption: 1.18G peak RSS (including heaptrack overhead): 2.76G total memory leaked: 11.88M suppressed leaks: 7.24K ```
ForsakenHarmony
pushed a commit
to vercel/next.js
that referenced
this pull request
Aug 1, 2024
) # Description This removes the weakref from each `Arc`, which we weren't using, saving 64 bits per unique `RcStr`. https://docs.rs/triomphe/latest/triomphe/struct.Arc.html The difference in memory is too small to easily measure with heaptrack, but it should be a small win regardless, and it's a drop-in replacement that doesn't make the code any more complicated. # Testing Instructions Built using ``` cargo build --release -p next-build-test ``` Ran heaptrack on a single page (`/sink`) in `shadcn/ui` with: ``` cd ~/ui/apps/www RUST_LOG=next_build_test=info heaptrack ~/nextpack/target/release/next-build-test run sequential 1 1 '/sink' ``` And analyzed the results with ``` heaptrack --analyze /home/bgw.linux/ui/apps/www/heaptrack.next-build-test.3066837.zst | tail -100 ``` ## Heaptrack Before PR ``` total runtime: 135.14s. calls to allocation functions: 49700807 (367770/s) temporary memory allocations: 4126130 (30532/s) peak heap memory consumption: 1.18G peak RSS (including heaptrack overhead): 2.76G total memory leaked: 11.93M suppressed leaks: 7.24K ``` ``` total runtime: 138.50s. calls to allocation functions: 49771465 (359355/s) temporary memory allocations: 4410274 (31842/s) peak heap memory consumption: 1.18G peak RSS (including heaptrack overhead): 2.75G total memory leaked: 12.09M suppressed leaks: 7.24K ``` ## Heaptrack After PR ``` total runtime: 145.38s. calls to allocation functions: 49714711 (341966/s) temporary memory allocations: 4164393 (28645/s) peak heap memory consumption: 1.18G peak RSS (including heaptrack overhead): 2.75G total memory leaked: 11.67M suppressed leaks: 7.24K ``` ``` total runtime: 128.03s. calls to allocation functions: 49739679 (388515/s) temporary memory allocations: 4111893 (32117/s) peak heap memory consumption: 1.18G peak RSS (including heaptrack overhead): 2.76G total memory leaked: 11.88M suppressed leaks: 7.24K ```
ForsakenHarmony
pushed a commit
to vercel/next.js
that referenced
this pull request
Aug 14, 2024
* vercel/turborepo#8686 <!-- Tobias Koppers - improve performance of the graph aggregation --> * vercel/turborepo#8694 <!-- Benjamin Woodruff - Replace `unreachable` with `bail` in `finalize_css` --> * vercel/turborepo#8661 <!-- Benjamin Woodruff - Create a minimum viable `ResolvedVc` type --> * vercel/turborepo#8662 <!-- Benjamin Woodruff - Add a minimum viable ResolvedValue marker trait --> * vercel/turborepo#8678 <!-- Benjamin Woodruff - Add derive macro for ResolvedValue --> * vercel/turborepo#8715 <!-- Benjamin Woodruff - Switch RcStr from std::sync::Arc to triomphe::Arc --> * vercel/turborepo#8699 <!-- Donny/강동윤 - perf: Merge multiple `EsmBinding` --> * vercel/turborepo#8720 <!-- Benjamin Woodruff - Add support for `#[turbo_tasks::value(resolved)]` and `#[turbo_tasks::value_trait(resolved)]` --> * vercel/turborepo#8706 <!-- Donny/강동윤 - build: Update `swc_core` to `v0.96.9` --> Update SWC crates to swc-project/swc@226617e and swc-project/plugins@b7658c3 To keep in sync - Closes #64890 - Closes #63104
ForsakenHarmony
pushed a commit
to vercel/next.js
that referenced
this pull request
Aug 15, 2024
* vercel/turborepo#8686 <!-- Tobias Koppers - improve performance of the graph aggregation --> * vercel/turborepo#8694 <!-- Benjamin Woodruff - Replace `unreachable` with `bail` in `finalize_css` --> * vercel/turborepo#8661 <!-- Benjamin Woodruff - Create a minimum viable `ResolvedVc` type --> * vercel/turborepo#8662 <!-- Benjamin Woodruff - Add a minimum viable ResolvedValue marker trait --> * vercel/turborepo#8678 <!-- Benjamin Woodruff - Add derive macro for ResolvedValue --> * vercel/turborepo#8715 <!-- Benjamin Woodruff - Switch RcStr from std::sync::Arc to triomphe::Arc --> * vercel/turborepo#8699 <!-- Donny/강동윤 - perf: Merge multiple `EsmBinding` --> * vercel/turborepo#8720 <!-- Benjamin Woodruff - Add support for `#[turbo_tasks::value(resolved)]` and `#[turbo_tasks::value_trait(resolved)]` --> * vercel/turborepo#8706 <!-- Donny/강동윤 - build: Update `swc_core` to `v0.96.9` --> Update SWC crates to swc-project/swc@226617e and swc-project/plugins@b7658c3 To keep in sync - Closes #64890 - Closes #63104
ForsakenHarmony
pushed a commit
to vercel/next.js
that referenced
this pull request
Aug 16, 2024
* vercel/turborepo#8686 <!-- Tobias Koppers - improve performance of the graph aggregation --> * vercel/turborepo#8694 <!-- Benjamin Woodruff - Replace `unreachable` with `bail` in `finalize_css` --> * vercel/turborepo#8661 <!-- Benjamin Woodruff - Create a minimum viable `ResolvedVc` type --> * vercel/turborepo#8662 <!-- Benjamin Woodruff - Add a minimum viable ResolvedValue marker trait --> * vercel/turborepo#8678 <!-- Benjamin Woodruff - Add derive macro for ResolvedValue --> * vercel/turborepo#8715 <!-- Benjamin Woodruff - Switch RcStr from std::sync::Arc to triomphe::Arc --> * vercel/turborepo#8699 <!-- Donny/강동윤 - perf: Merge multiple `EsmBinding` --> * vercel/turborepo#8720 <!-- Benjamin Woodruff - Add support for `#[turbo_tasks::value(resolved)]` and `#[turbo_tasks::value_trait(resolved)]` --> * vercel/turborepo#8706 <!-- Donny/강동윤 - build: Update `swc_core` to `v0.96.9` --> Update SWC crates to swc-project/swc@226617e and swc-project/plugins@b7658c3 To keep in sync - Closes #64890 - Closes #63104
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.
Description
This removes the weakref from each
Arc, which we weren't using, saving 64 bits per uniqueRcStr.https://docs.rs/triomphe/latest/triomphe/struct.Arc.html
The difference in memory is too small to easily measure with heaptrack, but it should be a small win regardless, and it's a drop-in replacement that doesn't make the code any more complicated.
Testing Instructions
Built using
Ran heaptrack on a single page (
/sink) inshadcn/uiwith:And analyzed the results with
Heaptrack Before PR
Heaptrack After PR