Skip to content

Conversation

@lukesandberg
Copy link
Contributor

@lukesandberg lukesandberg commented Feb 2, 2026

What?

Optimizes the change tracking mechanism used for persistent cache snapshots in turbo-tasks-backend.

Why?

The previous implementation used a FxDashMap<TaskId, ModifiedState> to track modified tasks, where ModifiedState was an enum with Modified and Snapshot(Option<Box<TaskStorage>>) variants. This had several inefficiencies:

  1. Memory overhead: The ModifiedState enum was 16 bytes to account for the rare 'snapshot race' case
  2. Datastructure overhead: the DashMap for the modified map was highly sharded leading to large amounts of memory overhead. Most acquisitions were extremely short lived and so this isn't necessary.
  3. Separate task cache log: New task types were tracked in a separate TaskCacheLog structure (Sharded<ChunkedVec<...>>), requiring additional synchronization and a separate persistence path.

How?

Restructured change tracking into two separate data structures:

  1. modified: Sharded<Vec<TaskId>> - A sharded append-only list for tracking modified task IDs. Since modifications are guarded by a transition in the any_modified flag on TaskStorage, each task is only added once, avoiding duplicates without additional synchronization.

  2. snapshots: FxDashMap<TaskId, Option<Box<TaskStorage>>> - A small, rarely-used map for the uncommon case where a task is modified during an active snapshot operation. Uses only 16 shards since this is rare (only during dev mode idle-callback persistence races).

Unified task cache persistence:

  • Removed the separate TaskCacheLog and ChunkedVec infrastructure
  • New tasks now include their task_type directly in the SnapshotItem struct
  • Task cache entries are written inline with task data during persistence, simplifying the write path

Other optimizations:

  • SmallVec::into_boxed_slice() now called directly instead of into_vec().into_boxed_slice() (avoids intermediate allocation)
  • Removed swap_retain utility (no longer needed)
  • Simplified snapshot iteration using scope_and_block with chunked shards, this reduces the number of 'scratch buffers' allocated to 'one per chunk' instead of 'one per original shard'
  • Fixed a very subtle race in end_snapshot where multiple racing save_snapshot calls could corrupt state.

Results on vercel-site

I ran a sequence of builds on vercel-site comparing canary and this branch

Cold Cache Builds

Metric Optimization (87071fe) Canary (479877d) Difference
Duration (avg) 82.98s 82.56s +0.51%
Duration (min) 82.44s 80.39s
Duration (max) 83.20s 87.24s
Duration (std dev) 0.27s 2.22s
Peak Memory (avg) 21.92 GB 23.07 GB -4.98%
Peak Memory (min) 21.83 GB 22.64 GB
Peak Memory (max) 22.06 GB 23.78 GB
Sample Size n=6 n=6

Warm Cache Builds

Metric Optimization (87071fe) Canary (479877d) Difference
Duration (avg) 13.71s 14.75s -7.02%
Duration (min) 10.65s 11.41s
Duration (max) 17.77s 19.84s
Peak Memory (avg) 11.10 GB 11.39 GB -2.49%
Peak Memory (min) 10.76 GB 10.68 GB
Peak Memory (max) 11.48 GB 12.06 GB
Sample Size n=12 n=12

The latency deltas are in the noise but there are substantial memory wins from this change

@nextjs-bot nextjs-bot added created-by: Turbopack team PRs by the Turbopack team. Turbopack Related to Turbopack with Next.js. labels Feb 2, 2026
Copy link
Contributor Author

lukesandberg commented Feb 2, 2026

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

@codspeed-hq
Copy link

codspeed-hq bot commented Feb 2, 2026

CodSpeed Performance Report

Merging this PR will improve performance by 3.76%

Comparing optimized_change_tracking (4dada51) with canary (7437fed)

Summary

⚡ 1 improved benchmark
✅ 16 untouched benchmarks
⏩ 3 skipped benchmarks1

Performance Changes

Mode Benchmark BASE HEAD Efficiency
Simulation jsonwebtoken.js[tracing] 280.6 ms 270.4 ms +3.76%

Footnotes

  1. 3 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports.

@nextjs-bot
Copy link
Collaborator

nextjs-bot commented Feb 2, 2026

Stats from current PR

🟢 1 improvement

Metric Canary PR Change Trend
Webpack Build Time 16.159s 14.536s 🟢 1.623s (-10%) ▁▁▃▁▅
📊 All Metrics
📖 Metrics Glossary

Dev Server Metrics:

  • Listen = TCP port starts accepting connections
  • First Request = HTTP server returns successful response
  • Cold = Fresh build (no cache)
  • Warm = With cached build artifacts

Build Metrics:

  • Fresh = Clean build (no .next directory)
  • Cached = With existing .next directory

Change Thresholds:

  • Time: Changes < 50ms AND < 10%, OR < 2% are insignificant
  • Size: Changes < 1KB AND < 1% are insignificant
  • All other changes are flagged to catch regressions

⚡ Dev Server

Metric Canary PR Change Trend
Cold (Listen) 506ms 505ms ▁▁▁▁▁
Cold (Ready in log) 461ms 460ms ▂▂▂▂▁
Cold (First Request) 829ms 834ms ▅▅▅▄▁
Warm (Listen) 458ms 456ms ▁▁▁▁▁
Warm (Ready in log) 459ms 456ms ▁▁▁▁▁
Warm (First Request) 363ms 364ms ▁▁▂▂▁
📦 Dev Server (Webpack) (Legacy)

📦 Dev Server (Webpack)

Metric Canary PR Change Trend
Cold (Listen) 507ms 506ms ▁█▅▅█
Cold (Ready in log) 475ms 467ms ▇▆█▇█
Cold (First Request) 1.963s 1.930s ▄▃▅▄▆
Warm (Listen) 458ms 458ms ▅▅▅▅█
Warm (Ready in log) 451ms 454ms ▇▆▆▆█
Warm (First Request) 1.907s 1.917s ▅▄▆▄▇

⚡ Production Builds

Metric Canary PR Change Trend
Fresh Build 4.194s 4.144s ▁▁▁▁▃
Cached Build 4.241s 4.201s ▁▁▁▁▃
📦 Production Builds (Webpack) (Legacy)

📦 Production Builds (Webpack)

Metric Canary PR Change Trend
Fresh Build 16.159s 14.536s 🟢 1.623s (-10%) ▁▁▃▁▅
Cached Build 14.925s 14.785s ▁▁▃▁▅
node_modules Size 464 MB 464 MB ▁▁▁▁▁
📦 Bundle Sizes

Bundle Sizes

⚡ Turbopack

Client

Main Bundles: **434 kB** → **435 kB** ⚠️ +131 B

81 files with content-based hashes (individual files not comparable between builds)

Server

Middleware
Canary PR Change
middleware-b..fest.js gzip 763 B 764 B
Total 763 B 764 B ⚠️ +1 B
Build Details
Build Manifests
Canary PR Change
_buildManifest.js gzip 449 B 451 B
Total 449 B 451 B ⚠️ +2 B

📦 Webpack

Client

Main Bundles
Canary PR Change
5528-HASH.js gzip 5.47 kB N/A -
6280-HASH.js gzip 54.5 kB N/A -
6335.HASH.js gzip 169 B N/A -
912-HASH.js gzip 4.53 kB N/A -
e8aec2e4-HASH.js gzip 62.5 kB N/A -
framework-HASH.js gzip 59.7 kB 59.7 kB
main-app-HASH.js gzip 255 B 254 B
main-HASH.js gzip 39 kB 39.1 kB
webpack-HASH.js gzip 1.68 kB 1.68 kB
262-HASH.js gzip N/A 4.52 kB -
2889.HASH.js gzip N/A 169 B -
5602-HASH.js gzip N/A 5.48 kB -
6948ada0-HASH.js gzip N/A 62.5 kB -
9544-HASH.js gzip N/A 55.2 kB -
Total 228 kB 229 kB ⚠️ +796 B
Polyfills
Canary PR Change
polyfills-HASH.js gzip 39.4 kB 39.4 kB
Total 39.4 kB 39.4 kB
Pages
Canary PR Change
_app-HASH.js gzip 194 B 194 B
_error-HASH.js gzip 183 B 180 B 🟢 3 B (-2%)
css-HASH.js gzip 331 B 330 B
dynamic-HASH.js gzip 1.81 kB 1.81 kB
edge-ssr-HASH.js gzip 256 B 256 B
head-HASH.js gzip 351 B 352 B
hooks-HASH.js gzip 384 B 383 B
image-HASH.js gzip 580 B 581 B
index-HASH.js gzip 260 B 260 B
link-HASH.js gzip 2.49 kB 2.49 kB
routerDirect..HASH.js gzip 320 B 319 B
script-HASH.js gzip 386 B 386 B
withRouter-HASH.js gzip 315 B 315 B
1afbb74e6ecf..834.css gzip 106 B 106 B
Total 7.97 kB 7.97 kB ✅ -1 B

Server

Edge SSR
Canary PR Change
edge-ssr.js gzip 126 kB 126 kB
page.js gzip 249 kB 249 kB
Total 375 kB 375 kB ⚠️ +403 B
Middleware
Canary PR Change
middleware-b..fest.js gzip 616 B 614 B
middleware-r..fest.js gzip 156 B 155 B
middleware.js gzip 33.1 kB 33.1 kB
edge-runtime..pack.js gzip 842 B 842 B
Total 34.7 kB 34.8 kB ⚠️ +56 B
Build Details
Build Manifests
Canary PR Change
_buildManifest.js gzip 732 B 736 B
Total 732 B 736 B ⚠️ +4 B
Build Cache
Canary PR Change
0.pack gzip 3.79 MB 3.81 MB 🔴 +15.3 kB (+0%)
index.pack gzip 104 kB 102 kB 🟢 2.03 kB (-2%)
index.pack.old gzip 103 kB 102 kB
Total 4 MB 4.01 MB ⚠️ +12.7 kB

🔄 Shared (bundler-independent)

Runtimes
Canary PR Change
app-page-exp...dev.js gzip 311 kB 311 kB
app-page-exp..prod.js gzip 166 kB 166 kB
app-page-tur...dev.js gzip 311 kB 311 kB
app-page-tur..prod.js gzip 166 kB 166 kB
app-page-tur...dev.js gzip 308 kB 308 kB
app-page-tur..prod.js gzip 164 kB 164 kB
app-page.run...dev.js gzip 308 kB 308 kB
app-page.run..prod.js gzip 164 kB 164 kB
app-route-ex...dev.js gzip 70.4 kB 70.5 kB
app-route-ex..prod.js gzip 48.9 kB 49 kB
app-route-tu...dev.js gzip 70.4 kB 70.5 kB
app-route-tu..prod.js gzip 49 kB 49 kB
app-route-tu...dev.js gzip 70 kB 70.1 kB
app-route-tu..prod.js gzip 48.7 kB 48.8 kB
app-route.ru...dev.js gzip 70 kB 70.1 kB
app-route.ru..prod.js gzip 48.7 kB 48.7 kB
dist_client_...dev.js gzip 324 B 324 B
dist_client_...dev.js gzip 326 B 326 B
dist_client_...dev.js gzip 318 B 318 B
dist_client_...dev.js gzip 317 B 317 B
pages-api-tu...dev.js gzip 43.1 kB 43.2 kB
pages-api-tu..prod.js gzip 32.9 kB 32.9 kB
pages-api.ru...dev.js gzip 43.1 kB 43.2 kB
pages-api.ru..prod.js gzip 32.8 kB 32.9 kB
pages-turbo....dev.js gzip 52.4 kB 52.5 kB
pages-turbo...prod.js gzip 39.4 kB 39.4 kB
pages.runtim...dev.js gzip 52.4 kB 52.5 kB
pages.runtim..prod.js gzip 39.3 kB 39.4 kB
server.runti..prod.js gzip 62.6 kB 62.6 kB
Total 2.77 MB 2.78 MB ⚠️ +1.01 kB
📝 Changed Files (25 files)

Files with changes:

  • app-page-exp..ntime.dev.js
  • app-page-exp..time.prod.js
  • app-page-tur..ntime.dev.js
  • app-page-tur..time.prod.js
  • app-page-tur..ntime.dev.js
  • app-page-tur..time.prod.js
  • app-page.runtime.dev.js
  • app-page.runtime.prod.js
  • app-route-ex..ntime.dev.js
  • app-route-ex..time.prod.js
  • app-route-tu..ntime.dev.js
  • app-route-tu..time.prod.js
  • app-route-tu..ntime.dev.js
  • app-route-tu..time.prod.js
  • app-route.runtime.dev.js
  • app-route.ru..time.prod.js
  • pages-api-tu..ntime.dev.js
  • pages-api-tu..time.prod.js
  • pages-api.runtime.dev.js
  • pages-api.ru..time.prod.js
  • ... and 5 more
View diffs
app-page-exp..ntime.dev.js
failed to diff
app-page-exp..time.prod.js

Diff too large to display

app-page-tur..ntime.dev.js
failed to diff
app-page-tur..time.prod.js

Diff too large to display

app-page-tur..ntime.dev.js
failed to diff
app-page-tur..time.prod.js

Diff too large to display

app-page.runtime.dev.js
failed to diff
app-page.runtime.prod.js

Diff too large to display

app-route-ex..ntime.dev.js

Diff too large to display

app-route-ex..time.prod.js

Diff too large to display

app-route-tu..ntime.dev.js

Diff too large to display

app-route-tu..time.prod.js

Diff too large to display

app-route-tu..ntime.dev.js

Diff too large to display

app-route-tu..time.prod.js

Diff too large to display

app-route.runtime.dev.js

Diff too large to display

app-route.ru..time.prod.js

Diff too large to display

pages-api-tu..ntime.dev.js

Diff too large to display

pages-api-tu..time.prod.js

Diff too large to display

pages-api.runtime.dev.js

Diff too large to display

pages-api.ru..time.prod.js

Diff too large to display

pages-turbo...ntime.dev.js

Diff too large to display

pages-turbo...time.prod.js

Diff too large to display

pages.runtime.dev.js

Diff too large to display

pages.runtime.prod.js

Diff too large to display

server.runtime.prod.js

Diff too large to display

@lukesandberg lukesandberg marked this pull request as ready for review February 3, 2026 05:23
@lukesandberg lukesandberg requested review from bgw and sokra February 3, 2026 05:23
@nextjs-bot
Copy link
Collaborator

Failing test suites

Commit: 4dada51 | About building and testing Next.js

pnpm test-dev test/e2e/app-dir-export/test/dynamicapiroute.test.ts (job)

  • app dir - with output export - dynamic api route > should work in prod with dynamicApiRoute 'force-static' > should work (DD)
Expand output

● app dir - with output export - dynamic api route › should work in prod with dynamicApiRoute 'force-static' › should work

expect(received).toContain(expected) // indexOf

Expected substring: "Another"
Received string:    "Home"

  306 |
  307 |       await retry(async () =>
> 308 |         expect(await browser.elementByCss('h1').text()).toContain('Another')
      |                                                         ^
  309 |       )
  310 |       expect(await browser.elementByCss(a(1)).text()).toBe(
  311 |         'Visit the home page'

  at toContain (e2e/app-dir-export/test/utils.ts:308:57)
  at retry (lib/next-test-utils.ts:831:14)
  at Object.<anonymous> (e2e/app-dir-export/test/utils.ts:307:7)

Comment on lines 1106 to 1112
let process = |task_id: TaskId,
(meta, data): (Option<TaskStorage>, Option<TaskStorage>),
(meta, data, task_type): (
Option<TaskStorage>,
Option<TaskStorage>,
Option<Arc<CachedTaskType>>,
),
buffer: &mut TurboBincodeBuffer| {
Copy link
Member

Choose a reason for hiding this comment

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

Funny formatting..

Maybe move it into a type X = to avoid that

Comment on lines -1141 to -1180
let meta = match meta {
Some(Ok(meta)) => {
#[cfg(feature = "print_cache_item_size")]
task_cache_stats
.lock()
.entry(self.debug_get_task_name(task_id))
.or_default()
.add_meta(&meta);
Some(meta)
}
None => None,
Some(Err(err)) => {
println!(
"Serializing task {} failed (meta): {:?}",
self.debug_get_task_description(task_id),
err
);
None
}
};
let data = match data {
Some(Ok(data)) => {
#[cfg(feature = "print_cache_item_size")]
task_cache_stats
.lock()
.entry(self.debug_get_task_name(task_id))
.or_default()
.add_data(&data);
Some(data)
}
None => None,
Some(Err(err)) => {
println!(
"Serializing task {} failed (data): {:?}",
self.debug_get_task_description(task_id),
err
);
None
}
};
Copy link
Member

Choose a reason for hiding this comment

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

Could you keep the print_cache_item_size logic?

&& !task.has_persistent_task_type()
{
let _ = task.set_persistent_task_type(task_type);
task.set_new_persistent_task(true);
Copy link
Member

Choose a reason for hiding this comment

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

I was a bit worried that this will add a race condition, because the task can be persisted before we set the new_persistent_task flag.
I think it's fine, as the task id can't be referenced by other tasks before that.
So in the very rare case of this happening we only end up with a dangling unused task.

process_snapshot,
scratch_buffer: TurboBincodeBuffer::with_capacity(SCRATCH_BUFFER_SIZE),
// Ideally these shards would be perfectly aligned with the dashmap so we could
// monolithically lock shards instead of axquiring a lock for each item. But doing this
Copy link
Member

Choose a reason for hiding this comment

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

axquiring?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

created-by: Turbopack team PRs by the Turbopack team. Turbopack Related to Turbopack with Next.js.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants