perf(pm): minimum-viable preload bundle (worker-pool + OnceMap + aws-lc-rs) by elrrrrrrr · Pull Request #2838 · utooland/utoo

elrrrrrrr · 2026-04-27T13:31:29Z

Summary

The minimum subset of #2818's 101 commits that, individually validated, capture the perf-driver synergy:

`feat(ruborist): add OnceMap util` — copy oncemap.rs from pm/util to ruborist/util so the resolver can use it (no usage yet).
`perf(ruborist): rustls with aws-lc-rs` — drop `ring` for BoringSSL primitives. TLS handshake CPU 162ms → 46ms p50 per measurement in perf(pm): sequential writes within tarball extraction #2818's commit msg.
`perf(ruborist): OnceMap dedup for concurrent manifest fetches` — wire OnceMap into UnifiedRegistry::resolve_full_manifest. When 50 packages all transitively pull `react`, only the first hits network; the rest await the shared Arc.
`perf(pm): preload multi-thread worker pool (tokio::spawn)` — N tokio::spawn workers + SegQueue + MaybeSend/MaybeSync shim for wasm compat.

4 commits, ~7 files. No tarball-extract / cap-raise / DNS-tuning / etc. from #2818.

Driver-hunt history

PR	scope	npmjs p1_resolve	takeaway
#2832	mt-pool only	4.59s ±1.66	mean drop, σ huge
#2835	aws-lc-rs only	6.13s ±1.00	nop
#2834	all 101 commits	2.62s ±0.07	-52%
#2836	#2834 minus wp	4.27s ±0.05	wp ≈ -30% of gain
#2837	#2836 minus OnceMap	5.95s ±1.46	OnceMap ≈ stability
this	minimum-viable bundle	TBD	aiming for #2834-level

The synergy hypothesis: aws-lc-rs unblocks TLS-CPU bottleneck → OnceMap removes duplicate fetches → worker-pool's N independent task slots finally have unique work to do. None alone shows the win.

Local data

ant-design `utoo deps` cold (mac, system proxy on — absolute numbers polluted but ratio is signal):

binary	wall
origin/next baseline	2m19s
this bundle	11s

12.6× ratio matches the order-of-magnitude expected from TLS 3.5× + OnceMap dedup ~2× + parse parallelism. CI bench-phases on clean network is the real signal.

Test plan

cargo fmt + clippy + test pass
cargo build --release pass
ant-design `utoo deps` macOS: exit 0, no silent crash
CI `pm-e2e-*` (4 platforms; linux is the silent-crash smoke from perf(pm): preload worker-pool replaces FuturesUnordered #2827)
CI `utooweb-ci-build-wasm` (validates MaybeSend shim)
CI `pm-bench-phases-*` head-to-head vs `utoo-next`

🤖 Generated with Claude Code

Copy the OnceMap util module from pm/util to ruborist/util so the resolver can use it for per-name dedup of concurrent manifest fetches. Foundation commit — no usage yet, that lands in the worker-pool wiring commit. The pm-side oncemap.rs stays for cloner/downloader users.

CI pcap analysis against npmjs.org revealed TLS CPU as a major preload bottleneck. Per-handshake timing on CI: utoo (ring) bun (BoringSSL) CH → SH (1 RTT) 11 ms 10 ms (network) CCS → first AppData 78 ms p50 12 ms p50 154 ms max 17 ms max TOTAL CH → AppData 162 ms 46 ms The "CCS → AppData" phase is dominated by post-ChangeCipherSpec client work (Finished MAC verify, state machine transition, request dispatch). Observed at CI pcap capture: utoo: Client CCS spread 154ms (first conn done at 0.975s, last at 1.129s), then first AppData *all* fire within 11ms at ~1.13s — classic CPU-saturation pattern where 128 parallel TLS handshakes serialise across 4 blocking threads and HTTP dispatch is starved until crypto drains. bun: CCS spread 51ms, AppData spread 43ms — dispatch flows smoothly as each conn completes. `ring` (rustls' default provider) is pure Rust + hand-tuned assembly; `aws-lc-rs` wraps BoringSSL's primitives which are more aggressively optimised for x86_64 AES-NI + SHA-NI. Reqwest 0.12's `rustls-tls-native-roots` feature pins `__rustls-ring` via Cargo's feature unification — no way to override. Swap to `rustls-tls-native-roots-no-provider`, add a direct `rustls` dep with the `aws-lc-rs` feature, load native root certs via `rustls-native-certs` and pass the `ClientConfig` into reqwest via `use_preconfigured_tls`. Local M2 (ARM, where ring's hand-tuned ARM assembly is already near-optimal) shows neutral perf (~2.9s both providers). Waiting on CI to confirm the x86_64 CCS→AppData gap closes. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Wire the OnceMap from #2832-foundation into UnifiedRegistry's resolve_full_manifest. When N packages all transitively pull on the same dep (e.g. 50 packages → react), only the first hits the network; the others await the shared Arc<result> on Notify and clone it on resolve. DashMap + tokio::sync::Notify keeps the fast path (cache hit) lock-free. cfg-gated on native; wasm path keeps the direct fetch since dashmap + tokio sync are unnecessary there. FullManifestResult derives Clone so OnceMap waiters can each take owned copies.

N long-lived `tokio::spawn` workers pull work from a shared lock-free `SegQueue`. Each worker is independent on tokio's multi-thread runtime, so when one worker is parsing manifest JSON (CPU-bound, simd_json), other workers continue driving their network IO and other parses run on different cores. Replaces `FuturesUnordered` where the main task owned all preload futures and polled them cooperatively — every per- future await continuation (including parse) ran serialised in that single task. The architectural fix alone (proven nop in PR #2830 with spawn_local) needs the network-layer drivers (aws-lc-rs + OnceMap, earlier in this stack) to actually express its benefit. With them: TLS handshake CPU is fast (aws-lc-rs), duplicate fetches are deduped (OnceMap), so worker-pool's N independent task slots are doing actual unique work. ## MaybeSend / MaybeSync shim `tokio::spawn` requires `Future: Send + 'static`, which forces `RegistryClient` trait surface to grow `+ Send` on its `impl Future` returns + `Self: Sync` / `Self::Error: Send` on default impls. Native uses real Send/Sync; wasm32 uses no-op marker traits because `JsFuture` is `!Send` and tasks run via `wasm_bindgen_futures::spawn_local`. `PreloadRegistry` trait alias bundles `RegistryClient + Clone + MaybeSend + MaybeSync + 'static` so the 5 resolver entry points (`build_deps*`, `resolve*`, `run_preload_phase`) read clean. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

gemini-code-assist

Code Review

This pull request refactors the manifest preloading system to use a multi-threaded worker pool with WASM support and introduces a OnceMap utility to deduplicate concurrent network requests. It also migrates the TLS implementation to aws-lc-rs for improved performance. Feedback identifies a race condition in the worker termination logic, the use of unstable let_chains syntax which will break stable Rust compilation, and an inefficient method for calculating processed statistics. Additionally, it is recommended to narrow the scope of the dead_code allowance.

gemini-code-assist · 2026-04-27T13:38:46Z

+                                pending.push(tdep);
+                                count += 1;
+                            }
+                        }
+                        if count > 0 {
+                            dispatched.fetch_add(count, Ordering::Release);


There is a race condition in the worker termination logic. The dispatched counter must be incremented before pushing new work items to the pending queue. If pending.push happens first, another worker could pop the item, finish it, and increment completed before the current worker increments dispatched. This would cause completed == dispatched to be true prematurely, leading to early shutdown of the preload phase.

let new_added = if let Ok(resolved) = &result { let mut to_push = Vec::new(); for tdep in extract_transitive_deps(&resolved.manifest, &config_for_worker) { let key = format!("{}@{}", tdep.0, tdep.1); if processed.insert(key) { to_push.push(tdep); } } let count = to_push.len(); if count > 0 { dispatched.fetch_add(count, Ordering::Release); for tdep in to_push { pending.push(tdep); } // Wake parked workers so they pick up the new // work before checking the termination condition. notify.notify_waiters(); } count } else { 0 };

gemini-code-assist · 2026-04-27T13:38:46Z

+                    if let Some(entry) = self.map.get(key)
+                        && matches!(entry.value(), Value::Done(_))
+                    {
+                        return;
+                    }


The use of let_chains (e.g., if let ... && ...) is an unstable Rust feature. Unless this project explicitly targets nightly and enables the feature flag, this will cause a compilation error on stable Rust. It's safer to use nested if let blocks.

Suggested change

if let Some(entry) = self.map.get(key)

&& matches!(entry.value(), Value::Done(_))

{

return;

}

if let Some(entry) = self.map.get(key) {

if matches!(entry.value(), Value::Done(_)) {

return;

}

}

gemini-code-assist · 2026-04-27T13:38:46Z

+                        if let Some(entry) = self.map.get(&key)
+                            && let Value::Done(result) = entry.value()
+                        {
+                            return Some(Arc::clone(result));
+                        }


The use of let_chains is unstable and will break compilation on stable Rust. Please use nested if let instead.

Suggested change

if let Some(entry) = self.map.get(&key)

&& let Value::Done(result) = entry.value()

{

return Some(Arc::clone(result));

}

if let Some(entry) = self.map.get(&key) {

if let Value::Done(result) = entry.value() {

return Some(Arc::clone(result));

}

}

gemini-code-assist · 2026-04-27T13:38:47Z

+    stats.total_processed = {
+        let mut set: HashSet<String> = HashSet::with_capacity(processed.len());
+        for entry in processed.iter() {
+            set.insert(entry.key().clone());
+        }
+        set.len()
+    };


Calculating total_processed by cloning the entire DashSet into a new HashSet is highly inefficient (O(N) time and space). DashSet::len() provides the count by summing shard lengths without allocating or copying the underlying data.

stats.total_processed = processed.len();

gemini-code-assist · 2026-04-27T13:38:47Z

@@ -0,0 +1,465 @@
+#![allow(dead_code)]


The #![allow(dead_code)] attribute at the crate level is too broad. It's better to allow it only on specific functions that are intended for future use, or remove it entirely if this is meant to be a public API.

aws-lc-sys cc_builder.rs:692 panics on wasm32 target (no support). Move the explicit rustls + aws-lc-rs dependency to native-only; reqwest's rustls-tls-native-roots-no-provider feature on wasm is a no-op since wasm uses the fetch API and the browser's TLS stack, not rustls.

github-actions · 2026-04-27T15:54:48Z

📊 pm-bench-phases · `33388e0` · linux (`ubuntu-latest`)

Workflow run — ant-design

PMs: utoo (this branch) · utoo-npm (latest published) · bun (latest)

npmjs.org

p0_full_cold

PM	wall	±σ	user	sys	RSS	pgMinor
bun	9.40s	0.45s	10.09s	10.15s	648M	315.8K
utoo-npm	10.11s	0.39s	11.60s	13.24s	1.32G	155.8K
utoo	9.53s	0.90s	12.47s	13.04s	1.33G	156.5K

PM	vCtx	iCtx	netRX	netTX	cache	node_mod	lock
bun	14.7K	17.0K	1.16G	6M	1.83G	1.72G	1M
utoo-npm	165.6K	153.0K	1.14G	4M	1.68G	1.68G	2M
utoo	131.9K	95.9K	1.14G	5M	1.68G	1.68G	2M

p1_resolve

PM	wall	±σ	user	sys	RSS	pgMinor
bun	2.17s	0.05s	3.91s	1.07s	502M	171.9K
utoo-npm	5.26s	0.06s	5.92s	1.19s	429M	67.9K
utoo	3.03s	0.11s	6.62s	1.95s	569M	83.1K

PM	vCtx	iCtx	netRX	netTX	cache	node_mod	lock
bun	9.7K	3.9K	200M	3M	104M	-	1M
utoo-npm	65.3K	3.0K	202M	2M	9M	5M	2M
utoo	39.4K	20.7K	198M	2M	9M	5M	2M

p3_cold_install

PM	wall	±σ	user	sys	RSS	pgMinor
bun	7.18s	0.77s	6.09s	9.88s	588M	197.5K
utoo-npm	7.29s	1.26s	5.57s	11.33s	991M	105.7K
utoo	6.40s	0.07s	5.47s	11.27s	913M	125.1K

PM	vCtx	iCtx	netRX	netTX	cache	node_mod	lock
bun	5.1K	6.9K	993M	4M	1.73G	1.73G	1M
utoo-npm	109.7K	72.0K	964M	3M	1.67G	1.67G	2M
utoo	104.0K	79.3K	964M	3M	1.67G	1.67G	2M

p4_warm_link

PM	wall	±σ	user	sys	RSS	pgMinor
bun	3.29s	0.16s	0.22s	2.34s	135M	32.5K
utoo-npm	2.58s	0.19s	0.65s	4.00s	82M	18.7K
utoo	2.37s	0.06s	0.65s	4.01s	82M	18.4K

PM	vCtx	iCtx	netRX	netTX	cache	node_mod	lock
bun	272	23	7M	52K	1.88G	1.72G	1M
utoo-npm	50.0K	21.5K	21K	10K	1.67G	1.67G	2M
utoo	50.7K	22.8K	21K	10K	1.68G	1.67G	2M

npmmirror.com

p0_full_cold

PM	wall	±σ	user	sys	RSS	pgMinor
bun	35.12s	11.91s	9.22s	9.56s	532M	381.8K
utoo-npm	26.95s	5.05s	8.66s	14.38s	776M	98.1K
utoo	51.32s	32.76s	8.58s	14.28s	782M	105.1K

PM	vCtx	iCtx	netRX	netTX	cache	node_mod	lock
bun	53.8K	4.6K	1.12G	9M	1.84G	1.72G	2M
utoo-npm	249.2K	107.1K	978M	8M	1.67G	1.68G	2M
utoo	236.8K	86.5K	981M	8M	1.67G	1.68G	2M

p1_resolve

PM	wall	±σ	user	sys	RSS	pgMinor
bun	1.48s	0.05s	3.97s	1.09s	607M	186.5K
utoo-npm	5.23s	0.11s	2.10s	0.59s	75M	16.1K
utoo	3.22s	1.83s	1.50s	0.54s	82M	17.3K

PM	vCtx	iCtx	netRX	netTX	cache	node_mod	lock
bun	5.0K	6.1K	151M	2M	106M	-	2M
utoo-npm	45.9K	751	12M	2M	-	4M	2M
utoo	26.6K	1.8K	16M	2M	-	4M	2M

p3_cold_install

PM	wall	±σ	user	sys	RSS	pgMinor
bun	23.97s	1.28s	5.43s	8.24s	243M	113.2K
utoo-npm	46.43s	30.54s	6.29s	12.68s	585M	89.0K
utoo	51.63s	30.07s	6.05s	12.89s	582M	92.6K

PM	vCtx	iCtx	netRX	netTX	cache	node_mod	lock
bun	32.7K	2.3K	984M	7M	1.70G	1.70G	2M
utoo-npm	190.7K	93.8K	965M	6M	1.67G	1.67G	2M
utoo	194.9K	94.6K	975M	6M	1.67G	1.67G	2M

p4_warm_link

PM	wall	±σ	user	sys	RSS	pgMinor
bun	15.01s	20.24s	0.18s	2.25s	134M	30.8K
utoo-npm	2.47s	0.26s	0.62s	3.86s	82M	18.9K
utoo	2.27s	0.09s	0.63s	3.90s	83M	19.4K

PM	vCtx	iCtx	netRX	netTX	cache	node_mod	lock
bun	633	22	3M	118K	1.82G	1.72G	2M
utoo-npm	48.1K	21.9K	86K	15K	1.67G	1.67G	2M
utoo	49.6K	22.0K	85K	13K	1.67G	1.67G	2M

github-actions · 2026-04-27T16:18:18Z

📊 pm-bench-phases · `33388e0` · mac (`macos-latest`)

Workflow run — ant-design

PMs: utoo (this branch) · utoo-npm (latest published) · bun (latest)

npmjs.org

p0_full_cold

PM	wall	±σ	user	sys	RSS	pgMinor
bun	17.54s	1.87s	6.20s	17.47s	774M	50.0K
utoo-npm	22.63s	6.56s	10.45s	23.24s	933M	96.9K
utoo	17.70s	3.92s	9.43s	20.66s	1.10G	109.9K

PM	vCtx	iCtx	netRX	netTX	cache	node_mod	lock
bun	16.1K	143.9K	-	-	1.79G	1.90G	1M
utoo-npm	13.9K	360.2K	-	-	1.63G	1.88G	2M
utoo	7.2K	352.2K	-	-	1.63G	1.87G	2M

p1_resolve

PM	wall	±σ	user	sys	RSS	pgMinor
bun	2.97s	0.12s	3.13s	1.66s	475M	30.8K
utoo-npm	5.71s	0.29s	4.50s	2.70s	549M	37.2K
utoo	5.43s	1.91s	5.26s	3.03s	609M	40.5K

PM	vCtx	iCtx	netRX	netTX	cache	node_mod	lock
bun	34	21.6K	-	-	110M	-	1M
utoo-npm	14	72.6K	-	-	28M	5M	2M
utoo	12	81.1K	-	-	28M	5M	2M

p3_cold_install

PM	wall	±σ	user	sys	RSS	pgMinor
bun	15.83s	5.17s	3.28s	15.01s	517M	33.7K
utoo-npm	14.65s	4.65s	3.78s	15.70s	791M	83.1K
utoo	12.77s	2.78s	3.84s	16.47s	747M	80.7K

PM	vCtx	iCtx	netRX	netTX	cache	node_mod	lock
bun	5.6K	132.3K	-	-	1.70G	1.94G	1M
utoo-npm	1.6K	238.7K	-	-	1.60G	1.83G	2M
utoo	1.4K	220.3K	-	-	1.60G	1.83G	2M

p4_warm_link

PM	wall	±σ	user	sys	RSS	pgMinor
bun	6.53s	1.56s	0.12s	2.53s	52M	3.9K
utoo-npm	5.53s	0.84s	0.69s	3.77s	97M	7.2K
utoo	3.78s	0.54s	0.51s	2.88s	92M	7.0K

PM	vCtx	iCtx	netRX	netTX	cache	node_mod	lock
bun	17.0K	1.0K	-	-	1.87G	1.91G	1M
utoo-npm	13.0K	76.5K	-	-	1.60G	1.83G	2M
utoo	13.5K	74.2K	-	-	1.63G	1.83G	2M

npmmirror.com

p0_full_cold

PM	wall	±σ	user	sys	RSS	pgMinor
bun	50.39s	36.38s	7.77s	24.95s	598M	38.6K
utoo-npm	51.90s	22.80s	8.93s	26.08s	778M	77.9K
utoo	27.55s	1.13s	7.30s	20.82s	820M	86.0K

PM	vCtx	iCtx	netRX	netTX	cache	node_mod	lock
bun	15.4K	151.5K	-	-	1.78G	1.90G	2M
utoo-npm	4.3K	429.4K	-	-	1.61G	1.87G	2M
utoo	898	413.5K	-	-	1.61G	1.86G	2M

p1_resolve

PM	wall	±σ	user	sys	RSS	pgMinor
bun	16.45s	22.35s	3.13s	1.88s	558M	36.3K
utoo-npm	4.61s	0.05s	1.79s	0.98s	79M	5.8K
utoo	4.20s	3.83s	1.45s	0.66s	83M	6.0K

PM	vCtx	iCtx	netRX	netTX	cache	node_mod	lock
bun	36	26.0K	-	-	111M	-	2M
utoo-npm	5	33.6K	-	-	-	4M	2M
utoo	28	17.9K	-	-	-	4M	2M

p3_cold_install

PM	wall	±σ	user	sys	RSS	pgMinor
bun	34.16s	16.69s	3.61s	15.05s	250M	16.5K
utoo-npm	54.67s	33.01s	5.62s	20.39s	641M	73.5K
utoo	32.00s	3.78s	4.84s	17.24s	673M	74.9K

PM	vCtx	iCtx	netRX	netTX	cache	node_mod	lock
bun	1.9K	128.2K	-	-	1.65G	1.91G	2M
utoo-npm	1.8K	374.4K	-	-	1.61G	1.87G	2M
utoo	1.6K	369.5K	-	-	1.61G	1.87G	2M

p4_warm_link

PM	wall	±σ	user	sys	RSS	pgMinor
bun	5.15s	0.54s	0.08s	2.14s	44M	3.4K
utoo-npm	3.42s	0.10s	0.49s	2.54s	91M	6.8K
utoo	3.75s	0.48s	0.49s	2.76s	87M	6.6K

PM	vCtx	iCtx	netRX	netTX	cache	node_mod	lock
bun	13.1K	444	-	-	1.75G	1.90G	2M
utoo-npm	12.8K	69.3K	-	-	1.61G	1.82G	2M
utoo	12.6K	71.5K	-	-	1.61G	1.82G	2M

elrrrrrrr and others added 4 commits April 27, 2026 21:04

elrrrrrrr added benchmark Run pm-bench on PR bench-phases Run pm-bench-phases workflow labels Apr 27, 2026

chore(pm): restore next.js submodule + tombi format

26ab021

gemini-code-assist Bot reviewed Apr 27, 2026

View reviewed changes

elrrrrrrr added 2 commits April 27, 2026 22:33

chore: tombi format Cargo.toml after aws-lc-rs cfg-gate

9f2f69a

elrrrrrrr mentioned this pull request Apr 27, 2026

perf(pm): probe — WP + OnceMap (no aws-lc-rs) #2840

Closed

fix(pm-bench-pcap): self-clone project if missing (dispatch-only path)

0c78862

elrrrrrrr mentioned this pull request Apr 28, 2026

perf(pm): iter1 — HTTP layer (gzip+brotli, DNS, conn pool, fire-and-forget) #2854

Closed

elrrrrrrr closed this Apr 28, 2026

elrrrrrrr mentioned this pull request Apr 29, 2026

perf(pm): preload worker-pool replaces FuturesUnordered #2869

Closed

4 of 6 tasks

Navigation Menu

Repository navigation

perf(pm): minimum-viable preload bundle (worker-pool + OnceMap + aws-lc-rs) - #2838

perf(pm): minimum-viable preload bundle (worker-pool + OnceMap + aws-lc-rs)#2838
elrrrrrrr wants to merge 8 commits into
nextfrom
perf/pm-perf-bundle

elrrrrrrr commented Apr 27, 2026

Uh oh!

gemini-code-assist Bot left a comment

Uh oh!

gemini-code-assist Bot Apr 27, 2026

Uh oh!

gemini-code-assist Bot Apr 27, 2026

Uh oh!

gemini-code-assist Bot Apr 27, 2026

Uh oh!

gemini-code-assist Bot Apr 27, 2026

Uh oh!

gemini-code-assist Bot Apr 27, 2026

Uh oh!

github-actions Bot commented Apr 27, 2026

Uh oh!

github-actions Bot commented Apr 27, 2026

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

elrrrrrrr commented Apr 27, 2026

Summary

Driver-hunt history

Local data

Test plan

Uh oh!

gemini-code-assist Bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

gemini-code-assist Bot Apr 27, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist Bot Apr 27, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist Bot Apr 27, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist Bot Apr 27, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist Bot Apr 27, 2026

Choose a reason for hiding this comment

Uh oh!

github-actions Bot commented Apr 27, 2026

📊 pm-bench-phases · 33388e0 · linux (ubuntu-latest)

npmjs.org

p0_full_cold

p1_resolve

p3_cold_install

p4_warm_link

npmmirror.com

p0_full_cold

p1_resolve

p3_cold_install

p4_warm_link

Uh oh!

github-actions Bot commented Apr 27, 2026

📊 pm-bench-phases · 33388e0 · mac (macos-latest)

npmjs.org

p0_full_cold

p1_resolve

p3_cold_install

p4_warm_link

npmmirror.com

p0_full_cold

p1_resolve

p3_cold_install

p4_warm_link

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

📊 pm-bench-phases · `33388e0` · linux (`ubuntu-latest`)

📊 pm-bench-phases · `33388e0` · mac (`macos-latest`)