perf(wasm-sdk): speed up release WASM optimization by ktechmidas · Pull Request #5147 · dashpay/platform

ktechmidas · 2026-09-28T17:56:49Z

Issue being fixed or feature implemented

The NPM release build is dominated by Binaryen. In the baseline job, yarn build took 87m24s, of which wasm-sdk optimization took 70m41s. There were two causes:

Binaryen sizes its thread pool from the host CPU count. Inside a Docker CPU quota that oversubscribes; profiling found 77% of sampled CPU in kernel futex spin locks.
The full recipe in build-wasm.sh ran -Oz four times with --converge, --flatten/--rereloop and extra passes. It first probed 14 optional flags by optimizing the entire module once per flag.

This consolidates #5152 into this PR.

What was done?

Default Binaryen to four threads in the NPM release build step (BINARYEN_CORES, which a runner-provided value overrides). Cargo's job budget is unchanged. The step logs the thread count and times yarn build.
One standard -Oz pass for full optimization, keeping producer stripping and the same enabled WASM features. Minimal (-O2), no-optimization and the wasm-sdk optimized.wasm copy are unchanged.

No Rust source, dependency, Cargo profile or binding changes.

Measurements (from #5152)

Same 47,166,333-byte pre-optimization wasm-sdk module, Binaryen 121, 20-CPU/32 GiB container:

Optimizer	Wall time	Output bytes	gzip-9 bytes
Original recipe, host-sized pool	~62m	23,682,944	8,175,680
Original recipe, 4 threads	16m41s	23,682,944 (byte-identical)	8,175,680
Single `-Oz`, 4 threads	2m18s	24,164,835	8,203,934

The single pass costs +2.0% raw / +0.35% gzip (~28 KB compressed). The optional -tnh (traps-never-happen) pass is no longer applied. A cold full yarn build on the self-hosted runner completed in 8m01s (single run, build stage only). The previous build's 441 SDK Node tests passed on both optimizer outputs, with identical WASM imports/exports.

How Has This Been Tested?

actionlint, shellcheck and the existing .github/scripts/tests suite pass.
The four applicable contract tests from perf(wasm-sdk): eliminate repeated release optimizer passes #5152 pass against the final build-wasm.sh: optimized.wasm copy, minimal/none modes, missing wasm-opt, and optimizer failure propagation.
Pending: the NPM runner-validation build on this PR is the first CI run of the single -Oz pass. Four threads was measured with the original recipe; it has not been re-tuned for the single pass.

Breaking Changes

None. Released WASM bytes change (slightly larger) because of the optimizer recipe.

Checklist:

I have performed a self-review of my own code
I have commented my code, particularly in hard-to-understand areas
I have added or updated relevant unit/integration/functional/e2e tests
I have added "!" to the title and described breaking changes in the corresponding section if my code contains any
I have made corresponding changes to the documentation if needed
If I added or changed GroveDB structure, I described it in the area's structure.rs, regenerated grovedb-structure.json, and checked the structure viewer link posted on this pull request

For repository code-owners and collaborators only

I have assigned this pull request to a milestone

🤖 Generated with Claude Code

PR Hygiene · `2806459`

Bots — coderabbitai not yet · thepastaclaw not yet — /skip-bots proceeds without the ones not yet reported
Self-review — address PastaPastaPasta left a review thread unresolved, then post /self-reviewed
Within your 5 open PRs
Build running
Approvals
- files with no dedicated owner (.github/actions/npm-release-build/action.yaml, packages/scripts/build-wasm.sh) — QuantumExplorer or shumkov

When every box is checked the PR Hygiene check passes and this can merge.

coderabbitai · 2026-09-28T17:56:58Z

Navigate logical layers of code changes, visualize relationships, and explore their blast radius.

Note

Currently processing new changes in this PR. This may take a few minutes, please wait...

⚙️ Run configuration

Configuration used: Repository: dashpay/platform/.coderabbit.yaml

Review profile: CHILL

Plan: Advanced

Run ID: 7a99ae27-d88b-4696-8ffc-e3a7c8898e51

📥 Commits

Reviewing files that changed from the base of the PR and between d22f8ab and 2806459.

📒 Files selected for processing (2)

.github/actions/npm-release-build/action.yaml
packages/scripts/build-wasm.sh

 _________________________________
< My bunny whiskers are tingling. >
 ---------------------------------
  \
   \   \
        \ /\
        ( )
      .( o ).

✨ Finishing Touches

📝 Generate docstrings

Commit to this branch
Create a new PR

🧪 Generate unit tests (beta)

Commit to this branch
Create a new PR

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

_{Comment @coderabbitai help to get the list of available commands.}

thepastaclaw · 2026-09-28T18:02:23Z

🕓 Review not started yet because this PR is a draft.

Request normal review — click when the PR is ready for review.
Request priority review — click to move this review to the front of the queue.

Commit 2806459. Normal review starts when eligible; priority review starts as soon as a slot is available.

PastaPastaPasta · 2026-09-28T19:47:06Z

+SPEC.loader.exec_module(MODULE)
+
+
+class NpmBuildBudgetTests(unittest.TestCase):


remove this; we don't need tests for a simple build script

Replace the Python launcher and its tests with a few lines in the existing build step: cap BINARYEN_CORES at min(nproc, cgroup v2 cpu.max quota), keep any runner-provided value, log it, and time `yarn build` with the exit status preserved. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

Four optimizer threads were measured ~3x faster than a host-sized pool with byte-identical output, which makes cgroup quota detection unnecessary. A runner-provided BINARYEN_CORES still takes precedence. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

The full-optimization recipe ran -Oz four times with --converge, --flatten/--rereloop and extra passes, and first probed 14 optional flags by optimizing the entire module once per flag. On wasm-sdk this took over an hour with a host-sized thread pool (16m41s at four threads). One standard -Oz pass with the same features and producer stripping takes 2m18s for +2.0% raw / +0.35% gzip size. Minimal (-O2) and no-optimization builds and the wasm-sdk optimized.wasm copy are unchanged. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

PastaPastaPasta

LGTM

Pick up #5156 (validate runner candidates using full commit statuses), #5153 and #5147. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

$@infraclaw-dash$

ci(wasm-sdk): bound optimizer threads independently of compilation

c135e7c

github-actions Bot added this to the v4.2.0 milestone Sep 28, 2026

$@infraclaw-dash$ infraclaw-dash mentioned this pull request Sep 28, 2026

perf(wasm-sdk): eliminate repeated release optimizer passes #5152

Closed

6 of 10 tasks

PastaPastaPasta reviewed Sep 28, 2026

View reviewed changes

PastaPastaPasta and others added 3 commits September 28, 2026 14:51

PastaPastaPasta changed the title ~~ci(wasm-sdk): bound optimizer threads independently of compilation~~ perf(wasm-sdk): speed up release WASM optimization Sep 28, 2026

PastaPastaPasta marked this pull request as ready for review September 28, 2026 20:13

github-actions Bot added the waiting-bots Waiting for the review bots to report on this head label Sep 28, 2026

PastaPastaPasta reviewed Sep 28, 2026

View reviewed changes

ktechmidas merged commit a0a7b3f into v4.2-dev Sep 28, 2026
62 of 64 checks passed

ktechmidas deleted the fix/ci-npm-binaryen-cpu-budget-20260928 branch September 28, 2026 20:14

PastaPastaPasta added a commit that referenced this pull request Sep 28, 2026

chore: merge v4.2-dev into v4.3-dev

5a3fa4a

Pick up #5156 (validate runner candidates using full commit statuses), #5153 and #5147. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

This was referenced Sep 28, 2026

chore!: merge v4.2-dev into v4.3-dev #5151

Closed

wasm-sdk: npm release build links the mocks feature; 25 MB wasm (8.5 MB gzip) dominates browser first load #5263

Open

Navigation Menu

Repository navigation

perf(wasm-sdk): speed up release WASM optimization - #5147

perf(wasm-sdk): speed up release WASM optimization#5147
ktechmidas merged 4 commits into
v4.2-devfrom
fix/ci-npm-binaryen-cpu-budget-20260928

ktechmidas commented Sep 28, 2026 •

edited by github-actions Bot

Loading

Uh oh!

coderabbitai Bot commented Sep 28, 2026 •

edited

Loading

Uh oh!

thepastaclaw commented Sep 28, 2026 •

edited

Loading

Uh oh!

PastaPastaPasta Sep 28, 2026

Uh oh!

PastaPastaPasta left a comment

Uh oh!

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

		SPEC.loader.exec_module(MODULE)


		class NpmBuildBudgetTests(unittest.TestCase):

Conversation

ktechmidas commented Sep 28, 2026 • edited by github-actions Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Issue being fixed or feature implemented

What was done?

Measurements (from #5152)

How Has This Been Tested?

Breaking Changes

Checklist:

PR Hygiene · 2806459

Uh oh!

coderabbitai Bot commented Sep 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

thepastaclaw commented Sep 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

PastaPastaPasta Sep 28, 2026

Choose a reason for hiding this comment

Uh oh!

PastaPastaPasta left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

ktechmidas commented Sep 28, 2026 •

edited by github-actions Bot

Loading

PR Hygiene · `2806459`

coderabbitai Bot commented Sep 28, 2026 •

edited

Loading

thepastaclaw commented Sep 28, 2026 •

edited

Loading