bench: add WASM init and startup timing benchmarks

[natemoo-re]

Two new benchmarks: createTerm and render-minimal, added as fixtures in bench/fixtures/. New CI job runs these on codspeed-macro with mode: walltime to measure the full E2E perf.

Updated benchmark.yml with a single build job where artifacts are shared between simulation + walltime runs. Also pinned SHAs.

[pkg-pr-new]

Open in StackBlitz

npm i https://pkg.pr.new/clayterm@52

commit: 266259e

[codspeed-hq]

Merging this PR will improve performance by 38.61%

⚠️ Different runtime environments detected

Some benchmarks with significant performance changes were compared across different runtime environments,
which may affect the accuracy of the results.

Open the report in CodSpeed to investigate

⚡ 4 improved benchmarks
✅ 14 untouched benchmarks
🆕 2 new benchmarks

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
🆕	WallTime	time to first render	N/A	271.9 ms	N/A
🆕	WallTime	createTerm	N/A	272.7 ms	N/A
⚡	Simulation	large list (50 items)	617.7 µs	409.8 µs	+50.74%
⚡	Simulation	diff render (second frame)	712.1 µs	640.6 µs	+11.17%
⚡	Simulation	printable ASCII (single char)	153.1 µs	88.6 µs	+72.76%
⚡	Simulation	long input burst (200 bytes)	1,249.6 µs	980.1 µs	+27.5%

Tip

Curious why this is faster? Comment @codspeedbot explain why this is faster on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing bench/startup-timing (266259e) with main (c55a136)}

[natemoo-re]

benchmark seems to be a V8 de-opt due to all the benchmarks running in a shared process? digging into it

Edit: 🤞 hopefully addressed by splitting the startup benchmark to a separate process in 6eb4995

Edit v2: yeah, fixed it!

[cowboyd]

Awesome! Including the whole process startup should benefit the brotli case which has smaller source.

[cowboyd]

On the first commit:

Co-Authored-By: Claude Sonnet 4.6 noreply@anthropic.com

Dunno how you feel about it, but for-profit companies whom we are paying good money to use their tools that permanently attach advertisements to our repo by injecting git metadata into commits is a practice I really find off-putting.

So much so that I added it to AGENTS.md https://github.com/bombshell-dev/clayterm/blob/main/AGENTS.md#commit-and-pr-conventions

😄

Any chance you can remove that when squashing?

[cowboyd]

Two more "first-to-x" benchmarks we should consider before splitting out the wasm modules

• create input

[natemoo-re]

Dunno how you feel about it, but for-profit companies whom we are paying good money to use their tools that permanently attach advertisements to our repo by injecting git metadata into commits is a practice I really find off-putting.

@cowboyd yes, fair enough! I apologize for not reading through that doc first. I normally take care to commit everything myself manually but didn't stop the auto-commit here. Bombshell hasn't made any official contribution guidelines for agent-assisted work yet, so this is good motivation to create a policy for the whole org. I'll spin up a thread with everyone in Discord.

[natemoo-re]

Co-author removed! I will add a new benchmark for createInput in the relevant branch.