docs: wasm32 SIMD is a shipped compile-time arm, not a future runtime tier

[AdaWorldAPI]

The drift

src/hpc/simd_dispatch.rs closed its runtime-tier table with:

On wasm32 (future): tier would be WASM SIMD (128-bit, +simd128).

Stale twice over, and the second half is the one that would have cost someone time.

It is not future

crate::simd re-exports the native v128 types from simd_wasm::wasm32_simd under cfg(all(target_arch = "wasm32", target_feature = "simd128")) (simd.rs:402-408), with a full scalar fallback when simd128 is absent (simd.rs:416-421). U32x16 — built as [U32x4; 4] — carries Add/BitXor/rotate_left, the ARX lane the ChaCha20 backend rides.

And it is CI-verified, not merely present: the wasm-simd/parity-node job (ci.yaml:114) builds those real +simd128 types via the excluded wasm-simd-parity cdylib and runs a lane-vs-scalar numeric selfcheck under node 22 (scripts/wasm-parity.sh), plus guards that the chacha20 fork's ndarray_simd backend keeps compiling for wasm32+simd128.

Nor would it ever be a tier in this table

The more useful correction. That table is runtime dispatch over simd_caps() behind a LazyLock. wasm SIMD is selected at compile time — there is no runtime feature probe to dispatch on, so nothing belongs in SimdTier. Phrasing it as "tier would be" invited a future session to add one and discover only afterwards that the mechanism doesn't fit.

The replacement says both things: why wasm is absent from the table by construction, and where the shipped arm actually lives.

Scope

Documentation only. No code, no cfg, no dispatch change — 11 insertions, 1 deletion, one doc comment.

Gates: cargo doc --no-deps -p ndarray produces zero warnings mentioning simd_dispatch (the 111 pre-existing lib-doc warnings are untouched and unrelated); cargo fmt -p ndarray --check clean.

One thing this does not claim

The node job proves the lane bit-exact, and the chacha20 step is compile-only. No keystream-vs-RFC-8439 vectors run on wasm, so a backend-selection bug yielding a valid-looking-but-wrong keystream would pass both. Out of scope here — recorded so it isn't mistaken for covered.

---

Generated by Claude Code

[coderabbitai]

Important

• 🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: efe49d4a-f7cc-4914-930e-07f15f2bce01

---

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

• X
• Mastodon
• Reddit
• LinkedIn

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

[cursor]

Bugbot couldn't run - usage limit reached

Bugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit.

A user or team admin can review and increase usage limits in the Cursor dashboard.

(requestId: serverGenReqId_a72c4e09-4b51-4f37-af9e-fca093fe8586)