simd: U64x8::mul_lo32 — widening lo32×lo32→u64 on every backend (argon2 BlaMka)

[AdaWorldAPI]

What this adds

U64x8::mul_lo32(self, rhs) computes, lane by lane, lo32(self) × lo32(rhs) as an exact u64.

• Semantics: the high 32 bits of each input lane are ignored, and the product can't overflow because (2³²−1)² < 2⁶⁴.
• Why: it's the multiply inside argon2's BlaMka function, a + b + 2·lo32(a)·lo32(b), which runs on a live path: ogar-encryption for a2ui session keys, and ogar-auth for logins. It's also the limb multiply that radix-2²⁶ Poly1305 and curve25519 arithmetic need.

backend	body
AVX-512	_mm512_mul_epu32 (VPMULUDQ zmm)
AVX2	_mm256_mul_epu32 on each 256-bit half
NEON	vmovn_u64 + vmull_u32 (UMULL) per quad
wasm	i32x4_shuffle::<0,2,0,2> + u64x2_extmul_low_u32x4 per v128
scalar	per-lane reference loop
nightly	(a & 0xFFFF_FFFF) * (b & 0xFFFF_FFFF) on core::simd::u64x8

AVX-512 and NEON name the intrinsic rather than relying on portable source. A codegen probe showed that inside BlaMka's add chain, LLVM lowered the portable form to VPMULLQ (the full 64-bit multiply) on x86-64-v4, and left it as a scalar madd on aarch64.

Tests

• simd::tests::u64x8_mul_lo32_matches_scalar:
  • checks every lane against the scalar definition;
  • every operand lane has non-zero high bits and is chosen so a full 64-bit multiply gives a different answer, and the test asserts that;
  • five lanes have products wider than 32 bits, so a truncating 32-bit multiply fails too;
  • also checks commutativity, and argon2's fBlaMka (RFC 9106 §3.5) built from mul_lo32.
• neon-simd-parity (qemu) and wasm-simd-parity (node) gain the same check, as return code 0x30F.

Run locally, all green:

• the new test on native AVX-512 and on pinned x86-64-v3;
• scripts/neon-parity.sh and scripts/wasm-parity.sh;
• full cargo test --lib on native: 2478 passed;
• clippy -D warnings on native and on v3.

Disable runs, each one red:

• AVX-512 changed to _mm512_mullo_epi64 → the lane 0 assertion fails;
• NEON changed to multiply the high halves via vshrn → rc 783 (0x30F);
• wasm shuffle changed to <1,3,1,3> → rc 783.

Not run locally: the nightly-simd arm; there's no nightly toolchain here, and CI's nightly rows cover it.

Note for reviewers

cargo --config X clippy does not apply X. clippy is an external subcommand, and cargo doesn't forward a --config given before it: rustc ran with target-cpu=native only. Use cargo clippy --config X. This is recorded on the blackboard.

🤖 Generated with Claude Code

https://claude.ai/code/session_019HnekoM1EidTwQLS3oFVFm

---

Generated by Claude Code

Summary by CodeRabbit

• New Features
  • Added lane-wise multiplication of the low 32 bits of each pair of 64-bit values, returning exact 64-bit products across supported SIMD implementations.
• Tests
  • Added checks against scalar results, including commutativity and wrapping-arithmetic cases. NEON and WebAssembly parity checks now cover the operation.

[coderabbitai]

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

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Essentials

Run ID: a4b7dafb-fd63-45dd-b7e0-5bb1487d1019

📥 Commits

Reviewing files that changed from the base of the PR and between a1ad0bc and 2cd4246.

📒 Files selected for processing (10)

• .claude/blackboard.md
• crates/neon-simd-parity/src/main.rs
• crates/wasm-simd-parity/src/lib.rs
• src/simd.rs
• src/simd_avx2.rs
• src/simd_avx512.rs
• src/simd_neon.rs
• src/simd_nightly/u_word_types.rs
• src/simd_scalar.rs
• src/simd_wasm.rs

Included review availability: 2 reviews are currently available. Your included PR review attempts over the past 7 days set your current allowance at 3 reviews per hour.

---

📝 Walkthrough

Walkthrough

U64x8::mul_lo32 is added to six SIMD implementations. It multiplies the low 32 bits of each pair of lanes and returns an exact 64-bit product. Tests and NEON and WebAssembly parity checks compare its results with scalar multiplication.

Changes

U64x8 multiplication

Layer / File(s)	Summary
Implement lane-wise multiplication src/simd_scalar.rs, src/simd_nightly/u_word_types.rs, src/simd_avx2.rs, src/simd_avx512.rs, src/simd_neon.rs, src/simd_wasm.rs	Adds U64x8::mul_lo32 to six SIMD implementations. Each implementation ignores the upper 32 bits and returns the exact 32×32-bit product in each 64-bit lane.
Validate multiplication results src/simd.rs, crates/neon-simd-parity/src/main.rs, crates/wasm-simd-parity/src/lib.rs, .claude/blackboard.md	Adds scalar comparison, commutativity, and BlaMka arithmetic tests. Adds lane-wise parity checks for NEON and WebAssembly. The blackboard records implementation and test details, a Cargo Clippy configuration caveat, and a proposed Argon2 follow-up.

Priority: ➖ Normal

Estimated code review effort: 3 (Moderate) | ~20 minutes

Change: Feature

Suggested reviewers: claude

Merge Risk: ⚪ Minimal · up to 2cd42

No identified defect blocks merging after normal checks.

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly and concisely describes the main change: adding widening low-32-bit multiplication for U64x8 across all SIMD backends, with its Argon2 BlaMka use case.
Docstring Coverage	✅ Passed	Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 18 functions across 9 files. (1 skipped: 1…
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

📝 Generate docstrings

• Commit to this branch
• Create a new PR

---

_{A rabbit checks each lane with care,
Low bits meet, and products pair.
Six paths return the same result,
Tests confirm each value built.
Then off I hop through fields of clover.}

_{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_2c1487f0-a405-4bb4-95c3-72ea456c61e9)