perf(encoder): Fast L1 +7.43% ratio gap vs donor on decodecorpus-z000033

[polaz]

Context

Follow-up to phase 3 of #198 (PR #219). With per-level Fast cParams threaded end-to-end and donor-shape 4-cursor kernel, L1 still loses ~7.43% on ratio vs donor C zstd on decodecorpus-z000033.

Current measurements (post-M8, b280b48)

Level	rust_bytes	ffi_bytes	Δ
L1 Fast	612905	570507	+7.43%

Sequence comparator (M7): 57.7% match rate vs donor, 6736 ffi_only seqs (matches donor finds we miss), 2805 rust_only.

What we tried

M7 added donor's missing current0+2 hash insertion after match emits (zstd_fast.c:407). Improved match rate from 43.1% → 57.7% but only saved ~3k bytes (612905 from 615899).

M8 removed RESERVED_PREFIX_BYTES dummy on hypothesis that byte-range alignment was off. Output identical to M7 — hypothesis wrong; positions are aligned with donor in both layouts.

Suspected residual causes

Things NOT investigated:

• Step doubling cadence (step++ at ip2 >= next_step) timing differences
• Bitwise vs logical AND in backward-extension bound
• Repcode-at-ip2 vs explicit-at-ip0 priority on tie
• Hash 32-bit truncation differences at very small mls values
• min_literals_to_compress / min_gain interaction at L1 specifically

Acceptance

Close L1 ratio gap to within ±2% of donor on decodecorpus-z000033.

Part of #198 phase 3 follow-up (PR #219 shipped M1-M8 without closing this last gap).

[coderabbitai]

🔗 Related PRs

#34 - fix(encoding): implement default compression level [merged]
#39 - perf(encoding): align fastest matcher with zstd fast path [merged]
#125 - perf(encoding): #124 Phase 4 — HC speculative tail check + MatcherStorage enum dispatch [merged]
#179 - perf(encoder): donor-parity greedy parse at L4 — ratio + speed win [merged]
#187 - perf(encoder): bump Fast window_log to 19 (donor parity, large-log-stream ratio fix) [merged]

---

📝 Issue Planner

_{Check the box below or use the @coderabbitai plan command to generate an implementation plan and prompts that you can use with your favorite coding assistant.}

• Create Plan

---

🧪 Issue enrichment is currently in open beta.

You can configure auto-planning by selecting labels in the issue_enrichment configuration.

To disable automatic issue enrichment, add the following to your .coderabbit.yaml:

issue_enrichment:
  auto_enrich:
    enabled: false

💬 Have feedback or questions? Drop into our discord!