release: v1.0.0

[avrabe]

Summary

Bump workspace version 0.9.0 → 1.0.0.

v1.0.0 — verifier-completion release. Lifts the verifier-side blocker that kept cross-call CSE dedup dormant for two releases, adds the size-threshold fallback that unblocks LOOM on large bodies, and ships a corpus-wide cargo bench with wasm-opt version pinning.

What landed

PR	Title
#115	feat(verify): PR-K3 model pure+no-trap Call as uninterpreted function + PR-K3.2 size-threshold fallback (Track A + Track B)
#116	feat(bench): criterion-based corpus baseline + wasm-opt version pinning (Track E)

What v1.0.0 means

v1.0.0 marks the point where LOOM's cross-call optimization infrastructure is end-to-end functional:

• PR-K (v0.8.0): IPA recognition of Expr::Call
• PR-K2 (v0.9.0): span-based replacement
• PR-K3 (v1.0.0): verifier-side uninterpreted-function encoding
• PR-K3.2 (v1.0.0): size-threshold fallback (closes 60min+ timeouts)
• PR-bench (v1.0.0): cargo bench harness + wasm-opt version pin

Combined with the measurement harness (v0.9.0 PR-P + PR-R), LOOM ships its first release where the optimization-vs-baseline story is fully measurable and the strategic moat (Component-Model adapter specialization) is empirically validated.

Tests

• 5 CSE tests now pass (was 3 pass + 2 ignored since v0.9.0):
  • test_cse_dedupes_repeated_pure_calls
  • test_cse_dedupes_pure_clamp_calls_via_span_replacement
  • 3 safety tests (impure / may-trap / different-args)
• 10 summary:: IPA tests pass.
• 330+ loom-core lib tests total.

Deferred from v1.0.0 sprint (planned for v1.0.1)

• Track C: directize MVP — agent stalled mid-task.
• Track D: real corpus fixtures (httparse, nom_numbers, etc.) — agent stalled.

The harness honest-marks the missing fixtures n/a.

Still tracked (future)

• PR-L3: power-of-2 mul/div → shift rules
• Pure-arithmetic-expression arms in canonicalize
• Cranelift-style acyclic ægraph mid-end
• Z3 verifier completeness (full exit-state equivalence, BrTable path predicates, float rustc_apfloat)

🤖 Generated with Claude Code