Tighten benchmark guardrail ratio to 0.05

Author: ealdent

.github/workflows/ci.yml

@@ -335,7 +335,7 @@ jobs:
 
     env:
       BENCH_RUNS: "1"
-      BENCH_RUST_TO_PURE_MAX_RATIO: "0.10"
+      BENCH_RUST_TO_PURE_MAX_RATIO: "0.05"
 
     steps:
       - uses: actions/checkout@v4

bin/check-rust-benchmark

@@ -16,7 +16,7 @@ def usage
       BENCH_TOPICS                    Number of topics (default: 8)
       BENCH_MAX_ITER                  Variational max iterations (default: 20)
       BENCH_EM_MAX_ITER               EM max iterations (default: 40)
-      BENCH_RUST_TO_PURE_MAX_RATIO    Maximum allowed rust_mean/pure_mean ratio (default: 0.10)
+      BENCH_RUST_TO_PURE_MAX_RATIO    Maximum allowed rust_mean/pure_mean ratio (default: 0.05)
   USAGE
 end
 
@@ -40,7 +40,7 @@ benchmark_env = {
   "BENCH_MAX_ITER" => ENV.fetch("BENCH_MAX_ITER", "20"),
   "BENCH_EM_MAX_ITER" => ENV.fetch("BENCH_EM_MAX_ITER", "40")
 }
-threshold = Float(ENV.fetch("BENCH_RUST_TO_PURE_MAX_RATIO", "0.10"))
+threshold = Float(ENV.fetch("BENCH_RUST_TO_PURE_MAX_RATIO", "0.05"))
 
 stdout, stderr, status = Open3.capture3(benchmark_env, *command, chdir: repo_root)
 unless status.success?

docs/modernization-handoff.md

@@ -179,7 +179,7 @@ Performance tracking:
 - `./bin/benchmark-backends`
 - `./bin/check-rust-benchmark`
 - `docs/rust-orchestration-guardrails.md`
-- CI currently enforces `BENCH_RUST_TO_PURE_MAX_RATIO=0.10` in `benchmark-guardrail`
+- CI currently enforces `BENCH_RUST_TO_PURE_MAX_RATIO=0.05` in `benchmark-guardrail`
 
 ## CI Jobs Expected
 

docs/porting-strategy.md

@@ -68,7 +68,7 @@ Completed in `codex/experiment-ruby3-modernization`:
 - Gem packaging now excludes local Rust cargo build artifacts (`target/**`) for clean release builds.
 - Backend benchmark driver added (`bin/benchmark-backends`) to track pure/native/rust runtime deltas.
 - Rust orchestration guardrail policy documented (`docs/rust-orchestration-guardrails.md`) with benchmark threshold checker (`bin/check-rust-benchmark`).
-- CI benchmark guardrail job added (`benchmark-guardrail`) to enforce Rust/pure runtime ratio on Ubuntu (currently `BENCH_RUST_TO_PURE_MAX_RATIO=0.10`).
+- CI benchmark guardrail job added (`benchmark-guardrail`) to enforce Rust/pure runtime ratio on Ubuntu (currently `BENCH_RUST_TO_PURE_MAX_RATIO=0.05`).
 - Source install path now has explicit Rust build policy via `LDA_RUBY_RUST_BUILD=auto|always|never`.
 - Docker install-policy matrix script added (`bin/docker-test-install-policies`) to verify source install behavior across environments.
 - CI now runs install-policy matrix checks on Ubuntu.

docs/rust-orchestration-guardrails.md

@@ -25,10 +25,10 @@ Run:
 
 Default benchmark policy:
 
-- `BENCH_RUST_TO_PURE_MAX_RATIO=0.10`
-  - i.e., Rust mean runtime must be no worse than 10% of pure mean runtime on the benchmark fixture/config.
+- `BENCH_RUST_TO_PURE_MAX_RATIO=0.05`
+  - i.e., Rust mean runtime must be no worse than 5% of pure mean runtime on the benchmark fixture/config.
 - CI benchmark guardrail job enforces the same ratio with `BENCH_RUNS=1` for runtime stability.
-- latest tightening evidence (2026-03-02): 12 consecutive successful CI benchmark-guardrail runs on `codex/rust-orchestration-phase8` (`22555725309` .. `22556698440`) observed Rust/Pure ratios in `[0.0252, 0.0288]`, supporting a tighter `0.10` threshold with headroom.
+- latest tightening evidence (2026-03-02): 17 consecutive successful CI benchmark-guardrail runs on `codex/rust-orchestration-phase8` (`22555725309` .. `22557953998`) observed Rust/Pure ratios in `[0.0252, 0.0288]`, supporting a tighter `0.05` threshold with headroom.
 
 Configurable environment knobs:
 
@@ -37,7 +37,7 @@ Configurable environment knobs:
 - `BENCH_TOPICS` (default `8`)
 - `BENCH_MAX_ITER` (default `20`)
 - `BENCH_EM_MAX_ITER` (default `40`)
-- `BENCH_RUST_TO_PURE_MAX_RATIO` (default `0.10`)
+- `BENCH_RUST_TO_PURE_MAX_RATIO` (default `0.05`)
 
 ## When to tighten thresholds