Skip to content

README.md

Latest commit

 

History

History
41 lines (29 loc) · 1.33 KB

README.md

File metadata and controls

41 lines (29 loc) · 1.33 KB
Raw

Opt-Level Evaluation Harness

This directory contains paired safe/unsafe crate copies for target-function-level optimization-sensitivity experiments.

Layout

  • crates/itertools-safe and crates/itertools-unsafe: kmerge path reaching sift_down.
  • crates/rand-safe and crates/rand-unsafe: partial_shuffle.
  • crates/ring-safe and crates/ring-unsafe: HKDF Okm::fill path reaching fill_okm.
  • crates/arrayvec-safe and crates/arrayvec-unsafe: swap_pop.
  • crates/bit-vec-safe and crates/bit-vec-unsafe: push.

Each crate pair has a bypasser_target Criterion benchmark, except ring, whose benchmark is in its existing bench package as bypasser_hkdf.

Running One Case

cd evaluation/crates/arrayvec-safe
CARGO_PROFILE_BENCH_OPT_LEVEL=2 cargo bench --bench bypasser_target -- --noplot

cd ../arrayvec-unsafe
CARGO_PROFILE_BENCH_OPT_LEVEL=2 cargo bench --bench bypasser_target -- --noplot

For ring, run from the nested benchmark package:

cd evaluation/crates/ring-safe/bench
CARGO_PROFILE_BENCH_OPT_LEVEL=2 cargo bench --bench bypasser_hkdf -- --noplot

Running All Configurations

bash evaluation/run_opt_level_benches.sh

The script evaluates opt-level=1, opt-level=2, and opt-level=3 for each safe/unsafe pair and stores raw Criterion logs under evaluation/results/.