These measurements are in-repository baselines for regression tracking, not performance claims across machines. New benchmark updates should append a dated row instead of overwriting prior entries.
cargo bench -p robustone-core --bench riscv_decodecargo bench -p robustone-cli --bench cli_end_to_end
| Date | Commit | Benchmark | Result | Notes |
|---|---|---|---|---|
| 2026-03-20 | 258a109 |
riscv32_decode_ir |
815.06 ns to 819.98 ns |
Pure shared-IR decode throughput baseline. |
| 2026-03-20 | 258a109 |
riscv32_detail_overhead |
1.0724 us to 1.0748 us |
Detail rendering overhead on top of decode. |
| 2026-03-20 | 258a109 |
riscv32_compat_text_render |
1.6670 us to 1.6872 us |
Capstone-style text rendering baseline. |
| 2026-03-20 | 258a109 |
riscv32_cli_end_to_end |
16.121 us to 16.617 us |
End-to-end CLI path through parsing, config, execution, and formatting. |
- Add a new dated row for each benchmark refresh.
- Keep the command name and the commit that produced the sample.
- Do not delete older rows unless the benchmark itself is removed from the repository, and document that removal in the same change.
Local smoke runs completed successfully on 2026-03-20 with:
cargo +nightly fuzz run decode_riscv -- -max_total_time=5cargo +nightly fuzz run format_riscv_json -- -max_total_time=5
Both runs completed without crashes in that environment.