Files
carbon-lang/scripts
Chandler Carruth 0b9d0c7042 Begin adding support for Bencher's BMF format and benchmark running (#8109)
This extends `scripts/bench_runner.py` with `--output=bmf` and other
flags, as well as fixing severaly other flags, all to allow it to
successfully output the schema consumed by Bencher.

Also adds the `bencher_suites.py` and `run_bencher.py` that prepare and
produce uploadable benchmark results. It addresses benchmark naming and
extra metrics collected for each run of the benchmark.

A major chunk of logic is selecting the suites and metrics for
benchmarks that we want to track long term. We define two tiers here,
one to track continuously that we can run reasonbly quickly, and a more
comprehensive set that we track only nightly on each nightly release. We
also make sure each metric from the benchmark is correctly recorded for
tracking, including its units.

Assisted-by: Antigravity with Gemini
2026-11-13 18:33:10 +00:00
..