suggested workflow for profiling benchmarks?

[BurntSushi]

A common workflow with criterion (and #[bench] before that) for me is to run the benchmarks without collecting results. This is criterion's --profile-time flag, for example. Then I can run them under a profiler like perf. I scanned the blog but couldn't find anything about profiling. I see #38, but I would prefer something that just lets me run the benchmark with the harness getting out of the way as much as possible. Then I can hook it up to any profiling tool.