Continuous benchmarking

[tarcieri]

We've had a couple notable performance regressions lately, namely in keccak and p256.

It'd be good to have a reusable solution for continuous benchmarking.

Some options:

• https://github.com/bencherdev/bencher
• https://github.com/bheisler/iai

[Slixe]

Criterion may be a good alternative too:

• https://github.com/bheisler/criterion.rs

EDIT: Looks like you're already using criterion crate in some parts of code.

[tarcieri]

Yeah, we already use criterion extensively.

For this particular problem though, we need to be able to produce reliable benchmarking results in a virtualized CI environment, which is not something criterion provides a solution for.

That particular problem is what makes something like iai interesting.

[Slixe]

Agree, iai looks promising, and the perfect choice for benchmarks of hashing algorithms.