thoughts on iteration count?

[rukai]

I've noticed that criterion will run far more iterations than divan, such that criterion benches will take multiple seconds to complete while divan benches will complete in 10-20ms

Do you have any idea why criterion takes such an aggressive approach while divan can get away with far fewer iterations?

Do you think criterion is being needlessly overcautious?
Or is divan doing something different to criterion such that it needs far fewer samples?