add script to run hash algorithm benchmark

[spencerschrock]

Summary

Builds upon the work in #306 and starts to define individual experiments. This one is aimed specifically at hashing algorithm.

As far as hashing is concerned, bytes are bytes. By generating our own bytes, we avoid I/O associated with reading models from disk. While we could read actual models into memory, recreating the filesystem seems unecessary for this benchmark.

$ hatch run +py=3.11 bench:hash --repeat 5 --methods sha256 blake2
algorithm: sha256, size: 1.0 KB, best time: 2.569984644651413e-06s
algorithm: blake2, size: 1.0 KB, best time: 3.1301751732826233e-06s
algorithm: sha256, size: 1.0 MB, best time: 0.0007016910240054131s
algorithm: blake2, size: 1.0 MB, best time: 0.001357788685709238s
algorithm: sha256, size: 512.0 MB, best time: 0.41755866911262274s
algorithm: blake2, size: 512.0 MB, best time: 0.7319729649461806s
algorithm: sha256, size: 1.0 GB, best time: 0.8285107491537929s
algorithm: blake2, size: 1.0 GB, best time: 1.466184071265161s
algorithm: sha256, size: 4.0 GB, best time: 3.375688728876412s
algorithm: blake2, size: 4.0 GB, best time: 5.8439111188054085s
algorithm: sha256, size: 16.0 GB, best time: 14.146526537369937s
algorithm: blake2, size: 16.0 GB, best time: 23.482591222040355s
algorithm: sha256, size: 32.0 GB, best time: 21.890016920864582s
algorithm: blake2, size: 32.0 GB, best time: 42.17644724994898s

Release Note

NONE

Documentation

NONE

[mihaimaruseac]

Thank you, it looks great. I have a few comments / discussion starters to make this useful both for humans and machines (plotting, comparing between runs)