Open mhdawson opened 9 years ago
IMO the most reliable things to micro-benchmark are native calls (e.g. Buffer#slice()
). Will definitely want to exclude micro calls (e.g. Buffer#writeUInt8()
).
May want to consider implementing some for the complex JS paths (e.g. several streams paths).
This issue is for discussion on the existing micro-benchmarks, what we do in order to improve and get the running again.
One suggestion so far is that we make sure they print and display time for single iteration (how many us to complete) were the time is both the wall time as well as the cycles used.
We also discussed coming up with a list of additional micro benchmarks that we might want to add