Links to nuggets of erlang performance wisdom.
Nuggets of erlang performance wisdom, from those kind enough to share.
eprof attaches to a running process and and can produces a flat list of functions that are most called, and took the most time.
fprof attaches to a running process, like eprof but gives a hierarchical breakdown of function calls counts and their total time. Useful when there are several calls throughout a process but it is not clear which one is taking the most time.
Loïc Hoguin achieves LOLSPEED in cowboy by ditching OTP behaviours and rolling his own, OTP compliant processes in critical parts of the code.
@ostinelli benchmarks message batching inside application code (as opposed to the TCP stack).
This is still a WIP.