Again on reaching optimal parallel scaling

[carstenbauer]

@tkf

Another way to profile the program is to look at the performance counters using GitHub - JuliaPerf/LinuxPerf.jl (or just perf stat command) if you use Linux.

FWIW, there also is GitHub - JuliaPerf/LIKWID.jl: Julia wrapper for the performance monitoring and benchmarking suite LIKWID. which has at least some documentation.