Measuring mutexes, spinlocks and how bad the Linux scheduler is

Re:Measuring mutexes, spinlocks and how bad the Linux schedu... by newnewpdro on Wednesday January 01, 2020 @01:47AM (#21926972)

I would like to see the measurements when the threads are explicitly assigned to their respective CPUs.

It surprised me to not see even a mention of CPU affinity.

[ Parent | on HN ]

Re:Measuring mutexes, spinlocks and how bad the Linux schedu... by dpc_pw on Thursday January 02, 2020 @12:12AM (#21933191)

For any sort of benchmarking like that you have to take care of affinities and cpufreq governor at minimum.

[ Parent | on HN ]

Re:Measuring mutexes, spinlocks and how bad the Linux schedu... by newnewpdro on Thursday January 02, 2020 @12:40AM (#21933262)

yep, I was assuming they at least didn't have any frequency scaling or powersave governor junk going on but you're right - it wasn't even mentioned so...

[ Parent | on HN ]