On 25.03.21 16:31, Andriy Gapon wrote: > [...] > There are some tools in tools/sched/ directory. > schedgraph.py can be used for visual inspection of scheduling traces collected > using KTR. The file has instructions on how to collect them. > Alternatively, schedgraph.d can be used to collect such traces. > If anyone affected can gather a short sample that captures the problem, then > there might be someone who would be willing to look at them. what should I tell? I've set "kern.sched.steal_thresh: 0 -> 2" and reliably got that micro-suttering back while watching YouTube videos (tearing test i.e.). Then I loaded kernel modules "dtrace.ko" and "dtraceall.ko" and ran ./schedgraph.d > /tmp/sched.out The micro-stuttering immediately went away while above DTrace-script was running. So it seems that the light load of the running DTrace-script was enough to eliminate any micro-stutterings. I've uploaded resulting .ktr- and .out-files (for both steal_thresh=2/0) here: kern.sched.steal_thresh=2 ------------------------- http://156.67.189.93:9080/sched.steal2.txz kern.sched.steal_thresh=0 ------------------------- http://156.67.189.93:9080/sched.steal0.txz Haven't tried your mentioned first KTR-variant yet... Thanks and BR, NilsReceived on Fri Mar 26 2021 - 08:55:50 UTC
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:41:27 UTC