Messages in this thread Patch in this message | | | Subject | Re: [PATCH v3 5/7] sched: SIS_CORE to disable idle core search | From | Parth Shah <> | Date | Thu, 4 Jul 2019 18:04:49 +0530 |
| |
On 7/2/19 2:07 AM, Subhra Mazumdar wrote: > >>>> Also, systems like POWER9 has sd_llc as a pair of core only. So it >>>> won't benefit from the limits and hence also hiding your code in select_idle_cpu >>>> behind static keys will be much preferred. >>> If it doesn't hurt then I don't see the point. >>> >> So these is the result from POWER9 system with your patches: >> System configuration: 2 Socket, 44 cores, 176 CPUs >> >> Experiment setup: >> =========== >> => Setup 1: >> - 44 tasks doing just while(1), this is to make select_idle_core return -1 most times >> - perf bench sched messaging -g 1 -l 1000000 >> +-----------+--------+--------------+--------+ >> | Baseline | stddev | Patch | stddev | >> +-----------+--------+--------------+--------+ >> | 135 | 3.21 | 158(-17.03%) | 4.69 | >> +-----------+--------+--------------+--------+ >> >> => Setup 2: >> - schbench -m44 -t 1 >> +=======+==========+=========+=========+==========+ >> | %ile | Baseline | stddev | patch | stddev | >> +=======+==========+=========+=========+==========+ >> | 50 | 10 | 3.49 | 10 | 2.29 | >> +-------+----------+---------+---------+----------+ >> | 95 | 467 | 4.47 | 469 | 0.81 | >> +-------+----------+---------+---------+----------+ >> | 99 | 571 | 21.32 | 584 | 18.69 | >> +-------+----------+---------+---------+----------+ >> | 99.5 | 629 | 30.05 | 641 | 20.95 | >> +-------+----------+---------+---------+----------+ >> | 99.9 | 780 | 40.38 | 773 | 44.2 | >> +-------+----------+---------+---------+----------+ >> >> I guess it doesn't make much difference in schbench results but hackbench (perf bench) >> seems to have an observable regression. >> >> >> Best, >> Parth >> > If POWER9 sd_llc has only 2 cores, the behavior shouldn't change much with > the select_idle_cpu changes as the limits are 1 and 2 core. Previously the > lower bound was 4 cpus and upper bound calculated by the prop. Now it is > 1 core (4 cpus on SMT4) and upper bound 2 cores. Could it be the extra > computation of cpumask_weight causing the regression rather than the > sliding window itself (one way to check this would be hardcode 4 in place > of topology_sibling_weight)? Or is it the L1 cache coherency? I am a bit > suprised because SPARC SMT8 which has more cores in sd_llc and L1 cache per > core showed improvement with Hackbench. >
Same experiment with hackbench and with perf analysis shows increase in L1 cache miss rate with these patches (Lower is better) Baseline(%) Patch(%) ----------------------- ------------- ----------- Total Cache miss rate 17.01 19(-11%) L1 icache miss rate 5.45 6.7(-22%)
So is is possible for idle_cpu search to try checking target_cpu first and then goto sliding window if not found. Below diff works as expected in IBM POWER9 system and resolves the problem of far wakeup upto large extent.
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index ff2e9b5c3ac5..fae035ce1162 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -6161,6 +6161,7 @@ static int select_idle_cpu(struct task_struct *p, struct sched_domain *sd, int t u64 time, cost; s64 delta; int cpu, limit, floor, target_tmp, nr = INT_MAX; + struct cpumask *cpus = this_cpu_cpumask_var_ptr(select_idle_mask); this_sd = rcu_dereference(*this_cpu_ptr(&sd_llc)); if (!this_sd) @@ -6198,16 +6199,22 @@ static int select_idle_cpu(struct task_struct *p, struct sched_domain *sd, int t time = local_clock(); - for_each_cpu_wrap(cpu, sched_domain_span(sd), target_tmp) { + cpumask_and(cpus, sched_domain_span(sd), &p->cpus_allowed); + for_each_cpu_wrap(cpu, cpu_smt_mask(target), target) { + __cpumask_clear_cpu(cpu, cpus); + if (available_idle_cpu(cpu)) + goto idle_cpu_exit; + } + + for_each_cpu_wrap(cpu, cpus, target_tmp) { per_cpu(next_cpu, target) = cpu; if (!--nr) return -1; - if (!cpumask_test_cpu(cpu, &p->cpus_allowed)) - continue; if (available_idle_cpu(cpu)) break; } +idle_cpu_exit: time = local_clock() - time; cost = this_sd->avg_scan_cost; delta = (s64)(time - cost) / 8;
Best, Parth
| |