Hi Patrick, Thank you for the patch! Yet something to improve:
[auto build test ERROR on tip/sched/core] [also build test ERROR on v4.17 next-20180604] [if your patch is applied to the wrong git tree, please drop us a note to help improve the system] url: https://github.com/0day-ci/linux/commits/Patrick-Bellasi/sched-fair-pelt-use-u32-for-util_avg/20180605-082640 config: i386-randconfig-s0-201822 (attached as .config) compiler: gcc-6 (Debian 6.4.0-9) 6.4.0 20171026 reproduce: # save the attached .config to linux build tree make ARCH=i386 All errors (new ones prefixed by >>): kernel/sched/fair.o: In function `post_init_entity_util_avg': >> kernel/sched/fair.c:761: undefined reference to `__udivdi3' vim +761 kernel/sched/fair.c 724 725 /* 726 * With new tasks being created, their initial util_avgs are extrapolated 727 * based on the cfs_rq's current util_avg: 728 * 729 * util_avg = cfs_rq->util_avg / (cfs_rq->load_avg + 1) * se.load.weight 730 * 731 * However, in many cases, the above util_avg does not give a desired 732 * value. Moreover, the sum of the util_avgs may be divergent, such 733 * as when the series is a harmonic series. 734 * 735 * To solve this problem, we also cap the util_avg of successive tasks to 736 * only 1/2 of the left utilization budget: 737 * 738 * util_avg_cap = (1024 - cfs_rq->avg.util_avg) / 2^n 739 * 740 * where n denotes the nth task. 741 * 742 * For example, a simplest series from the beginning would be like: 743 * 744 * task util_avg: 512, 256, 128, 64, 32, 16, 8, ... 745 * cfs_rq util_avg: 512, 768, 896, 960, 992, 1008, 1016, ... 746 * 747 * Finally, that extrapolated util_avg is clamped to the cap (util_avg_cap) 748 * if util_avg > util_avg_cap. 749 */ 750 void post_init_entity_util_avg(struct sched_entity *se) 751 { 752 struct cfs_rq *cfs_rq = cfs_rq_of(se); 753 long cap = (long)(SCHED_CAPACITY_SCALE - cfs_rq->avg.util_avg) / 2; 754 755 if (cap > 0) { 756 struct sched_avg *sa = &se->avg; 757 u64 util_avg = READ_ONCE(sa->util_avg); 758 759 if (cfs_rq->avg.util_avg != 0) { 760 util_avg = cfs_rq->avg.util_avg * se->load.weight; > 761 util_avg /= (cfs_rq->avg.load_avg + 1); 762 if (util_avg > cap) 763 util_avg = cap; 764 } else { 765 util_avg = cap; 766 } 767 768 WRITE_ONCE(sa->util_avg, util_avg); 769 } 770 771 if (entity_is_task(se)) { 772 struct task_struct *p = task_of(se); 773 if (p->sched_class != &fair_sched_class) { 774 /* 775 * For !fair tasks do: 776 * 777 update_cfs_rq_load_avg(now, cfs_rq); 778 attach_entity_load_avg(cfs_rq, se, 0); 779 switched_from_fair(rq, p); 780 * 781 * such that the next switched_to_fair() has the 782 * expected state. 783 */ 784 se->avg.last_update_time = cfs_rq_clock_task(cfs_rq); 785 return; 786 } 787 } 788 789 attach_entity_cfs_rq(se); 790 } 791 --- 0-DAY kernel test infrastructure Open Source Technology Center https://lists.01.org/pipermail/kbuild-all Intel Corporation
.config.gz
Description: application/gzip
