On 12/06/15 06:34, Amit Kucheria wrote:
On Thu, Jun 11, 2015 at 6:06 PM, Alex Shi alex.shi@linaro.org wrote:
On 04/28/2015 06:21 PM, Amit Kucheria wrote:
Hi Alex,
In 4.1-rc1, several patches (see 36ee28e4 onwards) related to cpu capacity consolidation were merged.
Hi, Amit,
Is this the only serial need to pick up? or still sth I missed?
d4573c3 sched: Improve load balancing in the presence of idle CPUs dfbca41 sched: Optimize freq invariant accounting 1aaf90a sched: Move CFS tasks to CPUs with higher capacity caff37e sched: Add SD_PREFER_SIBLING for SMT level dc7ff76 sched: Remove unused struct sched_group_capacity::capacity_orig ea67821 sched: Replace capacity_factor by usage 8bb5b00 sched: Calculate CPU's usage statistic and put it into struct sg_lb_stats::group_usage ca6d75e sched: Add struct rq::cpu_capacity_orig b5b4860 sched: Make scale_rt invariant with frequency 0c1dc6b sched: Make sched entity usage tracking scale-invariant a8faa8f sched: Remove frequency scaling from cpu_capacity 21f4486 sched: Track group sched_entity usage contributions 36ee28e sched: Add sched_avg::utilization_avg_contrib
Yes, that looks correct.
Looking closely, perhaps we should also backport dfbca41f34. Juri?
Yes, we need that too (dfbca41f3479 "sched: Optimize freq invariant accounting"), but is it not already in Alex's list above?
I would also suggest backporting following commits, that make EAS patches apply more cleanly:
638476007d13 "sched/fair: Fix the dealing with decay_count in __synchronize_entity_decay()" cb0b9f2445cd "sched/fair: Make update_sd_pick_busiest() return 'true' on a busier sd" 6c1d9410f007 "sched: Move p->nr_cpus_allowed check to select_task_rq()"
Best,
- Juri