Messages in this thread | | | Subject | Re: [PATCH v3 9/9] sched/topology: Define and use shortcut pointers for wakeup sd_flag scan | From | Dietmar Eggemann <> | Date | Thu, 16 Apr 2020 15:04:22 +0200 |
| |
On 16.04.20 12:24, Valentin Schneider wrote: > > On 16/04/20 08:46, Vincent Guittot wrote: >>> @@ -6657,7 +6646,19 @@ select_task_rq_fair(struct task_struct *p, int prev_cpu, int wake_flags) >>> >>> rcu_read_lock(); >>> >>> - sd = highest_flag_domain(cpu, sd_flag); >>> + switch (wake_flags & (WF_TTWU | WF_FORK | WF_EXEC)) { >>> + case WF_TTWU: >>> + sd_flag = SD_BALANCE_WAKE; >>> + sd = rcu_dereference(per_cpu(sd_balance_wake, cpu)); >> >> It's worth having a direct pointer for the fast path which we always >> try to keep short but the other paths are already slow and will not >> get any benefit of this per cpu pointer. >> We should keep the loop for the slow paths >> > > Which fast/slow paths are you referring to here? want_affine vs > !want_affine? If so, do you then mean that we should do the switch case > only when !want_affine, and otherwise look for the domain via the > for_each_domain() loop?
Coming back to the v2 discussion on this patch
https://lore.kernel.org/r/20200311181601.18314-10-valentin.schneider@arm.com
SD_BALANCE_WAKE is not used in mainline anymore, so wakeups are always fast today.
I.e. you wouldn't need a per_cpu(sd_balance_wake, cpu) since it's always NULL.
I.e. want_affine logic and the 'for_each_domain(cpu, tmp)' isn't needed anymore.
This will dramatically simplify the code in select_task_rq_fair().
But I guess Vincent wants to keep the functionality so we're able to enable SD_BALANCE_WAKE on certain sd's?
| |