Messages in this thread | | | From | Vincent Guittot <> | Date | Tue, 2 Apr 2024 10:49:56 +0200 | Subject | Re: [PATCH v2] sched/fair: fix initial util_avg calculation |
| |
On Tue, 2 Apr 2024 at 10:44, Vishal Chourasia <vishalc@linux.ibm.com> wrote: > > On 02/04/24 11:17 am, Dawei Li wrote: > > Hi Vishal > > > > Thanks for the comment! > > Do you suggest using scale_load_down() in place of se_weight()? > scale_load_down should be better.
se_weight is used for computing sched_entity's pelt signal so keep using it looks better but all this clearly just nitpick because that doesn't make any difference
> > It's a soft bug we should fix one way or another before what the > > comment mentions really happens. > IIUC, We should be moving towards using full load resolution > for all the calculations. In that case, we need not worry about scaling load at > all. Maybe someone could provide context here. > > > I am actually confused that we have both se_weight() and > > scale_load_down(), and they do the same thing. > > > > Best regards, > > Dawei > > > > On Mon, Apr 1, 2024 at 3:36 AM Vishal Chourasia <vishalc@linux.ibm.com> wrote: > >> > >> On Thu, Mar 14, 2024 at 06:59:16PM -0700, Dawei Li wrote: > >>> Change se->load.weight to se_weight(se) in the calculation for the > >>> initial util_avg to avoid unnecessarily inflating the util_avg by 1024 > >>> times. > >>> > >>> The reason is that se->load.weight has the unit/scale as the scaled-up > >>> load, while cfs_rg->avg.load_avg has the unit/scale as the true task > >>> weight (as mapped directly from the task's nice/priority value). With > >>> CONFIG_32BIT, the scaled-up load is equal to the true task weight. With > >>> CONFIG_64BIT, the scaled-up load is 1024 times the true task weight. > >>> Thus, the current code may inflate the util_avg by 1024 times. The > >>> follow-up capping will not allow the util_avg value to go wild. But the > >>> calculation should have the correct logic. > >>> > >>> Signed-off-by: Dawei Li <daweilics@gmail.com> > >>> --- > >>> Changes in v2: > >>> - update the commit message > >>> --- > >>> kernel/sched/fair.c | 5 +++-- > >>> 1 file changed, 3 insertions(+), 2 deletions(-) > >>> > >>> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c > >>> index a19ea290b790..5f98f639bdb9 100644 > >>> --- a/kernel/sched/fair.c > >>> +++ b/kernel/sched/fair.c > >>> @@ -1031,7 +1031,8 @@ void init_entity_runnable_average(struct sched_entity *se) > >>> * With new tasks being created, their initial util_avgs are extrapolated > >>> * based on the cfs_rq's current util_avg: > >>> * > >>> - * util_avg = cfs_rq->util_avg / (cfs_rq->load_avg + 1) * se.load.weight > >>> + * util_avg = cfs_rq->avg.util_avg / (cfs_rq->avg.load_avg + 1) > >>> + * * se_weight(se) > >>> * > >>> * However, in many cases, the above util_avg does not give a desired > >>> * value. Moreover, the sum of the util_avgs may be divergent, such > >>> @@ -1078,7 +1079,7 @@ void post_init_entity_util_avg(struct task_struct *p) > >>> > >>> if (cap > 0) { > >>> if (cfs_rq->avg.util_avg != 0) { > >>> - sa->util_avg = cfs_rq->avg.util_avg * se->load.weight; > >>> + sa->util_avg = cfs_rq->avg.util_avg * se_weight(se); > >> Hi, > >> > >> The comment above the declaration of se_weight function says we should be > >> using full load resolution and get rid of this helper. > >> > >> Should we be adding new user of the helper? > >> > >> /* > >> * XXX we want to get rid of these helpers and use the full load resolution. > >> */ > >> static inline long se_weight(struct sched_entity *se) > >> { > >> return scale_load_down(se->load.weight); > >> } > >> > >> > >>> sa->util_avg /= (cfs_rq->avg.load_avg + 1); > >>> > >>> if (sa->util_avg > cap) > >>> -- > >>> 2.40.1 > >>> >
| |