Re: [RFC PATCH 4/5] sched/fair: Rework inter-NUMA newidle balancing

[K Prateek Nayak]

On 4/10/2025 3:44 PM, Peter Zijlstra wrote:

[..snip..]

> > +				continue;
> > +
> > +			scoped_guard (rq_lock, overloaded_rq) {
> > +				update_rq_clock(overloaded_rq);
> > +
> > +				if (!has_pushable_tasks(overloaded_rq))
> > +					break;
>
> You can skip the clock update if there aren't any tasks to grab.

Ack.

> > +
> > +				env.src_cpu = overloaded_cpu;
> > +				env.src_rq = overloaded_rq;
> > +
> > +				p = detach_one_task(&env);
>
> Yep, detach_one_task() uses can_migrate_task() which checks
> task_on_cpu(), so that's all good :-)
>
> > +			}
> > +
> > +			if (!p)
> > +				continue;
> > +
> > +			attach_one_task(this_rq, p);
> > +			return 1;
> > +		}
> > +
> > +		cpumask_andnot(cpus, cpus, sched_domain_span(cpu_llc));
> > +	}

--
Thanks and Regards,
Prateek