Re: [RFC] sched: Add a new lockless wake-from-idle implementation

[Frederic Weisbecker]

2014-02-13 2:40 GMT+01:00 Andy Lutomirski <luto@xxxxxxxxxxxxxx>:
> This is a strawman proposal to simplify the idle implementation, eliminate
> a race

Please describe the race in question.

>
> Benefits over current code:
>  - ttwu_queue_remote doesn't use an IPI unless needed
>  - The diffstat should speak for itself :)

Actually referring to the diffstat alone sounds dangerous here. Sure
this simplifies the code, I'm all for negative diffstat, and probably
it avoids a few IPIs, but this comes at the cost of added non-cheap
atomic_xchg() calls in some critical fastpath like resched_task()
path, the timer enqueue path and the inner idle loop.

So it's not like this all comes for free. I'm not saying we don't want
it but at least some serious measurements is needed.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/