Re: [RFC PATCH 01/14] sched: track the runnable average on a per-taskentitiy basis

[Paul Turner]

On Wed, Feb 15, 2012 at 3:37 PM, Peter Zijlstra <a.p.zijlstra@xxxxxxxxx> wrote:
> On Wed, 2012-02-01 at 17:38 -0800, Paul Turner wrote:
>> +       /*
>> +        * Use 1024us as the unit of measurement since it's a reasonable
>> +        * approximation of 1ms and fast to compute.
>> +        */
>> +       delta >>= 10;
>
> ns -> us ?, text talks about ms, slightly confusing

Yes, comment was incorrect; I'd noticed this also on a re-read -- it's fixed.

>
>> +       if (!delta)
>> +               return 0;
>> +       sa->last_runnable_update = now;
>> +
>> +       delta_w = sa->runnable_avg_period % 1024;
>
> so delta_w is the chunk of this p we already accounted.
>
>> +       if (delta + delta_w >= 1024) {
>
> if delta pushes us over 1024*1024 ns (~1ms) we roll a window.
>
>> +               /* period roll-over */
>> +               decayed = 1;
>> +
>> +               delta_w = 1024 - delta_w;
>
> The distance we need to reach the next window.
>
>> +               BUG_ON(delta_w > delta);
>
> somehow reading this code took forever, this suggests clarification,
> either through better variable names or more comments. Could also mean
> I'm a moron and should get more sleep or so :-)

No... this bit is definitely fiddly, I definitely got it wrong more
than once writing it down.  And then got it again wrong later when
"optimizing" :-).

It deserves more comments, I will add.
--
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/