Re: [PATCH 1/4] Change cpustat fields to an array.

[Glauber Costa]

Ok, I missed the other comments in this patch.

Here it goes:

On 11/16/2011 09:25 AM, Glauber Costa wrote:
> > > for_each_possible_cpu(i) {
> > > - user = cputime64_add(user, kstat_cpu(i).cpustat.user);
> > > - nice = cputime64_add(nice, kstat_cpu(i).cpustat.nice);
> > > - system = cputime64_add(system, kstat_cpu(i).cpustat.system);
> > > - idle = cputime64_add(idle, get_idle_time(i));
> > > - iowait = cputime64_add(iowait, get_iowait_time(i));
> > > - irq = cputime64_add(irq, kstat_cpu(i).cpustat.irq);
> > > - softirq = cputime64_add(softirq, kstat_cpu(i).cpustat.softirq);
> > > - steal = cputime64_add(steal, kstat_cpu(i).cpustat.steal);
> > > - guest = cputime64_add(guest, kstat_cpu(i).cpustat.guest);
> > > - guest_nice = cputime64_add(guest_nice,
> > > - kstat_cpu(i).cpustat.guest_nice);
> > > - sum += kstat_cpu_irqs_sum(i);
> > > - sum += arch_irq_stat_cpu(i);
> > > + user += kstat_cpu(i).cpustat[USER];
> >
> > Half the time cputime64_add is preserved, half the time this patch
> > converts it to a naked '+='. Admittedly no one seems to usefully define
> > cputime64_add but why the conversion / inconsistency?
>
> Because at least in this patchset of mine, cputime conversion is not a
> goal, but a side effect. I wanted cpustat to be an array of u64, so I
> converted users. There are some places in the code that still uses
> cputime64 for other variables, and those I kept untouched.
>
> That's to make the patch focused. The other variables can be easily
> converted if we see value on it.
Outside of sched.c I did miss some. Those I can chase and convert.

> > > +enum cpu_usage_stat {
> > > + USER,
> > > + NICE,
> > > + SYSTEM,
> > > + SOFTIRQ,
> > > + IRQ,
> > > + IDLE,
> > > + IOWAIT,
> > > + STEAL,
> > > + GUEST,
> > > + GUEST_NICE,
> > > + NR_STATS,
> > > };
> >
> > I suspect we want a more descriptive prefix here, e.g. CPUTIME_USER

Ok.

> > > /* Add user time to cpustat. */
> > > tmp = cputime_to_cputime64(cputime);
> > > +
> > > if (TASK_NICE(p)> 0)
> >
> > We now that these are actually fields this could be:
> > field = TASK_NICE(p) > 0 ? CPUTIME_NICE : CPUTIME_USER;

Yes, absolutely.

> > > @@ -3925,9 +3926,9 @@ static void account_guest_time(struct
> > > task_struct *p, cputime_t cputime,
> > > */
> > > static inline
> > > void __account_system_time(struct task_struct *p, cputime_t cputime,
> > > - cputime_t cputime_scaled, cputime64_t *target_cputime64)
> > > + cputime_t cputime_scaled, u64 *target_cputime64)
> >
> > Having cpustat be an array means we can drop the pointer here and pass
> > the id.

And that's precisely what I do on a later patch, with account_field. I 
can move that code here if you prefer.

--
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/