Re: [PATCH 2/3 v2] proc: Export statistics for softirq to /proc

[kobayashi . kk]

2009/1/12 KOSAKI Motohiro <kosaki.motohiro@xxxxxxxxxxxxxx>:
>> >>> You can calcurate per_irq_sum here.
>> >>> Typically, # of possible cpu are very big.
>> >>>
>> >>> So, I don't like unnecessary twrice looping.
>> >>
>> >> I was about to send these patches to Linus, but it seems that this
>> >> optmisation hasn't been addressed?
>> >
>> > Thanks for your notice, Andrew.
>> > Of course, thanks for your advice, Kosaki-san.
>> > I didn't check carefully about this point...
>> >
>> > Now I am on a business trip.
>> > So, I'll post the fix next week.
>>
>> No problem.
>> I've made it three hour ago. I'll post soon.
>
> Done.
>
>
> ==
> Subject: [PATCH] proc: remove redundunt for_each_possible_cpu() loop
> Impact: cleanup
>
> Almost distro set NR_CPUS very large number and at that time for_each_possible_cpu() is
> a bit costly.
>
> Therefore, To remove unnecessary twice loop is better.
>
>
> Note. we can remove softirq's redundunt loop, but we can't remove irq's.
> because NR_SOFTIRQS ~= 10, NR_IRQS ~= 4000 (in x86 case).
>
>
> Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@xxxxxxxxxxxxxx>
> Cc: Keika Kobayashi <kobayashi.kk@xxxxxxxxxxxxxx>
> Cc: Hiroshi Shimamoto <h-shimamoto@xxxxxxxxxxxxx>
> Cc: Ingo Molnar <mingo@xxxxxxx>
> Cc: Eric Dumazet <dada1@xxxxxxxxxxxxx>
> Cc: Alexey Dobriyan <adobriyan@xxxxxxxxx>
> Cc: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
> ---
>  fs/proc/stat.c |   16 ++++++++--------
>  1 files changed, 8 insertions(+), 8 deletions(-)
>
> diff --git a/fs/proc/stat.c b/fs/proc/stat.c
> index f95d73a..cb839e8 100644
> --- a/fs/proc/stat.c
> +++ b/fs/proc/stat.c
> @@ -27,6 +27,7 @@ static int show_stat(struct seq_file *p, void *v)
>        cputime64_t guest;
>        u64 sum = 0;
>        u64 sum_softirq = 0;
> +       unsigned int per_softirq_sums[NR_SOFTIRQS] = {0};
>        struct timespec boottime;
>        unsigned int per_irq_sum;
>
> @@ -51,8 +52,12 @@ static int show_stat(struct seq_file *p, void *v)
>                }
>                sum += arch_irq_stat_cpu(i);
>
> -               for (j = 0; j < NR_SOFTIRQS; j++)
> -                       sum_softirq += kstat_softirqs_cpu(j, i);
> +               for (j = 0; j < NR_SOFTIRQS; j++) {
> +                       unsigned int softirq_stat = kstat_softirqs_cpu(j, i);
> +
> +                       per_softirq_sums[j] += softirq_stat;
> +                       sum_softirq += softirq_stat;
> +               }
>
>        }
>        sum += arch_irq_stat();
> @@ -118,12 +123,7 @@ static int show_stat(struct seq_file *p, void *v)
>        seq_printf(p, "softirq %llu", (unsigned long long)sum_softirq);
>
>        for (i = 0; i < NR_SOFTIRQS; i++) {
> -               per_irq_sum = 0;
> -
> -               for_each_possible_cpu(j)
> -                       per_irq_sum += kstat_softirqs_cpu(i, j);
> -
> -               seq_printf(p, " %u", per_irq_sum);
> +               seq_printf(p, " %u", per_softirq_sums[i]);
>        }
>        seq_printf(p, "\n");
>

Thank you very much, Kosaki-san!
It looks good to me.

// Keika Kobayashi
--
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/