On Thu, Mar 14, 2024 at 9:47 PM Alexei Starovoitov
<alexei.starovoitov@xxxxxxxxx> wrote:
>
> On Thu, Mar 14, 2024 at 8:56 PM Andrii Nakryiko
> <andrii.nakryiko@xxxxxxxxx> wrote:
> >
> > On Thu, Mar 14, 2024 at 7:33 PM Alexei Starovoitov
> > <alexei.starovoitov@xxxxxxxxx> wrote:
> > >
> > > On Tue, Mar 12, 2024 at 3:23 PM Andrii Nakryiko <andrii@xxxxxxxxxx> wrote:
> > > >
> > > >
> > > > static void trigger_base_measure(struct bench_res *res)
> > > > {
> > > > - res->hits = atomic_swap(&base_hits.value, 0);
> > > > + res->hits = sum_counters(base_hits);
> > > > }
> > > >
> > > > static void *trigger_producer(void *input)
> > > > @@ -42,7 +72,7 @@ static void *trigger_producer(void *input)
> > > >
> > > > static void trigger_measure(struct bench_res *res)
> > > > {
> > > > - res->hits = atomic_swap(&ctx.skel->bss->hits, 0);
> > > > + res->hits = sum_counters(ctx.skel->bss->hits);
> > > > }
> > >
> > > It was zeroing counters before.
> > > Do you need to zero them now?
> > >
> >
> > sum_counters() does zero them out, it does the same atomic_swap(0).
>
> I simply missed the swap in there.
> Could you rename it to sum_and_zero_counters() or
> something, so it's obvious?
> A helper that says 'sum_counters' isn't expected to have such side effects.
Yep, sure. I should have called it sum_and_reset_counters(). BTW, I
have since added another set of benchmarks to measure fentry/fexit and
kprobe/kretprobe throughput, but this time minimizing the amount of
syscalls. I'll bundle both together and send them as v2. This new
benchmark would definitely be bottlenecked on counting, so it depends
on these distributed counters for better counting.
