Re: [PATCH v5 6/8] perf report: Add --latency flag

[Dmitry Vyukov]

On Fri, 7 Feb 2025 at 04:53, Namhyung Kim <namhyung@xxxxxxxxxx> wrote:
>
> On Wed, Feb 05, 2025 at 05:27:45PM +0100, Dmitry Vyukov wrote:
> > Add record/report --latency flag that allows to capture and show
> > latency-centric profiles rather than the default CPU-consumption-centric
> > profiles. For latency profiles record captures context switch events,
> > and report shows Latency as the first column.
> >
> > Signed-off-by: Dmitry Vyukov <dvyukov@xxxxxxxxxx>
> > Cc: Namhyung Kim <namhyung@xxxxxxxxxx>
> > Cc: Arnaldo Carvalho de Melo <acme@xxxxxxxxxx>
> > Cc: Ian Rogers <irogers@xxxxxxxxxx>
> > Cc: linux-perf-users@xxxxxxxxxxxxxxx
> > Cc: linux-kernel@xxxxxxxxxxxxxxx
> >
> > ---
> [SNIP]
> > +void perf_hpp__cancel_latency(void)
> > +{
> > +     struct perf_hpp_fmt *fmt, *lat, *acc, *tmp;
> > +
> > +     if (is_strict_order(field_order) || is_strict_order(sort_order))
> > +             return;
>
> This also needs to be changed since you call setup_overhead even if you
> have a strict sort_order.
>
> For example, it's strange these two are different.
>
> With default sort order:
>
>   $ perf record --latency -- ls /
>
>   $ perf report --stdio
>   ...
>   # Overhead  Command  Shared Object         Symbol
>   # ........  .......  ....................  ........................
>   #
>       64.50%  ls       ld-linux-x86-64.so.2  [.] do_lookup_x
>       33.41%  ls       [kernel.kallsyms]     [k] chacha_block_generic
>        2.00%  perf-ex  [kernel.kallsyms]     [k] put_ctx
>        0.09%  perf-ex  [kernel.kallsyms]     [k] native_write_msr
>
> Same sort order, but explicitly:
>
>   $ perf report --stdio -s comm,dso,sym
>   ...
>   # Overhead   Latency  Command  Shared Object         Symbol
>   # ........  ........  .......  ....................  ........................
>   #
>       64.50%    64.50%  ls       ld-linux-x86-64.so.2  [.] do_lookup_x
>       33.41%    33.41%  ls       [kernel.kallsyms]     [k] chacha_block_generic
>        2.00%     2.00%  perf-ex  [kernel.kallsyms]     [k] put_ctx
>        0.09%     0.09%  perf-ex  [kernel.kallsyms]     [k] native_write_msr
>
> Maybe you want to cancel the latency field even if sort key is given
> (unless it has 'latency').
>
> Something like this?

Makes sense! Done in v6.

> ---8<---
> @@ -714,7 +715,9 @@ void perf_hpp__cancel_latency(void)
>  {
>         struct perf_hpp_fmt *fmt, *lat, *acc, *tmp;
>
> -       if (is_strict_order(field_order) || is_strict_order(sort_order))
> +       if (is_strict_order(field_order))
> +               return;
> +       if (is_strict_order(sort_order) && strstr(sort_order, "latency"))
>                 return;
>
>         lat = &perf_hpp__format[PERF_HPP__LATENCY];
> ---8<---
>
> Thanks,
> Namhyung
>
> > +
> > +     lat = &perf_hpp__format[PERF_HPP__LATENCY];
> > +     acc = &perf_hpp__format[PERF_HPP__LATENCY_ACC];
> > +
> > +     perf_hpp_list__for_each_format_safe(&perf_hpp_list, fmt, tmp) {
> > +             if (fmt_equal(lat, fmt) || fmt_equal(acc, fmt))
> > +                     perf_hpp__column_unregister(fmt);
> > +     }
> > +}
> > +
> >  void perf_hpp__setup_output_field(struct perf_hpp_list *list)
> >  {
> >       struct perf_hpp_fmt *fmt;