Re: [PATCH V1 20/23] perf stat: Add requires_cpu flag for uncore

[Ian Rogers]

On Thu, May 5, 2022 at 9:58 AM Adrian Hunter <adrian.hunter@xxxxxxxxx> wrote:
>
> Uncore events require a CPU i.e. it cannot be -1.
>
> The evsel system_wide flag is intended for events that should be on every
> CPU, which does not make sense for uncore events because uncore events do
> not map one-to-one with CPUs.
>
> These 2 requirements are not exactly the same, so introduce a new flag
> 'requires_cpu' the uncore case.

nit: typo missing 'for' the uncore case.

> Signed-off-by: Adrian Hunter <adrian.hunter@xxxxxxxxx>
> ---
>  tools/lib/perf/evlist.c                 | 4 +++-
>  tools/lib/perf/include/internal/evsel.h | 1 +
>  tools/perf/builtin-stat.c               | 5 +----
>  tools/perf/util/evsel.c                 | 1 +
>  tools/perf/util/parse-events.c          | 2 +-
>  5 files changed, 7 insertions(+), 6 deletions(-)
>
> diff --git a/tools/lib/perf/evlist.c b/tools/lib/perf/evlist.c
> index 8a82b4b94b99..3bf77f9617b6 100644
> --- a/tools/lib/perf/evlist.c
> +++ b/tools/lib/perf/evlist.c
> @@ -43,7 +43,9 @@ static void __perf_evlist__propagate_maps(struct perf_evlist *evlist,
>         if (!evsel->own_cpus || evlist->has_user_cpus) {
>                 perf_cpu_map__put(evsel->cpus);
>                 evsel->cpus = perf_cpu_map__get(evlist->user_requested_cpus);
> -       } else if (!evsel->system_wide && perf_cpu_map__empty(evlist->user_requested_cpus)) {
> +       } else if (!evsel->system_wide &&
> +                  !evsel->requires_cpu &&
> +                  perf_cpu_map__empty(evlist->user_requested_cpus)) {
>                 perf_cpu_map__put(evsel->cpus);
>                 evsel->cpus = perf_cpu_map__get(evlist->user_requested_cpus);
>         } else if (evsel->cpus != evsel->own_cpus) {
> diff --git a/tools/lib/perf/include/internal/evsel.h b/tools/lib/perf/include/internal/evsel.h
> index cfc9ebd7968e..77fbb8b97e5c 100644
> --- a/tools/lib/perf/include/internal/evsel.h
> +++ b/tools/lib/perf/include/internal/evsel.h
> @@ -50,6 +50,7 @@ struct perf_evsel {
>         /* parse modifier helper */
>         int                      nr_members;

>From the commit message, could we add the comment:

/* True for events that should be on every CPU. */

However, shouldn't this be captured by the CPU map? Perhaps we can
have an invariant check that they agree.

>         bool                     system_wide;

/* True for events that cannot be the any CPU (-1), for example,
uncore events. */

> +       bool                     requires_cpu;
>         int                      idx;
>  };
>
> diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
> index 1b96636df01e..c049533f74e4 100644
> --- a/tools/perf/builtin-stat.c
> +++ b/tools/perf/builtin-stat.c
> @@ -385,9 +385,6 @@ static int read_counter_cpu(struct evsel *counter, struct timespec *rs, int cpu_
>         if (!counter->supported)
>                 return -ENOENT;
>
> -       if (counter->core.system_wide)
> -               nthreads = 1;
> -
>         for (thread = 0; thread < nthreads; thread++) {
>                 struct perf_counts_values *count;
>
> @@ -2264,7 +2261,7 @@ static void setup_system_wide(int forks)
>                 struct evsel *counter;
>
>                 evlist__for_each_entry(evsel_list, counter) {
> -                       if (!counter->core.system_wide &&
> +                       if (!counter->core.requires_cpu &&
>                             strcmp(counter->name, "duration_time")) {

Off-topic of the patch: this test is a tool event check, it'd be better to do:

if (!counter->core.requires_cpu && counter->tool_event != PERF_TOOL_NONE)

Otherwise changes like:
https://lore.kernel.org/linux-perf-users/20220420102354.468173-1-florian.fischer@muhq.space/

Will break the test. I'll make a note to do a round of clean up.

Thanks,
Ian


>                                 return;
>                         }
> diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c
> index d38722560e80..12346293706b 100644
> --- a/tools/perf/util/evsel.c
> +++ b/tools/perf/util/evsel.c
> @@ -382,6 +382,7 @@ struct evsel *evsel__clone(struct evsel *orig)
>         evsel->core.threads = perf_thread_map__get(orig->core.threads);
>         evsel->core.nr_members = orig->core.nr_members;
>         evsel->core.system_wide = orig->core.system_wide;
> +       evsel->core.requires_cpu = orig->core.requires_cpu;
>
>         if (orig->name) {
>                 evsel->name = strdup(orig->name);
> diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-events.c
> index 937f6c9434a2..5227174099b5 100644
> --- a/tools/perf/util/parse-events.c
> +++ b/tools/perf/util/parse-events.c
> @@ -365,7 +365,7 @@ __add_event(struct list_head *list, int *idx,
>         (*idx)++;
>         evsel->core.cpus = cpus;
>         evsel->core.own_cpus = perf_cpu_map__get(cpus);
> -       evsel->core.system_wide = pmu ? pmu->is_uncore : false;
> +       evsel->core.requires_cpu = pmu ? pmu->is_uncore : false;
>         evsel->auto_merge_stats = auto_merge_stats;
>
>         if (name)
> --
> 2.25.1
>