Messages in this thread | | | From | Ian Rogers <> | Date | Wed, 24 Jan 2024 10:58:50 -0800 | Subject | Re: [RFC PATCH v3 11/18] perf stat: Add function to handle special events in hardware-grouping |
| |
On Tue, Dec 12, 2023 at 3:03 PM <weilin.wang@intel.com> wrote: > > From: Weilin Wang <weilin.wang@intel.com> > > There are some special events like topdown events and TSC that are not > decribed in pmu-event JSON files. Add support to handle this type of
nit: described.
> events. This should be considered as a temporary solution because including > these events in JSON files would be a better solution.
Agreed on this being temporary. We have a similar hack with arch_evsel__must_be_in_group, could that be reused here? https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/arch/x86/util/evsel.c?h=perf-tools-next#n41 It doesn't cover TSC but parse-events makes TSC a separate group during parsing as the PMU (msr) differs from the PMU of the events (typically cpu). Fixing the metrics and events to be json is very do-able. I see on Icelake, for example, TOPDOWN.BACKEND_BOUND_SLOTS is available while frontend isn't.
As an aside, we may want to treat msr as a software pmu: https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/util/pmu.c?h=perf-tools-next#n1761 which may be a metric regression.
Thanks, Ian
> Signed-off-by: Weilin Wang <weilin.wang@intel.com> > --- > tools/perf/util/metricgroup.c | 25 +++++++++++++++++++++++-- > tools/perf/util/metricgroup.h | 15 +++++++++++++++ > 2 files changed, 38 insertions(+), 2 deletions(-) > > diff --git a/tools/perf/util/metricgroup.c b/tools/perf/util/metricgroup.c > index 5d9ed22963be..03383f820e06 100644 > --- a/tools/perf/util/metricgroup.c > +++ b/tools/perf/util/metricgroup.c > @@ -2023,6 +2023,15 @@ static int create_grouping(struct list_head *pmu_info_list, > return ret; > }; > > +static bool is_special_event(const char *id) > +{ > + for (int i = 0; i < SPECIAL_EVENT_MAX; i++) { > + if (!strncmp(id, special_event_names[i], strlen(special_event_names[i]))) > + return true; > + } > + return false; > +} > + > /** > * hw_aware_build_grouping - Build event groupings by reading counter > * requirement of the events and counter available on the system from > @@ -2047,7 +2056,17 @@ static int hw_aware_build_grouping(struct expr_parse_ctx *ctx __maybe_unused, > hashmap__for_each_entry(ctx->ids, cur, bkt) { > const char *id = cur->pkey; > > - pr_debug("found event %s\n", id); > + if (is_special_event(id)) { > + struct metricgroup__event_info *event; > + > + event = event_info__new(id, "default_core", "0", > + /*free_counter=*/true); > + if (!event) > + goto err_out; > + > + list_add(&event->nd, &event_info_list); > + continue; > + } > ret = get_metricgroup_events(id, etable, &event_info_list); > if (ret) > goto err_out; > @@ -2519,8 +2538,10 @@ int metricgroup__parse_groups(struct evlist *perf_evlist, > ret = hw_aware_parse_groups(perf_evlist, pmu, str, > metric_no_threshold, user_requested_cpu_list, system_wide, > /*fake_pmu=*/NULL, metric_events, table); > - if (!ret) > + if (!ret) { > + pr_info("Hardware aware grouping completed\n"); > return 0; > + } > } > > return parse_groups(perf_evlist, pmu, str, metric_no_group, metric_no_merge, > diff --git a/tools/perf/util/metricgroup.h b/tools/perf/util/metricgroup.h > index 51596e4b4341..37fdf8c5ba77 100644 > --- a/tools/perf/util/metricgroup.h > +++ b/tools/perf/util/metricgroup.h > @@ -79,6 +79,21 @@ struct metric_expr { > #define TRANSFER_FIRST_BYTE(pos) (7 - pos) > #define TRANSFER_SEC_BYTE(pos) (23 - pos) > > +/* Special events that are not described in pmu-event JSON files. > + * topdown-* and TSC use dedicated registers, set as free > + * counter for grouping purpose > + */ > +enum special_events { > + TOPDOWN = 0, > + TSC = 1, > + SPECIAL_EVENT_MAX, > +}; > + > +static const char *const special_event_names[SPECIAL_EVENT_MAX] = { > + "topdown-", > + "TSC", > +}; > + > /** > * An event used in a metric. This info is for metric grouping. > */ > -- > 2.39.3 >
| |