perf stat fixes and improvements
Ian Rogers
reviewerv4 cover perf stat fixes and improvements
A collection of fixes aiming to stabilize and make more reasonable measurements/metrics such as memory bandwidth.
Tool events are changed from getting a PMU cpu map of all online CPUs to either CPU 0 or all online CPUs. This avoids iterating over useless CPUs for events in particular duration_time. Fix a bug where duration_time didn't correctly use the previous raw counts and would skip values in interval mode.
Change how json metrics handle tool events. Use the counter value rather than using shared state with perf stat. A later patch changes it so that tool events are read last, so that if reading say memory bandwidth counters you don't divide by an earlier read time and exceed the theoretical maximum memory bandwidth.
Do some clean up around the shared state in stat-shadow that's no longer used.
Change how affinities work with evlist__for_each_cpu. Move the affinity code into the iterator to simplify setting it up. Detect when affinities will and won't be profitable, for example a tool event and a regular perf event (or read group) may face less delay from a single IPI for the event read than from a call to sched_setaffinity. Add a
--no-affinity flag to perf stat to allow affinities to be disabled.
v4: Rebase. Add patch to reduce scope of walltime_nsec_stats now that
the legacy metric code is no more. Minor tweak to the ru_stats clean up.
v3: Add affinity clean ups and read tool events last.
https://lore.kernel.org/lkml/[email protected]/
v2: Fixed an aggregation index issue:
https://lore.kernel.org/lkml/[email protected]/
v1:
https://lore.kernel.org/lkml/[email protected]/
Ian Rogers (10):
libperf cpumap: Reduce allocations and sorting in intersect perf pmu: perf_cpu_map__new_int to avoid parsing a string perf tool_pmu: Use old_count when computing count values for time
events
perf stat-shadow: Read tool events directly
perf stat: Reduce scope of ru_stats
perf stat: Reduce scope of walltime_nsecs_stats
perf tool_pmu: More accurately set the cpus for tool events perf evlist: Reduce affinity use and move into iterator, fix no
affinity
perf stat: Read tool events last
perf stat: Add no-affinity flag
tools/lib/perf/cpumap.c | 29 ++-- tools/perf/Documentation/perf-stat.txt | 4 + tools/perf/builtin-stat.c | 210 +++++++++++++++---------- tools/perf/tests/parse-metric.c | 2 - tools/perf/tests/pmu-events.c | 2 - tools/perf/util/config.c | 3 +- tools/perf/util/drm_pmu.c | 2 +- tools/perf/util/evlist.c | 156 +++++++++++------- tools/perf/util/evlist.h | 27 +++- tools/perf/util/hwmon_pmu.c | 2 +- tools/perf/util/parse-events.c | 9 +- tools/perf/util/pmu.c | 12 ++ tools/perf/util/pmu.h | 1 + tools/perf/util/stat-shadow.c | 152 ++++++++---------- tools/perf/util/stat.h | 23 --- tools/perf/util/tool_pmu.c | 78 ++++++--- tools/perf/util/tool_pmu.h | 1 + 17 files changed, 407 insertions(+), 306 deletions(-)
Signature · 1 line
2.51.2.1041.gc1ab5b90ca-goog
Ian Rogers
reviewerPATCH v4 1/10 libperf cpumap: Reduce allocations and sorting in intersect
On hybrid platforms the CPU maps are often disjoint. Rather than copy CPUs and trim, compute the number of common CPUs, if none early exit, otherwise copy in an sorted order. This avoids memory allocation in the disjoint case and avoids a second malloc and useless sort in the previous trim cases.
--- tools/lib/perf/cpumap.c | 29 +++++++++++++++++++---------- 1 file changed, 19 insertions(+), 10 deletions(-)
tools/lib/perf/cpumap.c +19 -10
@@ -453,21 +453,33 @@ int perf_cpu_map__merge(struct perf_cpu_map **orig, struct perf_cpu_map *other)struct perf_cpu_map *perf_cpu_map__intersect (struct perf_cpu_map *orig,struct perf_cpu_map *other) { -struct perf_cpu *tmp_cpus; -int tmp_len;int i, j, k; -struct perf_cpu_map *merged =NULL ; +struct perf_cpu_map *merged;if (perf_cpu_map__is_subset (other, orig))return perf_cpu_map__get (orig);if (perf_cpu_map__is_subset (orig, other))return perf_cpu_map__get (other); - tmp_len =max (__perf_cpu_map__nr (orig),__perf_cpu_map__nr (other)); - tmp_cpus =malloc (tmp_len *sizeof (struct perf_cpu )); -if (!tmp_cpus) + i = j = k =0 ; +while (i <__perf_cpu_map__nr (orig) && j <__perf_cpu_map__nr (other)) { +if (__perf_cpu_map__cpu (orig, i).cpu <__perf_cpu_map__cpu (other, j).cpu) + i++; +else if (__perf_cpu_map__cpu (orig, i).cpu >__perf_cpu_map__cpu (other, j).cpu) + j++; +else {/* CPUs match. */ + i++; + j++; + k++; + } + } +if (k ==0 )/* Maps are completely disjoint. */ return NULL ; + merged =perf_cpu_map__alloc (k); +if (!merged) +return NULL ; +/* Entries are added to merged in sorted order, so no need to sort again. */ i = j = k =0 ;while (i <__perf_cpu_map__nr (orig) && j <__perf_cpu_map__nr (other)) {if (__perf_cpu_map__cpu (orig, i).cpu <__perf_cpu_map__cpu (other, j).cpu)
@@ -476,11 +488,8 @@ struct perf_cpu_map *perf_cpu_map__intersect(struct perf_cpu_map *orig, j++; else { j++; - tmp_cpus[k++] =__perf_cpu_map__cpu (orig, i++); +RC_CHK_ACCESS (merged)->map[k++] =__perf_cpu_map__cpu (orig, i++); } } -if (k) - merged =cpu_map__trim_new (k, tmp_cpus); -free (tmp_cpus);return merged; }
Signature · 1 line
2.51.2.1041.gc1ab5b90ca-goog
Ian Rogers
reviewerPATCH v4 2/10 perf pmu: perf_cpu_map__new_int to avoid parsing a string
In reply to Ian RogersPrefer perf_cpu_map__new_int(0) to perf_cpu_map__new("0") as it avoids strings parsing.
--- tools/perf/util/drm_pmu.c | 2 +- tools/perf/util/hwmon_pmu.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-)
tools/perf/util/drm_pmu.c +1 -1
@@ -119,7 +119,7 @@ static struct drm_pmu *add_drm_pmu(struct list_head *pmus, char *line, size_t lireturn NULL ; } - drm->pmu.cpus =perf_cpu_map__new ("0" ); + drm->pmu.cpus =perf_cpu_map__new_int (0 );if (!drm->pmu.cpus) {perf_pmu__delete (&drm->pmu);return NULL ;
tools/perf/util/hwmon_pmu.c +1 -1
@@ -376,7 +376,7 @@ struct perf_pmu *hwmon_pmu__new(struct list_head *pmus, const char *hwmon_dir,perf_pmu__delete (&hwm->pmu);return NULL ; } - hwm->pmu.cpus =perf_cpu_map__new ("0" ); + hwm->pmu.cpus =perf_cpu_map__new_int (0 );if (!hwm->pmu.cpus) {perf_pmu__delete (&hwm->pmu);return NULL ;
Signature · 1 line
2.51.2.1041.gc1ab5b90ca-goog
Ian Rogers
reviewerPATCH v4 3/10 perf tool_pmu: Use old_count when computing count values for time events
In reply to Ian RogersWhen running in interval mode every third count of a time event isn't showing properly:
```
$ perf stat -e duration_time -a -I 1000
1.001082862 1,002,290,425 duration_time
2.004264262 1,003,183,516 duration_time
3.007381401 <not counted> duration_time
4.011160141 1,003,705,631 duration_time
5.014515385 1,003,290,110 duration_time
6.018539680 <not counted> duration_time
7.022065321 1,003,591,720 duration_time
```The regression came in with a different fix, found through bisection, commit 68cb1567439f ("perf tool_pmu: Fix aggregation on duration_time"). The issue is caused by the enabled and running time of the event matching the old_count's and creating a delta of 0, which is indicative of an error.
--- tools/perf/builtin-stat.c | 16 +++++++++-- tools/perf/util/tool_pmu.c | 59 +++++++++++++++++++++----------------- 2 files changed, 46 insertions(+), 29 deletions(-)
tools/perf/builtin-stat.c +13 -3
@@ -283,17 +283,27 @@ static int read_single_counter(struct evsel *counter, int cpu_map_idx, int threaif (err && cpu_map_idx ==0 && (evsel__tool_event (counter) ==TOOL_PMU__EVENT_USER_TIME ||evsel__tool_event (counter) ==TOOL_PMU__EVENT_SYSTEM_TIME )) { -u64 val, *start_time;struct perf_counts_values *count =perf_counts (counter->counts, cpu_map_idx, thread); +struct perf_counts_values *old_count =NULL ; +u64 val; + +if (counter->prev_raw_counts) + old_count =perf_counts (counter->prev_raw_counts, cpu_map_idx, thread); - start_time =xyarray__entry (counter->start_times, cpu_map_idx, thread);if (evsel__tool_event (counter) ==TOOL_PMU__EVENT_USER_TIME ) val = ru_stats.ru_utime_usec_stat.mean;else val = ru_stats.ru_stime_usec_stat.mean; - count->ena = count->run = *start_time + val; + count->val = val; +if (old_count) { + count->run = old_count->run +1 ; + count->ena = old_count->ena +1 ; + }else { + count->run++; + count->ena++; + }return 0 ; }return err;
tools/perf/util/tool_pmu.c +33 -26
103 lines not on this page · raw message
Signature · 1 line
2.51.2.1041.gc1ab5b90ca-goog
Ian Rogers
reviewerPATCH v4 4/10 perf stat-shadow: Read tool events directly
In reply to Ian RogersWhen reading time values for metrics don't use the globals updated in builtin-stat, just read the events as regular events. The only exception is for time events where nanoseconds need converting to seconds as metrics assume time metrics are in seconds.
--- tools/perf/util/stat-shadow.c | 149 +++++++++++++++------------------- 1 file changed, 66 insertions(+), 83 deletions(-)
tools/perf/util/stat-shadow.c +66 -83
@@ -26,7 +26,32 @@ void perf_stat__reset_shadow_stats(void)memset (&ru_stats,0 ,sizeof (ru_stats)); } -static int prepare_metric (const struct metric_expr *mexp, +static bool tool_pmu__is_time_event (const struct perf_stat_config *config, +const struct evsel *evsel,int *tool_aggr_idx) +{ +enum tool_pmu_event event =evsel__tool_event (evsel); +int aggr_idx; + +if (event !=TOOL_PMU__EVENT_DURATION_TIME && + event !=TOOL_PMU__EVENT_USER_TIME && + event !=TOOL_PMU__EVENT_SYSTEM_TIME ) +return false; + +if (config) { +cpu_aggr_map__for_each_idx (aggr_idx, config->aggr_map) { +if (config->aggr_map->map[aggr_idx].cpu.cpu ==0 ) { + *tool_aggr_idx = aggr_idx; +return true; + } + } +pr_debug ("Unexpected CPU0 missing in aggregation for tool event.\n" ); + } + *tool_aggr_idx =0 ;/* Assume the first aggregation index works. */ +return true; +} + +static int prepare_metric (struct perf_stat_config *config, +const struct metric_expr *mexp,const struct evsel *evsel,struct expr_parse_ctx *pctx,int aggr_idx)
@@ -36,93 +61,51 @@ static int prepare_metric(const struct metric_expr *mexp,int i;for (i =0 ; metric_events[i]; i++) { +int source_count =0 , tool_aggr_idx; +bool is_tool_time = +tool_pmu__is_time_event (config, metric_events[i], &tool_aggr_idx); +struct perf_stat_evsel *ps = metric_events[i]->stats; +struct perf_stat_aggr *aggr;char *n;double val; -int source_count =0 ; -if (evsel__is_tool (metric_events[i])) { -struct stats *stats; -double scale; - -switch (evsel__tool_event (metric_events[i])) { -case TOOL_PMU__EVENT_DURATION_TIME : - stats = &walltime_nsecs_stats; - scale =1e-9 ; -break ; -case TOOL_PMU__EVENT_USER_TIME : - stats = &ru_stats.ru_utime_usec_stat; - scale =1e-6 ; +/* +* If there are multiple uncore PMUs and we're not reading the +* leader's stats, determine the stats for the appropriate +* uncore PMU. +*/ +if (evsel && evsel->metric_leader && + evsel->pmu != evsel->metric_leader->pmu && + mexp->metric_events[i]->pmu == evsel->metric_leader->pmu) { +struct evsel *pos; + +evlist__for_each_entry (evsel->evlist, pos) { +if (pos->pmu != evsel->pmu) +continue ; +if (pos->metric_leader != mexp->metric_events[i]) +continue ; + ps = pos->stats; + source_count =1 ;break ; -case TOOL_PMU__EVENT_SYSTEM_TIME : - stats = &ru_stats.ru_stime_usec_stat; - scale =1e-6 ; -break ; -case TOOL_PMU__EVENT_NONE : -pr_err ("Invalid tool event 'none'" ); -abort (); -case TOOL_PMU__EVENT_MAX : -pr_err ("Invalid tool event 'max'" ); -abort (); -case TOOL_PMU__EVENT_HAS_PMEM : -case TOOL_PMU__EVENT_NUM_CORES : -case TOOL_PMU__EVENT_NUM_CPUS : -case TOOL_PMU__EVENT_NUM_CPUS_ONLINE : -case TOOL_PMU__EVENT_NUM_DIES : -case TOOL_PMU__EVENT_NUM_PACKAGES : -case TOOL_PMU__EVENT_SLOTS : -case TOOL_PMU__EVENT_SMT_ON : -case TOOL_PMU__EVENT_SYSTEM_TSC_FREQ : -case TOOL_PMU__EVENT_CORE_WIDE : -case TOOL_PMU__EVENT_TARGET_CPU : -default : -pr_err ("Unexpected tool event '%s'" ,evsel__name (metric_events[i])); -abort (); } - val =avg_stats (stats) * scale; - source_count =1 ; - }else { -struct perf_stat_evsel *ps = metric_events[i]->stats; -struct perf_stat_aggr *aggr; - + } +/* Time events are always on CPU0, the first aggregation index. */ + aggr = &ps->aggr[is_tool_time ? tool_aggr_idx : aggr_idx]; +if (!aggr || !metric_events[i]->supported) {/* -* If there are multiple uncore PMUs and we're not -* reading the leader's stats, determine the stats for -* the appropriate uncore PMU. +* Not supported events will have a count of 0, which +* can be confusing in a metric. Explicitly set the +* value to NAN. Not counted events (enable time of 0) +* are read as 0. */ -if (evsel && evsel->metric_leader && - evsel->pmu != evsel->metric_leader->pmu && - mexp->metric_events[i]->pmu == evsel->metric_leader->pmu) { -struct evsel *pos; - -evlist__for_each_entry (evsel->evlist, pos) { -if (pos->pmu != evsel->pmu) -continue ; -if (pos->metric_leader != mexp->metric_events[i]) -continue ; - ps = pos->stats; - source_count =1 ; -break ; - } - } - aggr = &ps->aggr[aggr_idx]; -if (!aggr) -break ; - -if (!metric_events[i]->supported) { -/* -* Not supported events will have a count of 0, -* which can be confusing in a -* metric. Explicitly set the value to NAN. Not -* counted events (enable time of 0) are read as -* 0. -*/ - val =NAN ; - source_count =0 ; - }else { - val = aggr->counts.val; -if (!source_count) - source_count =evsel__source_count (metric_events[i]); - } + val =NAN ; + source_count =0 ; + }else { + val = aggr->counts.val; +if (is_tool_time) + val *=1e-9 ;/* Convert time event nanoseconds to seconds. */ +if (!source_count) + source_count =evsel__source_count (metric_events[i]); } n =strdup (evsel__metric_id (metric_events[i]));if (!n)
@@ -168,7 +151,7 @@ static void generic_metric(struct perf_stat_config *config, pctx->sctx.user_requested_cpu_list =strdup (config->user_requested_cpu_list); pctx->sctx.runtime = runtime; pctx->sctx.system_wide = config->system_wide; - i =prepare_metric (mexp, evsel, pctx, aggr_idx); + i =prepare_metric (config, mexp, evsel, pctx, aggr_idx);if (i <0 ) {expr__ctx_free (pctx);return ;
@@ -229,7 +212,7 @@ double test_generic_metric(struct metric_expr *mexp, int aggr_idx)if (!pctx)return NAN ; -if (prepare_metric (mexp,/*evsel=*/ NULL , pctx, aggr_idx) <0 ) +if (prepare_metric (/*config=*/ NULL , mexp,/*evsel=*/ NULL , pctx, aggr_idx) <0 ) goto out;if (expr__parse (&ratio, pctx, mexp->metric_expr))
Signature · 1 line
2.51.2.1041.gc1ab5b90ca-goog
Namhyung Kim
maintainerRe: [PATCH v4 04/10] perf stat-shadow: Read tool events directly
Quoted 85 lines · Ian Rogers
When reading time values for metrics don't use the globals updated in builtin-stat, just read the events as regular events. The only exception is for time events where nanoseconds need converting to seconds as metrics assume time metrics are in seconds.
Signed-off-by Ian Rogers <[email protected]>--- tools/perf/util/stat-shadow.c | 149 +++++++++++++++------------------- 1 file changed, 66 insertions(+), 83 deletions(-)tools/perf/util/stat-shadow.c +43
-14@@ -26,7 +26,32 @@ void perf_stat__reset_shadow_stats(void)memset (&ru_stats,0 ,sizeof (ru_stats)); } -static int prepare_metric (const struct metric_expr *mexp, +static bool tool_pmu__is_time_event (const struct perf_stat_config *config, +const struct evsel *evsel,int *tool_aggr_idx) +{ +enum tool_pmu_event event =evsel__tool_event (evsel); +int aggr_idx; + +if (event !=TOOL_PMU__EVENT_DURATION_TIME && + event !=TOOL_PMU__EVENT_USER_TIME && + event !=TOOL_PMU__EVENT_SYSTEM_TIME ) +return false; + +if (config) { +cpu_aggr_map__for_each_idx (aggr_idx, config->aggr_map) { +if (config->aggr_map->map[aggr_idx].cpu.cpu ==0 ) { + *tool_aggr_idx = aggr_idx; +return true; + } + } +pr_debug ("Unexpected CPU0 missing in aggregation for tool event.\n" ); + } + *tool_aggr_idx =0 ;/* Assume the first aggregation index works. */ +return true; +} + +static int prepare_metric (struct perf_stat_config *config, +const struct metric_expr *mexp,const struct evsel *evsel,struct expr_parse_ctx *pctx,int aggr_idx)@@ -36,93 +61,51 @@ static int prepare_metric(const struct metric_expr *mexp,int i;for (i =0 ; metric_events[i]; i++) { +int source_count =0 , tool_aggr_idx; +bool is_tool_time = +tool_pmu__is_time_event (config, metric_events[i], &tool_aggr_idx); +struct perf_stat_evsel *ps = metric_events[i]->stats; +struct perf_stat_aggr *aggr;char *n;double val; -int source_count =0 ; -if (evsel__is_tool (metric_events[i])) { -struct stats *stats; -double scale; - -switch (evsel__tool_event (metric_events[i])) { -case TOOL_PMU__EVENT_DURATION_TIME : - stats = &walltime_nsecs_stats; - scale =1e-9 ; -break ; -case TOOL_PMU__EVENT_USER_TIME : - stats = &ru_stats.ru_utime_usec_stat; - scale =1e-6 ; +/* +* If there are multiple uncore PMUs and we're not reading the +* leader's stats, determine the stats for the appropriate +* uncore PMU. +*/ +if (evsel && evsel->metric_leader && + evsel->pmu != evsel->metric_leader->pmu && + mexp->metric_events[i]->pmu == evsel->metric_leader->pmu) { +struct evsel *pos; + +evlist__for_each_entry (evsel->evlist, pos) { +if (pos->pmu != evsel->pmu)
tools/perf/util/stat-shadow.c static int prepare_metric(const struct metric_expr *mexp, · line 59
Quoted 10 lines
tools/perf/util/stat-shadow.c +5
-4+continue ; +if (pos->metric_leader != mexp->metric_events[i]) +continue ; + ps = pos->stats; + source_count =1 ;break ; -case TOOL_PMU__EVENT_SYSTEM_TIME : - stats = &ru_stats.ru_stime_usec_stat; - scale =1e-6 ; -break ;
I think this was broken as it seems to be converted to nsec in the update_rusage_stats().
Quoted 70 lines
tools/perf/util/stat-shadow.c +8
-59-case TOOL_PMU__EVENT_NONE : -pr_err ("Invalid tool event 'none'" ); -abort (); -case TOOL_PMU__EVENT_MAX : -pr_err ("Invalid tool event 'max'" ); -abort (); -case TOOL_PMU__EVENT_HAS_PMEM : -case TOOL_PMU__EVENT_NUM_CORES : -case TOOL_PMU__EVENT_NUM_CPUS : -case TOOL_PMU__EVENT_NUM_CPUS_ONLINE : -case TOOL_PMU__EVENT_NUM_DIES : -case TOOL_PMU__EVENT_NUM_PACKAGES : -case TOOL_PMU__EVENT_SLOTS : -case TOOL_PMU__EVENT_SMT_ON : -case TOOL_PMU__EVENT_SYSTEM_TSC_FREQ : -case TOOL_PMU__EVENT_CORE_WIDE : -case TOOL_PMU__EVENT_TARGET_CPU : -default : -pr_err ("Unexpected tool event '%s'" ,evsel__name (metric_events[i])); -abort (); } - val =avg_stats (stats) * scale; - source_count =1 ; - }else { -struct perf_stat_evsel *ps = metric_events[i]->stats; -struct perf_stat_aggr *aggr; - + } +/* Time events are always on CPU0, the first aggregation index. */ + aggr = &ps->aggr[is_tool_time ? tool_aggr_idx : aggr_idx]; +if (!aggr || !metric_events[i]->supported) {/* -* If there are multiple uncore PMUs and we're not -* reading the leader's stats, determine the stats for -* the appropriate uncore PMU. +* Not supported events will have a count of 0, which +* can be confusing in a metric. Explicitly set the +* value to NAN. Not counted events (enable time of 0) +* are read as 0. */ -if (evsel && evsel->metric_leader && - evsel->pmu != evsel->metric_leader->pmu && - mexp->metric_events[i]->pmu == evsel->metric_leader->pmu) { -struct evsel *pos; - -evlist__for_each_entry (evsel->evlist, pos) { -if (pos->pmu != evsel->pmu) -continue ; -if (pos->metric_leader != mexp->metric_events[i]) -continue ; - ps = pos->stats; - source_count =1 ; -break ; - } - } - aggr = &ps->aggr[aggr_idx]; -if (!aggr) -break ; - -if (!metric_events[i]->supported) { -/* -* Not supported events will have a count of 0, -* which can be confusing in a -* metric. Explicitly set the value to NAN. Not -* counted events (enable time of 0) are read as -* 0. -*/ - val =NAN ; - source_count =0 ; - }else {
tools/perf/util/stat-shadow.c static int prepare_metric(const struct metric_expr *mexp, · line 106
Quoted 10 lines
tools/perf/util/stat-shadow.c +6
-4- val = aggr->counts.val; -if (!source_count) - source_count =evsel__source_count (metric_events[i]); - } + val =NAN ; + source_count =0 ; + }else { + val = aggr->counts.val; +if (is_tool_time) + val *=1e-9 ;/* Convert time event nanoseconds to seconds. */
And this code treats all time events are in nsec now.
Thanks,
Namhyung
Quoted 26 lines
tools/perf/util/stat-shadow.c +4
-2+if (!source_count) + source_count =evsel__source_count (metric_events[i]); } n =strdup (evsel__metric_id (metric_events[i]));if (!n)@@ -168,7 +151,7 @@ static void generic_metric(struct perf_stat_config *config, pctx->sctx.user_requested_cpu_list =strdup (config->user_requested_cpu_list); pctx->sctx.runtime = runtime; pctx->sctx.system_wide = config->system_wide; - i =prepare_metric (mexp, evsel, pctx, aggr_idx); + i =prepare_metric (config, mexp, evsel, pctx, aggr_idx);if (i <0 ) {expr__ctx_free (pctx);return ;@@ -229,7 +212,7 @@ double test_generic_metric(struct metric_expr *mexp, int aggr_idx)if (!pctx)return NAN ; -if (prepare_metric (mexp,/*evsel=*/ NULL , pctx, aggr_idx) <0 ) +if (prepare_metric (/*config=*/ NULL , mexp,/*evsel=*/ NULL , pctx, aggr_idx) <0 ) goto out;if (expr__parse (&ratio, pctx, mexp->metric_expr))Signature · 1 line
2.51.2.1041.gc1ab5b90ca-goog
Ian Rogers
reviewerRe: [PATCH v4 04/10] perf stat-shadow: Read tool events directly
Quoted 92 lines · Namhyung Kim
Quoted 90 lines · Ian Rogers
When reading time values for metrics don't use the globals updated in builtin-stat, just read the events as regular events. The only exception is for time events where nanoseconds need converting to seconds as metrics assume time metrics are in seconds.
Signed-off-by Ian Rogers <[email protected]>--- tools/perf/util/stat-shadow.c | 149 +++++++++++++++------------------- 1 file changed, 66 insertions(+), 83 deletions(-)tools/perf/util/stat-shadow.c +48
-14@@ -26,7 +26,32 @@ void perf_stat__reset_shadow_stats(void)memset (&ru_stats,0 ,sizeof (ru_stats)); } -static int prepare_metric (const struct metric_expr *mexp, +static bool tool_pmu__is_time_event (const struct perf_stat_config *config, +const struct evsel *evsel,int *tool_aggr_idx) +{ +enum tool_pmu_event event =evsel__tool_event (evsel); +int aggr_idx; + +if (event !=TOOL_PMU__EVENT_DURATION_TIME && + event !=TOOL_PMU__EVENT_USER_TIME && + event !=TOOL_PMU__EVENT_SYSTEM_TIME ) +return false; + +if (config) { +cpu_aggr_map__for_each_idx (aggr_idx, config->aggr_map) { +if (config->aggr_map->map[aggr_idx].cpu.cpu ==0 ) { + *tool_aggr_idx = aggr_idx; +return true; + } + } +pr_debug ("Unexpected CPU0 missing in aggregation for tool event.\n" ); + } + *tool_aggr_idx =0 ;/* Assume the first aggregation index works. */ +return true; +} + +static int prepare_metric (struct perf_stat_config *config, +const struct metric_expr *mexp,const struct evsel *evsel,struct expr_parse_ctx *pctx,int aggr_idx)@@ -36,93 +61,51 @@ static int prepare_metric(const struct metric_expr *mexp,int i;for (i =0 ; metric_events[i]; i++) { +int source_count =0 , tool_aggr_idx; +bool is_tool_time = +tool_pmu__is_time_event (config, metric_events[i], &tool_aggr_idx); +struct perf_stat_evsel *ps = metric_events[i]->stats; +struct perf_stat_aggr *aggr;char *n;double val; -int source_count =0 ; -if (evsel__is_tool (metric_events[i])) { -struct stats *stats; -double scale; - -switch (evsel__tool_event (metric_events[i])) { -case TOOL_PMU__EVENT_DURATION_TIME : - stats = &walltime_nsecs_stats; - scale =1e-9 ; -break ; -case TOOL_PMU__EVENT_USER_TIME : - stats = &ru_stats.ru_utime_usec_stat; - scale =1e-6 ; +/* +* If there are multiple uncore PMUs and we're not reading the +* leader's stats, determine the stats for the appropriate +* uncore PMU. +*/ +if (evsel && evsel->metric_leader && + evsel->pmu != evsel->metric_leader->pmu && + mexp->metric_events[i]->pmu == evsel->metric_leader->pmu) { +struct evsel *pos; + +evlist__for_each_entry (evsel->evlist, pos) { +if (pos->pmu != evsel->pmu) +continue ; +if (pos->metric_leader != mexp->metric_events[i]) +continue ; + ps = pos->stats; + source_count =1 ;
Quoted 8 lines
tools/perf/util/stat-shadow.c static int prepare_metric(const struct metric_expr *mexp, · line 59
Quoted 5 lines
tools/perf/util/stat-shadow.c +0
-4break ; -case TOOL_PMU__EVENT_SYSTEM_TIME : - stats = &ru_stats.ru_stime_usec_stat; - scale =1e-6 ; -break ;I think this was broken as it seems to be converted to nsec in the
update_rusage_stats().
The global utime/stime is in usecs as in the name. In the tool_pmu it is converted to nanoseconds, for consistency with duration_time. In the new code the counters are all in nanoseconds and so the same scaling factor can be applied.
Quoted 74 lines
Quoted 74 lines
tools/perf/util/stat-shadow.c +8
-63-case TOOL_PMU__EVENT_NONE : -pr_err ("Invalid tool event 'none'" ); -abort (); -case TOOL_PMU__EVENT_MAX : -pr_err ("Invalid tool event 'max'" ); -abort (); -case TOOL_PMU__EVENT_HAS_PMEM : -case TOOL_PMU__EVENT_NUM_CORES : -case TOOL_PMU__EVENT_NUM_CPUS : -case TOOL_PMU__EVENT_NUM_CPUS_ONLINE : -case TOOL_PMU__EVENT_NUM_DIES : -case TOOL_PMU__EVENT_NUM_PACKAGES : -case TOOL_PMU__EVENT_SLOTS : -case TOOL_PMU__EVENT_SMT_ON : -case TOOL_PMU__EVENT_SYSTEM_TSC_FREQ : -case TOOL_PMU__EVENT_CORE_WIDE : -case TOOL_PMU__EVENT_TARGET_CPU : -default : -pr_err ("Unexpected tool event '%s'" ,evsel__name (metric_events[i])); -abort (); } - val =avg_stats (stats) * scale; - source_count =1 ; - }else { -struct perf_stat_evsel *ps = metric_events[i]->stats; -struct perf_stat_aggr *aggr; - + } +/* Time events are always on CPU0, the first aggregation index. */ + aggr = &ps->aggr[is_tool_time ? tool_aggr_idx : aggr_idx]; +if (!aggr || !metric_events[i]->supported) {/* -* If there are multiple uncore PMUs and we're not -* reading the leader's stats, determine the stats for -* the appropriate uncore PMU. +* Not supported events will have a count of 0, which +* can be confusing in a metric. Explicitly set the +* value to NAN. Not counted events (enable time of 0) +* are read as 0. */ -if (evsel && evsel->metric_leader && - evsel->pmu != evsel->metric_leader->pmu && - mexp->metric_events[i]->pmu == evsel->metric_leader->pmu) { -struct evsel *pos; - -evlist__for_each_entry (evsel->evlist, pos) { -if (pos->pmu != evsel->pmu) -continue ; -if (pos->metric_leader != mexp->metric_events[i]) -continue ; - ps = pos->stats; - source_count =1 ; -break ; - } - } - aggr = &ps->aggr[aggr_idx]; -if (!aggr) -break ; - -if (!metric_events[i]->supported) { -/* -* Not supported events will have a count of 0, -* which can be confusing in a -* metric. Explicitly set the value to NAN. Not -* counted events (enable time of 0) are read as -* 0. -*/ - val =NAN ; - source_count =0 ; - }else { - val = aggr->counts.val; -if (!source_count) - source_count =evsel__source_count (metric_events[i]); - }
Quoted 8 lines
tools/perf/util/stat-shadow.c static int prepare_metric(const struct metric_expr *mexp, · line 106
Quoted 6 lines
tools/perf/util/stat-shadow.c +6
-0+ val =NAN ; + source_count =0 ; + }else { + val = aggr->counts.val; +if (is_tool_time) + val *=1e-9 ;/* Convert time event nanoseconds to seconds. */ And this code treats all time events are in nsec now.
That's correct: user_time, system_time and duration_time are all counters that give a count in nanoseconds. However, in the metrics, I guess as the counts are doubles, the time has always been in whole seconds.
Thanks,
Ian
Quoted 30 lines
Thanks,
NamhyungQuoted 26 lines
tools/perf/util/stat-shadow.c +4
-3+if (!source_count) + source_count =evsel__source_count (metric_events[i]); } n =strdup (evsel__metric_id (metric_events[i]));if (!n)@@ -168,7 +151,7 @@ static void generic_metric(struct perf_stat_config *config, pctx->sctx.user_requested_cpu_list =strdup (config->user_requested_cpu_list); pctx->sctx.runtime = runtime; pctx->sctx.system_wide = config->system_wide; - i =prepare_metric (mexp, evsel, pctx, aggr_idx); + i =prepare_metric (config, mexp, evsel, pctx, aggr_idx);if (i <0 ) {expr__ctx_free (pctx);return ;@@ -229,7 +212,7 @@ double test_generic_metric(struct metric_expr *mexp, int aggr_idx)if (!pctx)return NAN ; -if (prepare_metric (mexp,/*evsel=*/ NULL , pctx, aggr_idx) <0 ) +if (prepare_metric (/*config=*/ NULL , mexp,/*evsel=*/ NULL , pctx, aggr_idx) <0 ) goto out;if (expr__parse (&ratio, pctx, mexp->metric_expr)) --2.51.2.1041.gc1ab5b90ca-goog
Namhyung Kim
maintainerRe: [PATCH v4 04/10] perf stat-shadow: Read tool events directly
Quoted 10 lines · Ian Rogers
Quoted 9 lines · Namhyung Kim
Quoted 7 lines · Ian Rogers
When reading time values for metrics don't use the globals updated in builtin-stat, just read the events as regular events. The only exception is for time events where nanoseconds need converting to seconds as metrics assume time metrics are in seconds.
Signed-off-by Ian Rogers <[email protected]>---
[SNIP]
Quoted 5 lines
Quoted 5 lines
Quoted 4 lines
tools/perf/util/stat-shadow.c +0
-4-case TOOL_PMU__EVENT_SYSTEM_TIME : - stats = &ru_stats.ru_stime_usec_stat; - scale =1e-6 ; -break ;
Quoted 7 lines
Quoted 2 lines
I think this was broken as it seems to be converted to nsec in the
update_rusage_stats().The global utime/stime is in usecs as in the name. In the
tool_pmuit is converted to nanoseconds, for consistency withduration_time. In the new code the counters are all in nanoseconds and so the same scaling factor can be applied.
What do you mean by global utime/stime?
Quoted 84 lines
Quoted 75 lines
Quoted 75 lines
tools/perf/util/stat-shadow.c +8
-63-case TOOL_PMU__EVENT_NONE : -pr_err ("Invalid tool event 'none'" ); -abort (); -case TOOL_PMU__EVENT_MAX : -pr_err ("Invalid tool event 'max'" ); -abort (); -case TOOL_PMU__EVENT_HAS_PMEM : -case TOOL_PMU__EVENT_NUM_CORES : -case TOOL_PMU__EVENT_NUM_CPUS : -case TOOL_PMU__EVENT_NUM_CPUS_ONLINE : -case TOOL_PMU__EVENT_NUM_DIES : -case TOOL_PMU__EVENT_NUM_PACKAGES : -case TOOL_PMU__EVENT_SLOTS : -case TOOL_PMU__EVENT_SMT_ON : -case TOOL_PMU__EVENT_SYSTEM_TSC_FREQ : -case TOOL_PMU__EVENT_CORE_WIDE : -case TOOL_PMU__EVENT_TARGET_CPU : -default : -pr_err ("Unexpected tool event '%s'" ,evsel__name (metric_events[i])); -abort (); } - val =avg_stats (stats) * scale; - source_count =1 ; - }else { -struct perf_stat_evsel *ps = metric_events[i]->stats; -struct perf_stat_aggr *aggr; - + } +/* Time events are always on CPU0, the first aggregation index. */ + aggr = &ps->aggr[is_tool_time ? tool_aggr_idx : aggr_idx]; +if (!aggr || !metric_events[i]->supported) {/* -* If there are multiple uncore PMUs and we're not -* reading the leader's stats, determine the stats for -* the appropriate uncore PMU. +* Not supported events will have a count of 0, which +* can be confusing in a metric. Explicitly set the +* value to NAN. Not counted events (enable time of 0) +* are read as 0. */ -if (evsel && evsel->metric_leader && - evsel->pmu != evsel->metric_leader->pmu && - mexp->metric_events[i]->pmu == evsel->metric_leader->pmu) { -struct evsel *pos; - -evlist__for_each_entry (evsel->evlist, pos) { -if (pos->pmu != evsel->pmu) -continue ; -if (pos->metric_leader != mexp->metric_events[i]) -continue ; - ps = pos->stats; - source_count =1 ; -break ; - } - } - aggr = &ps->aggr[aggr_idx]; -if (!aggr) -break ; - -if (!metric_events[i]->supported) { -/* -* Not supported events will have a count of 0, -* which can be confusing in a -* metric. Explicitly set the value to NAN. Not -* counted events (enable time of 0) are read as -* 0. -*/ - val =NAN ; - source_count =0 ; - }else { - val = aggr->counts.val; -if (!source_count) - source_count =evsel__source_count (metric_events[i]); - }Quoted 8 lines
tools/perf/util/stat-shadow.c static int prepare_metric(const struct metric_expr *mexp, · line 106
Quoted 6 lines
tools/perf/util/stat-shadow.c +6
-0+ val =NAN ; + source_count =0 ; + }else { + val = aggr->counts.val; +if (is_tool_time) + val *=1e-9 ;/* Convert time event nanoseconds to seconds. */ And this code treats all time events are in nsec now.
Quoted 4 lines
That's correct:
user_time,system_timeandduration_timeare all counters that give a count in nanoseconds. However, in the metrics, I guess as the counts are doubles, the time has always been in whole seconds.
Yep, I think it's ok.
Thanks,
Namhyung
Ian Rogers
reviewerPATCH v4 5/10 perf stat: Reduce scope of ru_stats
In reply to Ian RogersThe ru_stats are used to capture user and system time stats when a process exits. These are then applied to user and system time tool events if their reads fail due to the process terminating. Reduce the scope now the metric code no longer reads these values.
--- tools/perf/builtin-stat.c | 19 ++++++++++++++++++- tools/perf/util/config.c | 1 - tools/perf/util/stat-shadow.c | 2 -- tools/perf/util/stat.h | 21 --------------------- 4 files changed, 18 insertions(+), 25 deletions(-)
tools/perf/builtin-stat.c +18 -1
@@ -104,6 +104,11 @@#define DEFAULT_SEPARATOR " "#define FREEZE_ON_SMI_PATH "bus/event_source/devices/cpu/freeze_on_smi" +struct rusage_stats { +struct stats ru_utime_usec_stat; +struct stats ru_stime_usec_stat; +}; +static void print_counters (struct timespec *ts,int argc,const char **argv);static struct evlist *evsel_list;
@@ -133,6 +138,7 @@ static bool interval_count;static const char *output_name;static int output_fd;static char *metrics; +static struct rusage_stats ru_stats;struct perf_stat {bool record;
@@ -730,6 +736,17 @@ static int create_perf_stat_counter(struct evsel *evsel, evsel->core.threads); } +static void update_rusage_stats (const struct rusage *rusage) +{ +const u64 us_to_ns =1000 ; +const u64 s_to_ns =1000000000 ; + +update_stats (&ru_stats.ru_utime_usec_stat, + (rusage->ru_utime.tv_usec * us_to_ns + rusage->ru_utime.tv_sec * s_to_ns)); +update_stats (&ru_stats.ru_stime_usec_stat, + (rusage->ru_stime.tv_usec * us_to_ns + rusage->ru_stime.tv_sec * s_to_ns)); +} +static int __run_perf_stat (int argc,const char **argv,int run_idx) {int interval = stat_config.interval;
@@ -979,7 +996,7 @@ static int __run_perf_stat(int argc, const char **argv, int run_idx)evlist__reset_aggr_stats (evsel_list); }else {update_stats (&walltime_nsecs_stats, t1 - t0); -update_rusage_stats (&ru_stats, &stat_config.ru_data); +update_rusage_stats (&stat_config.ru_data); }/*
tools/perf/util/config.c +0 -1
@@ -45,7 +45,6 @@ struct perf_stat_config stat_config = { .run_count = 1, .metric_only_len = METRIC_ONLY_LEN, .walltime_nsecs_stats = &walltime_nsecs_stats, - .ru_stats = &ru_stats, .big_num = true, .ctl_fd = -1, .ctl_fd_ack = -1,
tools/perf/util/stat-shadow.c +0 -2
@@ -18,12 +18,10 @@#include "tool_pmu.h" struct stats walltime_nsecs_stats; -struct rusage_stats ru_stats;void perf_stat__reset_shadow_stats (void ) {memset (&walltime_nsecs_stats,0 ,sizeof (walltime_nsecs_stats)); -memset (&ru_stats,0 ,sizeof (ru_stats)); } static bool tool_pmu__is_time_event(const struct perf_stat_config *config,
tools/perf/util/stat.h +0 -21
@@ -56,11 +56,6 @@ enum aggr_mode {AGGR_MAX }; -struct rusage_stats { -struct stats ru_utime_usec_stat; -struct stats ru_stime_usec_stat; -}; -typedef struct aggr_cpu_id (*aggr_get_id_t )(struct perf_stat_config *config,struct perf_cpu cpu);struct perf_stat_config {
@@ -102,7 +97,6 @@ struct perf_stat_config {const char *csv_sep;struct stats *walltime_nsecs_stats;struct rusage ru_data; -struct rusage_stats *ru_stats;struct cpu_aggr_map *aggr_map;aggr_get_id_t aggr_get_id;struct cpu_aggr_map *cpus_aggr_map;
@@ -132,25 +126,10 @@ static inline void init_stats(struct stats *stats) stats->max =0 ; } -static inline void init_rusage_stats (struct rusage_stats *ru_stats) { -init_stats (&ru_stats->ru_utime_usec_stat); -init_stats (&ru_stats->ru_stime_usec_stat); -} - -static inline void update_rusage_stats (struct rusage_stats *ru_stats,struct rusage * rusage) { -const u64 us_to_ns =1000 ; -const u64 s_to_ns =1000000000 ; -update_stats (&ru_stats->ru_utime_usec_stat, - (rusage->ru_utime.tv_usec * us_to_ns + rusage->ru_utime.tv_sec * s_to_ns)); -update_stats (&ru_stats->ru_stime_usec_stat, - (rusage->ru_stime.tv_usec * us_to_ns + rusage->ru_stime.tv_sec * s_to_ns)); -} -struct evsel ;struct evlist ;extern struct stats walltime_nsecs_stats; -extern struct rusage_stats ru_stats;enum metric_threshold_classify {METRIC_THRESHOLD_UNKNOWN ,
Signature · 1 line
2.51.2.1041.gc1ab5b90ca-goog
Namhyung Kim
maintainerRe: [PATCH v4 05/10] perf stat: Reduce scope of ru_stats
Quoted 18 lines · Ian Rogers
The
ru_statsare used to capture user and system time stats when a process exits. These are then applied to user and system time tool events if their reads fail due to the process terminating. Reduce the scope now the metric code no longer reads these values.Signed-off-by Ian Rogers <[email protected]>--- tools/perf/builtin-stat.c | 19 ++++++++++++++++++- tools/perf/util/config.c | 1 - tools/perf/util/stat-shadow.c | 2 -- tools/perf/util/stat.h | 21 --------------------- 4 files changed, 18 insertions(+), 25 deletions(-)tools/perf/builtin-stat.c +0
-0
tools/perf/builtin-stat.c line 109
Quoted 6 lines
tools/perf/builtin-stat.c +3
-0@@ -104,6 +104,11 @@#define DEFAULT_SEPARATOR " "#define FREEZE_ON_SMI_PATH "bus/event_source/devices/cpu/freeze_on_smi" +struct rusage_stats { +struct stats ru_utime_usec_stat; +struct stats ru_stime_usec_stat;
Maybe we need to rename it to ru_[us]time_nsec_stat now? But it can be a separate change.
Thanks,
Namhyung
Quoted 122 lines
tools/perf/builtin-stat.c +15
-1+}; +static void print_counters (struct timespec *ts,int argc,const char **argv);static struct evlist *evsel_list;@@ -133,6 +138,7 @@ static bool interval_count;static const char *output_name;static int output_fd;static char *metrics; +static struct rusage_stats ru_stats;struct perf_stat {bool record;@@ -730,6 +736,17 @@ static int create_perf_stat_counter(struct evsel *evsel, evsel->core.threads); } +static void update_rusage_stats (const struct rusage *rusage) +{ +const u64 us_to_ns =1000 ; +const u64 s_to_ns =1000000000 ; + +update_stats (&ru_stats.ru_utime_usec_stat, + (rusage->ru_utime.tv_usec * us_to_ns + rusage->ru_utime.tv_sec * s_to_ns)); +update_stats (&ru_stats.ru_stime_usec_stat, + (rusage->ru_stime.tv_usec * us_to_ns + rusage->ru_stime.tv_sec * s_to_ns)); +} +static int __run_perf_stat (int argc,const char **argv,int run_idx) {int interval = stat_config.interval;@@ -979,7 +996,7 @@ static int __run_perf_stat(int argc, const char **argv, int run_idx)evlist__reset_aggr_stats (evsel_list); }else {update_stats (&walltime_nsecs_stats, t1 - t0); -update_rusage_stats (&ru_stats, &stat_config.ru_data); +update_rusage_stats (&stat_config.ru_data); }/* tools/perf/util/config.c +0
-1@@ -45,7 +45,6 @@ struct perf_stat_config stat_config = { .run_count = 1, .metric_only_len = METRIC_ONLY_LEN, .walltime_nsecs_stats = &walltime_nsecs_stats, - .ru_stats = &ru_stats, .big_num = true, .ctl_fd = -1, .ctl_fd_ack = -1,tools/perf/util/stat-shadow.c +0
-2@@ -18,12 +18,10 @@#include "tool_pmu.h" struct stats walltime_nsecs_stats; -struct rusage_stats ru_stats;void perf_stat__reset_shadow_stats (void ) {memset (&walltime_nsecs_stats,0 ,sizeof (walltime_nsecs_stats)); -memset (&ru_stats,0 ,sizeof (ru_stats)); } static bool tool_pmu__is_time_event(const struct perf_stat_config *config,tools/perf/util/stat.h +0
-21@@ -56,11 +56,6 @@ enum aggr_mode {AGGR_MAX }; -struct rusage_stats { -struct stats ru_utime_usec_stat; -struct stats ru_stime_usec_stat; -}; -typedef struct aggr_cpu_id (*aggr_get_id_t )(struct perf_stat_config *config,struct perf_cpu cpu);struct perf_stat_config {@@ -102,7 +97,6 @@ struct perf_stat_config {const char *csv_sep;struct stats *walltime_nsecs_stats;struct rusage ru_data; -struct rusage_stats *ru_stats;struct cpu_aggr_map *aggr_map;aggr_get_id_t aggr_get_id;struct cpu_aggr_map *cpus_aggr_map;@@ -132,25 +126,10 @@ static inline void init_stats(struct stats *stats) stats->max =0 ; } -static inline void init_rusage_stats (struct rusage_stats *ru_stats) { -init_stats (&ru_stats->ru_utime_usec_stat); -init_stats (&ru_stats->ru_stime_usec_stat); -} - -static inline void update_rusage_stats (struct rusage_stats *ru_stats,struct rusage * rusage) { -const u64 us_to_ns =1000 ; -const u64 s_to_ns =1000000000 ; -update_stats (&ru_stats->ru_utime_usec_stat, - (rusage->ru_utime.tv_usec * us_to_ns + rusage->ru_utime.tv_sec * s_to_ns)); -update_stats (&ru_stats->ru_stime_usec_stat, - (rusage->ru_stime.tv_usec * us_to_ns + rusage->ru_stime.tv_sec * s_to_ns)); -} -struct evsel ;struct evlist ;extern struct stats walltime_nsecs_stats; -extern struct rusage_stats ru_stats;enum metric_threshold_classify {METRIC_THRESHOLD_UNKNOWN ,Signature · 1 line
2.51.2.1041.gc1ab5b90ca-goog
Ian Rogers
reviewerPATCH v4 6/10 perf stat: Reduce scope of walltime_nsecs_stats
In reply to Ian Rogerswalltime_nsecs_stats is no longer used for counter values, move into that stat_config where it controls certain things like noise measurement.
--- tools/perf/builtin-stat.c | 16 ++++++++-------- tools/perf/tests/parse-metric.c | 2 -- tools/perf/tests/pmu-events.c | 2 -- tools/perf/util/config.c | 2 ++ tools/perf/util/stat-shadow.c | 7 ------- tools/perf/util/stat.h | 2 -- 6 files changed, 10 insertions(+), 21 deletions(-)
tools/perf/builtin-stat.c +8 -8
@@ -239,7 +239,7 @@ static inline void diff_timespec(struct timespec *r, struct timespec *a,static void perf_stat__reset_stats (void ) {evlist__reset_stats (evsel_list); -perf_stat__reset_shadow_stats (); +memset (stat_config.walltime_nsecs_stats,0 ,sizeof (*stat_config.walltime_nsecs_stats)); } static int process_synthesized_event(const struct perf_tool *tool __maybe_unused,
@@ -455,8 +455,8 @@ static void process_interval(void)pr_err ("failed to write stat round event\n" ); } -init_stats (&walltime_nsecs_stats); -update_stats (&walltime_nsecs_stats, stat_config.interval *1000000ULL ); +init_stats (stat_config.walltime_nsecs_stats); +update_stats (stat_config.walltime_nsecs_stats, stat_config.interval *1000000ULL );print_counters (&rs,0 ,NULL ); }
@@ -988,14 +988,14 @@ static int __run_perf_stat(int argc, const char **argv, int run_idx)if (interval && stat_config.summary) { stat_config.interval =0 ; stat_config.stop_read_counter = true; -init_stats (&walltime_nsecs_stats); -update_stats (&walltime_nsecs_stats, t1 - t0); +init_stats (stat_config.walltime_nsecs_stats); +update_stats (stat_config.walltime_nsecs_stats, t1 - t0);evlist__copy_prev_raw_counts (evsel_list);evlist__reset_prev_raw_counts (evsel_list);evlist__reset_aggr_stats (evsel_list); }else { -update_stats (&walltime_nsecs_stats, t1 - t0); +update_stats (stat_config.walltime_nsecs_stats, t1 - t0);update_rusage_stats (&stat_config.ru_data); }
@@ -2167,7 +2167,7 @@ static int process_stat_round_event(const struct perf_tool *tool __maybe_unused,process_counters ();if (stat_round->type ==PERF_STAT_ROUND_TYPE__FINAL ) -update_stats (&walltime_nsecs_stats, stat_round->time); +update_stats (stat_config.walltime_nsecs_stats, stat_round->time);if (stat_config.interval && stat_round->time) { tsh.tv_sec = stat_round->time /NSEC_PER_SEC ;
@@ -2975,7 +2975,7 @@ int cmd_stat(int argc, const char **argv) }if (!interval) { -if (WRITE_STAT_ROUND_EVENT (walltime_nsecs_stats.max,FINAL )) +if (WRITE_STAT_ROUND_EVENT (stat_config.walltime_nsecs_stats->max,FINAL ))pr_err ("failed to write stat round event\n" ); }
tools/perf/tests/parse-metric.c +0 -2
@@ -40,8 +40,6 @@ static void load_runtime_stat(struct evlist *evlist, struct value *vals) count =find_value (evsel->name, vals); evsel->supported = true; evsel->stats->aggr->counts.val = count; -if (evsel__name_is (evsel,"duration_time" )) -update_stats (&walltime_nsecs_stats, count); } }
tools/perf/tests/pmu-events.c +0 -2
@@ -872,8 +872,6 @@ static int test__parsing_callback(const struct pmu_metric *pm,evlist__alloc_aggr_stats (evlist,1 );evlist__for_each_entry (evlist, evsel) { evsel->stats->aggr->counts.val = k; -if (evsel__name_is (evsel,"duration_time" )) -update_stats (&walltime_nsecs_stats, k); k++; }evlist__for_each_entry (evlist, evsel) {
tools/perf/util/config.c +2 -0
@@ -37,6 +37,8 @@#define METRIC_ONLY_LEN 20 +static struct stats walltime_nsecs_stats; +struct perf_stat_config stat_config = { .aggr_mode =AGGR_GLOBAL , .aggr_level =MAX_CACHE_LVL +1 ,
tools/perf/util/stat-shadow.c +0 -7
@@ -17,13 +17,6 @@#include "util/hashmap.h" #include "tool_pmu.h" -struct stats walltime_nsecs_stats; - -void perf_stat__reset_shadow_stats (void ) -{ -memset (&walltime_nsecs_stats,0 ,sizeof (walltime_nsecs_stats)); -} -static bool tool_pmu__is_time_event (const struct perf_stat_config *config,const struct evsel *evsel,int *tool_aggr_idx) {
tools/perf/util/stat.h +0 -2
@@ -129,8 +129,6 @@ static inline void init_stats(struct stats *stats)struct evsel ;struct evlist ; -extern struct stats walltime_nsecs_stats; -enum metric_threshold_classify {METRIC_THRESHOLD_UNKNOWN ,METRIC_THRESHOLD_BAD ,
Signature · 1 line
2.51.2.1041.gc1ab5b90ca-goog
Ian Rogers
reviewerPATCH v4 7/10 perf tool_pmu: More accurately set the cpus for tool events
In reply to Ian RogersThe user and system time events can record on different CPUs, but for all other events a single CPU map of just CPU 0 makes sense. In parse-events detect a tool PMU and then pass the perf_event_attr so that the tool_pmu can return CPUs specific for the event. This avoids a CPU map of all online CPUs being used for events like duration_time. Avoiding this avoids the evlist CPUs containing CPUs for which duration_time just gives 0. Minimizing the evlist CPUs can remove unnecessary sched_setaffinity syscalls that delay metric calculations.
--- tools/perf/util/parse-events.c | 9 +++++++-- tools/perf/util/tool_pmu.c | 19 +++++++++++++++++++ tools/perf/util/tool_pmu.h | 1 + 3 files changed, 27 insertions(+), 2 deletions(-)
tools/perf/util/parse-events.c +7 -2
@@ -30,6 +30,7 @@#include "util/event.h" #include "util/bpf-filter.h" #include "util/stat.h" +#include "util/tool_pmu.h" #include "util/util.h" #include "tracepoint.h" #include <api/fs/tracing_path.h>
@@ -227,8 +228,12 @@ __add_event(struct list_head *list, int *idx,if (pmu) { is_pmu_core = pmu->is_core; pmu_cpus =perf_cpu_map__get (pmu->cpus); -if (perf_cpu_map__is_empty (pmu_cpus)) - pmu_cpus =cpu_map__online (); +if (perf_cpu_map__is_empty (pmu_cpus)) { +if (perf_pmu__is_tool (pmu)) + pmu_cpus =tool_pmu__cpus (attr); +else + pmu_cpus =cpu_map__online (); + } }else { is_pmu_core = (attr->type ==PERF_TYPE_HARDWARE || attr->type ==PERF_TYPE_HW_CACHE );
tools/perf/util/tool_pmu.c +19 -0
@@ -2,6 +2,7 @@#include "cgroup.h" #include "counts.h" #include "cputopo.h" +#include "debug.h" #include "evsel.h" #include "pmu.h" #include "print-events.h"
@@ -13,6 +14,7 @@#include <api/fs/fs.h> #include <api/io.h> #include <internal/threadmap.h> +#include <perf/cpumap.h> #include <perf/threadmap.h> #include <fcntl.h> #include <strings.h>
@@ -109,6 +111,23 @@ const char *evsel__tool_pmu_event_name(const struct evsel *evsel)return tool_pmu__event_to_str (evsel->core.attr.config); } +struct perf_cpu_map *tool_pmu__cpus (struct perf_event_attr *attr) +{ +static struct perf_cpu_map *cpu0_map; +enum tool_pmu_event event = (enum tool_pmu_event )attr->config; + +if (event <=TOOL_PMU__EVENT_NONE || event >=TOOL_PMU__EVENT_MAX ) { +pr_err ("Invalid tool PMU event config %llx\n" , attr->config); +return NULL ; + } +if (event ==TOOL_PMU__EVENT_USER_TIME || event ==TOOL_PMU__EVENT_SYSTEM_TIME ) +return cpu_map__online (); + +if (!cpu0_map) + cpu0_map =perf_cpu_map__new_int (0 ); +return perf_cpu_map__get (cpu0_map); +} +static bool read_until_char (struct io *io,char e) {int c;
tools/perf/util/tool_pmu.h +1 -0
@@ -46,6 +46,7 @@ bool tool_pmu__read_event(enum tool_pmu_event ev,u64 tool_pmu__cpu_slots_per_cycle (void );bool perf_pmu__is_tool (const struct perf_pmu *pmu); +struct perf_cpu_map *tool_pmu__cpus (struct perf_event_attr *attr);bool evsel__is_tool (const struct evsel *evsel);enum tool_pmu_event evsel__tool_event (const struct evsel *evsel);
Signature · 1 line
2.51.2.1041.gc1ab5b90ca-goog
Re: [PATCH v4 07/10] perf tool_pmu: More accurately set the cpus for tool events
Hi,
Quoted 9 lines · Ian Rogers
The user and system time events can record on different CPUs, but for all other events a single CPU map of just CPU 0 makes sense. In parse-events detect a tool PMU and then pass the
perf_event_attrso that thetool_pmucan return CPUs specific for the event. This avoids a CPU map of all online CPUs being used for events likeduration_time. Avoiding this avoids the evlist CPUs containing CPUs for whichduration_timejust gives 0. Minimizing the evlist CPUs can remove unnecessarysched_setaffinitysyscalls that delay metric calculations.
I was just testing v6.19-rc* and noticed that
perf stat -C $somecpu sleep 1
segfaults.
I bisected that down to this change (d8d8a0b3603a9a8fa207cf9e4f292e81dc5d1008).
$ git describe v6.18-rc1-116-gd8d8a0b3603a9
$ gdb --args perf stat -C11 sleep 1 (gdb) r Starting program: /home/andres/bin/bin/perf stat -C11 sleep 1 [Thread debugging using libthread_db enabled] Using host libthread_db library "/usr/lib/x86_64-linux-gnu/libthread_db.so.1". [Detaching after fork from child process 843870]
Performance counter stats for 'CPU(s) 11':
16 context-switches # 16.0 cs/sec cs_per_second
1,001.66 msec cpu-clock
Program received signal SIGSEGV, Segmentation fault.prepare_metric (config=config@entry=0x55555603bd80 <stat_config>, mexp=mexp@entry=0x5555560a40f0, evsel=evsel@entry=0x555556093e20,
pctx=pctx@entry=0x5555560a7150, aggr_idx=aggr_idx@entry=0) at util/stat-shadow.c:85
85 aggr = &ps->aggr[is_tool_time ? tool_aggr_idx : aggr_idx];
(gdb) bt
#0 prepare_metric (config=config@entry=0x55555603bd80 <stat_config>, mexp=mexp@entry=0x5555560a40f0, evsel=evsel@entry=0x555556093e20,
pctx=pctx@entry=0x5555560a7150, aggr_idx=aggr_idx@entry=0) at util/stat-shadow.c:85
#1 0x00005555557be0eb in generic_metric (config=<optimized out>, mexp=0x5555560a40f0, evsel=<optimized out>, aggr_idx=<optimized out>, out=0x7fffffff76c0)
at util/stat-shadow.c:146
#2 perf_stat__print_shadow_stats_metricgroup (config=config@entry=0x55555603bd80 <stat_config>, evsel=<optimized out>, aggr_idx=0,
num=num@entry=0x7fffffff7604, from=from@entry=0x0, out=0x7fffffff76c0) at util/stat-shadow.c:307
#3 0x00005555557be4c7 in perf_stat__print_shadow_stats (config=config@entry=0x55555603bd80 <stat_config>, evsel=evsel@entry=0x555556093e20,
aggr_idx=aggr_idx@entry=0, out=out@entry=0x7fffffff76c0) at util/stat-shadow.c:325
#4 0x00005555557c03a3 in printout (config=<optimized out>, os=0x7fffffff78b0, uval=<optimized out>, run=1001660763, ena=<optimized out>,
noise=<optimized out>, aggr_idx=0) at util/stat-display.c:874
#5 0x00005555557c1424 in print_counter_aggrdata (config=0x55555603bd80 <stat_config>, counter=0x555556093e20, aggr_idx=0, os=<optimized out>)
at util/stat-display.c:1013
#6 0x00005555557c3051 in print_counter (config=<optimized out>, counter=<optimized out>, os=<optimized out>) at util/stat-display.c:1127
#7 print_counter (config=<optimized out>, counter=<optimized out>, os=<optimized out>) at util/stat-display.c:1117
#8 evlist__print_counters (evlist=<optimized out>, config=config@entry=0x55555603bd80 <stat_config>, _target=_target@entry=0x5555560412e0 <target>,
ts=ts@entry=0x0, argc=argc@entry=2, argv=argv@entry=0x7fffffffdb20) at util/stat-display.c:1600#9 0x00005555555d0d27 in print_counters (ts=0x0, argc=2, argv=0x7fffffffdb20) at builtin-stat.c:1070
#10 print_counters (ts=0x0, argc=2, argv=0x7fffffffdb20) at builtin-stat.c:1062
#11 cmd_stat (argc=2, argv=0x7fffffffdb20) at builtin-stat.c:2949
#12 0x000055555562fee2 in run_builtin (p=p@entry=0x55555602df48 <commands+360>, argc=argc@entry=4, argv=argv@entry=0x7fffffffdb20) at perf.c:349 #13 0x00005555556301ce in handle_internal_command (argc=argc@entry=4, argv=argv@entry=0x7fffffffdb20) at perf.c:401
#14 0x00005555555a8d33 in run_argv (argcp=<synthetic pointer>, argv=<synthetic pointer>) at perf.c:445
#15 main (argc=<optimized out>, argv=0x7fffffffdb20) at perf.c:553
(gdb) bt full
#0 prepare_metric (config=config@entry=0x55555603bd80 <stat_config>, mexp=mexp@entry=0x5555560a40f0, evsel=evsel@entry=0x555556093e20, pctx=pctx@entry=0x5555560a7150, aggr_idx=aggr_idx@entry=0) at util/stat-shadow.c:85
val = <optimized out>
source_count = 0
tool_aggr_idx = 0
is_tool_time = true
ps = 0x0
aggr = <optimized out>
n = <optimized out>
metric_events = <optimized out>
metric_refs = 0x0
i = 1
#1 0x00005555557be0eb in generic_metric (config=<optimized out>, mexp=0x5555560a40f0, evsel=<optimized out>, aggr_idx=<optimized out>, out=0x7fffffff76c0) at util/stat-shadow.c:146
print_metric = 0x5555557c1b70 <print_metric_std>
metric_name = 0x5555560ae550 "CPUs_utilized"
metric_expr = 0x555555ea5bb5 "(software@cpu\\-clock\\,name\\=cpu\\-clock@ if #target_cpu else software@task\\-clock\\,name\\=task\\-clock@) / (duration_time * 1e9)"
metric_threshold = 0x0
metric_unit = 0x555555ea5c4d "1CPUs"
metric_events = 0x5555560ae490
runtime = 0
pctx = 0x5555560a7150
ratio = 15.973479072960373
scale = 1
threshold = 4.9406564584124654e-324
i = <optimized out>
ctxp = 0x7fffffff78b0
thresh = METRIC_THRESHOLD_UNKNOWN
#2 perf_stat__print_shadow_stats_metricgroup (config=config@entry=0x55555603bd80 <stat_config>, evsel=<optimized out>, aggr_idx=0, num=num@entry=0x7fffffff7604, from=from@entry=0x0, out=0x7fffffff76c0) at util/stat-shadow.c:307
me = 0x5555560877d0
mexp = 0x5555560a40f0
ctxp = 0x7fffffff78b0
header_printed = false
name = 0x555555879b28 ""
metric_events = <optimized out>
#3 0x00005555557be4c7 in perf_stat__print_shadow_stats (config=config@entry=0x55555603bd80 <stat_config>, evsel=evsel@entry=0x555556093e20, aggr_idx=aggr_idx@entry=0, out=out@entry=0x7fffffff76c0) at util/stat-shadow.c:325
print_metric = 0x5555557c1b70 <print_metric_std>
ctxp = 0x7fffffff78b0
num = 1#4 0x00005555557c03a3 in printout (config=<optimized out>, os=0x7fffffff78b0, uval=<optimized out>, run=1001660763, ena=<optimized out>, noise=<optimized out>, aggr_idx=0) at util/stat-display.c:874
out = {ctx = 0x7fffffff78b0, print_metric = 0x5555557c1b70 <print_metric_std>, new_line = 0x5555557be5a0 <new_line_std>, print_metricgroup_header = 0x5555557bea60 <print_metricgroup_header_std>, force_header = false} pm = <optimized out>
nl = <optimized out>
pmh = <optimized out>
ok = <optimized out>
counter = 0x555556093e20
#5 0x00005555557c1424 in print_counter_aggrdata (config=0x55555603bd80 <stat_config>, counter=0x555556093e20, aggr_idx=0, os=<optimized out>) at util/stat-display.c:1013
output = 0x7ffff69f24e0 <_IO_2_1_stderr_>
ena = 1001660763
run = <optimized out>
val = <optimized out>
uval = <optimized out>
ps = <optimized out>
aggr = <optimized out>
id = {thread_idx = -1, node = -1, socket = -1, die = -1, cluster = -1, cache_lvl = -1, cache = -1, core = -1, cpu = {cpu = 0}}
avg = <optimized out>
metric_only = false
...
whereas on
$ git describe v6.18-rc1-115-gd702c0f4af6e0
$ perf stat -C11 sleep 1
Performance counter stats for 'CPU(s) 11':
8 context-switches # 8.0 cs/sec cs_per_second
1,001.48 msec cpu-clock # 1.0 CPUs CPUs_utilized
0 cpu-migrations # 0.0 migrations/sec migrations_per_second
3 page-faults # 3.0 faults/sec page_faults_per_second
277,825 branch-misses # 0.3 % branch_miss_rate (33.11%)
102,546,153 branches # 102.4 M/sec branch_frequency (33.51%)
2,501,242,351 cpu-cycles # 2.5 GHz cycles_frequency (33.55%)
317,348,809 instructions # 0.1 instructions insn_per_cycle (33.55%)1.001449562 seconds time elapsed
it works without a problem.
Greetings,
Andres Freund
Ian Rogers
reviewerRe: [PATCH v4 07/10] perf tool_pmu: More accurately set the cpus for tool events
Quoted 140 lines · Andres Freund
Hi,
Quoted 9 lines · Ian Rogers
The user and system time events can record on different CPUs, but for all other events a single CPU map of just CPU 0 makes sense. In parse-events detect a tool PMU and then pass the
perf_event_attrso that thetool_pmucan return CPUs specific for the event. This avoids a CPU map of all online CPUs being used for events likeduration_time. Avoiding this avoids the evlist CPUs containing CPUs for whichduration_timejust gives 0. Minimizing the evlist CPUs can remove unnecessarysched_setaffinitysyscalls that delay metric calculations.I was just testing v6.19-rc* and noticed that
perf stat -C $somecpu sleep 1
segfaults.I bisected that down to this change (
d8d8a0b3603a9a8fa207cf9e4f292e81dc5d1008).$ git describe v6.18-rc1-116-gd8d8a0b3603a9$ gdb --args perf stat -C11 sleep 1 (gdb) r Starting program: /home/andres/bin/bin/perf stat -C11 sleep 1 [Thread debugging using libthread_db enabled] Using host libthread_db library "/usr/lib/x86_64-linux-gnu/libthread_db.so.1". [Detaching after fork from child process 843870]Performance counter stats for 'CPU(s) 11':
16 context-switches # 16.0 cs/sec
cs_per_second
1,001.66 msec cpu-clock
Program received signal SIGSEGV, Segmentation fault.prepare_metric(config=config@entry=0x55555603bd80 <stat_config>, mexp=mexp@entry=0x5555560a40f0, evsel=evsel@entry=0x555556093e20,pctx=pctx@entry=0x5555560a7150, aggr_idx=aggr_idx@entry=0) at util/stat-shadow.c:85 85 aggr = &ps->aggr[is_tool_time ? tool_aggr_idx : aggr_idx]; (gdb) bt #0 prepare_metric (config=config@entry=0x55555603bd80 <stat_config>, mexp=mexp@entry=0x5555560a40f0, evsel=evsel@entry=0x555556093e20, pctx=pctx@entry=0x5555560a7150, aggr_idx=aggr_idx@entry=0) at util/stat-shadow.c:85 #1 0x00005555557be0eb in generic_metric (config=<optimized out>, mexp=0x5555560a40f0, evsel=<optimized out>, aggr_idx=<optimized out>, out=0x7fffffff76c0) at util/stat-shadow.c:146 #2 perf_stat__print_shadow_stats_metricgroup (config=config@entry=0x55555603bd80 <stat_config>, evsel=<optimized out>, aggr_idx=0, num=num@entry=0x7fffffff7604, from=from@entry=0x0, out=0x7fffffff76c0) at util/stat-shadow.c:307 #3 0x00005555557be4c7 in perf_stat__print_shadow_stats (config=config@entry=0x55555603bd80 <stat_config>, evsel=evsel@entry=0x555556093e20, aggr_idx=aggr_idx@entry=0, out=out@entry=0x7fffffff76c0) at util/stat-shadow.c:325 #4 0x00005555557c03a3 in printout (config=<optimized out>, os=0x7fffffff78b0, uval=<optimized out>, run=1001660763, ena=<optimized out>, noise=<optimized out>, aggr_idx=0) at util/stat-display.c:874 #5 0x00005555557c1424 in print_counter_aggrdata (config=0x55555603bd80 <stat_config>, counter=0x555556093e20, aggr_idx=0, os=<optimized out>) at util/stat-display.c:1013 #6 0x00005555557c3051 in print_counter (config=<optimized out>, counter=<optimized out>, os=<optimized out>) at util/stat-display.c:1127 #7 print_counter (config=<optimized out>, counter=<optimized out>, os=<optimized out>) at util/stat-display.c:1117 #8 evlist__print_counters (evlist=<optimized out>, config=config@entry=0x55555603bd80 <stat_config>, _target=_target@entry=0x5555560412e0 <target>, ts=ts@entry=0x0, argc=argc@entry=2, argv=argv@entry=0x7fffffffdb20) at util/stat-display.c:1600#9
0x00005555555d0d27inprint_counters(ts=0x0, argc=2, argv=0x7fffffffdb20) at builtin-stat.c:1070
#10print_counters(ts=0x0, argc=2, argv=0x7fffffffdb20) at builtin-stat.c:1062
#11cmd_stat(argc=2, argv=0x7fffffffdb20) at builtin-stat.c:2949
#120x000055555562fee2inrun_builtin(p=p@entry=0x55555602df48 <commands+360>, argc=argc@entry=4, argv=argv@entry=0x7fffffffdb20) at perf.c:349 #130x00005555556301ceinhandle_internal_command(argc=argc@entry=4, argv=argv@entry=0x7fffffffdb20) at perf.c:401
#140x00005555555a8d33inrun_argv(argcp=<synthetic pointer>, argv=<synthetic pointer>) at perf.c:445
#15 main (argc=<optimized out>, argv=0x7fffffffdb20) at perf.c:553(gdb) bt full
#0prepare_metric(config=config@entry=0x55555603bd80 <stat_config>, mexp=mexp@entry=0x5555560a40f0, evsel=evsel@entry=0x555556093e20, pctx=pctx@entry=0x5555560a7150, aggr_idx=aggr_idx@entry=0) atutil/stat-shadow.c:85
val = <optimized out>
source_count= 0
tool_aggr_idx= 0
is_tool_time= true
ps =0x0
aggr = <optimized out>
n = <optimized out>
metric_events= <optimized out>
metric_refs=0x0
i = 1
#10x00005555557be0ebingeneric_metric(config=<optimized out>, mexp=0x5555560a40f0, evsel=<optimized out>, aggr_idx=<optimized out>, out=0x7fffffff76c0) atutil/stat-shadow.c:146print_metric = 0x5555557c1b70 <print_metric_std> metric_name = 0x5555560ae550 "CPUs_utilized" metric_expr = 0x555555ea5bb5 "(software@cpu\\-clock\\,name\\=cpu\\-clock@ if #target_cpu else software@task\\-clock\\,name\\=task\\-clock@) / (duration_time * 1e9)" metric_threshold = 0x0 metric_unit = 0x555555ea5c4d "1CPUs" metric_events = 0x5555560ae490 runtime = 0 pctx = 0x5555560a7150 ratio = 15.973479072960373 scale = 1 threshold = 4.9406564584124654e-324 i = <optimized out> ctxp = 0x7fffffff78b0 thresh = METRIC_THRESHOLD_UNKNOWN #2 perf_stat__print_shadow_stats_metricgroup (config=config@entry=0x55555603bd80 <stat_config>, evsel=<optimized out>, aggr_idx=0, num=num@entry=0x7fffffff7604, from=from@entry=0x0, out=0x7fffffff76c0) at util/stat-shadow.c:307 me = 0x5555560877d0 mexp = 0x5555560a40f0 ctxp = 0x7fffffff78b0 header_printed = false name = 0x555555879b28 "" metric_events = <optimized out> #3 0x00005555557be4c7 in perf_stat__print_shadow_stats (config=config@entry=0x55555603bd80 <stat_config>, evsel=evsel@entry=0x555556093e20, aggr_idx=aggr_idx@entry=0, out=out@entry=0x7fffffff76c0) at util/stat-shadow.c:325 print_metric = 0x5555557c1b70 <print_metric_std> ctxp = 0x7fffffff78b0 num = 1#4
0x00005555557c03a3in printout (config=<optimized out>, os=0x7fffffff78b0, uval=<optimized out>, run=1001660763, ena=<optimized out>, noise=<optimized out>, aggr_idx=0) atutil/stat-display.c:874
out = {ctx =0x7fffffff78b0,print_metric=0x5555557c1b70<print_metric_std>,new_line=0x5555557be5a0<new_line_std>,print_metricgroup_header=0x5555557bea60<print_metricgroup_header_std>,force_header= false} pm = <optimized out>
nl = <optimized out>
pmh = <optimized out>
ok = <optimized out>
counter =0x555556093e20
#50x00005555557c1424inprint_counter_aggrdata(config=0x55555603bd80 <stat_config>, counter=0x555556093e20, aggr_idx=0, os=<optimized out>) atutil/stat-display.c:1013
output =0x7ffff69f24e0<_IO_2_1_stderr_>
ena = 1001660763
run = <optimized out>
val = <optimized out>
uval = <optimized out>
ps = <optimized out>
aggr = <optimized out>
id = {thread_idx= -1, node = -1, socket = -1, die = -1, cluster = -1,cache_lvl= -1, cache = -1, core = -1, cpu = {cpu = 0}}
avg = <optimized out>
metric_only= false
...whereas on
$ git describe v6.18-rc1-115-gd702c0f4af6e0$ perf stat -C11 sleep 1Performance counter stats for 'CPU(s) 11':
8 context-switches # 8.0 cs/sec cs_per_second 1,001.48 msec cpu-clock # 1.0 CPUs CPUs_utilized 0 cpu-migrations # 0.0 migrations/sec migrations_per_second 3 page-faults # 3.0 faults/sec page_faults_per_second 277,825 branch-misses # 0.3 % branch_miss_rate (33.11%) 102,546,153 branches # 102.4 M/sec branch_frequency (33.51%) 2,501,242,351 cpu-cycles # 2.5 GHz cycles_frequency (33.55%) 317,348,809 instructions # 0.1 instructions insn_per_cycle (33.55%)
Quoted 3 lines
1.001449562 seconds time elapsedit works without a problem.
Thanks for the bug report! There were changes in v6.19 to change how metrics were computed hence the fault in the metric computation. The bisected patch is at fault, it aims to compute the CPU maps for tool events in a way to avoid unnecessary sched setaffinity calls. By having a smaller CPU map it set up a situation with user CPUs where the tool events could have no CPUs, removed from the list of events and the later seg fault occurred. I sent out a new patch series:
https://lore.kernel.org/linux-perf-users/[email protected]/
that has a revert of this patch, a fix to correctly avoid the segfault in the prepare_metric function, and a new patch to properly reduce the CPU maps of the tool events. The remaining affinity reduction patches are still part of the series and I believe ready to land.
Thanks,
Ian
Quoted 3 lines
Greetings,
Andres Freund
Re: [PATCH v4 07/10] perf tool_pmu: More accurately set the cpus for tool events
Hi,
Quoted 7 lines · Ian Rogers
Thanks for the bug report! There were changes in v6.19 to change how metrics were computed hence the fault in the metric computation. The bisected patch is at fault, it aims to compute the CPU maps for tool events in a way to avoid unnecessary sched setaffinity calls. By having a smaller CPU map it set up a situation with user CPUs where the tool events could have no CPUs, removed from the list of events and the later seg fault occurred.
Thanks for the quick reply!
Quoted 6 lines
I sent out a new patch series:
https://lore.kernel.org/linux-perf-users/[email protected]/
that has a revert of this patch, a fix to correctly avoid the segfault in theprepare_metricfunction, and a new patch to properly reduce the CPU maps of the tool events. The remaining affinity reduction patches are still part of the series and I believe ready to land.
That sounds like it's targeted for 6.20? I assume the revert on its own needs to land in 6.19 (or stable if too late for 6.19) somehow?
Greetings,
Andres Freund
Ian Rogers
reviewerRe: [PATCH v4 07/10] perf tool_pmu: More accurately set the cpus for tool events
Quoted 21 lines · Andres Freund
Hi,
Quoted 7 lines · Ian Rogers
Thanks for the bug report! There were changes in v6.19 to change how metrics were computed hence the fault in the metric computation. The bisected patch is at fault, it aims to compute the CPU maps for tool events in a way to avoid unnecessary sched setaffinity calls. By having a smaller CPU map it set up a situation with user CPUs where the tool events could have no CPUs, removed from the list of events and the later seg fault occurred.
Thanks for the quick reply!
Quoted 6 lines
I sent out a new patch series:
https://lore.kernel.org/linux-perf-users/[email protected]/
that has a revert of this patch, a fix to correctly avoid the segfault in theprepare_metricfunction, and a new patch to properly reduce the CPU maps of the tool events. The remaining affinity reduction patches are still part of the series and I believe ready to land.
Quoted 2 lines
That sounds like it's targeted for 6.20? I assume the revert on its own needs to land in 6.19 (or stable if too late for 6.19) somehow?
Yeah, the maintainers are the ones creating the pull requests and can do this.
Thanks,
Ian
Quoted 3 lines
Greetings,
Andres Freund
Ian Rogers
reviewerPATCH v4 8/10 perf evlist: Reduce affinity use and move into iterator, fix no affinity
In reply to Ian RogersThe evlist__for_each_cpu iterator will call sched_setaffitinity when moving between CPUs to avoid IPIs. If only 1 IPI is saved then this may be unprofitable as the delay to get scheduled may be considerable. This may be particularly true if reading an event group in perf stat in interval mode.
Move the affinity handling completely into the iterator so that a single evlist__use_affinity can determine whether CPU affinities will be used. For perf record the change is minimal as the dummy event and the real event will always make the use of affinities the thing to do. In perf stat, tool events are ignored and affinities only used if >1 event on the same CPU occur. Determining if affinities are useful is done by per-event in a new PMU benefits from affinity function.
Fix a bug where when there are no affinities that the CPU map iterator may reference a CPU not present in the initial evsel. Fix by making the iterator and non-iterator code common.
Fix a bug where closing events on an evlist wasn't closing TPEBS events.
--- tools/perf/builtin-stat.c | 128 ++++++++++++++---------------- tools/perf/util/evlist.c | 160 ++++++++++++++++++++++++-------------- tools/perf/util/evlist.h | 26 +++++-- tools/perf/util/pmu.c | 12 +++ tools/perf/util/pmu.h | 1 + 5 files changed, 189 insertions(+), 138 deletions(-)
tools/perf/builtin-stat.c +57 -71
@@ -366,22 +366,14 @@ static int read_counter_cpu(struct evsel *counter, int cpu_map_idx)return 0 ; } -static int read_affinity_counters (void ) +static int read_counters_with_affinity (void ) {struct evlist_cpu_iterator evlist_cpu_itr; -struct affinity saved_affinity, *affinity;if (all_counters_use_bpf)return 0 ; -if (!target__has_cpu (&target) ||target__has_per_thread (&target)) - affinity =NULL ; -else if (affinity__setup (&saved_affinity) <0 ) -return -1 ; -else - affinity = &saved_affinity; - -evlist__for_each_cpu (evlist_cpu_itr, evsel_list, affinity) { +evlist__for_each_cpu (evlist_cpu_itr, evsel_list) {struct evsel *counter = evlist_cpu_itr.evsel;if (evsel__is_bpf (counter))
@@ -390,8 +382,6 @@ static int read_affinity_counters(void)if (!counter->err) counter->err =read_counter_cpu (counter, evlist_cpu_itr.cpu_map_idx); } -if (affinity) -affinity__cleanup (&saved_affinity);return 0 ; }
@@ -414,12 +404,18 @@ static int read_bpf_map_counters(void)static int read_counters (void ) { -if (!stat_config.stop_read_counter) { -if (read_bpf_map_counters () || -read_affinity_counters ()) -return -1 ; - } -return 0 ; +int ret; + +if (stat_config.stop_read_counter) +return 0 ; + +// Read all BPF counters first. + ret =read_bpf_map_counters (); +if (ret) +return ret; + +// Read non-BPF and non-tool counters next. +return read_counters_with_affinity (); }static void process_counters (void )
@@ -760,7 +756,6 @@ static int __run_perf_stat(int argc, const char **argv, int run_idx)const bool forks = (argc >0 );bool is_pipe =STAT_RECORD ? perf_stat.data.is_pipe : false;struct evlist_cpu_iterator evlist_cpu_itr; -struct affinity saved_affinity, *affinity =NULL ;int err, open_err =0 ;bool second_pass = false, has_supported_counters;
@@ -772,14 +767,6 @@ static int __run_perf_stat(int argc, const char **argv, int run_idx) child_pid = evsel_list->workload.pid; } -if (!cpu_map__is_dummy (evsel_list->core.user_requested_cpus)) { -if (affinity__setup (&saved_affinity) <0 ) { - err =-1 ; - goto err_out; - } - affinity = &saved_affinity; - } -evlist__for_each_entry (evsel_list, counter) { counter->reset_group = false;if (bpf_counter__load (counter, &target)) {
@@ -792,49 +779,48 @@ static int __run_perf_stat(int argc, const char **argv, int run_idx)evlist__reset_aggr_stats (evsel_list); -evlist__for_each_cpu (evlist_cpu_itr, evsel_list, affinity) { - counter = evlist_cpu_itr.evsel; +/* +* bperf calls evsel__open_per_cpu() in bperf__load(), so +* no need to call it again here. +*/ +if (!target.use_bpf) { +evlist__for_each_cpu (evlist_cpu_itr, evsel_list) { + counter = evlist_cpu_itr.evsel; -/* -* bperf calls evsel__open_per_cpu() in bperf__load(), so -* no need to call it again here. -*/ -if (target.use_bpf) -break ; +if (counter->reset_group || !counter->supported) +continue ; +if (evsel__is_bperf (counter)) +continue ; -if (counter->reset_group || !counter->supported) -continue ; -if (evsel__is_bperf (counter)) -continue ; +while (true) { +if (create_perf_stat_counter (counter, &stat_config, + evlist_cpu_itr.cpu_map_idx) ==0 ) +break ; -while (true) { -if (create_perf_stat_counter (counter, &stat_config, - evlist_cpu_itr.cpu_map_idx) ==0 ) -break ; + open_err = errno; +/* +* Weak group failed. We cannot just undo this +* here because earlier CPUs might be in group +* mode, and the kernel doesn't support mixing +* group and non group reads. Defer it to later. +* Don't close here because we're in the wrong +* affinity. +*/ +if ((open_err ==EINVAL || open_err ==EBADF ) && +evsel__leader (counter) != counter && + counter->weak_group) { +evlist__reset_weak_group (evsel_list, counter, false); +assert (counter->reset_group); + counter->supported = true; + second_pass = true; +break ; + } - open_err = errno; -/* -* Weak group failed. We cannot just undo this here -* because earlier CPUs might be in group mode, and the kernel -* doesn't support mixing group and non group reads. Defer -* it to later. -* Don't close here because we're in the wrong affinity. -*/ -if ((open_err ==EINVAL || open_err ==EBADF ) && -evsel__leader (counter) != counter && - counter->weak_group) { -evlist__reset_weak_group (evsel_list, counter, false); -assert (counter->reset_group); - counter->supported = true; - second_pass = true; -break ; +if (stat_handle_error (counter, open_err) !=COUNTER_RETRY ) +break ; } - -if (stat_handle_error (counter, open_err) !=COUNTER_RETRY ) -break ; } } -if (second_pass) {/* * Now redo all the weak group after closing them,
@@ -842,7 +828,7 @@ static int __run_perf_stat(int argc, const char **argv, int run_idx)*/ /* First close errored or weak retry */ -evlist__for_each_cpu (evlist_cpu_itr, evsel_list, affinity) { +evlist__for_each_cpu (evlist_cpu_itr, evsel_list) { counter = evlist_cpu_itr.evsel;if (!counter->reset_group && counter->supported)
@@ -851,7 +837,7 @@ static int __run_perf_stat(int argc, const char **argv, int run_idx)perf_evsel__close_cpu (&counter->core, evlist_cpu_itr.cpu_map_idx); }/* Now reopen weak */ -evlist__for_each_cpu (evlist_cpu_itr, evsel_list, affinity) { +evlist__for_each_cpu (evlist_cpu_itr, evsel_list) { counter = evlist_cpu_itr.evsel;if (!counter->reset_group)
@@ -860,17 +846,18 @@ static int __run_perf_stat(int argc, const char **argv, int run_idx)while (true) {pr_debug2 ("reopening weak %s\n" ,evsel__name (counter));if (create_perf_stat_counter (counter, &stat_config, - evlist_cpu_itr.cpu_map_idx) ==0 ) + evlist_cpu_itr.cpu_map_idx) ==0 ) { +evlist_cpu_iterator__exit (&evlist_cpu_itr);break ; - + } open_err = errno; -if (stat_handle_error (counter, open_err) !=COUNTER_RETRY ) +if (stat_handle_error (counter, open_err) !=COUNTER_RETRY ) { +evlist_cpu_iterator__exit (&evlist_cpu_itr);break ; + } } } } -affinity__cleanup (affinity); - affinity =NULL ; has_supported_counters = false;evlist__for_each_entry (evsel_list, counter) {
@@ -1021,7 +1008,6 @@ static int __run_perf_stat(int argc, const char **argv, int run_idx)if (forks)evlist__cancel_workload (evsel_list); -affinity__cleanup (affinity);return err; }
tools/perf/util/evlist.c +100 -60
227 lines not on this page · raw message
tools/perf/util/evlist.h +19 -7
@@ -10,6 +10,7 @@#include <internal/evlist.h> #include <internal/evsel.h> #include <perf/evlist.h> +#include "affinity.h" #include "events_stats.h" #include "evsel.h" #include "rblist.h"
@@ -361,6 +362,8 @@ struct evlist_cpu_iterator {struct perf_cpu cpu;/** If present, used to set the affinity when switching between CPUs. */ struct affinity *affinity; +/** Maybe be used to hold affinity state prior to iterating. */ +struct affinity saved_affinity; };/**
@@ -368,22 +371,31 @@ struct evlist_cpu_iterator {* affinity, iterate over all CPUs and then the evlist * for each evsel on that CPU. When switching between * CPUs the affinity is set to the CPU to avoid IPIs -* during syscalls. +* during syscalls. The affinity is set up and removed +* automatically, if the loop is broken a call to +* evlist_cpu_iterator__exit is necessary. * @evlist_cpu_itr: the iterator instance. * @evlist: evlist instance to iterate. -* @affinity: NULL or used to set the affinity to the current CPU. */ -#define evlist__for_each_cpu (evlist_cpu_itr, evlist, affinity) \ - for ((evlist_cpu_itr) = evlist__cpu_begin(evlist, affinity); \ +#define evlist__for_each_cpu (evlist_cpu_itr, evlist) \ + for (evlist_cpu_iterator__init(&(evlist_cpu_itr), evlist); \ !evlist_cpu_iterator__end(&evlist_cpu_itr); \ evlist_cpu_iterator__next(&evlist_cpu_itr)) -/** Returns an iterator set to the first CPU/evsel of evlist. */ -struct evlist_cpu_iterator evlist__cpu_begin (struct evlist *evlist,struct affinity *affinity); +/** Setup an iterator set to the first CPU/evsel of evlist. */ +void evlist_cpu_iterator__init (struct evlist_cpu_iterator *itr,struct evlist *evlist); +/** +* Cleans up the iterator, automatically done by evlist_cpu_iterator__next when +* the end of the list is reached. Multiple calls are safe. +*/ +void evlist_cpu_iterator__exit (struct evlist_cpu_iterator *itr);/** Move to next element in iterator, updating CPU, evsel and the affinity. */ void evlist_cpu_iterator__next (struct evlist_cpu_iterator *evlist_cpu_itr);/** Returns true when iterator is at the end of the CPUs and evlist. */ -bool evlist_cpu_iterator__end (const struct evlist_cpu_iterator *evlist_cpu_itr); +static inline bool evlist_cpu_iterator__end (const struct evlist_cpu_iterator *evlist_cpu_itr) +{ +return evlist_cpu_itr->evlist_cpu_map_idx >= evlist_cpu_itr->evlist_cpu_map_nr; +}struct evsel *evlist__get_tracking_event (struct evlist *evlist);void evlist__set_tracking_event (struct evlist *evlist,struct evsel *tracking_evsel);
tools/perf/util/pmu.c +12 -0
@@ -2410,6 +2410,18 @@ bool perf_pmu__is_software(const struct perf_pmu *pmu)return false; } +bool perf_pmu__benefits_from_affinity (struct perf_pmu *pmu) +{ +if (!pmu) +return true;/* Assume is core. */ + +/* +* All perf event PMUs should benefit from accessing the perf event +* contexts on the local CPU. +*/ +return pmu->type <=PERF_PMU_TYPE_PE_END ; +} +FILE *perf_pmu__open_file (const struct perf_pmu *pmu,const char *name) {char path[PATH_MAX ];
tools/perf/util/pmu.h +1 -0
@@ -259,6 +259,7 @@ bool perf_pmu__name_no_suffix_match(const struct perf_pmu *pmu, const char *to_m* perf_sw_context in the kernel? */ bool perf_pmu__is_software (const struct perf_pmu *pmu); +bool perf_pmu__benefits_from_affinity (struct perf_pmu *pmu);FILE *perf_pmu__open_file (const struct perf_pmu *pmu,const char *name);FILE *perf_pmu__open_file_at (const struct perf_pmu *pmu,int dirfd,const char *name);
Signature · 1 line
2.51.2.1041.gc1ab5b90ca-goog
Ian Rogers
reviewerPATCH v4 9/10 perf stat: Read tool events last
In reply to Ian RogersWhen reading a metric like memory bandwidth on multiple sockets, the additional sockets will be on CPUS > 0. Because of the affinity reading, the counters are read on CPU 0 along with the time, then the later sockets are read. This can lead to the later sockets having a bandwidth larger than is possible for the period of time. To avoid this moving the reading of tool events to occur after all other events are read.
--- tools/perf/builtin-stat.c | 29 ++++++++++++++++++++++++++++- tools/perf/util/evlist.c | 4 ---- 2 files changed, 28 insertions(+), 5 deletions(-)
tools/perf/builtin-stat.c +28 -1
@@ -379,6 +379,9 @@ static int read_counters_with_affinity(void)if (evsel__is_bpf (counter))continue ; +if (evsel__is_tool (counter)) +continue ; +if (!counter->err) counter->err =read_counter_cpu (counter, evlist_cpu_itr.cpu_map_idx); }
@@ -402,6 +405,24 @@ static int read_bpf_map_counters(void)return 0 ; } +static int read_tool_counters (void ) +{ +struct evsel *counter; + +evlist__for_each_entry (evsel_list, counter) { +int idx; + +if (!evsel__is_tool (counter)) +continue ; + +perf_cpu_map__for_each_idx (idx, counter->core.cpus) { +if (!counter->err) + counter->err =read_counter_cpu (counter, idx); + } + } +return 0 ; +} +static int read_counters (void ) {int ret;
@@ -415,7 +436,13 @@ static int read_counters(void)return ret;// Read non-BPF and non-tool counters next. -return read_counters_with_affinity (); + ret =read_counters_with_affinity (); +if (ret) +return ret; + +// Read the tool counters last. This way the duration_time counter +// should always be greater than any other counter's enabled time. +return read_tool_counters (); }static void process_counters (void )
tools/perf/util/evlist.c +0 -4
@@ -368,10 +368,6 @@ static bool evlist__use_affinity(struct evlist *evlist)struct perf_cpu_map *used_cpus =NULL ;bool ret = false; -/* -* With perf record core.user_requested_cpus is usually NULL. -* Use the old method to handle this for now. -*/ if (!evlist->core.user_requested_cpus ||cpu_map__is_dummy (evlist->core.user_requested_cpus))return false;
Signature · 1 line
2.51.2.1041.gc1ab5b90ca-goog
Namhyung Kim
maintainerRe: [PATCH v4 09/10] perf stat: Read tool events last
Quoted 7 lines · Ian Rogers
When reading a metric like memory bandwidth on multiple sockets, the additional sockets will be on CPUS > 0. Because of the affinity reading, the counters are read on CPU 0 along with the time, then the later sockets are read. This can lead to the later sockets having a bandwidth larger than is possible for the period of time. To avoid this moving the reading of tool events to occur after all other events are read.
Can you move this change before the affinity updates? I think it's straight-forward and can be applied independently.
Thanks,
Namhyung
Quoted 79 lines
Signed-off-by Ian Rogers <[email protected]>--- tools/perf/builtin-stat.c | 29 ++++++++++++++++++++++++++++- tools/perf/util/evlist.c | 4 ---- 2 files changed, 28 insertions(+), 5 deletions(-)tools/perf/builtin-stat.c +28
-1@@ -379,6 +379,9 @@ static int read_counters_with_affinity(void)if (evsel__is_bpf (counter))continue ; +if (evsel__is_tool (counter)) +continue ; +if (!counter->err) counter->err =read_counter_cpu (counter, evlist_cpu_itr.cpu_map_idx); }@@ -402,6 +405,24 @@ static int read_bpf_map_counters(void)return 0 ; } +static int read_tool_counters (void ) +{ +struct evsel *counter; + +evlist__for_each_entry (evsel_list, counter) { +int idx; + +if (!evsel__is_tool (counter)) +continue ; + +perf_cpu_map__for_each_idx (idx, counter->core.cpus) { +if (!counter->err) + counter->err =read_counter_cpu (counter, idx); + } + } +return 0 ; +} +static int read_counters (void ) {int ret;@@ -415,7 +436,13 @@ static int read_counters(void)return ret;// Read non-BPF and non-tool counters next. -return read_counters_with_affinity (); + ret =read_counters_with_affinity (); +if (ret) +return ret; + +// Read the tool counters last. This way the duration_time counter +// should always be greater than any other counter's enabled time. +return read_tool_counters (); }static void process_counters (void )tools/perf/util/evlist.c +0
-4@@ -368,10 +368,6 @@ static bool evlist__use_affinity(struct evlist *evlist)struct perf_cpu_map *used_cpus =NULL ;bool ret = false; -/* -* With perf record core.user_requested_cpus is usually NULL. -* Use the old method to handle this for now. -*/ if (!evlist->core.user_requested_cpus ||cpu_map__is_dummy (evlist->core.user_requested_cpus))return false;Signature · 1 line
2.51.2.1041.gc1ab5b90ca-goog
Ian Rogers
reviewerRe: [PATCH v4 09/10] perf stat: Read tool events last
Quoted 10 lines · Namhyung Kim
Quoted 7 lines · Ian Rogers
When reading a metric like memory bandwidth on multiple sockets, the additional sockets will be on CPUS > 0. Because of the affinity reading, the counters are read on CPU 0 along with the time, then the later sockets are read. This can lead to the later sockets having a bandwidth larger than is possible for the period of time. To avoid this moving the reading of tool events to occur after all other events are read.
Quoted 2 lines
Can you move this change before the affinity updates? I think it's straight-forward and can be applied independently.
It is straightforward but will require changes to read_counters_with_affinity. I can do a v5 once I know what to do with the other patches.
Thanks,
Ian
Quoted 82 lines
Thanks,
NamhyungQuoted 79 lines
Signed-off-by Ian Rogers <[email protected]>--- tools/perf/builtin-stat.c | 29 ++++++++++++++++++++++++++++- tools/perf/util/evlist.c | 4 ---- 2 files changed, 28 insertions(+), 5 deletions(-)tools/perf/builtin-stat.c +28
-1@@ -379,6 +379,9 @@ static int read_counters_with_affinity(void)if (evsel__is_bpf (counter))continue ; +if (evsel__is_tool (counter)) +continue ; +if (!counter->err) counter->err =read_counter_cpu (counter, evlist_cpu_itr.cpu_map_idx); }@@ -402,6 +405,24 @@ static int read_bpf_map_counters(void)return 0 ; } +static int read_tool_counters (void ) +{ +struct evsel *counter; + +evlist__for_each_entry (evsel_list, counter) { +int idx; + +if (!evsel__is_tool (counter)) +continue ; + +perf_cpu_map__for_each_idx (idx, counter->core.cpus) { +if (!counter->err) + counter->err =read_counter_cpu (counter, idx); + } + } +return 0 ; +} +static int read_counters (void ) {int ret;@@ -415,7 +436,13 @@ static int read_counters(void)return ret;// Read non-BPF and non-tool counters next. -return read_counters_with_affinity (); + ret =read_counters_with_affinity (); +if (ret) +return ret; + +// Read the tool counters last. This way the duration_time counter +// should always be greater than any other counter's enabled time. +return read_tool_counters (); }static void process_counters (void )tools/perf/util/evlist.c +0
-5@@ -368,10 +368,6 @@ static bool evlist__use_affinity(struct evlist *evlist)struct perf_cpu_map *used_cpus =NULL ;bool ret = false; -/* -* With perf record core.user_requested_cpus is usually NULL. -* Use the old method to handle this for now. -*/ if (!evlist->core.user_requested_cpus ||cpu_map__is_dummy (evlist->core.user_requested_cpus))return false; --2.51.2.1041.gc1ab5b90ca-goog
Namhyung Kim
maintainerRe: [PATCH v4 09/10] perf stat: Read tool events last
Quoted 11 lines · Ian Rogers
Quoted 10 lines · Namhyung Kim
Quoted 7 lines · Ian Rogers
When reading a metric like memory bandwidth on multiple sockets, the additional sockets will be on CPUS > 0. Because of the affinity reading, the counters are read on CPU 0 along with the time, then the later sockets are read. This can lead to the later sockets having a bandwidth larger than is possible for the period of time. To avoid this moving the reading of tool events to occur after all other events are read.
Quoted 6 lines
Quoted 2 lines
Can you move this change before the affinity updates? I think it's straight-forward and can be applied independently.
It is straightforward but will require changes to
read_counters_with_affinity. I can do a v5 once I know what to do with the other patches.
I plan to merge patch 1 to 7 first. And I think it'd good to have this one as well.
Thanks,
Namhyung
Quoted 80 lines
Quoted 80 lines
Quoted 80 lines
Signed-off-by Ian Rogers <[email protected]>--- tools/perf/builtin-stat.c | 29 ++++++++++++++++++++++++++++- tools/perf/util/evlist.c | 4 ---- 2 files changed, 28 insertions(+), 5 deletions(-)tools/perf/builtin-stat.c +28
-1@@ -379,6 +379,9 @@ static int read_counters_with_affinity(void)if (evsel__is_bpf (counter))continue ; +if (evsel__is_tool (counter)) +continue ; +if (!counter->err) counter->err =read_counter_cpu (counter, evlist_cpu_itr.cpu_map_idx); }@@ -402,6 +405,24 @@ static int read_bpf_map_counters(void)return 0 ; } +static int read_tool_counters (void ) +{ +struct evsel *counter; + +evlist__for_each_entry (evsel_list, counter) { +int idx; + +if (!evsel__is_tool (counter)) +continue ; + +perf_cpu_map__for_each_idx (idx, counter->core.cpus) { +if (!counter->err) + counter->err =read_counter_cpu (counter, idx); + } + } +return 0 ; +} +static int read_counters (void ) {int ret;@@ -415,7 +436,13 @@ static int read_counters(void)return ret;// Read non-BPF and non-tool counters next. -return read_counters_with_affinity (); + ret =read_counters_with_affinity (); +if (ret) +return ret; + +// Read the tool counters last. This way the duration_time counter +// should always be greater than any other counter's enabled time. +return read_tool_counters (); }static void process_counters (void )tools/perf/util/evlist.c +0
-5@@ -368,10 +368,6 @@ static bool evlist__use_affinity(struct evlist *evlist)struct perf_cpu_map *used_cpus =NULL ;bool ret = false; -/* -* With perf record core.user_requested_cpus is usually NULL. -* Use the old method to handle this for now. -*/ if (!evlist->core.user_requested_cpus ||cpu_map__is_dummy (evlist->core.user_requested_cpus))return false; --2.51.2.1041.gc1ab5b90ca-goog
Ian Rogers
reviewerPATCH v4 10/10 perf stat: Add no-affinity flag
In reply to Ian RogersAdd flag that disables affinity behavior. Using sched_setaffinity to place a perf thread on a CPU can avoid certain interprocessor interrupts but may introduce a delay due to the scheduling, particularly on loaded machines. Add a command line option to disable the behavior. This behavior is less present in other tools like perf record, as it uses a ring buffer and doesn't make repeated system calls.
--- tools/perf/Documentation/perf-stat.txt | 4 ++++ tools/perf/builtin-stat.c | 6 ++++++ tools/perf/util/evlist.c | 2 +- tools/perf/util/evlist.h | 1 + 4 files changed, 12 insertions(+), 1 deletion(-)
tools/perf/Documentation/perf-stat.txt +4 -0
@@ -382,6 +382,10 @@ color the metric's computed value. Don't print output, warnings or messages. This is useful with perf stat record below to only write data to the perf.data file. +--no-affinity:: +Don't change scheduler affinities when iterating over CPUs. Disables +an optimization aimed at minimizing interprocessor interrupts. + STAT RECORD ----------- Stores stat data into perf data file.
tools/perf/builtin-stat.c +6 -0
@@ -2415,6 +2415,7 @@ static int parse_tpebs_mode(const struct option *opt, const char *str,int cmd_stat (int argc,const char **argv) {struct opt_aggr_mode opt_mode = {}; +bool no_affinity = false;struct option stat_options[] = {OPT_BOOLEAN ('T' ,"transaction" , &transaction_run,"hardware transaction statistics" ),
@@ -2543,6 +2544,8 @@ int cmd_stat(int argc, const char **argv) "don't print 'summary' for CSV summary output"),OPT_BOOLEAN (0 ,"quiet" , &quiet,"don't print any output, messages or warnings (useful with record)" ), +OPT_BOOLEAN (0 ,"no-affinity" , &no_affinity, +"don't allow affinity optimizations aimed at reducing IPIs" ), OPT_CALLBACK(0, "cputype", &evsel_list, "hybrid cpu type", "Only enable events on applying cpu with this type " "for hybrid platform (e.g. core or atom)",
@@ -2600,6 +2603,9 @@ int cmd_stat(int argc, const char **argv) }else stat_config.csv_sep =DEFAULT_SEPARATOR ; +if (no_affinity) + evsel_list->no_affinity = true; +if (argc &&strlen (argv[0 ]) >2 &&strstarts ("record" , argv[0 ])) { argc =__cmd_record (stat_options, &opt_mode, argc, argv);if (argc <0 )
tools/perf/util/evlist.c +1 -1
@@ -368,7 +368,7 @@ static bool evlist__use_affinity(struct evlist *evlist)struct perf_cpu_map *used_cpus =NULL ;bool ret = false; -if (!evlist->core.user_requested_cpus || +if (evlist->no_affinity || !evlist->core.user_requested_cpus ||cpu_map__is_dummy (evlist->core.user_requested_cpus))return false;
tools/perf/util/evlist.h +1 -0
@@ -59,6 +59,7 @@ struct event_enable_timer;struct evlist {struct perf_evlist core;bool enabled; +bool no_affinity;int id_pos;int is_pos;int nr_br_cntr;
Signature · 1 line
2.51.2.1041.gc1ab5b90ca-goog
Namhyung Kim
maintainerRe: [PATCH v4 10/10] perf stat: Add no-affinity flag
Quoted 44 lines · Ian Rogers
Add flag that disables affinity behavior. Using
sched_setaffinityto place a perf thread on a CPU can avoid certain interprocessor interrupts but may introduce a delay due to the scheduling, particularly on loaded machines. Add a command line option to disable the behavior. This behavior is less present in other tools likeperf record, as it uses a ring buffer and doesn't make repeated system calls.Signed-off-by Ian Rogers <[email protected]>--- tools/perf/Documentation/perf-stat.txt | 4 ++++ tools/perf/builtin-stat.c | 6 ++++++ tools/perf/util/evlist.c | 2 +- tools/perf/util/evlist.h | 1 + 4 files changed, 12 insertions(+), 1 deletion(-)tools/perf/Documentation/perf-stat.txt +4
-0@@ -382,6 +382,10 @@ color the metric's computed value. Don't print output, warnings or messages. This is useful with perf stat record below to only write data to the perf.data file. +--no-affinity:: +Don't change scheduler affinities when iterating over CPUs. Disables +an optimization aimed at minimizing interprocessor interrupts. + STAT RECORD ----------- Stores stat data into perf data file.tools/perf/builtin-stat.c +1
-0@@ -2415,6 +2415,7 @@ static int parse_tpebs_mode(const struct option *opt, const char *str,int cmd_stat (int argc,const char **argv) {struct opt_aggr_mode opt_mode = {}; +bool no_affinity = false;struct option stat_options[] = {OPT_BOOLEAN ('T' ,"transaction" , &transaction_run,"hardware transaction statistics" ),
tools/perf/builtin-stat.c int cmd_stat(int argc, const char **argv) · line 2548
Quoted 5 lines
tools/perf/builtin-stat.c +2
-0@@ -2543,6 +2544,8 @@ int cmd_stat(int argc, const char **argv) "don't print 'summary' for CSV summary output"),OPT_BOOLEAN (0 ,"quiet" , &quiet,"don't print any output, messages or warnings (useful with record)" ), +OPT_BOOLEAN (0 ,"no-affinity" , &no_affinity, + "don't allow affinity optimizations aimed at reducing IPIs"),
I know you want to add an option to disable the behaivor, but I think it'd better to have a positive option like just '--affinity'. Then we will have '--no-affinity' for free. :) The current form will allow '--no-no-affinity'.
Then the variable also can be 'enable_affinity' or so.
You can mention --no-affinity in the help message and the man page document so that users can discover the intention.
Thanks,
Namhyung
Quoted 41 lines
tools/perf/builtin-stat.c +3
-0OPT_CALLBACK(0, "cputype", &evsel_list, "hybrid cpu type", "Only enable events on applying cpu with this type " "for hybrid platform (e.g. core or atom)",@@ -2600,6 +2603,9 @@ int cmd_stat(int argc, const char **argv) }else stat_config.csv_sep =DEFAULT_SEPARATOR ; +if (no_affinity) + evsel_list->no_affinity = true; +if (argc &&strlen (argv[0 ]) >2 &&strstarts ("record" , argv[0 ])) { argc =__cmd_record (stat_options, &opt_mode, argc, argv);if (argc <0 )tools/perf/util/evlist.c +1
-1@@ -368,7 +368,7 @@ static bool evlist__use_affinity(struct evlist *evlist)struct perf_cpu_map *used_cpus =NULL ;bool ret = false; -if (!evlist->core.user_requested_cpus || +if (evlist->no_affinity || !evlist->core.user_requested_cpus ||cpu_map__is_dummy (evlist->core.user_requested_cpus))return false;tools/perf/util/evlist.h +1
-0@@ -59,6 +59,7 @@ struct event_enable_timer;struct evlist {struct perf_evlist core;bool enabled; +bool no_affinity;int id_pos;int is_pos;int nr_br_cntr;Signature · 1 line
2.51.2.1041.gc1ab5b90ca-goog
Ian Rogers
reviewerRe: [PATCH v4 10/10] perf stat: Add no-affinity flag
Quoted 57 lines · Namhyung Kim
Quoted 44 lines · Ian Rogers
Add flag that disables affinity behavior. Using
sched_setaffinityto place a perf thread on a CPU can avoid certain interprocessor interrupts but may introduce a delay due to the scheduling, particularly on loaded machines. Add a command line option to disable the behavior. This behavior is less present in other tools likeperf record, as it uses a ring buffer and doesn't make repeated system calls.Signed-off-by Ian Rogers <[email protected]>--- tools/perf/Documentation/perf-stat.txt | 4 ++++ tools/perf/builtin-stat.c | 6 ++++++ tools/perf/util/evlist.c | 2 +- tools/perf/util/evlist.h | 1 + 4 files changed, 12 insertions(+), 1 deletion(-)tools/perf/Documentation/perf-stat.txt +4
-0@@ -382,6 +382,10 @@ color the metric's computed value. Don't print output, warnings or messages. This is useful with perf stat record below to only write data to the perf.data file. +--no-affinity:: +Don't change scheduler affinities when iterating over CPUs. Disables +an optimization aimed at minimizing interprocessor interrupts. + STAT RECORD ----------- Stores stat data into perf data file.tools/perf/builtin-stat.c +1
-0@@ -2415,6 +2415,7 @@ static int parse_tpebs_mode(const struct option *opt, const char *str,int cmd_stat (int argc,const char **argv) {struct opt_aggr_mode opt_mode = {}; +bool no_affinity = false;struct option stat_options[] = {OPT_BOOLEAN ('T' ,"transaction" , &transaction_run,"hardware transaction statistics" ),tools/perf/builtin-stat.c int cmd_stat(int argc, const char **argv) · line 2548
Quoted 5 lines
tools/perf/builtin-stat.c +2
-0@@ -2543,6 +2544,8 @@ int cmd_stat(int argc, const char **argv) "don't print 'summary' for CSV summary output"),OPT_BOOLEAN (0 ,"quiet" , &quiet,"don't print any output, messages or warnings (useful with record)" ), +OPT_BOOLEAN (0 ,"no-affinity" , &no_affinity, + "don't allow affinity optimizations aimed at reducing IPIs"),I know you want to add an option to disable the behaivor, but I think it'd better to have a positive option like just '--affinity'. Then we will have '--no-affinity' for free. :) The current form will allow '--no-no-affinity'.
Quoted 4 lines
Then the variable also can be '
enable_affinity' or so.You can mention --no-affinity in the help message and the man page document so that users can discover the intention.
I was trying to keep the code the same as the flag. I don't want to have an affinity flag you need to set to true in say evlist, as that will need updating in all users of the evlist or become a behavioral change. We can have the evlist with no_affinity and invert the flag, it just looks awkward imo.
Thanks,
Ian
Quoted 45 lines
Thanks,
NamhyungQuoted 41 lines
tools/perf/builtin-stat.c +3
-0OPT_CALLBACK(0, "cputype", &evsel_list, "hybrid cpu type", "Only enable events on applying cpu with this type " "for hybrid platform (e.g. core or atom)",@@ -2600,6 +2603,9 @@ int cmd_stat(int argc, const char **argv) }else stat_config.csv_sep =DEFAULT_SEPARATOR ; +if (no_affinity) + evsel_list->no_affinity = true; +if (argc &&strlen (argv[0 ]) >2 &&strstarts ("record" , argv[0 ])) { argc =__cmd_record (stat_options, &opt_mode, argc, argv);if (argc <0 )tools/perf/util/evlist.c +1
-1@@ -368,7 +368,7 @@ static bool evlist__use_affinity(struct evlist *evlist)struct perf_cpu_map *used_cpus =NULL ;bool ret = false; -if (!evlist->core.user_requested_cpus || +if (evlist->no_affinity || !evlist->core.user_requested_cpus ||cpu_map__is_dummy (evlist->core.user_requested_cpus))return false;tools/perf/util/evlist.h +1
-1@@ -59,6 +59,7 @@ struct event_enable_timer;struct evlist {struct perf_evlist core;bool enabled; +bool no_affinity;int id_pos;int is_pos;int nr_br_cntr; --2.51.2.1041.gc1ab5b90ca-goog
Namhyung Kim
maintainerRe: [PATCH v4 10/10] perf stat: Add no-affinity flag
Quoted 60 lines · Ian Rogers
Quoted 59 lines · Namhyung Kim
Quoted 44 lines · Ian Rogers
Add flag that disables affinity behavior. Using
sched_setaffinityto place a perf thread on a CPU can avoid certain interprocessor interrupts but may introduce a delay due to the scheduling, particularly on loaded machines. Add a command line option to disable the behavior. This behavior is less present in other tools likeperf record, as it uses a ring buffer and doesn't make repeated system calls.Signed-off-by Ian Rogers <[email protected]>--- tools/perf/Documentation/perf-stat.txt | 4 ++++ tools/perf/builtin-stat.c | 6 ++++++ tools/perf/util/evlist.c | 2 +- tools/perf/util/evlist.h | 1 + 4 files changed, 12 insertions(+), 1 deletion(-)tools/perf/Documentation/perf-stat.txt +4
-0@@ -382,6 +382,10 @@ color the metric's computed value. Don't print output, warnings or messages. This is useful with perf stat record below to only write data to the perf.data file. +--no-affinity:: +Don't change scheduler affinities when iterating over CPUs. Disables +an optimization aimed at minimizing interprocessor interrupts. + STAT RECORD ----------- Stores stat data into perf data file.tools/perf/builtin-stat.c +1
-0@@ -2415,6 +2415,7 @@ static int parse_tpebs_mode(const struct option *opt, const char *str,int cmd_stat (int argc,const char **argv) {struct opt_aggr_mode opt_mode = {}; +bool no_affinity = false;struct option stat_options[] = {OPT_BOOLEAN ('T' ,"transaction" , &transaction_run,"hardware transaction statistics" ),tools/perf/builtin-stat.c int cmd_stat(int argc, const char **argv) · line 2548
Quoted 5 lines
tools/perf/builtin-stat.c +2
-0@@ -2543,6 +2544,8 @@ int cmd_stat(int argc, const char **argv) "don't print 'summary' for CSV summary output"),OPT_BOOLEAN (0 ,"quiet" , &quiet,"don't print any output, messages or warnings (useful with record)" ), +OPT_BOOLEAN (0 ,"no-affinity" , &no_affinity, + "don't allow affinity optimizations aimed at reducing IPIs"),I know you want to add an option to disable the behaivor, but I think it'd better to have a positive option like just '--affinity'. Then we will have '--no-affinity' for free. :) The current form will allow '--no-no-affinity'.
Then the variable also can be '
enable_affinity' or so.
Quoted 8 lines
Quoted 2 lines
You can mention --no-affinity in the help message and the man page document so that users can discover the intention.
I was trying to keep the code the same as the flag. I don't want to have an affinity flag you need to set to true in say evlist, as that will need updating in all users of the evlist or become a behavioral change. We can have the evlist with
no_affinityand invert the flag, it just looks awkward imo.
You can make the default to true and just update the users that want different behaviors.
Thanks,
Namhyung
Namhyung Kim
maintainerQuoted 9 lines · Ian Rogers
A collection of fixes aiming to stabilize and make more reasonable measurements/metrics such as memory bandwidth.
Tool events are changed from getting a PMU cpu map of all online CPUs to either CPU 0 or all online CPUs. This avoids iterating over useless CPUs for events in particular
duration_time. Fix a bug whereduration_timedidn't correctly use the previous raw counts and would skip values in interval mode.
Quoted 5 lines
Change how json metrics handle tool events. Use the counter value rather than using shared state with perf stat. A later patch changes it so that tool events are read last, so that if reading say memory bandwidth counters you don't divide by an earlier read time and exceed the theoretical maximum memory bandwidth.
[...]
Applied the first 7 patches to perf-tools-next, thanks!
Best regards,
Namhyung