linux-stable/tools/perf/util/stat-display.c

1272 lines
31 KiB
C
Raw Normal View History

#include <stdlib.h>
#include <stdio.h>
#include <inttypes.h>
#include <linux/string.h>
#include <linux/time64.h>
#include <math.h>
#include "color.h"
#include "counts.h"
#include "evlist.h"
#include "evsel.h"
#include "stat.h"
#include "top.h"
#include "thread_map.h"
#include "cpumap.h"
#include "string2.h"
#include <linux/ctype.h>
#include "cgroup.h"
#include <api/fs/fs.h>
#include "util.h"
#define CNTR_NOT_SUPPORTED "<not supported>"
#define CNTR_NOT_COUNTED "<not counted>"
static void print_running(struct perf_stat_config *config,
u64 run, u64 ena)
{
if (config->csv_output) {
fprintf(config->output, "%s%" PRIu64 "%s%.2f",
config->csv_sep,
run,
config->csv_sep,
ena ? 100.0 * run / ena : 100.0);
} else if (run != ena) {
fprintf(config->output, " (%.2f%%)", 100.0 * run / ena);
}
}
static void print_noise_pct(struct perf_stat_config *config,
double total, double avg)
{
double pct = rel_stddev_stats(total, avg);
if (config->csv_output)
fprintf(config->output, "%s%.2f%%", config->csv_sep, pct);
else if (pct)
fprintf(config->output, " ( +-%6.2f%% )", pct);
}
static void print_noise(struct perf_stat_config *config,
struct evsel *evsel, double avg)
{
struct perf_stat_evsel *ps;
if (config->run_count == 1)
return;
ps = evsel->stats;
print_noise_pct(config, stddev_stats(&ps->res_stats[0]), avg);
}
static void print_cgroup(struct perf_stat_config *config, struct evsel *evsel)
{
if (nr_cgroups) {
const char *cgrp_name = evsel->cgrp ? evsel->cgrp->name : "";
fprintf(config->output, "%s%s", config->csv_sep, cgrp_name);
}
}
static void aggr_printout(struct perf_stat_config *config,
struct evsel *evsel, struct aggr_cpu_id id, int nr)
{
switch (config->aggr_mode) {
case AGGR_CORE:
fprintf(config->output, "S%d-D%d-C%*d%s%*d%s",
perf stat aggregation: Add separate socket member Add socket as a separate member so that it doesn't have to be packed into the int value. When the socket ID was larger than 8 bits the output appeared corrupted or incomplete. For example, here on ThunderX2 'perf stat' reports a socket of -1 and an invalid die number: ./perf stat -a --per-die The socket id number is too big. Performance counter stats for 'system wide': S-1-D255 128 687.99 msec cpu-clock # 57.240 CPUs utilized ... S36-D0 128 842.34 msec cpu-clock # 70.081 CPUs utilized ... And with --per-core there is an entry with an invalid core ID: ./perf stat record -a --per-core The socket id number is too big. Performance counter stats for 'system wide': S-1-D255-C65535 128 671.04 msec cpu-clock # 54.112 CPUs utilized ... S36-D0-C0 4 28.27 msec cpu-clock # 2.279 CPUs utilized ... This fixes the "Session topology" self test on ThunderX2. After this fix the output contains the correct socket and die IDs and no longer prints a warning about the size of the socket ID: ./perf stat --per-die -a Performance counter stats for 'system wide': S36-D0 128 169,869.39 msec cpu-clock # 127.501 CPUs utilized ... S3612-D0 128 169,733.05 msec cpu-clock # 127.398 CPUs utilized Signed-off-by: James Clark <james.clark@arm.com> Acked-by: Namhyung Kim <namhyung@kernel.org> Acked-by: Jiri Olsa <jolsa@redhat.com> Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com> Tested-by: John Garry <john.garry@huawei.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: Mark Rutland <mark.rutland@arm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Richter <tmricht@linux.ibm.com> Link: https://lore.kernel.org/r/20201126141328.6509-10-james.clark@arm.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2020-11-26 14:13:25 +00:00
id.socket,
id.die,
config->csv_output ? 0 : -8,
id.core,
config->csv_sep,
config->csv_output ? 0 : 4,
nr,
config->csv_sep);
break;
case AGGR_DIE:
fprintf(config->output, "S%d-D%*d%s%*d%s",
perf stat aggregation: Add separate socket member Add socket as a separate member so that it doesn't have to be packed into the int value. When the socket ID was larger than 8 bits the output appeared corrupted or incomplete. For example, here on ThunderX2 'perf stat' reports a socket of -1 and an invalid die number: ./perf stat -a --per-die The socket id number is too big. Performance counter stats for 'system wide': S-1-D255 128 687.99 msec cpu-clock # 57.240 CPUs utilized ... S36-D0 128 842.34 msec cpu-clock # 70.081 CPUs utilized ... And with --per-core there is an entry with an invalid core ID: ./perf stat record -a --per-core The socket id number is too big. Performance counter stats for 'system wide': S-1-D255-C65535 128 671.04 msec cpu-clock # 54.112 CPUs utilized ... S36-D0-C0 4 28.27 msec cpu-clock # 2.279 CPUs utilized ... This fixes the "Session topology" self test on ThunderX2. After this fix the output contains the correct socket and die IDs and no longer prints a warning about the size of the socket ID: ./perf stat --per-die -a Performance counter stats for 'system wide': S36-D0 128 169,869.39 msec cpu-clock # 127.501 CPUs utilized ... S3612-D0 128 169,733.05 msec cpu-clock # 127.398 CPUs utilized Signed-off-by: James Clark <james.clark@arm.com> Acked-by: Namhyung Kim <namhyung@kernel.org> Acked-by: Jiri Olsa <jolsa@redhat.com> Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com> Tested-by: John Garry <john.garry@huawei.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: Mark Rutland <mark.rutland@arm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Richter <tmricht@linux.ibm.com> Link: https://lore.kernel.org/r/20201126141328.6509-10-james.clark@arm.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2020-11-26 14:13:25 +00:00
id.socket,
config->csv_output ? 0 : -8,
id.die,
config->csv_sep,
config->csv_output ? 0 : 4,
nr,
config->csv_sep);
break;
case AGGR_SOCKET:
fprintf(config->output, "S%*d%s%*d%s",
config->csv_output ? 0 : -5,
perf stat aggregation: Add separate socket member Add socket as a separate member so that it doesn't have to be packed into the int value. When the socket ID was larger than 8 bits the output appeared corrupted or incomplete. For example, here on ThunderX2 'perf stat' reports a socket of -1 and an invalid die number: ./perf stat -a --per-die The socket id number is too big. Performance counter stats for 'system wide': S-1-D255 128 687.99 msec cpu-clock # 57.240 CPUs utilized ... S36-D0 128 842.34 msec cpu-clock # 70.081 CPUs utilized ... And with --per-core there is an entry with an invalid core ID: ./perf stat record -a --per-core The socket id number is too big. Performance counter stats for 'system wide': S-1-D255-C65535 128 671.04 msec cpu-clock # 54.112 CPUs utilized ... S36-D0-C0 4 28.27 msec cpu-clock # 2.279 CPUs utilized ... This fixes the "Session topology" self test on ThunderX2. After this fix the output contains the correct socket and die IDs and no longer prints a warning about the size of the socket ID: ./perf stat --per-die -a Performance counter stats for 'system wide': S36-D0 128 169,869.39 msec cpu-clock # 127.501 CPUs utilized ... S3612-D0 128 169,733.05 msec cpu-clock # 127.398 CPUs utilized Signed-off-by: James Clark <james.clark@arm.com> Acked-by: Namhyung Kim <namhyung@kernel.org> Acked-by: Jiri Olsa <jolsa@redhat.com> Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com> Tested-by: John Garry <john.garry@huawei.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: Mark Rutland <mark.rutland@arm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Richter <tmricht@linux.ibm.com> Link: https://lore.kernel.org/r/20201126141328.6509-10-james.clark@arm.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2020-11-26 14:13:25 +00:00
id.socket,
config->csv_sep,
config->csv_output ? 0 : 4,
nr,
config->csv_sep);
break;
perf stat: Add --per-node agregation support Adding new --per-node option to aggregate counts per NUMA nodes for system-wide mode measurements. You can specify --per-node in live mode: # perf stat -a -I 1000 -e cycles --per-node # time node cpus counts unit events 1.000542550 N0 20 6,202,097 cycles 1.000542550 N1 20 639,559 cycles 2.002040063 N0 20 7,412,495 cycles 2.002040063 N1 20 2,185,577 cycles 3.003451699 N0 20 6,508,917 cycles 3.003451699 N1 20 765,607 cycles ... Or in the record/report stat session: # perf stat record -a -I 1000 -e cycles # time counts unit events 1.000536937 10,008,468 cycles 2.002090152 9,578,539 cycles 3.003625233 7,647,869 cycles 4.005135036 7,032,086 cycles ^C 4.340902364 3,923,893 cycles # perf stat report --per-node # time node cpus counts unit events 1.000536937 N0 20 9,355,086 cycles 1.000536937 N1 20 653,382 cycles 2.002090152 N0 20 7,712,838 cycles 2.002090152 N1 20 1,865,701 cycles 3.003625233 N0 20 6,604,441 cycles 3.003625233 N1 20 1,043,428 cycles 4.005135036 N0 20 6,350,522 cycles 4.005135036 N1 20 681,564 cycles 4.340902364 N0 20 3,403,188 cycles 4.340902364 N1 20 520,705 cycles Signed-off-by: Jiri Olsa <jolsa@kernel.org> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Alexey Budankov <alexey.budankov@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Joe Mario <jmario@redhat.com> Cc: Kan Liang <kan.liang@linux.intel.com> Cc: Michael Petlan <mpetlan@redhat.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/20190904073415.723-4-jolsa@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2019-08-28 08:17:43 +00:00
case AGGR_NODE:
fprintf(config->output, "N%*d%s%*d%s",
config->csv_output ? 0 : -5,
id.node,
perf stat: Add --per-node agregation support Adding new --per-node option to aggregate counts per NUMA nodes for system-wide mode measurements. You can specify --per-node in live mode: # perf stat -a -I 1000 -e cycles --per-node # time node cpus counts unit events 1.000542550 N0 20 6,202,097 cycles 1.000542550 N1 20 639,559 cycles 2.002040063 N0 20 7,412,495 cycles 2.002040063 N1 20 2,185,577 cycles 3.003451699 N0 20 6,508,917 cycles 3.003451699 N1 20 765,607 cycles ... Or in the record/report stat session: # perf stat record -a -I 1000 -e cycles # time counts unit events 1.000536937 10,008,468 cycles 2.002090152 9,578,539 cycles 3.003625233 7,647,869 cycles 4.005135036 7,032,086 cycles ^C 4.340902364 3,923,893 cycles # perf stat report --per-node # time node cpus counts unit events 1.000536937 N0 20 9,355,086 cycles 1.000536937 N1 20 653,382 cycles 2.002090152 N0 20 7,712,838 cycles 2.002090152 N1 20 1,865,701 cycles 3.003625233 N0 20 6,604,441 cycles 3.003625233 N1 20 1,043,428 cycles 4.005135036 N0 20 6,350,522 cycles 4.005135036 N1 20 681,564 cycles 4.340902364 N0 20 3,403,188 cycles 4.340902364 N1 20 520,705 cycles Signed-off-by: Jiri Olsa <jolsa@kernel.org> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Alexey Budankov <alexey.budankov@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Joe Mario <jmario@redhat.com> Cc: Kan Liang <kan.liang@linux.intel.com> Cc: Michael Petlan <mpetlan@redhat.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/20190904073415.723-4-jolsa@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2019-08-28 08:17:43 +00:00
config->csv_sep,
config->csv_output ? 0 : 4,
nr,
config->csv_sep);
break;
case AGGR_NONE:
perf stat: Show percore counts in per CPU output We have supported the event modifier "percore" which sums up the event counts for all hardware threads in a core and show the counts per core. For example, # perf stat -e cpu/event=cpu-cycles,percore/ -a -A -- sleep 1 Performance counter stats for 'system wide': S0-D0-C0 395,072 cpu/event=cpu-cycles,percore/ S0-D0-C1 851,248 cpu/event=cpu-cycles,percore/ S0-D0-C2 954,226 cpu/event=cpu-cycles,percore/ S0-D0-C3 1,233,659 cpu/event=cpu-cycles,percore/ This patch provides a new option "--percore-show-thread". It is used with event modifier "percore" together to sum up the event counts for all hardware threads in a core but show the counts per hardware thread. This is essentially a replacement for the any bit (which is gone in Icelake). Per core counts are useful for some formulas, e.g. CoreIPC. The original percore version was inconvenient to post process. This variant matches the output of the any bit. With this patch, for example, # perf stat -e cpu/event=cpu-cycles,percore/ -a -A --percore-show-thread -- sleep 1 Performance counter stats for 'system wide': CPU0 2,453,061 cpu/event=cpu-cycles,percore/ CPU1 1,823,921 cpu/event=cpu-cycles,percore/ CPU2 1,383,166 cpu/event=cpu-cycles,percore/ CPU3 1,102,652 cpu/event=cpu-cycles,percore/ CPU4 2,453,061 cpu/event=cpu-cycles,percore/ CPU5 1,823,921 cpu/event=cpu-cycles,percore/ CPU6 1,383,166 cpu/event=cpu-cycles,percore/ CPU7 1,102,652 cpu/event=cpu-cycles,percore/ We can see counts are duplicated in CPU pairs (CPU0/CPU4, CPU1/CPU5, CPU2/CPU6, CPU3/CPU7). The interval mode also works. For example, # perf stat -e cpu/event=cpu-cycles,percore/ -a -A --percore-show-thread -I 1000 # time CPU counts unit events 1.000425421 CPU0 925,032 cpu/event=cpu-cycles,percore/ 1.000425421 CPU1 430,202 cpu/event=cpu-cycles,percore/ 1.000425421 CPU2 436,843 cpu/event=cpu-cycles,percore/ 1.000425421 CPU3 1,192,504 cpu/event=cpu-cycles,percore/ 1.000425421 CPU4 925,032 cpu/event=cpu-cycles,percore/ 1.000425421 CPU5 430,202 cpu/event=cpu-cycles,percore/ 1.000425421 CPU6 436,843 cpu/event=cpu-cycles,percore/ 1.000425421 CPU7 1,192,504 cpu/event=cpu-cycles,percore/ If we offline CPU5, the result is: # perf stat -e cpu/event=cpu-cycles,percore/ -a -A --percore-show-thread -- sleep 1 Performance counter stats for 'system wide': CPU0 2,752,148 cpu/event=cpu-cycles,percore/ CPU1 1,009,312 cpu/event=cpu-cycles,percore/ CPU2 2,784,072 cpu/event=cpu-cycles,percore/ CPU3 2,427,922 cpu/event=cpu-cycles,percore/ CPU4 2,752,148 cpu/event=cpu-cycles,percore/ CPU6 2,784,072 cpu/event=cpu-cycles,percore/ CPU7 2,427,922 cpu/event=cpu-cycles,percore/ 1.001416041 seconds time elapsed v4: --- Ravi Bangoria reports an issue in v3. Once we offline a CPU, the output is not correct. The issue is we should use the cpu idx in print_percore_thread rather than using the cpu value. v3: --- 1. Fix the interval mode output error 2. Use cpu value (not cpu index) in config->aggr_get_id(). 3. Refine the code according to Jiri's comments. v2: --- Add the explanation in change log. This is essentially a replacement for the any bit. No code change. Signed-off-by: Jin Yao <yao.jin@linux.intel.com> Tested-by: Ravi Bangoria <ravi.bangoria@linux.ibm.com> Acked-by: Jiri Olsa <jolsa@kernel.org> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Kan Liang <kan.liang@linux.intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lore.kernel.org/lkml/20200214080452.26402-1-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2020-02-14 08:04:52 +00:00
if (evsel->percore && !config->percore_show_thread) {
fprintf(config->output, "S%d-D%d-C%*d%s",
perf stat aggregation: Add separate socket member Add socket as a separate member so that it doesn't have to be packed into the int value. When the socket ID was larger than 8 bits the output appeared corrupted or incomplete. For example, here on ThunderX2 'perf stat' reports a socket of -1 and an invalid die number: ./perf stat -a --per-die The socket id number is too big. Performance counter stats for 'system wide': S-1-D255 128 687.99 msec cpu-clock # 57.240 CPUs utilized ... S36-D0 128 842.34 msec cpu-clock # 70.081 CPUs utilized ... And with --per-core there is an entry with an invalid core ID: ./perf stat record -a --per-core The socket id number is too big. Performance counter stats for 'system wide': S-1-D255-C65535 128 671.04 msec cpu-clock # 54.112 CPUs utilized ... S36-D0-C0 4 28.27 msec cpu-clock # 2.279 CPUs utilized ... This fixes the "Session topology" self test on ThunderX2. After this fix the output contains the correct socket and die IDs and no longer prints a warning about the size of the socket ID: ./perf stat --per-die -a Performance counter stats for 'system wide': S36-D0 128 169,869.39 msec cpu-clock # 127.501 CPUs utilized ... S3612-D0 128 169,733.05 msec cpu-clock # 127.398 CPUs utilized Signed-off-by: James Clark <james.clark@arm.com> Acked-by: Namhyung Kim <namhyung@kernel.org> Acked-by: Jiri Olsa <jolsa@redhat.com> Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com> Tested-by: John Garry <john.garry@huawei.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: Mark Rutland <mark.rutland@arm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Richter <tmricht@linux.ibm.com> Link: https://lore.kernel.org/r/20201126141328.6509-10-james.clark@arm.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2020-11-26 14:13:25 +00:00
id.socket,
id.die,
perf stat: Align the output for interval aggregation mode There is a slight misalignment in -A -I output. For example: # perf stat -e cpu/event=cpu-cycles/ -a -A -I 1000 # time CPU counts unit events 1.000440863 CPU0 1,068,388 cpu/event=cpu-cycles/ 1.000440863 CPU1 875,954 cpu/event=cpu-cycles/ 1.000440863 CPU2 3,072,538 cpu/event=cpu-cycles/ 1.000440863 CPU3 4,026,870 cpu/event=cpu-cycles/ 1.000440863 CPU4 5,919,630 cpu/event=cpu-cycles/ 1.000440863 CPU5 2,714,260 cpu/event=cpu-cycles/ 1.000440863 CPU6 2,219,240 cpu/event=cpu-cycles/ 1.000440863 CPU7 1,299,232 cpu/event=cpu-cycles/ The value of counts is not aligned with the column "counts" and the event name is not aligned with the column "events". With this patch, the output is, # perf stat -e cpu/event=cpu-cycles/ -a -A -I 1000 # time CPU counts unit events 1.000423009 CPU0 997,421 cpu/event=cpu-cycles/ 1.000423009 CPU1 1,422,042 cpu/event=cpu-cycles/ 1.000423009 CPU2 484,651 cpu/event=cpu-cycles/ 1.000423009 CPU3 525,791 cpu/event=cpu-cycles/ 1.000423009 CPU4 1,370,100 cpu/event=cpu-cycles/ 1.000423009 CPU5 442,072 cpu/event=cpu-cycles/ 1.000423009 CPU6 205,643 cpu/event=cpu-cycles/ 1.000423009 CPU7 1,302,250 cpu/event=cpu-cycles/ Now output is aligned. Signed-off-by: Jin Yao <yao.jin@linux.intel.com> Acked-by: Jiri Olsa <jolsa@kernel.org> Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Kan Liang <kan.liang@linux.intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lore.kernel.org/lkml/20200218071614.25736-1-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2020-02-18 07:16:14 +00:00
config->csv_output ? 0 : -3,
id.core, config->csv_sep);
} else if (id.core > -1) {
perf stat: Align the output for interval aggregation mode There is a slight misalignment in -A -I output. For example: # perf stat -e cpu/event=cpu-cycles/ -a -A -I 1000 # time CPU counts unit events 1.000440863 CPU0 1,068,388 cpu/event=cpu-cycles/ 1.000440863 CPU1 875,954 cpu/event=cpu-cycles/ 1.000440863 CPU2 3,072,538 cpu/event=cpu-cycles/ 1.000440863 CPU3 4,026,870 cpu/event=cpu-cycles/ 1.000440863 CPU4 5,919,630 cpu/event=cpu-cycles/ 1.000440863 CPU5 2,714,260 cpu/event=cpu-cycles/ 1.000440863 CPU6 2,219,240 cpu/event=cpu-cycles/ 1.000440863 CPU7 1,299,232 cpu/event=cpu-cycles/ The value of counts is not aligned with the column "counts" and the event name is not aligned with the column "events". With this patch, the output is, # perf stat -e cpu/event=cpu-cycles/ -a -A -I 1000 # time CPU counts unit events 1.000423009 CPU0 997,421 cpu/event=cpu-cycles/ 1.000423009 CPU1 1,422,042 cpu/event=cpu-cycles/ 1.000423009 CPU2 484,651 cpu/event=cpu-cycles/ 1.000423009 CPU3 525,791 cpu/event=cpu-cycles/ 1.000423009 CPU4 1,370,100 cpu/event=cpu-cycles/ 1.000423009 CPU5 442,072 cpu/event=cpu-cycles/ 1.000423009 CPU6 205,643 cpu/event=cpu-cycles/ 1.000423009 CPU7 1,302,250 cpu/event=cpu-cycles/ Now output is aligned. Signed-off-by: Jin Yao <yao.jin@linux.intel.com> Acked-by: Jiri Olsa <jolsa@kernel.org> Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Kan Liang <kan.liang@linux.intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lore.kernel.org/lkml/20200218071614.25736-1-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2020-02-18 07:16:14 +00:00
fprintf(config->output, "CPU%*d%s",
config->csv_output ? 0 : -7,
evsel__cpus(evsel)->map[id.core],
perf stat: Support 'percore' event qualifier With this patch, we can use the 'percore' event qualifier in perf-stat. root@skl:/tmp# perf stat -e cpu/event=0,umask=0x3,percore=1/,cpu/event=0,umask=0x3/ -a -A -I1000 1.000773050 S0-C0 98,352,832 cpu/event=0,umask=0x3,percore=1/ (50.01%) 1.000773050 S0-C1 103,763,057 cpu/event=0,umask=0x3,percore=1/ (50.02%) 1.000773050 S0-C2 196,776,995 cpu/event=0,umask=0x3,percore=1/ (50.02%) 1.000773050 S0-C3 176,493,779 cpu/event=0,umask=0x3,percore=1/ (50.02%) 1.000773050 CPU0 47,699,641 cpu/event=0,umask=0x3/ (50.02%) 1.000773050 CPU1 49,052,451 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU2 102,771,422 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU3 100,784,662 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU4 43,171,342 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU5 54,152,158 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU6 93,618,410 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU7 74,477,589 cpu/event=0,umask=0x3/ (49.99%) In this example, we count the event 'ref-cycles' per-core and per-CPU in one perf stat command-line. From the output, we can see: S0-C0 = CPU0 + CPU4 S0-C1 = CPU1 + CPU5 S0-C2 = CPU2 + CPU6 S0-C3 = CPU3 + CPU7 So the result is expected (tiny difference is ignored). Note that, the 'percore' event qualifier needs to use with option '-A'. Signed-off-by: Jin Yao <yao.jin@linux.intel.com> Tested-by: Ravi Bangoria <ravi.bangoria@linux.ibm.com> Acked-by: Jiri Olsa <jolsa@kernel.org> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Jin Yao <yao.jin@intel.com> Cc: Kan Liang <kan.liang@linux.intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1555077590-27664-4-git-send-email-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2019-04-12 13:59:49 +00:00
config->csv_sep);
}
break;
case AGGR_THREAD:
fprintf(config->output, "%*s-%*d%s",
config->csv_output ? 0 : 16,
perf_thread_map__comm(evsel->core.threads, id.thread),
config->csv_output ? 0 : -8,
perf_thread_map__pid(evsel->core.threads, id.thread),
config->csv_sep);
break;
case AGGR_GLOBAL:
case AGGR_UNSET:
default:
break;
}
}
struct outstate {
FILE *fh;
bool newline;
const char *prefix;
int nfields;
int nr;
struct aggr_cpu_id id;
struct evsel *evsel;
};
#define METRIC_LEN 35
static void new_line_std(struct perf_stat_config *config __maybe_unused,
void *ctx)
{
struct outstate *os = ctx;
os->newline = true;
}
static void do_new_line_std(struct perf_stat_config *config,
struct outstate *os)
{
fputc('\n', os->fh);
fputs(os->prefix, os->fh);
aggr_printout(config, os->evsel, os->id, os->nr);
if (config->aggr_mode == AGGR_NONE)
fprintf(os->fh, " ");
fprintf(os->fh, " ");
}
static void print_metric_std(struct perf_stat_config *config,
void *ctx, const char *color, const char *fmt,
const char *unit, double val)
{
struct outstate *os = ctx;
FILE *out = os->fh;
int n;
bool newline = os->newline;
os->newline = false;
if (unit == NULL || fmt == NULL) {
fprintf(out, "%-*s", METRIC_LEN, "");
return;
}
if (newline)
do_new_line_std(config, os);
n = fprintf(out, " # ");
if (color)
n += color_fprintf(out, color, fmt, val);
else
n += fprintf(out, fmt, val);
fprintf(out, " %-*s", METRIC_LEN - n - 1, unit);
}
static void new_line_csv(struct perf_stat_config *config, void *ctx)
{
struct outstate *os = ctx;
int i;
fputc('\n', os->fh);
if (os->prefix)
fprintf(os->fh, "%s%s", os->prefix, config->csv_sep);
aggr_printout(config, os->evsel, os->id, os->nr);
for (i = 0; i < os->nfields; i++)
fputs(config->csv_sep, os->fh);
}
static void print_metric_csv(struct perf_stat_config *config __maybe_unused,
void *ctx,
const char *color __maybe_unused,
const char *fmt, const char *unit, double val)
{
struct outstate *os = ctx;
FILE *out = os->fh;
char buf[64], *vals, *ends;
if (unit == NULL || fmt == NULL) {
fprintf(out, "%s%s", config->csv_sep, config->csv_sep);
return;
}
snprintf(buf, sizeof(buf), fmt, val);
ends = vals = skip_spaces(buf);
while (isdigit(*ends) || *ends == '.')
ends++;
*ends = 0;
fprintf(out, "%s%s%s%s", config->csv_sep, vals, config->csv_sep, skip_spaces(unit));
}
/* Filter out some columns that don't work well in metrics only mode */
static bool valid_only_metric(const char *unit)
{
if (!unit)
return false;
if (strstr(unit, "/sec") ||
strstr(unit, "CPUs utilized"))
return false;
return true;
}
static const char *fixunit(char *buf, struct evsel *evsel,
const char *unit)
{
if (!strncmp(unit, "of all", 6)) {
snprintf(buf, 1024, "%s %s", evsel__name(evsel),
unit);
return buf;
}
return unit;
}
static void print_metric_only(struct perf_stat_config *config,
void *ctx, const char *color, const char *fmt,
const char *unit, double val)
{
struct outstate *os = ctx;
FILE *out = os->fh;
char buf[1024], str[1024];
unsigned mlen = config->metric_only_len;
if (!valid_only_metric(unit))
return;
unit = fixunit(buf, os->evsel, unit);
if (mlen < strlen(unit))
mlen = strlen(unit) + 1;
if (color)
mlen += strlen(color) + sizeof(PERF_COLOR_RESET) - 1;
color_snprintf(str, sizeof(str), color ?: "", fmt, val);
fprintf(out, "%*s ", mlen, str);
}
static void print_metric_only_csv(struct perf_stat_config *config __maybe_unused,
void *ctx, const char *color __maybe_unused,
const char *fmt,
const char *unit, double val)
{
struct outstate *os = ctx;
FILE *out = os->fh;
char buf[64], *vals, *ends;
char tbuf[1024];
if (!valid_only_metric(unit))
return;
unit = fixunit(tbuf, os->evsel, unit);
snprintf(buf, sizeof buf, fmt, val);
ends = vals = skip_spaces(buf);
while (isdigit(*ends) || *ends == '.')
ends++;
*ends = 0;
fprintf(out, "%s%s", vals, config->csv_sep);
}
static void new_line_metric(struct perf_stat_config *config __maybe_unused,
void *ctx __maybe_unused)
{
}
static void print_metric_header(struct perf_stat_config *config,
void *ctx, const char *color __maybe_unused,
const char *fmt __maybe_unused,
const char *unit, double val __maybe_unused)
{
struct outstate *os = ctx;
char tbuf[1024];
if (!valid_only_metric(unit))
return;
unit = fixunit(tbuf, os->evsel, unit);
if (config->csv_output)
fprintf(os->fh, "%s%s", unit, config->csv_sep);
else
fprintf(os->fh, "%*s ", config->metric_only_len, unit);
}
static int first_shadow_cpu(struct perf_stat_config *config,
struct evsel *evsel, struct aggr_cpu_id id)
{
struct evlist *evlist = evsel->evlist;
int i;
if (config->aggr_mode == AGGR_NONE)
return id.core;
perf stat: Use proper cpu for shadow stats Currently perf stat shows some metrics (like IPC) for defined events. But when no aggregation mode is used (-A option), it shows incorrect values since it used a value from a different cpu. Before: $ perf stat -aA -e cycles,instructions sleep 1 Performance counter stats for 'system wide': CPU0 116,057,380 cycles CPU1 86,084,722 cycles CPU2 99,423,125 cycles CPU3 98,272,994 cycles CPU0 53,369,217 instructions # 0.46 insn per cycle CPU1 33,378,058 instructions # 0.29 insn per cycle CPU2 58,150,086 instructions # 0.50 insn per cycle CPU3 40,029,703 instructions # 0.34 insn per cycle 1.001816971 seconds time elapsed So the IPC for CPU1 should be 0.38 (= 33,378,058 / 86,084,722) but it was 0.29 (= 33,378,058 / 116,057,380) and so on. After: $ perf stat -aA -e cycles,instructions sleep 1 Performance counter stats for 'system wide': CPU0 109,621,384 cycles CPU1 159,026,454 cycles CPU2 99,460,366 cycles CPU3 124,144,142 cycles CPU0 44,396,706 instructions # 0.41 insn per cycle CPU1 120,195,425 instructions # 0.76 insn per cycle CPU2 44,763,978 instructions # 0.45 insn per cycle CPU3 69,049,079 instructions # 0.56 insn per cycle 1.001910444 seconds time elapsed Fixes: 44d49a600259 ("perf stat: Support metrics in --per-core/socket mode") Reported-by: Sam Xi <xyzsam@google.com> Signed-off-by: Namhyung Kim <namhyung@kernel.org> Reviewed-by: Andi Kleen <ak@linux.intel.com> Acked-by: Jiri Olsa <jolsa@redhat.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Ian Rogers <irogers@google.com> Cc: Mark Rutland <mark.rutland@arm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Stephane Eranian <eranian@google.com> Link: http://lore.kernel.org/lkml/20201127041404.390276-1-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2020-11-27 04:14:03 +00:00
if (!config->aggr_get_id)
return 0;
for (i = 0; i < evsel__nr_cpus(evsel); i++) {
int cpu2 = evsel__cpus(evsel)->map[i];
if (cpu_map__compare_aggr_cpu_id(
config->aggr_get_id(config, evlist->core.cpus, cpu2),
id)) {
return cpu2;
}
}
return 0;
}
static void abs_printout(struct perf_stat_config *config,
struct aggr_cpu_id id, int nr, struct evsel *evsel, double avg)
{
FILE *output = config->output;
double sc = evsel->scale;
const char *fmt;
if (config->csv_output) {
fmt = floor(sc) != sc ? "%.2f%s" : "%.0f%s";
} else {
if (config->big_num)
fmt = floor(sc) != sc ? "%'18.2f%s" : "%'18.0f%s";
else
fmt = floor(sc) != sc ? "%18.2f%s" : "%18.0f%s";
}
aggr_printout(config, evsel, id, nr);
fprintf(output, fmt, avg, config->csv_sep);
if (evsel->unit)
fprintf(output, "%-*s%s",
config->csv_output ? 0 : config->unit_width,
evsel->unit, config->csv_sep);
fprintf(output, "%-*s", config->csv_output ? 0 : 25, evsel__name(evsel));
print_cgroup(config, evsel);
}
static bool is_mixed_hw_group(struct evsel *counter)
{
struct evlist *evlist = counter->evlist;
libperf: Move perf_event_attr field from perf's evsel to libperf's perf_evsel Move the perf_event_attr struct fron 'struct evsel' to 'struct perf_evsel'. Committer notes: Fixed up these: tools/perf/arch/arm/util/auxtrace.c tools/perf/arch/arm/util/cs-etm.c tools/perf/arch/arm64/util/arm-spe.c tools/perf/arch/s390/util/auxtrace.c tools/perf/util/cs-etm.c Also cc1: warnings being treated as errors tests/sample-parsing.c: In function 'do_test': tests/sample-parsing.c:162: error: missing initializer tests/sample-parsing.c:162: error: (near initialization for 'evsel.core.cpus') struct evsel evsel = { .needs_swap = false, - .core.attr = { - .sample_type = sample_type, - .read_format = read_format, + .core = { + . attr = { + .sample_type = sample_type, + .read_format = read_format, + }, [perfbuilder@a70e4eeb5549 /]$ gcc --version |& head -1 gcc (GCC) 4.4.7 Also we don't need to include perf_event.h in tools/perf/lib/include/perf/evsel.h, forward declaring 'struct perf_event_attr' is enough. And this even fixes the build in some systems where things are used somewhere down the include path from perf_event.h without defining __always_inline. Signed-off-by: Jiri Olsa <jolsa@kernel.org> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Alexey Budankov <alexey.budankov@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Michael Petlan <mpetlan@redhat.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/20190721112506.12306-43-jolsa@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2019-07-21 11:24:29 +00:00
u32 pmu_type = counter->core.attr.type;
struct evsel *pos;
if (counter->core.nr_members < 2)
return false;
evlist__for_each_entry(evlist, pos) {
/* software events can be part of any hardware group */
libperf: Move perf_event_attr field from perf's evsel to libperf's perf_evsel Move the perf_event_attr struct fron 'struct evsel' to 'struct perf_evsel'. Committer notes: Fixed up these: tools/perf/arch/arm/util/auxtrace.c tools/perf/arch/arm/util/cs-etm.c tools/perf/arch/arm64/util/arm-spe.c tools/perf/arch/s390/util/auxtrace.c tools/perf/util/cs-etm.c Also cc1: warnings being treated as errors tests/sample-parsing.c: In function 'do_test': tests/sample-parsing.c:162: error: missing initializer tests/sample-parsing.c:162: error: (near initialization for 'evsel.core.cpus') struct evsel evsel = { .needs_swap = false, - .core.attr = { - .sample_type = sample_type, - .read_format = read_format, + .core = { + . attr = { + .sample_type = sample_type, + .read_format = read_format, + }, [perfbuilder@a70e4eeb5549 /]$ gcc --version |& head -1 gcc (GCC) 4.4.7 Also we don't need to include perf_event.h in tools/perf/lib/include/perf/evsel.h, forward declaring 'struct perf_event_attr' is enough. And this even fixes the build in some systems where things are used somewhere down the include path from perf_event.h without defining __always_inline. Signed-off-by: Jiri Olsa <jolsa@kernel.org> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Alexey Budankov <alexey.budankov@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Michael Petlan <mpetlan@redhat.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/20190721112506.12306-43-jolsa@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2019-07-21 11:24:29 +00:00
if (pos->core.attr.type == PERF_TYPE_SOFTWARE)
continue;
if (pmu_type == PERF_TYPE_SOFTWARE) {
libperf: Move perf_event_attr field from perf's evsel to libperf's perf_evsel Move the perf_event_attr struct fron 'struct evsel' to 'struct perf_evsel'. Committer notes: Fixed up these: tools/perf/arch/arm/util/auxtrace.c tools/perf/arch/arm/util/cs-etm.c tools/perf/arch/arm64/util/arm-spe.c tools/perf/arch/s390/util/auxtrace.c tools/perf/util/cs-etm.c Also cc1: warnings being treated as errors tests/sample-parsing.c: In function 'do_test': tests/sample-parsing.c:162: error: missing initializer tests/sample-parsing.c:162: error: (near initialization for 'evsel.core.cpus') struct evsel evsel = { .needs_swap = false, - .core.attr = { - .sample_type = sample_type, - .read_format = read_format, + .core = { + . attr = { + .sample_type = sample_type, + .read_format = read_format, + }, [perfbuilder@a70e4eeb5549 /]$ gcc --version |& head -1 gcc (GCC) 4.4.7 Also we don't need to include perf_event.h in tools/perf/lib/include/perf/evsel.h, forward declaring 'struct perf_event_attr' is enough. And this even fixes the build in some systems where things are used somewhere down the include path from perf_event.h without defining __always_inline. Signed-off-by: Jiri Olsa <jolsa@kernel.org> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Alexey Budankov <alexey.budankov@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Michael Petlan <mpetlan@redhat.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/20190721112506.12306-43-jolsa@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2019-07-21 11:24:29 +00:00
pmu_type = pos->core.attr.type;
continue;
}
libperf: Move perf_event_attr field from perf's evsel to libperf's perf_evsel Move the perf_event_attr struct fron 'struct evsel' to 'struct perf_evsel'. Committer notes: Fixed up these: tools/perf/arch/arm/util/auxtrace.c tools/perf/arch/arm/util/cs-etm.c tools/perf/arch/arm64/util/arm-spe.c tools/perf/arch/s390/util/auxtrace.c tools/perf/util/cs-etm.c Also cc1: warnings being treated as errors tests/sample-parsing.c: In function 'do_test': tests/sample-parsing.c:162: error: missing initializer tests/sample-parsing.c:162: error: (near initialization for 'evsel.core.cpus') struct evsel evsel = { .needs_swap = false, - .core.attr = { - .sample_type = sample_type, - .read_format = read_format, + .core = { + . attr = { + .sample_type = sample_type, + .read_format = read_format, + }, [perfbuilder@a70e4eeb5549 /]$ gcc --version |& head -1 gcc (GCC) 4.4.7 Also we don't need to include perf_event.h in tools/perf/lib/include/perf/evsel.h, forward declaring 'struct perf_event_attr' is enough. And this even fixes the build in some systems where things are used somewhere down the include path from perf_event.h without defining __always_inline. Signed-off-by: Jiri Olsa <jolsa@kernel.org> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Alexey Budankov <alexey.budankov@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Michael Petlan <mpetlan@redhat.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/20190721112506.12306-43-jolsa@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2019-07-21 11:24:29 +00:00
if (pmu_type != pos->core.attr.type)
return true;
}
return false;
}
static void printout(struct perf_stat_config *config, struct aggr_cpu_id id, int nr,
struct evsel *counter, double uval,
char *prefix, u64 run, u64 ena, double noise,
struct runtime_stat *st)
{
struct perf_stat_output_ctx out;
struct outstate os = {
.fh = config->output,
.prefix = prefix ? prefix : "",
.id = id,
.nr = nr,
.evsel = counter,
};
print_metric_t pm = print_metric_std;
new_line_t nl;
if (config->metric_only) {
nl = new_line_metric;
if (config->csv_output)
pm = print_metric_only_csv;
else
pm = print_metric_only;
} else
nl = new_line_std;
if (config->csv_output && !config->metric_only) {
static int aggr_fields[] = {
[AGGR_GLOBAL] = 0,
[AGGR_THREAD] = 1,
[AGGR_NONE] = 1,
[AGGR_SOCKET] = 2,
[AGGR_DIE] = 2,
[AGGR_CORE] = 2,
};
pm = print_metric_csv;
nl = new_line_csv;
os.nfields = 3;
os.nfields += aggr_fields[config->aggr_mode];
if (counter->cgrp)
os.nfields++;
}
if (run == 0 || ena == 0 || counter->counts->scaled == -1) {
if (config->metric_only) {
pm(config, &os, NULL, "", "", 0);
return;
}
aggr_printout(config, counter, id, nr);
fprintf(config->output, "%*s%s",
config->csv_output ? 0 : 18,
counter->supported ? CNTR_NOT_COUNTED : CNTR_NOT_SUPPORTED,
config->csv_sep);
if (counter->supported) {
config->print_free_counters_hint = 1;
if (is_mixed_hw_group(counter))
config->print_mixed_hw_group_error = 1;
}
fprintf(config->output, "%-*s%s",
config->csv_output ? 0 : config->unit_width,
counter->unit, config->csv_sep);
fprintf(config->output, "%*s",
config->csv_output ? 0 : -25, evsel__name(counter));
print_cgroup(config, counter);
if (!config->csv_output)
pm(config, &os, NULL, NULL, "", 0);
print_noise(config, counter, noise);
print_running(config, run, ena);
if (config->csv_output)
pm(config, &os, NULL, NULL, "", 0);
return;
}
if (!config->metric_only)
abs_printout(config, id, nr, counter, uval);
out.print_metric = pm;
out.new_line = nl;
out.ctx = &os;
out.force_header = false;
if (config->csv_output && !config->metric_only) {
print_noise(config, counter, noise);
print_running(config, run, ena);
}
perf_stat__print_shadow_stats(config, counter, uval,
first_shadow_cpu(config, counter, id),
&out, &config->metric_events, st);
if (!config->csv_output && !config->metric_only) {
print_noise(config, counter, noise);
print_running(config, run, ena);
}
}
static void aggr_update_shadow(struct perf_stat_config *config,
struct evlist *evlist)
{
int cpu, s;
struct aggr_cpu_id s2, id;
u64 val;
struct evsel *counter;
for (s = 0; s < config->aggr_map->nr; s++) {
id = config->aggr_map->map[s];
evlist__for_each_entry(evlist, counter) {
val = 0;
for (cpu = 0; cpu < evsel__nr_cpus(counter); cpu++) {
s2 = config->aggr_get_id(config, evlist->core.cpus, cpu);
if (!cpu_map__compare_aggr_cpu_id(s2, id))
continue;
val += perf_counts(counter->counts, cpu, 0)->val;
}
perf_stat__update_shadow_stats(counter, val,
first_shadow_cpu(config, counter, id),
&rt_stat);
}
}
}
static void uniquify_event_name(struct evsel *counter)
{
char *new_name;
char *config;
if (counter->uniquified_name ||
!counter->pmu_name || !strncmp(counter->name, counter->pmu_name,
strlen(counter->pmu_name)))
return;
config = strchr(counter->name, '/');
if (config) {
if (asprintf(&new_name,
"%s%s", counter->pmu_name, config) > 0) {
free(counter->name);
counter->name = new_name;
}
} else {
if (asprintf(&new_name,
"%s [%s]", counter->name, counter->pmu_name) > 0) {
free(counter->name);
counter->name = new_name;
}
}
counter->uniquified_name = true;
}
static void collect_all_aliases(struct perf_stat_config *config, struct evsel *counter,
void (*cb)(struct perf_stat_config *config, struct evsel *counter, void *data,
bool first),
void *data)
{
struct evlist *evlist = counter->evlist;
struct evsel *alias;
alias = list_prepare_entry(counter, &(evlist->core.entries), core.node);
list_for_each_entry_continue (alias, &evlist->core.entries, core.node) {
if (strcmp(evsel__name(alias), evsel__name(counter)) ||
alias->scale != counter->scale ||
alias->cgrp != counter->cgrp ||
strcmp(alias->unit, counter->unit) ||
evsel__is_clock(alias) != evsel__is_clock(counter) ||
!strcmp(alias->pmu_name, counter->pmu_name))
break;
alias->merged_stat = true;
cb(config, alias, data, false);
}
}
static bool collect_data(struct perf_stat_config *config, struct evsel *counter,
void (*cb)(struct perf_stat_config *config, struct evsel *counter, void *data,
bool first),
void *data)
{
if (counter->merged_stat)
return false;
cb(config, counter, data, true);
if (config->no_merge)
uniquify_event_name(counter);
else if (counter->auto_merge_stats)
collect_all_aliases(config, counter, cb, data);
return true;
}
struct aggr_data {
u64 ena, run, val;
struct aggr_cpu_id id;
int nr;
int cpu;
};
static void aggr_cb(struct perf_stat_config *config,
struct evsel *counter, void *data, bool first)
{
struct aggr_data *ad = data;
int cpu;
struct aggr_cpu_id s2;
for (cpu = 0; cpu < evsel__nr_cpus(counter); cpu++) {
struct perf_counts_values *counts;
s2 = config->aggr_get_id(config, evsel__cpus(counter), cpu);
if (!cpu_map__compare_aggr_cpu_id(s2, ad->id))
continue;
if (first)
ad->nr++;
counts = perf_counts(counter->counts, cpu, 0);
/*
* When any result is bad, make them all to give
* consistent output in interval mode.
*/
if (counts->ena == 0 || counts->run == 0 ||
counter->counts->scaled == -1) {
ad->ena = 0;
ad->run = 0;
break;
}
ad->val += counts->val;
ad->ena += counts->ena;
ad->run += counts->run;
}
}
static void print_counter_aggrdata(struct perf_stat_config *config,
struct evsel *counter, int s,
char *prefix, bool metric_only,
perf stat: Show percore counts in per CPU output We have supported the event modifier "percore" which sums up the event counts for all hardware threads in a core and show the counts per core. For example, # perf stat -e cpu/event=cpu-cycles,percore/ -a -A -- sleep 1 Performance counter stats for 'system wide': S0-D0-C0 395,072 cpu/event=cpu-cycles,percore/ S0-D0-C1 851,248 cpu/event=cpu-cycles,percore/ S0-D0-C2 954,226 cpu/event=cpu-cycles,percore/ S0-D0-C3 1,233,659 cpu/event=cpu-cycles,percore/ This patch provides a new option "--percore-show-thread". It is used with event modifier "percore" together to sum up the event counts for all hardware threads in a core but show the counts per hardware thread. This is essentially a replacement for the any bit (which is gone in Icelake). Per core counts are useful for some formulas, e.g. CoreIPC. The original percore version was inconvenient to post process. This variant matches the output of the any bit. With this patch, for example, # perf stat -e cpu/event=cpu-cycles,percore/ -a -A --percore-show-thread -- sleep 1 Performance counter stats for 'system wide': CPU0 2,453,061 cpu/event=cpu-cycles,percore/ CPU1 1,823,921 cpu/event=cpu-cycles,percore/ CPU2 1,383,166 cpu/event=cpu-cycles,percore/ CPU3 1,102,652 cpu/event=cpu-cycles,percore/ CPU4 2,453,061 cpu/event=cpu-cycles,percore/ CPU5 1,823,921 cpu/event=cpu-cycles,percore/ CPU6 1,383,166 cpu/event=cpu-cycles,percore/ CPU7 1,102,652 cpu/event=cpu-cycles,percore/ We can see counts are duplicated in CPU pairs (CPU0/CPU4, CPU1/CPU5, CPU2/CPU6, CPU3/CPU7). The interval mode also works. For example, # perf stat -e cpu/event=cpu-cycles,percore/ -a -A --percore-show-thread -I 1000 # time CPU counts unit events 1.000425421 CPU0 925,032 cpu/event=cpu-cycles,percore/ 1.000425421 CPU1 430,202 cpu/event=cpu-cycles,percore/ 1.000425421 CPU2 436,843 cpu/event=cpu-cycles,percore/ 1.000425421 CPU3 1,192,504 cpu/event=cpu-cycles,percore/ 1.000425421 CPU4 925,032 cpu/event=cpu-cycles,percore/ 1.000425421 CPU5 430,202 cpu/event=cpu-cycles,percore/ 1.000425421 CPU6 436,843 cpu/event=cpu-cycles,percore/ 1.000425421 CPU7 1,192,504 cpu/event=cpu-cycles,percore/ If we offline CPU5, the result is: # perf stat -e cpu/event=cpu-cycles,percore/ -a -A --percore-show-thread -- sleep 1 Performance counter stats for 'system wide': CPU0 2,752,148 cpu/event=cpu-cycles,percore/ CPU1 1,009,312 cpu/event=cpu-cycles,percore/ CPU2 2,784,072 cpu/event=cpu-cycles,percore/ CPU3 2,427,922 cpu/event=cpu-cycles,percore/ CPU4 2,752,148 cpu/event=cpu-cycles,percore/ CPU6 2,784,072 cpu/event=cpu-cycles,percore/ CPU7 2,427,922 cpu/event=cpu-cycles,percore/ 1.001416041 seconds time elapsed v4: --- Ravi Bangoria reports an issue in v3. Once we offline a CPU, the output is not correct. The issue is we should use the cpu idx in print_percore_thread rather than using the cpu value. v3: --- 1. Fix the interval mode output error 2. Use cpu value (not cpu index) in config->aggr_get_id(). 3. Refine the code according to Jiri's comments. v2: --- Add the explanation in change log. This is essentially a replacement for the any bit. No code change. Signed-off-by: Jin Yao <yao.jin@linux.intel.com> Tested-by: Ravi Bangoria <ravi.bangoria@linux.ibm.com> Acked-by: Jiri Olsa <jolsa@kernel.org> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Kan Liang <kan.liang@linux.intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lore.kernel.org/lkml/20200214080452.26402-1-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2020-02-14 08:04:52 +00:00
bool *first, int cpu)
{
struct aggr_data ad;
FILE *output = config->output;
u64 ena, run, val;
int nr;
struct aggr_cpu_id id;
double uval;
ad.id = id = config->aggr_map->map[s];
ad.val = ad.ena = ad.run = 0;
ad.nr = 0;
if (!collect_data(config, counter, aggr_cb, &ad))
return;
nr = ad.nr;
ena = ad.ena;
run = ad.run;
val = ad.val;
if (*first && metric_only) {
*first = false;
aggr_printout(config, counter, id, nr);
}
if (prefix && !metric_only)
fprintf(output, "%s", prefix);
uval = val * counter->scale;
if (cpu != -1) {
id = cpu_map__empty_aggr_cpu_id();
id.core = cpu;
}
printout(config, id, nr, counter, uval,
prefix, run, ena, 1.0, &rt_stat);
if (!metric_only)
fputc('\n', output);
}
static void print_aggr(struct perf_stat_config *config,
struct evlist *evlist,
char *prefix)
{
bool metric_only = config->metric_only;
FILE *output = config->output;
struct evsel *counter;
int s;
bool first;
if (!config->aggr_map || !config->aggr_get_id)
return;
aggr_update_shadow(config, evlist);
/*
* With metric_only everything is on a single line.
* Without each counter has its own line.
*/
for (s = 0; s < config->aggr_map->nr; s++) {
if (prefix && metric_only)
fprintf(output, "%s", prefix);
first = true;
evlist__for_each_entry(evlist, counter) {
print_counter_aggrdata(config, counter, s,
prefix, metric_only,
perf stat: Show percore counts in per CPU output We have supported the event modifier "percore" which sums up the event counts for all hardware threads in a core and show the counts per core. For example, # perf stat -e cpu/event=cpu-cycles,percore/ -a -A -- sleep 1 Performance counter stats for 'system wide': S0-D0-C0 395,072 cpu/event=cpu-cycles,percore/ S0-D0-C1 851,248 cpu/event=cpu-cycles,percore/ S0-D0-C2 954,226 cpu/event=cpu-cycles,percore/ S0-D0-C3 1,233,659 cpu/event=cpu-cycles,percore/ This patch provides a new option "--percore-show-thread". It is used with event modifier "percore" together to sum up the event counts for all hardware threads in a core but show the counts per hardware thread. This is essentially a replacement for the any bit (which is gone in Icelake). Per core counts are useful for some formulas, e.g. CoreIPC. The original percore version was inconvenient to post process. This variant matches the output of the any bit. With this patch, for example, # perf stat -e cpu/event=cpu-cycles,percore/ -a -A --percore-show-thread -- sleep 1 Performance counter stats for 'system wide': CPU0 2,453,061 cpu/event=cpu-cycles,percore/ CPU1 1,823,921 cpu/event=cpu-cycles,percore/ CPU2 1,383,166 cpu/event=cpu-cycles,percore/ CPU3 1,102,652 cpu/event=cpu-cycles,percore/ CPU4 2,453,061 cpu/event=cpu-cycles,percore/ CPU5 1,823,921 cpu/event=cpu-cycles,percore/ CPU6 1,383,166 cpu/event=cpu-cycles,percore/ CPU7 1,102,652 cpu/event=cpu-cycles,percore/ We can see counts are duplicated in CPU pairs (CPU0/CPU4, CPU1/CPU5, CPU2/CPU6, CPU3/CPU7). The interval mode also works. For example, # perf stat -e cpu/event=cpu-cycles,percore/ -a -A --percore-show-thread -I 1000 # time CPU counts unit events 1.000425421 CPU0 925,032 cpu/event=cpu-cycles,percore/ 1.000425421 CPU1 430,202 cpu/event=cpu-cycles,percore/ 1.000425421 CPU2 436,843 cpu/event=cpu-cycles,percore/ 1.000425421 CPU3 1,192,504 cpu/event=cpu-cycles,percore/ 1.000425421 CPU4 925,032 cpu/event=cpu-cycles,percore/ 1.000425421 CPU5 430,202 cpu/event=cpu-cycles,percore/ 1.000425421 CPU6 436,843 cpu/event=cpu-cycles,percore/ 1.000425421 CPU7 1,192,504 cpu/event=cpu-cycles,percore/ If we offline CPU5, the result is: # perf stat -e cpu/event=cpu-cycles,percore/ -a -A --percore-show-thread -- sleep 1 Performance counter stats for 'system wide': CPU0 2,752,148 cpu/event=cpu-cycles,percore/ CPU1 1,009,312 cpu/event=cpu-cycles,percore/ CPU2 2,784,072 cpu/event=cpu-cycles,percore/ CPU3 2,427,922 cpu/event=cpu-cycles,percore/ CPU4 2,752,148 cpu/event=cpu-cycles,percore/ CPU6 2,784,072 cpu/event=cpu-cycles,percore/ CPU7 2,427,922 cpu/event=cpu-cycles,percore/ 1.001416041 seconds time elapsed v4: --- Ravi Bangoria reports an issue in v3. Once we offline a CPU, the output is not correct. The issue is we should use the cpu idx in print_percore_thread rather than using the cpu value. v3: --- 1. Fix the interval mode output error 2. Use cpu value (not cpu index) in config->aggr_get_id(). 3. Refine the code according to Jiri's comments. v2: --- Add the explanation in change log. This is essentially a replacement for the any bit. No code change. Signed-off-by: Jin Yao <yao.jin@linux.intel.com> Tested-by: Ravi Bangoria <ravi.bangoria@linux.ibm.com> Acked-by: Jiri Olsa <jolsa@kernel.org> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Kan Liang <kan.liang@linux.intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lore.kernel.org/lkml/20200214080452.26402-1-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2020-02-14 08:04:52 +00:00
&first, -1);
}
if (metric_only)
fputc('\n', output);
}
}
static int cmp_val(const void *a, const void *b)
{
return ((struct perf_aggr_thread_value *)b)->val -
((struct perf_aggr_thread_value *)a)->val;
}
static struct perf_aggr_thread_value *sort_aggr_thread(
struct evsel *counter,
int nthreads, int ncpus,
int *ret,
struct target *_target)
{
int cpu, thread, i = 0;
double uval;
struct perf_aggr_thread_value *buf;
buf = calloc(nthreads, sizeof(struct perf_aggr_thread_value));
if (!buf)
return NULL;
for (thread = 0; thread < nthreads; thread++) {
u64 ena = 0, run = 0, val = 0;
for (cpu = 0; cpu < ncpus; cpu++) {
val += perf_counts(counter->counts, cpu, thread)->val;
ena += perf_counts(counter->counts, cpu, thread)->ena;
run += perf_counts(counter->counts, cpu, thread)->run;
}
uval = val * counter->scale;
/*
* Skip value 0 when enabling --per-thread globally,
* otherwise too many 0 output.
*/
if (uval == 0.0 && target__has_per_thread(_target))
continue;
buf[i].counter = counter;
buf[i].id = cpu_map__empty_aggr_cpu_id();
buf[i].id.thread = thread;
buf[i].uval = uval;
buf[i].val = val;
buf[i].run = run;
buf[i].ena = ena;
i++;
}
qsort(buf, i, sizeof(struct perf_aggr_thread_value), cmp_val);
if (ret)
*ret = i;
return buf;
}
static void print_aggr_thread(struct perf_stat_config *config,
struct target *_target,
struct evsel *counter, char *prefix)
{
FILE *output = config->output;
int nthreads = perf_thread_map__nr(counter->core.threads);
int ncpus = perf_cpu_map__nr(counter->core.cpus);
int thread, sorted_threads;
struct aggr_cpu_id id;
struct perf_aggr_thread_value *buf;
buf = sort_aggr_thread(counter, nthreads, ncpus, &sorted_threads, _target);
if (!buf) {
perror("cannot sort aggr thread");
return;
}
for (thread = 0; thread < sorted_threads; thread++) {
if (prefix)
fprintf(output, "%s", prefix);
id = buf[thread].id;
if (config->stats)
printout(config, id, 0, buf[thread].counter, buf[thread].uval,
prefix, buf[thread].run, buf[thread].ena, 1.0,
&config->stats[id.thread]);
else
printout(config, id, 0, buf[thread].counter, buf[thread].uval,
prefix, buf[thread].run, buf[thread].ena, 1.0,
&rt_stat);
fputc('\n', output);
}
free(buf);
}
struct caggr_data {
double avg, avg_enabled, avg_running;
};
static void counter_aggr_cb(struct perf_stat_config *config __maybe_unused,
struct evsel *counter, void *data,
bool first __maybe_unused)
{
struct caggr_data *cd = data;
struct perf_stat_evsel *ps = counter->stats;
cd->avg += avg_stats(&ps->res_stats[0]);
cd->avg_enabled += avg_stats(&ps->res_stats[1]);
cd->avg_running += avg_stats(&ps->res_stats[2]);
}
/*
* Print out the results of a single counter:
* aggregated counts in system-wide mode
*/
static void print_counter_aggr(struct perf_stat_config *config,
struct evsel *counter, char *prefix)
{
bool metric_only = config->metric_only;
FILE *output = config->output;
double uval;
struct caggr_data cd = { .avg = 0.0 };
if (!collect_data(config, counter, counter_aggr_cb, &cd))
return;
if (prefix && !metric_only)
fprintf(output, "%s", prefix);
uval = cd.avg * counter->scale;
printout(config, cpu_map__empty_aggr_cpu_id(), 0, counter, uval, prefix, cd.avg_running,
cd.avg_enabled, cd.avg, &rt_stat);
if (!metric_only)
fprintf(output, "\n");
}
static void counter_cb(struct perf_stat_config *config __maybe_unused,
struct evsel *counter, void *data,
bool first __maybe_unused)
{
struct aggr_data *ad = data;
ad->val += perf_counts(counter->counts, ad->cpu, 0)->val;
ad->ena += perf_counts(counter->counts, ad->cpu, 0)->ena;
ad->run += perf_counts(counter->counts, ad->cpu, 0)->run;
}
/*
* Print out the results of a single counter:
* does not use aggregated count in system-wide
*/
static void print_counter(struct perf_stat_config *config,
struct evsel *counter, char *prefix)
{
FILE *output = config->output;
u64 ena, run, val;
double uval;
int cpu;
struct aggr_cpu_id id;
for (cpu = 0; cpu < evsel__nr_cpus(counter); cpu++) {
struct aggr_data ad = { .cpu = cpu };
if (!collect_data(config, counter, counter_cb, &ad))
return;
val = ad.val;
ena = ad.ena;
run = ad.run;
if (prefix)
fprintf(output, "%s", prefix);
uval = val * counter->scale;
id = cpu_map__empty_aggr_cpu_id();
id.core = cpu;
printout(config, id, 0, counter, uval, prefix,
run, ena, 1.0, &rt_stat);
fputc('\n', output);
}
}
static void print_no_aggr_metric(struct perf_stat_config *config,
struct evlist *evlist,
char *prefix)
{
int cpu;
int nrcpus = 0;
struct evsel *counter;
u64 ena, run, val;
double uval;
struct aggr_cpu_id id;
nrcpus = evlist->core.cpus->nr;
for (cpu = 0; cpu < nrcpus; cpu++) {
bool first = true;
if (prefix)
fputs(prefix, config->output);
evlist__for_each_entry(evlist, counter) {
id = cpu_map__empty_aggr_cpu_id();
id.core = cpu;
if (first) {
aggr_printout(config, counter, id, 0);
first = false;
}
val = perf_counts(counter->counts, cpu, 0)->val;
ena = perf_counts(counter->counts, cpu, 0)->ena;
run = perf_counts(counter->counts, cpu, 0)->run;
uval = val * counter->scale;
printout(config, id, 0, counter, uval, prefix,
run, ena, 1.0, &rt_stat);
}
fputc('\n', config->output);
}
}
static int aggr_header_lens[] = {
[AGGR_CORE] = 24,
[AGGR_DIE] = 18,
[AGGR_SOCKET] = 12,
[AGGR_NONE] = 6,
[AGGR_THREAD] = 24,
[AGGR_GLOBAL] = 0,
};
static const char *aggr_header_csv[] = {
[AGGR_CORE] = "core,cpus,",
[AGGR_DIE] = "die,cpus",
[AGGR_SOCKET] = "socket,cpus",
[AGGR_NONE] = "cpu,",
[AGGR_THREAD] = "comm-pid,",
[AGGR_GLOBAL] = ""
};
static void print_metric_headers(struct perf_stat_config *config,
struct evlist *evlist,
const char *prefix, bool no_indent)
{
struct perf_stat_output_ctx out;
struct evsel *counter;
struct outstate os = {
.fh = config->output
};
if (prefix)
fprintf(config->output, "%s", prefix);
if (!config->csv_output && !no_indent)
fprintf(config->output, "%*s",
aggr_header_lens[config->aggr_mode], "");
if (config->csv_output) {
if (config->interval)
fputs("time,", config->output);
fputs(aggr_header_csv[config->aggr_mode], config->output);
}
/* Print metrics headers only */
evlist__for_each_entry(evlist, counter) {
os.evsel = counter;
out.ctx = &os;
out.print_metric = print_metric_header;
out.new_line = new_line_metric;
out.force_header = true;
perf_stat__print_shadow_stats(config, counter, 0,
0,
&out,
&config->metric_events,
&rt_stat);
}
fputc('\n', config->output);
}
static void print_interval(struct perf_stat_config *config,
struct evlist *evlist,
char *prefix, struct timespec *ts)
{
bool metric_only = config->metric_only;
unsigned int unit_width = config->unit_width;
FILE *output = config->output;
static int num_print_interval;
if (config->interval_clear)
puts(CONSOLE_CLEAR);
sprintf(prefix, "%6lu.%09lu%s", ts->tv_sec, ts->tv_nsec, config->csv_sep);
if ((num_print_interval == 0 && !config->csv_output) || config->interval_clear) {
switch (config->aggr_mode) {
perf stat: Add --per-node agregation support Adding new --per-node option to aggregate counts per NUMA nodes for system-wide mode measurements. You can specify --per-node in live mode: # perf stat -a -I 1000 -e cycles --per-node # time node cpus counts unit events 1.000542550 N0 20 6,202,097 cycles 1.000542550 N1 20 639,559 cycles 2.002040063 N0 20 7,412,495 cycles 2.002040063 N1 20 2,185,577 cycles 3.003451699 N0 20 6,508,917 cycles 3.003451699 N1 20 765,607 cycles ... Or in the record/report stat session: # perf stat record -a -I 1000 -e cycles # time counts unit events 1.000536937 10,008,468 cycles 2.002090152 9,578,539 cycles 3.003625233 7,647,869 cycles 4.005135036 7,032,086 cycles ^C 4.340902364 3,923,893 cycles # perf stat report --per-node # time node cpus counts unit events 1.000536937 N0 20 9,355,086 cycles 1.000536937 N1 20 653,382 cycles 2.002090152 N0 20 7,712,838 cycles 2.002090152 N1 20 1,865,701 cycles 3.003625233 N0 20 6,604,441 cycles 3.003625233 N1 20 1,043,428 cycles 4.005135036 N0 20 6,350,522 cycles 4.005135036 N1 20 681,564 cycles 4.340902364 N0 20 3,403,188 cycles 4.340902364 N1 20 520,705 cycles Signed-off-by: Jiri Olsa <jolsa@kernel.org> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Alexey Budankov <alexey.budankov@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Joe Mario <jmario@redhat.com> Cc: Kan Liang <kan.liang@linux.intel.com> Cc: Michael Petlan <mpetlan@redhat.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/20190904073415.723-4-jolsa@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2019-08-28 08:17:43 +00:00
case AGGR_NODE:
fprintf(output, "# time node cpus");
if (!metric_only)
fprintf(output, " counts %*s events\n", unit_width, "unit");
break;
case AGGR_SOCKET:
fprintf(output, "# time socket cpus");
if (!metric_only)
fprintf(output, " counts %*s events\n", unit_width, "unit");
break;
case AGGR_DIE:
fprintf(output, "# time die cpus");
if (!metric_only)
fprintf(output, " counts %*s events\n", unit_width, "unit");
break;
case AGGR_CORE:
fprintf(output, "# time core cpus");
if (!metric_only)
fprintf(output, " counts %*s events\n", unit_width, "unit");
break;
case AGGR_NONE:
fprintf(output, "# time CPU ");
if (!metric_only)
fprintf(output, " counts %*s events\n", unit_width, "unit");
break;
case AGGR_THREAD:
fprintf(output, "# time comm-pid");
if (!metric_only)
fprintf(output, " counts %*s events\n", unit_width, "unit");
break;
case AGGR_GLOBAL:
default:
fprintf(output, "# time");
if (!metric_only)
fprintf(output, " counts %*s events\n", unit_width, "unit");
case AGGR_UNSET:
break;
}
}
if ((num_print_interval == 0 || config->interval_clear) && metric_only)
print_metric_headers(config, evlist, " ", true);
if (++num_print_interval == 25)
num_print_interval = 0;
}
static void print_header(struct perf_stat_config *config,
struct target *_target,
int argc, const char **argv)
{
FILE *output = config->output;
int i;
fflush(stdout);
if (!config->csv_output) {
fprintf(output, "\n");
fprintf(output, " Performance counter stats for ");
if (_target->system_wide)
fprintf(output, "\'system wide");
else if (_target->cpu_list)
fprintf(output, "\'CPU(s) %s", _target->cpu_list);
else if (!target__has_task(_target)) {
fprintf(output, "\'%s", argv ? argv[0] : "pipe");
for (i = 1; argv && (i < argc); i++)
fprintf(output, " %s", argv[i]);
} else if (_target->pid)
fprintf(output, "process id \'%s", _target->pid);
else
fprintf(output, "thread id \'%s", _target->tid);
fprintf(output, "\'");
if (config->run_count > 1)
fprintf(output, " (%d runs)", config->run_count);
fprintf(output, ":\n\n");
}
}
static int get_precision(double num)
{
if (num > 1)
return 0;
return lround(ceil(-log10(num)));
}
static void print_table(struct perf_stat_config *config,
FILE *output, int precision, double avg)
{
char tmp[64];
int idx, indent = 0;
scnprintf(tmp, 64, " %17.*f", precision, avg);
while (tmp[indent] == ' ')
indent++;
fprintf(output, "%*s# Table of individual measurements:\n", indent, "");
for (idx = 0; idx < config->run_count; idx++) {
double run = (double) config->walltime_run[idx] / NSEC_PER_SEC;
int h, n = 1 + abs((int) (100.0 * (run - avg)/run) / 5);
fprintf(output, " %17.*f (%+.*f) ",
precision, run, precision, run - avg);
for (h = 0; h < n; h++)
fprintf(output, "#");
fprintf(output, "\n");
}
fprintf(output, "\n%*s# Final result:\n", indent, "");
}
static double timeval2double(struct timeval *t)
{
return t->tv_sec + (double) t->tv_usec/USEC_PER_SEC;
}
static void print_footer(struct perf_stat_config *config)
{
double avg = avg_stats(config->walltime_nsecs_stats) / NSEC_PER_SEC;
FILE *output = config->output;
if (!config->null_run)
fprintf(output, "\n");
if (config->run_count == 1) {
fprintf(output, " %17.9f seconds time elapsed", avg);
if (config->ru_display) {
double ru_utime = timeval2double(&config->ru_data.ru_utime);
double ru_stime = timeval2double(&config->ru_data.ru_stime);
fprintf(output, "\n\n");
fprintf(output, " %17.9f seconds user\n", ru_utime);
fprintf(output, " %17.9f seconds sys\n", ru_stime);
}
} else {
double sd = stddev_stats(config->walltime_nsecs_stats) / NSEC_PER_SEC;
/*
* Display at most 2 more significant
* digits than the stddev inaccuracy.
*/
int precision = get_precision(sd) + 2;
if (config->walltime_run_table)
print_table(config, output, precision, avg);
fprintf(output, " %17.*f +- %.*f seconds time elapsed",
precision, avg, precision, sd);
print_noise_pct(config, sd, avg);
}
fprintf(output, "\n\n");
if (config->print_free_counters_hint && sysctl__nmi_watchdog_enabled())
fprintf(output,
"Some events weren't counted. Try disabling the NMI watchdog:\n"
" echo 0 > /proc/sys/kernel/nmi_watchdog\n"
" perf stat ...\n"
" echo 1 > /proc/sys/kernel/nmi_watchdog\n");
if (config->print_mixed_hw_group_error)
fprintf(output,
"The events in group usually have to be from "
"the same PMU. Try reorganizing the group.\n");
}
perf stat: Show percore counts in per CPU output We have supported the event modifier "percore" which sums up the event counts for all hardware threads in a core and show the counts per core. For example, # perf stat -e cpu/event=cpu-cycles,percore/ -a -A -- sleep 1 Performance counter stats for 'system wide': S0-D0-C0 395,072 cpu/event=cpu-cycles,percore/ S0-D0-C1 851,248 cpu/event=cpu-cycles,percore/ S0-D0-C2 954,226 cpu/event=cpu-cycles,percore/ S0-D0-C3 1,233,659 cpu/event=cpu-cycles,percore/ This patch provides a new option "--percore-show-thread". It is used with event modifier "percore" together to sum up the event counts for all hardware threads in a core but show the counts per hardware thread. This is essentially a replacement for the any bit (which is gone in Icelake). Per core counts are useful for some formulas, e.g. CoreIPC. The original percore version was inconvenient to post process. This variant matches the output of the any bit. With this patch, for example, # perf stat -e cpu/event=cpu-cycles,percore/ -a -A --percore-show-thread -- sleep 1 Performance counter stats for 'system wide': CPU0 2,453,061 cpu/event=cpu-cycles,percore/ CPU1 1,823,921 cpu/event=cpu-cycles,percore/ CPU2 1,383,166 cpu/event=cpu-cycles,percore/ CPU3 1,102,652 cpu/event=cpu-cycles,percore/ CPU4 2,453,061 cpu/event=cpu-cycles,percore/ CPU5 1,823,921 cpu/event=cpu-cycles,percore/ CPU6 1,383,166 cpu/event=cpu-cycles,percore/ CPU7 1,102,652 cpu/event=cpu-cycles,percore/ We can see counts are duplicated in CPU pairs (CPU0/CPU4, CPU1/CPU5, CPU2/CPU6, CPU3/CPU7). The interval mode also works. For example, # perf stat -e cpu/event=cpu-cycles,percore/ -a -A --percore-show-thread -I 1000 # time CPU counts unit events 1.000425421 CPU0 925,032 cpu/event=cpu-cycles,percore/ 1.000425421 CPU1 430,202 cpu/event=cpu-cycles,percore/ 1.000425421 CPU2 436,843 cpu/event=cpu-cycles,percore/ 1.000425421 CPU3 1,192,504 cpu/event=cpu-cycles,percore/ 1.000425421 CPU4 925,032 cpu/event=cpu-cycles,percore/ 1.000425421 CPU5 430,202 cpu/event=cpu-cycles,percore/ 1.000425421 CPU6 436,843 cpu/event=cpu-cycles,percore/ 1.000425421 CPU7 1,192,504 cpu/event=cpu-cycles,percore/ If we offline CPU5, the result is: # perf stat -e cpu/event=cpu-cycles,percore/ -a -A --percore-show-thread -- sleep 1 Performance counter stats for 'system wide': CPU0 2,752,148 cpu/event=cpu-cycles,percore/ CPU1 1,009,312 cpu/event=cpu-cycles,percore/ CPU2 2,784,072 cpu/event=cpu-cycles,percore/ CPU3 2,427,922 cpu/event=cpu-cycles,percore/ CPU4 2,752,148 cpu/event=cpu-cycles,percore/ CPU6 2,784,072 cpu/event=cpu-cycles,percore/ CPU7 2,427,922 cpu/event=cpu-cycles,percore/ 1.001416041 seconds time elapsed v4: --- Ravi Bangoria reports an issue in v3. Once we offline a CPU, the output is not correct. The issue is we should use the cpu idx in print_percore_thread rather than using the cpu value. v3: --- 1. Fix the interval mode output error 2. Use cpu value (not cpu index) in config->aggr_get_id(). 3. Refine the code according to Jiri's comments. v2: --- Add the explanation in change log. This is essentially a replacement for the any bit. No code change. Signed-off-by: Jin Yao <yao.jin@linux.intel.com> Tested-by: Ravi Bangoria <ravi.bangoria@linux.ibm.com> Acked-by: Jiri Olsa <jolsa@kernel.org> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Kan Liang <kan.liang@linux.intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lore.kernel.org/lkml/20200214080452.26402-1-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2020-02-14 08:04:52 +00:00
static void print_percore_thread(struct perf_stat_config *config,
struct evsel *counter, char *prefix)
{
int s;
struct aggr_cpu_id s2, id;
perf stat: Show percore counts in per CPU output We have supported the event modifier "percore" which sums up the event counts for all hardware threads in a core and show the counts per core. For example, # perf stat -e cpu/event=cpu-cycles,percore/ -a -A -- sleep 1 Performance counter stats for 'system wide': S0-D0-C0 395,072 cpu/event=cpu-cycles,percore/ S0-D0-C1 851,248 cpu/event=cpu-cycles,percore/ S0-D0-C2 954,226 cpu/event=cpu-cycles,percore/ S0-D0-C3 1,233,659 cpu/event=cpu-cycles,percore/ This patch provides a new option "--percore-show-thread". It is used with event modifier "percore" together to sum up the event counts for all hardware threads in a core but show the counts per hardware thread. This is essentially a replacement for the any bit (which is gone in Icelake). Per core counts are useful for some formulas, e.g. CoreIPC. The original percore version was inconvenient to post process. This variant matches the output of the any bit. With this patch, for example, # perf stat -e cpu/event=cpu-cycles,percore/ -a -A --percore-show-thread -- sleep 1 Performance counter stats for 'system wide': CPU0 2,453,061 cpu/event=cpu-cycles,percore/ CPU1 1,823,921 cpu/event=cpu-cycles,percore/ CPU2 1,383,166 cpu/event=cpu-cycles,percore/ CPU3 1,102,652 cpu/event=cpu-cycles,percore/ CPU4 2,453,061 cpu/event=cpu-cycles,percore/ CPU5 1,823,921 cpu/event=cpu-cycles,percore/ CPU6 1,383,166 cpu/event=cpu-cycles,percore/ CPU7 1,102,652 cpu/event=cpu-cycles,percore/ We can see counts are duplicated in CPU pairs (CPU0/CPU4, CPU1/CPU5, CPU2/CPU6, CPU3/CPU7). The interval mode also works. For example, # perf stat -e cpu/event=cpu-cycles,percore/ -a -A --percore-show-thread -I 1000 # time CPU counts unit events 1.000425421 CPU0 925,032 cpu/event=cpu-cycles,percore/ 1.000425421 CPU1 430,202 cpu/event=cpu-cycles,percore/ 1.000425421 CPU2 436,843 cpu/event=cpu-cycles,percore/ 1.000425421 CPU3 1,192,504 cpu/event=cpu-cycles,percore/ 1.000425421 CPU4 925,032 cpu/event=cpu-cycles,percore/ 1.000425421 CPU5 430,202 cpu/event=cpu-cycles,percore/ 1.000425421 CPU6 436,843 cpu/event=cpu-cycles,percore/ 1.000425421 CPU7 1,192,504 cpu/event=cpu-cycles,percore/ If we offline CPU5, the result is: # perf stat -e cpu/event=cpu-cycles,percore/ -a -A --percore-show-thread -- sleep 1 Performance counter stats for 'system wide': CPU0 2,752,148 cpu/event=cpu-cycles,percore/ CPU1 1,009,312 cpu/event=cpu-cycles,percore/ CPU2 2,784,072 cpu/event=cpu-cycles,percore/ CPU3 2,427,922 cpu/event=cpu-cycles,percore/ CPU4 2,752,148 cpu/event=cpu-cycles,percore/ CPU6 2,784,072 cpu/event=cpu-cycles,percore/ CPU7 2,427,922 cpu/event=cpu-cycles,percore/ 1.001416041 seconds time elapsed v4: --- Ravi Bangoria reports an issue in v3. Once we offline a CPU, the output is not correct. The issue is we should use the cpu idx in print_percore_thread rather than using the cpu value. v3: --- 1. Fix the interval mode output error 2. Use cpu value (not cpu index) in config->aggr_get_id(). 3. Refine the code according to Jiri's comments. v2: --- Add the explanation in change log. This is essentially a replacement for the any bit. No code change. Signed-off-by: Jin Yao <yao.jin@linux.intel.com> Tested-by: Ravi Bangoria <ravi.bangoria@linux.ibm.com> Acked-by: Jiri Olsa <jolsa@kernel.org> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Kan Liang <kan.liang@linux.intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lore.kernel.org/lkml/20200214080452.26402-1-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2020-02-14 08:04:52 +00:00
bool first = true;
for (int i = 0; i < evsel__nr_cpus(counter); i++) {
perf stat: Show percore counts in per CPU output We have supported the event modifier "percore" which sums up the event counts for all hardware threads in a core and show the counts per core. For example, # perf stat -e cpu/event=cpu-cycles,percore/ -a -A -- sleep 1 Performance counter stats for 'system wide': S0-D0-C0 395,072 cpu/event=cpu-cycles,percore/ S0-D0-C1 851,248 cpu/event=cpu-cycles,percore/ S0-D0-C2 954,226 cpu/event=cpu-cycles,percore/ S0-D0-C3 1,233,659 cpu/event=cpu-cycles,percore/ This patch provides a new option "--percore-show-thread". It is used with event modifier "percore" together to sum up the event counts for all hardware threads in a core but show the counts per hardware thread. This is essentially a replacement for the any bit (which is gone in Icelake). Per core counts are useful for some formulas, e.g. CoreIPC. The original percore version was inconvenient to post process. This variant matches the output of the any bit. With this patch, for example, # perf stat -e cpu/event=cpu-cycles,percore/ -a -A --percore-show-thread -- sleep 1 Performance counter stats for 'system wide': CPU0 2,453,061 cpu/event=cpu-cycles,percore/ CPU1 1,823,921 cpu/event=cpu-cycles,percore/ CPU2 1,383,166 cpu/event=cpu-cycles,percore/ CPU3 1,102,652 cpu/event=cpu-cycles,percore/ CPU4 2,453,061 cpu/event=cpu-cycles,percore/ CPU5 1,823,921 cpu/event=cpu-cycles,percore/ CPU6 1,383,166 cpu/event=cpu-cycles,percore/ CPU7 1,102,652 cpu/event=cpu-cycles,percore/ We can see counts are duplicated in CPU pairs (CPU0/CPU4, CPU1/CPU5, CPU2/CPU6, CPU3/CPU7). The interval mode also works. For example, # perf stat -e cpu/event=cpu-cycles,percore/ -a -A --percore-show-thread -I 1000 # time CPU counts unit events 1.000425421 CPU0 925,032 cpu/event=cpu-cycles,percore/ 1.000425421 CPU1 430,202 cpu/event=cpu-cycles,percore/ 1.000425421 CPU2 436,843 cpu/event=cpu-cycles,percore/ 1.000425421 CPU3 1,192,504 cpu/event=cpu-cycles,percore/ 1.000425421 CPU4 925,032 cpu/event=cpu-cycles,percore/ 1.000425421 CPU5 430,202 cpu/event=cpu-cycles,percore/ 1.000425421 CPU6 436,843 cpu/event=cpu-cycles,percore/ 1.000425421 CPU7 1,192,504 cpu/event=cpu-cycles,percore/ If we offline CPU5, the result is: # perf stat -e cpu/event=cpu-cycles,percore/ -a -A --percore-show-thread -- sleep 1 Performance counter stats for 'system wide': CPU0 2,752,148 cpu/event=cpu-cycles,percore/ CPU1 1,009,312 cpu/event=cpu-cycles,percore/ CPU2 2,784,072 cpu/event=cpu-cycles,percore/ CPU3 2,427,922 cpu/event=cpu-cycles,percore/ CPU4 2,752,148 cpu/event=cpu-cycles,percore/ CPU6 2,784,072 cpu/event=cpu-cycles,percore/ CPU7 2,427,922 cpu/event=cpu-cycles,percore/ 1.001416041 seconds time elapsed v4: --- Ravi Bangoria reports an issue in v3. Once we offline a CPU, the output is not correct. The issue is we should use the cpu idx in print_percore_thread rather than using the cpu value. v3: --- 1. Fix the interval mode output error 2. Use cpu value (not cpu index) in config->aggr_get_id(). 3. Refine the code according to Jiri's comments. v2: --- Add the explanation in change log. This is essentially a replacement for the any bit. No code change. Signed-off-by: Jin Yao <yao.jin@linux.intel.com> Tested-by: Ravi Bangoria <ravi.bangoria@linux.ibm.com> Acked-by: Jiri Olsa <jolsa@kernel.org> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Kan Liang <kan.liang@linux.intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lore.kernel.org/lkml/20200214080452.26402-1-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2020-02-14 08:04:52 +00:00
s2 = config->aggr_get_id(config, evsel__cpus(counter), i);
for (s = 0; s < config->aggr_map->nr; s++) {
id = config->aggr_map->map[s];
if (cpu_map__compare_aggr_cpu_id(s2, id))
perf stat: Show percore counts in per CPU output We have supported the event modifier "percore" which sums up the event counts for all hardware threads in a core and show the counts per core. For example, # perf stat -e cpu/event=cpu-cycles,percore/ -a -A -- sleep 1 Performance counter stats for 'system wide': S0-D0-C0 395,072 cpu/event=cpu-cycles,percore/ S0-D0-C1 851,248 cpu/event=cpu-cycles,percore/ S0-D0-C2 954,226 cpu/event=cpu-cycles,percore/ S0-D0-C3 1,233,659 cpu/event=cpu-cycles,percore/ This patch provides a new option "--percore-show-thread". It is used with event modifier "percore" together to sum up the event counts for all hardware threads in a core but show the counts per hardware thread. This is essentially a replacement for the any bit (which is gone in Icelake). Per core counts are useful for some formulas, e.g. CoreIPC. The original percore version was inconvenient to post process. This variant matches the output of the any bit. With this patch, for example, # perf stat -e cpu/event=cpu-cycles,percore/ -a -A --percore-show-thread -- sleep 1 Performance counter stats for 'system wide': CPU0 2,453,061 cpu/event=cpu-cycles,percore/ CPU1 1,823,921 cpu/event=cpu-cycles,percore/ CPU2 1,383,166 cpu/event=cpu-cycles,percore/ CPU3 1,102,652 cpu/event=cpu-cycles,percore/ CPU4 2,453,061 cpu/event=cpu-cycles,percore/ CPU5 1,823,921 cpu/event=cpu-cycles,percore/ CPU6 1,383,166 cpu/event=cpu-cycles,percore/ CPU7 1,102,652 cpu/event=cpu-cycles,percore/ We can see counts are duplicated in CPU pairs (CPU0/CPU4, CPU1/CPU5, CPU2/CPU6, CPU3/CPU7). The interval mode also works. For example, # perf stat -e cpu/event=cpu-cycles,percore/ -a -A --percore-show-thread -I 1000 # time CPU counts unit events 1.000425421 CPU0 925,032 cpu/event=cpu-cycles,percore/ 1.000425421 CPU1 430,202 cpu/event=cpu-cycles,percore/ 1.000425421 CPU2 436,843 cpu/event=cpu-cycles,percore/ 1.000425421 CPU3 1,192,504 cpu/event=cpu-cycles,percore/ 1.000425421 CPU4 925,032 cpu/event=cpu-cycles,percore/ 1.000425421 CPU5 430,202 cpu/event=cpu-cycles,percore/ 1.000425421 CPU6 436,843 cpu/event=cpu-cycles,percore/ 1.000425421 CPU7 1,192,504 cpu/event=cpu-cycles,percore/ If we offline CPU5, the result is: # perf stat -e cpu/event=cpu-cycles,percore/ -a -A --percore-show-thread -- sleep 1 Performance counter stats for 'system wide': CPU0 2,752,148 cpu/event=cpu-cycles,percore/ CPU1 1,009,312 cpu/event=cpu-cycles,percore/ CPU2 2,784,072 cpu/event=cpu-cycles,percore/ CPU3 2,427,922 cpu/event=cpu-cycles,percore/ CPU4 2,752,148 cpu/event=cpu-cycles,percore/ CPU6 2,784,072 cpu/event=cpu-cycles,percore/ CPU7 2,427,922 cpu/event=cpu-cycles,percore/ 1.001416041 seconds time elapsed v4: --- Ravi Bangoria reports an issue in v3. Once we offline a CPU, the output is not correct. The issue is we should use the cpu idx in print_percore_thread rather than using the cpu value. v3: --- 1. Fix the interval mode output error 2. Use cpu value (not cpu index) in config->aggr_get_id(). 3. Refine the code according to Jiri's comments. v2: --- Add the explanation in change log. This is essentially a replacement for the any bit. No code change. Signed-off-by: Jin Yao <yao.jin@linux.intel.com> Tested-by: Ravi Bangoria <ravi.bangoria@linux.ibm.com> Acked-by: Jiri Olsa <jolsa@kernel.org> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Kan Liang <kan.liang@linux.intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lore.kernel.org/lkml/20200214080452.26402-1-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2020-02-14 08:04:52 +00:00
break;
}
print_counter_aggrdata(config, counter, s,
prefix, false,
&first, i);
}
}
perf stat: Support 'percore' event qualifier With this patch, we can use the 'percore' event qualifier in perf-stat. root@skl:/tmp# perf stat -e cpu/event=0,umask=0x3,percore=1/,cpu/event=0,umask=0x3/ -a -A -I1000 1.000773050 S0-C0 98,352,832 cpu/event=0,umask=0x3,percore=1/ (50.01%) 1.000773050 S0-C1 103,763,057 cpu/event=0,umask=0x3,percore=1/ (50.02%) 1.000773050 S0-C2 196,776,995 cpu/event=0,umask=0x3,percore=1/ (50.02%) 1.000773050 S0-C3 176,493,779 cpu/event=0,umask=0x3,percore=1/ (50.02%) 1.000773050 CPU0 47,699,641 cpu/event=0,umask=0x3/ (50.02%) 1.000773050 CPU1 49,052,451 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU2 102,771,422 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU3 100,784,662 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU4 43,171,342 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU5 54,152,158 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU6 93,618,410 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU7 74,477,589 cpu/event=0,umask=0x3/ (49.99%) In this example, we count the event 'ref-cycles' per-core and per-CPU in one perf stat command-line. From the output, we can see: S0-C0 = CPU0 + CPU4 S0-C1 = CPU1 + CPU5 S0-C2 = CPU2 + CPU6 S0-C3 = CPU3 + CPU7 So the result is expected (tiny difference is ignored). Note that, the 'percore' event qualifier needs to use with option '-A'. Signed-off-by: Jin Yao <yao.jin@linux.intel.com> Tested-by: Ravi Bangoria <ravi.bangoria@linux.ibm.com> Acked-by: Jiri Olsa <jolsa@kernel.org> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Jin Yao <yao.jin@intel.com> Cc: Kan Liang <kan.liang@linux.intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1555077590-27664-4-git-send-email-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2019-04-12 13:59:49 +00:00
static void print_percore(struct perf_stat_config *config,
struct evsel *counter, char *prefix)
perf stat: Support 'percore' event qualifier With this patch, we can use the 'percore' event qualifier in perf-stat. root@skl:/tmp# perf stat -e cpu/event=0,umask=0x3,percore=1/,cpu/event=0,umask=0x3/ -a -A -I1000 1.000773050 S0-C0 98,352,832 cpu/event=0,umask=0x3,percore=1/ (50.01%) 1.000773050 S0-C1 103,763,057 cpu/event=0,umask=0x3,percore=1/ (50.02%) 1.000773050 S0-C2 196,776,995 cpu/event=0,umask=0x3,percore=1/ (50.02%) 1.000773050 S0-C3 176,493,779 cpu/event=0,umask=0x3,percore=1/ (50.02%) 1.000773050 CPU0 47,699,641 cpu/event=0,umask=0x3/ (50.02%) 1.000773050 CPU1 49,052,451 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU2 102,771,422 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU3 100,784,662 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU4 43,171,342 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU5 54,152,158 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU6 93,618,410 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU7 74,477,589 cpu/event=0,umask=0x3/ (49.99%) In this example, we count the event 'ref-cycles' per-core and per-CPU in one perf stat command-line. From the output, we can see: S0-C0 = CPU0 + CPU4 S0-C1 = CPU1 + CPU5 S0-C2 = CPU2 + CPU6 S0-C3 = CPU3 + CPU7 So the result is expected (tiny difference is ignored). Note that, the 'percore' event qualifier needs to use with option '-A'. Signed-off-by: Jin Yao <yao.jin@linux.intel.com> Tested-by: Ravi Bangoria <ravi.bangoria@linux.ibm.com> Acked-by: Jiri Olsa <jolsa@kernel.org> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Jin Yao <yao.jin@intel.com> Cc: Kan Liang <kan.liang@linux.intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1555077590-27664-4-git-send-email-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2019-04-12 13:59:49 +00:00
{
bool metric_only = config->metric_only;
FILE *output = config->output;
int s;
bool first = true;
if (!config->aggr_map || !config->aggr_get_id)
perf stat: Support 'percore' event qualifier With this patch, we can use the 'percore' event qualifier in perf-stat. root@skl:/tmp# perf stat -e cpu/event=0,umask=0x3,percore=1/,cpu/event=0,umask=0x3/ -a -A -I1000 1.000773050 S0-C0 98,352,832 cpu/event=0,umask=0x3,percore=1/ (50.01%) 1.000773050 S0-C1 103,763,057 cpu/event=0,umask=0x3,percore=1/ (50.02%) 1.000773050 S0-C2 196,776,995 cpu/event=0,umask=0x3,percore=1/ (50.02%) 1.000773050 S0-C3 176,493,779 cpu/event=0,umask=0x3,percore=1/ (50.02%) 1.000773050 CPU0 47,699,641 cpu/event=0,umask=0x3/ (50.02%) 1.000773050 CPU1 49,052,451 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU2 102,771,422 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU3 100,784,662 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU4 43,171,342 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU5 54,152,158 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU6 93,618,410 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU7 74,477,589 cpu/event=0,umask=0x3/ (49.99%) In this example, we count the event 'ref-cycles' per-core and per-CPU in one perf stat command-line. From the output, we can see: S0-C0 = CPU0 + CPU4 S0-C1 = CPU1 + CPU5 S0-C2 = CPU2 + CPU6 S0-C3 = CPU3 + CPU7 So the result is expected (tiny difference is ignored). Note that, the 'percore' event qualifier needs to use with option '-A'. Signed-off-by: Jin Yao <yao.jin@linux.intel.com> Tested-by: Ravi Bangoria <ravi.bangoria@linux.ibm.com> Acked-by: Jiri Olsa <jolsa@kernel.org> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Jin Yao <yao.jin@intel.com> Cc: Kan Liang <kan.liang@linux.intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1555077590-27664-4-git-send-email-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2019-04-12 13:59:49 +00:00
return;
perf stat: Show percore counts in per CPU output We have supported the event modifier "percore" which sums up the event counts for all hardware threads in a core and show the counts per core. For example, # perf stat -e cpu/event=cpu-cycles,percore/ -a -A -- sleep 1 Performance counter stats for 'system wide': S0-D0-C0 395,072 cpu/event=cpu-cycles,percore/ S0-D0-C1 851,248 cpu/event=cpu-cycles,percore/ S0-D0-C2 954,226 cpu/event=cpu-cycles,percore/ S0-D0-C3 1,233,659 cpu/event=cpu-cycles,percore/ This patch provides a new option "--percore-show-thread". It is used with event modifier "percore" together to sum up the event counts for all hardware threads in a core but show the counts per hardware thread. This is essentially a replacement for the any bit (which is gone in Icelake). Per core counts are useful for some formulas, e.g. CoreIPC. The original percore version was inconvenient to post process. This variant matches the output of the any bit. With this patch, for example, # perf stat -e cpu/event=cpu-cycles,percore/ -a -A --percore-show-thread -- sleep 1 Performance counter stats for 'system wide': CPU0 2,453,061 cpu/event=cpu-cycles,percore/ CPU1 1,823,921 cpu/event=cpu-cycles,percore/ CPU2 1,383,166 cpu/event=cpu-cycles,percore/ CPU3 1,102,652 cpu/event=cpu-cycles,percore/ CPU4 2,453,061 cpu/event=cpu-cycles,percore/ CPU5 1,823,921 cpu/event=cpu-cycles,percore/ CPU6 1,383,166 cpu/event=cpu-cycles,percore/ CPU7 1,102,652 cpu/event=cpu-cycles,percore/ We can see counts are duplicated in CPU pairs (CPU0/CPU4, CPU1/CPU5, CPU2/CPU6, CPU3/CPU7). The interval mode also works. For example, # perf stat -e cpu/event=cpu-cycles,percore/ -a -A --percore-show-thread -I 1000 # time CPU counts unit events 1.000425421 CPU0 925,032 cpu/event=cpu-cycles,percore/ 1.000425421 CPU1 430,202 cpu/event=cpu-cycles,percore/ 1.000425421 CPU2 436,843 cpu/event=cpu-cycles,percore/ 1.000425421 CPU3 1,192,504 cpu/event=cpu-cycles,percore/ 1.000425421 CPU4 925,032 cpu/event=cpu-cycles,percore/ 1.000425421 CPU5 430,202 cpu/event=cpu-cycles,percore/ 1.000425421 CPU6 436,843 cpu/event=cpu-cycles,percore/ 1.000425421 CPU7 1,192,504 cpu/event=cpu-cycles,percore/ If we offline CPU5, the result is: # perf stat -e cpu/event=cpu-cycles,percore/ -a -A --percore-show-thread -- sleep 1 Performance counter stats for 'system wide': CPU0 2,752,148 cpu/event=cpu-cycles,percore/ CPU1 1,009,312 cpu/event=cpu-cycles,percore/ CPU2 2,784,072 cpu/event=cpu-cycles,percore/ CPU3 2,427,922 cpu/event=cpu-cycles,percore/ CPU4 2,752,148 cpu/event=cpu-cycles,percore/ CPU6 2,784,072 cpu/event=cpu-cycles,percore/ CPU7 2,427,922 cpu/event=cpu-cycles,percore/ 1.001416041 seconds time elapsed v4: --- Ravi Bangoria reports an issue in v3. Once we offline a CPU, the output is not correct. The issue is we should use the cpu idx in print_percore_thread rather than using the cpu value. v3: --- 1. Fix the interval mode output error 2. Use cpu value (not cpu index) in config->aggr_get_id(). 3. Refine the code according to Jiri's comments. v2: --- Add the explanation in change log. This is essentially a replacement for the any bit. No code change. Signed-off-by: Jin Yao <yao.jin@linux.intel.com> Tested-by: Ravi Bangoria <ravi.bangoria@linux.ibm.com> Acked-by: Jiri Olsa <jolsa@kernel.org> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Kan Liang <kan.liang@linux.intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lore.kernel.org/lkml/20200214080452.26402-1-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2020-02-14 08:04:52 +00:00
if (config->percore_show_thread)
return print_percore_thread(config, counter, prefix);
perf stat: Support 'percore' event qualifier With this patch, we can use the 'percore' event qualifier in perf-stat. root@skl:/tmp# perf stat -e cpu/event=0,umask=0x3,percore=1/,cpu/event=0,umask=0x3/ -a -A -I1000 1.000773050 S0-C0 98,352,832 cpu/event=0,umask=0x3,percore=1/ (50.01%) 1.000773050 S0-C1 103,763,057 cpu/event=0,umask=0x3,percore=1/ (50.02%) 1.000773050 S0-C2 196,776,995 cpu/event=0,umask=0x3,percore=1/ (50.02%) 1.000773050 S0-C3 176,493,779 cpu/event=0,umask=0x3,percore=1/ (50.02%) 1.000773050 CPU0 47,699,641 cpu/event=0,umask=0x3/ (50.02%) 1.000773050 CPU1 49,052,451 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU2 102,771,422 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU3 100,784,662 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU4 43,171,342 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU5 54,152,158 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU6 93,618,410 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU7 74,477,589 cpu/event=0,umask=0x3/ (49.99%) In this example, we count the event 'ref-cycles' per-core and per-CPU in one perf stat command-line. From the output, we can see: S0-C0 = CPU0 + CPU4 S0-C1 = CPU1 + CPU5 S0-C2 = CPU2 + CPU6 S0-C3 = CPU3 + CPU7 So the result is expected (tiny difference is ignored). Note that, the 'percore' event qualifier needs to use with option '-A'. Signed-off-by: Jin Yao <yao.jin@linux.intel.com> Tested-by: Ravi Bangoria <ravi.bangoria@linux.ibm.com> Acked-by: Jiri Olsa <jolsa@kernel.org> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Jin Yao <yao.jin@intel.com> Cc: Kan Liang <kan.liang@linux.intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1555077590-27664-4-git-send-email-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2019-04-12 13:59:49 +00:00
for (s = 0; s < config->aggr_map->nr; s++) {
if (prefix && metric_only)
fprintf(output, "%s", prefix);
print_counter_aggrdata(config, counter, s,
prefix, metric_only,
perf stat: Show percore counts in per CPU output We have supported the event modifier "percore" which sums up the event counts for all hardware threads in a core and show the counts per core. For example, # perf stat -e cpu/event=cpu-cycles,percore/ -a -A -- sleep 1 Performance counter stats for 'system wide': S0-D0-C0 395,072 cpu/event=cpu-cycles,percore/ S0-D0-C1 851,248 cpu/event=cpu-cycles,percore/ S0-D0-C2 954,226 cpu/event=cpu-cycles,percore/ S0-D0-C3 1,233,659 cpu/event=cpu-cycles,percore/ This patch provides a new option "--percore-show-thread". It is used with event modifier "percore" together to sum up the event counts for all hardware threads in a core but show the counts per hardware thread. This is essentially a replacement for the any bit (which is gone in Icelake). Per core counts are useful for some formulas, e.g. CoreIPC. The original percore version was inconvenient to post process. This variant matches the output of the any bit. With this patch, for example, # perf stat -e cpu/event=cpu-cycles,percore/ -a -A --percore-show-thread -- sleep 1 Performance counter stats for 'system wide': CPU0 2,453,061 cpu/event=cpu-cycles,percore/ CPU1 1,823,921 cpu/event=cpu-cycles,percore/ CPU2 1,383,166 cpu/event=cpu-cycles,percore/ CPU3 1,102,652 cpu/event=cpu-cycles,percore/ CPU4 2,453,061 cpu/event=cpu-cycles,percore/ CPU5 1,823,921 cpu/event=cpu-cycles,percore/ CPU6 1,383,166 cpu/event=cpu-cycles,percore/ CPU7 1,102,652 cpu/event=cpu-cycles,percore/ We can see counts are duplicated in CPU pairs (CPU0/CPU4, CPU1/CPU5, CPU2/CPU6, CPU3/CPU7). The interval mode also works. For example, # perf stat -e cpu/event=cpu-cycles,percore/ -a -A --percore-show-thread -I 1000 # time CPU counts unit events 1.000425421 CPU0 925,032 cpu/event=cpu-cycles,percore/ 1.000425421 CPU1 430,202 cpu/event=cpu-cycles,percore/ 1.000425421 CPU2 436,843 cpu/event=cpu-cycles,percore/ 1.000425421 CPU3 1,192,504 cpu/event=cpu-cycles,percore/ 1.000425421 CPU4 925,032 cpu/event=cpu-cycles,percore/ 1.000425421 CPU5 430,202 cpu/event=cpu-cycles,percore/ 1.000425421 CPU6 436,843 cpu/event=cpu-cycles,percore/ 1.000425421 CPU7 1,192,504 cpu/event=cpu-cycles,percore/ If we offline CPU5, the result is: # perf stat -e cpu/event=cpu-cycles,percore/ -a -A --percore-show-thread -- sleep 1 Performance counter stats for 'system wide': CPU0 2,752,148 cpu/event=cpu-cycles,percore/ CPU1 1,009,312 cpu/event=cpu-cycles,percore/ CPU2 2,784,072 cpu/event=cpu-cycles,percore/ CPU3 2,427,922 cpu/event=cpu-cycles,percore/ CPU4 2,752,148 cpu/event=cpu-cycles,percore/ CPU6 2,784,072 cpu/event=cpu-cycles,percore/ CPU7 2,427,922 cpu/event=cpu-cycles,percore/ 1.001416041 seconds time elapsed v4: --- Ravi Bangoria reports an issue in v3. Once we offline a CPU, the output is not correct. The issue is we should use the cpu idx in print_percore_thread rather than using the cpu value. v3: --- 1. Fix the interval mode output error 2. Use cpu value (not cpu index) in config->aggr_get_id(). 3. Refine the code according to Jiri's comments. v2: --- Add the explanation in change log. This is essentially a replacement for the any bit. No code change. Signed-off-by: Jin Yao <yao.jin@linux.intel.com> Tested-by: Ravi Bangoria <ravi.bangoria@linux.ibm.com> Acked-by: Jiri Olsa <jolsa@kernel.org> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Kan Liang <kan.liang@linux.intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lore.kernel.org/lkml/20200214080452.26402-1-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2020-02-14 08:04:52 +00:00
&first, -1);
perf stat: Support 'percore' event qualifier With this patch, we can use the 'percore' event qualifier in perf-stat. root@skl:/tmp# perf stat -e cpu/event=0,umask=0x3,percore=1/,cpu/event=0,umask=0x3/ -a -A -I1000 1.000773050 S0-C0 98,352,832 cpu/event=0,umask=0x3,percore=1/ (50.01%) 1.000773050 S0-C1 103,763,057 cpu/event=0,umask=0x3,percore=1/ (50.02%) 1.000773050 S0-C2 196,776,995 cpu/event=0,umask=0x3,percore=1/ (50.02%) 1.000773050 S0-C3 176,493,779 cpu/event=0,umask=0x3,percore=1/ (50.02%) 1.000773050 CPU0 47,699,641 cpu/event=0,umask=0x3/ (50.02%) 1.000773050 CPU1 49,052,451 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU2 102,771,422 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU3 100,784,662 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU4 43,171,342 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU5 54,152,158 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU6 93,618,410 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU7 74,477,589 cpu/event=0,umask=0x3/ (49.99%) In this example, we count the event 'ref-cycles' per-core and per-CPU in one perf stat command-line. From the output, we can see: S0-C0 = CPU0 + CPU4 S0-C1 = CPU1 + CPU5 S0-C2 = CPU2 + CPU6 S0-C3 = CPU3 + CPU7 So the result is expected (tiny difference is ignored). Note that, the 'percore' event qualifier needs to use with option '-A'. Signed-off-by: Jin Yao <yao.jin@linux.intel.com> Tested-by: Ravi Bangoria <ravi.bangoria@linux.ibm.com> Acked-by: Jiri Olsa <jolsa@kernel.org> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Jin Yao <yao.jin@intel.com> Cc: Kan Liang <kan.liang@linux.intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1555077590-27664-4-git-send-email-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2019-04-12 13:59:49 +00:00
}
if (metric_only)
fputc('\n', output);
}
void evlist__print_counters(struct evlist *evlist, struct perf_stat_config *config,
struct target *_target, struct timespec *ts, int argc, const char **argv)
{
bool metric_only = config->metric_only;
int interval = config->interval;
struct evsel *counter;
char buf[64], *prefix = NULL;
if (interval)
print_interval(config, evlist, prefix = buf, ts);
else
print_header(config, _target, argc, argv);
if (metric_only) {
static int num_print_iv;
if (num_print_iv == 0 && !interval)
print_metric_headers(config, evlist, prefix, false);
if (num_print_iv++ == 25)
num_print_iv = 0;
if (config->aggr_mode == AGGR_GLOBAL && prefix)
fprintf(config->output, "%s", prefix);
}
switch (config->aggr_mode) {
case AGGR_CORE:
case AGGR_DIE:
case AGGR_SOCKET:
perf stat: Add --per-node agregation support Adding new --per-node option to aggregate counts per NUMA nodes for system-wide mode measurements. You can specify --per-node in live mode: # perf stat -a -I 1000 -e cycles --per-node # time node cpus counts unit events 1.000542550 N0 20 6,202,097 cycles 1.000542550 N1 20 639,559 cycles 2.002040063 N0 20 7,412,495 cycles 2.002040063 N1 20 2,185,577 cycles 3.003451699 N0 20 6,508,917 cycles 3.003451699 N1 20 765,607 cycles ... Or in the record/report stat session: # perf stat record -a -I 1000 -e cycles # time counts unit events 1.000536937 10,008,468 cycles 2.002090152 9,578,539 cycles 3.003625233 7,647,869 cycles 4.005135036 7,032,086 cycles ^C 4.340902364 3,923,893 cycles # perf stat report --per-node # time node cpus counts unit events 1.000536937 N0 20 9,355,086 cycles 1.000536937 N1 20 653,382 cycles 2.002090152 N0 20 7,712,838 cycles 2.002090152 N1 20 1,865,701 cycles 3.003625233 N0 20 6,604,441 cycles 3.003625233 N1 20 1,043,428 cycles 4.005135036 N0 20 6,350,522 cycles 4.005135036 N1 20 681,564 cycles 4.340902364 N0 20 3,403,188 cycles 4.340902364 N1 20 520,705 cycles Signed-off-by: Jiri Olsa <jolsa@kernel.org> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Alexey Budankov <alexey.budankov@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Joe Mario <jmario@redhat.com> Cc: Kan Liang <kan.liang@linux.intel.com> Cc: Michael Petlan <mpetlan@redhat.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/20190904073415.723-4-jolsa@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2019-08-28 08:17:43 +00:00
case AGGR_NODE:
print_aggr(config, evlist, prefix);
break;
case AGGR_THREAD:
evlist__for_each_entry(evlist, counter) {
print_aggr_thread(config, _target, counter, prefix);
}
break;
case AGGR_GLOBAL:
evlist__for_each_entry(evlist, counter) {
print_counter_aggr(config, counter, prefix);
}
if (metric_only)
fputc('\n', config->output);
break;
case AGGR_NONE:
if (metric_only)
print_no_aggr_metric(config, evlist, prefix);
else {
evlist__for_each_entry(evlist, counter) {
perf stat: Support 'percore' event qualifier With this patch, we can use the 'percore' event qualifier in perf-stat. root@skl:/tmp# perf stat -e cpu/event=0,umask=0x3,percore=1/,cpu/event=0,umask=0x3/ -a -A -I1000 1.000773050 S0-C0 98,352,832 cpu/event=0,umask=0x3,percore=1/ (50.01%) 1.000773050 S0-C1 103,763,057 cpu/event=0,umask=0x3,percore=1/ (50.02%) 1.000773050 S0-C2 196,776,995 cpu/event=0,umask=0x3,percore=1/ (50.02%) 1.000773050 S0-C3 176,493,779 cpu/event=0,umask=0x3,percore=1/ (50.02%) 1.000773050 CPU0 47,699,641 cpu/event=0,umask=0x3/ (50.02%) 1.000773050 CPU1 49,052,451 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU2 102,771,422 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU3 100,784,662 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU4 43,171,342 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU5 54,152,158 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU6 93,618,410 cpu/event=0,umask=0x3/ (49.98%) 1.000773050 CPU7 74,477,589 cpu/event=0,umask=0x3/ (49.99%) In this example, we count the event 'ref-cycles' per-core and per-CPU in one perf stat command-line. From the output, we can see: S0-C0 = CPU0 + CPU4 S0-C1 = CPU1 + CPU5 S0-C2 = CPU2 + CPU6 S0-C3 = CPU3 + CPU7 So the result is expected (tiny difference is ignored). Note that, the 'percore' event qualifier needs to use with option '-A'. Signed-off-by: Jin Yao <yao.jin@linux.intel.com> Tested-by: Ravi Bangoria <ravi.bangoria@linux.ibm.com> Acked-by: Jiri Olsa <jolsa@kernel.org> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Jin Yao <yao.jin@intel.com> Cc: Kan Liang <kan.liang@linux.intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1555077590-27664-4-git-send-email-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2019-04-12 13:59:49 +00:00
if (counter->percore)
print_percore(config, counter, prefix);
else
print_counter(config, counter, prefix);
}
}
break;
case AGGR_UNSET:
default:
break;
}
if (!interval && !config->csv_output)
print_footer(config);
fflush(config->output);
}