linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* [PATCH v3] perf stat: Append to default list if use -e event
@ 2021-01-22  2:08 Jin Yao
  0 siblings, 0 replies; only message in thread
From: Jin Yao @ 2021-01-22  2:08 UTC (permalink / raw)
  To: acme, jolsa, peterz, mingo, alexander.shishkin
  Cc: Linux-kernel, ak, kan.liang, yao.jin, Jin Yao

The default event list includes the most common events which are widely
used by users. But with -e option, the current perf only counts the events
assigned by -e option. Users may want to collect some extra events with
the default list. For this case, users have to manually add all the events
from the default list. It's inconvenient. Also, users may don't know how to
get the default list.

Currently perf-stat has supported '-d' option which prints more detailed
statistics such as L1-dcache-loads, L1-dcache-load-misses. We just need
one line change to let '-d' support printing the default list.

Before:

root@kbl-ppc:~# ./perf stat -e power/energy-pkg/ -a -- sleep 1

 Performance counter stats for 'system wide':

              2.04 Joules power/energy-pkg/

       1.000863884 seconds time elapsed

After:

root@kbl-ppc:~# ./perf stat -e power/energy-pkg/ -a -d -- sleep 1

 Performance counter stats for 'system wide':

              2.24 Joules power/energy-pkg/         #    0.000 K/sec
          8,009.72 msec   cpu-clock                 #    7.995 CPUs utilized
               288        context-switches          #    0.036 K/sec
                10        cpu-migrations            #    0.001 K/sec
                 2        page-faults               #    0.000 K/sec
       125,889,836        cycles                    #    0.016 GHz                      (49.60%)
        23,216,112        instructions              #    0.18  insn per cycle           (62.40%)
         4,571,447        branches                  #    0.571 M/sec                    (62.80%)
           652,350        branch-misses             #   14.27% of all branches          (62.85%)
         6,564,491        L1-dcache-loads           #    0.820 M/sec                    (62.85%)
         1,030,251        L1-dcache-load-misses     #   15.69% of all L1-dcache accesses  (62.78%)
           567,451        LLC-loads                 #    0.071 M/sec                    (49.58%)
            52,983        LLC-load-misses           #    9.34% of all LL-cache accesses  (49.53%)

       1.001854149 seconds time elapsed

And it can append the metrics to the default event list easily.

root@kbl-ppc:~# ./perf stat -M Page_Walks_Utilization -a -d -- sleep 1

 Performance counter stats for 'system wide':

         1,332,948      itlb_misses.walk_pending  #    0.166 M/sec
                                                  #     0.05 Page_Walks_Utilization   (30.68%)
         1,144,389      dtlb_store_misses.walk_pending #    0.143 M/sec                    (31.09%)
       123,661,801      cycles                    #    0.015 GHz                      (31.49%)
         8,772,539      dtlb_load_misses.walk_pending #    1.096 M/sec                    (31.67%)
                 0      ept.walk_pending          #    0.000 K/sec                    (31.36%)
          8,007.56 msec cpu-clock                 #    7.992 CPUs utilized
               270      context-switches          #    0.034 K/sec
                 8      cpu-migrations            #    0.001 K/sec
                 3      page-faults               #    0.000 K/sec
       124,703,802      cycles                    #    0.016 GHz                      (30.95%)
        23,239,045      instructions              #    0.19  insn per cycle           (38.54%)
         4,500,430      branches                  #    0.562 M/sec                    (38.14%)
           658,460      branch-misses             #   14.63% of all branches          (37.96%)
         6,296,710      L1-dcache-loads           #    0.786 M/sec                    (37.96%)
         1,027,741      L1-dcache-load-misses     #   16.32% of all L1-dcache accesses  (37.96%)
           551,872      LLC-loads                 #    0.069 M/sec                    (30.37%)
            48,296      LLC-load-misses           #    8.75% of all LL-cache accesses  (30.37%)

       1.001971812 seconds time elapsed

Signed-off-by: Jin Yao <yao.jin@linux.intel.com>
---
 v3:
   Improve '-d' to let it support printing default event list.
   
 v2:
   We just use a simple syntax: -e +event to append the -e event
   to default list.

 v1:
   Create a new option '--add-default' to append -e event to
   default list.

 tools/perf/builtin-stat.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index a380e38ca937..24f9ee90dbf0 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -1812,7 +1812,7 @@ static int add_default_attributes(void)
 		free(str);
 	}
 
-	if (!evsel_list->core.nr_entries) {
+	if (!evsel_list->core.nr_entries || detailed_run) {
 		if (target__has_cpu(&target))
 			default_attrs0[0].config = PERF_COUNT_SW_CPU_CLOCK;
 
-- 
2.17.1


^ permalink raw reply related	[flat|nested] only message in thread

only message in thread, other threads:[~2021-01-22  2:11 UTC | newest]

Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-01-22  2:08 [PATCH v3] perf stat: Append to default list if use -e event Jin Yao

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).