perf stat: Don't print bogus data on -e instructions
authorRamkumar Ramachandra <artagnon@gmail.com>
Tue, 1 Oct 2013 08:36:44 +0000 (14:06 +0530)
committerArnaldo Carvalho de Melo <acme@redhat.com>
Fri, 11 Oct 2013 15:17:35 +0000 (12:17 -0300)
When only the instructions event is requested:

$ perf stat -e instructions git s
M  builtin-stat.c

 Performance counter stats for 'git s':

       917,453,420 instructions              #    0.00  insns per cycle

       0.213002926 seconds time elapsed

The 0.00 insns per cycle comment in the output is totally bogus and
misleading. It happens because update_shadow_stats() doesn't touch
runtime_cycles_stats when only the instructions event is requested. So,
omit printing the bogus data altogether.

Signed-off-by: Ramkumar Ramachandra <artagnon@gmail.com>
Acked-by: Ingo Molnar <mingo@kernel.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1380616604-4077-1-git-send-email-artagnon@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
tools/perf/builtin-stat.c

index ce2266c..fb02b53 100644 (file)
@@ -930,11 +930,10 @@ static void abs_printout(int cpu, int nr, struct perf_evsel *evsel, double avg)
 
        if (perf_evsel__match(evsel, HARDWARE, HW_INSTRUCTIONS)) {
                total = avg_stats(&runtime_cycles_stats[cpu]);
-               if (total)
+               if (total) {
                        ratio = avg / total;
-
-               fprintf(output, " #   %5.2f  insns per cycle        ", ratio);
-
+                       fprintf(output, " #   %5.2f  insns per cycle        ", ratio);
+               }
                total = avg_stats(&runtime_stalled_cycles_front_stats[cpu]);
                total = max(total, avg_stats(&runtime_stalled_cycles_back_stats[cpu]));