[tip:perf/stat] perf stat: Print cache misses as percentage

[tip:perf/stat] perf stat: Print cache misses as percentage

[tip-bot for Ingo Molnar]

Commit-ID:  d58f4c82fed69fdd4a79fa54fe17fd14d98c27aa
Gitweb:     http://git.kernel.org/tip/d58f4c82fed69fdd4a79fa54fe17fd14d98c27aa
Author:     Ingo Molnar <mingo@elte.hu>
AuthorDate: Wed, 27 Apr 2011 03:42:18 +0200
Committer:  Ingo Molnar <mingo@elte.hu>
CommitDate: Tue, 26 Apr 2011 20:04:54 +0200

perf stat: Print cache misses as percentage

Before:

       113,393,041 cache-references         #     83.636 M/sec
         7,052,454 cache-misses             #      5.202 M/sec

After:

       112,589,441 cache-references         #     87.925 M/sec
         6,556,354 cache-misses             #      5.823 %

misses/hits percentages are more expressive than absolute numbers
or rates.

(Also prettify the CPUs printout line to not have a trailing whitespace.)

Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/n/tip-axm28f43x439bl41zkvfzd63@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@elte.hu>
---
 tools/perf/builtin-stat.c |   18 +++++++++++++++---
 1 files changed, 15 insertions(+), 3 deletions(-)

diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index 3a29041..0de3a20 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -157,6 +157,7 @@ static double stddev_stats(struct stats *stats)
 struct stats			runtime_nsecs_stats[MAX_NR_CPUS];
 struct stats			runtime_cycles_stats[MAX_NR_CPUS];
 struct stats			runtime_branches_stats[MAX_NR_CPUS];
+struct stats			runtime_cacherefs_stats[MAX_NR_CPUS];
 struct stats			walltime_nsecs_stats;
 
 static int create_perf_stat_counter(struct perf_evsel *evsel)
@@ -219,10 +220,12 @@ static int read_counter_aggr(struct perf_evsel *counter)
 	 */
 	if (perf_evsel__match(counter, SOFTWARE, SW_TASK_CLOCK))
 		update_stats(&runtime_nsecs_stats[0], count[0]);
-	if (perf_evsel__match(counter, HARDWARE, HW_CPU_CYCLES))
+	else if (perf_evsel__match(counter, HARDWARE, HW_CPU_CYCLES))
 		update_stats(&runtime_cycles_stats[0], count[0]);
-	if (perf_evsel__match(counter, HARDWARE, HW_BRANCH_INSTRUCTIONS))
+	else if (perf_evsel__match(counter, HARDWARE, HW_BRANCH_INSTRUCTIONS))
 		update_stats(&runtime_branches_stats[0], count[0]);
+	else if (perf_evsel__match(counter, HARDWARE, HW_CACHE_REFERENCES))
+		update_stats(&runtime_cacherefs_stats[0], count[0]);
 
 	return 0;
 }
@@ -404,7 +407,7 @@ static void nsec_printout(int cpu, struct perf_evsel *evsel, double avg)
 		return;
 
 	if (perf_evsel__match(evsel, SOFTWARE, SW_TASK_CLOCK))
-		fprintf(stderr, " # %10.3f CPUs ",
+		fprintf(stderr, " # %10.3f CPUs",
 				avg / avg_stats(&walltime_nsecs_stats));
 }
 
@@ -452,6 +455,15 @@ static void abs_printout(int cpu, struct perf_evsel *evsel, double avg)
 
 		fprintf(stderr, " # %10.3f %%", ratio);
 
+	} else if (perf_evsel__match(evsel, HARDWARE, HW_CACHE_MISSES) &&
+			runtime_cacherefs_stats[cpu].n != 0) {
+		total = avg_stats(&runtime_cacherefs_stats[cpu]);
+
+		if (total)
+			ratio = avg * 100 / total;
+
+		fprintf(stderr, " # %10.3f %%", ratio);
+
 	} else if (runtime_nsecs_stats[cpu].n != 0) {
 		total = avg_stats(&runtime_nsecs_stats[cpu]);