Messages in this thread Patch in this message | | | Date | Wed, 8 Apr 2015 22:28:33 +0900 | From | Namhyung Kim <> | Subject | Re: [PATCH 2/6] perf stat: Fix metrics calculation with event qualifiers |
| |
Hi Jiri and Andi,
On Tue, Apr 07, 2015 at 11:25:15PM +0200, Jiri Olsa wrote: > From: Andi Kleen <ak@linux.intel.com> > > Currently in perf IPC and other metrics cannot be directly shown > separately for both user and kernel in a single run. The problem was > that the metrics matching code did not check event qualifiers. > > With this patch the following case works correctly. > > % perf stat -e cycles:k,cycles:u,instructions:k,instructions:u true > > Performance counter stats for 'true': > > 531,718 cycles:k > 203,895 cycles:u > 338,151 instructions:k # 0.64 insns per cycle > 105,961 instructions:u # 0.52 insns per cycle > > 0.002989739 seconds time elapsed > > Previously it would misreport the ratios because they were matching > the wrong value.
This patch reminds me of following change:
From a8ae843bf7cff7b68dc39b6f088c43669d140d00 Mon Sep 17 00:00:00 2001 From: Namhyung Kim <namhyung@kernel.org> Date: Wed, 8 Apr 2015 21:56:12 +0900 Subject: [RFC] perf stat: Add -U/--user option
The -U/--user option is to control event modifier for all events easily. It has same effect as if adding 'u' modifier to every events.
An example follows:
$ perf stat -U true
Performance counter stats for 'true':
0.722461 task-clock:uH (msec) # 0.520 CPUs utilized 0 context-switches:uH # 0.000 K/sec 0 cpu-migrations:uH # 0.000 K/sec 41 page-faults:uH # 0.057 M/sec 145,341 cycles:uH # 0.201 GHz 517,317 stalled-cycles-frontend:uH # 355.93% frontend cycles idle 497,570 stalled-cycles-backend:uH # 342.35% backend cycles idle 98,846 instructions:uH # 0.68 insns per cycle # 5.23 stalled cycles per insn 19,552 branches:uH # 27.063 M/sec <not counted> branch-misses:uH
0.001389613 seconds time elapsed
Cc: Andi Kleen <ak@linux.intel.com> Signed-off-by: Namhyung Kim <namhyung@kernel.org> --- tools/perf/Documentation/perf-stat.txt | 5 ++++- tools/perf/builtin-stat.c | 11 +++++++++++ 2 files changed, 15 insertions(+), 1 deletion(-)
diff --git a/tools/perf/Documentation/perf-stat.txt b/tools/perf/Documentation/perf-stat.txt index 04e150d83e7d..54b0734028a6 100644 --- a/tools/perf/Documentation/perf-stat.txt +++ b/tools/perf/Documentation/perf-stat.txt @@ -151,9 +151,12 @@ filter out the startup phase of the program, which is often very different. -T:: --transaction:: - Print statistics of transactional execution if supported. +-U:: +--user:: +Collect events only in user mode execution + EXAMPLES -------- diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c index fec089f1c364..d191af942cb4 100644 --- a/tools/perf/builtin-stat.c +++ b/tools/perf/builtin-stat.c @@ -141,6 +141,7 @@ static unsigned int interval = 0; static unsigned int initial_delay = 0; static unsigned int unit_width = 4; /* strlen("unit") */ static bool forever = false; +static bool mod_user = false; static struct timespec ref_time; static struct cpu_map *aggr_map; static int (*aggr_get_id)(struct cpu_map *m, int cpu); @@ -1809,6 +1810,7 @@ int cmd_stat(int argc, const char **argv, const char *prefix __maybe_unused) "aggregate counts per physical processor core", AGGR_CORE), OPT_UINTEGER('D', "delay", &initial_delay, "ms to wait before starting measurement after program start"), + OPT_BOOLEAN('U', "user", &mod_user, "Count event only in user mode"), OPT_END() }; const char * const stat_usage[] = { @@ -1913,6 +1915,15 @@ int cmd_stat(int argc, const char **argv, const char *prefix __maybe_unused) if (add_default_attributes()) goto out; + if (mod_user) { + struct perf_evsel *evsel; + + evlist__for_each(evsel_list, evsel) { + evsel->attr.exclude_kernel = 1; + evsel->attr.exclude_hv = 1; + } + } + target__validate(&target); if (perf_evlist__create_maps(evsel_list, &target) < 0) { -- 2.3.5
| |