[PATCH 07/11] perf script: Allow computing 'perf stat' style metrics

From: Arnaldo Carvalho de Melo
Date: Fri Nov 24 2017 - 10:04:10 EST

Next message: Arnaldo Carvalho de Melo: "[PATCH 05/11] perf record: Synthesize unit/scale/... in event update"
Previous message: Arnaldo Carvalho de Melo: "[PATCH 09/11] perf report: Fix -D output for user metadata events"
In reply to: Arnaldo Carvalho de Melo: "[PATCH 09/11] perf report: Fix -D output for user metadata events"
Next in thread: Arnaldo Carvalho de Melo: "[PATCH 05/11] perf record: Synthesize unit/scale/... in event update"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

From: Andi Kleen <ak@xxxxxxxxxxxxxxx>

Add support for computing 'perf stat' style metrics in 'perf script'.

When using leader sampling we can get metrics for each sampling period
by computing formulas over the values of the different group members.

This allows things like fine grained IPC tracking through sampling, much
more fine grained than with 'perf stat'.

The metric is still averaged over the sampling period, it is not just
for the sampling point.

This patch adds a new metric output field for 'perf script' that uses
the existing 'perf stat' metrics infrastructure to compute any metrics
supported by 'perf stat'.

For example to sample IPC:

$ perf record -e '{ref-cycles,cycles,instructions}:S' -a sleep 1
$ perf script -F metric,ip,sym,time,cpu,comm
...
alsa-sink-ALC32 [000] 42815.856074: 7fd65937d6cc [unknown]
alsa-sink-ALC32 [000] 42815.856074: 7fd65937d6cc [unknown]
alsa-sink-ALC32 [000] 42815.856074: 7fd65937d6cc [unknown]
alsa-sink-ALC32 [000] 42815.856074: metric: 0.13 insn per cycle
swapper [000] 42815.857961: ffffffff81655df0 __schedule
swapper [000] 42815.857961: ffffffff81655df0 __schedule
swapper [000] 42815.857961: ffffffff81655df0 __schedule
swapper [000] 42815.857961: metric: 0.23 insn per cycle
qemu-system-x86 [000] 42815.858130: ffffffff8165ad0e _raw_spin_unlock_irqrestore
qemu-system-x86 [000] 42815.858130: ffffffff8165ad0e _raw_spin_unlock_irqrestore
qemu-system-x86 [000] 42815.858130: ffffffff8165ad0e _raw_spin_unlock_irqrestore
qemu-system-x86 [000] 42815.858130: metric: 0.46 insn per cycle
:4972 [000] 42815.858312: ffffffffa080e5f2 vmx_vcpu_run
:4972 [000] 42815.858312: ffffffffa080e5f2 vmx_vcpu_run
:4972 [000] 42815.858312: ffffffffa080e5f2 vmx_vcpu_run
:4972 [000] 42815.858312: metric: 0.45 insn per cycle

TopDown:

This requires disabling SMT if you have it enabled, because SMT would
require sampling per core, which is not supported.

$ perf record -e '{ref-cycles,topdown-fetch-bubbles,\
topdown-recovery-bubbles,\
topdown-slots-retired,topdown-total-slots,\
topdown-slots-issued}:S' -a sleep 1
$ perf script --header -I -F cpu,ip,sym,event,metric,period
...
[000] 121108 ref-cycles: ffffffff8165222e copy_user_enhanced_fast_string
[000] 190350 topdown-fetch-bubbles: ffffffff8165222e copy_user_enhanced_fast_string
[000] 2055 topdown-recovery-bubbles: ffffffff8165222e copy_user_enhanced_fast_string
[000] 148729 topdown-slots-retired: ffffffff8165222e copy_user_enhanced_fast_string
[000] 144324 topdown-total-slots: ffffffff8165222e copy_user_enhanced_fast_string
[000] 160852 topdown-slots-issued: ffffffff8165222e copy_user_enhanced_fast_string
[000] metric: 33.0% frontend bound
[000] metric: 3.5% bad speculation
[000] metric: 25.8% retiring
[000] metric: 37.7% backend bound
[000] 112112 ref-cycles: ffffffff8165aec8 _raw_spin_lock_irqsave
[000] 357222 topdown-fetch-bubbles: ffffffff8165aec8 _raw_spin_lock_irqsave
[000] 3325 topdown-recovery-bubbles: ffffffff8165aec8 _raw_spin_lock_irqsave
[000] 323553 topdown-slots-retired: ffffffff8165aec8 _raw_spin_lock_irqsave
[000] 270507 topdown-total-slots: ffffffff8165aec8 _raw_spin_lock_irqsave
[000] 341226 topdown-slots-issued: ffffffff8165aec8 _raw_spin_lock_irqsave
[000] metric: 33.0% frontend bound
[000] metric: 2.9% bad speculation
[000] metric: 29.9% retiring
[000] metric: 34.2% backend bound
...

v2:
Use evsel->priv for new fields
Port to new base line, support fp output.
Handle stats in ->stats, not ->priv
Minor cleanups

Extra explanation about the use of the term 'averaging', from Andi in the
thread in the Link: tag below:

<quote Andi>
The current samples contains the sum of event counts for a sampling period.

EventA-1 EventA-2 EventA-3 EventA-4
EventB-1 EventB-2 EventC-3

gap with no events overflow
|-----------------------------------------------------------------|
period-start period-end
^ ^
| |
previous sample current sample

So EventA = 4 and EventB = 3 at the sample point

I generate a metric, let's say EventA / EventB. It applies to the whole period.

But the metric is over a longer time which does not have the same behavior. For
example the gap above doesn't have any events, while they are clustered at the
beginning and end of the sample period.

But we're summing everything together. The metric doesn't know that the gap is
different than the busy period.

That's what I'm trying to express with averaging.
</quote>

Signed-off-by: Andi Kleen <ak@xxxxxxxxxxxxxxx>
Acked-by: Jiri Olsa <jolsa@xxxxxxxxxx>
Link: http://lkml.kernel.org/r/20171117214300.32746-4-andi@xxxxxxxxxxxxxx
Signed-off-by: Arnaldo Carvalho de Melo <acme@xxxxxxxxxx>
---
tools/perf/Documentation/perf-script.txt | 10 +++-
tools/perf/builtin-script.c | 97 +++++++++++++++++++++++++++++++-
tools/perf/util/metricgroup.c | 4 ++
3 files changed, 108 insertions(+), 3 deletions(-)

diff --git a/tools/perf/Documentation/perf-script.txt b/tools/perf/Documentation/perf-script.txt
index 2811fcf684cb..974ceb12c7f3 100644
--- a/tools/perf/Documentation/perf-script.txt
+++ b/tools/perf/Documentation/perf-script.txt
@@ -117,7 +117,7 @@ OPTIONS
Comma separated list of fields to print. Options are:
comm, tid, pid, time, cpu, event, trace, ip, sym, dso, addr, symoff,
srcline, period, iregs, uregs, brstack, brstacksym, flags, bpf-output, brstackinsn,
- brstackoff, callindent, insn, insnlen, synth, phys_addr.
+ brstackoff, callindent, insn, insnlen, synth, phys_addr, metric.
Field list can be prepended with the type, trace, sw or hw,
to indicate to which event type the field list applies.
e.g., -F sw:comm,tid,time,ip,sym and -F trace:time,cpu,trace
@@ -217,6 +217,14 @@ OPTIONS

The brstackoff field will print an offset into a specific dso/binary.

+ With the metric option perf script can compute metrics for
+ sampling periods, similar to perf stat. This requires
+ specifying a group with multiple metrics with the :S option
+ for perf record. perf will sample on the first event, and
+ compute metrics for all the events in the group. Please note
+ that the metric computed is averaged over the whole sampling
+ period, not just for the sample point.
+
-k::
--vmlinux=<file>::
vmlinux pathname
diff --git a/tools/perf/builtin-script.c b/tools/perf/builtin-script.c
index ee7c7aaaae72..39d8b55f0db3 100644
--- a/tools/perf/builtin-script.c
+++ b/tools/perf/builtin-script.c
@@ -22,6 +22,7 @@
#include "util/cpumap.h"
#include "util/thread_map.h"
#include "util/stat.h"
+#include "util/color.h"
#include "util/string2.h"
#include "util/thread-stack.h"
#include "util/time-utils.h"
@@ -90,6 +91,7 @@ enum perf_output_field {
PERF_OUTPUT_SYNTH = 1U << 25,
PERF_OUTPUT_PHYS_ADDR = 1U << 26,
PERF_OUTPUT_UREGS = 1U << 27,
+ PERF_OUTPUT_METRIC = 1U << 28,
};

struct output_option {
@@ -124,6 +126,7 @@ struct output_option {
{.str = "brstackoff", .field = PERF_OUTPUT_BRSTACKOFF},
{.str = "synth", .field = PERF_OUTPUT_SYNTH},
{.str = "phys_addr", .field = PERF_OUTPUT_PHYS_ADDR},
+ {.str = "metric", .field = PERF_OUTPUT_METRIC},
};

enum {
@@ -215,12 +218,20 @@ struct perf_evsel_script {
char *filename;
FILE *fp;
u64 samples;
+ /* For metric output */
+ u64 val;
+ int gnum;
};

+static inline struct perf_evsel_script *evsel_script(struct perf_evsel *evsel)
+{
+ return (struct perf_evsel_script *)evsel->priv;
+}
+
static struct perf_evsel_script *perf_evsel_script__new(struct perf_evsel *evsel,
struct perf_data *data)
{
- struct perf_evsel_script *es = malloc(sizeof(*es));
+ struct perf_evsel_script *es = zalloc(sizeof(*es));

if (es != NULL) {
if (asprintf(&es->filename, "%s.%s.dump", data->file.path, perf_evsel__name(evsel)) < 0)
@@ -228,7 +239,6 @@ static struct perf_evsel_script *perf_evsel_script__new(struct perf_evsel *evsel
es->fp = fopen(es->filename, "w");
if (es->fp == NULL)
goto out_free_filename;
- es->samples = 0;
}

return es;
@@ -1472,6 +1482,86 @@ static int data_src__fprintf(u64 data_src, FILE *fp)
return fprintf(fp, "%-*s", maxlen, out);
}

+struct metric_ctx {
+ struct perf_sample *sample;
+ struct thread *thread;
+ struct perf_evsel *evsel;
+ FILE *fp;
+};
+
+static void script_print_metric(void *ctx, const char *color,
+ const char *fmt,
+ const char *unit, double val)
+{
+ struct metric_ctx *mctx = ctx;
+
+ if (!fmt)
+ return;
+ perf_sample__fprintf_start(mctx->sample, mctx->thread, mctx->evsel,
+ mctx->fp);
+ fputs("\tmetric: ", mctx->fp);
+ if (color)
+ color_fprintf(mctx->fp, color, fmt, val);
+ else
+ printf(fmt, val);
+ fprintf(mctx->fp, " %s\n", unit);
+}
+
+static void script_new_line(void *ctx)
+{
+ struct metric_ctx *mctx = ctx;
+
+ perf_sample__fprintf_start(mctx->sample, mctx->thread, mctx->evsel,
+ mctx->fp);
+ fputs("\tmetric: ", mctx->fp);
+}
+
+static void perf_sample__fprint_metric(struct perf_script *script,
+ struct thread *thread,
+ struct perf_evsel *evsel,
+ struct perf_sample *sample,
+ FILE *fp)
+{
+ struct perf_stat_output_ctx ctx = {
+ .print_metric = script_print_metric,
+ .new_line = script_new_line,
+ .ctx = &(struct metric_ctx) {
+ .sample = sample,
+ .thread = thread,
+ .evsel = evsel,
+ .fp = fp,
+ },
+ .force_header = false,
+ };
+ struct perf_evsel *ev2;
+ static bool init;
+ u64 val;
+
+ if (!init) {
+ perf_stat__init_shadow_stats();
+ init = true;
+ }
+ if (!evsel->stats)
+ perf_evlist__alloc_stats(script->session->evlist, false);
+ if (evsel_script(evsel->leader)->gnum++ == 0)
+ perf_stat__reset_shadow_stats();
+ val = sample->period * evsel->scale;
+ perf_stat__update_shadow_stats(evsel,
+ val,
+ sample->cpu);
+ evsel_script(evsel)->val = val;
+ if (evsel_script(evsel->leader)->gnum == evsel->leader->nr_members) {
+ for_each_group_member (ev2, evsel->leader) {
+ perf_stat__print_shadow_stats(ev2,
+ evsel_script(ev2)->val,
+ sample->cpu,
+ &ctx,
+ NULL);
+ }
+ evsel_script(evsel->leader)->gnum = 0;
+ }
+}
+
static void process_event(struct perf_script *script,
struct perf_sample *sample, struct perf_evsel *evsel,
struct addr_location *al,
@@ -1559,6 +1649,9 @@ static void process_event(struct perf_script *script,
if (PRINT_FIELD(PHYS_ADDR))
fprintf(fp, "%16" PRIx64, sample->phys_addr);
fprintf(fp, "\n");
+
+ if (PRINT_FIELD(METRIC))
+ perf_sample__fprint_metric(script, thread, evsel, sample, fp);
}

static struct scripting_ops *scripting_ops;
diff --git a/tools/perf/util/metricgroup.c b/tools/perf/util/metricgroup.c
index 0ddd9c199227..6fd709017bbc 100644
--- a/tools/perf/util/metricgroup.c
+++ b/tools/perf/util/metricgroup.c
@@ -38,6 +38,10 @@ struct metric_event *metricgroup__lookup(struct rblist *metric_events,
struct metric_event me = {
.evsel = evsel
};
+
+ if (!metric_events)
+ return NULL;
+
nd = rblist__find(metric_events, &me);
if (nd)
return container_of(nd, struct metric_event, nd);
--
2.13.6

Next message: Arnaldo Carvalho de Melo: "[PATCH 05/11] perf record: Synthesize unit/scale/... in event update"
Previous message: Arnaldo Carvalho de Melo: "[PATCH 09/11] perf report: Fix -D output for user metadata events"
In reply to: Arnaldo Carvalho de Melo: "[PATCH 09/11] perf report: Fix -D output for user metadata events"
Next in thread: Arnaldo Carvalho de Melo: "[PATCH 05/11] perf record: Synthesize unit/scale/... in event update"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]