Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]
Groups > linux.kernel > #1503425 > unrolled thread
| Started by | Andi Kleen <andi@firstfloor.org> |
|---|---|
| First post | 2016-10-19 01:20 +0200 |
| Last post | 2016-10-19 01:20 +0200 |
| Articles | 3 — 1 participant |
Back to article view | Back to linux.kernel
Support Intel uncore event lists v2 Andi Kleen <andi@firstfloor.org> - 2016-10-19 01:20 +0200
[PATCH 08/11] perf, tools: Collapse identically named events in perf stat Andi Kleen <andi@firstfloor.org> - 2016-10-19 01:20 +0200
[PATCH 04/11] perf, tools: Add support for parsing uncore json files Andi Kleen <andi@firstfloor.org> - 2016-10-19 01:20 +0200
| From | Andi Kleen <andi@firstfloor.org> |
|---|---|
| Date | 2016-10-19 01:20 +0200 |
| Subject | Support Intel uncore event lists v2 |
| Message-ID | <stLG9-vI-3@gated-at.bofh.it> |
This adds uncore support on top of the recently merged JSON event list infrastructure for core events. Uncore is everything outside the core, including memory controllers, PCI, interconnect etc. Uncore is more complicated to handle than core events because it uses many duplicated PMUs, which leads to long event lists and verbose duplicated outputs. In fact previously it was nearly unusable for many cases without special tools to generate event list and aggregate data (such as https://github.com/andikleen/pmu-tools/tree/master/ucevent) With this patchkit we add: - Basic support for uncore events in JSON events - Support aliases that get duplicated over many PMUs transparently - Support summing up duplicated PMUs per socket - Support extending the perf stat builtin metrics with simple ratios specified in the event list. This covers the vast majority of useful metrics. So far mainly servers are supported. Also this is not using full event lists (which are full of very obscure events) but only for a smaller subset of curated useful and understandable metrics. The actual event lists are not posted, but available at git://git.kernel.org/pub/scm/linux/kernel/git/ak/linux-misc perf/intel-uncore-json-files-1 The code is available here git://git.kernel.org/pub/scm/linux/kernel/git/ak/linux-misc perf/builtin-json-17 v1: Initial post v2: Address review feedback. See changelog in commits. -Andi
[toc] | [next] | [standalone]
| From | Andi Kleen <andi@firstfloor.org> |
|---|---|
| Date | 2016-10-19 01:20 +0200 |
| Subject | [PATCH 08/11] perf, tools: Collapse identically named events in perf stat |
| Message-ID | <stLPQ-zw-11@gated-at.bofh.it> |
| In reply to | #1503425 |
From: Andi Kleen <ak@linux.intel.com>
The uncore PMU has a lot of duplicated PMUs for different subsystems.
When expanding an uncore alias we usually end up with a large
number of identically named aliases, which makes perf stat
output difficult to read.
Automatically sum them up in perf stat, unless --no-merge is specified.
This can be default because only the uncores generally have duplicated
aliases. Other PMUs have unique names.
Before:
% perf stat --no-merge -a -e unc_c_llc_lookup.any sleep 1
Performance counter stats for 'system wide':
694,976 Bytes unc_c_llc_lookup.any
706,304 Bytes unc_c_llc_lookup.any
956,608 Bytes unc_c_llc_lookup.any
782,720 Bytes unc_c_llc_lookup.any
605,696 Bytes unc_c_llc_lookup.any
442,816 Bytes unc_c_llc_lookup.any
659,328 Bytes unc_c_llc_lookup.any
509,312 Bytes unc_c_llc_lookup.any
263,936 Bytes unc_c_llc_lookup.any
592,448 Bytes unc_c_llc_lookup.any
672,448 Bytes unc_c_llc_lookup.any
608,640 Bytes unc_c_llc_lookup.any
641,024 Bytes unc_c_llc_lookup.any
856,896 Bytes unc_c_llc_lookup.any
808,832 Bytes unc_c_llc_lookup.any
684,864 Bytes unc_c_llc_lookup.any
710,464 Bytes unc_c_llc_lookup.any
538,304 Bytes unc_c_llc_lookup.any
1.002577660 seconds time elapsed
After:
% perf stat -a -e unc_c_llc_lookup.any sleep 1
Performance counter stats for 'system wide':
2,685,120 Bytes unc_c_llc_lookup.any
1.002648032 seconds time elapsed
v2: Split collect_aliases. Rename alias flag.
Signed-off-by: Andi Kleen <ak@linux.intel.com>
---
tools/perf/Documentation/perf-stat.txt | 3 +
tools/perf/builtin-stat.c | 132 +++++++++++++++++++++++++++------
tools/perf/util/evsel.h | 1 +
3 files changed, 113 insertions(+), 23 deletions(-)
diff --git a/tools/perf/Documentation/perf-stat.txt b/tools/perf/Documentation/perf-stat.txt
index d96ccd4844df..320d8020bc5b 100644
--- a/tools/perf/Documentation/perf-stat.txt
+++ b/tools/perf/Documentation/perf-stat.txt
@@ -237,6 +237,9 @@ To interpret the results it is usually needed to know on which
CPUs the workload runs on. If needed the CPUs can be forced using
taskset.
+--no-merge::
+Do not merge results from same PMUs.
+
EXAMPLES
--------
diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index 688dea7cb08f..501d58a4925a 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -140,6 +140,7 @@ static unsigned int unit_width = 4; /* strlen("unit") */
static bool forever = false;
static bool metric_only = false;
static bool force_metric_only = false;
+static bool no_merge = false;
static struct timespec ref_time;
static struct cpu_map *aggr_map;
static aggr_get_id_t aggr_get_id;
@@ -1178,11 +1179,66 @@ static void aggr_update_shadow(void)
}
}
+static void collect_all_aliases(struct perf_evsel *counter,
+ void (*cb)(struct perf_evsel *counter, void *data,
+ bool first),
+ void *data)
+{
+ struct perf_evsel *alias;
+
+ alias = list_prepare_entry(counter, &(evsel_list->entries), node);
+ list_for_each_entry_continue (alias, &evsel_list->entries, node) {
+ if (strcmp(perf_evsel__name(alias), perf_evsel__name(counter)) ||
+ alias->scale != counter->scale ||
+ alias->cgrp != counter->cgrp ||
+ strcmp(alias->unit, counter->unit) ||
+ nsec_counter(alias) != nsec_counter(counter))
+ break;
+ alias->merged_stat = true;
+ cb(alias, data, false);
+ }
+}
+
+static void collect_aliases(struct perf_evsel *counter,
+ void (*cb)(struct perf_evsel *counter, void *data,
+ bool first),
+ void *data)
+{
+ cb(counter, data, true);
+ if (!no_merge)
+ collect_all_aliases(counter, cb, data);
+}
+
+struct aggr_data {
+ u64 ena, run, val;
+ int id;
+ int nr;
+ int cpu;
+};
+
+static void aggr_cb(struct perf_evsel *counter, void *data, bool first)
+{
+ struct aggr_data *ad = data;
+ int cpu, cpu2, s2;
+
+ for (cpu = 0; cpu < perf_evsel__nr_cpus(counter); cpu++) {
+ cpu2 = perf_evsel__cpus(counter)->map[cpu];
+ s2 = aggr_get_id(evsel_list->cpus, cpu2);
+ if (s2 != ad->id)
+ continue;
+ ad->val += perf_counts(counter->counts, cpu, 0)->val;
+ ad->ena += perf_counts(counter->counts, cpu, 0)->ena;
+ ad->run += perf_counts(counter->counts, cpu, 0)->run;
+ if (first)
+ ad->nr++;
+ }
+}
+
static void print_aggr(char *prefix)
{
FILE *output = stat_config.output;
struct perf_evsel *counter;
- int cpu, s, s2, id, nr;
+ int s, id, nr;
double uval;
u64 ena, run, val;
bool first;
@@ -1197,23 +1253,22 @@ static void print_aggr(char *prefix)
* Without each counter has its own line.
*/
for (s = 0; s < aggr_map->nr; s++) {
+ struct aggr_data ad;
if (prefix && metric_only)
fprintf(output, "%s", prefix);
- id = aggr_map->map[s];
+ ad.id = id = aggr_map->map[s];
first = true;
evlist__for_each_entry(evsel_list, counter) {
- val = ena = run = 0;
- nr = 0;
- for (cpu = 0; cpu < perf_evsel__nr_cpus(counter); cpu++) {
- s2 = aggr_get_id(perf_evsel__cpus(counter), cpu);
- if (s2 != id)
- continue;
- val += perf_counts(counter->counts, cpu, 0)->val;
- ena += perf_counts(counter->counts, cpu, 0)->ena;
- run += perf_counts(counter->counts, cpu, 0)->run;
- nr++;
- }
+ if (counter->merged_stat)
+ continue;
+ ad.val = ad.ena = ad.run = 0;
+ ad.nr = 0;
+ collect_aliases(counter, aggr_cb, &ad);
+ nr = ad.nr;
+ ena = ad.ena;
+ run = ad.run;
+ val = ad.val;
if (first && metric_only) {
first = false;
aggr_printout(counter, id, nr);
@@ -1257,6 +1312,21 @@ static void print_aggr_thread(struct perf_evsel *counter, char *prefix)
}
}
+struct caggr_data {
+ double avg, avg_enabled, avg_running;
+};
+
+static void counter_aggr_cb(struct perf_evsel *counter, void *data,
+ bool first __maybe_unused)
+{
+ struct caggr_data *cd = data;
+ struct perf_stat_evsel *ps = counter->priv;
+
+ cd->avg += avg_stats(&ps->res_stats[0]);
+ cd->avg_enabled += avg_stats(&ps->res_stats[1]);
+ cd->avg_running += avg_stats(&ps->res_stats[2]);
+}
+
/*
* Print out the results of a single counter:
* aggregated counts in system-wide mode
@@ -1264,23 +1334,32 @@ static void print_aggr_thread(struct perf_evsel *counter, char *prefix)
static void print_counter_aggr(struct perf_evsel *counter, char *prefix)
{
FILE *output = stat_config.output;
- struct perf_stat_evsel *ps = counter->priv;
- double avg = avg_stats(&ps->res_stats[0]);
double uval;
- double avg_enabled, avg_running;
+ struct caggr_data cd = { .avg = 0.0 };
- avg_enabled = avg_stats(&ps->res_stats[1]);
- avg_running = avg_stats(&ps->res_stats[2]);
+ if (counter->merged_stat)
+ return;
+ collect_aliases(counter, counter_aggr_cb, &cd);
if (prefix && !metric_only)
fprintf(output, "%s", prefix);
- uval = avg * counter->scale;
- printout(-1, 0, counter, uval, prefix, avg_running, avg_enabled, avg);
+ uval = cd.avg * counter->scale;
+ printout(-1, 0, counter, uval, prefix, cd.avg_running, cd.avg_enabled, cd.avg);
if (!metric_only)
fprintf(output, "\n");
}
+static void counter_cb(struct perf_evsel *counter, void *data,
+ bool first __maybe_unused)
+{
+ struct aggr_data *ad = data;
+
+ ad->val += perf_counts(counter->counts, ad->cpu, 0)->val;
+ ad->ena += perf_counts(counter->counts, ad->cpu, 0)->ena;
+ ad->run += perf_counts(counter->counts, ad->cpu, 0)->run;
+}
+
/*
* Print out the results of a single counter:
* does not use aggregated count in system-wide
@@ -1292,10 +1371,16 @@ static void print_counter(struct perf_evsel *counter, char *prefix)
double uval;
int cpu;
+ if (counter->merged_stat)
+ return;
+
for (cpu = 0; cpu < perf_evsel__nr_cpus(counter); cpu++) {
- val = perf_counts(counter->counts, cpu, 0)->val;
- ena = perf_counts(counter->counts, cpu, 0)->ena;
- run = perf_counts(counter->counts, cpu, 0)->run;
+ struct aggr_data ad = { .cpu = cpu };
+
+ collect_aliases(counter, counter_cb, &ad);
+ val = ad.val;
+ ena = ad.ena;
+ run = ad.run;
if (prefix)
fprintf(output, "%s", prefix);
@@ -1633,6 +1718,7 @@ static const struct option stat_options[] = {
"list of cpus to monitor in system-wide"),
OPT_SET_UINT('A', "no-aggr", &stat_config.aggr_mode,
"disable CPU count aggregation", AGGR_NONE),
+ OPT_BOOLEAN(0, "no-merge", &no_merge, "Do not merge identical named events"),
OPT_STRING('x', "field-separator", &csv_sep, "separator",
"print counts with custom separator"),
OPT_CALLBACK('G', "cgroup", &evsel_list, "name",
diff --git a/tools/perf/util/evsel.h b/tools/perf/util/evsel.h
index b1503b0ecdff..aae496a9bec1 100644
--- a/tools/perf/util/evsel.h
+++ b/tools/perf/util/evsel.h
@@ -128,6 +128,7 @@ struct perf_evsel {
bool cmdline_group_boundary;
struct list_head config_terms;
int bpf_fd;
+ bool merged_stat;
};
union u64_swap {
--
2.5.5
[toc] | [prev] | [next] | [standalone]
| From | Andi Kleen <andi@firstfloor.org> |
|---|---|
| Date | 2016-10-19 01:20 +0200 |
| Subject | [PATCH 04/11] perf, tools: Add support for parsing uncore json files |
| Message-ID | <stLPQ-zw-13@gated-at.bofh.it> |
| In reply to | #1503425 |
From: Andi Kleen <ak@linux.intel.com>
Handle the Unit field, which is needed to find the right PMU for
an event. We call it "pmu" and convert it to the perf pmu name
with an uncore prefix.
Handle the ExtSel field, which just extends the event mask with
an additional bit.
Handle the Filter field which adds parameters to the main event
to configure filtering.
Handle the Unit field which declares the unit the values
should be scaled too (similar to what the kernel exports)
Set up the perpkg field for uncore events so that perf
knows they are per package (similar to what the kernel exports)
Then output the fields into the pmu-events data structures which
are compiled into perf.
Filter out zero fields, except for the event itself.
v2: Fix compilation. Add uncore_ prefix at pre-processing time.
Move eventcode change to separate patch.
Signed-off-by: Andi Kleen <ak@linux.intel.com>
---
tools/perf/pmu-events/jevents.c | 74 +++++++++++++++++++++++++++++++++++---
tools/perf/pmu-events/jevents.h | 4 ++-
tools/perf/pmu-events/pmu-events.h | 3 ++
tools/perf/util/pmu.c | 27 +++++++++-----
4 files changed, 95 insertions(+), 13 deletions(-)
diff --git a/tools/perf/pmu-events/jevents.c b/tools/perf/pmu-events/jevents.c
index d343156b1dea..3a3ab5b17fc5 100644
--- a/tools/perf/pmu-events/jevents.c
+++ b/tools/perf/pmu-events/jevents.c
@@ -191,6 +191,27 @@ static struct msrmap *lookup_msr(char *map, jsmntok_t *val)
return NULL;
}
+static struct map {
+ const char *json;
+ const char *perf;
+} unit_to_pmu[] = {
+ { "CBO", "uncore_cbox" },
+ { "QPI LL", "uncore_qpi" },
+ { "SBO", "uncore_sbox" },
+ {}
+};
+
+static const char *field_to_perf(struct map *table, char *map, jsmntok_t *val)
+{
+ int i;
+
+ for (i = 0; table[i].json; i++) {
+ if (json_streq(map, val, table[i].json))
+ return table[i].perf;
+ }
+ return NULL;
+}
+
#define EXPECT(e, t, m) do { if (!(e)) { \
jsmntok_t *loc = (t); \
if (!(t)->start && (t) > tokens) \
@@ -272,7 +293,8 @@ static void print_events_table_prefix(FILE *fp, const char *tblname)
}
static int print_events_table_entry(void *data, char *name, char *event,
- char *desc, char *long_desc)
+ char *desc, char *long_desc,
+ char *pmu, char *unit, char *perpkg)
{
struct perf_entry_data *pd = data;
FILE *outfp = pd->outfp;
@@ -290,7 +312,12 @@ static int print_events_table_entry(void *data, char *name, char *event,
fprintf(outfp, "\t.topic = \"%s\",\n", topic);
if (long_desc && long_desc[0])
fprintf(outfp, "\t.long_desc = \"%s\",\n", long_desc);
-
+ if (pmu)
+ fprintf(outfp, "\t.pmu = \"%s\",\n", pmu);
+ if (unit)
+ fprintf(outfp, "\t.unit = \"%s\",\n", unit);
+ if (perpkg)
+ fprintf(outfp, "\t.perpkg = \"%s\",\n", perpkg);
fprintf(outfp, "},\n");
return 0;
@@ -337,7 +364,8 @@ static char *real_event(const char *name, char *event)
/* Call func with each event in the json file */
int json_events(const char *fn,
int (*func)(void *data, char *name, char *event, char *desc,
- char *long_desc),
+ char *long_desc,
+ char *pmu, char *unit, char *perpkg),
void *data)
{
int err = -EIO;
@@ -359,6 +387,10 @@ int json_events(const char *fn,
char *event = NULL, *desc = NULL, *name = NULL;
char *long_desc = NULL;
char *extra_desc = NULL;
+ char *pmu = NULL;
+ char *filter = NULL;
+ char *perpkg = NULL;
+ char *unit = NULL;
unsigned long long eventcode = 0;
struct msrmap *msr = NULL;
jsmntok_t *msrval = NULL;
@@ -385,6 +417,11 @@ int json_events(const char *fn,
addfield(map, &code, "", "", val);
eventcode |= strtoul(code, NULL, 0);
free(code);
+ } else if (json_streq(map, field, "ExtSel")) {
+ char *code = NULL;
+ addfield(map, &code, "", "", val);
+ eventcode |= strtoul(code, NULL, 0) << 21;
+ free(code);
} else if (json_streq(map, field, "EventName")) {
addfield(map, &name, "", "", val);
} else if (json_streq(map, field, "BriefDescription")) {
@@ -408,6 +445,28 @@ int json_events(const char *fn,
addfield(map, &extra_desc, ". ",
" Supports address when precise",
NULL);
+ } else if (json_streq(map, field, "Unit")) {
+ const char *ppmu;
+ char *s;
+
+ ppmu = field_to_perf(unit_to_pmu, map, val);
+ if (ppmu) {
+ pmu = strdup(ppmu);
+ } else {
+ if (!pmu)
+ pmu = strdup("uncore_");
+ addfield(map, &pmu, "", "", val);
+ for (s = pmu; *s; s++)
+ *s = tolower(*s);
+ }
+ addfield(map, &desc, ". ", "Unit: ", NULL);
+ addfield(map, &desc, "", pmu, NULL);
+ } else if (json_streq(map, field, "Filter")) {
+ addfield(map, &filter, "", "", val);
+ } else if (json_streq(map, field, "ScaleUnit")) {
+ addfield(map, &unit, "", "", val);
+ } else if (json_streq(map, field, "PerPkg")) {
+ addfield(map, &perpkg, "", "", val);
}
/* ignore unknown fields */
}
@@ -425,16 +484,23 @@ int json_events(const char *fn,
addfield(map, &desc, " ", extra_desc, NULL);
if (long_desc && extra_desc)
addfield(map, &long_desc, " ", extra_desc, NULL);
+ if (filter)
+ addfield(map, &event, ",", filter, NULL);
if (msr != NULL)
addfield(map, &event, ",", msr->pname, msrval);
fixname(name);
- err = func(data, name, real_event(name, event), desc, long_desc);
+ err = func(data, name, real_event(name, event), desc, long_desc,
+ pmu, unit, perpkg);
free(event);
free(desc);
free(name);
free(long_desc);
free(extra_desc);
+ free(pmu);
+ free(filter);
+ free(perpkg);
+ free(unit);
if (err)
break;
tok += j;
diff --git a/tools/perf/pmu-events/jevents.h b/tools/perf/pmu-events/jevents.h
index b0eb2744b498..71e13de31092 100644
--- a/tools/perf/pmu-events/jevents.h
+++ b/tools/perf/pmu-events/jevents.h
@@ -3,7 +3,9 @@
int json_events(const char *fn,
int (*func)(void *data, char *name, char *event, char *desc,
- char *long_desc),
+ char *long_desc,
+ char *pmu,
+ char *unit, char *perpkg),
void *data);
char *get_cpu_str(void);
diff --git a/tools/perf/pmu-events/pmu-events.h b/tools/perf/pmu-events/pmu-events.h
index 2eaef595d8a0..c669a3cdb9f0 100644
--- a/tools/perf/pmu-events/pmu-events.h
+++ b/tools/perf/pmu-events/pmu-events.h
@@ -10,6 +10,9 @@ struct pmu_event {
const char *desc;
const char *topic;
const char *long_desc;
+ const char *pmu;
+ const char *unit;
+ const char *perpkg;
};
/*
diff --git a/tools/perf/util/pmu.c b/tools/perf/util/pmu.c
index 739e8aa63b78..6edcba323543 100644
--- a/tools/perf/util/pmu.c
+++ b/tools/perf/util/pmu.c
@@ -229,11 +229,13 @@ static int perf_pmu__parse_snapshot(struct perf_pmu_alias *alias,
}
static int __perf_pmu__new_alias(struct list_head *list, char *dir, char *name,
- char *desc, char *val, char *long_desc,
- char *topic)
+ char *desc __maybe_unused, char *val,
+ char *long_desc, char *topic,
+ char *unit, char *perpkg)
{
struct perf_pmu_alias *alias;
int ret;
+ int num;
alias = malloc(sizeof(*alias));
if (!alias)
@@ -267,7 +269,12 @@ static int __perf_pmu__new_alias(struct list_head *list, char *dir, char *name,
alias->long_desc = long_desc ? strdup(long_desc) :
desc ? strdup(desc) : NULL;
alias->topic = topic ? strdup(topic) : NULL;
-
+ if (unit) {
+ if (convert_scale(unit, &unit, &alias->scale) < 0)
+ return -1;
+ snprintf(alias->unit, sizeof(alias->unit), "%s", unit);
+ }
+ alias->per_pkg = perpkg && sscanf(perpkg, "%d", &num) == 1 && num == 1;
list_add_tail(&alias->list, list);
return 0;
@@ -284,7 +291,8 @@ static int perf_pmu__new_alias(struct list_head *list, char *dir, char *name, FI
buf[ret] = 0;
- return __perf_pmu__new_alias(list, dir, name, NULL, buf, NULL, NULL);
+ return __perf_pmu__new_alias(list, dir, name, NULL, buf, NULL, NULL, NULL,
+ NULL);
}
static inline bool pmu_alias_info_file(char *name)
@@ -504,7 +512,7 @@ char * __weak get_cpuid_str(void)
* to the current running CPU. Then, add all PMU events from that table
* as aliases.
*/
-static void pmu_add_cpu_aliases(struct list_head *head)
+static void pmu_add_cpu_aliases(struct list_head *head, const char *name)
{
int i;
struct pmu_events_map *map;
@@ -544,10 +552,14 @@ static void pmu_add_cpu_aliases(struct list_head *head)
if (!pe->name)
break;
+ if (pe->pmu && strncmp(pe->pmu, name, strlen(pe->pmu)))
+ continue;
+
/* need type casts to override 'const' */
__perf_pmu__new_alias(head, NULL, (char *)pe->name,
(char *)pe->desc, (char *)pe->event,
- (char *)pe->long_desc, (char *)pe->topic);
+ (char *)pe->long_desc, (char *)pe->topic,
+ (char *)pe->unit, (char *)pe->perpkg);
}
out:
@@ -578,8 +590,7 @@ static struct perf_pmu *pmu_lookup(const char *name)
if (pmu_aliases(name, &aliases))
return NULL;
- if (!strcmp(name, "cpu"))
- pmu_add_cpu_aliases(&aliases);
+ pmu_add_cpu_aliases(&aliases, name);
if (pmu_type(name, &type))
return NULL;
--
2.5.5
[toc] | [prev] | [standalone]
Back to top | Article view | linux.kernel
csiph-web