Skip to contentSkip to conversation

Freq/CPU%/CORE_BUSY% support

PATCH Superseded [email protected] 3 participants 13 messages Sep 3, 2015 – Sep 8, 2015 perf events

v8 of 10 Superseded Latest is v10 →

lore
Share & tools
lore
#0 lore raw

v8 cover Freq/CPU%/CORE_BUSY% support

From: Kan Liang <[email protected]>

This patch set supports per-sample freq/CPU%/CORE_BUSY% print in perf report -D and --stdio.
For printing these information, the perf.data file must have been obtained by group read and using special events cycles, ref-cycles, msr/tsc/, msr/aperf/ or msr/mperf/.

- Freq (MHz): The frequency during the sample interval. Needs cycles ref-cycles event.
- CPU%: CPU utilization during the sample interval. Needs ref-cycles and msr/tsc/ events.
- CORE_BUSY%: actual percent performance (APERF/MPERF%) during the sample interval. Needs msr/aperf/ and msr/mperf/ events.

Here is an example:

$ perf record --freq-perf ~/tchain_edit
$ perf report --stdio --group --show-freq-perf

Overhead FREQ MHz CPU% CORE_BUSY%
Command Shared Object Symbol
........................................ ......... ..... ..........
........... ................ ......................

    99.54%  99.54%  99.53%  99.53%  99.53%       2301     96         99
tchain_edit  tchain_edit       [.] f3
     0.20%   0.20%   0.20%   0.20%   0.20%       2301     98         99
tchain_edit  tchain_edit       [.] f2
     0.05%   0.05%   0.05%   0.05%   0.05%       2300     98         99
tchain_edit  [kernel.vmlinux]  [k] read_tsc

Changes since V1:
- Save cpu max freq to header when recording
- Read cpu max freq and msr type from header when reporting

Changes since V2:
- Introduce generic FEAT for CPU related data stored
- Make cpu max freq and msr type part of perf_session_env
- rename cpu_u to cpu_util
- Don't save sample value in perf_sample and discards new iterator. Calculating the freq_perf_info in add_entry_cb callback
- Introduce symbol_conf.freq_perf_type for related hpp column visibility

Changes since V3:
- add a identifier 'tag' for CPU attributes, max frequency.
- add backpointers to evlist for env, and evsel for evlist.
- Use bitmask for freq_perf_type
- Replace macros by functions to caculate freq, cpu_util and core_busy
- Move all caculation codes under symbol_conf.show_freq_perf condition.

Changes since V4:
- Store cpu attributes id as tag and more readable cpu_attr

Changes since V5:
- Rename freq to max_freq and use it
- Add a loop in process_cpu_attributes to facility future extension

Changes since V6:
- Split rename perf_session_env and add backpointer to evlist patches

Changes since V7:
- Introduce --freq-perf option for perf record
- Factor out fix for dump_sample

Kan Liang (7):
perf,tools: introduce generic FEAT for CPU attributes perf,tools: read msr pmu type from header. perf,record: introduce --freq-perf option perf,tools: fix dump_sample crash when evsel is NULL. perf,tools: Dump per-sample freq/CPU%/CORE_BUSY% in report -D perf,tools: caculate and save freq/CPU%/CORE_BUSY% in he_stat perf,tools: Show freq/CPU%/CORE_BUSY% in perf report --stdio

 tools/perf/Documentation/perf-record.txt |  4 ++
 tools/perf/Documentation/perf-report.txt | 12 ++++++
 tools/perf/builtin-record.c              | 39 +++++++++++++++++-
 tools/perf/builtin-report.c              | 56 +++++++++++++++++++++++++
 tools/perf/ui/hist.c                     | 71 +++++++++++++++++++++++++++++---
 tools/perf/util/cpumap.c                 | 32 ++++++++++++++
 tools/perf/util/cpumap.h                 |  1 +
 tools/perf/util/header.c                 | 62 ++++++++++++++++++++++++++++
 tools/perf/util/header.h                 | 13 ++++++
 tools/perf/util/hist.h                   |  3 ++
 tools/perf/util/session.c                | 33 +++++++++++++--
 tools/perf/util/session.h                | 64 ++++++++++++++++++++++++++++
 tools/perf/util/sort.c                   |  3 ++
 tools/perf/util/sort.h                   |  3 ++
 tools/perf/util/symbol.h                 | 12 +++++-
 15 files changed, 397 insertions(+), 11 deletions(-)
Signature · 1 line
1.8.3.1
#1 lore raw

PATCH v8 1/7 perf,tools: introduce generic FEAT for CPU attributes

From: Kan Liang <[email protected]>

This patch introduces generic FEAT for CPU attributes. For the patch set, we only need cpu max frequency. But it can be easily extented to support more other CPU attributes.
The cpu max frequency is from the first online cpu.

---
 tools/perf/util/cpumap.c | 32 ++++++++++++++++++++++++++
 tools/perf/util/cpumap.h |  1 +
 tools/perf/util/header.c | 59 ++++++++++++++++++++++++++++++++++++++++++++++++
 tools/perf/util/header.h | 12 ++++++++++
 4 files changed, 104 insertions(+)
tools/perf/util/cpumap.c +32 -0
index a05d76a..671ee83 100644
@@ -514,3 +514,35 @@ int cpu__setup_cpunode_map(void)
 	closedir(dir1);
 	return 0;
 }
+
+u64 get_cpu_max_freq(void)
+{
+	const char *mnt;
+	char path[PATH_MAX], tmp;
+	FILE *fp;
+	u64 freq;
+	int cpu = 0;
+	int ret;
+
+	mnt = sysfs__mountpoint();
+	if (!mnt)
+		return 0;
+
+	snprintf(path, PATH_MAX, "%s/devices/system/cpu/online", mnt);
+	fp = fopen(path, "r");
+	if (fp) {
+		ret = fscanf(fp, "%u%c", &cpu, &tmp);
+		fclose(fp);
+		if (ret < 1)
+			return 0;
+	}
+
+	snprintf(path, PATH_MAX, "%s/devices/system/cpu/cpu%d/cpufreq/cpuinfo_max_freq", mnt, cpu);
+	fp = fopen(path, "r");
+	if (!fp)
+		return 0;
+	ret = fscanf(fp, "%lu", &freq);
+	fclose(fp);
+
+	return (ret == 1) ? freq : 0;
+}
tools/perf/util/cpumap.h +1 -0
index 8982d53..06cd2c4 100644
@@ -60,6 +60,7 @@ int max_node_num;
 int *cpunode_map;
 
 int cpu__setup_cpunode_map(void);
+u64 get_cpu_max_freq(void);
 
 static inline int cpu__max_node(void)
 {
tools/perf/util/header.c +59 -0

83 lines not on this page · raw message

tools/perf/util/header.h +12 -0
index 975d803..dd9f6b0 100644
@@ -31,6 +31,7 @@ enum {
 	HEADER_PMU_MAPPINGS,
 	HEADER_GROUP_DESC,
 	HEADER_AUXTRACE,
+	HEADER_CPU_ATTR,
 	HEADER_LAST_FEATURE,
 	HEADER_FEAT_BITS	= 256,
 };
@@ -71,6 +72,11 @@ struct cpu_topology_map {
 	int	core_id;
 };
 
+enum perf_header_cpu_attr {
+	PERF_HEADER_CPU_MAX_FREQ	= 0,
+	PERF_HEADER_CPU_ATTR_MAX,
+};
+
 struct perf_env {
 	char			*hostname;
 	char			*os_release;
@@ -95,6 +101,12 @@ struct perf_env {
 	char			*numa_nodes;
 	char			*pmu_mappings;
 	struct cpu_topology_map	*cpu;
+	union  {
+		u64		cpu_attr[PERF_HEADER_CPU_ATTR_MAX];
+		struct {
+			u64	max_freq;
+		} cpuattr;
+	};
 };
 
 struct perf_header {
Signature · 1 line
1.8.3.1
#2 lore raw

PATCH v8 2/7 perf,tools: read msr pmu type from header.

In reply to [email protected]

From: Kan Liang <[email protected]>

Get msr pmu type when processing pmu_mappings

---
 tools/perf/util/header.c | 3 +++
 tools/perf/util/header.h | 1 +
 2 files changed, 4 insertions(+)
tools/perf/util/header.c +3 -0
index 3535dcb..04f415d 100644
@@ -1879,6 +1879,9 @@ static int process_pmu_mappings(struct perf_file_section *section __maybe_unused
 		/* include a NULL character at the end */
 		strbuf_add(&sb, "", 1);
 
+		if (!strcmp(name, "msr"))
+			ph->env.msr_pmu_type = type;
+
 		free(name);
 		pmu_num--;
 	}
tools/perf/util/header.h +1 -0
index dd9f6b0..d9034e8 100644
@@ -107,6 +107,7 @@ struct perf_env {
 			u64	max_freq;
 		} cpuattr;
 	};
+	unsigned int		msr_pmu_type;
 };
 
 struct perf_header {
Signature · 1 line
1.8.3.1
#3 lore raw

PATCH v8 3/7 perf, record: introduce --freq-perf option

In reply to [email protected]

From: Kan Liang <[email protected]>

To generate the frequency and performance output, perf must sample read special events like cycles, ref-cycles, msr/tsc/, msr/aperf/ or msr/mperf/.
With the --freq-perf option, perf record can automatically check and add those event into evlist for sampling read.

---
 tools/perf/Documentation/perf-record.txt |  4 ++++
 tools/perf/builtin-record.c              | 39 +++++++++++++++++++++++++++++++-
 tools/perf/util/session.h                | 10 ++++++++
 3 files changed, 52 insertions(+), 1 deletion(-)
tools/perf/Documentation/perf-record.txt +4 -0
index 2e9ce77..3f40712 100644
@@ -308,6 +308,10 @@ This option sets the time out limit. The default value is 500 ms.
 Record context switch events i.e. events of type PERF_RECORD_SWITCH or
 PERF_RECORD_SWITCH_CPU_WIDE.
 
+--freq-perf::
+Add frequency and performance related events to do sample read.
+These events include cycles, ref-cycles, msr/tsc/, msr/aperf/ and msr/mperf/.
+
 SEE ALSO
 --------
 linkperf:perf-stat[1], linkperf:perf-list[1]
tools/perf/builtin-record.c +38 -1
index 142eeb3..e87dda3 100644
@@ -13,7 +13,7 @@
 #include "util/util.h"
 #include "util/parse-options.h"
 #include "util/parse-events.h"
-
+#include "util/pmu.h"
 #include "util/callchain.h"
 #include "util/cgroup.h"
 #include "util/header.h"
@@ -50,6 +50,7 @@ struct record {
 	bool			no_buildid;
 	bool			no_buildid_cache;
 	long			samples;
+	bool			freq_perf;
 };
 
 static int record__write(struct record *rec, void *bf, size_t size)
@@ -948,6 +949,35 @@ out_free:
 	return ret;
 }
 
+const char *freq_perf_events[FREQ_PERF_MAX][3] = {
+	{ "msr", "tsc", "msr/tsc/" },
+	{ "msr", "aperf", "msr/aperf/" },
+	{ "msr", "mperf", "msr/mperf/" },
+	{ NULL, "cycles", "cycles" },
+	{ NULL, "ref-cycles", "ref-cycles" },
+};
+
+static int
+record_add_freq_perf_events(struct perf_evlist *evlist)
+{
+	int i;
+	char freq_perf_attrs[100];
+
+	strcpy(freq_perf_attrs, "{cycles,ref-cycles");
+	for (i = 0; i < FREQ_PERF_MAX; i++) {
+		if ((i == FREQ_PERF_CYCLES) ||
+		    (i == FREQ_PERF_REF_CYCLES))
+			continue;
+		if (pmu_have_event(freq_perf_events[i][0], freq_perf_events[i][1])) {
+			strcat(freq_perf_attrs, ",");
+			strcat(freq_perf_attrs, freq_perf_events[i][2]);
+		}
+	}
+	strcat(freq_perf_attrs, "}:S");
+
+	return parse_events(evlist, freq_perf_attrs, NULL);
+}
+
 static const char * const __record_usage[] = {
 	"perf record [<options>] [<command>]",
 	"perf record [<options>] -- <command> [<options>]",
@@ -1096,6 +1126,8 @@ struct option __record_options[] = {
 			"per thread proc mmap processing timeout in ms"),
 	OPT_BOOLEAN(0, "switch-events", &record.opts.record_switch_events,
 		    "Record context switch events"),
+	OPT_BOOLEAN(0, "freq-perf", &record.freq_perf,
+		    "Add frequency and performance related events to do sample read."),
 	OPT_END()
 };
 
@@ -1157,6 +1189,11 @@ int cmd_record(int argc, const char **argv, const char *prefix __maybe_unused)
 	if (rec->no_buildid_cache || rec->no_buildid)
 		disable_buildid_cache();
 
+	if (rec->freq_perf && record_add_freq_perf_events(rec->evlist)) {
+		pr_err("Cannot set up freq and performance events\n");
+		goto out_symbol_exit;
+	}
+
 	if (rec->evlist->nr_entries == 0 &&
 	    perf_evlist__add_default(rec->evlist) < 0) {
 		pr_err("Not enough memory for event selector list\n");
tools/perf/util/session.h +10 -0
index b44afc7..3915be7 100644
@@ -42,6 +42,16 @@ struct perf_session {
 #define PRINT_IP_OPT_ONELINE	(1<<4)
 #define PRINT_IP_OPT_SRCLINE	(1<<5)
 
+enum perf_freq_perf_index {
+	FREQ_PERF_TSC		= 0,
+	FREQ_PERF_APERF		= 1,
+	FREQ_PERF_MPERF		= 2,
+	FREQ_PERF_CYCLES	= 3,
+	FREQ_PERF_REF_CYCLES	= 4,
+
+	FREQ_PERF_MAX
+};
+
 struct perf_tool;
 
 struct perf_session *perf_session__new(struct perf_data_file *file,
Signature · 1 line
1.8.3.1
#4 lore raw

Re: [PATCH V8 3/7] perf, record: introduce --freq-perf option

Quoted 7 lines · On Thu, Sep 03, 2015 at 08:30:59AM -0400, [email protected] wrote:

From: Kan Liang <[email protected]>

To generate the frequency and performance output, perf must sample read special events like cycles, ref-cycles, msr/tsc/, msr/aperf/ or msr/mperf/.
With the --freq-perf option, perf record can automatically check and add those event into evlist for sampling read.

hum, not sure this is a big or small thing:

[jolsa@ibm-x3650m4-01 perf]$ ./perf record --freq-perf -e cache-misses ls

SNIP

[ perf record: Woken up 13 times to write data ]
[ perf record: Captured and wrote 0.014 MB perf.data (147 samples) ] [jolsa@ibm-x3650m4-01 perf]$ ./perf report -D
non matching read_format[jolsa@ibm-x3650m4-01 perf]$

jirka

Quoted 125 lines
---
 tools/perf/Documentation/perf-record.txt |  4 ++++
 tools/perf/builtin-record.c              | 39 +++++++++++++++++++++++++++++++-
 tools/perf/util/session.h                | 10 ++++++++
 3 files changed, 52 insertions(+), 1 deletion(-)
tools/perf/Documentation/perf-record.txt +4 -0
index 2e9ce77..3f40712 100644
@@ -308,6 +308,10 @@ This option sets the time out limit. The default value is 500 ms.
 Record context switch events i.e. events of type PERF_RECORD_SWITCH or
 PERF_RECORD_SWITCH_CPU_WIDE.
 
+--freq-perf::
+Add frequency and performance related events to do sample read.
+These events include cycles, ref-cycles, msr/tsc/, msr/aperf/ and msr/mperf/.
+
 SEE ALSO
 --------
 linkperf:perf-stat[1], linkperf:perf-list[1]
tools/perf/builtin-record.c +38 -1
index 142eeb3..e87dda3 100644
@@ -13,7 +13,7 @@
 #include "util/util.h"
 #include "util/parse-options.h"
 #include "util/parse-events.h"
-
+#include "util/pmu.h"
 #include "util/callchain.h"
 #include "util/cgroup.h"
 #include "util/header.h"
@@ -50,6 +50,7 @@ struct record {
 	bool			no_buildid;
 	bool			no_buildid_cache;
 	long			samples;
+	bool			freq_perf;
 };
 
 static int record__write(struct record *rec, void *bf, size_t size)
@@ -948,6 +949,35 @@ out_free:
 	return ret;
 }
 
+const char *freq_perf_events[FREQ_PERF_MAX][3] = {
+	{ "msr", "tsc", "msr/tsc/" },
+	{ "msr", "aperf", "msr/aperf/" },
+	{ "msr", "mperf", "msr/mperf/" },
+	{ NULL, "cycles", "cycles" },
+	{ NULL, "ref-cycles", "ref-cycles" },
+};
+
+static int
+record_add_freq_perf_events(struct perf_evlist *evlist)
+{
+	int i;
+	char freq_perf_attrs[100];
+
+	strcpy(freq_perf_attrs, "{cycles,ref-cycles");
+	for (i = 0; i < FREQ_PERF_MAX; i++) {
+		if ((i == FREQ_PERF_CYCLES) ||
+		    (i == FREQ_PERF_REF_CYCLES))
+			continue;
+		if (pmu_have_event(freq_perf_events[i][0], freq_perf_events[i][1])) {
+			strcat(freq_perf_attrs, ",");
+			strcat(freq_perf_attrs, freq_perf_events[i][2]);
+		}
+	}
+	strcat(freq_perf_attrs, "}:S");
+
+	return parse_events(evlist, freq_perf_attrs, NULL);
+}
+
 static const char * const __record_usage[] = {
 	"perf record [<options>] [<command>]",
 	"perf record [<options>] -- <command> [<options>]",
@@ -1096,6 +1126,8 @@ struct option __record_options[] = {
 			"per thread proc mmap processing timeout in ms"),
 	OPT_BOOLEAN(0, "switch-events", &record.opts.record_switch_events,
 		    "Record context switch events"),
+	OPT_BOOLEAN(0, "freq-perf", &record.freq_perf,
+		    "Add frequency and performance related events to do sample read."),
 	OPT_END()
 };
 
@@ -1157,6 +1189,11 @@ int cmd_record(int argc, const char **argv, const char *prefix __maybe_unused)
 	if (rec->no_buildid_cache || rec->no_buildid)
 		disable_buildid_cache();
 
+	if (rec->freq_perf && record_add_freq_perf_events(rec->evlist)) {
+		pr_err("Cannot set up freq and performance events\n");
+		goto out_symbol_exit;
+	}
+
 	if (rec->evlist->nr_entries == 0 &&
 	    perf_evlist__add_default(rec->evlist) < 0) {
 		pr_err("Not enough memory for event selector list\n");
tools/perf/util/session.h +10 -0
index b44afc7..3915be7 100644
@@ -42,6 +42,16 @@ struct perf_session {
 #define PRINT_IP_OPT_ONELINE	(1<<4)
 #define PRINT_IP_OPT_SRCLINE	(1<<5)
 
+enum perf_freq_perf_index {
+	FREQ_PERF_TSC		= 0,
+	FREQ_PERF_APERF		= 1,
+	FREQ_PERF_MPERF		= 2,
+	FREQ_PERF_CYCLES	= 3,
+	FREQ_PERF_REF_CYCLES	= 4,
+
+	FREQ_PERF_MAX
+};
+
 struct perf_tool;
 
 struct perf_session *perf_session__new(struct perf_data_file *file,
Signature · 1 line
1.8.3.1
author
#5 lore raw

RE: [PATCH V8 3/7] perf, record: introduce --freq-perf option

Quoted 14 lines · On Thu, Sep 03, 2015 at 08:30:59AM -0400, [email protected] wrote:
Quoted 7 lines

From: Kan Liang <[email protected]>

To generate the frequency and performance output, perf must sample read special events like cycles, ref-cycles, msr/tsc/, msr/aperf/ or msr/mperf/.
With the --freq-perf option, perf record can automatically check and add those event into evlist for sampling read.

hum, not sure this is a big or small thing:

Quoted 10 lines

[jolsa@ibm-x3650m4-01 perf]$ ./perf record --freq-perf -e cache-misses ls

SNIP

[ perf record: Woken up 13 times to write data ] [ perf record: Captured and wrote 0.014 MB perf.data (147 samples) ] [jolsa@ibm-x3650m4-01 perf]$ ./perf report -D non matching read_format[jolsa@ibm-x3650m4-01 perf]$

The issue is not caused by this patch. It's also an issue in current codes.

In perf_session__open, it will check sample_type, sample_id and read_format for each event by comparing it with the first event. If the first events group are sample read and the next event is normal event, it must error out.

I think it's not a bug. Because we should not allow to record events with different format.

Thanks,
Kan

Quoted 135 lines

jirka

Quoted 5 lines
---
 tools/perf/Documentation/perf-record.txt |  4 ++++
 tools/perf/builtin-record.c              | 39

+++++++++++++++++++++++++++++++-

tools/perf/Documentation/perf-record.txt This option sets the time out limit. The default

Quoted 9 lines
 tools/perf/util/session.h                | 10 ++++++++
 3 files changed, 52 insertions(+), 1 deletion(-)
tools/perf/Documentation/perf-record.txt +0 -0
index 2e9ce77..3f40712 100644
@@ -308,6 +308,10 @@ This option sets the time out limit. The default

value is 500 ms.

tools/perf/Documentation/perf-record.txt

Quoted 6 lines
tools/perf/Documentation/perf-record.txt +0 -0
 Record context switch events i.e. events of type PERF_RECORD_SWITCH

or PERF_RECORD_SWITCH_CPU_WIDE.

+--freq-perf::

tools/perf/Documentation/perf-record.txt +2 -0
+Add frequency and performance related events to do sample read.
+These events include cycles, ref-cycles, msr/tsc/, msr/aperf/ and

msr/mperf/.

Quoted 37 lines

+
SEE ALSO
--------
linkperf:perf-stat[1], linkperf:perf-list[1] diff --git
a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 142eeb3..e87dda3 100644

tools/perf/builtin-record.c +10 -1
@@ -13,7 +13,7 @@
 #include "util/util.h"
 #include "util/parse-options.h"
 #include "util/parse-events.h"
-
+#include "util/pmu.h"
 #include "util/callchain.h"
 #include "util/cgroup.h"
 #include "util/header.h"
@@ -50,6 +50,7 @@ struct record {
 	bool			no_buildid;
 	bool			no_buildid_cache;
 	long			samples;
+	bool			freq_perf;
 };
 
 static int record__write(struct record *rec, void *bf, size_t size)
@@ -948,6 +949,35 @@ out_free:
 	return ret;
 }
 
+const char *freq_perf_events[FREQ_PERF_MAX][3] = {
+	{ "msr", "tsc", "msr/tsc/" },
+	{ "msr", "aperf", "msr/aperf/" },
+	{ "msr", "mperf", "msr/mperf/" },
+	{ NULL, "cycles", "cycles" },
+	{ NULL, "ref-cycles", "ref-cycles" }, };
+
+static int

tools/perf/builtin-record.c out_free: · line 969

Quoted 10 lines
tools/perf/builtin-record.c +10 -0
+record_add_freq_perf_events(struct perf_evlist *evlist) {
+	int i;
+	char freq_perf_attrs[100];
+
+	strcpy(freq_perf_attrs, "{cycles,ref-cycles");
+	for (i = 0; i < FREQ_PERF_MAX; i++) {
+		if ((i == FREQ_PERF_CYCLES) ||
+		    (i == FREQ_PERF_REF_CYCLES))
+			continue;
+		if (pmu_have_event(freq_perf_events[i][0],

freq_perf_events[i][1])) {

Quoted 4 lines
tools/perf/builtin-record.c +4 -0
+			strcat(freq_perf_attrs, ",");
+			strcat(freq_perf_attrs, freq_perf_events[i][2]);
+		}
+	}

tools/perf/builtin-record.c out_free:

Quoted 10 lines
tools/perf/builtin-record.c +5 -0
+	strcat(freq_perf_attrs, "}:S");
+
+	return parse_events(evlist, freq_perf_attrs, NULL); }
+
 static const char * const __record_usage[] = {
 	"perf record [<options>] [<command>]",
 	"perf record [<options>] -- <command> [<options>]", @@ -1096,6
+1126,8 @@ struct option __record_options[] = {
 			"per thread proc mmap processing timeout in ms"),
 	OPT_BOOLEAN(0, "switch-events",

&record.opts.record_switch_events,

tools/perf/builtin-record.c struct option __record_options[] = {

Quoted 3 lines
tools/perf/builtin-record.c +2 -0
 		    "Record context switch events"),
+	OPT_BOOLEAN(0, "freq-perf", &record.freq_perf,
+		    "Add frequency and performance related events to do

sample

tools/perf/builtin-record.c int cmd_record(int argc, const char **argv,

Quoted 5 lines
tools/perf/builtin-record.c +1 -0
+read."),
 	OPT_END()
 };
@@ -1157,6 +1189,11 @@ int cmd_record(int argc, const char **argv,

const char *prefix __maybe_unused)

tools/perf/builtin-record.c int cmd_record(int argc, const char **argv, const char *prefix __maybe_unused)

Quoted 4 lines
tools/perf/builtin-record.c +1 -0
 	if (rec->no_buildid_cache || rec->no_buildid)
 		disable_buildid_cache();
 
+	if (rec->freq_perf && record_add_freq_perf_events(rec->evlist))

{

tools/perf/builtin-record.c int cmd_record(int argc, const char **argv, const char *prefix __maybe_unused) · line 1199

Quoted 7 lines
tools/perf/builtin-record.c +4 -0
+		pr_err("Cannot set up freq and performance events\n");
+		goto out_symbol_exit;
+	}
+
 	if (rec->evlist->nr_entries == 0 &&
 	    perf_evlist__add_default(rec->evlist) < 0) {
 		pr_err("Not enough memory for event selector list\n");

diff --git

Quoted 24 lines

a/tools/perf/util/session.h b/tools/perf/util/session.h index b44afc7..3915be7 100644

tools/perf/util/session.h +10 -1
@@ -42,6 +42,16 @@ struct perf_session {
 #define PRINT_IP_OPT_ONELINE	(1<<4)
 #define PRINT_IP_OPT_SRCLINE	(1<<5)
 
+enum perf_freq_perf_index {
+	FREQ_PERF_TSC		= 0,
+	FREQ_PERF_APERF		= 1,
+	FREQ_PERF_MPERF		= 2,
+	FREQ_PERF_CYCLES	= 3,
+	FREQ_PERF_REF_CYCLES	= 4,
+
+	FREQ_PERF_MAX
+};
+
 struct perf_tool;
 
 struct perf_session *perf_session__new(struct perf_data_file *file,
--

1.8.3.1

#6 lore raw

PATCH v8 4/7 perf,tools: fix dump_sample crash when evsel is NULL.

In reply to [email protected]

From: Kan Liang <[email protected]>

Need to check evsel before using in dump_sample.

---
 tools/perf/util/session.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/util/session.c +1 -1
index 61669be..23fed17 100644
@@ -1080,11 +1080,11 @@ static int machines__deliver_event(struct machines *machines,
 
 	switch (event->header.type) {
 	case PERF_RECORD_SAMPLE:
-		dump_sample(evsel, event, sample);
 		if (evsel == NULL) {
 			++evlist->stats.nr_unknown_id;
 			return 0;
 		}
+		dump_sample(evsel, event, sample);
 		if (machine == NULL) {
 			++evlist->stats.nr_unprocessable_samples;
 			return 0;
Signature · 1 line
1.8.3.1
#7 lore raw

[tip:perf/core] perf session: Don't call dump_sample() when evsel is NULL

perf session: Don't call dump_sample() when evsel is NULL
Commit-ID:  1b29ac59b1d692c06ec543a5f35e0d9ebb98e003
Gitweb:     http://git.kernel.org/tip/1b29ac59b1d692c06ec543a5f35e0d9ebb98e003
Author:     Kan Liang <[email protected]>
AuthorDate: Thu, 3 Sep 2015 08:31:00 -0400
Committer:  Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Fri, 4 Sep 2015 12:01:04 -0300

perf session: Don't call dump_sample() when evsel is NULL

Need to check evsel before passing it to dump_sample().

Cc Adrian Hunter <[email protected]>
Cc Andi Kleen <[email protected]>
Cc Andy Lutomirski <[email protected]>
Cc Jiri Olsa <[email protected]>
Cc Mark Rutland <[email protected]>
Cc Namhyung Kim <[email protected]>
Cc Peter Zijlstra <[email protected]>
Cc Stephane Eranian <[email protected]>
---
 tools/perf/util/session.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/util/session.c +1 -1
index 61669be..23fed17 100644
@@ -1080,11 +1080,11 @@ static int machines__deliver_event(struct machines *machines,
 
 	switch (event->header.type) {
 	case PERF_RECORD_SAMPLE:
-		dump_sample(evsel, event, sample);
 		if (evsel == NULL) {
 			++evlist->stats.nr_unknown_id;
 			return 0;
 		}
+		dump_sample(evsel, event, sample);
 		if (machine == NULL) {
 			++evlist->stats.nr_unprocessable_samples;
 			return 0;
#8 lore raw

PATCH v8 5/7 perf,tools: Dump per-sample freq/CPU%/CORE_BUSY% in report -D

In reply to [email protected]

From: Kan Liang <[email protected]>

The group read results from cycles/ref-cycles/TSC/ASTATE/MSTATE event can be used to calculate the frequency, CPU Utilization and percent performance during each sampling period.
This patch shows them in report -D.

Here is an example:

$ perf record --freq-perf  ~/tchain_edit

Here is one sample from perf report -D

1972044565107 0x3498 [0x88]: PERF_RECORD_SAMPLE(IP, 0x2): 10608/10608:
0x4005fd period: 564686 addr: 0
... sample_read:
.... group nr 5
..... id 0000000000000012, value 0000000002143901
..... id 0000000000000052, value 0000000002143896
..... id 0000000000000094, value 00000000021e443d
..... id 00000000000000d4, value 00000000021db984
..... id 0000000000000114, value 00000000021db964
..... Freq 2301 MHz
..... CPU% 98%
..... CORE_BUSY% 99%

---
 tools/perf/util/session.c | 28 +++++++++++++++++++++---
 tools/perf/util/session.h | 54 +++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 79 insertions(+), 3 deletions(-)
tools/perf/util/session.c +25 -3
index 23fed17..ba763f7 100644
@@ -7,7 +7,6 @@
 #include <sys/mman.h>
 
 #include "evlist.h"
-#include "evsel.h"
 #include "session.h"
 #include "tool.h"
 #include "sort.h"
@@ -878,8 +877,15 @@ static void perf_evlist__print_tstamp(struct perf_evlist *evlist,
 		printf("%" PRIu64 " ", sample->time);
 }
 
-static void sample_read__printf(struct perf_sample *sample, u64 read_format)
+static void sample_read__printf(struct perf_sample *sample,
+				struct perf_evsel *evsel)
 {
+	u64 read_format = evsel->attr.read_format;
+	struct perf_evlist *evlist = evsel->evlist;
+	struct perf_sample_id *sid;
+	perf_freq_t data = { 0 };
+	u64 cpu_max_freq = evlist->env->cpuattr.max_freq;
+
 	printf("... sample_read:\n");
 
 	if (read_format & PERF_FORMAT_TOTAL_TIME_ENABLED)
@@ -902,10 +908,26 @@ static void sample_read__printf(struct perf_sample *sample, u64 read_format)
 			printf("..... id %016" PRIx64
 			       ", value %016" PRIx64 "\n",
 			       value->id, value->value);
+
+			sid = perf_evlist__id2sid(evlist, value->id);
+			evsel = sid->evsel;
+			if (evsel != NULL)
+				perf_freq__init(evlist->env->msr_pmu_type,
+						evsel, data, value->value);
 		}
 	} else
 		printf("..... id %016" PRIx64 ", value %016" PRIx64 "\n",
 			sample->read.one.id, sample->read.one.value);
+
+	if (perf_freq__has_freq(data))
+		printf("..... Freq %lu MHz\n",
+		       perf_freq__get_freq(data, cpu_max_freq/1000));
+	if (perf_freq__has_cpu_util(data))
+		printf("..... CPU%% %lu%%\n",
+		       perf_freq__get_cpu_util(data));
+	if (perf_freq__has_core_busy(data))
+		printf("..... CORE_BUSY%% %lu%%\n",
+		       perf_freq__get_core_busy(data));
 }
 
 static void dump_event(struct perf_evlist *evlist, union perf_event *event,
@@ -965,7 +987,7 @@ static void dump_sample(struct perf_evsel *evsel, union perf_event *event,
 		printf("... transaction: %" PRIx64 "\n", sample->transaction);
 
 	if (sample_type & PERF_SAMPLE_READ)
-		sample_read__printf(sample, evsel->attr.read_format);
+		sample_read__printf(sample, evsel);
 }
 
 static struct machine *machines__find_for_cpumode(struct machines *machines,
tools/perf/util/session.h +54 -0

72 lines not on this page · raw message

Signature · 1 line
1.8.3.1
#9 lore raw

PATCH v8 6/7 perf,tools: caculate and save freq/CPU%/CORE_BUSY% in he_stat

In reply to [email protected]

From: Kan Liang <[email protected]>

Caculate freq/CPU%/CORE_BUSY% in add_entry_cb, and update the value in he_stat.

---
 tools/perf/builtin-report.c | 36 ++++++++++++++++++++++++++++++++++++
 tools/perf/util/sort.h      |  3 +++
 2 files changed, 39 insertions(+)
tools/perf/builtin-report.c +36 -0
index 62b285e..7673f1c 100644
@@ -89,6 +89,38 @@ static int report__config(const char *var, const char *value, void *cb)
 	return perf_default_config(var, value, cb);
 }
 
+static void set_he_freq_perf(struct perf_session *session,
+			     struct hist_entry_iter *iter)
+{
+	struct hist_entry *he = iter->he;
+	struct perf_evsel *evsel = iter->evsel;
+	struct perf_evsel *leader = evsel;
+	struct perf_sample *sample = iter->sample;
+	struct perf_evlist *evlist = session->evlist;
+	u64 cpu_max_freq = session->header.env.cpuattr.max_freq;
+	perf_freq_t data = { 0 };
+	u64 nr = 0;
+
+	perf_freq__init(session->header.env.msr_pmu_type,
+			evsel, data,
+			sample->read.group.values[nr].value);
+	evlist__for_each_continue(evlist, evsel) {
+		if ((evsel->leader != leader) ||
+		     (++nr >= sample->read.group.nr))
+			break;
+		perf_freq__init(session->header.env.msr_pmu_type,
+				evsel, data,
+				sample->read.group.values[nr].value);
+	}
+
+	if (perf_freq__has_freq(data))
+		he->stat.freq = perf_freq__get_freq(data, cpu_max_freq/1000);
+	if (perf_freq__has_cpu_util(data))
+		he->stat.cpu_util = perf_freq__get_cpu_util(data);
+	if (perf_freq__has_core_busy(data))
+		he->stat.core_busy = perf_freq__get_core_busy(data);
+}
+
 static int hist_iter__report_callback(struct hist_entry_iter *iter,
 				      struct addr_location *al, bool single,
 				      void *arg)
@@ -100,6 +132,10 @@ static int hist_iter__report_callback(struct hist_entry_iter *iter,
 	struct mem_info *mi;
 	struct branch_info *bi;
 
+	if ((iter->ops == &hist_iter_normal) &&
+	    perf_evsel__is_group_leader(evsel))
+		set_he_freq_perf(rep->session, iter);
+
 	if (!ui__has_annotation())
 		return 0;
 
tools/perf/util/sort.h +3 -0
index 3c2a399..9ed52e6 100644
@@ -54,6 +54,9 @@ struct he_stat {
 	u64			period_guest_us;
 	u64			weight;
 	u32			nr_events;
+	u64			freq;
+	u64			cpu_util;
+	u64			core_busy;
 };
 
 struct hist_entry_diff {
Signature · 1 line
1.8.3.1
#10 lore raw

PATCH v8 7/7 perf,tools: Show freq/CPU%/CORE_BUSY% in perf report --stdio

In reply to [email protected]

From: Kan Liang <[email protected]>

Show frequency, CPU Utilization and percent performance for each symbol in perf report by --stdio --show-freq-perf

In sampling group, only group leader do sampling. So only need to print group leader's freq in --group.

Here is an example.

$ perf report --stdio --group --show-freq-perf

Overhead FREQ MHz CPU% CORE_BUSY%
Command Shared Object Symbol
........................................ ......... ..... ..........
........... ................ ......................

    99.54%  99.54%  99.53%  99.53%  99.53%       2301     96         99
tchain_edit  tchain_edit       [.] f3
     0.20%   0.20%   0.20%   0.20%   0.20%       2301     98         99
tchain_edit  tchain_edit       [.] f2
     0.05%   0.05%   0.05%   0.05%   0.05%       2300     98         99
tchain_edit  [kernel.vmlinux]  [k] read_tsc
---
 tools/perf/Documentation/perf-report.txt | 12 ++++++
 tools/perf/builtin-report.c              | 22 +++++++++-
 tools/perf/ui/hist.c                     | 71 +++++++++++++++++++++++++++++---
 tools/perf/util/hist.h                   |  3 ++
 tools/perf/util/session.c                | 33 ++++++++-------
 tools/perf/util/sort.c                   |  3 ++
 tools/perf/util/symbol.h                 | 12 +++++-
 7 files changed, 134 insertions(+), 22 deletions(-)
tools/perf/Documentation/perf-report.txt +12 -0
index 9c7981b..6280e22 100644
@@ -306,6 +306,18 @@ OPTIONS
 	special event -e cpu/mem-loads/ or -e cpu/mem-stores/. See
 	'perf mem' for simpler access.
 
+--show-freq-perf::
+	Show CPU frequency and performance result from sample read.
+	To generate the frequency and performance output, the perf.data file
+	must have been obtained by group read and using special events cycles,
+	ref-cycles, msr/tsc/, msr/aperf/ or msr/mperf/
+	Freq MHz: The frequency during the sample interval. Needs cycles and
+		  ref-cycles event.
+	CPU%: CPU utilization during the sample interval. Needs ref-cycles and
+	      msr/tsc/ events.
+	CORE_BUSY%: actual percent performance (APERF/MPERF%) during the
+		    sample interval. Needs msr/aperf/ and msr/mperf/ events.
+
 --percent-limit::
 	Do not show entries which have an overhead under that percent.
 	(Default: 0).
tools/perf/builtin-report.c +21 -1
index 7673f1c..ebcbe4c 100644
@@ -133,7 +133,8 @@ static int hist_iter__report_callback(struct hist_entry_iter *iter,
 	struct branch_info *bi;
 
 	if ((iter->ops == &hist_iter_normal) &&
-	    perf_evsel__is_group_leader(evsel))
+	    perf_evsel__is_group_leader(evsel) &&
+	    symbol_conf.show_freq_perf)
 		set_he_freq_perf(rep->session, iter);
 
 	if (!ui__has_annotation())
@@ -772,6 +773,8 @@ int cmd_report(int argc, const char **argv, const char *prefix __maybe_unused)
 	OPT_BOOLEAN(0, "demangle-kernel", &symbol_conf.demangle_kernel,
 		    "Enable kernel symbol demangling"),
 	OPT_BOOLEAN(0, "mem-mode", &report.mem_mode, "mem access profile"),
+	OPT_BOOLEAN(0, "show-freq-perf", &symbol_conf.show_freq_perf,
+		    "show CPU freqency and performance info"),
 	OPT_CALLBACK(0, "percent-limit", &report, "percent",
 		     "Don't show entries under that percent", parse_percent_limit),
 	OPT_CALLBACK(0, "percentage", NULL, "relative|absolute",
@@ -788,7 +791,9 @@ int cmd_report(int argc, const char **argv, const char *prefix __maybe_unused)
 	struct perf_data_file file = {
 		.mode  = PERF_DATA_MODE_READ,
 	};
+	struct perf_evsel *pos;
 	int ret = hists__init();
+	perf_freq_t freq_data_status = { 0 };
 
 	if (ret < 0)
 		return ret;
@@ -873,6 +878,21 @@ repeat:
 		symbol_conf.cumulate_callchain = false;
 	}
 
+
+	if (symbol_conf.show_freq_perf) {
+		symbol_conf.freq_perf_type = 0;
+		evlist__for_each(session->evlist, pos) {
+			perf_freq__init(session->header.env.msr_pmu_type,
+					pos, freq_data_status, 1);
+		}
+		if (perf_freq__has_freq(freq_data_status))
+			symbol_conf.freq_perf_type |= 1U << DISPLAY_FREQ;
+		if (perf_freq__has_cpu_util(freq_data_status))
+			symbol_conf.freq_perf_type |= 1U << DISPLAY_CPU_UTIL;
+		if (perf_freq__has_core_busy(freq_data_status))
+			symbol_conf.freq_perf_type |= 1U << DISPLAY_CORE_BUSY;
+	}
+
 	if (setup_sorting() < 0) {
 		if (sort_order)
 			parse_options_usage(report_usage, options, "s", 1);
tools/perf/ui/hist.c +66 -5

140 lines not on this page · raw message

tools/perf/util/hist.h +3 -0
index de6d58e..df07f28 100644
@@ -237,6 +237,9 @@ enum {
 	PERF_HPP__OVERHEAD_ACC,
 	PERF_HPP__SAMPLES,
 	PERF_HPP__PERIOD,
+	PERF_HPP__FREQ,
+	PERF_HPP__CPU_UTIL,
+	PERF_HPP__CORE_BUSY,
 
 	PERF_HPP__MAX_INDEX
 };
tools/perf/util/session.c +18 -15
index ba763f7..071fe65 100644
@@ -908,26 +908,29 @@ static void sample_read__printf(struct perf_sample *sample,
 			printf("..... id %016" PRIx64
 			       ", value %016" PRIx64 "\n",
 			       value->id, value->value);
-
-			sid = perf_evlist__id2sid(evlist, value->id);
-			evsel = sid->evsel;
-			if (evsel != NULL)
-				perf_freq__init(evlist->env->msr_pmu_type,
-						evsel, data, value->value);
+			if (symbol_conf.show_freq_perf) {
+				sid = perf_evlist__id2sid(evlist, value->id);
+				evsel = sid->evsel;
+				if (evsel != NULL)
+					perf_freq__init(evlist->env->msr_pmu_type,
+							evsel, data, value->value);
+			}
 		}
 	} else
 		printf("..... id %016" PRIx64 ", value %016" PRIx64 "\n",
 			sample->read.one.id, sample->read.one.value);
 
-	if (perf_freq__has_freq(data))
-		printf("..... Freq %lu MHz\n",
-		       perf_freq__get_freq(data, cpu_max_freq/1000));
-	if (perf_freq__has_cpu_util(data))
-		printf("..... CPU%% %lu%%\n",
-		       perf_freq__get_cpu_util(data));
-	if (perf_freq__has_core_busy(data))
-		printf("..... CORE_BUSY%% %lu%%\n",
-		       perf_freq__get_core_busy(data));
+	if (symbol_conf.show_freq_perf) {
+		if (perf_freq__has_freq(data))
+			printf("..... Freq %lu MHz\n",
+			       perf_freq__get_freq(data, cpu_max_freq/1000));
+		if (perf_freq__has_cpu_util(data))
+			printf("..... CPU%% %lu%%\n",
+			       perf_freq__get_cpu_util(data));
+		if (perf_freq__has_core_busy(data))
+			printf("..... CORE_BUSY%% %lu%%\n",
+			       perf_freq__get_core_busy(data));
+	}
 }
 
 static void dump_event(struct perf_evlist *evlist, union perf_event *event,
tools/perf/util/sort.c +3 -0
index a97bcee..1d5c7d8 100644
@@ -1303,6 +1303,9 @@ static struct hpp_dimension hpp_sort_dimensions[] = {
 	DIM(PERF_HPP__OVERHEAD_ACC, "overhead_children"),
 	DIM(PERF_HPP__SAMPLES, "sample"),
 	DIM(PERF_HPP__PERIOD, "period"),
+	DIM(PERF_HPP__FREQ, "freq"),
+	DIM(PERF_HPP__CPU_UTIL, "cpu_u"),
+	DIM(PERF_HPP__CORE_BUSY, "core_busy"),
 };
 
 #undef DIM
tools/perf/util/symbol.h +11 -1
index 440ba8a..8b21fb4 100644
@@ -80,6 +80,14 @@ static inline size_t symbol__size(const struct symbol *sym)
 struct strlist;
 struct intlist;
 
+enum freq_perf_type_index {
+	DISPLAY_FREQ		= 0,
+	DISPLAY_CPU_UTIL,
+	DISPLAY_CORE_BUSY,
+
+	DISPLAY_MAX
+};
+
 struct symbol_conf {
 	unsigned short	priv_size;
 	unsigned short	nr_events;
@@ -107,7 +115,8 @@ struct symbol_conf {
 			show_hist_headers,
 			branch_callstack,
 			has_filter,
-			show_ref_callgraph;
+			show_ref_callgraph,
+			show_freq_perf;
 	const char	*vmlinux_name,
 			*kallsyms_name,
 			*source_prefix,
@@ -132,6 +141,7 @@ struct symbol_conf {
 	struct intlist	*pid_list,
 			*tid_list;
 	const char	*symfs;
+	u64		freq_perf_type;
 };
 
 extern struct symbol_conf symbol_conf;
Signature · 1 line
1.8.3.1
#11 lore raw

Re: [PATCH V8 7/7] perf,tools: Show freq/CPU%/CORE_BUSY% in perf report --stdio

Quoted 11 lines · On Thu, Sep 03, 2015 at 08:31:03AM -0400, [email protected] wrote:

From: Kan Liang <[email protected]>

Show frequency, CPU Utilization and percent performance for each symbol in perf report by --stdio --show-freq-perf

In sampling group, only group leader do sampling. So only need to print group leader's freq in --group.

Here is an example.

$ perf report --stdio --group --show-freq-perf

Ithink the --show-freq-perf option should imply --group, so user would type:

  perf record --freq-perf
  perf report --show-freq-perf

I wonder we should use --freq-perf for report as well 'perf report --freq-perf'

jirka

Quoted 374 lines

Overhead FREQ MHz CPU% CORE_BUSY%
Command Shared Object Symbol
........................................ ......... ..... ..........
........... ................ ......................

    99.54%  99.54%  99.53%  99.53%  99.53%       2301     96         99
tchain_edit  tchain_edit       [.] f3
     0.20%   0.20%   0.20%   0.20%   0.20%       2301     98         99
tchain_edit  tchain_edit       [.] f2
     0.05%   0.05%   0.05%   0.05%   0.05%       2300     98         99
tchain_edit  [kernel.vmlinux]  [k] read_tsc
---
 tools/perf/Documentation/perf-report.txt | 12 ++++++
 tools/perf/builtin-report.c              | 22 +++++++++-
 tools/perf/ui/hist.c                     | 71 +++++++++++++++++++++++++++++---
 tools/perf/util/hist.h                   |  3 ++
 tools/perf/util/session.c                | 33 ++++++++-------
 tools/perf/util/sort.c                   |  3 ++
 tools/perf/util/symbol.h                 | 12 +++++-
 7 files changed, 134 insertions(+), 22 deletions(-)
tools/perf/Documentation/perf-report.txt +12 -0
index 9c7981b..6280e22 100644
@@ -306,6 +306,18 @@ OPTIONS
 	special event -e cpu/mem-loads/ or -e cpu/mem-stores/. See
 	'perf mem' for simpler access.
 
+--show-freq-perf::
+	Show CPU frequency and performance result from sample read.
+	To generate the frequency and performance output, the perf.data file
+	must have been obtained by group read and using special events cycles,
+	ref-cycles, msr/tsc/, msr/aperf/ or msr/mperf/
+	Freq MHz: The frequency during the sample interval. Needs cycles and
+		  ref-cycles event.
+	CPU%: CPU utilization during the sample interval. Needs ref-cycles and
+	      msr/tsc/ events.
+	CORE_BUSY%: actual percent performance (APERF/MPERF%) during the
+		    sample interval. Needs msr/aperf/ and msr/mperf/ events.
+
 --percent-limit::
 	Do not show entries which have an overhead under that percent.
 	(Default: 0).
tools/perf/builtin-report.c +21 -1
index 7673f1c..ebcbe4c 100644
@@ -133,7 +133,8 @@ static int hist_iter__report_callback(struct hist_entry_iter *iter,
 	struct branch_info *bi;
 
 	if ((iter->ops == &hist_iter_normal) &&
-	    perf_evsel__is_group_leader(evsel))
+	    perf_evsel__is_group_leader(evsel) &&
+	    symbol_conf.show_freq_perf)
 		set_he_freq_perf(rep->session, iter);
 
 	if (!ui__has_annotation())
@@ -772,6 +773,8 @@ int cmd_report(int argc, const char **argv, const char *prefix __maybe_unused)
 	OPT_BOOLEAN(0, "demangle-kernel", &symbol_conf.demangle_kernel,
 		    "Enable kernel symbol demangling"),
 	OPT_BOOLEAN(0, "mem-mode", &report.mem_mode, "mem access profile"),
+	OPT_BOOLEAN(0, "show-freq-perf", &symbol_conf.show_freq_perf,
+		    "show CPU freqency and performance info"),
 	OPT_CALLBACK(0, "percent-limit", &report, "percent",
 		     "Don't show entries under that percent", parse_percent_limit),
 	OPT_CALLBACK(0, "percentage", NULL, "relative|absolute",
@@ -788,7 +791,9 @@ int cmd_report(int argc, const char **argv, const char *prefix __maybe_unused)
 	struct perf_data_file file = {
 		.mode  = PERF_DATA_MODE_READ,
 	};
+	struct perf_evsel *pos;
 	int ret = hists__init();
+	perf_freq_t freq_data_status = { 0 };
 
 	if (ret < 0)
 		return ret;
@@ -873,6 +878,21 @@ repeat:
 		symbol_conf.cumulate_callchain = false;
 	}
 
+
+	if (symbol_conf.show_freq_perf) {
+		symbol_conf.freq_perf_type = 0;
+		evlist__for_each(session->evlist, pos) {
+			perf_freq__init(session->header.env.msr_pmu_type,
+					pos, freq_data_status, 1);
+		}
+		if (perf_freq__has_freq(freq_data_status))
+			symbol_conf.freq_perf_type |= 1U << DISPLAY_FREQ;
+		if (perf_freq__has_cpu_util(freq_data_status))
+			symbol_conf.freq_perf_type |= 1U << DISPLAY_CPU_UTIL;
+		if (perf_freq__has_core_busy(freq_data_status))
+			symbol_conf.freq_perf_type |= 1U << DISPLAY_CORE_BUSY;
+	}
+
 	if (setup_sorting() < 0) {
 		if (sort_order)
 			parse_options_usage(report_usage, options, "s", 1);
tools/perf/ui/hist.c +66 -5
index 25d6083..c2be455 100644
@@ -17,7 +17,7 @@
 
 static int __hpp__fmt(struct perf_hpp *hpp, struct hist_entry *he,
 		      hpp_field_fn get_field, const char *fmt, int len,
-		      hpp_snprint_fn print_fn, bool fmt_percent)
+		      hpp_snprint_fn print_fn, bool fmt_percent, bool single)
 {
 	int ret;
 	struct hists *hists = he->hists;
@@ -36,7 +36,7 @@ static int __hpp__fmt(struct perf_hpp *hpp, struct hist_entry *he,
 	} else
 		ret = hpp__call_print_fn(hpp, print_fn, fmt, len, get_field(he));
 
-	if (perf_evsel__is_group_event(evsel)) {
+	if (perf_evsel__is_group_event(evsel) && !single) {
 		int prev_idx, idx_delta;
 		struct hist_entry *pair;
 		int nr_members = evsel->nr_members;
@@ -109,10 +109,16 @@ int hpp__fmt(struct perf_hpp_fmt *fmt, struct perf_hpp *hpp,
 	     const char *fmtstr, hpp_snprint_fn print_fn, bool fmt_percent)
 {
 	int len = fmt->user_len ?: fmt->len;
+	bool single = false;
+
+	if (((fmt == &perf_hpp__format[PERF_HPP__FREQ]) ||
+	     (fmt == &perf_hpp__format[PERF_HPP__CPU_UTIL]) ||
+	     (fmt == &perf_hpp__format[PERF_HPP__CORE_BUSY])))
+		single = true;
 
 	if (symbol_conf.field_sep) {
 		return __hpp__fmt(hpp, he, get_field, fmtstr, 1,
-				  print_fn, fmt_percent);
+				  print_fn, fmt_percent, single);
 	}
 
 	if (fmt_percent)
@@ -120,7 +126,7 @@ int hpp__fmt(struct perf_hpp_fmt *fmt, struct perf_hpp *hpp,
 	else
 		len -= 1;
 
-	return  __hpp__fmt(hpp, he, get_field, fmtstr, len, print_fn, fmt_percent);
+	return  __hpp__fmt(hpp, he, get_field, fmtstr, len, print_fn, fmt_percent, single);
 }
 
 int hpp__fmt_acc(struct perf_hpp_fmt *fmt, struct perf_hpp *hpp,
@@ -234,6 +240,30 @@ static int hpp__header_fn(struct perf_hpp_fmt *fmt, struct perf_hpp *hpp,
 	return scnprintf(hpp->buf, hpp->size, "%*s", len, fmt->name);
 }
 
+static int hpp__single_width_fn(struct perf_hpp_fmt *fmt,
+			 struct perf_hpp *hpp __maybe_unused,
+			 struct perf_evsel *evsel)
+{
+	int len = fmt->user_len ?: fmt->len;
+
+	if (symbol_conf.event_group && !symbol_conf.show_freq_perf)
+		len = max(len, evsel->nr_members * fmt->len);
+
+	if (len < (int)strlen(fmt->name))
+		len = strlen(fmt->name);
+
+	return len;
+}
+
+static int hpp__single_header_fn(struct perf_hpp_fmt *fmt, struct perf_hpp *hpp,
+			  struct perf_evsel *evsel)
+{
+	int len = hpp__single_width_fn(fmt, hpp, evsel);
+
+	return scnprintf(hpp->buf, hpp->size, "%*s", len, fmt->name);
+}
+
+
 static int hpp_color_scnprintf(struct perf_hpp *hpp, const char *fmt, ...)
 {
 	va_list args;
@@ -363,6 +393,9 @@ HPP_PERCENT_ACC_FNS(overhead_acc, period)
 
 HPP_RAW_FNS(samples, nr_events)
 HPP_RAW_FNS(period, period)
+HPP_RAW_FNS(freq, freq)
+HPP_RAW_FNS(cpu_util, cpu_util)
+HPP_RAW_FNS(core_busy, core_busy)
 
 static int64_t hpp__nop_cmp(struct perf_hpp_fmt *fmt __maybe_unused,
 			    struct hist_entry *a __maybe_unused,
@@ -395,6 +428,17 @@ static int64_t hpp__nop_cmp(struct perf_hpp_fmt *fmt __maybe_unused,
 		.sort	= hpp__sort_ ## _fn,		\
 	}
 
+#define HPP__SINGLE_PRINT_FNS(_name, _fn)		\
+	{						\
+		.name   = _name,			\
+		.header	= hpp__single_header_fn,	\
+		.width	= hpp__single_width_fn,		\
+		.entry	= hpp__entry_ ## _fn,		\
+		.cmp	= hpp__nop_cmp,			\
+		.collapse = hpp__nop_cmp,		\
+		.sort	= hpp__sort_ ## _fn,		\
+	}
+
 #define HPP__PRINT_FNS(_name, _fn)			\
 	{						\
 		.name   = _name,			\
@@ -414,7 +458,10 @@ struct perf_hpp_fmt perf_hpp__format[] = {
 	HPP__COLOR_PRINT_FNS("guest usr", overhead_guest_us),
 	HPP__COLOR_ACC_PRINT_FNS("Children", overhead_acc),
 	HPP__PRINT_FNS("Samples", samples),
-	HPP__PRINT_FNS("Period", period)
+	HPP__PRINT_FNS("Period", period),
+	HPP__SINGLE_PRINT_FNS("FREQ MHz", freq),
+	HPP__SINGLE_PRINT_FNS("CPU%", cpu_util),
+	HPP__SINGLE_PRINT_FNS("CORE_BUSY%", core_busy)
 };
 
 LIST_HEAD(perf_hpp__list);
@@ -485,6 +532,15 @@ void perf_hpp__init(void)
 	if (symbol_conf.show_total_period)
 		perf_hpp__column_enable(PERF_HPP__PERIOD);
 
+	if (symbol_conf.show_freq_perf) {
+		if (symbol_conf.freq_perf_type & (1U << DISPLAY_FREQ))
+			perf_hpp__column_enable(PERF_HPP__FREQ);
+		if (symbol_conf.freq_perf_type & (1U << DISPLAY_CPU_UTIL))
+			perf_hpp__column_enable(PERF_HPP__CPU_UTIL);
+		if (symbol_conf.freq_perf_type & (1U << DISPLAY_CORE_BUSY))
+			perf_hpp__column_enable(PERF_HPP__CORE_BUSY);
+	}
+
 	/* prepend overhead field for backward compatiblity.  */
 	list = &perf_hpp__format[PERF_HPP__OVERHEAD].sort_list;
 	if (list_empty(list))
@@ -652,6 +708,9 @@ void perf_hpp__reset_width(struct perf_hpp_fmt *fmt, struct hists *hists)
 		return;
 
 	switch (idx) {
+	case PERF_HPP__CPU_UTIL:
+		fmt->len = 5;
+		break;
 	case PERF_HPP__OVERHEAD:
 	case PERF_HPP__OVERHEAD_SYS:
 	case PERF_HPP__OVERHEAD_US:
@@ -661,6 +720,8 @@ void perf_hpp__reset_width(struct perf_hpp_fmt *fmt, struct hists *hists)
 
 	case PERF_HPP__OVERHEAD_GUEST_SYS:
 	case PERF_HPP__OVERHEAD_GUEST_US:
+	case PERF_HPP__FREQ:
+	case PERF_HPP__CORE_BUSY:
 		fmt->len = 9;
 		break;
 
tools/perf/util/hist.h +3 -0
index de6d58e..df07f28 100644
@@ -237,6 +237,9 @@ enum {
 	PERF_HPP__OVERHEAD_ACC,
 	PERF_HPP__SAMPLES,
 	PERF_HPP__PERIOD,
+	PERF_HPP__FREQ,
+	PERF_HPP__CPU_UTIL,
+	PERF_HPP__CORE_BUSY,
 
 	PERF_HPP__MAX_INDEX
 };
tools/perf/util/session.c +18 -15
index ba763f7..071fe65 100644
@@ -908,26 +908,29 @@ static void sample_read__printf(struct perf_sample *sample,
 			printf("..... id %016" PRIx64
 			       ", value %016" PRIx64 "\n",
 			       value->id, value->value);
-
-			sid = perf_evlist__id2sid(evlist, value->id);
-			evsel = sid->evsel;
-			if (evsel != NULL)
-				perf_freq__init(evlist->env->msr_pmu_type,
-						evsel, data, value->value);
+			if (symbol_conf.show_freq_perf) {
+				sid = perf_evlist__id2sid(evlist, value->id);
+				evsel = sid->evsel;
+				if (evsel != NULL)
+					perf_freq__init(evlist->env->msr_pmu_type,
+							evsel, data, value->value);
+			}
 		}
 	} else
 		printf("..... id %016" PRIx64 ", value %016" PRIx64 "\n",
 			sample->read.one.id, sample->read.one.value);
 
-	if (perf_freq__has_freq(data))
-		printf("..... Freq %lu MHz\n",
-		       perf_freq__get_freq(data, cpu_max_freq/1000));
-	if (perf_freq__has_cpu_util(data))
-		printf("..... CPU%% %lu%%\n",
-		       perf_freq__get_cpu_util(data));
-	if (perf_freq__has_core_busy(data))
-		printf("..... CORE_BUSY%% %lu%%\n",
-		       perf_freq__get_core_busy(data));
+	if (symbol_conf.show_freq_perf) {
+		if (perf_freq__has_freq(data))
+			printf("..... Freq %lu MHz\n",
+			       perf_freq__get_freq(data, cpu_max_freq/1000));
+		if (perf_freq__has_cpu_util(data))
+			printf("..... CPU%% %lu%%\n",
+			       perf_freq__get_cpu_util(data));
+		if (perf_freq__has_core_busy(data))
+			printf("..... CORE_BUSY%% %lu%%\n",
+			       perf_freq__get_core_busy(data));
+	}
 }
 
 static void dump_event(struct perf_evlist *evlist, union perf_event *event,
tools/perf/util/sort.c +3 -0
index a97bcee..1d5c7d8 100644
@@ -1303,6 +1303,9 @@ static struct hpp_dimension hpp_sort_dimensions[] = {
 	DIM(PERF_HPP__OVERHEAD_ACC, "overhead_children"),
 	DIM(PERF_HPP__SAMPLES, "sample"),
 	DIM(PERF_HPP__PERIOD, "period"),
+	DIM(PERF_HPP__FREQ, "freq"),
+	DIM(PERF_HPP__CPU_UTIL, "cpu_u"),
+	DIM(PERF_HPP__CORE_BUSY, "core_busy"),
 };
 
 #undef DIM
tools/perf/util/symbol.h +11 -1
index 440ba8a..8b21fb4 100644
@@ -80,6 +80,14 @@ static inline size_t symbol__size(const struct symbol *sym)
 struct strlist;
 struct intlist;
 
+enum freq_perf_type_index {
+	DISPLAY_FREQ		= 0,
+	DISPLAY_CPU_UTIL,
+	DISPLAY_CORE_BUSY,
+
+	DISPLAY_MAX
+};
+
 struct symbol_conf {
 	unsigned short	priv_size;
 	unsigned short	nr_events;
@@ -107,7 +115,8 @@ struct symbol_conf {
 			show_hist_headers,
 			branch_callstack,
 			has_filter,
-			show_ref_callgraph;
+			show_ref_callgraph,
+			show_freq_perf;
 	const char	*vmlinux_name,
 			*kallsyms_name,
 			*source_prefix,
@@ -132,6 +141,7 @@ struct symbol_conf {
 	struct intlist	*pid_list,
 			*tid_list;
 	const char	*symfs;
+	u64		freq_perf_type;
 };
 
 extern struct symbol_conf symbol_conf;
Signature · 1 line
1.8.3.1
#12 lore raw
In reply to [email protected]
Quoted 8 lines · On Thu, Sep 03, 2015 at 08:30:56AM -0400, [email protected] wrote:

From: Kan Liang <[email protected]>

This patch set supports per-sample freq/CPU%/CORE_BUSY% print in perf report -D and --stdio.
For printing these information, the perf.data file must have been obtained by group read and using special events cycles, ref-cycles, msr/tsc/, msr/aperf/ or msr/mperf/.

Quoted 8 lines

- Freq (MHz): The frequency during the sample interval. Needs cycles ref-cycles event.
- CPU%: CPU utilization during the sample interval. Needs ref-cycles and msr/tsc/ events.
- CORE_BUSY%: actual percent performance (APERF/MPERF%) during the sample interval. Needs msr/aperf/ and msr/mperf/ events.

Here is an example:

other than my latest 2 comments, I'm ok with the patchset

jirka