Skip to contentSkip to conversation

perf stat: enable 1ms interval for printing event counters values

PATCH Applied Alexey Budankov 4 participants 4 messages Apr 3, 2018 – Apr 16, 2018 perf events

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 perf stat: enable 1ms interval for printing event counters values

Currently print count interval for performance counters values is limited by 10ms so reading the values at frequencies higher than 100Hz is restricted by the tool.

This change makes perf stat -I possible on frequencies up to 1KHz and, to some extent, makes perf stat -I to be on-par with perf record sampling profiling.

When running perf stat -I for monitoring e.g. PCIe uncore counters and at the same time profiling some I/O workload by perf record e.g. for cpu-cycles and context switches, it is then possible to observe consolidated CPU/OS/IO(Uncore) performance picture for that workload.

Tool overhead warning printed when specifying -v option can be missed due to screen scrolling in case you have output to the console so message is moved into help available by running perf stat -h.

---

Changes in v2:
- updated minimum value to 1ms at perf-stat.txt manual

 tools/perf/Documentation/perf-stat.txt |  2 +-
 tools/perf/builtin-stat.c              | 14 ++------------
 2 files changed, 3 insertions(+), 13 deletions(-)
tools/perf/Documentation/perf-stat.txt +1 -1
index f15b306be183..e6c3b4e555c2 100644
@@ -153,7 +153,7 @@ perf stat --repeat 10 --null --sync --pre 'make -s O=defconfig-build/clean' -- m
 
 -I msecs::
 --interval-print msecs::
-Print count deltas every N milliseconds (minimum: 10ms)
+Print count deltas every N milliseconds (minimum: 1ms)
 The overhead percentage could be high in some cases, for instance with small, sub 100ms intervals.  Use with caution.
 	example: 'perf stat -I 1000 -e cycles -a sleep 5'
 
tools/perf/builtin-stat.c +2 -12
index f5c454855908..147a27e8c937 100644
@@ -1943,7 +1943,8 @@ static const struct option stat_options[] = {
 	OPT_STRING(0, "post", &post_cmd, "command",
 			"command to run after to the measured command"),
 	OPT_UINTEGER('I', "interval-print", &stat_config.interval,
-		    "print counts at regular interval in ms (>= 10)"),
+		    "print counts at regular interval in ms "
+		    "(overhead is possible for values <= 100ms)"),
 	OPT_INTEGER(0, "interval-count", &stat_config.times,
 		    "print counts for fixed number of times"),
 	OPT_UINTEGER(0, "timeout", &stat_config.timeout,
@@ -2923,17 +2924,6 @@ int cmd_stat(int argc, const char **argv)
 		}
 	}
 
-	if (interval && interval < 100) {
-		if (interval < 10) {
-			pr_err("print interval must be >= 10ms\n");
-			parse_options_usage(stat_usage, stat_options, "I", 1);
-			goto out;
-		} else
-			pr_warning("print interval < 100ms. "
-				   "The overhead percentage could be high in some cases. "
-				   "Please proceed with caution.\n");
-	}
-
 	if (stat_config.times && interval)
 		interval_count = true;
 	else if (stat_config.times && !interval) {
Acked-by on v2 1/1
#1 lore raw
Quoted 18 lines · Alexey Budankov

Currently print count interval for performance counters values is limited by 10ms so reading the values at frequencies higher than 100Hz is restricted by the tool.

This change makes perf stat -I possible on frequencies up to 1KHz and, to some extent, makes perf stat -I to be on-par with perf record sampling profiling.

When running perf stat -I for monitoring e.g. PCIe uncore counters and at the same time profiling some I/O workload by perf record e.g. for cpu-cycles and context switches, it is then possible to observe consolidated CPU/OS/IO(Uncore) performance picture for that workload.

Tool overhead warning printed when specifying -v option can be missed due to screen scrolling in case you have output to the console so message is moved into help available by running perf stat -h.

Quoted 5 lines

---

Changes in v2:
- updated minimum value to 1ms at perf-stat.txt manual

thanks,
jirka

maintainer Applied
#2 lore raw

Em Wed, Apr 04, 2018 at 09:05:16AM +0200, Jiri Olsa escreveu:

Quoted 19 lines · Alexey Budankov
Quoted 18 lines

Currently print count interval for performance counters values is limited by 10ms so reading the values at frequencies higher than 100Hz is restricted by the tool.

This change makes perf stat -I possible on frequencies up to 1KHz and, to some extent, makes perf stat -I to be on-par with perf record sampling profiling.

When running perf stat -I for monitoring e.g. PCIe uncore counters and at the same time profiling some I/O workload by perf record e.g. for cpu-cycles and context switches, it is then possible to observe consolidated CPU/OS/IO(Uncore) performance picture for that workload.

Tool overhead warning printed when specifying -v option can be missed due to screen scrolling in case you have output to the console so message is moved into help available by running perf stat -h.

Quoted 7 lines
Quoted 5 lines

---

Changes in v2:
- updated minimum value to 1ms at perf-stat.txt manual

Thanks, applied to perf/core.

- Arnaldo

#3 lore raw

[tip:perf/urgent] perf stat: Enable 1ms interval for printing event counters values

In reply to Alexey Budankov
perf stat: Enable 1ms interval for printing event counters values
Commit-ID:  9dc9a95f03a69ab926d9ff1986ab2087f34a5dce
Gitweb:     https://git.kernel.org/tip/9dc9a95f03a69ab926d9ff1986ab2087f34a5dce
Author:     Alexey Budankov <[email protected]>
AuthorDate: Tue, 3 Apr 2018 21:18:33 +0300
Committer:  Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Thu, 12 Apr 2018 09:29:31 -0300

perf stat: Enable 1ms interval for printing event counters values

Currently print count interval for performance counters values is limited by 10ms so reading the values at frequencies higher than 100Hz is restricted by the tool.

This change makes perf stat -I possible on frequencies up to 1KHz and, to some extent, makes perf stat -I to be on-par with perf record sampling profiling.

When running perf stat -I for monitoring e.g. PCIe uncore counters and at the same time profiling some I/O workload by perf record e.g. for cpu-cycles and context switches, it is then possible to observe consolidated CPU/OS/IO(Uncore) performance picture for that workload.

Tool overhead warning printed when specifying -v option can be missed due to screen scrolling in case you have output to the console so message is moved into help available by running perf stat -h.

---
 tools/perf/Documentation/perf-stat.txt |  2 +-
 tools/perf/builtin-stat.c              | 14 ++------------
 2 files changed, 3 insertions(+), 13 deletions(-)
tools/perf/Documentation/perf-stat.txt +1 -1
index f15b306be183..e6c3b4e555c2 100644
@@ -153,7 +153,7 @@ perf stat --repeat 10 --null --sync --pre 'make -s O=defconfig-build/clean' -- m
 
 -I msecs::
 --interval-print msecs::
-Print count deltas every N milliseconds (minimum: 10ms)
+Print count deltas every N milliseconds (minimum: 1ms)
 The overhead percentage could be high in some cases, for instance with small, sub 100ms intervals.  Use with caution.
 	example: 'perf stat -I 1000 -e cycles -a sleep 5'
 
tools/perf/builtin-stat.c +2 -12
index f5c454855908..147a27e8c937 100644
@@ -1943,7 +1943,8 @@ static const struct option stat_options[] = {
 	OPT_STRING(0, "post", &post_cmd, "command",
 			"command to run after to the measured command"),
 	OPT_UINTEGER('I', "interval-print", &stat_config.interval,
-		    "print counts at regular interval in ms (>= 10)"),
+		    "print counts at regular interval in ms "
+		    "(overhead is possible for values <= 100ms)"),
 	OPT_INTEGER(0, "interval-count", &stat_config.times,
 		    "print counts for fixed number of times"),
 	OPT_UINTEGER(0, "timeout", &stat_config.timeout,
@@ -2923,17 +2924,6 @@ int cmd_stat(int argc, const char **argv)
 		}
 	}
 
-	if (interval && interval < 100) {
-		if (interval < 10) {
-			pr_err("print interval must be >= 10ms\n");
-			parse_options_usage(stat_usage, stat_options, "I", 1);
-			goto out;
-		} else
-			pr_warning("print interval < 100ms. "
-				   "The overhead percentage could be high in some cases. "
-				   "Please proceed with caution.\n");
-	}
-
 	if (stat_config.times && interval)
 		interval_count = true;
 	else if (stat_config.times && !interval) {