Skip to contentSkip to conversation

perf stat: Disable NMI watchdog message on hybrid

PATCH Applied Jin Yao 3 participants 4 messages Jun 10, 2021 – Jul 7, 2021 perf events

lore
Share & tools
lore

Jin Yao

author
#0 lore raw

PATCH v3 1/1 perf stat: Disable NMI watchdog message on hybrid

If we run a single workload that only runs on big core, there is always a ugly message about disabling the NMI watchdog because the atom is not counted.

Before:

  # ./perf stat true

Performance counter stats for 'true':

                0.43 msec task-clock                #    0.396 CPUs utilized
                   0      context-switches          #    0.000 /sec
                   0      cpu-migrations            #    0.000 /sec
                  45      page-faults               #  103.918 K/sec
             639,634      cpu_core/cycles/          #    1.477 G/sec
       <not counted>      cpu_atom/cycles/                                              (0.00%)
             643,498      cpu_core/instructions/    #    1.486 G/sec
       <not counted>      cpu_atom/instructions/                                        (0.00%)
             123,715      cpu_core/branches/        #  285.694 M/sec
       <not counted>      cpu_atom/branches/                                            (0.00%)
               4,094      cpu_core/branch-misses/   #    9.454 M/sec
       <not counted>      cpu_atom/branch-misses/                                       (0.00%)
         0.001092407 seconds time elapsed

         0.001144000 seconds user
         0.000000000 seconds sys

Some events weren't counted. Try disabling the NMI watchdog:
echo 0 > /proc/sys/kernel/nmi_watchdog perf stat ...
echo 1 > /proc/sys/kernel/nmi_watchdog

  # ./perf stat -e '{cpu_atom/cycles/,msr/tsc/}' true

Performance counter stats for 'true':

       <not counted>      cpu_atom/cycles/                                              (0.00%)
       <not counted>      msr/tsc/                                                      (0.00%)
         0.001904106 seconds time elapsed

         0.001947000 seconds user
         0.000000000 seconds sys

Some events weren't counted. Try disabling the NMI watchdog:
echo 0 > /proc/sys/kernel/nmi_watchdog
perf stat ...
echo 1 > /proc/sys/kernel/nmi_watchdog
The events in group usually have to be from the same PMU. Try reorganizing the group.

Now we disable the NMI watchdog message on hybrid, otherwise there are too many false positives.

After:

  # ./perf stat true

Performance counter stats for 'true':

                0.79 msec task-clock                #    0.419 CPUs utilized
                   0      context-switches          #    0.000 /sec
                   0      cpu-migrations            #    0.000 /sec
                  48      page-faults               #   60.889 K/sec
             777,692      cpu_core/cycles/          #  986.519 M/sec
       <not counted>      cpu_atom/cycles/                                              (0.00%)
             669,147      cpu_core/instructions/    #  848.828 M/sec
       <not counted>      cpu_atom/instructions/                                        (0.00%)
             128,635      cpu_core/branches/        #  163.176 M/sec
       <not counted>      cpu_atom/branches/                                            (0.00%)
               4,089      cpu_core/branch-misses/   #    5.187 M/sec
       <not counted>      cpu_atom/branch-misses/                                       (0.00%)
         0.001880649 seconds time elapsed

         0.001935000 seconds user
         0.000000000 seconds sys
  # ./perf stat -e '{cpu_atom/cycles/,msr/tsc/}' true

Performance counter stats for 'true':

       <not counted>      cpu_atom/cycles/                                              (0.00%)
       <not counted>      msr/tsc/                                                      (0.00%)
         0.000963319 seconds time elapsed

         0.000999000 seconds user
         0.000000000 seconds sys

---
v3:
- Use evlist__has_hybrid() to check there is at least one hybrid event in evlist.

v2:
- If the group was mixed with hybrid event and non-hybrid event, the NMI watchdog message was still reported. V2 adds checking for hybrid event mixed group.

 tools/perf/util/stat-display.c | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)
tools/perf/util/stat-display.c +6 -3
index b759dfd633b4..55ae0d1705a9 100644
@@ -19,6 +19,7 @@
 #include "util.h"
 #include "iostat.h"
 #include "pmu-hybrid.h"
+#include "evlist-hybrid.h"
 
 #define CNTR_NOT_SUPPORTED	"<not supported>"
 #define CNTR_NOT_COUNTED	"<not counted>"
@@ -465,9 +466,11 @@ static void printout(struct perf_stat_config *config, struct aggr_cpu_id id, int
 			config->csv_sep);
 
 		if (counter->supported) {
-			config->print_free_counters_hint = 1;
-			if (is_mixed_hw_group(counter))
-				config->print_mixed_hw_group_error = 1;
+			if (!evlist__has_hybrid(counter->evlist)) {
+				config->print_free_counters_hint = 1;
+				if (is_mixed_hw_group(counter))
+					config->print_mixed_hw_group_error = 1;
+			}
 		}
 
 		fprintf(config->output, "%-*s%s",
Signature · 1 line
2.17.1
author
#1 lore raw

Hi Jiri,

Quoted 122 lines · Jin Yao

If we run a single workload that only runs on big core, there is always a ugly message about disabling the NMI watchdog because the atom is not counted.

Before:

   # ./perf stat true

Performance counter stats for 'true':

0.43 msec task-clock # 0.396 CPUs utilized
0 context-switches # 0.000 /sec
0 cpu-migrations # 0.000 /sec
45 page-faults # 103.918 K/sec
639,634 cpu_core/cycles/ # 1.477 G/sec
<not counted> cpu_atom/cycles/ (0.00%)
643,498 cpu_core/instructions/ # 1.486 G/sec
<not counted> cpu_atom/instructions/ (0.00%)
123,715 cpu_core/branches/ # 285.694 M/sec
<not counted> cpu_atom/branches/ (0.00%)
4,094 cpu_core/branch-misses/ # 9.454 M/sec
<not counted> cpu_atom/branch-misses/ (0.00%)

0.001092407 seconds time elapsed

0.001144000 seconds user
0.000000000 seconds sys

Some events weren't counted. Try disabling the NMI watchdog:
echo 0 > /proc/sys/kernel/nmi_watchdog perf stat ...
echo 1 > /proc/sys/kernel/nmi_watchdog

   # ./perf stat -e '{cpu_atom/cycles/,msr/tsc/}' true

Performance counter stats for 'true':

<not counted> cpu_atom/cycles/ (0.00%)
<not counted> msr/tsc/ (0.00%)

0.001904106 seconds time elapsed

0.001947000 seconds user
0.000000000 seconds sys

Some events weren't counted. Try disabling the NMI watchdog:
echo 0 > /proc/sys/kernel/nmi_watchdog
perf stat ...
echo 1 > /proc/sys/kernel/nmi_watchdog
The events in group usually have to be from the same PMU. Try reorganizing the group.

Now we disable the NMI watchdog message on hybrid, otherwise there are too many false positives.

After:

   # ./perf stat true

Performance counter stats for 'true':

0.79 msec task-clock # 0.419 CPUs utilized
0 context-switches # 0.000 /sec
0 cpu-migrations # 0.000 /sec
48 page-faults # 60.889 K/sec
777,692 cpu_core/cycles/ # 986.519 M/sec
<not counted> cpu_atom/cycles/ (0.00%)
669,147 cpu_core/instructions/ # 848.828 M/sec
<not counted> cpu_atom/instructions/ (0.00%)
128,635 cpu_core/branches/ # 163.176 M/sec
<not counted> cpu_atom/branches/ (0.00%)
4,089 cpu_core/branch-misses/ # 5.187 M/sec
<not counted> cpu_atom/branch-misses/ (0.00%)

0.001880649 seconds time elapsed

0.001935000 seconds user
0.000000000 seconds sys

   # ./perf stat -e '{cpu_atom/cycles/,msr/tsc/}' true

Performance counter stats for 'true':

<not counted> cpu_atom/cycles/ (0.00%)
<not counted> msr/tsc/ (0.00%)

0.000963319 seconds time elapsed

0.000999000 seconds user
0.000000000 seconds sys

---
v3:
- Use evlist__has_hybrid() to check there is at least one hybrid event in evlist.

v2:
- If the group was mixed with hybrid event and non-hybrid event, the NMI watchdog message was still reported. V2 adds checking for hybrid event mixed group.

  tools/perf/util/stat-display.c | 9 ++++++---
  1 file changed, 6 insertions(+), 3 deletions(-)
tools/perf/util/stat-display.c +1 -1
index b759dfd633b4..55ae0d1705a9 100644
@@ -19,6 +19,7 @@
  #include "util.h"
  #include "iostat.h"
  #include "pmu-hybrid.h"
+#include "evlist-hybrid.h"
 
  #define CNTR_NOT_SUPPORTED	"<not supported>"
  #define CNTR_NOT_COUNTED	"<not counted>"
@@ -465,9 +466,11 @@ static void printout(struct perf_stat_config *config, struct aggr_cpu_id id, int
  			config->csv_sep);
 
  		if (counter->supported) {
-			config->print_free_counters_hint = 1;

tools/perf/util/stat-display.c static void printout(struct perf_stat_config *config, struct aggr_cpu_id id, int · line 476

Quoted 10 lines
tools/perf/util/stat-display.c +5 -2
-			if (is_mixed_hw_group(counter))
-				config->print_mixed_hw_group_error = 1;
+			if (!evlist__has_hybrid(counter->evlist)) {
+				config->print_free_counters_hint = 1;
+				if (is_mixed_hw_group(counter))
+					config->print_mixed_hw_group_error = 1;
+			}
  		}
 
  		fprintf(config->output, "%-*s%s",

Any comments for v3?

Thanks
Jin Yao

Acked-by
#2 lore raw
Quoted 103 lines · Yao

Hi Jiri,

Quoted 100 lines · Jin Yao

If we run a single workload that only runs on big core, there is always a ugly message about disabling the NMI watchdog because the atom is not counted.

Before:

   # ./perf stat true

Performance counter stats for 'true':

0.43 msec task-clock # 0.396 CPUs utilized
0 context-switches # 0.000 /sec
0 cpu-migrations # 0.000 /sec
45 page-faults # 103.918 K/sec
639,634 cpu_core/cycles/ # 1.477 G/sec
<not counted> cpu_atom/cycles/ (0.00%)
643,498 cpu_core/instructions/ # 1.486 G/sec
<not counted> cpu_atom/instructions/ (0.00%)
123,715 cpu_core/branches/ # 285.694 M/sec
<not counted> cpu_atom/branches/ (0.00%)
4,094 cpu_core/branch-misses/ # 9.454 M/sec
<not counted> cpu_atom/branch-misses/ (0.00%)

0.001092407 seconds time elapsed

0.001144000 seconds user
0.000000000 seconds sys

Some events weren't counted. Try disabling the NMI watchdog:
echo 0 > /proc/sys/kernel/nmi_watchdog perf stat ...
echo 1 > /proc/sys/kernel/nmi_watchdog

   # ./perf stat -e '{cpu_atom/cycles/,msr/tsc/}' true

Performance counter stats for 'true':

<not counted> cpu_atom/cycles/ (0.00%)
<not counted> msr/tsc/ (0.00%)

0.001904106 seconds time elapsed

0.001947000 seconds user
0.000000000 seconds sys

Some events weren't counted. Try disabling the NMI watchdog:
echo 0 > /proc/sys/kernel/nmi_watchdog
perf stat ...
echo 1 > /proc/sys/kernel/nmi_watchdog
The events in group usually have to be from the same PMU. Try reorganizing the group.

Now we disable the NMI watchdog message on hybrid, otherwise there are too many false positives.

After:

   # ./perf stat true

Performance counter stats for 'true':

0.79 msec task-clock # 0.419 CPUs utilized
0 context-switches # 0.000 /sec
0 cpu-migrations # 0.000 /sec
48 page-faults # 60.889 K/sec
777,692 cpu_core/cycles/ # 986.519 M/sec
<not counted> cpu_atom/cycles/ (0.00%)
669,147 cpu_core/instructions/ # 848.828 M/sec
<not counted> cpu_atom/instructions/ (0.00%)
128,635 cpu_core/branches/ # 163.176 M/sec
<not counted> cpu_atom/branches/ (0.00%)
4,089 cpu_core/branch-misses/ # 5.187 M/sec
<not counted> cpu_atom/branch-misses/ (0.00%)

0.001880649 seconds time elapsed

0.001935000 seconds user
0.000000000 seconds sys

   # ./perf stat -e '{cpu_atom/cycles/,msr/tsc/}' true

Performance counter stats for 'true':

<not counted> cpu_atom/cycles/ (0.00%)
<not counted> msr/tsc/ (0.00%)

0.000963319 seconds time elapsed

0.000999000 seconds user
0.000000000 seconds sys

---
v3:
- Use evlist__has_hybrid() to check there is at least one hybrid event in evlist.

v2:
- If the group was mixed with hybrid event and non-hybrid event, the NMI watchdog message was still reported. V2 adds checking for hybrid event mixed group.

thanks,
jirka

Quoted 36 lines
Quoted 19 lines
  tools/perf/util/stat-display.c | 9 ++++++---
  1 file changed, 6 insertions(+), 3 deletions(-)
tools/perf/util/stat-display.c +1 -1
index b759dfd633b4..55ae0d1705a9 100644
@@ -19,6 +19,7 @@
  #include "util.h"
  #include "iostat.h"
  #include "pmu-hybrid.h"
+#include "evlist-hybrid.h"
  #define CNTR_NOT_SUPPORTED	"<not supported>"
  #define CNTR_NOT_COUNTED	"<not counted>"
@@ -465,9 +466,11 @@ static void printout(struct perf_stat_config *config, struct aggr_cpu_id id, int
  			config->csv_sep);
  		if (counter->supported) {
-			config->print_free_counters_hint = 1;

tools/perf/util/stat-display.c static void printout(struct perf_stat_config *config, struct aggr_cpu_id id, int · line 475

Quoted 10 lines
tools/perf/util/stat-display.c +5 -2
-			if (is_mixed_hw_group(counter))
-				config->print_mixed_hw_group_error = 1;
+			if (!evlist__has_hybrid(counter->evlist)) {
+				config->print_free_counters_hint = 1;
+				if (is_mixed_hw_group(counter))
+					config->print_mixed_hw_group_error = 1;
+			}
  		}
  		fprintf(config->output, "%-*s%s",
 

Any comments for v3?

Thanks
Jin Yao

maintainer Applied
#3 lore raw

Em Tue, Jul 06, 2021 at 09:54:15PM +0200, Jiri Olsa escreveu:

Quoted 11 lines · Yao
Quoted 8 lines
Quoted 8 lines

v3:
- Use evlist__has_hybrid() to check there is at least one hybrid event in evlist.

v2:
- If the group was mixed with hybrid event and non-hybrid event, the NMI watchdog message was still reported. V2 adds checking for hybrid event mixed group.

Thanks, applied.

- Arnaldo