The default event list includes the most common events which are widely
used by users. But with -e option, the current perf only counts the events
assigned by -e option. Users may want to collect some extra events with
the default list. For this case, users have to manually add all the events
from the default list. It's inconvenient. Also, users may don't know how to
get the default list.
Currently perf-stat has supported '-d' option which prints more detailed
statistics such as L1-dcache-loads, L1-dcache-load-misses. We just need
one line change to let '-d' support printing the default list.
Before:
root@kbl-ppc:~# ./perf stat -e power/energy-pkg/ -a -- sleep 1
Performance counter stats for 'system wide':
2.04 Joules power/energy-pkg/
1.000863884 seconds time elapsed
After:
root@kbl-ppc:~# ./perf stat -e power/energy-pkg/ -a -d -- sleep 1
Performance counter stats for 'system wide':
2.24 Joules power/energy-pkg/ # 0.000 K/sec
8,009.72 msec cpu-clock # 7.995 CPUs utilized
288 context-switches # 0.036 K/sec
10 cpu-migrations # 0.001 K/sec
2 page-faults # 0.000 K/sec
125,889,836 cycles # 0.016 GHz (49.60%)
23,216,112 instructions # 0.18 insn per cycle (62.40%)
4,571,447 branches # 0.571 M/sec (62.80%)
652,350 branch-misses # 14.27% of all branches (62.85%)
6,564,491 L1-dcache-loads # 0.820 M/sec (62.85%)
1,030,251 L1-dcache-load-misses # 15.69% of all L1-dcache accesses (62.78%)
567,451 LLC-loads # 0.071 M/sec (49.58%)
52,983 LLC-load-misses # 9.34% of all LL-cache accesses (49.53%)
1.001854149 seconds time elapsed
And it can append the metrics to the default event list easily.
root@kbl-ppc:~# ./perf stat -M Page_Walks_Utilization -a -d -- sleep 1
Performance counter stats for 'system wide':
1,332,948 itlb_misses.walk_pending # 0.166 M/sec
# 0.05 Page_Walks_Utilization (30.68%)
1,144,389 dtlb_store_misses.walk_pending # 0.143 M/sec (31.09%)
123,661,801 cycles # 0.015 GHz (31.49%)
8,772,539 dtlb_load_misses.walk_pending # 1.096 M/sec (31.67%)
0 ept.walk_pending # 0.000 K/sec (31.36%)
8,007.56 msec cpu-clock # 7.992 CPUs utilized
270 context-switches # 0.034 K/sec
8 cpu-migrations # 0.001 K/sec
3 page-faults # 0.000 K/sec
124,703,802 cycles # 0.016 GHz (30.95%)
23,239,045 instructions # 0.19 insn per cycle (38.54%)
4,500,430 branches # 0.562 M/sec (38.14%)
658,460 branch-misses # 14.63% of all branches (37.96%)
6,296,710 L1-dcache-loads # 0.786 M/sec (37.96%)
1,027,741 L1-dcache-load-misses # 16.32% of all L1-dcache accesses (37.96%)
551,872 LLC-loads # 0.069 M/sec (30.37%)
48,296 LLC-load-misses # 8.75% of all LL-cache accesses (30.37%)
1.001971812 seconds time elapsed
Signed-off-by: Jin Yao <[email protected]>
---
v3:
Improve '-d' to let it support printing default event list.
v2:
We just use a simple syntax: -e +event to append the -e event
to default list.
v1:
Create a new option '--add-default' to append -e event to
default list.
tools/perf/builtin-stat.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index a380e38ca937..24f9ee90dbf0 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -1812,7 +1812,7 @@ static int add_default_attributes(void)
free(str);
}
- if (!evsel_list->core.nr_entries) {
+ if (!evsel_list->core.nr_entries || detailed_run) {
if (target__has_cpu(&target))
default_attrs0[0].config = PERF_COUNT_SW_CPU_CLOCK;
--
2.17.1