Skip to contentSkip to conversation

perf report: Fix sample number stats for branch entry mode

PATCH Superseded Thomas Falcon 1 participant 2 messages Feb 19, 2025 – Feb 20, 2025 perf events

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 perf report: Fix sample number stats for branch entry mode

Currently, stats->nr_samples is incremented per entry in the sample's branch stack instead of per sample taken. As a result, statistics of samples taken during perf record in --branch-filter mode do not seem correct. Call hists__inc_nr_samples() only for each sample taken.

Before:

$ ./perf record -e cycles:u -b -c 10000000000 ./tchain_edit
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.005 MB perf.data (2 samples) ]
$ perf report -D | tail -n 16
Aggregated stats:
               TOTAL events:         16
                COMM events:          2  (12.5%)
                EXIT events:          1  ( 6.2%)
              SAMPLE events:          2  (12.5%)
               MMAP2 events:          2  (12.5%)
             KSYMBOL events:          1  ( 6.2%)
      FINISHED_ROUND events:          1  ( 6.2%)
            ID_INDEX events:          1  ( 6.2%)
          THREAD_MAP events:          1  ( 6.2%)
             CPU_MAP events:          1  ( 6.2%)
        EVENT_UPDATE events:          2  (12.5%)
           TIME_CONV events:          1  ( 6.2%)
       FINISHED_INIT events:          1  ( 6.2%)
cpu_core/cycles/u stats:
              SAMPLE events:         64

After:

$ ./perf report -D | tail -n 16
Aggregated stats:
               TOTAL events:         16
                COMM events:          2  (12.5%)
                EXIT events:          1  ( 6.2%)
              SAMPLE events:          2  (12.5%)
               MMAP2 events:          2  (12.5%)
             KSYMBOL events:          1  ( 6.2%)
      FINISHED_ROUND events:          1  ( 6.2%)
            ID_INDEX events:          1  ( 6.2%)
          THREAD_MAP events:          1  ( 6.2%)
             CPU_MAP events:          1  ( 6.2%)
        EVENT_UPDATE events:          2  (12.5%)
           TIME_CONV events:          1  ( 6.2%)
       FINISHED_INIT events:          1  ( 6.2%)
cpu_core/cycles/u stats:
              SAMPLE events:          2
---
 tools/perf/util/hist.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)
tools/perf/util/hist.c +5 -2
index 0f30f843c566..4ae56a2a4d34 100644
@@ -975,8 +975,6 @@ iter_add_next_branch_entry(struct hist_entry_iter *iter, struct addr_location *a
 	if (he == NULL)
 		return -ENOMEM;
 
-	hists__inc_nr_samples(hists, he->filtered);
-
 out:
 	iter->he = he;
 	iter->curr++;
@@ -995,9 +993,14 @@ static int
 iter_finish_branch_entry(struct hist_entry_iter *iter,
 			 struct addr_location *al __maybe_unused)
 {
+	struct evsel *evsel = iter->evsel;
+	struct hists *hists = evsel__hists(evsel);
+
 	for (int i = 0; i < iter->total; i++)
 		branch_info__exit(&iter->bi[i]);
 
+	hists__inc_nr_samples(hists, iter->he->filtered);
+
 	zfree(&iter->bi);
 	iter->he = NULL;
 
Signature · 1 line
2.48.1
#1 lore raw
Quoted 77 lines · Thomas Falcon

Currently, stats->nr_samples is incremented per entry in the sample's branch stack instead of per sample taken. As a result, statistics of samples taken during perf record in --branch-filter mode do not seem correct. Call hists__inc_nr_samples() only for each sample taken.

Before:

$ ./perf record -e cycles:u -b -c 10000000000 ./tchain_edit
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.005 MB perf.data (2 samples) ]
$ perf report -D | tail -n 16
Aggregated stats:
               TOTAL events:         16
                COMM events:          2  (12.5%)
                EXIT events:          1  ( 6.2%)
              SAMPLE events:          2  (12.5%)
               MMAP2 events:          2  (12.5%)
             KSYMBOL events:          1  ( 6.2%)
      FINISHED_ROUND events:          1  ( 6.2%)
            ID_INDEX events:          1  ( 6.2%)
          THREAD_MAP events:          1  ( 6.2%)
             CPU_MAP events:          1  ( 6.2%)
        EVENT_UPDATE events:          2  (12.5%)
           TIME_CONV events:          1  ( 6.2%)
       FINISHED_INIT events:          1  ( 6.2%)
cpu_core/cycles/u stats:
              SAMPLE events:         64

After:

$ ./perf report -D | tail -n 16
Aggregated stats:
               TOTAL events:         16
                COMM events:          2  (12.5%)
                EXIT events:          1  ( 6.2%)
              SAMPLE events:          2  (12.5%)
               MMAP2 events:          2  (12.5%)
             KSYMBOL events:          1  ( 6.2%)
      FINISHED_ROUND events:          1  ( 6.2%)
            ID_INDEX events:          1  ( 6.2%)
          THREAD_MAP events:          1  ( 6.2%)
             CPU_MAP events:          1  ( 6.2%)
        EVENT_UPDATE events:          2  (12.5%)
           TIME_CONV events:          1  ( 6.2%)
       FINISHED_INIT events:          1  ( 6.2%)
cpu_core/cycles/u stats:
              SAMPLE events:          2

---
 tools/perf/util/hist.c | 7 +++++--

 1 file changed, 5 insertions(+), 2 deletions(-)
tools/perf/util/hist.c +0 -0
index 0f30f843c566..4ae56a2a4d34 100644
@@ -975,8 +975,6 @@ iter_add_next_branch_entry(struct hist_entry_iter
 *iter, struct addr_location *a
  	if (he == NULL)
  		return -ENOMEM;

tools/perf/util/hist.c +0 -2
-	hists__inc_nr_samples(hists, he->filtered);
-

 out:
  iter->he = he;
  iter->curr++;

tools/perf/util/hist.c +3 -0
@@ -995,9 +993,14 @@ static int
  iter_finish_branch_entry(struct hist_entry_iter *iter,
  			 struct addr_location *al __maybe_unused)
  {
+	struct evsel *evsel = iter->evsel;
+	struct hists *hists = evsel__hists(evsel);
+
  	for (int i = 0; i < iter->total; i++)
  		branch_info__exit(&iter->bi[i]);

tools/perf/util/hist.c static int · line 1003

Quoted 2 lines
tools/perf/util/hist.c +2 -0
+	hists__inc_nr_samples(hists, iter->he->filtered);
+

Sorry, I missed a null check for iter->he here I think, will send a v2.

Tom

Quoted 3 lines

  zfree(&iter->bi);
  iter->he = NULL;