Skip to contentSkip to conversation

perf report: Fix incorrectly added dimensions as switch perf data file

PATCH Applied Jin Yao 4 participants 10 messages Dec 20, 2019 – Jan 20, 2020 perf events

lore
Share & tools
lore

Jin Yao

author
#0 lore raw

PATCH v6 1/4 perf report: Fix incorrectly added dimensions as switch perf data file

We observed an issue that was some extra columns displayed after switching perf data file in browser. The steps to reproduce:

1. perf record -a -e cycles,instructions -- sleep 3
2. perf report --group
3. In browser, we use hotkey 's' to switch to another perf.data
4. Now in browser, the extra columns 'Self' and 'Children' are displayed.

The issue is setup_sorting() executed again after repeat path, so dimensions are added again.

This patch checks the last key returned from __cmd_report(). If it's K_SWITCH_INPUT_DATA, skips the setup_sorting().

v6:
---
No change.

v5:
---
New patch in v5.

---
 tools/perf/builtin-report.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
tools/perf/builtin-report.c +4 -1
index 387311c67264..de988589d99b 100644
@@ -1076,6 +1076,7 @@ int cmd_report(int argc, const char **argv)
 	struct stat st;
 	bool has_br_stack = false;
 	int branch_mode = -1;
+	int last_key = 0;
 	bool branch_call_mode = false;
 #define CALLCHAIN_DEFAULT_OPT  "graph,0.5,caller,function,percent"
 	static const char report_callchain_help[] = "Display call graph (stack chain/backtrace):\n\n"
@@ -1450,7 +1451,8 @@ int cmd_report(int argc, const char **argv)
 		sort_order = sort_tmp;
 	}
 
-	if (setup_sorting(session->evlist) < 0) {
+	if ((last_key != K_SWITCH_INPUT_DATA) &&
+	    (setup_sorting(session->evlist) < 0)) {
 		if (sort_order)
 			parse_options_usage(report_usage, options, "s", 1);
 		if (field_order)
@@ -1530,6 +1532,7 @@ int cmd_report(int argc, const char **argv)
 	ret = __cmd_report(&report);
 	if (ret == K_SWITCH_INPUT_DATA) {
 		perf_session__delete(session);
+		last_key = K_SWITCH_INPUT_DATA;
 		goto repeat;
 	} else
 		ret = 0;
Signature · 1 line
2.17.1
author
#1 lore raw

PATCH v6 2/4 perf report: Change sort order by a specified event in group

When performing "perf report --group", it shows the event group information together. By default, the output is sorted by the first event in group.

It would be nice for user to select any event for sorting. This patch introduces a new option "--group-sort-idx" to sort the output by the event at the index n in event group.

For example,

Before:

  # perf report --group --stdio

# To display the perf.data header info, please use --header/--header-only options.
#
#
# Total Lost Samples: 0
#
# Samples: 12K of events 'cpu/instructions,period=2000003/, cpu/cpu-cycles,period=200003/, BR_MISP_RETIRED.ALL_BRANCHES:pp, cpu/event=0xc0,umask=1,cmask=1, # Event count (approx.): 6451235635
#

  #                         Overhead  Command    Shared Object            Symbol
  # ................................  .........  .......................  ...................................
  #
      92.19%  98.68%   0.00%  93.30%  mgen       mgen                     [.] LOOP1
       3.12%   0.29%   0.00%   0.16%  gsd-color  libglib-2.0.so.0.5600.4  [.] 0x0000000000049515
       1.56%   0.03%   0.00%   0.04%  gsd-color  libglib-2.0.so.0.5600.4  [.] 0x00000000000494b7
       1.56%   0.01%   0.00%   0.00%  gsd-color  libglib-2.0.so.0.5600.4  [.] 0x00000000000494ce
       1.56%   0.00%   0.00%   0.00%  mgen       [kernel.kallsyms]        [k] task_tick_fair
       0.00%   0.15%   0.00%   0.04%  perf       [kernel.kallsyms]        [k] smp_call_function_single
       0.00%   0.13%   0.00%   6.08%  swapper    [kernel.kallsyms]        [k] intel_idle
       0.00%   0.03%   0.00%   0.00%  gsd-color  libglib-2.0.so.0.5600.4  [.] g_main_context_check
       0.00%   0.03%   0.00%   0.00%  swapper    [kernel.kallsyms]        [k] apic_timer_interrupt
       ...

After:

  # perf report --group --stdio --group-sort-idx 3

# To display the perf.data header info, please use --header/--header-only options.
#
#
# Total Lost Samples: 0
#
# Samples: 12K of events 'cpu/instructions,period=2000003/, cpu/cpu-cycles,period=200003/, BR_MISP_RETIRED.ALL_BRANCHES:pp, cpu/event=0xc0,umask=1,cmask=1, # Event count (approx.): 6451235635
#

  #                         Overhead  Command    Shared Object            Symbol
  # ................................  .........  .......................  ...................................
  #
      92.19%  98.68%   0.00%  93.30%  mgen       mgen                     [.] LOOP1
       0.00%   0.13%   0.00%   6.08%  swapper    [kernel.kallsyms]        [k] intel_idle
       3.12%   0.29%   0.00%   0.16%  gsd-color  libglib-2.0.so.0.5600.4  [.] 0x0000000000049515
       0.00%   0.00%   0.00%   0.06%  swapper    [kernel.kallsyms]        [k] hrtimer_start_range_ns
       1.56%   0.03%   0.00%   0.04%  gsd-color  libglib-2.0.so.0.5600.4  [.] 0x00000000000494b7
       0.00%   0.15%   0.00%   0.04%  perf       [kernel.kallsyms]        [k] smp_call_function_single
       0.00%   0.00%   0.00%   0.02%  mgen       [kernel.kallsyms]        [k] update_curr
       0.00%   0.00%   0.00%   0.02%  mgen       [kernel.kallsyms]        [k] apic_timer_interrupt
       0.00%   0.00%   0.00%   0.02%  mgen       [kernel.kallsyms]        [k] native_apic_msr_eoi_write
       0.00%   0.00%   0.00%   0.02%  mgen       [kernel.kallsyms]        [k] __update_load_avg_se
       0.00%   0.00%   0.00%   0.02%  mgen       [kernel.kallsyms]        [k] scheduler_tick

Now the output is sorted by the fourth event in group.

v6:
---
No change.

v5:
---
No change in v5.
v4 has been ACKed by Jiri.

v4:
---
1. Update Documentation/perf-report.txt to mention '--group-sort-idx' support multiple groups with different amount of events and it should be used on grouped events.

2. Update __hpp__group_sort_idx(), just return when the idx is out of limit.

3. Return failure on symbol_conf.group_sort_idx && !session->evlist->nr_groups. So now we don't need to use together with --group.

v3:
---
Refine the code in __hpp__group_sort_idx().

Before:

   for (i = 1; i < nr_members; i++) {
        if (i == idx) {
                ret = field_cmp(fields_a[i], fields_b[i]);
                if (ret)
                        goto out;
        }

}

After:

   if (idx >= 1 && idx < nr_members) {
        ret = field_cmp(fields_a[idx], fields_b[idx]);
        if (ret)
                goto out;

}

v2:
---
No change

---
 tools/perf/Documentation/perf-report.txt |   5 ++
 tools/perf/builtin-report.c              |  10 +++
 tools/perf/ui/hist.c                     | 104 +++++++++++++++++++----
 tools/perf/util/symbol_conf.h            |   1 +
 4 files changed, 105 insertions(+), 15 deletions(-)
tools/perf/Documentation/perf-report.txt +5 -0
index 8dbe2119686a..fbb715bcb512 100644
@@ -371,6 +371,11 @@ OPTIONS
 	Show event group information together. It forces group output also
 	if there are no groups defined in data file.
 
+--group-sort-idx::
+	Sort the output by the event at the index n in group. If n is invalid,
+	sort by the first event. It can support multiple groups with different
+	amount of events. WARNING: This should be used on grouped events.
+
 --demangle::
 	Demangle symbol names to human readable form. It's enabled by default,
 	disable with --no-demangle.
tools/perf/builtin-report.c +10 -0
index de988589d99b..4c80a3ba73c3 100644
@@ -1211,6 +1211,10 @@ int cmd_report(int argc, const char **argv)
 		    "Show a column with the sum of periods"),
 	OPT_BOOLEAN_SET(0, "group", &symbol_conf.event_group, &report.group_set,
 		    "Show event group information together"),
+	OPT_INTEGER(0, "group-sort-idx", &symbol_conf.group_sort_idx,
+		    "Sort the output by the event at the index n in group. "
+		    "If n is invalid, sort by the first event. "
+		    "WARNING: should be used on grouped events."),
 	OPT_CALLBACK_NOOPT('b', "branch-stack", &branch_mode, "",
 		    "use branch records for per branch histogram filling",
 		    parse_branch_mode),
@@ -1350,6 +1354,12 @@ int cmd_report(int argc, const char **argv)
 
 	setup_forced_leader(&report, session->evlist);
 
+	if (symbol_conf.group_sort_idx && !session->evlist->nr_groups) {
+		parse_options_usage(NULL, options, "group-sort-idx", 0);
+		ret = -EINVAL;
+		goto error;
+	}
+
 	if (itrace_synth_opts.last_branch)
 		has_br_stack = true;
 
tools/perf/ui/hist.c +89 -15

126 lines not on this page · raw message

tools/perf/util/symbol_conf.h +1 -0
index 10f1ec3e0349..b916afb95ec5 100644
@@ -73,6 +73,7 @@ struct symbol_conf {
 	const char	*symfs;
 	int		res_sample;
 	int		pad_output_len_dso;
+	int		group_sort_idx;
 };
 
 extern struct symbol_conf symbol_conf;
Signature · 1 line
2.17.1
author
#2 lore raw

Re: [PATCH v6 2/4] perf report: Change sort order by a specified event in group

Hi Arnaldo,

I see the patch 1/4 has been accpeted. Can the remaining 3 patches be picked up? :)

Thanks
Jin Yao

Quoted 306 lines · Jin Yao

When performing "perf report --group", it shows the event group information together. By default, the output is sorted by the first event in group.

It would be nice for user to select any event for sorting. This patch introduces a new option "--group-sort-idx" to sort the output by the event at the index n in event group.

For example,

Before:

   # perf report --group --stdio
   # To display the perf.data header info, please use --header/--header-only options.
   #
   #
   # Total Lost Samples: 0
   #
   # Samples: 12K of events 'cpu/instructions,period=2000003/, cpu/cpu-cycles,period=200003/, BR_MISP_RETIRED.ALL_BRANCHES:pp, cpu/event=0xc0,umask=1,cmask=1,
   # Event count (approx.): 6451235635
   #
   #                         Overhead  Command    Shared Object            Symbol
   # ................................  .........  .......................  ...................................
   #
       92.19%  98.68%   0.00%  93.30%  mgen       mgen                     [.] LOOP1
        3.12%   0.29%   0.00%   0.16%  gsd-color  libglib-2.0.so.0.5600.4  [.] 0x0000000000049515
        1.56%   0.03%   0.00%   0.04%  gsd-color  libglib-2.0.so.0.5600.4  [.] 0x00000000000494b7
        1.56%   0.01%   0.00%   0.00%  gsd-color  libglib-2.0.so.0.5600.4  [.] 0x00000000000494ce
        1.56%   0.00%   0.00%   0.00%  mgen       [kernel.kallsyms]        [k] task_tick_fair
        0.00%   0.15%   0.00%   0.04%  perf       [kernel.kallsyms]        [k] smp_call_function_single
        0.00%   0.13%   0.00%   6.08%  swapper    [kernel.kallsyms]        [k] intel_idle
        0.00%   0.03%   0.00%   0.00%  gsd-color  libglib-2.0.so.0.5600.4  [.] g_main_context_check
        0.00%   0.03%   0.00%   0.00%  swapper    [kernel.kallsyms]        [k] apic_timer_interrupt
        ...

After:

   # perf report --group --stdio --group-sort-idx 3
   # To display the perf.data header info, please use --header/--header-only options.
   #
   #
   # Total Lost Samples: 0
   #
   # Samples: 12K of events 'cpu/instructions,period=2000003/, cpu/cpu-cycles,period=200003/, BR_MISP_RETIRED.ALL_BRANCHES:pp, cpu/event=0xc0,umask=1,cmask=1,
   # Event count (approx.): 6451235635
   #
   #                         Overhead  Command    Shared Object            Symbol
   # ................................  .........  .......................  ...................................
   #
       92.19%  98.68%   0.00%  93.30%  mgen       mgen                     [.] LOOP1
        0.00%   0.13%   0.00%   6.08%  swapper    [kernel.kallsyms]        [k] intel_idle
        3.12%   0.29%   0.00%   0.16%  gsd-color  libglib-2.0.so.0.5600.4  [.] 0x0000000000049515
        0.00%   0.00%   0.00%   0.06%  swapper    [kernel.kallsyms]        [k] hrtimer_start_range_ns
        1.56%   0.03%   0.00%   0.04%  gsd-color  libglib-2.0.so.0.5600.4  [.] 0x00000000000494b7
        0.00%   0.15%   0.00%   0.04%  perf       [kernel.kallsyms]        [k] smp_call_function_single
        0.00%   0.00%   0.00%   0.02%  mgen       [kernel.kallsyms]        [k] update_curr
        0.00%   0.00%   0.00%   0.02%  mgen       [kernel.kallsyms]        [k] apic_timer_interrupt
        0.00%   0.00%   0.00%   0.02%  mgen       [kernel.kallsyms]        [k] native_apic_msr_eoi_write
        0.00%   0.00%   0.00%   0.02%  mgen       [kernel.kallsyms]        [k] __update_load_avg_se
        0.00%   0.00%   0.00%   0.02%  mgen       [kernel.kallsyms]        [k] scheduler_tick

Now the output is sorted by the fourth event in group.

v6:
---
No change.

v5:
---
No change in v5.
v4 has been ACKed by Jiri.

v4:
---
1. Update Documentation/perf-report.txt to mention '--group-sort-idx' support multiple groups with different amount of events and it should be used on grouped events.

2. Update __hpp__group_sort_idx(), just return when the idx is out of limit.

3. Return failure on symbol_conf.group_sort_idx && !session->evlist->nr_groups. So now we don't need to use together with --group.

v3:
---
Refine the code in __hpp__group_sort_idx().

  Before:
    for (i = 1; i < nr_members; i++) {
         if (i == idx) {
                 ret = field_cmp(fields_a[i], fields_b[i]);
                 if (ret)
                         goto out;
         }
    }

  After:
    if (idx >= 1 && idx < nr_members) {
         ret = field_cmp(fields_a[idx], fields_b[idx]);
         if (ret)
                 goto out;
    }

v2:
---
No change

---
  tools/perf/Documentation/perf-report.txt |   5 ++
  tools/perf/builtin-report.c              |  10 +++
  tools/perf/ui/hist.c                     | 104 +++++++++++++++++++----
  tools/perf/util/symbol_conf.h            |   1 +
  4 files changed, 105 insertions(+), 15 deletions(-)
tools/perf/Documentation/perf-report.txt +5 -0
index 8dbe2119686a..fbb715bcb512 100644
@@ -371,6 +371,11 @@ OPTIONS
  	Show event group information together. It forces group output also
  	if there are no groups defined in data file.
 
+--group-sort-idx::
+	Sort the output by the event at the index n in group. If n is invalid,
+	sort by the first event. It can support multiple groups with different
+	amount of events. WARNING: This should be used on grouped events.
+
  --demangle::
  	Demangle symbol names to human readable form. It's enabled by default,
  	disable with --no-demangle.
tools/perf/builtin-report.c +10 -0
index de988589d99b..4c80a3ba73c3 100644
@@ -1211,6 +1211,10 @@ int cmd_report(int argc, const char **argv)
  		    "Show a column with the sum of periods"),
  	OPT_BOOLEAN_SET(0, "group", &symbol_conf.event_group, &report.group_set,
  		    "Show event group information together"),
+	OPT_INTEGER(0, "group-sort-idx", &symbol_conf.group_sort_idx,
+		    "Sort the output by the event at the index n in group. "
+		    "If n is invalid, sort by the first event. "
+		    "WARNING: should be used on grouped events."),
  	OPT_CALLBACK_NOOPT('b', "branch-stack", &branch_mode, "",
  		    "use branch records for per branch histogram filling",
  		    parse_branch_mode),
@@ -1350,6 +1354,12 @@ int cmd_report(int argc, const char **argv)
 
  	setup_forced_leader(&report, session->evlist);
 
+	if (symbol_conf.group_sort_idx && !session->evlist->nr_groups) {
+		parse_options_usage(NULL, options, "group-sort-idx", 0);
+		ret = -EINVAL;
+		goto error;
+	}
+
  	if (itrace_synth_opts.last_branch)
  		has_br_stack = true;
 
tools/perf/ui/hist.c +89 -15
index f73675500061..35224b366305 100644
@@ -151,15 +151,100 @@ static int field_cmp(u64 field_a, u64 field_b)
  	return 0;
  }
 
+static int pair_fields_alloc(struct hist_entry *a, struct hist_entry *b,
+			     hpp_field_fn get_field, int nr_members,
+			     u64 **fields_a, u64 **fields_b)
+{
+	struct evsel *evsel;
+	struct hist_entry *pair;
+	u64 *fa, *fb;
+	int ret = -1;
+
+	fa = calloc(nr_members, sizeof(*fa));
+	fb = calloc(nr_members, sizeof(*fb));
+
+	if (!fa || !fb)
+		goto out;
+
+	list_for_each_entry(pair, &a->pairs.head, pairs.node) {
+		evsel = hists_to_evsel(pair->hists);
+		fa[perf_evsel__group_idx(evsel)] = get_field(pair);
+	}
+
+	list_for_each_entry(pair, &b->pairs.head, pairs.node) {
+		evsel = hists_to_evsel(pair->hists);
+		fb[perf_evsel__group_idx(evsel)] = get_field(pair);
+	}
+
+	*fields_a = fa;
+	*fields_b = fb;
+	ret = 0;
+
+out:
+	if (ret != 0) {
+		free(fa);
+		free(fb);
+		*fields_a = NULL;
+		*fields_b = NULL;
+	}
+
+	return ret;
+}
+
+static int __hpp__group_sort_idx(struct hist_entry *a, struct hist_entry *b,
+				 hpp_field_fn get_field, int idx)
+{
+	struct evsel *evsel = hists_to_evsel(a->hists);
+	u64 *fields_a, *fields_b;
+	int cmp, nr_members, ret, i;
+
+	cmp = field_cmp(get_field(a), get_field(b));
+	if (!perf_evsel__is_group_event(evsel))
+		return cmp;
+
+	nr_members = evsel->core.nr_members;
+	if (idx < 1 || idx >= nr_members)
+		return cmp;
+
+	ret = pair_fields_alloc(a, b, get_field, nr_members,
+				&fields_a, &fields_b);
+	if (ret) {
+		ret = cmp;
+		goto out;
+	}
+
+	ret = field_cmp(fields_a[idx], fields_b[idx]);
+	if (ret)
+		goto out;
+
+	for (i = 1; i < nr_members; i++) {
+		if (i != idx) {
+			ret = field_cmp(fields_a[i], fields_b[i]);
+			if (ret)
+				goto out;
+		}
+	}
+
+out:
+	free(fields_a);
+	free(fields_b);
+
+	return ret;
+}
+
  static int __hpp__sort(struct hist_entry *a, struct hist_entry *b,
  		       hpp_field_fn get_field)
  {
  	s64 ret;
  	int i, nr_members;
  	struct evsel *evsel;
-	struct hist_entry *pair;
  	u64 *fields_a, *fields_b;
 
+	if (symbol_conf.group_sort_idx && symbol_conf.event_group) {
+		return __hpp__group_sort_idx(a, b, get_field,
+					     symbol_conf.group_sort_idx);
+	}
+
  	ret = field_cmp(get_field(a), get_field(b));
  	if (ret || !symbol_conf.event_group)
  		return ret;
@@ -169,22 +254,11 @@ static int __hpp__sort(struct hist_entry *a, struct hist_entry *b,
  		return ret;
 
  	nr_members = evsel->core.nr_members;
-	fields_a = calloc(nr_members, sizeof(*fields_a));
-	fields_b = calloc(nr_members, sizeof(*fields_b));
-
-	if (!fields_a || !fields_b)
+	i = pair_fields_alloc(a, b, get_field, nr_members,
+			      &fields_a, &fields_b);
+	if (i)
  		goto out;
 
-	list_for_each_entry(pair, &a->pairs.head, pairs.node) {
-		evsel = hists_to_evsel(pair->hists);
-		fields_a[perf_evsel__group_idx(evsel)] = get_field(pair);
-	}
-
-	list_for_each_entry(pair, &b->pairs.head, pairs.node) {
-		evsel = hists_to_evsel(pair->hists);
-		fields_b[perf_evsel__group_idx(evsel)] = get_field(pair);
-	}
-
  	for (i = 1; i < nr_members; i++) {
  		ret = field_cmp(fields_a[i], fields_b[i]);
  		if (ret)
tools/perf/util/symbol_conf.h +1 -0
index 10f1ec3e0349..b916afb95ec5 100644
@@ -73,6 +73,7 @@ struct symbol_conf {
  	const char	*symfs;
  	int		res_sample;
  	int		pad_output_len_dso;
+	int		group_sort_idx;
  };
 
  extern struct symbol_conf symbol_conf;
author
#3 lore raw

PATCH v6 3/4 perf report: Support a new key to reload the browser

In reply to Jin Yao

Sometimes we may need to reload the browser to update the output since some options are changed.

This patch creates a new key K_RELOAD. Once the __cmd_report() returns K_RELOAD, it would repeat the whole process, such as, read samples from data file, sort the data and display in the browser.

v6:
---
No change.

v5:
---
1. Fix the 'make NO_SLANG=1' error. Define K_RELOAD in util/hist.h.
2. Skip setup_sorting() in repeat path if last key is K_RELOAD.

v4:
---
Need to quit in perf_evsel_menu__run if key is K_RELOAD.

v3:
---
No change.

v2:
---
This is a new patch created in v2.

---
 tools/perf/builtin-report.c    | 6 +++---
 tools/perf/ui/browsers/hists.c | 1 +
 tools/perf/ui/keysyms.h        | 1 +
 tools/perf/util/hist.h         | 1 +
 4 files changed, 6 insertions(+), 3 deletions(-)
tools/perf/builtin-report.c +3 -3
index 4c80a3ba73c3..6f7c7cf061d6 100644
@@ -635,7 +635,7 @@ static int report__browse_hists(struct report *rep)
 		 * Usually "ret" is the last pressed key, and we only
 		 * care if the key notifies us to switch data file.
 		 */
-		if (ret != K_SWITCH_INPUT_DATA)
+		if (ret != K_SWITCH_INPUT_DATA && ret != K_RELOAD)
 			ret = 0;
 		break;
 	case 2:
@@ -1461,7 +1461,7 @@ int cmd_report(int argc, const char **argv)
 		sort_order = sort_tmp;
 	}
 
-	if ((last_key != K_SWITCH_INPUT_DATA) &&
+	if ((last_key != K_SWITCH_INPUT_DATA && last_key != K_RELOAD) &&
 	    (setup_sorting(session->evlist) < 0)) {
 		if (sort_order)
 			parse_options_usage(report_usage, options, "s", 1);
@@ -1540,7 +1540,7 @@ int cmd_report(int argc, const char **argv)
 	sort__setup_elide(stdout);
 
 	ret = __cmd_report(&report);
-	if (ret == K_SWITCH_INPUT_DATA) {
+	if (ret == K_SWITCH_INPUT_DATA || ret == K_RELOAD) {
 		perf_session__delete(session);
 		last_key = K_SWITCH_INPUT_DATA;
 		goto repeat;
tools/perf/ui/browsers/hists.c +1 -0
index d4d3558fdef4..096b3bc0fbef 100644
@@ -3335,6 +3335,7 @@ static int perf_evsel_menu__run(struct evsel_menu *menu,
 					pos = perf_evsel__prev(pos);
 				goto browse_hists;
 			case K_SWITCH_INPUT_DATA:
+			case K_RELOAD:
 			case 'q':
 			case CTRL('c'):
 				goto out;
tools/perf/ui/keysyms.h +1 -0
index fbfac29077f2..04cc4e5c031f 100644
@@ -25,5 +25,6 @@
 #define K_ERROR	 -2
 #define K_RESIZE -3
 #define K_SWITCH_INPUT_DATA -4
+#define K_RELOAD -5
 
 #endif /* _PERF_KEYSYMS_H_ */
tools/perf/util/hist.h +1 -0
index 45286900aacb..93ea84009393 100644
@@ -536,6 +536,7 @@ static inline int block_hists_tui_browse(struct block_hist *bh __maybe_unused,
 #define K_LEFT  -1000
 #define K_RIGHT -2000
 #define K_SWITCH_INPUT_DATA -3000
+#define K_RELOAD -4000
 #endif
 
 unsigned int hists__sort_list_width(struct hists *hists);
Signature · 1 line
2.17.1
author
#4 lore raw

PATCH v6 4/4 perf report: support hotkey to let user select any event for sorting

In reply to Jin Yao

When performing "perf report --group", it shows the event group information together. In previous patch, we have supported a new option "--group-sort-idx" to sort the output by the event at the index n in event group.

It would be nice if we can use a hotkey in browser to select a event to sort.

For example,

  # perf report --group

Samples: 12K of events 'cpu/instructions,period=2000003/, cpu/cpu-cycles,period=200003/, ...

                        Overhead  Command    Shared Object            Symbol
  92.19%  98.68%   0.00%  93.30%  mgen       mgen                     [.] LOOP1
   3.12%   0.29%   0.00%   0.16%  gsd-color  libglib-2.0.so.0.5600.4  [.] 0x0000000000049515
   1.56%   0.03%   0.00%   0.04%  gsd-color  libglib-2.0.so.0.5600.4  [.] 0x00000000000494b7
   1.56%   0.01%   0.00%   0.00%  gsd-color  libglib-2.0.so.0.5600.4  [.] 0x00000000000494ce
   1.56%   0.00%   0.00%   0.00%  mgen       [kernel.kallsyms]        [k] task_tick_fair
   0.00%   0.15%   0.00%   0.04%  perf       [kernel.kallsyms]        [k] smp_call_function_single
   0.00%   0.13%   0.00%   6.08%  swapper    [kernel.kallsyms]        [k] intel_idle
   0.00%   0.03%   0.00%   0.00%  gsd-color  libglib-2.0.so.0.5600.4  [.] g_main_context_check
   0.00%   0.03%   0.00%   0.00%  swapper    [kernel.kallsyms]        [k] apic_timer_interrupt
   0.00%   0.03%   0.00%   0.00%  swapper    [kernel.kallsyms]        [k] check_preempt_curr

When user press hotkey '3' (event index, starting from 0), it indicates to sort output by the forth event in group.

Samples: 12K of events 'cpu/instructions,period=2000003/, cpu/cpu-cycles,period=200003/, ...

                        Overhead  Command    Shared Object            Symbol
  92.19%  98.68%   0.00%  93.30%  mgen       mgen                     [.] LOOP1
   0.00%   0.13%   0.00%   6.08%  swapper    [kernel.kallsyms]        [k] intel_idle
   3.12%   0.29%   0.00%   0.16%  gsd-color  libglib-2.0.so.0.5600.4  [.] 0x0000000000049515
   0.00%   0.00%   0.00%   0.06%  swapper    [kernel.kallsyms]        [k] hrtimer_start_range_ns
   1.56%   0.03%   0.00%   0.04%  gsd-color  libglib-2.0.so.0.5600.4  [.] 0x00000000000494b7
   0.00%   0.15%   0.00%   0.04%  perf       [kernel.kallsyms]        [k] smp_call_function_single
   0.00%   0.00%   0.00%   0.02%  mgen       [kernel.kallsyms]        [k] update_curr
   0.00%   0.00%   0.00%   0.02%  mgen       [kernel.kallsyms]        [k] apic_timer_interrupt
   0.00%   0.00%   0.00%   0.02%  mgen       [kernel.kallsyms]        [k] native_apic_msr_eoi_write
   0.00%   0.00%   0.00%   0.02%  mgen       [kernel.kallsyms]        [k] __update_load_avg_se

v6:
---
Jiri provided a good improvement to eliminate unneeded refresh. This improvement is added to v6.

v5:
---
No change

v4:
---
No change

v3:
---
No change

v2:
---
1. Report warning at helpline when index is invalid.
2. Report warning at helpline when it's not group event.
3. Use "case '0' ... '9'" to refine the code
4. Split K_RELOAD implementation to another patch.

---
 tools/perf/ui/browsers/hists.c | 28 +++++++++++++++++++++++++++-
 1 file changed, 27 insertions(+), 1 deletion(-)
tools/perf/ui/browsers/hists.c +27 -1
index 096b3bc0fbef..9f5dd48500a2 100644
@@ -2876,7 +2876,8 @@ static int perf_evsel__hists_browse(struct evsel *evsel, int nr_events,
 	"s             Switch to another data file in PWD\n"
 	"t             Zoom into current Thread\n"
 	"V             Verbose (DSO names in callchains, etc)\n"
-	"/             Filter symbol by name";
+	"/             Filter symbol by name\n"
+	"0-9           Sort by event n in group";
 	static const char top_help[] = HIST_BROWSER_HELP_COMMON
 	"P             Print histograms to perf.hist.N\n"
 	"t             Zoom into current Thread\n"
@@ -2937,6 +2938,31 @@ static int perf_evsel__hists_browse(struct evsel *evsel, int nr_events,
 			 * go to the next or previous
 			 */
 			goto out_free_stack;
+		case '0' ... '9':
+			if (!symbol_conf.event_group ||
+			    evsel->core.nr_members < 2) {
+				snprintf(buf, sizeof(buf),
+					 "Sort by index only available with group events!");
+				helpline = buf;
+				continue;
+			}
+
+			if (key - '0' == symbol_conf.group_sort_idx)
+				continue;
+
+			symbol_conf.group_sort_idx = key - '0';
+
+			if (symbol_conf.group_sort_idx >= evsel->core.nr_members) {
+				snprintf(buf, sizeof(buf),
+					 "Max event group index to sort is %d (index from 0 to %d)",
+					 evsel->core.nr_members - 1,
+					 evsel->core.nr_members - 1);
+				helpline = buf;
+				continue;
+			}
+
+			key = K_RELOAD;
+			goto out_free_stack;
 		case 'a':
 			if (!hists__has(hists, sym)) {
 				ui_browser__warning(&browser->b, delay_secs * 2,
Signature · 1 line
2.17.1
Acked-by on v6 1/4
#5 lore raw
In reply to Jin Yao
Quoted 18 lines · Jin Yao

We observed an issue that was some extra columns displayed after switching perf data file in browser. The steps to reproduce:

1. perf record -a -e cycles,instructions -- sleep 3
2. perf report --group
3. In browser, we use hotkey 's' to switch to another perf.data
4. Now in browser, the extra columns 'Self' and 'Children' are displayed.

The issue is setup_sorting() executed again after repeat path, so dimensions are added again.

This patch checks the last key returned from __cmd_report(). If it's K_SWITCH_INPUT_DATA, skips the setup_sorting().

v6:
---
No change.

Quoted 5 lines

v5:
---
New patch in v5.

for all 4 patches:

thanks,
jirka

#6 lore raw
In reply to Jin Yao

Em Fri, Dec 20, 2019 at 09:37:19AM +0800, Jin Yao escreveu:

Quoted 8 lines

We observed an issue that was some extra columns displayed after switching perf data file in browser. The steps to reproduce:

1. perf record -a -e cycles,instructions -- sleep 3
2. perf report --group
3. In browser, we use hotkey 's' to switch to another perf.data
4. Now in browser, the extra columns 'Self' and 'Children' are displayed.

Quoted 5 lines

The issue is setup_sorting() executed again after repeat path, so dimensions are added again.

This patch checks the last key returned from __cmd_report(). If it's K_SWITCH_INPUT_DATA, skips the setup_sorting().

you forgot to add this right?

Cc Feng Tang <[email protected]>
Fixes ad0de0971b7f ("perf report: Enable the runtime switching of perf data file")
Quoted 45 lines

v6:
---
No change.

v5:
---
New patch in v5.

---
 tools/perf/builtin-report.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
tools/perf/builtin-report.c +4 -1
index 387311c67264..de988589d99b 100644
@@ -1076,6 +1076,7 @@ int cmd_report(int argc, const char **argv)
 	struct stat st;
 	bool has_br_stack = false;
 	int branch_mode = -1;
+	int last_key = 0;
 	bool branch_call_mode = false;
 #define CALLCHAIN_DEFAULT_OPT  "graph,0.5,caller,function,percent"
 	static const char report_callchain_help[] = "Display call graph (stack chain/backtrace):\n\n"
@@ -1450,7 +1451,8 @@ int cmd_report(int argc, const char **argv)
 		sort_order = sort_tmp;
 	}
 
-	if (setup_sorting(session->evlist) < 0) {
+	if ((last_key != K_SWITCH_INPUT_DATA) &&
+	    (setup_sorting(session->evlist) < 0)) {
 		if (sort_order)
 			parse_options_usage(report_usage, options, "s", 1);
 		if (field_order)
@@ -1530,6 +1532,7 @@ int cmd_report(int argc, const char **argv)
 	ret = __cmd_report(&report);
 	if (ret == K_SWITCH_INPUT_DATA) {
 		perf_session__delete(session);
+		last_key = K_SWITCH_INPUT_DATA;
 		goto repeat;
 	} else
 		ret = 0;
Signature · 1 line
2.17.1
Signature · 1 line
- Arnaldo
author
#7 lore raw
Quoted 12 lines · Arnaldo Carvalho de Melo

Em Fri, Dec 20, 2019 at 09:37:19AM +0800, Jin Yao escreveu:

Quoted 11 lines

We observed an issue that was some extra columns displayed after switching perf data file in browser. The steps to reproduce:

1. perf record -a -e cycles,instructions -- sleep 3
2. perf report --group
3. In browser, we use hotkey 's' to switch to another perf.data
4. Now in browser, the extra columns 'Self' and 'Children' are displayed.

The issue is setup_sorting() executed again after repeat path, so dimensions are added again.

Quoted 8 lines
Quoted 2 lines

This patch checks the last key returned from __cmd_report(). If it's K_SWITCH_INPUT_DATA, skips the setup_sorting().

you forgot to add this right?

Cc Feng Tang <[email protected]>
Fixes ad0de0971b7f ("perf report: Enable the runtime switching of perf data file")

Yes, I should add this in patch description. Thanks for reminding

Do you need me to resend this patch-set (v7)?

Thanks
Jin Yao

Quoted 46 lines
Quoted 45 lines

v6:
---
No change.

v5:
---
New patch in v5.

---
  tools/perf/builtin-report.c | 5 ++++-
  1 file changed, 4 insertions(+), 1 deletion(-)
tools/perf/builtin-report.c +4 -1
index 387311c67264..de988589d99b 100644
@@ -1076,6 +1076,7 @@ int cmd_report(int argc, const char **argv)
  	struct stat st;
  	bool has_br_stack = false;
  	int branch_mode = -1;
+	int last_key = 0;
  	bool branch_call_mode = false;
  #define CALLCHAIN_DEFAULT_OPT  "graph,0.5,caller,function,percent"
  	static const char report_callchain_help[] = "Display call graph (stack chain/backtrace):\n\n"
@@ -1450,7 +1451,8 @@ int cmd_report(int argc, const char **argv)
  		sort_order = sort_tmp;
  	}
 
-	if (setup_sorting(session->evlist) < 0) {
+	if ((last_key != K_SWITCH_INPUT_DATA) &&
+	    (setup_sorting(session->evlist) < 0)) {
  		if (sort_order)
  			parse_options_usage(report_usage, options, "s", 1);
  		if (field_order)
@@ -1530,6 +1532,7 @@ int cmd_report(int argc, const char **argv)
  	ret = __cmd_report(&report);
  	if (ret == K_SWITCH_INPUT_DATA) {
  		perf_session__delete(session);
+		last_key = K_SWITCH_INPUT_DATA;
  		goto repeat;
  	} else
  		ret = 0;
Signature · 1 line
2.17.1
#8 lore raw

Em Mon, Dec 23, 2019 at 08:55:00AM +0800, Jin, Yao escreveu:

Quoted 18 lines · Arnaldo Carvalho de Melo
Quoted 15 lines

Em Fri, Dec 20, 2019 at 09:37:19AM +0800, Jin Yao escreveu:

Quoted 13 lines

We observed an issue that was some extra columns displayed after switching perf data file in browser. The steps to reproduce:

1. perf record -a -e cycles,instructions -- sleep 3
2. perf report --group
3. In browser, we use hotkey 's' to switch to another perf.data
4. Now in browser, the extra columns 'Self' and 'Children' are displayed.

The issue is setup_sorting() executed again after repeat path, so dimensions are added again.

This patch checks the last key returned from __cmd_report(). If it's K_SWITCH_INPUT_DATA, skips the setup_sorting().

Quoted 8 lines
Quoted 4 lines

you forgot to add this right?

Cc Feng Tang <[email protected]>
Fixes ad0de0971b7f ("perf report: Enable the runtime switching of perf data file")

Yes, I should add this in patch description. Thanks for reminding

Do you need me to resend this patch-set (v7)?

Nope, I added it myself, thanks.

- Arnaldo

Quoted 48 lines

Thanks
Jin Yao

Quoted 45 lines
Quoted 44 lines

v6:
---
No change.

v5:
---
New patch in v5.

---
  tools/perf/builtin-report.c | 5 ++++-
  1 file changed, 4 insertions(+), 1 deletion(-)
tools/perf/builtin-report.c +4 -1
index 387311c67264..de988589d99b 100644
@@ -1076,6 +1076,7 @@ int cmd_report(int argc, const char **argv)
  	struct stat st;
  	bool has_br_stack = false;
  	int branch_mode = -1;
+	int last_key = 0;
  	bool branch_call_mode = false;
  #define CALLCHAIN_DEFAULT_OPT  "graph,0.5,caller,function,percent"
  	static const char report_callchain_help[] = "Display call graph (stack chain/backtrace):\n\n"
@@ -1450,7 +1451,8 @@ int cmd_report(int argc, const char **argv)
  		sort_order = sort_tmp;
  	}
-	if (setup_sorting(session->evlist) < 0) {
+	if ((last_key != K_SWITCH_INPUT_DATA) &&
+	    (setup_sorting(session->evlist) < 0)) {
  		if (sort_order)
  			parse_options_usage(report_usage, options, "s", 1);
  		if (field_order)
@@ -1530,6 +1532,7 @@ int cmd_report(int argc, const char **argv)
  	ret = __cmd_report(&report);
  	if (ret == K_SWITCH_INPUT_DATA) {
  		perf_session__delete(session);
+		last_key = K_SWITCH_INPUT_DATA;
  		goto repeat;
  	} else
  		ret = 0;
Signature · 1 line
2.17.1
Signature · 1 line
- Arnaldo
merge bot Applied
#9 lore raw
In reply to Jin Yao
The following commit has been merged into the perf/urgent branch of tip:

The following commit has been merged into the perf/urgent branch of tip:

Commit-ID:     0feba17bd7ee3b7e03d141f119049dcc23efa94e
Gitweb:        https://git.kernel.org/tip/0feba17bd7ee3b7e03d141f119049dcc23efa94e
Author:        Jin Yao <[email protected]>
AuthorDate:    Fri, 20 Dec 2019 09:37:19 +08:00
Committer:     Arnaldo Carvalho de Melo <[email protected]>
CommitterDate: Fri, 20 Dec 2019 18:49:27 -03:00

perf report: Fix incorrectly added dimensions as switch perf data file

We observed an issue that was some extra columns displayed after switching perf data file in browser. The steps to reproduce:

1. perf record -a -e cycles,instructions -- sleep 3
2. perf report --group
3. In browser, we use hotkey 's' to switch to another perf.data
4. Now in browser, the extra columns 'Self' and 'Children' are displayed.

The issue is setup_sorting() executed again after repeat path, so dimensions are added again.

This patch checks the last key returned from __cmd_report(). If it's K_SWITCH_INPUT_DATA, skips the setup_sorting().

Fixes ad0de0971b7f ("perf report: Enable the runtime switching of perf data file")
Cc Alexander Shishkin <[email protected]>
Cc Andi Kleen <[email protected]>
Cc Feng Tang <[email protected]>
Cc Jin Yao <[email protected]>
Cc Kan Liang <[email protected]>
Cc Peter Zijlstra <[email protected]>
---
 tools/perf/builtin-report.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
tools/perf/builtin-report.c +4 -1
index 387311c..de98858 100644
@@ -1076,6 +1076,7 @@ int cmd_report(int argc, const char **argv)
 	struct stat st;
 	bool has_br_stack = false;
 	int branch_mode = -1;
+	int last_key = 0;
 	bool branch_call_mode = false;
 #define CALLCHAIN_DEFAULT_OPT  "graph,0.5,caller,function,percent"
 	static const char report_callchain_help[] = "Display call graph (stack chain/backtrace):\n\n"
@@ -1450,7 +1451,8 @@ repeat:
 		sort_order = sort_tmp;
 	}
 
-	if (setup_sorting(session->evlist) < 0) {
+	if ((last_key != K_SWITCH_INPUT_DATA) &&
+	    (setup_sorting(session->evlist) < 0)) {
 		if (sort_order)
 			parse_options_usage(report_usage, options, "s", 1);
 		if (field_order)
@@ -1530,6 +1532,7 @@ repeat:
 	ret = __cmd_report(&report);
 	if (ret == K_SWITCH_INPUT_DATA) {
 		perf_session__delete(session);
+		last_key = K_SWITCH_INPUT_DATA;
 		goto repeat;
 	} else
 		ret = 0;