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

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

[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.

 v5:
 ---
 New patch in v5.

Signed-off-by: Jin Yao <yao.jin@linux.intel.com>
---
 tools/perf/builtin-report.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c
index 387311c67264..de988589d99b 100644
--- a/tools/perf/builtin-report.c
+++ b/tools/perf/builtin-report.c
@@ -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;
-- 
2.17.1

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

[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

Signed-off-by: Jin Yao <yao.jin@linux.intel.com>
---
 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(-)

diff --git a/tools/perf/Documentation/perf-report.txt b/tools/perf/Documentation/perf-report.txt
index 8dbe2119686a..fbb715bcb512 100644
--- a/tools/perf/Documentation/perf-report.txt
+++ b/tools/perf/Documentation/perf-report.txt
@@ -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.
diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c
index de988589d99b..4c80a3ba73c3 100644
--- a/tools/perf/builtin-report.c
+++ b/tools/perf/builtin-report.c
@@ -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;
 
diff --git a/tools/perf/ui/hist.c b/tools/perf/ui/hist.c
index f73675500061..35224b366305 100644
--- a/tools/perf/ui/hist.c
+++ b/tools/perf/ui/hist.c
@@ -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)
diff --git a/tools/perf/util/symbol_conf.h b/tools/perf/util/symbol_conf.h
index 10f1ec3e0349..b916afb95ec5 100644
--- a/tools/perf/util/symbol_conf.h
+++ b/tools/perf/util/symbol_conf.h
@@ -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;
-- 
2.17.1

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

[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.

Signed-off-by: Jin Yao <yao.jin@linux.intel.com>
---
 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(-)

diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c
index 4c80a3ba73c3..6f7c7cf061d6 100644
--- a/tools/perf/builtin-report.c
+++ b/tools/perf/builtin-report.c
@@ -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;
diff --git a/tools/perf/ui/browsers/hists.c b/tools/perf/ui/browsers/hists.c
index d4d3558fdef4..096b3bc0fbef 100644
--- a/tools/perf/ui/browsers/hists.c
+++ b/tools/perf/ui/browsers/hists.c
@@ -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;
diff --git a/tools/perf/ui/keysyms.h b/tools/perf/ui/keysyms.h
index fbfac29077f2..04cc4e5c031f 100644
--- a/tools/perf/ui/keysyms.h
+++ b/tools/perf/ui/keysyms.h
@@ -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_ */
diff --git a/tools/perf/util/hist.h b/tools/perf/util/hist.h
index 45286900aacb..93ea84009393 100644
--- a/tools/perf/util/hist.h
+++ b/tools/perf/util/hist.h
@@ -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);
-- 
2.17.1

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

[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.

Signed-off-by: Jin Yao <yao.jin@linux.intel.com>
---
 tools/perf/ui/browsers/hists.c | 28 +++++++++++++++++++++++++++-
 1 file changed, 27 insertions(+), 1 deletion(-)

diff --git a/tools/perf/ui/browsers/hists.c b/tools/perf/ui/browsers/hists.c
index 096b3bc0fbef..9f5dd48500a2 100644
--- a/tools/perf/ui/browsers/hists.c
+++ b/tools/perf/ui/browsers/hists.c
@@ -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,
-- 
2.17.1

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

[Jiri Olsa]

On Fri, Dec 20, 2019 at 09:37:19AM +0800, Jin Yao wrote:
> 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.
> 
> Signed-off-by: Jin Yao <yao.jin@linux.intel.com>

for all 4 patches:

Acked-by: Jiri Olsa <jolsa@redhat.com>

thanks,
jirka

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

[Arnaldo Carvalho de Melo]

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

you forgot to add this right?

Cc: Feng Tang <feng.tang@intel.com>
Fixes: ad0de0971b7f ("perf report: Enable the runtime switching of perf data file")
 
>  v6:
>  ---
>  No change.
> 
>  v5:
>  ---
>  New patch in v5.
> 
> Signed-off-by: Jin Yao <yao.jin@linux.intel.com>
> ---
>  tools/perf/builtin-report.c | 5 ++++-
>  1 file changed, 4 insertions(+), 1 deletion(-)
> 
> diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c
> index 387311c67264..de988589d99b 100644
> --- a/tools/perf/builtin-report.c
> +++ b/tools/perf/builtin-report.c
> @@ -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;
> -- 
> 2.17.1

-- 

- Arnaldo

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

[Jin, Yao]

On 12/21/2019 12:34 AM, Arnaldo Carvalho de Melo wrote:
> Em Fri, Dec 20, 2019 at 09:37:19AM +0800, Jin Yao escreveu:
>> 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().
> 
> you forgot to add this right?
> 
> Cc: Feng Tang <feng.tang@intel.com>
> 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

>>   v6:
>>   ---
>>   No change.
>>
>>   v5:
>>   ---
>>   New patch in v5.
>>
>> Signed-off-by: Jin Yao <yao.jin@linux.intel.com>
>> ---
>>   tools/perf/builtin-report.c | 5 ++++-
>>   1 file changed, 4 insertions(+), 1 deletion(-)
>>
>> diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c
>> index 387311c67264..de988589d99b 100644
>> --- a/tools/perf/builtin-report.c
>> +++ b/tools/perf/builtin-report.c
>> @@ -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;
>> -- 
>> 2.17.1
> 

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

[Arnaldo Carvalho de Melo]

Em Mon, Dec 23, 2019 at 08:55:00AM +0800, Jin, Yao escreveu:
> 
> 
> On 12/21/2019 12:34 AM, Arnaldo Carvalho de Melo wrote:
> > Em Fri, Dec 20, 2019 at 09:37:19AM +0800, Jin Yao escreveu:
> > > 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().
> > 
> > you forgot to add this right?
> > 
> > Cc: Feng Tang <feng.tang@intel.com>
> > 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
 
> Thanks
> Jin Yao
> 
> > >   v6:
> > >   ---
> > >   No change.
> > > 
> > >   v5:
> > >   ---
> > >   New patch in v5.
> > > 
> > > Signed-off-by: Jin Yao <yao.jin@linux.intel.com>
> > > ---
> > >   tools/perf/builtin-report.c | 5 ++++-
> > >   1 file changed, 4 insertions(+), 1 deletion(-)
> > > 
> > > diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c
> > > index 387311c67264..de988589d99b 100644
> > > --- a/tools/perf/builtin-report.c
> > > +++ b/tools/perf/builtin-report.c
> > > @@ -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;
> > > -- 
> > > 2.17.1
> > 

-- 

- Arnaldo

[tip: perf/urgent] perf report: Fix incorrectly added dimensions as switch perf data file

[tip-bot2 for Jin Yao]

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 <yao.jin@linux.intel.com>
AuthorDate:    Fri, 20 Dec 2019 09:37:19 +08:00
Committer:     Arnaldo Carvalho de Melo <acme@redhat.com>
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")
Signed-off-by: Jin Yao <yao.jin@linux.intel.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Acked-by: Jiri Olsa <jolsa@redhat.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Feng Tang <feng.tang@intel.com>
Cc: Jin Yao <yao.jin@intel.com>
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lore.kernel.org/lkml/20191220013722.20592-1-yao.jin@linux.intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/builtin-report.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c
index 387311c..de98858 100644
--- a/tools/perf/builtin-report.c
+++ b/tools/perf/builtin-report.c
@@ -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;

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

[Jin, Yao]

Hi Arnaldo,

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

Thanks
Jin Yao

On 12/20/2019 9:37 AM, Jin Yao wrote:
> 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
> 
> Signed-off-by: Jin Yao <yao.jin@linux.intel.com>
> ---
>   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(-)
> 
> diff --git a/tools/perf/Documentation/perf-report.txt b/tools/perf/Documentation/perf-report.txt
> index 8dbe2119686a..fbb715bcb512 100644
> --- a/tools/perf/Documentation/perf-report.txt
> +++ b/tools/perf/Documentation/perf-report.txt
> @@ -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.
> diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c
> index de988589d99b..4c80a3ba73c3 100644
> --- a/tools/perf/builtin-report.c
> +++ b/tools/perf/builtin-report.c
> @@ -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;
>   
> diff --git a/tools/perf/ui/hist.c b/tools/perf/ui/hist.c
> index f73675500061..35224b366305 100644
> --- a/tools/perf/ui/hist.c
> +++ b/tools/perf/ui/hist.c
> @@ -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)
> diff --git a/tools/perf/util/symbol_conf.h b/tools/perf/util/symbol_conf.h
> index 10f1ec3e0349..b916afb95ec5 100644
> --- a/tools/perf/util/symbol_conf.h
> +++ b/tools/perf/util/symbol_conf.h
> @@ -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;
>