Skip to contentSkip to conversation

perf report: Add -F option for specifying output fields (v5)

PATCH In review Namhyung Kim 3 participants 41 messages May 12, 2014 – May 19, 2014 perf events

lore
Share & tools
lore

Namhyung Kim

author maintainer
#0 lore raw

cover perf report: Add -F option for specifying output fields (v5)

Hello,

This is a patchset implementing -F/--fields option to setup output field/column as Ingo requested.

The -F option can receive any sort keys that -s option recognize, plus following fields (name can be changed):

overhead, overhead_sys, overhead_us, sample, period

The overhead_guest_sys and overhead_guest_us might be avaiable when you profile guest machines.

Output will be sorted by in order of fields and sort keys passed by -s option will be added to the output field list automatically. If you want to change the order of sorting you can give -s option in addition to -F option. To support old behavior, it'll also prepend 'overhead' field to the sort keys unless you give -F option explicitly.

  $ perf report -s dso,sym
  ...
  # Overhead  Shared Object                      Symbol
  # ........  .............  ..........................
  #
      13.75%  ld-2.17.so     [.] strcmp                
      10.00%  abc            [.] a                     
      10.00%  abc            [.] b                     
      10.00%  abc            [.] c                     
       8.75%  abc            [.] main                  
       7.50%  libc-2.17.so   [.] _setjmp               
       6.25%  abc            [.] _init                 
       6.25%  abc            [.] frame_dummy           
       5.00%  abc            [.] __libc_csu_init       
       5.00%  ld-2.17.so     [.] _dl_name_match_p      
       3.75%  libc-2.17.so   [.] __new_exitfn          
       2.50%  libc-2.17.so   [.] __cxa_atexit          
       1.25%  ld-2.17.so     [.] _dl_check_map_versions
       1.25%  ld-2.17.so     [.] _dl_setup_hash        
       1.25%  ld-2.17.so     [.] _dl_sysdep_start      
       1.25%  ld-2.17.so     [.] brk                   
       1.25%  ld-2.17.so     [.] calloc@plt            
       1.25%  ld-2.17.so     [.] dl_main               
       1.25%  ld-2.17.so     [.] match_symbol          
       1.25%  ld-2.17.so     [.] sbrk                  
       1.25%  ld-2.17.so     [.] strlen                
  $ perf report -F sym,sample,overhead
  ...
  #                     Symbol       Samples  Overhead
  # ..........................  ............  ........
  #
    [.] __cxa_atexit                       2     2.50%
    [.] __libc_csu_init                    4     5.00%
    [.] __new_exitfn                       3     3.75%
    [.] _dl_check_map_versions             1     1.25%
    [.] _dl_name_match_p                   4     5.00%
    [.] _dl_setup_hash                     1     1.25%
    [.] _dl_sysdep_start                   1     1.25%
    [.] _init                              5     6.25%
    [.] _setjmp                            6     7.50%
    [.] a                                  8    10.00%
    [.] b                                  8    10.00%
    [.] brk                                1     1.25%
    [.] c                                  8    10.00%
    [.] calloc@plt                         1     1.25%
    [.] dl_main                            1     1.25%
    [.] frame_dummy                        5     6.25%
    [.] main                               7     8.75%
    [.] match_symbol                       1     1.25%
    [.] sbrk                               1     1.25%
    [.] strcmp                            11    13.75%
    [.] strlen                             1     1.25%
  $ perf report -F sym,sample -s overhead
  ...
  #                     Symbol       Samples  Overhead
  # ..........................  ............  ........
  #
    [.] strcmp                            11    13.75%
    [.] a                                  8    10.00%
    [.] b                                  8    10.00%
    [.] c                                  8    10.00%
    [.] main                               7     8.75%
    [.] _setjmp                            6     7.50%
    [.] _init                              5     6.25%
    [.] frame_dummy                        5     6.25%
    [.] __libc_csu_init                    4     5.00%
    [.] _dl_name_match_p                   4     5.00%
    [.] __new_exitfn                       3     3.75%
    [.] __cxa_atexit                       2     2.50%
    [.] _dl_check_map_versions             1     1.25%
    [.] _dl_setup_hash                     1     1.25%
    [.] _dl_sysdep_start                   1     1.25%
    [.] brk                                1     1.25%
    [.] calloc@plt                         1     1.25%
    [.] dl_main                            1     1.25%
    [.] match_symbol                       1     1.25%
    [.] sbrk                               1     1.25%
    [.] strlen                             1     1.25%

* changes in v5:
- add a testcase for hist output sorting

* changes in v4:
- fix a tui navigation bug
- fix a bug in output change of perf diff
- move call to perf_hpp__init() out of setup_browser()
- fix alignment of some output fields on stdio

* changes in v3:
- rename to --fields option for consistency (David)
- prevent to add same keys multiple times
- change dso sorting to show unknown dsos last
- fix minor bugs

* changes in v2:
- add a cleanup patch using ui__has_annotation()
- cleanup default sort order managment
- support perf top also
- handle elided sort entries properly
- add Acked-by's from Ingo

I pushed the patch series on the 'perf/field-v5' branch in my tree

git://git.kernel.org/pub/scm/linux/kernel/git/namhyung/linux-perf.git

Any comments are welcome, please test!

Thanks,
Namhyung

Namhyung Kim (20):
perf tools: Add ->cmp(), ->collapse() and ->sort() to perf_hpp_fmt perf tools: Convert sort entries to hpp formats
perf tools: Use hpp formats to sort hist entries
perf tools: Support event grouping in hpp ->sort()
perf tools: Use hpp formats to sort final output
perf tools: Consolidate output field handling to hpp format routines perf ui: Get rid of callback from __hpp__fmt()
perf tools: Allow hpp fields to be sort keys
perf tools: Consolidate management of default sort orders perf tools: Call perf_hpp__init() before setting up GUI browsers perf report: Add -F option to specify output fields
perf tools: Add ->sort() member to struct sort_entry
perf report/tui: Fix a bug when --fields/sort is given perf top: Add --fields option to specify output fields perf diff: Add missing setup_output_field()
perf tools: Skip elided sort entries
perf hists: Reset width of output fields with header length perf tools: Introduce reset_output_field()
perf tests: Factor out print_hists_*()
perf tests: Add a testcase for histogram output sorting

 tools/perf/Documentation/perf-report.txt |  10 +
 tools/perf/Documentation/perf-top.txt    |   9 +
 tools/perf/Makefile.perf                 |   1 +
 tools/perf/builtin-diff.c                |   3 +
 tools/perf/builtin-report.c              |  31 +-
 tools/perf/builtin-top.c                 |  12 +-
 tools/perf/tests/builtin-test.c          |   4 +
 tools/perf/tests/hists_common.c          |  57 +++
 tools/perf/tests/hists_common.h          |   3 +
 tools/perf/tests/hists_filter.c          |  37 +-
 tools/perf/tests/hists_link.c            |  29 +-
 tools/perf/tests/hists_output.c          | 628 +++++++++++++++++++++++++++++++
 tools/perf/tests/tests.h                 |   1 +
 tools/perf/ui/browsers/hists.c           |  76 ++--
 tools/perf/ui/gtk/hists.c                |  41 +-
 tools/perf/ui/hist.c                     | 224 +++++++++--
 tools/perf/ui/setup.c                    |   2 -
 tools/perf/ui/stdio/hist.c               |  54 +--
 tools/perf/util/hist.c                   |  83 ++--
 tools/perf/util/hist.h                   |  27 +-
 tools/perf/util/sort.c                   | 382 ++++++++++++++++++-
 tools/perf/util/sort.h                   |   5 +
 22 files changed, 1415 insertions(+), 304 deletions(-)
 create mode 100644 tools/perf/tests/hists_output.c
Signature · 1 line
1.9.2
author maintainer
#1 lore raw

PATCH 1/20 perf tools: Add ->cmp(), ->collapse() and ->sort() to perf_hpp_fmt

Those function pointers will be used to sort report output based on the selected fields. This is a preparation of later change.

---
 tools/perf/ui/hist.c   | 39 +++++++++++++++++++++++++++++++++++----
 tools/perf/util/hist.h |  3 +++
 2 files changed, 38 insertions(+), 4 deletions(-)
tools/perf/ui/hist.c +35 -4
index 0912805c08f4..d4a4f2e7eb43 100644
@@ -192,6 +192,14 @@ static int hpp__entry_##_type(struct perf_hpp_fmt *_fmt __maybe_unused,		\
 			  hpp_entry_scnprintf, true);				\
 }
 
+#define __HPP_SORT_FN(_type, _field)						\
+static int64_t hpp__sort_##_type(struct hist_entry *a, struct hist_entry *b)	\
+{										\
+	s64 __a = he_get_##_field(a);						\
+	s64 __b = he_get_##_field(b);						\
+	return __a - __b;							\
+}
+
 #define __HPP_ENTRY_RAW_FN(_type, _field)					\
 static u64 he_get_raw_##_field(struct hist_entry *he)				\
 {										\
@@ -206,16 +214,27 @@ static int hpp__entry_##_type(struct perf_hpp_fmt *_fmt __maybe_unused,		\
 			  hpp_entry_scnprintf, false);				\
 }
 
+#define __HPP_SORT_RAW_FN(_type, _field)					\
+static int64_t hpp__sort_##_type(struct hist_entry *a, struct hist_entry *b)	\
+{										\
+	s64 __a = he_get_raw_##_field(a);					\
+	s64 __b = he_get_raw_##_field(b);					\
+	return __a - __b;							\
+}
+
+
 #define HPP_PERCENT_FNS(_type, _str, _field, _min_width, _unit_width)	\
 __HPP_HEADER_FN(_type, _str, _min_width, _unit_width)			\
 __HPP_WIDTH_FN(_type, _min_width, _unit_width)				\
 __HPP_COLOR_PERCENT_FN(_type, _field)					\
-__HPP_ENTRY_PERCENT_FN(_type, _field)
+__HPP_ENTRY_PERCENT_FN(_type, _field)					\
+__HPP_SORT_FN(_type, _field)
 
 #define HPP_RAW_FNS(_type, _str, _field, _min_width, _unit_width)	\
 __HPP_HEADER_FN(_type, _str, _min_width, _unit_width)			\
 __HPP_WIDTH_FN(_type, _min_width, _unit_width)				\
-__HPP_ENTRY_RAW_FN(_type, _field)
+__HPP_ENTRY_RAW_FN(_type, _field)					\
+__HPP_SORT_RAW_FN(_type, _field)
 
 
 HPP_PERCENT_FNS(overhead, "Overhead", period, 8, 8)
@@ -227,19 +246,31 @@ HPP_PERCENT_FNS(overhead_guest_us, "guest usr", period_guest_us, 9, 8)
 HPP_RAW_FNS(samples, "Samples", nr_events, 12, 12)
 HPP_RAW_FNS(period, "Period", period, 12, 12)
 
+static int64_t hpp__nop_cmp(struct hist_entry *a __maybe_unused,
+			    struct hist_entry *b __maybe_unused)
+{
+	return 0;
+}
+
 #define HPP__COLOR_PRINT_FNS(_name)			\
 	{						\
 		.header	= hpp__header_ ## _name,	\
 		.width	= hpp__width_ ## _name,		\
 		.color	= hpp__color_ ## _name,		\
-		.entry	= hpp__entry_ ## _name		\
+		.entry	= hpp__entry_ ## _name,		\
+		.cmp	= hpp__nop_cmp,			\
+		.collapse = hpp__nop_cmp,		\
+		.sort	= hpp__sort_ ## _name,		\
 	}
 
 #define HPP__PRINT_FNS(_name)				\
 	{						\
 		.header	= hpp__header_ ## _name,	\
 		.width	= hpp__width_ ## _name,		\
-		.entry	= hpp__entry_ ## _name		\
+		.entry	= hpp__entry_ ## _name,		\
+		.cmp	= hpp__nop_cmp,			\
+		.collapse = hpp__nop_cmp,		\
+		.sort	= hpp__sort_ ## _name,		\
 	}
 
 struct perf_hpp_fmt perf_hpp__format[] = {
tools/perf/util/hist.h +3 -0
index 38c3e874c164..36dbe00e3cc8 100644
@@ -160,6 +160,9 @@ struct perf_hpp_fmt {
 		     struct hist_entry *he);
 	int (*entry)(struct perf_hpp_fmt *fmt, struct perf_hpp *hpp,
 		     struct hist_entry *he);
+	int64_t (*cmp)(struct hist_entry *a, struct hist_entry *b);
+	int64_t (*collapse)(struct hist_entry *a, struct hist_entry *b);
+	int64_t (*sort)(struct hist_entry *a, struct hist_entry *b);
 
 	struct list_head list;
 };
Signature · 1 line
1.9.2
author maintainer
#2 lore raw

PATCH 2/20 perf tools: Convert sort entries to hpp formats

In reply to Namhyung Kim

This is a preparation of consolidating management of output field and sort keys.

---
 tools/perf/ui/hist.c   |  6 ++++
 tools/perf/util/hist.h |  6 ++++
 tools/perf/util/sort.c | 80 +++++++++++++++++++++++++++++++++++++++++++++++---
 3 files changed, 88 insertions(+), 4 deletions(-)
tools/perf/ui/hist.c +6 -0
index d4a4f2e7eb43..a6eea666b443 100644
@@ -284,6 +284,7 @@ struct perf_hpp_fmt perf_hpp__format[] = {
 };
 
 LIST_HEAD(perf_hpp__list);
+LIST_HEAD(perf_hpp__sort_list);
 
 
 #undef HPP__COLOR_PRINT_FNS
@@ -325,6 +326,11 @@ void perf_hpp__column_register(struct perf_hpp_fmt *format)
 	list_add_tail(&format->list, &perf_hpp__list);
 }
 
+void perf_hpp__register_sort_field(struct perf_hpp_fmt *format)
+{
+	list_add_tail(&format->sort_list, &perf_hpp__sort_list);
+}
+
 void perf_hpp__column_enable(unsigned col)
 {
 	BUG_ON(col >= PERF_HPP__MAX_INDEX);
tools/perf/util/hist.h +6 -0
index 36dbe00e3cc8..eee154a41723 100644
@@ -165,13 +165,18 @@ struct perf_hpp_fmt {
 	int64_t (*sort)(struct hist_entry *a, struct hist_entry *b);
 
 	struct list_head list;
+	struct list_head sort_list;
 };
 
 extern struct list_head perf_hpp__list;
+extern struct list_head perf_hpp__sort_list;
 
 #define perf_hpp__for_each_format(format) \
 	list_for_each_entry(format, &perf_hpp__list, list)
 
+#define perf_hpp__for_each_sort_list(format) \
+	list_for_each_entry(format, &perf_hpp__sort_list, sort_list)
+
 extern struct perf_hpp_fmt perf_hpp__format[];
 
 enum {
@@ -190,6 +195,7 @@ enum {
 void perf_hpp__init(void);
 void perf_hpp__column_register(struct perf_hpp_fmt *format);
 void perf_hpp__column_enable(unsigned col);
+void perf_hpp__register_sort_field(struct perf_hpp_fmt *format);
 
 typedef u64 (*hpp_field_fn)(struct hist_entry *he);
 typedef int (*hpp_callback_fn)(struct perf_hpp *hpp, bool front);
tools/perf/util/sort.c +76 -4

106 lines not on this page · raw message

Signature · 1 line
1.9.2
author maintainer
#3 lore raw

PATCH 3/20 perf tools: Use hpp formats to sort hist entries

In reply to Namhyung Kim

It wrapped sort entries to hpp functions, so using the hpp sort list to sort entries.

---
 tools/perf/util/hist.c | 16 ++++++----------
 1 file changed, 6 insertions(+), 10 deletions(-)
tools/perf/util/hist.c +6 -10
index 7f0236cea4fe..38373c986e97 100644
@@ -432,11 +432,11 @@ struct hist_entry *__hists__add_entry(struct hists *hists,
 int64_t
 hist_entry__cmp(struct hist_entry *left, struct hist_entry *right)
 {
-	struct sort_entry *se;
+	struct perf_hpp_fmt *fmt;
 	int64_t cmp = 0;
 
-	list_for_each_entry(se, &hist_entry__sort_list, list) {
-		cmp = se->se_cmp(left, right);
+	perf_hpp__for_each_sort_list(fmt) {
+		cmp = fmt->cmp(left, right);
 		if (cmp)
 			break;
 	}
@@ -447,15 +447,11 @@ hist_entry__cmp(struct hist_entry *left, struct hist_entry *right)
 int64_t
 hist_entry__collapse(struct hist_entry *left, struct hist_entry *right)
 {
-	struct sort_entry *se;
+	struct perf_hpp_fmt *fmt;
 	int64_t cmp = 0;
 
-	list_for_each_entry(se, &hist_entry__sort_list, list) {
-		int64_t (*f)(struct hist_entry *, struct hist_entry *);
-
-		f = se->se_collapse ?: se->se_cmp;
-
-		cmp = f(left, right);
+	perf_hpp__for_each_sort_list(fmt) {
+		cmp = fmt->collapse(left, right);
 		if (cmp)
 			break;
 	}
Signature · 1 line
1.9.2
author maintainer
#4 lore raw

PATCH 4/20 perf tools: Support event grouping in hpp ->sort()

In reply to Namhyung Kim

Move logic of hist_entry__sort_on_period to __hpp__sort() in order to support event group report.

---
 tools/perf/ui/hist.c | 64 +++++++++++++++++++++++++++++++++++++++++++++++-----
 1 file changed, 58 insertions(+), 6 deletions(-)
tools/perf/ui/hist.c +58 -6
index a6eea666b443..c65a7fd744c6 100644
@@ -116,6 +116,62 @@ int __hpp__fmt(struct perf_hpp *hpp, struct hist_entry *he,
 	return ret;
 }
 
+static int field_cmp(u64 field_a, u64 field_b)
+{
+	if (field_a > field_b)
+		return 1;
+	if (field_a < field_b)
+		return -1;
+	return 0;
+}
+
+static int __hpp__sort(struct hist_entry *a, struct hist_entry *b,
+		       hpp_field_fn get_field)
+{
+	s64 ret;
+	int i, nr_members;
+	struct perf_evsel *evsel;
+	struct hist_entry *pair;
+	u64 *fields_a, *fields_b;
+
+	ret = field_cmp(get_field(a), get_field(b));
+	if (ret || !symbol_conf.event_group)
+		return ret;
+
+	evsel = hists_to_evsel(a->hists);
+	if (!perf_evsel__is_group_event(evsel))
+		return ret;
+
+	nr_members = evsel->nr_members;
+	fields_a = calloc(sizeof(*fields_a), nr_members);
+	fields_b = calloc(sizeof(*fields_b), nr_members);
+
+	if (!fields_a || !fields_b)
+		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 = fields_a[i] - fields_b[i];
+		if (ret)
+			break;
+	}
+
+out:
+	free(fields_a);
+	free(fields_b);
+
+	return ret;
+}
+
 #define __HPP_HEADER_FN(_type, _str, _min_width, _unit_width) 		\
 static int hpp__header_##_type(struct perf_hpp_fmt *fmt __maybe_unused,	\
 			       struct perf_hpp *hpp,			\
@@ -195,9 +251,7 @@ static int hpp__entry_##_type(struct perf_hpp_fmt *_fmt __maybe_unused,		\
 #define __HPP_SORT_FN(_type, _field)						\
 static int64_t hpp__sort_##_type(struct hist_entry *a, struct hist_entry *b)	\
 {										\
-	s64 __a = he_get_##_field(a);						\
-	s64 __b = he_get_##_field(b);						\
-	return __a - __b;							\
+	return __hpp__sort(a, b, he_get_##_field);				\
 }
 
 #define __HPP_ENTRY_RAW_FN(_type, _field)					\
@@ -217,9 +271,7 @@ static int hpp__entry_##_type(struct perf_hpp_fmt *_fmt __maybe_unused,		\
 #define __HPP_SORT_RAW_FN(_type, _field)					\
 static int64_t hpp__sort_##_type(struct hist_entry *a, struct hist_entry *b)	\
 {										\
-	s64 __a = he_get_raw_##_field(a);					\
-	s64 __b = he_get_raw_##_field(b);					\
-	return __a - __b;							\
+	return __hpp__sort(a, b, he_get_raw_##_field);				\
 }
 
 
Signature · 1 line
1.9.2
#5 lore raw

Re: [PATCH 04/20] perf tools: Support event grouping in hpp ->sort()

On Mon, May 12, 2014 at 03:28:37PM +0900, Namhyung Kim wrote:

SNIP

Quoted 26 lines
tools/perf/ui/hist.c +26 -0
+static int __hpp__sort(struct hist_entry *a, struct hist_entry *b,
+		       hpp_field_fn get_field)
+{
+	s64 ret;
+	int i, nr_members;
+	struct perf_evsel *evsel;
+	struct hist_entry *pair;
+	u64 *fields_a, *fields_b;
+
+	ret = field_cmp(get_field(a), get_field(b));
+	if (ret || !symbol_conf.event_group)
+		return ret;
+
+	evsel = hists_to_evsel(a->hists);
+	if (!perf_evsel__is_group_event(evsel))
+		return ret;
+
+	nr_members = evsel->nr_members;
+	fields_a = calloc(sizeof(*fields_a), nr_members);
+	fields_b = calloc(sizeof(*fields_b), nr_members);
+
+	if (!fields_a || !fields_b)
+		goto out;
+
+	list_for_each_entry(pair, &a->pairs.head, pairs.node) {
+		evsel = hists_to_evsel(pair->hists);

tools/perf/ui/hist.c int __hpp__fmt(struct perf_hpp *hpp, struct hist_entry *he, · line 163

Quoted 10 lines
tools/perf/ui/hist.c +10 -0
+		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 = fields_a[i] - fields_b[i];

should we call here ^^^call field_cmp ?

jirka

#6 lore raw

Re: [PATCH 04/20] perf tools: Support event grouping in hpp ->sort()

Hi Jiri,

Quoted 34 lines · Jiri Olsa

On Mon, May 12, 2014 at 03:28:37PM +0900, Namhyung Kim wrote:

SNIP

Quoted 30 lines
tools/perf/ui/hist.c +30 -0
+static int __hpp__sort(struct hist_entry *a, struct hist_entry *b,
+		       hpp_field_fn get_field)
+{
+	s64 ret;
+	int i, nr_members;
+	struct perf_evsel *evsel;
+	struct hist_entry *pair;
+	u64 *fields_a, *fields_b;
+
+	ret = field_cmp(get_field(a), get_field(b));
+	if (ret || !symbol_conf.event_group)
+		return ret;
+
+	evsel = hists_to_evsel(a->hists);
+	if (!perf_evsel__is_group_event(evsel))
+		return ret;
+
+	nr_members = evsel->nr_members;
+	fields_a = calloc(sizeof(*fields_a), nr_members);
+	fields_b = calloc(sizeof(*fields_b), nr_members);
+
+	if (!fields_a || !fields_b)
+		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) {
Quoted 8 lines

tools/perf/ui/hist.c int __hpp__fmt(struct perf_hpp *hpp, struct hist_entry *he, · line 163

Quoted 6 lines
tools/perf/ui/hist.c +6 -0
+		evsel = hists_to_evsel(pair->hists);
+		fields_b[perf_evsel__group_idx(evsel)] = get_field(pair);
+	}
+
+	for (i = 1; i < nr_members; i++) {
+		ret = fields_a[i] - fields_b[i];

should we call here ^^^call field_cmp ?

Right! Will fix.

Thanks,
Namhyung

author maintainer
#7 lore raw

PATCH 5/20 perf tools: Use hpp formats to sort final output

In reply to Namhyung Kim

Convert output sorting function to use ->sort hpp functions.

---
 tools/perf/util/hist.c | 62 +++++++-------------------------------------------
 1 file changed, 8 insertions(+), 54 deletions(-)
tools/perf/util/hist.c +8 -54
index 38373c986e97..c99ae4dd973e 100644
@@ -564,64 +564,18 @@ void hists__collapse_resort(struct hists *hists, struct ui_progress *prog)
 	}
 }
 
-/*
- * reverse the map, sort on period.
- */
-
-static int period_cmp(u64 period_a, u64 period_b)
+static int hist_entry__sort(struct hist_entry *a, struct hist_entry *b)
 {
-	if (period_a > period_b)
-		return 1;
-	if (period_a < period_b)
-		return -1;
-	return 0;
-}
-
-static int hist_entry__sort_on_period(struct hist_entry *a,
-				      struct hist_entry *b)
-{
-	int ret;
-	int i, nr_members;
-	struct perf_evsel *evsel;
-	struct hist_entry *pair;
-	u64 *periods_a, *periods_b;
-
-	ret = period_cmp(a->stat.period, b->stat.period);
-	if (ret || !symbol_conf.event_group)
-		return ret;
-
-	evsel = hists_to_evsel(a->hists);
-	nr_members = evsel->nr_members;
-	if (nr_members <= 1)
-		return ret;
-
-	periods_a = zalloc(sizeof(periods_a) * nr_members);
-	periods_b = zalloc(sizeof(periods_b) * nr_members);
-
-	if (!periods_a || !periods_b)
-		goto out;
-
-	list_for_each_entry(pair, &a->pairs.head, pairs.node) {
-		evsel = hists_to_evsel(pair->hists);
-		periods_a[perf_evsel__group_idx(evsel)] = pair->stat.period;
-	}
-
-	list_for_each_entry(pair, &b->pairs.head, pairs.node) {
-		evsel = hists_to_evsel(pair->hists);
-		periods_b[perf_evsel__group_idx(evsel)] = pair->stat.period;
-	}
+	struct perf_hpp_fmt *fmt;
+	int64_t cmp = 0;
 
-	for (i = 1; i < nr_members; i++) {
-		ret = period_cmp(periods_a[i], periods_b[i]);
-		if (ret)
+	perf_hpp__for_each_format(fmt) {
+		cmp = fmt->sort(a, b);
+		if (cmp)
 			break;
 	}
 
-out:
-	free(periods_a);
-	free(periods_b);
-
-	return ret;
+	return cmp;
 }
 
 static void hists__reset_filter_stats(struct hists *hists)
@@ -669,7 +623,7 @@ static void __hists__insert_output_entry(struct rb_root *entries,
 		parent = *p;
 		iter = rb_entry(parent, struct hist_entry, rb_node);
 
-		if (hist_entry__sort_on_period(he, iter) > 0)
+		if (hist_entry__sort(he, iter) > 0)
 			p = &(*p)->rb_left;
 		else
 			p = &(*p)->rb_right;
Signature · 1 line
1.9.2
author maintainer
#8 lore raw

PATCH 6/20 perf tools: Consolidate output field handling to hpp format routines

In reply to Namhyung Kim

Until now the hpp and sort functions do similar jobs different ways. Since the sort functions converted/wrapped to hpp formats it can do the job in a uniform way.

The perf_hpp__sort_list has a list of hpp formats to sort entries and the perf_hpp__list has a list of hpp formats to print output result.

To have a backward compatiblity, it automatically adds 'overhead' field in front of sort list. And then all of fields in sort list added to the output list (if it's not already there).

---
 tools/perf/ui/browsers/hists.c |  4 ++--
 tools/perf/ui/gtk/hists.c      | 31 -------------------------
 tools/perf/ui/hist.c           | 28 +++++++++++++++++++++++
 tools/perf/ui/stdio/hist.c     | 52 +++++++++++++-----------------------------
 tools/perf/util/hist.c         |  2 +-
 5 files changed, 47 insertions(+), 70 deletions(-)
tools/perf/ui/browsers/hists.c +2 -2
index b0861e3e50a5..7bd8c0e81658 100644
@@ -760,8 +760,8 @@ static int hist_browser__show_entry(struct hist_browser *browser,
 		if (!browser->b.navkeypressed)
 			width += 1;
 
-		hist_entry__sort_snprintf(entry, s, sizeof(s), browser->hists);
-		slsmg_write_nstring(s, width);
+		slsmg_write_nstring("", width);
+
 		++row;
 		++printed;
 	} else
tools/perf/ui/gtk/hists.c +0 -31
index 91f10f3f6dd1..d5c336e1bb14 100644
@@ -153,7 +153,6 @@ static void perf_gtk__show_hists(GtkWidget *window, struct hists *hists,
 	struct perf_hpp_fmt *fmt;
 	GType col_types[MAX_COLUMNS];
 	GtkCellRenderer *renderer;
-	struct sort_entry *se;
 	GtkTreeStore *store;
 	struct rb_node *nd;
 	GtkWidget *view;
@@ -172,16 +171,6 @@ static void perf_gtk__show_hists(GtkWidget *window, struct hists *hists,
 	perf_hpp__for_each_format(fmt)
 		col_types[nr_cols++] = G_TYPE_STRING;
 
-	list_for_each_entry(se, &hist_entry__sort_list, list) {
-		if (se->elide)
-			continue;
-
-		if (se == &sort_sym)
-			sym_col = nr_cols;
-
-		col_types[nr_cols++] = G_TYPE_STRING;
-	}
-
 	store = gtk_tree_store_newv(nr_cols, col_types);
 
 	view = gtk_tree_view_new();
@@ -199,16 +188,6 @@ static void perf_gtk__show_hists(GtkWidget *window, struct hists *hists,
 							    col_idx++, NULL);
 	}
 
-	list_for_each_entry(se, &hist_entry__sort_list, list) {
-		if (se->elide)
-			continue;
-
-		gtk_tree_view_insert_column_with_attributes(GTK_TREE_VIEW(view),
-							    -1, se->se_header,
-							    renderer, "text",
-							    col_idx++, NULL);
-	}
-
 	for (col_idx = 0; col_idx < nr_cols; col_idx++) {
 		GtkTreeViewColumn *column;
 
@@ -253,16 +232,6 @@ static void perf_gtk__show_hists(GtkWidget *window, struct hists *hists,
 			gtk_tree_store_set(store, &iter, col_idx++, s, -1);
 		}
 
-		list_for_each_entry(se, &hist_entry__sort_list, list) {
-			if (se->elide)
-				continue;
-
-			se->se_snprintf(h, s, ARRAY_SIZE(s),
-					hists__col_len(hists, se->se_width_idx));
-
-			gtk_tree_store_set(store, &iter, col_idx++, s, -1);
-		}
-
 		if (symbol_conf.use_callchain && sort__has_sym) {
 			if (callchain_param.mode == CHAIN_GRAPH_REL)
 				total = h->stat.period;
tools/perf/ui/hist.c +28 -0
index c65a7fd744c6..32d2dfd794d9 100644
@@ -352,8 +352,18 @@ LIST_HEAD(perf_hpp__sort_list);
 #undef __HPP_ENTRY_RAW_FN
 
 
+void perf_hpp__setup_output_field(void);
+
 void perf_hpp__init(void)
 {
+	struct list_head *list;
+	int i;
+
+	for (i = 0; i < PERF_HPP__MAX_INDEX; i++) {
+		INIT_LIST_HEAD(&perf_hpp__format[i].list);
+		INIT_LIST_HEAD(&perf_hpp__format[i].sort_list);
+	}
+
 	perf_hpp__column_enable(PERF_HPP__OVERHEAD);
 
 	if (symbol_conf.show_cpu_utilization) {
@@ -371,6 +381,13 @@ void perf_hpp__init(void)
 
 	if (symbol_conf.show_total_period)
 		perf_hpp__column_enable(PERF_HPP__PERIOD);
+
+	/* prepend overhead field for backward compatiblity.  */
+	list = &perf_hpp__format[PERF_HPP__OVERHEAD].sort_list;
+	if (list_empty(list))
+		list_add(list, &perf_hpp__sort_list);
+
+	perf_hpp__setup_output_field();
 }
 
 void perf_hpp__column_register(struct perf_hpp_fmt *format)
@@ -389,6 +406,17 @@ void perf_hpp__column_enable(unsigned col)
 	perf_hpp__column_register(&perf_hpp__format[col]);
 }
 
+void perf_hpp__setup_output_field(void)
+{
+	struct perf_hpp_fmt *fmt;
+
+	/* append sort keys to output field */
+	perf_hpp__for_each_sort_list(fmt) {
+		if (list_empty(&fmt->list))
+			perf_hpp__column_register(fmt);
+	}
+}
+
 int hist_entry__sort_snprintf(struct hist_entry *he, char *s, size_t size,
 			      struct hists *hists)
 {
tools/perf/ui/stdio/hist.c +16 -36

79 lines not on this page · raw message

tools/perf/util/hist.c +1 -1
index c99ae4dd973e..ae13c2dbd27a 100644
@@ -569,7 +569,7 @@ static int hist_entry__sort(struct hist_entry *a, struct hist_entry *b)
 	struct perf_hpp_fmt *fmt;
 	int64_t cmp = 0;
 
-	perf_hpp__for_each_format(fmt) {
+	perf_hpp__for_each_sort_list(fmt) {
 		cmp = fmt->sort(a, b);
 		if (cmp)
 			break;
Signature · 1 line
1.9.2
#9 lore raw

Re: [PATCH 06/20] perf tools: Consolidate output field handling to hpp format routines

On Mon, May 12, 2014 at 03:28:39PM +0900, Namhyung Kim wrote:

SNIP

Quoted 10 lines
tools/perf/ui/gtk/hists.c +0 -2
-		}
-
 		if (symbol_conf.use_callchain && sort__has_sym) {
 			if (callchain_param.mode == CHAIN_GRAPH_REL)
 				total = h->stat.period;
tools/perf/ui/hist.c +0 -0
index c65a7fd744c6..32d2dfd794d9 100644

tools/perf/ui/hist.c LIST_HEAD(perf_hpp__sort_list); · line 355

Quoted 4 lines
tools/perf/ui/hist.c +1 -0
@@ -352,8 +352,18 @@ LIST_HEAD(perf_hpp__sort_list);
 #undef __HPP_ENTRY_RAW_FN
 
 
+void perf_hpp__setup_output_field(void);

should go to util/hist.h

jirka

tools/perf/ui/stdio/hist.c static int hist_entry__fprintf(struct hist_entry *he, size_t size, · line 356

Quoted 12 lines
tools/perf/ui/hist.c +0 -0
 {
tools/perf/ui/stdio/hist.c +1 -2
index 9eccf7f4f367..e6920d124c60 100644
@@ -353,8 +353,7 @@ static int hist_entry__fprintf(struct hist_entry *he, size_t size,
 	if (size == 0 || size > bfsz)
 		size = hpp.size = bfsz;
 
-	ret = hist_entry__period_snprintf(&hpp, he);
-	hist_entry__sort_snprintf(he, bf + ret, size - ret, hists);
+	hist_entry__period_snprintf(&hpp, he);

so it's no longer just period.. should we rename it? like
hist_entry__hpp_snprintf
hist_entry__snprintf

jirka

#10 lore raw

Re: [PATCH 06/20] perf tools: Consolidate output field handling to hpp format routines

Quoted 17 lines · Jiri Olsa

On Mon, May 12, 2014 at 03:28:39PM +0900, Namhyung Kim wrote:

SNIP

Quoted 13 lines
tools/perf/ui/gtk/hists.c +0 -2
-		}
-
 		if (symbol_conf.use_callchain && sort__has_sym) {
 			if (callchain_param.mode == CHAIN_GRAPH_REL)
 				total = h->stat.period;
tools/perf/ui/hist.c +0 -0
index c65a7fd744c6..32d2dfd794d9 100644
@@ -352,8 +352,18 @@ LIST_HEAD(perf_hpp__sort_list);
 #undef __HPP_ENTRY_RAW_FN
 
 
Quoted 3 lines

tools/perf/ui/hist.c LIST_HEAD(perf_hpp__sort_list); · line 355

Quoted 1 line
tools/perf/ui/hist.c +1 -0
+void perf_hpp__setup_output_field(void);

should go to util/hist.h

Okay.

Quoted 10 lines
Quoted 10 lines
tools/perf/ui/hist.c +0 -0
 {
tools/perf/ui/stdio/hist.c +0 -0
index 9eccf7f4f367..e6920d124c60 100644
@@ -353,8 +353,7 @@ static int hist_entry__fprintf(struct hist_entry *he, size_t size,
 	if (size == 0 || size > bfsz)
 		size = hpp.size = bfsz;
 
Quoted 7 lines

tools/perf/ui/stdio/hist.c static int hist_entry__fprintf(struct hist_entry *he, size_t size, · line 356

Quoted 3 lines
tools/perf/ui/stdio/hist.c +1 -2
-	ret = hist_entry__period_snprintf(&hpp, he);
-	hist_entry__sort_snprintf(he, bf + ret, size - ret, hists);
+	hist_entry__period_snprintf(&hpp, he);

so it's no longer just period.. should we rename it? like
hist_entry__hpp_snprintf
hist_entry__snprintf

The hist_entry__snprintf() looks better to me.

Thanks,
Namhyung

#11 lore raw

Re: [PATCH 06/20] perf tools: Consolidate output field handling to hpp format routines

In reply to Namhyung Kim

On Mon, May 12, 2014 at 03:28:39PM +0900, Namhyung Kim wrote:

SNIP

tools/perf/ui/stdio/hist.c static int hist_entry__fprintf(struct hist_entry *he, size_t size, · line 356

Quoted 11 lines
tools/perf/ui/stdio/hist.c +1 -2
index 9eccf7f4f367..e6920d124c60 100644
@@ -353,8 +353,7 @@ static int hist_entry__fprintf(struct hist_entry *he, size_t size,
 	if (size == 0 || size > bfsz)
 		size = hpp.size = bfsz;
 
-	ret = hist_entry__period_snprintf(&hpp, he);
-	hist_entry__sort_snprintf(he, bf + ret, size - ret, hists);
+	hist_entry__period_snprintf(&hpp, he);

hm, how about hist_browser__fprintf_entry.. looks like it could use just hist_entry__period_snprintf as well and we could get rid hist_entry__sort_snprintf completely

thanks,
jirka

#12 lore raw

Re: [PATCH 06/20] perf tools: Consolidate output field handling to hpp format routines

Quoted 12 lines · Jiri Olsa

On Mon, May 12, 2014 at 03:28:39PM +0900, Namhyung Kim wrote:

SNIP

Quoted 8 lines
tools/perf/ui/stdio/hist.c +0 -0
index 9eccf7f4f367..e6920d124c60 100644
@@ -353,8 +353,7 @@ static int hist_entry__fprintf(struct hist_entry *he, size_t size,
 	if (size == 0 || size > bfsz)
 		size = hpp.size = bfsz;
 
Quoted 5 lines

tools/perf/ui/stdio/hist.c static int hist_entry__fprintf(struct hist_entry *he, size_t size, · line 356

Quoted 3 lines
tools/perf/ui/stdio/hist.c +1 -2
-	ret = hist_entry__period_snprintf(&hpp, he);
-	hist_entry__sort_snprintf(he, bf + ret, size - ret, hists);
+	hist_entry__period_snprintf(&hpp, he);

hm, how about hist_browser__fprintf_entry.. looks like it could

The term 'hist_browser' is for some browsable UI windows like in TUI or GUI IMHO. So I think simply call it as hist_entry__snprintf() is better as you originally suggested.

Quoted 2 lines

use just hist_entry__period_snprintf as well and we could get rid hist_entry__sort_snprintf completely

Hmm.. right. It'll be gone eventually.

Thanks,
Namhyung

author maintainer
#13 lore raw

PATCH 7/20 perf ui: Get rid of callback from __hpp__fmt()

In reply to Namhyung Kim

The callback was used by TUI for determining color of folded sign using percent of first field/column. But it cannot be used anymore since it now support dynamic reording of output field.

So move the logic to the hist_browser__show_entry().

---
 tools/perf/ui/browsers/hists.c | 62 ++++++++++++++++--------------------------
 tools/perf/ui/gtk/hists.c      |  2 +-
 tools/perf/ui/hist.c           | 28 ++++++-------------
 tools/perf/util/hist.h         |  4 +--
 4 files changed, 34 insertions(+), 62 deletions(-)
tools/perf/ui/browsers/hists.c +23 -39
index 7bd8c0e81658..3ed9212d2a63 100644
@@ -616,35 +616,6 @@ struct hpp_arg {
 	bool current_entry;
 };
 
-static int __hpp__overhead_callback(struct perf_hpp *hpp, bool front)
-{
-	struct hpp_arg *arg = hpp->ptr;
-
-	if (arg->current_entry && arg->b->navkeypressed)
-		ui_browser__set_color(arg->b, HE_COLORSET_SELECTED);
-	else
-		ui_browser__set_color(arg->b, HE_COLORSET_NORMAL);
-
-	if (front) {
-		if (!symbol_conf.use_callchain)
-			return 0;
-
-		slsmg_printf("%c ", arg->folded_sign);
-		return 2;
-	}
-
-	return 0;
-}
-
-static int __hpp__color_callback(struct perf_hpp *hpp, bool front __maybe_unused)
-{
-	struct hpp_arg *arg = hpp->ptr;
-
-	if (!arg->current_entry || !arg->b->navkeypressed)
-		ui_browser__set_color(arg->b, HE_COLORSET_NORMAL);
-	return 0;
-}
-
 static int __hpp__slsmg_color_printf(struct perf_hpp *hpp, const char *fmt, ...)
 {
 	struct hpp_arg *arg = hpp->ptr;
@@ -665,7 +636,7 @@ static int __hpp__slsmg_color_printf(struct perf_hpp *hpp, const char *fmt, ...)
 	return ret;
 }
 
-#define __HPP_COLOR_PERCENT_FN(_type, _field, _cb)			\
+#define __HPP_COLOR_PERCENT_FN(_type, _field)				\
 static u64 __hpp_get_##_field(struct hist_entry *he)			\
 {									\
 	return he->stat._field;						\
@@ -676,15 +647,15 @@ hist_browser__hpp_color_##_type(struct perf_hpp_fmt *fmt __maybe_unused,\
 				struct perf_hpp *hpp,			\
 				struct hist_entry *he)			\
 {									\
-	return __hpp__fmt(hpp, he, __hpp_get_##_field, _cb, " %6.2f%%",	\
+	return __hpp__fmt(hpp, he, __hpp_get_##_field, " %6.2f%%",	\
 			  __hpp__slsmg_color_printf, true);		\
 }
 
-__HPP_COLOR_PERCENT_FN(overhead, period, __hpp__overhead_callback)
-__HPP_COLOR_PERCENT_FN(overhead_sys, period_sys, __hpp__color_callback)
-__HPP_COLOR_PERCENT_FN(overhead_us, period_us, __hpp__color_callback)
-__HPP_COLOR_PERCENT_FN(overhead_guest_sys, period_guest_sys, __hpp__color_callback)
-__HPP_COLOR_PERCENT_FN(overhead_guest_us, period_guest_us, __hpp__color_callback)
+__HPP_COLOR_PERCENT_FN(overhead, period)
+__HPP_COLOR_PERCENT_FN(overhead_sys, period_sys)
+__HPP_COLOR_PERCENT_FN(overhead_us, period_us)
+__HPP_COLOR_PERCENT_FN(overhead_guest_sys, period_guest_sys)
+__HPP_COLOR_PERCENT_FN(overhead_guest_us, period_guest_us)
 
 #undef __HPP_COLOR_PERCENT_FN
 
@@ -729,7 +700,7 @@ static int hist_browser__show_entry(struct hist_browser *browser,
 
 	if (row_offset == 0) {
 		struct hpp_arg arg = {
-			.b 		= &browser->b,
+			.b		= &browser->b,
 			.folded_sign	= folded_sign,
 			.current_entry	= current_entry,
 		};
@@ -742,11 +713,24 @@ static int hist_browser__show_entry(struct hist_browser *browser,
 		ui_browser__gotorc(&browser->b, row, 0);
 
 		perf_hpp__for_each_format(fmt) {
-			if (!first) {
+			if (current_entry && browser->b.navkeypressed) {
+				ui_browser__set_color(&browser->b,
+						      HE_COLORSET_SELECTED);
+			} else {
+				ui_browser__set_color(&browser->b,
+						      HE_COLORSET_NORMAL);
+			}
+
+			if (first) {
+				if (symbol_conf.use_callchain) {
+					slsmg_printf("%c ", folded_sign);
+					width -= 2;
+				}
+				first = false;
+			} else {
 				slsmg_printf("  ");
 				width -= 2;
 			}
-			first = false;
 
 			if (fmt->color) {
 				width -= fmt->color(fmt, &hpp, entry);
tools/perf/ui/gtk/hists.c +1 -1
index d5c336e1bb14..2237245bfac0 100644
@@ -43,7 +43,7 @@ static int perf_gtk__hpp_color_##_type(struct perf_hpp_fmt *fmt __maybe_unused,
 				       struct perf_hpp *hpp,			\
 				       struct hist_entry *he)			\
 {										\
-	return __hpp__fmt(hpp, he, he_get_##_field, NULL, " %6.2f%%",		\
+	return __hpp__fmt(hpp, he, he_get_##_field, " %6.2f%%",			\
 			  __percent_color_snprintf, true);			\
 }
 
tools/perf/ui/hist.c +8 -20
index 32d2dfd794d9..400dad8c41e4 100644
@@ -16,20 +16,15 @@
 })
 
 int __hpp__fmt(struct perf_hpp *hpp, struct hist_entry *he,
-	       hpp_field_fn get_field, hpp_callback_fn callback,
-	       const char *fmt, hpp_snprint_fn print_fn, bool fmt_percent)
+	       hpp_field_fn get_field, const char *fmt,
+	       hpp_snprint_fn print_fn, bool fmt_percent)
 {
-	int ret = 0;
+	int ret;
 	struct hists *hists = he->hists;
 	struct perf_evsel *evsel = hists_to_evsel(hists);
 	char *buf = hpp->buf;
 	size_t size = hpp->size;
 
-	if (callback) {
-		ret = callback(hpp, true);
-		advance_hpp(hpp, ret);
-	}
-
 	if (fmt_percent) {
 		double percent = 0.0;
 		u64 total = hists__total_period(hists);
@@ -37,9 +32,9 @@ int __hpp__fmt(struct perf_hpp *hpp, struct hist_entry *he,
 		if (total)
 			percent = 100.0 * get_field(he) / total;
 
-		ret += hpp__call_print_fn(hpp, print_fn, fmt, percent);
+		ret = hpp__call_print_fn(hpp, print_fn, fmt, percent);
 	} else
-		ret += hpp__call_print_fn(hpp, print_fn, fmt, get_field(he));
+		ret = hpp__call_print_fn(hpp, print_fn, fmt, get_field(he));
 
 	if (perf_evsel__is_group_event(evsel)) {
 		int prev_idx, idx_delta;
@@ -99,13 +94,6 @@ int __hpp__fmt(struct perf_hpp *hpp, struct hist_entry *he,
 		}
 	}
 
-	if (callback) {
-		int __ret = callback(hpp, false);
-
-		advance_hpp(hpp, __ret);
-		ret += __ret;
-	}
-
 	/*
 	 * Restore original buf and size as it's where caller expects
 	 * the result will be saved.
@@ -235,7 +223,7 @@ static u64 he_get_##_field(struct hist_entry *he)				\
 static int hpp__color_##_type(struct perf_hpp_fmt *fmt __maybe_unused,		\
 			      struct perf_hpp *hpp, struct hist_entry *he) 	\
 {										\
-	return __hpp__fmt(hpp, he, he_get_##_field, NULL, " %6.2f%%",		\
+	return __hpp__fmt(hpp, he, he_get_##_field, " %6.2f%%",			\
 			  hpp_color_scnprintf, true);				\
 }
 
@@ -244,7 +232,7 @@ static int hpp__entry_##_type(struct perf_hpp_fmt *_fmt __maybe_unused,		\
 			      struct perf_hpp *hpp, struct hist_entry *he) 	\
 {										\
 	const char *fmt = symbol_conf.field_sep ? " %.2f" : " %6.2f%%";		\
-	return __hpp__fmt(hpp, he, he_get_##_field, NULL, fmt,			\
+	return __hpp__fmt(hpp, he, he_get_##_field, fmt,			\
 			  hpp_entry_scnprintf, true);				\
 }
 
@@ -264,7 +252,7 @@ static int hpp__entry_##_type(struct perf_hpp_fmt *_fmt __maybe_unused,		\
 			      struct perf_hpp *hpp, struct hist_entry *he) 	\
 {										\
 	const char *fmt = symbol_conf.field_sep ? " %"PRIu64 : " %11"PRIu64;	\
-	return __hpp__fmt(hpp, he, he_get_raw_##_field, NULL, fmt,		\
+	return __hpp__fmt(hpp, he, he_get_raw_##_field, fmt,			\
 			  hpp_entry_scnprintf, false);				\
 }
 
tools/perf/util/hist.h +2 -2
index eee154a41723..78855d373381 100644
@@ -202,8 +202,8 @@ typedef int (*hpp_callback_fn)(struct perf_hpp *hpp, bool front);
 typedef int (*hpp_snprint_fn)(struct perf_hpp *hpp, const char *fmt, ...);
 
 int __hpp__fmt(struct perf_hpp *hpp, struct hist_entry *he,
-	       hpp_field_fn get_field, hpp_callback_fn callback,
-	       const char *fmt, hpp_snprint_fn print_fn, bool fmt_percent);
+	       hpp_field_fn get_field, const char *fmt,
+	       hpp_snprint_fn print_fn, bool fmt_percent);
 
 static inline void advance_hpp(struct perf_hpp *hpp, int inc)
 {
Signature · 1 line
1.9.2
author maintainer
#14 lore raw

PATCH 8/20 perf tools: Allow hpp fields to be sort keys

In reply to Namhyung Kim

Add overhead{,_sys,_us,_guest_sys,_guest_us}, sample and period sort keys so that they can be selected with --sort/-s option.

  $ perf report -s period,comm --stdio
  ...
  # Overhead        Period          Command
  # ........  ............  ...............
  #
      47.06%           152          swapper
      13.93%            45  qemu-system-arm
      12.38%            40         synergys
       3.72%            12          firefox
       2.48%             8            xchat
---
 tools/perf/ui/hist.c   |  9 +++++++--
 tools/perf/util/sort.c | 39 +++++++++++++++++++++++++++++++++++++++
 2 files changed, 46 insertions(+), 2 deletions(-)
tools/perf/ui/hist.c +7 -2
index 400dad8c41e4..f3e96463550b 100644
@@ -348,8 +348,13 @@ void perf_hpp__init(void)
 	int i;
 
 	for (i = 0; i < PERF_HPP__MAX_INDEX; i++) {
-		INIT_LIST_HEAD(&perf_hpp__format[i].list);
-		INIT_LIST_HEAD(&perf_hpp__format[i].sort_list);
+		struct perf_hpp_fmt *fmt = &perf_hpp__format[i];
+
+		INIT_LIST_HEAD(&fmt->list);
+
+		/* sort_list may be linked by setup_sorting() */
+		if (fmt->sort_list.next == NULL)
+			INIT_LIST_HEAD(&fmt->sort_list);
 	}
 
 	perf_hpp__column_enable(PERF_HPP__OVERHEAD);
tools/perf/util/sort.c +39 -0
index b2829f947053..916652af8304 100644
@@ -1028,6 +1028,26 @@ static struct sort_dimension memory_sort_dimensions[] = {
 
 #undef DIM
 
+struct hpp_dimension {
+	const char		*name;
+	struct perf_hpp_fmt	*fmt;
+	int			taken;
+};
+
+#define DIM(d, n) { .name = n, .fmt = &perf_hpp__format[d], }
+
+static struct hpp_dimension hpp_sort_dimensions[] = {
+	DIM(PERF_HPP__OVERHEAD, "overhead"),
+	DIM(PERF_HPP__OVERHEAD_SYS, "overhead_sys"),
+	DIM(PERF_HPP__OVERHEAD_US, "overhead_us"),
+	DIM(PERF_HPP__OVERHEAD_GUEST_SYS, "overhead_guest_sys"),
+	DIM(PERF_HPP__OVERHEAD_GUEST_US, "overhead_guest_us"),
+	DIM(PERF_HPP__SAMPLES, "sample"),
+	DIM(PERF_HPP__PERIOD, "period"),
+};
+
+#undef DIM
+
 struct hpp_sort_entry {
 	struct perf_hpp_fmt hpp;
 	struct sort_entry *se;
@@ -1115,6 +1135,16 @@ static int __sort_dimension__add(struct sort_dimension *sd, enum sort_type idx)
 	return 0;
 }
 
+static int __hpp_dimension__add(struct hpp_dimension *hd)
+{
+	if (!hd->taken) {
+		hd->taken = 1;
+
+		perf_hpp__register_sort_field(hd->fmt);
+	}
+	return 0;
+}
+
 int sort_dimension__add(const char *tok)
 {
 	unsigned int i;
@@ -1144,6 +1174,15 @@ int sort_dimension__add(const char *tok)
 		return __sort_dimension__add(sd, i);
 	}
 
+	for (i = 0; i < ARRAY_SIZE(hpp_sort_dimensions); i++) {
+		struct hpp_dimension *hd = &hpp_sort_dimensions[i];
+
+		if (strncasecmp(tok, hd->name, strlen(tok)))
+			continue;
+
+		return __hpp_dimension__add(hd);
+	}
+
 	for (i = 0; i < ARRAY_SIZE(bstack_sort_dimensions); i++) {
 		struct sort_dimension *sd = &bstack_sort_dimensions[i];
 
Signature · 1 line
1.9.2
#15 lore raw

Re: [PATCH 08/20] perf tools: Allow hpp fields to be sort keys

On Mon, May 12, 2014 at 03:28:41PM +0900, Namhyung Kim wrote:

SNIP

tools/perf/util/sort.c static struct sort_dimension memory_sort_dimensions[] = { · line 1047

Quoted 9 lines
tools/perf/util/sort.c +9 -0
+static struct hpp_dimension hpp_sort_dimensions[] = {
+	DIM(PERF_HPP__OVERHEAD, "overhead"),
+	DIM(PERF_HPP__OVERHEAD_SYS, "overhead_sys"),
+	DIM(PERF_HPP__OVERHEAD_US, "overhead_us"),
+	DIM(PERF_HPP__OVERHEAD_GUEST_SYS, "overhead_guest_sys"),
+	DIM(PERF_HPP__OVERHEAD_GUEST_US, "overhead_guest_us"),
+	DIM(PERF_HPP__SAMPLES, "sample"),
+	DIM(PERF_HPP__PERIOD, "period"),
+};

just a nit.. overhead_guest_us|sys misplace are not properly aligned

[jolsa@krava perf]$ ./perf report -s overhead_guest_us,period,dso,sym --stdio # To display the perf.data header info, please use --header/--header-only options. #
# Samples: 9 of event 'cycles'
# Event count (approx.): 3808201
#
# Overhead guest usr Period Shared Object Symbol

# ........  .........  ............  .................  ..........................
#
    43.05%     0.00%       1639423  ld-2.17.so         [.] _dl_map_object_deps   
    36.86%     0.00%       1403626  [kernel.kallsyms]  [k] _raw_spin_lock_irqsave

....

jirka

#16 lore raw

Re: [PATCH 08/20] perf tools: Allow hpp fields to be sort keys

Quoted 4 lines · Jiri Olsa

On Mon, May 12, 2014 at 03:28:41PM +0900, Namhyung Kim wrote:

SNIP

Quoted 11 lines

tools/perf/util/sort.c static struct sort_dimension memory_sort_dimensions[] = { · line 1047

Quoted 9 lines
tools/perf/util/sort.c +9 -0
+static struct hpp_dimension hpp_sort_dimensions[] = {
+	DIM(PERF_HPP__OVERHEAD, "overhead"),
+	DIM(PERF_HPP__OVERHEAD_SYS, "overhead_sys"),
+	DIM(PERF_HPP__OVERHEAD_US, "overhead_us"),
+	DIM(PERF_HPP__OVERHEAD_GUEST_SYS, "overhead_guest_sys"),
+	DIM(PERF_HPP__OVERHEAD_GUEST_US, "overhead_guest_us"),
+	DIM(PERF_HPP__SAMPLES, "sample"),
+	DIM(PERF_HPP__PERIOD, "period"),
+};

just a nit.. overhead_guest_us|sys misplace are not properly aligned

Okay. Will fix.

Thanks,
Namhyung

author maintainer
#17 lore raw

Re: [PATCH 08/20] perf tools: Allow hpp fields to be sort keys

Hi Jiri,

Quoted 5 lines · Namhyung Kim
Quoted 4 lines · Jiri Olsa

On Mon, May 12, 2014 at 03:28:41PM +0900, Namhyung Kim wrote:

SNIP

Quoted 11 lines
Quoted 11 lines

tools/perf/util/sort.c static struct sort_dimension memory_sort_dimensions[] = { · line 1047

Quoted 9 lines
tools/perf/util/sort.c +9 -0
+static struct hpp_dimension hpp_sort_dimensions[] = {
+	DIM(PERF_HPP__OVERHEAD, "overhead"),
+	DIM(PERF_HPP__OVERHEAD_SYS, "overhead_sys"),
+	DIM(PERF_HPP__OVERHEAD_US, "overhead_us"),
+	DIM(PERF_HPP__OVERHEAD_GUEST_SYS, "overhead_guest_sys"),
+	DIM(PERF_HPP__OVERHEAD_GUEST_US, "overhead_guest_us"),
+	DIM(PERF_HPP__SAMPLES, "sample"),
+	DIM(PERF_HPP__PERIOD, "period"),
+};

just a nit.. overhead_guest_us|sys misplace are not properly aligned

Ah.. this is a pre-existing bug and I plan to fix it (along with other related things) in a different series soon. So let's skip this for now. ;-p

Thanks,
Namhyung

#18 lore raw

Re: [PATCH 08/20] perf tools: Allow hpp fields to be sort keys

In reply to Namhyung Kim
Quoted 14 lines · Namhyung Kim

Add overhead{,_sys,_us,_guest_sys,_guest_us}, sample and period sort keys so that they can be selected with --sort/-s option.

  $ perf report -s period,comm --stdio
  ...
  # Overhead        Period          Command
  # ........  ............  ...............
  #
      47.06%           152          swapper
      13.93%            45  qemu-system-arm
      12.38%            40         synergys
       3.72%            12          firefox
       2.48%             8            xchat
Quoted 5 lines
---
 tools/perf/ui/hist.c   |  9 +++++++--
 tools/perf/util/sort.c | 39 +++++++++++++++++++++++++++++++++++++++

no documentation update.. Documentation/perf-report.txt

I wonder we want to update report's '-s' option help string.. it's already big ;-)

jirka

#19 lore raw

Re: [PATCH 08/20] perf tools: Allow hpp fields to be sort keys

Quoted 15 lines · Jiri Olsa
Quoted 14 lines · Namhyung Kim

Add overhead{,_sys,_us,_guest_sys,_guest_us}, sample and period sort keys so that they can be selected with --sort/-s option.

  $ perf report -s period,comm --stdio
  ...
  # Overhead        Period          Command
  # ........  ............  ...............
  #
      47.06%           152          swapper
      13.93%            45  qemu-system-arm
      12.38%            40         synergys
       3.72%            12          firefox
       2.48%             8            xchat
Quoted 7 lines
Quoted 5 lines
---
 tools/perf/ui/hist.c   |  9 +++++++--
 tools/perf/util/sort.c | 39 +++++++++++++++++++++++++++++++++++++++

no documentation update.. Documentation/perf-report.txt

Will add.

Quoted 2 lines

I wonder we want to update report's '-s' option help string.. it's already big ;-)

Right. What about just listing basic/common keys and description like "please refer the man page to the complete list." :)

Thanks,
Namhyung

author maintainer
#20 lore raw

PATCH 9/20 perf tools: Consolidate management of default sort orders

In reply to Namhyung Kim

The perf uses different default sort orders for different use-cases, and this was scattered throughout the code. Add get_default_sort_ order() function to handle this and change initial value of sort_order to NULL to distinguish it from user-given one.

Cc Stephane Eranian <[email protected]>
---
 tools/perf/builtin-report.c | 18 ------------------
 tools/perf/builtin-top.c    |  3 +--
 tools/perf/util/sort.c      | 25 +++++++++++++++++++++++--
 tools/perf/util/sort.h      |  1 +
 4 files changed, 25 insertions(+), 22 deletions(-)
tools/perf/builtin-report.c +0 -18
index 89c95289fd51..8c9fbbdc6505 100644
@@ -807,30 +807,12 @@ repeat:
 	if (branch_mode == -1 && has_br_stack)
 		sort__mode = SORT_MODE__BRANCH;
 
-	/* sort__mode could be NORMAL if --no-branch-stack */
-	if (sort__mode == SORT_MODE__BRANCH) {
-		/*
-		 * if no sort_order is provided, then specify
-		 * branch-mode specific order
-		 */
-		if (sort_order == default_sort_order)
-			sort_order = "comm,dso_from,symbol_from,"
-				     "dso_to,symbol_to";
-
-	}
 	if (report.mem_mode) {
 		if (sort__mode == SORT_MODE__BRANCH) {
 			pr_err("branch and mem mode incompatible\n");
 			goto error;
 		}
 		sort__mode = SORT_MODE__MEMORY;
-
-		/*
-		 * if no sort_order is provided, then specify
-		 * branch-mode specific order
-		 */
-		if (sort_order == default_sort_order)
-			sort_order = "local_weight,mem,sym,dso,symbol_daddr,dso_daddr,snoop,tlb,locked";
 	}
 
 	if (setup_sorting() < 0) {
tools/perf/builtin-top.c +1 -2
index 37d30460bada..bb2aa6645a7e 100644
@@ -1137,8 +1137,7 @@ int cmd_top(int argc, const char **argv, const char *prefix __maybe_unused)
 	if (argc)
 		usage_with_options(top_usage, options);
 
-	if (sort_order == default_sort_order)
-		sort_order = "dso,symbol";
+	sort__mode = SORT_MODE__TOP;
 
 	if (setup_sorting() < 0) {
 		parse_options_usage(top_usage, options, "s", 1);
tools/perf/util/sort.c +23 -2
index 916652af8304..2f83965ab2c0 100644
@@ -8,7 +8,10 @@ regex_t		parent_regex;
 const char	default_parent_pattern[] = "^sys_|^do_page_fault";
 const char	*parent_pattern = default_parent_pattern;
 const char	default_sort_order[] = "comm,dso,symbol";
-const char	*sort_order = default_sort_order;
+const char	default_branch_sort_order[] = "comm,dso_from,symbol_from,dso_to,symbol_to";
+const char	default_mem_sort_order[] = "local_weight,mem,sym,dso,symbol_daddr,dso_daddr,snoop,tlb,locked";
+const char	default_top_sort_order[] = "dso,symbol";
+const char	*sort_order;
 regex_t		ignore_callees_regex;
 int		have_ignore_callees = 0;
 int		sort__need_collapse = 0;
@@ -1218,11 +1221,29 @@ int sort_dimension__add(const char *tok)
 	return -ESRCH;
 }
 
+static const char *get_default_sort_order(void)
+{
+	const char *default_sort_orders[] = {
+		default_sort_order,
+		default_branch_sort_order,
+		default_mem_sort_order,
+		default_top_sort_order,
+	};
+
+	BUG_ON(sort__mode > ARRAY_SIZE(default_sort_orders));
+
+	return default_sort_orders[sort__mode];
+}
+
 int setup_sorting(void)
 {
-	char *tmp, *tok, *str = strdup(sort_order);
+	char *tmp, *tok, *str;
 	int ret = 0;
 
+	if (sort_order == NULL)
+		sort_order = get_default_sort_order();
+
+	str = strdup(sort_order);
 	if (str == NULL) {
 		error("Not enough memory to setup sort keys");
 		return -ENOMEM;
tools/perf/util/sort.h +1 -0
index 43e5ff42a609..35b53cc56feb 100644
@@ -133,6 +133,7 @@ enum sort_mode {
 	SORT_MODE__NORMAL,
 	SORT_MODE__BRANCH,
 	SORT_MODE__MEMORY,
+	SORT_MODE__TOP,
 };
 
 enum sort_type {
Signature · 1 line
1.9.2
#21 lore raw

Re: [PATCH 09/20] perf tools: Consolidate management of default sort orders

On Mon, May 12, 2014 at 03:28:42PM +0900, Namhyung Kim wrote:

SNIP

tools/perf/util/sort.c int sort_dimension__add(const char *tok) · line 1233

Quoted 11 lines
tools/perf/util/sort.c +10 -0
+static const char *get_default_sort_order(void)
+{
+	const char *default_sort_orders[] = {
+		default_sort_order,
+		default_branch_sort_order,
+		default_mem_sort_order,
+		default_top_sort_order,
+	};
+
+	BUG_ON(sort__mode > ARRAY_SIZE(default_sort_orders));

it should be >= in here ^^^

jirka

author maintainer
#22 lore raw

PATCH 10/20 perf tools: Call perf_hpp__init() before setting up GUI browsers

In reply to Namhyung Kim

So that it can be set properly prior to set up output fields. That makes easy to handle/warn errors during the setup since it doesn't need to be bothered with the GUI.

---
 tools/perf/builtin-report.c    | 6 +++---
 tools/perf/builtin-top.c       | 2 ++
 tools/perf/ui/browsers/hists.c | 2 --
 tools/perf/ui/gtk/hists.c      | 2 --
 tools/perf/ui/setup.c          | 2 --
 5 files changed, 5 insertions(+), 9 deletions(-)
tools/perf/builtin-report.c +3 -3
index 8c9fbbdc6505..76d8d0b4f7f5 100644
@@ -825,16 +825,16 @@ repeat:
 			goto error;
 	}
 
+	perf_hpp__init();
+
 	/* Force tty output for header output. */
 	if (report.header || report.header_only)
 		use_browser = 0;
 
 	if (strcmp(input_name, "-") != 0)
 		setup_browser(true);
-	else {
+	else
 		use_browser = 0;
-		perf_hpp__init();
-	}
 
 	if (report.header || report.header_only) {
 		perf_session__fprintf_info(session, stdout,
tools/perf/builtin-top.c +2 -0
index bb2aa6645a7e..9309629394dd 100644
@@ -1147,6 +1147,8 @@ int cmd_top(int argc, const char **argv, const char *prefix __maybe_unused)
 	/* display thread wants entries to be collapsed in a different tree */
 	sort__need_collapse = 1;
 
+	perf_hpp__init();
+
 	if (top.use_stdio)
 		use_browser = 0;
 	else if (top.use_tui)
tools/perf/ui/browsers/hists.c +0 -2
index 3ed9212d2a63..69c2b0e536ab 100644
@@ -661,8 +661,6 @@ __HPP_COLOR_PERCENT_FN(overhead_guest_us, period_guest_us)
 
 void hist_browser__init_hpp(void)
 {
-	perf_hpp__init();
-
 	perf_hpp__format[PERF_HPP__OVERHEAD].color =
 				hist_browser__hpp_color_overhead;
 	perf_hpp__format[PERF_HPP__OVERHEAD_SYS].color =
tools/perf/ui/gtk/hists.c +0 -2
index 2237245bfac0..fd52669018ee 100644
@@ -58,8 +58,6 @@ __HPP_COLOR_PERCENT_FN(overhead_guest_us, period_guest_us)
 
 void perf_gtk__init_hpp(void)
 {
-	perf_hpp__init();
-
 	perf_hpp__format[PERF_HPP__OVERHEAD].color =
 				perf_gtk__hpp_color_overhead;
 	perf_hpp__format[PERF_HPP__OVERHEAD_SYS].color =
tools/perf/ui/setup.c +0 -2
index 5df5140a9f29..ba51fa8a1176 100644
@@ -86,8 +86,6 @@ void setup_browser(bool fallback_to_pager)
 		use_browser = 0;
 		if (fallback_to_pager)
 			setup_pager();
-
-		perf_hpp__init();
 		break;
 	}
 }
Signature · 1 line
1.9.2
author maintainer
#23 lore raw

PATCH 11/20 perf report: Add -F option to specify output fields

In reply to Namhyung Kim

The -F/--fields option is to allow user setup output field in any order. It can recieve any sort keys and following (hpp) fields:

overhead, overhead_sys, overhead_us, sample and period

If guest profiling is enabled, overhead_guest_{sys,us} will be available too.

The output fields also affect sort order unless you give -s/--sort option. And any keys specified on -s option, will also be added to the output field list automatically.

  $ perf report -F sym,sample,overhead
  ...
  #                     Symbol       Samples  Overhead
  # ..........................  ............  ........
  #
    [.] __cxa_atexit                       2     2.50%
    [.] __libc_csu_init                    4     5.00%
    [.] __new_exitfn                       3     3.75%
    [.] _dl_check_map_versions             1     1.25%
    [.] _dl_name_match_p                   4     5.00%
    [.] _dl_setup_hash                     1     1.25%
    [.] _dl_sysdep_start                   1     1.25%
    [.] _init                              5     6.25%
    [.] _setjmp                            6     7.50%
    [.] a                                  8    10.00%
    [.] b                                  8    10.00%
    [.] brk                                1     1.25%
    [.] c                                  8    10.00%

Note that, the example output above is captured after applying next patch which fixes sort/comparing behavior.

---
 tools/perf/Documentation/perf-report.txt |  10 ++
 tools/perf/builtin-report.c              |   7 ++
 tools/perf/ui/hist.c                     |  61 +++++++++--
 tools/perf/util/hist.h                   |   5 +
 tools/perf/util/sort.c                   | 180 ++++++++++++++++++++++++++++++-
 tools/perf/util/sort.h                   |   2 +
 6 files changed, 255 insertions(+), 10 deletions(-)
tools/perf/Documentation/perf-report.txt +10 -0
index 09af66298564..8adbadf34b37 100644
@@ -98,6 +98,16 @@ OPTIONS
 	And default sort keys are changed to comm, dso_from, symbol_from, dso_to
 	and symbol_to, see '--branch-stack'.
 
+-F::
+--fields=::
+	Specify output field - multiple keys can be specified in CSV format.
+	Following fields are available:
+	overhead, overhead_sys, overhead_us, sample and period.
+	Also it can contain any sort key(s).
+
+	By default, every sort keys not specified in -F will be appended
+	automatically.
+
 -p::
 --parent=<regex>::
         A regex filter to identify parent. The parent is a caller of this
tools/perf/builtin-report.c +7 -0
index 76d8d0b4f7f5..39c9b3d2054c 100644
@@ -703,6 +703,8 @@ int cmd_report(int argc, const char **argv, const char *prefix __maybe_unused)
 		   " dso_to, dso_from, symbol_to, symbol_from, mispredict,"
 		   " weight, local_weight, mem, symbol_daddr, dso_daddr, tlb, "
 		   "snoop, locked, abort, in_tx, transaction"),
+	OPT_STRING('F', "fields", &field_order, "key[,keys...]",
+		   "output field(s): overhead, period, sample plus all of sort keys"),
 	OPT_BOOLEAN(0, "showcpuutilization", &symbol_conf.show_cpu_utilization,
 		    "Show sample percentage for different cpu modes"),
 	OPT_STRING('p', "parent", &parent_pattern, "regex",
@@ -827,6 +829,11 @@ repeat:
 
 	perf_hpp__init();
 
+	if (setup_output_field() < 0) {
+		parse_options_usage(report_usage, options, "F", 1);
+		goto error;
+	}
+
 	/* Force tty output for header output. */
 	if (report.header || report.header_only)
 		use_browser = 0;
tools/perf/ui/hist.c +55 -6

85 lines not on this page · raw message

tools/perf/util/hist.h +5 -0
index 78855d373381..f3713b79742d 100644
@@ -196,6 +196,11 @@ void perf_hpp__init(void);
 void perf_hpp__column_register(struct perf_hpp_fmt *format);
 void perf_hpp__column_enable(unsigned col);
 void perf_hpp__register_sort_field(struct perf_hpp_fmt *format);
+void perf_hpp__setup_output_field(void);
+void perf_hpp__append_sort_keys(void);
+
+bool perf_hpp__is_sort_entry(struct perf_hpp_fmt *format);
+bool perf_hpp__same_sort_entry(struct perf_hpp_fmt *a, struct perf_hpp_fmt *b);
 
 typedef u64 (*hpp_field_fn)(struct hist_entry *he);
 typedef int (*hpp_callback_fn)(struct perf_hpp *hpp, bool front);
tools/perf/util/sort.c +176 -4

235 lines not on this page · raw message

tools/perf/util/sort.h +2 -0
index 35b53cc56feb..02706c9766d6 100644
@@ -25,6 +25,7 @@
 
 extern regex_t parent_regex;
 extern const char *sort_order;
+extern const char *field_order;
 extern const char default_parent_pattern[];
 extern const char *parent_pattern;
 extern const char default_sort_order[];
@@ -190,6 +191,7 @@ extern struct sort_entry sort_thread;
 extern struct list_head hist_entry__sort_list;
 
 int setup_sorting(void);
+int setup_output_field(void);
 extern int sort_dimension__add(const char *);
 void sort__setup_elide(FILE *fp);
 
Signature · 1 line
1.9.2
#24 lore raw

Re: [PATCH 11/20] perf report: Add -F option to specify output fields

On Mon, May 12, 2014 at 03:28:44PM +0900, Namhyung Kim wrote:

SNIP

tools/perf/util/sort.c void sort__setup_elide(FILE *output) · line 1473

Quoted 10 lines
tools/perf/util/sort.c +10 -0
+
+int setup_output_field(void)
+{
+	char *tmp, *tok, *str;
+	int ret = 0;
+
+	if (field_order == NULL)
+		goto out;
+
+	reset_dimensions();

Do we want to call this in setup_sorting as well? In case we mix the order of setup_output_field and setup_sorting in the future?

jirka

#25 lore raw

Re: [PATCH 11/20] perf report: Add -F option to specify output fields

Quoted 4 lines · Jiri Olsa

On Mon, May 12, 2014 at 03:28:44PM +0900, Namhyung Kim wrote:

SNIP

Quoted 14 lines

tools/perf/util/sort.c void sort__setup_elide(FILE *output) · line 1473

Quoted 10 lines
tools/perf/util/sort.c +10 -0
+
+int setup_output_field(void)
+{
+	char *tmp, *tok, *str;
+	int ret = 0;
+
+	if (field_order == NULL)
+		goto out;
+
+	reset_dimensions();

Do we want to call this in setup_sorting as well? In case we mix the order of setup_output_field and setup_sorting in the future?

Yeah, looks like a good idea.

Thanks,
Namhyung

author maintainer
#26 lore raw

Re: [PATCH 11/20] perf report: Add -F option to specify output fields

Quoted 20 lines · Namhyung Kim
Quoted 4 lines · Jiri Olsa

On Mon, May 12, 2014 at 03:28:44PM +0900, Namhyung Kim wrote:

SNIP

Quoted 14 lines

tools/perf/util/sort.c void sort__setup_elide(FILE *output) · line 1473

Quoted 10 lines
tools/perf/util/sort.c +10 -0
+
+int setup_output_field(void)
+{
+	char *tmp, *tok, *str;
+	int ret = 0;
+
+	if (field_order == NULL)
+		goto out;
+
+	reset_dimensions();

Do we want to call this in setup_sorting as well? In case we mix the order of setup_output_field and setup_sorting in the future?

Quoted 1 line

Yeah, looks like a good idea.

Hmm.. I misunderstood your point. The thing is that the initialization step needs to be done in a specific order (due to their implicit dependency). So I change it to be called in one place, setup_sorting().

Thanks,
Namhyung

author maintainer
#27 lore raw

PATCH 12/20 perf tools: Add ->sort() member to struct sort_entry

In reply to Namhyung Kim

Currently, what the sort_entry does is just identifying hist entries so that they can be grouped properly. However, with -F option support, it indeed needs to sort entries appropriately to be shown to users. So add ->sort() member to do it.

---
 tools/perf/util/sort.c | 27 ++++++++++++++++++++++-----
 tools/perf/util/sort.h |  1 +
 2 files changed, 23 insertions(+), 5 deletions(-)
tools/perf/util/sort.c +22 -5
index 639dd49f2884..1e7b80e517d5 100644
@@ -98,6 +98,12 @@ sort__comm_collapse(struct hist_entry *left, struct hist_entry *right)
 	return comm__str(right->comm) - comm__str(left->comm);
 }
 
+static int64_t
+sort__comm_sort(struct hist_entry *left, struct hist_entry *right)
+{
+	return strcmp(comm__str(right->comm), comm__str(left->comm));
+}
+
 static int hist_entry__comm_snprintf(struct hist_entry *he, char *bf,
 				     size_t size, unsigned int width)
 {
@@ -108,6 +114,7 @@ struct sort_entry sort_comm = {
 	.se_header	= "Command",
 	.se_cmp		= sort__comm_cmp,
 	.se_collapse	= sort__comm_collapse,
+	.se_sort	= sort__comm_sort,
 	.se_snprintf	= hist_entry__comm_snprintf,
 	.se_width_idx	= HISTC_COMM,
 };
@@ -121,7 +128,7 @@ static int64_t _sort__dso_cmp(struct map *map_l, struct map *map_r)
 	const char *dso_name_l, *dso_name_r;
 
 	if (!dso_l || !dso_r)
-		return cmp_null(dso_l, dso_r);
+		return cmp_null(dso_r, dso_l);
 
 	if (verbose) {
 		dso_name_l = dso_l->long_name;
@@ -137,7 +144,7 @@ static int64_t _sort__dso_cmp(struct map *map_l, struct map *map_r)
 static int64_t
 sort__dso_cmp(struct hist_entry *left, struct hist_entry *right)
 {
-	return _sort__dso_cmp(left->ms.map, right->ms.map);
+	return _sort__dso_cmp(right->ms.map, left->ms.map);
 }
 
 static int _hist_entry__dso_snprintf(struct map *map, char *bf,
@@ -209,6 +216,15 @@ sort__sym_cmp(struct hist_entry *left, struct hist_entry *right)
 	return _sort__sym_cmp(left->ms.sym, right->ms.sym);
 }
 
+static int64_t
+sort__sym_sort(struct hist_entry *left, struct hist_entry *right)
+{
+	if (!left->ms.sym || !right->ms.sym)
+		return cmp_null(left->ms.sym, right->ms.sym);
+
+	return strcmp(right->ms.sym->name, left->ms.sym->name);
+}
+
 static int _hist_entry__sym_snprintf(struct map *map, struct symbol *sym,
 				     u64 ip, char level, char *bf, size_t size,
 				     unsigned int width)
@@ -255,6 +271,7 @@ static int hist_entry__sym_snprintf(struct hist_entry *he, char *bf,
 struct sort_entry sort_sym = {
 	.se_header	= "Symbol",
 	.se_cmp		= sort__sym_cmp,
+	.se_sort	= sort__sym_sort,
 	.se_snprintf	= hist_entry__sym_snprintf,
 	.se_width_idx	= HISTC_SYMBOL,
 };
@@ -282,7 +299,7 @@ sort__srcline_cmp(struct hist_entry *left, struct hist_entry *right)
 					    map__rip_2objdump(map, right->ip));
 		}
 	}
-	return strcmp(left->srcline, right->srcline);
+	return strcmp(right->srcline, left->srcline);
 }
 
 static int hist_entry__srcline_snprintf(struct hist_entry *he, char *bf,
@@ -310,7 +327,7 @@ sort__parent_cmp(struct hist_entry *left, struct hist_entry *right)
 	if (!sym_l || !sym_r)
 		return cmp_null(sym_l, sym_r);
 
-	return strcmp(sym_l->name, sym_r->name);
+	return strcmp(sym_r->name, sym_l->name);
 }
 
 static int hist_entry__parent_snprintf(struct hist_entry *he, char *bf,
@@ -1125,7 +1142,7 @@ __sort_dimension__alloc_hpp(struct sort_dimension *sd)
 
 	hse->hpp.cmp = sd->entry->se_cmp;
 	hse->hpp.collapse = sd->entry->se_collapse ? : sd->entry->se_cmp;
-	hse->hpp.sort = hse->hpp.collapse;
+	hse->hpp.sort = sd->entry->se_sort ? : hse->hpp.collapse;
 
 	INIT_LIST_HEAD(&hse->hpp.list);
 	INIT_LIST_HEAD(&hse->hpp.sort_list);
tools/perf/util/sort.h +1 -0
index 02706c9766d6..cd679a56c81d 100644
@@ -181,6 +181,7 @@ struct sort_entry {
 
 	int64_t (*se_cmp)(struct hist_entry *, struct hist_entry *);
 	int64_t (*se_collapse)(struct hist_entry *, struct hist_entry *);
+	int64_t	(*se_sort)(struct hist_entry *, struct hist_entry *);
 	int	(*se_snprintf)(struct hist_entry *he, char *bf, size_t size,
 			       unsigned int width);
 	u8	se_width_idx;
Signature · 1 line
1.9.2
#28 lore raw

Re: [PATCH 12/20] perf tools: Add ->sort() member to struct sort_entry

On Mon, May 12, 2014 at 03:28:45PM +0900, Namhyung Kim wrote:

SNIP

tools/perf/util/sort.c sort__sym_cmp(struct hist_entry *left, struct hist_entry *right) · line 226

Quoted 8 lines
tools/perf/util/sort.c +8 -0
+static int64_t
+sort__sym_sort(struct hist_entry *left, struct hist_entry *right)
+{
+	if (!left->ms.sym || !right->ms.sym)
+		return cmp_null(left->ms.sym, right->ms.sym);
+
+	return strcmp(right->ms.sym->name, left->ms.sym->name);
+}

why do we need just string comparison for sort?

tools/perf/util/sort.c static int hist_entry__sym_snprintf(struct hist_entry *he, char *bf, · line 276

Quoted 11 lines
tools/perf/util/sort.c +2 -0
+
 static int _hist_entry__sym_snprintf(struct map *map, struct symbol *sym,
 				     u64 ip, char level, char *bf, size_t size,
 				     unsigned int width)
@@ -255,6 +271,7 @@ static int hist_entry__sym_snprintf(struct hist_entry *he, char *bf,
 struct sort_entry sort_sym = {
 	.se_header	= "Symbol",
 	.se_cmp		= sort__sym_cmp,
+	.se_sort	= sort__sym_sort,
 	.se_snprintf	= hist_entry__sym_snprintf,
 	.se_width_idx	= HISTC_SYMBOL,

so when we insert entries into hists we use se_cmp to group them via hist_entry__cmp

the we sort this hists based on output and use se_sort to group them via __hists__insert_output_entry

why can't we use just cmp?

jirka

#29 lore raw

Re: [PATCH 12/20] perf tools: Add ->sort() member to struct sort_entry

Quoted 4 lines · Jiri Olsa

On Mon, May 12, 2014 at 03:28:45PM +0900, Namhyung Kim wrote:

SNIP

Quoted 10 lines

tools/perf/util/sort.c sort__sym_cmp(struct hist_entry *left, struct hist_entry *right) · line 226

Quoted 8 lines
tools/perf/util/sort.c +8 -0
+static int64_t
+sort__sym_sort(struct hist_entry *left, struct hist_entry *right)
+{
+	if (!left->ms.sym || !right->ms.sym)
+		return cmp_null(left->ms.sym, right->ms.sym);
+
+	return strcmp(right->ms.sym->name, left->ms.sym->name);
+}

why do we need just string comparison for sort?

Because we want to sort them in alphabetical order.

Quoted 16 lines

tools/perf/util/sort.c static int hist_entry__sym_snprintf(struct hist_entry *he, char *bf, · line 276

Quoted 12 lines
tools/perf/util/sort.c +2 -0
+
 static int _hist_entry__sym_snprintf(struct map *map, struct symbol *sym,
 				     u64 ip, char level, char *bf, size_t size,
 				     unsigned int width)
@@ -255,6 +271,7 @@ static int hist_entry__sym_snprintf(struct hist_entry *he, char *bf,
 struct sort_entry sort_sym = {
 	.se_header	= "Symbol",
 	.se_cmp		= sort__sym_cmp,
+	.se_sort	= sort__sym_sort,
 	.se_snprintf	= hist_entry__sym_snprintf,
 	.se_width_idx	= HISTC_SYMBOL,

so when we insert entries into hists we use se_cmp to group them via hist_entry__cmp

Quoted 4 lines

the we sort this hists based on output and use se_sort to group them via __hists__insert_output_entry

why can't we use just cmp?

As I said, the output should be sorted by symbol name. IOW, in order to group entries in input stage, putting them into a correct group is important, but for output stage, the order is also important.

Thanks,
Namhyung

author maintainer
#30 lore raw

PATCH 13/20 perf report/tui: Fix a bug when --fields/sort is given

In reply to Namhyung Kim

The hists__filter_entries() function is called when down arrow key is pressed for navigating through the entries in TUI. It has a check for filtering out entries that have very small overhead (under min_pcnt).

However it just assumed the entries are sorted by the overhead so when it saw such a small overheaded entry, it just stopped navigating as an optimization. But it's not true anymore due to new --fields and --sort optoin behavior and this case users cannot go down to a next entry if ther's an entry with small overhead in-between.

---
 tools/perf/ui/browsers/hists.c | 5 +----
 1 file changed, 1 insertion(+), 4 deletions(-)
tools/perf/ui/browsers/hists.c +1 -4
index 69c2b0e536ab..20b200f88129 100644
@@ -812,10 +812,7 @@ static struct rb_node *hists__filter_entries(struct rb_node *nd,
 		if (total)
 			percent = h->stat.period * 100.0 / total;
 
-		if (percent < min_pcnt)
-			return NULL;
-
-		if (!h->filtered)
+		if (!h->filtered && percent >= min_pcnt)
 			return nd;
 
 		nd = rb_next(nd);
Signature · 1 line
1.9.2
author maintainer
#31 lore raw

PATCH 14/20 perf top: Add --fields option to specify output fields

In reply to Namhyung Kim

The --fields option is to allow user setup output field in any order. It can recieve any sort keys and following (hpp) fields:

overhead, overhead_sys, overhead_us, sample and period

If guest profiling is enabled, overhead_guest_{sys,us} will be available too.

More more information, please see previous patch "perf report:
Add -F option to specify output fields"

---
 tools/perf/Documentation/perf-top.txt | 9 +++++++++
 tools/perf/builtin-top.c              | 7 +++++++
 2 files changed, 16 insertions(+)
tools/perf/Documentation/perf-top.txt +9 -0
index 64ed79c43639..feac28017419 100644
@@ -115,6 +115,15 @@ Default is to monitor all CPUS.
 	Sort by key(s): pid, comm, dso, symbol, parent, srcline, weight,
 	local_weight, abort, in_tx, transaction
 
+--fields=::
+	Specify output field - multiple keys can be specified in CSV format.
+	Following fields are available:
+	overhead, overhead_sys, overhead_us, sample and period.
+	Also it can contain any sort key(s).
+
+	By default, every sort keys not specified in --field will be appended
+	automatically.
+
 -n::
 --show-nr-samples::
 	Show a column with the number of samples.
tools/perf/builtin-top.c +7 -0
index 9309629394dd..7d133dff5e15 100644
@@ -1085,6 +1085,8 @@ int cmd_top(int argc, const char **argv, const char *prefix __maybe_unused)
 	OPT_STRING('s', "sort", &sort_order, "key[,key2...]",
 		   "sort by key(s): pid, comm, dso, symbol, parent, weight, local_weight,"
 		   " abort, in_tx, transaction"),
+	OPT_STRING(0, "fields", &field_order, "key[,keys...]",
+		   "output field(s): overhead, period, sample plus all of sort keys"),
 	OPT_BOOLEAN('n', "show-nr-samples", &symbol_conf.show_nr_samples,
 		    "Show a column with the number of samples"),
 	OPT_CALLBACK_NOOPT('g', NULL, &top.record_opts,
@@ -1149,6 +1151,11 @@ int cmd_top(int argc, const char **argv, const char *prefix __maybe_unused)
 
 	perf_hpp__init();
 
+	if (setup_output_field() < 0) {
+		parse_options_usage(top_usage, options, "fields", 0);
+		goto out_delete_evlist;
+	}
+
 	if (top.use_stdio)
 		use_browser = 0;
 	else if (top.use_tui)
Signature · 1 line
1.9.2
author maintainer
#32 lore raw

PATCH 15/20 perf diff: Add missing setup_output_field()

In reply to Namhyung Kim

The perf diff command itself doesn't make use of the --fields option, it still needs to call the function since the output only work with that way.

---
 tools/perf/builtin-diff.c | 3 +++
 1 file changed, 3 insertions(+)
tools/perf/builtin-diff.c +3 -0
index f3b10dcf6838..670d191bec31 100644
@@ -1155,6 +1155,9 @@ int cmd_diff(int argc, const char **argv, const char *prefix __maybe_unused)
 	if (setup_sorting() < 0)
 		usage_with_options(diff_usage, options);
 
+	if (setup_output_field() < 0)
+		usage_with_options(diff_usage, options);
+
 	setup_pager();
 
 	sort__setup_elide(NULL);
Signature · 1 line
1.9.2
#33 lore raw

Re: [PATCH 15/20] perf diff: Add missing setup_output_field()

Quoted 14 lines · Namhyung Kim

The perf diff command itself doesn't make use of the --fields option, it still needs to call the function since the output only work with that way.

---
 tools/perf/builtin-diff.c | 3 +++
 1 file changed, 3 insertions(+)
tools/perf/builtin-diff.c +0 -0
index f3b10dcf6838..670d191bec31 100644

tools/perf/builtin-diff.c int cmd_diff(int argc, const char **argv, const char *prefix __maybe_unused) · line 1160

Quoted 6 lines
tools/perf/builtin-diff.c +3 -0
@@ -1155,6 +1155,9 @@ int cmd_diff(int argc, const char **argv, const char *prefix __maybe_unused)
 	if (setup_sorting() < 0)
 		usage_with_options(diff_usage, options);
 
+	if (setup_output_field() < 0)
+		usage_with_options(diff_usage, options);
+

looks like this breaks perf diff even worse than it is now ;-)

[jolsa@krava perf]$ ./perf diff
# Event 'cycles'
#
# Baseline Delta

# ........  .......
#
    28.03%         
            +18.93%
    31.06%         
            +36.86%
    35.06%         
     0.38%         
     0.03%   +0.03%
             +1.11%
     5.44%         
            +43.05%

anyway, as I said it was already broken, I plan to check/fix this soon..

jirka

author maintainer
#34 lore raw

PATCH 16/20 perf tools: Skip elided sort entries

In reply to Namhyung Kim

When it converted sort entries to hpp formats, it missed se->elide handling, so add it for compatibility.

---
 tools/perf/ui/browsers/hists.c |  3 +++
 tools/perf/ui/gtk/hists.c      |  6 ++++++
 tools/perf/ui/stdio/hist.c     |  9 +++++++++
 tools/perf/util/hist.c         |  9 +++++++++
 tools/perf/util/hist.h         |  1 +
 tools/perf/util/sort.c         | 11 +++++++++++
 6 files changed, 39 insertions(+)
tools/perf/ui/browsers/hists.c +3 -0
index 20b200f88129..fa46e592b588 100644
@@ -711,6 +711,9 @@ static int hist_browser__show_entry(struct hist_browser *browser,
 		ui_browser__gotorc(&browser->b, row, 0);
 
 		perf_hpp__for_each_format(fmt) {
+			if (perf_hpp__should_skip(fmt))
+				continue;
+
 			if (current_entry && browser->b.navkeypressed) {
 				ui_browser__set_color(&browser->b,
 						      HE_COLORSET_SELECTED);
tools/perf/ui/gtk/hists.c +6 -0
index fd52669018ee..9d90683914d4 100644
@@ -178,6 +178,9 @@ static void perf_gtk__show_hists(GtkWidget *window, struct hists *hists,
 	col_idx = 0;
 
 	perf_hpp__for_each_format(fmt) {
+		if (perf_hpp__should_skip(fmt))
+			continue;
+
 		fmt->header(fmt, &hpp, hists_to_evsel(hists));
 
 		gtk_tree_view_insert_column_with_attributes(GTK_TREE_VIEW(view),
@@ -222,6 +225,9 @@ static void perf_gtk__show_hists(GtkWidget *window, struct hists *hists,
 		col_idx = 0;
 
 		perf_hpp__for_each_format(fmt) {
+			if (perf_hpp__should_skip(fmt))
+				continue;
+
 			if (fmt->color)
 				fmt->color(fmt, &hpp, h);
 			else
tools/perf/ui/stdio/hist.c +9 -0
index e6920d124c60..d2934659fd07 100644
@@ -319,6 +319,9 @@ static int hist_entry__period_snprintf(struct perf_hpp *hpp,
 		return 0;
 
 	perf_hpp__for_each_format(fmt) {
+		if (perf_hpp__should_skip(fmt))
+			continue;
+
 		/*
 		 * If there's no field_sep, we still need
 		 * to display initial '  '.
@@ -408,6 +411,9 @@ size_t hists__fprintf(struct hists *hists, bool show_header, int max_rows,
 	fprintf(fp, "# ");
 
 	perf_hpp__for_each_format(fmt) {
+		if (perf_hpp__should_skip(fmt))
+			continue;
+
 		if (!first)
 			fprintf(fp, "%s", sep ?: "  ");
 		else
@@ -431,6 +437,9 @@ size_t hists__fprintf(struct hists *hists, bool show_header, int max_rows,
 	perf_hpp__for_each_format(fmt) {
 		unsigned int i;
 
+		if (perf_hpp__should_skip(fmt))
+			continue;
+
 		if (!first)
 			fprintf(fp, "%s", sep ?: "  ");
 		else
tools/perf/util/hist.c +9 -0
index ae13c2dbd27a..b262b44b7a65 100644
@@ -436,6 +436,9 @@ hist_entry__cmp(struct hist_entry *left, struct hist_entry *right)
 	int64_t cmp = 0;
 
 	perf_hpp__for_each_sort_list(fmt) {
+		if (perf_hpp__should_skip(fmt))
+			continue;
+
 		cmp = fmt->cmp(left, right);
 		if (cmp)
 			break;
@@ -451,6 +454,9 @@ hist_entry__collapse(struct hist_entry *left, struct hist_entry *right)
 	int64_t cmp = 0;
 
 	perf_hpp__for_each_sort_list(fmt) {
+		if (perf_hpp__should_skip(fmt))
+			continue;
+
 		cmp = fmt->collapse(left, right);
 		if (cmp)
 			break;
@@ -570,6 +576,9 @@ static int hist_entry__sort(struct hist_entry *a, struct hist_entry *b)
 	int64_t cmp = 0;
 
 	perf_hpp__for_each_sort_list(fmt) {
+		if (perf_hpp__should_skip(fmt))
+			continue;
+
 		cmp = fmt->sort(a, b);
 		if (cmp)
 			break;
tools/perf/util/hist.h +1 -0
index f3713b79742d..f67feb432a44 100644
@@ -201,6 +201,7 @@ void perf_hpp__append_sort_keys(void);
 
 bool perf_hpp__is_sort_entry(struct perf_hpp_fmt *format);
 bool perf_hpp__same_sort_entry(struct perf_hpp_fmt *a, struct perf_hpp_fmt *b);
+bool perf_hpp__should_skip(struct perf_hpp_fmt *format);
 
 typedef u64 (*hpp_field_fn)(struct hist_entry *he);
 typedef int (*hpp_callback_fn)(struct perf_hpp *hpp, bool front);
tools/perf/util/sort.c +11 -0
index 1e7b80e517d5..9dc33df4f9a6 100644
@@ -1355,6 +1355,17 @@ int setup_sorting(void)
 	return ret;
 }
 
+bool perf_hpp__should_skip(struct perf_hpp_fmt *format)
+{
+	if (perf_hpp__is_sort_entry(format)) {
+		struct hpp_sort_entry *hse;
+
+		hse = container_of(format, struct hpp_sort_entry, hpp);
+		return hse->se->elide;
+	}
+	return false;
+}
+
 static void sort_entry__setup_elide(struct sort_entry *se,
 				    struct strlist *list,
 				    const char *list_name, FILE *fp)
Signature · 1 line
1.9.2
author maintainer
#35 lore raw

PATCH 17/20 perf hists: Reset width of output fields with header length

In reply to Namhyung Kim

Some fields missed to set default column length so it broke align in --stdio output. Add perf_hpp__reset_width() to set it to a sane default value.

Note that this change will ignore -w/--column-widths option for now.

Before:

  $ perf report -F cpu,comm,overhead --stdio
  ...
  # CPU          Command  Overhead
  #   ...............  ........
  #
    0          firefox     2.65%
    0      kworker/0:0     1.45%
    0          swapper     5.52%
    0         synergys     0.92%
    1          firefox     4.54%

After:
# CPU Command Overhead

  # ...  ...............  ........
  #
      0          firefox     2.65%
      0      kworker/0:0     1.45%
      0          swapper     5.52%
      0         synergys     0.92%
      1          firefox     4.54%
---
 tools/perf/ui/stdio/hist.c | 21 +++------------------
 tools/perf/util/hist.h     |  1 +
 tools/perf/util/sort.c     | 12 ++++++++++++
 3 files changed, 16 insertions(+), 18 deletions(-)
tools/perf/ui/stdio/hist.c +3 -18
index d2934659fd07..0363b19930ed 100644
@@ -370,12 +370,10 @@ size_t hists__fprintf(struct hists *hists, bool show_header, int max_rows,
 		      int max_cols, float min_pcnt, FILE *fp)
 {
 	struct perf_hpp_fmt *fmt;
-	struct sort_entry *se;
 	struct rb_node *nd;
 	size_t ret = 0;
 	unsigned int width;
 	const char *sep = symbol_conf.field_sep;
-	const char *col_width = symbol_conf.col_width_list_str;
 	int nr_rows = 0;
 	char bf[96];
 	struct perf_hpp dummy_hpp = {
@@ -388,22 +386,9 @@ size_t hists__fprintf(struct hists *hists, bool show_header, int max_rows,
 
 	init_rem_hits();
 
-	list_for_each_entry(se, &hist_entry__sort_list, list) {
-		if (se->elide)
-			continue;
-		width = strlen(se->se_header);
-		if (symbol_conf.col_width_list_str) {
-			if (col_width) {
-				hists__set_col_len(hists, se->se_width_idx,
-						   atoi(col_width));
-				col_width = strchr(col_width, ',');
-				if (col_width)
-					++col_width;
-			}
-		}
-		if (!hists__new_col_len(hists, se->se_width_idx, width))
-			width = hists__col_len(hists, se->se_width_idx);
-	}
+
+	perf_hpp__for_each_format(fmt)
+		perf_hpp__reset_width(fmt, hists);
 
 	if (!show_header)
 		goto print_entries;
tools/perf/util/hist.h +1 -0
index f67feb432a44..034db761630e 100644
@@ -202,6 +202,7 @@ void perf_hpp__append_sort_keys(void);
 bool perf_hpp__is_sort_entry(struct perf_hpp_fmt *format);
 bool perf_hpp__same_sort_entry(struct perf_hpp_fmt *a, struct perf_hpp_fmt *b);
 bool perf_hpp__should_skip(struct perf_hpp_fmt *format);
+void perf_hpp__reset_width(struct perf_hpp_fmt *fmt, struct hists *hists);
 
 typedef u64 (*hpp_field_fn)(struct hist_entry *he);
 typedef int (*hpp_callback_fn)(struct perf_hpp *hpp, bool front);
tools/perf/util/sort.c +12 -0
index 9dc33df4f9a6..d4502db36cba 100644
@@ -1088,6 +1088,18 @@ bool perf_hpp__same_sort_entry(struct perf_hpp_fmt *a, struct perf_hpp_fmt *b)
 	return hse_a->se == hse_b->se;
 }
 
+void perf_hpp__reset_width(struct perf_hpp_fmt *fmt, struct hists *hists)
+{
+	struct hpp_sort_entry *hse;
+
+	if (!perf_hpp__is_sort_entry(fmt))
+		return;
+
+	hse = container_of(fmt, struct hpp_sort_entry, hpp);
+	hists__new_col_len(hists, hse->se->se_width_idx,
+			   strlen(hse->se->se_header));
+}
+
 static int __sort__hpp_header(struct perf_hpp_fmt *fmt, struct perf_hpp *hpp,
 			      struct perf_evsel *evsel)
 {
Signature · 1 line
1.9.2
author maintainer
#36 lore raw

PATCH 18/20 perf tools: Introduce reset_output_field()

In reply to Namhyung Kim

The reset_output_field() function is for clearing output field settings and will be used for test code in later patch.

---
 tools/perf/ui/hist.c   | 17 +++++++++++++++++
 tools/perf/util/hist.h |  7 +++++++
 tools/perf/util/sort.c | 18 ++++++++++++++++++
 tools/perf/util/sort.h |  1 +
 4 files changed, 43 insertions(+)
tools/perf/ui/hist.c +17 -0
index f51cba43e9e7..265c8a17c4ca 100644
@@ -459,6 +459,23 @@ next:
 	}
 }
 
+void perf_hpp__reset_output_field(void)
+{
+	struct perf_hpp_fmt *fmt, *tmp;
+
+	/* reset output fields */
+	perf_hpp__for_each_format_safe(fmt, tmp) {
+		list_del_init(&fmt->list);
+		list_del_init(&fmt->sort_list);
+	}
+
+	/* reset sort keys */
+	perf_hpp__for_each_sort_list_safe(fmt, tmp) {
+		list_del_init(&fmt->list);
+		list_del_init(&fmt->sort_list);
+	}
+}
+
 int hist_entry__sort_snprintf(struct hist_entry *he, char *s, size_t size,
 			      struct hists *hists)
 {
tools/perf/util/hist.h +7 -0
index 034db761630e..a8418d19808d 100644
@@ -174,9 +174,15 @@ extern struct list_head perf_hpp__sort_list;
 #define perf_hpp__for_each_format(format) \
 	list_for_each_entry(format, &perf_hpp__list, list)
 
+#define perf_hpp__for_each_format_safe(format, tmp)	\
+	list_for_each_entry_safe(format, tmp, &perf_hpp__list, list)
+
 #define perf_hpp__for_each_sort_list(format) \
 	list_for_each_entry(format, &perf_hpp__sort_list, sort_list)
 
+#define perf_hpp__for_each_sort_list_safe(format, tmp)	\
+	list_for_each_entry_safe(format, tmp, &perf_hpp__sort_list, sort_list)
+
 extern struct perf_hpp_fmt perf_hpp__format[];
 
 enum {
@@ -197,6 +203,7 @@ void perf_hpp__column_register(struct perf_hpp_fmt *format);
 void perf_hpp__column_enable(unsigned col);
 void perf_hpp__register_sort_field(struct perf_hpp_fmt *format);
 void perf_hpp__setup_output_field(void);
+void perf_hpp__reset_output_field(void);
 void perf_hpp__append_sort_keys(void);
 
 bool perf_hpp__is_sort_entry(struct perf_hpp_fmt *format);
tools/perf/util/sort.c +18 -0
index d4502db36cba..aa52066b9a74 100644
@@ -1540,3 +1540,21 @@ out:
 
 	return ret;
 }
+
+void reset_output_field(void)
+{
+	struct sort_entry *pos, *tmp;
+
+	sort__need_collapse = 0;
+	sort__has_parent = 0;
+	sort__has_sym = 0;
+	sort__has_dso = 0;
+
+	sort__first_dimension = 0;
+
+	list_for_each_entry_safe(pos, tmp, &hist_entry__sort_list, list)
+		list_del_init(&pos->list);
+
+	reset_dimensions();
+	perf_hpp__reset_output_field();
+}
tools/perf/util/sort.h +1 -0
index cd679a56c81d..f7acf9dedf57 100644
@@ -193,6 +193,7 @@ extern struct list_head hist_entry__sort_list;
 
 int setup_sorting(void);
 int setup_output_field(void);
+void reset_output_field(void);
 extern int sort_dimension__add(const char *);
 void sort__setup_elide(FILE *fp);
 
Signature · 1 line
1.9.2
author maintainer
#37 lore raw

PATCH 19/20 perf tests: Factor out print_hists_*()

In reply to Namhyung Kim

Those print helper functions can be reused by later hist test cases so factor them out to a common location.

---
 tools/perf/tests/hists_common.c | 57 +++++++++++++++++++++++++++++++++++++++++
 tools/perf/tests/hists_common.h |  3 +++
 tools/perf/tests/hists_filter.c | 37 ++++----------------------
 tools/perf/tests/hists_link.c   | 29 +--------------------
 4 files changed, 66 insertions(+), 60 deletions(-)
tools/perf/tests/hists_common.c +57 -0
index 44655b395bb9..040a85b17aee 100644
@@ -146,3 +146,60 @@ out:
 	machine__delete(machine);
 	return NULL;
 }
+
+void print_hists_in(struct hists *hists)
+{
+	int i = 0;
+	struct rb_root *root;
+	struct rb_node *node;
+
+	if (sort__need_collapse)
+		root = &hists->entries_collapsed;
+	else
+		root = hists->entries_in;
+
+	pr_info("----- %s --------\n", __func__);
+	node = rb_first(root);
+	while (node) {
+		struct hist_entry *he;
+
+		he = rb_entry(node, struct hist_entry, rb_node_in);
+
+		if (!he->filtered) {
+			pr_info("%2d: entry: %-8s [%-8s] %20s: period = %"PRIu64"\n",
+				i, thread__comm_str(he->thread),
+				he->ms.map->dso->short_name,
+				he->ms.sym->name, he->stat.period);
+		}
+
+		i++;
+		node = rb_next(node);
+	}
+}
+
+void print_hists_out(struct hists *hists)
+{
+	int i = 0;
+	struct rb_root *root;
+	struct rb_node *node;
+
+	root = &hists->entries;
+
+	pr_info("----- %s --------\n", __func__);
+	node = rb_first(root);
+	while (node) {
+		struct hist_entry *he;
+
+		he = rb_entry(node, struct hist_entry, rb_node);
+
+		if (!he->filtered) {
+			pr_info("%2d: entry: %-8s [%-8s] %20s: period = %"PRIu64"\n",
+				i, thread__comm_str(he->thread),
+				he->ms.map->dso->short_name,
+				he->ms.sym->name, he->stat.period);
+		}
+
+		i++;
+		node = rb_next(node);
+	}
+}
tools/perf/tests/hists_common.h +3 -0
index 2528b8fc105a..1415ae69d7b6 100644
@@ -41,4 +41,7 @@ struct machines;
  */
 struct machine *setup_fake_machine(struct machines *machines);
 
+void print_hists_in(struct hists *hists);
+void print_hists_out(struct hists *hists);
+
 #endif /* __PERF_TESTS__HISTS_COMMON_H__ */
tools/perf/tests/hists_filter.c +5 -32
index 4617a8bee29b..13c8cf49225e 100644
@@ -98,33 +98,6 @@ out:
 	return TEST_FAIL;
 }
 
-static void print_hists(struct hists *hists)
-{
-	int i = 0;
-	struct rb_root *root;
-	struct rb_node *node;
-
-	root = &hists->entries;
-
-	pr_info("----- %s --------\n", __func__);
-	node = rb_first(root);
-	while (node) {
-		struct hist_entry *he;
-
-		he = rb_entry(node, struct hist_entry, rb_node);
-
-		if (!he->filtered) {
-			pr_info("%2d: entry: %-8s [%-8s] %20s: period = %"PRIu64"\n",
-				i, thread__comm_str(he->thread),
-				he->ms.map->dso->short_name,
-				he->ms.sym->name, he->stat.period);
-		}
-
-		i++;
-		node = rb_next(node);
-	}
-}
-
 int test__hists_filter(void)
 {
 	int err = TEST_FAIL;
@@ -169,7 +142,7 @@ int test__hists_filter(void)
 
 		if (verbose > 2) {
 			pr_info("Normal histogram\n");
-			print_hists(hists);
+			print_hists_out(hists);
 		}
 
 		TEST_ASSERT_VAL("Invalid nr samples",
@@ -193,7 +166,7 @@ int test__hists_filter(void)
 
 		if (verbose > 2) {
 			pr_info("Histogram for thread filter\n");
-			print_hists(hists);
+			print_hists_out(hists);
 		}
 
 		/* normal stats should be invariant */
@@ -222,7 +195,7 @@ int test__hists_filter(void)
 
 		if (verbose > 2) {
 			pr_info("Histogram for dso filter\n");
-			print_hists(hists);
+			print_hists_out(hists);
 		}
 
 		/* normal stats should be invariant */
@@ -257,7 +230,7 @@ int test__hists_filter(void)
 
 		if (verbose > 2) {
 			pr_info("Histogram for symbol filter\n");
-			print_hists(hists);
+			print_hists_out(hists);
 		}
 
 		/* normal stats should be invariant */
@@ -284,7 +257,7 @@ int test__hists_filter(void)
 
 		if (verbose > 2) {
 			pr_info("Histogram for all filters\n");
-			print_hists(hists);
+			print_hists_out(hists);
 		}
 
 		/* normal stats should be invariant */
tools/perf/tests/hists_link.c +1 -28
index b009bbf440d9..4e783db60bba 100644
@@ -268,33 +268,6 @@ static int validate_link(struct hists *leader, struct hists *other)
 	return __validate_link(leader, 0) || __validate_link(other, 1);
 }
 
-static void print_hists(struct hists *hists)
-{
-	int i = 0;
-	struct rb_root *root;
-	struct rb_node *node;
-
-	if (sort__need_collapse)
-		root = &hists->entries_collapsed;
-	else
-		root = hists->entries_in;
-
-	pr_info("----- %s --------\n", __func__);
-	node = rb_first(root);
-	while (node) {
-		struct hist_entry *he;
-
-		he = rb_entry(node, struct hist_entry, rb_node_in);
-
-		pr_info("%2d: entry: %-8s [%-8s] %20s: period = %"PRIu64"\n",
-			i, thread__comm_str(he->thread), he->ms.map->dso->short_name,
-			he->ms.sym->name, he->stat.period);
-
-		i++;
-		node = rb_next(node);
-	}
-}
-
 int test__hists_link(void)
 {
 	int err = -1;
@@ -336,7 +309,7 @@ int test__hists_link(void)
 		hists__collapse_resort(&evsel->hists, NULL);
 
 		if (verbose > 2)
-			print_hists(&evsel->hists);
+			print_hists_in(&evsel->hists);
 	}
 
 	first = perf_evlist__first(evlist);
Signature · 1 line
1.9.2
author maintainer
#38 lore raw

PATCH 20/20 perf tests: Add a testcase for histogram output sorting

In reply to Namhyung Kim

With new output fields option, its internal implementation was changed so add a new testcase to verify whether it breaks things.

---
 tools/perf/Makefile.perf        |   1 +
 tools/perf/tests/builtin-test.c |   4 +
 tools/perf/tests/hists_common.c |   4 +-
 tools/perf/tests/hists_output.c | 628 ++++++++++++++++++++++++++++++++++++++++
 tools/perf/tests/tests.h        |   1 +
 5 files changed, 636 insertions(+), 2 deletions(-)
 create mode 100644 tools/perf/tests/hists_output.c
tools/perf/Makefile.perf +1 -0
index 2baf61cec7ff..25a5d46eb08c 100644
@@ -399,6 +399,7 @@ LIB_OBJS += $(OUTPUT)tests/pmu.o
 LIB_OBJS += $(OUTPUT)tests/hists_common.o
 LIB_OBJS += $(OUTPUT)tests/hists_link.o
 LIB_OBJS += $(OUTPUT)tests/hists_filter.o
+LIB_OBJS += $(OUTPUT)tests/hists_output.o
 LIB_OBJS += $(OUTPUT)tests/python-use.o
 LIB_OBJS += $(OUTPUT)tests/bp_signal.o
 LIB_OBJS += $(OUTPUT)tests/bp_signal_overflow.o
tools/perf/tests/builtin-test.c +4 -0
index 0d5afaf72944..6f39cb80fdc2 100644
@@ -136,6 +136,10 @@ static struct test {
 		.func = test__thread_mg_share,
 	},
 	{
+		.desc = "Test output sorting of hist entries",
+		.func = test__hists_output,
+	},
+	{
 		.func = NULL,
 	},
 };
tools/perf/tests/hists_common.c +2 -2
index 040a85b17aee..e4e01aadc3be 100644
@@ -193,8 +193,8 @@ void print_hists_out(struct hists *hists)
 		he = rb_entry(node, struct hist_entry, rb_node);
 
 		if (!he->filtered) {
-			pr_info("%2d: entry: %-8s [%-8s] %20s: period = %"PRIu64"\n",
-				i, thread__comm_str(he->thread),
+			pr_info("%2d: entry: %8s:%5d [%-8s] %20s: period = %"PRIu64"\n",
+				i, thread__comm_str(he->thread), he->thread->tid,
 				he->ms.map->dso->short_name,
 				he->ms.sym->name, he->stat.period);
 		}
tools/perf/tests/hists_output.c new file +628 -0

628 lines not on this page · raw message

tools/perf/tests/tests.h +1 -0
index a9d7cb019f9e..edad4cfc5e06 100644
@@ -44,6 +44,7 @@ int test__dwarf_unwind(void);
 int test__hists_filter(void);
 int test__mmap_thread_lookup(void);
 int test__thread_mg_share(void);
+int test__hists_output(void);
 
 #if defined(__x86_64__) || defined(__i386__)
 #ifdef HAVE_DWARF_UNWIND_SUPPORT
Signature · 1 line
1.9.2
#39 lore raw
In reply to Namhyung Kim

On Mon, May 12, 2014 at 03:28:33PM +0900, Namhyung Kim wrote:

SNIP

Quoted 28 lines
  $ perf report -F sym,sample -s overhead
  ...
  #                     Symbol       Samples  Overhead
  # ..........................  ............  ........
  #
    [.] strcmp                            11    13.75%
    [.] a                                  8    10.00%
    [.] b                                  8    10.00%
    [.] c                                  8    10.00%
    [.] main                               7     8.75%
    [.] _setjmp                            6     7.50%
    [.] _init                              5     6.25%
    [.] frame_dummy                        5     6.25%
    [.] __libc_csu_init                    4     5.00%
    [.] _dl_name_match_p                   4     5.00%
    [.] __new_exitfn                       3     3.75%
    [.] __cxa_atexit                       2     2.50%
    [.] _dl_check_map_versions             1     1.25%
    [.] _dl_setup_hash                     1     1.25%
    [.] _dl_sysdep_start                   1     1.25%
    [.] brk                                1     1.25%
    [.] calloc@plt                         1     1.25%
    [.] dl_main                            1     1.25%
    [.] match_symbol                       1     1.25%
    [.] sbrk                               1     1.25%
    [.] strlen                             1     1.25%
Quoted 2 lines

* changes in v5:
- add a testcase for hist output sorting

cool, had just few questions.. otherwise it seems ok to me

thanks,
jirka

#40 lore raw
Quoted 32 lines · Jiri Olsa

On Mon, May 12, 2014 at 03:28:33PM +0900, Namhyung Kim wrote:

SNIP

Quoted 28 lines
  $ perf report -F sym,sample -s overhead
  ...
  #                     Symbol       Samples  Overhead
  # ..........................  ............  ........
  #
    [.] strcmp                            11    13.75%
    [.] a                                  8    10.00%
    [.] b                                  8    10.00%
    [.] c                                  8    10.00%
    [.] main                               7     8.75%
    [.] _setjmp                            6     7.50%
    [.] _init                              5     6.25%
    [.] frame_dummy                        5     6.25%
    [.] __libc_csu_init                    4     5.00%
    [.] _dl_name_match_p                   4     5.00%
    [.] __new_exitfn                       3     3.75%
    [.] __cxa_atexit                       2     2.50%
    [.] _dl_check_map_versions             1     1.25%
    [.] _dl_setup_hash                     1     1.25%
    [.] _dl_sysdep_start                   1     1.25%
    [.] brk                                1     1.25%
    [.] calloc@plt                         1     1.25%
    [.] dl_main                            1     1.25%
    [.] match_symbol                       1     1.25%
    [.] sbrk                               1     1.25%
    [.] strlen                             1     1.25%
Quoted 4 lines
Quoted 2 lines

* changes in v5:
- add a testcase for hist output sorting

cool, had just few questions.. otherwise it seems ok to me

Thanks for your review. I'll address your points and resend v6 next week.

Thanks,
Namhyung