Skip to contentSkip to conversation

perf hists: Changes on hists__{match,link} (v4)

PATCH In review Namhyung Kim 4 participants 13 messages Dec 10, 2012 – Jan 25, 2013 perf events

lore
Share & tools
lore

Namhyung Kim

author maintainer
#0 lore raw

cover perf hists: Changes on hists__{match,link} (v4)

Hi,

This is a new version of hists link patchset.

In this version, I've addressed Jiri's comment and added his acked-by's. Also added a comment on hist_entry__cmp() in order to clarify the reason of the change of the argument order exchange as Arnaldo suggested. Hoping that I didn't miss any of remaining issues on this.

I also pushed it on 'perf/link-v4' branch at:

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

Please have a look, thanks.
Namhyung

Namhyung Kim (4):
perf hists: Exchange order of comparing items when collapsing hists perf hists: Link hist entries before inserting to an output tree perf diff: Use internal rb tree for compute resort
perf test: Add a test case for hists__{match,link}

 tools/perf/Makefile             |   1 +
 tools/perf/builtin-diff.c       |  92 +++-----
 tools/perf/tests/builtin-test.c |   4 +
 tools/perf/tests/hists_link.c   | 502 ++++++++++++++++++++++++++++++++++++++++
 tools/perf/tests/tests.h        |   1 +
 tools/perf/util/hist.c          |  59 +++--
 tools/perf/util/hist.h          |   1 +
 tools/perf/util/machine.h       |   1 +
 tools/perf/util/session.c       |   2 +-
 9 files changed, 592 insertions(+), 71 deletions(-)
 create mode 100644 tools/perf/tests/hists_link.c
Signature · 1 line
1.7.11.7
author maintainer
#1 lore raw

PATCH 1/4 perf hists: Exchange order of comparing items when collapsing hists

From: Namhyung Kim <[email protected]>

When comparing entries for collapsing put the given entry first, and then the iterated entry. This is not the case of hist_entry__cmp() when called if given sort keys don't require collapsing. So change the order for the sake of consistency. It will be required for matching and/or linking multiple hist entries.

Cc Jiri Olsa <[email protected]>
Cc Stephane Eranian <[email protected]>
---
 tools/perf/builtin-diff.c |  2 +-
 tools/perf/util/hist.c    | 12 +++++++++---
 2 files changed, 10 insertions(+), 4 deletions(-)
tools/perf/builtin-diff.c +1 -1
index b2e7d39f099b..4dda6f4dc618 100644
@@ -285,7 +285,7 @@ static void insert_hist_entry_by_name(struct rb_root *root,
 	while (*p != NULL) {
 		parent = *p;
 		iter = rb_entry(parent, struct hist_entry, rb_node);
-		if (hist_entry__cmp(he, iter) < 0)
+		if (hist_entry__cmp(iter, he) < 0)
 			p = &(*p)->rb_left;
 		else
 			p = &(*p)->rb_right;
tools/perf/util/hist.c +9 -3
index 82df1b26f0d4..3a9ccd09835d 100644
@@ -285,7 +285,13 @@ static struct hist_entry *add_hist_entry(struct hists *hists,
 		parent = *p;
 		he = rb_entry(parent, struct hist_entry, rb_node_in);
 
-		cmp = hist_entry__cmp(entry, he);
+		/*
+		 * Make sure that it receives arguments in a same order as
+		 * hist_entry__collapse() so that we can use an appropriate
+		 * function when searching an entry regardless which sort
+		 * keys were used.
+		 */
+		cmp = hist_entry__cmp(he, entry);
 
 		if (!cmp) {
 			he_stat__add_period(&he->stat, period);
@@ -729,7 +735,7 @@ static struct hist_entry *hists__add_dummy_entry(struct hists *hists,
 		parent = *p;
 		he = rb_entry(parent, struct hist_entry, rb_node);
 
-		cmp = hist_entry__cmp(pair, he);
+		cmp = hist_entry__cmp(he, pair);
 
 		if (!cmp)
 			goto out;
@@ -759,7 +765,7 @@ static struct hist_entry *hists__find_entry(struct hists *hists,
 
 	while (n) {
 		struct hist_entry *iter = rb_entry(n, struct hist_entry, rb_node);
-		int64_t cmp = hist_entry__cmp(he, iter);
+		int64_t cmp = hist_entry__cmp(iter, he);
 
 		if (cmp < 0)
 			n = n->rb_left;
Signature · 1 line
1.7.11.7
Acked-by on v1 1/4
#2 lore raw

Re: [PATCH 1/4] perf hists: Exchange order of comparing items when collapsing hists

Quoted 11 lines · Namhyung Kim

From: Namhyung Kim <[email protected]>

When comparing entries for collapsing put the given entry first, and then the iterated entry. This is not the case of hist_entry__cmp() when called if given sort keys don't require collapsing. So change the order for the sake of consistency. It will be required for matching and/or linking multiple hist entries.

Cc Jiri Olsa <[email protected]>
Cc Stephane Eranian <[email protected]>
#3 lore raw

Re: [PATCH 1/4] perf hists: Exchange order of comparing items when collapsing hists

Em Mon, Dec 10, 2012 at 02:44:34PM +0100, Jiri Olsa escreveu:

Quoted 8 lines · Namhyung Kim
Quoted 5 lines

When comparing entries for collapsing put the given entry first, and then the iterated entry. This is not the case of hist_entry__cmp() when called if given sort keys don't require collapsing. So change the order for the sake of consistency. It will be required for matching and/or linking multiple hist entries.

Hey,

Before:

[root@sandy ~]# perf diff
# Event 'cycles'
#
# Baseline Delta Shared Object Symbol

# ........  .......  .................  ..................................
#
    17.41%  -17.41%  [kernel.kallsyms]  [k] memset                        
     0.38%   -0.38%  [kernel.kallsyms]  [k] strlcpy                       
            +48.19%  [kernel.kallsyms]  [k] __blocking_notifier_call_chain
            +46.96%  [kernel.kallsyms]  [k] radix_tree_lookup_slot        
             +4.66%  [kernel.kallsyms]  [k] do_close_on_exec              
    16.17%  -16.17%  [kernel.kallsyms]  [k] final_putname                 
     4.17%   -4.17%  [kernel.kallsyms]  [k] shift_arg_pages               
    30.74%  -30.74%  [kernel.kallsyms]  [k] kfree                         
    14.93%  -14.93%  [kernel.kallsyms]  [k] unlock_page                   
     0.03%   +0.15%  [kernel.kallsyms]  [k] native_write_msr_safe         
    16.17%  -16.17%  libc-2.12.so       [.] _dl_addr                      

After:

[root@sandy ~]# perf diff
# Event 'cycles'
#
# Baseline Delta Shared Object Symbol

# ........  .......  .................  ..................................
#
    16.17%  -16.17%  libc-2.12.so       [.] _dl_addr                      
            +48.19%  [kernel.kallsyms]  [k] __blocking_notifier_call_chain
    14.93%  -14.93%  [kernel.kallsyms]  [k] unlock_page                   
    30.74%  -30.74%  [kernel.kallsyms]  [k] kfree                         
     4.17%   -4.17%  [kernel.kallsyms]  [k] shift_arg_pages               
    16.17%  -16.17%  [kernel.kallsyms]  [k] final_putname                 
            +46.96%  [kernel.kallsyms]  [k] radix_tree_lookup_slot        
             +4.66%  [kernel.kallsyms]  [k] do_close_on_exec              
     0.03%   +0.15%  [kernel.kallsyms]  [k] native_write_msr_safe         
     0.38%   -0.38%  [kernel.kallsyms]  [k] strlcpy                       
    17.41%  -17.41%  [kernel.kallsyms]  [k] memset                        
[root@sandy ~]#

Got inverted.

The order was arbitrary, and was arbitrarily reversed, are you guys sure this is the only side effect?

Jiri, IIRC you mentioned that there would be some patch with an --order switch, would that allow sorting by the entry that had the biggest change, etc?

Holding on this for now.

- Arnaldo

#4 lore raw

Re: [PATCH 1/4] perf hists: Exchange order of comparing items when collapsing hists

Quoted 53 lines · Arnaldo Carvalho de Melo

Em Mon, Dec 10, 2012 at 02:44:34PM +0100, Jiri Olsa escreveu:

Quoted 8 lines · Namhyung Kim
Quoted 5 lines

When comparing entries for collapsing put the given entry first, and then the iterated entry. This is not the case of hist_entry__cmp() when called if given sort keys don't require collapsing. So change the order for the sake of consistency. It will be required for matching and/or linking multiple hist entries.

Hey,

Before:

[root@sandy ~]# perf diff
# Event 'cycles'
#
# Baseline Delta Shared Object Symbol

# ........  .......  .................  ..................................
#
    17.41%  -17.41%  [kernel.kallsyms]  [k] memset                        
     0.38%   -0.38%  [kernel.kallsyms]  [k] strlcpy                       
            +48.19%  [kernel.kallsyms]  [k] __blocking_notifier_call_chain
            +46.96%  [kernel.kallsyms]  [k] radix_tree_lookup_slot        
             +4.66%  [kernel.kallsyms]  [k] do_close_on_exec              
    16.17%  -16.17%  [kernel.kallsyms]  [k] final_putname                 
     4.17%   -4.17%  [kernel.kallsyms]  [k] shift_arg_pages               
    30.74%  -30.74%  [kernel.kallsyms]  [k] kfree                         
    14.93%  -14.93%  [kernel.kallsyms]  [k] unlock_page                   
     0.03%   +0.15%  [kernel.kallsyms]  [k] native_write_msr_safe         
    16.17%  -16.17%  libc-2.12.so       [.] _dl_addr                      

After:

[root@sandy ~]# perf diff
# Event 'cycles'
#
# Baseline Delta Shared Object Symbol

# ........  .......  .................  ..................................
#
    16.17%  -16.17%  libc-2.12.so       [.] _dl_addr                      
            +48.19%  [kernel.kallsyms]  [k] __blocking_notifier_call_chain
    14.93%  -14.93%  [kernel.kallsyms]  [k] unlock_page                   
    30.74%  -30.74%  [kernel.kallsyms]  [k] kfree                         
     4.17%   -4.17%  [kernel.kallsyms]  [k] shift_arg_pages               
    16.17%  -16.17%  [kernel.kallsyms]  [k] final_putname                 
            +46.96%  [kernel.kallsyms]  [k] radix_tree_lookup_slot        
             +4.66%  [kernel.kallsyms]  [k] do_close_on_exec              
     0.03%   +0.15%  [kernel.kallsyms]  [k] native_write_msr_safe         
     0.38%   -0.38%  [kernel.kallsyms]  [k] strlcpy                       
    17.41%  -17.41%  [kernel.kallsyms]  [k] memset                        
[root@sandy ~]#
Quoted 4 lines

Got inverted.

The order was arbitrary, and was arbitrarily reversed, are you guys sure this is the only side effect?

right,

currently (without namhyung's changes) there's initial 'name' sorting broken by hists_link

after the change, it gets sorted 'but' there are dummy entries added which brake the sort again.. this is sorted out within the multi diff patches

so the bottom line is that currently the sorting is broken anyway, and the fix is comming in upcomming patchset ;)

Quoted 3 lines

Jiri, IIRC you mentioned that there would be some patch with an --order switch, would that allow sorting by the entry that had the biggest change, etc?

by default it'll sorted by baseline and you can specify -o X to sort by specified file

thanks,
jirka

#5 lore raw

Re: [PATCH 1/4] perf hists: Exchange order of comparing items when collapsing hists

Em Wed, Dec 12, 2012 at 04:56:32PM +0100, Jiri Olsa escreveu:

Quoted 11 lines · Arnaldo Carvalho de Melo
Quoted 4 lines

Got inverted.

The order was arbitrary, and was arbitrarily reversed, are you guys sure this is the only side effect?

right,

currently (without namhyung's changes) there's initial 'name' sorting broken by hists_link

Quoted 6 lines

after the change, it gets sorted 'but' there are dummy entries added which brake the sort again.. this is sorted out within the multi diff patches

so the bottom line is that currently the sorting is broken anyway, and the fix is comming in upcomming patchset ;)

Ok, applied to my perf/core branch and pushed out.

Quoted 5 lines
Quoted 2 lines

Jiri, IIRC you mentioned that there would be some patch with an --order switch, would that allow sorting by the entry that had the biggest change, etc?

by default it'll sorted by baseline and you can specify -o X to sort by specified file

great.

- Arnaldo

#6 lore raw

[tip:perf/core] perf hists: Exchange order of comparing items when collapsing hists

In reply to Namhyung Kim
Commit-ID:  9afcf930b1fa1158b0878afeba3eff299300dc65
Gitweb:     http://git.kernel.org/tip/9afcf930b1fa1158b0878afeba3eff299300dc65
Author:     Namhyung Kim <[email protected]>
AuthorDate: Mon, 10 Dec 2012 17:29:54 +0900
Committer:  Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Thu, 24 Jan 2013 16:40:05 -0300

perf hists: Exchange order of comparing items when collapsing hists

When comparing entries for collapsing put the given entry first, and then the iterated entry. This is not the case of hist_entry__cmp() when called if given sort keys don't require collapsing. So change the order for the sake of consistency. It will be required for matching and/or linking multiple hist entries.

---
 tools/perf/builtin-diff.c |  2 +-
 tools/perf/util/hist.c    | 12 +++++++++---
 2 files changed, 10 insertions(+), 4 deletions(-)
tools/perf/builtin-diff.c +1 -1
index b2e7d39..4dda6f4 100644
@@ -285,7 +285,7 @@ static void insert_hist_entry_by_name(struct rb_root *root,
 	while (*p != NULL) {
 		parent = *p;
 		iter = rb_entry(parent, struct hist_entry, rb_node);
-		if (hist_entry__cmp(he, iter) < 0)
+		if (hist_entry__cmp(iter, he) < 0)
 			p = &(*p)->rb_left;
 		else
 			p = &(*p)->rb_right;
tools/perf/util/hist.c +9 -3
index 82df1b2..3a9ccd0 100644
@@ -285,7 +285,13 @@ static struct hist_entry *add_hist_entry(struct hists *hists,
 		parent = *p;
 		he = rb_entry(parent, struct hist_entry, rb_node_in);
 
-		cmp = hist_entry__cmp(entry, he);
+		/*
+		 * Make sure that it receives arguments in a same order as
+		 * hist_entry__collapse() so that we can use an appropriate
+		 * function when searching an entry regardless which sort
+		 * keys were used.
+		 */
+		cmp = hist_entry__cmp(he, entry);
 
 		if (!cmp) {
 			he_stat__add_period(&he->stat, period);
@@ -729,7 +735,7 @@ static struct hist_entry *hists__add_dummy_entry(struct hists *hists,
 		parent = *p;
 		he = rb_entry(parent, struct hist_entry, rb_node);
 
-		cmp = hist_entry__cmp(pair, he);
+		cmp = hist_entry__cmp(he, pair);
 
 		if (!cmp)
 			goto out;
@@ -759,7 +765,7 @@ static struct hist_entry *hists__find_entry(struct hists *hists,
 
 	while (n) {
 		struct hist_entry *iter = rb_entry(n, struct hist_entry, rb_node);
-		int64_t cmp = hist_entry__cmp(he, iter);
+		int64_t cmp = hist_entry__cmp(iter, he);
 
 		if (cmp < 0)
 			n = n->rb_left;
author maintainer
#7 lore raw

PATCH 2/4 perf hists: Link hist entries before inserting to an output tree

In reply to Namhyung Kim

From: Namhyung Kim <[email protected]>

For matching and/or linking hist entries, they need to be sorted by given sort keys. However current hists__match/link did this on the output trees, so that the entries in the output tree need to be resort before doing it.

This looks not so good since we have trees for collecting or collapsing entries before passing them to an output tree and they're already sorted by the given sort keys. Since we don't need to print anything at the time of matching/linking, we can use these internal trees directly instead of bothering with double resort on the output tree.

Its only user - at the time of this writing - perf diff can be easily converted to use the internal tree and can save some lines too by getting rid of unnecessary resorting codes.

---
 tools/perf/builtin-diff.c | 65 +++++++++++++----------------------------------
 tools/perf/util/hist.c    | 49 ++++++++++++++++++++++++++---------
 2 files changed, 54 insertions(+), 60 deletions(-)
tools/perf/builtin-diff.c +17 -48
index 4dda6f4dc618..8b896f5eded0 100644
@@ -275,43 +275,6 @@ static struct perf_tool tool = {
 	.ordering_requires_timestamps = true,
 };
 
-static void insert_hist_entry_by_name(struct rb_root *root,
-				      struct hist_entry *he)
-{
-	struct rb_node **p = &root->rb_node;
-	struct rb_node *parent = NULL;
-	struct hist_entry *iter;
-
-	while (*p != NULL) {
-		parent = *p;
-		iter = rb_entry(parent, struct hist_entry, rb_node);
-		if (hist_entry__cmp(iter, he) < 0)
-			p = &(*p)->rb_left;
-		else
-			p = &(*p)->rb_right;
-	}
-
-	rb_link_node(&he->rb_node, parent, p);
-	rb_insert_color(&he->rb_node, root);
-}
-
-static void hists__name_resort(struct hists *self)
-{
-	struct rb_root tmp = RB_ROOT;
-	struct rb_node *next = rb_first(&self->entries);
-
-	while (next != NULL) {
-		struct hist_entry *n = rb_entry(next, struct hist_entry, rb_node);
-
-		next = rb_next(&n->rb_node);
-
-		rb_erase(&n->rb_node, &self->entries);
-		insert_hist_entry_by_name(&tmp, n);
-	}
-
-	self->entries = tmp;
-}
-
 static struct perf_evsel *evsel_match(struct perf_evsel *evsel,
 				      struct perf_evlist *evlist)
 {
@@ -324,30 +287,34 @@ static struct perf_evsel *evsel_match(struct perf_evsel *evsel,
 	return NULL;
 }
 
-static void perf_evlist__resort_hists(struct perf_evlist *evlist, bool name)
+static void perf_evlist__collapse_resort(struct perf_evlist *evlist)
 {
 	struct perf_evsel *evsel;
 
 	list_for_each_entry(evsel, &evlist->entries, node) {
 		struct hists *hists = &evsel->hists;
 
-		hists__output_resort(hists);
-
-		if (name)
-			hists__name_resort(hists);
+		hists__collapse_resort(hists);
 	}
 }
 
 static void hists__baseline_only(struct hists *hists)
 {
-	struct rb_node *next = rb_first(&hists->entries);
+	struct rb_root *root;
+	struct rb_node *next;
+
+	if (sort__need_collapse)
+		root = &hists->entries_collapsed;
+	else
+		root = hists->entries_in;
 
+	next = rb_first(root);
 	while (next != NULL) {
-		struct hist_entry *he = rb_entry(next, struct hist_entry, rb_node);
+		struct hist_entry *he = rb_entry(next, struct hist_entry, rb_node_in);
 
-		next = rb_next(&he->rb_node);
+		next = rb_next(&he->rb_node_in);
 		if (!hist_entry__next_pair(he)) {
-			rb_erase(&he->rb_node, &hists->entries);
+			rb_erase(&he->rb_node_in, root);
 			hist_entry__free(he);
 		}
 	}
@@ -471,6 +438,8 @@ static void hists__process(struct hists *old, struct hists *new)
 	else
 		hists__link(new, old);
 
+	hists__output_resort(new);
+
 	if (sort_compute) {
 		hists__precompute(new);
 		hists__compute_resort(new);
@@ -505,8 +474,8 @@ static int __cmd_diff(void)
 	evlist_old = older->evlist;
 	evlist_new = newer->evlist;
 
-	perf_evlist__resort_hists(evlist_old, true);
-	perf_evlist__resort_hists(evlist_new, false);
+	perf_evlist__collapse_resort(evlist_old);
+	perf_evlist__collapse_resort(evlist_new);
 
 	list_for_each_entry(evsel, &evlist_new->entries, node) {
 		struct perf_evsel *evsel_old;
tools/perf/util/hist.c +37 -12

94 lines not on this page · raw message

Signature · 1 line
1.7.11.7
#8 lore raw

[tip:perf/core] perf hists: Link hist entries before inserting to an output tree

Commit-ID:  ce74f60eab3cc8b7a3b0cb9c29ec9b1e1abac7d2
Gitweb:     http://git.kernel.org/tip/ce74f60eab3cc8b7a3b0cb9c29ec9b1e1abac7d2
Author:     Namhyung Kim <[email protected]>
AuthorDate: Mon, 10 Dec 2012 17:29:55 +0900
Committer:  Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Thu, 24 Jan 2013 16:40:06 -0300

perf hists: Link hist entries before inserting to an output tree

For matching and/or linking hist entries, they need to be sorted by given sort keys. However current hists__match/link did this on the output trees, so that the entries in the output tree need to be resort before doing it.

This looks not so good since we have trees for collecting or collapsing entries before passing them to an output tree and they're already sorted by the given sort keys. Since we don't need to print anything at the time of matching/linking, we can use these internal trees directly instead of bothering with double resort on the output tree.

Its only user - at the time of this writing - perf diff can be easily converted to use the internal tree and can save some lines too by getting rid of unnecessary resorting codes.

---
 tools/perf/builtin-diff.c | 65 +++++++++++++----------------------------------
 tools/perf/util/hist.c    | 49 ++++++++++++++++++++++++++---------
 2 files changed, 54 insertions(+), 60 deletions(-)
tools/perf/builtin-diff.c +17 -48
index 4dda6f4..8b896f5 100644
@@ -275,43 +275,6 @@ static struct perf_tool tool = {
 	.ordering_requires_timestamps = true,
 };
 
-static void insert_hist_entry_by_name(struct rb_root *root,
-				      struct hist_entry *he)
-{
-	struct rb_node **p = &root->rb_node;
-	struct rb_node *parent = NULL;
-	struct hist_entry *iter;
-
-	while (*p != NULL) {
-		parent = *p;
-		iter = rb_entry(parent, struct hist_entry, rb_node);
-		if (hist_entry__cmp(iter, he) < 0)
-			p = &(*p)->rb_left;
-		else
-			p = &(*p)->rb_right;
-	}
-
-	rb_link_node(&he->rb_node, parent, p);
-	rb_insert_color(&he->rb_node, root);
-}
-
-static void hists__name_resort(struct hists *self)
-{
-	struct rb_root tmp = RB_ROOT;
-	struct rb_node *next = rb_first(&self->entries);
-
-	while (next != NULL) {
-		struct hist_entry *n = rb_entry(next, struct hist_entry, rb_node);
-
-		next = rb_next(&n->rb_node);
-
-		rb_erase(&n->rb_node, &self->entries);
-		insert_hist_entry_by_name(&tmp, n);
-	}
-
-	self->entries = tmp;
-}
-
 static struct perf_evsel *evsel_match(struct perf_evsel *evsel,
 				      struct perf_evlist *evlist)
 {
@@ -324,30 +287,34 @@ static struct perf_evsel *evsel_match(struct perf_evsel *evsel,
 	return NULL;
 }
 
-static void perf_evlist__resort_hists(struct perf_evlist *evlist, bool name)
+static void perf_evlist__collapse_resort(struct perf_evlist *evlist)
 {
 	struct perf_evsel *evsel;
 
 	list_for_each_entry(evsel, &evlist->entries, node) {
 		struct hists *hists = &evsel->hists;
 
-		hists__output_resort(hists);
-
-		if (name)
-			hists__name_resort(hists);
+		hists__collapse_resort(hists);
 	}
 }
 
 static void hists__baseline_only(struct hists *hists)
 {
-	struct rb_node *next = rb_first(&hists->entries);
+	struct rb_root *root;
+	struct rb_node *next;
+
+	if (sort__need_collapse)
+		root = &hists->entries_collapsed;
+	else
+		root = hists->entries_in;
 
+	next = rb_first(root);
 	while (next != NULL) {
-		struct hist_entry *he = rb_entry(next, struct hist_entry, rb_node);
+		struct hist_entry *he = rb_entry(next, struct hist_entry, rb_node_in);
 
-		next = rb_next(&he->rb_node);
+		next = rb_next(&he->rb_node_in);
 		if (!hist_entry__next_pair(he)) {
-			rb_erase(&he->rb_node, &hists->entries);
+			rb_erase(&he->rb_node_in, root);
 			hist_entry__free(he);
 		}
 	}
@@ -471,6 +438,8 @@ static void hists__process(struct hists *old, struct hists *new)
 	else
 		hists__link(new, old);
 
+	hists__output_resort(new);
+
 	if (sort_compute) {
 		hists__precompute(new);
 		hists__compute_resort(new);
@@ -505,8 +474,8 @@ static int __cmd_diff(void)
 	evlist_old = older->evlist;
 	evlist_new = newer->evlist;
 
-	perf_evlist__resort_hists(evlist_old, true);
-	perf_evlist__resort_hists(evlist_new, false);
+	perf_evlist__collapse_resort(evlist_old);
+	perf_evlist__collapse_resort(evlist_new);
 
 	list_for_each_entry(evsel, &evlist_new->entries, node) {
 		struct perf_evsel *evsel_old;
tools/perf/util/hist.c +37 -12

94 lines not on this page · raw message

author maintainer
#9 lore raw

PATCH 3/4 perf diff: Use internal rb tree for compute resort

In reply to Namhyung Kim

From: Namhyung Kim <[email protected]>

There's no reason to run hists_compute_resort() using output tree. Convert it to use internal tree so that it can remove unnecessary _output_resort.

---
 tools/perf/builtin-diff.c | 31 +++++++++++++++++++++----------
 tools/perf/util/hist.c    |  2 +-
 tools/perf/util/hist.h    |  1 +
 3 files changed, 23 insertions(+), 11 deletions(-)
tools/perf/builtin-diff.c +21 -10
index 8b896f5eded0..4af0b580b046 100644
@@ -414,19 +414,30 @@ static void insert_hist_entry_by_compute(struct rb_root *root,
 
 static void hists__compute_resort(struct hists *hists)
 {
-	struct rb_root tmp = RB_ROOT;
-	struct rb_node *next = rb_first(&hists->entries);
+	struct rb_root *root;
+	struct rb_node *next;
+
+	if (sort__need_collapse)
+		root = &hists->entries_collapsed;
+	else
+		root = hists->entries_in;
+
+	hists->entries = RB_ROOT;
+	next = rb_first(root);
+
+	hists->nr_entries = 0;
+	hists->stats.total_period = 0;
+	hists__reset_col_len(hists);
 
 	while (next != NULL) {
-		struct hist_entry *he = rb_entry(next, struct hist_entry, rb_node);
+		struct hist_entry *he;
 
-		next = rb_next(&he->rb_node);
+		he = rb_entry(next, struct hist_entry, rb_node_in);
+		next = rb_next(&he->rb_node_in);
 
-		rb_erase(&he->rb_node, &hists->entries);
-		insert_hist_entry_by_compute(&tmp, he, compute);
+		insert_hist_entry_by_compute(&hists->entries, he, compute);
+		hists__inc_nr_entries(hists, he);
 	}
-
-	hists->entries = tmp;
 }
 
 static void hists__process(struct hists *old, struct hists *new)
@@ -438,11 +449,11 @@ static void hists__process(struct hists *old, struct hists *new)
 	else
 		hists__link(new, old);
 
-	hists__output_resort(new);
-
 	if (sort_compute) {
 		hists__precompute(new);
 		hists__compute_resort(new);
+	} else {
+		hists__output_resort(new);
 	}
 
 	hists__fprintf(new, true, 0, 0, stdout);
tools/perf/util/hist.c +1 -1
index 8ff3c2f8a5dd..37179af74409 100644
@@ -251,7 +251,7 @@ static struct hist_entry *hist_entry__new(struct hist_entry *template)
 	return he;
 }
 
-static void hists__inc_nr_entries(struct hists *hists, struct hist_entry *h)
+void hists__inc_nr_entries(struct hists *hists, struct hist_entry *h)
 {
 	if (!h->filtered) {
 		hists__calc_col_len(hists, h);
tools/perf/util/hist.h +1 -0
index 5b3b0075be64..d3664abea6d6 100644
@@ -96,6 +96,7 @@ void hists__decay_entries_threaded(struct hists *hists, bool zap_user,
 				   bool zap_kernel);
 void hists__output_recalc_col_len(struct hists *hists, int max_rows);
 
+void hists__inc_nr_entries(struct hists *hists, struct hist_entry *h);
 void hists__inc_nr_events(struct hists *self, u32 type);
 size_t hists__fprintf_nr_events(struct hists *self, FILE *fp);
 
Signature · 1 line
1.7.11.7
#10 lore raw

[tip:perf/core] perf diff: Use internal rb tree for compute resort

Commit-ID:  66f97ed3ac44c24958171bbc5cc04896147752b7
Gitweb:     http://git.kernel.org/tip/66f97ed3ac44c24958171bbc5cc04896147752b7
Author:     Namhyung Kim <[email protected]>
AuthorDate: Mon, 10 Dec 2012 17:29:56 +0900
Committer:  Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Thu, 24 Jan 2013 16:40:06 -0300

perf diff: Use internal rb tree for compute resort

There's no reason to run hists_compute_resort() using output tree. Convert it to use internal tree so that it can remove unnecessary _output_resort.

---
 tools/perf/builtin-diff.c | 31 +++++++++++++++++++++----------
 tools/perf/util/hist.c    |  2 +-
 tools/perf/util/hist.h    |  1 +
 3 files changed, 23 insertions(+), 11 deletions(-)
tools/perf/builtin-diff.c +21 -10
index 8b896f5..4af0b58 100644
@@ -414,19 +414,30 @@ static void insert_hist_entry_by_compute(struct rb_root *root,
 
 static void hists__compute_resort(struct hists *hists)
 {
-	struct rb_root tmp = RB_ROOT;
-	struct rb_node *next = rb_first(&hists->entries);
+	struct rb_root *root;
+	struct rb_node *next;
+
+	if (sort__need_collapse)
+		root = &hists->entries_collapsed;
+	else
+		root = hists->entries_in;
+
+	hists->entries = RB_ROOT;
+	next = rb_first(root);
+
+	hists->nr_entries = 0;
+	hists->stats.total_period = 0;
+	hists__reset_col_len(hists);
 
 	while (next != NULL) {
-		struct hist_entry *he = rb_entry(next, struct hist_entry, rb_node);
+		struct hist_entry *he;
 
-		next = rb_next(&he->rb_node);
+		he = rb_entry(next, struct hist_entry, rb_node_in);
+		next = rb_next(&he->rb_node_in);
 
-		rb_erase(&he->rb_node, &hists->entries);
-		insert_hist_entry_by_compute(&tmp, he, compute);
+		insert_hist_entry_by_compute(&hists->entries, he, compute);
+		hists__inc_nr_entries(hists, he);
 	}
-
-	hists->entries = tmp;
 }
 
 static void hists__process(struct hists *old, struct hists *new)
@@ -438,11 +449,11 @@ static void hists__process(struct hists *old, struct hists *new)
 	else
 		hists__link(new, old);
 
-	hists__output_resort(new);
-
 	if (sort_compute) {
 		hists__precompute(new);
 		hists__compute_resort(new);
+	} else {
+		hists__output_resort(new);
 	}
 
 	hists__fprintf(new, true, 0, 0, stdout);
tools/perf/util/hist.c +1 -1
index 8ff3c2f..37179af 100644
@@ -251,7 +251,7 @@ static struct hist_entry *hist_entry__new(struct hist_entry *template)
 	return he;
 }
 
-static void hists__inc_nr_entries(struct hists *hists, struct hist_entry *h)
+void hists__inc_nr_entries(struct hists *hists, struct hist_entry *h)
 {
 	if (!h->filtered) {
 		hists__calc_col_len(hists, h);
tools/perf/util/hist.h +1 -0
index 5b3b007..d3664ab 100644
@@ -96,6 +96,7 @@ void hists__decay_entries_threaded(struct hists *hists, bool zap_user,
 				   bool zap_kernel);
 void hists__output_recalc_col_len(struct hists *hists, int max_rows);
 
+void hists__inc_nr_entries(struct hists *hists, struct hist_entry *h);
 void hists__inc_nr_events(struct hists *self, u32 type);
 size_t hists__fprintf_nr_events(struct hists *self, FILE *fp);
 
author maintainer
#11 lore raw

PATCH 4/4 perf test: Add a test case for hists__{match,link}

In reply to Namhyung Kim

From: Namhyung Kim <[email protected]>

As they are used from diff and event group report, add a test case to verify their behaviors.

In this test I made a fake machine and two evsel. Each evsel got 10 samples (so hist entries) - 5 are common and the rests are not. So after hists__match() both of them will have 5 entries with pair set.

And the second evsel has a collapsed entry so that the total number is 9 - I made it in order to simulate more realistic case. Thus after hists__link the first entry will have 14 entries - 5 are common (w/ pair), 5 are unmatch (w/o pair) and 4 are dummy (w/ pair). And the second entry will have 9 entries all have its pair.

Cc Jiri Olsa <[email protected]>
Cc Stephane Eranian <[email protected]>
---
 tools/perf/Makefile             |   1 +
 tools/perf/tests/builtin-test.c |   4 +
 tools/perf/tests/hists_link.c   | 502 ++++++++++++++++++++++++++++++++++++++++
 tools/perf/tests/tests.h        |   1 +
 tools/perf/util/machine.h       |   1 +
 tools/perf/util/session.c       |   2 +-
 6 files changed, 510 insertions(+), 1 deletion(-)
 create mode 100644 tools/perf/tests/hists_link.c
tools/perf/Makefile +1 -0
index 75785bb98c30..3628065cda20 100644
@@ -460,6 +460,7 @@ LIB_OBJS += $(OUTPUT)tests/evsel-roundtrip-name.o
 LIB_OBJS += $(OUTPUT)tests/evsel-tp-sched.o
 LIB_OBJS += $(OUTPUT)tests/pmu.o
 LIB_OBJS += $(OUTPUT)tests/util.o
+LIB_OBJS += $(OUTPUT)tests/hists_link.o
 
 BUILTIN_OBJS += $(OUTPUT)builtin-annotate.o
 BUILTIN_OBJS += $(OUTPUT)builtin-bench.o
tools/perf/tests/builtin-test.c +4 -0
index 186f67535494..479d10484a74 100644
@@ -69,6 +69,10 @@ static struct test {
 		.func = test__attr,
 	},
 	{
+		.desc = "Test matching and linking mutliple hists",
+		.func = test__hists_link,
+	},
+	{
 		.func = NULL,
 	},
 };
tools/perf/tests/hists_link.c new file +502 -0

502 lines not on this page · raw message

tools/perf/tests/tests.h +1 -0
index fc121edab016..eddf1ca8cec9 100644
@@ -15,6 +15,7 @@ int test__pmu(void);
 int test__attr(void);
 int test__dso_data(void);
 int test__parse_events(void);
+int test__hists_link(void);
 
 /* Util */
 int trace_event__id(const char *evname);
tools/perf/util/machine.h +1 -0
index b7cde7467d55..166c93ccea22 100644
@@ -89,6 +89,7 @@ static inline bool machine__is_host(struct machine *machine)
 
 struct thread *machine__findnew_thread(struct machine *machine, pid_t pid);
 void machine__remove_thread(struct machine *machine, struct thread *th);
+void machine__delete_threads(struct machine *machine);
 
 size_t machine__fprintf(struct machine *machine, FILE *fp);
 
tools/perf/util/session.c +1 -1
index aa5e58255cba..3f25862e3ab9 100644
@@ -177,7 +177,7 @@ static void perf_session__delete_dead_threads(struct perf_session *session)
 	machine__delete_dead_threads(&session->host_machine);
 }
 
-static void machine__delete_threads(struct machine *self)
+void machine__delete_threads(struct machine *self)
 {
 	struct rb_node *nd = rb_first(&self->threads);
 
Signature · 1 line
1.7.11.7
#12 lore raw

[tip:perf/core] perf test: Add a test case for hists__{match, link}

Commit-ID:  f8ebb0cdf30fec2c2a5a364cdda8e6758a44026c
Gitweb:     http://git.kernel.org/tip/f8ebb0cdf30fec2c2a5a364cdda8e6758a44026c
Author:     Namhyung Kim <[email protected]>
AuthorDate: Mon, 10 Dec 2012 17:29:57 +0900
Committer:  Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Thu, 24 Jan 2013 16:40:07 -0300

perf test: Add a test case for hists__{match,link}

As they are used from diff and event group report, add a test case to verify their behaviors.

In this test I made a fake machine and two evsel. Each evsel got 10 samples (so hist entries) - 5 are common and the rests are not. So after hists__match() both of them will have 5 entries with pair set.

And the second evsel has a collapsed entry so that the total number is 9
- I made it in order to simulate more realistic case. Thus after hists__link the first entry will have 14 entries - 5 are common (w/ pair), 5 are unmatch (w/o pair) and 4 are dummy (w/ pair). And the second entry will have 9 entries all have its pair.

Cc Ingo Molnar <[email protected]>
Cc Jiri Olsa <[email protected]>
Cc Paul Mackerras <[email protected]>
Cc Peter Zijlstra <[email protected]>
Cc Stephane Eranian <[email protected]>
[ committer note: fixed up clashes with cset that moved methods to machine.h ]
---
 tools/perf/Makefile             |   1 +
 tools/perf/tests/builtin-test.c |   4 +
 tools/perf/tests/hists_link.c   | 502 ++++++++++++++++++++++++++++++++++++++++
 tools/perf/tests/tests.h        |   1 +
 4 files changed, 508 insertions(+)
tools/perf/Makefile +1 -0
index 2cbaad8..c9f72b1 100644
@@ -487,6 +487,7 @@ LIB_OBJS += $(OUTPUT)tests/rdpmc.o
 LIB_OBJS += $(OUTPUT)tests/evsel-roundtrip-name.o
 LIB_OBJS += $(OUTPUT)tests/evsel-tp-sched.o
 LIB_OBJS += $(OUTPUT)tests/pmu.o
+LIB_OBJS += $(OUTPUT)tests/hists_link.o
 
 BUILTIN_OBJS += $(OUTPUT)builtin-annotate.o
 BUILTIN_OBJS += $(OUTPUT)builtin-bench.o
tools/perf/tests/builtin-test.c +4 -0
index 186f675..479d104 100644
@@ -69,6 +69,10 @@ static struct test {
 		.func = test__attr,
 	},
 	{
+		.desc = "Test matching and linking mutliple hists",
+		.func = test__hists_link,
+	},
+	{
 		.func = NULL,
 	},
 };
tools/perf/tests/hists_link.c new file +502 -0

502 lines not on this page · raw message

tools/perf/tests/tests.h +1 -0
index 0fd94657..0bf1069 100644
@@ -15,5 +15,6 @@ int test__pmu(void);
 int test__attr(void);
 int test__dso_data(void);
 int test__parse_events(void);
+int test__hists_link(void);
 
 #endif /* TESTS_H */