aboutsummaryrefslogtreecommitdiffstats
path: root/tools
diff options
context:
space:
mode:
authorArnaldo Carvalho de Melo <acme@redhat.com>2009-12-14 17:09:31 -0500
committerIngo Molnar <mingo@elte.hu>2009-12-15 02:50:29 -0500
commit86a9eee047ba09a714c3b8e27c9df2bbf715393a (patch)
tree09c096e92b583f42480ac1dbe7dfa45eae1a7f31 /tools
parentb38d34645cc52136b6c99623fef7ded26742e573 (diff)
perf diff: Introduce tool to show performance difference
I guess it is enough to show some examples: [root@doppio linux-2.6-tip]# rm -f perf.data* [root@doppio linux-2.6-tip]# ls -la perf.data* ls: cannot access perf.data*: No such file or directory [root@doppio linux-2.6-tip]# perf record -f find / > /dev/null [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.062 MB perf.data (~2699 samples) ] [root@doppio linux-2.6-tip]# ls -la perf.data* -rw------- 1 root root 74440 2009-12-14 20:03 perf.data [root@doppio linux-2.6-tip]# perf record -f find / > /dev/null [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.062 MB perf.data (~2692 samples) ] [root@doppio linux-2.6-tip]# ls -la perf.data* -rw------- 1 root root 74280 2009-12-14 20:03 perf.data -rw------- 1 root root 74440 2009-12-14 20:03 perf.data.old [root@doppio linux-2.6-tip]# perf diff | head -5 1 -34994580 /lib64/libc-2.10.1.so _IO_vfprintf_internal 2 -15307806 [kernel.kallsyms] __kmalloc 3 +1 +3665941 /lib64/libc-2.10.1.so __GI_memmove 4 +4 +23508995 /lib64/libc-2.10.1.so _int_malloc 5 +7 +38538813 [kernel.kallsyms] __d_lookup [root@doppio linux-2.6-tip]# perf diff -p | head -5 1 +1.00% /lib64/libc-2.10.1.so _IO_vfprintf_internal 2 [kernel.kallsyms] __kmalloc 3 +1 /lib64/libc-2.10.1.so __GI_memmove 4 +4 /lib64/libc-2.10.1.so _int_malloc 5 +7 -1.00% [kernel.kallsyms] __d_lookup [root@doppio linux-2.6-tip]# perf diff -v | head -5 1 361449551 326454971 -34994580 /lib64/libc-2.10.1.so _IO_vfprintf_internal 2 151009241 135701435 -15307806 [kernel.kallsyms] __kmalloc 3 +1 101805328 105471269 +3665941 /lib64/libc-2.10.1.so __GI_memmove 4 +4 78041440 101550435 +23508995 /lib64/libc-2.10.1.so _int_malloc 5 +7 59536172 98074985 +38538813 [kernel.kallsyms] __d_lookup [root@doppio linux-2.6-tip]# perf diff -vp | head -5 1 9.00% 8.00% +1.00% /lib64/libc-2.10.1.so _IO_vfprintf_internal 2 3.00% 3.00% [kernel.kallsyms] __kmalloc 3 +1 2.00% 2.00% /lib64/libc-2.10.1.so __GI_memmove 4 +4 2.00% 2.00% /lib64/libc-2.10.1.so _int_malloc 5 +7 1.00% 2.00% -1.00% [kernel.kallsyms] __d_lookup [root@doppio linux-2.6-tip]# This should be enough for diffs where the system is non volatile, i.e. when one doesn't updates binaries. For volatile environments, stay tuned for the next perf tool feature: a buildid cache populated by 'perf record', managed by 'perf buildid-cache' a-la ccache, and used by all the report tools. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com> Cc: Stephen Hemminger <shemminger@vyatta.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> LKML-Reference: <1260828571-3613-3-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
Diffstat (limited to 'tools')
-rw-r--r--tools/perf/Documentation/perf-diff.txt31
-rw-r--r--tools/perf/Makefile1
-rw-r--r--tools/perf/builtin-diff.c288
-rw-r--r--tools/perf/builtin.h1
-rw-r--r--tools/perf/command-list.txt1
-rw-r--r--tools/perf/perf.c1
-rw-r--r--tools/perf/util/sort.h8
7 files changed, 329 insertions, 2 deletions
diff --git a/tools/perf/Documentation/perf-diff.txt b/tools/perf/Documentation/perf-diff.txt
new file mode 100644
index 000000000000..bd1ee55cef6a
--- /dev/null
+++ b/tools/perf/Documentation/perf-diff.txt
@@ -0,0 +1,31 @@
1perf-diff(1)
2==============
3
4NAME
5----
6perf-diff - Read perf.data (created by perf record) and display the profile
7
8SYNOPSIS
9--------
10[verse]
11'perf diff' [oldfile] [newfile]
12
13DESCRIPTION
14-----------
15This command displays the performance difference among two perf.data files
16captured via perf record.
17
18If no parameters are passed it will assume perf.data.old and perf.data.
19
20OPTIONS
21-------
22-p::
23--percentage::
24 Show percentages instead of raw counters
25-v::
26--verbose::
27 Be verbose, for instance, show the raw counters in addition to the
28 diff.
29SEE ALSO
30--------
31linkperf:perf-record[1]
diff --git a/tools/perf/Makefile b/tools/perf/Makefile
index a4cb79255383..87a424e4cdae 100644
--- a/tools/perf/Makefile
+++ b/tools/perf/Makefile
@@ -427,6 +427,7 @@ BUILTIN_OBJS += bench/sched-messaging.o
427BUILTIN_OBJS += bench/sched-pipe.o 427BUILTIN_OBJS += bench/sched-pipe.o
428BUILTIN_OBJS += bench/mem-memcpy.o 428BUILTIN_OBJS += bench/mem-memcpy.o
429 429
430BUILTIN_OBJS += builtin-diff.o
430BUILTIN_OBJS += builtin-help.o 431BUILTIN_OBJS += builtin-help.o
431BUILTIN_OBJS += builtin-sched.o 432BUILTIN_OBJS += builtin-sched.o
432BUILTIN_OBJS += builtin-buildid-list.o 433BUILTIN_OBJS += builtin-buildid-list.o
diff --git a/tools/perf/builtin-diff.c b/tools/perf/builtin-diff.c
new file mode 100644
index 000000000000..0d528018ffb8
--- /dev/null
+++ b/tools/perf/builtin-diff.c
@@ -0,0 +1,288 @@
1/*
2 * builtin-diff.c
3 *
4 * Builtin diff command: Analyze two perf.data input files, look up and read
5 * DSOs and symbol information, sort them and produce a diff.
6 */
7#include "builtin.h"
8
9#include "util/debug.h"
10#include "util/event.h"
11#include "util/hist.h"
12#include "util/session.h"
13#include "util/sort.h"
14#include "util/symbol.h"
15#include "util/util.h"
16
17#include <stdlib.h>
18
19static char const *input_old = "perf.data.old",
20 *input_new = "perf.data";
21static int force;
22static bool show_percent;
23
24struct symbol_conf symbol_conf;
25
26static int perf_session__add_hist_entry(struct perf_session *self,
27 struct addr_location *al, u64 count)
28{
29 bool hit;
30 struct hist_entry *he = __perf_session__add_hist_entry(self, al, NULL,
31 count, &hit);
32 if (he == NULL)
33 return -ENOMEM;
34
35 if (hit)
36 he->count += count;
37
38 return 0;
39}
40
41static int diff__process_sample_event(event_t *event, struct perf_session *session)
42{
43 struct addr_location al;
44 struct sample_data data = { .period = 1, };
45
46 dump_printf("(IP, %d): %d: %p\n", event->header.misc,
47 event->ip.pid, (void *)(long)event->ip.ip);
48
49 if (event__preprocess_sample(event, session, &al, NULL) < 0) {
50 pr_warning("problem processing %d event, skipping it.\n",
51 event->header.type);
52 return -1;
53 }
54
55 event__parse_sample(event, session->sample_type, &data);
56
57 if (al.sym && perf_session__add_hist_entry(session, &al, data.period)) {
58 pr_warning("problem incrementing symbol count, skipping event\n");
59 return -1;
60 }
61
62 session->events_stats.total += data.period;
63 return 0;
64}
65
66static struct perf_event_ops event_ops = {
67 .process_sample_event = diff__process_sample_event,
68 .process_mmap_event = event__process_mmap,
69 .process_comm_event = event__process_comm,
70 .process_exit_event = event__process_task,
71 .process_fork_event = event__process_task,
72 .process_lost_event = event__process_lost,
73};
74
75static void perf_session__insert_hist_entry_by_name(struct rb_root *root,
76 struct hist_entry *he)
77{
78 struct rb_node **p = &root->rb_node;
79 struct rb_node *parent = NULL;
80 struct hist_entry *iter;
81
82 while (*p != NULL) {
83 int cmp;
84 parent = *p;
85 iter = rb_entry(parent, struct hist_entry, rb_node);
86
87 cmp = strcmp(he->map->dso->name, iter->map->dso->name);
88 if (cmp > 0)
89 p = &(*p)->rb_left;
90 else if (cmp < 0)
91 p = &(*p)->rb_right;
92 else {
93 cmp = strcmp(he->sym->name, iter->sym->name);
94 if (cmp > 0)
95 p = &(*p)->rb_left;
96 else
97 p = &(*p)->rb_right;
98 }
99 }
100
101 rb_link_node(&he->rb_node, parent, p);
102 rb_insert_color(&he->rb_node, root);
103}
104
105static void perf_session__resort_by_name(struct perf_session *self)
106{
107 unsigned long position = 1;
108 struct rb_root tmp = RB_ROOT;
109 struct rb_node *next = rb_first(&self->hists);
110
111 while (next != NULL) {
112 struct hist_entry *n = rb_entry(next, struct hist_entry, rb_node);
113
114 next = rb_next(&n->rb_node);
115 rb_erase(&n->rb_node, &self->hists);
116 n->position = position++;
117 perf_session__insert_hist_entry_by_name(&tmp, n);
118 }
119
120 self->hists = tmp;
121}
122
123static struct hist_entry *
124perf_session__find_hist_entry_by_name(struct perf_session *self,
125 struct hist_entry *he)
126{
127 struct rb_node *n = self->hists.rb_node;
128
129 while (n) {
130 struct hist_entry *iter = rb_entry(n, struct hist_entry, rb_node);
131 int cmp = strcmp(he->map->dso->name, iter->map->dso->name);
132
133 if (cmp > 0)
134 n = n->rb_left;
135 else if (cmp < 0)
136 n = n->rb_right;
137 else {
138 cmp = strcmp(he->sym->name, iter->sym->name);
139 if (cmp > 0)
140 n = n->rb_left;
141 else if (cmp < 0)
142 n = n->rb_right;
143 else
144 return iter;
145 }
146 }
147
148 return NULL;
149}
150
151static void perf_session__match_hists(struct perf_session *old_session,
152 struct perf_session *new_session)
153{
154 struct rb_node *nd;
155
156 perf_session__resort_by_name(old_session);
157
158 for (nd = rb_first(&new_session->hists); nd; nd = rb_next(nd)) {
159 struct hist_entry *pos = rb_entry(nd, struct hist_entry, rb_node);
160 pos->pair = perf_session__find_hist_entry_by_name(old_session, pos);
161 }
162}
163
164static size_t hist_entry__fprintf_matched(struct hist_entry *self,
165 unsigned long pos,
166 struct perf_session *session,
167 struct perf_session *pair_session,
168 FILE *fp)
169{
170 u64 old_count = 0;
171 char displacement[16];
172 size_t printed;
173
174 if (self->pair != NULL) {
175 long pdiff = (long)self->pair->position - (long)pos;
176 old_count = self->pair->count;
177 if (pdiff == 0)
178 goto blank;
179 snprintf(displacement, sizeof(displacement), "%+4ld", pdiff);
180 } else {
181blank: memset(displacement, ' ', sizeof(displacement));
182 }
183
184 printed = fprintf(fp, "%4lu %5.5s ", pos, displacement);
185
186 if (show_percent) {
187 double old_percent = (old_count * 100) / pair_session->events_stats.total,
188 new_percent = (self->count * 100) / session->events_stats.total;
189 double diff = old_percent - new_percent;
190
191 if (verbose)
192 printed += fprintf(fp, " %3.2f%% %3.2f%%", old_percent, new_percent);
193
194 if ((u64)diff != 0)
195 printed += fprintf(fp, " %+4.2F%%", diff);
196 else
197 printed += fprintf(fp, " ");
198 } else {
199 if (verbose)
200 printed += fprintf(fp, " %9Lu %9Lu", old_count, self->count);
201 printed += fprintf(fp, " %+9Ld", (s64)self->count - (s64)old_count);
202 }
203
204 return printed + fprintf(fp, " %25.25s %s\n",
205 self->map->dso->name, self->sym->name);
206}
207
208static size_t perf_session__fprintf_matched_hists(struct perf_session *self,
209 struct perf_session *pair,
210 FILE *fp)
211{
212 struct rb_node *nd;
213 size_t printed = 0;
214 unsigned long pos = 1;
215
216 for (nd = rb_first(&self->hists); nd; nd = rb_next(nd)) {
217 struct hist_entry *he = rb_entry(nd, struct hist_entry, rb_node);
218 printed += hist_entry__fprintf_matched(he, pos++, self, pair, fp);
219 }
220
221 return printed;
222}
223
224static int __cmd_diff(void)
225{
226 int ret, i;
227 struct perf_session *session[2];
228
229 session[0] = perf_session__new(input_old, O_RDONLY, force, &symbol_conf);
230 session[1] = perf_session__new(input_new, O_RDONLY, force, &symbol_conf);
231 if (session[0] == NULL || session[1] == NULL)
232 return -ENOMEM;
233
234 for (i = 0; i < 2; ++i) {
235 ret = perf_session__process_events(session[i], &event_ops);
236 if (ret)
237 goto out_delete;
238 perf_session__output_resort(session[i], session[i]->events_stats.total);
239 }
240
241 perf_session__match_hists(session[0], session[1]);
242 perf_session__fprintf_matched_hists(session[1], session[0], stdout);
243out_delete:
244 for (i = 0; i < 2; ++i)
245 perf_session__delete(session[i]);
246 return ret;
247}
248
249static const char *const diff_usage[] = {
250 "perf diff [<options>] [old_file] [new_file]",
251};
252
253static const struct option options[] = {
254 OPT_BOOLEAN('v', "verbose", &verbose,
255 "be more verbose (show symbol address, etc)"),
256 OPT_BOOLEAN('D', "dump-raw-trace", &dump_trace,
257 "dump raw trace in ASCII"),
258 OPT_BOOLEAN('f', "force", &force, "don't complain, do it"),
259 OPT_BOOLEAN('m', "modules", &symbol_conf.use_modules,
260 "load module symbols - WARNING: use only with -k and LIVE kernel"),
261 OPT_BOOLEAN('p', "percentages", &show_percent,
262 "Don't shorten the pathnames taking into account the cwd"),
263 OPT_BOOLEAN('P', "full-paths", &event_ops.full_paths,
264 "Don't shorten the pathnames taking into account the cwd"),
265 OPT_END()
266};
267
268int cmd_diff(int argc, const char **argv, const char *prefix __used)
269{
270 if (symbol__init(&symbol_conf) < 0)
271 return -1;
272
273 setup_sorting(diff_usage, options);
274
275 argc = parse_options(argc, argv, options, diff_usage, 0);
276 if (argc) {
277 if (argc > 2)
278 usage_with_options(diff_usage, options);
279 if (argc == 2) {
280 input_old = argv[0];
281 input_new = argv[1];
282 } else
283 input_new = argv[0];
284 }
285
286 setup_pager();
287 return __cmd_diff();
288}
diff --git a/tools/perf/builtin.h b/tools/perf/builtin.h
index a3d8bf65f26c..18035b1f16c7 100644
--- a/tools/perf/builtin.h
+++ b/tools/perf/builtin.h
@@ -17,6 +17,7 @@ extern int check_pager_config(const char *cmd);
17extern int cmd_annotate(int argc, const char **argv, const char *prefix); 17extern int cmd_annotate(int argc, const char **argv, const char *prefix);
18extern int cmd_bench(int argc, const char **argv, const char *prefix); 18extern int cmd_bench(int argc, const char **argv, const char *prefix);
19extern int cmd_buildid_list(int argc, const char **argv, const char *prefix); 19extern int cmd_buildid_list(int argc, const char **argv, const char *prefix);
20extern int cmd_diff(int argc, const char **argv, const char *prefix);
20extern int cmd_help(int argc, const char **argv, const char *prefix); 21extern int cmd_help(int argc, const char **argv, const char *prefix);
21extern int cmd_sched(int argc, const char **argv, const char *prefix); 22extern int cmd_sched(int argc, const char **argv, const char *prefix);
22extern int cmd_list(int argc, const char **argv, const char *prefix); 23extern int cmd_list(int argc, const char **argv, const char *prefix);
diff --git a/tools/perf/command-list.txt b/tools/perf/command-list.txt
index 02b09ea17a3e..71dc7c3fe7b2 100644
--- a/tools/perf/command-list.txt
+++ b/tools/perf/command-list.txt
@@ -5,6 +5,7 @@
5perf-annotate mainporcelain common 5perf-annotate mainporcelain common
6perf-bench mainporcelain common 6perf-bench mainporcelain common
7perf-buildid-list mainporcelain common 7perf-buildid-list mainporcelain common
8perf-diff mainporcelain common
8perf-list mainporcelain common 9perf-list mainporcelain common
9perf-sched mainporcelain common 10perf-sched mainporcelain common
10perf-record mainporcelain common 11perf-record mainporcelain common
diff --git a/tools/perf/perf.c b/tools/perf/perf.c
index cf64049bc9bd..873e55fab375 100644
--- a/tools/perf/perf.c
+++ b/tools/perf/perf.c
@@ -286,6 +286,7 @@ static void handle_internal_command(int argc, const char **argv)
286 const char *cmd = argv[0]; 286 const char *cmd = argv[0];
287 static struct cmd_struct commands[] = { 287 static struct cmd_struct commands[] = {
288 { "buildid-list", cmd_buildid_list, 0 }, 288 { "buildid-list", cmd_buildid_list, 0 },
289 { "diff", cmd_diff, 0 },
289 { "help", cmd_help, 0 }, 290 { "help", cmd_help, 0 },
290 { "list", cmd_list, 0 }, 291 { "list", cmd_list, 0 },
291 { "record", cmd_record, 0 }, 292 { "record", cmd_record, 0 },
diff --git a/tools/perf/util/sort.h b/tools/perf/util/sort.h
index cb6151c026f2..925f083e1eee 100644
--- a/tools/perf/util/sort.h
+++ b/tools/perf/util/sort.h
@@ -49,9 +49,13 @@ struct hist_entry {
49 struct symbol *sym; 49 struct symbol *sym;
50 u64 ip; 50 u64 ip;
51 char level; 51 char level;
52 struct symbol *parent; 52 struct symbol *parent;
53 struct callchain_node callchain; 53 struct callchain_node callchain;
54 struct rb_root sorted_chain; 54 union {
55 unsigned long position;
56 struct hist_entry *pair;
57 struct rb_root sorted_chain;
58 };
55}; 59};
56 60
57enum sort_type { 61enum sort_type {