aha/tools/perf/util/thread.c

300 lines
6.3 KiB
C
Raw Normal View History

#include "../perf.h"
#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include "thread.h"
#include "util.h"
#include "debug.h"
static struct rb_root threads;
static struct thread *last_match;
void thread__init(struct thread *self, pid_t pid)
{
int i;
self->pid = pid;
self->comm = NULL;
for (i = 0; i < MAP__NR_TYPES; ++i) {
self->maps[i] = RB_ROOT;
INIT_LIST_HEAD(&self->removed_maps[i]);
}
}
static struct thread *thread__new(pid_t pid)
{
struct thread *self = zalloc(sizeof(*self));
if (self != NULL) {
thread__init(self, pid);
self->comm = malloc(32);
if (self->comm)
snprintf(self->comm, 32, ":%d", self->pid);
}
return self;
}
int thread__set_comm(struct thread *self, const char *comm)
{
if (self->comm)
free(self->comm);
self->comm = strdup(comm);
return self->comm ? 0 : -ENOMEM;
}
perf tools: Bind callchains to the first sort dimension column Currently, the callchains are displayed using a constant left margin. So depending on the current sort dimension configuration, callchains may appear to be well attached to the first sort dimension column field which is mostly the case, except when the first dimension of sorting is done by comm, because these are right aligned. This patch binds the callchain to the first letter in the first column, whatever type of column it is (dso, comm, symbol). Before: 0.80% perf [k] __lock_acquire __lock_acquire lock_acquire | |--58.33%-- _spin_lock | | | |--28.57%-- inotify_should_send_event | | fsnotify | | __fsnotify_parent After: 0.80% perf [k] __lock_acquire __lock_acquire lock_acquire | |--58.33%-- _spin_lock | | | |--28.57%-- inotify_should_send_event | | fsnotify | | __fsnotify_parent Also, for clarity, we don't put anymore the callchain as is but: - If we have a top level ancestor in the callchain, start it with a first ascii hook. Before: 0.80% perf [kernel] [k] __lock_acquire __lock_acquire lock_acquire | |--58.33%-- _spin_lock | | | |--28.57%-- inotify_should_send_event | | fsnotify [..] [..] After: 0.80% perf [kernel] [k] __lock_acquire | --- __lock_acquire lock_acquire | |--58.33%-- _spin_lock | | | |--28.57%-- inotify_should_send_event | | fsnotify [..] [..] - Otherwise, if we have several top level ancestors, then display these like we did before: 1.69% Xorg | |--21.21%-- vread_hpet | 0x7fffd85b46fc | 0x7fffd85b494d | 0x7f4fafb4e54d | |--15.15%-- exaOffscreenAlloc | |--9.09%-- I830WaitLpRing Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Anton Blanchard <anton@samba.org> LKML-Reference: <1256246604-17156-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-22 21:23:23 +00:00
int thread__comm_len(struct thread *self)
{
if (!self->comm_len) {
if (!self->comm)
return 0;
self->comm_len = strlen(self->comm);
}
return self->comm_len;
}
static const char *map_type__name[MAP__NR_TYPES] = {
[MAP__FUNCTION] = "Functions",
};
static size_t __thread__fprintf_maps(struct thread *self,
enum map_type type, FILE *fp)
{
size_t printed = fprintf(fp, "%s:\n", map_type__name[type]);
perf tools: Use rb_tree for maps Threads can have many and kernel modules will be represented as a tree of maps as well. Ah, and for a perf.data with 146607 samples: Before: [root@doppio ~]# perf stat -r 5 perf report > /dev/null Performance counter stats for 'perf report' (5 runs): 699.823680 task-clock-msecs # 0.991 CPUs ( +- 0.454% ) 74 context-switches # 0.000 M/sec ( +- 1.709% ) 2 CPU-migrations # 0.000 M/sec ( +- 17.008% ) 23114 page-faults # 0.033 M/sec ( +- 0.000% ) 1381257019 cycles # 1973.721 M/sec ( +- 0.290% ) 1456894438 instructions # 1.055 IPC ( +- 0.007% ) 18779818 cache-references # 26.835 M/sec ( +- 0.380% ) 641799 cache-misses # 0.917 M/sec ( +- 1.200% ) 0.705972729 seconds time elapsed ( +- 0.501% ) [root@doppio ~]# After Performance counter stats for 'perf report' (5 runs): 691.261451 task-clock-msecs # 0.993 CPUs ( +- 0.307% ) 72 context-switches # 0.000 M/sec ( +- 0.829% ) 6 CPU-migrations # 0.000 M/sec ( +- 18.409% ) 23127 page-faults # 0.033 M/sec ( +- 0.000% ) 1366395876 cycles # 1976.670 M/sec ( +- 0.153% ) 1443136016 instructions # 1.056 IPC ( +- 0.012% ) 17956402 cache-references # 25.976 M/sec ( +- 0.325% ) 661924 cache-misses # 0.958 M/sec ( +- 1.335% ) 0.696127275 seconds time elapsed ( +- 0.377% ) I.e. we see some speedup too. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: "H. Peter Anvin" <hpa@zytor.com> LKML-Reference: <20090928174846.GA3361@ghostprotocols.net> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-09-28 17:48:46 +00:00
struct rb_node *nd;
for (nd = rb_first(&self->maps[type]); nd; nd = rb_next(nd)) {
struct map *pos = rb_entry(nd, struct map, rb_node);
printed += fprintf(fp, "Map:");
printed += map__fprintf(pos, fp);
if (verbose > 1) {
printed += dso__fprintf(pos->dso, type, fp);
printed += fprintf(fp, "--\n");
}
perf tools: Use rb_tree for maps Threads can have many and kernel modules will be represented as a tree of maps as well. Ah, and for a perf.data with 146607 samples: Before: [root@doppio ~]# perf stat -r 5 perf report > /dev/null Performance counter stats for 'perf report' (5 runs): 699.823680 task-clock-msecs # 0.991 CPUs ( +- 0.454% ) 74 context-switches # 0.000 M/sec ( +- 1.709% ) 2 CPU-migrations # 0.000 M/sec ( +- 17.008% ) 23114 page-faults # 0.033 M/sec ( +- 0.000% ) 1381257019 cycles # 1973.721 M/sec ( +- 0.290% ) 1456894438 instructions # 1.055 IPC ( +- 0.007% ) 18779818 cache-references # 26.835 M/sec ( +- 0.380% ) 641799 cache-misses # 0.917 M/sec ( +- 1.200% ) 0.705972729 seconds time elapsed ( +- 0.501% ) [root@doppio ~]# After Performance counter stats for 'perf report' (5 runs): 691.261451 task-clock-msecs # 0.993 CPUs ( +- 0.307% ) 72 context-switches # 0.000 M/sec ( +- 0.829% ) 6 CPU-migrations # 0.000 M/sec ( +- 18.409% ) 23127 page-faults # 0.033 M/sec ( +- 0.000% ) 1366395876 cycles # 1976.670 M/sec ( +- 0.153% ) 1443136016 instructions # 1.056 IPC ( +- 0.012% ) 17956402 cache-references # 25.976 M/sec ( +- 0.325% ) 661924 cache-misses # 0.958 M/sec ( +- 1.335% ) 0.696127275 seconds time elapsed ( +- 0.377% ) I.e. we see some speedup too. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: "H. Peter Anvin" <hpa@zytor.com> LKML-Reference: <20090928174846.GA3361@ghostprotocols.net> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-09-28 17:48:46 +00:00
}
return printed;
}
size_t thread__fprintf_maps(struct thread *self, FILE *fp)
{
size_t printed = 0, i;
for (i = 0; i < MAP__NR_TYPES; ++i)
printed += __thread__fprintf_maps(self, i, fp);
return printed;
}
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 06:29:58 +00:00
static size_t __thread__fprintf_removed_maps(struct thread *self,
enum map_type type, FILE *fp)
{
struct map *pos;
size_t printed = 0;
list_for_each_entry(pos, &self->removed_maps[type], node) {
printed += fprintf(fp, "Map:");
printed += map__fprintf(pos, fp);
if (verbose > 1) {
printed += dso__fprintf(pos->dso, type, fp);
printed += fprintf(fp, "--\n");
}
}
return printed;
}
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 06:29:58 +00:00
static size_t thread__fprintf_removed_maps(struct thread *self, FILE *fp)
{
size_t printed = 0, i;
for (i = 0; i < MAP__NR_TYPES; ++i)
printed += __thread__fprintf_removed_maps(self, i, fp);
return printed;
}
static size_t thread__fprintf(struct thread *self, FILE *fp)
{
size_t printed = fprintf(fp, "Thread %d %s\n", self->pid, self->comm);
printed += thread__fprintf_removed_maps(self, fp);
printed += fprintf(fp, "Removed maps:\n");
return printed + thread__fprintf_removed_maps(self, fp);
}
struct thread *threads__findnew(pid_t pid)
{
struct rb_node **p = &threads.rb_node;
struct rb_node *parent = NULL;
struct thread *th;
/*
* Font-end cache - PID lookups come in blocks,
* so most of the time we dont have to look up
* the full rbtree:
*/
if (last_match && last_match->pid == pid)
return last_match;
while (*p != NULL) {
parent = *p;
th = rb_entry(parent, struct thread, rb_node);
if (th->pid == pid) {
last_match = th;
return th;
}
if (pid < th->pid)
p = &(*p)->rb_left;
else
p = &(*p)->rb_right;
}
th = thread__new(pid);
if (th != NULL) {
rb_link_node(&th->rb_node, parent, p);
rb_insert_color(&th->rb_node, &threads);
last_match = th;
}
return th;
}
struct thread *register_idle_thread(void)
{
struct thread *thread = threads__findnew(0);
if (!thread || thread__set_comm(thread, "swapper")) {
fprintf(stderr, "problem inserting idle task.\n");
exit(-1);
}
return thread;
}
perf tools: Use rb_tree for maps Threads can have many and kernel modules will be represented as a tree of maps as well. Ah, and for a perf.data with 146607 samples: Before: [root@doppio ~]# perf stat -r 5 perf report > /dev/null Performance counter stats for 'perf report' (5 runs): 699.823680 task-clock-msecs # 0.991 CPUs ( +- 0.454% ) 74 context-switches # 0.000 M/sec ( +- 1.709% ) 2 CPU-migrations # 0.000 M/sec ( +- 17.008% ) 23114 page-faults # 0.033 M/sec ( +- 0.000% ) 1381257019 cycles # 1973.721 M/sec ( +- 0.290% ) 1456894438 instructions # 1.055 IPC ( +- 0.007% ) 18779818 cache-references # 26.835 M/sec ( +- 0.380% ) 641799 cache-misses # 0.917 M/sec ( +- 1.200% ) 0.705972729 seconds time elapsed ( +- 0.501% ) [root@doppio ~]# After Performance counter stats for 'perf report' (5 runs): 691.261451 task-clock-msecs # 0.993 CPUs ( +- 0.307% ) 72 context-switches # 0.000 M/sec ( +- 0.829% ) 6 CPU-migrations # 0.000 M/sec ( +- 18.409% ) 23127 page-faults # 0.033 M/sec ( +- 0.000% ) 1366395876 cycles # 1976.670 M/sec ( +- 0.153% ) 1443136016 instructions # 1.056 IPC ( +- 0.012% ) 17956402 cache-references # 25.976 M/sec ( +- 0.325% ) 661924 cache-misses # 0.958 M/sec ( +- 1.335% ) 0.696127275 seconds time elapsed ( +- 0.377% ) I.e. we see some speedup too. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: "H. Peter Anvin" <hpa@zytor.com> LKML-Reference: <20090928174846.GA3361@ghostprotocols.net> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-09-28 17:48:46 +00:00
static void thread__remove_overlappings(struct thread *self, struct map *map)
{
struct rb_root *root = &self->maps[map->type];
struct rb_node *next = rb_first(root);
perf tools: Use rb_tree for maps Threads can have many and kernel modules will be represented as a tree of maps as well. Ah, and for a perf.data with 146607 samples: Before: [root@doppio ~]# perf stat -r 5 perf report > /dev/null Performance counter stats for 'perf report' (5 runs): 699.823680 task-clock-msecs # 0.991 CPUs ( +- 0.454% ) 74 context-switches # 0.000 M/sec ( +- 1.709% ) 2 CPU-migrations # 0.000 M/sec ( +- 17.008% ) 23114 page-faults # 0.033 M/sec ( +- 0.000% ) 1381257019 cycles # 1973.721 M/sec ( +- 0.290% ) 1456894438 instructions # 1.055 IPC ( +- 0.007% ) 18779818 cache-references # 26.835 M/sec ( +- 0.380% ) 641799 cache-misses # 0.917 M/sec ( +- 1.200% ) 0.705972729 seconds time elapsed ( +- 0.501% ) [root@doppio ~]# After Performance counter stats for 'perf report' (5 runs): 691.261451 task-clock-msecs # 0.993 CPUs ( +- 0.307% ) 72 context-switches # 0.000 M/sec ( +- 0.829% ) 6 CPU-migrations # 0.000 M/sec ( +- 18.409% ) 23127 page-faults # 0.033 M/sec ( +- 0.000% ) 1366395876 cycles # 1976.670 M/sec ( +- 0.153% ) 1443136016 instructions # 1.056 IPC ( +- 0.012% ) 17956402 cache-references # 25.976 M/sec ( +- 0.325% ) 661924 cache-misses # 0.958 M/sec ( +- 1.335% ) 0.696127275 seconds time elapsed ( +- 0.377% ) I.e. we see some speedup too. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: "H. Peter Anvin" <hpa@zytor.com> LKML-Reference: <20090928174846.GA3361@ghostprotocols.net> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-09-28 17:48:46 +00:00
while (next) {
struct map *pos = rb_entry(next, struct map, rb_node);
next = rb_next(&pos->rb_node);
if (!map__overlap(pos, map))
continue;
if (verbose >= 2) {
fputs("overlapping maps:\n", stderr);
map__fprintf(map, stderr);
map__fprintf(pos, stderr);
perf tools: Use rb_tree for maps Threads can have many and kernel modules will be represented as a tree of maps as well. Ah, and for a perf.data with 146607 samples: Before: [root@doppio ~]# perf stat -r 5 perf report > /dev/null Performance counter stats for 'perf report' (5 runs): 699.823680 task-clock-msecs # 0.991 CPUs ( +- 0.454% ) 74 context-switches # 0.000 M/sec ( +- 1.709% ) 2 CPU-migrations # 0.000 M/sec ( +- 17.008% ) 23114 page-faults # 0.033 M/sec ( +- 0.000% ) 1381257019 cycles # 1973.721 M/sec ( +- 0.290% ) 1456894438 instructions # 1.055 IPC ( +- 0.007% ) 18779818 cache-references # 26.835 M/sec ( +- 0.380% ) 641799 cache-misses # 0.917 M/sec ( +- 1.200% ) 0.705972729 seconds time elapsed ( +- 0.501% ) [root@doppio ~]# After Performance counter stats for 'perf report' (5 runs): 691.261451 task-clock-msecs # 0.993 CPUs ( +- 0.307% ) 72 context-switches # 0.000 M/sec ( +- 0.829% ) 6 CPU-migrations # 0.000 M/sec ( +- 18.409% ) 23127 page-faults # 0.033 M/sec ( +- 0.000% ) 1366395876 cycles # 1976.670 M/sec ( +- 0.153% ) 1443136016 instructions # 1.056 IPC ( +- 0.012% ) 17956402 cache-references # 25.976 M/sec ( +- 0.325% ) 661924 cache-misses # 0.958 M/sec ( +- 1.335% ) 0.696127275 seconds time elapsed ( +- 0.377% ) I.e. we see some speedup too. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: "H. Peter Anvin" <hpa@zytor.com> LKML-Reference: <20090928174846.GA3361@ghostprotocols.net> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-09-28 17:48:46 +00:00
}
rb_erase(&pos->rb_node, root);
perf tools: Rewrite and improve support for kernel modules Representing modules as struct map entries, backed by a DSO, etc, using /proc/modules to find where the module is loaded. DSOs now can have a short and long name, so that in verbose mode we can show exactly which .ko or vmlinux image was used. As kernel modules now are a DSO separate from the kernel, we can ask for just the hits for a particular set of kernel modules, just like we can do with shared libraries: [root@doppio linux-2.6-tip]# perf report -n --vmlinux /home/acme/git/build/tip-recvmmsg/vmlinux --modules --dsos \[drm\] | head -15 84.58% 13266 Xorg [k] drm_clflush_pages 4.02% 630 Xorg [k] trace_kmalloc.clone.0 3.95% 619 Xorg [k] drm_ioctl 2.07% 324 Xorg [k] drm_addbufs 1.68% 263 Xorg [k] drm_gem_close_ioctl 0.77% 120 Xorg [k] drm_setmaster_ioctl 0.70% 110 Xorg [k] drm_lastclose 0.68% 106 Xorg [k] drm_open 0.54% 85 Xorg [k] drm_mm_search_free [root@doppio linux-2.6-tip]# Specifying --dsos /lib/modules/2.6.31-tip/kernel/drivers/gpu/drm/drm.ko would have the same effect. Allowing specifying just 'drm.ko' is left for another patch. Processing kallsyms so that per kernel module struct map are instantiated was also left for another patch. That will allow removing the module name from each of its symbols. struct symbol was reduced by removing the ->module backpointer and moving it (well now the map) to struct symbol_entry in perf top, that is its only user right now. The total linecount went down by ~500 lines. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Avi Kivity <avi@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-10-02 06:29:58 +00:00
/*
* We may have references to this map, for instance in some
* hist_entry instances, so just move them to a separate
* list.
*/
list_add_tail(&pos->node, &self->removed_maps[map->type]);
}
perf tools: Use rb_tree for maps Threads can have many and kernel modules will be represented as a tree of maps as well. Ah, and for a perf.data with 146607 samples: Before: [root@doppio ~]# perf stat -r 5 perf report > /dev/null Performance counter stats for 'perf report' (5 runs): 699.823680 task-clock-msecs # 0.991 CPUs ( +- 0.454% ) 74 context-switches # 0.000 M/sec ( +- 1.709% ) 2 CPU-migrations # 0.000 M/sec ( +- 17.008% ) 23114 page-faults # 0.033 M/sec ( +- 0.000% ) 1381257019 cycles # 1973.721 M/sec ( +- 0.290% ) 1456894438 instructions # 1.055 IPC ( +- 0.007% ) 18779818 cache-references # 26.835 M/sec ( +- 0.380% ) 641799 cache-misses # 0.917 M/sec ( +- 1.200% ) 0.705972729 seconds time elapsed ( +- 0.501% ) [root@doppio ~]# After Performance counter stats for 'perf report' (5 runs): 691.261451 task-clock-msecs # 0.993 CPUs ( +- 0.307% ) 72 context-switches # 0.000 M/sec ( +- 0.829% ) 6 CPU-migrations # 0.000 M/sec ( +- 18.409% ) 23127 page-faults # 0.033 M/sec ( +- 0.000% ) 1366395876 cycles # 1976.670 M/sec ( +- 0.153% ) 1443136016 instructions # 1.056 IPC ( +- 0.012% ) 17956402 cache-references # 25.976 M/sec ( +- 0.325% ) 661924 cache-misses # 0.958 M/sec ( +- 1.335% ) 0.696127275 seconds time elapsed ( +- 0.377% ) I.e. we see some speedup too. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: "H. Peter Anvin" <hpa@zytor.com> LKML-Reference: <20090928174846.GA3361@ghostprotocols.net> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-09-28 17:48:46 +00:00
}
void maps__insert(struct rb_root *maps, struct map *map)
{
struct rb_node **p = &maps->rb_node;
struct rb_node *parent = NULL;
const u64 ip = map->start;
struct map *m;
while (*p != NULL) {
parent = *p;
m = rb_entry(parent, struct map, rb_node);
if (ip < m->start)
p = &(*p)->rb_left;
else
p = &(*p)->rb_right;
}
rb_link_node(&map->rb_node, parent, p);
rb_insert_color(&map->rb_node, maps);
}
struct map *maps__find(struct rb_root *maps, u64 ip)
{
struct rb_node **p = &maps->rb_node;
struct rb_node *parent = NULL;
struct map *m;
while (*p != NULL) {
parent = *p;
m = rb_entry(parent, struct map, rb_node);
if (ip < m->start)
p = &(*p)->rb_left;
else if (ip > m->end)
p = &(*p)->rb_right;
else
return m;
}
return NULL;
}
perf tools: Use rb_tree for maps Threads can have many and kernel modules will be represented as a tree of maps as well. Ah, and for a perf.data with 146607 samples: Before: [root@doppio ~]# perf stat -r 5 perf report > /dev/null Performance counter stats for 'perf report' (5 runs): 699.823680 task-clock-msecs # 0.991 CPUs ( +- 0.454% ) 74 context-switches # 0.000 M/sec ( +- 1.709% ) 2 CPU-migrations # 0.000 M/sec ( +- 17.008% ) 23114 page-faults # 0.033 M/sec ( +- 0.000% ) 1381257019 cycles # 1973.721 M/sec ( +- 0.290% ) 1456894438 instructions # 1.055 IPC ( +- 0.007% ) 18779818 cache-references # 26.835 M/sec ( +- 0.380% ) 641799 cache-misses # 0.917 M/sec ( +- 1.200% ) 0.705972729 seconds time elapsed ( +- 0.501% ) [root@doppio ~]# After Performance counter stats for 'perf report' (5 runs): 691.261451 task-clock-msecs # 0.993 CPUs ( +- 0.307% ) 72 context-switches # 0.000 M/sec ( +- 0.829% ) 6 CPU-migrations # 0.000 M/sec ( +- 18.409% ) 23127 page-faults # 0.033 M/sec ( +- 0.000% ) 1366395876 cycles # 1976.670 M/sec ( +- 0.153% ) 1443136016 instructions # 1.056 IPC ( +- 0.012% ) 17956402 cache-references # 25.976 M/sec ( +- 0.325% ) 661924 cache-misses # 0.958 M/sec ( +- 1.335% ) 0.696127275 seconds time elapsed ( +- 0.377% ) I.e. we see some speedup too. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: "H. Peter Anvin" <hpa@zytor.com> LKML-Reference: <20090928174846.GA3361@ghostprotocols.net> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-09-28 17:48:46 +00:00
void thread__insert_map(struct thread *self, struct map *map)
{
thread__remove_overlappings(self, map);
maps__insert(&self->maps[map->type], map);
}
static int thread__clone_maps(struct thread *self, struct thread *parent,
enum map_type type)
{
perf tools: Use rb_tree for maps Threads can have many and kernel modules will be represented as a tree of maps as well. Ah, and for a perf.data with 146607 samples: Before: [root@doppio ~]# perf stat -r 5 perf report > /dev/null Performance counter stats for 'perf report' (5 runs): 699.823680 task-clock-msecs # 0.991 CPUs ( +- 0.454% ) 74 context-switches # 0.000 M/sec ( +- 1.709% ) 2 CPU-migrations # 0.000 M/sec ( +- 17.008% ) 23114 page-faults # 0.033 M/sec ( +- 0.000% ) 1381257019 cycles # 1973.721 M/sec ( +- 0.290% ) 1456894438 instructions # 1.055 IPC ( +- 0.007% ) 18779818 cache-references # 26.835 M/sec ( +- 0.380% ) 641799 cache-misses # 0.917 M/sec ( +- 1.200% ) 0.705972729 seconds time elapsed ( +- 0.501% ) [root@doppio ~]# After Performance counter stats for 'perf report' (5 runs): 691.261451 task-clock-msecs # 0.993 CPUs ( +- 0.307% ) 72 context-switches # 0.000 M/sec ( +- 0.829% ) 6 CPU-migrations # 0.000 M/sec ( +- 18.409% ) 23127 page-faults # 0.033 M/sec ( +- 0.000% ) 1366395876 cycles # 1976.670 M/sec ( +- 0.153% ) 1443136016 instructions # 1.056 IPC ( +- 0.012% ) 17956402 cache-references # 25.976 M/sec ( +- 0.325% ) 661924 cache-misses # 0.958 M/sec ( +- 1.335% ) 0.696127275 seconds time elapsed ( +- 0.377% ) I.e. we see some speedup too. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <efault@gmx.de> Cc: "H. Peter Anvin" <hpa@zytor.com> LKML-Reference: <20090928174846.GA3361@ghostprotocols.net> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-09-28 17:48:46 +00:00
struct rb_node *nd;
for (nd = rb_first(&parent->maps[type]); nd; nd = rb_next(nd)) {
struct map *map = rb_entry(nd, struct map, rb_node);
struct map *new = map__clone(map);
if (new == NULL)
return -ENOMEM;
thread__insert_map(self, new);
}
return 0;
}
int thread__fork(struct thread *self, struct thread *parent)
{
int i;
if (self->comm)
free(self->comm);
self->comm = strdup(parent->comm);
if (!self->comm)
return -ENOMEM;
for (i = 0; i < MAP__NR_TYPES; ++i)
if (thread__clone_maps(self, parent, i) < 0)
return -ENOMEM;
return 0;
}
size_t threads__fprintf(FILE *fp)
{
size_t ret = 0;
struct rb_node *nd;
for (nd = rb_first(&threads); nd; nd = rb_next(nd)) {
struct thread *pos = rb_entry(nd, struct thread, rb_node);
ret += thread__fprintf(pos, fp);
}
return ret;
}
perf tools: Consolidate symbol resolving across all tools Now we have a very high level routine for simple tools to process IP sample events: int event__preprocess_sample(const event_t *self, struct addr_location *al, symbol_filter_t filter) It receives the event itself and will insert new threads in the global threads list and resolve the map and symbol, filling all this info into the new addr_location struct, so that tools like annotate and report can further process the event by creating hist_entries in their specific way (with or without callgraphs, etc). It in turn uses the new next layer function: void thread__find_addr_location(struct thread *self, u8 cpumode, enum map_type type, u64 addr, struct addr_location *al, symbol_filter_t filter) This one will, given a thread (userspace or the kernel kthread one), will find the given type (MAP__FUNCTION now, MAP__VARIABLE too in the near future) at the given cpumode, taking vdsos into account (userspace hit, but kernel symbol) and will fill all these details in the addr_location given. Tools that need a more compact API for plain function resolution, like 'kmem', can use this other one: struct symbol *thread__find_function(struct thread *self, u64 addr, symbol_filter_t filter) So, to resolve a kernel symbol, that is all the 'kmem' tool needs, its just a matter of calling: sym = thread__find_function(kthread, addr, NULL); The 'filter' parameter is needed because we do lazy parsing/loading of ELF symtabs or /proc/kallsyms. With this we remove more code duplication all around, which is always good, huh? :-) Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: John Kacur <jkacur@redhat.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <1259346563-12568-12-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-27 18:29:23 +00:00
struct symbol *thread__find_symbol(struct thread *self,
enum map_type type, u64 addr,
symbol_filter_t filter)
{
struct map *map = thread__find_map(self, type, addr);
if (map != NULL)
return map__find_symbol(map, map->map_ip(map, addr), filter);
return NULL;
}