linux/tools/perf/util/hist.c

1219 lines
28 KiB
C
Raw Normal View History

#include "annotate.h"
#include "util.h"
#include "build-id.h"
#include "hist.h"
#include "session.h"
#include "sort.h"
perf diff: Percent calcs should use double values Otherwise we do integer math and the delta values round up to multiples of 1.0%. Also, calculate absolute values. Things look precise now: $ perf report -i perf.data.old --sort dso,symbol | head -13 9.02% libc-2.10.1.so [.] _IO_vfprintf_internal 4.88% find [.] 0x00000000014af0 2.91% [kernel] [k] __kmalloc 2.85% [kernel] [k] ext4_htree_store_dirent 2.50% libc-2.10.1.so [.] __GI_memmove 2.44% [kernel] [k] half_md4_transform 2.43% [kernel] [k] _spin_lock 2.33% [kernel] [k] system_call $ perf report -i perf.data --sort dso,symbol | head -13 8.55% libc-2.10.1.so [.] _IO_vfprintf_internal 3.11% [kernel] [k] __kmalloc 3.07% [kernel] [k] ext4_htree_store_dirent 2.66% find [.] 0x00000000016bcf 2.61% [kernel] [k] _atomic_dec_and_lock 2.46% [kernel] [k] half_md4_transform 2.41% libc-2.10.1.so [.] __GI_memmove 2.30% find [.] 0x00000000009219 $ perf diff | head -13 9.02% -0.47% libc-2.10.1.so [.] _IO_vfprintf_internal 2.91% +0.20% [kernel] [k] __kmalloc 2.85% +0.23% [kernel] [k] ext4_htree_store_dirent 1.99% +0.62% [kernel] [k] _atomic_dec_and_lock 2.44% +0.02% [kernel] [k] half_md4_transform 2.50% -0.09% libc-2.10.1.so [.] __GI_memmove 1.88% +0.01% [kernel] [k] __d_lookup 2.43% -0.75% [kernel] [k] _spin_lock 0.97% +0.62% [kernel] [k] path_get 1.99% -0.42% libc-2.10.1.so [.] _int_malloc $ Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.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> LKML-Reference: <1260981109-2621-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-17 00:31:49 +08:00
#include <math.h>
static bool hists__filter_entry_by_dso(struct hists *hists,
struct hist_entry *he);
static bool hists__filter_entry_by_thread(struct hists *hists,
struct hist_entry *he);
enum hist_filter {
HIST_FILTER__DSO,
HIST_FILTER__THREAD,
HIST_FILTER__PARENT,
};
struct callchain_param callchain_param = {
.mode = CHAIN_GRAPH_REL,
.min_percent = 0.5,
.order = ORDER_CALLEE
};
u16 hists__col_len(struct hists *hists, enum hist_column col)
{
return hists->col_len[col];
}
void hists__set_col_len(struct hists *hists, enum hist_column col, u16 len)
{
hists->col_len[col] = len;
}
bool hists__new_col_len(struct hists *hists, enum hist_column col, u16 len)
{
if (len > hists__col_len(hists, col)) {
hists__set_col_len(hists, col, len);
return true;
}
return false;
}
static void hists__reset_col_len(struct hists *hists)
{
enum hist_column col;
for (col = 0; col < HISTC_NR_COLS; ++col)
hists__set_col_len(hists, col, 0);
}
static void hists__calc_col_len(struct hists *hists, struct hist_entry *h)
{
u16 len;
if (h->ms.sym)
hists__new_col_len(hists, HISTC_SYMBOL, h->ms.sym->namelen);
else {
const unsigned int unresolved_col_width = BITS_PER_LONG / 4;
if (hists__col_len(hists, HISTC_DSO) < unresolved_col_width &&
!symbol_conf.col_width_list_str && !symbol_conf.field_sep &&
!symbol_conf.dso_list)
hists__set_col_len(hists, HISTC_DSO,
unresolved_col_width);
}
len = thread__comm_len(h->thread);
if (hists__new_col_len(hists, HISTC_COMM, len))
hists__set_col_len(hists, HISTC_THREAD, len + 6);
if (h->ms.map) {
len = dso__name_len(h->ms.map->dso);
hists__new_col_len(hists, HISTC_DSO, len);
}
}
static void hist_entry__add_cpumode_period(struct hist_entry *he,
unsigned int cpumode, u64 period)
{
switch (cpumode) {
case PERF_RECORD_MISC_KERNEL:
he->period_sys += period;
break;
case PERF_RECORD_MISC_USER:
he->period_us += period;
break;
case PERF_RECORD_MISC_GUEST_KERNEL:
he->period_guest_sys += period;
break;
case PERF_RECORD_MISC_GUEST_USER:
he->period_guest_us += period;
break;
default:
break;
}
}
perf top: Reuse the 'report' hist_entry/hists classes This actually fixes several problems we had in the old 'perf top': 1. Unresolved symbols not show, limitation that came from the old "KernelTop" codebase, to solve it we would need to do changes that would make sym_entry have most of the hist_entry fields. 2. It was using the number of samples, not the sum of sample->period. And brings the --sort code that allows us to have all the views in 'perf report', for instance: [root@emilia ~]# perf top --sort dso PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs) ------------------------------------------------------------------------------ 31.59% libcrypto.so.1.0.0 21.55% [kernel] 18.57% libpython2.6.so.1.0 7.04% libc-2.12.so 6.99% _backend_agg.so 4.72% sshd 1.48% multiarray.so 1.39% libfreetype.so.6.3.22 1.37% perf 0.71% libgobject-2.0.so.0.2200.5 0.53% [tg3] 0.48% libglib-2.0.so.0.2200.5 0.44% libstdc++.so.6.0.13 0.40% libcairo.so.2.10800.8 0.38% libm-2.12.so 0.34% umath.so 0.30% libgdk-x11-2.0.so.0.1800.9 0.22% libpthread-2.12.so 0.20% libgtk-x11-2.0.so.0.1800.9 0.20% librt-2.12.so 0.15% _path.so 0.13% libpango-1.0.so.0.2800.1 0.11% libatlas.so.3.0 0.09% ft2font.so 0.09% libpangoft2-1.0.so.0.2800.1 0.08% libX11.so.6.3.0 0.07% [vdso] 0.06% cyclictest ^C All the filter lists can be used as well: --dsos, --comms, --symbols, etc. The 'perf report' TUI is also reused, being possible to apply all the zoom operations, do annotation, etc. This change will allow multiple simplifications in the symbol system as well, that will be detailed in upcoming changesets. Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Stephane Eranian <eranian@google.com> Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 06:16:15 +08:00
static void hist_entry__decay(struct hist_entry *he)
{
he->period = (he->period * 7) / 8;
he->nr_events = (he->nr_events * 7) / 8;
}
static bool hists__decay_entry(struct hists *hists, struct hist_entry *he)
{
u64 prev_period = he->period;
if (prev_period == 0)
return true;
perf top: Reuse the 'report' hist_entry/hists classes This actually fixes several problems we had in the old 'perf top': 1. Unresolved symbols not show, limitation that came from the old "KernelTop" codebase, to solve it we would need to do changes that would make sym_entry have most of the hist_entry fields. 2. It was using the number of samples, not the sum of sample->period. And brings the --sort code that allows us to have all the views in 'perf report', for instance: [root@emilia ~]# perf top --sort dso PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs) ------------------------------------------------------------------------------ 31.59% libcrypto.so.1.0.0 21.55% [kernel] 18.57% libpython2.6.so.1.0 7.04% libc-2.12.so 6.99% _backend_agg.so 4.72% sshd 1.48% multiarray.so 1.39% libfreetype.so.6.3.22 1.37% perf 0.71% libgobject-2.0.so.0.2200.5 0.53% [tg3] 0.48% libglib-2.0.so.0.2200.5 0.44% libstdc++.so.6.0.13 0.40% libcairo.so.2.10800.8 0.38% libm-2.12.so 0.34% umath.so 0.30% libgdk-x11-2.0.so.0.1800.9 0.22% libpthread-2.12.so 0.20% libgtk-x11-2.0.so.0.1800.9 0.20% librt-2.12.so 0.15% _path.so 0.13% libpango-1.0.so.0.2800.1 0.11% libatlas.so.3.0 0.09% ft2font.so 0.09% libpangoft2-1.0.so.0.2800.1 0.08% libX11.so.6.3.0 0.07% [vdso] 0.06% cyclictest ^C All the filter lists can be used as well: --dsos, --comms, --symbols, etc. The 'perf report' TUI is also reused, being possible to apply all the zoom operations, do annotation, etc. This change will allow multiple simplifications in the symbol system as well, that will be detailed in upcoming changesets. Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Stephane Eranian <eranian@google.com> Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 06:16:15 +08:00
hist_entry__decay(he);
if (!he->filtered)
hists->stats.total_period -= prev_period - he->period;
perf top: Reuse the 'report' hist_entry/hists classes This actually fixes several problems we had in the old 'perf top': 1. Unresolved symbols not show, limitation that came from the old "KernelTop" codebase, to solve it we would need to do changes that would make sym_entry have most of the hist_entry fields. 2. It was using the number of samples, not the sum of sample->period. And brings the --sort code that allows us to have all the views in 'perf report', for instance: [root@emilia ~]# perf top --sort dso PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs) ------------------------------------------------------------------------------ 31.59% libcrypto.so.1.0.0 21.55% [kernel] 18.57% libpython2.6.so.1.0 7.04% libc-2.12.so 6.99% _backend_agg.so 4.72% sshd 1.48% multiarray.so 1.39% libfreetype.so.6.3.22 1.37% perf 0.71% libgobject-2.0.so.0.2200.5 0.53% [tg3] 0.48% libglib-2.0.so.0.2200.5 0.44% libstdc++.so.6.0.13 0.40% libcairo.so.2.10800.8 0.38% libm-2.12.so 0.34% umath.so 0.30% libgdk-x11-2.0.so.0.1800.9 0.22% libpthread-2.12.so 0.20% libgtk-x11-2.0.so.0.1800.9 0.20% librt-2.12.so 0.15% _path.so 0.13% libpango-1.0.so.0.2800.1 0.11% libatlas.so.3.0 0.09% ft2font.so 0.09% libpangoft2-1.0.so.0.2800.1 0.08% libX11.so.6.3.0 0.07% [vdso] 0.06% cyclictest ^C All the filter lists can be used as well: --dsos, --comms, --symbols, etc. The 'perf report' TUI is also reused, being possible to apply all the zoom operations, do annotation, etc. This change will allow multiple simplifications in the symbol system as well, that will be detailed in upcoming changesets. Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Stephane Eranian <eranian@google.com> Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 06:16:15 +08:00
return he->period == 0;
}
static void __hists__decay_entries(struct hists *hists, bool zap_user,
bool zap_kernel, bool threaded)
perf top: Reuse the 'report' hist_entry/hists classes This actually fixes several problems we had in the old 'perf top': 1. Unresolved symbols not show, limitation that came from the old "KernelTop" codebase, to solve it we would need to do changes that would make sym_entry have most of the hist_entry fields. 2. It was using the number of samples, not the sum of sample->period. And brings the --sort code that allows us to have all the views in 'perf report', for instance: [root@emilia ~]# perf top --sort dso PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs) ------------------------------------------------------------------------------ 31.59% libcrypto.so.1.0.0 21.55% [kernel] 18.57% libpython2.6.so.1.0 7.04% libc-2.12.so 6.99% _backend_agg.so 4.72% sshd 1.48% multiarray.so 1.39% libfreetype.so.6.3.22 1.37% perf 0.71% libgobject-2.0.so.0.2200.5 0.53% [tg3] 0.48% libglib-2.0.so.0.2200.5 0.44% libstdc++.so.6.0.13 0.40% libcairo.so.2.10800.8 0.38% libm-2.12.so 0.34% umath.so 0.30% libgdk-x11-2.0.so.0.1800.9 0.22% libpthread-2.12.so 0.20% libgtk-x11-2.0.so.0.1800.9 0.20% librt-2.12.so 0.15% _path.so 0.13% libpango-1.0.so.0.2800.1 0.11% libatlas.so.3.0 0.09% ft2font.so 0.09% libpangoft2-1.0.so.0.2800.1 0.08% libX11.so.6.3.0 0.07% [vdso] 0.06% cyclictest ^C All the filter lists can be used as well: --dsos, --comms, --symbols, etc. The 'perf report' TUI is also reused, being possible to apply all the zoom operations, do annotation, etc. This change will allow multiple simplifications in the symbol system as well, that will be detailed in upcoming changesets. Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Stephane Eranian <eranian@google.com> Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 06:16:15 +08:00
{
struct rb_node *next = rb_first(&hists->entries);
struct hist_entry *n;
while (next) {
n = rb_entry(next, struct hist_entry, rb_node);
next = rb_next(&n->rb_node);
/*
* We may be annotating this, for instance, so keep it here in
* case some it gets new samples, we'll eventually free it when
* the user stops browsing and it agains gets fully decayed.
*/
if (((zap_user && n->level == '.') ||
(zap_kernel && n->level != '.') ||
hists__decay_entry(hists, n)) &&
!n->used) {
perf top: Reuse the 'report' hist_entry/hists classes This actually fixes several problems we had in the old 'perf top': 1. Unresolved symbols not show, limitation that came from the old "KernelTop" codebase, to solve it we would need to do changes that would make sym_entry have most of the hist_entry fields. 2. It was using the number of samples, not the sum of sample->period. And brings the --sort code that allows us to have all the views in 'perf report', for instance: [root@emilia ~]# perf top --sort dso PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs) ------------------------------------------------------------------------------ 31.59% libcrypto.so.1.0.0 21.55% [kernel] 18.57% libpython2.6.so.1.0 7.04% libc-2.12.so 6.99% _backend_agg.so 4.72% sshd 1.48% multiarray.so 1.39% libfreetype.so.6.3.22 1.37% perf 0.71% libgobject-2.0.so.0.2200.5 0.53% [tg3] 0.48% libglib-2.0.so.0.2200.5 0.44% libstdc++.so.6.0.13 0.40% libcairo.so.2.10800.8 0.38% libm-2.12.so 0.34% umath.so 0.30% libgdk-x11-2.0.so.0.1800.9 0.22% libpthread-2.12.so 0.20% libgtk-x11-2.0.so.0.1800.9 0.20% librt-2.12.so 0.15% _path.so 0.13% libpango-1.0.so.0.2800.1 0.11% libatlas.so.3.0 0.09% ft2font.so 0.09% libpangoft2-1.0.so.0.2800.1 0.08% libX11.so.6.3.0 0.07% [vdso] 0.06% cyclictest ^C All the filter lists can be used as well: --dsos, --comms, --symbols, etc. The 'perf report' TUI is also reused, being possible to apply all the zoom operations, do annotation, etc. This change will allow multiple simplifications in the symbol system as well, that will be detailed in upcoming changesets. Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Stephane Eranian <eranian@google.com> Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 06:16:15 +08:00
rb_erase(&n->rb_node, &hists->entries);
if (sort__need_collapse || threaded)
perf top: Reuse the 'report' hist_entry/hists classes This actually fixes several problems we had in the old 'perf top': 1. Unresolved symbols not show, limitation that came from the old "KernelTop" codebase, to solve it we would need to do changes that would make sym_entry have most of the hist_entry fields. 2. It was using the number of samples, not the sum of sample->period. And brings the --sort code that allows us to have all the views in 'perf report', for instance: [root@emilia ~]# perf top --sort dso PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs) ------------------------------------------------------------------------------ 31.59% libcrypto.so.1.0.0 21.55% [kernel] 18.57% libpython2.6.so.1.0 7.04% libc-2.12.so 6.99% _backend_agg.so 4.72% sshd 1.48% multiarray.so 1.39% libfreetype.so.6.3.22 1.37% perf 0.71% libgobject-2.0.so.0.2200.5 0.53% [tg3] 0.48% libglib-2.0.so.0.2200.5 0.44% libstdc++.so.6.0.13 0.40% libcairo.so.2.10800.8 0.38% libm-2.12.so 0.34% umath.so 0.30% libgdk-x11-2.0.so.0.1800.9 0.22% libpthread-2.12.so 0.20% libgtk-x11-2.0.so.0.1800.9 0.20% librt-2.12.so 0.15% _path.so 0.13% libpango-1.0.so.0.2800.1 0.11% libatlas.so.3.0 0.09% ft2font.so 0.09% libpangoft2-1.0.so.0.2800.1 0.08% libX11.so.6.3.0 0.07% [vdso] 0.06% cyclictest ^C All the filter lists can be used as well: --dsos, --comms, --symbols, etc. The 'perf report' TUI is also reused, being possible to apply all the zoom operations, do annotation, etc. This change will allow multiple simplifications in the symbol system as well, that will be detailed in upcoming changesets. Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Stephane Eranian <eranian@google.com> Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 06:16:15 +08:00
rb_erase(&n->rb_node_in, &hists->entries_collapsed);
hist_entry__free(n);
--hists->nr_entries;
}
}
}
void hists__decay_entries(struct hists *hists, bool zap_user, bool zap_kernel)
{
return __hists__decay_entries(hists, zap_user, zap_kernel, false);
}
void hists__decay_entries_threaded(struct hists *hists,
bool zap_user, bool zap_kernel)
{
return __hists__decay_entries(hists, zap_user, zap_kernel, true);
}
/*
* histogram, sorted on item, collects periods
*/
static struct hist_entry *hist_entry__new(struct hist_entry *template)
{
size_t callchain_size = symbol_conf.use_callchain ? sizeof(struct callchain_root) : 0;
struct hist_entry *he = malloc(sizeof(*he) + callchain_size);
if (he != NULL) {
*he = *template;
he->nr_events = 1;
if (he->ms.map)
he->ms.map->referenced = true;
if (symbol_conf.use_callchain)
callchain_init(he->callchain);
}
return he;
}
static void hists__inc_nr_entries(struct hists *hists, struct hist_entry *h)
{
if (!h->filtered) {
hists__calc_col_len(hists, h);
++hists->nr_entries;
hists->stats.total_period += h->period;
}
}
static u8 symbol__parent_filter(const struct symbol *parent)
{
if (symbol_conf.exclude_other && parent == NULL)
return 1 << HIST_FILTER__PARENT;
return 0;
}
struct hist_entry *__hists__add_entry(struct hists *hists,
2010-05-11 00:04:11 +08:00
struct addr_location *al,
struct symbol *sym_parent, u64 period)
{
struct rb_node **p;
struct rb_node *parent = NULL;
struct hist_entry *he;
struct hist_entry entry = {
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>
2009-11-28 02:29:23 +08:00
.thread = al->thread,
.ms = {
.map = al->map,
.sym = al->sym,
},
.cpu = al->cpu,
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>
2009-11-28 02:29:23 +08:00
.ip = al->addr,
.level = al->level,
.period = period,
.parent = sym_parent,
.filtered = symbol__parent_filter(sym_parent),
};
int cmp;
pthread_mutex_lock(&hists->lock);
p = &hists->entries_in->rb_node;
while (*p != NULL) {
parent = *p;
he = rb_entry(parent, struct hist_entry, rb_node_in);
cmp = hist_entry__cmp(&entry, he);
if (!cmp) {
he->period += period;
++he->nr_events;
goto out;
}
if (cmp < 0)
p = &(*p)->rb_left;
else
p = &(*p)->rb_right;
}
he = hist_entry__new(&entry);
if (!he)
goto out_unlock;
rb_link_node(&he->rb_node_in, parent, p);
rb_insert_color(&he->rb_node_in, hists->entries_in);
out:
hist_entry__add_cpumode_period(he, al->cpumode, period);
out_unlock:
pthread_mutex_unlock(&hists->lock);
return he;
}
int64_t
hist_entry__cmp(struct hist_entry *left, struct hist_entry *right)
{
struct sort_entry *se;
int64_t cmp = 0;
list_for_each_entry(se, &hist_entry__sort_list, list) {
cmp = se->se_cmp(left, right);
if (cmp)
break;
}
return cmp;
}
int64_t
hist_entry__collapse(struct hist_entry *left, struct hist_entry *right)
{
struct sort_entry *se;
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);
if (cmp)
break;
}
return cmp;
}
void hist_entry__free(struct hist_entry *he)
{
free(he);
}
/*
* collapse the histogram
*/
static bool hists__collapse_insert_entry(struct hists *hists,
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
struct rb_root *root,
struct hist_entry *he)
{
struct rb_node **p = &root->rb_node;
struct rb_node *parent = NULL;
struct hist_entry *iter;
int64_t cmp;
while (*p != NULL) {
parent = *p;
iter = rb_entry(parent, struct hist_entry, rb_node_in);
cmp = hist_entry__collapse(iter, he);
if (!cmp) {
iter->period += he->period;
iter->nr_events += he->nr_events;
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
if (symbol_conf.use_callchain) {
callchain_cursor_reset(&hists->callchain_cursor);
callchain_merge(&hists->callchain_cursor, iter->callchain,
perf callchain: Feed callchains into a cursor The callchains are fed with an array of a fixed size. As a result we iterate over each callchains three times: - 1st to resolve symbols - 2nd to filter out context boundaries - 3rd for the insertion into the tree This also involves some pairs of memory allocation/deallocation everytime we insert a callchain, for the filtered out array of addresses and for the array of symbols that comes along. Instead, feed the callchains through a linked list with persistent allocations. It brings several pros like: - Merge the 1st and 2nd iterations in one. That was possible before but in a way that would involve allocating an array slightly taller than necessary because we don't know in advance the number of context boundaries to filter out. - Much lesser allocations/deallocations. The linked list keeps persistent empty entries for the next usages and is extendable at will. - Makes it easier for multiple sources of callchains to feed a stacktrace together. This is deemed to pave the way for cfi based callchains wherein traditional frame pointer based kernel stacktraces will precede cfi based user ones, producing an overall callchain which size is hardly predictable. This requirement makes the static array obsolete and makes a linked list based iterator a much more flexible fit. Basic testing on a big perf file containing callchains (~ 176 MB) has shown a throughput gain of about 11% with perf report. Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-01-14 11:51:58 +08:00
he->callchain);
}
hist_entry__free(he);
return false;
}
if (cmp < 0)
p = &(*p)->rb_left;
else
p = &(*p)->rb_right;
}
rb_link_node(&he->rb_node_in, parent, p);
rb_insert_color(&he->rb_node_in, root);
return true;
}
static struct rb_root *hists__get_rotate_entries_in(struct hists *hists)
{
struct rb_root *root;
pthread_mutex_lock(&hists->lock);
root = hists->entries_in;
if (++hists->entries_in > &hists->entries_in_array[1])
hists->entries_in = &hists->entries_in_array[0];
pthread_mutex_unlock(&hists->lock);
return root;
}
static void hists__apply_filters(struct hists *hists, struct hist_entry *he)
{
hists__filter_entry_by_dso(hists, he);
hists__filter_entry_by_thread(hists, he);
}
static void __hists__collapse_resort(struct hists *hists, bool threaded)
{
struct rb_root *root;
struct rb_node *next;
struct hist_entry *n;
if (!sort__need_collapse && !threaded)
return;
root = hists__get_rotate_entries_in(hists);
next = rb_first(root);
while (next) {
n = rb_entry(next, struct hist_entry, rb_node_in);
next = rb_next(&n->rb_node_in);
rb_erase(&n->rb_node_in, root);
if (hists__collapse_insert_entry(hists, &hists->entries_collapsed, n)) {
/*
* If it wasn't combined with one of the entries already
* collapsed, we need to apply the filters that may have
* been set by, say, the hist_browser.
*/
hists__apply_filters(hists, n);
}
}
}
void hists__collapse_resort(struct hists *hists)
{
return __hists__collapse_resort(hists, false);
}
void hists__collapse_resort_threaded(struct hists *hists)
{
return __hists__collapse_resort(hists, true);
}
/*
* reverse the map, sort on period.
*/
2010-05-11 00:04:11 +08:00
static void __hists__insert_output_entry(struct rb_root *entries,
struct hist_entry *he,
u64 min_callchain_hits)
{
2010-05-11 00:04:11 +08:00
struct rb_node **p = &entries->rb_node;
struct rb_node *parent = NULL;
struct hist_entry *iter;
if (symbol_conf.use_callchain)
callchain_param.sort(&he->sorted_chain, he->callchain,
min_callchain_hits, &callchain_param);
while (*p != NULL) {
parent = *p;
iter = rb_entry(parent, struct hist_entry, rb_node);
if (he->period > iter->period)
p = &(*p)->rb_left;
else
p = &(*p)->rb_right;
}
rb_link_node(&he->rb_node, parent, p);
2010-05-11 00:04:11 +08:00
rb_insert_color(&he->rb_node, entries);
}
static void __hists__output_resort(struct hists *hists, bool threaded)
{
struct rb_root *root;
struct rb_node *next;
struct hist_entry *n;
u64 min_callchain_hits;
min_callchain_hits = hists->stats.total_period * (callchain_param.min_percent / 100);
if (sort__need_collapse || threaded)
root = &hists->entries_collapsed;
else
root = hists->entries_in;
next = rb_first(root);
hists->entries = RB_ROOT;
hists->nr_entries = 0;
hists->stats.total_period = 0;
hists__reset_col_len(hists);
while (next) {
n = rb_entry(next, struct hist_entry, rb_node_in);
next = rb_next(&n->rb_node_in);
__hists__insert_output_entry(&hists->entries, n, min_callchain_hits);
hists__inc_nr_entries(hists, n);
}
}
void hists__output_resort(struct hists *hists)
{
return __hists__output_resort(hists, false);
}
void hists__output_resort_threaded(struct hists *hists)
{
return __hists__output_resort(hists, true);
}
static size_t callchain__fprintf_left_margin(FILE *fp, int left_margin)
{
int i;
int ret = fprintf(fp, " ");
for (i = 0; i < left_margin; i++)
ret += fprintf(fp, " ");
return ret;
}
static size_t ipchain__fprintf_graph_line(FILE *fp, int depth, int depth_mask,
int left_margin)
{
int i;
size_t ret = callchain__fprintf_left_margin(fp, left_margin);
for (i = 0; i < depth; i++)
if (depth_mask & (1 << i))
ret += fprintf(fp, "| ");
else
ret += fprintf(fp, " ");
ret += fprintf(fp, "\n");
return ret;
}
static size_t ipchain__fprintf_graph(FILE *fp, struct callchain_list *chain,
int depth, int depth_mask, int period,
u64 total_samples, u64 hits,
int left_margin)
{
int i;
size_t ret = 0;
ret += callchain__fprintf_left_margin(fp, left_margin);
for (i = 0; i < depth; i++) {
if (depth_mask & (1 << i))
ret += fprintf(fp, "|");
else
ret += fprintf(fp, " ");
if (!period && i == depth - 1) {
double percent;
percent = hits * 100.0 / total_samples;
ret += percent_color_fprintf(fp, "--%2.2f%%-- ", percent);
} else
ret += fprintf(fp, "%s", " ");
}
if (chain->ms.sym)
ret += fprintf(fp, "%s\n", chain->ms.sym->name);
else
ret += fprintf(fp, "%p\n", (void *)(long)chain->ip);
return ret;
}
static struct symbol *rem_sq_bracket;
static struct callchain_list rem_hits;
static void init_rem_hits(void)
{
rem_sq_bracket = malloc(sizeof(*rem_sq_bracket) + 6);
if (!rem_sq_bracket) {
fprintf(stderr, "Not enough memory to display remaining hits\n");
return;
}
strcpy(rem_sq_bracket->name, "[...]");
rem_hits.ms.sym = rem_sq_bracket;
}
static size_t __callchain__fprintf_graph(FILE *fp, struct callchain_node *self,
u64 total_samples, int depth,
int depth_mask, int left_margin)
{
struct rb_node *node, *next;
struct callchain_node *child;
struct callchain_list *chain;
int new_depth_mask = depth_mask;
u64 new_total;
u64 remaining;
size_t ret = 0;
int i;
uint entries_printed = 0;
if (callchain_param.mode == CHAIN_GRAPH_REL)
new_total = self->children_hit;
else
new_total = total_samples;
remaining = new_total;
node = rb_first(&self->rb_root);
while (node) {
u64 cumul;
child = rb_entry(node, struct callchain_node, rb_node);
cumul = callchain_cumul_hits(child);
remaining -= cumul;
/*
* The depth mask manages the output of pipes that show
* the depth. We don't want to keep the pipes of the current
* level for the last child of this depth.
* Except if we have remaining filtered hits. They will
* supersede the last child
*/
next = rb_next(node);
if (!next && (callchain_param.mode != CHAIN_GRAPH_REL || !remaining))
new_depth_mask &= ~(1 << (depth - 1));
/*
* But we keep the older depth mask for the line separator
* to keep the level link until we reach the last child
*/
ret += ipchain__fprintf_graph_line(fp, depth, depth_mask,
left_margin);
i = 0;
list_for_each_entry(chain, &child->val, list) {
ret += ipchain__fprintf_graph(fp, chain, depth,
new_depth_mask, i++,
new_total,
cumul,
left_margin);
}
ret += __callchain__fprintf_graph(fp, child, new_total,
depth + 1,
new_depth_mask | (1 << depth),
left_margin);
node = next;
if (++entries_printed == callchain_param.print_limit)
break;
}
if (callchain_param.mode == CHAIN_GRAPH_REL &&
remaining && remaining != new_total) {
if (!rem_sq_bracket)
return ret;
new_depth_mask &= ~(1 << (depth - 1));
ret += ipchain__fprintf_graph(fp, &rem_hits, depth,
new_depth_mask, 0, new_total,
remaining, left_margin);
}
return ret;
}
static size_t callchain__fprintf_graph(FILE *fp, struct callchain_node *self,
u64 total_samples, int left_margin)
{
struct callchain_list *chain;
bool printed = false;
int i = 0;
int ret = 0;
u32 entries_printed = 0;
list_for_each_entry(chain, &self->val, list) {
if (!i++ && sort__first_dimension == SORT_SYM)
continue;
if (!printed) {
ret += callchain__fprintf_left_margin(fp, left_margin);
ret += fprintf(fp, "|\n");
ret += callchain__fprintf_left_margin(fp, left_margin);
ret += fprintf(fp, "---");
left_margin += 3;
printed = true;
} else
ret += callchain__fprintf_left_margin(fp, left_margin);
if (chain->ms.sym)
ret += fprintf(fp, " %s\n", chain->ms.sym->name);
else
ret += fprintf(fp, " %p\n", (void *)(long)chain->ip);
if (++entries_printed == callchain_param.print_limit)
break;
}
ret += __callchain__fprintf_graph(fp, self, total_samples, 1, 1, left_margin);
return ret;
}
static size_t callchain__fprintf_flat(FILE *fp, struct callchain_node *self,
u64 total_samples)
{
struct callchain_list *chain;
size_t ret = 0;
if (!self)
return 0;
ret += callchain__fprintf_flat(fp, self->parent, total_samples);
list_for_each_entry(chain, &self->val, list) {
if (chain->ip >= PERF_CONTEXT_MAX)
continue;
if (chain->ms.sym)
ret += fprintf(fp, " %s\n", chain->ms.sym->name);
else
ret += fprintf(fp, " %p\n",
(void *)(long)chain->ip);
}
return ret;
}
static size_t hist_entry_callchain__fprintf(struct hist_entry *he,
u64 total_samples, int left_margin,
FILE *fp)
{
struct rb_node *rb_node;
struct callchain_node *chain;
size_t ret = 0;
u32 entries_printed = 0;
rb_node = rb_first(&he->sorted_chain);
while (rb_node) {
double percent;
chain = rb_entry(rb_node, struct callchain_node, rb_node);
percent = chain->hit * 100.0 / total_samples;
switch (callchain_param.mode) {
case CHAIN_FLAT:
ret += percent_color_fprintf(fp, " %6.2f%%\n",
percent);
ret += callchain__fprintf_flat(fp, chain, total_samples);
break;
case CHAIN_GRAPH_ABS: /* Falldown */
case CHAIN_GRAPH_REL:
ret += callchain__fprintf_graph(fp, chain, total_samples,
left_margin);
case CHAIN_NONE:
default:
break;
}
ret += fprintf(fp, "\n");
if (++entries_printed == callchain_param.print_limit)
break;
rb_node = rb_next(rb_node);
}
return ret;
}
perf top: Reuse the 'report' hist_entry/hists classes This actually fixes several problems we had in the old 'perf top': 1. Unresolved symbols not show, limitation that came from the old "KernelTop" codebase, to solve it we would need to do changes that would make sym_entry have most of the hist_entry fields. 2. It was using the number of samples, not the sum of sample->period. And brings the --sort code that allows us to have all the views in 'perf report', for instance: [root@emilia ~]# perf top --sort dso PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs) ------------------------------------------------------------------------------ 31.59% libcrypto.so.1.0.0 21.55% [kernel] 18.57% libpython2.6.so.1.0 7.04% libc-2.12.so 6.99% _backend_agg.so 4.72% sshd 1.48% multiarray.so 1.39% libfreetype.so.6.3.22 1.37% perf 0.71% libgobject-2.0.so.0.2200.5 0.53% [tg3] 0.48% libglib-2.0.so.0.2200.5 0.44% libstdc++.so.6.0.13 0.40% libcairo.so.2.10800.8 0.38% libm-2.12.so 0.34% umath.so 0.30% libgdk-x11-2.0.so.0.1800.9 0.22% libpthread-2.12.so 0.20% libgtk-x11-2.0.so.0.1800.9 0.20% librt-2.12.so 0.15% _path.so 0.13% libpango-1.0.so.0.2800.1 0.11% libatlas.so.3.0 0.09% ft2font.so 0.09% libpangoft2-1.0.so.0.2800.1 0.08% libX11.so.6.3.0 0.07% [vdso] 0.06% cyclictest ^C All the filter lists can be used as well: --dsos, --comms, --symbols, etc. The 'perf report' TUI is also reused, being possible to apply all the zoom operations, do annotation, etc. This change will allow multiple simplifications in the symbol system as well, that will be detailed in upcoming changesets. Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Stephane Eranian <eranian@google.com> Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 06:16:15 +08:00
void hists__output_recalc_col_len(struct hists *hists, int max_rows)
{
struct rb_node *next = rb_first(&hists->entries);
struct hist_entry *n;
int row = 0;
hists__reset_col_len(hists);
while (next && row++ < max_rows) {
n = rb_entry(next, struct hist_entry, rb_node);
if (!n->filtered)
hists__calc_col_len(hists, n);
perf top: Reuse the 'report' hist_entry/hists classes This actually fixes several problems we had in the old 'perf top': 1. Unresolved symbols not show, limitation that came from the old "KernelTop" codebase, to solve it we would need to do changes that would make sym_entry have most of the hist_entry fields. 2. It was using the number of samples, not the sum of sample->period. And brings the --sort code that allows us to have all the views in 'perf report', for instance: [root@emilia ~]# perf top --sort dso PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs) ------------------------------------------------------------------------------ 31.59% libcrypto.so.1.0.0 21.55% [kernel] 18.57% libpython2.6.so.1.0 7.04% libc-2.12.so 6.99% _backend_agg.so 4.72% sshd 1.48% multiarray.so 1.39% libfreetype.so.6.3.22 1.37% perf 0.71% libgobject-2.0.so.0.2200.5 0.53% [tg3] 0.48% libglib-2.0.so.0.2200.5 0.44% libstdc++.so.6.0.13 0.40% libcairo.so.2.10800.8 0.38% libm-2.12.so 0.34% umath.so 0.30% libgdk-x11-2.0.so.0.1800.9 0.22% libpthread-2.12.so 0.20% libgtk-x11-2.0.so.0.1800.9 0.20% librt-2.12.so 0.15% _path.so 0.13% libpango-1.0.so.0.2800.1 0.11% libatlas.so.3.0 0.09% ft2font.so 0.09% libpangoft2-1.0.so.0.2800.1 0.08% libX11.so.6.3.0 0.07% [vdso] 0.06% cyclictest ^C All the filter lists can be used as well: --dsos, --comms, --symbols, etc. The 'perf report' TUI is also reused, being possible to apply all the zoom operations, do annotation, etc. This change will allow multiple simplifications in the symbol system as well, that will be detailed in upcoming changesets. Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Stephane Eranian <eranian@google.com> Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 06:16:15 +08:00
next = rb_next(&n->rb_node);
}
}
static int hist_entry__pcnt_snprintf(struct hist_entry *he, char *s,
size_t size, struct hists *pair_hists,
bool show_displacement, long displacement,
bool color, u64 total_period)
{
u64 period, total, period_sys, period_us, period_guest_sys, period_guest_us;
perf hists: Print number of samples, not the period sum So that we match the header where we state the number of events with the "Samples" column when using 'perf report -n/--show-nr-samples': [root@emilia ~]# perf record -a sleep 1 [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.111 MB perf.data (~4860 samples) ] [root@emilia ~]# perf report --stdio --show-nr-samples # Events: 11 cycles # # Overhead Samples Command Shared Object Symbol # ........ .......... ........... .................. ............................ # 16.65% 1 sleep [kernel.kallsyms] [k] unmap_vmas 16.10% 1 perf libpthread-2.12.so [.] __pthread_cleanup_push_defer 15.79% 2 perf [kernel.kallsyms] [k] format_decode 12.88% 1 kworker/1:2 [kernel.kallsyms] [k] cache_reap 10.69% 1 swapper [kernel.kallsyms] [k] _raw_spin_lock 7.55% 1 sleep [kernel.kallsyms] [k] prepare_exec_creds 6.00% 1 perf [jbd2] [k] start_this_handle 5.29% 1 perf [kernel.kallsyms] [k] seq_read 4.75% 1 perf [kernel.kallsyms] [k] get_pid_task 4.30% 1 perf [kernel.kallsyms] [k] _raw_spin_unlock_irqrestore # # (For a higher level overview, try: perf report --sort comm,dso) # [root@emilia ~]# Reported-by: Stephane Eranian <eranian@google.com> Acked-by: Stephane Eranian <eranian@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Stephane Eranian <eranian@google.com> Cc: Tom Zanussi <tzanussi@gmail.com> LKML-Reference: <new-submission> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-02-17 20:37:23 +08:00
u64 nr_events;
perf diff: Use perf_session__fprintf_hists just like 'perf record' That means that almost everything you can do with 'perf report' can be done with 'perf diff', for instance: $ 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) ] $ 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 (~2687 samples) ] perf diff | head -8 9.02% +1.00% find libc-2.10.1.so [.] _IO_vfprintf_internal 2.91% -1.00% find [kernel] [k] __kmalloc 2.85% -1.00% find [kernel] [k] ext4_htree_store_dirent 1.99% -1.00% find [kernel] [k] _atomic_dec_and_lock 2.44% find [kernel] [k] half_md4_transform $ So if you want to zoom into libc: $ perf diff --dsos libc-2.10.1.so | head -8 37.34% find [.] _IO_vfprintf_internal 10.34% find [.] __GI_memmove 8.25% +2.00% find [.] _int_malloc 5.07% -1.00% find [.] __GI_mempcpy 7.62% +2.00% find [.] _int_free $ And if there were multiple commands using libc, it is also possible to aggregate them all by using --sort symbol: $ perf diff --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% [.] __GI_mempcpy 7.62% +2.00% [.] _int_free $ The displacement column now is off by default, to use it: perf diff -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% +2 [.] __GI_mempcpy 7.62% +2.00% -1 [.] _int_free $ Using -t/--field-separator can be used for scripting: $ perf diff -t, -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34, , ,[.] _IO_vfprintf_internal 10.34, , ,[.] __GI_memmove 8.25,+2.00%, ,[.] _int_malloc 5.07,-1.00%, +2,[.] __GI_mempcpy 7.62,+2.00%, -1,[.] _int_free 6.99,+1.00%, -1,[.] _IO_new_file_xsputn 1.89,-2.00%, +4,[.] __readdir64 $ Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.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> LKML-Reference: <1260978567-550-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-16 23:49:27 +08:00
const char *sep = symbol_conf.field_sep;
int ret;
if (symbol_conf.exclude_other && !he->parent)
return 0;
2010-05-11 00:04:11 +08:00
if (pair_hists) {
period = he->pair ? he->pair->period : 0;
nr_events = he->pair ? he->pair->nr_events : 0;
total = pair_hists->stats.total_period;
period_sys = he->pair ? he->pair->period_sys : 0;
period_us = he->pair ? he->pair->period_us : 0;
period_guest_sys = he->pair ? he->pair->period_guest_sys : 0;
period_guest_us = he->pair ? he->pair->period_guest_us : 0;
perf diff: Use perf_session__fprintf_hists just like 'perf record' That means that almost everything you can do with 'perf report' can be done with 'perf diff', for instance: $ 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) ] $ 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 (~2687 samples) ] perf diff | head -8 9.02% +1.00% find libc-2.10.1.so [.] _IO_vfprintf_internal 2.91% -1.00% find [kernel] [k] __kmalloc 2.85% -1.00% find [kernel] [k] ext4_htree_store_dirent 1.99% -1.00% find [kernel] [k] _atomic_dec_and_lock 2.44% find [kernel] [k] half_md4_transform $ So if you want to zoom into libc: $ perf diff --dsos libc-2.10.1.so | head -8 37.34% find [.] _IO_vfprintf_internal 10.34% find [.] __GI_memmove 8.25% +2.00% find [.] _int_malloc 5.07% -1.00% find [.] __GI_mempcpy 7.62% +2.00% find [.] _int_free $ And if there were multiple commands using libc, it is also possible to aggregate them all by using --sort symbol: $ perf diff --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% [.] __GI_mempcpy 7.62% +2.00% [.] _int_free $ The displacement column now is off by default, to use it: perf diff -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% +2 [.] __GI_mempcpy 7.62% +2.00% -1 [.] _int_free $ Using -t/--field-separator can be used for scripting: $ perf diff -t, -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34, , ,[.] _IO_vfprintf_internal 10.34, , ,[.] __GI_memmove 8.25,+2.00%, ,[.] _int_malloc 5.07,-1.00%, +2,[.] __GI_mempcpy 7.62,+2.00%, -1,[.] _int_free 6.99,+1.00%, -1,[.] _IO_new_file_xsputn 1.89,-2.00%, +4,[.] __readdir64 $ Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.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> LKML-Reference: <1260978567-550-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-16 23:49:27 +08:00
} else {
period = he->period;
nr_events = he->nr_events;
total = total_period;
period_sys = he->period_sys;
period_us = he->period_us;
period_guest_sys = he->period_guest_sys;
period_guest_us = he->period_guest_us;
perf diff: Use perf_session__fprintf_hists just like 'perf record' That means that almost everything you can do with 'perf report' can be done with 'perf diff', for instance: $ 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) ] $ 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 (~2687 samples) ] perf diff | head -8 9.02% +1.00% find libc-2.10.1.so [.] _IO_vfprintf_internal 2.91% -1.00% find [kernel] [k] __kmalloc 2.85% -1.00% find [kernel] [k] ext4_htree_store_dirent 1.99% -1.00% find [kernel] [k] _atomic_dec_and_lock 2.44% find [kernel] [k] half_md4_transform $ So if you want to zoom into libc: $ perf diff --dsos libc-2.10.1.so | head -8 37.34% find [.] _IO_vfprintf_internal 10.34% find [.] __GI_memmove 8.25% +2.00% find [.] _int_malloc 5.07% -1.00% find [.] __GI_mempcpy 7.62% +2.00% find [.] _int_free $ And if there were multiple commands using libc, it is also possible to aggregate them all by using --sort symbol: $ perf diff --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% [.] __GI_mempcpy 7.62% +2.00% [.] _int_free $ The displacement column now is off by default, to use it: perf diff -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% +2 [.] __GI_mempcpy 7.62% +2.00% -1 [.] _int_free $ Using -t/--field-separator can be used for scripting: $ perf diff -t, -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34, , ,[.] _IO_vfprintf_internal 10.34, , ,[.] __GI_memmove 8.25,+2.00%, ,[.] _int_malloc 5.07,-1.00%, +2,[.] __GI_mempcpy 7.62,+2.00%, -1,[.] _int_free 6.99,+1.00%, -1,[.] _IO_new_file_xsputn 1.89,-2.00%, +4,[.] __readdir64 $ Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.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> LKML-Reference: <1260978567-550-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-16 23:49:27 +08:00
}
if (total) {
if (color)
ret = percent_color_snprintf(s, size,
sep ? "%.2f" : " %6.2f%%",
(period * 100.0) / total);
else
ret = snprintf(s, size, sep ? "%.2f" : " %6.2f%%",
(period * 100.0) / total);
if (symbol_conf.show_cpu_utilization) {
ret += percent_color_snprintf(s + ret, size - ret,
sep ? "%.2f" : " %6.2f%%",
(period_sys * 100.0) / total);
ret += percent_color_snprintf(s + ret, size - ret,
sep ? "%.2f" : " %6.2f%%",
(period_us * 100.0) / total);
if (perf_guest) {
ret += percent_color_snprintf(s + ret,
size - ret,
sep ? "%.2f" : " %6.2f%%",
(period_guest_sys * 100.0) /
total);
ret += percent_color_snprintf(s + ret,
size - ret,
sep ? "%.2f" : " %6.2f%%",
(period_guest_us * 100.0) /
total);
}
}
} else
ret = snprintf(s, size, sep ? "%" PRIu64 : "%12" PRIu64 " ", period);
if (symbol_conf.show_nr_samples) {
perf diff: Use perf_session__fprintf_hists just like 'perf record' That means that almost everything you can do with 'perf report' can be done with 'perf diff', for instance: $ 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) ] $ 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 (~2687 samples) ] perf diff | head -8 9.02% +1.00% find libc-2.10.1.so [.] _IO_vfprintf_internal 2.91% -1.00% find [kernel] [k] __kmalloc 2.85% -1.00% find [kernel] [k] ext4_htree_store_dirent 1.99% -1.00% find [kernel] [k] _atomic_dec_and_lock 2.44% find [kernel] [k] half_md4_transform $ So if you want to zoom into libc: $ perf diff --dsos libc-2.10.1.so | head -8 37.34% find [.] _IO_vfprintf_internal 10.34% find [.] __GI_memmove 8.25% +2.00% find [.] _int_malloc 5.07% -1.00% find [.] __GI_mempcpy 7.62% +2.00% find [.] _int_free $ And if there were multiple commands using libc, it is also possible to aggregate them all by using --sort symbol: $ perf diff --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% [.] __GI_mempcpy 7.62% +2.00% [.] _int_free $ The displacement column now is off by default, to use it: perf diff -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% +2 [.] __GI_mempcpy 7.62% +2.00% -1 [.] _int_free $ Using -t/--field-separator can be used for scripting: $ perf diff -t, -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34, , ,[.] _IO_vfprintf_internal 10.34, , ,[.] __GI_memmove 8.25,+2.00%, ,[.] _int_malloc 5.07,-1.00%, +2,[.] __GI_mempcpy 7.62,+2.00%, -1,[.] _int_free 6.99,+1.00%, -1,[.] _IO_new_file_xsputn 1.89,-2.00%, +4,[.] __readdir64 $ Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.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> LKML-Reference: <1260978567-550-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-16 23:49:27 +08:00
if (sep)
perf hists: Print number of samples, not the period sum So that we match the header where we state the number of events with the "Samples" column when using 'perf report -n/--show-nr-samples': [root@emilia ~]# perf record -a sleep 1 [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.111 MB perf.data (~4860 samples) ] [root@emilia ~]# perf report --stdio --show-nr-samples # Events: 11 cycles # # Overhead Samples Command Shared Object Symbol # ........ .......... ........... .................. ............................ # 16.65% 1 sleep [kernel.kallsyms] [k] unmap_vmas 16.10% 1 perf libpthread-2.12.so [.] __pthread_cleanup_push_defer 15.79% 2 perf [kernel.kallsyms] [k] format_decode 12.88% 1 kworker/1:2 [kernel.kallsyms] [k] cache_reap 10.69% 1 swapper [kernel.kallsyms] [k] _raw_spin_lock 7.55% 1 sleep [kernel.kallsyms] [k] prepare_exec_creds 6.00% 1 perf [jbd2] [k] start_this_handle 5.29% 1 perf [kernel.kallsyms] [k] seq_read 4.75% 1 perf [kernel.kallsyms] [k] get_pid_task 4.30% 1 perf [kernel.kallsyms] [k] _raw_spin_unlock_irqrestore # # (For a higher level overview, try: perf report --sort comm,dso) # [root@emilia ~]# Reported-by: Stephane Eranian <eranian@google.com> Acked-by: Stephane Eranian <eranian@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Stephane Eranian <eranian@google.com> Cc: Tom Zanussi <tzanussi@gmail.com> LKML-Reference: <new-submission> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-02-17 20:37:23 +08:00
ret += snprintf(s + ret, size - ret, "%c%" PRIu64, *sep, nr_events);
else
perf hists: Print number of samples, not the period sum So that we match the header where we state the number of events with the "Samples" column when using 'perf report -n/--show-nr-samples': [root@emilia ~]# perf record -a sleep 1 [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.111 MB perf.data (~4860 samples) ] [root@emilia ~]# perf report --stdio --show-nr-samples # Events: 11 cycles # # Overhead Samples Command Shared Object Symbol # ........ .......... ........... .................. ............................ # 16.65% 1 sleep [kernel.kallsyms] [k] unmap_vmas 16.10% 1 perf libpthread-2.12.so [.] __pthread_cleanup_push_defer 15.79% 2 perf [kernel.kallsyms] [k] format_decode 12.88% 1 kworker/1:2 [kernel.kallsyms] [k] cache_reap 10.69% 1 swapper [kernel.kallsyms] [k] _raw_spin_lock 7.55% 1 sleep [kernel.kallsyms] [k] prepare_exec_creds 6.00% 1 perf [jbd2] [k] start_this_handle 5.29% 1 perf [kernel.kallsyms] [k] seq_read 4.75% 1 perf [kernel.kallsyms] [k] get_pid_task 4.30% 1 perf [kernel.kallsyms] [k] _raw_spin_unlock_irqrestore # # (For a higher level overview, try: perf report --sort comm,dso) # [root@emilia ~]# Reported-by: Stephane Eranian <eranian@google.com> Acked-by: Stephane Eranian <eranian@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Stephane Eranian <eranian@google.com> Cc: Tom Zanussi <tzanussi@gmail.com> LKML-Reference: <new-submission> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-02-17 20:37:23 +08:00
ret += snprintf(s + ret, size - ret, "%11" PRIu64, nr_events);
perf diff: Use perf_session__fprintf_hists just like 'perf record' That means that almost everything you can do with 'perf report' can be done with 'perf diff', for instance: $ 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) ] $ 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 (~2687 samples) ] perf diff | head -8 9.02% +1.00% find libc-2.10.1.so [.] _IO_vfprintf_internal 2.91% -1.00% find [kernel] [k] __kmalloc 2.85% -1.00% find [kernel] [k] ext4_htree_store_dirent 1.99% -1.00% find [kernel] [k] _atomic_dec_and_lock 2.44% find [kernel] [k] half_md4_transform $ So if you want to zoom into libc: $ perf diff --dsos libc-2.10.1.so | head -8 37.34% find [.] _IO_vfprintf_internal 10.34% find [.] __GI_memmove 8.25% +2.00% find [.] _int_malloc 5.07% -1.00% find [.] __GI_mempcpy 7.62% +2.00% find [.] _int_free $ And if there were multiple commands using libc, it is also possible to aggregate them all by using --sort symbol: $ perf diff --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% [.] __GI_mempcpy 7.62% +2.00% [.] _int_free $ The displacement column now is off by default, to use it: perf diff -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% +2 [.] __GI_mempcpy 7.62% +2.00% -1 [.] _int_free $ Using -t/--field-separator can be used for scripting: $ perf diff -t, -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34, , ,[.] _IO_vfprintf_internal 10.34, , ,[.] __GI_memmove 8.25,+2.00%, ,[.] _int_malloc 5.07,-1.00%, +2,[.] __GI_mempcpy 7.62,+2.00%, -1,[.] _int_free 6.99,+1.00%, -1,[.] _IO_new_file_xsputn 1.89,-2.00%, +4,[.] __readdir64 $ Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.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> LKML-Reference: <1260978567-550-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-16 23:49:27 +08:00
}
if (symbol_conf.show_total_period) {
if (sep)
ret += snprintf(s + ret, size - ret, "%c%" PRIu64, *sep, period);
else
ret += snprintf(s + ret, size - ret, " %12" PRIu64, period);
}
2010-05-11 00:04:11 +08:00
if (pair_hists) {
perf diff: Use perf_session__fprintf_hists just like 'perf record' That means that almost everything you can do with 'perf report' can be done with 'perf diff', for instance: $ 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) ] $ 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 (~2687 samples) ] perf diff | head -8 9.02% +1.00% find libc-2.10.1.so [.] _IO_vfprintf_internal 2.91% -1.00% find [kernel] [k] __kmalloc 2.85% -1.00% find [kernel] [k] ext4_htree_store_dirent 1.99% -1.00% find [kernel] [k] _atomic_dec_and_lock 2.44% find [kernel] [k] half_md4_transform $ So if you want to zoom into libc: $ perf diff --dsos libc-2.10.1.so | head -8 37.34% find [.] _IO_vfprintf_internal 10.34% find [.] __GI_memmove 8.25% +2.00% find [.] _int_malloc 5.07% -1.00% find [.] __GI_mempcpy 7.62% +2.00% find [.] _int_free $ And if there were multiple commands using libc, it is also possible to aggregate them all by using --sort symbol: $ perf diff --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% [.] __GI_mempcpy 7.62% +2.00% [.] _int_free $ The displacement column now is off by default, to use it: perf diff -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% +2 [.] __GI_mempcpy 7.62% +2.00% -1 [.] _int_free $ Using -t/--field-separator can be used for scripting: $ perf diff -t, -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34, , ,[.] _IO_vfprintf_internal 10.34, , ,[.] __GI_memmove 8.25,+2.00%, ,[.] _int_malloc 5.07,-1.00%, +2,[.] __GI_mempcpy 7.62,+2.00%, -1,[.] _int_free 6.99,+1.00%, -1,[.] _IO_new_file_xsputn 1.89,-2.00%, +4,[.] __readdir64 $ Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.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> LKML-Reference: <1260978567-550-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-16 23:49:27 +08:00
char bf[32];
double old_percent = 0, new_percent = 0, diff;
if (total > 0)
old_percent = (period * 100.0) / total;
if (total_period > 0)
new_percent = (he->period * 100.0) / total_period;
perf diff: Use perf_session__fprintf_hists just like 'perf record' That means that almost everything you can do with 'perf report' can be done with 'perf diff', for instance: $ 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) ] $ 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 (~2687 samples) ] perf diff | head -8 9.02% +1.00% find libc-2.10.1.so [.] _IO_vfprintf_internal 2.91% -1.00% find [kernel] [k] __kmalloc 2.85% -1.00% find [kernel] [k] ext4_htree_store_dirent 1.99% -1.00% find [kernel] [k] _atomic_dec_and_lock 2.44% find [kernel] [k] half_md4_transform $ So if you want to zoom into libc: $ perf diff --dsos libc-2.10.1.so | head -8 37.34% find [.] _IO_vfprintf_internal 10.34% find [.] __GI_memmove 8.25% +2.00% find [.] _int_malloc 5.07% -1.00% find [.] __GI_mempcpy 7.62% +2.00% find [.] _int_free $ And if there were multiple commands using libc, it is also possible to aggregate them all by using --sort symbol: $ perf diff --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% [.] __GI_mempcpy 7.62% +2.00% [.] _int_free $ The displacement column now is off by default, to use it: perf diff -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% +2 [.] __GI_mempcpy 7.62% +2.00% -1 [.] _int_free $ Using -t/--field-separator can be used for scripting: $ perf diff -t, -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34, , ,[.] _IO_vfprintf_internal 10.34, , ,[.] __GI_memmove 8.25,+2.00%, ,[.] _int_malloc 5.07,-1.00%, +2,[.] __GI_mempcpy 7.62,+2.00%, -1,[.] _int_free 6.99,+1.00%, -1,[.] _IO_new_file_xsputn 1.89,-2.00%, +4,[.] __readdir64 $ Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.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> LKML-Reference: <1260978567-550-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-16 23:49:27 +08:00
perf diff: Percent calcs should use double values Otherwise we do integer math and the delta values round up to multiples of 1.0%. Also, calculate absolute values. Things look precise now: $ perf report -i perf.data.old --sort dso,symbol | head -13 9.02% libc-2.10.1.so [.] _IO_vfprintf_internal 4.88% find [.] 0x00000000014af0 2.91% [kernel] [k] __kmalloc 2.85% [kernel] [k] ext4_htree_store_dirent 2.50% libc-2.10.1.so [.] __GI_memmove 2.44% [kernel] [k] half_md4_transform 2.43% [kernel] [k] _spin_lock 2.33% [kernel] [k] system_call $ perf report -i perf.data --sort dso,symbol | head -13 8.55% libc-2.10.1.so [.] _IO_vfprintf_internal 3.11% [kernel] [k] __kmalloc 3.07% [kernel] [k] ext4_htree_store_dirent 2.66% find [.] 0x00000000016bcf 2.61% [kernel] [k] _atomic_dec_and_lock 2.46% [kernel] [k] half_md4_transform 2.41% libc-2.10.1.so [.] __GI_memmove 2.30% find [.] 0x00000000009219 $ perf diff | head -13 9.02% -0.47% libc-2.10.1.so [.] _IO_vfprintf_internal 2.91% +0.20% [kernel] [k] __kmalloc 2.85% +0.23% [kernel] [k] ext4_htree_store_dirent 1.99% +0.62% [kernel] [k] _atomic_dec_and_lock 2.44% +0.02% [kernel] [k] half_md4_transform 2.50% -0.09% libc-2.10.1.so [.] __GI_memmove 1.88% +0.01% [kernel] [k] __d_lookup 2.43% -0.75% [kernel] [k] _spin_lock 0.97% +0.62% [kernel] [k] path_get 1.99% -0.42% libc-2.10.1.so [.] _int_malloc $ Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.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> LKML-Reference: <1260981109-2621-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-17 00:31:49 +08:00
diff = new_percent - old_percent;
perf diff: Use perf_session__fprintf_hists just like 'perf record' That means that almost everything you can do with 'perf report' can be done with 'perf diff', for instance: $ 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) ] $ 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 (~2687 samples) ] perf diff | head -8 9.02% +1.00% find libc-2.10.1.so [.] _IO_vfprintf_internal 2.91% -1.00% find [kernel] [k] __kmalloc 2.85% -1.00% find [kernel] [k] ext4_htree_store_dirent 1.99% -1.00% find [kernel] [k] _atomic_dec_and_lock 2.44% find [kernel] [k] half_md4_transform $ So if you want to zoom into libc: $ perf diff --dsos libc-2.10.1.so | head -8 37.34% find [.] _IO_vfprintf_internal 10.34% find [.] __GI_memmove 8.25% +2.00% find [.] _int_malloc 5.07% -1.00% find [.] __GI_mempcpy 7.62% +2.00% find [.] _int_free $ And if there were multiple commands using libc, it is also possible to aggregate them all by using --sort symbol: $ perf diff --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% [.] __GI_mempcpy 7.62% +2.00% [.] _int_free $ The displacement column now is off by default, to use it: perf diff -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% +2 [.] __GI_mempcpy 7.62% +2.00% -1 [.] _int_free $ Using -t/--field-separator can be used for scripting: $ perf diff -t, -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34, , ,[.] _IO_vfprintf_internal 10.34, , ,[.] __GI_memmove 8.25,+2.00%, ,[.] _int_malloc 5.07,-1.00%, +2,[.] __GI_mempcpy 7.62,+2.00%, -1,[.] _int_free 6.99,+1.00%, -1,[.] _IO_new_file_xsputn 1.89,-2.00%, +4,[.] __readdir64 $ Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.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> LKML-Reference: <1260978567-550-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-16 23:49:27 +08:00
perf diff: Percent calcs should use double values Otherwise we do integer math and the delta values round up to multiples of 1.0%. Also, calculate absolute values. Things look precise now: $ perf report -i perf.data.old --sort dso,symbol | head -13 9.02% libc-2.10.1.so [.] _IO_vfprintf_internal 4.88% find [.] 0x00000000014af0 2.91% [kernel] [k] __kmalloc 2.85% [kernel] [k] ext4_htree_store_dirent 2.50% libc-2.10.1.so [.] __GI_memmove 2.44% [kernel] [k] half_md4_transform 2.43% [kernel] [k] _spin_lock 2.33% [kernel] [k] system_call $ perf report -i perf.data --sort dso,symbol | head -13 8.55% libc-2.10.1.so [.] _IO_vfprintf_internal 3.11% [kernel] [k] __kmalloc 3.07% [kernel] [k] ext4_htree_store_dirent 2.66% find [.] 0x00000000016bcf 2.61% [kernel] [k] _atomic_dec_and_lock 2.46% [kernel] [k] half_md4_transform 2.41% libc-2.10.1.so [.] __GI_memmove 2.30% find [.] 0x00000000009219 $ perf diff | head -13 9.02% -0.47% libc-2.10.1.so [.] _IO_vfprintf_internal 2.91% +0.20% [kernel] [k] __kmalloc 2.85% +0.23% [kernel] [k] ext4_htree_store_dirent 1.99% +0.62% [kernel] [k] _atomic_dec_and_lock 2.44% +0.02% [kernel] [k] half_md4_transform 2.50% -0.09% libc-2.10.1.so [.] __GI_memmove 1.88% +0.01% [kernel] [k] __d_lookup 2.43% -0.75% [kernel] [k] _spin_lock 0.97% +0.62% [kernel] [k] path_get 1.99% -0.42% libc-2.10.1.so [.] _int_malloc $ Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.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> LKML-Reference: <1260981109-2621-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-17 00:31:49 +08:00
if (fabs(diff) >= 0.01)
perf diff: Use perf_session__fprintf_hists just like 'perf record' That means that almost everything you can do with 'perf report' can be done with 'perf diff', for instance: $ 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) ] $ 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 (~2687 samples) ] perf diff | head -8 9.02% +1.00% find libc-2.10.1.so [.] _IO_vfprintf_internal 2.91% -1.00% find [kernel] [k] __kmalloc 2.85% -1.00% find [kernel] [k] ext4_htree_store_dirent 1.99% -1.00% find [kernel] [k] _atomic_dec_and_lock 2.44% find [kernel] [k] half_md4_transform $ So if you want to zoom into libc: $ perf diff --dsos libc-2.10.1.so | head -8 37.34% find [.] _IO_vfprintf_internal 10.34% find [.] __GI_memmove 8.25% +2.00% find [.] _int_malloc 5.07% -1.00% find [.] __GI_mempcpy 7.62% +2.00% find [.] _int_free $ And if there were multiple commands using libc, it is also possible to aggregate them all by using --sort symbol: $ perf diff --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% [.] __GI_mempcpy 7.62% +2.00% [.] _int_free $ The displacement column now is off by default, to use it: perf diff -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% +2 [.] __GI_mempcpy 7.62% +2.00% -1 [.] _int_free $ Using -t/--field-separator can be used for scripting: $ perf diff -t, -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34, , ,[.] _IO_vfprintf_internal 10.34, , ,[.] __GI_memmove 8.25,+2.00%, ,[.] _int_malloc 5.07,-1.00%, +2,[.] __GI_mempcpy 7.62,+2.00%, -1,[.] _int_free 6.99,+1.00%, -1,[.] _IO_new_file_xsputn 1.89,-2.00%, +4,[.] __readdir64 $ Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.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> LKML-Reference: <1260978567-550-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-16 23:49:27 +08:00
snprintf(bf, sizeof(bf), "%+4.2F%%", diff);
else
snprintf(bf, sizeof(bf), " ");
if (sep)
ret += snprintf(s + ret, size - ret, "%c%s", *sep, bf);
perf diff: Use perf_session__fprintf_hists just like 'perf record' That means that almost everything you can do with 'perf report' can be done with 'perf diff', for instance: $ 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) ] $ 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 (~2687 samples) ] perf diff | head -8 9.02% +1.00% find libc-2.10.1.so [.] _IO_vfprintf_internal 2.91% -1.00% find [kernel] [k] __kmalloc 2.85% -1.00% find [kernel] [k] ext4_htree_store_dirent 1.99% -1.00% find [kernel] [k] _atomic_dec_and_lock 2.44% find [kernel] [k] half_md4_transform $ So if you want to zoom into libc: $ perf diff --dsos libc-2.10.1.so | head -8 37.34% find [.] _IO_vfprintf_internal 10.34% find [.] __GI_memmove 8.25% +2.00% find [.] _int_malloc 5.07% -1.00% find [.] __GI_mempcpy 7.62% +2.00% find [.] _int_free $ And if there were multiple commands using libc, it is also possible to aggregate them all by using --sort symbol: $ perf diff --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% [.] __GI_mempcpy 7.62% +2.00% [.] _int_free $ The displacement column now is off by default, to use it: perf diff -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% +2 [.] __GI_mempcpy 7.62% +2.00% -1 [.] _int_free $ Using -t/--field-separator can be used for scripting: $ perf diff -t, -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34, , ,[.] _IO_vfprintf_internal 10.34, , ,[.] __GI_memmove 8.25,+2.00%, ,[.] _int_malloc 5.07,-1.00%, +2,[.] __GI_mempcpy 7.62,+2.00%, -1,[.] _int_free 6.99,+1.00%, -1,[.] _IO_new_file_xsputn 1.89,-2.00%, +4,[.] __readdir64 $ Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.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> LKML-Reference: <1260978567-550-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-16 23:49:27 +08:00
else
ret += snprintf(s + ret, size - ret, "%11.11s", bf);
perf diff: Use perf_session__fprintf_hists just like 'perf record' That means that almost everything you can do with 'perf report' can be done with 'perf diff', for instance: $ 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) ] $ 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 (~2687 samples) ] perf diff | head -8 9.02% +1.00% find libc-2.10.1.so [.] _IO_vfprintf_internal 2.91% -1.00% find [kernel] [k] __kmalloc 2.85% -1.00% find [kernel] [k] ext4_htree_store_dirent 1.99% -1.00% find [kernel] [k] _atomic_dec_and_lock 2.44% find [kernel] [k] half_md4_transform $ So if you want to zoom into libc: $ perf diff --dsos libc-2.10.1.so | head -8 37.34% find [.] _IO_vfprintf_internal 10.34% find [.] __GI_memmove 8.25% +2.00% find [.] _int_malloc 5.07% -1.00% find [.] __GI_mempcpy 7.62% +2.00% find [.] _int_free $ And if there were multiple commands using libc, it is also possible to aggregate them all by using --sort symbol: $ perf diff --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% [.] __GI_mempcpy 7.62% +2.00% [.] _int_free $ The displacement column now is off by default, to use it: perf diff -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% +2 [.] __GI_mempcpy 7.62% +2.00% -1 [.] _int_free $ Using -t/--field-separator can be used for scripting: $ perf diff -t, -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34, , ,[.] _IO_vfprintf_internal 10.34, , ,[.] __GI_memmove 8.25,+2.00%, ,[.] _int_malloc 5.07,-1.00%, +2,[.] __GI_mempcpy 7.62,+2.00%, -1,[.] _int_free 6.99,+1.00%, -1,[.] _IO_new_file_xsputn 1.89,-2.00%, +4,[.] __readdir64 $ Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.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> LKML-Reference: <1260978567-550-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-16 23:49:27 +08:00
if (show_displacement) {
if (displacement)
snprintf(bf, sizeof(bf), "%+4ld", displacement);
else
snprintf(bf, sizeof(bf), " ");
if (sep)
ret += snprintf(s + ret, size - ret, "%c%s", *sep, bf);
perf diff: Use perf_session__fprintf_hists just like 'perf record' That means that almost everything you can do with 'perf report' can be done with 'perf diff', for instance: $ 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) ] $ 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 (~2687 samples) ] perf diff | head -8 9.02% +1.00% find libc-2.10.1.so [.] _IO_vfprintf_internal 2.91% -1.00% find [kernel] [k] __kmalloc 2.85% -1.00% find [kernel] [k] ext4_htree_store_dirent 1.99% -1.00% find [kernel] [k] _atomic_dec_and_lock 2.44% find [kernel] [k] half_md4_transform $ So if you want to zoom into libc: $ perf diff --dsos libc-2.10.1.so | head -8 37.34% find [.] _IO_vfprintf_internal 10.34% find [.] __GI_memmove 8.25% +2.00% find [.] _int_malloc 5.07% -1.00% find [.] __GI_mempcpy 7.62% +2.00% find [.] _int_free $ And if there were multiple commands using libc, it is also possible to aggregate them all by using --sort symbol: $ perf diff --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% [.] __GI_mempcpy 7.62% +2.00% [.] _int_free $ The displacement column now is off by default, to use it: perf diff -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% +2 [.] __GI_mempcpy 7.62% +2.00% -1 [.] _int_free $ Using -t/--field-separator can be used for scripting: $ perf diff -t, -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34, , ,[.] _IO_vfprintf_internal 10.34, , ,[.] __GI_memmove 8.25,+2.00%, ,[.] _int_malloc 5.07,-1.00%, +2,[.] __GI_mempcpy 7.62,+2.00%, -1,[.] _int_free 6.99,+1.00%, -1,[.] _IO_new_file_xsputn 1.89,-2.00%, +4,[.] __readdir64 $ Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.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> LKML-Reference: <1260978567-550-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-16 23:49:27 +08:00
else
ret += snprintf(s + ret, size - ret, "%6.6s", bf);
perf diff: Use perf_session__fprintf_hists just like 'perf record' That means that almost everything you can do with 'perf report' can be done with 'perf diff', for instance: $ 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) ] $ 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 (~2687 samples) ] perf diff | head -8 9.02% +1.00% find libc-2.10.1.so [.] _IO_vfprintf_internal 2.91% -1.00% find [kernel] [k] __kmalloc 2.85% -1.00% find [kernel] [k] ext4_htree_store_dirent 1.99% -1.00% find [kernel] [k] _atomic_dec_and_lock 2.44% find [kernel] [k] half_md4_transform $ So if you want to zoom into libc: $ perf diff --dsos libc-2.10.1.so | head -8 37.34% find [.] _IO_vfprintf_internal 10.34% find [.] __GI_memmove 8.25% +2.00% find [.] _int_malloc 5.07% -1.00% find [.] __GI_mempcpy 7.62% +2.00% find [.] _int_free $ And if there were multiple commands using libc, it is also possible to aggregate them all by using --sort symbol: $ perf diff --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% [.] __GI_mempcpy 7.62% +2.00% [.] _int_free $ The displacement column now is off by default, to use it: perf diff -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% +2 [.] __GI_mempcpy 7.62% +2.00% -1 [.] _int_free $ Using -t/--field-separator can be used for scripting: $ perf diff -t, -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34, , ,[.] _IO_vfprintf_internal 10.34, , ,[.] __GI_memmove 8.25,+2.00%, ,[.] _int_malloc 5.07,-1.00%, +2,[.] __GI_mempcpy 7.62,+2.00%, -1,[.] _int_free 6.99,+1.00%, -1,[.] _IO_new_file_xsputn 1.89,-2.00%, +4,[.] __readdir64 $ Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.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> LKML-Reference: <1260978567-550-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-16 23:49:27 +08:00
}
}
return ret;
}
int hist_entry__snprintf(struct hist_entry *he, char *s, size_t size,
struct hists *hists)
{
const char *sep = symbol_conf.field_sep;
struct sort_entry *se;
int ret = 0;
list_for_each_entry(se, &hist_entry__sort_list, list) {
if (se->elide)
continue;
ret += snprintf(s + ret, size - ret, "%s", sep ?: " ");
ret += se->se_snprintf(he, s + ret, size - ret,
hists__col_len(hists, se->se_width_idx));
}
return ret;
}
static int hist_entry__fprintf(struct hist_entry *he, size_t size,
struct hists *hists, struct hists *pair_hists,
bool show_displacement, long displacement,
u64 total_period, FILE *fp)
{
char bf[512];
int ret;
if (size == 0 || size > sizeof(bf))
size = sizeof(bf);
ret = hist_entry__pcnt_snprintf(he, bf, size, pair_hists,
show_displacement, displacement,
true, total_period);
hist_entry__snprintf(he, bf + ret, size - ret, hists);
return fprintf(fp, "%s\n", bf);
}
static size_t hist_entry__fprintf_callchain(struct hist_entry *he,
struct hists *hists,
u64 total_period, FILE *fp)
{
int left_margin = 0;
if (sort__first_dimension == SORT_COMM) {
struct sort_entry *se = list_first_entry(&hist_entry__sort_list,
typeof(*se), list);
left_margin = hists__col_len(hists, se->se_width_idx);
left_margin -= thread__comm_len(he->thread);
}
return hist_entry_callchain__fprintf(he, total_period, left_margin, fp);
}
size_t hists__fprintf(struct hists *hists, struct hists *pair,
bool show_displacement, bool show_header, int max_rows,
int max_cols, FILE *fp)
{
struct sort_entry *se;
struct rb_node *nd;
size_t ret = 0;
u64 total_period;
perf diff: Use perf_session__fprintf_hists just like 'perf record' That means that almost everything you can do with 'perf report' can be done with 'perf diff', for instance: $ 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) ] $ 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 (~2687 samples) ] perf diff | head -8 9.02% +1.00% find libc-2.10.1.so [.] _IO_vfprintf_internal 2.91% -1.00% find [kernel] [k] __kmalloc 2.85% -1.00% find [kernel] [k] ext4_htree_store_dirent 1.99% -1.00% find [kernel] [k] _atomic_dec_and_lock 2.44% find [kernel] [k] half_md4_transform $ So if you want to zoom into libc: $ perf diff --dsos libc-2.10.1.so | head -8 37.34% find [.] _IO_vfprintf_internal 10.34% find [.] __GI_memmove 8.25% +2.00% find [.] _int_malloc 5.07% -1.00% find [.] __GI_mempcpy 7.62% +2.00% find [.] _int_free $ And if there were multiple commands using libc, it is also possible to aggregate them all by using --sort symbol: $ perf diff --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% [.] __GI_mempcpy 7.62% +2.00% [.] _int_free $ The displacement column now is off by default, to use it: perf diff -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% +2 [.] __GI_mempcpy 7.62% +2.00% -1 [.] _int_free $ Using -t/--field-separator can be used for scripting: $ perf diff -t, -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34, , ,[.] _IO_vfprintf_internal 10.34, , ,[.] __GI_memmove 8.25,+2.00%, ,[.] _int_malloc 5.07,-1.00%, +2,[.] __GI_mempcpy 7.62,+2.00%, -1,[.] _int_free 6.99,+1.00%, -1,[.] _IO_new_file_xsputn 1.89,-2.00%, +4,[.] __readdir64 $ Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.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> LKML-Reference: <1260978567-550-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-16 23:49:27 +08:00
unsigned long position = 1;
long displacement = 0;
unsigned int width;
perf diff: Use perf_session__fprintf_hists just like 'perf record' That means that almost everything you can do with 'perf report' can be done with 'perf diff', for instance: $ 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) ] $ 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 (~2687 samples) ] perf diff | head -8 9.02% +1.00% find libc-2.10.1.so [.] _IO_vfprintf_internal 2.91% -1.00% find [kernel] [k] __kmalloc 2.85% -1.00% find [kernel] [k] ext4_htree_store_dirent 1.99% -1.00% find [kernel] [k] _atomic_dec_and_lock 2.44% find [kernel] [k] half_md4_transform $ So if you want to zoom into libc: $ perf diff --dsos libc-2.10.1.so | head -8 37.34% find [.] _IO_vfprintf_internal 10.34% find [.] __GI_memmove 8.25% +2.00% find [.] _int_malloc 5.07% -1.00% find [.] __GI_mempcpy 7.62% +2.00% find [.] _int_free $ And if there were multiple commands using libc, it is also possible to aggregate them all by using --sort symbol: $ perf diff --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% [.] __GI_mempcpy 7.62% +2.00% [.] _int_free $ The displacement column now is off by default, to use it: perf diff -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% +2 [.] __GI_mempcpy 7.62% +2.00% -1 [.] _int_free $ Using -t/--field-separator can be used for scripting: $ perf diff -t, -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34, , ,[.] _IO_vfprintf_internal 10.34, , ,[.] __GI_memmove 8.25,+2.00%, ,[.] _int_malloc 5.07,-1.00%, +2,[.] __GI_mempcpy 7.62,+2.00%, -1,[.] _int_free 6.99,+1.00%, -1,[.] _IO_new_file_xsputn 1.89,-2.00%, +4,[.] __readdir64 $ Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.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> LKML-Reference: <1260978567-550-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-16 23:49:27 +08:00
const char *sep = symbol_conf.field_sep;
const char *col_width = symbol_conf.col_width_list_str;
int nr_rows = 0;
init_rem_hits();
if (!show_header)
goto print_entries;
perf diff: Use perf_session__fprintf_hists just like 'perf record' That means that almost everything you can do with 'perf report' can be done with 'perf diff', for instance: $ 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) ] $ 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 (~2687 samples) ] perf diff | head -8 9.02% +1.00% find libc-2.10.1.so [.] _IO_vfprintf_internal 2.91% -1.00% find [kernel] [k] __kmalloc 2.85% -1.00% find [kernel] [k] ext4_htree_store_dirent 1.99% -1.00% find [kernel] [k] _atomic_dec_and_lock 2.44% find [kernel] [k] half_md4_transform $ So if you want to zoom into libc: $ perf diff --dsos libc-2.10.1.so | head -8 37.34% find [.] _IO_vfprintf_internal 10.34% find [.] __GI_memmove 8.25% +2.00% find [.] _int_malloc 5.07% -1.00% find [.] __GI_mempcpy 7.62% +2.00% find [.] _int_free $ And if there were multiple commands using libc, it is also possible to aggregate them all by using --sort symbol: $ perf diff --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% [.] __GI_mempcpy 7.62% +2.00% [.] _int_free $ The displacement column now is off by default, to use it: perf diff -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% +2 [.] __GI_mempcpy 7.62% +2.00% -1 [.] _int_free $ Using -t/--field-separator can be used for scripting: $ perf diff -t, -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34, , ,[.] _IO_vfprintf_internal 10.34, , ,[.] __GI_memmove 8.25,+2.00%, ,[.] _int_malloc 5.07,-1.00%, +2,[.] __GI_mempcpy 7.62,+2.00%, -1,[.] _int_free 6.99,+1.00%, -1,[.] _IO_new_file_xsputn 1.89,-2.00%, +4,[.] __readdir64 $ Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.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> LKML-Reference: <1260978567-550-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-16 23:49:27 +08:00
fprintf(fp, "# %s", pair ? "Baseline" : "Overhead");
if (symbol_conf.show_cpu_utilization) {
if (sep) {
ret += fprintf(fp, "%csys", *sep);
ret += fprintf(fp, "%cus", *sep);
if (perf_guest) {
ret += fprintf(fp, "%cguest sys", *sep);
ret += fprintf(fp, "%cguest us", *sep);
}
} else {
ret += fprintf(fp, " sys ");
ret += fprintf(fp, " us ");
if (perf_guest) {
ret += fprintf(fp, " guest sys ");
ret += fprintf(fp, " guest us ");
}
}
}
if (symbol_conf.show_nr_samples) {
if (sep)
fprintf(fp, "%cSamples", *sep);
else
fputs(" Samples ", fp);
}
if (symbol_conf.show_total_period) {
if (sep)
ret += fprintf(fp, "%cPeriod", *sep);
else
ret += fprintf(fp, " Period ");
}
perf diff: Use perf_session__fprintf_hists just like 'perf record' That means that almost everything you can do with 'perf report' can be done with 'perf diff', for instance: $ 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) ] $ 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 (~2687 samples) ] perf diff | head -8 9.02% +1.00% find libc-2.10.1.so [.] _IO_vfprintf_internal 2.91% -1.00% find [kernel] [k] __kmalloc 2.85% -1.00% find [kernel] [k] ext4_htree_store_dirent 1.99% -1.00% find [kernel] [k] _atomic_dec_and_lock 2.44% find [kernel] [k] half_md4_transform $ So if you want to zoom into libc: $ perf diff --dsos libc-2.10.1.so | head -8 37.34% find [.] _IO_vfprintf_internal 10.34% find [.] __GI_memmove 8.25% +2.00% find [.] _int_malloc 5.07% -1.00% find [.] __GI_mempcpy 7.62% +2.00% find [.] _int_free $ And if there were multiple commands using libc, it is also possible to aggregate them all by using --sort symbol: $ perf diff --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% [.] __GI_mempcpy 7.62% +2.00% [.] _int_free $ The displacement column now is off by default, to use it: perf diff -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% +2 [.] __GI_mempcpy 7.62% +2.00% -1 [.] _int_free $ Using -t/--field-separator can be used for scripting: $ perf diff -t, -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34, , ,[.] _IO_vfprintf_internal 10.34, , ,[.] __GI_memmove 8.25,+2.00%, ,[.] _int_malloc 5.07,-1.00%, +2,[.] __GI_mempcpy 7.62,+2.00%, -1,[.] _int_free 6.99,+1.00%, -1,[.] _IO_new_file_xsputn 1.89,-2.00%, +4,[.] __readdir64 $ Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.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> LKML-Reference: <1260978567-550-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-16 23:49:27 +08:00
if (pair) {
if (sep)
ret += fprintf(fp, "%cDelta", *sep);
else
ret += fprintf(fp, " Delta ");
if (show_displacement) {
if (sep)
ret += fprintf(fp, "%cDisplacement", *sep);
else
ret += fprintf(fp, " Displ");
}
}
list_for_each_entry(se, &hist_entry__sort_list, list) {
if (se->elide)
continue;
perf diff: Use perf_session__fprintf_hists just like 'perf record' That means that almost everything you can do with 'perf report' can be done with 'perf diff', for instance: $ 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) ] $ 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 (~2687 samples) ] perf diff | head -8 9.02% +1.00% find libc-2.10.1.so [.] _IO_vfprintf_internal 2.91% -1.00% find [kernel] [k] __kmalloc 2.85% -1.00% find [kernel] [k] ext4_htree_store_dirent 1.99% -1.00% find [kernel] [k] _atomic_dec_and_lock 2.44% find [kernel] [k] half_md4_transform $ So if you want to zoom into libc: $ perf diff --dsos libc-2.10.1.so | head -8 37.34% find [.] _IO_vfprintf_internal 10.34% find [.] __GI_memmove 8.25% +2.00% find [.] _int_malloc 5.07% -1.00% find [.] __GI_mempcpy 7.62% +2.00% find [.] _int_free $ And if there were multiple commands using libc, it is also possible to aggregate them all by using --sort symbol: $ perf diff --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% [.] __GI_mempcpy 7.62% +2.00% [.] _int_free $ The displacement column now is off by default, to use it: perf diff -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% +2 [.] __GI_mempcpy 7.62% +2.00% -1 [.] _int_free $ Using -t/--field-separator can be used for scripting: $ perf diff -t, -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34, , ,[.] _IO_vfprintf_internal 10.34, , ,[.] __GI_memmove 8.25,+2.00%, ,[.] _int_malloc 5.07,-1.00%, +2,[.] __GI_mempcpy 7.62,+2.00%, -1,[.] _int_free 6.99,+1.00%, -1,[.] _IO_new_file_xsputn 1.89,-2.00%, +4,[.] __readdir64 $ Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.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> LKML-Reference: <1260978567-550-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-16 23:49:27 +08:00
if (sep) {
fprintf(fp, "%c%s", *sep, se->se_header);
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);
fprintf(fp, " %*s", width, se->se_header);
}
fprintf(fp, "\n");
if (max_rows && ++nr_rows >= max_rows)
goto out;
perf diff: Use perf_session__fprintf_hists just like 'perf record' That means that almost everything you can do with 'perf report' can be done with 'perf diff', for instance: $ 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) ] $ 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 (~2687 samples) ] perf diff | head -8 9.02% +1.00% find libc-2.10.1.so [.] _IO_vfprintf_internal 2.91% -1.00% find [kernel] [k] __kmalloc 2.85% -1.00% find [kernel] [k] ext4_htree_store_dirent 1.99% -1.00% find [kernel] [k] _atomic_dec_and_lock 2.44% find [kernel] [k] half_md4_transform $ So if you want to zoom into libc: $ perf diff --dsos libc-2.10.1.so | head -8 37.34% find [.] _IO_vfprintf_internal 10.34% find [.] __GI_memmove 8.25% +2.00% find [.] _int_malloc 5.07% -1.00% find [.] __GI_mempcpy 7.62% +2.00% find [.] _int_free $ And if there were multiple commands using libc, it is also possible to aggregate them all by using --sort symbol: $ perf diff --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% [.] __GI_mempcpy 7.62% +2.00% [.] _int_free $ The displacement column now is off by default, to use it: perf diff -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% +2 [.] __GI_mempcpy 7.62% +2.00% -1 [.] _int_free $ Using -t/--field-separator can be used for scripting: $ perf diff -t, -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34, , ,[.] _IO_vfprintf_internal 10.34, , ,[.] __GI_memmove 8.25,+2.00%, ,[.] _int_malloc 5.07,-1.00%, +2,[.] __GI_mempcpy 7.62,+2.00%, -1,[.] _int_free 6.99,+1.00%, -1,[.] _IO_new_file_xsputn 1.89,-2.00%, +4,[.] __readdir64 $ Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.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> LKML-Reference: <1260978567-550-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-16 23:49:27 +08:00
if (sep)
goto print_entries;
fprintf(fp, "# ........");
if (symbol_conf.show_cpu_utilization)
fprintf(fp, " ....... .......");
if (symbol_conf.show_nr_samples)
fprintf(fp, " ..........");
if (symbol_conf.show_total_period)
fprintf(fp, " ............");
perf diff: Use perf_session__fprintf_hists just like 'perf record' That means that almost everything you can do with 'perf report' can be done with 'perf diff', for instance: $ 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) ] $ 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 (~2687 samples) ] perf diff | head -8 9.02% +1.00% find libc-2.10.1.so [.] _IO_vfprintf_internal 2.91% -1.00% find [kernel] [k] __kmalloc 2.85% -1.00% find [kernel] [k] ext4_htree_store_dirent 1.99% -1.00% find [kernel] [k] _atomic_dec_and_lock 2.44% find [kernel] [k] half_md4_transform $ So if you want to zoom into libc: $ perf diff --dsos libc-2.10.1.so | head -8 37.34% find [.] _IO_vfprintf_internal 10.34% find [.] __GI_memmove 8.25% +2.00% find [.] _int_malloc 5.07% -1.00% find [.] __GI_mempcpy 7.62% +2.00% find [.] _int_free $ And if there were multiple commands using libc, it is also possible to aggregate them all by using --sort symbol: $ perf diff --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% [.] __GI_mempcpy 7.62% +2.00% [.] _int_free $ The displacement column now is off by default, to use it: perf diff -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% +2 [.] __GI_mempcpy 7.62% +2.00% -1 [.] _int_free $ Using -t/--field-separator can be used for scripting: $ perf diff -t, -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34, , ,[.] _IO_vfprintf_internal 10.34, , ,[.] __GI_memmove 8.25,+2.00%, ,[.] _int_malloc 5.07,-1.00%, +2,[.] __GI_mempcpy 7.62,+2.00%, -1,[.] _int_free 6.99,+1.00%, -1,[.] _IO_new_file_xsputn 1.89,-2.00%, +4,[.] __readdir64 $ Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.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> LKML-Reference: <1260978567-550-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-16 23:49:27 +08:00
if (pair) {
fprintf(fp, " ..........");
if (show_displacement)
fprintf(fp, " .....");
}
list_for_each_entry(se, &hist_entry__sort_list, list) {
unsigned int i;
if (se->elide)
continue;
fprintf(fp, " ");
width = hists__col_len(hists, se->se_width_idx);
if (width == 0)
width = strlen(se->se_header);
for (i = 0; i < width; i++)
fprintf(fp, ".");
}
fprintf(fp, "\n");
if (max_rows && ++nr_rows >= max_rows)
goto out;
fprintf(fp, "#\n");
if (max_rows && ++nr_rows >= max_rows)
goto out;
print_entries:
total_period = hists->stats.total_period;
for (nd = rb_first(&hists->entries); nd; nd = rb_next(nd)) {
perf diff: Use perf_session__fprintf_hists just like 'perf record' That means that almost everything you can do with 'perf report' can be done with 'perf diff', for instance: $ 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) ] $ 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 (~2687 samples) ] perf diff | head -8 9.02% +1.00% find libc-2.10.1.so [.] _IO_vfprintf_internal 2.91% -1.00% find [kernel] [k] __kmalloc 2.85% -1.00% find [kernel] [k] ext4_htree_store_dirent 1.99% -1.00% find [kernel] [k] _atomic_dec_and_lock 2.44% find [kernel] [k] half_md4_transform $ So if you want to zoom into libc: $ perf diff --dsos libc-2.10.1.so | head -8 37.34% find [.] _IO_vfprintf_internal 10.34% find [.] __GI_memmove 8.25% +2.00% find [.] _int_malloc 5.07% -1.00% find [.] __GI_mempcpy 7.62% +2.00% find [.] _int_free $ And if there were multiple commands using libc, it is also possible to aggregate them all by using --sort symbol: $ perf diff --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% [.] __GI_mempcpy 7.62% +2.00% [.] _int_free $ The displacement column now is off by default, to use it: perf diff -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% +2 [.] __GI_mempcpy 7.62% +2.00% -1 [.] _int_free $ Using -t/--field-separator can be used for scripting: $ perf diff -t, -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34, , ,[.] _IO_vfprintf_internal 10.34, , ,[.] __GI_memmove 8.25,+2.00%, ,[.] _int_malloc 5.07,-1.00%, +2,[.] __GI_mempcpy 7.62,+2.00%, -1,[.] _int_free 6.99,+1.00%, -1,[.] _IO_new_file_xsputn 1.89,-2.00%, +4,[.] __readdir64 $ Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.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> LKML-Reference: <1260978567-550-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-16 23:49:27 +08:00
struct hist_entry *h = rb_entry(nd, struct hist_entry, rb_node);
if (h->filtered)
continue;
perf diff: Use perf_session__fprintf_hists just like 'perf record' That means that almost everything you can do with 'perf report' can be done with 'perf diff', for instance: $ 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) ] $ 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 (~2687 samples) ] perf diff | head -8 9.02% +1.00% find libc-2.10.1.so [.] _IO_vfprintf_internal 2.91% -1.00% find [kernel] [k] __kmalloc 2.85% -1.00% find [kernel] [k] ext4_htree_store_dirent 1.99% -1.00% find [kernel] [k] _atomic_dec_and_lock 2.44% find [kernel] [k] half_md4_transform $ So if you want to zoom into libc: $ perf diff --dsos libc-2.10.1.so | head -8 37.34% find [.] _IO_vfprintf_internal 10.34% find [.] __GI_memmove 8.25% +2.00% find [.] _int_malloc 5.07% -1.00% find [.] __GI_mempcpy 7.62% +2.00% find [.] _int_free $ And if there were multiple commands using libc, it is also possible to aggregate them all by using --sort symbol: $ perf diff --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% [.] __GI_mempcpy 7.62% +2.00% [.] _int_free $ The displacement column now is off by default, to use it: perf diff -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34% [.] _IO_vfprintf_internal 10.34% [.] __GI_memmove 8.25% +2.00% [.] _int_malloc 5.07% -1.00% +2 [.] __GI_mempcpy 7.62% +2.00% -1 [.] _int_free $ Using -t/--field-separator can be used for scripting: $ perf diff -t, -m --dsos libc-2.10.1.so --sort symbol | head -8 37.34, , ,[.] _IO_vfprintf_internal 10.34, , ,[.] __GI_memmove 8.25,+2.00%, ,[.] _int_malloc 5.07,-1.00%, +2,[.] __GI_mempcpy 7.62,+2.00%, -1,[.] _int_free 6.99,+1.00%, -1,[.] _IO_new_file_xsputn 1.89,-2.00%, +4,[.] __readdir64 $ Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.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> LKML-Reference: <1260978567-550-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-16 23:49:27 +08:00
if (show_displacement) {
if (h->pair != NULL)
displacement = ((long)h->pair->position -
(long)position);
else
displacement = 0;
++position;
}
ret += hist_entry__fprintf(h, max_cols, hists, pair, show_displacement,
displacement, total_period, fp);
if (symbol_conf.use_callchain)
ret += hist_entry__fprintf_callchain(h, hists, total_period, fp);
if (max_rows && ++nr_rows >= max_rows)
goto out;
if (h->ms.map == NULL && verbose > 1) {
__map_groups__fprintf_maps(&h->thread->mg,
MAP__FUNCTION, verbose, fp);
fprintf(fp, "%.10s end\n", graph_dotted_line);
}
}
out:
free(rem_sq_bracket);
return ret;
}
/*
* See hists__fprintf to match the column widths
*/
unsigned int hists__sort_list_width(struct hists *hists)
{
struct sort_entry *se;
int ret = 9; /* total % */
if (symbol_conf.show_cpu_utilization) {
ret += 7; /* count_sys % */
ret += 6; /* count_us % */
if (perf_guest) {
ret += 13; /* count_guest_sys % */
ret += 12; /* count_guest_us % */
}
}
if (symbol_conf.show_nr_samples)
ret += 11;
if (symbol_conf.show_total_period)
ret += 13;
list_for_each_entry(se, &hist_entry__sort_list, list)
if (!se->elide)
ret += 2 + hists__col_len(hists, se->se_width_idx);
if (verbose) /* Addr + origin */
ret += 3 + BITS_PER_LONG / 4;
return ret;
}
static void hists__remove_entry_filter(struct hists *hists, struct hist_entry *h,
enum hist_filter filter)
{
h->filtered &= ~(1 << filter);
if (h->filtered)
return;
++hists->nr_entries;
if (h->ms.unfolded)
hists->nr_entries += h->nr_rows;
h->row_offset = 0;
hists->stats.total_period += h->period;
hists->stats.nr_events[PERF_RECORD_SAMPLE] += h->nr_events;
hists__calc_col_len(hists, h);
}
static bool hists__filter_entry_by_dso(struct hists *hists,
struct hist_entry *he)
{
if (hists->dso_filter != NULL &&
(he->ms.map == NULL || he->ms.map->dso != hists->dso_filter)) {
he->filtered |= (1 << HIST_FILTER__DSO);
return true;
}
return false;
}
void hists__filter_by_dso(struct hists *hists)
{
struct rb_node *nd;
hists->nr_entries = hists->stats.total_period = 0;
hists->stats.nr_events[PERF_RECORD_SAMPLE] = 0;
hists__reset_col_len(hists);
for (nd = rb_first(&hists->entries); nd; nd = rb_next(nd)) {
struct hist_entry *h = rb_entry(nd, struct hist_entry, rb_node);
if (symbol_conf.exclude_other && !h->parent)
continue;
if (hists__filter_entry_by_dso(hists, h))
continue;
hists__remove_entry_filter(hists, h, HIST_FILTER__DSO);
}
}
static bool hists__filter_entry_by_thread(struct hists *hists,
struct hist_entry *he)
{
if (hists->thread_filter != NULL &&
he->thread != hists->thread_filter) {
he->filtered |= (1 << HIST_FILTER__THREAD);
return true;
}
return false;
}
void hists__filter_by_thread(struct hists *hists)
{
struct rb_node *nd;
hists->nr_entries = hists->stats.total_period = 0;
hists->stats.nr_events[PERF_RECORD_SAMPLE] = 0;
hists__reset_col_len(hists);
for (nd = rb_first(&hists->entries); nd; nd = rb_next(nd)) {
struct hist_entry *h = rb_entry(nd, struct hist_entry, rb_node);
if (hists__filter_entry_by_thread(hists, h))
continue;
hists__remove_entry_filter(hists, h, HIST_FILTER__THREAD);
}
}
int hist_entry__inc_addr_samples(struct hist_entry *he, int evidx, u64 ip)
{
return symbol__inc_addr_samples(he->ms.sym, he->ms.map, evidx, ip);
}
int hist_entry__annotate(struct hist_entry *he, size_t privsize)
{
return symbol__annotate(he->ms.sym, he->ms.map, privsize);
}
void hists__inc_nr_events(struct hists *hists, u32 type)
{
++hists->stats.nr_events[0];
++hists->stats.nr_events[type];
}
size_t hists__fprintf_nr_events(struct hists *hists, FILE *fp)
{
int i;
size_t ret = 0;
for (i = 0; i < PERF_RECORD_HEADER_MAX; ++i) {
const char *name;
if (hists->stats.nr_events[i] == 0)
continue;
name = perf_event__name(i);
if (!strcmp(name, "UNKNOWN"))
continue;
ret += fprintf(fp, "%16s events: %10d\n", name,
hists->stats.nr_events[i]);
}
return ret;
}