linux/tools/perf/util/hist.c

1645 lines
38 KiB
C
Raw Normal View History

#include "util.h"
#include "build-id.h"
#include "hist.h"
#include "session.h"
#include "sort.h"
#include "evlist.h"
#include "evsel.h"
#include "annotate.h"
#include "ui/progress.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>
2009-12-16 16:31:49 +00:00
#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);
static bool hists__filter_entry_by_symbol(struct hists *hists,
struct hist_entry *he);
static bool hists__filter_entry_by_socket(struct hists *hists,
struct hist_entry *he);
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;
}
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__set_unres_dso_col_len(struct hists *hists, int dso)
{
const unsigned int unresolved_col_width = BITS_PER_LONG / 4;
if (hists__col_len(hists, dso) < unresolved_col_width &&
!symbol_conf.col_width_list_str && !symbol_conf.field_sep &&
!symbol_conf.dso_list)
hists__set_col_len(hists, dso, unresolved_col_width);
}
void hists__calc_col_len(struct hists *hists, struct hist_entry *h)
{
const unsigned int unresolved_col_width = BITS_PER_LONG / 4;
int symlen;
u16 len;
/*
* +4 accounts for '[x] ' priv level info
* +2 accounts for 0x prefix on raw addresses
* +3 accounts for ' y ' symtab origin info
*/
if (h->ms.sym) {
symlen = h->ms.sym->namelen + 4;
if (verbose)
symlen += BITS_PER_LONG / 4 + 2 + 3;
hists__new_col_len(hists, HISTC_SYMBOL, symlen);
} else {
symlen = unresolved_col_width + 4 + 2;
hists__new_col_len(hists, HISTC_SYMBOL, symlen);
hists__set_unres_dso_col_len(hists, HISTC_DSO);
}
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);
}
if (h->parent)
hists__new_col_len(hists, HISTC_PARENT, h->parent->namelen);
if (h->branch_info) {
if (h->branch_info->from.sym) {
symlen = (int)h->branch_info->from.sym->namelen + 4;
if (verbose)
symlen += BITS_PER_LONG / 4 + 2 + 3;
hists__new_col_len(hists, HISTC_SYMBOL_FROM, symlen);
symlen = dso__name_len(h->branch_info->from.map->dso);
hists__new_col_len(hists, HISTC_DSO_FROM, symlen);
} else {
symlen = unresolved_col_width + 4 + 2;
hists__new_col_len(hists, HISTC_SYMBOL_FROM, symlen);
hists__set_unres_dso_col_len(hists, HISTC_DSO_FROM);
}
if (h->branch_info->to.sym) {
symlen = (int)h->branch_info->to.sym->namelen + 4;
if (verbose)
symlen += BITS_PER_LONG / 4 + 2 + 3;
hists__new_col_len(hists, HISTC_SYMBOL_TO, symlen);
symlen = dso__name_len(h->branch_info->to.map->dso);
hists__new_col_len(hists, HISTC_DSO_TO, symlen);
} else {
symlen = unresolved_col_width + 4 + 2;
hists__new_col_len(hists, HISTC_SYMBOL_TO, symlen);
hists__set_unres_dso_col_len(hists, HISTC_DSO_TO);
}
}
if (h->mem_info) {
if (h->mem_info->daddr.sym) {
symlen = (int)h->mem_info->daddr.sym->namelen + 4
+ unresolved_col_width + 2;
hists__new_col_len(hists, HISTC_MEM_DADDR_SYMBOL,
symlen);
perf tools: Add dcacheline sort In perf's 'mem-mode', one can get access to a whole bunch of details specific to a particular sample instruction. A bunch of those details relate to the data address. One interesting thing you can do with data addresses is to convert them into a unique cacheline they belong too. Organizing these data cachelines into similar groups and sorting them can reveal cache contention. This patch creates an alogorithm based on various sample details that can help group entries together into data cachelines and allows 'perf report' to sort on it. The algorithm relies on having proper mmap2 support in the kernel to help determine if the memory map the data address belongs to is private to a pid or globally shared. The alogortithm is as follows: o group cpumodes together o group entries with discovered maps together o sort on major, minor, inode and inode generation numbers o if userspace anon, then sort on pid o sort on cachelines based on data addresses The 'dcacheline' sort option in 'perf report' only works in 'mem-mode'. Sample output: # # Samples: 206 of event 'cpu/mem-loads/pp' # Total weight : 2534 # Sort order : dcacheline,pid # # Overhead Samples Data Cacheline Command: Pid # ........ ............ ...................................................................... .................. # 13.22% 1 [k] 0xffff88042f08ebc0 swapper: 0 9.27% 1 [k] 0xffff88082e8cea80 swapper: 0 3.59% 2 [k] 0xffffffff819ba180 swapper: 0 0.32% 1 [k] arch_trigger_all_cpu_backtrace_handler_na.23901+0xffffffffffffffe0 swapper: 0 0.32% 1 [k] timekeeper_seq+0xfffffffffffffff8 swapper: 0 Note: Added a '+1' to symlen size in hists__calc_col_len to prevent the next column from prematurely tabbing over and mis-aligning. Not sure what the problem is. Signed-off-by: Don Zickus <dzickus@redhat.com> Link: http://lkml.kernel.org/r/1401208087-181977-8-git-send-email-dzickus@redhat.com Signed-off-by: Jiri Olsa <jolsa@kernel.org>
2014-06-01 13:38:29 +00:00
hists__new_col_len(hists, HISTC_MEM_DCACHELINE,
symlen + 1);
} else {
symlen = unresolved_col_width + 4 + 2;
hists__new_col_len(hists, HISTC_MEM_DADDR_SYMBOL,
symlen);
}
if (h->mem_info->iaddr.sym) {
symlen = (int)h->mem_info->iaddr.sym->namelen + 4
+ unresolved_col_width + 2;
hists__new_col_len(hists, HISTC_MEM_IADDR_SYMBOL,
symlen);
} else {
symlen = unresolved_col_width + 4 + 2;
hists__new_col_len(hists, HISTC_MEM_IADDR_SYMBOL,
symlen);
}
if (h->mem_info->daddr.map) {
symlen = dso__name_len(h->mem_info->daddr.map->dso);
hists__new_col_len(hists, HISTC_MEM_DADDR_DSO,
symlen);
} else {
symlen = unresolved_col_width + 4 + 2;
hists__set_unres_dso_col_len(hists, HISTC_MEM_DADDR_DSO);
}
} else {
symlen = unresolved_col_width + 4 + 2;
hists__new_col_len(hists, HISTC_MEM_DADDR_SYMBOL, symlen);
hists__new_col_len(hists, HISTC_MEM_IADDR_SYMBOL, symlen);
hists__set_unres_dso_col_len(hists, HISTC_MEM_DADDR_DSO);
}
hists__new_col_len(hists, HISTC_CPU, 3);
hists__new_col_len(hists, HISTC_SOCKET, 6);
hists__new_col_len(hists, HISTC_MEM_LOCKED, 6);
hists__new_col_len(hists, HISTC_MEM_TLB, 22);
hists__new_col_len(hists, HISTC_MEM_SNOOP, 12);
hists__new_col_len(hists, HISTC_MEM_LVL, 21 + 3);
hists__new_col_len(hists, HISTC_LOCAL_WEIGHT, 12);
hists__new_col_len(hists, HISTC_GLOBAL_WEIGHT, 12);
if (h->srcline)
hists__new_col_len(hists, HISTC_SRCLINE, strlen(h->srcline));
perf report: Add support for srcfile sort key In some cases it's useful to characterize samples by file. This is useful to get a higher level categorization, for example to map cost to subsystems. Add a srcfile sort key to perf report. It builds on top of the existing srcline support. Commiter notes: E.g.: # perf record -F 10000 usleep 1 [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.016 MB perf.data (13 samples) ] [root@zoo ~]# perf report -s srcfile --stdio # Total Lost Samples: 0 # # Samples: 13 of event 'cycles' # Event count (approx.): 869878 # # Overhead Source File # ........ ........... 60.99% . 20.62% paravirt.h 14.23% rmap.c 4.04% signal.c 0.11% msr.h # The first line is collecting all the files for which srcfiles couldn't somehow get resolved to: # perf report -s srcfile,dso --stdio # Total Lost Samples: 0 # # Samples: 13 of event 'cycles' # Event count (approx.): 869878 # # Overhead Source File Shared Object # ........ ........... ................ 40.97% . ld-2.20.so 20.62% paravirt.h [kernel.vmlinux] 20.02% . libc-2.20.so 14.23% rmap.c [kernel.vmlinux] 4.04% signal.c [kernel.vmlinux] 0.11% msr.h [kernel.vmlinux] # XXX: Investigate why that is not resolving on Fedora 21, Andi says he hasn't seen this on Fedora 22. Signed-off-by: Andi Kleen <ak@linux.intel.com> Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Link: http://lkml.kernel.org/r/1438988064-21834-1-git-send-email-andi@firstfloor.org [ Added column length update, from 0e65bdb3f90f ('perf hists: Update the column width for the "srcline" sort key') ] Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-08-07 22:54:24 +00:00
if (h->srcfile)
hists__new_col_len(hists, HISTC_SRCFILE, strlen(h->srcfile));
if (h->transaction)
hists__new_col_len(hists, HISTC_TRANSACTION,
hist_entry__transaction_len());
}
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);
next = rb_next(&n->rb_node);
}
}
static void he_stat__add_cpumode_period(struct he_stat *he_stat,
unsigned int cpumode, u64 period)
{
switch (cpumode) {
case PERF_RECORD_MISC_KERNEL:
he_stat->period_sys += period;
break;
case PERF_RECORD_MISC_USER:
he_stat->period_us += period;
break;
case PERF_RECORD_MISC_GUEST_KERNEL:
he_stat->period_guest_sys += period;
break;
case PERF_RECORD_MISC_GUEST_USER:
he_stat->period_guest_us += period;
break;
default:
break;
}
}
static void he_stat__add_period(struct he_stat *he_stat, u64 period,
u64 weight)
{
he_stat->period += period;
he_stat->weight += weight;
he_stat->nr_events += 1;
}
static void he_stat__add_stat(struct he_stat *dest, struct he_stat *src)
{
dest->period += src->period;
dest->period_sys += src->period_sys;
dest->period_us += src->period_us;
dest->period_guest_sys += src->period_guest_sys;
dest->period_guest_us += src->period_guest_us;
dest->nr_events += src->nr_events;
dest->weight += src->weight;
}
static void he_stat__decay(struct he_stat *he_stat)
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>
2011-10-05 22:16:15 +00:00
{
he_stat->period = (he_stat->period * 7) / 8;
he_stat->nr_events = (he_stat->nr_events * 7) / 8;
/* XXX need decay for weight too? */
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>
2011-10-05 22:16:15 +00:00
}
static bool hists__decay_entry(struct hists *hists, struct hist_entry *he)
{
u64 prev_period = he->stat.period;
u64 diff;
if (prev_period == 0)
return true;
he_stat__decay(&he->stat);
if (symbol_conf.cumulate_callchain)
he_stat__decay(he->stat_acc);
decay_callchain(he->callchain);
diff = prev_period - he->stat.period;
hists->stats.total_period -= diff;
if (!he->filtered)
hists->stats.total_non_filtered_period -= diff;
return he->stat.period == 0;
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>
2011-10-05 22:16:15 +00:00
}
static void hists__delete_entry(struct hists *hists, struct hist_entry *he)
{
rb_erase(&he->rb_node, &hists->entries);
if (sort__need_collapse)
rb_erase(&he->rb_node_in, &hists->entries_collapsed);
else
rb_erase(&he->rb_node_in, hists->entries_in);
--hists->nr_entries;
if (!he->filtered)
--hists->nr_non_filtered_entries;
hist_entry__delete(he);
}
void hists__decay_entries(struct hists *hists, bool zap_user, bool zap_kernel)
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>
2011-10-05 22:16:15 +00:00
{
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);
if (((zap_user && n->level == '.') ||
(zap_kernel && n->level != '.') ||
hists__decay_entry(hists, n))) {
hists__delete_entry(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>
2011-10-05 22:16:15 +00:00
}
}
}
void hists__delete_entries(struct hists *hists)
{
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);
hists__delete_entry(hists, n);
}
}
/*
* histogram, sorted on item, collects periods
*/
static struct hist_entry *hist_entry__new(struct hist_entry *template,
bool sample_self)
{
size_t callchain_size = 0;
struct hist_entry *he;
if (symbol_conf.use_callchain)
callchain_size = sizeof(struct callchain_root);
he = zalloc(sizeof(*he) + callchain_size);
if (he != NULL) {
*he = *template;
if (symbol_conf.cumulate_callchain) {
he->stat_acc = malloc(sizeof(he->stat));
if (he->stat_acc == NULL) {
free(he);
return NULL;
}
memcpy(he->stat_acc, &he->stat, sizeof(he->stat));
if (!sample_self)
memset(&he->stat, 0, sizeof(he->stat));
}
map__get(he->ms.map);
if (he->branch_info) {
/*
* This branch info is (a part of) allocated from
* sample__resolve_bstack() and will be freed after
* adding new entries. So we need to save a copy.
*/
he->branch_info = malloc(sizeof(*he->branch_info));
if (he->branch_info == NULL) {
map__zput(he->ms.map);
free(he->stat_acc);
free(he);
return NULL;
}
memcpy(he->branch_info, template->branch_info,
sizeof(*he->branch_info));
map__get(he->branch_info->from.map);
map__get(he->branch_info->to.map);
}
if (he->mem_info) {
map__get(he->mem_info->iaddr.map);
map__get(he->mem_info->daddr.map);
}
if (symbol_conf.use_callchain)
callchain_init(he->callchain);
if (he->raw_data) {
he->raw_data = memdup(he->raw_data, he->raw_size);
if (he->raw_data == NULL) {
map__put(he->ms.map);
if (he->branch_info) {
map__put(he->branch_info->from.map);
map__put(he->branch_info->to.map);
free(he->branch_info);
}
if (he->mem_info) {
map__put(he->mem_info->iaddr.map);
map__put(he->mem_info->daddr.map);
}
free(he->stat_acc);
free(he);
return NULL;
}
}
INIT_LIST_HEAD(&he->pairs.node);
thread__get(he->thread);
}
return he;
}
static u8 symbol__parent_filter(const struct symbol *parent)
{
if (symbol_conf.exclude_other && parent == NULL)
return 1 << HIST_FILTER__PARENT;
return 0;
}
static struct hist_entry *hists__findnew_entry(struct hists *hists,
struct hist_entry *entry,
struct addr_location *al,
bool sample_self)
{
struct rb_node **p;
struct rb_node *parent = NULL;
struct hist_entry *he;
int64_t cmp;
u64 period = entry->stat.period;
u64 weight = entry->stat.weight;
p = &hists->entries_in->rb_node;
while (*p != NULL) {
parent = *p;
he = rb_entry(parent, struct hist_entry, rb_node_in);
/*
* Make sure that it receives arguments in a same order as
* hist_entry__collapse() so that we can use an appropriate
* function when searching an entry regardless which sort
* keys were used.
*/
cmp = hist_entry__cmp(he, entry);
if (!cmp) {
if (sample_self)
he_stat__add_period(&he->stat, period, weight);
if (symbol_conf.cumulate_callchain)
he_stat__add_period(he->stat_acc, period, weight);
/*
* This mem info was allocated from sample__resolve_mem
* and will not be used anymore.
*/
zfree(&entry->mem_info);
/* If the map of an existing hist_entry has
* become out-of-date due to an exec() or
* similar, update it. Otherwise we will
* mis-adjust symbol addresses when computing
* the history counter to increment.
*/
if (he->ms.map != entry->ms.map) {
map__put(he->ms.map);
he->ms.map = map__get(entry->ms.map);
}
goto out;
}
if (cmp < 0)
p = &(*p)->rb_left;
else
p = &(*p)->rb_right;
}
he = hist_entry__new(entry, sample_self);
if (!he)
return NULL;
hists->nr_entries++;
rb_link_node(&he->rb_node_in, parent, p);
rb_insert_color(&he->rb_node_in, hists->entries_in);
out:
if (sample_self)
he_stat__add_cpumode_period(&he->stat, al->cpumode, period);
if (symbol_conf.cumulate_callchain)
he_stat__add_cpumode_period(he->stat_acc, al->cpumode, period);
return he;
}
struct hist_entry *__hists__add_entry(struct hists *hists,
struct addr_location *al,
struct symbol *sym_parent,
struct branch_info *bi,
struct mem_info *mi,
struct perf_sample *sample,
bool sample_self)
{
struct hist_entry entry = {
.thread = al->thread,
.comm = thread__comm(al->thread),
.ms = {
.map = al->map,
.sym = al->sym,
},
.socket = al->socket,
.cpu = al->cpu,
.cpumode = al->cpumode,
.ip = al->addr,
.level = al->level,
.stat = {
.nr_events = 1,
.period = sample->period,
.weight = sample->weight,
},
.parent = sym_parent,
.filtered = symbol__parent_filter(sym_parent) | al->filtered,
.hists = hists,
.branch_info = bi,
.mem_info = mi,
.transaction = sample->transaction,
.raw_data = sample->raw_data,
.raw_size = sample->raw_size,
};
return hists__findnew_entry(hists, &entry, al, sample_self);
}
static int
iter_next_nop_entry(struct hist_entry_iter *iter __maybe_unused,
struct addr_location *al __maybe_unused)
{
return 0;
}
static int
iter_add_next_nop_entry(struct hist_entry_iter *iter __maybe_unused,
struct addr_location *al __maybe_unused)
{
return 0;
}
static int
iter_prepare_mem_entry(struct hist_entry_iter *iter, struct addr_location *al)
{
struct perf_sample *sample = iter->sample;
struct mem_info *mi;
mi = sample__resolve_mem(sample, al);
if (mi == NULL)
return -ENOMEM;
iter->priv = mi;
return 0;
}
static int
iter_add_single_mem_entry(struct hist_entry_iter *iter, struct addr_location *al)
{
u64 cost;
struct mem_info *mi = iter->priv;
struct hists *hists = evsel__hists(iter->evsel);
struct perf_sample *sample = iter->sample;
struct hist_entry *he;
if (mi == NULL)
return -EINVAL;
cost = sample->weight;
if (!cost)
cost = 1;
/*
* must pass period=weight in order to get the correct
* sorting from hists__collapse_resort() which is solely
* based on periods. We want sorting be done on nr_events * weight
* and this is indirectly achieved by passing period=weight here
* and the he_stat__add_period() function.
*/
sample->period = cost;
he = __hists__add_entry(hists, al, iter->parent, NULL, mi,
sample, true);
if (!he)
return -ENOMEM;
iter->he = he;
return 0;
}
static int
iter_finish_mem_entry(struct hist_entry_iter *iter,
struct addr_location *al __maybe_unused)
{
struct perf_evsel *evsel = iter->evsel;
struct hists *hists = evsel__hists(evsel);
struct hist_entry *he = iter->he;
int err = -EINVAL;
if (he == NULL)
goto out;
hists__inc_nr_samples(hists, he->filtered);
err = hist_entry__append_callchain(he, iter->sample);
out:
/*
* We don't need to free iter->priv (mem_info) here since the mem info
* was either already freed in hists__findnew_entry() or passed to a
* new hist entry by hist_entry__new().
*/
iter->priv = NULL;
iter->he = NULL;
return err;
}
static int
iter_prepare_branch_entry(struct hist_entry_iter *iter, struct addr_location *al)
{
struct branch_info *bi;
struct perf_sample *sample = iter->sample;
bi = sample__resolve_bstack(sample, al);
if (!bi)
return -ENOMEM;
iter->curr = 0;
iter->total = sample->branch_stack->nr;
iter->priv = bi;
return 0;
}
static int
iter_add_single_branch_entry(struct hist_entry_iter *iter __maybe_unused,
struct addr_location *al __maybe_unused)
{
/* to avoid calling callback function */
iter->he = NULL;
return 0;
}
static int
iter_next_branch_entry(struct hist_entry_iter *iter, struct addr_location *al)
{
struct branch_info *bi = iter->priv;
int i = iter->curr;
if (bi == NULL)
return 0;
if (iter->curr >= iter->total)
return 0;
al->map = bi[i].to.map;
al->sym = bi[i].to.sym;
al->addr = bi[i].to.addr;
return 1;
}
static int
iter_add_next_branch_entry(struct hist_entry_iter *iter, struct addr_location *al)
{
struct branch_info *bi;
struct perf_evsel *evsel = iter->evsel;
struct hists *hists = evsel__hists(evsel);
struct perf_sample *sample = iter->sample;
struct hist_entry *he = NULL;
int i = iter->curr;
int err = 0;
bi = iter->priv;
if (iter->hide_unresolved && !(bi[i].from.sym && bi[i].to.sym))
goto out;
/*
* The report shows the percentage of total branches captured
* and not events sampled. Thus we use a pseudo period of 1.
*/
sample->period = 1;
sample->weight = bi->flags.cycles ? bi->flags.cycles : 1;
he = __hists__add_entry(hists, al, iter->parent, &bi[i], NULL,
sample, true);
if (he == NULL)
return -ENOMEM;
hists__inc_nr_samples(hists, he->filtered);
out:
iter->he = he;
iter->curr++;
return err;
}
static int
iter_finish_branch_entry(struct hist_entry_iter *iter,
struct addr_location *al __maybe_unused)
{
zfree(&iter->priv);
iter->he = NULL;
return iter->curr >= iter->total ? 0 : -1;
}
static int
iter_prepare_normal_entry(struct hist_entry_iter *iter __maybe_unused,
struct addr_location *al __maybe_unused)
{
return 0;
}
static int
iter_add_single_normal_entry(struct hist_entry_iter *iter, struct addr_location *al)
{
struct perf_evsel *evsel = iter->evsel;
struct perf_sample *sample = iter->sample;
struct hist_entry *he;
he = __hists__add_entry(evsel__hists(evsel), al, iter->parent, NULL, NULL,
sample, true);
if (he == NULL)
return -ENOMEM;
iter->he = he;
return 0;
}
static int
iter_finish_normal_entry(struct hist_entry_iter *iter,
struct addr_location *al __maybe_unused)
{
struct hist_entry *he = iter->he;
struct perf_evsel *evsel = iter->evsel;
struct perf_sample *sample = iter->sample;
if (he == NULL)
return 0;
iter->he = NULL;
hists__inc_nr_samples(evsel__hists(evsel), he->filtered);
return hist_entry__append_callchain(he, sample);
}
static int
iter_prepare_cumulative_entry(struct hist_entry_iter *iter,
struct addr_location *al __maybe_unused)
{
struct hist_entry **he_cache;
callchain_cursor_commit(&callchain_cursor);
/*
* This is for detecting cycles or recursions so that they're
* cumulated only one time to prevent entries more than 100%
* overhead.
*/
he_cache = malloc(sizeof(*he_cache) * (iter->max_stack + 1));
if (he_cache == NULL)
return -ENOMEM;
iter->priv = he_cache;
iter->curr = 0;
return 0;
}
static int
iter_add_single_cumulative_entry(struct hist_entry_iter *iter,
struct addr_location *al)
{
struct perf_evsel *evsel = iter->evsel;
struct hists *hists = evsel__hists(evsel);
struct perf_sample *sample = iter->sample;
struct hist_entry **he_cache = iter->priv;
struct hist_entry *he;
int err = 0;
he = __hists__add_entry(hists, al, iter->parent, NULL, NULL,
sample, true);
if (he == NULL)
return -ENOMEM;
iter->he = he;
he_cache[iter->curr++] = he;
hist_entry__append_callchain(he, sample);
/*
* We need to re-initialize the cursor since callchain_append()
* advanced the cursor to the end.
*/
callchain_cursor_commit(&callchain_cursor);
hists__inc_nr_samples(hists, he->filtered);
return err;
}
static int
iter_next_cumulative_entry(struct hist_entry_iter *iter,
struct addr_location *al)
{
struct callchain_cursor_node *node;
node = callchain_cursor_current(&callchain_cursor);
if (node == NULL)
return 0;
return fill_callchain_info(al, node, iter->hide_unresolved);
}
static int
iter_add_next_cumulative_entry(struct hist_entry_iter *iter,
struct addr_location *al)
{
struct perf_evsel *evsel = iter->evsel;
struct perf_sample *sample = iter->sample;
struct hist_entry **he_cache = iter->priv;
struct hist_entry *he;
struct hist_entry he_tmp = {
.hists = evsel__hists(evsel),
.cpu = al->cpu,
.thread = al->thread,
.comm = thread__comm(al->thread),
.ip = al->addr,
.ms = {
.map = al->map,
.sym = al->sym,
},
.parent = iter->parent,
.raw_data = sample->raw_data,
.raw_size = sample->raw_size,
};
int i;
struct callchain_cursor cursor;
callchain_cursor_snapshot(&cursor, &callchain_cursor);
callchain_cursor_advance(&callchain_cursor);
/*
* Check if there's duplicate entries in the callchain.
* It's possible that it has cycles or recursive calls.
*/
for (i = 0; i < iter->curr; i++) {
if (hist_entry__cmp(he_cache[i], &he_tmp) == 0) {
/* to avoid calling callback function */
iter->he = NULL;
return 0;
}
}
he = __hists__add_entry(evsel__hists(evsel), al, iter->parent, NULL, NULL,
sample, false);
if (he == NULL)
return -ENOMEM;
iter->he = he;
he_cache[iter->curr++] = he;
if (symbol_conf.use_callchain)
callchain_append(he->callchain, &cursor, sample->period);
return 0;
}
static int
iter_finish_cumulative_entry(struct hist_entry_iter *iter,
struct addr_location *al __maybe_unused)
{
zfree(&iter->priv);
iter->he = NULL;
return 0;
}
const struct hist_iter_ops hist_iter_mem = {
.prepare_entry = iter_prepare_mem_entry,
.add_single_entry = iter_add_single_mem_entry,
.next_entry = iter_next_nop_entry,
.add_next_entry = iter_add_next_nop_entry,
.finish_entry = iter_finish_mem_entry,
};
const struct hist_iter_ops hist_iter_branch = {
.prepare_entry = iter_prepare_branch_entry,
.add_single_entry = iter_add_single_branch_entry,
.next_entry = iter_next_branch_entry,
.add_next_entry = iter_add_next_branch_entry,
.finish_entry = iter_finish_branch_entry,
};
const struct hist_iter_ops hist_iter_normal = {
.prepare_entry = iter_prepare_normal_entry,
.add_single_entry = iter_add_single_normal_entry,
.next_entry = iter_next_nop_entry,
.add_next_entry = iter_add_next_nop_entry,
.finish_entry = iter_finish_normal_entry,
};
const struct hist_iter_ops hist_iter_cumulative = {
.prepare_entry = iter_prepare_cumulative_entry,
.add_single_entry = iter_add_single_cumulative_entry,
.next_entry = iter_next_cumulative_entry,
.add_next_entry = iter_add_next_cumulative_entry,
.finish_entry = iter_finish_cumulative_entry,
};
int hist_entry_iter__add(struct hist_entry_iter *iter, struct addr_location *al,
int max_stack_depth, void *arg)
{
int err, err2;
err = sample__resolve_callchain(iter->sample, &iter->parent,
iter->evsel, al, max_stack_depth);
if (err)
return err;
iter->max_stack = max_stack_depth;
err = iter->ops->prepare_entry(iter, al);
if (err)
goto out;
err = iter->ops->add_single_entry(iter, al);
if (err)
goto out;
if (iter->he && iter->add_entry_cb) {
err = iter->add_entry_cb(iter, al, true, arg);
if (err)
goto out;
}
while (iter->ops->next_entry(iter, al)) {
err = iter->ops->add_next_entry(iter, al);
if (err)
break;
if (iter->he && iter->add_entry_cb) {
err = iter->add_entry_cb(iter, al, false, arg);
if (err)
goto out;
}
}
out:
err2 = iter->ops->finish_entry(iter, al);
if (!err)
err = err2;
return err;
}
int64_t
hist_entry__cmp(struct hist_entry *left, struct hist_entry *right)
{
struct perf_hpp_fmt *fmt;
int64_t cmp = 0;
perf_hpp__for_each_sort_list(fmt) {
cmp = fmt->cmp(fmt, left, right);
if (cmp)
break;
}
return cmp;
}
int64_t
hist_entry__collapse(struct hist_entry *left, struct hist_entry *right)
{
struct perf_hpp_fmt *fmt;
int64_t cmp = 0;
perf_hpp__for_each_sort_list(fmt) {
cmp = fmt->collapse(fmt, left, right);
if (cmp)
break;
}
return cmp;
}
void hist_entry__delete(struct hist_entry *he)
{
thread__zput(he->thread);
map__zput(he->ms.map);
if (he->branch_info) {
map__zput(he->branch_info->from.map);
map__zput(he->branch_info->to.map);
zfree(&he->branch_info);
}
if (he->mem_info) {
map__zput(he->mem_info->iaddr.map);
map__zput(he->mem_info->daddr.map);
zfree(&he->mem_info);
}
zfree(&he->stat_acc);
free_srcline(he->srcline);
perf report: Add support for srcfile sort key In some cases it's useful to characterize samples by file. This is useful to get a higher level categorization, for example to map cost to subsystems. Add a srcfile sort key to perf report. It builds on top of the existing srcline support. Commiter notes: E.g.: # perf record -F 10000 usleep 1 [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.016 MB perf.data (13 samples) ] [root@zoo ~]# perf report -s srcfile --stdio # Total Lost Samples: 0 # # Samples: 13 of event 'cycles' # Event count (approx.): 869878 # # Overhead Source File # ........ ........... 60.99% . 20.62% paravirt.h 14.23% rmap.c 4.04% signal.c 0.11% msr.h # The first line is collecting all the files for which srcfiles couldn't somehow get resolved to: # perf report -s srcfile,dso --stdio # Total Lost Samples: 0 # # Samples: 13 of event 'cycles' # Event count (approx.): 869878 # # Overhead Source File Shared Object # ........ ........... ................ 40.97% . ld-2.20.so 20.62% paravirt.h [kernel.vmlinux] 20.02% . libc-2.20.so 14.23% rmap.c [kernel.vmlinux] 4.04% signal.c [kernel.vmlinux] 0.11% msr.h [kernel.vmlinux] # XXX: Investigate why that is not resolving on Fedora 21, Andi says he hasn't seen this on Fedora 22. Signed-off-by: Andi Kleen <ak@linux.intel.com> Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Link: http://lkml.kernel.org/r/1438988064-21834-1-git-send-email-andi@firstfloor.org [ Added column length update, from 0e65bdb3f90f ('perf hists: Update the column width for the "srcline" sort key') ] Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-08-07 22:54:24 +00:00
if (he->srcfile && he->srcfile[0])
free(he->srcfile);
free_callchain(he->callchain);
free(he->trace_output);
free(he->raw_data);
free(he);
}
/*
* collapse the histogram
*/
perf tools: Use __maybe_used for unused variables perf defines both __used and __unused variables to use for marking unused variables. The variable __used is defined to __attribute__((__unused__)), which contradicts the kernel definition to __attribute__((__used__)) for new gcc versions. On Android, __used is also defined in system headers and this leads to warnings like: warning: '__used__' attribute ignored __unused is not defined in the kernel and is not a standard definition. If __unused is included everywhere instead of __used, this leads to conflicts with glibc headers, since glibc has a variables with this name in its headers. The best approach is to use __maybe_unused, the definition used in the kernel for __attribute__((unused)). In this way there is only one definition in perf sources (instead of 2 definitions that point to the same thing: __used and __unused) and it works on both Linux and Android. This patch simply replaces all instances of __used and __unused with __maybe_unused. Signed-off-by: Irina Tirdea <irina.tirdea@intel.com> Acked-by: Pekka Enberg <penberg@kernel.org> Cc: David Ahern <dsahern@gmail.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: Namhyung Kim <namhyung.kim@lge.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Steven Rostedt <rostedt@goodmis.org> Link: http://lkml.kernel.org/r/1347315303-29906-7-git-send-email-irina.tirdea@intel.com [ committer note: fixed up conflict with a116e05 in builtin-sched.c ] Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2012-09-10 22:15:03 +00:00
static bool hists__collapse_insert_entry(struct hists *hists __maybe_unused,
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 03:51:58 +00:00
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) {
he_stat__add_stat(&iter->stat, &he->stat);
if (symbol_conf.cumulate_callchain)
he_stat__add_stat(iter->stat_acc, he->stat_acc);
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 03:51:58 +00:00
if (symbol_conf.use_callchain) {
callchain_cursor_reset(&callchain_cursor);
callchain_merge(&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 03:51:58 +00:00
he->callchain);
}
hist_entry__delete(he);
return false;
}
if (cmp < 0)
p = &(*p)->rb_left;
else
p = &(*p)->rb_right;
}
hists->nr_entries++;
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);
hists__filter_entry_by_symbol(hists, he);
hists__filter_entry_by_socket(hists, he);
}
void hists__collapse_resort(struct hists *hists, struct ui_progress *prog)
{
struct rb_root *root;
struct rb_node *next;
struct hist_entry *n;
if (!sort__need_collapse)
return;
hists->nr_entries = 0;
root = hists__get_rotate_entries_in(hists);
next = rb_first(root);
while (next) {
if (session_done())
break;
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);
}
if (prog)
ui_progress__update(prog, 1);
}
}
static int hist_entry__sort(struct hist_entry *a, struct hist_entry *b)
{
struct perf_hpp_fmt *fmt;
int64_t cmp = 0;
perf_hpp__for_each_sort_list(fmt) {
perf tools: Skip dynamic fields not defined for current event When there are multiple events, each dynamic sort key is defined just for one event. In this case other events will always show "N/A" for those fields. But they are meaningless and consume precious screen width. Let's skip those undefined dynamic fields. $ perf record -e kmem:kmalloc,kmem:kfree -a sleep 1 $ perf report -s 'comm,kmalloc.*' --stdio # To display the perf.data header info, please use --header/--header-only options. # # # Total Lost Samples: 0 # # Samples: 20K of event 'kmem:kmalloc' # Event count (approx.): 20533 # # Overhead Command call_site ptr bytes_req bytes_alloc gfp_flags # ........ ....... .................. .................. ......... ........... ................... # 99.89% perf ffffffffa01d4396 0xffff8803ffb79720 96 96 GFP_NOFS|GFP_ZERO 0.06% sleep ffffffff8114e1cd 0xffff8803d228a000 4096 4096 GFP_KERNEL 0.03% perf ffffffff811d6ae6 0xffff8803f7678f00 240 256 GFP_KERNEL|GFP_ZERO 0.00% perf ffffffff812263c1 0xffff880406172380 128 128 GFP_KERNEL 0.00% perf ffffffff812264b9 0xffff8803ffac1600 504 512 GFP_KERNEL 0.00% perf ffffffff81226634 0xffff880401dc5280 28 32 GFP_KERNEL 0.00% sleep ffffffff81226da9 0xffff8803ffac3a00 392 512 GFP_KERNEL # Samples: 20K of event 'kmem:kfree' # Event count (approx.): 20597 # # Overhead Command # ........ .............. # 99.63% perf 0.14% sleep 0.11% irq/36-iwlwifi 0.11% kworker/u16:0 0.01% Xorg 0.00% firefox Signed-off-by: Namhyung Kim <namhyung@kernel.org> Acked-by: Jiri Olsa <jolsa@kernel.org> Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Andi Kleen <andi@firstfloor.org> Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Wang Nan <wangnan0@huawei.com> Link: http://lkml.kernel.org/r/1450804030-29193-12-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-12-22 17:07:08 +00:00
if (perf_hpp__should_skip(fmt, a->hists))
continue;
cmp = fmt->sort(fmt, a, b);
if (cmp)
break;
}
return cmp;
}
static void hists__reset_filter_stats(struct hists *hists)
{
hists->nr_non_filtered_entries = 0;
hists->stats.total_non_filtered_period = 0;
}
void hists__reset_stats(struct hists *hists)
{
hists->nr_entries = 0;
hists->stats.total_period = 0;
hists__reset_filter_stats(hists);
}
static void hists__inc_filter_stats(struct hists *hists, struct hist_entry *h)
{
hists->nr_non_filtered_entries++;
hists->stats.total_non_filtered_period += h->stat.period;
}
void hists__inc_stats(struct hists *hists, struct hist_entry *h)
{
if (!h->filtered)
hists__inc_filter_stats(hists, h);
hists->nr_entries++;
hists->stats.total_period += h->stat.period;
}
2010-05-10 16:04:11 +00:00
static void __hists__insert_output_entry(struct rb_root *entries,
struct hist_entry *he,
perf callchain: Allow disabling call graphs per event This patch introduce "call-graph=no" to disable per-event callgraph. Here is an example. perf record -e 'cpu/cpu-cycles,call-graph=fp/,cpu/instructions,call-graph=no/' sleep 1 perf report --stdio # To display the perf.data header info, please use --header/--header-only options. # # # Total Lost Samples: 0 # # Samples: 6 of event 'cpu/cpu-cycles,call-graph=fp/' # Event count (approx.): 774218 # # Children Self Command Shared Object Symbol # ........ ........ ....... ................ ........................................ # 61.94% 0.00% sleep [kernel.vmlinux] [k] entry_SYSCALL_64_fastpath | ---entry_SYSCALL_64_fastpath | |--97.30%-- __brk | --2.70%-- mmap64 _dl_check_map_versions _dl_check_all_versions 61.94% 0.00% sleep [kernel.vmlinux] [k] perf_event_mmap | ---perf_event_mmap | |--97.30%-- do_brk | sys_brk | entry_SYSCALL_64_fastpath | __brk | --2.70%-- mmap_region do_mmap_pgoff vm_mmap_pgoff sys_mmap_pgoff sys_mmap entry_SYSCALL_64_fastpath mmap64 _dl_check_map_versions _dl_check_all_versions ...... # Samples: 6 of event 'cpu/instructions,call-graph=no/' # Event count (approx.): 359692 # # Children Self Command Shared Object Symbol # ........ ........ ....... ................ ................................. # 89.03% 0.00% sleep [unknown] [.] 0xffff6598ffff6598 89.03% 0.00% sleep ld-2.17.so [.] _dl_resolve_conflicts 89.03% 0.00% sleep [kernel.vmlinux] [k] page_fault Signed-off-by: Kan Liang <kan.liang@intel.com> Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Link: http://lkml.kernel.org/r/1439289050-40510-2-git-send-email-kan.liang@intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-08-11 10:30:48 +00:00
u64 min_callchain_hits,
bool use_callchain)
{
2010-05-10 16:04:11 +00:00
struct rb_node **p = &entries->rb_node;
struct rb_node *parent = NULL;
struct hist_entry *iter;
perf callchain: Allow disabling call graphs per event This patch introduce "call-graph=no" to disable per-event callgraph. Here is an example. perf record -e 'cpu/cpu-cycles,call-graph=fp/,cpu/instructions,call-graph=no/' sleep 1 perf report --stdio # To display the perf.data header info, please use --header/--header-only options. # # # Total Lost Samples: 0 # # Samples: 6 of event 'cpu/cpu-cycles,call-graph=fp/' # Event count (approx.): 774218 # # Children Self Command Shared Object Symbol # ........ ........ ....... ................ ........................................ # 61.94% 0.00% sleep [kernel.vmlinux] [k] entry_SYSCALL_64_fastpath | ---entry_SYSCALL_64_fastpath | |--97.30%-- __brk | --2.70%-- mmap64 _dl_check_map_versions _dl_check_all_versions 61.94% 0.00% sleep [kernel.vmlinux] [k] perf_event_mmap | ---perf_event_mmap | |--97.30%-- do_brk | sys_brk | entry_SYSCALL_64_fastpath | __brk | --2.70%-- mmap_region do_mmap_pgoff vm_mmap_pgoff sys_mmap_pgoff sys_mmap entry_SYSCALL_64_fastpath mmap64 _dl_check_map_versions _dl_check_all_versions ...... # Samples: 6 of event 'cpu/instructions,call-graph=no/' # Event count (approx.): 359692 # # Children Self Command Shared Object Symbol # ........ ........ ....... ................ ................................. # 89.03% 0.00% sleep [unknown] [.] 0xffff6598ffff6598 89.03% 0.00% sleep ld-2.17.so [.] _dl_resolve_conflicts 89.03% 0.00% sleep [kernel.vmlinux] [k] page_fault Signed-off-by: Kan Liang <kan.liang@intel.com> Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Link: http://lkml.kernel.org/r/1439289050-40510-2-git-send-email-kan.liang@intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-08-11 10:30:48 +00:00
if (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 (hist_entry__sort(he, iter) > 0)
p = &(*p)->rb_left;
else
p = &(*p)->rb_right;
}
rb_link_node(&he->rb_node, parent, p);
2010-05-10 16:04:11 +00:00
rb_insert_color(&he->rb_node, entries);
}
void hists__output_resort(struct hists *hists, struct ui_progress *prog)
{
struct rb_root *root;
struct rb_node *next;
struct hist_entry *n;
u64 min_callchain_hits;
perf callchain: Allow disabling call graphs per event This patch introduce "call-graph=no" to disable per-event callgraph. Here is an example. perf record -e 'cpu/cpu-cycles,call-graph=fp/,cpu/instructions,call-graph=no/' sleep 1 perf report --stdio # To display the perf.data header info, please use --header/--header-only options. # # # Total Lost Samples: 0 # # Samples: 6 of event 'cpu/cpu-cycles,call-graph=fp/' # Event count (approx.): 774218 # # Children Self Command Shared Object Symbol # ........ ........ ....... ................ ........................................ # 61.94% 0.00% sleep [kernel.vmlinux] [k] entry_SYSCALL_64_fastpath | ---entry_SYSCALL_64_fastpath | |--97.30%-- __brk | --2.70%-- mmap64 _dl_check_map_versions _dl_check_all_versions 61.94% 0.00% sleep [kernel.vmlinux] [k] perf_event_mmap | ---perf_event_mmap | |--97.30%-- do_brk | sys_brk | entry_SYSCALL_64_fastpath | __brk | --2.70%-- mmap_region do_mmap_pgoff vm_mmap_pgoff sys_mmap_pgoff sys_mmap entry_SYSCALL_64_fastpath mmap64 _dl_check_map_versions _dl_check_all_versions ...... # Samples: 6 of event 'cpu/instructions,call-graph=no/' # Event count (approx.): 359692 # # Children Self Command Shared Object Symbol # ........ ........ ....... ................ ................................. # 89.03% 0.00% sleep [unknown] [.] 0xffff6598ffff6598 89.03% 0.00% sleep ld-2.17.so [.] _dl_resolve_conflicts 89.03% 0.00% sleep [kernel.vmlinux] [k] page_fault Signed-off-by: Kan Liang <kan.liang@intel.com> Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Link: http://lkml.kernel.org/r/1439289050-40510-2-git-send-email-kan.liang@intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-08-11 10:30:48 +00:00
struct perf_evsel *evsel = hists_to_evsel(hists);
bool use_callchain;
if (evsel && symbol_conf.use_callchain && !symbol_conf.show_ref_callgraph)
use_callchain = evsel->attr.sample_type & PERF_SAMPLE_CALLCHAIN;
else
use_callchain = symbol_conf.use_callchain;
min_callchain_hits = hists->stats.total_period * (callchain_param.min_percent / 100);
if (sort__need_collapse)
root = &hists->entries_collapsed;
else
root = hists->entries_in;
next = rb_first(root);
hists->entries = RB_ROOT;
hists__reset_stats(hists);
hists__reset_col_len(hists);
while (next) {
n = rb_entry(next, struct hist_entry, rb_node_in);
next = rb_next(&n->rb_node_in);
perf callchain: Allow disabling call graphs per event This patch introduce "call-graph=no" to disable per-event callgraph. Here is an example. perf record -e 'cpu/cpu-cycles,call-graph=fp/,cpu/instructions,call-graph=no/' sleep 1 perf report --stdio # To display the perf.data header info, please use --header/--header-only options. # # # Total Lost Samples: 0 # # Samples: 6 of event 'cpu/cpu-cycles,call-graph=fp/' # Event count (approx.): 774218 # # Children Self Command Shared Object Symbol # ........ ........ ....... ................ ........................................ # 61.94% 0.00% sleep [kernel.vmlinux] [k] entry_SYSCALL_64_fastpath | ---entry_SYSCALL_64_fastpath | |--97.30%-- __brk | --2.70%-- mmap64 _dl_check_map_versions _dl_check_all_versions 61.94% 0.00% sleep [kernel.vmlinux] [k] perf_event_mmap | ---perf_event_mmap | |--97.30%-- do_brk | sys_brk | entry_SYSCALL_64_fastpath | __brk | --2.70%-- mmap_region do_mmap_pgoff vm_mmap_pgoff sys_mmap_pgoff sys_mmap entry_SYSCALL_64_fastpath mmap64 _dl_check_map_versions _dl_check_all_versions ...... # Samples: 6 of event 'cpu/instructions,call-graph=no/' # Event count (approx.): 359692 # # Children Self Command Shared Object Symbol # ........ ........ ....... ................ ................................. # 89.03% 0.00% sleep [unknown] [.] 0xffff6598ffff6598 89.03% 0.00% sleep ld-2.17.so [.] _dl_resolve_conflicts 89.03% 0.00% sleep [kernel.vmlinux] [k] page_fault Signed-off-by: Kan Liang <kan.liang@intel.com> Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Link: http://lkml.kernel.org/r/1439289050-40510-2-git-send-email-kan.liang@intel.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-08-11 10:30:48 +00:00
__hists__insert_output_entry(&hists->entries, n, min_callchain_hits, use_callchain);
hists__inc_stats(hists, n);
if (!n->filtered)
hists__calc_col_len(hists, n);
if (prog)
ui_progress__update(prog, 1);
}
}
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;
/* force fold unfiltered entry for simplicity */
h->unfolded = false;
h->row_offset = 0;
perf hists browser: Fix UI bug after zoom into thread/dso/symbol When zoom into thread/dso/symbol, the fold/unfold stat is cleared in hists__filter_by_thread/dso/symbol(), but h->nr_rows is not cleared. So if we toggle fold stat on the unfold entires, nr_entries got a wrong value. This bug can be reproduced as follows: $ perf record -g -e syscalls:sys_enter_open ls $ perf report Children Self Command Shared Object Symbol ================================================================ + 50.00% 0.00% ls ld64.so [.] _dl_get_ready_to_run - 50.00% 0.00% ls ld64.so [.] _dl_load_shared_library _dl_load_shared_library <= [Zoom into thread/dso] _dl_get_ready_to_run _start ... In the new thread hists, all entries reset to fold, if we unfold the same entry as we previously unfolded, nr_entries got wrong value, and we can't move down cursor to bottom row. Thread: ls Children Self Command Shared Object Symbol ================================================================ + 50.00% 0.00% ls ld64.so [.] _dl_get_ready_to_run - 50.00% 0.00% ls ld64.so [.] _dl_load_shared_library _dl_load_shared_library _dl_get_ready_to_run <= [cursor may stop here, can't move down] _start ... This patch clear h->nr_rows to fix this bug. Signed-off-by: He Kuang <hekuang@huawei.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Wang Nan <wangnan0@huawei.com> Link: http://lkml.kernel.org/r/1426077363-855-2-git-send-email-hekuang@huawei.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-03-11 12:36:03 +00:00
h->nr_rows = 0;
hists->stats.nr_non_filtered_samples += h->stat.nr_events;
hists__inc_filter_stats(hists, h);
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->stats.nr_non_filtered_samples = 0;
hists__reset_filter_stats(hists);
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->stats.nr_non_filtered_samples = 0;
hists__reset_filter_stats(hists);
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);
}
}
static bool hists__filter_entry_by_symbol(struct hists *hists,
struct hist_entry *he)
{
if (hists->symbol_filter_str != NULL &&
(!he->ms.sym || strstr(he->ms.sym->name,
hists->symbol_filter_str) == NULL)) {
he->filtered |= (1 << HIST_FILTER__SYMBOL);
return true;
}
return false;
}
void hists__filter_by_symbol(struct hists *hists)
{
struct rb_node *nd;
hists->stats.nr_non_filtered_samples = 0;
hists__reset_filter_stats(hists);
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_symbol(hists, h))
continue;
hists__remove_entry_filter(hists, h, HIST_FILTER__SYMBOL);
}
}
static bool hists__filter_entry_by_socket(struct hists *hists,
struct hist_entry *he)
{
if ((hists->socket_filter > -1) &&
(he->socket != hists->socket_filter)) {
he->filtered |= (1 << HIST_FILTER__SOCKET);
return true;
}
return false;
}
void hists__filter_by_socket(struct hists *hists)
{
struct rb_node *nd;
hists->stats.nr_non_filtered_samples = 0;
hists__reset_filter_stats(hists);
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_socket(hists, h))
continue;
hists__remove_entry_filter(hists, h, HIST_FILTER__SOCKET);
}
}
void events_stats__inc(struct events_stats *stats, u32 type)
{
++stats->nr_events[0];
++stats->nr_events[type];
}
void hists__inc_nr_events(struct hists *hists, u32 type)
{
events_stats__inc(&hists->stats, type);
}
void hists__inc_nr_samples(struct hists *hists, bool filtered)
{
events_stats__inc(&hists->stats, PERF_RECORD_SAMPLE);
if (!filtered)
hists->stats.nr_non_filtered_samples++;
}
static struct hist_entry *hists__add_dummy_entry(struct hists *hists,
struct hist_entry *pair)
{
struct rb_root *root;
struct rb_node **p;
struct rb_node *parent = NULL;
struct hist_entry *he;
int64_t cmp;
if (sort__need_collapse)
root = &hists->entries_collapsed;
else
root = hists->entries_in;
p = &root->rb_node;
while (*p != NULL) {
parent = *p;
he = rb_entry(parent, struct hist_entry, rb_node_in);
cmp = hist_entry__collapse(he, pair);
if (!cmp)
goto out;
if (cmp < 0)
p = &(*p)->rb_left;
else
p = &(*p)->rb_right;
}
he = hist_entry__new(pair, true);
if (he) {
memset(&he->stat, 0, sizeof(he->stat));
he->hists = hists;
rb_link_node(&he->rb_node_in, parent, p);
rb_insert_color(&he->rb_node_in, root);
hists__inc_stats(hists, he);
he->dummy = true;
}
out:
return he;
}
static struct hist_entry *hists__find_entry(struct hists *hists,
struct hist_entry *he)
{
struct rb_node *n;
if (sort__need_collapse)
n = hists->entries_collapsed.rb_node;
else
n = hists->entries_in->rb_node;
while (n) {
struct hist_entry *iter = rb_entry(n, struct hist_entry, rb_node_in);
int64_t cmp = hist_entry__collapse(iter, he);
if (cmp < 0)
n = n->rb_left;
else if (cmp > 0)
n = n->rb_right;
else
return iter;
}
return NULL;
}
/*
* Look for pairs to link to the leader buckets (hist_entries):
*/
void hists__match(struct hists *leader, struct hists *other)
{
struct rb_root *root;
struct rb_node *nd;
struct hist_entry *pos, *pair;
if (sort__need_collapse)
root = &leader->entries_collapsed;
else
root = leader->entries_in;
for (nd = rb_first(root); nd; nd = rb_next(nd)) {
pos = rb_entry(nd, struct hist_entry, rb_node_in);
pair = hists__find_entry(other, pos);
if (pair)
hist_entry__add_pair(pair, pos);
}
}
/*
* Look for entries in the other hists that are not present in the leader, if
* we find them, just add a dummy entry on the leader hists, with period=0,
* nr_events=0, to serve as the list header.
*/
int hists__link(struct hists *leader, struct hists *other)
{
struct rb_root *root;
struct rb_node *nd;
struct hist_entry *pos, *pair;
if (sort__need_collapse)
root = &other->entries_collapsed;
else
root = other->entries_in;
for (nd = rb_first(root); nd; nd = rb_next(nd)) {
pos = rb_entry(nd, struct hist_entry, rb_node_in);
if (!hist_entry__has_pairs(pos)) {
pair = hists__add_dummy_entry(leader, pos);
if (pair == NULL)
return -1;
hist_entry__add_pair(pos, pair);
}
}
return 0;
}
void hist__account_cycles(struct branch_stack *bs, struct addr_location *al,
struct perf_sample *sample, bool nonany_branch_mode)
{
struct branch_info *bi;
/* If we have branch cycles always annotate them. */
if (bs && bs->nr && bs->entries[0].flags.cycles) {
int i;
bi = sample__resolve_bstack(sample, al);
if (bi) {
struct addr_map_symbol *prev = NULL;
/*
* Ignore errors, still want to process the
* other entries.
*
* For non standard branch modes always
* force no IPC (prev == NULL)
*
* Note that perf stores branches reversed from
* program order!
*/
for (i = bs->nr - 1; i >= 0; i--) {
addr_map_symbol__account_cycles(&bi[i].from,
nonany_branch_mode ? NULL : prev,
bi[i].flags.cycles);
prev = &bi[i].to;
}
free(bi);
}
}
}
size_t perf_evlist__fprintf_nr_events(struct perf_evlist *evlist, FILE *fp)
{
struct perf_evsel *pos;
size_t ret = 0;
evlist__for_each(evlist, pos) {
ret += fprintf(fp, "%s stats:\n", perf_evsel__name(pos));
ret += events_stats__fprintf(&evsel__hists(pos)->stats, fp);
}
return ret;
}
u64 hists__total_period(struct hists *hists)
{
return symbol_conf.filter_relative ? hists->stats.total_non_filtered_period :
hists->stats.total_period;
}
int parse_filter_percentage(const struct option *opt __maybe_unused,
const char *arg, int unset __maybe_unused)
{
if (!strcmp(arg, "relative"))
symbol_conf.filter_relative = true;
else if (!strcmp(arg, "absolute"))
symbol_conf.filter_relative = false;
else
return -1;
return 0;
}
int perf_hist_config(const char *var, const char *value)
{
if (!strcmp(var, "hist.percentage"))
return parse_filter_percentage(NULL, value, 0);
return 0;
}
static int hists_evsel__init(struct perf_evsel *evsel)
{
struct hists *hists = evsel__hists(evsel);
memset(hists, 0, sizeof(*hists));
hists->entries_in_array[0] = hists->entries_in_array[1] = RB_ROOT;
hists->entries_in = &hists->entries_in_array[0];
hists->entries_collapsed = RB_ROOT;
hists->entries = RB_ROOT;
pthread_mutex_init(&hists->lock, NULL);
hists->socket_filter = -1;
return 0;
}
static void hists__delete_remaining_entries(struct rb_root *root)
{
struct rb_node *node;
struct hist_entry *he;
while (!RB_EMPTY_ROOT(root)) {
node = rb_first(root);
rb_erase(node, root);
he = rb_entry(node, struct hist_entry, rb_node_in);
hist_entry__delete(he);
}
}
static void hists__delete_all_entries(struct hists *hists)
{
hists__delete_entries(hists);
hists__delete_remaining_entries(&hists->entries_in_array[0]);
hists__delete_remaining_entries(&hists->entries_in_array[1]);
hists__delete_remaining_entries(&hists->entries_collapsed);
}
static void hists_evsel__exit(struct perf_evsel *evsel)
{
struct hists *hists = evsel__hists(evsel);
hists__delete_all_entries(hists);
}
/*
* XXX We probably need a hists_evsel__exit() to free the hist_entries
* stored in the rbtree...
*/
int hists__init(void)
{
int err = perf_evsel__object_config(sizeof(struct hists_evsel),
hists_evsel__init,
hists_evsel__exit);
if (err)
fputs("FATAL ERROR: Couldn't setup hists class\n", stderr);
return err;
}