perf, tools, report: Add support for srcfile sort key
PATCH 1/1 perf, tools, report: Add support for srcfile sort key
From: Andi Kleen <[email protected]>
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.
--- tools/perf/Documentation/perf-report.txt | 2 ++ tools/perf/util/hist.c | 2 ++ tools/perf/util/hist.h | 1 + tools/perf/util/sort.c | 52 ++++++++++++++++++++++++++++++++ tools/perf/util/sort.h | 2 ++ 5 files changed, 59 insertions(+)
tools/perf/Documentation/perf-report.txt +2 -0
@@ -81,6 +81,8 @@ OPTIONS - cpu: cpu number the task ran at the time of sample - srcline: filename and line number executed at the time of sample. The DWARF debugging info must be provided. + - srcfile: file name of the source file of the same. Requires dwarf + information. - weight: Event specific weight, e.g. memory latency or transaction abort cost. This is the global weight. - local_weight: Local weight version of the weight above.
tools/perf/util/hist.c +2 -0
@@ -944,6 +944,8 @@ void hist_entry__delete(struct hist_entry *he)zfree (&he->stat_acc);free_srcline (he->srcline); +if (he->srcfile && he->srcfile[0 ]) +free (he->srcfile);free_callchain (he->callchain);free (he); }
tools/perf/util/hist.h +1 -0
@@ -30,6 +30,7 @@ enum hist_column {HISTC_PARENT ,HISTC_CPU ,HISTC_SRCLINE , +HISTC_SRCFILE ,HISTC_MISPREDICT ,HISTC_IN_TX , HISTC_ABORT,
tools/perf/util/sort.c +52 -0
64 lines not on this page · raw message
tools/perf/util/sort.h +2 -0
@@ -114,6 +114,7 @@ struct hist_entry { }; };char *srcline; +char *srcfile;struct symbol *parent;struct rb_root sorted_chain;struct branch_info *branch_info;
@@ -172,6 +173,7 @@ enum sort_type {SORT_PARENT ,SORT_CPU ,SORT_SRCLINE , +SORT_SRCFILE ,SORT_LOCAL_WEIGHT ,SORT_GLOBAL_WEIGHT , SORT_TRANSACTION,
Signature · 1 line
2.4.3
Em Fri, Aug 07, 2015 at 03:54:24PM -0700, Andi Kleen escreveu:
Quoted 8 lines
From: Andi Kleen <[email protected]>
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.
Applied
Quoted 139 lines
Signed-off-by Andi Kleen <[email protected]>--- tools/perf/Documentation/perf-report.txt | 2 ++ tools/perf/util/hist.c | 2 ++ tools/perf/util/hist.h | 1 + tools/perf/util/sort.c | 52 ++++++++++++++++++++++++++++++++ tools/perf/util/sort.h | 2 ++ 5 files changed, 59 insertions(+)tools/perf/Documentation/perf-report.txt +2
-0@@ -81,6 +81,8 @@ OPTIONS - cpu: cpu number the task ran at the time of sample - srcline: filename and line number executed at the time of sample. The DWARF debugging info must be provided. + - srcfile: file name of the source file of the same. Requires dwarf + information. - weight: Event specific weight, e.g. memory latency or transaction abort cost. This is the global weight. - local_weight: Local weight version of the weight above.tools/perf/util/hist.c +2
-0@@ -944,6 +944,8 @@ void hist_entry__delete(struct hist_entry *he)zfree (&he->stat_acc);free_srcline (he->srcline); +if (he->srcfile && he->srcfile[0 ]) +free (he->srcfile);free_callchain (he->callchain);free (he); }tools/perf/util/hist.h +1
-0@@ -30,6 +30,7 @@ enum hist_column {HISTC_PARENT ,HISTC_CPU ,HISTC_SRCLINE , +HISTC_SRCFILE ,HISTC_MISPREDICT ,HISTC_IN_TX , HISTC_ABORT,tools/perf/util/sort.c +52
-0@@ -319,6 +319,57 @@ struct sort_entry sort_srcline = { .se_width_idx = HISTC_SRCLINE, }; +/* --sort srcfile */ + +static char no_srcfile[1 ]; + +static char *get_srcfile (struct hist_entry *e) +{ +char *sf, *p; +struct map *map = e->ms.map; + + sf =get_srcline (map->dso,map__rip_2objdump (map, e->ip), + e->ms.sym, true); + p =strchr (sf,':' ); +if (p && *sf) { + *p =0 ; +return sf; + } +free (sf); +return no_srcfile; +} + +static int64_t +sort__srcfile_cmp (struct hist_entry *left,struct hist_entry *right) +{ +if (!left->srcfile) { +if (!left->ms.map) + left->srcfile = no_srcfile; +else + left->srcfile =get_srcfile (left); + } +if (!right->srcfile) { +if (!right->ms.map) + right->srcfile = no_srcfile; +else + right->srcfile =get_srcfile (right); + } +return strcmp (right->srcfile, left->srcfile); +} + +static int hist_entry__srcfile_snprintf (struct hist_entry *he,char *bf, +size_t size,unsigned int width) +{ +return repsep_snprintf (bf, size,"%-*.*s" , width, width, he->srcfile); +} + +struct sort_entry sort_srcfile = { + .se_header ="Source File" , + .se_cmp = sort__srcfile_cmp, + .se_snprintf = hist_entry__srcfile_snprintf, + .se_width_idx =HISTC_SRCFILE , +}; +/* --sort parent */ static int64_t@@ -1173,6 +1224,7 @@ static struct sort_dimension common_sort_dimensions[] = {DIM (SORT_PARENT ,"parent" , sort_parent),DIM (SORT_CPU ,"cpu" , sort_cpu),DIM (SORT_SRCLINE ,"srcline" , sort_srcline), +DIM (SORT_SRCFILE ,"srcfile" , sort_srcfile),DIM (SORT_LOCAL_WEIGHT ,"local_weight" , sort_local_weight),DIM (SORT_GLOBAL_WEIGHT ,"weight" , sort_global_weight), DIM(SORT_TRANSACTION, "transaction", sort_transaction),tools/perf/util/sort.h +2
-0@@ -114,6 +114,7 @@ struct hist_entry { }; };char *srcline; +char *srcfile;struct symbol *parent;struct rb_root sorted_chain;struct branch_info *branch_info;@@ -172,6 +173,7 @@ enum sort_type {SORT_PARENT ,SORT_CPU ,SORT_SRCLINE , +SORT_SRCFILE ,SORT_LOCAL_WEIGHT ,SORT_GLOBAL_WEIGHT , SORT_TRANSACTION,Signature · 1 line
2.4.3
Arnaldo Carvalho de Melo
maintainerEm Fri, Aug 07, 2015 at 08:51:45PM -0300, Arnaldo Carvalho de Melo escreveu:
Quoted 11 lines
Em Fri, Aug 07, 2015 at 03:54:24PM -0700, Andi Kleen escreveu:
Quoted 8 lines
From: Andi Kleen <[email protected]>
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.
Applied
Humm, holding this up a bit, further testing showed some oddities, fedora21, the width of the column is being limited to the lenght of the header and there are some DWARF errors, have you noticed those?
[root@zoo ~]# rpm -q binutils-devel
binutils-devel-2.24-32.fc21.x86_64
# perf report --stdio -s srcfile
Failed to open /tmp/perf-2268.map, continuing without symbols
# To display the perf.data header info, please use --header/--header-only options.
#
BFD: Dwarf Error: Offset (124839) greater than or equal to .debug_str size (22876).
BFD: Dwarf Error: Offset (111062) greater than or equal to .debug_str size (22876).
BFD: Dwarf Error: Offset (111062) greater than or equal to .debug_str size (22876).
BFD: Dwarf Error: Offset (70405) greater than or equal to .debug_str size (22876).
BFD: Dwarf Error: Offset (113124) greater than or equal to .debug_str size (22876).
BFD: Dwarf Error: Offset (113124) greater than or equal to .debug_str size (22876).
BFD: Dwarf Error: Offset (113124) greater than or equal to .debug_str size (22876).
BFD: Dwarf Error: Offset (113124) greater than or equal to .debug_str size (22876).
BFD: Dwarf Error: Offset (113124) greater than or equal to .debug_str size (22876).
BFD: Dwarf Error: Offset (70405) greater than or equal to .debug_str size (22876).
BFD: Dwarf Error: Offset (70405) greater than or equal to .debug_str size (22876).
BFD: Dwarf Error: Offset (70405) greater than or equal to .debug_str size (22876).
BFD: Dwarf Error: Offset (70405) greater than or equal to .debug_str size (22876).
BFD: Dwarf Error: Offset (70405) greater than or equal to .debug_str size (22876).
BFD: Dwarf Error: Offset (124839) greater than or equal to .debug_str size (22876).
BFD: Dwarf Error: Offset (124839) greater than or equal to .debug_str size (22876).
BFD: Dwarf Error: Offset (113124) greater than or equal to .debug_str size (22876).
BFD: Dwarf Error: Offset (113124) greater than or equal to .debug_str size (22876).
BFD: Dwarf Error: Offset (113124) greater than or equal to .debug_str size (22876).
BFD: Dwarf Error: Offset (113124) greater than or equal to .debug_str size (22876).
BFD: Dwarf Error: Offset (113124) greater than or equal to .debug_str size (22876).
BFD: Dwarf Error: Offset (113124) greater than or equal to .debug_str size (22876).
BFD: Dwarf Error: Offset (113124) greater than or equal to .debug_str size (22876).
BFD: Dwarf Error: Offset (113124) greater than or equal to .debug_str size (22876).
BFD: Dwarf Error: Offset (113124) greater than or equal to .debug_str size (22876).
BFD: Dwarf Error: Offset (113124) greater than or equal to .debug_str size (22876).
BFD: Dwarf Error: Offset (113124) greater than or equal to .debug_str size (22876).
BFD: Dwarf Error: Offset (113124) greater than or equal to .debug_str size (22876).
BFD: Dwarf Error: Offset (113124) greater than or equal to .debug_str size (22876).
BFD: Dwarf Error: Offset (113124) greater than or equal to .debug_str size (22876).
BFD: Dwarf Error: Offset (113124) greater than or equal to .debug_str size (22876).
BFD: Dwarf Error: Offset (113124) greater than or equal to .debug_str size (22876).
BFD: Dwarf Error: Offset (113124) greater than or equal to .debug_str size (22876).
BFD: Dwarf Error: Offset (113124) greater than or equal to .debug_str size (22876).
BFD: Dwarf Error: Offset (124839) greater than or equal to .debug_str size (22876).
BFD: Dwarf Error: Offset (70405) greater than or equal to .debug_str size (22876).
BFD: Dwarf Error: Offset (113124) greater than or equal to .debug_str size (22876).
BFD: Dwarf Error: Offset (124839) greater than or equal to .debug_str size (22876).
BFD: Dwarf Error: Offset (113124) greater than or equal to .debug_str size (22876).
BFD: Dwarf Error: Offset (113124) greater than or equal to .debug_str size (58106).
BFD: Dwarf Error: Offset (113124) greater than or equal to .debug_str size (58106).
BFD: Dwarf Error: Offset (58380) greater than or equal to .debug_str size (17648).
BFD: Dwarf Error: Offset (124839) greater than or equal to .debug_str size (58106).
BFD: Dwarf Error: Offset (70405) greater than or equal to .debug_str size (58106).
BFD: Dwarf Error: Offset (124839) greater than or equal to .debug_str size (58106).
BFD: Dwarf Error: Offset (124839) greater than or equal to .debug_str size (58106).
BFD: Dwarf Error: Offset (70405) greater than or equal to .debug_str size (58106).
BFD: Dwarf Error: Offset (70405) greater than or equal to .debug_str size (58106).
BFD: Dwarf Error: Offset (70405) greater than or equal to .debug_str size (58106).
BFD: Dwarf Error: Offset (124839) greater than or equal to .debug_str size (58106).
BFD: Dwarf Error: Offset (70405) greater than or equal to .debug_str size (58106).
BFD: Dwarf Error: Offset (124839) greater than or equal to .debug_str size (58106).
#
# Total Lost Samples: 0
#
# Samples: 7K of event 'cycles'
# Event count (approx.): 1770111576
#
# Overhead Source File
# ........ ...........
#
31.99%
7.97% .
6.79% processor.h
3.69% fair.c
2.74% gmain.c
1.73% gtype.c
1.40% core.c
1.23% atomic64_64
1.22% qspinlock.h
1.16% i915_gem.c
1.09% ghash.c
1.07% gthread-pos
1.06% bitops.h
1.06% malloc.c
1.06% af_unix.c
1.03% select.c
0.99% gslice.c
0.80% hrtimer.c
0.78% socket.c
0.67% paravirt.h
0.66% entry_64.S
0.62% hooks.c
0.61% gsignal.c
0.57% intel_ringb
0.55% timekeeping
0.54% slub.c
0.53% timerqueue.
0.52% menu.c
0.52% file.c
0.50% tree.c
0.47% compiler.h
0.46% tick-sched.
0.46% msr.h
0.39% busy_poll.h
0.39% int_sqrt.c
0.36% atomic.h
0.35% gobject.c
0.35% read_write.
0.33% garray.c
0.32% xcb_io.c
0.32% wait.c
0.31% copy_user_6
0.30% dbus-marsha
0.29% rbtree.c
0.29% skbuff.c
0.28% dbus-marsha
0.27% spinlock.c
0.27% iov_iter.c
0.26% avc.c
0.26% list_debug.
0.26% giochannel.
0.24% cairo-gstat
0.24% dbus-string
0.23% cpuidle.c
0.23% page_alloc.
0.22% tsc.c
0.21% gvariant-co
0.20% process_64.
0.19% gbsearcharr
0.18% sock.c
0.18% timer.c
0.18% dbus-connec
0.18% dbus-marsha
0.17% fsnotify.c
0.17% iomap.c
0.17% gutf8.c
0.17% dbus-list.c
0.16% cairo-path-
0.16% gbitlock.c
0.16% srcu.c
0.16% gdataset.c
0.16% intel_uncor
0.16% gvarianttyp
0.15% futex.c
0.15% sched.h
0.15% gslist.c
0.15% clock.c
0.15% XlibInt.c
0.15% gclosure.c
0.14% gmem.c
0.14% n_tty.c
0.14% eventfd.c
0.14% datagram.c
0.14% security.c
0.14% mutex_64.h
0.14% sock.h
0.14% dbus-messag
0.14% idle.c
0.13% internal.h
0.13% page_64.h
0.12% current.h
0.12% vmalloc.c
0.12% notifier.c
0.12% gvariant.c
0.12% gdbusprivat
0.11% i915_drv.h
0.11% intel_idle.
0.11% rbtree_augm
0.11% mutex.c
0.11% gsimpleasyn
0.11% cairo-regio
0.10% cairo-hash.
0.10% gqueue.c
0.10% input.c
0.10% preempt.h
0.10% mm.h
0.10% cairo-clip-
0.10% dbus-mempoo
0.09% idle_task.c
0.09% rt.c
0.09% dbus-mainlo
0.09% tick-onesho
0.09% dbus-transp
0.09% softirq.c
0.09% cairo-xlib-
0.09% workqueue.c
0.09% cairo-matri
0.08% dbus-marsha
0.08% dbus-resour
0.08% seqlock.h
0.08% gvaluetypes
0.08% cairo-surfa
0.08% glist.c
0.08% connection.
0.08% giounix.c
0.08% tty_io.c
0.08% stats.h
0.08% i915_cmd_pa
0.08% cairo.c
0.08% gvalue.c
0.07% mmzone.h
0.07% apic.c
0.07% gvarianttyp
0.07% gdbusmessag
0.07% gdbusconnec
0.07% ktime.h
0.07% string3.h
0.07% mempolicy.c
0.07% list.h
0.07% desc.h
0.07% mmap.c
0.06% loadavg.c
0.06% tree_plugin
0.06% gstring.c
0.06% cputime.c
0.06% cairo-scale
0.06% perf_event_
0.06% mmu_context
0.06% FilterEv.c
0.06% signal.c
0.06% eventpoll.c
0.06% gparam.c
0.06% locking.c
0.06% garbage.c
0.06% uaccess.h
0.06% irq_work.c
0.06% namei.c
0.06% Pending.c
0.06% cairo-surfa
0.06% i915_gem_ex
0.06% nmi.c
0.06% dbus-timeou
0.06% avtab.c
0.05% fdtable.h
0.05% pango-layou
0.05% drm_crtc.c
0.05% cgroup.h
0.05% gquark.c
0.05% kthread.c
0.05% i915_gem_co
0.05% driver.c
0.05% i915_irq.c
0.05% rx.c
0.05% smp.c
0.05% scm.h
0.05% dcache.c
0.05% cairo-color
0.04% cairo-bentl
0.04% skbuff.h
0.04% gapplicatio
0.04% clockevents
0.04% file.h
0.04% drm_fops.c
0.04% slab_common
0.04% cairo-boxes
0.04% gwakeup.c
0.04% vmstat.c
0.04% crc32.c
0.04% find_bit.c
0.04% dbus-socket
0.04% qspinlock.c
0.04% memcpy_64.S
0.04% dbus-sysdep
0.04% vsprintf.c
0.04% file_table.
0.04% poll.h
0.04% cpuacct.c
0.04% stop_task.c
0.04% x86.c
0.04% fsnotify.h
0.04% policy.c
0.04% watchdog.c
0.04% math64.h
0.04% gsocket.c
0.04% gdbusinterf
0.04% ptthread.c
0.04% mmzone.c
0.04% time.c
0.04% cairoint.h
0.04% gdbusintros
0.03% s_sin-avx.c
0.03% spinlock.h
0.03% guniprop.c
0.03% unistd.h
0.03% gmarshal.c
0.03% cairo-traps
0.03% radix-tree.
0.03% signals.c
0.03% main.c
0.03% dbus-creden
0.03% hcd.c
0.03% gvariant-se
0.03% kernel_stat
0.03% bus.c
0.03% cairo-defau
0.03% cairo-freed
0.03% CrGC.c
0.03% posix-cpu-t
0.03% tty_ldsem.c
0.03% NextEvent.c
0.03% pty.c
0.03% JSAutoCompa
0.03% gprintf.c
0.03% poll2.h
0.03% context.h
0.03% pangofc-fon
0.03% af_netlink.
0.03% vsyscall_gt
0.03% hrtimer.h
0.03% printk.c
0.03% cairo-xlib-
0.03% seq_file.c
0.03% cairo-patte
0.03% process.c
0.03% deadline.c
0.03% intel_displ
0.03% jiffies.h
0.03% dl-lookup.c
0.03% flex_array.
0.03% services.c
0.02% prtpd.c
0.02% cairo-devic
0.02% dnotify.c
0.02% scan.c
0.02% cairo-clip.
0.02% profile.c
0.02% cairo-array
0.02% blkdev.h
0.02% uprobes.c
0.02% gfp.h
0.02% cairo-traps
0.02% cairo-strok
0.02% FillRect.c
0.02% mutex.h
0.02% ghook.c
0.02% dbus-transp
0.02% perf_event.
0.02% mprotect.c
0.02% pango-impl-
0.02% journal.c
0.02% rbtree.h
0.02% namespace.c
0.02% mwait.h
0.02% memset_64.S
0.02% gboxed.c
0.02% handle.c
0.02% Xrender.c
0.02% gthread.c
0.02% drm_irq.c
0.02% memory.c
0.02% ehci-hcd.c
0.02% dbus-errors
0.02% tlb.c
0.02% tsacct.c
0.02% hid-apple.c
0.02% qos.c
0.02% open.c
0.02% ErrHndlr.c
0.02% cpumask.h
0.02% gioerror.c
0.02% urb.c
0.02% pid.c
0.02% blk-core.c
0.02% cairo-bentl
0.02% wait.h
0.02% irq_regs.h
0.02% dma.c
0.02% topology.h
0.02% FreePix.c
0.02% selinux.c
0.02% intel_pstat
0.02% cairo-slope
0.02% pipe.c
0.02% fcntl.c
0.02% ntp.c
0.02% pangocairo-
0.02% atkplug.c
0.02% cfq-iosched
0.02% shmem.c
0.02% basic-fc.c
0.02% seccomp.c
0.02% ptrace.h
0.02% drm_gem.c
0.02% commoncap.c
0.02% cpumask.c
0.02% gtimer.c
0.02% dbus-thread
0.01% base.c
0.01% tty_buffer.
0.01% sysrq.c
0.01% iommu-helpe
0.01% cairo-boxes
0.01% drm_modeset
0.01% gdbusproxy.
0.01% atkobject.c
0.01% pgtable.c
0.01% ffi64.c
0.01% mac80211_if
0.01% cairo-gstat
0.01% cairo-arc.c
0.01% cairo-ft-fo
0.01% gbytes.c
0.01% hid-core.c
0.01% tick-broadc
0.01% scsi.c
0.01% dbus-datasl
0.01% page-io.c
0.01% intel_front
0.01% swap.c
0.01% XlibAsync.c
0.01% intel_pm.c
0.01% intel_ringb
0.01% sn-list.c
0.01% locks.c
0.01% putuser.S
0.01% dbus-signat
0.01% cairo-xlib-
0.01% cairo-misc.
0.01% highmem.h
0.01% blk-integri
0.01% random.c
0.01% cairo-recta
0.01% memmove_64.
0.01% gdk-pixbuf.
0.01% Picture.c
0.01% pango-attri
0.01% usercopy_64
0.01% hid-input.c
0.01% tty_ldisc.c
0.01% perf_event_
0.01% blk-throttl
0.01% time.h
0.01% gsourceclos
0.01% physaddr.c
0.01% drm_ioctl.c
0.01% pango-conte
0.01% drm_modes.c
0.01% avc.h
0.01% qrwlock.h
0.01% apic.h
0.01% spinlock_ap
0.01% crc16.c
0.01% i915_drv.c
0.01% cpufeature.
0.01% dbus-sysdep
0.01% mount.h
0.01% jump_label.
0.01% revoke.c
0.01% blk.h
0.01% signal.h
0.01% ehci-q.c
0.01% tlbflush.h
0.01% list.cc
0.01% dbus-gmain.
0.01% scsi_lib.c
0.01% dbus-hash.c
0.01% GetEventDat
0.01% mlme.c
0.01% iface.c
0.01% io.h
0.01% dbus-watch.
0.01% page-flags.
0.01% sn-monitor.
0.01% task_work.c
0.01% evdev.c
0.01% cairo-obser
0.01% Macros.c
0.01% drm_atomic_
0.01% dbus-socket
0.01% itimer.c
0.01% gtestutils.
0.01% spurious.c
0.01% super.c
0.01% inode.c
0.01% rtnetlink.c
0.01% physaddr.h
0.01% unix64.S Quoted 8 lines · Arnaldo Carvalho de Melo
Em Fri, Aug 07, 2015 at 08:51:45PM -0300, Arnaldo Carvalho de Melo escreveu:
Quoted 7 lines
Em Fri, Aug 07, 2015 at 03:54:24PM -0700, Andi Kleen escreveu:
Quoted 6 lines
From: Andi Kleen <[email protected]>
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.
Quoted 8 lines
Quoted 4 lines
Quoted 2 lines
Add a srcfile sort key to perf report. It builds on top of the existing srcline support.
Applied
Humm, holding this up a bit, further testing showed some oddities, fedora21, the width of the column is being limited to the lenght of the header
Yes I've seen that, I just use -w normally. It also happens with --sort srcline. The column sizing code could probably be somewhat smarter and always allow the last column to become as wide as needed. But that's something that should be done separately; I don't think it belongs into this patch.
Quoted 1 line
and there are some DWARF errors, have you noticed those?
No I didn't. Some generic issue, likely happening with srcline and perhaps even objdump -S too. Find out with strace what file it is.
-Andi
Arnaldo Carvalho de Melo
maintainerEm Sat, Aug 08, 2015 at 04:27:35AM +0200, Andi Kleen escreveu:
Quoted 14 lines · Arnaldo Carvalho de Melo
Quoted 13 lines
Em Fri, Aug 07, 2015 at 08:51:45PM -0300, Arnaldo Carvalho de Melo escreveu:
Quoted 11 lines
Em Fri, Aug 07, 2015 at 03:54:24PM -0700, Andi Kleen escreveu:
Quoted 8 lines
From: Andi Kleen <[email protected]>
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.
Applied
Quoted 7 lines
Quoted 3 lines
Humm, holding this up a bit, further testing showed some oddities, fedora21, the width of the column is being limited to the lenght of the header
Yes I've seen that, I just use -w normally. It also happens with --sort srcline. The column sizing code could probably be somewhat smarter and always allow the last column to become as wide as needed. But that's
Right, I'll check that, there is code for figuring that out, I'll see what is missing.
Quoted 2 lines
something that should be done separately; I don't think it belongs into this patch.
I can agree with that, will add it, as it provides value as-is.
Quoted 4 lines
Quoted 1 line
and there are some DWARF errors, have you noticed those?
No I didn't. Some generic issue, likely happening with srcline and perhaps even objdump -S too. Find out with strace what file it is.
Ok.
- Arnaldo
Namhyung Kim
maintainerHi Andi,
Quoted 18 lines · Andi Kleen
Quoted 8 lines · Arnaldo Carvalho de Melo
Em Fri, Aug 07, 2015 at 08:51:45PM -0300, Arnaldo Carvalho de Melo escreveu:
Quoted 7 lines
Em Fri, Aug 07, 2015 at 03:54:24PM -0700, Andi Kleen escreveu:
Quoted 6 lines
From: Andi Kleen <[email protected]>
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.
Quoted 8 lines
Quoted 4 lines
Quoted 2 lines
Add a srcfile sort key to perf report. It builds on top of the existing srcline support.
Applied
Humm, holding this up a bit, further testing showed some oddities, fedora21, the width of the column is being limited to the lenght of the header
Quoted 5 lines
Yes I've seen that, I just use -w normally. It also happens with --sort srcline. The column sizing code could probably be somewhat smarter and always allow the last column to become as wide as needed. But that's something that should be done separately; I don't think it belongs into this patch.
Maybe something like this?
Quoted 1 line
From
c052d17ae45ac08a381192191473ed3c4308c0c5Mon Sep 17 00:00:00 2001
From: Namhyung Kim <[email protected]>
Date: Sun, 9 Aug 2015 12:28:01 +0900
Subject: [PATCH] perf tools: Update srcline column length
It didn't calculate the column length so the default of header length used only. Update the length when we get the srcline actually.
--- tools/perf/util/sort.c | 4 ++++ 1 file changed, 4 insertions(+)
tools/perf/util/sort.c +4 -0
@@ -291,6 +291,8 @@ sort__srcline_cmp(struct hist_entry *left, struct hist_entry *right) left->srcline =get_srcline (map->dso,map__rip_2objdump (map, left->ip), left->ms.sym, true); +hists__new_col_len (left->hists,HISTC_SRCLINE , +strlen (left->srcline)); } }if (!right->srcline) {
@@ -301,6 +303,8 @@ sort__srcline_cmp(struct hist_entry *left, struct hist_entry *right) right->srcline =get_srcline (map->dso,map__rip_2objdump (map, right->ip), right->ms.sym, true); +hists__new_col_len (right->hists,HISTC_SRCLINE , +strlen (right->srcline)); } }return strcmp (right->srcline, left->srcline);
Signature · 1 line
2.5.0
Quoted 21 lines · Namhyung Kim
Hi Andi,
Quoted 18 lines · Andi Kleen
Quoted 16 lines · Arnaldo Carvalho de Melo
Em Fri, Aug 07, 2015 at 08:51:45PM -0300, Arnaldo Carvalho de Melo escreveu:
Quoted 11 lines
Em Fri, Aug 07, 2015 at 03:54:24PM -0700, Andi Kleen escreveu:
Quoted 8 lines
From: Andi Kleen <[email protected]>
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.
Applied
Humm, holding this up a bit, further testing showed some oddities, fedora21, the width of the column is being limited to the lenght of the header
Quoted 7 lines
Quoted 5 lines
Yes I've seen that, I just use -w normally. It also happens with --sort srcline. The column sizing code could probably be somewhat smarter and always allow the last column to become as wide as needed. But that's something that should be done separately; I don't think it belongs into this patch.
Maybe something like this?
Thanks. Should do the same thing for srcfile.
One potential issue is that the srclines/files can be potentially very long, so it may be needed to give them a limit if they're not the last column.
-Andi
Em Sun, Aug 09, 2015 at 06:03:35PM +0200, Andi Kleen escreveu:
Quoted 28 lines · Namhyung Kim
Quoted 27 lines
Hi Andi,
Quoted 23 lines · Andi Kleen
Quoted 8 lines · Arnaldo Carvalho de Melo
Em Fri, Aug 07, 2015 at 08:51:45PM -0300, Arnaldo Carvalho de Melo escreveu:
Quoted 7 lines
Em Fri, Aug 07, 2015 at 03:54:24PM -0700, Andi Kleen escreveu:
Quoted 6 lines
From: Andi Kleen <[email protected]>
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.
Quoted 8 lines
Quoted 4 lines
Quoted 2 lines
Add a srcfile sort key to perf report. It builds on top of the existing srcline support.
Applied
Humm, holding this up a bit, further testing showed some oddities, fedora21, the width of the column is being limited to the lenght of the header
Yes I've seen that, I just use -w normally. It also happens with --sort srcline. The column sizing code could probably be somewhat smarter and always allow the last column to become as wide as needed. But that's something that should be done separately; I don't think it belongs into this patch.
Quoted 3 lines
Quoted 1 line
Maybe something like this?
Thanks. Should do the same thing for srcfile.
Thanks, applied, with Ack from Andi, will fix this with a commiter note on Andi's patch for srcfile when applying it, now.
Quoted 3 lines
One potential issue is that the srclines/files can be potentially very long, so it may be needed to give them a limit if they're not the last column.
That can come later.
- Arnaldo
Arnaldo Carvalho de Melo
maintainerEm Sun, Aug 09, 2015 at 12:30:49PM +0900, Namhyung Kim escreveu:
Quoted 21 lines
Hi Andi,
Quoted 18 lines · Andi Kleen
Quoted 16 lines · Arnaldo Carvalho de Melo
Em Fri, Aug 07, 2015 at 08:51:45PM -0300, Arnaldo Carvalho de Melo escreveu:
Quoted 11 lines
Em Fri, Aug 07, 2015 at 03:54:24PM -0700, Andi Kleen escreveu:
Quoted 8 lines
From: Andi Kleen <[email protected]>
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.
Applied
Humm, holding this up a bit, further testing showed some oddities, fedora21, the width of the column is being limited to the lenght of the header
Quoted 7 lines
Quoted 5 lines
Yes I've seen that, I just use -w normally. It also happens with --sort srcline. The column sizing code could probably be somewhat smarter and always allow the last column to become as wide as needed. But that's something that should be done separately; I don't think it belongs into this patch.
Maybe something like this?
Maybe, but it crashes when right->hists == NULL, perhaps that is some sort of placeholder? See the value of right (0x7fffffffc0a0)? looks like some static guard for something, will continue after lunch...
(gdb) fr 2
#2 0x00000000004bb812 in sort__srcline_cmp (left=0x1bfa090, right=0x7fffffffc0a0) at util/sort.c:306
306 hists__new_col_len(right->hists, HISTC_SRCLINE,
(gdb) p right->hists
$7 = (struct hists *) 0x0
(gdb) p left->hists
$8 = (struct hists *) 0x18d66e0
(gdb)
Due to this both "-s srcline" and "-s srcfile" crashes here.
- Arnaldo
Quoted 40 lines
Quoted 3 lines
From
c052d17ae45ac08a381192191473ed3c4308c0c5Mon Sep 17 00:00:00 2001From: Namhyung Kim <[email protected]>
Date: Sun, 9 Aug 2015 12:28:01 +0900
Subject: [PATCH] perf tools: Update srcline column lengthIt didn't calculate the column length so the default of header length used only. Update the length when we get the srcline actually.
Cc Andi Kleen <[email protected]>Signed-off-by Namhyung Kim <[email protected]>--- tools/perf/util/sort.c | 4 ++++ 1 file changed, 4 insertions(+)tools/perf/util/sort.c +4
-0@@ -291,6 +291,8 @@ sort__srcline_cmp(struct hist_entry *left, struct hist_entry *right) left->srcline =get_srcline (map->dso,map__rip_2objdump (map, left->ip), left->ms.sym, true); +hists__new_col_len (left->hists,HISTC_SRCLINE , +strlen (left->srcline)); } }if (!right->srcline) {@@ -301,6 +303,8 @@ sort__srcline_cmp(struct hist_entry *left, struct hist_entry *right) right->srcline =get_srcline (map->dso,map__rip_2objdump (map, right->ip), right->ms.sym, true); +hists__new_col_len (right->hists,HISTC_SRCLINE , +strlen (right->srcline)); } }return strcmp (right->srcline, left->srcline);Signature · 1 line
2.5.0
Arnaldo Carvalho de Melo
maintainerEm Mon, Aug 10, 2015 at 01:12:41PM -0300, Arnaldo Carvalho de Melo escreveu:
Quoted 43 lines
Em Sun, Aug 09, 2015 at 12:30:49PM +0900, Namhyung Kim escreveu:
Quoted 21 lines
Hi Andi,
Quoted 18 lines · Andi Kleen
Quoted 16 lines · Arnaldo Carvalho de Melo
Em Fri, Aug 07, 2015 at 08:51:45PM -0300, Arnaldo Carvalho de Melo escreveu:
Quoted 11 lines
Em Fri, Aug 07, 2015 at 03:54:24PM -0700, Andi Kleen escreveu:
Quoted 8 lines
From: Andi Kleen <[email protected]>
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.
Applied
Humm, holding this up a bit, further testing showed some oddities, fedora21, the width of the column is being limited to the lenght of the header
Quoted 7 lines
Quoted 5 lines
Yes I've seen that, I just use -w normally. It also happens with --sort srcline. The column sizing code could probably be somewhat smarter and always allow the last column to become as wide as needed. But that's something that should be done separately; I don't think it belongs into this patch.
Maybe something like this?
Maybe, but it crashes when
right->hists== NULL, perhaps that is some sort of placeholder? See the value of right (0x7fffffffc0a0)? looks like some static guard for something, will continue after lunch...(gdb) fr 2
#20x00000000004bb812insort__srcline_cmp(left=0x1bfa090, right=0x7fffffffc0a0) atutil/sort.c:306
306 hists__new_col_len(right->hists,HISTC_SRCLINE,
(gdb) pright->hists
$7 = (struct hists *)0x0
(gdb) pleft->hists
$8 = (struct hists *)0x18d66e0
(gdb)
Quoted 3 lines
Due to this both "-s srcline" and "-s srcfile" crashes here.
- Arnaldo
Ok, this one needs to be applied before yours and Andi's:
tools/perf/util/hist.c +1 -0
@@ -761,6 +761,7 @@ iter_add_next_cumulative_entry(struct hist_entry_iter *iter,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),
---------
Because you will use that stack synthesized he_tmp entry as a parameter to the cmp() function, that will end up trying to access he_tmp->hists, that was NULL, b00m.
- Arnaldo
Quoted 40 lines
Quoted 40 lines
Quoted 3 lines
From
c052d17ae45ac08a381192191473ed3c4308c0c5Mon Sep 17 00:00:00 2001From: Namhyung Kim <[email protected]>
Date: Sun, 9 Aug 2015 12:28:01 +0900
Subject: [PATCH] perf tools: Update srcline column lengthIt didn't calculate the column length so the default of header length used only. Update the length when we get the srcline actually.
Cc Andi Kleen <[email protected]>Signed-off-by Namhyung Kim <[email protected]>--- tools/perf/util/sort.c | 4 ++++ 1 file changed, 4 insertions(+)tools/perf/util/sort.c +4
-0@@ -291,6 +291,8 @@ sort__srcline_cmp(struct hist_entry *left, struct hist_entry *right) left->srcline =get_srcline (map->dso,map__rip_2objdump (map, left->ip), left->ms.sym, true); +hists__new_col_len (left->hists,HISTC_SRCLINE , +strlen (left->srcline)); } }if (!right->srcline) {@@ -301,6 +303,8 @@ sort__srcline_cmp(struct hist_entry *left, struct hist_entry *right) right->srcline =get_srcline (map->dso,map__rip_2objdump (map, right->ip), right->ms.sym, true); +hists__new_col_len (right->hists,HISTC_SRCLINE , +strlen (right->srcline)); } }return strcmp (right->srcline, left->srcline);Signature · 1 line
2.5.0
Arnaldo Carvalho de Melo
maintainerEm Mon, Aug 10, 2015 at 03:32:47PM -0300, Arnaldo Carvalho de Melo escreveu:
Quoted 22 lines
Em Mon, Aug 10, 2015 at 01:12:41PM -0300, Arnaldo Carvalho de Melo escreveu:
Quoted 9 lines
(gdb) fr 2
#20x00000000004bb812insort__srcline_cmp(left=0x1bfa090, right=0x7fffffffc0a0) atutil/sort.c:306
306 hists__new_col_len(right->hists,HISTC_SRCLINE,
(gdb) pright->hists
$7 = (struct hists *)0x0
(gdb) pleft->hists
$8 = (struct hists *)0x18d66e0Due to this both "-s srcline" and "-s srcfile" crashes here.
Ok, this one needs to be applied before yours and Andi's:
+++
b/tools/perf/util/hist.ctools/perf/util/hist.c +1
-1@@ -761,6 +761,7 @@ iter_add_next_cumulative_entry(struct hist_entry_iter *iter,struct hist_entry **he_cache = iter->priv;struct hist_entry *he;struct hist_entry he_tmp = { + .hists =evsel__hists (evsel), .cpu = al->cpu, ---------
Quoted 3 lines
Because you will use that stack synthesized
he_tmpentry as a parameter to thecmp()function, that will end up trying to accesshe_tmp->hists, that was NULL, b00m.
all this is moot, as we need to update that column length elsewhere, together with all the other fileds, please see:
://git.kernel.org/cgit/linux/kernel/git/acme/linux.git/commit/?h=perf/core&id=e8e6d37e73e6b950c891c780745460b87f4755b6
And now, with it really updating the width, we get to another problem, in one case I got a line like:
0.00% JS_GetInstancePrivate(JSContext*, JSObject*, JSClass*, JS /usr/lib64/libmozjs-24.so 0x1a0bd1 d [.] JS_GetInstancePrivate(JSContext*, JSObject*, JSClass*, JS::Value*)
Which is really strange, i.e. it mixed up demangled stuff with the srcfile, i.e. normal lines for:
perf report -v -s srcfile,dso,symbol --stdio
Look like:
1.94% malloc.c /usr/lib64/libc-2.20.so 0x80f91 d [.] _int_malloc
Anyway, will continue this later, together with checking why it doesn't resolve many things, that one may be related, in some cases, to functions in assembly source files, for now it updates correctly the column width for "srcline" and "srcfile".
- Arnaldo
Quoted 42 lines
- Arnaldo
Quoted 40 lines
Quoted 40 lines
Quoted 3 lines
From
c052d17ae45ac08a381192191473ed3c4308c0c5Mon Sep 17 00:00:00 2001From: Namhyung Kim <[email protected]>
Date: Sun, 9 Aug 2015 12:28:01 +0900
Subject: [PATCH] perf tools: Update srcline column lengthIt didn't calculate the column length so the default of header length used only. Update the length when we get the srcline actually.
Cc Andi Kleen <[email protected]>Signed-off-by Namhyung Kim <[email protected]>--- tools/perf/util/sort.c | 4 ++++ 1 file changed, 4 insertions(+)tools/perf/util/sort.c +4
-0@@ -291,6 +291,8 @@ sort__srcline_cmp(struct hist_entry *left, struct hist_entry *right) left->srcline =get_srcline (map->dso,map__rip_2objdump (map, left->ip), left->ms.sym, true); +hists__new_col_len (left->hists,HISTC_SRCLINE , +strlen (left->srcline)); } }if (!right->srcline) {@@ -301,6 +303,8 @@ sort__srcline_cmp(struct hist_entry *left, struct hist_entry *right) right->srcline =get_srcline (map->dso,map__rip_2objdump (map, right->ip), right->ms.sym, true); +hists__new_col_len (right->hists,HISTC_SRCLINE , +strlen (right->srcline)); } }return strcmp (right->srcline, left->srcline);Signature · 1 line
2.5.0
Namhyung Kim
maintainerQuoted 23 lines · Arnaldo Carvalho de Melo
Em Mon, Aug 10, 2015 at 03:32:47PM -0300, Arnaldo Carvalho de Melo escreveu:
Quoted 21 lines
Em Mon, Aug 10, 2015 at 01:12:41PM -0300, Arnaldo Carvalho de Melo escreveu:
Quoted 9 lines
(gdb) fr 2
#20x00000000004bb812insort__srcline_cmp(left=0x1bfa090, right=0x7fffffffc0a0) atutil/sort.c:306
306 hists__new_col_len(right->hists,HISTC_SRCLINE,
(gdb) pright->hists
$7 = (struct hists *)0x0
(gdb) pleft->hists
$8 = (struct hists *)0x18d66e0Due to this both "-s srcline" and "-s srcfile" crashes here.
Ok, this one needs to be applied before yours and Andi's:
+++
b/tools/perf/util/hist.ctools/perf/util/hist.c +1
-1@@ -761,6 +761,7 @@ iter_add_next_cumulative_entry(struct hist_entry_iter *iter,struct hist_entry **he_cache = iter->priv;struct hist_entry *he;struct hist_entry he_tmp = { + .hists =evsel__hists (evsel), .cpu = al->cpu, ---------
Quoted 8 lines
Quoted 3 lines
Because you will use that stack synthesized
he_tmpentry as a parameter to thecmp()function, that will end up trying to accesshe_tmp->hists, that was NULL, b00m.all this is moot, as we need to update that column length elsewhere, together with all the other fileds, please see:
://git.kernel.org/cgit/linux/kernel/git/acme/linux.git/commit/?h=perf/core&id=e8e6d37e73e6b950c891c780745460b87f4755b6
Looks good to me.
Quoted 5 lines
And now, with it really updating the width, we get to another problem, in one case I got a line like:
0.00% JS_GetInstancePrivate(JSContext*, JSObject*, JSClass*, JS
/usr/lib64/libmozjs-24.so0x1a0bd1d [.]JS_GetInstancePrivate(JSContext*, JSObject*, JSClass*, JS::Value*)
It seems like ':' character in the symbol name affects the result for some reason.
Thanks,
Namhyung
Quoted 16 lines
Which is really strange, i.e. it mixed up demangled stuff with the srcfile, i.e. normal lines for:
perf report -v -s srcfile,dso,symbol --stdio
Look like:
1.94% malloc.c /usr/lib64/libc-2.20.so 0x80f91 d [.] _int_mallocAnyway, will continue this later, together with checking why it doesn't resolve many things, that one may be related, in some cases, to functions in assembly source files, for now it updates correctly the column width for "srcline" and "srcfile".
- Arnaldo
tip-bot for Andi Kleen
merge bot[tip:perf/core] perf report: Add support for srcfile sort key
In reply to Andi Kleenperf report: Add support for srcfile sort key
Commit-ID: 31191a85fb875cf123cea56bbfd34f4b941f3c79 Gitweb: http://git.kernel.org/tip/31191a85fb875cf123cea56bbfd34f4b941f3c79 Author: Andi Kleen <[email protected]> AuthorDate: Fri, 7 Aug 2015 15:54:24 -0700 Committer: Arnaldo Carvalho de Melo <[email protected]> CommitDate: Mon, 10 Aug 2015 17:20:25 -0300
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.
--- tools/perf/Documentation/perf-report.txt | 2 ++ tools/perf/util/hist.c | 5 +++ tools/perf/util/hist.h | 1 + tools/perf/util/sort.c | 52 ++++++++++++++++++++++++++++++++ tools/perf/util/sort.h | 2 ++ 5 files changed, 62 insertions(+)
tools/perf/Documentation/perf-report.txt +2 -0
@@ -81,6 +81,8 @@ OPTIONS - cpu: cpu number the task ran at the time of sample - srcline: filename and line number executed at the time of sample. The DWARF debugging info must be provided. + - srcfile: file name of the source file of the same. Requires dwarf + information. - weight: Event specific weight, e.g. memory latency or transaction abort cost. This is the global weight. - local_weight: Local weight version of the weight above.
tools/perf/util/hist.c +5 -0
@@ -154,6 +154,9 @@ void hists__calc_col_len(struct hists *hists, struct hist_entry *h)if (h->srcline)hists__new_col_len (hists,HISTC_SRCLINE ,strlen (h->srcline)); +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 ());
@@ -949,6 +952,8 @@ void hist_entry__delete(struct hist_entry *he)zfree (&he->stat_acc);free_srcline (he->srcline); +if (he->srcfile && he->srcfile[0 ]) +free (he->srcfile);free_callchain (he->callchain);free (he); }
tools/perf/util/hist.h +1 -0
@@ -30,6 +30,7 @@ enum hist_column {HISTC_PARENT ,HISTC_CPU ,HISTC_SRCLINE , +HISTC_SRCFILE ,HISTC_MISPREDICT ,HISTC_IN_TX , HISTC_ABORT,
tools/perf/util/sort.c +52 -0
64 lines not on this page · raw message
tools/perf/util/sort.h +2 -0
@@ -114,6 +114,7 @@ struct hist_entry { }; };char *srcline; +char *srcfile;struct symbol *parent;struct rb_root sorted_chain;struct branch_info *branch_info;
@@ -172,6 +173,7 @@ enum sort_type {SORT_PARENT ,SORT_CPU ,SORT_SRCLINE , +SORT_SRCFILE ,SORT_LOCAL_WEIGHT ,SORT_GLOBAL_WEIGHT , SORT_TRANSACTION,