Skip to contentSkip to conversation

perf top -g -U --sort=symbol --children == lalalalala?

DISCUSSION Mike Galbraith 4 participants 11 messages Sep 9, 2014 – Sep 12, 2014

lore
Share & tools
lore
#0 lore raw

Seems the now default on --children thingy doesn't like -U much.

Samples: 5K of event 'cycles', Event count (approx.): 2268660922

  Children      Self  Symbol
+   46.42%     0.04%  [k] system_call_fastpath
+   27.12%     0.80%  [k] apic_timer_interrupt
+   26.26%     0.23%  [k] smp_apic_timer_interrupt
+   20.43%     0.13%  [k] local_apic_timer_interrupt
+   20.20%     0.74%  [k] hrtimer_interrupt
+   18.01%     0.39%  [k] __run_hrtimer
+   14.43%     0.35%  [k] tick_sched_timer
+   13.60%     0.05%  [k] sys_poll
+   13.09%     1.16%  [k] do_sys_poll
+   12.16%     0.08%  [k] tick_sched_handle.isra.16
+   11.70%     0.32%  [k] update_process_times
+   11.41%     0.03%  [k] sys_read
+   11.17%     0.11%  [k] vfs_read
+   10.85%     0.09%  [k] seq_read
+    8.27%     0.09%  [k] proc_single_show
+    7.79%     1.19%  [k] page_fault
+    7.01%     0.25%  [k] scheduler_tick
+    6.26%     0.11%  [k] __do_page_fault
+    5.97%     1.33%  [k] do_task_stat
+    5.91%     0.23%  [k] handle_mm_fault
+    5.40%     0.20%  [k] do_sys_open
+    5.02%     0.09%  [k] ret_from_intr
+    4.94%     0.05%  [k] do_IRQ
+    4.70%     0.14%  [k] path_openat
+    4.69%     0.62%  [k] tty_poll
+    4.69%     0.16%  [k] irq_exit
+    4.49%     0.63%  [k] __do_softirq
+    4.27%     0.12%  [k] __alloc_pages_nodemask
+    3.95%     0.03%  [k] handle_irq_event_percpu
+    3.51%     0.71%  [k] __fget_light
+    3.47%     0.04%  [k] sys_select
+    3.39%     0.08%  [k] seq_printf
+    3.37%     0.07%  [k] sys_open
+    3.34%     0.11%  [k] seq_vprintf
+    3.22%     0.02%  [k] core_sys_select
+    3.15%     0.26%  [k] do_select
+    3.15%     3.15%  [k] _raw_spin_lock
#1 lore raw

Em Tue, Sep 09, 2014 at 01:54:01PM +0200, Mike Galbraith escreveu:

Quoted 1 line

Seems the now default on --children thingy doesn't like -U much.

Namhyung, can you please take a look at this?

- Arnaldo

Quoted 41 lines

Samples: 5K of event 'cycles', Event count (approx.): 2268660922
Children Self Symbol

(quoted diff) +37 -0
+   46.42%     0.04%  [k] system_call_fastpath
+   27.12%     0.80%  [k] apic_timer_interrupt
+   26.26%     0.23%  [k] smp_apic_timer_interrupt
+   20.43%     0.13%  [k] local_apic_timer_interrupt
+   20.20%     0.74%  [k] hrtimer_interrupt
+   18.01%     0.39%  [k] __run_hrtimer
+   14.43%     0.35%  [k] tick_sched_timer
+   13.60%     0.05%  [k] sys_poll
+   13.09%     1.16%  [k] do_sys_poll
+   12.16%     0.08%  [k] tick_sched_handle.isra.16
+   11.70%     0.32%  [k] update_process_times
+   11.41%     0.03%  [k] sys_read
+   11.17%     0.11%  [k] vfs_read
+   10.85%     0.09%  [k] seq_read
+    8.27%     0.09%  [k] proc_single_show
+    7.79%     1.19%  [k] page_fault
+    7.01%     0.25%  [k] scheduler_tick
+    6.26%     0.11%  [k] __do_page_fault
+    5.97%     1.33%  [k] do_task_stat
+    5.91%     0.23%  [k] handle_mm_fault
+    5.40%     0.20%  [k] do_sys_open
+    5.02%     0.09%  [k] ret_from_intr
+    4.94%     0.05%  [k] do_IRQ
+    4.70%     0.14%  [k] path_openat
+    4.69%     0.62%  [k] tty_poll
+    4.69%     0.16%  [k] irq_exit
+    4.49%     0.63%  [k] __do_softirq
+    4.27%     0.12%  [k] __alloc_pages_nodemask
+    3.95%     0.03%  [k] handle_irq_event_percpu
+    3.51%     0.71%  [k] __fget_light
+    3.47%     0.04%  [k] sys_select
+    3.39%     0.08%  [k] seq_printf
+    3.37%     0.07%  [k] sys_open
+    3.34%     0.11%  [k] seq_vprintf
+    3.22%     0.02%  [k] core_sys_select
+    3.15%     0.26%  [k] do_select
+    3.15%     3.15%  [k] _raw_spin_lock
#2 lore raw
Quoted 4 lines · Arnaldo Carvalho de Melo

Em Tue, Sep 09, 2014 at 01:54:01PM +0200, Mike Galbraith escreveu:

Quoted 1 line

Seems the now default on --children thingy doesn't like -U much.

Namhyung, can you please take a look at this?

So what is the problem here?

Quoted 42 lines
Quoted 42 lines

Samples: 5K of event 'cycles', Event count (approx.): 2268660922
Children Self Symbol

(quoted diff) +37 -0
+   46.42%     0.04%  [k] system_call_fastpath
+   27.12%     0.80%  [k] apic_timer_interrupt
+   26.26%     0.23%  [k] smp_apic_timer_interrupt
+   20.43%     0.13%  [k] local_apic_timer_interrupt
+   20.20%     0.74%  [k] hrtimer_interrupt
+   18.01%     0.39%  [k] __run_hrtimer
+   14.43%     0.35%  [k] tick_sched_timer
+   13.60%     0.05%  [k] sys_poll
+   13.09%     1.16%  [k] do_sys_poll
+   12.16%     0.08%  [k] tick_sched_handle.isra.16
+   11.70%     0.32%  [k] update_process_times
+   11.41%     0.03%  [k] sys_read
+   11.17%     0.11%  [k] vfs_read
+   10.85%     0.09%  [k] seq_read
+    8.27%     0.09%  [k] proc_single_show
+    7.79%     1.19%  [k] page_fault
+    7.01%     0.25%  [k] scheduler_tick
+    6.26%     0.11%  [k] __do_page_fault
+    5.97%     1.33%  [k] do_task_stat
+    5.91%     0.23%  [k] handle_mm_fault
+    5.40%     0.20%  [k] do_sys_open
+    5.02%     0.09%  [k] ret_from_intr
+    4.94%     0.05%  [k] do_IRQ
+    4.70%     0.14%  [k] path_openat
+    4.69%     0.62%  [k] tty_poll
+    4.69%     0.16%  [k] irq_exit
+    4.49%     0.63%  [k] __do_softirq
+    4.27%     0.12%  [k] __alloc_pages_nodemask
+    3.95%     0.03%  [k] handle_irq_event_percpu
+    3.51%     0.71%  [k] __fget_light
+    3.47%     0.04%  [k] sys_select
+    3.39%     0.08%  [k] seq_printf
+    3.37%     0.07%  [k] sys_open
+    3.34%     0.11%  [k] seq_vprintf
+    3.22%     0.02%  [k] core_sys_select
+    3.15%     0.26%  [k] do_select
+    3.15%     3.15%  [k] _raw_spin_lock
Signature · 2 lines
Thanks,
Namhyung Kim
#3 lore raw
Quoted 8 lines · Namhyung Kim
Quoted 4 lines · Arnaldo Carvalho de Melo

Em Tue, Sep 09, 2014 at 01:54:01PM +0200, Mike Galbraith escreveu:

Quoted 1 line

Seems the now default on --children thingy doesn't like -U much.

Namhyung, can you please take a look at this?

So what is the problem here?

Well, if you don't see anything wrong, I guess nothing at all.

Quoted 3 lines
Quoted 3 lines
Quoted 3 lines

Samples: 5K of event 'cycles', Event count (approx.): 2268660922
Children Self Symbol
+ 46.42% 0.04% [k] system_call_fastpath

I'll just turn it off until I figure out what cool stuff this is telling me. why that symbol becomes the number one hit, and why total% > 100.

To me, it looks like top smoked it's breakfast, went to lala land ;-)

-Mike

#4 lore raw

Em Thu, Sep 11, 2014 at 02:43:12PM +0200, Mike Galbraith escreveu:

Quoted 11 lines · Namhyung Kim
Quoted 8 lines
Quoted 4 lines · Arnaldo Carvalho de Melo

Em Tue, Sep 09, 2014 at 01:54:01PM +0200, Mike Galbraith escreveu:

Quoted 1 line

Seems the now default on --children thingy doesn't like -U much.

Namhyung, can you please take a look at this?

So what is the problem here?

Well, if you don't see anything wrong, I guess nothing at all.

:-)

I think that when we decide that it is so better to change defaults like we did this time, we should be required to add a big fat warning (a --tui popup, use the first lines on --stdio, etc) about why the default was changed and allow quick, easy opt out, restoring previous behaviour after the user, being warned, knows what to expect, tries it, and then is in a better position to decide if keeping the new default is what is desired.

- Arnaldo

Quoted 8 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines

Samples: 5K of event 'cycles', Event count (approx.): 2268660922
Children Self Symbol
+ 46.42% 0.04% [k] system_call_fastpath

I'll just turn it off until I figure out what cool stuff this is telling me. why that symbol becomes the number one hit, and why total% > 100.

To me, it looks like top smoked it's breakfast, went to lala land ;-)

Yeah, its confusing, I'll let Namhyung explain it ;-)

:-)

- Arnaldo

#5 lore raw

Em Thu, Sep 11, 2014 at 10:43:38AM -0300, Arnaldo Carvalho de Melo escreveu:

Quoted 15 lines

Em Thu, Sep 11, 2014 at 02:43:12PM +0200, Mike Galbraith escreveu:

Quoted 11 lines · Namhyung Kim
Quoted 8 lines
Quoted 4 lines · Arnaldo Carvalho de Melo

Em Tue, Sep 09, 2014 at 01:54:01PM +0200, Mike Galbraith escreveu:

Quoted 1 line

Seems the now default on --children thingy doesn't like -U much.

Namhyung, can you please take a look at this?

So what is the problem here?

Well, if you don't see anything wrong, I guess nothing at all.

:-)

Quoted 7 lines

I think that when we decide that it is so better to change defaults like we did this time, we should be required to add a big fat warning (a --tui popup, use the first lines on --stdio, etc) about why the default was changed and allow quick, easy opt out, restoring previous behaviour after the user, being warned, knows what to expect, tries it, and then is in a better position to decide if keeping the new default is what is desired.

Also, looking at the changelog entries and at tools/perf/Documentation/ the only description for --children, the default, is:

----------------
--children::
Accumulate callchain of children to parent entry so that then can show up in the output. The output will have a new "Children" column and will be sorted on the data. It requires callchains are recorded.
----------------

I think that a longer/clearer entry in the 'perf record' man page is required.

Perhaps the description got lost in a --cover-letter for the patch series implementing it?

- Arnaldo

Quoted 15 lines
Quoted 9 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines

Samples: 5K of event 'cycles', Event count (approx.): 2268660922
Children Self Symbol
+ 46.42% 0.04% [k] system_call_fastpath

I'll just turn it off until I figure out what cool stuff this is telling me. why that symbol becomes the number one hit, and why total% > 100.

To me, it looks like top smoked it's breakfast, went to lala land ;-)

Yeah, its confusing, I'll let Namhyung explain it ;-)

:-)

- Arnaldo

#6 lore raw
Quoted 9 lines · Arnaldo Carvalho de Melo

Also, looking at the changelog entries and at tools/perf/Documentation/ the only description for --children, the default, is:

----------------
--children::
Accumulate callchain of children to parent entry so that then can show up in the output. The output will have a new "Children" column and will be sorted on the data. It requires callchains are recorded.
----------------

grep of course found that, and git log found more, but nothing told me what the heck it's sweeping up that's so darn plentiful in my box that there's more than 100% of it laying about :)

Quoted 5 lines

I think that a longer/clearer entry in the 'perf record' man page is required.

Perhaps the description got lost in a --cover-letter for the patch series implementing it?

If it ever existed, I can't find it. A little blurb would be helpful.

-Mike

#7 lore raw
In reply to Arnaldo Carvalho de Melo

Hi Arnaldo and Mike,

Quoted 15 lines · Arnaldo Carvalho de Melo

Em Thu, Sep 11, 2014 at 02:43:12PM +0200, Mike Galbraith escreveu:

Quoted 11 lines · Namhyung Kim
Quoted 8 lines
Quoted 4 lines · Arnaldo Carvalho de Melo

Em Tue, Sep 09, 2014 at 01:54:01PM +0200, Mike Galbraith escreveu:

Quoted 1 line

Seems the now default on --children thingy doesn't like -U much.

Namhyung, can you please take a look at this?

So what is the problem here?

Well, if you don't see anything wrong, I guess nothing at all.

:-)

Quoted 7 lines

I think that when we decide that it is so better to change defaults like we did this time, we should be required to add a big fat warning (a --tui popup, use the first lines on --stdio, etc) about why the default was changed and allow quick, easy opt out, restoring previous behaviour after the user, being warned, knows what to expect, tries it, and then is in a better position to decide if keeping the new default is what is desired.

So the problem is that why it turned on --children option by default, right? I thought you mentioned there's a problem with -U option and I couldn't figure out what it is.

Quoted 11 lines
Quoted 9 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines

Samples: 5K of event 'cycles', Event count (approx.): 2268660922
Children Self Symbol
+ 46.42% 0.04% [k] system_call_fastpath

I'll just turn it off until I figure out what cool stuff this is telling me. why that symbol becomes the number one hit, and why total% > 100.

To me, it looks like top smoked it's breakfast, went to lala land ;-)

Yeah, its confusing, I'll let Namhyung explain it ;-)

Now I have three persion yell at me for this change. :)

When this change was developping, Ingo said it'd be better if it looks like output of sysprof as it's more popular for most (userland?) guys. You can see the discussion in the following links:

https://lkml.org/lkml/2013/10/31/97
https://lkml.org/lkml/2013/11/1/85

The children field is a cumulative total overhead (for its all children/callee) so sum of them would be more than 100%. And as Ingo requested it sorts the output entries using children overhead so that one can easily see higher level view of performance bottle-neck.

Thanks,
Namhyung

#8 lore raw
Quoted 49 lines · Namhyung Kim

Hi Arnaldo and Mike,

Quoted 15 lines · Arnaldo Carvalho de Melo

Em Thu, Sep 11, 2014 at 02:43:12PM +0200, Mike Galbraith escreveu:

Quoted 11 lines · Namhyung Kim
Quoted 8 lines
Quoted 4 lines · Arnaldo Carvalho de Melo

Em Tue, Sep 09, 2014 at 01:54:01PM +0200, Mike Galbraith escreveu:

Quoted 1 line

Seems the now default on --children thingy doesn't like -U much.

Namhyung, can you please take a look at this?

So what is the problem here?

Well, if you don't see anything wrong, I guess nothing at all.

:-)

Quoted 7 lines

I think that when we decide that it is so better to change defaults like we did this time, we should be required to add a big fat warning (a --tui popup, use the first lines on --stdio, etc) about why the default was changed and allow quick, easy opt out, restoring previous behaviour after the user, being warned, knows what to expect, tries it, and then is in a better position to decide if keeping the new default is what is desired.

So the problem is that why it turned on --children option by default, right? I thought you mentioned there's a problem with -U option and I couldn't figure out what it is.

Quoted 11 lines
Quoted 9 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines

Samples: 5K of event 'cycles', Event count (approx.): 2268660922
Children Self Symbol
+ 46.42% 0.04% [k] system_call_fastpath

I'll just turn it off until I figure out what cool stuff this is telling me. why that symbol becomes the number one hit, and why total% > 100.

To me, it looks like top smoked it's breakfast, went to lala land ;-)

Yeah, its confusing, I'll let Namhyung explain it ;-)

Now I have three persion yell at me for this change. :)

When this change was developping, Ingo said it'd be better if it looks like output of sysprof as it's more popular for most (userland?) guys. You can see the discussion in the following links:

Quoted 7 lines

https://lkml.org/lkml/2013/10/31/97
https://lkml.org/lkml/2013/11/1/85

The children field is a cumulative total overhead (for its all children/callee) so sum of them would be more than 100%. And as Ingo requested it sorts the output entries using children overhead so that one can easily see higher level view of performance bottle-neck.

I still don't get why my #1 _KERNEL_ (-U) overhead in the below changes from __schedule() to system_call(). The thing that's consuming the most cycles remains __schedule(). It doesn't really matter much I 'spose, turning it off in .perfconfig works fine.

(seems top wants to learn to ignore cpu_startup_entry too [see hmm below], and maybe grow an option to _not_ ignore idle stuff)

--no-children
- 6.52% [k] __schedule
- __schedule
- 56.95% schedule
99.88% pipe_wait
pipe_read
new_sync_read
vfs_read
sys_read
system_call
__read_nocancel
__libc_start_main
+ 40.63% schedule_preempt_disabled
+ 1.03% __read_nocancel
0.78% pipe_wait
pipe_read
new_sync_read
vfs_read
sys_read
system_call
__read_nocancel
__libc_start_main
+ 0.62% cpu_startup_entry <== hmm

+    5.13%  [k] native_sched_clock
+    4.80%  [k] system_call
+    4.36%  [k] resched_curr
+    3.59%  [k] _raw_spin_lock_irqsave
+    3.34%  [k] __switch_to

--children

+   68.49%     4.92%  [k] system_call
+   32.75%     0.42%  [k] sys_write
+   31.85%     0.86%  [k] vfs_write
+   30.20%     0.43%  [k] sys_read
+   29.34%     0.74%  [k] new_sync_write
+   29.22%     0.62%  [k] vfs_read
+   27.88%     0.70%  [k] new_sync_read
+   27.65%     1.70%  [k] pipe_write
+   26.62%     2.86%  [k] cpu_startup_entry  <== hmm
+   26.23%     2.05%  [k] pipe_read
+   22.38%     0.45%  [k] __wake_up_sync_key
-   21.92%     6.61%  [k] __schedule
   - __schedule
      - 62.95% schedule
           99.88% pipe_wait
              pipe_read
              new_sync_read
              vfs_read
              sys_read
              system_call
              __read_nocancel
      + 35.27% schedule_preempt_disabled
      + 1.38% __read_nocancel
+   20.53%     0.81%  [k] __wake_up_common
+   19.72%     0.17%  [k] autoremove_wake_function
+   19.55%     0.09%  [k] default_wake_function
+   19.01%     1.44%  [k] try_to_wake_up
+   18.65%     0.84%  [k] pipe_wait
+   15.07%     1.06%  [k] schedule
#9 lore raw

Em Fri, Sep 12, 2014 at 10:41:33AM +0200, Mike Galbraith escreveu:

Quoted 58 lines · Namhyung Kim
Quoted 49 lines

Hi Arnaldo and Mike,

Quoted 15 lines · Arnaldo Carvalho de Melo

Em Thu, Sep 11, 2014 at 02:43:12PM +0200, Mike Galbraith escreveu:

Quoted 11 lines · Namhyung Kim
Quoted 8 lines
Quoted 4 lines · Arnaldo Carvalho de Melo

Em Tue, Sep 09, 2014 at 01:54:01PM +0200, Mike Galbraith escreveu:

Quoted 1 line

Seems the now default on --children thingy doesn't like -U much.

Namhyung, can you please take a look at this?

So what is the problem here?

Well, if you don't see anything wrong, I guess nothing at all.

:-)

Quoted 7 lines

I think that when we decide that it is so better to change defaults like we did this time, we should be required to add a big fat warning (a --tui popup, use the first lines on --stdio, etc) about why the default was changed and allow quick, easy opt out, restoring previous behaviour after the user, being warned, knows what to expect, tries it, and then is in a better position to decide if keeping the new default is what is desired.

So the problem is that why it turned on --children option by default, right? I thought you mentioned there's a problem with -U option and I couldn't figure out what it is.

Quoted 11 lines
Quoted 9 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines

Samples: 5K of event 'cycles', Event count (approx.): 2268660922
Children Self Symbol
+ 46.42% 0.04% [k] system_call_fastpath

I'll just turn it off until I figure out what cool stuff this is telling me. why that symbol becomes the number one hit, and why total% > 100.

To me, it looks like top smoked it's breakfast, went to lala land ;-)

Yeah, its confusing, I'll let Namhyung explain it ;-)

Now I have three persion yell at me for this change. :)

When this change was developping, Ingo said it'd be better if it looks like output of sysprof as it's more popular for most (userland?) guys. You can see the discussion in the following links:

Quoted 7 lines

https://lkml.org/lkml/2013/10/31/97
https://lkml.org/lkml/2013/11/1/85

The children field is a cumulative total overhead (for its all children/callee) so sum of them would be more than 100%. And as Ingo requested it sorts the output entries using children overhead so that one can easily see higher level view of performance bottle-neck.

Quoted 7 lines

I still don't get why my #1 _KERNEL_ (-U) overhead in the below changes from __schedule() to system_call(). The thing that's consuming the most cycles remains __schedule(). It doesn't really matter much I 'spose, turning it off in .perfconfig works fine.

(seems top wants to learn to ignore cpu_startup_entry too [see hmm below], and maybe grow an option to _not_ ignore idle stuff)

yeah, ideally we would have some flag from the kernel stating that it is idle context, so that we don't have to keep this list of idle functions.

Until then we can try to get this more manageable by having a file where people can add idle functions and as well as not just discard entries for these functions, but instead mark them as idle.

And as well have a key in interactive tools to toggle them on/off and somethile like -U/-K to set the initial value of this toggle from the command line.

Quoted 61 lines

--no-children

(quoted diff) +22 -3
-    6.52%  [k] __schedule
   - __schedule
      - 56.95% schedule
           99.88% pipe_wait
              pipe_read
              new_sync_read
              vfs_read
              sys_read
              system_call
              __read_nocancel
              __libc_start_main
      + 40.63% schedule_preempt_disabled
      + 1.03% __read_nocancel
        0.78% pipe_wait
           pipe_read
           new_sync_read
           vfs_read
           sys_read
           system_call
           __read_nocancel
           __libc_start_main
      + 0.62% cpu_startup_entry <== hmm
+    5.13%  [k] native_sched_clock
+    4.80%  [k] system_call
+    4.36%  [k] resched_curr
+    3.59%  [k] _raw_spin_lock_irqsave
+    3.34%  [k] __switch_to
 
 
--children
+   68.49%     4.92%  [k] system_call
+   32.75%     0.42%  [k] sys_write
+   31.85%     0.86%  [k] vfs_write
+   30.20%     0.43%  [k] sys_read
+   29.34%     0.74%  [k] new_sync_write
+   29.22%     0.62%  [k] vfs_read
+   27.88%     0.70%  [k] new_sync_read
+   27.65%     1.70%  [k] pipe_write
+   26.62%     2.86%  [k] cpu_startup_entry  <== hmm
+   26.23%     2.05%  [k] pipe_read
+   22.38%     0.45%  [k] __wake_up_sync_key
-   21.92%     6.61%  [k] __schedule
   - __schedule
      - 62.95% schedule
           99.88% pipe_wait
              pipe_read
              new_sync_read
              vfs_read
              sys_read
              system_call
              __read_nocancel
      + 35.27% schedule_preempt_disabled
      + 1.38% __read_nocancel
+   20.53%     0.81%  [k] __wake_up_common
+   19.72%     0.17%  [k] autoremove_wake_function
+   19.55%     0.09%  [k] default_wake_function
+   19.01%     1.44%  [k] try_to_wake_up
+   18.65%     0.84%  [k] pipe_wait
+   15.07%     1.06%  [k] schedule
#10 lore raw
In reply to Namhyung Kim

Em Fri, Sep 12, 2014 at 04:24:54PM +0900, Namhyung Kim escreveu:

Quoted 26 lines

Hi Arnaldo and Mike,

Quoted 15 lines · Arnaldo Carvalho de Melo

Em Thu, Sep 11, 2014 at 02:43:12PM +0200, Mike Galbraith escreveu:

Quoted 11 lines · Namhyung Kim
Quoted 8 lines
Quoted 4 lines · Arnaldo Carvalho de Melo

Em Tue, Sep 09, 2014 at 01:54:01PM +0200, Mike Galbraith escreveu:

Quoted 1 line

Seems the now default on --children thingy doesn't like -U much.

Namhyung, can you please take a look at this?

So what is the problem here?

Well, if you don't see anything wrong, I guess nothing at all.

:-)

Quoted 7 lines

I think that when we decide that it is so better to change defaults like we did this time, we should be required to add a big fat warning (a --tui popup, use the first lines on --stdio, etc) about why the default was changed and allow quick, easy opt out, restoring previous behaviour after the user, being warned, knows what to expect, tries it, and then is in a better position to decide if keeping the new default is what is desired.

Quoted 3 lines

So the problem is that why it turned on --children option by default, right? I thought you mentioned there's a problem with -U option and I couldn't figure out what it is.

Yeah, at that point I thought the problem was that -U wasn't honoured in --children mode, but after I looked again at the output, it was all [k], so I thought that it was something else but didn't got back to the thread :-)

Quoted 4 lines
Quoted 4 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines

Samples: 5K of event 'cycles', Event count (approx.): 2268660922
Children Self Symbol
+ 46.42% 0.04% [k] system_call_fastpath

Quoted 8 lines
Quoted 6 lines
Quoted 4 lines

I'll just turn it off until I figure out what cool stuff this is telling me. why that symbol becomes the number one hit, and why total% > 100.

To me, it looks like top smoked it's breakfast, went to lala land ;-)

Yeah, its confusing, I'll let Namhyung explain it ;-)

Now I have three persion yell at me for this change. :)

Hey, its not "yell" or something, I just think that it is under documented, and hey, perf is world renowned for being under developed, we don't need any further efforts in that area :-P

Quoted 11 lines

When this change was developping, Ingo said it'd be better if it looks like output of sysprof as it's more popular for most (userland?) guys. You can see the discussion in the following links:

https://lkml.org/lkml/2013/10/31/97
https://lkml.org/lkml/2013/11/1/85

The children field is a cumulative total overhead (for its all children/callee) so sum of them would be more than 100%. And as Ingo requested it sorts the output entries using children overhead so that one can easily see higher level view of performance bottle-neck.

yeah, yeah, but please have all this in the documentation :-)

- Arnaldo