Menu
▾
▴
Re: how to explain the call graph
|
From: Ming Z. <mi...@el...> - 2005-08-16 12:13:45
|
cool. thx.
later that day i used similar step to get a good call graph. something
like this.
CPU: P4 / Xeon with 2 hyper-threads, speed 3391.74 MHz (estimated)
Counted GLOBAL_POWER_EVENTS events (time during which processor is not
stopped) with a unit mask of 0x01 (mandatory) count 100000
samples % image name app name
symbol name
-------------------------------------------------------------------------------
1701 0.1020 vmlinux vmlinux
copy_from_user
4375 0.2623 vmlinux vmlinux
__generic_file_aio_write_nolock
1661536 99.6227 vmlinux vmlinux
generic_file_buffered_write
1647498 5.9225 vmlinux vmlinux
__copy_from_user_ll
1647498 100.000 vmlinux vmlinux
__copy_from_user_ll [self]
-------------------------------------------------------------------------------
45313 100.000 tg3.ko tg3
tg3_start_xmit
1198501 4.3084 tg3.ko tg3
tg3_start_xmit
1198501 94.3050 tg3.ko tg3
tg3_start_xmit [self]
45313 3.5655 tg3.ko tg3
tg3_start_xmit
7196 0.5662 tg3.ko tg3
tg3_set_txd
5535 0.4355 vmlinux tg3
sub_preempt_count
4837 0.3806 vmlinux tg3
common_interrupt
3408 0.2682 vmlinux tg3
apic_timer_interrupt
1818 0.1431 vmlinux tg3
_spin_unlock_irqrestore
1431 0.1126 vmlinux tg3
add_preempt_count
-------------------------------------------------------------------------------
7320 0.6746 oprofiled oprofiled
opd_put_sample
195494 18.0170 oprofiled oprofiled
sfile_log_sample
882238 81.3084 oprofiled oprofiled
sfile_log_arc
1078682 3.8777 oprofiled oprofiled
odb_insert
1078682 99.9818 oprofiled oprofiled
odb_insert [self]
-------------------------------------------------------------------------------
2910851 100.000 tg3.ko tg3
tg3_poll
1056230 3.7970 tg3.ko tg3
tg3_rx
1545577 49.6682 vmlinux tg3
netif_receive_skb
1056230 33.9427 tg3.ko tg3
tg3_rx [self]
257219 8.2659 tg3.ko tg3
tg3_alloc_rx_skb
234289 7.5290 vmlinux tg3
eth_type_trans
4072 0.1309 vmlinux tg3
sub_preempt_count
3851 0.1238 vmlinux tg3
ip_rcv
On Tue, 2005-08-16 at 15:50 +0800, Lv Zheng wrote:
> You are welcome, but your data seems not containing backtrace
> informations. only samples marked 'self' are generated.
> I think you should confirm following things, these will cause the
> callgraph data missing problem:
>
> 1. make sure your kernel is compiled with "framepointer" enabled
> 2. make sure callgraph depth is set with none-zero value
> `cat /dev/oprofile/backtrace_depth`
> you can do this by opcontrol --callgraph=5 after executing
> opcontrol --vmlinux=XX (pay attention to the command sequence,
> vmlinux must be set first)
> 3. make sure you are not using an ARM kernel together with
> oprofile-0.9.0 (a bug fix can be found in 0.9.1)
> 4. use opreport -c to generate callgraph data for you samples.
>
> following is a sample for oprofile callgraph usage:
>
> VMLINUX="/vmlinux"
> BACKTRACE=5
> VERSION=`uname -r`
>
> opcontrol --shutdown &> /dev/null
> opcontrol --init
> opcontrol --reset
> opcontrol --vmlinux=${VMLINUX}
> opcontrol --callgraph=${BACKTRACE}
> opcontrol --separate=kernel
> opcontrol --event=GLOBAL_POWER_EVENTS:533333:1:1:1
> opcontrol --start
> # run something here
> opcontrol --dump
> opcontrol --shutdown
> opreport -c -l -o myreport_${VERSION}_${i}_${j}
>
> Best regards
>
> ----- Original Message -----
> From: "Ming Zhang" <mi...@el...>
> To: "Lv Zheng" <zh...@sh...>
> Cc: "oprofile" <opr...@li...>
> Sent: Monday, August 15, 2005 8:46 PM
> Subject: Re: how to explain the call graph
>
>
> > thanks a lot. i found it.
> >
> > i assumed opstack could give me this. but seems opreport --callgraph is
> > for this.
> >
> > Ming
> >
> > On Mon, 2005-08-15 at 09:58 +0800, Lv Zheng wrote:
> > > Hi, Ming
> > >
> > > Make sure following configuration items enabled for your kernel
> > > then run oprofile --callgraph again.
> > >
> > > oprofile:
> > > Profiling support
> > > -> Profiling support
> > > -> OProfile system profiling
> > > frame pointers:
> > > Kernel hacking
> > > -> Kernel debugging
> > > -> Compile the kernel with frame pointers
> > >
> > > Best regards
> > >
> > > ----- Original Message -----
> > > From: "Ming Zhang" <mi...@el...>
> > > To: "oprofile" <opr...@li...>
> > > Sent: Friday, August 12, 2005 2:40 AM
> > > Subject: how to explain the call graph
> > >
> > >
> > > > I wonder how to explain this call graph output.
> > > >
> > > > i have oprofile configured as this
> > > >
> > > > [root@sc420 root]# opcontrol --status
> > > > Daemon running: pid 3469
> > > > Separate options: library kernel
> > > > vmlinux file: /usr/src/linux-2.6.12.3/vmlinux
> > > > Image filter: none
> > > > Call-graph depth: 5
> > > >
> > > > then i ran this to get file x.
> > > >
> > > > opreport -t0.1 -l -c -p /lib/modules/2.6.12.3/kernel/ -o x
> > > >
> > > > for example, i saw this schedule take 5%. but then how can i tell who
> > > > and where this schedule() is called?
> > > >
> > >
> > --------------------------------------------------------------------------
> > > -----
> > > > 54820 5.6787 vmlinux vmlinux
> > > > schedule
> > > > 54820 100.000 vmlinux vmlinux
> > > > schedule [self]
> > > >
> > > >
> > > > thanks a lot!
> > > >
> > > > Ming
> > > >
> > > >
> > >
> >
> >
> >
> > -------------------------------------------------------
> > SF.Net email is Sponsored by the Better Software Conference & EXPO
> > September 19-22, 2005 * San Francisco, CA * Development Lifecycle
> Practices
> > Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
> > Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf
> > _______________________________________________
> > oprofile-list mailing list
> > opr...@li...
> > https://lists.sourceforge.net/lists/listinfo/oprofile-list
>
|