Maynard Johnson | 4 Jan 20:29 2010
Picon

Re: Updates to the oprofile-tests

William Cohen wrote:
> It has been a while since the oprofile testsuite has been updated. I 
> went through it Tuesday and added support for a number of architectures. 
> In particular added tests for the newer i386 and power variants. If it 
> looks okay, I will check it in.
Will, the POWER variants work fine.  Thanks!  Once Suravee gives her OK on the 
i386 part, the patch can be committed.  One favor . . . can you re-post the 
patch with a "Signed-off-by" line?  And if you are the one to commit the patch, 
please don't forget to update the ChangeLog.

Thanks.
-Maynard
> 
> -Will
> 
> 
> ------------------------------------------------------------------------
> 
> ------------------------------------------------------------------------------
> This SF.Net email is sponsored by the Verizon Developer Community
> Take advantage of Verizon's best-in-class app development support
> A streamlined, 14 day to market process makes app distribution fast and easy
> Join now and get one step closer to millions of Verizon customers
> http://p.sf.net/sfu/verizon-dev2dev 
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> oprofile-list mailing list
(Continue reading)

William Cohen | 4 Jan 20:49 2010
Picon

[PATCH] Re: Updates to the oprofile-tests

On 12/23/2009 09:42 AM, William Cohen wrote:
> It has been a while since the oprofile testsuite has been updated. I
> went through it Tuesday and added support for a number of architectures.
> In particular added tests for the newer i386 and power variants. If it
> looks okay, I will check it in.
>
> -Will
>
>
>
> ------------------------------------------------------------------------------
> This SF.Net email is sponsored by the Verizon Developer Community
> Take advantage of Verizon's best-in-class app development support
> A streamlined, 14 day to market process makes app distribution fast and easy
> Join now and get one step closer to millions of Verizon customers
> http://p.sf.net/sfu/verizon-dev2dev
>
>
>
> _______________________________________________
> oprofile-list mailing list
> oprofile-list <at> lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/oprofile-list

Reposting with the signoff and proposed ChangeLog entry. -Will

Signed-off-by: William Cohen <wcohen <at> redhat.com>
---

2010-01-04  William Cohen  <wcohen <at> redhat.com>
(Continue reading)

Myke Smith | 6 Jan 18:05 2010
Picon

question about prep_ jitdump in opcontrol

Hello,

I have encountered a problem in prep_jitdump which is due to the  
limited implementation of "find" in busybox.

> 	# VMs may already be running when profiling is started, so
> 	# remove only dump files that are not in use
> 	for I in $dumpdir/*; do
> 		test -f $I || continue;
> 		local pid=`basename $I .dump`;
> 		if test -d /proc/$pid; then
> 			local files=`find /proc/$pid/fd -lname $I`;
> 			test -n "$files" && continue;
> 		fi
> 		rm -f $I;
>

In busybox, the option "-lname" is not supported. My question, after  
looking at the code, is why is it necessary to do all of this? Isn't  
it sufficient to see that the pid is still defined in /proc and if not  
remove the file from $dumpdir?

thanks,

-myke

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
(Continue reading)

Maynard Johnson | 7 Jan 14:12 2010
Picon

Re: question about prep_ jitdump in opcontrol

Myke Smith wrote:
> Hello,
> 
> I have encountered a problem in prep_jitdump which is due to the  
> limited implementation of "find" in busybox.
> 
>> 	# VMs may already be running when profiling is started, so
>> 	# remove only dump files that are not in use
>> 	for I in $dumpdir/*; do
>> 		test -f $I || continue;
>> 		local pid=`basename $I .dump`;
>> 		if test -d /proc/$pid; then
>> 			local files=`find /proc/$pid/fd -lname $I`;
>> 			test -n "$files" && continue;
>> 		fi
>> 		rm -f $I;
>>
> 
> In busybox, the option "-lname" is not supported. My question, after  
> looking at the code, is why is it necessary to do all of this? Isn't  
> it sufficient to see that the pid is still defined in /proc and if not  
> remove the file from $dumpdir?
Daniel, you are probably the best person to answer this question since you committed this change.

-Maynard
> 
> thanks,
> 
> -myke
> 
(Continue reading)

Choi, David | 11 Jan 20:22 2010

oprofile on ARMv4

Hello,

 

I am new to oprofile. I like to use oprofile on ARMv4 based SoC with Linux 2.6.30 . I have 2 questions as followings;

 

Q1: Does oprofile supports ARMv4(ARM922T) architecture? If available, is it useful because from “ARM Architecture reference manual”, I can not any hardware performance counter? If I am wrong, please make me correct.

 

 

Q2: If available to use oprofile, how can I specify the path to kernel when issuing “opcontrol –vmlinux”?

            (Our Soc does not allocate a partition for vmlinux.)

 

Any suggestion will be highly appreciated.

 

Regards,

David J. Choi

 

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev 
_______________________________________________
oprofile-list mailing list
oprofile-list <at> lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oprofile-list
Gerrit Binnenmars | 11 Jan 19:30 2010
Picon

all samples are reported in cpu buffer overflow

Hello,

I am using linux-2.6.29-rt23 when using oprofile everything looks ok,
but opreport says there are no samples. In /var/lib/oprofile/samples/oprofiled.log 
it is shown that all samples received are also shown as cpu buffer overflow.
I have an intel core2 duo and am using the default event.

I think this is not a real buffer overflow but there is something else wrong?
I tried buffer-size and cpu-buffer-size, so far without success.

Any ideas?

With regards,

Gerrit Binnenmars

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev 
Maynard Johnson | 11 Jan 22:09 2010
Picon

[PATCH] fix up oprofile-tests to work properly with newer versions of tcl

When the oprofile testsuite is run using tcl 8.5 or later, it appears to go into 
an infinite loop.  But in fact, it just slows way down.  This is a result of how 
a spin loop is being implemented in testsuite/lib/op_util.exp.  As of tcl 8.5, 
integers of arbitrary length are supported, which means they *never* overflow. 
But since the arithmetic operations for the integer variables used within the 
spin loop rely on the overflowing behaviour of C-stle integers, the results of 
the operations were not as expected.  With the latest tcl, the size of the 
variables grows exponentially with the number of iterations, and so does the run 
time.

In order for the arithmetic operations within the spin loop to behave as 
expected, the calculations should
use modulo 2**32 = 4294967296.  See the attached patch for details.

*Will* -- can you please review this patch?

Signed-off-by:  Maynard Johnson <maynardj <at> us.ibm.com>
Attachment (op-tests.patch): text/x-diff, 1306 bytes
------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev 
_______________________________________________
oprofile-list mailing list
oprofile-list <at> lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oprofile-list
SourceForge.net | 12 Jan 00:12 2010
Picon
Picon

[ oprofile-Bugs-2930192 ] trouble with java support

Bugs item #2930192, was opened at 2010-01-11 15:12
Message generated for change (Tracker Item Submitted) made by jpritikin
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=116191&aid=2930192&group_id=16191

Please note that this message will contain a full copy of the comment thread,
including the initial issue submission, for this request,
not just the latest update.
Category: None
Group: None
Status: Open
Resolution: None
Priority: 5
Private: No
Submitted By: Joshua Pritikin (jpritikin)
Assigned to: Nobody/Anonymous (nobody)
Summary: trouble with java support

Initial Comment:
1. The host OS is Ubuntu karmic 2.6.31-17-generic x86_64. I am running
oprofile in a KVM Ubuntu Linux i686 guest with a freshly compiled 2.6.32
kernel. I am using oprofile 0.9.6 compiled from source.

2. I am using the timer=1 module option to persuade oprofile to sample
events with the KVM guest.

3. I am passing -agentpath:/usr/local/lib/oprofile/libjvmti_oprofile.so
to java. I verified that it is getting to java by adding "=debug" which
causes lots of cryptic debug spew.

4. Not sure if I need to invoke "opcontrol --dump" explicitly, but tried
it.

5. opreport -m all -l /usr/lib/jvm/java-6-sun/jre/bin/java

CPU: CPU with timer interrupt, speed 0 MHz (estimated)
Profiling through timer interrupt
samples  %        app name                 symbol name
54645    46.6493  anon (tgid:869 range:0xb4bc8000-0xb5498000) anon (tgid:869
range:0xb4bc8000-0xb5498000)
7212      6.1567  anon (tgid:869 range:0xb4bc8000-0xb5470000) anon (tgid:869
range:0xb4bc8000-0xb5470000)
3774      3.2218  libjvm.so                void
FastScanClosure::do_oop_work<oopDesc*>(oopDesc**)
2481      2.1180  anon (tgid:869 range:0xb4bc8000-0xb54e8000) anon (tgid:869
range:0xb4bc8000-0xb54e8000)
1301      1.1106  anon (tgid:2618 range:0xb4c3f000-0xb4e9f000) anon (tgid:2618
range:0xb4c3f000-0xb4e9f000)
1250      1.0671  libjvm.so
Runtime1::primitive_arraycopy(HeapWord*, HeapWord*, int)
1168      0.9971  libjvm.so
constantPoolKlass::oop_follow_contents(oopDesc*)
1158      0.9886  libjvm.so
MarkSweep::AdjustPointerClosure::do_oop(oopDesc**)
1074      0.9169  libjvm.so
DefNewGeneration::copy_to_survivor_space(oopDesc*)
...

I see C++ functions from the libjvm support library, but I do not see
any java functions in the listing comparable to:

http://oprofile.sourceforge.net/doc/getting-jit-reports.html

Am I missing something? Do I need -Djava.compiler=NONE or some such?

----------------------------------------------------------------------

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=116191&aid=2930192&group_id=16191

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev 
William Cohen | 12 Jan 02:47 2010
Picon

Re: [PATCH] fix up oprofile-tests to work properly with newer versions of tcl

On 01/11/2010 04:09 PM, Maynard Johnson wrote:
> As of tcl 8.5, integers of arbitrary length are supported

Hi Maynard,

The patch patch looks and works fine. Go ahead and check it in. -Will

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev 
Joshua Pritikin | 12 Jan 00:00 2010

help with oprofile & java

1. The host OS is Ubuntu karmic 2.6.31-17-generic x86_64. I am running
oprofile in a KVM Ubuntu Linux i686 guest with a freshly compiled 2.6.32
kernel. I am using oprofile 0.9.6 compiled from source.

2. I am using the timer=1 module option to persuade oprofile to sample
events with the KVM guest.

3. I am passing -agentpath:/usr/local/lib/oprofile/libjvmti_oprofile.so
to java. I verified that it is getting to java by adding "=debug" which
causes lots of cryptic debug spew.

4. Not sure if I need to invoke "opcontrol --dump" explicitly, but tried
it.

5. opreport -m all -l /usr/lib/jvm/java-6-sun/jre/bin/java

CPU: CPU with timer interrupt, speed 0 MHz (estimated)
Profiling through timer interrupt
samples  %        app name                 symbol name
54645    46.6493  anon (tgid:869 range:0xb4bc8000-0xb5498000) anon (tgid:869
range:0xb4bc8000-0xb5498000)
7212      6.1567  anon (tgid:869 range:0xb4bc8000-0xb5470000) anon (tgid:869
range:0xb4bc8000-0xb5470000)
3774      3.2218  libjvm.so                void
FastScanClosure::do_oop_work<oopDesc*>(oopDesc**)
2481      2.1180  anon (tgid:869 range:0xb4bc8000-0xb54e8000) anon (tgid:869
range:0xb4bc8000-0xb54e8000)
1301      1.1106  anon (tgid:2618 range:0xb4c3f000-0xb4e9f000) anon (tgid:2618
range:0xb4c3f000-0xb4e9f000)
1250      1.0671  libjvm.so
Runtime1::primitive_arraycopy(HeapWord*, HeapWord*, int)
1168      0.9971  libjvm.so
constantPoolKlass::oop_follow_contents(oopDesc*)
1158      0.9886  libjvm.so
MarkSweep::AdjustPointerClosure::do_oop(oopDesc**)
1074      0.9169  libjvm.so
DefNewGeneration::copy_to_survivor_space(oopDesc*)
...

I see C++ functions from the libjvm support library, but I do not see
any java functions in the listing comparable to:

http://oprofile.sourceforge.net/doc/getting-jit-reports.html

Am I missing something? Do I need -Djava.compiler=NONE or some such?

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev 

Gmane