Hello Carole,

Sorry for the delay in responding.

On Sat, Jul 9, 2011 at 7:23 AM, Carole Wu <cwu...@gmail.com> wrote:
> Hi,
>
> I would like to use perfmon2 to analyze parallel application performance. To
> collect the number of instructions for the serial version, I run a parallel
> application from the PARSEC benchmark suite with the following command.
>
> carole@carole-software-lab-nehalem:~/parsec-2.1# pfmon --verbose
> --no-cmd-output -eINSTRUCTIONS_RETIRED --follow-all --
> /home/carole/parsec-2.1/pkgs/kernels/canneal/inst/amd64-linux.gcc/bin/canneal
> 1 100 200 100000.nets 32
>
> measuring at user privilege level ONLY
> 1 event set(s) defined
> long  sampling periods(val/mask/seed): 0/0x0/0
> short sampling periods(val/mask/seed): 0/0x0/0
> unavailable_pmcs=0xffffffffbff0fff0
> unavailable_pmds=0x7ff8fff0
> [FIXED_CTRL(pmc16)=0xa pmi0=1 en0=0x2 any0=0 pmi1=1 en1=0x0 any1=0 pmi2=1
> en2=0x0 any2=0] INSTRUCTIONS_RETIRED
> [FIXED_CTR0(pmd16)]
> pmd setup for event set0:
> [pmd16 set=0 ival=0x0 long_rate=0x0 short_rate=0x0 mask=0x0 seed=0
> randomize=n]
> exec-pattern=*
> [19672] started task:
> /home/carole/parsec-2.1/pkgs/kernels/canneal/inst/amd64-linux.gcc/bin/canneal
> 1 100 200 100000.nets 32
> follow_exec=y follow_vfork=y follow_fork=y follow_pthread=y
> [19672] 64-bit binary
> [19672] monitoring
> /home/carole/parsec-2.1/pkgs/kernels/canneal/inst/amd64-linux.gcc/bin/canneal...
> [19672] results are on terminal
> [19672] activating monitoring
> measurements started at Sat Jul  9 10:20:43 2011
>
> [19672] task terminated by signal %
> [19672] task terminated by signal 6, results may be incorrect
>                    1523248 INSTRUCTIONS_RETIRED

Most likely, this means that your program created some child threads
or processes and
that process 19672 exited because it received a signal. Pfmon warns
you because that
process may not have run to completion.

As you probably know, perfmon is now obsolete. I suggest you start
using perf_events
and the perf tool instead. They allow comparable measurements. The
difference is that
perf_events is now part of the standard kernels, incl. distro kernels.

> /home/carole/parsec-2.1/pkgs/kernels/canneal/inst/amd64-linux.gcc/bin/canneal
> (19672,19672,19671)
> [19672] detached
> sdesc chain to process:
> created tasks        : 0
> maximum tasks        : 1
> maximum active tasks : 1
>
> The output says "..., results may be incorrect"
>
> Can anyone help to understand what's done wrong here? How can I get more
> meaningful results for parallel application measurement?
>
> Thanks in advance,
> Carole
>
>
> ------------------------------------------------------------------------------
> All of the data generated in your IT infrastructure is seriously valuable.
> Why? It contains a definitive record of application performance, security
> threats, fraudulent activity, and more. Splunk takes this data and makes
> sense of it. IT sense. And common sense.
> http://p.sf.net/sfu/splunk-d2d-c2
> _______________________________________________
> perfmon2-devel mailing list
> perfmon2-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/perfmon2-devel
>
>

------------------------------------------------------------------------------
10 Tips for Better Web Security
Learn 10 ways to better secure your business today. Topics covered include:
Web security, SSL, hacker attacks & Denial of Service (DoS), private keys,
security Microsoft Exchange, secure Instant Messaging, and much more.
http://www.accelacomm.com/jaw/sfnl/114/51426210/
_______________________________________________
perfmon2-devel mailing list
perfmon2-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/perfmon2-devel

Reply via email to