Re: [perfmon2] perfmon2 problem on Cell

[stephane e. <er...@go...>]

Vince,


There is clearly a problem. I don't have a PS3 myself so I cannot test this.
I suggest you look at the perfmon2 stats. You need debugfs to do this.
Once you mount the filesystem, look at perfmon/cpu/cpu1 (if testing on CPU1).
There are a few files covering the interrupt handler. It should be
interesting to
look at the counts reported compared to the expected value.


On Wed, Aug 20, 2008 at 8:31 PM, Vince Weaver <vi...@cs...> wrote:
> Hello
>
> I've been trying to use perfmon2 on long-running programs on the Cell
> architecture (a Playstation3).
>
> The problem is that pfmon never returns counts higher than 2^32.  On
> kernel 2.6.24 it seemed that the counters were just overflowing and the
> virtual 64-bit counter just not updated.  I updated the kernel to 2.6.26
> (which was a lot of pain, turns out there is something broken with USB on
> kernels more recent than 2.6.24) and the problem is changed a bit but
> still there.  Now the counts seem to max out at 4294967295
>
> For example, I get results like this:
>
>> taskset 1 pfmon -e ppc_inst_commit_th0,ppc_inst_commit_th1,cycles long_running_program
> 4294967295 PPC_INST_COMMIT_TH0
>          0 PPC_INST_COMMIT_TH1
> 4294967295 CYCLES
>
> Is this a known problem?
>
> Vince
>
>
> -------------------------------------------------------------------------
> This SF.Net email is sponsored by the Moblin Your Move Developer's challenge
> Build the coolest Linux based applications with Moblin SDK & win great prizes
> Grand prize is a trip for two to an Open Source event anywhere in the world
> http://moblin-contest.org/redirect.php?banner_id=100&url=/
> _______________________________________________
> perfmon2-devel mailing list
> per...@li...
> https://lists.sourceforge.net/lists/listinfo/perfmon2-devel
>