Hi Stephane,
On 09/02/2010 06:58 AM, stephane eranian wrote:
> Corey,
>
>
> I have heavily restructured the powerpc support inside libpfm4
> to align it with the way we do things for x86 now. In particular, I wanted
> the ability to access every event table even if not running on the
> matching HW host. That meant that each PMU model needed to have
> its owns pfmlib_pmu_support entry. I made all the changes. I have
> tested on x86 (with a little hack) and it seems to work fine. I would
> appreciate if you could try on actual HW with the perf_examples. I
> could only play with the generic examples (showevtinfo, check_events)
>
> With the new structure, I can do:
> $ showevtinfo
> Supported PMU models:
> [41, ppc970, "PPC970"]
> [42, ppc970mp, "PPC970MP"]
> [44, power4, "POWER4"]
> [45, power5, "POWER5"]
> [46, power5p, "POWER5+"]
> [47, power6, "POWER6"]
> [48, power7, "POWER7"]
> Detected PMU models:
> Total events: 2716 available, 0 supported
>
>
> $ check_events power7::pm_nest_4
> Supported PMU models:
> [41, ppc970, "PPC970"]
> [42, ppc970mp, "PPC970MP"]
> [44, power4, "POWER4"]
> [45, power5, "POWER5"]
> [46, power5p, "POWER5+"]
> [47, power6, "POWER6"]
> [48, power7, "POWER7"]
> Detected PMU models:
> Total events: 2716
> Requested Event: power7::pm_nest_4
> Actual Event: power7::PM_NEST_4
> PMU : POWER7
> IDX : 805306368
> Codes : 0x87
>
>
> Just pull from git.
From all of the commits recently, it looks like you've been very busy
with both libpfm and perf_events!
Thanks for making the changes to Power arch to support this new structure.
I tested the latest code (pulled 5 minutes ago) on a Power5 machine. It
built flawlessly, and I tried hardware several events using the task
example, and they worked fine. Check events appears to give the correct
output as well:
% ./check_events
Supported PMU models:
[41, ppc970, "PPC970"]
[42, ppc970mp, "PPC970MP"]
[44, power4, "POWER4"]
[45, power5, "POWER5"]
[46, power5p, "POWER5+"]
[47, power6, "POWER6"]
[48, power7, "POWER7"]
[49, perf, "perf_events generic PMU"]
Detected PMU models:
[45, power5, "POWER5"]
[49, perf, "perf_events generic PMU"]
Total events: 2736
Requested Event: PERF_COUNT_HW_CPU_CYCLES
Actual Event: perf::PERF_COUNT_HW_CPU_CYCLES
PMU : perf_events generic PMU
IDX : 822083584
Codes : 0x0
Requested Event: PERF_COUNT_HW_INSTRUCTIONS
Actual Event: perf::PERF_COUNT_HW_INSTRUCTIONS
PMU : perf_events generic PMU
IDX : 822083585
Codes : 0x1
So, it looks good. Let me know if there's anything else you'd like me
to try.
Thanks again,
- Corey
------------------------------------------------------------------------------
This SF.net Dev2Dev email is sponsored by:
Show off your parallel programming skills.
Enter the Intel(R) Threading Challenge 2010.
http://p.sf.net/sfu/intel-thread-sfd
_______________________________________________
perfmon2-devel mailing list
perfmon2-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/perfmon2-devel
