On 03/09/2015 03:53 PM, Ken McDonell wrote:
> On 07/03/15 08:17, Frank Ch. Eigler wrote:
>> ...
>> Yeah, the papi-tools weren't always packaged with older debian, though
>> see <http://packages.debian.org/search?keywords=papi-tools>.
>
> I've implemented a common check for papi-ness of a platform in common.check
> that avoids any dependence on papi-tools.
>
> And this works for the platforms that were failing the papi PMDA tests, and
> on the platforms where these tests were being run (real h/w none of this VM
> pretend nonsense), they still run and pass.
>
> _______________________________________________
> pcp mailing list
> pcp@xxxxxxxxxxx
> http://oss.sgi.com/mailman/listinfo/pcp
Hi Ken,
One of the things that I have seen on guest VMs is that the cpuid information
provided to the guest VM is not the actual values from the physical processor.
PAPI is really particular about the processor CPU family and module numbers.
This information is used to determine what performance monitoring hardware is
available. If the guest VM CPUID instruction provides bogus cpuid information,
the kernel and papi may say this is an unsupport processor or try to use the
wrong events.
It is possible to make qemu use specific cpuid info:
http://wiki.qemu.org/Features/CPUModels
-Will
|