Re: [PATCH V1 0/4] KVM vPMU support for x86

[Wei Huang <wei@redhat.com>]

On 11/03/2014 12:39 PM, Radim Krčmář wrote:
> 2014-11-03 12:23-0600, Wei Huang:
>>
>>
>> On 11/03/2014 11:56 AM, Radim Krčmář wrote:
>>> 2014-10-31 12:05-0400, Wei Huang:
>>>> Currently KVM only supports vPMU for Intel platforms. This patch set 
>>>> enable vPMU support for AMD platform by creating a common PMU
>>>> interface for x86. The PMU calls from guest VMs are dispatched
>>>> to corresponding functions defined in arch specific files.
>>>
>>> The functionality looks good, so I just want verify the basic design:
>>> why don't we emulate AMD PMU on Intel, and vice versa?
>>> (Underlying PERF_COUNTs are identical in both.)
>>
>> Thanks. The underlining perf counters can be very different between AMD
>> and Intel. I think we can emulate AMD on Intel, or vice versa, for some
>> common perfmon_events (such as PERF_COUNT_HW_CPU_CYCLES). But as soon as
>> guest VMs access raw counters (see PERF_TYPE_RAW), we can't emulate them
>> anymore.
> 
> Thanks, I guess raw counters are used more than I thought, so code
> complexity would overshadow the gain of having at least something.
To be honest I did try. But it became a big rat-hole as soon as I tried
to abstract them. This problem also applies to Intel CPUs between
generations, if perf counters are not arch_counters.

-Wei

> 
> And then, there is the always perfect, "who cares" :)


> --
> To unsubscribe from this list: send the line "unsubscribe kvm" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>