Suggestions for testing VAPIC / TPR patching

Suggestions for testing VAPIC / TPR patching

[Jan Kiszka]

Hi again,

what is the best way to check if the TPR patching feature of qemu-kvm is
working and performs efficiently? I suppose Windows XP guests, e.g.? How
to measure this best? I'm about to start porting it to upstream and like
to prepare a good test case.

Thanks,
Jan

PS: After the /kvm removal, I'm currently counting a bit more than 7000
LOC difference in qemu-kvm. We are converging.

-- 
Siemens AG, Corporate Technology, CT T DE IT 1
Corporate Competence Center Embedded Linux

Re: Suggestions for testing VAPIC / TPR patching

[Gleb Natapov]

On Thu, Jan 19, 2012 at 09:35:47AM +0100, Jan Kiszka wrote:
> Hi again,
> 
> what is the best way to check if the TPR patching feature of qemu-kvm is
> working and performs efficiently? I suppose Windows XP guests, e.g.? How
> to measure this best? I'm about to start porting it to upstream and like
> to prepare a good test case.
> 
Run WindowsXP on AMD HW without vapic and see number of TPR access exits.
Than run it with vapic and check again. Don't forget to check that
reboot et al work.

> Thanks,
> Jan
> 
> PS: After the /kvm removal, I'm currently counting a bit more than 7000
> LOC difference in qemu-kvm. We are converging.
> 
> -- 
> Siemens AG, Corporate Technology, CT T DE IT 1
> Corporate Competence Center Embedded Linux
> --
> 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

--
			Gleb.

Re: Suggestions for testing VAPIC / TPR patching

[Jan Kiszka]

On 2012-01-19 10:02, Gleb Natapov wrote:
> On Thu, Jan 19, 2012 at 09:35:47AM +0100, Jan Kiszka wrote:
>> Hi again,
>>
>> what is the best way to check if the TPR patching feature of qemu-kvm is
>> working and performs efficiently? I suppose Windows XP guests, e.g.? How
>> to measure this best? I'm about to start porting it to upstream and like
>> to prepare a good test case.
>>
> Run WindowsXP on AMD HW without vapic and see number of TPR access exits.
> Than run it with vapic and check again. Don't forget to check that
> reboot et al work.

So is this an optimization only for AMD CPUs or can it be reproduced on
Intel as well? Any CPU feature dependencies?

Jan

-- 
Siemens AG, Corporate Technology, CT T DE IT 1
Corporate Competence Center Embedded Linux

Re: Suggestions for testing VAPIC / TPR patching

[Gleb Natapov]

On Thu, Jan 19, 2012 at 10:14:53AM +0100, Jan Kiszka wrote:
> On 2012-01-19 10:02, Gleb Natapov wrote:
> > On Thu, Jan 19, 2012 at 09:35:47AM +0100, Jan Kiszka wrote:
> >> Hi again,
> >>
> >> what is the best way to check if the TPR patching feature of qemu-kvm is
> >> working and performs efficiently? I suppose Windows XP guests, e.g.? How
> >> to measure this best? I'm about to start porting it to upstream and like
> >> to prepare a good test case.
> >>
> > Run WindowsXP on AMD HW without vapic and see number of TPR access exits.
> > Than run it with vapic and check again. Don't forget to check that
> > reboot et al work.
> 
> So is this an optimization only for AMD CPUs or can it be reproduced on
> Intel as well? Any CPU feature dependencies?
> 
It can be reproduced on Intel as well, but older once. There is
flexpriority module option that you can disable on loading, but I think
it is not enough and you also need to disable tpr_shadow, but there is
not option for that.

--
			Gleb.