Re: [gentoo-user] Recommendation for CPU type in QEMU?

[Kerin Millar]

On 03/08/2013 15:55, Marc Joliet wrote:
> Am Wed, 31 Jul 2013 13:12:01 +0100
> schrieb Kerin Millar <kerframil@×××××××××××.uk>:
>
>> On 31/07/2013 12:31, Marc Joliet wrote:
>>
>> [snip]
>>
>>>
>>> There's also "-cpu host", which simply passes your CPU through to the guest.
>>> That's what I use for my 32 bit WinXP VM. You can use it if you don't mind not
>>> being able to migrate your guest, but it sounds to me like you're doing this on
>>> a desktop machine, so I suspect guest migration doesn't matter to you.
>>>
>>
>> I thought the same until very recently but it's not the case. The -cpu
>> host feature exposes all feature bits supported by qemu. Those may
>> include features that aren't supported in hardware by the host CPU, in
>> which case qemu has to resort to (slow) emulation if they are used.
>>
>> --Kerin
>
> Just a follow up: the most authoritative answer I could find is this:
>
>    http://thread.gmane.org/gmane.comp.emulators.kvm.devel/84227/focus=90541
>
> Furthermore, the Linux KVM tuning page also defines "-cpu host" as I understand
> it:
>
>    http://www.linux-kvm.org/page/Tuning_KVM
>
>  From the above I conclude that "-cpu host" should *not* activate CPU features
> that the host CPU does not support.
>
> Otherwise I could only find out the following:
>
> - the Gentoo and Arch wikis both recommend "-cpu host" in conjunction with KVM
>    (see, e.g., http://wiki.gentoo.org/wiki/QEMU/Options)
> - in contrast, http://wiki.qemu.org/Features/CPUModels#-cpu_host_vs_-cpu_best
>    seems to match your statement
> - some guy on serverfault.com says this
>    (http://serverfault.com/questions/404195/kvm-which-cpu-features-make-vms-run-better):
>
>    "Qemu doesn't work in the same way many other hypervisors do. For starters, it
>    can provide full emulation. That means you can run x86 code on an ARM
>    processor, for example. When in KVM mode, as you're using it, it doesn't
>    actually do that... the processor is exposed no matter what, but what is
>    reported to the OS will be changed by the -cpu flag."
>
>    If that's correct, "-cpu host" might mean different things when in KVM
>    mode vs. when not. However I'm not going to blindly trust that statement.
>
> How/where did you find out that "-cpu host" also exposes non-host CPU features?
>

I checked the code and you're right. I had obtained the information from 
the qemu wiki but can now only assume that the content was discussing 
the feature before its implementation became concrete. Lesson to self: 
don't believe everything one reads in wikis (even official ones).

--Kerin