Re: [Xen-devel] How expensive is a VM EXIT?

[Keir Fraser <keir.fraser@xxxxxxxxxxxxx>]

Either of those can give you profiling info about common paths through the
hypervisor, and time spent in the hypervisor. For the specific
microbenchmark of time taken for a processor to execute a vmexit, I'd hack
in custom tracing to get a precise measurement.

 -- Keir

On 04/03/2009 18:39, "Nadolski, Ed" <Ed.Nadolski@xxxxxxx> wrote:

> Is there a good tool to look at for this -- xentrace, xenoprofile, other?
> 
> Thanks,
> Ed
> 
> -----Original Message-----
> From: xen-devel-bounces@xxxxxxxxxxxxxxxxxxx
> [mailto:xen-devel-bounces@xxxxxxxxxxxxxxxxxxx] On Behalf Of Keir Fraser
> Sent: Wednesday, March 04, 2009 1:46 AM
> To: NAHieu
> Cc: xen-devel@xxxxxxxxxxxxxxxxxxx
> Subject: Re: [Xen-devel] How expensive is a VM EXIT?
> 
> On 04/03/2009 00:15, "NAHieu" <nahieu@xxxxxxxxx> wrote:
> 
>> On Wed, Mar 4, 2009 at 1:34 AM, Keir Fraser <keir.fraser@xxxxxxxxxxxxx>
>> wrote:
>>> For just the VMEXIT itself, it's probably about 1000-1500 cycles on the
>>> latest processors. I haven't measured it very recently though.
>>> 
>> 
>> Could you explain how do you measure the cycles for VMEXIT?
> 
> RDTSC in the guest immediately before triggering a VMEXIT and RDTSC in the
> hypervisor on entry.
> 
>  -- Keir
> 
> 
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@xxxxxxxxxxxxxxxxxxx
> http://lists.xensource.com/xen-devel



_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel