Xen devel - Nov 2006 - RE: [Xen-changelog] [xen-unstable] [PV-ON-HVM] Don''t generate lots ofspurious interrupts when using event

>From: Steven Smith
>Sent: 2006年11月2日 19:38
>
>> 	This does solve the problem, however it adds unnecessary
>overhead
>> (one more trap into xen at the end of each event handler). IMO, the
>real
>> cause should be in pic_intack, where pending irr is converted into isr.
>I''m inclined to agree here.  The patch I checked in certainly
wasn''t
>very pretty, and this is the best alternative I''ve heard so far.
>However, it looks like it''s actually slower than what
we''ve got at the
>moment: NPtcp between dom0 and domU on my test box reports a
>latency
>and maximum bandwidth of 36.42us and 1961.58Mbps without this
>patch, or
>45.94us and 1874.69Mbps with.  All measurements are best-of-three.
>
>Looking at the patch, it ought to have been slightly faster, so these
>results rather surprised me.  Did you do any benchmarks yourself?
>
>Steven.
No, I didn''t because when I reading your patch it simply indicated a 
spurious interrupt issue to be fixed there. Based on the description, 
I came up above as an alternative. It''s quite interesting to see such
fix
with so obvious performance degradation. BTW, is it a typo? This logic 
should only matter for hvm guest, and why throughout between dom0 
and domU is affected which is more odd if true?

Thanks,
Kevin

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

On 3/11/06 1:24 am, "Tian, Kevin" <kevin.tian@intel.com> wrote:
> No, I didn''t because when I reading your patch it simply indicated
a
> spurious interrupt issue to be fixed there. Based on the description,
> I came up above as an alternative. It''s quite interesting to see
such fix
> with so obvious performance degradation. BTW, is it a typo? This logic
> should only matter for hvm guest, and why throughout between dom0
> and domU is affected which is more odd if true?
I think this is performance between dom0 and a Linux HVM domU using PV
drivers. Otherwise, as you say, it would make no sense!

 -- Keir



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

> >Looking at the patch, it ought to have been slightly faster, so these
> >results rather surprised me.  Did you do any benchmarks yourself?
> No, I didn''t because when I reading your patch it simply indicated
a
> spurious interrupt issue to be fixed there. Based on the description, 
> I came up above as an alternative.
Fair enough.
> It''s quite interesting to see such fix with so obvious performance
> degradation.  BTW, is it a typo? This logic should only matter for
> hvm guest, and why throughout between dom0 and domU is affected
> which is more odd if true?
Sorry, this was between dom0 and an HVM domU.

Steven.


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