Xen devel - May 2007 - vmx status report against cset 15072 - one performance issue

Hi,
In our today''s testing, we still met the issue below.
1. Xen0 hang when destroy VMX guest
http://bugzilla.xensource.com/bugzilla/show_bug.cgi?id=3D981

And found one more performance issue on IA32e:
2. Performance of HVM guest downgraded much.
Compares with cset 15017, kernel build on HVM against 15072 downgraded
about 35%.
And even the execution time of our nightly testing has been doubled.

We suspect the 1th issue is related with gcc version.
We found the issue on the binary compiled with gcc 3.4.3.
If uses gcc 4.1.0, the problem will be gone.

Thanks
Yunfeng

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

> 2. Performance of HVM guest downgraded much.
> Compares with cset 15017, kernel build on HVM against 15072 downgraded
> about 35%.
> And even the execution time of our nightly testing has been doubled.
Please can you try and pin down the range of changesets for this
regression -- there''s several possibilities in this range of 55.

Thanks,
Ian

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

We suspect it may be caused by the update of qemu.
IA64 part has also been found the same performance issue.

Thanks
Yunfeng
>-----Original Message-----
>From: Ian Pratt [mailto:Ian.Pratt@cl.cam.ac.uk]
>Sent: 2007年5月18日 19:46
>To: Zhao, Yunfeng; xen-devel@lists.xensource.com
>Cc: ian.pratt@cl.cam.ac.uk
>Subject: RE: [Xen-devel] vmx status report against cset 15072 - one
>performanceissue
>
>> 2. Performance of HVM guest downgraded much.
>> Compares with cset 15017, kernel build on HVM against 15072 downgraded
>> about 35%.
>> And even the execution time of our nightly testing has been doubled.
>
>Please can you try and pin down the range of changesets for this
>regression -- there''s several possibilities in this range of 55.
>
>Thanks,
>Ian
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

We did some regression tests for the performance issue.
And we found that performance issue was introduced by set 15021.
On cset 15020, performance of hvm guest is OK. 

Thanks
Yunfeng
>-----Original Message-----
>From: Ian Pratt [mailto:Ian.Pratt@cl.cam.ac.uk]
>Sent: 2007年5月18日 19:46
>To: Zhao, Yunfeng; xen-devel@lists.xensource.com
>Cc: ian.pratt@cl.cam.ac.uk
>Subject: RE: [Xen-devel] vmx status report against cset 15072 - one
>performanceissue
>
>> 2. Performance of HVM guest downgraded much.
>> Compares with cset 15017, kernel build on HVM against 15072 downgraded
>> about 35%.
>> And even the execution time of our nightly testing has been doubled.
>
>Please can you try and pin down the range of changesets for this
>regression -- there''s several possibilities in this range of 55.
>
>Thanks,
>Ian
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel