Re: 2.6.25-rc3-git3: Reported regressions from 2.6.24

["Zhang, Yanmin" <yanmin_zhang@linux.intel.com>]

On Wed, 2008-03-05 at 10:06 +0800, Zhang, Yanmin wrote:
> On Tue, 2008-03-04 at 16:57 +0530, Dhaval Giani wrote:
> > > 
> > > Bug-Entry	: http://bugzilla.kernel.org/show_bug.cgi?id=9978
> > > Subject		: 2.6.25-rc1: volanoMark 45% regression
> > > Submitter	: Zhang, Yanmin <yanmin_zhang@linux.intel.com>
> > > Date		: 2008-02-13 10:30
> > > References	: http://lkml.org/lkml/2008/2/13/128
> > > Handled-By	: Srivatsa Vaddagiri <vatsa@linux.vnet.ibm.com>
> > > 		  Balbir Singh <balbir@linux.vnet.ibm.com>
> >  
> > Peter's revert of the load balance patches should fix this one. Yanmin,
> > could you please confirm if the patch at
> > http://lkml.org/lkml/2008/2/25/202 helps?
> I tested it against 2.6.25-rc3 on my 16-core tigerton machine. It really improves
> volano result although it doesn't recover all the result.
> Comparing with 2.6.24, without the patch, volanoMark has about 50% regression
> with 2.6.25-rc3. With the patch, volanoMark has about 15% regression.
One more update on the reverted patch: Comparing with 2.6.24, cpu2000-fp has about
4% regression with kernel 2.6.25-rc on my madison IPF machine. As you know, cpu2000-fp
consists of many sub-testing. The most regression looks relevant to a couple of testing
in the middle step. But if I ran the sub-testing manually, I couldn't see any regression.
If I started kernel with boot parameter maxcpus=1, the regression becomes 1%.

If I apply Peter's revert patch to 2.6.25-rc3, the regression also becomes 1%.

I don't know what causes the last 1% regression.

-yanmin