Re: Discrepancies between Oprofile and vmstat II

[Andi K. <an...@fi...>]

On Mon, Apr 27, 2009 at 03:15:27PM -0700, Styner, Douglas W wrote:
> 
> > I believe so, but will confirm.
> >
> > > opcontrol -e=CPU_CLK_UNHALTED:80000 -e=LLC_MISSES:6000
> > > 
> > > Using another profiling tool to confirm, we see 74.784% user, 25.174% > > kernel.
> >
> > Just verifying -- you also see it when you use a shorter period than 80000 > right?
> 
> Confirmed.  Changes in profile appear to be due to increased sampling by oprofile with lower value.

You mean "with higher value", right ?

It's still strange that it suddenly changed in 2.6.27 though. There shouldn't have been
a difference there. So I still think it's a regression.

-Andi