Re: [PATCH] kernel <linux-2.6.11.10> kernel/sched.c

[Lee Revell <rlrevell@joe-job.com>]

On Thu, 2005-05-19 at 21:17 -0700, chen Shang wrote:
> > Hi Chen,
> > With the added branch and the extra icache footprint, it isn't clear
> > that this would be a win.
> > 
> > Also, you didn't say where your statistics came from (what workload).
> > 
> > So you really need to start by demonstrating some increase on some workload.
> > 
> > Also, minor comments on the patch: please work against mm kernels,
> > please follow
> > kernel coding style, and don't change schedstat output format in the
> > same patch
> > (makes it easier for those with schedstat parsing tools).
> > 
> Hi Nick,
> 
> Thank you very much for your comments. This is the first time of my
> kernel hacking. I will reduce the lines of changes as much as
> possible. As regard to the statistics, there are just count, ie, the
> total number of priority-recalculations vs. the number of priority
> changed from the former recalculation.

A kernel profile (check list archives for oprofile) would easily
demonstrate any performance gain.  On my system the residency of
schedule() is around 1% so this will be easy to spot.

Lee