hi,

just for grins, you could build a kernel with both debug and profiling,
and see if there's any obvious place where most of the slowdown comes from.
if it would be easy to make debug kernels faster without making them
harder to debug, that would be a fine change that we could make.

-Chuck


On Wed, Dec 18, 2002 at 12:59:17PM -0600, Peter Seebach wrote:
> Okay, a useful data point:  Removing DEBUG from my kernel has made a HUGE
> difference in interactive performance while running CVS.  This makes the
> difference between refreshing Mozilla's page in 1/10th second when swapping
> virtual desktops, and refreshing it in 3-4 seconds.
> 
> -s