More ULE bugs fixed.

[Bruce Evans]

On Sun, 26 Oct 2003, Jon Mini wrote:

> Jeff Roberson [jroberson at chesapeake.net] wrote :
>
> > On Fri, 17 Oct 2003, Bruce Evans wrote:
> >
> > How would one test if it was an improvement on the 4BSD scheduler?  It
> > is not even competitive in my simple tests.
>
> What were your simple tests?

Er, they were in the original mail.  Just do parts of buildworld with -j16
on an SMP system.  ULE was 2.4 times slower for make depend and 2.1 times
slower for make obj.  Something must have been very wrong, since make obj,
especially, should be completely i/o bound so it shouldn't be affected
by the scheduler.  Also, run a bunch of CPU hog processes with various
nicenesses and look at top output to check that they are given reasonable
amounts of CPU.

Bruce