Re: SMP doesn?t work the way i want.

On Fri, Feb 07, 2003 at 04:33:09PM -0500, John Franklin wrote:
> On Fri, Feb 07, 2003 at 09:06:45AM +0100, Martin Husemann wrote:
> > On Fri, Feb 07, 2003 at 08:45:43AM +0100, Patrik Hansson wrote:
> > 
> > > Wy can't i see both in systat ?
> > 
> > Because systat has not been updated for multiple CPU display yet.
> > It should, I guess.
> > 
> > > when i recompile a kernel only one cpu gets the load, and top
> > > reports 50% idle. Is tha normal ?
> > 
> > Yes. If the system only has one process (or more precise: thread) to run, it
> > can't realy use the second CPU. But for all makefiles in the NetBSD source
> > treee, you can add a "-j X" argument to make, i.e. "make -j 4", to make it
> > run multiple processes (compilers) at the same time.
> > 
> > If you have enough memory, this works great.
> 
> Not so much for me.  I have have no end of problems with it, in
> particular if something fails during the build, the build scripts skip
> to the "install" phase instead of dying at the point of failure.
> 
> Sometimes this is aggravated by build race-conditions.  I have run into
> them in the past, but I don't have any examples.  Typically, it
> manifests as an attempt to link to a library that's not yet finished
> building.  I'll give it another try this weekend and see if I can't find
> some concrete examples to send-pr.
> 
> If this has been corrected, then toolchain/18573 should be closed.

Well, I think most of them have been corrected. I've done multiple
build.sh -j4 on a dual 1Ghz system without problems (other than usual
breakages on current ;)

-- 
Manuel Bouyer <[email protected]>
     NetBSD: 24 ans d'experience feront toujours la difference
--