Jon Buller wrote:
> > thank's for your really marvelous idea. I changed your patch to handle
> > immediate constants. The code knows about constants in the -7...8 range
> > and ommits the addition of the low half if it is 0.
> 
> So if I type say a + 2**33 at GCC (with proper expansion of 2**33
> of course) it will do a single addqd?  Cool, I definately would
> not have bothered with that, I was just going to do the -7..8 stuff.

Right. It was not hard to do and came in almost for free when I did the
-7...8 stuff.

> > PS: If you wish to send this to the fsf, afaik you have to send the patch
> >     with a ChangeLog entry to Richard Kenner (kenner@vlsi1.ultra.nyu.edu).
> > 
> > Here comes my version of your patch (I did a prototype ChangeLog entry for
> > you):
> 
> Thanks, I did not know what needed to be done on this one.
> Have you noticed any improvements with the new mods?

The kernel is a little bit smaller. I did recompile all libraries and installed
them and the system still seems to work :-)

> I've been having
> problems getting NetBSD-current to compile.  (makefiles not being
> able to cd into directories, etc.)  So I haven't really gotten
> anything much done with it yet.

Yes. It took me some time to recompile gcc again. I just missed that I'd
have to do a "make depend" again.

	Matthias
-- 
leo@dachau.marco.de			in real life: Matthias Pfaller
marco GmbH, 85221 Dachau, Germany	tel: +49 8131 516142