On Thu, 21 Dec 2000, Luke Mewburn wrote:

> On Wed, Dec 20, 2000 at 10:18:56AM -0800, Jason R Thorpe wrote:
> > It sounds like what we need is:
> > 
> > 	sys/arch/arm
> > 	sys/arch/arm/arm	<- shared by all arm cpus
> > 	sys/arch/arm/arm32	<- arm32 cpus
> > 	sys/arch/arm/arm26	<- arm26 cpus
> > 
> > ...and then:
> > 
> > 	sys/arch/hpcarm32	<- hpcarm32 port
> > 	sys/arch/dnard		<- dnard port
> 
> How about making this the prototype case for a source reorganisation
> that we've mumbled about for years - /sys/cpu.
> 
> Then we could have:
> 
> 	sys/cpu/arm-shared	<- shared by all arm cpus
> 	sys/cpu/arm32		<- arm32 cpus
> 	sys/cpu/arm26		<- arm26 cpus
> 
> ...and then:
> 
> 	sys/arch/hpcarm32	<- hpcarm32 port
> 	sys/arch/shark		<- shark port	(dnard is a horrid name ;-)

An awkward aspect of both of these from my point of view is that they
force me to split my code in half, since we presumably wouldn't want all
of arm26 in the cpu (or arch/arm) directory.  This seems a somewhat
artificial split.  It also means I need to rename my port
(NetBSD/archimedes?), but that's hardly valid grounds for objection.

> I think this could be the perfect opportunity to experiment with this layout.

Why does it need to be experimented with?  As far as I can see, it's a
simple renaming exercise, so if it's thought to be a good idea, it should
just be done.  Moving one CPU at a time will just give you confusing
intermediate states.

Incidentally, I think if we went for sys/cpu/*, it should all be bundled
away under sys/cpu/arm, so as to save clutter in sys/cpu.

-- 
Ben Harris                                                   <bjh21@netbsd.org>
Portmaster, NetBSD/arm26               <URL:http://www.netbsd.org/Ports/arm26/>