Re: Porting to IBM Risc 6000

[Chuck Silvers <chuq%chuq.com@localhost>]

On Sun, Aug 11, 2002 at 12:22:05PM -0700, Matt Thomas wrote:
> At 11:22 AM 8/11/2002, Jochen Kunz wrote:
> >On 2002.08.09 16:34 David Edelsohn wrote:
> >
> >>       PReP is not the same as CHRP.
> >Yes, of course. My question was more like "Are they similar enough to
> >support them with a single NetBSD port?" I don't know PReP and CHRP
> >internas yet. I started to read the CHRP spec to get an overview...
> 
> With enough indirection/abstraction we could have one port.  The
> question is whether it's worth it.

I always like to have fewer kernel binaries (or at least have the
ability to configure such) to cover the spectrum of supported hardware.
that can make things easier for administrators.


> >> PReP machines do not have the same
> >> Open Firmware device tree, client services, and RTAS.
> >[...]
> >>       RPA is the IBM-specific continuation of CHRP.
> >>  It supports Open Firmware (device tree, client services, RTAS),
> >> but it may have different devices than specified in the CHRP
> >> specification.  It may use different interrupt controllers
> >> (not OpenPIC), etc.
> >So we would end with port-prep for the PReP only machines and port-chrp
> >(port-rpa?) for CHRP and RPA machines? And perhaps an additional
> >port-chrp64 for 64 bit CHRP/RPA machines? Analogus to port-sparc and
> >port-sparc64? (I know I can ask a lot of stupid questions. ;-) )
> 
> My inclination is to have to move to a rs6k port and have
> GENERIC and GENERIC64 for 32bit and 64bit PPC platforms.
> The real question is do we want a single kernel which can work
> on 32bit and 64bit platforms?

again, I'll say that it's useful to allow that.

the other interesting configuration is to allow running 64-bit applications
under a 32-bit kernel.  AIX supports this.  the advantage is that 32-bit
code is faster, and if you don't need the larger address space in the kernel,
then there's no real drawback (other than it being more code to maintain).

-Chuck