> The Great Mr. Kurtz writes:
> > > No one else is targetting
> > > Apple PowerPC-based hardware (as far as I know) because of the
> > > lack of documentation.
> > 
> > Hmmm.  I'm sure the PREP compliance stuff is out there.  After all, there 
> > are two or three companies already building clones.  I've got a reference 
> > manual on my shelf for the PPC 601/603/604/620 procesor.  It includes an 
> > entire section on OS development for the PPC.  What other information do 
> > you need?  Is it just stuff about Apple's implementation?  Talk to the 
> > clone manufacturers.  They'd be a lot more likely to help.  :-)
> 
> Unfortunately, you need a lot more than a processor ref. manual to
> write an OS, in the same way that you need a lot more than a processor
> to build a computer.  That is what the PREP spec is for.
> Unfortuantely, the PowerMacs are not PREP-compliant (as you point out
> below, they are essentially quadras) and, apparently, Apple is less
> than enthusiastic about divulging their technical details.  The clone
> makers may or may not be bound by some sort of non-disclosure
> agreement, but they'd be worth asking.

Except that CHRP has come out. I've seen two books so far, a spec
covering about half the problem (from IBM), and a book from Appl
describing the MacIO chip! This chip is essentially the IO chip in an
8500, or there abouts.  There have been a few removals (like the second
SCSI port and the Ethernet and something else, which is mentioned), but
the doc describes the hardware in a rough manner. There's also supposed
to be an I/O book, for writing drivers. I haven't found it.

Check out the OpenMac pages on puma. There's a whole campaign to get
info out of Apple, for PPC NetBSD, Linux, and others.

> > Anybody put together a cross-compiler for PPC yet?  I was just thinking 
> > if someone got an AIX copy of gcc (to get the right instruction set) and 
> > patched it with the standard netbsd/mac68k includes with (possibly) some 
> > tweaks to the big endian/litle endian stuff, could we at least get a 
> > kernel that would work as a serial terminal in native mode?  At least 
> > with the nubus PowerMacs, the internal architecture is nauseatingly 
> > similar to the Quadras, but with a few changes in the memory management 
> > and the display and its interface with the memory.
> 
> Last I saw, GCC and GAS could compile and assemble PPC code, but GNU
> ld couldn't link it.  Has this changed in the latest binutils?  I
> haven't checked.
> 
> If it does work, it would be immediately useful for MacOS development
> too, since MacOS can run standard AIX XCOFF-format binaries (that is,
> XCOFF format, but making MacOS, not AIX, traps.)

I think the Linux/PPC people have made some success here. They're
working hard right now on the tools, and I bet this is where they're
concentrating.

Big endian/Little endian patches? Isn't the PPC primarily a big-endian
machine, which can make little-endian accesses, and can (I was told)
generate little-endian stack frames?

Take care,

Bill