I know that MOL (Mac-On-Linux) can run Mac OS 9.2 (and higher), and it runs under Linux/PPC.

And QEMU can emulate a PowerPC and run Linux for PPC.

So, can we run MOL under QEMU, and eventually run Mac OS 9.2 ??

Hmm. Try it. It might be quite slow though.
Why is this in the SheepShaver forum?

As far as I know, the QEMU project hasn't yet gotten a full powerpc OS to load. Last I heard they got the bootloader to run, but not much more than that. (If I'm mistaken, please correct me.) Once they do, I'd certainly look into this possibility.

I've also seen that PearPC should be capable of running a powerpc linux distro, so I wonder if this same method is possible with that?

Could someone move this into the General Macintosh Emulation forum (since we no longer have a PPCemulation forum)?

the main problem with running 9.1 to my understanding is the requirement of an mmu. In previous versions (up to 9.0.4) no mmu was "required". the emulator faked the os into beliving there was an mmu. I believe 9.1 requires an mmu(aka it actually uses it). This is also the reason pearpc is so slow. Every time it needs something in memory it has to call a function. The way to make it really fast would be if you could figure out a way to use the pc's mmu for purposes of storing the mac memory. The mmus are so different that it is hard to implement this!

Hope this helps,
Mike

This was a random thought I just had. I was reading a bit into some things here about using the GPU as something of a co-processor. Would it be at all possible, no matter how difficult, to send off the MMU requests to the graphics card, provided it had a powerful enough GPU? Let the CPU deal with the PPC-x86 translations, and use the GPU to calculate all the MMU conversions. I need to read a few more books on modern PC architectures and programming languages, but something in the back of my head says this is concievable.

I can move stuff!

All PowerPC MacOSes need an MMU. The tricky part is to determine what usage MacOS makes of it and circumvent it or provide enough virtualization of that.

Even if it were possible, you'd be far better off letting the native MMU of the host handle most of this task for you... Memory conflicts between reserved areas of the host and guest OS aside, I think some of the only problems you'd have is limiting the page size to that of the host OS, and when emulating MMUs that have no-execute protection on hosts that don't, you'd probably only be able easily protect pages that haven't been re-complied yet.

That said, I wonder if any available GPUs would be able to pick up some of the slack in altivec emulation.

What about 'mapping' some of the host pc's memory to the guest mac and then just making an addition e.g. mac's memory address c34242 will be 1bc1f85 when the mac's memory begining address will be f8dd43.

i think that would be possible, and if so, will speed things up a great deal.