PearPC

[Ilwrath]

JIT Emulation of PPC on a 3000Mhz x86 CPU would be approximately equivalent to a 1200 MHz emulated G3 CPU using the PearPC code, as it's supposed to be 40% slower than the host CPU.

WHOA!!!! Where did you get that figure??  The real author says it's about 500x slower on a cycle-for-cycle basis.  That seems to me to be quite an exaggeration.  

On my 3ghz P4, I got a somewhat usable Mac OSX emulation, yet quite slow (at least compared to THAT figure!).  Probably about what I'd expect from an early (unsupported) PowerMac.  Maybe a 75-100mhz G3 emulation.  (It took about 2.5 hours to install Mac OSX 10.3 Panther!) Still, this would put the emulation at a level with some of the slower Cyberstorm PPC and Blizzard PPC cards....  and probably have better video support.  The video and IO of PearPC seem quite fast.  It's the processor emulation core that's the bottleneck...  But still, I'm impressed for a 0.1 beta.  :-)

I'm hoping that there's some people who want to work this into UAE!  That would be awesome.  :-)  I've never actually owned a real PPC Amiga...

[Ilwrath]

[...]using the PearPC code, as it's supposed to be 40% slower than the host CPU.

This can be achieved when you don't need to emulate the CPU [...]

This makes sense...  Like I said, the video and IO portion seems to work fairly quickly.  So, an unsupported PowerPC platform can run OSX at 40% slower than a real Mac of the same clock runs it.  That's good to know.  So, an A1 or Pegasos could use this and get quite a usable Mac emulation.  (Usable is a word I've never heard MoL described as.  I can't pass judgement on it myself, though, as I don't have a fast enough PPC to try it.)

But, back to PearPC....  If you use the processor emulation, you certainly get no where near losing only 40%.  In fact, I don't think a PPC emulation is even possible to that tight of performance...  I mean, think about it.  That's a lot tighter than any 680x0 emulation routines, and how much simpler of a processor is the 680x0 compared to a G3....

Anyhow, I had only read the x86 docs on PearPC, which was why I questioned the origins of that 40% figure.  ;-)