> > 
> > It certainly would be cute to have the companion cpu run m68k binaries,
> > but I predict that the difficulties:
> > 
> > - synchronizing caches on cross-cpu context switch
> > - translating syscalls (ok, the fake-netbsd-m68k kernel would solely
> > consist of the syscll translation mechanism)
> > 
> > will prevent its easy implementation. 
> 
> Symmetric Multi Vendor Multi Processing, not in my lifetime...

Uh, no, what was called for was Asymmetric Multi Processing (AMP).

	-is