Alan,

> You'll love this:
>  >>>>>>>>>>>
> patchInterpGCCPPC: fileName
>     "Interpreter patchInterpGCCPPC: 'Squeak copy 1'"
>     "This will patch out the unneccesary range check (a compare
>      and branch) in the inner interpreter dispatch loop. for the PPC 
> version of the GCC compiled
>     version of Squeak under MPW"
> <<<<<<<<<

  Well, this one is not exactly a C-macro, but close to what Andreas
suggested; to separate the hardware dependent part from the
independent part.  In this case, it is executed manually as a post
pass by the VM developer and there is zero chance that this code slows
down the VMs on the other platforms.

  I remember that I wrote a Emacs-Lisp function to do the thing for
SH-3 assembly.  This gave me 6% bc/s improvement.

-- Yoshiki