John,

> If you can gunify & inline the bytecode fetch that would make things a  
> lot faster for a modest increase in size.

    Tim also noticed there are places where inline made no sense.

    Can we just use the 'inline' specifier and leave the dirty job of actually
    expanding the code where it makes sense to the C/C++ compiler ?

    The C/C++ compiler can even be instructed to expand always and 
    the code is still always more readable ;-)

    Cheers,

    PhiHo.