pseudo-register addressing

[Stephen Weeks]

> There might be an advantage to using pseudo-registers because of the cache.
> I.e., the top of stack might not be in the cache (if you are looping around
> doing non-tail calls) but the pseudo-registers would be.  This might never
> happen realistically though.

That seems farfetched, because if you are doing nontail calls, then stack is in
the cache because the arguments and return address were just written there.