Why is the Mill CPU good for Lisp? The Mill requires the compiler to schedule for cache and memory latency, but I would expect this to be difficult to predict for Lisp programs.
You wouldn't predict it but if you assume decent code/data locality, which isn't an odd assumption in this case, the Mill's concept of scope and execution is close to a hardware implementation of Lisp's apply.
FYI; I'm at the limits of my understanding here, so if you're a compiler or CPU architect and you doubt what I'm saying here is true then you're probably right ;)
FYI; I'm at the limits of my understanding here, so if you're a compiler or CPU architect and you doubt what I'm saying here is true then you're probably right ;)