Just a small and irrelevant remark: Z80 did have blockmoves (LDIR/LDDR), and one would certainly not call it a DSP... :-) <p>

Ah, my bad (I'm still somewhat sad about not getting into assembly programming in the 80's ... when I was still young and things would have been even easier to learn. So I've never done anything in Z80 or 65xx assembly. I've had a project where the "old version" was completely in 68k assembly though, and had to work through it.).

However, I was not really referring to standard blockmoves (many architectures have those), but to explicit "delay" commands (copy a[i] to a[i+1]) that for example TI C54x DSPs have - even to the point where the delay is actually part of a single-cycle MAC instruction and thus does not create any overhead. But then again, this architecture also has hardware support for circular buffers, which is really neat.