1) Looks like i didn't understand the co-processor division, very well. Thought i/o co-proc was "link i/o" such as, keyboard, mice, paddles, serial, parallel, net...

2) No reason other than idealized throughput to argue against "polled i/o". Most of my experience says "interrupt i/o" is smoother, easier to strategize, and allows more foreground "free-board".

I was thinking of the case where video and audio might play off the disk. With interrupts you could compensate for the lack of DMA by using buffer interrupts to simply route data to Video/Audio.

3)If a mag disk co-proc implements, it can handle the stepping and seeking, and real-time buffering. But sooner or later it's buffer's fill and unless you unloading them (throughput) someone is stalling somewhere. It's not going to be "code or data" into memory that will stall it would be animation and sound that would test the connection capacity.

4)Many co-proc architectures that handle bulk data use DMA. The '51 doesn't have a integrated DREQ/DACK architecture so dual-ports would be the easiest way to implement this. They can also be costly.

In the absense of some form of "dual porting", the CPU performs the role that the accumulator does in the '51 itself; "bottleneck generation". With an interrupt you could appear to handle data more smoothly by setting up background transfers.

A less founded argument could be coded into a question; what else would you be doing with the interrupt?

Question about the audio... what kind is it?

regards,
p