Upper limits on performance are pretty high, and lower limits on available resources, in currently popular, hence, cost-effective technology, are ample.

As Erik states, many of us have had the experience, while attempting to make a standard IC do what we want, of thinking, "if only it worked this way ..."

With programmable logic, there's no reason it can't work exactly as you want. For example, I've never been satisfied with the handshaking capabilities of the 8255. While I can't, just now, remember the exact details, the handshake seldom worked as I needed. With a programmable device, it will work exactly as I design it. What's more, I don't have to start with an 8255 or 6821 core. I can design my parallel I/O in any way I see fit. It can have clocked handshaking, interrupt generation, bitwise direction controls, differential outputs, it can interface serially or in parallel, wth positive negative logic, or whatever. It can be jumper configurable for handshake "sense" or it can operate from an external clock with double buffering, again, whatever I want.

If I don't need the timer/counter, or need a different one, supporting a mode only available in some other MCU, I can put that in there, too. If I need the serial port to modulate/demodulate serial data as BPSK, it can do that, too.

If you don't look at ALTERA, ACTEL, LATTICE, and XILINX from time to time just to know what they're offering you, you're missing a lot of potentially viable options. It's also a good idea to keep up on what "free" IP is available.

RE