It's a tough situation you're in, with that venerable old PIO. A CPLD would solve your problem, I suspect, though most of them are no longer 5-volt compatible and I doubt you'll be able to meet the PLCC-44 pinout of the 82C55. Current FPGA's are just too costly for such a task, as they require something to boot them on power-up and the smallest of them would be 100x too big, in addition to which they use 2 or 3 power supplies, not one of which is 5 volts.

You should probably ask yourself why, aside from that's what you've been doing, you want to use the 8255. If you use a 5-volt CPLD such as a 44-pin XC9500 or a similar Altera part, e.g. something from their MAX7000A series, you can probably meet your needs. However, if you want canned IP, you have to buy it from someone, e.g. CAST, Inc. or someone similar, not that I recommend their products, else you have to create what you need yourself, which is what I'd recommend. It's clear that you don't need the entire functionality of the 8255, so there's no need to provide it all. With a little thought, you can probably reduce what you need to a small part of what the 8255 can do, and that may well fit into a small CPLD, such as a MAX7032 or an XC9536. However, since you're using a large number of boards, you should run the numbers to see whether it's worth leaving the PCB alone and building a CPLD adapter instead. I'd say not, in the general case, since you're talking 5K boards per year. A reroute of the board to accomodate pinout changes, etc, as mandated by the use of one CPLD or another will cost <<$1k, including the board fab setup costs. You probably won't need to change the drill tool, so it's just a reroute, which means plotting new films. An additional cost will be the modification and testing of the board-resident code. Only you can determine what the associated cost of that will be.

Another alternative, of course, is to skull out an approach that uses discrete MSI ports and MSI/SSI handshaking. Maybe that will work better for you, since that requires no programmable logic and support tools, including learning curve. If your product will be in use for a considerable time yet, it will probably be worth making whatever changes are needed. Though I've had some experience with the 8255, I've never thought that it did things better than the SSI/MSI solution did it. Once you decide on a mode in which you're going to use the thing, it's much easier to conjure up some circuitry that does things better than the venerable 8255.

If you need some specifics, then contact me directly.

RE