The MC6821 and MCS6520 both have bit-by-bit control over I/O direction.

If you're "into" old chips, there are some 8085-compatible (by implication, 805x-compatible as well) that have port direction control as well. I don't see the advantage, however. If you program a port (byte-wise) as output and set some of the bits to input, you still have to "mask off" (with an AND) the output bits on an input cycle, else yuo have to track the values you've last written to the port, which you may be able to do with a read-modify-write instruction of one sort or another.

I freely admit that I never learned to like the MCS6522, which was considered "cool" by the uniitiated, but which turned out to be a high-cost, low-benefit device that did little better or more effectively than its much lower-costing cousins.

The problem you're likely to encounter is 68xx/65xx-peripheral performance. If you consider what the phase-2 clock (call it 'E' if you like) on the 6502 or 680x did, and how it was timed, you'll see that addresses were valid before it became true, but that data became valid later on writes and was not required to be valid until the falling edge of phase-0 (see manufacturers' spec's for details). Caution! There is a required setup time and hold-time. If you operate your 805x slowly enough, you can use the NAND (negative logic OR) of nWR and nRD to generate a useable 'E' or Phase-2 to the 68xx/65xx peripheral. I doubt you'll be willing to do that, even with the fastest available 68xx/65xx peripherals of that era.

RE