hi,



Indeed, this is classic way. The only one note: the common suggestion is that you should not consider this way as "parallel port" but as "memory mapped I/O" instead. I mean that you connect MCU to CPLD with next scheme:
- P0 as data bus;
- /WR and /RD as strobe signals for write and read;
- optional ALE and/or P2 pins for address latch/decoding.
Then you just write and read data to/from CPLD with MOVX instructions. The time of MOVX very depends on the derivative and its settings. Most modern 8051 allow you to adjust MOVX speed. For example,
- AT89S51 does MOVX for 24 OSC periods, i.e. its time is 1us @24MHz;
- AT89C51RD2 with x2 feature is twice speed up;
- C8051F120 from SiLabs may do MOVX for 5 system clocks and it takes only 50ns for MOVX access. In theory it gives you bandwidth about 20Mb/s. I said "in theory" because you need prepare/store data during write/read as well.

By the way: usage of the concept "parallel port" instead "memory mapped I/O" decreases datarate dramaticaly. You will be need with setting port 1 pins to read/write mode, put/get data to port, manipulate with read and write strobes pins etc.



So you need not with high-speed CPLD access. In any case the slower MCU-to-PC connection determine total communication speed. Indeed, it may be improved with buffering but in general, serial link between MCU and PC is the thing which lights requirements. For example, if you use serial speed 115200 then it is about 11 kbytes per second. By other words you need to read/write a byte from/to CPLD at each 90us. Even slowest AT89S51 at 11MHz is able to do such task.

Regards,
Oleg