With the technology available today the most effective multi-processor implementation is to use following guidelines....

1) Utilize a fixed program object for each processor. In other words commit a particular functional slice of the function of the product for each processor.

2) Utilize current high integration processors with onboard FLASH, local RAM, and timers to implement the programming for the functional slice.

3) Place all processors at a location that is suitable to the configuration of the product. These could all be on the same board, multiple boards in a cage or stack, or each located in separate chassis enclosures.

4) Devise a communications scheme for the multiple processors so that they can coordinate their functions and share status information. Always dedicate one processor as the master and all others as slaves. The communcations should be initiated by the master and responded to by the slaves. Unsolicited communications from slave to master should be generally discouraged.

5) Implement the communications in one of two ways. Either parallel or serial.

5A) Parallel communications is best implemented using dual ported SRAM chips, I recommend one dual port chip for each master to slave communications path. The master can see the master side of the SRAMs either by linear address decoding or by bank switching.

5B) Serial communications can be implemented for the lowest cost overhead for those applications where shared information flow and throughput is not exceedingly high. This scheme can use any of RS485, I2C, or CAN protocol.

Good Luck
Michael Karas