Coduino

The Coduino (Coprocessor for Arduino systems) is 4 Arduino-compatible systems on a shield, to enable multiprocessor stacks.

Project Status

2013-05: Going to call this completed after all, decided against offering it as a product as I don't think there's a big enough market there. All working though, and the design is all open if anyone wants to make one :)
2012-06: Putting into dormant status, as I won't have time until after Burning Man as I'm putting all my time into Beacon4.
2012-04: Boards are here, awaiting time to test.
2012-03-10: Started/finished schematic for baseboard and submitted to Laen.
2012-03-04: Ordered connectors to prototype 'base board' from Digikey.
2012-03-03: Wrote a token-passing demo, base Arduino injects a token which the 4 CPUs on the Coduino pass around in a ring. I2C bus proved.
2012-02-24: First board assembled, runs 'blink' on all 4 CPUs successfully.
2012-02-22: Parts arrived.
2012-02-21: Ordered parts from Digikey.
2011-10-18: PCBs received.
2011-09-30: Completed routing board, added all required silkscreen, submitted to Laen for prototype manufacturing. Need to order parts now.
2011-09-29: Almost done routing board.
2011-09-26: Started routing board.
2011-09-25: Added master reset button and power LED. Schematic passes checklist.
2011-09-23: Started working on schematic, now 90% complete. Need to add a master reset button and a power LED.
Pre-Sept 2011: Put specs and BOM together.

Uses ATMega328p for compatibility with the Duemilanove board.
Standard boot loader will be used.
Serial only in the first revision, no USB-Serial on the board itself (FTDI cables are easy to get now).
Communicates (optionally) with the master Arduino (bottom of the stack) via I2C (with jumpers to disconnect this), leaving all other pins available.
Main headers just 'stack-through' the master Arduino, the I/O from the Coduino is broken out on separate headers (maybe with a 'base board' that can connect to stack shields on the Coduino via a cable?).
Board has 4 processors, could stuff half of them for a lower-cost version, though I don't think it would lower the cost that much in the end
Each processor has a local reset button, and can also be optionally linked to a master reset using jumpers
All 4 processors run off a single oscillator using a 4:1 fanout buffer, so they will stay synced up instruction-by-instruction