My long term goal is to make parallela (+addon-board) an all-in-one reprap controller.
Step 1 would be to make the coordinate transformation for complex printer geometries (like this:
http://www.youtube.com/watch?v=af7XkgxzmXE ) in realtime on the Epiphany. Since a microcontroller can't handle that (at least not an Arduino) the current workaround is to use a G-Code preprocessor on the PC. This approach has many limitations. For the Epiphany that's peanuts, probably even one core of those sixteen is already enough. If not, this can be time-multiplexed to several cores. So for this step I want to put the parallela between the Arduino (outputting coordinates, if bandwidth allows that or stepper-pulses as a workaround) and the motor drivers. That needs a parallella breakout board of course.
That's in the near term. Further Ideas:
Step 2 would be to design an open source stepper driver that performs better than the "stepstick"s currently in use. Main motivation is noise reduction. Microchip Application Note 1307 is the blueprint, the implementation would have to be done in VHDL on the Zynq. I can't do this alone though, because I'm not competent to design the board with the ADCs and H-Bridge drivers necessary. The breakout board is necessary for that also.
Step 3: Eliminate the Arduino, porting current firmware to parallella. There are at least four ways to accomplish that: (a) port some firmware to ZPUino, (b) put an AVR core in the FPGA, (c) dedicate one Cortex-A9 core to the task, (d) use real-time-linux kernel.
Step 4: Eliminate the Raspberry Pi many are using for network connectivity. That's the simplest one: just install Octoprint on parallella.
Having all these things integrated, the parallella would actually be cost competitive. There would be a lot of headroom for the further evolution in reprap. For example, if the design and slicing toolchain would go from line- to spline-based formats, the Epiphany has enough power to process that in realtime.