Stockfish on github

[JuLieN]

No, SF in itself can run on very modest hardware (at least for the CPU). I guess you problems are more:

- How much RAM would SF need for a decent use, with acceptable hash tables?
- How much flash storage would you need to store SF, its book, and a minimal Os to run it (I guess you'll be using a minimal stripped-down Linux?)

What hardware specs do you intend to have for your device, if that's not secret? Maybe it's still under discussion?

EDIT: btw, the Raspberry Pi has 128 MB of RAM, an ARM CPU with a powerful graphics, handle networks and many other thing, and is sold for $25...

[bturner]

Well, the issue we're grappling with is what firmware/hardware we need. We're looking at a Novag Citrine type machine.

My understanding may be incomplete. It is:

SF is a chess engine that can be used with a UCI and GUI (xboard, winboard) to show to a computer screen, OR just compiled to a high-enough-level chip and electronics built for it to work on a dedicated chess computer.

Is my understanding way out?

[JuLieN]

No, you understood well. If you run SF on a minimal Linux you would just need to write a proxy adapter for communications between the hardware and SF : this adapter would get the input from the board, translate this to uci instruction, feed them to SF and get back SF's output to whatever your hardware display.

But you probably know all this very well yet.

[kinderchocolate]

It depends on the hardware requirements for your system. Stockfish requires about 30MB RAM to run on iPhone. If you can afford 30MB or more, there should't be any issue.

Furthermore, you can reduce the RAM by reducing the default hash size. You can reduce it to around 10MB easily.

I think you just need to process piping the stockfish inputs and outputs, and you will be fine.

If you can't even give the chip even 10MB, consider MicroMax, shouldn't take more than a MB.

[bturner]

Thanks enormously Julien and Tim for your comments - I really appreciate your help