Okay, you get the idea.

I suspect if your logic contained states (i.e flip flops) then it could be Turing but I am not an expert here. It a bit like the days of analog computers. There was a need when digital computers where a dream.

Anyway, I thought that it added a tool to my tool box. I have a number of "demo" like projects (a Turing machine being one of them). So the idea of converting a simple problem and an algorithm into a hardware demonstration is something worth keeping.

AlanX

Okay, perhaps I am wrong, but you appear to be describing a hardware greedy and/or iterative algorithm to solve your Sudoku problem. As such, your machine is still subject to the normal rules of algorithms (P!=NP etc) and your logic may not scale very well or even find a solution. Such is P!=NP. Unfortunately, Sudoku (the 9x9 case) is a know hard problem to solve.

Some thoughts:
* You can model many simple iterative algorithms with your approach, one example from my past was to model an underground ventilation system (a flow network) using "Kirchhoff's current law" as the iterative algorithm.
* Rather than a discrete hardware logic and finite state machine, you could a simple one-bit computer (as these work well in this domain).

* So there are many real world problem you can solve, that would make fun demonstration units. That is demonstrations units that don't use a microprocessor for control. The demonstration unit could feature the iterative algorithm and the real world problem it solves.

Any, best of luck with your project, regards AlanX

Sure, that's why this project was abandoned - it was a nice theory, however without practical applications it makes not much sense, especially being that hardware greedy (*possibly it can be done using FPGA as alternative*)

Someone told me, that it looks a lot like reservoir computation device, from my perspective it's more accurate, of course it's iterative, but algorithm as a sequence of instructions make sense only within the boundaries of blocks (e.g. numbers in case of sudoku), as you define rules using logic elements. It's hard to predict how it works in general and how blocks affect each other, no synchronization present, as well as no algorithm describes how to solve a task. It's funny that it works with all those ridiculous conditions.

As a non-Turing way to complete tasks it makes sense, is it fast or not - open for discussion. With my <suspicious> measurements it was impressive, but task wasn't horribly hard either.

Maybe at some point I will continue doing something with that concept, which was very curious. Can make a guess: there would be something to do with cryptography , )