Implement generalized circuits

[engboris]

There're attempts here at describing boolean circuits: https://github.com/engboris/stellogen/blob/master/examples/circuits.sg

This can easily be converted to arithmetic circuits by changing the "semantic" plugged to doors.

The problem is that doors wait for their inputs to be computed before continuing. We need a nice way to enforce such a synchronization. Currently, computation with constellation is very "free". Constellations will try to do weird things when computation is not guided. Sometimes looping forever when they could just wait for a response from the inputs. They are impatient.

[jpt4]

For synchronization, asynchronous logic has many ideas to mine:
https://en.wikipedia.org/wiki/Asynchronous_circuit

In particularly, the Communicating Hardware Processes calculus:
https://www.cs.yale.edu/flint/cs428/notes/chp_introduction.pdf

[engboris]

My problem is quite simple.

'inputs
-0(X) +c0(X);
-0(X) +c1(X).
'layer 1
-c0(X) -not(X R) +c2(R);
-c1(X) -not(X R) +c3(R).
'layer 2
-c2(X) -c3(Y) -and(X Y R) +c4(R).
'output
-c4(R) R.

I have stars corresponding to doors. They're like hyperedges for a hypergraph representing a circuit. Doors then call logical operations outside.

The problem is that it can produce a very chaotic tiling of doors. When stars cannot directly compute they will do symbolic reasoning and "guess", leading to chaotic choices.

Layered computation (occurring in circuits) is an enforced order of computation.

[engboris]

For synchronization, asynchronous logic has many ideas to mine: https://en.wikipedia.org/wiki/Asynchronous_circuit

In particularly, the Communicating Hardware Processes calculus: https://www.cs.yale.edu/flint/cs428/notes/chp_introduction.pdf

Looks interesting! I'll take a look at it :)

[engboris]

I just remembered about boolean proof-nets: https://www.kurims.kyoto-u.ac.jp/~terui/pn.pdf

This is something I found fascinating when I was a student. This could be a possible path.