Define base primitives for digital back-end

[mtrberzi]

Issue by mtrberzi
Thursday Jul 03, 2014 at 16:37 GMT
Originally opened as manifold-lang/manifold#117

---

One point of contention will be how to represent these primitives in a way that the front-end can understand and that is compatible with how schematics are read.

I support keeping these separate from the back-end, in a JSON file similar to the schematic but that only defines node types/connection types/etc. The back-end would have a "hard-coded" representation of the types that it expects to see (and this could -- very possibly but not necessarily for our final demo -- result in automatic generation of this JSON file).

[mtrberzi]

Comment by mtrberzi
Thursday Jul 03, 2014 at 22:46 GMT

---

The primitives are probably going to be the following:

Node types:

• register (ports: in, out, clock, reset) (attributes: initial value, reset polarity, reset asynchronous, clock polarity)
• and (ports: in0, in1, out)
• or
• not (ports: in, out)
  (optionally: nand, nor, xor, xnor)
• inputPin (ports: out)
• outputPin (ports: in)

Connection types:

• digitalWire

Port types:

• digitalIn
• digitalOut

[mtrberzi]

Comment by lucaswoj
Friday Jul 04, 2014 at 14:31 GMT

---

I'm ok with everything except the exact of these primitives. At the risk of sounding pedantic, all the abbreviations and different naming conventions make this look bad.

A few high level points:

• Do we really need to indicate within identifiers that wires, ports, and pins are "digital"? I would think that everything in the digital backend would be digital. Either everything should be prefixed with "digital" (undesirable) or nothing should be.
• Register ports could be named "in", "out", "clock", and reset
• Do we have to indicate the arity of our and/or/not/xor gates? Can we assume 2 until we see the need to support higher arity gates?
• not1 ports could be named "in" and "out"
• digitalInputPin could be renamed to inputPin (same for digitalWire, digitalIn, digitalOut)

[mtrberzi]

Comment by mtrberzi
Friday Jul 04, 2014 at 14:42 GMT

---

This was a rough outline, but I appreciate the feedback. I'll definitely pick better naming conventions in release, and update that comment accordingly.

[mtrberzi]

Comment by mtrberzi
Friday Jul 04, 2014 at 14:50 GMT

---

The reason wires and ports are called "digital" is because of potential namespace clashes with, for example, mixed-signal circuits that have both digital and microfluidic inputs and outputs.

[mtrberzi]

Comment by mtrberzi
Thursday Jul 10, 2014 at 14:43 GMT

---

Okay, I will provide JSON type definitions for these.