[circt-synth] Implemented AIG node balancing algorithm

[Max-astro]

This implementation is based on a tree-balancing algorithm described in this paper: "https://ieeexplore.ieee.org/abstract/document/6105357".

Here is an easy-to-understand example (Which has been added as a test case):

The reference algorithm has two stages, "Covering" and "Tree-balancing," divided into maximum-and-cover and aig-balance-variadic passes.

A typical usage is: circt-synth %s --maximum-and-cover --aig-balance-variadic --cse. The maximum-and-cover pass should be always applied before the aig-balance-variadic pass.

[uenoku]

Looks really great! Thank you for working on this!! It's really nice to separate the balancing into two separate passes. What do you think replacing the existing LowerVariadic pass with your BlanceVariadic pass? I think it makes sense to me always run if it's always beneficial

[Max-astro]

I appreciate your feedback and suggestions. Thank you for taking the time to share your review!

What do you think replacing the existing LowerVariadic pass with your BlanceVariadic pass? I think it makes sense to me always run if it's always beneficial

The BlanceVariadic pass needs to calculate each node's depth during lowering, it might have runtime issues when the input design has very deep combinational logic. In those kinds of cases, using the existing LowerVariadic will be a good solution to get a usable result in less time. So, I think we should keep the LowerVariadic pass in case anyone has this kind of need.

[uenoku]

This logic is slightly unclear to me. For example I think we should consider %0 is also PO.

 %0 = aig.and_inv %port1, not %port2: i1
 %1 = aig.and_inv %0, not %port : i1
  hw.output %0, %1

[Max-astro]

You're right; I didn't consider this situation before.
Will change the for loop code like this:

    bool isPO = false;
    for (auto *user : op->getUsers()) {
      if (!isa<AndInverterOp>(user)) {
        isPO = true;
        break;
      }
    }
    if (isPO)
      po.push_back(op);

Is this a bug-free solution?

[uenoku]

I think it's necessary to check if the signal is negated.

[Max-astro]

Thanks for pointing it out. I now add an extra inverter node if the signal is negated. Is it better to push the inverter to its output if the uses are also AIG nodes? (By modifying all uses' inverted array)