Better Greedy Extraction Metric #75
Description
Currently our greedy extraction metric can be good, but runs into issues for example when there are LUTs which may share inputs. I believe our current greedy extraction may count these as different. ILP is unfortunately too slow right now.
This leads to ideas such as storing a map of the current seen e-nodes or something like that to thwart this double counting. Of course this isn't the only idea, there are a lot which we could try.
The goal of this issue should be to find a good greedy extraction metric (with some argument why we should choose it) or implement a system for a user to choose different metrics (or both!).