This is an initial cycles solve at 8l 4r - the latency feels very improvable by removing Four Billion Regrabs but also the rejection/projection is very limiting in cycles

will probably optimize more. if this is being read on stream then got lazy.

predict that min is 6 latency and 26 cycles

edit: did not optimize more. tried but didn't commit enough time to it. pretty sure min is 7 latency and 27 cycles because rejection+projection is slow.