ThisIf low latency is the most efficient vertical layoutmore important than space, you can go straight up in a mostly 1x2 space, with the least amounta small section of latency possible2x3:
Most of theIf space is only 1x2more important than low latency, you can replace the repeater with a small section ofvertical double NOT gate, which would lower the maximum space from 2x3 to accommodate the repeater. I anticipate this would work with your 128 block pillar.1x3: