Why can't we just extend the Shift module and make S=1? I would also rename the parameters so that N is the number of inputs and D is the depth of the shift register

Oops, didn't realize that would type check. Fixed now

Nice! As we start building this higher-level abstractions, we should also think of some optimizations to do at the Filament level. For example, ConcatWire and SplitWire definitions can be inlined into Shift for basically no cost and for the N=1 case, eliminate a bunch of wires. Those wires are probably not that expensive to begin with but still worth thinking about building up optimization infrastructure.