Here is a theoretical method I plan on using, and I'd like to know what you think of it:
Upon the base assumption that most lag in DF computers comes from the flowing of water in typical logic gates and secondmost from active pumps checking for liquid in their source tile, I have tried to eleminate those two features from my design as much as possible. Indeed, I have already designed an adder (with carry) that has only one pump and has no flowing water when not changing state (and it keeps its state unless told to reset). For instance, I use an OR gate that conists of a hatch directly above a floodgate. This style also uses less power.
In order to effect this, I designed essentially a quantum computer; that is, a computer designed to mesure and utilize individual quanta of water. A quanta of water is a single full tile, and since there is only one tile full there is no flowing. I can fill a quanta with an inf water source emptying into the single tile via hatch or pump, and to delete a quanta I simply pulse a door or floodgate next to it, which turns the quanta into a half-quanta. Also, by having the state of either quanta or half-quanta, I can more easily generate NOT gates because I can set the "LOW" state to output an "OPEN" signal.
There are some drawbacks, for instance I have to manually send "CLEAR" commands instead of merely having it shut off when the input stops, and I actually have to pulse the clear gates or that extra water level will bounce around. I am also in need of a good pulse generator that will clear its state when a signal ceases to be input yet not involve flowing water. The activation signal for the pulse generator can be either "OPEN" or "CLOSE", I can provide either.