Reversible Computing - Reversible Circuits

Research into reversible computing has concentrated upon a model in which inputs pass through a series of gates to yield an answer.

To be Turing complete, the set of gates must be complete (e.g. CCNOT) and we need additional constant inputs to serve as scratch memory.

To be reversible, each gate is reversible, and we need as many outputs as inputs.

All the outputs are needed to be able to reverse the calculation and reform the inputs.

However unwanted outputs (garbage bits) can be deleted once the calculation is completed. This is an irreversible step (hence must lead to entropy gain).

Some schemes minimise the number of additional wires by allowing the deletion of intermediate results during the calculation.