Quantum Carry-Save Arithmetic

This paper shows how to design efficient arithmetic elements out of quantum gates using "carry-save" techniques borrowed from classical computer design. This allows bit-parallel evaluation of all the arithmetic elements required for Shor's algorithm, including modular arithmetic, deferring all carry propagation until the end of the entire computation. This reduces the quantum gate delay from O(N^3) to O(N log N) at a cost of increasing the number of qubits required from O(N) to O(N^2).