Time/Space Trade-Offs for Reversible Computation
Charles H. Bennett · SIAM Journal on Computing · 1989
A reversible Turing machine is one whose transition function is $1:1$, so that no instantaneous description (ID) has more than one predecessor. Using a pebbling argument, this paper shows that, for any $\varepsilon > 0$, ordinary multitape Turing machines using time T and space S can be simulated by reversible ones using time $O(T^{1 + \varepsilon } )$ and space $O(S\log T)$ or in linear time and space $O(ST^\varepsilon )$. The former result implies in particular that reversible machines can simulate ordinary ones in quadratic space. These results refer to reversible machines that save their input, thereby insuring a global $1:1$ relation between initial and final IDs, even when the function being computed is many-to-one. Reversible machines that instead erase their input can of course compute only $1:1$ partial recursive functions and indeed provide a Godel numbering of such functions. The time/space cost of computing a $1:1$ function on such a machine is equal within a small polynomial to the cost of computing the function and its inverse on an ordinary Turing machine.