Reversible computing

Reversible computing is a model of computation in which every computational operation is logically reversible — meaning that each step can be uniquely undone, and no information is ever erased. In 1973, Charles Bennett proved that any classical computation can be performed reversibly, provided the computation retains all intermediate results and then "uncomputes" them by running the logic backward. This eliminates the thermodynamic cost of information erasure that Landauer's principle imposes on irreversible computation.

The tradeoff is not energy versus accuracy, but energy versus memory. A reversible computation requires storage proportional to its depth — all intermediate states must be preserved until they can be uncomputed. The thermodynamic cost of a computation thus becomes a design choice: the architect decides how much heat to dissipate by choosing how much to remember and how much to forget.

Reversible computing has direct relevance to quantum computing, where unitary evolution is inherently reversible. The theoretical foundations — including the Fredkin gate and the Toffoli gate, universal reversible logic gates — were laid by Edward Fredkin and Tommaso Toffoli in the 1970s and 1980s. These gates establish that any Boolean function can be computed without information loss, making reversibility a universal property of classical computation, not a special case.

The practical question is whether the energy savings of reversibility can ever be realized at scale. Current silicon technology dissipates roughly 10⁶ times the Landauer limit per operation, so reversible design is irrelevant to today's engineering. But at the fundamental limit — as computing approaches the atomic scale and the single-electron regime — the Landauer bound becomes unavoidable, and reversible architecture may be the only path forward.