On the Complexity of the Evaluation of Transient Extensions of Boolean Functions

Transient algebra is a multi-valued algebra for hazard detection in gate circuits. Sequences of alternating 0’s and 1’s, called transients, represent signal values, and gates are modeled by extensions of boolean functions to transients. Formulas for computing the output transient of a gate from the input transients are known for NOT, AND, OR} and XOR gates and their complements, but, in general, even the problem of deciding whether the length of the output transient exceeds a given bound is NP-complete. We propose a method of evaluating extensions of general boolean functions. We introduce and study a class of functions with the following property: Instead of evaluating an extension of a boolean function on a given set of transients, it is possible to get the same value by using transients derived from the given ones, but having length at most 3. We prove that all functions of three variables, as well as certain other functions, have this property, and can be efficiently evaluated.

💡 Research Summary

The paper addresses the computational difficulty of evaluating transient extensions of Boolean functions, a problem that arises in hazard detection for digital gate circuits. In transient algebra, a signal that alternates between 0 and 1 is modeled as a finite word called a transient. Logical gates are represented by extending ordinary Boolean functions to operate on transients, producing an output transient from a set of input transients. While closed‑form formulas exist for the basic gates NOT, AND, OR, XOR and their complements, the general problem of determining the length of the output transient (or whether it exceeds a given bound) is shown to be NP‑complete. This result follows from a polynomial‑time reduction from SAT, establishing that dynamic hazard analysis cannot be performed efficiently for arbitrary Boolean functions using naïve simulation.

To overcome this barrier, the authors introduce a new class of Boolean functions that possess a “length‑3 reducibility” property. A Boolean function f is said to be length‑3 reducible if, for any collection of input transients, one can replace each input by a derived transient of length at most three—while preserving the exact output transient produced by the extension of f. In other words, the original (potentially very long) input transients can be compressed to short representatives without changing the result of the transient extension.

The core theoretical contribution is a proof that all three‑variable Boolean functions satisfy this property. The proof proceeds by exhaustive examination of the eight possible input assignments, constructing for each case a set of length‑≤3 transients that emulate the behavior of any longer input. The authors also identify broader families of functions—such as symmetric functions and certain compositions of AND/OR—that are also length‑3 reducible under specific structural conditions.

Based on this property, the paper proposes an evaluation algorithm with the following steps:

1. Normalization – each input transient is transformed into a canonical transient of length ≤3. The transformation respects logical equivalence and can be performed in linear time with respect to the number of variables.
2. Lookup/Computation – for the normalized inputs, the output transient is obtained either from a pre‑computed lookup table (feasible because the domain size is bounded by 3^k) or by evaluating a polynomial‑time expression derived from the Boolean function’s algebraic normal form.
3. Reconstruction – the final output transient is assembled directly from the result of step 2; no further expansion is required.

The algorithm’s runtime is O(k), independent of the original lengths of the input transients, and its space requirement is polynomial in the number of variables. Consequently, for any length‑3 reducible function, the previously intractable hazard‑analysis problem becomes efficiently solvable.

The authors validate their approach theoretically by establishing upper bounds on the output transient length (linear in the number of inputs) and empirically by simulating a suite of benchmark circuits. In the experiments, the proposed method achieves speed‑ups of an order of magnitude compared with conventional transient simulation, while producing identical output transients.

The paper concludes by acknowledging that not all Boolean functions are known to be length‑3 reducible; a complete classification remains open. Future work includes extending the reducibility concept to longer bounded lengths, integrating the technique into commercial EDA tools, and exploring its applicability to multi‑clock and asynchronous designs where transient interactions are more complex.

Overall, the work provides a significant theoretical insight—identifying a broad class of functions for which transient evaluation is tractable—and translates this insight into a practical algorithm that can be directly applied to improve dynamic hazard detection in digital circuit design.