This paper introduces a new synthesis approach called Products of EXOR-sums (POE) and demonstrates its utility when employed in an Exclusive-Or-Sum-of-Products (ESOP) based algorithm for reversible and permutative quantum circuits restricted to single output functions. Unlike previous ESOP approaches such as EXORCISM-4 which realize functions as an EXOR-ed sum of product literals, this new algorithm realizes functions as an EXOR-ed sum of POEs (EPOE). A comparison of EXORCISM-4 and EPOE circuits shows that the latter approach often produces circuits with significantly lower quantum costs.
This paper introduces a new synthesis approach called Products of EXOR-sums (POE) and demonstrates its utility when employed in an Exclusive-Or-Sum-of-Products (ESOP) based algorithm for reversible and permutative quantum circuits restricted to single output functions. Unlike previous ESOP approaches such as EXORCISM-4 which realize functions as an EXOR-ed sum of product literals, this new algorithm realizes functions as an EXOR-ed sum of POEs (EPOE). A comparison of EXORCISM-4 and EPOE circuits shows that the latter approach often produces circuits with significantly lower quantum costs.
This paper introduces new algorithms to synthesize reversible functions using EXOR-sum of Products-of-EXOR-sums (EPOE) structures. The motivation for using these structures is to reduce the number of as well as the sizes of multiple controlled Toffoli gates, and thus the quantum cost. To achieve these reductions the paper generalizes from existing 2-level AND-EXOR structures (ESOP) commonly used in reversible logic to a mixture of 3-level EXOR-AND-EXOR structures and ESOPs. Our approach can be applied to reversible and permutative quantum circuits to synthesize single output functions on to an output line, with no additional ancilla bits. A comparison of the ESOP minimizer EXORCISM-4 and two variants of the EPOE minimizer, called EPOEM-1s and EPOEM-2, is presented. The results show that EPOE circuits do in fact achieve the above-stated cost reductions, in particular when expressed in terms of Maslov's quantum cost, the metric commonly used in quantum circuit synthesis.
Previously, the synthesis of reversible and quantum permutative circuits based on products of Exclusive-OR (EXOR) sums of inputs (POE) was explored as an approach to reduce quantum cost. Herein, the mathematical representation used in POE expressions is formally detailed and contrasted with the pseudoproduct-based theory of Luccio and Pagli. The theory and practice of product transformation, or computation of equivalent POE expressions, is presented. To demonstrate the generality of the mathematical representations outside of the reversible domain, a new method to synthesize incompletely specified functions as EXOR-AND-OR circuits is introduced. The method is heuristic. It first uses spectral information from the Walsh-Hadamard transform to compute a set of POE implicants for a function, and then uses product transformation to minimize both the total number of literals and unique EXOR sums (ignoring polarity) used in the POE implicants. The product transformation minimization extends to multioutput functions. The new method has no performance guarantees, i.e., it does not have a fallback mode of minimum or near-minimum sum of products; therefore, it is suited to a tournament synthesis approach. When synthesizing benchmark functions of up to 26 inputs, the new method is shown to typically produce lower cost circuits than heuristic pseudoproduct-based methods.
Today's rapid advances in quantum computing hardware call for scalable synthesis methods to map combinational logic represented as multi-level Boolean networks (e.g., an and-inverter graph, AIG) to quantum circuits. Such synthesis process must yield reversible logic function since quantum circuits are reversible. Thus, logic representations using exclusive sum-of-products (ESOP) are advantageous because of their natural relation to Toffoli gates, one of the primitives in reversible logic. This motivates developing effective methods to collapse AIG logic networks into ESOPs. In this work, we present two state-of-the-art methods to collapse an AIG into an ESOP expression, describe their shortcomings and introduce a new approach based on the divide-and-conquer paradigm. We demonstrate the effectiveness of our method in collapsing IEEE-compliant half precision floating point networks. Results show that our method can collapse designs-which were previously not solvable within a week-in less than 5 minutes. We also describe a technique capable of taking advantage of this new method to generate quantum circuits with up to 50% fewer T gates compared to state-of-the-art methods.
Emerging technologies with asymptotic zero power dissipation, such as quantum computing, require the logical operations to be done in a reversible manner. In recent years, the problem of synthesizing Boolean functions in the reversible logic domain has gained significant research attention. The efficiency of the synthesis methods is measured in terms of quantum cost, gate cost, garbage lines, logic depth and speed of synthesis. In this paper, we present an approach based on Exclusive sum-of-Products (ESOP), which allows the user to explore the trade-off between quantum cost and garbage lines. The proposed technique adds a new dimension to the reversible logic synthesis solutions. We demonstrate by detailed experiments that controlled improvement in quantum cost and gate count by increasing garbage count can be achieved. In some cases, improved quantum cost and gate count compared to state-of-the-art synthesis methods are reported. Furthermore, we propose a novel rule-based approach to achieve ancilla-free reversible logic synthesis starting from an ESOP formulation.
The Exclusive-OR Sum-of-Product (ESOP) minimization problem has long been of interest to the research community because of its importance in classical logic design (including low-power design and design for test), reversible logic synthesis, and knowledge discovery, among other applications. However, no exact minimal minimization method has been presented for more than seven variables on arbitrary functions. This paper presents a novel quantum-classical hybrid algorithm for the exact minimal ESOP minimization of incompletely specified Boolean functions. This algorithm constructs oracles from sets of constraints and leverages the quantum speedup offered by Grover's algorithm to find solutions to these oracles, thereby improving over classical algorithms. Improved encoding of ESOP expressions results in substantially fewer decision variables compared to many existing algorithms for many classes of Boolean functions. This paper also extends the idea of exact minimal ESOP minimization to additionally minimize the cost of realizing an ESOP expression as a quantum circuit. To the extent of the authors' knowledge, such a method has never been published. This algorithm was tested on completely and incompletely specified Boolean functions via quantum simulation.
This paper introduces a new logic structure for reversible quantum circuit synthesis. Our synthesis method aims to minimize the quantum cost of reversible quantum circuits with decoders. In this method, multi-valued input, binary output (MVI) functions are utilized as a mathematical concept only, but the circuits are binary. We introduce the new concept of ``Multi-Valued Input Fixed Polarity Reed-Muller (MVI-RM)" forms. Our decoder-based circuit uses three logical levels in contrast to commonly-used methods based on Exclusive-or Sum of Products (ESOP) with two levels (AND-XOR expressions), realized by Toffoli gates. In general, the high number of input qubits in the resulting Toffoli gates is a problem that greatly impacts the quantum cost. Using decoders decreases the number of input qubits in these Toffoli gates. We present two practical algorithms for three-level circuit synthesis by finding the MVI-FPRM: products-matching and the newly developed butterfly diagrams. The best MVI-FPRM forms are factorized and reduced to approximate Multi-Valued Input Generalized Reed-Muller (MVI-GRM) forms.