High-Performance Multithreaded Acceleration of High-Order Discrete Racah Polynomials on Multi-Core Architectures

Discrete Racah polynomials (DRPs) are considered essential for several applications such as approximation theory, quantum physics, and digital signal processing. However, the computation of high-order DRPs presents significant numerical and computational challenges. Even though the recent advancements in algorithms have improved numerical stability, the inherently sequential nature of these algorithms continues to be a bottleneck for large-scale implementations. This paper proposes a new multithreading-based algorithm for computing the DRP coefficients. The proposed algorithm is designed to accelerate the computation of high-order DRPs by considering the advantages of the independence of different tasks and the ability to process multiple coefficients simultaneously. The algorithm introduces distinct multithreading strategies for both zero-valued and non-zero-valued parameters, which include the zero-parameter balanced algorithm (ZPBA), the zero-parameter unbalanced algorithm (ZPUA), and the valued-parameter algorithm (VPA). The experimental results show that the proposed multithreading algorithms noticeably reduce computation time when compared to the state-of-the-art baseline algorithm. Notably, both ZPBA and VPA exhibit exceptional scalability across modern multicore architectures, delivering speedup factors exceeding 10× for large polynomial sizes. This work effectively bridges the gap between mathematical theory and high-performance computing, making the generation of large-scale DRP matrices computationally viable for advanced engineering applications.