Research on Parallel Algorithm for 3D Seismic Wave Simulation

With advancements in computational capabilities and algorithmic sophistication, 3D seismic simulations have become pivotal in geophysics, earthquake engineering, and disaster prevention efforts. However, the application of fast scanning methods in 3D seismic simulations presents significant challenges due to strong data dependencies, which hinder the parallelism in solving equations within the programmed functions. This limitation often precludes practical engineering applications. To overcome these constraints, we have developed an innovative approach that transcends the traditional 2D methodologies by extending these methods to three dimensions. We introduce a novel 3D diagonal direction chunking algorithm that utilizes Manhattan distance, accompanied by a foundational integral optimization algorithm. Experimental evaluations indicate that our proposed method enhances performance, achieving an optimization acceleration ratio of 1.47 on a single core. When scaled to a 56-core single node, the parallel efficiency reaches $\mathbf{4 2 \%}$. In practical engineering computations, the implementation of 56 OpenMP threads facilitates an acceleration by a factor of 23, demonstrating substantial improvements over existing algorithms.