Accelerating Sparse Linear Solvers for Reservoir Simulations

This technical report discusses the development and implementation of SolverBR, a high-performance sparse linear solver specifically designed for reservoir simulation applications. The report outlines the motivation behind SolverBR, which arose from the need to enhance the speed of reservoir studies and improve prediction accuracy while reducing uncertainties associated with simulation times. It details the characteristics of reservoir problems that can be exploited to achieve superior performance compared to conventional solvers. The report describes the technical specifications of SolverBR, including its implementation in C++ on Linux, and its support for various parallelism models. It also highlights the integration of SolverBR with other computational tools and architectures, including ARM, to optimize performance and cost-efficiency in high-performance computing environments. The report concludes with a discussion of the results from experiments conducted using SolverBR on various reservoir models, demonstrating its effectiveness in handling complex reservoir simulations.