Performance evaluation of preconditioning method for in-body current density analysis using numerical human model

抄録

Iterative methods such as the COCG method and the COCR method take too long to solve large-scale linear equations obtained by the SPFD method, which is a current density analysis method that uses a numerical human model. Therefore, in the present study, in order to accelerate these methods, we introduce iterative methods with various preconditioning methods. The results indicate that the block decomposition method can improve the parallelism of the IC and SSOR preconditioning. In addition, SSOR preconditioning and block SSOR preconditioning for a single thread and block IC decomposition preconditioning for multiple threads are the fastest methods used in the present study.