In this paper, we introduce a parallel implementation of a numerical verification method for generalized Hermitian eigenvalue problems to compute rigorous error bounds for eigenvectors and eigenvalues in a prescribed domain. The verification method is based on the Sakurai-Sugiura method that has hierarchical parallelism and fully utilizes the performance by parallel implementation. We implement a rigorous version of the Sakurai-Sugiura method using the Julia language and evaluate its performance. This paper provides the development in the fields of mechanics and electromagnetics etc. because it reduces the computation time, while computes error bounds rigorously for eigenvectors and eigenvalues.

错误:搜索内容不能为空,请输入英文关键词
错误:关键词超出字数限制,请精简
高级检索

Parallel Implementation of a Verified Contour Integral Based Eigensolver and Its Performance Evaluation Using the Julia Language

  • Shota Seto,
  • Akitoshi Takayasu

摘要

In this paper, we introduce a parallel implementation of a numerical verification method for generalized Hermitian eigenvalue problems to compute rigorous error bounds for eigenvectors and eigenvalues in a prescribed domain. The verification method is based on the Sakurai-Sugiura method that has hierarchical parallelism and fully utilizes the performance by parallel implementation. We implement a rigorous version of the Sakurai-Sugiura method using the Julia language and evaluate its performance. This paper provides the development in the fields of mechanics and electromagnetics etc. because it reduces the computation time, while computes error bounds rigorously for eigenvectors and eigenvalues.