The aim of this project is to implement eigenvalue problem solvers for sparse matrices in Julia in order to reduce the dependency of Julia language on ARPACK. During the GSoC program, my goal is to create a drop-in replacement for the current ’eigs’ function in pure Julia. The focus will be on nonsymmetric matrices, but if there is time, the implementation could be extended to cover symmetric matrices as well. As a part of this project, I will provide benchmarks comparing the performance of the new implementation of ’eigs’ versus the ARPACK’s implementation of ’eigs’ that is currently in use. The aim is to get this new method into the package IterativeSolvers.jl.

Organization

Student

Lauri Nyman

Mentors

  • Mike Innes
  • Harmen Stoppels
  • Christopher Rackauckas
close

2018