PRACE
Published October 19, 2012 | Version v1
Working paper Open

Multicore parallelization of block cyclic reduction algorithm

Creators

  • 1. IDRIS, Bat 506 BP 167, 91403 ORSAY CEDEX, FRANCE

Contributors

  • 1. IDRIS, Bat 506 BP 167, 91403 ORSAY CEDEX, FRANCE
  • 2. Laboratoire Jacques-Louis Lions, 4 place Jussieu, 75252 PARIS CEDEX 5, FRANCE

Description

The goal of this work is to evaluate how to parallelize the block cyclic reduction using MPI and OpenMP. This algorithm is used to solve elliptic problems much faster than the traditional iterative methods. We explain the parallelism that we exhibit and show the performance of the MPI and the OpenMP version that we have made.

Files

MulticoreParallelizationBlockCyclicReduction.pdf

Files (2.1 MB)

Additional details

Funding

PRACE-2IP – PRACE - Second Implementation Phase Project 283493
European Commission