Master Thesis

FZJ-2021-00911

http://join2-wiki.gsi.de/foswiki/pub/Main/Artwork/join2_logo100x88.png Parareal in Julia with Singularity

Heisters, I. (Corresponding author)FZJ*

2020

Please use a persistent id in citations: http://hdl.handle.net/2128/27103

Abstract: In this thesis a Julia version of the parallel-in-time method Parareal is introduced. Parareal decomposes the time as an approach for parallelization. Parareal is an hierarchical, iterative algorithm that uses a coarse, cheap integrator to propagate information quickly forward in time in order to provide initial values for the parallelized original time integration scheme. The problems here used to test the Parareal algorithm are the Lorenz equation, the heat equation and the Allen-Cahn equation. Julia is a programming language that was specifically designed to be used for numerical applications and parallelization. Julia is becoming more popular due to the fact that it is easier to implement than C but has a better runtime than Python. Julia is a new language and not available on most host systems. Singularity is a container solution to create the necessities for scientific application-driven workloads. By using a container the user can configure the environment in which the application can run independently of the host system and its software specifications.This thesis shows how a Singularity container for the Parareal algorithm implemented in Julia can be built. The container is portable to different hosts like HPC and cloud systems without having a runtime overhead in comparison to the runtime without a container.

Note: Masterarbeit, FH Aachen, 2020

---

Contributing Institute(s):

1. Jülich Supercomputing Center (JSC)

Research Program(s):

1. 511 - Computational Science and Mathematical Methods (POF3-511) (POF3-511)
2. DFG project 450829162 - Raum-Zeit-parallele Simulation multimodale Energiesystemen (450829162) (450829162)

Appears in the scientific report 2021

---

Database coverage:

---