Implementation of the BiCGSTAB method for the Helmholtz Equation on a Maxeler Data Flow Machine

Title

Implementation of the BiCGSTAB method for the Helmholtz Equation on a Maxeler Data Flow Machine

 Author

Meijers, Onno (TU Delft Electrical Engineering, Mathematics and Computer Science; TU Delft Delft Institute of Applied Mathematics)

Contributor

Vuik, Kees (mentor)
Gaydadjiev, Georgi (graduation committee)

Degree granting institution

Delft University of Technology

 Date

2017-08-25

 Abstract

To get a clear picture of the earth crust the Helmholtz equation needs to be
solved and a high wave number is needed in order to provide fine grain details.
The Stabilised BiConjugate Gradient method needs to be implemented to solve
the Helmholtz equation. This is done on a Maxeler data flow machine to improve
calculation times. The algorithm is split into 3 parts and all of them use the same
calculation kernel. An improvement in calculation time of 2.4 times faster than in
the literature is achieved and a two times faster calculation is expected with more
modern hardware. Maxeler’s data flow machine shows great possibilities to decrease calculation times for high performance computing problems.

 Subject

Maxeler
Data Flow
Helmholtz equation
BiCGSTAB

 To reference this document use:

http://resolver.tudelft.nl/uuid:adc4ad46-230f-4c29-be9e-6e914e8ea452

 Part of collection

Student theses

 Document type

master thesis

 Rights

© 2017 Onno Meijers
Files
PDF thesis_Onno_Meijers_25_08_17.pdf 1.02 MB
Close viewer