Evaluation of parallelization of an image processing algorithm for an embedded multicore platform using manual parallelization and the OpenMP parallel framework

Author

Prokesch, Richard

Author_Institution

Safety & Security Dept., Austrian Inst. of Technol. GmbH, Vienna, Austria

fYear

2013

fDate

10-13 Nov. 2013

Firstpage

2256

Lastpage

2260

Abstract

In this paper manual parallelization is compared with parallelization using the Open Multi-Processing (OpenMP) software framework on an embedded eight core Digital Signal Processor (DSP) developed by Texas Instruments. In order to compare these two methods an in-house developed 2D image registration algorithm is parallelized. Furthermore considerations regarding the parallelization on this DSP are presented. Finally both parallelization methods in respect of performance and real time behavior are measured and analyzed.

Keywords

digital signal processing chips; embedded systems; image registration; multiprocessing systems; parallel processing; DSP; OpenMP parallel framework; Texas Instruments; embedded eight core digital signal processor; embedded multicore platform; image processing algorithm parallelization; in-house developed 2D image registration algorithm parallelization; manual parallelization; open multiprocessing software framework; Digital signal processing; Image registration; Manuals; Message systems; Multicore processing; Vectors; OpenMP; digital signal processor; embedded system; image processing algorithm; manual parallelization; performance comparison;

fLanguage

English

Publisher

ieee

Conference_Titel

Industrial Electronics Society, IECON 2013 - 39th Annual Conference of the IEEE

Conference_Location

Vienna

ISSN

1553-572X

Type

conf

DOI

10.1109/IECON.2013.6699482

Filename

6699482