Code Parallelization for Multi-Core Software Defined Radio Platforms with OpenMP |
| |
Authors: | Michael Schwall Stefan Nagel Friedrich K Jondral |
| |
Affiliation: | (1) Karlsruhe Institute of Technology, Karlsruhe, Germany;(2) GIGATRONIK Technologies GmbH, Ulm, Germany |
| |
Abstract: | Since the number of processing cores in a General Purpose Processor (GPP) increases steadily, parallelization of algorithms
is a well known topic in computer science. Algorithms have to be adapted to this new processor architecture to fully exploit
the available processing power. This development equally affects the Software Defined Radio (SDR) technology because the GPP
has become an important processor for SDR platforms. To make use of the entire processing power of a multi-core GPP and hence
to avoid system inefficiency, this work provides an approach to parallelize C/C+ + code using OpenMP. This application programming
interface provides a rapid way to parallelize code using compiler directives inserted at appropriate positions in the code.
The processing load can be shared between all available cores. We use Matlab Simulink as a framework for a model-based design
and evaluate the processing gain of embedded handwritten C-code blocks with OpenMP support.We will show that with OpenMP the
core utilization is increased. Compared to a single-core GPP, we will present the increase of the processing speed depending
on the number of cores. We will also highlight the limitations of code parallelization. In our results, we will show that
a straightforward implementation of algorithms without multi-core consideration will cause an underutilized system. |
| |
Keywords: | |
本文献已被 SpringerLink 等数据库收录! |
|