Parallel Quick Sort Algorithms Analysis using OpenMP 3.0 in Embedded System

A multi-core system is commonly used in the general PC (Personal Computer) environment. There is a growing need for multi-core system in embedded environments with increasing usage of multi-core systems. Parallel programming is an important issue of a multi-core system, since it enables full use of the core system. Especially, OpenMP API provides an easy low-burdensome method for threading applications. However, it is more difficult to develop parallel programming in the embedded level due to the memory size limitation and number of cores. For example, limitation of cache memory and data sharing architecture induce frequent data movement overhead between cache and main memory. Use of an excessive number of threads, regardless of number of cores, induces the 'Heavy Contention' problem. In this paper, we parallelize the a lot of kinds of quick sort algorithm using OpenMP 3.0 in the OMAP-4430 embedded environment and analyze the effect of parallelization considering memory size and number of cores with the 'GNU Gprof' profiler.