Task Partition Comparison between Multi-core System and GPU

Abstract

parallel computing becomes more and more popular and the scale of software improves accordingly. A large application is made up of many small tasks. Some statistics indicate that some tasks of 80% software can support parallel execution. The difficulties are how to decompose one application into many tasks and exploit the parallelization among these tasks. In this paper, we compare some types of parallel programs based on multi-core system and GPU. From the experiment results, some advice is given about how to decompose one application into different tasks based on different computing architecture.