Memory Mapping and Task Scheduling Techniques for Computation Models of Image Processing on Many-Core Platforms

Many-core technology is proposed as a solution to improve the performance of modern computer systems. To obtain good performance on a many-core system, exploiting parallelism in arithmetic level is not enough. Due to the contention of shared hardware resource, the speedup ratio of a many-core system is usually much lower than the number of processor units. In this paper, the contention of shared memory resource is addressed. An algorithm is developed to perform memory mapping and task scheduling for many-core systems. According to experimental results, the proposed algorithm can effectively improve the performance by 64.77% in average. 43.55X speedup ratio can be achieved when 48 processor units are activated. The performance loss ratio is less than 10%.