论文部分内容阅读
针对嵌入式片上多处理器MPSoC(multiple processor system on chip)平台下任务并行化分配的问题,从理论上对任务调度进行了建模,针对模型中的任务间依赖问题,给出了层次任务图的分析模型;结合量化后函数的开销和OpenMP并行化思想,提出了基于复制分治调度策略的并行方案;以例子滤波算法为例,对任务并行化进行验证,实验结果分析表明,本文提出的并行化方案,合理的对任务进行分配,改善了多处理器的负载平衡,降低了处理器间通讯开销,具有较大的加速比,满足嵌入式多核平台下任务并行化的需求.
In order to solve the problem of parallel assignment of tasks on embedded multiprocessor MPSoC platform, the task scheduling is modeled theoretically. In order to solve the problem of interdependent tasks in the model, a hierarchical task graph Based on the analysis of the post-quantification function and the OpenMP parallelism, a parallel scheme based on replication-based scheduling strategy is proposed. Taking the example filter algorithm as an example, the task parallelization is verified. The experimental results show that the proposed scheme Parallelization scheme can distribute the tasks reasonably, improve the load balance of multi-processor, reduce the communication overhead between processors, have a large speed-up ratio and meet the demand of parallelization of tasks under embedded multi-core platform.