A New Parallel Scheduling Algorithm Based on MPI

Abstract

GPU is widely used in many fields with the development of machine vision, but after much research that parallel algorithm in traditional is inefficient to paralleling task directly. MPI (Message Passing Interface)is used in this paper to improve computational performance, and a new task scheduling algorithm WFM (Working Forever with MPI)is proposed to improve the computational performance by reducing the idle time of each group of CPU's task and GPU's task. Our experimental results show that the same SR task can be speed up by two times with the performance of WFM on the same Jetson TX2.