讨论了在SMP机群系统上用于通信优化的并行任务调度问题 .首先对该问题进行了形式化描述 ,并将其转化为图划分问题MMP ,证明了该问题是NP完全问题 ,然后提出了启发式算法MMP Solver.实验结果表明 ,针对SMP机群系统对并行程序进行任务调度用以优化通信是很有意义的 ,MMP Solver的优化效果优于现有算法 The parallel task scheduling problem for communication optimization in SMP cluster system is discussed.Firstly, the problem is formally described and transformed into a map partitioning problem matrix (MMP), which proves that the problem is NP-complete and then proposes the heuristic Algorithm MMP Solver. The experimental results show that it is significant for SMP cluster system to schedule tasks for parallel programs to optimize communication. The optimization of MMP Solver is superior to the existing algorithms