面向异构多核系统的多任务流并行编程模型设计  

Design of Multi-task Stream Parallel Programming Model for Heterogeneous Multi-core System

作　　者：唐维贤 杨锐 TANG Weixian;YANG Rui(School of Mathematics and Computer Science,Yan’an University,Yan’an 716000,China)

机构地区：[1]延安大学,数学与计算机科学学院,陕西延安716000

出　　处：《微型电脑应用》2022年第12期128-131,共4页Microcomputer Applications

摘　　要：目前的多任务流并行编程模型难以探测到异构多核系统中存的慢任务和落后任务,存在调用开销高、计算开销高和编程加速比低的问题。为此,设计了新的面向异构多核系统的多任务流并行编程模型。通过寄存器重定向、指令迁移、分离发射和线程切换4个部分组建多任务流并行编程模型,并设计了任务探测算法,探测异构多核系统中存在的慢任务和落后任务,提高了异构多核系统在多任务并行处理过程中的编程灵活性。测试结果表明,该模型的调用开销和计算开销均较低,且编程加速比高。The current parallel programming model of multi-task flow is difficult to detect the slow and backward tasks existing in heterogeneous multi-core system, which has the problems of high invocation overhead, high computation overhead and low programming acceleration ratio. Therefore, a new parallel programming model of multi-task flow for heterogeneous multi-core systems is designed in this paper. The parallel programming model of multi-task flow is established through register redirection, instruction migration, separate emission and thread switching, and the task detection algorithm is designed to detect the slow and backward tasks in the heterogeneous multi-core system, which improves the programming flexibility of the heterogeneous multi-core system in the parallel multi-task processing. The test results show that the model has low invocation and computation overhead, and high programming acceleration ratio.

关 键 词：异构多核系统 多任务流 编程模型 任务探测 线程切换 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]