基于时空情境的可重构高性能任务调度体系  

Research on Task Schedule Architecture of Reconfigurable High Performance Computing Based on Time and Area Context

作　　者：涂春萍[1] 柴亚辉[1] 黄卫春[1] 熊李艳[1] 

机构地区：[1]华东交通大学信息工程学院,江西南昌330013

出　　处：《实验室研究与探索》2015年第3期116-121,197,共7页Research and Exploration In Laboratory

基　　金：江西省自然科学基金项目(2014BAB201028);华东交通大学校立基金(11XX04)

摘　　要：针对基于FPGA计算加速的异构高性能计算平台上的硬件任务调度需求,根据硬件任务的空间属性与时间属性,提出了一个基于时空情境CBTA的任务调度体系与相关算法。高性能计算平台的分级调度将CBTA的调度分为平台任务调度与节点任务调度。通过将任务与资源划分为不同的时空情境,提出了一个基于情境转化的自适应任务调度策略,并给出了CBTA调度算法的并行优化策略。最后通过实验来说明了算法在对用户响应时间、负载均衡以及任务拒绝率上的优势。According to the requirement of hardware task scheduling of heterogeneous high performance computing platform, based on FPGA computing accelerate, a task scheduling system is developed based on CBTA of time and area context. The development contains system architecture and related algorithms. The multi-level of task schedule on high performance platform divides the CBTA into platform task schedule and node task schedule. Classifying the task and resource into different time and area contexts, a self-adaptive task schedule strategy based on the resource context transform is proposed. Then a parallel optimization strategy is put forward. Finally through the experiment, the advantages of CBTA on response time, load balance and task reject ratio are shown.

关 键 词：FPGA计算加速 硬件任务调度 时空情境 自适应调度 

分 类 号：TP316[自动化与计算机技术—计算机软件与理论]