【摘要】 复杂嵌入式应用领域的具体计算任务不仅需要计算平台具有嵌入式高效能计算能力,而且应具有与应用特点匹配的计算资源结构;面向复杂嵌入式应用领域研究了众核处理机体系架构,在GPU基本计算形态的基础上,依据NVIDIA Kepler架构的特性,进一步研究、构建了相应的众核组织与处理模式,将其形成3种基本计算模式:单任务并行计算、多任务并行计算、多任务流处理计算,有效地提高了嵌入式高性能计算的效能和灵活性。更多还原

【Abstract】 The computing tasks of complex embedded application area not only need the ability of embedded high-performance computing,but also need the structure of computing resource which matches the feature of the application.The architecture of many-core computer for complex embedded system is presented in this paper.Furthermore,this paper researches and constructs the organization and processing mode of many-core processor based on the NVIDIA Kepler architecture,brings forward three basic compute modes,single task parallel computing,multi-task parallel computing and multi-task stream processing.This is beneficial to improve the efficiency and flexibility of the embedded high-performance computing.更多还原