Logo image
Programming Model and System Software for embedded Multicore Processors
Dissertation

Programming Model and System Software for embedded Multicore Processors

Hsieh, Brian Kun-Yuan
Doctor of Philosophy (PHD), 國立清華大學, 資訊工程學系
2009

Abstract

程式設計模型 多核心系統 平行程式核心 作業系統 函式庫 系統工具軟體 資料串流技術 programming model multicore architecture parallel programming operating system design library design system software data streaming technology
Trends of embedded system development recently point toward multicore architecture with increasingly parallel hardwares. Efficient system software is becoming a critical aspect of developing multicore applications for embedded systems. There are challenges associates to the system software design and implementation for embedded multicore processors, effective programming model, efficient runtime system, and testing of concurrent softwares. This dissertation tackles the challenges of designing and implementing high performance system software for multicore processors. First, it examines the issues of parallelizing application on multicore processors and then presents an effective programming model to extract the parallelism in the applications. The programming environment considers not only the programmability but also the performance and maintenance of the applications. The programming model contains a data streaming technique for optimizing multimedia applications to reduce the communication overhead. The programming environment also contains design patterns that demonstrates the usage of the programing model. Second, this dissertation illustrates the design and implementation of the operating system and communication library. The operating system adapts a technique of reducing the context switch overhead during task switching for optimizing the performance of operating system. Finally, this dissertation introduce an automatic testing and verification tool, pTest, for testing the concurrent multicore softwares. pTest contains a formal representation to generate testing patterns for testing the runtime execution behavior. The experiments demonstrate the testing tool by evaluating the development flow of the programing environment. System software is a key factor to the efficiency of embedded multicore applications. The technique presented in this article address the issues of designing the programming environment, optimizing the overhead of runtime system, and verifying the current applications. This dissertation presents the design and implementation of efficient system software for multicore processors and discusses issues and solution of building efficient system software for multicore processors.

Metrics

1 Record Views

Details

Logo image