嵌入式多處理器系統之軟硬體分割方法論

Hardware-Software Partitioning Methodology for Embedded Multiprocessor Systems

范揚興(Yang-Hsin Fan)

指導教授 ： 李宗演

國立臺北科技大學/電資學院/電腦與通訊研究所/博士(2009年)

https://doi.org/10.6841/NTUT.2009.00160

摘要

隨著多功能、高速運算及即時服務的迫切需求，造成單一處理器架構的嵌入式系統無法滿足設計上的需求，此一趨勢的解決方案是採用運算功能強大及較具彈性的嵌入式多處理器系統進行設計，然而，嵌入式多處理器系統有軟硬體分割問題，例如，大量增加的軟硬體元件加劇軟硬體協同運作的困難度，還有，互斥的系統限制如低能源消耗與快速執行時間、高硬體配置使用率與低記憶體使用率及平行處理與處理器個數問題很難同時滿足，另外，還有系統資源配置的效率問題，此外，低能源消耗、高執行效能及高硬體資源使用率的問題，最後是軟硬體分割技術是否能快速評估各種不同系統限制的問題。
本論文中提出GHO (Genetic and Hardware-Oriented partitioning)軟硬體分割方法論來解決嵌入式多處理器系統之軟硬體分割問題。GHO分割方法包含兩個主要階段，分別是基因分割階段(Genetic partitioning phase)與硬體導向階段(Hardware-oriented phase)，應用GHO分割嵌入式多處理器系統之軟硬體元件可獲致以下優點，首先是GHO可以依照系統限制分割每一個工作為軟體或硬體元件；其次是GHO的軟硬體分割解可同時滿足多個系統限制，這些限制包含系統執行時間(Execution time)、記憶體大小(Memory size)、slice容量 (Slice capacity)、能源消耗(Energy dissipation)及處理器的個數；再者，GHO改善軟硬體資源使用率以達資源有效配置的目的；此外，GHO可獲得能源消耗較低、執行時間較快及硬體配置較高的軟硬體分割解；最後，GHO具備快速評估各種不同系統規格的軟硬體分割解的能力。
我們使用三個設計實例來驗證GHO分割方法的有效性，分別為自適應差分脈衝編解碼調製系統(ADPCM encoder/decoder system)、影像壓縮編碼系統(JPEG encoder system)及Purnaprajna benchmark，系統限制以執行時間(Execution time)、硬體配置(Slice utilization)及能源消耗(Energy dissipation)分別進行軟硬體分割實驗，GHO方法與相關方法的實驗結果比較顯示，GHO分割方法在執行時間平均減少39.47%、硬體配置使用率平均提高30.73%及能源消耗平均減少79.38%。

關鍵字

軟硬體分割 ； 嵌入式多處理器 ； 軟硬體共同設計 ； 基因分割演算法 ； 硬體導向分割演算法

並列摘要

Embedded systems have increasingly diverse functions, as well as powerful computational capabilities and real-time services, resulting in a situation in which embedded systems with only one processor can not complete designs. Alternatively, embedded multiprocessor systems provide powerful and flexible hardware and software architecture capabilities to satisfy design requirements. However, embedded multiprocessor systems have several hardware-software partitioning problems. For instance, the significantly increasing number of hardware and software tasks is difficult to coordinate when hardware and software interact with each other. Additionally, system constraints have multiple trade-off problems, ranging from low energy dissipation to fast execution time, high slice utilization to minimal memory usage, concurrence and the number of processors. Moreover, system resources are efficiently allocated after hardware-software partitioning. Furthermore, hardware-software partitioning achieves low energy dissipation, a fast execution time, and high slice utilization. Finally, a hardware-software partitioning approach can evaluate rapidly various system constraints.
This dissertation proposes a novel hardware-software partitioning methodology, Genetic and Hardware-Oriented partitioning (GHO), that can solve hardware-software partitioning problems of embedded multiprocessor systems. The GHO can determine each task to be implemented as either a hardware or software component from hundreds of thousands of hardware-software partitioning candidates. Additionally, the partitioning results of GHO can meet simultaneously the criteria of energy dissipation, execution time, memory size, slice capacity, and the number of processors. Also, GHO allocates resources efficiently for slice capacity and memory usage. Furthermore, the GHO obtains low energy dissipation, fast execution time and high slice utilization. Specifically, the GHO can rapidly assess various specifications of system constraints that enable the design of embedded multiprocessor systems to comply with time to market delivery requirements.
Three real design examples, i.e. ADPCM encoder/decoder system, JPEG encoder system and Purnaprajna benchmark, demonstrate the effectiveness of the proposed GHO. Each design example is partitioned by the system constraints of execution time, slice utilization and energy dissipation, respectively. Experimental results indicate that the proposed GHO decreases execution time by an average of 39.47%, increases slice utilization by 30.73%, as well as reduces energy dissipation by 79.38%.

並列關鍵字

Hardware-software partitioning ； embedded multiprocessor system ； hardware-software codesign ； genetic partitioning algorithm ； hardware-oriented partitioning algorithm

參考文獻

[1] P. A. Hsiung, S. W. Lin, C. H. Tseng, T. Y. Lee, J. M. Fu, and W. B. See, “VERTAF: An Application Framework for the Design and Verification of Embedded Real-Time Software,” IEEE Trans. on Software Engineering, vol. 30, no. 10, pp. 656-674, Oct. 2004.

[2] T. Y. Lee and P. A. Hsiung, “Embedded Software Synthesis and Prototyping,” IEEE Trans. on Consumer Electronics, vol. 50, no. 1, pp. 386-392, Feb. 2004.

[3] Y. L. Lin, C. P. Young and A. W. Y. Su, “Versatile PC/FPGA-Based Verification/Fast Prototyping Platform with Multimedia Applications,” IEEE Trans. on Instrumentation and Measurement, vol. 56, no. 6, pp. 2425-2434, Dec. 2007.

[4] C. L. Chuang, W. H. Cheng, D. J. Lu and C. N. J. Liu, “Hybrid Approach to Faster Functional Verification with Full Visibility,” IEEE Design & Test Computers, vol. 24, no. 2, pp. 154-162, Mar. 2007.

[8] W. Wolf, “A Decade of Hardware/Software Codesign,” IEEE Computer, vol. 36, no. 4, pp. 38-43, Apr. 2003.