非同步多核心處理器之訊息傳遞閉鎖機制設計

Message-Passing Lock Mechanism Design for Asynchronous Multicore Processor

蕭培穎(Pei-Ying Hsiao)

指導教授 ： 陳昌居

國立交通大學/資訊學院/資訊科學與工程研究所/碩士(2011年)

https://doi.org/10.6842/NCTU.2011.01014

摘要

近年來，由於單核心處理器的效能成長已趨近極限，多核心處理器架構與平行運算開始成為主流。一旦處理器的核心數變多，系統資源(如記憶體、輸入輸出等)的使用核心從只有一個變成多個，就必須有能確保資源使用時互斥的機制，以防止資源同時被多個核心存取而造成錯誤，此即所謂的鎖，而使用訊息傳遞的鎖又較使用共享記憶體的鎖更有可擴充性。現今的多核心處理器系統多使用晶片內網路連結各個核心及資源，因此可支援訊息傳遞，另因非同步系統比起需要時脈的同步系統更能節省功耗，因此本篇論文以一非同步雙道超大指令字組多核心處理器為基礎，修改其指令集架構，並於其上的晶片內網路加入一個鎖管理單元。我們將這個鎖管理單元以ModelSim 6.0模擬及驗證，使用TSMC 0.13微米的元件資料庫以Synopsys Design Compiler來做合成。

關鍵字

閉鎖 ； 共享記憶體 ； 訊息傳遞 ； 非同步電路設計

並列摘要

In recent years, because the performance growth of single chip processor has reached limit, multi-core processor architecture and parallel computing starts to become mainstream. Once the number of cores in a processor grows, the number of user of system resources(such as memory、I/O…etc) grows from only one to become more. In order to prevent errors caused by simultaneous accesses to a single resource from multi-core, a mutual exclusion mechanism of resource usage must be ensured, which is so-called lock. Using message-passing is more scalable than using shared-memory for lock design. Network-On-Chip is used to link processor-cores and resources in modern multi-core processor systems, so message-passing is supported. And asynchronous systems are more power-saving than clocked synchronous systems, so in this study, we modify the instruction set architecture based on a asynchronous two-way VLIW multi-core processor and add a lock management unit into its network. Functionality is simulated and verified by using ModelSim 6.0, and is synthesized with Synopsys Design Compiler using TSMC 0.13μm process library.

並列關鍵字

lock ； shared-memory ； message passing ； asynchronous circuit design

參考文獻

[30] C.J Chen, W.M. Cheng, H.Y. Tsai, and J.C. Wu, A Quasi-Delay-Insensitive Mircoprocessor Core Implementation for Microcontrollers, Journal of Information Science and Engineering, Vol. 25, No.2, pp. 543-557, March 2009.

[1] Edsger W. Dijkstra, “Cooperating Sequential Processes”, Technical Report, Technological University, Eindhoven, the Netherlands, 1965.

[4] Larry Rudolph and Zary Segall, Dynamic Decentralized Cache Schemes for MIMD Parallel Processors, Proceedings of the 11th Annual International Symposium on Computer Architecture, pp.340-347, June 1984.

[5] T. E. Anderson, “The Performance of Spin Lock Alternatives for Shared-

[6] J. M. Mellor-Crummey and M. L. Scott, “Algorithms for Scalable Synchronization on Shared-Memory Multiprocessors”, ACM Transactions on Computer Systems, Volume 9, Issue 1, February 1991.