| 研究生: |
李易修 Lee, I-Shiou |
|---|---|
| 論文名稱: |
使用於AMBA匯流排系統之橋接器之設計與分析 Design and analysis of a bridge system for AMBA bus |
| 指導教授: |
陳中和
Chen, Chung-Ho |
| 學位類別: |
碩士 Master |
| 系所名稱: |
電機資訊學院 - 電機工程學系 Department of Electrical Engineering |
| 論文出版年: | 2004 |
| 畢業學年度: | 92 |
| 語文別: | 中文 |
| 論文頁數: | 128 |
| 中文關鍵詞: | 緩衝記憶體 、橋接器 |
| 外文關鍵詞: | MPEG, burst length, buffer size, AHB-PCI bridge |
| 相關次數: | 點閱:81 下載:2 |
| 分享至: |
| 查詢本校圖書館目錄 查詢臺灣博碩士論文知識加值系統 勘誤回報 |
本論文針對應用層的不同行為模式,將一個AHB對PCI匯流排的橋接器(AHB-to-PCI Bridge)以優化緩衝記憶體大小的方式,提出改善傳輸效能的方案。我們將應用層的行為模式分為二大類:多媒體傳輸和隨機(一般)傳輸。在多媒體傳輸模式下,AHB和PCI兩側的諸多MASTER有其固定的頻寬要求;而隨機傳輸模式則是描述一般系統的行為。
我們的實驗結果顯示,在多媒體傳輸模式下,MASTER的行為和緩衝區大小影響傳輸效能,當緩衝區大小(至少)等於MASTER的BURST LENGTH時可以得到最佳效能。在隨機傳輸模式下,效能則受到系統讀取和寫入動作比例的影響:寫入動作的比例越高,則效能正比於緩衝區大小,我們發現最小給定二倍於BURST LENGTH的緩衝區可以在效能和成本上帶來最佳效益。另外,對於二種模式而言,橋接器兩側的工作頻率比例不同時,寫入動作的比例對系統的影響也不同。本論文在架構上提出一種整合讀取和寫入緩衝區的方案,使得兩者共用同樣的記憶體,進而縮小晶片面積,並且同時使用時脈閂的技術來降低能源消耗。
This thesis proposes schemes that optimize buffer usage in an AHB-to-PCI bridge for different application models. Application behaviors are classified into two major models: multi-media and random accesses. In the multi-media model, transfer bandwidths of masters tend to be fixed and predictable. The random access model describes a general-purpose system behavior.
The results of our experiment show that the master behavior and buffer size dominate the performance in the multi-media model. Optimal performance can be obtained if a buffer size is the same as the burst length. In the random access model, performance is dominated by the write ratio of the system. Larger buffer is required when write ratio increases, and the buffer size should be kept at twice the size of the burst length in any case to maintain optimal performance. For both models, our experiment also shows that the write ratio affects system performance in different ways when the clock rate at either side of the bridge changes. This thesis proposes a special scheme that combines read and write buffers into one single memory space and also introduces clock gating technology, which helps to save die area and energy consumption.
[1] Open Core: http://www.opencores.org/.
[2] Peter Luoma, “Resolving Access Conflicts & Refining Incompatible Interface,” G612 Hardware/Software Codesign, Chapter 8.4-8.5.1.
[3] Chung-Ho Chen and Arun K.Somani, “An Easy to Use Approach for Practical Bus-Based System Design,”IEEE Trans. Computers, vol.48, no. 8, pp. 780-793, August 1999.
[4] Mary K. Vernon and Edward D. Lazwardska, “An Accurate and Efficient Performance Analysis Technique for Multiprocessor Snooping Cache-Consistency Protocols,” Proc. 15th Int'l Symp. Computer Architecture, pp. 308-315, May 1988.
[5] Wang Zhonghai, YE Yizheng, and Wang Jinxiang, “Designing AHB/PCI Bridge,” Proc. 4th Int'l Conf.ASIC, pp. 578-580, October 2001.
[6] PLDapplications (http://www.plda.com ), “AHB-PCI User Guide,”22 August 2002.
[7] CAST (http://www.plda.com/ ), “PCI-M64 Datasheet”.
[8] PCI Special Interest Group, PCI Local Bus Specification Reversion 2.2, (PCI SIG, 1998).
[9] ARM Ltd, AMBA Specification Reversion 2.0, (ARM Ltd, 1999).
[10] ARM Ltd, AMBA University Kit Technical Reference Manual, (ARM Ltd, 2001).
[11] Miha Dolenc, PCI IP Core Specification Reversion 0.6,( OpenCores January 28,2002).
[12] Miha Dolenc, PCI IP Core Design Document Reversion 0.1, (OpenCores January 28, 2002).
[13] Kevin A. Kettler, “Modeling Bus Scheduling Policies for Real-time Systems,” Proc. 16th Int'l Symp. Real Time System, pp. 242-253, December 1995.