利用繪圖處理器平行運算實現基頻通道模擬與通訊接收機

Baseband Channel Simulation and Communication Receiver Implementation Using GPU Acceleration

陳語煙(Yu-Yen Chen)

指導教授 ： 闕志達

國立臺灣大學/電機資訊學院/電子工程學研究所/碩士(2009年)

https://doi.org/10.6342/NTU.2009.00914

摘要

繪圖處理器(Graphics Processing Unit)經過數代的演進，由只具備固定功能的運算單元漸漸發展而具有高度可程式化的能力。由於繪圖處理器通常具有相當高的記憶體頻寬，以及極佳的浮點運算能力，因此開始有利用繪圖處理器來加速一些非圖形處理計算工作的想法，即通用運算繪圖處理器(General Purpose Computing on Graphics Processing Units)。
然而，並非所有運算皆適合使用繪圖處理器加速，於繪圖處理器處理之運算需為重複且低資料相依性，即具備高平行化程度之特性。本論文中，將使用統一計算架構(Compute Unified Device Architecture)，也就是NVIDIA公司通用運算繪圖處理器的模型，平行化地實現正交分頻多工(Orthogonal Frequency-Division Multiplexing)通訊系統中常見的快速傅立葉轉換(Fast Fourier Transform)，以及修正傳送錯誤常用的維特比解碼器(Viterbi Decoder)。同時也將平行化的概念，應用到基頻通道模擬器以及通訊接收機中，將需要耗費大量時間且能夠平行化實現之運算，由中央處理器(CPU)移轉至繪圖處理器(GPU)中處理，而達到加速的目的。於本論文中，除了呈現使用繪圖處理器加速之方法與成果外，也將分享繪圖處理器程式設計(GPU Programming)之經驗。

關鍵字

通用運算繪圖處理器 ； 統一計算設備架構 ； 歐規數位電視 ； 基頻數位通訊

並列摘要

The Graphics Processing Unit (GPU) has been evolving for various generations, from a fixed-funcion graphics pipeline to a programmable parallel processor. Due to the fact that GPU has a high memory bandwidth and a high Giga FLoating points Operations Per Second (GFLOPS) performance, it is used to accelerate the computations of non-graphics data, which is also referred as General Purpose Computing on Graphics Processing Units (GPGPU).
However, not all algorithms benefit equally from this technique. The GPU computing is only suitable for algorithms with repeating computation and low data dependency, which means with a high degree of parallel computing characteristics. There are mainly two groups of communication applications analyzed and implemented in this thesis, which are all implemented and simulated using general personal computer under the NVIDIA CUDA environment. First, the method and result of the baseband channel simulation implementation using parallel computing concept is presented in the thesis. Then the parallel computing concepts are being used to implement the communication receiver. Both of the speedup performance and the GPU programming experience will be covered in this thesis.

並列關鍵字

GPGPU ； CUDA ； DVB-T ； Digital Baseband Communication

參考文獻

[5] E. Lindholm, J. Nickolls, S. Oberman, and J. Montrym, “NVIDIA Tesla: A Unified Graphics and Computing Architecture,” Micro. IEEE, vol. 28, no. 2, pp. 39-55, Mar.-Apr. 2008.

[6] M. Garland, S. L. Grand, J. Nickolls, J. Anderson, S. Morton, E. Phillips, Y. Zhang, and V. Volkov, “Parallel Computing Experiences with CUDA,” Micro. IEEE, vol. 28, no. 4, pp. 13-27, Jul.-Aug. 2008.

[11] Tzi-Dar Chiueh, and Pei-Yun Tsai, OFDM Baseband Receiver Design for Wireless Communications. Taiwan, John Wiley & Sons, Inc., 2007.

[14] N. K. Govindaraju, B. Lloyd, Y. Dotsenko, B. Smith, and J. Manferdelli, “High Performance Discrete Fourier Transforms on Graphics Processors,” ACM/IEEE Conference on Supercomputing, Nov. 15-21, 2008.

[16] V. Balu, J. P. Walters, S. Kompalli, and V. Chaudhary, “Evaluating the Use of GPUs for Life Science Application,” Tech. Report, University at Buffalo, New York, May 22, 2008.