Summary: | 碩士 === 國立臺灣大學 === 電子工程學研究所 === 101 === Wireless baseband processing, which usually requires high computation complexity and high data throughput, is regarded as the most challenging issue for Software-Defined Radio (SDR) systems. To relieve the difficulty in SDR systems, a modern Graphics Processing Unit (GPU) is chosen as implementation platform due to its numerous powerful arithmetic logic units. However, because of the lack of the universal parallel programming framework for SDR systems, it is difficult to take advantage of the GPU architecture. To overcome this problem, in this thesis, we propose the different levels of parallelism that can be exploited on GPU platforms for most baseband functions. The parallel approaches of each baseband function implemented on GPU platform can not only enhance the performance of baseband signal processing to meet the real-time requirement, but also gives an SDR developer the faster solutions to those time-consuming baseband functions if GPU is used to prototype a novel wireless protocol. In our experimental tests, a GPU platform can support about 20Mbps data rate for baseband signal processing at most. This result can also verify the effectiveness of our proposed GPU framework for SDR systems.
|