A Study Of Signal Processing Acceleration Technology Based On Cell Broadband Engine Processor

Heterogeneous multi-core processing technology is an important direction of the modern multi-core processor. Cell processor is a representative of the heterogeneous multi-core processors. It contains a Power Processor Element and eight Synergistic Processor Elements. Cell processor is more powerful than Single-core processors and homogeneous multi-core processors, especially in signal processing area. The programming mode is totally different from the traditional processors. The thesis presents a Cell processor-based programming model to simplify the programming process on the platform.This thesis attempts to enhance the performance of the signal processing program on the Cell processor from the following two aspects: firstly, we improve the speed of the program on a single SPE through SIMD and loop-unrolling technologies; secondly, we map computation stages to different SPEs then use DMA transfer intermediate results from SPE to SPE in pipeline fashion. After fully parallel the computation stages, we defined a SPE pipelined programming model. In order to enhance the portability of the program, the thesis implement a software middleware on the Cell processor.The thesis implement pulse compression function as an example of the realization of multi-core processors by the optimizational program model. The result shows that the program is far more faster than the traditional one. The research result of the thesis is useful for the future research on the Cell cluster.