Research And Design Ofstream Rocessor Based On TTA

Stream Processor is based on the stream architecture. In accordance with stream calculation model, data and calculation are separated and re-organized into a pipeline-based computing chain. By developing of multiple levels of parallelism such as data, instruction and task level, and taking full advantage of localization of all levels of storage, the architecture shows good performance especially in compute-intensive applications with highly parallelized data and less global reused data. In my thesis the computing unit in stream architecture is based on transport triggered architecture which is TTA for short. Operation in TTA is triggered by data transportation which means when data is transferred through network and stored in the trigger register of function unit, operation will start. TTA can overcome the limitation of current ASIP design and could get the processor which satisfies target application as fast as possible.The application in the thesis is intra-prediction arithmetic of H.264. Via evaluation, the processor consumes less energy and area, and the performance is better than the usual DSP core. Also I make optimization to the stream processor of intra-prediction arithmetic. I propose a fast algorithm for intra prediction algorithm based on the character of the stream processor to further improve the system performance.The thesis presents a highly efficient processor architecture design methodology for local compute-intensive applications. It can be implement some more complex algorithms like other arithmetic of H.264 through the approach.