Performance analysis and tuning for a single-chip multiprocessor DSP

A new generation of high performance programmable digital signal processors has a highly integrated parallel architecture, incorporating special purpose hardware features, on-chip memory, and multiple processors into a single chip. Such single chip multiprocessor DSPs, however, require a sophisticated performance monitoring tool to achieve maximum performance. The authors discuss the requirements and functionality of performance monitoring tools suitable for single chip multiprocessor DSPs. Specifically, they describe a performance monitoring tool that satisfies these requirements and functionality: the MVP Performance Monitor, developed for the Texas Instruments´ TMS320C80 Multimedia Video Processor. They present an overview of the MPM and a general performance tuning approach using the MPM. An 8×8 block based 2D discrete cosine transform (DCT) implementation demonstrates the MPM´s effectiveness. The authors achieved an overall speedup of 4.67 by tuning the performance based on the monitoring results from the MPM