The Execution Optimization Of RISC Oriented To Digital Signal Processing

As the increasing widespread applications of synthectical voice analysis, streaming media process and intelligent image identification rise in micro-devices such as cell phone and PDA, developers hopes that their limited system resources can deal with multi-type transactions effective during the research of development involving various fields of subject. The system is design to handle not only intensive complex computations, but also interactive scheduling missions. The ordinary solutions prefer the cooperative processing system containing both RISC and DSP which make the best of the performance in respective different areas. However, the design might also risk the properties of system complexity, total cost, power consumption and reliability.Based on the architecture and function, the paper analyzes the respective characteristics of RISC and DSP, and proposes the idea that we can implement DSP function in RISC, in order to substitute the co-system with a single RISC. Because the idea would surely result in performance degradation, especially for intensive computation performance, the paper presents the resolution for execution optimization, and also gives approaches of performance prediction for optimization result. The methods introduced here can be regard as suggestion and solution for the digital signal processing application implemented with RISC.The research in paper is oriented to digital signal processing field, which tries to figure out the difference and the performance gap between RISC and DSP. The paper implements on RISC the basic operations of multiplication, division, MAC, saturation and repeating execution, and optimizes the approaches. Analysises are given to quantify the optimization result. Finally, a post-filter module of G. 723.1 audio signal compressing algorithm is adopted for RISC performance test in digital-signal-processing tasks. The final result indicates that the RISC performance with optimization is close to DSP, which can be taken as a suggestion whether the DSP could be omitted in a practical system. Meanwhile, this will be helpful for the development of high-performance, low-cost and low-power-consumption real time system.