Optimized parallel implementation of polynomial approximation math functions on a DSP processor

This paper presents a general method to implement polynomial approximation math functions on TMS320C67X architecture with multiple parallel execution units. Our method consists of grain packing, mapping and scheduling to reduce data dependency overhead and fully utilize delay slots. Experimental results of our method on TMS320C67x have achieved up to 70.2% performance improvement over 'C67x library functions.