Effective programming of very long instruction word digital signal processors

The paper deals with an optimal approaches used for programming of time consuming algorithms in digital signal processors. More precisely, the Texas Instruments' DSP with Very Long Instruction Word architecture is taking into account. The principle of the optimal programming is based on new auxiliary tool which ensures the maximal functional units allocation. The properties of the programming method are proved by application of Fast Fourier Transform and they are compared with possibilities of Texas Instruments digital signal processing library. The proposed FFT implementation was tested on C6747 floating-point DSP and it is more efficient than official Texas Instruments library. Therefore, new functions can be used in high-performance computing systems.