Fast and efficient algorithm for the multiplierless realisation of linear DSP transforms

A fast algorithm having a pseudopolynomial run-time and memory requirement in the worst case is developed to generate multiplierless architectures at all wordlengths for constant multiplications in linear DSP transforms. It is also re-emphasised that indefinitely reducing operators for multiplierless architectures is not sufficient to reduce the final chip area. For a major reduction, techniques like resource folding must be used. Simple techniques for improving the results are also presented