Bjorn Hori et al.
Berkeley Design Technology (BDTI)

Digital signal processing has become a key component in many consumer, communications, medical, and industrial products. These products use a variety of hardware approaches to implement the required signal processing, ranging from off-the-shelf general-purpose microprocessors (GPPs) to application-specific integrated circuits (ASICs) to fixed-function chips.

While both GPPs and DSPs provide sufficient processing power for many signal processing applications, each type of processor brings with it important strengths and weaknesses that can have an enormous influence on how effectively the processor meets the needs of a particular application. In this paper, we examine the key attributes that distinguish DSP algorithms and applications and show how these attributes have led to the architectural specialization found in DSP processors. We then investigate how GPP and DSP architectures compare with regard to these key DSP algorithm attributes. In addition, we examine the architectural techniques employed in high-performance GPPs and DSPs, which have led to rapid gains in signal processing performance in both classes of processor.