Bridging the Gap between ASIC and GPP: A High-Performance and C-Programmable ASIP for Image Processing

Abstract

Emerging digital signal applications nowadays require not only massive computational load but also a significant level of flexibility and programmability. ASIC (Application Specific Integrated Circuit) and GPP (General Purpose Processor) have been widely used as the two main countermeasures to fulfill those requirements. Unfortunately, the gap between these two mainstreams is deliberately big. While ASIC can fulfill the requirement of ultra-high performance, it lacks the flexibility and programmability. On the other hand, the performance of GPP is not competitive despite it offering an excellent level of flexibility and programmability. In this paper, we present ASIP (Application Specific Instruction-set Processor) for image processing as a solution in bridging the gap between ASIC and GPP. The designed ASIP constitutes the appropriate solution for fulfilling both the flexibility and performance constraints of emerging and future image processing applications. Based on the simulation result, we found that our ASIP can achieve 16 times better performance of conventional RISC (Reduced Instruction Set Computing) Processor as well as providing a high level of flexibility and programmability that ASIC lacks.

Keywords: microprocessor, ASIP, embedded system, reconfigurable system, high-level synthesis