Application-specific instruction processors (ASIPs) have great potential to meet the challenging performance, cost, and power demands of the next generation pervasive computing platforms. Three critical issues with ASIPs are design time, design cost, and post-programmability. Automatic design is the key to simultaneously addressing these issues and enabling ASIPs to proliferate. The central vision of this research is to perform architecture synthesis based on compiler technology. Compilers employ sophisticated analyses that are adept at understanding, modeling, and transforming applications. These technologies will be reshaped from producing software customized for a predefined processor to producing hardware specialized to a software application. In this research, an automatic compiler-based synthesis system for ASIPs is being developed, including advances in architecture description technologies, instruction set, datapath, and memory synthesis techniques, and compiler scheduling and code generation algorithms.
The broad significance of this work is potentially reshaping the world of pervasive computers. Currently, a select few companies have the necessary resources, expertise, and ability to accomplish a completely designed and verified hardware solution. By creating the enabling technology to go from a software prototype to a specialized-programmable hardware solution, individuals without the necessary hardware expertise will gain the ability to innovate in the pervasive computing domain
