Accurate and Efficient Three Level Design Space Exploration Based on Constraints Satisfaction Optimization Problem Solver

Abstract

In this paper, we propose an efficient and effective there level Design Space Exploration (DSE) method for mapping a system consisting of a number of DSP functions onto an RTL or lower level model using constraint programming methodology. The design space has three dimensions: a) function execution schedule (when the functions should execute), b) function implementation assignment (how the execution of functions are assigned to physical kernels) and c) implementation architecture (how many arithmetic units are deployed in each kernel). The DSE has been formulated as a Constraints Satisfaction Optimization Problem (CSOP) and solved by the constraint programming solver in Google's OR-Tools.