A quantitative approach to development and validation of synthetic benchmarks for behavioral synthesis

Abstract

We present a quantitative approach to development and validation of synthetic benchmarks for behavioral synthesis systems. The approach is built on the idea of quantitative benchmark selection. We briefly explain the idea and present experimental results on the quantitative selection and validation of benchmarks. We develop a synthetic design example generator which composes the behavioral level specification of a design having the properties given by a set of numerical parameters. Experimental generation and application of synthetic design examples demonstrate the effectiveness of the proposed approach and the developed algorithms.