This article presented a methodology to enable fast cost-benefit estimation of a DNN accelerator on a given DNN model and mapping, which consists of a compiler-friendly data-centric representation of mappings and an analytical cost-benefit estimation framework that exploits the explicit data reuse in data space in data-centric representations.