论文部分内容阅读
In this paper,we propose a novel ECOMT predictive modeling approach to present delay and area models of FPGA.Semi-supervised Learning is employed in this approach to automatically generate the model with respect to architectural parameters.The process of obtaining model is time-efficient meanwhile the model produced is of high accuracy.According to experimental results,the model generated shows an ability of accurate prediction with MRE(mean relative error)less than 5%.Moreover,the comprehensive model,combined with nonlinear programming,can be further utilized to explore the FPGA design space.The optimum architectural parameters gained will guide FPGA architects to design an optimized FPGA under a specific design specification.