Deep Learning Creativity in EDA
Cheng‐Kuang Lee · 2020
Computing power brings new technologies into human life. NVIDIA devoted to accelerating parallel computing through CPU-GPU cooperating architecture since we invented CUDA in 2007. In the last decade, one of the most important technologies, deep learning, became feasible and realistic because GPU accelerates the optimizing process of neural network over 60x, which means that the model training time shortens from weeks to hours. DL not only performed superior to human in some computer vision tasks like ImageNet but also made significant progress of many fields like medical, autonomous, manufacturing, finance, electronic design automation (EDA) etc.In this paper, we introduced two deep learning innovations in EDA field including (1) Graph Convolutional Network in Testability Analysis and (2) DREAMPlace. In (1), graph convolutional network can predict observation point candidates in a netlist more efficiently compared to commercial tools. In (2), DREAMPlace significantly accelerates the VLSI placement process by using deep learning framework of GPU for optimizing process.