Algorithm Selection Framework for Legalization Using Deep Convolutional Neural Networks and Transfer Learning

International audienceMachine learning models have been used to improve the quality of different physical design steps, such as timing analysis, clock tree synthesis and routing. However, so far very few works have addressed the problem of algorithm selection during physical design, which can drastically reduce the computational effort of some steps. This work proposes a legalization algorithm selection framework using deep convolutional neural networks. To extract features, we used snapshots of circuit placements and used transfer learning to train the models using pre-trained weights of the Squeezenet architecture. By doing so we can greatly reduce the training time and required data even though the pre-trained weights come from a different problem. We performed extensive experimental analysis of machine learning models, providing details on how we chose the parameters of our model, such as convolutional neural network architecture, learning rate and number of epochs. We evaluated the proposed framework by training a model to select between different legalization algorithms according to cell displacement and wirelength variation. The trained models achieved an average F-score of 0.98 when predicting cell displacement and 0.83 when predicting wirelength variation. When integrated into the physical design flow, the cell displacement model achieved the best results on 15 out of 16 designs, while the wirelength variation model achieved that for 10 out of 16 designs, being better than any individual legalization algorithm. Finally, using the proposed machine learning model for algorithm selection resulted in a speedup of up to 10x compared to running all the algorithms separately