RouteNet

Abstract

Early routability prediction helps designers and tools perform preventive measures so that design rule violations can be avoided in a proactive manner. However, it is a huge challenge to have a predictor that is both accurate and fast. In this work, we study how to leverage convolutional neural network to address this challenge. The proposed method, called RouteNet, can either evaluate the overall routability of cell placement solutions without global routing or predict the locations of DRC (Design Rule Checking) hotspots. In both cases, large macros in mixed-size designs are taken into consideration. Experiments on benchmark circuits show that RouteNet can forecast overall routability with accuracy similar to that of global router while using substantially less runtime. For DRC hotspot prediction, RouteNet improves accuracy by 50% compared to global routing. It also significantly outperforms other machine learning approaches such as support vector machine and logistic regression.

Machine Learning Basado en La Red Para La Predecibilidad Del Tiempo Y Diafonía De Circuitos Digitales

“…Las redes se pueden dividir en redes de largo alcance y redes de corto alcance según su HPWL. En (Xie et al, 2018), se muestra que la congestión de enrutamiento tiene una correlación más fuerte con las redes de largo alcance que con las más cortas. en este sentido, también se extrae una función longrangerudy al considerar solo las redes de largo alcance al calcular el volumen total del cable.…”

A Deep Learning framework for simulation and defect prediction applied in microelectronics

“…The work of [33] uses a 3DCNN to model and predict surface deterioration phenomena on metallic materials used in artwork based on previous 3D scans of artificially aged reference samples. In [36], the authors propose a methodology based on convolutional neural networks, called RouteNet that performs preventive measures in early routability prediction so as to avoid design rule violation.…”

Application of Machine Learning to Physical Design

“…Deep learning is introduced to provide quick and precise routability prediction because to its great performance in computer vision and other related applications. Prior to completing the routing, [33] forecast the number and position of design rule violations. This work takes advantage of the 2D placement's similarity to the well-studied image classification problem, which can also be represented as an image.…”

“…This enables them to perform transfer learning from a different domain in addition to using a similar model (CNN). A fully convolutional network (FCN) is proposed by RouteNet [33] to predict the position of DRV hotspots, and a standard CNN to predict the quantity of design rule violations (DRVs) for routability forecast. (Fig.…”

“…Predictor is a deep learning system that uses a specified netlist to find small violations without the use of a global router. In contrast to [33], where raw images of the placed designs are provided to the NN during training, [34] train a NN but place a heavy emphasis on feature engineering. According to the experimental findings, the proposed detection method can anticipate, on average, 90% of the shorts on previously unobserved data with just 5% false alarms (FAs), and it can also significantly cut down on computation time.…”