PDNNet: PDN-Aware GNN-CNN Heterogeneous Network for Dynamic IR Drop Prediction

Yuxiang Zhao,Zhuomin Chai,Xun Jiang,Yibo Lin,Runsheng Wang,Ru Huang

2024-03-27

Abstract:IR drop on the power delivery network (PDN) is closely related to PDN's configuration and cell current consumption. As the integrated circuit (IC) design is growing larger, dynamic IR drop simulation becomes computationally unaffordable and machine learning based IR drop prediction has been explored as a promising solution. Although CNN-based methods have been adapted to IR drop prediction task in several works, the shortcomings of overlooking PDN configuration is non-negligible. In this paper, we consider not only how to properly represent cell-PDN relation, but also how to model IR drop following its physical nature in the feature aggregation procedure. Thus, we propose a novel graph structure, PDNGraph, to unify the representations of the PDN structure and the fine-grained cell-PDN relation. We further propose a dual-branch heterogeneous network, PDNNet, incorporating two parallel GNN-CNN branches to favorably capture the above features during the learning process. Several key designs are presented to make the dynamic IR drop prediction highly effective and interpretable. We are the first work to apply graph structure to deep-learning based dynamic IR drop prediction method. Experiments show that PDNNet outperforms the state-of-the-art CNN-based methods by up to 39.3% reduction in prediction error and achieves 545x speedup compared to the commercial tool, which demonstrates the superiority of our method.

Machine Learning,Artificial Intelligence

What problem does this paper attempt to address?

### Problems Addressed by the Paper This paper aims to address the issue of dynamic IR drop prediction in integrated circuit (IC) design. Specifically: 1. **Problem Background**: - Dynamic IR drop refers to the voltage drop phenomenon in the power delivery network (PDN) caused by current flow. - As the scale of IC design increases, traditional simulation-based methods become computationally infeasible, necessitating a new, efficient method for predicting dynamic IR drop. 2. **Limitations of Existing Methods**: - Current methods based on convolutional neural networks (CNN) have achieved some success but overlook the importance of PDN structure. - Existing methods mainly focus on current load patterns and fail to fully explore the PDN structure and its relationship with the cells. 3. **Contributions of the Paper**: - Proposes a new graph structure—PDNGraph, which comprehensively represents the PDN structure and the relationship between the cells and the PDN. - Designs a dual-branch heterogeneous network, PDNNet, combining the advantages of graph neural networks (GNN) and CNN to capture PDN architecture and dynamic IR drop variations. - Experimental results show that PDNNet reduces prediction error by up to 39.3% compared to state-of-the-art CNN methods and improves speed by 545 times. In summary, this paper improves existing dynamic IR drop prediction methods by introducing a graph structure and a dual-branch network, enhancing prediction accuracy and efficiency.

PDNNet: PDN-Aware GNN-CNN Heterogeneous Network for Dynamic IR Drop Prediction

Worst-Case Dynamic Power Distribution Network Noise Prediction Using Convolutional Neural Network

Worst-case Power Integrity Prediction Using Convolutional Neural Network.

PowerNet: Transferable Dynamic IR Drop Estimation via Maximum Convolutional Neural Network

A novel interactive deep cascade spectral graph convolutional network with multi-relational graphs for disease prediction

Global and Local Attention-based Inception U-Net for Static IR Drop Prediction

Circuits Physics Constrained Predictor of Static IR Drop with Limited Data

Static IR Drop Prediction with Attention U-Net and Saliency-Based Explainability

Global and Local Attention-based Inception U-Net for Static IR Drop Estimation

PDFormer: Propagation Delay-Aware Dynamic Long-Range Transformer for Traffic Flow Prediction

Deeper and Wider Networks for Performance Metrics Prediction in Communication Networks

NeuralPower: Predict and Deploy Energy-Efficient Convolutional Neural Networks

Dynamic directed graph convolution network based ultra‐short‐term forecasting method of distributed photovoltaic power to enhance the resilience and flexibility of distribution network

HTNet: Dynamic WLAN Performance Prediction using Heterogenous Temporal GNN

Ultra-short-term operation situation prediction method of active distribution network based on convolutional neural network long short term memory

Adaptive Hybrid Spatial-Temporal Graph Neural Network for Cellular Traffic Prediction

PN-HGNN: Precipitation Nowcasting Network Via Hypergraph Neural Networks

Information Cascade Prediction of complex networks based on Physics-informed Graph Convolutional Network

Fast Dynamic IR-Drop Prediction Using Machine Learning in Bulk FinFET Technologies

A Novel Cellular Network Traffic Prediction Algorithm Based on Graph Convolution Neural Networks and Long Short-Term Memory through Extraction of Spatial-Temporal Characteristics

Rapid prediction for the transient dispersion of leaked airborne pollutant in urban environment based on graph neural networks