Summary of Preroutgnn For Timing Prediction with Order Preserving Partition: Global Circuit Pre-training, Local Delay Learning and Attentional Cell Modeling, by Ruizhe Zhong et al.
PreRoutGNN for Timing Prediction with Order Preserving Partition: Global Circuit Pre-training, Local Delay Learning and Attentional Cell Modeling
by Ruizhe Zhong, Junjie Ye, Zhentao Tang, Shixiong Kai, Mingxuan Yuan, Jianye Hao, Junchi Yan
First submitted to arxiv on: 27 Feb 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Hardware Architecture (cs.AR)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper proposes a two-stage approach for pre-routing timing prediction in chip design. The method involves directly estimating timing metrics without time-consuming routing, but often suffers from signal decay and error accumulation due to long timing paths. To address these challenges, the authors propose global circuit training using graph auto-encoders to learn the global graph embedding from circuit netlist. They then use a novel node updating scheme for message passing on GCN, which residually models local time delay between adjacent pins and extracts lookup table information inside each cell via attention mechanism. To handle large-scale circuits efficiently, an order preserving partition scheme is introduced, reducing memory consumption while maintaining topological dependencies. The proposed method achieves a new state-of-the-art (SOTA) R2 of 0.93 for slack prediction, surpassing the previous SOTA method with 0.59. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper proposes a new way to predict timing in chip design without having to do all the complicated routing first. It’s like trying to figure out how long it will take to get somewhere by looking at a map of the roads, rather than actually driving on them. The method uses special computer programs called graph auto-encoders and GCN (Graph Convolutional Networks) to learn about the circuit and make predictions. This is important because timing prediction can help designers make better chips that are faster and more efficient. |
Keywords
* Artificial intelligence * Attention * Embedding * Gcn
