Physics-Guided Geometric Diffusion for Macro Placement Generation

2026-05-15

AI Summary

This work addresses the challenge of jointly optimizing topological connectivity, physical constraints, and sequential dependencies in macro-cell placement. To this end, the authors propose MacroDiff+, a dual-domain denoising diffusion framework that integrates heterogeneous graph neural networks to encode topological structure with Transformer-based modeling of global geometric context. During inference, a gradient-based physically guided sampling strategy is introduced to simultaneously ensure statistical plausibility and physical validity throughout the generation process. The approach significantly enhances stability and scalability for large-scale designs. Evaluated on the ISPD2005 MMS benchmark, MacroDiff+ achieves a 6.1–6.2% reduction in wirelength, outperforming existing methods while demonstrating superior convergence properties.

Abstract

Macro placement is a pivotal stage in VLSI physical design, fundamentally determining the overall chip performance. Recent data-driven placement methods have demonstrated significant potential, yet they often struggle to handle sequential dependencies and to balance topological connectivity with physical constraints. To bridge this gap, we propose MacroDiff+, a physics-guided geometric diffusion framework. Specifically, we design a dual-domain denoising architecture that couples topological connectivity encoded by heterogeneous GNNs with global geometric context modeled by a Transformer. Furthermore, we introduce Physics-Guided Sampling, an inference strategy that actively steers the generation using explicit gradients to ensure both statistical plausibility and physical validity. On the ISPD2005 MMS benchmarks, MacroDiff+ outperforms state-of-the-art baselines with a 6.1-6.2% reduction in wirelength. Notably, it exhibits superior stability and scalability on large-scale designs where prior methods fail to converge. The source code is available at https://github.com/jhy00n/MacroDiff-plus.

Problem

• macro placement
• VLSI physical design
• topological connectivity
• physical constraints
• sequential dependencies

Innovation

• physics-guided diffusion
• geometric context
• heterogeneous GNN
• Transformer
• macro placement