DiffPattern-Flex: Efficient Layout Pattern Generation via Discrete Diffusion

TL;DR

DiffPattern-Flex addresses the challenge of generating large-scale, legally compliant layout pattern libraries by decoupling topology synthesis from legalization and employing a discrete diffusion process to produce lossless topology tensors encoded as Deep Squish Patterns. A white-box 2D nonlinear system assigns geometric vectors to topology tensors to guarantee adherence to design rules, achieving 100% legality while preserving high pattern diversity through robust augmentation and pre-filtering. The framework delivers substantial speedups via fast-sampling and optimized legalization, with empirical results showing state-of-the-art diversity and perfect legality on ICCAD-2014-inspired data, and demonstrated flexibility to adapt to changing design rules without retraining. The work has practical impact for lithography simulation, hotspot detection, and other downstream EDA tasks that require scalable, reliable pattern libraries.

Abstract

Recent advancements in layout pattern generation have been dominated by deep generative models. However, relying solely on neural networks for legality guarantees raises concerns in many practical applications. In this paper, we present \tool{DiffPattern}-Flex, a novel approach designed to generate reliable layout patterns efficiently. \tool{DiffPattern}-Flex incorporates a new method for generating diverse topologies using a discrete diffusion model while maintaining a lossless and compute-efficient layout representation. To ensure legal pattern generation, we employ {an} optimization-based, white-box pattern assessment process based on specific design rules. Furthermore, fast sampling and efficient legalization technologies are employed to accelerate the generation process. Experimental results across various benchmarks demonstrate that \tool{DiffPattern}-Flex significantly outperforms existing methods and excels at producing reliable layout patterns.

Figures (14)

• Figure 1: Illustration of denoising diffusion process.
• Figure 2: Squish Pattern Representation.
• Figure 3: Illustration of design rules.
• Figure 4: An illustration of the $\mathsf{Diffpattern}$-Flex framework for reliable layout pattern generation.
• Figure 5: An illustration of the Deep Squish Pattern Encoding. The Topology Tensor provides a compact and lossless representation of the topology matrix. Simple bit concatenation leads to unequal power distribution across bits and causes exponential growth in the state space.

Theorems & Definitions (2)

• Definition 1
• Definition 2: Pattern Legality