TOPCELL: Topology Optimization of Standard Cell via LLMs

2026-04-15 · Source: Machine Learning · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Semiconductor Design & Optimization · Depth: Expert, quick

Summary

TOPCELL is a new framework designed to optimize transistor topology in standard cell design, addressing the computational bottleneck of traditional exhaustive search methods. It reframes high-dimensional topology exploration as a generative task, leveraging Large Language Models (LLMs). The framework uses Group Relative Policy Optimization (GRPO) to fine-tune the LLM, ensuring alignment with both logical circuit constraints and spatial layout requirements. Tested in an industrial flow for a 2nm technology node, TOPCELL demonstrated superior performance over foundation models in generating routable, physically-aware topologies. When integrated into a state-of-the-art automation flow for 7nm library generation, TOPCELL achieved layout quality comparable to exhaustive solvers with an 85.91x speedup and robust zero-shot generalization.

Key takeaway

For research scientists and design engineers working on advanced node standard cell design, TOPCELL offers a compelling alternative to traditional exhaustive search. You should consider integrating this LLM-based generative approach to achieve substantial speedups, specifically an 85.91x improvement, while maintaining or exceeding layout quality for 2nm and 7nm technology nodes. This could significantly accelerate library generation and design cycles.

Key insights

TOPCELL uses LLMs and GRPO to optimize transistor topology, achieving significant speedup and quality in standard cell design.

Principles

• Reformulate complex search as a generative task.
• Align generative models with physical and logical constraints.

Method

TOPCELL reformulates topology optimization as an LLM generative task, fine-tuned with Group Relative Policy Optimization (GRPO) to satisfy circuit and layout constraints.

In practice

• Apply LLMs to high-dimensional design space exploration.
• Use GRPO for constraint-aware model fine-tuning.

Topics

• TOPCELL
• Topology Optimization
• Large Language Models
• Standard Cell Design
• Group Relative Policy Optimization

Best for: Research Scientist, AI Scientist, AI Hardware Engineer, Machine Learning Engineer

Related on AIssential

• FusionCell: Cross-Attentive Fusion of Layout Geometry and Netlist Topology for Standard-Cell Performance Prediction — Topology-guided fusion of layout geometry and netlist topology accurately predicts standard-cell performance.
• What Makes an LLM a Good Optimizer? A Trajectory Analysis of LLM-Guided Evolutionary Search — Strong LLM optimizers act as local refiners, making incremental improvements within localized semantic search spaces.
• Hierarchical Task Network Planning with LLM-Generated Heuristics — LLMs can generate effective search heuristics for HTN planning, improving efficiency.
• Sandwich: Joint Configuration Search and Hot-Switching for Efficient CPU LLM Serving — Optimizing CPU LLM serving requires separate execution plans for prefill and decode phases due to their distinct computational and memory demands.
• Building Blocks of LLMs: Decoding, Generation Parameters, and the LLM Application Lifecycle — LLMs generate text by converting token probabilities into sequences using various decoding strategies.
• How I topped the Open LLM Leaderboard using 2x 4090 GPUs - Research notes in Blog form — Duplicating specific Transformer layer blocks can significantly boost LLM performance without weight modification.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine Learning.