DeepCircuitX provides a holistic, multi-level dataset for training LLMs to understand and generate hardware description code, bridging a critical gap in AI-assisted circuit design.

• First repository-level RTL code dataset with hierarchical structures from repositories down to code blocks
• Enables more nuanced training of models for code understanding, generation, and optimization
• Supports power-performance-area (PPA) analysis, a critical aspect of hardware design
• Creates new opportunities for automated design optimization and verification

This breakthrough addresses a fundamental challenge in hardware engineering automation, potentially accelerating design cycles and improving efficiency for complex circuit development.

DeepCircuitX: A Comprehensive Repository-Level Dataset for RTL Code Understanding, Generation, and PPA Analysis