EdgeDiff : Multi-modal Few-step Diffusion Model Accelerator
본문
Overview
This paper presents EdgeDiff, energy efficient diffusion model accelerator for mobile devices with 3 key features. 1) CRMP reorder activation for mixed precision and group quantization, and achieves ~2.72× energy efficiency. 2) CAA PE save 36.6% of MAC power with new PE structure and reduced toggle by BST. 3) TAU and GQU minimize hardware cost of the FP logic for group quantization. Finally, EdgeDiff demonstrates end-to-end multi-modal diffusion model with 3.3~6.8× lower energy than prior works.
Implementation results
Performance Comparison
Architecture
Features
- Condition-aware Reordered Group Mixed Precision (CRMP)
- Compress-and-Add (CAA) PE
- Tiered Accumulation Unit (TAU)
- Grid-based Quantization Unit (GQU)
Related Papers
- ISSCC 2025