A 28nm 4.96 TOPS/W End-to-End Diffusion Accelerator with Reconfigurable Hyper-Precision and Unified Non-Matrix Processing Engine

This paper presents Picasso, an end-to-end diffusion accelerator. Picasso proposes a novel hyper-precision data type and reconfigurable architecture that can maximize hardware efficiency with extended dynamic range, with no compromise in accuracy. Picasso also proposes a unified engine operating all non-matrix operations in a streamlined processing flow and minimizes the end-to-end latency by sub-block pipeline scheduling. The accelerator is fabricated in 28 nm CMOS technology and achieves an energy efficiency of 4.96 TOPS/W and a peak performance of 9.83 TOPS. Compared with prior works, Picasso achieves speedups of $8.4 \times-26.8 \times$ while improving energy and area efficiency by $1.1 \times-2.8 \times$ and $3.6 \times-30.5 \times$, respectively.