Temperature-Aware Monolithic 3D DNN Accelerators for Biomedical Applications
Temperature-Aware Monolithic 3D DNN Accelerators for Biomedical Applications
复制标题
DOI:
10.48550/arxiv.2203.15874
复制
发表时间:
2022-03
期刊:
影响因子:
--
通讯作者:
Prachi Shukla;V. Pavlidis;E. Salman;A. Coskun
中科院分区:
文献类型:
--
作者:
Prachi Shukla;V. Pavlidis;E. Salman;A. Coskun
In this paper, we focus on temperature-aware Monolithic 3D (Mono3D) deep neural network (DNN) inference accelerators for biomedical applications. We develop an optimizer that tunes aspect ratios and footprint of the accelerator under user-defined performance and thermal constraints, and generates near-optimal configurations. Using the proposed Mono3D optimizer, we demonstrate up to 61% improvement in energy efficiency for biomedical applications over a performance-optimized accelerator.