DNN accelerators are often developed and evaluated in isolation without considering the cross-stack, system-level effects in real-world environments. This makes it difficult to appreciate the impact of System-on-Chip (SoC) resource contention, OS overheads, and programming-stack inefficiencies on overall performance/energy-efficiency. To address this challenge, we present Gemmini, an open-source*, full-stack DNN accelerator generator. Gemmini generates a wide design-space of efficient ASIC accelerators from a flexible architectural template, together with flexible programming stacks and full SoCs with shared resources that capture system-level effects. Gemmini-generated accelerators have also been fabricated, delivering up to three orders-of-magnitude speedups over high-performance CPUs on various DNN benchmarks. * https://github.com/ucb-bar/gemmini
@article{arxiv.1911.09925,
title = {Gemmini: Enabling Systematic Deep-Learning Architecture Evaluation via Full-Stack Integration},
author = {Hasan Genc and Seah Kim and Alon Amid and Ameer Haj-Ali and Vighnesh Iyer and Pranav Prakash and Jerry Zhao and Daniel Grubb and Harrison Liew and Howard Mao and Albert Ou and Colin Schmidt and Samuel Steffl and John Wright and Ion Stoica and Jonathan Ragan-Kelley and Krste Asanovic and Borivoje Nikolic and Yakun Sophia Shao},
journal= {arXiv preprint arXiv:1911.09925},
year = {2021}
}
Comments
To appear at the 58th IEEE/ACM Design Automation Conference (DAC), December 2021, San Francisco, CA, USA