Gemmini: Enabling Systematic Deep-Learning Architecture Evaluation via Full-Stack Integration

• Published in: 2021 58th ACM/IEEE Design Automation Conference (DAC) pp. 769 - 774
• Main Authors: Genc, Hasan, Kim, Seah, Amid, Alon, Haj-Ali, Ameer, Iyer, Vighnesh, Prakash, Pranav, Zhao, Jerry, Grubb, Daniel, Liew, Harrison, Mao, Howard, Ou, Albert, Schmidt, Colin, Steffl, Samuel, Wright, John, Stoica, Ion, Ragan-Kelley, Jonathan, Asanovic, Krste, Nikolic, Borivoje, Shao, Yakun Sophia
• Format: Conference Proceeding
• Language: English
• Published: IEEE 05.12.2021
• Subjects: Accelerator architectures; Benchmark testing; Design automation; Operating systems; Productivity; Programming; Systematics
• DOI: 10.1109/DAC18074.2021.9586216

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 Systemon-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.