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

TL;DR

Gemmini is a comprehensive framework for designing, evaluating, and fabricating DNN accelerators with system-level considerations, enabling more realistic performance and energy-efficiency assessments.

Contribution

It introduces Gemmini, an open-source full-stack DNN accelerator generator that captures system-level effects for more accurate evaluation.

Findings

Achieved up to 1000x speedups over CPUs on DNN benchmarks.

Generated diverse ASIC accelerators from a flexible template.

Fabricated accelerators demonstrating practical performance benefits.

Abstract

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