QADAM: Quantization-Aware DNN Accelerator Modeling for Pareto-Optimality

TL;DR

QADAM is a flexible modeling framework that incorporates quantization effects into DNN accelerator design, enabling efficient exploration of design trade-offs and identifying Pareto-optimal configurations for energy and performance.

Contribution

The paper introduces QADAM, a comprehensive quantization-aware modeling framework for DNN accelerators that supports design space exploration with accurate power, performance, and area estimations.

Findings

Different quantization levels significantly affect performance and energy efficiency.

Lightweight processing elements (LightPEs) achieve Pareto-optimal results.

LightPEs outperform INT16 designs by up to 5.7x in performance per area and energy.

Abstract

As the machine learning and systems communities strive to achieve higher energy-efficiency through custom deep neural network (DNN) accelerators, varied bit precision or quantization levels, there is a need for design space exploration frameworks that incorporate quantization-aware processing elements (PE) into the accelerator design space while having accurate and fast power, performance, and area models. In this work, we present QADAM, a highly parameterized quantization-aware power, performance, and area modeling framework for DNN accelerators. Our framework can facilitate future research on design space exploration and Pareto-efficiency of DNN accelerators for various design choices such as bit precision, PE type, scratchpad sizes of PEs, global buffer size, number of total PEs, and DNN configurations. Our results show that different bit precisions and PE types lead to significant differences in terms of performance per area and energy. Specifically, our framework identifies a wide range of design points where performance per area and energy varies more than 5x and 35x, respectively. We also show that the proposed lightweight processing elements (LightPEs) consistently achieve Pareto-optimal results in terms of accuracy and hardware-efficiency. With the proposed framework, we show that LightPEs achieve on par accuracy results and up to 5.7x more performance per area and energy improvement when compared to the best INT16 based design.