A TensorFlow and System Simulator Integration Approach to Estimate Hardware Metrics of Convolution Accelerators

Abstract

GPUs became the reference platform for both training and inference phases of Convolutional Neural Networks (CNN), due to their tailored architecture to the CNN operators. However, GPUs are power-hungry architectures. A path to enable the deployment of CNNs in energy-constrained devices is adopting hardware accelerators for the inference phase. The design space exploration of CNNs using standard approaches, such as RTL, is limited due to their complexity. Thus, designers need frameworks enabling design space exploration that delivers accurate hardware estimation metrics do deploy CNNs. This work aims to propose a framework to explore hardware accelerators' design space, providing power, performance, and area (PPA) estimations. The heart of the framework is a system simulator with TensorFlow as front-end and as back-end performance estimations obtained from the physical synthesis. Results evaluate the energy trade-off varying the number of convolutional layers, showing that it can be necessary to spend approximately 40% more energy to increase 0.02% in the accuracy.

A High-Level Modeling Framework for Estimating Hardware Metrics of CNN Accelerators

“…This article extends the authors' previous work [18], which shares the proposal of a framework for PPA evaluation of CNNs. This work brings updated related work, refinement of the framework description, use of two hardware accelerators types (Section IV), and a set of new results, including the synthesis of accelerators synthesis (Sections V-B and V-C), and design space exploration (Section V-D), summarized in Figure 8.…”

“…2. TensorFlow code example [18]. it is possible to measure the power of real CNN architectures using actual inputs.…”

Automatic Algorithm-Based Fault Tolerance (AABFT) of Stencil Computations

“…Transient silent errors that originate in the hardware and manifest as silent data corruption pose a serious concern for software reliability. Such errors occur due to phenomena such as cosmic radiation [14,15] and hardware component aging and degradation [16,17]. This is a problem of both size and scale.…”