NetPU: Prototyping a Generic Reconfigurable Neural Network Accelerator Architecture

FPGA-based Neural Network (NN) accelerator is a rapidly advancing subject in recent research. Related works can be classified as two hardware architectures: i) Heterogeneous Streaming Dataflow (HSD) architecture and ii) Processing Element Matrix (PEM) architecture. HSD architecture explores the reco...

Full description

Saved in:
Bibliographic Details
Published in2022 International Conference on Field-Programmable Technology (ICFPT) p. 1
Main Authors Liu, Yuhao, Rai, Shubham, Ullah, Salim, Kumar, Akash
Format Conference Proceeding
LanguageEnglish
Published IEEE 05.12.2022
Subjects
Computer architecture
Field programmable gate arrays
Hardware
Microcontrollers
Neurons
Runtime
Software
Online AccessGet full text

Cover

More Information
Summary:FPGA-based Neural Network (NN) accelerator is a rapidly advancing subject in recent research. Related works can be classified as two hardware architectures: i) Heterogeneous Streaming Dataflow (HSD) architecture and ii) Processing Element Matrix (PEM) architecture. HSD architecture explores the reconfigurability of FPGAs to support the customization and optimization of hardware design to implement a complete network on FPGA for one given trained model. PEM architecture achieves relatively generic support for different network models, essentially implementing the neuron processing modules on the FPGA scheduled by the runtime software environment. In summary, the HSD architecture requires more resources with simplified runtime software control. The PEM architecture consumes fewer resources than the HSD architecture. However, the runtime software environment can be a heavy payload for lightweight systems, such as the low-power microcontroller of IoT or edge devices.
DOI:10.1109/ICFPT56656.2022.9974206