HYDRA: Hybrid Data Multiplexing and Run-time Layer Configurable DNN Accelerator

Document detail

ID

oai:arXiv.org:2409.04976

Topic

Computer Science - Hardware Archit... Computer Science - Artificial Inte... Computer Science - Computer Vision... Electrical Engineering and Systems...

Author

Kumar, Sonu Gupta, Komal Raut, Gopal Lokhande, Mukul Vishvakarma, Santosh Kumar

Year

2024

listing date

1/1/2025

Keywords

configurable proposed dnn hydra architecture science computer

Metrics

Abstract

Deep neural networks (DNNs) offer plenty of challenges in executing efficient computation at edge nodes, primarily due to the huge hardware resource demands.

The article proposes HYDRA, hybrid data multiplexing, and runtime layer configurable DNN accelerators to overcome the drawbacks.

The work proposes a layer-multiplexed approach, which further reuses a single activation function within the execution of a single layer with improved Fused-Multiply-Accumulate (FMA).

The proposed approach works in iterative mode to reuse the same hardware and execute different layers in a configurable fashion.

The proposed architectures achieve reductions over 90% of power consumption and resource utilization improvements of state-of-the-art works, with 35.21 TOPSW.

The proposed architecture reduces the area overhead (N-1) times required in bandwidth, AF and layer architecture.

This work shows HYDRA architecture supports optimal DNN computations while improving performance on resource-constrained edge devices.

Kumar, Sonu,Gupta, Komal,Raut, Gopal,Lokhande, Mukul,Vishvakarma, Santosh Kumar, 2024, HYDRA: Hybrid Data Multiplexing and Run-time Layer Configurable DNN Accelerator