High-Performance FPGA-Based CNN Accelerator With Block-Floating-Point Arithmetic
Autor: | Xiangyang Ji, Zhourui Song, Xiaocong Lian, Wei Zhou, Jiwu Dai, Zhenyu Liu |
---|---|
Rok vydání: | 2019 |
Předmět: |
Floating point
Artificial neural network Computer science Quantization (signal processing) Rounding 02 engineering and technology Convolutional neural network 020202 computer hardware & architecture Memory management Hardware and Architecture 0202 electrical engineering electronic engineering information engineering Electrical and Electronic Engineering Block floating-point Arithmetic Field-programmable gate array Software |
Zdroj: | IEEE Transactions on Very Large Scale Integration (VLSI) Systems. 27:1874-1885 |
ISSN: | 1557-9999 1063-8210 |
Popis: | Convolutional neural networks (CNNs) are widely used and have achieved great success in computer vision and speech processing applications. However, deploying the large-scale CNN model in the embedded system is subject to the constraints of computation and memory. An optimized block-floating-point (BFP) arithmetic is adopted in our accelerator for efficient inference of deep neural networks in this paper. The feature maps and model parameters are represented in 16-bit and 8-bit formats, respectively, in the off-chip memory, which can reduce memory and off-chip bandwidth requirements by 50% and 75% compared to the 32-bit FP counterpart. The proposed 8-bit BFP arithmetic with optimized rounding and shifting-operation-based quantization schemes improves the energy and hardware efficiency by three times. One CNN model can be deployed in our accelerator without retraining at the cost of an accuracy loss of not more than 0.12%. The proposed reconfigurable accelerator with three parallelism dimensions, ping-pong off-chip DDR3 memory access, and an optimized on-chip buffer group is implemented on the Xilinx VC709 evaluation board. Our accelerator achieves a performance of 760.83 GOP/s and 82.88 GOP/s/W under a 200-MHz working frequency, significantly outperforming previous accelerators. |
Databáze: | OpenAIRE |
Externí odkaz: |