Reconfigurable Architecture for Real-time Decoding of Canonical Huffman Codes

Abstract

Data compression is an important algorithm which has found its use in modern day algorithms such as Convolutional Neural Networks (CNNs). Reconfigurable platforms (like FPGAs) have strong capabilities to implement time complex tasks like CNNs, however, these algorithms present a big challenge due to high resource demand. Data compression is one of the most utilized techniques to reduce memory utilization in FPGAs. The weights of CNN architecture are usually encoded to store in FPGA. In this paper, we propose design of an efficient decoder based on Canonical Huffman that can be utilized for the efficient decompression of weights in CNN. The proposed design makes use of Hash functions to effectively decode the weights eliminating the need for searching dictionary. The proposed design decodes a single weight in a single clock cycle. Our proposed design has a maximum frequency of 408.97MHz utilizing 1% of system LUTs when tested for Aritix 7 platform.