@inproceedings{91aae368232141abb24cedb8d6d92a7b,
title = "An OpenCL{\texttrademark} implementation of WebP accelerator on FPGAs",
abstract = "With the development of cloud computing, the super-large scale of image data has bring severe challenges for the storage cost and network bandwidth in data centers. In order to alleviate the present situation effectively, WebP has replaced the current mainstream image file format due to its better compression efficiency. In this paper, we provide an OpenCL implementation of WebP accelerator on FPGAs to optimize the performance of WebP Lossy Compression Algorithm. Our accelerator makes use of a heavily-pipelined custom hardware implementation to achieve a high throughput ~450MPixel/s. The performance-per-watt of our OpenCL implementation on Intel{\textquoteright}s Arria 10 device is 8.32x better than a highly-tuned CPU implementation on Intel Xeon E5-2690v3 with 24 thread cores. Additionally, the delay time per image can be reduced to ~90% by the data parallelism and macroblock pipelining on FPGAs. Finally, our OpenCL{\texttrademark} implementation of WebP accelerator on FPGAs is more competitive for data centers to achieve higher performance and lower cost.",
keywords = "Accelerator, FPGAs, OpenCL, WebP",
author = "Zhenhua Guo and Baoyu Fan and Yaqian Zhao and Xuelei Li and Shixin Wei and Long Li",
note = "Publisher Copyright: {\textcopyright} Springer International Publishing AG, part of Springer Nature 2018.; 14th International Symposium on Applied Reconfigurable Computing, ARC 2018 ; Conference date: 02-05-2018 Through 04-05-2018",
year = "2018",
doi = "10.1007/978-3-319-78890-6_46",
language = "English",
isbn = "9783319788890",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "578--589",
editor = "Nikolaos Voros and Georgios Keramidas and Christos Antonopoulos and Michael Huebner and Diniz, {Pedro C.} and Diana Goehringer",
booktitle = "Applied Reconfigurable Computing",
address = "Germany",
}