Adaptive loss-less data compressionmethod optimized for GPU decompression

摘要

There is no doubt that data compression is very important in computer engineering. However,rnmost lossless data compression and decompression algorithms are very hard to parallelize,rnbecause they use dictionaries updated sequentially. The main contribution of this paper is tornpresent a new lossless data compression method that we call adaptive loss-less (ALL) data compression.rnIt is designed so that the data compression ratio is moderate, but decompression canrnbe performed very efficiently on the graphics processing unit (GPU). This makes sense for applicationsrnsuch as training of deep learning, in which compressed archived data are decompressedrnmany times.To show the potentiality of ALL data compression method,we have evaluated the runningrntime using five images and five text data and comparedALLwith previously published losslessrndata compression methods implemented in the GPU, Gompresso, CULZSS, and LZW. The datarncompression ratio of ALL data compression is better than the others for eight data out of thesern10 data. Also, our GPU implementation on GeForce GTX 1080 GPU for ALL decompression runsrn84.0 to 231 times faster than theCPU implementation onCore i7-4790CPU. Further, it runs 1.22rnto 23.5 times faster than Gompresso, CULZSS, and LZWrunning on the same GPU.