A GPU-based implementation of predictive partitioned vector quantization for compression of ultraspectral sounder data

Abstract

Recently there is a boom on the use of graphic processor units (GPU) for speedup of scientific computations. By identifying the time dominant portions of the code that can be executed in parallel, significant speedup can be achieved by a GPU-based implementation. For the voluminous ultraspectral sounder data, lossless compression is desirable to save storage space and transmission time without losing precision in retrieval of geophysical parameters. Predictive partitioned vector quantization (PPVQ) has been proven to be an effective lossless compression scheme for ultraspectral sounder data. It consists of linear prediction, bit partition, vector quantization, and entropy coding. Two most time consuming stages of linear prediction and vector quantization are chosen for GPU-based implementation. By exploiting the data parallel characteristics of these two stages, a speedup of 42x has been achieved in our GPU-based implementation of the PPVQ compression scheme.