Floating-point Parallel Compression Based on GPGPU

In recent years, the advance of information technology rapidly produces huge collection of ∞oating-point data, and it becomes necessary to design efiective compression algorithm for the storage and transmission of these data. However, most of current schemes focus on serial computing and thus are not e‐cient enough for real-time applications. In this paper, a parallel compression method is proposed for lossless compression of ∞oating-point data. Two policies are adopted in the compression pipeline. The flrst is organizing multiple ∞oating-point values into a prediction block for prediction, and the second is splitting big multi-dimension data into multiple cuboids. The result shows that the proposed method are either more time e‐cient or getting better compression ratios than previous method. Our method can be applied in specifled application scenario, such as scientiflc data compression, streaming transformation, etc.