Falcons: Pushing the Speed-Ratio Limits of Lossy Compression for Parallel Scientific Applications

DOI10.48550/ARXIV.2610.41209MaRDI QIDQ7409319arXivFDO

Authors Bing Luo, Jin-Yang Liu, Xinyang Ma, Zedong Liu, Wenjing Huang, Yida Guo, Xiaofeng Hou, Guang-Ming Tan, Minyi Guo, Dingwen Tao, Bing Lu, Jinyang Liu, Quantong Fu, Chenwei Hou, Yida Gu, Xingchen Liu, Zheng Wei, Anbang Wu, Chao Li, Guangming Tan

Publication date 30 October 2026

Published in arXiv

Tldr We present Falcons, a broadly deployable error-bounded lossy compressor that achieves high compression ratios with ultra-fast throughput, minimizing overhead in parallel scientific applications by integrating memory-regularized interpolation, distribution-adaptive lossless coding, and application-oriented optimizations for multiple parallelization schemes. Across evaluations on cutting-edge supercomputers, Falcons outperforms state-of-the-art compressors by up to 29% compression ratio improvement and 255% throughput improvement.

Full work available at URL https://doi.org/10.48550/ARXIV.2610.41209

arXiv classification

cs.DC

Cites work

• A public turbulence database cluster and applications to study Lagrangian evolution of velocity increments in turbulence
• Asymmetric numeral systems: entropy coding combining speed of Huffman coding with compression rate of arithmetic coding
• MGARD+: Optimizing Multilevel Methods for Error-Bounded Scientific Data Reduction
• Multilevel techniques for compression and reduction of scientific data -- the univariate case
• Universal codeword sets and representations of the integers

This page was built for publication: Falcons: Pushing the Speed-Ratio Limits of Lossy Compression for Parallel Scientific Applications

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7409319)