Existing fast GPU error-bounded lossy compressors have achieved high throughput through pure-GPU single-kernel designs, but their compression ratios remain limited because they typically apply a fixed first-order predictor on independent blocks. We propose FSZ, a GPU error-bounded lossy compressor that redesigns the prediction stage with three mutually reinforcing algorithmic innovations to achieve both higher compression ratios and higher throughput within a single CUDA kernel: (1) cross-block prediction state carries Lorenzo prediction state across block boundaries within 256-element tiles, eliminating 7 out of 8 boundary residuals that inflate encoding rates; (2) per-tile adaptive multi-order prediction and centering adaptively selects the best compression strategy per tile from first-order, second-order, and centering variants; and (3) a single-pass four-way evaluation exploits a mathematical property of finite differences to evaluate all variants from a single data read, enabling richer prediction within the same bandwidth budget as a fixed predictor. Experiments on NVIDIA GH200 GPU with 8 real-world application datasets show that FSZ outperforms cuSZp-P by up to 10.95x and the state-of-the-art cuSZp-O by up to 2.92x in compression ratio. Notably, these gains come with no throughput penalty: FSZ simultaneously achieves the highest average throughput (676 GB/s compression, 785 GB/s decompression) among all evaluated compressors.
Scientific applications generate massive amounts of data, necessitating effective data reduction techniques like error-bounded lossy compression. While state-of-the-art error-bounded lossy compressors effectively reduce storage footprint, they typically treat the compressed data as monolithic blobs, forcing analysts to decompress large volumes just to access specific data modalities. Conversely, existing storage solutions designed for flexible access lack the efficiency of modern compression. To bridge this gap between efficient compression and flexible data retrieval, we propose OPAL, a unified system that harmonizes high-performance compression with flexible, on-demand data access. Our first contribution is the OPAL universal framework, an adaptive architecture that endows generic lossy compressors with on-demand retrieval capabilities, enabling flexible and progressive access across multiple spatial regions, resolutions, and precisions. Our second contribution is OPALI, a customized compression algorithm designed natively with flexible data access in mind. By leveraging novel mechanisms such as shadow anchors, hybrid storage, and grouped bit-plane coding, OPALI eliminates the overhead inherent in generic wrappers. Extensive evaluations on real-world scientific datasets demonstrate that OPAL transforms standard compressors into flexible retrieval engines, while OPALI achieves superior compression ratios and retrieval performance, significantly accelerating scientific discovery.
Longtao Zhang, Ruoyu Li, Zhuoxun Yang et al.· IEEE International Symposium...· 1 citation
