L3: A GPU-Native Co-Designed Data Format for Learned Lossless Lightweight Compression
Proceedings of the ACM on Management of Data · 2026-05-18 · 4 (3) · 1-27
DOI: 10.1145/3802078 ↗At a glance
Journal article in Proceedings of the ACM on Management of Data, published online on 2026-05-18. Written by Youyang Xia and 6 co-authors. 73 references. Funded by: Beijing Science and Technology Project.
Abstract (original language)
Learned Compression achieves strong CPU performance but lacks a GPU-native format, limiting its use in GPU analytics. We present L3, a GPU-native Learned Lossless Lightweight Compression format that enables end-to-end on-device processing with efficient compression, high-throughput decompression, and fast random access on GPU. On NVIDIA GPUs, a warp is a group of 32 threads; we refer to each thread as a lane (lane id 0–31), and call a layout lane-major when each lane's words are stored contiguously. L3 introduces three tightly coupled components built around the SLAP Vertical layout. First, the L3 Storage Layout (SLAP) stores bit-packed residual streams in a lane-major organization, i.e., residual words are laid out lane by lane so each warp lane consumes a contiguous word sequence in memory, exploiting the GPU L1 sector cache for implicit prefetching and high reuse during unpacking. Second, the Warp-Cooperative Learned Decompression Module maps each partition to one thread block and decodes warp tiles using per-lane bit readers, branchless bit extraction, and a bit-exact no-FMA FP64 finite-difference predictor. Third, the GPU-Native Learned Compression Pipeline builds adaptive partitions via bulk delta-bits analysis, scan/compaction, and an odd-even GPU merge loop, then packs residuals directly into the final SLAP Vertical layout on the device. L3 achieves high performance on modern GPUs. It encodes 3–6× faster than Tile and FastLanes-GPU and sustains 1.08–1.90 TB/s decompression throughput, comparable to the fastest lightweight GPU codecs. On correlated datasets, L3 reaches up to 77× compression while remaining competitive on weakly correlated inputs. For random access, L3 maintains 1.2–2.6 Billion queries/s and outperforms Tile-DFOR/Tile-RFOR by 5–10×. On SSB with unified query plans, L3 achieves the lowest average latency (1.14 ms), matching or outperforming state-of-the-art GPU baselines.
Funding
- Beijing Science and Technology Project
Cite
APA
Xia, Y., Zhang, F., Pan, J., Liu, Y., Guan, J., Zhang, H., & Du, X. (2026). L3: A GPU-Native Co-Designed Data Format for Learned Lossless Lightweight Compression. Proceedings of the ACM on Management of Data, 4(3), 1-27. https://doi.org/10.1145/3802078
Vancouver
Xia Y, Zhang F, Pan J, Liu Y, Guan J, Zhang H, et al. L3: A GPU-Native Co-Designed Data Format for Learned Lossless Lightweight Compression. Proceedings of the ACM on Management of Data. 2026;4(3):1-27. doi:10.1145/3802078
BibTeX
@article{Xia2026,
author = {Xia, Youyang and Zhang, Feng and Pan, Junda and Liu, Yihao and Guan, Jiawei and Zhang, Huanchen and Du, Xiaoyong},
title = {L3: A GPU-Native Co-Designed Data Format for Learned Lossless Lightweight Compression},
journal = {Proceedings of the ACM on Management of Data},
year = {2026},
volume = {4},
number = {3},
pages = {1--27},
doi = {10.1145/3802078}
}
EUVIMED