cuSZp2: A GPU Lossy Compressor with Extreme Throughput and Optimized Compression Ratio

Yafan Huang; Sheng Di; Guanpeng Li; Franck Cappello

doi:10.1109/SC41406.2024.00021

Back

Conference proceeding

cuSZp2: A GPU Lossy Compressor with Extreme Throughput and Optimized Compression Ratio

Yafan Huang, Sheng Di, Guanpeng Li and Franck Cappello

Proceedings of the International Conference for High Performance Computing, Networking, Storage, and Analysis, pp.1-18

ACM Conferences

SC '24: The International Conference for High Performance Computing, Networking, Storage, and Analysis

11/17/2024

DOI: 10.1109/SC41406.2024.00021

View Online

Abstract

Existing GPU lossy compressors suffer from expensive data movement overheads, inefficient memory access patterns, and high synchronization latency, resulting in limited throughput. This work proposes cuSZp2, a generic single-kernel error-bounded lossy compressor purely on GPUs designed for applications that require high speed, such as large-scale GPU simulation and large language model training. In particular, cuSZp2 proposes a novel lossless encoding method, optimizes memory access patterns, and hides synchronization latency, achieving extreme end-to-end throughput and optimized compression ratio. Experiments on NVIDIA A100 GPU with 9 real-world HPC datasets demonstrate that, even with higher compression ratios and data quality, cuSZp2 can deliver on average 332.42 and 513.04 GB/s end-to-end throughput for compression and decompression, respectively, which is around 2× of existing pure-GPU compressors and 200× of CPU-GPU hybrid compressors.

Computer systems organization

Computer systems organization -- Architectures

Computer systems organization -- Architectures -- Parallel architectures

Computer systems organization -- Architectures -- Parallel architectures -- Single instruction, multiple data

Computing methodologies

Computing methodologies -- Computer graphics

Computing methodologies -- Computer graphics -- Graphics systems and interfaces

Computing methodologies -- Computer graphics -- Graphics systems and interfaces -- Graphics processors

Computing methodologies -- Computer graphics -- Image compression

Information systems

Information systems -- Data management systems

Information systems -- Data management systems -- Data structures

Information systems -- Data management systems -- Data structures -- Data layout

Information systems -- Data management systems -- Data structures -- Data layout -- Data compression

Theory of computation

Theory of computation -- Design and analysis of algorithms

Theory of computation -- Design and analysis of algorithms -- Data structures design and analysis

Theory of computation -- Design and analysis of algorithms -- Data structures design and analysis -- Data compression

Details

Title: Subtitle: cuSZp2: A GPU Lossy Compressor with Extreme Throughput and Optimized Compression Ratio
Creators: Yafan Huang - Computer Science Department, University of Iowa, Iowa City, IA, USA
Sheng Di - Argonne National Laboratory
Guanpeng Li - Computer Science Department, University of Iowa, Iowa City, IA, USA
Franck Cappello - Argonne National Laboratory
Resource Type: Conference proceeding
Publication Details: Proceedings of the International Conference for High Performance Computing, Networking, Storage, and Analysis, pp.1-18
Conference: SC '24: The International Conference for High Performance Computing, Networking, Storage, and Analysis
Series: ACM Conferences
DOI: 10.1109/SC41406.2024.00021
Publisher: IEEE Press
Grant note: U.S. Department of Energy, Office of Science, Advanced Scientific Computing Research (ASCR): DE-AC02-06CH11357, DE-SC0024559 National Science Foundation: OAC-2003709, OAC-2104023, OAC-2311875, OAC-2211538
This material was based upon work supported by the U.S. Department of Energy, Office of Science, Advanced Scientific Computing Research (ASCR), under contract DE-AC02-06CH11357 and DE-SC0024559. The material was also supported by the National Science Foundation under Grant OAC-2003709, OAC-2104023, OAC-2311875, and OAC-2211538. The experimental resource for this paper was provided by the Laboratory Computing Resource Center on the Swing cluster at Argonne National Laboratory.
Language: English
Date published: 11/17/2024
Academic Unit: Computer Science
Record Identifier: 9984748158502771

Metrics

22 Record Views