Onesweep: A Faster Least Significant Digit Radix Sort for GPUs
NVIDIA Corporation
arXiv:2206.01784 [cs.DC], (3 Jun 2022)
@misc{https://doi.org/10.48550/arxiv.2206.01784,
doi={10.48550/ARXIV.2206.01784},
url={https://arxiv.org/abs/2206.01784},
author={Adinets, Andy and Merrill, Duane},
keywords={Distributed, Parallel, and Cluster Computing (cs.DC), Data Structures and Algorithms (cs.DS), FOS: Computer and information sciences, FOS: Computer and information sciences, F.2.2; D.1.3},
title={Onesweep: A Faster Least Significant Digit Radix Sort for GPUs},
publisher={arXiv},
year={2022},
copyright={Creative Commons Attribution Non Commercial No Derivatives 4.0 International}
}
We present Onesweep, a least-significant digit (LSD) radix sorting algorithm for large GPU sorting problems residing in global memory. Our parallel algorithm employs a method of single-pass prefix sum that only requires ~2n global read/write operations for each digit-binning iteration. This exhibits a significant reduction in last-level memory traffic versus contemporary GPU radix sorting implementations, where each iteration of digit binning requires two passes through the dataset totaling ~3n global memory operations. On the NVIDIA A100 GPU, our approach achieves 29.4 GKey/s when sorting 256M random 32-bit keys. Compared to CUB, the current state-of-the-art GPU LSD radix sort, our approach provides a speedup of ~1.5x. For 32-bit keys with varied distributions, our approach provides more consistent performance compared to HRS, the current state-of-the-art GPU MSD radix sort, and outperforms it in almost all cases.
June 12, 2022 by hgpu
