Sorting and scanning are two fundamental primitives for constructing highly parallel algorithms. A number of libraries now provide implementations of these primitives for GPUs, but there is relatively little information about the performance of these implementations. We benchmark seven libraries for 32-bit integer scan and sort, and sorting 32-bit values by 32-bit integer keys. We show that there is a large variation in performance between the libraries, and that no one library has both optimal performance and portability.
@article{arxiv.1601.03144,
title = {A Performance Comparison of Sort and Scan Libraries for GPUs},
author = {Bruce Merry},
journal= {arXiv preprint arXiv:1601.03144},
year = {2016}
}
Comments
Preprint of an article published in Parallel Processing Letters, 25, 4, 2015, 1550007 DOI: 10.1142/S0129626415500073 \copyright\ copyright World Scientific Publishing Company <http://www.worldscientific.com/worldscinet/ppl>