Weighted Random Sampling on GPUs

TL;DR

This paper adapts the PSA algorithm for alias table construction to GPUs, achieving significant speedups and energy efficiency improvements for weighted random sampling compared to CPU implementations.

Contribution

It introduces a GPU-based parallel construction algorithm for alias tables, with optimized memory access and search techniques, outperforming CPU methods in speed and energy efficiency.

Findings

17x faster alias table construction on GPU

Up to 24x higher sampling throughput

Multiple times less energy consumption

Abstract

An alias table is a data structure that allows for efficiently drawing weighted random samples in constant time and can be constructed in linear time. The PSA algorithm by H\"ubschle-Schneider and Sanders is able to construct alias tables in parallel on the CPU. In this report, we transfer the PSA algorithm to the GPU. Our construction algorithm achieves a speedup of 17 on a consumer GPU in comparison to the PSA method on a 16-core high-end desktop CPU. For sampling, we achieve an up to 24 times higher throughput. Both operations also require several times less energy than on the CPU. Adaptations helping to achieve this include changing memory access patterns to do coalesced access. Where this is not possible, we first copy data to the faster shared memory using coalesced access. We also enhance a generalization of binary search enabling to search for a range of items in parallel. Besides naive sampling, we also give improved batched sampling algorithms.