A Memory Bandwidth-Efficient Hybrid Radix Sort on GPUs

TL;DR

This paper introduces a memory bandwidth-efficient hybrid radix sort algorithm for GPUs that significantly reduces memory transfers, leading to substantial speed-ups in sorting large datasets compared to previous GPU and CPU methods.

Contribution

It presents a novel GPU radix sort approach that nearly halves memory transfers, boosting sorting performance, and extends it with a pipelined heterogeneous algorithm for larger or off-GPU data.

Findings

Achieves 2.32x faster sorting of 2GB data over state-of-the-art GPU radix sort.

Maintains at least 1.66x speed-up on skewed distributions.

Improves end-to-end sorting of 64GB data by over 2x compared to CPU-based radix sort.

Abstract

Sorting is at the core of many database operations, such as index creation, sort-merge joins, and user-requested output sorting. As GPUs are emerging as a promising platform to accelerate various operations, sorting on GPUs becomes a viable endeavour. Over the past few years, several improvements have been proposed for sorting on GPUs, leading to the first radix sort implementations that achieve a sorting rate of over one billion 32-bit keys per second. Yet, state-of-the-art approaches are heavily memory bandwidth-bound, as they require substantially more memory transfers than their CPU-based counterparts. Our work proposes a novel approach that almost halves the amount of memory transfers and, therefore, considerably lifts the memory bandwidth limitation. Being able to sort two gigabytes of eight-byte records in as little as 50 milliseconds, our approach achieves a 2.32-fold improvement over the state-of-the-art GPU-based radix sort for uniform distributions, sustaining a minimum speed-up of no less than a factor of 1.66 for skewed distributions. To address inputs that either do not reside on the GPU or exceed the available device memory, we build on our efficient GPU sorting approach with a pipelined heterogeneous sorting algorithm that mitigates the overhead associated with PCIe data transfers. Comparing the end-to-end sorting performance to the state-of-the-art CPU-based radix sort running 16 threads, our heterogeneous approach achieves a 2.06-fold and a 1.53-fold improvement for sorting 64 GB key-value pairs with a skewed and a uniform distribution, respectively.