A Benchmark Set of Highly-efficient CUDA and OpenCL Kernels and its Dynamic Autotuning with Kernel Tuning Toolkit

TL;DR

This paper presents a benchmark set of autotunable CUDA and OpenCL kernels and demonstrates that dynamic autotuning during runtime can optimize performance across diverse hardware and input conditions, enhancing portability.

Contribution

It introduces a new benchmark suite of ten kernels and a dynamic autotuning approach integrated into the Kernel Tuning Toolkit for real-time performance optimization.

Findings

Most kernels reach near-peak performance with autotuning.

Dynamic tuning effectively adapts performance during application runtime.

Rationally designed tuning spaces enable feasible real-time autotuning.

Abstract

Autotuning of performance-relevant source-code parameters allows to automatically tune applications without hard coding optimizations and thus helps with keeping the performance portable. In this paper, we introduce a benchmark set of ten autotunable kernels for important computational problems implemented in OpenCL or CUDA. Using our Kernel Tuning Toolkit, we show that with autotuning most of the kernels reach near-peak performance on various GPUs and outperform baseline implementations on CPUs and Xeon Phis. Our evaluation also demonstrates that autotuning is key to performance portability. In addition to offline tuning, we also introduce dynamic autotuning of code optimization parameters during application runtime. With dynamic tuning, the Kernel Tuning Toolkit enables applications to re-tune performance-critical kernels at runtime whenever needed, for example, when input data changes. Although it is generally believed that autotuning spaces tend to be too large to be searched during application runtime, we show that it is not necessarily the case when tuning spaces are designed rationally. Many of our kernels reach near peak-performance with moderately sized tuning spaces that can be searched at runtime with acceptable overhead. Finally we demonstrate, how dynamic performance tuning can be integrated into a real-world application from cryo-electron microscopy domain.