Understanding the efficiency of ray traversal on GPUs

TLDR

GPU ray tracing performance is poorly understood, despite many fast methods, and observations may generalize beyond NVIDIA GPUs. The study aims to determine whether current GPU ray tracing methods reach theoretical limits and to identify causes of any performance gap. The authors map ray tracing operations onto SIMD/SIMT GPUs, compare measurements to a performance simulator, and propose a solution to close the gap. They find existing methods lag 1.5–2.5× behind theoretical optimum mainly due to hardware work‑distribution inefficiencies, and their solution yields the fastest GPU ray tracer, achieving strong results on primary, ambient occlusion, and diffuse interreflection rays.

Abstract

We discuss the mapping of elementary ray tracing operations---acceleration structure traversal and primitive intersection---onto wide SIMD/SIMT machines. Our focus is on NVIDIA GPUs, but some of the observations should be valid for other wide machines as well. While several fast GPU tracing methods have been published, very little is actually understood about their performance. Nobody knows whether the methods are anywhere near the theoretically obtainable limits, and if not, what might be causing the discrepancy. We study this question by comparing the measurements against a simulator that tells the upper bound of performance for a given kernel. We observe that previously known methods are a factor of 1.5--2.5X off from theoretical optimum, and most of the gap is not explained by memory bandwidth, but rather by previously unidentified inefficiencies in hardware work distribution. We then propose a simple solution that significantly narrows the gap between simulation and measurement. This results in the fastest GPU ray tracer to date. We provide results for primary, ambient occlusion and diffuse interreflection rays.