Data parallel multi-GPU path tracing using ray queue cycling

Abstract

We propose a novel approach to data-parallel path tracing on single-node/multi-GPU hardware that builds on ray forwarding,but which aims—above all else—at generality and practicability. We do this byavoidingany attempts at reducing the numberof traces or forward operations performed, and instead focus on always using all GPUs’ aggregate compute and bandwidth toeffectively trace each ray on every GPU. We show that—counter-intuitively—this is both feasible and desirable; and that whenrun on typical data-center/cloud hardware, the resulting framework not only achieves good performance and scalability, butalso comes with significantly fewer limitations, assumptions, or preprocessing requirements than existing techniques.