Open-source graphics processing unit–accelerated ray tracer for optical simulation

Abstract

Ray tracing still is the workhorse in optical design and simulation. Its basic principle, propagating light as a set of mutually independent rays, implies a linear dependency of the computational effort and the number of rays involved in the problem. At the same time, the mutual independence of the light rays bears a huge potential for parallelization of the computational load. This potential has recently been recognized in the visualization community, where graphics processing unit (GPU)-accelerated ray tracing is used to render photorealistic images. However, precision requirements in optical simulation are substantially higher than in visualization, and therefore performance results known from visualization cannot be expected to transfer to optical simulation one-to-one. In this contribution, we present an open-source implementation of a GPU-accelerated ray tracer, based on nVidias acceleration engine OptiX, that traces in double precision and exploits the massively parallel architecture of modern graphics cards. We compare its performance to a CPU-based tracer that has been developed in parallel. © The Authors. Published by SPIE under a Creative Commons Attribution 3.0 Unported License.