Interactive indirect lighting computed in the cloud.

In this talk we present two different ways to map indirect lighting computation onto the new "pipeline" represented by the Cloud. Some of this new pipeline is familiar, but some is quite different. Given the vast space of possible hardware configurations and the many known algorithms for light transport, we explore two quite different approaches to indirect lighting and empirically examine their performance on the Cloud. As this examination is largely empirical, we use test scenes with significant geometric and texture complexity. We present detailed optimizations and measurements of bandwidth and latency performance. We further examine the amortization of indirect lighting across multiple users, a key potential benefit of Cloud-based systems.