Validation of Gpu Lighting Simulation in Naturally and Artificially Lit Spaces

Daylight in buildings is both aesthetically pleasing and a sustainable means of offsetting costs for space conditioning and electric lighting. However, poor use of daylight can cause glare that impedes worker productivity. Traditional means of predicting lighting levels for indoor spaces through simulation are timeconsuming, which inhibits exploration of daylighting potential in new buildings. This paper compares results from two lighting simulation engines, Radiance and Accelerad, with measurements taken in physical spaces. Radiance is an established backward ray tracer that runs on the CPU, whereas Accelerad is a recent porting of Radiance’s algorithms for the GPU. Vertical eye illuminance, daylight glare probability, and monitor contrast ratio serve as metrics for comparison. Radiance and Accelerad produce similar errors in visual comfort metrics of around 10%, and Accelerad generates solutions between 3 and 24 times faster than Radiance in the tested scenes. These speedups are expected to scale up on new generations of graphics hardware.