MIT Sustainable Design Lab
Accelerad | Daylighting Simulation on the GPU

Nathaniel Jones and Christoph Reinhart

Accelerad has a new home!

https://nljones.github.io/Accelerad/

Accelerad is a free suite of programs for lighting and daylighting analysis and visualization. The suite uses physically-based backward ray tracing algorithms inspired by Lawrence Berkeley National Laboratory’s popular Radiance software suite by Greg Ward. These algorithms are accelerated up to twenty times faster using OptiX™, a ray tracing engine built for the graphics processor unit (GPU). The acceleration factor scales with the number of available GPUs and is expected to increase on new generations of hardware. In order to allow for smooth adoption among Radiance users and software developers, Accelerad maintains compatibility with Radiance scene and output file formats and uses a subset of Radiance’s material modifiers and command-line arguments.

Accelerad is designed to make new types of simulation possible that are currently too time-consuming for computation on the CPU. It is ideal for rpict renderings with large numbers of ambient bounces or rtrace calculations over many thousands of sensor points.

RTRACE speedup

In rtrace, a 20-fold speed increase is possible using the GPU.

RPICT single image speedup

Our implementation of Radiance on the GPU shows a 7-fold speed increase in a simple scene without irradaince caching.

RPICT multiple image speedup

Further speedups are possible by batching multiple images in the same scene.

RPICT irradiance caching

Irradiance caching on the GPU produces a twenty-fold speedup.

RPICT irradiance caching

Smoother gradients are produced by creating the irradiance cache prior to the final gather.


Sneak Peek

These demonstration videos preview Accelerad's real-time interface for glare and visual discomfort prediction. This interface, AcceleradRT, is the first major departure from Accelerad's method of porting Radiance functionality to the GPU. Instead, AcceleradRT adds new functions and combines the capabilities of other programs such as rpict, rvu, evalglare, and wxfalsecolor with GPU acceleration.

AcceleradRT includes an experimental feature, AcceleradVR, for experiencing the luminous environment in virtual reality. AcceleradVR uses third party software to stream AcceleradRT's real-time rendering to a smart phone. The stereo image can be viewed with any VR headset, including Google cardboard. Use head movement, your keyboard, or any USB game controller to move through your virtual space.

Put on your 3D glasses! This early test video will take you on a walk-through and fly-through of Gund Hall on the Harvard campus. Use a virtual reality headset to best experience the real-time falsecolor rendering.