Problem
Rendering research needs an implementation environment where new techniques can be compared, integrated and investigated. Separate prototypes make it harder to understand how an idea interacts with the rest of the frame.
Constraints
The framework serves multiple workloads: acceleration-structure building, 3D graphics processing, ray tracing and machine-learning training and inference. The implementation has to support investigation across those boundaries.
What we did
Traverse built the Breda framework as an in-house collection of rendering and computational tools. Publicly described work includes wavefront and pipeline-based path tracers, dynamic global illumination, micro-polygon rendering and volumetric transport.
The dynamic global-illumination path evaluates material properties rather than requiring artists to assign proxy bounce colours. The volumetric path tracer explores multiple scattering with up to 16 light bounces.
Results
The published framework demonstrates these techniques in a shared implementation. Per-device timing and memory comparisons are not published here, so no performance uplift is claimed.
