Lobachevsky State University of Nizhni Novgorod

Interactive simulation of global illumination on the GPU for animated heterogeneous scenes


General features

  1. Accelerating structures are built entirely on the GPU
  2. All calculations are performed entirely on the GPU
  3. Global illumination of heterogeneous scene is calculated without first tessellation nonpolygonal objects
  4. The maximum size of the stage for the GPU with 3GB GRAM is about 30 000 000 triangles


Performance


San-Miguel (Upper Floor), 10500K, 2.5-6 fps.
GPU NVIDIA GeForce 580, Ubuntu Linux 10.10

Performance comparisons with analogues: NVIDIA OptiX

NVIDIA GeForce GTX 680 4 Гб, Ubuntu Linux 12




Materials




Project summary

As part of this work has developed a system for the visualization of hybrid scenes with the support of the global illumination and dynamic geometry. Hybrid means combination of polygonal models with objects other types, such as the quadric surface, implicity functions and volumetric data. At the heart of system is a flexible graphics pipeline, which can be configured for various tasks with the balance speed and rendering quality. Optimized cross-platform version of the system based on NVIDIA CUDA, which provides a high performance building accelerating structures and visualization of heterogeneous scenes was developed.

Real-Time SAH BVH Construction for Ray Tracing Dynamic Scenes

This stage was aimed at the development of effective algorithms for building of full SAH BVH trees on GPU in real-time. In this work it is presupposed that all the scene objects are represented by a number of triangles (the so-called “triangle soup”), at the same time the arbitrary changes in the geometry are allowed in the process of rendering. In contrast with the other interactive BVH build methods we don’t use the simple heuristics similar to “Median Splits” and “LBVH” because they lead to decline in the quality of the generated data structure and consequently to the worse rendering performance. This work proposes a number of methods for the effective mapping of the general algorithm to the architecture of modern GPUs what allowed to accelerate the tree building up to 10 times compared to the best known GPU implementation.

Participants

  1. Denis Bogolepov
  2. Dmitry Sopin
  3. Danila Ulyanov

Project site

https://gitorious.org/+rtrt-dev

Publications

  1. Боголепов Д.К., Турлапов В.Е. Моделирование каустик в реальном времени на основе комбинированных возможностей OpenCL и шейдеров. Вестник ННГУ им.Н.И.Лобачевского, вып.3 (2), 2011. С. 180–186
  2. Боголепов Д.К., Сопин Д.П., Ульянов Д.Я., Турлапов В.Е. Построение SAH BVH деревьев для трассировки лучей с использованием графических процессоров. Ж. «Известия вузов. Приборостроение», 2011, Том 54, №10. Изд-во: С.-Пб. НИУ информационных технологий, механики и оптики (С.-ПбГУ ИТМО). С. 92-94
  3. Сопин Д.П., Боголепов Д.К., Ульянов Д.Я., Турлапов В.Е. Построение SAH BVH деревьев для трассировки лучей на GPU в реальном времени. Материалы XI Всероссийской конференции «Высокопроизводительные параллельные вычисления на кластерных системах», c. 298-303, 17-21 ноября, 2011, Нижний Новгород, Россия
  4. Denis Bogolepov, Dmitry Sopin and Danila Ulyanov. Real-Time SAH BVH Construction for Ray Tracing Dynamic Scenes. To appear in Proceedings of the 21th International Conference on Computer Graphics and Vision (GraphiCon), 2011