Graphics and Media Lab

Review

Photorealistic rendering used in many spheres such as cinema, architecture design, light design, computer games (light maps) and others. The speed is a central problem of photorealistic rendering. If films, for example, it may take about several hours or every a day to synthesize one frame (with complex scenes and large resolutions). That’s why, acceleration of photorealistic rendering is very important.

Êîëè÷åñòâî îïåðàöèé ñ ïëàâàþùåé òî÷êîé â ñåêóíäó	Ïðîïóñêíàÿ ñïîñîáíîñòü ïàìÿòè

Ñðàâíåíèå CPU è GPU.

On the other hand, Graphics Processing Units (GPUs) became very powerful at present (pic. 1). But the price for that power is the universality. Programming for GPUs is very specific. You need to take into account architecture specifics and constraints when you're dealing with GPUs. And it is very hard to get close to pick GPU performance. So, one of our tasks is to efficiently utilize GPU resources when coding some algorithm.

Project description

The main goal of our project is to accelerate photorealistic rendering using GPUs power. We develop a system that support followed algorithms:

• Backward ray tracing
• Path tracing
• Photon mapping

See http://ray-tracing.ru/en for more information.

Conference Room (281K triangles); 640x480; GTX260; shadows from one point light; 27 FPS.

spheres; 640x480; GTX260; one specular bounce; 20 FPS.

Soft shadows and glossy reflections. Modified path tracing.

Conference Room (280K triangles); 260K photons; 2 FPS.	Conference Room (280K triangles); 2M photons; 3.22 seconds per frame.

Team

• Vladimir Frolov

Publications

• Ôðîëîâ , Èãíàòåíêî . Èíòåðàêòèâíàÿ òðàññèðîâêà ëó÷åé íà ãðàôè÷åñêîì ïðîöåññîðå ñ ïðèìåíåíèåì òåõíîëîãèè CUDA. GraphiCon'2008, 2008 (download)
• Ôðîëîâ , Èãíàòåíêî . Èíòåðàêòèâíàÿ òðàññèðîâêà ëó÷åé è ôîòîííûå êàðòû íà GPU. GraphiCon'2009, 2009 (download)