A Parallel Ray Tracer

The graphics industry has advanced very far from the days of pixelated 2D images on small monitors. The main focus of graphics hardware work has been achieving the most realistic images while maintaining the interactive nature of realtime applications. This focus has matured and today’s modern graphics card is able to process millions of triangles and vertices to form complex 3D images with a variety of realistic effects. However the realism achieved is mainly due to a bunch of ”hacks.” Many of the algorithms used in realtime graphics applications favor speed over the quality of work produced. These ”hacks” allow us to just barely get by without increasing the computation time significantly. To better meet the needs of todays consumers, the graphics industry will need to spend time updating the poorly looked over quality metric. Using ray tracing, we can upgrade quality. Current graphics APIs such as OpenGL and DirectX use an overly simple lighting model to compute the effects of light on a scene. Ray tracing is a way to more accurately compute lighting effects on surfaces by simulating how rays are transported around the scene. This is more realistic since it not only takes into account how the initial light from a source bounces off an object, but also takes contributions of light which have been reflected off of objects. This enables ray tracing to be very good at rendering highly reflective and shadowed surfaces [1]. Many algorithms have been developed to implement ray tracing efficiently on a standard sequential CPU. Most ray tracers are implemented by shooting out rays of light from points on the screen into the scene. This way only rays of light which contribute to the rendered image will be computed. This still happens to be very computationally intensive for any CPU today. The computation time prevents ray tracing techniques from getting into main stream graphics cards since images can no longer be rendered in real-time. However there is nothing inherently stopping this computation from happening in parallel [1]. The ”embarrassing amount of parallelism” which can be taken advantage of in ray tracing is the focus of my project. Today many companies are moving towards multiprocessor systems and multicore chips. This provides much more hardware and computatoin power to use in parallel for ray tracing applications. So it is important to see how the feasibility of using ray tracing for more applications changes with these advancements in technology.