Granted, real-time path tracing has been a long standing promise, with many different ideas about what the term means exactly. Depending on who you ask, it is already a reality or it will never see the light of day (or something in between). Undisputed is the fact that path tracing, for any kind of image quality, requires a lot of computation, the more the better. This (and the pretty output of course) makes it the ideal playground for HPC enthusiasts such as myself.

The raPT project is about bringing real-time path tracing to big cluster architectures with massive parallelism, while at the same time optimizing kernel algorithms / implementations down to the last SIMD instruction. The motivation is to reach path tracing performance sufficient for applications like interactive movie rendering (for people who own several KW of compute power).

The raPT project is embedded into my doctoral thesis at the Fraunhofer ITWM HPC department in cooperation with the University of Kaiserslautern Computer Graphics Group.

Thanks, Valentin

Publications