Skip to Main Content
Due to the serial nature and non dependent data of the ray tracing algorithm, in particular, POV-Ray's algorithm, there is a high potential for parallelism for image rendering. Hence, speed up can be achieved by breaking the serial control flow. Breaking this control flow using threads is the optimal solution, but creating and maintaining threads is overhead and also due to POV-Ray's code structure, using the thread approach was not feasible. Therefore, another solution is to use tightly coupled and loosely coupled symmetric multiprocessor where power of multiple processors can be utilized achieving concurrency as well as parallelism. With the server-worker-client model the whole image can be splited in multithreaded and multiprocessor environment and rendering time can substantially decreased.
Date of Conference: 28-30 Oct. 2010