Imprecise Computation and Load Sharing in Computer Generated Imaging Systems

The fast rendering of complex scenes IS a fundamental problem in imaging systems. Imprecise computation is a two level scheme for executing screen update tasks . The full level specifies the maximum computation requirement of a rendering task while the reduced level specifies the minimum computation requirement which will render a region imprecisely but acceptably. Load sharing is a task allocation scheme which shares the working load of the system among all processors, better utilizing each processor and thereby reducing overall response time. This paper describes an integration of imprecise computation and load sharing techniques which reduce image generation time while maintaining high image quality. No one has formally studied this integrated system and we do this with an analytical model which evaluates and compares CGI systems using only one of the techniques with systems which use both techniques. Our study demonstrated that the coupled system features faster task response time and high image quality even when the system has been over-loaded .