The authors describe a fully distributed, parallel algorithm forray-tracing problem. Load balancing is achieved through the use of combdistribution to roughly assign the same amount of pixels to eachprocessor first, and then dynamically redistribute excessive loads amongprocessors to keep each processor busy. In this model, there is no needfor a master node to be responsible for dynamic scheduling. When eachnode finishes its job, it just requests an extra job from one of itsneighbors. The authors implement their algorithm on Intel DeltaTouchstone machine with 2-D mesh network topology and provide simulationresults. With their scheme, they can get good speedup and highefficiency without much communication overhead
展开▼