Scheduling Time-Critical Graphics on Multiple Processors * H95- 25870

This paper describes an algorithm for the scheduling of time-critical rendering and computation tasks on single- and multiple-processor architectures, with minimal pipelining. It was developed to man- age scientific visualization scenes consisting of hundreds of objects, each of which can be computed and displayed at thousands of pos- sible resolution levels. The algorithm generates the time-critical schedule using progressive-refinement techniques; it always returns a feasible schedule and, when allowed to run to completion, pro- duces a near-optimal schedule which takes advantage of almost the entire multiple-processor system.