Scheduling nonlinear computational loads

A scheduling model for a tree network is studied where the computation time for each node is nonlinear in the size of the assigned load. Optimal load allocation and speedup for simultaneous load distribution for a quadratic nonlinearity are obtained using simple equations. An iterative solution for sequential load distribution is presented for a nonlinearity of arbitrary power. Superlinear speedup is possible when computational complexity is nonlinear in the size of assigned loads. Aerospace applications include spectrum computation, radar and sensor data processing, and satellite image processing.