Non-preemptive multi-constrain scheduling for multiprocessor with hopfield neural network

In this paper, task scheduling for non-preemptive multi-constrained multi-processor systems was presented. The proposed model based on discrete Hopfield Neural network augmented with a methodology for weighting constrains to form overall network energy function. The network augmented with a layer to handle network re-initialization, based on min-max algorithm, case of local minima trapped without an acceptable solution. The proposed neural network solution does not require a predetermined scheduling length. Constrains included in the study are: task time, precedence, resources conflict, task dead time, and favoring tasks of the same setup to run on the same processor to suit reconfigurable hardware.