Implementing multi-agent systems on GPU 435
step of all agents in the model, to ensure equity, and to minimize the numbers
of synchronizations. The minimization is done by factoring the distribution of
carbon in the model in two separated steps at the beginning and the end of
each iteration rather than at multiple points of the execution.
18.5.2 MAS execution workflow
Many MAS simulations decompose their execution process into discrete
evolution steps where each step represents a quantum of time (minimal unit of
time described). At the end of each step many global data, graphical displays
or output files are updated. This execution model may not correctly fit on
GPU platforms as they assume more or less a batch-like workflow model. The
execution model