Evaluating conflicts impact over shared last-level cache using public goods game on cellular automata

Regardless the fact that micro-architecture design has reached a very high level of complexity to overcome the restrains of Moore's Law, memory systems remain a bottleneck for processing systems. Taking that into account, this paper presents a model that simulates the conflict between cores for the use of shared resources, i.e. cache memory, on a multi-core processor and illustrates the impact of that conflict on the global performance of the system, using an example of Game Theory, namely Public Goods Game. In this context, Cellular Automata are the computational tool selected, in order to encapsulate the local dynamics realized on a multi-core system. Moreover, in order to validate the applicability of the model, the behavior of multicore applications running on a regular multi-core system, were profiled and compared with the results obtained from the model. This comparison illustrated that the performance of system simulated by the model is in good agreement with the profiling results of the real system configuration.