On the Analysis of the Discrete Element Method in Multi-core Processors
Luiz Carlos Facundo Sanches1; Ronaldo Carrion2; Carlos Héracles Moraes de Lima1
1 São Paulo State University - UNESP; 2 University of São Paulo - USP
doi:10.20906/CPS/CILAMCE2015-0321
Resumo
The Discrete Element Method (DEM) has been commonly used to simulate particle flows, though it is becoming a popular method to represent discontinuous medium lately. In computational fluids or solids mechanics, DEM is computationally expensive; however, reconfigurable computing can be used to make it less demanding. Recently, multi-core Central Processing Unit (CPU) and Graphics Processing Unit (GPU) have attracted a lot of attention in order to accelerate computer simulations in various applications. In fact, there exist architectures for multi thread parallel computation of DEM that accelerate the computation, while making effective use of the available memory. Although preliminaries results in literature show that multi-thread parallel programming performs similarly in GPUs and CPUs platforms, it is well known that DEM algorithms have higher scalability in the later. This paper describes the feasibility of using parallel computing to implement DEM on Personal Computer (PC) multi-core processors for granular hopper flow problem in silos. Preliminary results showed a speedup and almost linear scalability for scheme parallelization utilized. The values are in agreement with those encountered in literature.
Palavras-chave: Discrete Element Method; Multi-core processors; Hopper flow; Parallel computing