The ways of computational performance scaling for the numerical solution of a typical element of many CFD problems, which is also one of the most demanding to the computer resources – the Poisson equation are considered. Special attention was paid to the efficient usage of MPI algorithms together with multi-core architecture of nodes of modern cluster systems, as well as evaluating additional productivity increasing due to Intel Hyper-threading technology. The obtained results demonstrate the dependence of Hyper-threading efficiency for the Poisson equation solving on the dimensionality of the problem and the Hyper-threading ability to provide the acceleration in 3D case, which is close to the use of real cores.
parallel multithreaded and distributed computations, Hyper-Threading, Poisson equation
"Efektyvnist vykorystannia tekhnolohii virtualnoi bahatopotochnosti pry paralelnomu rozviazanni rivniannia Puassona" ,
Information Processing Systems,