TIPS: Cost of Implicit Nonlinear Analysis
Parallel Computing in Marc Marc supports parallelisation of both the element loops and the equation solution. They are controlled by two different mechanisms – one of them is “free”, the other requires an additional license (or FEATURE used from MSC One tokens). Which one is most effective for a particular analysis will be the subject of another blog entry. The following sections show the two controlling options in the Mentat GUI, and then the results of a benchmark to show the speed obtained using different solvers in combination with parallelisation. Enabling Parallel Processing There are two parts to control that are displayed in Mentat: In the flowchart below, the yellow boundary show the element loops that are affected by DDM, and the green boundary show the matrix solution phase that is controlled by solvers such as Pardiso: Alternatively, to access through the solver command line: DDM via –nps run_marc option Solver via –nthread run_marc...