There are improvements in overall stability and performance while solving
a simulation, especially for linear static studies with multiple load cases.
A major driver of the performance improvements is based on the fact that the solution
formulation is restructured for optimized performance. The optimized reconfiguration is
limited to the Intel Direct Sparse solver only. Noticeable improvements in solution time
are observed for linear static studies with up to 25 load cases. Note the following:
- The optimized performance of the Intel Direct Sparse solver is
realized for Pressure, Force, and Torque load types. All other load types,
connectors, and non-zero prescribed displacements cannot vary from one load case to
another, or else you will not observe any performance benefit during the
solution.
- There are not any performance improvements for these cases: linear
static studies with load cases defined with no-penetration contacts, virtual wall
contacts, distributed beam loads which vary across load cases, loads on beam
vertices, and extension only or compression only springs.
- For linear static studies with multiple load cases, the performance
benefit comes from the factorization of the global stiffness matrix. The global
stiffness matrix factorization (which consumes the majority of the total solution
time) is done only once, as the stiffness matrix is exactly the same for every load
case.