For integral viscoelastic calculations, the flow pathlines are computed by integrating the velocity field. In order to avoid tracking problems (for example, "out of mesh by a wall" error message, which causes a reduction of the S step), you must ensure that the mesh is adequately resolved (that is, not too coarse).

When solving an integral viscoelastic flow problem, the viscoelastic extra stresses are uncoupled from the pressure and velocity variables, and most of the CPU time (more than 95%) is spent tracking the deformation history at each Gauss point. An uncoupled scheme means that convergence cannot be quadratic, as is the case for Newton-Raphson schemes. Since tracking the past deformation history involves many element-configuration checks, integral viscoelastic flow problems will not vectorize. Performance on vector machines will therefore not be as good as that achieved for differential viscoelastic calculations. Performance on parallel machines will be similar to parallel performance for differential viscoelastic calculations.