Initial Guesses for Sequences of Linear Systems in a GPU-Accelerated Incompressible Flow Solver

We consider several methods for generating initial guesses when iteratively solving sequences of linear systems, showing that they can be implemented efficiently in GPU-accelerated PDE solvers, specifically solvers incompressible flow. propose new guess based on stabilized polynomial extrapolation and compare them to the projection method Fischer [Comput. Methods Appl. Mech. Engrg., 163 (1998), pp. 193--204], are generally competitive with schemes despite requiring only half storage performing considerably less data movement communication. Our implementations these algorithms freely available as part \tt libParanumal collection flow solvers.