Search for: All records

Interactions between an evolving solid and inviscid flow can result in substantial computational complexity, particularly in circumstances involving varied boundary conditions between the solid and fluid phases. Examples of such interactions include melting, sublimation, and deflagration, all of which exhibit bidirectional coupling, mass/heat transfer, and topological change of the solid–fluid interface. The diffuse interface method is a powerful technique that has been used to describe a wide range of solid-phase interface-driven phenomena. The implicit treatment of the interface eliminates the need for cumbersome interface tracking, and advances in adaptive mesh refinement have provided a way to sufficiently resolve diffuse interfaces without excessive computational cost. However, the general scale-invariant coupling of these techniques to flow solvers has been relatively unexplored. In this work, a robust method is presented for treating diffuse solid–fluid interfaces with arbitrary boundary conditions. Source terms defined over the diffuse region mimic boundary conditions at the solid–fluid interface, and it is demonstrated that the diffuse length scale has no adverse effects. To show the efficacy of the method, a one-dimensional implementation is introduced and tested for three types of boundaries: mass flux through the boundary, a moving boundary, and passive interaction of the boundary with an incident acoustic wave. Two-dimensional results are presented as well these demonstrate expected behavior in all cases. Convergence analysis is also performed and compared against the sharp-interface solution, and linear convergence is observed. This method lays the groundwork for the extension to viscous flow and the solution of problems involving time-varying mass-flux boundaries.

 

Stencil computations are widely used in the scientific simulation domain, and their performance is critical to the overall efficiency of many large-scale numerical applications. Many optimization techniques, most of them varying strategies of tiling and parallelization, exist to systematically enhance the efficiency of stencil computations. However, the effective- ness of these optimizations vary significantly depending on the wide range of properties demonstrated by the different stencils. This paper studies several well-known optimization strategies for stencils and presents a new approach to effectively guide the composition of these optimizations, by modeling their interactions with four domain-level proper- ties of stencils: spatial dimensionality, temporal order, order of accuracy, and directional dependence. When using our prediction model to guide optimizations for five real-world stencil applications, we were able to identify optimization strategies that outperformed two highly optimized stencil libraries by an average of 2.4x.