Performance Analysis of Data Structures for Unstructured Grid Solutions on Finite Volume Simulations

Abstract. The efficient storage and access of grids by data structures using topological operators is one of the most important goals of the geometric modeling research field. It usually requires less time to load grid files and allows efficient query operations. Topological data structures index elements of grid representing incidence and adjacency properties. One of most common aspects from problems solved by computational fluid mechanics is the complexity of the domain’s geometry. The usage of data structures to index computational grids is of great importance because it performs queries efficiently and places all grid operations on a unique module, allowing its extension to aid diverse applications. This work aims at exploring the coupling of a topological data structure, the Mate Face, into a solver module, by controlling all grid access by providing operators and iterators that perform complex neighbor queries. The solver module solves the governing equations from fluid mechanics though the finite volume technique with a formulation that sets the property values to the control volume centroids, using high order methods the ENO and WENO schemes. The two dimensional Euler equations are considered to represent the flows of interest. The coupling of the Mate Face data structure to the solver module was achieved by a creation of a library that acts as an interface layer between modules, the Mate Face and the solver, which had been implemented using different programming languages. Therefore, all Mate Face class methods are available to the solver module though the interface library in the form of procedures. Experiments measuring the computational cost of using the Mate Face data structure comparing with the traditional tables representation of grids show that the Mate Face is faster to build and presents less memory occupation rate when indexing grids with a higher number of elements. Moreover, experiments were performed to validate the usage of the Mate Face along with numeric methods. Finally, the data structure can aid the fluid flow simulation tools by managing the grid file and providing efficient query operators.