A multi-block viscous flow solver based on GPU parallel methodology

A multi-block viscous flow solver for steady and unsteady turbulent flows based on GPU parallel methodology under the finite volume frame is presented in this paper. Both numerical accuracy and computational efficiency are concerned. Numerical flux scheme for all speeds SLAU is adopted because of its wide adaptability and strong robustness; high order reconstruction schemes like MLP and WENO are chosen to evaluate the inviscid terms while a set of fully conservative 4th-order central differencing schemes are utilized to deal with the viscous terms. Second-order temporal accuracy is forfeited for unsteady simulations by coupling DP-LUR with dual time-stepping strategy. Furthermore, heterogeneous multiple CPU + GPU coprocessing system is well established with CUDA and MPI methodology. Design details about GPU implementation are analyzed and discussed. Impressive speedup factor is achieved on our GPU platform compared with CPU indicating the bright feature of these algorithms. Numerical results of several complex configurations have demonstrated the validity and reliability for aerospace engineering applications.

A family of high-order targeted ENO schemes for compressible-fluid simulations

Although classical WENO schemes have achieved great success and are widely accepted, they exhibit several shortcomings. They are too dissipative for direct simulations of turbulence and lack robustness when very-high-order versions are applied to complex flows. In this paper, we propose a family of high-order targeted ENO schemes which are applicable for compressible-fluid simulations involving a wide range of flow scales. In order to increase the numerical robustness as compared to very-high-order classical WENO schemes, the reconstruction dynamically assembles a set of low-order candidate stencils with incrementally increasing width. While discontinuities and small-scale fluctuations are efficiently separated, the numerical dissipation is significantly diminished by an ENO-like stencil selection, which either applies a candidate stencil with its original linear weight, or removes its contribution when it is crossed by a discontinuity. The background linear scheme is optimized under the constraint of preserving an approximate dispersion–dissipation relation. By means of quasi-linear analyses and practical numerical experiments, a set of case-independent parameters is determined. The general formulation of arbitrarily high-order schemes is presented in a straightforward way. A variety of benchmark-test problems, including broadband waves, strong shock and contact discontinuities are studied. Compared to well-established classical WENO schemes, the present schemes exhibit significantly improved robustness, low numerical dissipation and sharp discontinuity capturing. They are particularly suitable for DNS and LES of shock–turbulence interactions.

Happy new year !