Bulletin of the American Physical Society
69th Annual Meeting of the APS Division of Fluid Dynamics
Volume 61, Number 20
Sunday–Tuesday, November 20–22, 2016; Portland, Oregon
Session E27: CFD: High Performance Computing |
Hide Abstracts |
Chair: Michael Dodd, University of Washington Room: E147-148 |
Sunday, November 20, 2016 5:37PM - 5:50PM |
E27.00001: PSH3D fast Poisson solver for petascale DNS Darren Adams, Michael Dodd, Antonino Ferrante Direct numerical simulation (DNS) of high Reynolds number, ${Re} \ge O(10^5)$, turbulent flows requires computational meshes $\ge O(10^{12})$ grid points, and, thus, the use of petascale supercomputers. DNS often requires the solution of a Helmholtz (or Poisson) equation for pressure, which constitutes the bottleneck of the solver. We have developed a parallel solver of the Helmholtz equation in 3D, PSH3D. The numerical method underlying PSH3D combines a parallel 2D Fast Fourier transform in two spatial directions, and a parallel linear solver in the third direction. For computational meshes up to $8192^{3}$ grid points, our numerical results show that PSH3D scales up to at least 262k cores of Cray {XT5} (Blue Waters). PSH3D has a peak performance 6$\times$ faster than 3D FFT-based methods when used with the `partial-global' optimization, and for a $8192^3$ mesh solves the Poisson equation in 1~sec using 128k cores. Also, we have verified that the use of PSH3D with the `partial-global' optimization in our DNS solver does not reduce the accuracy of the numerical solution of the incompressible Navier-Stokes equations. [Preview Abstract] |
Sunday, November 20, 2016 5:50PM - 6:03PM |
E27.00002: Towards mitigating Asynchronous Computing effects in largescale simulations Ankita Mittal, Sharath Girimaji Synchronization of processing elements (PEs) in massively parallel simulations has shown to significantly affect scalability of scientific applications. Relaxing this synchronization among PEs (asynchronous) conserves the stability condition but severely affects the accuracy reducing the average error to first-order regardless of the original scheme. At the present time, several approaches are under consideration to improve the order of asynchronous computations. In this work, we propose to modify the original governing equation to obtain a Proxy-Equation which when solved asynchronously recovers the order of accuracy of the original numerical scheme. Performing 1D simulations for the Advection Diffusion Equation, we observe that the wave speed and the viscosity must be increased in the vicinity of PE boundaries in order to counteract the effect of asynchrony. In addition to recovering accuracy, this method shows lower magnitudes of average error when compared to existing asynchrony–tolerant methods. Similar results are also presented for a 1D viscous Burgers equation. [Preview Abstract] |
Sunday, November 20, 2016 6:03PM - 6:16PM |
E27.00003: Unstable phenomena of low speed compressible natural convection with open boundaries by multi-GPU implementation. Wei-Hsiang Wang, Wu-Shung Fu, Makoto Tsubokura Unstable phenomena of low speed compressible natural convection are investigated numerically. Geometry contains parallel square plates or single heated plate with open boundaries is taken into consideration. Numerical methods of the Roe scheme, preconditioning and dual time stepping matching the DP-LUR method are used for low speed compressible flow. The absorbing boundary condition and modified LODI method is adopted to solve open boundary problems. High performance parallel computation is achieved by multi-GPU implementation with CUDA platform. The effects of natural convection by isothermal plates facing upwards in air is then carried out by the methods mentioned above Unstable behaviors appeared upon certain Rayleigh number with characteristic length respect to the width of plates or height between plates. [Preview Abstract] |
Sunday, November 20, 2016 6:16PM - 6:29PM |
E27.00004: CODE BLUE: Three dimensional massively-parallel simulation of multi-scale configurations Damir Juric, Lyes Kahouadji, Jalel Chergui, Seungwon Shin, Richard Craster, Omar Matar We present recent progress on BLUE, a solver for massively parallel simulations of fully three-dimensional multiphase flows which runs on a variety of computer architectures from laptops to supercomputers and on 131072 threads or more (limited only by the availability to us of more threads). The code is wholly written in Fortran 2003 and uses a domain decomposition strategy for parallelization with MPI. The fluid interface solver is based on a parallel implementation of a hybrid Front Tracking/Level Set method designed to handle highly deforming interfaces with complex topology changes. We developed parallel GMRES and multigrid iterative solvers suited to the linear systems arising from the implicit solution for the fluid velocities and pressure in the presence of strong density and viscosity discontinuities across fluid phases. Particular attention is drawn to the details and performance of the parallel Multigrid solver. [Preview Abstract] |
Sunday, November 20, 2016 6:29PM - 6:42PM |
E27.00005: Discrete kinetic and lattice Boltzmann formulations for reaction cross-diffusion systems and their hyperbolic extensions in chemotaxis Paul Dellar We present discrete kinetic and lattice Boltzmann formulations for reaction cross-diffusion systems, as commonly used to model microbiological chemotaxis and macroscopic predator-prey interactions, and their hyperbolic extensions with fluid-like persistence terms. For example, the canonical Patlak--Keller--Segal model for chemotaxis involves a flux of cells up the gradient of a chemical secreted by the cells, in addition to the usual down-gradient diffusive fluxes. Existing lattice Boltzmann approaches for such systems use finite difference approximations to compute the flux of cells due to the chemical gradient. The resulting coupling between, and necessary synchronisation of the evolution of, adjacent grid points greatly complicates boundary conditions, and efficient implementation on graphical processing units (GPUs). We present a kinetic formulation using cross-collisions between bases of moments for the two sets of distribution functions to couple the fluxes of the two species, from which we construct lattice Boltzmann algorithms using second-order Strang splitting. We demonstrate an efficient GPU implementation, and verify second-order spatial convergence towards spectral solutions for benchmark problems such as the finite-time blow-up in the Patlak--Keller--Segal model. [Preview Abstract] |
Sunday, November 20, 2016 6:42PM - 6:55PM |
E27.00006: LES of a bluff-body stabilized premixed flame using discontinuous Galerkin scheme Yu Lv, Matthias Ihme This talk focuses on the development of a high-order discontinuous Galerkin (DG) method for application to chemically reacting flows. To enable these simulations, several algorithmic aspects are addressed, including the time-integration of multi-step chemical reactions, the incorporation of detailed thermo-viscous transport properties, and the stabilization of high-order solution representation. This DG solver is applied in implicit LES of a turbulent bluff-body stabilized propane/air premixed flame. The simulation results for cold-flow and reacting conditions are reported and compared to experimental data. [Preview Abstract] |
Follow Us |
Engage
Become an APS Member |
My APS
Renew Membership |
Information for |
About APSThe American Physical Society (APS) is a non-profit membership organization working to advance the knowledge of physics. |
© 2024 American Physical Society
| All rights reserved | Terms of Use
| Contact Us
Headquarters
1 Physics Ellipse, College Park, MD 20740-3844
(301) 209-3200
Editorial Office
100 Motor Pkwy, Suite 110, Hauppauge, NY 11788
(631) 591-4000
Office of Public Affairs
529 14th St NW, Suite 1050, Washington, D.C. 20045-2001
(202) 662-8700