Conference Agenda

Overview and details of the sessions of this conference. Please select a date or location to show only sessions at that day or location. Please select a single session for detailed view (with abstracts and downloads if available).

Please note that all times are shown in the time zone of the conference. The current conference time is: 1st Dec 2022, 06:56:28am CET

 
Only Sessions at Location/Venue 
 
 
Session Overview
Session
MS 25: Software for high-order methods
Time:
Tuesday, 13/July/2021:
4:00pm - 6:00pm

Session Chair: Chris Cantwell
Session Chair: David Moxey
Virtual location: Zoom 6


Session Abstract

High-order methods have been demonstrated to provide advantageous numerical properties, achiev-

ing improved accuracy per degree of freedom over low-order methods, and can make more efficient

use of modern computer hardware. Interest in them has grown in recent years and they are already

enhancing cutting-edge scientific applications, particularly in the areas of fluid dynamics. However,

their complexity means they are more challenging to implement than traditional finite element

techniques, particularly in an efficient and maintainable way. This has previously hampered their

uptake by both academia and industry.

This minisymposium focuses on the latest developments in a growing ecosystem of software tools

that is overcoming this challenge and making high-order methods accessible to a broader commu-

nity of academic and industrial users. These tools include both application solvers and the pre-

and post-processing software, such as high-order mesh generation and visualisation, which are all

critical components for these methods to impact society across a broad range of scientific domains

and support translation of these techniques into industrial practice.


Show help for 'Increase or decrease the abstract text size'
Presentations
4:00pm - 4:30pm

An industry-relevant implicit LES via spectral/hp element methods

Gianmarco Mengaldo1, David Moxey2, Michael Turner3, Rodrigo C Moura4, Ayad Jassim5, Mark Taylor6, Joaquim Peiro3, Spencer J Sherwin3

1National University of Singapore, Singapore; 2University of Exeter, UK; 3Imperial College London, UK; 4Instituto Tecnologico de Aeronautica, Brazil; 5Hewlett Packard Enterprise, UK; 6London Computational Solutions, UK

We present the successful deployment of high-fidelity Large-Eddy Simulation (LES) technologies based on spectral/hp element methods to a real automotive car, namely the Elemental Rp1 model [1]. The simulation presents the common challenges of an industry-relevant simulation, namely high Reynolds number and complex geometry. To the best of the authors' knowledge, this simulation represents the first fifth-order accurate transient LES of an entire real car geometry. Moreover, this constitutes a key milestone towards considerably expanding the computational design envelope currently allowed in industry, where steady-state modelling remains the standard. In this talk, we highlight the key developments that were required to achieve the simulation, from mesh generation to improvements in solver and numerical technology.



4:30pm - 5:00pm

Cache Blocking Strategies Applied to Flux Reconstruction

Semih Akkurt1, Freddie Witherden2, Peter Vincent1

1Imperial College London, United Kingdom; 2Texas A&M University, College Station, TX, United States

On modern hardware architectures, the performance of Flux Reconstruction (FR) methods for tensor product elements can be limited by memory bandwidth. In general, these methods are implemented as a chain of distinct kernels. Often, a dataset which has just been written to main memory by a kernel is read back immediately by the next kernel. The accepted solution for such a redundant expenditure of memory bandwidth is kernel fusion. However, on a practical level kernel fusion requires that the source for all kernels be available, thus preventing calls to certain third-party library functions. Moreover, it can add substantial complexity to a codebase. An alternative to full kernel fusion is using cache blocking, which has become practically possible in recent years due to growth in the size of CPU L2 cache. In this approach, kernels remain distinct, and are executed one after another on small chunks of data that can fit in the cache, as opposed to on full datasets. These chunks of data stay in the cache and whenever a kernel requests access to data that is already in the cache, memory bandwidth is saved. In this talk, a variety of kernel grouping configurations will be presented for an FR based Navier-Stokes solver, alongside associated theoretical memory bandwidth savings, and actual achieved performance gains when implemented in the PyFR solver. A Taylor-Green Vortex test case is used as a benchmark, and the most performant strategy leads to a speedup of approximately 3x.



5:00pm - 5:30pm

Multirate timestepping for the incompressible Navier-Stokes equations in overlapping grids

Ketan Mittal1, Som Dutta2, Paul Fischer3

1Lawrence Livermore National Laboratory, 7000 East Avenue, Livermore, CA 94550; 2Mechanical & Aerospace Engineering, Utah State University, 4130 Old Main Hill, Logan, UT 84332; 3Computer Science, University of Illinois at Urbana-Champaign, 201 N. Goodwin Ave., Urbana, IL 61801

We develop a multirate timestepper for semi-implicit solutions of the unsteady incompressible Navier-Stokes equations (INSE) based on a recently-developed multidomain spectral element method. For incompressible flows, multirate timestepping (MTS) is particularly challenging because of the tight coupling implied by the incompressibility constraint, which manifests as an elliptic subproblem for the pressure at each timestep. The novelty of our approach stems from the development of a stable overlapping Schwarz method applied directly to the Navier-Stokes equations, rather than to the convective, viscous, and pressure substeps that are at the heart of most INSE solvers. This MTS approach is based on a predictor-corrector strategy that preserves the temporal convergence of the underlying semi-implicit timestepper, and scales to an arbitrarily high time-step ratio between overlapping subdomains. We present numerical results demonstrating that this approach accurately models complex turbulent flow phenomenon and improves computational efficiency in comparison to singlerate timestepping-based calculations

Performed under the auspices of the U.S. Department of Energy under Contract DE-AC52-07NA27344 (LLNL-ABS-806352)



5:30pm - 6:00pm

ViZiR 4: pixel-exact rendering visualization software for high-order meshes and solutions

Matthieu Maunoury, RĂ©mi Feuillet, Adrien Loseille

GAMMA Team, Inria Saclay, France

With the increasing use of high-order methods and high-order meshes, scientific visualization softwares need to adapt themselves to reliably render the associated meshes and numerical solutions. However, classic visualization softwares were originally developed to handle affine functions on degree one elements. Therefore, most of standard visualization softwares cannot be used to render high-order meshes or high-order solutions. We are developing a tool able to answer to this visualization challenge.

ViZiR 4 is a light, simple and interactive high-order meshes and solutions visualization software using OpenGL 4 graphic pipeline. The use of OpenGL Shading Language (GLSL) allows to perform pixel exact rendering of high order solutions on flat elements and almost pixel exact rendering on curved elements (high-order meshes). ViZiR 4 is the new version of ViZiR and therefore enables the representation of high order meshes (up to degree 4) and high order solutions (up to degree 10) with a pixel exact rendering. Furthermore, the OpenGL 4 version of ViZiR is way more faster than the old version based on OpenGL legacy thanks to the GPU rendering used based on shaders. Many post-processing tools, such as picking, hidding surfaces, isolines, clipping, capping, are integrated to enable on the fly the analysis of the numerical results.

In this talk, this novel approach to render high-order solution and meshes is presented and several examples are shown to illustrate the potential of ViZiR 4 in various applications.



 
Contact and Legal Notice · Contact Address:
Privacy Statement · Conference: ICOSAHOM2020
Conference Software - ConfTool Pro 2.6.145+CC
© 2001–2022 by Dr. H. Weinreich, Hamburg, Germany