TPSAS-NF1676L-31456-DND
Explore the source record for details and available documents.
Cara A.C. Leckey↗
Engineering topics
Eric J. Nielsen
Publications and source records attributed to Eric J. Nielsen.
Performance and Portability of a Linear Solver Across Emerging Architectures
A linear solver algorithm used by a large-scale unstructured-grid computational fluid dynamics application is examined for a broad range of familiar and emerging architectures. Efficient implementation of a linear solver is challenging on recent CPUs offering vector architectures. Vector loads and stores are essential to effectively utilize available memory bandwidth on CPUs, and maintaining performance across different CPUs can be difficult in the face of varying vector lengths offered by each. A similar challenge occurs on GPU architectures, where it is essential to have coalesced memory accesses to utilize memory bandwidth effectively. In this work, we demonstrate that restructuring a computation, and possibly data layout, with regard to architecture is essential to achieve optimal performance by establishing a performance benchmark for each target architecture in a low level language such as vector intrinsics or CUDA. In doing so, we demonstrate how a linear solver kernel can be mapped to Intel® Xeon™ and Xeon Phi™, Marvell® ThunderX2®, NEC® SX-Aurora™ TSUBASA Vector Engine, and NVIDIA® and AMD® GPUs. We further demonstrate that the required code restructuring can be achieved in higher level programming environments such as OpenACC, OCCA, and Intel® OneAPI™/SYCL, and that each generally results in optimal performance on the target architecture. Relative performance metrics for all implementations are shown, and subjective ratings for ease of implementation and optimization are suggested.
Programming models↗
Improvements in Iterative Convergence of FUN3D Solutions
This paper presents a hierarchical adaptive nonlinear iteration method (HANIM) implemented in the NASA computational fluid dynamics code, FUN3D, to improve robustness and computational efficiency. In contrast to the legacy FUN3D iterative solver that relies on an approximate Jacobian, a simple multicolor Gauss-Seidel point-implicit iteration scheme, and linear Courant-Friedrichs-Lewy number (CFL) ramping, HANIM is based upon a hierarchy of modules including preconditioner, generalized conjugate residual, realizability check, nonlinear control, and CFL adaption modules. HANIM performance is systematically compared with the performance of the legacy solver of FUN3D and a baseline solver based on a preconditioner alone. Iterative solutions are compared for three benchmark cases: a subsonic separated flow around a hemisphere cylinder, a supersonic flow through a long duct, and a subsonic flow over the NASA wing-fuselage juncture model. Two Reynolds-averaged Navier-Stokes turbulence models are used in these computations, namely, the negative variant of the linear one-equation Spalart-Allmaras model and its nonlinear extension based on quadratic constitutive relations.
CFD↗
The Influence of Computer Architecture on Performance and Scaling for Hypersonic Flow Simulations
It is critical to understand how hypersonic simulation tools perform on a range of computational platforms. This information will aid in the acquisition of appropriate hardware and the potential refactoring of hypersonic codes to run on different systems. In this paper, we consider two representative high-speed reacting flow cases: a model Mach 8 hypersonic waverider glide vehicle and a model hydrocarbon-fueled hypersonic ramjet propulsion system. In both scenarios, the flow fields are in chemical non-equilibrium and are modeled by the multi-species reacting Navier-Stokes equations. For these simulations we use several hypersonic simulation tools, including US3D, Kestrel, FUN3D, and the JENRER© flow solver. We explore several high performance computing systems containing IntelR© XeonR© Platinum processors, AMD EPYCTM7702 processors, and NVIDIAR© Tesla V100 devices. We compare performance and strong scaling between the different systems.
CPU↗
Wall-Modeled Large-Eddy Simulations for High-Lift Configurations Using FUN3D
Explore the source record for details and available documents.
Wall model↗
Edge-Based Viscous Method for Mixed-Element Node-Centered Finite-Volume Solvers
A novel, efficient, edge-based viscous (EBV) discretization method has been recently developed, implemented in a practical, unstructured-grid, node-centered, finite-volume flow solver, and applied to viscous-kernel computations that include evaluations of meanflow viscous fluxes, turbulence-model and chemistry-model diffusion terms, and the corresponding Jacobian contributions. Initially, the EBV method had been implemented for tetrahedral grids and demonstrated multifold acceleration of all viscous-kernel computations. This paper presents an extension of the EBV method for mixed-element grids. In addition to the primal edges of a given mixed-element grid, virtual edges are introduced to connect cell nodes that are not connected by a primal edge. The EBV method uses an efficient loop over all (primal and virtual) edges and features a compact discretization stencil based on the nearest neighbors. This study verifies the EBV method and assesses its efficiency on mixed-element grids by comparing the EBV solution accuracy and iterative convergence with those of well-established solutions obtained using a cell-based viscous (CBV) discretization method. The EBV solver’s memory footprint is optimized and often smaller than the memory footprint of the CBV solver. A multifold speedup is demonstrated for all viscous-kernel computations resulting in significant reduction of the time to solutions for several benchmark mixed-element-grid computations, including simulations of a flow around NASA’s juncture-flow model and a hypersonic, chemically reacting flow around a blunt body.
CFD↗
Exergy-based Sensitivity Analysis of the Generic Hypersonic Vehicle using FUN3D
In this paper, the implementation of an exergy-based objective function and its adjoint gradient into NASA’s FUN3D solver is discussed and verified. In order to verify that the exergy-based functional is properly implemented, it is used to predict the drag of the Generic Hypersonic Vehicle (GHV), which is then compared to more traditional force-based drag predictions. In addition to the functional implementation, FUN3D’s adjoint capability was extended to obtain sensitivities. Results were verified using FUN3D’s native complex step method for di↵erentiation using a generic wing configuration. The complex and adjoint gradients yielded discrete agreement demonstrating correct implementation and that the functional can be used for gradient-based multidisciplinary analysis and optimization. Next, various trade studies are conducted on the GHV to understand the design space of the vehicle. Finally, an inverse design problem is solved to verify the utilized design optimization framework which is ready to be deployed for exergy-based optimizations in future work.
Neal L. Novotny↗
Assessment of Edge-Based Viscous Method for Corner-Flow Solutions on Graphics Processing Units
A highly efficient, edge-based viscous (EBV) discretization method has been recently implemented in a practical, unstructured-grid, node-centered, finite-volume flow solver and evaluated for Reynolds-averaged Navier-Stokes (RANS) formulations. In comparison to a well-established cell-based viscous (CBV) method, the EBV method has demonstrated multifold acceleration of all viscous-kernel computations on general unstructured mixed-element grids. The viscous kernels include evaluation of viscous fluxes, diffusion terms in turbulence models, and the corresponding Jacobian terms. In this paper, an EBV implementation of a nonlinear extension of the Spalart-Allmaras turbulence model, SA-neg-QCR2000, is presented and verified. The SA-neg-QCR2000 model is used for simulating turbulent corner flows. Previously reported EBV computations have been conducted on traditional computing architectures based on central processing units (CPU). This paper assesses benefits of the EBV method on modern high-performance computing architectures based on graphics processing units (GPU). The GPU implementations of the CBV and EBV methods are verified by comparing solutions and iterative convergence with those observed in CPU computations on the same grids. A comprehensive assessment of the EBV speedup on CPU and GPU architectures is presented for established benchmark corner flows, namely, a supersonic flow through a long square duct and a subsonic flow around a NASA juncture flow model.
CFD↗
Assessment of UMUSCL Scheme for DNS of Turbulent Flows
Direct numerical simulation (DNS) is performed using an unstructured node-based finite-volume methodology for canonical turbulent flow configurations to assess the effect of the upwinding parameter (κ) in the unstructured monotonic upstream scheme for conservation laws (UMUSCL) reconstruction scheme. The NASA FUN3D solver is used, and four turbulent validation cases are considered: periodic Taylor-Green vortex, (2) flow through a plane channel, (3) flow through a channel with a constriction, and (4) flow over a flat plate. The results are presented for κ = 0.5, 0.9 and 1 on hexahedral grids, with the last value corresponding to a zero-numerical dissipation scheme. Detailed comparisons with available reference data indicate that for the smallest length-scales simulated, κ = 0.5 is too dissipative while κ = 1 could lead to an energy build-up. Overall, κ =0.9 yields satisfactory results in terms of accuracy and robustness for all the cases considered in this study.
Direct Numerical Simulation↗
Closed-Loop Simulations of Human-Scale Mars Lander Descent Trajectories on Frontier
A computational campaign was performed to run high-fidelity, free-flight simulations of a human-scale Mars lander concept vehicle decelerating under retropropulsion through the Martian atmosphere with closed-loop flight control. A novel approach is used to couple computational fluid dynamics (CFD) software with a mature flight mechanics package, where the two applications communicate in real-time across two geographically-dispersed computational facilities. The CFD is performed on the Frontier exascale system located at Oak Ridge National Laboratory, and the flight mechanics are executed on a system located at NASA Langley Research Center. In the current campaign, CFD is performed using finite-rate chemistry to account for the interactions between the LOXCH 4 engines and the CO 2 Martian atmosphere. A simulation of a closed-loop main engine throttling and RCS actuation is presented, demonstrating that the vehicle and model are able to maintain stability in a long-duration CFD-in-the-loop flight simulation. Comparisons are made to a reduced order model ignoring aero-propulsive interactions.
CFD↗