We present an updated version of the open-source Hypersonics Task-based Research (HTR) solver for hypersonic aerothermodynamics. The solver, whose first version was presented in Di Renzo et al. (2020), is designed for direct numerical simulation (DNS) of canonical hypersonic flows at high Reynolds numbers in which thermo-chemical effects induced by high temperatures are relevant. The solver relies on high-order spatial discretization on structured meshes and efficient time integrators for stiff systems within the Regent/Legion software stack, which makes the code highly portable and scalable in CPU and GPU-based supercomputers. The new version herein presented includes several optimizations and new tools for data analysis, along with novel user option for hybrid skew-symmetric/targeted essentially non-oscillatory numerics, to offer higher computational efficiency and lower numerical dissipation at moderate Mach numbers, inclusion of a new combustion mechanism for methane and oxygen, and new recycling–rescaling inlet boundary conditions targeted to the simulation of fully developed turbulent boundary layers. New version program summary: Program Title: Hypersonics Task-based Research solver CPC Library link to program files: http://dx.doi.org/10.17632/9zsxjtzfr7.2 Developer's repository link: https://github.com/stanfordhpccenter/HTR-solver.git Licensing provisions: BSD 2-clause Programming language: Regent, C++, and CUDA Journal Reference of previous version: Di Renzo, M., Fu, L., & Urzay, J. (2020). HTR solver: An open-source exascale-oriented task-based multi-GPU high-order code for hypersonic aerothermodynamics. Computer Physics Communications, 107262. Does the new version supersede the previous version?: Yes Reasons for the new version: Release of new features Summary of revisions: • New optional sixth-order hybrid scheme has been implemented (activated by the flag “hybridScheme” in the input file). The scheme combines the energy-preserving properties of a sixth-order skew-symmetric central difference scheme [1] in smooth flow regions with the shock-capturing properties of a sixth-order targeted essentially non-oscillatory (TENO) scheme at points where shocks are involved. The switch between the two schemes is controlled by a TENO sensor whose cutoff value is adapted based on the maximum value of a modified Ducros sensor [2] across the reconstruction stencil. If the flag “hybridScheme” is set to false, the numerical scheme will revert to the TENO6-A scheme released in the previous version of the solver [3]; • New recycling–rescaling inflow boundary conditions [4,5] for the simulation of turbulent compressible boundary layers are now available to the user; • Support for Legion tracing, which significantly improves the strong scalability of the solver, has been implemented; • A diagnostic tool to monitor the time evolution of the flow variables in a subvolume of the computational domain is now available; • A single-step chemistry mechanism for methane/oxygen combustion has been added to the mixtures handled by the HTR solver; • Sample scripts for strong scaling have been added to the “testcases” directory; • Unit test and regression test suites have been added to the repository; • The input file scheme has been modified in order to reduce verbosity and increase flexibility in specifying the boundary conditions and type of gas mixture; • Hyperbolic sine stretching functions have been made available to users during the grid generation process; • Computationally intensive tasks have been ported to C++ and CUDA in order to achieve higher efficiency on all hardware; • Several optimizations of the tasks body and mapper have been implemented in order to increase the computational efficiency and reduce the memory footprint. Nature of problem: This code solves the Navier–Stokes equations at hypersonic Mach numbers including finite-rate chemistry for dissociating air and multicomponent transport. The solver is designed for direct numerical simulations (DNS) of transitional and turbulent hypersonic turbulent flows under high-enthalpy conditions, and it accounts for thermochemical effects (vibrational excitation and chemical dissociation). Solution method: This code uses a low-dissipation sixth-order schemes for the spatial discretization of the conservation equations on Cartesian stretched meshes. Time advancement is carried out by either an explicit method if chemistry is slow, hence not introducing additional stiffness, or by an operator-splitting algorithm whereby chemical production rates are handled implicitly. Additional comments including restrictions and unusual features: The HTR solver builds on the runtime Legion [6,7] and is written in the programming language Regent [8,9] developed at Stanford University. Instructions for installation of the components are provided in the README file enclosed with the HTR solver and in the Legion repository [6]. References [1] S. Pirozzoli, Journal of Computational Physics 229 (2010) 7180–7190. https://doi.org/10.1016/j.jcp.2010.06.006. F. Ducros, F. Laporte, T. Soulères, V. Guinot, P. Moinat, B. Caruelle, Journal of Computational Physics 161 (2000) 114–139. https://doi.org/10.1006/jcph.2000.6492. M. Di Renzo, L. Fu, J. Urzay, Computer Physics Communications 255 (2020) 107262. https://doi.org/10.1016/j.cpc.2020.107262. T. S. Lund, X. Wu, K. D. Squires, Journal of Computational Physics 140 (1996) 233–258. https://doi.org/10.1006/jcph.1998.5882. S. Pirozzoli, M. Bernardini, F. Grasso, Journal of Fluid Mechanics 657 (2010) 361–393. https://doi.org/10.1017/S0022112010001710. Legion web page, 2020. URL:https://legion.stanford.edu. M. Bauer, S. Treichler, E. Slaughter, A. Aiken, Legion: Expressing locality and independence with logical regions, International Conference for High Performance Computing, Networking, Storage and Analysis, SC (2012), IEEE. Regent web page, 2020. URL:http://regent-lang.org. E. Slaughter, W. Lee, S. Treichler, M. Bauer, A. Aiken, SC ’15: Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis (2015) 1–12,. https://doi.org/10.1145/2807591.2807629.
HTR-1.2 solver: Hypersonic Task-based Research solver version 1.2
Di Renzo M.
Primo
;
2021-01-01
Abstract
We present an updated version of the open-source Hypersonics Task-based Research (HTR) solver for hypersonic aerothermodynamics. The solver, whose first version was presented in Di Renzo et al. (2020), is designed for direct numerical simulation (DNS) of canonical hypersonic flows at high Reynolds numbers in which thermo-chemical effects induced by high temperatures are relevant. The solver relies on high-order spatial discretization on structured meshes and efficient time integrators for stiff systems within the Regent/Legion software stack, which makes the code highly portable and scalable in CPU and GPU-based supercomputers. The new version herein presented includes several optimizations and new tools for data analysis, along with novel user option for hybrid skew-symmetric/targeted essentially non-oscillatory numerics, to offer higher computational efficiency and lower numerical dissipation at moderate Mach numbers, inclusion of a new combustion mechanism for methane and oxygen, and new recycling–rescaling inlet boundary conditions targeted to the simulation of fully developed turbulent boundary layers. New version program summary: Program Title: Hypersonics Task-based Research solver CPC Library link to program files: http://dx.doi.org/10.17632/9zsxjtzfr7.2 Developer's repository link: https://github.com/stanfordhpccenter/HTR-solver.git Licensing provisions: BSD 2-clause Programming language: Regent, C++, and CUDA Journal Reference of previous version: Di Renzo, M., Fu, L., & Urzay, J. (2020). HTR solver: An open-source exascale-oriented task-based multi-GPU high-order code for hypersonic aerothermodynamics. Computer Physics Communications, 107262. Does the new version supersede the previous version?: Yes Reasons for the new version: Release of new features Summary of revisions: • New optional sixth-order hybrid scheme has been implemented (activated by the flag “hybridScheme” in the input file). The scheme combines the energy-preserving properties of a sixth-order skew-symmetric central difference scheme [1] in smooth flow regions with the shock-capturing properties of a sixth-order targeted essentially non-oscillatory (TENO) scheme at points where shocks are involved. The switch between the two schemes is controlled by a TENO sensor whose cutoff value is adapted based on the maximum value of a modified Ducros sensor [2] across the reconstruction stencil. If the flag “hybridScheme” is set to false, the numerical scheme will revert to the TENO6-A scheme released in the previous version of the solver [3]; • New recycling–rescaling inflow boundary conditions [4,5] for the simulation of turbulent compressible boundary layers are now available to the user; • Support for Legion tracing, which significantly improves the strong scalability of the solver, has been implemented; • A diagnostic tool to monitor the time evolution of the flow variables in a subvolume of the computational domain is now available; • A single-step chemistry mechanism for methane/oxygen combustion has been added to the mixtures handled by the HTR solver; • Sample scripts for strong scaling have been added to the “testcases” directory; • Unit test and regression test suites have been added to the repository; • The input file scheme has been modified in order to reduce verbosity and increase flexibility in specifying the boundary conditions and type of gas mixture; • Hyperbolic sine stretching functions have been made available to users during the grid generation process; • Computationally intensive tasks have been ported to C++ and CUDA in order to achieve higher efficiency on all hardware; • Several optimizations of the tasks body and mapper have been implemented in order to increase the computational efficiency and reduce the memory footprint. Nature of problem: This code solves the Navier–Stokes equations at hypersonic Mach numbers including finite-rate chemistry for dissociating air and multicomponent transport. The solver is designed for direct numerical simulations (DNS) of transitional and turbulent hypersonic turbulent flows under high-enthalpy conditions, and it accounts for thermochemical effects (vibrational excitation and chemical dissociation). Solution method: This code uses a low-dissipation sixth-order schemes for the spatial discretization of the conservation equations on Cartesian stretched meshes. Time advancement is carried out by either an explicit method if chemistry is slow, hence not introducing additional stiffness, or by an operator-splitting algorithm whereby chemical production rates are handled implicitly. Additional comments including restrictions and unusual features: The HTR solver builds on the runtime Legion [6,7] and is written in the programming language Regent [8,9] developed at Stanford University. Instructions for installation of the components are provided in the README file enclosed with the HTR solver and in the Legion repository [6]. References [1] S. Pirozzoli, Journal of Computational Physics 229 (2010) 7180–7190. https://doi.org/10.1016/j.jcp.2010.06.006. F. Ducros, F. Laporte, T. Soulères, V. Guinot, P. Moinat, B. Caruelle, Journal of Computational Physics 161 (2000) 114–139. https://doi.org/10.1006/jcph.2000.6492. M. Di Renzo, L. Fu, J. Urzay, Computer Physics Communications 255 (2020) 107262. https://doi.org/10.1016/j.cpc.2020.107262. T. S. Lund, X. Wu, K. D. Squires, Journal of Computational Physics 140 (1996) 233–258. https://doi.org/10.1006/jcph.1998.5882. S. Pirozzoli, M. Bernardini, F. Grasso, Journal of Fluid Mechanics 657 (2010) 361–393. https://doi.org/10.1017/S0022112010001710. Legion web page, 2020. URL:https://legion.stanford.edu. M. Bauer, S. Treichler, E. Slaughter, A. Aiken, Legion: Expressing locality and independence with logical regions, International Conference for High Performance Computing, Networking, Storage and Analysis, SC (2012), IEEE. Regent web page, 2020. URL:http://regent-lang.org. E. Slaughter, W. Lee, S. Treichler, M. Bauer, A. Aiken, SC ’15: Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis (2015) 1–12,. https://doi.org/10.1145/2807591.2807629.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.