MGLET-base flow solver

MGLET-base is an incompressible Navier-Stokes flow solver written in Fortran, provided by KM Turbulenz GmbH.

Relationship to commercial MGLET

The present MGLET-base is a subset of the commercial product MGLET developed and sold by KM Turbulenz GmbH since 2005. This open source code base was established in 2023 to provide a framework for contributions and interactions with academic user communities. The present MGLET-base contains the following features:

• Incompressible flow solver
• Ghost cell immersed boundary method for DNS and low-Re wall-resolved LES of stationary geometries

The commercial MGLET contains the following additional features and advantages:

• Unique cut-cell immersed boundary method for high-Re and wall-modeled simulations, with the possibility of doing moving geometries
• Aeroacoustic solver capable of predicting flow noise, with a huge variety of options on boundary conditions, loudspeakers, damping materials, acoustic feedback and various acoustic evaluation tools
• Structural solver for acoustic-fluid-structure-interaction
• Advanced and efficient grid-generator with many refinement options such as refinement boxes, surfaces and volumes, giving large flexibility in grid shapes and setups. With this tool, grids are generated in seconds.
• A comprehensive set of integrated run-time sampling and postprocessing options, including surface loads, wall-shear stresses, Fourier transforms of 3-D volumetric fields, sound sources and more
• Postprocessing options with integration and output to Paraview, generation of volumetric- and surface-fields, automatic image generation, spectra generation and plotting
• A comprehensive report tool to generate automatic reports from a simulation
• Workflow management solutions that streamline the configuration, execution and post-processing of a simulation
• A finished distribution package ready to use on any HPC system including comprehensive documentation
• User training and support

If you need a turnkey solution for flow simulations, KM Turbulenz GmbH provides MGLET as a ready-to-use commercial flow and aeroacoustic solver including everything listed above and more. Please contact us for a discussion about the possibilities MGLET will give you.

Purpose of MGLET-base

The present code is a new and modern re-implementation of the "traditional" and well-known MGLET numerics that date back to the 1980's. The code is re-implemented Fortran 2018 making use of modern programming techniques to improve flexibility and expandability.

The purpose of the MGLET-base is to provide high-quality building blocks for academic work in flow simulations. The code is built to be easily expandable for specific projects and needs. Implementing new physical models for any transport phenomena should be easy. Although the present code compiles and runs certain basic flow cases, it is not intended as the final tool for any specific task or need. Instead, it serves as a starting point for further developments of specific features.

Brief history

• 1980's: Initial version developed by Heinrich Werner within the scope of his dissertation "Grobstruktursimulation der turbulenten Strömung über eine querliegende Rippe in einem Plattenkanal bei hoher Reynoldszahl". Performing high-Reynolds number large-eddy simulations of turbulent channel flows with obstacles on Cray X-MP and Y-MP vector computers.

• 1990's: Michael Manhart introduced message passing with the MPI library and the multi-grid approach, including local grid refinement. Since this time the code was named Multi-Grid-Large-Eddy-Turbulence.

• 2000's: Frederic Tremblay, Nikolaus Peller and Johannes Kreuzinger: Ghost-cell immersed boundary implementation.

• 2005: KM Turbulenz GmbH is founded and acquired necessary rights to the MGLET code from TU Munich.

• 2005 -> present: KM Turbulenz develops MGLET into a commercial state-of-the art flow and aeroacoustics solver

• 2023: MGLET-base open source core released

Numerical properties

MGLET uses staggered, Cartesian grids organized in a overlapping, hierarchical multi-level pattern allowing local grid refinement.

The staggered arrangement of the flow variables allows for energy conserving spatial schemes. Coupled with explicit Runge-Kutta time integration methods it is tailored for DNS and LES simulations of flows at a wide range of Reynolds numbers.

Performance and scaling

MGLET's handling of many independent Cartesian grids is very efficient, and simulations have been conducted with over 250 000 grids of size 24x24x24 - that is nearly 3.5 billion grid cells. It has also been proven to run on tens of thousands of MPI ranks in big supercomputers throughout Europe.

Building MGLET-base

The most important build requirement is a set with C, C++ and Fortran compiler with Fortran 2008 + TS 29113 support. Then you will need an MPI library and a HDF5 library with MPI support.

The code is currently tested with GNU compilers version >= 11.2 and Intel compilers from the oneAPI toolkits version >= 2022.2.1. Older compilers than these will typically not work.

Since version 7.2 build 7213, the NAG Fortran compiler also compiles a working binary that runs all testcases successfully. This is only tested on x64 Linux systems.

The LLVM flang compiler (together with clang and clang++), built from Github sources, also works. Released versions prior to version 20 (expected released in spring 2025) will probably not work.

Dependencies

• An MPI implementation of your choice that provides the MPI_f08 Fortran bindings. Intel MPI, OpenMPI and MPICH >= 4.2.2 are known to work well.
• HDF5: https://github.com/HDFGroup/hdf5
• CMake: https://cmake.org/download/

The following dependencies are fetched and built automatically by CMake:

• Nlohman JSON: https://github.com/nlohmann/json
• Exprtk: https://www.partow.net/programming/exprtk/

Build instructions

MGLET-base make use of CMake presets to store a default set of build settings for the most common environments. There are currently pre-defined presets for the following compilers/toolchains:

• gnu-debug and gnu-release: GNU compilers gcc, g++ and gfortran.
• intel-debug and intel-release: Intel compilers icx, icpx and ifx.
• nag-debug and nag-release: NAG Fortran compiler nagfor with GNU C and C++ compilers gcc and g++. The debug configuration nag-debug is very effective of discovering errors that the Intel and GNU compilers cannot detect.
• llvm-debug and llvm-release: LLVM compilers clang, clang++ and flang (formerly known as flang-new).
• cray-debug and cray-release: Cray compilers cc, CC and ftn.

In order to build MGLET there are a few simple steps to follow:

1. Check out the source code

2. Create a separate build directory in the source code root and enter that directory: mkdir build && cd build

3. Run CMake: cmake --preset=gnu-release .. (replace the preset with your desired one)

4. Compile: make

5. Run tests: ctest --output-on-failure --test-dir tests

License and trademarks

MGLET is a registered trademark of Kreuzinger und Manhart Turbulenz GmbH.

The present code is licensed under the Apache License, Version 2.0 (the "License"). A copy of the license is contained in this repository or at:

http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.