NEMoSys

The Nuclear Energy Modeling System is a modular, extensible resource designed for use in typical application development systems as well as distributed web-services environments. The project focus is providing a framework for robust, automated mesh generation, mesh quality analysis, adaptive mesh refinement, and data transfer between arbitrary meshes. Python bindings to the NEMoSys library can also be enabled.

Getting Started

To acquire NEMoSys, you can download it from Illinois Rocstar's GitHub or clone it with the following command:

$ git clone git@git.illinois.rocstar:Nemosys/Nemosys.git

Build Options

The following table contains all available CMake options to configure NEMoSys functionality. The necessary third-party library is listed in the notes section.

Option name	Option description	Default	Notes
ENABLE_MPI	Enable MPI support	OFF	Requires MPI compiler
ENABLE_TESTING	Enable testing	ON
ENABLE_BUILD_UTILS	Build utilities	OFF
ENABLE_PYTHON_BINDINGS	Enable Python bindings	OFF	Requires Python 3 and SWIG 3
ENABLE_CFMSH	Enable cfMesh Meshing engine	OFF	Requires OpenFOAM
ENABLE_CGNS	Enable CGNS extensions	OFF	Requires CGNS
ENABLE_CONSRV_SURFACE_TRANSFER	Enable conservative surface transfer	OFF	Requires IMPACT
ENABLE_CONSRV_VOLUME_TRANSFER	Enable conservative volume transfer	OFF	Requires MPI
ENABLE_EPIC	Enable EPIC preprocessor	OFF
ENABLE_HDF5	Enable HDF5 extensions	OFF	Requires HDF5
ENABLE_METIS	Enable Metis partitioner	ON	Requires METIS
ENABLE_NETGEN	Enable Netgen meshing engine	ON	Requires Netgen
ENABLE_OMEGAH_CUDA	Enable GPU for Omega_h	OFF	Requires Kokkos
ENABLE_OPENCASCADE	Enable OpenCASACADE support	ON	Requires OpenCASCADE (OCCT)
ENABLE_TEMPLATE_MESH	Enable meshing templates	ON
ENABLE_MLAMR	Enable machine learning based AMR	OFF	Requires Frugally-deep library

Enabling cfMesh

cfMesh is an open-source meshing engine implemented on top of OpenFOAM. NEMoSys comes with a fully integrated cfMesh-based meshing module. To enable the capability, the NEMoSys should be compiled with ENABLE_CFMSH=ON.

Note: cfMesh depends on OpenFOAM, so before starting the compile process make sure to load OpenFOAM environment variables. Depending on the version, OpenFOAM can be loaded by sourcing the bashrc, or cshrc scripts provided in the OpenFoam-x.y/etc/. Refer to the OpenFOAM documentation for further instructions. After OpenFOAM environment is loaded, enable cfMesh build by adding this line to the cmake command:

-DENABLE_CFMSH=ON

Enabling machine learning based AMR

NEMoSys adaptive mesh refinement module for CFD is now equipped with machine learning support. This module allows users to use trained machine learning models for adaptive mesh refinement. To enable this capability, the NEMoSys should be compiled with ENABLE_MLAMR=ON.

User will also need a header only library frugally-deep, which loads python trained ML models in C++. User will need to provide installation path for this library using -DCMAKE_PREFIX_PATH flag along with other dependecies of NEMoSys. frugally-deep library also requires Tensorflow 2.1.0 installed.

-DENABLE_MLAMR=ON
-DCMAKE_PREFIX_PATH="\
${NEMOSYS_DEPS_INSTALL_PATH}/opencascade;\
${NEMOSYS_DEPS_INSTALL_PATH}/gmsh;\
${NEMOSYS_DEPS_INSTALL_PATH}/vtk;\
${NEMOSYS_DEPS_INSTALL_PATH}/netgen; \
/Install/path/to/frugally-deep" \

Enabling CUDA for Omega_h

Omega_h can be built with CUDA support using the Kokkos backend, assuming Kokkos is built with CUDA support. Currently, only Kokkos version 2 is supported. To enable this, make sure that the following flag is set:

-DENABLE_OMEGAH_CUDA=ON

and that CMAKE_PREFIX_PATH (or $PATH) contains ${NEMOSYS_DEPS_INSTALL_PATH}/kokkos/lib/CMake. Note that both Kokkos and NEMoSys must be built as shared libraries (-DBUILD_SHARED_LIBS=ON).

Unix Building Instructions

Build Dependencies

You will need to apt install at least the following dependencies:

build-essential
cmake
gfortran
libopenmpi-dev (or other MPI compiler)
zlib1g-dev
libfreetype6-dev
libfltk1.3-dev
libxmu-dev
libxi-dev
libhdf5-dev
liblapack-dev
libjpeg-dev
libcgns-dev
libmetis-dev
libexodusii-dev

Optional dependencies for additional functionality:

libpython3-dev
swig3.0

Note

We no longer maintain the build script. The file and the archive of TPLs in contribs folder are now deprecated. Instead, we publish the latest tested versions of TPLs used in the project. The list includes following items:

Gmsh 4.5.1
Netgen v6.2.1905 (commit hash 099acc9)
OpenCASCADE 7.3.0
VTK 8.2.0
Boost 1.68.0
OpenFoam version 4, 5, 6, or 7
Kokkos 2

Some of the TPLs need to be compiled in specific configurations. Directions for such TPLs are provided in the Manually Build Third Party Libraries section.

Build NEMoSys

Now, we can compile the NEMoSys library, create its Python bindings, and build other utilities:

$ NEMOSYS_INSTALL_PATH=/full/path/to/Nemosys/install
$ cd ${NEMOSYS_PROJECT_PATH}
$ mkdir build && cd build
$ cmake .. \
        -DCMAKE_PREFIX_PATH="\
${NEMOSYS_DEPS_INSTALL_PATH}/opencascade;\
${NEMOSYS_DEPS_INSTALL_PATH}/gmsh;\
${NEMOSYS_DEPS_INSTALL_PATH}/vtk;\
${NEMOSYS_DEPS_INSTALL_PATH}/netgen" \
        -DCMAKE_INSTALL_PREFIX=${NEMOSYS_INSTALL_PATH} \
        -DENABLE_BUILD_UTILS=ON \
        -DENABLE_TESTING=ON \
        -DBUILD_SHARED_LIBS=ON \
        -DENABLE_PYTHON_BINDINGS=ON \
        -DCMAKE_BUILD_TYPE=Release 
$ make -j$(nproc) # (or however many threads you'd like to use)
$ make install # (sudo if install location requires it)

Executing the commands above will build all libraries, executables, and bindings. The libraries are installed in $NEMOSYS_INSTALL_PATH/lib. Executables are installed in $NEMOSYS_INSTALL_PATH/bin. If Python bindings are enabled, the pyNemosys module is installed for the user. The pyNemosys module can be imported in Python as import pyNemosys. The build configuration can be modified through the CMake Curses interface, ccmake, or by passing the command line options to cmake.

Manually Build Third Party Libraries

Throughout this section, we assume NEMOSYS_DEPS_INSTALL_PATH is pointing to the location of the installation of the TPLs.

Building OpenCASCADE

Checkout proper version of OpenCASCADE, build and install the project by running the following commands:

$ mkdir build
$ cd build
$ cmake .. \
        -DCMAKE_INSTALL_PREFIX=${NEMOSYS_DEPS_INSTALL_PATH}/opencascade \
        -DBUILD_DOC_Overview=OFF \
        -DBUILD_SHARED_LIBS=ON
$ make -j$(nproc)
$ make install

Building Gmsh

Gmsh depends on OpenCASCADE. Checkout proper version of Gmsh, build and install by running the following commands:

$ mkdir build
$ cd build
$ cmake .. \
        -DCMAKE_INSTALL_PREFIX=${NEMOSYS_DEPS_INSTALL_PATH}/gmsh \
        -DCMAKE_PREFIX_PATH=${NEMOSYS_DEPS_INSTALL_PATH}/opencascade \
        -DENABLE_BUILD_LIB=OFF -DENABLE_BUILD_SHARED=ON -DENABLE_PRIVATE_API=ON \
        -DDEFAULT=ON -DENABLE_CGNS=OFF -DENABLE_NETGEN=OFF -DENABLE_HXT=ON \
        -DENABLE_FLTK=ON -DENABLE_BUILD_DYNAMIC=ON \
        -DENABLE_OPENMP=ON
$ make -j$(nproc)
$ make install -j$(nproc)

Building VTK

Build and install VTK by running the following commands:

$ mkdir build
$ cd build
$ cmake .. \
        -DCMAKE_INSTALL_PREFIX=${NEMOSYS_DEPS_INSTALL_PATH}/vtk \
        -DVTK_USE_SYSTEM_HDF5=ON \
        -DVTK_USE_SYSTEM_NETCDF=ON
$ make -j$(nproc)
$ make install

Building Netgen

Build and install Netgen by running the following commands:

$ mkdir build && cd build
$ cmake .. -DCMAKE_INSTALL_PREFIX=${NEMOSYS_DEPS_INSTALL_PATH}/netgen \
        -DUSE_GUI=OFF -DUSE_PYTHON=OFF -DUSE_SUPERBUILD=OFF -DUSE_OCC=ON \
        -DOCC_INCLUDE_DIR=${NEMOSYS_DEPS_INSTALL_PATH}/opencascade/include/opencascade \
        -DOCC_LIBRARY=${NEMOSYS_DEPS_INSTALL_PATH}/opencascade/lib/libTKernel.so
$ make -j$(nproc)
$ make install

Building Kokkos

Build and install Kokkos version 2 with CUDA backend using the following:

$ cd build
$ cmake ${KOKKOS_SRC} \
$       -DCMAKE_CXX_COMPILER=${KOKKOS_SRC}/bin/nvcc_wrapper \
$       -DCMAKE_INSTALL_PREFIX=${NEMOSYS_DEPS_INSTALL_PATH}/kokkos \
$       -DCMAKE_POSITION_INDEPENDENT_CODE=ON \
$       -DCKOKKOS_ARCH=${CUDA_ARCH_CC} \
$       -DKOKKOS_ENABLE_CUDA=ON \
$       -DKOKKOS_ENABLE_CUDA_LAMBDA=ON \
$       -DBUILD_SHARED_LIBS=ON
$ make install

where ${KOKKOS_SRC} is the source directory and ${CUDA_ARCH_CC} refers to the architecture and compute capability of your GPU, from the list Kepler30 Kepler32 Kepler35 Kepler37 Maxwell50 Maxwell52 Maxwell53 Pascal60 Pascal61 Volta70 Volta72.

Windows Build Instructions

The dependencies are similar to a UNIX build of NEMoSys with the addition of boost. An archive of pre-built Windows dependencies is available with some exceptions:

Note: When downloading pre-built libraries, ensure to select the version matching the bitness (32 or 64) and the MSVC compiler version (14.1, 14.0, etc.) used.

boost: https://sourceforge.net/projects/boost/files/boost-binaries/

netCDF (only if Exodus/Epic support is enabled): https://www.unidata.ucar.edu/downloads/netcdf/index.jsp

SWIG (only if Python bindings are enabled): http://swig.org/download.html

boost Dependency

boost will install by default to C:\local\boost_#_##_# where the # are the boost version. The boost location must be specified to CMake with the BOOST_ROOT variable.

netCDF Dependency

The netCDF installation is not detected automatically. Pass the location to the build system through the CMAKE_PREFIX_PATH variable.

Archived Dependencies

Extract the archive to a custom location (%TPL_DIR%). The dependencies must be given to CMake through the CMAKE_PREFIX_PATH` variable.

-DCMAKE_PREFIX_PATH="%TPL_DIR%\cgns;%TPL_DIR%\exodusii;%TPL_DIR%\gmsh;%TPL_DIR%\hdf5\cmake\hdf5;%TPL_DIR%\metis;%TPL_DIR%\netgen;%TPL_DIR%\vtk;%TPL_DIR%\zlib"

Exclude any you wish to use a custom installation.

Python on Windows

Python bindings are generated using SWIG. A SWIG installer is needed on the system and must be available to CMake and available to the Python environment.

This is accomplished in two parts: (1) add the SWIG_EXECUTABLE variable pointing to the SWIG executable (swig.exe) for CMake and (2) add the folder containing the SWIG executable to the system PATH for Python. A pre-built Windows executable can be downloaded from the SWIG website: http://swig.org/download.html

Note: On a system with multiple Python installations, the Cmake variable PYTHON_EXECUTABLE can be set to the python.exe executable to force the specific environment.

Build NEMoSys

With the dependencies specified above installed, we can compile NEMoSys with Python bindings and build tools with the following command from a MSVC command prompt:

> set NETCDF_LOCATION=C:\full\path\to\netCDF\install
> set BOOST_LOCATION=C:\full\path\to\boost\install
> set SWIG_EXE=C:\full\path\to\SWIG\executable\swig.exe
> set NEMOSYS_INSTALL_PATH=C:\full\path\to\Nemosys\install
> cd %NEMOSYS_PROJECT_PATH%
> md build && cd build
> cmake .. ^
        -G "Ninja" ^
        -DBOOST_ROOT="%BOOST_LOCATION%" ^
        -DSWIG_EXECUTABLE="%SWIG_EXE_LOCATION%" ^
        -DCMAKE_INSTALL_PREFIX="%NEMOSYS_INSTALL_PATH%" ^
        -DBUILD_SHARED_LIBS=ON ^
        -DENABLE_TESTING=ON ^
        -DENABLE_BUILD_UTILS=ON ^
        -DENABLE_PYTHON_BINDINGS=ON ^
        -DCMAKE_PREFIX_PATH="%TPL_DIR%\cgns;%TPL_DIR%\exodusii;%TPL_DIR%\gmsh;%TPL_DIR%\hdf5\cmake\hdf5;%TPL_DIR%\metis;%TPL_DIR%\netgen;%TPL_DIR%\vtk;%TPL_DIR%\zlib;%NETCDF_LOCATION%"
> ninja -j %NUMBER_OF_PROCESSORS% (or however many threads you'd like to use)
> ninja install

Executing the commands above will build all libraries, executables, and bindings. The libraries are installed in %NEMOSYS_INSTALL_PATH%\lib. Executables are installed in %NEMOSYS_INSTALL_PATH%\bin.

Testing NEMoSys

From the build directory, execute the following command to test the installation:

UNIX:
$ make test
WINDOWS:
> ninja test

This will execute several tests found in $NEMOSYS_PROJECT_PATH/testing.

Repository Installation

Note : This is for internal IR user only. You can install experimental built of the latest NEMoSys for Ubuntu 18 and CentOS7 using following directions.

Ubuntu 18.04 LTS

1- Install following packages/dependencies

sudo apt update
sudo apt install apt-transport-https ca-certificates curl software-properties-common wget

2- Add repository

curl -fsSL http://nemosys-repository.illinois.rocstar/nemosys-repository-pub.gpg | sudo apt-key add -
sudo add-apt-repository "deb http://nemosys-repository.illinois.rocstar/ bionic main"

3- Install OpenFoam

sudo sh -c "wget -O - https://dl.openfoam.org/gpg.key | apt-key add -"
sudo add-apt-repository http://dl.openfoam.org/ubuntu

4- Install Nemosys

sudo apt-get install nemosys

5- Test installation

source /opt/openfoam7/etc/bachrc
nemosysRun

you should see:

Usage: nemosysRun input.json

CentOS 7

1- Add NEMoSys Repository

echo "[NEMoSys]
name=NEMoSys
baseurl=http://nemosys-rpm-repository.illinois.rocstar
enabled=1
gpgcheck=0
" >> /etc/yum.repos.d/nemosys.repo

2- Install Nemosys

sudo yum install nemosys

3- Test installation

source /opt/openfoam7/etc/bachrc
nemosysRun

you should see:

Usage: nemosysRun input.json

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

README_IR.md

README_IR.md

NEMoSys

Getting Started

Build Options

Enabling cfMesh

Enabling machine learning based AMR

Enabling CUDA for Omega_h

Unix Building Instructions

Build Dependencies

Note

Build NEMoSys

Manually Build Third Party Libraries

Building OpenCASCADE

Building Gmsh

Building VTK

Building Netgen

Building Kokkos

Windows Build Instructions

boost Dependency

netCDF Dependency

Archived Dependencies

Python on Windows

Build NEMoSys

Testing NEMoSys

Repository Installation

Ubuntu 18.04 LTS

CentOS 7

Files

README_IR.md

Latest commit

History

README_IR.md

File metadata and controls

NEMoSys

Getting Started

Build Options

Enabling cfMesh

Enabling machine learning based AMR

Enabling CUDA for Omega_h

Unix Building Instructions

Build Dependencies

Note

Build NEMoSys

Manually Build Third Party Libraries

Building OpenCASCADE

Building Gmsh

Building VTK

Building Netgen

Building Kokkos

Windows Build Instructions

boost Dependency

netCDF Dependency

Archived Dependencies

Python on Windows

Build NEMoSys

Testing NEMoSys

Repository Installation

Ubuntu 18.04 LTS

CentOS 7