Quickstart

Dependencies

• HDF5 (1.8+)
• A C++ compiler compatible with C++17
• CMake (3.16+)

Cloning the repo

To clone FV2D, you need to clone the repository and the submodules. To do so from scratch:

git clone --recurse-submodules git@github.com:mdelorme/fv2d.git

If you have already cloned the repo but forgot to add the --recurse-submodules, you can still get the submodules by typing at the root of fv2d the following commands:

git submodule init
git submodule update

Building

Building relies on CMake. As FV2D is built with Kokkos included as a submodule, all Kokkos option have to be passed here. For more information on how to indicate which architecture/compiler you wish to use with Kokkos, please refer to the Kokkos documentation. Here are a few examples of build commands for FV2D depending on the architecture :

Configuring on CPU with Clang

mkdir build_cpu
cd build_cpu
CC=clang CXX=clang++ cmake -DKokkos_ENABLE_OPENMP=ON -DCMAKE_BUILD_TYPE=Release ..

Configuring on an NVidia GPU with Clang

mkdir build_nvgpu
cd build_nvgpu
CC=clang CXX=clang++ cmake -DKokkos_ENABLE_CUDA=ON -DCMAKE_BUILD_TYPE=Release ..

Note: using nvcc and not clang for compiling with CUDA might lead to problems because of an incompatibility between nvcc and HighFive the hdf5 wrapper used in this project.

Configuring on an AMD GPU with HIP

mkdir build_amdgpu
cd build_amdgpu
CC=hipcc CXX=hipcc cmake -DKokkos_ENABLE_HIP=ON -DCMAKE_BUILD_TYPE=Release ..

Compiling

Once the program has been successfully configured, you can build the code :

make -j

At the end of the compilation, an executable fv2d should be created in the build folder.

Usage

To use fv2d simply run the fv2d executable and provide a .ini file on the command line. For instance, if you are in a build folder located right below the root of fv2d, you can run a test by typing :

./fv2d ../settings/blast.ini

The code will start outputting information :

░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░
░        ░░   ░░░░░░░░░   ░░░░░░░░░░░░░░░░░   ░
▒   ▒▒▒▒▒▒▒▒   ▒▒▒▒▒▒▒   ▒▒▒   ▒  ▒▒▒▒▒▒▒▒▒   ▒
▒   ▒▒▒▒▒▒▒▒▒   ▒▒▒▒▒   ▒▒▒  ▒▒▒▒▒   ▒▒▒▒▒▒   ▒
▓       ▓▓▓▓▓▓   ▓▓▓   ▓▓▓▓▓▓▓▓▓   ▓▓▓▓   ▓   ▓
▓   ▓▓▓▓▓▓▓▓▓▓▓   ▓   ▓▓▓▓▓▓▓▓   ▓▓▓▓▓  ▓▓▓   ▓
▓   ▓▓▓▓▓▓▓▓▓▓▓▓     ▓▓▓▓▓▓▓   ▓▓▓▓▓▓▓  ▓▓▓   ▓
█   █████████████   ███████         ███   █   █
███████████████████████████████████████████████

Computing dts at (t=0) : dt_hyp=0.000521995; dt_TC=1e+06; dt_visc=1e+06
 - Saving at time 0
Computing dts at (t=0.000551063) : dt_hyp=0.000480941; dt_TC=1e+06; dt_visc=1e+06
Computing dts at (t=0.00106851) : dt_hyp=0.000459815; dt_TC=1e+06; dt_visc=1e+06
...

Each reporting line indicate the current time t and the timesteps associated with each physical process. dt_hyp is the hyperbolic time step associated with the hydro update, dt_TC the time step associated with thermal conduction, and finally dt_visc the viscous time step.

Once the execution is finished, the code displays a "goodbye" message :

    █     ▀██  ▀██         ▀██                              ▄█▄ 
   ███     ██   ██       ▄▄ ██    ▄▄▄   ▄▄ ▄▄▄     ▄▄▄▄     ███ 
  █  ██    ██   ██     ▄▀  ▀██  ▄█  ▀█▄  ██  ██  ▄█▄▄▄██    ▀█▀ 
 ▄▀▀▀▀█▄   ██   ██     █▄   ██  ██   ██  ██  ██  ██          █  
▄█▄  ▄██▄ ▄██▄ ▄██▄    ▀█▄▄▀██▄  ▀█▄▄█▀ ▄██▄ ██▄  ▀█▄▄▄▀     ▄  
                                                            ▀█▀