nscepi/idefix

A new finite volume code designed to run on many architectures, such as GPU, CPU and manycores, using Kokkos.

• Download:
• Installation:
• Compile an example:
• Running
  • serial (gpu/cpu), openMP (cpu)
  • With MPI (cpu)
  • With MPI (gpu)
• Profiling
• Debugging
• Code Validation
• Contributing

Download:

Assuming you want to use https to get idefix (easiest option):

git clone https://github.com/idefix-code/idefix.git idefix
cd idefix
git submodule init
git submodule update

Installation:

Set the IDEFIX_DIR environment variable to the absolute path of the directory

export IDEFIX_DIR=<idefix main folder>

Add this line to ~/.<shell_rc_file> for a permanent install.

Compile an example:

Go to the example directory. For instance:

cd test/HD/sod

Configure the code launching cmake (version >= 3.16) in the example directory:

cmake $IDEFIX_DIR

Several options can be enabled from the command line (a complete list is available with cmake $IDEFIX_DIR -LH). For instance: -DIdefix_RECONSTRUCTION=Parabolic (enable PPM reconstruction), -DIdefix_MPI=ON (enable mpi), -DKokkos_ENABLE_OPENMP=ON (enable openmp parallelisation), etc... For more complex target architectures, it is recommended to use cmake GUI launching ccmake $IDEFIX_DIR in place of cmake and then switching on the required options.

Optional xdmf(hdf5+xmf) file dumping feature has been added to Idefix. This uses either serial or parallel implementation of hdf5 library which needs to be made available. These xdmf file pairs can be easily visualized in ParaView or VisIt by loading the xmf files. The hdf5 files can also be loaded easily in python (using h5py) for post-processing and post-run analysis. One can turn on xdmf data dumps by using -DIdefix_HDF5=ON. The [Output] block of .ini file is checked during runtime for a xdmf entry whih controls the frequency of xdmf file dumps during code execution.

One can then compile the code:

make -j8

Running

serial (gpu/cpu), openMP (cpu)

simple launch the executable

./idefix

With MPI (cpu)

-dec can be used to specify a domain decomposition manually.

It can be omitted for 1D problems, or if NX, NY, NZ and nproc are all powers of 2. Otherwise, -dec is mandatory. For instance, in 2D, using a 2x2 domain decomposition:

mpirun -np 4 ./idefix -dec 2 2

or in 3D, using a 1x2x4 decomposition:

mpirun -np 8 ./idefix -dec 1 2 4

With MPI (gpu)

The same rules for cpu domain decomposition applies for gpus. In addition, one should manually specify how many GPU devices one wants to use per node. Example, in a run with 2 nodes, 4 gpu per node, one would launch idefix with

mpirun -np 8 ./idefix -dec 1 2 4 --kokkos-num-devices=4

Profiling

use the embedded profiling tool by adding "-profile" when calling idefix (no need to recompile)

./idefix -profile

Debugging

Add -DIdefix_DEBUG=ON when calling cmake, or activate the Idefix_DEBUG option in ccmake, and recompile. Note that this option triggers a lot of outputs and memory access checks which significantly slow down the code.

Code Validation

Most of tests provided in the test/ directory can be validated against analytical solution (standard test) and/or pre-computed solutions (non-regression tests). Note that the validation relies on large reference files that are stored in the separate idefix-code/reference repository that is cloned as a submodule.

Ensure that reference files were properly downloaded (in the reference/ directory of the root of idefix) before attempting to validate the code.

In order to do a full validation of a particular test (with all of the possible combination of algorithms), use the script testme.py with the -all option, as in e.g.:

cd $IDEFIX_DIR/test/HD/sod
./testme.py -all

Tests require Python 3 along with some third party dependencies to be installed. To install those deps, run

pip install -r test/python_requirements.txt

Contributing

Idefix is developed with the help of the pre-commit framework. We use cpplint to validate code style, mostly following the Google standards for C++, and several pre-commit hooks to automatically fix some coding bad practices.

It is recommended (though not mandatory) to install pre-commit by running the following script from the top level of the repo

python3 -m pip install pre-commit
pre-commit install

Then, as one checks in their contribution with git commit, pre-commit hooks may perform changes in situ. One then needs to re-add and enter the git commit command again for the commit to be validated. Note that an important hook that does not perform auto-fixes is cpplint, so contributors need to accomodate for this one by hand.

Note that if for any reason you do not wish, or are unable to install pre-commit in your environment, formatting errors will be caught by our CI after you open a merge-request.