Configuring

Configuration Script

The configuration script (configure.py in the code root directory) is written in Python. Python 2.7 or above is required to run it (older versions may work if the argparse module is included, but they not officially supported). The configuration script selects the source code and generates a Makefile according to specified options. In order to see the list of available options, within the code root directory use:

> python configure.py -h

Options

The following option flags are available. --option [...] means a parameter is required, and the possible choices for each option are shown in the help message.

• -h, --help : show the help message
• --prob [problem_generator] : select a problem generator (from src/pgen/ matching filename)
• --coord [coordinates] : select a coordinate system (from src/coordinates/ matching filename)
• --eos [eos] : select an equation-of-state (adiabatic, isothermal, or general)
• --flux [riemann_solver] : select a Riemann solver
• --nghost [nghost] : set NGHOST to some nonnegative integer
• --nscalars [nscalars] : set NSCALARS to some nonnegative integer for number of Passive Scalars species
• -b : enable magnetic field
• -s : enable special relativity
• -g : enable general relativity
• -t : enable interface frame transformations for GR
• -debug : enable debug flags; override other compiler options
• -float : enable single precision (default is double precision)
• -mpi : enable MPI parallelization
• -omp : enable OpenMP parallelization
• -hdf5 : enable HDF5 Output (requires the HDF5 library)
• --hdf5_path [path] : path to HDF5 libraries
• -h5double : write HDF5 floating point output as H5T_NATIVE_DOUBLE (default is H5T_NATIVE_REAL)
• -fft : enable FFT capabilities (requires the FFTW library)
• --fftw_path [path] : path to FFTW libraries
• --grav [grav_solver] : select a self-gravity solver
• --cxx [compiler] : select a C++ compiler and predefined compiler flags (works with or without -mpi)
• --ccmd [compiler command] : set a compiler command overriding --cxx (when -mpi is not used)
• --mpiccmd [compiler command] : set a compiler command overriding --cxx and -mpi wrapper compiler syntax
• --cflag="[compiler flags]" : add additional flags to be given to compiler (appended, allowing for previous flags to be overridden)
• --include [path] : use -Ipath when compiling object files
• --lib_path [path] : use -Lpath when linking binary executable
• --lib [library] : use -llibrary when linking binary executable

Note, the argparse module allows each long option and parameter to be specified via --option=parameter or --option parameter.

Some combinations are prohibited. For example, the HLLD approximate Riemann solver cannot be used without enabling magnetic fields (-b). In most (but not necessarily all!) such cases, the script will issue a warning and quit. The order of options does not matter. Note that certain sets of options are required for some problem generators (e.g. an MHD problem requires that magnetic fields be enabled), but the script does not check this automatically.

Because mesh refinement (both static and adaptive) is fully integrated into the underlying algorithms, options are not needed to use it.

Some environments have compiler commands different from the standard ones. For example, on some Cray supercomputers, the compiler command is always CC no matter which compiler is selected. In this case, you can override the compiler command by configuring the code with --ccmd CC. You may also want to add flags (overriding incompatible, earlier flags generated by the configure script), for example to specify a target architecture. This can be done with the --cflag option. Note that this option should be set with an = sign to avoid the script trying to interpret any dashes. For example, --cflag="-O2 -xCORE-AVX512".

Example 1:

The linear-wave propagation test on a single core without magnetic fields using the HLLC approximate Riemann solver:

> python configure.py --prob linear_wave --flux hllc

Example 2:

The Orszag-Tang test (a typical 2D MHD test problem) in parallel using MPI on IBM BlueGene/Q, for debugging:

> python configure.py --prob orszag-tang -b --flux hlld --cxx bgxl -mpi -d

Example 3:

An MHD torus problem (similar to Stone & Pringle 2001) in spherical-polar coordinates, with hybrid parallelization and HDF5 output, using the Intel C++ Compiler.

> python configure.py --prob sphtorus --coord spherical_polar -b --flux hlld --cxx icc -mpi -omp -hdf5

Recommended options

While Athena++ supports many different algorithms, some are better than others. For hydrodynamics without magnetic fields, we recommend the HLLC (hllc) or Roe's (roe) approximate Riemann solvers, because they are more accurate. For MHD, either the HLLD (hlld) or Roe's solver are recommended. The HLLD solver is almost as accurate as Roe's, but it is somewhat faster and more robust in most situations.

Major changes from Athena 4.2

• + OpenMP parallelization
• + Flexible coordinate systems
• + HDF5 and Parallel IO
• - Corner-Transport-Upwind (CTU) integrator (incompatible with relativity)