Compiling

Makefile

The configuration script creates a Makefile in the top-level directory, and you can build the code using standard make tools. Before building the code, it is strongly recommended that any stale or temporary files created by previous compilations be cleaned up.

> make clean

Then simply,

> make

Depending on the compiler and level of optimization, the compilation may take a while. You can speed it up using parallel bmake

> make -j

This process creates the executable bin/athena.

Compatibility

Because Athena++ is written in standard C++11 (although we do not use most of its new features), it should be able to be compiled using any compiler supporting the standard. This includes the GNU C++ compiler (g++), Intel C++ compiler (icc/icpc), Clang, Cray C++ Compiler, PGI C++ Compiler, IBM XL C++ Compiler, etc.. Athena++ uses #pragma simd defined in Intel Cilk Plus, but the pragma statements should not cause compatibility problems. Our configuration script supports GNU, Intel, Cray and IBM. If you want to use other compilers, configure the code for g++ and modify Makefile to specify appropriate options, flags, and libraries (see below).

Athena++ uses OpenMP for shared-memory parallelization and MPI-2 for distributed-memory parallelization. Most of the latest compilers support OpenMP used in Athena++, but if you are using Clang, we recommend the latest version as the older versions did not support OpenMP. For distributed-memory parallelization, any MPI library supporting the MPI-2 standard should work, including OpenMPI, MPICH, MVAPICH, Intel, Cray, IBM, etc..

Modifying the Makefile

On some machines, compilers may have different names, or may have different options. For example, on some Cray systems, 'CC' is used no matter which compiler environment is selected. In this case, the Makefile may be edited accordingly.

For best performance

On Intel-based systems, we strongly recommend use of the Intel C++ Compiler suite (icc, version 15 or later) because it results in a significantly faster executable than other compilers. If you want to try more aggressive optimization, try the "-fast" option. Also, static linking (-static) may improve the performance of highly parallel simulations.