Installation: Machine Configuration File

The machine configuration file is located under configs. It specifies the library paths, compiler types, compilation flags, and GPU compute capability.

Tip

Check the configs directory to see if a configuration file is already available for your machine.

Set Up the Machine Configuration File

To set up your machine configuration file, go to configs and make a copy of template.config to modify:

cd configs
cp template.config your_machine.config

Please refer to the following sections to set up your machine configuration file.

0. Rules of the configuration file

• Comments must start with #.
• The variable name and its value must be separated by space(s).

1. Library paths

For example, MPI_PATH can be set by:

MPI_PATH /usr/local/mpich-3.2

Please also check out the installation of External Libraries.

2. Compilers

There are two compilers in this section: C++ (CXX) and MPI (CXX_MPI).

Note

MPI_PATH/bin/ will be combined with CXX_MPI automatically

3. Compilation flags

For example, CXXFLAG can be set by:

CXXFLAG -g -O2

or

CXXFLAG -g
CXXFLAG -O2

Here is a table of all the available flag variables:

Flag name	Description
CXXFLAG	Flags for compiler CXX and CXX_MPI
OPENMPFLAG	Flags for OpenMP
LIBFLAG	Flags for all libraries
NVCCFLAG_COM	Flags for nvcc compiler
NVCCFLAG_FLU	Flags for fluid solver files
NVCCFLAG_POT	Flags for Poisson/gravity solvers files

4. GPU compute capability

The GPU compute capability can be calculated by major_verison*100 + minor_version*10. For example, for GeForce RTX 4090, set GPU_COMPUTE_CAPABILITY 890 (8*100 + 9*10).

Tip

• You can also set GPU_COMPUTE_CAPABILITY to -1 to determine the value automatically using get_gpu_compute_capability() in configure.py.
• Check your GPU compute capability:
  1. https://developer.nvidia.com/cuda-gpus
  2. https://en.wikipedia.org/wiki/CUDA#GPUs_supported

Links

• Option List
• External Libraries
• Back to the main page of Installation