Building the model

Building the model#

Before running the model, it needs to be compiled into an executable file. There are a number of ways to do this, but all require the following to be installed:

  • A modern Fortran compiler, such as GFortran ≥ 7 or Intel Fortran.

  • NetCDF Fortran

If you are on Linux or Mac, installing using apt or brew is easy: apt install gfortran libnetcdf-dev libnetcdff-dev or brew install gcc netcdf. If you are on Windows, we recommend using Windows Subsystem for Linux (in which case, installation is the same as for Linux using apt) or Cygwin (install package libnetcdf-fortran-devel).

Caution

If your system has multiple versions of GFortran installed, make sure that the NetCDF libraries are compiled using the same compiler version as you compile the NanoFASE model with. Issues can occur if, for instance, you have an older version of GFortran installed that is run using the gfortran command, and version 9 installed that is run using gfortran-9. apt will default to compiling NetCDF using the gfortran command, and thus the compiled libraries will not be compatible with NanoFASE model code compiled using gfortran-9.

Getting the code#

First, get a copy of the code by cloning the repo, making sure to specify --recurse-submodules to pull in code for the submodules in the vendor/ directory:

$ git clone https://github.com/nerc-ceh/nanofase.git --recurse-submodules
$ cd nanofase

If you forget the --recurse-submodule flag, then run the following from the repo directory:

$ git submodule update --init --recursive

Using fpm (Fortran Package Manager)#

The easiest way to build the model is use fpm (Fortran Package Manger). You can install this easily using Conda (conda install -c conda-forge fpm). To build the model without installing:

$ fpm @build

Note

If you are using Windows PowerShell, the “@” character is a special character and must be escaped using a backtick: fpm \@build`.

The model binary will be placed in a subdirectory of the build folder. The fpm @run command can be used to run this executable (see Runing the model). Alternatively, you can install the model:

$ fpm @install

Optionally, you can provide an install directory using --prefix <path-to-installation-dir>, otherwise the model will install to a default location for your system (e.g. /home/<user>/.local on Linux)

Using CMake#

The model can also be built using CMake:

$ mkdir build && cd build
$ cmake ..
$ make

After running these commands, a executable file named nanofase (nanofase.exe on Windows) will be available in the build directory. Use the -DCMAKE_BUILD_TYPE=Debug|Release flag to control whether you want to build the debug (default) or release version (e.g. cmake .. -DCMAKE_BUILD_TYPE=Release for the release version). Unless you are developing the model, use the release version as the model is significantly faster.

Caution

Note that, whilst CMake can be used to compile the model using the ifort or ifx compilers (e.g. by using the -DCMAKE_Fortran_COMPILER=ifort flag), issues can arise when the installed version of NetCDF that the CMake script finds isn’t a version compiled by an Intel compiler (such as those installed by apt and brew). Therefore, we recommend compiling using the compiler that your NetCDF installation used. As a workaround, you can explicitly point CMake to the correct version of NetCDF by declaring NETCDF_INCLUDES (include directories) and NETCDF_LIBRARIES_F90 (link libraries) (e.g. by specifying -DNETCDF_INCLUDES=.. -DNETCDF_LIBRARIES_F90=..).

Using Make#

The model can also be built using make, and an example Makefile is provided to help. This Makefile only uses GFortran compiler flags, and so you will need to extend this if you wish to use with another compiler. To use, simply take a copy of this file and run make:

$ cp Makefile.example
$ mkdir build
$ make

The default target builds the model with debugging enabled. To turn on optimisations, build using make release (for -O3 optimisations) or make fast (for -Ofast optimsations).

The model will be built into the build directory by default (a custom directory can be specified with the $BUILD_DIR option, e.g. make BUILD_DIR=makebuild). An executable file named nanofase will now be available in this directory.

Contents