= Compiling the PDAF library = The library file of PDAF can be compiled on its own or in connection with an example implementation from `tutorial/`, `models/`, or `testsuite/`. In order to build the library file, you need a Fortran-2003 compatible compiler and 'make'. In addition, the libraries 'BLAS', 'LAPACK', and 'MPI' are required. Here, we describe the stand-alone compilation, which one can use for example to compile some model code with added coupling for PDAF. == Specify the machine-spefic compile settings == To compile the PDAF library first choose a suitable include file for the make process and/or edit one. In the directory `make.arch/` several include files are provided. The name of the include files as well as the header comment in the file described the intended architecture. The choice `linux_gfortran_openmpi.h` is a quite generic choice that should work on a wide range of computers. The environment variable '''PDAF_ARCH''' specifies for which architecture you compile PDAF according to your choice above. You need to specify the file name without '.h'. You can specify PDAF_ARCH in the make command line like `make PDAF_ARCH=`. Alternatively, you can set the environment variable $PDAF_ARCH in the shell to the name of the include file (without ending .h). For a (t)csh shell one can set the variable by `setenv PDAF_ARCH ` while for bash one uses `export PDAF_ARCH=`. == Compiling PDAF for ensemble filters and smoothers == To compile PDAF for ensemble filters and smoothers, do the following: {{{ cd src/ make [PDAF_ARCH=] }}} The compilation with generate the PDAF library file '''PDAFROOT/lib/libpdaf-d.a'''. In addition several Fortran module files `*.mod` are generated in the directory `PDAFROOT/include/`. These will be used when compiling the application program coupled to PDAF. == Compiling PDAF for 3D-Var methods == The 3D-Var methods implemented in PDAF use solver algorithms from external libraries (from the directory `external/`). To compile PDAF with full support for the analysis with 3D-Var schemes do the following: {{{ cd src/ make pdaf-var [PDAF_ARCH=] }}} The compilation with generate the PDAF library file '''PDAFROOT/lib/libpdaf-var.a'''. This library also includes the ensemble filters and smoothers. In addition several Fortran module files `*.mod` are generated in the directory `PDAFROOT/include/`. These will be used when compiling the application program coupled to PDAF. == Linking the PDAF library == To point to the PDAF include directory to ensure that the modules `pdaf_interface_module.mod` and `PDAFomi.mod` are found, one should specificy {{{ -I${PDAFROOT}/include }}} Here `${PDAFROOT}` is the directory of the PDAF package. It can either be directly specified by replacing ${PDAFROOT} or one speficies PDAFROOT as a variable. For linking the PDAF library for the ensemble schemes use {{{ -L${PDAFROOT}/lib -lpdaf-d }}} while in case of the PDAF library supporting the 3D-Var schemes use {{{ -L${PDAFROOT}/lib -lpdaf-var }}} Apart from the PDAF library, one also needs to specify the location of the BLAS and LAPACK libraries (for example by setting `-lblas -llapack` if the libraries are installed in standard directories) == Notes == * '''Note on parallelization:''' PDAF is generally intended for parallel computing using MPI. Before PDAF V2.0 it was also possible to compile PDAF without MPI by using a stub library that simulated the behavior for MPI for a single process that we provided with PDAF. With PDAF V2., we revised the MPI implementation and removed this option. Given that it is today extremely easy to install an MPI library (it is virtually always available as a package in any Linux distribution and standard on cluster computers) we prefer not to invest in maintaining and upgrading the MPI stub library. * '''Note on precision of floating point variables:''' PDAF is designed to use floating point variables of double precision. However, for flexibility the variables are decared in the source code without a 'KIND' specification. * For the compilation one has to specify that the compiler treats all 'real' variables with double precision accuracy. This is done, e.g. for gfortran by setting `-fdefault-real-8` or for ifort by setting `-r8`. In the provided include files in `make.arch/` these specifications are included. * PDAF also supports single precision. To enable it in PDAF, one has to use the preprocessor definition `-DSNGLPREC`. Then one has to compile without the double-precision settings.