Configuring the build

The material on the SWAN website provides a Makefile and two Perl scripts (platform.pl and switch.pl) that enables the user or developer to install SWAN on the computer in a proper manner. For this, the following platforms, operating systems and compilers are supported (and tested):

platform	OS	F90 compiler
Intel/AMD desktop/laptop	Linux	gfortran
Intel Core desktop/laptop	Linux	Intel
Intel Xeon desktop/laptop	Linux	Intel
x86-64 processor-based system	Linux	Portland Group
Intel/AMD desktop/laptop	Linux	Lahey
Intel Core desktop/laptop	MS Windows	Intel
Intel Xeon desktop/laptop	MS Windows	Intel
Intel/AMD desktop/laptop	MS Windows	Lahey
MacBook	macOS	gfortran
MacBook	macOS	Intel
SGI Origin 3000 (Silicon Graphics)	IRIX	SGI
IBM SP	AIX	IBM
Compaq True 64 Alpha (DEC ALFA)	OSF1	Compaq
Sun SPARC	Solaris	Sun
PA-RISC (HP 9000 series 700/800)	HP-UX v11	HP
IBM Power6 (pSeries 575)	Linux	IBM
Power Mac G4	Mac OS X	IBM

If your computer and available compiler is mentioned in the table, you may consult Section 3.2.2 for a complete build of the software without making any modifications. If desired, you may install SWAN manually; see Section 3.2.3.


Note that for a successful installation, a Perl package must be available on your computer. Usually, it is available for macOS, Linux and a UNIX-like operating system. Check it by typing perl -v. You can download Perl for MS Windows from Strawberry Perl. The Perl version should be at least 5.0.0 or higher!


Before starting the build process, the user may first decide how to run the SWAN program. There are three run modes:

• serial runs,
• parallel runs on shared-memory systems or
• parallel runs on distributed-memory machines.

For stationary and small-scale computations, it may be sufficient to choose the serial mode, i.e. one SWAN program running on one processor. However, for relatively large CPU-demanding calculations (e.g., instationary or nesting ones), two ways of parallelism for reducing the turn-around time are available:

• The SWAN code contains a number of so-called OpenMP directives that enables the compiler to generate multi-threaded code on a shared-memory computer. For this, you need a Fortran 90 compiler supporting OpenMP. The performance is good for a limited number of threads. This type of parallelism can be used e.g., on laptops containing multi-core processors.
• If the user want to run SWAN on a relative large number of processors, a message passing model is a good alternative. It is based on independent processors which do not share any memory but are connected via an interconnection network (e.g. cluster of Linux PC's connected via fast Ethernet switches). The implementation is based on the Message Passing Interface (MPI) standard (e.g., MPICH and OpenMPI). The SWAN code contains a set of generic subroutines that call a number of MPI routines, meant for local data exchange, gathering data, global reductions, etc. This technique is beneficial for larger simulations only, such that the communication times are relatively small compared to the computing times.
• For a proper installation of MPI-based application on Windows, please consult Section 3.2.4.