Message passing interface (MPI) is a flexible parallel programming paradigm, and
it is the dominant method of parallelizing codes in scientific computing.
MPI is suitable for both distributed
memory computers and shared memory architectures. In MPI, each task has an address space in memory that other tasks cannot
directly access. When exchange of data is needed between tasks, the tasks send 'messages' each other.
Compiling and linking MPI programs on Vuori
There are compiler wrappers mpif90, mpicc and mpiCC for compiling Fortran, C and C++ MPI programs, respectively. These wrappers take care of compiler and linker directives for compiling MPI programs. Users do not need to specify include file locations, MPI libraries or their locations.
For example, a MPI program written with the C language ('my_mpi_prog.c') is compiled as
mpicc my_mpi_prog.c -o my_mpi_prog
The -show option displays the actual compilation/linking command generated by the wrapper. For example, try
In Fortran programs use