MPMD MPI decorator

The @mpmd_mpi decorator can be used to define Multiple Program Multiple Data (MPMD) MPI tasks as shown in the following example (Code 75):

Code 75 MPMD MPI task example

from pycompss.api.mpmd_mpi import mpmd_mpi

@mpmd_mpi(runner="mpirun",
          programs=[
               dict(binary="hostname", processes=2),
               dict(binary="date", processes=2)
          ])
@task()
def basic():
    pass

The definition implies that MPMD MPI command will be run by ‘mpirun’, and will execute 2 processes for ‘hostname’, and 2 processes to show the ‘ date’. It’s not mandatory to specify total number of programs as long as they are added inside programs list of dictionaries argument.

Each of the MPMD MPI programs must at least have binary, but also can have processes and args string (Code 76):

Code 76 MPMD MPI task example

from pycompss.api.mpmd_mpi import mpmd_mpi

@mpmd_mpi(runner="mpirun",
          programs=[
               dict(binary="date", processes=2, args="-d {{first}}"),
               dict(binary="date", processes=4, args="-d {{second}}")
          ])
@task()
def task_args(first, second):
    pass

def print_monday_friday(self):
    task_args("next monday", "next friday")
    compss_barrier()

When executed, this MPMD MPI program would invoke 2 MPI processes to print the date of next Monday, and 4 processes for next Friday. “args” string replaces every parameter that is ‘called’ between double curly braces with their real value. This allows using multiple FILE_IN parameters for multiple MPI programs. Moreover, output of the full MPMD MPI programs can be forwarded to an FILE_OUT_STDOUT param:

Code 77 MPMD MPI task example

from pycompss.api.mpmd_mpi import mpmd_mpi

@mpmd_mpi(runner="mpirun",
          programs=[
               dict(binary="grep", args="{{keyword}} {{in_file_1}}"),
               dict(binary="grep", args="{{keyword}} {{in_file_2}}"),
          ])
@task(in_file=FILE_IN, result={Type: FILE_OUT_STDOUT})
def std_out(keyword, in_file_1, in_file_2, result):
    pass

Other parameters of @mpmd_mpi decorator such as working_dir, fail_by_exit_value, processes_per_node, have the same behaviors as in @mpi.