cdds_initmpix - initialize the DDS system for MPIX parallelization

SYNOPSIS

#include "cdds.h"

int cdds_initmpix(node, nodes);

int *node
int *nodes

DESCRIPTION

This function is used to initialize the DDS system for MPIX parallelization (where MPI is not used if np=1 so programs can be piped when using a single node). Makefiles should define "MPIX" instead of "MPI". It determines the number of MPI nodes used and the node number (0 through nodes-1) for the current process. It checks the value of "np=" from the command line to determine if MPI is needed. If so and if np > 1, it checks if MPI is available on the current system and initializes MPI with the MPI_Init call. The master node then passes all the command line parameters to the slave nodes to ensure all nodes "see" the exact same parameters (since MPI does not guarantee that). If MPI is not available, it simply checks that "np=1". The "par:" dictionary is set for scanning upon exiting this routine. Any slave nodes will close any dictionaries opened for printing. The return value is alway 0.

NOTE: you will have to define this function for yourself because it cannot be defined in cdds.h (because it would try to use the parallel routines for all codes). So add "int cdds_initmpix" to any routine that uses it.

SEE ALSO

      cdds_initmpi Initialize DDS for MPI
      cdds_initopenmp Initialize DDS for OpenMP

AUTHOR

Jerry Ehlers, BP America (Aug 2007)