1: !
  2: !     This introductory example illustrates running PETSc on a subset
  3: !     of processes
  4: !
  5: !/*T
  6: !   Concepts: introduction to PETSc;
  7: !   Concepts: process^subset set PETSC_COMM_WORLD
  8: !   Processors: 2
  9: !T*/
 10: ! -----------------------------------------------------------------------

 12:       program main
 13: #include <petsc/finclude/petscsys.h>
 14:       use petscmpi  ! or mpi or mpi_f08
 15:       use petscsys
 16:       implicit none
 17:       PetscErrorCode ierr
 18:       PetscMPIInt    rank, size,grank,zero,two
 19:       PetscReal globalrank

 21: !     We must call MPI_Init() first, making us, not PETSc, responsible
 22: !     for MPI

 24:       call MPI_Init(ierr)
 25: #if defined(PETSC_HAVE_ELEMENTAL)
 26:       call PetscElementalInitializePackage(ierr)
 27: #endif
 28: !     We can now change the communicator universe for PETSc

 30:       zero = 0
 31:       two = 2
 32:       call MPI_Comm_rank(MPI_COMM_WORLD,rank,ierr)
 33:       call MPI_Comm_split(MPI_COMM_WORLD,mod(rank,two),zero,            &
 34:      &     PETSC_COMM_WORLD,ierr)

 36: !     Every PETSc routine should begin with the PetscInitialize()
 37: !     routine.
 38:       call PetscInitializeNoArguments(ierr)
 39:       if (ierr .ne. 0) then
 40:          print*,'Unable to initialize PETSc'
 41:          stop
 42:       endif

 44: !     The following MPI calls return the number of processes being used
 45: !     and the rank of this process in the group.

 47:       call MPI_Comm_size(PETSC_COMM_WORLD,size,ierr)
 48:       call MPI_Comm_rank(PETSC_COMM_WORLD,rank,ierr)


 51: !     Here we would like to print only one message that represents all
 52: !     the processes in the group. Sleep so that IO from different ranks
 53: !     don't get mixed up. Note this is not an ideal solution
 54:       call MPI_Comm_rank(MPI_COMM_WORLD,grank,ierr)
 55:       globalrank = grank
 56:       call PetscSleep(globalrank,ierr)
 57:       if (rank .eq. 0) write(6,100) size,rank
 58:  100  format('No of Procs = ',i4,' rank = ',i4)

 60: !     Always call PetscFinalize() before exiting a program.  This
 61: !     routine - finalizes the PETSc libraries as well as MPI - provides
 62: !     summary and diagnostic information if certain runtime options are
 63: !     chosen (e.g., -log_view).  See PetscFinalize() manpage for more
 64: !     information.

 66:       call PetscFinalize(ierr)
 67:       call MPI_Comm_free(PETSC_COMM_WORLD,ierr)
 68: #if defined(PETSC_HAVE_ELEMENTAL)
 69:       call PetscElementalFinalizePackage(ierr)
 70: #endif


 73: !     Since we initialized MPI, we must call MPI_Finalize()

 75:       call  MPI_Finalize(ierr)
 76:       end

 78: !/*TEST
 79: !
 80: !   test:
 81: !      nsize: 5
 82: !      filter: sort -b
 83: !      filter_output: sort -b
 84: !      requires: !cuda
 85: !
 86: !TEST*/