1! 2! This introductory example illustrates running PETSc on a subset 3! of processes 4! 5! ----------------------------------------------------------------------- 6 7 program main 8#include <petsc/finclude/petscsys.h> 9 use petscmpi ! or mpi or mpi_f08 10 use petscsys 11 implicit none 12 PetscErrorCode ierr 13 PetscMPIInt rank, size,grank,zero,two 14 PetscReal globalrank 15 16! We must call MPI_Init() first, making us, not PETSc, responsible for MPI 17 18 PetscCallMPIA(MPI_Init(ierr)) 19#if defined(PETSC_HAVE_ELEMENTAL) 20 PetscCallA(PetscElementalInitializePackage(ierr)) 21#endif 22! We can now change the communicator universe for PETSc 23 24 zero = 0 25 two = 2 26 PetscCallMPIA(MPI_Comm_rank(MPI_COMM_WORLD,rank,ierr)) 27 PetscCallMPIA(MPI_Comm_split(MPI_COMM_WORLD,mod(rank,two),zero,PETSC_COMM_WORLD,ierr)) 28 29! Every PETSc routine should begin with the PetscInitialize() 30! routine. 31 PetscCallA(PetscInitializeNoArguments(ierr)) 32 33! The following MPI calls return the number of processes being used 34! and the rank of this process in the group. 35 36 PetscCallMPIA(MPI_Comm_size(PETSC_COMM_WORLD,size,ierr)) 37 PetscCallMPIA(MPI_Comm_rank(PETSC_COMM_WORLD,rank,ierr)) 38 39! Here we would like to print only one message that represents all 40! the processes in the group. Sleep so that IO from different ranks 41! don't get mixed up. Note this is not an ideal solution 42 PetscCallMPIA(MPI_Comm_rank(MPI_COMM_WORLD,grank,ierr)) 43 globalrank = grank 44 PetscCallA(PetscSleep(globalrank,ierr)) 45 if (rank .eq. 0) write(6,100) size,rank 46 100 format('No of Procs = ',i4,' rank = ',i4) 47 48! Always call PetscFinalize() before exiting a program. This 49! routine - finalizes the PETSc libraries as well as MPI - provides 50! summary and diagnostic information if certain runtime options are 51! chosen (e.g., -log_view). See PetscFinalize() manpage for more 52! information. 53 54 PetscCallA(PetscFinalize(ierr)) 55 PetscCallMPIA(MPI_Comm_free(PETSC_COMM_WORLD,ierr)) 56#if defined(PETSC_HAVE_ELEMENTAL) 57 PetscCallA(PetscElementalFinalizePackage(ierr)) 58#endif 59 60! Since we initialized MPI, we must call MPI_Finalize() 61 62 PetscCallMPIA(MPI_Finalize(ierr)) 63 end 64 65!/*TEST 66! 67! test: 68! nsize: 5 69! filter: sort -b 70! filter_output: sort -b 71! requires: !cuda !saws 72! 73!TEST*/ 74