| 1 |
C $Header: /u/gcmpack/MITgcm/eesupp/src/gather_vector.F,v 1.4 2006/10/19 06:54:23 dimitri Exp $ |
| 2 |
C $Name: $ |
| 3 |
|
| 4 |
#include "CPP_OPTIONS.h" |
| 5 |
|
| 6 |
SUBROUTINE GATHER_VECTOR( lprint, length, global, local, myThid ) |
| 7 |
C Gather elements of a vector from all mpi processes to process 0. |
| 8 |
IMPLICIT NONE |
| 9 |
#include "SIZE.h" |
| 10 |
#include "EEPARAMS.h" |
| 11 |
#include "EESUPPORT.h" |
| 12 |
C mythid - thread number for this instance of the routine. |
| 13 |
C global,local - working arrays used to transfer 2-D fields |
| 14 |
logical lprint |
| 15 |
INTEGER mythid |
| 16 |
INTEGER length |
| 17 |
Real*8 global(length*nPx*nPy) |
| 18 |
_RL local(length) |
| 19 |
|
| 20 |
INTEGER iG,jG,lG, l |
| 21 |
#ifdef ALLOW_USE_MPI |
| 22 |
_RL temp(length) |
| 23 |
INTEGER istatus(MPI_STATUS_SIZE), ierr |
| 24 |
INTEGER lbuff, idest, itag, npe, ready_to_receive |
| 25 |
#endif /* ALLOW_USE_MPI */ |
| 26 |
|
| 27 |
C-- Make everyone wait except for master thread. |
| 28 |
_BARRIER |
| 29 |
_BEGIN_MASTER( myThid ) |
| 30 |
|
| 31 |
#ifndef ALLOW_USE_MPI |
| 32 |
|
| 33 |
DO l=1,length |
| 34 |
iG=1+(myXGlobalLo-1)/sNx ! Kludge until unstructered tiles |
| 35 |
jG=1+(myYGlobalLo-1)/sNy ! Kludge until unstructered tiles |
| 36 |
lG= (jG-1)*nPx*length + (iG-1)*length + l |
| 37 |
global(lG) = local(l) |
| 38 |
ENDDO |
| 39 |
|
| 40 |
#else /* ALLOW_USE_MPI */ |
| 41 |
|
| 42 |
lbuff = length |
| 43 |
idest = 0 |
| 44 |
itag = 0 |
| 45 |
ready_to_receive = 0 |
| 46 |
|
| 47 |
IF( mpiMyId .EQ. 0 ) THEN |
| 48 |
|
| 49 |
C-- Process 0 fills-in its local data |
| 50 |
npe = 0 |
| 51 |
iG=mpi_myXGlobalLo(npe+1)/sNx+1 |
| 52 |
jG=mpi_myYGlobalLo(npe+1)/sNy+1 |
| 53 |
DO l=1,length |
| 54 |
lG= (jG-1)*nPx*length + (iG-1)*length + l |
| 55 |
global(lG) = local(l) |
| 56 |
ENDDO |
| 57 |
|
| 58 |
C-- Process 0 polls and receives data from each process in turn |
| 59 |
DO npe = 1, numberOfProcs-1 |
| 60 |
#ifndef DISABLE_MPI_READY_TO_RECEIVE |
| 61 |
CALL MPI_SEND (ready_to_receive, 1, MPI_INTEGER, |
| 62 |
& npe, itag, MPI_COMM_MODEL, ierr) |
| 63 |
#endif |
| 64 |
CALL MPI_RECV (temp, lbuff, MPI_DOUBLE_PRECISION, |
| 65 |
& npe, itag, MPI_COMM_MODEL, istatus, ierr) |
| 66 |
|
| 67 |
C-- Process 0 gathers the local arrays into a global array. |
| 68 |
iG=mpi_myXGlobalLo(npe+1)/sNx+1 |
| 69 |
jG=mpi_myYGlobalLo(npe+1)/sNy+1 |
| 70 |
cph( |
| 71 |
cph if (lprint) then |
| 72 |
cph print *, 'ph-gather A ', npe, |
| 73 |
cph & mpi_myXGlobalLo(npe+1), mpi_myYGlobalLo(npe+1) |
| 74 |
cph print *, 'ph-gather B ', npe, iG, jG |
| 75 |
cph endif |
| 76 |
cph) |
| 77 |
DO l=1,length |
| 78 |
lG= (jG-1)*nPx*length + (iG-1)*length + l |
| 79 |
global(lG) = temp(l) |
| 80 |
ENDDO |
| 81 |
|
| 82 |
|
| 83 |
ENDDO |
| 84 |
|
| 85 |
ELSE |
| 86 |
|
| 87 |
C-- All proceses except 0 wait to be polled then send local array |
| 88 |
#ifndef DISABLE_MPI_READY_TO_RECEIVE |
| 89 |
CALL MPI_RECV (ready_to_receive, 1, MPI_INTEGER, |
| 90 |
& idest, itag, MPI_COMM_MODEL, istatus, ierr) |
| 91 |
#endif |
| 92 |
CALL MPI_SEND (local, lbuff, MPI_DOUBLE_PRECISION, |
| 93 |
& idest, itag, MPI_COMM_MODEL, ierr) |
| 94 |
|
| 95 |
ENDIF |
| 96 |
|
| 97 |
#endif /* ALLOW_USE_MPI */ |
| 98 |
|
| 99 |
_END_MASTER( myThid ) |
| 100 |
_BARRIER |
| 101 |
|
| 102 |
RETURN |
| 103 |
END |
Parent Directory
| 
Revision Log
| 
Revision Graph