| 1 |
C $Header: /u/gcmpack/MITgcm/eesupp/src/cumulsum_z_tile.F,v 1.1 2011/07/01 18:18:31 jmc Exp $ |
| 2 |
C $Name: $ |
| 3 |
|
| 4 |
#include "CPP_EEOPTIONS.h" |
| 5 |
#include "W2_OPTIONS.h" |
| 6 |
|
| 7 |
C-- File w2_cumulsum_z_tile.F: Routines that perform cumulated sum |
| 8 |
C on a tiled array, corner grid-cell location |
| 9 |
C Contents |
| 10 |
C o W2_CUMULSUM_Z_TILE_RL |
| 11 |
C o W2_CUMULSUM_Z_TILE_RS <- not yet coded |
| 12 |
|
| 13 |
C---+----1----+----2----+----3----+----4----+----5----+----6----+----7-|--+----| |
| 14 |
CBOP |
| 15 |
C !ROUTINE: W2_CUMULSUM_Z_TILE_RL |
| 16 |
|
| 17 |
C !INTERFACE: |
| 18 |
SUBROUTINE W2_CUMULSUM_Z_TILE_RL( |
| 19 |
O psiZ, psiLoc, |
| 20 |
I dPsiX, dPsiY, myThid ) |
| 21 |
|
| 22 |
C !DESCRIPTION: |
| 23 |
C *==========================================================* |
| 24 |
C | SUBROUTINE W2\_CUMULSUM\_Z\_TILE\_RL |
| 25 |
C | o Handle cumulated sum for _RL tile data. |
| 26 |
C *==========================================================* |
| 27 |
C | Cumulate sum on tiled array, corner grid-cell location: |
| 28 |
C | Starts from 1rst tile and, going through all tiles & all |
| 29 |
C | the processes, add increment in both directions |
| 30 |
C *==========================================================* |
| 31 |
|
| 32 |
C !USES: |
| 33 |
IMPLICIT NONE |
| 34 |
|
| 35 |
C == Global data == |
| 36 |
#include "SIZE.h" |
| 37 |
#include "EEPARAMS.h" |
| 38 |
#include "EESUPPORT.h" |
| 39 |
#include "W2_EXCH2_SIZE.h" |
| 40 |
#include "W2_EXCH2_TOPOLOGY.h" |
| 41 |
#include "CUMULSUM.h" |
| 42 |
|
| 43 |
C !INPUT/OUTPUT PARAMETERS: |
| 44 |
C == Routine arguments == |
| 45 |
C psiZ :: results of cumulated sum, corresponds to tile South-East corner |
| 46 |
C psiLoc :: cumulated sum at special locations |
| 47 |
C dPsiX :: tile increment in X direction |
| 48 |
C dPsiY :: tile increment in Y direction |
| 49 |
C myThid :: my Thread Id. number |
| 50 |
_RL psiZ (nSx,nSy) |
| 51 |
_RL psiLoc(2) |
| 52 |
_RL dPsiX (nSx,nSy) |
| 53 |
_RL dPsiY (nSx,nSy) |
| 54 |
INTEGER myThid |
| 55 |
|
| 56 |
C !LOCAL VARIABLES: |
| 57 |
C == Local variables == |
| 58 |
C bi,bj :: tile indices |
| 59 |
C- type declaration of: loc[1,2]Buf and shareBufCS[1,2]_R8 : |
| 60 |
C all 4 needs to have the same length as MPI_DOUBLE_PRECISION |
| 61 |
INTEGER bi,bj |
| 62 |
INTEGER tN, tS |
| 63 |
Real*8 globalBuf(3,W2_maxNbTiles) |
| 64 |
#ifdef ALLOW_USE_MPI |
| 65 |
INTEGER npe, np1 |
| 66 |
INTEGER lbuf1, lbuf2, idest, itag, ready_to_receive |
| 67 |
INTEGER istatus(MPI_STATUS_SIZE), ierr |
| 68 |
Real*8 loc1Buf (nSx,nSy) |
| 69 |
Real*8 loc2Buf(2,nSx,nSy) |
| 70 |
#endif /* ALLOW_USE_MPI */ |
| 71 |
CEOP |
| 72 |
|
| 73 |
C-- Initialise to zero: |
| 74 |
psiLoc(1) = 0. |
| 75 |
psiLoc(2) = 0. |
| 76 |
DO tN = 1,exch2_nTiles |
| 77 |
globalBuf(1,tN) = 0. |
| 78 |
globalBuf(2,tN) = 0. |
| 79 |
globalBuf(3,tN) = 0. |
| 80 |
ENDDO |
| 81 |
|
| 82 |
C-- write input into shared-buffer array |
| 83 |
DO bj = myByLo(myThid), myByHi(myThid) |
| 84 |
DO bi = myBxLo(myThid), myBxHi(myThid) |
| 85 |
shareBufCS2_R8(1,bi,bj) = dPsiX(bi,bj) |
| 86 |
shareBufCS2_R8(2,bi,bj) = dPsiY(bi,bj) |
| 87 |
ENDDO |
| 88 |
ENDDO |
| 89 |
|
| 90 |
C-- Master thread cannot start until everyone is ready: |
| 91 |
CALL BAR2( myThid ) |
| 92 |
_BEGIN_MASTER( myThid ) |
| 93 |
|
| 94 |
#ifdef ALLOW_USE_MPI |
| 95 |
#ifndef ALWAYS_USE_MPI |
| 96 |
IF ( usingMPI ) THEN |
| 97 |
#endif |
| 98 |
|
| 99 |
lbuf1 = nSx*nSy |
| 100 |
lbuf2 = 2*lbuf1 |
| 101 |
idest = 0 |
| 102 |
itag = 0 |
| 103 |
ready_to_receive = 0 |
| 104 |
|
| 105 |
IF ( mpiMyId.NE.0 ) THEN |
| 106 |
|
| 107 |
C-- All proceses except 0 wait to be polled then send local array |
| 108 |
#ifndef DISABLE_MPI_READY_TO_RECEIVE |
| 109 |
CALL MPI_RECV (ready_to_receive, 1, MPI_INTEGER, |
| 110 |
& idest, itag, MPI_COMM_MODEL, istatus, ierr) |
| 111 |
#endif |
| 112 |
CALL MPI_SEND (shareBufCS2_R8, lbuf2, MPI_DOUBLE_PRECISION, |
| 113 |
& idest, itag, MPI_COMM_MODEL, ierr) |
| 114 |
|
| 115 |
C-- All proceses except 0 receive result from process 0 |
| 116 |
CALL MPI_RECV (shareBufCS1_R8, lbuf1, MPI_DOUBLE_PRECISION, |
| 117 |
& idest, itag, MPI_COMM_MODEL, istatus, ierr) |
| 118 |
|
| 119 |
ELSE |
| 120 |
|
| 121 |
C-- Process 0 polls and receives data from each process in turn |
| 122 |
DO npe = 1, numberOfProcs-1 |
| 123 |
#ifndef DISABLE_MPI_READY_TO_RECEIVE |
| 124 |
CALL MPI_SEND (ready_to_receive, 1, MPI_INTEGER, |
| 125 |
& npe, itag, MPI_COMM_MODEL, ierr) |
| 126 |
#endif |
| 127 |
CALL MPI_RECV (loc2Buf, lbuf2, MPI_DOUBLE_PRECISION, |
| 128 |
& npe, itag, MPI_COMM_MODEL, istatus, ierr) |
| 129 |
|
| 130 |
C-- Process 0 gathers the local arrays into a global array. |
| 131 |
np1 = npe + 1 |
| 132 |
DO bj=1,nSy |
| 133 |
DO bi=1,nSx |
| 134 |
tN = W2_procTileList(bi,bj,np1) |
| 135 |
globalBuf(1,tN) = loc2Buf(1,bi,bj) |
| 136 |
globalBuf(2,tN) = loc2Buf(2,bi,bj) |
| 137 |
ENDDO |
| 138 |
ENDDO |
| 139 |
ENDDO |
| 140 |
|
| 141 |
C-- end if process not 0 / else = 0 |
| 142 |
ENDIF |
| 143 |
|
| 144 |
#ifndef ALWAYS_USE_MPI |
| 145 |
ENDIF |
| 146 |
#endif |
| 147 |
#endif /* ALLOW_USE_MPI */ |
| 148 |
|
| 149 |
IF ( myProcId.EQ.0 ) THEN |
| 150 |
|
| 151 |
C-- Process 0 fills-in its local data |
| 152 |
DO bj=1,nSy |
| 153 |
DO bi=1,nSx |
| 154 |
tN = W2_myTileList(bi,bj) |
| 155 |
globalBuf(1,tN) = shareBufCS2_R8(1,bi,bj) |
| 156 |
globalBuf(2,tN) = shareBufCS2_R8(2,bi,bj) |
| 157 |
ENDDO |
| 158 |
ENDDO |
| 159 |
|
| 160 |
C-- Cumulate Sum over all tiles: |
| 161 |
DO tN = 1,exch2_nTiles |
| 162 |
globalBuf(3,tN) = 0. |
| 163 |
DO tS = 1,exch2_nTiles |
| 164 |
globalBuf(3,tN) = globalBuf(3,tN) |
| 165 |
& + W2_cumSum_tiles(1,tS,tN)*globalBuf(1,tS) |
| 166 |
& + W2_cumSum_tiles(2,tS,tN)*globalBuf(2,tS) |
| 167 |
ENDDO |
| 168 |
ENDDO |
| 169 |
C- Value at Special location (e.g., Missing-Corner values) |
| 170 |
IF ( W2_tMC1.GE.1 ) |
| 171 |
& psiLoc(1) = globalBuf(3,W2_tMC1) + globalBuf(2,W2_tMC1) |
| 172 |
IF ( W2_tMC2.GE.1 ) |
| 173 |
& psiLoc(2) = globalBuf(3,W2_tMC2) + globalBuf(1,W2_tMC2) |
| 174 |
|
| 175 |
C-- Process 0 fills-in its local data |
| 176 |
DO bj=1,nSy |
| 177 |
DO bi=1,nSx |
| 178 |
tN = W2_myTileList(bi,bj) |
| 179 |
shareBufCS1_R8(bi,bj) = globalBuf(3,tN) |
| 180 |
ENDDO |
| 181 |
ENDDO |
| 182 |
|
| 183 |
#ifdef ALLOW_USE_MPI |
| 184 |
#ifndef ALWAYS_USE_MPI |
| 185 |
IF ( usingMPI ) THEN |
| 186 |
#endif |
| 187 |
C-- Process 0 sends result to all other processes |
| 188 |
DO npe = 1, numberOfProcs-1 |
| 189 |
C- fill local array with relevant portion of global array |
| 190 |
np1 = npe + 1 |
| 191 |
DO bj=1,nSy |
| 192 |
DO bi=1,nSx |
| 193 |
tN = W2_procTileList(bi,bj,np1) |
| 194 |
loc1Buf(bi,bj) = globalBuf(3,tN) |
| 195 |
ENDDO |
| 196 |
ENDDO |
| 197 |
CALL MPI_SEND (loc1Buf, lbuf1, MPI_DOUBLE_PRECISION, |
| 198 |
& npe, itag, MPI_COMM_MODEL, ierr) |
| 199 |
ENDDO |
| 200 |
|
| 201 |
#ifndef ALWAYS_USE_MPI |
| 202 |
ENDIF |
| 203 |
#endif |
| 204 |
#endif /* ALLOW_USE_MPI */ |
| 205 |
|
| 206 |
C-- end if process 0 |
| 207 |
ENDIF |
| 208 |
|
| 209 |
_END_MASTER( myThid ) |
| 210 |
C-- Everyone wait for Master thread to be ready |
| 211 |
CALL BAR2( myThid ) |
| 212 |
|
| 213 |
C-- set result for every threads |
| 214 |
DO bj = myByLo(myThid), myByHi(myThid) |
| 215 |
DO bi = myBxLo(myThid), myBxHi(myThid) |
| 216 |
psiZ(bi,bj) = shareBufCS1_R8(bi,bj) |
| 217 |
ENDDO |
| 218 |
ENDDO |
| 219 |
|
| 220 |
RETURN |
| 221 |
END |
Parent Directory
| 
Revision Log
| 
Revision Graph