| 1 |
jmc |
1.1 |
C $Header: /u/gcmpack/MITgcm/eesupp/src/global_sum_tile.F,v 1.3 2009/06/10 03:47:03 jmc Exp $ |
| 2 |
|
|
C $Name: $ |
| 3 |
|
|
|
| 4 |
|
|
#include "PACKAGES_CONFIG.h" |
| 5 |
|
|
#include "CPP_EEOPTIONS.h" |
| 6 |
|
|
|
| 7 |
|
|
C-- File 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 CUMULSUM_Z_TILE_RL |
| 11 |
|
|
C o CUMULSUM_Z_TILE_RS <- not yet coded |
| 12 |
|
|
|
| 13 |
|
|
C---+----1----+----2----+----3----+----4----+----5----+----6----+----7-|--+----| |
| 14 |
|
|
CBOP |
| 15 |
|
|
C !ROUTINE: CUMULSUM_Z_TILE_RL |
| 16 |
|
|
|
| 17 |
|
|
C !INTERFACE: |
| 18 |
|
|
SUBROUTINE CUMULSUM_Z_TILE_RL( |
| 19 |
|
|
O psiZ, psiLoc, |
| 20 |
|
|
I dPsiX, dPsiY, myThid ) |
| 21 |
|
|
|
| 22 |
|
|
C !DESCRIPTION: |
| 23 |
|
|
C *==========================================================* |
| 24 |
|
|
C | SUBROUTINE 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 |
|
|
c#include "CUMULSUM.h" |
| 40 |
|
|
COMMON / CUMULSUM_R8 / shareBufCS1_R8, shareBufCS2_R8 |
| 41 |
|
|
Real*8 shareBufCS1_R8 (nSx,nSy) |
| 42 |
|
|
Real*8 shareBufCS2_R8(2,nSx,nSy) |
| 43 |
|
|
|
| 44 |
|
|
C !INPUT/OUTPUT PARAMETERS: |
| 45 |
|
|
C == Routine arguments == |
| 46 |
|
|
C psiZ :: results of cumulated sum, corresponds to tile South-East corner |
| 47 |
|
|
C psiLoc :: cumulated sum at special locations |
| 48 |
|
|
C dPsiX :: tile increment in X direction |
| 49 |
|
|
C dPsiY :: tile increment in Y direction |
| 50 |
|
|
C myThid :: my Thread Id. number |
| 51 |
|
|
_RL psiZ (nSx,nSy) |
| 52 |
|
|
_RL psiLoc(2) |
| 53 |
|
|
_RL dPsiX (nSx,nSy) |
| 54 |
|
|
_RL dPsiY (nSx,nSy) |
| 55 |
|
|
INTEGER myThid |
| 56 |
|
|
|
| 57 |
|
|
C !LOCAL VARIABLES: |
| 58 |
|
|
#ifndef ALLOW_EXCH2 |
| 59 |
|
|
C == Local variables == |
| 60 |
|
|
C bi,bj :: tile indices |
| 61 |
|
|
C- type declaration of: loc[1,2]Buf and shareBufCS[1,2]_R8 : |
| 62 |
|
|
C all 4 needs to have the same length as MPI_DOUBLE_PRECISION |
| 63 |
|
|
INTEGER bi,bj |
| 64 |
|
|
INTEGER nf |
| 65 |
|
|
#ifdef ALLOW_USE_MPI |
| 66 |
|
|
INTEGER biG, bjG, npe, np1 |
| 67 |
|
|
INTEGER lbuf1, lbuf2, idest, itag, ready_to_receive |
| 68 |
|
|
INTEGER istatus(MPI_STATUS_SIZE), ierr |
| 69 |
|
|
Real*8 loc1Buf (nSx,nSy) |
| 70 |
|
|
Real*8 loc2Buf(2,nSx,nSy) |
| 71 |
|
|
Real*8 globalBuf(3,nSx*nPx,nSy*nPy) |
| 72 |
|
|
#endif /* ALLOW_USE_MPI */ |
| 73 |
|
|
#else |
| 74 |
|
|
CHARACTER*(MAX_LEN_MBUF) msgBuf |
| 75 |
|
|
#endif /* ALLOW_EXCH2 */ |
| 76 |
|
|
CEOP |
| 77 |
|
|
|
| 78 |
|
|
#ifdef ALLOW_EXCH2 |
| 79 |
|
|
c CALL EXCH2_CUMULSUM_Z_TILE_RL( |
| 80 |
|
|
c O psiZ, psiLoc, |
| 81 |
|
|
c I dPsiX, dPsiY, myThid ) |
| 82 |
|
|
WRITE(msgBuf,'(A)') |
| 83 |
|
|
& 'CUMULSUM_Z_TILE_RL: missing call to EXCH2 S/R' |
| 84 |
|
|
CALL PRINT_ERROR( msgBuf, myThid ) |
| 85 |
|
|
RETURN |
| 86 |
|
|
#else /* ALLOW_EXCH2 */ |
| 87 |
|
|
|
| 88 |
|
|
C-- write input into shared-buffer array |
| 89 |
|
|
DO bj = myByLo(myThid), myByHi(myThid) |
| 90 |
|
|
DO bi = myBxLo(myThid), myBxHi(myThid) |
| 91 |
|
|
shareBufCS2_R8(1,bi,bj) = dPsiX(bi,bj) |
| 92 |
|
|
shareBufCS2_R8(2,bi,bj) = dPsiY(bi,bj) |
| 93 |
|
|
ENDDO |
| 94 |
|
|
ENDDO |
| 95 |
|
|
psiLoc(1) = 0. |
| 96 |
|
|
psiLoc(2) = 0. |
| 97 |
|
|
|
| 98 |
|
|
C-- Master thread cannot start until everyone is ready: |
| 99 |
|
|
CALL BAR2( myThid ) |
| 100 |
|
|
_BEGIN_MASTER( myThid ) |
| 101 |
|
|
|
| 102 |
|
|
#ifdef ALLOW_USE_MPI |
| 103 |
|
|
#ifdef ALWAYS_USE_MPI |
| 104 |
|
|
IF ( .TRUE. ) THEN |
| 105 |
|
|
#else |
| 106 |
|
|
IF ( usingMPI ) THEN |
| 107 |
|
|
#endif |
| 108 |
|
|
|
| 109 |
|
|
lbuf1 = nSx*nSy |
| 110 |
|
|
lbuf2 = 2*lbuf1 |
| 111 |
|
|
idest = 0 |
| 112 |
|
|
itag = 0 |
| 113 |
|
|
ready_to_receive = 0 |
| 114 |
|
|
|
| 115 |
|
|
IF ( mpiMyId.NE.0 ) THEN |
| 116 |
|
|
|
| 117 |
|
|
C-- All proceses except 0 wait to be polled then send local array |
| 118 |
|
|
#ifndef DISABLE_MPI_READY_TO_RECEIVE |
| 119 |
|
|
CALL MPI_RECV (ready_to_receive, 1, MPI_INTEGER, |
| 120 |
|
|
& idest, itag, MPI_COMM_MODEL, istatus, ierr) |
| 121 |
|
|
#endif |
| 122 |
|
|
CALL MPI_SEND (shareBufCS2_R8, lbuf2, MPI_DOUBLE_PRECISION, |
| 123 |
|
|
& idest, itag, MPI_COMM_MODEL, ierr) |
| 124 |
|
|
|
| 125 |
|
|
C-- All proceses except 0 receive result from process 0 |
| 126 |
|
|
CALL MPI_RECV (shareBufCS1_R8, lbuf1, MPI_DOUBLE_PRECISION, |
| 127 |
|
|
& idest, itag, MPI_COMM_MODEL, istatus, ierr) |
| 128 |
|
|
|
| 129 |
|
|
ELSE |
| 130 |
|
|
|
| 131 |
|
|
C-- Process 0 fills-in its local data |
| 132 |
|
|
np1 = 1 |
| 133 |
|
|
DO bj=1,nSy |
| 134 |
|
|
DO bi=1,nSx |
| 135 |
|
|
biG = (mpi_myXGlobalLo(np1)-1)/sNx+bi |
| 136 |
|
|
bjG = (mpi_myYGlobalLo(np1)-1)/sNy+bj |
| 137 |
|
|
globalBuf(1,biG,bjG) = shareBufCS2_R8(1,bi,bj) |
| 138 |
|
|
globalBuf(2,biG,bjG) = shareBufCS2_R8(2,bi,bj) |
| 139 |
|
|
ENDDO |
| 140 |
|
|
ENDDO |
| 141 |
|
|
|
| 142 |
|
|
C-- Process 0 polls and receives data from each process in turn |
| 143 |
|
|
DO npe = 1, numberOfProcs-1 |
| 144 |
|
|
#ifndef DISABLE_MPI_READY_TO_RECEIVE |
| 145 |
|
|
CALL MPI_SEND (ready_to_receive, 1, MPI_INTEGER, |
| 146 |
|
|
& npe, itag, MPI_COMM_MODEL, ierr) |
| 147 |
|
|
#endif |
| 148 |
|
|
CALL MPI_RECV (loc2Buf, lbuf2, MPI_DOUBLE_PRECISION, |
| 149 |
|
|
& npe, itag, MPI_COMM_MODEL, istatus, ierr) |
| 150 |
|
|
|
| 151 |
|
|
C-- Process 0 gathers the local arrays into a global array. |
| 152 |
|
|
np1 = npe + 1 |
| 153 |
|
|
DO bj=1,nSy |
| 154 |
|
|
DO bi=1,nSx |
| 155 |
|
|
biG = (mpi_myXGlobalLo(np1)-1)/sNx+bi |
| 156 |
|
|
bjG = (mpi_myYGlobalLo(np1)-1)/sNy+bj |
| 157 |
|
|
globalBuf(1,biG,bjG) = loc2Buf(1,bi,bj) |
| 158 |
|
|
globalBuf(2,biG,bjG) = loc2Buf(2,bi,bj) |
| 159 |
|
|
ENDDO |
| 160 |
|
|
ENDDO |
| 161 |
|
|
ENDDO |
| 162 |
|
|
|
| 163 |
|
|
C-- Cumulate Sum over all tiles: |
| 164 |
|
|
globalBuf(3,1,1) = 0. |
| 165 |
|
|
bj = 1 |
| 166 |
|
|
DO bi = 1,nSx*nPx-1 |
| 167 |
|
|
globalBuf(3,1+bi,bj) = globalBuf(3,bi,bj) |
| 168 |
|
|
& + globalBuf(1,bi,bj) |
| 169 |
|
|
ENDDO |
| 170 |
|
|
DO bj = 1,nSy*nPy-1 |
| 171 |
|
|
DO bi = 1,nSx*nPx |
| 172 |
|
|
globalBuf(3,bi,1+bj) = globalBuf(3,bi,bj) |
| 173 |
|
|
& + globalBuf(2,bi,bj) |
| 174 |
|
|
ENDDO |
| 175 |
|
|
ENDDO |
| 176 |
|
|
|
| 177 |
|
|
C-- Process 0 fills-in its local data |
| 178 |
|
|
np1 = 1 |
| 179 |
|
|
DO bj=1,nSy |
| 180 |
|
|
DO bi=1,nSx |
| 181 |
|
|
biG = (mpi_myXGlobalLo(np1)-1)/sNx+bi |
| 182 |
|
|
bjG = (mpi_myYGlobalLo(np1)-1)/sNy+bj |
| 183 |
|
|
shareBufCS1_R8(bi,bj) = globalBuf(3,biG,bjG) |
| 184 |
|
|
ENDDO |
| 185 |
|
|
ENDDO |
| 186 |
|
|
|
| 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 |
|
|
biG = (mpi_myXGlobalLo(np1)-1)/sNx+bi |
| 194 |
|
|
bjG = (mpi_myYGlobalLo(np1)-1)/sNy+bj |
| 195 |
|
|
loc1Buf(bi,bj) = globalBuf(3,biG,bjG) |
| 196 |
|
|
ENDDO |
| 197 |
|
|
ENDDO |
| 198 |
|
|
CALL MPI_SEND (loc1Buf, lbuf1, MPI_DOUBLE_PRECISION, |
| 199 |
|
|
& npe, itag, MPI_COMM_MODEL, ierr) |
| 200 |
|
|
|
| 201 |
|
|
ENDDO |
| 202 |
|
|
|
| 203 |
|
|
ENDIF |
| 204 |
|
|
|
| 205 |
|
|
ELSEIF (useCubedSphereExchange) THEN |
| 206 |
|
|
#else /* not USE_MPI */ |
| 207 |
|
|
IF (useCubedSphereExchange) THEN |
| 208 |
|
|
#endif /* ALLOW_USE_MPI */ |
| 209 |
|
|
|
| 210 |
|
|
C-- assume 1 tile / face, from bi=1 to 6, no MPI |
| 211 |
|
|
shareBufCS1_R8(1,1) = 0. |
| 212 |
|
|
bj = 1 |
| 213 |
|
|
DO bi = 1,nSx-1 |
| 214 |
|
|
nf = 1 + MOD(1+bi,2) |
| 215 |
|
|
shareBufCS1_R8(1+bi,bj) = shareBufCS1_R8(bi,bj) |
| 216 |
|
|
& + shareBufCS2_R8(nf,bi,bj) |
| 217 |
|
|
ENDDO |
| 218 |
|
|
C- fill in missing corner: 1 = North-West corner of face 1 |
| 219 |
|
|
C- 2 = South-East corner of face 2 |
| 220 |
|
|
bi = 1 |
| 221 |
|
|
psiLoc(1) = shareBufCS1_R8(bi,bj) + shareBufCS2_R8(2,bi,bj) |
| 222 |
|
|
bi = MIN(2,nSx) |
| 223 |
|
|
psiLoc(2) = shareBufCS1_R8(bi,bj) + shareBufCS2_R8(1,bi,bj) |
| 224 |
|
|
|
| 225 |
|
|
ELSE |
| 226 |
|
|
|
| 227 |
|
|
C-- Cumulate Sum over all tiles: |
| 228 |
|
|
shareBufCS1_R8(1,1) = 0. |
| 229 |
|
|
bj = 1 |
| 230 |
|
|
DO bi = 1,nSx-1 |
| 231 |
|
|
shareBufCS1_R8(1+bi,bj) = shareBufCS1_R8(bi,bj) |
| 232 |
|
|
& + shareBufCS2_R8(1,bi,bj) |
| 233 |
|
|
ENDDO |
| 234 |
|
|
DO bj = 1,nSy-1 |
| 235 |
|
|
DO bi = 1,nSx |
| 236 |
|
|
shareBufCS1_R8(bi,1+bj) = shareBufCS1_R8(bi,bj) |
| 237 |
|
|
& + shareBufCS2_R8(2,bi,bj) |
| 238 |
|
|
ENDDO |
| 239 |
|
|
ENDDO |
| 240 |
|
|
|
| 241 |
|
|
ENDIF |
| 242 |
|
|
|
| 243 |
|
|
_END_MASTER( myThid ) |
| 244 |
|
|
C-- Everyone wait for Master thread to be ready |
| 245 |
|
|
CALL BAR2( myThid ) |
| 246 |
|
|
|
| 247 |
|
|
C-- set result for every threads |
| 248 |
|
|
DO bj = myByLo(myThid), myByHi(myThid) |
| 249 |
|
|
DO bi = myBxLo(myThid), myBxHi(myThid) |
| 250 |
|
|
psiZ(bi,bj) = shareBufCS1_R8(bi,bj) |
| 251 |
|
|
ENDDO |
| 252 |
|
|
ENDDO |
| 253 |
|
|
|
| 254 |
|
|
RETURN |
| 255 |
|
|
#endif /* ALLOW_EXCH2 */ |
| 256 |
|
|
END |
Parent Directory
| 
Revision Log
| 
Revision Graph