| 1 |
C $Header: /u/gcmpack/MITgcm/eesupp/src/global_vec_sum.F,v 1.4 2006/08/12 03:10:26 edhill Exp $ |
| 2 |
C $Name: $ |
| 3 |
#include "CPP_EEOPTIONS.h" |
| 4 |
|
| 5 |
C-- File global_vec_sum.F: Perform a global sum on a tiled-array of vectors. |
| 6 |
C-- Contents |
| 7 |
C-- o GLOBAL_VEC_SUM_R4 |
| 8 |
C-- o GLOBAL_VEC_SUM_R8 |
| 9 |
C-- o GLOBAL_VEC_SUM_INT |
| 10 |
|
| 11 |
C---+----1----+----2----+----3----+----4----+----5----+----6----+----7-|--+----| |
| 12 |
CBOP |
| 13 |
C !ROUTINE: GLOBAL_VEC_SUM_R4 |
| 14 |
|
| 15 |
C !INTERFACE: |
| 16 |
SUBROUTINE GLOBAL_VEC_SUM_R4( |
| 17 |
I ndim, nval, |
| 18 |
U sumPhi, |
| 19 |
I myThid ) |
| 20 |
|
| 21 |
C !DESCRIPTION: |
| 22 |
C Sum the vector over tiles and then sum the result over all MPI |
| 23 |
C processes. Within a process only one thread (Master) does the sum. |
| 24 |
C The same thread also does the inter-process sum for example with MPI |
| 25 |
C and then writes the result into a shared location. All threads wait |
| 26 |
C until the sum is available. |
| 27 |
C Warning: Only works if argument array "sumPhi" is shared by all threads. |
| 28 |
|
| 29 |
C !USES: |
| 30 |
IMPLICIT NONE |
| 31 |
#include "SIZE.h" |
| 32 |
#include "EEPARAMS.h" |
| 33 |
#include "EESUPPORT.h" |
| 34 |
c#include "GLOBAL_SUM.h" |
| 35 |
|
| 36 |
C !INPUT/OUTPUT PARAMETERS: |
| 37 |
C sumPhi :: input/output array |
| 38 |
C myThid :: thread ID |
| 39 |
INTEGER ndim, nval, myThid |
| 40 |
Real*4 sumPhi(ndim,nSx,nSy) |
| 41 |
CEOP |
| 42 |
|
| 43 |
C !LOCAL VARIABLES: |
| 44 |
C mpiRC :: MPI return code |
| 45 |
INTEGER i, bi,bj |
| 46 |
Real*4 tmp1(nval), tmp2(nval) |
| 47 |
#ifdef ALLOW_USE_MPI |
| 48 |
INTEGER mpiRC |
| 49 |
#endif /* ALLOW_USE_MPI */ |
| 50 |
|
| 51 |
_BARRIER |
| 52 |
_BEGIN_MASTER( myThid ) |
| 53 |
|
| 54 |
C Sum over all tiles |
| 55 |
DO i = 1,nval |
| 56 |
tmp1(i) = 0. |
| 57 |
ENDDO |
| 58 |
DO bj = 1,nSy |
| 59 |
DO bi = 1,nSx |
| 60 |
DO i = 1,nval |
| 61 |
tmp1(i) = tmp1(i) + sumPhi( i, bi,bj ) |
| 62 |
ENDDO |
| 63 |
ENDDO |
| 64 |
ENDDO |
| 65 |
|
| 66 |
C Invoke MPI if necessary |
| 67 |
IF ( usingMPI ) THEN |
| 68 |
|
| 69 |
#ifdef ALLOW_USE_MPI |
| 70 |
CALL MPI_Allreduce(tmp1,tmp2,nval,MPI_REAL, |
| 71 |
& MPI_SUM,MPI_COMM_MODEL,mpiRC) |
| 72 |
#endif /* ALLOW_USE_MPI */ |
| 73 |
|
| 74 |
C Copy the results to the first location of the input array |
| 75 |
DO i = 1,nval |
| 76 |
sumPhi( i, 1,1 ) = tmp2(i) |
| 77 |
ENDDO |
| 78 |
|
| 79 |
ELSE |
| 80 |
C Copy the results to the first location of the input array |
| 81 |
DO i = 1,nval |
| 82 |
sumPhi( i, 1,1 ) = tmp1(i) |
| 83 |
ENDDO |
| 84 |
|
| 85 |
ENDIF |
| 86 |
|
| 87 |
_END_MASTER( myThid ) |
| 88 |
_BARRIER |
| 89 |
|
| 90 |
RETURN |
| 91 |
END |
| 92 |
|
| 93 |
C---+----1----+----2----+----3----+----4----+----5----+----6----+----7-|--+----| |
| 94 |
CBOP |
| 95 |
C !ROUTINE: GLOBAL_VEC_SUM_R8 |
| 96 |
|
| 97 |
C !INTERFACE: |
| 98 |
SUBROUTINE GLOBAL_VEC_SUM_R8( |
| 99 |
I ndim, nval, |
| 100 |
U sumPhi, |
| 101 |
I myThid ) |
| 102 |
|
| 103 |
C !DESCRIPTION: |
| 104 |
C Sum the vector over tiles and then sum the result over all MPI |
| 105 |
C processes. Within a process only one thread (Master) does the sum. |
| 106 |
C The same thread also does the inter-process sum for example with MPI |
| 107 |
C and then writes the result into a shared location. All threads wait |
| 108 |
C until the sum is available. |
| 109 |
C Warning: Only works if argument array "sumPhi" is shared by all threads. |
| 110 |
|
| 111 |
C !USES: |
| 112 |
IMPLICIT NONE |
| 113 |
#include "SIZE.h" |
| 114 |
#include "EEPARAMS.h" |
| 115 |
#include "EESUPPORT.h" |
| 116 |
c#include "GLOBAL_SUM.h" |
| 117 |
|
| 118 |
C !INPUT/OUTPUT PARAMETERS: |
| 119 |
C sumPhi :: input/output array |
| 120 |
C myThid :: thread ID |
| 121 |
INTEGER ndim, nval, myThid |
| 122 |
Real*8 sumPhi(ndim,nSx,nSy) |
| 123 |
CEOP |
| 124 |
|
| 125 |
C !LOCAL VARIABLES: |
| 126 |
C mpiRC :: MPI return code |
| 127 |
INTEGER i, bi,bj |
| 128 |
Real*8 tmp1(nval), tmp2(nval) |
| 129 |
#ifdef ALLOW_USE_MPI |
| 130 |
INTEGER mpiRC |
| 131 |
#endif /* ALLOW_USE_MPI */ |
| 132 |
|
| 133 |
_BARRIER |
| 134 |
_BEGIN_MASTER( myThid ) |
| 135 |
|
| 136 |
C Sum over all tiles |
| 137 |
DO i = 1,nval |
| 138 |
tmp1(i) = 0. |
| 139 |
ENDDO |
| 140 |
DO bj = 1,nSy |
| 141 |
DO bi = 1,nSx |
| 142 |
DO i = 1,nval |
| 143 |
tmp1(i) = tmp1(i) + sumPhi( i, bi,bj ) |
| 144 |
ENDDO |
| 145 |
ENDDO |
| 146 |
ENDDO |
| 147 |
|
| 148 |
C Invoke MPI if necessary |
| 149 |
IF ( usingMPI ) THEN |
| 150 |
|
| 151 |
#ifdef ALLOW_USE_MPI |
| 152 |
CALL MPI_Allreduce(tmp1,tmp2,nval,MPI_DOUBLE_PRECISION, |
| 153 |
& MPI_SUM,MPI_COMM_MODEL,mpiRC) |
| 154 |
#endif /* ALLOW_USE_MPI */ |
| 155 |
|
| 156 |
C Copy the results to the first location of the input array |
| 157 |
DO i = 1,nval |
| 158 |
sumPhi( i, 1,1 ) = tmp2(i) |
| 159 |
ENDDO |
| 160 |
|
| 161 |
ELSE |
| 162 |
C Copy the results to the first location of the input array |
| 163 |
DO i = 1,nval |
| 164 |
sumPhi( i, 1,1 ) = tmp1(i) |
| 165 |
ENDDO |
| 166 |
|
| 167 |
ENDIF |
| 168 |
|
| 169 |
_END_MASTER( myThid ) |
| 170 |
_BARRIER |
| 171 |
|
| 172 |
RETURN |
| 173 |
END |
| 174 |
|
| 175 |
C---+----1----+----2----+----3----+----4----+----5----+----6----+----7-|--+----| |
| 176 |
CBOP |
| 177 |
C !ROUTINE: GLOBAL_VEC_SUM_INT |
| 178 |
|
| 179 |
C !INTERFACE: |
| 180 |
SUBROUTINE GLOBAL_VEC_SUM_INT( |
| 181 |
I ndim, nval, |
| 182 |
U sumPhi, |
| 183 |
I myThid ) |
| 184 |
|
| 185 |
C !DESCRIPTION: |
| 186 |
C Sum the vector over tiles and then sum the result over all MPI |
| 187 |
C processes. Within a process only one thread (Master) does the sum. |
| 188 |
C The same thread also does the inter-process sum for example with MPI |
| 189 |
C and then writes the result into a shared location. All threads wait |
| 190 |
C until the sum is available. |
| 191 |
C Warning: Only works if argument array "sumPhi" is shared by all threads. |
| 192 |
|
| 193 |
C !USES: |
| 194 |
IMPLICIT NONE |
| 195 |
#include "SIZE.h" |
| 196 |
#include "EEPARAMS.h" |
| 197 |
#include "EESUPPORT.h" |
| 198 |
c#include "GLOBAL_SUM.h" |
| 199 |
|
| 200 |
C !INPUT/OUTPUT PARAMETERS: |
| 201 |
C sumPhi :: input/output array |
| 202 |
C myThid :: thread ID |
| 203 |
INTEGER ndim, nval, myThid |
| 204 |
INTEGER sumPhi(ndim,nSx,nSy) |
| 205 |
CEOP |
| 206 |
|
| 207 |
C !LOCAL VARIABLES: |
| 208 |
C mpiRC :: MPI return code |
| 209 |
INTEGER i, bi,bj |
| 210 |
INTEGER tmp1(nval), tmp2(nval) |
| 211 |
#ifdef ALLOW_USE_MPI |
| 212 |
INTEGER mpiRC |
| 213 |
#endif /* ALLOW_USE_MPI */ |
| 214 |
|
| 215 |
_BARRIER |
| 216 |
_BEGIN_MASTER( myThid ) |
| 217 |
|
| 218 |
C Sum over all tiles |
| 219 |
DO i = 1,nval |
| 220 |
tmp1(i) = 0 |
| 221 |
ENDDO |
| 222 |
DO bj = 1,nSy |
| 223 |
DO bi = 1,nSx |
| 224 |
DO i = 1,nval |
| 225 |
tmp1(i) = tmp1(i) + sumPhi( i, bi,bj ) |
| 226 |
ENDDO |
| 227 |
ENDDO |
| 228 |
ENDDO |
| 229 |
|
| 230 |
C Invoke MPI if necessary |
| 231 |
IF ( usingMPI ) THEN |
| 232 |
|
| 233 |
#ifdef ALLOW_USE_MPI |
| 234 |
CALL MPI_Allreduce(tmp1,tmp2,nval,MPI_INTEGER, |
| 235 |
& MPI_SUM,MPI_COMM_MODEL,mpiRC) |
| 236 |
#endif /* ALLOW_USE_MPI */ |
| 237 |
|
| 238 |
C Copy the results to the first location of the input array |
| 239 |
DO i = 1,nval |
| 240 |
sumPhi( i, 1,1 ) = tmp2(i) |
| 241 |
ENDDO |
| 242 |
|
| 243 |
ELSE |
| 244 |
C Copy the results to the first location of the input array |
| 245 |
DO i = 1,nval |
| 246 |
sumPhi( i, 1,1 ) = tmp1(i) |
| 247 |
ENDDO |
| 248 |
|
| 249 |
ENDIF |
| 250 |
|
| 251 |
_END_MASTER( myThid ) |
| 252 |
_BARRIER |
| 253 |
|
| 254 |
RETURN |
| 255 |
END |
| 256 |
|
| 257 |
C---+----1----+----2----+----3----+----4----+----5----+----6----+----7-|--+----| |
Parent Directory
| 
Revision Log
| 
Revision Graph