[1] | 1 | SUBROUTINE cpu_statistics |
---|
| 2 | |
---|
[1036] | 3 | !--------------------------------------------------------------------------------! |
---|
| 4 | ! This file is part of PALM. |
---|
| 5 | ! |
---|
| 6 | ! PALM is free software: you can redistribute it and/or modify it under the terms |
---|
| 7 | ! of the GNU General Public License as published by the Free Software Foundation, |
---|
| 8 | ! either version 3 of the License, or (at your option) any later version. |
---|
| 9 | ! |
---|
| 10 | ! PALM is distributed in the hope that it will be useful, but WITHOUT ANY |
---|
| 11 | ! WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR |
---|
| 12 | ! A PARTICULAR PURPOSE. See the GNU General Public License for more details. |
---|
| 13 | ! |
---|
| 14 | ! You should have received a copy of the GNU General Public License along with |
---|
| 15 | ! PALM. If not, see <http://www.gnu.org/licenses/>. |
---|
| 16 | ! |
---|
| 17 | ! Copyright 1997-2012 Leibniz University Hannover |
---|
| 18 | !--------------------------------------------------------------------------------! |
---|
| 19 | ! |
---|
[484] | 20 | ! Current revisions: |
---|
[1] | 21 | ! ----------------- |
---|
| 22 | ! |
---|
[1017] | 23 | ! |
---|
[1] | 24 | ! Former revisions: |
---|
| 25 | ! ----------------- |
---|
[3] | 26 | ! $Id: cpu_statistics.f90 1037 2012-10-22 14:10:22Z raasch $ |
---|
[83] | 27 | ! |
---|
[1037] | 28 | ! 1036 2012-10-22 13:43:42Z raasch |
---|
| 29 | ! code put under GPL (PALM 3.9) |
---|
| 30 | ! |
---|
[1017] | 31 | ! 1015 2012-09-27 09:23:24Z raasch |
---|
| 32 | ! output of accelerator board information |
---|
| 33 | ! |
---|
[1015] | 34 | ! 683 2011-02-09 14:25:15Z raasch |
---|
| 35 | ! output of handling of ghostpoint exchange |
---|
| 36 | ! |
---|
[623] | 37 | ! 622 2010-12-10 08:08:13Z raasch |
---|
| 38 | ! output of handling of collective operations |
---|
| 39 | ! |
---|
[226] | 40 | ! 222 2009-01-12 16:04:16Z letzel |
---|
| 41 | ! Bugfix for nonparallel execution |
---|
| 42 | ! |
---|
[198] | 43 | ! 197 2008-09-16 15:29:03Z raasch |
---|
| 44 | ! Format adjustments in order to allow CPU# > 999, |
---|
| 45 | ! data are collected from PE0 in an ordered sequence which seems to avoid |
---|
| 46 | ! hanging of processes on SGI-ICE |
---|
| 47 | ! |
---|
[83] | 48 | ! 82 2007-04-16 15:40:52Z raasch |
---|
| 49 | ! Preprocessor directives for old systems removed |
---|
| 50 | ! |
---|
[3] | 51 | ! RCS Log replace by Id keyword, revision history cleaned up |
---|
| 52 | ! |
---|
[1] | 53 | ! Revision 1.13 2006/04/26 12:10:51 raasch |
---|
| 54 | ! Output of number of threads per task, max = min in case of 1 PE |
---|
| 55 | ! |
---|
| 56 | ! Revision 1.1 1997/07/24 11:11:11 raasch |
---|
| 57 | ! Initial revision |
---|
| 58 | ! |
---|
| 59 | ! |
---|
| 60 | ! Description: |
---|
| 61 | ! ------------ |
---|
| 62 | ! Analysis and output of the cpu-times measured. All PE results are collected |
---|
| 63 | ! on PE0 in order to calculate the mean cpu-time over all PEs and other |
---|
| 64 | ! statistics. The output is sorted according to the amount of cpu-time consumed |
---|
| 65 | ! and output on PE0. |
---|
[3] | 66 | !------------------------------------------------------------------------------! |
---|
[1] | 67 | |
---|
| 68 | USE cpulog |
---|
| 69 | USE pegrid |
---|
| 70 | USE control_parameters |
---|
| 71 | |
---|
| 72 | IMPLICIT NONE |
---|
| 73 | |
---|
| 74 | INTEGER :: i, ii(1), iii, lp, sender |
---|
| 75 | REAL, SAVE :: norm = 1.0 |
---|
| 76 | REAL, DIMENSION(:), ALLOCATABLE :: pe_max, pe_min, pe_rms, sum |
---|
| 77 | REAL, DIMENSION(:,:), ALLOCATABLE :: pe_log_points |
---|
| 78 | |
---|
| 79 | |
---|
| 80 | ! |
---|
| 81 | !-- Compute cpu-times in seconds |
---|
| 82 | log_point%mtime = log_point%mtime / norm |
---|
| 83 | log_point%sum = log_point%sum / norm |
---|
| 84 | log_point%vector = log_point%vector / norm |
---|
| 85 | WHERE ( log_point%counts /= 0 ) |
---|
| 86 | log_point%mean = log_point%sum / log_point%counts |
---|
| 87 | END WHERE |
---|
| 88 | |
---|
| 89 | |
---|
| 90 | ! |
---|
| 91 | !-- Collect cpu-times from all PEs and calculate statistics |
---|
| 92 | IF ( myid == 0 ) THEN |
---|
| 93 | ! |
---|
| 94 | !-- Allocate and initialize temporary arrays needed for statistics |
---|
| 95 | ALLOCATE( pe_max( SIZE( log_point ) ), pe_min( SIZE( log_point ) ), & |
---|
| 96 | pe_rms( SIZE( log_point ) ), & |
---|
| 97 | pe_log_points( SIZE( log_point ), 0:numprocs-1 ) ) |
---|
| 98 | pe_min = log_point%sum |
---|
| 99 | pe_max = log_point%sum ! need to be set in case of 1 PE |
---|
| 100 | pe_rms = 0.0 |
---|
| 101 | |
---|
| 102 | #if defined( __parallel ) |
---|
| 103 | ! |
---|
| 104 | !-- Receive data from all PEs |
---|
| 105 | DO i = 1, numprocs-1 |
---|
| 106 | CALL MPI_RECV( pe_max(1), SIZE( log_point ), MPI_REAL, & |
---|
[197] | 107 | i, i, comm2d, status, ierr ) |
---|
[1] | 108 | sender = status(MPI_SOURCE) |
---|
| 109 | pe_log_points(:,sender) = pe_max |
---|
| 110 | ENDDO |
---|
| 111 | pe_log_points(:,0) = log_point%sum ! Results from PE0 |
---|
| 112 | ! |
---|
| 113 | !-- Calculate mean of all PEs, store it on log_point%sum |
---|
| 114 | !-- and find minimum and maximum |
---|
| 115 | DO iii = 1, SIZE( log_point ) |
---|
| 116 | DO i = 1, numprocs-1 |
---|
| 117 | log_point(iii)%sum = log_point(iii)%sum + pe_log_points(iii,i) |
---|
| 118 | pe_min(iii) = MIN( pe_min(iii), pe_log_points(iii,i) ) |
---|
| 119 | pe_max(iii) = MAX( pe_max(iii), pe_log_points(iii,i) ) |
---|
| 120 | ENDDO |
---|
| 121 | log_point(iii)%sum = log_point(iii)%sum / numprocs |
---|
| 122 | ! |
---|
| 123 | !-- Calculate rms |
---|
| 124 | DO i = 0, numprocs-1 |
---|
| 125 | pe_rms(iii) = pe_rms(iii) + ( & |
---|
| 126 | pe_log_points(iii,i) - log_point(iii)%sum & |
---|
| 127 | )**2 |
---|
| 128 | ENDDO |
---|
| 129 | pe_rms(iii) = SQRT( pe_rms(iii) / numprocs ) |
---|
| 130 | ENDDO |
---|
| 131 | ELSE |
---|
| 132 | ! |
---|
| 133 | !-- Send data to PE0 (pe_max is used as temporary storage to send |
---|
| 134 | !-- the data in order to avoid sending the data type log) |
---|
| 135 | ALLOCATE( pe_max( SIZE( log_point ) ) ) |
---|
| 136 | pe_max = log_point%sum |
---|
[197] | 137 | CALL MPI_SEND( pe_max(1), SIZE( log_point ), MPI_REAL, 0, myid, comm2d, & |
---|
[1] | 138 | ierr ) |
---|
| 139 | #endif |
---|
| 140 | |
---|
| 141 | ENDIF |
---|
| 142 | |
---|
| 143 | ! |
---|
| 144 | !-- Write cpu-times |
---|
| 145 | IF ( myid == 0 ) THEN |
---|
| 146 | ! |
---|
| 147 | !-- Re-store sums |
---|
| 148 | ALLOCATE( sum( SIZE( log_point ) ) ) |
---|
| 149 | WHERE ( log_point%counts /= 0 ) |
---|
| 150 | sum = log_point%sum |
---|
| 151 | ELSEWHERE |
---|
| 152 | sum = -1.0 |
---|
| 153 | ENDWHERE |
---|
| 154 | |
---|
| 155 | ! |
---|
| 156 | !-- Write cpu-times sorted by size |
---|
| 157 | CALL check_open( 18 ) |
---|
[222] | 158 | #if defined( __parallel ) |
---|
[1] | 159 | WRITE ( 18, 100 ) TRIM( run_description_header ), & |
---|
[181] | 160 | numprocs * threads_per_task, pdims(1), pdims(2), & |
---|
[1] | 161 | threads_per_task |
---|
[1015] | 162 | IF ( num_acc_per_node /= 0 ) WRITE ( 18, 108 ) num_acc_per_node |
---|
| 163 | WRITE ( 18, 110 ) |
---|
[222] | 164 | #else |
---|
| 165 | WRITE ( 18, 100 ) TRIM( run_description_header ), & |
---|
| 166 | numprocs * threads_per_task, 1, 1, & |
---|
| 167 | threads_per_task |
---|
[1015] | 168 | IF ( num_acc_per_node /= 0 ) WRITE ( 18, 109 ) num_acc_per_node |
---|
| 169 | WRITE ( 18, 110 ) |
---|
[222] | 170 | #endif |
---|
[1] | 171 | DO |
---|
| 172 | ii = MAXLOC( sum ) |
---|
| 173 | i = ii(1) |
---|
| 174 | IF ( sum(i) /= -1.0 ) THEN |
---|
| 175 | WRITE ( 18, 102 ) & |
---|
| 176 | log_point(i)%place, log_point(i)%sum, & |
---|
| 177 | log_point(i)%sum / log_point(1)%sum * 100.0, & |
---|
| 178 | log_point(i)%counts, pe_min(i), pe_max(i), pe_rms(i) |
---|
| 179 | sum(i) = -1.0 |
---|
| 180 | ELSE |
---|
| 181 | EXIT |
---|
| 182 | ENDIF |
---|
| 183 | ENDDO |
---|
| 184 | ENDIF |
---|
| 185 | |
---|
| 186 | |
---|
| 187 | ! |
---|
| 188 | !-- The same procedure again for the individual measurements. |
---|
| 189 | ! |
---|
| 190 | !-- Compute cpu-times in seconds |
---|
| 191 | log_point_s%mtime = log_point_s%mtime / norm |
---|
| 192 | log_point_s%sum = log_point_s%sum / norm |
---|
| 193 | log_point_s%vector = log_point_s%vector / norm |
---|
| 194 | WHERE ( log_point_s%counts /= 0 ) |
---|
| 195 | log_point_s%mean = log_point_s%sum / log_point_s%counts |
---|
| 196 | END WHERE |
---|
| 197 | |
---|
| 198 | ! |
---|
| 199 | !-- Collect cpu-times from all PEs and calculate statistics |
---|
| 200 | #if defined( __parallel ) |
---|
| 201 | ! |
---|
| 202 | !-- Set barrier in order to avoid that PE0 receives log_point_s-data |
---|
| 203 | !-- while still busy with receiving log_point-data (see above) |
---|
| 204 | CALL MPI_BARRIER( comm2d, ierr ) |
---|
| 205 | #endif |
---|
| 206 | IF ( myid == 0 ) THEN |
---|
| 207 | ! |
---|
| 208 | !-- Initialize temporary arrays needed for statistics |
---|
| 209 | pe_min = log_point_s%sum |
---|
| 210 | pe_max = log_point_s%sum ! need to be set in case of 1 PE |
---|
| 211 | pe_rms = 0.0 |
---|
| 212 | |
---|
| 213 | #if defined( __parallel ) |
---|
| 214 | ! |
---|
| 215 | !-- Receive data from all PEs |
---|
| 216 | DO i = 1, numprocs-1 |
---|
| 217 | CALL MPI_RECV( pe_max(1), SIZE( log_point ), MPI_REAL, & |
---|
| 218 | MPI_ANY_SOURCE, MPI_ANY_TAG, comm2d, status, ierr ) |
---|
| 219 | sender = status(MPI_SOURCE) |
---|
| 220 | pe_log_points(:,sender) = pe_max |
---|
| 221 | ENDDO |
---|
| 222 | pe_log_points(:,0) = log_point_s%sum ! Results from PE0 |
---|
| 223 | ! |
---|
| 224 | !-- Calculate mean of all PEs, store it on log_point_s%sum |
---|
| 225 | !-- and find minimum and maximum |
---|
| 226 | DO iii = 1, SIZE( log_point ) |
---|
| 227 | DO i = 1, numprocs-1 |
---|
| 228 | log_point_s(iii)%sum = log_point_s(iii)%sum + pe_log_points(iii,i) |
---|
| 229 | pe_min(iii) = MIN( pe_min(iii), pe_log_points(iii,i) ) |
---|
| 230 | pe_max(iii) = MAX( pe_max(iii), pe_log_points(iii,i) ) |
---|
| 231 | ENDDO |
---|
| 232 | log_point_s(iii)%sum = log_point_s(iii)%sum / numprocs |
---|
| 233 | ! |
---|
| 234 | !-- Calculate rms |
---|
| 235 | DO i = 0, numprocs-1 |
---|
| 236 | pe_rms(iii) = pe_rms(iii) + ( & |
---|
| 237 | pe_log_points(iii,i) - log_point_s(iii)%sum & |
---|
| 238 | )**2 |
---|
| 239 | ENDDO |
---|
| 240 | pe_rms(iii) = SQRT( pe_rms(iii) / numprocs ) |
---|
| 241 | ENDDO |
---|
| 242 | ELSE |
---|
| 243 | ! |
---|
| 244 | !-- Send data to PE0 (pe_max is used as temporary storage to send |
---|
| 245 | !-- the data in order to avoid sending the data type log) |
---|
| 246 | pe_max = log_point_s%sum |
---|
| 247 | CALL MPI_SEND( pe_max(1), SIZE( log_point ), MPI_REAL, 0, 0, comm2d, & |
---|
| 248 | ierr ) |
---|
| 249 | #endif |
---|
| 250 | |
---|
| 251 | ENDIF |
---|
| 252 | |
---|
| 253 | ! |
---|
| 254 | !-- Write cpu-times |
---|
| 255 | IF ( myid == 0 ) THEN |
---|
| 256 | ! |
---|
| 257 | !-- Re-store sums |
---|
| 258 | WHERE ( log_point_s%counts /= 0 ) |
---|
| 259 | sum = log_point_s%sum |
---|
| 260 | ELSEWHERE |
---|
| 261 | sum = -1.0 |
---|
| 262 | ENDWHERE |
---|
| 263 | |
---|
| 264 | ! |
---|
| 265 | !-- Write cpu-times sorted by size |
---|
| 266 | WRITE ( 18, 101 ) |
---|
| 267 | DO |
---|
| 268 | ii = MAXLOC( sum ) |
---|
| 269 | i = ii(1) |
---|
| 270 | IF ( sum(i) /= -1.0 ) THEN |
---|
| 271 | WRITE ( 18, 102 ) & |
---|
| 272 | log_point_s(i)%place, log_point_s(i)%sum, & |
---|
| 273 | log_point_s(i)%sum / log_point(1)%sum * 100.0, & |
---|
| 274 | log_point_s(i)%counts, pe_min(i), pe_max(i), pe_rms(i) |
---|
| 275 | sum(i) = -1.0 |
---|
| 276 | ELSE |
---|
| 277 | EXIT |
---|
| 278 | ENDIF |
---|
| 279 | ENDDO |
---|
| 280 | |
---|
| 281 | ! |
---|
[683] | 282 | !-- Output of handling of MPI operations |
---|
[622] | 283 | IF ( collective_wait ) THEN |
---|
| 284 | WRITE ( 18, 103 ) |
---|
| 285 | ELSE |
---|
| 286 | WRITE ( 18, 104 ) |
---|
| 287 | ENDIF |
---|
[683] | 288 | IF ( synchronous_exchange ) THEN |
---|
| 289 | WRITE ( 18, 105 ) |
---|
| 290 | ELSE |
---|
| 291 | WRITE ( 18, 106 ) |
---|
| 292 | ENDIF |
---|
[622] | 293 | |
---|
| 294 | ! |
---|
[1] | 295 | !-- Empty lines in order to create a gap to the results of the model |
---|
| 296 | !-- continuation runs |
---|
[683] | 297 | WRITE ( 18, 107 ) |
---|
[1] | 298 | |
---|
| 299 | ! |
---|
| 300 | !-- Unit 18 is not needed anymore |
---|
| 301 | CALL close_file( 18 ) |
---|
| 302 | |
---|
| 303 | ENDIF |
---|
| 304 | |
---|
| 305 | |
---|
[181] | 306 | 100 FORMAT (A/11('-')//'CPU measures for ',I5,' PEs (',I5,'(x) * ',I5,'(y', & |
---|
[1015] | 307 | &') tasks *',I5,' threads):') |
---|
[1] | 308 | |
---|
| 309 | 101 FORMAT (/'special measures:'/ & |
---|
| 310 | &'-----------------------------------------------------------', & |
---|
| 311 | &'--------------------') |
---|
| 312 | |
---|
| 313 | 102 FORMAT (A20,2X,F9.3,2X,F7.2,1X,I7,3(1X,F9.3)) |
---|
[622] | 314 | 103 FORMAT (/'Barriers are set in front of collective operations') |
---|
| 315 | 104 FORMAT (/'No barriers are set in front of collective operations') |
---|
[683] | 316 | 105 FORMAT (/'Exchange of ghostpoints via MPI_SENDRCV') |
---|
| 317 | 106 FORMAT (/'Exchange of ghostpoints via MPI_ISEND/MPI_IRECV') |
---|
| 318 | 107 FORMAT (//) |
---|
[1015] | 319 | 108 FORMAT ('Accelerator boards per node: ',I2) |
---|
| 320 | 109 FORMAT ('Accelerator boards: ',I2) |
---|
| 321 | 110 FORMAT ('----------------------------------------------------------', & |
---|
| 322 | &'------------'//& |
---|
| 323 | &'place: mean counts min ', & |
---|
| 324 | &' max rms'/ & |
---|
| 325 | &' sec. % sec. ', & |
---|
| 326 | &' sec. sec.'/ & |
---|
| 327 | &'-----------------------------------------------------------', & |
---|
| 328 | &'-------------------') |
---|
[1] | 329 | |
---|
| 330 | END SUBROUTINE cpu_statistics |
---|
| 331 | |
---|