[1] | 1 | SUBROUTINE check_for_restart |
---|
| 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 | ! |
---|
[1310] | 17 | ! Copyright 1997-2014 Leibniz Universitaet Hannover |
---|
[1036] | 18 | !--------------------------------------------------------------------------------! |
---|
| 19 | ! |
---|
[247] | 20 | ! Current revisions: |
---|
[1] | 21 | ! ----------------- |
---|
[1354] | 22 | ! |
---|
[1510] | 23 | ! |
---|
[1321] | 24 | ! Former revisions: |
---|
| 25 | ! ----------------- |
---|
| 26 | ! $Id: check_for_restart.f90 1510 2014-12-16 08:57:29Z suehring $ |
---|
| 27 | ! |
---|
[1510] | 28 | ! 1509 2014-12-16 08:56:46Z heinze |
---|
| 29 | ! bugfix: prevent infinite loop in case of automatic restarts |
---|
| 30 | ! |
---|
[1469] | 31 | ! 1468 2014-09-24 14:06:57Z maronga |
---|
| 32 | ! Added support for unscheduled job termination using the flag files |
---|
| 33 | ! DO_STOP_NOW and DO_RESTART_NOW |
---|
| 34 | ! |
---|
[1354] | 35 | ! 1353 2014-04-08 15:21:23Z heinze |
---|
| 36 | ! REAL constants provided with KIND-attribute |
---|
| 37 | ! |
---|
[1321] | 38 | ! 1320 2014-03-20 08:40:49Z raasch |
---|
[1320] | 39 | ! ONLY-attribute added to USE-statements, |
---|
| 40 | ! kind-parameters added to all INTEGER and REAL declaration statements, |
---|
| 41 | ! kinds are defined in new module kinds, |
---|
| 42 | ! revision history before 2012 removed, |
---|
| 43 | ! comment fields (!:) to be used for variable explanations added to |
---|
| 44 | ! all variable declaration statements |
---|
[1] | 45 | ! |
---|
[1037] | 46 | ! 1036 2012-10-22 13:43:42Z raasch |
---|
| 47 | ! code put under GPL (PALM 3.9) |
---|
| 48 | ! |
---|
[1033] | 49 | ! 1032 2012-10-21 13:03:21Z letzel |
---|
| 50 | ! minor reformatting |
---|
| 51 | ! |
---|
[1] | 52 | ! Revision 1.1 1998/03/18 20:06:51 raasch |
---|
| 53 | ! Initial revision |
---|
| 54 | ! |
---|
| 55 | ! |
---|
| 56 | ! Description: |
---|
| 57 | ! ------------ |
---|
| 58 | ! Set stop flag, if restart is neccessary because of expiring cpu-time or |
---|
| 59 | ! if it is forced by user |
---|
| 60 | !------------------------------------------------------------------------------! |
---|
| 61 | |
---|
[1320] | 62 | USE control_parameters, & |
---|
| 63 | ONLY: coupling_mode, dt_restart, end_time, message_string, & |
---|
| 64 | run_description_header, simulated_time, terminate_coupled, & |
---|
| 65 | terminate_coupled_remote, terminate_run, & |
---|
| 66 | termination_time_needed, time_restart, & |
---|
| 67 | time_since_reference_point, write_binary |
---|
| 68 | USE kinds |
---|
[1] | 69 | USE pegrid |
---|
| 70 | |
---|
| 71 | IMPLICIT NONE |
---|
| 72 | |
---|
| 73 | |
---|
[1468] | 74 | LOGICAL :: terminate_run_l !: |
---|
| 75 | LOGICAL :: do_stop_now = .FALSE. !: |
---|
| 76 | LOGICAL :: do_restart_now = .FALSE. !: |
---|
[1] | 77 | |
---|
[1320] | 78 | REAL(wp) :: remaining_time !: |
---|
[1] | 79 | |
---|
[1320] | 80 | |
---|
[1] | 81 | ! |
---|
| 82 | !-- Check remaining CPU-time |
---|
| 83 | CALL local_tremain( remaining_time ) |
---|
| 84 | |
---|
| 85 | ! |
---|
| 86 | !-- If necessary set a flag to stop the model run |
---|
| 87 | terminate_run_l = .FALSE. |
---|
[1320] | 88 | IF ( remaining_time <= termination_time_needed .AND. & |
---|
[1] | 89 | write_binary(1:4) == 'true' ) THEN |
---|
| 90 | |
---|
| 91 | terminate_run_l = .TRUE. |
---|
| 92 | ENDIF |
---|
| 93 | |
---|
| 94 | #if defined( __parallel ) |
---|
| 95 | ! |
---|
| 96 | !-- Make a logical OR for all processes. Stop the model run if at least |
---|
| 97 | !-- one processor has reached the time limit. |
---|
[622] | 98 | IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr ) |
---|
[1320] | 99 | CALL MPI_ALLREDUCE( terminate_run_l, terminate_run, 1, MPI_LOGICAL, & |
---|
[1] | 100 | MPI_LOR, comm2d, ierr ) |
---|
| 101 | #else |
---|
| 102 | terminate_run = terminate_run_l |
---|
| 103 | #endif |
---|
| 104 | |
---|
| 105 | ! |
---|
| 106 | !-- Output that job will be terminated |
---|
| 107 | IF ( terminate_run .AND. myid == 0 ) THEN |
---|
[1320] | 108 | WRITE( message_string, * ) 'run will be terminated because it is ', & |
---|
| 109 | 'running out of job cpu limit & ', & |
---|
| 110 | 'remaining time: ', remaining_time, ' s', & |
---|
[274] | 111 | 'termination time needed:', termination_time_needed, ' s' |
---|
[247] | 112 | CALL message( 'check_for_restart', 'PA0163', 0, 1, 0, 6, 0 ) |
---|
[1] | 113 | ENDIF |
---|
| 114 | |
---|
| 115 | ! |
---|
[108] | 116 | !-- In case of coupled runs inform the remote model of the termination |
---|
| 117 | !-- and its reason, provided the remote model has not already been |
---|
| 118 | !-- informed of another termination reason (terminate_coupled > 0) before, |
---|
| 119 | !-- or vice versa (terminate_coupled_remote > 0). |
---|
[1320] | 120 | IF ( terminate_run .AND. TRIM( coupling_mode ) /= 'uncoupled' .AND. & |
---|
[110] | 121 | terminate_coupled == 0 .AND. terminate_coupled_remote == 0 ) THEN |
---|
| 122 | |
---|
[108] | 123 | terminate_coupled = 3 |
---|
[1468] | 124 | |
---|
[222] | 125 | #if defined( __parallel ) |
---|
[667] | 126 | IF ( myid == 0 ) THEN |
---|
[1032] | 127 | CALL MPI_SENDRECV( terminate_coupled, 1, MPI_INTEGER, & |
---|
| 128 | target_id, 0, & |
---|
| 129 | terminate_coupled_remote, 1, MPI_INTEGER, & |
---|
| 130 | target_id, 0, & |
---|
[667] | 131 | comm_inter, status, ierr ) |
---|
| 132 | ENDIF |
---|
[1320] | 133 | CALL MPI_BCAST( terminate_coupled_remote, 1, MPI_INTEGER, 0, comm2d, & |
---|
[1032] | 134 | ierr ) |
---|
[222] | 135 | #endif |
---|
[108] | 136 | ENDIF |
---|
| 137 | |
---|
[1468] | 138 | |
---|
[108] | 139 | ! |
---|
[1468] | 140 | !-- Check if a flag file exists that forces a termination of the model |
---|
| 141 | IF ( myid == 0 ) THEN |
---|
| 142 | INQUIRE(FILE="DO_STOP_NOW", EXIST=do_stop_now) |
---|
| 143 | INQUIRE(FILE="DO_RESTART_NOW", EXIST=do_restart_now) |
---|
| 144 | |
---|
| 145 | IF ( do_stop_now .OR. do_restart_now ) THEN |
---|
| 146 | |
---|
| 147 | terminate_run_l = .TRUE. |
---|
| 148 | |
---|
| 149 | WRITE( message_string, * ) 'run will be terminated because user ', & |
---|
| 150 | 'forced a job finialization using a flag', & |
---|
| 151 | 'file:', & |
---|
| 152 | '&DO_STOP_NOW: ', do_stop_now, & |
---|
| 153 | '&DO_RESTART_NOW: ', do_restart_now |
---|
| 154 | CALL message( 'check_for_restart', 'PA0398', 0, 0, 0, 6, 0 ) |
---|
| 155 | |
---|
| 156 | ENDIF |
---|
| 157 | ENDIF |
---|
| 158 | |
---|
| 159 | |
---|
| 160 | #if defined( __parallel ) |
---|
| 161 | ! |
---|
| 162 | !-- Make a logical OR for all processes. Stop the model run if at least |
---|
| 163 | !-- one processor has reached the time limit. |
---|
| 164 | IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr ) |
---|
| 165 | CALL MPI_ALLREDUCE( terminate_run_l, terminate_run, 1, MPI_LOGICAL, & |
---|
| 166 | MPI_LOR, comm2d, ierr ) |
---|
| 167 | #else |
---|
| 168 | terminate_run = terminate_run_l |
---|
| 169 | #endif |
---|
| 170 | |
---|
| 171 | ! |
---|
| 172 | !-- In case of coupled runs inform the remote model of the termination |
---|
| 173 | !-- and its reason, provided the remote model has not already been |
---|
| 174 | !-- informed of another termination reason (terminate_coupled > 0) before, |
---|
| 175 | !-- or vice versa (terminate_coupled_remote > 0). |
---|
| 176 | IF ( terminate_run .AND. coupling_mode /= 'uncoupled' .AND. & |
---|
| 177 | terminate_coupled == 0 .AND. terminate_coupled_remote == 0 ) THEN |
---|
| 178 | |
---|
| 179 | terminate_coupled = 6 |
---|
| 180 | |
---|
| 181 | #if defined( __parallel ) |
---|
| 182 | IF ( myid == 0 ) THEN |
---|
| 183 | CALL MPI_SENDRECV( terminate_coupled, 1, MPI_INTEGER, & |
---|
| 184 | target_id, 0, & |
---|
| 185 | terminate_coupled_remote, 1, MPI_INTEGER, & |
---|
| 186 | target_id, 0, & |
---|
| 187 | comm_inter, status, ierr ) |
---|
| 188 | ENDIF |
---|
| 189 | CALL MPI_BCAST( terminate_coupled_remote, 1, MPI_INTEGER, 0, & |
---|
| 190 | comm2d, ierr ) |
---|
| 191 | #endif |
---|
| 192 | |
---|
| 193 | ENDIF |
---|
| 194 | |
---|
| 195 | ! |
---|
| 196 | !-- Set the stop flag also, if restart is forced by user settings |
---|
| 197 | IF ( time_restart /= 9999999.9_wp .AND. & |
---|
[291] | 198 | time_restart < time_since_reference_point ) THEN |
---|
| 199 | |
---|
[1] | 200 | ! |
---|
| 201 | !-- Restart is not neccessary, if the end time of the run (given by |
---|
| 202 | !-- the user) has been reached |
---|
| 203 | IF ( simulated_time < end_time ) THEN |
---|
| 204 | terminate_run = .TRUE. |
---|
| 205 | ! |
---|
| 206 | !-- Increment restart time, if forced by user, otherwise set restart |
---|
| 207 | !-- time to default (no user restart) |
---|
[1353] | 208 | IF ( dt_restart /= 9999999.9_wp ) THEN |
---|
[1] | 209 | time_restart = time_restart + dt_restart |
---|
| 210 | ELSE |
---|
[1353] | 211 | time_restart = 9999999.9_wp |
---|
[1] | 212 | ENDIF |
---|
| 213 | |
---|
[1320] | 214 | WRITE( message_string, * ) 'run will be terminated due to user ', & |
---|
| 215 | 'settings of', & |
---|
| 216 | '&restart_time / dt_restart', & |
---|
[274] | 217 | '&new restart time is: ', time_restart, ' s' |
---|
| 218 | CALL message( 'check_for_restart', 'PA0164', 0, 0, 0, 6, 0 ) |
---|
[247] | 219 | |
---|
[108] | 220 | ! |
---|
| 221 | !-- In case of coupled runs inform the remote model of the termination |
---|
| 222 | !-- and its reason, provided the remote model has not already been |
---|
| 223 | !-- informed of another termination reason (terminate_coupled > 0) before, |
---|
| 224 | !-- or vice versa (terminate_coupled_remote > 0). |
---|
[1320] | 225 | IF ( coupling_mode /= 'uncoupled' .AND. terminate_coupled == 0 & |
---|
[206] | 226 | .AND. terminate_coupled_remote == 0 ) THEN |
---|
[110] | 227 | |
---|
[1353] | 228 | IF ( dt_restart /= 9999999.9_wp ) THEN |
---|
[108] | 229 | terminate_coupled = 4 |
---|
| 230 | ELSE |
---|
| 231 | terminate_coupled = 5 |
---|
| 232 | ENDIF |
---|
[222] | 233 | #if defined( __parallel ) |
---|
[667] | 234 | IF ( myid == 0 ) THEN |
---|
[1032] | 235 | CALL MPI_SENDRECV( terminate_coupled, 1, MPI_INTEGER, & |
---|
| 236 | target_id, 0, & |
---|
| 237 | terminate_coupled_remote, 1, MPI_INTEGER, & |
---|
| 238 | target_id, 0, & |
---|
[667] | 239 | comm_inter, status, ierr ) |
---|
| 240 | ENDIF |
---|
[1320] | 241 | CALL MPI_BCAST( terminate_coupled_remote, 1, MPI_INTEGER, 0, & |
---|
[1032] | 242 | comm2d, ierr ) |
---|
[222] | 243 | #endif |
---|
[108] | 244 | ENDIF |
---|
| 245 | ELSE |
---|
[1353] | 246 | time_restart = 9999999.9_wp |
---|
[1] | 247 | ENDIF |
---|
| 248 | ENDIF |
---|
| 249 | |
---|
| 250 | ! |
---|
| 251 | !-- If the run is stopped, set a flag file which is necessary to initiate |
---|
[1468] | 252 | !-- the start of a continuation run, except if the user forced to stop the |
---|
| 253 | !-- run without restart |
---|
| 254 | IF ( terminate_run .AND. myid == 0 .AND. .NOT. do_stop_now) THEN |
---|
[1] | 255 | |
---|
| 256 | OPEN ( 90, FILE='CONTINUE_RUN', FORM='FORMATTED' ) |
---|
| 257 | WRITE ( 90, '(A)' ) TRIM( run_description_header ) |
---|
| 258 | CLOSE ( 90 ) |
---|
| 259 | |
---|
| 260 | ENDIF |
---|
| 261 | |
---|
| 262 | |
---|
| 263 | END SUBROUTINE check_for_restart |
---|