| 1 | !> @file check_for_restart.f90 |
|---|
| 2 | !--------------------------------------------------------------------------------! |
|---|
| 3 | ! This file is part of PALM. |
|---|
| 4 | ! |
|---|
| 5 | ! PALM is free software: you can redistribute it and/or modify it under the terms |
|---|
| 6 | ! of the GNU General Public License as published by the Free Software Foundation, |
|---|
| 7 | ! either version 3 of the License, or (at your option) any later version. |
|---|
| 8 | ! |
|---|
| 9 | ! PALM is distributed in the hope that it will be useful, but WITHOUT ANY |
|---|
| 10 | ! WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR |
|---|
| 11 | ! A PARTICULAR PURPOSE. See the GNU General Public License for more details. |
|---|
| 12 | ! |
|---|
| 13 | ! You should have received a copy of the GNU General Public License along with |
|---|
| 14 | ! PALM. If not, see <http://www.gnu.org/licenses/>. |
|---|
| 15 | ! |
|---|
| 16 | ! Copyright 1997-2014 Leibniz Universitaet Hannover |
|---|
| 17 | !--------------------------------------------------------------------------------! |
|---|
| 18 | ! |
|---|
| 19 | ! Current revisions: |
|---|
| 20 | ! ----------------- |
|---|
| 21 | ! |
|---|
| 22 | ! |
|---|
| 23 | ! Former revisions: |
|---|
| 24 | ! ----------------- |
|---|
| 25 | ! $Id: check_for_restart.f90 1683 2015-10-07 23:57:51Z gronemeier $ |
|---|
| 26 | ! |
|---|
| 27 | ! 1682 2015-10-07 23:56:08Z knoop |
|---|
| 28 | ! Code annotations made doxygen readable |
|---|
| 29 | ! |
|---|
| 30 | ! 1509 2014-12-16 08:56:46Z heinze |
|---|
| 31 | ! bugfix: prevent infinite loop in case of automatic restarts |
|---|
| 32 | ! |
|---|
| 33 | ! 1468 2014-09-24 14:06:57Z maronga |
|---|
| 34 | ! Added support for unscheduled job termination using the flag files |
|---|
| 35 | ! DO_STOP_NOW and DO_RESTART_NOW |
|---|
| 36 | ! |
|---|
| 37 | ! 1353 2014-04-08 15:21:23Z heinze |
|---|
| 38 | ! REAL constants provided with KIND-attribute |
|---|
| 39 | ! |
|---|
| 40 | ! 1320 2014-03-20 08:40:49Z raasch |
|---|
| 41 | ! ONLY-attribute added to USE-statements, |
|---|
| 42 | ! kind-parameters added to all INTEGER and REAL declaration statements, |
|---|
| 43 | ! kinds are defined in new module kinds, |
|---|
| 44 | ! revision history before 2012 removed, |
|---|
| 45 | ! comment fields (!:) to be used for variable explanations added to |
|---|
| 46 | ! all variable declaration statements |
|---|
| 47 | ! |
|---|
| 48 | ! 1036 2012-10-22 13:43:42Z raasch |
|---|
| 49 | ! code put under GPL (PALM 3.9) |
|---|
| 50 | ! |
|---|
| 51 | ! 1032 2012-10-21 13:03:21Z letzel |
|---|
| 52 | ! minor reformatting |
|---|
| 53 | ! |
|---|
| 54 | ! Revision 1.1 1998/03/18 20:06:51 raasch |
|---|
| 55 | ! Initial revision |
|---|
| 56 | ! |
|---|
| 57 | ! |
|---|
| 58 | ! Description: |
|---|
| 59 | ! ------------ |
|---|
| 60 | !> Set stop flag, if restart is neccessary because of expiring cpu-time or |
|---|
| 61 | !> if it is forced by user |
|---|
| 62 | !------------------------------------------------------------------------------! |
|---|
| 63 | SUBROUTINE check_for_restart |
|---|
| 64 | |
|---|
| 65 | |
|---|
| 66 | USE control_parameters, & |
|---|
| 67 | ONLY: coupling_mode, dt_restart, end_time, message_string, & |
|---|
| 68 | run_description_header, simulated_time, terminate_coupled, & |
|---|
| 69 | terminate_coupled_remote, terminate_run, & |
|---|
| 70 | termination_time_needed, time_restart, & |
|---|
| 71 | time_since_reference_point, write_binary |
|---|
| 72 | USE kinds |
|---|
| 73 | USE pegrid |
|---|
| 74 | |
|---|
| 75 | IMPLICIT NONE |
|---|
| 76 | |
|---|
| 77 | |
|---|
| 78 | LOGICAL :: terminate_run_l !< |
|---|
| 79 | LOGICAL :: do_stop_now = .FALSE. !< |
|---|
| 80 | LOGICAL :: do_restart_now = .FALSE. !< |
|---|
| 81 | |
|---|
| 82 | REAL(wp) :: remaining_time !< |
|---|
| 83 | |
|---|
| 84 | |
|---|
| 85 | ! |
|---|
| 86 | !-- Check remaining CPU-time |
|---|
| 87 | CALL local_tremain( remaining_time ) |
|---|
| 88 | |
|---|
| 89 | ! |
|---|
| 90 | !-- If necessary set a flag to stop the model run |
|---|
| 91 | terminate_run_l = .FALSE. |
|---|
| 92 | IF ( remaining_time <= termination_time_needed .AND. & |
|---|
| 93 | write_binary(1:4) == 'true' ) THEN |
|---|
| 94 | |
|---|
| 95 | terminate_run_l = .TRUE. |
|---|
| 96 | ENDIF |
|---|
| 97 | |
|---|
| 98 | #if defined( __parallel ) |
|---|
| 99 | ! |
|---|
| 100 | !-- Make a logical OR for all processes. Stop the model run if at least |
|---|
| 101 | !-- one processor has reached the time limit. |
|---|
| 102 | IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr ) |
|---|
| 103 | CALL MPI_ALLREDUCE( terminate_run_l, terminate_run, 1, MPI_LOGICAL, & |
|---|
| 104 | MPI_LOR, comm2d, ierr ) |
|---|
| 105 | #else |
|---|
| 106 | terminate_run = terminate_run_l |
|---|
| 107 | #endif |
|---|
| 108 | |
|---|
| 109 | ! |
|---|
| 110 | !-- Output that job will be terminated |
|---|
| 111 | IF ( terminate_run .AND. myid == 0 ) THEN |
|---|
| 112 | WRITE( message_string, * ) 'run will be terminated because it is ', & |
|---|
| 113 | 'running out of job cpu limit & ', & |
|---|
| 114 | 'remaining time: ', remaining_time, ' s', & |
|---|
| 115 | 'termination time needed:', termination_time_needed, ' s' |
|---|
| 116 | CALL message( 'check_for_restart', 'PA0163', 0, 1, 0, 6, 0 ) |
|---|
| 117 | ENDIF |
|---|
| 118 | |
|---|
| 119 | ! |
|---|
| 120 | !-- In case of coupled runs inform the remote model of the termination |
|---|
| 121 | !-- and its reason, provided the remote model has not already been |
|---|
| 122 | !-- informed of another termination reason (terminate_coupled > 0) before, |
|---|
| 123 | !-- or vice versa (terminate_coupled_remote > 0). |
|---|
| 124 | IF ( terminate_run .AND. TRIM( coupling_mode ) /= 'uncoupled' .AND. & |
|---|
| 125 | terminate_coupled == 0 .AND. terminate_coupled_remote == 0 ) THEN |
|---|
| 126 | |
|---|
| 127 | terminate_coupled = 3 |
|---|
| 128 | |
|---|
| 129 | #if defined( __parallel ) |
|---|
| 130 | IF ( myid == 0 ) THEN |
|---|
| 131 | CALL MPI_SENDRECV( terminate_coupled, 1, MPI_INTEGER, & |
|---|
| 132 | target_id, 0, & |
|---|
| 133 | terminate_coupled_remote, 1, MPI_INTEGER, & |
|---|
| 134 | target_id, 0, & |
|---|
| 135 | comm_inter, status, ierr ) |
|---|
| 136 | ENDIF |
|---|
| 137 | CALL MPI_BCAST( terminate_coupled_remote, 1, MPI_INTEGER, 0, comm2d, & |
|---|
| 138 | ierr ) |
|---|
| 139 | #endif |
|---|
| 140 | ENDIF |
|---|
| 141 | |
|---|
| 142 | |
|---|
| 143 | ! |
|---|
| 144 | !-- Check if a flag file exists that forces a termination of the model |
|---|
| 145 | IF ( myid == 0 ) THEN |
|---|
| 146 | INQUIRE(FILE="DO_STOP_NOW", EXIST=do_stop_now) |
|---|
| 147 | INQUIRE(FILE="DO_RESTART_NOW", EXIST=do_restart_now) |
|---|
| 148 | |
|---|
| 149 | IF ( do_stop_now .OR. do_restart_now ) THEN |
|---|
| 150 | |
|---|
| 151 | terminate_run_l = .TRUE. |
|---|
| 152 | |
|---|
| 153 | WRITE( message_string, * ) 'run will be terminated because user ', & |
|---|
| 154 | 'forced a job finialization using a flag', & |
|---|
| 155 | 'file:', & |
|---|
| 156 | '&DO_STOP_NOW: ', do_stop_now, & |
|---|
| 157 | '&DO_RESTART_NOW: ', do_restart_now |
|---|
| 158 | CALL message( 'check_for_restart', 'PA0398', 0, 0, 0, 6, 0 ) |
|---|
| 159 | |
|---|
| 160 | ENDIF |
|---|
| 161 | ENDIF |
|---|
| 162 | |
|---|
| 163 | |
|---|
| 164 | #if defined( __parallel ) |
|---|
| 165 | ! |
|---|
| 166 | !-- Make a logical OR for all processes. Stop the model run if at least |
|---|
| 167 | !-- one processor has reached the time limit. |
|---|
| 168 | IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr ) |
|---|
| 169 | CALL MPI_ALLREDUCE( terminate_run_l, terminate_run, 1, MPI_LOGICAL, & |
|---|
| 170 | MPI_LOR, comm2d, ierr ) |
|---|
| 171 | #else |
|---|
| 172 | terminate_run = terminate_run_l |
|---|
| 173 | #endif |
|---|
| 174 | |
|---|
| 175 | ! |
|---|
| 176 | !-- In case of coupled runs inform the remote model of the termination |
|---|
| 177 | !-- and its reason, provided the remote model has not already been |
|---|
| 178 | !-- informed of another termination reason (terminate_coupled > 0) before, |
|---|
| 179 | !-- or vice versa (terminate_coupled_remote > 0). |
|---|
| 180 | IF ( terminate_run .AND. coupling_mode /= 'uncoupled' .AND. & |
|---|
| 181 | terminate_coupled == 0 .AND. terminate_coupled_remote == 0 ) THEN |
|---|
| 182 | |
|---|
| 183 | terminate_coupled = 6 |
|---|
| 184 | |
|---|
| 185 | #if defined( __parallel ) |
|---|
| 186 | IF ( myid == 0 ) THEN |
|---|
| 187 | CALL MPI_SENDRECV( terminate_coupled, 1, MPI_INTEGER, & |
|---|
| 188 | target_id, 0, & |
|---|
| 189 | terminate_coupled_remote, 1, MPI_INTEGER, & |
|---|
| 190 | target_id, 0, & |
|---|
| 191 | comm_inter, status, ierr ) |
|---|
| 192 | ENDIF |
|---|
| 193 | CALL MPI_BCAST( terminate_coupled_remote, 1, MPI_INTEGER, 0, & |
|---|
| 194 | comm2d, ierr ) |
|---|
| 195 | #endif |
|---|
| 196 | |
|---|
| 197 | ENDIF |
|---|
| 198 | |
|---|
| 199 | ! |
|---|
| 200 | !-- Set the stop flag also, if restart is forced by user settings |
|---|
| 201 | IF ( time_restart /= 9999999.9_wp .AND. & |
|---|
| 202 | time_restart < time_since_reference_point ) THEN |
|---|
| 203 | |
|---|
| 204 | ! |
|---|
| 205 | !-- Restart is not neccessary, if the end time of the run (given by |
|---|
| 206 | !-- the user) has been reached |
|---|
| 207 | IF ( simulated_time < end_time ) THEN |
|---|
| 208 | terminate_run = .TRUE. |
|---|
| 209 | ! |
|---|
| 210 | !-- Increment restart time, if forced by user, otherwise set restart |
|---|
| 211 | !-- time to default (no user restart) |
|---|
| 212 | IF ( dt_restart /= 9999999.9_wp ) THEN |
|---|
| 213 | time_restart = time_restart + dt_restart |
|---|
| 214 | ELSE |
|---|
| 215 | time_restart = 9999999.9_wp |
|---|
| 216 | ENDIF |
|---|
| 217 | |
|---|
| 218 | WRITE( message_string, * ) 'run will be terminated due to user ', & |
|---|
| 219 | 'settings of', & |
|---|
| 220 | '&restart_time / dt_restart', & |
|---|
| 221 | '&new restart time is: ', time_restart, ' s' |
|---|
| 222 | CALL message( 'check_for_restart', 'PA0164', 0, 0, 0, 6, 0 ) |
|---|
| 223 | |
|---|
| 224 | ! |
|---|
| 225 | !-- In case of coupled runs inform the remote model of the termination |
|---|
| 226 | !-- and its reason, provided the remote model has not already been |
|---|
| 227 | !-- informed of another termination reason (terminate_coupled > 0) before, |
|---|
| 228 | !-- or vice versa (terminate_coupled_remote > 0). |
|---|
| 229 | IF ( coupling_mode /= 'uncoupled' .AND. terminate_coupled == 0 & |
|---|
| 230 | .AND. terminate_coupled_remote == 0 ) THEN |
|---|
| 231 | |
|---|
| 232 | IF ( dt_restart /= 9999999.9_wp ) THEN |
|---|
| 233 | terminate_coupled = 4 |
|---|
| 234 | ELSE |
|---|
| 235 | terminate_coupled = 5 |
|---|
| 236 | ENDIF |
|---|
| 237 | #if defined( __parallel ) |
|---|
| 238 | IF ( myid == 0 ) THEN |
|---|
| 239 | CALL MPI_SENDRECV( terminate_coupled, 1, MPI_INTEGER, & |
|---|
| 240 | target_id, 0, & |
|---|
| 241 | terminate_coupled_remote, 1, MPI_INTEGER, & |
|---|
| 242 | target_id, 0, & |
|---|
| 243 | comm_inter, status, ierr ) |
|---|
| 244 | ENDIF |
|---|
| 245 | CALL MPI_BCAST( terminate_coupled_remote, 1, MPI_INTEGER, 0, & |
|---|
| 246 | comm2d, ierr ) |
|---|
| 247 | #endif |
|---|
| 248 | ENDIF |
|---|
| 249 | ELSE |
|---|
| 250 | time_restart = 9999999.9_wp |
|---|
| 251 | ENDIF |
|---|
| 252 | ENDIF |
|---|
| 253 | |
|---|
| 254 | ! |
|---|
| 255 | !-- If the run is stopped, set a flag file which is necessary to initiate |
|---|
| 256 | !-- the start of a continuation run, except if the user forced to stop the |
|---|
| 257 | !-- run without restart |
|---|
| 258 | IF ( terminate_run .AND. myid == 0 .AND. .NOT. do_stop_now) THEN |
|---|
| 259 | |
|---|
| 260 | OPEN ( 90, FILE='CONTINUE_RUN', FORM='FORMATTED' ) |
|---|
| 261 | WRITE ( 90, '(A)' ) TRIM( run_description_header ) |
|---|
| 262 | CLOSE ( 90 ) |
|---|
| 263 | |
|---|
| 264 | ENDIF |
|---|
| 265 | |
|---|
| 266 | |
|---|
| 267 | END SUBROUTINE check_for_restart |
|---|
