Context Navigation

mpi_mod.f90 @ f55fdce

10.4.1_peseiGFS_025bugfixes+enhancementsdevrelease-10release-10.4.1scaling-bugunivie

Last change on this file since f55fdce was f55fdce, checked in by Espen Sollum ATMOS <eso@…>, 9 years ago
Added a modification so thah number of particles in simulation reported to standard output (MPI version) is exact instead of estimated
Property mode set to `100644`
File size: 76.7 KB

Line
1	!**********************************************************************
2	! Copyright 1998,1999,2000,2001,2002,2005,2007,2008,2009,2010 *
3	! Andreas Stohl, Petra Seibert, A. Frank, Gerhard Wotawa, *
4	! Caroline Forster, Sabine Eckhardt, John Burkhart, Harald Sodemann *
5	! *
6	! This file is part of FLEXPART. *
7	! *
8	! FLEXPART is free software: you can redistribute it and/or modify *
9	! it under the terms of the GNU General Public License as published by*
10	! the Free Software Foundation, either version 3 of the License, or *
11	! (at your option) any later version. *
12	! *
13	! FLEXPART is distributed in the hope that it will be useful, *
14	! but WITHOUT ANY WARRANTY; without even the implied warranty of *
15	! MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the *
16	! GNU General Public License for more details. *
17	! *
18	! You should have received a copy of the GNU General Public License *
19	! along with FLEXPART. If not, see <http://www.gnu.org/licenses/>. *
20	!**********************************************************************
21
22	module mpi_mod
23
24	!*****************************************************************************
25	! *
26	! DESCRIPTION *
27	! This module contains subroutines and common variables used for the *
28	! MPI parallelization of FLEXPART. *
29	! *
30	! NOTE *
31	! Depending on the MPI library installed on your system (e.g. mpich2, *
32	! OpenMPI) you may need to choose below in this file between *
33	! use mpi *
34	! (if the MPI library comes with the file 'mpi.mod'); or *
35	! include 'mpif.h' *
36	! *
37	! *
38	!*****************************************************************************
39	! *
40	! Variables: *
41	! *
42	! mp_ierr MPI error code *
43	! mp_np Number of MPI processes *
44	! mp_pid Process ID of each MPI process *
45	! mp_seed Parameter for random number seed *
46	! read_grp_min Minimum number of processes at which one will be *
47	! used as reader *
48	! numpart_mpi, Number of particles per node *
49	! maxpart_mpi *
50	! mp_partid MPI process ID for particle calculation *
51	! mp_partgroup_ Refers to the subset of processors performing *
52	! loops over particles. Will be all processes *
53	! unless a dedicated process runs getfields/readwind *
54	! lmp_sync If .false., use asynchronous MPI *
55	! *
56	! *
57	! *
58	! *
59	!*****************************************************************************
60
61	use mpi
62	use par_mod, only: dp, sp
63	use com_mod, only: lroot
64
65	implicit none
66
67	! include 'mpif.h'
68
69	public
70
71	! Set aside a process for reading windfields if using at least these many processes
72	!==================================================
73	integer, parameter, private :: read_grp_min=4
74	!==================================================
75
76	! Variables for each MPI process in the world group
77	integer :: mp_ierr, mp_np, mp_pid, mp_partid
78	integer, private :: world_group_id
79
80	! Variables for MPI processes in the 'particle' group
81	integer, allocatable, dimension(:) :: mp_partgroup_rank
82	integer :: mp_partgroup_comm, mp_partgroup_pid, mp_partgroup_np
83
84	integer :: mp_seed=0
85	integer, parameter :: mp_sp=MPI_REAL4, mp_dp=MPI_REAL8
86	integer, parameter :: mp_pp=mp_sp
87	integer, parameter :: id_root=0 ! master process
88
89	! MPI tags/requests for send/receive operation
90	integer :: tm1
91	integer, parameter :: nvar_async=27 !29 :DBG:
92	!integer, dimension(:), allocatable :: tags
93	integer, dimension(:), allocatable :: reqs
94
95
96	integer :: id_read ! readwind/getfield process
97	integer :: numpart_mpi,maxpart_mpi ! number of particles per node
98	integer :: tot_numpart=0
99	integer :: mp_comm_used ! global or subgroup communicator
100
101	logical :: lmpreader=.false. ! is set to true for reading process(es) only.
102	logical :: lmp_use_reader=.false. ! true if separate readwind process is used
103
104	! true if only using synchronous MPI send/recv:
105	! If setting this to .false., numwfmem must be set to 3
106	!===============================================================================
107	logical :: lmp_sync=.true.
108	!===============================================================================
109
110	! mp_dbg_mode Used for debugging MPI.
111	! mp_dev_mode various checks related to debugging the parallel code
112	! mp_dbg_out write some arrays to data file for debugging
113	! mp_measure_time Measure cpu/wall time, write out at end of run
114	! mp_time_barrier Measure MPI barrier time
115	! mp_exact_numpart Use an extra MPI communication to give the exact number of particles
116	! to standard output (this does not otherwise affect the simulation)
117	logical, parameter :: mp_dbg_mode = .false.
118	logical, parameter :: mp_dev_mode = .false.
119	logical, parameter :: mp_dbg_out = .false.
120	logical, parameter :: mp_time_barrier=.true.
121	logical, parameter :: mp_measure_time=.false.
122	logical, parameter :: mp_exact_numpart=.true.
123
124	! for measuring CPU/Wall time
125	real(sp) :: mp_comm_time_beg, mp_comm_time_end, mp_comm_time_total=0.
126	real(dp) :: mp_comm_wtime_beg, mp_comm_wtime_end, mp_comm_wtime_total=0.
127	real(sp) :: mp_root_time_beg, mp_root_time_end, mp_root_time_total=0.
128	real(dp) :: mp_root_wtime_beg, mp_root_wtime_end, mp_root_wtime_total=0.
129	real(sp) :: mp_barrier_time_beg, mp_barrier_time_end, mp_barrier_time_total=0.
130	real(dp) :: mp_barrier_wtime_beg, mp_barrier_wtime_end, mp_barrier_wtime_total=0.
131	real(sp) :: tm_nploop_beg, tm_nploop_end, tm_nploop_total=0.
132	real(sp) :: tm_tot_beg, tm_tot_end, tm_tot_total=0.
133	real(dp) :: mp_getfields_wtime_beg, mp_getfields_wtime_end, mp_getfields_wtime_total=0.
134	real(sp) :: mp_getfields_time_beg, mp_getfields_time_end, mp_getfields_time_total=0.
135	real(dp) :: mp_readwind_wtime_beg, mp_readwind_wtime_end, mp_readwind_wtime_total=0.
136	real(sp) :: mp_readwind_time_beg, mp_readwind_time_end, mp_readwind_time_total=0.
137	real(dp) :: mp_io_wtime_beg, mp_io_wtime_end, mp_io_wtime_total=0.
138	real(sp) :: mp_io_time_beg, mp_io_time_end, mp_io_time_total=0.
139	real(dp) :: mp_wetdepo_wtime_beg, mp_wetdepo_wtime_end, mp_wetdepo_wtime_total=0.
140	real(sp) :: mp_wetdepo_time_beg, mp_wetdepo_time_end, mp_wetdepo_time_total=0.
141	real(dp) :: mp_advance_wtime_beg, mp_advance_wtime_end, mp_advance_wtime_total=0.
142	real(dp) :: mp_conccalc_time_beg, mp_conccalc_time_end, mp_conccalc_time_total=0.
143	real(dp) :: mp_total_wtime_beg, mp_total_wtime_end, mp_total_wtime_total=0.
144	real(dp) :: mp_vt_wtime_beg, mp_vt_wtime_end, mp_vt_wtime_total
145	real(sp) :: mp_vt_time_beg, mp_vt_time_end, mp_vt_time_total
146
147	! dat_lun logical unit number for i/o
148	integer, private :: dat_lun
149
150	contains
151
152	subroutine mpif_init
153	!***********************************************************************
154	! DESCRIPTION
155	! Initialize MPI.
156	!
157	! Create the global communicator MPI_COMM_WORLD
158	! If using a separate MPI process for getfields/readwind, a subgroup
159	! is created for the other processes.
160	!
161	! VARIABLES
162	! mpi_mode default 0, set to 2/3 if running MPI version
163	! mp_np number of running processes, decided at run-time
164	!***********************************************************************
165	use par_mod, only: maxpart, numwfmem
166	use com_mod, only: mpi_mode
167
168	implicit none
169
170	integer :: i,j,s,addmaxpart=0
171
172	! each process gets an ID (mp_pid) in the range 0,..,mp_np-1
173	call MPI_INIT(mp_ierr)
174	if (mp_ierr /= 0) goto 100
175	call MPI_COMM_RANK(MPI_COMM_WORLD, mp_pid, mp_ierr)
176	if (mp_ierr /= 0) goto 100
177	call MPI_COMM_SIZE(MPI_COMM_WORLD, mp_np, mp_ierr)
178	if (mp_ierr /= 0) goto 100
179
180
181	! this variable is used to handle subroutines common to parallel/serial version
182	if (lmp_sync) then
183	mpi_mode=2 ! hold 2 windfields in memory
184	else
185	mpi_mode=3 ! hold 3 windfields in memory
186	end if
187
188	if (mp_pid.ne.0) then
189	lroot = .false.
190	end if
191
192	! Check for sensible combination of parameters
193	!*********************************************
194
195	if (.not.lmp_sync.and.numwfmem.ne.3.and.lroot) then
196	write(*,FMT='(80("#"))')
197	write(,) '#### mpi_mod::mpif_init> ERROR: ', &
198	& 'numwfmem must be set to 3 for asyncronous reading ####'
199	write(*,FMT='(80("#"))')
200	stop
201	else if (lmp_sync.and.numwfmem.ne.2.and.lroot) then
202	write(*,FMT='(80("#"))')
203	write(,) '#### mpi_mod::mpif_init> WARNING: ', &
204	& 'numwfmem should be set to 2 for syncronous reading'
205	write(*,FMT='(80("#"))')
206	! Force "syncronized" version if all processes will call getfields
207	else if (.not.lmp_sync.and.mp_np.lt.read_grp_min) then
208	if (lroot) then
209	write(*,FMT='(80("#"))')
210	write(,) '#### mpi_mod::mpif_init> WARNING: ', &
211	& 'all procs call getfields. Setting lmp_sync=.true.'
212	write(*,FMT='(80("#"))')
213	end if
214	lmp_sync=.true. ! :DBG: eso fix this...
215	end if
216	! TODO: Add warnings for unimplemented flexpart features
217
218	! Set ID of process that calls getfield/readwind.
219	! Using the last process in the group ensures statistical identical results
220	! as running with one process less but not using separate read process
221	!**********************************************************************
222
223	! id_read = min(mp_np-1, 1)
224	id_read = mp_np-1
225
226	if (mp_pid.eq.id_read) lmpreader=.true.
227
228	call MPI_Comm_group (MPI_COMM_WORLD, world_group_id, mp_ierr)
229
230	! Create a MPI group/communicator that will calculate trajectories.
231	! Skip this step if program is run with only a few processes
232	!************************************************************************
233	allocate(mp_partgroup_rank(0:mp_np-2))
234
235	! This allows checking for allocation of arrays when no subgroup is used
236	mp_partgroup_pid=0
237
238	if (read_grp_min.lt.2) then
239	write(,) '#### mpi_mod::mpif_init> ERROR ####', &
240	& 'read_grp_min should be at least 2. Exiting'
241	stop
242	end if
243
244	if (mp_np.ge.read_grp_min) then
245	lmp_use_reader = .true.
246
247	! Map the subgroup IDs= 0,1,2,...,mp_np-2, skipping reader process
248	j=-1
249	do i=0, mp_np-2 !loop over all (subgroup) IDs
250	j=j+1
251	if (i.eq.id_read) then
252	j=j+1
253	end if
254	mp_partgroup_rank(i) = j
255	end do
256
257	call MPI_Group_incl (world_group_id, mp_np-1, mp_partgroup_rank, &
258	&mp_partgroup_pid, mp_ierr)
259	if (mp_ierr /= 0) goto 100
260	call MPI_Comm_create (MPI_COMM_WORLD, mp_partgroup_pid, mp_partgroup_comm, mp_ierr)
261	if (mp_ierr /= 0) goto 100
262
263	if (mp_pid.ne.id_read) then
264	call MPI_Comm_rank (mp_partgroup_comm, mp_partgroup_pid, mp_ierr)
265	if (mp_ierr /= 0) goto 100
266
267	! Get size of new group (=mp_np-1)
268	call MPI_COMM_SIZE(mp_partgroup_comm, mp_partgroup_np, mp_ierr)
269	if (mp_ierr /= 0) goto 100
270	if (mp_partgroup_np.ne.mp_np-1) then
271	write(,) '#### mpi_mod:: mpif_init> ERROR ####&
272	& mp_partgroup_np.ne.mp_np-1'
273	stop
274	endif
275
276	else
277	mp_partgroup_pid = -1
278	end if
279	end if
280
281	if (lmp_use_reader) then
282	mp_comm_used = mp_partgroup_comm
283	mp_partid = mp_partgroup_pid
284	mp_partgroup_np=mp_np-1
285	else
286	mp_comm_used = MPI_COMM_WORLD
287	mp_partgroup_np = mp_np
288	mp_partid = mp_pid
289	end if
290
291	! Set maxpart per process
292	if (mp_partid.lt.mod(maxpart,mp_partgroup_np)) addmaxpart=1
293	maxpart_mpi=int(maxpart/mp_partgroup_np)+addmaxpart
294
295	! Do not allocate particle data arrays for readwind process
296	if (lmpreader.and.lmp_use_reader) then
297	maxpart_mpi=0
298	end if
299
300	! Set random seed for each non-root process
301	if (mp_pid.gt.0) then
302	! if (mp_pid.ge.0) then
303	! call system_clock(s)
304	s = 244
305	mp_seed = -abs(mod((s181)((mp_pid-83)*359), 104729))
306	end if
307	if (mp_dev_mode) write(,) 'PID, mp_seed: ',mp_pid, mp_seed
308	if (mp_dbg_mode) then
309	! :DBG: For debugging, set all seed to 0 and maxrand to e.g. 20
310	mp_seed=0
311	if (lroot) write(,) 'MPI: setting seed=0'
312	end if
313
314	! Allocate request array for asynchronous MPI
315	if (.not.lmp_sync) then
316	allocate(reqs(0:nvar_async*mp_np-1))
317	reqs(:)=MPI_REQUEST_NULL
318	else
319	allocate(reqs(0:1))
320	reqs(:)=MPI_REQUEST_NULL
321	end if
322
323	goto 101
324
325	100 write(,) '#### mpi_mod::mpif_init> ERROR ####', mp_ierr
326	stop
327
328	101 end subroutine mpif_init
329
330
331	subroutine mpif_mpi_barrier
332	!***********************************************************************
333	! Set MPI barrier (all processes are synchronized here), with the
334	! possible exception of a separate 'readwind' process.
335	! Optionally measure cpu/wall time.
336	!
337	!***********************************************************************
338	implicit none
339
340	if (mp_time_barrier) then
341	call cpu_time(mp_barrier_time_beg)
342	mp_barrier_wtime_beg = mpi_wtime()
343	endif
344
345	call MPI_BARRIER(mp_comm_used, mp_ierr)
346	if (mp_ierr /= 0) goto 100
347
348	if (mp_time_barrier) then
349	call cpu_time(mp_barrier_time_end)
350	mp_barrier_wtime_end = mpi_wtime()
351
352	mp_barrier_time_total=mp_barrier_time_total+&
353	&(mp_barrier_time_end-mp_barrier_time_beg)
354	mp_barrier_wtime_total=mp_barrier_wtime_total+&
355	&(mp_barrier_wtime_end-mp_barrier_wtime_beg)
356	end if
357
358	goto 101
359
360	100 write(,) '#### mpi_mod::mpif_mpi_barrier> ERROR ####', mp_ierr
361	stop
362
363	101 end subroutine mpif_mpi_barrier
364
365
366	subroutine mpif_com_mod_allocate
367	!*******************************************************************************
368	! Dynamic allocation of arrays (in serial code these have size maxpart)
369	!
370	!*******************************************************************************
371	use com_mod
372
373	implicit none
374
375	integer :: array_size
376
377	array_size = maxpart_mpi
378
379	allocate(itra1(array_size),npoint(array_size),&
380	& nclass(array_size),idt(array_size),&
381	& itramem(array_size),itrasplit(array_size),&
382	& xtra1(array_size),ytra1(array_size),&
383	& ztra1(array_size),xmass1(array_size, maxspec))
384
385	allocate(uap(array_size),ucp(array_size),&
386	& uzp(array_size),us(array_size),&
387	& vs(array_size),&
388	& ws(array_size),cbt(array_size))
389
390
391	end subroutine mpif_com_mod_allocate
392
393
394	subroutine mpif_tm_send_dims
395	!***********************************************************************
396	! Distribute array dimensions from pid0 to all processes.
397	! numpart_mpi/numpart is sent to allow dynamic allocation
398	!
399	! Currently not in use.
400	!
401	! Variables of similar type (integer, real) are placed in an array
402	! and sent collectively, to avoid the overhead associated with individual
403	! MPI send/receives
404	!
405	!
406	!***********************************************************************
407	use com_mod
408
409	implicit none
410
411	integer :: add_part=0
412
413	call MPI_Bcast(numpart, 1, MPI_INTEGER, id_root, mp_comm_used, mp_ierr)
414
415	! MPI subgroup does the particle-loop
416	if (lmp_use_reader) then
417	if (mod(numpart,mp_partgroup_np).ne.0) add_part=1
418	numpart_mpi=int(numpart/mp_partgroup_np)+add_part
419	else
420	if (mod(numpart,mp_np).ne.0) add_part=1
421	numpart_mpi=int(numpart/mp_np)+add_part
422	end if
423
424
425	! redefine numpart as 'numpart per process' throughout the code
426	!**************************************************************
427	numpart = numpart_mpi
428
429	end subroutine mpif_tm_send_dims
430
431
432	subroutine mpif_tm_send_vars
433	!***********************************************************************
434	! Distribute particle variables from pid0 to all processes.
435	! Called from timemanager
436	!
437	!
438	!***********************************************************************
439	use com_mod
440
441	implicit none
442
443	integer :: i
444
445	! Time for MPI communications
446	!****************************
447	if (mp_measure_time) call mpif_mtime('commtime',0)
448
449	! Distribute variables, send from pid 0 to other processes
450	!**********************************************************************
451
452	! integers
453	if (lroot) then
454	call MPI_SCATTER(npoint,numpart_mpi,MPI_INTEGER,MPI_IN_PLACE,&
455	&numpart_mpi,MPI_INTEGER,id_root,mp_comm_used,mp_ierr)
456	if (mp_ierr /= 0) goto 600
457	call MPI_SCATTER(idt,numpart_mpi,MPI_INTEGER,MPI_IN_PLACE,&
458	&numpart_mpi,MPI_INTEGER,id_root,mp_comm_used,mp_ierr)
459	if (mp_ierr /= 0) goto 600
460	call MPI_SCATTER(itra1,numpart_mpi,MPI_INTEGER,MPI_IN_PLACE,&
461	&numpart_mpi,MPI_INTEGER,id_root,mp_comm_used,mp_ierr)
462	if (mp_ierr /= 0) goto 600
463	call MPI_SCATTER(nclass,numpart_mpi,MPI_INTEGER,MPI_IN_PLACE,&
464	&numpart_mpi,MPI_INTEGER,id_root,mp_comm_used,mp_ierr)
465	if (mp_ierr /= 0) goto 600
466	call MPI_SCATTER(itramem,numpart_mpi,MPI_INTEGER,MPI_IN_PLACE,&
467	&numpart_mpi,MPI_INTEGER,id_root,mp_comm_used,mp_ierr)
468	if (mp_ierr /= 0) goto 600
469
470	! int2
471	call MPI_SCATTER(cbt,numpart_mpi,MPI_INTEGER2,MPI_IN_PLACE,&
472	&numpart_mpi,MPI_INTEGER2,id_root,mp_comm_used,mp_ierr)
473	if (mp_ierr /= 0) goto 600
474
475	! real
476	call MPI_SCATTER(uap,numpart_mpi,mp_pp,MPI_IN_PLACE,&
477	&numpart_mpi,mp_pp,id_root,mp_comm_used,mp_ierr)
478	if (mp_ierr /= 0) goto 600
479	call MPI_SCATTER(ucp,numpart_mpi,mp_pp,MPI_IN_PLACE,&
480	&numpart_mpi,mp_pp,id_root,mp_comm_used,mp_ierr)
481	if (mp_ierr /= 0) goto 600
482	call MPI_SCATTER(uzp,numpart_mpi,mp_pp,MPI_IN_PLACE,&
483	&numpart_mpi,mp_pp,id_root,mp_comm_used,mp_ierr)
484	if (mp_ierr /= 0) goto 600
485
486	call MPI_SCATTER(us,numpart_mpi,mp_pp,MPI_IN_PLACE,&
487	&numpart_mpi,mp_pp,id_root,mp_comm_used,mp_ierr)
488	if (mp_ierr /= 0) goto 600
489	call MPI_SCATTER(vs,numpart_mpi,mp_pp,MPI_IN_PLACE,&
490	&numpart_mpi,mp_pp,id_root,mp_comm_used,mp_ierr)
491	if (mp_ierr /= 0) goto 600
492	call MPI_SCATTER(ws,numpart_mpi,mp_pp,MPI_IN_PLACE,&
493	&numpart_mpi,mp_pp,id_root,mp_comm_used,mp_ierr)
494	if (mp_ierr /= 0) goto 600
495
496	call MPI_SCATTER(xtra1,numpart_mpi,mp_dp,MPI_IN_PLACE,&
497	&numpart_mpi,mp_dp,id_root,mp_comm_used,mp_ierr)
498	if (mp_ierr /= 0) goto 600
499	call MPI_SCATTER(ytra1,numpart_mpi,mp_dp,MPI_IN_PLACE,&
500	&numpart_mpi,mp_dp,id_root,mp_comm_used,mp_ierr)
501	if (mp_ierr /= 0) goto 600
502	call MPI_SCATTER(ztra1,numpart_mpi,mp_pp,MPI_IN_PLACE,&
503	&numpart_mpi,mp_pp,id_root,mp_comm_used,mp_ierr)
504	if (mp_ierr /= 0) goto 600
505
506	do i=1,nspec
507	call MPI_SCATTER(xmass1(:,i),numpart_mpi,mp_pp,MPI_IN_PLACE,&
508	&numpart_mpi,mp_pp,id_root,mp_comm_used,mp_ierr)
509	if (mp_ierr /= 0) goto 600
510	end do
511
512	else ! (mp_pid >= 1)
513	! integers
514	call MPI_SCATTER(npoint,numpart_mpi,MPI_INTEGER,npoint,&
515	&numpart_mpi,MPI_INTEGER,id_root,mp_comm_used,mp_ierr)
516	if (mp_ierr /= 0) goto 600
517	call MPI_SCATTER(idt,numpart_mpi,MPI_INTEGER,idt,&
518	&numpart_mpi,MPI_INTEGER,id_root,mp_comm_used,mp_ierr)
519	if (mp_ierr /= 0) goto 600
520	call MPI_SCATTER(itra1,numpart_mpi,MPI_INTEGER,itra1,&
521	&numpart_mpi,MPI_INTEGER,id_root,mp_comm_used,mp_ierr)
522	if (mp_ierr /= 0) goto 600
523	call MPI_SCATTER(nclass,numpart_mpi,MPI_INTEGER,nclass,&
524	&numpart_mpi,MPI_INTEGER,id_root,mp_comm_used,mp_ierr)
525	if (mp_ierr /= 0) goto 600
526	call MPI_SCATTER(itramem,numpart_mpi,MPI_INTEGER,itramem,&
527	&numpart_mpi,MPI_INTEGER,id_root,mp_comm_used,mp_ierr)
528	if (mp_ierr /= 0) goto 600
529
530	! int2
531	call MPI_SCATTER(cbt,numpart_mpi,MPI_INTEGER2,cbt,&
532	&numpart_mpi,MPI_INTEGER2,id_root,mp_comm_used,mp_ierr)
533	if (mp_ierr /= 0) goto 600
534
535	! reals
536	call MPI_SCATTER(uap,numpart_mpi,mp_pp,uap,&
537	&numpart_mpi,mp_pp,id_root,mp_comm_used,mp_ierr)
538	if (mp_ierr /= 0) goto 600
539	call MPI_SCATTER(ucp,numpart_mpi,mp_pp,ucp,&
540	&numpart_mpi,mp_pp,id_root,mp_comm_used,mp_ierr)
541	if (mp_ierr /= 0) goto 600
542	call MPI_SCATTER(uzp,numpart_mpi,mp_pp,uzp,&
543	&numpart_mpi,mp_pp,id_root,mp_comm_used,mp_ierr)
544	if (mp_ierr /= 0) goto 600
545
546	call MPI_SCATTER(us,numpart_mpi,mp_pp,us,&
547	&numpart_mpi,mp_pp,id_root,mp_comm_used,mp_ierr)
548	if (mp_ierr /= 0) goto 600
549	call MPI_SCATTER(vs,numpart_mpi,mp_pp,vs,&
550	&numpart_mpi,mp_pp,id_root,mp_comm_used,mp_ierr)
551	if (mp_ierr /= 0) goto 600
552	call MPI_SCATTER(ws,numpart_mpi,mp_pp,ws,&
553	&numpart_mpi,mp_pp,id_root,mp_comm_used,mp_ierr)
554	if (mp_ierr /= 0) goto 600
555
556	call MPI_SCATTER(xtra1,numpart_mpi,mp_dp,xtra1,&
557	&numpart_mpi,mp_dp,id_root,mp_comm_used,mp_ierr)
558	if (mp_ierr /= 0) goto 600
559	call MPI_SCATTER(ytra1,numpart_mpi,mp_dp,ytra1,&
560	&numpart_mpi,mp_dp,id_root,mp_comm_used,mp_ierr)
561	if (mp_ierr /= 0) goto 600
562	call MPI_SCATTER(ztra1,numpart_mpi,mp_pp,ztra1,&
563	&numpart_mpi,mp_pp,id_root,mp_comm_used,mp_ierr)
564	if (mp_ierr /= 0) goto 600
565
566	do i=1,nspec
567	call MPI_SCATTER(xmass1(:,i),numpart_mpi,mp_pp,xmass1(:,i),&
568	&numpart_mpi,mp_pp,id_root,mp_comm_used,mp_ierr)
569	if (mp_ierr /= 0) goto 600
570	end do
571
572	end if !(lroot)
573
574	if (mp_measure_time) call mpif_mtime('commtime',1)
575
576	goto 601
577
578	600 write(,) "mpi_mod> mp_ierr \= 0", mp_ierr
579	stop
580	601 end subroutine mpif_tm_send_vars
581
582
583	subroutine mpif_tm_collect_vars
584	!*******************************************************************************
585	! Collect particle data to PID 0 from all processes. *
586	! *
587	! This can be called from timemanager_mpi, before caclulating *
588	! concentrations/writing output grids. *
589	! *
590	! Currently not in use, as each process calculates concentrations *
591	! separately. *
592	! *
593	! To use this routine (together with mpif_tm_send_vars) to transfer data *
594	! to the MPI root process (useful for debugging), insert code like this *
595	! (in timemanager_mpi.f90) *
596	! *
597	! if (lroot) tot_numpart=numpart ! root stores total numpart *
598	! call mpif_tm_send_dims *
599	! if (numpart>1) then *
600	! call mpif_tm_send_vars *
601	! end if *
602	! *
603	! To collect data from all processes to root process after a parallel *
604	! region: *
605	! *
606	! if (numpart>0) then *
607	! if (lroot) numpart = tot_numpart *
608	! call mpif_tm_collect_vars *
609	! end if *
610	! *
611	! Then a section that begins with "if (lroot) ..." will behave like *
612	! serial flexpart *
613	! *
614	!*******************************************************************************
615	use com_mod !, only: numpart, nspec, itra1, npoint, nclass
616
617	implicit none
618
619	integer :: i
620
621	logical :: use_mp_vars = .false.! use mp_... type allocated vars
622	logical :: use_in_place = .true.! use MPI_IN_PLACE to collect vars.
623
624
625	! Time for MPI communications
626	if (mp_measure_time) call mpif_mtime('commtime',0)
627
628	! Distribute variables, send from pid 0 to other processes
629	!**********************************************************************
630
631	if (.not. use_mp_vars.and..not.use_in_place) then
632
633	! Integers:
634	call MPI_GATHER(npoint, numpart_mpi, MPI_INTEGER, npoint, &
635	&numpart_mpi, MPI_INTEGER, id_root, mp_comm_used, mp_ierr)
636	if (mp_ierr /= 0) goto 600
637	call MPI_GATHER(idt, numpart_mpi, MPI_INTEGER, idt, &
638	&numpart_mpi, MPI_INTEGER, id_root, mp_comm_used, mp_ierr)
639	if (mp_ierr /= 0) goto 600
640	call MPI_GATHER(itra1, numpart_mpi, MPI_INTEGER, itra1, &
641	&numpart_mpi, MPI_INTEGER, id_root, mp_comm_used, mp_ierr)
642	if (mp_ierr /= 0) goto 600
643	call MPI_GATHER(nclass, numpart_mpi, MPI_INTEGER, nclass, &
644	&numpart_mpi, MPI_INTEGER, id_root, mp_comm_used, mp_ierr)
645	if (mp_ierr /= 0) goto 600
646	call MPI_GATHER(itramem, numpart_mpi, MPI_INTEGER, itramem, &
647	&numpart_mpi, MPI_INTEGER, id_root, mp_comm_used, mp_ierr)
648	if (mp_ierr /= 0) goto 600
649
650	! int2
651	call MPI_GATHER(cbt, numpart_mpi, MPI_INTEGER2, cbt, &
652	&numpart_mpi, MPI_INTEGER2, id_root, mp_comm_used, mp_ierr)
653	if (mp_ierr /= 0) goto 600
654
655	! Reals:
656	call MPI_GATHER(uap, numpart_mpi, mp_pp, uap, &
657	&numpart_mpi, mp_pp, id_root, mp_comm_used, mp_ierr)
658	if (mp_ierr /= 0) goto 600
659	call MPI_GATHER(ucp, numpart_mpi, mp_pp, ucp, &
660	&numpart_mpi, mp_pp, id_root, mp_comm_used, mp_ierr)
661	if (mp_ierr /= 0) goto 600
662	call MPI_GATHER(uzp, numpart_mpi, mp_pp, uzp, &
663	&numpart_mpi, mp_pp, id_root, mp_comm_used, mp_ierr)
664	if (mp_ierr /= 0) goto 600
665
666	call MPI_GATHER(us, numpart_mpi, mp_pp, us, &
667	&numpart_mpi, mp_pp, id_root, mp_comm_used, mp_ierr)
668	if (mp_ierr /= 0) goto 600
669	call MPI_GATHER(vs, numpart_mpi, mp_pp, vs, &
670	&numpart_mpi, mp_pp, id_root, mp_comm_used, mp_ierr)
671	if (mp_ierr /= 0) goto 600
672	call MPI_GATHER(ws, numpart_mpi, mp_pp, ws, &
673	&numpart_mpi, mp_pp, id_root, mp_comm_used, mp_ierr)
674	if (mp_ierr /= 0) goto 600
675
676
677	call MPI_GATHER(xtra1, numpart_mpi, mp_dp, xtra1, &
678	&numpart_mpi, mp_dp, id_root, mp_comm_used, mp_ierr)
679	if (mp_ierr /= 0) goto 600
680	call MPI_GATHER(ytra1, numpart_mpi, mp_dp, ytra1, &
681	&numpart_mpi, mp_dp, id_root, mp_comm_used, mp_ierr)
682	if (mp_ierr /= 0) goto 600
683	call MPI_GATHER(ztra1, numpart_mpi, mp_pp, ztra1, &
684	&numpart_mpi, mp_pp, id_root, mp_comm_used, mp_ierr)
685	if (mp_ierr /= 0) goto 600
686
687	do i=1, nspec
688	call MPI_GATHER(xmass1(:,i), numpart_mpi, mp_pp, xmass1(:,i), &
689	&numpart_mpi, mp_pp, id_root, mp_comm_used, mp_ierr)
690	if (mp_ierr /= 0) goto 600
691
692	end do
693
694	! Use variables npoint etc directly for communications
695	!***********************************************************************
696	else if (use_in_place.and..not.use_mp_vars) then
697	if (lroot) then
698	! Integers:
699	call MPI_GATHER(MPI_IN_PLACE, numpart_mpi, MPI_INTEGER, npoint, &
700	&numpart_mpi, MPI_INTEGER, id_root, mp_comm_used, mp_ierr)
701	if (mp_ierr /= 0) goto 600
702	call MPI_GATHER(MPI_IN_PLACE, numpart_mpi, MPI_INTEGER, idt, &
703	&numpart_mpi, MPI_INTEGER, id_root, mp_comm_used, mp_ierr)
704	if (mp_ierr /= 0) goto 600
705	call MPI_GATHER(MPI_IN_PLACE, numpart_mpi, MPI_INTEGER, itra1, &
706	&numpart_mpi, MPI_INTEGER, id_root, mp_comm_used, mp_ierr)
707	if (mp_ierr /= 0) goto 600
708	call MPI_GATHER(MPI_IN_PLACE, numpart_mpi, MPI_INTEGER, nclass, &
709	&numpart_mpi, MPI_INTEGER, id_root, mp_comm_used, mp_ierr)
710	if (mp_ierr /= 0) goto 600
711	call MPI_GATHER(MPI_IN_PLACE, numpart_mpi, MPI_INTEGER, itramem, &
712	&numpart_mpi, MPI_INTEGER, id_root, mp_comm_used, mp_ierr)
713	if (mp_ierr /= 0) goto 600
714
715	! int2
716	call MPI_GATHER(MPI_IN_PLACE, numpart_mpi, MPI_INTEGER2, cbt, &
717	&numpart_mpi, MPI_INTEGER2, id_root, mp_comm_used, mp_ierr)
718	if (mp_ierr /= 0) goto 600
719
720	! Reals:
721	call MPI_GATHER(MPI_IN_PLACE, numpart_mpi, mp_pp, uap, &
722	&numpart_mpi, mp_pp, id_root, mp_comm_used, mp_ierr)
723	if (mp_ierr /= 0) goto 600
724	call MPI_GATHER(MPI_IN_PLACE, numpart_mpi, mp_pp, ucp, &
725	&numpart_mpi, mp_pp, id_root, mp_comm_used, mp_ierr)
726	if (mp_ierr /= 0) goto 600
727	call MPI_GATHER(MPI_IN_PLACE, numpart_mpi, mp_pp, uzp, &
728	&numpart_mpi, mp_pp, id_root, mp_comm_used, mp_ierr)
729	if (mp_ierr /= 0) goto 600
730
731	call MPI_GATHER(MPI_IN_PLACE, numpart_mpi, mp_pp, us, &
732	&numpart_mpi, mp_pp, id_root, mp_comm_used, mp_ierr)
733	if (mp_ierr /= 0) goto 600
734	call MPI_GATHER(MPI_IN_PLACE, numpart_mpi, mp_pp, vs, &
735	&numpart_mpi, mp_pp, id_root, mp_comm_used, mp_ierr)
736	if (mp_ierr /= 0) goto 600
737	call MPI_GATHER(MPI_IN_PLACE, numpart_mpi, mp_pp, ws, &
738	&numpart_mpi, mp_pp, id_root, mp_comm_used, mp_ierr)
739	if (mp_ierr /= 0) goto 600
740
741
742	call MPI_GATHER(MPI_IN_PLACE, numpart_mpi, mp_dp, xtra1, &
743	&numpart_mpi, mp_dp, id_root, mp_comm_used, mp_ierr)
744	if (mp_ierr /= 0) goto 600
745	call MPI_GATHER(MPI_IN_PLACE, numpart_mpi, mp_dp, ytra1, &
746	&numpart_mpi, mp_dp, id_root, mp_comm_used, mp_ierr)
747	if (mp_ierr /= 0) goto 600
748	call MPI_GATHER(MPI_IN_PLACE, numpart_mpi, mp_pp, ztra1, &
749	&numpart_mpi, mp_pp, id_root, mp_comm_used, mp_ierr)
750	if (mp_ierr /= 0) goto 600
751
752	do i=1, nspec
753	call MPI_GATHER(MPI_IN_PLACE, numpart_mpi, mp_pp, xmass1(:,i), &
754	&numpart_mpi, mp_pp, id_root, mp_comm_used, mp_ierr)
755	if (mp_ierr /= 0) goto 600
756	end do
757
758	else ! (for mp_pid >= 1)
759
760	! Integers:
761	call MPI_GATHER(npoint, numpart_mpi, MPI_INTEGER, npoint, &
762	&numpart_mpi, MPI_INTEGER, id_root, mp_comm_used, mp_ierr)
763	if (mp_ierr /= 0) goto 600
764	call MPI_GATHER(idt, numpart_mpi, MPI_INTEGER, idt, &
765	&numpart_mpi, MPI_INTEGER, id_root, mp_comm_used, mp_ierr)
766	if (mp_ierr /= 0) goto 600
767	call MPI_GATHER(itra1, numpart_mpi, MPI_INTEGER, itra1, &
768	&numpart_mpi, MPI_INTEGER, id_root, mp_comm_used, mp_ierr)
769	if (mp_ierr /= 0) goto 600
770	call MPI_GATHER(nclass, numpart_mpi, MPI_INTEGER, nclass, &
771	&numpart_mpi, MPI_INTEGER, id_root, mp_comm_used, mp_ierr)
772	if (mp_ierr /= 0) goto 600
773	call MPI_GATHER(itramem, numpart_mpi, MPI_INTEGER, itramem, &
774	&numpart_mpi, MPI_INTEGER, id_root, mp_comm_used, mp_ierr)
775	if (mp_ierr /= 0) goto 600
776
777	! int2
778	call MPI_GATHER(cbt, numpart_mpi, MPI_INTEGER2, cbt, &
779	&numpart_mpi, MPI_INTEGER2, id_root, mp_comm_used, mp_ierr)
780	if (mp_ierr /= 0) goto 600
781
782	! Reals:
783	call MPI_GATHER(uap, numpart_mpi, mp_pp, uap, &
784	&numpart_mpi, mp_pp, id_root, mp_comm_used, mp_ierr)
785	if (mp_ierr /= 0) goto 600
786	call MPI_GATHER(ucp, numpart_mpi, mp_pp, ucp, &
787	&numpart_mpi, mp_pp, id_root, mp_comm_used, mp_ierr)
788	if (mp_ierr /= 0) goto 600
789	call MPI_GATHER(uzp, numpart_mpi, mp_pp, uzp, &
790	&numpart_mpi, mp_pp, id_root, mp_comm_used, mp_ierr)
791	if (mp_ierr /= 0) goto 600
792
793	call MPI_GATHER(us, numpart_mpi, mp_pp, us, &
794	&numpart_mpi, mp_pp, id_root, mp_comm_used, mp_ierr)
795	if (mp_ierr /= 0) goto 600
796	call MPI_GATHER(vs, numpart_mpi, mp_pp, vs, &
797	&numpart_mpi, mp_pp, id_root, mp_comm_used, mp_ierr)
798	if (mp_ierr /= 0) goto 600
799	call MPI_GATHER(ws, numpart_mpi, mp_pp, ws, &
800	&numpart_mpi, mp_pp, id_root, mp_comm_used, mp_ierr)
801	if (mp_ierr /= 0) goto 600
802
803
804	call MPI_GATHER(xtra1, numpart_mpi, mp_dp, xtra1, &
805	&numpart_mpi, mp_dp, id_root, mp_comm_used, mp_ierr)
806	if (mp_ierr /= 0) goto 600
807	call MPI_GATHER(ytra1, numpart_mpi, mp_dp, ytra1, &
808	&numpart_mpi, mp_dp, id_root, mp_comm_used, mp_ierr)
809	if (mp_ierr /= 0) goto 600
810	call MPI_GATHER(ztra1, numpart_mpi, mp_pp, ztra1, &
811	&numpart_mpi, mp_pp, id_root, mp_comm_used, mp_ierr)
812	if (mp_ierr /= 0) goto 600
813
814	do i=1, nspec
815	call MPI_GATHER(xmass1(:,i), numpart_mpi, mp_pp, xmass1(:,i), &
816	&numpart_mpi, mp_pp, id_root, mp_comm_used, mp_ierr)
817	if (mp_ierr /= 0) goto 600
818	end do
819	end if ! (lroot)
820	end if ! (use_in_place)
821
822	if (mp_measure_time) call mpif_mtime('commtime',1)
823
824	goto 601
825
826	600 write(,) "mpi_mod> mp_ierr \= 0", mp_ierr
827	stop
828	601 end subroutine mpif_tm_collect_vars
829
830
831	subroutine mpif_gf_send_vars(memstat)
832	!*******************************************************************************
833	! DESCRIPTION
834	! Distribute 'getfield' variables from reader process
835	!
836	! Called from timemanager
837	!
838	! NOTE
839	! This subroutine distributes windfields read from the reader process to
840	! all other processes. Usually one set of fields is transfered, but at
841	! first timestep when there are no fields in memory, both are transfered.
842	! MPI_Bcast is used, so implicitly all processes are synchronized at this
843	! step
844	!
845	! TODO
846	! GFS version
847	!
848	!*******************************************************************************
849	use com_mod
850	use par_mod,only: numwfmem
851
852	implicit none
853
854	integer, intent(in) :: memstat
855
856	! Common array sizes used for communications
857	integer, parameter :: d3_size1 = nxmaxnymaxnzmax
858	integer, parameter :: d3_size2 = nxmaxnymaxnuvzmax
859	integer, parameter :: d2_size1 = nxmax*nymax
860	integer, parameter :: d2_size2 = nxmaxnymaxmaxspec
861	integer, parameter :: d2_size3 = nxmax*nymax
862	integer, parameter :: d1_size1 = maxwf
863
864	integer :: d3s1,d3s2,d2s1,d2s2
865
866	! li,ui lower,upper indices of fields to be transfered (1-3, ui>=li)
867	integer :: li,ui
868
869	! First time routine is called the unchangeable fields will be transfered
870	logical :: first_call=.true.
871
872	!**********************************************************************
873
874	! Sizes of arrays transfered
875	d3s1=d3_size1
876	d3s2=d3_size2
877	d2s1=d2_size1
878	d2s2=d2_size2
879
880	! Decide which field(s) need to be transfered
881	if (memstat.ge.32) then ! distribute 2 fields, to 1st/2nd indices
882	li=1; ui=2
883	d3s1=2*d3_size1
884	d3s2=2*d3_size2
885	d2s1=2*d2_size1
886	d2s2=2*d2_size2
887	else if (memstat.eq.17) then ! distribute 1 field, place on 1st index
888	li=1; ui=1
889	else if (memstat.eq.18) then ! distribute 1 field, place on 2nd index
890	li=2; ui=2
891	else if (memstat.eq.19) then ! distribute 1 field, place on 3nd index
892	li=3; ui=3
893	else
894	write(,) '#### mpi_mod::mpif_gf_send_vars> ERROR: ', &
895	& 'wrong value of memstat, exiting ####', memstat
896	stop
897	end if
898
899
900	! Time for MPI communications
901	if (mp_measure_time) call mpif_mtime('commtime',0)
902
903	! Send variables from getfield process (id_read) to other processes
904	!**********************************************************************
905
906	! The non-reader processes need to know if clouds were read.
907	! TODO: only at first step or always?
908	call MPI_Bcast(readclouds,1,MPI_LOGICAL,id_read,MPI_COMM_WORLD,mp_ierr)
909	if (mp_ierr /= 0) goto 600
910
911	! Static fields/variables sent only at startup
912	if (first_call) then
913	call MPI_Bcast(oro(:,:),d2_size3,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
914	if (mp_ierr /= 0) goto 600
915	call MPI_Bcast(excessoro(:,:),d2_size3,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
916	if (mp_ierr /= 0) goto 600
917	call MPI_Bcast(lsm(:,:),d2_size3,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
918	if (mp_ierr /= 0) goto 600
919	call MPI_Bcast(xlanduse(:,:,:),d2_size3*numclass,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
920	if (mp_ierr /= 0) goto 600
921	call MPI_Bcast(wftime,d1_size1,MPI_INTEGER,id_read,MPI_COMM_WORLD,mp_ierr)
922	if (mp_ierr /= 0) goto 600
923	call MPI_Bcast(numbwf,1,MPI_INTEGER,id_read,MPI_COMM_WORLD,mp_ierr)
924	if (mp_ierr /= 0) goto 600
925	call MPI_Bcast(nmixz,1,MPI_INTEGER,id_read,MPI_COMM_WORLD,mp_ierr)
926	if (mp_ierr /= 0) goto 600
927	call MPI_Bcast(height,nzmax,MPI_INTEGER,id_read,MPI_COMM_WORLD,mp_ierr)
928	if (mp_ierr /= 0) goto 600
929
930	first_call=.false.
931	endif
932
933	call MPI_Bcast(uu(:,:,:,li:ui),d3s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
934	if (mp_ierr /= 0) goto 600
935	call MPI_Bcast(vv(:,:,:,li:ui),d3s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
936	if (mp_ierr /= 0) goto 600
937	call MPI_Bcast(uupol(:,:,:,li:ui),d3s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
938	if (mp_ierr /= 0) goto 600
939	call MPI_Bcast(vvpol(:,:,:,li:ui),d3s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
940	if (mp_ierr /= 0) goto 600
941	call MPI_Bcast(ww(:,:,:,li:ui),d3s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
942	if (mp_ierr /= 0) goto 600
943	call MPI_Bcast(tt(:,:,:,li:ui),d3s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
944	if (mp_ierr /= 0) goto 600
945	call MPI_Bcast(rho(:,:,:,li:ui),d3s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
946	if (mp_ierr /= 0) goto 600
947	call MPI_Bcast(drhodz(:,:,:,li:ui),d3s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
948	if (mp_ierr /= 0) goto 600
949	call MPI_Bcast(tth(:,:,:,li:ui),d3s2,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
950	if (mp_ierr /= 0) goto 600
951	call MPI_Bcast(qvh(:,:,:,li:ui),d3s2,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
952	if (mp_ierr /= 0) goto 600
953	call MPI_Bcast(qv(:,:,:,li:ui),d3s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
954	if (mp_ierr /= 0) goto 600
955	call MPI_Bcast(pv(:,:,:,li:ui),d3s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
956	if (mp_ierr /= 0) goto 600
957	call MPI_Bcast(clouds(:,:,:,li:ui),d3s1,MPI_INTEGER1,id_read,MPI_COMM_WORLD,mp_ierr)
958	if (mp_ierr /= 0) goto 600
959
960	! cloud water/ice:
961	if (readclouds) then
962	call MPI_Bcast(clwc(:,:,:,li:ui),d3s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
963	if (mp_ierr /= 0) goto 600
964	call MPI_Bcast(ciwc(:,:,:,li:ui),d3s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
965	if (mp_ierr /= 0) goto 600
966	end if
967
968	! 2D fields
969	call MPI_Bcast(cloudsh(:,:,li:ui),d2s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
970	if (mp_ierr /= 0) goto 600
971	call MPI_Bcast(vdep(:,:,:,li:ui),d2s2,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
972	if (mp_ierr /= 0) goto 600
973	call MPI_Bcast(ps(:,:,:,li:ui),d2s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
974	if (mp_ierr /= 0) goto 600
975	call MPI_Bcast(sd(:,:,:,li:ui),d2s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
976	if (mp_ierr /= 0) goto 600
977	call MPI_Bcast(tcc(:,:,:,li:ui),d2s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
978	if (mp_ierr /= 0) goto 600
979	call MPI_Bcast(tt2(:,:,:,li:ui),d2s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
980	if (mp_ierr /= 0) goto 600
981	call MPI_Bcast(td2(:,:,:,li:ui),d2s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
982	if (mp_ierr /= 0) goto 600
983	call MPI_Bcast(lsprec(:,:,:,li:ui),d2s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
984	if (mp_ierr /= 0) goto 600
985	call MPI_Bcast(convprec(:,:,:,li:ui),d2s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
986	if (mp_ierr /= 0) goto 600
987	call MPI_Bcast(ustar(:,:,:,li:ui),d2s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
988	if (mp_ierr /= 0) goto 600
989	call MPI_Bcast(wstar(:,:,:,li:ui),d2s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
990	if (mp_ierr /= 0) goto 600
991	call MPI_Bcast(hmix(:,:,:,li:ui),d2s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
992	if (mp_ierr /= 0) goto 600
993	call MPI_Bcast(tropopause(:,:,:,li:ui),d2s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
994	if (mp_ierr /= 0) goto 600
995	call MPI_Bcast(oli(:,:,:,li:ui),d2s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
996	if (mp_ierr /= 0) goto 600
997
998	call MPI_Bcast(z0,numclass,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
999	if (mp_ierr /= 0) goto 600
1000
1001	if (mp_measure_time) call mpif_mtime('commtime',1)
1002
1003	goto 601
1004
1005	600 write(,) "mpi_mod> mp_ierr \= 0", mp_ierr
1006	stop
1007
1008	601 end subroutine mpif_gf_send_vars
1009
1010
1011	subroutine mpif_gf_send_vars_nest(memstat)
1012	!***********************************************************************
1013	! DESCRIPTION
1014	! Distribute 'getfield' variables from reader process to all processes.
1015	! For nested fields
1016	!
1017	! Called from timemanager
1018	!
1019	! NOTE
1020	! This subroutine distributes nested windfields from the reader process to
1021	! all other processes. Usually one set of fields is transfered, but at
1022	! first timestep when there are no fields in memory, both are transfered.
1023	! MPI_Bcast is used, so implicitly all processes are synchronized at this
1024	! step
1025	!
1026	! TODO
1027	! Transfer cloud water/ice if and when available for nested
1028	!
1029	!***********************************************************************
1030	use com_mod
1031	use par_mod, only: numwfmem
1032
1033	implicit none
1034
1035	integer, intent(in) :: memstat
1036	integer :: i
1037	! li,ui lower,upper indices of fields to be transfered (1-3, ui>=li)
1038	integer :: li,ui
1039
1040	! Common array sizes used for communications
1041	integer :: d3_size1 = nxmaxnnymaxnnzmax*maxnests
1042	integer :: d3_size2 = nxmaxnnymaxnnuvzmax*maxnests
1043	integer :: d2_size1 = nxmaxnnymaxnmaxnests
1044	integer :: d2_size2 = nxmaxnnymaxnmaxspec*maxnests
1045	integer :: d2_size3 = nxmaxnnymaxnmaxnests
1046
1047	integer :: d3s1,d3s2,d2s1,d2s2
1048
1049	! First time routine is called the unchangeable fields will be transfered
1050	logical :: first_call=.true.
1051
1052	!**********************************************************************
1053
1054	! Sizes of arrays transfered
1055	d3s1=d3_size1
1056	d3s2=d3_size2
1057	d2s1=d2_size1
1058	d2s2=d2_size2
1059
1060	! Decide which field(s) need to be transfered
1061	if (memstat.ge.32) then ! distribute 2 fields
1062	li=1; ui=2
1063	d3s1=2*d3_size1
1064	d3s2=2*d3_size2
1065	d2s1=2*d2_size1
1066	d2s2=2*d2_size2
1067	else if (memstat.eq.17) then ! distribute 1 field, on 1st index
1068	li=1; ui=1
1069	else if (memstat.eq.18) then ! distribute 1 field, on 2nd index
1070	li=2; ui=2
1071	else if (memstat.eq.19) then ! distribute 1 field, on 3nd index
1072	li=3; ui=3
1073	else
1074	write(,) '#### mpi_mod::mpif_gf_send_vars_nest> ERROR: ', &
1075	& 'wrong value of memstat, exiting ####', memstat
1076	stop
1077	end if
1078
1079
1080	! Time for MPI communications
1081	if (mp_measure_time) call mpif_mtime('commtime',0)
1082
1083	! Distribute variables, send from getfield process (id_read) to other
1084	! processes
1085	!**********************************************************************
1086
1087	! Static fields/variables sent only at startup
1088	if (first_call) then
1089	call MPI_Bcast(oron(:,:,:),d2_size3,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
1090	if (mp_ierr /= 0) goto 600
1091	call MPI_Bcast(excessoron(:,:,:),d2_size3,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
1092	if (mp_ierr /= 0) goto 600
1093	call MPI_Bcast(lsmn(:,:,:),d2_size3,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
1094	if (mp_ierr /= 0) goto 600
1095	call MPI_Bcast(xlandusen(:,:,:,:),d2_size3*numclass,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
1096	if (mp_ierr /= 0) goto 600
1097	first_call=.false.
1098	end if
1099
1100	! MPI prefers contiguous arrays for sending (else a buffer is created),
1101	! hence the loop
1102
1103	do i=1, numbnests
1104	! 3D fields
1105	call MPI_Bcast(uun(:,:,:,li:ui,i),d3s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
1106	if (mp_ierr /= 0) goto 600
1107	call MPI_Bcast(vvn(:,:,:,li:ui,i),d3s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
1108	if (mp_ierr /= 0) goto 600
1109	call MPI_Bcast(wwn(:,:,:,li:ui,i),d3s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
1110	if (mp_ierr /= 0) goto 600
1111	call MPI_Bcast(ttn(:,:,:,li:ui,i),d3s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
1112	if (mp_ierr /= 0) goto 600
1113	call MPI_Bcast(rhon(:,:,:,li:ui,i),d3s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
1114	if (mp_ierr /= 0) goto 600
1115	call MPI_Bcast(drhodzn(:,:,:,li:ui,i),d3s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
1116	if (mp_ierr /= 0) goto 600
1117	call MPI_Bcast(tthn(:,:,:,li:ui,i),d3s2,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
1118	if (mp_ierr /= 0) goto 600
1119	call MPI_Bcast(qvhn(:,:,:,li:ui,i),d3s2,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
1120	if (mp_ierr /= 0) goto 600
1121	call MPI_Bcast(qvn(:,:,:,li:ui,i),d3s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
1122	if (mp_ierr /= 0) goto 600
1123	call MPI_Bcast(pvn(:,:,:,li:ui,i),d3s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
1124	if (mp_ierr /= 0) goto 600
1125	call MPI_Bcast(cloudsn(:,:,:,li:ui,i),d3s1,MPI_INTEGER1,id_read,MPI_COMM_WORLD,mp_ierr)
1126	if (mp_ierr /= 0) goto 600
1127
1128	! 2D fields
1129	call MPI_Bcast(cloudsnh(:,:,li:ui,i),d2s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
1130	if (mp_ierr /= 0) goto 600
1131	call MPI_Bcast(vdepn(:,:,:,li:ui,i),d2s2,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
1132	if (mp_ierr /= 0) goto 600
1133	call MPI_Bcast(psn(:,:,:,li:ui,i),d2s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
1134	if (mp_ierr /= 0) goto 600
1135	call MPI_Bcast(sdn(:,:,:,li:ui,i),d2s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
1136	if (mp_ierr /= 0) goto 600
1137	call MPI_Bcast(tccn(:,:,:,li:ui,i),d2s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
1138	if (mp_ierr /= 0) goto 600
1139	call MPI_Bcast(tt2n(:,:,:,li:ui,i),d2s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
1140	if (mp_ierr /= 0) goto 600
1141	call MPI_Bcast(td2n(:,:,:,li:ui,i),d2s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
1142	if (mp_ierr /= 0) goto 600
1143	call MPI_Bcast(lsprecn(:,:,:,li:ui,i),d2s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
1144	if (mp_ierr /= 0) goto 600
1145	call MPI_Bcast(convprecn(:,:,:,li:ui,i),d2s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
1146	if (mp_ierr /= 0) goto 600
1147	call MPI_Bcast(ustarn(:,:,:,li:ui,i),d2s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
1148	if (mp_ierr /= 0) goto 600
1149	call MPI_Bcast(wstarn(:,:,:,li:ui,i),d2s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
1150	if (mp_ierr /= 0) goto 600
1151	call MPI_Bcast(olin(:,:,:,li:ui,i),d2s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
1152	if (mp_ierr /= 0) goto 600
1153	call MPI_Bcast(hmixn(:,:,:,li:ui,i),d2s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
1154	if (mp_ierr /= 0) goto 600
1155	call MPI_Bcast(tropopausen(:,:,:,li:ui,i),d2s1,mp_pp,id_read,MPI_COMM_WORLD,mp_ierr)
1156	if (mp_ierr /= 0) goto 600
1157	end do
1158
1159	if (mp_measure_time) call mpif_mtime('commtime',1)
1160
1161	goto 601
1162
1163	600 write(,) "mpi_mod> mp_ierr \= 0",mp_ierr
1164	stop
1165
1166	601 end subroutine mpif_gf_send_vars_nest
1167
1168
1169	subroutine mpif_gf_send_vars_async(memstat)
1170	!*******************************************************************************
1171	! DESCRIPTION
1172	! Distribute 'getfield' variables from reader process to all processes.
1173	! Called from timemanager by reader process only
1174	!
1175	! NOTE
1176	! This version uses asynchronious sends. The newest fields are sent in the
1177	! background, so calculations can continue while
1178	! MPI communications are performed.
1179	!
1180	! The way the MPI tags/requests are sequenced means that this routine must
1181	! carefully match routine 'mpif_gf_recv_vars_async'
1182	!
1183	! VARIABLES
1184	! memstat -- input, for resolving pointer to windfield index being read
1185	! mind -- index where to place new fields
1186	!
1187	! TODO
1188	! Transfer cloud water/ice
1189	!
1190	!*******************************************************************************
1191	use com_mod
1192
1193	implicit none
1194
1195	integer, intent(in) :: memstat
1196	integer :: mind
1197	integer :: dest,i
1198
1199	! Common array sizes used for communications
1200	integer :: d3s1 = nxmaxnymaxnzmax
1201	integer :: d3s2 = nxmaxnymaxnuvzmax
1202	integer :: d2s1 = nxmax*nymax
1203	integer :: d2s2 = nxmaxnymaxmaxspec
1204	integer :: d1s1 = maxwf
1205
1206	!*******************************************************************************
1207
1208	! At the time the send is posted, the reader process is one step further
1209	! in the permutation of memstat compared with the receiving processes
1210
1211	if (memstat.ge.32) then
1212	! last read was synchronous, to indices 1 and 2, 3 is free
1213	write(,) "#### mpi_mod::mpif_gf_send_vars_async> ERROR: &
1214	& memstat>=32 should never happen here."
1215	stop
1216	else if (memstat.eq.17) then
1217	! old fields on 1,2, send 3
1218	mind=3
1219	else if (memstat.eq.18) then
1220	! old fields on 2,3, send 1
1221	mind=1
1222	else if (memstat.eq.19) then
1223	! old fields on 3,1, send 2
1224	mind=2
1225	else
1226	write(,) "#### mpi_mod::mpif_gf_send_vars_async> ERROR: &
1227	& invalid memstat"
1228	end if
1229
1230	if (mp_dev_mode) then
1231	if (mp_pid.ne.id_read) then
1232	write(,) 'MPI_DEV_MODE: error calling mpif_gf_send_vars_async'
1233	end if
1234	end if
1235
1236	if (mp_dev_mode) write(,) '## in mpif_gf_send_vars_async, memstat:', memstat
1237
1238	! Time for MPI communications
1239	if (mp_measure_time) call mpif_mtime('commtime',0)
1240
1241	! Loop over receiving processes, initiate data sending
1242	!*****************************************************
1243
1244	do dest=0,mp_np-1 ! mp_np-2 will also work if last proc reserved for reading
1245	! TODO: use mp_partgroup_np here
1246	if (dest.eq.id_read) cycle
1247	i=dest*nvar_async
1248	call MPI_Isend(uu(:,:,:,mind),d3s1,mp_pp,dest,tm1,MPI_COMM_WORLD,reqs(i),mp_ierr)
1249	if (mp_ierr /= 0) goto 600
1250	i=i+1
1251	call MPI_Isend(vv(:,:,:,mind),d3s1,mp_pp,dest,tm1,MPI_COMM_WORLD,reqs(i),mp_ierr)
1252	if (mp_ierr /= 0) goto 600
1253	i=i+1
1254	call MPI_Isend(uupol(:,:,:,mind),d3s1,mp_pp,dest,tm1,MPI_COMM_WORLD,reqs(i),mp_ierr)
1255	if (mp_ierr /= 0) goto 600
1256	i=i+1
1257	call MPI_Isend(vvpol(:,:,:,mind),d3s1,mp_pp,dest,tm1,MPI_COMM_WORLD,reqs(i),mp_ierr)
1258	if (mp_ierr /= 0) goto 600
1259	i=i+1
1260	call MPI_Isend(ww(:,:,:,mind),d3s1,mp_pp,dest,tm1,MPI_COMM_WORLD,reqs(i),mp_ierr)
1261	if (mp_ierr /= 0) goto 600
1262	i=i+1
1263	call MPI_Isend(tt(:,:,:,mind),d3s1,mp_pp,dest,tm1,MPI_COMM_WORLD,reqs(i),mp_ierr)
1264	if (mp_ierr /= 0) goto 600
1265	i=i+1
1266	call MPI_Isend(rho(:,:,:,mind),d3s1,mp_pp,dest,tm1,MPI_COMM_WORLD,reqs(i),mp_ierr)
1267	if (mp_ierr /= 0) goto 600
1268	i=i+1
1269	call MPI_Isend(drhodz(:,:,:,mind),d3s1,mp_pp,dest,tm1,MPI_COMM_WORLD,reqs(i),mp_ierr)
1270	if (mp_ierr /= 0) goto 600
1271	i=i+1
1272	call MPI_Isend(tth(:,:,:,mind),d3s2,mp_pp,dest,tm1,MPI_COMM_WORLD,reqs(i),mp_ierr)
1273	if (mp_ierr /= 0) goto 600
1274	i=i+1
1275	call MPI_Isend(qvh(:,:,:,mind),d3s2,mp_pp,dest,tm1,MPI_COMM_WORLD,reqs(i),mp_ierr)
1276	if (mp_ierr /= 0) goto 600
1277	i=i+1
1278	call MPI_Isend(qv(:,:,:,mind),d3s1,mp_pp,dest,tm1,MPI_COMM_WORLD,reqs(i),mp_ierr)
1279	if (mp_ierr /= 0) goto 600
1280	i=i+1
1281	call MPI_Isend(pv(:,:,:,mind),d3s1,mp_pp,dest,tm1,MPI_COMM_WORLD,reqs(i),mp_ierr)
1282	if (mp_ierr /= 0) goto 600
1283	i=i+1
1284	call MPI_Isend(clouds(:,:,:,mind),d3s1,MPI_INTEGER1,dest,tm1,MPI_COMM_WORLD,reqs(i),mp_ierr)
1285	i=i+1
1286	if (mp_ierr /= 0) goto 600
1287
1288	call MPI_Isend(cloudsh(:,:,mind),d2s1,mp_pp,dest,tm1,MPI_COMM_WORLD,reqs(i),mp_ierr)
1289	if (mp_ierr /= 0) goto 600
1290	i=i+1
1291	call MPI_Isend(vdep(:,:,:,mind),d2s2,mp_pp,dest,tm1,MPI_COMM_WORLD,reqs(i),mp_ierr)
1292	if (mp_ierr /= 0) goto 600
1293	i=i+1
1294	call MPI_Isend(ps(:,:,:,mind),d2s1,mp_pp,dest,tm1,MPI_COMM_WORLD,reqs(i),mp_ierr)
1295	if (mp_ierr /= 0) goto 600
1296	i=i+1
1297	call MPI_Isend(sd(:,:,:,mind),d2s1,mp_pp,dest,tm1,MPI_COMM_WORLD,reqs(i),mp_ierr)
1298	if (mp_ierr /= 0) goto 600
1299	i=i+1
1300	call MPI_Isend(tcc(:,:,:,mind),d2s1,mp_pp,dest,tm1,MPI_COMM_WORLD,reqs(i),mp_ierr)
1301	if (mp_ierr /= 0) goto 600
1302	i=i+1
1303	call MPI_Isend(tt2(:,:,:,mind),d2s1,mp_pp,dest,tm1,MPI_COMM_WORLD,reqs(i),mp_ierr)
1304	if (mp_ierr /= 0) goto 600
1305	i=i+1
1306	call MPI_Isend(td2(:,:,:,mind),d2s1,mp_pp,dest,tm1,MPI_COMM_WORLD,reqs(i),mp_ierr)
1307	if (mp_ierr /= 0) goto 600
1308	i=i+1
1309	call MPI_Isend(lsprec(:,:,:,mind),d2s1,mp_pp,dest,tm1,MPI_COMM_WORLD,reqs(i),mp_ierr)
1310	if (mp_ierr /= 0) goto 600
1311	i=i+1
1312	call MPI_Isend(convprec(:,:,:,mind),d2s1,mp_pp,dest,tm1,MPI_COMM_WORLD,reqs(i),mp_ierr)
1313	if (mp_ierr /= 0) goto 600
1314	i=i+1
1315	call MPI_Isend(ustar(:,:,:,mind),d2s1,mp_pp,dest,tm1,MPI_COMM_WORLD,reqs(i),mp_ierr)
1316	if (mp_ierr /= 0) goto 600
1317	i=i+1
1318	call MPI_Isend(wstar(:,:,:,mind),d2s1,mp_pp,dest,tm1,MPI_COMM_WORLD,reqs(i),mp_ierr)
1319	if (mp_ierr /= 0) goto 600
1320	i=i+1
1321	call MPI_Isend(hmix(:,:,:,mind),d2s1,mp_pp,dest,tm1,MPI_COMM_WORLD,reqs(i),mp_ierr)
1322	if (mp_ierr /= 0) goto 600
1323	i=i+1
1324	call MPI_Isend(tropopause(:,:,:,mind),d2s1,mp_pp,dest,tm1,MPI_COMM_WORLD,reqs(i),mp_ierr)
1325	if (mp_ierr /= 0) goto 600
1326	i=i+1
1327	call MPI_Isend(oli(:,:,:,mind),d2s1,mp_pp,dest,tm1,MPI_COMM_WORLD,reqs(i),mp_ierr)
1328	if (mp_ierr /= 0) goto 600
1329
1330	! Send cloud water if it exists. Increment counter always (as on receiving end)
1331	if (readclouds) then
1332	i=i+1
1333	call MPI_Isend(clwc(:,:,:,mind),d3s1,mp_pp,dest,tm1,&
1334	&MPI_COMM_WORLD,reqs(i),mp_ierr)
1335	if (mp_ierr /= 0) goto 600
1336	i=i+1
1337
1338	call MPI_Isend(ciwc(:,:,:,mind),d3s1,mp_pp,dest,tm1,&
1339	&MPI_COMM_WORLD,reqs(i),mp_ierr)
1340	if (mp_ierr /= 0) goto 600
1341	! else
1342	! i=i+2
1343	end if
1344
1345	end do
1346
1347	if (mp_measure_time) call mpif_mtime('commtime',1)
1348
1349	goto 601
1350
1351	600 write(,) "#### mpi_mod::mpif_gf_send_vars_async> mp_ierr \= 0", mp_ierr
1352	stop
1353
1354	601 end subroutine mpif_gf_send_vars_async
1355
1356
1357	subroutine mpif_gf_recv_vars_async(memstat)
1358	!*******************************************************************************
1359	! DESCRIPTION
1360	! Receive 'getfield' variables from reader process.
1361	! Called from timemanager by all processes except reader
1362	!
1363	! NOTE
1364	! This version uses asynchronious communications.
1365	!
1366	! VARIABLES
1367	! memstat -- input, used to resolve windfield index being received
1368	!
1369	! TODO
1370	! Transfer cloud water/ice
1371	!
1372	!*******************************************************************************
1373	use com_mod
1374
1375	implicit none
1376
1377	integer, intent(in) :: memstat
1378	integer :: mind,j
1379
1380	! Common array sizes used for communications
1381	integer :: d3s1 = nxmaxnymaxnzmax
1382	integer :: d3s2 = nxmaxnymaxnuvzmax
1383	integer :: d2s1 = nxmax*nymax
1384	integer :: d2s2 = nxmaxnymaxmaxspec
1385	integer :: d1_size1 = maxwf
1386
1387	! integer :: d3s1,d3s2,d2s1,d2s2
1388	!*******************************************************************************
1389
1390	! :TODO: don't need these
1391	! d3s1=d3_size1
1392	! d3s2=d3_size2
1393	! d2s1=d2_size1
1394	! d2s2=d2_size2
1395
1396	! At the time this immediate receive is posted, memstat is the state of
1397	! windfield indices at the previous/current time. From this, the future
1398	! state is deduced.
1399	if (memstat.eq.32) then
1400	! last read was synchronous to indices 1 and 2, 3 is free
1401	mind=3
1402	else if (memstat.eq.17) then
1403	! last read was asynchronous to index 3, 1 is free
1404	mind=1
1405	else if (memstat.eq.18) then
1406	! last read was asynchronous to index 1, 2 is free
1407	mind=2
1408	else if (memstat.eq.19) then
1409	! last read was asynchronous to index 2, 3 is free
1410	mind=3
1411	end if
1412
1413	if (mp_dev_mode) then
1414	if (mp_pid.eq.id_read) then
1415	write(,) 'MPI_DEV_MODE: error calling mpif_gf_recv_vars_async'
1416	end if
1417	end if
1418
1419	! Time for MPI communications
1420	if (mp_measure_time) call mpif_mtime('commtime',0)
1421
1422	if (mp_dev_mode) write(,) '## in mpif_gf_send_vars_async, memstat:', memstat
1423
1424	! Initiate receiving of data
1425	!***************************
1426
1427	! Get MPI tags/requests for communications
1428	j=mp_pid*nvar_async
1429	call MPI_Irecv(uu(:,:,:,mind),d3s1,mp_pp,id_read,MPI_ANY_TAG,&
1430	&MPI_COMM_WORLD,reqs(j),mp_ierr)
1431	if (mp_ierr /= 0) goto 600
1432	j=j+1
1433	call MPI_Irecv(vv(:,:,:,mind),d3s1,mp_pp,id_read,MPI_ANY_TAG,&
1434	&MPI_COMM_WORLD,reqs(j),mp_ierr)
1435	if (mp_ierr /= 0) goto 600
1436	j=j+1
1437	call MPI_Irecv(uupol(:,:,:,mind),d3s1,mp_pp,id_read,MPI_ANY_TAG,&
1438	&MPI_COMM_WORLD,reqs(j),mp_ierr)
1439	if (mp_ierr /= 0) goto 600
1440	j=j+1
1441	call MPI_Irecv(vvpol(:,:,:,mind),d3s1,mp_pp,id_read,MPI_ANY_TAG,&
1442	&MPI_COMM_WORLD,reqs(j),mp_ierr)
1443	if (mp_ierr /= 0) goto 600
1444	j=j+1
1445	call MPI_Irecv(ww(:,:,:,mind),d3s1,mp_pp,id_read,MPI_ANY_TAG,&
1446	&MPI_COMM_WORLD,reqs(j),mp_ierr)
1447	if (mp_ierr /= 0) goto 600
1448	j=j+1
1449	call MPI_Irecv(tt(:,:,:,mind),d3s1,mp_pp,id_read,MPI_ANY_TAG,&
1450	&MPI_COMM_WORLD,reqs(j),mp_ierr)
1451	if (mp_ierr /= 0) goto 600
1452	j=j+1
1453	call MPI_Irecv(rho(:,:,:,mind),d3s1,mp_pp,id_read,MPI_ANY_TAG,&
1454	&MPI_COMM_WORLD,reqs(j),mp_ierr)
1455	if (mp_ierr /= 0) goto 600
1456	j=j+1
1457	call MPI_Irecv(drhodz(:,:,:,mind),d3s1,mp_pp,id_read,MPI_ANY_TAG,&
1458	&MPI_COMM_WORLD,reqs(j),mp_ierr)
1459	if (mp_ierr /= 0) goto 600
1460	j=j+1
1461	call MPI_Irecv(tth(:,:,:,mind),d3s2,mp_pp,id_read,MPI_ANY_TAG,&
1462	&MPI_COMM_WORLD,reqs(j),mp_ierr)
1463	if (mp_ierr /= 0) goto 600
1464	j=j+1
1465	call MPI_Irecv(qvh(:,:,:,mind),d3s2,mp_pp,id_read,MPI_ANY_TAG,&
1466	&MPI_COMM_WORLD,reqs(j),mp_ierr)
1467	if (mp_ierr /= 0) goto 600
1468	j=j+1
1469
1470	call MPI_Irecv(qv(:,:,:,mind),d3s1,mp_pp,id_read,MPI_ANY_TAG,&
1471	&MPI_COMM_WORLD,reqs(j),mp_ierr)
1472	if (mp_ierr /= 0) goto 600
1473	j=j+1
1474	call MPI_Irecv(pv(:,:,:,mind),d3s1,mp_pp,id_read,MPI_ANY_TAG,&
1475	&MPI_COMM_WORLD,reqs(j),mp_ierr)
1476	if (mp_ierr /= 0) goto 600
1477	j=j+1
1478	call MPI_Irecv(clouds(:,:,:,mind),d3s1,MPI_INTEGER1,id_read,MPI_ANY_TAG,&
1479	&MPI_COMM_WORLD,reqs(j),mp_ierr)
1480	if (mp_ierr /= 0) goto 600
1481	j=j+1
1482
1483	call MPI_Irecv(cloudsh(:,:,mind),d2s1,mp_pp,id_read,MPI_ANY_TAG,&
1484	&MPI_COMM_WORLD,reqs(j),mp_ierr)
1485	if (mp_ierr /= 0) goto 600
1486	j=j+1
1487	call MPI_Irecv(vdep(:,:,:,mind),d2s2,mp_pp,id_read,MPI_ANY_TAG,&
1488	&MPI_COMM_WORLD,reqs(j),mp_ierr)
1489	if (mp_ierr /= 0) goto 600
1490	j=j+1
1491	call MPI_Irecv(ps(:,:,:,mind),d2s1,mp_pp,id_read,MPI_ANY_TAG,&
1492	&MPI_COMM_WORLD,reqs(j),mp_ierr)
1493	if (mp_ierr /= 0) goto 600
1494	j=j+1
1495	call MPI_Irecv(sd(:,:,:,mind),d2s1,mp_pp,id_read,MPI_ANY_TAG,&
1496	&MPI_COMM_WORLD,reqs(j),mp_ierr)
1497	if (mp_ierr /= 0) goto 600
1498	j=j+1
1499	call MPI_Irecv(tcc(:,:,:,mind),d2s1,mp_pp,id_read,MPI_ANY_TAG,&
1500	&MPI_COMM_WORLD,reqs(j),mp_ierr)
1501	if (mp_ierr /= 0) goto 600
1502	j=j+1
1503	call MPI_Irecv(tt2(:,:,:,mind),d2s1,mp_pp,id_read,MPI_ANY_TAG,&
1504	&MPI_COMM_WORLD,reqs(j),mp_ierr)
1505	if (mp_ierr /= 0) goto 600
1506	j=j+1
1507	call MPI_Irecv(td2(:,:,:,mind),d2s1,mp_pp,id_read,MPI_ANY_TAG,&
1508	&MPI_COMM_WORLD,reqs(j),mp_ierr)
1509	if (mp_ierr /= 0) goto 600
1510	j=j+1
1511	call MPI_Irecv(lsprec(:,:,:,mind),d2s1,mp_pp,id_read,MPI_ANY_TAG,&
1512	&MPI_COMM_WORLD,reqs(j),mp_ierr)
1513	if (mp_ierr /= 0) goto 600
1514	j=j+1
1515	call MPI_Irecv(convprec(:,:,:,mind),d2s1,mp_pp,id_read,MPI_ANY_TAG,&
1516	&MPI_COMM_WORLD,reqs(j),mp_ierr)
1517	if (mp_ierr /= 0) goto 600
1518
1519	call MPI_Irecv(ustar(:,:,:,mind),d2s1,mp_pp,id_read,MPI_ANY_TAG,&
1520	&MPI_COMM_WORLD,reqs(j),mp_ierr)
1521	if (mp_ierr /= 0) goto 600
1522	j=j+1
1523	call MPI_Irecv(wstar(:,:,:,mind),d2s1,mp_pp,id_read,MPI_ANY_TAG,&
1524	&MPI_COMM_WORLD,reqs(j),mp_ierr)
1525	if (mp_ierr /= 0) goto 600
1526	j=j+1
1527	call MPI_Irecv(hmix(:,:,:,mind),d2s1,mp_pp,id_read,MPI_ANY_TAG,&
1528	&MPI_COMM_WORLD,reqs(j),mp_ierr)
1529	if (mp_ierr /= 0) goto 600
1530	j=j+1
1531	call MPI_Irecv(tropopause(:,:,:,mind),d2s1,mp_pp,id_read,MPI_ANY_TAG,&
1532	&MPI_COMM_WORLD,reqs(j),mp_ierr)
1533	if (mp_ierr /= 0) goto 600
1534	j=j+1
1535	call MPI_Irecv(oli(:,:,:,mind),d2s1,mp_pp,id_read,MPI_ANY_TAG,&
1536	&MPI_COMM_WORLD,reqs(j),mp_ierr)
1537	if (mp_ierr /= 0) goto 600
1538
1539
1540	! Post request for clwc. These data are possibly not sent, request must then be cancelled
1541	! For now assume that data at all steps either have or do not have water
1542	if (readclouds) then
1543	j=j+1
1544	call MPI_Irecv(clwc(:,:,:,mind),d3s1,mp_pp,id_read,MPI_ANY_TAG,&
1545	&MPI_COMM_WORLD,reqs(j),mp_ierr)
1546	if (mp_ierr /= 0) goto 600
1547	j=j+1
1548	call MPI_Irecv(ciwc(:,:,:,mind),d3s1,mp_pp,id_read,MPI_ANY_TAG,&
1549	&MPI_COMM_WORLD,reqs(j),mp_ierr)
1550	if (mp_ierr /= 0) goto 600
1551	end if
1552
1553
1554	if (mp_measure_time) call mpif_mtime('commtime',1)
1555
1556	goto 601
1557
1558	600 write(,) "#### mpi_mod::mpif_gf_recv_vars_async> MPI ERROR ", mp_ierr
1559	stop
1560
1561	601 end subroutine mpif_gf_recv_vars_async
1562
1563
1564	subroutine mpif_gf_request
1565	!*******************************************************************************
1566	! DESCRIPTION
1567	! Check for completion of MPI_Isend/MPI_Irecv data transfer.
1568	!
1569	!
1570	! NOTE
1571	! implicit synchronisation between all processes takes place here
1572	!
1573	!
1574	!*******************************************************************************
1575	use com_mod, only: readclouds
1576
1577	implicit none
1578
1579
1580	integer :: n_req
1581	integer :: i
1582
1583	!***********************************************************************
1584
1585	n_req=nvar_async*mp_np
1586
1587	if (mp_measure_time) call mpif_mtime('commtime',0)
1588
1589	! call MPI_Wait(rm1,MPI_STATUS_IGNORE,mp_ierr)
1590
1591	! TODO: cancel recv request if readclouds=.false.
1592	! if (readclouds) then
1593	call MPI_Waitall(n_req,reqs,MPI_STATUSES_IGNORE,mp_ierr)
1594	! endif
1595	! else
1596	! do i = 0, nvar_async*mp_np-1
1597	! if (mod(i,27).eq.0 .or. mod(i,28).eq.0) then
1598	! call MPI_Cancel(reqs(i),mp_ierr)
1599	! cycle
1600	! end if
1601	! call MPI_Wait(reqs(i),MPI_STATUS_IGNORE,mp_ierr)
1602	! end do
1603	! end if
1604
1605	if (mp_ierr /= 0) goto 600
1606
1607	if (mp_measure_time) call mpif_mtime('commtime',1)
1608
1609	goto 601
1610
1611	600 write(,) "#### mpi_mod::mpif_gf_request> MPI ERROR ", mp_ierr
1612	stop
1613
1614	601 end subroutine mpif_gf_request
1615
1616
1617	subroutine mpif_tm_reduce_grid
1618	!***********************************************************************
1619	! Collect grid variable to PID 0, adding from all processes.
1620	!
1621	! NOTE
1622	! - Older MPI implementations may lack the MPI_IN_PLACE operation.
1623	! If so use 1) below and change unc_mod
1624	!
1625	!***********************************************************************
1626	use com_mod
1627	use unc_mod
1628	use par_mod
1629
1630	implicit none
1631
1632	integer :: grid_size2d,grid_size3d
1633	integer, parameter :: rcpt_size=maxreceptor*maxspec
1634
1635	!**********************************************************************
1636	grid_size2d=numxgridnumygridmaxspec* &
1637	& maxpointspec_actnclassuncmaxageclass
1638	grid_size3d=numxgridnumygridnumzgridmaxspec &
1639	& maxpointspec_actnclassuncmaxageclass
1640
1641
1642	! Time for MPI communications
1643	if (mp_measure_time) call mpif_mtime('commtime',0)
1644
1645	! 1) Using a separate grid (gridunc0) for received values
1646	! call MPI_Reduce(gridunc, gridunc0, grid_size3d, mp_pp, MPI_SUM, id_root, &
1647	! & mp_comm_used, mp_ierr)
1648	! if (mp_ierr /= 0) goto 600
1649
1650	! 2) Using in-place reduction
1651	if (lroot) then
1652	call MPI_Reduce(MPI_IN_PLACE, gridunc, grid_size3d, mp_pp, MPI_SUM, id_root, &
1653	& mp_comm_used, mp_ierr)
1654	if (mp_ierr /= 0) goto 600
1655	else
1656	call MPI_Reduce(gridunc, gridunc, grid_size3d, mp_pp, MPI_SUM, id_root, &
1657	& mp_comm_used, mp_ierr)
1658	end if
1659
1660	if ((WETDEP).and.(ldirect.gt.0)) then
1661	call MPI_Reduce(wetgridunc, wetgridunc0, grid_size2d, mp_pp, MPI_SUM, id_root, &
1662	& mp_comm_used, mp_ierr)
1663	if (mp_ierr /= 0) goto 600
1664	end if
1665
1666	if ((DRYDEP).and.(ldirect.gt.0)) then
1667	call MPI_Reduce(drygridunc, drygridunc0, grid_size2d, mp_pp, MPI_SUM, id_root, &
1668	& mp_comm_used, mp_ierr)
1669	if (mp_ierr /= 0) goto 600
1670	end if
1671
1672	! Receptor concentrations
1673	if (lroot) then
1674	call MPI_Reduce(MPI_IN_PLACE,creceptor,rcpt_size,mp_pp,MPI_SUM,id_root, &
1675	& mp_comm_used,mp_ierr)
1676	if (mp_ierr /= 0) goto 600
1677	else
1678	call MPI_Reduce(creceptor,creceptor,rcpt_size,mp_pp,MPI_SUM,id_root, &
1679	& mp_comm_used,mp_ierr)
1680	end if
1681
1682	if (mp_measure_time) call mpif_mtime('commtime',1)
1683
1684	goto 601
1685
1686	600 write(,) "mpi_mod> mp_ierr \= 0", mp_ierr
1687	stop
1688
1689	601 end subroutine mpif_tm_reduce_grid
1690
1691
1692	subroutine mpif_tm_reduce_grid_nest
1693	!***********************************************************************
1694	! Collect nested grids to PID 0, adding from all processes.
1695	!
1696	! NOTE
1697	! - Compiling with 'fcheck=all' (gfortran) will cause a run-time error
1698	! as wetgriduncn0 is only allocated for root process. Possibly valid
1699	! MPI code but invalid gfortran.
1700	!
1701	!***********************************************************************
1702	use com_mod
1703	use unc_mod
1704	use par_mod
1705
1706	implicit none
1707
1708	integer :: grid_size2d,grid_size3d
1709
1710	!**********************************************************************
1711
1712	grid_size3d=numxgridnnumygridnnumzgridmaxspec &
1713	& maxpointspec_actnclassuncmaxageclass
1714	grid_size2d=numxgridnnumygridnmaxspec* &
1715	& maxpointspec_actnclassuncmaxageclass
1716
1717
1718	! Time for MPI communications
1719	if (mp_measure_time) call mpif_mtime('commtime',0)
1720
1721	! Using a separate grid (gridunc0) for received values, for debugging
1722	! call MPI_Reduce(griduncn, griduncn0, grid_size3d, mp_pp, MPI_SUM, id_root, &
1723	! & mp_comm_used, mp_ierr)
1724	! if (mp_ierr /= 0) goto 600
1725
1726	! Using in-place reduction
1727	if (lroot) then
1728	call MPI_Reduce(MPI_IN_PLACE, griduncn, grid_size3d, mp_pp, MPI_SUM, id_root, &
1729	& mp_comm_used, mp_ierr)
1730	if (mp_ierr /= 0) goto 600
1731	else
1732	call MPI_Reduce(griduncn, griduncn, grid_size3d, mp_pp, MPI_SUM, id_root, &
1733	& mp_comm_used, mp_ierr)
1734	end if
1735
1736	if ((WETDEP).and.(ldirect.gt.0)) then
1737	call MPI_Reduce(wetgriduncn, wetgriduncn0, grid_size2d, mp_pp, MPI_SUM, id_root, &
1738	& mp_comm_used, mp_ierr)
1739	if (mp_ierr /= 0) goto 600
1740	end if
1741
1742	if ((DRYDEP).and.(ldirect.gt.0)) then
1743	call MPI_Reduce(drygriduncn, drygriduncn0, grid_size2d, mp_pp, MPI_SUM, id_root, &
1744	& mp_comm_used, mp_ierr)
1745	if (mp_ierr /= 0) goto 600
1746	end if
1747
1748	if (mp_measure_time) call mpif_mtime('commtime',1)
1749
1750	goto 601
1751
1752	600 write(,) "mpi_mod> mp_ierr \= 0", mp_ierr
1753	stop
1754
1755	601 end subroutine mpif_tm_reduce_grid_nest
1756
1757
1758	subroutine mpif_mtime(ident,imode)
1759	!***********************************************************************
1760	! Measure CPU/Wall time in various parts of the code
1761	!
1762	! VARIABLES
1763	! ident character, identifies routine to measure
1764	! imode integer, 0:start clock(s) 1: stop clock(s)
1765	!
1766	!***********************************************************************
1767	implicit none
1768
1769	character(LEN=*), intent(in) :: ident
1770	integer, intent(in) :: imode
1771
1772	!***********************************************************************
1773
1774	select case(ident)
1775
1776	case ('timemanager')
1777	if (imode.eq.0) then
1778	call cpu_time(tm_tot_beg)
1779	else
1780	call cpu_time(tm_tot_end)
1781	tm_tot_total = tm_tot_total + (tm_tot_end-tm_tot_beg)
1782	end if
1783
1784	case ('wetdepo')
1785	if (imode.eq.0) then
1786	mp_wetdepo_wtime_beg = mpi_wtime()
1787	call cpu_time(mp_wetdepo_time_beg)
1788	else
1789	mp_wetdepo_wtime_end = mpi_wtime()
1790	call cpu_time(mp_wetdepo_time_end)
1791
1792	mp_wetdepo_wtime_total = mp_wetdepo_wtime_total + &
1793	&(mp_wetdepo_wtime_end - mp_wetdepo_wtime_beg)
1794	mp_wetdepo_time_total = mp_wetdepo_time_total + &
1795	&(mp_wetdepo_time_end - mp_wetdepo_time_beg)
1796	end if
1797
1798	case ('getfields')
1799	if (imode.eq.0) then
1800	mp_getfields_wtime_beg = mpi_wtime()
1801	call cpu_time(mp_getfields_time_beg)
1802	else
1803	mp_getfields_wtime_end = mpi_wtime()
1804	call cpu_time(mp_getfields_time_end)
1805
1806	mp_getfields_wtime_total = mp_getfields_wtime_total + &
1807	&(mp_getfields_wtime_end - mp_getfields_wtime_beg)
1808	mp_getfields_time_total = mp_getfields_time_total + &
1809	&(mp_getfields_time_end - mp_getfields_time_beg)
1810	end if
1811
1812	case ('partloop1')
1813	if (imode.eq.0) then
1814	call cpu_time(tm_nploop_beg)
1815	else
1816	call cpu_time(tm_nploop_end)
1817	tm_nploop_total = tm_nploop_total + (tm_nploop_end - tm_nploop_beg)
1818	end if
1819
1820	case ('conccalc')
1821	if (imode.eq.0) then
1822	call cpu_time(mp_conccalc_time_beg)
1823	else
1824	call cpu_time(mp_conccalc_time_end)
1825	mp_conccalc_time_total = mp_conccalc_time_total + mp_conccalc_time_end - &
1826	&mp_conccalc_time_beg
1827	end if
1828	case ('rootonly')
1829	if (imode.eq.0) then
1830	call cpu_time(mp_root_time_beg)
1831	mp_root_wtime_beg = mpi_wtime()
1832	else
1833	call cpu_time(mp_root_time_end)
1834	mp_root_wtime_end = mpi_wtime()
1835	mp_root_time_total = mp_root_time_total + &
1836	&(mp_root_time_end - mp_root_time_beg)
1837	mp_root_wtime_total = mp_root_wtime_total + &
1838	&(mp_root_wtime_end - mp_root_wtime_beg)
1839	end if
1840
1841	case ('iotime')
1842	if (imode.eq.0) then
1843	mp_io_wtime_beg = mpi_wtime()
1844	call cpu_time(mp_io_time_beg)
1845	else
1846	mp_io_wtime_end = mpi_wtime()
1847	call cpu_time(mp_io_time_end)
1848
1849	mp_io_wtime_total = mp_io_wtime_total + (mp_io_wtime_end - &
1850	& mp_io_wtime_beg)
1851	mp_io_time_total = mp_io_time_total + (mp_io_time_end - &
1852	& mp_io_time_beg)
1853	end if
1854
1855	case ('verttransform')
1856	if (imode.eq.0) then
1857	mp_vt_wtime_beg = mpi_wtime()
1858	call cpu_time(mp_vt_time_beg)
1859	else
1860	mp_vt_wtime_end = mpi_wtime()
1861	call cpu_time(mp_vt_time_end)
1862
1863	mp_vt_wtime_total = mp_vt_wtime_total + (mp_vt_wtime_end - &
1864	& mp_vt_wtime_beg)
1865	mp_vt_time_total = mp_vt_time_total + (mp_vt_time_end - &
1866	& mp_vt_time_beg)
1867	end if
1868
1869	case ('readwind')
1870	if (imode.eq.0) then
1871	call cpu_time(mp_readwind_time_beg)
1872	mp_readwind_wtime_beg = mpi_wtime()
1873	else
1874	call cpu_time(mp_readwind_time_end)
1875	mp_readwind_wtime_end = mpi_wtime()
1876
1877	mp_readwind_time_total = mp_readwind_time_total + &
1878	&(mp_readwind_time_end - mp_readwind_time_beg)
1879	mp_readwind_wtime_total = mp_readwind_wtime_total + &
1880	&(mp_readwind_wtime_end - mp_readwind_wtime_beg)
1881	end if
1882
1883	case ('commtime')
1884	if (imode.eq.0) then
1885	call cpu_time(mp_comm_time_beg)
1886	mp_comm_wtime_beg = mpi_wtime()
1887	else
1888	call cpu_time(mp_comm_time_end)
1889	mp_comm_wtime_end = mpi_wtime()
1890	mp_comm_time_total = mp_comm_time_total + &
1891	&(mp_comm_time_end - mp_comm_time_beg)
1892	mp_comm_wtime_total = mp_comm_wtime_total + &
1893	&(mp_comm_wtime_end - mp_comm_wtime_beg)
1894	end if
1895
1896	case ('flexpart')
1897	if (imode.eq.0) then
1898	mp_total_wtime_beg=mpi_wtime()
1899	else
1900	mp_total_wtime_end=mpi_wtime()
1901	mp_total_wtime_total = mp_total_wtime_end-mp_total_wtime_beg
1902	end if
1903
1904	case default
1905	write(,) 'mpi_mod::mpif_mtime> WARNING: unknown case identifier'
1906
1907	end select
1908
1909
1910	end subroutine mpif_mtime
1911
1912
1913	subroutine mpif_finalize
1914	!***********************************************************************
1915	! Finalize MPI
1916	! Optionally print detailed time diagnostics
1917	!***********************************************************************
1918	implicit none
1919
1920	integer :: ip,j,r
1921
1922	!***********************************************************************
1923
1924	IF (mp_measure_time) THEN
1925	do ip=0, mp_np-1
1926	call MPI_BARRIER(MPI_COMM_WORLD, mp_ierr)
1927
1928	if (ip == mp_pid) then
1929	write(*,FMT='(72("#"))')
1930	write(*,FMT='(A60,I3)') 'STATISTICS FOR MPI PROCESS:', mp_pid
1931	if (ip == 0) then
1932	write(*,FMT='(A60,TR1,F9.2)') 'TOTAL CPU TIME FOR ROOT-ONLY (PID 0) &
1933	&SECTIONS: ', mp_root_time_total
1934	write(*,FMT='(A60,TR1,F9.2)') 'TOTAL WALL TIME FOR ROOT-ONLY (PID 0) &
1935	&SECTIONS:', mp_root_wtime_total
1936	end if
1937	write(*,FMT='(A60,TR1,F9.2)') 'TOTAL WALL TIME FOR FLEXPART: ' &
1938	&, mp_total_wtime_total
1939	write(*,FMT='(A60,TR1,F9.2)') 'TOTAL CPU TIME FOR MPI COMMUNICATIONS:',&
1940	& mp_comm_time_total
1941	write(*,FMT='(A60,TR1,F9.2)') 'TOTAL WALL TIME FOR MPI COMMUNICATIONS:',&
1942	& mp_comm_wtime_total
1943	write(*,FMT='(A60,TR1,F9.2)') 'TOTAL CPU TIME FOR MPI BARRIERS:',&
1944	& mp_barrier_time_total
1945	write(*,FMT='(A60,TR1,F9.2)') 'TOTAL WALL TIME FOR MPI BARRIERS:',&
1946	& mp_barrier_wtime_total
1947	write(*,FMT='(A60,TR1,F9.2)') 'TOTAL CPU TIME FOR TIMEMANAGER:',&
1948	& tm_tot_total
1949	write(*,FMT='(A60,TR1,F9.2)') 'TOTAL CPU TIME FOR TIMEMANAGER/NUMPART LOOP:',&
1950	& tm_nploop_total
1951	write(*,FMT='(A60,TR1,F9.2)') 'TOTAL WALL TIME FOR ADVANCE:',&
1952	& mp_advance_wtime_total
1953	write(*,FMT='(A60,TR1,F9.2)') 'TOTAL WALL TIME FOR GETFIELDS:',&
1954	& mp_getfields_wtime_total
1955	write(*,FMT='(A60,TR1,F9.2)') 'TOTAL CPU TIME FOR GETFIELDS:',&
1956	& mp_getfields_time_total
1957	write(*,FMT='(A60,TR1,F9.2)') 'TOTAL WALL TIME FOR READWIND:',&
1958	& mp_readwind_wtime_total
1959	write(*,FMT='(A60,TR1,F9.2)') 'TOTAL CPU TIME FOR READWIND:',&
1960	& mp_readwind_time_total
1961	write(*,FMT='(A60,TR1,F9.2)') 'TOTAL WALL TIME FOR FILE IO:',&
1962	& mp_io_wtime_total
1963	write(*,FMT='(A60,TR1,F9.2)') 'TOTAL CPU TIME FOR FILE IO:',&
1964	& mp_io_time_total
1965	write(*,FMT='(A60,TR1,F9.2)') 'TOTAL WALL TIME FOR WETDEPO:',&
1966	& mp_wetdepo_wtime_total
1967	write(*,FMT='(A60,TR1,F9.2)') 'TOTAL CPU TIME FOR WETDEPO:',&
1968	& mp_wetdepo_time_total
1969	write(*,FMT='(A60,TR1,F9.2)') 'TOTAL WALL TIME FOR CONCCALC:',&
1970	& mp_conccalc_time_total
1971	! write(*,FMT='(A60,TR1,F9.2)') 'TOTAL WALL TIME FOR VERTTRANSFORM:',&
1972	! & mp_vt_wtime_total
1973	! write(*,FMT='(A60,TR1,F9.2)') 'TOTAL CPU TIME FOR VERTTRANSFORM:',&
1974	! & mp_vt_time_total
1975	! NB: the 'flush' function is possibly a gfortran-specific extension
1976	call flush()
1977	end if
1978	end do
1979	end if
1980
1981	! This call to barrier is for correctly formatting output
1982	call MPI_BARRIER(MPI_COMM_WORLD, mp_ierr)
1983
1984	if (lroot.and.mp_measure_time) then
1985	write(*,FMT='(72("#"))')
1986	WRITE(,) "To turn off output of time measurements, set "
1987	WRITE(,) " mp_measure_time=.false."
1988	WRITE(,) "in file mpi_mod.f90"
1989	write(*,FMT='(72("#"))')
1990	end if
1991
1992	! j=mp_pid*nvar_async
1993	! In the implementation with 3 fields, the processes may have posted
1994	! MPI_Irecv requests that should be cancelled here
1995	!! TODO:
1996	! if (.not.lmp_sync) then
1997	! r=mp_pid*nvar_async
1998	! do j=r,r+nvar_async-1
1999	! call MPI_Cancel(j,mp_ierr)
2000	! if (mp_ierr /= 0) write(,) '#### mpif_finalize::MPI_Cancel> ERROR ####'
2001	! end do
2002	! end if
2003
2004	call MPI_FINALIZE(mp_ierr)
2005	if (mp_ierr /= 0) then
2006	write(,) '#### mpif_finalize::MPI_FINALIZE> MPI ERROR ', mp_ierr, ' ####'
2007	stop
2008	end if
2009
2010
2011	end subroutine mpif_finalize
2012
2013
2014	subroutine get_lun(my_lun)
2015	!***********************************************************************
2016	! get_lun:
2017	! Starting from 100, get next free logical unit number
2018	!***********************************************************************
2019
2020	implicit none
2021
2022	integer, intent(inout) :: my_lun
2023	integer, save :: free_lun=100
2024	logical :: exists, iopen
2025
2026	!***********************************************************************
2027
2028	loop1: do
2029	inquire(UNIT=free_lun, EXIST=exists, OPENED=iopen)
2030	if (exists .and. .not.iopen) exit loop1
2031	free_lun = free_lun+1
2032	end do loop1
2033	my_lun = free_lun
2034
2035	end subroutine get_lun
2036
2037
2038	subroutine write_data_dbg(array_in, array_name, tstep, ident)
2039	!***********************************************************************
2040	! Write one-dimensional arrays to file (for debugging purposes)
2041	!***********************************************************************
2042	implicit none
2043
2044	real, intent(in), dimension(:) :: array_in
2045	integer, intent(in) :: tstep
2046	integer :: lios
2047	character(LEN=*), intent(in) :: ident, array_name
2048
2049	character(LEN=8) :: c_ts
2050	character(LEN=40) :: fn_1, fn_2
2051
2052	!***********************************************************************
2053
2054	write(c_ts, FMT='(I8.8,BZ)') tstep
2055	fn_1='-'//trim(adjustl(c_ts))//'-'//trim(ident)
2056	write(c_ts, FMT='(I2.2,BZ)') mp_np
2057	fn_2= trim(adjustl(array_name))//trim(adjustl(fn_1))//'-np'//trim(adjustl(c_ts))//'.dat'
2058
2059	call get_lun(dat_lun)
2060	open(UNIT=dat_lun, FILE=fn_2, IOSTAT=lios, ACTION='WRITE', &
2061	FORM='UNFORMATTED', STATUS='REPLACE')
2062	write(UNIT=dat_lun, IOSTAT=lios) array_in
2063	close(UNIT=dat_lun)
2064
2065	end subroutine write_data_dbg
2066
2067
2068	end module mpi_mod

Note: See TracBrowser for help on using the repository browser.

Download in other formats:

Original Format