-
Notifications
You must be signed in to change notification settings - Fork 22
/
Copy pathmp_wave.f90
723 lines (554 loc) · 21.5 KB
/
mp_wave.f90
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
!
! Copyright (C) 2002-2008 Quantum ESPRESSO group
! This file is distributed under the terms of the
! GNU General Public License. See the file `License'
! in the root directory of the present distribution,
! or http://www.gnu.org/copyleft/gpl.txt .
!
!
MODULE mp_wave
IMPLICIT NONE
SAVE
CONTAINS
SUBROUTINE mergewf ( pw, pwt, ngwl, ig_l2g, mpime, nproc, root, comm )
! ... This subroutine merges the pieces of a wave functions (pw) splitted across
! ... processors into a total wave function (pwt) containing al the components
! ... in a pre-defined order (the same as if only one processor is used)
USE kinds
USE parallel_include
IMPLICIT NONE
COMPLEX(DP), intent(in) :: PW(:)
COMPLEX(DP), intent(out) :: PWT(:)
INTEGER, INTENT(IN) :: mpime ! index of the calling processor ( starting from 0 )
INTEGER, INTENT(IN) :: nproc ! number of processors
INTEGER, INTENT(IN) :: root ! root processor ( the one that should receive the data )
INTEGER, INTENT(IN) :: comm ! communicator
INTEGER, INTENT(IN) :: ig_l2g(:)
INTEGER, INTENT(IN) :: ngwl
INTEGER, ALLOCATABLE :: ig_ip(:)
COMPLEX(DP), ALLOCATABLE :: pw_ip(:)
INTEGER :: ierr, i, ip, ngw_ip, ngw_lmax, itmp, igwx, gid
#if defined __MPI
INTEGER :: istatus(MPI_STATUS_SIZE)
#endif
!
! ... Subroutine Body
!
igwx = MAXVAL( ig_l2g(1:ngwl) )
#if defined __MPI
gid = comm
! ... Get local and global wavefunction dimensions
CALL MPI_ALLREDUCE( ngwl, ngw_lmax, 1, MPI_INTEGER, MPI_MAX, gid, IERR )
CALL MPI_ALLREDUCE( igwx, itmp, 1, MPI_INTEGER, MPI_MAX, gid, IERR )
igwx = itmp
#endif
IF( igwx > SIZE( pwt ) ) &
CALL errore(' mergewf ',' wrong size for pwt ',SIZE(pwt) )
#if defined __MPI
DO ip = 1, nproc
IF( (ip-1) /= root ) THEN
! ... In turn each processors send to root the wave components and their indexes in the
! ... global array
IF ( mpime == (ip-1) ) THEN
CALL MPI_SEND( ig_l2g, ngwl, MPI_INTEGER, ROOT, IP, gid, IERR )
CALL MPI_SEND( pw(1), ngwl, MPI_DOUBLE_COMPLEX, ROOT, IP+NPROC, gid, IERR )
END IF
IF ( mpime == root) THEN
ALLOCATE(ig_ip(ngw_lmax))
ALLOCATE(pw_ip(ngw_lmax))
CALL MPI_RECV( ig_ip, ngw_lmax, MPI_INTEGER, (ip-1), IP, gid, istatus, IERR )
CALL MPI_RECV( pw_ip, ngw_lmax, MPI_DOUBLE_COMPLEX, (ip-1), IP+NPROC, gid, istatus, IERR )
CALL MPI_GET_COUNT( istatus, MPI_DOUBLE_COMPLEX, ngw_ip, ierr )
DO I = 1, ngw_ip
PWT(ig_ip(i)) = pw_ip(i)
END DO
DEALLOCATE(ig_ip)
DEALLOCATE(pw_ip)
END IF
ELSE
IF(mpime == root) THEN
DO I = 1, ngwl
PWT(ig_l2g(i)) = pw(i)
END DO
END IF
END IF
CALL MPI_BARRIER( gid, IERR )
END DO
#elif ! defined __MPI
DO I = 1, ngwl
! WRITE( stdout,*) 'MW ', ig_l2g(i), i
PWT( ig_l2g(i) ) = pw(i)
END DO
#else
CALL errore(' MERGEWF ',' no communication protocol ',0)
#endif
RETURN
END SUBROUTINE mergewf
!=----------------------------------------------------------------------------=!
SUBROUTINE splitwf ( pw, pwt, ngwl, ig_l2g, mpime, nproc, root, comm )
! ... This subroutine splits a total wave function (pwt) containing al the components
! ... in a pre-defined order (the same as if only one processor is used), across
! ... processors (pw).
USE kinds
USE parallel_include
IMPLICIT NONE
COMPLEX(DP), INTENT(OUT) :: PW(:)
COMPLEX(DP), INTENT(IN) :: PWT(:)
INTEGER, INTENT(IN) :: mpime, nproc, root
INTEGER, INTENT(IN) :: comm ! communicator
INTEGER, INTENT(IN) :: ig_l2g(:)
INTEGER, INTENT(IN) :: ngwl
INTEGER, ALLOCATABLE :: ig_ip(:)
COMPLEX(DP), ALLOCATABLE :: pw_ip(:)
INTEGER ierr, i, ngw_ip, ip, ngw_lmax, gid, igwx, itmp
#if defined __MPI
integer istatus(MPI_STATUS_SIZE)
#endif
!
! ... Subroutine Body
!
igwx = MAXVAL( ig_l2g(1:ngwl) )
#if defined __MPI
gid = comm
! ... Get local and global wavefunction dimensions
CALL MPI_ALLREDUCE(ngwl, ngw_lmax, 1, MPI_INTEGER, MPI_MAX, gid, IERR )
CALL MPI_ALLREDUCE(igwx, itmp , 1, MPI_INTEGER, MPI_MAX, gid, IERR )
igwx = itmp
#endif
IF( igwx > SIZE( pwt ) ) &
CALL errore(' splitwf ',' wrong size for pwt ',SIZE(pwt) )
#if defined __MPI
DO ip = 1, nproc
! ... In turn each processor send to root the the indexes of its wavefunction conponents
! ... Root receive the indexes and send the componens of the wavefunction read from the disk (pwt)
IF ( (ip-1) /= root ) THEN
IF ( mpime == (ip-1) ) THEN
CALL MPI_SEND( ig_l2g, ngwl, MPI_INTEGER, ROOT, IP, gid,IERR)
CALL MPI_RECV( pw(1), ngwl, MPI_DOUBLE_COMPLEX, ROOT, IP+NPROC, gid, istatus, IERR )
END IF
IF ( mpime == root ) THEN
ALLOCATE(ig_ip(ngw_lmax))
ALLOCATE(pw_ip(ngw_lmax))
CALL MPI_RECV( ig_ip, ngw_lmax, MPI_INTEGER, (ip-1), IP, gid, istatus, IERR )
CALL MPI_GET_COUNT(istatus, MPI_INTEGER, ngw_ip, ierr)
DO i = 1, ngw_ip
pw_ip(i) = PWT(ig_ip(i))
END DO
CALL MPI_SEND( pw_ip, ngw_ip, MPI_DOUBLE_COMPLEX, (ip-1), IP+NPROC, gid, IERR )
DEALLOCATE(ig_ip)
DEALLOCATE(pw_ip)
END IF
ELSE
IF ( mpime == root ) THEN
DO i = 1, ngwl
pw(i) = PWT(ig_l2g(i))
END DO
END IF
END IF
CALL MPI_BARRIER(gid, IERR)
END DO
#elif ! defined __MPI
DO I = 1, ngwl
pw(i) = pwt( ig_l2g(i) )
END DO
#else
CALL errore(' SPLITWF ',' no communication protocol ',0)
#endif
RETURN
END SUBROUTINE splitwf
SUBROUTINE mergeig(igl, igtot, ngl, mpime, nproc, root, comm)
! ... This subroutine merges the pieces of a vector splitted across
! ... processors into a total vector (igtot) containing al the components
! ... in a pre-defined order (the same as if only one processor is used)
USE kinds
USE parallel_include
IMPLICIT NONE
INTEGER, intent(in) :: igl(:)
INTEGER, intent(out) :: igtot(:)
INTEGER, INTENT(IN) :: mpime ! index of the calling processor ( starting from 0 )
INTEGER, INTENT(IN) :: nproc ! number of processors
INTEGER, INTENT(IN) :: root ! root processor ( the one that should receive the data )
INTEGER, INTENT(IN) :: comm ! communicator
INTEGER, INTENT(IN) :: ngl
INTEGER, ALLOCATABLE :: ig_ip(:)
INTEGER :: ierr, i, ip, ng_ip, ng_lmax, ng_g, gid, igs
#if defined __MPI
INTEGER :: istatus(MPI_STATUS_SIZE)
#endif
#if defined __MPI
gid = comm
! ... Get local and global wavefunction dimensions
CALL MPI_ALLREDUCE( ngl, ng_lmax, 1, MPI_INTEGER, MPI_MAX, gid, IERR )
CALL MPI_ALLREDUCE( ngl, ng_g , 1, MPI_INTEGER, MPI_SUM, gid, IERR )
IF( ng_g > SIZE( igtot ) ) THEN
CALL errore(' mergeig ',' wrong size for igtot ',SIZE(igtot) )
END IF
igs = 1
DO ip = 1, nproc
IF( (ip-1) /= root ) THEN
! ... In turn each processors send to root the wave components and their indexes in the
! ... global array
IF ( mpime == (ip-1) ) THEN
CALL MPI_SEND( igl(1), ngl, MPI_INTEGER, ROOT, IP, gid, IERR )
END IF
IF ( mpime == root) THEN
ALLOCATE( ig_ip(ng_lmax) )
CALL MPI_RECV( ig_ip, ng_lmax, MPI_INTEGER, (ip-1), IP, gid, istatus, IERR )
CALL MPI_GET_COUNT( istatus, MPI_INTEGER, ng_ip, ierr )
DO i = 1, ng_ip
igtot( igs + i - 1 ) = ig_ip( i )
END DO
DEALLOCATE(ig_ip)
END IF
ELSE
IF(mpime == root) THEN
ng_ip = ngl
DO i = 1, ngl
igtot( igs + i - 1 ) = igl( i )
END DO
END IF
END IF
IF(mpime == root) THEN
igs = igs + ng_ip
END IF
CALL MPI_BARRIER( gid, IERR )
END DO
#elif ! defined __MPI
igtot( 1:ngl ) = igl( 1:ngl )
#else
CALL errore(' mergeig ',' no communication protocol ',0)
#endif
RETURN
END SUBROUTINE mergeig
!=----------------------------------------------------------------------------=!
SUBROUTINE splitig(igl, igtot, ngl, mpime, nproc, root, comm)
! ... This subroutine splits a replicated vector (igtot) stored on the root proc
! ... across processors (igl).
USE kinds
USE parallel_include
IMPLICIT NONE
INTEGER, INTENT(OUT) :: igl(:)
INTEGER, INTENT(IN) :: igtot(:)
INTEGER, INTENT(IN) :: mpime, nproc, root
INTEGER, INTENT(IN) :: comm ! communicator
INTEGER, INTENT(IN) :: ngl
INTEGER ierr, i, ng_ip, ip, ng_lmax, ng_g, gid, igs
#if defined __MPI
integer istatus(MPI_STATUS_SIZE)
#endif
INTEGER, ALLOCATABLE :: ig_ip(:)
#if defined __MPI
gid = comm
! ... Get local and global wavefunction dimensions
CALL MPI_ALLREDUCE(ngl, ng_lmax, 1, MPI_INTEGER, MPI_MAX, gid, IERR )
CALL MPI_ALLREDUCE(ngl, ng_g , 1, MPI_INTEGER, MPI_SUM, gid, IERR )
IF( ng_g > SIZE( igtot ) ) THEN
CALL errore(' splitig ',' wrong size for igtot ', SIZE(igtot) )
END IF
igs = 1
DO ip = 1, nproc
! ... In turn each processor sends to root the indices of its wavefunction components
! ... Root receives the indices and sends the components of the wavefunction read from the disk (pwt)
IF ( (ip-1) /= root ) THEN
IF ( mpime == (ip-1) ) THEN
CALL MPI_SEND( ngl, 1 , MPI_INTEGER, ROOT, IP, gid,IERR)
CALL MPI_RECV( igl, ngl, MPI_INTEGER, ROOT, IP+NPROC, gid, istatus, IERR )
END IF
IF ( mpime == root ) THEN
ALLOCATE(ig_ip(ng_lmax))
CALL MPI_RECV( ng_ip, 1, MPI_INTEGER, (ip-1), IP, gid, istatus, IERR )
DO i = 1, ng_ip
ig_ip(i) = igtot( igs + i - 1)
END DO
CALL MPI_SEND( ig_ip, ng_ip, MPI_INTEGER, (ip-1), IP+NPROC, gid, IERR )
DEALLOCATE(ig_ip)
END IF
ELSE
IF ( mpime == root ) THEN
ng_ip = ngl
DO i = 1, ng_ip
igl(i) = igtot( igs + i - 1)
END DO
END IF
END IF
IF( mpime == root ) igs = igs + ng_ip
CALL MPI_BARRIER(gid, IERR)
END DO
#elif ! defined __MPI
igl( 1:ngl ) = igtot( 1:ngl )
#else
CALL errore(' splitig ',' no communication protocol ',0)
#endif
RETURN
END SUBROUTINE splitig
!=----------------------------------------------------------------------------=!
SUBROUTINE pwscatter( c, ctmp, ngw, indi_l, sour_indi, dest_indi, &
n_indi_rcv, n_indi_snd, icntix, mpime, nproc, group )
USE kinds
USE parallel_include
implicit none
integer :: indi_l(:) ! list of G-vec index to be exchanged
integer :: sour_indi(:) ! the list of source processors
integer :: dest_indi(:) ! the list of destination processors
integer :: n_indi_rcv ! number of G-vectors to be received
integer :: n_indi_snd ! number of G-vectors to be sent
integer :: icntix ! total number of G-vec to be exchanged
INTEGER, INTENT(IN) :: nproc, mpime, group
COMPLEX(DP) :: c(:)
COMPLEX(DP) :: ctmp(:)
integer :: ngw
integer :: ig, icsize
INTEGER :: me, idest, isour, ierr
COMPLEX(DP), ALLOCATABLE :: my_buffer( : )
COMPLEX(DP), ALLOCATABLE :: mp_snd_buffer( : )
COMPLEX(DP), ALLOCATABLE :: mp_rcv_buffer( : )
INTEGER, ALLOCATABLE :: ibuf(:)
!
! ... SUBROUTINE BODY
!
me = mpime + 1
if( icntix .lt. 1 ) then
icsize = 1
else
icsize = icntix
endif
ALLOCATE( mp_snd_buffer( icsize * nproc ) )
ALLOCATE( mp_rcv_buffer( icsize * nproc ) )
ALLOCATE( my_buffer( ngw ) )
ALLOCATE( ibuf( nproc ) )
ctmp = ( 0.0_DP, 0.0_DP )
! WRITE( stdout,*) 'D: ', nproc, mpime, group
ibuf = 0
DO IG = 1, n_indi_snd
idest = dest_indi(ig)
ibuf(idest) = ibuf(idest) + 1;
if(idest .ne. me) then
mp_snd_buffer( ibuf(idest) + (idest-1)*icsize ) = C( indi_l( ig ) )
else
my_buffer(ibuf(idest)) = C(indi_l(ig))
end if
end do
#if defined __MPI
call MPI_ALLTOALL( mp_snd_buffer(1), icsize, MPI_DOUBLE_COMPLEX, &
mp_rcv_buffer(1), icsize, MPI_DOUBLE_COMPLEX, &
group, ierr)
#else
CALL errore(' pwscatter ',' no communication protocol ',0)
#endif
ibuf = 0
DO IG = 1, n_indi_rcv
isour = sour_indi(ig)
if(isour.gt.0 .and. isour.ne.me) then
ibuf(isour) = ibuf(isour) + 1
CTMP(ig) = mp_rcv_buffer(ibuf(isour) + (isour-1)*icsize)
else if(isour.gt.0) then
ibuf(isour) = ibuf(isour) + 1
CTMP(ig) = my_buffer(ibuf(isour))
else
CTMP(ig) = (0.0_DP,0.0_DP)
end if
end do
DEALLOCATE( mp_snd_buffer )
DEALLOCATE( mp_rcv_buffer )
DEALLOCATE( my_buffer )
DEALLOCATE( ibuf )
RETURN
END SUBROUTINE pwscatter
!=----------------------------------------------------------------------------=!
SUBROUTINE redistwf( c_dist_pw, c_dist_st, npw_p, nst_p, comm, idir )
!
! Redistribute wave function.
! c_dist_pw are the wave functions with plane waves distributed over processors
! c_dist_st are the wave functions with electronic states distributed over processors
!
USE kinds
USE parallel_include
implicit none
COMPLEX(DP) :: c_dist_pw(:,:)
COMPLEX(DP) :: c_dist_st(:,:)
INTEGER, INTENT(IN) :: npw_p(:) ! the number of plane wave on each processor
INTEGER, INTENT(IN) :: nst_p(:) ! the number of states on each processor
INTEGER, INTENT(IN) :: comm ! group communicator
INTEGER, INTENT(IN) :: idir ! direction of the redistribution
! idir > 0 c_dist_pw --> c_dist_st
! idir < 0 c_dist_pw <-- c_dist_st
INTEGER :: mpime, nproc, ierr, npw_t, nst_t, proc, i, j, ngpww, ii
INTEGER, ALLOCATABLE :: rdispls(:), recvcount(:)
INTEGER, ALLOCATABLE :: sendcount(:), sdispls(:)
COMPLEX(DP), ALLOCATABLE :: ctmp( : )
#if defined(__MPI)
CALL mpi_comm_rank( comm, mpime, ierr )
IF( ierr /= 0 ) CALL errore( ' wf_redist ', ' mpi_comm_rank ', ierr )
CALL mpi_comm_size( comm, nproc, ierr )
IF( ierr /= 0 ) CALL errore( ' wf_redist ', ' mpi_comm_size ', ierr )
ALLOCATE( rdispls( nproc ), recvcount( nproc ), sendcount( nproc ), sdispls( nproc ) )
npw_t = 0
nst_t = 0
DO proc=1,nproc
sendcount(proc) = npw_p(mpime+1) * nst_p(proc)
recvcount(proc) = npw_p(proc) * nst_p(mpime+1)
npw_t = npw_t + npw_p(proc)
nst_t = nst_t + nst_p(proc)
END DO
sdispls(1)=0
rdispls(1)=0
DO proc=2,nproc
sdispls(proc) = sdispls(proc-1) + sendcount(proc-1)
rdispls(proc) = rdispls(proc-1) + recvcount(proc-1)
END DO
ALLOCATE( ctmp( npw_t * nst_p( mpime + 1 ) ) )
IF( idir > 0 ) THEN
!
! ... Step 1. Communicate to all Procs so that each proc has all
! ... G-vectors and some states instead of all states and some
! ... G-vectors. This information is stored in the 1-d array ctmp.
!
CALL MPI_BARRIER( comm, ierr )
IF( ierr /= 0 ) CALL errore( ' wf_redist ', ' mpi_barrier ', ierr )
!
CALL MPI_ALLTOALLV( c_dist_pw, sendcount, sdispls, MPI_DOUBLE_COMPLEX, &
& ctmp, recvcount, rdispls, MPI_DOUBLE_COMPLEX, comm, ierr)
IF( ierr /= 0 ) CALL errore( ' wf_redist ', ' mpi_alltoallv ', ierr )
!
! Step 2. Convert the 1-d array ctmp into a 2-d array consistent with the
! original notation c(ngw,nbsp). Psitot contains ntot = SUM_Procs(ngw) G-vecs
! and nstat states instead of all nbsp states
!
ngpww = 0
DO proc = 1, nproc
DO i = 1, nst_p(mpime+1)
ii = (i-1) * npw_p(proc)
DO j = 1, npw_p(proc)
c_dist_st( j + ngpww, i ) = ctmp( rdispls(proc) + j + ii )
END DO
END DO
ngpww = ngpww + npw_p(proc)
END DO
ELSE
!
! Step 4. Convert the 2-d array c_dist_st into 1-d array
!
ngpww = 0
DO proc = 1, nproc
DO i = 1, nst_p(mpime+1)
ii = (i-1) * npw_p(proc)
DO j = 1, npw_p(proc)
ctmp( rdispls(proc) + j + ii ) = c_dist_st( j + ngpww, i )
END DO
END DO
ngpww = ngpww + npw_p(proc)
END DO
!
! Step 5. Redistribute among processors. The result is stored in 2-d
! array c_dist_pw consistent with the notation c(ngw,nbsp)
!
CALL MPI_BARRIER( comm, ierr )
IF( ierr /= 0 ) CALL errore( ' wf_redist ', ' mpi_barrier ', ierr )
CALL MPI_ALLTOALLV( ctmp, recvcount, rdispls, MPI_DOUBLE_COMPLEX, &
& c_dist_pw, sendcount , sdispls, MPI_DOUBLE_COMPLEX, comm, ierr )
IF( ierr /= 0 ) CALL errore( ' wf_redist ', ' mpi_alltoallv ', ierr )
END IF
DEALLOCATE( ctmp )
DEALLOCATE( rdispls, recvcount, sendcount, sdispls )
#endif
RETURN
END SUBROUTINE redistwf
!=----------------------------------------------------------------------------=!
SUBROUTINE redistwfr( c_dist_pw, c_dist_st, npw_p, nst_p, comm, idir )
!
! Redistribute wave function.
! c_dist_pw are the wave functions with plane waves distributed over processors
! c_dist_st are the wave functions with electronic states distributed over processors
!
USE kinds
USE parallel_include
implicit none
REAL(DP) :: c_dist_pw(:,:)
REAL(DP) :: c_dist_st(:,:)
INTEGER, INTENT(IN) :: npw_p(:) ! the number of plane wave on each processor
INTEGER, INTENT(IN) :: nst_p(:) ! the number of states on each processor
INTEGER, INTENT(IN) :: comm ! group communicator
INTEGER, INTENT(IN) :: idir ! direction of the redistribution
! idir > 0 c_dist_pw --> c_dist_st
! idir < 0 c_dist_pw <-- c_dist_st
INTEGER :: mpime, nproc, ierr, npw_t, nst_t, proc, i, j, ngpww
INTEGER, ALLOCATABLE :: rdispls(:), recvcount(:)
INTEGER, ALLOCATABLE :: sendcount(:), sdispls(:)
REAL(DP), ALLOCATABLE :: ctmp( : )
#if defined(__MPI)
CALL mpi_comm_rank( comm, mpime, ierr )
IF( ierr /= 0 ) CALL errore( ' wf_redist ', ' mpi_comm_rank ', ierr )
CALL mpi_comm_size( comm, nproc, ierr )
IF( ierr /= 0 ) CALL errore( ' wf_redist ', ' mpi_comm_size ', ierr )
ALLOCATE( rdispls( nproc ), recvcount( nproc ), sendcount( nproc ), sdispls( nproc ) )
npw_t = 0
nst_t = 0
DO proc=1,nproc
sendcount(proc) = npw_p(mpime+1) * nst_p(proc)
recvcount(proc) = npw_p(proc) * nst_p(mpime+1)
npw_t = npw_t + npw_p(proc)
nst_t = nst_t + nst_p(proc)
END DO
sdispls(1)=0
rdispls(1)=0
DO proc=2,nproc
sdispls(proc) = sdispls(proc-1) + sendcount(proc-1)
rdispls(proc) = rdispls(proc-1) + recvcount(proc-1)
END DO
ALLOCATE( ctmp( npw_t * nst_p( mpime + 1 ) ) )
IF( idir > 0 ) THEN
!
! ... Step 1. Communicate to all Procs so that each proc has all
! ... G-vectors and some states instead of all states and some
! ... G-vectors. This information is stored in the 1-d array ctmp.
!
CALL MPI_BARRIER( comm, ierr )
IF( ierr /= 0 ) CALL errore( ' wf_redist ', ' mpi_barrier ', ierr )
!
CALL MPI_ALLTOALLV( c_dist_pw, sendcount, sdispls, MPI_DOUBLE_PRECISION, &
& ctmp, recvcount, rdispls, MPI_DOUBLE_PRECISION, comm, ierr)
IF( ierr /= 0 ) CALL errore( ' wf_redist ', ' mpi_alltoallv ', ierr )
!
! Step 2. Convert the 1-d array ctmp into a 2-d array consistent with the
! original notation c(ngw,nbsp). Psitot contains ntot = SUM_Procs(ngw) G-vecs
! and nstat states instead of all nbsp states
!
ngpww = 0
DO proc = 1, nproc
DO i = 1, nst_p(mpime+1)
DO j = 1, npw_p(proc)
c_dist_st( j + ngpww, i ) = ctmp( rdispls(proc) + j + (i-1) * npw_p(proc) )
END DO
END DO
ngpww = ngpww + npw_p(proc)
END DO
ELSE
!
! Step 4. Convert the 2-d array c_dist_st into 1-d array
!
ngpww = 0
DO proc = 1, nproc
DO i = 1, nst_p(mpime+1)
DO j = 1, npw_p(proc)
ctmp( rdispls(proc) + j + (i-1) * npw_p(proc) ) = c_dist_st( j + ngpww, i )
END DO
END DO
ngpww = ngpww + npw_p(proc)
END DO
!
! Step 5. Redistribute among processors. The result is stored in 2-d
! array c_dist_pw consistent with the notation c(ngw,nbsp)
!
CALL MPI_BARRIER( comm, ierr )
IF( ierr /= 0 ) CALL errore( ' wf_redist ', ' mpi_barrier ', ierr )
CALL MPI_ALLTOALLV( ctmp, recvcount, rdispls, MPI_DOUBLE_PRECISION, &
& c_dist_pw, sendcount , sdispls, MPI_DOUBLE_PRECISION, comm, ierr )
IF( ierr /= 0 ) CALL errore( ' wf_redist ', ' mpi_alltoallv ', ierr )
END IF
DEALLOCATE( ctmp )
DEALLOCATE( rdispls, recvcount, sendcount, sdispls )
#endif
RETURN
END SUBROUTINE redistwfr
!=----------------------------------------------------------------------------=!
END MODULE mp_wave