forked from cp2k/cp2k
-
Notifications
You must be signed in to change notification settings - Fork 0
/
bse.F
646 lines (473 loc) · 25.4 KB
/
bse.F
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
!--------------------------------------------------------------------------------------------------!
! CP2K: A general program to perform molecular dynamics simulations !
! Copyright 2000-2024 CP2K developers group <https://cp2k.org> !
! !
! SPDX-License-Identifier: GPL-2.0-or-later !
!--------------------------------------------------------------------------------------------------!
! **************************************************************************************************
!> \brief Routines for GW + Bethe-Salpeter for computing electronic excitations
!> \par History
!> 04.2017 created [Jan Wilhelm]
! **************************************************************************************************
MODULE bse
USE cp_fm_basic_linalg, ONLY: cp_fm_upper_to_full
USE cp_fm_cholesky, ONLY: cp_fm_cholesky_decompose,&
cp_fm_cholesky_invert
USE cp_fm_types, ONLY: cp_fm_create,&
cp_fm_get_info,&
cp_fm_release,&
cp_fm_set_all,&
cp_fm_to_fm,&
cp_fm_type
USE group_dist_types, ONLY: get_group_dist,&
group_dist_d1_type
USE kinds, ONLY: dp
USE message_passing, ONLY: mp_para_env_type,&
mp_request_type
USE mp2_types, ONLY: integ_mat_buffer_type
USE parallel_gemm_api, ONLY: parallel_gemm
USE rpa_communication, ONLY: communicate_buffer
#include "./base/base_uses.f90"
IMPLICIT NONE
PRIVATE
CHARACTER(len=*), PARAMETER, PRIVATE :: moduleN = 'bse'
PUBLIC :: mult_B_with_W_and_fill_local_3c_arrays, do_subspace_iterations
CONTAINS
! **************************************************************************************************
!> \brief ...
!> \param B_bar_ijQ_bse_local ...
!> \param B_abQ_bse_local ...
!> \param B_bar_iaQ_bse_local ...
!> \param B_iaQ_bse_local ...
!> \param homo ...
!> \param virtual ...
!> \param num_Z_vectors ...
!> \param max_iter ...
!> \param threshold_min_trans ...
!> \param Eigenval ...
!> \param para_env ...
! **************************************************************************************************
SUBROUTINE do_subspace_iterations(B_bar_ijQ_bse_local, B_abQ_bse_local, B_bar_iaQ_bse_local, &
B_iaQ_bse_local, homo, virtual, num_Z_vectors, &
max_iter, threshold_min_trans, Eigenval, para_env)
REAL(KIND=dp), ALLOCATABLE, DIMENSION(:, :, :) :: B_bar_ijQ_bse_local, B_abQ_bse_local, &
B_bar_iaQ_bse_local, B_iaQ_bse_local
INTEGER :: homo, virtual, num_Z_vectors, max_iter
REAL(KIND=dp) :: threshold_min_trans
REAL(KIND=dp), DIMENSION(:) :: Eigenval
TYPE(mp_para_env_type), INTENT(IN) :: para_env
CHARACTER(LEN=*), PARAMETER :: routineN = 'do_subspace_iterations'
INTEGER :: handle, i_iter, local_RI_size
REAL(KIND=dp), ALLOCATABLE, DIMENSION(:, :) :: M_ia_tmp, M_ji_tmp, RI_vector
REAL(KIND=dp), ALLOCATABLE, DIMENSION(:, :, :) :: AZ, BZ, Z_vectors
CALL timeset(routineN, handle)
! JW hack 2del
threshold_min_trans = 0.01_dp
ALLOCATE (Z_vectors(homo, virtual, num_Z_vectors))
Z_vectors = 0.0_dp
ALLOCATE (AZ(homo, virtual, num_Z_vectors))
AZ = 0.0_dp
ALLOCATE (BZ(homo, virtual, num_Z_vectors))
BZ = 0.0_dp
local_RI_size = SIZE(B_iaQ_bse_local, 3)
ALLOCATE (M_ia_tmp(homo, virtual))
M_ia_tmp = 0.0_dp
ALLOCATE (M_ji_tmp(homo, homo))
M_ji_tmp = 0.0_dp
ALLOCATE (RI_vector(local_RI_size, num_Z_vectors))
RI_vector = 0.0_dp
CALL initial_guess_Z_vectors(Z_vectors, Eigenval, num_Z_vectors, homo, virtual)
DO i_iter = 1, max_iter
CALL compute_AZ(AZ, Z_vectors, B_iaQ_bse_local, B_bar_ijQ_bse_local, B_abQ_bse_local, &
M_ia_tmp, RI_vector, Eigenval, homo, virtual, num_Z_vectors, local_RI_size, &
para_env)
CALL compute_BZ(BZ, Z_vectors, B_iaQ_bse_local, B_bar_iaQ_bse_local, &
M_ji_tmp, RI_vector, homo, virtual, num_Z_vectors, local_RI_size, &
para_env)
END DO
DEALLOCATE (AZ, BZ, Z_vectors, M_ia_tmp, M_ji_tmp, RI_vector)
CALL timestop(handle)
END SUBROUTINE
! **************************************************************************************************
!> \brief ...
!> \param BZ ...
!> \param Z_vectors ...
!> \param B_iaQ_bse_local ...
!> \param B_bar_iaQ_bse_local ...
!> \param M_ji_tmp ...
!> \param RI_vector ...
!> \param homo ...
!> \param virtual ...
!> \param num_Z_vectors ...
!> \param local_RI_size ...
!> \param para_env ...
! **************************************************************************************************
SUBROUTINE compute_BZ(BZ, Z_vectors, B_iaQ_bse_local, B_bar_iaQ_bse_local, &
M_ji_tmp, RI_vector, homo, virtual, num_Z_vectors, local_RI_size, para_env)
REAL(KIND=dp), ALLOCATABLE, DIMENSION(:, :, :) :: BZ, Z_vectors, B_iaQ_bse_local, &
B_bar_iaQ_bse_local
REAL(KIND=dp), DIMENSION(:, :) :: M_ji_tmp, RI_vector
INTEGER :: homo, virtual, num_Z_vectors, &
local_RI_size
TYPE(mp_para_env_type), INTENT(IN) :: para_env
INTEGER :: i_Z_vector, LLL
BZ(:, :, :) = 0.0_dp
CALL compute_v_ia_jb_part(BZ, Z_vectors, B_iaQ_bse_local, RI_vector, local_RI_size, &
num_Z_vectors, homo, virtual)
DO i_Z_vector = 1, num_Z_vectors
DO LLL = 1, local_RI_size
! M_ji^P = sum_b Z_jb*B_bi^P
CALL DGEMM("N", "T", homo, homo, virtual, 1.0_dp, Z_vectors(:, :, i_Z_vector), homo, &
B_iaQ_bse_local(:, :, LLL), homo, 0.0_dp, M_ji_tmp, homo)
! (BZ)_ia = sum_jP M_ij^P*B^bar_ja^P
CALL DGEMM("T", "N", homo, virtual, homo, 1.0_dp, M_ji_tmp, homo, &
B_bar_iaQ_bse_local, homo, 1.0_dp, BZ(:, :, i_Z_vector), homo)
END DO
END DO
! we make the sum over all RI basis functions
CALL para_env%sum(BZ)
END SUBROUTINE
! **************************************************************************************************
!> \brief ...
!> \param AZ ...
!> \param Z_vectors ...
!> \param B_iaQ_bse_local ...
!> \param B_bar_ijQ_bse_local ...
!> \param B_abQ_bse_local ...
!> \param M_ia_tmp ...
!> \param RI_vector ...
!> \param Eigenval ...
!> \param homo ...
!> \param virtual ...
!> \param num_Z_vectors ...
!> \param local_RI_size ...
!> \param para_env ...
! **************************************************************************************************
SUBROUTINE compute_AZ(AZ, Z_vectors, B_iaQ_bse_local, B_bar_ijQ_bse_local, B_abQ_bse_local, M_ia_tmp, &
RI_vector, Eigenval, homo, virtual, num_Z_vectors, local_RI_size, para_env)
REAL(KIND=dp), ALLOCATABLE, DIMENSION(:, :, :) :: AZ, Z_vectors, B_iaQ_bse_local, &
B_bar_ijQ_bse_local, B_abQ_bse_local
REAL(KIND=dp), DIMENSION(:, :) :: M_ia_tmp, RI_vector
REAL(KIND=dp), DIMENSION(:) :: Eigenval
INTEGER :: homo, virtual, num_Z_vectors, &
local_RI_size
TYPE(mp_para_env_type), INTENT(IN) :: para_env
CHARACTER(LEN=*), PARAMETER :: routineN = 'compute_AZ'
INTEGER :: a_virt, handle, i_occ, i_Z_vector, LLL
REAL(KIND=dp) :: eigen_diff
CALL timeset(routineN, handle)
AZ(:, :, :) = 0.0_dp
CALL compute_v_ia_jb_part(AZ, Z_vectors, B_iaQ_bse_local, RI_vector, local_RI_size, &
num_Z_vectors, homo, virtual)
DO i_Z_vector = 1, num_Z_vectors
! JW TO DO: OMP PARALLELIZATION
DO LLL = 1, local_RI_size
! M_ja^P = sum_j Z_jb*B_ba^P
CALL DGEMM("N", "N", homo, virtual, virtual, 1.0_dp, Z_vectors(:, :, i_Z_vector), homo, &
B_abQ_bse_local(:, :, LLL), virtual, 0.0_dp, M_ia_tmp, homo)
! (AZ)_ia = sum_jP B_bar_ij^P*M_ja^P
CALL DGEMM("N", "N", homo, virtual, homo, 1.0_dp, B_bar_ijQ_bse_local(:, :, LLL), homo, &
M_ia_tmp, homo, 1.0_dp, AZ(:, :, i_Z_vector), homo)
END DO
END DO
! we make the sum over all RI basis functions
CALL para_env%sum(AZ)
! add (e_a-e_i)*Z_ia
DO i_occ = 1, homo
DO a_virt = 1, virtual
eigen_diff = Eigenval(a_virt + homo) - Eigenval(i_occ)
AZ(i_occ, a_virt, :) = AZ(i_occ, a_virt, :) + Z_vectors(i_occ, a_virt, :)*eigen_diff
END DO
END DO
CALL timestop(handle)
END SUBROUTINE
! **************************************************************************************************
!> \brief ...
!> \param AZ ...
!> \param Z_vectors ...
!> \param B_iaQ_bse_local ...
!> \param RI_vector ...
!> \param local_RI_size ...
!> \param num_Z_vectors ...
!> \param homo ...
!> \param virtual ...
! **************************************************************************************************
SUBROUTINE compute_v_ia_jb_part(AZ, Z_vectors, B_iaQ_bse_local, RI_vector, local_RI_size, &
num_Z_vectors, homo, virtual)
REAL(KIND=dp), DIMENSION(:, :, :), INTENT(INOUT) :: AZ, Z_vectors, B_iaQ_bse_local
REAL(KIND=dp), DIMENSION(:, :), INTENT(INOUT) :: RI_vector
INTEGER, INTENT(IN) :: local_RI_size, num_Z_vectors, homo, &
virtual
CHARACTER(LEN=*), PARAMETER :: routineN = 'compute_v_ia_jb_part'
INTEGER :: a_virt, handle, i_occ, i_Z_vector, LLL
CALL timeset(routineN, handle)
RI_vector = 0.0_dp
! v_P = sum_jb B_jb^P Z_jb
DO LLL = 1, local_RI_size
DO i_Z_vector = 1, num_Z_vectors
DO i_occ = 1, homo
DO a_virt = 1, virtual
RI_vector(LLL, i_Z_vector) = RI_vector(LLL, i_Z_vector) + &
Z_vectors(i_occ, a_virt, i_Z_vector)* &
B_iaQ_bse_local(i_occ, a_virt, LLL)
END DO
END DO
END DO
END DO
! AZ = sum_P B_ia^P*v_P + ...
DO LLL = 1, local_RI_size
DO i_Z_vector = 1, num_Z_vectors
DO i_occ = 1, homo
DO a_virt = 1, virtual
AZ(i_occ, a_virt, i_Z_vector) = AZ(i_occ, a_virt, i_Z_vector) + &
RI_vector(LLL, i_Z_vector)* &
B_iaQ_bse_local(i_occ, a_virt, LLL)
END DO
END DO
END DO
END DO
CALL timestop(handle)
END SUBROUTINE
! **************************************************************************************************
!> \brief ...
!> \param Z_vectors ...
!> \param Eigenval ...
!> \param num_Z_vectors ...
!> \param homo ...
!> \param virtual ...
! **************************************************************************************************
SUBROUTINE initial_guess_Z_vectors(Z_vectors, Eigenval, num_Z_vectors, homo, virtual)
REAL(KIND=dp), DIMENSION(:, :, :), INTENT(INOUT) :: Z_vectors
REAL(KIND=dp), DIMENSION(:), INTENT(IN) :: Eigenval
INTEGER, INTENT(IN) :: num_Z_vectors, homo, virtual
CHARACTER(LEN=*), PARAMETER :: routineN = 'initial_guess_Z_vectors'
INTEGER :: a_virt, handle, i_occ, i_Z_vector, &
min_loc(2)
REAL(KIND=dp), ALLOCATABLE, DIMENSION(:, :) :: eigen_diff_ia
CALL timeset(routineN, handle)
ALLOCATE (eigen_diff_ia(homo, virtual))
DO i_occ = 1, homo
DO a_virt = 1, virtual
eigen_diff_ia(i_occ, a_virt) = Eigenval(a_virt + homo) - Eigenval(i_occ)
END DO
END DO
DO i_Z_vector = 1, num_Z_vectors
min_loc = MINLOC(eigen_diff_ia)
Z_vectors(min_loc(1), min_loc(2), i_Z_vector) = 1.0_dp
eigen_diff_ia(min_loc(1), min_loc(2)) = 1.0E20_dp
END DO
DEALLOCATE (eigen_diff_ia)
CALL timestop(handle)
END SUBROUTINE
! **************************************************************************************************
!> \brief ...
!> \param fm_mat_S_ij_bse ...
!> \param fm_mat_S_ab_bse ...
!> \param fm_mat_S ...
!> \param fm_mat_Q_static_bse ...
!> \param fm_mat_Q_static_bse_gemm ...
!> \param B_bar_ijQ_bse_local ...
!> \param B_abQ_bse_local ...
!> \param B_bar_iaQ_bse_local ...
!> \param B_iaQ_bse_local ...
!> \param dimen_RI ...
!> \param homo ...
!> \param virtual ...
!> \param dimen_ia ...
!> \param gd_array ...
!> \param color_sub ...
!> \param para_env ...
! **************************************************************************************************
SUBROUTINE mult_B_with_W_and_fill_local_3c_arrays(fm_mat_S_ij_bse, fm_mat_S_ab_bse, fm_mat_S, fm_mat_Q_static_bse, &
fm_mat_Q_static_bse_gemm, &
B_bar_ijQ_bse_local, B_abQ_bse_local, B_bar_iaQ_bse_local, &
B_iaQ_bse_local, dimen_RI, homo, virtual, dimen_ia, &
gd_array, color_sub, para_env)
TYPE(cp_fm_type), INTENT(IN) :: fm_mat_S_ij_bse, fm_mat_S_ab_bse, &
fm_mat_S, fm_mat_Q_static_bse, &
fm_mat_Q_static_bse_gemm
REAL(KIND=dp), ALLOCATABLE, DIMENSION(:, :, :), &
INTENT(OUT) :: B_bar_ijQ_bse_local, B_abQ_bse_local, &
B_bar_iaQ_bse_local, B_iaQ_bse_local
INTEGER, INTENT(IN) :: dimen_RI, homo, virtual, dimen_ia
TYPE(group_dist_d1_type), INTENT(IN) :: gd_array
INTEGER, INTENT(IN) :: color_sub
TYPE(mp_para_env_type), INTENT(IN) :: para_env
CHARACTER(LEN=*), PARAMETER :: routineN = 'mult_B_with_W_and_fill_local_3c_arrays'
INTEGER :: handle, i_global, iiB, info_chol, &
j_global, jjB, ncol_local, nrow_local
INTEGER, DIMENSION(:), POINTER :: col_indices, row_indices
TYPE(cp_fm_type) :: fm_mat_S_bar_ia_bse, &
fm_mat_S_bar_ij_bse, fm_mat_work
CALL timeset(routineN, handle)
CALL cp_fm_create(fm_mat_S_bar_ia_bse, fm_mat_S%matrix_struct)
CALL cp_fm_to_fm(fm_mat_S, fm_mat_S_bar_ia_bse)
CALL cp_fm_set_all(fm_mat_S_bar_ia_bse, 0.0_dp)
CALL cp_fm_create(fm_mat_S_bar_ij_bse, fm_mat_S_ij_bse%matrix_struct)
CALL cp_fm_to_fm(fm_mat_S_ij_bse, fm_mat_S_bar_ij_bse)
CALL cp_fm_set_all(fm_mat_S_bar_ij_bse, 0.0_dp)
CALL cp_fm_create(fm_mat_work, fm_mat_Q_static_bse_gemm%matrix_struct)
CALL cp_fm_to_fm(fm_mat_Q_static_bse_gemm, fm_mat_work)
CALL cp_fm_set_all(fm_mat_work, 0.0_dp)
! get info of fm_mat_Q_static_bse and compute ((1+Q(0))^-1-1)
CALL cp_fm_get_info(matrix=fm_mat_Q_static_bse_gemm, &
nrow_local=nrow_local, &
ncol_local=ncol_local, &
row_indices=row_indices, &
col_indices=col_indices)
DO jjB = 1, ncol_local
j_global = col_indices(jjB)
DO iiB = 1, nrow_local
i_global = row_indices(iiB)
IF (j_global == i_global .AND. i_global <= dimen_RI) THEN
fm_mat_Q_static_bse_gemm%local_data(iiB, jjB) = fm_mat_Q_static_bse_gemm%local_data(iiB, jjB) + 1.0_dp
END IF
END DO
END DO
! calculate Trace(Log(Matrix)) as Log(DET(Matrix)) via cholesky decomposition
CALL cp_fm_cholesky_decompose(matrix=fm_mat_Q_static_bse_gemm, n=dimen_RI, info_out=info_chol)
CPASSERT(info_chol == 0)
! calculate [1+Q(i0)]^-1
CALL cp_fm_cholesky_invert(fm_mat_Q_static_bse_gemm)
! symmetrize the result
CALL cp_fm_upper_to_full(fm_mat_Q_static_bse_gemm, fm_mat_work)
CALL parallel_gemm(transa="N", transb="N", m=dimen_RI, n=homo**2, k=dimen_RI, alpha=1.0_dp, &
matrix_a=fm_mat_Q_static_bse, matrix_b=fm_mat_S_ij_bse, beta=0.0_dp, &
matrix_c=fm_mat_S_bar_ij_bse)
! fm_mat_S_bar_ia_bse has a different blacs_env as fm_mat_S_ij_bse since we take
! fm_mat_S from RPA. Therefore, we also need a different fm_mat_Q_static_bse_gemm
CALL parallel_gemm(transa="N", transb="N", m=dimen_RI, n=dimen_ia, k=dimen_RI, alpha=1.0_dp, &
matrix_a=fm_mat_Q_static_bse_gemm, matrix_b=fm_mat_S, beta=0.0_dp, &
matrix_c=fm_mat_S_bar_ia_bse)
CALL allocate_and_fill_local_array(B_iaQ_bse_local, fm_mat_S, gd_array, color_sub, homo, virtual, dimen_RI, para_env)
CALL allocate_and_fill_local_array(B_bar_iaQ_bse_local, fm_mat_S_bar_ia_bse, gd_array, color_sub, homo, virtual, &
dimen_RI, para_env)
CALL allocate_and_fill_local_array(B_bar_ijQ_bse_local, fm_mat_S_bar_ij_bse, gd_array, color_sub, homo, homo, &
dimen_RI, para_env)
CALL allocate_and_fill_local_array(B_abQ_bse_local, fm_mat_S_ab_bse, gd_array, color_sub, virtual, virtual, &
dimen_RI, para_env)
CALL cp_fm_release(fm_mat_S_bar_ia_bse)
CALL cp_fm_release(fm_mat_S_bar_ij_bse)
CALL cp_fm_release(fm_mat_work)
CALL timestop(handle)
END SUBROUTINE
! **************************************************************************************************
!> \brief ...
!> \param B_local ...
!> \param fm_mat_S ...
!> \param gd_array ...
!> \param color_sub ...
!> \param small_size ...
!> \param big_size ...
!> \param dimen_RI ...
!> \param para_env ...
! **************************************************************************************************
SUBROUTINE allocate_and_fill_local_array(B_local, fm_mat_S, gd_array, &
color_sub, small_size, big_size, dimen_RI, para_env)
REAL(KIND=dp), ALLOCATABLE, DIMENSION(:, :, :), &
INTENT(OUT) :: B_local
TYPE(cp_fm_type), INTENT(IN) :: fm_mat_S
TYPE(group_dist_d1_type), INTENT(IN) :: gd_array
INTEGER, INTENT(IN) :: color_sub, small_size, big_size, dimen_RI
TYPE(mp_para_env_type), INTENT(IN) :: para_env
CHARACTER(LEN=*), PARAMETER :: routineN = 'allocate_and_fill_local_array'
INTEGER :: combi_index, end_RI, handle, handle1, i_comm, i_entry, iiB, imepos, jjB, &
level_big_size, level_small_size, ncol_local, nrow_local, num_comm_cycles, RI_index, &
size_RI, start_RI
INTEGER, ALLOCATABLE, DIMENSION(:) :: entry_counter, mepos_from_RI_index, &
num_entries_rec, num_entries_send
INTEGER, DIMENSION(:), POINTER :: col_indices, row_indices
REAL(KIND=dp) :: matrix_el
TYPE(integ_mat_buffer_type), ALLOCATABLE, &
DIMENSION(:) :: buffer_rec, buffer_send
TYPE(mp_request_type), DIMENSION(:, :), POINTER :: req_array
CALL timeset(routineN, handle)
ALLOCATE (mepos_from_RI_index(dimen_RI))
mepos_from_RI_index = 0
DO imepos = 0, para_env%num_pe - 1
CALL get_group_dist(gd_array, pos=imepos, starts=start_RI, ends=end_RI)
mepos_from_RI_index(start_RI:end_RI) = imepos
END DO
! color_sub is automatically the number of the process since every subgroup has only one MPI rank
CALL get_group_dist(gd_array, color_sub, start_RI, end_RI, size_RI)
ALLOCATE (B_local(small_size, big_size, 1:size_RI))
ALLOCATE (num_entries_send(0:para_env%num_pe - 1))
ALLOCATE (num_entries_rec(0:para_env%num_pe - 1))
ALLOCATE (req_array(1:para_env%num_pe, 4))
ALLOCATE (entry_counter(0:para_env%num_pe - 1))
CALL cp_fm_get_info(matrix=fm_mat_S, &
nrow_local=nrow_local, &
ncol_local=ncol_local, &
row_indices=row_indices, &
col_indices=col_indices)
num_comm_cycles = 10
! communicate not all due to huge memory overhead, since for every number in fm_mat_S, we store
! three additional ones (RI index, first MO index, second MO index!!)
DO i_comm = 0, num_comm_cycles - 1
num_entries_send = 0
num_entries_rec = 0
! loop over RI index to get the number of sent entries
DO jjB = 1, nrow_local
RI_index = row_indices(jjB)
IF (MODULO(RI_index, num_comm_cycles) /= i_comm) CYCLE
imepos = mepos_from_RI_index(RI_index)
num_entries_send(imepos) = num_entries_send(imepos) + ncol_local
END DO
CALL para_env%alltoall(num_entries_send, num_entries_rec, 1)
ALLOCATE (buffer_rec(0:para_env%num_pe - 1))
ALLOCATE (buffer_send(0:para_env%num_pe - 1))
! allocate data message and corresponding indices
DO imepos = 0, para_env%num_pe - 1
ALLOCATE (buffer_rec(imepos)%msg(num_entries_rec(imepos)))
buffer_rec(imepos)%msg = 0.0_dp
ALLOCATE (buffer_send(imepos)%msg(num_entries_send(imepos)))
buffer_send(imepos)%msg = 0.0_dp
ALLOCATE (buffer_rec(imepos)%indx(num_entries_rec(imepos), 3))
buffer_rec(imepos)%indx = 0
ALLOCATE (buffer_send(imepos)%indx(num_entries_send(imepos), 3))
buffer_send(imepos)%indx = 0
END DO
entry_counter(:) = 0
! loop over RI index for filling the send-buffer
DO jjB = 1, nrow_local
RI_index = row_indices(jjB)
IF (MODULO(RI_index, num_comm_cycles) /= i_comm) CYCLE
imepos = mepos_from_RI_index(RI_index)
DO iiB = 1, ncol_local
combi_index = col_indices(iiB)
level_small_size = MAX(1, combi_index - 1)/big_size + 1
level_big_size = combi_index - (level_small_size - 1)*big_size
entry_counter(imepos) = entry_counter(imepos) + 1
buffer_send(imepos)%msg(entry_counter(imepos)) = fm_mat_S%local_data(jjB, iiB)
buffer_send(imepos)%indx(entry_counter(imepos), 1) = RI_index
buffer_send(imepos)%indx(entry_counter(imepos), 2) = level_small_size
buffer_send(imepos)%indx(entry_counter(imepos), 3) = level_big_size
END DO
END DO
CALL timeset("BSE_comm_data", handle1)
CALL communicate_buffer(para_env, num_entries_rec, num_entries_send, buffer_rec, buffer_send, req_array)
CALL timestop(handle1)
! fill B_local
DO imepos = 0, para_env%num_pe - 1
DO i_entry = 1, num_entries_rec(imepos)
RI_index = buffer_rec(imepos)%indx(i_entry, 1) - start_RI + 1
level_small_size = buffer_rec(imepos)%indx(i_entry, 2)
level_big_size = buffer_rec(imepos)%indx(i_entry, 3)
matrix_el = buffer_rec(imepos)%msg(i_entry)
B_local(level_small_size, level_big_size, RI_index) = matrix_el
END DO
END DO
DO imepos = 0, para_env%num_pe - 1
DEALLOCATE (buffer_send(imepos)%msg)
DEALLOCATE (buffer_send(imepos)%indx)
DEALLOCATE (buffer_rec(imepos)%msg)
DEALLOCATE (buffer_rec(imepos)%indx)
END DO
DEALLOCATE (buffer_rec, buffer_send)
END DO
DEALLOCATE (num_entries_send, num_entries_rec)
DEALLOCATE (mepos_from_RI_index)
DEALLOCATE (entry_counter, req_array)
CALL timestop(handle)
END SUBROUTINE
END MODULE bse