elpa2.F90 209 KB
Newer Older
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
!    This file is part of ELPA.
!
!    The ELPA library was originally created by the ELPA consortium,
!    consisting of the following organizations:
!
!    - Rechenzentrum Garching der Max-Planck-Gesellschaft (RZG),
!    - Bergische Universität Wuppertal, Lehrstuhl für angewandte
!      Informatik,
!    - Technische Universität München, Lehrstuhl für Informatik mit
!      Schwerpunkt Wissenschaftliches Rechnen ,
!    - Fritz-Haber-Institut, Berlin, Abt. Theorie,
!    - Max-Plack-Institut für Mathematik in den Naturwissenschaftrn,
!      Leipzig, Abt. Komplexe Strukutren in Biologie und Kognition,
!      and
!    - IBM Deutschland GmbH
!
!
!    More information can be found here:
!    http://elpa.rzg.mpg.de/
!
!    ELPA is free software: you can redistribute it and/or modify
!    it under the terms of the version 3 of the license of the
!    GNU Lesser General Public License as published by the Free
!    Software Foundation.
!
!    ELPA is distributed in the hope that it will be useful,
!    but WITHOUT ANY WARRANTY; without even the implied warranty of
!    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
!    GNU Lesser General Public License for more details.
!
!    You should have received a copy of the GNU Lesser General Public License
!    along with ELPA.  If not, see <http://www.gnu.org/licenses/>
!
!    ELPA reflects a substantial effort on the part of the original
!    ELPA consortium, and we ask you to respect the spirit of the
!    license that we chose: i.e., please contribute any changes you
!    may have back to the original ELPA library distribution, and keep
!    any derivatives of ELPA under the same license that we chose for
!    the original distribution, the GNU Lesser General Public License.
!
!
! ELPA1 -- Faster replacements for ScaLAPACK symmetric eigenvalue routines
!
! Copyright of the original code rests with the authors inside the ELPA
! consortium. The copyright of any additional modifications shall rest
! with their original authors, but shall adhere to the licensing terms
! distributed along with the original code in the file "COPYING".



! ELPA2 -- 2-stage solver for ELPA
!
! Copyright of the original code rests with the authors inside the ELPA
! consortium. The copyright of any additional modifications shall rest
! with their original authors, but shall adhere to the licensing terms
! distributed along with the original code in the file "COPYING".


#include "config-f90.h"

module ELPA2

! Version 1.1.2, 2011-02-21

65
  use elpa_utilities
66
  USE ELPA1
67
68
  use elpa2_utilities

69

70
71
72
#ifdef HAVE_ISO_FORTRAN_ENV
  use iso_fortran_env, only : error_unit
#endif
73
74
75

  use elpa_pdgeqrf

76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
  implicit none

  PRIVATE ! By default, all routines contained are private

  ! The following routines are public:

  public :: solve_evp_real_2stage
  public :: solve_evp_complex_2stage

  public :: bandred_real
  public :: tridiag_band_real
  public :: trans_ev_tridi_to_band_real
  public :: trans_ev_band_to_full_real

  public :: bandred_complex
  public :: tridiag_band_complex
  public :: trans_ev_tridi_to_band_complex
  public :: trans_ev_band_to_full_complex
94
95
96
97
#ifndef HAVE_ISO_FORTRAN_ENV
  integer, parameter :: error_unit = 6
#endif

98
99
100
101
102
103
  public :: band_band_real
  public :: divide_band

  integer, public :: which_qr_decomposition = 1     ! defines, which QR-decomposition algorithm will be used
                                                    ! 0 for unblocked
                                                    ! 1 for blocked (maxrank: nblk)
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
!-------------------------------------------------------------------------------

  ! The following array contains the Householder vectors of the
  ! transformation band -> tridiagonal.
  ! It is allocated and set in tridiag_band_real and used in
  ! trans_ev_tridi_to_band_real.
  ! It must be deallocated by the user after trans_ev_tridi_to_band_real!

  real*8, allocatable :: hh_trans_real(:,:)
  complex*16, allocatable :: hh_trans_complex(:,:)

!-------------------------------------------------------------------------------

  include 'mpif.h'


!******
contains
122

123
124
125
126
function solve_evp_real_2stage(na, nev, a, lda, ev, q, ldq, nblk,        &
                                 mpi_comm_rows, mpi_comm_cols,           &
                                 mpi_comm_all, THIS_REAL_ELPA_KERNEL_API,&
                                 useQR) result(success)
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

!-------------------------------------------------------------------------------
!  solve_evp_real_2stage: Solves the real eigenvalue problem with a 2 stage approach
!
!  Parameters
!
!  na          Order of matrix a
!
!  nev         Number of eigenvalues needed
!
!  a(lda,*)    Distributed matrix for which eigenvalues are to be computed.
!              Distribution is like in Scalapack.
!              The full matrix must be set (not only one half like in scalapack).
!              Destroyed on exit (upper and lower half).
!
!  lda         Leading dimension of a
!
!  ev(na)      On output: eigenvalues of a, every processor gets the complete set
!
!  q(ldq,*)    On output: Eigenvectors of a
!              Distribution is like in Scalapack.
!              Must be always dimensioned to the full size (corresponding to (na,na))
!              even if only a part of the eigenvalues is needed.
!
!  ldq         Leading dimension of q
!
!  nblk        blocksize of cyclic distribution, must be the same in both directions!
!
!  mpi_comm_rows
!  mpi_comm_cols
!              MPI-Communicators for rows/columns
!  mpi_comm_all
!              MPI-Communicator for the total processor set
!
!-------------------------------------------------------------------------------
162
163
164
#ifdef HAVE_DETAILED_TIMINGS
 use timings
#endif
165
   implicit none
166
167
   logical, intent(in), optional :: useQR
   logical                       :: useQRActual, useQREnvironment
Andreas Marek's avatar
Andreas Marek committed
168
   integer, intent(in), optional :: THIS_REAL_ELPA_KERNEL_API
169
   integer                       :: THIS_REAL_ELPA_KERNEL
170

171
   integer, intent(in)           :: na, nev, lda, ldq, mpi_comm_rows, &
172
                                    mpi_comm_cols, mpi_comm_all
173
   integer, intent(inout)        :: nblk
174
   real*8, intent(inout)         :: a(lda,*), ev(na), q(ldq,*)
175

176
177
178
179
180
181
   integer                       :: my_pe, n_pes, my_prow, my_pcol, np_rows, np_cols, mpierr
   integer                       :: nbw, num_blocks
   real*8, allocatable           :: tmat(:,:,:), e(:)
   real*8                        :: ttt0, ttt1, ttts
   integer                       :: i
   logical                       :: success
182
183
   logical, save                 :: firstCall = .true.
   logical                       :: wantDebug
184

185
186
187
#ifdef HAVE_DETAILED_TIMINGS
   call timer%start("solve_evp_real_2stage")
#endif
188
189
190
191
192
193
194
   call mpi_comm_rank(mpi_comm_all,my_pe,mpierr)
   call mpi_comm_size(mpi_comm_all,n_pes,mpierr)

   call mpi_comm_rank(mpi_comm_rows,my_prow,mpierr)
   call mpi_comm_size(mpi_comm_rows,np_rows,mpierr)
   call mpi_comm_rank(mpi_comm_cols,my_pcol,mpierr)
   call mpi_comm_size(mpi_comm_cols,np_cols,mpierr)
195

196
197
198
199
200
201
202
203

   wantDebug = .false.
   if (firstCall) then
     ! are debug messages desired?
     wantDebug = debug_messages_via_environment_variable()
     firstCall = .false.
   endif

204
205
   success = .true.

206
207
208
209
210
211
212
213
214
215
216
217
218
   useQRActual = .false.

   ! set usage of qr decomposition via API call
   if (present(useQR)) then
     if (useQR) useQRActual = .true.
     if (.not.(useQR)) useQRACtual = .false.
   endif

   ! overwrite this with environment variable settings
   if (qr_decomposition_via_environment_variable(useQREnvironment)) then
     useQRActual = useQREnvironment
   endif

219
   if (useQRActual) then
220
221
222
223
224
225
     if (mod(na,nblk) .ne. 0) then
       if (wantDebug) then
         write(error_unit,*) "solve_evp_real_2stage: QR-decomposition: blocksize does not fit with matrixsize"
       endif
     stop "Do not use QR-decomposition for this matrix and blocksize."
     endif
226
227
   endif

228

229
230
231
   if (present(THIS_REAL_ELPA_KERNEL_API)) then
     ! user defined kernel via the optional argument in the API call
     THIS_REAL_ELPA_KERNEL = THIS_REAL_ELPA_KERNEL_API
Andreas Marek's avatar
Andreas Marek committed
232
   else
233

234
235
236
     ! if kernel is not choosen via api
     ! check whether set by environment variable
     THIS_REAL_ELPA_KERNEL = get_actual_real_kernel()
Andreas Marek's avatar
Andreas Marek committed
237
238
239
240
   endif

   ! check whether choosen kernel is allowed
   if (check_allowed_real_kernels(THIS_REAL_ELPA_KERNEL)) then
241

242
243
244
245
246
247
248
249
250
251
252
     if (my_pe == 0) then
       write(error_unit,*) " "
       write(error_unit,*) "The choosen kernel ",REAL_ELPA_KERNEL_NAMES(THIS_REAL_ELPA_KERNEL)
       write(error_unit,*) "is not in the list of the allowed kernels!"
       write(error_unit,*) " "
       write(error_unit,*) "Allowed kernels are:"
       do i=1,size(REAL_ELPA_KERNEL_NAMES(:))
         if (AVAILABLE_REAL_ELPA_KERNELS(i) .ne. 0) then
           write(error_unit,*) REAL_ELPA_KERNEL_NAMES(i)
         endif
       enddo
Andreas Marek's avatar
Andreas Marek committed
253

254
255
256
257
       write(error_unit,*) " "
       write(error_unit,*) "The defaul kernel REAL_ELPA_KERNEL_GENERIC will be used !"
     endif
     THIS_REAL_ELPA_KERNEL = REAL_ELPA_KERNEL_GENERIC
Andreas Marek's avatar
Andreas Marek committed
258
259

   endif
260
261
262
263
264
265
266
267
268
269
270
271
272

   ! Choose bandwidth, must be a multiple of nblk, set to a value >= 32

   nbw = (31/nblk+1)*nblk

   num_blocks = (na-1)/nbw + 1

   allocate(tmat(nbw,nbw,num_blocks))

   ! Reduction full -> band

   ttt0 = MPI_Wtime()
   ttts = ttt0
273
   call bandred_real(na, a, lda, nblk, nbw, mpi_comm_rows, mpi_comm_cols, &
274
                     tmat, wantDebug, success, useQRActual)
275
   if (.not.(success)) return
276
   ttt1 = MPI_Wtime()
277
   if (my_prow==0 .and. my_pcol==0 .and. elpa_print_times) &
278
      write(error_unit,*) 'Time bandred_real               :',ttt1-ttt0
279
280
281
282
283
284

   ! Reduction band -> tridiagonal

   allocate(e(na))

   ttt0 = MPI_Wtime()
Andreas Marek's avatar
Andreas Marek committed
285
286
   call tridiag_band_real(na, nbw, nblk, a, lda, ev, e, mpi_comm_rows, &
                          mpi_comm_cols, mpi_comm_all)
287
   ttt1 = MPI_Wtime()
288
   if (my_prow==0 .and. my_pcol==0 .and. elpa_print_times) &
289
      write(error_unit,*) 'Time tridiag_band_real          :',ttt1-ttt0
290
291
292
293
294
295
296
297
298
299

   call mpi_bcast(ev,na,MPI_REAL8,0,mpi_comm_all,mpierr)
   call mpi_bcast(e,na,MPI_REAL8,0,mpi_comm_all,mpierr)

   ttt1 = MPI_Wtime()
   time_evp_fwd = ttt1-ttts

   ! Solve tridiagonal system

   ttt0 = MPI_Wtime()
300
   call solve_tridi(na, nev, ev, e, q, ldq, nblk, mpi_comm_rows,  &
301
                    mpi_comm_cols, wantDebug, success)
302
303
   if (.not.(success)) return

304
   ttt1 = MPI_Wtime()
305
306
   if (my_prow==0 .and. my_pcol==0 .and. elpa_print_times) &
     write(error_unit,*) 'Time solve_tridi                :',ttt1-ttt0
307
308
309
310
311
312
313
314
   time_evp_solve = ttt1-ttt0
   ttts = ttt1

   deallocate(e)

   ! Backtransform stage 1

   ttt0 = MPI_Wtime()
315
   call trans_ev_tridi_to_band_real(na, nev, nblk, nbw, q, ldq, mpi_comm_rows, &
316
                                    mpi_comm_cols, wantDebug, success, THIS_REAL_ELPA_KERNEL)
317
   if (.not.(success)) return
318
   ttt1 = MPI_Wtime()
319
   if (my_prow==0 .and. my_pcol==0 .and. elpa_print_times) &
320
      write(error_unit,*) 'Time trans_ev_tridi_to_band_real:',ttt1-ttt0
321
322
323
324
325
326
327

   ! We can now deallocate the stored householder vectors
   deallocate(hh_trans_real)

   ! Backtransform stage 2

   ttt0 = MPI_Wtime()
328
329
   call trans_ev_band_to_full_real(na, nev, nblk, nbw, a, lda, tmat, q, ldq, mpi_comm_rows, &
                                   mpi_comm_cols, useQRActual)
330
   ttt1 = MPI_Wtime()
331
   if (my_prow==0 .and. my_pcol==0 .and. elpa_print_times) &
332
      write(error_unit,*) 'Time trans_ev_band_to_full_real :',ttt1-ttt0
333
334
335
   time_evp_back = ttt1-ttts

   deallocate(tmat)
336
337
338
#ifdef HAVE_DETAILED_TIMINGS
   call timer%stop("solve_evp_real_2stage")
#endif
339
340
1  format(a,f10.3)

341
end function solve_evp_real_2stage
342
343
344
345
346

!-------------------------------------------------------------------------------

!-------------------------------------------------------------------------------

347
function solve_evp_complex_2stage(na, nev, a, lda, ev, q, ldq, nblk, &
Andreas Marek's avatar
Andreas Marek committed
348
                                    mpi_comm_rows, mpi_comm_cols,      &
349
                                    mpi_comm_all, THIS_COMPLEX_ELPA_KERNEL_API) result(success)
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

!-------------------------------------------------------------------------------
!  solve_evp_complex_2stage: Solves the complex eigenvalue problem with a 2 stage approach
!
!  Parameters
!
!  na          Order of matrix a
!
!  nev         Number of eigenvalues needed
!
!  a(lda,*)    Distributed matrix for which eigenvalues are to be computed.
!              Distribution is like in Scalapack.
!              The full matrix must be set (not only one half like in scalapack).
!              Destroyed on exit (upper and lower half).
!
!  lda         Leading dimension of a
!
!  ev(na)      On output: eigenvalues of a, every processor gets the complete set
!
!  q(ldq,*)    On output: Eigenvectors of a
!              Distribution is like in Scalapack.
!              Must be always dimensioned to the full size (corresponding to (na,na))
!              even if only a part of the eigenvalues is needed.
!
!  ldq         Leading dimension of q
!
!  nblk        blocksize of cyclic distribution, must be the same in both directions!
!
!  mpi_comm_rows
!  mpi_comm_cols
!              MPI-Communicators for rows/columns
!  mpi_comm_all
!              MPI-Communicator for the total processor set
!
!-------------------------------------------------------------------------------
385
386
387
#ifdef HAVE_DETAILED_TIMINGS
 use timings
#endif
388
   implicit none
Andreas Marek's avatar
Andreas Marek committed
389
390
   integer, intent(in), optional :: THIS_COMPLEX_ELPA_KERNEL_API
   integer                       :: THIS_COMPLEX_ELPA_KERNEL
391
392
393
394
395
396
397
398
399
400
   integer, intent(in)           :: na, nev, lda, ldq, nblk, mpi_comm_rows, mpi_comm_cols, mpi_comm_all
   complex*16, intent(inout)     :: a(lda,*), q(ldq,*)
   real*8, intent(inout)         :: ev(na)

   integer                       :: my_prow, my_pcol, np_rows, np_cols, mpierr, my_pe, n_pes
   integer                       :: l_cols, l_rows, l_cols_nev, nbw, num_blocks
   complex*16, allocatable       :: tmat(:,:,:)
   real*8, allocatable           :: q_real(:,:), e(:)
   real*8                        :: ttt0, ttt1, ttts
   integer                       :: i
401

402
403
404
   logical                       :: success, wantDebug
   logical, save                 :: firstCall = .true.

405
406
407
#ifdef HAVE_DETAILED_TIMINGS
   call timer%start("solve_evp_complex_2stage")
#endif
Andreas Marek's avatar
Andreas Marek committed
408
409
   call mpi_comm_rank(mpi_comm_all,my_pe,mpierr)
   call mpi_comm_size(mpi_comm_all,n_pes,mpierr)
410
411
412
413
414

   call mpi_comm_rank(mpi_comm_rows,my_prow,mpierr)
   call mpi_comm_size(mpi_comm_rows,np_rows,mpierr)
   call mpi_comm_rank(mpi_comm_cols,my_pcol,mpierr)
   call mpi_comm_size(mpi_comm_cols,np_cols,mpierr)
415

416
417
418
419
420
421
422
423
   wantDebug = .false.
   if (firstCall) then
     ! are debug messages desired?
     wantDebug = debug_messages_via_environment_variable()
     firstCall = .false.
   endif


424
425
   success = .true.

426
427
428
   if (present(THIS_COMPLEX_ELPA_KERNEL_API)) then
     ! user defined kernel via the optional argument in the API call
     THIS_COMPLEX_ELPA_KERNEL = THIS_COMPLEX_ELPA_KERNEL_API
Andreas Marek's avatar
Andreas Marek committed
429
   else
430
431
432
     ! if kernel is not choosen via api
     ! check whether set by environment variable
     THIS_COMPLEX_ELPA_KERNEL = get_actual_complex_kernel()
Andreas Marek's avatar
Andreas Marek committed
433
   endif
434

Andreas Marek's avatar
Andreas Marek committed
435
436
   ! check whether choosen kernel is allowed
   if (check_allowed_complex_kernels(THIS_COMPLEX_ELPA_KERNEL)) then
437

438
439
440
441
442
443
444
445
446
447
448
     if (my_pe == 0) then
       write(error_unit,*) " "
       write(error_unit,*) "The choosen kernel ",COMPLEX_ELPA_KERNEL_NAMES(THIS_COMPLEX_ELPA_KERNEL)
       write(error_unit,*) "is not in the list of the allowed kernels!"
       write(error_unit,*) " "
       write(error_unit,*) "Allowed kernels are:"
       do i=1,size(COMPLEX_ELPA_KERNEL_NAMES(:))
         if (AVAILABLE_COMPLEX_ELPA_KERNELS(i) .ne. 0) then
           write(error_unit,*) COMPLEX_ELPA_KERNEL_NAMES(i)
         endif
       enddo
Andreas Marek's avatar
Andreas Marek committed
449

450
451
452
453
       write(error_unit,*) " "
       write(error_unit,*) "The defaul kernel COMPLEX_ELPA_KERNEL_GENERIC will be used !"
     endif
     THIS_COMPLEX_ELPA_KERNEL = COMPLEX_ELPA_KERNEL_GENERIC
Andreas Marek's avatar
Andreas Marek committed
454
455
!      call MPI_ABORT(mpi_comm_all, mpierr)
   endif
456
457
458
459
460
461
462
463
464
465
466
467
   ! Choose bandwidth, must be a multiple of nblk, set to a value >= 32

   nbw = (31/nblk+1)*nblk

   num_blocks = (na-1)/nbw + 1

   allocate(tmat(nbw,nbw,num_blocks))

   ! Reduction full -> band

   ttt0 = MPI_Wtime()
   ttts = ttt0
468
   call bandred_complex(na, a, lda, nblk, nbw, mpi_comm_rows, mpi_comm_cols, &
469
                        tmat, wantDebug, success)
470
471
472
473
474
475
   if (.not.(success)) then
#ifdef HAVE_DETAILED_TIMINGS
     call timer%stop()
#endif
     return
   endif
476
   ttt1 = MPI_Wtime()
477
   if (my_prow==0 .and. my_pcol==0 .and. elpa_print_times) &
478
      write(error_unit,*) 'Time bandred_complex               :',ttt1-ttt0
479
480
481
482
483
484
485
486

   ! Reduction band -> tridiagonal

   allocate(e(na))

   ttt0 = MPI_Wtime()
   call tridiag_band_complex(na, nbw, nblk, a, lda, ev, e, mpi_comm_rows, mpi_comm_cols, mpi_comm_all)
   ttt1 = MPI_Wtime()
487
   if (my_prow==0 .and. my_pcol==0 .and. elpa_print_times) &
488
      write(error_unit,*) 'Time tridiag_band_complex          :',ttt1-ttt0
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504

   call mpi_bcast(ev,na,MPI_REAL8,0,mpi_comm_all,mpierr)
   call mpi_bcast(e,na,MPI_REAL8,0,mpi_comm_all,mpierr)

   ttt1 = MPI_Wtime()
   time_evp_fwd = ttt1-ttts

   l_rows = local_index(na, my_prow, np_rows, nblk, -1) ! Local rows of a and q
   l_cols = local_index(na, my_pcol, np_cols, nblk, -1) ! Local columns of q
   l_cols_nev = local_index(nev, my_pcol, np_cols, nblk, -1) ! Local columns corresponding to nev

   allocate(q_real(l_rows,l_cols))

   ! Solve tridiagonal system

   ttt0 = MPI_Wtime()
505
   call solve_tridi(na, nev, ev, e, q_real, ubound(q_real,1), nblk, &
506
                    mpi_comm_rows, mpi_comm_cols, wantDebug, success)
507
508
   if (.not.(success)) return

509
   ttt1 = MPI_Wtime()
510
   if (my_prow==0 .and. my_pcol==0 .and. elpa_print_times)  &
511
      write(error_unit,*) 'Time solve_tridi                   :',ttt1-ttt0
512
513
514
515
516
517
518
519
520
521
   time_evp_solve = ttt1-ttt0
   ttts = ttt1

   q(1:l_rows,1:l_cols_nev) = q_real(1:l_rows,1:l_cols_nev)

   deallocate(e, q_real)

   ! Backtransform stage 1

   ttt0 = MPI_Wtime()
Andreas Marek's avatar
Andreas Marek committed
522
   call trans_ev_tridi_to_band_complex(na, nev, nblk, nbw, q, ldq,  &
523
                                       mpi_comm_rows, mpi_comm_cols,&
524
                                       wantDebug, success,THIS_COMPLEX_ELPA_KERNEL)
525
   if (.not.(success)) return
526
   ttt1 = MPI_Wtime()
527
   if (my_prow==0 .and. my_pcol==0 .and. elpa_print_times) &
528
      write(error_unit,*) 'Time trans_ev_tridi_to_band_complex:',ttt1-ttt0
529
530
531
532
533
534
535
536
537

   ! We can now deallocate the stored householder vectors
   deallocate(hh_trans_complex)

   ! Backtransform stage 2

   ttt0 = MPI_Wtime()
   call trans_ev_band_to_full_complex(na, nev, nblk, nbw, a, lda, tmat, q, ldq, mpi_comm_rows, mpi_comm_cols)
   ttt1 = MPI_Wtime()
538
   if (my_prow==0 .and. my_pcol==0 .and. elpa_print_times) &
539
      write(error_unit,*) 'Time trans_ev_band_to_full_complex :',ttt1-ttt0
540
541
542
   time_evp_back = ttt1-ttts

   deallocate(tmat)
543
544
545
#ifdef HAVE_DETAILED_TIMINGS
   call timer%stop("solve_evp_complex_2stage")
#endif
546
547
548

1  format(a,f10.3)

549
end function solve_evp_complex_2stage
550
551
552

!-------------------------------------------------------------------------------

553
subroutine bandred_real(na, a, lda, nblk, nbw, mpi_comm_rows, mpi_comm_cols, &
554
                        tmat, wantDebug, success, useQR)
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

!-------------------------------------------------------------------------------
!  bandred_real: Reduces a distributed symmetric matrix to band form
!
!  Parameters
!
!  na          Order of matrix
!
!  a(lda,*)    Distributed matrix which should be reduced.
!              Distribution is like in Scalapack.
!              Opposed to Scalapack, a(:,:) must be set completely (upper and lower half)
!              a(:,:) is overwritten on exit with the band and the Householder vectors
!              in the upper half.
!
!  lda         Leading dimension of a
!
!  nblk        blocksize of cyclic distribution, must be the same in both directions!
!
!  nbw         semi bandwith of output matrix
!
!  mpi_comm_rows
!  mpi_comm_cols
!              MPI-Communicators for rows/columns
!
!  tmat(nbw,nbw,num_blocks)    where num_blocks = (na-1)/nbw + 1
!              Factors for the Householder vectors (returned), needed for back transformation
!
!-------------------------------------------------------------------------------
583
584
585
#ifdef HAVE_DETAILED_TIMINGS
 use timings
#endif
586
   implicit none
587
   
588
589
   integer             :: na, lda, nblk, nbw, mpi_comm_rows, mpi_comm_cols
   real*8              :: a(lda,*), tmat(nbw,nbw,*)
590

591
592
593
594
595
   integer             :: my_prow, my_pcol, np_rows, np_cols, mpierr
   integer             :: l_cols, l_rows
   integer             :: i, j, lcs, lce, lre, lc, lr, cur_pcol, n_cols, nrow
   integer             :: istep, ncol, lch, lcx, nlc
   integer             :: tile_size, l_rows_tile, l_cols_tile
596

597
   real*8              :: vnorm2, xf, aux1(nbw), aux2(nbw), vrl, tau, vav(nbw,nbw)
598

599
   real*8, allocatable :: tmp(:,:), vr(:), vmr(:,:), umc(:,:)
600

601
   integer             :: pcol, prow
602
603
604
605
606
607

   ! needed for blocked QR decomposition
   integer             :: PQRPARAM(11), work_size
   real*8              :: dwork_size(1)
   real*8, allocatable :: work_blocked(:), tauvector(:), blockheuristic(:)

608
609
610
   pcol(i) = MOD((i-1)/nblk,np_cols) !Processor col for global col number
   prow(i) = MOD((i-1)/nblk,np_rows) !Processor row for global row number

611
   logical, intent(in) :: wantDebug
612
613
   logical, intent(out):: success

614
615
   logical, intent(in) :: useQR

616
617
618
#ifdef HAVE_DETAILED_TIMINGS
   call timer%start("bandred_real")
#endif
619
620
621
622
   call mpi_comm_rank(mpi_comm_rows,my_prow,mpierr)
   call mpi_comm_size(mpi_comm_rows,np_rows,mpierr)
   call mpi_comm_rank(mpi_comm_cols,my_pcol,mpierr)
   call mpi_comm_size(mpi_comm_cols,np_cols,mpierr)
623
   success = .true.
624
625


626
   ! Semibandwith nbw must be a multiple of blocksize nblk
627
628
   if (mod(nbw,nblk)/=0) then
     if (my_prow==0 .and. my_pcol==0) then
629
630
631
632
       if (wantDebug) then
         write(error_unit,*) 'ELPA2_bandred_real: ERROR: nbw=',nbw,', nblk=',nblk
         write(error_unit,*) 'ELPA2_bandred_real: ELPA2 works only for nbw==n*nblk'
       endif
633
       success = .false.
634
!           call mpi_abort(mpi_comm_world,0,mpierr)
635
     endif
636
637
638
639
640
641
642
643
644
645
   endif

   ! Matrix is split into tiles; work is done only for tiles on the diagonal or above

   tile_size = nblk*least_common_multiple(np_rows,np_cols) ! minimum global tile size
   tile_size = ((128*max(np_rows,np_cols)-1)/tile_size+1)*tile_size ! make local tiles at least 128 wide

   l_rows_tile = tile_size/np_rows ! local rows of a tile
   l_cols_tile = tile_size/np_cols ! local cols of a tile

646
647
648
649
650
651
652
   if (useQR) then
     if (which_qr_decomposition == 1) then
       call qr_pqrparam_init(pqrparam,    nblk,'M',0,   nblk,'M',0,   nblk,'M',1,'s')
       allocate(tauvector(na))
       allocate(blockheuristic(nblk))
       l_rows = local_index(na, my_prow, np_rows, nblk, -1)
       allocate(vmr(max(l_rows,1),na))
653

654
       call qr_pdgeqrf_2dcomm(a, lda, vmr, max(l_rows,1), tauvector(1), tmat(1,1,1), nbw, dwork_size(1), -1, na, &
655
                             nbw, nblk, nblk, na, na, 1, 0, PQRPARAM, mpi_comm_rows, mpi_comm_cols, blockheuristic)
656
657
       work_size = dwork_size(1)
       allocate(work_blocked(work_size))
658

659
660
661
       work_blocked = 0.0d0
       deallocate(vmr)
     endif
662
663
   endif

664
665
   do istep = (na-1)/nbw, 1, -1

666
     n_cols = MIN(na,(istep+1)*nbw) - istep*nbw ! Number of columns in current step
667

668
669
670
     ! Number of local columns/rows of remaining matrix
     l_cols = local_index(istep*nbw, my_pcol, np_cols, nblk, -1)
     l_rows = local_index(istep*nbw, my_prow, np_rows, nblk, -1)
671

672
     ! Allocate vmr and umc to their exact sizes so that they can be used in bcasts and reduces
673

674
675
     allocate(vmr(max(l_rows,1),2*n_cols))
     allocate(umc(max(l_cols,1),2*n_cols))
676

677
     allocate(vr(l_rows+1))
678

679
680
681
     vmr(1:l_rows,1:n_cols) = 0.
     vr(:) = 0
     tmat(:,:,istep) = 0
682

683
     ! Reduce current block to lower triangular form
684
685
686
687
688
689
690
691
692
693

     if (useQR) then
       if (which_qr_decomposition == 1) then
         call qr_pdgeqrf_2dcomm(a, lda, vmr, max(l_rows,1), tauvector(1), &
                                  tmat(1,1,istep), nbw, work_blocked,       &
                                  work_size, na, n_cols, nblk, nblk,        &
                                  istep*nbw+n_cols-nbw, istep*nbw+n_cols, 1,&
                                  0, PQRPARAM, mpi_comm_rows, mpi_comm_cols,&
                                  blockheuristic)
       endif
694
     else
695

696
       do lc = n_cols, 1, -1
697

698
699
         ncol = istep*nbw + lc ! absolute column number of householder vector
         nrow = ncol - nbw ! Absolute number of pivot row
700

701
702
         lr  = local_index(nrow, my_prow, np_rows, nblk, -1) ! current row length
         lch = local_index(ncol, my_pcol, np_cols, nblk, -1) ! HV local column number
703

704
         tau = 0
705

706
         if (nrow == 1) exit ! Nothing to do
707

708
         cur_pcol = pcol(ncol) ! Processor column owning current block
709

710
         if (my_pcol==cur_pcol) then
711

712
713
           ! Get vector to be transformed; distribute last element and norm of
           ! remaining elements to all procs in current column
714

715
           vr(1:lr) = a(1:lr,lch) ! vector to be transformed
716

717
718
719
720
721
722
723
           if (my_prow==prow(nrow)) then
             aux1(1) = dot_product(vr(1:lr-1),vr(1:lr-1))
             aux1(2) = vr(lr)
           else
             aux1(1) = dot_product(vr(1:lr),vr(1:lr))
             aux1(2) = 0.
           endif
724

725
           call mpi_allreduce(aux1,aux2,2,MPI_REAL8,MPI_SUM,mpi_comm_rows,mpierr)
726

727
728
           vnorm2 = aux2(1)
           vrl    = aux2(2)
729

730
           ! Householder transformation
731

732
           call hh_transform_real(vrl, vnorm2, xf, tau)
733

734
           ! Scale vr and store Householder vector for back transformation
735

736
737
738
739
740
741
742
           vr(1:lr) = vr(1:lr) * xf
           if (my_prow==prow(nrow)) then
             a(1:lr-1,lch) = vr(1:lr-1)
             a(lr,lch) = vrl
             vr(lr) = 1.
           else
             a(1:lr,lch) = vr(1:lr)
743
           endif
744

745
         endif
746

747
         ! Broadcast Householder vector and tau along columns
748

749
750
751
752
753
         vr(lr+1) = tau
         call MPI_Bcast(vr,lr+1,MPI_REAL8,cur_pcol,mpi_comm_cols,mpierr)
         vmr(1:lr,lc) = vr(1:lr)
         tau = vr(lr+1)
         tmat(lc,lc,istep) = tau ! Store tau in diagonal of tmat
754

755
756
         ! Transform remaining columns in current block with Householder vector
         ! Local dot product
757

758
         aux1 = 0
759

760
761
762
763
764
765
766
767
         nlc = 0 ! number of local columns
         do j=1,lc-1
           lcx = local_index(istep*nbw+j, my_pcol, np_cols, nblk, 0)
           if (lcx>0) then
             nlc = nlc+1
             if (lr>0) aux1(nlc) = dot_product(vr(1:lr),a(1:lr,lcx))
           endif
         enddo
768

769
770
         ! Get global dot products
         if (nlc>0) call mpi_allreduce(aux1,aux2,nlc,MPI_REAL8,MPI_SUM,mpi_comm_rows,mpierr)
771

772
         ! Transform
773

774
775
776
777
778
779
780
781
782
783
         nlc = 0
         do j=1,lc-1
           lcx = local_index(istep*nbw+j, my_pcol, np_cols, nblk, 0)
           if (lcx>0) then
             nlc = nlc+1
             a(1:lr,lcx) = a(1:lr,lcx) - tau*aux2(nlc)*vr(1:lr)
           endif
         enddo

       enddo
784

785
786
       ! Calculate scalar products of stored Householder vectors.
       ! This can be done in different ways, we use dsyrk
787

788
789
       vav = 0
       if (l_rows>0) &
790
           call dsyrk('U','T',n_cols,l_rows,1.d0,vmr,ubound(vmr,1),0.d0,vav,ubound(vav,1))
791
       call symm_matrix_allreduce(n_cols,vav,ubound(vav,1),mpi_comm_rows)
792

793
       ! Calculate triangular matrix T for block Householder Transformation
794

795
796
797
798
799
800
801
       do lc=n_cols,1,-1
         tau = tmat(lc,lc,istep)
         if (lc<n_cols) then
           call dtrmv('U','T','N',n_cols-lc,tmat(lc+1,lc+1,istep),ubound(tmat,1),vav(lc+1,lc),1)
           tmat(lc,lc+1:n_cols,istep) = -tau * vav(lc+1:n_cols,lc)
         endif
       enddo
802
     endif
803

804
    ! Transpose vmr -> vmc (stored in umc, second half)
805

806
    call elpa_transpose_vectors  (vmr, ubound(vmr,1), mpi_comm_rows, &
807
808
809
                                    umc(1,n_cols+1), ubound(umc,1), mpi_comm_cols, &
                                    1, istep*nbw, n_cols, nblk)

810
811
812
813
    ! Calculate umc = A**T * vmr
    ! Note that the distributed A has to be transposed
    ! Opposed to direct tridiagonalization there is no need to use the cache locality
    ! of the tiles, so we can use strips of the matrix
814

815
816
817
818
    umc(1:l_cols,1:n_cols) = 0.d0
    vmr(1:l_rows,n_cols+1:2*n_cols) = 0
    if (l_cols>0 .and. l_rows>0) then
      do i=0,(istep*nbw-1)/tile_size
819

820
821
822
        lcs = i*l_cols_tile+1
        lce = min(l_cols,(i+1)*l_cols_tile)
        if (lce<lcs) cycle
823

824
825
826
        lre = min(l_rows,(i+1)*l_rows_tile)
        call DGEMM('T','N',lce-lcs+1,n_cols,lre,1.d0,a(1,lcs),ubound(a,1), &
                     vmr,ubound(vmr,1),1.d0,umc(lcs,1),ubound(umc,1))
827

828
829
830
831
832
833
        if (i==0) cycle
        lre = min(l_rows,i*l_rows_tile)
        call DGEMM('N','N',lre,n_cols,lce-lcs+1,1.d0,a(1,lcs),lda, &
                     umc(lcs,n_cols+1),ubound(umc,1),1.d0,vmr(1,n_cols+1),ubound(vmr,1))
      enddo
    endif
834

835
836
837
838
    ! Sum up all ur(:) parts along rows and add them to the uc(:) parts
    ! on the processors containing the diagonal
    ! This is only necessary if ur has been calculated, i.e. if the
    ! global tile size is smaller than the global remaining matrix
839

840
841
842
843
844
    if (tile_size < istep*nbw) then
       call elpa_reduce_add_vectors  (vmr(1,n_cols+1),ubound(vmr,1),mpi_comm_rows, &
                                      umc, ubound(umc,1), mpi_comm_cols, &
                                      istep*nbw, n_cols, nblk)
    endif
845

846
847
848
849
850
851
    if (l_cols>0) then
      allocate(tmp(l_cols,n_cols))
      call mpi_allreduce(umc,tmp,l_cols*n_cols,MPI_REAL8,MPI_SUM,mpi_comm_rows,mpierr)
      umc(1:l_cols,1:n_cols) = tmp(1:l_cols,1:n_cols)
      deallocate(tmp)
    endif
852

853
    ! U = U * Tmat**T
854

855
    call dtrmm('Right','Upper','Trans','Nonunit',l_cols,n_cols,1.d0,tmat(1,1,istep),ubound(tmat,1),umc,ubound(umc,1))
856

857
    ! VAV = Tmat * V**T * A * V * Tmat**T = (U*Tmat**T)**T * V * Tmat**T
858

859
860
    call dgemm('T','N',n_cols,n_cols,l_cols,1.d0,umc,ubound(umc,1),umc(1,n_cols+1),ubound(umc,1),0.d0,vav,ubound(vav,1))
    call dtrmm('Right','Upper','Trans','Nonunit',n_cols,n_cols,1.d0,tmat(1,1,istep),ubound(tmat,1),vav,ubound(vav,1))
861

862
    call symm_matrix_allreduce(n_cols,vav,ubound(vav,1),mpi_comm_cols)
863

864
865
    ! U = U - 0.5 * V * VAV
    call dgemm('N','N',l_cols,n_cols,n_cols,-0.5d0,umc(1,n_cols+1),ubound(umc,1),vav,ubound(vav,1),1.d0,umc,ubound(umc,1))
866

867
    ! Transpose umc -> umr (stored in vmr, second half)
868

869
870
871
    call elpa_transpose_vectors  (umc, ubound(umc,1), mpi_comm_cols, &
                                   vmr(1,n_cols+1), ubound(vmr,1), mpi_comm_rows, &
                                   1, istep*nbw, n_cols, nblk)
872

873
    ! A = A - V*U**T - U*V**T
874

875
876
877
878
879
880
881
882
883
    do i=0,(istep*nbw-1)/tile_size
      lcs = i*l_cols_tile+1
      lce = min(l_cols,(i+1)*l_cols_tile)
      lre = min(l_rows,(i+1)*l_rows_tile)
      if (lce<lcs .or. lre<1) cycle
      call dgemm('N','T',lre,lce-lcs+1,2*n_cols,-1.d0, &
                  vmr,ubound(vmr,1),umc(lcs,1),ubound(umc,1), &
                  1.d0,a(1,lcs),lda)
    enddo
884

885
    deallocate(vmr, umc, vr)
886

887
  enddo
888

889
890
891
892
893
  if (useQR) then
    if (which_qr_decomposition == 1) then
      deallocate(work_blocked)
      deallocate(tauvector)
    endif
894
  endif
895

Andreas Marek's avatar
Andreas Marek committed
896
897
898
#ifdef HAVE_DETAILED_TIMINGS
  call timer%stop("bandred_real")
#endif
899
900
901
902
903
904
905
906
907
908
909
end subroutine bandred_real

!-------------------------------------------------------------------------------

subroutine symm_matrix_allreduce(n,a,lda,comm)

!-------------------------------------------------------------------------------
!  symm_matrix_allreduce: Does an mpi_allreduce for a symmetric matrix A.
!  On entry, only the upper half of A needs to be set
!  On exit, the complete matrix is set
!-------------------------------------------------------------------------------
Andreas Marek's avatar
Andreas Marek committed
910
911
912
#ifdef HAVE_DETAILED_TIMINGS
 use timings
#endif
913
   implicit none
Andreas Marek's avatar
Andreas Marek committed
914
915
916
917
918
   integer  :: n, lda, comm
   real*8   :: a(lda,*)

   integer  :: i, nc, mpierr
   real*8   :: h1(n*n), h2(n*n)
919

Andreas Marek's avatar
Andreas Marek committed
920
921
922
#ifdef HAVE_DETAILED_TIMINGS
  call timer%start("symm_matrix_allreduce")
#endif
923
924
925

   nc = 0
   do i=1,n
926
927
     h1(nc+1:nc+i) = a(1:i,i)
     nc = nc+i
928
929
930
931
932
933
   enddo

   call mpi_allreduce(h1,h2,nc,MPI_REAL8,MPI_SUM,comm,mpierr)

   nc = 0
   do i=1,n
934
935
936
     a(1:i,i) = h2(nc+1:nc+i)
     a(i,1:i-1) = a(1:i-1,i)
     nc = nc+i
937
938
   enddo

Andreas Marek's avatar
Andreas Marek committed
939
940
941
942
#ifdef HAVE_DETAILED_TIMINGS
  call timer%stop("symm_matrix_allreduce")
#endif

943
944
945
946
end subroutine symm_matrix_allreduce

!-------------------------------------------------------------------------------

947
948
subroutine trans_ev_band_to_full_real(na, nqc, nblk, nbw, a, lda, tmat, q, ldq, mpi_comm_rows, &
                                      mpi_comm_cols, useQR)
949

Andreas Marek's avatar
Andreas Marek committed
950

951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
!-------------------------------------------------------------------------------
!  trans_ev_band_to_full_real:
!  Transforms the eigenvectors of a band matrix back to the eigenvectors of the original matrix
!
!  Parameters
!
!  na          Order of matrix a, number of rows of matrix q
!
!  nqc         Number of columns of matrix q
!
!  nblk        blocksize of cyclic distribution, must be the same in both directions!
!
!  nbw         semi bandwith
!
!  a(lda,*)    Matrix containing the Householder vectors (i.e. matrix a after bandred_real)
!              Distribution is like in Scalapack.
!
!  lda         Leading dimension of a
!
!  tmat(nbw,nbw,.) Factors returned by bandred_real
!
!  q           On input: Eigenvectors of band matrix
!              On output: Transformed eigenvectors
!              Distribution is like in Scalapack.
!
!  ldq         Leading dimension of q
!
!  mpi_comm_rows
!  mpi_comm_cols
!              MPI-Communicators for rows/columns
!
!-------------------------------------------------------------------------------
983
984
985
#ifdef HAVE_DETAILED_TIMINGS
 use timings
#endif
986
987
   implicit none

988
989
   integer              :: na, nqc, lda, ldq, nblk, nbw, mpi_comm_rows, mpi_comm_cols
   real*8               :: a(lda,*), q(ldq,*), tmat(nbw, nbw, *)
990

991
992
993
994
995
   integer              :: my_prow, my_pcol, np_rows, np_cols, mpierr
   integer              :: max_blocks_row, max_blocks_col, max_local_rows, &
                           max_local_cols
   integer              :: l_cols, l_rows, l_colh, n_cols
   integer              :: istep, lc, ncol, nrow, nb, ns
996

997
   real*8, allocatable  :: tmp1(:), tmp2(:), hvb(:), hvm(:,:)
998

999
   integer              :: pcol, prow, i
1000
1001

   real*8, allocatable  :: tmat_complete(:,:), t_tmp(:,:), t_tmp2(:,:)
1002
1003
1004
   integer              :: cwy_blocking, t_blocking, t_cols, t_rows
   logical, intent(in)  :: useQR

1005
1006
1007
   pcol(i) = MOD((i-1)/nblk,np_cols) !Processor col for global col number
   prow(i) = MOD((i-1)/nblk,np_rows) !Processor row for global row number

1008
1009
1010
#ifdef HAVE_DETAILED_TIMINGS
   call timer%start("trans_ev_band_to_full_real")
#endif
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022

   call mpi_comm_rank(mpi_comm_rows,my_prow,mpierr)
   call mpi_comm_size(mpi_comm_rows,np_rows,mpierr)
   call mpi_comm_rank(mpi_comm_cols,my_pcol,mpierr)
   call mpi_comm_size(mpi_comm_cols,np_cols,mpierr)

   max_blocks_row = ((na -1)/nblk)/np_rows + 1  ! Rows of A
   max_blocks_col = ((nqc-1)/nblk)/np_cols + 1  ! Columns of q!

   max_local_rows = max_blocks_row*nblk
   max_local_cols = max_blocks_col*nblk

1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
   if (useQR) then
     t_blocking = 2 ! number of matrices T (tmat) which are aggregated into a new (larger) T matrix (tmat_complete) and applied at once
     cwy_blocking = t_blocking * nbw

     allocate(tmp1(max_local_cols*cwy_blocking))
     allocate(tmp2(max_local_cols*cwy_blocking))
     allocate(hvb(max_local_rows*cwy_blocking))
     allocate(hvm(max_local_rows,cwy_blocking))
     allocate(tmat_complete(cwy_blocking,cwy_blocking))
     allocate(t_tmp(cwy_blocking,nbw))
     allocate(t_tmp2(cwy_blocking,nbw))
   else
     allocate(tmp1(max_local_cols*nbw))
     allocate(tmp2(max_local_cols*nbw))
     allocate(hvb(max_local_rows*nbw))
     allocate(hvm(max_local_rows,nbw))
   endif
1040
1041
1042
1043
1044
1045

   hvm = 0   ! Must be set to 0 !!!
   hvb = 0   ! Safety only

   l_cols = local_index(nqc, my_pcol, np_cols, nblk, -1) ! Local columns of q

1046
   if (useQR) then
1047

1048
1049
     do istep=1,((na-1)/nbw-1)/t_blocking + 1
       n_cols = MIN(na,istep*cwy_blocking+nbw) - (istep-1)*cwy_blocking - nbw ! Number of columns in current step
1050

1051
       ! Broadcast all Householder vectors for current step compressed in hvb
1052

1053
1054
       nb = 0
       ns = 0
1055

1056
1057
1058
       do lc = 1, n_cols
         ncol = (istep-1)*cwy_blocking + nbw + lc ! absolute column number of householder vector
         nrow = ncol - nbw ! absolute number of pivot row
1059

1060
1061
         l_rows = local_index(nrow-1, my_prow, np_rows, nblk, -1) ! row length for bcast
         l_colh = local_index(ncol  , my_pcol, np_cols, nblk, -1) ! HV local column number
1062

1063
         if (my_pcol==pcol(ncol)) hvb(nb+1:nb+l_rows) = a(1:l_rows,l_colh)
1064

1065
         nb = nb+l_rows
1066

1067
1068
1069
1070
1071
         if (lc==n_cols .or. mod(ncol,nblk)==0) then
           call MPI_Bcast(hvb(ns+1),nb-ns,MPI_REAL8,pcol(ncol),mpi_comm_cols,mpierr)
           ns = nb
         endif
       enddo
1072

1073
       ! Expand compressed Householder vectors into matrix hvm
1074

1075
1076
1077
1078
       nb = 0
       do lc = 1, n_cols
         nrow = (istep-1)*cwy_blocking + lc ! absolute number of pivot row
         l_rows = local_index(nrow-1, my_prow, np_rows, nblk, -1) ! row length for bcast
1079

1080
1081
         hvm(1:l_rows,lc) = hvb(nb+1:nb+l_rows)
         if (my_prow==prow(nrow)) hvm(l_rows+1,lc) = 1.
1082

1083
1084
         nb = nb+l_rows
       enddo
1085

1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
       l_rows = local_index(MIN(na,(istep+1)*cwy_blocking), my_prow, np_rows, nblk, -1)

       ! compute tmat2 out of tmat(:,:,)
       tmat_complete = 0
       do i = 1, t_blocking
         t_cols = MIN(nbw, n_cols - (i-1)*nbw)
         if (t_cols <= 0) exit
         t_rows = (i - 1) * nbw
         tmat_complete(t_rows+1:t_rows+t_cols,t_rows+1:t_rows+t_cols) = tmat(1:t_cols,1:t_cols,(istep-1)*t_blocking + i)
         if (i > 1) then
           call dgemm('T', 'N', t_rows, t_cols, l_rows, 1.d0, hvm(1,1), max_local_rows, hvm(1,(i-1)*nbw+1), &
1097
                     max_local_rows, 0.d0, t_tmp, cwy_blocking)
1098
1099
1100
1101
1102
1103
           call mpi_allreduce(t_tmp,t_tmp2,cwy_blocking*nbw,MPI_REAL8,MPI_SUM,mpi_comm_rows,mpierr)
           call dtrmm('L','U','N','N',t_rows,t_cols,1.0d0,tmat_complete,cwy_blocking,t_tmp2,cwy_blocking)
           call dtrmm('R','U','N','N',t_rows,t_cols,-1.0d0,tmat_complete(t_rows+1,t_rows+1),cwy_blocking,t_tmp2,cwy_blocking)
           tmat_complete(1:t_rows,t_rows+1:t_rows+t_cols) = t_tmp2(1:t_rows,1:t_cols)
         endif
       enddo
1104

1105
       ! Q = Q - V * T**T * V**T * Q
1106

1107
       if (l_rows>0) then
1108
1109
         call dgemm('T','N',n_cols,l_cols,l_rows,1.d0,hvm,ubound(hvm,1), &
                    q,ldq,0.d0,tmp1,n_cols)
1110
       else
1111
         tmp1(1:l_cols*n_cols) = 0
1112
1113
1114
1115
1116
       endif
       call mpi_allreduce(tmp1,tmp2,n_cols*l_cols,MPI_REAL8,MPI_SUM,mpi_comm_rows,mpierr)


       if (l_rows>0) then
1117
1118
         call dtrmm('L','U','T','N',n_cols,l_cols,1.0d0,tmat_complete,cwy_blocking,tmp2,n_cols)
         call dgemm('N','N',l_rows,l_cols,n_cols,-1.d0,hvm,ubound(hvm,1), tmp2,n_cols,1.d0,q,ldq)
1119
1120
       endif
     enddo
1121

1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146