legacy_single_toeplitz.F90 11.9 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
!    This file is part of ELPA.
!
!    The ELPA library was originally created by the ELPA consortium,
!    consisting of the following organizations:
!
!    - Max Planck Computing and Data Facility (MPCDF), formerly known as
!      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 Naturwissenschaften,
!      Leipzig, Abt. Komplexe Strukutren in Biologie und Kognition,
!      and
!    - IBM Deutschland GmbH
!
!
!    More information can be found here:
!    http://elpa.mpcdf.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.
!
!
#include "config-f90.h"
!>
!> Fortran test programm to demonstrates the use of
!> the elpa_solve_tridi library function
!>
program test_solve_tridi_single

!
! 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".
!
!-------------------------------------------------------------------------------
   use precision
58
   use elpa1
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
   use elpa_utilities
#ifdef WITH_OPENMP
   use test_util
#endif

   use mod_read_input_parameters
   use mod_check_correctness
   use mod_setup_mpi
   use mod_blacs_infrastructure
   use mod_prepare_matrix

   use elpa_mpi
#ifdef HAVE_REDIRECT
   use redirect
#endif
#ifdef HAVE_DETAILED_TIMINGS
  use timings
#endif
  use output_types

   implicit none

   !-------------------------------------------------------------------------------
   ! Please set system size parameters below!
   ! na:   System size
   ! nev:  Number of eigenvectors to be calculated
   ! nblk: Blocking factor in block cyclic distribution
   !-------------------------------------------------------------------------------
   integer(kind=ik)            :: nblk
   integer(kind=ik)            :: na, nev

   integer(kind=ik)            :: np_rows, np_cols, na_rows, na_cols

   integer(kind=ik)            :: myid, nprocs, my_prow, my_pcol, mpi_comm_rows, mpi_comm_cols
   integer(kind=ik)            :: i, mpierr, my_blacs_ctxt, sc_desc(9), info, nprow, npcol

   integer, external           :: numroc

   real(kind=rk4), allocatable :: a(:,:), d(:), e(:), ev_analytic(:), ev(:)
   real(kind=rk4)              :: diagonalELement, subdiagonalElement

100
   real(kind=rk4), allocatable :: as(:,:)
101
102
103
104
105
106
107
108
   real(kind=rk4)              :: tmp
   integer(kind=ik)            :: loctmp ,rowLocal, colLocal


   real(kind=rk4)              :: maxerr

   logical                     :: wantDebug

109
   real(kind=rk4), parameter   :: pi = 3.141592653589793238462643383279_rk4
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133

   integer(kind=ik)            :: STATUS
#ifdef WITH_OPENMP
   integer(kind=ik)            :: omp_get_max_threads,  required_mpi_thread_level, &
                                  provided_mpi_thread_level
#endif
   type(output_t)              :: write_to_file
   logical                     :: success
   character(len=8)            :: task_suffix
   integer(kind=ik)            :: j
   !-------------------------------------------------------------------------------

   success = .true.

   call read_input_parameters(na, nev, nblk, write_to_file)

   !-------------------------------------------------------------------------------
   !  MPI Initialization
   call setup_mpi(myid, nprocs)

   STATUS = 0

!#define DATATYPE REAL
!#define ELPA1
Andreas Marek's avatar
Andreas Marek committed
134
!#include "../../elpa_print_headers.X90"
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

#ifdef HAVE_DETAILED_TIMINGS

   ! initialise the timing functionality

#ifdef HAVE_LIBPAPI
   call timer%measure_flops(.true.)
#endif

   call timer%measure_allocated_memory(.true.)
   call timer%measure_virtual_memory(.true.)
   call timer%measure_max_allocated_memory(.true.)

   call timer%set_print_options(&
#ifdef HAVE_LIBPAPI
                print_flop_count=.true., &
                print_flop_rate=.true., &
#endif
                print_allocated_memory = .true. , &
                print_virtual_memory=.true., &
                print_max_allocated_memory=.true.)


  call timer%enable()

  call timer%start("program")
#endif

   do np_cols = NINT(SQRT(REAL(nprocs))),2,-1
      if(mod(nprocs,np_cols) == 0 ) exit
   enddo

   ! at the end of the above loop, nprocs is always divisible by np_cols

   np_rows = nprocs/np_cols

   if(myid==0) then
      print *
      print '(a)','Test program for elpa_solve_tridi with a Toeplitz matrix'
      print *
      print '(3(a,i0))','Matrix size=',na,', Number of eigenvectors=',nev,', Block size=',nblk
      print '(3(a,i0))','Number of processor rows=',np_rows,', cols=',np_cols,', total=',nprocs
      print *
   endif

   !-------------------------------------------------------------------------------
   ! Set up BLACS context and MPI communicators
   !
   ! The BLACS context is only necessary for using Scalapack.
   !
   ! For ELPA, the MPI communicators along rows/cols are sufficient,
   ! and the grid setup may be done in an arbitrary way as long as it is
   ! consistent (i.e. 0<=my_prow<np_rows, 0<=my_pcol<np_cols and every
   ! process has a unique (my_prow,my_pcol) pair).

   call set_up_blacsgrid(mpi_comm_world, my_blacs_ctxt, np_rows, np_cols, &
                         nprow, npcol, my_prow, my_pcol)

   if (myid==0) then
     print '(a)','| Past BLACS_Gridinfo.'
   end if

   ! All ELPA routines need MPI communicators for communicating within
198
   ! rows or columns of processes, these are set in elpa_get_communicators.
199

200
   mpierr = elpa_get_communicators(mpi_comm_world, my_prow, my_pcol, &
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
                                   mpi_comm_rows, mpi_comm_cols)

   if (myid==0) then
     print '(a)','| Past split communicator setup for rows and columns.'
   end if

   call set_up_blacs_descriptor(na ,nblk, my_prow, my_pcol, np_rows, np_cols, &
                                na_rows, na_cols, sc_desc, my_blacs_ctxt, info)

   if (myid==0) then
     print '(a)','| Past scalapack descriptor setup.'
   end if

   !-------------------------------------------------------------------------------
   ! Allocate matrices and set up a test toeplitz matrix for solve_tridi

#ifdef HAVE_DETAILED_TIMINGS
   call timer%start("set up matrix")
#endif
   allocate(a (na_rows,na_cols))
   allocate(as(na_rows,na_cols))

   allocate(d (na))
   allocate(e (na))
   allocate(ev_analytic(na))
   allocate(ev(na))

   a(:,:) = 0.0_rk4


   ! changeable numbers here would be nice
   diagonalElement = 0.45_rk4
   subdiagonalElement =  0.78_rk4

   d(:) = diagonalElement
   e(:) = subdiagonalElement


   ! set up the diagonal and subdiagonals (for general solver test)
   do i=1, na ! for diagonal elements
     if (map_global_array_index_to_local_index(i, i, rowLocal, colLocal, nblk, np_rows, np_cols, my_prow, my_pcol)) then
       a(rowLocal,colLocal) = diagonalElement
     endif
   enddo

   do i=1, na-1
     if (map_global_array_index_to_local_index(i, i+1, rowLocal, colLocal, nblk, np_rows, np_cols, my_prow, my_pcol)) then
       a(rowLocal,colLocal) = subdiagonalElement
     endif
   enddo

   do i=2, na
     if (map_global_array_index_to_local_index(i, i-1, rowLocal, colLocal, nblk, np_rows, np_cols, my_prow, my_pcol)) then
       a(rowLocal,colLocal) = subdiagonalElement
     endif
   enddo

   as = a

#ifdef HAVE_DETAILED_TIMINGS
   call timer%stop("set up matrix")
#endif


   !-------------------------------------------------------------------------------
   ! Calculate eigenvalues/eigenvectors

   if (myid==0) then
     print '(a)','| Entering elpa_solve_tridi ... '
     print *
   end if

#ifdef WITH_MPI
   call mpi_barrier(mpi_comm_world, mpierr) ! for correct timings only
#endif

   success = elpa_solve_tridi_single(na, nev, d, e, a, na_rows, nblk, na_cols, mpi_comm_rows, &
                                     mpi_comm_cols, wantDebug)

   if (.not.(success)) then
      write(error_unit,*) "elpa_solve_tridi produced an error! Aborting..."
#ifdef WITH_MPI
      call MPI_ABORT(mpi_comm_world, 1, mpierr)
#endif
   endif

   if (myid==0) then
     print '(a)','| elpa_solve_tridi complete.'
     print *
   end if


   ev = d

!
!   if(myid == 0) print *,'Time tridiag_real     :',time_evp_fwd
!   if(myid == 0) print *,'Time solve_tridi      :',time_evp_solve
!   if(myid == 0) print *,'Time trans_ev_real    :',time_evp_back
!   if(myid == 0) print *,'Total time (sum above):',time_evp_back+time_evp_solve+time_evp_fwd
!
!   if(write_to_file%eigenvectors) then
!     write(unit = task_suffix, fmt = '(i8.8)') myid
!     open(17,file="EVs_real_out_task_"//task_suffix(1:8)//".txt",form='formatted',status='new')
!     write(17,*) "Part of eigenvectors: na_rows=",na_rows,"of na=",na," na_cols=",na_cols," of na=",na
!
!     do i=1,na_rows
!       do j=1,na_cols
!         write(17,*) "row=",i," col=",j," element of eigenvector=",z(i,j)
!       enddo
!     enddo
!     close(17)
!   endif
!
!   if(write_to_file%eigenvalues) then
!      if (myid == 0) then
!         open(17,file="Eigenvalues_real_out.txt",form='formatted',status='new')
!         do i=1,na
!            write(17,*) i,ev(i)
!         enddo
!         close(17)
!      endif
!   endif
!
!
   !-------------------------------------------------------------------------------

   ! analytic solution
   do i=1, na
     ev_analytic(i) = diagonalElement + 2.0_rk4 * subdiagonalElement *cos( pi*real(i,kind=rk4)/ real(na+1,kind=rk4) )
   enddo

   ! sort analytic solution:

   ! this hack is neihter elegant, nor optimized: for huge matrixes it might be expensive
   ! a proper sorting algorithmus might be implemented here

   tmp    = minval(ev_analytic)
   loctmp = minloc(ev_analytic, 1)

   ev_analytic(loctmp) = ev_analytic(1)
   ev_analytic(1) = tmp

   do i=2, na
     tmp = ev_analytic(i)
     do j= i, na
       if (ev_analytic(j) .lt. tmp) then
         tmp    = ev_analytic(j)
         loctmp = j
       endif
     enddo
     ev_analytic(loctmp) = ev_analytic(i)
     ev_analytic(i) = tmp
   enddo

   ! compute a simple error max of eigenvalues
   maxerr = 0.0_rk4
   maxerr = maxval( (d(:) - ev_analytic(:))/ev_analytic(:) , 1)

   if (maxerr .gt. 8.e-13_rk4) then
     if (myid .eq. 0) then
       print *,"Eigenvalues differ from analytic solution: maxerr = ",maxerr
     endif

   status = 1

   endif

   ! Test correctness of result (using plain scalapack routines)
369
   status = check_correctness(na, nev, as, a, ev, sc_desc, myid)
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

   deallocate(a)
   deallocate(as)
   deallocate(d)
   deallocate(e)
   deallocate(ev)
   deallocate(ev_analytic)

#ifdef HAVE_DETAILED_TIMINGS
   call timer%stop("program")
   print *," "
   print *,"Timings program:"
   print *," "
   call timer%print("program")
   print *," "
   print *,"End timings program"
   print *," "
#endif

#ifdef WITH_MPI
   call blacs_gridexit(my_blacs_ctxt)
   call mpi_finalize(mpierr)
#endif

   call EXIT(STATUS)


end

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