test_check_correctness_template.F90 25.3 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
!    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.
!
! Author: A. Marek, MPCDF

    function check_correctness_&
    &MATH_DATATYPE&
    &_&
    &PRECISION&
Pavel Kus's avatar
Pavel Kus committed
48
    & (na, nev, as, z, ev, sc_desc, nblk, myid, np_rows, np_cols, my_prow, my_pcol) result(status)
49
      implicit none
50
#include "../../src/general/precision_kinds.F90"
51
      integer(kind=ik)                 :: status
Pavel Kus's avatar
Pavel Kus committed
52
      integer(kind=ik), intent(in)     :: na, nev, nblk, myid, np_rows, np_cols, my_prow, my_pcol
53
#if REALCASE == 1
54
55
56
57
      real(kind=rck), intent(in)     :: as(:,:), z(:,:)
      real(kind=rck)                 :: ev(:)
      real(kind=rck), dimension(size(as,dim=1),size(as,dim=2)) :: tmp1, tmp2
      real(kind=rck)                :: xc
58

59
60
61
#ifndef WITH_MPI

#ifdef DOUBLE_PRECISION_REAL
62
      real(kind=rck)                 :: dnrm2
63
#else
64
      real(kind=rck)                 :: snrm2
65
66
67
68
69
70
#endif

#endif
#endif /* REALCASE */

#if COMPLEXCASE == 1
71
72
73
74
      complex(kind=rck), intent(in)    :: as(:,:), z(:,:)
      real(kind=rck)                   :: ev(:)
      complex(kind=rck), dimension(size(as,dim=1),size(as,dim=2)) :: tmp1, tmp2
      complex(kind=rck)                :: xc
75
76
#ifdef DOUBLE_PRECISION_COMPLEX
#ifndef WITH_MPI
77
      complex(kind=rck)                :: zdotc, cdotc
78
79
80
81
#endif

#else /* DOUBLE_PRECISION_COMPLEX */
#ifndef WITH_MPI
82
      complex(kind=rck)                :: zdotc, cdotc
83
84
85
86
87
88
89
90
#endif

#endif /* DOUBLE_PRECISION_COMPLEX */

#endif /* COMPLEXCASE */

      integer(kind=ik)                 :: sc_desc(:)

Pavel Kus's avatar
Pavel Kus committed
91
      integer(kind=ik)                 :: i, rowLocal, colLocal
92
      real(kind=rck)                   :: err, errmax
93
94
95
96
97
98
99
100
101
102

      integer :: mpierr

      status = 0

      ! 1. Residual (maximum of || A*Zi - Zi*EVi ||)
      ! tmp1 =  A * Z
      ! as is original stored matrix, Z are the EVs

#ifdef WITH_MPI
103
104
      call scal_PRECISION_GEMM('N', 'N', na, nev, na, ONE, as, 1, 1, sc_desc, &
                  z, 1, 1, sc_desc, ZERO, tmp1, 1, 1, sc_desc)
105
#else /* WITH_MPI */
106
      call PRECISION_GEMM('N','N',na,nev,na,ONE,as,na,z,na,ZERO,tmp1,na)
107
108
109
110
111
112
#endif /* WITH_MPI */


      ! tmp2 = Zi*EVi
      tmp2(:,:) = z(:,:)
      do i=1,nev
Pavel Kus's avatar
Pavel Kus committed
113
        xc = ev(i)
114
115
116
117
#if REALCASE == 1
#ifdef WITH_MPI

#ifdef DOUBLE_PRECISION_REAL
Pavel Kus's avatar
Pavel Kus committed
118
        call pdscal(na, xc, tmp2, 1, i, sc_desc, 1)
119
#else
Pavel Kus's avatar
Pavel Kus committed
120
        call psscal(na, xc, tmp2, 1, i, sc_desc, 1)
121
122
123
124
125
#endif

#else /* WITH_MPI */

#ifdef DOUBLE_PRECISION_REAL
Pavel Kus's avatar
Pavel Kus committed
126
        call dscal(na,xc,tmp2(:,i),1)
127
#else
Pavel Kus's avatar
Pavel Kus committed
128
        call sscal(na,xc,tmp2(:,i),1)
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
#endif

#endif /* WITH_MPI */
#endif /* REALCASE */

#if COMPLEXCASE == 1
#ifdef WITH_MPI

#ifdef DOUBLE_PRECISION_COMPLEX
        call pzscal(na, xc, tmp2, 1, i, sc_desc, 1)
#else
        call pcscal(na, xc, tmp2, 1, i, sc_desc, 1)
#endif

#else /* WITH_MPI */

#ifdef DOUBLE_PRECISION_COMPLEX
        call zscal(na,xc,tmp2(1,i),1)
#else
        call cscal(na,xc,tmp2(1,i),1)
#endif

#endif /* WITH_MPI */
#endif /* COMPLEXCASE */
      enddo

      !  tmp1 = A*Zi - Zi*EVi
      tmp1(:,:) =  tmp1(:,:) - tmp2(:,:)

      ! Get maximum norm of columns of tmp1
159
      errmax = 0.0_rk
160
161
162

      do i=1,nev
#if REALCASE == 1
163
        err = 0.0_rk
164
#ifdef WITH_MPI
Pavel Kus's avatar
Pavel Kus committed
165
        call scal_PRECISION_NRM2(na, err, tmp1, 1, i, sc_desc, 1)
166
#else /* WITH_MPI */
Pavel Kus's avatar
Pavel Kus committed
167
        err = PRECISION_NRM2(na,tmp1(1,i),1)
168
169
170
171
172
173
174
#endif /* WITH_MPI */
        errmax = max(errmax, err)
#endif /* REALCASE */

#if COMPLEXCASE == 1
        xc = 0
#ifdef WITH_MPI
Pavel Kus's avatar
Pavel Kus committed
175
        call scal_PRECISION_DOTC(na, xc, tmp1, 1, i, sc_desc, 1, tmp1, 1, i, sc_desc, 1)
176
#else /* WITH_MPI */
Pavel Kus's avatar
Pavel Kus committed
177
        xc = PRECISION_DOTC(na,tmp1,1,tmp1,1)
178
#endif /* WITH_MPI */
179
        errmax = max(errmax, sqrt(real(xc,kind=REAL_DATATYPE)))
180
181
182
183
184
185
#endif /* COMPLEXCASE */
      enddo

      ! Get maximum error norm over all processors
      err = errmax
#ifdef WITH_MPI
186
      call mpi_allreduce(err, errmax, 1, MPI_REAL_PRECISION, MPI_MAX, MPI_COMM_WORLD, mpierr)
187
188
189
190
191
192
#else /* WITH_MPI */
      errmax = err
#endif /* WITH_MPI */

      if (myid==0) print *,'Error Residual     :',errmax
#if REALCASE == 1
193
      if (nev .ge. 2) then
194
#ifdef DOUBLE_PRECISION_REAL
195
        if (errmax .gt. 5e-12_rk8 .or. errmax .eq. 0.0_rk8) then
196
#else
197
        if (errmax .gt. 3e-3_rk4 .or. errmax .eq. 0.0_rk4) then
198
#endif
199
200
201
202
203
204
205
206
207
208
209
          status = 1
        endif
      else
#ifdef DOUBLE_PRECISION_REAL
        if (errmax .gt. 5e-12_rk8) then
#else
        if (errmax .gt. 3e-3_rk4) then
#endif
          status = 1
        endif
      endif
210
211
#endif
#if COMPLEXCASE == 1
212
      if (nev .gt. 2) then
213
#ifdef DOUBLE_PRECISION_COMPLEX
214
        if (errmax .gt. 5e-12_rk8 .or. errmax .eq. 0.0_rk8) then
215
#else
216
        if (errmax .gt. 3e-3_rk4 .or. errmax .eq. 0.0_rk4) then
217
#endif
218
219
220
221
222
223
224
          status =1
        endif
      else
#ifdef DOUBLE_PRECISION_COMPLEX
        if (errmax .gt. 5e-12_rk8) then
#else
        if (errmax .gt. 3e-3_rk4) then
225
#endif
226
227
          status = 1
        endif
228
      endif
229
#endif
230
231
232
233
234
235

      ! 2. Eigenvector orthogonality

      ! tmp1 = Z**T * Z
      tmp1 = 0
#ifdef WITH_MPI
236
237
      call scal_PRECISION_GEMM(BLAS_TRANS_OR_CONJ, 'N', nev, nev, na, ONE, z, 1, 1, &
                        sc_desc, z, 1, 1, sc_desc, ZERO, tmp1, 1, 1, sc_desc)
238
#else /* WITH_MPI */
239
      call PRECISION_GEMM(BLAS_TRANS_OR_CONJ,'N',nev,nev,na,ONE,z,na,z,na,ZERO,tmp1,na)
240
#endif /* WITH_MPI */
Pavel Kus's avatar
Pavel Kus committed
241
242
      !TODO for the C interface, not all information is passed (zeros instead)
      !TODO than this part of the test cannot be done
243
      !TODO either we will not have this part of test at all, or it will be improved
Pavel Kus's avatar
Pavel Kus committed
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
      if(nblk > 0) then
        ! First check, whether the elements on diagonal are 1 .. "normality" of the vectors
        err = CONST_REAL_0_0
        do i=1, nev
          if (map_global_array_index_to_local_index(i, i, rowLocal, colLocal, nblk, np_rows, np_cols, my_prow, my_pcol)) then
             err = max(err, abs(tmp1(rowLocal,colLocal) - CONST_REAL_1_0))
           endif
        end do
#ifdef WITH_MPI
        call mpi_allreduce(err, errmax, 1, MPI_REAL_PRECISION, MPI_MAX, MPI_COMM_WORLD, mpierr)
#else /* WITH_MPI */
        errmax = err
#endif /* WITH_MPI */
        if (myid==0) print *,'Maximal error in eigenvector lengths:',errmax
      end if
259

Pavel Kus's avatar
Pavel Kus committed
260
      ! Second, find the maximal error in the whole Z**T * Z matrix (its diference from identity matrix)
261
262
263
      ! Initialize tmp2 to unit matrix
      tmp2 = 0
#ifdef WITH_MPI
Pavel Kus's avatar
Pavel Kus committed
264
      call scal_PRECISION_LASET('A', nev, nev, ZERO, ONE, tmp2, 1, 1, sc_desc)
265
#else /* WITH_MPI */
Pavel Kus's avatar
Pavel Kus committed
266
      call PRECISION_LASET('A',nev,nev,ZERO,ONE,tmp2,na)
267
268
269
270
271
272
273
274
#endif /* WITH_MPI */

      !      ! tmp1 = Z**T * Z - Unit Matrix
      tmp1(:,:) =  tmp1(:,:) - tmp2(:,:)

      ! Get maximum error (max abs value in tmp1)
      err = maxval(abs(tmp1))
#ifdef WITH_MPI
275
      call mpi_allreduce(err, errmax, 1, MPI_REAL_PRECISION, MPI_MAX, MPI_COMM_WORLD, mpierr)
276
277
278
279
280
281
#else /* WITH_MPI */
      errmax = err
#endif /* WITH_MPI */

      if (myid==0) print *,'Error Orthogonality:',errmax
#if REALCASE == 1
282
      if (nev .ge. 2) then
283
#ifdef DOUBLE_PRECISION_REAL
284
        if (errmax .gt. 5e-12_rk8 .or. errmax .eq. 0.0_rk8) then
285
#else
286
        if (errmax .gt. 9e-4_rk4 .or. errmax .eq. 0.0_rk4) then
287
#endif
288
289
290
291
292
293
294
295
296
297
298
          status = 1
        endif
      else
#ifdef DOUBLE_PRECISION_REAL
        if (errmax .gt. 5e-12_rk8) then
#else
        if (errmax .gt. 9e-4_rk4) then
#endif
          status = 1
        endif
      endif
299
300
#endif
#if COMPLEXCASE == 1
301
302
303
304
305
306
307
308
309
      if (nev .ge. 2) then
#ifdef DOUBLE_PRECISION_COMPLEX
        if (errmax .gt. 5e-12_rk8 .or. errmax .eq. 0.0_rk8) then
#else
        if (errmax .gt. 9e-4_rk4 .or. errmax .eq. 0.0_rk4) then
#endif
          status = 1
        endif
      else
310
#ifdef DOUBLE_PRECISION_COMPLEX
311
        if (errmax .gt. 5e-12_rk8) then
312
#else
313
        if (errmax .gt. 9e-4_rk4) then
314
#endif
315
316
317
          status = 1
        endif
      endif
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
#endif

    end function


#if REALCASE == 1

#ifdef DOUBLE_PRECISION_REAL
    !c> int check_correctness_real_double_f(int na, int nev, int na_rows, int na_cols,
    !c>                                         double *as, double *z, double *ev,
    !c>                                         int sc_desc[9], int myid);
#else
    !c> int check_correctness_real_single_f(int na, int nev, int na_rows, int na_cols,
    !c>                                         float *as, float *z, float *ev,
    !c>                                         int sc_desc[9], int myid);
#endif

#endif /* REALCASE */

#if COMPLEXCASE == 1
#ifdef DOUBLE_PRECISION_COMPLEX
    !c> int check_correctness_complex_double_f(int na, int nev, int na_rows, int na_cols,
    !c>                                         complex double *as, complex double *z, double *ev,
    !c>                                         int sc_desc[9], int myid);
#else
    !c> int check_correctness_complex_single_f(int na, int nev, int na_rows, int na_cols,
Andreas Marek's avatar
Andreas Marek committed
344
    !c>                                         complex float *as, complex float *z, float *ev,
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
    !c>                                         int sc_desc[9], int myid);
#endif
#endif /* COMPLEXCASE */

function check_correctness_&
&MATH_DATATYPE&
&_&
&PRECISION&
&_f (na, nev, na_rows, na_cols, as, z, ev, sc_desc, myid) result(status) &
      bind(C,name="check_correctness_&
      &MATH_DATATYPE&
      &_&
      &PRECISION&
      &_f")

      use iso_c_binding

      implicit none
363
#include "../../src/general/precision_kinds.F90"
364
365
366

      integer(kind=c_int)            :: status
      integer(kind=c_int), value     :: na, nev, myid, na_rows, na_cols
367
368
      MATH_DATATYPE(kind=rck)     :: as(1:na_rows,1:na_cols), z(1:na_rows,1:na_cols)
      real(kind=rck)    :: ev(1:na)
369
370
      integer(kind=c_int)            :: sc_desc(1:9)

Pavel Kus's avatar
Pavel Kus committed
371
372
373
      ! TODO: I did not want to add all the variables to the C interface as well
      ! TODO: I think that we should find a better way to pass this information
      ! TODO: to all the functions anyway (get it from sc_desc, pass elpa_t, etc..)
374
375
376
377
      status = check_correctness_&
      &MATH_DATATYPE&
      &_&
      &PRECISION&
Pavel Kus's avatar
Pavel Kus committed
378
      & (na, nev, as, z, ev, sc_desc, 0, myid, 0, 0, 0, 0)
379
380
381

    end function

Andreas Marek's avatar
Andreas Marek committed
382
383
384
385
386
387
388
    function check_correctness_eigenvalues_toeplitz_&
    &MATH_DATATYPE&
    &_&
    &PRECISION&
    & (na, diagonalElement, subdiagonalElement, ev, z, myid) result(status)
      use iso_c_binding
      implicit none
389
#include "../../src/general/precision_kinds.F90"
Andreas Marek's avatar
Andreas Marek committed
390
391
392

      integer               :: status, ii, j, myid
      integer, intent(in)   :: na
393
394
395
      real(kind=rck) :: diagonalElement, subdiagonalElement
      real(kind=rck) :: ev_analytic(na), ev(na)
      MATH_DATATYPE(kind=rck) :: z(:,:)
Andreas Marek's avatar
Andreas Marek committed
396
397

#if defined(DOUBLE_PRECISION_REAL) || defined(DOUBLE_PRECISION_COMPLEX)
398
      real(kind=rck), parameter   :: pi = 3.141592653589793238462643383279_c_double
Andreas Marek's avatar
Andreas Marek committed
399
#else
400
      real(kind=rck), parameter   :: pi = 3.1415926535897932_c_float
Andreas Marek's avatar
Andreas Marek committed
401
#endif
402
      real(kind=rck)              :: tmp, maxerr
Andreas Marek's avatar
Andreas Marek committed
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
      integer                                 :: loctmp
      status = 0

     ! analytic solution
     do ii=1, na
#if defined(DOUBLE_PRECISION_REAL) || defined(DOUBLE_PRECISION_COMPLEX)
       ev_analytic(ii) = diagonalElement + 2.0_c_double * &
                         subdiagonalElement *cos( pi*real(ii,kind=c_double)/ &
			 real(na+1,kind=c_double) )
#else
       ev_analytic(ii) = diagonalElement + 2.0_c_float * &
                         subdiagonalElement *cos( pi*real(ii,kind=c_float)/ &
                         real(na+1,kind=c_float) )
#endif
     enddo

     ! sort analytic solution:

     ! this hack is neither 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 ii=2, na
       tmp = ev_analytic(ii)
       do j= ii, na
         if (ev_analytic(j) .lt. tmp) then
           tmp    = ev_analytic(j)
           loctmp = j
         endif
       enddo
       ev_analytic(loctmp) = ev_analytic(ii)
       ev_analytic(ii) = tmp
     enddo

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

#if defined(DOUBLE_PRECISION_REAL) || defined(DOUBLE_PRECISION_COMPLEX)
     if (maxerr .gt. 8.e-13_c_double) then
#else
     if (maxerr .gt. 8.e-4_c_float) then
#endif
       status = 1
       if (myid .eq. 0) then
         print *,"Eigenvalues differ from analytic solution: maxerr = ",maxerr
       endif
     endif
    end function
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
    function check_correctness_cholesky_&
    &MATH_DATATYPE&
    &_&
    &PRECISION&
    & (na, a, as, na_rows, sc_desc, myid) result(status)
      implicit none
#include "../../src/general/precision_kinds.F90"
      integer(kind=ik)                 :: status
      integer(kind=ik), intent(in)     :: na, myid, na_rows
#if REALCASE == 1
      real(kind=rck), intent(in)       :: a(:,:), as(:,:)
      real(kind=rck), dimension(size(as,dim=1),size(as,dim=2)) :: tmp1, tmp2
      real(kind=rck)                   :: norm, normmax

#ifdef WITH_MPI
#ifdef DOUBLE_PRECISION_REAL
      real(kind=rck)                   :: pdlange
#else
      real(kind=rck)                   :: pslange
#endif

#else /* WITH_MPI */

#ifdef DOUBLE_PRECISION_REAL
      real(kind=rck)                   :: dlange
#else
      real(kind=rck)                   :: slange
#endif

#endif /* WITH_MPI */

#endif /* REALCASE */

#if COMPLEXCASE == 1
      complex(kind=rck), intent(in)    :: a(:,:), as(:,:)
      complex(kind=rck), dimension(size(as,dim=1),size(as,dim=2)) :: tmp1, tmp2
      real(kind=rck)                :: norm, normmax
#ifdef DOUBLE_PRECISION_COMPLEX
      complex(kind=ck8), parameter   :: CZERO = (0.0_rk8,0.0_rk8), CONE = (1.0_rk8,0.0_rk8)
#else
      complex(kind=ck4), parameter   :: CZERO = (0.0_rk4,0.0_rk4), CONE = (1.0_rk4,0.0_rk8)
#endif

#ifdef WITH_MPI
#ifdef DOUBLE_PRECISION_COMPLEX
      complex(kind=rck)                   :: pzlange
#else
      complex(kind=rck)                   :: pclange
#endif

#else /* WITH_MPI */

#ifdef DOUBLE_PRECISION_COMPLEX
      complex(kind=rck)                   :: zlange
#else
      complex(kind=rck)                   :: clange
#endif

#endif /* WITH_MPI */
#endif /* COMPLEXCASE */

      integer(kind=ik)                 :: sc_desc(:)

      real(kind=rck)                   :: err, errmax

      integer :: mpierr

      status = 0

#if REALCASE == 1
      tmp1(:,:) = 0.0_rk8
#endif
#if COMPLEXCASE == 1
      tmp1(:,:) = 0.0_ck8
#endif


#if REALCASE == 1
      ! tmp1 = a**T
#ifdef WITH_MPI
#ifdef DOUBLE_PRECISION_REAL
      call pdtran(na, na, 1.0_rk8, a, 1, 1, sc_desc, 0.0_rk8, tmp1, 1, 1, sc_desc)
#else

      call pstran(na, na, 1.0_rk4, a, 1, 1, sc_desc, 0.0_rk4, tmp1, 1, 1, sc_desc)
#endif

#else /* WITH_MPI */
      tmp1 = transpose(a)
#endif /* WITH_MPI */

      ! tmp2 = a * a**T
#ifdef WITH_MPI
#ifdef DOUBLE_PRECISION_REAL
      call pdgemm("N","N", na, na, na, 1.0_rk8, a, 1, 1, sc_desc, tmp1, 1, 1, &
               sc_desc, 0.0_rk8, tmp2, 1, 1, sc_desc)
#else
      call psgemm("N","N", na, na, na, 1.0_rk4, a, 1, 1, sc_desc, tmp1, 1, 1, &
               sc_desc, 0.0_rk4, tmp2, 1, 1, sc_desc)

#endif

#else /* WITH_MPI */

#ifdef DOUBLE_PRECISION_REAL
      call dgemm("N","N", na, na, na, 1.0_rk8, a, na, tmp1, na, 0.0_rk8, tmp2, na)
#else
      call sgemm("N","N", na, na, na, 1.0_rk4, a, na, tmp1, na, 0.0_rk4, tmp2, na)
#endif
#endif /* WITH_MPI */


#endif /* REALCASE == 1 */

#if COMPLEXCASE == 1
      ! tmp1 = a**H
#ifdef WITH_MPI
#ifdef DOUBLE_PRECISION_COMPLEX
      call pztranc(na, na, CONE, a, 1, 1, sc_desc, CZERO, tmp1, 1, 1, sc_desc)
#else

      call pctranc(na, na, CONE, a, 1, 1, sc_desc, CZERO, tmp1, 1, 1, sc_desc)
#endif

#else /* WITH_MPI */
      tmp1 = transpose(conjg(a))
#endif /* WITH_MPI */

      ! tmp2 = a * a**T
#ifdef WITH_MPI
#ifdef DOUBLE_PRECISION_COMPLEX
      call pzgemm("N","N", na, na, na, CONE, a, 1, 1, sc_desc, tmp1, 1, 1, &
               sc_desc, CZERO, tmp2, 1, 1, sc_desc)
#else
      call pcgemm("N","N", na, na, na, CONE, a, 1, 1, sc_desc, tmp1, 1, 1, &
               sc_desc, CZERO, tmp2, 1, 1, sc_desc)

#endif

#else /* WITH_MPI */

#ifdef DOUBLE_PRECISION_COMPLEX
      call zgemm("N","N", na, na, na, CONE, a, na, tmp1, na, CZERO, tmp2, na)
#else
      call cgemm("N","N", na, na, na, CONE, a, na, tmp1, na, CZERO, tmp2, na)
#endif
#endif /* WITH_MPI */


#endif /* COMPLEXCASE == 1 */

      ! compare tmp2 with original matrix
      tmp2(:,:) = tmp2(:,:) - as(:,:)
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
#if REALCASE == 1

#ifdef WITH_MPI
#ifdef DOUBLE_PRECISION_REAL

      norm = pdlange("M",na, na, tmp2, 1, 1, sc_desc, tmp1)
#else
      norm = pslange("M",na, na, tmp2, 1, 1, sc_desc, tmp1)
#endif

#else /* WITH_MPI */
#ifdef DOUBLE_PRECISION_REAL
      norm = dlange("M", na, na, tmp2, na_rows, tmp1)
#else
      norm = slange("M", na, na, tmp2, na_rows, tmp1)

#endif
#endif /* WITH_MPI */


#ifdef WITH_MPI
#ifdef DOUBLE_PRECISION_REAL
      call mpi_allreduce(norm,normmax,1,MPI_REAL8,MPI_MAX,MPI_COMM_WORLD,mpierr)
#else
      call mpi_allreduce(norm,normmax,1,MPI_REAL4,MPI_MAX,MPI_COMM_WORLD,mpierr)
#endif
#else /* WITH_MPI */
      normmax = norm
#endif /* WITH_MPI */


#endif /* REALCASE == 1 */

#if COMPLEXCASE == 1

#ifdef WITH_MPI
#ifdef DOUBLE_PRECISION_COMPLEX

      norm = pzlange("M",na, na, tmp2, 1, 1, sc_desc, tmp1)
#else
      norm = pclange("M",na, na, tmp2, 1, 1, sc_desc, tmp1)
#endif

#else /* WITH_MPI */
#ifdef DOUBLE_PRECISION_COMPLEX
      norm = zlange("M", na, na, tmp2, na_rows, tmp1)
#else
      norm = clange("M", na, na, tmp2, na_rows, tmp1)

#endif
#endif /* WITH_MPI */


#ifdef WITH_MPI
#ifdef DOUBLE_PRECISION_COMPLEX
      call mpi_allreduce(norm,normmax,1,MPI_REAL8,MPI_MAX,MPI_COMM_WORLD,mpierr)
#else
      call mpi_allreduce(norm,normmax,1,MPI_REAL4,MPI_MAX,MPI_COMM_WORLD,mpierr)
#endif
#else /* WITH_MPI */
      normmax = norm
#endif /* WITH_MPI */


#endif /* REALCASE == 1 */

      if (myid .eq. 0) then
        print *," Maximum error of result: ", normmax
      endif

#ifdef DOUBLE_PRECISION_REAL
      if (normmax .gt. 5e-12_rk8) then
        status = 1
      endif
#else
      if (normmax .gt. 5e-4_rk4) then
        status = 1
      endif
#endif

    end function

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
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
    function check_correctness_hermitian_multiply_&
    &MATH_DATATYPE&
    &_&
    &PRECISION&
    & (na, a, b, c, na_rows, sc_desc, myid) result(status)
      implicit none
#include "../../src/general/precision_kinds.F90"
      integer(kind=ik)                 :: status
      integer(kind=ik), intent(in)     :: na, myid, na_rows
#if REALCASE == 1
      real(kind=rck), intent(in)       :: a(:,:), b(:,:), c(:,:)
      real(kind=rck), dimension(size(a,dim=1),size(a,dim=2)) :: tmp1, tmp2
      real(kind=rck)                   :: norm, normmax

#ifdef WITH_MPI
#ifdef DOUBLE_PRECISION_REAL
      real(kind=rck)                   :: pdlange
#else
      real(kind=rck)                   :: pslange
#endif

#else /* WITH_MPI */

#ifdef DOUBLE_PRECISION_REAL
      real(kind=rck)                   :: dlange
#else
      real(kind=rck)                   :: slange
#endif

#endif /* WITH_MPI */

#endif /* REALCASE */

#if COMPLEXCASE == 1
      complex(kind=rck), intent(in)    :: a(:,:), b(:,:), c(:,:)
      complex(kind=rck), dimension(size(a,dim=1),size(a,dim=2)) :: tmp1, tmp2
      real(kind=rck)                :: norm, normmax
#ifdef DOUBLE_PRECISION_COMPLEX
      complex(kind=ck8), parameter   :: CZERO = (0.0_rk8,0.0_rk8), CONE = (1.0_rk8,0.0_rk8)
#else
      complex(kind=ck4), parameter   :: CZERO = (0.0_rk4,0.0_rk4), CONE = (1.0_rk4,0.0_rk8)
#endif

#ifdef WITH_MPI
#ifdef DOUBLE_PRECISION_COMPLEX
      complex(kind=rck)                   :: pzlange
#else
      complex(kind=rck)                   :: pclange
#endif

#else /* WITH_MPI */

#ifdef DOUBLE_PRECISION_COMPLEX
      complex(kind=rck)                   :: zlange
#else
      complex(kind=rck)                   :: clange
#endif

#endif /* WITH_MPI */
#endif /* COMPLEXCASE */

      integer(kind=ik)                 :: sc_desc(:)

      real(kind=rck)                   :: err, errmax

      integer :: mpierr

      status = 0

#if REALCASE == 1

#ifdef DOUBLE_PRECISION_REAL
      tmp1(:,:) = 0.0_rk8
#else
      tmp1(:,:) = 0.0_rk4
#endif

#endif /* REALCASE */
#if COMPLEXCASE == 1
#ifdef DOUBLE_PRECISION_COMPLEX
      tmp1(:,:) = 0.0_ck8
#else
      tmp1(:,:) = 0.0_ck4
#endif
#endif /* COMPLEXCASE */


#if REALCASE == 1
      ! tmp1 = a**T
#ifdef WITH_MPI
#ifdef DOUBLE_PRECISION_REAL
      call pdtran(na, na, 1.0_rk8, a, 1, 1, sc_desc, 0.0_rk8, tmp1, 1, 1, sc_desc)
#else

      call pstran(na, na, 1.0_rk4, a, 1, 1, sc_desc, 0.0_rk4, tmp1, 1, 1, sc_desc)
#endif

#else /* WITH_MPI */
      tmp1 = transpose(a)
#endif /* WITH_MPI */

   ! tmp2 = tmp1 * b
#ifdef DOUBLE_PRECISION_REAL
#ifdef WITH_MPI
   call pdgemm("N","N", na, na, na, 1.0_rk8, tmp1, 1, 1, sc_desc, b, 1, 1, &
               sc_desc, 0.0_rk8, tmp2, 1, 1, sc_desc)
#else
   call dgemm("N","N", na, na, na, 1.0_rk8, tmp1, na, b, na, 0.0_rk8, tmp2, na)
#endif

#else /* DOUBLE_PRECISION_REAL */
#ifdef WITH_MPI
   call psgemm("N","N", na, na, na, 1.0_rk4, tmp1, 1, 1, sc_desc, b, 1, 1, &
               sc_desc, 0.0_rk4, tmp2, 1, 1, sc_desc)
#else
   call sgemm("N","N", na, na, na, 1.0_rk4, tmp1, na, b, na, 0.0_rk4, tmp2, na)
#endif

#endif /* DOUBLE_PRECISION_REAL */

#endif /* REALCASE == 1 */

#if COMPLEXCASE == 1
      ! tmp1 = a**H
#ifdef WITH_MPI
#ifdef DOUBLE_PRECISION_COMPLEX
      call pztranc(na, na, CONE, a, 1, 1, sc_desc, CZERO, tmp1, 1, 1, sc_desc)
#else

      call pctranc(na, na, CONE, a, 1, 1, sc_desc, CZERO, tmp1, 1, 1, sc_desc)
#endif

#else /* WITH_MPI */
      tmp1 = transpose(conjg(a))
#endif /* WITH_MPI */

   ! tmp2 = tmp1 * b
#ifdef DOUBLE_PRECISION_COMPLEX
#ifdef WITH_MPI
   call pzgemm("N","N", na, na, na, CONE, tmp1, 1, 1, sc_desc, b, 1, 1, &
               sc_desc, CZERO, tmp2, 1, 1, sc_desc)
#else
   call zgemm("N","N", na, na, na, CONE, tmp1, na, b, na, CZERO, tmp2, na)
#endif

#else /* DOUBLE_PRECISION_COMPLEX */

#ifdef WITH_MPI
   call pcgemm("N","N", na, na, na, CONE, tmp1, 1, 1, sc_desc, b, 1, 1, &
               sc_desc, CZERO, tmp2, 1, 1, sc_desc)
#else
   call cgemm("N","N", na, na, na, CONE, tmp1, na, b, na, CZERO, tmp2, na)
#endif
#endif /* DOUBLE_PRECISION_COMPLEX */

#endif /* COMPLEXCASE == 1 */

      ! compare tmp2 with c
      tmp2(:,:) = tmp2(:,:) - c(:,:)
#if REALCASE == 1

#ifdef WITH_MPI
#ifdef DOUBLE_PRECISION_REAL

      norm = pdlange("M",na, na, tmp2, 1, 1, sc_desc, tmp1)
#else
      norm = pslange("M",na, na, tmp2, 1, 1, sc_desc, tmp1)
#endif

#else /* WITH_MPI */
#ifdef DOUBLE_PRECISION_REAL
      norm = dlange("M", na, na, tmp2, na_rows, tmp1)
#else
      norm = slange("M", na, na, tmp2, na_rows, tmp1)

#endif
#endif /* WITH_MPI */


#ifdef WITH_MPI
#ifdef DOUBLE_PRECISION_REAL
      call mpi_allreduce(norm,normmax,1,MPI_REAL8,MPI_MAX,MPI_COMM_WORLD,mpierr)
#else
      call mpi_allreduce(norm,normmax,1,MPI_REAL4,MPI_MAX,MPI_COMM_WORLD,mpierr)
#endif
#else /* WITH_MPI */
      normmax = norm
#endif /* WITH_MPI */


#endif /* REALCASE == 1 */

#if COMPLEXCASE == 1

#ifdef WITH_MPI
#ifdef DOUBLE_PRECISION_COMPLEX

      norm = pzlange("M",na, na, tmp2, 1, 1, sc_desc, tmp1)
#else
      norm = pclange("M",na, na, tmp2, 1, 1, sc_desc, tmp1)
#endif

#else /* WITH_MPI */
#ifdef DOUBLE_PRECISION_COMPLEX
      norm = zlange("M", na, na, tmp2, na_rows, tmp1)
#else
      norm = clange("M", na, na, tmp2, na_rows, tmp1)

#endif
#endif /* WITH_MPI */


#ifdef WITH_MPI
#ifdef DOUBLE_PRECISION_COMPLEX
      call mpi_allreduce(norm,normmax,1,MPI_REAL8,MPI_MAX,MPI_COMM_WORLD,mpierr)
#else
      call mpi_allreduce(norm,normmax,1,MPI_REAL4,MPI_MAX,MPI_COMM_WORLD,mpierr)
#endif
#else /* WITH_MPI */
      normmax = norm
#endif /* WITH_MPI */


#endif /* REALCASE == 1 */

      if (myid .eq. 0) then
        print *," Maximum error of result: ", normmax
      endif

#ifdef DOUBLE_PRECISION_REAL
      if (normmax .gt. 5e-12_rk8) then
        status = 1
      endif
#else
      if (normmax .gt. 5e-4_rk4) then
        status = 1
      endif
#endif

    end function


935
936


937
! vim: syntax=fortran