elpa_c_interface.F90 96.5 KB
Newer Older
Andreas Marek's avatar
Andreas Marek committed
1
2
3
4
5
!    This file is part of ELPA.
!
!    The ELPA library was originally created by the ELPA consortium,
!    consisting of the following organizations:
!
6
7
!    - Max Planck Computing and Data Facility (MPCDF), formerly known as
!      Rechenzentrum Garching der Max-Planck-Gesellschaft (RZG),
Andreas Marek's avatar
Andreas Marek committed
8
9
10
11
12
!    - Bergische Universität Wuppertal, Lehrstuhl für angewandte
!      Informatik,
!    - Technische Universität München, Lehrstuhl für Informatik mit
!      Schwerpunkt Wissenschaftliches Rechnen ,
!    - Fritz-Haber-Institut, Berlin, Abt. Theorie,
13
!    - Max-Plack-Institut für Mathematik in den Naturwissenschaften,
Andreas Marek's avatar
Andreas Marek committed
14
15
16
17
18
19
!      Leipzig, Abt. Komplexe Strukutren in Biologie und Kognition,
!      and
!    - IBM Deutschland GmbH
!
!
!    More information can be found here:
20
!    http://elpa.mpcdf.mpg.de/
Andreas Marek's avatar
Andreas Marek committed
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
!
!    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.
!
42
! Author: Andreas Marek, MCPDF
Andreas Marek's avatar
Andreas Marek committed
43
#include "config-f90.h"
Andreas Marek's avatar
Andreas Marek committed
44
  !c> #include <complex.h>
Andreas Marek's avatar
Andreas Marek committed
45

46
  !c> /*! \brief C old, deprecated interface, will be deleted. Use "elpa_get_communicators"
47
48
49
50
51
52
53
  !c> *
  !c> * \param mpi_comm_word    MPI global communicator (in)
  !c> * \param my_prow          Row coordinate of the calling process in the process grid (in)
  !c> * \param my_pcol          Column coordinate of the calling process in the process grid (in)
  !c> * \param mpi_comm_rows    Communicator for communicating within rows of processes (out)
  !c> * \result int             integer error value of mpi_comm_split function
  !c> */
54
  !c> int get_elpa_row_col_comms(int mpi_comm_world, int my_prow, int my_pcol, int *mpi_comm_rows, int *mpi_comm_cols);
55
  function get_elpa_row_col_comms_wrapper_c_name1(mpi_comm_world, my_prow, my_pcol, &
Andreas Marek's avatar
Andreas Marek committed
56
                                          mpi_comm_rows, mpi_comm_cols)     &
57
                                          result(mpierr) bind(C,name="get_elpa_row_col_comms")
Andreas Marek's avatar
Andreas Marek committed
58
59
60
    use, intrinsic :: iso_c_binding
    use elpa1, only : get_elpa_row_col_comms

Andreas Marek's avatar
Andreas Marek committed
61
    implicit none
Andreas Marek's avatar
Andreas Marek committed
62
63
64
65
66
67
68
69
    integer(kind=c_int)         :: mpierr
    integer(kind=c_int), value  :: mpi_comm_world, my_prow, my_pcol
    integer(kind=c_int)         :: mpi_comm_rows, mpi_comm_cols

    mpierr = get_elpa_row_col_comms(mpi_comm_world, my_prow, my_pcol, &
                                    mpi_comm_rows, mpi_comm_cols)

  end function
70
71
  !c> #include <complex.h>

72
  !c> /*! \brief C old, deprecated interface, will be deleted. Use "elpa_get_communicators"
73
74
75
76
77
78
79
80
81
82
83
84
  !c> *
  !c> * \param mpi_comm_word    MPI global communicator (in)
  !c> * \param my_prow          Row coordinate of the calling process in the process grid (in)
  !c> * \param my_pcol          Column coordinate of the calling process in the process grid (in)
  !c> * \param mpi_comm_rows    Communicator for communicating within rows of processes (out)
  !c> * \result int             integer error value of mpi_comm_split function
  !c> */
  !c> int get_elpa_communicators(int mpi_comm_world, int my_prow, int my_pcol, int *mpi_comm_rows, int *mpi_comm_cols);
  function get_elpa_row_col_comms_wrapper_c_name2(mpi_comm_world, my_prow, my_pcol, &
                                          mpi_comm_rows, mpi_comm_cols)     &
                                          result(mpierr) bind(C,name="get_elpa_communicators")
    use, intrinsic :: iso_c_binding
85
    use elpa1, only : get_elpa_communicators
86
87
88
89
90
91

    implicit none
    integer(kind=c_int)         :: mpierr
    integer(kind=c_int), value  :: mpi_comm_world, my_prow, my_pcol
    integer(kind=c_int)         :: mpi_comm_rows, mpi_comm_cols

92
    mpierr = get_elpa_communicators(mpi_comm_world, my_prow, my_pcol, &
93
94
95
96
                                    mpi_comm_rows, mpi_comm_cols)

  end function

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
  !c> #include <complex.h>

  !c> /*! \brief C interface to create ELPA communicators
  !c> *
  !c> * \param mpi_comm_word    MPI global communicator (in)
  !c> * \param my_prow          Row coordinate of the calling process in the process grid (in)
  !c> * \param my_pcol          Column coordinate of the calling process in the process grid (in)
  !c> * \param mpi_comm_rows    Communicator for communicating within rows of processes (out)
  !c> * \result int             integer error value of mpi_comm_split function
  !c> */
  !c> int elpa_get_communicators(int mpi_comm_world, int my_prow, int my_pcol, int *mpi_comm_rows, int *mpi_comm_cols);
  function elpa_get_communicators_wrapper_c(mpi_comm_world, my_prow, my_pcol, &
                                          mpi_comm_rows, mpi_comm_cols)     &
                                          result(mpierr) bind(C,name="elpa_get_communicators")
    use, intrinsic :: iso_c_binding
    use elpa1, only : elpa_get_communicators

    implicit none
    integer(kind=c_int)         :: mpierr
    integer(kind=c_int), value  :: mpi_comm_world, my_prow, my_pcol
    integer(kind=c_int)         :: mpi_comm_rows, mpi_comm_cols

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

  end function
123
124


125
  !c>  /*! \brief C interface to solve the double-precision real eigenvalue problem with 1-stage solver
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
  !c>  *
  !c> *  \param  na                   Order of matrix a
  !c> *  \param  nev                  Number of eigenvalues needed.
  !c> *                               The smallest nev eigenvalues/eigenvectors are calculated.
  !c> *  \param  a                    Distributed matrix for which eigenvalues are to be computed.
  !c> *                               Distribution is like in Scalapack.
  !c> *                               The full matrix must be set (not only one half like in scalapack).
  !c> *  \param lda                   Leading dimension of a
  !c> *  \param ev(na)                On output: eigenvalues of a, every processor gets the complete set
  !c> *  \param q                     On output: Eigenvectors of a
  !c> *                               Distribution is like in Scalapack.
  !c> *                               Must be always dimensioned to the full size (corresponding to (na,na))
  !c> *                               even if only a part of the eigenvalues is needed.
  !c> *  \param ldq                   Leading dimension of q
  !c> *  \param nblk                  blocksize of cyclic distribution, must be the same in both directions!
  !c> *  \param matrixCols           distributed number of matrix columns
  !c> *  \param mpi_comm_rows        MPI-Communicator for rows
  !c> *  \param mpi_comm_cols        MPI-Communicator for columns
  !c> *
  !c> *  \result                     int: 1 if error occured, otherwise 0
  !c>*/
147
#define DOUBLE_PRECISION_REAL 1
148
149
150
151
152
#ifdef DOUBLE_PRECISION_REAL
  !c> int elpa_solve_evp_real_1stage_double_precision(int na, int nev, double *a, int lda, double *ev, double *q, int ldq, int nblk, int matrixCols, int mpi_comm_rows, int mpi_comm_cols);
#else
  !c> int elpa_solve_evp_real_1stage_single_precision(int na, int nev, float *a, int lda, float *ev, float *q, int ldq, int nblk, int matrixCols, int mpi_comm_rows, int mpi_comm_cols);
#endif
153

154
#ifdef DOUBLE_PRECISION_REAL
155
156
  function solve_elpa1_evp_real_wrapper_double(na, nev, a, lda, ev, q, ldq, nblk, &
                                  matrixCols, mpi_comm_rows, mpi_comm_cols)      &
157
158
                                  result(success) bind(C,name="elpa_solve_evp_real_1stage_double_precision")
#else
159
160
  function solve_elpa1_evp_real_wrapper_single(na, nev, a, lda, ev, q, ldq, nblk, &
                                  matrixCols, mpi_comm_rows, mpi_comm_cols)      &
161
162
                                  result(success) bind(C,name="elpa_solve_evp_real_1stage_single_precision")
#endif
163

Andreas Marek's avatar
Andreas Marek committed
164
    use, intrinsic :: iso_c_binding
165
    use elpa1
Andreas Marek's avatar
Andreas Marek committed
166

Andreas Marek's avatar
Andreas Marek committed
167
    implicit none
Andreas Marek's avatar
Andreas Marek committed
168
    integer(kind=c_int)                    :: success
169
    integer(kind=c_int), value, intent(in) :: na, nev, lda, ldq, nblk, matrixCols, mpi_comm_cols, mpi_comm_rows
170
#ifdef DOUBLE_PRECISION_REAL
171
    real(kind=c_double)                    :: ev(1:na)
172
#ifdef USE_ASSUMED_SIZE
173
174
175
    real(kind=c_double)                    :: a(lda,*), q(ldq,*)
#else
    real(kind=c_double)                    :: a(1:lda,1:matrixCols), q(1:ldq,1:matrixCols)
176
177
178
179
180
#endif

#else /* SINGLE_PRECISION */
    real(kind=c_float)                     :: ev(1:na)

181
#ifdef USE_ASSUMED_SIZE
182
    real(kind=c_float)                     :: a(lda,*), q(ldq,*)
183
184
#else
    real(kind=c_float)                     :: a(1:lda,1:matrixCols), ev(1:na), q(1:ldq,1:matrixCols)
185
186
#endif

187
#endif
Andreas Marek's avatar
Andreas Marek committed
188
189
    logical                                :: successFortran

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
#ifdef DOUBLE_PRECISION_REAL
    successFortran = solve_evp_real_1stage_double(na, nev, a, lda, ev, q, ldq, nblk, matrixCols, mpi_comm_rows, mpi_comm_cols)
#else
    successFortran = solve_evp_real_1stage_single(na, nev, a, lda, ev, q, ldq, nblk, matrixCols, mpi_comm_rows, mpi_comm_cols)
#endif
    if (successFortran) then
      success = 1
    else
      success = 0
    endif

  end function

#ifdef WANT_SINGLE_PRECISION_REAL
#undef DOUBLE_PRECISION_REAL
  !c>  /*! \brief C interface to solve the single-precision real eigenvalue problem with 1-stage solver
  !c>  *
  !c> *  \param  na                   Order of matrix a
  !c> *  \param  nev                  Number of eigenvalues needed.
  !c> *                               The smallest nev eigenvalues/eigenvectors are calculated.
  !c> *  \param  a                    Distributed matrix for which eigenvalues are to be computed.
  !c> *                               Distribution is like in Scalapack.
  !c> *                               The full matrix must be set (not only one half like in scalapack).
  !c> *  \param lda                   Leading dimension of a
  !c> *  \param ev(na)                On output: eigenvalues of a, every processor gets the complete set
  !c> *  \param q                     On output: Eigenvectors of a
  !c> *                               Distribution is like in Scalapack.
  !c> *                               Must be always dimensioned to the full size (corresponding to (na,na))
  !c> *                               even if only a part of the eigenvalues is needed.
  !c> *  \param ldq                   Leading dimension of q
  !c> *  \param nblk                  blocksize of cyclic distribution, must be the same in both directions!
  !c> *  \param matrixCols           distributed number of matrix columns
  !c> *  \param mpi_comm_rows        MPI-Communicator for rows
  !c> *  \param mpi_comm_cols        MPI-Communicator for columns
  !c> *
  !c> *  \result                     int: 1 if error occured, otherwise 0
  !c>*/
#ifdef DOUBLE_PRECISION_REAL
  !c> int elpa_solve_evp_real_1stage_double_precision(int na, int nev, double *a, int lda, double *ev, double *q, int ldq, int nblk, int matrixCols, int mpi_comm_rows, int mpi_comm_cols);
#else
  !c> int elpa_solve_evp_real_1stage_single_precision(int na, int nev, float *a, int lda, float *ev, float *q, int ldq, int nblk, int matrixCols, int mpi_comm_rows, int mpi_comm_cols);
#endif

#ifdef DOUBLE_PRECISION_REAL
  function solve_elpa1_evp_real_wrapper_double(na, nev, a, lda, ev, q, ldq, nblk, &
                                  matrixCols, mpi_comm_rows, mpi_comm_cols)      &
                                  result(success) bind(C,name="elpa_solve_evp_real_1stage_double_precision")
#else
  function solve_elpa1_evp_real_wrapper_single(na, nev, a, lda, ev, q, ldq, nblk, &
                                  matrixCols, mpi_comm_rows, mpi_comm_cols)      &
                                  result(success) bind(C,name="elpa_solve_evp_real_1stage_single_precision")
#endif
    use, intrinsic :: iso_c_binding
    use elpa1

    implicit none
    integer(kind=c_int)                    :: success
    integer(kind=c_int), value, intent(in) :: na, nev, lda, ldq, nblk, matrixCols, mpi_comm_cols, mpi_comm_rows
#ifdef DOUBLE_PRECISION_REAL
    real(kind=c_double)                    :: a(1:lda,1:matrixCols), ev(1:na), q(1:ldq,1:matrixCols)
#else
    real(kind=c_float)                     :: a(1:lda,1:matrixCols), ev(1:na), q(1:ldq,1:matrixCols)
#endif
    logical                                :: successFortran
Andreas Marek's avatar
Andreas Marek committed
254

255
256
257
258
259
#ifdef DOUBLE_PRECISION_REAL
    successFortran = solve_evp_real_1stage_double(na, nev, a, lda, ev, q, ldq, nblk, matrixCols, mpi_comm_rows, mpi_comm_cols)
#else
    successFortran = solve_evp_real_1stage_single(na, nev, a, lda, ev, q, ldq, nblk, matrixCols, mpi_comm_rows, mpi_comm_cols)
#endif
Andreas Marek's avatar
Andreas Marek committed
260
261
262
263
264
265
266
    if (successFortran) then
      success = 1
    else
      success = 0
    endif

  end function
267

268
269
#endif /* WANT_SINGLE_PRECISION_REAL */

270

271
272

  !c> /*! \brief C interface to solve the double-precision complex eigenvalue problem with 1-stage solver
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
  !c> *
  !c> *  \param  na                   Order of matrix a
  !c> *  \param  nev                  Number of eigenvalues needed.
  !c> *                               The smallest nev eigenvalues/eigenvectors are calculated.
  !c> *  \param  a                    Distributed matrix for which eigenvalues are to be computed.
  !c> *                               Distribution is like in Scalapack.
  !c> *                               The full matrix must be set (not only one half like in scalapack).
  !c> *  \param lda                   Leading dimension of a
  !c> *  \param ev(na)                On output: eigenvalues of a, every processor gets the complete set
  !c> *  \param q                     On output: Eigenvectors of a
  !c> *                               Distribution is like in Scalapack.
  !c> *                               Must be always dimensioned to the full size (corresponding to (na,na))
  !c> *                               even if only a part of the eigenvalues is needed.
  !c> *  \param ldq                   Leading dimension of q
  !c> *  \param nblk                  blocksize of cyclic distribution, must be the same in both directions!
  !c> *  \param matrixCols           distributed number of matrix columns
  !c> *  \param mpi_comm_rows        MPI-Communicator for rows
  !c> *  \param mpi_comm_cols        MPI-Communicator for columns
  !c> *
  !c> *  \result                     int: 1 if error occured, otherwise 0
  !c> */
294
#define DOUBLE_PRECISION_COMPLEX 1
295
296
297
298
299
300
301
#ifdef DOUBLE_PRECISION_COMPLEX
  !c> int elpa_solve_evp_complex_1stage_double_precision(int na, int nev, double complex *a, int lda, double *ev, double complex *q, int ldq, int nblk, int matrixCols, int mpi_comm_rows, int mpi_comm_cols);
#else
  !c> int elpa_solve_evp_complex_1stage_single_precision(int na, int nev,  complex *a, int lda, float *ev, complex *q, int ldq, int nblk, int matrixCols, int mpi_comm_rows, int mpi_comm_cols);
#endif

#ifdef DOUBLE_PRECISION_COMPLEX
302
303
  function solve_evp_real_wrapper_double(na, nev, a, lda, ev, q, ldq, nblk, &
                                  matrixCols, mpi_comm_rows, mpi_comm_cols)      &
304
305
                                  result(success) bind(C,name="elpa_solve_evp_complex_1stage_double_precision")
#else
306
307
  function solve_evp_real_wrapper_single(na, nev, a, lda, ev, q, ldq, nblk, &
                                  matrixCols, mpi_comm_rows, mpi_comm_cols)      &
308
309
                                  result(success) bind(C,name="elpa_solve_evp_complex_1stage_single_precision")
#endif
Andreas Marek's avatar
Andreas Marek committed
310
    use, intrinsic :: iso_c_binding
311
    use elpa1
Andreas Marek's avatar
Andreas Marek committed
312

Andreas Marek's avatar
Andreas Marek committed
313
    implicit none
Andreas Marek's avatar
Andreas Marek committed
314
    integer(kind=c_int)                    :: success
315
    integer(kind=c_int), value, intent(in) :: na, nev, lda, ldq, nblk, matrixCols, mpi_comm_cols, mpi_comm_rows
316
#ifdef DOUBLE_PRECISION_COMPLEX
Andreas Marek's avatar
Andreas Marek committed
317
    real(kind=c_double)                    :: ev(1:na)
318
#ifdef USE_ASSUMED_SIZE
319
    complex(kind=c_double_complex)         :: a(lda,*), q(ldq,*)
320
#else
321
    complex(kind=c_double_complex)         :: a(1:lda,1:matrixCols), q(1:ldq,1:matrixCols)
322
#endif
323
324

#else /* SINGLE_PRECISION */
325
    real(kind=c_float)                     :: ev(1:na)
326
#ifdef USE_ASSUMED_SIZE
327
328
329
330
331
    complex(kind=c_float_complex)          :: a(lda,*), q(ldq,*)
#else
    complex(kind=c_float_complex)          :: a(1:lda,1:matrixCols), q(1:ldq,1:matrixCols)
#endif

332
#endif
Andreas Marek's avatar
Andreas Marek committed
333
334
335

    logical                                :: successFortran

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
#ifdef DOUBLE_PRECISION_COMPLEX
    successFortran = solve_evp_complex_1stage_double(na, nev, a, lda, ev, q, ldq, nblk, matrixCols, mpi_comm_rows, mpi_comm_cols)
#else
    successFortran = solve_evp_complex_1stage_single(na, nev, a, lda, ev, q, ldq, nblk, matrixCols, mpi_comm_rows, mpi_comm_cols)
#endif
    if (successFortran) then
      success = 1
    else
      success = 0
    endif

  end function

#ifdef WANT_SINGLE_PRECISION_COMPLEX

  !c> /*! \brief C interface to solve the single-precision complex eigenvalue problem with 1-stage solver
  !c> *
  !c> *  \param  na                   Order of matrix a
  !c> *  \param  nev                  Number of eigenvalues needed.
  !c> *                               The smallest nev eigenvalues/eigenvectors are calculated.
  !c> *  \param  a                    Distributed matrix for which eigenvalues are to be computed.
  !c> *                               Distribution is like in Scalapack.
  !c> *                               The full matrix must be set (not only one half like in scalapack).
  !c> *  \param lda                   Leading dimension of a
  !c> *  \param ev(na)                On output: eigenvalues of a, every processor gets the complete set
  !c> *  \param q                     On output: Eigenvectors of a
  !c> *                               Distribution is like in Scalapack.
  !c> *                               Must be always dimensioned to the full size (corresponding to (na,na))
  !c> *                               even if only a part of the eigenvalues is needed.
  !c> *  \param ldq                   Leading dimension of q
  !c> *  \param nblk                  blocksize of cyclic distribution, must be the same in both directions!
  !c> *  \param matrixCols           distributed number of matrix columns
  !c> *  \param mpi_comm_rows        MPI-Communicator for rows
  !c> *  \param mpi_comm_cols        MPI-Communicator for columns
  !c> *
  !c> *  \result                     int: 1 if error occured, otherwise 0
  !c> */
#undef DOUBLE_PRECISION_COMPLEX
#ifdef DOUBLE_PRECISION_COMPLEX
  !c> int elpa_solve_evp_complex_1stage_double_precision(int na, int nev, double complex *a, int lda, double *ev, double complex *q, int ldq, int nblk, int matrixCols, int mpi_comm_rows, int mpi_comm_cols);
#else
  !c> int elpa_solve_evp_complex_1stage_single_precision(int na, int nev,  complex *a, int lda, float *ev, complex *q, int ldq, int nblk, int matrixCols, int mpi_comm_rows, int mpi_comm_cols);
#endif

#ifdef DOUBLE_PRECISION_COMPLEX
  function solve_evp_real_wrapper_double(na, nev, a, lda, ev, q, ldq, nblk, &
                                  matrixCols, mpi_comm_rows, mpi_comm_cols)      &
                                  result(success) bind(C,name="elpa_solve_evp_complex_1stage_double_precision")
#else
  function solve_evp_real_wrapper_single(na, nev, a, lda, ev, q, ldq, nblk, &
                                  matrixCols, mpi_comm_rows, mpi_comm_cols)      &
                                  result(success) bind(C,name="elpa_solve_evp_complex_1stage_single_precision")
#endif
    use, intrinsic :: iso_c_binding
    use elpa1

    implicit none
    integer(kind=c_int)                    :: success
    integer(kind=c_int), value, intent(in) :: na, nev, lda, ldq, nblk, matrixCols, mpi_comm_cols, mpi_comm_rows
#ifdef DOUBLE_PRECISION_COMPLEX
    complex(kind=c_double_complex)         :: a(1:lda,1:matrixCols), q(1:ldq,1:matrixCols)
    real(kind=c_double)                    :: ev(1:na)
#else
    complex(kind=c_float_complex)          :: a(1:lda,1:matrixCols), q(1:ldq,1:matrixCols)
    real(kind=c_float)                     :: ev(1:na)
#endif

    logical                                :: successFortran
Andreas Marek's avatar
Andreas Marek committed
404

405
406
407
408
409
#ifdef DOUBLE_PRECISION_COMPLEX
    successFortran = solve_evp_complex_1stage_double(na, nev, a, lda, ev, q, ldq, nblk, matrixCols, mpi_comm_rows, mpi_comm_cols)
#else
    successFortran = solve_evp_complex_1stage_single(na, nev, a, lda, ev, q, ldq, nblk, matrixCols, mpi_comm_rows, mpi_comm_cols)
#endif
Andreas Marek's avatar
Andreas Marek committed
410
411
412
413
414
415
416
    if (successFortran) then
      success = 1
    else
      success = 0
    endif

  end function
417
418
419
420
421

#endif /* WANT_SINGLE_PRECISION_COMPLEX */


  !c> /*! \brief C interface to solve the double-precision real eigenvalue problem with 2-stage solver
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
  !c> *
  !c> *  \param  na                        Order of matrix a
  !c> *  \param  nev                       Number of eigenvalues needed.
  !c> *                                    The smallest nev eigenvalues/eigenvectors are calculated.
  !c> *  \param  a                         Distributed matrix for which eigenvalues are to be computed.
  !c> *                                    Distribution is like in Scalapack.
  !c> *                                    The full matrix must be set (not only one half like in scalapack).
  !c> *  \param lda                        Leading dimension of a
  !c> *  \param ev(na)                     On output: eigenvalues of a, every processor gets the complete set
  !c> *  \param q                          On output: Eigenvectors of a
  !c> *                                    Distribution is like in Scalapack.
  !c> *                                    Must be always dimensioned to the full size (corresponding to (na,na))
  !c> *                                    even if only a part of the eigenvalues is needed.
  !c> *  \param ldq                        Leading dimension of q
  !c> *  \param nblk                       blocksize of cyclic distribution, must be the same in both directions!
  !c> *  \param matrixCols                 distributed number of matrix columns
  !c> *  \param mpi_comm_rows              MPI-Communicator for rows
  !c> *  \param mpi_comm_cols              MPI-Communicator for columns
  !c> *  \param mpi_coll_all               MPI communicator for the total processor set
  !c> *  \param THIS_REAL_ELPA_KERNEL_API  specify used ELPA2 kernel via API
  !c> *  \param use_qr                     use QR decomposition 1 = yes, 0 = no
  !c> *
  !c> *  \result                     int: 1 if error occured, otherwise 0
  !c> */
446
#define DOUBLE_PRECISION_REAL 1
447
448
449
450
451
452
#ifdef DOUBLE_PRECISION_REAL
  !c> int elpa_solve_evp_real_2stage_double_precision(int na, int nev, double *a, int lda, double *ev, double *q, int ldq, int nblk, int matrixCols, int mpi_comm_rows, int mpi_comm_cols, int mpi_comm_all, int THIS_REAL_ELPA_KERNEL_API, int useQR);
#else
  !c> int elpa_solve_evp_real_2stage_single_precision(int na, int nev, float *a, int lda, float *ev, float *q, int ldq, int nblk, int matrixCols, int mpi_comm_rows, int mpi_comm_cols, int mpi_comm_all, int THIS_REAL_ELPA_KERNEL_API, int useQR);
#endif

453
454
#ifdef DOUBLE_PRECISION_REAL
  function solve_elpa2_evp_real_wrapper_double(na, nev, a, lda, ev, q, ldq, nblk,    &
455
                                  matrixCols, mpi_comm_rows, mpi_comm_cols, mpi_comm_all, &
Andreas Marek's avatar
Andreas Marek committed
456
                                  THIS_REAL_ELPA_KERNEL_API, useQR)           &
457
458
                                  result(success) bind(C,name="elpa_solve_evp_real_2stage_double_precision")
#else
459
460
461
462
463
  function solve_elpa2_evp_real_wrapper_single(na, nev, a, lda, ev, q, ldq, nblk,    &
                                  matrixCols, mpi_comm_rows, mpi_comm_cols, mpi_comm_all, &
                                  THIS_REAL_ELPA_KERNEL_API, useQR)           &
                                  result(success) bind(C,name="elpa_solve_evp_real_2stage_double_precision")

464
465
                                  result(success) bind(C,name="elpa_solve_evp_real_2stage_single_precision")
#endif
Andreas Marek's avatar
Andreas Marek committed
466
    use, intrinsic :: iso_c_binding
467
    use elpa2
Andreas Marek's avatar
Andreas Marek committed
468

Andreas Marek's avatar
Andreas Marek committed
469
    implicit none
Andreas Marek's avatar
Andreas Marek committed
470
    integer(kind=c_int)                    :: success
471
    integer(kind=c_int), value, intent(in) :: na, nev, lda, ldq, nblk, matrixCols, mpi_comm_cols, mpi_comm_rows, &
Andreas Marek's avatar
Andreas Marek committed
472
473
                                              mpi_comm_all
    integer(kind=c_int), value, intent(in) :: THIS_REAL_ELPA_KERNEL_API, useQR
474
#ifdef DOUBLE_PRECISION_REAL
475
    real(kind=c_double)                    :: ev(1:na)
476
477
#ifdef USE_ASSUMED_SIZE
    real(kind=c_double)                    :: a(lda,*), q(ldq,*)
478
#else
479
480
    real(kind=c_double)                    :: a(1:lda,1:matrixCols), q(1:ldq,1:matrixCols)
#endif
481
482
483
484

#else /* SINGLE_PRECISION */

    real(kind=c_float)                     :: ev(1:na)
485
#ifdef USE_ASSUMED_SIZE
486
487
488
489
490
    real(kind=c_float)                     :: a(1:lda,*), q(1:ldq,*)
#else
    real(kind=c_float)                     :: a(1:lda,1:matrixCols), q(1:ldq,1:matrixCols)
#endif

491
#endif
Andreas Marek's avatar
Andreas Marek committed
492
493
494
495
496
497
498
499
500

    logical                                :: successFortran, useQRFortran

    if (useQR .eq. 0) then
      useQRFortran =.false.
    else
      useQRFortran = .true.
    endif

501
502
#ifdef DOUBLE_PRECISION_REAL
    successFortran = solve_evp_real_2stage_double(na, nev, a, lda, ev, q, ldq, nblk, matrixCols, mpi_comm_rows, &
Andreas Marek's avatar
Andreas Marek committed
503
                                           mpi_comm_cols, mpi_comm_all,                                  &
Andreas Marek's avatar
Andreas Marek committed
504
                                           THIS_REAL_ELPA_KERNEL_API, useQRFortran)
505
506
507
508
509
510
511
512
513
514
#else
    successFortran = solve_evp_real_2stage_single(na, nev, a, lda, ev, q, ldq, nblk, matrixCols, mpi_comm_rows, &
                                           mpi_comm_cols, mpi_comm_all,                                  &
                                           THIS_REAL_ELPA_KERNEL_API, useQRFortran)
#endif
    if (successFortran) then
      success = 1
    else
      success = 0
    endif
Andreas Marek's avatar
Andreas Marek committed
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
  end function

#ifdef WANT_SINGLE_PRECISION_REAL

  !c> /*! \brief C interface to solve the single-precision real eigenvalue problem with 2-stage solver
  !c> *
  !c> *  \param  na                        Order of matrix a
  !c> *  \param  nev                       Number of eigenvalues needed.
  !c> *                                    The smallest nev eigenvalues/eigenvectors are calculated.
  !c> *  \param  a                         Distributed matrix for which eigenvalues are to be computed.
  !c> *                                    Distribution is like in Scalapack.
  !c> *                                    The full matrix must be set (not only one half like in scalapack).
  !c> *  \param lda                        Leading dimension of a
  !c> *  \param ev(na)                     On output: eigenvalues of a, every processor gets the complete set
  !c> *  \param q                          On output: Eigenvectors of a
  !c> *                                    Distribution is like in Scalapack.
  !c> *                                    Must be always dimensioned to the full size (corresponding to (na,na))
  !c> *                                    even if only a part of the eigenvalues is needed.
  !c> *  \param ldq                        Leading dimension of q
  !c> *  \param nblk                       blocksize of cyclic distribution, must be the same in both directions!
  !c> *  \param matrixCols                 distributed number of matrix columns
  !c> *  \param mpi_comm_rows              MPI-Communicator for rows
  !c> *  \param mpi_comm_cols              MPI-Communicator for columns
  !c> *  \param mpi_coll_all               MPI communicator for the total processor set
  !c> *  \param THIS_REAL_ELPA_KERNEL_API  specify used ELPA2 kernel via API
  !c> *  \param use_qr                     use QR decomposition 1 = yes, 0 = no
  !c> *
  !c> *  \result                     int: 1 if error occured, otherwise 0
  !c> */
#undef DOUBLE_PRECISION_REAL
#ifdef DOUBLE_PRECISION_REAL
  !c> int elpa_solve_evp_real_2stage_double_precision(int na, int nev, double *a, int lda, double *ev, double *q, int ldq, int nblk, int matrixCols, int mpi_comm_rows, int mpi_comm_cols, int mpi_comm_all, int THIS_REAL_ELPA_KERNEL_API, int useQR);
#else
  !c> int elpa_solve_evp_real_2stage_single_precision(int na, int nev, float *a, int lda, float *ev, float *q, int ldq, int nblk, int matrixCols, int mpi_comm_rows, int mpi_comm_cols, int mpi_comm_all, int THIS_REAL_ELPA_KERNEL_API, int useQR);
#endif

#ifdef DOUBLE_PRECISION_REAL
  function solve_elpa2_evp_real_wrapper_double(na, nev, a, lda, ev, q, ldq, nblk,    &
                                  matrixCols, mpi_comm_rows, mpi_comm_cols, mpi_comm_all, &
                                  THIS_REAL_ELPA_KERNEL_API, useQR)           &
                                  result(success) bind(C,name="elpa_solve_evp_real_2stage_double_precision")
#else
  function solve_elpa2_evp_real_wrapper_single(na, nev, a, lda, ev, q, ldq, nblk,    &
                                  matrixCols, mpi_comm_rows, mpi_comm_cols, mpi_comm_all, &
                                  THIS_REAL_ELPA_KERNEL_API, useQR)           &
                                  result(success) bind(C,name="elpa_solve_evp_real_2stage_single_precision")
#endif
    use, intrinsic :: iso_c_binding
    use elpa2

    implicit none
    integer(kind=c_int)                    :: success
    integer(kind=c_int), value, intent(in) :: na, nev, lda, ldq, nblk, matrixCols, mpi_comm_cols, mpi_comm_rows, &
                                              mpi_comm_all
    integer(kind=c_int), value, intent(in) :: THIS_REAL_ELPA_KERNEL_API, useQR
#ifdef DOUBLE_PRECISION_REAL
572
    real(kind=c_double)                    ::  ev(1:na)
573
#ifdef USE_ASSUMED_SIZE
574
    real(kind=c_double)                    :: a(1:lda,*), q(1:ldq,*)
575
#else
576
577
    real(kind=c_double)                    :: a(1:lda,1:matrixCols), q(1:ldq,1:matrixCols)
#endif
Andreas Marek's avatar
Andreas Marek committed
578

579
580
581
#else /* SINGLE_PRECISION */

    real(kind=c_float)                     :: ev(1:na)
582
#ifdef USE_ASSUMED_SIZE
583
584
585
    real(kind=c_float)                     :: a(1:lda,*), q(1:ldq,*)
#else
    real(kind=c_float)                     :: a(1:lda,1:matrixCols), q(1:ldq,1:matrixCols)
586
587
#endif

588
#endif
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
    logical                                :: successFortran, useQRFortran

    if (useQR .eq. 0) then
      useQRFortran =.false.
    else
      useQRFortran = .true.
    endif

#ifdef DOUBLE_PRECISION_REAL
    successFortran = solve_evp_real_2stage_double(na, nev, a, lda, ev, q, ldq, nblk, matrixCols, mpi_comm_rows, &
                                           mpi_comm_cols, mpi_comm_all,                                  &
                                           THIS_REAL_ELPA_KERNEL_API, useQRFortran)
#else
    successFortran = solve_evp_real_2stage_single(na, nev, a, lda, ev, q, ldq, nblk, matrixCols, mpi_comm_rows, &
                                           mpi_comm_cols, mpi_comm_all,                                  &
                                           THIS_REAL_ELPA_KERNEL_API, useQRFortran)
#endif
Andreas Marek's avatar
Andreas Marek committed
606
607
608
609
610
611
612
613
    if (successFortran) then
      success = 1
    else
      success = 0
    endif

  end function

614
#endif /* WANT_SINGLE_PRECISION_REAL */
615

616
  !c> /*! \brief C interface to solve the double-precision complex eigenvalue problem with 2-stage solver
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
  !c> *
  !c> *  \param  na                        Order of matrix a
  !c> *  \param  nev                       Number of eigenvalues needed.
  !c> *                                    The smallest nev eigenvalues/eigenvectors are calculated.
  !c> *  \param  a                         Distributed matrix for which eigenvalues are to be computed.
  !c> *                                    Distribution is like in Scalapack.
  !c> *                                    The full matrix must be set (not only one half like in scalapack).
  !c> *  \param lda                        Leading dimension of a
  !c> *  \param ev(na)                     On output: eigenvalues of a, every processor gets the complete set
  !c> *  \param q                          On output: Eigenvectors of a
  !c> *                                    Distribution is like in Scalapack.
  !c> *                                    Must be always dimensioned to the full size (corresponding to (na,na))
  !c> *                                    even if only a part of the eigenvalues is needed.
  !c> *  \param ldq                        Leading dimension of q
  !c> *  \param nblk                       blocksize of cyclic distribution, must be the same in both directions!
  !c> *  \param matrixCols                 distributed number of matrix columns
  !c> *  \param mpi_comm_rows              MPI-Communicator for rows
  !c> *  \param mpi_comm_cols              MPI-Communicator for columns
  !c> *  \param mpi_coll_all               MPI communicator for the total processor set
636
  !c> *  \param THIS_COMPLEX_ELPA_KERNEL_API  specify used ELPA2 kernel via API
637
638
639
  !c> *
  !c> *  \result                     int: 1 if error occured, otherwise 0
  !c> */
640
641
#define DOUBLE_PRECISION_COMPLEX 1

642
643
644
645
646
#ifdef DOUBLE_PRECISION_COMPLEX
  !c> int elpa_solve_evp_complex_2stage_double_precision(int na, int nev, double complex *a, int lda, double *ev, double complex *q, int ldq, int nblk, int matrixCols, int mpi_comm_rows, int mpi_comm_cols, int mpi_comm_all, int THIS_COMPLEX_ELPA_KERNEL_API);
#else
  !c> int elpa_solve_evp_complex_2stage_single_precision(int na, int nev, complex *a, int lda, float *ev, complex *q, int ldq, int nblk, int matrixCols, int mpi_comm_rows, int mpi_comm_cols, int mpi_comm_all, int THIS_COMPLEX_ELPA_KERNEL_API);
#endif
647
648
649

#ifdef DOUBLE_PRECISION_COMPLEX
  function solve_elpa2_evp_complex_wrapper_double(na, nev, a, lda, ev, q, ldq, nblk,    &
650
                                  matrixCols, mpi_comm_rows, mpi_comm_cols, mpi_comm_all,    &
Andreas Marek's avatar
Andreas Marek committed
651
                                  THIS_COMPLEX_ELPA_KERNEL_API)                  &
652
653
                                  result(success) bind(C,name="elpa_solve_evp_complex_2stage_double_precision")
#else
654
655
656
  function solve_elpa2_evp_complex_wrapper_single(na, nev, a, lda, ev, q, ldq, nblk,    &
                                  matrixCols, mpi_comm_rows, mpi_comm_cols, mpi_comm_all,    &
                                  THIS_COMPLEX_ELPA_KERNEL_API)                  &
657
658
                                  result(success) bind(C,name="elpa_solve_evp_complex_2stage_single_precision")
#endif
Andreas Marek's avatar
Andreas Marek committed
659
660

    use, intrinsic :: iso_c_binding
661
    use elpa2
Andreas Marek's avatar
Andreas Marek committed
662

Andreas Marek's avatar
Andreas Marek committed
663
    implicit none
Andreas Marek's avatar
Andreas Marek committed
664
    integer(kind=c_int)                    :: success
665
    integer(kind=c_int), value, intent(in) :: na, nev, lda, ldq, nblk, matrixCols, mpi_comm_cols, mpi_comm_rows, &
Andreas Marek's avatar
Andreas Marek committed
666
667
                                              mpi_comm_all
    integer(kind=c_int), value, intent(in) :: THIS_COMPLEX_ELPA_KERNEL_API
668
#ifdef DOUBLE_PRECISION_COMPLEX
Andreas Marek's avatar
Andreas Marek committed
669
    real(kind=c_double)                    :: ev(1:na)
670
#ifdef USE_ASSUMED_SIZE
671
    complex(kind=c_double_complex)         :: a(lda,*), q(ldq,*)
672
#else
673
    complex(kind=c_double_complex)         :: a(1:lda,1:matrixCols), q(1:ldq,1:matrixCols)
674
#endif
675
676

#else /* SINGLE_PRECISION */
677
    real(kind=c_float)                     :: ev(1:na)
678
#ifdef USE_ASSUMED_SIZE
679
680
681
682
683
    complex(kind=c_float_complex)          ::  a(lda,*), q(ldq,*)
#else
    complex(kind=c_float_complex)          :: a(1:lda,1:matrixCols), q(1:ldq,1:matrixCols)
#endif

684
#endif
Andreas Marek's avatar
Andreas Marek committed
685
686
    logical                                :: successFortran

687
688
#ifdef DOUBLE_PRECISION_COMPLEX
    successFortran = solve_evp_complex_2stage_double(na, nev, a, lda, ev, q, ldq, nblk, matrixCols, mpi_comm_rows, mpi_comm_cols, &
Andreas Marek's avatar
Andreas Marek committed
689
                                              mpi_comm_all, THIS_COMPLEX_ELPA_KERNEL_API)
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
#else
    successFortran = solve_evp_complex_2stage_single(na, nev, a, lda, ev, q, ldq, nblk, matrixCols, mpi_comm_rows, mpi_comm_cols, &
                                              mpi_comm_all, THIS_COMPLEX_ELPA_KERNEL_API)
#endif
    if (successFortran) then
      success = 1
    else
      success = 0
    endif

  end function

#ifdef WANT_SINGLE_PRECISION_COMPLEX

  !c> /*! \brief C interface to solve the single-precision complex eigenvalue problem with 2-stage solver
  !c> *
  !c> *  \param  na                        Order of matrix a
  !c> *  \param  nev                       Number of eigenvalues needed.
  !c> *                                    The smallest nev eigenvalues/eigenvectors are calculated.
  !c> *  \param  a                         Distributed matrix for which eigenvalues are to be computed.
  !c> *                                    Distribution is like in Scalapack.
  !c> *                                    The full matrix must be set (not only one half like in scalapack).
  !c> *  \param lda                        Leading dimension of a
  !c> *  \param ev(na)                     On output: eigenvalues of a, every processor gets the complete set
  !c> *  \param q                          On output: Eigenvectors of a
  !c> *                                    Distribution is like in Scalapack.
  !c> *                                    Must be always dimensioned to the full size (corresponding to (na,na))
  !c> *                                    even if only a part of the eigenvalues is needed.
  !c> *  \param ldq                        Leading dimension of q
  !c> *  \param nblk                       blocksize of cyclic distribution, must be the same in both directions!
  !c> *  \param matrixCols                 distributed number of matrix columns
  !c> *  \param mpi_comm_rows              MPI-Communicator for rows
  !c> *  \param mpi_comm_cols              MPI-Communicator for columns
  !c> *  \param mpi_coll_all               MPI communicator for the total processor set
  !c> *  \param THIS_REAL_ELPA_KERNEL_API  specify used ELPA2 kernel via API
  !c> *  \param use_qr                     use QR decomposition 1 = yes, 0 = no
  !c> *
  !c> *  \result                     int: 1 if error occured, otherwise 0
  !c> */
#undef DOUBLE_PRECISION_COMPLEX
Andreas Marek's avatar
Andreas Marek committed
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
#ifdef DOUBLE_PRECISION_COMPLEX
  !c> int elpa_solve_evp_complex_2stage_double_precision(int na, int nev, double complex *a, int lda, double *ev, double complex *q, int ldq, int nblk, int matrixCols, int mpi_comm_rows, int mpi_comm_cols, int mpi_comm_all, int THIS_COMPLEX_ELPA_KERNEL_API);
#else
  !c> int elpa_solve_evp_complex_2stage_single_precision(int na, int nev, complex *a, int lda, float *ev, complex *q, int ldq, int nblk, int matrixCols, int mpi_comm_rows, int mpi_comm_cols, int mpi_comm_all, int THIS_COMPLEX_ELPA_KERNEL_API);
#endif

#ifdef DOUBLE_PRECISION_COMPLEX
  function solve_elpa2_evp_complex_wrapper_double(na, nev, a, lda, ev, q, ldq, nblk,    &
                                  matrixCols, mpi_comm_rows, mpi_comm_cols, mpi_comm_all,    &
                                  THIS_COMPLEX_ELPA_KERNEL_API)                  &
                                  result(success) bind(C,name="elpa_solve_evp_complex_2stage_double_precision")
#else
  function solve_elpa2_evp_complex_wrapper_single(na, nev, a, lda, ev, q, ldq, nblk,    &
                                  matrixCols, mpi_comm_rows, mpi_comm_cols, mpi_comm_all,    &
                                  THIS_COMPLEX_ELPA_KERNEL_API)                  &
                                  result(success) bind(C,name="elpa_solve_evp_complex_2stage_single_precision")
#endif

    use, intrinsic :: iso_c_binding
    use elpa2

    implicit none
    integer(kind=c_int)                    :: success
    integer(kind=c_int), value, intent(in) :: na, nev, lda, ldq, nblk, matrixCols, mpi_comm_cols, mpi_comm_rows, &
                                              mpi_comm_all
    integer(kind=c_int), value, intent(in) :: THIS_COMPLEX_ELPA_KERNEL_API
#ifdef DOUBLE_PRECISION_COMPLEX
    complex(kind=c_double_complex)         :: a(1:lda,1:matrixCols), q(1:ldq,1:matrixCols)
    real(kind=c_double)                    :: ev(1:na)
#else
    complex(kind=c_float_complex)          :: a(1:lda,1:matrixCols), q(1:ldq,1:matrixCols)
    real(kind=c_float)                     :: ev(1:na)
#endif
    logical                                :: successFortran

#ifdef DOUBLE_PRECISION_COMPLEX
    successFortran = solve_evp_complex_2stage_double(na, nev, a, lda, ev, q, ldq, nblk, matrixCols, mpi_comm_rows, mpi_comm_cols, &
                                              mpi_comm_all, THIS_COMPLEX_ELPA_KERNEL_API)
#else
    successFortran = solve_evp_complex_2stage_single(na, nev, a, lda, ev, q, ldq, nblk, matrixCols, mpi_comm_rows, mpi_comm_cols, &
                                              mpi_comm_all, THIS_COMPLEX_ELPA_KERNEL_API)
#endif
Andreas Marek's avatar
Andreas Marek committed
773
774
775
776
777
778
779
    if (successFortran) then
      success = 1
    else
      success = 0
    endif

  end function
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
#endif /* WANT_SINGLE_PRECISION_COMPLEX */

  !c> /*! \brief C interface to driver function "elpa_solve_evp_real_double"
  !c> *
  !c> *  \param  na                        Order of matrix a
  !c> *  \param  nev                       Number of eigenvalues needed.
  !c> *                                    The smallest nev eigenvalues/eigenvectors are calculated.
  !c> *  \param  a                         Distributed matrix for which eigenvalues are to be computed.
  !c> *                                    Distribution is like in Scalapack.
  !c> *                                    The full matrix must be set (not only one half like in scalapack).
  !c> *  \param lda                        Leading dimension of a
  !c> *  \param ev(na)                     On output: eigenvalues of a, every processor gets the complete set
  !c> *  \param q                          On output: Eigenvectors of a
  !c> *                                    Distribution is like in Scalapack.
  !c> *                                    Must be always dimensioned to the full size (corresponding to (na,na))
  !c> *                                    even if only a part of the eigenvalues is needed.
  !c> *  \param ldq                        Leading dimension of q
  !c> *  \param nblk                       blocksize of cyclic distribution, must be the same in both directions!
  !c> *  \param matrixCols                 distributed number of matrix columns
  !c> *  \param mpi_comm_rows              MPI-Communicator for rows
  !c> *  \param mpi_comm_cols              MPI-Communicator for columns
  !c> *  \param mpi_coll_all               MPI communicator for the total processor set
  !c> *  \param THIS_REAL_ELPA_KERNEL_API  specify used ELPA2 kernel via API
  !c> *  \param use_qr                     use QR decomposition 1 = yes, 0 = no
  !c> *  \param method                      choose whether to use ELPA 1stage or 2stage solver
  !c> *                                     possible values: "1stage" => use ELPA 1stage solver
  !c> *                                                      "2stage" => use ELPA 2stage solver
  !c> *                                                       "auto"   => (at the moment) use ELPA 2stage solver
  !c> *
  !c> *  \result                     int: 1 if error occured, otherwise 0
  !c> */
  !c> int elpa_solve_evp_real_double(int na, int nev, double *a, int lda, double *ev, double *q, int ldq, int nblk, int matrixCols, int mpi_comm_rows, int mpi_comm_cols, int mpi_comm_all, int THIS_REAL_ELPA_KERNEL_API, int useQR, char *method);
  function elpa_solve_evp_real_wrapper_double(na, nev, a, lda, ev, q, ldq, nblk,    &
                                  matrixCols, mpi_comm_rows, mpi_comm_cols, mpi_comm_all, &
                                  THIS_REAL_ELPA_KERNEL_API, useQR, method)           &
                                  result(success) bind(C,name="elpa_solve_evp_real_double")

    use, intrinsic :: iso_c_binding
    use elpa, only : elpa_solve_evp_real_double

    implicit none
    integer(kind=c_int)                      :: success
    integer(kind=c_int), value, intent(in)   :: na, nev, lda, ldq, nblk, matrixCols, mpi_comm_cols, mpi_comm_rows, &
                                                mpi_comm_all
    integer(kind=c_int), value, intent(in)   :: THIS_REAL_ELPA_KERNEL_API, useQR
    real(kind=c_double)                      :: ev(1:na)
#ifdef USE_ASSUMED_SIZE
    real(kind=c_double)                      :: a(lda,*), q(ldq,*)
#else
    real(kind=c_double)                      :: a(1:lda,1:matrixCols), q(1:ldq,1:matrixCols)
#endif
    logical                                  :: successFortran, useQRFortran
    character(kind=c_char,len=1), intent(in) :: method(*)
    character(len=6)                         :: methodFortran
    integer(kind=c_int)                      :: charCount

    if (useQR .eq. 0) then
      useQRFortran =.false.
    else
      useQRFortran = .true.
    endif

    charCount = 1
    do
      if (method(charCount) == c_null_char) exit
      charCount = charCount + 1
    enddo
    charCount = charCount - 1

    if (charCount .ge. 1)  then
      methodFortran(1:charCount) = transfer(method(1:charCount), methodFortran)

      successFortran = elpa_solve_evp_real_double(na, nev, a, lda, ev, q, ldq, nblk, matrixCols, mpi_comm_rows, &
                                           mpi_comm_cols, mpi_comm_all,                                  &
                                           THIS_REAL_ELPA_KERNEL_API, useQRFortran, methodFortran)
    else
      successFortran = elpa_solve_evp_real_double(na, nev, a, lda, ev, q, ldq, nblk, matrixCols, mpi_comm_rows, &
                                           mpi_comm_cols, mpi_comm_all,                                  &
                                           THIS_REAL_ELPA_KERNEL_API, useQRFortran)
    endif

    if (successFortran) then
      success = 1
    else
      success = 0
    endif

  end function
Andreas Marek's avatar
Andreas Marek committed
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
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
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
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
#ifdef WANT_SINGLE_PRECISION_REAL
  !c> /*! \brief C interface to driver function "elpa_solve_evp_real_single"
  !c> *
  !c> *  \param  na                        Order of matrix a
  !c> *  \param  nev                       Number of eigenvalues needed.
  !c> *                                    The smallest nev eigenvalues/eigenvectors are calculated.
  !c> *  \param  a                         Distributed matrix for which eigenvalues are to be computed.
  !c> *                                    Distribution is like in Scalapack.
  !c> *                                    The full matrix must be set (not only one half like in scalapack).
  !c> *  \param lda                        Leading dimension of a
  !c> *  \param ev(na)                     On output: eigenvalues of a, every processor gets the complete set
  !c> *  \param q                          On output: Eigenvectors of a
  !c> *                                    Distribution is like in Scalapack.
  !c> *                                    Must be always dimensioned to the full size (corresponding to (na,na))
  !c> *                                    even if only a part of the eigenvalues is needed.
  !c> *  \param ldq                        Leading dimension of q
  !c> *  \param nblk                       blocksize of cyclic distribution, must be the same in both directions!
  !c> *  \param matrixCols                 distributed number of matrix columns
  !c> *  \param mpi_comm_rows              MPI-Communicator for rows
  !c> *  \param mpi_comm_cols              MPI-Communicator for columns
  !c> *  \param mpi_coll_all               MPI communicator for the total processor set
  !c> *  \param THIS_REAL_ELPA_KERNEL_API  specify used ELPA2 kernel via API
  !c> *  \param use_qr                     use QR decomposition 1 = yes, 0 = no
  !c> *  \param method                      choose whether to use ELPA 1stage or 2stage solver
  !c> *                                     possible values: "1stage" => use ELPA 1stage solver
  !c> *                                                      "2stage" => use ELPA 2stage solver
  !c> *                                                       "auto"   => (at the moment) use ELPA 2stage solver
  !c> *
  !c> *  \result                     int: 1 if error occured, otherwise 0
  !c> */
  !c> int elpa_solve_evp_real_single(int na, int nev, float *a, int lda, float *ev, float *q, int ldq, int nblk, int matrixCols, int mpi_comm_rows, int mpi_comm_cols, int mpi_comm_all, int THIS_REAL_ELPA_KERNEL_API, int useQR, char *method);
  function elpa_solve_evp_real_wrapper_single(na, nev, a, lda, ev, q, ldq, nblk,    &
                                  matrixCols, mpi_comm_rows, mpi_comm_cols, mpi_comm_all, &
                                  THIS_REAL_ELPA_KERNEL_API, useQR, method)           &
                                  result(success) bind(C,name="elpa_solve_evp_real_single")

    use, intrinsic :: iso_c_binding
    use elpa, only : elpa_solve_evp_real_single

    implicit none
    integer(kind=c_int)                      :: success
    integer(kind=c_int), value, intent(in)   :: na, nev, lda, ldq, nblk, matrixCols, mpi_comm_cols, mpi_comm_rows, &
                                                mpi_comm_all
    integer(kind=c_int), value, intent(in)   :: THIS_REAL_ELPA_KERNEL_API, useQR
    real(kind=c_float)                       :: ev(1:na)
#ifdef USE_ASSUMED_SIZE
    real(kind=c_float)                       :: a(lda,*), q(ldq,*)
#else
    real(kind=c_float)                       :: a(1:lda,1:matrixCols), q(1:ldq,1:matrixCols)
#endif
    logical                                  :: successFortran, useQRFortran
    character(kind=c_char,len=1), intent(in) :: method(*)
    character(len=6)                         :: methodFortran
    integer(kind=c_int)                      :: charCount

    if (useQR .eq. 0) then
      useQRFortran =.false.
    else
      useQRFortran = .true.
    endif

    charCount = 1
    do
      if (method(charCount) == c_null_char) exit
      charCount = charCount + 1
    enddo
    charCount = charCount - 1

    if (charCount .ge. 1)  then
      methodFortran(1:charCount) = transfer(method(1:charCount), methodFortran)

      successFortran = elpa_solve_evp_real_single(na, nev, a, lda, ev, q, ldq, nblk, matrixCols, mpi_comm_rows, &
                                           mpi_comm_cols, mpi_comm_all,                                  &
                                           THIS_REAL_ELPA_KERNEL_API, useQRFortran, methodFortran)
    else
      successFortran = elpa_solve_evp_real_single(na, nev, a, lda, ev, q, ldq, nblk, matrixCols, mpi_comm_rows, &
                                           mpi_comm_cols, mpi_comm_all,                                  &
                                           THIS_REAL_ELPA_KERNEL_API, useQRFortran)
    endif

    if (successFortran) then
      success = 1
    else
      success = 0
    endif

  end function
#endif /* WANT_SINGLE_PRECISION_REAL */

  !c> /*! \brief C interface to driver function "elpa_solve_evp_complex_double"
  !c> *
  !c> *  \param  na                           Order of matrix a
  !c> *  \param  nev                          Number of eigenvalues needed.
  !c> *                                       The smallest nev eigenvalues/eigenvectors are calculated.
  !c> *  \param  a                            Distributed matrix for which eigenvalues are to be computed.
  !c> *                                       Distribution is like in Scalapack.
  !c> *                                       The full matrix must be set (not only one half like in scalapack).
  !c> *  \param lda                           Leading dimension of a
  !c> *  \param ev(na)                        On output: eigenvalues of a, every processor gets the complete set
  !c> *  \param q                             On output: Eigenvectors of a
  !c> *                                       Distribution is like in Scalapack.
  !c> *                                       Must be always dimensioned to the full size (corresponding to (na,na))
  !c> *                                       even if only a part of the eigenvalues is needed.
  !c> *  \param ldq                           Leading dimension of q
  !c> *  \param nblk                          blocksize of cyclic distribution, must be the same in both directions!
  !c> *  \param matrixCols                    distributed number of matrix columns
  !c> *  \param mpi_comm_rows                 MPI-Communicator for rows
  !c> *  \param mpi_comm_cols                 MPI-Communicator for columns
  !c> *  \param mpi_coll_all                  MPI communicator for the total processor set
  !c> *  \param THIS_COMPLEX_ELPA_KERNEL_API  specify used ELPA2 kernel via API
  !c> *  \param method                        choose whether to use ELPA 1stage or 2stage solver
  !c> *                                       possible values: "1stage" => use ELPA 1stage solver
  !c> *                                                        "2stage" => use ELPA 2stage solver
  !c> *                                                         "auto"   => (at the moment) use ELPA 2stage solver
  !c> *
  !c> *  \result                     int: 1 if error occured, otherwise 0
  !c> */
  !c> int elpa_solve_evp_complex_double(int na, int nev, double complex *a, int lda, double *ev, double complex *q, int ldq, int nblk, int matrixCols, int mpi_comm_rows, int mpi_comm_cols, int mpi_comm_all, int THIS_COMPLEX_ELPA_KERNEL_API, char *method);
  function elpa_solve_evp_complex_wrapper_double(na, nev, a, lda, ev, q, ldq, nblk,    &
                                  matrixCols, mpi_comm_rows, mpi_comm_cols, mpi_comm_all,    &
                                  THIS_COMPLEX_ELPA_KERNEL_API, method)                  &
                                  result(success) bind(C,name="elpa_solve_evp_complex_double")

    use, intrinsic :: iso_c_binding
    use elpa, only : elpa_solve_evp_complex_double

    implicit none
    integer(kind=c_int)                      :: success
    integer(kind=c_int), value, intent(in)   :: na, nev, lda, ldq, nblk, matrixCols, mpi_comm_cols, mpi_comm_rows, &
                                                mpi_comm_all
    integer(kind=c_int), value, intent(in)   :: THIS_COMPLEX_ELPA_KERNEL_API
#ifdef USE_ASSUMED_SIZE
    complex(kind=c_double_complex)           :: a(lda,*), q(ldq,*)
#else
    complex(kind=c_double_complex)           :: a(1:lda,1:matrixCols), q(1:ldq,1:matrixCols)
#endif
    real(kind=c_double)                      :: ev(1:na)
    character(kind=c_char,len=1), intent(in) :: method(*)
    character(len=6)                         :: methodFortran
    integer(kind=c_int)                      :: charCount

    logical                                  :: successFortran


    charCount = 1
    do
      if (method(charCount) == c_null_char) exit
      charCount = charCount + 1
    enddo
    charCount = charCount - 1

    if (charCount .ge. 1)  then
      methodFortran(1:charCount) = transfer(method(1:charCount), methodFortran)
      successFortran = elpa_solve_evp_complex_double(na, nev, a, lda, ev, q, ldq, nblk, matrixCols, mpi_comm_rows, mpi_comm_cols, &
                                              mpi_comm_all, THIS_COMPLEX_ELPA_KERNEL_API, methodFortran)
    else
      successFortran = elpa_solve_evp_complex_double(na, nev, a, lda, ev, q, ldq, nblk, matrixCols, mpi_comm_rows, mpi_comm_cols, &
                                              mpi_comm_all, THIS_COMPLEX_ELPA_KERNEL_API)
    endif

    if (successFortran) then
      success = 1
    else
      success = 0
    endif

  end function

#ifdef WANT_SINGLE_PRECISION_COMPLEX
  !c> /*! \brief C interface to driver function "elpa_solve_evp_complex_single"
  !c> *
  !c> *  \param  na                           Order of matrix a
  !c> *  \param  nev                          Number of eigenvalues needed.
  !c> *                                       The smallest nev eigenvalues/eigenvectors are calculated.
  !c> *  \param  a                            Distributed matrix for which eigenvalues are to be computed.
  !c> *                                       Distribution is like in Scalapack.
  !c> *                                       The full matrix must be set (not only one half like in scalapack).
  !c> *  \param lda                           Leading dimension of a
  !c> *  \param ev(na)                        On output: eigenvalues of a, every processor gets the complete set
  !c> *  \param q                             On output: Eigenvectors of a
  !c> *                                       Distribution is like in Scalapack.
  !c> *                                       Must be always dimensioned to the full size (corresponding to (na,na))
  !c> *                                       even if only a part of the eigenvalues is needed.
  !c> *  \param ldq                           Leading dimension of q
  !c> *  \param nblk                          blocksize of cyclic distribution, must be the same in both directions!
  !c> *  \param matrixCols                    distributed number of matrix columns
  !c> *  \param mpi_comm_rows                 MPI-Communicator for rows
  !c> *  \param mpi_comm_cols                 MPI-Communicator for columns
  !c> *  \param mpi_coll_all                  MPI communicator for the total processor set
  !c> *  \param THIS_COMPLEX_ELPA_KERNEL_API  specify used ELPA2 kernel via API
  !c> *  \param method                        choose whether to use ELPA 1stage or 2stage solver
  !c> *                                       possible values: "1stage" => use ELPA 1stage solver
  !c> *                                                        "2stage" => use ELPA 2stage solver
  !c> *                                                         "auto"   => (at the moment) use ELPA 2stage solver
  !c> *
  !c> *  \result                     int: 1 if error occured, otherwise 0
  !c> */
  !c> int elpa_solve_evp_complex_single(int na, int nev, complex *a, int lda, float *ev, complex *q, int ldq, int nblk, int matrixCols, int mpi_comm_rows, int mpi_comm_cols, int mpi_comm_all, int THIS_COMPLEX_ELPA_KERNEL_API, char *method);
  function elpa_solve_evp_complex_wrapper_single(na, nev, a, lda, ev, q, ldq, nblk,    &
                                  matrixCols, mpi_comm_rows, mpi_comm_cols, mpi_comm_all,    &
                                  THIS_COMPLEX_ELPA_KERNEL_API, method)                  &
                                  result(success) bind(C,name="elpa_solve_evp_complex_single")

    use, intrinsic :: iso_c_binding
    use elpa, only : elpa_solve_evp_complex_single

    implicit none
    integer(kind=c_int)                      :: success
    integer(kind=c_int), value, intent(in)   :: na, nev, lda, ldq, nblk, matrixCols, mpi_comm_cols, mpi_comm_rows, &
                                                mpi_comm_all
    integer(kind=c_int), value, intent(in)   :: THIS_COMPLEX_ELPA_KERNEL_API
#ifdef USE_ASSUMED_SIZE
    complex(kind=c_float_complex)            :: a(lda,*), q(ldq,*)
#else
    complex(kind=c_float_complex)            :: a(1:lda,1:matrixCols), q(1:ldq,1:matrixCols)
#endif
    real(kind=c_float)                       :: ev(1:na)
    character(kind=c_char,len=1), intent(in) :: method(*)
    character(len=6)                         :: methodFortran
    integer(kind=c_int)                      :: charCount

    logical                                  :: successFortran


    charCount = 1
    do
      if (method(charCount) == c_null_char) exit
      charCount = charCount + 1
    enddo
    charCount = charCount - 1

    if (charCount .ge. 1)  then
      methodFortran(1:charCount) = transfer(method(1:charCount), methodFortran)
      successFortran = elpa_solve_evp_complex_single(na, nev, a, lda, ev, q, ldq, nblk, matrixCols, mpi_comm_rows, mpi_comm_cols, &
                                              mpi_comm_all, THIS_COMPLEX_ELPA_KERNEL_API, methodFortran)
    else
      successFortran = elpa_solve_evp_complex_single(na, nev, a, lda, ev, q, ldq, nblk, matrixCols, mpi_comm_rows, mpi_comm_cols, &
                                              mpi_comm_all, THIS_COMPLEX_ELPA_KERNEL_API)
    endif

    if (successFortran) then
      success = 1
    else
      success = 0
    endif

  end function
1116
1117
#endif /* WANT_SINGLE_PRECISION_COMPLEX */

1118
  !c> /*
1119
  !c> \brief  C interface to solve double-precision tridiagonal eigensystem with divide and conquer method
1120
1121
  !c> \details
  !c>
Andreas Marek's avatar
Andreas Marek committed
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
  !c> *\param na                    Matrix dimension
  !c> *\param nev                   number of eigenvalues/vectors to be computed
  !c> *\param d                     array d(na) on input diagonal elements of tridiagonal matrix, on
  !c> *                             output the eigenvalues in ascending order
  !c> *\param e                     array e(na) on input subdiagonal elements of matrix, on exit destroyed
  !c> *\param q                     on exit : matrix q(ldq,matrixCols) contains the eigenvectors
  !c> *\param ldq                   leading dimension of matrix q
  !c> *\param nblk                  blocksize of cyclic distribution, must be the same in both directions!
  !c> *\param matrixCols            columns of matrix q
  !c> *\param mpi_comm_rows         MPI communicator for rows
  !c> *\param mpi_comm_cols         MPI communicator for columns
  !c> *\param wantDebug             give more debug information if 1, else 0
  !c> *\result success              int 1 on success, else 0
1135
  !c> */
1136
1137
1138
  !c> int elpa_solve_tridi_double(int na, int nev, double *d, double *e, double *q, int ldq, int nblk, int matrixCols, int mpi_comm_rows, int mpi_comm_cols, int wantDebug);
  function elpa_solve_tridi_wrapper_double(na, nev, d, e, q, ldq, nblk, matrixCols, mpi_comm_rows, mpi_comm_cols, wantDebug) &
           result(success) bind(C,name="elpa_solve_tridi_double")
1139
1140

    use, intrinsic :: iso_c_binding
1141
    use elpa1_auxiliary, only : elpa_solve_tridi_double
1142
1143
1144
1145
1146

    implicit none
    integer(kind=c_int)                    :: success
    integer(kind=c_int), value, intent(in) :: na, nev, ldq, nblk, matrixCols,  mpi_comm_cols, mpi_comm_rows
    integer(kind=c_int), value             :: wantDebug
1147
    real(kind=c_double)                    :: d(1:na), e(1:na)
1148
#ifdef USE_ASSUMED_SIZE
1149
1150
1151
1152
    real(kind=c_double)                    :: q(ldq,*)
#else
    real(kind=c_double)                    :: q(1:ldq, 1:matrixCols)
#endif
1153
1154
1155
1156
1157
1158
1159
1160
    logical                                :: successFortran, wantDebugFortran

    if (wantDebug .ne. 0) then
      wantDebugFortran = .true.
    else
      wantDebugFortran = .false.
    endif

1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
    successFortran = elpa_solve_tridi_double(na, nev, d, e, q, ldq, nblk, matrixCols, mpi_comm_rows, mpi_comm_cols, &
                      wantDebugFortran)

    if (successFortran) then
      success = 1
    else
      success = 0
    endif

  end function

#ifdef WANT_SINGLE_PRECISION_REAL

  !c> /*
  !c> \brief  C interface to solve single-precision tridiagonal eigensystem with divide and conquer method
  !c> \details
  !c>
  !c> \param na                    Matrix dimension
  !c> \param nev                   number of eigenvalues/vectors to be computed
  !c> \param d                     array d(na) on input diagonal elements of tridiagonal matrix, on
  !c>                              output the eigenvalues in ascending order
  !c> \param e                     array e(na) on input subdiagonal elements of matrix, on exit destroyed
  !c> \param q                     on exit : matrix q(ldq,matrixCols) contains the eigenvectors
  !c> \param ldq                   leading dimension of matrix q
  !c> \param nblk                  blocksize of cyclic distribution, must be the same in both directions!
  !c> \param matrixCols            columns of matrix q
  !c> \param mpi_comm_rows         MPI communicator for rows
  !c> \param mpi_comm_cols         MPI communicator for columns
  !c> \param wantDebug             give more debug information if 1, else 0
  !c> \result success              int 1 on success, else 0
  !c> */
  !c> int elpa_solve_tridi_single(int na, int nev, float *d, float *e, float *q, int ldq, int nblk, int matrixCols, int mpi_comm_rows, int mpi_comm_cols, int wantDebug);
  function elpa_solve_tridi_wrapper_single(na, nev, d, e, q, ldq, nblk, matrixCols, mpi_comm_rows, mpi_comm_cols, wantDebug) &
           result(success) bind(C,name="elpa_solve_tridi_single")

    use, intrinsic :: iso_c_binding
    use elpa1_auxiliary, only : elpa_solve_tridi_single

    implicit none
    integer(kind=c_int)                    :: success
    integer(kind=c_int), value, intent(in) :: na, nev, ldq, nblk, matrixCols,  mpi_comm_cols, mpi_comm_rows
    integer(kind=c_int), value             :: wantDebug
    real(kind=c_float)                     :: d(1:na), e(1:na), q(1:ldq, 1:matrixCols)
    logical                                :: successFortran, wantDebugFortran

    if (wantDebug .ne. 0) then
      wantDebugFortran = .true.
    else
      wantDebugFortran = .false.
    endif

1212
1213
    successFortran = elpa_solve_tridi_single(na, nev, d, e, q, ldq, nblk, matrixCols, mpi_comm_rows, &
                                             mpi_comm_cols, wantDebugFortran)
1214
1215
1216
1217
1218
1219
1220
1221
1222

    if (successFortran) then
      success = 1
    else
      success = 0
    endif

  end function

1223
1224
#endif /* WANT_SINGLE_PRECISION_REAL */

1225
  !c> /*
1226
  !c> \brief  C interface for elpa_mult_at_b_real_double: Performs C : = A**T * B for double-precision matrices
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
  !c>         where   A is a square matrix (na,na) which is optionally upper or lower triangular
  !c>                 B is a (na,ncb) matrix
  !c>                 C is a (na,ncb) matrix where optionally only the upper or lower
  !c>                   triangle may be computed
  !c> \details
  !c> \param  uplo_a               'U' if A is upper triangular
  !c>                              'L' if A is lower triangular
  !c>                              anything else if A is a full matrix
  !c>                              Please note: This pertains to the original A (as set in the calling program)
  !c>                                           whereas the transpose of A is used for calculations
  !c>                              If uplo_a is 'U' or 'L', the other triangle is not used at all,
  !c>                              i.e. it may contain arbitrary numbers
  !c> \param uplo_c                'U' if only the upper diagonal part of C is needed
  !c>                              'L' if only the upper diagonal part of C is needed
  !c>                              anything else if the full matrix C is needed
  !c>                              Please note: Even when uplo_c is 'U' or 'L', the other triangle may be
  !c>                                            written to a certain extent, i.e. one shouldn't rely on the content there!
  !c> \param na                    Number of rows/columns of A, number of rows of B and C
  !c> \param ncb                   Number of columns  of B and C
  !c> \param a                     matrix a
  !c> \param lda                   leading dimension of matrix a
1248
  !c> \param ldaCols               columns of matrix a
1249
1250
  !c> \param b                     matrix b
  !c> \param ldb                   leading dimension of matrix b
1251
  !c> \param ldbCols               columns of matrix b
1252
1253
1254
1255
1256
  !c> \param nblk                  blocksize of cyclic distribution, must be the same in both directions!
  !c> \param  mpi_comm_rows        MPI communicator for rows
  !c> \param  mpi_comm_cols        MPI communicator for columns
  !c> \param c                     matrix c
  !c> \param ldc                   leading dimension of matrix c
1257
  !c> \param ldcCols               columns of matrix c
1258
1259
1260
  !c> \result success              int report success (1) or failure (0)
  !c> */

1261
1262
1263
  !c> int elpa_mult_at_b_real_double(char uplo_a, char uplo_c, int na, int ncb, double *a, int lda, int ldaCols, double *b, int ldb, int ldbCols, int nlbk, int mpi_comm_rows, int mpi_comm_cols, double *c, int ldc, int ldcCols);
  function elpa_mult_at_b_real_wrapper_double(uplo_a, uplo_c, na, ncb, a, lda, ldaCols, b, ldb, ldbCols, &
                                              nblk, mpi_comm_rows, mpi_comm_cols, c, ldc, ldcCols) &
1264
                                              bind(C,name="elpa_mult_at_b_real_double") result(success)
1265
    use, intrinsic :: iso_c_binding
1266
    use elpa1_auxiliary, only : elpa_mult_at_b_real_double
1267
1268
1269
1270

    implicit none

    character(1,C_CHAR), value  :: uplo_a, uplo_c
1271
1272
    integer(kind=c_int), value  :: na, ncb, lda, ldb, nblk, mpi_comm_rows, mpi_comm_cols, ldc, &
                                   ldaCols, ldbCols, ldcCols
1273
    integer(kind=c_int)         :: success
1274
#ifdef USE_ASSUMED_SIZE
1275
    real(kind=c_double)         :: a(lda,*), b(ldb,*), c(ldc,*)
1276
1277
1278
#else
    real(kind=c_double)         :: a(lda,ldaCols), b(ldb,ldbCols), c(ldc,ldcCols)
#endif
1279
1280
    logical                     :: successFortran

1281
1282
    successFortran = elpa_mult_at_b_real_double(uplo_a, uplo_c, na, ncb, a, lda, ldaCols, b, ldb, ldbCols, &
                                                nblk, mpi_comm_rows, mpi_comm_cols, c, ldc, ldcCols)
1283
1284
1285
1286
1287
1288
1289
1290
1291

    if (successFortran) then
      success = 1
    else
      success = 0
    endif

  end function

1292
#ifdef WANT_SINGLE_PRECISION_REAL
1293
  !c> /*
1294
  !c> \brief  C interface for elpa_mult_at_b_real_single: Performs C : = A**T * B for single-precision matrices
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
  !c>         where   A is a square matrix (na,na) which is optionally upper or lower triangular
  !c>                 B is a (na,ncb) matrix
  !c>                 C is a (na,ncb) matrix where optionally only the upper or lower
  !c>                   triangle may be computed
  !c> \details
  !c> \param  uplo_a               'U' if A is upper triangular
  !c>                              'L' if A is lower triangular
  !c>                              anything else if A is a full matrix
  !c>                              Please note: This pertains to the original A (as set in the calling program)
  !c>                                           whereas the transpose of A is used for calculations
  !c>                              If uplo_a is 'U' or 'L', the other triangle is not used at all,
  !c>                              i.e. it may contain arbitrary numbers
  !c> \param uplo_c                'U' if only the upper diagonal part of C is needed
  !c>                              'L' if only the upper diagonal part of C is needed
  !c>                              anything else if the full matrix C is needed
  !c>                              Please note: Even when uplo_c is 'U' or 'L', the other triangle may be
  !c>                                            written to a certain extent, i.e. one shouldn't rely on the content there!
  !c> \param na                    Number of rows/columns of A, number of rows of B and C
  !c> \param ncb                   Number of columns  of B and C
  !c> \param a                     matrix a
  !c> \param lda                   leading dimension of matrix a
1316
  !c> \param ldaCols               columns of matrix a
1317
1318
  !c> \param b                     matrix b
  !c> \param ldb                   leading dimension of matrix b
1319
  !c> \param ldbCols               columns of matrix b
1320
1321
1322
1323
1324
  !c> \param nblk                  blocksize of cyclic distribution, must be the same in both directions!
  !c> \param  mpi_comm_rows        MPI communicator for rows
  !c> \param  mpi_comm_cols        MPI communicator for columns
  !c> \param c                     matrix c
  !c> \param ldc                   leading dimension of matrix c
1325
  !c> \result success              int report success (1) or failure (0)
1326
1327
  !c> */

1328
1329
1330
1331
  !c> int elpa_mult_at_b_real_single(char uplo_a, char uplo_c, int na, int ncb, float *a, int lda, int ldaCols, float *b, int ldb, int ldbCols, int nlbk, int mpi_comm_rows, int mpi_comm_cols, float *c, int ldc, int ldcCols);
  function elpa_mult_at_b_real_wrapper_float(uplo_a, uplo_c, na, ncb, a, lda, ldaCols, b, ldb, ldbCols, &
                                             nblk, mpi_comm_rows, mpi_comm_cols, c, ldc, ldcCols) &
    bind(C,name="elpa_mult_at_b_real_float") result(success)
1332
    use, intrinsic :: iso_c_binding
1333
    use elpa1_auxiliary, only : elpa_mult_at_b_real_single
1334
1335
1336

    implicit none

1337
1338
1339
1340
    character(1,C_CHAR), value  :: uplo_a, uplo_c
    integer(kind=c_int), value  :: na, ncb, lda, ldb, nblk, mpi_comm_rows, mpi_comm_cols, ldc
    integer(kind=c_int)         :: success
    integer(kind=c_int), value  :: ldaCols, ldbCols, ldCcols
1341
#ifdef USE_ASSUMED_SIZE
1342
1343
1344
1345
1346
1347
    real(kind=c_float)          :: a(lda,*), b(ldb,*), c(ldc,*)
#else
    real(kind=c_float)          :: a(lda,ldaCols), b(ldb,ldbCols), c(ldc,ldcCols)
#endif
    logical                     :: successFortran

1348
    successFortran = elpa_mult_at_b_real_single(uplo_a, uplo_c, na, ncb, a, lda, ldaCols, b, ldb, ldbCols, &
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
                                               nblk, mpi_comm_rows, mpi_comm_cols, c, ldc, ldcCols)

    if (successFortran) then
      success = 1
    else
      success = 0
    endif

  end function

#endif /* WANT_SINGLE_PRECISION_REAL */

  !c> /*
  !c> \brief C interface for elpa_mult_ah_b_complex_double: Performs C : = A**H * B for double-precision matrices
  !c>         where   A is a square matrix (na,na) which is optionally upper or lower triangular
  !c>                 B is a (na,ncb) matrix
  !c>                 C is a (na,ncb) matrix where optionally only the upper or lower
  !c>                   triangle may be computed
  !c> \details
  !c>
  !c> \param  uplo_a               'U' if A is upper triangular
  !c>                              'L' if A is lower triangular
  !c>                              anything else if A is a full matrix
  !c>                              Please note: This pertains to the original A (as set in the calling program)
  !c>                                           whereas the transpose of A is used for calculations
  !c>                              If uplo_a is 'U' or 'L', the other triangle is not used at all,
  !c>                              i.e. it may contain arbitrary numbers
  !c> \param uplo_c                'U' if only the upper diagonal part of C is needed
  !c>                              'L' if only the upper diagonal part of C is needed
  !c>                              anything else if the full matrix C is needed
  !c>                              Please note: Even when uplo_c is 'U' or 'L', the other triangle may be
  !c>                                            written to a certain extent, i.e. one shouldn't rely on the content there!
  !c> \param na                    Number of rows/columns of A, number of rows of B and C
  !c> \param ncb                   Number of columns  of B and C
  !c> \param a                     matrix a
  !c> \param lda                   leading dimension of matrix a
  !c> \param b                     matrix b
  !c> \param ldb                   leading dimension of matrix b
  !c> \param nblk                  blocksize of cyclic distribution, must be the same in both directions!
  !c> \param  mpi_comm_rows        MPI communicator for rows
  !c> \param  mpi_comm_cols        MPI communicator for columns
  !c> \param c                     matrix c
  !c> \param ldc                   leading dimension of matrix c
  !c> \result success              int reports success (1) or failure (0)
  !c> */

  !c> int elpa_mult_ah_b_complex_double(char uplo_a, char uplo_c, int na, int ncb, double complex *a, int lda, int ldaCols, double complex *b, int ldb, int ldbCols, int nblk, int mpi_comm_rows, int mpi_comm_cols, double complex *c, int ldc, int ldcCols);
  function elpa_mult_ah_b_complex_wrapper_double( uplo_a, uplo_c, na, ncb, a, lda, ldaCols, b, ldb, ldbCols, &
                                                  nblk, mpi_comm_rows, &
                                                mpi_comm_cols, c, ldc, ldcCols) result(success) &
                                                bind(C,name="elpa_mult_ah_b_complex_double")
    use, intrinsic :: iso_c_binding
    use elpa1_auxiliary, only : elpa_mult_ah_b_complex_double

    implicit none

    character(1,C_CHAR), value     :: uplo_a, uplo_c
    integer(kind=c_int), value     :: na, ncb, lda, ldb, nblk, mpi_comm_rows, mpi_comm_cols, ldc
    integer(kind=c_int)            :: success
    integer(kind=c_int), value     :: ldaCols, ldbCols, ldcCols
1409
#ifdef USE_ASSUMED_SIZE
1410
    complex(kind=c_double_complex) :: a(lda,*), b(ldb,*), c(ldc,*)
1411
1412
1413
#else
    complex(kind=c_double_complex) :: a(lda,ldaCols), b(ldb,ldbCols), c(ldc,ldcCols)
#endif
1414
1415
    logical                        :: successFortran

1416
1417
    successFortran = elpa_mult_ah_b_complex_double(uplo_a, uplo_c, na, ncb, a, lda, ldaCols, b, ldb, ldbCols, nblk, &
                                                   mpi_comm_rows, mpi_comm_cols, c, ldc, ldcCols)
1418
1419
1420
1421
1422
1423
1424
1425
1426

    if (successFortran) then
      success = 1
    else
      success = 0
     endif

  end function

1427
1428
#ifdef WANT_SINGLE_PRECISION_COMPLEX

1429
  !c> /*
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
  !c> \brief C interface for elpa_mult_ah_b_complex_single: Performs C : = A**H * B for single-precision matrices
  !c>         where   A is a square matrix (na,na) which is optionally upper or lower triangular
  !c>                 B is a (na,ncb) matrix
  !c>                 C is a (na,ncb) matrix where optionally only the upper or lower
  !c>                   triangle may be computed
  !c> \details
  !c>
  !c> \param  uplo_a               'U' if A is upper triangular
  !c>                              'L' if A is lower triangular
  !c>                              anything else if A is a full matrix
  !c>                              Please note: This pertains to the original A (as set in the calling program)
  !c>                                           whereas the transpose of A is used for calculations
  !c>                              If uplo_a is 'U' or 'L', the other triangle is not used at all,
  !c>                              i.e. it may contain arbitrary numbers
  !c> \param uplo_c                'U' if only the upper diagonal part of C is needed
  !c>                              'L' if only the upper diagonal part of C is needed
  !c>                              anything else if the full matrix C is needed
  !c>                              Please note: Even when uplo_c is 'U' or 'L', the other triangle may be
  !c>                                            written to a certain extent, i.e. one shouldn't rely on the content there!
  !c> \param na                    Number of rows/columns of A, number of rows of B and C
  !c> \param ncb                   Number of columns  of B and C
  !c> \param a                     matrix a
  !c> \param lda                   leading dimension of matrix a
  !c> \param b                     matrix b
  !c> \param ldb                   leading dimension of matrix b
  !c> \param nblk                  blocksize of cyclic distribution, must be the same in both directions!
  !c> \param  mpi_comm_rows        MPI communicator for rows
  !c> \param  mpi_comm_cols        MPI communicator for columns
  !c> \param c                     matrix c
  !c> \param ldc                   leading dimension of matrix c
  !c> \result success              int reports success (1) or failure (0)
  !c> */

  !c> int elpa_mult_ah_b_complex_single(char uplo_a, char uplo_c, int na, int ncb, complex *a, int lda, int ldaCols, complex *b, int ldb, int ldbCols, int nblk, int mpi_comm_rows, int mpi_comm_cols, complex *c, int ldc, int ldcCols);
  function elpa_mult_ah_b_complex_wrapper_single( uplo_a, uplo_c, na, ncb, a, lda, ldaCols, b, ldb, ldbCols, &
                                                 nblk, mpi_comm_rows, mpi_comm_cols, c, ldc, ldcCols) &
    result(success) bind(C,name="elpa_mult_ah_b_complex_single")
    use, intrinsic :: iso_c_binding
    use elpa1_auxiliary, only : elpa_mult_ah_b_complex_single

    implicit none

    character(1,C_CHAR), value     :: uplo_a, uplo_c
    integer(kind=c_int), value     :: na, ncb, lda, ldb, nblk, mpi_comm_rows, mpi_comm_cols, ldc
    integer(kind=c_int)            :: success
    integer(kind=c_int), value     :: ldaCols, ldbCols, ldcCols