elpa_c_interface.F90 82.4 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
  !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
144
  !c> *  \param useGPU               use GPU (1=yes, 0=No)
145
146
147
  !c> *
  !c> *  \result                     int: 1 if error occured, otherwise 0
  !c>*/
148
149
150
151
152
#define REALCASE 1
#define DOUBLE_PRECISION 1
#include "precision_macros.h"

#if DOUBLE_PRECISION == 1
153
  !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, int mpi_comm_all, int useGPU);
154
#else
155
  !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, int mpi_comm_all, int useGPU);
156
#endif
157

158
159
160
#include "elpa1_c_interface_template.X90"
#undef REALCASE
#undef DOUBLE_PRECISION
161
162

#ifdef WANT_SINGLE_PRECISION_REAL
163

164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
  !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
183
  !c> *  \param useGPU               use GPU (1=yes, 0=No)
184
185
186
  !c> *
  !c> *  \result                     int: 1 if error occured, otherwise 0
  !c>*/
187
188
189
190
191
192
#define REALCASE 1
#undef DOUBLE_PRECISION
#define SINGLE_PRECISION 1
#include "precision_macros.h"

#if DOUBLE_PRECISION == 1
193
  !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, int mpi_comm_all, int useGPU);
194
#else
195
  !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, int mpi_comm_all, int useGPU);
196
197
#endif

198
199
200
#include "elpa1_c_interface_template.X90"
#undef SINGLE_PRECISION
#undef REALCASE
201
202
203
#endif /* WANT_SINGLE_PRECISION_REAL */

  !c> /*! \brief C interface to solve the double-precision complex eigenvalue problem with 1-stage solver
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
  !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
222
  !c> *  \param useGPU               use GPU (1=yes, 0=No)
223
224
225
  !c> *
  !c> *  \result                     int: 1 if error occured, otherwise 0
  !c> */
226

227
228
229
#define COMPLEXCASE 1
#define DOUBLE_PRECISION 1
#include "precision_macros.h"
Andreas Marek's avatar
Andreas Marek committed
230

231
232
#if DOUBLE_PRECISION == 1
  !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, int mpi_comm_all, int useGPU);
233
#else
234
  !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, int mpi_comm_all, int useGPU);
235
#endif
Andreas Marek's avatar
Andreas Marek committed
236

237
238
239
#include "elpa1_c_interface_template.X90"
#undef COMPLEXCASE
#undef DOUBLE_PRECISION
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261

#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
262
  !c> *  \param useGPU               use GPU (1=yes, 0=No)
263
264
265
  !c> *
  !c> *  \result                     int: 1 if error occured, otherwise 0
  !c> */
266
267
268
269
270
271
#define COMPLEXCASE 1
#undef DOUBLE_PRECISION
#define SINGLE_PRECISION
#include "precision_macros.h"

#if DOUBLE_PRECISION == 1
272
  !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, int mpi_comm_all, int useGPU);
273
#else
274
  !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, int mpi_comm_all, int useGPU);
275
276
#endif

277
#include "elpa1_c_interface_template.X90"
278

279
280
#undef SINGLE_PRECISION
#undef COMPLEXCASE
281
282
283
284
#endif /* WANT_SINGLE_PRECISION_COMPLEX */


  !c> /*! \brief C interface to solve the double-precision real eigenvalue problem with 2-stage solver
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
  !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
305
306
  !c> *  \param useQR                      use QR decomposition 1 = yes, 0 = no
  !c> *  \param useGPU                     use GPU (1=yes, 0=No)
307
308
309
  !c> *
  !c> *  \result                     int: 1 if error occured, otherwise 0
  !c> */
310
311
312
313
#define REALCASE 1
#define DOUBLE_PRECISION 1

#if DOUBLE_PRECISION == 1
314
  !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, int useGPU);
315
#else
316
  !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, int useGPU);
317
318
#endif

319
320
321
322
#include "precision_macros.h"
#include "elpa2_c_interface_template.X90"
#undef DOUBLE_PRECISION
#undef REALCASE
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346

#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
347
348
  !c> *  \param useQR                      use QR decomposition 1 = yes, 0 = no
  !c> *  \param useGPU                     use GPU (1=yes, 0=No)
349
350
351
  !c> *
  !c> *  \result                     int: 1 if error occured, otherwise 0
  !c> */
352
353
354
355
356
#define REALCASE 1
#define SINGLE_PRECISION 1
#undef DOUBLE_PRECISION

#if DOUBLE_PRECISION == 1
357
  !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, int useGPU);
358
#else
359
  !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, int useGPU);
360
361
#endif

362
363
364
365
366
#include "precision_macros.h"
#include "elpa2_c_interface_template.X90"
#undef DOUBLE_PRECISION
#undef SINGLE_PRECISION
#undef REALCASE
Andreas Marek's avatar
Andreas Marek committed
367

368
#endif /* WANT_SINGLE_PRECISION_REAL */
369

370
  !c> /*! \brief C interface to solve the double-precision complex eigenvalue problem with 2-stage solver
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
  !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
390
  !c> *  \param THIS_COMPLEX_ELPA_KERNEL_API  specify used ELPA2 kernel via API
391
  !c> *  \param useGPU                     use GPU (1=yes, 0=No)
392
393
394
  !c> *
  !c> *  \result                     int: 1 if error occured, otherwise 0
  !c> */
395

396
397
398
399
#define COMPLEXCASE 1
#define DOUBLE_PRECISION  1

#if DOUBLE_PRECISION == 1
400
  !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, int useGPU);
401
#else
402
  !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, int useGPU);
403
#endif
404

405
406
407
408
#include "precision_macros.h"
#include "elpa2_c_interface_template.X90"
#undef DOUBLE_PRECISION
#undef COMPLEXCASE
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432

#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
433
  !c> *  \param useGPU                     use GPU (1=yes, 0=No)
434
435
436
  !c> *
  !c> *  \result                     int: 1 if error occured, otherwise 0
  !c> */
Andreas Marek's avatar
Andreas Marek committed
437

438
439
440
441
#define COMPLEXCASE 1
#undef DOUBLE_PRECISION
#define SINGLE_PRECISION 1
#if DOUBLE_PRECISION == 1
442
  !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, int useGPU);
443
#else
444
  !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, int useGPU);
445
446
#endif

447
448
449
450
451
#include "precision_macros.h"
#include "elpa2_c_interface_template.X90"
#undef DOUBLE_PRECISION
#undef SINGLE_PRECISION
#undef COMPLEXCASE
Andreas Marek's avatar
Andreas Marek committed
452

453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
#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
476
477
478
479
  !c> *  \param useQR                      use QR decomposition 1 = yes, 0 = no
  !c> *  \param useGPU                     use GPU (1=yes, 0=No)
  !c> *  \param method                     choose whether to use ELPA 1stage or 2stage solver
  !c> *                                    possible values: "1stage" => use ELPA 1stage solver
480
481
482
483
484
  !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> */
485
  !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, int useGPU, char *method);
486
487
  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, &
488
                                  THIS_REAL_ELPA_KERNEL_API, useQR, useGPU, method)           &
489
490
491
492
493
494
495
496
497
                                  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
498
    integer(kind=c_int), value, intent(in)   :: THIS_REAL_ELPA_KERNEL_API, useQR, useGPU
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
    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,                                  &
528
                                           THIS_REAL_ELPA_KERNEL_API, useQRFortran, useGPU == 1, methodFortran)
529
530
531
    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,                                  &
532
                                           THIS_REAL_ELPA_KERNEL_API, useQRFortran, useGPU == 1)
533
534
535
536
537
538
539
540
541
    endif

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

  end function
Andreas Marek's avatar
Andreas Marek committed
542

543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
#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
565
566
567
568
  !c> *  \param useQR                      use QR decomposition 1 = yes, 0 = no
  !c> *  \param useGPU                     use GPU (1=yes, 0=No)
  !c> *  \param method                     choose whether to use ELPA 1stage or 2stage solver
  !c> *                                    possible values: "1stage" => use ELPA 1stage solver
569
570
571
572
573
  !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> */
574
  !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, int useGPU, char *method);
575
576
  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, &
577
                                  THIS_REAL_ELPA_KERNEL_API, useQR, useGPU, method)           &
578
579
580
581
582
583
584
585
586
                                  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
587
    integer(kind=c_int), value, intent(in)   :: THIS_REAL_ELPA_KERNEL_API, useQR, useGPU
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
    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,                                  &
617
                                           THIS_REAL_ELPA_KERNEL_API, useQRFortran, useGPU == 1, methodFortran)
618
619
620
    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,                                  &
621
                                           THIS_REAL_ELPA_KERNEL_API, useQRFortran, useGPU == 1)
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
    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
654
  !c> *  \param useGPU                        use GPU (1=yes, 0=No)
655
656
657
658
659
660
661
  !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> */
662
  !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, int useGPU, char *method);
663
664
  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,    &
665
                                  THIS_COMPLEX_ELPA_KERNEL_API, useGPU, method)                  &
666
667
668
669
670
671
672
673
674
                                  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
675
    integer(kind=c_int), value, intent(in)   :: THIS_COMPLEX_ELPA_KERNEL_API, useGPU
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
#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, &
699
                                              mpi_comm_all, THIS_COMPLEX_ELPA_KERNEL_API, useGPU == 1, methodFortran)
700
701
    else
      successFortran = elpa_solve_evp_complex_double(na, nev, a, lda, ev, q, ldq, nblk, matrixCols, mpi_comm_rows, mpi_comm_cols, &
702
                                              mpi_comm_all, THIS_COMPLEX_ELPA_KERNEL_API, useGPU == 1)
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
    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
735
  !c> *  \param useGPU                        use GPU (1=yes, 0=No)
736
737
738
739
740
741
742
  !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> */
743
  !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, int useGPU, char *method);
744
745
  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,    &
746
                                  THIS_COMPLEX_ELPA_KERNEL_API, useGPU, method)                  &
747
748
749
750
751
752
753
754
755
                                  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
756
    integer(kind=c_int), value, intent(in)   :: THIS_COMPLEX_ELPA_KERNEL_API, useGPU
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
#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, &
780
                                              mpi_comm_all, THIS_COMPLEX_ELPA_KERNEL_API, useGPU == 1, methodFortran)
781
782
    else
      successFortran = elpa_solve_evp_complex_single(na, nev, a, lda, ev, q, ldq, nblk, matrixCols, mpi_comm_rows, mpi_comm_cols, &
783
                                              mpi_comm_all, THIS_COMPLEX_ELPA_KERNEL_API, useGPU ==1)
784
785
786
787
788
789
790
791
792
    endif

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

  end function
793
794
#endif /* WANT_SINGLE_PRECISION_COMPLEX */

795
  !c> /*
796
  !c> \brief  C interface to solve double-precision tridiagonal eigensystem with divide and conquer method
797
798
  !c> \details
  !c>
Andreas Marek's avatar
Andreas Marek committed
799
800
801
802
803
804
805
806
807
808
809
810
811
  !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
812
  !c> */
813
814
815
  !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")
816
817

    use, intrinsic :: iso_c_binding
818
    use elpa1_auxiliary, only : elpa_solve_tridi_double
819
820
821
822
823

    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
824
    real(kind=c_double)                    :: d(1:na), e(1:na)
825
#ifdef USE_ASSUMED_SIZE
826
827
828
829
    real(kind=c_double)                    :: q(ldq,*)
#else
    real(kind=c_double)                    :: q(1:ldq, 1:matrixCols)
#endif
830
831
832
833
834
835
836
837
    logical                                :: successFortran, wantDebugFortran

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

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
    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

889
890
    successFortran = elpa_solve_tridi_single(na, nev, d, e, q, ldq, nblk, matrixCols, mpi_comm_rows, &
                                             mpi_comm_cols, wantDebugFortran)
891
892
893
894
895
896
897
898
899

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

  end function

900
901
#endif /* WANT_SINGLE_PRECISION_REAL */

902
  !c> /*
903
  !c> \brief  C interface for elpa_mult_at_b_real_double: Performs C : = A**T * B for double-precision matrices
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
  !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
925
  !c> \param ldaCols               columns of matrix a
926
927
  !c> \param b                     matrix b
  !c> \param ldb                   leading dimension of matrix b
928
  !c> \param ldbCols               columns of matrix b
929
930
931
932
933
  !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
934
  !c> \param ldcCols               columns of matrix c
935
936
937
  !c> \result success              int report success (1) or failure (0)
  !c> */

938
939
940
  !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) &
941
                                              bind(C,name="elpa_mult_at_b_real_double") result(success)
942
    use, intrinsic :: iso_c_binding
943
    use elpa1_auxiliary, only : elpa_mult_at_b_real_double
944
945
946
947

    implicit none

    character(1,C_CHAR), value  :: uplo_a, uplo_c
948
949
    integer(kind=c_int), value  :: na, ncb, lda, ldb, nblk, mpi_comm_rows, mpi_comm_cols, ldc, &
                                   ldaCols, ldbCols, ldcCols
950
    integer(kind=c_int)         :: success
951
#ifdef USE_ASSUMED_SIZE
952
    real(kind=c_double)         :: a(lda,*), b(ldb,*), c(ldc,*)
953
954
955
#else
    real(kind=c_double)         :: a(lda,ldaCols), b(ldb,ldbCols), c(ldc,ldcCols)
#endif
956
957
    logical                     :: successFortran

958
959
    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)
960
961
962
963
964
965
966
967
968

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

  end function

969
#ifdef WANT_SINGLE_PRECISION_REAL
970
  !c> /*
971
  !c> \brief  C interface for elpa_mult_at_b_real_single: Performs C : = A**T * B for single-precision matrices
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
  !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
993
  !c> \param ldaCols               columns of matrix a
994
995
  !c> \param b                     matrix b
  !c> \param ldb                   leading dimension of matrix b
996
  !c> \param ldbCols               columns of matrix b
997
998
999
1000
1001
  !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
1002
  !c> \result success              int report success (1) or failure (0)
1003
1004
  !c> */

1005
1006
1007
1008
  !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)
1009
    use, intrinsic :: iso_c_binding
1010
    use elpa1_auxiliary, only : elpa_mult_at_b_real_single
1011
1012
1013

    implicit none

1014
1015
1016
1017
    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
1018
#ifdef USE_ASSUMED_SIZE
1019
1020
1021
1022
1023
1024
    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

1025
    successFortran = elpa_mult_at_b_real_single(uplo_a, uplo_c, na, ncb, a, lda, ldaCols, b, ldb, ldbCols, &
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
                                               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
1086
#ifdef USE_ASSUMED_SIZE
1087
    complex(kind=c_double_complex) :: a(lda,*), b(ldb,*), c(ldc,*)
1088
1089
1090
#else
    complex(kind=c_double_complex) :: a(lda,ldaCols), b(ldb,ldbCols), c(ldc,ldcCols)
#endif
1091
1092
    logical                        :: successFortran

1093
1094
    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)
1095
1096
1097
1098
1099
1100
1101
1102
1103

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

  end function

1104
1105
#ifdef WANT_SINGLE_PRECISION_COMPLEX

1106
  !c> /*
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
  !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
1153
#ifdef USE_ASSUMED_SIZE
1154
1155
1156
1157
1158
1159
    complex(kind=c_float_complex)  :: a(lda,*), b(ldb,*), c(ldc,*)
#else
    complex(kind=c_float_complex)  :: a(lda,ldaCols), b(ldb,ldbCols), c(ldc,ldcCols)
#endif
    logical                        :: successFortran

1160
    successFortran = elpa_mult_ah_b_complex_single(uplo_a, uplo_c, na, ncb, a, lda, ldaCols, b, ldb, ldbCols, &
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
                                                  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_COMPLEX */

  !c> /*
  !c> \brief  C interface to elpa_invert_trm_real_double: Inverts a real double-precision upper triangular matrix
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
  !c> \details
  !c> \param  na                   Order of matrix
  !c> \param  a(lda,matrixCols)    Distributed matrix which should be inverted
  !c>                              Distribution is like in Scalapack.
  !c>                              Only upper triangle is needs to be set.
  !c>                              The lower triangle is not referenced.
  !c> \param  lda                  Leading dimension of a
  !c> \param                       matrixCols  local columns of matrix a
  !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 wantDebug             int more debug information on failure if 1, else 0
  !c> \result succes               int reports success (1) or failure (0)
  !c> */

1190
1191
1192
  !c> int elpa_invert_trm_real_double(int na, double *a, int lda, int nblk, int matrixCols, int mpi_comm_rows, int mpi_comm_cols, int wantDebug);
  function elpa_invert_trm_real_wrapper_double(na, a, lda, nblk, matrixCols, mpi_comm_rows, mpi_comm_cols, wantDebug) &
        result(success) bind(C,name="elpa_invert_trm_real_double")
1193
   use, intrinsic :: iso_c_binding
1194
   use elpa1_auxiliary, only : elpa_invert_trm_real_double
1195
1196
1197
1198
1199
1200

   implicit none

   integer(kind=c_int), value  :: na, lda, nblk, matrixCols, mpi_comm_rows, mpi_comm_cols
   integer(kind=c_int), value  :: wantDebug
   integer(kind=c_int)         :: success
1201
#ifdef USE_ASSUMED_SIZE
1202
1203
   real(kind=c_double)         :: a(lda,*)
#else
1204
   real(kind=c_double)         :: a(lda,matrixCols)
1205
#endif
1206
1207
1208
1209
1210
1211
1212
1213
   logical                     :: wantDebugFortran, successFortran

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

1214
   successFortran = elpa_invert_trm_real_double(na, a, lda, nblk, matrixCols, mpi_comm_rows, mpi_comm_cols, wantDebugFortran)
1215
1216
1217
1218
1219
1220
1221
1222
1223

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

 end function

1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
#ifdef WANT_SINGLE_PRECISION_REAL

  !c> /*
  !c> \brief  C interface to elpa_invert_trm_real_single: Inverts a real single-precision upper triangular matrix
  !c> \details
  !c> \param  na                   Order of matrix
  !c> \param  a(lda,matrixCols)    Distributed matrix which should be inverted
  !c>                              Distribution is like in Scalapack.
  !c>                              Only upper triangle is needs to be set.
  !c>                              The lower triangle is not referenced.
  !c> \param  lda                  Leading dimension of a
  !c> \param                       matrixCols  local columns of matrix a
  !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 wantDebug             int more debug information on failure if 1, else 0
  !c> \result succes               int reports success (1) or failure (0)
  !c> */

  !c> int elpa_invert_trm_real_single(int na, double *a, int lda, int nblk, int matrixCols, int mpi_comm_rows, int mpi_comm_cols, int wantDebug);
  function elpa_invert_trm_real_wrapper_single(na, a, lda, nblk, matrixCols, mpi_comm_rows, mpi_comm_cols, wantDebug) &
        result(success) bind(C,name="elpa_invert_trm_real_single")
   use, intrinsic :: iso_c_binding
   use elpa1_auxiliary, only : elpa_invert_trm_real_single

   implicit none

   integer(kind=c_int), value  :: na, lda, nblk, matrixCols, mpi_comm_rows, mpi_comm_cols
   integer(kind=c_int), value  :: wantDebug
   integer(kind=c_int)         :: success
   real(kind=c_float)          :: a(lda,matrixCols)

   logical                     :: wantDebugFortran, successFortran

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

   successFortran = elpa_invert_trm_real_single(na, a, lda, nblk, matrixCols, mpi_comm_rows, mpi_comm_cols, wantDebugFortran)

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

 end function

#endif /* WANT_SINGLE_PRECISION_REAL */

1276
 !c> /*
1277
 !c> \brief  C interface to elpa_invert_trm_complex_double: Inverts a double-precision complex upper triangular matrix
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
 !c> \details
 !c> \param  na                   Order of matrix
 !c> \param  a(lda,matrixCols)    Distributed matrix which should be inverted
 !c>                              Distribution is like in Scalapack.
 !c>                              Only upper triangle is needs to be set.
 !c>                              The lower triangle is not referenced.
 !c> \param  lda                  Leading dimension of a
 !c> \param                       matrixCols  local columns of matrix a
 !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 wantDebug             int more debug information on failure if 1, else 0
 !c> \result succes               int reports success (1) or failure (0)
 !c> */

1293
1294
1295
1296
 !c> int elpa_invert_trm_complex_double(int na, double complex *a, int lda, int nblk, int matrixCols, int mpi_comm_rows, int mpi_comm_cols, int wantDebug);
 function elpa_invert_trm_complex_wrapper_double(na, a, lda, nblk, matrixCols, mpi_comm_rows, &
                                                 mpi_comm_cols, wantDebug) result(success) &
   bind(C,name="elpa_invert_trm_complex_double")
1297
1298

   use, intrinsic :: iso_c_binding
1299
   use elpa1_auxiliary, only : elpa_invert_trm_complex_double
1300
1301
1302
1303
1304
1305

   implicit none

   integer(kind=c_int), value     :: na, lda, nblk, matrixCols, mpi_comm_rows, mpi_comm_cols
   integer(kind=c_int), value     :: wantDebug
   integer(kind=c_int)            :: success
1306
#ifdef USE_ASSUMED_SIZE
1307
1308
   complex(kind=c_double_complex) :: a(lda, *)
#else
1309
   complex(kind=c_double_complex) :: a(lda, matrixCols)
1310
#endif
1311
1312
1313
1314
1315
1316
1317
1318
1319
   logical                        :: successFortran, wantDebugFortran


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

1320
1321
   successFortran = elpa_invert_trm_complex_double(na, a, lda, nblk, matrixCols, mpi_comm_rows, &
                                                   mpi_comm_cols, wantDebugFortran)
1322
1323
1324
1325
1326
1327
1328
1329

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

1330
#ifdef WANT_SINGLE_PRECISION_COMPLEX
1331
 !c> /*
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
 !c> \brief  C interface to elpa_invert_trm_complex_single: Inverts a single-precision complex upper triangular matrix
 !c> \details
 !c> \param  na                   Order of matrix
 !c> \param  a(lda,matrixCols)    Distributed matrix which should be inverted
 !c>                              Distribution is like in Scalapack.
 !c>                              Only upper triangle is needs to be set.
 !c>                              The lower triangle is not referenced.
 !c> \param  lda                  Leading dimension of a
 !c> \param                       matrixCols  local columns of matrix a
 !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 wantDebug             int more debug information on failure if 1, else 0
 !c> \result succes               int reports success (1) or failure (0)
 !c> */

 !c> int elpa_invert_trm_complex_single(int na, complex *a, int lda, int nblk, int matrixCols, int mpi_comm_rows, int mpi_comm_cols, int wantDebug);
1349
1350
 function elpa_invert_trm_complex_wrapper_single(na, a, lda, nblk, matrixCols, mpi_comm_rows, &
                                                 mpi_comm_cols, wantDebug) result(success) &
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
   bind(C,name="elpa_invert_trm_complex_single")

   use, intrinsic :: iso_c_binding
   use elpa1_auxiliary, only : elpa_invert_trm_complex_single

   implicit none

   integer(kind=c_int), value     :: na, lda, nblk, matrixCols, mpi_comm_rows, mpi_comm_cols
   integer(kind=c_int), value     :: wantDebug
   integer(kind=c_int)            :: success
   complex(kind=c_float_complex)  :: a(lda, matrixCols)

   logical                        :: successFortran, wantDebugFortran


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

   successFortran = elpa_invert_trm_complex_single(na, a, lda, nblk, matrixCols, mpi_comm_rows, mpi_comm_cols, wantDebugFortran)

   if (successFortran) then
     success = 1
   else
     success = 0
   endif
 end function
1380
1381
1382

#endif /* WANT_SINGLE_PRECISION_COMPLEX */

1383
1384
 !c> /*
 !c> \brief  elpa_cholesky_real_double: Cholesky factorization of a double-precision real symmetric matrix
1385
1386
 !c> \details
 !c>
Andreas Marek's avatar
Andreas Marek committed
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
 !c> *\param  na                   Order of matrix
 !c> *\param  a(lda,matrixCols)    Distributed matrix which should be factorized.
 !c> *                             Distribution is like in Scalapack.
 !c> *                             Only upper triangle is needs to be set.
 !c> *                             On return, the upper triangle contains the Cholesky factor
 !c> *                             and the lower triangle is set to 0.
 !c> *\param  lda                  Leading dimension of a
 !c> *\param  matrixCols           local columns of matrix a
 !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 wantDebug             int more debug information on failure if 1, else 0
 !c> *\result succes               int reports success (1) or failure (0)
1400
1401
 !c> */

1402
1403
1404
 !c> int elpa_cholesky_real_double(int na, double *a, int lda, int nblk, int matrixCols, int mpi_comm_rows, int mpi_comm_cols, int wantDebug);
 function elpa_cholesky_real_wrapper_double(na, a, lda, nblk, matrixCols, mpi_comm_rows, mpi_comm_cols, wantDebug) result(success) &
       bind(C,name="elpa_cholesky_real_double")
1405
1406

   use, intrinsic :: iso_c_binding
1407
   use elpa1_auxiliary, only : elpa_cholesky_real_double
1408
1409
1410
1411
1412

   implicit none

   integer(kind=c_int), value :: na, lda, nblk, matrixCols,  mpi_comm_rows, mpi_comm_cols, wantDebug
   integer(kind=c_int)        :: success
1413
#ifdef USE_ASSUMED_SIZE
1414
1415
   real(kind=c_double)        :: a(lda,*)
#else
1416
   real(kind=c_double)        :: a(lda,matrixCols)
1417
#endif
1418
1419
1420
1421
1422
1423
1424
1425
   logical                    :: successFortran, wantDebugFortran

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

1426
1427
1428
1429
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 </