zgels_param.c 15.8 KB
Newer Older
1
/**
2 3
 *
 * @file zgels_param.c
4
 *
Mathieu Faverge's avatar
Mathieu Faverge committed
5 6
 * @copyright 2009-2014 The University of Tennessee and The University of
 *                      Tennessee Research Foundation. All rights reserved.
7 8
 * @copyright 2012-2017 Bordeaux INP, CNRS (LaBRI UMR 5800), Inria,
 *                      Univ. Bordeaux. All rights reserved.
9
 *
Mathieu Faverge's avatar
Mathieu Faverge committed
10
 ***
11 12 13 14 15 16
 *
 *
 *  MORSE computational routines
 *  MORSE is a software package provided by Univ. of Tennessee,
 *  Univ. of California Berkeley and Univ. of Colorado Denver
 *
BOUCHERIE Raphael's avatar
BOUCHERIE Raphael committed
17 18
 * @version 1.0.0
 * @author Raphael Boucherie
19
 * @author Mathieu Faverge
BOUCHERIE Raphael's avatar
BOUCHERIE Raphael committed
20
 * @date 2017-05-17
21 22 23 24
 * @precisions normal z -> s d c
 *
 **/
#include "control/common.h"
Mathieu Faverge's avatar
Mathieu Faverge committed
25
#include <stdlib.h>
26

27 28
/**
 *******************************************************************************
29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47
 *
 * @ingroup MORSE_Complex64_t
 *
 *  MORSE_zgels_param - solves overdetermined or underdetermined linear systems involving an M-by-N
 *  matrix A using the QR or the LQ factorization of A.  It is assumed that A has full rank.
 *  The following options are provided:
 *
 *  # trans = MorseNoTrans and M >= N: find the least squares solution of an overdetermined
 *    system, i.e., solve the least squares problem: minimize || B - A*X ||.
 *
 *  # trans = MorseNoTrans and M < N:  find the minimum norm solution of an underdetermined
 *    system A * X = B.
 *
 *  Several right hand side vectors B and solution vectors X can be handled in a single call;
 *  they are stored as the columns of the M-by-NRHS right hand side matrix B and the N-by-NRHS
 *  solution matrix X.
 *
 *******************************************************************************
 *
BOUCHERIE Raphael's avatar
BOUCHERIE Raphael committed
48 49 50
 * @param[in] qrtree
 *          The tree used for the factorization
 *
51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77
 * @param[in] trans
 *          Intended usage:
 *          = MorseNoTrans:   the linear system involves A;
 *          = MorseConjTrans: the linear system involves A**H.
 *          Currently only MorseNoTrans is supported.
 *
 * @param[in] M
 *          The number of rows of the matrix A. M >= 0.
 *
 * @param[in] N
 *          The number of columns of the matrix A. N >= 0.
 *
 * @param[in] NRHS
 *          The number of right hand sides, i.e., the number of columns of the matrices B and X.
 *          NRHS >= 0.
 *
 * @param[in,out] A
 *          On entry, the M-by-N matrix A.
 *          On exit,
 *          if M >= N, A is overwritten by details of its QR factorization as returned by
 *                     MORSE_zgeqrf;
 *          if M < N, A is overwritten by details of its LQ factorization as returned by
 *                      MORSE_zgelqf.
 *
 * @param[in] LDA
 *          The leading dimension of the array A. LDA >= max(1,M).
 *
BOUCHERIE Raphael's avatar
BOUCHERIE Raphael committed
78 79 80 81
 * @param[out] descTS
 *          On exit, auxiliary factorization data.
 *
 * @param[out] descTT
82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110
 *          On exit, auxiliary factorization data.
 *
 * @param[in,out] B
 *          On entry, the M-by-NRHS matrix B of right hand side vectors, stored columnwise;
 *          On exit, if return value = 0, B is overwritten by the solution vectors, stored
 *          columnwise:
 *          if M >= N, rows 1 to N of B contain the least squares solution vectors; the residual
 *          sum of squares for the solution in each column is given by the sum of squares of the
 *          modulus of elements N+1 to M in that column;
 *          if M < N, rows 1 to N of B contain the minimum norm solution vectors;
 *
 * @param[in] LDB
 *          The leading dimension of the array B. LDB >= MAX(1,M,N).
 *
 *******************************************************************************
 *
 * @return
 *          \retval MORSE_SUCCESS successful exit
 *          \retval <0 if -i, the i-th argument had an illegal value
 *
 *******************************************************************************
 *
 * @sa MORSE_zgels_param_Tile
 * @sa MORSE_zgels_param_Tile_Async
 * @sa MORSE_cgels
 * @sa MORSE_dgels
 * @sa MORSE_sgels
 *
 ******************************************************************************/
111 112 113 114
int MORSE_zgels_param( const libhqr_tree_t *qrtree, MORSE_enum trans, int M, int N, int NRHS,
                       MORSE_Complex64_t *A, int LDA,
                       MORSE_desc_t *descTS, MORSE_desc_t *descTT,
                       MORSE_Complex64_t *B, int LDB )
115 116 117 118 119 120 121
{
    int i, j;
    int NB;
    int status;
    MORSE_context_t *morse;
    MORSE_sequence_t *sequence = NULL;
    MORSE_request_t request = MORSE_REQUEST_INITIALIZER;
122 123
    MORSE_desc_t descAl, descAt;
    MORSE_desc_t descBl, descBt;
124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173

    morse = morse_context_self();
    if (morse == NULL) {
        morse_fatal_error("MORSE_zgels_param", "MORSE not initialized");
        return MORSE_ERR_NOT_INITIALIZED;
    }

    /* Check input arguments */
    if (trans != MorseNoTrans) {
        morse_error("MORSE_zgels_param", "only MorseNoTrans supported");
        return MORSE_ERR_NOT_SUPPORTED;
    }
    if (M < 0) {
        morse_error("MORSE_zgels_param", "illegal value of M");
        return -2;
    }
    if (N < 0) {
        morse_error("MORSE_zgels_param", "illegal value of N");
        return -3;
    }
    if (NRHS < 0) {
        morse_error("MORSE_zgels_param", "illegal value of NRHS");
        return -4;
    }
    if (LDA < chameleon_max(1, M)) {
        morse_error("MORSE_zgels_param", "illegal value of LDA");
        return -6;
    }
    if (LDB < chameleon_max(1, chameleon_max(M, N))) {
        morse_error("MORSE_zgels_param", "illegal value of LDB");
        return -9;
    }
    /* Quick return */
    if (chameleon_min(M, chameleon_min(N, NRHS)) == 0) {
        for (i = 0; i < chameleon_max(M, N); i++)
            for (j = 0; j < NRHS; j++)
                B[j*LDB+i] = 0.0;
        return MORSE_SUCCESS;
    }

    /* Tune NB & IB depending on M, N & NRHS; Set NBNB */
    status = morse_tune(MORSE_FUNC_ZGELS, M, N, NRHS);
    if (status != MORSE_SUCCESS) {
        morse_error("MORSE_zgels_param", "morse_tune() failed");
        return status;
    }

    /* Set NT */
    NB = MORSE_NB;

174
    morse_sequence_create( morse, &sequence );
175

176
    /* Submit the matrix conversion */
177
    if ( M >= N ) {
178
        morse_zlap2tile( morse, &descAl, &descAt, MorseDescInout, MorseUpperLower,
179
                         A, NB, NB, LDA, N, M, N, sequence, &request );
180
        morse_zlap2tile( morse, &descBl, &descBt, MorseDescInout, MorseUpperLower,
181
                         B, NB, NB, LDB, NRHS, M, NRHS, sequence, &request );
182
    } else {
183
        morse_zlap2tile( morse, &descAl, &descAt, MorseDescInout, MorseUpperLower,
184
                         A, NB, NB, LDA, N, M, N, sequence, &request );
185
        morse_zlap2tile( morse, &descBl, &descBt, MorseDescInout, MorseUpperLower,
186
                         B, NB, NB, LDB, NRHS, N, NRHS, sequence, &request );
187 188 189
    }

    /* Call the tile interface */
190
    MORSE_zgels_param_Tile_Async( qrtree, MorseNoTrans, &descAt, descTS, descTT, &descBt, sequence, &request );
191

Mathieu Faverge's avatar
Mathieu Faverge committed
192
    /* Submit the matrix conversion back */
Mathieu Faverge's avatar
Mathieu Faverge committed
193
    morse_ztile2lap( morse, &descAl, &descAt,
194
                     MorseDescInout, MorseUpperLower, sequence, &request );
Mathieu Faverge's avatar
Mathieu Faverge committed
195
    morse_ztile2lap( morse, &descBl, &descBt,
196
                     MorseDescInout, MorseUpperLower, sequence, &request );
197 198
    MORSE_Desc_Flush( descTS, sequence );
    MORSE_Desc_Flush( descTT, sequence );
Mathieu Faverge's avatar
Mathieu Faverge committed
199

200
    morse_sequence_wait( morse, sequence );
Mathieu Faverge's avatar
Mathieu Faverge committed
201

Mathieu Faverge's avatar
Mathieu Faverge committed
202 203 204
    /* Cleanup the temporary data */
    morse_ztile2lap_cleanup( morse, &descAl, &descAt );
    morse_ztile2lap_cleanup( morse, &descBl, &descBt );
205 206

    status = sequence->status;
207
    morse_sequence_destroy( morse, sequence );
208 209 210
    return status;
}

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
 *
 * @ingroup MORSE_Complex64_t_Tile
 *
 *  MORSE_zgels_param_Tile - Solves overdetermined or underdetermined linear system of equations
 *  using the tile QR or the tile LQ factorization.
 *  Tile equivalent of MORSE_zgels_param().
 *  Operates on matrices stored by tiles.
 *  All matrices are passed through descriptors.
 *  All dimensions are taken from the descriptors.
 *
 *******************************************************************************
 *
 * @param[in] trans
 *          Intended usage:
 *          = MorseNoTrans:   the linear system involves A;
 *          = MorseConjTrans: the linear system involves A**H.
 *          Currently only MorseNoTrans is supported.
 *
 * @param[in,out] A
 *          On entry, the M-by-N matrix A.
 *          On exit,
 *          if M >= N, A is overwritten by details of its QR factorization as returned by
 *                     MORSE_zgeqrf;
 *          if M < N, A is overwritten by details of its LQ factorization as returned by
 *                      MORSE_zgelqf.
 *
BOUCHERIE Raphael's avatar
BOUCHERIE Raphael committed
239 240 241 242
 * @param[out] TS
 *          On exit, auxiliary factorization data.
 *
 * @param[out] TT
243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267
 *          On exit, auxiliary factorization data.
 *
 * @param[in,out] B
 *          On entry, the M-by-NRHS matrix B of right hand side vectors, stored columnwise;
 *          On exit, if return value = 0, B is overwritten by the solution vectors, stored
 *          columnwise:
 *          if M >= N, rows 1 to N of B contain the least squares solution vectors; the residual
 *          sum of squares for the solution in each column is given by the sum of squares of the
 *          modulus of elements N+1 to M in that column;
 *          if M < N, rows 1 to N of B contain the minimum norm solution vectors;
 *
 *******************************************************************************
 *
 * @return
 *          \return MORSE_SUCCESS successful exit
 *
 *******************************************************************************
 *
 * @sa MORSE_zgels_param
 * @sa MORSE_zgels_param_Tile_Async
 * @sa MORSE_cgels_Tile
 * @sa MORSE_dgels_Tile
 * @sa MORSE_sgels_Tile
 *
 ******************************************************************************/
268 269
int MORSE_zgels_param_Tile( const libhqr_tree_t *qrtree, MORSE_enum trans, MORSE_desc_t *A,
                            MORSE_desc_t *TS, MORSE_desc_t *TT, MORSE_desc_t *B )
270 271 272 273 274 275 276 277 278 279 280
{
    MORSE_context_t *morse;
    MORSE_sequence_t *sequence = NULL;
    MORSE_request_t request = MORSE_REQUEST_INITIALIZER;
    int status;

    morse = morse_context_self();
    if (morse == NULL) {
        morse_fatal_error("MORSE_zgels_param_Tile", "MORSE not initialized");
        return MORSE_ERR_NOT_INITIALIZED;
    }
281 282
    morse_sequence_create( morse, &sequence );

283
    MORSE_zgels_param_Tile_Async( qrtree, trans, A, TS, TT, B, sequence, &request );
284

285
    MORSE_Desc_Flush( A, sequence );
286 287
    MORSE_Desc_Flush( TS, sequence );
    MORSE_Desc_Flush( TT, sequence );
288
    MORSE_Desc_Flush( B, sequence );
Mathieu Faverge's avatar
Mathieu Faverge committed
289

290
    morse_sequence_wait( morse, sequence );
291
    status = sequence->status;
292
    morse_sequence_destroy( morse, sequence );
293 294 295
    return status;
}

296 297
/**
 *******************************************************************************
298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324
 *
 * @ingroup MORSE_Complex64_t_Tile_Async
 *
 *  MORSE_zgels_param_Tile_Async - Solves overdetermined or underdetermined linear
 *  system of equations using the tile QR or the tile LQ factorization.
 *  Non-blocking equivalent of MORSE_zgels_param_Tile().
 *  May return before the computation is finished.
 *  Allows for pipelining of operations at runtime.
 *
 *******************************************************************************
 *
 * @param[in] sequence
 *          Identifies the sequence of function calls that this call belongs to
 *          (for completion checks and exception handling purposes).
 *
 * @param[out] request
 *          Identifies this function call (for exception handling purposes).
 *
 *******************************************************************************
 *
 * @sa MORSE_zgels_param
 * @sa MORSE_zgels_param_Tile
 * @sa MORSE_cgels_Tile_Async
 * @sa MORSE_dgels_Tile_Async
 * @sa MORSE_sgels_Tile_Async
 *
 ******************************************************************************/
325 326 327
int MORSE_zgels_param_Tile_Async( const libhqr_tree_t *qrtree, MORSE_enum trans, MORSE_desc_t *A,
                                  MORSE_desc_t *TS, MORSE_desc_t *TT, MORSE_desc_t *B,
                                  MORSE_sequence_t *sequence, MORSE_request_t *request )
328 329 330 331
{
    MORSE_desc_t *subA;
    MORSE_desc_t *subB;
    MORSE_context_t *morse;
Mathieu Faverge's avatar
Mathieu Faverge committed
332
    MORSE_desc_t D, *Dptr = NULL;
333

334 335 336 337 338 339 340 341 342 343 344 345 346 347
    morse = morse_context_self();
    if (morse == NULL) {
        morse_fatal_error("MORSE_zgels_param_Tile", "MORSE not initialized");
        return MORSE_ERR_NOT_INITIALIZED;
    }
    if (sequence == NULL) {
        morse_fatal_error("MORSE_zgels_param_Tile", "NULL sequence");
        return MORSE_ERR_UNALLOCATED;
    }
    if (request == NULL) {
        morse_fatal_error("MORSE_zgels_param_Tile", "NULL request");
        return MORSE_ERR_UNALLOCATED;
    }
    /* Check sequence status */
Mathieu Faverge's avatar
Mathieu Faverge committed
348
    if (sequence->status == MORSE_SUCCESS) {
349
        request->status = MORSE_SUCCESS;
Mathieu Faverge's avatar
Mathieu Faverge committed
350 351
    }
    else {
352
        return morse_request_fail(sequence, request, MORSE_ERR_SEQUENCE_FLUSHED);
Mathieu Faverge's avatar
Mathieu Faverge committed
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

    /* Check descriptors for correctness */
    if (morse_desc_check(A) != MORSE_SUCCESS) {
        morse_error("MORSE_zgels_param_Tile", "invalid first descriptor");
        return morse_request_fail(sequence, request, MORSE_ERR_ILLEGAL_VALUE);
    }
    if (morse_desc_check(TS) != MORSE_SUCCESS) {
        morse_error("MORSE_zgels_param_Tile", "invalid second descriptor");
        return morse_request_fail(sequence, request, MORSE_ERR_ILLEGAL_VALUE);
    }
    if (morse_desc_check(TT) != MORSE_SUCCESS) {
        morse_error("MORSE_zgels_param_Tile", "invalid third descriptor");
        return morse_request_fail(sequence, request, MORSE_ERR_ILLEGAL_VALUE);
    }
    if (morse_desc_check(B) != MORSE_SUCCESS) {
        morse_error("MORSE_zgels_param_Tile", "invalid fourth descriptor");
        return morse_request_fail(sequence, request, MORSE_ERR_ILLEGAL_VALUE);
    }
    /* Check input arguments */
    if (A->nb != A->mb || B->nb != B->mb) {
        morse_error("MORSE_zgels_param_Tile", "only square tiles supported");
        return morse_request_fail(sequence, request, MORSE_ERR_ILLEGAL_VALUE);
    }
    if (trans != MorseNoTrans) {
        morse_error("MORSE_zgels_param_Tile", "only MorseNoTrans supported");
        return morse_request_fail(sequence, request, MORSE_ERR_NOT_SUPPORTED);
    }
    /* Quick return  - currently NOT equivalent to LAPACK's:
Mathieu Faverge's avatar
Mathieu Faverge committed
382 383 384 385 386 387
     if (chameleon_min(M, chameleon_min(N, NRHS)) == 0) {
     for (i = 0; i < chameleon_max(M, N); i++)
     for (j = 0; j < NRHS; j++)
     B[j*LDB+i] = 0.0;
     return MORSE_SUCCESS;
     }
388 389 390
     */
    if (A->m >= A->n) {

BOUCHERIE Raphael's avatar
BOUCHERIE Raphael committed
391
#if defined(CHAMELEON_COPY_DIAG)
Mathieu Faverge's avatar
Mathieu Faverge committed
392 393 394 395 396
        {
            int n = chameleon_min(A->mt, A->nt) * A->nb;
            morse_zdesc_alloc(D, A->mb, A->nb, A->m, n, 0, 0, A->m, n, );
            Dptr = &D;
        }
BOUCHERIE Raphael's avatar
BOUCHERIE Raphael committed
397
#endif
398

BOUCHERIE Raphael's avatar
BOUCHERIE Raphael committed
399 400
        subB = morse_desc_submatrix(B, 0, 0, A->n, B->n);
        subA = morse_desc_submatrix(A, 0, 0, A->n, A->n);
Mathieu Faverge's avatar
Mathieu Faverge committed
401

Mathieu Faverge's avatar
Mathieu Faverge committed
402 403 404
        morse_pzgeqrf_param( qrtree, A, TS, TT, Dptr, sequence, request );
        morse_pzunmqr_param( qrtree, MorseLeft, MorseConjTrans, A, B, TS, TT, Dptr, sequence, request );
        morse_pztrsm( MorseLeft, MorseUpper, MorseNoTrans, MorseNonUnit, 1.0, subA, subB, sequence, request );
405 406
    }
    else {
BOUCHERIE Raphael's avatar
BOUCHERIE Raphael committed
407
#if defined(CHAMELEON_COPY_DIAG)
Mathieu Faverge's avatar
Mathieu Faverge committed
408 409 410 411 412 413 414
        {
            int m = chameleon_min(A->mt, A->nt) * A->mb;
            morse_zdesc_alloc(D, A->mb, A->nb, m, A->n, 0, 0, m, A->n, );
            Dptr = &D;
        }
#endif

BOUCHERIE Raphael's avatar
BOUCHERIE Raphael committed
415 416
        subB = morse_desc_submatrix(B, 0, 0, A->m, B->n);
        subA = morse_desc_submatrix(A, 0, 0, A->m, A->m);
417

Mathieu Faverge's avatar
Mathieu Faverge committed
418 419 420
        morse_pzgelqf_param( qrtree, A, TS, TT, Dptr, sequence, request );
        morse_pztrsm( MorseLeft, MorseLower, MorseNoTrans, MorseNonUnit, 1.0, subA, subB, sequence, request );
        morse_pzunmlq_param( qrtree, MorseLeft, MorseConjTrans, A, B, TS, TT, Dptr, sequence, request );
421
    }
BOUCHERIE Raphael's avatar
BOUCHERIE Raphael committed
422

Mathieu Faverge's avatar
Mathieu Faverge committed
423 424 425 426
    free(subA);
    free(subB);

    if (Dptr != NULL) {
427
        MORSE_Desc_Flush( A, sequence );
428
        MORSE_Desc_Flush( B, sequence );
429 430 431 432
        MORSE_Desc_Flush( TS, sequence );
        MORSE_Desc_Flush( TT, sequence );
        MORSE_Desc_Flush( Dptr, sequence );
        morse_sequence_wait( morse, sequence );
Mathieu Faverge's avatar
Mathieu Faverge committed
433
        morse_desc_mat_free( Dptr );
Mathieu Faverge's avatar
Mathieu Faverge committed
434
    }
435
    (void)D;
436 437
    return MORSE_SUCCESS;
}