Skip Navigation Links
Numerical Libraries
Linear Algebra
Differential Equations
Optimization
Samples
Skip Navigation Links
Linear Algebra
CSLapack
CSBlas
   1:  #region Translated by Jose Antonio De Santiago-Castillo.
   2:   
   3:  //Translated by Jose Antonio De Santiago-Castillo. 
   4:  //E-mail:JAntonioDeSantiago@gmail.com
   5:  //Web: www.DotNumerics.com
   6:  //
   7:  //Fortran to C# Translation.
   8:  //Translated by:
   9:  //F2CSharp Version 0.71 (November 10, 2009)
  10:  //Code Optimizations: None
  11:  //
  12:  #endregion
  13:   
  14:  using System;
  15:  using DotNumerics.FortranLibrary;
  16:   
  17:  namespace DotNumerics.CSLapack
  18:  {
  19:      /// <summary>
  20:      /// -- LAPACK driver routine (version 3.1) --
  21:      /// Univ. of Tennessee, Univ. of California Berkeley and NAG Ltd..
  22:      /// November 2006
  23:      /// Purpose
  24:      /// =======
  25:      /// 
  26:      /// DGGGLM solves a general Gauss-Markov linear model (GLM) problem:
  27:      /// 
  28:      /// minimize || y ||_2   subject to   d = A*x + B*y
  29:      /// x
  30:      /// 
  31:      /// where A is an N-by-M matrix, B is an N-by-P matrix, and d is a
  32:      /// given N-vector. It is assumed that M .LE. N .LE. M+P, and
  33:      /// 
  34:      /// rank(A) = M    and    rank( A B ) = N.
  35:      /// 
  36:      /// Under these assumptions, the constrained equation is always
  37:      /// consistent, and there is a unique solution x and a minimal 2-norm
  38:      /// solution y, which is obtained using a generalized QR factorization
  39:      /// of the matrices (A, B) given by
  40:      /// 
  41:      /// A = Q*(R),   B = Q*T*Z.
  42:      /// (0)
  43:      /// 
  44:      /// In particular, if matrix B is square nonsingular, then the problem
  45:      /// GLM is equivalent to the following weighted linear least squares
  46:      /// problem
  47:      /// 
  48:      /// minimize || inv(B)*(d-A*x) ||_2
  49:      /// x
  50:      /// 
  51:      /// where inv(B) denotes the inverse of B.
  52:      /// 
  53:      ///</summary>
  54:      public class DGGGLM
  55:      {
  56:      
  57:   
  58:          #region Dependencies
  59:          
  60:          DCOPY _dcopy; DGEMV _dgemv; DGGQRF _dggqrf; DORMQR _dormqr; DORMRQ _dormrq; DTRTRS _dtrtrs; XERBLA _xerbla; 
  61:          ILAENV _ilaenv;
  62:   
  63:          #endregion
  64:   
  65:   
  66:          #region Fields
  67:          
  68:          const double ZERO = 0.0E+0; const double ONE = 1.0E+0; bool LQUERY = false; int I = 0; int LOPT = 0; int LWKMIN = 0; 
  69:          int LWKOPT = 0;int NB = 0; int NB1 = 0; int NB2 = 0; int NB3 = 0; int NB4 = 0; int NP = 0; 
  70:   
  71:          #endregion
  72:   
  73:          public DGGGLM(DCOPY dcopy, DGEMV dgemv, DGGQRF dggqrf, DORMQR dormqr, DORMRQ dormrq, DTRTRS dtrtrs, XERBLA xerbla, ILAENV ilaenv)
  74:          {
  75:      
  76:   
  77:              #region Set Dependencies
  78:              
  79:              this._dcopy = dcopy; this._dgemv = dgemv; this._dggqrf = dggqrf; this._dormqr = dormqr; this._dormrq = dormrq; 
  80:              this._dtrtrs = dtrtrs;this._xerbla = xerbla; this._ilaenv = ilaenv; 
  81:   
  82:              #endregion
  83:   
  84:          }
  85:      
  86:          public DGGGLM()
  87:          {
  88:      
  89:   
  90:              #region Dependencies (Initialization)
  91:              
  92:              DCOPY dcopy = new DCOPY();
  93:              LSAME lsame = new LSAME();
  94:              XERBLA xerbla = new XERBLA();
  95:              DLAMC3 dlamc3 = new DLAMC3();
  96:              DLAPY2 dlapy2 = new DLAPY2();
  97:              DNRM2 dnrm2 = new DNRM2();
  98:              DSCAL dscal = new DSCAL();
  99:              IEEECK ieeeck = new IEEECK();
 100:              IPARMQ iparmq = new IPARMQ();
 101:              DGEMV dgemv = new DGEMV(lsame, xerbla);
 102:              DGER dger = new DGER(xerbla);
 103:              DLARF dlarf = new DLARF(dgemv, dger, lsame);
 104:              DLAMC1 dlamc1 = new DLAMC1(dlamc3);
 105:              DLAMC4 dlamc4 = new DLAMC4(dlamc3);
 106:              DLAMC5 dlamc5 = new DLAMC5(dlamc3);
 107:              DLAMC2 dlamc2 = new DLAMC2(dlamc3, dlamc1, dlamc4, dlamc5);
 108:              DLAMCH dlamch = new DLAMCH(lsame, dlamc2);
 109:              DLARFG dlarfg = new DLARFG(dlamch, dlapy2, dnrm2, dscal);
 110:              DGEQR2 dgeqr2 = new DGEQR2(dlarf, dlarfg, xerbla);
 111:              DGEMM dgemm = new DGEMM(lsame, xerbla);
 112:              DTRMM dtrmm = new DTRMM(lsame, xerbla);
 113:              DLARFB dlarfb = new DLARFB(lsame, dcopy, dgemm, dtrmm);
 114:              DTRMV dtrmv = new DTRMV(lsame, xerbla);
 115:              DLARFT dlarft = new DLARFT(dgemv, dtrmv, lsame);
 116:              ILAENV ilaenv = new ILAENV(ieeeck, iparmq);
 117:              DGEQRF dgeqrf = new DGEQRF(dgeqr2, dlarfb, dlarft, xerbla, ilaenv);
 118:              DGERQ2 dgerq2 = new DGERQ2(dlarf, dlarfg, xerbla);
 119:              DGERQF dgerqf = new DGERQF(dgerq2, dlarfb, dlarft, xerbla, ilaenv);
 120:              DORM2R dorm2r = new DORM2R(lsame, dlarf, xerbla);
 121:              DORMQR dormqr = new DORMQR(lsame, ilaenv, dlarfb, dlarft, dorm2r, xerbla);
 122:              DGGQRF dggqrf = new DGGQRF(dgeqrf, dgerqf, dormqr, xerbla, ilaenv);
 123:              DORMR2 dormr2 = new DORMR2(lsame, dlarf, xerbla);
 124:              DORMRQ dormrq = new DORMRQ(lsame, ilaenv, dlarfb, dlarft, dormr2, xerbla);
 125:              DTRSM dtrsm = new DTRSM(lsame, xerbla);
 126:              DTRTRS dtrtrs = new DTRTRS(lsame, dtrsm, xerbla);
 127:   
 128:              #endregion
 129:   
 130:   
 131:              #region Set Dependencies
 132:              
 133:              this._dcopy = dcopy; this._dgemv = dgemv; this._dggqrf = dggqrf; this._dormqr = dormqr; this._dormrq = dormrq; 
 134:              this._dtrtrs = dtrtrs;this._xerbla = xerbla; this._ilaenv = ilaenv; 
 135:   
 136:              #endregion
 137:   
 138:          }
 139:          /// <summary>
 140:          /// Purpose
 141:          /// =======
 142:          /// 
 143:          /// DGGGLM solves a general Gauss-Markov linear model (GLM) problem:
 144:          /// 
 145:          /// minimize || y ||_2   subject to   d = A*x + B*y
 146:          /// x
 147:          /// 
 148:          /// where A is an N-by-M matrix, B is an N-by-P matrix, and d is a
 149:          /// given N-vector. It is assumed that M .LE. N .LE. M+P, and
 150:          /// 
 151:          /// rank(A) = M    and    rank( A B ) = N.
 152:          /// 
 153:          /// Under these assumptions, the constrained equation is always
 154:          /// consistent, and there is a unique solution x and a minimal 2-norm
 155:          /// solution y, which is obtained using a generalized QR factorization
 156:          /// of the matrices (A, B) given by
 157:          /// 
 158:          /// A = Q*(R),   B = Q*T*Z.
 159:          /// (0)
 160:          /// 
 161:          /// In particular, if matrix B is square nonsingular, then the problem
 162:          /// GLM is equivalent to the following weighted linear least squares
 163:          /// problem
 164:          /// 
 165:          /// minimize || inv(B)*(d-A*x) ||_2
 166:          /// x
 167:          /// 
 168:          /// where inv(B) denotes the inverse of B.
 169:          /// 
 170:          ///</summary>
 171:          /// <param name="N">
 172:          /// (input) INTEGER
 173:          /// The number of rows of the matrices A and B.  N .GE. 0.
 174:          ///</param>
 175:          /// <param name="M">
 176:          /// (input) INTEGER
 177:          /// The number of columns of the matrix A.  0 .LE. M .LE. N.
 178:          ///</param>
 179:          /// <param name="P">
 180:          /// (input) INTEGER
 181:          /// The number of columns of the matrix B.  P .GE. N-M.
 182:          ///</param>
 183:          /// <param name="A">
 184:          /// = Q*(R),   B = Q*T*Z.
 185:          /// (0)
 186:          ///</param>
 187:          /// <param name="LDA">
 188:          /// (input) INTEGER
 189:          /// The leading dimension of the array A. LDA .GE. max(1,N).
 190:          ///</param>
 191:          /// <param name="B">
 192:          /// (input/output) DOUBLE PRECISION array, dimension (LDB,P)
 193:          /// On entry, the N-by-P matrix B.
 194:          /// On exit, if N .LE. P, the upper triangle of the subarray
 195:          /// B(1:N,P-N+1:P) contains the N-by-N upper triangular matrix T;
 196:          /// if N .GT. P, the elements on and above the (N-P)th subdiagonal
 197:          /// contain the N-by-P upper trapezoidal matrix T.
 198:          ///</param>
 199:          /// <param name="LDB">
 200:          /// (input) INTEGER
 201:          /// The leading dimension of the array B. LDB .GE. max(1,N).
 202:          ///</param>
 203:          /// <param name="D">
 204:          /// (input/output) DOUBLE PRECISION array, dimension (N)
 205:          /// On entry, D is the left hand side of the GLM equation.
 206:          /// On exit, D is destroyed.
 207:          ///</param>
 208:          /// <param name="X">
 209:          /// (output) DOUBLE PRECISION array, dimension (M)
 210:          ///</param>
 211:          /// <param name="Y">
 212:          /// (output) DOUBLE PRECISION array, dimension (P)
 213:          /// On exit, X and Y are the solutions of the GLM problem.
 214:          ///</param>
 215:          /// <param name="WORK">
 216:          /// (workspace/output) DOUBLE PRECISION array, dimension (MAX(1,LWORK))
 217:          /// On exit, if INFO = 0, WORK(1) returns the optimal LWORK.
 218:          ///</param>
 219:          /// <param name="LWORK">
 220:          /// (input) INTEGER
 221:          /// The dimension of the array WORK. LWORK .GE. max(1,N+M+P).
 222:          /// For optimum performance, LWORK .GE. M+min(N,P)+max(N,P)*NB,
 223:          /// where NB is an upper bound for the optimal blocksizes for
 224:          /// DGEQRF, SGERQF, DORMQR and SORMRQ.
 225:          /// 
 226:          /// If LWORK = -1, then a workspace query is assumed; the routine
 227:          /// only calculates the optimal size of the WORK array, returns
 228:          /// this value as the first entry of the WORK array, and no error
 229:          /// message related to LWORK is issued by XERBLA.
 230:          ///</param>
 231:          /// <param name="INFO">
 232:          /// (output) INTEGER
 233:          /// = 0:  successful exit.
 234:          /// .LT. 0:  if INFO = -i, the i-th argument had an illegal value.
 235:          /// = 1:  the upper triangular factor R associated with A in the
 236:          /// generalized QR factorization of the pair (A, B) is
 237:          /// singular, so that rank(A) .LT. M; the least squares
 238:          /// solution could not be computed.
 239:          /// = 2:  the bottom (N-M) by (N-M) part of the upper trapezoidal
 240:          /// factor T associated with B in the generalized QR
 241:          /// factorization of the pair (A, B) is singular, so that
 242:          /// rank( A B ) .LT. N; the least squares solution could not
 243:          /// be computed.
 244:          ///</param>
 245:          public void Run(int N, int M, int P, ref double[] A, int offset_a, int LDA, ref double[] B, int offset_b
 246:                           , int LDB, ref double[] D, int offset_d, ref double[] X, int offset_x, ref double[] Y, int offset_y, ref double[] WORK, int offset_work, int LWORK
 247:                           , ref int INFO)
 248:          {
 249:   
 250:              #region Array Index Correction
 251:              
 252:               int o_a = -1 - LDA + offset_a;  int o_b = -1 - LDB + offset_b;  int o_d = -1 + offset_d;  int o_x = -1 + offset_x; 
 253:               int o_y = -1 + offset_y; int o_work = -1 + offset_work; 
 254:   
 255:              #endregion
 256:   
 257:   
 258:              #region Prolog
 259:              
 260:              // *
 261:              // *  -- LAPACK driver routine (version 3.1) --
 262:              // *     Univ. of Tennessee, Univ. of California Berkeley and NAG Ltd..
 263:              // *     November 2006
 264:              // *
 265:              // *     .. Scalar Arguments ..
 266:              // *     ..
 267:              // *     .. Array Arguments ..
 268:              // *     ..
 269:              // *
 270:              // *  Purpose
 271:              // *  =======
 272:              // *
 273:              // *  DGGGLM solves a general Gauss-Markov linear model (GLM) problem:
 274:              // *
 275:              // *          minimize || y ||_2   subject to   d = A*x + B*y
 276:              // *              x
 277:              // *
 278:              // *  where A is an N-by-M matrix, B is an N-by-P matrix, and d is a
 279:              // *  given N-vector. It is assumed that M <= N <= M+P, and
 280:              // *
 281:              // *             rank(A) = M    and    rank( A B ) = N.
 282:              // *
 283:              // *  Under these assumptions, the constrained equation is always
 284:              // *  consistent, and there is a unique solution x and a minimal 2-norm
 285:              // *  solution y, which is obtained using a generalized QR factorization
 286:              // *  of the matrices (A, B) given by
 287:              // *
 288:              // *     A = Q*(R),   B = Q*T*Z.
 289:              // *           (0)
 290:              // *
 291:              // *  In particular, if matrix B is square nonsingular, then the problem
 292:              // *  GLM is equivalent to the following weighted linear least squares
 293:              // *  problem
 294:              // *
 295:              // *               minimize || inv(B)*(d-A*x) ||_2
 296:              // *                   x
 297:              // *
 298:              // *  where inv(B) denotes the inverse of B.
 299:              // *
 300:              // *  Arguments
 301:              // *  =========
 302:              // *
 303:              // *  N       (input) INTEGER
 304:              // *          The number of rows of the matrices A and B.  N >= 0.
 305:              // *
 306:              // *  M       (input) INTEGER
 307:              // *          The number of columns of the matrix A.  0 <= M <= N.
 308:              // *
 309:              // *  P       (input) INTEGER
 310:              // *          The number of columns of the matrix B.  P >= N-M.
 311:              // *
 312:              // *  A       (input/output) DOUBLE PRECISION array, dimension (LDA,M)
 313:              // *          On entry, the N-by-M matrix A.
 314:              // *          On exit, the upper triangular part of the array A contains
 315:              // *          the M-by-M upper triangular matrix R.
 316:              // *
 317:              // *  LDA     (input) INTEGER
 318:              // *          The leading dimension of the array A. LDA >= max(1,N).
 319:              // *
 320:              // *  B       (input/output) DOUBLE PRECISION array, dimension (LDB,P)
 321:              // *          On entry, the N-by-P matrix B.
 322:              // *          On exit, if N <= P, the upper triangle of the subarray
 323:              // *          B(1:N,P-N+1:P) contains the N-by-N upper triangular matrix T;
 324:              // *          if N > P, the elements on and above the (N-P)th subdiagonal
 325:              // *          contain the N-by-P upper trapezoidal matrix T.
 326:              // *
 327:              // *  LDB     (input) INTEGER
 328:              // *          The leading dimension of the array B. LDB >= max(1,N).
 329:              // *
 330:              // *  D       (input/output) DOUBLE PRECISION array, dimension (N)
 331:              // *          On entry, D is the left hand side of the GLM equation.
 332:              // *          On exit, D is destroyed.
 333:              // *
 334:              // *  X       (output) DOUBLE PRECISION array, dimension (M)
 335:              // *  Y       (output) DOUBLE PRECISION array, dimension (P)
 336:              // *          On exit, X and Y are the solutions of the GLM problem.
 337:              // *
 338:              // *  WORK    (workspace/output) DOUBLE PRECISION array, dimension (MAX(1,LWORK))
 339:              // *          On exit, if INFO = 0, WORK(1) returns the optimal LWORK.
 340:              // *
 341:              // *  LWORK   (input) INTEGER
 342:              // *          The dimension of the array WORK. LWORK >= max(1,N+M+P).
 343:              // *          For optimum performance, LWORK >= M+min(N,P)+max(N,P)*NB,
 344:              // *          where NB is an upper bound for the optimal blocksizes for
 345:              // *          DGEQRF, SGERQF, DORMQR and SORMRQ.
 346:              // *
 347:              // *          If LWORK = -1, then a workspace query is assumed; the routine
 348:              // *          only calculates the optimal size of the WORK array, returns
 349:              // *          this value as the first entry of the WORK array, and no error
 350:              // *          message related to LWORK is issued by XERBLA.
 351:              // *
 352:              // *  INFO    (output) INTEGER
 353:              // *          = 0:  successful exit.
 354:              // *          < 0:  if INFO = -i, the i-th argument had an illegal value.
 355:              // *          = 1:  the upper triangular factor R associated with A in the
 356:              // *                generalized QR factorization of the pair (A, B) is
 357:              // *                singular, so that rank(A) < M; the least squares
 358:              // *                solution could not be computed.
 359:              // *          = 2:  the bottom (N-M) by (N-M) part of the upper trapezoidal
 360:              // *                factor T associated with B in the generalized QR
 361:              // *                factorization of the pair (A, B) is singular, so that
 362:              // *                rank( A B ) < N; the least squares solution could not
 363:              // *                be computed.
 364:              // *
 365:              // *  ===================================================================
 366:              // *
 367:              // *     .. Parameters ..
 368:              // *     ..
 369:              // *     .. Local Scalars ..
 370:              // *     ..
 371:              // *     .. External Subroutines ..
 372:              // *     ..
 373:              // *     .. External Functions ..
 374:              // *     ..
 375:              // *     .. Intrinsic Functions ..
 376:              //      INTRINSIC          INT, MAX, MIN;
 377:              // *     ..
 378:              // *     .. Executable Statements ..
 379:              // *
 380:              // *     Test the input parameters
 381:              // *
 382:   
 383:              #endregion
 384:   
 385:   
 386:              #region Body
 387:              
 388:              INFO = 0;
 389:              NP = Math.Min(N, P);
 390:              LQUERY = (LWORK ==  - 1);
 391:              if (N < 0)
 392:              {
 393:                  INFO =  - 1;
 394:              }
 395:              else
 396:              {
 397:                  if (M < 0 || M > N)
 398:                  {
 399:                      INFO =  - 2;
 400:                  }
 401:                  else
 402:                  {
 403:                      if (P < 0 || P < N - M)
 404:                      {
 405:                          INFO =  - 3;
 406:                      }
 407:                      else
 408:                      {
 409:                          if (LDA < Math.Max(1, N))
 410:                          {
 411:                              INFO =  - 5;
 412:                          }
 413:                          else
 414:                          {
 415:                              if (LDB < Math.Max(1, N))
 416:                              {
 417:                                  INFO =  - 7;
 418:                              }
 419:                          }
 420:                      }
 421:                  }
 422:              }
 423:              // *
 424:              // *     Calculate workspace
 425:              // *
 426:              if (INFO == 0)
 427:              {
 428:                  if (N == 0)
 429:                  {
 430:                      LWKMIN = 1;
 431:                      LWKOPT = 1;
 432:                  }
 433:                  else
 434:                  {
 435:                      NB1 = this._ilaenv.Run(1, "DGEQRF", " ", N, M,  - 1,  - 1);
 436:                      NB2 = this._ilaenv.Run(1, "DGERQF", " ", N, M,  - 1,  - 1);
 437:                      NB3 = this._ilaenv.Run(1, "DORMQR", " ", N, M, P,  - 1);
 438:                      NB4 = this._ilaenv.Run(1, "DORMRQ", " ", N, M, P,  - 1);
 439:                      NB = Math.Max(NB1, Math.Max(NB2, Math.Max(NB3, NB4)));
 440:                      LWKMIN = M + N + P;
 441:                      LWKOPT = M + NP + Math.Max(N, P) * NB;
 442:                  }
 443:                  WORK[1 + o_work] = LWKOPT;
 444:                  // *
 445:                  if (LWORK < LWKMIN && !LQUERY)
 446:                  {
 447:                      INFO =  - 12;
 448:                  }
 449:              }
 450:              // *
 451:              if (INFO != 0)
 452:              {
 453:                  this._xerbla.Run("DGGGLM",  - INFO);
 454:                  return;
 455:              }
 456:              else
 457:              {
 458:                  if (LQUERY)
 459:                  {
 460:                      return;
 461:                  }
 462:              }
 463:              // *
 464:              // *     Quick return if possible
 465:              // *
 466:              if (N == 0) return;
 467:              // *
 468:              // *     Compute the GQR factorization of matrices A and B:
 469:              // *
 470:              // *            Q'*A = ( R11 ) M,    Q'*B*Z' = ( T11   T12 ) M
 471:              // *                   (  0  ) N-M             (  0    T22 ) N-M
 472:              // *                      M                     M+P-N  N-M
 473:              // *
 474:              // *     where R11 and T22 are upper triangular, and Q and Z are
 475:              // *     orthogonal.
 476:              // *
 477:              this._dggqrf.Run(N, M, P, ref A, offset_a, LDA, ref WORK, offset_work
 478:                               , ref B, offset_b, LDB, ref WORK, M + 1 + o_work, ref WORK, M + NP + 1 + o_work, LWORK - M - NP, ref INFO);
 479:              LOPT = (int)WORK[M + NP + 1 + o_work];
 480:              // *
 481:              // *     Update left-hand-side vector d = Q'*d = ( d1 ) M
 482:              // *                                             ( d2 ) N-M
 483:              // *
 484:              this._dormqr.Run("Left", "Transpose", N, 1, M, ref A, offset_a
 485:                               , LDA, WORK, offset_work, ref D, offset_d, Math.Max(1, N), ref WORK, M + NP + 1 + o_work, LWORK - M - NP
 486:                               , ref INFO);
 487:              LOPT = Math.Max(LOPT, Convert.ToInt32(Math.Truncate(WORK[M + NP + 1 + o_work])));
 488:              // *
 489:              // *     Solve T22*y2 = d2 for y2
 490:              // *
 491:              if (N > M)
 492:              {
 493:                  this._dtrtrs.Run("Upper", "No transpose", "Non unit", N - M, 1, B, M + 1+(M + P - N + 1) * LDB + o_b
 494:                                   , LDB, ref D, M + 1 + o_d, N - M, ref INFO);
 495:                  // *
 496:                  if (INFO > 0)
 497:                  {
 498:                      INFO = 1;
 499:                      return;
 500:                  }
 501:                  // *
 502:                  this._dcopy.Run(N - M, D, M + 1 + o_d, 1, ref Y, M + P - N + 1 + o_y, 1);
 503:              }
 504:              // *
 505:              // *     Set y1 = 0
 506:              // *
 507:              for (I = 1; I <= M + P - N; I++)
 508:              {
 509:                  Y[I + o_y] = ZERO;
 510:              }
 511:              // *
 512:              // *     Update d1 = d1 - T12*y2
 513:              // *
 514:              this._dgemv.Run("No transpose", M, N - M,  - ONE, B, 1+(M + P - N + 1) * LDB + o_b, LDB
 515:                              , Y, M + P - N + 1 + o_y, 1, ONE, ref D, offset_d, 1);
 516:              // *
 517:              // *     Solve triangular system: R11*x = d1
 518:              // *
 519:              if (M > 0)
 520:              {
 521:                  this._dtrtrs.Run("Upper", "No Transpose", "Non unit", M, 1, A, offset_a
 522:                                   , LDA, ref D, offset_d, M, ref INFO);
 523:                  // *
 524:                  if (INFO > 0)
 525:                  {
 526:                      INFO = 2;
 527:                      return;
 528:                  }
 529:                  // *
 530:                  // *        Copy D to X
 531:                  // *
 532:                  this._dcopy.Run(M, D, offset_d, 1, ref X, offset_x, 1);
 533:              }
 534:              // *
 535:              // *     Backward transformation y = Z'*y
 536:              // *
 537:              this._dormrq.Run("Left", "Transpose", P, 1, NP, ref B, Math.Max(1, N - P + 1)+1 * LDB + o_b
 538:                               , LDB, WORK, M + 1 + o_work, ref Y, offset_y, Math.Max(1, P), ref WORK, M + NP + 1 + o_work, LWORK - M - NP
 539:                               , ref INFO);
 540:              WORK[1 + o_work] = M + NP + Math.Max(LOPT, Convert.ToInt32(Math.Truncate(WORK[M + NP + 1 + o_work])));
 541:              // *
 542:              return;
 543:              // *
 544:              // *     End of DGGGLM
 545:              // *
 546:   
 547:              #endregion
 548:   
 549:          }
 550:      }
 551:  }