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 auxiliary routine (version 3.1) --
  21:      /// Univ. of Tennessee, Univ. of California Berkeley and NAG Ltd..
  22:      /// November 2006
  23:      /// Purpose
  24:      /// =======
  25:      /// 
  26:      /// DLASD8 finds the square roots of the roots of the secular equation,
  27:      /// as defined by the values in DSIGMA and Z. It makes the appropriate
  28:      /// calls to DLASD4, and stores, for each  element in D, the distance
  29:      /// to its two nearest poles (elements in DSIGMA). It also updates
  30:      /// the arrays VF and VL, the first and last components of all the
  31:      /// right singular vectors of the original bidiagonal matrix.
  32:      /// 
  33:      /// DLASD8 is called from DLASD6.
  34:      /// 
  35:      ///</summary>
  36:      public class DLASD8
  37:      {
  38:      
  39:   
  40:          #region Dependencies
  41:          
  42:          DCOPY _dcopy; DLASCL _dlascl; DLASD4 _dlasd4; DLASET _dlaset; XERBLA _xerbla; DDOT _ddot; DLAMC3 _dlamc3; DNRM2 _dnrm2; 
  43:   
  44:          #endregion
  45:   
  46:   
  47:          #region Fields
  48:          
  49:          const double ONE = 1.0E+0; int I = 0; int IWK1 = 0; int IWK2 = 0; int IWK2I = 0; int IWK3 = 0; int IWK3I = 0; int J = 0; 
  50:          double DIFLJ = 0;double DIFRJ = 0; double DJ = 0; double DSIGJ = 0; double DSIGJP = 0; double RHO = 0; double TEMP = 0; 
  51:   
  52:          #endregion
  53:   
  54:          public DLASD8(DCOPY dcopy, DLASCL dlascl, DLASD4 dlasd4, DLASET dlaset, XERBLA xerbla, DDOT ddot, DLAMC3 dlamc3, DNRM2 dnrm2)
  55:          {
  56:      
  57:   
  58:              #region Set Dependencies
  59:              
  60:              this._dcopy = dcopy; this._dlascl = dlascl; this._dlasd4 = dlasd4; this._dlaset = dlaset; this._xerbla = xerbla; 
  61:              this._ddot = ddot;this._dlamc3 = dlamc3; this._dnrm2 = dnrm2; 
  62:   
  63:              #endregion
  64:   
  65:          }
  66:      
  67:          public DLASD8()
  68:          {
  69:      
  70:   
  71:              #region Dependencies (Initialization)
  72:              
  73:              DCOPY dcopy = new DCOPY();
  74:              LSAME lsame = new LSAME();
  75:              DLAMC3 dlamc3 = new DLAMC3();
  76:              XERBLA xerbla = new XERBLA();
  77:              DLASD5 dlasd5 = new DLASD5();
  78:              DDOT ddot = new DDOT();
  79:              DNRM2 dnrm2 = new DNRM2();
  80:              DLAMC1 dlamc1 = new DLAMC1(dlamc3);
  81:              DLAMC4 dlamc4 = new DLAMC4(dlamc3);
  82:              DLAMC5 dlamc5 = new DLAMC5(dlamc3);
  83:              DLAMC2 dlamc2 = new DLAMC2(dlamc3, dlamc1, dlamc4, dlamc5);
  84:              DLAMCH dlamch = new DLAMCH(lsame, dlamc2);
  85:              DLASCL dlascl = new DLASCL(lsame, dlamch, xerbla);
  86:              DLAED6 dlaed6 = new DLAED6(dlamch);
  87:              DLASD4 dlasd4 = new DLASD4(dlaed6, dlasd5, dlamch);
  88:              DLASET dlaset = new DLASET(lsame);
  89:   
  90:              #endregion
  91:   
  92:   
  93:              #region Set Dependencies
  94:              
  95:              this._dcopy = dcopy; this._dlascl = dlascl; this._dlasd4 = dlasd4; this._dlaset = dlaset; this._xerbla = xerbla; 
  96:              this._ddot = ddot;this._dlamc3 = dlamc3; this._dnrm2 = dnrm2; 
  97:   
  98:              #endregion
  99:   
 100:          }
 101:          /// <summary>
 102:          /// Purpose
 103:          /// =======
 104:          /// 
 105:          /// DLASD8 finds the square roots of the roots of the secular equation,
 106:          /// as defined by the values in DSIGMA and Z. It makes the appropriate
 107:          /// calls to DLASD4, and stores, for each  element in D, the distance
 108:          /// to its two nearest poles (elements in DSIGMA). It also updates
 109:          /// the arrays VF and VL, the first and last components of all the
 110:          /// right singular vectors of the original bidiagonal matrix.
 111:          /// 
 112:          /// DLASD8 is called from DLASD6.
 113:          /// 
 114:          ///</summary>
 115:          /// <param name="ICOMPQ">
 116:          /// (input) INTEGER
 117:          /// Specifies whether singular vectors are to be computed in
 118:          /// factored form in the calling routine:
 119:          /// = 0: Compute singular values only.
 120:          /// = 1: Compute singular vectors in factored form as well.
 121:          ///</param>
 122:          /// <param name="K">
 123:          /// (input) INTEGER
 124:          /// The number of terms in the rational function to be solved
 125:          /// by DLASD4.  K .GE. 1.
 126:          ///</param>
 127:          /// <param name="D">
 128:          /// (output) DOUBLE PRECISION array, dimension ( K )
 129:          /// On output, D contains the updated singular values.
 130:          ///</param>
 131:          /// <param name="Z">
 132:          /// (input) DOUBLE PRECISION array, dimension ( K )
 133:          /// The first K elements of this array contain the components
 134:          /// of the deflation-adjusted updating row vector.
 135:          ///</param>
 136:          /// <param name="VF">
 137:          /// (input/output) DOUBLE PRECISION array, dimension ( K )
 138:          /// On entry, VF contains  information passed through DBEDE8.
 139:          /// On exit, VF contains the first K components of the first
 140:          /// components of all right singular vectors of the bidiagonal
 141:          /// matrix.
 142:          ///</param>
 143:          /// <param name="VL">
 144:          /// (input/output) DOUBLE PRECISION array, dimension ( K )
 145:          /// On entry, VL contains  information passed through DBEDE8.
 146:          /// On exit, VL contains the first K components of the last
 147:          /// components of all right singular vectors of the bidiagonal
 148:          /// matrix.
 149:          ///</param>
 150:          /// <param name="DIFL">
 151:          /// (output) DOUBLE PRECISION array, dimension ( K )
 152:          /// On exit, DIFL(I) = D(I) - DSIGMA(I).
 153:          ///</param>
 154:          /// <param name="DIFR">
 155:          /// (output) DOUBLE PRECISION array,
 156:          /// dimension ( LDDIFR, 2 ) if ICOMPQ = 1 and
 157:          /// dimension ( K ) if ICOMPQ = 0.
 158:          /// On exit, DIFR(I,1) = D(I) - DSIGMA(I+1), DIFR(K,1) is not
 159:          /// defined and will not be referenced.
 160:          /// 
 161:          /// If ICOMPQ = 1, DIFR(1:K,2) is an array containing the
 162:          /// normalizing factors for the right singular vector matrix.
 163:          ///</param>
 164:          /// <param name="LDDIFR">
 165:          /// (input) INTEGER
 166:          /// The leading dimension of DIFR, must be at least K.
 167:          ///</param>
 168:          /// <param name="DSIGMA">
 169:          /// (input) DOUBLE PRECISION array, dimension ( K )
 170:          /// The first K elements of this array contain the old roots
 171:          /// of the deflated updating problem.  These are the poles
 172:          /// of the secular equation.
 173:          ///</param>
 174:          /// <param name="WORK">
 175:          /// (workspace) DOUBLE PRECISION array, dimension at least 3 * K
 176:          ///</param>
 177:          /// <param name="INFO">
 178:          /// (output) INTEGER
 179:          /// = 0:  successful exit.
 180:          /// .LT. 0:  if INFO = -i, the i-th argument had an illegal value.
 181:          /// .GT. 0:  if INFO = 1, an singular value did not converge
 182:          ///</param>
 183:          public void Run(int ICOMPQ, int K, ref double[] D, int offset_d, ref double[] Z, int offset_z, ref double[] VF, int offset_vf, ref double[] VL, int offset_vl
 184:                           , ref double[] DIFL, int offset_difl, ref double[] DIFR, int offset_difr, int LDDIFR, ref double[] DSIGMA, int offset_dsigma, ref double[] WORK, int offset_work, ref int INFO)
 185:          {
 186:   
 187:              #region Array Index Correction
 188:              
 189:               int o_d = -1 + offset_d;  int o_z = -1 + offset_z;  int o_vf = -1 + offset_vf;  int o_vl = -1 + offset_vl; 
 190:               int o_difl = -1 + offset_difl; int o_difr = -1 - LDDIFR + offset_difr;  int o_dsigma = -1 + offset_dsigma; 
 191:               int o_work = -1 + offset_work;
 192:   
 193:              #endregion
 194:   
 195:   
 196:              #region Prolog
 197:              
 198:              // *
 199:              // *  -- LAPACK auxiliary routine (version 3.1) --
 200:              // *     Univ. of Tennessee, Univ. of California Berkeley and NAG Ltd..
 201:              // *     November 2006
 202:              // *
 203:              // *     .. Scalar Arguments ..
 204:              // *     ..
 205:              // *     .. Array Arguments ..
 206:              // *     ..
 207:              // *
 208:              // *  Purpose
 209:              // *  =======
 210:              // *
 211:              // *  DLASD8 finds the square roots of the roots of the secular equation,
 212:              // *  as defined by the values in DSIGMA and Z. It makes the appropriate
 213:              // *  calls to DLASD4, and stores, for each  element in D, the distance
 214:              // *  to its two nearest poles (elements in DSIGMA). It also updates
 215:              // *  the arrays VF and VL, the first and last components of all the
 216:              // *  right singular vectors of the original bidiagonal matrix.
 217:              // *
 218:              // *  DLASD8 is called from DLASD6.
 219:              // *
 220:              // *  Arguments
 221:              // *  =========
 222:              // *
 223:              // *  ICOMPQ  (input) INTEGER
 224:              // *          Specifies whether singular vectors are to be computed in
 225:              // *          factored form in the calling routine:
 226:              // *          = 0: Compute singular values only.
 227:              // *          = 1: Compute singular vectors in factored form as well.
 228:              // *
 229:              // *  K       (input) INTEGER
 230:              // *          The number of terms in the rational function to be solved
 231:              // *          by DLASD4.  K >= 1.
 232:              // *
 233:              // *  D       (output) DOUBLE PRECISION array, dimension ( K )
 234:              // *          On output, D contains the updated singular values.
 235:              // *
 236:              // *  Z       (input) DOUBLE PRECISION array, dimension ( K )
 237:              // *          The first K elements of this array contain the components
 238:              // *          of the deflation-adjusted updating row vector.
 239:              // *
 240:              // *  VF      (input/output) DOUBLE PRECISION array, dimension ( K )
 241:              // *          On entry, VF contains  information passed through DBEDE8.
 242:              // *          On exit, VF contains the first K components of the first
 243:              // *          components of all right singular vectors of the bidiagonal
 244:              // *          matrix.
 245:              // *
 246:              // *  VL      (input/output) DOUBLE PRECISION array, dimension ( K )
 247:              // *          On entry, VL contains  information passed through DBEDE8.
 248:              // *          On exit, VL contains the first K components of the last
 249:              // *          components of all right singular vectors of the bidiagonal
 250:              // *          matrix.
 251:              // *
 252:              // *  DIFL    (output) DOUBLE PRECISION array, dimension ( K )
 253:              // *          On exit, DIFL(I) = D(I) - DSIGMA(I).
 254:              // *
 255:              // *  DIFR    (output) DOUBLE PRECISION array,
 256:              // *                   dimension ( LDDIFR, 2 ) if ICOMPQ = 1 and
 257:              // *                   dimension ( K ) if ICOMPQ = 0.
 258:              // *          On exit, DIFR(I,1) = D(I) - DSIGMA(I+1), DIFR(K,1) is not
 259:              // *          defined and will not be referenced.
 260:              // *
 261:              // *          If ICOMPQ = 1, DIFR(1:K,2) is an array containing the
 262:              // *          normalizing factors for the right singular vector matrix.
 263:              // *
 264:              // *  LDDIFR  (input) INTEGER
 265:              // *          The leading dimension of DIFR, must be at least K.
 266:              // *
 267:              // *  DSIGMA  (input) DOUBLE PRECISION array, dimension ( K )
 268:              // *          The first K elements of this array contain the old roots
 269:              // *          of the deflated updating problem.  These are the poles
 270:              // *          of the secular equation.
 271:              // *
 272:              // *  WORK    (workspace) DOUBLE PRECISION array, dimension at least 3 * K
 273:              // *
 274:              // *  INFO    (output) INTEGER
 275:              // *          = 0:  successful exit.
 276:              // *          < 0:  if INFO = -i, the i-th argument had an illegal value.
 277:              // *          > 0:  if INFO = 1, an singular value did not converge
 278:              // *
 279:              // *  Further Details
 280:              // *  ===============
 281:              // *
 282:              // *  Based on contributions by
 283:              // *     Ming Gu and Huan Ren, Computer Science Division, University of
 284:              // *     California at Berkeley, USA
 285:              // *
 286:              // *  =====================================================================
 287:              // *
 288:              // *     .. Parameters ..
 289:              // *     ..
 290:              // *     .. Local Scalars ..
 291:              // *     ..
 292:              // *     .. External Subroutines ..
 293:              // *     ..
 294:              // *     .. External Functions ..
 295:              // *     ..
 296:              // *     .. Intrinsic Functions ..
 297:              //      INTRINSIC          ABS, SIGN, SQRT;
 298:              // *     ..
 299:              // *     .. Executable Statements ..
 300:              // *
 301:              // *     Test the input parameters.
 302:              // *
 303:   
 304:              #endregion
 305:   
 306:   
 307:              #region Body
 308:              
 309:              INFO = 0;
 310:              // *
 311:              if ((ICOMPQ < 0) || (ICOMPQ > 1))
 312:              {
 313:                  INFO =  - 1;
 314:              }
 315:              else
 316:              {
 317:                  if (K < 1)
 318:                  {
 319:                      INFO =  - 2;
 320:                  }
 321:                  else
 322:                  {
 323:                      if (LDDIFR < K)
 324:                      {
 325:                          INFO =  - 9;
 326:                      }
 327:                  }
 328:              }
 329:              if (INFO != 0)
 330:              {
 331:                  this._xerbla.Run("DLASD8",  - INFO);
 332:                  return;
 333:              }
 334:              // *
 335:              // *     Quick return if possible
 336:              // *
 337:              if (K == 1)
 338:              {
 339:                  D[1 + o_d] = Math.Abs(Z[1 + o_z]);
 340:                  DIFL[1 + o_difl] = D[1 + o_d];
 341:                  if (ICOMPQ == 1)
 342:                  {
 343:                      DIFL[2 + o_difl] = ONE;
 344:                      DIFR[1+2 * LDDIFR + o_difr] = ONE;
 345:                  }
 346:                  return;
 347:              }
 348:              // *
 349:              // *     Modify values DSIGMA(i) to make sure all DSIGMA(i)-DSIGMA(j) can
 350:              // *     be computed with high relative accuracy (barring over/underflow).
 351:              // *     This is a problem on machines without a guard digit in
 352:              // *     add/subtract (Cray XMP, Cray YMP, Cray C 90 and Cray 2).
 353:              // *     The following code replaces DSIGMA(I) by 2*DSIGMA(I)-DSIGMA(I),
 354:              // *     which on any of these machines zeros out the bottommost
 355:              // *     bit of DSIGMA(I) if it is 1; this makes the subsequent
 356:              // *     subtractions DSIGMA(I)-DSIGMA(J) unproblematic when cancellation
 357:              // *     occurs. On binary machines with a guard digit (almost all
 358:              // *     machines) it does not change DSIGMA(I) at all. On hexadecimal
 359:              // *     and decimal machines with a guard digit, it slightly
 360:              // *     changes the bottommost bits of DSIGMA(I). It does not account
 361:              // *     for hexadecimal or decimal machines without guard digits
 362:              // *     (we know of none). We use a subroutine call to compute
 363:              // *     2*DSIGMA(I) to prevent optimizing compilers from eliminating
 364:              // *     this code.
 365:              // *
 366:              for (I = 1; I <= K; I++)
 367:              {
 368:                  DSIGMA[I + o_dsigma] = this._dlamc3.Run(DSIGMA[I + o_dsigma], DSIGMA[I + o_dsigma]) - DSIGMA[I + o_dsigma];
 369:              }
 370:              // *
 371:              // *     Book keeping.
 372:              // *
 373:              IWK1 = 1;
 374:              IWK2 = IWK1 + K;
 375:              IWK3 = IWK2 + K;
 376:              IWK2I = IWK2 - 1;
 377:              IWK3I = IWK3 - 1;
 378:              // *
 379:              // *     Normalize Z.
 380:              // *
 381:              RHO = this._dnrm2.Run(K, Z, offset_z, 1);
 382:              this._dlascl.Run("G", 0, 0, RHO, ONE, K
 383:                               , 1, ref Z, offset_z, K, ref INFO);
 384:              RHO = RHO * RHO;
 385:              // *
 386:              // *     Initialize WORK(IWK3).
 387:              // *
 388:              this._dlaset.Run("A", K, 1, ONE, ONE, ref WORK, IWK3 + o_work
 389:                               , K);
 390:              // *
 391:              // *     Compute the updated singular values, the arrays DIFL, DIFR,
 392:              // *     and the updated Z.
 393:              // *
 394:              for (J = 1; J <= K; J++)
 395:              {
 396:                  this._dlasd4.Run(K, J, DSIGMA, offset_dsigma, Z, offset_z, ref WORK, IWK1 + o_work, RHO
 397:                                   , ref D[J + o_d], ref WORK, IWK2 + o_work, ref INFO);
 398:                  // *
 399:                  // *        If the root finder fails, the computation is terminated.
 400:                  // *
 401:                  if (INFO != 0)
 402:                  {
 403:                      return;
 404:                  }
 405:                  WORK[IWK3I + J + o_work] = WORK[IWK3I + J + o_work] * WORK[J + o_work] * WORK[IWK2I + J + o_work];
 406:                  DIFL[J + o_difl] =  - WORK[J + o_work];
 407:                  DIFR[J+1 * LDDIFR + o_difr] =  - WORK[J + 1 + o_work];
 408:                  for (I = 1; I <= J - 1; I++)
 409:                  {
 410:                      WORK[IWK3I + I + o_work] = WORK[IWK3I + I + o_work] * WORK[I + o_work] * WORK[IWK2I + I + o_work] / (DSIGMA[I + o_dsigma] - DSIGMA[J + o_dsigma]) / (DSIGMA[I + o_dsigma] + DSIGMA[J + o_dsigma]);
 411:                  }
 412:                  for (I = J + 1; I <= K; I++)
 413:                  {
 414:                      WORK[IWK3I + I + o_work] = WORK[IWK3I + I + o_work] * WORK[I + o_work] * WORK[IWK2I + I + o_work] / (DSIGMA[I + o_dsigma] - DSIGMA[J + o_dsigma]) / (DSIGMA[I + o_dsigma] + DSIGMA[J + o_dsigma]);
 415:                  }
 416:              }
 417:              // *
 418:              // *     Compute updated Z.
 419:              // *
 420:              for (I = 1; I <= K; I++)
 421:              {
 422:                  Z[I + o_z] = FortranLib.Sign(Math.Sqrt(Math.Abs(WORK[IWK3I + I + o_work])),Z[I + o_z]);
 423:              }
 424:              // *
 425:              // *     Update VF and VL.
 426:              // *
 427:              for (J = 1; J <= K; J++)
 428:              {
 429:                  DIFLJ = DIFL[J + o_difl];
 430:                  DJ = D[J + o_d];
 431:                  DSIGJ =  - DSIGMA[J + o_dsigma];
 432:                  if (J < K)
 433:                  {
 434:                      DIFRJ =  - DIFR[J+1 * LDDIFR + o_difr];
 435:                      DSIGJP =  - DSIGMA[J + 1 + o_dsigma];
 436:                  }
 437:                  WORK[J + o_work] =  - Z[J + o_z] / DIFLJ / (DSIGMA[J + o_dsigma] + DJ);
 438:                  for (I = 1; I <= J - 1; I++)
 439:                  {
 440:                      WORK[I + o_work] = Z[I + o_z] / (this._dlamc3.Run(DSIGMA[I + o_dsigma], DSIGJ) - DIFLJ) / (DSIGMA[I + o_dsigma] + DJ);
 441:                  }
 442:                  for (I = J + 1; I <= K; I++)
 443:                  {
 444:                      WORK[I + o_work] = Z[I + o_z] / (this._dlamc3.Run(DSIGMA[I + o_dsigma], DSIGJP) + DIFRJ) / (DSIGMA[I + o_dsigma] + DJ);
 445:                  }
 446:                  TEMP = this._dnrm2.Run(K, WORK, offset_work, 1);
 447:                  WORK[IWK2I + J + o_work] = this._ddot.Run(K, WORK, offset_work, 1, VF, offset_vf, 1) / TEMP;
 448:                  WORK[IWK3I + J + o_work] = this._ddot.Run(K, WORK, offset_work, 1, VL, offset_vl, 1) / TEMP;
 449:                  if (ICOMPQ == 1)
 450:                  {
 451:                      DIFR[J+2 * LDDIFR + o_difr] = TEMP;
 452:                  }
 453:              }
 454:              // *
 455:              this._dcopy.Run(K, WORK, IWK2 + o_work, 1, ref VF, offset_vf, 1);
 456:              this._dcopy.Run(K, WORK, IWK3 + o_work, 1, ref VL, offset_vl, 1);
 457:              // *
 458:              return;
 459:              // *
 460:              // *     End of DLASD8
 461:              // *
 462:   
 463:              #endregion
 464:   
 465:          }
 466:      }
 467:  }