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:      /// DLAQP2 computes a QR factorization with column pivoting of
  27:      /// the block A(OFFSET+1:M,1:N).
  28:      /// The block A(1:OFFSET,1:N) is accordingly pivoted, but not factorized.
  29:      /// 
  30:      ///</summary>
  31:      public class DLAQP2
  32:      {
  33:      
  34:   
  35:          #region Dependencies
  36:          
  37:          DLARF _dlarf; DLARFG _dlarfg; DSWAP _dswap; IDAMAX _idamax; DLAMCH _dlamch; DNRM2 _dnrm2; 
  38:   
  39:          #endregion
  40:   
  41:   
  42:          #region Fields
  43:          
  44:          const double ZERO = 0.0E+0; const double ONE = 1.0E+0; int I = 0; int ITEMP = 0; int J = 0; int MN = 0; int OFFPI = 0; 
  45:          int PVT = 0;double AII = 0; double TEMP = 0; double TEMP2 = 0; double TOL3Z = 0; 
  46:   
  47:          #endregion
  48:   
  49:          public DLAQP2(DLARF dlarf, DLARFG dlarfg, DSWAP dswap, IDAMAX idamax, DLAMCH dlamch, DNRM2 dnrm2)
  50:          {
  51:      
  52:   
  53:              #region Set Dependencies
  54:              
  55:              this._dlarf = dlarf; this._dlarfg = dlarfg; this._dswap = dswap; this._idamax = idamax; this._dlamch = dlamch; 
  56:              this._dnrm2 = dnrm2;
  57:   
  58:              #endregion
  59:   
  60:          }
  61:      
  62:          public DLAQP2()
  63:          {
  64:      
  65:   
  66:              #region Dependencies (Initialization)
  67:              
  68:              LSAME lsame = new LSAME();
  69:              XERBLA xerbla = new XERBLA();
  70:              DLAMC3 dlamc3 = new DLAMC3();
  71:              DLAPY2 dlapy2 = new DLAPY2();
  72:              DNRM2 dnrm2 = new DNRM2();
  73:              DSCAL dscal = new DSCAL();
  74:              DSWAP dswap = new DSWAP();
  75:              IDAMAX idamax = new IDAMAX();
  76:              DGEMV dgemv = new DGEMV(lsame, xerbla);
  77:              DGER dger = new DGER(xerbla);
  78:              DLARF dlarf = new DLARF(dgemv, dger, lsame);
  79:              DLAMC1 dlamc1 = new DLAMC1(dlamc3);
  80:              DLAMC4 dlamc4 = new DLAMC4(dlamc3);
  81:              DLAMC5 dlamc5 = new DLAMC5(dlamc3);
  82:              DLAMC2 dlamc2 = new DLAMC2(dlamc3, dlamc1, dlamc4, dlamc5);
  83:              DLAMCH dlamch = new DLAMCH(lsame, dlamc2);
  84:              DLARFG dlarfg = new DLARFG(dlamch, dlapy2, dnrm2, dscal);
  85:   
  86:              #endregion
  87:   
  88:   
  89:              #region Set Dependencies
  90:              
  91:              this._dlarf = dlarf; this._dlarfg = dlarfg; this._dswap = dswap; this._idamax = idamax; this._dlamch = dlamch; 
  92:              this._dnrm2 = dnrm2;
  93:   
  94:              #endregion
  95:   
  96:          }
  97:          /// <summary>
  98:          /// Purpose
  99:          /// =======
 100:          /// 
 101:          /// DLAQP2 computes a QR factorization with column pivoting of
 102:          /// the block A(OFFSET+1:M,1:N).
 103:          /// The block A(1:OFFSET,1:N) is accordingly pivoted, but not factorized.
 104:          /// 
 105:          ///</summary>
 106:          /// <param name="M">
 107:          /// (input) INTEGER
 108:          /// The number of rows of the matrix A. M .GE. 0.
 109:          ///</param>
 110:          /// <param name="N">
 111:          /// (input) INTEGER
 112:          /// The number of columns of the matrix A. N .GE. 0.
 113:          ///</param>
 114:          /// <param name="OFFSET">
 115:          /// (input) INTEGER
 116:          /// The number of rows of the matrix A that must be pivoted
 117:          /// but no factorized. OFFSET .GE. 0.
 118:          ///</param>
 119:          /// <param name="A">
 120:          /// (input/output) DOUBLE PRECISION array, dimension (LDA,N)
 121:          /// On entry, the M-by-N matrix A.
 122:          /// On exit, the upper triangle of block A(OFFSET+1:M,1:N) is 
 123:          /// the triangular factor obtained; the elements in block
 124:          /// A(OFFSET+1:M,1:N) below the diagonal, together with the
 125:          /// array TAU, represent the orthogonal matrix Q as a product of
 126:          /// elementary reflectors. Block A(1:OFFSET,1:N) has been
 127:          /// accordingly pivoted, but no factorized.
 128:          ///</param>
 129:          /// <param name="LDA">
 130:          /// (input) INTEGER
 131:          /// The leading dimension of the array A. LDA .GE. max(1,M).
 132:          ///</param>
 133:          /// <param name="JPVT">
 134:          /// (input/output) INTEGER array, dimension (N)
 135:          /// On entry, if JPVT(i) .ne. 0, the i-th column of A is permuted
 136:          /// to the front of A*P (a leading column); if JPVT(i) = 0,
 137:          /// the i-th column of A is a free column.
 138:          /// On exit, if JPVT(i) = k, then the i-th column of A*P
 139:          /// was the k-th column of A.
 140:          ///</param>
 141:          /// <param name="TAU">
 142:          /// (output) DOUBLE PRECISION array, dimension (min(M,N))
 143:          /// The scalar factors of the elementary reflectors.
 144:          ///</param>
 145:          /// <param name="VN1">
 146:          /// (input/output) DOUBLE PRECISION array, dimension (N)
 147:          /// The vector with the partial column norms.
 148:          ///</param>
 149:          /// <param name="VN2">
 150:          /// (input/output) DOUBLE PRECISION array, dimension (N)
 151:          /// The vector with the exact column norms.
 152:          ///</param>
 153:          /// <param name="WORK">
 154:          /// (workspace) DOUBLE PRECISION array, dimension (N)
 155:          ///</param>
 156:          public void Run(int M, int N, int OFFSET, ref double[] A, int offset_a, int LDA, ref int[] JPVT, int offset_jpvt
 157:                           , ref double[] TAU, int offset_tau, ref double[] VN1, int offset_vn1, ref double[] VN2, int offset_vn2, ref double[] WORK, int offset_work)
 158:          {
 159:   
 160:              #region Array Index Correction
 161:              
 162:               int o_a = -1 - LDA + offset_a;  int o_jpvt = -1 + offset_jpvt;  int o_tau = -1 + offset_tau; 
 163:               int o_vn1 = -1 + offset_vn1; int o_vn2 = -1 + offset_vn2;  int o_work = -1 + offset_work; 
 164:   
 165:              #endregion
 166:   
 167:   
 168:              #region Prolog
 169:              
 170:              // *
 171:              // *  -- LAPACK auxiliary routine (version 3.1) --
 172:              // *     Univ. of Tennessee, Univ. of California Berkeley and NAG Ltd..
 173:              // *     November 2006
 174:              // *
 175:              // *     .. Scalar Arguments ..
 176:              // *     ..
 177:              // *     .. Array Arguments ..
 178:              // *     ..
 179:              // *
 180:              // *  Purpose
 181:              // *  =======
 182:              // *
 183:              // *  DLAQP2 computes a QR factorization with column pivoting of
 184:              // *  the block A(OFFSET+1:M,1:N).
 185:              // *  The block A(1:OFFSET,1:N) is accordingly pivoted, but not factorized.
 186:              // *
 187:              // *  Arguments
 188:              // *  =========
 189:              // *
 190:              // *  M       (input) INTEGER
 191:              // *          The number of rows of the matrix A. M >= 0.
 192:              // *
 193:              // *  N       (input) INTEGER
 194:              // *          The number of columns of the matrix A. N >= 0.
 195:              // *
 196:              // *  OFFSET  (input) INTEGER
 197:              // *          The number of rows of the matrix A that must be pivoted
 198:              // *          but no factorized. OFFSET >= 0.
 199:              // *
 200:              // *  A       (input/output) DOUBLE PRECISION array, dimension (LDA,N)
 201:              // *          On entry, the M-by-N matrix A.
 202:              // *          On exit, the upper triangle of block A(OFFSET+1:M,1:N) is 
 203:              // *          the triangular factor obtained; the elements in block
 204:              // *          A(OFFSET+1:M,1:N) below the diagonal, together with the
 205:              // *          array TAU, represent the orthogonal matrix Q as a product of
 206:              // *          elementary reflectors. Block A(1:OFFSET,1:N) has been
 207:              // *          accordingly pivoted, but no factorized.
 208:              // *
 209:              // *  LDA     (input) INTEGER
 210:              // *          The leading dimension of the array A. LDA >= max(1,M).
 211:              // *
 212:              // *  JPVT    (input/output) INTEGER array, dimension (N)
 213:              // *          On entry, if JPVT(i) .ne. 0, the i-th column of A is permuted
 214:              // *          to the front of A*P (a leading column); if JPVT(i) = 0,
 215:              // *          the i-th column of A is a free column.
 216:              // *          On exit, if JPVT(i) = k, then the i-th column of A*P
 217:              // *          was the k-th column of A.
 218:              // *
 219:              // *  TAU     (output) DOUBLE PRECISION array, dimension (min(M,N))
 220:              // *          The scalar factors of the elementary reflectors.
 221:              // *
 222:              // *  VN1     (input/output) DOUBLE PRECISION array, dimension (N)
 223:              // *          The vector with the partial column norms.
 224:              // *
 225:              // *  VN2     (input/output) DOUBLE PRECISION array, dimension (N)
 226:              // *          The vector with the exact column norms.
 227:              // *
 228:              // *  WORK    (workspace) DOUBLE PRECISION array, dimension (N)
 229:              // *
 230:              // *  Further Details
 231:              // *  ===============
 232:              // *
 233:              // *  Based on contributions by
 234:              // *    G. Quintana-Orti, Depto. de Informatica, Universidad Jaime I, Spain
 235:              // *    X. Sun, Computer Science Dept., Duke University, USA
 236:              // *
 237:              // *  Partial column norm updating strategy modified by
 238:              // *    Z. Drmac and Z. Bujanovic, Dept. of Mathematics,
 239:              // *    University of Zagreb, Croatia.
 240:              // *    June 2006.
 241:              // *  For more details see LAPACK Working Note 176.
 242:              // *  =====================================================================
 243:              // *
 244:              // *     .. Parameters ..
 245:              // *     ..
 246:              // *     .. Local Scalars ..
 247:              // *     ..
 248:              // *     .. External Subroutines ..
 249:              // *     ..
 250:              // *     .. Intrinsic Functions ..
 251:              //      INTRINSIC          ABS, MAX, MIN, SQRT;
 252:              // *     ..
 253:              // *     .. External Functions ..
 254:              // *     ..
 255:              // *     .. Executable Statements ..
 256:              // *
 257:   
 258:              #endregion
 259:   
 260:   
 261:              #region Body
 262:              
 263:              MN = Math.Min(M - OFFSET, N);
 264:              TOL3Z = Math.Sqrt(this._dlamch.Run("Epsilon"));
 265:              // *
 266:              // *     Compute factorization.
 267:              // *
 268:              for (I = 1; I <= MN; I++)
 269:              {
 270:                  // *
 271:                  OFFPI = OFFSET + I;
 272:                  // *
 273:                  // *        Determine ith pivot column and swap if necessary.
 274:                  // *
 275:                  PVT = (I - 1) + this._idamax.Run(N - I + 1, VN1, I + o_vn1, 1);
 276:                  // *
 277:                  if (PVT != I)
 278:                  {
 279:                      this._dswap.Run(M, ref A, 1+PVT * LDA + o_a, 1, ref A, 1+I * LDA + o_a, 1);
 280:                      ITEMP = JPVT[PVT + o_jpvt];
 281:                      JPVT[PVT + o_jpvt] = JPVT[I + o_jpvt];
 282:                      JPVT[I + o_jpvt] = ITEMP;
 283:                      VN1[PVT + o_vn1] = VN1[I + o_vn1];
 284:                      VN2[PVT + o_vn2] = VN2[I + o_vn2];
 285:                  }
 286:                  // *
 287:                  // *        Generate elementary reflector H(i).
 288:                  // *
 289:                  if (OFFPI < M)
 290:                  {
 291:                      this._dlarfg.Run(M - OFFPI + 1, ref A[OFFPI+I * LDA + o_a], ref A, OFFPI + 1+I * LDA + o_a, 1, ref TAU[I + o_tau]);
 292:                  }
 293:                  else
 294:                  {
 295:                      this._dlarfg.Run(1, ref A[M+I * LDA + o_a], ref A, M+I * LDA + o_a, 1, ref TAU[I + o_tau]);
 296:                  }
 297:                  // *
 298:                  if (I < N)
 299:                  {
 300:                      // *
 301:                      // *           Apply H(i)' to A(offset+i:m,i+1:n) from the left.
 302:                      // *
 303:                      AII = A[OFFPI+I * LDA + o_a];
 304:                      A[OFFPI+I * LDA + o_a] = ONE;
 305:                      this._dlarf.Run("Left", M - OFFPI + 1, N - I, A, OFFPI+I * LDA + o_a, 1, TAU[I + o_tau]
 306:                                      , ref A, OFFPI+(I + 1) * LDA + o_a, LDA, ref WORK, 1 + o_work);
 307:                      A[OFFPI+I * LDA + o_a] = AII;
 308:                  }
 309:                  // *
 310:                  // *        Update partial column norms.
 311:                  // *
 312:                  for (J = I + 1; J <= N; J++)
 313:                  {
 314:                      if (VN1[J + o_vn1] != ZERO)
 315:                      {
 316:                          // *
 317:                          // *              NOTE: The following 4 lines follow from the analysis in
 318:                          // *              Lapack Working Note 176.
 319:                          // *
 320:                          TEMP = ONE - Math.Pow(Math.Abs(A[OFFPI+J * LDA + o_a]) / VN1[J + o_vn1],2);
 321:                          TEMP = Math.Max(TEMP, ZERO);
 322:                          TEMP2 = TEMP * Math.Pow(VN1[J + o_vn1] / VN2[J + o_vn2],2);
 323:                          if (TEMP2 <= TOL3Z)
 324:                          {
 325:                              if (OFFPI < M)
 326:                              {
 327:                                  VN1[J + o_vn1] = this._dnrm2.Run(M - OFFPI, A, OFFPI + 1+J * LDA + o_a, 1);
 328:                                  VN2[J + o_vn2] = VN1[J + o_vn1];
 329:                              }
 330:                              else
 331:                              {
 332:                                  VN1[J + o_vn1] = ZERO;
 333:                                  VN2[J + o_vn2] = ZERO;
 334:                              }
 335:                          }
 336:                          else
 337:                          {
 338:                              VN1[J + o_vn1] = VN1[J + o_vn1] * Math.Sqrt(TEMP);
 339:                          }
 340:                      }
 341:                  }
 342:                  // *
 343:              }
 344:              // *
 345:              return;
 346:              // *
 347:              // *     End of DLAQP2
 348:              // *
 349:   
 350:              #endregion
 351:   
 352:          }
 353:      }
 354:  }