dlaed9.c
Go to the documentation of this file.
00001 /* dlaed9.f -- translated by f2c (version 20061008).
00002    You must link the resulting object file with libf2c:
00003         on Microsoft Windows system, link with libf2c.lib;
00004         on Linux or Unix systems, link with .../path/to/libf2c.a -lm
00005         or, if you install libf2c.a in a standard place, with -lf2c -lm
00006         -- in that order, at the end of the command line, as in
00007                 cc *.o -lf2c -lm
00008         Source for libf2c is in /netlib/f2c/libf2c.zip, e.g.,
00009 
00010                 http://www.netlib.org/f2c/libf2c.zip
00011 */
00012 
00013 #include "f2c.h"
00014 #include "blaswrap.h"
00015 
00016 /* Table of constant values */
00017 
00018 static integer c__1 = 1;
00019 
00020 /* Subroutine */ int dlaed9_(integer *k, integer *kstart, integer *kstop, 
00021         integer *n, doublereal *d__, doublereal *q, integer *ldq, doublereal *
00022         rho, doublereal *dlamda, doublereal *w, doublereal *s, integer *lds, 
00023         integer *info)
00024 {
00025     /* System generated locals */
00026     integer q_dim1, q_offset, s_dim1, s_offset, i__1, i__2;
00027     doublereal d__1;
00028 
00029     /* Builtin functions */
00030     double sqrt(doublereal), d_sign(doublereal *, doublereal *);
00031 
00032     /* Local variables */
00033     integer i__, j;
00034     doublereal temp;
00035     extern doublereal dnrm2_(integer *, doublereal *, integer *);
00036     extern /* Subroutine */ int dcopy_(integer *, doublereal *, integer *, 
00037             doublereal *, integer *), dlaed4_(integer *, integer *, 
00038             doublereal *, doublereal *, doublereal *, doublereal *, 
00039             doublereal *, integer *);
00040     extern doublereal dlamc3_(doublereal *, doublereal *);
00041     extern /* Subroutine */ int xerbla_(char *, integer *);
00042 
00043 
00044 /*  -- LAPACK routine (version 3.2) -- */
00045 /*     Univ. of Tennessee, Univ. of California Berkeley and NAG Ltd.. */
00046 /*     November 2006 */
00047 
00048 /*     .. Scalar Arguments .. */
00049 /*     .. */
00050 /*     .. Array Arguments .. */
00051 /*     .. */
00052 
00053 /*  Purpose */
00054 /*  ======= */
00055 
00056 /*  DLAED9 finds the roots of the secular equation, as defined by the */
00057 /*  values in D, Z, and RHO, between KSTART and KSTOP.  It makes the */
00058 /*  appropriate calls to DLAED4 and then stores the new matrix of */
00059 /*  eigenvectors for use in calculating the next level of Z vectors. */
00060 
00061 /*  Arguments */
00062 /*  ========= */
00063 
00064 /*  K       (input) INTEGER */
00065 /*          The number of terms in the rational function to be solved by */
00066 /*          DLAED4.  K >= 0. */
00067 
00068 /*  KSTART  (input) INTEGER */
00069 /*  KSTOP   (input) INTEGER */
00070 /*          The updated eigenvalues Lambda(I), KSTART <= I <= KSTOP */
00071 /*          are to be computed.  1 <= KSTART <= KSTOP <= K. */
00072 
00073 /*  N       (input) INTEGER */
00074 /*          The number of rows and columns in the Q matrix. */
00075 /*          N >= K (delation may result in N > K). */
00076 
00077 /*  D       (output) DOUBLE PRECISION array, dimension (N) */
00078 /*          D(I) contains the updated eigenvalues */
00079 /*          for KSTART <= I <= KSTOP. */
00080 
00081 /*  Q       (workspace) DOUBLE PRECISION array, dimension (LDQ,N) */
00082 
00083 /*  LDQ     (input) INTEGER */
00084 /*          The leading dimension of the array Q.  LDQ >= max( 1, N ). */
00085 
00086 /*  RHO     (input) DOUBLE PRECISION */
00087 /*          The value of the parameter in the rank one update equation. */
00088 /*          RHO >= 0 required. */
00089 
00090 /*  DLAMDA  (input) DOUBLE PRECISION array, dimension (K) */
00091 /*          The first K elements of this array contain the old roots */
00092 /*          of the deflated updating problem.  These are the poles */
00093 /*          of the secular equation. */
00094 
00095 /*  W       (input) DOUBLE PRECISION array, dimension (K) */
00096 /*          The first K elements of this array contain the components */
00097 /*          of the deflation-adjusted updating vector. */
00098 
00099 /*  S       (output) DOUBLE PRECISION array, dimension (LDS, K) */
00100 /*          Will contain the eigenvectors of the repaired matrix which */
00101 /*          will be stored for subsequent Z vector calculation and */
00102 /*          multiplied by the previously accumulated eigenvectors */
00103 /*          to update the system. */
00104 
00105 /*  LDS     (input) INTEGER */
00106 /*          The leading dimension of S.  LDS >= max( 1, K ). */
00107 
00108 /*  INFO    (output) INTEGER */
00109 /*          = 0:  successful exit. */
00110 /*          < 0:  if INFO = -i, the i-th argument had an illegal value. */
00111 /*          > 0:  if INFO = 1, an eigenvalue did not converge */
00112 
00113 /*  Further Details */
00114 /*  =============== */
00115 
00116 /*  Based on contributions by */
00117 /*     Jeff Rutter, Computer Science Division, University of California */
00118 /*     at Berkeley, USA */
00119 
00120 /*  ===================================================================== */
00121 
00122 /*     .. Local Scalars .. */
00123 /*     .. */
00124 /*     .. External Functions .. */
00125 /*     .. */
00126 /*     .. External Subroutines .. */
00127 /*     .. */
00128 /*     .. Intrinsic Functions .. */
00129 /*     .. */
00130 /*     .. Executable Statements .. */
00131 
00132 /*     Test the input parameters. */
00133 
00134     /* Parameter adjustments */
00135     --d__;
00136     q_dim1 = *ldq;
00137     q_offset = 1 + q_dim1;
00138     q -= q_offset;
00139     --dlamda;
00140     --w;
00141     s_dim1 = *lds;
00142     s_offset = 1 + s_dim1;
00143     s -= s_offset;
00144 
00145     /* Function Body */
00146     *info = 0;
00147 
00148     if (*k < 0) {
00149         *info = -1;
00150     } else if (*kstart < 1 || *kstart > max(1,*k)) {
00151         *info = -2;
00152     } else if (max(1,*kstop) < *kstart || *kstop > max(1,*k)) {
00153         *info = -3;
00154     } else if (*n < *k) {
00155         *info = -4;
00156     } else if (*ldq < max(1,*k)) {
00157         *info = -7;
00158     } else if (*lds < max(1,*k)) {
00159         *info = -12;
00160     }
00161     if (*info != 0) {
00162         i__1 = -(*info);
00163         xerbla_("DLAED9", &i__1);
00164         return 0;
00165     }
00166 
00167 /*     Quick return if possible */
00168 
00169     if (*k == 0) {
00170         return 0;
00171     }
00172 
00173 /*     Modify values DLAMDA(i) to make sure all DLAMDA(i)-DLAMDA(j) can */
00174 /*     be computed with high relative accuracy (barring over/underflow). */
00175 /*     This is a problem on machines without a guard digit in */
00176 /*     add/subtract (Cray XMP, Cray YMP, Cray C 90 and Cray 2). */
00177 /*     The following code replaces DLAMDA(I) by 2*DLAMDA(I)-DLAMDA(I), */
00178 /*     which on any of these machines zeros out the bottommost */
00179 /*     bit of DLAMDA(I) if it is 1; this makes the subsequent */
00180 /*     subtractions DLAMDA(I)-DLAMDA(J) unproblematic when cancellation */
00181 /*     occurs. On binary machines with a guard digit (almost all */
00182 /*     machines) it does not change DLAMDA(I) at all. On hexadecimal */
00183 /*     and decimal machines with a guard digit, it slightly */
00184 /*     changes the bottommost bits of DLAMDA(I). It does not account */
00185 /*     for hexadecimal or decimal machines without guard digits */
00186 /*     (we know of none). We use a subroutine call to compute */
00187 /*     2*DLAMBDA(I) to prevent optimizing compilers from eliminating */
00188 /*     this code. */
00189 
00190     i__1 = *n;
00191     for (i__ = 1; i__ <= i__1; ++i__) {
00192         dlamda[i__] = dlamc3_(&dlamda[i__], &dlamda[i__]) - dlamda[i__];
00193 /* L10: */
00194     }
00195 
00196     i__1 = *kstop;
00197     for (j = *kstart; j <= i__1; ++j) {
00198         dlaed4_(k, &j, &dlamda[1], &w[1], &q[j * q_dim1 + 1], rho, &d__[j], 
00199                 info);
00200 
00201 /*        If the zero finder fails, the computation is terminated. */
00202 
00203         if (*info != 0) {
00204             goto L120;
00205         }
00206 /* L20: */
00207     }
00208 
00209     if (*k == 1 || *k == 2) {
00210         i__1 = *k;
00211         for (i__ = 1; i__ <= i__1; ++i__) {
00212             i__2 = *k;
00213             for (j = 1; j <= i__2; ++j) {
00214                 s[j + i__ * s_dim1] = q[j + i__ * q_dim1];
00215 /* L30: */
00216             }
00217 /* L40: */
00218         }
00219         goto L120;
00220     }
00221 
00222 /*     Compute updated W. */
00223 
00224     dcopy_(k, &w[1], &c__1, &s[s_offset], &c__1);
00225 
00226 /*     Initialize W(I) = Q(I,I) */
00227 
00228     i__1 = *ldq + 1;
00229     dcopy_(k, &q[q_offset], &i__1, &w[1], &c__1);
00230     i__1 = *k;
00231     for (j = 1; j <= i__1; ++j) {
00232         i__2 = j - 1;
00233         for (i__ = 1; i__ <= i__2; ++i__) {
00234             w[i__] *= q[i__ + j * q_dim1] / (dlamda[i__] - dlamda[j]);
00235 /* L50: */
00236         }
00237         i__2 = *k;
00238         for (i__ = j + 1; i__ <= i__2; ++i__) {
00239             w[i__] *= q[i__ + j * q_dim1] / (dlamda[i__] - dlamda[j]);
00240 /* L60: */
00241         }
00242 /* L70: */
00243     }
00244     i__1 = *k;
00245     for (i__ = 1; i__ <= i__1; ++i__) {
00246         d__1 = sqrt(-w[i__]);
00247         w[i__] = d_sign(&d__1, &s[i__ + s_dim1]);
00248 /* L80: */
00249     }
00250 
00251 /*     Compute eigenvectors of the modified rank-1 modification. */
00252 
00253     i__1 = *k;
00254     for (j = 1; j <= i__1; ++j) {
00255         i__2 = *k;
00256         for (i__ = 1; i__ <= i__2; ++i__) {
00257             q[i__ + j * q_dim1] = w[i__] / q[i__ + j * q_dim1];
00258 /* L90: */
00259         }
00260         temp = dnrm2_(k, &q[j * q_dim1 + 1], &c__1);
00261         i__2 = *k;
00262         for (i__ = 1; i__ <= i__2; ++i__) {
00263             s[i__ + j * s_dim1] = q[i__ + j * q_dim1] / temp;
00264 /* L100: */
00265         }
00266 /* L110: */
00267     }
00268 
00269 L120:
00270     return 0;
00271 
00272 /*     End of DLAED9 */
00273 
00274 } /* dlaed9_ */


swiftnav
Author(s):
autogenerated on Sat Jun 8 2019 18:55:46