sspsvx.c
Go to the documentation of this file.
00001 /* sspsvx.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 sspsvx_(char *fact, char *uplo, integer *n, integer *
00021         nrhs, real *ap, real *afp, integer *ipiv, real *b, integer *ldb, real 
00022         *x, integer *ldx, real *rcond, real *ferr, real *berr, real *work, 
00023         integer *iwork, integer *info)
00024 {
00025     /* System generated locals */
00026     integer b_dim1, b_offset, x_dim1, x_offset, i__1;
00027 
00028     /* Local variables */
00029     extern logical lsame_(char *, char *);
00030     real anorm;
00031     extern /* Subroutine */ int scopy_(integer *, real *, integer *, real *, 
00032             integer *);
00033     extern doublereal slamch_(char *);
00034     logical nofact;
00035     extern /* Subroutine */ int xerbla_(char *, integer *), slacpy_(
00036             char *, integer *, integer *, real *, integer *, real *, integer *
00037 );
00038     extern doublereal slansp_(char *, char *, integer *, real *, real *);
00039     extern /* Subroutine */ int sspcon_(char *, integer *, real *, integer *, 
00040             real *, real *, real *, integer *, integer *), ssprfs_(
00041             char *, integer *, integer *, real *, real *, integer *, real *, 
00042             integer *, real *, integer *, real *, real *, real *, integer *, 
00043             integer *), ssptrf_(char *, integer *, real *, integer *, 
00044             integer *), ssptrs_(char *, integer *, integer *, real *, 
00045             integer *, real *, integer *, integer *);
00046 
00047 
00048 /*  -- LAPACK driver routine (version 3.2) -- */
00049 /*     Univ. of Tennessee, Univ. of California Berkeley and NAG Ltd.. */
00050 /*     November 2006 */
00051 
00052 /*     .. Scalar Arguments .. */
00053 /*     .. */
00054 /*     .. Array Arguments .. */
00055 /*     .. */
00056 
00057 /*  Purpose */
00058 /*  ======= */
00059 
00060 /*  SSPSVX uses the diagonal pivoting factorization A = U*D*U**T or */
00061 /*  A = L*D*L**T to compute the solution to a real system of linear */
00062 /*  equations A * X = B, where A is an N-by-N symmetric matrix stored */
00063 /*  in packed format and X and B are N-by-NRHS matrices. */
00064 
00065 /*  Error bounds on the solution and a condition estimate are also */
00066 /*  provided. */
00067 
00068 /*  Description */
00069 /*  =========== */
00070 
00071 /*  The following steps are performed: */
00072 
00073 /*  1. If FACT = 'N', the diagonal pivoting method is used to factor A as */
00074 /*        A = U * D * U**T,  if UPLO = 'U', or */
00075 /*        A = L * D * L**T,  if UPLO = 'L', */
00076 /*     where U (or L) is a product of permutation and unit upper (lower) */
00077 /*     triangular matrices and D is symmetric and block diagonal with */
00078 /*     1-by-1 and 2-by-2 diagonal blocks. */
00079 
00080 /*  2. If some D(i,i)=0, so that D is exactly singular, then the routine */
00081 /*     returns with INFO = i. Otherwise, the factored form of A is used */
00082 /*     to estimate the condition number of the matrix A.  If the */
00083 /*     reciprocal of the condition number is less than machine precision, */
00084 /*     INFO = N+1 is returned as a warning, but the routine still goes on */
00085 /*     to solve for X and compute error bounds as described below. */
00086 
00087 /*  3. The system of equations is solved for X using the factored form */
00088 /*     of A. */
00089 
00090 /*  4. Iterative refinement is applied to improve the computed solution */
00091 /*     matrix and calculate error bounds and backward error estimates */
00092 /*     for it. */
00093 
00094 /*  Arguments */
00095 /*  ========= */
00096 
00097 /*  FACT    (input) CHARACTER*1 */
00098 /*          Specifies whether or not the factored form of A has been */
00099 /*          supplied on entry. */
00100 /*          = 'F':  On entry, AFP and IPIV contain the factored form of */
00101 /*                  A.  AP, AFP and IPIV will not be modified. */
00102 /*          = 'N':  The matrix A will be copied to AFP and factored. */
00103 
00104 /*  UPLO    (input) CHARACTER*1 */
00105 /*          = 'U':  Upper triangle of A is stored; */
00106 /*          = 'L':  Lower triangle of A is stored. */
00107 
00108 /*  N       (input) INTEGER */
00109 /*          The number of linear equations, i.e., the order of the */
00110 /*          matrix A.  N >= 0. */
00111 
00112 /*  NRHS    (input) INTEGER */
00113 /*          The number of right hand sides, i.e., the number of columns */
00114 /*          of the matrices B and X.  NRHS >= 0. */
00115 
00116 /*  AP      (input) REAL array, dimension (N*(N+1)/2) */
00117 /*          The upper or lower triangle of the symmetric matrix A, packed */
00118 /*          columnwise in a linear array.  The j-th column of A is stored */
00119 /*          in the array AP as follows: */
00120 /*          if UPLO = 'U', AP(i + (j-1)*j/2) = A(i,j) for 1<=i<=j; */
00121 /*          if UPLO = 'L', AP(i + (j-1)*(2*n-j)/2) = A(i,j) for j<=i<=n. */
00122 /*          See below for further details. */
00123 
00124 /*  AFP     (input or output) REAL array, dimension */
00125 /*                            (N*(N+1)/2) */
00126 /*          If FACT = 'F', then AFP is an input argument and on entry */
00127 /*          contains the block diagonal matrix D and the multipliers used */
00128 /*          to obtain the factor U or L from the factorization */
00129 /*          A = U*D*U**T or A = L*D*L**T as computed by SSPTRF, stored as */
00130 /*          a packed triangular matrix in the same storage format as A. */
00131 
00132 /*          If FACT = 'N', then AFP is an output argument and on exit */
00133 /*          contains the block diagonal matrix D and the multipliers used */
00134 /*          to obtain the factor U or L from the factorization */
00135 /*          A = U*D*U**T or A = L*D*L**T as computed by SSPTRF, stored as */
00136 /*          a packed triangular matrix in the same storage format as A. */
00137 
00138 /*  IPIV    (input or output) INTEGER array, dimension (N) */
00139 /*          If FACT = 'F', then IPIV is an input argument and on entry */
00140 /*          contains details of the interchanges and the block structure */
00141 /*          of D, as determined by SSPTRF. */
00142 /*          If IPIV(k) > 0, then rows and columns k and IPIV(k) were */
00143 /*          interchanged and D(k,k) is a 1-by-1 diagonal block. */
00144 /*          If UPLO = 'U' and IPIV(k) = IPIV(k-1) < 0, then rows and */
00145 /*          columns k-1 and -IPIV(k) were interchanged and D(k-1:k,k-1:k) */
00146 /*          is a 2-by-2 diagonal block.  If UPLO = 'L' and IPIV(k) = */
00147 /*          IPIV(k+1) < 0, then rows and columns k+1 and -IPIV(k) were */
00148 /*          interchanged and D(k:k+1,k:k+1) is a 2-by-2 diagonal block. */
00149 
00150 /*          If FACT = 'N', then IPIV is an output argument and on exit */
00151 /*          contains details of the interchanges and the block structure */
00152 /*          of D, as determined by SSPTRF. */
00153 
00154 /*  B       (input) REAL array, dimension (LDB,NRHS) */
00155 /*          The N-by-NRHS right hand side matrix B. */
00156 
00157 /*  LDB     (input) INTEGER */
00158 /*          The leading dimension of the array B.  LDB >= max(1,N). */
00159 
00160 /*  X       (output) REAL array, dimension (LDX,NRHS) */
00161 /*          If INFO = 0 or INFO = N+1, the N-by-NRHS solution matrix X. */
00162 
00163 /*  LDX     (input) INTEGER */
00164 /*          The leading dimension of the array X.  LDX >= max(1,N). */
00165 
00166 /*  RCOND   (output) REAL */
00167 /*          The estimate of the reciprocal condition number of the matrix */
00168 /*          A.  If RCOND is less than the machine precision (in */
00169 /*          particular, if RCOND = 0), the matrix is singular to working */
00170 /*          precision.  This condition is indicated by a return code of */
00171 /*          INFO > 0. */
00172 
00173 /*  FERR    (output) REAL array, dimension (NRHS) */
00174 /*          The estimated forward error bound for each solution vector */
00175 /*          X(j) (the j-th column of the solution matrix X). */
00176 /*          If XTRUE is the true solution corresponding to X(j), FERR(j) */
00177 /*          is an estimated upper bound for the magnitude of the largest */
00178 /*          element in (X(j) - XTRUE) divided by the magnitude of the */
00179 /*          largest element in X(j).  The estimate is as reliable as */
00180 /*          the estimate for RCOND, and is almost always a slight */
00181 /*          overestimate of the true error. */
00182 
00183 /*  BERR    (output) REAL array, dimension (NRHS) */
00184 /*          The componentwise relative backward error of each solution */
00185 /*          vector X(j) (i.e., the smallest relative change in */
00186 /*          any element of A or B that makes X(j) an exact solution). */
00187 
00188 /*  WORK    (workspace) REAL array, dimension (3*N) */
00189 
00190 /*  IWORK   (workspace) INTEGER array, dimension (N) */
00191 
00192 /*  INFO    (output) INTEGER */
00193 /*          = 0: successful exit */
00194 /*          < 0: if INFO = -i, the i-th argument had an illegal value */
00195 /*          > 0:  if INFO = i, and i is */
00196 /*                <= N:  D(i,i) is exactly zero.  The factorization */
00197 /*                       has been completed but the factor D is exactly */
00198 /*                       singular, so the solution and error bounds could */
00199 /*                       not be computed. RCOND = 0 is returned. */
00200 /*                = N+1: D is nonsingular, but RCOND is less than machine */
00201 /*                       precision, meaning that the matrix is singular */
00202 /*                       to working precision.  Nevertheless, the */
00203 /*                       solution and error bounds are computed because */
00204 /*                       there are a number of situations where the */
00205 /*                       computed solution can be more accurate than the */
00206 /*                       value of RCOND would suggest. */
00207 
00208 /*  Further Details */
00209 /*  =============== */
00210 
00211 /*  The packed storage scheme is illustrated by the following example */
00212 /*  when N = 4, UPLO = 'U': */
00213 
00214 /*  Two-dimensional storage of the symmetric matrix A: */
00215 
00216 /*     a11 a12 a13 a14 */
00217 /*         a22 a23 a24 */
00218 /*             a33 a34     (aij = aji) */
00219 /*                 a44 */
00220 
00221 /*  Packed storage of the upper triangle of A: */
00222 
00223 /*  AP = [ a11, a12, a22, a13, a23, a33, a14, a24, a34, a44 ] */
00224 
00225 /*  ===================================================================== */
00226 
00227 /*     .. Parameters .. */
00228 /*     .. */
00229 /*     .. Local Scalars .. */
00230 /*     .. */
00231 /*     .. External Functions .. */
00232 /*     .. */
00233 /*     .. External Subroutines .. */
00234 /*     .. */
00235 /*     .. Intrinsic Functions .. */
00236 /*     .. */
00237 /*     .. Executable Statements .. */
00238 
00239 /*     Test the input parameters. */
00240 
00241     /* Parameter adjustments */
00242     --ap;
00243     --afp;
00244     --ipiv;
00245     b_dim1 = *ldb;
00246     b_offset = 1 + b_dim1;
00247     b -= b_offset;
00248     x_dim1 = *ldx;
00249     x_offset = 1 + x_dim1;
00250     x -= x_offset;
00251     --ferr;
00252     --berr;
00253     --work;
00254     --iwork;
00255 
00256     /* Function Body */
00257     *info = 0;
00258     nofact = lsame_(fact, "N");
00259     if (! nofact && ! lsame_(fact, "F")) {
00260         *info = -1;
00261     } else if (! lsame_(uplo, "U") && ! lsame_(uplo, 
00262             "L")) {
00263         *info = -2;
00264     } else if (*n < 0) {
00265         *info = -3;
00266     } else if (*nrhs < 0) {
00267         *info = -4;
00268     } else if (*ldb < max(1,*n)) {
00269         *info = -9;
00270     } else if (*ldx < max(1,*n)) {
00271         *info = -11;
00272     }
00273     if (*info != 0) {
00274         i__1 = -(*info);
00275         xerbla_("SSPSVX", &i__1);
00276         return 0;
00277     }
00278 
00279     if (nofact) {
00280 
00281 /*        Compute the factorization A = U*D*U' or A = L*D*L'. */
00282 
00283         i__1 = *n * (*n + 1) / 2;
00284         scopy_(&i__1, &ap[1], &c__1, &afp[1], &c__1);
00285         ssptrf_(uplo, n, &afp[1], &ipiv[1], info);
00286 
00287 /*        Return if INFO is non-zero. */
00288 
00289         if (*info > 0) {
00290             *rcond = 0.f;
00291             return 0;
00292         }
00293     }
00294 
00295 /*     Compute the norm of the matrix A. */
00296 
00297     anorm = slansp_("I", uplo, n, &ap[1], &work[1]);
00298 
00299 /*     Compute the reciprocal of the condition number of A. */
00300 
00301     sspcon_(uplo, n, &afp[1], &ipiv[1], &anorm, rcond, &work[1], &iwork[1], 
00302             info);
00303 
00304 /*     Compute the solution vectors X. */
00305 
00306     slacpy_("Full", n, nrhs, &b[b_offset], ldb, &x[x_offset], ldx);
00307     ssptrs_(uplo, n, nrhs, &afp[1], &ipiv[1], &x[x_offset], ldx, info);
00308 
00309 /*     Use iterative refinement to improve the computed solutions and */
00310 /*     compute error bounds and backward error estimates for them. */
00311 
00312     ssprfs_(uplo, n, nrhs, &ap[1], &afp[1], &ipiv[1], &b[b_offset], ldb, &x[
00313             x_offset], ldx, &ferr[1], &berr[1], &work[1], &iwork[1], info);
00314 
00315 /*     Set INFO = N+1 if the matrix is singular to working precision. */
00316 
00317     if (*rcond < slamch_("Epsilon")) {
00318         *info = *n + 1;
00319     }
00320 
00321     return 0;
00322 
00323 /*     End of SSPSVX */
00324 
00325 } /* sspsvx_ */


swiftnav
Author(s):
autogenerated on Sat Jun 8 2019 18:56:13