2010-07-16 14:54:53 +02:00
|
|
|
/* dlasyf.f -- translated by f2c (version 20061008).
|
|
|
|
You must link the resulting object file with libf2c:
|
|
|
|
on Microsoft Windows system, link with libf2c.lib;
|
|
|
|
on Linux or Unix systems, link with .../path/to/libf2c.a -lm
|
|
|
|
or, if you install libf2c.a in a standard place, with -lf2c -lm
|
|
|
|
-- in that order, at the end of the command line, as in
|
|
|
|
cc *.o -lf2c -lm
|
|
|
|
Source for libf2c is in /netlib/f2c/libf2c.zip, e.g.,
|
|
|
|
|
|
|
|
http://www.netlib.org/f2c/libf2c.zip
|
|
|
|
*/
|
|
|
|
|
2010-05-11 19:44:00 +02:00
|
|
|
#include "clapack.h"
|
|
|
|
|
2010-07-16 14:54:53 +02:00
|
|
|
|
|
|
|
/* Table of constant values */
|
|
|
|
|
|
|
|
static integer c__1 = 1;
|
|
|
|
static doublereal c_b8 = -1.;
|
|
|
|
static doublereal c_b9 = 1.;
|
|
|
|
|
|
|
|
/* Subroutine */ int dlasyf_(char *uplo, integer *n, integer *nb, integer *kb,
|
2010-05-11 19:44:00 +02:00
|
|
|
doublereal *a, integer *lda, integer *ipiv, doublereal *w, integer *
|
|
|
|
ldw, integer *info)
|
|
|
|
{
|
|
|
|
/* System generated locals */
|
|
|
|
integer a_dim1, a_offset, w_dim1, w_offset, i__1, i__2, i__3, i__4, i__5;
|
|
|
|
doublereal d__1, d__2, d__3;
|
2010-07-16 14:54:53 +02:00
|
|
|
|
2010-05-11 19:44:00 +02:00
|
|
|
/* Builtin functions */
|
|
|
|
double sqrt(doublereal);
|
2010-07-16 14:54:53 +02:00
|
|
|
|
2010-05-11 19:44:00 +02:00
|
|
|
/* Local variables */
|
2010-07-16 14:54:53 +02:00
|
|
|
integer j, k;
|
|
|
|
doublereal t, r1, d11, d21, d22;
|
|
|
|
integer jb, jj, kk, jp, kp, kw, kkw, imax, jmax;
|
|
|
|
doublereal alpha;
|
2010-05-11 19:44:00 +02:00
|
|
|
extern /* Subroutine */ int dscal_(integer *, doublereal *, doublereal *,
|
|
|
|
integer *), dgemm_(char *, char *, integer *, integer *, integer *
|
|
|
|
, doublereal *, doublereal *, integer *, doublereal *, integer *,
|
|
|
|
doublereal *, doublereal *, integer *);
|
|
|
|
extern logical lsame_(char *, char *);
|
|
|
|
extern /* Subroutine */ int dgemv_(char *, integer *, integer *,
|
|
|
|
doublereal *, doublereal *, integer *, doublereal *, integer *,
|
|
|
|
doublereal *, doublereal *, integer *), dcopy_(integer *,
|
|
|
|
doublereal *, integer *, doublereal *, integer *), dswap_(integer
|
|
|
|
*, doublereal *, integer *, doublereal *, integer *);
|
2010-07-16 14:54:53 +02:00
|
|
|
integer kstep;
|
|
|
|
doublereal absakk;
|
2010-05-11 19:44:00 +02:00
|
|
|
extern integer idamax_(integer *, doublereal *, integer *);
|
2010-07-16 14:54:53 +02:00
|
|
|
doublereal colmax, rowmax;
|
|
|
|
|
|
|
|
|
|
|
|
/* -- LAPACK routine (version 3.2) -- */
|
|
|
|
/* Univ. of Tennessee, Univ. of California Berkeley and NAG Ltd.. */
|
|
|
|
/* November 2006 */
|
|
|
|
|
|
|
|
/* .. Scalar Arguments .. */
|
|
|
|
/* .. */
|
|
|
|
/* .. Array Arguments .. */
|
|
|
|
/* .. */
|
|
|
|
|
|
|
|
/* Purpose */
|
|
|
|
/* ======= */
|
|
|
|
|
|
|
|
/* DLASYF computes a partial factorization of a real symmetric matrix A */
|
|
|
|
/* using the Bunch-Kaufman diagonal pivoting method. The partial */
|
|
|
|
/* factorization has the form: */
|
|
|
|
|
|
|
|
/* A = ( I U12 ) ( A11 0 ) ( I 0 ) if UPLO = 'U', or: */
|
|
|
|
/* ( 0 U22 ) ( 0 D ) ( U12' U22' ) */
|
|
|
|
|
|
|
|
/* A = ( L11 0 ) ( D 0 ) ( L11' L21' ) if UPLO = 'L' */
|
|
|
|
/* ( L21 I ) ( 0 A22 ) ( 0 I ) */
|
|
|
|
|
|
|
|
/* where the order of D is at most NB. The actual order is returned in */
|
|
|
|
/* the argument KB, and is either NB or NB-1, or N if N <= NB. */
|
|
|
|
|
|
|
|
/* DLASYF is an auxiliary routine called by DSYTRF. It uses blocked code */
|
|
|
|
/* (calling Level 3 BLAS) to update the submatrix A11 (if UPLO = 'U') or */
|
|
|
|
/* A22 (if UPLO = 'L'). */
|
|
|
|
|
|
|
|
/* Arguments */
|
|
|
|
/* ========= */
|
|
|
|
|
|
|
|
/* UPLO (input) CHARACTER*1 */
|
|
|
|
/* Specifies whether the upper or lower triangular part of the */
|
|
|
|
/* symmetric matrix A is stored: */
|
|
|
|
/* = 'U': Upper triangular */
|
|
|
|
/* = 'L': Lower triangular */
|
|
|
|
|
|
|
|
/* N (input) INTEGER */
|
|
|
|
/* The order of the matrix A. N >= 0. */
|
|
|
|
|
|
|
|
/* NB (input) INTEGER */
|
|
|
|
/* The maximum number of columns of the matrix A that should be */
|
|
|
|
/* factored. NB should be at least 2 to allow for 2-by-2 pivot */
|
|
|
|
/* blocks. */
|
|
|
|
|
|
|
|
/* KB (output) INTEGER */
|
|
|
|
/* The number of columns of A that were actually factored. */
|
|
|
|
/* KB is either NB-1 or NB, or N if N <= NB. */
|
|
|
|
|
|
|
|
/* A (input/output) DOUBLE PRECISION array, dimension (LDA,N) */
|
|
|
|
/* On entry, the symmetric matrix A. If UPLO = 'U', the leading */
|
|
|
|
/* n-by-n upper triangular part of A contains the upper */
|
|
|
|
/* triangular part of the matrix A, and the strictly lower */
|
|
|
|
/* triangular part of A is not referenced. If UPLO = 'L', the */
|
|
|
|
/* leading n-by-n lower triangular part of A contains the lower */
|
|
|
|
/* triangular part of the matrix A, and the strictly upper */
|
|
|
|
/* triangular part of A is not referenced. */
|
|
|
|
/* On exit, A contains details of the partial factorization. */
|
|
|
|
|
|
|
|
/* LDA (input) INTEGER */
|
|
|
|
/* The leading dimension of the array A. LDA >= max(1,N). */
|
|
|
|
|
|
|
|
/* IPIV (output) INTEGER array, dimension (N) */
|
|
|
|
/* Details of the interchanges and the block structure of D. */
|
|
|
|
/* If UPLO = 'U', only the last KB elements of IPIV are set; */
|
|
|
|
/* if UPLO = 'L', only the first KB elements are set. */
|
|
|
|
|
|
|
|
/* If IPIV(k) > 0, then rows and columns k and IPIV(k) were */
|
|
|
|
/* interchanged and D(k,k) is a 1-by-1 diagonal block. */
|
|
|
|
/* If UPLO = 'U' and IPIV(k) = IPIV(k-1) < 0, then rows and */
|
|
|
|
/* columns k-1 and -IPIV(k) were interchanged and D(k-1:k,k-1:k) */
|
|
|
|
/* is a 2-by-2 diagonal block. If UPLO = 'L' and IPIV(k) = */
|
|
|
|
/* IPIV(k+1) < 0, then rows and columns k+1 and -IPIV(k) were */
|
|
|
|
/* interchanged and D(k:k+1,k:k+1) is a 2-by-2 diagonal block. */
|
|
|
|
|
|
|
|
/* W (workspace) DOUBLE PRECISION array, dimension (LDW,NB) */
|
|
|
|
|
|
|
|
/* LDW (input) INTEGER */
|
|
|
|
/* The leading dimension of the array W. LDW >= max(1,N). */
|
|
|
|
|
|
|
|
/* INFO (output) INTEGER */
|
|
|
|
/* = 0: successful exit */
|
|
|
|
/* > 0: if INFO = k, D(k,k) is exactly zero. The factorization */
|
|
|
|
/* has been completed, but the block diagonal matrix D is */
|
|
|
|
/* exactly singular. */
|
|
|
|
|
|
|
|
/* ===================================================================== */
|
|
|
|
|
|
|
|
/* .. Parameters .. */
|
|
|
|
/* .. */
|
|
|
|
/* .. Local Scalars .. */
|
|
|
|
/* .. */
|
|
|
|
/* .. External Functions .. */
|
|
|
|
/* .. */
|
|
|
|
/* .. External Subroutines .. */
|
|
|
|
/* .. */
|
|
|
|
/* .. Intrinsic Functions .. */
|
|
|
|
/* .. */
|
|
|
|
/* .. Executable Statements .. */
|
|
|
|
|
|
|
|
/* Parameter adjustments */
|
2010-05-11 19:44:00 +02:00
|
|
|
a_dim1 = *lda;
|
|
|
|
a_offset = 1 + a_dim1;
|
|
|
|
a -= a_offset;
|
|
|
|
--ipiv;
|
|
|
|
w_dim1 = *ldw;
|
|
|
|
w_offset = 1 + w_dim1;
|
|
|
|
w -= w_offset;
|
|
|
|
|
|
|
|
/* Function Body */
|
|
|
|
*info = 0;
|
|
|
|
|
|
|
|
/* Initialize ALPHA for use in choosing pivot block size. */
|
|
|
|
|
|
|
|
alpha = (sqrt(17.) + 1.) / 8.;
|
|
|
|
|
|
|
|
if (lsame_(uplo, "U")) {
|
|
|
|
|
2010-07-16 14:54:53 +02:00
|
|
|
/* Factorize the trailing columns of A using the upper triangle */
|
|
|
|
/* of A and working backwards, and compute the matrix W = U12*D */
|
|
|
|
/* for use in updating A11 */
|
2010-05-11 19:44:00 +02:00
|
|
|
|
2010-07-16 14:54:53 +02:00
|
|
|
/* K is the main loop index, decreasing from N in steps of 1 or 2 */
|
2010-05-11 19:44:00 +02:00
|
|
|
|
2010-07-16 14:54:53 +02:00
|
|
|
/* KW is the column of W which corresponds to column K of A */
|
2010-05-11 19:44:00 +02:00
|
|
|
|
|
|
|
k = *n;
|
|
|
|
L10:
|
|
|
|
kw = *nb + k - *n;
|
|
|
|
|
|
|
|
/* Exit from loop */
|
|
|
|
|
|
|
|
if (k <= *n - *nb + 1 && *nb < *n || k < 1) {
|
|
|
|
goto L30;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Copy column K of A to column KW of W and update it */
|
|
|
|
|
|
|
|
dcopy_(&k, &a[k * a_dim1 + 1], &c__1, &w[kw * w_dim1 + 1], &c__1);
|
|
|
|
if (k < *n) {
|
|
|
|
i__1 = *n - k;
|
|
|
|
dgemv_("No transpose", &k, &i__1, &c_b8, &a[(k + 1) * a_dim1 + 1],
|
|
|
|
lda, &w[k + (kw + 1) * w_dim1], ldw, &c_b9, &w[kw *
|
|
|
|
w_dim1 + 1], &c__1);
|
|
|
|
}
|
|
|
|
|
|
|
|
kstep = 1;
|
|
|
|
|
2010-07-16 14:54:53 +02:00
|
|
|
/* Determine rows and columns to be interchanged and whether */
|
|
|
|
/* a 1-by-1 or 2-by-2 pivot block will be used */
|
2010-05-11 19:44:00 +02:00
|
|
|
|
|
|
|
absakk = (d__1 = w[k + kw * w_dim1], abs(d__1));
|
|
|
|
|
2010-07-16 14:54:53 +02:00
|
|
|
/* IMAX is the row-index of the largest off-diagonal element in */
|
|
|
|
/* column K, and COLMAX is its absolute value */
|
2010-05-11 19:44:00 +02:00
|
|
|
|
|
|
|
if (k > 1) {
|
|
|
|
i__1 = k - 1;
|
|
|
|
imax = idamax_(&i__1, &w[kw * w_dim1 + 1], &c__1);
|
|
|
|
colmax = (d__1 = w[imax + kw * w_dim1], abs(d__1));
|
|
|
|
} else {
|
|
|
|
colmax = 0.;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (max(absakk,colmax) == 0.) {
|
|
|
|
|
|
|
|
/* Column K is zero: set INFO and continue */
|
|
|
|
|
|
|
|
if (*info == 0) {
|
|
|
|
*info = k;
|
|
|
|
}
|
|
|
|
kp = k;
|
|
|
|
} else {
|
|
|
|
if (absakk >= alpha * colmax) {
|
|
|
|
|
|
|
|
/* no interchange, use 1-by-1 pivot block */
|
|
|
|
|
|
|
|
kp = k;
|
|
|
|
} else {
|
|
|
|
|
|
|
|
/* Copy column IMAX to column KW-1 of W and update it */
|
|
|
|
|
|
|
|
dcopy_(&imax, &a[imax * a_dim1 + 1], &c__1, &w[(kw - 1) *
|
|
|
|
w_dim1 + 1], &c__1);
|
|
|
|
i__1 = k - imax;
|
|
|
|
dcopy_(&i__1, &a[imax + (imax + 1) * a_dim1], lda, &w[imax +
|
|
|
|
1 + (kw - 1) * w_dim1], &c__1);
|
|
|
|
if (k < *n) {
|
|
|
|
i__1 = *n - k;
|
|
|
|
dgemv_("No transpose", &k, &i__1, &c_b8, &a[(k + 1) *
|
|
|
|
a_dim1 + 1], lda, &w[imax + (kw + 1) * w_dim1],
|
|
|
|
ldw, &c_b9, &w[(kw - 1) * w_dim1 + 1], &c__1);
|
|
|
|
}
|
|
|
|
|
2010-07-16 14:54:53 +02:00
|
|
|
/* JMAX is the column-index of the largest off-diagonal */
|
|
|
|
/* element in row IMAX, and ROWMAX is its absolute value */
|
2010-05-11 19:44:00 +02:00
|
|
|
|
|
|
|
i__1 = k - imax;
|
|
|
|
jmax = imax + idamax_(&i__1, &w[imax + 1 + (kw - 1) * w_dim1],
|
|
|
|
&c__1);
|
|
|
|
rowmax = (d__1 = w[jmax + (kw - 1) * w_dim1], abs(d__1));
|
|
|
|
if (imax > 1) {
|
|
|
|
i__1 = imax - 1;
|
|
|
|
jmax = idamax_(&i__1, &w[(kw - 1) * w_dim1 + 1], &c__1);
|
|
|
|
/* Computing MAX */
|
|
|
|
d__2 = rowmax, d__3 = (d__1 = w[jmax + (kw - 1) * w_dim1],
|
|
|
|
abs(d__1));
|
|
|
|
rowmax = max(d__2,d__3);
|
|
|
|
}
|
|
|
|
|
|
|
|
if (absakk >= alpha * colmax * (colmax / rowmax)) {
|
|
|
|
|
|
|
|
/* no interchange, use 1-by-1 pivot block */
|
|
|
|
|
|
|
|
kp = k;
|
|
|
|
} else if ((d__1 = w[imax + (kw - 1) * w_dim1], abs(d__1)) >=
|
|
|
|
alpha * rowmax) {
|
|
|
|
|
2010-07-16 14:54:53 +02:00
|
|
|
/* interchange rows and columns K and IMAX, use 1-by-1 */
|
|
|
|
/* pivot block */
|
2010-05-11 19:44:00 +02:00
|
|
|
|
|
|
|
kp = imax;
|
|
|
|
|
|
|
|
/* copy column KW-1 of W to column KW */
|
|
|
|
|
|
|
|
dcopy_(&k, &w[(kw - 1) * w_dim1 + 1], &c__1, &w[kw *
|
|
|
|
w_dim1 + 1], &c__1);
|
|
|
|
} else {
|
|
|
|
|
2010-07-16 14:54:53 +02:00
|
|
|
/* interchange rows and columns K-1 and IMAX, use 2-by-2 */
|
|
|
|
/* pivot block */
|
2010-05-11 19:44:00 +02:00
|
|
|
|
|
|
|
kp = imax;
|
|
|
|
kstep = 2;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
kk = k - kstep + 1;
|
|
|
|
kkw = *nb + kk - *n;
|
|
|
|
|
|
|
|
/* Updated column KP is already stored in column KKW of W */
|
|
|
|
|
|
|
|
if (kp != kk) {
|
|
|
|
|
|
|
|
/* Copy non-updated column KK to column KP */
|
|
|
|
|
|
|
|
a[kp + k * a_dim1] = a[kk + k * a_dim1];
|
|
|
|
i__1 = k - 1 - kp;
|
|
|
|
dcopy_(&i__1, &a[kp + 1 + kk * a_dim1], &c__1, &a[kp + (kp +
|
|
|
|
1) * a_dim1], lda);
|
|
|
|
dcopy_(&kp, &a[kk * a_dim1 + 1], &c__1, &a[kp * a_dim1 + 1], &
|
|
|
|
c__1);
|
|
|
|
|
|
|
|
/* Interchange rows KK and KP in last KK columns of A and W */
|
|
|
|
|
|
|
|
i__1 = *n - kk + 1;
|
|
|
|
dswap_(&i__1, &a[kk + kk * a_dim1], lda, &a[kp + kk * a_dim1],
|
|
|
|
lda);
|
|
|
|
i__1 = *n - kk + 1;
|
|
|
|
dswap_(&i__1, &w[kk + kkw * w_dim1], ldw, &w[kp + kkw *
|
|
|
|
w_dim1], ldw);
|
|
|
|
}
|
|
|
|
|
|
|
|
if (kstep == 1) {
|
|
|
|
|
2010-07-16 14:54:53 +02:00
|
|
|
/* 1-by-1 pivot block D(k): column KW of W now holds */
|
2010-05-11 19:44:00 +02:00
|
|
|
|
2010-07-16 14:54:53 +02:00
|
|
|
/* W(k) = U(k)*D(k) */
|
2010-05-11 19:44:00 +02:00
|
|
|
|
2010-07-16 14:54:53 +02:00
|
|
|
/* where U(k) is the k-th column of U */
|
2010-05-11 19:44:00 +02:00
|
|
|
|
2010-07-16 14:54:53 +02:00
|
|
|
/* Store U(k) in column k of A */
|
2010-05-11 19:44:00 +02:00
|
|
|
|
|
|
|
dcopy_(&k, &w[kw * w_dim1 + 1], &c__1, &a[k * a_dim1 + 1], &
|
|
|
|
c__1);
|
|
|
|
r1 = 1. / a[k + k * a_dim1];
|
|
|
|
i__1 = k - 1;
|
|
|
|
dscal_(&i__1, &r1, &a[k * a_dim1 + 1], &c__1);
|
|
|
|
} else {
|
|
|
|
|
2010-07-16 14:54:53 +02:00
|
|
|
/* 2-by-2 pivot block D(k): columns KW and KW-1 of W now */
|
|
|
|
/* hold */
|
2010-05-11 19:44:00 +02:00
|
|
|
|
2010-07-16 14:54:53 +02:00
|
|
|
/* ( W(k-1) W(k) ) = ( U(k-1) U(k) )*D(k) */
|
2010-05-11 19:44:00 +02:00
|
|
|
|
2010-07-16 14:54:53 +02:00
|
|
|
/* where U(k) and U(k-1) are the k-th and (k-1)-th columns */
|
|
|
|
/* of U */
|
2010-05-11 19:44:00 +02:00
|
|
|
|
|
|
|
if (k > 2) {
|
|
|
|
|
|
|
|
/* Store U(k) and U(k-1) in columns k and k-1 of A */
|
|
|
|
|
|
|
|
d21 = w[k - 1 + kw * w_dim1];
|
|
|
|
d11 = w[k + kw * w_dim1] / d21;
|
|
|
|
d22 = w[k - 1 + (kw - 1) * w_dim1] / d21;
|
|
|
|
t = 1. / (d11 * d22 - 1.);
|
|
|
|
d21 = t / d21;
|
|
|
|
i__1 = k - 2;
|
|
|
|
for (j = 1; j <= i__1; ++j) {
|
|
|
|
a[j + (k - 1) * a_dim1] = d21 * (d11 * w[j + (kw - 1)
|
|
|
|
* w_dim1] - w[j + kw * w_dim1]);
|
|
|
|
a[j + k * a_dim1] = d21 * (d22 * w[j + kw * w_dim1] -
|
|
|
|
w[j + (kw - 1) * w_dim1]);
|
|
|
|
/* L20: */
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Copy D(k) to A */
|
|
|
|
|
|
|
|
a[k - 1 + (k - 1) * a_dim1] = w[k - 1 + (kw - 1) * w_dim1];
|
|
|
|
a[k - 1 + k * a_dim1] = w[k - 1 + kw * w_dim1];
|
|
|
|
a[k + k * a_dim1] = w[k + kw * w_dim1];
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Store details of the interchanges in IPIV */
|
|
|
|
|
|
|
|
if (kstep == 1) {
|
|
|
|
ipiv[k] = kp;
|
|
|
|
} else {
|
|
|
|
ipiv[k] = -kp;
|
|
|
|
ipiv[k - 1] = -kp;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Decrease K and return to the start of the main loop */
|
|
|
|
|
|
|
|
k -= kstep;
|
|
|
|
goto L10;
|
|
|
|
|
|
|
|
L30:
|
|
|
|
|
2010-07-16 14:54:53 +02:00
|
|
|
/* Update the upper triangle of A11 (= A(1:k,1:k)) as */
|
2010-05-11 19:44:00 +02:00
|
|
|
|
2010-07-16 14:54:53 +02:00
|
|
|
/* A11 := A11 - U12*D*U12' = A11 - U12*W' */
|
2010-05-11 19:44:00 +02:00
|
|
|
|
2010-07-16 14:54:53 +02:00
|
|
|
/* computing blocks of NB columns at a time */
|
2010-05-11 19:44:00 +02:00
|
|
|
|
|
|
|
i__1 = -(*nb);
|
|
|
|
for (j = (k - 1) / *nb * *nb + 1; i__1 < 0 ? j >= 1 : j <= 1; j +=
|
|
|
|
i__1) {
|
|
|
|
/* Computing MIN */
|
|
|
|
i__2 = *nb, i__3 = k - j + 1;
|
|
|
|
jb = min(i__2,i__3);
|
|
|
|
|
|
|
|
/* Update the upper triangle of the diagonal block */
|
|
|
|
|
|
|
|
i__2 = j + jb - 1;
|
|
|
|
for (jj = j; jj <= i__2; ++jj) {
|
|
|
|
i__3 = jj - j + 1;
|
|
|
|
i__4 = *n - k;
|
|
|
|
dgemv_("No transpose", &i__3, &i__4, &c_b8, &a[j + (k + 1) *
|
|
|
|
a_dim1], lda, &w[jj + (kw + 1) * w_dim1], ldw, &c_b9,
|
|
|
|
&a[j + jj * a_dim1], &c__1);
|
|
|
|
/* L40: */
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Update the rectangular superdiagonal block */
|
|
|
|
|
|
|
|
i__2 = j - 1;
|
|
|
|
i__3 = *n - k;
|
|
|
|
dgemm_("No transpose", "Transpose", &i__2, &jb, &i__3, &c_b8, &a[(
|
|
|
|
k + 1) * a_dim1 + 1], lda, &w[j + (kw + 1) * w_dim1], ldw,
|
|
|
|
&c_b9, &a[j * a_dim1 + 1], lda);
|
|
|
|
/* L50: */
|
|
|
|
}
|
|
|
|
|
2010-07-16 14:54:53 +02:00
|
|
|
/* Put U12 in standard form by partially undoing the interchanges */
|
|
|
|
/* in columns k+1:n */
|
2010-05-11 19:44:00 +02:00
|
|
|
|
|
|
|
j = k + 1;
|
|
|
|
L60:
|
|
|
|
jj = j;
|
|
|
|
jp = ipiv[j];
|
|
|
|
if (jp < 0) {
|
|
|
|
jp = -jp;
|
|
|
|
++j;
|
|
|
|
}
|
|
|
|
++j;
|
|
|
|
if (jp != jj && j <= *n) {
|
|
|
|
i__1 = *n - j + 1;
|
|
|
|
dswap_(&i__1, &a[jp + j * a_dim1], lda, &a[jj + j * a_dim1], lda);
|
|
|
|
}
|
|
|
|
if (j <= *n) {
|
|
|
|
goto L60;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Set KB to the number of columns factorized */
|
|
|
|
|
|
|
|
*kb = *n - k;
|
|
|
|
|
|
|
|
} else {
|
|
|
|
|
2010-07-16 14:54:53 +02:00
|
|
|
/* Factorize the leading columns of A using the lower triangle */
|
|
|
|
/* of A and working forwards, and compute the matrix W = L21*D */
|
|
|
|
/* for use in updating A22 */
|
2010-05-11 19:44:00 +02:00
|
|
|
|
2010-07-16 14:54:53 +02:00
|
|
|
/* K is the main loop index, increasing from 1 in steps of 1 or 2 */
|
2010-05-11 19:44:00 +02:00
|
|
|
|
|
|
|
k = 1;
|
|
|
|
L70:
|
|
|
|
|
|
|
|
/* Exit from loop */
|
|
|
|
|
|
|
|
if (k >= *nb && *nb < *n || k > *n) {
|
|
|
|
goto L90;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Copy column K of A to column K of W and update it */
|
|
|
|
|
|
|
|
i__1 = *n - k + 1;
|
|
|
|
dcopy_(&i__1, &a[k + k * a_dim1], &c__1, &w[k + k * w_dim1], &c__1);
|
|
|
|
i__1 = *n - k + 1;
|
|
|
|
i__2 = k - 1;
|
|
|
|
dgemv_("No transpose", &i__1, &i__2, &c_b8, &a[k + a_dim1], lda, &w[k
|
|
|
|
+ w_dim1], ldw, &c_b9, &w[k + k * w_dim1], &c__1);
|
|
|
|
|
|
|
|
kstep = 1;
|
|
|
|
|
2010-07-16 14:54:53 +02:00
|
|
|
/* Determine rows and columns to be interchanged and whether */
|
|
|
|
/* a 1-by-1 or 2-by-2 pivot block will be used */
|
2010-05-11 19:44:00 +02:00
|
|
|
|
|
|
|
absakk = (d__1 = w[k + k * w_dim1], abs(d__1));
|
|
|
|
|
2010-07-16 14:54:53 +02:00
|
|
|
/* IMAX is the row-index of the largest off-diagonal element in */
|
|
|
|
/* column K, and COLMAX is its absolute value */
|
2010-05-11 19:44:00 +02:00
|
|
|
|
|
|
|
if (k < *n) {
|
|
|
|
i__1 = *n - k;
|
|
|
|
imax = k + idamax_(&i__1, &w[k + 1 + k * w_dim1], &c__1);
|
|
|
|
colmax = (d__1 = w[imax + k * w_dim1], abs(d__1));
|
|
|
|
} else {
|
|
|
|
colmax = 0.;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (max(absakk,colmax) == 0.) {
|
|
|
|
|
|
|
|
/* Column K is zero: set INFO and continue */
|
|
|
|
|
|
|
|
if (*info == 0) {
|
|
|
|
*info = k;
|
|
|
|
}
|
|
|
|
kp = k;
|
|
|
|
} else {
|
|
|
|
if (absakk >= alpha * colmax) {
|
|
|
|
|
|
|
|
/* no interchange, use 1-by-1 pivot block */
|
|
|
|
|
|
|
|
kp = k;
|
|
|
|
} else {
|
|
|
|
|
|
|
|
/* Copy column IMAX to column K+1 of W and update it */
|
|
|
|
|
|
|
|
i__1 = imax - k;
|
|
|
|
dcopy_(&i__1, &a[imax + k * a_dim1], lda, &w[k + (k + 1) *
|
|
|
|
w_dim1], &c__1);
|
|
|
|
i__1 = *n - imax + 1;
|
|
|
|
dcopy_(&i__1, &a[imax + imax * a_dim1], &c__1, &w[imax + (k +
|
|
|
|
1) * w_dim1], &c__1);
|
|
|
|
i__1 = *n - k + 1;
|
|
|
|
i__2 = k - 1;
|
|
|
|
dgemv_("No transpose", &i__1, &i__2, &c_b8, &a[k + a_dim1],
|
|
|
|
lda, &w[imax + w_dim1], ldw, &c_b9, &w[k + (k + 1) *
|
|
|
|
w_dim1], &c__1);
|
|
|
|
|
2010-07-16 14:54:53 +02:00
|
|
|
/* JMAX is the column-index of the largest off-diagonal */
|
|
|
|
/* element in row IMAX, and ROWMAX is its absolute value */
|
2010-05-11 19:44:00 +02:00
|
|
|
|
|
|
|
i__1 = imax - k;
|
|
|
|
jmax = k - 1 + idamax_(&i__1, &w[k + (k + 1) * w_dim1], &c__1)
|
|
|
|
;
|
|
|
|
rowmax = (d__1 = w[jmax + (k + 1) * w_dim1], abs(d__1));
|
|
|
|
if (imax < *n) {
|
|
|
|
i__1 = *n - imax;
|
|
|
|
jmax = imax + idamax_(&i__1, &w[imax + 1 + (k + 1) *
|
|
|
|
w_dim1], &c__1);
|
|
|
|
/* Computing MAX */
|
|
|
|
d__2 = rowmax, d__3 = (d__1 = w[jmax + (k + 1) * w_dim1],
|
|
|
|
abs(d__1));
|
|
|
|
rowmax = max(d__2,d__3);
|
|
|
|
}
|
|
|
|
|
|
|
|
if (absakk >= alpha * colmax * (colmax / rowmax)) {
|
|
|
|
|
|
|
|
/* no interchange, use 1-by-1 pivot block */
|
|
|
|
|
|
|
|
kp = k;
|
|
|
|
} else if ((d__1 = w[imax + (k + 1) * w_dim1], abs(d__1)) >=
|
|
|
|
alpha * rowmax) {
|
|
|
|
|
2010-07-16 14:54:53 +02:00
|
|
|
/* interchange rows and columns K and IMAX, use 1-by-1 */
|
|
|
|
/* pivot block */
|
2010-05-11 19:44:00 +02:00
|
|
|
|
|
|
|
kp = imax;
|
|
|
|
|
|
|
|
/* copy column K+1 of W to column K */
|
|
|
|
|
|
|
|
i__1 = *n - k + 1;
|
|
|
|
dcopy_(&i__1, &w[k + (k + 1) * w_dim1], &c__1, &w[k + k *
|
|
|
|
w_dim1], &c__1);
|
|
|
|
} else {
|
|
|
|
|
2010-07-16 14:54:53 +02:00
|
|
|
/* interchange rows and columns K+1 and IMAX, use 2-by-2 */
|
|
|
|
/* pivot block */
|
2010-05-11 19:44:00 +02:00
|
|
|
|
|
|
|
kp = imax;
|
|
|
|
kstep = 2;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
kk = k + kstep - 1;
|
|
|
|
|
|
|
|
/* Updated column KP is already stored in column KK of W */
|
|
|
|
|
|
|
|
if (kp != kk) {
|
|
|
|
|
|
|
|
/* Copy non-updated column KK to column KP */
|
|
|
|
|
|
|
|
a[kp + k * a_dim1] = a[kk + k * a_dim1];
|
|
|
|
i__1 = kp - k - 1;
|
|
|
|
dcopy_(&i__1, &a[k + 1 + kk * a_dim1], &c__1, &a[kp + (k + 1)
|
|
|
|
* a_dim1], lda);
|
|
|
|
i__1 = *n - kp + 1;
|
|
|
|
dcopy_(&i__1, &a[kp + kk * a_dim1], &c__1, &a[kp + kp *
|
|
|
|
a_dim1], &c__1);
|
|
|
|
|
|
|
|
/* Interchange rows KK and KP in first KK columns of A and W */
|
|
|
|
|
|
|
|
dswap_(&kk, &a[kk + a_dim1], lda, &a[kp + a_dim1], lda);
|
|
|
|
dswap_(&kk, &w[kk + w_dim1], ldw, &w[kp + w_dim1], ldw);
|
|
|
|
}
|
|
|
|
|
|
|
|
if (kstep == 1) {
|
|
|
|
|
2010-07-16 14:54:53 +02:00
|
|
|
/* 1-by-1 pivot block D(k): column k of W now holds */
|
2010-05-11 19:44:00 +02:00
|
|
|
|
2010-07-16 14:54:53 +02:00
|
|
|
/* W(k) = L(k)*D(k) */
|
2010-05-11 19:44:00 +02:00
|
|
|
|
2010-07-16 14:54:53 +02:00
|
|
|
/* where L(k) is the k-th column of L */
|
2010-05-11 19:44:00 +02:00
|
|
|
|
2010-07-16 14:54:53 +02:00
|
|
|
/* Store L(k) in column k of A */
|
2010-05-11 19:44:00 +02:00
|
|
|
|
|
|
|
i__1 = *n - k + 1;
|
|
|
|
dcopy_(&i__1, &w[k + k * w_dim1], &c__1, &a[k + k * a_dim1], &
|
|
|
|
c__1);
|
|
|
|
if (k < *n) {
|
|
|
|
r1 = 1. / a[k + k * a_dim1];
|
|
|
|
i__1 = *n - k;
|
|
|
|
dscal_(&i__1, &r1, &a[k + 1 + k * a_dim1], &c__1);
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
|
2010-07-16 14:54:53 +02:00
|
|
|
/* 2-by-2 pivot block D(k): columns k and k+1 of W now hold */
|
2010-05-11 19:44:00 +02:00
|
|
|
|
2010-07-16 14:54:53 +02:00
|
|
|
/* ( W(k) W(k+1) ) = ( L(k) L(k+1) )*D(k) */
|
2010-05-11 19:44:00 +02:00
|
|
|
|
2010-07-16 14:54:53 +02:00
|
|
|
/* where L(k) and L(k+1) are the k-th and (k+1)-th columns */
|
|
|
|
/* of L */
|
2010-05-11 19:44:00 +02:00
|
|
|
|
|
|
|
if (k < *n - 1) {
|
|
|
|
|
|
|
|
/* Store L(k) and L(k+1) in columns k and k+1 of A */
|
|
|
|
|
|
|
|
d21 = w[k + 1 + k * w_dim1];
|
|
|
|
d11 = w[k + 1 + (k + 1) * w_dim1] / d21;
|
|
|
|
d22 = w[k + k * w_dim1] / d21;
|
|
|
|
t = 1. / (d11 * d22 - 1.);
|
|
|
|
d21 = t / d21;
|
|
|
|
i__1 = *n;
|
|
|
|
for (j = k + 2; j <= i__1; ++j) {
|
|
|
|
a[j + k * a_dim1] = d21 * (d11 * w[j + k * w_dim1] -
|
|
|
|
w[j + (k + 1) * w_dim1]);
|
|
|
|
a[j + (k + 1) * a_dim1] = d21 * (d22 * w[j + (k + 1) *
|
|
|
|
w_dim1] - w[j + k * w_dim1]);
|
|
|
|
/* L80: */
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Copy D(k) to A */
|
|
|
|
|
|
|
|
a[k + k * a_dim1] = w[k + k * w_dim1];
|
|
|
|
a[k + 1 + k * a_dim1] = w[k + 1 + k * w_dim1];
|
|
|
|
a[k + 1 + (k + 1) * a_dim1] = w[k + 1 + (k + 1) * w_dim1];
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Store details of the interchanges in IPIV */
|
|
|
|
|
|
|
|
if (kstep == 1) {
|
|
|
|
ipiv[k] = kp;
|
|
|
|
} else {
|
|
|
|
ipiv[k] = -kp;
|
|
|
|
ipiv[k + 1] = -kp;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Increase K and return to the start of the main loop */
|
|
|
|
|
|
|
|
k += kstep;
|
|
|
|
goto L70;
|
|
|
|
|
|
|
|
L90:
|
|
|
|
|
2010-07-16 14:54:53 +02:00
|
|
|
/* Update the lower triangle of A22 (= A(k:n,k:n)) as */
|
2010-05-11 19:44:00 +02:00
|
|
|
|
2010-07-16 14:54:53 +02:00
|
|
|
/* A22 := A22 - L21*D*L21' = A22 - L21*W' */
|
2010-05-11 19:44:00 +02:00
|
|
|
|
2010-07-16 14:54:53 +02:00
|
|
|
/* computing blocks of NB columns at a time */
|
2010-05-11 19:44:00 +02:00
|
|
|
|
|
|
|
i__1 = *n;
|
|
|
|
i__2 = *nb;
|
|
|
|
for (j = k; i__2 < 0 ? j >= i__1 : j <= i__1; j += i__2) {
|
|
|
|
/* Computing MIN */
|
|
|
|
i__3 = *nb, i__4 = *n - j + 1;
|
|
|
|
jb = min(i__3,i__4);
|
|
|
|
|
|
|
|
/* Update the lower triangle of the diagonal block */
|
|
|
|
|
|
|
|
i__3 = j + jb - 1;
|
|
|
|
for (jj = j; jj <= i__3; ++jj) {
|
|
|
|
i__4 = j + jb - jj;
|
|
|
|
i__5 = k - 1;
|
|
|
|
dgemv_("No transpose", &i__4, &i__5, &c_b8, &a[jj + a_dim1],
|
|
|
|
lda, &w[jj + w_dim1], ldw, &c_b9, &a[jj + jj * a_dim1]
|
|
|
|
, &c__1);
|
|
|
|
/* L100: */
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Update the rectangular subdiagonal block */
|
|
|
|
|
|
|
|
if (j + jb <= *n) {
|
|
|
|
i__3 = *n - j - jb + 1;
|
|
|
|
i__4 = k - 1;
|
|
|
|
dgemm_("No transpose", "Transpose", &i__3, &jb, &i__4, &c_b8,
|
|
|
|
&a[j + jb + a_dim1], lda, &w[j + w_dim1], ldw, &c_b9,
|
|
|
|
&a[j + jb + j * a_dim1], lda);
|
|
|
|
}
|
|
|
|
/* L110: */
|
|
|
|
}
|
|
|
|
|
2010-07-16 14:54:53 +02:00
|
|
|
/* Put L21 in standard form by partially undoing the interchanges */
|
|
|
|
/* in columns 1:k-1 */
|
2010-05-11 19:44:00 +02:00
|
|
|
|
|
|
|
j = k - 1;
|
|
|
|
L120:
|
|
|
|
jj = j;
|
|
|
|
jp = ipiv[j];
|
|
|
|
if (jp < 0) {
|
|
|
|
jp = -jp;
|
|
|
|
--j;
|
|
|
|
}
|
|
|
|
--j;
|
|
|
|
if (jp != jj && j >= 1) {
|
|
|
|
dswap_(&j, &a[jp + a_dim1], lda, &a[jj + a_dim1], lda);
|
|
|
|
}
|
|
|
|
if (j >= 1) {
|
|
|
|
goto L120;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Set KB to the number of columns factorized */
|
|
|
|
|
|
|
|
*kb = k - 1;
|
|
|
|
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
/* End of DLASYF */
|
|
|
|
|
|
|
|
} /* dlasyf_ */
|