CFD/hypre
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
2dd7550c83
hypre/lapack/dsteqr.c

629 lines
15 KiB
C
Raw Blame History

/*BHEADER**********************************************************************
* Copyright (c) 2006 The Regents of the University of California.
* Produced at the Lawrence Livermore National Laboratory.
* Written by the HYPRE team. UCRL-CODE-222953.
* All rights reserved.
*
* This file is part of HYPRE (see http://www.llnl.gov/CASC/hypre/).
* Please see the COPYRIGHT_and_LICENSE file for the copyright notice,
* disclaimer, contact information and the GNU Lesser General Public License.
*
* HYPRE is free software; you can redistribute it and/or modify it under the
* terms of the GNU General Public License (as published by the Free Software
* Foundation) version 2.1 dated February 1999.
*
* HYPRE is distributed in the hope that it will be useful, but WITHOUT ANY
* WARRANTY; without even the IMPLIED WARRANTY OF MERCHANTABILITY or FITNESS
* FOR A PARTICULAR PURPOSE. See the terms and conditions of the GNU General
* Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public License
* along with this program; if not, write to the Free Software Foundation,
* Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*
* $Revision$
***********************************************************************EHEADER*/
#include "../blas/hypre_blas.h"
#include "hypre_lapack.h"
#include "f2c.h"
/* Subroutine */ int dsteqr_(char *compz, integer *n, doublereal *d__,
doublereal *e, doublereal *z__, integer *ldz, doublereal *work,
integer *info)
{
/* -- LAPACK routine (version 3.0) --
Univ. of Tennessee, Univ. of California Berkeley, NAG Ltd.,
Courant Institute, Argonne National Lab, and Rice University
September 30, 1994
Purpose
=======
DSTEQR computes all eigenvalues and, optionally, eigenvectors of a
symmetric tridiagonal matrix using the implicit QL or QR method.
The eigenvectors of a full or band symmetric matrix can also be found
if DSYTRD or DSPTRD or DSBTRD has been used to reduce this matrix to
tridiagonal form.
Arguments
=========
COMPZ (input) CHARACTER*1
= 'N': Compute eigenvalues only.
= 'V': Compute eigenvalues and eigenvectors of the original
symmetric matrix. On entry, Z must contain the
orthogonal matrix used to reduce the original matrix
to tridiagonal form.
= 'I': Compute eigenvalues and eigenvectors of the
tridiagonal matrix. Z is initialized to the identity
matrix.
N (input) INTEGER
The order of the matrix. N >= 0.
D (input/output) DOUBLE PRECISION array, dimension (N)
On entry, the diagonal elements of the tridiagonal matrix.
On exit, if INFO = 0, the eigenvalues in ascending order.
E (input/output) DOUBLE PRECISION array, dimension (N-1)
On entry, the (n-1) subdiagonal elements of the tridiagonal
matrix.
On exit, E has been destroyed.
Z (input/output) DOUBLE PRECISION array, dimension (LDZ, N)
On entry, if COMPZ = 'V', then Z contains the orthogonal
matrix used in the reduction to tridiagonal form.
On exit, if INFO = 0, then if COMPZ = 'V', Z contains the
orthonormal eigenvectors of the original symmetric matrix,
and if COMPZ = 'I', Z contains the orthonormal eigenvectors
of the symmetric tridiagonal matrix.
If COMPZ = 'N', then Z is not referenced.
LDZ (input) INTEGER
The leading dimension of the array Z. LDZ >= 1, and if
eigenvectors are desired, then LDZ >= max(1,N).
WORK (workspace) DOUBLE PRECISION array, dimension (max(1,2*N-2))
If COMPZ = 'N', then WORK is not referenced.
INFO (output) INTEGER
= 0: successful exit
< 0: if INFO = -i, the i-th argument had an illegal value
> 0: the algorithm has failed to find all the eigenvalues in
a total of 30*N iterations; if INFO = i, then i
elements of E have not converged to zero; on exit, D
and E contain the elements of a symmetric tridiagonal
matrix which is orthogonally similar to the original
matrix.
=====================================================================
Test the input parameters.
Parameter adjustments */
/* Table of constant values */
static doublereal c_b9 = 0.;
static doublereal c_b10 = 1.;
static integer c__0 = 0;
static integer c__1 = 1;
static integer c__2 = 2;
/* System generated locals */
integer z_dim1, z_offset, i__1, i__2;
doublereal d__1, d__2;
/* Builtin functions */
double sqrt(doublereal), d_sign(doublereal *, doublereal *);
/* Local variables */
static integer lend, jtot;
extern /* Subroutine */ int dlae2_(doublereal *, doublereal *, doublereal
*, doublereal *, doublereal *);
static doublereal b, c__, f, g;
static integer i__, j, k, l, m;
static doublereal p, r__, s;
extern logical lsame_(char *, char *);
extern /* Subroutine */ int dlasr_(char *, char *, char *, integer *,
integer *, doublereal *, doublereal *, doublereal *, integer *);
static doublereal anorm;
extern /* Subroutine */ int dswap_(integer *, doublereal *, integer *,
doublereal *, integer *);
static integer l1;
extern /* Subroutine */ int dlaev2_(doublereal *, doublereal *,
doublereal *, doublereal *, doublereal *, doublereal *,
doublereal *);
static integer lendm1, lendp1;
extern doublereal dlapy2_(doublereal *, doublereal *);
static integer ii;
extern doublereal dlamch_(char *);
static integer mm, iscale;
extern /* Subroutine */ int dlascl_(char *, integer *, integer *,
doublereal *, doublereal *, integer *, integer *, doublereal *,
integer *, integer *), dlaset_(char *, integer *, integer
*, doublereal *, doublereal *, doublereal *, integer *);
static doublereal safmin;
extern /* Subroutine */ int dlartg_(doublereal *, doublereal *,
doublereal *, doublereal *, doublereal *);
static doublereal safmax;
extern /* Subroutine */ int xerbla_(char *, integer *);
extern doublereal dlanst_(char *, integer *, doublereal *, doublereal *);
extern /* Subroutine */ int dlasrt_(char *, integer *, doublereal *,
integer *);
static integer lendsv;
static doublereal ssfmin;
static integer nmaxit, icompz;
static doublereal ssfmax;
static integer lm1, mm1, nm1;
static doublereal rt1, rt2, eps;
static integer lsv;
static doublereal tst, eps2;
#define z___ref(a_1,a_2) z__[(a_2)*z_dim1 + a_1]
--d__;
--e;
z_dim1 = *ldz;
z_offset = 1 + z_dim1 * 1;
z__ -= z_offset;
--work;
/* Function Body */
*info = 0;
if (lsame_(compz, "N")) {
icompz = 0;
} else if (lsame_(compz, "V")) {
icompz = 1;
} else if (lsame_(compz, "I")) {
icompz = 2;
} else {
icompz = -1;
}
if (icompz < 0) {
*info = -1;
} else if (*n < 0) {
*info = -2;
} else if ((*ldz < 1) || ((icompz > 0) && (*ldz < max(1,*n)))) {
*info = -6;
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("DSTEQR", &i__1);
return 0;
}
/* Quick return if possible */
if (*n == 0) {
return 0;
}
if (*n == 1) {
if (icompz == 2) {
z___ref(1, 1) = 1.;
}
return 0;
}
/* Determine the unit roundoff and over/underflow thresholds. */
eps = dlamch_("E");
/* Computing 2nd power */
d__1 = eps;
eps2 = d__1 * d__1;
safmin = dlamch_("S");
safmax = 1. / safmin;
ssfmax = sqrt(safmax) / 3.;
ssfmin = sqrt(safmin) / eps2;
/* Compute the eigenvalues and eigenvectors of the tridiagonal
matrix. */
if (icompz == 2) {
dlaset_("Full", n, n, &c_b9, &c_b10, &z__[z_offset], ldz);
}
nmaxit = *n * 30;
jtot = 0;
/* Determine where the matrix splits and choose QL or QR iteration
for each block, according to whether top or bottom diagonal
element is smaller. */
l1 = 1;
nm1 = *n - 1;
L10:
if (l1 > *n) {
goto L160;
}
if (l1 > 1) {
e[l1 - 1] = 0.;
}
if (l1 <= nm1) {
i__1 = nm1;
for (m = l1; m <= i__1; ++m) {
tst = (d__1 = e[m], abs(d__1));
if (tst == 0.) {
goto L30;
}
if (tst <= sqrt((d__1 = d__[m], abs(d__1))) * sqrt((d__2 = d__[m
+ 1], abs(d__2))) * eps) {
e[m] = 0.;
goto L30;
}
/* L20: */
}
}
m = *n;
L30:
l = l1;
lsv = l;
lend = m;
lendsv = lend;
l1 = m + 1;
if (lend == l) {
goto L10;
}
/* Scale submatrix in rows and columns L to LEND */
i__1 = lend - l + 1;
anorm = dlanst_("I", &i__1, &d__[l], &e[l]);
iscale = 0;
if (anorm == 0.) {
goto L10;
}
if (anorm > ssfmax) {
iscale = 1;
i__1 = lend - l + 1;
dlascl_("G", &c__0, &c__0, &anorm, &ssfmax, &i__1, &c__1, &d__[l], n,
info);
i__1 = lend - l;
dlascl_("G", &c__0, &c__0, &anorm, &ssfmax, &i__1, &c__1, &e[l], n,
info);
} else if (anorm < ssfmin) {
iscale = 2;
i__1 = lend - l + 1;
dlascl_("G", &c__0, &c__0, &anorm, &ssfmin, &i__1, &c__1, &d__[l], n,
info);
i__1 = lend - l;
dlascl_("G", &c__0, &c__0, &anorm, &ssfmin, &i__1, &c__1, &e[l], n,
info);
}
/* Choose between QL and QR iteration */
if ((d__1 = d__[lend], abs(d__1)) < (d__2 = d__[l], abs(d__2))) {
lend = lsv;
l = lendsv;
}
if (lend > l) {
/* QL Iteration
Look for small subdiagonal element. */
L40:
if (l != lend) {
lendm1 = lend - 1;
i__1 = lendm1;
for (m = l; m <= i__1; ++m) {
/* Computing 2nd power */
d__2 = (d__1 = e[m], abs(d__1));
tst = d__2 * d__2;
if (tst <= eps2 * (d__1 = d__[m], abs(d__1)) * (d__2 = d__[m
+ 1], abs(d__2)) + safmin) {
goto L60;
}
/* L50: */
}
}
m = lend;
L60:
if (m < lend) {
e[m] = 0.;
}
p = d__[l];
if (m == l) {
goto L80;
}
/* If remaining matrix is 2-by-2, use DLAE2 or SLAEV2
to compute its eigensystem. */
if (m == l + 1) {
if (icompz > 0) {
dlaev2_(&d__[l], &e[l], &d__[l + 1], &rt1, &rt2, &c__, &s);
work[l] = c__;
work[*n - 1 + l] = s;
dlasr_("R", "V", "B", n, &c__2, &work[l], &work[*n - 1 + l], &
z___ref(1, l), ldz);
} else {
dlae2_(&d__[l], &e[l], &d__[l + 1], &rt1, &rt2);
}
d__[l] = rt1;
d__[l + 1] = rt2;
e[l] = 0.;
l += 2;
if (l <= lend) {
goto L40;
}
goto L140;
}
if (jtot == nmaxit) {
goto L140;
}
++jtot;
/* Form shift. */
g = (d__[l + 1] - p) / (e[l] * 2.);
r__ = dlapy2_(&g, &c_b10);
g = d__[m] - p + e[l] / (g + d_sign(&r__, &g));
s = 1.;
c__ = 1.;
p = 0.;
/* Inner loop */
mm1 = m - 1;
i__1 = l;
for (i__ = mm1; i__ >= i__1; --i__) {
f = s * e[i__];
b = c__ * e[i__];
dlartg_(&g, &f, &c__, &s, &r__);
if (i__ != m - 1) {
e[i__ + 1] = r__;
}
g = d__[i__ + 1] - p;
r__ = (d__[i__] - g) * s + c__ * 2. * b;
p = s * r__;
d__[i__ + 1] = g + p;
g = c__ * r__ - b;
/* If eigenvectors are desired, then save rotations. */
if (icompz > 0) {
work[i__] = c__;
work[*n - 1 + i__] = -s;
}
/* L70: */
}
/* If eigenvectors are desired, then apply saved rotations. */
if (icompz > 0) {
mm = m - l + 1;
dlasr_("R", "V", "B", n, &mm, &work[l], &work[*n - 1 + l], &
z___ref(1, l), ldz);
}
d__[l] -= p;
e[l] = g;
goto L40;
/* Eigenvalue found. */
L80:
d__[l] = p;
++l;
if (l <= lend) {
goto L40;
}
goto L140;
} else {
/* QR Iteration
Look for small superdiagonal element. */
L90:
if (l != lend) {
lendp1 = lend + 1;
i__1 = lendp1;
for (m = l; m >= i__1; --m) {
/* Computing 2nd power */
d__2 = (d__1 = e[m - 1], abs(d__1));
tst = d__2 * d__2;
if (tst <= eps2 * (d__1 = d__[m], abs(d__1)) * (d__2 = d__[m
- 1], abs(d__2)) + safmin) {
goto L110;
}
/* L100: */
}
}
m = lend;
L110:
if (m > lend) {
e[m - 1] = 0.;
}
p = d__[l];
if (m == l) {
goto L130;
}
/* If remaining matrix is 2-by-2, use DLAE2 or SLAEV2
to compute its eigensystem. */
if (m == l - 1) {
if (icompz > 0) {
dlaev2_(&d__[l - 1], &e[l - 1], &d__[l], &rt1, &rt2, &c__, &s)
;
work[m] = c__;
work[*n - 1 + m] = s;
dlasr_("R", "V", "F", n, &c__2, &work[m], &work[*n - 1 + m], &
z___ref(1, l - 1), ldz);
} else {
dlae2_(&d__[l - 1], &e[l - 1], &d__[l], &rt1, &rt2);
}
d__[l - 1] = rt1;
d__[l] = rt2;
e[l - 1] = 0.;
l += -2;
if (l >= lend) {
goto L90;
}
goto L140;
}
if (jtot == nmaxit) {
goto L140;
}
++jtot;
/* Form shift. */
g = (d__[l - 1] - p) / (e[l - 1] * 2.);
r__ = dlapy2_(&g, &c_b10);
g = d__[m] - p + e[l - 1] / (g + d_sign(&r__, &g));
s = 1.;
c__ = 1.;
p = 0.;
/* Inner loop */
lm1 = l - 1;
i__1 = lm1;
for (i__ = m; i__ <= i__1; ++i__) {
f = s * e[i__];
b = c__ * e[i__];
dlartg_(&g, &f, &c__, &s, &r__);
if (i__ != m) {
e[i__ - 1] = r__;
}
g = d__[i__] - p;
r__ = (d__[i__ + 1] - g) * s + c__ * 2. * b;
p = s * r__;
d__[i__] = g + p;
g = c__ * r__ - b;
/* If eigenvectors are desired, then save rotations. */
if (icompz > 0) {
work[i__] = c__;
work[*n - 1 + i__] = s;
}
/* L120: */
}
/* If eigenvectors are desired, then apply saved rotations. */
if (icompz > 0) {
mm = l - m + 1;
dlasr_("R", "V", "F", n, &mm, &work[m], &work[*n - 1 + m], &
z___ref(1, m), ldz);
}
d__[l] -= p;
e[lm1] = g;
goto L90;
/* Eigenvalue found. */
L130:
d__[l] = p;
--l;
if (l >= lend) {
goto L90;
}
goto L140;
}
/* Undo scaling if necessary */
L140:
if (iscale == 1) {
i__1 = lendsv - lsv + 1;
dlascl_("G", &c__0, &c__0, &ssfmax, &anorm, &i__1, &c__1, &d__[lsv],
n, info);
i__1 = lendsv - lsv;
dlascl_("G", &c__0, &c__0, &ssfmax, &anorm, &i__1, &c__1, &e[lsv], n,
info);
} else if (iscale == 2) {
i__1 = lendsv - lsv + 1;
dlascl_("G", &c__0, &c__0, &ssfmin, &anorm, &i__1, &c__1, &d__[lsv],
n, info);
i__1 = lendsv - lsv;
dlascl_("G", &c__0, &c__0, &ssfmin, &anorm, &i__1, &c__1, &e[lsv], n,
info);
}
/* Check for no convergence to an eigenvalue after a total
of N*MAXIT iterations. */
if (jtot < nmaxit) {
goto L10;
}
i__1 = *n - 1;
for (i__ = 1; i__ <= i__1; ++i__) {
if (e[i__] != 0.) {
++(*info);
}
/* L150: */
}
goto L190;
/* Order eigenvalues and eigenvectors. */
L160:
if (icompz == 0) {
/* Use Quick Sort */
dlasrt_("I", n, &d__[1], info);
} else {
/* Use Selection Sort to minimize swaps of eigenvectors */
i__1 = *n;
for (ii = 2; ii <= i__1; ++ii) {
i__ = ii - 1;
k = i__;
p = d__[i__];
i__2 = *n;
for (j = ii; j <= i__2; ++j) {
if (d__[j] < p) {
k = j;
p = d__[j];
}
/* L170: */
}
if (k != i__) {
d__[k] = d__[i__];
d__[i__] = p;
dswap_(n, &z___ref(1, i__), &c__1, &z___ref(1, k), &c__1);
}
/* L180: */
}
}
L190:
return 0;
/* End of DSTEQR */
} /* dsteqr_ */
#undef z___ref
Reference in New Issue View Git Blame Copy Permalink