1 *> \brief \b SLAGTF computes an LU factorization of a matrix T-λI, where T is a general tridiagonal matrix, and λ a scalar, using partial pivoting with row interchanges.
3 * =========== DOCUMENTATION ===========
5 * Online html documentation available at
6 * http://www.netlib.org/lapack/explore-html/
9 *> Download SLAGTF + dependencies
10 *> <a href="http://www.netlib.org/cgi-bin/netlibfiles.tgz?format=tgz&filename=/lapack/lapack_routine/slagtf.f">
12 *> <a href="http://www.netlib.org/cgi-bin/netlibfiles.zip?format=zip&filename=/lapack/lapack_routine/slagtf.f">
14 *> <a href="http://www.netlib.org/cgi-bin/netlibfiles.txt?format=txt&filename=/lapack/lapack_routine/slagtf.f">
21 * SUBROUTINE SLAGTF( N, A, LAMBDA, B, C, TOL, D, IN, INFO )
23 * .. Scalar Arguments ..
27 * .. Array Arguments ..
29 * REAL A( * ), B( * ), C( * ), D( * )
38 *> SLAGTF factorizes the matrix (T - lambda*I), where T is an n by n
39 *> tridiagonal matrix and lambda is a scalar, as
41 *> T - lambda*I = PLU,
43 *> where P is a permutation matrix, L is a unit lower tridiagonal matrix
44 *> with at most one non-zero sub-diagonal elements per column and U is
45 *> an upper triangular matrix with at most two non-zero super-diagonal
46 *> elements per column.
48 *> The factorization is obtained by Gaussian elimination with partial
49 *> pivoting and implicit row scaling.
51 *> The parameter LAMBDA is included in the routine so that SLAGTF may
52 *> be used, in conjunction with SLAGTS, to obtain eigenvectors of T by
62 *> The order of the matrix T.
67 *> A is REAL array, dimension (N)
68 *> On entry, A must contain the diagonal elements of T.
70 *> On exit, A is overwritten by the n diagonal elements of the
71 *> upper triangular matrix U of the factorization of T.
77 *> On entry, the scalar lambda.
82 *> B is REAL array, dimension (N-1)
83 *> On entry, B must contain the (n-1) super-diagonal elements of
86 *> On exit, B is overwritten by the (n-1) super-diagonal
87 *> elements of the matrix U of the factorization of T.
92 *> C is REAL array, dimension (N-1)
93 *> On entry, C must contain the (n-1) sub-diagonal elements of
96 *> On exit, C is overwritten by the (n-1) sub-diagonal elements
97 *> of the matrix L of the factorization of T.
103 *> On entry, a relative tolerance used to indicate whether or
104 *> not the matrix (T - lambda*I) is nearly singular. TOL should
105 *> normally be chose as approximately the largest relative error
106 *> in the elements of T. For example, if the elements of T are
107 *> correct to about 4 significant figures, then TOL should be
108 *> set to about 5*10**(-4). If TOL is supplied as less than eps,
109 *> where eps is the relative machine precision, then the value
110 *> eps is used in place of TOL.
115 *> D is REAL array, dimension (N-2)
116 *> On exit, D is overwritten by the (n-2) second super-diagonal
117 *> elements of the matrix U of the factorization of T.
122 *> IN is INTEGER array, dimension (N)
123 *> On exit, IN contains details of the permutation matrix P. If
124 *> an interchange occurred at the kth step of the elimination,
125 *> then IN(k) = 1, otherwise IN(k) = 0. The element IN(n)
126 *> returns the smallest positive integer j such that
128 *> abs( u(j,j) ).le. norm( (T - lambda*I)(j) )*TOL,
130 *> where norm( A(j) ) denotes the sum of the absolute values of
131 *> the jth row of the matrix A. If no such j exists then IN(n)
132 *> is returned as zero. If IN(n) is returned as positive, then a
133 *> diagonal element of U is small, indicating that
134 *> (T - lambda*I) is singular or nearly singular,
140 *> = 0 : successful exit
141 *> .lt. 0: if INFO = -k, the kth argument had an illegal value
147 *> \author Univ. of Tennessee
148 *> \author Univ. of California Berkeley
149 *> \author Univ. of Colorado Denver
152 *> \date September 2012
154 *> \ingroup auxOTHERcomputational
156 * =====================================================================
157 SUBROUTINE SLAGTF( N, A, LAMBDA, B, C, TOL, D, IN, INFO )
159 * -- LAPACK computational routine (version 3.4.2) --
160 * -- LAPACK is a software package provided by Univ. of Tennessee, --
161 * -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..--
164 * .. Scalar Arguments ..
168 * .. Array Arguments ..
170 REAL A( * ), B( * ), C( * ), D( * )
173 * =====================================================================
177 PARAMETER ( ZERO = 0.0E+0 )
179 * .. Local Scalars ..
181 REAL EPS, MULT, PIV1, PIV2, SCALE1, SCALE2, TEMP, TL
183 * .. Intrinsic Functions ..
186 * .. External Functions ..
190 * .. External Subroutines ..
193 * .. Executable Statements ..
198 CALL XERBLA( 'SLAGTF', -INFO )
205 A( 1 ) = A( 1 ) - LAMBDA
213 EPS = SLAMCH( 'Epsilon' )
216 SCALE1 = ABS( A( 1 ) ) + ABS( B( 1 ) )
218 A( K+1 ) = A( K+1 ) - LAMBDA
219 SCALE2 = ABS( C( K ) ) + ABS( A( K+1 ) )
221 $ SCALE2 = SCALE2 + ABS( B( K+1 ) )
222 IF( A( K ).EQ.ZERO ) THEN
225 PIV1 = ABS( A( K ) ) / SCALE1
227 IF( C( K ).EQ.ZERO ) THEN
234 PIV2 = ABS( C( K ) ) / SCALE2
235 IF( PIV2.LE.PIV1 ) THEN
238 C( K ) = C( K ) / A( K )
239 A( K+1 ) = A( K+1 ) - C( K )*B( K )
244 MULT = A( K ) / C( K )
247 A( K+1 ) = B( K ) - MULT*TEMP
248 IF( K.LT.( N-1 ) ) THEN
250 B( K+1 ) = -MULT*D( K )
256 IF( ( MAX( PIV1, PIV2 ).LE.TL ) .AND. ( IN( N ).EQ.0 ) )
259 IF( ( ABS( A( N ) ).LE.SCALE1*TL ) .AND. ( IN( N ).EQ.0 ) )