1 *> \brief \b DLATRD reduces the first nb rows and columns of a symmetric/Hermitian matrix A to real tridiagonal form by an orthogonal similarity transformation.
3 * =========== DOCUMENTATION ===========
5 * Online html documentation available at
6 * http://www.netlib.org/lapack/explore-html/
9 *> Download DLATRD + dependencies
10 *> <a href="http://www.netlib.org/cgi-bin/netlibfiles.tgz?format=tgz&filename=/lapack/lapack_routine/dlatrd.f">
12 *> <a href="http://www.netlib.org/cgi-bin/netlibfiles.zip?format=zip&filename=/lapack/lapack_routine/dlatrd.f">
14 *> <a href="http://www.netlib.org/cgi-bin/netlibfiles.txt?format=txt&filename=/lapack/lapack_routine/dlatrd.f">
21 * SUBROUTINE DLATRD( UPLO, N, NB, A, LDA, E, TAU, W, LDW )
23 * .. Scalar Arguments ..
25 * INTEGER LDA, LDW, N, NB
27 * .. Array Arguments ..
28 * DOUBLE PRECISION A( LDA, * ), E( * ), TAU( * ), W( LDW, * )
37 *> DLATRD reduces NB rows and columns of a real symmetric matrix A to
38 *> symmetric tridiagonal form by an orthogonal similarity
39 *> transformation Q**T * A * Q, and returns the matrices V and W which are
40 *> needed to apply the transformation to the unreduced part of A.
42 *> If UPLO = 'U', DLATRD reduces the last NB rows and columns of a
43 *> matrix, of which the upper triangle is supplied;
44 *> if UPLO = 'L', DLATRD reduces the first NB rows and columns of a
45 *> matrix, of which the lower triangle is supplied.
47 *> This is an auxiliary routine called by DSYTRD.
55 *> UPLO is CHARACTER*1
56 *> Specifies whether the upper or lower triangular part of the
57 *> symmetric matrix A is stored:
58 *> = 'U': Upper triangular
59 *> = 'L': Lower triangular
65 *> The order of the matrix A.
71 *> The number of rows and columns to be reduced.
76 *> A is DOUBLE PRECISION array, dimension (LDA,N)
77 *> On entry, the symmetric matrix A. If UPLO = 'U', the leading
78 *> n-by-n upper triangular part of A contains the upper
79 *> triangular part of the matrix A, and the strictly lower
80 *> triangular part of A is not referenced. If UPLO = 'L', the
81 *> leading n-by-n lower triangular part of A contains the lower
82 *> triangular part of the matrix A, and the strictly upper
83 *> triangular part of A is not referenced.
85 *> if UPLO = 'U', the last NB columns have been reduced to
86 *> tridiagonal form, with the diagonal elements overwriting
87 *> the diagonal elements of A; the elements above the diagonal
88 *> with the array TAU, represent the orthogonal matrix Q as a
89 *> product of elementary reflectors;
90 *> if UPLO = 'L', the first NB columns have been reduced to
91 *> tridiagonal form, with the diagonal elements overwriting
92 *> the diagonal elements of A; the elements below the diagonal
93 *> with the array TAU, represent the orthogonal matrix Q as a
94 *> product of elementary reflectors.
95 *> See Further Details.
101 *> The leading dimension of the array A. LDA >= (1,N).
106 *> E is DOUBLE PRECISION array, dimension (N-1)
107 *> If UPLO = 'U', E(n-nb:n-1) contains the superdiagonal
108 *> elements of the last NB columns of the reduced matrix;
109 *> if UPLO = 'L', E(1:nb) contains the subdiagonal elements of
110 *> the first NB columns of the reduced matrix.
115 *> TAU is DOUBLE PRECISION array, dimension (N-1)
116 *> The scalar factors of the elementary reflectors, stored in
117 *> TAU(n-nb:n-1) if UPLO = 'U', and in TAU(1:nb) if UPLO = 'L'.
118 *> See Further Details.
123 *> W is DOUBLE PRECISION array, dimension (LDW,NB)
124 *> The n-by-nb matrix W required to update the unreduced part
131 *> The leading dimension of the array W. LDW >= max(1,N).
137 *> \author Univ. of Tennessee
138 *> \author Univ. of California Berkeley
139 *> \author Univ. of Colorado Denver
142 *> \date September 2012
144 *> \ingroup doubleOTHERauxiliary
146 *> \par Further Details:
147 * =====================
151 *> If UPLO = 'U', the matrix Q is represented as a product of elementary
154 *> Q = H(n) H(n-1) . . . H(n-nb+1).
156 *> Each H(i) has the form
158 *> H(i) = I - tau * v * v**T
160 *> where tau is a real scalar, and v is a real vector with
161 *> v(i:n) = 0 and v(i-1) = 1; v(1:i-1) is stored on exit in A(1:i-1,i),
162 *> and tau in TAU(i-1).
164 *> If UPLO = 'L', the matrix Q is represented as a product of elementary
167 *> Q = H(1) H(2) . . . H(nb).
169 *> Each H(i) has the form
171 *> H(i) = I - tau * v * v**T
173 *> where tau is a real scalar, and v is a real vector with
174 *> v(1:i) = 0 and v(i+1) = 1; v(i+1:n) is stored on exit in A(i+1:n,i),
175 *> and tau in TAU(i).
177 *> The elements of the vectors v together form the n-by-nb matrix V
178 *> which is needed, with W, to apply the transformation to the unreduced
179 *> part of the matrix, using a symmetric rank-2k update of the form:
180 *> A := A - V*W**T - W*V**T.
182 *> The contents of A on exit are illustrated by the following examples
183 *> with n = 5 and nb = 2:
185 *> if UPLO = 'U': if UPLO = 'L':
187 *> ( a a a v4 v5 ) ( d )
188 *> ( a a v4 v5 ) ( 1 d )
189 *> ( a 1 v5 ) ( v1 1 a )
190 *> ( d 1 ) ( v1 v2 a a )
191 *> ( d ) ( v1 v2 a a a )
193 *> where d denotes a diagonal element of the reduced matrix, a denotes
194 *> an element of the original matrix that is unchanged, and vi denotes
195 *> an element of the vector defining H(i).
198 * =====================================================================
199 SUBROUTINE DLATRD( UPLO, N, NB, A, LDA, E, TAU, W, LDW )
201 * -- LAPACK auxiliary routine (version 3.4.2) --
202 * -- LAPACK is a software package provided by Univ. of Tennessee, --
203 * -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..--
206 * .. Scalar Arguments ..
208 INTEGER LDA, LDW, N, NB
210 * .. Array Arguments ..
211 DOUBLE PRECISION A( LDA, * ), E( * ), TAU( * ), W( LDW, * )
214 * =====================================================================
217 DOUBLE PRECISION ZERO, ONE, HALF
218 PARAMETER ( ZERO = 0.0D+0, ONE = 1.0D+0, HALF = 0.5D+0 )
220 * .. Local Scalars ..
222 DOUBLE PRECISION ALPHA
224 * .. External Subroutines ..
225 EXTERNAL DAXPY, DGEMV, DLARFG, DSCAL, DSYMV
227 * .. External Functions ..
229 DOUBLE PRECISION DDOT
232 * .. Intrinsic Functions ..
235 * .. Executable Statements ..
237 * Quick return if possible
242 IF( LSAME( UPLO, 'U' ) ) THEN
244 * Reduce last NB columns of upper triangle
246 DO 10 I = N, N - NB + 1, -1
252 CALL DGEMV( 'No transpose', I, N-I, -ONE, A( 1, I+1 ),
253 $ LDA, W( I, IW+1 ), LDW, ONE, A( 1, I ), 1 )
254 CALL DGEMV( 'No transpose', I, N-I, -ONE, W( 1, IW+1 ),
255 $ LDW, A( I, I+1 ), LDA, ONE, A( 1, I ), 1 )
259 * Generate elementary reflector H(i) to annihilate
262 CALL DLARFG( I-1, A( I-1, I ), A( 1, I ), 1, TAU( I-1 ) )
263 E( I-1 ) = A( I-1, I )
268 CALL DSYMV( 'Upper', I-1, ONE, A, LDA, A( 1, I ), 1,
269 $ ZERO, W( 1, IW ), 1 )
271 CALL DGEMV( 'Transpose', I-1, N-I, ONE, W( 1, IW+1 ),
272 $ LDW, A( 1, I ), 1, ZERO, W( I+1, IW ), 1 )
273 CALL DGEMV( 'No transpose', I-1, N-I, -ONE,
274 $ A( 1, I+1 ), LDA, W( I+1, IW ), 1, ONE,
276 CALL DGEMV( 'Transpose', I-1, N-I, ONE, A( 1, I+1 ),
277 $ LDA, A( 1, I ), 1, ZERO, W( I+1, IW ), 1 )
278 CALL DGEMV( 'No transpose', I-1, N-I, -ONE,
279 $ W( 1, IW+1 ), LDW, W( I+1, IW ), 1, ONE,
282 CALL DSCAL( I-1, TAU( I-1 ), W( 1, IW ), 1 )
283 ALPHA = -HALF*TAU( I-1 )*DDOT( I-1, W( 1, IW ), 1,
285 CALL DAXPY( I-1, ALPHA, A( 1, I ), 1, W( 1, IW ), 1 )
291 * Reduce first NB columns of lower triangle
297 CALL DGEMV( 'No transpose', N-I+1, I-1, -ONE, A( I, 1 ),
298 $ LDA, W( I, 1 ), LDW, ONE, A( I, I ), 1 )
299 CALL DGEMV( 'No transpose', N-I+1, I-1, -ONE, W( I, 1 ),
300 $ LDW, A( I, 1 ), LDA, ONE, A( I, I ), 1 )
303 * Generate elementary reflector H(i) to annihilate
306 CALL DLARFG( N-I, A( I+1, I ), A( MIN( I+2, N ), I ), 1,
313 CALL DSYMV( 'Lower', N-I, ONE, A( I+1, I+1 ), LDA,
314 $ A( I+1, I ), 1, ZERO, W( I+1, I ), 1 )
315 CALL DGEMV( 'Transpose', N-I, I-1, ONE, W( I+1, 1 ), LDW,
316 $ A( I+1, I ), 1, ZERO, W( 1, I ), 1 )
317 CALL DGEMV( 'No transpose', N-I, I-1, -ONE, A( I+1, 1 ),
318 $ LDA, W( 1, I ), 1, ONE, W( I+1, I ), 1 )
319 CALL DGEMV( 'Transpose', N-I, I-1, ONE, A( I+1, 1 ), LDA,
320 $ A( I+1, I ), 1, ZERO, W( 1, I ), 1 )
321 CALL DGEMV( 'No transpose', N-I, I-1, -ONE, W( I+1, 1 ),
322 $ LDW, W( 1, I ), 1, ONE, W( I+1, I ), 1 )
323 CALL DSCAL( N-I, TAU( I ), W( I+1, I ), 1 )
324 ALPHA = -HALF*TAU( I )*DDOT( N-I, W( I+1, I ), 1,
326 CALL DAXPY( N-I, ALPHA, A( I+1, I ), 1, W( I+1, I ), 1 )