root / src / lapack / double / dgebrd.f @ 2
Historique | Voir | Annoter | Télécharger (9,05 ko)
1 |
SUBROUTINE DGEBRD( M, N, A, LDA, D, E, TAUQ, TAUP, WORK, LWORK, |
---|---|
2 |
$ INFO ) |
3 |
* |
4 |
* -- LAPACK routine (version 3.2) -- |
5 |
* -- LAPACK is a software package provided by Univ. of Tennessee, -- |
6 |
* -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..-- |
7 |
* November 2006 |
8 |
* |
9 |
* .. Scalar Arguments .. |
10 |
INTEGER INFO, LDA, LWORK, M, N |
11 |
* .. |
12 |
* .. Array Arguments .. |
13 |
DOUBLE PRECISION A( LDA, * ), D( * ), E( * ), TAUP( * ), |
14 |
$ TAUQ( * ), WORK( * ) |
15 |
* .. |
16 |
* |
17 |
* Purpose |
18 |
* ======= |
19 |
* |
20 |
* DGEBRD reduces a general real M-by-N matrix A to upper or lower |
21 |
* bidiagonal form B by an orthogonal transformation: Q**T * A * P = B. |
22 |
* |
23 |
* If m >= n, B is upper bidiagonal; if m < n, B is lower bidiagonal. |
24 |
* |
25 |
* Arguments |
26 |
* ========= |
27 |
* |
28 |
* M (input) INTEGER |
29 |
* The number of rows in the matrix A. M >= 0. |
30 |
* |
31 |
* N (input) INTEGER |
32 |
* The number of columns in the matrix A. N >= 0. |
33 |
* |
34 |
* A (input/output) DOUBLE PRECISION array, dimension (LDA,N) |
35 |
* On entry, the M-by-N general matrix to be reduced. |
36 |
* On exit, |
37 |
* if m >= n, the diagonal and the first superdiagonal are |
38 |
* overwritten with the upper bidiagonal matrix B; the |
39 |
* elements below the diagonal, with the array TAUQ, represent |
40 |
* the orthogonal matrix Q as a product of elementary |
41 |
* reflectors, and the elements above the first superdiagonal, |
42 |
* with the array TAUP, represent the orthogonal matrix P as |
43 |
* a product of elementary reflectors; |
44 |
* if m < n, the diagonal and the first subdiagonal are |
45 |
* overwritten with the lower bidiagonal matrix B; the |
46 |
* elements below the first subdiagonal, with the array TAUQ, |
47 |
* represent the orthogonal matrix Q as a product of |
48 |
* elementary reflectors, and the elements above the diagonal, |
49 |
* with the array TAUP, represent the orthogonal matrix P as |
50 |
* a product of elementary reflectors. |
51 |
* See Further Details. |
52 |
* |
53 |
* LDA (input) INTEGER |
54 |
* The leading dimension of the array A. LDA >= max(1,M). |
55 |
* |
56 |
* D (output) DOUBLE PRECISION array, dimension (min(M,N)) |
57 |
* The diagonal elements of the bidiagonal matrix B: |
58 |
* D(i) = A(i,i). |
59 |
* |
60 |
* E (output) DOUBLE PRECISION array, dimension (min(M,N)-1) |
61 |
* The off-diagonal elements of the bidiagonal matrix B: |
62 |
* if m >= n, E(i) = A(i,i+1) for i = 1,2,...,n-1; |
63 |
* if m < n, E(i) = A(i+1,i) for i = 1,2,...,m-1. |
64 |
* |
65 |
* TAUQ (output) DOUBLE PRECISION array dimension (min(M,N)) |
66 |
* The scalar factors of the elementary reflectors which |
67 |
* represent the orthogonal matrix Q. See Further Details. |
68 |
* |
69 |
* TAUP (output) DOUBLE PRECISION array, dimension (min(M,N)) |
70 |
* The scalar factors of the elementary reflectors which |
71 |
* represent the orthogonal matrix P. See Further Details. |
72 |
* |
73 |
* WORK (workspace/output) DOUBLE PRECISION array, dimension (MAX(1,LWORK)) |
74 |
* On exit, if INFO = 0, WORK(1) returns the optimal LWORK. |
75 |
* |
76 |
* LWORK (input) INTEGER |
77 |
* The length of the array WORK. LWORK >= max(1,M,N). |
78 |
* For optimum performance LWORK >= (M+N)*NB, where NB |
79 |
* is the optimal blocksize. |
80 |
* |
81 |
* If LWORK = -1, then a workspace query is assumed; the routine |
82 |
* only calculates the optimal size of the WORK array, returns |
83 |
* this value as the first entry of the WORK array, and no error |
84 |
* message related to LWORK is issued by XERBLA. |
85 |
* |
86 |
* INFO (output) INTEGER |
87 |
* = 0: successful exit |
88 |
* < 0: if INFO = -i, the i-th argument had an illegal value. |
89 |
* |
90 |
* Further Details |
91 |
* =============== |
92 |
* |
93 |
* The matrices Q and P are represented as products of elementary |
94 |
* reflectors: |
95 |
* |
96 |
* If m >= n, |
97 |
* |
98 |
* Q = H(1) H(2) . . . H(n) and P = G(1) G(2) . . . G(n-1) |
99 |
* |
100 |
* Each H(i) and G(i) has the form: |
101 |
* |
102 |
* H(i) = I - tauq * v * v' and G(i) = I - taup * u * u' |
103 |
* |
104 |
* where tauq and taup are real scalars, and v and u are real vectors; |
105 |
* v(1:i-1) = 0, v(i) = 1, and v(i+1:m) is stored on exit in A(i+1:m,i); |
106 |
* u(1:i) = 0, u(i+1) = 1, and u(i+2:n) is stored on exit in A(i,i+2:n); |
107 |
* tauq is stored in TAUQ(i) and taup in TAUP(i). |
108 |
* |
109 |
* If m < n, |
110 |
* |
111 |
* Q = H(1) H(2) . . . H(m-1) and P = G(1) G(2) . . . G(m) |
112 |
* |
113 |
* Each H(i) and G(i) has the form: |
114 |
* |
115 |
* H(i) = I - tauq * v * v' and G(i) = I - taup * u * u' |
116 |
* |
117 |
* where tauq and taup are real scalars, and v and u are real vectors; |
118 |
* v(1:i) = 0, v(i+1) = 1, and v(i+2:m) is stored on exit in A(i+2:m,i); |
119 |
* u(1:i-1) = 0, u(i) = 1, and u(i+1:n) is stored on exit in A(i,i+1:n); |
120 |
* tauq is stored in TAUQ(i) and taup in TAUP(i). |
121 |
* |
122 |
* The contents of A on exit are illustrated by the following examples: |
123 |
* |
124 |
* m = 6 and n = 5 (m > n): m = 5 and n = 6 (m < n): |
125 |
* |
126 |
* ( d e u1 u1 u1 ) ( d u1 u1 u1 u1 u1 ) |
127 |
* ( v1 d e u2 u2 ) ( e d u2 u2 u2 u2 ) |
128 |
* ( v1 v2 d e u3 ) ( v1 e d u3 u3 u3 ) |
129 |
* ( v1 v2 v3 d e ) ( v1 v2 e d u4 u4 ) |
130 |
* ( v1 v2 v3 v4 d ) ( v1 v2 v3 e d u5 ) |
131 |
* ( v1 v2 v3 v4 v5 ) |
132 |
* |
133 |
* where d and e denote diagonal and off-diagonal elements of B, vi |
134 |
* denotes an element of the vector defining H(i), and ui an element of |
135 |
* the vector defining G(i). |
136 |
* |
137 |
* ===================================================================== |
138 |
* |
139 |
* .. Parameters .. |
140 |
DOUBLE PRECISION ONE |
141 |
PARAMETER ( ONE = 1.0D+0 ) |
142 |
* .. |
143 |
* .. Local Scalars .. |
144 |
LOGICAL LQUERY |
145 |
INTEGER I, IINFO, J, LDWRKX, LDWRKY, LWKOPT, MINMN, NB, |
146 |
$ NBMIN, NX |
147 |
DOUBLE PRECISION WS |
148 |
* .. |
149 |
* .. External Subroutines .. |
150 |
EXTERNAL DGEBD2, DGEMM, DLABRD, XERBLA |
151 |
* .. |
152 |
* .. Intrinsic Functions .. |
153 |
INTRINSIC DBLE, MAX, MIN |
154 |
* .. |
155 |
* .. External Functions .. |
156 |
INTEGER ILAENV |
157 |
EXTERNAL ILAENV |
158 |
* .. |
159 |
* .. Executable Statements .. |
160 |
* |
161 |
* Test the input parameters |
162 |
* |
163 |
INFO = 0 |
164 |
NB = MAX( 1, ILAENV( 1, 'DGEBRD', ' ', M, N, -1, -1 ) ) |
165 |
LWKOPT = ( M+N )*NB |
166 |
WORK( 1 ) = DBLE( LWKOPT ) |
167 |
LQUERY = ( LWORK.EQ.-1 ) |
168 |
IF( M.LT.0 ) THEN |
169 |
INFO = -1 |
170 |
ELSE IF( N.LT.0 ) THEN |
171 |
INFO = -2 |
172 |
ELSE IF( LDA.LT.MAX( 1, M ) ) THEN |
173 |
INFO = -4 |
174 |
ELSE IF( LWORK.LT.MAX( 1, M, N ) .AND. .NOT.LQUERY ) THEN |
175 |
INFO = -10 |
176 |
END IF |
177 |
IF( INFO.LT.0 ) THEN |
178 |
CALL XERBLA( 'DGEBRD', -INFO ) |
179 |
RETURN |
180 |
ELSE IF( LQUERY ) THEN |
181 |
RETURN |
182 |
END IF |
183 |
* |
184 |
* Quick return if possible |
185 |
* |
186 |
MINMN = MIN( M, N ) |
187 |
IF( MINMN.EQ.0 ) THEN |
188 |
WORK( 1 ) = 1 |
189 |
RETURN |
190 |
END IF |
191 |
* |
192 |
WS = MAX( M, N ) |
193 |
LDWRKX = M |
194 |
LDWRKY = N |
195 |
* |
196 |
IF( NB.GT.1 .AND. NB.LT.MINMN ) THEN |
197 |
* |
198 |
* Set the crossover point NX. |
199 |
* |
200 |
NX = MAX( NB, ILAENV( 3, 'DGEBRD', ' ', M, N, -1, -1 ) ) |
201 |
* |
202 |
* Determine when to switch from blocked to unblocked code. |
203 |
* |
204 |
IF( NX.LT.MINMN ) THEN |
205 |
WS = ( M+N )*NB |
206 |
IF( LWORK.LT.WS ) THEN |
207 |
* |
208 |
* Not enough work space for the optimal NB, consider using |
209 |
* a smaller block size. |
210 |
* |
211 |
NBMIN = ILAENV( 2, 'DGEBRD', ' ', M, N, -1, -1 ) |
212 |
IF( LWORK.GE.( M+N )*NBMIN ) THEN |
213 |
NB = LWORK / ( M+N ) |
214 |
ELSE |
215 |
NB = 1 |
216 |
NX = MINMN |
217 |
END IF |
218 |
END IF |
219 |
END IF |
220 |
ELSE |
221 |
NX = MINMN |
222 |
END IF |
223 |
* |
224 |
DO 30 I = 1, MINMN - NX, NB |
225 |
* |
226 |
* Reduce rows and columns i:i+nb-1 to bidiagonal form and return |
227 |
* the matrices X and Y which are needed to update the unreduced |
228 |
* part of the matrix |
229 |
* |
230 |
CALL DLABRD( M-I+1, N-I+1, NB, A( I, I ), LDA, D( I ), E( I ), |
231 |
$ TAUQ( I ), TAUP( I ), WORK, LDWRKX, |
232 |
$ WORK( LDWRKX*NB+1 ), LDWRKY ) |
233 |
* |
234 |
* Update the trailing submatrix A(i+nb:m,i+nb:n), using an update |
235 |
* of the form A := A - V*Y' - X*U' |
236 |
* |
237 |
CALL DGEMM( 'No transpose', 'Transpose', M-I-NB+1, N-I-NB+1, |
238 |
$ NB, -ONE, A( I+NB, I ), LDA, |
239 |
$ WORK( LDWRKX*NB+NB+1 ), LDWRKY, ONE, |
240 |
$ A( I+NB, I+NB ), LDA ) |
241 |
CALL DGEMM( 'No transpose', 'No transpose', M-I-NB+1, N-I-NB+1, |
242 |
$ NB, -ONE, WORK( NB+1 ), LDWRKX, A( I, I+NB ), LDA, |
243 |
$ ONE, A( I+NB, I+NB ), LDA ) |
244 |
* |
245 |
* Copy diagonal and off-diagonal elements of B back into A |
246 |
* |
247 |
IF( M.GE.N ) THEN |
248 |
DO 10 J = I, I + NB - 1 |
249 |
A( J, J ) = D( J ) |
250 |
A( J, J+1 ) = E( J ) |
251 |
10 CONTINUE |
252 |
ELSE |
253 |
DO 20 J = I, I + NB - 1 |
254 |
A( J, J ) = D( J ) |
255 |
A( J+1, J ) = E( J ) |
256 |
20 CONTINUE |
257 |
END IF |
258 |
30 CONTINUE |
259 |
* |
260 |
* Use unblocked code to reduce the remainder of the matrix |
261 |
* |
262 |
CALL DGEBD2( M-I+1, N-I+1, A( I, I ), LDA, D( I ), E( I ), |
263 |
$ TAUQ( I ), TAUP( I ), WORK, IINFO ) |
264 |
WORK( 1 ) = WS |
265 |
RETURN |
266 |
* |
267 |
* End of DGEBRD |
268 |
* |
269 |
END |