summaryrefslogtreecommitdiff
path: root/SRC/zlaqps.f
blob: c142e8c69d88d4f1dff579bd74fa59efc1caddd8 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
*> \brief \b ZLAQPS computes a step of QR factorization with column pivoting of a real m-by-n matrix A by using BLAS level 3.
*
*  =========== DOCUMENTATION ===========
*
* Online html documentation available at
*            http://www.netlib.org/lapack/explore-html/
*
*> \htmlonly
*> Download ZLAQPS + dependencies
*> <a href="http://www.netlib.org/cgi-bin/netlibfiles.tgz?format=tgz&filename=/lapack/lapack_routine/zlaqps.f">
*> [TGZ]</a>
*> <a href="http://www.netlib.org/cgi-bin/netlibfiles.zip?format=zip&filename=/lapack/lapack_routine/zlaqps.f">
*> [ZIP]</a>
*> <a href="http://www.netlib.org/cgi-bin/netlibfiles.txt?format=txt&filename=/lapack/lapack_routine/zlaqps.f">
*> [TXT]</a>
*> \endhtmlonly
*
*  Definition:
*  ===========
*
*       SUBROUTINE ZLAQPS( M, N, OFFSET, NB, KB, A, LDA, JPVT, TAU, VN1,
*                          VN2, AUXV, F, LDF )
*
*       .. Scalar Arguments ..
*       INTEGER            KB, LDA, LDF, M, N, NB, OFFSET
*       ..
*       .. Array Arguments ..
*       INTEGER            JPVT( * )
*       DOUBLE PRECISION   VN1( * ), VN2( * )
*       COMPLEX*16         A( LDA, * ), AUXV( * ), F( LDF, * ), TAU( * )
*       ..
*
*
*> \par Purpose:
*  =============
*>
*> \verbatim
*>
*> ZLAQPS computes a step of QR factorization with column pivoting
*> of a complex M-by-N matrix A by using Blas-3.  It tries to factorize
*> NB columns from A starting from the row OFFSET+1, and updates all
*> of the matrix with Blas-3 xGEMM.
*>
*> In some cases, due to catastrophic cancellations, it cannot
*> factorize NB columns.  Hence, the actual number of factorized
*> columns is returned in KB.
*>
*> Block A(1:OFFSET,1:N) is accordingly pivoted, but not factorized.
*> \endverbatim
*
*  Arguments:
*  ==========
*
*> \param[in] M
*> \verbatim
*>          M is INTEGER
*>          The number of rows of the matrix A. M >= 0.
*> \endverbatim
*>
*> \param[in] N
*> \verbatim
*>          N is INTEGER
*>          The number of columns of the matrix A. N >= 0
*> \endverbatim
*>
*> \param[in] OFFSET
*> \verbatim
*>          OFFSET is INTEGER
*>          The number of rows of A that have been factorized in
*>          previous steps.
*> \endverbatim
*>
*> \param[in] NB
*> \verbatim
*>          NB is INTEGER
*>          The number of columns to factorize.
*> \endverbatim
*>
*> \param[out] KB
*> \verbatim
*>          KB is INTEGER
*>          The number of columns actually factorized.
*> \endverbatim
*>
*> \param[in,out] A
*> \verbatim
*>          A is COMPLEX*16 array, dimension (LDA,N)
*>          On entry, the M-by-N matrix A.
*>          On exit, block A(OFFSET+1:M,1:KB) is the triangular
*>          factor obtained and block A(1:OFFSET,1:N) has been
*>          accordingly pivoted, but no factorized.
*>          The rest of the matrix, block A(OFFSET+1:M,KB+1:N) has
*>          been updated.
*> \endverbatim
*>
*> \param[in] LDA
*> \verbatim
*>          LDA is INTEGER
*>          The leading dimension of the array A. LDA >= max(1,M).
*> \endverbatim
*>
*> \param[in,out] JPVT
*> \verbatim
*>          JPVT is INTEGER array, dimension (N)
*>          JPVT(I) = K <==> Column K of the full matrix A has been
*>          permuted into position I in AP.
*> \endverbatim
*>
*> \param[out] TAU
*> \verbatim
*>          TAU is COMPLEX*16 array, dimension (KB)
*>          The scalar factors of the elementary reflectors.
*> \endverbatim
*>
*> \param[in,out] VN1
*> \verbatim
*>          VN1 is DOUBLE PRECISION array, dimension (N)
*>          The vector with the partial column norms.
*> \endverbatim
*>
*> \param[in,out] VN2
*> \verbatim
*>          VN2 is DOUBLE PRECISION array, dimension (N)
*>          The vector with the exact column norms.
*> \endverbatim
*>
*> \param[in,out] AUXV
*> \verbatim
*>          AUXV is COMPLEX*16 array, dimension (NB)
*>          Auxiliar vector.
*> \endverbatim
*>
*> \param[in,out] F
*> \verbatim
*>          F is COMPLEX*16 array, dimension (LDF,NB)
*>          Matrix F**H = L * Y**H * A.
*> \endverbatim
*>
*> \param[in] LDF
*> \verbatim
*>          LDF is INTEGER
*>          The leading dimension of the array F. LDF >= max(1,N).
*> \endverbatim
*
*  Authors:
*  ========
*
*> \author Univ. of Tennessee
*> \author Univ. of California Berkeley
*> \author Univ. of Colorado Denver
*> \author NAG Ltd.
*
*> \date December 2016
*
*> \ingroup complex16OTHERauxiliary
*
*> \par Contributors:
*  ==================
*>
*>    G. Quintana-Orti, Depto. de Informatica, Universidad Jaime I, Spain
*>    X. Sun, Computer Science Dept., Duke University, USA
*> \n
*>  Partial column norm updating strategy modified on April 2011
*>    Z. Drmac and Z. Bujanovic, Dept. of Mathematics,
*>    University of Zagreb, Croatia.
*
*> \par References:
*  ================
*>
*> LAPACK Working Note 176
*
*> \htmlonly
*> <a href="http://www.netlib.org/lapack/lawnspdf/lawn176.pdf">[PDF]</a>
*> \endhtmlonly
*
*  =====================================================================
      SUBROUTINE ZLAQPS( M, N, OFFSET, NB, KB, A, LDA, JPVT, TAU, VN1,
     $                   VN2, AUXV, F, LDF )
*
*  -- LAPACK auxiliary routine (version 3.7.0) --
*  -- LAPACK is a software package provided by Univ. of Tennessee,    --
*  -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..--
*     December 2016
*
*     .. Scalar Arguments ..
      INTEGER            KB, LDA, LDF, M, N, NB, OFFSET
*     ..
*     .. Array Arguments ..
      INTEGER            JPVT( * )
      DOUBLE PRECISION   VN1( * ), VN2( * )
      COMPLEX*16         A( LDA, * ), AUXV( * ), F( LDF, * ), TAU( * )
*     ..
*
*  =====================================================================
*
*     .. Parameters ..
      DOUBLE PRECISION   ZERO, ONE
      COMPLEX*16         CZERO, CONE
      PARAMETER          ( ZERO = 0.0D+0, ONE = 1.0D+0,
     $                   CZERO = ( 0.0D+0, 0.0D+0 ),
     $                   CONE = ( 1.0D+0, 0.0D+0 ) )
*     ..
*     .. Local Scalars ..
      INTEGER            ITEMP, J, K, LASTRK, LSTICC, PVT, RK
      DOUBLE PRECISION   TEMP, TEMP2, TOL3Z
      COMPLEX*16         AKK
*     ..
*     .. External Subroutines ..
      EXTERNAL           ZGEMM, ZGEMV, ZLARFG, ZSWAP
*     ..
*     .. Intrinsic Functions ..
      INTRINSIC          ABS, DBLE, DCONJG, MAX, MIN, NINT, SQRT
*     ..
*     .. External Functions ..
      INTEGER            IDAMAX
      DOUBLE PRECISION   DLAMCH, DZNRM2
      EXTERNAL           IDAMAX, DLAMCH, DZNRM2
*     ..
*     .. Executable Statements ..
*
      LASTRK = MIN( M, N+OFFSET )
      LSTICC = 0
      K = 0
      TOL3Z = SQRT(DLAMCH('Epsilon'))
*
*     Beginning of while loop.
*
   10 CONTINUE
      IF( ( K.LT.NB ) .AND. ( LSTICC.EQ.0 ) ) THEN
         K = K + 1
         RK = OFFSET + K
*
*        Determine ith pivot column and swap if necessary
*
         PVT = ( K-1 ) + IDAMAX( N-K+1, VN1( K ), 1 )
         IF( PVT.NE.K ) THEN
            CALL ZSWAP( M, A( 1, PVT ), 1, A( 1, K ), 1 )
            CALL ZSWAP( K-1, F( PVT, 1 ), LDF, F( K, 1 ), LDF )
            ITEMP = JPVT( PVT )
            JPVT( PVT ) = JPVT( K )
            JPVT( K ) = ITEMP
            VN1( PVT ) = VN1( K )
            VN2( PVT ) = VN2( K )
         END IF
*
*        Apply previous Householder reflectors to column K:
*        A(RK:M,K) := A(RK:M,K) - A(RK:M,1:K-1)*F(K,1:K-1)**H.
*
         IF( K.GT.1 ) THEN
            DO 20 J = 1, K - 1
               F( K, J ) = DCONJG( F( K, J ) )
   20       CONTINUE
            CALL ZGEMV( 'No transpose', M-RK+1, K-1, -CONE, A( RK, 1 ),
     $                  LDA, F( K, 1 ), LDF, CONE, A( RK, K ), 1 )
            DO 30 J = 1, K - 1
               F( K, J ) = DCONJG( F( K, J ) )
   30       CONTINUE
         END IF
*
*        Generate elementary reflector H(k).
*
         IF( RK.LT.M ) THEN
            CALL ZLARFG( M-RK+1, A( RK, K ), A( RK+1, K ), 1, TAU( K ) )
         ELSE
            CALL ZLARFG( 1, A( RK, K ), A( RK, K ), 1, TAU( K ) )
         END IF
*
         AKK = A( RK, K )
         A( RK, K ) = CONE
*
*        Compute Kth column of F:
*
*        Compute  F(K+1:N,K) := tau(K)*A(RK:M,K+1:N)**H*A(RK:M,K).
*
         IF( K.LT.N ) THEN
            CALL ZGEMV( 'Conjugate transpose', M-RK+1, N-K, TAU( K ),
     $                  A( RK, K+1 ), LDA, A( RK, K ), 1, CZERO,
     $                  F( K+1, K ), 1 )
         END IF
*
*        Padding F(1:K,K) with zeros.
*
         DO 40 J = 1, K
            F( J, K ) = CZERO
   40    CONTINUE
*
*        Incremental updating of F:
*        F(1:N,K) := F(1:N,K) - tau(K)*F(1:N,1:K-1)*A(RK:M,1:K-1)**H
*                    *A(RK:M,K).
*
         IF( K.GT.1 ) THEN
            CALL ZGEMV( 'Conjugate transpose', M-RK+1, K-1, -TAU( K ),
     $                  A( RK, 1 ), LDA, A( RK, K ), 1, CZERO,
     $                  AUXV( 1 ), 1 )
*
            CALL ZGEMV( 'No transpose', N, K-1, CONE, F( 1, 1 ), LDF,
     $                  AUXV( 1 ), 1, CONE, F( 1, K ), 1 )
         END IF
*
*        Update the current row of A:
*        A(RK,K+1:N) := A(RK,K+1:N) - A(RK,1:K)*F(K+1:N,1:K)**H.
*
         IF( K.LT.N ) THEN
            CALL ZGEMM( 'No transpose', 'Conjugate transpose', 1, N-K,
     $                  K, -CONE, A( RK, 1 ), LDA, F( K+1, 1 ), LDF,
     $                  CONE, A( RK, K+1 ), LDA )
         END IF
*
*        Update partial column norms.
*
         IF( RK.LT.LASTRK ) THEN
            DO 50 J = K + 1, N
               IF( VN1( J ).NE.ZERO ) THEN
*
*                 NOTE: The following 4 lines follow from the analysis in
*                 Lapack Working Note 176.
*
                  TEMP = ABS( A( RK, J ) ) / VN1( J )
                  TEMP = MAX( ZERO, ( ONE+TEMP )*( ONE-TEMP ) )
                  TEMP2 = TEMP*( VN1( J ) / VN2( J ) )**2
                  IF( TEMP2 .LE. TOL3Z ) THEN
                     VN2( J ) = DBLE( LSTICC )
                     LSTICC = J
                  ELSE
                     VN1( J ) = VN1( J )*SQRT( TEMP )
                  END IF
               END IF
   50       CONTINUE
         END IF
*
         A( RK, K ) = AKK
*
*        End of while loop.
*
         GO TO 10
      END IF
      KB = K
      RK = OFFSET + KB
*
*     Apply the block reflector to the rest of the matrix:
*     A(OFFSET+KB+1:M,KB+1:N) := A(OFFSET+KB+1:M,KB+1:N) -
*                         A(OFFSET+KB+1:M,1:KB)*F(KB+1:N,1:KB)**H.
*
      IF( KB.LT.MIN( N, M-OFFSET ) ) THEN
         CALL ZGEMM( 'No transpose', 'Conjugate transpose', M-RK, N-KB,
     $               KB, -CONE, A( RK+1, 1 ), LDA, F( KB+1, 1 ), LDF,
     $               CONE, A( RK+1, KB+1 ), LDA )
      END IF
*
*     Recomputation of difficult columns.
*
   60 CONTINUE
      IF( LSTICC.GT.0 ) THEN
         ITEMP = NINT( VN2( LSTICC ) )
         VN1( LSTICC ) = DZNRM2( M-RK, A( RK+1, LSTICC ), 1 )
*
*        NOTE: The computation of VN1( LSTICC ) relies on the fact that
*        SNRM2 does not fail on vectors with norm below the value of
*        SQRT(DLAMCH('S'))
*
         VN2( LSTICC ) = VN1( LSTICC )
         LSTICC = ITEMP
         GO TO 60
      END IF
*
      RETURN
*
*     End of ZLAQPS
*
      END