LAPACK  3.4.2
LAPACK: Linear Algebra PACKage
 All Files Functions Groups
chpsvx.f
Go to the documentation of this file.
1 *> \brief <b> CHPSVX computes the solution to system of linear equations A * X = B for OTHER matrices</b>
2 *
3 * =========== DOCUMENTATION ===========
4 *
5 * Online html documentation available at
6 * http://www.netlib.org/lapack/explore-html/
7 *
8 *> \htmlonly
9 *> Download CHPSVX + dependencies
10 *> <a href="http://www.netlib.org/cgi-bin/netlibfiles.tgz?format=tgz&filename=/lapack/lapack_routine/chpsvx.f">
11 *> [TGZ]</a>
12 *> <a href="http://www.netlib.org/cgi-bin/netlibfiles.zip?format=zip&filename=/lapack/lapack_routine/chpsvx.f">
13 *> [ZIP]</a>
14 *> <a href="http://www.netlib.org/cgi-bin/netlibfiles.txt?format=txt&filename=/lapack/lapack_routine/chpsvx.f">
15 *> [TXT]</a>
16 *> \endhtmlonly
17 *
18 * Definition:
19 * ===========
20 *
21 * SUBROUTINE CHPSVX( FACT, UPLO, N, NRHS, AP, AFP, IPIV, B, LDB, X,
22 * LDX, RCOND, FERR, BERR, WORK, RWORK, INFO )
23 *
24 * .. Scalar Arguments ..
25 * CHARACTER FACT, UPLO
26 * INTEGER INFO, LDB, LDX, N, NRHS
27 * REAL RCOND
28 * ..
29 * .. Array Arguments ..
30 * INTEGER IPIV( * )
31 * REAL BERR( * ), FERR( * ), RWORK( * )
32 * COMPLEX AFP( * ), AP( * ), B( LDB, * ), WORK( * ),
33 * $ X( LDX, * )
34 * ..
35 *
36 *
37 *> \par Purpose:
38 * =============
39 *>
40 *> \verbatim
41 *>
42 *> CHPSVX uses the diagonal pivoting factorization A = U*D*U**H or
43 *> A = L*D*L**H to compute the solution to a complex system of linear
44 *> equations A * X = B, where A is an N-by-N Hermitian matrix stored
45 *> in packed format and X and B are N-by-NRHS matrices.
46 *>
47 *> Error bounds on the solution and a condition estimate are also
48 *> provided.
49 *> \endverbatim
50 *
51 *> \par Description:
52 * =================
53 *>
54 *> \verbatim
55 *>
56 *> The following steps are performed:
57 *>
58 *> 1. If FACT = 'N', the diagonal pivoting method is used to factor A as
59 *> A = U * D * U**H, if UPLO = 'U', or
60 *> A = L * D * L**H, if UPLO = 'L',
61 *> where U (or L) is a product of permutation and unit upper (lower)
62 *> triangular matrices and D is Hermitian and block diagonal with
63 *> 1-by-1 and 2-by-2 diagonal blocks.
64 *>
65 *> 2. If some D(i,i)=0, so that D is exactly singular, then the routine
66 *> returns with INFO = i. Otherwise, the factored form of A is used
67 *> to estimate the condition number of the matrix A. If the
68 *> reciprocal of the condition number is less than machine precision,
69 *> INFO = N+1 is returned as a warning, but the routine still goes on
70 *> to solve for X and compute error bounds as described below.
71 *>
72 *> 3. The system of equations is solved for X using the factored form
73 *> of A.
74 *>
75 *> 4. Iterative refinement is applied to improve the computed solution
76 *> matrix and calculate error bounds and backward error estimates
77 *> for it.
78 *> \endverbatim
79 *
80 * Arguments:
81 * ==========
82 *
83 *> \param[in] FACT
84 *> \verbatim
85 *> FACT is CHARACTER*1
86 *> Specifies whether or not the factored form of A has been
87 *> supplied on entry.
88 *> = 'F': On entry, AFP and IPIV contain the factored form of
89 *> A. AFP and IPIV will not be modified.
90 *> = 'N': The matrix A will be copied to AFP and factored.
91 *> \endverbatim
92 *>
93 *> \param[in] UPLO
94 *> \verbatim
95 *> UPLO is CHARACTER*1
96 *> = 'U': Upper triangle of A is stored;
97 *> = 'L': Lower triangle of A is stored.
98 *> \endverbatim
99 *>
100 *> \param[in] N
101 *> \verbatim
102 *> N is INTEGER
103 *> The number of linear equations, i.e., the order of the
104 *> matrix A. N >= 0.
105 *> \endverbatim
106 *>
107 *> \param[in] NRHS
108 *> \verbatim
109 *> NRHS is INTEGER
110 *> The number of right hand sides, i.e., the number of columns
111 *> of the matrices B and X. NRHS >= 0.
112 *> \endverbatim
113 *>
114 *> \param[in] AP
115 *> \verbatim
116 *> AP is COMPLEX array, dimension (N*(N+1)/2)
117 *> The upper or lower triangle of the Hermitian matrix A, packed
118 *> columnwise in a linear array. The j-th column of A is stored
119 *> in the array AP as follows:
120 *> if UPLO = 'U', AP(i + (j-1)*j/2) = A(i,j) for 1<=i<=j;
121 *> if UPLO = 'L', AP(i + (j-1)*(2*n-j)/2) = A(i,j) for j<=i<=n.
122 *> See below for further details.
123 *> \endverbatim
124 *>
125 *> \param[in,out] AFP
126 *> \verbatim
127 *> AFP is COMPLEX array, dimension (N*(N+1)/2)
128 *> If FACT = 'F', then AFP is an input argument and on entry
129 *> contains the block diagonal matrix D and the multipliers used
130 *> to obtain the factor U or L from the factorization
131 *> A = U*D*U**H or A = L*D*L**H as computed by CHPTRF, stored as
132 *> a packed triangular matrix in the same storage format as A.
133 *>
134 *> If FACT = 'N', then AFP is an output argument and on exit
135 *> contains the block diagonal matrix D and the multipliers used
136 *> to obtain the factor U or L from the factorization
137 *> A = U*D*U**H or A = L*D*L**H as computed by CHPTRF, stored as
138 *> a packed triangular matrix in the same storage format as A.
139 *> \endverbatim
140 *>
141 *> \param[in,out] IPIV
142 *> \verbatim
143 *> IPIV is INTEGER array, dimension (N)
144 *> If FACT = 'F', then IPIV is an input argument and on entry
145 *> contains details of the interchanges and the block structure
146 *> of D, as determined by CHPTRF.
147 *> If IPIV(k) > 0, then rows and columns k and IPIV(k) were
148 *> interchanged and D(k,k) is a 1-by-1 diagonal block.
149 *> If UPLO = 'U' and IPIV(k) = IPIV(k-1) < 0, then rows and
150 *> columns k-1 and -IPIV(k) were interchanged and D(k-1:k,k-1:k)
151 *> is a 2-by-2 diagonal block. If UPLO = 'L' and IPIV(k) =
152 *> IPIV(k+1) < 0, then rows and columns k+1 and -IPIV(k) were
153 *> interchanged and D(k:k+1,k:k+1) is a 2-by-2 diagonal block.
154 *>
155 *> If FACT = 'N', then IPIV is an output argument and on exit
156 *> contains details of the interchanges and the block structure
157 *> of D, as determined by CHPTRF.
158 *> \endverbatim
159 *>
160 *> \param[in] B
161 *> \verbatim
162 *> B is COMPLEX array, dimension (LDB,NRHS)
163 *> The N-by-NRHS right hand side matrix B.
164 *> \endverbatim
165 *>
166 *> \param[in] LDB
167 *> \verbatim
168 *> LDB is INTEGER
169 *> The leading dimension of the array B. LDB >= max(1,N).
170 *> \endverbatim
171 *>
172 *> \param[out] X
173 *> \verbatim
174 *> X is COMPLEX array, dimension (LDX,NRHS)
175 *> If INFO = 0 or INFO = N+1, the N-by-NRHS solution matrix X.
176 *> \endverbatim
177 *>
178 *> \param[in] LDX
179 *> \verbatim
180 *> LDX is INTEGER
181 *> The leading dimension of the array X. LDX >= max(1,N).
182 *> \endverbatim
183 *>
184 *> \param[out] RCOND
185 *> \verbatim
186 *> RCOND is REAL
187 *> The estimate of the reciprocal condition number of the matrix
188 *> A. If RCOND is less than the machine precision (in
189 *> particular, if RCOND = 0), the matrix is singular to working
190 *> precision. This condition is indicated by a return code of
191 *> INFO > 0.
192 *> \endverbatim
193 *>
194 *> \param[out] FERR
195 *> \verbatim
196 *> FERR is REAL array, dimension (NRHS)
197 *> The estimated forward error bound for each solution vector
198 *> X(j) (the j-th column of the solution matrix X).
199 *> If XTRUE is the true solution corresponding to X(j), FERR(j)
200 *> is an estimated upper bound for the magnitude of the largest
201 *> element in (X(j) - XTRUE) divided by the magnitude of the
202 *> largest element in X(j). The estimate is as reliable as
203 *> the estimate for RCOND, and is almost always a slight
204 *> overestimate of the true error.
205 *> \endverbatim
206 *>
207 *> \param[out] BERR
208 *> \verbatim
209 *> BERR is REAL array, dimension (NRHS)
210 *> The componentwise relative backward error of each solution
211 *> vector X(j) (i.e., the smallest relative change in
212 *> any element of A or B that makes X(j) an exact solution).
213 *> \endverbatim
214 *>
215 *> \param[out] WORK
216 *> \verbatim
217 *> WORK is COMPLEX array, dimension (2*N)
218 *> \endverbatim
219 *>
220 *> \param[out] RWORK
221 *> \verbatim
222 *> RWORK is REAL array, dimension (N)
223 *> \endverbatim
224 *>
225 *> \param[out] INFO
226 *> \verbatim
227 *> INFO is INTEGER
228 *> = 0: successful exit
229 *> < 0: if INFO = -i, the i-th argument had an illegal value
230 *> > 0: if INFO = i, and i is
231 *> <= N: D(i,i) is exactly zero. The factorization
232 *> has been completed but the factor D is exactly
233 *> singular, so the solution and error bounds could
234 *> not be computed. RCOND = 0 is returned.
235 *> = N+1: D is nonsingular, but RCOND is less than machine
236 *> precision, meaning that the matrix is singular
237 *> to working precision. Nevertheless, the
238 *> solution and error bounds are computed because
239 *> there are a number of situations where the
240 *> computed solution can be more accurate than the
241 *> value of RCOND would suggest.
242 *> \endverbatim
243 *
244 * Authors:
245 * ========
246 *
247 *> \author Univ. of Tennessee
248 *> \author Univ. of California Berkeley
249 *> \author Univ. of Colorado Denver
250 *> \author NAG Ltd.
251 *
252 *> \date April 2012
253 *
254 *> \ingroup complexOTHERsolve
255 *
256 *> \par Further Details:
257 * =====================
258 *>
259 *> \verbatim
260 *>
261 *> The packed storage scheme is illustrated by the following example
262 *> when N = 4, UPLO = 'U':
263 *>
264 *> Two-dimensional storage of the Hermitian matrix A:
265 *>
266 *> a11 a12 a13 a14
267 *> a22 a23 a24
268 *> a33 a34 (aij = conjg(aji))
269 *> a44
270 *>
271 *> Packed storage of the upper triangle of A:
272 *>
273 *> AP = [ a11, a12, a22, a13, a23, a33, a14, a24, a34, a44 ]
274 *> \endverbatim
275 *>
276 * =====================================================================
277  SUBROUTINE chpsvx( FACT, UPLO, N, NRHS, AP, AFP, IPIV, B, LDB, X,
278  $ ldx, rcond, ferr, berr, work, rwork, info )
279 *
280 * -- LAPACK driver routine (version 3.4.1) --
281 * -- LAPACK is a software package provided by Univ. of Tennessee, --
282 * -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..--
283 * April 2012
284 *
285 * .. Scalar Arguments ..
286  CHARACTER fact, uplo
287  INTEGER info, ldb, ldx, n, nrhs
288  REAL rcond
289 * ..
290 * .. Array Arguments ..
291  INTEGER ipiv( * )
292  REAL berr( * ), ferr( * ), rwork( * )
293  COMPLEX afp( * ), ap( * ), b( ldb, * ), work( * ),
294  $ x( ldx, * )
295 * ..
296 *
297 * =====================================================================
298 *
299 * .. Parameters ..
300  REAL zero
301  parameter( zero = 0.0e+0 )
302 * ..
303 * .. Local Scalars ..
304  LOGICAL nofact
305  REAL anorm
306 * ..
307 * .. External Functions ..
308  LOGICAL lsame
309  REAL clanhp, slamch
310  EXTERNAL lsame, clanhp, slamch
311 * ..
312 * .. External Subroutines ..
313  EXTERNAL ccopy, chpcon, chprfs, chptrf, chptrs, clacpy,
314  $ xerbla
315 * ..
316 * .. Intrinsic Functions ..
317  INTRINSIC max
318 * ..
319 * .. Executable Statements ..
320 *
321 * Test the input parameters.
322 *
323  info = 0
324  nofact = lsame( fact, 'N' )
325  IF( .NOT.nofact .AND. .NOT.lsame( fact, 'F' ) ) THEN
326  info = -1
327  ELSE IF( .NOT.lsame( uplo, 'U' ) .AND. .NOT.lsame( uplo, 'L' ) )
328  $ THEN
329  info = -2
330  ELSE IF( n.LT.0 ) THEN
331  info = -3
332  ELSE IF( nrhs.LT.0 ) THEN
333  info = -4
334  ELSE IF( ldb.LT.max( 1, n ) ) THEN
335  info = -9
336  ELSE IF( ldx.LT.max( 1, n ) ) THEN
337  info = -11
338  END IF
339  IF( info.NE.0 ) THEN
340  CALL xerbla( 'CHPSVX', -info )
341  return
342  END IF
343 *
344  IF( nofact ) THEN
345 *
346 * Compute the factorization A = U*D*U**H or A = L*D*L**H.
347 *
348  CALL ccopy( n*( n+1 ) / 2, ap, 1, afp, 1 )
349  CALL chptrf( uplo, n, afp, ipiv, info )
350 *
351 * Return if INFO is non-zero.
352 *
353  IF( info.GT.0 )THEN
354  rcond = zero
355  return
356  END IF
357  END IF
358 *
359 * Compute the norm of the matrix A.
360 *
361  anorm = clanhp( 'I', uplo, n, ap, rwork )
362 *
363 * Compute the reciprocal of the condition number of A.
364 *
365  CALL chpcon( uplo, n, afp, ipiv, anorm, rcond, work, info )
366 *
367 * Compute the solution vectors X.
368 *
369  CALL clacpy( 'Full', n, nrhs, b, ldb, x, ldx )
370  CALL chptrs( uplo, n, nrhs, afp, ipiv, x, ldx, info )
371 *
372 * Use iterative refinement to improve the computed solutions and
373 * compute error bounds and backward error estimates for them.
374 *
375  CALL chprfs( uplo, n, nrhs, ap, afp, ipiv, b, ldb, x, ldx, ferr,
376  $ berr, work, rwork, info )
377 *
378 * Set INFO = N+1 if the matrix is singular to working precision.
379 *
380  IF( rcond.LT.slamch( 'Epsilon' ) )
381  $ info = n + 1
382 *
383  return
384 *
385 * End of CHPSVX
386 *
387  END