1       SUBROUTINE ZPBSVX( FACT, UPLO, N, KD, NRHS, AB, LDAB, AFB, LDAFB,
  2      $                   EQUED, S, B, LDB, X, LDX, RCOND, FERR, BERR,
  3      $                   WORK, RWORK, INFO )
  4 *
  5 *  -- LAPACK driver routine (version 3.3.1) --
  6 *  -- LAPACK is a software package provided by Univ. of Tennessee,    --
  7 *  -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..--
  8 *  -- April 2011                                                      --
  9 *
 10 *     .. Scalar Arguments ..
 11       CHARACTER          EQUED, FACT, UPLO
 12       INTEGER            INFO, KD, LDAB, LDAFB, LDB, LDX, N, NRHS
 13       DOUBLE PRECISION   RCOND
 14 *     ..
 15 *     .. Array Arguments ..
 16       DOUBLE PRECISION   BERR( * ), FERR( * ), RWORK( * ), S( * )
 17       COMPLEX*16         AB( LDAB, * ), AFB( LDAFB, * ), B( LDB, * ),
 18      $                   WORK( * ), X( LDX, * )
 19 *     ..
 20 *
 21 *  Purpose
 22 *  =======
 23 *
 24 *  ZPBSVX uses the Cholesky factorization A = U**H*U or A = L*L**H to
 25 *  compute the solution to a complex system of linear equations
 26 *     A * X = B,
 27 *  where A is an N-by-N Hermitian positive definite band matrix and X
 28 *  and B are N-by-NRHS matrices.
 29 *
 30 *  Error bounds on the solution and a condition estimate are also
 31 *  provided.
 32 *
 33 *  Description
 34 *  ===========
 35 *
 36 *  The following steps are performed:
 37 *
 38 *  1. If FACT = 'E', real scaling factors are computed to equilibrate
 39 *     the system:
 40 *        diag(S) * A * diag(S) * inv(diag(S)) * X = diag(S) * B
 41 *     Whether or not the system will be equilibrated depends on the
 42 *     scaling of the matrix A, but if equilibration is used, A is
 43 *     overwritten by diag(S)*A*diag(S) and B by diag(S)*B.
 44 *
 45 *  2. If FACT = 'N' or 'E', the Cholesky decomposition is used to
 46 *     factor the matrix A (after equilibration if FACT = 'E') as
 47 *        A = U**H * U,  if UPLO = 'U', or
 48 *        A = L * L**H,  if UPLO = 'L',
 49 *     where U is an upper triangular band matrix, and L is a lower
 50 *     triangular band matrix.
 51 *
 52 *  3. If the leading i-by-i principal minor is not positive definite,
 53 *     then the routine returns with INFO = i. Otherwise, the factored
 54 *     form of A is used to estimate the condition number of the matrix
 55 *     A.  If the reciprocal of the condition number is less than machine
 56 *     precision, INFO = N+1 is returned as a warning, but the routine
 57 *     still goes on to solve for X and compute error bounds as
 58 *     described below.
 59 *
 60 *  4. The system of equations is solved for X using the factored form
 61 *     of A.
 62 *
 63 *  5. Iterative refinement is applied to improve the computed solution
 64 *     matrix and calculate error bounds and backward error estimates
 65 *     for it.
 66 *
 67 *  6. If equilibration was used, the matrix X is premultiplied by
 68 *     diag(S) so that it solves the original system before
 69 *     equilibration.
 70 *
 71 *  Arguments
 72 *  =========
 73 *
 74 *  FACT    (input) CHARACTER*1
 75 *          Specifies whether or not the factored form of the matrix A is
 76 *          supplied on entry, and if not, whether the matrix A should be
 77 *          equilibrated before it is factored.
 78 *          = 'F':  On entry, AFB contains the factored form of A.
 79 *                  If EQUED = 'Y', the matrix A has been equilibrated
 80 *                  with scaling factors given by S.  AB and AFB will not
 81 *                  be modified.
 82 *          = 'N':  The matrix A will be copied to AFB and factored.
 83 *          = 'E':  The matrix A will be equilibrated if necessary, then
 84 *                  copied to AFB and factored.
 85 *
 86 *  UPLO    (input) CHARACTER*1
 87 *          = 'U':  Upper triangle of A is stored;
 88 *          = 'L':  Lower triangle of A is stored.
 89 *
 90 *  N       (input) INTEGER
 91 *          The number of linear equations, i.e., the order of the
 92 *          matrix A.  N >= 0.
 93 *
 94 *  KD      (input) INTEGER
 95 *          The number of superdiagonals of the matrix A if UPLO = 'U',
 96 *          or the number of subdiagonals if UPLO = 'L'.  KD >= 0.
 97 *
 98 *  NRHS    (input) INTEGER
 99 *          The number of right-hand sides, i.e., the number of columns
100 *          of the matrices B and X.  NRHS >= 0.
101 *
102 *  AB      (input/output) COMPLEX*16 array, dimension (LDAB,N)
103 *          On entry, the upper or lower triangle of the Hermitian band
104 *          matrix A, stored in the first KD+1 rows of the array, except
105 *          if FACT = 'F' and EQUED = 'Y', then A must contain the
106 *          equilibrated matrix diag(S)*A*diag(S).  The j-th column of A
107 *          is stored in the j-th column of the array AB as follows:
108 *          if UPLO = 'U', AB(KD+1+i-j,j) = A(i,j) for max(1,j-KD)<=i<=j;
109 *          if UPLO = 'L', AB(1+i-j,j)    = A(i,j) for j<=i<=min(N,j+KD).
110 *          See below for further details.
111 *
112 *          On exit, if FACT = 'E' and EQUED = 'Y', A is overwritten by
113 *          diag(S)*A*diag(S).
114 *
115 *  LDAB    (input) INTEGER
116 *          The leading dimension of the array A.  LDAB >= KD+1.
117 *
118 *  AFB     (input or output) COMPLEX*16 array, dimension (LDAFB,N)
119 *          If FACT = 'F', then AFB is an input argument and on entry
120 *          contains the triangular factor U or L from the Cholesky
121 *          factorization A = U**H *U or A = L*L**H of the band matrix
122 *          A, in the same storage format as A (see AB).  If EQUED = 'Y',
123 *          then AFB is the factored form of the equilibrated matrix A.
124 *
125 *          If FACT = 'N', then AFB is an output argument and on exit
126 *          returns the triangular factor U or L from the Cholesky
127 *          factorization A = U**H *U or A = L*L**H.
128 *
129 *          If FACT = 'E', then AFB is an output argument and on exit
130 *          returns the triangular factor U or L from the Cholesky
131 *          factorization A = U**H *U or A = L*L**H of the equilibrated
132 *          matrix A (see the description of A for the form of the
133 *          equilibrated matrix).
134 *
135 *  LDAFB   (input) INTEGER
136 *          The leading dimension of the array AFB.  LDAFB >= KD+1.
137 *
138 *  EQUED   (input or output) CHARACTER*1
139 *          Specifies the form of equilibration that was done.
140 *          = 'N':  No equilibration (always true if FACT = 'N').
141 *          = 'Y':  Equilibration was done, i.e., A has been replaced by
142 *                  diag(S) * A * diag(S).
143 *          EQUED is an input argument if FACT = 'F'; otherwise, it is an
144 *          output argument.
145 *
146 *  S       (input or output) DOUBLE PRECISION array, dimension (N)
147 *          The scale factors for A; not accessed if EQUED = 'N'.  S is
148 *          an input argument if FACT = 'F'; otherwise, S is an output
149 *          argument.  If FACT = 'F' and EQUED = 'Y', each element of S
150 *          must be positive.
151 *
152 *  B       (input/output) COMPLEX*16 array, dimension (LDB,NRHS)
153 *          On entry, the N-by-NRHS right hand side matrix B.
154 *          On exit, if EQUED = 'N', B is not modified; if EQUED = 'Y',
155 *          B is overwritten by diag(S) * B.
156 *
157 *  LDB     (input) INTEGER
158 *          The leading dimension of the array B.  LDB >= max(1,N).
159 *
160 *  X       (output) COMPLEX*16 array, dimension (LDX,NRHS)
161 *          If INFO = 0 or INFO = N+1, the N-by-NRHS solution matrix X to
162 *          the original system of equations.  Note that if EQUED = 'Y',
163 *          A and B are modified on exit, and the solution to the
164 *          equilibrated system is inv(diag(S))*X.
165 *
166 *  LDX     (input) INTEGER
167 *          The leading dimension of the array X.  LDX >= max(1,N).
168 *
169 *  RCOND   (output) DOUBLE PRECISION
170 *          The estimate of the reciprocal condition number of the matrix
171 *          A after equilibration (if done).  If RCOND is less than the
172 *          machine precision (in particular, if RCOND = 0), the matrix
173 *          is singular to working precision.  This condition is
174 *          indicated by a return code of INFO > 0.
175 *
176 *  FERR    (output) DOUBLE PRECISION array, dimension (NRHS)
177 *          The estimated forward error bound for each solution vector
178 *          X(j) (the j-th column of the solution matrix X).
179 *          If XTRUE is the true solution corresponding to X(j), FERR(j)
180 *          is an estimated upper bound for the magnitude of the largest
181 *          element in (X(j) - XTRUE) divided by the magnitude of the
182 *          largest element in X(j).  The estimate is as reliable as
183 *          the estimate for RCOND, and is almost always a slight
184 *          overestimate of the true error.
185 *
186 *  BERR    (output) DOUBLE PRECISION array, dimension (NRHS)
187 *          The componentwise relative backward error of each solution
188 *          vector X(j) (i.e., the smallest relative change in
189 *          any element of A or B that makes X(j) an exact solution).
190 *
191 *  WORK    (workspace) COMPLEX*16 array, dimension (2*N)
192 *
193 *  RWORK   (workspace) DOUBLE PRECISION array, dimension (N)
194 *
195 *  INFO    (output) INTEGER
196 *          = 0: successful exit
197 *          < 0: if INFO = -i, the i-th argument had an illegal value
198 *          > 0: if INFO = i, and i is
199 *                <= N:  the leading minor of order i of A is
200 *                       not positive definite, so the factorization
201 *                       could not be completed, and the solution has not
202 *                       been computed. RCOND = 0 is returned.
203 *                = N+1: U is nonsingular, but RCOND is less than machine
204 *                       precision, meaning that the matrix is singular
205 *                       to working precision.  Nevertheless, the
206 *                       solution and error bounds are computed because
207 *                       there are a number of situations where the
208 *                       computed solution can be more accurate than the
209 *                       value of RCOND would suggest.
210 *
211 *  Further Details
212 *  ===============
213 *
214 *  The band storage scheme is illustrated by the following example, when
215 *  N = 6, KD = 2, and UPLO = 'U':
216 *
217 *  Two-dimensional storage of the Hermitian matrix A:
218 *
219 *     a11  a12  a13
220 *          a22  a23  a24
221 *               a33  a34  a35
222 *                    a44  a45  a46
223 *                         a55  a56
224 *     (aij=conjg(aji))         a66
225 *
226 *  Band storage of the upper triangle of A:
227 *
228 *      *    *   a13  a24  a35  a46
229 *      *   a12  a23  a34  a45  a56
230 *     a11  a22  a33  a44  a55  a66
231 *
232 *  Similarly, if UPLO = 'L' the format of A is as follows:
233 *
234 *     a11  a22  a33  a44  a55  a66
235 *     a21  a32  a43  a54  a65   *
236 *     a31  a42  a53  a64   *    *
237 *
238 *  Array elements marked * are not used by the routine.
239 *
240 *  =====================================================================
241 *
242 *     .. Parameters ..
243       DOUBLE PRECISION   ZERO, ONE
244       PARAMETER          ( ZERO = 0.0D+0, ONE = 1.0D+0 )
245 *     ..
246 *     .. Local Scalars ..
247       LOGICAL            EQUIL, NOFACT, RCEQU, UPPER
248       INTEGER            I, INFEQU, J, J1, J2
249       DOUBLE PRECISION   AMAX, ANORM, BIGNUM, SCOND, SMAX, SMIN, SMLNUM
250 *     ..
251 *     .. External Functions ..
252       LOGICAL            LSAME
253       DOUBLE PRECISION   DLAMCH, ZLANHB
254       EXTERNAL           LSAME, DLAMCH, ZLANHB
255 *     ..
256 *     .. External Subroutines ..
257       EXTERNAL           XERBLA, ZCOPY, ZLACPY, ZLAQHB, ZPBCON, ZPBEQU,
258      $                   ZPBRFS, ZPBTRF, ZPBTRS
259 *     ..
260 *     .. Intrinsic Functions ..
261       INTRINSIC          MAXMIN
262 *     ..
263 *     .. Executable Statements ..
264 *
265       INFO = 0
266       NOFACT = LSAME( FACT, 'N' )
267       EQUIL = LSAME( FACT, 'E' )
268       UPPER = LSAME( UPLO, 'U' )
269       IF( NOFACT .OR. EQUIL ) THEN
270          EQUED = 'N'
271          RCEQU = .FALSE.
272       ELSE
273          RCEQU = LSAME( EQUED, 'Y' )
274          SMLNUM = DLAMCH( 'Safe minimum' )
275          BIGNUM = ONE / SMLNUM
276       END IF
277 *
278 *     Test the input parameters.
279 *
280       IF.NOT.NOFACT .AND. .NOT.EQUIL .AND. .NOT.LSAME( FACT, 'F' ) )
281      $     THEN
282          INFO = -1
283       ELSE IF.NOT.UPPER .AND. .NOT.LSAME( UPLO, 'L' ) ) THEN
284          INFO = -2
285       ELSE IF( N.LT.0 ) THEN
286          INFO = -3
287       ELSE IF( KD.LT.0 ) THEN
288          INFO = -4
289       ELSE IF( NRHS.LT.0 ) THEN
290          INFO = -5
291       ELSE IF( LDAB.LT.KD+1 ) THEN
292          INFO = -7
293       ELSE IF( LDAFB.LT.KD+1 ) THEN
294          INFO = -9
295       ELSE IF( LSAME( FACT, 'F' ) .AND. .NOT.
296      $         ( RCEQU .OR. LSAME( EQUED, 'N' ) ) ) THEN
297          INFO = -10
298       ELSE
299          IF( RCEQU ) THEN
300             SMIN = BIGNUM
301             SMAX = ZERO
302             DO 10 J = 1, N
303                SMIN = MIN( SMIN, S( J ) )
304                SMAX = MAX( SMAX, S( J ) )
305    10       CONTINUE
306             IF( SMIN.LE.ZERO ) THEN
307                INFO = -11
308             ELSE IF( N.GT.0 ) THEN
309                SCOND = MAX( SMIN, SMLNUM ) / MIN( SMAX, BIGNUM )
310             ELSE
311                SCOND = ONE
312             END IF
313          END IF
314          IF( INFO.EQ.0 ) THEN
315             IF( LDB.LT.MAX1, N ) ) THEN
316                INFO = -13
317             ELSE IF( LDX.LT.MAX1, N ) ) THEN
318                INFO = -15
319             END IF
320          END IF
321       END IF
322 *
323       IF( INFO.NE.0 ) THEN
324          CALL XERBLA( 'ZPBSVX'-INFO )
325          RETURN
326       END IF
327 *
328       IF( EQUIL ) THEN
329 *
330 *        Compute row and column scalings to equilibrate the matrix A.
331 *
332          CALL ZPBEQU( UPLO, N, KD, AB, LDAB, S, SCOND, AMAX, INFEQU )
333          IF( INFEQU.EQ.0 ) THEN
334 *
335 *           Equilibrate the matrix.
336 *
337             CALL ZLAQHB( UPLO, N, KD, AB, LDAB, S, SCOND, AMAX, EQUED )
338             RCEQU = LSAME( EQUED, 'Y' )
339          END IF
340       END IF
341 *
342 *     Scale the right-hand side.
343 *
344       IF( RCEQU ) THEN
345          DO 30 J = 1, NRHS
346             DO 20 I = 1, N
347                B( I, J ) = S( I )*B( I, J )
348    20       CONTINUE
349    30    CONTINUE
350       END IF
351 *
352       IF( NOFACT .OR. EQUIL ) THEN
353 *
354 *        Compute the Cholesky factorization A = U**H *U or A = L*L**H.
355 *
356          IF( UPPER ) THEN
357             DO 40 J = 1, N
358                J1 = MAX( J-KD, 1 )
359                CALL ZCOPY( J-J1+1, AB( KD+1-J+J1, J ), 1,
360      $                     AFB( KD+1-J+J1, J ), 1 )
361    40       CONTINUE
362          ELSE
363             DO 50 J = 1, N
364                J2 = MIN( J+KD, N )
365                CALL ZCOPY( J2-J+1, AB( 1, J ), 1, AFB( 1, J ), 1 )
366    50       CONTINUE
367          END IF
368 *
369          CALL ZPBTRF( UPLO, N, KD, AFB, LDAFB, INFO )
370 *
371 *        Return if INFO is non-zero.
372 *
373          IF( INFO.GT.0 )THEN
374             RCOND = ZERO
375             RETURN
376          END IF
377       END IF
378 *
379 *     Compute the norm of the matrix A.
380 *
381       ANORM = ZLANHB( '1', UPLO, N, KD, AB, LDAB, RWORK )
382 *
383 *     Compute the reciprocal of the condition number of A.
384 *
385       CALL ZPBCON( UPLO, N, KD, AFB, LDAFB, ANORM, RCOND, WORK, RWORK,
386      $             INFO )
387 *
388 *     Compute the solution matrix X.
389 *
390       CALL ZLACPY( 'Full', N, NRHS, B, LDB, X, LDX )
391       CALL ZPBTRS( UPLO, N, KD, NRHS, AFB, LDAFB, X, LDX, INFO )
392 *
393 *     Use iterative refinement to improve the computed solution and
394 *     compute error bounds and backward error estimates for it.
395 *
396       CALL ZPBRFS( UPLO, N, KD, NRHS, AB, LDAB, AFB, LDAFB, B, LDB, X,
397      $             LDX, FERR, BERR, WORK, RWORK, INFO )
398 *
399 *     Transform the solution matrix X to a solution of the original
400 *     system.
401 *
402       IF( RCEQU ) THEN
403          DO 70 J = 1, NRHS
404             DO 60 I = 1, N
405                X( I, J ) = S( I )*X( I, J )
406    60       CONTINUE
407    70    CONTINUE
408          DO 80 J = 1, NRHS
409             FERR( J ) = FERR( J ) / SCOND
410    80    CONTINUE
411       END IF
412 *
413 *     Set INFO = N+1 if the matrix is singular to working precision.
414 *
415       IF( RCOND.LT.DLAMCH( 'Epsilon' ) )
416      $   INFO = N + 1
417 *
418       RETURN
419 *
420 *     End of ZPBSVX
421 *
422       END