LAPACK  3.5.0
LAPACK: Linear Algebra PACKage
 All Classes Files Functions Variables Typedefs Macros
cdrvsg.f
Go to the documentation of this file.
1 *> \brief \b CDRVSG
2 *
3 * =========== DOCUMENTATION ===========
4 *
5 * Online html documentation available at
6 * http://www.netlib.org/lapack/explore-html/
7 *
8 * Definition:
9 * ===========
10 *
11 * SUBROUTINE CDRVSG( NSIZES, NN, NTYPES, DOTYPE, ISEED, THRESH,
12 * NOUNIT, A, LDA, B, LDB, D, Z, LDZ, AB, BB, AP,
13 * BP, WORK, NWORK, RWORK, LRWORK, IWORK, LIWORK,
14 * RESULT, INFO )
15 *
16 * .. Scalar Arguments ..
17 * INTEGER INFO, LDA, LDB, LDZ, LIWORK, LRWORK, NOUNIT,
18 * $ NSIZES, NTYPES, NWORK
19 * REAL THRESH
20 * ..
21 * .. Array Arguments ..
22 * LOGICAL DOTYPE( * )
23 * INTEGER ISEED( 4 ), IWORK( * ), NN( * )
24 * REAL D( * ), RESULT( * ), RWORK( * )
25 * COMPLEX A( LDA, * ), AB( LDA, * ), AP( * ),
26 * $ B( LDB, * ), BB( LDB, * ), BP( * ), WORK( * ),
27 * $ Z( LDZ, * )
28 * ..
29 *
30 *
31 *> \par Purpose:
32 * =============
33 *>
34 *> \verbatim
35 *>
36 *> CDRVSG checks the complex Hermitian generalized eigenproblem
37 *> drivers.
38 *>
39 *> CHEGV computes all eigenvalues and, optionally,
40 *> eigenvectors of a complex Hermitian-definite generalized
41 *> eigenproblem.
42 *>
43 *> CHEGVD computes all eigenvalues and, optionally,
44 *> eigenvectors of a complex Hermitian-definite generalized
45 *> eigenproblem using a divide and conquer algorithm.
46 *>
47 *> CHEGVX computes selected eigenvalues and, optionally,
48 *> eigenvectors of a complex Hermitian-definite generalized
49 *> eigenproblem.
50 *>
51 *> CHPGV computes all eigenvalues and, optionally,
52 *> eigenvectors of a complex Hermitian-definite generalized
53 *> eigenproblem in packed storage.
54 *>
55 *> CHPGVD computes all eigenvalues and, optionally,
56 *> eigenvectors of a complex Hermitian-definite generalized
57 *> eigenproblem in packed storage using a divide and
58 *> conquer algorithm.
59 *>
60 *> CHPGVX computes selected eigenvalues and, optionally,
61 *> eigenvectors of a complex Hermitian-definite generalized
62 *> eigenproblem in packed storage.
63 *>
64 *> CHBGV computes all eigenvalues and, optionally,
65 *> eigenvectors of a complex Hermitian-definite banded
66 *> generalized eigenproblem.
67 *>
68 *> CHBGVD computes all eigenvalues and, optionally,
69 *> eigenvectors of a complex Hermitian-definite banded
70 *> generalized eigenproblem using a divide and conquer
71 *> algorithm.
72 *>
73 *> CHBGVX computes selected eigenvalues and, optionally,
74 *> eigenvectors of a complex Hermitian-definite banded
75 *> generalized eigenproblem.
76 *>
77 *> When CDRVSG is called, a number of matrix "sizes" ("n's") and a
78 *> number of matrix "types" are specified. For each size ("n")
79 *> and each type of matrix, one matrix A of the given type will be
80 *> generated; a random well-conditioned matrix B is also generated
81 *> and the pair (A,B) is used to test the drivers.
82 *>
83 *> For each pair (A,B), the following tests are performed:
84 *>
85 *> (1) CHEGV with ITYPE = 1 and UPLO ='U':
86 *>
87 *> | A Z - B Z D | / ( |A| |Z| n ulp )
88 *>
89 *> (2) as (1) but calling CHPGV
90 *> (3) as (1) but calling CHBGV
91 *> (4) as (1) but with UPLO = 'L'
92 *> (5) as (4) but calling CHPGV
93 *> (6) as (4) but calling CHBGV
94 *>
95 *> (7) CHEGV with ITYPE = 2 and UPLO ='U':
96 *>
97 *> | A B Z - Z D | / ( |A| |Z| n ulp )
98 *>
99 *> (8) as (7) but calling CHPGV
100 *> (9) as (7) but with UPLO = 'L'
101 *> (10) as (9) but calling CHPGV
102 *>
103 *> (11) CHEGV with ITYPE = 3 and UPLO ='U':
104 *>
105 *> | B A Z - Z D | / ( |A| |Z| n ulp )
106 *>
107 *> (12) as (11) but calling CHPGV
108 *> (13) as (11) but with UPLO = 'L'
109 *> (14) as (13) but calling CHPGV
110 *>
111 *> CHEGVD, CHPGVD and CHBGVD performed the same 14 tests.
112 *>
113 *> CHEGVX, CHPGVX and CHBGVX performed the above 14 tests with
114 *> the parameter RANGE = 'A', 'N' and 'I', respectively.
115 *>
116 *> The "sizes" are specified by an array NN(1:NSIZES); the value of
117 *> each element NN(j) specifies one size.
118 *> The "types" are specified by a logical array DOTYPE( 1:NTYPES );
119 *> if DOTYPE(j) is .TRUE., then matrix type "j" will be generated.
120 *> This type is used for the matrix A which has half-bandwidth KA.
121 *> B is generated as a well-conditioned positive definite matrix
122 *> with half-bandwidth KB (<= KA).
123 *> Currently, the list of possible types for A is:
124 *>
125 *> (1) The zero matrix.
126 *> (2) The identity matrix.
127 *>
128 *> (3) A diagonal matrix with evenly spaced entries
129 *> 1, ..., ULP and random signs.
130 *> (ULP = (first number larger than 1) - 1 )
131 *> (4) A diagonal matrix with geometrically spaced entries
132 *> 1, ..., ULP and random signs.
133 *> (5) A diagonal matrix with "clustered" entries 1, ULP, ..., ULP
134 *> and random signs.
135 *>
136 *> (6) Same as (4), but multiplied by SQRT( overflow threshold )
137 *> (7) Same as (4), but multiplied by SQRT( underflow threshold )
138 *>
139 *> (8) A matrix of the form U* D U, where U is unitary and
140 *> D has evenly spaced entries 1, ..., ULP with random signs
141 *> on the diagonal.
142 *>
143 *> (9) A matrix of the form U* D U, where U is unitary and
144 *> D has geometrically spaced entries 1, ..., ULP with random
145 *> signs on the diagonal.
146 *>
147 *> (10) A matrix of the form U* D U, where U is unitary and
148 *> D has "clustered" entries 1, ULP,..., ULP with random
149 *> signs on the diagonal.
150 *>
151 *> (11) Same as (8), but multiplied by SQRT( overflow threshold )
152 *> (12) Same as (8), but multiplied by SQRT( underflow threshold )
153 *>
154 *> (13) Hermitian matrix with random entries chosen from (-1,1).
155 *> (14) Same as (13), but multiplied by SQRT( overflow threshold )
156 *> (15) Same as (13), but multiplied by SQRT( underflow threshold )
157 *>
158 *> (16) Same as (8), but with KA = 1 and KB = 1
159 *> (17) Same as (8), but with KA = 2 and KB = 1
160 *> (18) Same as (8), but with KA = 2 and KB = 2
161 *> (19) Same as (8), but with KA = 3 and KB = 1
162 *> (20) Same as (8), but with KA = 3 and KB = 2
163 *> (21) Same as (8), but with KA = 3 and KB = 3
164 *> \endverbatim
165 *
166 * Arguments:
167 * ==========
168 *
169 *> \verbatim
170 *> NSIZES INTEGER
171 *> The number of sizes of matrices to use. If it is zero,
172 *> CDRVSG does nothing. It must be at least zero.
173 *> Not modified.
174 *>
175 *> NN INTEGER array, dimension (NSIZES)
176 *> An array containing the sizes to be used for the matrices.
177 *> Zero values will be skipped. The values must be at least
178 *> zero.
179 *> Not modified.
180 *>
181 *> NTYPES INTEGER
182 *> The number of elements in DOTYPE. If it is zero, CDRVSG
183 *> does nothing. It must be at least zero. If it is MAXTYP+1
184 *> and NSIZES is 1, then an additional type, MAXTYP+1 is
185 *> defined, which is to use whatever matrix is in A. This
186 *> is only useful if DOTYPE(1:MAXTYP) is .FALSE. and
187 *> DOTYPE(MAXTYP+1) is .TRUE. .
188 *> Not modified.
189 *>
190 *> DOTYPE LOGICAL array, dimension (NTYPES)
191 *> If DOTYPE(j) is .TRUE., then for each size in NN a
192 *> matrix of that size and of type j will be generated.
193 *> If NTYPES is smaller than the maximum number of types
194 *> defined (PARAMETER MAXTYP), then types NTYPES+1 through
195 *> MAXTYP will not be generated. If NTYPES is larger
196 *> than MAXTYP, DOTYPE(MAXTYP+1) through DOTYPE(NTYPES)
197 *> will be ignored.
198 *> Not modified.
199 *>
200 *> ISEED INTEGER array, dimension (4)
201 *> On entry ISEED specifies the seed of the random number
202 *> generator. The array elements should be between 0 and 4095;
203 *> if not they will be reduced mod 4096. Also, ISEED(4) must
204 *> be odd. The random number generator uses a linear
205 *> congruential sequence limited to small integers, and so
206 *> should produce machine independent random numbers. The
207 *> values of ISEED are changed on exit, and can be used in the
208 *> next call to CDRVSG to continue the same random number
209 *> sequence.
210 *> Modified.
211 *>
212 *> THRESH REAL
213 *> A test will count as "failed" if the "error", computed as
214 *> described above, exceeds THRESH. Note that the error
215 *> is scaled to be O(1), so THRESH should be a reasonably
216 *> small multiple of 1, e.g., 10 or 100. In particular,
217 *> it should not depend on the precision (single vs. double)
218 *> or the size of the matrix. It must be at least zero.
219 *> Not modified.
220 *>
221 *> NOUNIT INTEGER
222 *> The FORTRAN unit number for printing out error messages
223 *> (e.g., if a routine returns IINFO not equal to 0.)
224 *> Not modified.
225 *>
226 *> A COMPLEX array, dimension (LDA , max(NN))
227 *> Used to hold the matrix whose eigenvalues are to be
228 *> computed. On exit, A contains the last matrix actually
229 *> used.
230 *> Modified.
231 *>
232 *> LDA INTEGER
233 *> The leading dimension of A. It must be at
234 *> least 1 and at least max( NN ).
235 *> Not modified.
236 *>
237 *> B COMPLEX array, dimension (LDB , max(NN))
238 *> Used to hold the Hermitian positive definite matrix for
239 *> the generailzed problem.
240 *> On exit, B contains the last matrix actually
241 *> used.
242 *> Modified.
243 *>
244 *> LDB INTEGER
245 *> The leading dimension of B. It must be at
246 *> least 1 and at least max( NN ).
247 *> Not modified.
248 *>
249 *> D REAL array, dimension (max(NN))
250 *> The eigenvalues of A. On exit, the eigenvalues in D
251 *> correspond with the matrix in A.
252 *> Modified.
253 *>
254 *> Z COMPLEX array, dimension (LDZ, max(NN))
255 *> The matrix of eigenvectors.
256 *> Modified.
257 *>
258 *> LDZ INTEGER
259 *> The leading dimension of ZZ. It must be at least 1 and
260 *> at least max( NN ).
261 *> Not modified.
262 *>
263 *> AB COMPLEX array, dimension (LDA, max(NN))
264 *> Workspace.
265 *> Modified.
266 *>
267 *> BB COMPLEX array, dimension (LDB, max(NN))
268 *> Workspace.
269 *> Modified.
270 *>
271 *> AP COMPLEX array, dimension (max(NN)**2)
272 *> Workspace.
273 *> Modified.
274 *>
275 *> BP COMPLEX array, dimension (max(NN)**2)
276 *> Workspace.
277 *> Modified.
278 *>
279 *> WORK COMPLEX array, dimension (NWORK)
280 *> Workspace.
281 *> Modified.
282 *>
283 *> NWORK INTEGER
284 *> The number of entries in WORK. This must be at least
285 *> 2*N + N**2 where N = max( NN(j), 2 ).
286 *> Not modified.
287 *>
288 *> RWORK REAL array, dimension (LRWORK)
289 *> Workspace.
290 *> Modified.
291 *>
292 *> LRWORK INTEGER
293 *> The number of entries in RWORK. This must be at least
294 *> max( 7*N, 1 + 4*N + 2*N*lg(N) + 3*N**2 ) where
295 *> N = max( NN(j) ) and lg( N ) = smallest integer k such
296 *> that 2**k >= N .
297 *> Not modified.
298 *>
299 *> IWORK INTEGER array, dimension (LIWORK))
300 *> Workspace.
301 *> Modified.
302 *>
303 *> LIWORK INTEGER
304 *> The number of entries in IWORK. This must be at least
305 *> 2 + 5*max( NN(j) ).
306 *> Not modified.
307 *>
308 *> RESULT REAL array, dimension (70)
309 *> The values computed by the 70 tests described above.
310 *> Modified.
311 *>
312 *> INFO INTEGER
313 *> If 0, then everything ran OK.
314 *> -1: NSIZES < 0
315 *> -2: Some NN(j) < 0
316 *> -3: NTYPES < 0
317 *> -5: THRESH < 0
318 *> -9: LDA < 1 or LDA < NMAX, where NMAX is max( NN(j) ).
319 *> -16: LDZ < 1 or LDZ < NMAX.
320 *> -21: NWORK too small.
321 *> -23: LRWORK too small.
322 *> -25: LIWORK too small.
323 *> If CLATMR, CLATMS, CHEGV, CHPGV, CHBGV, CHEGVD, CHPGVD,
324 *> CHPGVD, CHEGVX, CHPGVX, CHBGVX returns an error code,
325 *> the absolute value of it is returned.
326 *> Modified.
327 *>
328 *>-----------------------------------------------------------------------
329 *>
330 *> Some Local Variables and Parameters:
331 *> ---- ----- --------- --- ----------
332 *> ZERO, ONE Real 0 and 1.
333 *> MAXTYP The number of types defined.
334 *> NTEST The number of tests that have been run
335 *> on this matrix.
336 *> NTESTT The total number of tests for this call.
337 *> NMAX Largest value in NN.
338 *> NMATS The number of matrices generated so far.
339 *> NERRS The number of tests which have exceeded THRESH
340 *> so far (computed by SLAFTS).
341 *> COND, IMODE Values to be passed to the matrix generators.
342 *> ANORM Norm of A; passed to matrix generators.
343 *>
344 *> OVFL, UNFL Overflow and underflow thresholds.
345 *> ULP, ULPINV Finest relative precision and its inverse.
346 *> RTOVFL, RTUNFL Square roots of the previous 2 values.
347 *> The following four arrays decode JTYPE:
348 *> KTYPE(j) The general type (1-10) for type "j".
349 *> KMODE(j) The MODE value to be passed to the matrix
350 *> generator for type "j".
351 *> KMAGN(j) The order of magnitude ( O(1),
352 *> O(overflow^(1/2) ), O(underflow^(1/2) )
353 *> \endverbatim
354 *
355 * Authors:
356 * ========
357 *
358 *> \author Univ. of Tennessee
359 *> \author Univ. of California Berkeley
360 *> \author Univ. of Colorado Denver
361 *> \author NAG Ltd.
362 *
363 *> \date November 2011
364 *
365 *> \ingroup complex_eig
366 *
367 * =====================================================================
368  SUBROUTINE cdrvsg( NSIZES, NN, NTYPES, DOTYPE, ISEED, THRESH,
369  $ nounit, a, lda, b, ldb, d, z, ldz, ab, bb, ap,
370  $ bp, work, nwork, rwork, lrwork, iwork, liwork,
371  $ result, info )
372 *
373 * -- LAPACK test routine (version 3.4.0) --
374 * -- LAPACK is a software package provided by Univ. of Tennessee, --
375 * -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..--
376 * November 2011
377 *
378 * .. Scalar Arguments ..
379  INTEGER info, lda, ldb, ldz, liwork, lrwork, nounit,
380  $ nsizes, ntypes, nwork
381  REAL thresh
382 * ..
383 * .. Array Arguments ..
384  LOGICAL dotype( * )
385  INTEGER iseed( 4 ), iwork( * ), nn( * )
386  REAL d( * ), result( * ), rwork( * )
387  COMPLEX a( lda, * ), ab( lda, * ), ap( * ),
388  $ b( ldb, * ), bb( ldb, * ), bp( * ), work( * ),
389  $ z( ldz, * )
390 * ..
391 *
392 * =====================================================================
393 *
394 * .. Parameters ..
395  REAL zero, one, ten
396  parameter( zero = 0.0e+0, one = 1.0e+0, ten = 10.0e+0 )
397  COMPLEX czero, cone
398  parameter( czero = ( 0.0e+0, 0.0e+0 ),
399  $ cone = ( 1.0e+0, 0.0e+0 ) )
400  INTEGER maxtyp
401  parameter( maxtyp = 21 )
402 * ..
403 * .. Local Scalars ..
404  LOGICAL badnn
405  CHARACTER uplo
406  INTEGER i, ibtype, ibuplo, iinfo, ij, il, imode, itemp,
407  $ itype, iu, j, jcol, jsize, jtype, ka, ka9, kb,
408  $ kb9, m, mtypes, n, nerrs, nmats, nmax, ntest,
409  $ ntestt
410  REAL abstol, aninv, anorm, cond, ovfl, rtovfl,
411  $ rtunfl, ulp, ulpinv, unfl, vl, vu
412 * ..
413 * .. Local Arrays ..
414  INTEGER idumma( 1 ), ioldsd( 4 ), iseed2( 4 ),
415  $ kmagn( maxtyp ), kmode( maxtyp ),
416  $ ktype( maxtyp )
417 * ..
418 * .. External Functions ..
419  LOGICAL lsame
420  REAL slamch, slarnd
421  EXTERNAL lsame, slamch, slarnd
422 * ..
423 * .. External Subroutines ..
424  EXTERNAL chbgv, chbgvd, chbgvx, chegv, chegvd, chegvx,
427 * ..
428 * .. Intrinsic Functions ..
429  INTRINSIC abs, max, min, REAL, sqrt
430 * ..
431 * .. Data statements ..
432  DATA ktype / 1, 2, 5*4, 5*5, 3*8, 6*9 /
433  DATA kmagn / 2*1, 1, 1, 1, 2, 3, 1, 1, 1, 2, 3, 1,
434  $ 2, 3, 6*1 /
435  DATA kmode / 2*0, 4, 3, 1, 4, 4, 4, 3, 1, 4, 4, 0,
436  $ 0, 0, 6*4 /
437 * ..
438 * .. Executable Statements ..
439 *
440 * 1) Check for errors
441 *
442  ntestt = 0
443  info = 0
444 *
445  badnn = .false.
446  nmax = 0
447  DO 10 j = 1, nsizes
448  nmax = max( nmax, nn( j ) )
449  IF( nn( j ).LT.0 )
450  $ badnn = .true.
451  10 CONTINUE
452 *
453 * Check for errors
454 *
455  IF( nsizes.LT.0 ) THEN
456  info = -1
457  ELSE IF( badnn ) THEN
458  info = -2
459  ELSE IF( ntypes.LT.0 ) THEN
460  info = -3
461  ELSE IF( lda.LE.1 .OR. lda.LT.nmax ) THEN
462  info = -9
463  ELSE IF( ldz.LE.1 .OR. ldz.LT.nmax ) THEN
464  info = -16
465  ELSE IF( 2*max( nmax, 2 )**2.GT.nwork ) THEN
466  info = -21
467  ELSE IF( 2*max( nmax, 2 )**2.GT.lrwork ) THEN
468  info = -23
469  ELSE IF( 2*max( nmax, 2 )**2.GT.liwork ) THEN
470  info = -25
471  END IF
472 *
473  IF( info.NE.0 ) THEN
474  CALL xerbla( 'CDRVSG', -info )
475  RETURN
476  END IF
477 *
478 * Quick return if possible
479 *
480  IF( nsizes.EQ.0 .OR. ntypes.EQ.0 )
481  $ RETURN
482 *
483 * More Important constants
484 *
485  unfl = slamch( 'Safe minimum' )
486  ovfl = slamch( 'Overflow' )
487  CALL slabad( unfl, ovfl )
488  ulp = slamch( 'Epsilon' )*slamch( 'Base' )
489  ulpinv = one / ulp
490  rtunfl = sqrt( unfl )
491  rtovfl = sqrt( ovfl )
492 *
493  DO 20 i = 1, 4
494  iseed2( i ) = iseed( i )
495  20 CONTINUE
496 *
497 * Loop over sizes, types
498 *
499  nerrs = 0
500  nmats = 0
501 *
502  DO 650 jsize = 1, nsizes
503  n = nn( jsize )
504  aninv = one / REAL( MAX( 1, N ) )
505 *
506  IF( nsizes.NE.1 ) THEN
507  mtypes = min( maxtyp, ntypes )
508  ELSE
509  mtypes = min( maxtyp+1, ntypes )
510  END IF
511 *
512  ka9 = 0
513  kb9 = 0
514  DO 640 jtype = 1, mtypes
515  IF( .NOT.dotype( jtype ) )
516  $ go to 640
517  nmats = nmats + 1
518  ntest = 0
519 *
520  DO 30 j = 1, 4
521  ioldsd( j ) = iseed( j )
522  30 CONTINUE
523 *
524 * 2) Compute "A"
525 *
526 * Control parameters:
527 *
528 * KMAGN KMODE KTYPE
529 * =1 O(1) clustered 1 zero
530 * =2 large clustered 2 identity
531 * =3 small exponential (none)
532 * =4 arithmetic diagonal, w/ eigenvalues
533 * =5 random log hermitian, w/ eigenvalues
534 * =6 random (none)
535 * =7 random diagonal
536 * =8 random hermitian
537 * =9 banded, w/ eigenvalues
538 *
539  IF( mtypes.GT.maxtyp )
540  $ go to 90
541 *
542  itype = ktype( jtype )
543  imode = kmode( jtype )
544 *
545 * Compute norm
546 *
547  go to( 40, 50, 60 )kmagn( jtype )
548 *
549  40 CONTINUE
550  anorm = one
551  go to 70
552 *
553  50 CONTINUE
554  anorm = ( rtovfl*ulp )*aninv
555  go to 70
556 *
557  60 CONTINUE
558  anorm = rtunfl*n*ulpinv
559  go to 70
560 *
561  70 CONTINUE
562 *
563  iinfo = 0
564  cond = ulpinv
565 *
566 * Special Matrices -- Identity & Jordan block
567 *
568  IF( itype.EQ.1 ) THEN
569 *
570 * Zero
571 *
572  ka = 0
573  kb = 0
574  CALL claset( 'Full', lda, n, czero, czero, a, lda )
575 *
576  ELSE IF( itype.EQ.2 ) THEN
577 *
578 * Identity
579 *
580  ka = 0
581  kb = 0
582  CALL claset( 'Full', lda, n, czero, czero, a, lda )
583  DO 80 jcol = 1, n
584  a( jcol, jcol ) = anorm
585  80 CONTINUE
586 *
587  ELSE IF( itype.EQ.4 ) THEN
588 *
589 * Diagonal Matrix, [Eigen]values Specified
590 *
591  ka = 0
592  kb = 0
593  CALL clatms( n, n, 'S', iseed, 'H', rwork, imode, cond,
594  $ anorm, 0, 0, 'N', a, lda, work, iinfo )
595 *
596  ELSE IF( itype.EQ.5 ) THEN
597 *
598 * Hermitian, eigenvalues specified
599 *
600  ka = max( 0, n-1 )
601  kb = ka
602  CALL clatms( n, n, 'S', iseed, 'H', rwork, imode, cond,
603  $ anorm, n, n, 'N', a, lda, work, iinfo )
604 *
605  ELSE IF( itype.EQ.7 ) THEN
606 *
607 * Diagonal, random eigenvalues
608 *
609  ka = 0
610  kb = 0
611  CALL clatmr( n, n, 'S', iseed, 'H', work, 6, one, cone,
612  $ 'T', 'N', work( n+1 ), 1, one,
613  $ work( 2*n+1 ), 1, one, 'N', idumma, 0, 0,
614  $ zero, anorm, 'NO', a, lda, iwork, iinfo )
615 *
616  ELSE IF( itype.EQ.8 ) THEN
617 *
618 * Hermitian, random eigenvalues
619 *
620  ka = max( 0, n-1 )
621  kb = ka
622  CALL clatmr( n, n, 'S', iseed, 'H', work, 6, one, cone,
623  $ 'T', 'N', work( n+1 ), 1, one,
624  $ work( 2*n+1 ), 1, one, 'N', idumma, n, n,
625  $ zero, anorm, 'NO', a, lda, iwork, iinfo )
626 *
627  ELSE IF( itype.EQ.9 ) THEN
628 *
629 * Hermitian banded, eigenvalues specified
630 *
631 * The following values are used for the half-bandwidths:
632 *
633 * ka = 1 kb = 1
634 * ka = 2 kb = 1
635 * ka = 2 kb = 2
636 * ka = 3 kb = 1
637 * ka = 3 kb = 2
638 * ka = 3 kb = 3
639 *
640  kb9 = kb9 + 1
641  IF( kb9.GT.ka9 ) THEN
642  ka9 = ka9 + 1
643  kb9 = 1
644  END IF
645  ka = max( 0, min( n-1, ka9 ) )
646  kb = max( 0, min( n-1, kb9 ) )
647  CALL clatms( n, n, 'S', iseed, 'H', rwork, imode, cond,
648  $ anorm, ka, ka, 'N', a, lda, work, iinfo )
649 *
650  ELSE
651 *
652  iinfo = 1
653  END IF
654 *
655  IF( iinfo.NE.0 ) THEN
656  WRITE( nounit, fmt = 9999 )'Generator', iinfo, n, jtype,
657  $ ioldsd
658  info = abs( iinfo )
659  RETURN
660  END IF
661 *
662  90 CONTINUE
663 *
664  abstol = unfl + unfl
665  IF( n.LE.1 ) THEN
666  il = 1
667  iu = n
668  ELSE
669  il = 1 + ( n-1 )*slarnd( 1, iseed2 )
670  iu = 1 + ( n-1 )*slarnd( 1, iseed2 )
671  IF( il.GT.iu ) THEN
672  itemp = il
673  il = iu
674  iu = itemp
675  END IF
676  END IF
677 *
678 * 3) Call CHEGV, CHPGV, CHBGV, CHEGVD, CHPGVD, CHBGVD,
679 * CHEGVX, CHPGVX and CHBGVX, do tests.
680 *
681 * loop over the three generalized problems
682 * IBTYPE = 1: A*x = (lambda)*B*x
683 * IBTYPE = 2: A*B*x = (lambda)*x
684 * IBTYPE = 3: B*A*x = (lambda)*x
685 *
686  DO 630 ibtype = 1, 3
687 *
688 * loop over the setting UPLO
689 *
690  DO 620 ibuplo = 1, 2
691  IF( ibuplo.EQ.1 )
692  $ uplo = 'U'
693  IF( ibuplo.EQ.2 )
694  $ uplo = 'L'
695 *
696 * Generate random well-conditioned positive definite
697 * matrix B, of bandwidth not greater than that of A.
698 *
699  CALL clatms( n, n, 'U', iseed, 'P', rwork, 5, ten,
700  $ one, kb, kb, uplo, b, ldb, work( n+1 ),
701  $ iinfo )
702 *
703 * Test CHEGV
704 *
705  ntest = ntest + 1
706 *
707  CALL clacpy( ' ', n, n, a, lda, z, ldz )
708  CALL clacpy( uplo, n, n, b, ldb, bb, ldb )
709 *
710  CALL chegv( ibtype, 'V', uplo, n, z, ldz, bb, ldb, d,
711  $ work, nwork, rwork, iinfo )
712  IF( iinfo.NE.0 ) THEN
713  WRITE( nounit, fmt = 9999 )'CHEGV(V,' // uplo //
714  $ ')', iinfo, n, jtype, ioldsd
715  info = abs( iinfo )
716  IF( iinfo.LT.0 ) THEN
717  RETURN
718  ELSE
719  result( ntest ) = ulpinv
720  go to 100
721  END IF
722  END IF
723 *
724 * Do Test
725 *
726  CALL csgt01( ibtype, uplo, n, n, a, lda, b, ldb, z,
727  $ ldz, d, work, rwork, result( ntest ) )
728 *
729 * Test CHEGVD
730 *
731  ntest = ntest + 1
732 *
733  CALL clacpy( ' ', n, n, a, lda, z, ldz )
734  CALL clacpy( uplo, n, n, b, ldb, bb, ldb )
735 *
736  CALL chegvd( ibtype, 'V', uplo, n, z, ldz, bb, ldb, d,
737  $ work, nwork, rwork, lrwork, iwork,
738  $ liwork, iinfo )
739  IF( iinfo.NE.0 ) THEN
740  WRITE( nounit, fmt = 9999 )'CHEGVD(V,' // uplo //
741  $ ')', iinfo, n, jtype, ioldsd
742  info = abs( iinfo )
743  IF( iinfo.LT.0 ) THEN
744  RETURN
745  ELSE
746  result( ntest ) = ulpinv
747  go to 100
748  END IF
749  END IF
750 *
751 * Do Test
752 *
753  CALL csgt01( ibtype, uplo, n, n, a, lda, b, ldb, z,
754  $ ldz, d, work, rwork, result( ntest ) )
755 *
756 * Test CHEGVX
757 *
758  ntest = ntest + 1
759 *
760  CALL clacpy( ' ', n, n, a, lda, ab, lda )
761  CALL clacpy( uplo, n, n, b, ldb, bb, ldb )
762 *
763  CALL chegvx( ibtype, 'V', 'A', uplo, n, ab, lda, bb,
764  $ ldb, vl, vu, il, iu, abstol, m, d, z,
765  $ ldz, work, nwork, rwork, iwork( n+1 ),
766  $ iwork, iinfo )
767  IF( iinfo.NE.0 ) THEN
768  WRITE( nounit, fmt = 9999 )'CHEGVX(V,A' // uplo //
769  $ ')', iinfo, n, jtype, ioldsd
770  info = abs( iinfo )
771  IF( iinfo.LT.0 ) THEN
772  RETURN
773  ELSE
774  result( ntest ) = ulpinv
775  go to 100
776  END IF
777  END IF
778 *
779 * Do Test
780 *
781  CALL csgt01( ibtype, uplo, n, n, a, lda, b, ldb, z,
782  $ ldz, d, work, rwork, result( ntest ) )
783 *
784  ntest = ntest + 1
785 *
786  CALL clacpy( ' ', n, n, a, lda, ab, lda )
787  CALL clacpy( uplo, n, n, b, ldb, bb, ldb )
788 *
789 * since we do not know the exact eigenvalues of this
790 * eigenpair, we just set VL and VU as constants.
791 * It is quite possible that there are no eigenvalues
792 * in this interval.
793 *
794  vl = zero
795  vu = anorm
796  CALL chegvx( ibtype, 'V', 'V', uplo, n, ab, lda, bb,
797  $ ldb, vl, vu, il, iu, abstol, m, d, z,
798  $ ldz, work, nwork, rwork, iwork( n+1 ),
799  $ iwork, iinfo )
800  IF( iinfo.NE.0 ) THEN
801  WRITE( nounit, fmt = 9999 )'CHEGVX(V,V,' //
802  $ uplo // ')', iinfo, n, jtype, ioldsd
803  info = abs( iinfo )
804  IF( iinfo.LT.0 ) THEN
805  RETURN
806  ELSE
807  result( ntest ) = ulpinv
808  go to 100
809  END IF
810  END IF
811 *
812 * Do Test
813 *
814  CALL csgt01( ibtype, uplo, n, m, a, lda, b, ldb, z,
815  $ ldz, d, work, rwork, result( ntest ) )
816 *
817  ntest = ntest + 1
818 *
819  CALL clacpy( ' ', n, n, a, lda, ab, lda )
820  CALL clacpy( uplo, n, n, b, ldb, bb, ldb )
821 *
822  CALL chegvx( ibtype, 'V', 'I', uplo, n, ab, lda, bb,
823  $ ldb, vl, vu, il, iu, abstol, m, d, z,
824  $ ldz, work, nwork, rwork, iwork( n+1 ),
825  $ iwork, iinfo )
826  IF( iinfo.NE.0 ) THEN
827  WRITE( nounit, fmt = 9999 )'CHEGVX(V,I,' //
828  $ uplo // ')', iinfo, n, jtype, ioldsd
829  info = abs( iinfo )
830  IF( iinfo.LT.0 ) THEN
831  RETURN
832  ELSE
833  result( ntest ) = ulpinv
834  go to 100
835  END IF
836  END IF
837 *
838 * Do Test
839 *
840  CALL csgt01( ibtype, uplo, n, m, a, lda, b, ldb, z,
841  $ ldz, d, work, rwork, result( ntest ) )
842 *
843  100 CONTINUE
844 *
845 * Test CHPGV
846 *
847  ntest = ntest + 1
848 *
849 * Copy the matrices into packed storage.
850 *
851  IF( lsame( uplo, 'U' ) ) THEN
852  ij = 1
853  DO 120 j = 1, n
854  DO 110 i = 1, j
855  ap( ij ) = a( i, j )
856  bp( ij ) = b( i, j )
857  ij = ij + 1
858  110 CONTINUE
859  120 CONTINUE
860  ELSE
861  ij = 1
862  DO 140 j = 1, n
863  DO 130 i = j, n
864  ap( ij ) = a( i, j )
865  bp( ij ) = b( i, j )
866  ij = ij + 1
867  130 CONTINUE
868  140 CONTINUE
869  END IF
870 *
871  CALL chpgv( ibtype, 'V', uplo, n, ap, bp, d, z, ldz,
872  $ work, rwork, iinfo )
873  IF( iinfo.NE.0 ) THEN
874  WRITE( nounit, fmt = 9999 )'CHPGV(V,' // uplo //
875  $ ')', iinfo, n, jtype, ioldsd
876  info = abs( iinfo )
877  IF( iinfo.LT.0 ) THEN
878  RETURN
879  ELSE
880  result( ntest ) = ulpinv
881  go to 310
882  END IF
883  END IF
884 *
885 * Do Test
886 *
887  CALL csgt01( ibtype, uplo, n, n, a, lda, b, ldb, z,
888  $ ldz, d, work, rwork, result( ntest ) )
889 *
890 * Test CHPGVD
891 *
892  ntest = ntest + 1
893 *
894 * Copy the matrices into packed storage.
895 *
896  IF( lsame( uplo, 'U' ) ) THEN
897  ij = 1
898  DO 160 j = 1, n
899  DO 150 i = 1, j
900  ap( ij ) = a( i, j )
901  bp( ij ) = b( i, j )
902  ij = ij + 1
903  150 CONTINUE
904  160 CONTINUE
905  ELSE
906  ij = 1
907  DO 180 j = 1, n
908  DO 170 i = j, n
909  ap( ij ) = a( i, j )
910  bp( ij ) = b( i, j )
911  ij = ij + 1
912  170 CONTINUE
913  180 CONTINUE
914  END IF
915 *
916  CALL chpgvd( ibtype, 'V', uplo, n, ap, bp, d, z, ldz,
917  $ work, nwork, rwork, lrwork, iwork,
918  $ liwork, iinfo )
919  IF( iinfo.NE.0 ) THEN
920  WRITE( nounit, fmt = 9999 )'CHPGVD(V,' // uplo //
921  $ ')', iinfo, n, jtype, ioldsd
922  info = abs( iinfo )
923  IF( iinfo.LT.0 ) THEN
924  RETURN
925  ELSE
926  result( ntest ) = ulpinv
927  go to 310
928  END IF
929  END IF
930 *
931 * Do Test
932 *
933  CALL csgt01( ibtype, uplo, n, n, a, lda, b, ldb, z,
934  $ ldz, d, work, rwork, result( ntest ) )
935 *
936 * Test CHPGVX
937 *
938  ntest = ntest + 1
939 *
940 * Copy the matrices into packed storage.
941 *
942  IF( lsame( uplo, 'U' ) ) THEN
943  ij = 1
944  DO 200 j = 1, n
945  DO 190 i = 1, j
946  ap( ij ) = a( i, j )
947  bp( ij ) = b( i, j )
948  ij = ij + 1
949  190 CONTINUE
950  200 CONTINUE
951  ELSE
952  ij = 1
953  DO 220 j = 1, n
954  DO 210 i = j, n
955  ap( ij ) = a( i, j )
956  bp( ij ) = b( i, j )
957  ij = ij + 1
958  210 CONTINUE
959  220 CONTINUE
960  END IF
961 *
962  CALL chpgvx( ibtype, 'V', 'A', uplo, n, ap, bp, vl,
963  $ vu, il, iu, abstol, m, d, z, ldz, work,
964  $ rwork, iwork( n+1 ), iwork, info )
965  IF( iinfo.NE.0 ) THEN
966  WRITE( nounit, fmt = 9999 )'CHPGVX(V,A' // uplo //
967  $ ')', iinfo, n, jtype, ioldsd
968  info = abs( iinfo )
969  IF( iinfo.LT.0 ) THEN
970  RETURN
971  ELSE
972  result( ntest ) = ulpinv
973  go to 310
974  END IF
975  END IF
976 *
977 * Do Test
978 *
979  CALL csgt01( ibtype, uplo, n, n, a, lda, b, ldb, z,
980  $ ldz, d, work, rwork, result( ntest ) )
981 *
982  ntest = ntest + 1
983 *
984 * Copy the matrices into packed storage.
985 *
986  IF( lsame( uplo, 'U' ) ) THEN
987  ij = 1
988  DO 240 j = 1, n
989  DO 230 i = 1, j
990  ap( ij ) = a( i, j )
991  bp( ij ) = b( i, j )
992  ij = ij + 1
993  230 CONTINUE
994  240 CONTINUE
995  ELSE
996  ij = 1
997  DO 260 j = 1, n
998  DO 250 i = j, n
999  ap( ij ) = a( i, j )
1000  bp( ij ) = b( i, j )
1001  ij = ij + 1
1002  250 CONTINUE
1003  260 CONTINUE
1004  END IF
1005 *
1006  vl = zero
1007  vu = anorm
1008  CALL chpgvx( ibtype, 'V', 'V', uplo, n, ap, bp, vl,
1009  $ vu, il, iu, abstol, m, d, z, ldz, work,
1010  $ rwork, iwork( n+1 ), iwork, info )
1011  IF( iinfo.NE.0 ) THEN
1012  WRITE( nounit, fmt = 9999 )'CHPGVX(V,V' // uplo //
1013  $ ')', iinfo, n, jtype, ioldsd
1014  info = abs( iinfo )
1015  IF( iinfo.LT.0 ) THEN
1016  RETURN
1017  ELSE
1018  result( ntest ) = ulpinv
1019  go to 310
1020  END IF
1021  END IF
1022 *
1023 * Do Test
1024 *
1025  CALL csgt01( ibtype, uplo, n, m, a, lda, b, ldb, z,
1026  $ ldz, d, work, rwork, result( ntest ) )
1027 *
1028  ntest = ntest + 1
1029 *
1030 * Copy the matrices into packed storage.
1031 *
1032  IF( lsame( uplo, 'U' ) ) THEN
1033  ij = 1
1034  DO 280 j = 1, n
1035  DO 270 i = 1, j
1036  ap( ij ) = a( i, j )
1037  bp( ij ) = b( i, j )
1038  ij = ij + 1
1039  270 CONTINUE
1040  280 CONTINUE
1041  ELSE
1042  ij = 1
1043  DO 300 j = 1, n
1044  DO 290 i = j, n
1045  ap( ij ) = a( i, j )
1046  bp( ij ) = b( i, j )
1047  ij = ij + 1
1048  290 CONTINUE
1049  300 CONTINUE
1050  END IF
1051 *
1052  CALL chpgvx( ibtype, 'V', 'I', uplo, n, ap, bp, vl,
1053  $ vu, il, iu, abstol, m, d, z, ldz, work,
1054  $ rwork, iwork( n+1 ), iwork, info )
1055  IF( iinfo.NE.0 ) THEN
1056  WRITE( nounit, fmt = 9999 )'CHPGVX(V,I' // uplo //
1057  $ ')', iinfo, n, jtype, ioldsd
1058  info = abs( iinfo )
1059  IF( iinfo.LT.0 ) THEN
1060  RETURN
1061  ELSE
1062  result( ntest ) = ulpinv
1063  go to 310
1064  END IF
1065  END IF
1066 *
1067 * Do Test
1068 *
1069  CALL csgt01( ibtype, uplo, n, m, a, lda, b, ldb, z,
1070  $ ldz, d, work, rwork, result( ntest ) )
1071 *
1072  310 CONTINUE
1073 *
1074  IF( ibtype.EQ.1 ) THEN
1075 *
1076 * TEST CHBGV
1077 *
1078  ntest = ntest + 1
1079 *
1080 * Copy the matrices into band storage.
1081 *
1082  IF( lsame( uplo, 'U' ) ) THEN
1083  DO 340 j = 1, n
1084  DO 320 i = max( 1, j-ka ), j
1085  ab( ka+1+i-j, j ) = a( i, j )
1086  320 CONTINUE
1087  DO 330 i = max( 1, j-kb ), j
1088  bb( kb+1+i-j, j ) = b( i, j )
1089  330 CONTINUE
1090  340 CONTINUE
1091  ELSE
1092  DO 370 j = 1, n
1093  DO 350 i = j, min( n, j+ka )
1094  ab( 1+i-j, j ) = a( i, j )
1095  350 CONTINUE
1096  DO 360 i = j, min( n, j+kb )
1097  bb( 1+i-j, j ) = b( i, j )
1098  360 CONTINUE
1099  370 CONTINUE
1100  END IF
1101 *
1102  CALL chbgv( 'V', uplo, n, ka, kb, ab, lda, bb, ldb,
1103  $ d, z, ldz, work, rwork, iinfo )
1104  IF( iinfo.NE.0 ) THEN
1105  WRITE( nounit, fmt = 9999 )'CHBGV(V,' //
1106  $ uplo // ')', iinfo, n, jtype, ioldsd
1107  info = abs( iinfo )
1108  IF( iinfo.LT.0 ) THEN
1109  RETURN
1110  ELSE
1111  result( ntest ) = ulpinv
1112  go to 620
1113  END IF
1114  END IF
1115 *
1116 * Do Test
1117 *
1118  CALL csgt01( ibtype, uplo, n, n, a, lda, b, ldb, z,
1119  $ ldz, d, work, rwork, result( ntest ) )
1120 *
1121 * TEST CHBGVD
1122 *
1123  ntest = ntest + 1
1124 *
1125 * Copy the matrices into band storage.
1126 *
1127  IF( lsame( uplo, 'U' ) ) THEN
1128  DO 400 j = 1, n
1129  DO 380 i = max( 1, j-ka ), j
1130  ab( ka+1+i-j, j ) = a( i, j )
1131  380 CONTINUE
1132  DO 390 i = max( 1, j-kb ), j
1133  bb( kb+1+i-j, j ) = b( i, j )
1134  390 CONTINUE
1135  400 CONTINUE
1136  ELSE
1137  DO 430 j = 1, n
1138  DO 410 i = j, min( n, j+ka )
1139  ab( 1+i-j, j ) = a( i, j )
1140  410 CONTINUE
1141  DO 420 i = j, min( n, j+kb )
1142  bb( 1+i-j, j ) = b( i, j )
1143  420 CONTINUE
1144  430 CONTINUE
1145  END IF
1146 *
1147  CALL chbgvd( 'V', uplo, n, ka, kb, ab, lda, bb,
1148  $ ldb, d, z, ldz, work, nwork, rwork,
1149  $ lrwork, iwork, liwork, iinfo )
1150  IF( iinfo.NE.0 ) THEN
1151  WRITE( nounit, fmt = 9999 )'CHBGVD(V,' //
1152  $ uplo // ')', iinfo, n, jtype, ioldsd
1153  info = abs( iinfo )
1154  IF( iinfo.LT.0 ) THEN
1155  RETURN
1156  ELSE
1157  result( ntest ) = ulpinv
1158  go to 620
1159  END IF
1160  END IF
1161 *
1162 * Do Test
1163 *
1164  CALL csgt01( ibtype, uplo, n, n, a, lda, b, ldb, z,
1165  $ ldz, d, work, rwork, result( ntest ) )
1166 *
1167 * Test CHBGVX
1168 *
1169  ntest = ntest + 1
1170 *
1171 * Copy the matrices into band storage.
1172 *
1173  IF( lsame( uplo, 'U' ) ) THEN
1174  DO 460 j = 1, n
1175  DO 440 i = max( 1, j-ka ), j
1176  ab( ka+1+i-j, j ) = a( i, j )
1177  440 CONTINUE
1178  DO 450 i = max( 1, j-kb ), j
1179  bb( kb+1+i-j, j ) = b( i, j )
1180  450 CONTINUE
1181  460 CONTINUE
1182  ELSE
1183  DO 490 j = 1, n
1184  DO 470 i = j, min( n, j+ka )
1185  ab( 1+i-j, j ) = a( i, j )
1186  470 CONTINUE
1187  DO 480 i = j, min( n, j+kb )
1188  bb( 1+i-j, j ) = b( i, j )
1189  480 CONTINUE
1190  490 CONTINUE
1191  END IF
1192 *
1193  CALL chbgvx( 'V', 'A', uplo, n, ka, kb, ab, lda,
1194  $ bb, ldb, bp, max( 1, n ), vl, vu, il,
1195  $ iu, abstol, m, d, z, ldz, work, rwork,
1196  $ iwork( n+1 ), iwork, iinfo )
1197  IF( iinfo.NE.0 ) THEN
1198  WRITE( nounit, fmt = 9999 )'CHBGVX(V,A' //
1199  $ uplo // ')', iinfo, n, jtype, ioldsd
1200  info = abs( iinfo )
1201  IF( iinfo.LT.0 ) THEN
1202  RETURN
1203  ELSE
1204  result( ntest ) = ulpinv
1205  go to 620
1206  END IF
1207  END IF
1208 *
1209 * Do Test
1210 *
1211  CALL csgt01( ibtype, uplo, n, n, a, lda, b, ldb, z,
1212  $ ldz, d, work, rwork, result( ntest ) )
1213 *
1214  ntest = ntest + 1
1215 *
1216 * Copy the matrices into band storage.
1217 *
1218  IF( lsame( uplo, 'U' ) ) THEN
1219  DO 520 j = 1, n
1220  DO 500 i = max( 1, j-ka ), j
1221  ab( ka+1+i-j, j ) = a( i, j )
1222  500 CONTINUE
1223  DO 510 i = max( 1, j-kb ), j
1224  bb( kb+1+i-j, j ) = b( i, j )
1225  510 CONTINUE
1226  520 CONTINUE
1227  ELSE
1228  DO 550 j = 1, n
1229  DO 530 i = j, min( n, j+ka )
1230  ab( 1+i-j, j ) = a( i, j )
1231  530 CONTINUE
1232  DO 540 i = j, min( n, j+kb )
1233  bb( 1+i-j, j ) = b( i, j )
1234  540 CONTINUE
1235  550 CONTINUE
1236  END IF
1237 *
1238  vl = zero
1239  vu = anorm
1240  CALL chbgvx( 'V', 'V', uplo, n, ka, kb, ab, lda,
1241  $ bb, ldb, bp, max( 1, n ), vl, vu, il,
1242  $ iu, abstol, m, d, z, ldz, work, rwork,
1243  $ iwork( n+1 ), iwork, iinfo )
1244  IF( iinfo.NE.0 ) THEN
1245  WRITE( nounit, fmt = 9999 )'CHBGVX(V,V' //
1246  $ uplo // ')', iinfo, n, jtype, ioldsd
1247  info = abs( iinfo )
1248  IF( iinfo.LT.0 ) THEN
1249  RETURN
1250  ELSE
1251  result( ntest ) = ulpinv
1252  go to 620
1253  END IF
1254  END IF
1255 *
1256 * Do Test
1257 *
1258  CALL csgt01( ibtype, uplo, n, m, a, lda, b, ldb, z,
1259  $ ldz, d, work, rwork, result( ntest ) )
1260 *
1261  ntest = ntest + 1
1262 *
1263 * Copy the matrices into band storage.
1264 *
1265  IF( lsame( uplo, 'U' ) ) THEN
1266  DO 580 j = 1, n
1267  DO 560 i = max( 1, j-ka ), j
1268  ab( ka+1+i-j, j ) = a( i, j )
1269  560 CONTINUE
1270  DO 570 i = max( 1, j-kb ), j
1271  bb( kb+1+i-j, j ) = b( i, j )
1272  570 CONTINUE
1273  580 CONTINUE
1274  ELSE
1275  DO 610 j = 1, n
1276  DO 590 i = j, min( n, j+ka )
1277  ab( 1+i-j, j ) = a( i, j )
1278  590 CONTINUE
1279  DO 600 i = j, min( n, j+kb )
1280  bb( 1+i-j, j ) = b( i, j )
1281  600 CONTINUE
1282  610 CONTINUE
1283  END IF
1284 *
1285  CALL chbgvx( 'V', 'I', uplo, n, ka, kb, ab, lda,
1286  $ bb, ldb, bp, max( 1, n ), vl, vu, il,
1287  $ iu, abstol, m, d, z, ldz, work, rwork,
1288  $ iwork( n+1 ), iwork, iinfo )
1289  IF( iinfo.NE.0 ) THEN
1290  WRITE( nounit, fmt = 9999 )'CHBGVX(V,I' //
1291  $ uplo // ')', iinfo, n, jtype, ioldsd
1292  info = abs( iinfo )
1293  IF( iinfo.LT.0 ) THEN
1294  RETURN
1295  ELSE
1296  result( ntest ) = ulpinv
1297  go to 620
1298  END IF
1299  END IF
1300 *
1301 * Do Test
1302 *
1303  CALL csgt01( ibtype, uplo, n, m, a, lda, b, ldb, z,
1304  $ ldz, d, work, rwork, result( ntest ) )
1305 *
1306  END IF
1307 *
1308  620 CONTINUE
1309  630 CONTINUE
1310 *
1311 * End of Loop -- Check for RESULT(j) > THRESH
1312 *
1313  ntestt = ntestt + ntest
1314  CALL slafts( 'CSG', n, n, jtype, ntest, result, ioldsd,
1315  $ thresh, nounit, nerrs )
1316  640 CONTINUE
1317  650 CONTINUE
1318 *
1319 * Summary
1320 *
1321  CALL slasum( 'CSG', nounit, nerrs, ntestt )
1322 *
1323  RETURN
1324 *
1325  9999 FORMAT( ' CDRVSG: ', a, ' returned INFO=', i6, '.', / 9x, 'N=',
1326  $ i6, ', JTYPE=', i6, ', ISEED=(', 3( i5, ',' ), i5, ')' )
1327 *
1328 * End of CDRVSG
1329 *
1330  END