LAPACK  3.8.0 LAPACK: Linear Algebra PACKage

## ◆ dchksy_aa_2stage()

 subroutine dchksy_aa_2stage ( logical, dimension( * ) DOTYPE, integer NN, integer, dimension( * ) NVAL, integer NNB, integer, dimension( * ) NBVAL, integer NNS, integer, dimension( * ) NSVAL, double precision THRESH, logical TSTERR, integer NMAX, double precision, dimension( * ) A, double precision, dimension( * ) AFAC, double precision, dimension( * ) AINV, double precision, dimension( * ) B, double precision, dimension( * ) X, double precision, dimension( * ) XACT, double precision, dimension( * ) WORK, double precision, dimension( * ) RWORK, integer, dimension( * ) IWORK, integer NOUT )

DCHKSY_AA_2STAGE

Purpose:
` DCHKSY_AA_2STAGE tests DSYTRF_AA_2STAGE, -TRS_AA_2STAGE.`
Parameters
 [in] DOTYPE ``` DOTYPE is LOGICAL array, dimension (NTYPES) The matrix types to be used for testing. Matrices of type j (for 1 <= j <= NTYPES) are used for testing if DOTYPE(j) = .TRUE.; if DOTYPE(j) = .FALSE., then type j is not used.``` [in] NN ``` NN is INTEGER The number of values of N contained in the vector NVAL.``` [in] NVAL ``` NVAL is INTEGER array, dimension (NN) The values of the matrix dimension N.``` [in] NNB ``` NNB is INTEGER The number of values of NB contained in the vector NBVAL.``` [in] NBVAL ``` NBVAL is INTEGER array, dimension (NBVAL) The values of the blocksize NB.``` [in] NNS ``` NNS is INTEGER The number of values of NRHS contained in the vector NSVAL.``` [in] NSVAL ``` NSVAL is INTEGER array, dimension (NNS) The values of the number of right hand sides NRHS.``` [in] THRESH ``` THRESH is DOUBLE PRECISION The threshold value for the test ratios. A result is included in the output file if RESULT >= THRESH. To have every test ratio printed, use THRESH = 0.``` [in] TSTERR ``` TSTERR is LOGICAL Flag that indicates whether error exits are to be tested.``` [in] NMAX ``` NMAX is INTEGER The maximum value permitted for N, used in dimensioning the work arrays.``` [out] A ` A is DOUBLE PRECISION array, dimension (NMAX*NMAX)` [out] AFAC ` AFAC is DOUBLE PRECISION array, dimension (NMAX*NMAX)` [out] AINV ` AINV is DOUBLE PRECISION array, dimension (NMAX*NMAX)` [out] B ``` B is DOUBLE PRECISION array, dimension (NMAX*NSMAX) where NSMAX is the largest entry in NSVAL.``` [out] X ` X is DOUBLE PRECISION array, dimension (NMAX*NSMAX)` [out] XACT ` XACT is DOUBLE PRECISION array, dimension (NMAX*NSMAX)` [out] WORK ` WORK is DOUBLE PRECISION array, dimension (NMAX*max(3,NSMAX))` [out] RWORK ` RWORK is DOUBLE PRECISION array, dimension (max(NMAX,2*NSMAX))` [out] IWORK ` IWORK is INTEGER array, dimension (2*NMAX)` [in] NOUT ``` NOUT is INTEGER The unit number for output.```
Date
November 2017

Definition at line 175 of file dchksy_aa_2stage.f.

175 *
176 * -- LAPACK test routine (version 3.8.0) --
177 * -- LAPACK is a software package provided by Univ. of Tennessee, --
178 * -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..--
179 * November 2017
180 *
181  IMPLICIT NONE
182 *
183 * .. Scalar Arguments ..
184  LOGICAL tsterr
185  INTEGER nn, nnb, nns, nmax, nout
186  DOUBLE PRECISION thresh
187 * ..
188 * .. Array Arguments ..
189  LOGICAL dotype( * )
190  INTEGER iwork( * ), nbval( * ), nsval( * ), nval( * )
191  DOUBLE PRECISION a( * ), afac( * ), ainv( * ), b( * ),
192  \$ rwork( * ), work( * ), x( * ), xact( * )
193 * ..
194 *
195 * =====================================================================
196 *
197 * .. Parameters ..
198  DOUBLE PRECISION zero, one
199  parameter( zero = 0.0d+0, one = 1.0d+0 )
200  INTEGER ntypes
201  parameter( ntypes = 10 )
202  INTEGER ntests
203  parameter( ntests = 9 )
204 * ..
205 * .. Local Scalars ..
206  LOGICAL zerot
207  CHARACTER dist, TYPE, uplo, xtype
208  CHARACTER*3 path, matpath
209  INTEGER i, i1, i2, imat, in, inb, info, ioff, irhs,
210  \$ iuplo, izero, j, k, kl, ku, lda, lwork, mode,
211  \$ n, nb, nerrs, nfail, nimat, nrhs, nrun, nt
212  DOUBLE PRECISION anorm, cndnum
213 * ..
214 * .. Local Arrays ..
215  CHARACTER uplos( 2 )
216  INTEGER iseed( 4 ), iseedy( 4 )
217  DOUBLE PRECISION result( ntests )
218 * ..
219 * .. External Subroutines ..
220  EXTERNAL alaerh, alahd, alasum, derrsy, dlacpy, dlarhs,
223 * ..
224 * .. Intrinsic Functions ..
225  INTRINSIC max, min
226 * ..
227 * .. Scalars in Common ..
228  LOGICAL lerr, ok
229  CHARACTER*32 srnamt
230  INTEGER infot, nunit
231 * ..
232 * .. Common blocks ..
233  COMMON / infoc / infot, nunit, ok, lerr
234  COMMON / srnamc / srnamt
235 * ..
236 * .. Data statements ..
237  DATA iseedy / 1988, 1989, 1990, 1991 /
238  DATA uplos / 'U', 'L' /
239 * ..
240 * .. Executable Statements ..
241 *
242 * Initialize constants and the random number seed.
243 *
244 * Test path
245 *
246  path( 1: 1 ) = 'Double precision'
247  path( 2: 3 ) = 'S2'
248 *
249 * Path to generate matrices
250 *
251  matpath( 1: 1 ) = 'Double precision'
252  matpath( 2: 3 ) = 'SY'
253  nrun = 0
254  nfail = 0
255  nerrs = 0
256  DO 10 i = 1, 4
257  iseed( i ) = iseedy( i )
258  10 CONTINUE
259 *
260 * Test the error exits
261 *
262  IF( tsterr )
263  \$ CALL derrsy( path, nout )
264  infot = 0
265 *
266 * Set the minimum block size for which the block routine should
267 * be used, which will be later returned by ILAENV
268 *
269  CALL xlaenv( 2, 2 )
270 *
271 * Do for each value of N in NVAL
272 *
273  DO 180 in = 1, nn
274  n = nval( in )
275  IF( n .GT. nmax ) THEN
276  nfail = nfail + 1
277  WRITE(nout, 9995) 'M ', n, nmax
278  GO TO 180
279  END IF
280  lda = max( n, 1 )
281  xtype = 'N'
282  nimat = ntypes
283  IF( n.LE.0 )
284  \$ nimat = 1
285 *
286  izero = 0
287 *
288 * Do for each value of matrix type IMAT
289 *
290  DO 170 imat = 1, nimat
291 *
292 * Do the tests only if DOTYPE( IMAT ) is true.
293 *
294  IF( .NOT.dotype( imat ) )
295  \$ GO TO 170
296 *
297 * Skip types 3, 4, 5, or 6 if the matrix size is too small.
298 *
299  zerot = imat.GE.3 .AND. imat.LE.6
300  IF( zerot .AND. n.LT.imat-2 )
301  \$ GO TO 170
302 *
303 * Do first for UPLO = 'U', then for UPLO = 'L'
304 *
305  DO 160 iuplo = 1, 2
306  uplo = uplos( iuplo )
307 *
308 * Begin generate the test matrix A.
309 *
310 *
311 * Set up parameters with DLATB4 for the matrix generator
312 * based on the type of matrix to be generated.
313 *
314  CALL dlatb4( matpath, imat, n, n, TYPE, kl, ku,
315  \$ anorm, mode, cndnum, dist )
316 *
317 * Generate a matrix with DLATMS.
318 *
319  srnamt = 'DLATMS'
320  CALL dlatms( n, n, dist, iseed, TYPE, rwork, mode,
321  \$ cndnum, anorm, kl, ku, uplo, a, lda, work,
322  \$ info )
323 *
324 * Check error code from DLATMS and handle error.
325 *
326  IF( info.NE.0 ) THEN
327  CALL alaerh( path, 'DLATMS', info, 0, uplo, n, n, -1,
328  \$ -1, -1, imat, nfail, nerrs, nout )
329 *
330 * Skip all tests for this generated matrix
331 *
332  GO TO 160
333  END IF
334 *
335 * For matrix types 3-6, zero one or more rows and
336 * columns of the matrix to test that INFO is returned
337 * correctly.
338 *
339  IF( zerot ) THEN
340  IF( imat.EQ.3 ) THEN
341  izero = 1
342  ELSE IF( imat.EQ.4 ) THEN
343  izero = n
344  ELSE
345  izero = n / 2 + 1
346  END IF
347 *
348  IF( imat.LT.6 ) THEN
349 *
350 * Set row and column IZERO to zero.
351 *
352  IF( iuplo.EQ.1 ) THEN
353  ioff = ( izero-1 )*lda
354  DO 20 i = 1, izero - 1
355  a( ioff+i ) = zero
356  20 CONTINUE
357  ioff = ioff + izero
358  DO 30 i = izero, n
359  a( ioff ) = zero
360  ioff = ioff + lda
361  30 CONTINUE
362  ELSE
363  ioff = izero
364  DO 40 i = 1, izero - 1
365  a( ioff ) = zero
366  ioff = ioff + lda
367  40 CONTINUE
368  ioff = ioff - izero
369  DO 50 i = izero, n
370  a( ioff+i ) = zero
371  50 CONTINUE
372  END IF
373  ELSE
374  IF( iuplo.EQ.1 ) THEN
375 *
376 * Set the first IZERO rows and columns to zero.
377 *
378  ioff = 0
379  DO 70 j = 1, n
380  i2 = min( j, izero )
381  DO 60 i = 1, i2
382  a( ioff+i ) = zero
383  60 CONTINUE
384  ioff = ioff + lda
385  70 CONTINUE
386  izero = 1
387  ELSE
388 *
389 * Set the last IZERO rows and columns to zero.
390 *
391  ioff = 0
392  DO 90 j = 1, n
393  i1 = max( j, izero )
394  DO 80 i = i1, n
395  a( ioff+i ) = zero
396  80 CONTINUE
397  ioff = ioff + lda
398  90 CONTINUE
399  END IF
400  END IF
401  ELSE
402  izero = 0
403  END IF
404 *
405 * End generate the test matrix A.
406 *
407 * Do for each value of NB in NBVAL
408 *
409  DO 150 inb = 1, nnb
410 *
411 * Set the optimal blocksize, which will be later
412 * returned by ILAENV.
413 *
414  nb = nbval( inb )
415  CALL xlaenv( 1, nb )
416 *
417 * Copy the test matrix A into matrix AFAC which
418 * will be factorized in place. This is needed to
419 * preserve the test matrix A for subsequent tests.
420 *
421  CALL dlacpy( uplo, n, n, a, lda, afac, lda )
422 *
423 * Compute the L*D*L**T or U*D*U**T factorization of the
424 * matrix. IWORK stores details of the interchanges and
425 * the block structure of D. AINV is a work array for
426 * block factorization, LWORK is the length of AINV.
427 *
428  srnamt = 'DSYTRF_AA_2STAGE'
429  lwork = min(n*nb, 3*nmax*nmax)
430  CALL dsytrf_aa_2stage( uplo, n, afac, lda,
431  \$ ainv, (3*nb+1)*n,
432  \$ iwork, iwork( 1+n ),
433  \$ work, lwork,
434  \$ info )
435 *
436 * Adjust the expected value of INFO to account for
437 * pivoting.
438 *
439  IF( izero.GT.0 ) THEN
440  j = 1
441  k = izero
442  100 CONTINUE
443  IF( j.EQ.k ) THEN
444  k = iwork( j )
445  ELSE IF( iwork( j ).EQ.k ) THEN
446  k = j
447  END IF
448  IF( j.LT.k ) THEN
449  j = j + 1
450  GO TO 100
451  END IF
452  ELSE
453  k = 0
454  END IF
455 *
456 * Check error code from DSYTRF and handle error.
457 *
458  IF( info.NE.k ) THEN
459  CALL alaerh( path, 'DSYTRF_AA_2STAGE', info, k,
460  \$ uplo, n, n, -1, -1, nb, imat, nfail,
461  \$ nerrs, nout )
462  END IF
463 *
464 *+ TEST 1
465 * Reconstruct matrix from factors and compute residual.
466 *
467 c CALL DSYT01_AA( UPLO, N, A, LDA, AFAC, LDA, IWORK,
468 c \$ AINV, LDA, RWORK, RESULT( 1 ) )
469 c NT = 1
470  nt = 0
471 *
472 *
473 * Print information about the tests that did not pass
474 * the threshold.
475 *
476  DO 110 k = 1, nt
477  IF( result( k ).GE.thresh ) THEN
478  IF( nfail.EQ.0 .AND. nerrs.EQ.0 )
479  \$ CALL alahd( nout, path )
480  WRITE( nout, fmt = 9999 )uplo, n, nb, imat, k,
481  \$ result( k )
482  nfail = nfail + 1
483  END IF
484  110 CONTINUE
485  nrun = nrun + nt
486 *
487 * Skip solver test if INFO is not 0.
488 *
489  IF( info.NE.0 ) THEN
490  GO TO 140
491  END IF
492 *
493 * Do for each value of NRHS in NSVAL.
494 *
495  DO 130 irhs = 1, nns
496  nrhs = nsval( irhs )
497 *
498 *+ TEST 2 (Using TRS)
499 * Solve and compute residual for A * X = B.
500 *
501 * Choose a set of NRHS random solution vectors
502 * stored in XACT and set up the right hand side B
503 *
504  srnamt = 'DLARHS'
505  CALL dlarhs( matpath, xtype, uplo, ' ', n, n,
506  \$ kl, ku, nrhs, a, lda, xact, lda,
507  \$ b, lda, iseed, info )
508  CALL dlacpy( 'Full', n, nrhs, b, lda, x, lda )
509 *
510  srnamt = 'DSYTRS_AA_2STAGE'
511  lwork = max( 1, 3*n-2 )
512  CALL dsytrs_aa_2stage( uplo, n, nrhs, afac, lda,
513  \$ ainv, (3*nb+1)*n, iwork, iwork( 1+n ),
514  \$ x, lda, info )
515 *
516 * Check error code from DSYTRS and handle error.
517 *
518  IF( info.NE.0 ) THEN
519  IF( izero.EQ.0 ) THEN
520  CALL alaerh( path, 'DSYTRS_AA_2STAGE',
521  \$ info, 0, uplo, n, n, -1, -1,
522  \$ nrhs, imat, nfail, nerrs, nout )
523  END IF
524  ELSE
525  CALL dlacpy( 'Full', n, nrhs, b, lda, work, lda
526  \$ )
527 *
528 * Compute the residual for the solution
529 *
530  CALL dpot02( uplo, n, nrhs, a, lda, x, lda,
531  \$ work, lda, rwork, result( 2 ) )
532 *
533 *
534 * Print information about the tests that did not pass
535 * the threshold.
536 *
537  DO 120 k = 2, 2
538  IF( result( k ).GE.thresh ) THEN
539  IF( nfail.EQ.0 .AND. nerrs.EQ.0 )
540  \$ CALL alahd( nout, path )
541  WRITE( nout, fmt = 9998 )uplo, n, nrhs,
542  \$ imat, k, result( k )
543  nfail = nfail + 1
544  END IF
545  120 CONTINUE
546  END IF
547  nrun = nrun + 1
548 *
549 * End do for each value of NRHS in NSVAL.
550 *
551  130 CONTINUE
552  140 CONTINUE
553  150 CONTINUE
554  160 CONTINUE
555  170 CONTINUE
556  180 CONTINUE
557 *
558 * Print a summary of the results.
559 *
560  CALL alasum( path, nout, nfail, nrun, nerrs )
561 *
562  9999 FORMAT( ' UPLO = ''', a1, ''', N =', i5, ', NB =', i4, ', type ',
563  \$ i2, ', test ', i2, ', ratio =', g12.5 )
564  9998 FORMAT( ' UPLO = ''', a1, ''', N =', i5, ', NRHS=', i3, ', type ',
565  \$ i2, ', test(', i2, ') =', g12.5 )
566  9995 FORMAT( ' Invalid input value: ', a4, '=', i6, '; must be <=',
567  \$ i6 )
568  RETURN
569 *
570 * End of DCHKSY_AA_2STAGE
571 *
subroutine dlatms(M, N, DIST, ISEED, SYM, D, MODE, COND, DMAX, KL, KU, PACK, A, LDA, WORK, INFO)
DLATMS
Definition: dlatms.f:323
subroutine xlaenv(ISPEC, NVALUE)
XLAENV
Definition: xlaenv.f:83
subroutine dsytrf_aa_2stage(UPLO, N, A, LDA, TB, LTB, IPIV, IPIV2, WORK, LWORK, INFO)
DSYTRF_AA_2STAGE
subroutine derrsy(PATH, NUNIT)
DERRSY
Definition: derrsy.f:57
subroutine alasum(TYPE, NOUT, NFAIL, NRUN, NERRS)
ALASUM
Definition: alasum.f:75
subroutine alahd(IOUNIT, PATH)
ALAHD
Definition: alahd.f:107
subroutine dsytrs_aa_2stage(UPLO, N, NRHS, A, LDA, TB, LTB, IPIV, IPIV2, B, LDB, INFO)
DSYTRS_AA_2STAGE
subroutine dlatb4(PATH, IMAT, M, N, TYPE, KL, KU, ANORM, MODE, CNDNUM, DIST)
DLATB4
Definition: dlatb4.f:122
subroutine dlarhs(PATH, XTYPE, UPLO, TRANS, M, N, KL, KU, NRHS, A, LDA, X, LDX, B, LDB, ISEED, INFO)
DLARHS
Definition: dlarhs.f:206
subroutine alaerh(PATH, SUBNAM, INFO, INFOE, OPTS, M, N, KL, KU, N5, IMAT, NFAIL, NERRS, NOUT)
ALAERH
Definition: alaerh.f:149
subroutine dlacpy(UPLO, M, N, A, LDA, B, LDB)
DLACPY copies all or part of one two-dimensional array to another.
Definition: dlacpy.f:105
subroutine dpot02(UPLO, N, NRHS, A, LDA, X, LDX, B, LDB, RWORK, RESID)
DPOT02
Definition: dpot02.f:129
Here is the call graph for this function:
Here is the caller graph for this function: