*> \brief \b DDRVAC * * =========== DOCUMENTATION =========== * * Online html documentation available at * http://www.netlib.org/lapack/explore-html/ * * Definition * ========== * * SUBROUTINE DDRVAC( DOTYPE, NM, MVAL, NNS, NSVAL, THRESH, NMAX, * A, AFAC, B, X, WORK, * RWORK, SWORK, NOUT ) * * .. Scalar Arguments .. * INTEGER NMAX, NM, NNS, NOUT * DOUBLE PRECISION THRESH * .. * .. Array Arguments .. * LOGICAL DOTYPE( * ) * INTEGER MVAL( * ), NSVAL( * ) * REAL SWORK(*) * DOUBLE PRECISION A( * ), AFAC( * ), B( * ), * $ RWORK( * ), WORK( * ), X( * ) * .. * * Purpose * ======= * *>\details \b Purpose: *>\verbatim *> *> DDRVAC tests DSPOSV. *> *>\endverbatim * * Arguments * ========= * *> \param[in] DOTYPE *> \verbatim *> 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. *> \endverbatim *> *> \param[in] NM *> \verbatim *> NM is INTEGER *> The number of values of N contained in the vector MVAL. *> \endverbatim *> *> \param[in] MVAL *> \verbatim *> MVAL is INTEGER array, dimension (NM) *> The values of the matrix dimension N. *> \endverbatim *> *> \param[in] NNS *> \verbatim *> NNS is INTEGER *> The number of values of NRHS contained in the vector NSVAL. *> \endverbatim *> *> \param[in] NSVAL *> \verbatim *> NSVAL is INTEGER array, dimension (NNS) *> The values of the number of right hand sides NRHS. *> \endverbatim *> *> \param[in] THRESH *> \verbatim *> 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. *> \endverbatim *> *> \param[in] NMAX *> \verbatim *> NMAX is INTEGER *> The maximum value permitted for N, used in dimensioning the *> work arrays. *> \endverbatim *> *> \param[out] A *> \verbatim *> A is DOUBLE PRECISION array, dimension (NMAX*NMAX) *> \endverbatim *> *> \param[out] AFAC *> \verbatim *> AFAC is DOUBLE PRECISION array, dimension (NMAX*NMAX) *> \endverbatim *> *> \param[out] B *> \verbatim *> B is DOUBLE PRECISION array, dimension (NMAX*NSMAX) *> \endverbatim *> *> \param[out] X *> \verbatim *> X is DOUBLE PRECISION array, dimension (NMAX*NSMAX) *> \endverbatim *> *> \param[out] WORK *> \verbatim *> WORK is DOUBLE PRECISION array, dimension *> (NMAX*max(3,NSMAX)) *> \endverbatim *> *> \param[out] RWORK *> \verbatim *> RWORK is DOUBLE PRECISION array, dimension *> (max(2*NMAX,2*NSMAX+NWORK)) *> \endverbatim *> *> \param[out] SWORK *> \verbatim *> SWORK is REAL array, dimension *> (NMAX*(NSMAX+NMAX)) *> \endverbatim *> *> \param[in] NOUT *> \verbatim *> NOUT is INTEGER *> The unit number for output. *> \endverbatim *> * * Authors * ======= * *> \author Univ. of Tennessee *> \author Univ. of California Berkeley *> \author Univ. of Colorado Denver *> \author NAG Ltd. * *> \date November 2011 * *> \ingroup double_lin * * ===================================================================== SUBROUTINE DDRVAC( DOTYPE, NM, MVAL, NNS, NSVAL, THRESH, NMAX, $ A, AFAC, B, X, WORK, $ RWORK, SWORK, NOUT ) * * -- LAPACK test routine (version 3.1.2) -- * -- LAPACK is a software package provided by Univ. of Tennessee, -- * -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..-- * November 2011 * * .. Scalar Arguments .. INTEGER NMAX, NM, NNS, NOUT DOUBLE PRECISION THRESH * .. * .. Array Arguments .. LOGICAL DOTYPE( * ) INTEGER MVAL( * ), NSVAL( * ) REAL SWORK(*) DOUBLE PRECISION A( * ), AFAC( * ), B( * ), $ RWORK( * ), WORK( * ), X( * ) * .. * * ===================================================================== * * .. Parameters .. DOUBLE PRECISION ONE, ZERO PARAMETER ( ONE = 1.0D+0, ZERO = 0.0D+0 ) INTEGER NTYPES PARAMETER ( NTYPES = 9 ) INTEGER NTESTS PARAMETER ( NTESTS = 1 ) * .. * .. Local Scalars .. LOGICAL ZEROT CHARACTER DIST, TYPE, UPLO, XTYPE CHARACTER*3 PATH INTEGER I, IM, IMAT, INFO, IOFF, IRHS, IUPLO, $ IZERO, KL, KU, LDA, MODE, N, $ NERRS, NFAIL, NIMAT, NRHS, NRUN DOUBLE PRECISION ANORM, CNDNUM * .. * .. Local Arrays .. CHARACTER UPLOS( 2 ) INTEGER ISEED( 4 ), ISEEDY( 4 ) DOUBLE PRECISION RESULT( NTESTS ) * .. * .. Local Variables .. INTEGER ITER, KASE * .. * .. External Functions .. LOGICAL LSAME EXTERNAL LSAME * .. * .. External Subroutines .. EXTERNAL ALAERH, DLACPY, $ DLARHS, DLASET, DLATB4, DLATMS, $ DPOT06, DSPOSV * .. * .. Intrinsic Functions .. INTRINSIC DBLE, MAX, SQRT * .. * .. Scalars in Common .. LOGICAL LERR, OK CHARACTER*32 SRNAMT INTEGER INFOT, NUNIT * .. * .. Common blocks .. COMMON / INFOC / INFOT, NUNIT, OK, LERR COMMON / SRNAMC / SRNAMT * .. * .. Data statements .. DATA ISEEDY / 1988, 1989, 1990, 1991 / DATA UPLOS / 'U', 'L' / * .. * .. Executable Statements .. * * Initialize constants and the random number seed. * KASE = 0 PATH( 1: 1 ) = 'Double precision' PATH( 2: 3 ) = 'PO' NRUN = 0 NFAIL = 0 NERRS = 0 DO 10 I = 1, 4 ISEED( I ) = ISEEDY( I ) 10 CONTINUE * INFOT = 0 * * Do for each value of N in MVAL * DO 120 IM = 1, NM N = MVAL( IM ) LDA = MAX( N, 1 ) NIMAT = NTYPES IF( N.LE.0 ) $ NIMAT = 1 * DO 110 IMAT = 1, NIMAT * * Do the tests only if DOTYPE( IMAT ) is true. * IF( .NOT.DOTYPE( IMAT ) ) $ GO TO 110 * * Skip types 3, 4, or 5 if the matrix size is too small. * ZEROT = IMAT.GE.3 .AND. IMAT.LE.5 IF( ZEROT .AND. N.LT.IMAT-2 ) $ GO TO 110 * * Do first for UPLO = 'U', then for UPLO = 'L' * DO 100 IUPLO = 1, 2 UPLO = UPLOS( IUPLO ) * * Set up parameters with DLATB4 and generate a test matrix * with DLATMS. * CALL DLATB4( PATH, IMAT, N, N, TYPE, KL, KU, ANORM, MODE, $ CNDNUM, DIST ) * SRNAMT = 'DLATMS' CALL DLATMS( N, N, DIST, ISEED, TYPE, RWORK, MODE, $ CNDNUM, ANORM, KL, KU, UPLO, A, LDA, WORK, $ INFO ) * * Check error code from DLATMS. * IF( INFO.NE.0 ) THEN CALL ALAERH( PATH, 'DLATMS', INFO, 0, UPLO, N, N, -1, $ -1, -1, IMAT, NFAIL, NERRS, NOUT ) GO TO 100 END IF * * For types 3-5, zero one row and column of the matrix to * test that INFO is returned correctly. * IF( ZEROT ) THEN IF( IMAT.EQ.3 ) THEN IZERO = 1 ELSE IF( IMAT.EQ.4 ) THEN IZERO = N ELSE IZERO = N / 2 + 1 END IF IOFF = ( IZERO-1 )*LDA * * Set row and column IZERO of A to 0. * IF( IUPLO.EQ.1 ) THEN DO 20 I = 1, IZERO - 1 A( IOFF+I ) = ZERO 20 CONTINUE IOFF = IOFF + IZERO DO 30 I = IZERO, N A( IOFF ) = ZERO IOFF = IOFF + LDA 30 CONTINUE ELSE IOFF = IZERO DO 40 I = 1, IZERO - 1 A( IOFF ) = ZERO IOFF = IOFF + LDA 40 CONTINUE IOFF = IOFF - IZERO DO 50 I = IZERO, N A( IOFF+I ) = ZERO 50 CONTINUE END IF ELSE IZERO = 0 END IF * DO 60 IRHS = 1, NNS NRHS = NSVAL( IRHS ) XTYPE = 'N' * * Form an exact solution and set the right hand side. * SRNAMT = 'DLARHS' CALL DLARHS( PATH, XTYPE, UPLO, ' ', N, N, KL, KU, $ NRHS, A, LDA, X, LDA, B, LDA, $ ISEED, INFO ) * * Compute the L*L' or U'*U factorization of the * matrix and solve the system. * SRNAMT = 'DSPOSV ' KASE = KASE + 1 * CALL DLACPY( 'All', N, N, A, LDA, AFAC, LDA) * CALL DSPOSV( UPLO, N, NRHS, AFAC, LDA, B, LDA, X, LDA, $ WORK, SWORK, ITER, INFO ) IF (ITER.LT.0) THEN CALL DLACPY( 'All', N, N, A, LDA, AFAC, LDA ) ENDIF * * Check error code from DSPOSV . * IF( INFO.NE.IZERO ) THEN * IF( NFAIL.EQ.0 .AND. NERRS.EQ.0 ) $ CALL ALAHD( NOUT, PATH ) NERRS = NERRS + 1 * IF( INFO.NE.IZERO .AND. IZERO.NE.0 ) THEN WRITE( NOUT, FMT = 9988 )'DSPOSV',INFO,IZERO,N, $ IMAT ELSE WRITE( NOUT, FMT = 9975 )'DSPOSV',INFO,N,IMAT END IF END IF * * Skip the remaining test if the matrix is singular. * IF( INFO.NE.0 ) $ GO TO 110 * * Check the quality of the solution * CALL DLACPY( 'All', N, NRHS, B, LDA, WORK, LDA ) * CALL DPOT06( UPLO, N, NRHS, A, LDA, X, LDA, WORK, $ LDA, RWORK, RESULT( 1 ) ) * * Check if the test passes the tesing. * Print information about the tests that did not * pass the testing. * * If iterative refinement has been used and claimed to * be successful (ITER>0), we want * NORM1(B - A*X)/(NORM1(A)*NORM1(X)*EPS*SRQT(N)) < 1 * * If double precision has been used (ITER<0), we want * NORM1(B - A*X)/(NORM1(A)*NORM1(X)*EPS) < THRES * (Cf. the linear solver testing routines) * IF ((THRESH.LE.0.0E+00) $ .OR.((ITER.GE.0).AND.(N.GT.0) $ .AND.(RESULT(1).GE.SQRT(DBLE(N)))) $ .OR.((ITER.LT.0).AND.(RESULT(1).GE.THRESH))) THEN * IF( NFAIL.EQ.0 .AND. NERRS.EQ.0 ) THEN WRITE( NOUT, FMT = 8999 )'DPO' WRITE( NOUT, FMT = '( '' Matrix types:'' )' ) WRITE( NOUT, FMT = 8979 ) WRITE( NOUT, FMT = '( '' Test ratios:'' )' ) WRITE( NOUT, FMT = 8960 )1 WRITE( NOUT, FMT = '( '' Messages:'' )' ) END IF * WRITE( NOUT, FMT = 9998 )UPLO, N, NRHS, IMAT, 1, $ RESULT( 1 ) * NFAIL = NFAIL + 1 * END IF * NRUN = NRUN + 1 * 60 CONTINUE 100 CONTINUE 110 CONTINUE 120 CONTINUE * * Print a summary of the results. * IF( NFAIL.GT.0 ) THEN WRITE( NOUT, FMT = 9996 )'DSPOSV', NFAIL, NRUN ELSE WRITE( NOUT, FMT = 9995 )'DSPOSV', NRUN END IF IF( NERRS.GT.0 ) THEN WRITE( NOUT, FMT = 9994 )NERRS END IF * 9998 FORMAT( ' UPLO=''', A1, ''', N =', I5, ', NRHS=', I3, ', type ', $ I2, ', test(', I2, ') =', G12.5 ) 9996 FORMAT( 1X, A6, ': ', I6, ' out of ', I6, $ ' tests failed to pass the threshold' ) 9995 FORMAT( /1X, 'All tests for ', A6, $ ' routines passed the threshold ( ', I6, ' tests run)' ) 9994 FORMAT( 6X, I6, ' error messages recorded' ) * * SUBNAM, INFO, INFOE, N, IMAT * 9988 FORMAT( ' *** ', A6, ' returned with INFO =', I5, ' instead of ', $ I5, / ' ==> N =', I5, ', type ', $ I2 ) * * SUBNAM, INFO, N, IMAT * 9975 FORMAT( ' *** Error code from ', A6, '=', I5, ' for M=', I5, $ ', type ', I2 ) 8999 FORMAT( / 1X, A3, ': positive definite dense matrices' ) 8979 FORMAT( 4X, '1. Diagonal', 24X, '7. Last n/2 columns zero', / 4X, $ '2. Upper triangular', 16X, $ '8. Random, CNDNUM = sqrt(0.1/EPS)', / 4X, $ '3. Lower triangular', 16X, '9. Random, CNDNUM = 0.1/EPS', $ / 4X, '4. Random, CNDNUM = 2', 13X, $ '10. Scaled near underflow', / 4X, '5. First column zero', $ 14X, '11. Scaled near overflow', / 4X, $ '6. Last column zero' ) 8960 FORMAT( 3X, I2, ': norm_1( B - A * X ) / ', $ '( norm_1(A) * norm_1(X) * EPS * SQRT(N) ) > 1 if ITERREF', $ / 4x, 'or norm_1( B - A * X ) / ', $ '( norm_1(A) * norm_1(X) * EPS ) > THRES if DPOTRF' ) RETURN * * End of DDRVAC * END