1       SUBROUTINE CTBT02( UPLO, TRANS, DIAG, N, KD, NRHS, AB, LDAB, X,
  2      $                   LDX, B, LDB, WORK, RWORK, RESID )
  3 *
  4 *  -- LAPACK test routine (version 3.1) --
  5 *     Univ. of Tennessee, Univ. of California Berkeley and NAG Ltd..
  6 *     November 2006
  7 *
  8 *     .. Scalar Arguments ..
  9       CHARACTER          DIAG, TRANS, UPLO
 10       INTEGER            KD, LDAB, LDB, LDX, N, NRHS
 11       REAL               RESID
 12 *     ..
 13 *     .. Array Arguments ..
 14       REAL               RWORK( * )
 15       COMPLEX            AB( LDAB, * ), B( LDB, * ), WORK( * ),
 16      $                   X( LDX, * )
 17 *     ..
 18 *
 19 *  Purpose
 20 *  =======
 21 *
 22 *  CTBT02 computes the residual for the computed solution to a
 23 *  triangular system of linear equations  A*x = b,  A**T *x = b,  or
 24 *  A**H *x = b  when A is a triangular band matrix.  Here A**T denotes
 25 *  the transpose of A, A**H denotes the conjugate transpose of A, and
 26 *  x and b are N by NRHS matrices.  The test ratio is the maximum over
 27 *  the number of right hand sides of
 28 *     norm(b - op(A)*x) / ( norm(op(A)) * norm(x) * EPS ),
 29 *  where op(A) denotes A, A**T, or A**H, and EPS is the machine epsilon.
 30 *
 31 *  Arguments
 32 *  =========
 33 *
 34 *  UPLO    (input) CHARACTER*1
 35 *          Specifies whether the matrix A is upper or lower triangular.
 36 *          = 'U':  Upper triangular
 37 *          = 'L':  Lower triangular
 38 *
 39 *  TRANS   (input) CHARACTER*1
 40 *          Specifies the operation applied to A.
 41 *          = 'N':  A *x = b     (No transpose)
 42 *          = 'T':  A**T *x = b  (Transpose)
 43 *          = 'C':  A**H *x = b  (Conjugate transpose)
 44 *
 45 *  DIAG    (input) CHARACTER*1
 46 *          Specifies whether or not the matrix A is unit triangular.
 47 *          = 'N':  Non-unit triangular
 48 *          = 'U':  Unit triangular
 49 *
 50 *  N       (input) INTEGER
 51 *          The order of the matrix A.  N >= 0.
 52 *
 53 *  KD      (input) INTEGER
 54 *          The number of superdiagonals or subdiagonals of the
 55 *          triangular band matrix A.  KD >= 0.
 56 *
 57 *  NRHS    (input) INTEGER
 58 *          The number of right hand sides, i.e., the number of columns
 59 *          of the matrices X and B.  NRHS >= 0.
 60 *
 61 *  AB      (input) COMPLEX array, dimension (LDA,N)
 62 *          The upper or lower triangular band matrix A, stored in the
 63 *          first kd+1 rows of the array. The j-th column of A is stored
 64 *          in the j-th column of the array AB as follows:
 65 *          if UPLO = 'U', AB(kd+1+i-j,j) = A(i,j) for max(1,j-kd)<=i<=j;
 66 *          if UPLO = 'L', AB(1+i-j,j)    = A(i,j) for j<=i<=min(n,j+kd).
 67 *
 68 *  LDAB    (input) INTEGER
 69 *          The leading dimension of the array AB.  LDAB >= max(1,KD+1).
 70 *
 71 *  X       (input) COMPLEX array, dimension (LDX,NRHS)
 72 *          The computed solution vectors for the system of linear
 73 *          equations.
 74 *
 75 *  LDX     (input) INTEGER
 76 *          The leading dimension of the array X.  LDX >= max(1,N).
 77 *
 78 *  B       (input) COMPLEX array, dimension (LDB,NRHS)
 79 *          The right hand side vectors for the system of linear
 80 *          equations.
 81 *
 82 *  LDB     (input) INTEGER
 83 *          The leading dimension of the array B.  LDB >= max(1,N).
 84 *
 85 *  WORK    (workspace) COMPLEX array, dimension (N)
 86 *
 87 *  RWORK   (workspace) REAL array, dimension (N)
 88 *
 89 *  RESID   (output) REAL
 90 *          The maximum over the number of right hand sides of
 91 *          norm(op(A)*x - b) / ( norm(op(A)) * norm(x) * EPS ).
 92 *
 93 *  =====================================================================
 94 *
 95 *     .. Parameters ..
 96       REAL               ZERO, ONE
 97       PARAMETER          ( ZERO = 0.0E+0, ONE = 1.0E+0 )
 98 *     ..
 99 *     .. Local Scalars ..
100       INTEGER            J
101       REAL               ANORM, BNORM, EPS, XNORM
102 *     ..
103 *     .. External Functions ..
104       LOGICAL            LSAME
105       REAL               CLANTB, SCASUM, SLAMCH
106       EXTERNAL           LSAME, CLANTB, SCASUM, SLAMCH
107 *     ..
108 *     .. External Subroutines ..
109       EXTERNAL           CAXPY, CCOPY, CTBMV
110 *     ..
111 *     .. Intrinsic Functions ..
112       INTRINSIC          CMPLXMAX
113 *     ..
114 *     .. Executable Statements ..
115 *
116 *     Quick exit if N = 0 or NRHS = 0
117 *
118       IF( N.LE.0 .OR. NRHS.LE.0 ) THEN
119          RESID = ZERO
120          RETURN
121       END IF
122 *
123 *     Compute the 1-norm of A or A'.
124 *
125       IF( LSAME( TRANS, 'N' ) ) THEN
126          ANORM = CLANTB( '1', UPLO, DIAG, N, KD, AB, LDAB, RWORK )
127       ELSE
128          ANORM = CLANTB( 'I', UPLO, DIAG, N, KD, AB, LDAB, RWORK )
129       END IF
130 *
131 *     Exit with RESID = 1/EPS if ANORM = 0.
132 *
133       EPS = SLAMCH( 'Epsilon' )
134       IF( ANORM.LE.ZERO ) THEN
135          RESID = ONE / EPS
136          RETURN
137       END IF
138 *
139 *     Compute the maximum over the number of right hand sides of
140 *        norm(op(A)*x - b) / ( norm(op(A)) * norm(x) * EPS ).
141 *
142       RESID = ZERO
143       DO 10 J = 1, NRHS
144          CALL CCOPY( N, X( 1, J ), 1, WORK, 1 )
145          CALL CTBMV( UPLO, TRANS, DIAG, N, KD, AB, LDAB, WORK, 1 )
146          CALL CAXPY( N, CMPLX-ONE ), B( 1, J ), 1, WORK, 1 )
147          BNORM = SCASUM( N, WORK, 1 )
148          XNORM = SCASUM( N, X( 1, J ), 1 )
149          IF( XNORM.LE.ZERO ) THEN
150             RESID = ONE / EPS
151          ELSE
152             RESID = MAX( RESID, ( ( BNORM / ANORM ) / XNORM ) / EPS )
153          END IF
154    10 CONTINUE
155 *
156       RETURN
157 *
158 *     End of CTBT02
159 *
160       END