1       DOUBLE PRECISION FUNCTION DLANSY( NORM, UPLO, N, A, LDA, WORK )
  2 *
  3 *  -- LAPACK auxiliary routine (version 3.2) --
  4 *  -- LAPACK is a software package provided by Univ. of Tennessee,    --
  5 *  -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..--
  6 *     November 2006
  7 *
  8 *     .. Scalar Arguments ..
  9       CHARACTER          NORM, UPLO
 10       INTEGER            LDA, N
 11 *     ..
 12 *     .. Array Arguments ..
 13       DOUBLE PRECISION   A( LDA, * ), WORK( * )
 14 *     ..
 15 *
 16 *  Purpose
 17 *  =======
 18 *
 19 *  DLANSY  returns the value of the one norm,  or the Frobenius norm, or
 20 *  the  infinity norm,  or the  element of  largest absolute value  of a
 21 *  real symmetric matrix A.
 22 *
 23 *  Description
 24 *  ===========
 25 *
 26 *  DLANSY returns the value
 27 *
 28 *     DLANSY = ( max(abs(A(i,j))), NORM = 'M' or 'm'
 29 *              (
 30 *              ( norm1(A),         NORM = '1', 'O' or 'o'
 31 *              (
 32 *              ( normI(A),         NORM = 'I' or 'i'
 33 *              (
 34 *              ( normF(A),         NORM = 'F', 'f', 'E' or 'e'
 35 *
 36 *  where  norm1  denotes the  one norm of a matrix (maximum column sum),
 37 *  normI  denotes the  infinity norm  of a matrix  (maximum row sum) and
 38 *  normF  denotes the  Frobenius norm of a matrix (square root of sum of
 39 *  squares).  Note that  max(abs(A(i,j)))  is not a consistent matrix norm.
 40 *
 41 *  Arguments
 42 *  =========
 43 *
 44 *  NORM    (input) CHARACTER*1
 45 *          Specifies the value to be returned in DLANSY as described
 46 *          above.
 47 *
 48 *  UPLO    (input) CHARACTER*1
 49 *          Specifies whether the upper or lower triangular part of the
 50 *          symmetric matrix A is to be referenced.
 51 *          = 'U':  Upper triangular part of A is referenced
 52 *          = 'L':  Lower triangular part of A is referenced
 53 *
 54 *  N       (input) INTEGER
 55 *          The order of the matrix A.  N >= 0.  When N = 0, DLANSY is
 56 *          set to zero.
 57 *
 58 *  A       (input) DOUBLE PRECISION array, dimension (LDA,N)
 59 *          The symmetric matrix A.  If UPLO = 'U', the leading n by n
 60 *          upper triangular part of A contains the upper triangular part
 61 *          of the matrix A, and the strictly lower triangular part of A
 62 *          is not referenced.  If UPLO = 'L', the leading n by n lower
 63 *          triangular part of A contains the lower triangular part of
 64 *          the matrix A, and the strictly upper triangular part of A is
 65 *          not referenced.
 66 *
 67 *  LDA     (input) INTEGER
 68 *          The leading dimension of the array A.  LDA >= max(N,1).
 69 *
 70 *  WORK    (workspace) DOUBLE PRECISION array, dimension (MAX(1,LWORK)),
 71 *          where LWORK >= N when NORM = 'I' or '1' or 'O'; otherwise,
 72 *          WORK is not referenced.
 73 *
 74 * =====================================================================
 75 *
 76 *     .. Parameters ..
 77       DOUBLE PRECISION   ONE, ZERO
 78       PARAMETER          ( ONE = 1.0D+0, ZERO = 0.0D+0 )
 79 *     ..
 80 *     .. Local Scalars ..
 81       INTEGER            I, J
 82       DOUBLE PRECISION   ABSA, SCALESUMVALUE
 83 *     ..
 84 *     .. External Subroutines ..
 85       EXTERNAL           DLASSQ
 86 *     ..
 87 *     .. External Functions ..
 88       LOGICAL            LSAME
 89       EXTERNAL           LSAME
 90 *     ..
 91 *     .. Intrinsic Functions ..
 92       INTRINSIC          ABSMAXSQRT
 93 *     ..
 94 *     .. Executable Statements ..
 95 *
 96       IF( N.EQ.0 ) THEN
 97          VALUE = ZERO
 98       ELSE IF( LSAME( NORM, 'M' ) ) THEN
 99 *
100 *        Find max(abs(A(i,j))).
101 *
102          VALUE = ZERO
103          IF( LSAME( UPLO, 'U' ) ) THEN
104             DO 20 J = 1, N
105                DO 10 I = 1, J
106                   VALUE = MAXVALUEABS( A( I, J ) ) )
107    10          CONTINUE
108    20       CONTINUE
109          ELSE
110             DO 40 J = 1, N
111                DO 30 I = J, N
112                   VALUE = MAXVALUEABS( A( I, J ) ) )
113    30          CONTINUE
114    40       CONTINUE
115          END IF
116       ELSE IF( ( LSAME( NORM, 'I' ) ) .OR. ( LSAME( NORM, 'O' ) ) .OR.
117      $         ( NORM.EQ.'1' ) ) THEN
118 *
119 *        Find normI(A) ( = norm1(A), since A is symmetric).
120 *
121          VALUE = ZERO
122          IF( LSAME( UPLO, 'U' ) ) THEN
123             DO 60 J = 1, N
124                SUM = ZERO
125                DO 50 I = 1, J - 1
126                   ABSA = ABS( A( I, J ) )
127                   SUM = SUM + ABSA
128                   WORK( I ) = WORK( I ) + ABSA
129    50          CONTINUE
130                WORK( J ) = SUM + ABS( A( J, J ) )
131    60       CONTINUE
132             DO 70 I = 1, N
133                VALUE = MAXVALUE, WORK( I ) )
134    70       CONTINUE
135          ELSE
136             DO 80 I = 1, N
137                WORK( I ) = ZERO
138    80       CONTINUE
139             DO 100 J = 1, N
140                SUM = WORK( J ) + ABS( A( J, J ) )
141                DO 90 I = J + 1, N
142                   ABSA = ABS( A( I, J ) )
143                   SUM = SUM + ABSA
144                   WORK( I ) = WORK( I ) + ABSA
145    90          CONTINUE
146                VALUE = MAXVALUESUM )
147   100       CONTINUE
148          END IF
149       ELSE IF( ( LSAME( NORM, 'F' ) ) .OR. ( LSAME( NORM, 'E' ) ) ) THEN
150 *
151 *        Find normF(A).
152 *
153          SCALE = ZERO
154          SUM = ONE
155          IF( LSAME( UPLO, 'U' ) ) THEN
156             DO 110 J = 2, N
157                CALL DLASSQ( J-1, A( 1, J ), 1SCALESUM )
158   110       CONTINUE
159          ELSE
160             DO 120 J = 1, N - 1
161                CALL DLASSQ( N-J, A( J+1, J ), 1SCALESUM )
162   120       CONTINUE
163          END IF
164          SUM = 2*SUM
165          CALL DLASSQ( N, A, LDA+1SCALESUM )
166          VALUE = SCALE*SQRTSUM )
167       END IF
168 *
169       DLANSY = VALUE
170       RETURN
171 *
172 *     End of DLANSY
173 *
174       END