1       SUBROUTINE ZLA_LIN_BERR ( N, NZ, NRHS, RES, AYB, BERR )
 2 *
 3 *     -- LAPACK routine (version 3.2.2)                                 --
 4 *     -- Contributed by James Demmel, Deaglan Halligan, Yozo Hida and --
 5 *     -- Jason Riedy of Univ. of California Berkeley.                 --
 6 *     -- June 2010                                                    --
 7 *
 8 *     -- LAPACK is a software package provided by Univ. of Tennessee, --
 9 *     -- Univ. of California Berkeley and NAG Ltd.                    --
10 *
11       IMPLICIT NONE
12 *     ..
13 *     .. Scalar Arguments ..
14       INTEGER            N, NZ, NRHS
15 *     ..
16 *     .. Array Arguments ..
17       DOUBLE PRECISION   AYB( N, NRHS ), BERR( NRHS )
18       COMPLEX*16         RES( N, NRHS )
19 *     ..
20 *
21 *  Purpose
22 *  =======
23 *
24 *     ZLA_LIN_BERR computes componentwise relative backward error from
25 *     the formula
26 *         max(i) ( abs(R(i)) / ( abs(op(A_s))*abs(Y) + abs(B_s) )(i) )
27 *     where abs(Z) is the componentwise absolute value of the matrix
28 *     or vector Z.
29 *
30 *     N       (input) INTEGER
31 *     The number of linear equations, i.e., the order of the
32 *     matrix A.  N >= 0.
33 *
34 *     NZ      (input) INTEGER
35 *     We add (NZ+1)*SLAMCH( 'Safe minimum' ) to R(i) in the numerator to
36 *     guard against spuriously zero residuals. Default value is N.
37 *
38 *     NRHS    (input) INTEGER
39 *     The number of right hand sides, i.e., the number of columns
40 *     of the matrices AYB, RES, and BERR.  NRHS >= 0.
41 *
42 *     RES    (input) DOUBLE PRECISION array, dimension (N,NRHS)
43 *     The residual matrix, i.e., the matrix R in the relative backward
44 *     error formula above.
45 *
46 *     AYB    (input) DOUBLE PRECISION array, dimension (N, NRHS)
47 *     The denominator in the relative backward error formula above, i.e.,
48 *     the matrix abs(op(A_s))*abs(Y) + abs(B_s). The matrices A, Y, and B
49 *     are from iterative refinement (see zla_gerfsx_extended.f).
50 *     
51 *     BERR   (output) COMPLEX*16 array, dimension (NRHS)
52 *     The componentwise relative backward error from the formula above.
53 *
54 *  =====================================================================
55 *
56 *     .. Local Scalars ..
57       DOUBLE PRECISION   TMP
58       INTEGER            I, J
59       COMPLEX*16         CDUM
60 *     ..
61 *     .. Intrinsic Functions ..
62       INTRINSIC          ABS, REAL, DIMAGMAX
63 *     ..
64 *     .. External Functions ..
65       EXTERNAL           DLAMCH
66       DOUBLE PRECISION   DLAMCH
67       DOUBLE PRECISION   SAFE1
68 *     ..
69 *     .. Statement Functions ..
70       COMPLEX*16         CABS1
71 *     ..
72 *     .. Statement Function Definitions ..
73       CABS1( CDUM ) = ABSDBLE( CDUM ) ) + ABSDIMAG( CDUM ) )
74 *     ..
75 *     .. Executable Statements ..
76 *
77 *     Adding SAFE1 to the numerator guards against spuriously zero
78 *     residuals.  A similar safeguard is in the CLA_yyAMV routine used
79 *     to compute AYB.
80 *
81       SAFE1 = DLAMCH( 'Safe minimum' )
82       SAFE1 = (NZ+1)*SAFE1
83 
84       DO J = 1, NRHS
85          BERR(J) = 0.0D+0
86          DO I = 1, N
87             IF (AYB(I,J) .NE. 0.0D+0THEN
88                TMP = (SAFE1 + CABS1(RES(I,J)))/AYB(I,J)
89                BERR(J) = MAX( BERR(J), TMP )
90             END IF
91 *
92 *     If AYB is exactly 0.0 (and if computed by CLA_yyAMV), then we know
93 *     the true residual also must be exactly 0.0.
94 *
95          END DO
96       END DO
97       END