LAPACK 3.12.1
LAPACK: Linear Algebra PACKage
All Classes Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Modules Pages
dla_lin_berr.f
Go to the documentation of this file.
1*> \brief \b DLA_LIN_BERR computes a component-wise relative backward error.
2*
3* =========== DOCUMENTATION ===========
4*
5* Online html documentation available at
6* http://www.netlib.org/lapack/explore-html/
7*
8*> Download DLA_LIN_BERR + dependencies
9*> <a href="http://www.netlib.org/cgi-bin/netlibfiles.tgz?format=tgz&filename=/lapack/lapack_routine/dla_lin_berr.f">
10*> [TGZ]</a>
11*> <a href="http://www.netlib.org/cgi-bin/netlibfiles.zip?format=zip&filename=/lapack/lapack_routine/dla_lin_berr.f">
12*> [ZIP]</a>
13*> <a href="http://www.netlib.org/cgi-bin/netlibfiles.txt?format=txt&filename=/lapack/lapack_routine/dla_lin_berr.f">
14*> [TXT]</a>
15*
16* Definition:
17* ===========
18*
19* SUBROUTINE DLA_LIN_BERR( N, NZ, NRHS, RES, AYB, BERR )
20*
21* .. Scalar Arguments ..
22* INTEGER N, NZ, NRHS
23* ..
24* .. Array Arguments ..
25* DOUBLE PRECISION AYB( N, NRHS ), BERR( NRHS )
26* DOUBLE PRECISION RES( N, NRHS )
27* ..
28*
29*
30*> \par Purpose:
31* =============
32*>
33*> \verbatim
34*>
35*> DLA_LIN_BERR computes component-wise relative backward error from
36*> the formula
37*> max(i) ( abs(R(i)) / ( abs(op(A_s))*abs(Y) + abs(B_s) )(i) )
38*> where abs(Z) is the component-wise absolute value of the matrix
39*> or vector Z.
40*> \endverbatim
41*
42* Arguments:
43* ==========
44*
45*> \param[in] N
46*> \verbatim
47*> N is INTEGER
48*> The number of linear equations, i.e., the order of the
49*> matrix A. N >= 0.
50*> \endverbatim
51*>
52*> \param[in] NZ
53*> \verbatim
54*> NZ is INTEGER
55*> We add (NZ+1)*SLAMCH( 'Safe minimum' ) to R(i) in the numerator to
56*> guard against spuriously zero residuals. Default value is N.
57*> \endverbatim
58*>
59*> \param[in] NRHS
60*> \verbatim
61*> NRHS is INTEGER
62*> The number of right hand sides, i.e., the number of columns
63*> of the matrices AYB, RES, and BERR. NRHS >= 0.
64*> \endverbatim
65*>
66*> \param[in] RES
67*> \verbatim
68*> RES is DOUBLE PRECISION array, dimension (N,NRHS)
69*> The residual matrix, i.e., the matrix R in the relative backward
70*> error formula above.
71*> \endverbatim
72*>
73*> \param[in] AYB
74*> \verbatim
75*> AYB is DOUBLE PRECISION array, dimension (N, NRHS)
76*> The denominator in the relative backward error formula above, i.e.,
77*> the matrix abs(op(A_s))*abs(Y) + abs(B_s). The matrices A, Y, and B
78*> are from iterative refinement (see dla_gerfsx_extended.f).
79*> \endverbatim
80*>
81*> \param[out] BERR
82*> \verbatim
83*> BERR is DOUBLE PRECISION array, dimension (NRHS)
84*> The component-wise relative backward error from the formula above.
85*> \endverbatim
86*
87* Authors:
88* ========
89*
90*> \author Univ. of Tennessee
91*> \author Univ. of California Berkeley
92*> \author Univ. of Colorado Denver
93*> \author NAG Ltd.
94*
95*> \ingroup la_lin_berr
96*
97* =====================================================================
98 SUBROUTINE dla_lin_berr ( N, NZ, NRHS, RES, AYB, BERR )
99*
100* -- LAPACK computational routine --
101* -- LAPACK is a software package provided by Univ. of Tennessee, --
102* -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..--
103*
104* .. Scalar Arguments ..
105 INTEGER N, NZ, NRHS
106* ..
107* .. Array Arguments ..
108 DOUBLE PRECISION AYB( N, NRHS ), BERR( NRHS )
109 DOUBLE PRECISION RES( N, NRHS )
110* ..
111*
112* =====================================================================
113*
114* .. Local Scalars ..
115 DOUBLE PRECISION TMP
116 INTEGER I, J
117* ..
118* .. Intrinsic Functions ..
119 INTRINSIC abs, max
120* ..
121* .. External Functions ..
122 EXTERNAL dlamch
123 DOUBLE PRECISION DLAMCH
124 DOUBLE PRECISION SAFE1
125* ..
126* .. Executable Statements ..
127*
128* Adding SAFE1 to the numerator guards against spuriously zero
129* residuals. A similar safeguard is in the SLA_yyAMV routine used
130* to compute AYB.
131*
132 safe1 = dlamch( 'Safe minimum' )
133 safe1 = (nz+1)*safe1
134
135 DO j = 1, nrhs
136 berr(j) = 0.0d+0
137 DO i = 1, n
138 IF (ayb(i,j) .NE. 0.0d+0) THEN
139 tmp = (safe1+abs(res(i,j)))/ayb(i,j)
140 berr(j) = max( berr(j), tmp )
141 END IF
142*
143* If AYB is exactly 0.0 (and if computed by SLA_yyAMV), then we know
144* the true residual also must be exactly 0.0.
145*
146 END DO
147 END DO
148*
149* End of DLA_LIN_BERR
150*
151 END
subroutine dla_lin_berr(n, nz, nrhs, res, ayb, berr)
DLA_LIN_BERR computes a component-wise relative backward error.
double precision function dlamch(cmach)
DLAMCH
Definition dlamch.f:69