1 *> \brief \b DLA_LIN_BERR computes a component-wise relative backward error.
3 * =========== DOCUMENTATION ===========
5 * Online html documentation available at
6 * http://www.netlib.org/lapack/explore-html/
9 *> Download DLA_LIN_BERR + dependencies
10 *> <a href="http://www.netlib.org/cgi-bin/netlibfiles.tgz?format=tgz&filename=/lapack/lapack_routine/dla_lin_berr.f">
12 *> <a href="http://www.netlib.org/cgi-bin/netlibfiles.zip?format=zip&filename=/lapack/lapack_routine/dla_lin_berr.f">
14 *> <a href="http://www.netlib.org/cgi-bin/netlibfiles.txt?format=txt&filename=/lapack/lapack_routine/dla_lin_berr.f">
21 * SUBROUTINE DLA_LIN_BERR ( N, NZ, NRHS, RES, AYB, BERR )
23 * .. Scalar Arguments ..
26 * .. Array Arguments ..
27 * DOUBLE PRECISION AYB( N, NRHS ), BERR( NRHS )
28 * DOUBLE PRECISION RES( N, NRHS )
37 *> DLA_LIN_BERR computes component-wise relative backward error from
39 *> max(i) ( abs(R(i)) / ( abs(op(A_s))*abs(Y) + abs(B_s) )(i) )
40 *> where abs(Z) is the component-wise absolute value of the matrix
50 *> The number of linear equations, i.e., the order of the
57 *> We add (NZ+1)*SLAMCH( 'Safe minimum' ) to R(i) in the numerator to
58 *> guard against spuriously zero residuals. Default value is N.
64 *> The number of right hand sides, i.e., the number of columns
65 *> of the matrices AYB, RES, and BERR. NRHS >= 0.
70 *> RES is DOUBLE PRECISION array, dimension (N,NRHS)
71 *> The residual matrix, i.e., the matrix R in the relative backward
72 *> error formula above.
77 *> AYB is DOUBLE PRECISION array, dimension (N, NRHS)
78 *> The denominator in the relative backward error formula above, i.e.,
79 *> the matrix abs(op(A_s))*abs(Y) + abs(B_s). The matrices A, Y, and B
80 *> are from iterative refinement (see dla_gerfsx_extended.f).
85 *> BERR is DOUBLE PRECISION array, dimension (NRHS)
86 *> The component-wise relative backward error from the formula above.
92 *> \author Univ. of Tennessee
93 *> \author Univ. of California Berkeley
94 *> \author Univ. of Colorado Denver
97 *> \date September 2012
99 *> \ingroup doubleOTHERcomputational
101 * =====================================================================
102 SUBROUTINE DLA_LIN_BERR ( N, NZ, NRHS, RES, AYB, BERR )
104 * -- LAPACK computational routine (version 3.4.2) --
105 * -- LAPACK is a software package provided by Univ. of Tennessee, --
106 * -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..--
109 * .. Scalar Arguments ..
112 * .. Array Arguments ..
113 DOUBLE PRECISION AYB( N, NRHS ), BERR( NRHS )
114 DOUBLE PRECISION RES( N, NRHS )
117 * =====================================================================
119 * .. Local Scalars ..
123 * .. Intrinsic Functions ..
126 * .. External Functions ..
128 DOUBLE PRECISION DLAMCH
129 DOUBLE PRECISION SAFE1
131 * .. Executable Statements ..
133 * Adding SAFE1 to the numerator guards against spuriously zero
134 * residuals. A similar safeguard is in the SLA_yyAMV routine used
137 SAFE1 = DLAMCH( 'Safe minimum' )
143 IF (AYB(I,J) .NE. 0.0D+0) THEN
144 TMP = (SAFE1+ABS(RES(I,J)))/AYB(I,J)
145 BERR(J) = MAX( BERR(J), TMP )
148 * If AYB is exactly 0.0 (and if computed by SLA_yyAMV), then we know
149 * the true residual also must be exactly 0.0.