![]() |
LAPACK
3.4.1
LAPACK: Linear Algebra PACKage
|
00001 *> \brief \b DLA_LIN_BERR 00002 * 00003 * =========== DOCUMENTATION =========== 00004 * 00005 * Online html documentation available at 00006 * http://www.netlib.org/lapack/explore-html/ 00007 * 00008 *> \htmlonly 00009 *> Download DLA_LIN_BERR + dependencies 00010 *> <a href="http://www.netlib.org/cgi-bin/netlibfiles.tgz?format=tgz&filename=/lapack/lapack_routine/dla_lin_berr.f"> 00011 *> [TGZ]</a> 00012 *> <a href="http://www.netlib.org/cgi-bin/netlibfiles.zip?format=zip&filename=/lapack/lapack_routine/dla_lin_berr.f"> 00013 *> [ZIP]</a> 00014 *> <a href="http://www.netlib.org/cgi-bin/netlibfiles.txt?format=txt&filename=/lapack/lapack_routine/dla_lin_berr.f"> 00015 *> [TXT]</a> 00016 *> \endhtmlonly 00017 * 00018 * Definition: 00019 * =========== 00020 * 00021 * SUBROUTINE DLA_LIN_BERR ( N, NZ, NRHS, RES, AYB, BERR ) 00022 * 00023 * .. Scalar Arguments .. 00024 * INTEGER N, NZ, NRHS 00025 * .. 00026 * .. Array Arguments .. 00027 * DOUBLE PRECISION AYB( N, NRHS ), BERR( NRHS ) 00028 * DOUBLE PRECISION RES( N, NRHS ) 00029 * .. 00030 * 00031 * 00032 *> \par Purpose: 00033 * ============= 00034 *> 00035 *> \verbatim 00036 *> 00037 *> DLA_LIN_BERR computes component-wise relative backward error from 00038 *> the formula 00039 *> max(i) ( abs(R(i)) / ( abs(op(A_s))*abs(Y) + abs(B_s) )(i) ) 00040 *> where abs(Z) is the component-wise absolute value of the matrix 00041 *> or vector Z. 00042 *> \endverbatim 00043 * 00044 * Arguments: 00045 * ========== 00046 * 00047 *> \param[in] N 00048 *> \verbatim 00049 *> N is INTEGER 00050 *> The number of linear equations, i.e., the order of the 00051 *> matrix A. N >= 0. 00052 *> \endverbatim 00053 *> 00054 *> \param[in] NZ 00055 *> \verbatim 00056 *> NZ is INTEGER 00057 *> We add (NZ+1)*SLAMCH( 'Safe minimum' ) to R(i) in the numerator to 00058 *> guard against spuriously zero residuals. Default value is N. 00059 *> \endverbatim 00060 *> 00061 *> \param[in] NRHS 00062 *> \verbatim 00063 *> NRHS is INTEGER 00064 *> The number of right hand sides, i.e., the number of columns 00065 *> of the matrices AYB, RES, and BERR. NRHS >= 0. 00066 *> \endverbatim 00067 *> 00068 *> \param[in] RES 00069 *> \verbatim 00070 *> RES is DOUBLE PRECISION array, dimension (N,NRHS) 00071 *> The residual matrix, i.e., the matrix R in the relative backward 00072 *> error formula above. 00073 *> \endverbatim 00074 *> 00075 *> \param[in] AYB 00076 *> \verbatim 00077 *> AYB is DOUBLE PRECISION array, dimension (N, NRHS) 00078 *> The denominator in the relative backward error formula above, i.e., 00079 *> the matrix abs(op(A_s))*abs(Y) + abs(B_s). The matrices A, Y, and B 00080 *> are from iterative refinement (see dla_gerfsx_extended.f). 00081 *> \endverbatim 00082 *> 00083 *> \param[out] BERR 00084 *> \verbatim 00085 *> BERR is DOUBLE PRECISION array, dimension (NRHS) 00086 *> The component-wise relative backward error from the formula above. 00087 *> \endverbatim 00088 * 00089 * Authors: 00090 * ======== 00091 * 00092 *> \author Univ. of Tennessee 00093 *> \author Univ. of California Berkeley 00094 *> \author Univ. of Colorado Denver 00095 *> \author NAG Ltd. 00096 * 00097 *> \date November 2011 00098 * 00099 *> \ingroup doubleOTHERcomputational 00100 * 00101 * ===================================================================== 00102 SUBROUTINE DLA_LIN_BERR ( N, NZ, NRHS, RES, AYB, BERR ) 00103 * 00104 * -- LAPACK computational routine (version 3.4.0) -- 00105 * -- LAPACK is a software package provided by Univ. of Tennessee, -- 00106 * -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..-- 00107 * November 2011 00108 * 00109 * .. Scalar Arguments .. 00110 INTEGER N, NZ, NRHS 00111 * .. 00112 * .. Array Arguments .. 00113 DOUBLE PRECISION AYB( N, NRHS ), BERR( NRHS ) 00114 DOUBLE PRECISION RES( N, NRHS ) 00115 * .. 00116 * 00117 * ===================================================================== 00118 * 00119 * .. Local Scalars .. 00120 DOUBLE PRECISION TMP 00121 INTEGER I, J 00122 * .. 00123 * .. Intrinsic Functions .. 00124 INTRINSIC ABS, MAX 00125 * .. 00126 * .. External Functions .. 00127 EXTERNAL DLAMCH 00128 DOUBLE PRECISION DLAMCH 00129 DOUBLE PRECISION SAFE1 00130 * .. 00131 * .. Executable Statements .. 00132 * 00133 * Adding SAFE1 to the numerator guards against spuriously zero 00134 * residuals. A similar safeguard is in the SLA_yyAMV routine used 00135 * to compute AYB. 00136 * 00137 SAFE1 = DLAMCH( 'Safe minimum' ) 00138 SAFE1 = (NZ+1)*SAFE1 00139 00140 DO J = 1, NRHS 00141 BERR(J) = 0.0D+0 00142 DO I = 1, N 00143 IF (AYB(I,J) .NE. 0.0D+0) THEN 00144 TMP = (SAFE1+ABS(RES(I,J)))/AYB(I,J) 00145 BERR(J) = MAX( BERR(J), TMP ) 00146 END IF 00147 * 00148 * If AYB is exactly 0.0 (and if computed by SLA_yyAMV), then we know 00149 * the true residual also must be exactly 0.0. 00150 * 00151 END DO 00152 END DO 00153 END