LAPACK 3.3.0
|
00001 SUBROUTINE DSYT22( ITYPE, UPLO, N, M, KBAND, A, LDA, D, E, U, LDU, 00002 $ V, LDV, TAU, WORK, RESULT ) 00003 * 00004 * -- LAPACK test routine (version 3.1) -- 00005 * Univ. of Tennessee, Univ. of California Berkeley and NAG Ltd.. 00006 * November 2006 00007 * 00008 * .. Scalar Arguments .. 00009 CHARACTER UPLO 00010 INTEGER ITYPE, KBAND, LDA, LDU, LDV, M, N 00011 * .. 00012 * .. Array Arguments .. 00013 DOUBLE PRECISION A( LDA, * ), D( * ), E( * ), RESULT( 2 ), 00014 $ TAU( * ), U( LDU, * ), V( LDV, * ), WORK( * ) 00015 * .. 00016 * 00017 * Purpose 00018 * ======= 00019 * 00020 * DSYT22 generally checks a decomposition of the form 00021 * 00022 * A U = U S 00023 * 00024 * where A is symmetric, the columns of U are orthonormal, and S 00025 * is diagonal (if KBAND=0) or symmetric tridiagonal (if 00026 * KBAND=1). If ITYPE=1, then U is represented as a dense matrix, 00027 * otherwise the U is expressed as a product of Householder 00028 * transformations, whose vectors are stored in the array "V" and 00029 * whose scaling constants are in "TAU"; we shall use the letter 00030 * "V" to refer to the product of Householder transformations 00031 * (which should be equal to U). 00032 * 00033 * Specifically, if ITYPE=1, then: 00034 * 00035 * RESULT(1) = | U' A U - S | / ( |A| m ulp ) *and* 00036 * RESULT(2) = | I - U'U | / ( m ulp ) 00037 * 00038 * Arguments 00039 * ========= 00040 * 00041 * ITYPE INTEGER 00042 * Specifies the type of tests to be performed. 00043 * 1: U expressed as a dense orthogonal matrix: 00044 * RESULT(1) = | A - U S U' | / ( |A| n ulp ) *and* 00045 * RESULT(2) = | I - UU' | / ( n ulp ) 00046 * 00047 * UPLO CHARACTER 00048 * If UPLO='U', the upper triangle of A will be used and the 00049 * (strictly) lower triangle will not be referenced. If 00050 * UPLO='L', the lower triangle of A will be used and the 00051 * (strictly) upper triangle will not be referenced. 00052 * Not modified. 00053 * 00054 * N INTEGER 00055 * The size of the matrix. If it is zero, DSYT22 does nothing. 00056 * It must be at least zero. 00057 * Not modified. 00058 * 00059 * M INTEGER 00060 * The number of columns of U. If it is zero, DSYT22 does 00061 * nothing. It must be at least zero. 00062 * Not modified. 00063 * 00064 * KBAND INTEGER 00065 * The bandwidth of the matrix. It may only be zero or one. 00066 * If zero, then S is diagonal, and E is not referenced. If 00067 * one, then S is symmetric tri-diagonal. 00068 * Not modified. 00069 * 00070 * A DOUBLE PRECISION array, dimension (LDA , N) 00071 * The original (unfactored) matrix. It is assumed to be 00072 * symmetric, and only the upper (UPLO='U') or only the lower 00073 * (UPLO='L') will be referenced. 00074 * Not modified. 00075 * 00076 * LDA INTEGER 00077 * The leading dimension of A. It must be at least 1 00078 * and at least N. 00079 * Not modified. 00080 * 00081 * D DOUBLE PRECISION array, dimension (N) 00082 * The diagonal of the (symmetric tri-) diagonal matrix. 00083 * Not modified. 00084 * 00085 * E DOUBLE PRECISION array, dimension (N) 00086 * The off-diagonal of the (symmetric tri-) diagonal matrix. 00087 * E(1) is ignored, E(2) is the (1,2) and (2,1) element, etc. 00088 * Not referenced if KBAND=0. 00089 * Not modified. 00090 * 00091 * U DOUBLE PRECISION array, dimension (LDU, N) 00092 * If ITYPE=1 or 3, this contains the orthogonal matrix in 00093 * the decomposition, expressed as a dense matrix. If ITYPE=2, 00094 * then it is not referenced. 00095 * Not modified. 00096 * 00097 * LDU INTEGER 00098 * The leading dimension of U. LDU must be at least N and 00099 * at least 1. 00100 * Not modified. 00101 * 00102 * V DOUBLE PRECISION array, dimension (LDV, N) 00103 * If ITYPE=2 or 3, the lower triangle of this array contains 00104 * the Householder vectors used to describe the orthogonal 00105 * matrix in the decomposition. If ITYPE=1, then it is not 00106 * referenced. 00107 * Not modified. 00108 * 00109 * LDV INTEGER 00110 * The leading dimension of V. LDV must be at least N and 00111 * at least 1. 00112 * Not modified. 00113 * 00114 * TAU DOUBLE PRECISION array, dimension (N) 00115 * If ITYPE >= 2, then TAU(j) is the scalar factor of 00116 * v(j) v(j)' in the Householder transformation H(j) of 00117 * the product U = H(1)...H(n-2) 00118 * If ITYPE < 2, then TAU is not referenced. 00119 * Not modified. 00120 * 00121 * WORK DOUBLE PRECISION array, dimension (2*N**2) 00122 * Workspace. 00123 * Modified. 00124 * 00125 * RESULT DOUBLE PRECISION array, dimension (2) 00126 * The values computed by the two tests described above. The 00127 * values are currently limited to 1/ulp, to avoid overflow. 00128 * RESULT(1) is always modified. RESULT(2) is modified only 00129 * if LDU is at least N. 00130 * Modified. 00131 * 00132 * ===================================================================== 00133 * 00134 * .. Parameters .. 00135 DOUBLE PRECISION ZERO, ONE 00136 PARAMETER ( ZERO = 0.0D0, ONE = 1.0D0 ) 00137 * .. 00138 * .. Local Scalars .. 00139 INTEGER J, JJ, JJ1, JJ2, NN, NNP1 00140 DOUBLE PRECISION ANORM, ULP, UNFL, WNORM 00141 * .. 00142 * .. External Functions .. 00143 DOUBLE PRECISION DLAMCH, DLANSY 00144 EXTERNAL DLAMCH, DLANSY 00145 * .. 00146 * .. External Subroutines .. 00147 EXTERNAL DGEMM, DORT01, DSYMM 00148 * .. 00149 * .. Intrinsic Functions .. 00150 INTRINSIC DBLE, MAX, MIN 00151 * .. 00152 * .. Executable Statements .. 00153 * 00154 RESULT( 1 ) = ZERO 00155 RESULT( 2 ) = ZERO 00156 IF( N.LE.0 .OR. M.LE.0 ) 00157 $ RETURN 00158 * 00159 UNFL = DLAMCH( 'Safe minimum' ) 00160 ULP = DLAMCH( 'Precision' ) 00161 * 00162 * Do Test 1 00163 * 00164 * Norm of A: 00165 * 00166 ANORM = MAX( DLANSY( '1', UPLO, N, A, LDA, WORK ), UNFL ) 00167 * 00168 * Compute error matrix: 00169 * 00170 * ITYPE=1: error = U' A U - S 00171 * 00172 CALL DSYMM( 'L', UPLO, N, M, ONE, A, LDA, U, LDU, ZERO, WORK, N ) 00173 NN = N*N 00174 NNP1 = NN + 1 00175 CALL DGEMM( 'T', 'N', M, M, N, ONE, U, LDU, WORK, N, ZERO, 00176 $ WORK( NNP1 ), N ) 00177 DO 10 J = 1, M 00178 JJ = NN + ( J-1 )*N + J 00179 WORK( JJ ) = WORK( JJ ) - D( J ) 00180 10 CONTINUE 00181 IF( KBAND.EQ.1 .AND. N.GT.1 ) THEN 00182 DO 20 J = 2, M 00183 JJ1 = NN + ( J-1 )*N + J - 1 00184 JJ2 = NN + ( J-2 )*N + J 00185 WORK( JJ1 ) = WORK( JJ1 ) - E( J-1 ) 00186 WORK( JJ2 ) = WORK( JJ2 ) - E( J-1 ) 00187 20 CONTINUE 00188 END IF 00189 WNORM = DLANSY( '1', UPLO, M, WORK( NNP1 ), N, WORK( 1 ) ) 00190 * 00191 IF( ANORM.GT.WNORM ) THEN 00192 RESULT( 1 ) = ( WNORM / ANORM ) / ( M*ULP ) 00193 ELSE 00194 IF( ANORM.LT.ONE ) THEN 00195 RESULT( 1 ) = ( MIN( WNORM, M*ANORM ) / ANORM ) / ( M*ULP ) 00196 ELSE 00197 RESULT( 1 ) = MIN( WNORM / ANORM, DBLE( M ) ) / ( M*ULP ) 00198 END IF 00199 END IF 00200 * 00201 * Do Test 2 00202 * 00203 * Compute U'U - I 00204 * 00205 IF( ITYPE.EQ.1 ) 00206 $ CALL DORT01( 'Columns', N, M, U, LDU, WORK, 2*N*N, 00207 $ RESULT( 2 ) ) 00208 * 00209 RETURN 00210 * 00211 * End of DSYT22 00212 * 00213 END