| 1 |
SUBROUTINE DGGHRD( COMPQ, COMPZ, N, ILO, IHI, A, LDA, B, LDB, Q, |
| 2 |
$ LDQ, Z, LDZ, INFO ) |
| 3 |
* |
| 4 |
* -- LAPACK routine (version 3.0) -- |
| 5 |
* Univ. of Tennessee, Univ. of California Berkeley, NAG Ltd., |
| 6 |
* Courant Institute, Argonne National Lab, and Rice University |
| 7 |
* September 30, 1994 |
| 8 |
* |
| 9 |
* .. Scalar Arguments .. |
| 10 |
CHARACTER COMPQ, COMPZ |
| 11 |
INTEGER IHI, ILO, INFO, LDA, LDB, LDQ, LDZ, N |
| 12 |
* .. |
| 13 |
* .. Array Arguments .. |
| 14 |
DOUBLE PRECISION A( LDA, * ), B( LDB, * ), Q( LDQ, * ), |
| 15 |
$ Z( LDZ, * ) |
| 16 |
* .. |
| 17 |
* |
| 18 |
* Purpose |
| 19 |
* ======= |
| 20 |
* |
| 21 |
* DGGHRD reduces a pair of real matrices (A,B) to generalized upper |
| 22 |
* Hessenberg form using orthogonal transformations, where A is a |
| 23 |
* general matrix and B is upper triangular: Q' * A * Z = H and |
| 24 |
* Q' * B * Z = T, where H is upper Hessenberg, T is upper triangular, |
| 25 |
* and Q and Z are orthogonal, and ' means transpose. |
| 26 |
* |
| 27 |
* The orthogonal matrices Q and Z are determined as products of Givens |
| 28 |
* rotations. They may either be formed explicitly, or they may be |
| 29 |
* postmultiplied into input matrices Q1 and Z1, so that |
| 30 |
* |
| 31 |
* Q1 * A * Z1' = (Q1*Q) * H * (Z1*Z)' |
| 32 |
* Q1 * B * Z1' = (Q1*Q) * T * (Z1*Z)' |
| 33 |
* |
| 34 |
* Arguments |
| 35 |
* ========= |
| 36 |
* |
| 37 |
* COMPQ (input) CHARACTER*1 |
| 38 |
* = 'N': do not compute Q; |
| 39 |
* = 'I': Q is initialized to the unit matrix, and the |
| 40 |
* orthogonal matrix Q is returned; |
| 41 |
* = 'V': Q must contain an orthogonal matrix Q1 on entry, |
| 42 |
* and the product Q1*Q is returned. |
| 43 |
* |
| 44 |
* COMPZ (input) CHARACTER*1 |
| 45 |
* = 'N': do not compute Z; |
| 46 |
* = 'I': Z is initialized to the unit matrix, and the |
| 47 |
* orthogonal matrix Z is returned; |
| 48 |
* = 'V': Z must contain an orthogonal matrix Z1 on entry, |
| 49 |
* and the product Z1*Z is returned. |
| 50 |
* |
| 51 |
* N (input) INTEGER |
| 52 |
* The order of the matrices A and B. N >= 0. |
| 53 |
* |
| 54 |
* ILO (input) INTEGER |
| 55 |
* IHI (input) INTEGER |
| 56 |
* It is assumed that A is already upper triangular in rows and |
| 57 |
* columns 1:ILO-1 and IHI+1:N. ILO and IHI are normally set |
| 58 |
* by a previous call to DGGBAL; otherwise they should be set |
| 59 |
* to 1 and N respectively. |
| 60 |
* 1 <= ILO <= IHI <= N, if N > 0; ILO=1 and IHI=0, if N=0. |
| 61 |
* |
| 62 |
* A (input/output) DOUBLE PRECISION array, dimension (LDA, N) |
| 63 |
* On entry, the N-by-N general matrix to be reduced. |
| 64 |
* On exit, the upper triangle and the first subdiagonal of A |
| 65 |
* are overwritten with the upper Hessenberg matrix H, and the |
| 66 |
* rest is set to zero. |
| 67 |
* |
| 68 |
* LDA (input) INTEGER |
| 69 |
* The leading dimension of the array A. LDA >= max(1,N). |
| 70 |
* |
| 71 |
* B (input/output) DOUBLE PRECISION array, dimension (LDB, N) |
| 72 |
* On entry, the N-by-N upper triangular matrix B. |
| 73 |
* On exit, the upper triangular matrix T = Q' B Z. The |
| 74 |
* elements below the diagonal are set to zero. |
| 75 |
* |
| 76 |
* LDB (input) INTEGER |
| 77 |
* The leading dimension of the array B. LDB >= max(1,N). |
| 78 |
* |
| 79 |
* Q (input/output) DOUBLE PRECISION array, dimension (LDQ, N) |
| 80 |
* If COMPQ='N': Q is not referenced. |
| 81 |
* If COMPQ='I': on entry, Q need not be set, and on exit it |
| 82 |
* contains the orthogonal matrix Q, where Q' |
| 83 |
* is the product of the Givens transformations |
| 84 |
* which are applied to A and B on the left. |
| 85 |
* If COMPQ='V': on entry, Q must contain an orthogonal matrix |
| 86 |
* Q1, and on exit this is overwritten by Q1*Q. |
| 87 |
* |
| 88 |
* LDQ (input) INTEGER |
| 89 |
* The leading dimension of the array Q. |
| 90 |
* LDQ >= N if COMPQ='V' or 'I'; LDQ >= 1 otherwise. |
| 91 |
* |
| 92 |
* Z (input/output) DOUBLE PRECISION array, dimension (LDZ, N) |
| 93 |
* If COMPZ='N': Z is not referenced. |
| 94 |
* If COMPZ='I': on entry, Z need not be set, and on exit it |
| 95 |
* contains the orthogonal matrix Z, which is |
| 96 |
* the product of the Givens transformations |
| 97 |
* which are applied to A and B on the right. |
| 98 |
* If COMPZ='V': on entry, Z must contain an orthogonal matrix |
| 99 |
* Z1, and on exit this is overwritten by Z1*Z. |
| 100 |
* |
| 101 |
* LDZ (input) INTEGER |
| 102 |
* The leading dimension of the array Z. |
| 103 |
* LDZ >= N if COMPZ='V' or 'I'; LDZ >= 1 otherwise. |
| 104 |
* |
| 105 |
* INFO (output) INTEGER |
| 106 |
* = 0: successful exit. |
| 107 |
* < 0: if INFO = -i, the i-th argument had an illegal value. |
| 108 |
* |
| 109 |
* Further Details |
| 110 |
* =============== |
| 111 |
* |
| 112 |
* This routine reduces A to Hessenberg and B to triangular form by |
| 113 |
* an unblocked reduction, as described in _Matrix_Computations_, |
| 114 |
* by Golub and Van Loan (Johns Hopkins Press.) |
| 115 |
* |
| 116 |
* ===================================================================== |
| 117 |
* |
| 118 |
* .. Parameters .. |
| 119 |
DOUBLE PRECISION ONE, ZERO |
| 120 |
PARAMETER ( ONE = 1.0D+0, ZERO = 0.0D+0 ) |
| 121 |
* .. |
| 122 |
* .. Local Scalars .. |
| 123 |
LOGICAL ILQ, ILZ |
| 124 |
INTEGER ICOMPQ, ICOMPZ, JCOL, JROW |
| 125 |
DOUBLE PRECISION C, S, TEMP |
| 126 |
* .. |
| 127 |
* .. External Functions .. |
| 128 |
LOGICAL LSAME |
| 129 |
EXTERNAL LSAME |
| 130 |
* .. |
| 131 |
* .. External Subroutines .. |
| 132 |
EXTERNAL DLARTG, DLASET, DROT, XERBLA |
| 133 |
* .. |
| 134 |
* .. Intrinsic Functions .. |
| 135 |
INTRINSIC MAX |
| 136 |
* .. |
| 137 |
* .. Executable Statements .. |
| 138 |
* |
| 139 |
* Decode COMPQ |
| 140 |
* |
| 141 |
IF( LSAME( COMPQ, 'N' ) ) THEN |
| 142 |
ILQ = .FALSE. |
| 143 |
ICOMPQ = 1 |
| 144 |
ELSE IF( LSAME( COMPQ, 'V' ) ) THEN |
| 145 |
ILQ = .TRUE. |
| 146 |
ICOMPQ = 2 |
| 147 |
ELSE IF( LSAME( COMPQ, 'I' ) ) THEN |
| 148 |
ILQ = .TRUE. |
| 149 |
ICOMPQ = 3 |
| 150 |
ELSE |
| 151 |
ICOMPQ = 0 |
| 152 |
END IF |
| 153 |
* |
| 154 |
* Decode COMPZ |
| 155 |
* |
| 156 |
IF( LSAME( COMPZ, 'N' ) ) THEN |
| 157 |
ILZ = .FALSE. |
| 158 |
ICOMPZ = 1 |
| 159 |
ELSE IF( LSAME( COMPZ, 'V' ) ) THEN |
| 160 |
ILZ = .TRUE. |
| 161 |
ICOMPZ = 2 |
| 162 |
ELSE IF( LSAME( COMPZ, 'I' ) ) THEN |
| 163 |
ILZ = .TRUE. |
| 164 |
ICOMPZ = 3 |
| 165 |
ELSE |
| 166 |
ICOMPZ = 0 |
| 167 |
END IF |
| 168 |
* |
| 169 |
* Test the input parameters. |
| 170 |
* |
| 171 |
INFO = 0 |
| 172 |
IF( ICOMPQ.LE.0 ) THEN |
| 173 |
INFO = -1 |
| 174 |
ELSE IF( ICOMPZ.LE.0 ) THEN |
| 175 |
INFO = -2 |
| 176 |
ELSE IF( N.LT.0 ) THEN |
| 177 |
INFO = -3 |
| 178 |
ELSE IF( ILO.LT.1 ) THEN |
| 179 |
INFO = -4 |
| 180 |
ELSE IF( IHI.GT.N .OR. IHI.LT.ILO-1 ) THEN |
| 181 |
INFO = -5 |
| 182 |
ELSE IF( LDA.LT.MAX( 1, N ) ) THEN |
| 183 |
INFO = -7 |
| 184 |
ELSE IF( LDB.LT.MAX( 1, N ) ) THEN |
| 185 |
INFO = -9 |
| 186 |
ELSE IF( ( ILQ .AND. LDQ.LT.N ) .OR. LDQ.LT.1 ) THEN |
| 187 |
INFO = -11 |
| 188 |
ELSE IF( ( ILZ .AND. LDZ.LT.N ) .OR. LDZ.LT.1 ) THEN |
| 189 |
INFO = -13 |
| 190 |
END IF |
| 191 |
IF( INFO.NE.0 ) THEN |
| 192 |
CALL XERBLA( 'DGGHRD', -INFO ) |
| 193 |
RETURN |
| 194 |
END IF |
| 195 |
* |
| 196 |
* Initialize Q and Z if desired. |
| 197 |
* |
| 198 |
IF( ICOMPQ.EQ.3 ) |
| 199 |
$ CALL DLASET( 'Full', N, N, ZERO, ONE, Q, LDQ ) |
| 200 |
IF( ICOMPZ.EQ.3 ) |
| 201 |
$ CALL DLASET( 'Full', N, N, ZERO, ONE, Z, LDZ ) |
| 202 |
* |
| 203 |
* Quick return if possible |
| 204 |
* |
| 205 |
IF( N.LE.1 ) |
| 206 |
$ RETURN |
| 207 |
* |
| 208 |
* Zero out lower triangle of B |
| 209 |
* |
| 210 |
DO 20 JCOL = 1, N - 1 |
| 211 |
DO 10 JROW = JCOL + 1, N |
| 212 |
B( JROW, JCOL ) = ZERO |
| 213 |
10 CONTINUE |
| 214 |
20 CONTINUE |
| 215 |
* |
| 216 |
* Reduce A and B |
| 217 |
* |
| 218 |
DO 40 JCOL = ILO, IHI - 2 |
| 219 |
* |
| 220 |
DO 30 JROW = IHI, JCOL + 2, -1 |
| 221 |
* |
| 222 |
* Step 1: rotate rows JROW-1, JROW to kill A(JROW,JCOL) |
| 223 |
* |
| 224 |
TEMP = A( JROW-1, JCOL ) |
| 225 |
CALL DLARTG( TEMP, A( JROW, JCOL ), C, S, |
| 226 |
$ A( JROW-1, JCOL ) ) |
| 227 |
A( JROW, JCOL ) = ZERO |
| 228 |
CALL DROT( N-JCOL, A( JROW-1, JCOL+1 ), LDA, |
| 229 |
$ A( JROW, JCOL+1 ), LDA, C, S ) |
| 230 |
CALL DROT( N+2-JROW, B( JROW-1, JROW-1 ), LDB, |
| 231 |
$ B( JROW, JROW-1 ), LDB, C, S ) |
| 232 |
IF( ILQ ) |
| 233 |
$ CALL DROT( N, Q( 1, JROW-1 ), 1, Q( 1, JROW ), 1, C, S ) |
| 234 |
* |
| 235 |
* Step 2: rotate columns JROW, JROW-1 to kill B(JROW,JROW-1) |
| 236 |
* |
| 237 |
TEMP = B( JROW, JROW ) |
| 238 |
CALL DLARTG( TEMP, B( JROW, JROW-1 ), C, S, |
| 239 |
$ B( JROW, JROW ) ) |
| 240 |
B( JROW, JROW-1 ) = ZERO |
| 241 |
CALL DROT( IHI, A( 1, JROW ), 1, A( 1, JROW-1 ), 1, C, S ) |
| 242 |
CALL DROT( JROW-1, B( 1, JROW ), 1, B( 1, JROW-1 ), 1, C, |
| 243 |
$ S ) |
| 244 |
IF( ILZ ) |
| 245 |
$ CALL DROT( N, Z( 1, JROW ), 1, Z( 1, JROW-1 ), 1, C, S ) |
| 246 |
30 CONTINUE |
| 247 |
40 CONTINUE |
| 248 |
* |
| 249 |
RETURN |
| 250 |
* |
| 251 |
* End of DGGHRD |
| 252 |
* |
| 253 |
END |