2 SUBROUTINE DGBFA
(ABD
, LDA
, N
, ML
, MU
, IPVT
, INFO
)
3 C***BEGIN PROLOGUE DGBFA
4 C***PURPOSE Factor a band matrix using Gaussian elimination.
6 C***TYPE DOUBLE PRECISION (SGBFA-S, DGBFA-D, CGBFA-C)
7 C***KEYWORDS BANDED, LINEAR ALGEBRA, LINPACK, MATRIX FACTORIZATION
8 C***AUTHOR Moler, C. B., (U. of New Mexico)
11 C DGBFA factors a double precision band matrix by elimination.
13 C DGBFA is usually called by DGBCO, but it can be called
14 C directly with a saving in time if RCOND is not needed.
18 C ABD DOUBLE PRECISION(LDA, N)
19 C contains the matrix in band storage. The columns
20 C of the matrix are stored in the columns of ABD and
21 C the diagonals of the matrix are stored in rows
22 C ML+1 through 2*ML+MU+1 of ABD .
23 C See the comments below for details.
26 C the leading dimension of the array ABD .
27 C LDA must be .GE. 2*ML + MU + 1 .
30 C the order of the original matrix.
33 C number of diagonals below the main diagonal.
37 C number of diagonals above the main diagonal.
39 C More efficient if ML .LE. MU .
42 C ABD an upper triangular matrix in band storage and
43 C the multipliers which were used to obtain it.
44 C The factorization can be written A = L*U where
45 C L is a product of permutation and unit lower
46 C triangular matrices and U is upper triangular.
49 C an integer vector of pivot indices.
53 C = K if U(K,K) .EQ. 0.0 . This is not an error
54 C condition for this subroutine, but it does
55 C indicate that DGBSL will divide by zero if
56 C called. Use RCOND in DGBCO for a reliable
57 C indication of singularity.
61 C If A is a band matrix, the following program segment
62 C will set up the input.
64 C ML = (band width below the diagonal)
65 C MU = (band width above the diagonal)
76 C This uses rows ML+1 through 2*ML+MU+1 of ABD .
77 C In addition, the first ML rows in ABD are used for
78 C elements generated during the triangularization.
79 C The total number of rows needed in ABD is 2*ML+MU+1 .
80 C The ML+MU by ML+MU upper left triangle and the
81 C ML by ML lower right triangle are not referenced.
83 C***REFERENCES J. J. Dongarra, J. R. Bunch, C. B. Moler, and G. W.
84 C Stewart, LINPACK Users' Guide, SIAM, 1979.
85 C***ROUTINES CALLED DAXPY, DSCAL, IDAMAX
86 C***REVISION HISTORY (YYMMDD)
88 C 890531 Changed all specific intrinsics to generic. (WRB)
89 C 890831 Modified array declarations. (WRB)
90 C 890831 REVISION DATE from Version 3.2
91 C 891214 Prologue converted to Version 4.0 format. (BAB)
92 C 900326 Removed duplicate information from DESCRIPTION section.
94 C 920501 Reformatted the REFERENCES section. (WRB)
95 C***END PROLOGUE DGBFA
96 INTEGER LDA
,N
,ML
,MU
,IPVT
(*),INFO
97 DOUBLE PRECISION ABD
(LDA
,*)
100 INTEGER I
,IDAMAX
,I0
,J
,JU
,JZ
,J0
,J1
,K
,KP1
,L
,LM
,M
,MM
,NM1
102 C***FIRST EXECUTABLE STATEMENT DGBFA
106 C ZERO INITIAL FILL-IN COLUMNS
110 IF (J1
.LT
. J0
) GO TO 30
121 C GAUSSIAN ELIMINATION WITH PARTIAL PIVOTING
124 IF (NM1
.LT
. 1) GO TO 130
128 C ZERO NEXT FILL-IN COLUMN
131 IF (JZ
.GT
. N
) GO TO 50
132 IF (ML
.LT
. 1) GO TO 50
138 C FIND L = PIVOT INDEX
141 L
= IDAMAX
(LM
+1,ABD
(M
,K
),1) + M
- 1
144 C ZERO PIVOT IMPLIES THIS COLUMN ALREADY TRIANGULARIZED
146 IF (ABD
(L
,K
) .EQ
. 0.0D0
) GO TO 100
148 C INTERCHANGE IF NECESSARY
150 IF (L
.EQ
. M
) GO TO 60
156 C COMPUTE MULTIPLIERS
159 CALL DSCAL
(LM
,T
,ABD
(M
+1,K
),1)
161 C ROW ELIMINATION WITH COLUMN INDEXING
163 JU
= MIN
(MAX
(JU
,MU
+IPVT
(K
)),N
)
165 IF (JU
.LT
. KP1
) GO TO 90
170 IF (L
.EQ
. MM
) GO TO 70
174 CALL DAXPY
(LM
,T
,ABD
(M
+1,K
),1,ABD
(MM
+1,J
),1)
184 IF (ABD
(M
,N
) .EQ
. 0.0D0
) INFO
= N