1 |
|
2 |
|
3 |
|
4 |
|
5 |
subroutine sposl(a,lda,n,b) |
6 |
integer lda,n |
7 |
real a(lda,1),b(1) |
8 |
c |
9 |
c sposl solves the real symmetric positive definite system |
10 |
c a * x = b |
11 |
c using the factors computed by spoco or spofa. |
12 |
c |
13 |
c on entry |
14 |
c |
15 |
c a real(lda, n) |
16 |
c the output from spoco or spofa. |
17 |
c |
18 |
c lda integer |
19 |
c the leading dimension of the array a . |
20 |
c |
21 |
c n integer |
22 |
c the order of the matrix a . |
23 |
c |
24 |
c b real(n) |
25 |
c the right hand side vector. |
26 |
c |
27 |
c on return |
28 |
c |
29 |
c b the solution vector x . |
30 |
c |
31 |
c error condition |
32 |
c |
33 |
c a division by zero will occur if the input factor contains |
34 |
c a zero on the diagonal. technically this indicates |
35 |
c singularity but it is usually caused by improper subroutine |
36 |
c arguments. it will not occur if the subroutines are called |
37 |
c correctly and info .eq. 0 . |
38 |
c |
39 |
c to compute inverse(a) * c where c is a matrix |
40 |
c with p columns |
41 |
c call spoco(a,lda,n,rcond,z,info) |
42 |
c if (rcond is too small .or. info .ne. 0) go to ... |
43 |
c do 10 j = 1, p |
44 |
c call sposl(a,lda,n,c(1,j)) |
45 |
c 10 continue |
46 |
c |
47 |
c linpack. this version dated 08/14/78 . |
48 |
c cleve moler, university of new mexico, argonne national lab. |
49 |
c |
50 |
c subroutines and functions |
51 |
c |
52 |
c blas saxpy,sdot |
53 |
c |
54 |
c internal variables |
55 |
c |
56 |
real sdot,t |
57 |
integer k,kb |
58 |
c |
59 |
c solve trans(r)*y = b |
60 |
c |
61 |
do 10 k = 1, n |
62 |
t = sdot(k-1,a(1,k),1,b(1),1) |
63 |
b(k) = (b(k) - t)/a(k,k) |
64 |
10 continue |
65 |
c |
66 |
c solve r*x = y |
67 |
c |
68 |
do 20 kb = 1, n |
69 |
k = n + 1 - kb |
70 |
b(k) = b(k)/a(k,k) |
71 |
t = -b(k) |
72 |
call saxpy(k-1,t,a(1,k),1,b(1),1) |
73 |
20 continue |
74 |
return |
75 |
end |