Debian Codegraph @

Modules/ThirdParty/VNL/src/vxl/v3p/netlib/eispack/otqlrat.f

raw
1 subroutine tqlrat(n,d,e2,ierr)
2 c
3 integer i,j,l,m,n,ii,l1,mml,ierr
4 double precision d(n),e2(n)
5 double precision b,c,f,g,h,p,r,s,t,epslon,pythag
6 c
7 c this subroutine is a translation of the algol procedure tqlrat,
8 c algorithm 464, comm. acm 16, 689(1973) by reinsch.
9 c
10 c this subroutine finds the eigenvalues of a symmetric
11 c tridiagonal matrix by the rational ql method.
12 c
13 c on input
14 c
15 c n is the order of the matrix.
16 c
17 c d contains the diagonal elements of the input matrix.
18 c
19 c e2 contains the squares of the subdiagonal elements of the
20 c input matrix in its last n-1 positions. e2(1) is arbitrary.
21 c
22 c on output
23 c
24 c d contains the eigenvalues in ascending order. if an
25 c error exit is made, the eigenvalues are correct and
26 c ordered for indices 1,2,...ierr-1, but may not be
27 c the smallest eigenvalues.
28 c
29 c e2 has been destroyed.
30 c
31 c ierr is set to
32 c zero for normal return,
33 c j if the j-th eigenvalue has not been
34 c determined after 30 iterations.
35 c
36 c calls pythag for dsqrt(a*a + b*b) .
37 c
38 c questions and comments should be directed to burton s. garbow,
39 c mathematics and computer science div, argonne national laboratory
40 c
41 c this version dated august 1983.
42 c
43 c ------------------------------------------------------------------
44 c
45 ierr = 0
46 if (n .eq. 1) go to 1001
47 c
48 do 100 i = 2, n
49 100 e2(i-1) = e2(i)
50 c
51 f = 0.0d0
52 t = 0.0d0
53 e2(n) = 0.0d0
54 c
55 do 290 l = 1, n
56 j = 0
57 h = dabs(d(l)) + dsqrt(e2(l))
58 if (t .gt. h) go to 105
59 t = h
60 b = epslon(t)
61 c = b * b
62 c .......... look for small squared sub-diagonal element ..........
63 105 do 110 m = l, n
64 if (e2(m) .le. c) go to 120
65 c .......... e2(n) is always zero, so there is no exit
66 c through the bottom of the loop ..........
67 110 continue
68 c
69 120 if (m .eq. l) go to 210
70 130 if (j .eq. 30) go to 1000
71 j = j + 1
72 c .......... form shift ..........
73 l1 = l + 1
74 s = dsqrt(e2(l))
75 g = d(l)
76 p = (d(l1) - g) / (2.0d0 * s)
77 r = pythag(p,1.0d0)
78 d(l) = s / (p + dsign(r,p))
79 h = g - d(l)
80 c
81 do 140 i = l1, n
82 140 d(i) = d(i) - h
83 c
84 f = f + h
85 c .......... rational ql transformation ..........
86 g = d(m)
87 if (g .eq. 0.0d0) g = b
88 h = g
89 s = 0.0d0
90 mml = m - l
91 c .......... for i=m-1 step -1 until l do -- ..........
92 do 200 ii = 1, mml
93 i = m - ii
94 p = g * h
95 r = p + e2(i)
96 e2(i+1) = s * r
97 s = e2(i) / r
98 d(i+1) = h + s * (h + d(i))
99 g = d(i) - e2(i) / g
100 if (g .eq. 0.0d0) g = b
101 h = g * p / r
102 200 continue
103 c
104 e2(l) = s * g
105 d(l) = h
106 c .......... guard against underflow in convergence test ..........
107 if (h .eq. 0.0d0) go to 210
108 if (dabs(e2(l)) .le. dabs(c/h)) go to 210
109 e2(l) = h * e2(l)
110 if (e2(l) .ne. 0.0d0) go to 130
111 210 p = d(l) + f
112 c .......... order eigenvalues ..........
113 if (l .eq. 1) go to 250
114 c .......... for i=l step -1 until 2 do -- ..........
115 do 230 ii = 2, l
116 i = l + 2 - ii
117 if (p .ge. d(i-1)) go to 270
118 d(i) = d(i-1)
119 230 continue
120 c
121 250 i = 1
122 270 d(i) = p
123 290 continue
124 c
125 go to 1001
126 c .......... set error -- no convergence to an
127 c eigenvalue after 30 iterations ..........
128 1000 ierr = l
129 1001 return
130 end