1 |
149 |
jeremybenn |
c { dg-do compile }
|
2 |
|
|
C To: egcs-bugs@cygnus.com
|
3 |
|
|
C Subject: -fPIC problem showing up with fortran on x86
|
4 |
|
|
C From: Dave Love <d.love@dl.ac.uk>
|
5 |
|
|
C Date: 19 Dec 1997 19:31:41 +0000
|
6 |
|
|
C
|
7 |
|
|
C
|
8 |
|
|
C This illustrates a long-standing problem noted at the end of the g77
|
9 |
|
|
C `Actual Bugs' info node and thought to be in the back end. Although
|
10 |
|
|
C the report is against gcc 2.7 I can reproduce it (specifically on
|
11 |
|
|
C redhat 4.2) with the 971216 egcs snapshot.
|
12 |
|
|
C
|
13 |
|
|
C g77 version 0.5.21
|
14 |
|
|
C gcc -v -fnull-version -o /tmp/gfa00415 -xf77-cpp-input /tmp/gfa00415.f -xnone
|
15 |
|
|
C -lf2c -lm
|
16 |
|
|
C
|
17 |
|
|
|
18 |
|
|
C ------------
|
19 |
|
|
subroutine dqage(f,a,b,epsabs,epsrel,limit,result,abserr,
|
20 |
|
|
* neval,ier,alist,blist,rlist,elist,iord,last)
|
21 |
|
|
C --------------------------------------------------
|
22 |
|
|
C
|
23 |
|
|
C Modified Feb 1989 by Barry W. Brown to eliminate key
|
24 |
|
|
C as argument (use key=1) and to eliminate all Fortran
|
25 |
|
|
C output.
|
26 |
|
|
C
|
27 |
|
|
C Purpose: to make this routine usable from within S.
|
28 |
|
|
C
|
29 |
|
|
C --------------------------------------------------
|
30 |
|
|
c***begin prologue dqage
|
31 |
|
|
c***date written 800101 (yymmdd)
|
32 |
|
|
c***revision date 830518 (yymmdd)
|
33 |
|
|
c***category no. h2a1a1
|
34 |
|
|
c***keywords automatic integrator, general-purpose,
|
35 |
|
|
c integrand examinator, globally adaptive,
|
36 |
|
|
c gauss-kronrod
|
37 |
|
|
c***author piessens,robert,appl. math. & progr. div. - k.u.leuven
|
38 |
|
|
c de doncker,elise,appl. math. & progr. div. - k.u.leuven
|
39 |
|
|
c***purpose the routine calculates an approximation result to a given
|
40 |
|
|
c definite integral i = integral of f over (a,b),
|
41 |
|
|
c hopefully satisfying following claim for accuracy
|
42 |
|
|
c abs(i-reslt).le.max(epsabs,epsrel*abs(i)).
|
43 |
|
|
c***description
|
44 |
|
|
c
|
45 |
|
|
c computation of a definite integral
|
46 |
|
|
c standard fortran subroutine
|
47 |
|
|
c double precision version
|
48 |
|
|
c
|
49 |
|
|
c parameters
|
50 |
|
|
c on entry
|
51 |
|
|
c f - double precision
|
52 |
|
|
c function subprogram defining the integrand
|
53 |
|
|
c function f(x). the actual name for f needs to be
|
54 |
|
|
c declared e x t e r n a l in the driver program.
|
55 |
|
|
c
|
56 |
|
|
c a - double precision
|
57 |
|
|
c lower limit of integration
|
58 |
|
|
c
|
59 |
|
|
c b - double precision
|
60 |
|
|
c upper limit of integration
|
61 |
|
|
c
|
62 |
|
|
c epsabs - double precision
|
63 |
|
|
c absolute accuracy requested
|
64 |
|
|
c epsrel - double precision
|
65 |
|
|
c relative accuracy requested
|
66 |
|
|
c if epsabs.le.0
|
67 |
|
|
c and epsrel.lt.max(50*rel.mach.acc.,0.5d-28),
|
68 |
|
|
c the routine will end with ier = 6.
|
69 |
|
|
c
|
70 |
|
|
c key - integer
|
71 |
|
|
c key for choice of local integration rule
|
72 |
|
|
c a gauss-kronrod pair is used with
|
73 |
|
|
c 7 - 15 points if key.lt.2,
|
74 |
|
|
c 10 - 21 points if key = 2,
|
75 |
|
|
c 15 - 31 points if key = 3,
|
76 |
|
|
c 20 - 41 points if key = 4,
|
77 |
|
|
c 25 - 51 points if key = 5,
|
78 |
|
|
c 30 - 61 points if key.gt.5.
|
79 |
|
|
c
|
80 |
|
|
c limit - integer
|
81 |
|
|
c gives an upperbound on the number of subintervals
|
82 |
|
|
c in the partition of (a,b), limit.ge.1.
|
83 |
|
|
c
|
84 |
|
|
c on return
|
85 |
|
|
c result - double precision
|
86 |
|
|
c approximation to the integral
|
87 |
|
|
c
|
88 |
|
|
c abserr - double precision
|
89 |
|
|
c estimate of the modulus of the absolute error,
|
90 |
|
|
c which should equal or exceed abs(i-result)
|
91 |
|
|
c
|
92 |
|
|
c neval - integer
|
93 |
|
|
c number of integrand evaluations
|
94 |
|
|
c
|
95 |
|
|
c ier - integer
|
96 |
|
|
c ier = 0 normal and reliable termination of the
|
97 |
|
|
c routine. it is assumed that the requested
|
98 |
|
|
c accuracy has been achieved.
|
99 |
|
|
c ier.gt.0 abnormal termination of the routine
|
100 |
|
|
c the estimates for result and error are
|
101 |
|
|
c less reliable. it is assumed that the
|
102 |
|
|
c requested accuracy has not been achieved.
|
103 |
|
|
c error messages
|
104 |
|
|
c ier = 1 maximum number of subdivisions allowed
|
105 |
|
|
c has been achieved. one can allow more
|
106 |
|
|
c subdivisions by increasing the value
|
107 |
|
|
c of limit.
|
108 |
|
|
c however, if this yields no improvement it
|
109 |
|
|
c is rather advised to analyze the integrand
|
110 |
|
|
c in order to determine the integration
|
111 |
|
|
c difficulties. if the position of a local
|
112 |
|
|
c difficulty can be determined(e.g.
|
113 |
|
|
c singularity, discontinuity within the
|
114 |
|
|
c interval) one will probably gain from
|
115 |
|
|
c splitting up the interval at this point
|
116 |
|
|
c and calling the integrator on the
|
117 |
|
|
c subranges. if possible, an appropriate
|
118 |
|
|
c special-purpose integrator should be used
|
119 |
|
|
c which is designed for handling the type of
|
120 |
|
|
c difficulty involved.
|
121 |
|
|
c = 2 the occurrence of roundoff error is
|
122 |
|
|
c detected, which prevents the requested
|
123 |
|
|
c tolerance from being achieved.
|
124 |
|
|
c = 3 extremely bad integrand behavior occurs
|
125 |
|
|
c at some points of the integration
|
126 |
|
|
c interval.
|
127 |
|
|
c = 6 the input is invalid, because
|
128 |
|
|
c (epsabs.le.0 and
|
129 |
|
|
c epsrel.lt.max(50*rel.mach.acc.,0.5d-28),
|
130 |
|
|
c result, abserr, neval, last, rlist(1) ,
|
131 |
|
|
c elist(1) and iord(1) are set to zero.
|
132 |
|
|
c alist(1) and blist(1) are set to a and b
|
133 |
|
|
c respectively.
|
134 |
|
|
c
|
135 |
|
|
c alist - double precision
|
136 |
|
|
c vector of dimension at least limit, the first
|
137 |
|
|
c last elements of which are the left
|
138 |
|
|
c end points of the subintervals in the partition
|
139 |
|
|
c of the given integration range (a,b)
|
140 |
|
|
c
|
141 |
|
|
c blist - double precision
|
142 |
|
|
c vector of dimension at least limit, the first
|
143 |
|
|
c last elements of which are the right
|
144 |
|
|
c end points of the subintervals in the partition
|
145 |
|
|
c of the given integration range (a,b)
|
146 |
|
|
c
|
147 |
|
|
c rlist - double precision
|
148 |
|
|
c vector of dimension at least limit, the first
|
149 |
|
|
c last elements of which are the
|
150 |
|
|
c integral approximations on the subintervals
|
151 |
|
|
c
|
152 |
|
|
c elist - double precision
|
153 |
|
|
c vector of dimension at least limit, the first
|
154 |
|
|
c last elements of which are the moduli of the
|
155 |
|
|
c absolute error estimates on the subintervals
|
156 |
|
|
c
|
157 |
|
|
c iord - integer
|
158 |
|
|
c vector of dimension at least limit, the first k
|
159 |
|
|
c elements of which are pointers to the
|
160 |
|
|
c error estimates over the subintervals,
|
161 |
|
|
c such that elist(iord(1)), ...,
|
162 |
|
|
c elist(iord(k)) form a decreasing sequence,
|
163 |
|
|
c with k = last if last.le.(limit/2+2), and
|
164 |
|
|
c k = limit+1-last otherwise
|
165 |
|
|
c
|
166 |
|
|
c last - integer
|
167 |
|
|
c number of subintervals actually produced in the
|
168 |
|
|
c subdivision process
|
169 |
|
|
c
|
170 |
|
|
c***references (none)
|
171 |
|
|
c***routines called d1mach,dqk15,dqk21,dqk31,
|
172 |
|
|
c dqk41,dqk51,dqk61,dqpsrt
|
173 |
|
|
c***end prologue dqage
|
174 |
|
|
c
|
175 |
|
|
double precision a,abserr,alist,area,area1,area12,area2,a1,a2,b,
|
176 |
|
|
* blist,b1,b2,dabs,defabs,defab1,defab2,dmax1,d1mach,elist,epmach,
|
177 |
|
|
* epsabs,epsrel,errbnd,errmax,error1,error2,erro12,errsum,f,
|
178 |
|
|
* resabs,result,rlist,uflow
|
179 |
|
|
integer ier,iord,iroff1,iroff2,k,last,limit,maxerr,neval,
|
180 |
|
|
* nrmax
|
181 |
|
|
c
|
182 |
|
|
dimension alist(limit),blist(limit),elist(limit),iord(limit),
|
183 |
|
|
* rlist(limit)
|
184 |
|
|
c
|
185 |
|
|
external f
|
186 |
|
|
c
|
187 |
|
|
c list of major variables
|
188 |
|
|
c -----------------------
|
189 |
|
|
c
|
190 |
|
|
c alist - list of left end points of all subintervals
|
191 |
|
|
c considered up to now
|
192 |
|
|
c blist - list of right end points of all subintervals
|
193 |
|
|
c considered up to now
|
194 |
|
|
c rlist(i) - approximation to the integral over
|
195 |
|
|
c (alist(i),blist(i))
|
196 |
|
|
c elist(i) - error estimate applying to rlist(i)
|
197 |
|
|
c maxerr - pointer to the interval with largest
|
198 |
|
|
c error estimate
|
199 |
|
|
c errmax - elist(maxerr)
|
200 |
|
|
c area - sum of the integrals over the subintervals
|
201 |
|
|
c errsum - sum of the errors over the subintervals
|
202 |
|
|
c errbnd - requested accuracy max(epsabs,epsrel*
|
203 |
|
|
c abs(result))
|
204 |
|
|
c *****1 - variable for the left subinterval
|
205 |
|
|
c *****2 - variable for the right subinterval
|
206 |
|
|
c last - index for subdivision
|
207 |
|
|
c
|
208 |
|
|
c
|
209 |
|
|
c machine dependent constants
|
210 |
|
|
c ---------------------------
|
211 |
|
|
c
|
212 |
|
|
c epmach is the largest relative spacing.
|
213 |
|
|
c uflow is the smallest positive magnitude.
|
214 |
|
|
c
|
215 |
|
|
c***first executable statement dqage
|
216 |
|
|
epmach = d1mach(4)
|
217 |
|
|
uflow = d1mach(1)
|
218 |
|
|
c
|
219 |
|
|
c test on validity of parameters
|
220 |
|
|
c ------------------------------
|
221 |
|
|
c
|
222 |
|
|
ier = 0
|
223 |
|
|
neval = 0
|
224 |
|
|
last = 0
|
225 |
|
|
result = 0.0d+00
|
226 |
|
|
abserr = 0.0d+00
|
227 |
|
|
alist(1) = a
|
228 |
|
|
blist(1) = b
|
229 |
|
|
rlist(1) = 0.0d+00
|
230 |
|
|
elist(1) = 0.0d+00
|
231 |
|
|
iord(1) = 0
|
232 |
|
|
if(epsabs.le.0.0d+00.and.
|
233 |
|
|
* epsrel.lt.dmax1(0.5d+02*epmach,0.5d-28)) ier = 6
|
234 |
|
|
if(ier.eq.6) go to 999
|
235 |
|
|
c
|
236 |
|
|
c first approximation to the integral
|
237 |
|
|
c -----------------------------------
|
238 |
|
|
c
|
239 |
|
|
neval = 0
|
240 |
|
|
call dqk15(f,a,b,result,abserr,defabs,resabs)
|
241 |
|
|
last = 1
|
242 |
|
|
rlist(1) = result
|
243 |
|
|
elist(1) = abserr
|
244 |
|
|
iord(1) = 1
|
245 |
|
|
c
|
246 |
|
|
c test on accuracy.
|
247 |
|
|
c
|
248 |
|
|
errbnd = dmax1(epsabs,epsrel*dabs(result))
|
249 |
|
|
if(abserr.le.0.5d+02*epmach*defabs.and.abserr.gt.errbnd) ier = 2
|
250 |
|
|
if(limit.eq.1) ier = 1
|
251 |
|
|
if(ier.ne.0.or.(abserr.le.errbnd.and.abserr.ne.resabs)
|
252 |
|
|
* .or.abserr.eq.0.0d+00) go to 60
|
253 |
|
|
c
|
254 |
|
|
c initialization
|
255 |
|
|
c --------------
|
256 |
|
|
c
|
257 |
|
|
c
|
258 |
|
|
errmax = abserr
|
259 |
|
|
maxerr = 1
|
260 |
|
|
area = result
|
261 |
|
|
errsum = abserr
|
262 |
|
|
nrmax = 1
|
263 |
|
|
iroff1 = 0
|
264 |
|
|
iroff2 = 0
|
265 |
|
|
c
|
266 |
|
|
c main do-loop
|
267 |
|
|
c ------------
|
268 |
|
|
c
|
269 |
|
|
do 30 last = 2,limit
|
270 |
|
|
c
|
271 |
|
|
c bisect the subinterval with the largest error estimate.
|
272 |
|
|
c
|
273 |
|
|
a1 = alist(maxerr)
|
274 |
|
|
b1 = 0.5d+00*(alist(maxerr)+blist(maxerr))
|
275 |
|
|
a2 = b1
|
276 |
|
|
b2 = blist(maxerr)
|
277 |
|
|
call dqk15(f,a1,b1,area1,error1,resabs,defab1)
|
278 |
|
|
call dqk15(f,a2,b2,area2,error2,resabs,defab2)
|
279 |
|
|
c
|
280 |
|
|
c improve previous approximations to integral
|
281 |
|
|
c and error and test for accuracy.
|
282 |
|
|
c
|
283 |
|
|
neval = neval+1
|
284 |
|
|
area12 = area1+area2
|
285 |
|
|
erro12 = error1+error2
|
286 |
|
|
errsum = errsum+erro12-errmax
|
287 |
|
|
area = area+area12-rlist(maxerr)
|
288 |
|
|
if(defab1.eq.error1.or.defab2.eq.error2) go to 5
|
289 |
|
|
if(dabs(rlist(maxerr)-area12).le.0.1d-04*dabs(area12)
|
290 |
|
|
* .and.erro12.ge.0.99d+00*errmax) iroff1 = iroff1+1
|
291 |
|
|
if(last.gt.10.and.erro12.gt.errmax) iroff2 = iroff2+1
|
292 |
|
|
5 rlist(maxerr) = area1
|
293 |
|
|
rlist(last) = area2
|
294 |
|
|
errbnd = dmax1(epsabs,epsrel*dabs(area))
|
295 |
|
|
if(errsum.le.errbnd) go to 8
|
296 |
|
|
c
|
297 |
|
|
c test for roundoff error and eventually set error flag.
|
298 |
|
|
c
|
299 |
|
|
if(iroff1.ge.6.or.iroff2.ge.20) ier = 2
|
300 |
|
|
c
|
301 |
|
|
c set error flag in the case that the number of subintervals
|
302 |
|
|
c equals limit.
|
303 |
|
|
c
|
304 |
|
|
if(last.eq.limit) ier = 1
|
305 |
|
|
c
|
306 |
|
|
c set error flag in the case of bad integrand behavior
|
307 |
|
|
c at a point of the integration range.
|
308 |
|
|
c
|
309 |
|
|
if(dmax1(dabs(a1),dabs(b2)).le.(0.1d+01+0.1d+03*
|
310 |
|
|
* epmach)*(dabs(a2)+0.1d+04*uflow)) ier = 3
|
311 |
|
|
c
|
312 |
|
|
c append the newly-created intervals to the list.
|
313 |
|
|
c
|
314 |
|
|
8 if(error2.gt.error1) go to 10
|
315 |
|
|
alist(last) = a2
|
316 |
|
|
blist(maxerr) = b1
|
317 |
|
|
blist(last) = b2
|
318 |
|
|
elist(maxerr) = error1
|
319 |
|
|
elist(last) = error2
|
320 |
|
|
go to 20
|
321 |
|
|
10 alist(maxerr) = a2
|
322 |
|
|
alist(last) = a1
|
323 |
|
|
blist(last) = b1
|
324 |
|
|
rlist(maxerr) = area2
|
325 |
|
|
rlist(last) = area1
|
326 |
|
|
elist(maxerr) = error2
|
327 |
|
|
elist(last) = error1
|
328 |
|
|
c
|
329 |
|
|
c call subroutine dqpsrt to maintain the descending ordering
|
330 |
|
|
c in the list of error estimates and select the subinterval
|
331 |
|
|
c with the largest error estimate (to be bisected next).
|
332 |
|
|
c
|
333 |
|
|
20 call dqpsrt(limit,last,maxerr,errmax,elist,iord,nrmax)
|
334 |
|
|
c ***jump out of do-loop
|
335 |
|
|
if(ier.ne.0.or.errsum.le.errbnd) go to 40
|
336 |
|
|
30 continue
|
337 |
|
|
c
|
338 |
|
|
c compute final result.
|
339 |
|
|
c ---------------------
|
340 |
|
|
c
|
341 |
|
|
40 result = 0.0d+00
|
342 |
|
|
do 50 k=1,last
|
343 |
|
|
result = result+rlist(k)
|
344 |
|
|
50 continue
|
345 |
|
|
abserr = errsum
|
346 |
|
|
60 neval = 30*neval+15
|
347 |
|
|
999 return
|
348 |
|
|
end
|