Improve the translation of notequal
[maxima.git] / share / hompack / fortran / fixpns.f
blob484bf7ce9201697546e50a3145f8abbe905aa7e4
1 SUBROUTINE FIXPNS(N,Y,IFLAG,ARCRE,ARCAE,ANSRE,ANSAE,TRACE,A,
2 $ NFE,ARCLEN,YP,YOLD,YPOLD,QR,LENQR,PIVOT,WORK,SSPAR,
3 $ PAR,IPAR)
5 C SUBROUTINE FIXPNS FINDS A FIXED POINT OR ZERO OF THE
6 C N-DIMENSIONAL VECTOR FUNCTION F(X), OR TRACKS A ZERO CURVE
7 C OF A GENERAL HOMOTOPY MAP RHO(A,X,LAMBDA). FOR THE FIXED
8 C POINT PROBLEM F(X) IS ASSUMED TO BE A C2 MAP OF SOME BALL
9 C INTO ITSELF. THE EQUATION X = F(X) IS SOLVED BY
10 C FOLLOWING THE ZERO CURVE OF THE HOMOTOPY MAP
12 C LAMBDA*(X - F(X)) + (1 - LAMBDA)*(X - A) ,
14 C STARTING FROM LAMBDA = 0, X = A. THE CURVE IS PARAMETERIZED
15 C BY ARC LENGTH S, AND IS FOLLOWED BY SOLVING THE ORDINARY
16 C DIFFERENTIAL EQUATION D(HOMOTOPY MAP)/DS = 0 FOR
17 C Y(S) = (X(S), LAMBDA(S)) USING A HERMITE CUBIC PREDICTOR AND A
18 C CORRECTOR WHICH RETURNS TO THE ZERO CURVE ALONG THE FLOW NORMAL
19 C TO THE DAVIDENKO FLOW (WHICH CONSISTS OF THE INTEGRAL CURVES OF
20 C D(HOMOTOPY MAP)/DS ).
22 C FOR THE ZERO FINDING PROBLEM F(X) IS ASSUMED TO BE A C2 MAP
23 C SUCH THAT FOR SOME R > 0, X*F(X) >= 0 WHENEVER NORM(X) = R.
24 C THE EQUATION F(X) = 0 IS SOLVED BY FOLLOWING THE ZERO CURVE
25 C OF THE HOMOTOPY MAP
27 C LAMBDA*F(X) + (1 - LAMBDA)*(X - A)
29 C EMANATING FROM LAMBDA = 0, X = A.
31 C A MUST BE AN INTERIOR POINT OF THE ABOVE MENTIONED BALLS.
33 C FOR THE CURVE TRACKING PROBLEM RHO(A,X,LAMBDA) IS ASSUMED TO
34 C BE A C2 MAP FROM E**M X E**N X [0,1) INTO E**N, WHICH FOR
35 C ALMOST ALL PARAMETER VECTORS A IN SOME NONEMPTY OPEN SUBSET
36 C OF E**M SATISFIES
38 C RANK [D RHO(A,X,LAMBDA)/DX , D RHO(A,X,LAMBDA)/D LAMBDA] = N
40 C FOR ALL POINTS (X,LAMBDA) SUCH THAT RHO(A,X,LAMBDA)=0. IT IS
41 C FURTHER ASSUMED THAT
43 C RANK [ D RHO(A,X0,0)/DX ] = N .
45 C WITH A FIXED, THE ZERO CURVE OF RHO(A,X,LAMBDA) EMANATING
46 C FROM LAMBDA = 0, X = X0 IS TRACKED UNTIL LAMBDA = 1 BY
47 C SOLVING THE ORDINARY DIFFERENTIAL EQUATION
48 C D RHO(A,X(S),LAMBDA(S))/DS = 0 FOR Y(S) = (X(S), LAMBDA(S)),
49 C WHERE S IS ARC LENGTH ALONG THE ZERO CURVE. ALSO THE HOMOTOPY
50 C MAP RHO(A,X,LAMBDA) IS ASSUMED TO BE CONSTRUCTED SUCH THAT
52 C D LAMBDA(0)/DS > 0 .
55 C FOR THE FIXED POINT AND ZERO FINDING PROBLEMS, THE USER
56 C MUST SUPPLY A SUBROUTINE F(X,V) WHICH EVALUATES F(X) AT X
57 C AND RETURNS THE VECTOR F(X) IN V, AND A SUBROUTINE
58 C FJACS(X,QR,LENQR,PIVOT) WHICH EVALUATES THE (SYMMETRIC)
59 C JACOBIAN MATRIX OF F(X) AT X, AND RETURNS THE SYMMETRIC
60 C JACOBIAN MATRIX IN PACKED SKYLINE STORAGE FORMAT IN QR. LENQR
61 C AND PIVOT DESCRIBE THE DATA STRUCTURE IN QR. FOR THE CURVE
62 C TRACKING PROBLEM, THE USER MUST SUPPLY A SUBROUTINE
63 C RHO(A,LAMBDA,X,V,PAR,IPAR) WHICH EVALUATES THE HOMOTOPY MAP RHO
64 C AT (A,X,LAMBDA) AND RETURNS THE VECTOR RHO(A,X,LAMBDA) IN V, AND
65 C A SUBROUTINE RHOJS(A,LAMBDA,X,QR,LENQR,PIVOT,PP,PAR,IPAR) WHICH
66 C RETURNS IN QR THE SYMMETRIC N X N JACOBIAN MATRIX [D RHO/DX]
67 C EVALUATED AT (A,X,LAMBDA) AND STORED IN PACKED SKYLINE FORMAT, AND
68 C RETURNS IN PP THE VECTOR -(D RHO/D LAMBDA) EVALUATED AT
69 C (A,X,LAMBDA). LENQR AND PIVOT DESCRIBE THE DATA STRUCTURE
70 C IN QR.
71 C *** NOTE THE MINUS SIGN IN THE DEFINITION OF PP. ***
74 C FUNCTIONS AND SUBROUTINES DIRECTLY OR INDIRECTLY CALLED BY FIXPDS:
75 C D1MACH , F (OR RHO ), FJACS (OR RHOJS ), GMFADS , MFACDS ,
76 C MULTDS , PCGDS , PCGNS , QIMUDS , ROOT , ROOTNS , SOLVDS ,
77 C STEPNS , TANGNS , AND THE BLAS FUNCTIONS DAXPY , DCOPY , DDOT ,
78 C DNRM2 , DSCAL , IDAMAX . ONLY D1MACH CONTAINS MACHINE DEPENDENT
79 C CONSTANTS. NO OTHER MODIFICATIONS BY THE USER ARE REQUIRED.
82 C ON INPUT:
84 C N IS THE DIMENSION OF X, F(X), AND RHO(A,X,LAMBDA).
86 C Y IS AN ARRRAY OF LENGTH N + 1. (Y(1),...,Y(N)) = A IS THE
87 C STARTING POINT FOR THE ZERO CURVE FOR THE FIXED POINT AND
88 C ZERO FINDING PROBLEMS. (Y(1),...,Y(N)) = X0 FOR THE CURVE
89 C TRACKING PROBLEM.
91 C IFLAG CAN BE -2, -1, 0, 2, OR 3. IFLAG SHOULD BE 0 ON THE
92 C FIRST CALL TO FIXPNS FOR THE PROBLEM X=F(X), -1 FOR THE
93 C PROBLEM F(X)=0, AND -2 FOR THE PROBLEM RHO(A,X,LAMBDA)=0.
94 C IN CERTAIN SITUATIONS IFLAG IS SET TO 2 OR 3 BY FIXPNS,
95 C AND FIXPNS CAN BE CALLED AGAIN WITHOUT CHANGING IFLAG.
97 C ARCRE , ARCAE ARE THE RELATIVE AND ABSOLUTE ERRORS, RESPECTIVELY,
98 C ALLOWED THE NORMAL FLOW ITERATION ALONG THE ZERO CURVE. IF
99 C ARC?E .LE. 0.0 ON INPUT IT IS RESET TO .5*SQRT(ANS?E) .
100 C NORMALLY ARC?E SHOULD BE CONSIDERABLY LARGER THAN ANS?E .
102 C ANSRE , ANSAE ARE THE RELATIVE AND ABSOLUTE ERROR VALUES USED FOR
103 C THE ANSWER AT LAMBDA = 1. THE ACCEPTED ANSWER Y = (X, LAMBDA)
104 C SATISFIES
106 C |Y(NP1) - 1| .LE. ANSRE + ANSAE .AND.
108 C ||Z|| .LE. ANSRE*||X|| + ANSAE WHERE
110 C (Z,.) IS THE NEWTON STEP TO Y.
112 C TRACE IS AN INTEGER SPECIFYING THE LOGICAL I/O UNIT FOR
113 C INTERMEDIATE OUTPUT. IF TRACE .GT. 0 THE POINTS COMPUTED ON
114 C THE ZERO CURVE ARE WRITTEN TO I/O UNIT TRACE .
116 C A(1:*) CONTAINS THE PARAMETER VECTOR A . FOR THE FIXED POINT
117 C AND ZERO FINDING PROBLEMS, A NEED NOT BE INITIALIZED BY THE
118 C USER, AND IS ASSUMED TO HAVE LENGTH N. FOR THE CURVE
119 C TRACKING PROBLEM, A MUST BE INITIALIZED BY THE USER.
121 C YP(1:N+1) IS A WORK ARRAY CONTAINING THE TANGENT VECTOR TO
122 C THE ZERO CURVE AT THE CURRENT POINT Y .
124 C YOLD(1:N+1) IS A WORK ARRAY CONTAINING THE PREVIOUS POINT FOUND
125 C ON THE ZERO CURVE.
127 C YPOLD(1:N+1) IS A WORK ARRAY CONTAINING THE TANGENT VECTOR TO
128 C THE ZERO CURVE AT YOLD .
130 C QR(1:LENQR) IS A WORK ARRAY CONTAINING THE N X N SYMMETRIC
131 C JACOBIAN MATRIX WITH RESPECT TO X STORED IN PACKED SKYLINE
132 C STORAGE FORMAT. LENQR AND PIVOT DESCRIBE THE DATA
133 C STRUCTURE IN QR .
135 C LENQR IS THE LENGTH OF THE ONE-DIMENSIONAL ARRAY QR .
137 C PIVOT(1:N+2) IS A WORK ARRAY CONTAINING THE INDICES OF THE
138 C DIAGONAL ELEMENTS OF THE N X N SYMMETRIC JACOBIAN MATRIX
139 C (WITH RESPECT TO X) WITHIN QR .
141 C WORK(1:13*(N+1)+2*N+LENQR) IS A WORK ARRAY SPLIT UP AND USED
142 C FOR THE CALCULATION OF THE JACOBIAN MATRIX KERNEL, THE
143 C NEWTON STEP, INTERPOLATION, AND THE ESTIMATION OF THE OPTIMAL
144 C STEP SIZE H .
146 C SSPAR(1:8) = (LIDEAL, RIDEAL, DIDEAL, HMIN, HMAX, BMIN, BMAX, P) IS
147 C A VECTOR OF PARAMETERS USED FOR THE OPTIMAL STEP SIZE ESTIMATION.
148 C IF SSPAR(J) .LE. 0.0 ON INPUT, IT IS RESET TO A DEFAULT VALUE
149 C BY FIXPNS . OTHERWISE THE INPUT VALUE OF SSPAR(J) IS USED.
150 C SEE THE COMMENTS BELOW AND IN STEPNS FOR MORE INFORMATION ABOUT
151 C THESE CONSTANTS.
153 C PAR(1:*) AND IPAR(1:*) ARE ARRAYS FOR (OPTIONAL) USER PARAMETERS,
154 C WHICH ARE SIMPLY PASSED THROUGH TO THE USER WRITTEN SUBROUTINES
155 C RHO, RHOJS.
158 C ON OUTPUT:
160 C N , TRACE , A ARE UNCHANGED.
162 C (Y(1),...,Y(N)) = X, Y(NP1) = LAMBDA, AND Y IS AN APPROXIMATE
163 C ZERO OF THE HOMOTOPY MAP. NORMALLY LAMBDA = 1 AND X IS A
164 C FIXED POINT(ZERO) OF F(X). IN ABNORMAL SITUATIONS LAMBDA
165 C MAY ONLY BE NEAR 1 AND X IS NEAR A FIXED POINT(ZERO).
167 C IFLAG =
168 C -2 CAUSES FIXPNS TO INITIALIZE EVERYTHING FOR THE PROBLEM
169 C RHO(A,X,LAMBDA) = 0 (USE ON FIRST CALL).
171 C -1 CAUSES FIXPNS TO INITIALIZE EVERYTHING FOR THE PROBLEM
172 C F(X) = 0 (USE ON FIRST CALL).
174 C 0 CAUSES FIXPNS TO INITIALIZE EVERYTHING FOR THE PROBLEM
175 C X = F(X) (USE ON FIRST CALL).
177 C 1 NORMAL RETURN.
179 C 2 SPECIFIED ERROR TOLERANCE CANNOT BE MET. SOME OR ALL OF
180 C ARCRE , ARCAE , ANSRE , ANSAE HAVE BEEN INCREASED TO
181 C SUITABLE VALUES. TO CONTINUE, JUST CALL FIXPNS AGAIN
182 C WITHOUT CHANGING ANY PARAMETERS.
184 C 3 STEPNS HAS BEEN CALLED 1000 TIMES. TO CONTINUE, CALL
185 C FIXPNS AGAIN WITHOUT CHANGING ANY PARAMETERS.
187 C 4 THE PRECONDITIONED CONJUGATE GRADIENT ITERATION FAILED TO
188 C CONVERGE (PROBABLY BECAUSE THE JACOBIAN MATRIX DID NOT HAVE
189 C FULL RANK). THE ALGORITHM HAS FAILED (THE ZERO CURVE OF
190 C THE HOMOTOPY MAP CANNOT BE FOLLOWED ANY FURTHER).
192 C 5 THE TRACKING ALGORITHM HAS LOST THE ZERO CURVE OF THE
193 C HOMOTOPY MAP AND IS NOT MAKING PROGRESS. THE ERROR TOLERANCES
194 C ARC?E AND ANS?E WERE TOO LENIENT. THE PROBLEM SHOULD BE
195 C RESTARTED BY CALLING FIXPNS WITH SMALLER ERROR TOLERANCES
196 C AND IFLAG = 0 (-1, -2).
198 C 6 THE NORMAL FLOW NEWTON ITERATION IN STEPNS OR ROOTNS
199 C FAILED TO CONVERGE. THE ERROR TOLERANCES ANS?E MAY BE TOO
200 C STRINGENT.
202 C 7 ILLEGAL INPUT PARAMETERS, A FATAL ERROR.
204 C ARCRE , ARCAE , ANSRE , ANSAE ARE UNCHANGED AFTER A NORMAL RETURN
205 C (IFLAG = 1). THEY ARE INCREASED TO APPROPRIATE VALUES ON THE
206 C RETURN IFLAG = 2 .
208 C NFE IS THE NUMBER OF FUNCTION EVALUATIONS (= NUMBER OF
209 C JACOBIAN EVALUATIONS).
211 C ARCLEN IS THE LENGTH OF THE PATH FOLLOWED.
215 DOUBLE PRECISION ABSERR,ANSAE,ANSRE,ARCAE,ARCLEN,ARCRE,
216 1 CURSW,CURTOL,D1MACH,DNRM2,H,HOLD,RELERR,S
217 INTEGER IFLAG,IFLAGC,IPP,IRHO,ITANGW,ITER,ITZ,IW,IWP,
218 1 IZ0,IZ1,JW,LENQR,LIMIT,LIMITD,N,NC,NFE,NFEC,NP1,TRACE
219 LOGICAL START,CRASH
221 C ***** ARRAY DECLARATIONS. *****
223 DOUBLE PRECISION Y(N+1),YP(N+1),YOLD(N+1),YPOLD(N+1),A(N),
224 $ QR(LENQR),WORK(13*(N+1)+2*N+LENQR),SSPAR(8),PAR(1)
225 INTEGER PIVOT(N+2),IPAR(1)
227 C ***** END OF DIMENSIONAL INFORMATION. *****
229 SAVE
231 C LIMITD IS AN UPPER BOUND ON THE NUMBER OF STEPS. IT MAY BE
232 C CHANGED BY CHANGING THE FOLLOWING PARAMETER STATEMENT:
233 PARAMETER (LIMITD=1000)
235 C SWITCH FROM THE TOLERANCE ARC?E TO THE (FINER) TOLERANCE ANS?E IF
236 C THE CURVATURE OF ANY COMPONENT OF Y EXCEEDS CURSW.
237 PARAMETER (CURSW=10.0)
241 C : : : : : : : : : : : : : : : : : : : : : : : :
242 IF (N .LE. 0 .OR. ANSRE .LE. 0.0 .OR. ANSAE .LT. 0.0)
243 $ IFLAG=7
244 IF (IFLAG .GE. -2 .AND. IFLAG .LE. 0) GO TO 20
245 IF (IFLAG .EQ. 2) GO TO 120
246 IF (IFLAG .EQ. 3) GO TO 90
247 C ONLY VALID INPUT FOR IFLAG IS -2, -1, 0, 2, 3.
248 IFLAG=7
249 RETURN
251 C ***** INITIALIZATION BLOCK. *****
253 20 ARCLEN=0.0
254 IF (ARCRE .LE. 0.0) ARCRE=.5*SQRT(ANSRE)
255 IF (ARCAE .LE. 0.0) ARCAE=.5*SQRT(ANSAE)
256 NC=N
257 NFEC=0
258 IFLAGC=IFLAG
259 NP1=N+1
260 C SET INDICES FOR SPLITTING UP WORK ARRAY.
261 IPP=1
262 IRHO=N+1
263 IW=IRHO+N
264 IWP=IW+NP1
265 ITZ=IWP+NP1
266 IZ0=ITZ+NP1
267 IZ1=IZ0+NP1
268 ITANGW=IZ1+NP1
269 C SET INITIAL CONDITIONS FOR FIRST CALL TO STEPNS .
270 START=.TRUE.
271 CRASH=.FALSE.
272 HOLD=1.0
273 H=.1
274 S=0.0
275 YPOLD(NP1)=1.0
276 YP(NP1)=1.0
277 Y(NP1)=0.0
278 DO 40 JW=1,N
279 YPOLD(JW)=0.0
280 YP(JW)=0.0
281 40 CONTINUE
282 DO 50 JW=ITANGW,ITANGW+NP1+N
283 WORK(JW)=0.0
284 50 CONTINUE
285 C SET OPTIMAL STEP SIZE ESTIMATION PARAMETERS.
286 C LET Z[K] DENOTE THE NEWTON ITERATES ALONG THE FLOW NORMAL TO THE
287 C DAVIDENKO FLOW AND Y THEIR LIMIT.
288 C IDEAL CONTRACTION FACTOR: ||Z[2] - Z[1]|| / ||Z[1] - Z[0]||
289 IF (SSPAR(1) .LE. 0.0) SSPAR(1)= .5
290 C IDEAL RESIDUAL FACTOR: ||RHO(A, Z[1])|| / ||RHO(A, Z[0])||
291 IF (SSPAR(2) .LE. 0.0) SSPAR(2)= .01
292 C IDEAL DISTANCE FACTOR: ||Z[1] - Y|| / ||Z[0] - Y||
293 IF (SSPAR(3) .LE. 0.0) SSPAR(3)= .5
294 C MINIMUM STEP SIZE HMIN .
295 IF (SSPAR(4) .LE. 0.0) SSPAR(4)= (SQRT(N+1.0)+4.0)*D1MACH(4)
296 C MAXIMUM STEP SIZE HMAX .
297 IF (SSPAR(5) .LE. 0.0) SSPAR(5)= 1.0
298 C MINIMUM STEP SIZE REDUCTION FACTOR BMIN .
299 IF (SSPAR(6) .LE. 0.0) SSPAR(6)= .1
300 C MAXIMUM STEP SIZE EXPANSION FACTOR BMAX .
301 IF (SSPAR(7) .LE. 0.0) SSPAR(7)= 3.0
302 C ASSUMED OPERATING ORDER P .
303 IF (SSPAR(8) .LE. 0.0) SSPAR(8)= 2.0
305 C LOAD A FOR THE FIXED POINT AND ZERO FINDING PROBLEMS.
306 IF (IFLAGC .GE. -1) THEN
307 CALL DCOPY(N,Y,1,A,1)
308 ENDIF
309 90 LIMIT=LIMITD
311 C ***** END OF INITIALIZATION BLOCK. *****
314 C ***** MAIN LOOP. *****
316 120 DO 400 ITER=1,LIMIT
317 IF (Y(NP1) .LT. 0.0) THEN
318 ARCLEN=S
319 IFLAG=5
320 RETURN
321 ENDIF
323 C SET DIFFERENT ERROR TOLERANCE IF THE TRAJECTORY Y(S) HAS ANY HIGH
324 C CURVATURE COMPONENTS.
325 140 CURTOL=CURSW*HOLD
326 RELERR=ARCRE
327 ABSERR=ARCAE
328 DO 160 JW=1,NP1
329 IF (ABS(YP(JW)-YPOLD(JW)) .GT. CURTOL) THEN
330 RELERR=ANSRE
331 ABSERR=ANSAE
332 GO TO 200
333 ENDIF
334 160 CONTINUE
336 C TAKE A STEP ALONG THE CURVE.
337 200 CALL STEPNS(NC,NFEC,IFLAGC,START,CRASH,HOLD,H,RELERR,ABSERR,
338 + S,Y,YP,YOLD,YPOLD,A,QR,LENQR,PIVOT,WORK,SSPAR,PAR,IPAR)
339 C PRINT LATEST POINT ON CURVE IF REQUESTED.
340 IF (TRACE .GT. 0) THEN
341 WRITE (TRACE,217) ITER,NFEC,S,Y(NP1),(Y(JW),JW=1,NC)
342 217 FORMAT(/' STEP',I5,3X,'NFE =',I5,3X,'ARC LENGTH =',F9.4,3X,
343 $ 'LAMBDA =',F7.4,5X,'X vector:'/1P,(1X,6E12.4))
344 ENDIF
345 NFE=NFEC
346 C CHECK IF THE STEP WAS SUCCESSFUL.
347 IF (IFLAGC .GT. 0) THEN
348 ARCLEN=S
349 IFLAG=IFLAGC
350 RETURN
351 ENDIF
352 IF (CRASH) THEN
353 C RETURN CODE FOR ERROR TOLERANCE TOO SMALL.
354 IFLAG=2
355 C CHANGE ERROR TOLERANCES.
356 IF (ARCRE .LT. RELERR) ARCRE=RELERR
357 IF (ANSRE .LT. RELERR) ANSRE=RELERR
358 IF (ARCAE .LT. ABSERR) ARCAE=ABSERR
359 IF (ANSAE .LT. ABSERR) ANSAE=ABSERR
360 C CHANGE LIMIT ON NUMBER OF ITERATIONS.
361 LIMIT=LIMIT-ITER
362 RETURN
363 ENDIF
365 IF (Y(NP1) .GE. 1.0) THEN
367 C USE HERMITE CUBIC INTERPOLATION AND NEWTON ITERATION TO GET THE
368 C ANSWER AT LAMBDA = 1.0 .
370 C SAVE YOLD FOR ARC LENGTH CALCULATION LATER.
371 CALL DCOPY(NP1,YOLD,1,WORK(IZ0),1)
373 CALL ROOTNS(NC,NFEC,IFLAGC,ANSRE,ANSAE,Y,YP,YOLD,YPOLD,
374 $ A,QR,LENQR,PIVOT,WORK,PAR,IPAR)
376 NFE=NFEC
377 IFLAG=1
378 C SET ERROR FLAG IF ROOTNS COULD NOT GET THE POINT ON THE ZERO
379 C CURVE AT LAMBDA = 1.0 .
380 IF (IFLAGC .GT. 0) IFLAG=IFLAGC
381 C CALCULATE FINAL ARC LENGTH.
382 DO 290 JW=1,NP1
383 WORK(JW)=Y(JW) - WORK(IZ0+JW-1)
384 290 CONTINUE
385 ARCLEN=S - HOLD + DNRM2(NP1,WORK,1)
386 RETURN
387 ENDIF
389 400 CONTINUE
391 C ***** END OF MAIN LOOP. *****
393 C LAMBDA HAS NOT REACHED 1 IN 1000 STEPS.
394 IFLAG=3
395 ARCLEN=S
396 RETURN