1 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~! |
---|
2 | ! Discrete adjoints of Rosenbrock, ! |
---|
3 | ! for several Rosenbrock methods: ! |
---|
4 | ! * Ros2 ! |
---|
5 | ! * Ros3 ! |
---|
6 | ! * Ros4 ! |
---|
7 | ! * Rodas3 ! |
---|
8 | ! * Rodas4 ! |
---|
9 | ! By default the code employs the KPP sparse linear algebra routines ! |
---|
10 | ! Compile with -DFULL_ALGEBRA to use full linear algebra (LAPACK) ! |
---|
11 | ! ! |
---|
12 | ! (C) Adrian Sandu, August 2004 ! |
---|
13 | ! Virginia Polytechnic Institute and State University ! |
---|
14 | ! Contact: sandu@cs.vt.edu ! |
---|
15 | ! Revised by Philipp Miehe and Adrian Sandu, May 2006 ! |
---|
16 | ! This implementation is part of KPP - the Kinetic PreProcessor ! |
---|
17 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~! |
---|
18 | |
---|
19 | MODULE KPP_ROOT_Integrator |
---|
20 | |
---|
21 | USE KPP_ROOT_Precision |
---|
22 | USE KPP_ROOT_Parameters |
---|
23 | USE KPP_ROOT_Global |
---|
24 | USE KPP_ROOT_LinearAlgebra |
---|
25 | USE KPP_ROOT_Rates |
---|
26 | USE KPP_ROOT_Function |
---|
27 | USE KPP_ROOT_Jacobian |
---|
28 | USE KPP_ROOT_Hessian |
---|
29 | USE KPP_ROOT_Util |
---|
30 | |
---|
31 | IMPLICIT NONE |
---|
32 | PUBLIC |
---|
33 | SAVE |
---|
34 | |
---|
35 | !~~~> Statistics on the work performed by the Rosenbrock method |
---|
36 | INTEGER, PARAMETER :: Nfun=1, Njac=2, Nstp=3, Nacc=4, & |
---|
37 | Nrej=5, Ndec=6, Nsol=7, Nsng=8, & |
---|
38 | Ntexit=1, Nhexit=2, Nhnew = 3 |
---|
39 | |
---|
40 | |
---|
41 | CONTAINS ! Routines in the module KPP_ROOT_Integrator |
---|
42 | |
---|
43 | |
---|
44 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
45 | SUBROUTINE INTEGRATE_ADJ( NADJ, Y, Lambda, TIN, TOUT, & |
---|
46 | ATOL_adj, RTOL_adj, & |
---|
47 | ICNTRL_U, RCNTRL_U, ISTATUS_U, RSTATUS_U ) |
---|
48 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
49 | |
---|
50 | IMPLICIT NONE |
---|
51 | |
---|
52 | !~~~> Y - Concentrations |
---|
53 | KPP_REAL :: Y(NVAR) |
---|
54 | !~~~> NADJ - No. of cost functionals for which adjoints |
---|
55 | ! are evaluated simultaneously |
---|
56 | ! If single cost functional is considered (like in |
---|
57 | ! most applications) simply set NADJ = 1 |
---|
58 | INTEGER NADJ |
---|
59 | !~~~> Lambda - Sensitivities w.r.t. concentrations |
---|
60 | ! Note: Lambda (1:NVAR,j) contains sensitivities of |
---|
61 | ! the j-th cost functional w.r.t. Y(1:NVAR), j=1...NADJ |
---|
62 | KPP_REAL :: Lambda(NVAR,NADJ) |
---|
63 | KPP_REAL, INTENT(IN) :: TIN ! TIN - Start Time |
---|
64 | KPP_REAL, INTENT(IN) :: TOUT ! TOUT - End Time |
---|
65 | !~~~> Tolerances for adjoint calculations |
---|
66 | ! (used only for full continuous adjoint) |
---|
67 | KPP_REAL, INTENT(IN) :: ATOL_adj(NVAR,NADJ), RTOL_adj(NVAR,NADJ) |
---|
68 | !~~~> Optional input parameters and statistics |
---|
69 | INTEGER, INTENT(IN), OPTIONAL :: ICNTRL_U(20) |
---|
70 | KPP_REAL, INTENT(IN), OPTIONAL :: RCNTRL_U(20) |
---|
71 | INTEGER, INTENT(OUT), OPTIONAL :: ISTATUS_U(20) |
---|
72 | KPP_REAL, INTENT(OUT), OPTIONAL :: RSTATUS_U(20) |
---|
73 | |
---|
74 | KPP_REAL :: RCNTRL(20), RSTATUS(20) |
---|
75 | INTEGER :: ICNTRL(20), ISTATUS(20), IERR |
---|
76 | |
---|
77 | INTEGER, SAVE :: Ntotal |
---|
78 | |
---|
79 | |
---|
80 | ICNTRL(1:20) = 0 |
---|
81 | RCNTRL(1:20) = 0.0_dp |
---|
82 | ISTATUS(1:20) = 0 |
---|
83 | RSTATUS(1:20) = 0.0_dp |
---|
84 | |
---|
85 | |
---|
86 | !~~~> fine-tune the integrator: |
---|
87 | ! ICNTRL(1) = 0 ! 0 = non-autonomous, 1 = autonomous |
---|
88 | ! ICNTRL(2) = 1 ! 0 = scalar, 1 = vector tolerances |
---|
89 | ! RCNTRL(3) = STEPMIN ! starting step |
---|
90 | ! ICNTRL(3) = 5 ! choice of the method for forward integration |
---|
91 | ! ICNTRL(6) = 1 ! choice of the method for continuous adjoint |
---|
92 | ! ICNTRL(7) = 2 ! 1=none, 2=discrete, 3=full continuous, 4=simplified continuous adjoint |
---|
93 | ! ICNTRL(8) = 1 ! Save fwd LU factorization: 0 = *don't* save, 1 = save |
---|
94 | |
---|
95 | |
---|
96 | ! if optional parameters are given, and if they are >=0, then they overwrite default settings |
---|
97 | IF (PRESENT(ICNTRL_U)) THEN |
---|
98 | WHERE(ICNTRL_U(:) >= 0) ICNTRL(:) = ICNTRL_U(:) |
---|
99 | END IF |
---|
100 | IF (PRESENT(RCNTRL_U)) THEN |
---|
101 | WHERE(RCNTRL_U(:) >= 0) RCNTRL(:) = RCNTRL_U(:) |
---|
102 | END IF |
---|
103 | |
---|
104 | |
---|
105 | CALL RosenbrockADJ(Y, NADJ, Lambda, & |
---|
106 | TIN, TOUT, & |
---|
107 | ATOL, RTOL, ATOL_adj, RTOL_adj, & |
---|
108 | RCNTRL, ICNTRL, RSTATUS, ISTATUS, IERR) |
---|
109 | |
---|
110 | |
---|
111 | !~~~> Debug option: show number of steps |
---|
112 | ! Ntotal = Ntotal + ISTATUS(Nstp) |
---|
113 | ! WRITE(6,777) ISTATUS(Nstp),Ntotal,VAR(ind_O3),VAR(ind_NO2) |
---|
114 | !777 FORMAT('NSTEPS=',I6,' (',I6,') O3=',E24.14,' NO2=',E24.14) |
---|
115 | |
---|
116 | IF (IERR < 0) THEN |
---|
117 | print *,'RosenbrockADJ: Unsucessful step at T=', & |
---|
118 | TIN,' (IERR=',IERR,')' |
---|
119 | END IF |
---|
120 | |
---|
121 | STEPMIN = RSTATUS(Nhexit) |
---|
122 | ! if optional parameters are given for output |
---|
123 | ! copy to them to return information |
---|
124 | IF (PRESENT(ISTATUS_U)) ISTATUS_U(:) = ISTATUS(:) |
---|
125 | IF (PRESENT(RSTATUS_U)) RSTATUS_U(:) = RSTATUS(:) |
---|
126 | |
---|
127 | END SUBROUTINE INTEGRATE_ADJ |
---|
128 | |
---|
129 | |
---|
130 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
131 | SUBROUTINE RosenbrockADJ( Y, NADJ, Lambda, & |
---|
132 | Tstart, Tend, & |
---|
133 | AbsTol, RelTol, AbsTol_adj, RelTol_adj, & |
---|
134 | RCNTRL, ICNTRL, RSTATUS, ISTATUS, IERR) |
---|
135 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
136 | ! |
---|
137 | ! ADJ = Adjoint of the Tangent Linear Model of a Rosenbrock Method |
---|
138 | ! |
---|
139 | ! Solves the system y'=F(t,y) using a RosenbrockADJ method defined by: |
---|
140 | ! |
---|
141 | ! G = 1/(H*gamma(1)) - Jac(t0,Y0) |
---|
142 | ! T_i = t0 + Alpha(i)*H |
---|
143 | ! Y_i = Y0 + \sum_{j=1}^{i-1} A(i,j)*K_j |
---|
144 | ! G * K_i = Fun( T_i, Y_i ) + \sum_{j=1}^S C(i,j)/H * K_j + |
---|
145 | ! gamma(i)*dF/dT(t0, Y0) |
---|
146 | ! Y1 = Y0 + \sum_{j=1}^S M(j)*K_j |
---|
147 | ! |
---|
148 | ! For details on RosenbrockADJ methods and their implementation consult: |
---|
149 | ! E. Hairer and G. Wanner |
---|
150 | ! "Solving ODEs II. Stiff and differential-algebraic problems". |
---|
151 | ! Springer series in computational mathematics, Springer-Verlag, 1996. |
---|
152 | ! The codes contained in the book inspired this implementation. |
---|
153 | ! |
---|
154 | ! (C) Adrian Sandu, August 2004 |
---|
155 | ! Virginia Polytechnic Institute and State University |
---|
156 | ! Contact: sandu@cs.vt.edu |
---|
157 | ! Revised by Philipp Miehe and Adrian Sandu, May 2006 |
---|
158 | ! This implementation is part of KPP - the Kinetic PreProcessor |
---|
159 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
160 | ! |
---|
161 | !~~~> INPUT ARGUMENTS: |
---|
162 | ! |
---|
163 | !- Y(NVAR) = vector of initial conditions (at T=Tstart) |
---|
164 | ! NADJ -> dimension of linearized system, |
---|
165 | ! i.e. the number of sensitivity coefficients |
---|
166 | !- Lambda(NVAR,NADJ) -> vector of initial sensitivity conditions (at T=Tstart) |
---|
167 | !- [Tstart,Tend] = time range of integration |
---|
168 | ! (if Tstart>Tend the integration is performed backwards in time) |
---|
169 | !- RelTol, AbsTol = user precribed accuracy |
---|
170 | !- SUBROUTINE Fun( T, Y, Ydot ) = ODE function, |
---|
171 | ! returns Ydot = Y' = F(T,Y) |
---|
172 | !- SUBROUTINE Jac( T, Y, Jcb ) = Jacobian of the ODE function, |
---|
173 | ! returns Jcb = dF/dY |
---|
174 | !- ICNTRL(1:10) = integer inputs parameters |
---|
175 | !- RCNTRL(1:10) = real inputs parameters |
---|
176 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
177 | ! |
---|
178 | !~~~> OUTPUT ARGUMENTS: |
---|
179 | ! |
---|
180 | !- Y(NVAR) -> vector of final states (at T->Tend) |
---|
181 | !- Lambda(NVAR,NADJ) -> vector of final sensitivities (at T=Tend) |
---|
182 | !- ICNTRL(11:20) -> integer output parameters |
---|
183 | !- RCNTRL(11:20) -> real output parameters |
---|
184 | !- IERR -> job status upon return |
---|
185 | ! - succes (positive value) or failure (negative value) - |
---|
186 | ! = 1 : Success |
---|
187 | ! = -1 : Improper value for maximal no of steps |
---|
188 | ! = -2 : Selected RosenbrockADJ method not implemented |
---|
189 | ! = -3 : Hmin/Hmax/Hstart must be positive |
---|
190 | ! = -4 : FacMin/FacMax/FacRej must be positive |
---|
191 | ! = -5 : Improper tolerance values |
---|
192 | ! = -6 : No of steps exceeds maximum bound |
---|
193 | ! = -7 : Step size too small |
---|
194 | ! = -8 : Matrix is repeatedly singular |
---|
195 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
196 | ! |
---|
197 | !~~~> INPUT PARAMETERS: |
---|
198 | ! |
---|
199 | ! Note: For input parameters equal to zero the default values of the |
---|
200 | ! corresponding variables are used. |
---|
201 | ! |
---|
202 | ! ICNTRL(1) = 1: F = F(y) Independent of T (AUTONOMOUS) |
---|
203 | ! = 0: F = F(t,y) Depends on T (NON-AUTONOMOUS) |
---|
204 | ! |
---|
205 | ! ICNTRL(2) = 0: AbsTol, RelTol are NVAR-dimensional vectors |
---|
206 | ! = 1: AbsTol, RelTol are scalars |
---|
207 | ! |
---|
208 | ! ICNTRL(3) -> selection of a particular Rosenbrock method |
---|
209 | ! = 0 : default method is Rodas3 |
---|
210 | ! = 1 : method is Ros2 |
---|
211 | ! = 2 : method is Ros3 |
---|
212 | ! = 3 : method is Ros4 |
---|
213 | ! = 4 : method is Rodas3 |
---|
214 | ! = 5: method is Rodas4 |
---|
215 | ! |
---|
216 | ! ICNTRL(4) -> maximum number of integration steps |
---|
217 | ! For ICNTRL(4)=0) the default value of BUFSIZE is used |
---|
218 | ! |
---|
219 | ! ICNTRL(6) -> selection of a particular Rosenbrock method for the |
---|
220 | ! continuous adjoint integration - for cts adjoint it |
---|
221 | ! can be different than the forward method ICNTRL(3) |
---|
222 | ! Note 1: to avoid interpolation errors (which can be huge!) |
---|
223 | ! it is recommended to use only ICNTRL(7) = 2 or 4 |
---|
224 | ! Note 2: the performance of the full continuous adjoint |
---|
225 | ! strongly depends on the forward solution accuracy Abs/RelTol |
---|
226 | ! |
---|
227 | ! ICNTRL(7) -> Type of adjoint algorithm |
---|
228 | ! = 0 : default is discrete adjoint ( of method ICNTRL(3) ) |
---|
229 | ! = 1 : no adjoint |
---|
230 | ! = 2 : discrete adjoint ( of method ICNTRL(3) ) |
---|
231 | ! = 3 : fully adaptive continuous adjoint ( with method ICNTRL(6) ) |
---|
232 | ! = 4 : simplified continuous adjoint ( with method ICNTRL(6) ) |
---|
233 | ! |
---|
234 | ! ICNTRL(8) -> checkpointing the LU factorization at each step: |
---|
235 | ! ICNTRL(8)=0 : do *not* save LU factorization (the default) |
---|
236 | ! ICNTRL(8)=1 : save LU factorization |
---|
237 | ! Note: if ICNTRL(7)=1 the LU factorization is *not* saved |
---|
238 | ! |
---|
239 | !~~~> Real input parameters: |
---|
240 | ! |
---|
241 | ! RCNTRL(1) -> Hmin, lower bound for the integration step size |
---|
242 | ! It is strongly recommended to keep Hmin = ZERO |
---|
243 | ! |
---|
244 | ! RCNTRL(2) -> Hmax, upper bound for the integration step size |
---|
245 | ! |
---|
246 | ! RCNTRL(3) -> Hstart, starting value for the integration step size |
---|
247 | ! |
---|
248 | ! RCNTRL(4) -> FacMin, lower bound on step decrease factor (default=0.2) |
---|
249 | ! |
---|
250 | ! RCNTRL(5) -> FacMax, upper bound on step increase factor (default=6) |
---|
251 | ! |
---|
252 | ! RCNTRL(6) -> FacRej, step decrease factor after multiple rejections |
---|
253 | ! (default=0.1) |
---|
254 | ! |
---|
255 | ! RCNTRL(7) -> FacSafe, by which the new step is slightly smaller |
---|
256 | ! than the predicted value (default=0.9) |
---|
257 | ! |
---|
258 | ! RCNTRL(8) -> ThetaMin. If Newton convergence rate smaller |
---|
259 | ! than ThetaMin the Jacobian is not recomputed; |
---|
260 | ! (default=0.001) |
---|
261 | ! |
---|
262 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
263 | ! |
---|
264 | !~~~> OUTPUT PARAMETERS: |
---|
265 | ! |
---|
266 | ! Note: each call to RosenbrockADJ adds the corrent no. of fcn calls |
---|
267 | ! to previous value of ISTATUS(1), and similar for the other params. |
---|
268 | ! Set ISTATUS(1:10) = 0 before call to avoid this accumulation. |
---|
269 | ! |
---|
270 | ! ISTATUS(1) = No. of function calls |
---|
271 | ! ISTATUS(2) = No. of jacobian calls |
---|
272 | ! ISTATUS(3) = No. of steps |
---|
273 | ! ISTATUS(4) = No. of accepted steps |
---|
274 | ! ISTATUS(5) = No. of rejected steps (except at the beginning) |
---|
275 | ! ISTATUS(6) = No. of LU decompositions |
---|
276 | ! ISTATUS(7) = No. of forward/backward substitutions |
---|
277 | ! ISTATUS(8) = No. of singular matrix decompositions |
---|
278 | ! |
---|
279 | ! RSTATUS(1) -> Texit, the time corresponding to the |
---|
280 | ! computed Y upon return |
---|
281 | ! RSTATUS(2) -> Hexit, last accepted step before exit |
---|
282 | ! For multiple restarts, use Hexit as Hstart in the following run |
---|
283 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
284 | |
---|
285 | IMPLICIT NONE |
---|
286 | |
---|
287 | !~~~> Arguments |
---|
288 | KPP_REAL, INTENT(INOUT) :: Y(NVAR) |
---|
289 | INTEGER, INTENT(IN) :: NADJ |
---|
290 | KPP_REAL, INTENT(INOUT) :: Lambda(NVAR,NADJ) |
---|
291 | KPP_REAL, INTENT(IN) :: Tstart,Tend |
---|
292 | KPP_REAL, INTENT(IN) :: AbsTol(NVAR),RelTol(NVAR) |
---|
293 | KPP_REAL, INTENT(IN) :: AbsTol_adj(NVAR,NADJ), RelTol_adj(NVAR,NADJ) |
---|
294 | INTEGER, INTENT(IN) :: ICNTRL(20) |
---|
295 | KPP_REAL, INTENT(IN) :: RCNTRL(20) |
---|
296 | INTEGER, INTENT(INOUT) :: ISTATUS(20) |
---|
297 | KPP_REAL, INTENT(INOUT) :: RSTATUS(20) |
---|
298 | INTEGER, INTENT(OUT) :: IERR |
---|
299 | !~~~> Parameters of the Rosenbrock method, up to 6 stages |
---|
300 | INTEGER :: ros_S, rosMethod |
---|
301 | INTEGER, PARAMETER :: RS2=1, RS3=2, RS4=3, RD3=4, RD4=5 |
---|
302 | KPP_REAL :: ros_A(15), ros_C(15), ros_M(6), ros_E(6), & |
---|
303 | ros_Alpha(6), ros_Gamma(6), ros_ELO |
---|
304 | LOGICAL :: ros_NewF(6) |
---|
305 | CHARACTER(LEN=12) :: ros_Name |
---|
306 | !~~~> Types of Adjoints Implemented |
---|
307 | INTEGER, PARAMETER :: Adj_none = 1, Adj_discrete = 2, & |
---|
308 | Adj_continuous = 3, Adj_simple_continuous = 4 |
---|
309 | !~~~> Checkpoints in memory |
---|
310 | INTEGER, PARAMETER :: bufsize = 200000 |
---|
311 | INTEGER :: stack_ptr = 0 ! last written entry |
---|
312 | KPP_REAL, DIMENSION(:), POINTER :: chk_H, chk_T |
---|
313 | KPP_REAL, DIMENSION(:,:), POINTER :: chk_Y, chk_K, chk_J |
---|
314 | KPP_REAL, DIMENSION(:,:), POINTER :: chk_dY, chk_d2Y |
---|
315 | !~~~> Local variables |
---|
316 | KPP_REAL :: Roundoff, FacMin, FacMax, FacRej, FacSafe |
---|
317 | KPP_REAL :: Hmin, Hmax, Hstart |
---|
318 | KPP_REAL :: Texit |
---|
319 | INTEGER :: i, UplimTol, Max_no_steps |
---|
320 | INTEGER :: AdjointType, CadjMethod |
---|
321 | LOGICAL :: Autonomous, VectorTol, SaveLU |
---|
322 | !~~~> Parameters |
---|
323 | KPP_REAL, PARAMETER :: ZERO = 0.0d0, ONE = 1.0d0 |
---|
324 | KPP_REAL, PARAMETER :: DeltaMin = 1.0d-5 |
---|
325 | |
---|
326 | !~~~> Initialize statistics |
---|
327 | ISTATUS(1:20) = 0 |
---|
328 | RSTATUS(1:20) = ZERO |
---|
329 | |
---|
330 | !~~~> Autonomous or time dependent ODE. Default is time dependent. |
---|
331 | Autonomous = .NOT.(ICNTRL(1) == 0) |
---|
332 | |
---|
333 | !~~~> For Scalar tolerances (ICNTRL(2).NE.0) the code uses AbsTol(1) and RelTol(1) |
---|
334 | ! For Vector tolerances (ICNTRL(2) == 0) the code uses AbsTol(1:NVAR) and RelTol(1:NVAR) |
---|
335 | IF (ICNTRL(2) == 0) THEN |
---|
336 | VectorTol = .TRUE. |
---|
337 | UplimTol = NVAR |
---|
338 | ELSE |
---|
339 | VectorTol = .FALSE. |
---|
340 | UplimTol = 1 |
---|
341 | END IF |
---|
342 | |
---|
343 | !~~~> Initialize the particular Rosenbrock method selected |
---|
344 | SELECT CASE (ICNTRL(3)) |
---|
345 | CASE (1) |
---|
346 | CALL Ros2 |
---|
347 | CASE (2) |
---|
348 | CALL Ros3 |
---|
349 | CASE (3) |
---|
350 | CALL Ros4 |
---|
351 | CASE (0,4) |
---|
352 | CALL Rodas3 |
---|
353 | CASE (5) |
---|
354 | CALL Rodas4 |
---|
355 | CASE DEFAULT |
---|
356 | PRINT * , 'Unknown Rosenbrock method: ICNTRL(3)=',ICNTRL(3) |
---|
357 | CALL ros_ErrorMsg(-2,Tstart,ZERO,IERR) |
---|
358 | RETURN |
---|
359 | END SELECT |
---|
360 | |
---|
361 | !~~~> The maximum number of steps admitted |
---|
362 | IF (ICNTRL(4) == 0) THEN |
---|
363 | Max_no_steps = bufsize - 1 |
---|
364 | ELSEIF (Max_no_steps > 0) THEN |
---|
365 | Max_no_steps=ICNTRL(4) |
---|
366 | ELSE |
---|
367 | PRINT * ,'User-selected max no. of steps: ICNTRL(4)=',ICNTRL(4) |
---|
368 | CALL ros_ErrorMsg(-1,Tstart,ZERO,IERR) |
---|
369 | RETURN |
---|
370 | END IF |
---|
371 | |
---|
372 | !~~~> The particular Rosenbrock method chosen for integrating the cts adjoint |
---|
373 | IF (ICNTRL(6) == 0) THEN |
---|
374 | CadjMethod = 4 |
---|
375 | ELSEIF ( (ICNTRL(6) >= 1).AND.(ICNTRL(6) <= 5) ) THEN |
---|
376 | CadjMethod = ICNTRL(6) |
---|
377 | ELSE |
---|
378 | PRINT * , 'Unknown CADJ Rosenbrock method: ICNTRL(6)=', CadjMethod |
---|
379 | CALL ros_ErrorMsg(-2,Tstart,ZERO,IERR) |
---|
380 | RETURN |
---|
381 | END IF |
---|
382 | |
---|
383 | !~~~> Discrete or continuous adjoint formulation |
---|
384 | IF ( ICNTRL(7) == 0 ) THEN |
---|
385 | AdjointType = Adj_discrete |
---|
386 | ELSEIF ( (ICNTRL(7) >= 1).AND.(ICNTRL(7) <= 4) ) THEN |
---|
387 | AdjointType = ICNTRL(7) |
---|
388 | ELSE |
---|
389 | PRINT * , 'User-selected adjoint type: ICNTRL(7)=', AdjointType |
---|
390 | CALL ros_ErrorMsg(-9,Tstart,ZERO,IERR) |
---|
391 | RETURN |
---|
392 | END IF |
---|
393 | |
---|
394 | !~~~> Save or not the forward LU factorization |
---|
395 | SaveLU = (ICNTRL(8) /= 0) |
---|
396 | |
---|
397 | |
---|
398 | !~~~> Unit roundoff (1+Roundoff>1) |
---|
399 | Roundoff = WLAMCH('E') |
---|
400 | |
---|
401 | !~~~> Lower bound on the step size: (positive value) |
---|
402 | IF (RCNTRL(1) == ZERO) THEN |
---|
403 | Hmin = ZERO |
---|
404 | ELSEIF (RCNTRL(1) > ZERO) THEN |
---|
405 | Hmin = RCNTRL(1) |
---|
406 | ELSE |
---|
407 | PRINT * , 'User-selected Hmin: RCNTRL(1)=', RCNTRL(1) |
---|
408 | CALL ros_ErrorMsg(-3,Tstart,ZERO,IERR) |
---|
409 | RETURN |
---|
410 | END IF |
---|
411 | !~~~> Upper bound on the step size: (positive value) |
---|
412 | IF (RCNTRL(2) == ZERO) THEN |
---|
413 | Hmax = ABS(Tend-Tstart) |
---|
414 | ELSEIF (RCNTRL(2) > ZERO) THEN |
---|
415 | Hmax = MIN(ABS(RCNTRL(2)),ABS(Tend-Tstart)) |
---|
416 | ELSE |
---|
417 | PRINT * , 'User-selected Hmax: RCNTRL(2)=', RCNTRL(2) |
---|
418 | CALL ros_ErrorMsg(-3,Tstart,ZERO,IERR) |
---|
419 | RETURN |
---|
420 | END IF |
---|
421 | !~~~> Starting step size: (positive value) |
---|
422 | IF (RCNTRL(3) == ZERO) THEN |
---|
423 | Hstart = MAX(Hmin,DeltaMin) |
---|
424 | ELSEIF (RCNTRL(3) > ZERO) THEN |
---|
425 | Hstart = MIN(ABS(RCNTRL(3)),ABS(Tend-Tstart)) |
---|
426 | ELSE |
---|
427 | PRINT * , 'User-selected Hstart: RCNTRL(3)=', RCNTRL(3) |
---|
428 | CALL ros_ErrorMsg(-3,Tstart,ZERO,IERR) |
---|
429 | RETURN |
---|
430 | END IF |
---|
431 | !~~~> Step size can be changed s.t. FacMin < Hnew/Hold < FacMax |
---|
432 | IF (RCNTRL(4) == ZERO) THEN |
---|
433 | FacMin = 0.2d0 |
---|
434 | ELSEIF (RCNTRL(4) > ZERO) THEN |
---|
435 | FacMin = RCNTRL(4) |
---|
436 | ELSE |
---|
437 | PRINT * , 'User-selected FacMin: RCNTRL(4)=', RCNTRL(4) |
---|
438 | CALL ros_ErrorMsg(-4,Tstart,ZERO,IERR) |
---|
439 | RETURN |
---|
440 | END IF |
---|
441 | IF (RCNTRL(5) == ZERO) THEN |
---|
442 | FacMax = 6.0d0 |
---|
443 | ELSEIF (RCNTRL(5) > ZERO) THEN |
---|
444 | FacMax = RCNTRL(5) |
---|
445 | ELSE |
---|
446 | PRINT * , 'User-selected FacMax: RCNTRL(5)=', RCNTRL(5) |
---|
447 | CALL ros_ErrorMsg(-4,Tstart,ZERO,IERR) |
---|
448 | RETURN |
---|
449 | END IF |
---|
450 | !~~~> FacRej: Factor to decrease step after 2 succesive rejections |
---|
451 | IF (RCNTRL(6) == ZERO) THEN |
---|
452 | FacRej = 0.1d0 |
---|
453 | ELSEIF (RCNTRL(6) > ZERO) THEN |
---|
454 | FacRej = RCNTRL(6) |
---|
455 | ELSE |
---|
456 | PRINT * , 'User-selected FacRej: RCNTRL(6)=', RCNTRL(6) |
---|
457 | CALL ros_ErrorMsg(-4,Tstart,ZERO,IERR) |
---|
458 | RETURN |
---|
459 | END IF |
---|
460 | !~~~> FacSafe: Safety Factor in the computation of new step size |
---|
461 | IF (RCNTRL(7) == ZERO) THEN |
---|
462 | FacSafe = 0.9d0 |
---|
463 | ELSEIF (RCNTRL(7) > ZERO) THEN |
---|
464 | FacSafe = RCNTRL(7) |
---|
465 | ELSE |
---|
466 | PRINT * , 'User-selected FacSafe: RCNTRL(7)=', RCNTRL(7) |
---|
467 | CALL ros_ErrorMsg(-4,Tstart,ZERO,IERR) |
---|
468 | RETURN |
---|
469 | END IF |
---|
470 | !~~~> Check if tolerances are reasonable |
---|
471 | DO i=1,UplimTol |
---|
472 | IF ( (AbsTol(i) <= ZERO) .OR. (RelTol(i) <= 10.d0*Roundoff) & |
---|
473 | .OR. (RelTol(i) >= 1.0d0) ) THEN |
---|
474 | PRINT * , ' AbsTol(',i,') = ',AbsTol(i) |
---|
475 | PRINT * , ' RelTol(',i,') = ',RelTol(i) |
---|
476 | CALL ros_ErrorMsg(-5,Tstart,ZERO,IERR) |
---|
477 | RETURN |
---|
478 | END IF |
---|
479 | END DO |
---|
480 | |
---|
481 | |
---|
482 | !~~~> Allocate checkpoint space or open checkpoint files |
---|
483 | IF (AdjointType == Adj_discrete) THEN |
---|
484 | CALL ros_AllocateDBuffers( ros_S ) |
---|
485 | ELSEIF ( (AdjointType == Adj_continuous).OR. & |
---|
486 | (AdjointType == Adj_simple_continuous) ) THEN |
---|
487 | CALL ros_AllocateCBuffers |
---|
488 | END IF |
---|
489 | |
---|
490 | !~~~> CALL Forward Rosenbrock method |
---|
491 | CALL ros_FwdInt(Y,Tstart,Tend,Texit, & |
---|
492 | AbsTol, RelTol, & |
---|
493 | ! Error indicator |
---|
494 | IERR) |
---|
495 | |
---|
496 | PRINT*,'FORWARD STATISTICS' |
---|
497 | PRINT*,'Step=',Nstp,' Acc=',Nacc, & |
---|
498 | ' Rej=',Nrej, ' Singular=',Nsng |
---|
499 | |
---|
500 | !~~~> If Forward integration failed return |
---|
501 | IF (IERR<0) RETURN |
---|
502 | |
---|
503 | !~~~> Initialize the particular Rosenbrock method for continuous adjoint |
---|
504 | IF ( (AdjointType == Adj_continuous).OR. & |
---|
505 | (AdjointType == Adj_simple_continuous) ) THEN |
---|
506 | SELECT CASE (CadjMethod) |
---|
507 | CASE (1) |
---|
508 | CALL Ros2 |
---|
509 | CASE (2) |
---|
510 | CALL Ros3 |
---|
511 | CASE (3) |
---|
512 | CALL Ros4 |
---|
513 | CASE (4) |
---|
514 | CALL Rodas3 |
---|
515 | CASE (5) |
---|
516 | CALL Rodas4 |
---|
517 | CASE DEFAULT |
---|
518 | PRINT * , 'Unknown Rosenbrock method: ICNTRL(3)=', ICNTRL(3) |
---|
519 | CALL ros_ErrorMsg(-2,Tstart,ZERO,IERR) |
---|
520 | RETURN |
---|
521 | END SELECT |
---|
522 | END IF |
---|
523 | |
---|
524 | SELECT CASE (AdjointType) |
---|
525 | CASE (Adj_discrete) |
---|
526 | CALL ros_DadjInt ( & |
---|
527 | NADJ, Lambda, & |
---|
528 | Tstart, Tend, Texit, & |
---|
529 | IERR ) |
---|
530 | CASE (Adj_continuous) |
---|
531 | CALL ros_CadjInt ( & |
---|
532 | NADJ, Lambda, & |
---|
533 | Tend, Tstart, Texit, & |
---|
534 | AbsTol_adj, RelTol_adj, & |
---|
535 | IERR ) |
---|
536 | CASE (Adj_simple_continuous) |
---|
537 | CALL ros_SimpleCadjInt ( & |
---|
538 | NADJ, Lambda, & |
---|
539 | Tstart, Tend, Texit, & |
---|
540 | IERR ) |
---|
541 | END SELECT ! AdjointType |
---|
542 | |
---|
543 | PRINT*,'ADJOINT STATISTICS' |
---|
544 | PRINT*,'Step=',Nstp,' Acc=',Nacc, & |
---|
545 | ' Rej=',Nrej, ' Singular=',Nsng |
---|
546 | |
---|
547 | !~~~> Free checkpoint space or close checkpoint files |
---|
548 | IF (AdjointType == Adj_discrete) THEN |
---|
549 | CALL ros_FreeDBuffers |
---|
550 | ELSEIF ( (AdjointType == Adj_continuous) .OR. & |
---|
551 | (AdjointType == Adj_simple_continuous) ) THEN |
---|
552 | CALL ros_FreeCBuffers |
---|
553 | END IF |
---|
554 | |
---|
555 | |
---|
556 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
557 | CONTAINS ! Procedures internal to RosenbrockADJ |
---|
558 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
559 | |
---|
560 | |
---|
561 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
562 | SUBROUTINE ros_AllocateDBuffers( S ) |
---|
563 | !~~~> Allocate buffer space for discrete adjoint |
---|
564 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
565 | INTEGER :: i, S |
---|
566 | |
---|
567 | ALLOCATE( chk_H(bufsize), STAT=i ) |
---|
568 | IF (i/=0) THEN |
---|
569 | PRINT*,'Failed allocation of buffer H'; STOP |
---|
570 | END IF |
---|
571 | ALLOCATE( chk_T(bufsize), STAT=i ) |
---|
572 | IF (i/=0) THEN |
---|
573 | PRINT*,'Failed allocation of buffer T'; STOP |
---|
574 | END IF |
---|
575 | ALLOCATE( chk_Y(NVAR*S,bufsize), STAT=i ) |
---|
576 | IF (i/=0) THEN |
---|
577 | PRINT*,'Failed allocation of buffer Y'; STOP |
---|
578 | END IF |
---|
579 | ALLOCATE( chk_K(NVAR*S,bufsize), STAT=i ) |
---|
580 | IF (i/=0) THEN |
---|
581 | PRINT*,'Failed allocation of buffer K'; STOP |
---|
582 | END IF |
---|
583 | IF (SaveLU) THEN |
---|
584 | #ifdef FULL_ALGEBRA |
---|
585 | ALLOCATE( chk_J(NVAR*NVAR,bufsize), STAT=i ) |
---|
586 | #else |
---|
587 | ALLOCATE( chk_J(LU_NONZERO,bufsize), STAT=i ) |
---|
588 | #endif |
---|
589 | IF (i/=0) THEN |
---|
590 | PRINT*,'Failed allocation of buffer J'; STOP |
---|
591 | END IF |
---|
592 | END IF |
---|
593 | |
---|
594 | END SUBROUTINE ros_AllocateDBuffers |
---|
595 | |
---|
596 | |
---|
597 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
598 | SUBROUTINE ros_FreeDBuffers |
---|
599 | !~~~> Dallocate buffer space for discrete adjoint |
---|
600 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
601 | INTEGER :: i |
---|
602 | |
---|
603 | DEALLOCATE( chk_H, STAT=i ) |
---|
604 | IF (i/=0) THEN |
---|
605 | PRINT*,'Failed deallocation of buffer H'; STOP |
---|
606 | END IF |
---|
607 | DEALLOCATE( chk_T, STAT=i ) |
---|
608 | IF (i/=0) THEN |
---|
609 | PRINT*,'Failed deallocation of buffer T'; STOP |
---|
610 | END IF |
---|
611 | DEALLOCATE( chk_Y, STAT=i ) |
---|
612 | IF (i/=0) THEN |
---|
613 | PRINT*,'Failed deallocation of buffer Y'; STOP |
---|
614 | END IF |
---|
615 | DEALLOCATE( chk_K, STAT=i ) |
---|
616 | IF (i/=0) THEN |
---|
617 | PRINT*,'Failed deallocation of buffer K'; STOP |
---|
618 | END IF |
---|
619 | IF (SaveLU) THEN |
---|
620 | DEALLOCATE( chk_J, STAT=i ) |
---|
621 | IF (i/=0) THEN |
---|
622 | PRINT*,'Failed deallocation of buffer J'; STOP |
---|
623 | END IF |
---|
624 | END IF |
---|
625 | |
---|
626 | END SUBROUTINE ros_FreeDBuffers |
---|
627 | |
---|
628 | |
---|
629 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
630 | SUBROUTINE ros_AllocateCBuffers |
---|
631 | !~~~> Allocate buffer space for continuous adjoint |
---|
632 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
633 | INTEGER :: i |
---|
634 | |
---|
635 | ALLOCATE( chk_H(bufsize), STAT=i ) |
---|
636 | IF (i/=0) THEN |
---|
637 | PRINT*,'Failed allocation of buffer H'; STOP |
---|
638 | END IF |
---|
639 | ALLOCATE( chk_T(bufsize), STAT=i ) |
---|
640 | IF (i/=0) THEN |
---|
641 | PRINT*,'Failed allocation of buffer T'; STOP |
---|
642 | END IF |
---|
643 | ALLOCATE( chk_Y(NVAR,bufsize), STAT=i ) |
---|
644 | IF (i/=0) THEN |
---|
645 | PRINT*,'Failed allocation of buffer Y'; STOP |
---|
646 | END IF |
---|
647 | ALLOCATE( chk_dY(NVAR,bufsize), STAT=i ) |
---|
648 | IF (i/=0) THEN |
---|
649 | PRINT*,'Failed allocation of buffer dY'; STOP |
---|
650 | END IF |
---|
651 | ALLOCATE( chk_d2Y(NVAR,bufsize), STAT=i ) |
---|
652 | IF (i/=0) THEN |
---|
653 | PRINT*,'Failed allocation of buffer d2Y'; STOP |
---|
654 | END IF |
---|
655 | |
---|
656 | END SUBROUTINE ros_AllocateCBuffers |
---|
657 | |
---|
658 | |
---|
659 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
660 | SUBROUTINE ros_FreeCBuffers |
---|
661 | !~~~> Dallocate buffer space for continuous adjoint |
---|
662 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
663 | INTEGER :: i |
---|
664 | |
---|
665 | DEALLOCATE( chk_H, STAT=i ) |
---|
666 | IF (i/=0) THEN |
---|
667 | PRINT*,'Failed deallocation of buffer H'; STOP |
---|
668 | END IF |
---|
669 | DEALLOCATE( chk_T, STAT=i ) |
---|
670 | IF (i/=0) THEN |
---|
671 | PRINT*,'Failed deallocation of buffer T'; STOP |
---|
672 | END IF |
---|
673 | DEALLOCATE( chk_Y, STAT=i ) |
---|
674 | IF (i/=0) THEN |
---|
675 | PRINT*,'Failed deallocation of buffer Y'; STOP |
---|
676 | END IF |
---|
677 | DEALLOCATE( chk_dY, STAT=i ) |
---|
678 | IF (i/=0) THEN |
---|
679 | PRINT*,'Failed deallocation of buffer dY'; STOP |
---|
680 | END IF |
---|
681 | DEALLOCATE( chk_d2Y, STAT=i ) |
---|
682 | IF (i/=0) THEN |
---|
683 | PRINT*,'Failed deallocation of buffer d2Y'; STOP |
---|
684 | END IF |
---|
685 | |
---|
686 | END SUBROUTINE ros_FreeCBuffers |
---|
687 | |
---|
688 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
689 | SUBROUTINE ros_DPush( S, T, H, Ystage, K, E, P ) |
---|
690 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
691 | !~~~> Saves the next trajectory snapshot for discrete adjoints |
---|
692 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
693 | INTEGER :: S ! no of stages |
---|
694 | KPP_REAL :: T, H, Ystage(NVAR*S), K(NVAR*S) |
---|
695 | INTEGER :: P(NVAR) |
---|
696 | #ifdef FULL_ALGEBRA |
---|
697 | KPP_REAL :: E(NVAR,NVAR) |
---|
698 | #else |
---|
699 | KPP_REAL :: E(LU_NONZERO) |
---|
700 | #endif |
---|
701 | |
---|
702 | stack_ptr = stack_ptr + 1 |
---|
703 | IF ( stack_ptr > bufsize ) THEN |
---|
704 | PRINT*,'Push failed: buffer overflow' |
---|
705 | STOP |
---|
706 | END IF |
---|
707 | chk_H( stack_ptr ) = H |
---|
708 | chk_T( stack_ptr ) = T |
---|
709 | !CALL WCOPY(NVAR*S,Ystage,1,chk_Y(1,stack_ptr),1) |
---|
710 | !CALL WCOPY(NVAR*S,K,1,chk_K(1,stack_ptr),1) |
---|
711 | chk_Y(1:NVAR*S,stack_ptr) = Ystage(1:NVAR*S) |
---|
712 | chk_K(1:NVAR*S,stack_ptr) = K(1:NVAR*S) |
---|
713 | IF (SaveLU) THEN |
---|
714 | #ifdef FULL_ALGEBRA |
---|
715 | chk_J(1:NVAR,1:NVAR,stack_ptr) = E(1:NVAR,1:NVAR) |
---|
716 | chk_P(1:NVAR,stack_ptr) = P(1:NVAR) |
---|
717 | #else |
---|
718 | chk_J(1:LU_NONZERO,stack_ptr) = E(1:LU_NONZERO) |
---|
719 | #endif |
---|
720 | END IF |
---|
721 | |
---|
722 | END SUBROUTINE ros_DPush |
---|
723 | |
---|
724 | |
---|
725 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
726 | SUBROUTINE ros_DPop( S, T, H, Ystage, K, E, P ) |
---|
727 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
728 | !~~~> Retrieves the next trajectory snapshot for discrete adjoints |
---|
729 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
730 | |
---|
731 | INTEGER :: S ! no of stages |
---|
732 | KPP_REAL :: T, H, Ystage(NVAR*S), K(NVAR*S) |
---|
733 | INTEGER :: P(NVAR) |
---|
734 | #ifdef FULL_ALGEBRA |
---|
735 | KPP_REAL :: E(NVAR,NVAR) |
---|
736 | #else |
---|
737 | KPP_REAL :: E(LU_NONZERO) |
---|
738 | #endif |
---|
739 | |
---|
740 | IF ( stack_ptr <= 0 ) THEN |
---|
741 | PRINT*,'Pop failed: empty buffer' |
---|
742 | STOP |
---|
743 | END IF |
---|
744 | H = chk_H( stack_ptr ) |
---|
745 | T = chk_T( stack_ptr ) |
---|
746 | !CALL WCOPY(NVAR*S,chk_Y(1,stack_ptr),1,Ystage,1) |
---|
747 | !CALL WCOPY(NVAR*S,chk_K(1,stack_ptr),1,K,1) |
---|
748 | Ystage(1:NVAR*S) = chk_Y(1:NVAR*S,stack_ptr) |
---|
749 | K(1:NVAR*S) = chk_K(1:NVAR*S,stack_ptr) |
---|
750 | !CALL WCOPY(LU_NONZERO,chk_J(1,stack_ptr),1,Jcb,1) |
---|
751 | IF (SaveLU) THEN |
---|
752 | #ifdef FULL_ALGEBRA |
---|
753 | E(1:NVAR,1:NVAR) = chk_J(1:NVAR,1:NVAR,stack_ptr) |
---|
754 | P(1:NVAR) = chk_P(1:NVAR,stack_ptr) |
---|
755 | #else |
---|
756 | E(1:LU_NONZERO) = chk_J(1:LU_NONZERO,stack_ptr) |
---|
757 | #endif |
---|
758 | END IF |
---|
759 | |
---|
760 | stack_ptr = stack_ptr - 1 |
---|
761 | |
---|
762 | END SUBROUTINE ros_DPop |
---|
763 | |
---|
764 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
765 | SUBROUTINE ros_CPush( T, H, Y, dY, d2Y ) |
---|
766 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
767 | !~~~> Saves the next trajectory snapshot for discrete adjoints |
---|
768 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
769 | |
---|
770 | KPP_REAL :: T, H, Y(NVAR), dY(NVAR), d2Y(NVAR) |
---|
771 | |
---|
772 | stack_ptr = stack_ptr + 1 |
---|
773 | IF ( stack_ptr > bufsize ) THEN |
---|
774 | PRINT*,'Push failed: buffer overflow' |
---|
775 | STOP |
---|
776 | END IF |
---|
777 | chk_H( stack_ptr ) = H |
---|
778 | chk_T( stack_ptr ) = T |
---|
779 | !CALL WCOPY(NVAR,Y,1,chk_Y(1,stack_ptr),1) |
---|
780 | !CALL WCOPY(NVAR,dY,1,chk_dY(1,stack_ptr),1) |
---|
781 | !CALL WCOPY(NVAR,d2Y,1,chk_d2Y(1,stack_ptr),1) |
---|
782 | chk_Y(1:NVAR,stack_ptr) = Y(1:NVAR) |
---|
783 | chk_dY(1:NVAR,stack_ptr) = dY(1:NVAR) |
---|
784 | chk_d2Y(1:NVAR,stack_ptr) = d2Y(1:NVAR) |
---|
785 | END SUBROUTINE ros_CPush |
---|
786 | |
---|
787 | |
---|
788 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
789 | SUBROUTINE ros_CPop( T, H, Y, dY, d2Y ) |
---|
790 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
791 | !~~~> Retrieves the next trajectory snapshot for discrete adjoints |
---|
792 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
793 | |
---|
794 | KPP_REAL :: T, H, Y(NVAR), dY(NVAR), d2Y(NVAR) |
---|
795 | |
---|
796 | IF ( stack_ptr <= 0 ) THEN |
---|
797 | PRINT*,'Pop failed: empty buffer' |
---|
798 | STOP |
---|
799 | END IF |
---|
800 | H = chk_H( stack_ptr ) |
---|
801 | T = chk_T( stack_ptr ) |
---|
802 | !CALL WCOPY(NVAR,chk_Y(1,stack_ptr),1,Y,1) |
---|
803 | !CALL WCOPY(NVAR,chk_dY(1,stack_ptr),1,dY,1) |
---|
804 | !CALL WCOPY(NVAR,chk_d2Y(1,stack_ptr),1,d2Y,1) |
---|
805 | Y(1:NVAR) = chk_Y(1:NVAR,stack_ptr) |
---|
806 | dY(1:NVAR) = chk_dY(1:NVAR,stack_ptr) |
---|
807 | d2Y(1:NVAR) = chk_d2Y(1:NVAR,stack_ptr) |
---|
808 | |
---|
809 | stack_ptr = stack_ptr - 1 |
---|
810 | |
---|
811 | END SUBROUTINE ros_CPop |
---|
812 | |
---|
813 | |
---|
814 | |
---|
815 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
816 | SUBROUTINE ros_ErrorMsg(Code,T,H,IERR) |
---|
817 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
818 | ! Handles all error messages |
---|
819 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
820 | |
---|
821 | KPP_REAL, INTENT(IN) :: T, H |
---|
822 | INTEGER, INTENT(IN) :: Code |
---|
823 | INTEGER, INTENT(OUT) :: IERR |
---|
824 | |
---|
825 | IERR = Code |
---|
826 | PRINT * , & |
---|
827 | 'Forced exit from RosenbrockADJ due to the following error:' |
---|
828 | |
---|
829 | SELECT CASE (Code) |
---|
830 | CASE (-1) |
---|
831 | PRINT * , '--> Improper value for maximal no of steps' |
---|
832 | CASE (-2) |
---|
833 | PRINT * , '--> Selected RosenbrockADJ method not implemented' |
---|
834 | CASE (-3) |
---|
835 | PRINT * , '--> Hmin/Hmax/Hstart must be positive' |
---|
836 | CASE (-4) |
---|
837 | PRINT * , '--> FacMin/FacMax/FacRej must be positive' |
---|
838 | CASE (-5) |
---|
839 | PRINT * , '--> Improper tolerance values' |
---|
840 | CASE (-6) |
---|
841 | PRINT * , '--> No of steps exceeds maximum buffer bound' |
---|
842 | CASE (-7) |
---|
843 | PRINT * , '--> Step size too small: T + 10*H = T', & |
---|
844 | ' or H < Roundoff' |
---|
845 | CASE (-8) |
---|
846 | PRINT * , '--> Matrix is repeatedly singular' |
---|
847 | CASE (-9) |
---|
848 | PRINT * , '--> Improper type of adjoint selected' |
---|
849 | CASE DEFAULT |
---|
850 | PRINT *, 'Unknown Error code: ', Code |
---|
851 | END SELECT |
---|
852 | |
---|
853 | PRINT *, "T=", T, "and H=", H |
---|
854 | |
---|
855 | END SUBROUTINE ros_ErrorMsg |
---|
856 | |
---|
857 | |
---|
858 | |
---|
859 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
860 | SUBROUTINE ros_FwdInt (Y, & |
---|
861 | Tstart, Tend, T, & |
---|
862 | AbsTol, RelTol, & |
---|
863 | !~~~> Error indicator |
---|
864 | IERR ) |
---|
865 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
866 | ! Template for the implementation of a generic RosenbrockADJ method |
---|
867 | ! defined by ros_S (no of stages) |
---|
868 | ! and its coefficients ros_{A,C,M,E,Alpha,Gamma} |
---|
869 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
870 | |
---|
871 | IMPLICIT NONE |
---|
872 | |
---|
873 | !~~~> Input: the initial condition at Tstart; Output: the solution at T |
---|
874 | KPP_REAL, INTENT(INOUT) :: Y(NVAR) |
---|
875 | !~~~> Input: integration interval |
---|
876 | KPP_REAL, INTENT(IN) :: Tstart,Tend |
---|
877 | !~~~> Output: time at which the solution is returned (T=Tend if success) |
---|
878 | KPP_REAL, INTENT(OUT) :: T |
---|
879 | !~~~> Input: tolerances |
---|
880 | KPP_REAL, INTENT(IN) :: AbsTol(NVAR), RelTol(NVAR) |
---|
881 | !~~~> Output: Error indicator |
---|
882 | INTEGER, INTENT(OUT) :: IERR |
---|
883 | ! ~~~~ Local variables |
---|
884 | KPP_REAL :: Ynew(NVAR), Fcn0(NVAR), Fcn(NVAR) |
---|
885 | KPP_REAL :: K(NVAR*ros_S), dFdT(NVAR) |
---|
886 | KPP_REAL, DIMENSION(:), POINTER :: Ystage |
---|
887 | #ifdef FULL_ALGEBRA |
---|
888 | KPP_REAL :: Jac0(NVAR,NVAR), Ghimj(NVAR,NVAR) |
---|
889 | #else |
---|
890 | KPP_REAL :: Jac0(LU_NONZERO), Ghimj(LU_NONZERO) |
---|
891 | #endif |
---|
892 | KPP_REAL :: H, Hnew, HC, HG, Fac, Tau |
---|
893 | KPP_REAL :: Err, Yerr(NVAR) |
---|
894 | INTEGER :: Pivot(NVAR), Direction, ioffset, i, j, istage |
---|
895 | LOGICAL :: RejectLastH, RejectMoreH, Singular |
---|
896 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
897 | |
---|
898 | !~~~> Allocate stage vector buffer if needed |
---|
899 | IF (AdjointType == Adj_discrete) THEN |
---|
900 | ALLOCATE(Ystage(NVAR*ros_S), STAT=i) |
---|
901 | ! Uninitialized Ystage may lead to NaN on some compilers |
---|
902 | Ystage = 0.0d0 |
---|
903 | IF (i/=0) THEN |
---|
904 | PRINT*,'Allocation of Ystage failed' |
---|
905 | STOP |
---|
906 | END IF |
---|
907 | END IF |
---|
908 | |
---|
909 | !~~~> Initial preparations |
---|
910 | T = Tstart |
---|
911 | RSTATUS(Nhexit) = ZERO |
---|
912 | H = MIN( MAX(ABS(Hmin),ABS(Hstart)) , ABS(Hmax) ) |
---|
913 | IF (ABS(H) <= 10.0_dp*Roundoff) H = DeltaMin |
---|
914 | |
---|
915 | IF (Tend >= Tstart) THEN |
---|
916 | Direction = +1 |
---|
917 | ELSE |
---|
918 | Direction = -1 |
---|
919 | END IF |
---|
920 | H = Direction*H |
---|
921 | |
---|
922 | RejectLastH=.FALSE. |
---|
923 | RejectMoreH=.FALSE. |
---|
924 | |
---|
925 | !~~~> Time loop begins below |
---|
926 | |
---|
927 | TimeLoop: DO WHILE ( (Direction > 0).AND.((T-Tend)+Roundoff <= ZERO) & |
---|
928 | .OR. (Direction < 0).AND.((Tend-T)+Roundoff <= ZERO) ) |
---|
929 | |
---|
930 | IF ( ISTATUS(Nstp) > Max_no_steps ) THEN ! Too many steps |
---|
931 | CALL ros_ErrorMsg(-6,T,H,IERR) |
---|
932 | RETURN |
---|
933 | END IF |
---|
934 | IF ( ((T+0.1d0*H) == T).OR.(H <= Roundoff) ) THEN ! Step size too small |
---|
935 | CALL ros_ErrorMsg(-7,T,H,IERR) |
---|
936 | RETURN |
---|
937 | END IF |
---|
938 | |
---|
939 | !~~~> Limit H if necessary to avoid going beyond Tend |
---|
940 | RSTATUS(Nhexit) = H |
---|
941 | H = MIN(H,ABS(Tend-T)) |
---|
942 | |
---|
943 | !~~~> Compute the function at current time |
---|
944 | CALL FunTemplate(T,Y,Fcn0) |
---|
945 | ISTATUS(Nfun) = ISTATUS(Nfun) + 1 |
---|
946 | |
---|
947 | !~~~> Compute the function derivative with respect to T |
---|
948 | IF (.NOT.Autonomous) THEN |
---|
949 | CALL ros_FunTimeDerivative ( T, Roundoff, Y, & |
---|
950 | Fcn0, dFdT ) |
---|
951 | END IF |
---|
952 | |
---|
953 | !~~~> Compute the Jacobian at current time |
---|
954 | CALL JacTemplate(T,Y,Jac0) |
---|
955 | ISTATUS(Njac) = ISTATUS(Njac) + 1 |
---|
956 | |
---|
957 | !~~~> Repeat step calculation until current step accepted |
---|
958 | UntilAccepted: DO |
---|
959 | |
---|
960 | CALL ros_PrepareMatrix(H,Direction,ros_Gamma(1), & |
---|
961 | Jac0,Ghimj,Pivot,Singular) |
---|
962 | IF (Singular) THEN ! More than 5 consecutive failed decompositions |
---|
963 | CALL ros_ErrorMsg(-8,T,H,IERR) |
---|
964 | RETURN |
---|
965 | END IF |
---|
966 | |
---|
967 | !~~~> Compute the stages |
---|
968 | Stage: DO istage = 1, ros_S |
---|
969 | |
---|
970 | ! Current istage offset. Current istage vector is K(ioffset+1:ioffset+NVAR) |
---|
971 | ioffset = NVAR*(istage-1) |
---|
972 | |
---|
973 | ! For the 1st istage the function has been computed previously |
---|
974 | IF ( istage == 1 ) THEN |
---|
975 | CALL WCOPY(NVAR,Fcn0,1,Fcn,1) |
---|
976 | IF (AdjointType == Adj_discrete) THEN ! Save stage solution |
---|
977 | ! CALL WCOPY(NVAR,Y,1,Ystage(1),1) |
---|
978 | Ystage(1:NVAR) = Y(1:NVAR) |
---|
979 | CALL WCOPY(NVAR,Y,1,Ynew,1) |
---|
980 | END IF |
---|
981 | ! istage>1 and a new function evaluation is needed at the current istage |
---|
982 | ELSEIF ( ros_NewF(istage) ) THEN |
---|
983 | CALL WCOPY(NVAR,Y,1,Ynew,1) |
---|
984 | DO j = 1, istage-1 |
---|
985 | CALL WAXPY(NVAR,ros_A((istage-1)*(istage-2)/2+j), & |
---|
986 | K(NVAR*(j-1)+1),1,Ynew,1) |
---|
987 | END DO |
---|
988 | Tau = T + ros_Alpha(istage)*Direction*H |
---|
989 | CALL FunTemplate(Tau,Ynew,Fcn) |
---|
990 | ISTATUS(Nfun) = ISTATUS(Nfun) + 1 |
---|
991 | END IF ! if istage == 1 elseif ros_NewF(istage) |
---|
992 | ! save stage solution every time even if ynew is not updated |
---|
993 | IF ( ( istage > 1 ).AND.(AdjointType == Adj_discrete) ) THEN |
---|
994 | ! CALL WCOPY(NVAR,Ynew,1,Ystage(ioffset+1),1) |
---|
995 | Ystage(ioffset+1:ioffset+NVAR) = Ynew(1:NVAR) |
---|
996 | END IF |
---|
997 | CALL WCOPY(NVAR,Fcn,1,K(ioffset+1),1) |
---|
998 | DO j = 1, istage-1 |
---|
999 | HC = ros_C((istage-1)*(istage-2)/2+j)/(Direction*H) |
---|
1000 | CALL WAXPY(NVAR,HC,K(NVAR*(j-1)+1),1,K(ioffset+1),1) |
---|
1001 | END DO |
---|
1002 | IF ((.NOT. Autonomous).AND.(ros_Gamma(istage).NE.ZERO)) THEN |
---|
1003 | HG = Direction*H*ros_Gamma(istage) |
---|
1004 | CALL WAXPY(NVAR,HG,dFdT,1,K(ioffset+1),1) |
---|
1005 | END IF |
---|
1006 | CALL ros_Solve('N', Ghimj, Pivot, K(ioffset+1)) |
---|
1007 | |
---|
1008 | END DO Stage |
---|
1009 | |
---|
1010 | |
---|
1011 | !~~~> Compute the new solution |
---|
1012 | CALL WCOPY(NVAR,Y,1,Ynew,1) |
---|
1013 | DO j=1,ros_S |
---|
1014 | CALL WAXPY(NVAR,ros_M(j),K(NVAR*(j-1)+1),1,Ynew,1) |
---|
1015 | END DO |
---|
1016 | |
---|
1017 | !~~~> Compute the error estimation |
---|
1018 | CALL WSCAL(NVAR,ZERO,Yerr,1) |
---|
1019 | DO j=1,ros_S |
---|
1020 | CALL WAXPY(NVAR,ros_E(j),K(NVAR*(j-1)+1),1,Yerr,1) |
---|
1021 | END DO |
---|
1022 | Err = ros_ErrorNorm ( Y, Ynew, Yerr, AbsTol, RelTol, VectorTol ) |
---|
1023 | |
---|
1024 | !~~~> New step size is bounded by FacMin <= Hnew/H <= FacMax |
---|
1025 | Fac = MIN(FacMax,MAX(FacMin,FacSafe/Err**(ONE/ros_ELO))) |
---|
1026 | Hnew = H*Fac |
---|
1027 | |
---|
1028 | !~~~> Check the error magnitude and adjust step size |
---|
1029 | ISTATUS(Nstp) = ISTATUS(Nstp) + 1 |
---|
1030 | IF ( (Err <= ONE).OR.(H <= Hmin) ) THEN !~~~> Accept step |
---|
1031 | ISTATUS(Nacc) = ISTATUS(Nacc) + 1 |
---|
1032 | IF (AdjointType == Adj_discrete) THEN ! Save current state |
---|
1033 | CALL ros_DPush( ros_S, T, H, Ystage, K, Ghimj, Pivot ) |
---|
1034 | ELSEIF ( (AdjointType == Adj_continuous) .OR. & |
---|
1035 | (AdjointType == Adj_simple_continuous) ) THEN |
---|
1036 | #ifdef FULL_ALGEBRA |
---|
1037 | K = MATMUL(Jac0,Fcn0) |
---|
1038 | #else |
---|
1039 | CALL Jac_SP_Vec( Jac0, Fcn0, K(1) ) |
---|
1040 | #endif |
---|
1041 | IF (.NOT. Autonomous) THEN |
---|
1042 | CALL WAXPY(NVAR,ONE,dFdT,1,K(1),1) |
---|
1043 | END IF |
---|
1044 | CALL ros_CPush( T, H, Y, Fcn0, K(1) ) |
---|
1045 | END IF |
---|
1046 | CALL WCOPY(NVAR,Ynew,1,Y,1) |
---|
1047 | T = T + Direction*H |
---|
1048 | Hnew = MAX(Hmin,MIN(Hnew,Hmax)) |
---|
1049 | IF (RejectLastH) THEN ! No step size increase after a rejected step |
---|
1050 | Hnew = MIN(Hnew,H) |
---|
1051 | END IF |
---|
1052 | RSTATUS(Nhexit) = H |
---|
1053 | RSTATUS(Nhnew) = Hnew |
---|
1054 | RSTATUS(Ntexit) = T |
---|
1055 | RejectLastH = .FALSE. |
---|
1056 | RejectMoreH = .FALSE. |
---|
1057 | H = Hnew |
---|
1058 | EXIT UntilAccepted ! EXIT THE LOOP: WHILE STEP NOT ACCEPTED |
---|
1059 | ELSE !~~~> Reject step |
---|
1060 | IF (RejectMoreH) THEN |
---|
1061 | Hnew = H*FacRej |
---|
1062 | END IF |
---|
1063 | RejectMoreH = RejectLastH |
---|
1064 | RejectLastH = .TRUE. |
---|
1065 | H = Hnew |
---|
1066 | IF (ISTATUS(Nacc) >= 1) THEN |
---|
1067 | ISTATUS(Nrej) = ISTATUS(Nrej) + 1 |
---|
1068 | END IF |
---|
1069 | END IF ! Err <= 1 |
---|
1070 | |
---|
1071 | END DO UntilAccepted |
---|
1072 | |
---|
1073 | END DO TimeLoop |
---|
1074 | |
---|
1075 | !~~~> Save last state: only needed for continuous adjoint |
---|
1076 | IF ( (AdjointType == Adj_continuous) .OR. & |
---|
1077 | (AdjointType == Adj_simple_continuous) ) THEN |
---|
1078 | CALL FunTemplate(T,Y,Fcn0) |
---|
1079 | ISTATUS(Nfun) = ISTATUS(Nfun) + 1 |
---|
1080 | CALL JacTemplate(T,Y,Jac0) |
---|
1081 | ISTATUS(Njac) = ISTATUS(Njac) + 1 |
---|
1082 | #ifdef FULL_ALGEBRA |
---|
1083 | K = MATMUL(Jac0,Fcn0) |
---|
1084 | #else |
---|
1085 | CALL Jac_SP_Vec( Jac0, Fcn0, K(1) ) |
---|
1086 | #endif |
---|
1087 | IF (.NOT. Autonomous) THEN |
---|
1088 | CALL ros_FunTimeDerivative ( T, Roundoff, Y, & |
---|
1089 | Fcn0, dFdT ) |
---|
1090 | CALL WAXPY(NVAR,ONE,dFdT,1,K(1),1) |
---|
1091 | END IF |
---|
1092 | CALL ros_CPush( T, H, Y, Fcn0, K(1) ) |
---|
1093 | !~~~> Deallocate stage buffer: only needed for discrete adjoint |
---|
1094 | ELSEIF (AdjointType == Adj_discrete) THEN |
---|
1095 | DEALLOCATE(Ystage, STAT=i) |
---|
1096 | IF (i/=0) THEN |
---|
1097 | PRINT*,'Deallocation of Ystage failed' |
---|
1098 | STOP |
---|
1099 | END IF |
---|
1100 | END IF |
---|
1101 | |
---|
1102 | !~~~> Succesful exit |
---|
1103 | IERR = 1 !~~~> The integration was successful |
---|
1104 | |
---|
1105 | END SUBROUTINE ros_FwdInt |
---|
1106 | |
---|
1107 | |
---|
1108 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
1109 | SUBROUTINE ros_DadjInt ( & |
---|
1110 | NADJ, Lambda, & |
---|
1111 | Tstart, Tend, T, & |
---|
1112 | !~~~> Error indicator |
---|
1113 | IERR ) |
---|
1114 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
1115 | ! Template for the implementation of a generic RosenbrockSOA method |
---|
1116 | ! defined by ros_S (no of stages) |
---|
1117 | ! and its coefficients ros_{A,C,M,E,Alpha,Gamma} |
---|
1118 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
1119 | |
---|
1120 | IMPLICIT NONE |
---|
1121 | |
---|
1122 | !~~~> Input: the initial condition at Tstart; Output: the solution at T |
---|
1123 | INTEGER, INTENT(IN) :: NADJ |
---|
1124 | !~~~> First order adjoint |
---|
1125 | KPP_REAL, INTENT(INOUT) :: Lambda(NVAR,NADJ) |
---|
1126 | !!~~~> Input: integration interval |
---|
1127 | KPP_REAL, INTENT(IN) :: Tstart,Tend |
---|
1128 | !~~~> Output: time at which the solution is returned (T=Tend if success) |
---|
1129 | KPP_REAL, INTENT(OUT) :: T |
---|
1130 | !~~~> Output: Error indicator |
---|
1131 | INTEGER, INTENT(OUT) :: IERR |
---|
1132 | ! ~~~~ Local variables |
---|
1133 | KPP_REAL :: Ystage(NVAR*ros_S), K(NVAR*ros_S) |
---|
1134 | KPP_REAL :: U(NVAR*ros_S,NADJ), V(NVAR*ros_S,NADJ) |
---|
1135 | #ifdef FULL_ALGEBRA |
---|
1136 | KPP_REAL, DIMENSION(NVAR,NVAR) :: Jac, dJdT, Ghimj |
---|
1137 | #else |
---|
1138 | KPP_REAL, DIMENSION(LU_NONZERO) :: Jac, dJdT, Ghimj |
---|
1139 | #endif |
---|
1140 | KPP_REAL :: Hes0(NHESS) |
---|
1141 | KPP_REAL :: Tmp(NVAR), Tmp2(NVAR) |
---|
1142 | KPP_REAL :: H, HC, HA, Tau |
---|
1143 | INTEGER :: Pivot(NVAR), Direction |
---|
1144 | INTEGER :: i, j, m, istage, istart, jstart |
---|
1145 | !~~~> Local parameters |
---|
1146 | KPP_REAL, PARAMETER :: ZERO = 0.0d0, ONE = 1.0d0 |
---|
1147 | KPP_REAL, PARAMETER :: DeltaMin = 1.0d-5 |
---|
1148 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
1149 | |
---|
1150 | |
---|
1151 | |
---|
1152 | IF (Tend >= Tstart) THEN |
---|
1153 | Direction = +1 |
---|
1154 | ELSE |
---|
1155 | Direction = -1 |
---|
1156 | END IF |
---|
1157 | |
---|
1158 | !~~~> Time loop begins below |
---|
1159 | TimeLoop: DO WHILE ( stack_ptr > 0 ) |
---|
1160 | |
---|
1161 | !~~~> Recover checkpoints for stage values and vectors |
---|
1162 | CALL ros_DPop( ros_S, T, H, Ystage, K, Ghimj, Pivot ) |
---|
1163 | |
---|
1164 | ! ISTATUS(Nstp) = ISTATUS(Nstp) + 1 |
---|
1165 | |
---|
1166 | !~~~> Compute LU decomposition |
---|
1167 | IF (.NOT.SaveLU) THEN |
---|
1168 | CALL JacTemplate(T,Ystage(1),Ghimj) |
---|
1169 | ISTATUS(Njac) = ISTATUS(Njac) + 1 |
---|
1170 | Tau = ONE/(Direction*H*ros_Gamma(1)) |
---|
1171 | #ifdef FULL_ALGEBRA |
---|
1172 | Ghimj(1:NVAR,1:NVAR) = -Ghimj(1:NVAR,1:NVAR) |
---|
1173 | DO i=1,NVAR |
---|
1174 | Ghimj(i,i) = Ghimj(i,i)+Tau |
---|
1175 | END DO |
---|
1176 | #else |
---|
1177 | CALL WSCAL(LU_NONZERO,(-ONE),Ghimj,1) |
---|
1178 | DO i=1,NVAR |
---|
1179 | Ghimj(LU_DIAG(i)) = Ghimj(LU_DIAG(i))+Tau |
---|
1180 | END DO |
---|
1181 | #endif |
---|
1182 | CALL ros_Decomp( Ghimj, Pivot, j ) |
---|
1183 | END IF |
---|
1184 | |
---|
1185 | !~~~> Compute Hessian at the beginning of the interval |
---|
1186 | CALL HessTemplate(T,Ystage(1),Hes0) |
---|
1187 | |
---|
1188 | !~~~> Compute the stages |
---|
1189 | Stage: DO istage = ros_S, 1, -1 |
---|
1190 | |
---|
1191 | !~~~> Current istage first entry |
---|
1192 | istart = NVAR*(istage-1) + 1 |
---|
1193 | |
---|
1194 | !~~~> Compute U |
---|
1195 | DO m = 1,NADJ |
---|
1196 | CALL WCOPY(NVAR,Lambda(1,m),1,U(istart,m),1) |
---|
1197 | CALL WSCAL(NVAR,ros_M(istage),U(istart,m),1) |
---|
1198 | END DO ! m=1:NADJ |
---|
1199 | DO j = istage+1, ros_S |
---|
1200 | jstart = NVAR*(j-1) + 1 |
---|
1201 | HA = ros_A((j-1)*(j-2)/2+istage) |
---|
1202 | HC = ros_C((j-1)*(j-2)/2+istage)/(Direction*H) |
---|
1203 | DO m = 1,NADJ |
---|
1204 | CALL WAXPY(NVAR,HA,V(jstart,m),1,U(istart,m),1) |
---|
1205 | CALL WAXPY(NVAR,HC,U(jstart,m),1,U(istart,m),1) |
---|
1206 | END DO ! m=1:NADJ |
---|
1207 | END DO |
---|
1208 | DO m = 1,NADJ |
---|
1209 | CALL ros_Solve('T', Ghimj, Pivot, U(istart,m)) |
---|
1210 | END DO ! m=1:NADJ |
---|
1211 | !~~~> Compute V |
---|
1212 | Tau = T + ros_Alpha(istage)*Direction*H |
---|
1213 | CALL JacTemplate(Tau,Ystage(istart),Jac) |
---|
1214 | ISTATUS(Njac) = ISTATUS(Njac) + 1 |
---|
1215 | DO m = 1,NADJ |
---|
1216 | #ifdef FULL_ALGEBRA |
---|
1217 | V(istart:istart+NVAR-1,m) = MATMUL(TRANSPOSE(Jac),U(istart:istart+NVAR-1,m)) |
---|
1218 | #else |
---|
1219 | CALL JacTR_SP_Vec(Jac,U(istart,m),V(istart,m)) |
---|
1220 | #endif |
---|
1221 | END DO ! m=1:NADJ |
---|
1222 | |
---|
1223 | END DO Stage |
---|
1224 | |
---|
1225 | IF (.NOT.Autonomous) THEN |
---|
1226 | !~~~> Compute the Jacobian derivative with respect to T. |
---|
1227 | ! Last "Jac" computed for stage 1 |
---|
1228 | CALL ros_JacTimeDerivative ( T, Roundoff, Ystage(1), Jac, dJdT ) |
---|
1229 | END IF |
---|
1230 | |
---|
1231 | !~~~> Compute the new solution |
---|
1232 | |
---|
1233 | !~~~> Compute Lambda |
---|
1234 | DO istage=1,ros_S |
---|
1235 | istart = NVAR*(istage-1) + 1 |
---|
1236 | DO m = 1,NADJ |
---|
1237 | ! Add V_i |
---|
1238 | CALL WAXPY(NVAR,ONE,V(istart,m),1,Lambda(1,m),1) |
---|
1239 | ! Add (H0xK_i)^T * U_i |
---|
1240 | CALL HessTR_Vec ( Hes0, U(istart,m), K(istart), Tmp ) |
---|
1241 | CALL WAXPY(NVAR,ONE,Tmp,1,Lambda(1,m),1) |
---|
1242 | END DO ! m=1:NADJ |
---|
1243 | END DO |
---|
1244 | ! Add H * dJac_dT_0^T * \sum(gamma_i U_i) |
---|
1245 | ! Tmp holds sum gamma_i U_i |
---|
1246 | IF (.NOT.Autonomous) THEN |
---|
1247 | DO m = 1,NADJ |
---|
1248 | Tmp(1:NVAR) = ZERO |
---|
1249 | DO istage = 1, ros_S |
---|
1250 | istart = NVAR*(istage-1) + 1 |
---|
1251 | CALL WAXPY(NVAR,ros_Gamma(istage),U(istart,m),1,Tmp,1) |
---|
1252 | END DO |
---|
1253 | #ifdef FULL_ALGEBRA |
---|
1254 | Tmp2 = MATMUL(TRANSPOSE(dJdT),Tmp) |
---|
1255 | #else |
---|
1256 | CALL JacTR_SP_Vec(dJdT,Tmp,Tmp2) |
---|
1257 | #endif |
---|
1258 | CALL WAXPY(NVAR,H,Tmp2,1,Lambda(1,m),1) |
---|
1259 | END DO ! m=1:NADJ |
---|
1260 | END IF ! .NOT.Autonomous |
---|
1261 | |
---|
1262 | |
---|
1263 | END DO TimeLoop |
---|
1264 | |
---|
1265 | !~~~> Save last state |
---|
1266 | |
---|
1267 | !~~~> Succesful exit |
---|
1268 | IERR = 1 !~~~> The integration was successful |
---|
1269 | |
---|
1270 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
1271 | END SUBROUTINE ros_DadjInt |
---|
1272 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
1273 | |
---|
1274 | |
---|
1275 | |
---|
1276 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
1277 | SUBROUTINE ros_CadjInt ( & |
---|
1278 | NADJ, Y, & |
---|
1279 | Tstart, Tend, T, & |
---|
1280 | AbsTol_adj, RelTol_adj, & |
---|
1281 | !~~~> Error indicator |
---|
1282 | IERR ) |
---|
1283 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
1284 | ! Template for the implementation of a generic RosenbrockADJ method |
---|
1285 | ! defined by ros_S (no of stages) |
---|
1286 | ! and its coefficients ros_{A,C,M,E,Alpha,Gamma} |
---|
1287 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
1288 | |
---|
1289 | IMPLICIT NONE |
---|
1290 | |
---|
1291 | !~~~> Input: the initial condition at Tstart; Output: the solution at T |
---|
1292 | INTEGER, INTENT(IN) :: NADJ |
---|
1293 | KPP_REAL, INTENT(INOUT) :: Y(NVAR,NADJ) |
---|
1294 | !~~~> Input: integration interval |
---|
1295 | KPP_REAL, INTENT(IN) :: Tstart,Tend |
---|
1296 | !~~~> Input: adjoint tolerances |
---|
1297 | KPP_REAL, INTENT(IN) :: AbsTol_adj(NVAR,NADJ), RelTol_adj(NVAR,NADJ) |
---|
1298 | !~~~> Output: time at which the solution is returned (T=Tend if success) |
---|
1299 | KPP_REAL, INTENT(OUT) :: T |
---|
1300 | !~~~> Output: Error indicator |
---|
1301 | INTEGER, INTENT(OUT) :: IERR |
---|
1302 | ! ~~~~ Local variables |
---|
1303 | KPP_REAL :: Y0(NVAR) |
---|
1304 | KPP_REAL :: Ynew(NVAR,NADJ), Fcn0(NVAR,NADJ), Fcn(NVAR,NADJ) |
---|
1305 | KPP_REAL :: K(NVAR*ros_S,NADJ), dFdT(NVAR,NADJ) |
---|
1306 | #ifdef FULL_ALGEBRA |
---|
1307 | KPP_REAL, DIMENSION(NVAR,NVAR) :: Jac0, Ghimj, Jac, dJdT |
---|
1308 | #else |
---|
1309 | KPP_REAL, DIMENSION(LU_NONZERO) :: Jac0, Ghimj, Jac, dJdT |
---|
1310 | #endif |
---|
1311 | KPP_REAL :: H, Hnew, HC, HG, Fac, Tau |
---|
1312 | KPP_REAL :: Err, Yerr(NVAR,NADJ) |
---|
1313 | INTEGER :: Pivot(NVAR), Direction, ioffset, j, istage, iadj |
---|
1314 | LOGICAL :: RejectLastH, RejectMoreH, Singular |
---|
1315 | !~~~> Local parameters |
---|
1316 | KPP_REAL, PARAMETER :: ZERO = 0.0d0, ONE = 1.0d0 |
---|
1317 | KPP_REAL, PARAMETER :: DeltaMin = 1.0d-5 |
---|
1318 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
1319 | |
---|
1320 | |
---|
1321 | !~~~> Initial preparations |
---|
1322 | T = Tstart |
---|
1323 | RSTATUS(Nhexit) = 0.0_dp |
---|
1324 | H = MIN( MAX(ABS(Hmin),ABS(Hstart)) , ABS(Hmax) ) |
---|
1325 | IF (ABS(H) <= 10.0_dp*Roundoff) H = DeltaMin |
---|
1326 | |
---|
1327 | IF (Tend >= Tstart) THEN |
---|
1328 | Direction = +1 |
---|
1329 | ELSE |
---|
1330 | Direction = -1 |
---|
1331 | END IF |
---|
1332 | H = Direction*H |
---|
1333 | |
---|
1334 | RejectLastH=.FALSE. |
---|
1335 | RejectMoreH=.FALSE. |
---|
1336 | |
---|
1337 | !~~~> Time loop begins below |
---|
1338 | |
---|
1339 | TimeLoop: DO WHILE ( (Direction > 0).AND.((T-Tend)+Roundoff <= ZERO) & |
---|
1340 | .OR. (Direction < 0).AND.((Tend-T)+Roundoff <= ZERO) ) |
---|
1341 | |
---|
1342 | IF ( ISTATUS(Nstp) > Max_no_steps ) THEN ! Too many steps |
---|
1343 | CALL ros_ErrorMsg(-6,T,H,IERR) |
---|
1344 | RETURN |
---|
1345 | END IF |
---|
1346 | IF ( ((T+0.1d0*H) == T).OR.(H <= Roundoff) ) THEN ! Step size too small |
---|
1347 | CALL ros_ErrorMsg(-7,T,H,IERR) |
---|
1348 | RETURN |
---|
1349 | END IF |
---|
1350 | |
---|
1351 | !~~~> Limit H if necessary to avoid going beyond Tend |
---|
1352 | RSTATUS(Nhexit) = H |
---|
1353 | H = MIN(H,ABS(Tend-T)) |
---|
1354 | |
---|
1355 | !~~~> Interpolate forward solution |
---|
1356 | CALL ros_cadj_Y( T, Y0 ) |
---|
1357 | !~~~> Compute the Jacobian at current time |
---|
1358 | CALL JacTemplate(T, Y0, Jac0) |
---|
1359 | ISTATUS(Njac) = ISTATUS(Njac) + 1 |
---|
1360 | |
---|
1361 | !~~~> Compute the function derivative with respect to T |
---|
1362 | IF (.NOT.Autonomous) THEN |
---|
1363 | CALL ros_JacTimeDerivative ( T, Roundoff, Y0, & |
---|
1364 | Jac0, dJdT ) |
---|
1365 | DO iadj = 1, NADJ |
---|
1366 | #ifdef FULL_ALGEBRA |
---|
1367 | dFdT(1:NVAR,iadj) = MATMUL(TRANSPOSE(dJdT),Y(1:NVAR,iadj)) |
---|
1368 | #else |
---|
1369 | CALL JacTR_SP_Vec(dJdT,Y(1,iadj),dFdT(1,iadj)) |
---|
1370 | #endif |
---|
1371 | CALL WSCAL(NVAR,(-ONE),dFdT(1,iadj),1) |
---|
1372 | END DO |
---|
1373 | END IF |
---|
1374 | |
---|
1375 | !~~~> Ydot = -J^T*Y |
---|
1376 | #ifdef FULL_ALGEBRA |
---|
1377 | Jac0(1:NVAR,1:NVAR) = -Jac0(1:NVAR,1:NVAR) |
---|
1378 | #else |
---|
1379 | CALL WSCAL(LU_NONZERO,(-ONE),Jac0,1) |
---|
1380 | #endif |
---|
1381 | DO iadj = 1, NADJ |
---|
1382 | #ifdef FULL_ALGEBRA |
---|
1383 | Fcn0(1:NVAR,iadj) = MATMUL(TRANSPOSE(Jac0),Y(1:NVAR,iadj)) |
---|
1384 | #else |
---|
1385 | CALL JacTR_SP_Vec(Jac0,Y(1,iadj),Fcn0(1,iadj)) |
---|
1386 | #endif |
---|
1387 | END DO |
---|
1388 | |
---|
1389 | !~~~> Repeat step calculation until current step accepted |
---|
1390 | UntilAccepted: DO |
---|
1391 | |
---|
1392 | CALL ros_PrepareMatrix(H,Direction,ros_Gamma(1), & |
---|
1393 | Jac0,Ghimj,Pivot,Singular) |
---|
1394 | IF (Singular) THEN ! More than 5 consecutive failed decompositions |
---|
1395 | CALL ros_ErrorMsg(-8,T,H,IERR) |
---|
1396 | RETURN |
---|
1397 | END IF |
---|
1398 | |
---|
1399 | !~~~> Compute the stages |
---|
1400 | Stage: DO istage = 1, ros_S |
---|
1401 | |
---|
1402 | ! Current istage offset. Current istage vector is K(ioffset+1:ioffset+NVAR) |
---|
1403 | ioffset = NVAR*(istage-1) |
---|
1404 | |
---|
1405 | ! For the 1st istage the function has been computed previously |
---|
1406 | IF ( istage == 1 ) THEN |
---|
1407 | DO iadj = 1, NADJ |
---|
1408 | CALL WCOPY(NVAR,Fcn0(1,iadj),1,Fcn(1,iadj),1) |
---|
1409 | END DO |
---|
1410 | ! istage>1 and a new function evaluation is needed at the current istage |
---|
1411 | ELSEIF ( ros_NewF(istage) ) THEN |
---|
1412 | CALL WCOPY(NVAR*NADJ,Y,1,Ynew,1) |
---|
1413 | DO j = 1, istage-1 |
---|
1414 | DO iadj = 1, NADJ |
---|
1415 | CALL WAXPY(NVAR,ros_A((istage-1)*(istage-2)/2+j), & |
---|
1416 | K(NVAR*(j-1)+1,iadj),1,Ynew(1,iadj),1) |
---|
1417 | END DO |
---|
1418 | END DO |
---|
1419 | Tau = T + ros_Alpha(istage)*Direction*H |
---|
1420 | ! CALL FunTemplate(Tau,Ynew,Fcn) |
---|
1421 | ! ISTATUS(Nfun) = ISTATUS(Nfun) + 1 |
---|
1422 | CALL ros_cadj_Y( Tau, Y0 ) |
---|
1423 | CALL JacTemplate(Tau, Y0, Jac) |
---|
1424 | ISTATUS(Njac) = ISTATUS(Njac) + 1 |
---|
1425 | #ifdef FULL_ALGEBRA |
---|
1426 | Jac(1:NVAR,1:NVAR) = -Jac(1:NVAR,1:NVAR) |
---|
1427 | #else |
---|
1428 | CALL WSCAL(LU_NONZERO,(-ONE),Jac,1) |
---|
1429 | #endif |
---|
1430 | DO iadj = 1, NADJ |
---|
1431 | #ifdef FULL_ALGEBRA |
---|
1432 | Fcn(1:NVAR,iadj) = MATMUL(TRANSPOSE(Jac),Ynew(1:NVAR,iadj)) |
---|
1433 | #else |
---|
1434 | CALL JacTR_SP_Vec(Jac,Ynew(1,iadj),Fcn(1,iadj)) |
---|
1435 | #endif |
---|
1436 | !CALL WSCAL(NVAR,(-ONE),Fcn(1,iadj),1) |
---|
1437 | END DO |
---|
1438 | END IF ! if istage == 1 elseif ros_NewF(istage) |
---|
1439 | |
---|
1440 | DO iadj = 1, NADJ |
---|
1441 | CALL WCOPY(NVAR,Fcn(1,iadj),1,K(ioffset+1,iadj),1) |
---|
1442 | END DO |
---|
1443 | DO j = 1, istage-1 |
---|
1444 | HC = ros_C((istage-1)*(istage-2)/2+j)/(Direction*H) |
---|
1445 | DO iadj = 1, NADJ |
---|
1446 | CALL WAXPY(NVAR,HC,K(NVAR*(j-1)+1,iadj),1, & |
---|
1447 | K(ioffset+1,iadj),1) |
---|
1448 | END DO |
---|
1449 | END DO |
---|
1450 | IF ((.NOT. Autonomous).AND.(ros_Gamma(istage).NE.ZERO)) THEN |
---|
1451 | HG = Direction*H*ros_Gamma(istage) |
---|
1452 | DO iadj = 1, NADJ |
---|
1453 | CALL WAXPY(NVAR,HG,dFdT(1,iadj),1,K(ioffset+1,iadj),1) |
---|
1454 | END DO |
---|
1455 | END IF |
---|
1456 | DO iadj = 1, NADJ |
---|
1457 | CALL ros_Solve('T', Ghimj, Pivot, K(ioffset+1,iadj)) |
---|
1458 | END DO |
---|
1459 | |
---|
1460 | END DO Stage |
---|
1461 | |
---|
1462 | |
---|
1463 | !~~~> Compute the new solution |
---|
1464 | DO iadj = 1, NADJ |
---|
1465 | CALL WCOPY(NVAR,Y(1,iadj),1,Ynew(1,iadj),1) |
---|
1466 | DO j=1,ros_S |
---|
1467 | CALL WAXPY(NVAR,ros_M(j),K(NVAR*(j-1)+1,iadj),1,Ynew(1,iadj),1) |
---|
1468 | END DO |
---|
1469 | END DO |
---|
1470 | |
---|
1471 | !~~~> Compute the error estimation |
---|
1472 | CALL WSCAL(NVAR*NADJ,ZERO,Yerr,1) |
---|
1473 | DO j=1,ros_S |
---|
1474 | DO iadj = 1, NADJ |
---|
1475 | CALL WAXPY(NVAR,ros_E(j),K(NVAR*(j-1)+1,iadj),1,Yerr(1,iadj),1) |
---|
1476 | END DO |
---|
1477 | END DO |
---|
1478 | !~~~> Max error among all adjoint components |
---|
1479 | iadj = 1 |
---|
1480 | Err = ros_ErrorNorm ( Y(1,iadj), Ynew(1,iadj), Yerr(1,iadj), & |
---|
1481 | AbsTol_adj(1,iadj), RelTol_adj(1,iadj), VectorTol ) |
---|
1482 | |
---|
1483 | !~~~> New step size is bounded by FacMin <= Hnew/H <= FacMax |
---|
1484 | Fac = MIN(FacMax,MAX(FacMin,FacSafe/Err**(ONE/ros_ELO))) |
---|
1485 | Hnew = H*Fac |
---|
1486 | |
---|
1487 | !~~~> Check the error magnitude and adjust step size |
---|
1488 | ! ISTATUS(Nstp) = ISTATUS(Nstp) + 1 |
---|
1489 | IF ( (Err <= ONE).OR.(H <= Hmin) ) THEN !~~~> Accept step |
---|
1490 | ISTATUS(Nacc) = ISTATUS(Nacc) + 1 |
---|
1491 | CALL WCOPY(NVAR*NADJ,Ynew,1,Y,1) |
---|
1492 | T = T + Direction*H |
---|
1493 | Hnew = MAX(Hmin,MIN(Hnew,Hmax)) |
---|
1494 | IF (RejectLastH) THEN ! No step size increase after a rejected step |
---|
1495 | Hnew = MIN(Hnew,H) |
---|
1496 | END IF |
---|
1497 | RSTATUS(Nhexit) = H |
---|
1498 | RSTATUS(Nhnew) = Hnew |
---|
1499 | RSTATUS(Ntexit) = T |
---|
1500 | RejectLastH = .FALSE. |
---|
1501 | RejectMoreH = .FALSE. |
---|
1502 | H = Hnew |
---|
1503 | EXIT UntilAccepted ! EXIT THE LOOP: WHILE STEP NOT ACCEPTED |
---|
1504 | ELSE !~~~> Reject step |
---|
1505 | IF (RejectMoreH) THEN |
---|
1506 | Hnew = H*FacRej |
---|
1507 | END IF |
---|
1508 | RejectMoreH = RejectLastH |
---|
1509 | RejectLastH = .TRUE. |
---|
1510 | H = Hnew |
---|
1511 | IF (ISTATUS(Nacc) >= 1) THEN |
---|
1512 | ISTATUS(Nrej) = ISTATUS(Nrej) + 1 |
---|
1513 | END IF |
---|
1514 | END IF ! Err <= 1 |
---|
1515 | |
---|
1516 | END DO UntilAccepted |
---|
1517 | |
---|
1518 | END DO TimeLoop |
---|
1519 | |
---|
1520 | !~~~> Succesful exit |
---|
1521 | IERR = 1 !~~~> The integration was successful |
---|
1522 | |
---|
1523 | END SUBROUTINE ros_CadjInt |
---|
1524 | |
---|
1525 | |
---|
1526 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
1527 | SUBROUTINE ros_SimpleCadjInt ( & |
---|
1528 | NADJ, Y, & |
---|
1529 | Tstart, Tend, T, & |
---|
1530 | !~~~> Error indicator |
---|
1531 | IERR ) |
---|
1532 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
1533 | ! Template for the implementation of a generic RosenbrockADJ method |
---|
1534 | ! defined by ros_S (no of stages) |
---|
1535 | ! and its coefficients ros_{A,C,M,E,Alpha,Gamma} |
---|
1536 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
1537 | |
---|
1538 | IMPLICIT NONE |
---|
1539 | |
---|
1540 | !~~~> Input: the initial condition at Tstart; Output: the solution at T |
---|
1541 | INTEGER, INTENT(IN) :: NADJ |
---|
1542 | KPP_REAL, INTENT(INOUT) :: Y(NVAR,NADJ) |
---|
1543 | !~~~> Input: integration interval |
---|
1544 | KPP_REAL, INTENT(IN) :: Tstart,Tend |
---|
1545 | !~~~> Output: time at which the solution is returned (T=Tend if success) |
---|
1546 | KPP_REAL, INTENT(OUT) :: T |
---|
1547 | !~~~> Output: Error indicator |
---|
1548 | INTEGER, INTENT(OUT) :: IERR |
---|
1549 | ! ~~~~ Local variables |
---|
1550 | KPP_REAL :: Y0(NVAR) |
---|
1551 | KPP_REAL :: Ynew(NVAR,NADJ), Fcn0(NVAR,NADJ), Fcn(NVAR,NADJ) |
---|
1552 | KPP_REAL :: K(NVAR*ros_S,NADJ), dFdT(NVAR,NADJ) |
---|
1553 | #ifdef FULL_ALGEBRA |
---|
1554 | KPP_REAL,DIMENSION(NVAR,NVAR) :: Jac0, Ghimj, Jac, dJdT |
---|
1555 | #else |
---|
1556 | KPP_REAL,DIMENSION(LU_NONZERO) :: Jac0, Ghimj, Jac, dJdT |
---|
1557 | #endif |
---|
1558 | KPP_REAL :: H, HC, HG, Tau |
---|
1559 | KPP_REAL :: ghinv |
---|
1560 | INTEGER :: Pivot(NVAR), Direction, ioffset, i, j, istage, iadj |
---|
1561 | INTEGER :: istack |
---|
1562 | !~~~> Local parameters |
---|
1563 | KPP_REAL, PARAMETER :: ZERO = 0.0d0, ONE = 1.0d0 |
---|
1564 | KPP_REAL, PARAMETER :: DeltaMin = 1.0d-5 |
---|
1565 | !~~~> Locally called functions |
---|
1566 | ! KPP_REAL WLAMCH |
---|
1567 | ! EXTERNAL WLAMCH |
---|
1568 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
1569 | |
---|
1570 | |
---|
1571 | !~~~> INITIAL PREPARATIONS |
---|
1572 | |
---|
1573 | IF (Tend >= Tstart) THEN |
---|
1574 | Direction = -1 |
---|
1575 | ELSE |
---|
1576 | Direction = +1 |
---|
1577 | END IF |
---|
1578 | |
---|
1579 | !~~~> Time loop begins below |
---|
1580 | TimeLoop: DO istack = stack_ptr,2,-1 |
---|
1581 | |
---|
1582 | T = chk_T(istack) |
---|
1583 | H = chk_H(istack-1) |
---|
1584 | !CALL WCOPY(NVAR,chk_Y(1,istack),1,Y0,1) |
---|
1585 | Y0(1:NVAR) = chk_Y(1:NVAR,istack) |
---|
1586 | |
---|
1587 | !~~~> Compute the Jacobian at current time |
---|
1588 | CALL JacTemplate(T, Y0, Jac0) |
---|
1589 | ISTATUS(Njac) = ISTATUS(Njac) + 1 |
---|
1590 | |
---|
1591 | !~~~> Compute the function derivative with respect to T |
---|
1592 | IF (.NOT.Autonomous) THEN |
---|
1593 | CALL ros_JacTimeDerivative ( T, Roundoff, Y0, & |
---|
1594 | Jac0, dJdT ) |
---|
1595 | DO iadj = 1, NADJ |
---|
1596 | #ifdef FULL_ALGEBRA |
---|
1597 | dFdT(1:NVAR,iadj) = MATMUL(TRANSPOSE(dJdT),Y(1:NVAR,iadj)) |
---|
1598 | #else |
---|
1599 | CALL JacTR_SP_Vec(dJdT,Y(1,iadj),dFdT(1,iadj)) |
---|
1600 | #endif |
---|
1601 | CALL WSCAL(NVAR,(-ONE),dFdT(1,iadj),1) |
---|
1602 | END DO |
---|
1603 | END IF |
---|
1604 | |
---|
1605 | !~~~> Ydot = -J^T*Y |
---|
1606 | #ifdef FULL_ALGEBRA |
---|
1607 | Jac0(1:NVAR,1:NVAR) = -Jac0(1:NVAR,1:NVAR) |
---|
1608 | #else |
---|
1609 | CALL WSCAL(LU_NONZERO,(-ONE),Jac0,1) |
---|
1610 | #endif |
---|
1611 | DO iadj = 1, NADJ |
---|
1612 | #ifdef FULL_ALGEBRA |
---|
1613 | Fcn0(1:NVAR,iadj) = MATMUL(TRANSPOSE(Jac0),Y(1:NVAR,iadj)) |
---|
1614 | #else |
---|
1615 | CALL JacTR_SP_Vec(Jac0,Y(1,iadj),Fcn0(1,iadj)) |
---|
1616 | #endif |
---|
1617 | END DO |
---|
1618 | |
---|
1619 | !~~~> Construct Ghimj = 1/(H*ham) - Jac0 |
---|
1620 | ghinv = ONE/(Direction*H*ros_Gamma(1)) |
---|
1621 | #ifdef FULL_ALGEBRA |
---|
1622 | Ghimj(1:NVAR,1:NVAR) = -Jac0(1:NVAR,1:NVAR) |
---|
1623 | DO i=1,NVAR |
---|
1624 | Ghimj(i,i) = Ghimj(i,i)+ghinv |
---|
1625 | END DO |
---|
1626 | #else |
---|
1627 | CALL WCOPY(LU_NONZERO,Jac0,1,Ghimj,1) |
---|
1628 | CALL WSCAL(LU_NONZERO,(-ONE),Ghimj,1) |
---|
1629 | DO i=1,NVAR |
---|
1630 | Ghimj(LU_DIAG(i)) = Ghimj(LU_DIAG(i))+ghinv |
---|
1631 | END DO |
---|
1632 | #endif |
---|
1633 | !~~~> Compute LU decomposition |
---|
1634 | CALL ros_Decomp( Ghimj, Pivot, j ) |
---|
1635 | IF (j /= 0) THEN |
---|
1636 | CALL ros_ErrorMsg(-8,T,H,IERR) |
---|
1637 | PRINT*,' The matrix is singular !' |
---|
1638 | STOP |
---|
1639 | END IF |
---|
1640 | |
---|
1641 | !~~~> Compute the stages |
---|
1642 | Stage: DO istage = 1, ros_S |
---|
1643 | |
---|
1644 | ! Current istage offset. Current istage vector is K(ioffset+1:ioffset+NVAR) |
---|
1645 | ioffset = NVAR*(istage-1) |
---|
1646 | |
---|
1647 | ! For the 1st istage the function has been computed previously |
---|
1648 | IF ( istage == 1 ) THEN |
---|
1649 | DO iadj = 1, NADJ |
---|
1650 | CALL WCOPY(NVAR,Fcn0(1,iadj),1,Fcn(1,iadj),1) |
---|
1651 | END DO |
---|
1652 | ! istage>1 and a new function evaluation is needed at the current istage |
---|
1653 | ELSEIF ( ros_NewF(istage) ) THEN |
---|
1654 | CALL WCOPY(NVAR*NADJ,Y,1,Ynew,1) |
---|
1655 | DO j = 1, istage-1 |
---|
1656 | DO iadj = 1, NADJ |
---|
1657 | CALL WAXPY(NVAR,ros_A((istage-1)*(istage-2)/2+j), & |
---|
1658 | K(NVAR*(j-1)+1,iadj),1,Ynew(1,iadj),1) |
---|
1659 | END DO |
---|
1660 | END DO |
---|
1661 | Tau = T + ros_Alpha(istage)*Direction*H |
---|
1662 | CALL ros_Hermite3( chk_T(istack-1), chk_T(istack), Tau, & |
---|
1663 | chk_Y(1:NVAR,istack-1), chk_Y(1:NVAR,istack), & |
---|
1664 | chk_dY(1:NVAR,istack-1), chk_dY(1:NVAR,istack), Y0 ) |
---|
1665 | CALL JacTemplate(Tau, Y0, Jac) |
---|
1666 | ISTATUS(Njac) = ISTATUS(Njac) + 1 |
---|
1667 | #ifdef FULL_ALGEBRA |
---|
1668 | Jac(1:NVAR,1:NVAR) = -Jac(1:NVAR,1:NVAR) |
---|
1669 | #else |
---|
1670 | CALL WSCAL(LU_NONZERO,(-ONE),Jac,1) |
---|
1671 | #endif |
---|
1672 | DO iadj = 1, NADJ |
---|
1673 | #ifdef FULL_ALGEBRA |
---|
1674 | Fcn(1:NVAR,iadj) = MATMUL(TRANSPOSE(Jac),Ynew(1:NVAR,iadj)) |
---|
1675 | #else |
---|
1676 | CALL JacTR_SP_Vec(Jac,Ynew(1,iadj),Fcn(1,iadj)) |
---|
1677 | #endif |
---|
1678 | END DO |
---|
1679 | END IF ! if istage == 1 elseif ros_NewF(istage) |
---|
1680 | |
---|
1681 | DO iadj = 1, NADJ |
---|
1682 | CALL WCOPY(NVAR,Fcn(1,iadj),1,K(ioffset+1,iadj),1) |
---|
1683 | END DO |
---|
1684 | DO j = 1, istage-1 |
---|
1685 | HC = ros_C((istage-1)*(istage-2)/2+j)/(Direction*H) |
---|
1686 | DO iadj = 1, NADJ |
---|
1687 | CALL WAXPY(NVAR,HC,K(NVAR*(j-1)+1,iadj),1, & |
---|
1688 | K(ioffset+1,iadj),1) |
---|
1689 | END DO |
---|
1690 | END DO |
---|
1691 | IF ((.NOT. Autonomous).AND.(ros_Gamma(istage).NE.ZERO)) THEN |
---|
1692 | HG = Direction*H*ros_Gamma(istage) |
---|
1693 | DO iadj = 1, NADJ |
---|
1694 | CALL WAXPY(NVAR,HG,dFdT(1,iadj),1,K(ioffset+1,iadj),1) |
---|
1695 | END DO |
---|
1696 | END IF |
---|
1697 | DO iadj = 1, NADJ |
---|
1698 | CALL ros_Solve('T', Ghimj, Pivot, K(ioffset+1,iadj)) |
---|
1699 | END DO |
---|
1700 | |
---|
1701 | END DO Stage |
---|
1702 | |
---|
1703 | |
---|
1704 | !~~~> Compute the new solution |
---|
1705 | DO iadj = 1, NADJ |
---|
1706 | DO j=1,ros_S |
---|
1707 | CALL WAXPY(NVAR,ros_M(j),K(NVAR*(j-1)+1,iadj),1,Y(1,iadj),1) |
---|
1708 | END DO |
---|
1709 | END DO |
---|
1710 | |
---|
1711 | END DO TimeLoop |
---|
1712 | |
---|
1713 | !~~~> Succesful exit |
---|
1714 | IERR = 1 !~~~> The integration was successful |
---|
1715 | |
---|
1716 | END SUBROUTINE ros_SimpleCadjInt |
---|
1717 | |
---|
1718 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
1719 | KPP_REAL FUNCTION ros_ErrorNorm ( Y, Ynew, Yerr, & |
---|
1720 | AbsTol, RelTol, VectorTol ) |
---|
1721 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
1722 | !~~~> Computes the "scaled norm" of the error vector Yerr |
---|
1723 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
1724 | IMPLICIT NONE |
---|
1725 | |
---|
1726 | ! Input arguments |
---|
1727 | KPP_REAL, INTENT(IN) :: Y(NVAR), Ynew(NVAR), & |
---|
1728 | Yerr(NVAR), AbsTol(NVAR), RelTol(NVAR) |
---|
1729 | LOGICAL, INTENT(IN) :: VectorTol |
---|
1730 | ! Local variables |
---|
1731 | KPP_REAL :: Err, Scale, Ymax |
---|
1732 | INTEGER :: i |
---|
1733 | |
---|
1734 | Err = ZERO |
---|
1735 | DO i=1,NVAR |
---|
1736 | Ymax = MAX(ABS(Y(i)),ABS(Ynew(i))) |
---|
1737 | IF (VectorTol) THEN |
---|
1738 | Scale = AbsTol(i)+RelTol(i)*Ymax |
---|
1739 | ELSE |
---|
1740 | Scale = AbsTol(1)+RelTol(1)*Ymax |
---|
1741 | END IF |
---|
1742 | Err = Err+(Yerr(i)/Scale)**2 |
---|
1743 | END DO |
---|
1744 | Err = SQRT(Err/NVAR) |
---|
1745 | |
---|
1746 | ros_ErrorNorm = MAX(Err,1.0d-10) |
---|
1747 | |
---|
1748 | END FUNCTION ros_ErrorNorm |
---|
1749 | |
---|
1750 | |
---|
1751 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
1752 | SUBROUTINE ros_FunTimeDerivative ( T, Roundoff, Y, Fcn0, dFdT ) |
---|
1753 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
1754 | !~~~> The time partial derivative of the function by finite differences |
---|
1755 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
1756 | IMPLICIT NONE |
---|
1757 | |
---|
1758 | !~~~> Input arguments |
---|
1759 | KPP_REAL, INTENT(IN) :: T, Roundoff, Y(NVAR), Fcn0(NVAR) |
---|
1760 | !~~~> Output arguments |
---|
1761 | KPP_REAL, INTENT(OUT) :: dFdT(NVAR) |
---|
1762 | !~~~> Local variables |
---|
1763 | KPP_REAL :: Delta |
---|
1764 | KPP_REAL, PARAMETER :: ONE = 1.0d0, DeltaMin = 1.0d-6 |
---|
1765 | |
---|
1766 | Delta = SQRT(Roundoff)*MAX(DeltaMin,ABS(T)) |
---|
1767 | CALL FunTemplate(T+Delta,Y,dFdT) |
---|
1768 | ISTATUS(Nfun) = ISTATUS(Nfun) + 1 |
---|
1769 | CALL WAXPY(NVAR,(-ONE),Fcn0,1,dFdT,1) |
---|
1770 | CALL WSCAL(NVAR,(ONE/Delta),dFdT,1) |
---|
1771 | |
---|
1772 | END SUBROUTINE ros_FunTimeDerivative |
---|
1773 | |
---|
1774 | |
---|
1775 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
1776 | SUBROUTINE ros_JacTimeDerivative ( T, Roundoff, Y, & |
---|
1777 | Jac0, dJdT ) |
---|
1778 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
1779 | !~~~> The time partial derivative of the Jacobian by finite differences |
---|
1780 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
1781 | IMPLICIT NONE |
---|
1782 | |
---|
1783 | !~~~> Arguments |
---|
1784 | KPP_REAL, INTENT(IN) :: T, Roundoff, Y(NVAR) |
---|
1785 | #ifdef FULL_ALGEBRA |
---|
1786 | KPP_REAL, INTENT(IN) :: Jac0(NVAR,NVAR) |
---|
1787 | KPP_REAL, INTENT(OUT) :: dJdT(NVAR,NVAR) |
---|
1788 | #else |
---|
1789 | KPP_REAL, INTENT(IN) :: Jac0(LU_NONZERO) |
---|
1790 | KPP_REAL, INTENT(OUT) :: dJdT(LU_NONZERO) |
---|
1791 | #endif |
---|
1792 | !~~~> Local variables |
---|
1793 | KPP_REAL :: Delta |
---|
1794 | |
---|
1795 | Delta = SQRT(Roundoff)*MAX(DeltaMin,ABS(T)) |
---|
1796 | CALL JacTemplate(T+Delta,Y,dJdT) |
---|
1797 | ISTATUS(Njac) = ISTATUS(Njac) + 1 |
---|
1798 | #ifdef FULL_ALGEBRA |
---|
1799 | CALL WAXPY(NVAR*NVAR,(-ONE),Jac0,1,dJdT,1) |
---|
1800 | CALL WSCAL(NVAR*NVAR,(ONE/Delta),dJdT,1) |
---|
1801 | #else |
---|
1802 | CALL WAXPY(LU_NONZERO,(-ONE),Jac0,1,dJdT,1) |
---|
1803 | CALL WSCAL(LU_NONZERO,(ONE/Delta),dJdT,1) |
---|
1804 | #endif |
---|
1805 | |
---|
1806 | END SUBROUTINE ros_JacTimeDerivative |
---|
1807 | |
---|
1808 | |
---|
1809 | |
---|
1810 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
1811 | SUBROUTINE ros_PrepareMatrix ( H, Direction, gam, & |
---|
1812 | Jac0, Ghimj, Pivot, Singular ) |
---|
1813 | ! --- --- --- --- --- --- --- --- --- --- --- --- --- |
---|
1814 | ! Prepares the LHS matrix for stage calculations |
---|
1815 | ! 1. Construct Ghimj = 1/(H*gam) - Jac0 |
---|
1816 | ! "(Gamma H) Inverse Minus Jacobian" |
---|
1817 | ! 2. Repeat LU decomposition of Ghimj until successful. |
---|
1818 | ! -half the step size if LU decomposition fails and retry |
---|
1819 | ! -exit after 5 consecutive fails |
---|
1820 | ! --- --- --- --- --- --- --- --- --- --- --- --- --- |
---|
1821 | IMPLICIT NONE |
---|
1822 | |
---|
1823 | !~~~> Input arguments |
---|
1824 | #ifdef FULL_ALGEBRA |
---|
1825 | KPP_REAL, INTENT(IN) :: Jac0(NVAR,NVAR) |
---|
1826 | #else |
---|
1827 | KPP_REAL, INTENT(IN) :: Jac0(LU_NONZERO) |
---|
1828 | #endif |
---|
1829 | KPP_REAL, INTENT(IN) :: gam |
---|
1830 | INTEGER, INTENT(IN) :: Direction |
---|
1831 | !~~~> Output arguments |
---|
1832 | #ifdef FULL_ALGEBRA |
---|
1833 | KPP_REAL, INTENT(OUT) :: Ghimj(NVAR,NVAR) |
---|
1834 | #else |
---|
1835 | KPP_REAL, INTENT(OUT) :: Ghimj(LU_NONZERO) |
---|
1836 | #endif |
---|
1837 | LOGICAL, INTENT(OUT) :: Singular |
---|
1838 | INTEGER, INTENT(OUT) :: Pivot(NVAR) |
---|
1839 | !~~~> Inout arguments |
---|
1840 | KPP_REAL, INTENT(INOUT) :: H ! step size is decreased when LU fails |
---|
1841 | !~~~> Local variables |
---|
1842 | INTEGER :: i, ISING, Nconsecutive |
---|
1843 | KPP_REAL :: ghinv |
---|
1844 | KPP_REAL, PARAMETER :: ONE = 1.0_dp, HALF = 0.5_dp |
---|
1845 | |
---|
1846 | Nconsecutive = 0 |
---|
1847 | Singular = .TRUE. |
---|
1848 | |
---|
1849 | DO WHILE (Singular) |
---|
1850 | |
---|
1851 | !~~~> Construct Ghimj = 1/(H*gam) - Jac0 |
---|
1852 | #ifdef FULL_ALGEBRA |
---|
1853 | CALL WCOPY(NVAR*NVAR,Jac0,1,Ghimj,1) |
---|
1854 | CALL WSCAL(NVAR*NVAR,(-ONE),Ghimj,1) |
---|
1855 | ghinv = ONE/(Direction*H*gam) |
---|
1856 | DO i=1,NVAR |
---|
1857 | Ghimj(i,i) = Ghimj(i,i)+ghinv |
---|
1858 | END DO |
---|
1859 | #else |
---|
1860 | CALL WCOPY(LU_NONZERO,Jac0,1,Ghimj,1) |
---|
1861 | CALL WSCAL(LU_NONZERO,(-ONE),Ghimj,1) |
---|
1862 | ghinv = ONE/(Direction*H*gam) |
---|
1863 | DO i=1,NVAR |
---|
1864 | Ghimj(LU_DIAG(i)) = Ghimj(LU_DIAG(i))+ghinv |
---|
1865 | END DO |
---|
1866 | #endif |
---|
1867 | !~~~> Compute LU decomposition |
---|
1868 | CALL ros_Decomp( Ghimj, Pivot, ISING ) |
---|
1869 | IF (ISING == 0) THEN |
---|
1870 | !~~~> If successful done |
---|
1871 | Singular = .FALSE. |
---|
1872 | ELSE ! ISING .ne. 0 |
---|
1873 | !~~~> If unsuccessful half the step size; if 5 consecutive fails then return |
---|
1874 | ISTATUS(Nsng) = ISTATUS(Nsng) + 1 |
---|
1875 | Nconsecutive = Nconsecutive+1 |
---|
1876 | Singular = .TRUE. |
---|
1877 | PRINT*,'Warning: LU Decomposition returned ISING = ',ISING |
---|
1878 | IF (Nconsecutive <= 5) THEN ! Less than 5 consecutive failed decompositions |
---|
1879 | H = H*HALF |
---|
1880 | ELSE ! More than 5 consecutive failed decompositions |
---|
1881 | RETURN |
---|
1882 | END IF ! Nconsecutive |
---|
1883 | END IF ! ISING |
---|
1884 | |
---|
1885 | END DO ! WHILE Singular |
---|
1886 | |
---|
1887 | END SUBROUTINE ros_PrepareMatrix |
---|
1888 | |
---|
1889 | |
---|
1890 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
1891 | SUBROUTINE ros_Decomp( A, Pivot, ISING ) |
---|
1892 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
1893 | ! Template for the LU decomposition |
---|
1894 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
1895 | IMPLICIT NONE |
---|
1896 | !~~~> Inout variables |
---|
1897 | #ifdef FULL_ALGEBRA |
---|
1898 | KPP_REAL, INTENT(INOUT) :: A(NVAR,NVAR) |
---|
1899 | #else |
---|
1900 | KPP_REAL, INTENT(INOUT) :: A(LU_NONZERO) |
---|
1901 | #endif |
---|
1902 | !~~~> Output variables |
---|
1903 | INTEGER, INTENT(OUT) :: Pivot(NVAR), ISING |
---|
1904 | |
---|
1905 | #ifdef FULL_ALGEBRA |
---|
1906 | CALL DGETRF( NVAR, NVAR, A, NVAR, Pivot, ISING ) |
---|
1907 | #else |
---|
1908 | CALL KppDecomp ( A, ISING ) |
---|
1909 | Pivot(1) = 1 |
---|
1910 | #endif |
---|
1911 | ISTATUS(Ndec) = ISTATUS(Ndec) + 1 |
---|
1912 | |
---|
1913 | END SUBROUTINE ros_Decomp |
---|
1914 | |
---|
1915 | |
---|
1916 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
1917 | SUBROUTINE ros_Solve( How, A, Pivot, b ) |
---|
1918 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
1919 | ! Template for the forward/backward substitution (using pre-computed LU decomposition) |
---|
1920 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
1921 | IMPLICIT NONE |
---|
1922 | !~~~> Input variables |
---|
1923 | CHARACTER, INTENT(IN) :: How |
---|
1924 | #ifdef FULL_ALGEBRA |
---|
1925 | KPP_REAL, INTENT(IN) :: A(NVAR,NVAR) |
---|
1926 | INTEGER :: ISING |
---|
1927 | #else |
---|
1928 | KPP_REAL, INTENT(IN) :: A(LU_NONZERO) |
---|
1929 | #endif |
---|
1930 | INTEGER, INTENT(IN) :: Pivot(NVAR) |
---|
1931 | !~~~> InOut variables |
---|
1932 | KPP_REAL, INTENT(INOUT) :: b(NVAR) |
---|
1933 | |
---|
1934 | SELECT CASE (How) |
---|
1935 | CASE ('N') |
---|
1936 | #ifdef FULL_ALGEBRA |
---|
1937 | CALL DGETRS( 'N', NVAR , 1, A, NVAR, Pivot, b, NVAR, ISING ) |
---|
1938 | #else |
---|
1939 | CALL KppSolve( A, b ) |
---|
1940 | #endif |
---|
1941 | CASE ('T') |
---|
1942 | #ifdef FULL_ALGEBRA |
---|
1943 | CALL DGETRS( 'T', NVAR , 1, A, NVAR, Pivot, b, NVAR, ISING ) |
---|
1944 | #else |
---|
1945 | CALL KppSolveTR( A, b, b ) |
---|
1946 | #endif |
---|
1947 | CASE DEFAULT |
---|
1948 | PRINT*,'Error: unknown argument in ros_Solve: How=',How |
---|
1949 | STOP |
---|
1950 | END SELECT |
---|
1951 | ISTATUS(Nsol) = ISTATUS(Nsol) + 1 |
---|
1952 | |
---|
1953 | END SUBROUTINE ros_Solve |
---|
1954 | |
---|
1955 | |
---|
1956 | |
---|
1957 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
1958 | SUBROUTINE ros_cadj_Y( T, Y ) |
---|
1959 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
1960 | ! Finds the solution Y at T by interpolating the stored forward trajectory |
---|
1961 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
1962 | IMPLICIT NONE |
---|
1963 | !~~~> Input variables |
---|
1964 | KPP_REAL, INTENT(IN) :: T |
---|
1965 | !~~~> Output variables |
---|
1966 | KPP_REAL, INTENT(OUT) :: Y(NVAR) |
---|
1967 | !~~~> Local variables |
---|
1968 | INTEGER :: i |
---|
1969 | KPP_REAL, PARAMETER :: ONE = 1.0d0 |
---|
1970 | |
---|
1971 | ! chk_H, chk_T, chk_Y, chk_dY, chk_d2Y |
---|
1972 | |
---|
1973 | IF( (T < chk_T(1)).OR.(T> chk_T(stack_ptr)) ) THEN |
---|
1974 | PRINT*,'Cannot locate solution at T = ',T |
---|
1975 | PRINT*,'Stored trajectory is between Tstart = ',chk_T(1) |
---|
1976 | PRINT*,' and Tend = ',chk_T(stack_ptr) |
---|
1977 | STOP |
---|
1978 | END IF |
---|
1979 | DO i = 1, stack_ptr-1 |
---|
1980 | IF( (T>= chk_T(i)).AND.(T<= chk_T(i+1)) ) EXIT |
---|
1981 | END DO |
---|
1982 | |
---|
1983 | |
---|
1984 | ! IF (.FALSE.) THEN |
---|
1985 | ! |
---|
1986 | ! CALL ros_Hermite5( chk_T(i), chk_T(i+1), T, & |
---|
1987 | ! chk_Y(1,i), chk_Y(1,i+1), & |
---|
1988 | ! chk_dY(1,i), chk_dY(1,i+1), & |
---|
1989 | ! chk_d2Y(1,i), chk_d2Y(1,i+1), Y ) |
---|
1990 | ! |
---|
1991 | ! ELSE |
---|
1992 | |
---|
1993 | CALL ros_Hermite3( chk_T(i), chk_T(i+1), T, & |
---|
1994 | chk_Y(1:NVAR,i), chk_Y(1:NVAR,i+1), & |
---|
1995 | chk_dY(1:NVAR,i), chk_dY(1:NVAR,i+1), & |
---|
1996 | Y ) |
---|
1997 | |
---|
1998 | ! |
---|
1999 | ! END IF |
---|
2000 | |
---|
2001 | END SUBROUTINE ros_cadj_Y |
---|
2002 | |
---|
2003 | |
---|
2004 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
2005 | SUBROUTINE ros_Hermite3( a, b, T, Ya, Yb, Ja, Jb, Y ) |
---|
2006 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
2007 | ! Template for Hermite interpolation of order 5 on the interval [a,b] |
---|
2008 | ! P = c(1) + c(2)*(x-a) + ... + c(4)*(x-a)^3 |
---|
2009 | ! P[a,b] = [Ya,Yb], P'[a,b] = [Ja,Jb] |
---|
2010 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
2011 | IMPLICIT NONE |
---|
2012 | !~~~> Input variables |
---|
2013 | KPP_REAL, INTENT(IN) :: a, b, T, Ya(NVAR), Yb(NVAR) |
---|
2014 | KPP_REAL, INTENT(IN) :: Ja(NVAR), Jb(NVAR) |
---|
2015 | !~~~> Output variables |
---|
2016 | KPP_REAL, INTENT(OUT) :: Y(NVAR) |
---|
2017 | !~~~> Local variables |
---|
2018 | KPP_REAL :: Tau, amb(3), C(NVAR,4) |
---|
2019 | KPP_REAL, PARAMETER :: ZERO = 0.0d0 |
---|
2020 | INTEGER :: i, j |
---|
2021 | |
---|
2022 | amb(1) = 1.0d0/(a-b) |
---|
2023 | DO i=2,3 |
---|
2024 | amb(i) = amb(i-1)*amb(1) |
---|
2025 | END DO |
---|
2026 | |
---|
2027 | |
---|
2028 | ! c(1) = ya; |
---|
2029 | CALL WCOPY(NVAR,Ya,1,C(1,1),1) |
---|
2030 | ! c(2) = ja; |
---|
2031 | CALL WCOPY(NVAR,Ja,1,C(1,2),1) |
---|
2032 | ! c(3) = 2/(a-b)*ja + 1/(a-b)*jb - 3/(a - b)^2*ya + 3/(a - b)^2*yb ; |
---|
2033 | CALL WCOPY(NVAR,Ya,1,C(1,3),1) |
---|
2034 | CALL WSCAL(NVAR,-3.0*amb(2),C(1,3),1) |
---|
2035 | CALL WAXPY(NVAR,3.0*amb(2),Yb,1,C(1,3),1) |
---|
2036 | CALL WAXPY(NVAR,2.0*amb(1),Ja,1,C(1,3),1) |
---|
2037 | CALL WAXPY(NVAR,amb(1),Jb,1,C(1,3),1) |
---|
2038 | ! c(4) = 1/(a-b)^2*ja + 1/(a-b)^2*jb - 2/(a-b)^3*ya + 2/(a-b)^3*yb ; |
---|
2039 | CALL WCOPY(NVAR,Ya,1,C(1,4),1) |
---|
2040 | CALL WSCAL(NVAR,-2.0*amb(3),C(1,4),1) |
---|
2041 | CALL WAXPY(NVAR,2.0*amb(3),Yb,1,C(1,4),1) |
---|
2042 | CALL WAXPY(NVAR,amb(2),Ja,1,C(1,4),1) |
---|
2043 | CALL WAXPY(NVAR,amb(2),Jb,1,C(1,4),1) |
---|
2044 | |
---|
2045 | Tau = T - a |
---|
2046 | CALL WCOPY(NVAR,C(1,4),1,Y,1) |
---|
2047 | CALL WSCAL(NVAR,Tau**3,Y,1) |
---|
2048 | DO j = 3,1,-1 |
---|
2049 | CALL WAXPY(NVAR,TAU**(j-1),C(1,j),1,Y,1) |
---|
2050 | END DO |
---|
2051 | |
---|
2052 | END SUBROUTINE ros_Hermite3 |
---|
2053 | |
---|
2054 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
2055 | SUBROUTINE ros_Hermite5( a, b, T, Ya, Yb, Ja, Jb, Ha, Hb, Y ) |
---|
2056 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
2057 | ! Template for Hermite interpolation of order 5 on the interval [a,b] |
---|
2058 | ! P = c(1) + c(2)*(x-a) + ... + c(6)*(x-a)^5 |
---|
2059 | ! P[a,b] = [Ya,Yb], P'[a,b] = [Ja,Jb], P"[a,b] = [Ha,Hb] |
---|
2060 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
2061 | IMPLICIT NONE |
---|
2062 | !~~~> Input variables |
---|
2063 | KPP_REAL, INTENT(IN) :: a, b, T, Ya(NVAR), Yb(NVAR) |
---|
2064 | KPP_REAL, INTENT(IN) :: Ja(NVAR), Jb(NVAR), Ha(NVAR), Hb(NVAR) |
---|
2065 | !~~~> Output variables |
---|
2066 | KPP_REAL, INTENT(OUT) :: Y(NVAR) |
---|
2067 | !~~~> Local variables |
---|
2068 | KPP_REAL :: Tau, amb(5), C(NVAR,6) |
---|
2069 | KPP_REAL, PARAMETER :: ZERO = 0.0d0, HALF = 0.5d0 |
---|
2070 | INTEGER :: i, j |
---|
2071 | |
---|
2072 | amb(1) = 1.0d0/(a-b) |
---|
2073 | DO i=2,5 |
---|
2074 | amb(i) = amb(i-1)*amb(1) |
---|
2075 | END DO |
---|
2076 | |
---|
2077 | ! c(1) = ya; |
---|
2078 | CALL WCOPY(NVAR,Ya,1,C(1,1),1) |
---|
2079 | ! c(2) = ja; |
---|
2080 | CALL WCOPY(NVAR,Ja,1,C(1,2),1) |
---|
2081 | ! c(3) = ha/2; |
---|
2082 | CALL WCOPY(NVAR,Ha,1,C(1,3),1) |
---|
2083 | CALL WSCAL(NVAR,HALF,C(1,3),1) |
---|
2084 | |
---|
2085 | ! c(4) = 10*amb(3)*ya - 10*amb(3)*yb - 6*amb(2)*ja - 4*amb(2)*jb + 1.5*amb(1)*ha - 0.5*amb(1)*hb ; |
---|
2086 | CALL WCOPY(NVAR,Ya,1,C(1,4),1) |
---|
2087 | CALL WSCAL(NVAR,10.0*amb(3),C(1,4),1) |
---|
2088 | CALL WAXPY(NVAR,-10.0*amb(3),Yb,1,C(1,4),1) |
---|
2089 | CALL WAXPY(NVAR,-6.0*amb(2),Ja,1,C(1,4),1) |
---|
2090 | CALL WAXPY(NVAR,-4.0*amb(2),Jb,1,C(1,4),1) |
---|
2091 | CALL WAXPY(NVAR, 1.5*amb(1),Ha,1,C(1,4),1) |
---|
2092 | CALL WAXPY(NVAR,-0.5*amb(1),Hb,1,C(1,4),1) |
---|
2093 | |
---|
2094 | ! c(5) = 15*amb(4)*ya - 15*amb(4)*yb - 8.*amb(3)*ja - 7*amb(3)*jb + 1.5*amb(2)*ha - 1*amb(2)*hb ; |
---|
2095 | CALL WCOPY(NVAR,Ya,1,C(1,5),1) |
---|
2096 | CALL WSCAL(NVAR, 15.0*amb(4),C(1,5),1) |
---|
2097 | CALL WAXPY(NVAR,-15.0*amb(4),Yb,1,C(1,5),1) |
---|
2098 | CALL WAXPY(NVAR,-8.0*amb(3),Ja,1,C(1,5),1) |
---|
2099 | CALL WAXPY(NVAR,-7.0*amb(3),Jb,1,C(1,5),1) |
---|
2100 | CALL WAXPY(NVAR,1.5*amb(2),Ha,1,C(1,5),1) |
---|
2101 | CALL WAXPY(NVAR,-amb(2),Hb,1,C(1,5),1) |
---|
2102 | |
---|
2103 | ! c(6) = 6*amb(5)*ya - 6*amb(5)*yb - 3.*amb(4)*ja - 3.*amb(4)*jb + 0.5*amb(3)*ha -0.5*amb(3)*hb ; |
---|
2104 | CALL WCOPY(NVAR,Ya,1,C(1,6),1) |
---|
2105 | CALL WSCAL(NVAR, 6.0*amb(5),C(1,6),1) |
---|
2106 | CALL WAXPY(NVAR,-6.0*amb(5),Yb,1,C(1,6),1) |
---|
2107 | CALL WAXPY(NVAR,-3.0*amb(4),Ja,1,C(1,6),1) |
---|
2108 | CALL WAXPY(NVAR,-3.0*amb(4),Jb,1,C(1,6),1) |
---|
2109 | CALL WAXPY(NVAR, 0.5*amb(3),Ha,1,C(1,6),1) |
---|
2110 | CALL WAXPY(NVAR,-0.5*amb(3),Hb,1,C(1,6),1) |
---|
2111 | |
---|
2112 | Tau = T - a |
---|
2113 | CALL WCOPY(NVAR,C(1,6),1,Y,1) |
---|
2114 | DO j = 5,1,-1 |
---|
2115 | CALL WSCAL(NVAR,Tau,Y,1) |
---|
2116 | CALL WAXPY(NVAR,ONE,C(1,j),1,Y,1) |
---|
2117 | END DO |
---|
2118 | |
---|
2119 | END SUBROUTINE ros_Hermite5 |
---|
2120 | |
---|
2121 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
2122 | SUBROUTINE Ros2 |
---|
2123 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
2124 | ! --- AN L-STABLE METHOD, 2 stages, order 2 |
---|
2125 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
2126 | |
---|
2127 | IMPLICIT NONE |
---|
2128 | DOUBLE PRECISION g |
---|
2129 | |
---|
2130 | g = 1.0d0 + 1.0d0/SQRT(2.0d0) |
---|
2131 | |
---|
2132 | rosMethod = RS2 |
---|
2133 | !~~~> Name of the method |
---|
2134 | ros_Name = 'ROS-2' |
---|
2135 | !~~~> Number of stages |
---|
2136 | ros_S = 2 |
---|
2137 | |
---|
2138 | !~~~> The coefficient matrices A and C are strictly lower triangular. |
---|
2139 | ! The lower triangular (subdiagonal) elements are stored in row-wise order: |
---|
2140 | ! A(2,1) = ros_A(1), A(3,1)=ros_A(2), A(3,2)=ros_A(3), etc. |
---|
2141 | ! The general mapping formula is: |
---|
2142 | ! A(i,j) = ros_A( (i-1)*(i-2)/2 + j ) |
---|
2143 | ! C(i,j) = ros_C( (i-1)*(i-2)/2 + j ) |
---|
2144 | |
---|
2145 | ros_A(1) = (1.d0)/g |
---|
2146 | ros_C(1) = (-2.d0)/g |
---|
2147 | !~~~> Does the stage i require a new function evaluation (ros_NewF(i)=TRUE) |
---|
2148 | ! or does it re-use the function evaluation from stage i-1 (ros_NewF(i)=FALSE) |
---|
2149 | ros_NewF(1) = .TRUE. |
---|
2150 | ros_NewF(2) = .TRUE. |
---|
2151 | !~~~> M_i = Coefficients for new step solution |
---|
2152 | ros_M(1)= (3.d0)/(2.d0*g) |
---|
2153 | ros_M(2)= (1.d0)/(2.d0*g) |
---|
2154 | ! E_i = Coefficients for error estimator |
---|
2155 | ros_E(1) = 1.d0/(2.d0*g) |
---|
2156 | ros_E(2) = 1.d0/(2.d0*g) |
---|
2157 | !~~~> ros_ELO = estimator of local order - the minimum between the |
---|
2158 | ! main and the embedded scheme orders plus one |
---|
2159 | ros_ELO = 2.0d0 |
---|
2160 | !~~~> Y_stage_i ~ Y( T + H*Alpha_i ) |
---|
2161 | ros_Alpha(1) = 0.0d0 |
---|
2162 | ros_Alpha(2) = 1.0d0 |
---|
2163 | !~~~> Gamma_i = \sum_j gamma_{i,j} |
---|
2164 | ros_Gamma(1) = g |
---|
2165 | ros_Gamma(2) =-g |
---|
2166 | |
---|
2167 | END SUBROUTINE Ros2 |
---|
2168 | |
---|
2169 | |
---|
2170 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
2171 | SUBROUTINE Ros3 |
---|
2172 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
2173 | ! --- AN L-STABLE METHOD, 3 stages, order 3, 2 function evaluations |
---|
2174 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
2175 | |
---|
2176 | IMPLICIT NONE |
---|
2177 | |
---|
2178 | rosMethod = RS3 |
---|
2179 | !~~~> Name of the method |
---|
2180 | ros_Name = 'ROS-3' |
---|
2181 | !~~~> Number of stages |
---|
2182 | ros_S = 3 |
---|
2183 | |
---|
2184 | !~~~> The coefficient matrices A and C are strictly lower triangular. |
---|
2185 | ! The lower triangular (subdiagonal) elements are stored in row-wise order: |
---|
2186 | ! A(2,1) = ros_A(1), A(3,1)=ros_A(2), A(3,2)=ros_A(3), etc. |
---|
2187 | ! The general mapping formula is: |
---|
2188 | ! A(i,j) = ros_A( (i-1)*(i-2)/2 + j ) |
---|
2189 | ! C(i,j) = ros_C( (i-1)*(i-2)/2 + j ) |
---|
2190 | |
---|
2191 | ros_A(1)= 1.d0 |
---|
2192 | ros_A(2)= 1.d0 |
---|
2193 | ros_A(3)= 0.d0 |
---|
2194 | |
---|
2195 | ros_C(1) = -0.10156171083877702091975600115545d+01 |
---|
2196 | ros_C(2) = 0.40759956452537699824805835358067d+01 |
---|
2197 | ros_C(3) = 0.92076794298330791242156818474003d+01 |
---|
2198 | !~~~> Does the stage i require a new function evaluation (ros_NewF(i)=TRUE) |
---|
2199 | ! or does it re-use the function evaluation from stage i-1 (ros_NewF(i)=FALSE) |
---|
2200 | ros_NewF(1) = .TRUE. |
---|
2201 | ros_NewF(2) = .TRUE. |
---|
2202 | ros_NewF(3) = .FALSE. |
---|
2203 | !~~~> M_i = Coefficients for new step solution |
---|
2204 | ros_M(1) = 0.1d+01 |
---|
2205 | ros_M(2) = 0.61697947043828245592553615689730d+01 |
---|
2206 | ros_M(3) = -0.42772256543218573326238373806514d+00 |
---|
2207 | ! E_i = Coefficients for error estimator |
---|
2208 | ros_E(1) = 0.5d+00 |
---|
2209 | ros_E(2) = -0.29079558716805469821718236208017d+01 |
---|
2210 | ros_E(3) = 0.22354069897811569627360909276199d+00 |
---|
2211 | !~~~> ros_ELO = estimator of local order - the minimum between the |
---|
2212 | ! main and the embedded scheme orders plus 1 |
---|
2213 | ros_ELO = 3.0d0 |
---|
2214 | !~~~> Y_stage_i ~ Y( T + H*Alpha_i ) |
---|
2215 | ros_Alpha(1)= 0.0d+00 |
---|
2216 | ros_Alpha(2)= 0.43586652150845899941601945119356d+00 |
---|
2217 | ros_Alpha(3)= 0.43586652150845899941601945119356d+00 |
---|
2218 | !~~~> Gamma_i = \sum_j gamma_{i,j} |
---|
2219 | ros_Gamma(1)= 0.43586652150845899941601945119356d+00 |
---|
2220 | ros_Gamma(2)= 0.24291996454816804366592249683314d+00 |
---|
2221 | ros_Gamma(3)= 0.21851380027664058511513169485832d+01 |
---|
2222 | |
---|
2223 | END SUBROUTINE Ros3 |
---|
2224 | |
---|
2225 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
2226 | |
---|
2227 | |
---|
2228 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
2229 | SUBROUTINE Ros4 |
---|
2230 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
2231 | ! L-STABLE ROSENBROCK METHOD OF ORDER 4, WITH 4 STAGES |
---|
2232 | ! L-STABLE EMBEDDED ROSENBROCK METHOD OF ORDER 3 |
---|
2233 | ! |
---|
2234 | ! E. HAIRER AND G. WANNER, SOLVING ORDINARY DIFFERENTIAL |
---|
2235 | ! EQUATIONS II. STIFF AND DIFFERENTIAL-ALGEBRAIC PROBLEMS. |
---|
2236 | ! SPRINGER SERIES IN COMPUTATIONAL MATHEMATICS, |
---|
2237 | ! SPRINGER-VERLAG (1990) |
---|
2238 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
2239 | |
---|
2240 | IMPLICIT NONE |
---|
2241 | |
---|
2242 | rosMethod = RS4 |
---|
2243 | !~~~> Name of the method |
---|
2244 | ros_Name = 'ROS-4' |
---|
2245 | !~~~> Number of stages |
---|
2246 | ros_S = 4 |
---|
2247 | |
---|
2248 | !~~~> The coefficient matrices A and C are strictly lower triangular. |
---|
2249 | ! The lower triangular (subdiagonal) elements are stored in row-wise order: |
---|
2250 | ! A(2,1) = ros_A(1), A(3,1)=ros_A(2), A(3,2)=ros_A(3), etc. |
---|
2251 | ! The general mapping formula is: |
---|
2252 | ! A(i,j) = ros_A( (i-1)*(i-2)/2 + j ) |
---|
2253 | ! C(i,j) = ros_C( (i-1)*(i-2)/2 + j ) |
---|
2254 | |
---|
2255 | ros_A(1) = 0.2000000000000000d+01 |
---|
2256 | ros_A(2) = 0.1867943637803922d+01 |
---|
2257 | ros_A(3) = 0.2344449711399156d+00 |
---|
2258 | ros_A(4) = ros_A(2) |
---|
2259 | ros_A(5) = ros_A(3) |
---|
2260 | ros_A(6) = 0.0D0 |
---|
2261 | |
---|
2262 | ros_C(1) =-0.7137615036412310d+01 |
---|
2263 | ros_C(2) = 0.2580708087951457d+01 |
---|
2264 | ros_C(3) = 0.6515950076447975d+00 |
---|
2265 | ros_C(4) =-0.2137148994382534d+01 |
---|
2266 | ros_C(5) =-0.3214669691237626d+00 |
---|
2267 | ros_C(6) =-0.6949742501781779d+00 |
---|
2268 | !~~~> Does the stage i require a new function evaluation (ros_NewF(i)=TRUE) |
---|
2269 | ! or does it re-use the function evaluation from stage i-1 (ros_NewF(i)=FALSE) |
---|
2270 | ros_NewF(1) = .TRUE. |
---|
2271 | ros_NewF(2) = .TRUE. |
---|
2272 | ros_NewF(3) = .TRUE. |
---|
2273 | ros_NewF(4) = .FALSE. |
---|
2274 | !~~~> M_i = Coefficients for new step solution |
---|
2275 | ros_M(1) = 0.2255570073418735d+01 |
---|
2276 | ros_M(2) = 0.2870493262186792d+00 |
---|
2277 | ros_M(3) = 0.4353179431840180d+00 |
---|
2278 | ros_M(4) = 0.1093502252409163d+01 |
---|
2279 | !~~~> E_i = Coefficients for error estimator |
---|
2280 | ros_E(1) =-0.2815431932141155d+00 |
---|
2281 | ros_E(2) =-0.7276199124938920d-01 |
---|
2282 | ros_E(3) =-0.1082196201495311d+00 |
---|
2283 | ros_E(4) =-0.1093502252409163d+01 |
---|
2284 | !~~~> ros_ELO = estimator of local order - the minimum between the |
---|
2285 | ! main and the embedded scheme orders plus 1 |
---|
2286 | ros_ELO = 4.0d0 |
---|
2287 | !~~~> Y_stage_i ~ Y( T + H*Alpha_i ) |
---|
2288 | ros_Alpha(1) = 0.D0 |
---|
2289 | ros_Alpha(2) = 0.1145640000000000d+01 |
---|
2290 | ros_Alpha(3) = 0.6552168638155900d+00 |
---|
2291 | ros_Alpha(4) = ros_Alpha(3) |
---|
2292 | !~~~> Gamma_i = \sum_j gamma_{i,j} |
---|
2293 | ros_Gamma(1) = 0.5728200000000000d+00 |
---|
2294 | ros_Gamma(2) =-0.1769193891319233d+01 |
---|
2295 | ros_Gamma(3) = 0.7592633437920482d+00 |
---|
2296 | ros_Gamma(4) =-0.1049021087100450d+00 |
---|
2297 | |
---|
2298 | END SUBROUTINE Ros4 |
---|
2299 | |
---|
2300 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
2301 | SUBROUTINE Rodas3 |
---|
2302 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
2303 | ! --- A STIFFLY-STABLE METHOD, 4 stages, order 3 |
---|
2304 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
2305 | |
---|
2306 | IMPLICIT NONE |
---|
2307 | |
---|
2308 | rosMethod = RD3 |
---|
2309 | !~~~> Name of the method |
---|
2310 | ros_Name = 'RODAS-3' |
---|
2311 | !~~~> Number of stages |
---|
2312 | ros_S = 4 |
---|
2313 | |
---|
2314 | !~~~> The coefficient matrices A and C are strictly lower triangular. |
---|
2315 | ! The lower triangular (subdiagonal) elements are stored in row-wise order: |
---|
2316 | ! A(2,1) = ros_A(1), A(3,1)=ros_A(2), A(3,2)=ros_A(3), etc. |
---|
2317 | ! The general mapping formula is: |
---|
2318 | ! A(i,j) = ros_A( (i-1)*(i-2)/2 + j ) |
---|
2319 | ! C(i,j) = ros_C( (i-1)*(i-2)/2 + j ) |
---|
2320 | |
---|
2321 | ros_A(1) = 0.0d+00 |
---|
2322 | ros_A(2) = 2.0d+00 |
---|
2323 | ros_A(3) = 0.0d+00 |
---|
2324 | ros_A(4) = 2.0d+00 |
---|
2325 | ros_A(5) = 0.0d+00 |
---|
2326 | ros_A(6) = 1.0d+00 |
---|
2327 | |
---|
2328 | ros_C(1) = 4.0d+00 |
---|
2329 | ros_C(2) = 1.0d+00 |
---|
2330 | ros_C(3) =-1.0d+00 |
---|
2331 | ros_C(4) = 1.0d+00 |
---|
2332 | ros_C(5) =-1.0d+00 |
---|
2333 | ros_C(6) =-(8.0d+00/3.0d+00) |
---|
2334 | |
---|
2335 | !~~~> Does the stage i require a new function evaluation (ros_NewF(i)=TRUE) |
---|
2336 | ! or does it re-use the function evaluation from stage i-1 (ros_NewF(i)=FALSE) |
---|
2337 | ros_NewF(1) = .TRUE. |
---|
2338 | ros_NewF(2) = .FALSE. |
---|
2339 | ros_NewF(3) = .TRUE. |
---|
2340 | ros_NewF(4) = .TRUE. |
---|
2341 | !~~~> M_i = Coefficients for new step solution |
---|
2342 | ros_M(1) = 2.0d+00 |
---|
2343 | ros_M(2) = 0.0d+00 |
---|
2344 | ros_M(3) = 1.0d+00 |
---|
2345 | ros_M(4) = 1.0d+00 |
---|
2346 | !~~~> E_i = Coefficients for error estimator |
---|
2347 | ros_E(1) = 0.0d+00 |
---|
2348 | ros_E(2) = 0.0d+00 |
---|
2349 | ros_E(3) = 0.0d+00 |
---|
2350 | ros_E(4) = 1.0d+00 |
---|
2351 | !~~~> ros_ELO = estimator of local order - the minimum between the |
---|
2352 | ! main and the embedded scheme orders plus 1 |
---|
2353 | ros_ELO = 3.0d+00 |
---|
2354 | !~~~> Y_stage_i ~ Y( T + H*Alpha_i ) |
---|
2355 | ros_Alpha(1) = 0.0d+00 |
---|
2356 | ros_Alpha(2) = 0.0d+00 |
---|
2357 | ros_Alpha(3) = 1.0d+00 |
---|
2358 | ros_Alpha(4) = 1.0d+00 |
---|
2359 | !~~~> Gamma_i = \sum_j gamma_{i,j} |
---|
2360 | ros_Gamma(1) = 0.5d+00 |
---|
2361 | ros_Gamma(2) = 1.5d+00 |
---|
2362 | ros_Gamma(3) = 0.0d+00 |
---|
2363 | ros_Gamma(4) = 0.0d+00 |
---|
2364 | |
---|
2365 | END SUBROUTINE Rodas3 |
---|
2366 | |
---|
2367 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
2368 | SUBROUTINE Rodas4 |
---|
2369 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
2370 | ! STIFFLY-STABLE ROSENBROCK METHOD OF ORDER 4, WITH 6 STAGES |
---|
2371 | ! |
---|
2372 | ! E. HAIRER AND G. WANNER, SOLVING ORDINARY DIFFERENTIAL |
---|
2373 | ! EQUATIONS II. STIFF AND DIFFERENTIAL-ALGEBRAIC PROBLEMS. |
---|
2374 | ! SPRINGER SERIES IN COMPUTATIONAL MATHEMATICS, |
---|
2375 | ! SPRINGER-VERLAG (1996) |
---|
2376 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
2377 | |
---|
2378 | IMPLICIT NONE |
---|
2379 | |
---|
2380 | rosMethod = RD4 |
---|
2381 | !~~~> Name of the method |
---|
2382 | ros_Name = 'RODAS-4' |
---|
2383 | !~~~> Number of stages |
---|
2384 | ros_S = 6 |
---|
2385 | |
---|
2386 | !~~~> Y_stage_i ~ Y( T + H*Alpha_i ) |
---|
2387 | ros_Alpha(1) = 0.000d0 |
---|
2388 | ros_Alpha(2) = 0.386d0 |
---|
2389 | ros_Alpha(3) = 0.210d0 |
---|
2390 | ros_Alpha(4) = 0.630d0 |
---|
2391 | ros_Alpha(5) = 1.000d0 |
---|
2392 | ros_Alpha(6) = 1.000d0 |
---|
2393 | |
---|
2394 | !~~~> Gamma_i = \sum_j gamma_{i,j} |
---|
2395 | ros_Gamma(1) = 0.2500000000000000d+00 |
---|
2396 | ros_Gamma(2) =-0.1043000000000000d+00 |
---|
2397 | ros_Gamma(3) = 0.1035000000000000d+00 |
---|
2398 | ros_Gamma(4) =-0.3620000000000023d-01 |
---|
2399 | ros_Gamma(5) = 0.0d0 |
---|
2400 | ros_Gamma(6) = 0.0d0 |
---|
2401 | |
---|
2402 | !~~~> The coefficient matrices A and C are strictly lower triangular. |
---|
2403 | ! The lower triangular (subdiagonal) elements are stored in row-wise order: |
---|
2404 | ! A(2,1) = ros_A(1), A(3,1)=ros_A(2), A(3,2)=ros_A(3), etc. |
---|
2405 | ! The general mapping formula is: A(i,j) = ros_A( (i-1)*(i-2)/2 + j ) |
---|
2406 | ! C(i,j) = ros_C( (i-1)*(i-2)/2 + j ) |
---|
2407 | |
---|
2408 | ros_A(1) = 0.1544000000000000d+01 |
---|
2409 | ros_A(2) = 0.9466785280815826d+00 |
---|
2410 | ros_A(3) = 0.2557011698983284d+00 |
---|
2411 | ros_A(4) = 0.3314825187068521d+01 |
---|
2412 | ros_A(5) = 0.2896124015972201d+01 |
---|
2413 | ros_A(6) = 0.9986419139977817d+00 |
---|
2414 | ros_A(7) = 0.1221224509226641d+01 |
---|
2415 | ros_A(8) = 0.6019134481288629d+01 |
---|
2416 | ros_A(9) = 0.1253708332932087d+02 |
---|
2417 | ros_A(10) =-0.6878860361058950d+00 |
---|
2418 | ros_A(11) = ros_A(7) |
---|
2419 | ros_A(12) = ros_A(8) |
---|
2420 | ros_A(13) = ros_A(9) |
---|
2421 | ros_A(14) = ros_A(10) |
---|
2422 | ros_A(15) = 1.0d+00 |
---|
2423 | |
---|
2424 | ros_C(1) =-0.5668800000000000d+01 |
---|
2425 | ros_C(2) =-0.2430093356833875d+01 |
---|
2426 | ros_C(3) =-0.2063599157091915d+00 |
---|
2427 | ros_C(4) =-0.1073529058151375d+00 |
---|
2428 | ros_C(5) =-0.9594562251023355d+01 |
---|
2429 | ros_C(6) =-0.2047028614809616d+02 |
---|
2430 | ros_C(7) = 0.7496443313967647d+01 |
---|
2431 | ros_C(8) =-0.1024680431464352d+02 |
---|
2432 | ros_C(9) =-0.3399990352819905d+02 |
---|
2433 | ros_C(10) = 0.1170890893206160d+02 |
---|
2434 | ros_C(11) = 0.8083246795921522d+01 |
---|
2435 | ros_C(12) =-0.7981132988064893d+01 |
---|
2436 | ros_C(13) =-0.3152159432874371d+02 |
---|
2437 | ros_C(14) = 0.1631930543123136d+02 |
---|
2438 | ros_C(15) =-0.6058818238834054d+01 |
---|
2439 | |
---|
2440 | !~~~> M_i = Coefficients for new step solution |
---|
2441 | ros_M(1) = ros_A(7) |
---|
2442 | ros_M(2) = ros_A(8) |
---|
2443 | ros_M(3) = ros_A(9) |
---|
2444 | ros_M(4) = ros_A(10) |
---|
2445 | ros_M(5) = 1.0d+00 |
---|
2446 | ros_M(6) = 1.0d+00 |
---|
2447 | |
---|
2448 | !~~~> E_i = Coefficients for error estimator |
---|
2449 | ros_E(1) = 0.0d+00 |
---|
2450 | ros_E(2) = 0.0d+00 |
---|
2451 | ros_E(3) = 0.0d+00 |
---|
2452 | ros_E(4) = 0.0d+00 |
---|
2453 | ros_E(5) = 0.0d+00 |
---|
2454 | ros_E(6) = 1.0d+00 |
---|
2455 | |
---|
2456 | !~~~> Does the stage i require a new function evaluation (ros_NewF(i)=TRUE) |
---|
2457 | ! or does it re-use the function evaluation from stage i-1 (ros_NewF(i)=FALSE) |
---|
2458 | ros_NewF(1) = .TRUE. |
---|
2459 | ros_NewF(2) = .TRUE. |
---|
2460 | ros_NewF(3) = .TRUE. |
---|
2461 | ros_NewF(4) = .TRUE. |
---|
2462 | ros_NewF(5) = .TRUE. |
---|
2463 | ros_NewF(6) = .TRUE. |
---|
2464 | |
---|
2465 | !~~~> ros_ELO = estimator of local order - the minimum between the |
---|
2466 | ! main and the embedded scheme orders plus 1 |
---|
2467 | ros_ELO = 4.0d0 |
---|
2468 | |
---|
2469 | END SUBROUTINE Rodas4 |
---|
2470 | |
---|
2471 | |
---|
2472 | END SUBROUTINE RosenbrockADJ ! and its internal procedures |
---|
2473 | |
---|
2474 | |
---|
2475 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
2476 | SUBROUTINE FunTemplate( T, Y, Ydot ) |
---|
2477 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
2478 | ! Template for the ODE function call. |
---|
2479 | ! Updates the rate coefficients (and possibly the fixed species) at each call |
---|
2480 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
2481 | |
---|
2482 | !~~~> Input variables |
---|
2483 | KPP_REAL, INTENT(IN) :: T, Y(NVAR) |
---|
2484 | !~~~> Output variables |
---|
2485 | KPP_REAL, INTENT(OUT) :: Ydot(NVAR) |
---|
2486 | !~~~> Local variables |
---|
2487 | KPP_REAL :: Told |
---|
2488 | |
---|
2489 | Told = TIME |
---|
2490 | TIME = T |
---|
2491 | CALL Update_SUN() |
---|
2492 | CALL Update_RCONST() |
---|
2493 | CALL Fun( Y, FIX, RCONST, Ydot ) |
---|
2494 | TIME = Told |
---|
2495 | |
---|
2496 | END SUBROUTINE FunTemplate |
---|
2497 | |
---|
2498 | |
---|
2499 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
2500 | SUBROUTINE JacTemplate( T, Y, Jcb ) |
---|
2501 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
2502 | ! Template for the ODE Jacobian call. |
---|
2503 | ! Updates the rate coefficients (and possibly the fixed species) at each call |
---|
2504 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
2505 | |
---|
2506 | !~~~> Input variables |
---|
2507 | KPP_REAL :: T, Y(NVAR) |
---|
2508 | !~~~> Output variables |
---|
2509 | #ifdef FULL_ALGEBRA |
---|
2510 | KPP_REAL :: JV(LU_NONZERO), Jcb(NVAR,NVAR) |
---|
2511 | #else |
---|
2512 | KPP_REAL :: Jcb(LU_NONZERO) |
---|
2513 | #endif |
---|
2514 | !~~~> Local variables |
---|
2515 | KPP_REAL :: Told |
---|
2516 | #ifdef FULL_ALGEBRA |
---|
2517 | INTEGER :: i, j |
---|
2518 | #endif |
---|
2519 | |
---|
2520 | Told = TIME |
---|
2521 | TIME = T |
---|
2522 | CALL Update_SUN() |
---|
2523 | CALL Update_RCONST() |
---|
2524 | #ifdef FULL_ALGEBRA |
---|
2525 | CALL Jac_SP(Y, FIX, RCONST, JV) |
---|
2526 | DO j=1,NVAR |
---|
2527 | DO i=1,NVAR |
---|
2528 | Jcb(i,j) = 0.0_dp |
---|
2529 | END DO |
---|
2530 | END DO |
---|
2531 | DO i=1,LU_NONZERO |
---|
2532 | Jcb(LU_IROW(i),LU_ICOL(i)) = JV(i) |
---|
2533 | END DO |
---|
2534 | #else |
---|
2535 | CALL Jac_SP( Y, FIX, RCONST, Jcb ) |
---|
2536 | #endif |
---|
2537 | TIME = Told |
---|
2538 | |
---|
2539 | END SUBROUTINE JacTemplate |
---|
2540 | |
---|
2541 | |
---|
2542 | |
---|
2543 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
2544 | SUBROUTINE HessTemplate( T, Y, Hes ) |
---|
2545 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
2546 | ! Template for the ODE Hessian call. |
---|
2547 | ! Updates the rate coefficients (and possibly the fixed species) at each call |
---|
2548 | !~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
---|
2549 | |
---|
2550 | !~~~> Input variables |
---|
2551 | KPP_REAL, INTENT(IN) :: T, Y(NVAR) |
---|
2552 | !~~~> Output variables |
---|
2553 | KPP_REAL, INTENT(OUT) :: Hes(NHESS) |
---|
2554 | !~~~> Local variables |
---|
2555 | KPP_REAL :: Told |
---|
2556 | |
---|
2557 | Told = TIME |
---|
2558 | TIME = T |
---|
2559 | CALL Update_SUN() |
---|
2560 | CALL Update_RCONST() |
---|
2561 | CALL Hessian( Y, FIX, RCONST, Hes ) |
---|
2562 | TIME = Told |
---|
2563 | |
---|
2564 | END SUBROUTINE HessTemplate |
---|
2565 | |
---|
2566 | END MODULE KPP_ROOT_Integrator |
---|
2567 | |
---|
2568 | |
---|
2569 | |
---|
2570 | |
---|