xref: /petsc/src/ts/impls/implicit/theta/theta.c (revision 3fd522052c1b8f4bac4b06348665e2a89d1ff5dc)
1 /*
2   Code for timestepping with implicit Theta method
3 */
4 #include <petsc-private/tsimpl.h>                /*I   "petscts.h"   I*/
5 #include <petscsnes.h>
6 #include <petscdm.h>
7 #include <petscmat.h>
8 
9 typedef struct {
10   Vec          X,Xdot;                   /* Storage for one stage */
11   Vec          X0;                       /* work vector to store X0 */
12   Vec          affine;                   /* Affine vector needed for residual at beginning of step */
13   Vec          *VecDeltaLam;             /* Increment of the adjoint sensitivity w.r.t IC at stage*/
14   Vec          *VecDeltaMu;              /* Increment of the adjoint sensitivity w.r.t P at stage*/
15   Vec          *VecSensiTemp;            /* Vector to be timed with Jacobian transpose*/
16   PetscBool    extrapolate;
17   PetscBool    endpoint;
18   PetscReal    Theta;
19   PetscReal    stage_time;
20   TSStepStatus status;
21   char         *name;
22   PetscInt     order;
23   PetscReal    ccfl;               /* Placeholder for CFL coefficient relative to forward Euler */
24   PetscBool    adapt;  /* use time-step adaptivity ? */
25 } TS_Theta;
26 
27 #undef __FUNCT__
28 #define __FUNCT__ "TSThetaGetX0AndXdot"
29 static PetscErrorCode TSThetaGetX0AndXdot(TS ts,DM dm,Vec *X0,Vec *Xdot)
30 {
31   TS_Theta       *th = (TS_Theta*)ts->data;
32   PetscErrorCode ierr;
33 
34   PetscFunctionBegin;
35   if (X0) {
36     if (dm && dm != ts->dm) {
37       ierr = DMGetNamedGlobalVector(dm,"TSTheta_X0",X0);CHKERRQ(ierr);
38     } else *X0 = ts->vec_sol;
39   }
40   if (Xdot) {
41     if (dm && dm != ts->dm) {
42       ierr = DMGetNamedGlobalVector(dm,"TSTheta_Xdot",Xdot);CHKERRQ(ierr);
43     } else *Xdot = th->Xdot;
44   }
45   PetscFunctionReturn(0);
46 }
47 
48 
49 #undef __FUNCT__
50 #define __FUNCT__ "TSThetaRestoreX0AndXdot"
51 static PetscErrorCode TSThetaRestoreX0AndXdot(TS ts,DM dm,Vec *X0,Vec *Xdot)
52 {
53   PetscErrorCode ierr;
54 
55   PetscFunctionBegin;
56   if (X0) {
57     if (dm && dm != ts->dm) {
58       ierr = DMRestoreNamedGlobalVector(dm,"TSTheta_X0",X0);CHKERRQ(ierr);
59     }
60   }
61   if (Xdot) {
62     if (dm && dm != ts->dm) {
63       ierr = DMRestoreNamedGlobalVector(dm,"TSTheta_Xdot",Xdot);CHKERRQ(ierr);
64     }
65   }
66   PetscFunctionReturn(0);
67 }
68 
69 #undef __FUNCT__
70 #define __FUNCT__ "DMCoarsenHook_TSTheta"
71 static PetscErrorCode DMCoarsenHook_TSTheta(DM fine,DM coarse,void *ctx)
72 {
73 
74   PetscFunctionBegin;
75   PetscFunctionReturn(0);
76 }
77 
78 #undef __FUNCT__
79 #define __FUNCT__ "DMRestrictHook_TSTheta"
80 static PetscErrorCode DMRestrictHook_TSTheta(DM fine,Mat restrct,Vec rscale,Mat inject,DM coarse,void *ctx)
81 {
82   TS             ts = (TS)ctx;
83   PetscErrorCode ierr;
84   Vec            X0,Xdot,X0_c,Xdot_c;
85 
86   PetscFunctionBegin;
87   ierr = TSThetaGetX0AndXdot(ts,fine,&X0,&Xdot);CHKERRQ(ierr);
88   ierr = TSThetaGetX0AndXdot(ts,coarse,&X0_c,&Xdot_c);CHKERRQ(ierr);
89   ierr = MatRestrict(restrct,X0,X0_c);CHKERRQ(ierr);
90   ierr = MatRestrict(restrct,Xdot,Xdot_c);CHKERRQ(ierr);
91   ierr = VecPointwiseMult(X0_c,rscale,X0_c);CHKERRQ(ierr);
92   ierr = VecPointwiseMult(Xdot_c,rscale,Xdot_c);CHKERRQ(ierr);
93   ierr = TSThetaRestoreX0AndXdot(ts,fine,&X0,&Xdot);CHKERRQ(ierr);
94   ierr = TSThetaRestoreX0AndXdot(ts,coarse,&X0_c,&Xdot_c);CHKERRQ(ierr);
95   PetscFunctionReturn(0);
96 }
97 
98 #undef __FUNCT__
99 #define __FUNCT__ "DMSubDomainHook_TSTheta"
100 static PetscErrorCode DMSubDomainHook_TSTheta(DM dm,DM subdm,void *ctx)
101 {
102 
103   PetscFunctionBegin;
104   PetscFunctionReturn(0);
105 }
106 
107 #undef __FUNCT__
108 #define __FUNCT__ "DMSubDomainRestrictHook_TSTheta"
109 static PetscErrorCode DMSubDomainRestrictHook_TSTheta(DM dm,VecScatter gscat,VecScatter lscat,DM subdm,void *ctx)
110 {
111   TS             ts = (TS)ctx;
112   PetscErrorCode ierr;
113   Vec            X0,Xdot,X0_sub,Xdot_sub;
114 
115   PetscFunctionBegin;
116   ierr = TSThetaGetX0AndXdot(ts,dm,&X0,&Xdot);CHKERRQ(ierr);
117   ierr = TSThetaGetX0AndXdot(ts,subdm,&X0_sub,&Xdot_sub);CHKERRQ(ierr);
118 
119   ierr = VecScatterBegin(gscat,X0,X0_sub,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr);
120   ierr = VecScatterEnd(gscat,X0,X0_sub,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr);
121 
122   ierr = VecScatterBegin(gscat,Xdot,Xdot_sub,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr);
123   ierr = VecScatterEnd(gscat,Xdot,Xdot_sub,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr);
124 
125   ierr = TSThetaRestoreX0AndXdot(ts,dm,&X0,&Xdot);CHKERRQ(ierr);
126   ierr = TSThetaRestoreX0AndXdot(ts,subdm,&X0_sub,&Xdot_sub);CHKERRQ(ierr);
127   PetscFunctionReturn(0);
128 }
129 
130 #undef __FUNCT__
131 #define __FUNCT__ "TSEvaluateStep_Theta"
132 static PetscErrorCode TSEvaluateStep_Theta(TS ts,PetscInt order,Vec U,PetscBool *done)
133 {
134   PetscErrorCode ierr;
135   TS_Theta       *th = (TS_Theta*)ts->data;
136 
137   PetscFunctionBegin;
138   if (order == 0) SETERRQ(PetscObjectComm((PetscObject)ts),PETSC_ERR_USER,"No time-step adaptivity implemented for 1st order theta method; Run with -ts_adapt_type none");
139   if (order == th->order) {
140     if (th->endpoint) {
141       ierr = VecCopy(th->X,U);CHKERRQ(ierr);
142     } else {
143       PetscReal shift = 1./(th->Theta*ts->time_step);
144       ierr = VecAXPBYPCZ(th->Xdot,-shift,shift,0,U,th->X);CHKERRQ(ierr);
145       ierr = VecAXPY(U,ts->time_step,th->Xdot);CHKERRQ(ierr);
146     }
147   } else if (order == th->order-1 && order) {
148     ierr = VecWAXPY(U,ts->time_step,th->Xdot,th->X0);CHKERRQ(ierr);
149   }
150   PetscFunctionReturn(0);
151 }
152 
153 #undef __FUNCT__
154 #define __FUNCT__ "TSRollBack_Theta"
155 static PetscErrorCode TSRollBack_Theta(TS ts)
156 {
157   TS_Theta       *th = (TS_Theta*)ts->data;
158   PetscErrorCode ierr;
159 
160   PetscFunctionBegin;
161   ierr = VecCopy(th->X0,ts->vec_sol);CHKERRQ(ierr);
162   th->status    = TS_STEP_INCOMPLETE;
163   PetscFunctionReturn(0);
164 }
165 
166 #undef __FUNCT__
167 #define __FUNCT__ "TSStep_Theta"
168 static PetscErrorCode TSStep_Theta(TS ts)
169 {
170   TS_Theta       *th = (TS_Theta*)ts->data;
171   PetscInt       its,lits,reject,next_scheme;
172   PetscReal      next_time_step;
173   TSAdapt        adapt;
174   PetscBool      stageok,accept = PETSC_TRUE;
175   PetscErrorCode ierr;
176 
177   PetscFunctionBegin;
178   th->status = TS_STEP_INCOMPLETE;
179   ierr = VecCopy(ts->vec_sol,th->X0);CHKERRQ(ierr);
180   for (reject=0; !ts->reason && th->status != TS_STEP_COMPLETE; ts->reject++) {
181     PetscReal shift = 1./(th->Theta*ts->time_step);
182     th->stage_time = ts->ptime + (th->endpoint ? 1. : th->Theta)*ts->time_step;
183     ierr = TSPreStep(ts);CHKERRQ(ierr);
184     ierr = TSPreStage(ts,th->stage_time);CHKERRQ(ierr);
185 
186     if (th->endpoint) {           /* This formulation assumes linear time-independent mass matrix */
187       ierr = VecZeroEntries(th->Xdot);CHKERRQ(ierr);
188       if (!th->affine) {ierr = VecDuplicate(ts->vec_sol,&th->affine);CHKERRQ(ierr);}
189       ierr = TSComputeIFunction(ts,ts->ptime,ts->vec_sol,th->Xdot,th->affine,PETSC_FALSE);CHKERRQ(ierr);
190       ierr = VecScale(th->affine,(th->Theta-1.)/th->Theta);CHKERRQ(ierr);
191     }
192     if (th->extrapolate) {
193       ierr = VecWAXPY(th->X,1./shift,th->Xdot,ts->vec_sol);CHKERRQ(ierr);
194     } else {
195       ierr = VecCopy(ts->vec_sol,th->X);CHKERRQ(ierr);
196     }
197     ierr = SNESSolve(ts->snes,th->affine,th->X);CHKERRQ(ierr);
198     ierr = SNESGetIterationNumber(ts->snes,&its);CHKERRQ(ierr);
199     ierr = SNESGetLinearSolveIterations(ts->snes,&lits);CHKERRQ(ierr);
200     ts->snes_its += its; ts->ksp_its += lits;
201     ierr = TSPostStage(ts,th->stage_time,0,&(th->X));CHKERRQ(ierr);
202     ierr = TSGetAdapt(ts,&adapt);CHKERRQ(ierr);
203     ierr = TSAdaptCheckStage(adapt,ts,&stageok);CHKERRQ(ierr);
204     if (!stageok) {accept = PETSC_FALSE; goto reject_step;}
205 
206     ierr = TSEvaluateStep(ts,th->order,ts->vec_sol,NULL);CHKERRQ(ierr);
207     th->status = TS_STEP_PENDING;
208     /* Register only the current method as a candidate because we're not supporting multiple candidates yet. */
209     ierr = TSGetAdapt(ts,&adapt);CHKERRQ(ierr);
210     ierr = TSAdaptCandidatesClear(adapt);CHKERRQ(ierr);
211     ierr = TSAdaptCandidateAdd(adapt,NULL,th->order,1,th->ccfl,1.0,PETSC_TRUE);CHKERRQ(ierr);
212     ierr = TSAdaptChoose(adapt,ts,ts->time_step,&next_scheme,&next_time_step,&accept);CHKERRQ(ierr);
213     if (!accept) {           /* Roll back the current step */
214       ts->ptime += next_time_step; /* This will be undone in rollback */
215       th->status = TS_STEP_INCOMPLETE;
216       ierr = TSRollBack(ts);CHKERRQ(ierr);
217       goto reject_step;
218     }
219 
220     /* ignore next_scheme for now */
221     ts->ptime    += ts->time_step;
222     ts->time_step = next_time_step;
223     ts->steps++;
224     th->status = TS_STEP_COMPLETE;
225     break;
226 
227 reject_step:
228     if (!ts->reason && ++reject > ts->max_reject && ts->max_reject >= 0) {
229       ts->reason = TS_DIVERGED_STEP_REJECTED;
230       ierr = PetscInfo2(ts,"Step=%D, step rejections %D greater than current TS allowed, stopping solve\n",ts->steps,reject);CHKERRQ(ierr);
231     }
232     continue;
233   }
234   PetscFunctionReturn(0);
235 }
236 
237 #undef __FUNCT__
238 #define __FUNCT__ "TSStepAdj_Theta"
239 static PetscErrorCode TSStepAdj_Theta(TS ts)
240 {
241   TS_Theta            *th = (TS_Theta*)ts->data;
242   Vec                 *VecDeltaLam = th->VecDeltaLam,*VecDeltaMu = th->VecDeltaMu,*VecSensiTemp = th->VecSensiTemp;
243   PetscInt            nadj;
244   PetscErrorCode      ierr;
245   Mat                 J,Jp;
246   KSP                 ksp;
247   PetscReal           shift;
248 
249   PetscFunctionBegin;
250 
251   th->status = TS_STEP_INCOMPLETE;
252   ierr = SNESGetKSP(ts->snes,&ksp);
253   ierr = TSGetIJacobian(ts,&J,&Jp,NULL,NULL);CHKERRQ(ierr);
254   th->stage_time = ts->ptime + (th->endpoint ? ts->time_step : (1.-th->Theta)*ts->time_step); /* time_step is negative*/
255 
256   ierr = TSPreStep(ts);CHKERRQ(ierr);
257 
258   /* Build RHS */
259   for (nadj=0; nadj<ts->numberadjs; nadj++) {
260     ierr = VecCopy(ts->vecs_sensi[nadj],VecSensiTemp[nadj]);CHKERRQ(ierr);
261     ierr = VecScale(VecSensiTemp[nadj],-1./(th->Theta*ts->time_step));CHKERRQ(ierr);
262   }
263 
264   /* Build LHS */
265   shift = -1./(th->Theta*ts->time_step);
266   if (th->endpoint) {
267     ierr = TSComputeIJacobian(ts,ts->ptime,ts->vec_sol,th->Xdot,shift,J,Jp,PETSC_FALSE);CHKERRQ(ierr);
268   }else {
269     ierr = TSComputeIJacobian(ts,th->stage_time,th->X,th->Xdot,shift,J,Jp,PETSC_FALSE);CHKERRQ(ierr);
270   }
271   ierr = KSPSetOperators(ksp,J,Jp);CHKERRQ(ierr);
272 
273   /* Solve LHS X = RHS */
274   for (nadj=0; nadj<ts->numberadjs; nadj++) {
275     ierr = KSPSolveTranspose(ksp,VecSensiTemp[nadj],VecDeltaLam[nadj]);CHKERRQ(ierr);
276   }
277 
278   /* Update sensitivities */
279   if(th->endpoint && th->Theta!=1.) { /* two-stage case */
280     shift = -1./((th->Theta-1.)*ts->time_step);
281     ierr  = TSComputeIJacobian(ts,th->stage_time,th->X,th->Xdot,shift,J,Jp,PETSC_FALSE);CHKERRQ(ierr);
282     for (nadj=0; nadj<ts->numberadjs; nadj++) {
283       ierr = MatMultTranspose(J,VecDeltaLam[nadj],ts->vecs_sensi[nadj]);CHKERRQ(ierr);
284       ierr = VecScale(ts->vecs_sensi[nadj],1./shift);CHKERRQ(ierr);
285     }
286 
287     if (ts->vecs_sensip) { /* sensitivities wrt parameters */
288       ierr = TSRHSJacobianP(ts,ts->ptime,ts->vec_sol,ts->Jacp);CHKERRQ(ierr);
289       for (nadj=0; nadj<ts->numberadjs; nadj++) {
290         ierr = MatMultTranspose(ts->Jacp,VecDeltaLam[nadj],VecDeltaMu[nadj]);CHKERRQ(ierr);
291         ierr = VecAXPY(ts->vecs_sensip[nadj],-ts->time_step*th->Theta,VecDeltaMu[nadj]);CHKERRQ(ierr);
292       }
293       ierr = TSRHSJacobianP(ts,th->stage_time,th->X,ts->Jacp);CHKERRQ(ierr);
294       for (nadj=0; nadj<ts->numberadjs; nadj++) {
295         ierr = MatMultTranspose(ts->Jacp,VecDeltaLam[nadj],VecDeltaMu[nadj]);CHKERRQ(ierr);
296         ierr = VecAXPY(ts->vecs_sensip[nadj],-ts->time_step*(1.-th->Theta),VecDeltaMu[nadj]);CHKERRQ(ierr);
297       }
298     }
299   }else { /* one-stage case */
300     shift = 0.0;
301     ierr  = TSComputeIJacobian(ts,th->stage_time,th->X,th->Xdot,shift,J,Jp,PETSC_FALSE);CHKERRQ(ierr);
302      /* When th->endpoint is true and th->Theta==1 (beuler method), the Jacobian is supposed to be evaluated at ts->ptime like this:
303     if(th->endpoint) {
304       ierr  = TSComputeIJacobian(ts,ts->ptime,ts->vec_sol,th->Xdot,shift,J,Jp,PETSC_FALSE);CHKERRQ(ierr);
305     }
306     but ts->ptime and ts->vec_sol have the same values as th->stage_time and th->X in this case. So the code is simplified here.
307     */
308     for (nadj=0; nadj<ts->numberadjs; nadj++) {
309       ierr = MatMultTranspose(J,VecDeltaLam[nadj],VecSensiTemp[nadj]);CHKERRQ(ierr);
310       ierr = VecAXPY(ts->vecs_sensi[nadj],ts->time_step,VecSensiTemp[nadj]);CHKERRQ(ierr);
311     }
312     if (ts->vecs_sensip) {
313       ierr = TSRHSJacobianP(ts,th->stage_time,th->X,ts->Jacp);CHKERRQ(ierr);
314       for (nadj=0; nadj<ts->numberadjs; nadj++) {
315         ierr = MatMultTranspose(ts->Jacp,VecDeltaLam[nadj],VecDeltaMu[nadj]);CHKERRQ(ierr);
316         ierr = VecAXPY(ts->vecs_sensip[nadj],-ts->time_step,VecDeltaMu[nadj]);CHKERRQ(ierr);
317       }
318     }
319   }
320 
321   ts->ptime += ts->time_step;
322   ts->steps++;
323   th->status = TS_STEP_COMPLETE;
324   PetscFunctionReturn(0);
325 }
326 
327 #undef __FUNCT__
328 #define __FUNCT__ "TSInterpolate_Theta"
329 static PetscErrorCode TSInterpolate_Theta(TS ts,PetscReal t,Vec X)
330 {
331   TS_Theta       *th   = (TS_Theta*)ts->data;
332   PetscReal      alpha = t - ts->ptime;
333   PetscErrorCode ierr;
334 
335   PetscFunctionBegin;
336   ierr = VecCopy(ts->vec_sol,th->X);CHKERRQ(ierr);
337   if (th->endpoint) alpha *= th->Theta;
338   ierr = VecWAXPY(X,alpha,th->Xdot,th->X);CHKERRQ(ierr);
339   PetscFunctionReturn(0);
340 }
341 
342 /*------------------------------------------------------------*/
343 #undef __FUNCT__
344 #define __FUNCT__ "TSReset_Theta"
345 static PetscErrorCode TSReset_Theta(TS ts)
346 {
347   TS_Theta       *th = (TS_Theta*)ts->data;
348   PetscErrorCode ierr;
349 
350   PetscFunctionBegin;
351   ierr = VecDestroy(&th->X);CHKERRQ(ierr);
352   ierr = VecDestroy(&th->Xdot);CHKERRQ(ierr);
353   ierr = VecDestroy(&th->X0);CHKERRQ(ierr);
354   ierr = VecDestroy(&th->affine);CHKERRQ(ierr);
355   if(ts->reverse_mode) {
356     ierr = VecDestroyVecs(ts->numberadjs,&th->VecDeltaLam);CHKERRQ(ierr);
357     if(th->VecDeltaMu) {
358       ierr = VecDestroyVecs(ts->numberadjs,&th->VecDeltaMu);CHKERRQ(ierr);
359     }
360     ierr = VecDestroyVecs(ts->numberadjs,&th->VecSensiTemp);CHKERRQ(ierr);
361   }
362   PetscFunctionReturn(0);
363 }
364 
365 #undef __FUNCT__
366 #define __FUNCT__ "TSDestroy_Theta"
367 static PetscErrorCode TSDestroy_Theta(TS ts)
368 {
369   PetscErrorCode ierr;
370 
371   PetscFunctionBegin;
372   ierr = TSReset_Theta(ts);CHKERRQ(ierr);
373   ierr = PetscFree(ts->data);CHKERRQ(ierr);
374   ierr = PetscObjectComposeFunction((PetscObject)ts,"TSThetaGetTheta_C",NULL);CHKERRQ(ierr);
375   ierr = PetscObjectComposeFunction((PetscObject)ts,"TSThetaSetTheta_C",NULL);CHKERRQ(ierr);
376   ierr = PetscObjectComposeFunction((PetscObject)ts,"TSThetaGetEndpoint_C",NULL);CHKERRQ(ierr);
377   ierr = PetscObjectComposeFunction((PetscObject)ts,"TSThetaSetEndpoint_C",NULL);CHKERRQ(ierr);
378   PetscFunctionReturn(0);
379 }
380 
381 /*
382   This defines the nonlinear equation that is to be solved with SNES
383   G(U) = F[t0+Theta*dt, U, (U-U0)*shift] = 0
384 */
385 #undef __FUNCT__
386 #define __FUNCT__ "SNESTSFormFunction_Theta"
387 static PetscErrorCode SNESTSFormFunction_Theta(SNES snes,Vec x,Vec y,TS ts)
388 {
389   TS_Theta       *th = (TS_Theta*)ts->data;
390   PetscErrorCode ierr;
391   Vec            X0,Xdot;
392   DM             dm,dmsave;
393   PetscReal      shift = 1./(th->Theta*ts->time_step);
394 
395   PetscFunctionBegin;
396   ierr = SNESGetDM(snes,&dm);CHKERRQ(ierr);
397   /* When using the endpoint variant, this is actually 1/Theta * Xdot */
398   ierr = TSThetaGetX0AndXdot(ts,dm,&X0,&Xdot);CHKERRQ(ierr);
399   ierr = VecAXPBYPCZ(Xdot,-shift,shift,0,X0,x);CHKERRQ(ierr);
400 
401   /* DM monkey-business allows user code to call TSGetDM() inside of functions evaluated on levels of FAS */
402   dmsave = ts->dm;
403   ts->dm = dm;
404   ierr   = TSComputeIFunction(ts,th->stage_time,x,Xdot,y,PETSC_FALSE);CHKERRQ(ierr);
405   ts->dm = dmsave;
406   ierr   = TSThetaRestoreX0AndXdot(ts,dm,&X0,&Xdot);CHKERRQ(ierr);
407   PetscFunctionReturn(0);
408 }
409 
410 #undef __FUNCT__
411 #define __FUNCT__ "SNESTSFormJacobian_Theta"
412 static PetscErrorCode SNESTSFormJacobian_Theta(SNES snes,Vec x,Mat A,Mat B,TS ts)
413 {
414   TS_Theta       *th = (TS_Theta*)ts->data;
415   PetscErrorCode ierr;
416   Vec            Xdot;
417   DM             dm,dmsave;
418   PetscReal      shift = 1./(th->Theta*ts->time_step);
419 
420   PetscFunctionBegin;
421   ierr = SNESGetDM(snes,&dm);CHKERRQ(ierr);
422 
423   /* th->Xdot has already been computed in SNESTSFormFunction_Theta (SNES guarantees this) */
424   ierr = TSThetaGetX0AndXdot(ts,dm,NULL,&Xdot);CHKERRQ(ierr);
425 
426   dmsave = ts->dm;
427   ts->dm = dm;
428   ierr   = TSComputeIJacobian(ts,th->stage_time,x,Xdot,shift,A,B,PETSC_FALSE);CHKERRQ(ierr);
429   ts->dm = dmsave;
430   ierr   = TSThetaRestoreX0AndXdot(ts,dm,NULL,&Xdot);CHKERRQ(ierr);
431   PetscFunctionReturn(0);
432 }
433 
434 #undef __FUNCT__
435 #define __FUNCT__ "TSSetUp_Theta"
436 static PetscErrorCode TSSetUp_Theta(TS ts)
437 {
438   TS_Theta       *th = (TS_Theta*)ts->data;
439   PetscErrorCode ierr;
440   SNES           snes;
441   TSAdapt        adapt;
442   DM             dm;
443 
444   PetscFunctionBegin;
445   ierr = VecDuplicate(ts->vec_sol,&th->X);CHKERRQ(ierr);
446   ierr = VecDuplicate(ts->vec_sol,&th->Xdot);CHKERRQ(ierr);
447   ierr = VecDuplicate(ts->vec_sol,&th->X0);CHKERRQ(ierr);
448   ierr = TSGetSNES(ts,&snes);CHKERRQ(ierr);
449   ierr = TSGetDM(ts,&dm);CHKERRQ(ierr);
450   if (dm) {
451     ierr = DMCoarsenHookAdd(dm,DMCoarsenHook_TSTheta,DMRestrictHook_TSTheta,ts);CHKERRQ(ierr);
452     ierr = DMSubDomainHookAdd(dm,DMSubDomainHook_TSTheta,DMSubDomainRestrictHook_TSTheta,ts);CHKERRQ(ierr);
453   }
454   if (th->Theta == 0.5 && th->endpoint) th->order = 2;
455   else th->order = 1;
456 
457   ierr = TSGetAdapt(ts,&adapt);CHKERRQ(ierr);
458   if (!th->adapt) {
459     ierr = TSAdaptSetType(adapt,TSADAPTNONE);CHKERRQ(ierr);
460   }
461   if (ts->reverse_mode) {
462     ierr = VecDuplicateVecs(ts->vecs_sensi[0],ts->numberadjs,&th->VecDeltaLam);CHKERRQ(ierr);
463     if(ts->vecs_sensip) {
464       ierr = VecDuplicateVecs(ts->vecs_sensip[0],ts->numberadjs,&th->VecDeltaMu);CHKERRQ(ierr);
465     }
466     ierr = VecDuplicateVecs(ts->vecs_sensi[0],ts->numberadjs,&th->VecSensiTemp);CHKERRQ(ierr);
467   }
468   PetscFunctionReturn(0);
469 }
470 /*------------------------------------------------------------*/
471 
472 #undef __FUNCT__
473 #define __FUNCT__ "TSSetFromOptions_Theta"
474 static PetscErrorCode TSSetFromOptions_Theta(PetscOptions *PetscOptionsObject,TS ts)
475 {
476   TS_Theta       *th = (TS_Theta*)ts->data;
477   PetscErrorCode ierr;
478 
479   PetscFunctionBegin;
480   ierr = PetscOptionsHead(PetscOptionsObject,"Theta ODE solver options");CHKERRQ(ierr);
481   {
482     ierr = PetscOptionsReal("-ts_theta_theta","Location of stage (0<Theta<=1)","TSThetaSetTheta",th->Theta,&th->Theta,NULL);CHKERRQ(ierr);
483     ierr = PetscOptionsBool("-ts_theta_extrapolate","Extrapolate stage solution from previous solution (sometimes unstable)","TSThetaSetExtrapolate",th->extrapolate,&th->extrapolate,NULL);CHKERRQ(ierr);
484     ierr = PetscOptionsBool("-ts_theta_endpoint","Use the endpoint instead of midpoint form of the Theta method","TSThetaSetEndpoint",th->endpoint,&th->endpoint,NULL);CHKERRQ(ierr);
485     ierr = PetscOptionsBool("-ts_theta_adapt","Use time-step adaptivity with the Theta method","",th->adapt,&th->adapt,NULL);CHKERRQ(ierr);
486     ierr = SNESSetFromOptions(ts->snes);CHKERRQ(ierr);
487   }
488   ierr = PetscOptionsTail();CHKERRQ(ierr);
489   PetscFunctionReturn(0);
490 }
491 
492 #undef __FUNCT__
493 #define __FUNCT__ "TSView_Theta"
494 static PetscErrorCode TSView_Theta(TS ts,PetscViewer viewer)
495 {
496   TS_Theta       *th = (TS_Theta*)ts->data;
497   PetscBool      iascii;
498   PetscErrorCode ierr;
499 
500   PetscFunctionBegin;
501   ierr = PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERASCII,&iascii);CHKERRQ(ierr);
502   if (iascii) {
503     ierr = PetscViewerASCIIPrintf(viewer,"  Theta=%g\n",(double)th->Theta);CHKERRQ(ierr);
504     ierr = PetscViewerASCIIPrintf(viewer,"  Extrapolation=%s\n",th->extrapolate ? "yes" : "no");CHKERRQ(ierr);
505   }
506   if (ts->snes) {ierr = SNESView(ts->snes,viewer);CHKERRQ(ierr);}
507   PetscFunctionReturn(0);
508 }
509 
510 #undef __FUNCT__
511 #define __FUNCT__ "TSThetaGetTheta_Theta"
512 PetscErrorCode  TSThetaGetTheta_Theta(TS ts,PetscReal *theta)
513 {
514   TS_Theta *th = (TS_Theta*)ts->data;
515 
516   PetscFunctionBegin;
517   *theta = th->Theta;
518   PetscFunctionReturn(0);
519 }
520 
521 #undef __FUNCT__
522 #define __FUNCT__ "TSThetaSetTheta_Theta"
523 PetscErrorCode  TSThetaSetTheta_Theta(TS ts,PetscReal theta)
524 {
525   TS_Theta *th = (TS_Theta*)ts->data;
526 
527   PetscFunctionBegin;
528   if (theta <= 0 || 1 < theta) SETERRQ1(PetscObjectComm((PetscObject)ts),PETSC_ERR_ARG_OUTOFRANGE,"Theta %g not in range (0,1]",(double)theta);
529   th->Theta = theta;
530   PetscFunctionReturn(0);
531 }
532 
533 #undef __FUNCT__
534 #define __FUNCT__ "TSThetaGetEndpoint_Theta"
535 PetscErrorCode  TSThetaGetEndpoint_Theta(TS ts,PetscBool *endpoint)
536 {
537   TS_Theta *th = (TS_Theta*)ts->data;
538 
539   PetscFunctionBegin;
540   *endpoint = th->endpoint;
541   PetscFunctionReturn(0);
542 }
543 
544 #undef __FUNCT__
545 #define __FUNCT__ "TSThetaSetEndpoint_Theta"
546 PetscErrorCode  TSThetaSetEndpoint_Theta(TS ts,PetscBool flg)
547 {
548   TS_Theta *th = (TS_Theta*)ts->data;
549 
550   PetscFunctionBegin;
551   th->endpoint = flg;
552   PetscFunctionReturn(0);
553 }
554 
555 #if defined(PETSC_HAVE_COMPLEX)
556 #undef __FUNCT__
557 #define __FUNCT__ "TSComputeLinearStability_Theta"
558 static PetscErrorCode TSComputeLinearStability_Theta(TS ts,PetscReal xr,PetscReal xi,PetscReal *yr,PetscReal *yi)
559 {
560   PetscComplex z   = xr + xi*PETSC_i,f;
561   TS_Theta     *th = (TS_Theta*)ts->data;
562   const PetscReal one = 1.0;
563 
564   PetscFunctionBegin;
565   f   = (one + (one - th->Theta)*z)/(one - th->Theta*z);
566   *yr = PetscRealPartComplex(f);
567   *yi = PetscImaginaryPartComplex(f);
568   PetscFunctionReturn(0);
569 }
570 #endif
571 
572 #undef __FUNCT__
573 #define __FUNCT__ "TSGetStages_Theta"
574 static PetscErrorCode  TSGetStages_Theta(TS ts,PetscInt *ns,Vec **Y)
575 {
576   TS_Theta     *th = (TS_Theta*)ts->data;
577 
578   PetscFunctionBegin;
579   *ns = 1;
580   if(Y) {
581     *Y  = &(th->X);
582   }
583   PetscFunctionReturn(0);
584 }
585 
586 /* ------------------------------------------------------------ */
587 /*MC
588       TSTHETA - DAE solver using the implicit Theta method
589 
590    Level: beginner
591 
592    Options Database:
593       -ts_theta_theta <Theta> - Location of stage (0<Theta<=1)
594       -ts_theta_extrapolate <flg> Extrapolate stage solution from previous solution (sometimes unstable)
595       -ts_theta_endpoint <flag> - Use the endpoint (like Crank-Nicholson) instead of midpoint form of the Theta method
596 
597    Notes:
598 $  -ts_type theta -ts_theta_theta 1.0 corresponds to backward Euler (TSBEULER)
599 $  -ts_type theta -ts_theta_theta 0.5 corresponds to the implicit midpoint rule
600 $  -ts_type theta -ts_theta_theta 0.5 -ts_theta_endpoint corresponds to Crank-Nicholson (TSCN)
601 
602 
603 
604    This method can be applied to DAE.
605 
606    This method is cast as a 1-stage implicit Runge-Kutta method.
607 
608 .vb
609   Theta | Theta
610   -------------
611         |  1
612 .ve
613 
614    For the default Theta=0.5, this is also known as the implicit midpoint rule.
615 
616    When the endpoint variant is chosen, the method becomes a 2-stage method with first stage explicit:
617 
618 .vb
619   0 | 0         0
620   1 | 1-Theta   Theta
621   -------------------
622     | 1-Theta   Theta
623 .ve
624 
625    For the default Theta=0.5, this is the trapezoid rule (also known as Crank-Nicolson, see TSCN).
626 
627    To apply a diagonally implicit RK method to DAE, the stage formula
628 
629 $  Y_i = X + h sum_j a_ij Y'_j
630 
631    is interpreted as a formula for Y'_i in terms of Y_i and known values (Y'_j, j<i)
632 
633 .seealso:  TSCreate(), TS, TSSetType(), TSCN, TSBEULER, TSThetaSetTheta(), TSThetaSetEndpoint()
634 
635 M*/
636 #undef __FUNCT__
637 #define __FUNCT__ "TSCreate_Theta"
638 PETSC_EXTERN PetscErrorCode TSCreate_Theta(TS ts)
639 {
640   TS_Theta       *th;
641   PetscErrorCode ierr;
642 
643   PetscFunctionBegin;
644   ts->ops->reset          = TSReset_Theta;
645   ts->ops->destroy        = TSDestroy_Theta;
646   ts->ops->view           = TSView_Theta;
647   ts->ops->setup          = TSSetUp_Theta;
648   ts->ops->step           = TSStep_Theta;
649   ts->ops->interpolate    = TSInterpolate_Theta;
650   ts->ops->evaluatestep   = TSEvaluateStep_Theta;
651   ts->ops->rollback       = TSRollBack_Theta;
652   ts->ops->setfromoptions = TSSetFromOptions_Theta;
653   ts->ops->snesfunction   = SNESTSFormFunction_Theta;
654   ts->ops->snesjacobian   = SNESTSFormJacobian_Theta;
655 #if defined(PETSC_HAVE_COMPLEX)
656   ts->ops->linearstability = TSComputeLinearStability_Theta;
657 #endif
658   ts->ops->getstages      = TSGetStages_Theta;
659   ts->ops->stepadj        = TSStepAdj_Theta;
660 
661   ierr = PetscNewLog(ts,&th);CHKERRQ(ierr);
662   ts->data = (void*)th;
663 
664   th->extrapolate = PETSC_FALSE;
665   th->Theta       = 0.5;
666   th->ccfl        = 1.0;
667   th->adapt       = PETSC_FALSE;
668   ierr = PetscObjectComposeFunction((PetscObject)ts,"TSThetaGetTheta_C",TSThetaGetTheta_Theta);CHKERRQ(ierr);
669   ierr = PetscObjectComposeFunction((PetscObject)ts,"TSThetaSetTheta_C",TSThetaSetTheta_Theta);CHKERRQ(ierr);
670   ierr = PetscObjectComposeFunction((PetscObject)ts,"TSThetaGetEndpoint_C",TSThetaGetEndpoint_Theta);CHKERRQ(ierr);
671   ierr = PetscObjectComposeFunction((PetscObject)ts,"TSThetaSetEndpoint_C",TSThetaSetEndpoint_Theta);CHKERRQ(ierr);
672   PetscFunctionReturn(0);
673 }
674 
675 #undef __FUNCT__
676 #define __FUNCT__ "TSThetaGetTheta"
677 /*@
678   TSThetaGetTheta - Get the abscissa of the stage in (0,1].
679 
680   Not Collective
681 
682   Input Parameter:
683 .  ts - timestepping context
684 
685   Output Parameter:
686 .  theta - stage abscissa
687 
688   Note:
689   Use of this function is normally only required to hack TSTHETA to use a modified integration scheme.
690 
691   Level: Advanced
692 
693 .seealso: TSThetaSetTheta()
694 @*/
695 PetscErrorCode  TSThetaGetTheta(TS ts,PetscReal *theta)
696 {
697   PetscErrorCode ierr;
698 
699   PetscFunctionBegin;
700   PetscValidHeaderSpecific(ts,TS_CLASSID,1);
701   PetscValidPointer(theta,2);
702   ierr = PetscUseMethod(ts,"TSThetaGetTheta_C",(TS,PetscReal*),(ts,theta));CHKERRQ(ierr);
703   PetscFunctionReturn(0);
704 }
705 
706 #undef __FUNCT__
707 #define __FUNCT__ "TSThetaSetTheta"
708 /*@
709   TSThetaSetTheta - Set the abscissa of the stage in (0,1].
710 
711   Not Collective
712 
713   Input Parameter:
714 +  ts - timestepping context
715 -  theta - stage abscissa
716 
717   Options Database:
718 .  -ts_theta_theta <theta>
719 
720   Level: Intermediate
721 
722 .seealso: TSThetaGetTheta()
723 @*/
724 PetscErrorCode  TSThetaSetTheta(TS ts,PetscReal theta)
725 {
726   PetscErrorCode ierr;
727 
728   PetscFunctionBegin;
729   PetscValidHeaderSpecific(ts,TS_CLASSID,1);
730   ierr = PetscTryMethod(ts,"TSThetaSetTheta_C",(TS,PetscReal),(ts,theta));CHKERRQ(ierr);
731   PetscFunctionReturn(0);
732 }
733 
734 #undef __FUNCT__
735 #define __FUNCT__ "TSThetaGetEndpoint"
736 /*@
737   TSThetaGetEndpoint - Gets whether to use the endpoint variant of the method (e.g. trapezoid/Crank-Nicolson instead of midpoint rule).
738 
739   Not Collective
740 
741   Input Parameter:
742 .  ts - timestepping context
743 
744   Output Parameter:
745 .  endpoint - PETSC_TRUE when using the endpoint variant
746 
747   Level: Advanced
748 
749 .seealso: TSThetaSetEndpoint(), TSTHETA, TSCN
750 @*/
751 PetscErrorCode TSThetaGetEndpoint(TS ts,PetscBool *endpoint)
752 {
753   PetscErrorCode ierr;
754 
755   PetscFunctionBegin;
756   PetscValidHeaderSpecific(ts,TS_CLASSID,1);
757   PetscValidPointer(endpoint,2);
758   ierr = PetscTryMethod(ts,"TSThetaGetEndpoint_C",(TS,PetscBool*),(ts,endpoint));CHKERRQ(ierr);
759   PetscFunctionReturn(0);
760 }
761 
762 #undef __FUNCT__
763 #define __FUNCT__ "TSThetaSetEndpoint"
764 /*@
765   TSThetaSetEndpoint - Sets whether to use the endpoint variant of the method (e.g. trapezoid/Crank-Nicolson instead of midpoint rule).
766 
767   Not Collective
768 
769   Input Parameter:
770 +  ts - timestepping context
771 -  flg - PETSC_TRUE to use the endpoint variant
772 
773   Options Database:
774 .  -ts_theta_endpoint <flg>
775 
776   Level: Intermediate
777 
778 .seealso: TSTHETA, TSCN
779 @*/
780 PetscErrorCode TSThetaSetEndpoint(TS ts,PetscBool flg)
781 {
782   PetscErrorCode ierr;
783 
784   PetscFunctionBegin;
785   PetscValidHeaderSpecific(ts,TS_CLASSID,1);
786   ierr = PetscTryMethod(ts,"TSThetaSetEndpoint_C",(TS,PetscBool),(ts,flg));CHKERRQ(ierr);
787   PetscFunctionReturn(0);
788 }
789 
790 /*
791  * TSBEULER and TSCN are straightforward specializations of TSTHETA.
792  * The creation functions for these specializations are below.
793  */
794 
795 #undef __FUNCT__
796 #define __FUNCT__ "TSView_BEuler"
797 static PetscErrorCode TSView_BEuler(TS ts,PetscViewer viewer)
798 {
799   PetscErrorCode ierr;
800 
801   PetscFunctionBegin;
802   ierr = SNESView(ts->snes,viewer);CHKERRQ(ierr);
803   PetscFunctionReturn(0);
804 }
805 
806 /*MC
807       TSBEULER - ODE solver using the implicit backward Euler method
808 
809   Level: beginner
810 
811   Notes:
812   TSBEULER is equivalent to TSTHETA with Theta=1.0
813 
814 $  -ts_type theta -ts_theta_theta 1.
815 
816 .seealso:  TSCreate(), TS, TSSetType(), TSEULER, TSCN, TSTHETA
817 
818 M*/
819 #undef __FUNCT__
820 #define __FUNCT__ "TSCreate_BEuler"
821 PETSC_EXTERN PetscErrorCode TSCreate_BEuler(TS ts)
822 {
823   PetscErrorCode ierr;
824 
825   PetscFunctionBegin;
826   ierr = TSCreate_Theta(ts);CHKERRQ(ierr);
827   ierr = TSThetaSetTheta(ts,1.0);CHKERRQ(ierr);
828   ts->ops->view = TSView_BEuler;
829   PetscFunctionReturn(0);
830 }
831 
832 #undef __FUNCT__
833 #define __FUNCT__ "TSView_CN"
834 static PetscErrorCode TSView_CN(TS ts,PetscViewer viewer)
835 {
836   PetscErrorCode ierr;
837 
838   PetscFunctionBegin;
839   ierr = SNESView(ts->snes,viewer);CHKERRQ(ierr);
840   PetscFunctionReturn(0);
841 }
842 
843 /*MC
844       TSCN - ODE solver using the implicit Crank-Nicolson method.
845 
846   Level: beginner
847 
848   Notes:
849   TSCN is equivalent to TSTHETA with Theta=0.5 and the "endpoint" option set. I.e.
850 
851 $  -ts_type theta -ts_theta_theta 0.5 -ts_theta_endpoint
852 
853 .seealso:  TSCreate(), TS, TSSetType(), TSBEULER, TSTHETA
854 
855 M*/
856 #undef __FUNCT__
857 #define __FUNCT__ "TSCreate_CN"
858 PETSC_EXTERN PetscErrorCode TSCreate_CN(TS ts)
859 {
860   PetscErrorCode ierr;
861 
862   PetscFunctionBegin;
863   ierr = TSCreate_Theta(ts);CHKERRQ(ierr);
864   ierr = TSThetaSetTheta(ts,0.5);CHKERRQ(ierr);
865   ierr = TSThetaSetEndpoint(ts,PETSC_TRUE);CHKERRQ(ierr);
866   ts->ops->view = TSView_CN;
867   PetscFunctionReturn(0);
868 }
869