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 if (ts->vec_costintegral) { 221 /* Evolve ts->vec_costintegral to compute integrals */ 222 if (th->endpoint) { 223 ierr = TSAdjointComputeCostIntegrand(ts,ts->ptime,th->X0,ts->vec_costintegrand);CHKERRQ(ierr); 224 ierr = VecAXPY(ts->vec_costintegral,ts->time_step*(1.-th->Theta),ts->vec_costintegrand);CHKERRQ(ierr); 225 } 226 ierr = TSAdjointComputeCostIntegrand(ts,th->stage_time,th->X,ts->vec_costintegrand);CHKERRQ(ierr); 227 if (th->endpoint) { 228 ierr = VecAXPY(ts->vec_costintegral,ts->time_step*th->Theta,ts->vec_costintegrand);CHKERRQ(ierr); 229 }else { 230 ierr = VecAXPY(ts->vec_costintegral,ts->time_step,ts->vec_costintegrand);CHKERRQ(ierr); 231 } 232 } 233 234 /* ignore next_scheme for now */ 235 ts->ptime += ts->time_step; 236 ts->time_step = next_time_step; 237 ts->steps++; 238 th->status = TS_STEP_COMPLETE; 239 break; 240 241 reject_step: 242 if (!ts->reason && ++reject > ts->max_reject && ts->max_reject >= 0) { 243 ts->reason = TS_DIVERGED_STEP_REJECTED; 244 ierr = PetscInfo2(ts,"Step=%D, step rejections %D greater than current TS allowed, stopping solve\n",ts->steps,reject);CHKERRQ(ierr); 245 } 246 continue; 247 } 248 PetscFunctionReturn(0); 249 } 250 251 #undef __FUNCT__ 252 #define __FUNCT__ "TSAdjointStep_Theta" 253 static PetscErrorCode TSAdjointStep_Theta(TS ts) 254 { 255 TS_Theta *th = (TS_Theta*)ts->data; 256 Vec *VecDeltaLam = th->VecDeltaLam,*VecDeltaMu = th->VecDeltaMu,*VecSensiTemp = th->VecSensiTemp; 257 PetscInt nadj; 258 PetscErrorCode ierr; 259 Mat J,Jp; 260 KSP ksp; 261 PetscReal shift; 262 263 PetscFunctionBegin; 264 265 th->status = TS_STEP_INCOMPLETE; 266 ierr = SNESGetKSP(ts->snes,&ksp);CHKERRQ(ierr); 267 ierr = TSGetIJacobian(ts,&J,&Jp,NULL,NULL);CHKERRQ(ierr); 268 th->stage_time = ts->ptime + (th->endpoint ? ts->time_step : (1.-th->Theta)*ts->time_step); /* time_step is negative*/ 269 270 ierr = TSPreStep(ts);CHKERRQ(ierr); 271 272 /* Build RHS */ 273 if (ts->vec_costintegral) { /* Cost function has an integral term */ 274 if (th->endpoint) { 275 ierr = TSAdjointComputeDRDYFunction(ts,ts->ptime,ts->vec_sol,ts->vecs_drdy);CHKERRQ(ierr); 276 }else { 277 ierr = TSAdjointComputeDRDYFunction(ts,th->stage_time,th->X,ts->vecs_drdy);CHKERRQ(ierr); 278 } 279 } 280 for (nadj=0; nadj<ts->numcost; nadj++) { 281 ierr = VecCopy(ts->vecs_sensi[nadj],VecSensiTemp[nadj]);CHKERRQ(ierr); 282 ierr = VecScale(VecSensiTemp[nadj],-1./(th->Theta*ts->time_step));CHKERRQ(ierr); 283 if (ts->vec_costintegral) { 284 ierr = VecAXPY(VecSensiTemp[nadj],1.,ts->vecs_drdy[nadj]);CHKERRQ(ierr); 285 } 286 } 287 288 /* Build LHS */ 289 shift = -1./(th->Theta*ts->time_step); 290 if (th->endpoint) { 291 ierr = TSComputeIJacobian(ts,ts->ptime,ts->vec_sol,th->Xdot,shift,J,Jp,PETSC_FALSE);CHKERRQ(ierr); 292 }else { 293 ierr = TSComputeIJacobian(ts,th->stage_time,th->X,th->Xdot,shift,J,Jp,PETSC_FALSE);CHKERRQ(ierr); 294 } 295 ierr = KSPSetOperators(ksp,J,Jp);CHKERRQ(ierr); 296 297 /* Solve LHS X = RHS */ 298 for (nadj=0; nadj<ts->numcost; nadj++) { 299 ierr = KSPSolveTranspose(ksp,VecSensiTemp[nadj],VecDeltaLam[nadj]);CHKERRQ(ierr); 300 } 301 302 /* Update sensitivities, and evaluate integrals if there is any */ 303 if(th->endpoint && th->Theta!=1.) { /* two-stage case */ 304 shift = -1./((th->Theta-1.)*ts->time_step); 305 ierr = TSComputeIJacobian(ts,th->stage_time,th->X,th->Xdot,shift,J,Jp,PETSC_FALSE);CHKERRQ(ierr); 306 if (ts->vec_costintegral) { 307 ierr = TSAdjointComputeDRDYFunction(ts,th->stage_time,th->X,ts->vecs_drdy);CHKERRQ(ierr); 308 if (!ts->costintegralfwd) { 309 /* Evolve ts->vec_costintegral to compute integrals */ 310 ierr = TSAdjointComputeCostIntegrand(ts,ts->ptime,ts->vec_sol,ts->vec_costintegrand);CHKERRQ(ierr); 311 ierr = VecAXPY(ts->vec_costintegral,-ts->time_step*th->Theta,ts->vec_costintegrand);CHKERRQ(ierr); 312 ierr = TSAdjointComputeCostIntegrand(ts,th->stage_time,th->X,ts->vec_costintegrand);CHKERRQ(ierr); 313 ierr = VecAXPY(ts->vec_costintegral,ts->time_step*(th->Theta-1.),ts->vec_costintegrand);CHKERRQ(ierr); 314 } 315 } 316 for (nadj=0; nadj<ts->numcost; nadj++) { 317 ierr = MatMultTranspose(J,VecDeltaLam[nadj],ts->vecs_sensi[nadj]);CHKERRQ(ierr); 318 if (ts->vec_costintegral) { 319 ierr = VecAXPY(ts->vecs_sensi[nadj],-1.,ts->vecs_drdy[nadj]);CHKERRQ(ierr); 320 } 321 ierr = VecScale(ts->vecs_sensi[nadj],1./shift);CHKERRQ(ierr); 322 } 323 324 if (ts->vecs_sensip) { /* sensitivities wrt parameters */ 325 ierr = TSAdjointComputeRHSJacobian(ts,ts->ptime,ts->vec_sol,ts->Jacp);CHKERRQ(ierr); 326 for (nadj=0; nadj<ts->numcost; nadj++) { 327 ierr = MatMultTranspose(ts->Jacp,VecDeltaLam[nadj],VecDeltaMu[nadj]);CHKERRQ(ierr); 328 ierr = VecAXPY(ts->vecs_sensip[nadj],-ts->time_step*th->Theta,VecDeltaMu[nadj]);CHKERRQ(ierr); 329 } 330 ierr = TSAdjointComputeRHSJacobian(ts,th->stage_time,th->X,ts->Jacp);CHKERRQ(ierr); 331 for (nadj=0; nadj<ts->numcost; nadj++) { 332 ierr = MatMultTranspose(ts->Jacp,VecDeltaLam[nadj],VecDeltaMu[nadj]);CHKERRQ(ierr); 333 ierr = VecAXPY(ts->vecs_sensip[nadj],-ts->time_step*(1.-th->Theta),VecDeltaMu[nadj]);CHKERRQ(ierr); 334 } 335 if (ts->vec_costintegral) { 336 ierr = TSAdjointComputeDRDPFunction(ts,ts->ptime,ts->vec_sol,ts->vecs_drdp);CHKERRQ(ierr); 337 for (nadj=0; nadj<ts->numcost; nadj++) { 338 ierr = VecAXPY(ts->vecs_sensip[nadj],-ts->time_step*th->Theta,ts->vecs_drdp[nadj]);CHKERRQ(ierr); 339 } 340 ierr = TSAdjointComputeDRDPFunction(ts,th->stage_time,th->X,ts->vecs_drdp);CHKERRQ(ierr); 341 for (nadj=0; nadj<ts->numcost; nadj++) { 342 ierr = VecAXPY(ts->vecs_sensip[nadj],-ts->time_step*(1.-th->Theta),ts->vecs_drdp[nadj]);CHKERRQ(ierr); 343 } 344 } 345 } 346 }else { /* one-stage case */ 347 shift = 0.0; 348 ierr = TSComputeIJacobian(ts,th->stage_time,th->X,th->Xdot,shift,J,Jp,PETSC_FALSE);CHKERRQ(ierr); /* get -f_y */ 349 if (ts->vec_costintegral) { 350 ierr = TSAdjointComputeDRDYFunction(ts,th->stage_time,th->X,ts->vecs_drdy);CHKERRQ(ierr); 351 if (!ts->costintegralfwd) { 352 /* Evolve ts->vec_costintegral to compute integrals */ 353 ierr = TSAdjointComputeCostIntegrand(ts,th->stage_time,th->X,ts->vec_costintegrand);CHKERRQ(ierr); 354 ierr = VecAXPY(ts->vec_costintegral,-ts->time_step,ts->vec_costintegrand);CHKERRQ(ierr); 355 } 356 } 357 /* When th->endpoint is true and th->Theta==1 (beuler method), the Jacobian is supposed to be evaluated at ts->ptime like this: 358 if(th->endpoint) { 359 ierr = TSComputeIJacobian(ts,ts->ptime,ts->vec_sol,th->Xdot,shift,J,Jp,PETSC_FALSE);CHKERRQ(ierr); 360 } 361 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. 362 */ 363 for (nadj=0; nadj<ts->numcost; nadj++) { 364 ierr = MatMultTranspose(J,VecDeltaLam[nadj],VecSensiTemp[nadj]);CHKERRQ(ierr); 365 ierr = VecAXPY(ts->vecs_sensi[nadj],ts->time_step,VecSensiTemp[nadj]);CHKERRQ(ierr); 366 if (ts->vec_costintegral) { 367 ierr = VecAXPY(ts->vecs_sensi[nadj],-ts->time_step,ts->vecs_drdy[nadj]);CHKERRQ(ierr); 368 } 369 } 370 if (ts->vecs_sensip) { 371 ierr = TSAdjointComputeRHSJacobian(ts,th->stage_time,th->X,ts->Jacp);CHKERRQ(ierr); 372 for (nadj=0; nadj<ts->numcost; nadj++) { 373 ierr = MatMultTranspose(ts->Jacp,VecDeltaLam[nadj],VecDeltaMu[nadj]);CHKERRQ(ierr); 374 ierr = VecAXPY(ts->vecs_sensip[nadj],-ts->time_step,VecDeltaMu[nadj]);CHKERRQ(ierr); 375 } 376 if (ts->vec_costintegral) { 377 ierr = TSAdjointComputeDRDPFunction(ts,th->stage_time,th->X,ts->vecs_drdp);CHKERRQ(ierr); 378 for (nadj=0; nadj<ts->numcost; nadj++) { 379 ierr = VecAXPY(ts->vecs_sensip[nadj],-ts->time_step,ts->vecs_drdp[nadj]);CHKERRQ(ierr); 380 } 381 } 382 } 383 } 384 385 ts->ptime += ts->time_step; 386 ts->steps++; 387 th->status = TS_STEP_COMPLETE; 388 PetscFunctionReturn(0); 389 } 390 391 #undef __FUNCT__ 392 #define __FUNCT__ "TSInterpolate_Theta" 393 static PetscErrorCode TSInterpolate_Theta(TS ts,PetscReal t,Vec X) 394 { 395 TS_Theta *th = (TS_Theta*)ts->data; 396 PetscReal alpha = t - ts->ptime; 397 PetscErrorCode ierr; 398 399 PetscFunctionBegin; 400 ierr = VecCopy(ts->vec_sol,th->X);CHKERRQ(ierr); 401 if (th->endpoint) alpha *= th->Theta; 402 ierr = VecWAXPY(X,alpha,th->Xdot,th->X);CHKERRQ(ierr); 403 PetscFunctionReturn(0); 404 } 405 406 /*------------------------------------------------------------*/ 407 #undef __FUNCT__ 408 #define __FUNCT__ "TSReset_Theta" 409 static PetscErrorCode TSReset_Theta(TS ts) 410 { 411 TS_Theta *th = (TS_Theta*)ts->data; 412 PetscErrorCode ierr; 413 414 PetscFunctionBegin; 415 ierr = VecDestroy(&th->X);CHKERRQ(ierr); 416 ierr = VecDestroy(&th->Xdot);CHKERRQ(ierr); 417 ierr = VecDestroy(&th->X0);CHKERRQ(ierr); 418 ierr = VecDestroy(&th->affine);CHKERRQ(ierr); 419 ierr = VecDestroyVecs(ts->numcost,&th->VecDeltaLam);CHKERRQ(ierr); 420 ierr = VecDestroyVecs(ts->numcost,&th->VecDeltaMu);CHKERRQ(ierr); 421 ierr = VecDestroyVecs(ts->numcost,&th->VecSensiTemp);CHKERRQ(ierr); 422 PetscFunctionReturn(0); 423 } 424 425 #undef __FUNCT__ 426 #define __FUNCT__ "TSDestroy_Theta" 427 static PetscErrorCode TSDestroy_Theta(TS ts) 428 { 429 PetscErrorCode ierr; 430 431 PetscFunctionBegin; 432 ierr = TSReset_Theta(ts);CHKERRQ(ierr); 433 ierr = PetscFree(ts->data);CHKERRQ(ierr); 434 ierr = PetscObjectComposeFunction((PetscObject)ts,"TSThetaGetTheta_C",NULL);CHKERRQ(ierr); 435 ierr = PetscObjectComposeFunction((PetscObject)ts,"TSThetaSetTheta_C",NULL);CHKERRQ(ierr); 436 ierr = PetscObjectComposeFunction((PetscObject)ts,"TSThetaGetEndpoint_C",NULL);CHKERRQ(ierr); 437 ierr = PetscObjectComposeFunction((PetscObject)ts,"TSThetaSetEndpoint_C",NULL);CHKERRQ(ierr); 438 PetscFunctionReturn(0); 439 } 440 441 /* 442 This defines the nonlinear equation that is to be solved with SNES 443 G(U) = F[t0+Theta*dt, U, (U-U0)*shift] = 0 444 */ 445 #undef __FUNCT__ 446 #define __FUNCT__ "SNESTSFormFunction_Theta" 447 static PetscErrorCode SNESTSFormFunction_Theta(SNES snes,Vec x,Vec y,TS ts) 448 { 449 TS_Theta *th = (TS_Theta*)ts->data; 450 PetscErrorCode ierr; 451 Vec X0,Xdot; 452 DM dm,dmsave; 453 PetscReal shift = 1./(th->Theta*ts->time_step); 454 455 PetscFunctionBegin; 456 ierr = SNESGetDM(snes,&dm);CHKERRQ(ierr); 457 /* When using the endpoint variant, this is actually 1/Theta * Xdot */ 458 ierr = TSThetaGetX0AndXdot(ts,dm,&X0,&Xdot);CHKERRQ(ierr); 459 ierr = VecAXPBYPCZ(Xdot,-shift,shift,0,X0,x);CHKERRQ(ierr); 460 461 /* DM monkey-business allows user code to call TSGetDM() inside of functions evaluated on levels of FAS */ 462 dmsave = ts->dm; 463 ts->dm = dm; 464 ierr = TSComputeIFunction(ts,th->stage_time,x,Xdot,y,PETSC_FALSE);CHKERRQ(ierr); 465 ts->dm = dmsave; 466 ierr = TSThetaRestoreX0AndXdot(ts,dm,&X0,&Xdot);CHKERRQ(ierr); 467 PetscFunctionReturn(0); 468 } 469 470 #undef __FUNCT__ 471 #define __FUNCT__ "SNESTSFormJacobian_Theta" 472 static PetscErrorCode SNESTSFormJacobian_Theta(SNES snes,Vec x,Mat A,Mat B,TS ts) 473 { 474 TS_Theta *th = (TS_Theta*)ts->data; 475 PetscErrorCode ierr; 476 Vec Xdot; 477 DM dm,dmsave; 478 PetscReal shift = 1./(th->Theta*ts->time_step); 479 480 PetscFunctionBegin; 481 ierr = SNESGetDM(snes,&dm);CHKERRQ(ierr); 482 483 /* th->Xdot has already been computed in SNESTSFormFunction_Theta (SNES guarantees this) */ 484 ierr = TSThetaGetX0AndXdot(ts,dm,NULL,&Xdot);CHKERRQ(ierr); 485 486 dmsave = ts->dm; 487 ts->dm = dm; 488 ierr = TSComputeIJacobian(ts,th->stage_time,x,Xdot,shift,A,B,PETSC_FALSE);CHKERRQ(ierr); 489 ts->dm = dmsave; 490 ierr = TSThetaRestoreX0AndXdot(ts,dm,NULL,&Xdot);CHKERRQ(ierr); 491 PetscFunctionReturn(0); 492 } 493 494 #undef __FUNCT__ 495 #define __FUNCT__ "TSSetUp_Theta" 496 static PetscErrorCode TSSetUp_Theta(TS ts) 497 { 498 TS_Theta *th = (TS_Theta*)ts->data; 499 PetscErrorCode ierr; 500 SNES snes; 501 TSAdapt adapt; 502 DM dm; 503 504 PetscFunctionBegin; 505 if (!th->X) { 506 ierr = VecDuplicate(ts->vec_sol,&th->X);CHKERRQ(ierr); 507 } 508 if (!th->Xdot) { 509 ierr = VecDuplicate(ts->vec_sol,&th->Xdot);CHKERRQ(ierr); 510 } 511 if (!th->X0) { 512 ierr = VecDuplicate(ts->vec_sol,&th->X0);CHKERRQ(ierr); 513 } 514 ierr = TSGetSNES(ts,&snes);CHKERRQ(ierr); 515 ierr = TSGetDM(ts,&dm);CHKERRQ(ierr); 516 if (dm) { 517 ierr = DMCoarsenHookAdd(dm,DMCoarsenHook_TSTheta,DMRestrictHook_TSTheta,ts);CHKERRQ(ierr); 518 ierr = DMSubDomainHookAdd(dm,DMSubDomainHook_TSTheta,DMSubDomainRestrictHook_TSTheta,ts);CHKERRQ(ierr); 519 } 520 if (th->Theta == 0.5 && th->endpoint) th->order = 2; 521 else th->order = 1; 522 523 ierr = TSGetAdapt(ts,&adapt);CHKERRQ(ierr); 524 if (!th->adapt) { 525 ierr = TSAdaptSetType(adapt,TSADAPTNONE);CHKERRQ(ierr); 526 } 527 PetscFunctionReturn(0); 528 } 529 /*------------------------------------------------------------*/ 530 531 #undef __FUNCT__ 532 #define __FUNCT__ "TSAdjointSetUp_Theta" 533 static PetscErrorCode TSAdjointSetUp_Theta(TS ts) 534 { 535 TS_Theta *th = (TS_Theta*)ts->data; 536 PetscErrorCode ierr; 537 538 PetscFunctionBegin; 539 ierr = VecDuplicateVecs(ts->vecs_sensi[0],ts->numcost,&th->VecDeltaLam);CHKERRQ(ierr); 540 if(ts->vecs_sensip) { 541 ierr = VecDuplicateVecs(ts->vecs_sensip[0],ts->numcost,&th->VecDeltaMu);CHKERRQ(ierr); 542 } 543 ierr = VecDuplicateVecs(ts->vecs_sensi[0],ts->numcost,&th->VecSensiTemp);CHKERRQ(ierr); 544 PetscFunctionReturn(0); 545 } 546 /*------------------------------------------------------------*/ 547 548 #undef __FUNCT__ 549 #define __FUNCT__ "TSSetFromOptions_Theta" 550 static PetscErrorCode TSSetFromOptions_Theta(PetscOptions *PetscOptionsObject,TS ts) 551 { 552 TS_Theta *th = (TS_Theta*)ts->data; 553 PetscErrorCode ierr; 554 555 PetscFunctionBegin; 556 ierr = PetscOptionsHead(PetscOptionsObject,"Theta ODE solver options");CHKERRQ(ierr); 557 { 558 ierr = PetscOptionsReal("-ts_theta_theta","Location of stage (0<Theta<=1)","TSThetaSetTheta",th->Theta,&th->Theta,NULL);CHKERRQ(ierr); 559 ierr = PetscOptionsBool("-ts_theta_extrapolate","Extrapolate stage solution from previous solution (sometimes unstable)","TSThetaSetExtrapolate",th->extrapolate,&th->extrapolate,NULL);CHKERRQ(ierr); 560 ierr = PetscOptionsBool("-ts_theta_endpoint","Use the endpoint instead of midpoint form of the Theta method","TSThetaSetEndpoint",th->endpoint,&th->endpoint,NULL);CHKERRQ(ierr); 561 ierr = PetscOptionsBool("-ts_theta_adapt","Use time-step adaptivity with the Theta method","",th->adapt,&th->adapt,NULL);CHKERRQ(ierr); 562 ierr = SNESSetFromOptions(ts->snes);CHKERRQ(ierr); 563 } 564 ierr = PetscOptionsTail();CHKERRQ(ierr); 565 PetscFunctionReturn(0); 566 } 567 568 #undef __FUNCT__ 569 #define __FUNCT__ "TSView_Theta" 570 static PetscErrorCode TSView_Theta(TS ts,PetscViewer viewer) 571 { 572 TS_Theta *th = (TS_Theta*)ts->data; 573 PetscBool iascii; 574 PetscErrorCode ierr; 575 576 PetscFunctionBegin; 577 ierr = PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERASCII,&iascii);CHKERRQ(ierr); 578 if (iascii) { 579 ierr = PetscViewerASCIIPrintf(viewer," Theta=%g\n",(double)th->Theta);CHKERRQ(ierr); 580 ierr = PetscViewerASCIIPrintf(viewer," Extrapolation=%s\n",th->extrapolate ? "yes" : "no");CHKERRQ(ierr); 581 } 582 if (ts->snes) {ierr = SNESView(ts->snes,viewer);CHKERRQ(ierr);} 583 PetscFunctionReturn(0); 584 } 585 586 #undef __FUNCT__ 587 #define __FUNCT__ "TSThetaGetTheta_Theta" 588 PetscErrorCode TSThetaGetTheta_Theta(TS ts,PetscReal *theta) 589 { 590 TS_Theta *th = (TS_Theta*)ts->data; 591 592 PetscFunctionBegin; 593 *theta = th->Theta; 594 PetscFunctionReturn(0); 595 } 596 597 #undef __FUNCT__ 598 #define __FUNCT__ "TSThetaSetTheta_Theta" 599 PetscErrorCode TSThetaSetTheta_Theta(TS ts,PetscReal theta) 600 { 601 TS_Theta *th = (TS_Theta*)ts->data; 602 603 PetscFunctionBegin; 604 if (theta <= 0 || 1 < theta) SETERRQ1(PetscObjectComm((PetscObject)ts),PETSC_ERR_ARG_OUTOFRANGE,"Theta %g not in range (0,1]",(double)theta); 605 th->Theta = theta; 606 PetscFunctionReturn(0); 607 } 608 609 #undef __FUNCT__ 610 #define __FUNCT__ "TSThetaGetEndpoint_Theta" 611 PetscErrorCode TSThetaGetEndpoint_Theta(TS ts,PetscBool *endpoint) 612 { 613 TS_Theta *th = (TS_Theta*)ts->data; 614 615 PetscFunctionBegin; 616 *endpoint = th->endpoint; 617 PetscFunctionReturn(0); 618 } 619 620 #undef __FUNCT__ 621 #define __FUNCT__ "TSThetaSetEndpoint_Theta" 622 PetscErrorCode TSThetaSetEndpoint_Theta(TS ts,PetscBool flg) 623 { 624 TS_Theta *th = (TS_Theta*)ts->data; 625 626 PetscFunctionBegin; 627 th->endpoint = flg; 628 PetscFunctionReturn(0); 629 } 630 631 #if defined(PETSC_HAVE_COMPLEX) 632 #undef __FUNCT__ 633 #define __FUNCT__ "TSComputeLinearStability_Theta" 634 static PetscErrorCode TSComputeLinearStability_Theta(TS ts,PetscReal xr,PetscReal xi,PetscReal *yr,PetscReal *yi) 635 { 636 PetscComplex z = xr + xi*PETSC_i,f; 637 TS_Theta *th = (TS_Theta*)ts->data; 638 const PetscReal one = 1.0; 639 640 PetscFunctionBegin; 641 f = (one + (one - th->Theta)*z)/(one - th->Theta*z); 642 *yr = PetscRealPartComplex(f); 643 *yi = PetscImaginaryPartComplex(f); 644 PetscFunctionReturn(0); 645 } 646 #endif 647 648 #undef __FUNCT__ 649 #define __FUNCT__ "TSGetStages_Theta" 650 static PetscErrorCode TSGetStages_Theta(TS ts,PetscInt *ns,Vec **Y) 651 { 652 TS_Theta *th = (TS_Theta*)ts->data; 653 654 PetscFunctionBegin; 655 *ns = 1; 656 if(Y) { 657 *Y = &(th->X); 658 } 659 PetscFunctionReturn(0); 660 } 661 662 /* ------------------------------------------------------------ */ 663 /*MC 664 TSTHETA - DAE solver using the implicit Theta method 665 666 Level: beginner 667 668 Options Database: 669 -ts_theta_theta <Theta> - Location of stage (0<Theta<=1) 670 -ts_theta_extrapolate <flg> Extrapolate stage solution from previous solution (sometimes unstable) 671 -ts_theta_endpoint <flag> - Use the endpoint (like Crank-Nicholson) instead of midpoint form of the Theta method 672 673 Notes: 674 $ -ts_type theta -ts_theta_theta 1.0 corresponds to backward Euler (TSBEULER) 675 $ -ts_type theta -ts_theta_theta 0.5 corresponds to the implicit midpoint rule 676 $ -ts_type theta -ts_theta_theta 0.5 -ts_theta_endpoint corresponds to Crank-Nicholson (TSCN) 677 678 679 680 This method can be applied to DAE. 681 682 This method is cast as a 1-stage implicit Runge-Kutta method. 683 684 .vb 685 Theta | Theta 686 ------------- 687 | 1 688 .ve 689 690 For the default Theta=0.5, this is also known as the implicit midpoint rule. 691 692 When the endpoint variant is chosen, the method becomes a 2-stage method with first stage explicit: 693 694 .vb 695 0 | 0 0 696 1 | 1-Theta Theta 697 ------------------- 698 | 1-Theta Theta 699 .ve 700 701 For the default Theta=0.5, this is the trapezoid rule (also known as Crank-Nicolson, see TSCN). 702 703 To apply a diagonally implicit RK method to DAE, the stage formula 704 705 $ Y_i = X + h sum_j a_ij Y'_j 706 707 is interpreted as a formula for Y'_i in terms of Y_i and known values (Y'_j, j<i) 708 709 .seealso: TSCreate(), TS, TSSetType(), TSCN, TSBEULER, TSThetaSetTheta(), TSThetaSetEndpoint() 710 711 M*/ 712 #undef __FUNCT__ 713 #define __FUNCT__ "TSCreate_Theta" 714 PETSC_EXTERN PetscErrorCode TSCreate_Theta(TS ts) 715 { 716 TS_Theta *th; 717 PetscErrorCode ierr; 718 719 PetscFunctionBegin; 720 ts->ops->reset = TSReset_Theta; 721 ts->ops->destroy = TSDestroy_Theta; 722 ts->ops->view = TSView_Theta; 723 ts->ops->setup = TSSetUp_Theta; 724 ts->ops->adjointsetup = TSAdjointSetUp_Theta; 725 ts->ops->step = TSStep_Theta; 726 ts->ops->interpolate = TSInterpolate_Theta; 727 ts->ops->evaluatestep = TSEvaluateStep_Theta; 728 ts->ops->rollback = TSRollBack_Theta; 729 ts->ops->setfromoptions = TSSetFromOptions_Theta; 730 ts->ops->snesfunction = SNESTSFormFunction_Theta; 731 ts->ops->snesjacobian = SNESTSFormJacobian_Theta; 732 #if defined(PETSC_HAVE_COMPLEX) 733 ts->ops->linearstability = TSComputeLinearStability_Theta; 734 #endif 735 ts->ops->getstages = TSGetStages_Theta; 736 ts->ops->adjointstep = TSAdjointStep_Theta; 737 738 ierr = PetscNewLog(ts,&th);CHKERRQ(ierr); 739 ts->data = (void*)th; 740 741 th->extrapolate = PETSC_FALSE; 742 th->Theta = 0.5; 743 th->ccfl = 1.0; 744 th->adapt = PETSC_FALSE; 745 ierr = PetscObjectComposeFunction((PetscObject)ts,"TSThetaGetTheta_C",TSThetaGetTheta_Theta);CHKERRQ(ierr); 746 ierr = PetscObjectComposeFunction((PetscObject)ts,"TSThetaSetTheta_C",TSThetaSetTheta_Theta);CHKERRQ(ierr); 747 ierr = PetscObjectComposeFunction((PetscObject)ts,"TSThetaGetEndpoint_C",TSThetaGetEndpoint_Theta);CHKERRQ(ierr); 748 ierr = PetscObjectComposeFunction((PetscObject)ts,"TSThetaSetEndpoint_C",TSThetaSetEndpoint_Theta);CHKERRQ(ierr); 749 PetscFunctionReturn(0); 750 } 751 752 #undef __FUNCT__ 753 #define __FUNCT__ "TSThetaGetTheta" 754 /*@ 755 TSThetaGetTheta - Get the abscissa of the stage in (0,1]. 756 757 Not Collective 758 759 Input Parameter: 760 . ts - timestepping context 761 762 Output Parameter: 763 . theta - stage abscissa 764 765 Note: 766 Use of this function is normally only required to hack TSTHETA to use a modified integration scheme. 767 768 Level: Advanced 769 770 .seealso: TSThetaSetTheta() 771 @*/ 772 PetscErrorCode TSThetaGetTheta(TS ts,PetscReal *theta) 773 { 774 PetscErrorCode ierr; 775 776 PetscFunctionBegin; 777 PetscValidHeaderSpecific(ts,TS_CLASSID,1); 778 PetscValidPointer(theta,2); 779 ierr = PetscUseMethod(ts,"TSThetaGetTheta_C",(TS,PetscReal*),(ts,theta));CHKERRQ(ierr); 780 PetscFunctionReturn(0); 781 } 782 783 #undef __FUNCT__ 784 #define __FUNCT__ "TSThetaSetTheta" 785 /*@ 786 TSThetaSetTheta - Set the abscissa of the stage in (0,1]. 787 788 Not Collective 789 790 Input Parameter: 791 + ts - timestepping context 792 - theta - stage abscissa 793 794 Options Database: 795 . -ts_theta_theta <theta> 796 797 Level: Intermediate 798 799 .seealso: TSThetaGetTheta() 800 @*/ 801 PetscErrorCode TSThetaSetTheta(TS ts,PetscReal theta) 802 { 803 PetscErrorCode ierr; 804 805 PetscFunctionBegin; 806 PetscValidHeaderSpecific(ts,TS_CLASSID,1); 807 ierr = PetscTryMethod(ts,"TSThetaSetTheta_C",(TS,PetscReal),(ts,theta));CHKERRQ(ierr); 808 PetscFunctionReturn(0); 809 } 810 811 #undef __FUNCT__ 812 #define __FUNCT__ "TSThetaGetEndpoint" 813 /*@ 814 TSThetaGetEndpoint - Gets whether to use the endpoint variant of the method (e.g. trapezoid/Crank-Nicolson instead of midpoint rule). 815 816 Not Collective 817 818 Input Parameter: 819 . ts - timestepping context 820 821 Output Parameter: 822 . endpoint - PETSC_TRUE when using the endpoint variant 823 824 Level: Advanced 825 826 .seealso: TSThetaSetEndpoint(), TSTHETA, TSCN 827 @*/ 828 PetscErrorCode TSThetaGetEndpoint(TS ts,PetscBool *endpoint) 829 { 830 PetscErrorCode ierr; 831 832 PetscFunctionBegin; 833 PetscValidHeaderSpecific(ts,TS_CLASSID,1); 834 PetscValidPointer(endpoint,2); 835 ierr = PetscTryMethod(ts,"TSThetaGetEndpoint_C",(TS,PetscBool*),(ts,endpoint));CHKERRQ(ierr); 836 PetscFunctionReturn(0); 837 } 838 839 #undef __FUNCT__ 840 #define __FUNCT__ "TSThetaSetEndpoint" 841 /*@ 842 TSThetaSetEndpoint - Sets whether to use the endpoint variant of the method (e.g. trapezoid/Crank-Nicolson instead of midpoint rule). 843 844 Not Collective 845 846 Input Parameter: 847 + ts - timestepping context 848 - flg - PETSC_TRUE to use the endpoint variant 849 850 Options Database: 851 . -ts_theta_endpoint <flg> 852 853 Level: Intermediate 854 855 .seealso: TSTHETA, TSCN 856 @*/ 857 PetscErrorCode TSThetaSetEndpoint(TS ts,PetscBool flg) 858 { 859 PetscErrorCode ierr; 860 861 PetscFunctionBegin; 862 PetscValidHeaderSpecific(ts,TS_CLASSID,1); 863 ierr = PetscTryMethod(ts,"TSThetaSetEndpoint_C",(TS,PetscBool),(ts,flg));CHKERRQ(ierr); 864 PetscFunctionReturn(0); 865 } 866 867 /* 868 * TSBEULER and TSCN are straightforward specializations of TSTHETA. 869 * The creation functions for these specializations are below. 870 */ 871 872 #undef __FUNCT__ 873 #define __FUNCT__ "TSView_BEuler" 874 static PetscErrorCode TSView_BEuler(TS ts,PetscViewer viewer) 875 { 876 PetscErrorCode ierr; 877 878 PetscFunctionBegin; 879 ierr = SNESView(ts->snes,viewer);CHKERRQ(ierr); 880 PetscFunctionReturn(0); 881 } 882 883 /*MC 884 TSBEULER - ODE solver using the implicit backward Euler method 885 886 Level: beginner 887 888 Notes: 889 TSBEULER is equivalent to TSTHETA with Theta=1.0 890 891 $ -ts_type theta -ts_theta_theta 1. 892 893 .seealso: TSCreate(), TS, TSSetType(), TSEULER, TSCN, TSTHETA 894 895 M*/ 896 #undef __FUNCT__ 897 #define __FUNCT__ "TSCreate_BEuler" 898 PETSC_EXTERN PetscErrorCode TSCreate_BEuler(TS ts) 899 { 900 PetscErrorCode ierr; 901 902 PetscFunctionBegin; 903 ierr = TSCreate_Theta(ts);CHKERRQ(ierr); 904 ierr = TSThetaSetTheta(ts,1.0);CHKERRQ(ierr); 905 ts->ops->view = TSView_BEuler; 906 PetscFunctionReturn(0); 907 } 908 909 #undef __FUNCT__ 910 #define __FUNCT__ "TSView_CN" 911 static PetscErrorCode TSView_CN(TS ts,PetscViewer viewer) 912 { 913 PetscErrorCode ierr; 914 915 PetscFunctionBegin; 916 ierr = SNESView(ts->snes,viewer);CHKERRQ(ierr); 917 PetscFunctionReturn(0); 918 } 919 920 /*MC 921 TSCN - ODE solver using the implicit Crank-Nicolson method. 922 923 Level: beginner 924 925 Notes: 926 TSCN is equivalent to TSTHETA with Theta=0.5 and the "endpoint" option set. I.e. 927 928 $ -ts_type theta -ts_theta_theta 0.5 -ts_theta_endpoint 929 930 .seealso: TSCreate(), TS, TSSetType(), TSBEULER, TSTHETA 931 932 M*/ 933 #undef __FUNCT__ 934 #define __FUNCT__ "TSCreate_CN" 935 PETSC_EXTERN PetscErrorCode TSCreate_CN(TS ts) 936 { 937 PetscErrorCode ierr; 938 939 PetscFunctionBegin; 940 ierr = TSCreate_Theta(ts);CHKERRQ(ierr); 941 ierr = TSThetaSetTheta(ts,0.5);CHKERRQ(ierr); 942 ierr = TSThetaSetEndpoint(ts,PETSC_TRUE);CHKERRQ(ierr); 943 ts->ops->view = TSView_CN; 944 PetscFunctionReturn(0); 945 } 946