Actual source code: tssen.c
1: #include <petsc/private/tsimpl.h>
2: #include <petscdraw.h>
4: PetscLogEvent TS_AdjointStep, TS_ForwardStep, TS_JacobianPEval;
6: /* #define TSADJOINT_STAGE */
8: /* ------------------------ Sensitivity Context ---------------------------*/
10: /*@C
11: TSSetRHSJacobianP - Sets the function that computes the Jacobian of G w.r.t. the parameters P where U_t = G(U,P,t), as well as the location to store the matrix.
13: Logically Collective
15: Input Parameters:
16: + ts - `TS` context obtained from `TSCreate()`
17: . Amat - JacobianP matrix
18: . func - function
19: - ctx - [optional] user-defined function context
21: Level: intermediate
23: Note:
24: `Amat` has the same number of rows and the same row parallel layout as `u`, `Amat` has the same number of columns and parallel layout as `p`
26: .seealso: [](ch_ts), `TS`, `TSRHSJacobianP`, `TSGetRHSJacobianP()`
27: @*/
28: PetscErrorCode TSSetRHSJacobianP(TS ts, Mat Amat, TSRHSJacobianP func, void *ctx)
29: {
30: PetscFunctionBegin;
34: ts->rhsjacobianp = func;
35: ts->rhsjacobianpctx = ctx;
36: if (Amat) {
37: PetscCall(PetscObjectReference((PetscObject)Amat));
38: PetscCall(MatDestroy(&ts->Jacprhs));
39: ts->Jacprhs = Amat;
40: }
41: PetscFunctionReturn(PETSC_SUCCESS);
42: }
44: /*@C
45: TSGetRHSJacobianP - Gets the function that computes the Jacobian of G w.r.t. the parameters P where U_t = G(U,P,t), as well as the location to store the matrix.
47: Logically Collective
49: Input Parameter:
50: . ts - `TS` context obtained from `TSCreate()`
52: Output Parameters:
53: + Amat - JacobianP matrix
54: . func - function
55: - ctx - [optional] user-defined function context
57: Level: intermediate
59: Note:
60: `Amat` has the same number of rows and the same row parallel layout as `u`, `Amat` has the same number of columns and parallel layout as `p`
62: .seealso: [](ch_ts), `TSSetRHSJacobianP()`, `TS`, `TSRHSJacobianP`
63: @*/
64: PetscErrorCode TSGetRHSJacobianP(TS ts, Mat *Amat, TSRHSJacobianP *func, void **ctx)
65: {
66: PetscFunctionBegin;
67: if (func) *func = ts->rhsjacobianp;
68: if (ctx) *ctx = ts->rhsjacobianpctx;
69: if (Amat) *Amat = ts->Jacprhs;
70: PetscFunctionReturn(PETSC_SUCCESS);
71: }
73: /*@C
74: TSComputeRHSJacobianP - Runs the user-defined JacobianP function.
76: Collective
78: Input Parameters:
79: + ts - The `TS` context obtained from `TSCreate()`
80: . t - the time
81: - U - the solution at which to compute the Jacobian
83: Output Parameter:
84: . Amat - the computed Jacobian
86: Level: developer
88: .seealso: [](ch_ts), `TSSetRHSJacobianP()`, `TS`
89: @*/
90: PetscErrorCode TSComputeRHSJacobianP(TS ts, PetscReal t, Vec U, Mat Amat)
91: {
92: PetscFunctionBegin;
93: if (!Amat) PetscFunctionReturn(PETSC_SUCCESS);
97: if (ts->rhsjacobianp) PetscCallBack("TS callback JacobianP for sensitivity analysis", (*ts->rhsjacobianp)(ts, t, U, Amat, ts->rhsjacobianpctx));
98: else {
99: PetscBool assembled;
100: PetscCall(MatZeroEntries(Amat));
101: PetscCall(MatAssembled(Amat, &assembled));
102: if (!assembled) {
103: PetscCall(MatAssemblyBegin(Amat, MAT_FINAL_ASSEMBLY));
104: PetscCall(MatAssemblyEnd(Amat, MAT_FINAL_ASSEMBLY));
105: }
106: }
107: PetscFunctionReturn(PETSC_SUCCESS);
108: }
110: /*@C
111: TSSetIJacobianP - Sets the function that computes the Jacobian of F w.r.t. the parameters P where F(Udot,U,t) = G(U,P,t), as well as the location to store the matrix.
113: Logically Collective
115: Input Parameters:
116: + ts - `TS` context obtained from `TSCreate()`
117: . Amat - JacobianP matrix
118: . func - function
119: - ctx - [optional] user-defined function context
121: Calling sequence of `func`:
122: + ts - the `TS` context
123: . t - current timestep
124: . U - input vector (current ODE solution)
125: . Udot - time derivative of state vector
126: . shift - shift to apply, see note below
127: . A - output matrix
128: - ctx - [optional] user-defined function context
130: Level: intermediate
132: Note:
133: Amat has the same number of rows and the same row parallel layout as u, Amat has the same number of columns and parallel layout as p
135: .seealso: [](ch_ts), `TSSetRHSJacobianP()`, `TS`
136: @*/
137: PetscErrorCode TSSetIJacobianP(TS ts, Mat Amat, PetscErrorCode (*func)(TS ts, PetscReal t, Vec U, Vec Udot, PetscReal shift, Mat A, void *ctx), void *ctx)
138: {
139: PetscFunctionBegin;
143: ts->ijacobianp = func;
144: ts->ijacobianpctx = ctx;
145: if (Amat) {
146: PetscCall(PetscObjectReference((PetscObject)Amat));
147: PetscCall(MatDestroy(&ts->Jacp));
148: ts->Jacp = Amat;
149: }
150: PetscFunctionReturn(PETSC_SUCCESS);
151: }
153: /*@C
154: TSComputeIJacobianP - Runs the user-defined IJacobianP function.
156: Collective
158: Input Parameters:
159: + ts - the `TS` context
160: . t - current timestep
161: . U - state vector
162: . Udot - time derivative of state vector
163: . shift - shift to apply, see note below
164: - imex - flag indicates if the method is IMEX so that the RHSJacobianP should be kept separate
166: Output Parameter:
167: . Amat - Jacobian matrix
169: Level: developer
171: .seealso: [](ch_ts), `TS`, `TSSetIJacobianP()`
172: @*/
173: PetscErrorCode TSComputeIJacobianP(TS ts, PetscReal t, Vec U, Vec Udot, PetscReal shift, Mat Amat, PetscBool imex)
174: {
175: PetscFunctionBegin;
176: if (!Amat) PetscFunctionReturn(PETSC_SUCCESS);
181: PetscCall(PetscLogEventBegin(TS_JacobianPEval, ts, U, Amat, 0));
182: if (ts->ijacobianp) PetscCallBack("TS callback JacobianP for sensitivity analysis", (*ts->ijacobianp)(ts, t, U, Udot, shift, Amat, ts->ijacobianpctx));
183: if (imex) {
184: if (!ts->ijacobianp) { /* system was written as Udot = G(t,U) */
185: PetscBool assembled;
186: PetscCall(MatZeroEntries(Amat));
187: PetscCall(MatAssembled(Amat, &assembled));
188: if (!assembled) {
189: PetscCall(MatAssemblyBegin(Amat, MAT_FINAL_ASSEMBLY));
190: PetscCall(MatAssemblyEnd(Amat, MAT_FINAL_ASSEMBLY));
191: }
192: }
193: } else {
194: if (ts->rhsjacobianp) PetscCall(TSComputeRHSJacobianP(ts, t, U, ts->Jacprhs));
195: if (ts->Jacprhs == Amat) { /* No IJacobian, so we only have the RHS matrix */
196: PetscCall(MatScale(Amat, -1));
197: } else if (ts->Jacprhs) { /* Both IJacobian and RHSJacobian */
198: MatStructure axpy = DIFFERENT_NONZERO_PATTERN;
199: if (!ts->ijacobianp) { /* No IJacobianp provided, but we have a separate RHS matrix */
200: PetscCall(MatZeroEntries(Amat));
201: }
202: PetscCall(MatAXPY(Amat, -1, ts->Jacprhs, axpy));
203: }
204: }
205: PetscCall(PetscLogEventEnd(TS_JacobianPEval, ts, U, Amat, 0));
206: PetscFunctionReturn(PETSC_SUCCESS);
207: }
209: /*@C
210: TSSetCostIntegrand - Sets the routine for evaluating the integral term in one or more cost functions
212: Logically Collective
214: Input Parameters:
215: + ts - the `TS` context obtained from `TSCreate()`
216: . numcost - number of gradients to be computed, this is the number of cost functions
217: . costintegral - vector that stores the integral values
218: . rf - routine for evaluating the integrand function
219: . drduf - function that computes the gradients of the r's with respect to u
220: . drdpf - function that computes the gradients of the r's with respect to p, can be `NULL` if parametric sensitivity is not desired (`mu` = `NULL`)
221: . fwd - flag indicating whether to evaluate cost integral in the forward run or the adjoint run
222: - ctx - [optional] user-defined context for private data for the function evaluation routine (may be `NULL`)
224: Calling sequence of `rf`:
225: $ PetscErrorCode rf(TS ts, PetscReal t, Vec U, Vec F, oid *ctx)
227: Calling sequence of `drduf`:
228: $ PetscErroCode drduf(TS ts, PetscReal t, Vec U, Vec *dRdU, void *ctx)
230: Calling sequence of `drdpf`:
231: $ PetscErroCode drdpf(TS ts, PetscReal t, Vec U, Vec *dRdP, void *ctx)
233: Level: deprecated
235: Note:
236: For optimization there is usually a single cost function (numcost = 1). For sensitivities there may be multiple cost functions
238: .seealso: [](ch_ts), `TS`, `TSSetRHSJacobianP()`, `TSGetCostGradients()`, `TSSetCostGradients()`
239: @*/
240: PetscErrorCode TSSetCostIntegrand(TS ts, PetscInt numcost, Vec costintegral, PetscErrorCode (*rf)(TS, PetscReal, Vec, Vec, void *), PetscErrorCode (*drduf)(TS, PetscReal, Vec, Vec *, void *), PetscErrorCode (*drdpf)(TS, PetscReal, Vec, Vec *, void *), PetscBool fwd, void *ctx)
241: {
242: PetscFunctionBegin;
245: PetscCheck(!ts->numcost || ts->numcost == numcost, PetscObjectComm((PetscObject)ts), PETSC_ERR_USER, "The number of cost functions (2nd parameter of TSSetCostIntegrand()) is inconsistent with the one set by TSSetCostGradients() or TSForwardSetIntegralGradients()");
246: if (!ts->numcost) ts->numcost = numcost;
248: if (costintegral) {
249: PetscCall(PetscObjectReference((PetscObject)costintegral));
250: PetscCall(VecDestroy(&ts->vec_costintegral));
251: ts->vec_costintegral = costintegral;
252: } else {
253: if (!ts->vec_costintegral) { /* Create a seq vec if user does not provide one */
254: PetscCall(VecCreateSeq(PETSC_COMM_SELF, numcost, &ts->vec_costintegral));
255: } else {
256: PetscCall(VecSet(ts->vec_costintegral, 0.0));
257: }
258: }
259: if (!ts->vec_costintegrand) {
260: PetscCall(VecDuplicate(ts->vec_costintegral, &ts->vec_costintegrand));
261: } else {
262: PetscCall(VecSet(ts->vec_costintegrand, 0.0));
263: }
264: ts->costintegralfwd = fwd; /* Evaluate the cost integral in forward run if fwd is true */
265: ts->costintegrand = rf;
266: ts->costintegrandctx = ctx;
267: ts->drdufunction = drduf;
268: ts->drdpfunction = drdpf;
269: PetscFunctionReturn(PETSC_SUCCESS);
270: }
272: /*@C
273: TSGetCostIntegral - Returns the values of the integral term in the cost functions.
274: It is valid to call the routine after a backward run.
276: Not Collective
278: Input Parameter:
279: . ts - the `TS` context obtained from `TSCreate()`
281: Output Parameter:
282: . v - the vector containing the integrals for each cost function
284: Level: intermediate
286: .seealso: [](ch_ts), `TS`, `TSAdjointSolve()`, ``TSSetCostIntegrand()`
287: @*/
288: PetscErrorCode TSGetCostIntegral(TS ts, Vec *v)
289: {
290: TS quadts;
292: PetscFunctionBegin;
294: PetscAssertPointer(v, 2);
295: PetscCall(TSGetQuadratureTS(ts, NULL, &quadts));
296: *v = quadts->vec_sol;
297: PetscFunctionReturn(PETSC_SUCCESS);
298: }
300: /*@C
301: TSComputeCostIntegrand - Evaluates the integral function in the cost functions.
303: Input Parameters:
304: + ts - the `TS` context
305: . t - current time
306: - U - state vector, i.e. current solution
308: Output Parameter:
309: . Q - vector of size numcost to hold the outputs
311: Level: deprecated
313: Note:
314: Most users should not need to explicitly call this routine, as it
315: is used internally within the sensitivity analysis context.
317: .seealso: [](ch_ts), `TS`, `TSAdjointSolve()`, `TSSetCostIntegrand()`
318: @*/
319: PetscErrorCode TSComputeCostIntegrand(TS ts, PetscReal t, Vec U, Vec Q)
320: {
321: PetscFunctionBegin;
326: PetscCall(PetscLogEventBegin(TS_FunctionEval, ts, U, Q, 0));
327: if (ts->costintegrand) PetscCallBack("TS callback integrand in the cost function", (*ts->costintegrand)(ts, t, U, Q, ts->costintegrandctx));
328: else PetscCall(VecZeroEntries(Q));
329: PetscCall(PetscLogEventEnd(TS_FunctionEval, ts, U, Q, 0));
330: PetscFunctionReturn(PETSC_SUCCESS);
331: }
333: // PetscClangLinter pragma disable: -fdoc-*
334: /*@C
335: TSComputeDRDUFunction - Deprecated, use `TSGetQuadratureTS()` then `TSComputeRHSJacobian()`
337: Level: deprecated
339: @*/
340: PetscErrorCode TSComputeDRDUFunction(TS ts, PetscReal t, Vec U, Vec *DRDU)
341: {
342: PetscFunctionBegin;
343: if (!DRDU) PetscFunctionReturn(PETSC_SUCCESS);
347: PetscCallBack("TS callback DRDU for sensitivity analysis", (*ts->drdufunction)(ts, t, U, DRDU, ts->costintegrandctx));
348: PetscFunctionReturn(PETSC_SUCCESS);
349: }
351: // PetscClangLinter pragma disable: -fdoc-*
352: /*@C
353: TSComputeDRDPFunction - Deprecated, use `TSGetQuadratureTS()` then `TSComputeRHSJacobianP()`
355: Level: deprecated
357: @*/
358: PetscErrorCode TSComputeDRDPFunction(TS ts, PetscReal t, Vec U, Vec *DRDP)
359: {
360: PetscFunctionBegin;
361: if (!DRDP) PetscFunctionReturn(PETSC_SUCCESS);
365: PetscCallBack("TS callback DRDP for sensitivity analysis", (*ts->drdpfunction)(ts, t, U, DRDP, ts->costintegrandctx));
366: PetscFunctionReturn(PETSC_SUCCESS);
367: }
369: // PetscClangLinter pragma disable: -fdoc-param-list-func-parameter-documentation
370: // PetscClangLinter pragma disable: -fdoc-section-header-unknown
371: /*@C
372: TSSetIHessianProduct - Sets the function that computes the vector-Hessian-vector product. The Hessian is the second-order derivative of F (IFunction) w.r.t. the state variable.
374: Logically Collective
376: Input Parameters:
377: + ts - `TS` context obtained from `TSCreate()`
378: . ihp1 - an array of vectors storing the result of vector-Hessian-vector product for F_UU
379: . hessianproductfunc1 - vector-Hessian-vector product function for F_UU
380: . ihp2 - an array of vectors storing the result of vector-Hessian-vector product for F_UP
381: . hessianproductfunc2 - vector-Hessian-vector product function for F_UP
382: . ihp3 - an array of vectors storing the result of vector-Hessian-vector product for F_PU
383: . hessianproductfunc3 - vector-Hessian-vector product function for F_PU
384: . ihp4 - an array of vectors storing the result of vector-Hessian-vector product for F_PP
385: - hessianproductfunc4 - vector-Hessian-vector product function for F_PP
387: Calling sequence of `ihessianproductfunc1`:
388: + ts - the `TS` context
389: + t - current timestep
390: . U - input vector (current ODE solution)
391: . Vl - an array of input vectors to be left-multiplied with the Hessian
392: . Vr - input vector to be right-multiplied with the Hessian
393: . VHV - an array of output vectors for vector-Hessian-vector product
394: - ctx - [optional] user-defined function context
396: Level: intermediate
398: Notes:
399: All other functions have the same calling sequence as `rhhessianproductfunc1`, so their
400: descriptions are omitted for brevity.
402: The first Hessian function and the working array are required.
403: As an example to implement the callback functions, the second callback function calculates the vector-Hessian-vector product
404: $ Vl_n^T*F_UP*Vr
405: where the vector Vl_n (n-th element in the array Vl) and Vr are of size N and M respectively, and the Hessian F_UP is of size N x N x M.
406: Each entry of F_UP corresponds to the derivative
407: $ F_UP[i][j][k] = \frac{\partial^2 F[i]}{\partial U[j] \partial P[k]}.
408: The result of the vector-Hessian-vector product for Vl_n needs to be stored in vector VHV_n with the j-th entry being
409: $ VHV_n[j] = \sum_i \sum_k {Vl_n[i] * F_UP[i][j][k] * Vr[k]}
410: If the cost function is a scalar, there will be only one vector in Vl and VHV.
412: .seealso: [](ch_ts), `TS`
413: @*/
414: PetscErrorCode TSSetIHessianProduct(TS ts, Vec *ihp1, PetscErrorCode (*ihessianproductfunc1)(TS ts, PetscReal t, Vec U, Vec *Vl, Vec Vr, Vec *VHV, void *ctx), Vec *ihp2, PetscErrorCode (*ihessianproductfunc2)(TS, PetscReal, Vec, Vec *, Vec, Vec *, void *), Vec *ihp3, PetscErrorCode (*ihessianproductfunc3)(TS, PetscReal, Vec, Vec *, Vec, Vec *, void *), Vec *ihp4, PetscErrorCode (*ihessianproductfunc4)(TS, PetscReal, Vec, Vec *, Vec, Vec *, void *), void *ctx)
415: {
416: PetscFunctionBegin;
418: PetscAssertPointer(ihp1, 2);
420: ts->ihessianproductctx = ctx;
421: if (ihp1) ts->vecs_fuu = ihp1;
422: if (ihp2) ts->vecs_fup = ihp2;
423: if (ihp3) ts->vecs_fpu = ihp3;
424: if (ihp4) ts->vecs_fpp = ihp4;
425: ts->ihessianproduct_fuu = ihessianproductfunc1;
426: ts->ihessianproduct_fup = ihessianproductfunc2;
427: ts->ihessianproduct_fpu = ihessianproductfunc3;
428: ts->ihessianproduct_fpp = ihessianproductfunc4;
429: PetscFunctionReturn(PETSC_SUCCESS);
430: }
432: /*@C
433: TSComputeIHessianProductFunctionUU - Runs the user-defined vector-Hessian-vector product function for Fuu.
435: Collective
437: Input Parameters:
438: + ts - The `TS` context obtained from `TSCreate()`
439: . t - the time
440: . U - the solution at which to compute the Hessian product
441: . Vl - the array of input vectors to be multiplied with the Hessian from the left
442: - Vr - the input vector to be multiplied with the Hessian from the right
444: Output Parameter:
445: . VHV - the array of output vectors that store the Hessian product
447: Level: developer
449: Note:
450: `TSComputeIHessianProductFunctionUU()` is typically used for sensitivity implementation,
451: so most users would not generally call this routine themselves.
453: .seealso: [](ch_ts), `TSSetIHessianProduct()`
454: @*/
455: PetscErrorCode TSComputeIHessianProductFunctionUU(TS ts, PetscReal t, Vec U, Vec *Vl, Vec Vr, Vec *VHV)
456: {
457: PetscFunctionBegin;
458: if (!VHV) PetscFunctionReturn(PETSC_SUCCESS);
462: if (ts->ihessianproduct_fuu) PetscCallBack("TS callback IHessianProduct 1 for sensitivity analysis", (*ts->ihessianproduct_fuu)(ts, t, U, Vl, Vr, VHV, ts->ihessianproductctx));
464: /* does not consider IMEX for now, so either IHessian or RHSHessian will be calculated, using the same output VHV */
465: if (ts->rhshessianproduct_guu) {
466: PetscInt nadj;
467: PetscCall(TSComputeRHSHessianProductFunctionUU(ts, t, U, Vl, Vr, VHV));
468: for (nadj = 0; nadj < ts->numcost; nadj++) PetscCall(VecScale(VHV[nadj], -1));
469: }
470: PetscFunctionReturn(PETSC_SUCCESS);
471: }
473: /*@C
474: TSComputeIHessianProductFunctionUP - Runs the user-defined vector-Hessian-vector product function for Fup.
476: Collective
478: Input Parameters:
479: + ts - The `TS` context obtained from `TSCreate()`
480: . t - the time
481: . U - the solution at which to compute the Hessian product
482: . Vl - the array of input vectors to be multiplied with the Hessian from the left
483: - Vr - the input vector to be multiplied with the Hessian from the right
485: Output Parameter:
486: . VHV - the array of output vectors that store the Hessian product
488: Level: developer
490: Note:
491: `TSComputeIHessianProductFunctionUP()` is typically used for sensitivity implementation,
492: so most users would not generally call this routine themselves.
494: .seealso: [](ch_ts), `TSSetIHessianProduct()`
495: @*/
496: PetscErrorCode TSComputeIHessianProductFunctionUP(TS ts, PetscReal t, Vec U, Vec *Vl, Vec Vr, Vec *VHV)
497: {
498: PetscFunctionBegin;
499: if (!VHV) PetscFunctionReturn(PETSC_SUCCESS);
503: if (ts->ihessianproduct_fup) PetscCallBack("TS callback IHessianProduct 2 for sensitivity analysis", (*ts->ihessianproduct_fup)(ts, t, U, Vl, Vr, VHV, ts->ihessianproductctx));
505: /* does not consider IMEX for now, so either IHessian or RHSHessian will be calculated, using the same output VHV */
506: if (ts->rhshessianproduct_gup) {
507: PetscInt nadj;
508: PetscCall(TSComputeRHSHessianProductFunctionUP(ts, t, U, Vl, Vr, VHV));
509: for (nadj = 0; nadj < ts->numcost; nadj++) PetscCall(VecScale(VHV[nadj], -1));
510: }
511: PetscFunctionReturn(PETSC_SUCCESS);
512: }
514: /*@C
515: TSComputeIHessianProductFunctionPU - Runs the user-defined vector-Hessian-vector product function for Fpu.
517: Collective
519: Input Parameters:
520: + ts - The `TS` context obtained from `TSCreate()`
521: . t - the time
522: . U - the solution at which to compute the Hessian product
523: . Vl - the array of input vectors to be multiplied with the Hessian from the left
524: - Vr - the input vector to be multiplied with the Hessian from the right
526: Output Parameter:
527: . VHV - the array of output vectors that store the Hessian product
529: Level: developer
531: Note:
532: `TSComputeIHessianProductFunctionPU()` is typically used for sensitivity implementation,
533: so most users would not generally call this routine themselves.
535: .seealso: [](ch_ts), `TSSetIHessianProduct()`
536: @*/
537: PetscErrorCode TSComputeIHessianProductFunctionPU(TS ts, PetscReal t, Vec U, Vec *Vl, Vec Vr, Vec *VHV)
538: {
539: PetscFunctionBegin;
540: if (!VHV) PetscFunctionReturn(PETSC_SUCCESS);
544: if (ts->ihessianproduct_fpu) PetscCallBack("TS callback IHessianProduct 3 for sensitivity analysis", (*ts->ihessianproduct_fpu)(ts, t, U, Vl, Vr, VHV, ts->ihessianproductctx));
546: /* does not consider IMEX for now, so either IHessian or RHSHessian will be calculated, using the same output VHV */
547: if (ts->rhshessianproduct_gpu) {
548: PetscInt nadj;
549: PetscCall(TSComputeRHSHessianProductFunctionPU(ts, t, U, Vl, Vr, VHV));
550: for (nadj = 0; nadj < ts->numcost; nadj++) PetscCall(VecScale(VHV[nadj], -1));
551: }
552: PetscFunctionReturn(PETSC_SUCCESS);
553: }
555: /*@C
556: TSComputeIHessianProductFunctionPP - Runs the user-defined vector-Hessian-vector product function for Fpp.
558: Collective
560: Input Parameters:
561: + ts - The `TS` context obtained from `TSCreate()`
562: . t - the time
563: . U - the solution at which to compute the Hessian product
564: . Vl - the array of input vectors to be multiplied with the Hessian from the left
565: - Vr - the input vector to be multiplied with the Hessian from the right
567: Output Parameter:
568: . VHV - the array of output vectors that store the Hessian product
570: Level: developer
572: Note:
573: `TSComputeIHessianProductFunctionPP()` is typically used for sensitivity implementation,
574: so most users would not generally call this routine themselves.
576: .seealso: [](ch_ts), `TSSetIHessianProduct()`
577: @*/
578: PetscErrorCode TSComputeIHessianProductFunctionPP(TS ts, PetscReal t, Vec U, Vec *Vl, Vec Vr, Vec *VHV)
579: {
580: PetscFunctionBegin;
581: if (!VHV) PetscFunctionReturn(PETSC_SUCCESS);
585: if (ts->ihessianproduct_fpp) PetscCallBack("TS callback IHessianProduct 3 for sensitivity analysis", (*ts->ihessianproduct_fpp)(ts, t, U, Vl, Vr, VHV, ts->ihessianproductctx));
587: /* does not consider IMEX for now, so either IHessian or RHSHessian will be calculated, using the same output VHV */
588: if (ts->rhshessianproduct_gpp) {
589: PetscInt nadj;
590: PetscCall(TSComputeRHSHessianProductFunctionPP(ts, t, U, Vl, Vr, VHV));
591: for (nadj = 0; nadj < ts->numcost; nadj++) PetscCall(VecScale(VHV[nadj], -1));
592: }
593: PetscFunctionReturn(PETSC_SUCCESS);
594: }
596: // PetscClangLinter pragma disable: -fdoc-param-list-func-parameter-documentation
597: // PetscClangLinter pragma disable: -fdoc-section-header-unknown
598: /*@C
599: TSSetRHSHessianProduct - Sets the function that computes the vector-Hessian-vector
600: product. The Hessian is the second-order derivative of G (RHSFunction) w.r.t. the state
601: variable.
603: Logically Collective
605: Input Parameters:
606: + ts - `TS` context obtained from `TSCreate()`
607: . rhshp1 - an array of vectors storing the result of vector-Hessian-vector product for G_UU
608: . hessianproductfunc1 - vector-Hessian-vector product function for G_UU
609: . rhshp2 - an array of vectors storing the result of vector-Hessian-vector product for G_UP
610: . hessianproductfunc2 - vector-Hessian-vector product function for G_UP
611: . rhshp3 - an array of vectors storing the result of vector-Hessian-vector product for G_PU
612: . hessianproductfunc3 - vector-Hessian-vector product function for G_PU
613: . rhshp4 - an array of vectors storing the result of vector-Hessian-vector product for G_PP
614: . hessianproductfunc4 - vector-Hessian-vector product function for G_PP
615: - ctx - [optional] user-defined function context
617: Calling sequence of `rhshessianproductfunc1`:
618: + ts - the `TS` context
619: . t - current timestep
620: . U - input vector (current ODE solution)
621: . Vl - an array of input vectors to be left-multiplied with the Hessian
622: . Vr - input vector to be right-multiplied with the Hessian
623: . VHV - an array of output vectors for vector-Hessian-vector product
624: - ctx - [optional] user-defined function context
626: Level: intermediate
628: Notes:
629: All other functions have the same calling sequence as `rhhessianproductfunc1`, so their
630: descriptions are omitted for brevity.
632: The first Hessian function and the working array are required.
634: As an example to implement the callback functions, the second callback function calculates the vector-Hessian-vector product
635: $ Vl_n^T*G_UP*Vr
636: where the vector Vl_n (n-th element in the array Vl) and Vr are of size N and M respectively, and the Hessian G_UP is of size N x N x M.
637: Each entry of G_UP corresponds to the derivative
638: $ G_UP[i][j][k] = \frac{\partial^2 G[i]}{\partial U[j] \partial P[k]}.
639: The result of the vector-Hessian-vector product for Vl_n needs to be stored in vector VHV_n with j-th entry being
640: $ VHV_n[j] = \sum_i \sum_k {Vl_n[i] * G_UP[i][j][k] * Vr[k]}
641: If the cost function is a scalar, there will be only one vector in Vl and VHV.
643: .seealso: `TS`, `TSAdjoint`
644: @*/
645: PetscErrorCode TSSetRHSHessianProduct(TS ts, Vec *rhshp1, PetscErrorCode (*rhshessianproductfunc1)(TS ts, PetscReal t, Vec U, Vec *Vl, Vec Vr, Vec *VHV, void *ctx), Vec *rhshp2, PetscErrorCode (*rhshessianproductfunc2)(TS, PetscReal, Vec, Vec *, Vec, Vec *, void *), Vec *rhshp3, PetscErrorCode (*rhshessianproductfunc3)(TS, PetscReal, Vec, Vec *, Vec, Vec *, void *), Vec *rhshp4, PetscErrorCode (*rhshessianproductfunc4)(TS, PetscReal, Vec, Vec *, Vec, Vec *, void *), void *ctx)
646: {
647: PetscFunctionBegin;
649: PetscAssertPointer(rhshp1, 2);
651: ts->rhshessianproductctx = ctx;
652: if (rhshp1) ts->vecs_guu = rhshp1;
653: if (rhshp2) ts->vecs_gup = rhshp2;
654: if (rhshp3) ts->vecs_gpu = rhshp3;
655: if (rhshp4) ts->vecs_gpp = rhshp4;
656: ts->rhshessianproduct_guu = rhshessianproductfunc1;
657: ts->rhshessianproduct_gup = rhshessianproductfunc2;
658: ts->rhshessianproduct_gpu = rhshessianproductfunc3;
659: ts->rhshessianproduct_gpp = rhshessianproductfunc4;
660: PetscFunctionReturn(PETSC_SUCCESS);
661: }
663: /*@C
664: TSComputeRHSHessianProductFunctionUU - Runs the user-defined vector-Hessian-vector product function for Guu.
666: Collective
668: Input Parameters:
669: + ts - The `TS` context obtained from `TSCreate()`
670: . t - the time
671: . U - the solution at which to compute the Hessian product
672: . Vl - the array of input vectors to be multiplied with the Hessian from the left
673: - Vr - the input vector to be multiplied with the Hessian from the right
675: Output Parameter:
676: . VHV - the array of output vectors that store the Hessian product
678: Level: developer
680: Note:
681: `TSComputeRHSHessianProductFunctionUU()` is typically used for sensitivity implementation,
682: so most users would not generally call this routine themselves.
684: .seealso: [](ch_ts), `TS`, `TSSetRHSHessianProduct()`
685: @*/
686: PetscErrorCode TSComputeRHSHessianProductFunctionUU(TS ts, PetscReal t, Vec U, Vec *Vl, Vec Vr, Vec *VHV)
687: {
688: PetscFunctionBegin;
689: if (!VHV) PetscFunctionReturn(PETSC_SUCCESS);
693: PetscCallBack("TS callback RHSHessianProduct 1 for sensitivity analysis", (*ts->rhshessianproduct_guu)(ts, t, U, Vl, Vr, VHV, ts->rhshessianproductctx));
694: PetscFunctionReturn(PETSC_SUCCESS);
695: }
697: /*@C
698: TSComputeRHSHessianProductFunctionUP - Runs the user-defined vector-Hessian-vector product function for Gup.
700: Collective
702: Input Parameters:
703: + ts - The `TS` context obtained from `TSCreate()`
704: . t - the time
705: . U - the solution at which to compute the Hessian product
706: . Vl - the array of input vectors to be multiplied with the Hessian from the left
707: - Vr - the input vector to be multiplied with the Hessian from the right
709: Output Parameter:
710: . VHV - the array of output vectors that store the Hessian product
712: Level: developer
714: Note:
715: `TSComputeRHSHessianProductFunctionUP()` is typically used for sensitivity implementation,
716: so most users would not generally call this routine themselves.
718: .seealso: [](ch_ts), `TS`, `TSSetRHSHessianProduct()`
719: @*/
720: PetscErrorCode TSComputeRHSHessianProductFunctionUP(TS ts, PetscReal t, Vec U, Vec *Vl, Vec Vr, Vec *VHV)
721: {
722: PetscFunctionBegin;
723: if (!VHV) PetscFunctionReturn(PETSC_SUCCESS);
727: PetscCallBack("TS callback RHSHessianProduct 2 for sensitivity analysis", (*ts->rhshessianproduct_gup)(ts, t, U, Vl, Vr, VHV, ts->rhshessianproductctx));
728: PetscFunctionReturn(PETSC_SUCCESS);
729: }
731: /*@C
732: TSComputeRHSHessianProductFunctionPU - Runs the user-defined vector-Hessian-vector product function for Gpu.
734: Collective
736: Input Parameters:
737: + ts - The `TS` context obtained from `TSCreate()`
738: . t - the time
739: . U - the solution at which to compute the Hessian product
740: . Vl - the array of input vectors to be multiplied with the Hessian from the left
741: - Vr - the input vector to be multiplied with the Hessian from the right
743: Output Parameter:
744: . VHV - the array of output vectors that store the Hessian product
746: Level: developer
748: Note:
749: `TSComputeRHSHessianProductFunctionPU()` is typically used for sensitivity implementation,
750: so most users would not generally call this routine themselves.
752: .seealso: [](ch_ts), `TSSetRHSHessianProduct()`
753: @*/
754: PetscErrorCode TSComputeRHSHessianProductFunctionPU(TS ts, PetscReal t, Vec U, Vec *Vl, Vec Vr, Vec *VHV)
755: {
756: PetscFunctionBegin;
757: if (!VHV) PetscFunctionReturn(PETSC_SUCCESS);
761: PetscCallBack("TS callback RHSHessianProduct 3 for sensitivity analysis", (*ts->rhshessianproduct_gpu)(ts, t, U, Vl, Vr, VHV, ts->rhshessianproductctx));
762: PetscFunctionReturn(PETSC_SUCCESS);
763: }
765: /*@C
766: TSComputeRHSHessianProductFunctionPP - Runs the user-defined vector-Hessian-vector product function for Gpp.
768: Collective
770: Input Parameters:
771: + ts - The `TS` context obtained from `TSCreate()`
772: . t - the time
773: . U - the solution at which to compute the Hessian product
774: . Vl - the array of input vectors to be multiplied with the Hessian from the left
775: - Vr - the input vector to be multiplied with the Hessian from the right
777: Output Parameter:
778: . VHV - the array of output vectors that store the Hessian product
780: Level: developer
782: Note:
783: `TSComputeRHSHessianProductFunctionPP()` is typically used for sensitivity implementation,
784: so most users would not generally call this routine themselves.
786: .seealso: [](ch_ts), `TSSetRHSHessianProduct()`
787: @*/
788: PetscErrorCode TSComputeRHSHessianProductFunctionPP(TS ts, PetscReal t, Vec U, Vec *Vl, Vec Vr, Vec *VHV)
789: {
790: PetscFunctionBegin;
791: if (!VHV) PetscFunctionReturn(PETSC_SUCCESS);
795: PetscCallBack("TS callback RHSHessianProduct 3 for sensitivity analysis", (*ts->rhshessianproduct_gpp)(ts, t, U, Vl, Vr, VHV, ts->rhshessianproductctx));
796: PetscFunctionReturn(PETSC_SUCCESS);
797: }
799: /* --------------------------- Adjoint sensitivity ---------------------------*/
801: /*@
802: TSSetCostGradients - Sets the initial value of the gradients of the cost function w.r.t. initial values and w.r.t. the problem parameters
803: for use by the `TS` adjoint routines.
805: Logically Collective
807: Input Parameters:
808: + ts - the `TS` context obtained from `TSCreate()`
809: . numcost - number of gradients to be computed, this is the number of cost functions
810: . lambda - gradients with respect to the initial condition variables, the dimension and parallel layout of these vectors is the same as the ODE solution vector
811: - mu - gradients with respect to the parameters, the number of entries in these vectors is the same as the number of parameters
813: Level: beginner
815: Notes:
816: the entries in these vectors must be correctly initialized with the values lambda_i = df/dy|finaltime mu_i = df/dp|finaltime
818: After `TSAdjointSolve()` is called the lambda and the mu contain the computed sensitivities
820: .seealso: `TS`, `TSAdjointSolve()`, `TSGetCostGradients()`
821: @*/
822: PetscErrorCode TSSetCostGradients(TS ts, PetscInt numcost, Vec *lambda, Vec *mu)
823: {
824: PetscFunctionBegin;
826: PetscAssertPointer(lambda, 3);
827: ts->vecs_sensi = lambda;
828: ts->vecs_sensip = mu;
829: PetscCheck(!ts->numcost || ts->numcost == numcost, PetscObjectComm((PetscObject)ts), PETSC_ERR_USER, "The number of cost functions (2nd parameter of TSSetCostIntegrand()) is inconsistent with the one set by TSSetCostIntegrand");
830: ts->numcost = numcost;
831: PetscFunctionReturn(PETSC_SUCCESS);
832: }
834: /*@
835: TSGetCostGradients - Returns the gradients from the `TSAdjointSolve()`
837: Not Collective, but the vectors returned are parallel if `TS` is parallel
839: Input Parameter:
840: . ts - the `TS` context obtained from `TSCreate()`
842: Output Parameters:
843: + numcost - size of returned arrays
844: . lambda - vectors containing the gradients of the cost functions with respect to the ODE/DAE solution variables
845: - mu - vectors containing the gradients of the cost functions with respect to the problem parameters
847: Level: intermediate
849: .seealso: [](ch_ts), `TS`, `TSAdjointSolve()`, `TSSetCostGradients()`
850: @*/
851: PetscErrorCode TSGetCostGradients(TS ts, PetscInt *numcost, Vec **lambda, Vec **mu)
852: {
853: PetscFunctionBegin;
855: if (numcost) *numcost = ts->numcost;
856: if (lambda) *lambda = ts->vecs_sensi;
857: if (mu) *mu = ts->vecs_sensip;
858: PetscFunctionReturn(PETSC_SUCCESS);
859: }
861: /*@
862: TSSetCostHessianProducts - Sets the initial value of the Hessian-vector products of the cost function w.r.t. initial values and w.r.t. the problem parameters
863: for use by the `TS` adjoint routines.
865: Logically Collective
867: Input Parameters:
868: + ts - the `TS` context obtained from `TSCreate()`
869: . numcost - number of cost functions
870: . lambda2 - Hessian-vector product with respect to the initial condition variables, the dimension and parallel layout of these vectors is the same as the ODE solution vector
871: . mu2 - Hessian-vector product with respect to the parameters, the number of entries in these vectors is the same as the number of parameters
872: - dir - the direction vector that are multiplied with the Hessian of the cost functions
874: Level: beginner
876: Notes:
877: Hessian of the cost function is completely different from Hessian of the ODE/DAE system
879: For second-order adjoint, one needs to call this function and then `TSAdjointSetForward()` before `TSSolve()`.
881: After `TSAdjointSolve()` is called, the lambda2 and the mu2 will contain the computed second-order adjoint sensitivities, and can be used to produce Hessian-vector product (not the full Hessian matrix). Users must provide a direction vector; it is usually generated by an optimization solver.
883: Passing `NULL` for `lambda2` disables the second-order calculation.
885: .seealso: [](ch_ts), `TS`, `TSAdjointSolve()`, `TSAdjointSetForward()`
886: @*/
887: PetscErrorCode TSSetCostHessianProducts(TS ts, PetscInt numcost, Vec *lambda2, Vec *mu2, Vec dir)
888: {
889: PetscFunctionBegin;
891: PetscCheck(!ts->numcost || ts->numcost == numcost, PetscObjectComm((PetscObject)ts), PETSC_ERR_USER, "The number of cost functions (2nd parameter of TSSetCostIntegrand()) is inconsistent with the one set by TSSetCostIntegrand");
892: ts->numcost = numcost;
893: ts->vecs_sensi2 = lambda2;
894: ts->vecs_sensi2p = mu2;
895: ts->vec_dir = dir;
896: PetscFunctionReturn(PETSC_SUCCESS);
897: }
899: /*@
900: TSGetCostHessianProducts - Returns the gradients from the `TSAdjointSolve()`
902: Not Collective, but vectors returned are parallel if `TS` is parallel
904: Input Parameter:
905: . ts - the `TS` context obtained from `TSCreate()`
907: Output Parameters:
908: + numcost - number of cost functions
909: . lambda2 - Hessian-vector product with respect to the initial condition variables, the dimension and parallel layout of these vectors is the same as the ODE solution vector
910: . mu2 - Hessian-vector product with respect to the parameters, the number of entries in these vectors is the same as the number of parameters
911: - dir - the direction vector that are multiplied with the Hessian of the cost functions
913: Level: intermediate
915: .seealso: [](ch_ts), `TSAdjointSolve()`, `TSSetCostHessianProducts()`
916: @*/
917: PetscErrorCode TSGetCostHessianProducts(TS ts, PetscInt *numcost, Vec **lambda2, Vec **mu2, Vec *dir)
918: {
919: PetscFunctionBegin;
921: if (numcost) *numcost = ts->numcost;
922: if (lambda2) *lambda2 = ts->vecs_sensi2;
923: if (mu2) *mu2 = ts->vecs_sensi2p;
924: if (dir) *dir = ts->vec_dir;
925: PetscFunctionReturn(PETSC_SUCCESS);
926: }
928: /*@
929: TSAdjointSetForward - Trigger the tangent linear solver and initialize the forward sensitivities
931: Logically Collective
933: Input Parameters:
934: + ts - the `TS` context obtained from `TSCreate()`
935: - didp - the derivative of initial values w.r.t. parameters
937: Level: intermediate
939: Notes:
940: When computing sensitivities w.r.t. initial condition, set didp to `NULL` so that the solver will take it as an identity matrix mathematically.
941: `TSAdjoint` does not reset the tangent linear solver automatically, `TSAdjointResetForward()` should be called to reset the tangent linear solver.
943: .seealso: [](ch_ts), `TSAdjointSolve()`, `TSSetCostHessianProducts()`, `TSAdjointResetForward()`
944: @*/
945: PetscErrorCode TSAdjointSetForward(TS ts, Mat didp)
946: {
947: Mat A;
948: Vec sp;
949: PetscScalar *xarr;
950: PetscInt lsize;
952: PetscFunctionBegin;
953: ts->forward_solve = PETSC_TRUE; /* turn on tangent linear mode */
954: PetscCheck(ts->vecs_sensi2, PetscObjectComm((PetscObject)ts), PETSC_ERR_USER, "Must call TSSetCostHessianProducts() first");
955: PetscCheck(ts->vec_dir, PetscObjectComm((PetscObject)ts), PETSC_ERR_USER, "Directional vector is missing. Call TSSetCostHessianProducts() to set it.");
956: /* create a single-column dense matrix */
957: PetscCall(VecGetLocalSize(ts->vec_sol, &lsize));
958: PetscCall(MatCreateDense(PetscObjectComm((PetscObject)ts), lsize, PETSC_DECIDE, PETSC_DECIDE, 1, NULL, &A));
960: PetscCall(VecDuplicate(ts->vec_sol, &sp));
961: PetscCall(MatDenseGetColumn(A, 0, &xarr));
962: PetscCall(VecPlaceArray(sp, xarr));
963: if (ts->vecs_sensi2p) { /* tangent linear variable initialized as 2*dIdP*dir */
964: if (didp) {
965: PetscCall(MatMult(didp, ts->vec_dir, sp));
966: PetscCall(VecScale(sp, 2.));
967: } else {
968: PetscCall(VecZeroEntries(sp));
969: }
970: } else { /* tangent linear variable initialized as dir */
971: PetscCall(VecCopy(ts->vec_dir, sp));
972: }
973: PetscCall(VecResetArray(sp));
974: PetscCall(MatDenseRestoreColumn(A, &xarr));
975: PetscCall(VecDestroy(&sp));
977: PetscCall(TSForwardSetInitialSensitivities(ts, A)); /* if didp is NULL, identity matrix is assumed */
979: PetscCall(MatDestroy(&A));
980: PetscFunctionReturn(PETSC_SUCCESS);
981: }
983: /*@
984: TSAdjointResetForward - Reset the tangent linear solver and destroy the tangent linear context
986: Logically Collective
988: Input Parameter:
989: . ts - the `TS` context obtained from `TSCreate()`
991: Level: intermediate
993: .seealso: [](ch_ts), `TSAdjointSetForward()`
994: @*/
995: PetscErrorCode TSAdjointResetForward(TS ts)
996: {
997: PetscFunctionBegin;
998: ts->forward_solve = PETSC_FALSE; /* turn off tangent linear mode */
999: PetscCall(TSForwardReset(ts));
1000: PetscFunctionReturn(PETSC_SUCCESS);
1001: }
1003: /*@
1004: TSAdjointSetUp - Sets up the internal data structures for the later use
1005: of an adjoint solver
1007: Collective
1009: Input Parameter:
1010: . ts - the `TS` context obtained from `TSCreate()`
1012: Level: advanced
1014: .seealso: [](ch_ts), `TSCreate()`, `TSAdjointStep()`, `TSSetCostGradients()`
1015: @*/
1016: PetscErrorCode TSAdjointSetUp(TS ts)
1017: {
1018: TSTrajectory tj;
1019: PetscBool match;
1021: PetscFunctionBegin;
1023: if (ts->adjointsetupcalled) PetscFunctionReturn(PETSC_SUCCESS);
1024: PetscCheck(ts->vecs_sensi, PetscObjectComm((PetscObject)ts), PETSC_ERR_ARG_WRONGSTATE, "Must call TSSetCostGradients() first");
1025: PetscCheck(!ts->vecs_sensip || ts->Jacp || ts->Jacprhs, PetscObjectComm((PetscObject)ts), PETSC_ERR_ARG_WRONGSTATE, "Must call TSSetRHSJacobianP() or TSSetIJacobianP() first");
1026: PetscCall(TSGetTrajectory(ts, &tj));
1027: PetscCall(PetscObjectTypeCompare((PetscObject)tj, TSTRAJECTORYBASIC, &match));
1028: if (match) {
1029: PetscBool solution_only;
1030: PetscCall(TSTrajectoryGetSolutionOnly(tj, &solution_only));
1031: PetscCheck(!solution_only, PetscObjectComm((PetscObject)ts), PETSC_ERR_USER, "TSAdjoint cannot use the solution-only mode when choosing the Basic TSTrajectory type. Turn it off with -ts_trajectory_solution_only 0");
1032: }
1033: PetscCall(TSTrajectorySetUseHistory(tj, PETSC_FALSE)); /* not use TSHistory */
1035: if (ts->quadraturets) { /* if there is integral in the cost function */
1036: PetscCall(VecDuplicate(ts->vecs_sensi[0], &ts->vec_drdu_col));
1037: if (ts->vecs_sensip) PetscCall(VecDuplicate(ts->vecs_sensip[0], &ts->vec_drdp_col));
1038: }
1040: PetscTryTypeMethod(ts, adjointsetup);
1041: ts->adjointsetupcalled = PETSC_TRUE;
1042: PetscFunctionReturn(PETSC_SUCCESS);
1043: }
1045: /*@
1046: TSAdjointReset - Resets a `TS` adjoint context and removes any allocated `Vec`s and `Mat`s.
1048: Collective
1050: Input Parameter:
1051: . ts - the `TS` context obtained from `TSCreate()`
1053: Level: beginner
1055: .seealso: [](ch_ts), `TSCreate()`, `TSAdjointSetUp()`, `TSADestroy()`
1056: @*/
1057: PetscErrorCode TSAdjointReset(TS ts)
1058: {
1059: PetscFunctionBegin;
1061: PetscTryTypeMethod(ts, adjointreset);
1062: if (ts->quadraturets) { /* if there is integral in the cost function */
1063: PetscCall(VecDestroy(&ts->vec_drdu_col));
1064: if (ts->vecs_sensip) PetscCall(VecDestroy(&ts->vec_drdp_col));
1065: }
1066: ts->vecs_sensi = NULL;
1067: ts->vecs_sensip = NULL;
1068: ts->vecs_sensi2 = NULL;
1069: ts->vecs_sensi2p = NULL;
1070: ts->vec_dir = NULL;
1071: ts->adjointsetupcalled = PETSC_FALSE;
1072: PetscFunctionReturn(PETSC_SUCCESS);
1073: }
1075: /*@
1076: TSAdjointSetSteps - Sets the number of steps the adjoint solver should take backward in time
1078: Logically Collective
1080: Input Parameters:
1081: + ts - the `TS` context obtained from `TSCreate()`
1082: - steps - number of steps to use
1084: Level: intermediate
1086: Notes:
1087: Normally one does not call this and `TSAdjointSolve()` integrates back to the original timestep. One can call this
1088: so as to integrate back to less than the original timestep
1090: .seealso: [](ch_ts), `TSAdjointSolve()`, `TS`, `TSSetExactFinalTime()`
1091: @*/
1092: PetscErrorCode TSAdjointSetSteps(TS ts, PetscInt steps)
1093: {
1094: PetscFunctionBegin;
1097: PetscCheck(steps >= 0, PetscObjectComm((PetscObject)ts), PETSC_ERR_ARG_OUTOFRANGE, "Cannot step back a negative number of steps");
1098: PetscCheck(steps <= ts->steps, PetscObjectComm((PetscObject)ts), PETSC_ERR_ARG_OUTOFRANGE, "Cannot step back more than the total number of forward steps");
1099: ts->adjoint_max_steps = steps;
1100: PetscFunctionReturn(PETSC_SUCCESS);
1101: }
1103: // PetscClangLinter pragma disable: -fdoc-*
1104: /*@C
1105: TSAdjointSetRHSJacobian - Deprecated, use `TSSetRHSJacobianP()`
1107: Level: deprecated
1108: @*/
1109: PetscErrorCode TSAdjointSetRHSJacobian(TS ts, Mat Amat, PetscErrorCode (*func)(TS, PetscReal, Vec, Mat, void *), void *ctx)
1110: {
1111: PetscFunctionBegin;
1115: ts->rhsjacobianp = func;
1116: ts->rhsjacobianpctx = ctx;
1117: if (Amat) {
1118: PetscCall(PetscObjectReference((PetscObject)Amat));
1119: PetscCall(MatDestroy(&ts->Jacp));
1120: ts->Jacp = Amat;
1121: }
1122: PetscFunctionReturn(PETSC_SUCCESS);
1123: }
1125: // PetscClangLinter pragma disable: -fdoc-*
1126: /*@C
1127: TSAdjointComputeRHSJacobian - Deprecated, use `TSComputeRHSJacobianP()`
1129: Level: deprecated
1130: @*/
1131: PetscErrorCode TSAdjointComputeRHSJacobian(TS ts, PetscReal t, Vec U, Mat Amat)
1132: {
1133: PetscFunctionBegin;
1138: PetscCallBack("TS callback JacobianP for sensitivity analysis", (*ts->rhsjacobianp)(ts, t, U, Amat, ts->rhsjacobianpctx));
1139: PetscFunctionReturn(PETSC_SUCCESS);
1140: }
1142: // PetscClangLinter pragma disable: -fdoc-*
1143: /*@
1144: TSAdjointComputeDRDYFunction - Deprecated, use `TSGetQuadratureTS()` then `TSComputeRHSJacobian()`
1146: Level: deprecated
1147: @*/
1148: PetscErrorCode TSAdjointComputeDRDYFunction(TS ts, PetscReal t, Vec U, Vec *DRDU)
1149: {
1150: PetscFunctionBegin;
1154: PetscCallBack("TS callback DRDY for sensitivity analysis", (*ts->drdufunction)(ts, t, U, DRDU, ts->costintegrandctx));
1155: PetscFunctionReturn(PETSC_SUCCESS);
1156: }
1158: // PetscClangLinter pragma disable: -fdoc-*
1159: /*@
1160: TSAdjointComputeDRDPFunction - Deprecated, use `TSGetQuadratureTS()` then `TSComputeRHSJacobianP()`
1162: Level: deprecated
1163: @*/
1164: PetscErrorCode TSAdjointComputeDRDPFunction(TS ts, PetscReal t, Vec U, Vec *DRDP)
1165: {
1166: PetscFunctionBegin;
1170: PetscCallBack("TS callback DRDP for sensitivity analysis", (*ts->drdpfunction)(ts, t, U, DRDP, ts->costintegrandctx));
1171: PetscFunctionReturn(PETSC_SUCCESS);
1172: }
1174: // PetscClangLinter pragma disable: -fdoc-param-list-func-parameter-documentation
1175: /*@C
1176: TSAdjointMonitorSensi - monitors the first lambda sensitivity
1178: Level: intermediate
1180: .seealso: [](ch_ts), `TSAdjointMonitorSet()`
1181: @*/
1182: static PetscErrorCode TSAdjointMonitorSensi(TS ts, PetscInt step, PetscReal ptime, Vec v, PetscInt numcost, Vec *lambda, Vec *mu, PetscViewerAndFormat *vf)
1183: {
1184: PetscViewer viewer = vf->viewer;
1186: PetscFunctionBegin;
1188: PetscCall(PetscViewerPushFormat(viewer, vf->format));
1189: PetscCall(VecView(lambda[0], viewer));
1190: PetscCall(PetscViewerPopFormat(viewer));
1191: PetscFunctionReturn(PETSC_SUCCESS);
1192: }
1194: /*@C
1195: TSAdjointMonitorSetFromOptions - Sets a monitor function and viewer appropriate for the type indicated by the user
1197: Collective
1199: Input Parameters:
1200: + ts - `TS` object you wish to monitor
1201: . name - the monitor type one is seeking
1202: . help - message indicating what monitoring is done
1203: . manual - manual page for the monitor
1204: . monitor - the monitor function
1205: - monitorsetup - a function that is called once ONLY if the user selected this monitor that may set additional features of the `TS` or `PetscViewer` objects
1207: Level: developer
1209: .seealso: [](ch_ts), `PetscOptionsGetViewer()`, `PetscOptionsGetReal()`, `PetscOptionsHasName()`, `PetscOptionsGetString()`,
1210: `PetscOptionsGetIntArray()`, `PetscOptionsGetRealArray()`, `PetscOptionsBool()`
1211: `PetscOptionsInt()`, `PetscOptionsString()`, `PetscOptionsReal()`,
1212: `PetscOptionsName()`, `PetscOptionsBegin()`, `PetscOptionsEnd()`, `PetscOptionsHeadBegin()`,
1213: `PetscOptionsStringArray()`, `PetscOptionsRealArray()`, `PetscOptionsScalar()`,
1214: `PetscOptionsBoolGroupBegin()`, `PetscOptionsBoolGroup()`, `PetscOptionsBoolGroupEnd()`,
1215: `PetscOptionsFList()`, `PetscOptionsEList()`
1216: @*/
1217: PetscErrorCode TSAdjointMonitorSetFromOptions(TS ts, const char name[], const char help[], const char manual[], PetscErrorCode (*monitor)(TS, PetscInt, PetscReal, Vec, PetscInt, Vec *, Vec *, PetscViewerAndFormat *), PetscErrorCode (*monitorsetup)(TS, PetscViewerAndFormat *))
1218: {
1219: PetscViewer viewer;
1220: PetscViewerFormat format;
1221: PetscBool flg;
1223: PetscFunctionBegin;
1224: PetscCall(PetscOptionsGetViewer(PetscObjectComm((PetscObject)ts), ((PetscObject)ts)->options, ((PetscObject)ts)->prefix, name, &viewer, &format, &flg));
1225: if (flg) {
1226: PetscViewerAndFormat *vf;
1227: PetscCall(PetscViewerAndFormatCreate(viewer, format, &vf));
1228: PetscCall(PetscObjectDereference((PetscObject)viewer));
1229: if (monitorsetup) PetscCall((*monitorsetup)(ts, vf));
1230: PetscCall(TSAdjointMonitorSet(ts, (PetscErrorCode(*)(TS, PetscInt, PetscReal, Vec, PetscInt, Vec *, Vec *, void *))monitor, vf, (PetscErrorCode(*)(void **))PetscViewerAndFormatDestroy));
1231: }
1232: PetscFunctionReturn(PETSC_SUCCESS);
1233: }
1235: /*@C
1236: TSAdjointMonitorSet - Sets an ADDITIONAL function that is to be used at every
1237: timestep to display the iteration's progress.
1239: Logically Collective
1241: Input Parameters:
1242: + ts - the `TS` context obtained from `TSCreate()`
1243: . adjointmonitor - monitoring routine
1244: . adjointmctx - [optional] user-defined context for private data for the monitor routine
1245: (use `NULL` if no context is desired)
1246: - adjointmdestroy - [optional] routine that frees monitor context (may be `NULL`)
1248: Calling sequence of `adjointmonitor`:
1249: + ts - the `TS` context
1250: . steps - iteration number (after the final time step the monitor routine is called with
1251: a step of -1, this is at the final time which may have been interpolated to)
1252: . time - current time
1253: . u - current iterate
1254: . numcost - number of cost functionos
1255: . lambda - sensitivities to initial conditions
1256: . mu - sensitivities to parameters
1257: - adjointmctx - [optional] adjoint monitoring context
1259: Calling sequence of `adjointmdestroy`:
1260: . mctx - the monitor context to destroy
1262: Level: intermediate
1264: Note:
1265: This routine adds an additional monitor to the list of monitors that
1266: already has been loaded.
1268: Fortran Notes:
1269: Only a single monitor function can be set for each `TS` object
1271: .seealso: [](ch_ts), `TS`, `TSAdjointSolve()`, `TSAdjointMonitorCancel()`
1272: @*/
1273: PetscErrorCode TSAdjointMonitorSet(TS ts, PetscErrorCode (*adjointmonitor)(TS ts, PetscInt steps, PetscReal time, Vec u, PetscInt numcost, Vec *lambda, Vec *mu, void *adjointmctx), void *adjointmctx, PetscErrorCode (*adjointmdestroy)(void **mctx))
1274: {
1275: PetscInt i;
1276: PetscBool identical;
1278: PetscFunctionBegin;
1280: for (i = 0; i < ts->numbermonitors; i++) {
1281: PetscCall(PetscMonitorCompare((PetscErrorCode(*)(void))adjointmonitor, adjointmctx, adjointmdestroy, (PetscErrorCode(*)(void))ts->adjointmonitor[i], ts->adjointmonitorcontext[i], ts->adjointmonitordestroy[i], &identical));
1282: if (identical) PetscFunctionReturn(PETSC_SUCCESS);
1283: }
1284: PetscCheck(ts->numberadjointmonitors < MAXTSMONITORS, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Too many adjoint monitors set");
1285: ts->adjointmonitor[ts->numberadjointmonitors] = adjointmonitor;
1286: ts->adjointmonitordestroy[ts->numberadjointmonitors] = adjointmdestroy;
1287: ts->adjointmonitorcontext[ts->numberadjointmonitors++] = (void *)adjointmctx;
1288: PetscFunctionReturn(PETSC_SUCCESS);
1289: }
1291: /*@C
1292: TSAdjointMonitorCancel - Clears all the adjoint monitors that have been set on a time-step object.
1294: Logically Collective
1296: Input Parameter:
1297: . ts - the `TS` context obtained from `TSCreate()`
1299: Notes:
1300: There is no way to remove a single, specific monitor.
1302: Level: intermediate
1304: .seealso: [](ch_ts), `TS`, `TSAdjointSolve()`, `TSAdjointMonitorSet()`
1305: @*/
1306: PetscErrorCode TSAdjointMonitorCancel(TS ts)
1307: {
1308: PetscInt i;
1310: PetscFunctionBegin;
1312: for (i = 0; i < ts->numberadjointmonitors; i++) {
1313: if (ts->adjointmonitordestroy[i]) PetscCall((*ts->adjointmonitordestroy[i])(&ts->adjointmonitorcontext[i]));
1314: }
1315: ts->numberadjointmonitors = 0;
1316: PetscFunctionReturn(PETSC_SUCCESS);
1317: }
1319: /*@C
1320: TSAdjointMonitorDefault - the default monitor of adjoint computations
1322: Input Parameters:
1323: + ts - the `TS` context
1324: . step - iteration number (after the final time step the monitor routine is called with a
1325: step of -1, this is at the final time which may have been interpolated to)
1326: . time - current time
1327: . v - current iterate
1328: . numcost - number of cost functionos
1329: . lambda - sensitivities to initial conditions
1330: . mu - sensitivities to parameters
1331: - vf - the viewer and format
1333: Level: intermediate
1335: .seealso: [](ch_ts), `TS`, `TSAdjointSolve()`, `TSAdjointMonitorSet()`
1336: @*/
1337: PetscErrorCode TSAdjointMonitorDefault(TS ts, PetscInt step, PetscReal time, Vec v, PetscInt numcost, Vec *lambda, Vec *mu, PetscViewerAndFormat *vf)
1338: {
1339: PetscViewer viewer = vf->viewer;
1341: PetscFunctionBegin;
1342: (void)v;
1343: (void)numcost;
1344: (void)lambda;
1345: (void)mu;
1347: PetscCall(PetscViewerPushFormat(viewer, vf->format));
1348: PetscCall(PetscViewerASCIIAddTab(viewer, ((PetscObject)ts)->tablevel));
1349: PetscCall(PetscViewerASCIIPrintf(viewer, "%" PetscInt_FMT " TS dt %g time %g%s", step, (double)ts->time_step, (double)time, ts->steprollback ? " (r)\n" : "\n"));
1350: PetscCall(PetscViewerASCIISubtractTab(viewer, ((PetscObject)ts)->tablevel));
1351: PetscCall(PetscViewerPopFormat(viewer));
1352: PetscFunctionReturn(PETSC_SUCCESS);
1353: }
1355: /*@C
1356: TSAdjointMonitorDrawSensi - Monitors progress of the adjoint `TS` solvers by calling
1357: `VecView()` for the sensitivities to initial states at each timestep
1359: Collective
1361: Input Parameters:
1362: + ts - the `TS` context
1363: . step - current time-step
1364: . ptime - current time
1365: . u - current state
1366: . numcost - number of cost functions
1367: . lambda - sensitivities to initial conditions
1368: . mu - sensitivities to parameters
1369: - dummy - either a viewer or `NULL`
1371: Level: intermediate
1373: .seealso: [](ch_ts), `TSAdjointSolve()`, `TSAdjointMonitorSet()`, `TSAdjointMonitorDefault()`, `VecView()`
1374: @*/
1375: PetscErrorCode TSAdjointMonitorDrawSensi(TS ts, PetscInt step, PetscReal ptime, Vec u, PetscInt numcost, Vec *lambda, Vec *mu, void *dummy)
1376: {
1377: TSMonitorDrawCtx ictx = (TSMonitorDrawCtx)dummy;
1378: PetscDraw draw;
1379: PetscReal xl, yl, xr, yr, h;
1380: char time[32];
1382: PetscFunctionBegin;
1383: if (!(((ictx->howoften > 0) && (!(step % ictx->howoften))) || ((ictx->howoften == -1) && ts->reason))) PetscFunctionReturn(PETSC_SUCCESS);
1385: PetscCall(VecView(lambda[0], ictx->viewer));
1386: PetscCall(PetscViewerDrawGetDraw(ictx->viewer, 0, &draw));
1387: PetscCall(PetscSNPrintf(time, 32, "Timestep %d Time %g", (int)step, (double)ptime));
1388: PetscCall(PetscDrawGetCoordinates(draw, &xl, &yl, &xr, &yr));
1389: h = yl + .95 * (yr - yl);
1390: PetscCall(PetscDrawStringCentered(draw, .5 * (xl + xr), h, PETSC_DRAW_BLACK, time));
1391: PetscCall(PetscDrawFlush(draw));
1392: PetscFunctionReturn(PETSC_SUCCESS);
1393: }
1395: /*@C
1396: TSAdjointSetFromOptions - Sets various `TS` adjoint parameters from options database.
1398: Collective
1400: Input Parameters:
1401: + ts - the `TS` context
1402: - PetscOptionsObject - the options context
1404: Options Database Keys:
1405: + -ts_adjoint_solve <yes,no> - After solving the ODE/DAE solve the adjoint problem (requires `-ts_save_trajectory`)
1406: . -ts_adjoint_monitor - print information at each adjoint time step
1407: - -ts_adjoint_monitor_draw_sensi - monitor the sensitivity of the first cost function wrt initial conditions (lambda[0]) graphically
1409: Level: developer
1411: Note:
1412: This is not normally called directly by users
1414: .seealso: [](ch_ts), `TSSetSaveTrajectory()`, `TSTrajectorySetUp()`
1415: @*/
1416: PetscErrorCode TSAdjointSetFromOptions(TS ts, PetscOptionItems *PetscOptionsObject)
1417: {
1418: PetscBool tflg, opt;
1420: PetscFunctionBegin;
1422: PetscOptionsHeadBegin(PetscOptionsObject, "TS Adjoint options");
1423: tflg = ts->adjoint_solve ? PETSC_TRUE : PETSC_FALSE;
1424: PetscCall(PetscOptionsBool("-ts_adjoint_solve", "Solve the adjoint problem immediately after solving the forward problem", "", tflg, &tflg, &opt));
1425: if (opt) {
1426: PetscCall(TSSetSaveTrajectory(ts));
1427: ts->adjoint_solve = tflg;
1428: }
1429: PetscCall(TSAdjointMonitorSetFromOptions(ts, "-ts_adjoint_monitor", "Monitor adjoint timestep size", "TSAdjointMonitorDefault", TSAdjointMonitorDefault, NULL));
1430: PetscCall(TSAdjointMonitorSetFromOptions(ts, "-ts_adjoint_monitor_sensi", "Monitor sensitivity in the adjoint computation", "TSAdjointMonitorSensi", TSAdjointMonitorSensi, NULL));
1431: opt = PETSC_FALSE;
1432: PetscCall(PetscOptionsName("-ts_adjoint_monitor_draw_sensi", "Monitor adjoint sensitivities (lambda only) graphically", "TSAdjointMonitorDrawSensi", &opt));
1433: if (opt) {
1434: TSMonitorDrawCtx ctx;
1435: PetscInt howoften = 1;
1437: PetscCall(PetscOptionsInt("-ts_adjoint_monitor_draw_sensi", "Monitor adjoint sensitivities (lambda only) graphically", "TSAdjointMonitorDrawSensi", howoften, &howoften, NULL));
1438: PetscCall(TSMonitorDrawCtxCreate(PetscObjectComm((PetscObject)ts), NULL, NULL, PETSC_DECIDE, PETSC_DECIDE, 300, 300, howoften, &ctx));
1439: PetscCall(TSAdjointMonitorSet(ts, TSAdjointMonitorDrawSensi, ctx, (PetscErrorCode(*)(void **))TSMonitorDrawCtxDestroy));
1440: }
1441: PetscFunctionReturn(PETSC_SUCCESS);
1442: }
1444: /*@
1445: TSAdjointStep - Steps one time step backward in the adjoint run
1447: Collective
1449: Input Parameter:
1450: . ts - the `TS` context obtained from `TSCreate()`
1452: Level: intermediate
1454: .seealso: [](ch_ts), `TSAdjointSetUp()`, `TSAdjointSolve()`
1455: @*/
1456: PetscErrorCode TSAdjointStep(TS ts)
1457: {
1458: DM dm;
1460: PetscFunctionBegin;
1462: PetscCall(TSGetDM(ts, &dm));
1463: PetscCall(TSAdjointSetUp(ts));
1464: ts->steps--; /* must decrease the step index before the adjoint step is taken. */
1466: ts->reason = TS_CONVERGED_ITERATING;
1467: ts->ptime_prev = ts->ptime;
1468: PetscCall(PetscLogEventBegin(TS_AdjointStep, ts, 0, 0, 0));
1469: PetscUseTypeMethod(ts, adjointstep);
1470: PetscCall(PetscLogEventEnd(TS_AdjointStep, ts, 0, 0, 0));
1471: ts->adjoint_steps++;
1473: if (ts->reason < 0) {
1474: PetscCheck(!ts->errorifstepfailed, PetscObjectComm((PetscObject)ts), PETSC_ERR_NOT_CONVERGED, "TSAdjointStep has failed due to %s", TSConvergedReasons[ts->reason]);
1475: } else if (!ts->reason) {
1476: if (ts->adjoint_steps >= ts->adjoint_max_steps) ts->reason = TS_CONVERGED_ITS;
1477: }
1478: PetscFunctionReturn(PETSC_SUCCESS);
1479: }
1481: /*@
1482: TSAdjointSolve - Solves the discrete ajoint problem for an ODE/DAE
1484: Collective
1485: `
1487: Input Parameter:
1488: . ts - the `TS` context obtained from `TSCreate()`
1490: Options Database Key:
1491: . -ts_adjoint_view_solution <viewerinfo> - views the first gradient with respect to the initial values
1493: Level: intermediate
1495: Notes:
1496: This must be called after a call to `TSSolve()` that solves the forward problem
1498: By default this will integrate back to the initial time, one can use `TSAdjointSetSteps()` to step back to a later time
1500: .seealso: [](ch_ts), `TSCreate()`, `TSSetCostGradients()`, `TSSetSolution()`, `TSAdjointStep()`
1501: @*/
1502: PetscErrorCode TSAdjointSolve(TS ts)
1503: {
1504: static PetscBool cite = PETSC_FALSE;
1505: #if defined(TSADJOINT_STAGE)
1506: PetscLogStage adjoint_stage;
1507: #endif
1509: PetscFunctionBegin;
1511: PetscCall(PetscCitationsRegister("@article{Zhang2022tsadjoint,\n"
1512: " title = {{PETSc TSAdjoint: A Discrete Adjoint ODE Solver for First-Order and Second-Order Sensitivity Analysis}},\n"
1513: " author = {Zhang, Hong and Constantinescu, Emil M. and Smith, Barry F.},\n"
1514: " journal = {SIAM Journal on Scientific Computing},\n"
1515: " volume = {44},\n"
1516: " number = {1},\n"
1517: " pages = {C1-C24},\n"
1518: " doi = {10.1137/21M140078X},\n"
1519: " year = {2022}\n}\n",
1520: &cite));
1521: #if defined(TSADJOINT_STAGE)
1522: PetscCall(PetscLogStageRegister("TSAdjoint", &adjoint_stage));
1523: PetscCall(PetscLogStagePush(adjoint_stage));
1524: #endif
1525: PetscCall(TSAdjointSetUp(ts));
1527: /* reset time step and iteration counters */
1528: ts->adjoint_steps = 0;
1529: ts->ksp_its = 0;
1530: ts->snes_its = 0;
1531: ts->num_snes_failures = 0;
1532: ts->reject = 0;
1533: ts->reason = TS_CONVERGED_ITERATING;
1535: if (!ts->adjoint_max_steps) ts->adjoint_max_steps = ts->steps;
1536: if (ts->adjoint_steps >= ts->adjoint_max_steps) ts->reason = TS_CONVERGED_ITS;
1538: while (!ts->reason) {
1539: PetscCall(TSTrajectoryGet(ts->trajectory, ts, ts->steps, &ts->ptime));
1540: PetscCall(TSAdjointMonitor(ts, ts->steps, ts->ptime, ts->vec_sol, ts->numcost, ts->vecs_sensi, ts->vecs_sensip));
1541: PetscCall(TSAdjointEventHandler(ts));
1542: PetscCall(TSAdjointStep(ts));
1543: if ((ts->vec_costintegral || ts->quadraturets) && !ts->costintegralfwd) PetscCall(TSAdjointCostIntegral(ts));
1544: }
1545: if (!ts->steps) {
1546: PetscCall(TSTrajectoryGet(ts->trajectory, ts, ts->steps, &ts->ptime));
1547: PetscCall(TSAdjointMonitor(ts, ts->steps, ts->ptime, ts->vec_sol, ts->numcost, ts->vecs_sensi, ts->vecs_sensip));
1548: }
1549: ts->solvetime = ts->ptime;
1550: PetscCall(TSTrajectoryViewFromOptions(ts->trajectory, NULL, "-ts_trajectory_view"));
1551: PetscCall(VecViewFromOptions(ts->vecs_sensi[0], (PetscObject)ts, "-ts_adjoint_view_solution"));
1552: ts->adjoint_max_steps = 0;
1553: #if defined(TSADJOINT_STAGE)
1554: PetscCall(PetscLogStagePop());
1555: #endif
1556: PetscFunctionReturn(PETSC_SUCCESS);
1557: }
1559: /*@C
1560: TSAdjointMonitor - Runs all user-provided adjoint monitor routines set using `TSAdjointMonitorSet()`
1562: Collective
1564: Input Parameters:
1565: + ts - time stepping context obtained from `TSCreate()`
1566: . step - step number that has just completed
1567: . ptime - model time of the state
1568: . u - state at the current model time
1569: . numcost - number of cost functions (dimension of lambda or mu)
1570: . lambda - vectors containing the gradients of the cost functions with respect to the ODE/DAE solution variables
1571: - mu - vectors containing the gradients of the cost functions with respect to the problem parameters
1573: Level: developer
1575: Note:
1576: `TSAdjointMonitor()` is typically used automatically within the time stepping implementations.
1577: Users would almost never call this routine directly.
1579: .seealso: `TSAdjointMonitorSet()`, `TSAdjointSolve()`
1580: @*/
1581: PetscErrorCode TSAdjointMonitor(TS ts, PetscInt step, PetscReal ptime, Vec u, PetscInt numcost, Vec *lambda, Vec *mu)
1582: {
1583: PetscInt i, n = ts->numberadjointmonitors;
1585: PetscFunctionBegin;
1588: PetscCall(VecLockReadPush(u));
1589: for (i = 0; i < n; i++) PetscCall((*ts->adjointmonitor[i])(ts, step, ptime, u, numcost, lambda, mu, ts->adjointmonitorcontext[i]));
1590: PetscCall(VecLockReadPop(u));
1591: PetscFunctionReturn(PETSC_SUCCESS);
1592: }
1594: /*@
1595: TSAdjointCostIntegral - Evaluate the cost integral in the adjoint run.
1597: Collective
1599: Input Parameter:
1600: . ts - time stepping context
1602: Level: advanced
1604: Notes:
1605: This function cannot be called until `TSAdjointStep()` has been completed.
1607: .seealso: [](ch_ts), `TSAdjointSolve()`, `TSAdjointStep()`
1608: @*/
1609: PetscErrorCode TSAdjointCostIntegral(TS ts)
1610: {
1611: PetscFunctionBegin;
1613: PetscUseTypeMethod(ts, adjointintegral);
1614: PetscFunctionReturn(PETSC_SUCCESS);
1615: }
1617: /* ------------------ Forward (tangent linear) sensitivity ------------------*/
1619: /*@
1620: TSForwardSetUp - Sets up the internal data structures for the later use
1621: of forward sensitivity analysis
1623: Collective
1625: Input Parameter:
1626: . ts - the `TS` context obtained from `TSCreate()`
1628: Level: advanced
1630: .seealso: [](ch_ts), `TS`, `TSCreate()`, `TSDestroy()`, `TSSetUp()`
1631: @*/
1632: PetscErrorCode TSForwardSetUp(TS ts)
1633: {
1634: PetscFunctionBegin;
1636: if (ts->forwardsetupcalled) PetscFunctionReturn(PETSC_SUCCESS);
1637: PetscTryTypeMethod(ts, forwardsetup);
1638: PetscCall(VecDuplicate(ts->vec_sol, &ts->vec_sensip_col));
1639: ts->forwardsetupcalled = PETSC_TRUE;
1640: PetscFunctionReturn(PETSC_SUCCESS);
1641: }
1643: /*@
1644: TSForwardReset - Reset the internal data structures used by forward sensitivity analysis
1646: Collective
1648: Input Parameter:
1649: . ts - the `TS` context obtained from `TSCreate()`
1651: Level: advanced
1653: .seealso: [](ch_ts), `TSCreate()`, `TSDestroy()`, `TSForwardSetUp()`
1654: @*/
1655: PetscErrorCode TSForwardReset(TS ts)
1656: {
1657: TS quadts = ts->quadraturets;
1659: PetscFunctionBegin;
1661: PetscTryTypeMethod(ts, forwardreset);
1662: PetscCall(MatDestroy(&ts->mat_sensip));
1663: if (quadts) PetscCall(MatDestroy(&quadts->mat_sensip));
1664: PetscCall(VecDestroy(&ts->vec_sensip_col));
1665: ts->forward_solve = PETSC_FALSE;
1666: ts->forwardsetupcalled = PETSC_FALSE;
1667: PetscFunctionReturn(PETSC_SUCCESS);
1668: }
1670: /*@
1671: TSForwardSetIntegralGradients - Set the vectors holding forward sensitivities of the integral term.
1673: Input Parameters:
1674: + ts - the `TS` context obtained from `TSCreate()`
1675: . numfwdint - number of integrals
1676: - vp - the vectors containing the gradients for each integral w.r.t. parameters
1678: Level: deprecated
1680: .seealso: [](ch_ts), `TSForwardGetSensitivities()`, `TSForwardGetIntegralGradients()`, `TSForwardStep()`
1681: @*/
1682: PetscErrorCode TSForwardSetIntegralGradients(TS ts, PetscInt numfwdint, Vec *vp)
1683: {
1684: PetscFunctionBegin;
1686: PetscCheck(!ts->numcost || ts->numcost == numfwdint, PetscObjectComm((PetscObject)ts), PETSC_ERR_USER, "The number of cost functions (2nd parameter of TSSetCostIntegrand()) is inconsistent with the one set by TSSetCostIntegrand()");
1687: if (!ts->numcost) ts->numcost = numfwdint;
1689: ts->vecs_integral_sensip = vp;
1690: PetscFunctionReturn(PETSC_SUCCESS);
1691: }
1693: /*@
1694: TSForwardGetIntegralGradients - Returns the forward sensitivities of the integral term.
1696: Input Parameter:
1697: . ts - the `TS` context obtained from `TSCreate()`
1699: Output Parameters:
1700: + numfwdint - number of integrals
1701: - vp - the vectors containing the gradients for each integral w.r.t. parameters
1703: Level: deprecated
1705: .seealso: [](ch_ts), `TSForwardSetSensitivities()`, `TSForwardSetIntegralGradients()`, `TSForwardStep()`
1706: @*/
1707: PetscErrorCode TSForwardGetIntegralGradients(TS ts, PetscInt *numfwdint, Vec **vp)
1708: {
1709: PetscFunctionBegin;
1711: PetscAssertPointer(vp, 3);
1712: if (numfwdint) *numfwdint = ts->numcost;
1713: if (vp) *vp = ts->vecs_integral_sensip;
1714: PetscFunctionReturn(PETSC_SUCCESS);
1715: }
1717: /*@
1718: TSForwardStep - Compute the forward sensitivity for one time step.
1720: Collective
1722: Input Parameter:
1723: . ts - time stepping context
1725: Level: advanced
1727: Notes:
1728: This function cannot be called until `TSStep()` has been completed.
1730: .seealso: [](ch_ts), `TSForwardSetSensitivities()`, `TSForwardGetSensitivities()`, `TSForwardSetIntegralGradients()`, `TSForwardGetIntegralGradients()`, `TSForwardSetUp()`
1731: @*/
1732: PetscErrorCode TSForwardStep(TS ts)
1733: {
1734: PetscFunctionBegin;
1736: PetscCall(PetscLogEventBegin(TS_ForwardStep, ts, 0, 0, 0));
1737: PetscUseTypeMethod(ts, forwardstep);
1738: PetscCall(PetscLogEventEnd(TS_ForwardStep, ts, 0, 0, 0));
1739: PetscCheck(ts->reason >= 0 || !ts->errorifstepfailed, PetscObjectComm((PetscObject)ts), PETSC_ERR_NOT_CONVERGED, "TSFowardStep has failed due to %s", TSConvergedReasons[ts->reason]);
1740: PetscFunctionReturn(PETSC_SUCCESS);
1741: }
1743: /*@
1744: TSForwardSetSensitivities - Sets the initial value of the trajectory sensitivities of solution w.r.t. the problem parameters and initial values.
1746: Logically Collective
1748: Input Parameters:
1749: + ts - the `TS` context obtained from `TSCreate()`
1750: . nump - number of parameters
1751: - Smat - sensitivities with respect to the parameters, the number of entries in these vectors is the same as the number of parameters
1753: Level: beginner
1755: Notes:
1756: Forward sensitivity is also called 'trajectory sensitivity' in some fields such as power systems.
1757: This function turns on a flag to trigger `TSSolve()` to compute forward sensitivities automatically.
1758: You must call this function before `TSSolve()`.
1759: The entries in the sensitivity matrix must be correctly initialized with the values S = dy/dp|startingtime.
1761: .seealso: [](ch_ts), `TSForwardGetSensitivities()`, `TSForwardSetIntegralGradients()`, `TSForwardGetIntegralGradients()`, `TSForwardStep()`
1762: @*/
1763: PetscErrorCode TSForwardSetSensitivities(TS ts, PetscInt nump, Mat Smat)
1764: {
1765: PetscFunctionBegin;
1768: ts->forward_solve = PETSC_TRUE;
1769: if (nump == PETSC_DEFAULT) {
1770: PetscCall(MatGetSize(Smat, NULL, &ts->num_parameters));
1771: } else ts->num_parameters = nump;
1772: PetscCall(PetscObjectReference((PetscObject)Smat));
1773: PetscCall(MatDestroy(&ts->mat_sensip));
1774: ts->mat_sensip = Smat;
1775: PetscFunctionReturn(PETSC_SUCCESS);
1776: }
1778: /*@
1779: TSForwardGetSensitivities - Returns the trajectory sensitivities
1781: Not Collective, but Smat returned is parallel if ts is parallel
1783: Output Parameters:
1784: + ts - the `TS` context obtained from `TSCreate()`
1785: . nump - number of parameters
1786: - Smat - sensitivities with respect to the parameters, the number of entries in these vectors is the same as the number of parameters
1788: Level: intermediate
1790: .seealso: [](ch_ts), `TSForwardSetSensitivities()`, `TSForwardSetIntegralGradients()`, `TSForwardGetIntegralGradients()`, `TSForwardStep()`
1791: @*/
1792: PetscErrorCode TSForwardGetSensitivities(TS ts, PetscInt *nump, Mat *Smat)
1793: {
1794: PetscFunctionBegin;
1796: if (nump) *nump = ts->num_parameters;
1797: if (Smat) *Smat = ts->mat_sensip;
1798: PetscFunctionReturn(PETSC_SUCCESS);
1799: }
1801: /*@
1802: TSForwardCostIntegral - Evaluate the cost integral in the forward run.
1804: Collective
1806: Input Parameter:
1807: . ts - time stepping context
1809: Level: advanced
1811: Note:
1812: This function cannot be called until `TSStep()` has been completed.
1814: .seealso: [](ch_ts), `TS`, `TSSolve()`, `TSAdjointCostIntegral()`
1815: @*/
1816: PetscErrorCode TSForwardCostIntegral(TS ts)
1817: {
1818: PetscFunctionBegin;
1820: PetscUseTypeMethod(ts, forwardintegral);
1821: PetscFunctionReturn(PETSC_SUCCESS);
1822: }
1824: /*@
1825: TSForwardSetInitialSensitivities - Set initial values for tangent linear sensitivities
1827: Collective
1829: Input Parameters:
1830: + ts - the `TS` context obtained from `TSCreate()`
1831: - didp - parametric sensitivities of the initial condition
1833: Level: intermediate
1835: Notes:
1836: `TSSolve()` allows users to pass the initial solution directly to `TS`. But the tangent linear variables cannot be initialized in this way.
1837: This function is used to set initial values for tangent linear variables.
1839: .seealso: [](ch_ts), `TS`, `TSForwardSetSensitivities()`
1840: @*/
1841: PetscErrorCode TSForwardSetInitialSensitivities(TS ts, Mat didp)
1842: {
1843: PetscFunctionBegin;
1846: if (!ts->mat_sensip) PetscCall(TSForwardSetSensitivities(ts, PETSC_DEFAULT, didp));
1847: PetscFunctionReturn(PETSC_SUCCESS);
1848: }
1850: /*@
1851: TSForwardGetStages - Get the number of stages and the tangent linear sensitivities at the intermediate stages
1853: Input Parameter:
1854: . ts - the `TS` context obtained from `TSCreate()`
1856: Output Parameters:
1857: + ns - number of stages
1858: - S - tangent linear sensitivities at the intermediate stages
1860: Level: advanced
1862: .seealso: `TS`
1863: @*/
1864: PetscErrorCode TSForwardGetStages(TS ts, PetscInt *ns, Mat **S)
1865: {
1866: PetscFunctionBegin;
1869: if (!ts->ops->getstages) *S = NULL;
1870: else PetscUseTypeMethod(ts, forwardgetstages, ns, S);
1871: PetscFunctionReturn(PETSC_SUCCESS);
1872: }
1874: /*@
1875: TSCreateQuadratureTS - Create a sub-`TS` that evaluates integrals over time
1877: Input Parameters:
1878: + ts - the `TS` context obtained from `TSCreate()`
1879: - fwd - flag indicating whether to evaluate cost integral in the forward run or the adjoint run
1881: Output Parameter:
1882: . quadts - the child `TS` context
1884: Level: intermediate
1886: .seealso: [](ch_ts), `TSGetQuadratureTS()`
1887: @*/
1888: PetscErrorCode TSCreateQuadratureTS(TS ts, PetscBool fwd, TS *quadts)
1889: {
1890: char prefix[128];
1892: PetscFunctionBegin;
1894: PetscAssertPointer(quadts, 3);
1895: PetscCall(TSDestroy(&ts->quadraturets));
1896: PetscCall(TSCreate(PetscObjectComm((PetscObject)ts), &ts->quadraturets));
1897: PetscCall(PetscObjectIncrementTabLevel((PetscObject)ts->quadraturets, (PetscObject)ts, 1));
1898: PetscCall(PetscSNPrintf(prefix, sizeof(prefix), "%squad_", ((PetscObject)ts)->prefix ? ((PetscObject)ts)->prefix : ""));
1899: PetscCall(TSSetOptionsPrefix(ts->quadraturets, prefix));
1900: *quadts = ts->quadraturets;
1902: if (ts->numcost) {
1903: PetscCall(VecCreateSeq(PETSC_COMM_SELF, ts->numcost, &(*quadts)->vec_sol));
1904: } else {
1905: PetscCall(VecCreateSeq(PETSC_COMM_SELF, 1, &(*quadts)->vec_sol));
1906: }
1907: ts->costintegralfwd = fwd;
1908: PetscFunctionReturn(PETSC_SUCCESS);
1909: }
1911: /*@
1912: TSGetQuadratureTS - Return the sub-`TS` that evaluates integrals over time
1914: Input Parameter:
1915: . ts - the `TS` context obtained from `TSCreate()`
1917: Output Parameters:
1918: + fwd - flag indicating whether to evaluate cost integral in the forward run or the adjoint run
1919: - quadts - the child `TS` context
1921: Level: intermediate
1923: .seealso: [](ch_ts), `TSCreateQuadratureTS()`
1924: @*/
1925: PetscErrorCode TSGetQuadratureTS(TS ts, PetscBool *fwd, TS *quadts)
1926: {
1927: PetscFunctionBegin;
1929: if (fwd) *fwd = ts->costintegralfwd;
1930: if (quadts) *quadts = ts->quadraturets;
1931: PetscFunctionReturn(PETSC_SUCCESS);
1932: }
1934: /*@
1935: TSComputeSNESJacobian - Compute the Jacobian needed for the `SNESSolve()` in `TS`
1937: Collective
1939: Input Parameters:
1940: + ts - the `TS` context obtained from `TSCreate()`
1941: - x - state vector
1943: Output Parameters:
1944: + J - Jacobian matrix
1945: - Jpre - preconditioning matrix for J (may be same as J)
1947: Level: developer
1949: Note:
1950: Uses finite differencing when `TS` Jacobian is not available.
1952: .seealso: `SNES`, `TS`, `SNESSetJacobian()`, `TSSetRHSJacobian()`, `TSSetIJacobian()`
1953: @*/
1954: PetscErrorCode TSComputeSNESJacobian(TS ts, Vec x, Mat J, Mat Jpre)
1955: {
1956: SNES snes = ts->snes;
1957: PetscErrorCode (*jac)(SNES, Vec, Mat, Mat, void *) = NULL;
1959: PetscFunctionBegin;
1960: /*
1961: Unlike implicit methods, explicit methods do not have SNESMatFDColoring in the snes object
1962: because SNESSolve() has not been called yet; so querying SNESMatFDColoring does not work for
1963: explicit methods. Instead, we check the Jacobian compute function directly to determine if FD
1964: coloring is used.
1965: */
1966: PetscCall(SNESGetJacobian(snes, NULL, NULL, &jac, NULL));
1967: if (jac == SNESComputeJacobianDefaultColor) {
1968: Vec f;
1969: PetscCall(SNESSetSolution(snes, x));
1970: PetscCall(SNESGetFunction(snes, &f, NULL, NULL));
1971: /* Force MatFDColoringApply to evaluate the SNES residual function for the base vector */
1972: PetscCall(SNESComputeFunction(snes, x, f));
1973: }
1974: PetscCall(SNESComputeJacobian(snes, x, J, Jpre));
1975: PetscFunctionReturn(PETSC_SUCCESS);
1976: }