xref: /petsc/src/ts/tutorials/ex5.c (revision f97672e55eacc8688507b9471cd7ec2664d7f203)
1 
2 static char help[] ="Solves a simple time-dependent linear PDE (the heat equation).\n\
3 Input parameters include:\n\
4   -m <points>, where <points> = number of grid points\n\
5   -time_dependent_rhs : Treat the problem as having a time-dependent right-hand side\n\
6   -debug              : Activate debugging printouts\n\
7   -nox                : Deactivate x-window graphics\n\n";
8 
9 /* ------------------------------------------------------------------------
10 
11    This program solves the one-dimensional heat equation (also called the
12    diffusion equation),
13        u_t = u_xx,
14    on the domain 0 <= x <= 1, with the boundary conditions
15        u(t,0) = 1, u(t,1) = 1,
16    and the initial condition
17        u(0,x) = cos(6*pi*x) + 3*cos(2*pi*x).
18    This is a linear, second-order, parabolic equation.
19 
20    We discretize the right-hand side using finite differences with
21    uniform grid spacing h:
22        u_xx = (u_{i+1} - 2u_{i} + u_{i-1})/(h^2)
23    We then demonstrate time evolution using the various TS methods by
24    running the program via
25        ex3 -ts_type <timestepping solver>
26 
27    We compare the approximate solution with the exact solution, given by
28        u_exact(x,t) = exp(-36*pi*pi*t) * cos(6*pi*x) +
29                       3*exp(-4*pi*pi*t) * cos(2*pi*x)
30 
31    Notes:
32    This code demonstrates the TS solver interface to two variants of
33    linear problems, u_t = f(u,t), namely
34      - time-dependent f:   f(u,t) is a function of t
35      - time-independent f: f(u,t) is simply just f(u)
36 
37     The parallel version of this code is ts/tutorials/ex4.c
38 
39   ------------------------------------------------------------------------- */
40 
41 /*
42    Include "petscts.h" so that we can use TS solvers.  Note that this file
43    automatically includes:
44      petscsys.h       - base PETSc routines   petscvec.h  - vectors
45      petscmat.h  - matrices
46      petscis.h     - index sets            petscksp.h  - Krylov subspace methods
47      petscviewer.h - viewers               petscpc.h   - preconditioners
48      petscksp.h   - linear solvers        petscsnes.h - nonlinear solvers
49 */
50 #include <petscts.h>
51 #include <petscdraw.h>
52 
53 /*
54    User-defined application context - contains data needed by the
55    application-provided call-back routines.
56 */
57 typedef struct {
58   Vec         solution;          /* global exact solution vector */
59   PetscInt    m;                      /* total number of grid points */
60   PetscReal   h;                 /* mesh width h = 1/(m-1) */
61   PetscBool   debug;             /* flag (1 indicates activation of debugging printouts) */
62   PetscViewer viewer1,viewer2;  /* viewers for the solution and error */
63   PetscReal   norm_2,norm_max;  /* error norms */
64 } AppCtx;
65 
66 /*
67    User-defined routines
68 */
69 extern PetscErrorCode InitialConditions(Vec,AppCtx*);
70 extern PetscErrorCode RHSMatrixHeat(TS,PetscReal,Vec,Mat,Mat,void*);
71 extern PetscErrorCode Monitor(TS,PetscInt,PetscReal,Vec,void*);
72 extern PetscErrorCode ExactSolution(PetscReal,Vec,AppCtx*);
73 
74 int main(int argc,char **argv)
75 {
76   AppCtx         appctx;                 /* user-defined application context */
77   TS             ts;                     /* timestepping context */
78   Mat            A;                      /* matrix data structure */
79   Vec            u;                      /* approximate solution vector */
80   PetscReal      time_total_max = 100.0; /* default max total time */
81   PetscInt       time_steps_max = 100;   /* default max timesteps */
82   PetscDraw      draw;                   /* drawing context */
83   PetscInt       steps,m;
84   PetscMPIInt    size;
85   PetscBool      flg;
86   PetscReal      dt,ftime;
87 
88   /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
89      Initialize program and set problem parameters
90      - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
91 
92   PetscCall(PetscInitialize(&argc,&argv,(char*)0,help));
93   PetscCallMPI(MPI_Comm_size(PETSC_COMM_WORLD,&size));
94   PetscCheck(size == 1,PETSC_COMM_WORLD,PETSC_ERR_WRONG_MPI_SIZE,"This is a uniprocessor example only!");
95 
96   m               = 60;
97   PetscCall(PetscOptionsGetInt(NULL,NULL,"-m",&m,NULL));
98   PetscCall(PetscOptionsHasName(NULL,NULL,"-debug",&appctx.debug));
99   appctx.m        = m;
100   appctx.h        = 1.0/(m-1.0);
101   appctx.norm_2   = 0.0;
102   appctx.norm_max = 0.0;
103 
104   PetscCall(PetscPrintf(PETSC_COMM_SELF,"Solving a linear TS problem on 1 processor\n"));
105 
106   /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
107      Create vector data structures
108      - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
109 
110   /*
111      Create vector data structures for approximate and exact solutions
112   */
113   PetscCall(VecCreateSeq(PETSC_COMM_SELF,m,&u));
114   PetscCall(VecDuplicate(u,&appctx.solution));
115 
116   /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
117      Set up displays to show graphs of the solution and error
118      - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
119 
120   PetscCall(PetscViewerDrawOpen(PETSC_COMM_SELF,0,"",80,380,400,160,&appctx.viewer1));
121   PetscCall(PetscViewerDrawGetDraw(appctx.viewer1,0,&draw));
122   PetscCall(PetscDrawSetDoubleBuffer(draw));
123   PetscCall(PetscViewerDrawOpen(PETSC_COMM_SELF,0,"",80,0,400,160,&appctx.viewer2));
124   PetscCall(PetscViewerDrawGetDraw(appctx.viewer2,0,&draw));
125   PetscCall(PetscDrawSetDoubleBuffer(draw));
126 
127   /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
128      Create timestepping solver context
129      - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
130 
131   PetscCall(TSCreate(PETSC_COMM_SELF,&ts));
132   PetscCall(TSSetProblemType(ts,TS_LINEAR));
133 
134   /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
135      Set optional user-defined monitoring routine
136      - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
137 
138   PetscCall(TSMonitorSet(ts,Monitor,&appctx,NULL));
139 
140   /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
141 
142      Create matrix data structure; set matrix evaluation routine.
143      - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
144 
145   PetscCall(MatCreate(PETSC_COMM_SELF,&A));
146   PetscCall(MatSetSizes(A,PETSC_DECIDE,PETSC_DECIDE,m,m));
147   PetscCall(MatSetFromOptions(A));
148   PetscCall(MatSetUp(A));
149 
150   PetscCall(PetscOptionsHasName(NULL,NULL,"-time_dependent_rhs",&flg));
151   if (flg) {
152     /*
153        For linear problems with a time-dependent f(u,t) in the equation
154        u_t = f(u,t), the user provides the discretized right-hand-side
155        as a time-dependent matrix.
156     */
157     PetscCall(TSSetRHSFunction(ts,NULL,TSComputeRHSFunctionLinear,&appctx));
158     PetscCall(TSSetRHSJacobian(ts,A,A,RHSMatrixHeat,&appctx));
159   } else {
160     /*
161        For linear problems with a time-independent f(u) in the equation
162        u_t = f(u), the user provides the discretized right-hand-side
163        as a matrix only once, and then sets a null matrix evaluation
164        routine.
165     */
166     PetscCall(RHSMatrixHeat(ts,0.0,u,A,A,&appctx));
167     PetscCall(TSSetRHSFunction(ts,NULL,TSComputeRHSFunctionLinear,&appctx));
168     PetscCall(TSSetRHSJacobian(ts,A,A,TSComputeRHSJacobianConstant,&appctx));
169   }
170 
171   /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
172      Set solution vector and initial timestep
173      - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
174 
175   dt   = appctx.h*appctx.h/2.0;
176   PetscCall(TSSetTimeStep(ts,dt));
177   PetscCall(TSSetSolution(ts,u));
178 
179   /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
180      Customize timestepping solver:
181        - Set the solution method to be the Backward Euler method.
182        - Set timestepping duration info
183      Then set runtime options, which can override these defaults.
184      For example,
185           -ts_max_steps <maxsteps> -ts_max_time <maxtime>
186      to override the defaults set by TSSetMaxSteps()/TSSetMaxTime().
187      - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
188 
189   PetscCall(TSSetMaxSteps(ts,time_steps_max));
190   PetscCall(TSSetMaxTime(ts,time_total_max));
191   PetscCall(TSSetExactFinalTime(ts,TS_EXACTFINALTIME_STEPOVER));
192   PetscCall(TSSetFromOptions(ts));
193 
194   /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
195      Solve the problem
196      - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
197 
198   /*
199      Evaluate initial conditions
200   */
201   PetscCall(InitialConditions(u,&appctx));
202 
203   /*
204      Run the timestepping solver
205   */
206   PetscCall(TSSolve(ts,u));
207   PetscCall(TSGetSolveTime(ts,&ftime));
208   PetscCall(TSGetStepNumber(ts,&steps));
209 
210   /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
211      View timestepping solver info
212      - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
213 
214   PetscCall(PetscPrintf(PETSC_COMM_SELF,"avg. error (2 norm) = %g, avg. error (max norm) = %g\n",(double)(appctx.norm_2/steps),(double)(appctx.norm_max/steps)));
215   PetscCall(TSView(ts,PETSC_VIEWER_STDOUT_SELF));
216 
217   /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
218      Free work space.  All PETSc objects should be destroyed when they
219      are no longer needed.
220      - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
221 
222   PetscCall(TSDestroy(&ts));
223   PetscCall(MatDestroy(&A));
224   PetscCall(VecDestroy(&u));
225   PetscCall(PetscViewerDestroy(&appctx.viewer1));
226   PetscCall(PetscViewerDestroy(&appctx.viewer2));
227   PetscCall(VecDestroy(&appctx.solution));
228 
229   /*
230      Always call PetscFinalize() before exiting a program.  This routine
231        - finalizes the PETSc libraries as well as MPI
232        - provides summary and diagnostic information if certain runtime
233          options are chosen (e.g., -log_view).
234   */
235   PetscCall(PetscFinalize());
236   return 0;
237 }
238 /* --------------------------------------------------------------------- */
239 /*
240    InitialConditions - Computes the solution at the initial time.
241 
242    Input Parameter:
243    u - uninitialized solution vector (global)
244    appctx - user-defined application context
245 
246    Output Parameter:
247    u - vector with solution at initial time (global)
248 */
249 PetscErrorCode InitialConditions(Vec u,AppCtx *appctx)
250 {
251   PetscScalar    *u_localptr,h = appctx->h;
252   PetscInt       i;
253 
254   /*
255     Get a pointer to vector data.
256     - For default PETSc vectors, VecGetArray() returns a pointer to
257       the data array.  Otherwise, the routine is implementation dependent.
258     - You MUST call VecRestoreArray() when you no longer need access to
259       the array.
260     - Note that the Fortran interface to VecGetArray() differs from the
261       C version.  See the users manual for details.
262   */
263   PetscCall(VecGetArray(u,&u_localptr));
264 
265   /*
266      We initialize the solution array by simply writing the solution
267      directly into the array locations.  Alternatively, we could use
268      VecSetValues() or VecSetValuesLocal().
269   */
270   for (i=0; i<appctx->m; i++) u_localptr[i] = PetscCosScalar(PETSC_PI*i*6.*h) + 3.*PetscCosScalar(PETSC_PI*i*2.*h);
271 
272   /*
273      Restore vector
274   */
275   PetscCall(VecRestoreArray(u,&u_localptr));
276 
277   /*
278      Print debugging information if desired
279   */
280   if (appctx->debug) {
281     printf("initial guess vector\n");
282     PetscCall(VecView(u,PETSC_VIEWER_STDOUT_SELF));
283   }
284 
285   return 0;
286 }
287 /* --------------------------------------------------------------------- */
288 /*
289    ExactSolution - Computes the exact solution at a given time.
290 
291    Input Parameters:
292    t - current time
293    solution - vector in which exact solution will be computed
294    appctx - user-defined application context
295 
296    Output Parameter:
297    solution - vector with the newly computed exact solution
298 */
299 PetscErrorCode ExactSolution(PetscReal t,Vec solution,AppCtx *appctx)
300 {
301   PetscScalar    *s_localptr,h = appctx->h,ex1,ex2,sc1,sc2,tc = t;
302   PetscInt       i;
303 
304   /*
305      Get a pointer to vector data.
306   */
307   PetscCall(VecGetArray(solution,&s_localptr));
308 
309   /*
310      Simply write the solution directly into the array locations.
311      Alternatively, we culd use VecSetValues() or VecSetValuesLocal().
312   */
313   ex1 = PetscExpScalar(-36.*PETSC_PI*PETSC_PI*tc); ex2 = PetscExpScalar(-4.*PETSC_PI*PETSC_PI*tc);
314   sc1 = PETSC_PI*6.*h;                 sc2 = PETSC_PI*2.*h;
315   for (i=0; i<appctx->m; i++) s_localptr[i] = PetscCosScalar(sc1*(PetscReal)i)*ex1 + 3.*PetscCosScalar(sc2*(PetscReal)i)*ex2;
316 
317   /*
318      Restore vector
319   */
320   PetscCall(VecRestoreArray(solution,&s_localptr));
321   return 0;
322 }
323 /* --------------------------------------------------------------------- */
324 /*
325    Monitor - User-provided routine to monitor the solution computed at
326    each timestep.  This example plots the solution and computes the
327    error in two different norms.
328 
329    Input Parameters:
330    ts     - the timestep context
331    step   - the count of the current step (with 0 meaning the
332              initial condition)
333    time   - the current time
334    u      - the solution at this timestep
335    ctx    - the user-provided context for this monitoring routine.
336             In this case we use the application context which contains
337             information about the problem size, workspace and the exact
338             solution.
339 */
340 PetscErrorCode Monitor(TS ts,PetscInt step,PetscReal time,Vec u,void *ctx)
341 {
342   AppCtx         *appctx = (AppCtx*) ctx;   /* user-defined application context */
343   PetscReal      norm_2,norm_max;
344 
345   /*
346      View a graph of the current iterate
347   */
348   PetscCall(VecView(u,appctx->viewer2));
349 
350   /*
351      Compute the exact solution
352   */
353   PetscCall(ExactSolution(time,appctx->solution,appctx));
354 
355   /*
356      Print debugging information if desired
357   */
358   if (appctx->debug) {
359     printf("Computed solution vector\n");
360     PetscCall(VecView(u,PETSC_VIEWER_STDOUT_SELF));
361     printf("Exact solution vector\n");
362     PetscCall(VecView(appctx->solution,PETSC_VIEWER_STDOUT_SELF));
363   }
364 
365   /*
366      Compute the 2-norm and max-norm of the error
367   */
368   PetscCall(VecAXPY(appctx->solution,-1.0,u));
369   PetscCall(VecNorm(appctx->solution,NORM_2,&norm_2));
370   norm_2 = PetscSqrtReal(appctx->h)*norm_2;
371   PetscCall(VecNorm(appctx->solution,NORM_MAX,&norm_max));
372   if (norm_2   < 1e-14) norm_2   = 0;
373   if (norm_max < 1e-14) norm_max = 0;
374 
375   PetscCall(PetscPrintf(PETSC_COMM_WORLD,"Timestep %" PetscInt_FMT ": time = %g, 2-norm error = %g, max norm error = %g\n",step,(double)time,(double)norm_2,(double)norm_max));
376   appctx->norm_2   += norm_2;
377   appctx->norm_max += norm_max;
378 
379   /*
380      View a graph of the error
381   */
382   PetscCall(VecView(appctx->solution,appctx->viewer1));
383 
384   /*
385      Print debugging information if desired
386   */
387   if (appctx->debug) {
388     printf("Error vector\n");
389     PetscCall(VecView(appctx->solution,PETSC_VIEWER_STDOUT_SELF));
390   }
391 
392   return 0;
393 }
394 /* --------------------------------------------------------------------- */
395 /*
396    RHSMatrixHeat - User-provided routine to compute the right-hand-side
397    matrix for the heat equation.
398 
399    Input Parameters:
400    ts - the TS context
401    t - current time
402    global_in - global input vector
403    dummy - optional user-defined context, as set by TSetRHSJacobian()
404 
405    Output Parameters:
406    AA - Jacobian matrix
407    BB - optionally different preconditioning matrix
408    str - flag indicating matrix structure
409 
410   Notes:
411   Recall that MatSetValues() uses 0-based row and column numbers
412   in Fortran as well as in C.
413 */
414 PetscErrorCode RHSMatrixHeat(TS ts,PetscReal t,Vec X,Mat AA,Mat BB,void *ctx)
415 {
416   Mat            A       = AA;                /* Jacobian matrix */
417   AppCtx         *appctx = (AppCtx*)ctx;     /* user-defined application context */
418   PetscInt       mstart  = 0;
419   PetscInt       mend    = appctx->m;
420   PetscInt       i,idx[3];
421   PetscScalar    v[3],stwo = -2./(appctx->h*appctx->h),sone = -.5*stwo;
422 
423   /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
424      Compute entries for the locally owned part of the matrix
425      - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
426   /*
427      Set matrix rows corresponding to boundary data
428   */
429 
430   mstart = 0;
431   v[0]   = 1.0;
432   PetscCall(MatSetValues(A,1,&mstart,1,&mstart,v,INSERT_VALUES));
433   mstart++;
434 
435   mend--;
436   v[0] = 1.0;
437   PetscCall(MatSetValues(A,1,&mend,1,&mend,v,INSERT_VALUES));
438 
439   /*
440      Set matrix rows corresponding to interior data.  We construct the
441      matrix one row at a time.
442   */
443   v[0] = sone; v[1] = stwo; v[2] = sone;
444   for (i=mstart; i<mend; i++) {
445     idx[0] = i-1; idx[1] = i; idx[2] = i+1;
446     PetscCall(MatSetValues(A,1,&i,3,idx,v,INSERT_VALUES));
447   }
448 
449   /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
450      Complete the matrix assembly process and set some options
451      - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
452   /*
453      Assemble matrix, using the 2-step process:
454        MatAssemblyBegin(), MatAssemblyEnd()
455      Computations can be done while messages are in transition
456      by placing code between these two statements.
457   */
458   PetscCall(MatAssemblyBegin(A,MAT_FINAL_ASSEMBLY));
459   PetscCall(MatAssemblyEnd(A,MAT_FINAL_ASSEMBLY));
460 
461   /*
462      Set and option to indicate that we will never add a new nonzero location
463      to the matrix. If we do, it will generate an error.
464   */
465   PetscCall(MatSetOption(A,MAT_NEW_NONZERO_LOCATION_ERR,PETSC_TRUE));
466 
467   return 0;
468 }
469 
470 /*TEST
471 
472     test:
473       requires: x
474 
475     test:
476       suffix: nox
477       args: -nox
478 
479 TEST*/
480