xref: /petsc/src/ts/tests/ex3.c (revision ebead697dbf761eb322f829370bbe90b3bd93fa3)
1 
2 static char help[] = "Solves 1D heat equation with FEM formulation.\n\
3 Input arguments are\n\
4   -useAlhs: solve Alhs*U' =  (Arhs*U + g) \n\
5             otherwise, solve U' = inv(Alhs)*(Arhs*U + g) \n\n";
6 
7 /*--------------------------------------------------------------------------
8   Solves 1D heat equation U_t = U_xx with FEM formulation:
9                           Alhs*U' = rhs (= Arhs*U + g)
10   We thank Chris Cox <clcox@clemson.edu> for contributing the original code
11 ----------------------------------------------------------------------------*/
12 
13 #include <petscksp.h>
14 #include <petscts.h>
15 
16 /* special variable - max size of all arrays  */
17 #define num_z 10
18 
19 /*
20    User-defined application context - contains data needed by the
21    application-provided call-back routines.
22 */
23 typedef struct {
24   Mat         Amat;               /* left hand side matrix */
25   Vec         ksp_rhs,ksp_sol;    /* working vectors for formulating inv(Alhs)*(Arhs*U+g) */
26   int         max_probsz;         /* max size of the problem */
27   PetscBool   useAlhs;            /* flag (1 indicates solving Alhs*U' = Arhs*U+g */
28   int         nz;                 /* total number of grid points */
29   PetscInt    m;                  /* total number of interio grid points */
30   Vec         solution;           /* global exact ts solution vector */
31   PetscScalar *z;                 /* array of grid points */
32   PetscBool   debug;              /* flag (1 indicates activation of debugging printouts) */
33 } AppCtx;
34 
35 extern PetscScalar exact(PetscScalar,PetscReal);
36 extern PetscErrorCode Monitor(TS,PetscInt,PetscReal,Vec,void*);
37 extern PetscErrorCode Petsc_KSPSolve(AppCtx*);
38 extern PetscScalar bspl(PetscScalar*,PetscScalar,PetscInt,PetscInt,PetscInt[][2],PetscInt);
39 extern PetscErrorCode femBg(PetscScalar[][3],PetscScalar*,PetscInt,PetscScalar*,PetscReal);
40 extern PetscErrorCode femA(AppCtx*,PetscInt,PetscScalar*);
41 extern PetscErrorCode rhs(AppCtx*,PetscScalar*, PetscInt, PetscScalar*,PetscReal);
42 extern PetscErrorCode RHSfunction(TS,PetscReal,Vec,Vec,void*);
43 
44 int main(int argc,char **argv)
45 {
46   PetscInt       i,m,nz,steps,max_steps,k,nphase=1;
47   PetscScalar    zInitial,zFinal,val,*z;
48   PetscReal      stepsz[4],T,ftime;
49   TS             ts;
50   SNES           snes;
51   Mat            Jmat;
52   AppCtx         appctx;   /* user-defined application context */
53   Vec            init_sol; /* ts solution vector */
54   PetscMPIInt    size;
55 
56   PetscFunctionBeginUser;
57   PetscCall(PetscInitialize(&argc,&argv,(char*)0,help));
58   PetscCallMPI(MPI_Comm_size(PETSC_COMM_WORLD,&size));
59   PetscCheck(size == 1,PETSC_COMM_SELF,PETSC_ERR_WRONG_MPI_SIZE,"This is a uniprocessor example only");
60 
61   /* initializations */
62   zInitial  = 0.0;
63   zFinal    = 1.0;
64   nz        = num_z;
65   m         = nz-2;
66   appctx.nz = nz;
67   max_steps = (PetscInt)10000;
68 
69   appctx.m          = m;
70   appctx.max_probsz = nz;
71   appctx.debug      = PETSC_FALSE;
72   appctx.useAlhs    = PETSC_FALSE;
73 
74   PetscOptionsBegin(PETSC_COMM_WORLD,NULL,"","");
75   PetscCall(PetscOptionsName("-debug",NULL,NULL,&appctx.debug));
76   PetscCall(PetscOptionsName("-useAlhs",NULL,NULL,&appctx.useAlhs));
77   PetscCall(PetscOptionsRangeInt("-nphase",NULL,NULL,nphase,&nphase,NULL,1,3));
78   PetscOptionsEnd();
79   T = 0.014/nphase;
80 
81   /* create vector to hold ts solution */
82   /*-----------------------------------*/
83   PetscCall(VecCreate(PETSC_COMM_WORLD, &init_sol));
84   PetscCall(VecSetSizes(init_sol, PETSC_DECIDE, m));
85   PetscCall(VecSetFromOptions(init_sol));
86 
87   /* create vector to hold true ts soln for comparison */
88   PetscCall(VecDuplicate(init_sol, &appctx.solution));
89 
90   /* create LHS matrix Amat */
91   /*------------------------*/
92   PetscCall(MatCreateSeqAIJ(PETSC_COMM_WORLD, m, m, 3, NULL, &appctx.Amat));
93   PetscCall(MatSetFromOptions(appctx.Amat));
94   PetscCall(MatSetUp(appctx.Amat));
95   /* set space grid points - interio points only! */
96   PetscCall(PetscMalloc1(nz+1,&z));
97   for (i=0; i<nz; i++) z[i]=(i)*((zFinal-zInitial)/(nz-1));
98   appctx.z = z;
99   femA(&appctx,nz,z);
100 
101   /* create the jacobian matrix */
102   /*----------------------------*/
103   PetscCall(MatCreate(PETSC_COMM_WORLD, &Jmat));
104   PetscCall(MatSetSizes(Jmat,PETSC_DECIDE,PETSC_DECIDE,m,m));
105   PetscCall(MatSetFromOptions(Jmat));
106   PetscCall(MatSetUp(Jmat));
107 
108   /* create working vectors for formulating rhs=inv(Alhs)*(Arhs*U + g) */
109   PetscCall(VecDuplicate(init_sol,&appctx.ksp_rhs));
110   PetscCall(VecDuplicate(init_sol,&appctx.ksp_sol));
111 
112   /* set initial guess */
113   /*-------------------*/
114   for (i=0; i<nz-2; i++) {
115     val  = exact(z[i+1], 0.0);
116     PetscCall(VecSetValue(init_sol,i,(PetscScalar)val,INSERT_VALUES));
117   }
118   PetscCall(VecAssemblyBegin(init_sol));
119   PetscCall(VecAssemblyEnd(init_sol));
120 
121   /*create a time-stepping context and set the problem type */
122   /*--------------------------------------------------------*/
123   PetscCall(TSCreate(PETSC_COMM_WORLD, &ts));
124   PetscCall(TSSetProblemType(ts,TS_NONLINEAR));
125 
126   /* set time-step method */
127   PetscCall(TSSetType(ts,TSCN));
128 
129   /* Set optional user-defined monitoring routine */
130   PetscCall(TSMonitorSet(ts,Monitor,&appctx,NULL));
131   /* set the right hand side of U_t = RHSfunction(U,t) */
132   PetscCall(TSSetRHSFunction(ts,NULL,(PetscErrorCode (*)(TS,PetscScalar,Vec,Vec,void*))RHSfunction,&appctx));
133 
134   if (appctx.useAlhs) {
135     /* set the left hand side matrix of Amat*U_t = rhs(U,t) */
136 
137     /* Note: this approach is incompatible with the finite differenced Jacobian set below because we can't restore the
138      * Alhs matrix without making a copy.  Either finite difference the entire thing or use analytic Jacobians in both
139      * places.
140      */
141     PetscCall(TSSetIFunction(ts,NULL,TSComputeIFunctionLinear,&appctx));
142     PetscCall(TSSetIJacobian(ts,appctx.Amat,appctx.Amat,TSComputeIJacobianConstant,&appctx));
143   }
144 
145   /* use petsc to compute the jacobian by finite differences */
146   PetscCall(TSGetSNES(ts,&snes));
147   PetscCall(SNESSetJacobian(snes,Jmat,Jmat,SNESComputeJacobianDefault,NULL));
148 
149   /* get the command line options if there are any and set them */
150   PetscCall(TSSetFromOptions(ts));
151 
152 #if defined(PETSC_HAVE_SUNDIALS2)
153   {
154     TSType    type;
155     PetscBool sundialstype=PETSC_FALSE;
156     PetscCall(TSGetType(ts,&type));
157     PetscCall(PetscObjectTypeCompare((PetscObject)ts,TSSUNDIALS,&sundialstype));
158     PetscCheck(!sundialstype || !appctx.useAlhs,PETSC_COMM_SELF,PETSC_ERR_SUP,"Cannot use Alhs formulation for TSSUNDIALS type");
159   }
160 #endif
161   /* Sets the initial solution */
162   PetscCall(TSSetSolution(ts,init_sol));
163 
164   stepsz[0] = 1.0/(2.0*(nz-1)*(nz-1)); /* (mesh_size)^2/2.0 */
165   ftime     = 0.0;
166   for (k=0; k<nphase; k++) {
167     if (nphase > 1) PetscCall(PetscPrintf(PETSC_COMM_WORLD,"Phase %" PetscInt_FMT " initial time %g, stepsz %g, duration: %g\n",k,(double)ftime,(double)stepsz[k],(double)((k+1)*T)));
168     PetscCall(TSSetTime(ts,ftime));
169     PetscCall(TSSetTimeStep(ts,stepsz[k]));
170     PetscCall(TSSetMaxSteps(ts,max_steps));
171     PetscCall(TSSetMaxTime(ts,(k+1)*T));
172     PetscCall(TSSetExactFinalTime(ts,TS_EXACTFINALTIME_STEPOVER));
173 
174     /* loop over time steps */
175     /*----------------------*/
176     PetscCall(TSSolve(ts,init_sol));
177     PetscCall(TSGetSolveTime(ts,&ftime));
178     PetscCall(TSGetStepNumber(ts,&steps));
179     stepsz[k+1] = stepsz[k]*1.5; /* change step size for the next phase */
180   }
181 
182   /* free space */
183   PetscCall(TSDestroy(&ts));
184   PetscCall(MatDestroy(&appctx.Amat));
185   PetscCall(MatDestroy(&Jmat));
186   PetscCall(VecDestroy(&appctx.ksp_rhs));
187   PetscCall(VecDestroy(&appctx.ksp_sol));
188   PetscCall(VecDestroy(&init_sol));
189   PetscCall(VecDestroy(&appctx.solution));
190   PetscCall(PetscFree(z));
191 
192   PetscCall(PetscFinalize());
193   return 0;
194 }
195 
196 /*------------------------------------------------------------------------
197   Set exact solution
198   u(z,t) = sin(6*PI*z)*exp(-36.*PI*PI*t) + 3.*sin(2*PI*z)*exp(-4.*PI*PI*t)
199 --------------------------------------------------------------------------*/
200 PetscScalar exact(PetscScalar z,PetscReal t)
201 {
202   PetscScalar val, ex1, ex2;
203 
204   ex1 = PetscExpReal(-36.*PETSC_PI*PETSC_PI*t);
205   ex2 = PetscExpReal(-4.*PETSC_PI*PETSC_PI*t);
206   val = PetscSinScalar(6*PETSC_PI*z)*ex1 + 3.*PetscSinScalar(2*PETSC_PI*z)*ex2;
207   return val;
208 }
209 
210 /*
211    Monitor - User-provided routine to monitor the solution computed at
212    each timestep.  This example plots the solution and computes the
213    error in two different norms.
214 
215    Input Parameters:
216    ts     - the timestep context
217    step   - the count of the current step (with 0 meaning the
218              initial condition)
219    time   - the current time
220    u      - the solution at this timestep
221    ctx    - the user-provided context for this monitoring routine.
222             In this case we use the application context which contains
223             information about the problem size, workspace and the exact
224             solution.
225 */
226 PetscErrorCode Monitor(TS ts,PetscInt step,PetscReal time,Vec u,void *ctx)
227 {
228   AppCtx         *appctx = (AppCtx*)ctx;
229   PetscInt       i,m=appctx->m;
230   PetscReal      norm_2,norm_max,h=1.0/(m+1);
231   PetscScalar    *u_exact;
232 
233   /* Compute the exact solution */
234   PetscCall(VecGetArrayWrite(appctx->solution,&u_exact));
235   for (i=0; i<m; i++) u_exact[i] = exact(appctx->z[i+1],time);
236   PetscCall(VecRestoreArrayWrite(appctx->solution,&u_exact));
237 
238   /* Print debugging information if desired */
239   if (appctx->debug) {
240     PetscCall(PetscPrintf(PETSC_COMM_SELF,"Computed solution vector at time %g\n",(double)time));
241     PetscCall(VecView(u,PETSC_VIEWER_STDOUT_SELF));
242     PetscCall(PetscPrintf(PETSC_COMM_SELF,"Exact solution vector\n"));
243     PetscCall(VecView(appctx->solution,PETSC_VIEWER_STDOUT_SELF));
244   }
245 
246   /* Compute the 2-norm and max-norm of the error */
247   PetscCall(VecAXPY(appctx->solution,-1.0,u));
248   PetscCall(VecNorm(appctx->solution,NORM_2,&norm_2));
249 
250   norm_2 = PetscSqrtReal(h)*norm_2;
251   PetscCall(VecNorm(appctx->solution,NORM_MAX,&norm_max));
252   PetscCall(PetscPrintf(PETSC_COMM_SELF,"Timestep %" PetscInt_FMT ": time = %g, 2-norm error = %6.4f, max norm error = %6.4f\n",step,(double)time,(double)norm_2,(double)norm_max));
253 
254   /*
255      Print debugging information if desired
256   */
257   if (appctx->debug) {
258     PetscCall(PetscPrintf(PETSC_COMM_SELF,"Error vector\n"));
259     PetscCall(VecView(appctx->solution,PETSC_VIEWER_STDOUT_SELF));
260   }
261   return 0;
262 }
263 
264 /*%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
265       Function to solve a linear system using KSP
266 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%*/
267 
268 PetscErrorCode Petsc_KSPSolve(AppCtx *obj)
269 {
270   KSP            ksp;
271   PC             pc;
272 
273   /*create the ksp context and set the operators,that is, associate the system matrix with it*/
274   PetscCall(KSPCreate(PETSC_COMM_WORLD,&ksp));
275   PetscCall(KSPSetOperators(ksp,obj->Amat,obj->Amat));
276 
277   /*get the preconditioner context, set its type and the tolerances*/
278   PetscCall(KSPGetPC(ksp,&pc));
279   PetscCall(PCSetType(pc,PCLU));
280   PetscCall(KSPSetTolerances(ksp,1.e-7,PETSC_DEFAULT,PETSC_DEFAULT,PETSC_DEFAULT));
281 
282   /*get the command line options if there are any and set them*/
283   PetscCall(KSPSetFromOptions(ksp));
284 
285   /*get the linear system (ksp) solve*/
286   PetscCall(KSPSolve(ksp,obj->ksp_rhs,obj->ksp_sol));
287 
288   PetscCall(KSPDestroy(&ksp));
289   return 0;
290 }
291 
292 /***********************************************************************
293   Function to return value of basis function or derivative of basis function.
294  ***********************************************************************
295 
296         Arguments:
297           x       = array of xpoints or nodal values
298           xx      = point at which the basis function is to be
299                       evaluated.
300           il      = interval containing xx.
301           iq      = indicates which of the two basis functions in
302                       interval intrvl should be used
303           nll     = array containing the endpoints of each interval.
304           id      = If id ~= 2, the value of the basis function
305                       is calculated; if id = 2, the value of the
306                       derivative of the basis function is returned.
307  ***********************************************************************/
308 
309 PetscScalar bspl(PetscScalar *x, PetscScalar xx,PetscInt il,PetscInt iq,PetscInt nll[][2],PetscInt id)
310 {
311   PetscScalar x1,x2,bfcn;
312   PetscInt    i1,i2,iq1,iq2;
313 
314   /* Determine which basis function in interval intrvl is to be used in */
315   iq1 = iq;
316   if (iq1==0) iq2 = 1;
317   else iq2 = 0;
318 
319   /*    Determine endpoint of the interval intrvl   */
320   i1=nll[il][iq1];
321   i2=nll[il][iq2];
322 
323   /*   Determine nodal values at the endpoints of the interval intrvl   */
324   x1=x[i1];
325   x2=x[i2];
326 
327   /*   Evaluate basis function   */
328   if (id == 2) bfcn=(1.0)/(x1-x2);
329   else bfcn=(xx-x2)/(x1-x2);
330   return bfcn;
331 }
332 
333 /*---------------------------------------------------------
334   Function called by rhs function to get B and g
335 ---------------------------------------------------------*/
336 PetscErrorCode femBg(PetscScalar btri[][3],PetscScalar *f,PetscInt nz,PetscScalar *z, PetscReal t)
337 {
338   PetscInt    i,j,jj,il,ip,ipp,ipq,iq,iquad,iqq;
339   PetscInt    nli[num_z][2],indx[num_z];
340   PetscScalar dd,dl,zip,zipq,zz,b_z,bb_z,bij;
341   PetscScalar zquad[num_z][3],dlen[num_z],qdwt[3];
342 
343   /*  initializing everything - btri and f are initialized in rhs.c  */
344   for (i=0; i < nz; i++) {
345     nli[i][0]   = 0;
346     nli[i][1]   = 0;
347     indx[i]     = 0;
348     zquad[i][0] = 0.0;
349     zquad[i][1] = 0.0;
350     zquad[i][2] = 0.0;
351     dlen[i]     = 0.0;
352   } /*end for (i)*/
353 
354   /*  quadrature weights  */
355   qdwt[0] = 1.0/6.0;
356   qdwt[1] = 4.0/6.0;
357   qdwt[2] = 1.0/6.0;
358 
359   /* 1st and last nodes have Dirichlet boundary condition -
360      set indices there to -1 */
361 
362   for (i=0; i < nz-1; i++) indx[i] = i-1;
363   indx[nz-1] = -1;
364 
365   ipq = 0;
366   for (il=0; il < nz-1; il++) {
367     ip           = ipq;
368     ipq          = ip+1;
369     zip          = z[ip];
370     zipq         = z[ipq];
371     dl           = zipq-zip;
372     zquad[il][0] = zip;
373     zquad[il][1] = (0.5)*(zip+zipq);
374     zquad[il][2] = zipq;
375     dlen[il]     = PetscAbsScalar(dl);
376     nli[il][0]   = ip;
377     nli[il][1]   = ipq;
378   }
379 
380   for (il=0; il < nz-1; il++) {
381     for (iquad=0; iquad < 3; iquad++) {
382       dd = (dlen[il])*(qdwt[iquad]);
383       zz = zquad[il][iquad];
384 
385       for (iq=0; iq < 2; iq++) {
386         ip  = nli[il][iq];
387         b_z = bspl(z,zz,il,iq,nli,2);
388         i   = indx[ip];
389 
390         if (i > -1) {
391           for (iqq=0; iqq < 2; iqq++) {
392             ipp  = nli[il][iqq];
393             bb_z = bspl(z,zz,il,iqq,nli,2);
394             j    = indx[ipp];
395             bij  = -b_z*bb_z;
396 
397             if (j > -1) {
398               jj = 1+j-i;
399               btri[i][jj] += bij*dd;
400             } else {
401               f[i] += bij*dd*exact(z[ipp], t);
402               /* f[i] += 0.0; */
403               /* if (il==0 && j==-1) { */
404               /* f[i] += bij*dd*exact(zz,t); */
405               /* }*/ /*end if*/
406             } /*end else*/
407           } /*end for (iqq)*/
408         } /*end if (i>0)*/
409       } /*end for (iq)*/
410     } /*end for (iquad)*/
411   } /*end for (il)*/
412   return 0;
413 }
414 
415 PetscErrorCode femA(AppCtx *obj,PetscInt nz,PetscScalar *z)
416 {
417   PetscInt       i,j,il,ip,ipp,ipq,iq,iquad,iqq;
418   PetscInt       nli[num_z][2],indx[num_z];
419   PetscScalar    dd,dl,zip,zipq,zz,bb,bbb,aij;
420   PetscScalar    rquad[num_z][3],dlen[num_z],qdwt[3],add_term;
421 
422   /*  initializing everything  */
423   for (i=0; i < nz; i++) {
424     nli[i][0]   = 0;
425     nli[i][1]   = 0;
426     indx[i]     = 0;
427     rquad[i][0] = 0.0;
428     rquad[i][1] = 0.0;
429     rquad[i][2] = 0.0;
430     dlen[i]     = 0.0;
431   } /*end for (i)*/
432 
433   /*  quadrature weights  */
434   qdwt[0] = 1.0/6.0;
435   qdwt[1] = 4.0/6.0;
436   qdwt[2] = 1.0/6.0;
437 
438   /* 1st and last nodes have Dirichlet boundary condition -
439      set indices there to -1 */
440 
441   for (i=0; i < nz-1; i++) indx[i]=i-1;
442   indx[nz-1]=-1;
443 
444   ipq = 0;
445 
446   for (il=0; il < nz-1; il++) {
447     ip           = ipq;
448     ipq          = ip+1;
449     zip          = z[ip];
450     zipq         = z[ipq];
451     dl           = zipq-zip;
452     rquad[il][0] = zip;
453     rquad[il][1] = (0.5)*(zip+zipq);
454     rquad[il][2] = zipq;
455     dlen[il]     = PetscAbsScalar(dl);
456     nli[il][0]   = ip;
457     nli[il][1]   = ipq;
458   } /*end for (il)*/
459 
460   for (il=0; il < nz-1; il++) {
461     for (iquad=0; iquad < 3; iquad++) {
462       dd = (dlen[il])*(qdwt[iquad]);
463       zz = rquad[il][iquad];
464 
465       for (iq=0; iq < 2; iq++) {
466         ip = nli[il][iq];
467         bb = bspl(z,zz,il,iq,nli,1);
468         i = indx[ip];
469         if (i > -1) {
470           for (iqq=0; iqq < 2; iqq++) {
471             ipp = nli[il][iqq];
472             bbb = bspl(z,zz,il,iqq,nli,1);
473             j = indx[ipp];
474             aij = bb*bbb;
475             if (j > -1) {
476               add_term = aij*dd;
477               PetscCall(MatSetValue(obj->Amat,i,j,add_term,ADD_VALUES));
478             }/*endif*/
479           } /*end for (iqq)*/
480         } /*end if (i>0)*/
481       } /*end for (iq)*/
482     } /*end for (iquad)*/
483   } /*end for (il)*/
484   PetscCall(MatAssemblyBegin(obj->Amat,MAT_FINAL_ASSEMBLY));
485   PetscCall(MatAssemblyEnd(obj->Amat,MAT_FINAL_ASSEMBLY));
486   return 0;
487 }
488 
489 /*---------------------------------------------------------
490         Function to fill the rhs vector with
491         By + g values ****
492 ---------------------------------------------------------*/
493 PetscErrorCode rhs(AppCtx *obj,PetscScalar *y, PetscInt nz, PetscScalar *z, PetscReal t)
494 {
495   PetscInt       i,j,js,je,jj;
496   PetscScalar    val,g[num_z],btri[num_z][3],add_term;
497 
498   for (i=0; i < nz-2; i++) {
499     for (j=0; j <= 2; j++) btri[i][j]=0.0;
500     g[i] = 0.0;
501   }
502 
503   /*  call femBg to set the tri-diagonal b matrix and vector g  */
504   femBg(btri,g,nz,z,t);
505 
506   /*  setting the entries of the right hand side vector  */
507   for (i=0; i < nz-2; i++) {
508     val = 0.0;
509     js  = 0;
510     if (i == 0) js = 1;
511     je = 2;
512     if (i == nz-2) je = 1;
513 
514     for (jj=js; jj <= je; jj++) {
515       j    = i+jj-1;
516       val += (btri[i][jj])*(y[j]);
517     }
518     add_term = val + g[i];
519     PetscCall(VecSetValue(obj->ksp_rhs,(PetscInt)i,(PetscScalar)add_term,INSERT_VALUES));
520   }
521   PetscCall(VecAssemblyBegin(obj->ksp_rhs));
522   PetscCall(VecAssemblyEnd(obj->ksp_rhs));
523   return 0;
524 }
525 
526 /*%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
527 %%   Function to form the right hand side of the time-stepping problem.                       %%
528 %% -------------------------------------------------------------------------------------------%%
529   if (useAlhs):
530     globalout = By+g
531   else if (!useAlhs):
532     globalout = f(y,t)=Ainv(By+g),
533       in which the ksp solver to transform the problem A*ydot=By+g
534       to the problem ydot=f(y,t)=inv(A)*(By+g)
535 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%*/
536 
537 PetscErrorCode RHSfunction(TS ts,PetscReal t,Vec globalin,Vec globalout,void *ctx)
538 {
539   AppCtx            *obj = (AppCtx*)ctx;
540   PetscScalar       soln[num_z];
541   const PetscScalar *soln_ptr;
542   PetscInt          i,nz=obj->nz;
543   PetscReal         time;
544 
545   /* get the previous solution to compute updated system */
546   PetscCall(VecGetArrayRead(globalin,&soln_ptr));
547   for (i=0; i < num_z-2; i++) soln[i] = soln_ptr[i];
548   PetscCall(VecRestoreArrayRead(globalin,&soln_ptr));
549   soln[num_z-1] = 0.0;
550   soln[num_z-2] = 0.0;
551 
552   /* clear out the matrix and rhs for ksp to keep things straight */
553   PetscCall(VecSet(obj->ksp_rhs,(PetscScalar)0.0));
554 
555   time = t;
556   /* get the updated system */
557   rhs(obj,soln,nz,obj->z,time); /* setup of the By+g rhs */
558 
559   /* do a ksp solve to get the rhs for the ts problem */
560   if (obj->useAlhs) {
561     /* ksp_sol = ksp_rhs */
562     PetscCall(VecCopy(obj->ksp_rhs,globalout));
563   } else {
564     /* ksp_sol = inv(Amat)*ksp_rhs */
565     PetscCall(Petsc_KSPSolve(obj));
566     PetscCall(VecCopy(obj->ksp_sol,globalout));
567   }
568   return 0;
569 }
570 
571 /*TEST
572 
573     build:
574       requires: !complex
575 
576     test:
577       suffix: euler
578       output_file: output/ex3.out
579 
580     test:
581       suffix: 2
582       args:   -useAlhs
583       output_file: output/ex3.out
584       TODO: Broken because SNESComputeJacobianDefault is incompatible with TSComputeIJacobianConstant
585 
586 TEST*/
587