xref: /petsc/src/tao/bound/impls/bncg/bncg.c (revision 5000ebde5f407411787d5853133476ecf92a6e6c) 
1 #include <petsctaolinesearch.h>
2 #include <../src/tao/bound/impls/bncg/bncg.h>
3 
4 #define CG_FletcherReeves       0
5 #define CG_PolakRibiere         1
6 #define CG_PolakRibierePlus     2
7 #define CG_HestenesStiefel      3
8 #define CG_DaiYuan              4
9 #define CG_Types                5
10 
11 static const char *CG_Table[64] = {"fr", "pr", "prp", "hs", "dy"};
12 
13 #define CG_AS_NONE       0
14 #define CG_AS_BERTSEKAS  1
15 #define CG_AS_SIZE       2
16 
17 static const char *CG_AS_TYPE[64] = {"none", "bertsekas"};
18 
19 PetscErrorCode TaoBNCGSetRecycleFlag(Tao tao, PetscBool recycle)
20 {
21   TAO_BNCG                     *cg = (TAO_BNCG*)tao->data;
22 
23   PetscFunctionBegin;
24   cg->recycle = recycle;
25   PetscFunctionReturn(0);
26 }
27 
28 PetscErrorCode TaoBNCGEstimateActiveSet(Tao tao, PetscInt asType)
29 {
30   PetscErrorCode               ierr;
31   TAO_BNCG                     *cg = (TAO_BNCG *)tao->data;
32 
33   PetscFunctionBegin;
34   ierr = ISDestroy(&cg->inactive_old);CHKERRQ(ierr);
35   if (cg->inactive_idx) {
36     ierr = ISDuplicate(cg->inactive_idx, &cg->inactive_old);CHKERRQ(ierr);
37     ierr = ISCopy(cg->inactive_idx, cg->inactive_old);CHKERRQ(ierr);
38   }
39   switch (asType) {
40   case CG_AS_NONE:
41     ierr = ISDestroy(&cg->inactive_idx);CHKERRQ(ierr);
42     ierr = VecWhichInactive(tao->XL, tao->solution, cg->unprojected_gradient, tao->XU, PETSC_TRUE, &cg->inactive_idx);CHKERRQ(ierr);
43     ierr = ISDestroy(&cg->active_idx);CHKERRQ(ierr);
44     ierr = ISComplementVec(cg->inactive_idx, tao->solution, &cg->active_idx);CHKERRQ(ierr);
45     break;
46 
47   case CG_AS_BERTSEKAS:
48     /* Use gradient descent to estimate the active set */
49     ierr = VecCopy(cg->unprojected_gradient, cg->W);CHKERRQ(ierr);
50     ierr = VecScale(cg->W, -1.0);CHKERRQ(ierr);
51     ierr = TaoEstimateActiveBounds(tao->solution, tao->XL, tao->XU, cg->unprojected_gradient, cg->W, cg->as_step, &cg->as_tol, &cg->active_lower, &cg->active_upper, &cg->active_fixed, &cg->active_idx, &cg->inactive_idx);CHKERRQ(ierr);
52 
53   default:
54     break;
55   }
56   PetscFunctionReturn(0);
57 }
58 
59 PetscErrorCode TaoBNCGBoundStep(Tao tao, Vec step)
60 {
61   PetscErrorCode               ierr;
62   TAO_BNCG                     *cg = (TAO_BNCG *)tao->data;
63 
64   PetscFunctionBegin;
65   switch (cg->as_type) {
66   case CG_AS_NONE:
67     if (cg->active_idx) {ierr = VecISSet(step, cg->active_idx, 0.0);CHKERRQ(ierr);}
68     break;
69 
70   case CG_AS_BERTSEKAS:
71     ierr = TaoBoundStep(tao->solution, tao->XL, tao->XU, cg->active_lower, cg->active_upper, cg->active_fixed, step);CHKERRQ(ierr);
72     break;
73 
74   default:
75     break;
76   }
77   PetscFunctionReturn(0);
78 }
79 
80 static PetscErrorCode TaoSolve_BNCG(Tao tao)
81 {
82   TAO_BNCG                     *cg = (TAO_BNCG*)tao->data;
83   PetscErrorCode               ierr;
84   TaoLineSearchConvergedReason ls_status = TAOLINESEARCH_CONTINUE_ITERATING;
85   PetscReal                    step=1.0,gnorm,gnorm2,delta,gd,ginner,beta,dnorm,resnorm;
86   PetscReal                    gd_old,gnorm2_old,f_old;
87   PetscBool                    cg_restart;
88 
89   PetscFunctionBegin;
90   /*   Project the current point onto the feasible set */
91   ierr = TaoComputeVariableBounds(tao);CHKERRQ(ierr);
92   ierr = TaoLineSearchSetVariableBounds(tao->linesearch,tao->XL,tao->XU);CHKERRQ(ierr);
93 
94   /* Project the initial point onto the feasible region */
95   ierr = VecMedian(tao->XL,tao->solution,tao->XU,tao->solution);CHKERRQ(ierr);
96 
97   /*  Compute the objective function and criteria */
98   if (!cg->recycle) {
99     ierr = TaoComputeObjectiveAndGradient(tao, tao->solution, &cg->f, cg->unprojected_gradient);CHKERRQ(ierr);
100   }
101   ierr = VecNorm(cg->unprojected_gradient,NORM_2,&gnorm);CHKERRQ(ierr);
102   if (PetscIsInfOrNanReal(cg->f) || PetscIsInfOrNanReal(gnorm)) SETERRQ(PETSC_COMM_SELF,1, "User provided compute function generated Inf or NaN");
103 
104   /* Estimate the active set and compute the projected gradient */
105   ierr = VecCopy(cg->unprojected_gradient, cg->W);CHKERRQ(ierr);
106   ierr = VecScale(cg->W, -1.0);CHKERRQ(ierr);
107   ierr = TaoBNCGEstimateActiveSet(tao, cg->as_type);CHKERRQ(ierr);
108 
109   /* Project the gradient and calculate the norm */
110   ierr = VecCopy(cg->unprojected_gradient, tao->gradient);CHKERRQ(ierr);
111   ierr = VecISSet(tao->gradient, cg->active_idx, 0.0);CHKERRQ(ierr);
112   ierr = VecNorm(tao->gradient,NORM_2,&gnorm);CHKERRQ(ierr);
113   gnorm2 = gnorm*gnorm;
114 
115   /* Convergence check */
116   tao->niter = 0;
117   tao->reason = TAO_CONTINUE_ITERATING;
118   ierr = VecFischer(tao->solution, cg->unprojected_gradient, tao->XL, tao->XU, cg->W);CHKERRQ(ierr);
119   ierr = VecNorm(cg->W, NORM_2, &resnorm);CHKERRQ(ierr);
120   ierr = TaoLogConvergenceHistory(tao, cg->f, resnorm, 0.0, tao->ksp_its);CHKERRQ(ierr);
121   ierr = TaoMonitor(tao, tao->niter, cg->f, resnorm, 0.0, step);CHKERRQ(ierr);
122   ierr = (*tao->ops->convergencetest)(tao,tao->cnvP);CHKERRQ(ierr);
123   if (tao->reason != TAO_CONTINUE_ITERATING) PetscFunctionReturn(0);
124 
125   /* Start optimization iterations */
126   cg->ls_fails = cg->broken_ortho = cg->descent_error = 0;
127   cg->resets = -1;
128   while (tao->reason == TAO_CONTINUE_ITERATING) {
129     /* Check restart conditions for using steepest descent */
130     cg_restart = PETSC_FALSE;
131     ierr = VecDot(tao->gradient, cg->G_old, &ginner);CHKERRQ(ierr);
132     ierr = VecNorm(tao->stepdirection, NORM_2, &dnorm);CHKERRQ(ierr);
133     if (tao->niter == 0 && !cg->recycle && dnorm != 0.0) {
134       /* 1) First iteration, with recycle disabled, and a non-zero previous step */
135       cg_restart = PETSC_TRUE;
136     } else if (PetscAbsScalar(ginner) >= cg->eta * gnorm2) {
137       /* 2) Gradients are far from orthogonal */
138       cg_restart = PETSC_TRUE;
139       cg->broken_ortho++;
140     }
141 
142     /* Compute CG step */
143     if (cg_restart) {
144       beta = 0.0;
145       cg->resets++;
146     } else {
147       switch (cg->cg_type) {
148       case CG_FletcherReeves:
149         beta = gnorm2 / gnorm2_old;
150         break;
151 
152       case CG_PolakRibiere:
153         beta = (gnorm2 - ginner) / gnorm2_old;
154         break;
155 
156       case CG_PolakRibierePlus:
157         beta = PetscMax((gnorm2-ginner)/gnorm2_old, 0.0);
158         break;
159 
160       case CG_HestenesStiefel:
161         ierr = VecDot(tao->gradient, tao->stepdirection, &gd);CHKERRQ(ierr);
162         ierr = VecDot(cg->G_old, tao->stepdirection, &gd_old);CHKERRQ(ierr);
163         beta = (gnorm2 - ginner) / (gd - gd_old);
164         break;
165 
166       case CG_DaiYuan:
167         ierr = VecDot(tao->gradient, tao->stepdirection, &gd);CHKERRQ(ierr);
168         ierr = VecDot(cg->G_old, tao->stepdirection, &gd_old);CHKERRQ(ierr);
169         beta = gnorm2 / (gd - gd_old);
170         break;
171 
172       default:
173         beta = 0.0;
174         break;
175       }
176     }
177 
178     /*  Compute the direction d=-g + beta*d */
179     ierr = VecAXPBY(tao->stepdirection, -1.0, beta, tao->gradient);CHKERRQ(ierr);
180     ierr = TaoBNCGBoundStep(tao, tao->stepdirection);CHKERRQ(ierr);
181     if (cg->inactive_old) {
182       ierr = ISDestroy(&cg->new_inactives);CHKERRQ(ierr);
183       ierr = ISDifference(cg->inactive_old, cg->inactive_idx, &cg->new_inactives);
184       ierr = VecGetSubVector(tao->stepdirection, cg->new_inactives, &cg->inactive_step);CHKERRQ(ierr);
185       ierr = VecGetSubVector(tao->gradient, cg->new_inactives, &cg->inactive_grad);CHKERRQ(ierr);
186       ierr = VecCopy(cg->inactive_grad, cg->inactive_step);CHKERRQ(ierr);
187       ierr = VecScale(cg->inactive_step, -1.0);CHKERRQ(ierr);
188       ierr = VecRestoreSubVector(tao->stepdirection, cg->new_inactives, &cg->inactive_step);CHKERRQ(ierr);
189       ierr = VecRestoreSubVector(tao->gradient, cg->new_inactives, &cg->inactive_grad);CHKERRQ(ierr);
190     }
191 
192     /* Verify that this is a descent direction */
193     ierr = VecDot(tao->gradient, tao->stepdirection, &gd);CHKERRQ(ierr);
194     ierr = VecNorm(tao->stepdirection, NORM_2, &dnorm);
195     if (gd > -cg->rho*PetscPowReal(dnorm, cg->pow)) {
196       /* Not a descent direction, so we reset back to projected gradient descent */
197       ierr = VecAXPBY(tao->stepdirection, -1.0, 0.0, tao->gradient);CHKERRQ(ierr);
198       cg->resets++;
199       cg->descent_error++;
200     }
201 
202     /*  update initial steplength choice */
203     delta = 1.0;
204     delta = PetscMax(delta, cg->delta_min);
205     delta = PetscMin(delta, cg->delta_max);
206 
207     /* Store solution and gradient info before it changes */
208     ierr = VecCopy(tao->solution, cg->X_old);CHKERRQ(ierr);
209     ierr = VecCopy(tao->gradient, cg->G_old);CHKERRQ(ierr);
210     ierr = VecCopy(cg->unprojected_gradient, cg->unprojected_gradient_old);CHKERRQ(ierr);
211     gnorm2_old = gnorm2;
212     f_old = cg->f;
213 
214     /* Perform bounded line search */
215     ierr = TaoLineSearchSetInitialStepLength(tao->linesearch,delta);CHKERRQ(ierr);
216     ierr = TaoLineSearchApply(tao->linesearch, tao->solution, &cg->f, cg->unprojected_gradient, tao->stepdirection, &step, &ls_status);CHKERRQ(ierr);
217     ierr = TaoAddLineSearchCounts(tao);CHKERRQ(ierr);
218 
219     /*  Check linesearch failure */
220     if (ls_status != TAOLINESEARCH_SUCCESS && ls_status != TAOLINESEARCH_SUCCESS_USER) {
221       cg->ls_fails++;
222       /* Restore previous point */
223       gnorm2 = gnorm2_old;
224       cg->f = f_old;
225       ierr = VecCopy(cg->X_old, tao->solution);CHKERRQ(ierr);
226       ierr = VecCopy(cg->G_old, tao->gradient);CHKERRQ(ierr);
227       ierr = VecCopy(cg->unprojected_gradient_old, cg->unprojected_gradient);CHKERRQ(ierr);
228 
229       /* Fall back on the unscaled gradient step */
230       delta = 1.0;
231       ierr = VecCopy(tao->gradient, tao->stepdirection);CHKERRQ(ierr);
232       ierr = VecScale(tao->stepdirection, -1.0);CHKERRQ(ierr);
233       ierr = TaoBoundStep(tao->solution, tao->XL, tao->XU, cg->active_lower, cg->active_upper, cg->active_fixed, tao->stepdirection);CHKERRQ(ierr);
234 
235       ierr = TaoLineSearchSetInitialStepLength(tao->linesearch,delta);CHKERRQ(ierr);
236       ierr = TaoLineSearchApply(tao->linesearch, tao->solution, &cg->f, cg->unprojected_gradient, tao->stepdirection, &step, &ls_status);CHKERRQ(ierr);
237       ierr = TaoAddLineSearchCounts(tao);CHKERRQ(ierr);
238 
239       if (ls_status != TAOLINESEARCH_SUCCESS && ls_status != TAOLINESEARCH_SUCCESS_USER){
240         cg->ls_fails++;
241         /* Restore previous point */
242         gnorm2 = gnorm2_old;
243         cg->f = f_old;
244         ierr = VecCopy(cg->X_old, tao->solution);CHKERRQ(ierr);
245         ierr = VecCopy(cg->G_old, tao->gradient);CHKERRQ(ierr);
246         ierr = VecCopy(cg->unprojected_gradient_old, cg->unprojected_gradient);CHKERRQ(ierr);
247 
248         /* Nothing left to do but fail out of the optimization */
249         step = 0.0;
250         tao->reason = TAO_DIVERGED_LS_FAILURE;
251       }
252     }
253 
254     /* Estimate the active set at the new solution */
255     ierr = TaoBNCGEstimateActiveSet(tao, cg->as_type);CHKERRQ(ierr);
256 
257     /* Compute the projected gradient and its norm */
258     ierr = VecCopy(cg->unprojected_gradient, tao->gradient);CHKERRQ(ierr);
259     ierr = VecISSet(tao->gradient, cg->active_idx, 0.0);CHKERRQ(ierr);
260     ierr = VecNorm(tao->gradient,NORM_2,&gnorm);CHKERRQ(ierr);
261     gnorm2 = gnorm*gnorm;
262 
263     /* Convergence test */
264     tao->niter++;
265     ierr = VecFischer(tao->solution, cg->unprojected_gradient, tao->XL, tao->XU, cg->W);CHKERRQ(ierr);
266     ierr = VecNorm(cg->W, NORM_2, &resnorm);CHKERRQ(ierr);
267     ierr = TaoLogConvergenceHistory(tao, cg->f, resnorm, 0.0, tao->ksp_its);CHKERRQ(ierr);
268     ierr = TaoMonitor(tao, tao->niter, cg->f, resnorm, 0.0, step);CHKERRQ(ierr);
269     ierr = (*tao->ops->convergencetest)(tao,tao->cnvP);CHKERRQ(ierr);
270   }
271   PetscFunctionReturn(0);
272 }
273 
274 static PetscErrorCode TaoSetUp_BNCG(Tao tao)
275 {
276   TAO_BNCG         *cg = (TAO_BNCG*)tao->data;
277   PetscErrorCode ierr;
278 
279   PetscFunctionBegin;
280   if (!tao->gradient) {ierr = VecDuplicate(tao->solution,&tao->gradient);CHKERRQ(ierr);}
281   if (!tao->stepdirection) {ierr = VecDuplicate(tao->solution,&tao->stepdirection);CHKERRQ(ierr);}
282   if (!cg->W) {ierr = VecDuplicate(tao->solution,&cg->W);CHKERRQ(ierr);}
283   if (!cg->X_old) {ierr = VecDuplicate(tao->solution,&cg->X_old);CHKERRQ(ierr);}
284   if (!cg->G_old) {ierr = VecDuplicate(tao->gradient,&cg->G_old);CHKERRQ(ierr);}
285   if (!cg->unprojected_gradient) {ierr = VecDuplicate(tao->gradient,&cg->unprojected_gradient);CHKERRQ(ierr);}
286   if (!cg->unprojected_gradient_old) {ierr = VecDuplicate(tao->gradient,&cg->unprojected_gradient_old);CHKERRQ(ierr);}
287   PetscFunctionReturn(0);
288 }
289 
290 static PetscErrorCode TaoDestroy_BNCG(Tao tao)
291 {
292   TAO_BNCG       *cg = (TAO_BNCG*) tao->data;
293   PetscErrorCode ierr;
294 
295   PetscFunctionBegin;
296   if (tao->setupcalled) {
297     ierr = VecDestroy(&cg->W);CHKERRQ(ierr);
298     ierr = VecDestroy(&cg->X_old);CHKERRQ(ierr);
299     ierr = VecDestroy(&cg->G_old);CHKERRQ(ierr);
300     ierr = VecDestroy(&cg->unprojected_gradient);CHKERRQ(ierr);
301     ierr = VecDestroy(&cg->unprojected_gradient_old);CHKERRQ(ierr);
302   }
303   ierr = PetscFree(tao->data);CHKERRQ(ierr);
304   PetscFunctionReturn(0);
305 }
306 
307 static PetscErrorCode TaoSetFromOptions_BNCG(PetscOptionItems *PetscOptionsObject,Tao tao)
308  {
309     TAO_BNCG       *cg = (TAO_BNCG*)tao->data;
310     PetscErrorCode ierr;
311 
312     PetscFunctionBegin;
313     ierr = TaoLineSearchSetFromOptions(tao->linesearch);CHKERRQ(ierr);
314     ierr = PetscOptionsHead(PetscOptionsObject,"Nonlinear Conjugate Gradient method for unconstrained optimization");CHKERRQ(ierr);
315     ierr = PetscOptionsReal("-tao_bncg_eta","restart tolerance", "", cg->eta,&cg->eta,NULL);CHKERRQ(ierr);
316     ierr = PetscOptionsReal("-tao_bncg_rho","descent direction tolerance", "", cg->rho,&cg->rho,NULL);CHKERRQ(ierr);
317     ierr = PetscOptionsReal("-tao_bncg_pow","descent direction exponent", "", cg->pow,&cg->pow,NULL);CHKERRQ(ierr);
318     ierr = PetscOptionsEList("-tao_bncg_type","cg formula", "", CG_Table, CG_Types, CG_Table[cg->cg_type], &cg->cg_type,NULL);CHKERRQ(ierr);
319     ierr = PetscOptionsEList("-tao_bncg_as_type","active set estimation method", "", CG_AS_TYPE, CG_AS_SIZE, CG_AS_TYPE[cg->cg_type], &cg->cg_type,NULL);CHKERRQ(ierr);
320     ierr = PetscOptionsReal("-tao_bncg_delta_min","minimum delta value", "", cg->delta_min,&cg->delta_min,NULL);CHKERRQ(ierr);
321     ierr = PetscOptionsReal("-tao_bncg_delta_max","maximum delta value", "", cg->delta_max,&cg->delta_max,NULL);CHKERRQ(ierr);
322     ierr = PetscOptionsBool("-tao_bncg_recycle","enable recycling the existing solution and gradient at the start of a new solve","",cg->recycle,&cg->recycle,NULL);CHKERRQ(ierr);
323     ierr = PetscOptionsReal("-tao_bncg_as_tol", "initial tolerance used when estimating actively bounded variables", "", cg->as_tol, &cg->as_tol,NULL);CHKERRQ(ierr);
324     ierr = PetscOptionsReal("-tao_bncg_as_step", "step length used when estimating actively bounded variables", "", cg->as_step, &cg->as_step,NULL);CHKERRQ(ierr);
325    ierr = PetscOptionsTail();CHKERRQ(ierr);
326    PetscFunctionReturn(0);
327 }
328 
329 static PetscErrorCode TaoView_BNCG(Tao tao, PetscViewer viewer)
330 {
331   PetscBool      isascii;
332   TAO_BNCG       *cg = (TAO_BNCG*)tao->data;
333   PetscErrorCode ierr;
334 
335   PetscFunctionBegin;
336   ierr = PetscObjectTypeCompare((PetscObject)viewer, PETSCVIEWERASCII, &isascii);CHKERRQ(ierr);
337   if (isascii) {
338     ierr = PetscViewerASCIIPushTab(viewer);CHKERRQ(ierr);
339     ierr = PetscViewerASCIIPrintf(viewer, "CG Type: %s\n", CG_Table[cg->cg_type]);CHKERRQ(ierr);
340     ierr = PetscViewerASCIIPrintf(viewer, "Resets: %i\n", cg->resets);CHKERRQ(ierr);
341     ierr = PetscViewerASCIIPrintf(viewer, "  Broken ortho: %i\n", cg->broken_ortho);CHKERRQ(ierr);
342     ierr = PetscViewerASCIIPrintf(viewer, "  Not a descent dir.: %i\n", cg->descent_error);CHKERRQ(ierr);
343     ierr = PetscViewerASCIIPrintf(viewer, "Line search fails: %i\n", cg->ls_fails);CHKERRQ(ierr);
344     ierr = PetscViewerASCIIPopTab(viewer);CHKERRQ(ierr);
345   }
346   PetscFunctionReturn(0);
347 }
348 
349 /*MC
350      TAOBNCG -   Bound-constrained Nonlinear Conjugate Gradient method.
351 
352    Options Database Keys:
353 +      -tao_bncg_eta <r> - restart tolerance
354 .      -tao_bncg_type <taocg_type> - cg formula
355 .      -tao_bncg_delta_min <r> - minimum delta value
356 -      -tao_bncg_delta_max <r> - maximum delta value
357 
358   Notes:
359      CG formulas are:
360          "fr" - Fletcher-Reeves
361          "pr" - Polak-Ribiere
362          "prp" - Polak-Ribiere-Plus
363          "hs" - Hestenes-Steifel
364          "dy" - Dai-Yuan
365   Level: beginner
366 M*/
367 
368 
369 PETSC_EXTERN PetscErrorCode TaoCreate_BNCG(Tao tao)
370 {
371   TAO_BNCG       *cg;
372   const char     *morethuente_type = TAOLINESEARCHMT;
373   PetscErrorCode ierr;
374 
375   PetscFunctionBegin;
376   tao->ops->setup = TaoSetUp_BNCG;
377   tao->ops->solve = TaoSolve_BNCG;
378   tao->ops->view = TaoView_BNCG;
379   tao->ops->setfromoptions = TaoSetFromOptions_BNCG;
380   tao->ops->destroy = TaoDestroy_BNCG;
381 
382   /* Override default settings (unless already changed) */
383   if (!tao->max_it_changed) tao->max_it = 2000;
384   if (!tao->max_funcs_changed) tao->max_funcs = 4000;
385 
386   /*  Note: nondefault values should be used for nonlinear conjugate gradient  */
387   /*  method.  In particular, gtol should be less that 0.5; the value used in  */
388   /*  Nocedal and Wright is 0.10.  We use the default values for the  */
389   /*  linesearch because it seems to work better. */
390   ierr = TaoLineSearchCreate(((PetscObject)tao)->comm, &tao->linesearch);CHKERRQ(ierr);
391   ierr = PetscObjectIncrementTabLevel((PetscObject)tao->linesearch, (PetscObject)tao, 1);CHKERRQ(ierr);
392   ierr = TaoLineSearchSetType(tao->linesearch, morethuente_type);CHKERRQ(ierr);
393   ierr = TaoLineSearchUseTaoRoutines(tao->linesearch, tao);CHKERRQ(ierr);
394   ierr = TaoLineSearchSetOptionsPrefix(tao->linesearch,tao->hdr.prefix);CHKERRQ(ierr);
395 
396   ierr = PetscNewLog(tao,&cg);CHKERRQ(ierr);
397   tao->data = (void*)cg;
398   cg->rho = 1e-4;
399   cg->pow = 2.1;
400   cg->eta = 0.5;
401   cg->delta_min = 1e-7;
402   cg->delta_max = 100;
403   cg->as_step = 0.001;
404   cg->as_tol = 0.001;
405   cg->as_type = CG_AS_BERTSEKAS;
406   cg->cg_type = CG_DaiYuan;
407   cg->recycle = PETSC_FALSE;
408   PetscFunctionReturn(0);
409 }
410