xref: /petsc/src/ksp/pc/impls/mg/mg.c (revision 58cd72c3d0f0c5a63e84afc03ac54bb1bb84cd8d)
1 #define PETSCKSP_DLL
2 
3 /*
4     Defines the multigrid preconditioner interface.
5 */
6 #include "src/ksp/pc/impls/mg/mgimpl.h"                    /*I "petscmg.h" I*/
7 
8 
9 #undef __FUNCT__
10 #define __FUNCT__ "PCMGMCycle_Private"
11 PetscErrorCode PCMGMCycle_Private(PC_MG **mglevels,PetscTruth *converged)
12 {
13   PC_MG          *mg = *mglevels,*mgc;
14   PetscErrorCode ierr;
15   PetscInt       cycles = mg->cycles;
16 
17   PetscFunctionBegin;
18   if (converged) *converged = PETSC_FALSE;
19 
20   if (mg->eventsolve) {ierr = PetscLogEventBegin(mg->eventsolve,0,0,0,0);CHKERRQ(ierr);}
21   ierr = KSPSolve(mg->smoothd,mg->b,mg->x);CHKERRQ(ierr);
22   if (mg->eventsolve) {ierr = PetscLogEventEnd(mg->eventsolve,0,0,0,0);CHKERRQ(ierr);}
23   if (mg->level) {  /* not the coarsest grid */
24     ierr = (*mg->residual)(mg->A,mg->b,mg->x,mg->r);CHKERRQ(ierr);
25 
26     /* if on finest level and have convergence criteria set */
27     if (mg->level == mg->levels-1 && mg->ttol) {
28       PetscReal rnorm;
29       ierr = VecNorm(mg->r,NORM_2,&rnorm);CHKERRQ(ierr);
30       if (rnorm <= mg->ttol) {
31         *converged = PETSC_TRUE;
32         if (rnorm < mg->abstol) {
33           ierr = PetscInfo2(0,"Linear solver has converged. Residual norm %G is less than absolute tolerance %G\n",rnorm,mg->abstol);CHKERRQ(ierr);
34         } else {
35           ierr = PetscInfo2(0,"Linear solver has converged. Residual norm %G is less than relative tolerance times initial residual norm %G\n",rnorm,mg->ttol);CHKERRQ(ierr);
36         }
37         PetscFunctionReturn(0);
38       }
39     }
40 
41     mgc = *(mglevels - 1);
42     ierr = MatRestrict(mg->restrct,mg->r,mgc->b);CHKERRQ(ierr);
43     ierr = VecSet(mgc->x,0.0);CHKERRQ(ierr);
44     while (cycles--) {
45       ierr = PCMGMCycle_Private(mglevels-1,converged);CHKERRQ(ierr);
46     }
47     ierr = MatInterpolateAdd(mg->interpolate,mgc->x,mg->x,mg->x);CHKERRQ(ierr);
48     if (mg->eventsolve) {ierr = PetscLogEventBegin(mg->eventsolve,0,0,0,0);CHKERRQ(ierr);}
49     ierr = KSPSolve(mg->smoothu,mg->b,mg->x);CHKERRQ(ierr);
50     if (mg->eventsolve) {ierr = PetscLogEventEnd(mg->eventsolve,0,0,0,0);CHKERRQ(ierr);}
51   }
52   PetscFunctionReturn(0);
53 }
54 
55 /*
56        PCMGCreate_Private - Creates a PC_MG structure for use with the
57                multigrid code. Level 0 is the coarsest. (But the
58                finest level is stored first in the array).
59 
60 */
61 #undef __FUNCT__
62 #define __FUNCT__ "PCMGCreate_Private"
63 static PetscErrorCode PCMGCreate_Private(MPI_Comm comm,PetscInt levels,PC pc,MPI_Comm *comms,PC_MG ***result)
64 {
65   PC_MG          **mg;
66   PetscErrorCode ierr;
67   PetscInt       i;
68   PetscMPIInt    size;
69   const char     *prefix;
70   PC             ipc;
71 
72   PetscFunctionBegin;
73   ierr = PetscMalloc(levels*sizeof(PC_MG*),&mg);CHKERRQ(ierr);
74   ierr = PetscLogObjectMemory(pc,levels*(sizeof(PC_MG*)+sizeof(PC_MG)));CHKERRQ(ierr);
75 
76   ierr = PCGetOptionsPrefix(pc,&prefix);CHKERRQ(ierr);
77 
78   for (i=0; i<levels; i++) {
79     ierr = PetscNew(PC_MG,&mg[i]);CHKERRQ(ierr);
80     mg[i]->level           = i;
81     mg[i]->levels          = levels;
82     mg[i]->cycles          = 1;
83     mg[i]->galerkin        = PETSC_FALSE;
84     mg[i]->galerkinused    = PETSC_FALSE;
85     mg[i]->default_smoothu = 1;
86     mg[i]->default_smoothd = 1;
87 
88     if (comms) comm = comms[i];
89     ierr = KSPCreate(comm,&mg[i]->smoothd);CHKERRQ(ierr);
90     ierr = KSPSetTolerances(mg[i]->smoothd,PETSC_DEFAULT,PETSC_DEFAULT,PETSC_DEFAULT, mg[i]->default_smoothd);CHKERRQ(ierr);
91     ierr = KSPSetOptionsPrefix(mg[i]->smoothd,prefix);CHKERRQ(ierr);
92 
93     /* do special stuff for coarse grid */
94     if (!i && levels > 1) {
95       ierr = KSPAppendOptionsPrefix(mg[0]->smoothd,"mg_coarse_");CHKERRQ(ierr);
96 
97       /* coarse solve is (redundant) LU by default */
98       ierr = KSPSetType(mg[0]->smoothd,KSPPREONLY);CHKERRQ(ierr);
99       ierr = KSPGetPC(mg[0]->smoothd,&ipc);CHKERRQ(ierr);
100       ierr = MPI_Comm_size(comm,&size);CHKERRQ(ierr);
101       if (size > 1) {
102         ierr = PCSetType(ipc,PCREDUNDANT);CHKERRQ(ierr);
103         ierr = PCRedundantGetPC(ipc,&ipc);CHKERRQ(ierr);
104       }
105       ierr = PCSetType(ipc,PCLU);CHKERRQ(ierr);
106 
107     } else {
108       char tprefix[128];
109       sprintf(tprefix,"mg_levels_%d_",(int)i);
110       ierr = KSPAppendOptionsPrefix(mg[i]->smoothd,tprefix);CHKERRQ(ierr);
111     }
112     ierr = PetscLogObjectParent(pc,mg[i]->smoothd);CHKERRQ(ierr);
113     mg[i]->smoothu         = mg[i]->smoothd;
114     mg[i]->rtol = 0.0;
115     mg[i]->abstol = 0.0;
116     mg[i]->dtol = 0.0;
117     mg[i]->ttol = 0.0;
118     mg[i]->eventsetup = 0;
119     mg[i]->eventsolve = 0;
120   }
121   *result = mg;
122   PetscFunctionReturn(0);
123 }
124 
125 #undef __FUNCT__
126 #define __FUNCT__ "PCDestroy_MG"
127 static PetscErrorCode PCDestroy_MG(PC pc)
128 {
129   PC_MG          **mg = (PC_MG**)pc->data;
130   PetscErrorCode ierr;
131   PetscInt       i,n = mg[0]->levels;
132 
133   PetscFunctionBegin;
134   for (i=0; i<n-1; i++) {
135     if (mg[i+1]->r) {ierr = VecDestroy(mg[i+1]->r);CHKERRQ(ierr);}
136     if (mg[i]->b) {ierr = VecDestroy(mg[i]->b);CHKERRQ(ierr);}
137     if (mg[i]->x) {ierr = VecDestroy(mg[i]->x);CHKERRQ(ierr);}
138     if (mg[i+1]->restrct) {ierr = MatDestroy(mg[i+1]->restrct);CHKERRQ(ierr);}
139     if (mg[i+1]->interpolate) {ierr = MatDestroy(mg[i+1]->interpolate);CHKERRQ(ierr);}
140   }
141 
142   for (i=0; i<n; i++) {
143     if (mg[i]->smoothd != mg[i]->smoothu) {
144       ierr = KSPDestroy(mg[i]->smoothd);CHKERRQ(ierr);
145     }
146     ierr = KSPDestroy(mg[i]->smoothu);CHKERRQ(ierr);
147     ierr = PetscFree(mg[i]);CHKERRQ(ierr);
148   }
149   ierr = PetscFree(mg);CHKERRQ(ierr);
150   PetscFunctionReturn(0);
151 }
152 
153 
154 
155 EXTERN PetscErrorCode PCMGACycle_Private(PC_MG**);
156 EXTERN PetscErrorCode PCMGFCycle_Private(PC_MG**);
157 EXTERN PetscErrorCode PCMGKCycle_Private(PC_MG**);
158 
159 /*
160    PCApply_MG - Runs either an additive, multiplicative, Kaskadic
161              or full cycle of multigrid.
162 
163   Note:
164   A simple wrapper which calls PCMGMCycle(),PCMGACycle(), or PCMGFCycle().
165 */
166 #undef __FUNCT__
167 #define __FUNCT__ "PCApply_MG"
168 static PetscErrorCode PCApply_MG(PC pc,Vec b,Vec x)
169 {
170   PC_MG          **mg = (PC_MG**)pc->data;
171   PetscErrorCode ierr;
172   PetscInt       levels = mg[0]->levels;
173 
174   PetscFunctionBegin;
175   mg[levels-1]->b = b;
176   mg[levels-1]->x = x;
177   if (!mg[levels-1]->r && mg[0]->am != PC_MG_ADDITIVE && levels > 1) {
178     Vec tvec;
179     ierr = VecDuplicate(mg[levels-1]->b,&tvec);CHKERRQ(ierr);
180     ierr = PCMGSetR(pc,levels-1,tvec);CHKERRQ(ierr);
181     ierr = VecDestroy(tvec);CHKERRQ(ierr);
182   }
183   if (mg[0]->am == PC_MG_MULTIPLICATIVE) {
184     ierr = VecSet(x,0.0);CHKERRQ(ierr);
185     ierr = PCMGMCycle_Private(mg+levels-1,PETSC_NULL);CHKERRQ(ierr);
186   }
187   else if (mg[0]->am == PC_MG_ADDITIVE) {
188     ierr = PCMGACycle_Private(mg);CHKERRQ(ierr);
189   }
190   else if (mg[0]->am == PC_MG_KASKADE) {
191     ierr = PCMGKCycle_Private(mg);CHKERRQ(ierr);
192   }
193   else {
194     ierr = PCMGFCycle_Private(mg);CHKERRQ(ierr);
195   }
196   PetscFunctionReturn(0);
197 }
198 
199 #undef __FUNCT__
200 #define __FUNCT__ "PCApplyRichardson_MG"
201 static PetscErrorCode PCApplyRichardson_MG(PC pc,Vec b,Vec x,Vec w,PetscReal rtol,PetscReal abstol, PetscReal dtol,PetscInt its)
202 {
203   PC_MG          **mg = (PC_MG**)pc->data;
204   PetscErrorCode ierr;
205   PetscInt       levels = mg[0]->levels;
206   PetscTruth     converged = PETSC_FALSE;
207 
208   PetscFunctionBegin;
209   mg[levels-1]->b    = b;
210   mg[levels-1]->x    = x;
211 
212   mg[levels-1]->rtol = rtol;
213   mg[levels-1]->abstol = abstol;
214   mg[levels-1]->dtol = dtol;
215   if (rtol) {
216     /* compute initial residual norm for relative convergence test */
217     PetscReal rnorm;
218     ierr               = (*mg[levels-1]->residual)(mg[levels-1]->A,b,x,w);CHKERRQ(ierr);
219     ierr               = VecNorm(w,NORM_2,&rnorm);CHKERRQ(ierr);
220     mg[levels-1]->ttol = PetscMax(rtol*rnorm,abstol);
221   } else if (abstol) {
222     mg[levels-1]->ttol = abstol;
223   } else {
224     mg[levels-1]->ttol = 0.0;
225   }
226 
227   while (its-- && !converged) {
228     ierr = PCMGMCycle_Private(mg+levels-1,&converged);CHKERRQ(ierr);
229   }
230   PetscFunctionReturn(0);
231 }
232 
233 #undef __FUNCT__
234 #define __FUNCT__ "PCSetFromOptions_MG"
235 PetscErrorCode PCSetFromOptions_MG(PC pc)
236 {
237   PetscErrorCode ierr;
238   PetscInt       m,levels = 1;
239   PetscTruth     flg;
240   PC_MG          **mg = (PC_MG**)pc->data;
241   PCMGType       mgtype;
242 
243   PetscFunctionBegin;
244 
245   ierr = PetscOptionsHead("Multigrid options");CHKERRQ(ierr);
246     if (!pc->data) {
247       ierr = PetscOptionsInt("-pc_mg_levels","Number of Levels","PCMGSetLevels",levels,&levels,&flg);CHKERRQ(ierr);
248       ierr = PCMGSetLevels(pc,levels,PETSC_NULL);CHKERRQ(ierr);
249       mg = (PC_MG**)pc->data;
250     }
251     mgtype = mg[0]->am;
252     ierr = PetscOptionsInt("-pc_mg_cycles","1 for V cycle, 2 for W-cycle","PCMGSetCycles",1,&m,&flg);CHKERRQ(ierr);
253     if (flg) {
254       ierr = PCMGSetCycles(pc,m);CHKERRQ(ierr);
255     }
256     ierr = PetscOptionsName("-pc_mg_galerkin","Use Galerkin process to compute coarser operators","PCMGSetGalerkin",&flg);CHKERRQ(ierr);
257     if (flg) {
258       ierr = PCMGSetGalerkin(pc);CHKERRQ(ierr);
259     }
260     ierr = PetscOptionsInt("-pc_mg_smoothup","Number of post-smoothing steps","PCMGSetNumberSmoothUp",1,&m,&flg);CHKERRQ(ierr);
261     if (flg) {
262       ierr = PCMGSetNumberSmoothUp(pc,m);CHKERRQ(ierr);
263     }
264     ierr = PetscOptionsInt("-pc_mg_smoothdown","Number of pre-smoothing steps","PCMGSetNumberSmoothDown",1,&m,&flg);CHKERRQ(ierr);
265     if (flg) {
266       ierr = PCMGSetNumberSmoothDown(pc,m);CHKERRQ(ierr);
267     }
268     ierr = PetscOptionsEnum("-pc_mg_type","Multigrid type","PCMGSetType",PCMGTypes,(PetscEnum)mgtype,(PetscEnum*)&mgtype,&flg);CHKERRQ(ierr);
269     if (flg) {ierr = PCMGSetType(pc,mgtype);CHKERRQ(ierr);}
270     ierr = PetscOptionsName("-pc_mg_log","Log times for each multigrid level","None",&flg);CHKERRQ(ierr);
271     if (flg) {
272       PetscInt i;
273       char     eventname[128];
274       if (!mg) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Must set MG levels before calling");
275       levels = mg[0]->levels;
276       for (i=0; i<levels; i++) {
277         sprintf(eventname,"MSetup Level %d",(int)i);
278         ierr = PetscLogEventRegister(&mg[i]->eventsetup,eventname,pc->cookie);CHKERRQ(ierr);
279         sprintf(eventname,"MGSolve Level %d to 0",(int)i);
280         ierr = PetscLogEventRegister(&mg[i]->eventsolve,eventname,pc->cookie);CHKERRQ(ierr);
281       }
282     }
283   ierr = PetscOptionsTail();CHKERRQ(ierr);
284   PetscFunctionReturn(0);
285 }
286 
287 const char *PCMGTypes[] = {"MULTIPLICATIVE","ADDITIVE","FULL","KASKADE","PCMGType","PC_MG",0};
288 
289 #undef __FUNCT__
290 #define __FUNCT__ "PCView_MG"
291 static PetscErrorCode PCView_MG(PC pc,PetscViewer viewer)
292 {
293   PC_MG          **mg = (PC_MG**)pc->data;
294   PetscErrorCode ierr;
295   PetscInt       levels = mg[0]->levels,i;
296   PetscTruth     iascii;
297 
298   PetscFunctionBegin;
299   ierr = PetscTypeCompare((PetscObject)viewer,PETSC_VIEWER_ASCII,&iascii);CHKERRQ(ierr);
300   if (iascii) {
301     ierr = PetscViewerASCIIPrintf(viewer,"  MG: type is %s, levels=%D cycles=%D, pre-smooths=%D, post-smooths=%D\n",
302                       PCMGTypes[mg[0]->am],levels,mg[0]->cycles,mg[0]->default_smoothd,mg[0]->default_smoothu);CHKERRQ(ierr);
303     if (mg[0]->galerkin) {
304       ierr = PetscViewerASCIIPrintf(viewer,"    Using Galerkin computed coarse grid matrices\n");CHKERRQ(ierr);
305     }
306     for (i=0; i<levels; i++) {
307       if (!i) {
308         ierr = PetscViewerASCIIPrintf(viewer,"Coarse gride solver -- level %D -------------------------------\n",i);CHKERRQ(ierr);
309       } else {
310         ierr = PetscViewerASCIIPrintf(viewer,"Down solver (pre-smoother) on level %D -------------------------------\n",i);CHKERRQ(ierr);
311       }
312       ierr = PetscViewerASCIIPushTab(viewer);CHKERRQ(ierr);
313       ierr = KSPView(mg[i]->smoothd,viewer);CHKERRQ(ierr);
314       ierr = PetscViewerASCIIPopTab(viewer);CHKERRQ(ierr);
315       if (i && mg[i]->smoothd == mg[i]->smoothu) {
316         ierr = PetscViewerASCIIPrintf(viewer,"Up solver (post-smoother) same as down solver (pre-smoother)\n");CHKERRQ(ierr);
317       } else if (i){
318         ierr = PetscViewerASCIIPrintf(viewer,"Up solver (post-smoother) on level %D -------------------------------\n",i);CHKERRQ(ierr);
319         ierr = PetscViewerASCIIPushTab(viewer);CHKERRQ(ierr);
320         ierr = KSPView(mg[i]->smoothu,viewer);CHKERRQ(ierr);
321         ierr = PetscViewerASCIIPopTab(viewer);CHKERRQ(ierr);
322       }
323     }
324   } else {
325     SETERRQ1(PETSC_ERR_SUP,"Viewer type %s not supported for PCMG",((PetscObject)viewer)->type_name);
326   }
327   PetscFunctionReturn(0);
328 }
329 
330 /*
331     Calls setup for the KSP on each level
332 */
333 #undef __FUNCT__
334 #define __FUNCT__ "PCSetUp_MG"
335 static PetscErrorCode PCSetUp_MG(PC pc)
336 {
337   PC_MG                   **mg = (PC_MG**)pc->data;
338   PetscErrorCode          ierr;
339   PetscInt                i,n = mg[0]->levels;
340   PC                      cpc;
341   PetscTruth              preonly,lu,redundant,cholesky,monitor = PETSC_FALSE,dump,opsset;
342   PetscViewerASCIIMonitor ascii;
343   PetscViewer             viewer = PETSC_NULL;
344   MPI_Comm                comm;
345   Mat                     dA,dB;
346   MatStructure            uflag;
347   Vec                     tvec;
348 
349   PetscFunctionBegin;
350 
351   /* If user did not provide fine grid operators OR operator was not updated since last global KSPSetOperators() */
352   /* so use those from global PC */
353   /* Is this what we always want? What if user wants to keep old one? */
354   ierr = KSPGetOperatorsSet(mg[n-1]->smoothd,PETSC_NULL,&opsset);CHKERRQ(ierr);
355   ierr = KSPGetPC(mg[0]->smoothd,&cpc);CHKERRQ(ierr);
356   if (!opsset || cpc->setupcalled == 2) {
357     ierr = PetscInfo(pc,"Using outer operators to define finest grid operator \n  because PCMGGetSmoother(pc,nlevels-1,&ksp);KSPSetOperators(ksp,...); was not called.\n");CHKERRQ(ierr);
358     ierr = KSPSetOperators(mg[n-1]->smoothd,pc->mat,pc->pmat,pc->flag);CHKERRQ(ierr);
359   }
360 
361   if (mg[0]->galerkin) {
362     Mat B;
363     mg[0]->galerkinused = PETSC_TRUE;
364     /* currently only handle case where mat and pmat are the same on coarser levels */
365     ierr = KSPGetOperators(mg[n-1]->smoothd,&dA,&dB,&uflag);CHKERRQ(ierr);
366     if (!pc->setupcalled) {
367       for (i=n-2; i>-1; i--) {
368         ierr = MatPtAP(dB,mg[i+1]->interpolate,MAT_INITIAL_MATRIX,1.0,&B);CHKERRQ(ierr);
369         ierr = KSPSetOperators(mg[i]->smoothd,B,B,uflag);CHKERRQ(ierr);
370 	if (i != n-2) {ierr = PetscObjectDereference((PetscObject)dB);CHKERRQ(ierr);}
371         dB   = B;
372       }
373       ierr = PetscObjectDereference((PetscObject)dB);CHKERRQ(ierr);
374     } else {
375       for (i=n-2; i>-1; i--) {
376         ierr = KSPGetOperators(mg[i]->smoothd,PETSC_NULL,&B,PETSC_NULL);CHKERRQ(ierr);
377         ierr = MatPtAP(dB,mg[i+1]->interpolate,MAT_REUSE_MATRIX,1.0,&B);CHKERRQ(ierr);
378         ierr = KSPSetOperators(mg[i]->smoothd,B,B,uflag);CHKERRQ(ierr);
379         dB   = B;
380       }
381     }
382   }
383 
384   if (!pc->setupcalled) {
385     ierr = PetscOptionsHasName(0,"-pc_mg_monitor",&monitor);CHKERRQ(ierr);
386 
387     for (i=0; i<n; i++) {
388       if (monitor) {
389         ierr = PetscObjectGetComm((PetscObject)mg[i]->smoothd,&comm);CHKERRQ(ierr);
390         ierr = PetscViewerASCIIMonitorCreate(comm,"stdout",n-i,&ascii);CHKERRQ(ierr);
391         ierr = KSPMonitorSet(mg[i]->smoothd,KSPMonitorDefault,ascii,(PetscErrorCode(*)(void*))PetscViewerASCIIMonitorDestroy);CHKERRQ(ierr);
392       }
393       ierr = KSPSetFromOptions(mg[i]->smoothd);CHKERRQ(ierr);
394     }
395     for (i=1; i<n; i++) {
396       if (mg[i]->smoothu && (mg[i]->smoothu != mg[i]->smoothd)) {
397         if (monitor) {
398           ierr = PetscObjectGetComm((PetscObject)mg[i]->smoothu,&comm);CHKERRQ(ierr);
399           ierr = PetscViewerASCIIMonitorCreate(comm,"stdout",n-i,&ascii);CHKERRQ(ierr);
400           ierr = KSPMonitorSet(mg[i]->smoothu,KSPMonitorDefault,ascii,(PetscErrorCode(*)(void*))PetscViewerASCIIMonitorDestroy);CHKERRQ(ierr);
401         }
402         ierr = KSPSetFromOptions(mg[i]->smoothu);CHKERRQ(ierr);
403       }
404     }
405     for (i=1; i<n; i++) {
406       if (!mg[i]->residual) {
407         Mat mat;
408         ierr = KSPGetOperators(mg[i]->smoothd,PETSC_NULL,&mat,PETSC_NULL);CHKERRQ(ierr);
409         ierr = PCMGSetResidual(pc,i,PCMGDefaultResidual,mat);CHKERRQ(ierr);
410       }
411       if (mg[i]->restrct && !mg[i]->interpolate) {
412         ierr = PCMGSetInterpolate(pc,i,mg[i]->restrct);CHKERRQ(ierr);
413       }
414       if (!mg[i]->restrct && mg[i]->interpolate) {
415         ierr = PCMGSetRestriction(pc,i,mg[i]->interpolate);CHKERRQ(ierr);
416       }
417 #if defined(PETSC_USE_DEBUG)
418       if (!mg[i]->restrct || !mg[i]->interpolate) {
419         SETERRQ1(PETSC_ERR_ARG_WRONGSTATE,"Need to set restriction or interpolation on level %d",(int)i);
420       }
421 #endif
422     }
423     for (i=0; i<n-1; i++) {
424       if (!mg[i]->b) {
425         Vec *vec;
426         ierr = KSPGetVecs(mg[i]->smoothd,1,&vec,0,PETSC_NULL);CHKERRQ(ierr);
427         ierr = PCMGSetRhs(pc,i,*vec);CHKERRQ(ierr);
428         ierr = PetscFree(vec);CHKERRQ(ierr);
429       }
430       if (!mg[i]->r && i) {
431         ierr = VecDuplicate(mg[i]->b,&tvec);CHKERRQ(ierr);
432         ierr = PCMGSetR(pc,i,tvec);CHKERRQ(ierr);
433         ierr = VecDestroy(tvec);CHKERRQ(ierr);
434       }
435       if (!mg[i]->x) {
436         ierr = VecDuplicate(mg[i]->b,&tvec);CHKERRQ(ierr);
437         ierr = PCMGSetX(pc,i,tvec);CHKERRQ(ierr);
438         ierr = VecDestroy(tvec);CHKERRQ(ierr);
439       }
440     }
441   }
442 
443 
444   for (i=1; i<n; i++) {
445     if (mg[i]->smoothu == mg[i]->smoothd) {
446       /* if doing only down then initial guess is zero */
447       ierr = KSPSetInitialGuessNonzero(mg[i]->smoothd,PETSC_TRUE);CHKERRQ(ierr);
448     }
449     if (mg[i]->eventsetup) {ierr = PetscLogEventBegin(mg[i]->eventsetup,0,0,0,0);CHKERRQ(ierr);}
450     ierr = KSPSetUp(mg[i]->smoothd);CHKERRQ(ierr);
451     if (mg[i]->eventsetup) {ierr = PetscLogEventEnd(mg[i]->eventsetup,0,0,0,0);CHKERRQ(ierr);}
452   }
453   for (i=1; i<n; i++) {
454     if (mg[i]->smoothu && mg[i]->smoothu != mg[i]->smoothd) {
455       Mat          downmat,downpmat;
456       MatStructure matflag;
457       PetscTruth   opsset;
458 
459       /* check if operators have been set for up, if not use down operators to set them */
460       ierr = KSPGetOperatorsSet(mg[i]->smoothu,&opsset,PETSC_NULL);CHKERRQ(ierr);
461       if (!opsset) {
462         ierr = KSPGetOperators(mg[i]->smoothd,&downmat,&downpmat,&matflag);CHKERRQ(ierr);
463         ierr = KSPSetOperators(mg[i]->smoothu,downmat,downpmat,matflag);CHKERRQ(ierr);
464       }
465 
466       ierr = KSPSetInitialGuessNonzero(mg[i]->smoothu,PETSC_TRUE);CHKERRQ(ierr);
467       if (mg[i]->eventsetup) {ierr = PetscLogEventBegin(mg[i]->eventsetup,0,0,0,0);CHKERRQ(ierr);}
468       ierr = KSPSetUp(mg[i]->smoothu);CHKERRQ(ierr);
469       if (mg[i]->eventsetup) {ierr = PetscLogEventEnd(mg[i]->eventsetup,0,0,0,0);CHKERRQ(ierr);}
470     }
471   }
472 
473   /*
474       If coarse solver is not direct method then DO NOT USE preonly
475   */
476   ierr = PetscTypeCompare((PetscObject)mg[0]->smoothd,KSPPREONLY,&preonly);CHKERRQ(ierr);
477   if (preonly) {
478     ierr = PetscTypeCompare((PetscObject)cpc,PCLU,&lu);CHKERRQ(ierr);
479     ierr = PetscTypeCompare((PetscObject)cpc,PCREDUNDANT,&redundant);CHKERRQ(ierr);
480     ierr = PetscTypeCompare((PetscObject)cpc,PCCHOLESKY,&cholesky);CHKERRQ(ierr);
481     if (!lu && !redundant && !cholesky) {
482       ierr = KSPSetType(mg[0]->smoothd,KSPGMRES);CHKERRQ(ierr);
483     }
484   }
485 
486   if (!pc->setupcalled) {
487     if (monitor) {
488       ierr = PetscObjectGetComm((PetscObject)mg[0]->smoothd,&comm);CHKERRQ(ierr);
489       ierr = PetscViewerASCIIMonitorCreate(comm,"stdout",n,&ascii);CHKERRQ(ierr);
490       ierr = KSPMonitorSet(mg[0]->smoothd,KSPMonitorDefault,ascii,(PetscErrorCode(*)(void*))PetscViewerASCIIMonitorDestroy);CHKERRQ(ierr);
491     }
492     ierr = KSPSetFromOptions(mg[0]->smoothd);CHKERRQ(ierr);
493   }
494 
495   if (mg[0]->eventsetup) {ierr = PetscLogEventBegin(mg[0]->eventsetup,0,0,0,0);CHKERRQ(ierr);}
496   ierr = KSPSetUp(mg[0]->smoothd);CHKERRQ(ierr);
497   if (mg[0]->eventsetup) {ierr = PetscLogEventEnd(mg[0]->eventsetup,0,0,0,0);CHKERRQ(ierr);}
498 
499   /*
500      Dump the interpolation/restriction matrices plus the
501    Jacobian/stiffness on each level. This allows Matlab users to
502    easily check if the Galerkin condition A_c = R A_f R^T is satisfied.
503 
504    Only support one or the other at the same time.
505   */
506 #if defined(PETSC_USE_SOCKET_VIEWER)
507   ierr = PetscOptionsHasName(pc->prefix,"-pc_mg_dump_matlab",&dump);CHKERRQ(ierr);
508   if (dump) {
509     viewer = PETSC_VIEWER_SOCKET_(pc->comm);
510   }
511 #endif
512   ierr = PetscOptionsHasName(pc->prefix,"-pc_mg_dump_binary",&dump);CHKERRQ(ierr);
513   if (dump) {
514     viewer = PETSC_VIEWER_BINARY_(pc->comm);
515   }
516 
517   if (viewer) {
518     for (i=1; i<n; i++) {
519       ierr = MatView(mg[i]->restrct,viewer);CHKERRQ(ierr);
520     }
521     for (i=0; i<n; i++) {
522       ierr = KSPGetPC(mg[i]->smoothd,&pc);CHKERRQ(ierr);
523       ierr = MatView(pc->mat,viewer);CHKERRQ(ierr);
524     }
525   }
526   PetscFunctionReturn(0);
527 }
528 
529 /* -------------------------------------------------------------------------------------*/
530 
531 #undef __FUNCT__
532 #define __FUNCT__ "PCMGSetLevels"
533 /*@C
534    PCMGSetLevels - Sets the number of levels to use with MG.
535    Must be called before any other MG routine.
536 
537    Collective on PC
538 
539    Input Parameters:
540 +  pc - the preconditioner context
541 .  levels - the number of levels
542 -  comms - optional communicators for each level; this is to allow solving the coarser problems
543            on smaller sets of processors. Use PETSC_NULL_OBJECT for default in Fortran
544 
545    Level: intermediate
546 
547    Notes:
548      If the number of levels is one then the multigrid uses the -mg_levels prefix
549   for setting the level options rather than the -mg_coarse prefix.
550 
551 .keywords: MG, set, levels, multigrid
552 
553 .seealso: PCMGSetType(), PCMGGetLevels()
554 @*/
555 PetscErrorCode PETSCKSP_DLLEXPORT PCMGSetLevels(PC pc,PetscInt levels,MPI_Comm *comms)
556 {
557   PetscErrorCode ierr;
558   PC_MG          **mg=0;
559 
560   PetscFunctionBegin;
561   PetscValidHeaderSpecific(pc,PC_COOKIE,1);
562 
563   if (pc->data) {
564     SETERRQ(PETSC_ERR_ORDER,"Number levels already set for MG\n\
565     make sure that you call PCMGSetLevels() before KSPSetFromOptions()");
566   }
567   ierr                     = PCMGCreate_Private(pc->comm,levels,pc,comms,&mg);CHKERRQ(ierr);
568   mg[0]->am                = PC_MG_MULTIPLICATIVE;
569   pc->data                 = (void*)mg;
570   pc->ops->applyrichardson = PCApplyRichardson_MG;
571   PetscFunctionReturn(0);
572 }
573 
574 #undef __FUNCT__
575 #define __FUNCT__ "PCMGGetLevels"
576 /*@
577    PCMGGetLevels - Gets the number of levels to use with MG.
578 
579    Not Collective
580 
581    Input Parameter:
582 .  pc - the preconditioner context
583 
584    Output parameter:
585 .  levels - the number of levels
586 
587    Level: advanced
588 
589 .keywords: MG, get, levels, multigrid
590 
591 .seealso: PCMGSetLevels()
592 @*/
593 PetscErrorCode PETSCKSP_DLLEXPORT PCMGGetLevels(PC pc,PetscInt *levels)
594 {
595   PC_MG  **mg;
596 
597   PetscFunctionBegin;
598   PetscValidHeaderSpecific(pc,PC_COOKIE,1);
599   PetscValidIntPointer(levels,2);
600 
601   mg      = (PC_MG**)pc->data;
602   *levels = mg[0]->levels;
603   PetscFunctionReturn(0);
604 }
605 
606 #undef __FUNCT__
607 #define __FUNCT__ "PCMGSetType"
608 /*@
609    PCMGSetType - Determines the form of multigrid to use:
610    multiplicative, additive, full, or the Kaskade algorithm.
611 
612    Collective on PC
613 
614    Input Parameters:
615 +  pc - the preconditioner context
616 -  form - multigrid form, one of PC_MG_MULTIPLICATIVE, PC_MG_ADDITIVE,
617    PC_MG_FULL, PC_MG_KASKADE
618 
619    Options Database Key:
620 .  -pc_mg_type <form> - Sets <form>, one of multiplicative,
621    additive, full, kaskade
622 
623    Level: advanced
624 
625 .keywords: MG, set, method, multiplicative, additive, full, Kaskade, multigrid
626 
627 .seealso: PCMGSetLevels()
628 @*/
629 PetscErrorCode PETSCKSP_DLLEXPORT PCMGSetType(PC pc,PCMGType form)
630 {
631   PC_MG **mg;
632 
633   PetscFunctionBegin;
634   PetscValidHeaderSpecific(pc,PC_COOKIE,1);
635   mg = (PC_MG**)pc->data;
636 
637   if (!mg) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Must set MG levels before calling");
638   mg[0]->am = form;
639   if (form == PC_MG_MULTIPLICATIVE) pc->ops->applyrichardson = PCApplyRichardson_MG;
640   else pc->ops->applyrichardson = 0;
641   PetscFunctionReturn(0);
642 }
643 
644 #undef __FUNCT__
645 #define __FUNCT__ "PCMGSetCycles"
646 /*@
647    PCMGSetCycles - Sets the type cycles to use.  Use PCMGSetCyclesOnLevel() for more
648    complicated cycling.
649 
650    Collective on PC
651 
652    Input Parameters:
653 +  pc - the multigrid context
654 -  n - the number of cycles
655 
656    Options Database Key:
657 $  -pc_mg_cycles n - 1 denotes a V-cycle; 2 denotes a W-cycle.
658 
659    Level: advanced
660 
661 .keywords: MG, set, cycles, V-cycle, W-cycle, multigrid
662 
663 .seealso: PCMGSetCyclesOnLevel()
664 @*/
665 PetscErrorCode PETSCKSP_DLLEXPORT PCMGSetCycles(PC pc,PetscInt n)
666 {
667   PC_MG    **mg;
668   PetscInt i,levels;
669 
670   PetscFunctionBegin;
671   PetscValidHeaderSpecific(pc,PC_COOKIE,1);
672   mg     = (PC_MG**)pc->data;
673   if (!mg) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Must set MG levels before calling");
674   levels = mg[0]->levels;
675 
676   for (i=0; i<levels; i++) {
677     mg[i]->cycles  = n;
678   }
679   PetscFunctionReturn(0);
680 }
681 
682 #undef __FUNCT__
683 #define __FUNCT__ "PCMGSetGalerkin"
684 /*@
685    PCMGSetGalerkin - Causes the coarser grid matrices to be computed from the
686       finest grid via the Galerkin process: A_i-1 = r_i * A_i * r_i^t
687 
688    Collective on PC
689 
690    Input Parameters:
691 .  pc - the multigrid context
692 
693    Options Database Key:
694 $  -pc_mg_galerkin
695 
696    Level: intermediate
697 
698 .keywords: MG, set, Galerkin
699 
700 .seealso: PCMGGetGalerkin()
701 
702 @*/
703 PetscErrorCode PETSCKSP_DLLEXPORT PCMGSetGalerkin(PC pc)
704 {
705   PC_MG    **mg;
706   PetscInt i,levels;
707 
708   PetscFunctionBegin;
709   PetscValidHeaderSpecific(pc,PC_COOKIE,1);
710   mg     = (PC_MG**)pc->data;
711   if (!mg) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Must set MG levels before calling");
712   levels = mg[0]->levels;
713 
714   for (i=0; i<levels; i++) {
715     mg[i]->galerkin = PETSC_TRUE;
716   }
717   PetscFunctionReturn(0);
718 }
719 
720 #undef __FUNCT__
721 #define __FUNCT__ "PCMGGetGalerkin"
722 /*@
723    PCMGGetGalerkin - Checks if Galerkin multigrid is being used, i.e.
724       A_i-1 = r_i * A_i * r_i^t
725 
726    Not Collective
727 
728    Input Parameter:
729 .  pc - the multigrid context
730 
731    Output Parameter:
732 .  gelerkin - PETSC_TRUE or PETSC_FALSE
733 
734    Options Database Key:
735 $  -pc_mg_galerkin
736 
737    Level: intermediate
738 
739 .keywords: MG, set, Galerkin
740 
741 .seealso: PCMGSetGalerkin()
742 
743 @*/
744 PetscErrorCode PETSCKSP_DLLEXPORT PCMGGetGalerkin(PC pc,PetscTruth *galerkin)
745 {
746   PC_MG    **mg;
747 
748   PetscFunctionBegin;
749   PetscValidHeaderSpecific(pc,PC_COOKIE,1);
750   mg     = (PC_MG**)pc->data;
751   if (!mg) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Must set MG levels before calling");
752   *galerkin = mg[0]->galerkin;
753   PetscFunctionReturn(0);
754 }
755 
756 #undef __FUNCT__
757 #define __FUNCT__ "PCMGSetNumberSmoothDown"
758 /*@
759    PCMGSetNumberSmoothDown - Sets the number of pre-smoothing steps to
760    use on all levels. Use PCMGGetSmootherDown() to set different
761    pre-smoothing steps on different levels.
762 
763    Collective on PC
764 
765    Input Parameters:
766 +  mg - the multigrid context
767 -  n - the number of smoothing steps
768 
769    Options Database Key:
770 .  -pc_mg_smoothdown <n> - Sets number of pre-smoothing steps
771 
772    Level: advanced
773 
774 .keywords: MG, smooth, down, pre-smoothing, steps, multigrid
775 
776 .seealso: PCMGSetNumberSmoothUp()
777 @*/
778 PetscErrorCode PETSCKSP_DLLEXPORT PCMGSetNumberSmoothDown(PC pc,PetscInt n)
779 {
780   PC_MG          **mg;
781   PetscErrorCode ierr;
782   PetscInt       i,levels;
783 
784   PetscFunctionBegin;
785   PetscValidHeaderSpecific(pc,PC_COOKIE,1);
786   mg     = (PC_MG**)pc->data;
787   if (!mg) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Must set MG levels before calling");
788   levels = mg[0]->levels;
789 
790   for (i=1; i<levels; i++) {
791     /* make sure smoother up and down are different */
792     ierr = PCMGGetSmootherUp(pc,i,PETSC_NULL);CHKERRQ(ierr);
793     ierr = KSPSetTolerances(mg[i]->smoothd,PETSC_DEFAULT,PETSC_DEFAULT,PETSC_DEFAULT,n);CHKERRQ(ierr);
794     mg[i]->default_smoothd = n;
795   }
796   PetscFunctionReturn(0);
797 }
798 
799 #undef __FUNCT__
800 #define __FUNCT__ "PCMGSetNumberSmoothUp"
801 /*@
802    PCMGSetNumberSmoothUp - Sets the number of post-smoothing steps to use
803    on all levels. Use PCMGGetSmootherUp() to set different numbers of
804    post-smoothing steps on different levels.
805 
806    Collective on PC
807 
808    Input Parameters:
809 +  mg - the multigrid context
810 -  n - the number of smoothing steps
811 
812    Options Database Key:
813 .  -pc_mg_smoothup <n> - Sets number of post-smoothing steps
814 
815    Level: advanced
816 
817    Note: this does not set a value on the coarsest grid, since we assume that
818     there is no separate smooth up on the coarsest grid.
819 
820 .keywords: MG, smooth, up, post-smoothing, steps, multigrid
821 
822 .seealso: PCMGSetNumberSmoothDown()
823 @*/
824 PetscErrorCode PETSCKSP_DLLEXPORT PCMGSetNumberSmoothUp(PC pc,PetscInt n)
825 {
826   PC_MG          **mg;
827   PetscErrorCode ierr;
828   PetscInt       i,levels;
829 
830   PetscFunctionBegin;
831   PetscValidHeaderSpecific(pc,PC_COOKIE,1);
832   mg     = (PC_MG**)pc->data;
833   if (!mg) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Must set MG levels before calling");
834   levels = mg[0]->levels;
835 
836   for (i=1; i<levels; i++) {
837     /* make sure smoother up and down are different */
838     ierr = PCMGGetSmootherUp(pc,i,PETSC_NULL);CHKERRQ(ierr);
839     ierr = KSPSetTolerances(mg[i]->smoothu,PETSC_DEFAULT,PETSC_DEFAULT,PETSC_DEFAULT,n);CHKERRQ(ierr);
840     mg[i]->default_smoothu = n;
841   }
842   PetscFunctionReturn(0);
843 }
844 
845 /* ----------------------------------------------------------------------------------------*/
846 
847 /*MC
848    PCMG - Use multigrid preconditioning. This preconditioner requires you provide additional
849     information about the coarser grid matrices and restriction/interpolation operators.
850 
851    Options Database Keys:
852 +  -pc_mg_levels <nlevels> - number of levels including finest
853 .  -pc_mg_cycles 1 or 2 - for V or W-cycle
854 .  -pc_mg_smoothup <n> - number of smoothing steps after interpolation
855 .  -pc_mg_smoothdown <n> - number of smoothing steps before applying restriction operator
856 .  -pc_mg_type <additive,multiplicative,full,cascade> - multiplicative is the default
857 .  -pc_mg_log - log information about time spent on each level of the solver
858 .  -pc_mg_monitor - print information on the multigrid convergence
859 .  -pc_mg_galerkin - use Galerkin process to compute coarser operators
860 -  -pc_mg_dump_matlab - dumps the matrices for each level and the restriction/interpolation matrices
861                         to the Socket viewer for reading from Matlab.
862 
863    Notes:
864 
865    Level: intermediate
866 
867    Concepts: multigrid
868 
869 .seealso:  PCCreate(), PCSetType(), PCType (for list of available types), PC, PCMGType,
870            PCMGSetLevels(), PCMGGetLevels(), PCMGSetType(), PCMGSetCycles(), PCMGSetNumberSmoothDown(),
871            PCMGSetNumberSmoothUp(), PCMGGetCoarseSolve(), PCMGSetResidual(), PCMGSetInterpolation(),
872            PCMGSetRestriction(), PCMGGetSmoother(), PCMGGetSmootherUp(), PCMGGetSmootherDown(),
873            PCMGSetCyclesOnLevel(), PCMGSetRhs(), PCMGSetX(), PCMGSetR()
874 M*/
875 
876 EXTERN_C_BEGIN
877 #undef __FUNCT__
878 #define __FUNCT__ "PCCreate_MG"
879 PetscErrorCode PETSCKSP_DLLEXPORT PCCreate_MG(PC pc)
880 {
881   PetscFunctionBegin;
882   pc->ops->apply          = PCApply_MG;
883   pc->ops->setup          = PCSetUp_MG;
884   pc->ops->destroy        = PCDestroy_MG;
885   pc->ops->setfromoptions = PCSetFromOptions_MG;
886   pc->ops->view           = PCView_MG;
887 
888   pc->data                = (void*)0;
889   PetscFunctionReturn(0);
890 }
891 EXTERN_C_END
892