xref: /petsc/src/ksp/pc/impls/mg/mg.c (revision 794163968b27cf3436b159913eecff082c02cff0)
1 /*
2     Defines the multigrid preconditioner interface.
3 */
4 #include "src/ksp/pc/impls/mg/mgimpl.h"                    /*I "petscmg.h" I*/
5 
6 
7 /*
8        MGMCycle_Private - Given an MG structure created with MGCreate() runs
9                   one multiplicative cycle down through the levels and
10                   back up.
11 
12     Input Parameter:
13 .   mg - structure created with  MGCreate().
14 */
15 #undef __FUNCT__
16 #define __FUNCT__ "MGMCycle_Private"
17 PetscErrorCode MGMCycle_Private(MG *mglevels,PetscTruth *converged)
18 {
19   MG             mg = *mglevels,mgc;
20   PetscErrorCode ierr;
21   PetscInt       cycles = mg->cycles;
22   PetscScalar    zero = 0.0;
23 
24   PetscFunctionBegin;
25   if (converged) *converged = PETSC_FALSE;
26 
27   if (mg->eventsolve) {ierr = PetscLogEventBegin(mg->eventsolve,0,0,0,0);CHKERRQ(ierr);}
28   ierr = KSPSolve(mg->smoothd,mg->b,mg->x);CHKERRQ(ierr);
29   if (mg->eventsolve) {ierr = PetscLogEventEnd(mg->eventsolve,0,0,0,0);CHKERRQ(ierr);}
30   if (mg->level) {  /* not the coarsest grid */
31     ierr = (*mg->residual)(mg->A,mg->b,mg->x,mg->r);CHKERRQ(ierr);
32 
33     /* if on finest level and have convergence criteria set */
34     if (mg->level == mg->levels-1 && mg->ttol) {
35       PetscReal rnorm;
36       ierr = VecNorm(mg->r,NORM_2,&rnorm);CHKERRQ(ierr);
37       if (rnorm <= mg->ttol) {
38         *converged = PETSC_TRUE;
39         if (rnorm < mg->abstol) {
40           PetscLogInfo(0,"Linear solver has converged. Residual norm %g is less than absolute tolerance %g\n",rnorm,mg->abstol);
41         } else {
42           PetscLogInfo(0,"Linear solver has converged. Residual norm %g is less than relative tolerance times initial residual norm %g\n",rnorm,mg->ttol);
43         }
44         PetscFunctionReturn(0);
45       }
46     }
47 
48     mgc = *(mglevels - 1);
49     ierr = MatRestrict(mg->restrct,mg->r,mgc->b);CHKERRQ(ierr);
50     ierr = VecSet(&zero,mgc->x);CHKERRQ(ierr);
51     while (cycles--) {
52       ierr = MGMCycle_Private(mglevels-1,converged);CHKERRQ(ierr);
53     }
54     ierr = MatInterpolateAdd(mg->interpolate,mgc->x,mg->x,mg->x);CHKERRQ(ierr);
55     if (mg->eventsolve) {ierr = PetscLogEventBegin(mg->eventsolve,0,0,0,0);CHKERRQ(ierr);}
56     ierr = KSPSolve(mg->smoothu,mg->b,mg->x);CHKERRQ(ierr);
57     if (mg->eventsolve) {ierr = PetscLogEventEnd(mg->eventsolve,0,0,0,0);CHKERRQ(ierr);}
58   }
59   PetscFunctionReturn(0);
60 }
61 
62 /*
63        MGCreate_Private - Creates a MG structure for use with the
64                multigrid code. Level 0 is the coarsest. (But the
65                finest level is stored first in the array).
66 
67 */
68 #undef __FUNCT__
69 #define __FUNCT__ "MGCreate_Private"
70 static PetscErrorCode MGCreate_Private(MPI_Comm comm,PetscInt levels,PC pc,MPI_Comm *comms,MG **result)
71 {
72   MG             *mg;
73   PetscErrorCode ierr;
74   PetscInt       i;
75   PetscMPIInt    size;
76   char           *prefix;
77   PC             ipc;
78 
79   PetscFunctionBegin;
80   ierr = PetscMalloc(levels*sizeof(MG),&mg);CHKERRQ(ierr);
81   PetscLogObjectMemory(pc,levels*(sizeof(MG)+sizeof(struct _MG)));
82 
83   ierr = PCGetOptionsPrefix(pc,&prefix);CHKERRQ(ierr);
84 
85   for (i=0; i<levels; i++) {
86     ierr = PetscNew(struct _MG,&mg[i]);CHKERRQ(ierr);
87     ierr = PetscMemzero(mg[i],sizeof(struct _MG));CHKERRQ(ierr);
88     mg[i]->level  = i;
89     mg[i]->levels = levels;
90     mg[i]->cycles = 1;
91     mg[i]->default_smoothu = 1;
92     mg[i]->default_smoothd = 1;
93 
94     if (comms) comm = comms[i];
95     ierr = KSPCreate(comm,&mg[i]->smoothd);CHKERRQ(ierr);
96     ierr = KSPSetTolerances(mg[i]->smoothd,PETSC_DEFAULT,PETSC_DEFAULT,PETSC_DEFAULT, mg[i]->default_smoothd);CHKERRQ(ierr);
97     ierr = KSPSetOptionsPrefix(mg[i]->smoothd,prefix);CHKERRQ(ierr);
98 
99     /* do special stuff for coarse grid */
100     if (!i && levels > 1) {
101       ierr = KSPAppendOptionsPrefix(mg[0]->smoothd,"mg_coarse_");CHKERRQ(ierr);
102 
103       /* coarse solve is (redundant) LU by default */
104       ierr = KSPSetType(mg[0]->smoothd,KSPPREONLY);CHKERRQ(ierr);
105       ierr = KSPGetPC(mg[0]->smoothd,&ipc);CHKERRQ(ierr);
106       ierr = MPI_Comm_size(comm,&size);CHKERRQ(ierr);
107       if (size > 1) {
108         ierr = PCSetType(ipc,PCREDUNDANT);CHKERRQ(ierr);
109         ierr = PCRedundantGetPC(ipc,&ipc);CHKERRQ(ierr);
110       }
111       ierr = PCSetType(ipc,PCLU);CHKERRQ(ierr);
112 
113     } else {
114       ierr = KSPAppendOptionsPrefix(mg[i]->smoothd,"mg_levels_");CHKERRQ(ierr);
115     }
116     PetscLogObjectParent(pc,mg[i]->smoothd);
117     mg[i]->smoothu         = mg[i]->smoothd;
118     mg[i]->rtol = 0.0;
119     mg[i]->abstol = 0.0;
120     mg[i]->dtol = 0.0;
121     mg[i]->ttol = 0.0;
122     mg[i]->eventsetup = 0;
123     mg[i]->eventsolve = 0;
124   }
125   *result = mg;
126   PetscFunctionReturn(0);
127 }
128 
129 #undef __FUNCT__
130 #define __FUNCT__ "PCDestroy_MG"
131 static PetscErrorCode PCDestroy_MG(PC pc)
132 {
133   MG             *mg = (MG*)pc->data;
134   PetscErrorCode ierr;
135   PetscInt       i,n = mg[0]->levels;
136 
137   PetscFunctionBegin;
138   for (i=0; i<n; i++) {
139     if (mg[i]->smoothd != mg[i]->smoothu) {
140       ierr = KSPDestroy(mg[i]->smoothd);CHKERRQ(ierr);
141     }
142     ierr = KSPDestroy(mg[i]->smoothu);CHKERRQ(ierr);
143     ierr = PetscFree(mg[i]);CHKERRQ(ierr);
144   }
145   ierr = PetscFree(mg);CHKERRQ(ierr);
146   PetscFunctionReturn(0);
147 }
148 
149 
150 
151 EXTERN PetscErrorCode MGACycle_Private(MG*);
152 EXTERN PetscErrorCode MGFCycle_Private(MG*);
153 EXTERN PetscErrorCode MGKCycle_Private(MG*);
154 
155 /*
156    PCApply_MG - Runs either an additive, multiplicative, Kaskadic
157              or full cycle of multigrid.
158 
159   Note:
160   A simple wrapper which calls MGMCycle(),MGACycle(), or MGFCycle().
161 */
162 #undef __FUNCT__
163 #define __FUNCT__ "PCApply_MG"
164 static PetscErrorCode PCApply_MG(PC pc,Vec b,Vec x)
165 {
166   MG             *mg = (MG*)pc->data;
167   PetscScalar    zero = 0.0;
168   PetscErrorCode ierr;
169   PetscInt       levels = mg[0]->levels;
170 
171   PetscFunctionBegin;
172   mg[levels-1]->b = b;
173   mg[levels-1]->x = x;
174   if (mg[0]->am == MGMULTIPLICATIVE) {
175     ierr = VecSet(&zero,x);CHKERRQ(ierr);
176     ierr = MGMCycle_Private(mg+levels-1,PETSC_NULL);CHKERRQ(ierr);
177   }
178   else if (mg[0]->am == MGADDITIVE) {
179     ierr = MGACycle_Private(mg);CHKERRQ(ierr);
180   }
181   else if (mg[0]->am == MGKASKADE) {
182     ierr = MGKCycle_Private(mg);CHKERRQ(ierr);
183   }
184   else {
185     ierr = MGFCycle_Private(mg);CHKERRQ(ierr);
186   }
187   PetscFunctionReturn(0);
188 }
189 
190 #undef __FUNCT__
191 #define __FUNCT__ "PCApplyRichardson_MG"
192 static PetscErrorCode PCApplyRichardson_MG(PC pc,Vec b,Vec x,Vec w,PetscReal rtol,PetscReal abstol, PetscReal dtol,PetscInt its)
193 {
194   MG             *mg = (MG*)pc->data;
195   PetscErrorCode ierr;
196   PetscInt       levels = mg[0]->levels;
197   PetscTruth     converged = PETSC_FALSE;
198 
199   PetscFunctionBegin;
200   mg[levels-1]->b    = b;
201   mg[levels-1]->x    = x;
202 
203   mg[levels-1]->rtol = rtol;
204   mg[levels-1]->abstol = abstol;
205   mg[levels-1]->dtol = dtol;
206   if (rtol) {
207     /* compute initial residual norm for relative convergence test */
208     PetscReal rnorm;
209     ierr               = (*mg[levels-1]->residual)(mg[levels-1]->A,b,x,w);CHKERRQ(ierr);
210     ierr               = VecNorm(w,NORM_2,&rnorm);CHKERRQ(ierr);
211     mg[levels-1]->ttol = PetscMax(rtol*rnorm,abstol);
212   } else if (abstol) {
213     mg[levels-1]->ttol = abstol;
214   } else {
215     mg[levels-1]->ttol = 0.0;
216   }
217 
218   while (its-- && !converged) {
219     ierr = MGMCycle_Private(mg+levels-1,&converged);CHKERRQ(ierr);
220   }
221   PetscFunctionReturn(0);
222 }
223 
224 #undef __FUNCT__
225 #define __FUNCT__ "PCSetFromOptions_MG"
226 static PetscErrorCode PCSetFromOptions_MG(PC pc)
227 {
228   PetscErrorCode ierr;
229   PetscInt       indx,m,levels = 1;
230   PetscTruth     flg;
231   const char     *type[] = {"additive","multiplicative","full","cascade","kascade"};
232 
233   PetscFunctionBegin;
234 
235   ierr = PetscOptionsHead("Multigrid options");CHKERRQ(ierr);
236     if (!pc->data) {
237       ierr = PetscOptionsInt("-pc_mg_levels","Number of Levels","MGSetLevels",levels,&levels,&flg);CHKERRQ(ierr);
238       ierr = MGSetLevels(pc,levels,PETSC_NULL);CHKERRQ(ierr);
239     }
240     ierr = PetscOptionsInt("-pc_mg_cycles","1 for V cycle, 2 for W-cycle","MGSetCycles",1,&m,&flg);CHKERRQ(ierr);
241     if (flg) {
242       ierr = MGSetCycles(pc,m);CHKERRQ(ierr);
243     }
244     ierr = PetscOptionsInt("-pc_mg_smoothup","Number of post-smoothing steps","MGSetNumberSmoothUp",1,&m,&flg);CHKERRQ(ierr);
245     if (flg) {
246       ierr = MGSetNumberSmoothUp(pc,m);CHKERRQ(ierr);
247     }
248     ierr = PetscOptionsInt("-pc_mg_smoothdown","Number of pre-smoothing steps","MGSetNumberSmoothDown",1,&m,&flg);CHKERRQ(ierr);
249     if (flg) {
250       ierr = MGSetNumberSmoothDown(pc,m);CHKERRQ(ierr);
251     }
252     ierr = PetscOptionsEList("-pc_mg_type","Multigrid type","MGSetType",type,5,type[1],&indx,&flg);CHKERRQ(ierr);
253     if (flg) {
254       MGType mg = (MGType) 0;
255       switch (indx) {
256       case 0:
257         mg = MGADDITIVE;
258         break;
259       case 1:
260         mg = MGMULTIPLICATIVE;
261         break;
262       case 2:
263         mg = MGFULL;
264         break;
265       case 3:
266         mg = MGKASKADE;
267         break;
268       case 4:
269         mg = MGKASKADE;
270         break;
271       }
272       ierr = MGSetType(pc,mg);CHKERRQ(ierr);
273     }
274     ierr = PetscOptionsName("-pc_mg_log","Log times for each multigrid level","None",&flg);CHKERRQ(ierr);
275     if (flg) {
276       MG   *mg = (MG*)pc->data;
277       int  i;
278       char eventname[128];
279       if (!mg) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Must set MG levels before calling");
280       levels = mg[0]->levels;
281       for (i=0; i<levels; i++) {
282         sprintf(eventname,"MSetup Level %d",(int)i);
283         ierr = PetscLogEventRegister(&mg[i]->eventsetup,eventname,pc->cookie);CHKERRQ(ierr);
284         sprintf(eventname,"MGSolve Level %d",(int)i);
285         ierr = PetscLogEventRegister(&mg[i]->eventsolve,eventname,pc->cookie);CHKERRQ(ierr);
286       }
287     }
288   ierr = PetscOptionsTail();CHKERRQ(ierr);
289   PetscFunctionReturn(0);
290 }
291 
292 #undef __FUNCT__
293 #define __FUNCT__ "PCView_MG"
294 static PetscErrorCode PCView_MG(PC pc,PetscViewer viewer)
295 {
296   MG             *mg = (MG*)pc->data;
297   PetscErrorCode ierr;
298   PetscInt       levels = mg[0]->levels,i;
299   const char     *cstring;
300   PetscTruth     iascii;
301 
302   PetscFunctionBegin;
303   ierr = PetscTypeCompare((PetscObject)viewer,PETSC_VIEWER_ASCII,&iascii);CHKERRQ(ierr);
304   if (iascii) {
305     if (mg[0]->am == MGMULTIPLICATIVE) cstring = "multiplicative";
306     else if (mg[0]->am == MGADDITIVE)  cstring = "additive";
307     else if (mg[0]->am == MGFULL)      cstring = "full";
308     else if (mg[0]->am == MGKASKADE)   cstring = "Kaskade";
309     else cstring = "unknown";
310     ierr = PetscViewerASCIIPrintf(viewer,"  MG: type is %s, levels=%D cycles=%D, pre-smooths=%D, post-smooths=%D\n",
311                       cstring,levels,mg[0]->cycles,mg[0]->default_smoothd,mg[0]->default_smoothu);CHKERRQ(ierr);
312     for (i=0; i<levels; i++) {
313       ierr = PetscViewerASCIIPrintf(viewer,"Down solver (pre-smoother) on level %D -------------------------------\n",i);CHKERRQ(ierr);
314       ierr = PetscViewerASCIIPushTab(viewer);CHKERRQ(ierr);
315       ierr = KSPView(mg[i]->smoothd,viewer);CHKERRQ(ierr);
316       ierr = PetscViewerASCIIPopTab(viewer);CHKERRQ(ierr);
317       if (mg[i]->smoothd == mg[i]->smoothu) {
318         ierr = PetscViewerASCIIPrintf(viewer,"Up solver (post-smoother) same as down solver (pre-smoother)\n");CHKERRQ(ierr);
319       } else {
320         ierr = PetscViewerASCIIPrintf(viewer,"Up solver (post-smoother) on level %D -------------------------------\n",i);CHKERRQ(ierr);
321         ierr = PetscViewerASCIIPushTab(viewer);CHKERRQ(ierr);
322         ierr = KSPView(mg[i]->smoothu,viewer);CHKERRQ(ierr);
323         ierr = PetscViewerASCIIPopTab(viewer);CHKERRQ(ierr);
324       }
325     }
326   } else {
327     SETERRQ1(PETSC_ERR_SUP,"Viewer type %s not supported for PCMG",((PetscObject)viewer)->type_name);
328   }
329   PetscFunctionReturn(0);
330 }
331 
332 /*
333     Calls setup for the KSP on each level
334 */
335 #undef __FUNCT__
336 #define __FUNCT__ "PCSetUp_MG"
337 static PetscErrorCode PCSetUp_MG(PC pc)
338 {
339   MG             *mg = (MG*)pc->data;
340   PetscErrorCode ierr;
341   PetscInt       i,n = mg[0]->levels;
342   PC             cpc;
343   PetscTruth     preonly,lu,redundant,monitor = PETSC_FALSE,dump;
344   PetscViewer    ascii;
345   MPI_Comm       comm;
346 
347   PetscFunctionBegin;
348   if (!pc->setupcalled) {
349     ierr = PetscOptionsHasName(0,"-pc_mg_monitor",&monitor);CHKERRQ(ierr);
350 
351     for (i=1; i<n; i++) {
352       if (mg[i]->smoothd) {
353         if (monitor) {
354           ierr = PetscObjectGetComm((PetscObject)mg[i]->smoothd,&comm);CHKERRQ(ierr);
355           ierr = PetscViewerASCIIOpen(comm,"stdout",&ascii);CHKERRQ(ierr);
356           ierr = PetscViewerASCIISetTab(ascii,n-i);CHKERRQ(ierr);
357           ierr = KSPSetMonitor(mg[i]->smoothd,KSPDefaultMonitor,ascii,(PetscErrorCode(*)(void*))PetscViewerDestroy);CHKERRQ(ierr);
358         }
359         ierr = KSPSetFromOptions(mg[i]->smoothd);CHKERRQ(ierr);
360       }
361     }
362     for (i=0; i<n; i++) {
363       if (mg[i]->smoothu && mg[i]->smoothu != mg[i]->smoothd) {
364         if (monitor) {
365           ierr = PetscObjectGetComm((PetscObject)mg[i]->smoothu,&comm);CHKERRQ(ierr);
366           ierr = PetscViewerASCIIOpen(comm,"stdout",&ascii);CHKERRQ(ierr);
367           ierr = PetscViewerASCIISetTab(ascii,n-i);CHKERRQ(ierr);
368           ierr = KSPSetMonitor(mg[i]->smoothu,KSPDefaultMonitor,ascii,(PetscErrorCode(*)(void*))PetscViewerDestroy);CHKERRQ(ierr);
369         }
370         ierr = KSPSetFromOptions(mg[i]->smoothu);CHKERRQ(ierr);
371       }
372     }
373   }
374 
375   for (i=1; i<n; i++) {
376     if (mg[i]->smoothd) {
377       ierr = KSPSetInitialGuessNonzero(mg[i]->smoothd,PETSC_TRUE);CHKERRQ(ierr);
378       if (mg[i]->eventsetup) {ierr = PetscLogEventBegin(mg[i]->eventsetup,0,0,0,0);CHKERRQ(ierr);}
379       ierr = KSPSetUp(mg[i]->smoothd);CHKERRQ(ierr);
380       if (mg[i]->eventsetup) {ierr = PetscLogEventEnd(mg[i]->eventsetup,0,0,0,0);CHKERRQ(ierr);}
381     }
382   }
383   for (i=0; i<n; i++) {
384     if (mg[i]->smoothu && mg[i]->smoothu != mg[i]->smoothd) {
385         PC           downpc,uppc;
386         Mat          downmat,downpmat,upmat,uppmat;
387         MatStructure matflag;
388 
389       /* check if operators have been set for up, if not use down operators to set them */
390       ierr = KSPGetPC(mg[i]->smoothu,&uppc);CHKERRQ(ierr);
391       ierr = PCGetOperators(uppc,&upmat,&uppmat,PETSC_NULL);CHKERRQ(ierr);
392       if (!upmat) {
393         ierr = KSPGetPC(mg[i]->smoothd,&downpc);CHKERRQ(ierr);
394         ierr = PCGetOperators(downpc,&downmat,&downpmat,&matflag);CHKERRQ(ierr);
395         ierr = KSPSetOperators(mg[i]->smoothu,downmat,downpmat,matflag);CHKERRQ(ierr);
396       }
397 
398       ierr = KSPSetInitialGuessNonzero(mg[i]->smoothu,PETSC_TRUE);CHKERRQ(ierr);
399       if (mg[i]->eventsetup) {ierr = PetscLogEventBegin(mg[i]->eventsetup,0,0,0,0);CHKERRQ(ierr);}
400       ierr = KSPSetUp(mg[i]->smoothu);CHKERRQ(ierr);
401       if (mg[i]->eventsetup) {ierr = PetscLogEventEnd(mg[i]->eventsetup,0,0,0,0);CHKERRQ(ierr);}
402     }
403   }
404 
405   /*
406       If coarse solver is not direct method then DO NOT USE preonly
407   */
408   ierr = PetscTypeCompare((PetscObject)mg[0]->smoothd,KSPPREONLY,&preonly);CHKERRQ(ierr);
409   if (preonly) {
410     ierr = KSPGetPC(mg[0]->smoothd,&cpc);CHKERRQ(ierr);
411     ierr = PetscTypeCompare((PetscObject)cpc,PCLU,&lu);CHKERRQ(ierr);
412     ierr = PetscTypeCompare((PetscObject)cpc,PCREDUNDANT,&redundant);CHKERRQ(ierr);
413     if (!lu && !redundant) {
414       ierr = KSPSetType(mg[0]->smoothd,KSPGMRES);CHKERRQ(ierr);
415     }
416   }
417 
418   if (!pc->setupcalled) {
419     if (monitor) {
420       ierr = PetscObjectGetComm((PetscObject)mg[0]->smoothd,&comm);CHKERRQ(ierr);
421       ierr = PetscViewerASCIIOpen(comm,"stdout",&ascii);CHKERRQ(ierr);
422       ierr = PetscViewerASCIISetTab(ascii,n);CHKERRQ(ierr);
423       ierr = KSPSetMonitor(mg[0]->smoothd,KSPDefaultMonitor,ascii,(PetscErrorCode(*)(void*))PetscViewerDestroy);CHKERRQ(ierr);
424     }
425     ierr = KSPSetFromOptions(mg[0]->smoothd);CHKERRQ(ierr);
426   }
427 
428   if (mg[0]->eventsetup) {ierr = PetscLogEventBegin(mg[0]->eventsetup,0,0,0,0);CHKERRQ(ierr);}
429   ierr = KSPSetUp(mg[0]->smoothd);CHKERRQ(ierr);
430   if (mg[0]->eventsetup) {ierr = PetscLogEventEnd(mg[0]->eventsetup,0,0,0,0);CHKERRQ(ierr);}
431 
432   /*
433      Dump the interpolation/restriction matrices to matlab plus the
434    Jacobian/stiffness on each level. This allows Matlab users to
435    easily check if the Galerkin condition A_c = R A_f R^T is satisfied */
436   ierr = PetscOptionsHasName(pc->prefix,"-pc_mg_dump_matlab",&dump);CHKERRQ(ierr);
437   if (dump) {
438     for (i=1; i<n; i++) {
439       ierr = MatView(mg[i]->restrct,PETSC_VIEWER_SOCKET_(pc->comm));CHKERRQ(ierr);
440     }
441     for (i=0; i<n; i++) {
442       ierr = KSPGetPC(mg[i]->smoothd,&pc);CHKERRQ(ierr);
443       ierr = MatView(pc->mat,PETSC_VIEWER_SOCKET_(pc->comm));CHKERRQ(ierr);
444     }
445   }
446 
447   PetscFunctionReturn(0);
448 }
449 
450 /* -------------------------------------------------------------------------------------*/
451 
452 #undef __FUNCT__
453 #define __FUNCT__ "MGSetLevels"
454 /*@C
455    MGSetLevels - Sets the number of levels to use with MG.
456    Must be called before any other MG routine.
457 
458    Collective on PC
459 
460    Input Parameters:
461 +  pc - the preconditioner context
462 .  levels - the number of levels
463 -  comms - optional communicators for each level; this is to allow solving the coarser problems
464            on smaller sets of processors. Use PETSC_NULL_OBJECT for default in Fortran
465 
466    Level: intermediate
467 
468    Notes:
469      If the number of levels is one then the multigrid uses the -mg_levels prefix
470   for setting the level options rather than the -mg_coarse prefix.
471 
472 .keywords: MG, set, levels, multigrid
473 
474 .seealso: MGSetType(), MGGetLevels()
475 @*/
476 PetscErrorCode MGSetLevels(PC pc,int levels,MPI_Comm *comms)
477 {
478   PetscErrorCode ierr;
479   MG             *mg;
480 
481   PetscFunctionBegin;
482   PetscValidHeaderSpecific(pc,PC_COOKIE,1);
483 
484   if (pc->data) {
485     SETERRQ(PETSC_ERR_ORDER,"Number levels already set for MG\n\
486     make sure that you call MGSetLevels() before KSPSetFromOptions()");
487   }
488   ierr                     = MGCreate_Private(pc->comm,levels,pc,comms,&mg);CHKERRQ(ierr);
489   mg[0]->am                = MGMULTIPLICATIVE;
490   pc->data                 = (void*)mg;
491   pc->ops->applyrichardson = PCApplyRichardson_MG;
492   PetscFunctionReturn(0);
493 }
494 
495 #undef __FUNCT__
496 #define __FUNCT__ "MGGetLevels"
497 /*@
498    MGGetLevels - Gets the number of levels to use with MG.
499 
500    Not Collective
501 
502    Input Parameter:
503 .  pc - the preconditioner context
504 
505    Output parameter:
506 .  levels - the number of levels
507 
508    Level: advanced
509 
510 .keywords: MG, get, levels, multigrid
511 
512 .seealso: MGSetLevels()
513 @*/
514 PetscErrorCode MGGetLevels(PC pc,int *levels)
515 {
516   MG  *mg;
517 
518   PetscFunctionBegin;
519   PetscValidHeaderSpecific(pc,PC_COOKIE,1);
520   PetscValidIntPointer(levels,2);
521 
522   mg      = (MG*)pc->data;
523   *levels = mg[0]->levels;
524   PetscFunctionReturn(0);
525 }
526 
527 #undef __FUNCT__
528 #define __FUNCT__ "MGSetType"
529 /*@
530    MGSetType - Determines the form of multigrid to use:
531    multiplicative, additive, full, or the Kaskade algorithm.
532 
533    Collective on PC
534 
535    Input Parameters:
536 +  pc - the preconditioner context
537 -  form - multigrid form, one of MGMULTIPLICATIVE, MGADDITIVE,
538    MGFULL, MGKASKADE
539 
540    Options Database Key:
541 .  -pc_mg_type <form> - Sets <form>, one of multiplicative,
542    additive, full, kaskade
543 
544    Level: advanced
545 
546 .keywords: MG, set, method, multiplicative, additive, full, Kaskade, multigrid
547 
548 .seealso: MGSetLevels()
549 @*/
550 PetscErrorCode MGSetType(PC pc,MGType form)
551 {
552   MG *mg;
553 
554   PetscFunctionBegin;
555   PetscValidHeaderSpecific(pc,PC_COOKIE,1);
556   mg = (MG*)pc->data;
557 
558   if (!mg) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Must set MG levels before calling");
559   mg[0]->am = form;
560   if (form == MGMULTIPLICATIVE) pc->ops->applyrichardson = PCApplyRichardson_MG;
561   else pc->ops->applyrichardson = 0;
562   PetscFunctionReturn(0);
563 }
564 
565 #undef __FUNCT__
566 #define __FUNCT__ "MGSetCycles"
567 /*@
568    MGSetCycles - Sets the type cycles to use.  Use MGSetCyclesOnLevel() for more
569    complicated cycling.
570 
571    Collective on PC
572 
573    Input Parameters:
574 +  mg - the multigrid context
575 -  n - the number of cycles
576 
577    Options Database Key:
578 $  -pc_mg_cycles n - 1 denotes a V-cycle; 2 denotes a W-cycle.
579 
580    Level: advanced
581 
582 .keywords: MG, set, cycles, V-cycle, W-cycle, multigrid
583 
584 .seealso: MGSetCyclesOnLevel()
585 @*/
586 PetscErrorCode MGSetCycles(PC pc,PetscInt n)
587 {
588   MG       *mg;
589   PetscInt i,levels;
590 
591   PetscFunctionBegin;
592   PetscValidHeaderSpecific(pc,PC_COOKIE,1);
593   mg     = (MG*)pc->data;
594   if (!mg) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Must set MG levels before calling");
595   levels = mg[0]->levels;
596 
597   for (i=0; i<levels; i++) {
598     mg[i]->cycles  = n;
599   }
600   PetscFunctionReturn(0);
601 }
602 
603 #undef __FUNCT__
604 #define __FUNCT__ "MGCheck"
605 /*@
606    MGCheck - Checks that all components of the MG structure have
607    been set.
608 
609    Collective on PC
610 
611    Input Parameters:
612 .  mg - the MG structure
613 
614    Level: advanced
615 
616 .keywords: MG, check, set, multigrid
617 @*/
618 PetscErrorCode MGCheck(PC pc)
619 {
620   MG       *mg;
621   PetscInt i,n,count = 0;
622 
623   PetscFunctionBegin;
624   PetscValidHeaderSpecific(pc,PC_COOKIE,1);
625   mg = (MG*)pc->data;
626 
627   if (!mg) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Must set MG levels before calling");
628 
629   n = mg[0]->levels;
630 
631   for (i=1; i<n; i++) {
632     if (!mg[i]->restrct) {
633       (*PetscErrorPrintf)("No restrict set level %D \n",n-i); count++;
634     }
635     if (!mg[i]->interpolate) {
636       (*PetscErrorPrintf)("No interpolate set level %D \n",n-i); count++;
637     }
638     if (!mg[i]->residual) {
639       (*PetscErrorPrintf)("No residual set level %D \n",n-i); count++;
640     }
641     if (!mg[i]->smoothu) {
642       (*PetscErrorPrintf)("No smoothup set level %D \n",n-i); count++;
643     }
644     if (!mg[i]->smoothd) {
645       (*PetscErrorPrintf)("No smoothdown set level %D \n",n-i); count++;
646     }
647     if (!mg[i]->r) {
648       (*PetscErrorPrintf)("No r set level %D \n",n-i); count++;
649     }
650     if (!mg[i-1]->x) {
651       (*PetscErrorPrintf)("No x set level %D \n",n-i); count++;
652     }
653     if (!mg[i-1]->b) {
654       (*PetscErrorPrintf)("No b set level %D \n",n-i); count++;
655     }
656   }
657   PetscFunctionReturn(count);
658 }
659 
660 
661 #undef __FUNCT__
662 #define __FUNCT__ "MGSetNumberSmoothDown"
663 /*@
664    MGSetNumberSmoothDown - Sets the number of pre-smoothing steps to
665    use on all levels. Use MGGetSmootherDown() to set different
666    pre-smoothing steps on different levels.
667 
668    Collective on PC
669 
670    Input Parameters:
671 +  mg - the multigrid context
672 -  n - the number of smoothing steps
673 
674    Options Database Key:
675 .  -pc_mg_smoothdown <n> - Sets number of pre-smoothing steps
676 
677    Level: advanced
678 
679 .keywords: MG, smooth, down, pre-smoothing, steps, multigrid
680 
681 .seealso: MGSetNumberSmoothUp()
682 @*/
683 PetscErrorCode MGSetNumberSmoothDown(PC pc,PetscInt n)
684 {
685   MG             *mg;
686   PetscErrorCode ierr;
687   PetscInt       i,levels;
688 
689   PetscFunctionBegin;
690   PetscValidHeaderSpecific(pc,PC_COOKIE,1);
691   mg     = (MG*)pc->data;
692   if (!mg) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Must set MG levels before calling");
693   levels = mg[0]->levels;
694 
695   for (i=0; i<levels; i++) {
696     /* make sure smoother up and down are different */
697     ierr = MGGetSmootherUp(pc,i,PETSC_NULL);CHKERRQ(ierr);
698     ierr = KSPSetTolerances(mg[i]->smoothd,PETSC_DEFAULT,PETSC_DEFAULT,PETSC_DEFAULT,n);CHKERRQ(ierr);
699     mg[i]->default_smoothd = n;
700   }
701   PetscFunctionReturn(0);
702 }
703 
704 #undef __FUNCT__
705 #define __FUNCT__ "MGSetNumberSmoothUp"
706 /*@
707    MGSetNumberSmoothUp - Sets the number of post-smoothing steps to use
708    on all levels. Use MGGetSmootherUp() to set different numbers of
709    post-smoothing steps on different levels.
710 
711    Collective on PC
712 
713    Input Parameters:
714 +  mg - the multigrid context
715 -  n - the number of smoothing steps
716 
717    Options Database Key:
718 .  -pc_mg_smoothup <n> - Sets number of post-smoothing steps
719 
720    Level: advanced
721 
722    Note: this does not set a value on the coarsest grid, since we assume that
723     there is no seperate smooth up on the coarsest grid. If you want to have a
724     seperate smooth up on the coarsest grid then call MGGetSmoothUp(pc,0,&ksp);
725 
726 .keywords: MG, smooth, up, post-smoothing, steps, multigrid
727 
728 .seealso: MGSetNumberSmoothDown()
729 @*/
730 PetscErrorCode  MGSetNumberSmoothUp(PC pc,PetscInt n)
731 {
732   MG             *mg;
733   PetscErrorCode ierr;
734   PetscInt       i,levels;
735 
736   PetscFunctionBegin;
737   PetscValidHeaderSpecific(pc,PC_COOKIE,1);
738   mg     = (MG*)pc->data;
739   if (!mg) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Must set MG levels before calling");
740   levels = mg[0]->levels;
741 
742   for (i=1; i<levels; i++) {
743     /* make sure smoother up and down are different */
744     ierr = MGGetSmootherUp(pc,i,PETSC_NULL);CHKERRQ(ierr);
745     ierr = KSPSetTolerances(mg[i]->smoothu,PETSC_DEFAULT,PETSC_DEFAULT,PETSC_DEFAULT,n);CHKERRQ(ierr);
746     mg[i]->default_smoothu = n;
747   }
748   PetscFunctionReturn(0);
749 }
750 
751 /* ----------------------------------------------------------------------------------------*/
752 
753 /*MC
754    PCMG - Use geometric multigrid preconditioning. This preconditioner requires you provide additional
755     information about the coarser grid matrices and restriction/interpolation operators.
756 
757    Options Database Keys:
758 +  -pc_mg_levels <nlevels> - number of levels including finest
759 .  -pc_mg_cycles 1 or 2 - for V or W-cycle
760 .  -pc_mg_smoothup <n> - number of smoothing steps after interpolation
761 .  -pc_mg_smoothdown <n> - number of smoothing steps before applying restriction operator
762 .  -pc_mg_type <additive,multiplicative,full,cascade> - multiplicative is the default
763 .  -pc_mg_log - log information about time spent on each level of the solver
764 .  -pc_mg_monitor - print information on the multigrid convergence
765 -  -pc_mg_dump_matlab - dumps the matrices for each level and the restriction/interpolation matrices
766                         to the Socket viewer for reading from Matlab.
767 
768    Notes:
769 
770    Level: intermediate
771 
772    Concepts: multigrid
773 
774 .seealso:  PCCreate(), PCSetType(), PCType (for list of available types), PC, PCMGType,
775            MGSetLevels(), MGGetLevels(), MGSetType(), MPSetCycles(), MGSetNumberSmoothDown(),
776            MGSetNumberSmoothUp(), MGGetCoarseSolve(), MGSetResidual(), MGSetInterpolation(),
777            MGSetRestriction(), MGGetSmoother(), MGGetSmootherUp(), MGGetSmootherDown(),
778            MGSetCyclesOnLevel(), MGSetRhs(), MGSetX(), MGSetR()
779 M*/
780 
781 EXTERN_C_BEGIN
782 #undef __FUNCT__
783 #define __FUNCT__ "PCCreate_MG"
784 PetscErrorCode PCCreate_MG(PC pc)
785 {
786   PetscFunctionBegin;
787   pc->ops->apply          = PCApply_MG;
788   pc->ops->setup          = PCSetUp_MG;
789   pc->ops->destroy        = PCDestroy_MG;
790   pc->ops->setfromoptions = PCSetFromOptions_MG;
791   pc->ops->view           = PCView_MG;
792 
793   pc->data                = (void*)0;
794   PetscFunctionReturn(0);
795 }
796 EXTERN_C_END
797