xref: /petsc/src/ksp/pc/impls/gamg/classical.c (revision bcab024551cbaaaa7f49e357634a3584f91cb6d5) 
1 #include <../src/ksp/pc/impls/gamg/gamg.h>        /*I "petscpc.h" I*/
2 #include <petsc-private/kspimpl.h>
3 #include <petscsf.h>
4 
5 PetscFunctionList PCGAMGClassicalProlongatorList    = NULL;
6 PetscBool         PCGAMGClassicalPackageInitialized = PETSC_FALSE;
7 
8 typedef struct {
9   PetscReal interp_threshold; /* interpolation threshold */
10   char      prolongtype[256];
11   PetscInt  nsmooths;         /* number of jacobi smoothings on the prolongator */
12 } PC_GAMG_Classical;
13 
14 #undef __FUNCT__
15 #define __FUNCT__ "PCGAMGClassicalSetType"
16 /*@C
17    PCGAMGClassicalSetType - Sets the type of classical interpolation to use
18 
19    Collective on PC
20 
21    Input Parameters:
22 .  pc - the preconditioner context
23 
24    Options Database Key:
25 .  -pc_gamg_classical_type
26 
27    Level: intermediate
28 
29 .seealso: ()
30 @*/
31 PetscErrorCode PCGAMGClassicalSetType(PC pc, PCGAMGClassicalType type)
32 {
33   PetscErrorCode ierr;
34 
35   PetscFunctionBegin;
36   PetscValidHeaderSpecific(pc,PC_CLASSID,1);
37   ierr = PetscTryMethod(pc,"PCGAMGClassicalSetType_C",(PC,PCGAMGType),(pc,type));CHKERRQ(ierr);
38   PetscFunctionReturn(0);
39 }
40 
41 #undef __FUNCT__
42 #define __FUNCT__ "PCGAMGClassicalGetType"
43 /*@C
44    PCGAMGClassicalGetType - Gets the type of classical interpolation to use
45 
46    Collective on PC
47 
48    Input Parameter:
49 .  pc - the preconditioner context
50 
51    Output Parameter:
52 .  type - the type used
53 
54    Level: intermediate
55 
56 .seealso: ()
57 @*/
58 PetscErrorCode PCGAMGClassicalGetType(PC pc, PCGAMGClassicalType *type)
59 {
60   PetscErrorCode ierr;
61 
62   PetscFunctionBegin;
63   PetscValidHeaderSpecific(pc,PC_CLASSID,1);
64   ierr = PetscUseMethod(pc,"PCGAMGClassicalGetType_C",(PC,PCGAMGType*),(pc,type));CHKERRQ(ierr);
65   PetscFunctionReturn(0);
66 }
67 
68 #undef __FUNCT__
69 #define __FUNCT__ "PCGAMGClassicalSetType_GAMG"
70 static PetscErrorCode PCGAMGClassicalSetType_GAMG(PC pc, PCGAMGClassicalType type)
71 {
72   PetscErrorCode    ierr;
73   PC_MG             *mg          = (PC_MG*)pc->data;
74   PC_GAMG           *pc_gamg     = (PC_GAMG*)mg->innerctx;
75   PC_GAMG_Classical *cls         = (PC_GAMG_Classical*)pc_gamg->subctx;
76 
77   PetscFunctionBegin;
78   ierr = PetscStrcpy(cls->prolongtype,type);CHKERRQ(ierr);
79   PetscFunctionReturn(0);
80 }
81 
82 #undef __FUNCT__
83 #define __FUNCT__ "PCGAMGClassicalGetType_GAMG"
84 static PetscErrorCode PCGAMGClassicalGetType_GAMG(PC pc, PCGAMGClassicalType *type)
85 {
86   PC_MG             *mg          = (PC_MG*)pc->data;
87   PC_GAMG           *pc_gamg     = (PC_GAMG*)mg->innerctx;
88   PC_GAMG_Classical *cls         = (PC_GAMG_Classical*)pc_gamg->subctx;
89 
90   PetscFunctionBegin;
91   *type = cls->prolongtype;
92   PetscFunctionReturn(0);
93 }
94 
95 #undef __FUNCT__
96 #define __FUNCT__ "PCGAMGGraph_Classical"
97 PetscErrorCode PCGAMGGraph_Classical(PC pc,const Mat A,Mat *G)
98 {
99   PetscInt          s,f,n,idx,lidx,gidx;
100   PetscInt          r,c,ncols;
101   const PetscInt    *rcol;
102   const PetscScalar *rval;
103   PetscInt          *gcol;
104   PetscScalar       *gval;
105   PetscReal         rmax;
106   PetscInt          cmax = 0;
107   PC_MG             *mg;
108   PC_GAMG           *gamg;
109   PetscErrorCode    ierr;
110   PetscInt          *gsparse,*lsparse;
111   PetscScalar       *Amax;
112   MatType           mtype;
113 
114   PetscFunctionBegin;
115   mg   = (PC_MG *)pc->data;
116   gamg = (PC_GAMG *)mg->innerctx;
117 
118   ierr = MatGetOwnershipRange(A,&s,&f);CHKERRQ(ierr);
119   n=f-s;
120   ierr = PetscMalloc1(n,&lsparse);CHKERRQ(ierr);
121   ierr = PetscMalloc1(n,&gsparse);CHKERRQ(ierr);
122   ierr = PetscMalloc1(n,&Amax);CHKERRQ(ierr);
123 
124   for (r = 0;r < n;r++) {
125     lsparse[r] = 0;
126     gsparse[r] = 0;
127   }
128 
129   for (r = s;r < f;r++) {
130     /* determine the maximum off-diagonal in each row */
131     rmax = 0.;
132     ierr = MatGetRow(A,r,&ncols,&rcol,&rval);CHKERRQ(ierr);
133     for (c = 0; c < ncols; c++) {
134       if (PetscRealPart(-rval[c]) > rmax && rcol[c] != r) {
135         rmax = PetscRealPart(-rval[c]);
136       }
137     }
138     Amax[r-s] = rmax;
139     if (ncols > cmax) cmax = ncols;
140     lidx = 0;
141     gidx = 0;
142     /* create the local and global sparsity patterns */
143     for (c = 0; c < ncols; c++) {
144       if (PetscRealPart(-rval[c]) > gamg->threshold*PetscRealPart(Amax[r-s]) || rcol[c] == r) {
145         if (rcol[c] < f && rcol[c] >= s) {
146           lidx++;
147         } else {
148           gidx++;
149         }
150       }
151     }
152     ierr = MatRestoreRow(A,r,&ncols,&rcol,&rval);CHKERRQ(ierr);
153     lsparse[r-s] = lidx;
154     gsparse[r-s] = gidx;
155   }
156   ierr = PetscMalloc1(cmax,&gval);CHKERRQ(ierr);
157   ierr = PetscMalloc1(cmax,&gcol);CHKERRQ(ierr);
158 
159   ierr = MatCreate(PetscObjectComm((PetscObject)A),G); CHKERRQ(ierr);
160   ierr = MatGetType(A,&mtype);CHKERRQ(ierr);
161   ierr = MatSetType(*G,mtype);CHKERRQ(ierr);
162   ierr = MatSetSizes(*G,n,n,PETSC_DETERMINE,PETSC_DETERMINE);CHKERRQ(ierr);
163   ierr = MatMPIAIJSetPreallocation(*G,0,lsparse,0,gsparse);CHKERRQ(ierr);
164   ierr = MatSeqAIJSetPreallocation(*G,0,lsparse);CHKERRQ(ierr);
165   for (r = s;r < f;r++) {
166     ierr = MatGetRow(A,r,&ncols,&rcol,&rval);CHKERRQ(ierr);
167     idx = 0;
168     for (c = 0; c < ncols; c++) {
169       /* classical strength of connection */
170       if (PetscRealPart(-rval[c]) > gamg->threshold*PetscRealPart(Amax[r-s]) || rcol[c] == r) {
171         gcol[idx] = rcol[c];
172         gval[idx] = rval[c];
173         idx++;
174       }
175     }
176     ierr = MatSetValues(*G,1,&r,idx,gcol,gval,INSERT_VALUES);CHKERRQ(ierr);
177     ierr = MatRestoreRow(A,r,&ncols,&rcol,&rval);CHKERRQ(ierr);
178   }
179   ierr = MatAssemblyBegin(*G, MAT_FINAL_ASSEMBLY); CHKERRQ(ierr);
180   ierr = MatAssemblyEnd(*G, MAT_FINAL_ASSEMBLY);CHKERRQ(ierr);
181 
182   ierr = PetscFree(gval);CHKERRQ(ierr);
183   ierr = PetscFree(gcol);CHKERRQ(ierr);
184   ierr = PetscFree(lsparse);CHKERRQ(ierr);
185   ierr = PetscFree(gsparse);CHKERRQ(ierr);
186   ierr = PetscFree(Amax);CHKERRQ(ierr);
187   PetscFunctionReturn(0);
188 }
189 
190 
191 #undef __FUNCT__
192 #define __FUNCT__ "PCGAMGCoarsen_Classical"
193 PetscErrorCode PCGAMGCoarsen_Classical(PC pc,Mat *G,PetscCoarsenData **agg_lists)
194 {
195   PetscErrorCode   ierr;
196   MatCoarsen       crs;
197   MPI_Comm         fcomm = ((PetscObject)pc)->comm;
198 
199   PetscFunctionBegin;
200 
201 
202   /* construct the graph if necessary */
203   if (!G) {
204     SETERRQ(fcomm,PETSC_ERR_ARG_WRONGSTATE,"Must set Graph in PC in PCGAMG before coarsening");
205   }
206 
207   ierr = MatCoarsenCreate(fcomm,&crs);CHKERRQ(ierr);
208   ierr = MatCoarsenSetFromOptions(crs);CHKERRQ(ierr);
209   ierr = MatCoarsenSetAdjacency(crs,*G);CHKERRQ(ierr);
210   ierr = MatCoarsenSetStrictAggs(crs,PETSC_TRUE);CHKERRQ(ierr);
211   ierr = MatCoarsenApply(crs);CHKERRQ(ierr);
212   ierr = MatCoarsenGetData(crs,agg_lists);CHKERRQ(ierr);
213   ierr = MatCoarsenDestroy(&crs);CHKERRQ(ierr);
214 
215   PetscFunctionReturn(0);
216 }
217 
218 #undef __FUNCT__
219 #define __FUNCT__ "PCGAMGProlongator_Classical_Direct"
220 PetscErrorCode PCGAMGProlongator_Classical_Direct(PC pc, const Mat A, const Mat G, PetscCoarsenData *agg_lists,Mat *P)
221 {
222   PetscErrorCode    ierr;
223   PC_MG             *mg          = (PC_MG*)pc->data;
224   PC_GAMG           *gamg        = (PC_GAMG*)mg->innerctx;
225   PetscBool         iscoarse,isMPIAIJ,isSEQAIJ;
226   PetscInt          fn,cn,fs,fe,cs,ce,i,j,ncols,col,row_f,row_c,cmax=0,idx,noff;
227   PetscInt          *lcid,*gcid,*lsparse,*gsparse,*colmap,*pcols;
228   const PetscInt    *rcol;
229   PetscReal         *Amax_pos,*Amax_neg;
230   PetscScalar       g_pos,g_neg,a_pos,a_neg,diag,invdiag,alpha,beta,pij;
231   PetscScalar       *pvals;
232   const PetscScalar *rval;
233   Mat               lA,gA=NULL;
234   MatType           mtype;
235   Vec               C,lvec;
236   PetscLayout       clayout;
237   PetscSF           sf;
238   Mat_MPIAIJ        *mpiaij;
239 
240   PetscFunctionBegin;
241   ierr = MatGetOwnershipRange(A,&fs,&fe);CHKERRQ(ierr);
242   fn = fe-fs;
243   ierr = PetscObjectTypeCompare((PetscObject)A,MATMPIAIJ,&isMPIAIJ);CHKERRQ(ierr);
244   ierr = PetscObjectTypeCompare((PetscObject)A,MATSEQAIJ,&isSEQAIJ);CHKERRQ(ierr);
245   if (!isMPIAIJ && !isSEQAIJ) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONG,"Classical AMG requires MPIAIJ matrix");
246   if (isMPIAIJ) {
247     mpiaij = (Mat_MPIAIJ*)A->data;
248     lA = mpiaij->A;
249     gA = mpiaij->B;
250     lvec = mpiaij->lvec;
251     ierr = VecGetSize(lvec,&noff);CHKERRQ(ierr);
252     colmap = mpiaij->garray;
253     ierr = MatGetLayouts(A,NULL,&clayout);CHKERRQ(ierr);
254     ierr = PetscSFCreate(PetscObjectComm((PetscObject)A),&sf);CHKERRQ(ierr);
255     ierr = PetscSFSetGraphLayout(sf,clayout,noff,NULL,PETSC_COPY_VALUES,colmap);CHKERRQ(ierr);
256     ierr = PetscMalloc1(noff,&gcid);CHKERRQ(ierr);
257   } else {
258     lA = A;
259   }
260   ierr = PetscMalloc1(fn,&lsparse);CHKERRQ(ierr);
261   ierr = PetscMalloc1(fn,&gsparse);CHKERRQ(ierr);
262   ierr = PetscMalloc1(fn,&lcid);CHKERRQ(ierr);
263   ierr = PetscMalloc1(fn,&Amax_pos);CHKERRQ(ierr);
264   ierr = PetscMalloc1(fn,&Amax_neg);CHKERRQ(ierr);
265 
266   /* count the number of coarse unknowns */
267   cn = 0;
268   for (i=0;i<fn;i++) {
269     /* filter out singletons */
270     ierr = PetscCDEmptyAt(agg_lists,i,&iscoarse); CHKERRQ(ierr);
271     lcid[i] = -1;
272     if (!iscoarse) {
273       cn++;
274     }
275   }
276 
277    /* create the coarse vector */
278   ierr = VecCreateMPI(PetscObjectComm((PetscObject)A),cn,PETSC_DECIDE,&C);CHKERRQ(ierr);
279   ierr = VecGetOwnershipRange(C,&cs,&ce);CHKERRQ(ierr);
280 
281   cn = 0;
282   for (i=0;i<fn;i++) {
283     ierr = PetscCDEmptyAt(agg_lists,i,&iscoarse); CHKERRQ(ierr);
284     if (!iscoarse) {
285       lcid[i] = cs+cn;
286       cn++;
287     } else {
288       lcid[i] = -1;
289     }
290   }
291 
292   if (gA) {
293     ierr = PetscSFBcastBegin(sf,MPIU_INT,lcid,gcid);CHKERRQ(ierr);
294     ierr = PetscSFBcastEnd(sf,MPIU_INT,lcid,gcid);CHKERRQ(ierr);
295   }
296 
297   /* determine the biggest off-diagonal entries in each row */
298   for (i=fs;i<fe;i++) {
299     Amax_pos[i-fs] = 0.;
300     Amax_neg[i-fs] = 0.;
301     ierr = MatGetRow(A,i,&ncols,&rcol,&rval);CHKERRQ(ierr);
302     for(j=0;j<ncols;j++){
303       if ((PetscRealPart(-rval[j]) > Amax_neg[i-fs]) && i != rcol[j]) Amax_neg[i-fs] = PetscAbsScalar(rval[j]);
304       if ((PetscRealPart(rval[j])  > Amax_pos[i-fs]) && i != rcol[j]) Amax_pos[i-fs] = PetscAbsScalar(rval[j]);
305     }
306     if (ncols > cmax) cmax = ncols;
307     ierr = MatRestoreRow(A,i,&ncols,&rcol,&rval);CHKERRQ(ierr);
308   }
309   ierr = PetscMalloc1(cmax,&pcols);CHKERRQ(ierr);
310   ierr = PetscMalloc1(cmax,&pvals);CHKERRQ(ierr);
311   ierr = VecDestroy(&C);CHKERRQ(ierr);
312 
313   /* count the on and off processor sparsity patterns for the prolongator */
314   for (i=0;i<fn;i++) {
315     /* on */
316     lsparse[i] = 0;
317     gsparse[i] = 0;
318     if (lcid[i] >= 0) {
319       lsparse[i] = 1;
320       gsparse[i] = 0;
321     } else {
322       ierr = MatGetRow(lA,i,&ncols,&rcol,&rval);CHKERRQ(ierr);
323       for (j = 0;j < ncols;j++) {
324         col = rcol[j];
325         if (lcid[col] >= 0 && (PetscRealPart(rval[j]) > gamg->threshold*Amax_pos[i] || PetscRealPart(-rval[j]) > gamg->threshold*Amax_neg[i])) {
326           lsparse[i] += 1;
327         }
328       }
329       ierr = MatRestoreRow(lA,i,&ncols,&rcol,&rval);CHKERRQ(ierr);
330       /* off */
331       if (gA) {
332         ierr = MatGetRow(gA,i,&ncols,&rcol,&rval);CHKERRQ(ierr);
333         for (j = 0; j < ncols; j++) {
334           col = rcol[j];
335           if (gcid[col] >= 0 && (PetscRealPart(rval[j]) > gamg->threshold*Amax_pos[i] || PetscRealPart(-rval[j]) > gamg->threshold*Amax_neg[i])) {
336             gsparse[i] += 1;
337           }
338         }
339         ierr = MatRestoreRow(gA,i,&ncols,&rcol,&rval);CHKERRQ(ierr);
340       }
341     }
342   }
343 
344   /* preallocate and create the prolongator */
345   ierr = MatCreate(PetscObjectComm((PetscObject)A),P); CHKERRQ(ierr);
346   ierr = MatGetType(G,&mtype);CHKERRQ(ierr);
347   ierr = MatSetType(*P,mtype);CHKERRQ(ierr);
348 
349   ierr = MatSetSizes(*P,fn,cn,PETSC_DETERMINE,PETSC_DETERMINE);CHKERRQ(ierr);
350   ierr = MatMPIAIJSetPreallocation(*P,0,lsparse,0,gsparse);CHKERRQ(ierr);
351   ierr = MatSeqAIJSetPreallocation(*P,0,lsparse);CHKERRQ(ierr);
352 
353   /* loop over local fine nodes -- get the diagonal, the sum of positive and negative strong and weak weights, and set up the row */
354   for (i = 0;i < fn;i++) {
355     /* determine on or off */
356     row_f = i + fs;
357     row_c = lcid[i];
358     if (row_c >= 0) {
359       pij = 1.;
360       ierr = MatSetValues(*P,1,&row_f,1,&row_c,&pij,INSERT_VALUES);CHKERRQ(ierr);
361     } else {
362       g_pos = 0.;
363       g_neg = 0.;
364       a_pos = 0.;
365       a_neg = 0.;
366       diag = 0.;
367 
368       /* local connections */
369       ierr = MatGetRow(lA,i,&ncols,&rcol,&rval);CHKERRQ(ierr);
370       for (j = 0; j < ncols; j++) {
371         col = rcol[j];
372         if (lcid[col] >= 0 && (PetscRealPart(rval[j]) > gamg->threshold*Amax_pos[i] || PetscRealPart(-rval[j]) > gamg->threshold*Amax_neg[i])) {
373           if (PetscRealPart(rval[j]) > 0.) {
374             g_pos += rval[j];
375           } else {
376             g_neg += rval[j];
377           }
378         }
379         if (col != i) {
380           if (PetscRealPart(rval[j]) > 0.) {
381             a_pos += rval[j];
382           } else {
383             a_neg += rval[j];
384           }
385         } else {
386           diag = rval[j];
387         }
388       }
389       ierr = MatRestoreRow(lA,i,&ncols,&rcol,&rval);CHKERRQ(ierr);
390 
391       /* ghosted connections */
392       if (gA) {
393         ierr = MatGetRow(gA,i,&ncols,&rcol,&rval);CHKERRQ(ierr);
394         for (j = 0; j < ncols; j++) {
395           col = rcol[j];
396           if (gcid[col] >= 0 && (PetscRealPart(rval[j]) > gamg->threshold*Amax_pos[i] || PetscRealPart(-rval[j]) > gamg->threshold*Amax_neg[i])) {
397             if (PetscRealPart(rval[j]) > 0.) {
398               g_pos += rval[j];
399             } else {
400               g_neg += rval[j];
401             }
402           }
403           if (PetscRealPart(rval[j]) > 0.) {
404             a_pos += rval[j];
405           } else {
406             a_neg += rval[j];
407           }
408         }
409         ierr = MatRestoreRow(gA,i,&ncols,&rcol,&rval);CHKERRQ(ierr);
410       }
411 
412       if (g_neg == 0.) {
413         alpha = 0.;
414       } else {
415         alpha = -a_neg/g_neg;
416       }
417 
418       if (g_pos == 0.) {
419         diag += a_pos;
420         beta = 0.;
421       } else {
422         beta = -a_pos/g_pos;
423       }
424       if (diag == 0.) {
425         invdiag = 0.;
426       } else invdiag = 1. / diag;
427       /* on */
428       ierr = MatGetRow(lA,i,&ncols,&rcol,&rval);CHKERRQ(ierr);
429       idx = 0;
430       for (j = 0;j < ncols;j++) {
431         col = rcol[j];
432         if (lcid[col] >= 0 && (PetscRealPart(rval[j]) > gamg->threshold*Amax_pos[i] || PetscRealPart(-rval[j]) > gamg->threshold*Amax_neg[i])) {
433           row_f = i + fs;
434           row_c = lcid[col];
435           /* set the values for on-processor ones */
436           if (PetscRealPart(rval[j]) < 0.) {
437             pij = rval[j]*alpha*invdiag;
438           } else {
439             pij = rval[j]*beta*invdiag;
440           }
441           if (PetscAbsScalar(pij) != 0.) {
442             pvals[idx] = pij;
443             pcols[idx] = row_c;
444             idx++;
445           }
446         }
447       }
448       ierr = MatRestoreRow(lA,i,&ncols,&rcol,&rval);CHKERRQ(ierr);
449       /* off */
450       if (gA) {
451         ierr = MatGetRow(gA,i,&ncols,&rcol,&rval);CHKERRQ(ierr);
452         for (j = 0; j < ncols; j++) {
453           col = rcol[j];
454           if (gcid[col] >= 0 && (PetscRealPart(rval[j]) > gamg->threshold*Amax_pos[i] || PetscRealPart(-rval[j]) > gamg->threshold*Amax_neg[i])) {
455             row_f = i + fs;
456             row_c = gcid[col];
457             /* set the values for on-processor ones */
458             if (PetscRealPart(rval[j]) < 0.) {
459               pij = rval[j]*alpha*invdiag;
460             } else {
461               pij = rval[j]*beta*invdiag;
462             }
463             if (PetscAbsScalar(pij) != 0.) {
464               pvals[idx] = pij;
465               pcols[idx] = row_c;
466               idx++;
467             }
468           }
469         }
470         ierr = MatRestoreRow(gA,i,&ncols,&rcol,&rval);CHKERRQ(ierr);
471       }
472       ierr = MatSetValues(*P,1,&row_f,idx,pcols,pvals,INSERT_VALUES);CHKERRQ(ierr);
473     }
474   }
475 
476   ierr = MatAssemblyBegin(*P, MAT_FINAL_ASSEMBLY);CHKERRQ(ierr);
477   ierr = MatAssemblyEnd(*P, MAT_FINAL_ASSEMBLY);CHKERRQ(ierr);
478 
479   ierr = PetscFree(lsparse);CHKERRQ(ierr);
480   ierr = PetscFree(gsparse);CHKERRQ(ierr);
481   ierr = PetscFree(pcols);CHKERRQ(ierr);
482   ierr = PetscFree(pvals);CHKERRQ(ierr);
483   ierr = PetscFree(Amax_pos);CHKERRQ(ierr);
484   ierr = PetscFree(Amax_neg);CHKERRQ(ierr);
485   ierr = PetscFree(lcid);CHKERRQ(ierr);
486   if (gA) {
487     ierr = PetscSFDestroy(&sf);CHKERRQ(ierr);
488     ierr = PetscFree(gcid);CHKERRQ(ierr);
489   }
490 
491   PetscFunctionReturn(0);
492 }
493 
494 #undef __FUNCT__
495 #define __FUNCT__ "PCGAMGTruncateProlongator_Private"
496 PetscErrorCode PCGAMGTruncateProlongator_Private(PC pc,Mat *P)
497 {
498   PetscInt          j,i,ps,pf,pn,pcs,pcf,pcn,idx,cmax;
499   PetscErrorCode    ierr;
500   const PetscScalar *pval;
501   const PetscInt    *pcol;
502   PetscScalar       *pnval;
503   PetscInt          *pncol;
504   PetscInt          ncols;
505   Mat               Pnew;
506   PetscInt          *lsparse,*gsparse;
507   PetscReal         pmax_pos,pmax_neg,ptot_pos,ptot_neg,pthresh_pos,pthresh_neg;
508   PC_MG             *mg          = (PC_MG*)pc->data;
509   PC_GAMG           *pc_gamg     = (PC_GAMG*)mg->innerctx;
510   PC_GAMG_Classical *cls         = (PC_GAMG_Classical*)pc_gamg->subctx;
511 
512   PetscFunctionBegin;
513   /* trim and rescale with reallocation */
514   ierr = MatGetOwnershipRange(*P,&ps,&pf);CHKERRQ(ierr);
515   ierr = MatGetOwnershipRangeColumn(*P,&pcs,&pcf);CHKERRQ(ierr);
516   pn = pf-ps;
517   pcn = pcf-pcs;
518   ierr = PetscMalloc1(pn,&lsparse);CHKERRQ(ierr);
519   ierr = PetscMalloc1(pn,&gsparse);CHKERRQ(ierr);
520   /* allocate */
521   cmax = 0;
522   for (i=ps;i<pf;i++) {
523     lsparse[i-ps] = 0;
524     gsparse[i-ps] = 0;
525     ierr = MatGetRow(*P,i,&ncols,&pcol,&pval);CHKERRQ(ierr);
526     if (ncols > cmax) {
527       cmax = ncols;
528     }
529     pmax_pos = 0.;
530     pmax_neg = 0.;
531     for (j=0;j<ncols;j++) {
532       if (PetscRealPart(pval[j]) > pmax_pos) {
533         pmax_pos = PetscRealPart(pval[j]);
534       } else if (PetscRealPart(pval[j]) < pmax_neg) {
535         pmax_neg = PetscRealPart(pval[j]);
536       }
537     }
538     for (j=0;j<ncols;j++) {
539       if (PetscRealPart(pval[j]) >= pmax_pos*cls->interp_threshold || PetscRealPart(pval[j]) <= pmax_neg*cls->interp_threshold) {
540         if (pcol[j] >= pcs && pcol[j] < pcf) {
541           lsparse[i-ps]++;
542         } else {
543           gsparse[i-ps]++;
544         }
545       }
546     }
547     ierr = MatRestoreRow(*P,i,&ncols,&pcol,&pval);CHKERRQ(ierr);
548   }
549 
550   ierr = PetscMalloc1(cmax,&pnval);CHKERRQ(ierr);
551   ierr = PetscMalloc1(cmax,&pncol);CHKERRQ(ierr);
552 
553   ierr = MatCreate(PetscObjectComm((PetscObject)*P),&Pnew);CHKERRQ(ierr);
554   ierr = MatSetType(Pnew, MATAIJ);CHKERRQ(ierr);
555   ierr = MatSetSizes(Pnew,pn,pcn,PETSC_DETERMINE,PETSC_DETERMINE);CHKERRQ(ierr);
556   ierr = MatSeqAIJSetPreallocation(Pnew,0,lsparse);CHKERRQ(ierr);
557   ierr = MatMPIAIJSetPreallocation(Pnew,0,lsparse,0,gsparse);CHKERRQ(ierr);
558 
559   for (i=ps;i<pf;i++) {
560     ierr = MatGetRow(*P,i,&ncols,&pcol,&pval);CHKERRQ(ierr);
561     pmax_pos = 0.;
562     pmax_neg = 0.;
563     for (j=0;j<ncols;j++) {
564       if (PetscRealPart(pval[j]) > pmax_pos) {
565         pmax_pos = PetscRealPart(pval[j]);
566       } else if (PetscRealPart(pval[j]) < pmax_neg) {
567         pmax_neg = PetscRealPart(pval[j]);
568       }
569     }
570     pthresh_pos = 0.;
571     pthresh_neg = 0.;
572     ptot_pos = 0.;
573     ptot_neg = 0.;
574     for (j=0;j<ncols;j++) {
575       if (PetscRealPart(pval[j]) >= cls->interp_threshold*pmax_pos) {
576         pthresh_pos += PetscRealPart(pval[j]);
577       } else if (PetscRealPart(pval[j]) <= cls->interp_threshold*pmax_neg) {
578         pthresh_neg += PetscRealPart(pval[j]);
579       }
580       if (PetscRealPart(pval[j]) > 0.) {
581         ptot_pos += PetscRealPart(pval[j]);
582       } else {
583         ptot_neg += PetscRealPart(pval[j]);
584       }
585     }
586     if (PetscAbsReal(pthresh_pos) > 0.) ptot_pos /= pthresh_pos;
587     if (PetscAbsReal(pthresh_neg) > 0.) ptot_neg /= pthresh_neg;
588     idx=0;
589     for (j=0;j<ncols;j++) {
590       if (PetscRealPart(pval[j]) >= pmax_pos*cls->interp_threshold) {
591         pnval[idx] = ptot_pos*pval[j];
592         pncol[idx] = pcol[j];
593         idx++;
594       } else if (PetscRealPart(pval[j]) <= pmax_neg*cls->interp_threshold) {
595         pnval[idx] = ptot_neg*pval[j];
596         pncol[idx] = pcol[j];
597         idx++;
598       }
599     }
600     ierr = MatRestoreRow(*P,i,&ncols,&pcol,&pval);CHKERRQ(ierr);
601     ierr = MatSetValues(Pnew,1,&i,idx,pncol,pnval,INSERT_VALUES);CHKERRQ(ierr);
602   }
603 
604   ierr = MatAssemblyBegin(Pnew, MAT_FINAL_ASSEMBLY);CHKERRQ(ierr);
605   ierr = MatAssemblyEnd(Pnew, MAT_FINAL_ASSEMBLY);CHKERRQ(ierr);
606   ierr = MatDestroy(P);CHKERRQ(ierr);
607 
608   *P = Pnew;
609   ierr = PetscFree(lsparse);CHKERRQ(ierr);
610   ierr = PetscFree(gsparse);CHKERRQ(ierr);
611   ierr = PetscFree(pncol);CHKERRQ(ierr);
612   ierr = PetscFree(pnval);CHKERRQ(ierr);
613   PetscFunctionReturn(0);
614 }
615 
616 #undef __FUNCT__
617 #define __FUNCT__ "PCGAMGProlongator_Classical_Standard"
618 PetscErrorCode PCGAMGProlongator_Classical_Standard(PC pc, const Mat A, const Mat G, PetscCoarsenData *agg_lists,Mat *P)
619 {
620   PetscErrorCode    ierr;
621   Mat               lA,*lAs;
622   MatType           mtype;
623   Vec               cv;
624   PetscInt          *gcid,*lcid,*lsparse,*gsparse,*picol;
625   PetscInt          fs,fe,cs,ce,nl,i,j,k,li,lni,ci,ncols,maxcols,fn,cn,cid;
626   PetscMPIInt       size;
627   const PetscInt    *lidx,*icol,*gidx;
628   PetscBool         iscoarse;
629   PetscScalar       vi,pentry,pjentry;
630   PetscScalar       *pcontrib,*pvcol;
631   const PetscScalar *vcol;
632   PetscReal         diag,jdiag,jwttotal;
633   PetscInt          pncols;
634   PetscSF           sf;
635   PetscLayout       clayout;
636   IS                lis;
637 
638   PetscFunctionBegin;
639   ierr = MPI_Comm_size(PetscObjectComm((PetscObject)A),&size);CHKERRQ(ierr);
640   ierr = MatGetOwnershipRange(A,&fs,&fe);CHKERRQ(ierr);
641   fn = fe-fs;
642   ierr = ISCreateStride(PETSC_COMM_SELF,fe-fs,fs,1,&lis);CHKERRQ(ierr);
643   if (size > 1) {
644     ierr = MatGetLayouts(A,NULL,&clayout);CHKERRQ(ierr);
645     /* increase the overlap by two to get neighbors of neighbors */
646     ierr = MatIncreaseOverlap(A,1,&lis,2);CHKERRQ(ierr);
647     ierr = ISSort(lis);CHKERRQ(ierr);
648     /* get the local part of A */
649     ierr = MatGetSubMatrices(A,1,&lis,&lis,MAT_INITIAL_MATRIX,&lAs);CHKERRQ(ierr);
650     lA = lAs[0];
651     /* build an SF out of it */
652     ierr = ISGetLocalSize(lis,&nl);CHKERRQ(ierr);
653     ierr = PetscSFCreate(PetscObjectComm((PetscObject)A),&sf);CHKERRQ(ierr);
654     ierr = ISGetIndices(lis,&lidx);CHKERRQ(ierr);
655     ierr = PetscSFSetGraphLayout(sf,clayout,nl,NULL,PETSC_COPY_VALUES,lidx);CHKERRQ(ierr);
656     ierr = ISRestoreIndices(lis,&lidx);CHKERRQ(ierr);
657   } else {
658     lA = A;
659     nl = fn;
660   }
661   /* create a communication structure for the overlapped portion and transmit coarse indices */
662   ierr = PetscMalloc1(fn,&lsparse);CHKERRQ(ierr);
663   ierr = PetscMalloc1(fn,&gsparse);CHKERRQ(ierr);
664   ierr = PetscMalloc1(nl,&pcontrib);CHKERRQ(ierr);
665   /* create coarse vector */
666   cn = 0;
667   for (i=0;i<fn;i++) {
668     ierr = PetscCDEmptyAt(agg_lists,i,&iscoarse);CHKERRQ(ierr);
669     if (!iscoarse) {
670       cn++;
671     }
672   }
673   ierr = PetscMalloc1(fn,&gcid);CHKERRQ(ierr);
674   ierr = VecCreateMPI(PetscObjectComm((PetscObject)A),cn,PETSC_DECIDE,&cv);CHKERRQ(ierr);
675   ierr = VecGetOwnershipRange(cv,&cs,&ce);CHKERRQ(ierr);
676   cn = 0;
677   for (i=0;i<fn;i++) {
678     ierr = PetscCDEmptyAt(agg_lists,i,&iscoarse); CHKERRQ(ierr);
679     if (!iscoarse) {
680       gcid[i] = cs+cn;
681       cn++;
682     } else {
683       gcid[i] = -1;
684     }
685   }
686   if (size > 1) {
687     ierr = PetscMalloc1(nl,&lcid);CHKERRQ(ierr);
688     ierr = PetscSFBcastBegin(sf,MPIU_INT,gcid,lcid);CHKERRQ(ierr);
689     ierr = PetscSFBcastEnd(sf,MPIU_INT,gcid,lcid);CHKERRQ(ierr);
690   } else {
691     lcid = gcid;
692   }
693   /* count to preallocate the prolongator */
694   ierr = ISGetIndices(lis,&gidx);CHKERRQ(ierr);
695   maxcols = 0;
696   /* count the number of unique contributing coarse cells for each fine */
697   for (i=0;i<nl;i++) {
698     pcontrib[i] = 0.;
699     ierr = MatGetRow(lA,i,&ncols,&icol,NULL);CHKERRQ(ierr);
700     if (gidx[i] >= fs && gidx[i] < fe) {
701       li = gidx[i] - fs;
702       lsparse[li] = 0;
703       gsparse[li] = 0;
704       cid = lcid[i];
705       if (cid >= 0) {
706         lsparse[li] = 1;
707       } else {
708         for (j=0;j<ncols;j++) {
709           if (lcid[icol[j]] >= 0) {
710             pcontrib[icol[j]] = 1.;
711           } else {
712             ci = icol[j];
713             ierr = MatRestoreRow(lA,i,&ncols,&icol,NULL);CHKERRQ(ierr);
714             ierr = MatGetRow(lA,ci,&ncols,&icol,NULL);CHKERRQ(ierr);
715             for (k=0;k<ncols;k++) {
716               if (lcid[icol[k]] >= 0) {
717                 pcontrib[icol[k]] = 1.;
718               }
719             }
720             ierr = MatRestoreRow(lA,ci,&ncols,&icol,NULL);CHKERRQ(ierr);
721             ierr = MatGetRow(lA,i,&ncols,&icol,NULL);CHKERRQ(ierr);
722           }
723         }
724         for (j=0;j<ncols;j++) {
725           if (lcid[icol[j]] >= 0 && pcontrib[icol[j]] != 0.) {
726             lni = lcid[icol[j]];
727             if (lni >= cs && lni < ce) {
728               lsparse[li]++;
729             } else {
730               gsparse[li]++;
731             }
732             pcontrib[icol[j]] = 0.;
733           } else {
734             ci = icol[j];
735             ierr = MatRestoreRow(lA,i,&ncols,&icol,NULL);CHKERRQ(ierr);
736             ierr = MatGetRow(lA,ci,&ncols,&icol,NULL);CHKERRQ(ierr);
737             for (k=0;k<ncols;k++) {
738               if (lcid[icol[k]] >= 0 && pcontrib[icol[k]] != 0.) {
739                 lni = lcid[icol[k]];
740                 if (lni >= cs && lni < ce) {
741                   lsparse[li]++;
742                 } else {
743                   gsparse[li]++;
744                 }
745                 pcontrib[icol[k]] = 0.;
746               }
747             }
748             ierr = MatRestoreRow(lA,ci,&ncols,&icol,NULL);CHKERRQ(ierr);
749             ierr = MatGetRow(lA,i,&ncols,&icol,NULL);CHKERRQ(ierr);
750           }
751         }
752       }
753       if (lsparse[li] + gsparse[li] > maxcols) maxcols = lsparse[li]+gsparse[li];
754     }
755     ierr = MatRestoreRow(lA,i,&ncols,&icol,&vcol);CHKERRQ(ierr);
756   }
757   ierr = PetscMalloc1(maxcols,&picol);CHKERRQ(ierr);
758   ierr = PetscMalloc1(maxcols,&pvcol);CHKERRQ(ierr);
759   ierr = MatCreate(PetscObjectComm((PetscObject)A),P);CHKERRQ(ierr);
760   ierr = MatGetType(A,&mtype);CHKERRQ(ierr);
761   ierr = MatSetType(*P,mtype);CHKERRQ(ierr);
762   ierr = MatSetSizes(*P,fn,cn,PETSC_DETERMINE,PETSC_DETERMINE);CHKERRQ(ierr);
763   ierr = MatMPIAIJSetPreallocation(*P,0,lsparse,0,gsparse);CHKERRQ(ierr);
764   ierr = MatSeqAIJSetPreallocation(*P,0,lsparse);CHKERRQ(ierr);
765   for (i=0;i<nl;i++) {
766     diag = 0.;
767     if (gidx[i] >= fs && gidx[i] < fe) {
768       li = gidx[i] - fs;
769       pncols=0;
770       cid = lcid[i];
771       if (cid >= 0) {
772         pncols = 1;
773         picol[0] = cid;
774         pvcol[0] = 1.;
775       } else {
776         ierr = MatGetRow(lA,i,&ncols,&icol,&vcol);CHKERRQ(ierr);
777         for (j=0;j<ncols;j++) {
778           pentry = vcol[j];
779           if (lcid[icol[j]] >= 0) {
780             /* coarse neighbor */
781             pcontrib[icol[j]] += pentry;
782           } else if (icol[j] != i) {
783             /* the neighbor is a strongly connected fine node */
784             ci = icol[j];
785             vi = vcol[j];
786             ierr = MatRestoreRow(lA,i,&ncols,&icol,&vcol);CHKERRQ(ierr);
787             ierr = MatGetRow(lA,ci,&ncols,&icol,&vcol);CHKERRQ(ierr);
788             jwttotal=0.;
789             jdiag = 0.;
790             for (k=0;k<ncols;k++) {
791               if (ci == icol[k]) {
792                 jdiag = PetscRealPart(vcol[k]);
793               }
794             }
795             for (k=0;k<ncols;k++) {
796               if (lcid[icol[k]] >= 0 && jdiag*PetscRealPart(vcol[k]) < 0.) {
797                 pjentry = vcol[k];
798                 jwttotal += PetscRealPart(pjentry);
799               }
800             }
801             if (jwttotal != 0.) {
802               jwttotal = PetscRealPart(vi)/jwttotal;
803               for (k=0;k<ncols;k++) {
804                 if (lcid[icol[k]] >= 0 && jdiag*PetscRealPart(vcol[k]) < 0.) {
805                   pjentry = vcol[k]*jwttotal;
806                   pcontrib[icol[k]] += pjentry;
807                 }
808               }
809             } else {
810               diag += PetscRealPart(vi);
811             }
812             ierr = MatRestoreRow(lA,ci,&ncols,&icol,&vcol);CHKERRQ(ierr);
813             ierr = MatGetRow(lA,i,&ncols,&icol,&vcol);CHKERRQ(ierr);
814           } else {
815             diag += PetscRealPart(vcol[j]);
816           }
817         }
818         if (diag != 0.) {
819           diag = 1./diag;
820           for (j=0;j<ncols;j++) {
821             if (lcid[icol[j]] >= 0 && pcontrib[icol[j]] != 0.) {
822               /* the neighbor is a coarse node */
823               if (PetscAbsScalar(pcontrib[icol[j]]) > 0.0) {
824                 lni = lcid[icol[j]];
825                 pvcol[pncols] = -pcontrib[icol[j]]*diag;
826                 picol[pncols] = lni;
827                 pncols++;
828               }
829               pcontrib[icol[j]] = 0.;
830             } else {
831               /* the neighbor is a strongly connected fine node */
832               ci = icol[j];
833               ierr = MatRestoreRow(lA,i,&ncols,&icol,&vcol);CHKERRQ(ierr);
834               ierr = MatGetRow(lA,ci,&ncols,&icol,&vcol);CHKERRQ(ierr);
835               for (k=0;k<ncols;k++) {
836                 if (lcid[icol[k]] >= 0 && pcontrib[icol[k]] != 0.) {
837                   if (PetscAbsScalar(pcontrib[icol[k]]) > 0.0) {
838                     lni = lcid[icol[k]];
839                     pvcol[pncols] = -pcontrib[icol[k]]*diag;
840                     picol[pncols] = lni;
841                     pncols++;
842                   }
843                   pcontrib[icol[k]] = 0.;
844                 }
845               }
846               ierr = MatRestoreRow(lA,ci,&ncols,&icol,&vcol);CHKERRQ(ierr);
847               ierr = MatGetRow(lA,i,&ncols,&icol,&vcol);CHKERRQ(ierr);
848             }
849             pcontrib[icol[j]] = 0.;
850           }
851           ierr = MatRestoreRow(lA,i,&ncols,&icol,&vcol);CHKERRQ(ierr);
852         }
853       }
854       ci = gidx[i];
855       li = gidx[i] - fs;
856       if (pncols > 0) {
857         ierr = MatSetValues(*P,1,&ci,pncols,picol,pvcol,INSERT_VALUES);CHKERRQ(ierr);
858       }
859     }
860   }
861   ierr = ISRestoreIndices(lis,&gidx);CHKERRQ(ierr);
862   ierr = PetscFree(pcontrib);CHKERRQ(ierr);
863   ierr = PetscFree(picol);CHKERRQ(ierr);
864   ierr = PetscFree(pvcol);CHKERRQ(ierr);
865   ierr = PetscFree(lsparse);CHKERRQ(ierr);
866   ierr = PetscFree(gsparse);CHKERRQ(ierr);
867   ierr = ISDestroy(&lis);CHKERRQ(ierr);
868   ierr = PetscFree(gcid);CHKERRQ(ierr);
869   if (size > 1) {
870     ierr = PetscFree(lcid);CHKERRQ(ierr);
871     ierr = MatDestroyMatrices(1,&lAs);CHKERRQ(ierr);
872     ierr = PetscSFDestroy(&sf);CHKERRQ(ierr);
873   }
874   ierr = VecDestroy(&cv);CHKERRQ(ierr);
875   ierr = MatAssemblyBegin(*P, MAT_FINAL_ASSEMBLY);CHKERRQ(ierr);
876   ierr = MatAssemblyEnd(*P, MAT_FINAL_ASSEMBLY);CHKERRQ(ierr);
877   PetscFunctionReturn(0);
878 }
879 
880 #undef __FUNCT__
881 #define __FUNCT__ "PCGAMGOptProl_Classical_Jacobi"
882 PetscErrorCode PCGAMGOptProl_Classical_Jacobi(PC pc,const Mat A,Mat *P)
883 {
884 
885   PetscErrorCode    ierr;
886   PetscInt          f,s,n,cf,cs,i,idx;
887   PetscInt          *coarserows;
888   PetscInt          ncols;
889   const PetscInt    *pcols;
890   const PetscScalar *pvals;
891   Mat               Pnew;
892   Vec               diag;
893   PC_MG             *mg          = (PC_MG*)pc->data;
894   PC_GAMG           *pc_gamg     = (PC_GAMG*)mg->innerctx;
895   PC_GAMG_Classical *cls         = (PC_GAMG_Classical*)pc_gamg->subctx;
896 
897   PetscFunctionBegin;
898   if (cls->nsmooths == 0) {
899     ierr = PCGAMGTruncateProlongator_Private(pc,P);CHKERRQ(ierr);
900     PetscFunctionReturn(0);
901   }
902   ierr = MatGetOwnershipRange(*P,&s,&f);CHKERRQ(ierr);
903   n = f-s;
904   ierr = MatGetOwnershipRangeColumn(*P,&cs,&cf);CHKERRQ(ierr);
905   ierr = PetscMalloc1(n,&coarserows);CHKERRQ(ierr);
906   /* identify the rows corresponding to coarse unknowns */
907   idx = 0;
908   for (i=s;i<f;i++) {
909     ierr = MatGetRow(*P,i,&ncols,&pcols,&pvals);CHKERRQ(ierr);
910     /* assume, for now, that it's a coarse unknown if it has a single unit entry */
911     if (ncols == 1) {
912       if (pvals[0] == 1.) {
913         coarserows[idx] = i;
914         idx++;
915       }
916     }
917     ierr = MatRestoreRow(*P,i,&ncols,&pcols,&pvals);CHKERRQ(ierr);
918   }
919   ierr = MatCreateVecs(A,&diag,0);CHKERRQ(ierr);
920   ierr = MatGetDiagonal(A,diag);CHKERRQ(ierr);
921   ierr = VecReciprocal(diag);CHKERRQ(ierr);
922   for (i=0;i<cls->nsmooths;i++) {
923     ierr = MatMatMult(A,*P,MAT_INITIAL_MATRIX,PETSC_DEFAULT,&Pnew);CHKERRQ(ierr);
924     ierr = MatZeroRows(Pnew,idx,coarserows,0.,NULL,NULL);CHKERRQ(ierr);
925     ierr = MatDiagonalScale(Pnew,diag,0);CHKERRQ(ierr);
926     ierr = MatAYPX(Pnew,-1.0,*P,DIFFERENT_NONZERO_PATTERN);CHKERRQ(ierr);
927     ierr = MatDestroy(P);CHKERRQ(ierr);
928     *P  = Pnew;
929     Pnew = NULL;
930   }
931   ierr = VecDestroy(&diag);CHKERRQ(ierr);
932   ierr = PetscFree(coarserows);CHKERRQ(ierr);
933   ierr = PCGAMGTruncateProlongator_Private(pc,P);CHKERRQ(ierr);
934   PetscFunctionReturn(0);
935 }
936 
937 #undef __FUNCT__
938 #define __FUNCT__ "PCGAMGProlongator_Classical"
939 PetscErrorCode PCGAMGProlongator_Classical(PC pc, const Mat A, const Mat G, PetscCoarsenData *agg_lists,Mat *P)
940 {
941   PetscErrorCode    ierr;
942   PetscErrorCode    (*f)(PC,Mat,Mat,PetscCoarsenData*,Mat*);
943   PC_MG             *mg          = (PC_MG*)pc->data;
944   PC_GAMG           *pc_gamg     = (PC_GAMG*)mg->innerctx;
945   PC_GAMG_Classical *cls         = (PC_GAMG_Classical*)pc_gamg->subctx;
946 
947   PetscFunctionBegin;
948   ierr = PetscFunctionListFind(PCGAMGClassicalProlongatorList,cls->prolongtype,&f);CHKERRQ(ierr);
949   if (!f)SETERRQ(PetscObjectComm((PetscObject)pc),PETSC_ERR_ARG_WRONGSTATE,"Cannot find PCGAMG Classical prolongator type");
950   ierr = (*f)(pc,A,G,agg_lists,P);CHKERRQ(ierr);
951   PetscFunctionReturn(0);
952 }
953 
954 #undef __FUNCT__
955 #define __FUNCT__ "PCGAMGDestroy_Classical"
956 PetscErrorCode PCGAMGDestroy_Classical(PC pc)
957 {
958   PetscErrorCode ierr;
959   PC_MG          *mg          = (PC_MG*)pc->data;
960   PC_GAMG        *pc_gamg     = (PC_GAMG*)mg->innerctx;
961 
962   PetscFunctionBegin;
963   ierr = PetscFree(pc_gamg->subctx);CHKERRQ(ierr);
964   ierr = PetscObjectComposeFunction((PetscObject)pc,"PCGAMGClassicalSetType_C",NULL);CHKERRQ(ierr);
965   ierr = PetscObjectComposeFunction((PetscObject)pc,"PCGAMGClassicalGetType_C",NULL);CHKERRQ(ierr);
966   PetscFunctionReturn(0);
967 }
968 
969 #undef __FUNCT__
970 #define __FUNCT__ "PCGAMGSetFromOptions_Classical"
971 PetscErrorCode PCGAMGSetFromOptions_Classical(PC pc)
972 {
973   PC_MG             *mg          = (PC_MG*)pc->data;
974   PC_GAMG           *pc_gamg     = (PC_GAMG*)mg->innerctx;
975   PC_GAMG_Classical *cls         = (PC_GAMG_Classical*)pc_gamg->subctx;
976   char              tname[256];
977   PetscErrorCode    ierr;
978   PetscBool         flg;
979 
980   PetscFunctionBegin;
981   ierr = PetscOptionsHead("GAMG-Classical options");CHKERRQ(ierr);
982   ierr = PetscOptionsFList("-pc_gamg_classical_type","Type of Classical AMG prolongation","PCGAMGClassicalSetType",PCGAMGClassicalProlongatorList,cls->prolongtype, tname, sizeof(tname), &flg);CHKERRQ(ierr);
983   if (flg) {
984     ierr = PCGAMGClassicalSetType(pc,tname);CHKERRQ(ierr);
985   }
986   ierr = PetscOptionsReal("-pc_gamg_classical_interp_threshold","Threshold for classical interpolator entries","",cls->interp_threshold,&cls->interp_threshold,NULL);CHKERRQ(ierr);
987   ierr = PetscOptionsInt("-pc_gamg_classical_nsmooths","Threshold for classical interpolator entries","",cls->nsmooths,&cls->nsmooths,NULL);CHKERRQ(ierr);
988   ierr = PetscOptionsTail();CHKERRQ(ierr);
989   PetscFunctionReturn(0);
990 }
991 
992 #undef __FUNCT__
993 #define __FUNCT__ "PCGAMGSetData_Classical"
994 PetscErrorCode PCGAMGSetData_Classical(PC pc, Mat A)
995 {
996   PC_MG          *mg      = (PC_MG*)pc->data;
997   PC_GAMG        *pc_gamg = (PC_GAMG*)mg->innerctx;
998 
999   PetscFunctionBegin;
1000   /* no data for classical AMG */
1001   pc_gamg->data = NULL;
1002   pc_gamg->data_cell_cols = 0;
1003   pc_gamg->data_cell_rows = 0;
1004   pc_gamg->data_sz        = 0;
1005   PetscFunctionReturn(0);
1006 }
1007 
1008 
1009 #undef __FUNCT__
1010 #define __FUNCT__ "PCGAMGClassicalFinalizePackage"
1011 PetscErrorCode PCGAMGClassicalFinalizePackage(void)
1012 {
1013   PetscErrorCode ierr;
1014 
1015   PetscFunctionBegin;
1016   PCGAMGClassicalPackageInitialized = PETSC_FALSE;
1017   ierr = PetscFunctionListDestroy(&PCGAMGClassicalProlongatorList);CHKERRQ(ierr);
1018   PetscFunctionReturn(0);
1019 }
1020 
1021 #undef __FUNCT__
1022 #define __FUNCT__ "PCGAMGClassicalInitializePackage"
1023 PetscErrorCode PCGAMGClassicalInitializePackage(void)
1024 {
1025   PetscErrorCode ierr;
1026 
1027   PetscFunctionBegin;
1028   if (PCGAMGClassicalPackageInitialized) PetscFunctionReturn(0);
1029   ierr = PetscFunctionListAdd(&PCGAMGClassicalProlongatorList,PCGAMGCLASSICALDIRECT,PCGAMGProlongator_Classical_Direct);CHKERRQ(ierr);
1030   ierr = PetscFunctionListAdd(&PCGAMGClassicalProlongatorList,PCGAMGCLASSICALSTANDARD,PCGAMGProlongator_Classical_Standard);CHKERRQ(ierr);
1031   ierr = PetscRegisterFinalize(PCGAMGClassicalFinalizePackage);CHKERRQ(ierr);
1032   PetscFunctionReturn(0);
1033 }
1034 
1035 /* -------------------------------------------------------------------------- */
1036 /*
1037    PCCreateGAMG_Classical
1038 
1039 */
1040 #undef __FUNCT__
1041 #define __FUNCT__ "PCCreateGAMG_Classical"
1042 PetscErrorCode  PCCreateGAMG_Classical(PC pc)
1043 {
1044   PetscErrorCode ierr;
1045   PC_MG             *mg      = (PC_MG*)pc->data;
1046   PC_GAMG           *pc_gamg = (PC_GAMG*)mg->innerctx;
1047   PC_GAMG_Classical *pc_gamg_classical;
1048 
1049   PetscFunctionBegin;
1050   ierr = PCGAMGClassicalInitializePackage();
1051   if (pc_gamg->subctx) {
1052     /* call base class */
1053     ierr = PCDestroy_GAMG(pc);CHKERRQ(ierr);
1054   }
1055 
1056   /* create sub context for SA */
1057   ierr = PetscNewLog(pc,&pc_gamg_classical);CHKERRQ(ierr);
1058   pc_gamg->subctx = pc_gamg_classical;
1059   pc->ops->setfromoptions = PCGAMGSetFromOptions_Classical;
1060   /* reset does not do anything; setup not virtual */
1061 
1062   /* set internal function pointers */
1063   pc_gamg->ops->destroy        = PCGAMGDestroy_Classical;
1064   pc_gamg->ops->graph          = PCGAMGGraph_Classical;
1065   pc_gamg->ops->coarsen        = PCGAMGCoarsen_Classical;
1066   pc_gamg->ops->prolongator    = PCGAMGProlongator_Classical;
1067   pc_gamg->ops->optprol        = PCGAMGOptProl_Classical_Jacobi;
1068   pc_gamg->ops->setfromoptions = PCGAMGSetFromOptions_Classical;
1069 
1070   pc_gamg->ops->createdefaultdata = PCGAMGSetData_Classical;
1071   pc_gamg_classical->interp_threshold = 0.2;
1072   pc_gamg_classical->nsmooths         = 0;
1073   ierr = PetscObjectComposeFunction((PetscObject)pc,"PCGAMGClassicalSetType_C",PCGAMGClassicalSetType_GAMG);CHKERRQ(ierr);
1074   ierr = PetscObjectComposeFunction((PetscObject)pc,"PCGAMGClassicalGetType_C",PCGAMGClassicalGetType_GAMG);CHKERRQ(ierr);
1075   ierr = PCGAMGClassicalSetType(pc,PCGAMGCLASSICALSTANDARD);CHKERRQ(ierr);
1076   PetscFunctionReturn(0);
1077 }
1078