Actual source code: ml.c
1: #define PETSCKSP_DLL
3: /*
4: Provides an interface to the ML 4.0 smoothed Aggregation
5: */
6: #include private/pcimpl.h
7: #include src/ksp/pc/impls/mg/mgimpl.h
8: #include src/mat/impls/aij/seq/aij.h
9: #include src/mat/impls/aij/mpi/mpiaij.h
12: #include <math.h>
13: #include "ml_include.h"
16: /* The context (data structure) at each grid level */
17: typedef struct {
18: Vec x,b,r; /* global vectors */
19: Mat A,P,R;
20: KSP ksp;
21: } GridCtx;
23: /* The context used to input PETSc matrix into ML at fine grid */
24: typedef struct {
25: Mat A,Aloc;
26: Vec x,y;
27: ML_Operator *mlmat;
28: PetscScalar *pwork; /* tmp array used by PetscML_comm() */
29: } FineGridCtx;
31: /* The context associates a ML matrix with a PETSc shell matrix */
32: typedef struct {
33: Mat A; /* PETSc shell matrix associated with mlmat */
34: ML_Operator *mlmat; /* ML matrix assorciated with A */
35: Vec y;
36: } Mat_MLShell;
38: /* Private context for the ML preconditioner */
39: typedef struct {
40: ML *ml_object;
41: ML_Aggregate *agg_object;
42: GridCtx *gridctx;
43: FineGridCtx *PetscMLdata;
44: PetscInt fine_level,MaxNlevels,MaxCoarseSize,CoarsenScheme;
45: PetscReal Threshold,DampingFactor;
46: PetscTruth SpectralNormScheme_Anorm;
47: PetscMPIInt size;
48: PetscErrorCode (*PCSetUp)(PC);
49: PetscErrorCode (*PCDestroy)(PC);
50: } PC_ML;
53: int allocated_space,int columns[],double values[],int row_lengths[]);
64: /* -------------------------------------------------------------------------- */
65: /*
66: PCSetUp_ML - Prepares for the use of the ML preconditioner
67: by setting data structures and options.
69: Input Parameter:
70: . pc - the preconditioner context
72: Application Interface Routine: PCSetUp()
74: Notes:
75: The interface routine PCSetUp() is not usually called directly by
76: the user, but instead is called by PCApply() if necessary.
77: */
81: PetscErrorCode PCSetUp_ML(PC pc)
82: {
83: PetscErrorCode ierr;
84: PetscMPIInt size;
85: FineGridCtx *PetscMLdata;
86: ML *ml_object;
87: ML_Aggregate *agg_object;
88: ML_Operator *mlmat;
89: PetscInt nlocal_allcols,Nlevels,mllevel,level,level1,m,fine_level;
90: Mat A,Aloc;
91: GridCtx *gridctx;
92: PC_ML *pc_ml=PETSC_NULL;
93: PetscContainer container;
96: PetscObjectQuery((PetscObject)pc,"PC_ML",(PetscObject *)&container);
97: if (container) {
98: PetscContainerGetPointer(container,(void **)&pc_ml);
99: } else {
100: SETERRQ(PETSC_ERR_ARG_NULL,"Container does not exit");
101: }
102:
103: /* setup special features of PCML */
104: /*--------------------------------*/
105: /* covert A to Aloc to be used by ML at fine grid */
106: A = pc->pmat;
107: MPI_Comm_size(A->comm,&size);
108: pc_ml->size = size;
109: if (size > 1){
110: MatConvert_MPIAIJ_ML(A,PETSC_NULL,MAT_INITIAL_MATRIX,&Aloc);
111: } else {
112: Aloc = A;
113: }
115: /* create and initialize struct 'PetscMLdata' */
116: PetscNew(FineGridCtx,&PetscMLdata);
117: PetscMLdata->A = A;
118: PetscMLdata->Aloc = Aloc;
119: PetscMalloc((Aloc->cmap.n+1)*sizeof(PetscScalar),&PetscMLdata->pwork);
120: pc_ml->PetscMLdata = PetscMLdata;
122: VecCreate(PETSC_COMM_SELF,&PetscMLdata->x);
123: if (size == 1){
124: VecSetSizes(PetscMLdata->x,A->cmap.n,A->cmap.n);
125: } else {
126: VecSetSizes(PetscMLdata->x,Aloc->cmap.n,Aloc->cmap.n);
127: }
128: VecSetType(PetscMLdata->x,VECSEQ);
130: VecCreate(PETSC_COMM_SELF,&PetscMLdata->y);
131: VecSetSizes(PetscMLdata->y,A->rmap.n,PETSC_DECIDE);
132: VecSetType(PetscMLdata->y,VECSEQ);
133:
134: /* create ML discretization matrix at fine grid */
135: MatGetSize(Aloc,&m,&nlocal_allcols);
136: ML_Create(&ml_object,pc_ml->MaxNlevels);
137: ML_Init_Amatrix(ml_object,0,m,m,PetscMLdata);
138: ML_Set_Amatrix_Getrow(ml_object,0,PetscML_getrow,PetscML_comm,nlocal_allcols);
139: ML_Set_Amatrix_Matvec(ml_object,0,PetscML_matvec);
141: /* aggregation */
142: ML_Aggregate_Create(&agg_object);
143: ML_Aggregate_Set_MaxCoarseSize(agg_object,pc_ml->MaxCoarseSize);
144: /* set options */
145: switch (pc_ml->CoarsenScheme) {
146: case 1:
147: ML_Aggregate_Set_CoarsenScheme_Coupled(agg_object);break;
148: case 2:
149: ML_Aggregate_Set_CoarsenScheme_MIS(agg_object);break;
150: case 3:
151: ML_Aggregate_Set_CoarsenScheme_METIS(agg_object);break;
152: }
153: ML_Aggregate_Set_Threshold(agg_object,pc_ml->Threshold);
154: ML_Aggregate_Set_DampingFactor(agg_object,pc_ml->DampingFactor);
155: if (pc_ml->SpectralNormScheme_Anorm){
156: ML_Aggregate_Set_SpectralNormScheme_Anorm(agg_object);
157: }
158:
159: Nlevels = ML_Gen_MGHierarchy_UsingAggregation(ml_object,0,ML_INCREASING,agg_object);
160: if (Nlevels<=0) SETERRQ1(PETSC_ERR_ARG_OUTOFRANGE,"Nlevels %d must > 0",Nlevels);
161: PCMGSetLevels(pc,Nlevels,PETSC_NULL);
162: PCSetFromOptions_MG(pc); /* should be called in PCSetFromOptions_ML(), but cannot be called prior to PCMGSetLevels() */
163: pc_ml->ml_object = ml_object;
164: pc_ml->agg_object = agg_object;
166: PetscMalloc(Nlevels*sizeof(GridCtx),&gridctx);
167: fine_level = Nlevels - 1;
168: pc_ml->gridctx = gridctx;
169: pc_ml->fine_level = fine_level;
171: /* wrap ML matrices by PETSc shell matrices at coarsened grids.
172: Level 0 is the finest grid for ML, but coarsest for PETSc! */
173: gridctx[fine_level].A = A;
174: level = fine_level - 1;
175: if (size == 1){ /* convert ML P, R and A into seqaij format */
176: for (mllevel=1; mllevel<Nlevels; mllevel++){
177: mlmat = &(ml_object->Pmat[mllevel]);
178: MatWrapML_SeqAIJ(mlmat,&gridctx[level].P);
179: mlmat = &(ml_object->Amat[mllevel]);
180: MatWrapML_SeqAIJ(mlmat,&gridctx[level].A);
181: mlmat = &(ml_object->Rmat[mllevel-1]);
182: MatWrapML_SeqAIJ(mlmat,&gridctx[level].R);
183: level--;
184: }
185: } else { /* convert ML P and R into shell format, ML A into mpiaij format */
186: for (mllevel=1; mllevel<Nlevels; mllevel++){
187: mlmat = &(ml_object->Pmat[mllevel]);
188: MatWrapML_SHELL(mlmat,&gridctx[level].P);
189: mlmat = &(ml_object->Rmat[mllevel-1]);
190: MatWrapML_SHELL(mlmat,&gridctx[level].R);
191: mlmat = &(ml_object->Amat[mllevel]);
192: MatWrapML_MPIAIJ(mlmat,&gridctx[level].A);
193: level--;
194: }
195: }
197: /* create coarse level and the interpolation between the levels */
198: for (level=0; level<fine_level; level++){
199: VecCreate(gridctx[level].A->comm,&gridctx[level].x);
200: VecSetSizes(gridctx[level].x,gridctx[level].A->cmap.n,PETSC_DECIDE);
201: VecSetType(gridctx[level].x,VECMPI);
202: PCMGSetX(pc,level,gridctx[level].x);
203:
204: VecCreate(gridctx[level].A->comm,&gridctx[level].b);
205: VecSetSizes(gridctx[level].b,gridctx[level].A->rmap.n,PETSC_DECIDE);
206: VecSetType(gridctx[level].b,VECMPI);
207: PCMGSetRhs(pc,level,gridctx[level].b);
208:
209: level1 = level + 1;
210: VecCreate(gridctx[level1].A->comm,&gridctx[level1].r);
211: VecSetSizes(gridctx[level1].r,gridctx[level1].A->rmap.n,PETSC_DECIDE);
212: VecSetType(gridctx[level1].r,VECMPI);
213: PCMGSetR(pc,level1,gridctx[level1].r);
215: PCMGSetInterpolate(pc,level1,gridctx[level].P);
216: PCMGSetRestriction(pc,level1,gridctx[level].R);
218: if (level == 0){
219: PCMGGetCoarseSolve(pc,&gridctx[level].ksp);
220: } else {
221: PCMGGetSmoother(pc,level,&gridctx[level].ksp);
222: PCMGSetResidual(pc,level,PCMGDefaultResidual,gridctx[level].A);
223: }
224: KSPSetOperators(gridctx[level].ksp,gridctx[level].A,gridctx[level].A,DIFFERENT_NONZERO_PATTERN);
225: }
226: PCMGGetSmoother(pc,fine_level,&gridctx[fine_level].ksp);
227: PCMGSetResidual(pc,fine_level,PCMGDefaultResidual,gridctx[fine_level].A);
228: KSPSetOperators(gridctx[fine_level].ksp,gridctx[level].A,gridctx[fine_level].A,DIFFERENT_NONZERO_PATTERN);
229:
230: /* now call PCSetUp_MG() */
231: /*--------------------------------*/
232: (*pc_ml->PCSetUp)(pc);
233: return(0);
234: }
238: PetscErrorCode PetscContainerDestroy_PC_ML(void *ptr)
239: {
240: PetscErrorCode ierr;
241: PC_ML *pc_ml = (PC_ML*)ptr;
242: PetscInt level;
245: if (pc_ml->size > 1){MatDestroy(pc_ml->PetscMLdata->Aloc);}
246: ML_Aggregate_Destroy(&pc_ml->agg_object);
247: ML_Destroy(&pc_ml->ml_object);
249: PetscFree(pc_ml->PetscMLdata->pwork);
250: if (pc_ml->PetscMLdata->x){VecDestroy(pc_ml->PetscMLdata->x);}
251: if (pc_ml->PetscMLdata->y){VecDestroy(pc_ml->PetscMLdata->y);}
252: PetscFree(pc_ml->PetscMLdata);
254: for (level=0; level<pc_ml->fine_level; level++){
255: if (pc_ml->gridctx[level].A){MatDestroy(pc_ml->gridctx[level].A);}
256: if (pc_ml->gridctx[level].P){MatDestroy(pc_ml->gridctx[level].P);}
257: if (pc_ml->gridctx[level].R){MatDestroy(pc_ml->gridctx[level].R);}
258: if (pc_ml->gridctx[level].x){VecDestroy(pc_ml->gridctx[level].x);}
259: if (pc_ml->gridctx[level].b){VecDestroy(pc_ml->gridctx[level].b);}
260: if (pc_ml->gridctx[level+1].r){VecDestroy(pc_ml->gridctx[level+1].r);}
261: }
262: PetscFree(pc_ml->gridctx);
263: PetscFree(pc_ml);
264: return(0);
265: }
266: /* -------------------------------------------------------------------------- */
267: /*
268: PCDestroy_ML - Destroys the private context for the ML preconditioner
269: that was created with PCCreate_ML().
271: Input Parameter:
272: . pc - the preconditioner context
274: Application Interface Routine: PCDestroy()
275: */
278: PetscErrorCode PCDestroy_ML(PC pc)
279: {
280: PetscErrorCode ierr;
281: PC_ML *pc_ml=PETSC_NULL;
282: PetscContainer container;
285: PetscObjectQuery((PetscObject)pc,"PC_ML",(PetscObject *)&container);
286: if (container) {
287: PetscContainerGetPointer(container,(void **)&pc_ml);
288: pc->ops->destroy = pc_ml->PCDestroy;
289: } else {
290: SETERRQ(PETSC_ERR_ARG_NULL,"Container does not exit");
291: }
292: /* detach pc and PC_ML and dereference container */
293: PetscObjectCompose((PetscObject)pc,"PC_ML",0);
294: (*pc->ops->destroy)(pc);
296: PetscContainerDestroy(container);
297: return(0);
298: }
302: PetscErrorCode PCSetFromOptions_ML(PC pc)
303: {
304: PetscErrorCode ierr;
305: PetscInt indx,m,PrintLevel,MaxNlevels,MaxCoarseSize;
306: PetscReal Threshold,DampingFactor;
307: PetscTruth flg;
308: const char *scheme[] = {"Uncoupled","Coupled","MIS","METIS"};
309: PC_ML *pc_ml=PETSC_NULL;
310: PetscContainer container;
311: PCMGType mgtype;
314: PetscObjectQuery((PetscObject)pc,"PC_ML",(PetscObject *)&container);
315: if (container) {
316: PetscContainerGetPointer(container,(void **)&pc_ml);
317: } else {
318: SETERRQ(PETSC_ERR_ARG_NULL,"Container does not exit");
319: }
321: /* inherited MG options */
322: PetscOptionsHead("Multigrid options(inherited)");
323: PetscOptionsInt("-pc_mg_cycles","1 for V cycle, 2 for W-cycle","MGSetCycles",1,&m,&flg);
324: PetscOptionsInt("-pc_mg_smoothup","Number of post-smoothing steps","MGSetNumberSmoothUp",1,&m,&flg);
325: PetscOptionsInt("-pc_mg_smoothdown","Number of pre-smoothing steps","MGSetNumberSmoothDown",1,&m,&flg);
326: PetscOptionsEnum("-pc_mg_type","Multigrid type","PCMGSetType",PCMGTypes,(PetscEnum)PC_MG_MULTIPLICATIVE,(PetscEnum*)&mgtype,&flg);
327: PetscOptionsTail();
329: /* ML options */
330: PetscOptionsHead("ML options");
331: /* set defaults */
332: PrintLevel = 0;
333: MaxNlevels = 10;
334: MaxCoarseSize = 1;
335: indx = 0;
336: Threshold = 0.0;
337: DampingFactor = 4.0/3.0;
338:
339: PetscOptionsInt("-pc_ml_PrintLevel","Print level","ML_Set_PrintLevel",PrintLevel,&PrintLevel,PETSC_NULL);
340: ML_Set_PrintLevel(PrintLevel);
342: PetscOptionsInt("-pc_ml_maxNlevels","Maximum number of levels","None",MaxNlevels,&MaxNlevels,PETSC_NULL);
343: pc_ml->MaxNlevels = MaxNlevels;
345: PetscOptionsInt("-pc_ml_maxCoarseSize","Maximum coarsest mesh size","ML_Aggregate_Set_MaxCoarseSize",MaxCoarseSize,&MaxCoarseSize,PETSC_NULL);
346: pc_ml->MaxCoarseSize = MaxCoarseSize;
348: PetscOptionsEList("-pc_ml_CoarsenScheme","Aggregate Coarsen Scheme","ML_Aggregate_Set_CoarsenScheme_*",scheme,4,scheme[0],&indx,PETSC_NULL);
349: pc_ml->CoarsenScheme = indx;
351: PetscOptionsReal("-pc_ml_DampingFactor","P damping factor","ML_Aggregate_Set_DampingFactor",DampingFactor,&DampingFactor,PETSC_NULL);
352: pc_ml->DampingFactor = DampingFactor;
353:
354: PetscOptionsReal("-pc_ml_Threshold","Smoother drop tol","ML_Aggregate_Set_Threshold",Threshold,&Threshold,PETSC_NULL);
355: pc_ml->Threshold = Threshold;
357: PetscOptionsTruth("-pc_ml_SpectralNormScheme_Anorm","Method used for estimating spectral radius","ML_Aggregate_Set_SpectralNormScheme_Anorm",PETSC_FALSE,&pc_ml->SpectralNormScheme_Anorm,PETSC_NULL);
358:
359: PetscOptionsTail();
360: return(0);
361: }
363: /* -------------------------------------------------------------------------- */
364: /*
365: PCCreate_ML - Creates a ML preconditioner context, PC_ML,
366: and sets this as the private data within the generic preconditioning
367: context, PC, that was created within PCCreate().
369: Input Parameter:
370: . pc - the preconditioner context
372: Application Interface Routine: PCCreate()
373: */
375: /*MC
376: PCML - Use algebraic multigrid preconditioning. This preconditioner requires you provide
377: fine grid discretization matrix. The coarser grid matrices and restriction/interpolation
378: operators are computed by ML, with the matrices coverted to PETSc matrices in aij format
379: and the restriction/interpolation operators wrapped as PETSc shell matrices.
381: Options Database Key:
382: Multigrid options(inherited)
383: + -pc_mg_cycles <1>: 1 for V cycle, 2 for W-cycle (MGSetCycles)
384: . -pc_mg_smoothup <1>: Number of post-smoothing steps (MGSetNumberSmoothUp)
385: . -pc_mg_smoothdown <1>: Number of pre-smoothing steps (MGSetNumberSmoothDown)
386: - -pc_mg_type <multiplicative> (one of) additive multiplicative full cascade kascade
387:
388: ML options:
389: + -pc_ml_PrintLevel <0>: Print level (ML_Set_PrintLevel)
390: . -pc_ml_maxNlevels <10>: Maximum number of levels (None)
391: . -pc_ml_maxCoarseSize <1>: Maximum coarsest mesh size (ML_Aggregate_Set_MaxCoarseSize)
392: . -pc_ml_CoarsenScheme <Uncoupled> (one of) Uncoupled Coupled MIS METIS
393: . -pc_ml_DampingFactor <1.33333>: P damping factor (ML_Aggregate_Set_DampingFactor)
394: . -pc_ml_Threshold <0>: Smoother drop tol (ML_Aggregate_Set_Threshold)
395: - -pc_ml_SpectralNormScheme_Anorm: <false> Method used for estimating spectral radius (ML_Aggregate_Set_SpectralNormScheme_Anorm)
397: Level: intermediate
399: Concepts: multigrid
400:
401: .seealso: PCCreate(), PCSetType(), PCType (for list of available types), PC, PCMGType,
402: PCMGSetLevels(), PCMGGetLevels(), PCMGSetType(), MPSetCycles(), PCMGSetNumberSmoothDown(),
403: PCMGSetNumberSmoothUp(), PCMGGetCoarseSolve(), PCMGSetResidual(), PCMGSetInterpolation(),
404: PCMGSetRestriction(), PCMGGetSmoother(), PCMGGetSmootherUp(), PCMGGetSmootherDown(),
405: PCMGSetCyclesOnLevel(), PCMGSetRhs(), PCMGSetX(), PCMGSetR()
406: M*/
411: PetscErrorCode PCCreate_ML(PC pc)
412: {
413: PetscErrorCode ierr;
414: PC_ML *pc_ml;
415: PetscContainer container;
418: /* initialize pc as PCMG */
419: PCSetType(pc,PCMG); /* calls PCCreate_MG() and MGCreate_Private() */
421: /* create a supporting struct and attach it to pc */
422: PetscNew(PC_ML,&pc_ml);
423: PetscContainerCreate(PETSC_COMM_SELF,&container);
424: PetscContainerSetPointer(container,pc_ml);
425: PetscContainerSetUserDestroy(container,PetscContainerDestroy_PC_ML);
426: PetscObjectCompose((PetscObject)pc,"PC_ML",(PetscObject)container);
427:
428: pc_ml->PCSetUp = pc->ops->setup;
429: pc_ml->PCDestroy = pc->ops->destroy;
431: /* overwrite the pointers of PCMG by the functions of PCML */
432: pc->ops->setfromoptions = PCSetFromOptions_ML;
433: pc->ops->setup = PCSetUp_ML;
434: pc->ops->destroy = PCDestroy_ML;
435: return(0);
436: }
439: int PetscML_getrow(ML_Operator *ML_data, int N_requested_rows, int requested_rows[],
440: int allocated_space, int columns[], double values[], int row_lengths[])
441: {
443: Mat Aloc;
444: Mat_SeqAIJ *a;
445: PetscInt m,i,j,k=0,row,*aj;
446: PetscScalar *aa;
447: FineGridCtx *ml=(FineGridCtx*)ML_Get_MyGetrowData(ML_data);
449: Aloc = ml->Aloc;
450: a = (Mat_SeqAIJ*)Aloc->data;
451: MatGetSize(Aloc,&m,PETSC_NULL);
453: for (i = 0; i<N_requested_rows; i++) {
454: row = requested_rows[i];
455: row_lengths[i] = a->ilen[row];
456: if (allocated_space < k+row_lengths[i]) return(0);
457: if ( (row >= 0) || (row <= (m-1)) ) {
458: aj = a->j + a->i[row];
459: aa = a->a + a->i[row];
460: for (j=0; j<row_lengths[i]; j++){
461: columns[k] = aj[j];
462: values[k++] = aa[j];
463: }
464: }
465: }
466: return(1);
467: }
469: int PetscML_matvec(ML_Operator *ML_data,int in_length,double p[],int out_length,double ap[])
470: {
472: FineGridCtx *ml=(FineGridCtx*)ML_Get_MyMatvecData(ML_data);
473: Mat A=ml->A, Aloc=ml->Aloc;
474: PetscMPIInt size;
475: PetscScalar *pwork=ml->pwork;
476: PetscInt i;
478: MPI_Comm_size(A->comm,&size);
479: if (size == 1){
480: VecPlaceArray(ml->x,p);
481: } else {
482: for (i=0; i<in_length; i++) pwork[i] = p[i];
483: PetscML_comm(pwork,ml);
484: VecPlaceArray(ml->x,pwork);
485: }
486: VecPlaceArray(ml->y,ap);
487: MatMult(Aloc,ml->x,ml->y);
488: VecResetArray(ml->x);
489: VecResetArray(ml->y);
490: return 0;
491: }
493: int PetscML_comm(double p[],void *ML_data)
494: {
496: FineGridCtx *ml=(FineGridCtx*)ML_data;
497: Mat A=ml->A;
498: Mat_MPIAIJ *a = (Mat_MPIAIJ*)A->data;
499: PetscMPIInt size;
500: PetscInt i,in_length=A->rmap.n,out_length=ml->Aloc->cmap.n;
501: PetscScalar *array;
503: MPI_Comm_size(A->comm,&size);
504: if (size == 1) return 0;
505:
506: VecPlaceArray(ml->y,p);
507: VecScatterBegin(ml->y,a->lvec,INSERT_VALUES,SCATTER_FORWARD,a->Mvctx);
508: VecScatterEnd(ml->y,a->lvec,INSERT_VALUES,SCATTER_FORWARD,a->Mvctx);
509: VecResetArray(ml->y);
510: VecGetArray(a->lvec,&array);
511: for (i=in_length; i<out_length; i++){
512: p[i] = array[i-in_length];
513: }
514: VecRestoreArray(a->lvec,&array);
515: return 0;
516: }
519: PetscErrorCode MatMult_ML(Mat A,Vec x,Vec y)
520: {
521: PetscErrorCode ierr;
522: Mat_MLShell *shell;
523: PetscScalar *xarray,*yarray;
524: PetscInt x_length,y_length;
525:
527: MatShellGetContext(A,(void **)&shell);
528: VecGetArray(x,&xarray);
529: VecGetArray(y,&yarray);
530: x_length = shell->mlmat->invec_leng;
531: y_length = shell->mlmat->outvec_leng;
533: ML_Operator_Apply(shell->mlmat,x_length,xarray,y_length,yarray);
535: VecRestoreArray(x,&xarray);
536: VecRestoreArray(y,&yarray);
537: return(0);
538: }
539: /* MatMultAdd_ML - Compute y = w + A*x */
542: PetscErrorCode MatMultAdd_ML(Mat A,Vec x,Vec w,Vec y)
543: {
544: PetscErrorCode ierr;
545: Mat_MLShell *shell;
546: PetscScalar *xarray,*yarray;
547: PetscInt x_length,y_length;
548:
550: MatShellGetContext(A,(void **)&shell);
551: VecGetArray(x,&xarray);
552: VecGetArray(y,&yarray);
554: x_length = shell->mlmat->invec_leng;
555: y_length = shell->mlmat->outvec_leng;
557: ML_Operator_Apply(shell->mlmat,x_length,xarray,y_length,yarray);
559: VecRestoreArray(x,&xarray);
560: VecRestoreArray(y,&yarray);
561: VecAXPY(y,1.0,w);
563: return(0);
564: }
566: /* newtype is ignored because "ml" is not listed under Petsc MatType yet */
569: PetscErrorCode MatConvert_MPIAIJ_ML(Mat A,MatType newtype,MatReuse scall,Mat *Aloc)
570: {
571: PetscErrorCode ierr;
572: Mat_MPIAIJ *mpimat=(Mat_MPIAIJ*)A->data;
573: Mat_SeqAIJ *mat,*a=(Mat_SeqAIJ*)(mpimat->A)->data,*b=(Mat_SeqAIJ*)(mpimat->B)->data;
574: PetscInt *ai=a->i,*aj=a->j,*bi=b->i,*bj=b->j;
575: PetscScalar *aa=a->a,*ba=b->a,*ca;
576: PetscInt am=A->rmap.n,an=A->cmap.n,i,j,k;
577: PetscInt *ci,*cj,ncols;
580: if (am != an) SETERRQ2(PETSC_ERR_ARG_WRONG,"A must have a square diagonal portion, am: %d != an: %d",am,an);
582: if (scall == MAT_INITIAL_MATRIX){
583: PetscMalloc((1+am)*sizeof(PetscInt),&ci);
584: ci[0] = 0;
585: for (i=0; i<am; i++){
586: ci[i+1] = ci[i] + (ai[i+1] - ai[i]) + (bi[i+1] - bi[i]);
587: }
588: PetscMalloc((1+ci[am])*sizeof(PetscInt),&cj);
589: PetscMalloc((1+ci[am])*sizeof(PetscScalar),&ca);
591: k = 0;
592: for (i=0; i<am; i++){
593: /* diagonal portion of A */
594: ncols = ai[i+1] - ai[i];
595: for (j=0; j<ncols; j++) {
596: cj[k] = *aj++;
597: ca[k++] = *aa++;
598: }
599: /* off-diagonal portion of A */
600: ncols = bi[i+1] - bi[i];
601: for (j=0; j<ncols; j++) {
602: cj[k] = an + (*bj); bj++;
603: ca[k++] = *ba++;
604: }
605: }
606: if (k != ci[am]) SETERRQ2(PETSC_ERR_ARG_WRONG,"k: %d != ci[am]: %d",k,ci[am]);
608: /* put together the new matrix */
609: an = mpimat->A->cmap.n+mpimat->B->cmap.n;
610: MatCreateSeqAIJWithArrays(PETSC_COMM_SELF,am,an,ci,cj,ca,Aloc);
612: /* MatCreateSeqAIJWithArrays flags matrix so PETSc doesn't free the user's arrays. */
613: /* Since these are PETSc arrays, change flags to free them as necessary. */
614: mat = (Mat_SeqAIJ*)(*Aloc)->data;
615: mat->free_a = PETSC_TRUE;
616: mat->free_ij = PETSC_TRUE;
618: mat->nonew = 0;
619: } else if (scall == MAT_REUSE_MATRIX){
620: mat=(Mat_SeqAIJ*)(*Aloc)->data;
621: ci = mat->i; cj = mat->j; ca = mat->a;
622: for (i=0; i<am; i++) {
623: /* diagonal portion of A */
624: ncols = ai[i+1] - ai[i];
625: for (j=0; j<ncols; j++) *ca++ = *aa++;
626: /* off-diagonal portion of A */
627: ncols = bi[i+1] - bi[i];
628: for (j=0; j<ncols; j++) *ca++ = *ba++;
629: }
630: } else {
631: SETERRQ1(PETSC_ERR_ARG_WRONG,"Invalid MatReuse %d",(int)scall);
632: }
633: return(0);
634: }
638: PetscErrorCode MatDestroy_ML(Mat A)
639: {
641: Mat_MLShell *shell;
644: MatShellGetContext(A,(void **)&shell);
645: VecDestroy(shell->y);
646: PetscFree(shell);
647: MatDestroy_Shell(A);
648: PetscObjectChangeTypeName((PetscObject)A,0);
649: return(0);
650: }
654: PetscErrorCode MatWrapML_SeqAIJ(ML_Operator *mlmat,Mat *newmat)
655: {
656: struct ML_CSR_MSRdata *matdata = (struct ML_CSR_MSRdata *)mlmat->data;
657: PetscErrorCode ierr;
658: PetscInt m=mlmat->outvec_leng,n=mlmat->invec_leng,*nnz,nz_max;
659: PetscInt *ml_cols=matdata->columns,*aj,i,j,k;
660: PetscScalar *ml_vals=matdata->values,*aa;
661:
663: if ( mlmat->getrow == NULL) SETERRQ(PETSC_ERR_ARG_NULL,"mlmat->getrow = NULL");
664: if (m != n){ /* ML Pmat and Rmat are in CSR format. Pass array pointers into SeqAIJ matrix */
665: MatCreateSeqAIJWithArrays(PETSC_COMM_SELF,m,n,matdata->rowptr,ml_cols,ml_vals,newmat);
666: return(0);
667: }
669: /* ML Amat is in MSR format. Copy its data into SeqAIJ matrix */
670: MatCreate(PETSC_COMM_SELF,newmat);
671: MatSetSizes(*newmat,m,n,PETSC_DECIDE,PETSC_DECIDE);
672: MatSetType(*newmat,MATSEQAIJ);
673: PetscMalloc((m+1)*sizeof(PetscInt),&nnz);
675: nz_max = 0;
676: for (i=0; i<m; i++) {
677: nnz[i] = ml_cols[i+1] - ml_cols[i] + 1;
678: if (nnz[i] > nz_max) nz_max = nnz[i];
679: }
680: MatSeqAIJSetPreallocation(*newmat,0,nnz);
681: MatSetOption(*newmat,MAT_COLUMNS_SORTED); /* check! */
683: nz_max++;
684: PetscMalloc(nz_max*(sizeof(PetscInt)+sizeof(PetscScalar)),&aj);
685: aa = (PetscScalar*)(aj + nz_max);
687: for (i=0; i<m; i++){
688: k = 0;
689: /* diagonal entry */
690: aj[k] = i; aa[k++] = ml_vals[i];
691: /* off diagonal entries */
692: for (j=ml_cols[i]; j<ml_cols[i+1]; j++){
693: aj[k] = ml_cols[j]; aa[k++] = ml_vals[j];
694: }
695: /* sort aj and aa */
696: PetscSortIntWithScalarArray(nnz[i],aj,aa);
697: MatSetValues(*newmat,1,&i,nnz[i],aj,aa,INSERT_VALUES);
698: }
699: MatAssemblyBegin(*newmat,MAT_FINAL_ASSEMBLY);
700: MatAssemblyEnd(*newmat,MAT_FINAL_ASSEMBLY);
701: PetscFree(aj);
702: PetscFree(nnz);
703: return(0);
704: }
708: PetscErrorCode MatWrapML_SHELL(ML_Operator *mlmat,Mat *newmat)
709: {
711: PetscInt m,n;
712: ML_Comm *MLcomm;
713: Mat_MLShell *shellctx;
716: m = mlmat->outvec_leng;
717: n = mlmat->invec_leng;
718: if (!m || !n){
719: newmat = PETSC_NULL;
720: } else {
721: MLcomm = mlmat->comm;
722: PetscNew(Mat_MLShell,&shellctx);
723: MatCreateShell(MLcomm->USR_comm,m,n,PETSC_DETERMINE,PETSC_DETERMINE,shellctx,newmat);
724: MatShellSetOperation(*newmat,MATOP_MULT,(void(*)(void))MatMult_ML);
725: MatShellSetOperation(*newmat,MATOP_MULT_ADD,(void(*)(void))MatMultAdd_ML);
726: shellctx->A = *newmat;
727: shellctx->mlmat = mlmat;
728: VecCreate(PETSC_COMM_WORLD,&shellctx->y);
729: VecSetSizes(shellctx->y,m,PETSC_DECIDE);
730: VecSetFromOptions(shellctx->y);
731: (*newmat)->ops->destroy = MatDestroy_ML;
732: }
733: return(0);
734: }
738: PetscErrorCode MatWrapML_MPIAIJ(ML_Operator *mlmat,Mat *newmat)
739: {
740: struct ML_CSR_MSRdata *matdata = (struct ML_CSR_MSRdata *)mlmat->data;
741: PetscInt *ml_cols=matdata->columns,*aj;
742: PetscScalar *ml_vals=matdata->values,*aa;
743: PetscErrorCode ierr;
744: PetscInt i,j,k,*gordering;
745: PetscInt m=mlmat->outvec_leng,n,*nnzA,*nnzB,*nnz,nz_max,row;
746: Mat A;
749: if (mlmat->getrow == NULL) SETERRQ(PETSC_ERR_ARG_NULL,"mlmat->getrow = NULL");
750: n = mlmat->invec_leng;
751: if (m != n) SETERRQ2(PETSC_ERR_ARG_OUTOFRANGE,"m %d must equal to n %d",m,n);
753: MatCreate(mlmat->comm->USR_comm,&A);
754: MatSetSizes(A,m,n,PETSC_DECIDE,PETSC_DECIDE);
755: MatSetType(A,MATMPIAIJ);
756: PetscMalloc3(m,PetscInt,&nnzA,m,PetscInt,&nnzB,m,PetscInt,&nnz);
757:
758: nz_max = 0;
759: for (i=0; i<m; i++){
760: nnz[i] = ml_cols[i+1] - ml_cols[i] + 1;
761: if (nz_max < nnz[i]) nz_max = nnz[i];
762: nnzA[i] = 1; /* diag */
763: for (j=ml_cols[i]; j<ml_cols[i+1]; j++){
764: if (ml_cols[j] < m) nnzA[i]++;
765: }
766: nnzB[i] = nnz[i] - nnzA[i];
767: }
768: MatMPIAIJSetPreallocation(A,0,nnzA,0,nnzB);
770: /* insert mat values -- remap row and column indices */
771: nz_max++;
772: PetscMalloc(nz_max*(sizeof(PetscInt)+sizeof(PetscScalar)),&aj);
773: aa = (PetscScalar*)(aj + nz_max);
774: ML_build_global_numbering(mlmat,&gordering);
775: for (i=0; i<m; i++){
776: row = gordering[i];
777: k = 0;
778: /* diagonal entry */
779: aj[k] = row; aa[k++] = ml_vals[i];
780: /* off diagonal entries */
781: for (j=ml_cols[i]; j<ml_cols[i+1]; j++){
782: aj[k] = gordering[ml_cols[j]]; aa[k++] = ml_vals[j];
783: }
784: MatSetValues(A,1,&row,nnz[i],aj,aa,INSERT_VALUES);
785: }
786: MatAssemblyBegin(A,MAT_FINAL_ASSEMBLY);
787: MatAssemblyEnd(A,MAT_FINAL_ASSEMBLY);
788: *newmat = A;
790: PetscFree3(nnzA,nnzB,nnz);
791: PetscFree(aj);
792: return(0);
793: }