Actual source code: ml.c
petsc-3.5.4 2015-05-23
2: /*
3: Provides an interface to the ML smoothed Aggregation
4: Note: Something non-obvious breaks -pc_mg_type ADDITIVE for parallel runs
5: Jed Brown, see [PETSC #18321, #18449].
6: */
7: #include <petsc-private/pcimpl.h> /*I "petscpc.h" I*/
8: #include <../src/ksp/pc/impls/mg/mgimpl.h> /*I "petscksp.h" I*/
9: #include <../src/mat/impls/aij/seq/aij.h>
10: #include <../src/mat/impls/aij/mpi/mpiaij.h>
11: #include <petscdm.h> /* for DMDestroy(&pc->mg) hack */
13: EXTERN_C_BEGIN
14: /* HAVE_CONFIG_H flag is required by ML include files */
15: #if !defined(HAVE_CONFIG_H)
16: #define HAVE_CONFIG_H
17: #endif
18: #include <ml_include.h>
19: #include <ml_viz_stats.h>
20: EXTERN_C_END
22: typedef enum {PCML_NULLSPACE_AUTO,PCML_NULLSPACE_USER,PCML_NULLSPACE_BLOCK,PCML_NULLSPACE_SCALAR} PCMLNullSpaceType;
23: static const char *const PCMLNullSpaceTypes[] = {"AUTO","USER","BLOCK","SCALAR","PCMLNullSpaceType","PCML_NULLSPACE_",0};
25: /* The context (data structure) at each grid level */
26: typedef struct {
27: Vec x,b,r; /* global vectors */
28: Mat A,P,R;
29: KSP ksp;
30: Vec coords; /* projected by ML, if PCSetCoordinates is called; values packed by node */
31: } GridCtx;
33: /* The context used to input PETSc matrix into ML at fine grid */
34: typedef struct {
35: Mat A; /* Petsc matrix in aij format */
36: Mat Aloc; /* local portion of A to be used by ML */
37: Vec x,y;
38: ML_Operator *mlmat;
39: PetscScalar *pwork; /* tmp array used by PetscML_comm() */
40: } FineGridCtx;
42: /* The context associates a ML matrix with a PETSc shell matrix */
43: typedef struct {
44: Mat A; /* PETSc shell matrix associated with mlmat */
45: ML_Operator *mlmat; /* ML matrix assorciated with A */
46: Vec y, work;
47: } Mat_MLShell;
49: /* Private context for the ML preconditioner */
50: typedef struct {
51: ML *ml_object;
52: ML_Aggregate *agg_object;
53: GridCtx *gridctx;
54: FineGridCtx *PetscMLdata;
55: PetscInt Nlevels,MaxNlevels,MaxCoarseSize,CoarsenScheme,EnergyMinimization,MinPerProc,PutOnSingleProc,RepartitionType,ZoltanScheme;
56: PetscReal Threshold,DampingFactor,EnergyMinimizationDropTol,MaxMinRatio,AuxThreshold;
57: PetscBool SpectralNormScheme_Anorm,BlockScaling,EnergyMinimizationCheap,Symmetrize,OldHierarchy,KeepAggInfo,Reusable,Repartition,Aux;
58: PetscBool reuse_interpolation;
59: PCMLNullSpaceType nulltype;
60: PetscMPIInt size; /* size of communicator for pc->pmat */
61: PetscInt dim; /* data from PCSetCoordinates(_ML) */
62: PetscInt nloc;
63: PetscReal *coords; /* ML has a grid object for each level: the finest grid will point into coords */
64: } PC_ML;
68: static int PetscML_getrow(ML_Operator *ML_data, int N_requested_rows, int requested_rows[],int allocated_space, int columns[], double values[], int row_lengths[])
69: {
71: PetscInt m,i,j,k=0,row,*aj;
72: PetscScalar *aa;
73: FineGridCtx *ml=(FineGridCtx*)ML_Get_MyGetrowData(ML_data);
74: Mat_SeqAIJ *a = (Mat_SeqAIJ*)ml->Aloc->data;
76: MatGetSize(ml->Aloc,&m,NULL); if (ierr) return(0);
77: for (i = 0; i<N_requested_rows; i++) {
78: row = requested_rows[i];
79: row_lengths[i] = a->ilen[row];
80: if (allocated_space < k+row_lengths[i]) return(0);
81: if ((row >= 0) || (row <= (m-1))) {
82: aj = a->j + a->i[row];
83: aa = a->a + a->i[row];
84: for (j=0; j<row_lengths[i]; j++) {
85: columns[k] = aj[j];
86: values[k++] = aa[j];
87: }
88: }
89: }
90: return(1);
91: }
95: static PetscErrorCode PetscML_comm(double p[],void *ML_data)
96: {
98: FineGridCtx *ml = (FineGridCtx*)ML_data;
99: Mat A = ml->A;
100: Mat_MPIAIJ *a = (Mat_MPIAIJ*)A->data;
101: PetscMPIInt size;
102: PetscInt i,in_length=A->rmap->n,out_length=ml->Aloc->cmap->n;
103: PetscScalar *array;
106: MPI_Comm_size(PetscObjectComm((PetscObject)A),&size);
107: if (size == 1) return 0;
109: VecPlaceArray(ml->y,p);
110: VecScatterBegin(a->Mvctx,ml->y,a->lvec,INSERT_VALUES,SCATTER_FORWARD);
111: VecScatterEnd(a->Mvctx,ml->y,a->lvec,INSERT_VALUES,SCATTER_FORWARD);
112: VecResetArray(ml->y);
113: VecGetArray(a->lvec,&array);
114: for (i=in_length; i<out_length; i++) p[i] = array[i-in_length];
115: VecRestoreArray(a->lvec,&array);
116: return(0);
117: }
121: static int PetscML_matvec(ML_Operator *ML_data,int in_length,double p[],int out_length,double ap[])
122: {
124: FineGridCtx *ml = (FineGridCtx*)ML_Get_MyMatvecData(ML_data);
125: Mat A = ml->A, Aloc=ml->Aloc;
126: PetscMPIInt size;
127: PetscScalar *pwork=ml->pwork;
128: PetscInt i;
131: MPI_Comm_size(PetscObjectComm((PetscObject)A),&size);
132: if (size == 1) {
133: VecPlaceArray(ml->x,p);
134: } else {
135: for (i=0; i<in_length; i++) pwork[i] = p[i];
136: PetscML_comm(pwork,ml);
137: VecPlaceArray(ml->x,pwork);
138: }
139: VecPlaceArray(ml->y,ap);
140: MatMult(Aloc,ml->x,ml->y);
141: VecResetArray(ml->x);
142: VecResetArray(ml->y);
143: return(0);
144: }
148: static PetscErrorCode MatMult_ML(Mat A,Vec x,Vec y)
149: {
151: Mat_MLShell *shell;
152: PetscScalar *xarray,*yarray;
153: PetscInt x_length,y_length;
156: MatShellGetContext(A,(void**)&shell);
157: VecGetArray(x,&xarray);
158: VecGetArray(y,&yarray);
159: x_length = shell->mlmat->invec_leng;
160: y_length = shell->mlmat->outvec_leng;
161: PetscStackCall("ML_Operator_Apply",ML_Operator_Apply(shell->mlmat,x_length,xarray,y_length,yarray));
162: VecRestoreArray(x,&xarray);
163: VecRestoreArray(y,&yarray);
164: return(0);
165: }
169: /* Computes y = w + A * x
170: It is possible that w == y, but not x == y
171: */
172: static PetscErrorCode MatMultAdd_ML(Mat A,Vec x,Vec w,Vec y)
173: {
174: Mat_MLShell *shell;
175: PetscScalar *xarray,*yarray;
176: PetscInt x_length,y_length;
180: MatShellGetContext(A, (void**) &shell);
181: if (y == w) {
182: if (!shell->work) {
183: VecDuplicate(y, &shell->work);
184: }
185: VecGetArray(x, &xarray);
186: VecGetArray(shell->work, &yarray);
187: x_length = shell->mlmat->invec_leng;
188: y_length = shell->mlmat->outvec_leng;
189: PetscStackCall("ML_Operator_Apply",ML_Operator_Apply(shell->mlmat, x_length, xarray, y_length, yarray));
190: VecRestoreArray(x, &xarray);
191: VecRestoreArray(shell->work, &yarray);
192: VecAXPY(y, 1.0, shell->work);
193: } else {
194: VecGetArray(x, &xarray);
195: VecGetArray(y, &yarray);
196: x_length = shell->mlmat->invec_leng;
197: y_length = shell->mlmat->outvec_leng;
198: PetscStackCall("ML_Operator_Apply",ML_Operator_Apply(shell->mlmat, x_length, xarray, y_length, yarray));
199: VecRestoreArray(x, &xarray);
200: VecRestoreArray(y, &yarray);
201: VecAXPY(y, 1.0, w);
202: }
203: return(0);
204: }
206: /* newtype is ignored since only handles one case */
209: static PetscErrorCode MatConvert_MPIAIJ_ML(Mat A,MatType newtype,MatReuse scall,Mat *Aloc)
210: {
212: Mat_MPIAIJ *mpimat=(Mat_MPIAIJ*)A->data;
213: Mat_SeqAIJ *mat,*a=(Mat_SeqAIJ*)(mpimat->A)->data,*b=(Mat_SeqAIJ*)(mpimat->B)->data;
214: PetscInt *ai=a->i,*aj=a->j,*bi=b->i,*bj=b->j;
215: PetscScalar *aa=a->a,*ba=b->a,*ca;
216: PetscInt am =A->rmap->n,an=A->cmap->n,i,j,k;
217: PetscInt *ci,*cj,ncols;
220: if (am != an) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"A must have a square diagonal portion, am: %d != an: %d",am,an);
222: if (scall == MAT_INITIAL_MATRIX) {
223: PetscMalloc1((1+am),&ci);
224: ci[0] = 0;
225: for (i=0; i<am; i++) ci[i+1] = ci[i] + (ai[i+1] - ai[i]) + (bi[i+1] - bi[i]);
226: PetscMalloc1((1+ci[am]),&cj);
227: PetscMalloc1((1+ci[am]),&ca);
229: k = 0;
230: for (i=0; i<am; i++) {
231: /* diagonal portion of A */
232: ncols = ai[i+1] - ai[i];
233: for (j=0; j<ncols; j++) {
234: cj[k] = *aj++;
235: ca[k++] = *aa++;
236: }
237: /* off-diagonal portion of A */
238: ncols = bi[i+1] - bi[i];
239: for (j=0; j<ncols; j++) {
240: cj[k] = an + (*bj); bj++;
241: ca[k++] = *ba++;
242: }
243: }
244: if (k != ci[am]) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"k: %d != ci[am]: %d",k,ci[am]);
246: /* put together the new matrix */
247: an = mpimat->A->cmap->n+mpimat->B->cmap->n;
248: MatCreateSeqAIJWithArrays(PETSC_COMM_SELF,am,an,ci,cj,ca,Aloc);
250: /* MatCreateSeqAIJWithArrays flags matrix so PETSc doesn't free the user's arrays. */
251: /* Since these are PETSc arrays, change flags to free them as necessary. */
252: mat = (Mat_SeqAIJ*)(*Aloc)->data;
253: mat->free_a = PETSC_TRUE;
254: mat->free_ij = PETSC_TRUE;
256: mat->nonew = 0;
257: } else if (scall == MAT_REUSE_MATRIX) {
258: mat=(Mat_SeqAIJ*)(*Aloc)->data;
259: ci = mat->i; cj = mat->j; ca = mat->a;
260: for (i=0; i<am; i++) {
261: /* diagonal portion of A */
262: ncols = ai[i+1] - ai[i];
263: for (j=0; j<ncols; j++) *ca++ = *aa++;
264: /* off-diagonal portion of A */
265: ncols = bi[i+1] - bi[i];
266: for (j=0; j<ncols; j++) *ca++ = *ba++;
267: }
268: } else SETERRQ1(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONG,"Invalid MatReuse %d",(int)scall);
269: return(0);
270: }
272: extern PetscErrorCode MatDestroy_Shell(Mat);
275: static PetscErrorCode MatDestroy_ML(Mat A)
276: {
278: Mat_MLShell *shell;
281: MatShellGetContext(A,(void**)&shell);
282: VecDestroy(&shell->y);
283: if (shell->work) {VecDestroy(&shell->work);}
284: PetscFree(shell);
285: MatDestroy_Shell(A);
286: PetscObjectChangeTypeName((PetscObject)A,0);
287: return(0);
288: }
292: static PetscErrorCode MatWrapML_SeqAIJ(ML_Operator *mlmat,MatReuse reuse,Mat *newmat)
293: {
294: struct ML_CSR_MSRdata *matdata = (struct ML_CSR_MSRdata*)mlmat->data;
295: PetscErrorCode ierr;
296: PetscInt m =mlmat->outvec_leng,n=mlmat->invec_leng,*nnz = NULL,nz_max;
297: PetscInt *ml_cols=matdata->columns,*ml_rowptr=matdata->rowptr,*aj,i;
298: PetscScalar *ml_vals=matdata->values,*aa;
301: if (!mlmat->getrow) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_NULL,"mlmat->getrow = NULL");
302: if (m != n) { /* ML Pmat and Rmat are in CSR format. Pass array pointers into SeqAIJ matrix */
303: if (reuse) {
304: Mat_SeqAIJ *aij= (Mat_SeqAIJ*)(*newmat)->data;
305: aij->i = ml_rowptr;
306: aij->j = ml_cols;
307: aij->a = ml_vals;
308: } else {
309: /* sort ml_cols and ml_vals */
310: PetscMalloc1((m+1),&nnz);
311: for (i=0; i<m; i++) nnz[i] = ml_rowptr[i+1] - ml_rowptr[i];
312: aj = ml_cols; aa = ml_vals;
313: for (i=0; i<m; i++) {
314: PetscSortIntWithScalarArray(nnz[i],aj,aa);
315: aj += nnz[i]; aa += nnz[i];
316: }
317: MatCreateSeqAIJWithArrays(PETSC_COMM_SELF,m,n,ml_rowptr,ml_cols,ml_vals,newmat);
318: PetscFree(nnz);
319: }
320: return(0);
321: }
323: nz_max = PetscMax(1,mlmat->max_nz_per_row);
324: PetscMalloc2(nz_max,&aa,nz_max,&aj);
325: if (!reuse) {
326: MatCreate(PETSC_COMM_SELF,newmat);
327: MatSetSizes(*newmat,m,n,PETSC_DECIDE,PETSC_DECIDE);
328: MatSetType(*newmat,MATSEQAIJ);
329: /* keep track of block size for A matrices */
330: MatSetBlockSize (*newmat, mlmat->num_PDEs);
332: PetscMalloc1(m,&nnz);
333: for (i=0; i<m; i++) {
334: PetscStackCall("ML_Operator_Getrow",ML_Operator_Getrow(mlmat,1,&i,nz_max,aj,aa,&nnz[i]));
335: }
336: MatSeqAIJSetPreallocation(*newmat,0,nnz);
337: }
338: for (i=0; i<m; i++) {
339: PetscInt ncols;
341: PetscStackCall("ML_Operator_Getrow",ML_Operator_Getrow(mlmat,1,&i,nz_max,aj,aa,&ncols));
342: MatSetValues(*newmat,1,&i,ncols,aj,aa,INSERT_VALUES);
343: }
344: MatAssemblyBegin(*newmat,MAT_FINAL_ASSEMBLY);
345: MatAssemblyEnd(*newmat,MAT_FINAL_ASSEMBLY);
347: PetscFree2(aa,aj);
348: PetscFree(nnz);
349: return(0);
350: }
354: static PetscErrorCode MatWrapML_SHELL(ML_Operator *mlmat,MatReuse reuse,Mat *newmat)
355: {
357: PetscInt m,n;
358: ML_Comm *MLcomm;
359: Mat_MLShell *shellctx;
362: m = mlmat->outvec_leng;
363: n = mlmat->invec_leng;
365: if (reuse) {
366: MatShellGetContext(*newmat,(void**)&shellctx);
367: shellctx->mlmat = mlmat;
368: return(0);
369: }
371: MLcomm = mlmat->comm;
373: PetscNew(&shellctx);
374: MatCreateShell(MLcomm->USR_comm,m,n,PETSC_DETERMINE,PETSC_DETERMINE,shellctx,newmat);
375: MatShellSetOperation(*newmat,MATOP_MULT,(void(*)(void))MatMult_ML);
376: MatShellSetOperation(*newmat,MATOP_MULT_ADD,(void(*)(void))MatMultAdd_ML);
378: shellctx->A = *newmat;
379: shellctx->mlmat = mlmat;
380: shellctx->work = NULL;
382: VecCreate(MLcomm->USR_comm,&shellctx->y);
383: VecSetSizes(shellctx->y,m,PETSC_DECIDE);
384: VecSetType(shellctx->y,VECSTANDARD);
386: (*newmat)->ops->destroy = MatDestroy_ML;
387: return(0);
388: }
392: static PetscErrorCode MatWrapML_MPIAIJ(ML_Operator *mlmat,MatReuse reuse,Mat *newmat)
393: {
394: PetscInt *aj;
395: PetscScalar *aa;
397: PetscInt i,j,*gordering;
398: PetscInt m=mlmat->outvec_leng,n,nz_max,row;
399: Mat A;
402: if (!mlmat->getrow) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_NULL,"mlmat->getrow = NULL");
403: n = mlmat->invec_leng;
404: if (m != n) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"m %d must equal to n %d",m,n);
406: /* create global row numbering for a ML_Operator */
407: PetscStackCall("ML_build_global_numbering",ML_build_global_numbering(mlmat,&gordering,"rows"));
409: nz_max = PetscMax(1,mlmat->max_nz_per_row) + 1;
410: PetscMalloc2(nz_max,&aa,nz_max,&aj);
411: if (reuse) {
412: A = *newmat;
413: } else {
414: PetscInt *nnzA,*nnzB,*nnz;
415: PetscInt rstart;
416: MatCreate(mlmat->comm->USR_comm,&A);
417: MatSetSizes(A,m,n,PETSC_DECIDE,PETSC_DECIDE);
418: MatSetType(A,MATMPIAIJ);
419: /* keep track of block size for A matrices */
420: MatSetBlockSize (A,mlmat->num_PDEs);
421: PetscMalloc3(m,&nnzA,m,&nnzB,m,&nnz);
422: MPI_Scan(&m,&rstart,1,MPIU_INT,MPIU_SUM,mlmat->comm->USR_comm);
423: rstart -= m;
425: for (i=0; i<m; i++) {
426: row = gordering[i] - rstart;
427: PetscStackCall("ML_Operator_Getrow",ML_Operator_Getrow(mlmat,1,&i,nz_max,aj,aa,&nnz[i]));
428: nnzA[row] = 0;
429: for (j=0; j<nnz[i]; j++) {
430: if (aj[j] < m) nnzA[row]++;
431: }
432: nnzB[row] = nnz[i] - nnzA[row];
433: }
434: MatMPIAIJSetPreallocation(A,0,nnzA,0,nnzB);
435: PetscFree3(nnzA,nnzB,nnz);
436: }
437: for (i=0; i<m; i++) {
438: PetscInt ncols;
439: row = gordering[i];
441: PetscStackCall(",ML_Operator_Getrow",ML_Operator_Getrow(mlmat,1,&i,nz_max,aj,aa,&ncols));
442: for (j = 0; j < ncols; j++) aj[j] = gordering[aj[j]];
443: MatSetValues(A,1,&row,ncols,aj,aa,INSERT_VALUES);
444: }
445: PetscStackCall("ML_free",ML_free(gordering));
446: MatAssemblyBegin(A,MAT_FINAL_ASSEMBLY);
447: MatAssemblyEnd(A,MAT_FINAL_ASSEMBLY);
448: *newmat = A;
450: PetscFree2(aa,aj);
451: return(0);
452: }
454: /* -------------------------------------------------------------------------- */
455: /*
456: PCSetCoordinates_ML
458: Input Parameter:
459: . pc - the preconditioner context
460: */
463: static PetscErrorCode PCSetCoordinates_ML(PC pc, PetscInt ndm, PetscInt a_nloc, PetscReal *coords)
464: {
465: PC_MG *mg = (PC_MG*)pc->data;
466: PC_ML *pc_ml = (PC_ML*)mg->innerctx;
468: PetscInt arrsz,oldarrsz,bs,my0,kk,ii,nloc,Iend;
469: Mat Amat = pc->pmat;
471: /* this function copied and modified from PCSetCoordinates_GEO -TGI */
474: MatGetBlockSize(Amat, &bs);
476: MatGetOwnershipRange(Amat, &my0, &Iend);
477: nloc = (Iend-my0)/bs;
479: if (nloc!=a_nloc) SETERRQ2(PetscObjectComm((PetscObject)Amat),PETSC_ERR_ARG_WRONG, "Number of local blocks must locations = %d %d.",a_nloc,nloc);
480: if ((Iend-my0)%bs!=0) SETERRQ1(PetscObjectComm((PetscObject)Amat),PETSC_ERR_ARG_WRONG, "Bad local size %d.",nloc);
482: oldarrsz = pc_ml->dim * pc_ml->nloc;
483: pc_ml->dim = ndm;
484: pc_ml->nloc = a_nloc;
485: arrsz = ndm * a_nloc;
487: /* create data - syntactic sugar that should be refactored at some point */
488: if (pc_ml->coords==0 || (oldarrsz != arrsz)) {
489: PetscFree(pc_ml->coords);
490: PetscMalloc1((arrsz), &pc_ml->coords);
491: }
492: for (kk=0; kk<arrsz; kk++) pc_ml->coords[kk] = -999.;
493: /* copy data in - column oriented */
494: for (kk = 0; kk < nloc; kk++) {
495: for (ii = 0; ii < ndm; ii++) {
496: pc_ml->coords[ii*nloc + kk] = coords[kk*ndm + ii];
497: }
498: }
499: return(0);
500: }
502: /* -----------------------------------------------------------------------------*/
503: extern PetscErrorCode PCReset_MG(PC);
506: PetscErrorCode PCReset_ML(PC pc)
507: {
509: PC_MG *mg = (PC_MG*)pc->data;
510: PC_ML *pc_ml = (PC_ML*)mg->innerctx;
511: PetscInt level,fine_level=pc_ml->Nlevels-1,dim=pc_ml->dim;
514: if (dim) {
515: ML_Aggregate_Viz_Stats * grid_info = (ML_Aggregate_Viz_Stats*) pc_ml->ml_object->Grid[0].Grid;
517: for (level=0; level<=fine_level; level++) {
518: VecDestroy(&pc_ml->gridctx[level].coords);
519: }
521: grid_info->x = 0; /* do this so ML doesn't try to free coordinates */
522: grid_info->y = 0;
523: grid_info->z = 0;
525: PetscStackCall("ML_Operator_Getrow",ML_Aggregate_VizAndStats_Clean(pc_ml->ml_object));
526: }
527: PetscStackCall("ML_Aggregate_Destroy",ML_Aggregate_Destroy(&pc_ml->agg_object));
528: PetscStackCall("ML_Aggregate_Destroy",ML_Destroy(&pc_ml->ml_object));
530: if (pc_ml->PetscMLdata) {
531: PetscFree(pc_ml->PetscMLdata->pwork);
532: MatDestroy(&pc_ml->PetscMLdata->Aloc);
533: VecDestroy(&pc_ml->PetscMLdata->x);
534: VecDestroy(&pc_ml->PetscMLdata->y);
535: }
536: PetscFree(pc_ml->PetscMLdata);
538: if (pc_ml->gridctx) {
539: for (level=0; level<fine_level; level++) {
540: if (pc_ml->gridctx[level].A) {MatDestroy(&pc_ml->gridctx[level].A);}
541: if (pc_ml->gridctx[level].P) {MatDestroy(&pc_ml->gridctx[level].P);}
542: if (pc_ml->gridctx[level].R) {MatDestroy(&pc_ml->gridctx[level].R);}
543: if (pc_ml->gridctx[level].x) {VecDestroy(&pc_ml->gridctx[level].x);}
544: if (pc_ml->gridctx[level].b) {VecDestroy(&pc_ml->gridctx[level].b);}
545: if (pc_ml->gridctx[level+1].r) {VecDestroy(&pc_ml->gridctx[level+1].r);}
546: }
547: }
548: PetscFree(pc_ml->gridctx);
549: PetscFree(pc_ml->coords);
551: pc_ml->dim = 0;
552: pc_ml->nloc = 0;
553: PCReset_MG(pc);
554: return(0);
555: }
556: /* -------------------------------------------------------------------------- */
557: /*
558: PCSetUp_ML - Prepares for the use of the ML preconditioner
559: by setting data structures and options.
561: Input Parameter:
562: . pc - the preconditioner context
564: Application Interface Routine: PCSetUp()
566: Notes:
567: The interface routine PCSetUp() is not usually called directly by
568: the user, but instead is called by PCApply() if necessary.
569: */
570: extern PetscErrorCode PCSetFromOptions_MG(PC);
571: extern PetscErrorCode PCReset_MG(PC);
575: PetscErrorCode PCSetUp_ML(PC pc)
576: {
577: PetscErrorCode ierr;
578: PetscMPIInt size;
579: FineGridCtx *PetscMLdata;
580: ML *ml_object;
581: ML_Aggregate *agg_object;
582: ML_Operator *mlmat;
583: PetscInt nlocal_allcols,Nlevels,mllevel,level,level1,m,fine_level,bs;
584: Mat A,Aloc;
585: GridCtx *gridctx;
586: PC_MG *mg = (PC_MG*)pc->data;
587: PC_ML *pc_ml = (PC_ML*)mg->innerctx;
588: PetscBool isSeq, isMPI;
589: KSP smoother;
590: PC subpc;
591: PetscInt mesh_level, old_mesh_level;
592: MatInfo info;
593: static PetscBool cite = PETSC_FALSE;
596: PetscCitationsRegister("@TechReport{ml_users_guide,\n author = {M. Sala and J.J. Hu and R.S. Tuminaro},\n title = {{ML}3.1 {S}moothed {A}ggregation {U}ser's {G}uide},\n institution = {Sandia National Laboratories},\n number = {SAND2004-4821},\n year = 2004\n}\n",&cite);
597: A = pc->pmat;
598: MPI_Comm_size(PetscObjectComm((PetscObject)A),&size);
600: if (pc->setupcalled) {
601: if (pc->flag == SAME_NONZERO_PATTERN && pc_ml->reuse_interpolation) {
602: /*
603: Reuse interpolaton instead of recomputing aggregates and updating the whole hierarchy. This is less expensive for
604: multiple solves in which the matrix is not changing too quickly.
605: */
606: ml_object = pc_ml->ml_object;
607: gridctx = pc_ml->gridctx;
608: Nlevels = pc_ml->Nlevels;
609: fine_level = Nlevels - 1;
610: gridctx[fine_level].A = A;
612: PetscObjectTypeCompare((PetscObject) A, MATSEQAIJ, &isSeq);
613: PetscObjectTypeCompare((PetscObject) A, MATMPIAIJ, &isMPI);
614: if (isMPI) {
615: MatConvert_MPIAIJ_ML(A,NULL,MAT_INITIAL_MATRIX,&Aloc);
616: } else if (isSeq) {
617: Aloc = A;
618: PetscObjectReference((PetscObject)Aloc);
619: } else SETERRQ1(PetscObjectComm((PetscObject)pc),PETSC_ERR_ARG_WRONG, "Matrix type '%s' cannot be used with ML. ML can only handle AIJ matrices.",((PetscObject)A)->type_name);
621: MatGetSize(Aloc,&m,&nlocal_allcols);
622: PetscMLdata = pc_ml->PetscMLdata;
623: MatDestroy(&PetscMLdata->Aloc);
624: PetscMLdata->A = A;
625: PetscMLdata->Aloc = Aloc;
626: PetscStackCall("ML_Aggregate_Destroy",ML_Init_Amatrix(ml_object,0,m,m,PetscMLdata));
627: PetscStackCall("ML_Set_Amatrix_Matvec",ML_Set_Amatrix_Matvec(ml_object,0,PetscML_matvec));
629: mesh_level = ml_object->ML_finest_level;
630: while (ml_object->SingleLevel[mesh_level].Rmat->to) {
631: old_mesh_level = mesh_level;
632: mesh_level = ml_object->SingleLevel[mesh_level].Rmat->to->levelnum;
634: /* clean and regenerate A */
635: mlmat = &(ml_object->Amat[mesh_level]);
636: PetscStackCall("ML_Operator_Clean",ML_Operator_Clean(mlmat));
637: PetscStackCall("ML_Operator_Init",ML_Operator_Init(mlmat,ml_object->comm));
638: PetscStackCall("ML_Gen_AmatrixRAP",ML_Gen_AmatrixRAP(ml_object, old_mesh_level, mesh_level));
639: }
641: level = fine_level - 1;
642: if (size == 1) { /* convert ML P, R and A into seqaij format */
643: for (mllevel=1; mllevel<Nlevels; mllevel++) {
644: mlmat = &(ml_object->Amat[mllevel]);
645: MatWrapML_SeqAIJ(mlmat,MAT_REUSE_MATRIX,&gridctx[level].A);
646: level--;
647: }
648: } else { /* convert ML P and R into shell format, ML A into mpiaij format */
649: for (mllevel=1; mllevel<Nlevels; mllevel++) {
650: mlmat = &(ml_object->Amat[mllevel]);
651: MatWrapML_MPIAIJ(mlmat,MAT_REUSE_MATRIX,&gridctx[level].A);
652: level--;
653: }
654: }
656: for (level=0; level<fine_level; level++) {
657: if (level > 0) {
658: PCMGSetResidual(pc,level,PCMGResidualDefault,gridctx[level].A);
659: }
660: KSPSetOperators(gridctx[level].ksp,gridctx[level].A,gridctx[level].A);
661: }
662: PCMGSetResidual(pc,fine_level,PCMGResidualDefault,gridctx[fine_level].A);
663: KSPSetOperators(gridctx[fine_level].ksp,gridctx[level].A,gridctx[fine_level].A);
665: PCSetUp_MG(pc);
666: return(0);
667: } else {
668: /* since ML can change the size of vectors/matrices at any level we must destroy everything */
669: PCReset_ML(pc);
670: }
671: }
673: /* setup special features of PCML */
674: /*--------------------------------*/
675: /* covert A to Aloc to be used by ML at fine grid */
676: pc_ml->size = size;
677: PetscObjectTypeCompare((PetscObject) A, MATSEQAIJ, &isSeq);
678: PetscObjectTypeCompare((PetscObject) A, MATMPIAIJ, &isMPI);
679: if (isMPI) {
680: MatConvert_MPIAIJ_ML(A,NULL,MAT_INITIAL_MATRIX,&Aloc);
681: } else if (isSeq) {
682: Aloc = A;
683: PetscObjectReference((PetscObject)Aloc);
684: } else SETERRQ1(PetscObjectComm((PetscObject)pc),PETSC_ERR_ARG_WRONG, "Matrix type '%s' cannot be used with ML. ML can only handle AIJ matrices.",((PetscObject)A)->type_name);
686: /* create and initialize struct 'PetscMLdata' */
687: PetscNewLog(pc,&PetscMLdata);
688: pc_ml->PetscMLdata = PetscMLdata;
689: PetscMalloc1((Aloc->cmap->n+1),&PetscMLdata->pwork);
691: VecCreate(PETSC_COMM_SELF,&PetscMLdata->x);
692: VecSetSizes(PetscMLdata->x,Aloc->cmap->n,Aloc->cmap->n);
693: VecSetType(PetscMLdata->x,VECSEQ);
695: VecCreate(PETSC_COMM_SELF,&PetscMLdata->y);
696: VecSetSizes(PetscMLdata->y,A->rmap->n,PETSC_DECIDE);
697: VecSetType(PetscMLdata->y,VECSEQ);
698: PetscMLdata->A = A;
699: PetscMLdata->Aloc = Aloc;
700: if (pc_ml->dim) { /* create vecs around the coordinate data given */
701: PetscInt i,j,dim=pc_ml->dim;
702: PetscInt nloc = pc_ml->nloc,nlocghost;
703: PetscReal *ghostedcoords;
705: MatGetBlockSize(A,&bs);
706: nlocghost = Aloc->cmap->n / bs;
707: PetscMalloc1(dim*nlocghost,&ghostedcoords);
708: for (i = 0; i < dim; i++) {
709: /* copy coordinate values into first component of pwork */
710: for (j = 0; j < nloc; j++) {
711: PetscMLdata->pwork[bs * j] = pc_ml->coords[nloc * i + j];
712: }
713: /* get the ghost values */
714: PetscML_comm(PetscMLdata->pwork,PetscMLdata);
715: /* write into the vector */
716: for (j = 0; j < nlocghost; j++) {
717: ghostedcoords[i * nlocghost + j] = PetscMLdata->pwork[bs * j];
718: }
719: }
720: /* replace the original coords with the ghosted coords, because these are
721: * what ML needs */
722: PetscFree(pc_ml->coords);
723: pc_ml->coords = ghostedcoords;
724: }
726: /* create ML discretization matrix at fine grid */
727: /* ML requires input of fine-grid matrix. It determines nlevels. */
728: MatGetSize(Aloc,&m,&nlocal_allcols);
729: MatGetBlockSize(A,&bs);
730: PetscStackCall("ML_Create",ML_Create(&ml_object,pc_ml->MaxNlevels));
731: PetscStackCall("ML_Comm_Set_UsrComm",ML_Comm_Set_UsrComm(ml_object->comm,PetscObjectComm((PetscObject)A)));
732: pc_ml->ml_object = ml_object;
733: PetscStackCall("ML_Init_Amatrix",ML_Init_Amatrix(ml_object,0,m,m,PetscMLdata));
734: PetscStackCall("ML_Set_Amatrix_Getrow",ML_Set_Amatrix_Getrow(ml_object,0,PetscML_getrow,PetscML_comm,nlocal_allcols));
735: PetscStackCall("ML_Set_Amatrix_Matvec",ML_Set_Amatrix_Matvec(ml_object,0,PetscML_matvec));
737: PetscStackCall("ML_Set_Symmetrize",ML_Set_Symmetrize(ml_object,pc_ml->Symmetrize ? ML_YES : ML_NO));
739: /* aggregation */
740: PetscStackCall("ML_Aggregate_Create",ML_Aggregate_Create(&agg_object));
741: pc_ml->agg_object = agg_object;
743: {
744: MatNullSpace mnull;
745: MatGetNearNullSpace(A,&mnull);
746: if (pc_ml->nulltype == PCML_NULLSPACE_AUTO) {
747: if (mnull) pc_ml->nulltype = PCML_NULLSPACE_USER;
748: else if (bs > 1) pc_ml->nulltype = PCML_NULLSPACE_BLOCK;
749: else pc_ml->nulltype = PCML_NULLSPACE_SCALAR;
750: }
751: switch (pc_ml->nulltype) {
752: case PCML_NULLSPACE_USER: {
753: PetscScalar *nullvec;
754: const PetscScalar *v;
755: PetscBool has_const;
756: PetscInt i,j,mlocal,nvec,M;
757: const Vec *vecs;
759: if (!mnull) SETERRQ(PetscObjectComm((PetscObject)pc),PETSC_ERR_USER,"Must provide explicit null space using MatSetNearNullSpace() to use user-specified null space");
760: MatGetSize(A,&M,NULL);
761: MatGetLocalSize(Aloc,&mlocal,NULL);
762: MatNullSpaceGetVecs(mnull,&has_const,&nvec,&vecs);
763: PetscMalloc1((nvec+!!has_const)*mlocal,&nullvec);
764: if (has_const) for (i=0; i<mlocal; i++) nullvec[i] = 1.0/M;
765: for (i=0; i<nvec; i++) {
766: VecGetArrayRead(vecs[i],&v);
767: for (j=0; j<mlocal; j++) nullvec[(i+!!has_const)*mlocal + j] = v[j];
768: VecRestoreArrayRead(vecs[i],&v);
769: }
770: PetscStackCall("ML_Aggregate_Create",ML_Aggregate_Set_NullSpace(agg_object,bs,nvec+!!has_const,nullvec,mlocal);CHKERRQ(ierr));
771: PetscFree(nullvec);
772: } break;
773: case PCML_NULLSPACE_BLOCK:
774: PetscStackCall("ML_Aggregate_Set_NullSpace",ML_Aggregate_Set_NullSpace(agg_object,bs,bs,0,0);CHKERRQ(ierr));
775: break;
776: case PCML_NULLSPACE_SCALAR:
777: break;
778: default: SETERRQ(PetscObjectComm((PetscObject)pc),PETSC_ERR_SUP,"Unknown null space type");
779: }
780: }
781: PetscStackCall("ML_Aggregate_Set_MaxCoarseSize",ML_Aggregate_Set_MaxCoarseSize(agg_object,pc_ml->MaxCoarseSize));
782: /* set options */
783: switch (pc_ml->CoarsenScheme) {
784: case 1:
785: PetscStackCall("ML_Aggregate_Set_CoarsenScheme_Coupled",ML_Aggregate_Set_CoarsenScheme_Coupled(agg_object));break;
786: case 2:
787: PetscStackCall("ML_Aggregate_Set_CoarsenScheme_MIS",ML_Aggregate_Set_CoarsenScheme_MIS(agg_object));break;
788: case 3:
789: PetscStackCall("ML_Aggregate_Set_CoarsenScheme_METIS",ML_Aggregate_Set_CoarsenScheme_METIS(agg_object));break;
790: }
791: PetscStackCall("ML_Aggregate_Set_Threshold",ML_Aggregate_Set_Threshold(agg_object,pc_ml->Threshold));
792: PetscStackCall("ML_Aggregate_Set_DampingFactor",ML_Aggregate_Set_DampingFactor(agg_object,pc_ml->DampingFactor));
793: if (pc_ml->SpectralNormScheme_Anorm) {
794: PetscStackCall("ML_Set_SpectralNormScheme_Anorm",ML_Set_SpectralNormScheme_Anorm(ml_object));
795: }
796: agg_object->keep_agg_information = (int)pc_ml->KeepAggInfo;
797: agg_object->keep_P_tentative = (int)pc_ml->Reusable;
798: agg_object->block_scaled_SA = (int)pc_ml->BlockScaling;
799: agg_object->minimizing_energy = (int)pc_ml->EnergyMinimization;
800: agg_object->minimizing_energy_droptol = (double)pc_ml->EnergyMinimizationDropTol;
801: agg_object->cheap_minimizing_energy = (int)pc_ml->EnergyMinimizationCheap;
803: if (pc_ml->Aux) {
804: if (!pc_ml->dim) SETERRQ(PetscObjectComm((PetscObject)pc),PETSC_ERR_USER,"Auxiliary matrix requires coordinates");
805: ml_object->Amat[0].aux_data->threshold = pc_ml->AuxThreshold;
806: ml_object->Amat[0].aux_data->enable = 1;
807: ml_object->Amat[0].aux_data->max_level = 10;
808: ml_object->Amat[0].num_PDEs = bs;
809: }
811: MatGetInfo(A,MAT_LOCAL,&info);
812: ml_object->Amat[0].N_nonzeros = (int) info.nz_used;
814: if (pc_ml->dim) {
815: PetscInt i,dim = pc_ml->dim;
816: ML_Aggregate_Viz_Stats *grid_info;
817: PetscInt nlocghost;
819: MatGetBlockSize(A,&bs);
820: nlocghost = Aloc->cmap->n / bs;
822: PetscStackCall("ML_Aggregate_VizAndStats_Setup(",ML_Aggregate_VizAndStats_Setup(ml_object)); /* create ml info for coords */
823: grid_info = (ML_Aggregate_Viz_Stats*) ml_object->Grid[0].Grid;
824: for (i = 0; i < dim; i++) {
825: /* set the finest level coordinates to point to the column-order array
826: * in pc_ml */
827: /* NOTE: must point away before VizAndStats_Clean so ML doesn't free */
828: switch (i) {
829: case 0: grid_info->x = pc_ml->coords + nlocghost * i; break;
830: case 1: grid_info->y = pc_ml->coords + nlocghost * i; break;
831: case 2: grid_info->z = pc_ml->coords + nlocghost * i; break;
832: default: SETERRQ(PetscObjectComm((PetscObject)pc),PETSC_ERR_ARG_SIZ,"PCML coordinate dimension must be <= 3");
833: }
834: }
835: grid_info->Ndim = dim;
836: }
838: /* repartitioning */
839: if (pc_ml->Repartition) {
840: PetscStackCall("ML_Repartition_Activate",ML_Repartition_Activate(ml_object));
841: PetscStackCall("ML_Repartition_Set_LargestMinMaxRatio",ML_Repartition_Set_LargestMinMaxRatio(ml_object,pc_ml->MaxMinRatio));
842: PetscStackCall("ML_Repartition_Set_MinPerProc",ML_Repartition_Set_MinPerProc(ml_object,pc_ml->MinPerProc));
843: PetscStackCall("ML_Repartition_Set_PutOnSingleProc",ML_Repartition_Set_PutOnSingleProc(ml_object,pc_ml->PutOnSingleProc));
844: #if 0 /* Function not yet defined in ml-6.2 */
845: /* I'm not sure what compatibility issues might crop up if we partitioned
846: * on the finest level, so to be safe repartition starts on the next
847: * finest level (reflection default behavior in
848: * ml_MultiLevelPreconditioner) */
849: PetscStackCall("ML_Repartition_Set_StartLevel",ML_Repartition_Set_StartLevel(ml_object,1));
850: #endif
852: if (!pc_ml->RepartitionType) {
853: PetscInt i;
855: if (!pc_ml->dim) SETERRQ(PetscObjectComm((PetscObject)pc),PETSC_ERR_USER,"ML Zoltan repartitioning requires coordinates");
856: PetscStackCall("ML_Repartition_Set_Partitioner",ML_Repartition_Set_Partitioner(ml_object,ML_USEZOLTAN));
857: PetscStackCall("ML_Aggregate_Set_Dimensions",ML_Aggregate_Set_Dimensions(agg_object, pc_ml->dim));
859: for (i = 0; i < ml_object->ML_num_levels; i++) {
860: ML_Aggregate_Viz_Stats *grid_info = (ML_Aggregate_Viz_Stats*)ml_object->Grid[i].Grid;
861: grid_info->zoltan_type = pc_ml->ZoltanScheme + 1; /* ml numbers options 1, 2, 3 */
862: /* defaults from ml_agg_info.c */
863: grid_info->zoltan_estimated_its = 40; /* only relevant to hypergraph / fast hypergraph */
864: grid_info->zoltan_timers = 0;
865: grid_info->smoothing_steps = 4; /* only relevant to hypergraph / fast hypergraph */
866: }
867: } else {
868: PetscStackCall("ML_Repartition_Set_Partitioner",ML_Repartition_Set_Partitioner(ml_object,ML_USEPARMETIS));
869: }
870: }
872: if (pc_ml->OldHierarchy) {
873: PetscStackCall("ML_Gen_MGHierarchy_UsingAggregation",Nlevels = ML_Gen_MGHierarchy_UsingAggregation(ml_object,0,ML_INCREASING,agg_object));
874: } else {
875: PetscStackCall("ML_Gen_MultiLevelHierarchy_UsingAggregation",Nlevels = ML_Gen_MultiLevelHierarchy_UsingAggregation(ml_object,0,ML_INCREASING,agg_object));
876: }
877: if (Nlevels<=0) SETERRQ1(PetscObjectComm((PetscObject)pc),PETSC_ERR_ARG_OUTOFRANGE,"Nlevels %d must > 0",Nlevels);
878: pc_ml->Nlevels = Nlevels;
879: fine_level = Nlevels - 1;
881: PCMGSetLevels(pc,Nlevels,NULL);
882: /* set default smoothers */
883: for (level=1; level<=fine_level; level++) {
884: PCMGGetSmoother(pc,level,&smoother);
885: KSPSetType(smoother,KSPRICHARDSON);
886: KSPGetPC(smoother,&subpc);
887: PCSetType(subpc,PCSOR);
888: }
889: PetscObjectOptionsBegin((PetscObject)pc);
890: PCSetFromOptions_MG(pc); /* should be called in PCSetFromOptions_ML(), but cannot be called prior to PCMGSetLevels() */
891: PetscOptionsEnd();
893: PetscMalloc1(Nlevels,&gridctx);
895: pc_ml->gridctx = gridctx;
897: /* wrap ML matrices by PETSc shell matrices at coarsened grids.
898: Level 0 is the finest grid for ML, but coarsest for PETSc! */
899: gridctx[fine_level].A = A;
901: level = fine_level - 1;
902: if (size == 1) { /* convert ML P, R and A into seqaij format */
903: for (mllevel=1; mllevel<Nlevels; mllevel++) {
904: mlmat = &(ml_object->Pmat[mllevel]);
905: MatWrapML_SeqAIJ(mlmat,MAT_INITIAL_MATRIX,&gridctx[level].P);
906: mlmat = &(ml_object->Rmat[mllevel-1]);
907: MatWrapML_SeqAIJ(mlmat,MAT_INITIAL_MATRIX,&gridctx[level].R);
909: mlmat = &(ml_object->Amat[mllevel]);
910: MatWrapML_SeqAIJ(mlmat,MAT_INITIAL_MATRIX,&gridctx[level].A);
911: level--;
912: }
913: } else { /* convert ML P and R into shell format, ML A into mpiaij format */
914: for (mllevel=1; mllevel<Nlevels; mllevel++) {
915: mlmat = &(ml_object->Pmat[mllevel]);
916: MatWrapML_SHELL(mlmat,MAT_INITIAL_MATRIX,&gridctx[level].P);
917: mlmat = &(ml_object->Rmat[mllevel-1]);
918: MatWrapML_SHELL(mlmat,MAT_INITIAL_MATRIX,&gridctx[level].R);
920: mlmat = &(ml_object->Amat[mllevel]);
921: MatWrapML_MPIAIJ(mlmat,MAT_INITIAL_MATRIX,&gridctx[level].A);
922: level--;
923: }
924: }
926: /* create vectors and ksp at all levels */
927: for (level=0; level<fine_level; level++) {
928: level1 = level + 1;
929: VecCreate(((PetscObject)gridctx[level].A)->comm,&gridctx[level].x);
930: VecSetSizes(gridctx[level].x,gridctx[level].A->cmap->n,PETSC_DECIDE);
931: VecSetType(gridctx[level].x,VECMPI);
932: PCMGSetX(pc,level,gridctx[level].x);
934: VecCreate(((PetscObject)gridctx[level].A)->comm,&gridctx[level].b);
935: VecSetSizes(gridctx[level].b,gridctx[level].A->rmap->n,PETSC_DECIDE);
936: VecSetType(gridctx[level].b,VECMPI);
937: PCMGSetRhs(pc,level,gridctx[level].b);
939: VecCreate(((PetscObject)gridctx[level1].A)->comm,&gridctx[level1].r);
940: VecSetSizes(gridctx[level1].r,gridctx[level1].A->rmap->n,PETSC_DECIDE);
941: VecSetType(gridctx[level1].r,VECMPI);
942: PCMGSetR(pc,level1,gridctx[level1].r);
944: if (level == 0) {
945: PCMGGetCoarseSolve(pc,&gridctx[level].ksp);
946: } else {
947: PCMGGetSmoother(pc,level,&gridctx[level].ksp);
948: }
949: }
950: PCMGGetSmoother(pc,fine_level,&gridctx[fine_level].ksp);
952: /* create coarse level and the interpolation between the levels */
953: for (level=0; level<fine_level; level++) {
954: level1 = level + 1;
955: PCMGSetInterpolation(pc,level1,gridctx[level].P);
956: PCMGSetRestriction(pc,level1,gridctx[level].R);
957: if (level > 0) {
958: PCMGSetResidual(pc,level,PCMGResidualDefault,gridctx[level].A);
959: }
960: KSPSetOperators(gridctx[level].ksp,gridctx[level].A,gridctx[level].A);
961: }
962: PCMGSetResidual(pc,fine_level,PCMGResidualDefault,gridctx[fine_level].A);
963: KSPSetOperators(gridctx[fine_level].ksp,gridctx[level].A,gridctx[fine_level].A);
965: /* put coordinate info in levels */
966: if (pc_ml->dim) {
967: PetscInt i,j,dim = pc_ml->dim;
968: PetscInt bs, nloc;
969: PC subpc;
970: PetscReal *array;
972: level = fine_level;
973: for (mllevel = 0; mllevel < Nlevels; mllevel++) {
974: ML_Aggregate_Viz_Stats *grid_info = (ML_Aggregate_Viz_Stats*)ml_object->Amat[mllevel].to->Grid->Grid;
975: MPI_Comm comm = ((PetscObject)gridctx[level].A)->comm;
977: MatGetBlockSize (gridctx[level].A, &bs);
978: MatGetLocalSize (gridctx[level].A, NULL, &nloc);
979: nloc /= bs; /* number of local nodes */
981: VecCreate(comm,&gridctx[level].coords);
982: VecSetSizes(gridctx[level].coords,dim * nloc,PETSC_DECIDE);
983: VecSetType(gridctx[level].coords,VECMPI);
984: VecGetArray(gridctx[level].coords,&array);
985: for (j = 0; j < nloc; j++) {
986: for (i = 0; i < dim; i++) {
987: switch (i) {
988: case 0: array[dim * j + i] = grid_info->x[j]; break;
989: case 1: array[dim * j + i] = grid_info->y[j]; break;
990: case 2: array[dim * j + i] = grid_info->z[j]; break;
991: default: SETERRQ(PetscObjectComm((PetscObject)pc),PETSC_ERR_ARG_SIZ,"PCML coordinate dimension must be <= 3");
992: }
993: }
994: }
996: /* passing coordinates to smoothers/coarse solver, should they need them */
997: KSPGetPC(gridctx[level].ksp,&subpc);
998: PCSetCoordinates(subpc,dim,nloc,array);
999: VecRestoreArray(gridctx[level].coords,&array);
1000: level--;
1001: }
1002: }
1004: /* setupcalled is set to 0 so that MG is setup from scratch */
1005: pc->setupcalled = 0;
1006: PCSetUp_MG(pc);
1007: return(0);
1008: }
1010: /* -------------------------------------------------------------------------- */
1011: /*
1012: PCDestroy_ML - Destroys the private context for the ML preconditioner
1013: that was created with PCCreate_ML().
1015: Input Parameter:
1016: . pc - the preconditioner context
1018: Application Interface Routine: PCDestroy()
1019: */
1022: PetscErrorCode PCDestroy_ML(PC pc)
1023: {
1025: PC_MG *mg = (PC_MG*)pc->data;
1026: PC_ML *pc_ml= (PC_ML*)mg->innerctx;
1029: PCReset_ML(pc);
1030: PetscFree(pc_ml);
1031: PCDestroy_MG(pc);
1032: PetscObjectComposeFunction((PetscObject)pc,"PCSetCoordinates_C",NULL);
1033: return(0);
1034: }
1038: PetscErrorCode PCSetFromOptions_ML(PC pc)
1039: {
1041: PetscInt indx,PrintLevel,partindx;
1042: const char *scheme[] = {"Uncoupled","Coupled","MIS","METIS"};
1043: const char *part[] = {"Zoltan","ParMETIS"};
1044: #if defined(HAVE_ML_ZOLTAN)
1045: PetscInt zidx;
1046: const char *zscheme[] = {"RCB","hypergraph","fast_hypergraph"};
1047: #endif
1048: PC_MG *mg = (PC_MG*)pc->data;
1049: PC_ML *pc_ml = (PC_ML*)mg->innerctx;
1050: PetscMPIInt size;
1051: MPI_Comm comm;
1054: PetscObjectGetComm((PetscObject)pc,&comm);
1055: MPI_Comm_size(comm,&size);
1056: PetscOptionsHead("ML options");
1058: PrintLevel = 0;
1059: indx = 0;
1060: partindx = 0;
1062: PetscOptionsInt("-pc_ml_PrintLevel","Print level","ML_Set_PrintLevel",PrintLevel,&PrintLevel,NULL);
1063: PetscStackCall("ML_Set_PrintLeve",ML_Set_PrintLevel(PrintLevel));
1064: PetscOptionsInt("-pc_ml_maxNlevels","Maximum number of levels","None",pc_ml->MaxNlevels,&pc_ml->MaxNlevels,NULL);
1065: PetscOptionsInt("-pc_ml_maxCoarseSize","Maximum coarsest mesh size","ML_Aggregate_Set_MaxCoarseSize",pc_ml->MaxCoarseSize,&pc_ml->MaxCoarseSize,NULL);
1066: PetscOptionsEList("-pc_ml_CoarsenScheme","Aggregate Coarsen Scheme","ML_Aggregate_Set_CoarsenScheme_*",scheme,4,scheme[0],&indx,NULL);
1068: pc_ml->CoarsenScheme = indx;
1070: PetscOptionsReal("-pc_ml_DampingFactor","P damping factor","ML_Aggregate_Set_DampingFactor",pc_ml->DampingFactor,&pc_ml->DampingFactor,NULL);
1071: PetscOptionsReal("-pc_ml_Threshold","Smoother drop tol","ML_Aggregate_Set_Threshold",pc_ml->Threshold,&pc_ml->Threshold,NULL);
1072: PetscOptionsBool("-pc_ml_SpectralNormScheme_Anorm","Method used for estimating spectral radius","ML_Set_SpectralNormScheme_Anorm",pc_ml->SpectralNormScheme_Anorm,&pc_ml->SpectralNormScheme_Anorm,NULL);
1073: PetscOptionsBool("-pc_ml_Symmetrize","Symmetrize aggregation","ML_Set_Symmetrize",pc_ml->Symmetrize,&pc_ml->Symmetrize,NULL);
1074: PetscOptionsBool("-pc_ml_BlockScaling","Scale all dofs at each node together","None",pc_ml->BlockScaling,&pc_ml->BlockScaling,NULL);
1075: PetscOptionsEnum("-pc_ml_nullspace","Which type of null space information to use","None",PCMLNullSpaceTypes,(PetscEnum)pc_ml->nulltype,(PetscEnum*)&pc_ml->nulltype,NULL);
1076: PetscOptionsInt("-pc_ml_EnergyMinimization","Energy minimization norm type (0=no minimization; see ML manual for 1,2,3; -1 and 4 undocumented)","None",pc_ml->EnergyMinimization,&pc_ml->EnergyMinimization,NULL);
1077: PetscOptionsBool("-pc_ml_reuse_interpolation","Reuse the interpolation operators when possible (cheaper, weaker when matrix entries change a lot)","None",pc_ml->reuse_interpolation,&pc_ml->reuse_interpolation,NULL);
1078: /*
1079: The following checks a number of conditions. If we let this stuff slip by, then ML's error handling will take over.
1080: This is suboptimal because it amounts to calling exit(1) so we check for the most common conditions.
1082: We also try to set some sane defaults when energy minimization is activated, otherwise it's hard to find a working
1083: combination of options and ML's exit(1) explanations don't help matters.
1084: */
1085: if (pc_ml->EnergyMinimization < -1 || pc_ml->EnergyMinimization > 4) SETERRQ(comm,PETSC_ERR_ARG_OUTOFRANGE,"EnergyMinimization must be in range -1..4");
1086: if (pc_ml->EnergyMinimization == 4 && size > 1) SETERRQ(comm,PETSC_ERR_SUP,"Energy minimization type 4 does not work in parallel");
1087: if (pc_ml->EnergyMinimization == 4) {PetscInfo(pc,"Mandel's energy minimization scheme is experimental and broken in ML-6.2");}
1088: if (pc_ml->EnergyMinimization) {
1089: PetscOptionsReal("-pc_ml_EnergyMinimizationDropTol","Energy minimization drop tolerance","None",pc_ml->EnergyMinimizationDropTol,&pc_ml->EnergyMinimizationDropTol,NULL);
1090: }
1091: if (pc_ml->EnergyMinimization == 2) {
1092: /* According to ml_MultiLevelPreconditioner.cpp, this option is only meaningful for norm type (2) */
1093: PetscOptionsBool("-pc_ml_EnergyMinimizationCheap","Use cheaper variant of norm type 2","None",pc_ml->EnergyMinimizationCheap,&pc_ml->EnergyMinimizationCheap,NULL);
1094: }
1095: /* energy minimization sometimes breaks if this is turned off, the more classical stuff should be okay without it */
1096: if (pc_ml->EnergyMinimization) pc_ml->KeepAggInfo = PETSC_TRUE;
1097: PetscOptionsBool("-pc_ml_KeepAggInfo","Allows the preconditioner to be reused, or auxilliary matrices to be generated","None",pc_ml->KeepAggInfo,&pc_ml->KeepAggInfo,NULL);
1098: /* Option (-1) doesn't work at all (calls exit(1)) if the tentative restriction operator isn't stored. */
1099: if (pc_ml->EnergyMinimization == -1) pc_ml->Reusable = PETSC_TRUE;
1100: PetscOptionsBool("-pc_ml_Reusable","Store intermedaiate data structures so that the multilevel hierarchy is reusable","None",pc_ml->Reusable,&pc_ml->Reusable,NULL);
1101: /*
1102: ML's C API is severely underdocumented and lacks significant functionality. The C++ API calls
1103: ML_Gen_MultiLevelHierarchy_UsingAggregation() which is a modified copy (!?) of the documented function
1104: ML_Gen_MGHierarchy_UsingAggregation(). This modification, however, does not provide a strict superset of the
1105: functionality in the old function, so some users may still want to use it. Note that many options are ignored in
1106: this context, but ML doesn't provide a way to find out which ones.
1107: */
1108: PetscOptionsBool("-pc_ml_OldHierarchy","Use old routine to generate hierarchy","None",pc_ml->OldHierarchy,&pc_ml->OldHierarchy,NULL);
1109: PetscOptionsBool("-pc_ml_repartition", "Allow ML to repartition levels of the heirarchy","ML_Repartition_Activate",pc_ml->Repartition,&pc_ml->Repartition,NULL);
1110: if (pc_ml->Repartition) {
1111: PetscOptionsReal("-pc_ml_repartitionMaxMinRatio", "Acceptable ratio of repartitioned sizes","ML_Repartition_Set_LargestMinMaxRatio",pc_ml->MaxMinRatio,&pc_ml->MaxMinRatio,NULL);
1112: PetscOptionsInt("-pc_ml_repartitionMinPerProc", "Smallest repartitioned size","ML_Repartition_Set_MinPerProc",pc_ml->MinPerProc,&pc_ml->MinPerProc,NULL);
1113: PetscOptionsInt("-pc_ml_repartitionPutOnSingleProc", "Problem size automatically repartitioned to one processor","ML_Repartition_Set_PutOnSingleProc",pc_ml->PutOnSingleProc,&pc_ml->PutOnSingleProc,NULL);
1114: #if defined(HAVE_ML_ZOLTAN)
1115: partindx = 0;
1116: PetscOptionsEList("-pc_ml_repartitionType", "Repartitioning library to use","ML_Repartition_Set_Partitioner",part,2,part[0],&partindx,NULL);
1118: pc_ml->RepartitionType = partindx;
1119: if (!partindx) {
1120: PetscInt zindx = 0;
1122: PetscOptionsEList("-pc_ml_repartitionZoltanScheme", "Repartitioning scheme to use","None",zscheme,3,zscheme[0],&zindx,NULL);
1124: pc_ml->ZoltanScheme = zindx;
1125: }
1126: #else
1127: partindx = 1;
1128: PetscOptionsEList("-pc_ml_repartitionType", "Repartitioning library to use","ML_Repartition_Set_Partitioner",part,2,part[1],&partindx,NULL);
1129: pc_ml->RepartitionType = partindx;
1130: if (!partindx) SETERRQ(PetscObjectComm((PetscObject)pc),PETSC_ERR_SUP_SYS,"ML not compiled with Zoltan");
1131: #endif
1132: PetscOptionsBool("-pc_ml_Aux","Aggregate using auxiliary coordinate-based laplacian","None",pc_ml->Aux,&pc_ml->Aux,NULL);
1133: PetscOptionsReal("-pc_ml_AuxThreshold","Auxiliary smoother drop tol","None",pc_ml->AuxThreshold,&pc_ml->AuxThreshold,NULL);
1134: }
1135: PetscOptionsTail();
1136: return(0);
1137: }
1139: /* -------------------------------------------------------------------------- */
1140: /*
1141: PCCreate_ML - Creates a ML preconditioner context, PC_ML,
1142: and sets this as the private data within the generic preconditioning
1143: context, PC, that was created within PCCreate().
1145: Input Parameter:
1146: . pc - the preconditioner context
1148: Application Interface Routine: PCCreate()
1149: */
1151: /*MC
1152: PCML - Use algebraic multigrid preconditioning. This preconditioner requires you provide
1153: fine grid discretization matrix. The coarser grid matrices and restriction/interpolation
1154: operators are computed by ML, with the matrices coverted to PETSc matrices in aij format
1155: and the restriction/interpolation operators wrapped as PETSc shell matrices.
1157: Options Database Key:
1158: Multigrid options(inherited):
1159: + -pc_mg_cycles <1>: 1 for V cycle, 2 for W-cycle (MGSetCycles)
1160: . -pc_mg_smoothup <1>: Number of post-smoothing steps (MGSetNumberSmoothUp)
1161: . -pc_mg_smoothdown <1>: Number of pre-smoothing steps (MGSetNumberSmoothDown)
1162: - -pc_mg_type <multiplicative>: (one of) additive multiplicative full kascade
1163: ML options:
1164: + -pc_ml_PrintLevel <0>: Print level (ML_Set_PrintLevel)
1165: . -pc_ml_maxNlevels <10>: Maximum number of levels (None)
1166: . -pc_ml_maxCoarseSize <1>: Maximum coarsest mesh size (ML_Aggregate_Set_MaxCoarseSize)
1167: . -pc_ml_CoarsenScheme <Uncoupled>: (one of) Uncoupled Coupled MIS METIS
1168: . -pc_ml_DampingFactor <1.33333>: P damping factor (ML_Aggregate_Set_DampingFactor)
1169: . -pc_ml_Threshold <0>: Smoother drop tol (ML_Aggregate_Set_Threshold)
1170: . -pc_ml_SpectralNormScheme_Anorm <false>: Method used for estimating spectral radius (ML_Set_SpectralNormScheme_Anorm)
1171: . -pc_ml_repartition <false>: Allow ML to repartition levels of the heirarchy (ML_Repartition_Activate)
1172: . -pc_ml_repartitionMaxMinRatio <1.3>: Acceptable ratio of repartitioned sizes (ML_Repartition_Set_LargestMinMaxRatio)
1173: . -pc_ml_repartitionMinPerProc <512>: Smallest repartitioned size (ML_Repartition_Set_MinPerProc)
1174: . -pc_ml_repartitionPutOnSingleProc <5000>: Problem size automatically repartitioned to one processor (ML_Repartition_Set_PutOnSingleProc)
1175: . -pc_ml_repartitionType <Zoltan>: Repartitioning library to use (ML_Repartition_Set_Partitioner)
1176: . -pc_ml_repartitionZoltanScheme <RCB>: Repartitioning scheme to use (None)
1177: . -pc_ml_Aux <false>: Aggregate using auxiliary coordinate-based laplacian (None)
1178: - -pc_ml_AuxThreshold <0.0>: Auxiliary smoother drop tol (None)
1180: Level: intermediate
1182: Concepts: multigrid
1184: .seealso: PCCreate(), PCSetType(), PCType (for list of available types), PC, PCMGType,
1185: PCMGSetLevels(), PCMGGetLevels(), PCMGSetType(), MPSetCycles(), PCMGSetNumberSmoothDown(),
1186: PCMGSetNumberSmoothUp(), PCMGGetCoarseSolve(), PCMGSetResidual(), PCMGSetInterpolation(),
1187: PCMGSetRestriction(), PCMGGetSmoother(), PCMGGetSmootherUp(), PCMGGetSmootherDown(),
1188: PCMGSetCyclesOnLevel(), PCMGSetRhs(), PCMGSetX(), PCMGSetR()
1189: M*/
1193: PETSC_EXTERN PetscErrorCode PCCreate_ML(PC pc)
1194: {
1196: PC_ML *pc_ml;
1197: PC_MG *mg;
1200: /* PCML is an inherited class of PCMG. Initialize pc as PCMG */
1201: PCSetType(pc,PCMG); /* calls PCCreate_MG() and MGCreate_Private() */
1202: PetscObjectChangeTypeName((PetscObject)pc,PCML);
1203: /* Since PCMG tries to use DM assocated with PC must delete it */
1204: DMDestroy(&pc->dm);
1205: mg = (PC_MG*)pc->data;
1206: mg->galerkin = 2; /* Use Galerkin, but it is computed externally */
1208: /* create a supporting struct and attach it to pc */
1209: PetscNewLog(pc,&pc_ml);
1210: mg->innerctx = pc_ml;
1212: pc_ml->ml_object = 0;
1213: pc_ml->agg_object = 0;
1214: pc_ml->gridctx = 0;
1215: pc_ml->PetscMLdata = 0;
1216: pc_ml->Nlevels = -1;
1217: pc_ml->MaxNlevels = 10;
1218: pc_ml->MaxCoarseSize = 1;
1219: pc_ml->CoarsenScheme = 1;
1220: pc_ml->Threshold = 0.0;
1221: pc_ml->DampingFactor = 4.0/3.0;
1222: pc_ml->SpectralNormScheme_Anorm = PETSC_FALSE;
1223: pc_ml->size = 0;
1224: pc_ml->dim = 0;
1225: pc_ml->nloc = 0;
1226: pc_ml->coords = 0;
1227: pc_ml->Repartition = PETSC_FALSE;
1228: pc_ml->MaxMinRatio = 1.3;
1229: pc_ml->MinPerProc = 512;
1230: pc_ml->PutOnSingleProc = 5000;
1231: pc_ml->RepartitionType = 0;
1232: pc_ml->ZoltanScheme = 0;
1233: pc_ml->Aux = PETSC_FALSE;
1234: pc_ml->AuxThreshold = 0.0;
1236: /* allow for coordinates to be passed */
1237: PetscObjectComposeFunction((PetscObject)pc,"PCSetCoordinates_C",PCSetCoordinates_ML);
1239: /* overwrite the pointers of PCMG by the functions of PCML */
1240: pc->ops->setfromoptions = PCSetFromOptions_ML;
1241: pc->ops->setup = PCSetUp_ML;
1242: pc->ops->reset = PCReset_ML;
1243: pc->ops->destroy = PCDestroy_ML;
1244: return(0);
1245: }