Actual source code: mpidense.c
petsc-3.5.0 2014-06-30
2: /*
3: Basic functions for basic parallel dense matrices.
4: */
7: #include <../src/mat/impls/dense/mpi/mpidense.h> /*I "petscmat.h" I*/
8: #include <../src/mat/impls/aij/mpi/mpiaij.h>
12: /*@
14: MatDenseGetLocalMatrix - For a MATMPIDENSE or MATSEQDENSE matrix returns the sequential
15: matrix that represents the operator. For sequential matrices it returns itself.
17: Input Parameter:
18: . A - the Seq or MPI dense matrix
20: Output Parameter:
21: . B - the inner matrix
23: Level: intermediate
25: @*/
26: PetscErrorCode MatDenseGetLocalMatrix(Mat A,Mat *B)
27: {
28: Mat_MPIDense *mat = (Mat_MPIDense*)A->data;
30: PetscBool flg;
33: PetscObjectTypeCompare((PetscObject)A,MATMPIDENSE,&flg);
34: if (flg) *B = mat->A;
35: else *B = A;
36: return(0);
37: }
41: PetscErrorCode MatGetRow_MPIDense(Mat A,PetscInt row,PetscInt *nz,PetscInt **idx,PetscScalar **v)
42: {
43: Mat_MPIDense *mat = (Mat_MPIDense*)A->data;
45: PetscInt lrow,rstart = A->rmap->rstart,rend = A->rmap->rend;
48: if (row < rstart || row >= rend) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"only local rows");
49: lrow = row - rstart;
50: MatGetRow(mat->A,lrow,nz,(const PetscInt**)idx,(const PetscScalar**)v);
51: return(0);
52: }
56: PetscErrorCode MatRestoreRow_MPIDense(Mat mat,PetscInt row,PetscInt *nz,PetscInt **idx,PetscScalar **v)
57: {
61: if (idx) {PetscFree(*idx);}
62: if (v) {PetscFree(*v);}
63: return(0);
64: }
68: PetscErrorCode MatGetDiagonalBlock_MPIDense(Mat A,Mat *a)
69: {
70: Mat_MPIDense *mdn = (Mat_MPIDense*)A->data;
72: PetscInt m = A->rmap->n,rstart = A->rmap->rstart;
73: PetscScalar *array;
74: MPI_Comm comm;
75: Mat B;
78: if (A->rmap->N != A->cmap->N) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Only square matrices supported.");
80: PetscObjectQuery((PetscObject)A,"DiagonalBlock",(PetscObject*)&B);
81: if (!B) {
82: PetscObjectGetComm((PetscObject)(mdn->A),&comm);
83: MatCreate(comm,&B);
84: MatSetSizes(B,m,m,m,m);
85: MatSetType(B,((PetscObject)mdn->A)->type_name);
86: MatDenseGetArray(mdn->A,&array);
87: MatSeqDenseSetPreallocation(B,array+m*rstart);
88: MatDenseRestoreArray(mdn->A,&array);
89: MatAssemblyBegin(B,MAT_FINAL_ASSEMBLY);
90: MatAssemblyEnd(B,MAT_FINAL_ASSEMBLY);
91: PetscObjectCompose((PetscObject)A,"DiagonalBlock",(PetscObject)B);
92: *a = B;
93: MatDestroy(&B);
94: } else *a = B;
95: return(0);
96: }
100: PetscErrorCode MatSetValues_MPIDense(Mat mat,PetscInt m,const PetscInt idxm[],PetscInt n,const PetscInt idxn[],const PetscScalar v[],InsertMode addv)
101: {
102: Mat_MPIDense *A = (Mat_MPIDense*)mat->data;
104: PetscInt i,j,rstart = mat->rmap->rstart,rend = mat->rmap->rend,row;
105: PetscBool roworiented = A->roworiented;
108: for (i=0; i<m; i++) {
109: if (idxm[i] < 0) continue;
110: if (idxm[i] >= mat->rmap->N) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Row too large");
111: if (idxm[i] >= rstart && idxm[i] < rend) {
112: row = idxm[i] - rstart;
113: if (roworiented) {
114: MatSetValues(A->A,1,&row,n,idxn,v+i*n,addv);
115: } else {
116: for (j=0; j<n; j++) {
117: if (idxn[j] < 0) continue;
118: if (idxn[j] >= mat->cmap->N) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Column too large");
119: MatSetValues(A->A,1,&row,1,&idxn[j],v+i+j*m,addv);
120: }
121: }
122: } else if (!A->donotstash) {
123: mat->assembled = PETSC_FALSE;
124: if (roworiented) {
125: MatStashValuesRow_Private(&mat->stash,idxm[i],n,idxn,v+i*n,PETSC_FALSE);
126: } else {
127: MatStashValuesCol_Private(&mat->stash,idxm[i],n,idxn,v+i,m,PETSC_FALSE);
128: }
129: }
130: }
131: return(0);
132: }
136: PetscErrorCode MatGetValues_MPIDense(Mat mat,PetscInt m,const PetscInt idxm[],PetscInt n,const PetscInt idxn[],PetscScalar v[])
137: {
138: Mat_MPIDense *mdn = (Mat_MPIDense*)mat->data;
140: PetscInt i,j,rstart = mat->rmap->rstart,rend = mat->rmap->rend,row;
143: for (i=0; i<m; i++) {
144: if (idxm[i] < 0) continue; /* SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Negative row"); */
145: if (idxm[i] >= mat->rmap->N) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Row too large");
146: if (idxm[i] >= rstart && idxm[i] < rend) {
147: row = idxm[i] - rstart;
148: for (j=0; j<n; j++) {
149: if (idxn[j] < 0) continue; /* SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Negative column"); */
150: if (idxn[j] >= mat->cmap->N) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Column too large");
151: MatGetValues(mdn->A,1,&row,1,&idxn[j],v+i*n+j);
152: }
153: } else SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Only local values currently supported");
154: }
155: return(0);
156: }
160: PetscErrorCode MatDenseGetArray_MPIDense(Mat A,PetscScalar *array[])
161: {
162: Mat_MPIDense *a = (Mat_MPIDense*)A->data;
166: MatDenseGetArray(a->A,array);
167: return(0);
168: }
172: static PetscErrorCode MatGetSubMatrix_MPIDense(Mat A,IS isrow,IS iscol,MatReuse scall,Mat *B)
173: {
174: Mat_MPIDense *mat = (Mat_MPIDense*)A->data,*newmatd;
175: Mat_SeqDense *lmat = (Mat_SeqDense*)mat->A->data;
177: PetscInt i,j,rstart,rend,nrows,ncols,Ncols,nlrows,nlcols;
178: const PetscInt *irow,*icol;
179: PetscScalar *av,*bv,*v = lmat->v;
180: Mat newmat;
181: IS iscol_local;
184: ISAllGather(iscol,&iscol_local);
185: ISGetIndices(isrow,&irow);
186: ISGetIndices(iscol_local,&icol);
187: ISGetLocalSize(isrow,&nrows);
188: ISGetLocalSize(iscol,&ncols);
189: ISGetSize(iscol,&Ncols); /* global number of columns, size of iscol_local */
191: /* No parallel redistribution currently supported! Should really check each index set
192: to comfirm that it is OK. ... Currently supports only submatrix same partitioning as
193: original matrix! */
195: MatGetLocalSize(A,&nlrows,&nlcols);
196: MatGetOwnershipRange(A,&rstart,&rend);
198: /* Check submatrix call */
199: if (scall == MAT_REUSE_MATRIX) {
200: /* SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Reused submatrix wrong size"); */
201: /* Really need to test rows and column sizes! */
202: newmat = *B;
203: } else {
204: /* Create and fill new matrix */
205: MatCreate(PetscObjectComm((PetscObject)A),&newmat);
206: MatSetSizes(newmat,nrows,ncols,PETSC_DECIDE,Ncols);
207: MatSetType(newmat,((PetscObject)A)->type_name);
208: MatMPIDenseSetPreallocation(newmat,NULL);
209: }
211: /* Now extract the data pointers and do the copy, column at a time */
212: newmatd = (Mat_MPIDense*)newmat->data;
213: bv = ((Mat_SeqDense*)newmatd->A->data)->v;
215: for (i=0; i<Ncols; i++) {
216: av = v + ((Mat_SeqDense*)mat->A->data)->lda*icol[i];
217: for (j=0; j<nrows; j++) {
218: *bv++ = av[irow[j] - rstart];
219: }
220: }
222: /* Assemble the matrices so that the correct flags are set */
223: MatAssemblyBegin(newmat,MAT_FINAL_ASSEMBLY);
224: MatAssemblyEnd(newmat,MAT_FINAL_ASSEMBLY);
226: /* Free work space */
227: ISRestoreIndices(isrow,&irow);
228: ISRestoreIndices(iscol_local,&icol);
229: ISDestroy(&iscol_local);
230: *B = newmat;
231: return(0);
232: }
236: PetscErrorCode MatDenseRestoreArray_MPIDense(Mat A,PetscScalar *array[])
237: {
238: Mat_MPIDense *a = (Mat_MPIDense*)A->data;
242: MatDenseRestoreArray(a->A,array);
243: return(0);
244: }
248: PetscErrorCode MatAssemblyBegin_MPIDense(Mat mat,MatAssemblyType mode)
249: {
250: Mat_MPIDense *mdn = (Mat_MPIDense*)mat->data;
251: MPI_Comm comm;
253: PetscInt nstash,reallocs;
254: InsertMode addv;
257: PetscObjectGetComm((PetscObject)mat,&comm);
258: /* make sure all processors are either in INSERTMODE or ADDMODE */
259: MPI_Allreduce((PetscEnum*)&mat->insertmode,(PetscEnum*)&addv,1,MPIU_ENUM,MPI_BOR,comm);
260: if (addv == (ADD_VALUES|INSERT_VALUES)) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_WRONGSTATE,"Cannot mix adds/inserts on different procs");
261: mat->insertmode = addv; /* in case this processor had no cache */
263: MatStashScatterBegin_Private(mat,&mat->stash,mat->rmap->range);
264: MatStashGetInfo_Private(&mat->stash,&nstash,&reallocs);
265: PetscInfo2(mdn->A,"Stash has %D entries, uses %D mallocs.\n",nstash,reallocs);
266: return(0);
267: }
271: PetscErrorCode MatAssemblyEnd_MPIDense(Mat mat,MatAssemblyType mode)
272: {
273: Mat_MPIDense *mdn=(Mat_MPIDense*)mat->data;
275: PetscInt i,*row,*col,flg,j,rstart,ncols;
276: PetscMPIInt n;
277: PetscScalar *val;
278: InsertMode addv=mat->insertmode;
281: /* wait on receives */
282: while (1) {
283: MatStashScatterGetMesg_Private(&mat->stash,&n,&row,&col,&val,&flg);
284: if (!flg) break;
286: for (i=0; i<n;) {
287: /* Now identify the consecutive vals belonging to the same row */
288: for (j=i,rstart=row[j]; j<n; j++) {
289: if (row[j] != rstart) break;
290: }
291: if (j < n) ncols = j-i;
292: else ncols = n-i;
293: /* Now assemble all these values with a single function call */
294: MatSetValues_MPIDense(mat,1,row+i,ncols,col+i,val+i,addv);
295: i = j;
296: }
297: }
298: MatStashScatterEnd_Private(&mat->stash);
300: MatAssemblyBegin(mdn->A,mode);
301: MatAssemblyEnd(mdn->A,mode);
303: if (!mat->was_assembled && mode == MAT_FINAL_ASSEMBLY) {
304: MatSetUpMultiply_MPIDense(mat);
305: }
306: return(0);
307: }
311: PetscErrorCode MatZeroEntries_MPIDense(Mat A)
312: {
314: Mat_MPIDense *l = (Mat_MPIDense*)A->data;
317: MatZeroEntries(l->A);
318: return(0);
319: }
321: /* the code does not do the diagonal entries correctly unless the
322: matrix is square and the column and row owerships are identical.
323: This is a BUG. The only way to fix it seems to be to access
324: mdn->A and mdn->B directly and not through the MatZeroRows()
325: routine.
326: */
329: PetscErrorCode MatZeroRows_MPIDense(Mat A,PetscInt N,const PetscInt rows[],PetscScalar diag,Vec x,Vec b)
330: {
331: Mat_MPIDense *l = (Mat_MPIDense*)A->data;
332: PetscErrorCode ierr;
333: PetscInt i,*owners = A->rmap->range;
334: PetscInt *sizes,j,idx,nsends;
335: PetscInt nmax,*svalues,*starts,*owner,nrecvs;
336: PetscInt *rvalues,tag = ((PetscObject)A)->tag,count,base,slen,*source;
337: PetscInt *lens,*lrows,*values;
338: PetscMPIInt n,imdex,rank = l->rank,size = l->size;
339: MPI_Comm comm;
340: MPI_Request *send_waits,*recv_waits;
341: MPI_Status recv_status,*send_status;
342: PetscBool found;
343: const PetscScalar *xx;
344: PetscScalar *bb;
347: PetscObjectGetComm((PetscObject)A,&comm);
348: if (A->rmap->N != A->cmap->N) SETERRQ(comm,PETSC_ERR_SUP,"Only handles square matrices");
349: if (A->rmap->n != A->cmap->n) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Only handles matrices with identical column and row ownership");
350: /* first count number of contributors to each processor */
351: PetscCalloc1(2*size,&sizes);
352: PetscMalloc1(N+1,&owner); /* see note*/
353: for (i=0; i<N; i++) {
354: idx = rows[i];
355: found = PETSC_FALSE;
356: for (j=0; j<size; j++) {
357: if (idx >= owners[j] && idx < owners[j+1]) {
358: sizes[2*j]++; sizes[2*j+1] = 1; owner[i] = j; found = PETSC_TRUE; break;
359: }
360: }
361: if (!found) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Index out of range");
362: }
363: nsends = 0;
364: for (i=0; i<size; i++) nsends += sizes[2*i+1];
366: /* inform other processors of number of messages and max length*/
367: PetscMaxSum(comm,sizes,&nmax,&nrecvs);
369: /* post receives: */
370: PetscMalloc1((nrecvs+1)*(nmax+1),&rvalues);
371: PetscMalloc1((nrecvs+1),&recv_waits);
372: for (i=0; i<nrecvs; i++) {
373: MPI_Irecv(rvalues+nmax*i,nmax,MPIU_INT,MPI_ANY_SOURCE,tag,comm,recv_waits+i);
374: }
376: /* do sends:
377: 1) starts[i] gives the starting index in svalues for stuff going to
378: the ith processor
379: */
380: PetscMalloc1((N+1),&svalues);
381: PetscMalloc1((nsends+1),&send_waits);
382: PetscMalloc1((size+1),&starts);
384: starts[0] = 0;
385: for (i=1; i<size; i++) starts[i] = starts[i-1] + sizes[2*i-2];
386: for (i=0; i<N; i++) svalues[starts[owner[i]]++] = rows[i];
388: starts[0] = 0;
389: for (i=1; i<size+1; i++) starts[i] = starts[i-1] + sizes[2*i-2];
390: count = 0;
391: for (i=0; i<size; i++) {
392: if (sizes[2*i+1]) {
393: MPI_Isend(svalues+starts[i],sizes[2*i],MPIU_INT,i,tag,comm,send_waits+count++);
394: }
395: }
396: PetscFree(starts);
398: base = owners[rank];
400: /* wait on receives */
401: PetscMalloc2(nrecvs,&lens,nrecvs,&source);
402: count = nrecvs;
403: slen = 0;
404: while (count) {
405: MPI_Waitany(nrecvs,recv_waits,&imdex,&recv_status);
406: /* unpack receives into our local space */
407: MPI_Get_count(&recv_status,MPIU_INT,&n);
409: source[imdex] = recv_status.MPI_SOURCE;
410: lens[imdex] = n;
411: slen += n;
412: count--;
413: }
414: PetscFree(recv_waits);
416: /* move the data into the send scatter */
417: PetscMalloc1((slen+1),&lrows);
418: count = 0;
419: for (i=0; i<nrecvs; i++) {
420: values = rvalues + i*nmax;
421: for (j=0; j<lens[i]; j++) {
422: lrows[count++] = values[j] - base;
423: }
424: }
425: PetscFree(rvalues);
426: PetscFree2(lens,source);
427: PetscFree(owner);
428: PetscFree(sizes);
430: /* fix right hand side if needed */
431: if (x && b) {
432: VecGetArrayRead(x,&xx);
433: VecGetArray(b,&bb);
434: for (i=0; i<slen; i++) {
435: bb[lrows[i]] = diag*xx[lrows[i]];
436: }
437: VecRestoreArrayRead(x,&xx);
438: VecRestoreArray(b,&bb);
439: }
441: /* actually zap the local rows */
442: MatZeroRows(l->A,slen,lrows,0.0,0,0);
443: if (diag != 0.0) {
444: Mat_SeqDense *ll = (Mat_SeqDense*)l->A->data;
445: PetscInt m = ll->lda, i;
447: for (i=0; i<slen; i++) {
448: ll->v[lrows[i] + m*(A->cmap->rstart + lrows[i])] = diag;
449: }
450: }
451: PetscFree(lrows);
453: /* wait on sends */
454: if (nsends) {
455: PetscMalloc1(nsends,&send_status);
456: MPI_Waitall(nsends,send_waits,send_status);
457: PetscFree(send_status);
458: }
459: PetscFree(send_waits);
460: PetscFree(svalues);
461: return(0);
462: }
466: PetscErrorCode MatMult_MPIDense(Mat mat,Vec xx,Vec yy)
467: {
468: Mat_MPIDense *mdn = (Mat_MPIDense*)mat->data;
472: VecScatterBegin(mdn->Mvctx,xx,mdn->lvec,INSERT_VALUES,SCATTER_FORWARD);
473: VecScatterEnd(mdn->Mvctx,xx,mdn->lvec,INSERT_VALUES,SCATTER_FORWARD);
474: MatMult_SeqDense(mdn->A,mdn->lvec,yy);
475: return(0);
476: }
480: PetscErrorCode MatMultAdd_MPIDense(Mat mat,Vec xx,Vec yy,Vec zz)
481: {
482: Mat_MPIDense *mdn = (Mat_MPIDense*)mat->data;
486: VecScatterBegin(mdn->Mvctx,xx,mdn->lvec,INSERT_VALUES,SCATTER_FORWARD);
487: VecScatterEnd(mdn->Mvctx,xx,mdn->lvec,INSERT_VALUES,SCATTER_FORWARD);
488: MatMultAdd_SeqDense(mdn->A,mdn->lvec,yy,zz);
489: return(0);
490: }
494: PetscErrorCode MatMultTranspose_MPIDense(Mat A,Vec xx,Vec yy)
495: {
496: Mat_MPIDense *a = (Mat_MPIDense*)A->data;
498: PetscScalar zero = 0.0;
501: VecSet(yy,zero);
502: MatMultTranspose_SeqDense(a->A,xx,a->lvec);
503: VecScatterBegin(a->Mvctx,a->lvec,yy,ADD_VALUES,SCATTER_REVERSE);
504: VecScatterEnd(a->Mvctx,a->lvec,yy,ADD_VALUES,SCATTER_REVERSE);
505: return(0);
506: }
510: PetscErrorCode MatMultTransposeAdd_MPIDense(Mat A,Vec xx,Vec yy,Vec zz)
511: {
512: Mat_MPIDense *a = (Mat_MPIDense*)A->data;
516: VecCopy(yy,zz);
517: MatMultTranspose_SeqDense(a->A,xx,a->lvec);
518: VecScatterBegin(a->Mvctx,a->lvec,zz,ADD_VALUES,SCATTER_REVERSE);
519: VecScatterEnd(a->Mvctx,a->lvec,zz,ADD_VALUES,SCATTER_REVERSE);
520: return(0);
521: }
525: PetscErrorCode MatGetDiagonal_MPIDense(Mat A,Vec v)
526: {
527: Mat_MPIDense *a = (Mat_MPIDense*)A->data;
528: Mat_SeqDense *aloc = (Mat_SeqDense*)a->A->data;
530: PetscInt len,i,n,m = A->rmap->n,radd;
531: PetscScalar *x,zero = 0.0;
534: VecSet(v,zero);
535: VecGetArray(v,&x);
536: VecGetSize(v,&n);
537: if (n != A->rmap->N) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Nonconforming mat and vec");
538: len = PetscMin(a->A->rmap->n,a->A->cmap->n);
539: radd = A->rmap->rstart*m;
540: for (i=0; i<len; i++) {
541: x[i] = aloc->v[radd + i*m + i];
542: }
543: VecRestoreArray(v,&x);
544: return(0);
545: }
549: PetscErrorCode MatDestroy_MPIDense(Mat mat)
550: {
551: Mat_MPIDense *mdn = (Mat_MPIDense*)mat->data;
555: #if defined(PETSC_USE_LOG)
556: PetscLogObjectState((PetscObject)mat,"Rows=%D, Cols=%D",mat->rmap->N,mat->cmap->N);
557: #endif
558: MatStashDestroy_Private(&mat->stash);
559: MatDestroy(&mdn->A);
560: VecDestroy(&mdn->lvec);
561: VecScatterDestroy(&mdn->Mvctx);
563: PetscFree(mat->data);
564: PetscObjectChangeTypeName((PetscObject)mat,0);
565: PetscObjectComposeFunction((PetscObject)mat,"MatGetDiagonalBlock_C",NULL);
566: PetscObjectComposeFunction((PetscObject)mat,"MatMPIDenseSetPreallocation_C",NULL);
567: PetscObjectComposeFunction((PetscObject)mat,"MatMatMult_mpiaij_mpidense_C",NULL);
568: PetscObjectComposeFunction((PetscObject)mat,"MatMatMultSymbolic_mpiaij_mpidense_C",NULL);
569: PetscObjectComposeFunction((PetscObject)mat,"MatMatMultNumeric_mpiaij_mpidense_C",NULL);
570: return(0);
571: }
575: static PetscErrorCode MatView_MPIDense_Binary(Mat mat,PetscViewer viewer)
576: {
577: Mat_MPIDense *mdn = (Mat_MPIDense*)mat->data;
578: PetscErrorCode ierr;
579: PetscViewerFormat format;
580: int fd;
581: PetscInt header[4],mmax,N = mat->cmap->N,i,j,m,k;
582: PetscMPIInt rank,tag = ((PetscObject)viewer)->tag,size;
583: PetscScalar *work,*v,*vv;
584: Mat_SeqDense *a = (Mat_SeqDense*)mdn->A->data;
587: if (mdn->size == 1) {
588: MatView(mdn->A,viewer);
589: } else {
590: PetscViewerBinaryGetDescriptor(viewer,&fd);
591: MPI_Comm_rank(PetscObjectComm((PetscObject)mat),&rank);
592: MPI_Comm_size(PetscObjectComm((PetscObject)mat),&size);
594: PetscViewerGetFormat(viewer,&format);
595: if (format == PETSC_VIEWER_NATIVE) {
597: if (!rank) {
598: /* store the matrix as a dense matrix */
599: header[0] = MAT_FILE_CLASSID;
600: header[1] = mat->rmap->N;
601: header[2] = N;
602: header[3] = MATRIX_BINARY_FORMAT_DENSE;
603: PetscBinaryWrite(fd,header,4,PETSC_INT,PETSC_TRUE);
605: /* get largest work array needed for transposing array */
606: mmax = mat->rmap->n;
607: for (i=1; i<size; i++) {
608: mmax = PetscMax(mmax,mat->rmap->range[i+1] - mat->rmap->range[i]);
609: }
610: PetscMalloc1(mmax*N,&work);
612: /* write out local array, by rows */
613: m = mat->rmap->n;
614: v = a->v;
615: for (j=0; j<N; j++) {
616: for (i=0; i<m; i++) {
617: work[j + i*N] = *v++;
618: }
619: }
620: PetscBinaryWrite(fd,work,m*N,PETSC_SCALAR,PETSC_FALSE);
621: /* get largest work array to receive messages from other processes, excludes process zero */
622: mmax = 0;
623: for (i=1; i<size; i++) {
624: mmax = PetscMax(mmax,mat->rmap->range[i+1] - mat->rmap->range[i]);
625: }
626: PetscMalloc1(mmax*N,&vv);
627: for (k = 1; k < size; k++) {
628: v = vv;
629: m = mat->rmap->range[k+1] - mat->rmap->range[k];
630: MPIULong_Recv(v,m*N,MPIU_SCALAR,k,tag,PetscObjectComm((PetscObject)mat));
632: for (j = 0; j < N; j++) {
633: for (i = 0; i < m; i++) {
634: work[j + i*N] = *v++;
635: }
636: }
637: PetscBinaryWrite(fd,work,m*N,PETSC_SCALAR,PETSC_FALSE);
638: }
639: PetscFree(work);
640: PetscFree(vv);
641: } else {
642: MPIULong_Send(a->v,mat->rmap->n*mat->cmap->N,MPIU_SCALAR,0,tag,PetscObjectComm((PetscObject)mat));
643: }
644: } else SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"To store a parallel dense matrix you must first call PetscViewerSetFormat(viewer,PETSC_VIEWER_NATIVE)");
645: }
646: return(0);
647: }
649: extern PetscErrorCode MatView_SeqDense(Mat,PetscViewer);
650: #include <petscdraw.h>
653: static PetscErrorCode MatView_MPIDense_ASCIIorDraworSocket(Mat mat,PetscViewer viewer)
654: {
655: Mat_MPIDense *mdn = (Mat_MPIDense*)mat->data;
656: PetscErrorCode ierr;
657: PetscMPIInt rank = mdn->rank;
658: PetscViewerType vtype;
659: PetscBool iascii,isdraw;
660: PetscViewer sviewer;
661: PetscViewerFormat format;
664: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERASCII,&iascii);
665: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERDRAW,&isdraw);
666: if (iascii) {
667: PetscViewerGetType(viewer,&vtype);
668: PetscViewerGetFormat(viewer,&format);
669: if (format == PETSC_VIEWER_ASCII_INFO_DETAIL) {
670: MatInfo info;
671: MatGetInfo(mat,MAT_LOCAL,&info);
672: PetscViewerASCIISynchronizedAllow(viewer,PETSC_TRUE);
673: PetscViewerASCIISynchronizedPrintf(viewer," [%d] local rows %D nz %D nz alloced %D mem %D \n",rank,mat->rmap->n,(PetscInt)info.nz_used,(PetscInt)info.nz_allocated,(PetscInt)info.memory);
674: PetscViewerFlush(viewer);
675: PetscViewerASCIISynchronizedAllow(viewer,PETSC_FALSE);
676: VecScatterView(mdn->Mvctx,viewer);
677: return(0);
678: } else if (format == PETSC_VIEWER_ASCII_INFO) {
679: return(0);
680: }
681: } else if (isdraw) {
682: PetscDraw draw;
683: PetscBool isnull;
685: PetscViewerDrawGetDraw(viewer,0,&draw);
686: PetscDrawIsNull(draw,&isnull);
687: if (isnull) return(0);
688: }
690: {
691: /* assemble the entire matrix onto first processor. */
692: Mat A;
693: PetscInt M = mat->rmap->N,N = mat->cmap->N,m,row,i,nz;
694: PetscInt *cols;
695: PetscScalar *vals;
697: MatCreate(PetscObjectComm((PetscObject)mat),&A);
698: if (!rank) {
699: MatSetSizes(A,M,N,M,N);
700: } else {
701: MatSetSizes(A,0,0,M,N);
702: }
703: /* Since this is a temporary matrix, MATMPIDENSE instead of ((PetscObject)A)->type_name here is probably acceptable. */
704: MatSetType(A,MATMPIDENSE);
705: MatMPIDenseSetPreallocation(A,NULL);
706: PetscLogObjectParent((PetscObject)mat,(PetscObject)A);
708: /* Copy the matrix ... This isn't the most efficient means,
709: but it's quick for now */
710: A->insertmode = INSERT_VALUES;
712: row = mat->rmap->rstart;
713: m = mdn->A->rmap->n;
714: for (i=0; i<m; i++) {
715: MatGetRow_MPIDense(mat,row,&nz,&cols,&vals);
716: MatSetValues_MPIDense(A,1,&row,nz,cols,vals,INSERT_VALUES);
717: MatRestoreRow_MPIDense(mat,row,&nz,&cols,&vals);
718: row++;
719: }
721: MatAssemblyBegin(A,MAT_FINAL_ASSEMBLY);
722: MatAssemblyEnd(A,MAT_FINAL_ASSEMBLY);
723: PetscViewerGetSingleton(viewer,&sviewer);
724: if (!rank) {
725: MatView_SeqDense(((Mat_MPIDense*)(A->data))->A,sviewer);
726: }
727: PetscViewerRestoreSingleton(viewer,&sviewer);
728: PetscViewerFlush(viewer);
729: MatDestroy(&A);
730: }
731: return(0);
732: }
736: PetscErrorCode MatView_MPIDense(Mat mat,PetscViewer viewer)
737: {
739: PetscBool iascii,isbinary,isdraw,issocket;
742: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERASCII,&iascii);
743: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERBINARY,&isbinary);
744: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERSOCKET,&issocket);
745: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERDRAW,&isdraw);
747: if (iascii || issocket || isdraw) {
748: MatView_MPIDense_ASCIIorDraworSocket(mat,viewer);
749: } else if (isbinary) {
750: MatView_MPIDense_Binary(mat,viewer);
751: }
752: return(0);
753: }
757: PetscErrorCode MatGetInfo_MPIDense(Mat A,MatInfoType flag,MatInfo *info)
758: {
759: Mat_MPIDense *mat = (Mat_MPIDense*)A->data;
760: Mat mdn = mat->A;
762: PetscReal isend[5],irecv[5];
765: info->block_size = 1.0;
767: MatGetInfo(mdn,MAT_LOCAL,info);
769: isend[0] = info->nz_used; isend[1] = info->nz_allocated; isend[2] = info->nz_unneeded;
770: isend[3] = info->memory; isend[4] = info->mallocs;
771: if (flag == MAT_LOCAL) {
772: info->nz_used = isend[0];
773: info->nz_allocated = isend[1];
774: info->nz_unneeded = isend[2];
775: info->memory = isend[3];
776: info->mallocs = isend[4];
777: } else if (flag == MAT_GLOBAL_MAX) {
778: MPI_Allreduce(isend,irecv,5,MPIU_REAL,MPIU_MAX,PetscObjectComm((PetscObject)A));
780: info->nz_used = irecv[0];
781: info->nz_allocated = irecv[1];
782: info->nz_unneeded = irecv[2];
783: info->memory = irecv[3];
784: info->mallocs = irecv[4];
785: } else if (flag == MAT_GLOBAL_SUM) {
786: MPI_Allreduce(isend,irecv,5,MPIU_REAL,MPIU_SUM,PetscObjectComm((PetscObject)A));
788: info->nz_used = irecv[0];
789: info->nz_allocated = irecv[1];
790: info->nz_unneeded = irecv[2];
791: info->memory = irecv[3];
792: info->mallocs = irecv[4];
793: }
794: info->fill_ratio_given = 0; /* no parallel LU/ILU/Cholesky */
795: info->fill_ratio_needed = 0;
796: info->factor_mallocs = 0;
797: return(0);
798: }
802: PetscErrorCode MatSetOption_MPIDense(Mat A,MatOption op,PetscBool flg)
803: {
804: Mat_MPIDense *a = (Mat_MPIDense*)A->data;
808: switch (op) {
809: case MAT_NEW_NONZERO_LOCATIONS:
810: case MAT_NEW_NONZERO_LOCATION_ERR:
811: case MAT_NEW_NONZERO_ALLOCATION_ERR:
812: MatSetOption(a->A,op,flg);
813: break;
814: case MAT_ROW_ORIENTED:
815: a->roworiented = flg;
817: MatSetOption(a->A,op,flg);
818: break;
819: case MAT_NEW_DIAGONALS:
820: case MAT_KEEP_NONZERO_PATTERN:
821: case MAT_USE_HASH_TABLE:
822: PetscInfo1(A,"Option %s ignored\n",MatOptions[op]);
823: break;
824: case MAT_IGNORE_OFF_PROC_ENTRIES:
825: a->donotstash = flg;
826: break;
827: case MAT_SYMMETRIC:
828: case MAT_STRUCTURALLY_SYMMETRIC:
829: case MAT_HERMITIAN:
830: case MAT_SYMMETRY_ETERNAL:
831: case MAT_IGNORE_LOWER_TRIANGULAR:
832: PetscInfo1(A,"Option %s ignored\n",MatOptions[op]);
833: break;
834: default:
835: SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"unknown option %s",MatOptions[op]);
836: }
837: return(0);
838: }
843: PetscErrorCode MatDiagonalScale_MPIDense(Mat A,Vec ll,Vec rr)
844: {
845: Mat_MPIDense *mdn = (Mat_MPIDense*)A->data;
846: Mat_SeqDense *mat = (Mat_SeqDense*)mdn->A->data;
847: PetscScalar *l,*r,x,*v;
849: PetscInt i,j,s2a,s3a,s2,s3,m=mdn->A->rmap->n,n=mdn->A->cmap->n;
852: MatGetLocalSize(A,&s2,&s3);
853: if (ll) {
854: VecGetLocalSize(ll,&s2a);
855: if (s2a != s2) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Left scaling vector non-conforming local size, %d != %d.", s2a, s2);
856: VecGetArray(ll,&l);
857: for (i=0; i<m; i++) {
858: x = l[i];
859: v = mat->v + i;
860: for (j=0; j<n; j++) { (*v) *= x; v+= m;}
861: }
862: VecRestoreArray(ll,&l);
863: PetscLogFlops(n*m);
864: }
865: if (rr) {
866: VecGetLocalSize(rr,&s3a);
867: if (s3a != s3) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Right scaling vec non-conforming local size, %d != %d.", s3a, s3);
868: VecScatterBegin(mdn->Mvctx,rr,mdn->lvec,INSERT_VALUES,SCATTER_FORWARD);
869: VecScatterEnd(mdn->Mvctx,rr,mdn->lvec,INSERT_VALUES,SCATTER_FORWARD);
870: VecGetArray(mdn->lvec,&r);
871: for (i=0; i<n; i++) {
872: x = r[i];
873: v = mat->v + i*m;
874: for (j=0; j<m; j++) (*v++) *= x;
875: }
876: VecRestoreArray(mdn->lvec,&r);
877: PetscLogFlops(n*m);
878: }
879: return(0);
880: }
884: PetscErrorCode MatNorm_MPIDense(Mat A,NormType type,PetscReal *nrm)
885: {
886: Mat_MPIDense *mdn = (Mat_MPIDense*)A->data;
887: Mat_SeqDense *mat = (Mat_SeqDense*)mdn->A->data;
889: PetscInt i,j;
890: PetscReal sum = 0.0;
891: PetscScalar *v = mat->v;
894: if (mdn->size == 1) {
895: MatNorm(mdn->A,type,nrm);
896: } else {
897: if (type == NORM_FROBENIUS) {
898: for (i=0; i<mdn->A->cmap->n*mdn->A->rmap->n; i++) {
899: sum += PetscRealPart(PetscConj(*v)*(*v)); v++;
900: }
901: MPI_Allreduce(&sum,nrm,1,MPIU_REAL,MPIU_SUM,PetscObjectComm((PetscObject)A));
902: *nrm = PetscSqrtReal(*nrm);
903: PetscLogFlops(2.0*mdn->A->cmap->n*mdn->A->rmap->n);
904: } else if (type == NORM_1) {
905: PetscReal *tmp,*tmp2;
906: PetscMalloc2(A->cmap->N,&tmp,A->cmap->N,&tmp2);
907: PetscMemzero(tmp,A->cmap->N*sizeof(PetscReal));
908: PetscMemzero(tmp2,A->cmap->N*sizeof(PetscReal));
909: *nrm = 0.0;
910: v = mat->v;
911: for (j=0; j<mdn->A->cmap->n; j++) {
912: for (i=0; i<mdn->A->rmap->n; i++) {
913: tmp[j] += PetscAbsScalar(*v); v++;
914: }
915: }
916: MPI_Allreduce(tmp,tmp2,A->cmap->N,MPIU_REAL,MPIU_SUM,PetscObjectComm((PetscObject)A));
917: for (j=0; j<A->cmap->N; j++) {
918: if (tmp2[j] > *nrm) *nrm = tmp2[j];
919: }
920: PetscFree2(tmp,tmp);
921: PetscLogFlops(A->cmap->n*A->rmap->n);
922: } else if (type == NORM_INFINITY) { /* max row norm */
923: PetscReal ntemp;
924: MatNorm(mdn->A,type,&ntemp);
925: MPI_Allreduce(&ntemp,nrm,1,MPIU_REAL,MPIU_MAX,PetscObjectComm((PetscObject)A));
926: } else SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_SUP,"No support for two norm");
927: }
928: return(0);
929: }
933: PetscErrorCode MatTranspose_MPIDense(Mat A,MatReuse reuse,Mat *matout)
934: {
935: Mat_MPIDense *a = (Mat_MPIDense*)A->data;
936: Mat_SeqDense *Aloc = (Mat_SeqDense*)a->A->data;
937: Mat B;
938: PetscInt M = A->rmap->N,N = A->cmap->N,m,n,*rwork,rstart = A->rmap->rstart;
940: PetscInt j,i;
941: PetscScalar *v;
944: if (reuse == MAT_REUSE_MATRIX && A == *matout && M != N) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_SUP,"Supports square matrix only in-place");
945: if (reuse == MAT_INITIAL_MATRIX || A == *matout) {
946: MatCreate(PetscObjectComm((PetscObject)A),&B);
947: MatSetSizes(B,A->cmap->n,A->rmap->n,N,M);
948: MatSetType(B,((PetscObject)A)->type_name);
949: MatMPIDenseSetPreallocation(B,NULL);
950: } else {
951: B = *matout;
952: }
954: m = a->A->rmap->n; n = a->A->cmap->n; v = Aloc->v;
955: PetscMalloc1(m,&rwork);
956: for (i=0; i<m; i++) rwork[i] = rstart + i;
957: for (j=0; j<n; j++) {
958: MatSetValues(B,1,&j,m,rwork,v,INSERT_VALUES);
959: v += m;
960: }
961: PetscFree(rwork);
962: MatAssemblyBegin(B,MAT_FINAL_ASSEMBLY);
963: MatAssemblyEnd(B,MAT_FINAL_ASSEMBLY);
964: if (reuse == MAT_INITIAL_MATRIX || *matout != A) {
965: *matout = B;
966: } else {
967: MatHeaderMerge(A,B);
968: }
969: return(0);
970: }
973: static PetscErrorCode MatDuplicate_MPIDense(Mat,MatDuplicateOption,Mat*);
974: extern PetscErrorCode MatScale_MPIDense(Mat,PetscScalar);
978: PetscErrorCode MatSetUp_MPIDense(Mat A)
979: {
983: MatMPIDenseSetPreallocation(A,0);
984: return(0);
985: }
989: PetscErrorCode MatAXPY_MPIDense(Mat Y,PetscScalar alpha,Mat X,MatStructure str)
990: {
992: Mat_MPIDense *A = (Mat_MPIDense*)Y->data, *B = (Mat_MPIDense*)X->data;
995: MatAXPY(A->A,alpha,B->A,str);
996: PetscObjectStateIncrease((PetscObject)Y);
997: return(0);
998: }
1002: PetscErrorCode MatConjugate_MPIDense(Mat mat)
1003: {
1004: Mat_MPIDense *a = (Mat_MPIDense*)mat->data;
1008: MatConjugate(a->A);
1009: return(0);
1010: }
1014: PetscErrorCode MatRealPart_MPIDense(Mat A)
1015: {
1016: Mat_MPIDense *a = (Mat_MPIDense*)A->data;
1020: MatRealPart(a->A);
1021: return(0);
1022: }
1026: PetscErrorCode MatImaginaryPart_MPIDense(Mat A)
1027: {
1028: Mat_MPIDense *a = (Mat_MPIDense*)A->data;
1032: MatImaginaryPart(a->A);
1033: return(0);
1034: }
1036: extern PetscErrorCode MatGetColumnNorms_SeqDense(Mat,NormType,PetscReal*);
1039: PetscErrorCode MatGetColumnNorms_MPIDense(Mat A,NormType type,PetscReal *norms)
1040: {
1042: PetscInt i,n;
1043: Mat_MPIDense *a = (Mat_MPIDense*) A->data;
1044: PetscReal *work;
1047: MatGetSize(A,NULL,&n);
1048: PetscMalloc1(n,&work);
1049: MatGetColumnNorms_SeqDense(a->A,type,work);
1050: if (type == NORM_2) {
1051: for (i=0; i<n; i++) work[i] *= work[i];
1052: }
1053: if (type == NORM_INFINITY) {
1054: MPI_Allreduce(work,norms,n,MPIU_REAL,MPIU_MAX,A->hdr.comm);
1055: } else {
1056: MPI_Allreduce(work,norms,n,MPIU_REAL,MPIU_SUM,A->hdr.comm);
1057: }
1058: PetscFree(work);
1059: if (type == NORM_2) {
1060: for (i=0; i<n; i++) norms[i] = PetscSqrtReal(norms[i]);
1061: }
1062: return(0);
1063: }
1067: static PetscErrorCode MatSetRandom_MPIDense(Mat x,PetscRandom rctx)
1068: {
1069: Mat_MPIDense *d = (Mat_MPIDense*)x->data;
1071: PetscScalar *a;
1072: PetscInt m,n,i;
1075: MatGetSize(d->A,&m,&n);
1076: MatDenseGetArray(d->A,&a);
1077: for (i=0; i<m*n; i++) {
1078: PetscRandomGetValue(rctx,a+i);
1079: }
1080: MatDenseRestoreArray(d->A,&a);
1081: return(0);
1082: }
1084: /* -------------------------------------------------------------------*/
1085: static struct _MatOps MatOps_Values = { MatSetValues_MPIDense,
1086: MatGetRow_MPIDense,
1087: MatRestoreRow_MPIDense,
1088: MatMult_MPIDense,
1089: /* 4*/ MatMultAdd_MPIDense,
1090: MatMultTranspose_MPIDense,
1091: MatMultTransposeAdd_MPIDense,
1092: 0,
1093: 0,
1094: 0,
1095: /* 10*/ 0,
1096: 0,
1097: 0,
1098: 0,
1099: MatTranspose_MPIDense,
1100: /* 15*/ MatGetInfo_MPIDense,
1101: MatEqual_MPIDense,
1102: MatGetDiagonal_MPIDense,
1103: MatDiagonalScale_MPIDense,
1104: MatNorm_MPIDense,
1105: /* 20*/ MatAssemblyBegin_MPIDense,
1106: MatAssemblyEnd_MPIDense,
1107: MatSetOption_MPIDense,
1108: MatZeroEntries_MPIDense,
1109: /* 24*/ MatZeroRows_MPIDense,
1110: 0,
1111: 0,
1112: 0,
1113: 0,
1114: /* 29*/ MatSetUp_MPIDense,
1115: 0,
1116: 0,
1117: 0,
1118: 0,
1119: /* 34*/ MatDuplicate_MPIDense,
1120: 0,
1121: 0,
1122: 0,
1123: 0,
1124: /* 39*/ MatAXPY_MPIDense,
1125: MatGetSubMatrices_MPIDense,
1126: 0,
1127: MatGetValues_MPIDense,
1128: 0,
1129: /* 44*/ 0,
1130: MatScale_MPIDense,
1131: 0,
1132: 0,
1133: 0,
1134: /* 49*/ MatSetRandom_MPIDense,
1135: 0,
1136: 0,
1137: 0,
1138: 0,
1139: /* 54*/ 0,
1140: 0,
1141: 0,
1142: 0,
1143: 0,
1144: /* 59*/ MatGetSubMatrix_MPIDense,
1145: MatDestroy_MPIDense,
1146: MatView_MPIDense,
1147: 0,
1148: 0,
1149: /* 64*/ 0,
1150: 0,
1151: 0,
1152: 0,
1153: 0,
1154: /* 69*/ 0,
1155: 0,
1156: 0,
1157: 0,
1158: 0,
1159: /* 74*/ 0,
1160: 0,
1161: 0,
1162: 0,
1163: 0,
1164: /* 79*/ 0,
1165: 0,
1166: 0,
1167: 0,
1168: /* 83*/ MatLoad_MPIDense,
1169: 0,
1170: 0,
1171: 0,
1172: 0,
1173: 0,
1174: /* 89*/
1175: 0,
1176: 0,
1177: 0,
1178: 0,
1179: 0,
1180: /* 94*/ 0,
1181: 0,
1182: 0,
1183: 0,
1184: 0,
1185: /* 99*/ 0,
1186: 0,
1187: 0,
1188: MatConjugate_MPIDense,
1189: 0,
1190: /*104*/ 0,
1191: MatRealPart_MPIDense,
1192: MatImaginaryPart_MPIDense,
1193: 0,
1194: 0,
1195: /*109*/ 0,
1196: 0,
1197: 0,
1198: 0,
1199: 0,
1200: /*114*/ 0,
1201: 0,
1202: 0,
1203: 0,
1204: 0,
1205: /*119*/ 0,
1206: 0,
1207: 0,
1208: 0,
1209: 0,
1210: /*124*/ 0,
1211: MatGetColumnNorms_MPIDense,
1212: 0,
1213: 0,
1214: 0,
1215: /*129*/ 0,
1216: 0,
1217: 0,
1218: 0,
1219: 0,
1220: /*134*/ 0,
1221: 0,
1222: 0,
1223: 0,
1224: 0,
1225: /*139*/ 0,
1226: 0,
1227: 0
1228: };
1232: PetscErrorCode MatMPIDenseSetPreallocation_MPIDense(Mat mat,PetscScalar *data)
1233: {
1234: Mat_MPIDense *a;
1238: mat->preallocated = PETSC_TRUE;
1239: /* Note: For now, when data is specified above, this assumes the user correctly
1240: allocates the local dense storage space. We should add error checking. */
1242: a = (Mat_MPIDense*)mat->data;
1243: PetscLayoutSetUp(mat->rmap);
1244: PetscLayoutSetUp(mat->cmap);
1245: a->nvec = mat->cmap->n;
1247: MatCreate(PETSC_COMM_SELF,&a->A);
1248: MatSetSizes(a->A,mat->rmap->n,mat->cmap->N,mat->rmap->n,mat->cmap->N);
1249: MatSetType(a->A,MATSEQDENSE);
1250: MatSeqDenseSetPreallocation(a->A,data);
1251: PetscLogObjectParent((PetscObject)mat,(PetscObject)a->A);
1252: return(0);
1253: }
1257: PETSC_EXTERN PetscErrorCode MatCreate_MPIDense(Mat mat)
1258: {
1259: Mat_MPIDense *a;
1263: PetscNewLog(mat,&a);
1264: mat->data = (void*)a;
1265: PetscMemcpy(mat->ops,&MatOps_Values,sizeof(struct _MatOps));
1267: mat->insertmode = NOT_SET_VALUES;
1268: MPI_Comm_rank(PetscObjectComm((PetscObject)mat),&a->rank);
1269: MPI_Comm_size(PetscObjectComm((PetscObject)mat),&a->size);
1271: /* build cache for off array entries formed */
1272: a->donotstash = PETSC_FALSE;
1274: MatStashCreate_Private(PetscObjectComm((PetscObject)mat),1,&mat->stash);
1276: /* stuff used for matrix vector multiply */
1277: a->lvec = 0;
1278: a->Mvctx = 0;
1279: a->roworiented = PETSC_TRUE;
1281: PetscObjectComposeFunction((PetscObject)mat,"MatDenseGetArray_C",MatDenseGetArray_MPIDense);
1282: PetscObjectComposeFunction((PetscObject)mat,"MatDenseRestoreArray_C",MatDenseRestoreArray_MPIDense);
1284: PetscObjectComposeFunction((PetscObject)mat,"MatGetDiagonalBlock_C",MatGetDiagonalBlock_MPIDense);
1285: PetscObjectComposeFunction((PetscObject)mat,"MatMPIDenseSetPreallocation_C",MatMPIDenseSetPreallocation_MPIDense);
1286: PetscObjectComposeFunction((PetscObject)mat,"MatMatMult_mpiaij_mpidense_C",MatMatMult_MPIAIJ_MPIDense);
1287: PetscObjectComposeFunction((PetscObject)mat,"MatMatMultSymbolic_mpiaij_mpidense_C",MatMatMultSymbolic_MPIAIJ_MPIDense);
1288: PetscObjectComposeFunction((PetscObject)mat,"MatMatMultNumeric_mpiaij_mpidense_C",MatMatMultNumeric_MPIAIJ_MPIDense);
1290: PetscObjectComposeFunction((PetscObject)mat,"MatTransposeMatMult_mpiaij_mpidense_C",MatTransposeMatMult_MPIAIJ_MPIDense);
1291: PetscObjectComposeFunction((PetscObject)mat,"MatTransposeMatMultSymbolic_mpiaij_mpidense_C",MatTransposeMatMultSymbolic_MPIAIJ_MPIDense);
1292: PetscObjectComposeFunction((PetscObject)mat,"MatTransposeMatMultNumeric_mpiaij_mpidense_C",MatTransposeMatMultNumeric_MPIAIJ_MPIDense);
1293: PetscObjectChangeTypeName((PetscObject)mat,MATMPIDENSE);
1294: return(0);
1295: }
1297: /*MC
1298: MATDENSE - MATDENSE = "dense" - A matrix type to be used for dense matrices.
1300: This matrix type is identical to MATSEQDENSE when constructed with a single process communicator,
1301: and MATMPIDENSE otherwise.
1303: Options Database Keys:
1304: . -mat_type dense - sets the matrix type to "dense" during a call to MatSetFromOptions()
1306: Level: beginner
1309: .seealso: MatCreateMPIDense,MATSEQDENSE,MATMPIDENSE
1310: M*/
1314: /*@C
1315: MatMPIDenseSetPreallocation - Sets the array used to store the matrix entries
1317: Not collective
1319: Input Parameters:
1320: . A - the matrix
1321: - data - optional location of matrix data. Set data=NULL for PETSc
1322: to control all matrix memory allocation.
1324: Notes:
1325: The dense format is fully compatible with standard Fortran 77
1326: storage by columns.
1328: The data input variable is intended primarily for Fortran programmers
1329: who wish to allocate their own matrix memory space. Most users should
1330: set data=NULL.
1332: Level: intermediate
1334: .keywords: matrix,dense, parallel
1336: .seealso: MatCreate(), MatCreateSeqDense(), MatSetValues()
1337: @*/
1338: PetscErrorCode MatMPIDenseSetPreallocation(Mat mat,PetscScalar *data)
1339: {
1343: PetscTryMethod(mat,"MatMPIDenseSetPreallocation_C",(Mat,PetscScalar*),(mat,data));
1344: return(0);
1345: }
1349: /*@C
1350: MatCreateDense - Creates a parallel matrix in dense format.
1352: Collective on MPI_Comm
1354: Input Parameters:
1355: + comm - MPI communicator
1356: . m - number of local rows (or PETSC_DECIDE to have calculated if M is given)
1357: . n - number of local columns (or PETSC_DECIDE to have calculated if N is given)
1358: . M - number of global rows (or PETSC_DECIDE to have calculated if m is given)
1359: . N - number of global columns (or PETSC_DECIDE to have calculated if n is given)
1360: - data - optional location of matrix data. Set data=NULL (NULL_SCALAR for Fortran users) for PETSc
1361: to control all matrix memory allocation.
1363: Output Parameter:
1364: . A - the matrix
1366: Notes:
1367: The dense format is fully compatible with standard Fortran 77
1368: storage by columns.
1370: The data input variable is intended primarily for Fortran programmers
1371: who wish to allocate their own matrix memory space. Most users should
1372: set data=NULL (NULL_SCALAR for Fortran users).
1374: The user MUST specify either the local or global matrix dimensions
1375: (possibly both).
1377: Level: intermediate
1379: .keywords: matrix,dense, parallel
1381: .seealso: MatCreate(), MatCreateSeqDense(), MatSetValues()
1382: @*/
1383: PetscErrorCode MatCreateDense(MPI_Comm comm,PetscInt m,PetscInt n,PetscInt M,PetscInt N,PetscScalar *data,Mat *A)
1384: {
1386: PetscMPIInt size;
1389: MatCreate(comm,A);
1390: MatSetSizes(*A,m,n,M,N);
1391: MPI_Comm_size(comm,&size);
1392: if (size > 1) {
1393: MatSetType(*A,MATMPIDENSE);
1394: MatMPIDenseSetPreallocation(*A,data);
1395: } else {
1396: MatSetType(*A,MATSEQDENSE);
1397: MatSeqDenseSetPreallocation(*A,data);
1398: }
1399: return(0);
1400: }
1404: static PetscErrorCode MatDuplicate_MPIDense(Mat A,MatDuplicateOption cpvalues,Mat *newmat)
1405: {
1406: Mat mat;
1407: Mat_MPIDense *a,*oldmat = (Mat_MPIDense*)A->data;
1411: *newmat = 0;
1412: MatCreate(PetscObjectComm((PetscObject)A),&mat);
1413: MatSetSizes(mat,A->rmap->n,A->cmap->n,A->rmap->N,A->cmap->N);
1414: MatSetType(mat,((PetscObject)A)->type_name);
1415: a = (Mat_MPIDense*)mat->data;
1416: PetscMemcpy(mat->ops,A->ops,sizeof(struct _MatOps));
1418: mat->factortype = A->factortype;
1419: mat->assembled = PETSC_TRUE;
1420: mat->preallocated = PETSC_TRUE;
1422: a->size = oldmat->size;
1423: a->rank = oldmat->rank;
1424: mat->insertmode = NOT_SET_VALUES;
1425: a->nvec = oldmat->nvec;
1426: a->donotstash = oldmat->donotstash;
1428: PetscLayoutReference(A->rmap,&mat->rmap);
1429: PetscLayoutReference(A->cmap,&mat->cmap);
1431: MatSetUpMultiply_MPIDense(mat);
1432: MatDuplicate(oldmat->A,cpvalues,&a->A);
1433: PetscLogObjectParent((PetscObject)mat,(PetscObject)a->A);
1435: *newmat = mat;
1436: return(0);
1437: }
1441: PetscErrorCode MatLoad_MPIDense_DenseInFile(MPI_Comm comm,PetscInt fd,PetscInt M,PetscInt N,Mat newmat,PetscInt sizesset)
1442: {
1444: PetscMPIInt rank,size;
1445: PetscInt *rowners,i,m,nz,j;
1446: PetscScalar *array,*vals,*vals_ptr;
1449: MPI_Comm_rank(comm,&rank);
1450: MPI_Comm_size(comm,&size);
1452: /* determine ownership of all rows */
1453: if (newmat->rmap->n < 0) m = M/size + ((M % size) > rank);
1454: else m = newmat->rmap->n;
1455: PetscMalloc1((size+2),&rowners);
1456: MPI_Allgather(&m,1,MPIU_INT,rowners+1,1,MPIU_INT,comm);
1457: rowners[0] = 0;
1458: for (i=2; i<=size; i++) {
1459: rowners[i] += rowners[i-1];
1460: }
1462: if (!sizesset) {
1463: MatSetSizes(newmat,m,PETSC_DECIDE,M,N);
1464: }
1465: MatMPIDenseSetPreallocation(newmat,NULL);
1466: MatDenseGetArray(newmat,&array);
1468: if (!rank) {
1469: PetscMalloc1(m*N,&vals);
1471: /* read in my part of the matrix numerical values */
1472: PetscBinaryRead(fd,vals,m*N,PETSC_SCALAR);
1474: /* insert into matrix-by row (this is why cannot directly read into array */
1475: vals_ptr = vals;
1476: for (i=0; i<m; i++) {
1477: for (j=0; j<N; j++) {
1478: array[i + j*m] = *vals_ptr++;
1479: }
1480: }
1482: /* read in other processors and ship out */
1483: for (i=1; i<size; i++) {
1484: nz = (rowners[i+1] - rowners[i])*N;
1485: PetscBinaryRead(fd,vals,nz,PETSC_SCALAR);
1486: MPIULong_Send(vals,nz,MPIU_SCALAR,i,((PetscObject)(newmat))->tag,comm);
1487: }
1488: } else {
1489: /* receive numeric values */
1490: PetscMalloc1(m*N,&vals);
1492: /* receive message of values*/
1493: MPIULong_Recv(vals,m*N,MPIU_SCALAR,0,((PetscObject)(newmat))->tag,comm);
1495: /* insert into matrix-by row (this is why cannot directly read into array */
1496: vals_ptr = vals;
1497: for (i=0; i<m; i++) {
1498: for (j=0; j<N; j++) {
1499: array[i + j*m] = *vals_ptr++;
1500: }
1501: }
1502: }
1503: MatDenseRestoreArray(newmat,&array);
1504: PetscFree(rowners);
1505: PetscFree(vals);
1506: MatAssemblyBegin(newmat,MAT_FINAL_ASSEMBLY);
1507: MatAssemblyEnd(newmat,MAT_FINAL_ASSEMBLY);
1508: return(0);
1509: }
1513: PetscErrorCode MatLoad_MPIDense(Mat newmat,PetscViewer viewer)
1514: {
1515: PetscScalar *vals,*svals;
1516: MPI_Comm comm;
1517: MPI_Status status;
1518: PetscMPIInt rank,size,tag = ((PetscObject)viewer)->tag,*rowners,*sndcounts,m,maxnz;
1519: PetscInt header[4],*rowlengths = 0,M,N,*cols;
1520: PetscInt *ourlens,*procsnz = 0,*offlens,jj,*mycols,*smycols;
1521: PetscInt i,nz,j,rstart,rend,sizesset=1,grows,gcols;
1522: int fd;
1526: PetscObjectGetComm((PetscObject)viewer,&comm);
1527: MPI_Comm_size(comm,&size);
1528: MPI_Comm_rank(comm,&rank);
1529: if (!rank) {
1530: PetscViewerBinaryGetDescriptor(viewer,&fd);
1531: PetscBinaryRead(fd,(char*)header,4,PETSC_INT);
1532: if (header[0] != MAT_FILE_CLASSID) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_FILE_UNEXPECTED,"not matrix object");
1533: }
1534: if (newmat->rmap->n < 0 && newmat->rmap->N < 0 && newmat->cmap->n < 0 && newmat->cmap->N < 0) sizesset = 0;
1536: MPI_Bcast(header+1,3,MPIU_INT,0,comm);
1537: M = header[1]; N = header[2]; nz = header[3];
1539: /* If global rows/cols are set to PETSC_DECIDE, set it to the sizes given in the file */
1540: if (sizesset && newmat->rmap->N < 0) newmat->rmap->N = M;
1541: if (sizesset && newmat->cmap->N < 0) newmat->cmap->N = N;
1543: /* If global sizes are set, check if they are consistent with that given in the file */
1544: if (sizesset) {
1545: MatGetSize(newmat,&grows,&gcols);
1546: }
1547: if (sizesset && newmat->rmap->N != grows) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_FILE_UNEXPECTED, "Inconsistent # of rows:Matrix in file has (%d) and input matrix has (%d)",M,grows);
1548: if (sizesset && newmat->cmap->N != gcols) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_FILE_UNEXPECTED, "Inconsistent # of cols:Matrix in file has (%d) and input matrix has (%d)",N,gcols);
1550: /*
1551: Handle case where matrix is stored on disk as a dense matrix
1552: */
1553: if (nz == MATRIX_BINARY_FORMAT_DENSE) {
1554: MatLoad_MPIDense_DenseInFile(comm,fd,M,N,newmat,sizesset);
1555: return(0);
1556: }
1558: /* determine ownership of all rows */
1559: if (newmat->rmap->n < 0) {
1560: PetscMPIIntCast(M/size + ((M % size) > rank),&m);
1561: } else {
1562: PetscMPIIntCast(newmat->rmap->n,&m);
1563: }
1564: PetscMalloc1((size+2),&rowners);
1565: MPI_Allgather(&m,1,MPI_INT,rowners+1,1,MPI_INT,comm);
1566: rowners[0] = 0;
1567: for (i=2; i<=size; i++) {
1568: rowners[i] += rowners[i-1];
1569: }
1570: rstart = rowners[rank];
1571: rend = rowners[rank+1];
1573: /* distribute row lengths to all processors */
1574: PetscMalloc2(rend-rstart,&ourlens,rend-rstart,&offlens);
1575: if (!rank) {
1576: PetscMalloc1(M,&rowlengths);
1577: PetscBinaryRead(fd,rowlengths,M,PETSC_INT);
1578: PetscMalloc1(size,&sndcounts);
1579: for (i=0; i<size; i++) sndcounts[i] = rowners[i+1] - rowners[i];
1580: MPI_Scatterv(rowlengths,sndcounts,rowners,MPIU_INT,ourlens,rend-rstart,MPIU_INT,0,comm);
1581: PetscFree(sndcounts);
1582: } else {
1583: MPI_Scatterv(0,0,0,MPIU_INT,ourlens,rend-rstart,MPIU_INT,0,comm);
1584: }
1586: if (!rank) {
1587: /* calculate the number of nonzeros on each processor */
1588: PetscMalloc1(size,&procsnz);
1589: PetscMemzero(procsnz,size*sizeof(PetscInt));
1590: for (i=0; i<size; i++) {
1591: for (j=rowners[i]; j< rowners[i+1]; j++) {
1592: procsnz[i] += rowlengths[j];
1593: }
1594: }
1595: PetscFree(rowlengths);
1597: /* determine max buffer needed and allocate it */
1598: maxnz = 0;
1599: for (i=0; i<size; i++) {
1600: maxnz = PetscMax(maxnz,procsnz[i]);
1601: }
1602: PetscMalloc1(maxnz,&cols);
1604: /* read in my part of the matrix column indices */
1605: nz = procsnz[0];
1606: PetscMalloc1(nz,&mycols);
1607: PetscBinaryRead(fd,mycols,nz,PETSC_INT);
1609: /* read in every one elses and ship off */
1610: for (i=1; i<size; i++) {
1611: nz = procsnz[i];
1612: PetscBinaryRead(fd,cols,nz,PETSC_INT);
1613: MPI_Send(cols,nz,MPIU_INT,i,tag,comm);
1614: }
1615: PetscFree(cols);
1616: } else {
1617: /* determine buffer space needed for message */
1618: nz = 0;
1619: for (i=0; i<m; i++) {
1620: nz += ourlens[i];
1621: }
1622: PetscMalloc1((nz+1),&mycols);
1624: /* receive message of column indices*/
1625: MPI_Recv(mycols,nz,MPIU_INT,0,tag,comm,&status);
1626: MPI_Get_count(&status,MPIU_INT,&maxnz);
1627: if (maxnz != nz) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_FILE_UNEXPECTED,"something is wrong with file");
1628: }
1630: /* loop over local rows, determining number of off diagonal entries */
1631: PetscMemzero(offlens,m*sizeof(PetscInt));
1632: jj = 0;
1633: for (i=0; i<m; i++) {
1634: for (j=0; j<ourlens[i]; j++) {
1635: if (mycols[jj] < rstart || mycols[jj] >= rend) offlens[i]++;
1636: jj++;
1637: }
1638: }
1640: /* create our matrix */
1641: for (i=0; i<m; i++) ourlens[i] -= offlens[i];
1643: if (!sizesset) {
1644: MatSetSizes(newmat,m,PETSC_DECIDE,M,N);
1645: }
1646: MatMPIDenseSetPreallocation(newmat,NULL);
1647: for (i=0; i<m; i++) ourlens[i] += offlens[i];
1649: if (!rank) {
1650: PetscMalloc1(maxnz,&vals);
1652: /* read in my part of the matrix numerical values */
1653: nz = procsnz[0];
1654: PetscBinaryRead(fd,vals,nz,PETSC_SCALAR);
1656: /* insert into matrix */
1657: jj = rstart;
1658: smycols = mycols;
1659: svals = vals;
1660: for (i=0; i<m; i++) {
1661: MatSetValues(newmat,1,&jj,ourlens[i],smycols,svals,INSERT_VALUES);
1662: smycols += ourlens[i];
1663: svals += ourlens[i];
1664: jj++;
1665: }
1667: /* read in other processors and ship out */
1668: for (i=1; i<size; i++) {
1669: nz = procsnz[i];
1670: PetscBinaryRead(fd,vals,nz,PETSC_SCALAR);
1671: MPI_Send(vals,nz,MPIU_SCALAR,i,((PetscObject)newmat)->tag,comm);
1672: }
1673: PetscFree(procsnz);
1674: } else {
1675: /* receive numeric values */
1676: PetscMalloc1((nz+1),&vals);
1678: /* receive message of values*/
1679: MPI_Recv(vals,nz,MPIU_SCALAR,0,((PetscObject)newmat)->tag,comm,&status);
1680: MPI_Get_count(&status,MPIU_SCALAR,&maxnz);
1681: if (maxnz != nz) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_FILE_UNEXPECTED,"something is wrong with file");
1683: /* insert into matrix */
1684: jj = rstart;
1685: smycols = mycols;
1686: svals = vals;
1687: for (i=0; i<m; i++) {
1688: MatSetValues(newmat,1,&jj,ourlens[i],smycols,svals,INSERT_VALUES);
1689: smycols += ourlens[i];
1690: svals += ourlens[i];
1691: jj++;
1692: }
1693: }
1694: PetscFree2(ourlens,offlens);
1695: PetscFree(vals);
1696: PetscFree(mycols);
1697: PetscFree(rowners);
1699: MatAssemblyBegin(newmat,MAT_FINAL_ASSEMBLY);
1700: MatAssemblyEnd(newmat,MAT_FINAL_ASSEMBLY);
1701: return(0);
1702: }
1706: PetscErrorCode MatEqual_MPIDense(Mat A,Mat B,PetscBool *flag)
1707: {
1708: Mat_MPIDense *matB = (Mat_MPIDense*)B->data,*matA = (Mat_MPIDense*)A->data;
1709: Mat a,b;
1710: PetscBool flg;
1714: a = matA->A;
1715: b = matB->A;
1716: MatEqual(a,b,&flg);
1717: MPI_Allreduce(&flg,flag,1,MPIU_BOOL,MPI_LAND,PetscObjectComm((PetscObject)A));
1718: return(0);
1719: }