Actual source code: mpiaij.c
petsc-3.11.1 2019-04-12
3: #include <../src/mat/impls/aij/mpi/mpiaij.h>
4: #include <petsc/private/vecimpl.h>
5: #include <petsc/private/vecscatterimpl.h>
6: #include <petsc/private/isimpl.h>
7: #include <petscblaslapack.h>
8: #include <petscsf.h>
10: /*MC
11: MATAIJ - MATAIJ = "aij" - A matrix type to be used for sparse matrices.
13: This matrix type is identical to MATSEQAIJ when constructed with a single process communicator,
14: and MATMPIAIJ otherwise. As a result, for single process communicators,
15: MatSeqAIJSetPreallocation is supported, and similarly MatMPIAIJSetPreallocation is supported
16: for communicators controlling multiple processes. It is recommended that you call both of
17: the above preallocation routines for simplicity.
19: Options Database Keys:
20: . -mat_type aij - sets the matrix type to "aij" during a call to MatSetFromOptions()
22: Developer Notes:
23: Subclasses include MATAIJCUSP, MATAIJCUSPARSE, MATAIJPERM, MATAIJSELL, MATAIJMKL, MATAIJCRL, and also automatically switches over to use inodes when
24: enough exist.
26: Level: beginner
28: .seealso: MatCreateAIJ(), MatCreateSeqAIJ(), MATSEQAIJ, MATMPIAIJ
29: M*/
31: /*MC
32: MATAIJCRL - MATAIJCRL = "aijcrl" - A matrix type to be used for sparse matrices.
34: This matrix type is identical to MATSEQAIJCRL when constructed with a single process communicator,
35: and MATMPIAIJCRL otherwise. As a result, for single process communicators,
36: MatSeqAIJSetPreallocation() is supported, and similarly MatMPIAIJSetPreallocation() is supported
37: for communicators controlling multiple processes. It is recommended that you call both of
38: the above preallocation routines for simplicity.
40: Options Database Keys:
41: . -mat_type aijcrl - sets the matrix type to "aijcrl" during a call to MatSetFromOptions()
43: Level: beginner
45: .seealso: MatCreateMPIAIJCRL,MATSEQAIJCRL,MATMPIAIJCRL, MATSEQAIJCRL, MATMPIAIJCRL
46: M*/
48: PetscErrorCode MatSetBlockSizes_MPIAIJ(Mat M, PetscInt rbs, PetscInt cbs)
49: {
51: Mat_MPIAIJ *mat = (Mat_MPIAIJ*)M->data;
54: if (mat->A) {
55: MatSetBlockSizes(mat->A,rbs,cbs);
56: MatSetBlockSizes(mat->B,rbs,1);
57: }
58: return(0);
59: }
61: PetscErrorCode MatFindNonzeroRows_MPIAIJ(Mat M,IS *keptrows)
62: {
63: PetscErrorCode ierr;
64: Mat_MPIAIJ *mat = (Mat_MPIAIJ*)M->data;
65: Mat_SeqAIJ *a = (Mat_SeqAIJ*)mat->A->data;
66: Mat_SeqAIJ *b = (Mat_SeqAIJ*)mat->B->data;
67: const PetscInt *ia,*ib;
68: const MatScalar *aa,*bb;
69: PetscInt na,nb,i,j,*rows,cnt=0,n0rows;
70: PetscInt m = M->rmap->n,rstart = M->rmap->rstart;
73: *keptrows = 0;
74: ia = a->i;
75: ib = b->i;
76: for (i=0; i<m; i++) {
77: na = ia[i+1] - ia[i];
78: nb = ib[i+1] - ib[i];
79: if (!na && !nb) {
80: cnt++;
81: goto ok1;
82: }
83: aa = a->a + ia[i];
84: for (j=0; j<na; j++) {
85: if (aa[j] != 0.0) goto ok1;
86: }
87: bb = b->a + ib[i];
88: for (j=0; j <nb; j++) {
89: if (bb[j] != 0.0) goto ok1;
90: }
91: cnt++;
92: ok1:;
93: }
94: MPIU_Allreduce(&cnt,&n0rows,1,MPIU_INT,MPI_SUM,PetscObjectComm((PetscObject)M));
95: if (!n0rows) return(0);
96: PetscMalloc1(M->rmap->n-cnt,&rows);
97: cnt = 0;
98: for (i=0; i<m; i++) {
99: na = ia[i+1] - ia[i];
100: nb = ib[i+1] - ib[i];
101: if (!na && !nb) continue;
102: aa = a->a + ia[i];
103: for (j=0; j<na;j++) {
104: if (aa[j] != 0.0) {
105: rows[cnt++] = rstart + i;
106: goto ok2;
107: }
108: }
109: bb = b->a + ib[i];
110: for (j=0; j<nb; j++) {
111: if (bb[j] != 0.0) {
112: rows[cnt++] = rstart + i;
113: goto ok2;
114: }
115: }
116: ok2:;
117: }
118: ISCreateGeneral(PetscObjectComm((PetscObject)M),cnt,rows,PETSC_OWN_POINTER,keptrows);
119: return(0);
120: }
122: PetscErrorCode MatDiagonalSet_MPIAIJ(Mat Y,Vec D,InsertMode is)
123: {
124: PetscErrorCode ierr;
125: Mat_MPIAIJ *aij = (Mat_MPIAIJ*) Y->data;
126: PetscBool cong;
129: MatHasCongruentLayouts(Y,&cong);
130: if (Y->assembled && cong) {
131: MatDiagonalSet(aij->A,D,is);
132: } else {
133: MatDiagonalSet_Default(Y,D,is);
134: }
135: return(0);
136: }
138: PetscErrorCode MatFindZeroDiagonals_MPIAIJ(Mat M,IS *zrows)
139: {
140: Mat_MPIAIJ *aij = (Mat_MPIAIJ*)M->data;
142: PetscInt i,rstart,nrows,*rows;
145: *zrows = NULL;
146: MatFindZeroDiagonals_SeqAIJ_Private(aij->A,&nrows,&rows);
147: MatGetOwnershipRange(M,&rstart,NULL);
148: for (i=0; i<nrows; i++) rows[i] += rstart;
149: ISCreateGeneral(PetscObjectComm((PetscObject)M),nrows,rows,PETSC_OWN_POINTER,zrows);
150: return(0);
151: }
153: PetscErrorCode MatGetColumnNorms_MPIAIJ(Mat A,NormType type,PetscReal *norms)
154: {
156: Mat_MPIAIJ *aij = (Mat_MPIAIJ*)A->data;
157: PetscInt i,n,*garray = aij->garray;
158: Mat_SeqAIJ *a_aij = (Mat_SeqAIJ*) aij->A->data;
159: Mat_SeqAIJ *b_aij = (Mat_SeqAIJ*) aij->B->data;
160: PetscReal *work;
163: MatGetSize(A,NULL,&n);
164: PetscCalloc1(n,&work);
165: if (type == NORM_2) {
166: for (i=0; i<a_aij->i[aij->A->rmap->n]; i++) {
167: work[A->cmap->rstart + a_aij->j[i]] += PetscAbsScalar(a_aij->a[i]*a_aij->a[i]);
168: }
169: for (i=0; i<b_aij->i[aij->B->rmap->n]; i++) {
170: work[garray[b_aij->j[i]]] += PetscAbsScalar(b_aij->a[i]*b_aij->a[i]);
171: }
172: } else if (type == NORM_1) {
173: for (i=0; i<a_aij->i[aij->A->rmap->n]; i++) {
174: work[A->cmap->rstart + a_aij->j[i]] += PetscAbsScalar(a_aij->a[i]);
175: }
176: for (i=0; i<b_aij->i[aij->B->rmap->n]; i++) {
177: work[garray[b_aij->j[i]]] += PetscAbsScalar(b_aij->a[i]);
178: }
179: } else if (type == NORM_INFINITY) {
180: for (i=0; i<a_aij->i[aij->A->rmap->n]; i++) {
181: work[A->cmap->rstart + a_aij->j[i]] = PetscMax(PetscAbsScalar(a_aij->a[i]), work[A->cmap->rstart + a_aij->j[i]]);
182: }
183: for (i=0; i<b_aij->i[aij->B->rmap->n]; i++) {
184: work[garray[b_aij->j[i]]] = PetscMax(PetscAbsScalar(b_aij->a[i]),work[garray[b_aij->j[i]]]);
185: }
187: } else SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_WRONG,"Unknown NormType");
188: if (type == NORM_INFINITY) {
189: MPIU_Allreduce(work,norms,n,MPIU_REAL,MPIU_MAX,PetscObjectComm((PetscObject)A));
190: } else {
191: MPIU_Allreduce(work,norms,n,MPIU_REAL,MPIU_SUM,PetscObjectComm((PetscObject)A));
192: }
193: PetscFree(work);
194: if (type == NORM_2) {
195: for (i=0; i<n; i++) norms[i] = PetscSqrtReal(norms[i]);
196: }
197: return(0);
198: }
200: PetscErrorCode MatFindOffBlockDiagonalEntries_MPIAIJ(Mat A,IS *is)
201: {
202: Mat_MPIAIJ *a = (Mat_MPIAIJ*)A->data;
203: IS sis,gis;
204: PetscErrorCode ierr;
205: const PetscInt *isis,*igis;
206: PetscInt n,*iis,nsis,ngis,rstart,i;
209: MatFindOffBlockDiagonalEntries(a->A,&sis);
210: MatFindNonzeroRows(a->B,&gis);
211: ISGetSize(gis,&ngis);
212: ISGetSize(sis,&nsis);
213: ISGetIndices(sis,&isis);
214: ISGetIndices(gis,&igis);
216: PetscMalloc1(ngis+nsis,&iis);
217: PetscMemcpy(iis,igis,ngis*sizeof(PetscInt));
218: PetscMemcpy(iis+ngis,isis,nsis*sizeof(PetscInt));
219: n = ngis + nsis;
220: PetscSortRemoveDupsInt(&n,iis);
221: MatGetOwnershipRange(A,&rstart,NULL);
222: for (i=0; i<n; i++) iis[i] += rstart;
223: ISCreateGeneral(PetscObjectComm((PetscObject)A),n,iis,PETSC_OWN_POINTER,is);
225: ISRestoreIndices(sis,&isis);
226: ISRestoreIndices(gis,&igis);
227: ISDestroy(&sis);
228: ISDestroy(&gis);
229: return(0);
230: }
232: /*
233: Distributes a SeqAIJ matrix across a set of processes. Code stolen from
234: MatLoad_MPIAIJ(). Horrible lack of reuse. Should be a routine for each matrix type.
236: Only for square matrices
238: Used by a preconditioner, hence PETSC_EXTERN
239: */
240: PETSC_EXTERN PetscErrorCode MatDistribute_MPIAIJ(MPI_Comm comm,Mat gmat,PetscInt m,MatReuse reuse,Mat *inmat)
241: {
242: PetscMPIInt rank,size;
243: PetscInt *rowners,*dlens,*olens,i,rstart,rend,j,jj,nz = 0,*gmataj,cnt,row,*ld,bses[2];
245: Mat mat;
246: Mat_SeqAIJ *gmata;
247: PetscMPIInt tag;
248: MPI_Status status;
249: PetscBool aij;
250: MatScalar *gmataa,*ao,*ad,*gmataarestore=0;
253: MPI_Comm_rank(comm,&rank);
254: MPI_Comm_size(comm,&size);
255: if (!rank) {
256: PetscObjectTypeCompare((PetscObject)gmat,MATSEQAIJ,&aij);
257: if (!aij) SETERRQ1(PetscObjectComm((PetscObject)gmat),PETSC_ERR_SUP,"Currently no support for input matrix of type %s\n",((PetscObject)gmat)->type_name);
258: }
259: if (reuse == MAT_INITIAL_MATRIX) {
260: MatCreate(comm,&mat);
261: MatSetSizes(mat,m,m,PETSC_DETERMINE,PETSC_DETERMINE);
262: MatGetBlockSizes(gmat,&bses[0],&bses[1]);
263: MPI_Bcast(bses,2,MPIU_INT,0,comm);
264: MatSetBlockSizes(mat,bses[0],bses[1]);
265: MatSetType(mat,MATAIJ);
266: PetscMalloc1(size+1,&rowners);
267: PetscMalloc2(m,&dlens,m,&olens);
268: MPI_Allgather(&m,1,MPIU_INT,rowners+1,1,MPIU_INT,comm);
270: rowners[0] = 0;
271: for (i=2; i<=size; i++) rowners[i] += rowners[i-1];
272: rstart = rowners[rank];
273: rend = rowners[rank+1];
274: PetscObjectGetNewTag((PetscObject)mat,&tag);
275: if (!rank) {
276: gmata = (Mat_SeqAIJ*) gmat->data;
277: /* send row lengths to all processors */
278: for (i=0; i<m; i++) dlens[i] = gmata->ilen[i];
279: for (i=1; i<size; i++) {
280: MPI_Send(gmata->ilen + rowners[i],rowners[i+1]-rowners[i],MPIU_INT,i,tag,comm);
281: }
282: /* determine number diagonal and off-diagonal counts */
283: PetscMemzero(olens,m*sizeof(PetscInt));
284: PetscCalloc1(m,&ld);
285: jj = 0;
286: for (i=0; i<m; i++) {
287: for (j=0; j<dlens[i]; j++) {
288: if (gmata->j[jj] < rstart) ld[i]++;
289: if (gmata->j[jj] < rstart || gmata->j[jj] >= rend) olens[i]++;
290: jj++;
291: }
292: }
293: /* send column indices to other processes */
294: for (i=1; i<size; i++) {
295: nz = gmata->i[rowners[i+1]]-gmata->i[rowners[i]];
296: MPI_Send(&nz,1,MPIU_INT,i,tag,comm);
297: MPI_Send(gmata->j + gmata->i[rowners[i]],nz,MPIU_INT,i,tag,comm);
298: }
300: /* send numerical values to other processes */
301: for (i=1; i<size; i++) {
302: nz = gmata->i[rowners[i+1]]-gmata->i[rowners[i]];
303: MPI_Send(gmata->a + gmata->i[rowners[i]],nz,MPIU_SCALAR,i,tag,comm);
304: }
305: gmataa = gmata->a;
306: gmataj = gmata->j;
308: } else {
309: /* receive row lengths */
310: MPI_Recv(dlens,m,MPIU_INT,0,tag,comm,&status);
311: /* receive column indices */
312: MPI_Recv(&nz,1,MPIU_INT,0,tag,comm,&status);
313: PetscMalloc2(nz,&gmataa,nz,&gmataj);
314: MPI_Recv(gmataj,nz,MPIU_INT,0,tag,comm,&status);
315: /* determine number diagonal and off-diagonal counts */
316: PetscMemzero(olens,m*sizeof(PetscInt));
317: PetscCalloc1(m,&ld);
318: jj = 0;
319: for (i=0; i<m; i++) {
320: for (j=0; j<dlens[i]; j++) {
321: if (gmataj[jj] < rstart) ld[i]++;
322: if (gmataj[jj] < rstart || gmataj[jj] >= rend) olens[i]++;
323: jj++;
324: }
325: }
326: /* receive numerical values */
327: PetscMemzero(gmataa,nz*sizeof(PetscScalar));
328: MPI_Recv(gmataa,nz,MPIU_SCALAR,0,tag,comm,&status);
329: }
330: /* set preallocation */
331: for (i=0; i<m; i++) {
332: dlens[i] -= olens[i];
333: }
334: MatSeqAIJSetPreallocation(mat,0,dlens);
335: MatMPIAIJSetPreallocation(mat,0,dlens,0,olens);
337: for (i=0; i<m; i++) {
338: dlens[i] += olens[i];
339: }
340: cnt = 0;
341: for (i=0; i<m; i++) {
342: row = rstart + i;
343: MatSetValues(mat,1,&row,dlens[i],gmataj+cnt,gmataa+cnt,INSERT_VALUES);
344: cnt += dlens[i];
345: }
346: if (rank) {
347: PetscFree2(gmataa,gmataj);
348: }
349: PetscFree2(dlens,olens);
350: PetscFree(rowners);
352: ((Mat_MPIAIJ*)(mat->data))->ld = ld;
354: *inmat = mat;
355: } else { /* column indices are already set; only need to move over numerical values from process 0 */
356: Mat_SeqAIJ *Ad = (Mat_SeqAIJ*)((Mat_MPIAIJ*)((*inmat)->data))->A->data;
357: Mat_SeqAIJ *Ao = (Mat_SeqAIJ*)((Mat_MPIAIJ*)((*inmat)->data))->B->data;
358: mat = *inmat;
359: PetscObjectGetNewTag((PetscObject)mat,&tag);
360: if (!rank) {
361: /* send numerical values to other processes */
362: gmata = (Mat_SeqAIJ*) gmat->data;
363: MatGetOwnershipRanges(mat,(const PetscInt**)&rowners);
364: gmataa = gmata->a;
365: for (i=1; i<size; i++) {
366: nz = gmata->i[rowners[i+1]]-gmata->i[rowners[i]];
367: MPI_Send(gmataa + gmata->i[rowners[i]],nz,MPIU_SCALAR,i,tag,comm);
368: }
369: nz = gmata->i[rowners[1]]-gmata->i[rowners[0]];
370: } else {
371: /* receive numerical values from process 0*/
372: nz = Ad->nz + Ao->nz;
373: PetscMalloc1(nz,&gmataa); gmataarestore = gmataa;
374: MPI_Recv(gmataa,nz,MPIU_SCALAR,0,tag,comm,&status);
375: }
376: /* transfer numerical values into the diagonal A and off diagonal B parts of mat */
377: ld = ((Mat_MPIAIJ*)(mat->data))->ld;
378: ad = Ad->a;
379: ao = Ao->a;
380: if (mat->rmap->n) {
381: i = 0;
382: nz = ld[i]; PetscMemcpy(ao,gmataa,nz*sizeof(PetscScalar)); ao += nz; gmataa += nz;
383: nz = Ad->i[i+1] - Ad->i[i]; PetscMemcpy(ad,gmataa,nz*sizeof(PetscScalar)); ad += nz; gmataa += nz;
384: }
385: for (i=1; i<mat->rmap->n; i++) {
386: nz = Ao->i[i] - Ao->i[i-1] - ld[i-1] + ld[i]; PetscMemcpy(ao,gmataa,nz*sizeof(PetscScalar)); ao += nz; gmataa += nz;
387: nz = Ad->i[i+1] - Ad->i[i]; PetscMemcpy(ad,gmataa,nz*sizeof(PetscScalar)); ad += nz; gmataa += nz;
388: }
389: i--;
390: if (mat->rmap->n) {
391: nz = Ao->i[i+1] - Ao->i[i] - ld[i]; PetscMemcpy(ao,gmataa,nz*sizeof(PetscScalar));
392: }
393: if (rank) {
394: PetscFree(gmataarestore);
395: }
396: }
397: MatAssemblyBegin(mat,MAT_FINAL_ASSEMBLY);
398: MatAssemblyEnd(mat,MAT_FINAL_ASSEMBLY);
399: return(0);
400: }
402: /*
403: Local utility routine that creates a mapping from the global column
404: number to the local number in the off-diagonal part of the local
405: storage of the matrix. When PETSC_USE_CTABLE is used this is scalable at
406: a slightly higher hash table cost; without it it is not scalable (each processor
407: has an order N integer array but is fast to acess.
408: */
409: PetscErrorCode MatCreateColmap_MPIAIJ_Private(Mat mat)
410: {
411: Mat_MPIAIJ *aij = (Mat_MPIAIJ*)mat->data;
413: PetscInt n = aij->B->cmap->n,i;
416: if (!aij->garray) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_PLIB,"MPIAIJ Matrix was assembled but is missing garray");
417: #if defined(PETSC_USE_CTABLE)
418: PetscTableCreate(n,mat->cmap->N+1,&aij->colmap);
419: for (i=0; i<n; i++) {
420: PetscTableAdd(aij->colmap,aij->garray[i]+1,i+1,INSERT_VALUES);
421: }
422: #else
423: PetscCalloc1(mat->cmap->N+1,&aij->colmap);
424: PetscLogObjectMemory((PetscObject)mat,(mat->cmap->N+1)*sizeof(PetscInt));
425: for (i=0; i<n; i++) aij->colmap[aij->garray[i]] = i+1;
426: #endif
427: return(0);
428: }
430: #define MatSetValues_SeqAIJ_A_Private(row,col,value,addv,orow,ocol) \
431: { \
432: if (col <= lastcol1) low1 = 0; \
433: else high1 = nrow1; \
434: lastcol1 = col;\
435: while (high1-low1 > 5) { \
436: t = (low1+high1)/2; \
437: if (rp1[t] > col) high1 = t; \
438: else low1 = t; \
439: } \
440: for (_i=low1; _i<high1; _i++) { \
441: if (rp1[_i] > col) break; \
442: if (rp1[_i] == col) { \
443: if (addv == ADD_VALUES) ap1[_i] += value; \
444: else ap1[_i] = value; \
445: goto a_noinsert; \
446: } \
447: } \
448: if (value == 0.0 && ignorezeroentries && row != col) {low1 = 0; high1 = nrow1;goto a_noinsert;} \
449: if (nonew == 1) {low1 = 0; high1 = nrow1; goto a_noinsert;} \
450: if (nonew == -1) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Inserting a new nonzero at global row/column (%D, %D) into matrix", orow, ocol); \
451: MatSeqXAIJReallocateAIJ(A,am,1,nrow1,row,col,rmax1,aa,ai,aj,rp1,ap1,aimax,nonew,MatScalar); \
452: N = nrow1++ - 1; a->nz++; high1++; \
453: /* shift up all the later entries in this row */ \
454: for (ii=N; ii>=_i; ii--) { \
455: rp1[ii+1] = rp1[ii]; \
456: ap1[ii+1] = ap1[ii]; \
457: } \
458: rp1[_i] = col; \
459: ap1[_i] = value; \
460: A->nonzerostate++;\
461: a_noinsert: ; \
462: ailen[row] = nrow1; \
463: }
465: #define MatSetValues_SeqAIJ_B_Private(row,col,value,addv,orow,ocol) \
466: { \
467: if (col <= lastcol2) low2 = 0; \
468: else high2 = nrow2; \
469: lastcol2 = col; \
470: while (high2-low2 > 5) { \
471: t = (low2+high2)/2; \
472: if (rp2[t] > col) high2 = t; \
473: else low2 = t; \
474: } \
475: for (_i=low2; _i<high2; _i++) { \
476: if (rp2[_i] > col) break; \
477: if (rp2[_i] == col) { \
478: if (addv == ADD_VALUES) ap2[_i] += value; \
479: else ap2[_i] = value; \
480: goto b_noinsert; \
481: } \
482: } \
483: if (value == 0.0 && ignorezeroentries) {low2 = 0; high2 = nrow2; goto b_noinsert;} \
484: if (nonew == 1) {low2 = 0; high2 = nrow2; goto b_noinsert;} \
485: if (nonew == -1) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Inserting a new nonzero at global row/column (%D, %D) into matrix", orow, ocol); \
486: MatSeqXAIJReallocateAIJ(B,bm,1,nrow2,row,col,rmax2,ba,bi,bj,rp2,ap2,bimax,nonew,MatScalar); \
487: N = nrow2++ - 1; b->nz++; high2++; \
488: /* shift up all the later entries in this row */ \
489: for (ii=N; ii>=_i; ii--) { \
490: rp2[ii+1] = rp2[ii]; \
491: ap2[ii+1] = ap2[ii]; \
492: } \
493: rp2[_i] = col; \
494: ap2[_i] = value; \
495: B->nonzerostate++; \
496: b_noinsert: ; \
497: bilen[row] = nrow2; \
498: }
500: PetscErrorCode MatSetValuesRow_MPIAIJ(Mat A,PetscInt row,const PetscScalar v[])
501: {
502: Mat_MPIAIJ *mat = (Mat_MPIAIJ*)A->data;
503: Mat_SeqAIJ *a = (Mat_SeqAIJ*)mat->A->data,*b = (Mat_SeqAIJ*)mat->B->data;
505: PetscInt l,*garray = mat->garray,diag;
508: /* code only works for square matrices A */
510: /* find size of row to the left of the diagonal part */
511: MatGetOwnershipRange(A,&diag,0);
512: row = row - diag;
513: for (l=0; l<b->i[row+1]-b->i[row]; l++) {
514: if (garray[b->j[b->i[row]+l]] > diag) break;
515: }
516: PetscMemcpy(b->a+b->i[row],v,l*sizeof(PetscScalar));
518: /* diagonal part */
519: PetscMemcpy(a->a+a->i[row],v+l,(a->i[row+1]-a->i[row])*sizeof(PetscScalar));
521: /* right of diagonal part */
522: PetscMemcpy(b->a+b->i[row]+l,v+l+a->i[row+1]-a->i[row],(b->i[row+1]-b->i[row]-l)*sizeof(PetscScalar));
523: return(0);
524: }
526: PetscErrorCode MatSetValues_MPIAIJ(Mat mat,PetscInt m,const PetscInt im[],PetscInt n,const PetscInt in[],const PetscScalar v[],InsertMode addv)
527: {
528: Mat_MPIAIJ *aij = (Mat_MPIAIJ*)mat->data;
529: PetscScalar value;
531: PetscInt i,j,rstart = mat->rmap->rstart,rend = mat->rmap->rend;
532: PetscInt cstart = mat->cmap->rstart,cend = mat->cmap->rend,row,col;
533: PetscBool roworiented = aij->roworiented;
535: /* Some Variables required in the macro */
536: Mat A = aij->A;
537: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
538: PetscInt *aimax = a->imax,*ai = a->i,*ailen = a->ilen,*aj = a->j;
539: MatScalar *aa = a->a;
540: PetscBool ignorezeroentries = a->ignorezeroentries;
541: Mat B = aij->B;
542: Mat_SeqAIJ *b = (Mat_SeqAIJ*)B->data;
543: PetscInt *bimax = b->imax,*bi = b->i,*bilen = b->ilen,*bj = b->j,bm = aij->B->rmap->n,am = aij->A->rmap->n;
544: MatScalar *ba = b->a;
546: PetscInt *rp1,*rp2,ii,nrow1,nrow2,_i,rmax1,rmax2,N,low1,high1,low2,high2,t,lastcol1,lastcol2;
547: PetscInt nonew;
548: MatScalar *ap1,*ap2;
551: for (i=0; i<m; i++) {
552: if (im[i] < 0) continue;
553: #if defined(PETSC_USE_DEBUG)
554: if (im[i] >= mat->rmap->N) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Row too large: row %D max %D",im[i],mat->rmap->N-1);
555: #endif
556: if (im[i] >= rstart && im[i] < rend) {
557: row = im[i] - rstart;
558: lastcol1 = -1;
559: rp1 = aj + ai[row];
560: ap1 = aa + ai[row];
561: rmax1 = aimax[row];
562: nrow1 = ailen[row];
563: low1 = 0;
564: high1 = nrow1;
565: lastcol2 = -1;
566: rp2 = bj + bi[row];
567: ap2 = ba + bi[row];
568: rmax2 = bimax[row];
569: nrow2 = bilen[row];
570: low2 = 0;
571: high2 = nrow2;
573: for (j=0; j<n; j++) {
574: if (roworiented) value = v[i*n+j];
575: else value = v[i+j*m];
576: if (in[j] >= cstart && in[j] < cend) {
577: col = in[j] - cstart;
578: nonew = a->nonew;
579: if (ignorezeroentries && value == 0.0 && (addv == ADD_VALUES) && row != col) continue;
580: MatSetValues_SeqAIJ_A_Private(row,col,value,addv,im[i],in[j]);
581: } else if (in[j] < 0) continue;
582: #if defined(PETSC_USE_DEBUG)
583: else if (in[j] >= mat->cmap->N) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Column too large: col %D max %D",in[j],mat->cmap->N-1);
584: #endif
585: else {
586: if (mat->was_assembled) {
587: if (!aij->colmap) {
588: MatCreateColmap_MPIAIJ_Private(mat);
589: }
590: #if defined(PETSC_USE_CTABLE)
591: PetscTableFind(aij->colmap,in[j]+1,&col);
592: col--;
593: #else
594: col = aij->colmap[in[j]] - 1;
595: #endif
596: if (col < 0 && !((Mat_SeqAIJ*)(aij->B->data))->nonew) {
597: MatDisAssemble_MPIAIJ(mat);
598: col = in[j];
599: /* Reinitialize the variables required by MatSetValues_SeqAIJ_B_Private() */
600: B = aij->B;
601: b = (Mat_SeqAIJ*)B->data;
602: bimax = b->imax; bi = b->i; bilen = b->ilen; bj = b->j; ba = b->a;
603: rp2 = bj + bi[row];
604: ap2 = ba + bi[row];
605: rmax2 = bimax[row];
606: nrow2 = bilen[row];
607: low2 = 0;
608: high2 = nrow2;
609: bm = aij->B->rmap->n;
610: ba = b->a;
611: } else if (col < 0) {
612: if (1 == ((Mat_SeqAIJ*)(aij->B->data))->nonew) {
613: PetscInfo3(mat,"Skipping of insertion of new nonzero location in off-diagonal portion of matrix %g(%D,%D)\n",(double)PetscRealPart(value),im[i],in[j]);
614: } else SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Inserting a new nonzero at global row/column (%D, %D) into matrix", im[i], in[j]);
615: }
616: } else col = in[j];
617: nonew = b->nonew;
618: MatSetValues_SeqAIJ_B_Private(row,col,value,addv,im[i],in[j]);
619: }
620: }
621: } else {
622: if (mat->nooffprocentries) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"Setting off process row %D even though MatSetOption(,MAT_NO_OFF_PROC_ENTRIES,PETSC_TRUE) was set",im[i]);
623: if (!aij->donotstash) {
624: mat->assembled = PETSC_FALSE;
625: if (roworiented) {
626: MatStashValuesRow_Private(&mat->stash,im[i],n,in,v+i*n,(PetscBool)(ignorezeroentries && (addv == ADD_VALUES)));
627: } else {
628: MatStashValuesCol_Private(&mat->stash,im[i],n,in,v+i,m,(PetscBool)(ignorezeroentries && (addv == ADD_VALUES)));
629: }
630: }
631: }
632: }
633: return(0);
634: }
636: /*
637: This function sets the j and ilen arrays (of the diagonal and off-diagonal part) of an MPIAIJ-matrix.
638: The values in mat_i have to be sorted and the values in mat_j have to be sorted for each row (CSR-like).
639: No off-processor parts off the matrix are allowed here and mat->was_assembled has to be PETSC_FALSE.
640: */
641: PetscErrorCode MatSetValues_MPIAIJ_CopyFromCSRFormat_Symbolic(Mat mat,const PetscInt mat_j[],const PetscInt mat_i[])
642: {
643: Mat_MPIAIJ *aij = (Mat_MPIAIJ*)mat->data;
644: Mat A = aij->A; /* diagonal part of the matrix */
645: Mat B = aij->B; /* offdiagonal part of the matrix */
646: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
647: Mat_SeqAIJ *b = (Mat_SeqAIJ*)B->data;
648: PetscInt cstart = mat->cmap->rstart,cend = mat->cmap->rend,col;
649: PetscInt *ailen = a->ilen,*aj = a->j;
650: PetscInt *bilen = b->ilen,*bj = b->j;
651: PetscInt am = aij->A->rmap->n,j;
652: PetscInt diag_so_far = 0,dnz;
653: PetscInt offd_so_far = 0,onz;
656: /* Iterate over all rows of the matrix */
657: for (j=0; j<am; j++) {
658: dnz = onz = 0;
659: /* Iterate over all non-zero columns of the current row */
660: for (col=mat_i[j]; col<mat_i[j+1]; col++) {
661: /* If column is in the diagonal */
662: if (mat_j[col] >= cstart && mat_j[col] < cend) {
663: aj[diag_so_far++] = mat_j[col] - cstart;
664: dnz++;
665: } else { /* off-diagonal entries */
666: bj[offd_so_far++] = mat_j[col];
667: onz++;
668: }
669: }
670: ailen[j] = dnz;
671: bilen[j] = onz;
672: }
673: return(0);
674: }
676: /*
677: This function sets the local j, a and ilen arrays (of the diagonal and off-diagonal part) of an MPIAIJ-matrix.
678: The values in mat_i have to be sorted and the values in mat_j have to be sorted for each row (CSR-like).
679: No off-processor parts off the matrix are allowed here, they are set at a later point by MatSetValues_MPIAIJ.
680: Also, mat->was_assembled has to be false, otherwise the statement aj[rowstart_diag+dnz_row] = mat_j[col] - cstart;
681: would not be true and the more complex MatSetValues_MPIAIJ has to be used.
682: */
683: PetscErrorCode MatSetValues_MPIAIJ_CopyFromCSRFormat(Mat mat,const PetscInt mat_j[],const PetscInt mat_i[],const PetscScalar mat_a[])
684: {
685: Mat_MPIAIJ *aij = (Mat_MPIAIJ*)mat->data;
686: Mat A = aij->A; /* diagonal part of the matrix */
687: Mat B = aij->B; /* offdiagonal part of the matrix */
688: Mat_SeqAIJ *aijd =(Mat_SeqAIJ*)(aij->A)->data,*aijo=(Mat_SeqAIJ*)(aij->B)->data;
689: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
690: Mat_SeqAIJ *b = (Mat_SeqAIJ*)B->data;
691: PetscInt cstart = mat->cmap->rstart,cend = mat->cmap->rend;
692: PetscInt *ailen = a->ilen,*aj = a->j;
693: PetscInt *bilen = b->ilen,*bj = b->j;
694: PetscInt am = aij->A->rmap->n,j;
695: PetscInt *full_diag_i=aijd->i,*full_offd_i=aijo->i; /* These variables can also include non-local elements, which are set at a later point. */
696: PetscInt col,dnz_row,onz_row,rowstart_diag,rowstart_offd;
697: PetscScalar *aa = a->a,*ba = b->a;
700: /* Iterate over all rows of the matrix */
701: for (j=0; j<am; j++) {
702: dnz_row = onz_row = 0;
703: rowstart_offd = full_offd_i[j];
704: rowstart_diag = full_diag_i[j];
705: /* Iterate over all non-zero columns of the current row */
706: for (col=mat_i[j]; col<mat_i[j+1]; col++) {
707: /* If column is in the diagonal */
708: if (mat_j[col] >= cstart && mat_j[col] < cend) {
709: aj[rowstart_diag+dnz_row] = mat_j[col] - cstart;
710: aa[rowstart_diag+dnz_row] = mat_a[col];
711: dnz_row++;
712: } else { /* off-diagonal entries */
713: bj[rowstart_offd+onz_row] = mat_j[col];
714: ba[rowstart_offd+onz_row] = mat_a[col];
715: onz_row++;
716: }
717: }
718: ailen[j] = dnz_row;
719: bilen[j] = onz_row;
720: }
721: return(0);
722: }
724: PetscErrorCode MatGetValues_MPIAIJ(Mat mat,PetscInt m,const PetscInt idxm[],PetscInt n,const PetscInt idxn[],PetscScalar v[])
725: {
726: Mat_MPIAIJ *aij = (Mat_MPIAIJ*)mat->data;
728: PetscInt i,j,rstart = mat->rmap->rstart,rend = mat->rmap->rend;
729: PetscInt cstart = mat->cmap->rstart,cend = mat->cmap->rend,row,col;
732: for (i=0; i<m; i++) {
733: if (idxm[i] < 0) continue; /* SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Negative row: %D",idxm[i]);*/
734: if (idxm[i] >= mat->rmap->N) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Row too large: row %D max %D",idxm[i],mat->rmap->N-1);
735: if (idxm[i] >= rstart && idxm[i] < rend) {
736: row = idxm[i] - rstart;
737: for (j=0; j<n; j++) {
738: if (idxn[j] < 0) continue; /* SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Negative column: %D",idxn[j]); */
739: if (idxn[j] >= mat->cmap->N) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Column too large: col %D max %D",idxn[j],mat->cmap->N-1);
740: if (idxn[j] >= cstart && idxn[j] < cend) {
741: col = idxn[j] - cstart;
742: MatGetValues(aij->A,1,&row,1,&col,v+i*n+j);
743: } else {
744: if (!aij->colmap) {
745: MatCreateColmap_MPIAIJ_Private(mat);
746: }
747: #if defined(PETSC_USE_CTABLE)
748: PetscTableFind(aij->colmap,idxn[j]+1,&col);
749: col--;
750: #else
751: col = aij->colmap[idxn[j]] - 1;
752: #endif
753: if ((col < 0) || (aij->garray[col] != idxn[j])) *(v+i*n+j) = 0.0;
754: else {
755: MatGetValues(aij->B,1,&row,1,&col,v+i*n+j);
756: }
757: }
758: }
759: } else SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Only local values currently supported");
760: }
761: return(0);
762: }
764: extern PetscErrorCode MatMultDiagonalBlock_MPIAIJ(Mat,Vec,Vec);
766: PetscErrorCode MatAssemblyBegin_MPIAIJ(Mat mat,MatAssemblyType mode)
767: {
768: Mat_MPIAIJ *aij = (Mat_MPIAIJ*)mat->data;
770: PetscInt nstash,reallocs;
773: if (aij->donotstash || mat->nooffprocentries) return(0);
775: MatStashScatterBegin_Private(mat,&mat->stash,mat->rmap->range);
776: MatStashGetInfo_Private(&mat->stash,&nstash,&reallocs);
777: PetscInfo2(aij->A,"Stash has %D entries, uses %D mallocs.\n",nstash,reallocs);
778: return(0);
779: }
781: PetscErrorCode MatAssemblyEnd_MPIAIJ(Mat mat,MatAssemblyType mode)
782: {
783: Mat_MPIAIJ *aij = (Mat_MPIAIJ*)mat->data;
784: Mat_SeqAIJ *a = (Mat_SeqAIJ*)aij->A->data;
786: PetscMPIInt n;
787: PetscInt i,j,rstart,ncols,flg;
788: PetscInt *row,*col;
789: PetscBool other_disassembled;
790: PetscScalar *val;
792: /* do not use 'b = (Mat_SeqAIJ*)aij->B->data' as B can be reset in disassembly */
795: if (!aij->donotstash && !mat->nooffprocentries) {
796: while (1) {
797: MatStashScatterGetMesg_Private(&mat->stash,&n,&row,&col,&val,&flg);
798: if (!flg) break;
800: for (i=0; i<n; ) {
801: /* Now identify the consecutive vals belonging to the same row */
802: for (j=i,rstart=row[j]; j<n; j++) {
803: if (row[j] != rstart) break;
804: }
805: if (j < n) ncols = j-i;
806: else ncols = n-i;
807: /* Now assemble all these values with a single function call */
808: MatSetValues_MPIAIJ(mat,1,row+i,ncols,col+i,val+i,mat->insertmode);
810: i = j;
811: }
812: }
813: MatStashScatterEnd_Private(&mat->stash);
814: }
815: MatAssemblyBegin(aij->A,mode);
816: MatAssemblyEnd(aij->A,mode);
818: /* determine if any processor has disassembled, if so we must
819: also disassemble ourselfs, in order that we may reassemble. */
820: /*
821: if nonzero structure of submatrix B cannot change then we know that
822: no processor disassembled thus we can skip this stuff
823: */
824: if (!((Mat_SeqAIJ*)aij->B->data)->nonew) {
825: MPIU_Allreduce(&mat->was_assembled,&other_disassembled,1,MPIU_BOOL,MPI_PROD,PetscObjectComm((PetscObject)mat));
826: if (mat->was_assembled && !other_disassembled) {
827: MatDisAssemble_MPIAIJ(mat);
828: }
829: }
830: if (!mat->was_assembled && mode == MAT_FINAL_ASSEMBLY) {
831: MatSetUpMultiply_MPIAIJ(mat);
832: }
833: MatSetOption(aij->B,MAT_USE_INODES,PETSC_FALSE);
834: MatAssemblyBegin(aij->B,mode);
835: MatAssemblyEnd(aij->B,mode);
837: PetscFree2(aij->rowvalues,aij->rowindices);
839: aij->rowvalues = 0;
841: VecDestroy(&aij->diag);
842: if (a->inode.size) mat->ops->multdiagonalblock = MatMultDiagonalBlock_MPIAIJ;
844: /* if no new nonzero locations are allowed in matrix then only set the matrix state the first time through */
845: if ((!mat->was_assembled && mode == MAT_FINAL_ASSEMBLY) || !((Mat_SeqAIJ*)(aij->A->data))->nonew) {
846: PetscObjectState state = aij->A->nonzerostate + aij->B->nonzerostate;
847: MPIU_Allreduce(&state,&mat->nonzerostate,1,MPIU_INT64,MPI_SUM,PetscObjectComm((PetscObject)mat));
848: }
849: return(0);
850: }
852: PetscErrorCode MatZeroEntries_MPIAIJ(Mat A)
853: {
854: Mat_MPIAIJ *l = (Mat_MPIAIJ*)A->data;
858: MatZeroEntries(l->A);
859: MatZeroEntries(l->B);
860: return(0);
861: }
863: PetscErrorCode MatZeroRows_MPIAIJ(Mat A,PetscInt N,const PetscInt rows[],PetscScalar diag,Vec x,Vec b)
864: {
865: Mat_MPIAIJ *mat = (Mat_MPIAIJ *) A->data;
866: PetscInt *lrows;
867: PetscInt r, len;
868: PetscBool cong;
872: /* get locally owned rows */
873: MatZeroRowsMapLocal_Private(A,N,rows,&len,&lrows);
874: /* fix right hand side if needed */
875: if (x && b) {
876: const PetscScalar *xx;
877: PetscScalar *bb;
879: VecGetArrayRead(x, &xx);
880: VecGetArray(b, &bb);
881: for (r = 0; r < len; ++r) bb[lrows[r]] = diag*xx[lrows[r]];
882: VecRestoreArrayRead(x, &xx);
883: VecRestoreArray(b, &bb);
884: }
885: /* Must zero l->B before l->A because the (diag) case below may put values into l->B*/
886: MatZeroRows(mat->B, len, lrows, 0.0, NULL, NULL);
887: MatHasCongruentLayouts(A,&cong);
888: if ((diag != 0.0) && cong) {
889: MatZeroRows(mat->A, len, lrows, diag, NULL, NULL);
890: } else if (diag != 0.0) {
891: MatZeroRows(mat->A, len, lrows, 0.0, NULL, NULL);
892: if (((Mat_SeqAIJ *) mat->A->data)->nonew) SETERRQ(PETSC_COMM_SELF, PETSC_ERR_SUP, "MatZeroRows() on rectangular matrices cannot be used with the Mat options\nMAT_NEW_NONZERO_LOCATIONS,MAT_NEW_NONZERO_LOCATION_ERR,MAT_NEW_NONZERO_ALLOCATION_ERR");
893: for (r = 0; r < len; ++r) {
894: const PetscInt row = lrows[r] + A->rmap->rstart;
895: MatSetValues(A, 1, &row, 1, &row, &diag, INSERT_VALUES);
896: }
897: MatAssemblyBegin(A, MAT_FINAL_ASSEMBLY);
898: MatAssemblyEnd(A, MAT_FINAL_ASSEMBLY);
899: } else {
900: MatZeroRows(mat->A, len, lrows, 0.0, NULL, NULL);
901: }
902: PetscFree(lrows);
904: /* only change matrix nonzero state if pattern was allowed to be changed */
905: if (!((Mat_SeqAIJ*)(mat->A->data))->keepnonzeropattern) {
906: PetscObjectState state = mat->A->nonzerostate + mat->B->nonzerostate;
907: MPIU_Allreduce(&state,&A->nonzerostate,1,MPIU_INT64,MPI_SUM,PetscObjectComm((PetscObject)A));
908: }
909: return(0);
910: }
912: PetscErrorCode MatZeroRowsColumns_MPIAIJ(Mat A,PetscInt N,const PetscInt rows[],PetscScalar diag,Vec x,Vec b)
913: {
914: Mat_MPIAIJ *l = (Mat_MPIAIJ*)A->data;
915: PetscErrorCode ierr;
916: PetscMPIInt n = A->rmap->n;
917: PetscInt i,j,r,m,p = 0,len = 0;
918: PetscInt *lrows,*owners = A->rmap->range;
919: PetscSFNode *rrows;
920: PetscSF sf;
921: const PetscScalar *xx;
922: PetscScalar *bb,*mask;
923: Vec xmask,lmask;
924: Mat_SeqAIJ *aij = (Mat_SeqAIJ*)l->B->data;
925: const PetscInt *aj, *ii,*ridx;
926: PetscScalar *aa;
929: /* Create SF where leaves are input rows and roots are owned rows */
930: PetscMalloc1(n, &lrows);
931: for (r = 0; r < n; ++r) lrows[r] = -1;
932: PetscMalloc1(N, &rrows);
933: for (r = 0; r < N; ++r) {
934: const PetscInt idx = rows[r];
935: if (idx < 0 || A->rmap->N <= idx) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Row %D out of range [0,%D)",idx,A->rmap->N);
936: if (idx < owners[p] || owners[p+1] <= idx) { /* short-circuit the search if the last p owns this row too */
937: PetscLayoutFindOwner(A->rmap,idx,&p);
938: }
939: rrows[r].rank = p;
940: rrows[r].index = rows[r] - owners[p];
941: }
942: PetscSFCreate(PetscObjectComm((PetscObject) A), &sf);
943: PetscSFSetGraph(sf, n, N, NULL, PETSC_OWN_POINTER, rrows, PETSC_OWN_POINTER);
944: /* Collect flags for rows to be zeroed */
945: PetscSFReduceBegin(sf, MPIU_INT, (PetscInt *) rows, lrows, MPI_LOR);
946: PetscSFReduceEnd(sf, MPIU_INT, (PetscInt *) rows, lrows, MPI_LOR);
947: PetscSFDestroy(&sf);
948: /* Compress and put in row numbers */
949: for (r = 0; r < n; ++r) if (lrows[r] >= 0) lrows[len++] = r;
950: /* zero diagonal part of matrix */
951: MatZeroRowsColumns(l->A,len,lrows,diag,x,b);
952: /* handle off diagonal part of matrix */
953: MatCreateVecs(A,&xmask,NULL);
954: VecDuplicate(l->lvec,&lmask);
955: VecGetArray(xmask,&bb);
956: for (i=0; i<len; i++) bb[lrows[i]] = 1;
957: VecRestoreArray(xmask,&bb);
958: VecScatterBegin(l->Mvctx,xmask,lmask,ADD_VALUES,SCATTER_FORWARD);
959: VecScatterEnd(l->Mvctx,xmask,lmask,ADD_VALUES,SCATTER_FORWARD);
960: VecDestroy(&xmask);
961: if (x) {
962: VecScatterBegin(l->Mvctx,x,l->lvec,INSERT_VALUES,SCATTER_FORWARD);
963: VecScatterEnd(l->Mvctx,x,l->lvec,INSERT_VALUES,SCATTER_FORWARD);
964: VecGetArrayRead(l->lvec,&xx);
965: VecGetArray(b,&bb);
966: }
967: VecGetArray(lmask,&mask);
968: /* remove zeroed rows of off diagonal matrix */
969: ii = aij->i;
970: for (i=0; i<len; i++) {
971: PetscMemzero(aij->a + ii[lrows[i]],(ii[lrows[i]+1] - ii[lrows[i]])*sizeof(PetscScalar));
972: }
973: /* loop over all elements of off process part of matrix zeroing removed columns*/
974: if (aij->compressedrow.use) {
975: m = aij->compressedrow.nrows;
976: ii = aij->compressedrow.i;
977: ridx = aij->compressedrow.rindex;
978: for (i=0; i<m; i++) {
979: n = ii[i+1] - ii[i];
980: aj = aij->j + ii[i];
981: aa = aij->a + ii[i];
983: for (j=0; j<n; j++) {
984: if (PetscAbsScalar(mask[*aj])) {
985: if (b) bb[*ridx] -= *aa*xx[*aj];
986: *aa = 0.0;
987: }
988: aa++;
989: aj++;
990: }
991: ridx++;
992: }
993: } else { /* do not use compressed row format */
994: m = l->B->rmap->n;
995: for (i=0; i<m; i++) {
996: n = ii[i+1] - ii[i];
997: aj = aij->j + ii[i];
998: aa = aij->a + ii[i];
999: for (j=0; j<n; j++) {
1000: if (PetscAbsScalar(mask[*aj])) {
1001: if (b) bb[i] -= *aa*xx[*aj];
1002: *aa = 0.0;
1003: }
1004: aa++;
1005: aj++;
1006: }
1007: }
1008: }
1009: if (x) {
1010: VecRestoreArray(b,&bb);
1011: VecRestoreArrayRead(l->lvec,&xx);
1012: }
1013: VecRestoreArray(lmask,&mask);
1014: VecDestroy(&lmask);
1015: PetscFree(lrows);
1017: /* only change matrix nonzero state if pattern was allowed to be changed */
1018: if (!((Mat_SeqAIJ*)(l->A->data))->keepnonzeropattern) {
1019: PetscObjectState state = l->A->nonzerostate + l->B->nonzerostate;
1020: MPIU_Allreduce(&state,&A->nonzerostate,1,MPIU_INT64,MPI_SUM,PetscObjectComm((PetscObject)A));
1021: }
1022: return(0);
1023: }
1025: PetscErrorCode MatMult_MPIAIJ(Mat A,Vec xx,Vec yy)
1026: {
1027: Mat_MPIAIJ *a = (Mat_MPIAIJ*)A->data;
1029: PetscInt nt;
1030: VecScatter Mvctx = a->Mvctx;
1033: VecGetLocalSize(xx,&nt);
1034: if (nt != A->cmap->n) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Incompatible partition of A (%D) and xx (%D)",A->cmap->n,nt);
1036: VecScatterBegin(Mvctx,xx,a->lvec,INSERT_VALUES,SCATTER_FORWARD);
1037: (*a->A->ops->mult)(a->A,xx,yy);
1038: VecScatterEnd(Mvctx,xx,a->lvec,INSERT_VALUES,SCATTER_FORWARD);
1039: (*a->B->ops->multadd)(a->B,a->lvec,yy,yy);
1040: return(0);
1041: }
1043: PetscErrorCode MatMultDiagonalBlock_MPIAIJ(Mat A,Vec bb,Vec xx)
1044: {
1045: Mat_MPIAIJ *a = (Mat_MPIAIJ*)A->data;
1049: MatMultDiagonalBlock(a->A,bb,xx);
1050: return(0);
1051: }
1053: PetscErrorCode MatMultAdd_MPIAIJ(Mat A,Vec xx,Vec yy,Vec zz)
1054: {
1055: Mat_MPIAIJ *a = (Mat_MPIAIJ*)A->data;
1057: VecScatter Mvctx = a->Mvctx;
1060: if (a->Mvctx_mpi1_flg) Mvctx = a->Mvctx_mpi1;
1061: VecScatterBegin(Mvctx,xx,a->lvec,INSERT_VALUES,SCATTER_FORWARD);
1062: (*a->A->ops->multadd)(a->A,xx,yy,zz);
1063: VecScatterEnd(Mvctx,xx,a->lvec,INSERT_VALUES,SCATTER_FORWARD);
1064: (*a->B->ops->multadd)(a->B,a->lvec,zz,zz);
1065: return(0);
1066: }
1068: PetscErrorCode MatMultTranspose_MPIAIJ(Mat A,Vec xx,Vec yy)
1069: {
1070: Mat_MPIAIJ *a = (Mat_MPIAIJ*)A->data;
1074: /* do nondiagonal part */
1075: (*a->B->ops->multtranspose)(a->B,xx,a->lvec);
1076: /* do local part */
1077: (*a->A->ops->multtranspose)(a->A,xx,yy);
1078: /* add partial results together */
1079: VecScatterBegin(a->Mvctx,a->lvec,yy,ADD_VALUES,SCATTER_REVERSE);
1080: VecScatterEnd(a->Mvctx,a->lvec,yy,ADD_VALUES,SCATTER_REVERSE);
1081: return(0);
1082: }
1084: PetscErrorCode MatIsTranspose_MPIAIJ(Mat Amat,Mat Bmat,PetscReal tol,PetscBool *f)
1085: {
1086: MPI_Comm comm;
1087: Mat_MPIAIJ *Aij = (Mat_MPIAIJ*) Amat->data, *Bij;
1088: Mat Adia = Aij->A, Bdia, Aoff,Boff,*Aoffs,*Boffs;
1089: IS Me,Notme;
1091: PetscInt M,N,first,last,*notme,i;
1092: PetscBool lf;
1093: PetscMPIInt size;
1096: /* Easy test: symmetric diagonal block */
1097: Bij = (Mat_MPIAIJ*) Bmat->data; Bdia = Bij->A;
1098: MatIsTranspose(Adia,Bdia,tol,&lf);
1099: MPIU_Allreduce(&lf,f,1,MPIU_BOOL,MPI_LAND,PetscObjectComm((PetscObject)Amat));
1100: if (!*f) return(0);
1101: PetscObjectGetComm((PetscObject)Amat,&comm);
1102: MPI_Comm_size(comm,&size);
1103: if (size == 1) return(0);
1105: /* Hard test: off-diagonal block. This takes a MatCreateSubMatrix. */
1106: MatGetSize(Amat,&M,&N);
1107: MatGetOwnershipRange(Amat,&first,&last);
1108: PetscMalloc1(N-last+first,¬me);
1109: for (i=0; i<first; i++) notme[i] = i;
1110: for (i=last; i<M; i++) notme[i-last+first] = i;
1111: ISCreateGeneral(MPI_COMM_SELF,N-last+first,notme,PETSC_COPY_VALUES,&Notme);
1112: ISCreateStride(MPI_COMM_SELF,last-first,first,1,&Me);
1113: MatCreateSubMatrices(Amat,1,&Me,&Notme,MAT_INITIAL_MATRIX,&Aoffs);
1114: Aoff = Aoffs[0];
1115: MatCreateSubMatrices(Bmat,1,&Notme,&Me,MAT_INITIAL_MATRIX,&Boffs);
1116: Boff = Boffs[0];
1117: MatIsTranspose(Aoff,Boff,tol,f);
1118: MatDestroyMatrices(1,&Aoffs);
1119: MatDestroyMatrices(1,&Boffs);
1120: ISDestroy(&Me);
1121: ISDestroy(&Notme);
1122: PetscFree(notme);
1123: return(0);
1124: }
1126: PetscErrorCode MatIsSymmetric_MPIAIJ(Mat A,PetscReal tol,PetscBool *f)
1127: {
1131: MatIsTranspose_MPIAIJ(A,A,tol,f);
1132: return(0);
1133: }
1135: PetscErrorCode MatMultTransposeAdd_MPIAIJ(Mat A,Vec xx,Vec yy,Vec zz)
1136: {
1137: Mat_MPIAIJ *a = (Mat_MPIAIJ*)A->data;
1141: /* do nondiagonal part */
1142: (*a->B->ops->multtranspose)(a->B,xx,a->lvec);
1143: /* do local part */
1144: (*a->A->ops->multtransposeadd)(a->A,xx,yy,zz);
1145: /* add partial results together */
1146: VecScatterBegin(a->Mvctx,a->lvec,zz,ADD_VALUES,SCATTER_REVERSE);
1147: VecScatterEnd(a->Mvctx,a->lvec,zz,ADD_VALUES,SCATTER_REVERSE);
1148: return(0);
1149: }
1151: /*
1152: This only works correctly for square matrices where the subblock A->A is the
1153: diagonal block
1154: */
1155: PetscErrorCode MatGetDiagonal_MPIAIJ(Mat A,Vec v)
1156: {
1158: Mat_MPIAIJ *a = (Mat_MPIAIJ*)A->data;
1161: if (A->rmap->N != A->cmap->N) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_SUP,"Supports only square matrix where A->A is diag block");
1162: if (A->rmap->rstart != A->cmap->rstart || A->rmap->rend != A->cmap->rend) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"row partition must equal col partition");
1163: MatGetDiagonal(a->A,v);
1164: return(0);
1165: }
1167: PetscErrorCode MatScale_MPIAIJ(Mat A,PetscScalar aa)
1168: {
1169: Mat_MPIAIJ *a = (Mat_MPIAIJ*)A->data;
1173: MatScale(a->A,aa);
1174: MatScale(a->B,aa);
1175: return(0);
1176: }
1178: PetscErrorCode MatDestroy_MPIAIJ(Mat mat)
1179: {
1180: Mat_MPIAIJ *aij = (Mat_MPIAIJ*)mat->data;
1184: #if defined(PETSC_USE_LOG)
1185: PetscLogObjectState((PetscObject)mat,"Rows=%D, Cols=%D",mat->rmap->N,mat->cmap->N);
1186: #endif
1187: MatStashDestroy_Private(&mat->stash);
1188: VecDestroy(&aij->diag);
1189: MatDestroy(&aij->A);
1190: MatDestroy(&aij->B);
1191: #if defined(PETSC_USE_CTABLE)
1192: PetscTableDestroy(&aij->colmap);
1193: #else
1194: PetscFree(aij->colmap);
1195: #endif
1196: PetscFree(aij->garray);
1197: VecDestroy(&aij->lvec);
1198: VecScatterDestroy(&aij->Mvctx);
1199: if (aij->Mvctx_mpi1) {VecScatterDestroy(&aij->Mvctx_mpi1);}
1200: PetscFree2(aij->rowvalues,aij->rowindices);
1201: PetscFree(aij->ld);
1202: PetscFree(mat->data);
1204: PetscObjectChangeTypeName((PetscObject)mat,0);
1205: PetscObjectComposeFunction((PetscObject)mat,"MatStoreValues_C",NULL);
1206: PetscObjectComposeFunction((PetscObject)mat,"MatRetrieveValues_C",NULL);
1207: PetscObjectComposeFunction((PetscObject)mat,"MatIsTranspose_C",NULL);
1208: PetscObjectComposeFunction((PetscObject)mat,"MatMPIAIJSetPreallocation_C",NULL);
1209: PetscObjectComposeFunction((PetscObject)mat,"MatResetPreallocation_C",NULL);
1210: PetscObjectComposeFunction((PetscObject)mat,"MatMPIAIJSetPreallocationCSR_C",NULL);
1211: PetscObjectComposeFunction((PetscObject)mat,"MatDiagonalScaleLocal_C",NULL);
1212: PetscObjectComposeFunction((PetscObject)mat,"MatConvert_mpiaij_mpisbaij_C",NULL);
1213: #if defined(PETSC_HAVE_ELEMENTAL)
1214: PetscObjectComposeFunction((PetscObject)mat,"MatConvert_mpiaij_elemental_C",NULL);
1215: #endif
1216: #if defined(PETSC_HAVE_HYPRE)
1217: PetscObjectComposeFunction((PetscObject)mat,"MatConvert_mpiaij_hypre_C",NULL);
1218: PetscObjectComposeFunction((PetscObject)mat,"MatMatMatMult_transpose_mpiaij_mpiaij_C",NULL);
1219: #endif
1220: PetscObjectComposeFunction((PetscObject)mat,"MatConvert_mpiaij_is_C",NULL);
1221: PetscObjectComposeFunction((PetscObject)mat,"MatPtAP_is_mpiaij_C",NULL);
1222: return(0);
1223: }
1225: PetscErrorCode MatView_MPIAIJ_Binary(Mat mat,PetscViewer viewer)
1226: {
1227: Mat_MPIAIJ *aij = (Mat_MPIAIJ*)mat->data;
1228: Mat_SeqAIJ *A = (Mat_SeqAIJ*)aij->A->data;
1229: Mat_SeqAIJ *B = (Mat_SeqAIJ*)aij->B->data;
1231: PetscMPIInt rank,size,tag = ((PetscObject)viewer)->tag;
1232: int fd;
1233: PetscInt nz,header[4],*row_lengths,*range=0,rlen,i;
1234: PetscInt nzmax,*column_indices,j,k,col,*garray = aij->garray,cnt,cstart = mat->cmap->rstart,rnz = 0;
1235: PetscScalar *column_values;
1236: PetscInt message_count,flowcontrolcount;
1237: FILE *file;
1240: MPI_Comm_rank(PetscObjectComm((PetscObject)mat),&rank);
1241: MPI_Comm_size(PetscObjectComm((PetscObject)mat),&size);
1242: nz = A->nz + B->nz;
1243: PetscViewerBinaryGetDescriptor(viewer,&fd);
1244: if (!rank) {
1245: header[0] = MAT_FILE_CLASSID;
1246: header[1] = mat->rmap->N;
1247: header[2] = mat->cmap->N;
1249: MPI_Reduce(&nz,&header[3],1,MPIU_INT,MPI_SUM,0,PetscObjectComm((PetscObject)mat));
1250: PetscBinaryWrite(fd,header,4,PETSC_INT,PETSC_TRUE);
1251: /* get largest number of rows any processor has */
1252: rlen = mat->rmap->n;
1253: range = mat->rmap->range;
1254: for (i=1; i<size; i++) rlen = PetscMax(rlen,range[i+1] - range[i]);
1255: } else {
1256: MPI_Reduce(&nz,0,1,MPIU_INT,MPI_SUM,0,PetscObjectComm((PetscObject)mat));
1257: rlen = mat->rmap->n;
1258: }
1260: /* load up the local row counts */
1261: PetscMalloc1(rlen+1,&row_lengths);
1262: for (i=0; i<mat->rmap->n; i++) row_lengths[i] = A->i[i+1] - A->i[i] + B->i[i+1] - B->i[i];
1264: /* store the row lengths to the file */
1265: PetscViewerFlowControlStart(viewer,&message_count,&flowcontrolcount);
1266: if (!rank) {
1267: PetscBinaryWrite(fd,row_lengths,mat->rmap->n,PETSC_INT,PETSC_TRUE);
1268: for (i=1; i<size; i++) {
1269: PetscViewerFlowControlStepMaster(viewer,i,&message_count,flowcontrolcount);
1270: rlen = range[i+1] - range[i];
1271: MPIULong_Recv(row_lengths,rlen,MPIU_INT,i,tag,PetscObjectComm((PetscObject)mat));
1272: PetscBinaryWrite(fd,row_lengths,rlen,PETSC_INT,PETSC_TRUE);
1273: }
1274: PetscViewerFlowControlEndMaster(viewer,&message_count);
1275: } else {
1276: PetscViewerFlowControlStepWorker(viewer,rank,&message_count);
1277: MPIULong_Send(row_lengths,mat->rmap->n,MPIU_INT,0,tag,PetscObjectComm((PetscObject)mat));
1278: PetscViewerFlowControlEndWorker(viewer,&message_count);
1279: }
1280: PetscFree(row_lengths);
1282: /* load up the local column indices */
1283: nzmax = nz; /* th processor needs space a largest processor needs */
1284: MPI_Reduce(&nz,&nzmax,1,MPIU_INT,MPI_MAX,0,PetscObjectComm((PetscObject)mat));
1285: PetscMalloc1(nzmax+1,&column_indices);
1286: cnt = 0;
1287: for (i=0; i<mat->rmap->n; i++) {
1288: for (j=B->i[i]; j<B->i[i+1]; j++) {
1289: if ((col = garray[B->j[j]]) > cstart) break;
1290: column_indices[cnt++] = col;
1291: }
1292: for (k=A->i[i]; k<A->i[i+1]; k++) column_indices[cnt++] = A->j[k] + cstart;
1293: for (; j<B->i[i+1]; j++) column_indices[cnt++] = garray[B->j[j]];
1294: }
1295: if (cnt != A->nz + B->nz) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_LIB,"Internal PETSc error: cnt = %D nz = %D",cnt,A->nz+B->nz);
1297: /* store the column indices to the file */
1298: PetscViewerFlowControlStart(viewer,&message_count,&flowcontrolcount);
1299: if (!rank) {
1300: MPI_Status status;
1301: PetscBinaryWrite(fd,column_indices,nz,PETSC_INT,PETSC_TRUE);
1302: for (i=1; i<size; i++) {
1303: PetscViewerFlowControlStepMaster(viewer,i,&message_count,flowcontrolcount);
1304: MPI_Recv(&rnz,1,MPIU_INT,i,tag,PetscObjectComm((PetscObject)mat),&status);
1305: if (rnz > nzmax) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_LIB,"Internal PETSc error: nz = %D nzmax = %D",nz,nzmax);
1306: MPIULong_Recv(column_indices,rnz,MPIU_INT,i,tag,PetscObjectComm((PetscObject)mat));
1307: PetscBinaryWrite(fd,column_indices,rnz,PETSC_INT,PETSC_TRUE);
1308: }
1309: PetscViewerFlowControlEndMaster(viewer,&message_count);
1310: } else {
1311: PetscViewerFlowControlStepWorker(viewer,rank,&message_count);
1312: MPI_Send(&nz,1,MPIU_INT,0,tag,PetscObjectComm((PetscObject)mat));
1313: MPIULong_Send(column_indices,nz,MPIU_INT,0,tag,PetscObjectComm((PetscObject)mat));
1314: PetscViewerFlowControlEndWorker(viewer,&message_count);
1315: }
1316: PetscFree(column_indices);
1318: /* load up the local column values */
1319: PetscMalloc1(nzmax+1,&column_values);
1320: cnt = 0;
1321: for (i=0; i<mat->rmap->n; i++) {
1322: for (j=B->i[i]; j<B->i[i+1]; j++) {
1323: if (garray[B->j[j]] > cstart) break;
1324: column_values[cnt++] = B->a[j];
1325: }
1326: for (k=A->i[i]; k<A->i[i+1]; k++) column_values[cnt++] = A->a[k];
1327: for (; j<B->i[i+1]; j++) column_values[cnt++] = B->a[j];
1328: }
1329: if (cnt != A->nz + B->nz) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_PLIB,"Internal PETSc error: cnt = %D nz = %D",cnt,A->nz+B->nz);
1331: /* store the column values to the file */
1332: PetscViewerFlowControlStart(viewer,&message_count,&flowcontrolcount);
1333: if (!rank) {
1334: MPI_Status status;
1335: PetscBinaryWrite(fd,column_values,nz,PETSC_SCALAR,PETSC_TRUE);
1336: for (i=1; i<size; i++) {
1337: PetscViewerFlowControlStepMaster(viewer,i,&message_count,flowcontrolcount);
1338: MPI_Recv(&rnz,1,MPIU_INT,i,tag,PetscObjectComm((PetscObject)mat),&status);
1339: if (rnz > nzmax) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_LIB,"Internal PETSc error: nz = %D nzmax = %D",nz,nzmax);
1340: MPIULong_Recv(column_values,rnz,MPIU_SCALAR,i,tag,PetscObjectComm((PetscObject)mat));
1341: PetscBinaryWrite(fd,column_values,rnz,PETSC_SCALAR,PETSC_TRUE);
1342: }
1343: PetscViewerFlowControlEndMaster(viewer,&message_count);
1344: } else {
1345: PetscViewerFlowControlStepWorker(viewer,rank,&message_count);
1346: MPI_Send(&nz,1,MPIU_INT,0,tag,PetscObjectComm((PetscObject)mat));
1347: MPIULong_Send(column_values,nz,MPIU_SCALAR,0,tag,PetscObjectComm((PetscObject)mat));
1348: PetscViewerFlowControlEndWorker(viewer,&message_count);
1349: }
1350: PetscFree(column_values);
1352: PetscViewerBinaryGetInfoPointer(viewer,&file);
1353: if (file) fprintf(file,"-matload_block_size %d\n",(int)PetscAbs(mat->rmap->bs));
1354: return(0);
1355: }
1357: #include <petscdraw.h>
1358: PetscErrorCode MatView_MPIAIJ_ASCIIorDraworSocket(Mat mat,PetscViewer viewer)
1359: {
1360: Mat_MPIAIJ *aij = (Mat_MPIAIJ*)mat->data;
1361: PetscErrorCode ierr;
1362: PetscMPIInt rank = aij->rank,size = aij->size;
1363: PetscBool isdraw,iascii,isbinary;
1364: PetscViewer sviewer;
1365: PetscViewerFormat format;
1368: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERDRAW,&isdraw);
1369: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERASCII,&iascii);
1370: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERBINARY,&isbinary);
1371: if (iascii) {
1372: PetscViewerGetFormat(viewer,&format);
1373: if (format == PETSC_VIEWER_LOAD_BALANCE) {
1374: PetscInt i,nmax = 0,nmin = PETSC_MAX_INT,navg = 0,*nz,nzlocal = ((Mat_SeqAIJ*) (aij->A->data))->nz + ((Mat_SeqAIJ*) (aij->B->data))->nz;
1375: PetscMalloc1(size,&nz);
1376: MPI_Allgather(&nzlocal,1,MPIU_INT,nz,1,MPIU_INT,PetscObjectComm((PetscObject)mat));
1377: for (i=0; i<(PetscInt)size; i++) {
1378: nmax = PetscMax(nmax,nz[i]);
1379: nmin = PetscMin(nmin,nz[i]);
1380: navg += nz[i];
1381: }
1382: PetscFree(nz);
1383: navg = navg/size;
1384: PetscViewerASCIIPrintf(viewer,"Load Balance - Nonzeros: Min %D avg %D max %D\n",nmin,navg,nmax);
1385: return(0);
1386: }
1387: PetscViewerGetFormat(viewer,&format);
1388: if (format == PETSC_VIEWER_ASCII_INFO_DETAIL) {
1389: MatInfo info;
1390: PetscBool inodes;
1392: MPI_Comm_rank(PetscObjectComm((PetscObject)mat),&rank);
1393: MatGetInfo(mat,MAT_LOCAL,&info);
1394: MatInodeGetInodeSizes(aij->A,NULL,(PetscInt**)&inodes,NULL);
1395: PetscViewerASCIIPushSynchronized(viewer);
1396: if (!inodes) {
1397: PetscViewerASCIISynchronizedPrintf(viewer,"[%d] Local rows %D nz %D nz alloced %D mem %g, not using I-node routines\n",
1398: rank,mat->rmap->n,(PetscInt)info.nz_used,(PetscInt)info.nz_allocated,(double)info.memory);
1399: } else {
1400: PetscViewerASCIISynchronizedPrintf(viewer,"[%d] Local rows %D nz %D nz alloced %D mem %g, using I-node routines\n",
1401: rank,mat->rmap->n,(PetscInt)info.nz_used,(PetscInt)info.nz_allocated,(double)info.memory);
1402: }
1403: MatGetInfo(aij->A,MAT_LOCAL,&info);
1404: PetscViewerASCIISynchronizedPrintf(viewer,"[%d] on-diagonal part: nz %D \n",rank,(PetscInt)info.nz_used);
1405: MatGetInfo(aij->B,MAT_LOCAL,&info);
1406: PetscViewerASCIISynchronizedPrintf(viewer,"[%d] off-diagonal part: nz %D \n",rank,(PetscInt)info.nz_used);
1407: PetscViewerFlush(viewer);
1408: PetscViewerASCIIPopSynchronized(viewer);
1409: PetscViewerASCIIPrintf(viewer,"Information on VecScatter used in matrix-vector product: \n");
1410: VecScatterView(aij->Mvctx,viewer);
1411: return(0);
1412: } else if (format == PETSC_VIEWER_ASCII_INFO) {
1413: PetscInt inodecount,inodelimit,*inodes;
1414: MatInodeGetInodeSizes(aij->A,&inodecount,&inodes,&inodelimit);
1415: if (inodes) {
1416: PetscViewerASCIIPrintf(viewer,"using I-node (on process 0) routines: found %D nodes, limit used is %D\n",inodecount,inodelimit);
1417: } else {
1418: PetscViewerASCIIPrintf(viewer,"not using I-node (on process 0) routines\n");
1419: }
1420: return(0);
1421: } else if (format == PETSC_VIEWER_ASCII_FACTOR_INFO) {
1422: return(0);
1423: }
1424: } else if (isbinary) {
1425: if (size == 1) {
1426: PetscObjectSetName((PetscObject)aij->A,((PetscObject)mat)->name);
1427: MatView(aij->A,viewer);
1428: } else {
1429: MatView_MPIAIJ_Binary(mat,viewer);
1430: }
1431: return(0);
1432: } else if (iascii && size == 1) {
1433: PetscObjectSetName((PetscObject)aij->A,((PetscObject)mat)->name);
1434: MatView(aij->A,viewer);
1435: return(0);
1436: } else if (isdraw) {
1437: PetscDraw draw;
1438: PetscBool isnull;
1439: PetscViewerDrawGetDraw(viewer,0,&draw);
1440: PetscDrawIsNull(draw,&isnull);
1441: if (isnull) return(0);
1442: }
1444: { /* assemble the entire matrix onto first processor */
1445: Mat A = NULL, Av;
1446: IS isrow,iscol;
1448: ISCreateStride(PetscObjectComm((PetscObject)mat),!rank ? mat->rmap->N : 0,0,1,&isrow);
1449: ISCreateStride(PetscObjectComm((PetscObject)mat),!rank ? mat->cmap->N : 0,0,1,&iscol);
1450: MatCreateSubMatrix(mat,isrow,iscol,MAT_INITIAL_MATRIX,&A);
1451: MatMPIAIJGetSeqAIJ(A,&Av,NULL,NULL);
1452: /* The commented code uses MatCreateSubMatrices instead */
1453: /*
1454: Mat *AA, A = NULL, Av;
1455: IS isrow,iscol;
1457: ISCreateStride(PetscObjectComm((PetscObject)mat),!rank ? mat->rmap->N : 0,0,1,&isrow);
1458: ISCreateStride(PetscObjectComm((PetscObject)mat),!rank ? mat->cmap->N : 0,0,1,&iscol);
1459: MatCreateSubMatrices(mat,1,&isrow,&iscol,MAT_INITIAL_MATRIX,&AA);
1460: if (!rank) {
1461: PetscObjectReference((PetscObject)AA[0]);
1462: A = AA[0];
1463: Av = AA[0];
1464: }
1465: MatDestroySubMatrices(1,&AA);
1466: */
1467: ISDestroy(&iscol);
1468: ISDestroy(&isrow);
1469: /*
1470: Everyone has to call to draw the matrix since the graphics waits are
1471: synchronized across all processors that share the PetscDraw object
1472: */
1473: PetscViewerGetSubViewer(viewer,PETSC_COMM_SELF,&sviewer);
1474: if (!rank) {
1475: if (((PetscObject)mat)->name) {
1476: PetscObjectSetName((PetscObject)Av,((PetscObject)mat)->name);
1477: }
1478: MatView_SeqAIJ(Av,sviewer);
1479: }
1480: PetscViewerRestoreSubViewer(viewer,PETSC_COMM_SELF,&sviewer);
1481: PetscViewerFlush(viewer);
1482: MatDestroy(&A);
1483: }
1484: return(0);
1485: }
1487: PetscErrorCode MatView_MPIAIJ(Mat mat,PetscViewer viewer)
1488: {
1490: PetscBool iascii,isdraw,issocket,isbinary;
1493: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERASCII,&iascii);
1494: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERDRAW,&isdraw);
1495: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERBINARY,&isbinary);
1496: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERSOCKET,&issocket);
1497: if (iascii || isdraw || isbinary || issocket) {
1498: MatView_MPIAIJ_ASCIIorDraworSocket(mat,viewer);
1499: }
1500: return(0);
1501: }
1503: PetscErrorCode MatSOR_MPIAIJ(Mat matin,Vec bb,PetscReal omega,MatSORType flag,PetscReal fshift,PetscInt its,PetscInt lits,Vec xx)
1504: {
1505: Mat_MPIAIJ *mat = (Mat_MPIAIJ*)matin->data;
1507: Vec bb1 = 0;
1508: PetscBool hasop;
1511: if (flag == SOR_APPLY_UPPER) {
1512: (*mat->A->ops->sor)(mat->A,bb,omega,flag,fshift,lits,1,xx);
1513: return(0);
1514: }
1516: if (its > 1 || ~flag & SOR_ZERO_INITIAL_GUESS || flag & SOR_EISENSTAT) {
1517: VecDuplicate(bb,&bb1);
1518: }
1520: if ((flag & SOR_LOCAL_SYMMETRIC_SWEEP) == SOR_LOCAL_SYMMETRIC_SWEEP) {
1521: if (flag & SOR_ZERO_INITIAL_GUESS) {
1522: (*mat->A->ops->sor)(mat->A,bb,omega,flag,fshift,lits,1,xx);
1523: its--;
1524: }
1526: while (its--) {
1527: VecScatterBegin(mat->Mvctx,xx,mat->lvec,INSERT_VALUES,SCATTER_FORWARD);
1528: VecScatterEnd(mat->Mvctx,xx,mat->lvec,INSERT_VALUES,SCATTER_FORWARD);
1530: /* update rhs: bb1 = bb - B*x */
1531: VecScale(mat->lvec,-1.0);
1532: (*mat->B->ops->multadd)(mat->B,mat->lvec,bb,bb1);
1534: /* local sweep */
1535: (*mat->A->ops->sor)(mat->A,bb1,omega,SOR_SYMMETRIC_SWEEP,fshift,lits,1,xx);
1536: }
1537: } else if (flag & SOR_LOCAL_FORWARD_SWEEP) {
1538: if (flag & SOR_ZERO_INITIAL_GUESS) {
1539: (*mat->A->ops->sor)(mat->A,bb,omega,flag,fshift,lits,1,xx);
1540: its--;
1541: }
1542: while (its--) {
1543: VecScatterBegin(mat->Mvctx,xx,mat->lvec,INSERT_VALUES,SCATTER_FORWARD);
1544: VecScatterEnd(mat->Mvctx,xx,mat->lvec,INSERT_VALUES,SCATTER_FORWARD);
1546: /* update rhs: bb1 = bb - B*x */
1547: VecScale(mat->lvec,-1.0);
1548: (*mat->B->ops->multadd)(mat->B,mat->lvec,bb,bb1);
1550: /* local sweep */
1551: (*mat->A->ops->sor)(mat->A,bb1,omega,SOR_FORWARD_SWEEP,fshift,lits,1,xx);
1552: }
1553: } else if (flag & SOR_LOCAL_BACKWARD_SWEEP) {
1554: if (flag & SOR_ZERO_INITIAL_GUESS) {
1555: (*mat->A->ops->sor)(mat->A,bb,omega,flag,fshift,lits,1,xx);
1556: its--;
1557: }
1558: while (its--) {
1559: VecScatterBegin(mat->Mvctx,xx,mat->lvec,INSERT_VALUES,SCATTER_FORWARD);
1560: VecScatterEnd(mat->Mvctx,xx,mat->lvec,INSERT_VALUES,SCATTER_FORWARD);
1562: /* update rhs: bb1 = bb - B*x */
1563: VecScale(mat->lvec,-1.0);
1564: (*mat->B->ops->multadd)(mat->B,mat->lvec,bb,bb1);
1566: /* local sweep */
1567: (*mat->A->ops->sor)(mat->A,bb1,omega,SOR_BACKWARD_SWEEP,fshift,lits,1,xx);
1568: }
1569: } else if (flag & SOR_EISENSTAT) {
1570: Vec xx1;
1572: VecDuplicate(bb,&xx1);
1573: (*mat->A->ops->sor)(mat->A,bb,omega,(MatSORType)(SOR_ZERO_INITIAL_GUESS | SOR_LOCAL_BACKWARD_SWEEP),fshift,lits,1,xx);
1575: VecScatterBegin(mat->Mvctx,xx,mat->lvec,INSERT_VALUES,SCATTER_FORWARD);
1576: VecScatterEnd(mat->Mvctx,xx,mat->lvec,INSERT_VALUES,SCATTER_FORWARD);
1577: if (!mat->diag) {
1578: MatCreateVecs(matin,&mat->diag,NULL);
1579: MatGetDiagonal(matin,mat->diag);
1580: }
1581: MatHasOperation(matin,MATOP_MULT_DIAGONAL_BLOCK,&hasop);
1582: if (hasop) {
1583: MatMultDiagonalBlock(matin,xx,bb1);
1584: } else {
1585: VecPointwiseMult(bb1,mat->diag,xx);
1586: }
1587: VecAYPX(bb1,(omega-2.0)/omega,bb);
1589: MatMultAdd(mat->B,mat->lvec,bb1,bb1);
1591: /* local sweep */
1592: (*mat->A->ops->sor)(mat->A,bb1,omega,(MatSORType)(SOR_ZERO_INITIAL_GUESS | SOR_LOCAL_FORWARD_SWEEP),fshift,lits,1,xx1);
1593: VecAXPY(xx,1.0,xx1);
1594: VecDestroy(&xx1);
1595: } else SETERRQ(PetscObjectComm((PetscObject)matin),PETSC_ERR_SUP,"Parallel SOR not supported");
1597: VecDestroy(&bb1);
1599: matin->factorerrortype = mat->A->factorerrortype;
1600: return(0);
1601: }
1603: PetscErrorCode MatPermute_MPIAIJ(Mat A,IS rowp,IS colp,Mat *B)
1604: {
1605: Mat aA,aB,Aperm;
1606: const PetscInt *rwant,*cwant,*gcols,*ai,*bi,*aj,*bj;
1607: PetscScalar *aa,*ba;
1608: PetscInt i,j,m,n,ng,anz,bnz,*dnnz,*onnz,*tdnnz,*tonnz,*rdest,*cdest,*work,*gcdest;
1609: PetscSF rowsf,sf;
1610: IS parcolp = NULL;
1611: PetscBool done;
1615: MatGetLocalSize(A,&m,&n);
1616: ISGetIndices(rowp,&rwant);
1617: ISGetIndices(colp,&cwant);
1618: PetscMalloc3(PetscMax(m,n),&work,m,&rdest,n,&cdest);
1620: /* Invert row permutation to find out where my rows should go */
1621: PetscSFCreate(PetscObjectComm((PetscObject)A),&rowsf);
1622: PetscSFSetGraphLayout(rowsf,A->rmap,A->rmap->n,NULL,PETSC_OWN_POINTER,rwant);
1623: PetscSFSetFromOptions(rowsf);
1624: for (i=0; i<m; i++) work[i] = A->rmap->rstart + i;
1625: PetscSFReduceBegin(rowsf,MPIU_INT,work,rdest,MPIU_REPLACE);
1626: PetscSFReduceEnd(rowsf,MPIU_INT,work,rdest,MPIU_REPLACE);
1628: /* Invert column permutation to find out where my columns should go */
1629: PetscSFCreate(PetscObjectComm((PetscObject)A),&sf);
1630: PetscSFSetGraphLayout(sf,A->cmap,A->cmap->n,NULL,PETSC_OWN_POINTER,cwant);
1631: PetscSFSetFromOptions(sf);
1632: for (i=0; i<n; i++) work[i] = A->cmap->rstart + i;
1633: PetscSFReduceBegin(sf,MPIU_INT,work,cdest,MPIU_REPLACE);
1634: PetscSFReduceEnd(sf,MPIU_INT,work,cdest,MPIU_REPLACE);
1635: PetscSFDestroy(&sf);
1637: ISRestoreIndices(rowp,&rwant);
1638: ISRestoreIndices(colp,&cwant);
1639: MatMPIAIJGetSeqAIJ(A,&aA,&aB,&gcols);
1641: /* Find out where my gcols should go */
1642: MatGetSize(aB,NULL,&ng);
1643: PetscMalloc1(ng,&gcdest);
1644: PetscSFCreate(PetscObjectComm((PetscObject)A),&sf);
1645: PetscSFSetGraphLayout(sf,A->cmap,ng,NULL,PETSC_OWN_POINTER,gcols);
1646: PetscSFSetFromOptions(sf);
1647: PetscSFBcastBegin(sf,MPIU_INT,cdest,gcdest);
1648: PetscSFBcastEnd(sf,MPIU_INT,cdest,gcdest);
1649: PetscSFDestroy(&sf);
1651: PetscCalloc4(m,&dnnz,m,&onnz,m,&tdnnz,m,&tonnz);
1652: MatGetRowIJ(aA,0,PETSC_FALSE,PETSC_FALSE,&anz,&ai,&aj,&done);
1653: MatGetRowIJ(aB,0,PETSC_FALSE,PETSC_FALSE,&bnz,&bi,&bj,&done);
1654: for (i=0; i<m; i++) {
1655: PetscInt row = rdest[i],rowner;
1656: PetscLayoutFindOwner(A->rmap,row,&rowner);
1657: for (j=ai[i]; j<ai[i+1]; j++) {
1658: PetscInt cowner,col = cdest[aj[j]];
1659: PetscLayoutFindOwner(A->cmap,col,&cowner); /* Could build an index for the columns to eliminate this search */
1660: if (rowner == cowner) dnnz[i]++;
1661: else onnz[i]++;
1662: }
1663: for (j=bi[i]; j<bi[i+1]; j++) {
1664: PetscInt cowner,col = gcdest[bj[j]];
1665: PetscLayoutFindOwner(A->cmap,col,&cowner);
1666: if (rowner == cowner) dnnz[i]++;
1667: else onnz[i]++;
1668: }
1669: }
1670: PetscSFBcastBegin(rowsf,MPIU_INT,dnnz,tdnnz);
1671: PetscSFBcastEnd(rowsf,MPIU_INT,dnnz,tdnnz);
1672: PetscSFBcastBegin(rowsf,MPIU_INT,onnz,tonnz);
1673: PetscSFBcastEnd(rowsf,MPIU_INT,onnz,tonnz);
1674: PetscSFDestroy(&rowsf);
1676: MatCreateAIJ(PetscObjectComm((PetscObject)A),A->rmap->n,A->cmap->n,A->rmap->N,A->cmap->N,0,tdnnz,0,tonnz,&Aperm);
1677: MatSeqAIJGetArray(aA,&aa);
1678: MatSeqAIJGetArray(aB,&ba);
1679: for (i=0; i<m; i++) {
1680: PetscInt *acols = dnnz,*bcols = onnz; /* Repurpose now-unneeded arrays */
1681: PetscInt j0,rowlen;
1682: rowlen = ai[i+1] - ai[i];
1683: for (j0=j=0; j<rowlen; j0=j) { /* rowlen could be larger than number of rows m, so sum in batches */
1684: for ( ; j<PetscMin(rowlen,j0+m); j++) acols[j-j0] = cdest[aj[ai[i]+j]];
1685: MatSetValues(Aperm,1,&rdest[i],j-j0,acols,aa+ai[i]+j0,INSERT_VALUES);
1686: }
1687: rowlen = bi[i+1] - bi[i];
1688: for (j0=j=0; j<rowlen; j0=j) {
1689: for ( ; j<PetscMin(rowlen,j0+m); j++) bcols[j-j0] = gcdest[bj[bi[i]+j]];
1690: MatSetValues(Aperm,1,&rdest[i],j-j0,bcols,ba+bi[i]+j0,INSERT_VALUES);
1691: }
1692: }
1693: MatAssemblyBegin(Aperm,MAT_FINAL_ASSEMBLY);
1694: MatAssemblyEnd(Aperm,MAT_FINAL_ASSEMBLY);
1695: MatRestoreRowIJ(aA,0,PETSC_FALSE,PETSC_FALSE,&anz,&ai,&aj,&done);
1696: MatRestoreRowIJ(aB,0,PETSC_FALSE,PETSC_FALSE,&bnz,&bi,&bj,&done);
1697: MatSeqAIJRestoreArray(aA,&aa);
1698: MatSeqAIJRestoreArray(aB,&ba);
1699: PetscFree4(dnnz,onnz,tdnnz,tonnz);
1700: PetscFree3(work,rdest,cdest);
1701: PetscFree(gcdest);
1702: if (parcolp) {ISDestroy(&colp);}
1703: *B = Aperm;
1704: return(0);
1705: }
1707: PetscErrorCode MatGetGhosts_MPIAIJ(Mat mat,PetscInt *nghosts,const PetscInt *ghosts[])
1708: {
1709: Mat_MPIAIJ *aij = (Mat_MPIAIJ*)mat->data;
1713: MatGetSize(aij->B,NULL,nghosts);
1714: if (ghosts) *ghosts = aij->garray;
1715: return(0);
1716: }
1718: PetscErrorCode MatGetInfo_MPIAIJ(Mat matin,MatInfoType flag,MatInfo *info)
1719: {
1720: Mat_MPIAIJ *mat = (Mat_MPIAIJ*)matin->data;
1721: Mat A = mat->A,B = mat->B;
1723: PetscReal isend[5],irecv[5];
1726: info->block_size = 1.0;
1727: MatGetInfo(A,MAT_LOCAL,info);
1729: isend[0] = info->nz_used; isend[1] = info->nz_allocated; isend[2] = info->nz_unneeded;
1730: isend[3] = info->memory; isend[4] = info->mallocs;
1732: MatGetInfo(B,MAT_LOCAL,info);
1734: isend[0] += info->nz_used; isend[1] += info->nz_allocated; isend[2] += info->nz_unneeded;
1735: isend[3] += info->memory; isend[4] += info->mallocs;
1736: if (flag == MAT_LOCAL) {
1737: info->nz_used = isend[0];
1738: info->nz_allocated = isend[1];
1739: info->nz_unneeded = isend[2];
1740: info->memory = isend[3];
1741: info->mallocs = isend[4];
1742: } else if (flag == MAT_GLOBAL_MAX) {
1743: MPIU_Allreduce(isend,irecv,5,MPIU_REAL,MPIU_MAX,PetscObjectComm((PetscObject)matin));
1745: info->nz_used = irecv[0];
1746: info->nz_allocated = irecv[1];
1747: info->nz_unneeded = irecv[2];
1748: info->memory = irecv[3];
1749: info->mallocs = irecv[4];
1750: } else if (flag == MAT_GLOBAL_SUM) {
1751: MPIU_Allreduce(isend,irecv,5,MPIU_REAL,MPIU_SUM,PetscObjectComm((PetscObject)matin));
1753: info->nz_used = irecv[0];
1754: info->nz_allocated = irecv[1];
1755: info->nz_unneeded = irecv[2];
1756: info->memory = irecv[3];
1757: info->mallocs = irecv[4];
1758: }
1759: info->fill_ratio_given = 0; /* no parallel LU/ILU/Cholesky */
1760: info->fill_ratio_needed = 0;
1761: info->factor_mallocs = 0;
1762: return(0);
1763: }
1765: PetscErrorCode MatSetOption_MPIAIJ(Mat A,MatOption op,PetscBool flg)
1766: {
1767: Mat_MPIAIJ *a = (Mat_MPIAIJ*)A->data;
1771: switch (op) {
1772: case MAT_NEW_NONZERO_LOCATIONS:
1773: case MAT_NEW_NONZERO_ALLOCATION_ERR:
1774: case MAT_UNUSED_NONZERO_LOCATION_ERR:
1775: case MAT_KEEP_NONZERO_PATTERN:
1776: case MAT_NEW_NONZERO_LOCATION_ERR:
1777: case MAT_USE_INODES:
1778: case MAT_IGNORE_ZERO_ENTRIES:
1779: MatCheckPreallocated(A,1);
1780: MatSetOption(a->A,op,flg);
1781: MatSetOption(a->B,op,flg);
1782: break;
1783: case MAT_ROW_ORIENTED:
1784: MatCheckPreallocated(A,1);
1785: a->roworiented = flg;
1787: MatSetOption(a->A,op,flg);
1788: MatSetOption(a->B,op,flg);
1789: break;
1790: case MAT_NEW_DIAGONALS:
1791: PetscInfo1(A,"Option %s ignored\n",MatOptions[op]);
1792: break;
1793: case MAT_IGNORE_OFF_PROC_ENTRIES:
1794: a->donotstash = flg;
1795: break;
1796: /* Symmetry flags are handled directly by MatSetOption() and they don't affect preallocation */
1797: case MAT_SPD:
1798: case MAT_SYMMETRIC:
1799: case MAT_STRUCTURALLY_SYMMETRIC:
1800: case MAT_HERMITIAN:
1801: case MAT_SYMMETRY_ETERNAL:
1802: break;
1803: case MAT_SUBMAT_SINGLEIS:
1804: A->submat_singleis = flg;
1805: break;
1806: case MAT_STRUCTURE_ONLY:
1807: /* The option is handled directly by MatSetOption() */
1808: break;
1809: default:
1810: SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"unknown option %d",op);
1811: }
1812: return(0);
1813: }
1815: PetscErrorCode MatGetRow_MPIAIJ(Mat matin,PetscInt row,PetscInt *nz,PetscInt **idx,PetscScalar **v)
1816: {
1817: Mat_MPIAIJ *mat = (Mat_MPIAIJ*)matin->data;
1818: PetscScalar *vworkA,*vworkB,**pvA,**pvB,*v_p;
1820: PetscInt i,*cworkA,*cworkB,**pcA,**pcB,cstart = matin->cmap->rstart;
1821: PetscInt nztot,nzA,nzB,lrow,rstart = matin->rmap->rstart,rend = matin->rmap->rend;
1822: PetscInt *cmap,*idx_p;
1825: if (mat->getrowactive) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Already active");
1826: mat->getrowactive = PETSC_TRUE;
1828: if (!mat->rowvalues && (idx || v)) {
1829: /*
1830: allocate enough space to hold information from the longest row.
1831: */
1832: Mat_SeqAIJ *Aa = (Mat_SeqAIJ*)mat->A->data,*Ba = (Mat_SeqAIJ*)mat->B->data;
1833: PetscInt max = 1,tmp;
1834: for (i=0; i<matin->rmap->n; i++) {
1835: tmp = Aa->i[i+1] - Aa->i[i] + Ba->i[i+1] - Ba->i[i];
1836: if (max < tmp) max = tmp;
1837: }
1838: PetscMalloc2(max,&mat->rowvalues,max,&mat->rowindices);
1839: }
1841: if (row < rstart || row >= rend) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Only local rows");
1842: lrow = row - rstart;
1844: pvA = &vworkA; pcA = &cworkA; pvB = &vworkB; pcB = &cworkB;
1845: if (!v) {pvA = 0; pvB = 0;}
1846: if (!idx) {pcA = 0; if (!v) pcB = 0;}
1847: (*mat->A->ops->getrow)(mat->A,lrow,&nzA,pcA,pvA);
1848: (*mat->B->ops->getrow)(mat->B,lrow,&nzB,pcB,pvB);
1849: nztot = nzA + nzB;
1851: cmap = mat->garray;
1852: if (v || idx) {
1853: if (nztot) {
1854: /* Sort by increasing column numbers, assuming A and B already sorted */
1855: PetscInt imark = -1;
1856: if (v) {
1857: *v = v_p = mat->rowvalues;
1858: for (i=0; i<nzB; i++) {
1859: if (cmap[cworkB[i]] < cstart) v_p[i] = vworkB[i];
1860: else break;
1861: }
1862: imark = i;
1863: for (i=0; i<nzA; i++) v_p[imark+i] = vworkA[i];
1864: for (i=imark; i<nzB; i++) v_p[nzA+i] = vworkB[i];
1865: }
1866: if (idx) {
1867: *idx = idx_p = mat->rowindices;
1868: if (imark > -1) {
1869: for (i=0; i<imark; i++) {
1870: idx_p[i] = cmap[cworkB[i]];
1871: }
1872: } else {
1873: for (i=0; i<nzB; i++) {
1874: if (cmap[cworkB[i]] < cstart) idx_p[i] = cmap[cworkB[i]];
1875: else break;
1876: }
1877: imark = i;
1878: }
1879: for (i=0; i<nzA; i++) idx_p[imark+i] = cstart + cworkA[i];
1880: for (i=imark; i<nzB; i++) idx_p[nzA+i] = cmap[cworkB[i]];
1881: }
1882: } else {
1883: if (idx) *idx = 0;
1884: if (v) *v = 0;
1885: }
1886: }
1887: *nz = nztot;
1888: (*mat->A->ops->restorerow)(mat->A,lrow,&nzA,pcA,pvA);
1889: (*mat->B->ops->restorerow)(mat->B,lrow,&nzB,pcB,pvB);
1890: return(0);
1891: }
1893: PetscErrorCode MatRestoreRow_MPIAIJ(Mat mat,PetscInt row,PetscInt *nz,PetscInt **idx,PetscScalar **v)
1894: {
1895: Mat_MPIAIJ *aij = (Mat_MPIAIJ*)mat->data;
1898: if (!aij->getrowactive) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"MatGetRow() must be called first");
1899: aij->getrowactive = PETSC_FALSE;
1900: return(0);
1901: }
1903: PetscErrorCode MatNorm_MPIAIJ(Mat mat,NormType type,PetscReal *norm)
1904: {
1905: Mat_MPIAIJ *aij = (Mat_MPIAIJ*)mat->data;
1906: Mat_SeqAIJ *amat = (Mat_SeqAIJ*)aij->A->data,*bmat = (Mat_SeqAIJ*)aij->B->data;
1908: PetscInt i,j,cstart = mat->cmap->rstart;
1909: PetscReal sum = 0.0;
1910: MatScalar *v;
1913: if (aij->size == 1) {
1914: MatNorm(aij->A,type,norm);
1915: } else {
1916: if (type == NORM_FROBENIUS) {
1917: v = amat->a;
1918: for (i=0; i<amat->nz; i++) {
1919: sum += PetscRealPart(PetscConj(*v)*(*v)); v++;
1920: }
1921: v = bmat->a;
1922: for (i=0; i<bmat->nz; i++) {
1923: sum += PetscRealPart(PetscConj(*v)*(*v)); v++;
1924: }
1925: MPIU_Allreduce(&sum,norm,1,MPIU_REAL,MPIU_SUM,PetscObjectComm((PetscObject)mat));
1926: *norm = PetscSqrtReal(*norm);
1927: PetscLogFlops(2*amat->nz+2*bmat->nz);
1928: } else if (type == NORM_1) { /* max column norm */
1929: PetscReal *tmp,*tmp2;
1930: PetscInt *jj,*garray = aij->garray;
1931: PetscCalloc1(mat->cmap->N+1,&tmp);
1932: PetscMalloc1(mat->cmap->N+1,&tmp2);
1933: *norm = 0.0;
1934: v = amat->a; jj = amat->j;
1935: for (j=0; j<amat->nz; j++) {
1936: tmp[cstart + *jj++] += PetscAbsScalar(*v); v++;
1937: }
1938: v = bmat->a; jj = bmat->j;
1939: for (j=0; j<bmat->nz; j++) {
1940: tmp[garray[*jj++]] += PetscAbsScalar(*v); v++;
1941: }
1942: MPIU_Allreduce(tmp,tmp2,mat->cmap->N,MPIU_REAL,MPIU_SUM,PetscObjectComm((PetscObject)mat));
1943: for (j=0; j<mat->cmap->N; j++) {
1944: if (tmp2[j] > *norm) *norm = tmp2[j];
1945: }
1946: PetscFree(tmp);
1947: PetscFree(tmp2);
1948: PetscLogFlops(PetscMax(amat->nz+bmat->nz-1,0));
1949: } else if (type == NORM_INFINITY) { /* max row norm */
1950: PetscReal ntemp = 0.0;
1951: for (j=0; j<aij->A->rmap->n; j++) {
1952: v = amat->a + amat->i[j];
1953: sum = 0.0;
1954: for (i=0; i<amat->i[j+1]-amat->i[j]; i++) {
1955: sum += PetscAbsScalar(*v); v++;
1956: }
1957: v = bmat->a + bmat->i[j];
1958: for (i=0; i<bmat->i[j+1]-bmat->i[j]; i++) {
1959: sum += PetscAbsScalar(*v); v++;
1960: }
1961: if (sum > ntemp) ntemp = sum;
1962: }
1963: MPIU_Allreduce(&ntemp,norm,1,MPIU_REAL,MPIU_MAX,PetscObjectComm((PetscObject)mat));
1964: PetscLogFlops(PetscMax(amat->nz+bmat->nz-1,0));
1965: } else SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"No support for two norm");
1966: }
1967: return(0);
1968: }
1970: PetscErrorCode MatTranspose_MPIAIJ(Mat A,MatReuse reuse,Mat *matout)
1971: {
1972: Mat_MPIAIJ *a =(Mat_MPIAIJ*)A->data,*b;
1973: Mat_SeqAIJ *Aloc =(Mat_SeqAIJ*)a->A->data,*Bloc=(Mat_SeqAIJ*)a->B->data,*sub_B_diag;
1974: PetscInt M = A->rmap->N,N=A->cmap->N,ma,na,mb,nb,*ai,*aj,*bi,*bj,row,*cols,*cols_tmp,*B_diag_ilen,*B_diag_i,i,ncol,A_diag_ncol;
1976: Mat B,A_diag,*B_diag;
1977: MatScalar *array;
1980: ma = A->rmap->n; na = A->cmap->n; mb = a->B->rmap->n; nb = a->B->cmap->n;
1981: ai = Aloc->i; aj = Aloc->j;
1982: bi = Bloc->i; bj = Bloc->j;
1983: if (reuse == MAT_INITIAL_MATRIX || *matout == A) {
1984: PetscInt *d_nnz,*g_nnz,*o_nnz;
1985: PetscSFNode *oloc;
1986: PETSC_UNUSED PetscSF sf;
1988: PetscMalloc4(na,&d_nnz,na,&o_nnz,nb,&g_nnz,nb,&oloc);
1989: /* compute d_nnz for preallocation */
1990: PetscMemzero(d_nnz,na*sizeof(PetscInt));
1991: for (i=0; i<ai[ma]; i++) {
1992: d_nnz[aj[i]]++;
1993: }
1994: /* compute local off-diagonal contributions */
1995: PetscMemzero(g_nnz,nb*sizeof(PetscInt));
1996: for (i=0; i<bi[ma]; i++) g_nnz[bj[i]]++;
1997: /* map those to global */
1998: PetscSFCreate(PetscObjectComm((PetscObject)A),&sf);
1999: PetscSFSetGraphLayout(sf,A->cmap,nb,NULL,PETSC_USE_POINTER,a->garray);
2000: PetscSFSetFromOptions(sf);
2001: PetscMemzero(o_nnz,na*sizeof(PetscInt));
2002: PetscSFReduceBegin(sf,MPIU_INT,g_nnz,o_nnz,MPIU_SUM);
2003: PetscSFReduceEnd(sf,MPIU_INT,g_nnz,o_nnz,MPIU_SUM);
2004: PetscSFDestroy(&sf);
2006: MatCreate(PetscObjectComm((PetscObject)A),&B);
2007: MatSetSizes(B,A->cmap->n,A->rmap->n,N,M);
2008: MatSetBlockSizes(B,PetscAbs(A->cmap->bs),PetscAbs(A->rmap->bs));
2009: MatSetType(B,((PetscObject)A)->type_name);
2010: MatMPIAIJSetPreallocation(B,0,d_nnz,0,o_nnz);
2011: PetscFree4(d_nnz,o_nnz,g_nnz,oloc);
2012: } else {
2013: B = *matout;
2014: MatSetOption(B,MAT_NEW_NONZERO_ALLOCATION_ERR,PETSC_TRUE);
2015: }
2017: b = (Mat_MPIAIJ*)B->data;
2018: A_diag = a->A;
2019: B_diag = &b->A;
2020: sub_B_diag = (Mat_SeqAIJ*)(*B_diag)->data;
2021: A_diag_ncol = A_diag->cmap->N;
2022: B_diag_ilen = sub_B_diag->ilen;
2023: B_diag_i = sub_B_diag->i;
2025: /* Set ilen for diagonal of B */
2026: for (i=0; i<A_diag_ncol; i++) {
2027: B_diag_ilen[i] = B_diag_i[i+1] - B_diag_i[i];
2028: }
2030: /* Transpose the diagonal part of the matrix. In contrast to the offdiagonal part, this can be done
2031: very quickly (=without using MatSetValues), because all writes are local. */
2032: MatTranspose(A_diag,MAT_REUSE_MATRIX,B_diag);
2034: /* copy over the B part */
2035: PetscCalloc1(bi[mb],&cols);
2036: array = Bloc->a;
2037: row = A->rmap->rstart;
2038: for (i=0; i<bi[mb]; i++) cols[i] = a->garray[bj[i]];
2039: cols_tmp = cols;
2040: for (i=0; i<mb; i++) {
2041: ncol = bi[i+1]-bi[i];
2042: MatSetValues(B,ncol,cols_tmp,1,&row,array,INSERT_VALUES);
2043: row++;
2044: array += ncol; cols_tmp += ncol;
2045: }
2046: PetscFree(cols);
2048: MatAssemblyBegin(B,MAT_FINAL_ASSEMBLY);
2049: MatAssemblyEnd(B,MAT_FINAL_ASSEMBLY);
2050: if (reuse == MAT_INITIAL_MATRIX || reuse == MAT_REUSE_MATRIX) {
2051: *matout = B;
2052: } else {
2053: MatHeaderMerge(A,&B);
2054: }
2055: return(0);
2056: }
2058: PetscErrorCode MatDiagonalScale_MPIAIJ(Mat mat,Vec ll,Vec rr)
2059: {
2060: Mat_MPIAIJ *aij = (Mat_MPIAIJ*)mat->data;
2061: Mat a = aij->A,b = aij->B;
2063: PetscInt s1,s2,s3;
2066: MatGetLocalSize(mat,&s2,&s3);
2067: if (rr) {
2068: VecGetLocalSize(rr,&s1);
2069: if (s1!=s3) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"right vector non-conforming local size");
2070: /* Overlap communication with computation. */
2071: VecScatterBegin(aij->Mvctx,rr,aij->lvec,INSERT_VALUES,SCATTER_FORWARD);
2072: }
2073: if (ll) {
2074: VecGetLocalSize(ll,&s1);
2075: if (s1!=s2) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"left vector non-conforming local size");
2076: (*b->ops->diagonalscale)(b,ll,0);
2077: }
2078: /* scale the diagonal block */
2079: (*a->ops->diagonalscale)(a,ll,rr);
2081: if (rr) {
2082: /* Do a scatter end and then right scale the off-diagonal block */
2083: VecScatterEnd(aij->Mvctx,rr,aij->lvec,INSERT_VALUES,SCATTER_FORWARD);
2084: (*b->ops->diagonalscale)(b,0,aij->lvec);
2085: }
2086: return(0);
2087: }
2089: PetscErrorCode MatSetUnfactored_MPIAIJ(Mat A)
2090: {
2091: Mat_MPIAIJ *a = (Mat_MPIAIJ*)A->data;
2095: MatSetUnfactored(a->A);
2096: return(0);
2097: }
2099: PetscErrorCode MatEqual_MPIAIJ(Mat A,Mat B,PetscBool *flag)
2100: {
2101: Mat_MPIAIJ *matB = (Mat_MPIAIJ*)B->data,*matA = (Mat_MPIAIJ*)A->data;
2102: Mat a,b,c,d;
2103: PetscBool flg;
2107: a = matA->A; b = matA->B;
2108: c = matB->A; d = matB->B;
2110: MatEqual(a,c,&flg);
2111: if (flg) {
2112: MatEqual(b,d,&flg);
2113: }
2114: MPIU_Allreduce(&flg,flag,1,MPIU_BOOL,MPI_LAND,PetscObjectComm((PetscObject)A));
2115: return(0);
2116: }
2118: PetscErrorCode MatCopy_MPIAIJ(Mat A,Mat B,MatStructure str)
2119: {
2121: Mat_MPIAIJ *a = (Mat_MPIAIJ*)A->data;
2122: Mat_MPIAIJ *b = (Mat_MPIAIJ*)B->data;
2125: /* If the two matrices don't have the same copy implementation, they aren't compatible for fast copy. */
2126: if ((str != SAME_NONZERO_PATTERN) || (A->ops->copy != B->ops->copy)) {
2127: /* because of the column compression in the off-processor part of the matrix a->B,
2128: the number of columns in a->B and b->B may be different, hence we cannot call
2129: the MatCopy() directly on the two parts. If need be, we can provide a more
2130: efficient copy than the MatCopy_Basic() by first uncompressing the a->B matrices
2131: then copying the submatrices */
2132: MatCopy_Basic(A,B,str);
2133: } else {
2134: MatCopy(a->A,b->A,str);
2135: MatCopy(a->B,b->B,str);
2136: }
2137: PetscObjectStateIncrease((PetscObject)B);
2138: return(0);
2139: }
2141: PetscErrorCode MatSetUp_MPIAIJ(Mat A)
2142: {
2146: MatMPIAIJSetPreallocation(A,PETSC_DEFAULT,0,PETSC_DEFAULT,0);
2147: return(0);
2148: }
2150: /*
2151: Computes the number of nonzeros per row needed for preallocation when X and Y
2152: have different nonzero structure.
2153: */
2154: PetscErrorCode MatAXPYGetPreallocation_MPIX_private(PetscInt m,const PetscInt *xi,const PetscInt *xj,const PetscInt *xltog,const PetscInt *yi,const PetscInt *yj,const PetscInt *yltog,PetscInt *nnz)
2155: {
2156: PetscInt i,j,k,nzx,nzy;
2159: /* Set the number of nonzeros in the new matrix */
2160: for (i=0; i<m; i++) {
2161: const PetscInt *xjj = xj+xi[i],*yjj = yj+yi[i];
2162: nzx = xi[i+1] - xi[i];
2163: nzy = yi[i+1] - yi[i];
2164: nnz[i] = 0;
2165: for (j=0,k=0; j<nzx; j++) { /* Point in X */
2166: for (; k<nzy && yltog[yjj[k]]<xltog[xjj[j]]; k++) nnz[i]++; /* Catch up to X */
2167: if (k<nzy && yltog[yjj[k]]==xltog[xjj[j]]) k++; /* Skip duplicate */
2168: nnz[i]++;
2169: }
2170: for (; k<nzy; k++) nnz[i]++;
2171: }
2172: return(0);
2173: }
2175: /* This is the same as MatAXPYGetPreallocation_SeqAIJ, except that the local-to-global map is provided */
2176: static PetscErrorCode MatAXPYGetPreallocation_MPIAIJ(Mat Y,const PetscInt *yltog,Mat X,const PetscInt *xltog,PetscInt *nnz)
2177: {
2179: PetscInt m = Y->rmap->N;
2180: Mat_SeqAIJ *x = (Mat_SeqAIJ*)X->data;
2181: Mat_SeqAIJ *y = (Mat_SeqAIJ*)Y->data;
2184: MatAXPYGetPreallocation_MPIX_private(m,x->i,x->j,xltog,y->i,y->j,yltog,nnz);
2185: return(0);
2186: }
2188: PetscErrorCode MatAXPY_MPIAIJ(Mat Y,PetscScalar a,Mat X,MatStructure str)
2189: {
2191: Mat_MPIAIJ *xx = (Mat_MPIAIJ*)X->data,*yy = (Mat_MPIAIJ*)Y->data;
2192: PetscBLASInt bnz,one=1;
2193: Mat_SeqAIJ *x,*y;
2196: if (str == SAME_NONZERO_PATTERN) {
2197: PetscScalar alpha = a;
2198: x = (Mat_SeqAIJ*)xx->A->data;
2199: PetscBLASIntCast(x->nz,&bnz);
2200: y = (Mat_SeqAIJ*)yy->A->data;
2201: PetscStackCallBLAS("BLASaxpy",BLASaxpy_(&bnz,&alpha,x->a,&one,y->a,&one));
2202: x = (Mat_SeqAIJ*)xx->B->data;
2203: y = (Mat_SeqAIJ*)yy->B->data;
2204: PetscBLASIntCast(x->nz,&bnz);
2205: PetscStackCallBLAS("BLASaxpy",BLASaxpy_(&bnz,&alpha,x->a,&one,y->a,&one));
2206: PetscObjectStateIncrease((PetscObject)Y);
2207: } else if (str == SUBSET_NONZERO_PATTERN) { /* nonzeros of X is a subset of Y's */
2208: MatAXPY_Basic(Y,a,X,str);
2209: } else {
2210: Mat B;
2211: PetscInt *nnz_d,*nnz_o;
2212: PetscMalloc1(yy->A->rmap->N,&nnz_d);
2213: PetscMalloc1(yy->B->rmap->N,&nnz_o);
2214: MatCreate(PetscObjectComm((PetscObject)Y),&B);
2215: PetscObjectSetName((PetscObject)B,((PetscObject)Y)->name);
2216: MatSetSizes(B,Y->rmap->n,Y->cmap->n,Y->rmap->N,Y->cmap->N);
2217: MatSetBlockSizesFromMats(B,Y,Y);
2218: MatSetType(B,MATMPIAIJ);
2219: MatAXPYGetPreallocation_SeqAIJ(yy->A,xx->A,nnz_d);
2220: MatAXPYGetPreallocation_MPIAIJ(yy->B,yy->garray,xx->B,xx->garray,nnz_o);
2221: MatMPIAIJSetPreallocation(B,0,nnz_d,0,nnz_o);
2222: MatAXPY_BasicWithPreallocation(B,Y,a,X,str);
2223: MatHeaderReplace(Y,&B);
2224: PetscFree(nnz_d);
2225: PetscFree(nnz_o);
2226: }
2227: return(0);
2228: }
2230: extern PetscErrorCode MatConjugate_SeqAIJ(Mat);
2232: PetscErrorCode MatConjugate_MPIAIJ(Mat mat)
2233: {
2234: #if defined(PETSC_USE_COMPLEX)
2236: Mat_MPIAIJ *aij = (Mat_MPIAIJ*)mat->data;
2239: MatConjugate_SeqAIJ(aij->A);
2240: MatConjugate_SeqAIJ(aij->B);
2241: #else
2243: #endif
2244: return(0);
2245: }
2247: PetscErrorCode MatRealPart_MPIAIJ(Mat A)
2248: {
2249: Mat_MPIAIJ *a = (Mat_MPIAIJ*)A->data;
2253: MatRealPart(a->A);
2254: MatRealPart(a->B);
2255: return(0);
2256: }
2258: PetscErrorCode MatImaginaryPart_MPIAIJ(Mat A)
2259: {
2260: Mat_MPIAIJ *a = (Mat_MPIAIJ*)A->data;
2264: MatImaginaryPart(a->A);
2265: MatImaginaryPart(a->B);
2266: return(0);
2267: }
2269: PetscErrorCode MatGetRowMaxAbs_MPIAIJ(Mat A, Vec v, PetscInt idx[])
2270: {
2271: Mat_MPIAIJ *a = (Mat_MPIAIJ*)A->data;
2273: PetscInt i,*idxb = 0;
2274: PetscScalar *va,*vb;
2275: Vec vtmp;
2278: MatGetRowMaxAbs(a->A,v,idx);
2279: VecGetArray(v,&va);
2280: if (idx) {
2281: for (i=0; i<A->rmap->n; i++) {
2282: if (PetscAbsScalar(va[i])) idx[i] += A->cmap->rstart;
2283: }
2284: }
2286: VecCreateSeq(PETSC_COMM_SELF,A->rmap->n,&vtmp);
2287: if (idx) {
2288: PetscMalloc1(A->rmap->n,&idxb);
2289: }
2290: MatGetRowMaxAbs(a->B,vtmp,idxb);
2291: VecGetArray(vtmp,&vb);
2293: for (i=0; i<A->rmap->n; i++) {
2294: if (PetscAbsScalar(va[i]) < PetscAbsScalar(vb[i])) {
2295: va[i] = vb[i];
2296: if (idx) idx[i] = a->garray[idxb[i]];
2297: }
2298: }
2300: VecRestoreArray(v,&va);
2301: VecRestoreArray(vtmp,&vb);
2302: PetscFree(idxb);
2303: VecDestroy(&vtmp);
2304: return(0);
2305: }
2307: PetscErrorCode MatGetRowMinAbs_MPIAIJ(Mat A, Vec v, PetscInt idx[])
2308: {
2309: Mat_MPIAIJ *a = (Mat_MPIAIJ*)A->data;
2311: PetscInt i,*idxb = 0;
2312: PetscScalar *va,*vb;
2313: Vec vtmp;
2316: MatGetRowMinAbs(a->A,v,idx);
2317: VecGetArray(v,&va);
2318: if (idx) {
2319: for (i=0; i<A->cmap->n; i++) {
2320: if (PetscAbsScalar(va[i])) idx[i] += A->cmap->rstart;
2321: }
2322: }
2324: VecCreateSeq(PETSC_COMM_SELF,A->rmap->n,&vtmp);
2325: if (idx) {
2326: PetscMalloc1(A->rmap->n,&idxb);
2327: }
2328: MatGetRowMinAbs(a->B,vtmp,idxb);
2329: VecGetArray(vtmp,&vb);
2331: for (i=0; i<A->rmap->n; i++) {
2332: if (PetscAbsScalar(va[i]) > PetscAbsScalar(vb[i])) {
2333: va[i] = vb[i];
2334: if (idx) idx[i] = a->garray[idxb[i]];
2335: }
2336: }
2338: VecRestoreArray(v,&va);
2339: VecRestoreArray(vtmp,&vb);
2340: PetscFree(idxb);
2341: VecDestroy(&vtmp);
2342: return(0);
2343: }
2345: PetscErrorCode MatGetRowMin_MPIAIJ(Mat A, Vec v, PetscInt idx[])
2346: {
2347: Mat_MPIAIJ *mat = (Mat_MPIAIJ*) A->data;
2348: PetscInt n = A->rmap->n;
2349: PetscInt cstart = A->cmap->rstart;
2350: PetscInt *cmap = mat->garray;
2351: PetscInt *diagIdx, *offdiagIdx;
2352: Vec diagV, offdiagV;
2353: PetscScalar *a, *diagA, *offdiagA;
2354: PetscInt r;
2358: PetscMalloc2(n,&diagIdx,n,&offdiagIdx);
2359: VecCreateSeq(PetscObjectComm((PetscObject)A), n, &diagV);
2360: VecCreateSeq(PetscObjectComm((PetscObject)A), n, &offdiagV);
2361: MatGetRowMin(mat->A, diagV, diagIdx);
2362: MatGetRowMin(mat->B, offdiagV, offdiagIdx);
2363: VecGetArray(v, &a);
2364: VecGetArray(diagV, &diagA);
2365: VecGetArray(offdiagV, &offdiagA);
2366: for (r = 0; r < n; ++r) {
2367: if (PetscAbsScalar(diagA[r]) <= PetscAbsScalar(offdiagA[r])) {
2368: a[r] = diagA[r];
2369: idx[r] = cstart + diagIdx[r];
2370: } else {
2371: a[r] = offdiagA[r];
2372: idx[r] = cmap[offdiagIdx[r]];
2373: }
2374: }
2375: VecRestoreArray(v, &a);
2376: VecRestoreArray(diagV, &diagA);
2377: VecRestoreArray(offdiagV, &offdiagA);
2378: VecDestroy(&diagV);
2379: VecDestroy(&offdiagV);
2380: PetscFree2(diagIdx, offdiagIdx);
2381: return(0);
2382: }
2384: PetscErrorCode MatGetRowMax_MPIAIJ(Mat A, Vec v, PetscInt idx[])
2385: {
2386: Mat_MPIAIJ *mat = (Mat_MPIAIJ*) A->data;
2387: PetscInt n = A->rmap->n;
2388: PetscInt cstart = A->cmap->rstart;
2389: PetscInt *cmap = mat->garray;
2390: PetscInt *diagIdx, *offdiagIdx;
2391: Vec diagV, offdiagV;
2392: PetscScalar *a, *diagA, *offdiagA;
2393: PetscInt r;
2397: PetscMalloc2(n,&diagIdx,n,&offdiagIdx);
2398: VecCreateSeq(PETSC_COMM_SELF, n, &diagV);
2399: VecCreateSeq(PETSC_COMM_SELF, n, &offdiagV);
2400: MatGetRowMax(mat->A, diagV, diagIdx);
2401: MatGetRowMax(mat->B, offdiagV, offdiagIdx);
2402: VecGetArray(v, &a);
2403: VecGetArray(diagV, &diagA);
2404: VecGetArray(offdiagV, &offdiagA);
2405: for (r = 0; r < n; ++r) {
2406: if (PetscAbsScalar(diagA[r]) >= PetscAbsScalar(offdiagA[r])) {
2407: a[r] = diagA[r];
2408: idx[r] = cstart + diagIdx[r];
2409: } else {
2410: a[r] = offdiagA[r];
2411: idx[r] = cmap[offdiagIdx[r]];
2412: }
2413: }
2414: VecRestoreArray(v, &a);
2415: VecRestoreArray(diagV, &diagA);
2416: VecRestoreArray(offdiagV, &offdiagA);
2417: VecDestroy(&diagV);
2418: VecDestroy(&offdiagV);
2419: PetscFree2(diagIdx, offdiagIdx);
2420: return(0);
2421: }
2423: PetscErrorCode MatGetSeqNonzeroStructure_MPIAIJ(Mat mat,Mat *newmat)
2424: {
2426: Mat *dummy;
2429: MatCreateSubMatrix_MPIAIJ_All(mat,MAT_DO_NOT_GET_VALUES,MAT_INITIAL_MATRIX,&dummy);
2430: *newmat = *dummy;
2431: PetscFree(dummy);
2432: return(0);
2433: }
2435: PetscErrorCode MatInvertBlockDiagonal_MPIAIJ(Mat A,const PetscScalar **values)
2436: {
2437: Mat_MPIAIJ *a = (Mat_MPIAIJ*) A->data;
2441: MatInvertBlockDiagonal(a->A,values);
2442: A->factorerrortype = a->A->factorerrortype;
2443: return(0);
2444: }
2446: static PetscErrorCode MatSetRandom_MPIAIJ(Mat x,PetscRandom rctx)
2447: {
2449: Mat_MPIAIJ *aij = (Mat_MPIAIJ*)x->data;
2452: MatSetRandom(aij->A,rctx);
2453: MatSetRandom(aij->B,rctx);
2454: MatAssemblyBegin(x,MAT_FINAL_ASSEMBLY);
2455: MatAssemblyEnd(x,MAT_FINAL_ASSEMBLY);
2456: return(0);
2457: }
2459: PetscErrorCode MatMPIAIJSetUseScalableIncreaseOverlap_MPIAIJ(Mat A,PetscBool sc)
2460: {
2462: if (sc) A->ops->increaseoverlap = MatIncreaseOverlap_MPIAIJ_Scalable;
2463: else A->ops->increaseoverlap = MatIncreaseOverlap_MPIAIJ;
2464: return(0);
2465: }
2467: /*@
2468: MatMPIAIJSetUseScalableIncreaseOverlap - Determine if the matrix uses a scalable algorithm to compute the overlap
2470: Collective on Mat
2472: Input Parameters:
2473: + A - the matrix
2474: - sc - PETSC_TRUE indicates use the scalable algorithm (default is not to use the scalable algorithm)
2476: Level: advanced
2478: @*/
2479: PetscErrorCode MatMPIAIJSetUseScalableIncreaseOverlap(Mat A,PetscBool sc)
2480: {
2481: PetscErrorCode ierr;
2484: PetscTryMethod(A,"MatMPIAIJSetUseScalableIncreaseOverlap_C",(Mat,PetscBool),(A,sc));
2485: return(0);
2486: }
2488: PetscErrorCode MatSetFromOptions_MPIAIJ(PetscOptionItems *PetscOptionsObject,Mat A)
2489: {
2490: PetscErrorCode ierr;
2491: PetscBool sc = PETSC_FALSE,flg;
2494: PetscOptionsHead(PetscOptionsObject,"MPIAIJ options");
2495: if (A->ops->increaseoverlap == MatIncreaseOverlap_MPIAIJ_Scalable) sc = PETSC_TRUE;
2496: PetscOptionsBool("-mat_increase_overlap_scalable","Use a scalable algorithm to compute the overlap","MatIncreaseOverlap",sc,&sc,&flg);
2497: if (flg) {
2498: MatMPIAIJSetUseScalableIncreaseOverlap(A,sc);
2499: }
2500: PetscOptionsTail();
2501: return(0);
2502: }
2504: PetscErrorCode MatShift_MPIAIJ(Mat Y,PetscScalar a)
2505: {
2507: Mat_MPIAIJ *maij = (Mat_MPIAIJ*)Y->data;
2508: Mat_SeqAIJ *aij = (Mat_SeqAIJ*)maij->A->data;
2511: if (!Y->preallocated) {
2512: MatMPIAIJSetPreallocation(Y,1,NULL,0,NULL);
2513: } else if (!aij->nz) {
2514: PetscInt nonew = aij->nonew;
2515: MatSeqAIJSetPreallocation(maij->A,1,NULL);
2516: aij->nonew = nonew;
2517: }
2518: MatShift_Basic(Y,a);
2519: return(0);
2520: }
2522: PetscErrorCode MatMissingDiagonal_MPIAIJ(Mat A,PetscBool *missing,PetscInt *d)
2523: {
2524: Mat_MPIAIJ *a = (Mat_MPIAIJ*)A->data;
2528: if (A->rmap->n != A->cmap->n) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Only works for square matrices");
2529: MatMissingDiagonal(a->A,missing,d);
2530: if (d) {
2531: PetscInt rstart;
2532: MatGetOwnershipRange(A,&rstart,NULL);
2533: *d += rstart;
2535: }
2536: return(0);
2537: }
2539: PetscErrorCode MatInvertVariableBlockDiagonal_MPIAIJ(Mat A,PetscInt nblocks,const PetscInt *bsizes,PetscScalar *diag)
2540: {
2541: Mat_MPIAIJ *a = (Mat_MPIAIJ*)A->data;
2545: MatInvertVariableBlockDiagonal(a->A,nblocks,bsizes,diag);
2546: return(0);
2547: }
2549: /* -------------------------------------------------------------------*/
2550: static struct _MatOps MatOps_Values = {MatSetValues_MPIAIJ,
2551: MatGetRow_MPIAIJ,
2552: MatRestoreRow_MPIAIJ,
2553: MatMult_MPIAIJ,
2554: /* 4*/ MatMultAdd_MPIAIJ,
2555: MatMultTranspose_MPIAIJ,
2556: MatMultTransposeAdd_MPIAIJ,
2557: 0,
2558: 0,
2559: 0,
2560: /*10*/ 0,
2561: 0,
2562: 0,
2563: MatSOR_MPIAIJ,
2564: MatTranspose_MPIAIJ,
2565: /*15*/ MatGetInfo_MPIAIJ,
2566: MatEqual_MPIAIJ,
2567: MatGetDiagonal_MPIAIJ,
2568: MatDiagonalScale_MPIAIJ,
2569: MatNorm_MPIAIJ,
2570: /*20*/ MatAssemblyBegin_MPIAIJ,
2571: MatAssemblyEnd_MPIAIJ,
2572: MatSetOption_MPIAIJ,
2573: MatZeroEntries_MPIAIJ,
2574: /*24*/ MatZeroRows_MPIAIJ,
2575: 0,
2576: 0,
2577: 0,
2578: 0,
2579: /*29*/ MatSetUp_MPIAIJ,
2580: 0,
2581: 0,
2582: MatGetDiagonalBlock_MPIAIJ,
2583: 0,
2584: /*34*/ MatDuplicate_MPIAIJ,
2585: 0,
2586: 0,
2587: 0,
2588: 0,
2589: /*39*/ MatAXPY_MPIAIJ,
2590: MatCreateSubMatrices_MPIAIJ,
2591: MatIncreaseOverlap_MPIAIJ,
2592: MatGetValues_MPIAIJ,
2593: MatCopy_MPIAIJ,
2594: /*44*/ MatGetRowMax_MPIAIJ,
2595: MatScale_MPIAIJ,
2596: MatShift_MPIAIJ,
2597: MatDiagonalSet_MPIAIJ,
2598: MatZeroRowsColumns_MPIAIJ,
2599: /*49*/ MatSetRandom_MPIAIJ,
2600: 0,
2601: 0,
2602: 0,
2603: 0,
2604: /*54*/ MatFDColoringCreate_MPIXAIJ,
2605: 0,
2606: MatSetUnfactored_MPIAIJ,
2607: MatPermute_MPIAIJ,
2608: 0,
2609: /*59*/ MatCreateSubMatrix_MPIAIJ,
2610: MatDestroy_MPIAIJ,
2611: MatView_MPIAIJ,
2612: 0,
2613: MatMatMatMult_MPIAIJ_MPIAIJ_MPIAIJ,
2614: /*64*/ MatMatMatMultSymbolic_MPIAIJ_MPIAIJ_MPIAIJ,
2615: MatMatMatMultNumeric_MPIAIJ_MPIAIJ_MPIAIJ,
2616: 0,
2617: 0,
2618: 0,
2619: /*69*/ MatGetRowMaxAbs_MPIAIJ,
2620: MatGetRowMinAbs_MPIAIJ,
2621: 0,
2622: 0,
2623: 0,
2624: 0,
2625: /*75*/ MatFDColoringApply_AIJ,
2626: MatSetFromOptions_MPIAIJ,
2627: 0,
2628: 0,
2629: MatFindZeroDiagonals_MPIAIJ,
2630: /*80*/ 0,
2631: 0,
2632: 0,
2633: /*83*/ MatLoad_MPIAIJ,
2634: MatIsSymmetric_MPIAIJ,
2635: 0,
2636: 0,
2637: 0,
2638: 0,
2639: /*89*/ MatMatMult_MPIAIJ_MPIAIJ,
2640: MatMatMultSymbolic_MPIAIJ_MPIAIJ,
2641: MatMatMultNumeric_MPIAIJ_MPIAIJ,
2642: MatPtAP_MPIAIJ_MPIAIJ,
2643: MatPtAPSymbolic_MPIAIJ_MPIAIJ,
2644: /*94*/ MatPtAPNumeric_MPIAIJ_MPIAIJ,
2645: 0,
2646: 0,
2647: 0,
2648: 0,
2649: /*99*/ 0,
2650: 0,
2651: 0,
2652: MatConjugate_MPIAIJ,
2653: 0,
2654: /*104*/MatSetValuesRow_MPIAIJ,
2655: MatRealPart_MPIAIJ,
2656: MatImaginaryPart_MPIAIJ,
2657: 0,
2658: 0,
2659: /*109*/0,
2660: 0,
2661: MatGetRowMin_MPIAIJ,
2662: 0,
2663: MatMissingDiagonal_MPIAIJ,
2664: /*114*/MatGetSeqNonzeroStructure_MPIAIJ,
2665: 0,
2666: MatGetGhosts_MPIAIJ,
2667: 0,
2668: 0,
2669: /*119*/0,
2670: 0,
2671: 0,
2672: 0,
2673: MatGetMultiProcBlock_MPIAIJ,
2674: /*124*/MatFindNonzeroRows_MPIAIJ,
2675: MatGetColumnNorms_MPIAIJ,
2676: MatInvertBlockDiagonal_MPIAIJ,
2677: MatInvertVariableBlockDiagonal_MPIAIJ,
2678: MatCreateSubMatricesMPI_MPIAIJ,
2679: /*129*/0,
2680: MatTransposeMatMult_MPIAIJ_MPIAIJ,
2681: MatTransposeMatMultSymbolic_MPIAIJ_MPIAIJ,
2682: MatTransposeMatMultNumeric_MPIAIJ_MPIAIJ,
2683: 0,
2684: /*134*/0,
2685: 0,
2686: MatRARt_MPIAIJ_MPIAIJ,
2687: 0,
2688: 0,
2689: /*139*/MatSetBlockSizes_MPIAIJ,
2690: 0,
2691: 0,
2692: MatFDColoringSetUp_MPIXAIJ,
2693: MatFindOffBlockDiagonalEntries_MPIAIJ,
2694: /*144*/MatCreateMPIMatConcatenateSeqMat_MPIAIJ
2695: };
2697: /* ----------------------------------------------------------------------------------------*/
2699: PetscErrorCode MatStoreValues_MPIAIJ(Mat mat)
2700: {
2701: Mat_MPIAIJ *aij = (Mat_MPIAIJ*)mat->data;
2705: MatStoreValues(aij->A);
2706: MatStoreValues(aij->B);
2707: return(0);
2708: }
2710: PetscErrorCode MatRetrieveValues_MPIAIJ(Mat mat)
2711: {
2712: Mat_MPIAIJ *aij = (Mat_MPIAIJ*)mat->data;
2716: MatRetrieveValues(aij->A);
2717: MatRetrieveValues(aij->B);
2718: return(0);
2719: }
2721: PetscErrorCode MatMPIAIJSetPreallocation_MPIAIJ(Mat B,PetscInt d_nz,const PetscInt d_nnz[],PetscInt o_nz,const PetscInt o_nnz[])
2722: {
2723: Mat_MPIAIJ *b;
2727: PetscLayoutSetUp(B->rmap);
2728: PetscLayoutSetUp(B->cmap);
2729: b = (Mat_MPIAIJ*)B->data;
2731: #if defined(PETSC_USE_CTABLE)
2732: PetscTableDestroy(&b->colmap);
2733: #else
2734: PetscFree(b->colmap);
2735: #endif
2736: PetscFree(b->garray);
2737: VecDestroy(&b->lvec);
2738: VecScatterDestroy(&b->Mvctx);
2740: /* Because the B will have been resized we simply destroy it and create a new one each time */
2741: MatDestroy(&b->B);
2742: MatCreate(PETSC_COMM_SELF,&b->B);
2743: MatSetSizes(b->B,B->rmap->n,B->cmap->N,B->rmap->n,B->cmap->N);
2744: MatSetBlockSizesFromMats(b->B,B,B);
2745: MatSetType(b->B,MATSEQAIJ);
2746: PetscLogObjectParent((PetscObject)B,(PetscObject)b->B);
2748: if (!B->preallocated) {
2749: MatCreate(PETSC_COMM_SELF,&b->A);
2750: MatSetSizes(b->A,B->rmap->n,B->cmap->n,B->rmap->n,B->cmap->n);
2751: MatSetBlockSizesFromMats(b->A,B,B);
2752: MatSetType(b->A,MATSEQAIJ);
2753: PetscLogObjectParent((PetscObject)B,(PetscObject)b->A);
2754: }
2756: MatSeqAIJSetPreallocation(b->A,d_nz,d_nnz);
2757: MatSeqAIJSetPreallocation(b->B,o_nz,o_nnz);
2758: B->preallocated = PETSC_TRUE;
2759: B->was_assembled = PETSC_FALSE;
2760: B->assembled = PETSC_FALSE;
2761: return(0);
2762: }
2764: PetscErrorCode MatResetPreallocation_MPIAIJ(Mat B)
2765: {
2766: Mat_MPIAIJ *b;
2771: PetscLayoutSetUp(B->rmap);
2772: PetscLayoutSetUp(B->cmap);
2773: b = (Mat_MPIAIJ*)B->data;
2775: #if defined(PETSC_USE_CTABLE)
2776: PetscTableDestroy(&b->colmap);
2777: #else
2778: PetscFree(b->colmap);
2779: #endif
2780: PetscFree(b->garray);
2781: VecDestroy(&b->lvec);
2782: VecScatterDestroy(&b->Mvctx);
2784: MatResetPreallocation(b->A);
2785: MatResetPreallocation(b->B);
2786: B->preallocated = PETSC_TRUE;
2787: B->was_assembled = PETSC_FALSE;
2788: B->assembled = PETSC_FALSE;
2789: return(0);
2790: }
2792: PetscErrorCode MatDuplicate_MPIAIJ(Mat matin,MatDuplicateOption cpvalues,Mat *newmat)
2793: {
2794: Mat mat;
2795: Mat_MPIAIJ *a,*oldmat = (Mat_MPIAIJ*)matin->data;
2799: *newmat = 0;
2800: MatCreate(PetscObjectComm((PetscObject)matin),&mat);
2801: MatSetSizes(mat,matin->rmap->n,matin->cmap->n,matin->rmap->N,matin->cmap->N);
2802: MatSetBlockSizesFromMats(mat,matin,matin);
2803: MatSetType(mat,((PetscObject)matin)->type_name);
2804: a = (Mat_MPIAIJ*)mat->data;
2806: mat->factortype = matin->factortype;
2807: mat->assembled = PETSC_TRUE;
2808: mat->insertmode = NOT_SET_VALUES;
2809: mat->preallocated = PETSC_TRUE;
2811: a->size = oldmat->size;
2812: a->rank = oldmat->rank;
2813: a->donotstash = oldmat->donotstash;
2814: a->roworiented = oldmat->roworiented;
2815: a->rowindices = 0;
2816: a->rowvalues = 0;
2817: a->getrowactive = PETSC_FALSE;
2819: PetscLayoutReference(matin->rmap,&mat->rmap);
2820: PetscLayoutReference(matin->cmap,&mat->cmap);
2822: if (oldmat->colmap) {
2823: #if defined(PETSC_USE_CTABLE)
2824: PetscTableCreateCopy(oldmat->colmap,&a->colmap);
2825: #else
2826: PetscMalloc1(mat->cmap->N,&a->colmap);
2827: PetscLogObjectMemory((PetscObject)mat,(mat->cmap->N)*sizeof(PetscInt));
2828: PetscMemcpy(a->colmap,oldmat->colmap,(mat->cmap->N)*sizeof(PetscInt));
2829: #endif
2830: } else a->colmap = 0;
2831: if (oldmat->garray) {
2832: PetscInt len;
2833: len = oldmat->B->cmap->n;
2834: PetscMalloc1(len+1,&a->garray);
2835: PetscLogObjectMemory((PetscObject)mat,len*sizeof(PetscInt));
2836: if (len) { PetscMemcpy(a->garray,oldmat->garray,len*sizeof(PetscInt)); }
2837: } else a->garray = 0;
2839: VecDuplicate(oldmat->lvec,&a->lvec);
2840: PetscLogObjectParent((PetscObject)mat,(PetscObject)a->lvec);
2841: VecScatterCopy(oldmat->Mvctx,&a->Mvctx);
2842: PetscLogObjectParent((PetscObject)mat,(PetscObject)a->Mvctx);
2844: if (oldmat->Mvctx_mpi1) {
2845: VecScatterCopy(oldmat->Mvctx_mpi1,&a->Mvctx_mpi1);
2846: PetscLogObjectParent((PetscObject)mat,(PetscObject)a->Mvctx_mpi1);
2847: }
2849: MatDuplicate(oldmat->A,cpvalues,&a->A);
2850: PetscLogObjectParent((PetscObject)mat,(PetscObject)a->A);
2851: MatDuplicate(oldmat->B,cpvalues,&a->B);
2852: PetscLogObjectParent((PetscObject)mat,(PetscObject)a->B);
2853: PetscFunctionListDuplicate(((PetscObject)matin)->qlist,&((PetscObject)mat)->qlist);
2854: *newmat = mat;
2855: return(0);
2856: }
2858: PetscErrorCode MatLoad_MPIAIJ(Mat newMat, PetscViewer viewer)
2859: {
2860: PetscBool isbinary, ishdf5;
2866: /* force binary viewer to load .info file if it has not yet done so */
2867: PetscViewerSetUp(viewer);
2868: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERBINARY,&isbinary);
2869: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERHDF5, &ishdf5);
2870: if (isbinary) {
2871: MatLoad_MPIAIJ_Binary(newMat,viewer);
2872: } else if (ishdf5) {
2873: #if defined(PETSC_HAVE_HDF5)
2874: MatLoad_AIJ_HDF5(newMat,viewer);
2875: #else
2876: SETERRQ(PetscObjectComm((PetscObject)newMat),PETSC_ERR_SUP,"HDF5 not supported in this build.\nPlease reconfigure using --download-hdf5");
2877: #endif
2878: } else {
2879: SETERRQ2(PetscObjectComm((PetscObject)newMat),PETSC_ERR_SUP,"Viewer type %s not yet supported for reading %s matrices",((PetscObject)viewer)->type_name,((PetscObject)newMat)->type_name);
2880: }
2881: return(0);
2882: }
2884: PetscErrorCode MatLoad_MPIAIJ_Binary(Mat newMat, PetscViewer viewer)
2885: {
2886: PetscScalar *vals,*svals;
2887: MPI_Comm comm;
2889: PetscMPIInt rank,size,tag = ((PetscObject)viewer)->tag;
2890: PetscInt i,nz,j,rstart,rend,mmax,maxnz = 0;
2891: PetscInt header[4],*rowlengths = 0,M,N,m,*cols;
2892: PetscInt *ourlens = NULL,*procsnz = NULL,*offlens = NULL,jj,*mycols,*smycols;
2893: PetscInt cend,cstart,n,*rowners;
2894: int fd;
2895: PetscInt bs = newMat->rmap->bs;
2898: PetscObjectGetComm((PetscObject)viewer,&comm);
2899: MPI_Comm_size(comm,&size);
2900: MPI_Comm_rank(comm,&rank);
2901: PetscViewerBinaryGetDescriptor(viewer,&fd);
2902: if (!rank) {
2903: PetscBinaryRead(fd,(char*)header,4,PETSC_INT);
2904: if (header[0] != MAT_FILE_CLASSID) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_FILE_UNEXPECTED,"not matrix object");
2905: if (header[3] < 0) SETERRQ(PetscObjectComm((PetscObject)newMat),PETSC_ERR_FILE_UNEXPECTED,"Matrix stored in special format on disk,cannot load as MATMPIAIJ");
2906: }
2908: PetscOptionsBegin(comm,NULL,"Options for loading MATMPIAIJ matrix","Mat");
2909: PetscOptionsInt("-matload_block_size","Set the blocksize used to store the matrix","MatLoad",bs,&bs,NULL);
2910: PetscOptionsEnd();
2911: if (bs < 0) bs = 1;
2913: MPI_Bcast(header+1,3,MPIU_INT,0,comm);
2914: M = header[1]; N = header[2];
2916: /* If global sizes are set, check if they are consistent with that given in the file */
2917: if (newMat->rmap->N >= 0 && newMat->rmap->N != M) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_FILE_UNEXPECTED,"Inconsistent # of rows:Matrix in file has (%D) and input matrix has (%D)",newMat->rmap->N,M);
2918: if (newMat->cmap->N >=0 && newMat->cmap->N != N) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_FILE_UNEXPECTED,"Inconsistent # of cols:Matrix in file has (%D) and input matrix has (%D)",newMat->cmap->N,N);
2920: /* determine ownership of all (block) rows */
2921: if (M%bs) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_FILE_UNEXPECTED, "Inconsistent # of rows (%d) and block size (%d)",M,bs);
2922: if (newMat->rmap->n < 0) m = bs*((M/bs)/size + (((M/bs) % size) > rank)); /* PETSC_DECIDE */
2923: else m = newMat->rmap->n; /* Set by user */
2925: PetscMalloc1(size+1,&rowners);
2926: MPI_Allgather(&m,1,MPIU_INT,rowners+1,1,MPIU_INT,comm);
2928: /* First process needs enough room for process with most rows */
2929: if (!rank) {
2930: mmax = rowners[1];
2931: for (i=2; i<=size; i++) {
2932: mmax = PetscMax(mmax, rowners[i]);
2933: }
2934: } else mmax = -1; /* unused, but compilers complain */
2936: rowners[0] = 0;
2937: for (i=2; i<=size; i++) {
2938: rowners[i] += rowners[i-1];
2939: }
2940: rstart = rowners[rank];
2941: rend = rowners[rank+1];
2943: /* distribute row lengths to all processors */
2944: PetscMalloc2(m,&ourlens,m,&offlens);
2945: if (!rank) {
2946: PetscBinaryRead(fd,ourlens,m,PETSC_INT);
2947: PetscMalloc1(mmax,&rowlengths);
2948: PetscCalloc1(size,&procsnz);
2949: for (j=0; j<m; j++) {
2950: procsnz[0] += ourlens[j];
2951: }
2952: for (i=1; i<size; i++) {
2953: PetscBinaryRead(fd,rowlengths,rowners[i+1]-rowners[i],PETSC_INT);
2954: /* calculate the number of nonzeros on each processor */
2955: for (j=0; j<rowners[i+1]-rowners[i]; j++) {
2956: procsnz[i] += rowlengths[j];
2957: }
2958: MPIULong_Send(rowlengths,rowners[i+1]-rowners[i],MPIU_INT,i,tag,comm);
2959: }
2960: PetscFree(rowlengths);
2961: } else {
2962: MPIULong_Recv(ourlens,m,MPIU_INT,0,tag,comm);
2963: }
2965: if (!rank) {
2966: /* determine max buffer needed and allocate it */
2967: maxnz = 0;
2968: for (i=0; i<size; i++) {
2969: maxnz = PetscMax(maxnz,procsnz[i]);
2970: }
2971: PetscMalloc1(maxnz,&cols);
2973: /* read in my part of the matrix column indices */
2974: nz = procsnz[0];
2975: PetscMalloc1(nz,&mycols);
2976: PetscBinaryRead(fd,mycols,nz,PETSC_INT);
2978: /* read in every one elses and ship off */
2979: for (i=1; i<size; i++) {
2980: nz = procsnz[i];
2981: PetscBinaryRead(fd,cols,nz,PETSC_INT);
2982: MPIULong_Send(cols,nz,MPIU_INT,i,tag,comm);
2983: }
2984: PetscFree(cols);
2985: } else {
2986: /* determine buffer space needed for message */
2987: nz = 0;
2988: for (i=0; i<m; i++) {
2989: nz += ourlens[i];
2990: }
2991: PetscMalloc1(nz,&mycols);
2993: /* receive message of column indices*/
2994: MPIULong_Recv(mycols,nz,MPIU_INT,0,tag,comm);
2995: }
2997: /* determine column ownership if matrix is not square */
2998: if (N != M) {
2999: if (newMat->cmap->n < 0) n = N/size + ((N % size) > rank);
3000: else n = newMat->cmap->n;
3001: MPI_Scan(&n,&cend,1,MPIU_INT,MPI_SUM,comm);
3002: cstart = cend - n;
3003: } else {
3004: cstart = rstart;
3005: cend = rend;
3006: n = cend - cstart;
3007: }
3009: /* loop over local rows, determining number of off diagonal entries */
3010: PetscMemzero(offlens,m*sizeof(PetscInt));
3011: jj = 0;
3012: for (i=0; i<m; i++) {
3013: for (j=0; j<ourlens[i]; j++) {
3014: if (mycols[jj] < cstart || mycols[jj] >= cend) offlens[i]++;
3015: jj++;
3016: }
3017: }
3019: for (i=0; i<m; i++) {
3020: ourlens[i] -= offlens[i];
3021: }
3022: MatSetSizes(newMat,m,n,M,N);
3024: if (bs > 1) {MatSetBlockSize(newMat,bs);}
3026: MatMPIAIJSetPreallocation(newMat,0,ourlens,0,offlens);
3028: for (i=0; i<m; i++) {
3029: ourlens[i] += offlens[i];
3030: }
3032: if (!rank) {
3033: PetscMalloc1(maxnz+1,&vals);
3035: /* read in my part of the matrix numerical values */
3036: nz = procsnz[0];
3037: PetscBinaryRead(fd,vals,nz,PETSC_SCALAR);
3039: /* insert into matrix */
3040: jj = rstart;
3041: smycols = mycols;
3042: svals = vals;
3043: for (i=0; i<m; i++) {
3044: MatSetValues_MPIAIJ(newMat,1,&jj,ourlens[i],smycols,svals,INSERT_VALUES);
3045: smycols += ourlens[i];
3046: svals += ourlens[i];
3047: jj++;
3048: }
3050: /* read in other processors and ship out */
3051: for (i=1; i<size; i++) {
3052: nz = procsnz[i];
3053: PetscBinaryRead(fd,vals,nz,PETSC_SCALAR);
3054: MPIULong_Send(vals,nz,MPIU_SCALAR,i,((PetscObject)newMat)->tag,comm);
3055: }
3056: PetscFree(procsnz);
3057: } else {
3058: /* receive numeric values */
3059: PetscMalloc1(nz+1,&vals);
3061: /* receive message of values*/
3062: MPIULong_Recv(vals,nz,MPIU_SCALAR,0,((PetscObject)newMat)->tag,comm);
3064: /* insert into matrix */
3065: jj = rstart;
3066: smycols = mycols;
3067: svals = vals;
3068: for (i=0; i<m; i++) {
3069: MatSetValues_MPIAIJ(newMat,1,&jj,ourlens[i],smycols,svals,INSERT_VALUES);
3070: smycols += ourlens[i];
3071: svals += ourlens[i];
3072: jj++;
3073: }
3074: }
3075: PetscFree2(ourlens,offlens);
3076: PetscFree(vals);
3077: PetscFree(mycols);
3078: PetscFree(rowners);
3079: MatAssemblyBegin(newMat,MAT_FINAL_ASSEMBLY);
3080: MatAssemblyEnd(newMat,MAT_FINAL_ASSEMBLY);
3081: return(0);
3082: }
3084: /* Not scalable because of ISAllGather() unless getting all columns. */
3085: PetscErrorCode ISGetSeqIS_Private(Mat mat,IS iscol,IS *isseq)
3086: {
3088: IS iscol_local;
3089: PetscBool isstride;
3090: PetscMPIInt lisstride=0,gisstride;
3093: /* check if we are grabbing all columns*/
3094: PetscObjectTypeCompare((PetscObject)iscol,ISSTRIDE,&isstride);
3096: if (isstride) {
3097: PetscInt start,len,mstart,mlen;
3098: ISStrideGetInfo(iscol,&start,NULL);
3099: ISGetLocalSize(iscol,&len);
3100: MatGetOwnershipRangeColumn(mat,&mstart,&mlen);
3101: if (mstart == start && mlen-mstart == len) lisstride = 1;
3102: }
3104: MPIU_Allreduce(&lisstride,&gisstride,1,MPI_INT,MPI_MIN,PetscObjectComm((PetscObject)mat));
3105: if (gisstride) {
3106: PetscInt N;
3107: MatGetSize(mat,NULL,&N);
3108: ISCreateStride(PetscObjectComm((PetscObject)mat),N,0,1,&iscol_local);
3109: ISSetIdentity(iscol_local);
3110: PetscInfo(mat,"Optimizing for obtaining all columns of the matrix; skipping ISAllGather()\n");
3111: } else {
3112: PetscInt cbs;
3113: ISGetBlockSize(iscol,&cbs);
3114: ISAllGather(iscol,&iscol_local);
3115: ISSetBlockSize(iscol_local,cbs);
3116: }
3118: *isseq = iscol_local;
3119: return(0);
3120: }
3122: /*
3123: Used by MatCreateSubMatrix_MPIAIJ_SameRowColDist() to avoid ISAllGather() and global size of iscol_local
3124: (see MatCreateSubMatrix_MPIAIJ_nonscalable)
3126: Input Parameters:
3127: mat - matrix
3128: isrow - parallel row index set; its local indices are a subset of local columns of mat,
3129: i.e., mat->rstart <= isrow[i] < mat->rend
3130: iscol - parallel column index set; its local indices are a subset of local columns of mat,
3131: i.e., mat->cstart <= iscol[i] < mat->cend
3132: Output Parameter:
3133: isrow_d,iscol_d - sequential row and column index sets for retrieving mat->A
3134: iscol_o - sequential column index set for retrieving mat->B
3135: garray - column map; garray[i] indicates global location of iscol_o[i] in iscol
3136: */
3137: PetscErrorCode ISGetSeqIS_SameColDist_Private(Mat mat,IS isrow,IS iscol,IS *isrow_d,IS *iscol_d,IS *iscol_o,const PetscInt *garray[])
3138: {
3140: Vec x,cmap;
3141: const PetscInt *is_idx;
3142: PetscScalar *xarray,*cmaparray;
3143: PetscInt ncols,isstart,*idx,m,rstart,*cmap1,count;
3144: Mat_MPIAIJ *a=(Mat_MPIAIJ*)mat->data;
3145: Mat B=a->B;
3146: Vec lvec=a->lvec,lcmap;
3147: PetscInt i,cstart,cend,Bn=B->cmap->N;
3148: MPI_Comm comm;
3149: VecScatter Mvctx=a->Mvctx;
3152: PetscObjectGetComm((PetscObject)mat,&comm);
3153: ISGetLocalSize(iscol,&ncols);
3155: /* (1) iscol is a sub-column vector of mat, pad it with '-1.' to form a full vector x */
3156: MatCreateVecs(mat,&x,NULL);
3157: VecSet(x,-1.0);
3158: VecDuplicate(x,&cmap);
3159: VecSet(cmap,-1.0);
3161: /* Get start indices */
3162: MPI_Scan(&ncols,&isstart,1,MPIU_INT,MPI_SUM,comm);
3163: isstart -= ncols;
3164: MatGetOwnershipRangeColumn(mat,&cstart,&cend);
3166: ISGetIndices(iscol,&is_idx);
3167: VecGetArray(x,&xarray);
3168: VecGetArray(cmap,&cmaparray);
3169: PetscMalloc1(ncols,&idx);
3170: for (i=0; i<ncols; i++) {
3171: xarray[is_idx[i]-cstart] = (PetscScalar)is_idx[i];
3172: cmaparray[is_idx[i]-cstart] = i + isstart; /* global index of iscol[i] */
3173: idx[i] = is_idx[i]-cstart; /* local index of iscol[i] */
3174: }
3175: VecRestoreArray(x,&xarray);
3176: VecRestoreArray(cmap,&cmaparray);
3177: ISRestoreIndices(iscol,&is_idx);
3179: /* Get iscol_d */
3180: ISCreateGeneral(PETSC_COMM_SELF,ncols,idx,PETSC_OWN_POINTER,iscol_d);
3181: ISGetBlockSize(iscol,&i);
3182: ISSetBlockSize(*iscol_d,i);
3184: /* Get isrow_d */
3185: ISGetLocalSize(isrow,&m);
3186: rstart = mat->rmap->rstart;
3187: PetscMalloc1(m,&idx);
3188: ISGetIndices(isrow,&is_idx);
3189: for (i=0; i<m; i++) idx[i] = is_idx[i]-rstart;
3190: ISRestoreIndices(isrow,&is_idx);
3192: ISCreateGeneral(PETSC_COMM_SELF,m,idx,PETSC_OWN_POINTER,isrow_d);
3193: ISGetBlockSize(isrow,&i);
3194: ISSetBlockSize(*isrow_d,i);
3196: /* (2) Scatter x and cmap using aij->Mvctx to get their off-process portions (see MatMult_MPIAIJ) */
3197: VecScatterBegin(Mvctx,x,lvec,INSERT_VALUES,SCATTER_FORWARD);
3198: VecScatterEnd(Mvctx,x,lvec,INSERT_VALUES,SCATTER_FORWARD);
3200: VecDuplicate(lvec,&lcmap);
3202: VecScatterBegin(Mvctx,cmap,lcmap,INSERT_VALUES,SCATTER_FORWARD);
3203: VecScatterEnd(Mvctx,cmap,lcmap,INSERT_VALUES,SCATTER_FORWARD);
3205: /* (3) create sequential iscol_o (a subset of iscol) and isgarray */
3206: /* off-process column indices */
3207: count = 0;
3208: PetscMalloc1(Bn,&idx);
3209: PetscMalloc1(Bn,&cmap1);
3211: VecGetArray(lvec,&xarray);
3212: VecGetArray(lcmap,&cmaparray);
3213: for (i=0; i<Bn; i++) {
3214: if (PetscRealPart(xarray[i]) > -1.0) {
3215: idx[count] = i; /* local column index in off-diagonal part B */
3216: cmap1[count] = (PetscInt)PetscRealPart(cmaparray[i]); /* column index in submat */
3217: count++;
3218: }
3219: }
3220: VecRestoreArray(lvec,&xarray);
3221: VecRestoreArray(lcmap,&cmaparray);
3223: ISCreateGeneral(PETSC_COMM_SELF,count,idx,PETSC_COPY_VALUES,iscol_o);
3224: /* cannot ensure iscol_o has same blocksize as iscol! */
3226: PetscFree(idx);
3227: *garray = cmap1;
3229: VecDestroy(&x);
3230: VecDestroy(&cmap);
3231: VecDestroy(&lcmap);
3232: return(0);
3233: }
3235: /* isrow and iscol have same processor distribution as mat, output *submat is a submatrix of local mat */
3236: PetscErrorCode MatCreateSubMatrix_MPIAIJ_SameRowColDist(Mat mat,IS isrow,IS iscol,MatReuse call,Mat *submat)
3237: {
3239: Mat_MPIAIJ *a = (Mat_MPIAIJ*)mat->data,*asub;
3240: Mat M = NULL;
3241: MPI_Comm comm;
3242: IS iscol_d,isrow_d,iscol_o;
3243: Mat Asub = NULL,Bsub = NULL;
3244: PetscInt n;
3247: PetscObjectGetComm((PetscObject)mat,&comm);
3249: if (call == MAT_REUSE_MATRIX) {
3250: /* Retrieve isrow_d, iscol_d and iscol_o from submat */
3251: PetscObjectQuery((PetscObject)*submat,"isrow_d",(PetscObject*)&isrow_d);
3252: if (!isrow_d) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"isrow_d passed in was not used before, cannot reuse");
3254: PetscObjectQuery((PetscObject)*submat,"iscol_d",(PetscObject*)&iscol_d);
3255: if (!iscol_d) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"iscol_d passed in was not used before, cannot reuse");
3257: PetscObjectQuery((PetscObject)*submat,"iscol_o",(PetscObject*)&iscol_o);
3258: if (!iscol_o) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"iscol_o passed in was not used before, cannot reuse");
3260: /* Update diagonal and off-diagonal portions of submat */
3261: asub = (Mat_MPIAIJ*)(*submat)->data;
3262: MatCreateSubMatrix_SeqAIJ(a->A,isrow_d,iscol_d,PETSC_DECIDE,MAT_REUSE_MATRIX,&asub->A);
3263: ISGetLocalSize(iscol_o,&n);
3264: if (n) {
3265: MatCreateSubMatrix_SeqAIJ(a->B,isrow_d,iscol_o,PETSC_DECIDE,MAT_REUSE_MATRIX,&asub->B);
3266: }
3267: MatAssemblyBegin(*submat,MAT_FINAL_ASSEMBLY);
3268: MatAssemblyEnd(*submat,MAT_FINAL_ASSEMBLY);
3270: } else { /* call == MAT_INITIAL_MATRIX) */
3271: const PetscInt *garray;
3272: PetscInt BsubN;
3274: /* Create isrow_d, iscol_d, iscol_o and isgarray (replace isgarray with array?) */
3275: ISGetSeqIS_SameColDist_Private(mat,isrow,iscol,&isrow_d,&iscol_d,&iscol_o,&garray);
3277: /* Create local submatrices Asub and Bsub */
3278: MatCreateSubMatrix_SeqAIJ(a->A,isrow_d,iscol_d,PETSC_DECIDE,MAT_INITIAL_MATRIX,&Asub);
3279: MatCreateSubMatrix_SeqAIJ(a->B,isrow_d,iscol_o,PETSC_DECIDE,MAT_INITIAL_MATRIX,&Bsub);
3281: /* Create submatrix M */
3282: MatCreateMPIAIJWithSeqAIJ(comm,Asub,Bsub,garray,&M);
3284: /* If Bsub has empty columns, compress iscol_o such that it will retrieve condensed Bsub from a->B during reuse */
3285: asub = (Mat_MPIAIJ*)M->data;
3287: ISGetLocalSize(iscol_o,&BsubN);
3288: n = asub->B->cmap->N;
3289: if (BsubN > n) {
3290: /* This case can be tested using ~petsc/src/tao/bound/examples/tutorials/runplate2_3 */
3291: const PetscInt *idx;
3292: PetscInt i,j,*idx_new,*subgarray = asub->garray;
3293: PetscInfo2(M,"submatrix Bn %D != BsubN %D, update iscol_o\n",n,BsubN);
3295: PetscMalloc1(n,&idx_new);
3296: j = 0;
3297: ISGetIndices(iscol_o,&idx);
3298: for (i=0; i<n; i++) {
3299: if (j >= BsubN) break;
3300: while (subgarray[i] > garray[j]) j++;
3302: if (subgarray[i] == garray[j]) {
3303: idx_new[i] = idx[j++];
3304: } else SETERRQ4(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"subgarray[%D]=%D cannot < garray[%D]=%D",i,subgarray[i],j,garray[j]);
3305: }
3306: ISRestoreIndices(iscol_o,&idx);
3308: ISDestroy(&iscol_o);
3309: ISCreateGeneral(PETSC_COMM_SELF,n,idx_new,PETSC_OWN_POINTER,&iscol_o);
3311: } else if (BsubN < n) {
3312: SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Columns of Bsub cannot be smaller than B's",BsubN,asub->B->cmap->N);
3313: }
3315: PetscFree(garray);
3316: *submat = M;
3318: /* Save isrow_d, iscol_d and iscol_o used in processor for next request */
3319: PetscObjectCompose((PetscObject)M,"isrow_d",(PetscObject)isrow_d);
3320: ISDestroy(&isrow_d);
3322: PetscObjectCompose((PetscObject)M,"iscol_d",(PetscObject)iscol_d);
3323: ISDestroy(&iscol_d);
3325: PetscObjectCompose((PetscObject)M,"iscol_o",(PetscObject)iscol_o);
3326: ISDestroy(&iscol_o);
3327: }
3328: return(0);
3329: }
3331: PetscErrorCode MatCreateSubMatrix_MPIAIJ(Mat mat,IS isrow,IS iscol,MatReuse call,Mat *newmat)
3332: {
3334: IS iscol_local=NULL,isrow_d;
3335: PetscInt csize;
3336: PetscInt n,i,j,start,end;
3337: PetscBool sameRowDist=PETSC_FALSE,sameDist[2],tsameDist[2];
3338: MPI_Comm comm;
3341: /* If isrow has same processor distribution as mat,
3342: call MatCreateSubMatrix_MPIAIJ_SameRowDist() to avoid using a hash table with global size of iscol */
3343: if (call == MAT_REUSE_MATRIX) {
3344: PetscObjectQuery((PetscObject)*newmat,"isrow_d",(PetscObject*)&isrow_d);
3345: if (isrow_d) {
3346: sameRowDist = PETSC_TRUE;
3347: tsameDist[1] = PETSC_TRUE; /* sameColDist */
3348: } else {
3349: PetscObjectQuery((PetscObject)*newmat,"SubIScol",(PetscObject*)&iscol_local);
3350: if (iscol_local) {
3351: sameRowDist = PETSC_TRUE;
3352: tsameDist[1] = PETSC_FALSE; /* !sameColDist */
3353: }
3354: }
3355: } else {
3356: /* Check if isrow has same processor distribution as mat */
3357: sameDist[0] = PETSC_FALSE;
3358: ISGetLocalSize(isrow,&n);
3359: if (!n) {
3360: sameDist[0] = PETSC_TRUE;
3361: } else {
3362: ISGetMinMax(isrow,&i,&j);
3363: MatGetOwnershipRange(mat,&start,&end);
3364: if (i >= start && j < end) {
3365: sameDist[0] = PETSC_TRUE;
3366: }
3367: }
3369: /* Check if iscol has same processor distribution as mat */
3370: sameDist[1] = PETSC_FALSE;
3371: ISGetLocalSize(iscol,&n);
3372: if (!n) {
3373: sameDist[1] = PETSC_TRUE;
3374: } else {
3375: ISGetMinMax(iscol,&i,&j);
3376: MatGetOwnershipRangeColumn(mat,&start,&end);
3377: if (i >= start && j < end) sameDist[1] = PETSC_TRUE;
3378: }
3380: PetscObjectGetComm((PetscObject)mat,&comm);
3381: MPIU_Allreduce(&sameDist,&tsameDist,2,MPIU_BOOL,MPI_LAND,comm);
3382: sameRowDist = tsameDist[0];
3383: }
3385: if (sameRowDist) {
3386: if (tsameDist[1]) { /* sameRowDist & sameColDist */
3387: /* isrow and iscol have same processor distribution as mat */
3388: MatCreateSubMatrix_MPIAIJ_SameRowColDist(mat,isrow,iscol,call,newmat);
3389: return(0);
3390: } else { /* sameRowDist */
3391: /* isrow has same processor distribution as mat */
3392: if (call == MAT_INITIAL_MATRIX) {
3393: PetscBool sorted;
3394: ISGetSeqIS_Private(mat,iscol,&iscol_local);
3395: ISGetLocalSize(iscol_local,&n); /* local size of iscol_local = global columns of newmat */
3396: ISGetSize(iscol,&i);
3397: if (n != i) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"n %d != size of iscol %d",n,i);
3399: ISSorted(iscol_local,&sorted);
3400: if (sorted) {
3401: /* MatCreateSubMatrix_MPIAIJ_SameRowDist() requires iscol_local be sorted; it can have duplicate indices */
3402: MatCreateSubMatrix_MPIAIJ_SameRowDist(mat,isrow,iscol,iscol_local,MAT_INITIAL_MATRIX,newmat);
3403: return(0);
3404: }
3405: } else { /* call == MAT_REUSE_MATRIX */
3406: IS iscol_sub;
3407: PetscObjectQuery((PetscObject)*newmat,"SubIScol",(PetscObject*)&iscol_sub);
3408: if (iscol_sub) {
3409: MatCreateSubMatrix_MPIAIJ_SameRowDist(mat,isrow,iscol,NULL,call,newmat);
3410: return(0);
3411: }
3412: }
3413: }
3414: }
3416: /* General case: iscol -> iscol_local which has global size of iscol */
3417: if (call == MAT_REUSE_MATRIX) {
3418: PetscObjectQuery((PetscObject)*newmat,"ISAllGather",(PetscObject*)&iscol_local);
3419: if (!iscol_local) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Submatrix passed in was not used before, cannot reuse");
3420: } else {
3421: if (!iscol_local) {
3422: ISGetSeqIS_Private(mat,iscol,&iscol_local);
3423: }
3424: }
3426: ISGetLocalSize(iscol,&csize);
3427: MatCreateSubMatrix_MPIAIJ_nonscalable(mat,isrow,iscol_local,csize,call,newmat);
3429: if (call == MAT_INITIAL_MATRIX) {
3430: PetscObjectCompose((PetscObject)*newmat,"ISAllGather",(PetscObject)iscol_local);
3431: ISDestroy(&iscol_local);
3432: }
3433: return(0);
3434: }
3436: /*@C
3437: MatCreateMPIAIJWithSeqAIJ - creates a MPIAIJ matrix using SeqAIJ matrices that contain the "diagonal"
3438: and "off-diagonal" part of the matrix in CSR format.
3440: Collective on MPI_Comm
3442: Input Parameters:
3443: + comm - MPI communicator
3444: . A - "diagonal" portion of matrix
3445: . B - "off-diagonal" portion of matrix, may have empty columns, will be destroyed by this routine
3446: - garray - global index of B columns
3448: Output Parameter:
3449: . mat - the matrix, with input A as its local diagonal matrix
3450: Level: advanced
3452: Notes:
3453: See MatCreateAIJ() for the definition of "diagonal" and "off-diagonal" portion of the matrix.
3454: A becomes part of output mat, B is destroyed by this routine. The user cannot use A and B anymore.
3456: .seealso: MatCreateMPIAIJWithSplitArrays()
3457: @*/
3458: PetscErrorCode MatCreateMPIAIJWithSeqAIJ(MPI_Comm comm,Mat A,Mat B,const PetscInt garray[],Mat *mat)
3459: {
3461: Mat_MPIAIJ *maij;
3462: Mat_SeqAIJ *b=(Mat_SeqAIJ*)B->data,*bnew;
3463: PetscInt *oi=b->i,*oj=b->j,i,nz,col;
3464: PetscScalar *oa=b->a;
3465: Mat Bnew;
3466: PetscInt m,n,N;
3469: MatCreate(comm,mat);
3470: MatGetSize(A,&m,&n);
3471: if (m != B->rmap->N) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Am %D != Bm %D",m,B->rmap->N);
3472: if (A->rmap->bs != B->rmap->bs) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"A row bs %D != B row bs %D",A->rmap->bs,B->rmap->bs);
3473: /* remove check below; When B is created using iscol_o from ISGetSeqIS_SameColDist_Private(), its bs may not be same as A */
3474: /* if (A->cmap->bs != B->cmap->bs) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"A column bs %D != B column bs %D",A->cmap->bs,B->cmap->bs); */
3476: /* Get global columns of mat */
3477: MPIU_Allreduce(&n,&N,1,MPIU_INT,MPI_SUM,comm);
3479: MatSetSizes(*mat,m,n,PETSC_DECIDE,N);
3480: MatSetType(*mat,MATMPIAIJ);
3481: MatSetBlockSizes(*mat,A->rmap->bs,A->cmap->bs);
3482: maij = (Mat_MPIAIJ*)(*mat)->data;
3484: (*mat)->preallocated = PETSC_TRUE;
3486: PetscLayoutSetUp((*mat)->rmap);
3487: PetscLayoutSetUp((*mat)->cmap);
3489: /* Set A as diagonal portion of *mat */
3490: maij->A = A;
3492: nz = oi[m];
3493: for (i=0; i<nz; i++) {
3494: col = oj[i];
3495: oj[i] = garray[col];
3496: }
3498: /* Set Bnew as off-diagonal portion of *mat */
3499: MatCreateSeqAIJWithArrays(PETSC_COMM_SELF,m,N,oi,oj,oa,&Bnew);
3500: bnew = (Mat_SeqAIJ*)Bnew->data;
3501: bnew->maxnz = b->maxnz; /* allocated nonzeros of B */
3502: maij->B = Bnew;
3504: if (B->rmap->N != Bnew->rmap->N) SETERRQ2(PETSC_COMM_SELF,0,"BN %d != BnewN %d",B->rmap->N,Bnew->rmap->N);
3506: b->singlemalloc = PETSC_FALSE; /* B arrays are shared by Bnew */
3507: b->free_a = PETSC_FALSE;
3508: b->free_ij = PETSC_FALSE;
3509: MatDestroy(&B);
3511: bnew->singlemalloc = PETSC_TRUE; /* arrays will be freed by MatDestroy(&Bnew) */
3512: bnew->free_a = PETSC_TRUE;
3513: bnew->free_ij = PETSC_TRUE;
3515: /* condense columns of maij->B */
3516: MatSetOption(*mat,MAT_NO_OFF_PROC_ENTRIES,PETSC_TRUE);
3517: MatAssemblyBegin(*mat,MAT_FINAL_ASSEMBLY);
3518: MatAssemblyEnd(*mat,MAT_FINAL_ASSEMBLY);
3519: MatSetOption(*mat,MAT_NO_OFF_PROC_ENTRIES,PETSC_FALSE);
3520: MatSetOption(*mat,MAT_NEW_NONZERO_LOCATION_ERR,PETSC_TRUE);
3521: return(0);
3522: }
3524: extern PetscErrorCode MatCreateSubMatrices_MPIAIJ_SingleIS_Local(Mat,PetscInt,const IS[],const IS[],MatReuse,PetscBool,Mat*);
3526: PetscErrorCode MatCreateSubMatrix_MPIAIJ_SameRowDist(Mat mat,IS isrow,IS iscol,IS iscol_local,MatReuse call,Mat *newmat)
3527: {
3529: PetscInt i,m,n,rstart,row,rend,nz,j,bs,cbs;
3530: PetscInt *ii,*jj,nlocal,*dlens,*olens,dlen,olen,jend,mglobal;
3531: Mat_MPIAIJ *a=(Mat_MPIAIJ*)mat->data;
3532: Mat M,Msub,B=a->B;
3533: MatScalar *aa;
3534: Mat_SeqAIJ *aij;
3535: PetscInt *garray = a->garray,*colsub,Ncols;
3536: PetscInt count,Bn=B->cmap->N,cstart=mat->cmap->rstart,cend=mat->cmap->rend;
3537: IS iscol_sub,iscmap;
3538: const PetscInt *is_idx,*cmap;
3539: PetscBool allcolumns=PETSC_FALSE;
3540: MPI_Comm comm;
3543: PetscObjectGetComm((PetscObject)mat,&comm);
3545: if (call == MAT_REUSE_MATRIX) {
3546: PetscObjectQuery((PetscObject)*newmat,"SubIScol",(PetscObject*)&iscol_sub);
3547: if (!iscol_sub) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"SubIScol passed in was not used before, cannot reuse");
3548: ISGetLocalSize(iscol_sub,&count);
3550: PetscObjectQuery((PetscObject)*newmat,"Subcmap",(PetscObject*)&iscmap);
3551: if (!iscmap) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Subcmap passed in was not used before, cannot reuse");
3553: PetscObjectQuery((PetscObject)*newmat,"SubMatrix",(PetscObject*)&Msub);
3554: if (!Msub) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Submatrix passed in was not used before, cannot reuse");
3556: MatCreateSubMatrices_MPIAIJ_SingleIS_Local(mat,1,&isrow,&iscol_sub,MAT_REUSE_MATRIX,PETSC_FALSE,&Msub);
3558: } else { /* call == MAT_INITIAL_MATRIX) */
3559: PetscBool flg;
3561: ISGetLocalSize(iscol,&n);
3562: ISGetSize(iscol,&Ncols);
3564: /* (1) iscol -> nonscalable iscol_local */
3565: /* Check for special case: each processor gets entire matrix columns */
3566: ISIdentity(iscol_local,&flg);
3567: if (flg && n == mat->cmap->N) allcolumns = PETSC_TRUE;
3568: if (allcolumns) {
3569: iscol_sub = iscol_local;
3570: PetscObjectReference((PetscObject)iscol_local);
3571: ISCreateStride(PETSC_COMM_SELF,n,0,1,&iscmap);
3573: } else {
3574: /* (2) iscol_local -> iscol_sub and iscmap. Implementation below requires iscol_local be sorted, it can have duplicate indices */
3575: PetscInt *idx,*cmap1,k;
3576: PetscMalloc1(Ncols,&idx);
3577: PetscMalloc1(Ncols,&cmap1);
3578: ISGetIndices(iscol_local,&is_idx);
3579: count = 0;
3580: k = 0;
3581: for (i=0; i<Ncols; i++) {
3582: j = is_idx[i];
3583: if (j >= cstart && j < cend) {
3584: /* diagonal part of mat */
3585: idx[count] = j;
3586: cmap1[count++] = i; /* column index in submat */
3587: } else if (Bn) {
3588: /* off-diagonal part of mat */
3589: if (j == garray[k]) {
3590: idx[count] = j;
3591: cmap1[count++] = i; /* column index in submat */
3592: } else if (j > garray[k]) {
3593: while (j > garray[k] && k < Bn-1) k++;
3594: if (j == garray[k]) {
3595: idx[count] = j;
3596: cmap1[count++] = i; /* column index in submat */
3597: }
3598: }
3599: }
3600: }
3601: ISRestoreIndices(iscol_local,&is_idx);
3603: ISCreateGeneral(PETSC_COMM_SELF,count,idx,PETSC_OWN_POINTER,&iscol_sub);
3604: ISGetBlockSize(iscol,&cbs);
3605: ISSetBlockSize(iscol_sub,cbs);
3607: ISCreateGeneral(PetscObjectComm((PetscObject)iscol_local),count,cmap1,PETSC_OWN_POINTER,&iscmap);
3608: }
3610: /* (3) Create sequential Msub */
3611: MatCreateSubMatrices_MPIAIJ_SingleIS_Local(mat,1,&isrow,&iscol_sub,MAT_INITIAL_MATRIX,allcolumns,&Msub);
3612: }
3614: ISGetLocalSize(iscol_sub,&count);
3615: aij = (Mat_SeqAIJ*)(Msub)->data;
3616: ii = aij->i;
3617: ISGetIndices(iscmap,&cmap);
3619: /*
3620: m - number of local rows
3621: Ncols - number of columns (same on all processors)
3622: rstart - first row in new global matrix generated
3623: */
3624: MatGetSize(Msub,&m,NULL);
3626: if (call == MAT_INITIAL_MATRIX) {
3627: /* (4) Create parallel newmat */
3628: PetscMPIInt rank,size;
3629: PetscInt csize;
3631: MPI_Comm_size(comm,&size);
3632: MPI_Comm_rank(comm,&rank);
3634: /*
3635: Determine the number of non-zeros in the diagonal and off-diagonal
3636: portions of the matrix in order to do correct preallocation
3637: */
3639: /* first get start and end of "diagonal" columns */
3640: ISGetLocalSize(iscol,&csize);
3641: if (csize == PETSC_DECIDE) {
3642: ISGetSize(isrow,&mglobal);
3643: if (mglobal == Ncols) { /* square matrix */
3644: nlocal = m;
3645: } else {
3646: nlocal = Ncols/size + ((Ncols % size) > rank);
3647: }
3648: } else {
3649: nlocal = csize;
3650: }
3651: MPI_Scan(&nlocal,&rend,1,MPIU_INT,MPI_SUM,comm);
3652: rstart = rend - nlocal;
3653: if (rank == size - 1 && rend != Ncols) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Local column sizes %D do not add up to total number of columns %D",rend,Ncols);
3655: /* next, compute all the lengths */
3656: jj = aij->j;
3657: PetscMalloc1(2*m+1,&dlens);
3658: olens = dlens + m;
3659: for (i=0; i<m; i++) {
3660: jend = ii[i+1] - ii[i];
3661: olen = 0;
3662: dlen = 0;
3663: for (j=0; j<jend; j++) {
3664: if (cmap[*jj] < rstart || cmap[*jj] >= rend) olen++;
3665: else dlen++;
3666: jj++;
3667: }
3668: olens[i] = olen;
3669: dlens[i] = dlen;
3670: }
3672: ISGetBlockSize(isrow,&bs);
3673: ISGetBlockSize(iscol,&cbs);
3675: MatCreate(comm,&M);
3676: MatSetSizes(M,m,nlocal,PETSC_DECIDE,Ncols);
3677: MatSetBlockSizes(M,bs,cbs);
3678: MatSetType(M,((PetscObject)mat)->type_name);
3679: MatMPIAIJSetPreallocation(M,0,dlens,0,olens);
3680: PetscFree(dlens);
3682: } else { /* call == MAT_REUSE_MATRIX */
3683: M = *newmat;
3684: MatGetLocalSize(M,&i,NULL);
3685: if (i != m) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Previous matrix must be same size/layout as request");
3686: MatZeroEntries(M);
3687: /*
3688: The next two lines are needed so we may call MatSetValues_MPIAIJ() below directly,
3689: rather than the slower MatSetValues().
3690: */
3691: M->was_assembled = PETSC_TRUE;
3692: M->assembled = PETSC_FALSE;
3693: }
3695: /* (5) Set values of Msub to *newmat */
3696: PetscMalloc1(count,&colsub);
3697: MatGetOwnershipRange(M,&rstart,NULL);
3699: jj = aij->j;
3700: aa = aij->a;
3701: for (i=0; i<m; i++) {
3702: row = rstart + i;
3703: nz = ii[i+1] - ii[i];
3704: for (j=0; j<nz; j++) colsub[j] = cmap[jj[j]];
3705: MatSetValues_MPIAIJ(M,1,&row,nz,colsub,aa,INSERT_VALUES);
3706: jj += nz; aa += nz;
3707: }
3708: ISRestoreIndices(iscmap,&cmap);
3710: MatAssemblyBegin(M,MAT_FINAL_ASSEMBLY);
3711: MatAssemblyEnd(M,MAT_FINAL_ASSEMBLY);
3713: PetscFree(colsub);
3715: /* save Msub, iscol_sub and iscmap used in processor for next request */
3716: if (call == MAT_INITIAL_MATRIX) {
3717: *newmat = M;
3718: PetscObjectCompose((PetscObject)(*newmat),"SubMatrix",(PetscObject)Msub);
3719: MatDestroy(&Msub);
3721: PetscObjectCompose((PetscObject)(*newmat),"SubIScol",(PetscObject)iscol_sub);
3722: ISDestroy(&iscol_sub);
3724: PetscObjectCompose((PetscObject)(*newmat),"Subcmap",(PetscObject)iscmap);
3725: ISDestroy(&iscmap);
3727: if (iscol_local) {
3728: PetscObjectCompose((PetscObject)(*newmat),"ISAllGather",(PetscObject)iscol_local);
3729: ISDestroy(&iscol_local);
3730: }
3731: }
3732: return(0);
3733: }
3735: /*
3736: Not great since it makes two copies of the submatrix, first an SeqAIJ
3737: in local and then by concatenating the local matrices the end result.
3738: Writing it directly would be much like MatCreateSubMatrices_MPIAIJ()
3740: Note: This requires a sequential iscol with all indices.
3741: */
3742: PetscErrorCode MatCreateSubMatrix_MPIAIJ_nonscalable(Mat mat,IS isrow,IS iscol,PetscInt csize,MatReuse call,Mat *newmat)
3743: {
3745: PetscMPIInt rank,size;
3746: PetscInt i,m,n,rstart,row,rend,nz,*cwork,j,bs,cbs;
3747: PetscInt *ii,*jj,nlocal,*dlens,*olens,dlen,olen,jend,mglobal;
3748: Mat M,Mreuse;
3749: MatScalar *aa,*vwork;
3750: MPI_Comm comm;
3751: Mat_SeqAIJ *aij;
3752: PetscBool colflag,allcolumns=PETSC_FALSE;
3755: PetscObjectGetComm((PetscObject)mat,&comm);
3756: MPI_Comm_rank(comm,&rank);
3757: MPI_Comm_size(comm,&size);
3759: /* Check for special case: each processor gets entire matrix columns */
3760: ISIdentity(iscol,&colflag);
3761: ISGetLocalSize(iscol,&n);
3762: if (colflag && n == mat->cmap->N) allcolumns = PETSC_TRUE;
3764: if (call == MAT_REUSE_MATRIX) {
3765: PetscObjectQuery((PetscObject)*newmat,"SubMatrix",(PetscObject*)&Mreuse);
3766: if (!Mreuse) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Submatrix passed in was not used before, cannot reuse");
3767: MatCreateSubMatrices_MPIAIJ_SingleIS_Local(mat,1,&isrow,&iscol,MAT_REUSE_MATRIX,allcolumns,&Mreuse);
3768: } else {
3769: MatCreateSubMatrices_MPIAIJ_SingleIS_Local(mat,1,&isrow,&iscol,MAT_INITIAL_MATRIX,allcolumns,&Mreuse);
3770: }
3772: /*
3773: m - number of local rows
3774: n - number of columns (same on all processors)
3775: rstart - first row in new global matrix generated
3776: */
3777: MatGetSize(Mreuse,&m,&n);
3778: MatGetBlockSizes(Mreuse,&bs,&cbs);
3779: if (call == MAT_INITIAL_MATRIX) {
3780: aij = (Mat_SeqAIJ*)(Mreuse)->data;
3781: ii = aij->i;
3782: jj = aij->j;
3784: /*
3785: Determine the number of non-zeros in the diagonal and off-diagonal
3786: portions of the matrix in order to do correct preallocation
3787: */
3789: /* first get start and end of "diagonal" columns */
3790: if (csize == PETSC_DECIDE) {
3791: ISGetSize(isrow,&mglobal);
3792: if (mglobal == n) { /* square matrix */
3793: nlocal = m;
3794: } else {
3795: nlocal = n/size + ((n % size) > rank);
3796: }
3797: } else {
3798: nlocal = csize;
3799: }
3800: MPI_Scan(&nlocal,&rend,1,MPIU_INT,MPI_SUM,comm);
3801: rstart = rend - nlocal;
3802: if (rank == size - 1 && rend != n) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Local column sizes %D do not add up to total number of columns %D",rend,n);
3804: /* next, compute all the lengths */
3805: PetscMalloc1(2*m+1,&dlens);
3806: olens = dlens + m;
3807: for (i=0; i<m; i++) {
3808: jend = ii[i+1] - ii[i];
3809: olen = 0;
3810: dlen = 0;
3811: for (j=0; j<jend; j++) {
3812: if (*jj < rstart || *jj >= rend) olen++;
3813: else dlen++;
3814: jj++;
3815: }
3816: olens[i] = olen;
3817: dlens[i] = dlen;
3818: }
3819: MatCreate(comm,&M);
3820: MatSetSizes(M,m,nlocal,PETSC_DECIDE,n);
3821: MatSetBlockSizes(M,bs,cbs);
3822: MatSetType(M,((PetscObject)mat)->type_name);
3823: MatMPIAIJSetPreallocation(M,0,dlens,0,olens);
3824: PetscFree(dlens);
3825: } else {
3826: PetscInt ml,nl;
3828: M = *newmat;
3829: MatGetLocalSize(M,&ml,&nl);
3830: if (ml != m) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Previous matrix must be same size/layout as request");
3831: MatZeroEntries(M);
3832: /*
3833: The next two lines are needed so we may call MatSetValues_MPIAIJ() below directly,
3834: rather than the slower MatSetValues().
3835: */
3836: M->was_assembled = PETSC_TRUE;
3837: M->assembled = PETSC_FALSE;
3838: }
3839: MatGetOwnershipRange(M,&rstart,&rend);
3840: aij = (Mat_SeqAIJ*)(Mreuse)->data;
3841: ii = aij->i;
3842: jj = aij->j;
3843: aa = aij->a;
3844: for (i=0; i<m; i++) {
3845: row = rstart + i;
3846: nz = ii[i+1] - ii[i];
3847: cwork = jj; jj += nz;
3848: vwork = aa; aa += nz;
3849: MatSetValues_MPIAIJ(M,1,&row,nz,cwork,vwork,INSERT_VALUES);
3850: }
3852: MatAssemblyBegin(M,MAT_FINAL_ASSEMBLY);
3853: MatAssemblyEnd(M,MAT_FINAL_ASSEMBLY);
3854: *newmat = M;
3856: /* save submatrix used in processor for next request */
3857: if (call == MAT_INITIAL_MATRIX) {
3858: PetscObjectCompose((PetscObject)M,"SubMatrix",(PetscObject)Mreuse);
3859: MatDestroy(&Mreuse);
3860: }
3861: return(0);
3862: }
3864: PetscErrorCode MatMPIAIJSetPreallocationCSR_MPIAIJ(Mat B,const PetscInt Ii[],const PetscInt J[],const PetscScalar v[])
3865: {
3866: PetscInt m,cstart, cend,j,nnz,i,d;
3867: PetscInt *d_nnz,*o_nnz,nnz_max = 0,rstart,ii;
3868: const PetscInt *JJ;
3869: PetscScalar *values;
3871: PetscBool nooffprocentries;
3874: if (Ii && Ii[0]) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Ii[0] must be 0 it is %D",Ii[0]);
3876: PetscLayoutSetUp(B->rmap);
3877: PetscLayoutSetUp(B->cmap);
3878: m = B->rmap->n;
3879: cstart = B->cmap->rstart;
3880: cend = B->cmap->rend;
3881: rstart = B->rmap->rstart;
3883: PetscCalloc2(m,&d_nnz,m,&o_nnz);
3885: #if defined(PETSC_USE_DEBUG)
3886: for (i=0; i<m && Ii; i++) {
3887: nnz = Ii[i+1]- Ii[i];
3888: JJ = J + Ii[i];
3889: if (nnz < 0) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Local row %D has a negative %D number of columns",i,nnz);
3890: if (nnz && (JJ[0] < 0)) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Row %D starts with negative column index",i,JJ[0]);
3891: if (nnz && (JJ[nnz-1] >= B->cmap->N)) SETERRQ3(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Row %D ends with too large a column index %D (max allowed %D)",i,JJ[nnz-1],B->cmap->N);
3892: }
3893: #endif
3895: for (i=0; i<m && Ii; i++) {
3896: nnz = Ii[i+1]- Ii[i];
3897: JJ = J + Ii[i];
3898: nnz_max = PetscMax(nnz_max,nnz);
3899: d = 0;
3900: for (j=0; j<nnz; j++) {
3901: if (cstart <= JJ[j] && JJ[j] < cend) d++;
3902: }
3903: d_nnz[i] = d;
3904: o_nnz[i] = nnz - d;
3905: }
3906: MatMPIAIJSetPreallocation(B,0,d_nnz,0,o_nnz);
3907: PetscFree2(d_nnz,o_nnz);
3909: if (v) values = (PetscScalar*)v;
3910: else {
3911: PetscCalloc1(nnz_max+1,&values);
3912: }
3914: for (i=0; i<m && Ii; i++) {
3915: ii = i + rstart;
3916: nnz = Ii[i+1]- Ii[i];
3917: MatSetValues_MPIAIJ(B,1,&ii,nnz,J+Ii[i],values+(v ? Ii[i] : 0),INSERT_VALUES);
3918: }
3919: nooffprocentries = B->nooffprocentries;
3920: B->nooffprocentries = PETSC_TRUE;
3921: MatAssemblyBegin(B,MAT_FINAL_ASSEMBLY);
3922: MatAssemblyEnd(B,MAT_FINAL_ASSEMBLY);
3923: B->nooffprocentries = nooffprocentries;
3925: if (!v) {
3926: PetscFree(values);
3927: }
3928: MatSetOption(B,MAT_NEW_NONZERO_LOCATION_ERR,PETSC_TRUE);
3929: return(0);
3930: }
3932: /*@
3933: MatMPIAIJSetPreallocationCSR - Allocates memory for a sparse parallel matrix in AIJ format
3934: (the default parallel PETSc format).
3936: Collective on MPI_Comm
3938: Input Parameters:
3939: + B - the matrix
3940: . i - the indices into j for the start of each local row (starts with zero)
3941: . j - the column indices for each local row (starts with zero)
3942: - v - optional values in the matrix
3944: Level: developer
3946: Notes:
3947: The i, j, and v arrays ARE copied by this routine into the internal format used by PETSc;
3948: thus you CANNOT change the matrix entries by changing the values of v[] after you have
3949: called this routine. Use MatCreateMPIAIJWithSplitArrays() to avoid needing to copy the arrays.
3951: The i and j indices are 0 based, and i indices are indices corresponding to the local j array.
3953: The format which is used for the sparse matrix input, is equivalent to a
3954: row-major ordering.. i.e for the following matrix, the input data expected is
3955: as shown
3957: $ 1 0 0
3958: $ 2 0 3 P0
3959: $ -------
3960: $ 4 5 6 P1
3961: $
3962: $ Process0 [P0]: rows_owned=[0,1]
3963: $ i = {0,1,3} [size = nrow+1 = 2+1]
3964: $ j = {0,0,2} [size = 3]
3965: $ v = {1,2,3} [size = 3]
3966: $
3967: $ Process1 [P1]: rows_owned=[2]
3968: $ i = {0,3} [size = nrow+1 = 1+1]
3969: $ j = {0,1,2} [size = 3]
3970: $ v = {4,5,6} [size = 3]
3972: .keywords: matrix, aij, compressed row, sparse, parallel
3974: .seealso: MatCreate(), MatCreateSeqAIJ(), MatSetValues(), MatMPIAIJSetPreallocation(), MatCreateAIJ(), MATMPIAIJ,
3975: MatCreateSeqAIJWithArrays(), MatCreateMPIAIJWithSplitArrays()
3976: @*/
3977: PetscErrorCode MatMPIAIJSetPreallocationCSR(Mat B,const PetscInt i[],const PetscInt j[], const PetscScalar v[])
3978: {
3982: PetscTryMethod(B,"MatMPIAIJSetPreallocationCSR_C",(Mat,const PetscInt[],const PetscInt[],const PetscScalar[]),(B,i,j,v));
3983: return(0);
3984: }
3986: /*@C
3987: MatMPIAIJSetPreallocation - Preallocates memory for a sparse parallel matrix in AIJ format
3988: (the default parallel PETSc format). For good matrix assembly performance
3989: the user should preallocate the matrix storage by setting the parameters
3990: d_nz (or d_nnz) and o_nz (or o_nnz). By setting these parameters accurately,
3991: performance can be increased by more than a factor of 50.
3993: Collective on MPI_Comm
3995: Input Parameters:
3996: + B - the matrix
3997: . d_nz - number of nonzeros per row in DIAGONAL portion of local submatrix
3998: (same value is used for all local rows)
3999: . d_nnz - array containing the number of nonzeros in the various rows of the
4000: DIAGONAL portion of the local submatrix (possibly different for each row)
4001: or NULL (PETSC_NULL_INTEGER in Fortran), if d_nz is used to specify the nonzero structure.
4002: The size of this array is equal to the number of local rows, i.e 'm'.
4003: For matrices that will be factored, you must leave room for (and set)
4004: the diagonal entry even if it is zero.
4005: . o_nz - number of nonzeros per row in the OFF-DIAGONAL portion of local
4006: submatrix (same value is used for all local rows).
4007: - o_nnz - array containing the number of nonzeros in the various rows of the
4008: OFF-DIAGONAL portion of the local submatrix (possibly different for
4009: each row) or NULL (PETSC_NULL_INTEGER in Fortran), if o_nz is used to specify the nonzero
4010: structure. The size of this array is equal to the number
4011: of local rows, i.e 'm'.
4013: If the *_nnz parameter is given then the *_nz parameter is ignored
4015: The AIJ format (also called the Yale sparse matrix format or
4016: compressed row storage (CSR)), is fully compatible with standard Fortran 77
4017: storage. The stored row and column indices begin with zero.
4018: See Users-Manual: ch_mat for details.
4020: The parallel matrix is partitioned such that the first m0 rows belong to
4021: process 0, the next m1 rows belong to process 1, the next m2 rows belong
4022: to process 2 etc.. where m0,m1,m2... are the input parameter 'm'.
4024: The DIAGONAL portion of the local submatrix of a processor can be defined
4025: as the submatrix which is obtained by extraction the part corresponding to
4026: the rows r1-r2 and columns c1-c2 of the global matrix, where r1 is the
4027: first row that belongs to the processor, r2 is the last row belonging to
4028: the this processor, and c1-c2 is range of indices of the local part of a
4029: vector suitable for applying the matrix to. This is an mxn matrix. In the
4030: common case of a square matrix, the row and column ranges are the same and
4031: the DIAGONAL part is also square. The remaining portion of the local
4032: submatrix (mxN) constitute the OFF-DIAGONAL portion.
4034: If o_nnz, d_nnz are specified, then o_nz, and d_nz are ignored.
4036: You can call MatGetInfo() to get information on how effective the preallocation was;
4037: for example the fields mallocs,nz_allocated,nz_used,nz_unneeded;
4038: You can also run with the option -info and look for messages with the string
4039: malloc in them to see if additional memory allocation was needed.
4041: Example usage:
4043: Consider the following 8x8 matrix with 34 non-zero values, that is
4044: assembled across 3 processors. Lets assume that proc0 owns 3 rows,
4045: proc1 owns 3 rows, proc2 owns 2 rows. This division can be shown
4046: as follows:
4048: .vb
4049: 1 2 0 | 0 3 0 | 0 4
4050: Proc0 0 5 6 | 7 0 0 | 8 0
4051: 9 0 10 | 11 0 0 | 12 0
4052: -------------------------------------
4053: 13 0 14 | 15 16 17 | 0 0
4054: Proc1 0 18 0 | 19 20 21 | 0 0
4055: 0 0 0 | 22 23 0 | 24 0
4056: -------------------------------------
4057: Proc2 25 26 27 | 0 0 28 | 29 0
4058: 30 0 0 | 31 32 33 | 0 34
4059: .ve
4061: This can be represented as a collection of submatrices as:
4063: .vb
4064: A B C
4065: D E F
4066: G H I
4067: .ve
4069: Where the submatrices A,B,C are owned by proc0, D,E,F are
4070: owned by proc1, G,H,I are owned by proc2.
4072: The 'm' parameters for proc0,proc1,proc2 are 3,3,2 respectively.
4073: The 'n' parameters for proc0,proc1,proc2 are 3,3,2 respectively.
4074: The 'M','N' parameters are 8,8, and have the same values on all procs.
4076: The DIAGONAL submatrices corresponding to proc0,proc1,proc2 are
4077: submatrices [A], [E], [I] respectively. The OFF-DIAGONAL submatrices
4078: corresponding to proc0,proc1,proc2 are [BC], [DF], [GH] respectively.
4079: Internally, each processor stores the DIAGONAL part, and the OFF-DIAGONAL
4080: part as SeqAIJ matrices. for eg: proc1 will store [E] as a SeqAIJ
4081: matrix, ans [DF] as another SeqAIJ matrix.
4083: When d_nz, o_nz parameters are specified, d_nz storage elements are
4084: allocated for every row of the local diagonal submatrix, and o_nz
4085: storage locations are allocated for every row of the OFF-DIAGONAL submat.
4086: One way to choose d_nz and o_nz is to use the max nonzerors per local
4087: rows for each of the local DIAGONAL, and the OFF-DIAGONAL submatrices.
4088: In this case, the values of d_nz,o_nz are:
4089: .vb
4090: proc0 : dnz = 2, o_nz = 2
4091: proc1 : dnz = 3, o_nz = 2
4092: proc2 : dnz = 1, o_nz = 4
4093: .ve
4094: We are allocating m*(d_nz+o_nz) storage locations for every proc. This
4095: translates to 3*(2+2)=12 for proc0, 3*(3+2)=15 for proc1, 2*(1+4)=10
4096: for proc3. i.e we are using 12+15+10=37 storage locations to store
4097: 34 values.
4099: When d_nnz, o_nnz parameters are specified, the storage is specified
4100: for every row, coresponding to both DIAGONAL and OFF-DIAGONAL submatrices.
4101: In the above case the values for d_nnz,o_nnz are:
4102: .vb
4103: proc0: d_nnz = [2,2,2] and o_nnz = [2,2,2]
4104: proc1: d_nnz = [3,3,2] and o_nnz = [2,1,1]
4105: proc2: d_nnz = [1,1] and o_nnz = [4,4]
4106: .ve
4107: Here the space allocated is sum of all the above values i.e 34, and
4108: hence pre-allocation is perfect.
4110: Level: intermediate
4112: .keywords: matrix, aij, compressed row, sparse, parallel
4114: .seealso: MatCreate(), MatCreateSeqAIJ(), MatSetValues(), MatCreateAIJ(), MatMPIAIJSetPreallocationCSR(),
4115: MATMPIAIJ, MatGetInfo(), PetscSplitOwnership()
4116: @*/
4117: PetscErrorCode MatMPIAIJSetPreallocation(Mat B,PetscInt d_nz,const PetscInt d_nnz[],PetscInt o_nz,const PetscInt o_nnz[])
4118: {
4124: PetscTryMethod(B,"MatMPIAIJSetPreallocation_C",(Mat,PetscInt,const PetscInt[],PetscInt,const PetscInt[]),(B,d_nz,d_nnz,o_nz,o_nnz));
4125: return(0);
4126: }
4128: /*@
4129: MatCreateMPIAIJWithArrays - creates a MPI AIJ matrix using arrays that contain in standard
4130: CSR format the local rows.
4132: Collective on MPI_Comm
4134: Input Parameters:
4135: + comm - MPI communicator
4136: . m - number of local rows (Cannot be PETSC_DECIDE)
4137: . n - This value should be the same as the local size used in creating the
4138: x vector for the matrix-vector product y = Ax. (or PETSC_DECIDE to have
4139: calculated if N is given) For square matrices n is almost always m.
4140: . M - number of global rows (or PETSC_DETERMINE to have calculated if m is given)
4141: . N - number of global columns (or PETSC_DETERMINE to have calculated if n is given)
4142: . i - row indices; that is i[0] = 0, i[row] = i[row-1] + number of elements in that row of the matrix
4143: . j - column indices
4144: - a - matrix values
4146: Output Parameter:
4147: . mat - the matrix
4149: Level: intermediate
4151: Notes:
4152: The i, j, and a arrays ARE copied by this routine into the internal format used by PETSc;
4153: thus you CANNOT change the matrix entries by changing the values of a[] after you have
4154: called this routine. Use MatCreateMPIAIJWithSplitArrays() to avoid needing to copy the arrays.
4156: The i and j indices are 0 based, and i indices are indices corresponding to the local j array.
4158: The format which is used for the sparse matrix input, is equivalent to a
4159: row-major ordering.. i.e for the following matrix, the input data expected is
4160: as shown
4162: $ 1 0 0
4163: $ 2 0 3 P0
4164: $ -------
4165: $ 4 5 6 P1
4166: $
4167: $ Process0 [P0]: rows_owned=[0,1]
4168: $ i = {0,1,3} [size = nrow+1 = 2+1]
4169: $ j = {0,0,2} [size = 3]
4170: $ v = {1,2,3} [size = 3]
4171: $
4172: $ Process1 [P1]: rows_owned=[2]
4173: $ i = {0,3} [size = nrow+1 = 1+1]
4174: $ j = {0,1,2} [size = 3]
4175: $ v = {4,5,6} [size = 3]
4177: .keywords: matrix, aij, compressed row, sparse, parallel
4179: .seealso: MatCreate(), MatCreateSeqAIJ(), MatSetValues(), MatMPIAIJSetPreallocation(), MatMPIAIJSetPreallocationCSR(),
4180: MATMPIAIJ, MatCreateAIJ(), MatCreateMPIAIJWithSplitArrays()
4181: @*/
4182: PetscErrorCode MatCreateMPIAIJWithArrays(MPI_Comm comm,PetscInt m,PetscInt n,PetscInt M,PetscInt N,const PetscInt i[],const PetscInt j[],const PetscScalar a[],Mat *mat)
4183: {
4187: if (i && i[0]) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"i (row indices) must start with 0");
4188: if (m < 0) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"local number of rows (m) cannot be PETSC_DECIDE, or negative");
4189: MatCreate(comm,mat);
4190: MatSetSizes(*mat,m,n,M,N);
4191: /* MatSetBlockSizes(M,bs,cbs); */
4192: MatSetType(*mat,MATMPIAIJ);
4193: MatMPIAIJSetPreallocationCSR(*mat,i,j,a);
4194: return(0);
4195: }
4197: /*@C
4198: MatCreateAIJ - Creates a sparse parallel matrix in AIJ format
4199: (the default parallel PETSc format). For good matrix assembly performance
4200: the user should preallocate the matrix storage by setting the parameters
4201: d_nz (or d_nnz) and o_nz (or o_nnz). By setting these parameters accurately,
4202: performance can be increased by more than a factor of 50.
4204: Collective on MPI_Comm
4206: Input Parameters:
4207: + comm - MPI communicator
4208: . m - number of local rows (or PETSC_DECIDE to have calculated if M is given)
4209: This value should be the same as the local size used in creating the
4210: y vector for the matrix-vector product y = Ax.
4211: . n - This value should be the same as the local size used in creating the
4212: x vector for the matrix-vector product y = Ax. (or PETSC_DECIDE to have
4213: calculated if N is given) For square matrices n is almost always m.
4214: . M - number of global rows (or PETSC_DETERMINE to have calculated if m is given)
4215: . N - number of global columns (or PETSC_DETERMINE to have calculated if n is given)
4216: . d_nz - number of nonzeros per row in DIAGONAL portion of local submatrix
4217: (same value is used for all local rows)
4218: . d_nnz - array containing the number of nonzeros in the various rows of the
4219: DIAGONAL portion of the local submatrix (possibly different for each row)
4220: or NULL, if d_nz is used to specify the nonzero structure.
4221: The size of this array is equal to the number of local rows, i.e 'm'.
4222: . o_nz - number of nonzeros per row in the OFF-DIAGONAL portion of local
4223: submatrix (same value is used for all local rows).
4224: - o_nnz - array containing the number of nonzeros in the various rows of the
4225: OFF-DIAGONAL portion of the local submatrix (possibly different for
4226: each row) or NULL, if o_nz is used to specify the nonzero
4227: structure. The size of this array is equal to the number
4228: of local rows, i.e 'm'.
4230: Output Parameter:
4231: . A - the matrix
4233: It is recommended that one use the MatCreate(), MatSetType() and/or MatSetFromOptions(),
4234: MatXXXXSetPreallocation() paradgm instead of this routine directly.
4235: [MatXXXXSetPreallocation() is, for example, MatSeqAIJSetPreallocation]
4237: Notes:
4238: If the *_nnz parameter is given then the *_nz parameter is ignored
4240: m,n,M,N parameters specify the size of the matrix, and its partitioning across
4241: processors, while d_nz,d_nnz,o_nz,o_nnz parameters specify the approximate
4242: storage requirements for this matrix.
4244: If PETSC_DECIDE or PETSC_DETERMINE is used for a particular argument on one
4245: processor than it must be used on all processors that share the object for
4246: that argument.
4248: The user MUST specify either the local or global matrix dimensions
4249: (possibly both).
4251: The parallel matrix is partitioned across processors such that the
4252: first m0 rows belong to process 0, the next m1 rows belong to
4253: process 1, the next m2 rows belong to process 2 etc.. where
4254: m0,m1,m2,.. are the input parameter 'm'. i.e each processor stores
4255: values corresponding to [m x N] submatrix.
4257: The columns are logically partitioned with the n0 columns belonging
4258: to 0th partition, the next n1 columns belonging to the next
4259: partition etc.. where n0,n1,n2... are the input parameter 'n'.
4261: The DIAGONAL portion of the local submatrix on any given processor
4262: is the submatrix corresponding to the rows and columns m,n
4263: corresponding to the given processor. i.e diagonal matrix on
4264: process 0 is [m0 x n0], diagonal matrix on process 1 is [m1 x n1]
4265: etc. The remaining portion of the local submatrix [m x (N-n)]
4266: constitute the OFF-DIAGONAL portion. The example below better
4267: illustrates this concept.
4269: For a square global matrix we define each processor's diagonal portion
4270: to be its local rows and the corresponding columns (a square submatrix);
4271: each processor's off-diagonal portion encompasses the remainder of the
4272: local matrix (a rectangular submatrix).
4274: If o_nnz, d_nnz are specified, then o_nz, and d_nz are ignored.
4276: When calling this routine with a single process communicator, a matrix of
4277: type SEQAIJ is returned. If a matrix of type MPIAIJ is desired for this
4278: type of communicator, use the construction mechanism
4279: .vb
4280: MatCreate(...,&A); MatSetType(A,MATMPIAIJ); MatSetSizes(A, m,n,M,N); MatMPIAIJSetPreallocation(A,...);
4281: .ve
4283: $ MatCreate(...,&A);
4284: $ MatSetType(A,MATMPIAIJ);
4285: $ MatSetSizes(A, m,n,M,N);
4286: $ MatMPIAIJSetPreallocation(A,...);
4288: By default, this format uses inodes (identical nodes) when possible.
4289: We search for consecutive rows with the same nonzero structure, thereby
4290: reusing matrix information to achieve increased efficiency.
4292: Options Database Keys:
4293: + -mat_no_inode - Do not use inodes
4294: - -mat_inode_limit <limit> - Sets inode limit (max limit=5)
4298: Example usage:
4300: Consider the following 8x8 matrix with 34 non-zero values, that is
4301: assembled across 3 processors. Lets assume that proc0 owns 3 rows,
4302: proc1 owns 3 rows, proc2 owns 2 rows. This division can be shown
4303: as follows
4305: .vb
4306: 1 2 0 | 0 3 0 | 0 4
4307: Proc0 0 5 6 | 7 0 0 | 8 0
4308: 9 0 10 | 11 0 0 | 12 0
4309: -------------------------------------
4310: 13 0 14 | 15 16 17 | 0 0
4311: Proc1 0 18 0 | 19 20 21 | 0 0
4312: 0 0 0 | 22 23 0 | 24 0
4313: -------------------------------------
4314: Proc2 25 26 27 | 0 0 28 | 29 0
4315: 30 0 0 | 31 32 33 | 0 34
4316: .ve
4318: This can be represented as a collection of submatrices as
4320: .vb
4321: A B C
4322: D E F
4323: G H I
4324: .ve
4326: Where the submatrices A,B,C are owned by proc0, D,E,F are
4327: owned by proc1, G,H,I are owned by proc2.
4329: The 'm' parameters for proc0,proc1,proc2 are 3,3,2 respectively.
4330: The 'n' parameters for proc0,proc1,proc2 are 3,3,2 respectively.
4331: The 'M','N' parameters are 8,8, and have the same values on all procs.
4333: The DIAGONAL submatrices corresponding to proc0,proc1,proc2 are
4334: submatrices [A], [E], [I] respectively. The OFF-DIAGONAL submatrices
4335: corresponding to proc0,proc1,proc2 are [BC], [DF], [GH] respectively.
4336: Internally, each processor stores the DIAGONAL part, and the OFF-DIAGONAL
4337: part as SeqAIJ matrices. for eg: proc1 will store [E] as a SeqAIJ
4338: matrix, ans [DF] as another SeqAIJ matrix.
4340: When d_nz, o_nz parameters are specified, d_nz storage elements are
4341: allocated for every row of the local diagonal submatrix, and o_nz
4342: storage locations are allocated for every row of the OFF-DIAGONAL submat.
4343: One way to choose d_nz and o_nz is to use the max nonzerors per local
4344: rows for each of the local DIAGONAL, and the OFF-DIAGONAL submatrices.
4345: In this case, the values of d_nz,o_nz are
4346: .vb
4347: proc0 : dnz = 2, o_nz = 2
4348: proc1 : dnz = 3, o_nz = 2
4349: proc2 : dnz = 1, o_nz = 4
4350: .ve
4351: We are allocating m*(d_nz+o_nz) storage locations for every proc. This
4352: translates to 3*(2+2)=12 for proc0, 3*(3+2)=15 for proc1, 2*(1+4)=10
4353: for proc3. i.e we are using 12+15+10=37 storage locations to store
4354: 34 values.
4356: When d_nnz, o_nnz parameters are specified, the storage is specified
4357: for every row, coresponding to both DIAGONAL and OFF-DIAGONAL submatrices.
4358: In the above case the values for d_nnz,o_nnz are
4359: .vb
4360: proc0: d_nnz = [2,2,2] and o_nnz = [2,2,2]
4361: proc1: d_nnz = [3,3,2] and o_nnz = [2,1,1]
4362: proc2: d_nnz = [1,1] and o_nnz = [4,4]
4363: .ve
4364: Here the space allocated is sum of all the above values i.e 34, and
4365: hence pre-allocation is perfect.
4367: Level: intermediate
4369: .keywords: matrix, aij, compressed row, sparse, parallel
4371: .seealso: MatCreate(), MatCreateSeqAIJ(), MatSetValues(), MatMPIAIJSetPreallocation(), MatMPIAIJSetPreallocationCSR(),
4372: MATMPIAIJ, MatCreateMPIAIJWithArrays()
4373: @*/
4374: PetscErrorCode MatCreateAIJ(MPI_Comm comm,PetscInt m,PetscInt n,PetscInt M,PetscInt N,PetscInt d_nz,const PetscInt d_nnz[],PetscInt o_nz,const PetscInt o_nnz[],Mat *A)
4375: {
4377: PetscMPIInt size;
4380: MatCreate(comm,A);
4381: MatSetSizes(*A,m,n,M,N);
4382: MPI_Comm_size(comm,&size);
4383: if (size > 1) {
4384: MatSetType(*A,MATMPIAIJ);
4385: MatMPIAIJSetPreallocation(*A,d_nz,d_nnz,o_nz,o_nnz);
4386: } else {
4387: MatSetType(*A,MATSEQAIJ);
4388: MatSeqAIJSetPreallocation(*A,d_nz,d_nnz);
4389: }
4390: return(0);
4391: }
4393: PetscErrorCode MatMPIAIJGetSeqAIJ(Mat A,Mat *Ad,Mat *Ao,const PetscInt *colmap[])
4394: {
4395: Mat_MPIAIJ *a = (Mat_MPIAIJ*)A->data;
4396: PetscBool flg;
4400: PetscStrbeginswith(((PetscObject)A)->type_name,MATMPIAIJ,&flg);
4401: if (!flg) SETERRQ(PetscObjectComm((PetscObject)A),PETSC_ERR_SUP,"This function requires a MATMPIAIJ matrix as input");
4402: if (Ad) *Ad = a->A;
4403: if (Ao) *Ao = a->B;
4404: if (colmap) *colmap = a->garray;
4405: return(0);
4406: }
4408: PetscErrorCode MatCreateMPIMatConcatenateSeqMat_MPIAIJ(MPI_Comm comm,Mat inmat,PetscInt n,MatReuse scall,Mat *outmat)
4409: {
4411: PetscInt m,N,i,rstart,nnz,Ii;
4412: PetscInt *indx;
4413: PetscScalar *values;
4416: MatGetSize(inmat,&m,&N);
4417: if (scall == MAT_INITIAL_MATRIX) { /* symbolic phase */
4418: PetscInt *dnz,*onz,sum,bs,cbs;
4420: if (n == PETSC_DECIDE) {
4421: PetscSplitOwnership(comm,&n,&N);
4422: }
4423: /* Check sum(n) = N */
4424: MPIU_Allreduce(&n,&sum,1,MPIU_INT,MPI_SUM,comm);
4425: if (sum != N) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_INCOMP,"Sum of local columns %D != global columns %D",sum,N);
4427: MPI_Scan(&m, &rstart,1,MPIU_INT,MPI_SUM,comm);
4428: rstart -= m;
4430: MatPreallocateInitialize(comm,m,n,dnz,onz);
4431: for (i=0; i<m; i++) {
4432: MatGetRow_SeqAIJ(inmat,i,&nnz,&indx,NULL);
4433: MatPreallocateSet(i+rstart,nnz,indx,dnz,onz);
4434: MatRestoreRow_SeqAIJ(inmat,i,&nnz,&indx,NULL);
4435: }
4437: MatCreate(comm,outmat);
4438: MatSetSizes(*outmat,m,n,PETSC_DETERMINE,PETSC_DETERMINE);
4439: MatGetBlockSizes(inmat,&bs,&cbs);
4440: MatSetBlockSizes(*outmat,bs,cbs);
4441: MatSetType(*outmat,MATAIJ);
4442: MatSeqAIJSetPreallocation(*outmat,0,dnz);
4443: MatMPIAIJSetPreallocation(*outmat,0,dnz,0,onz);
4444: MatPreallocateFinalize(dnz,onz);
4445: }
4447: /* numeric phase */
4448: MatGetOwnershipRange(*outmat,&rstart,NULL);
4449: for (i=0; i<m; i++) {
4450: MatGetRow_SeqAIJ(inmat,i,&nnz,&indx,&values);
4451: Ii = i + rstart;
4452: MatSetValues(*outmat,1,&Ii,nnz,indx,values,INSERT_VALUES);
4453: MatRestoreRow_SeqAIJ(inmat,i,&nnz,&indx,&values);
4454: }
4455: MatAssemblyBegin(*outmat,MAT_FINAL_ASSEMBLY);
4456: MatAssemblyEnd(*outmat,MAT_FINAL_ASSEMBLY);
4457: return(0);
4458: }
4460: PetscErrorCode MatFileSplit(Mat A,char *outfile)
4461: {
4462: PetscErrorCode ierr;
4463: PetscMPIInt rank;
4464: PetscInt m,N,i,rstart,nnz;
4465: size_t len;
4466: const PetscInt *indx;
4467: PetscViewer out;
4468: char *name;
4469: Mat B;
4470: const PetscScalar *values;
4473: MatGetLocalSize(A,&m,0);
4474: MatGetSize(A,0,&N);
4475: /* Should this be the type of the diagonal block of A? */
4476: MatCreate(PETSC_COMM_SELF,&B);
4477: MatSetSizes(B,m,N,m,N);
4478: MatSetBlockSizesFromMats(B,A,A);
4479: MatSetType(B,MATSEQAIJ);
4480: MatSeqAIJSetPreallocation(B,0,NULL);
4481: MatGetOwnershipRange(A,&rstart,0);
4482: for (i=0; i<m; i++) {
4483: MatGetRow(A,i+rstart,&nnz,&indx,&values);
4484: MatSetValues(B,1,&i,nnz,indx,values,INSERT_VALUES);
4485: MatRestoreRow(A,i+rstart,&nnz,&indx,&values);
4486: }
4487: MatAssemblyBegin(B,MAT_FINAL_ASSEMBLY);
4488: MatAssemblyEnd(B,MAT_FINAL_ASSEMBLY);
4490: MPI_Comm_rank(PetscObjectComm((PetscObject)A),&rank);
4491: PetscStrlen(outfile,&len);
4492: PetscMalloc1(len+5,&name);
4493: sprintf(name,"%s.%d",outfile,rank);
4494: PetscViewerBinaryOpen(PETSC_COMM_SELF,name,FILE_MODE_APPEND,&out);
4495: PetscFree(name);
4496: MatView(B,out);
4497: PetscViewerDestroy(&out);
4498: MatDestroy(&B);
4499: return(0);
4500: }
4502: PetscErrorCode MatDestroy_MPIAIJ_SeqsToMPI(Mat A)
4503: {
4504: PetscErrorCode ierr;
4505: Mat_Merge_SeqsToMPI *merge;
4506: PetscContainer container;
4509: PetscObjectQuery((PetscObject)A,"MatMergeSeqsToMPI",(PetscObject*)&container);
4510: if (container) {
4511: PetscContainerGetPointer(container,(void**)&merge);
4512: PetscFree(merge->id_r);
4513: PetscFree(merge->len_s);
4514: PetscFree(merge->len_r);
4515: PetscFree(merge->bi);
4516: PetscFree(merge->bj);
4517: PetscFree(merge->buf_ri[0]);
4518: PetscFree(merge->buf_ri);
4519: PetscFree(merge->buf_rj[0]);
4520: PetscFree(merge->buf_rj);
4521: PetscFree(merge->coi);
4522: PetscFree(merge->coj);
4523: PetscFree(merge->owners_co);
4524: PetscLayoutDestroy(&merge->rowmap);
4525: PetscFree(merge);
4526: PetscObjectCompose((PetscObject)A,"MatMergeSeqsToMPI",0);
4527: }
4528: MatDestroy_MPIAIJ(A);
4529: return(0);
4530: }
4532: #include <../src/mat/utils/freespace.h>
4533: #include <petscbt.h>
4535: PetscErrorCode MatCreateMPIAIJSumSeqAIJNumeric(Mat seqmat,Mat mpimat)
4536: {
4537: PetscErrorCode ierr;
4538: MPI_Comm comm;
4539: Mat_SeqAIJ *a =(Mat_SeqAIJ*)seqmat->data;
4540: PetscMPIInt size,rank,taga,*len_s;
4541: PetscInt N=mpimat->cmap->N,i,j,*owners,*ai=a->i,*aj;
4542: PetscInt proc,m;
4543: PetscInt **buf_ri,**buf_rj;
4544: PetscInt k,anzi,*bj_i,*bi,*bj,arow,bnzi,nextaj;
4545: PetscInt nrows,**buf_ri_k,**nextrow,**nextai;
4546: MPI_Request *s_waits,*r_waits;
4547: MPI_Status *status;
4548: MatScalar *aa=a->a;
4549: MatScalar **abuf_r,*ba_i;
4550: Mat_Merge_SeqsToMPI *merge;
4551: PetscContainer container;
4554: PetscObjectGetComm((PetscObject)mpimat,&comm);
4555: PetscLogEventBegin(MAT_Seqstompinum,seqmat,0,0,0);
4557: MPI_Comm_size(comm,&size);
4558: MPI_Comm_rank(comm,&rank);
4560: PetscObjectQuery((PetscObject)mpimat,"MatMergeSeqsToMPI",(PetscObject*)&container);
4561: PetscContainerGetPointer(container,(void**)&merge);
4563: bi = merge->bi;
4564: bj = merge->bj;
4565: buf_ri = merge->buf_ri;
4566: buf_rj = merge->buf_rj;
4568: PetscMalloc1(size,&status);
4569: owners = merge->rowmap->range;
4570: len_s = merge->len_s;
4572: /* send and recv matrix values */
4573: /*-----------------------------*/
4574: PetscObjectGetNewTag((PetscObject)mpimat,&taga);
4575: PetscPostIrecvScalar(comm,taga,merge->nrecv,merge->id_r,merge->len_r,&abuf_r,&r_waits);
4577: PetscMalloc1(merge->nsend+1,&s_waits);
4578: for (proc=0,k=0; proc<size; proc++) {
4579: if (!len_s[proc]) continue;
4580: i = owners[proc];
4581: MPI_Isend(aa+ai[i],len_s[proc],MPIU_MATSCALAR,proc,taga,comm,s_waits+k);
4582: k++;
4583: }
4585: if (merge->nrecv) {MPI_Waitall(merge->nrecv,r_waits,status);}
4586: if (merge->nsend) {MPI_Waitall(merge->nsend,s_waits,status);}
4587: PetscFree(status);
4589: PetscFree(s_waits);
4590: PetscFree(r_waits);
4592: /* insert mat values of mpimat */
4593: /*----------------------------*/
4594: PetscMalloc1(N,&ba_i);
4595: PetscMalloc3(merge->nrecv,&buf_ri_k,merge->nrecv,&nextrow,merge->nrecv,&nextai);
4597: for (k=0; k<merge->nrecv; k++) {
4598: buf_ri_k[k] = buf_ri[k]; /* beginning of k-th recved i-structure */
4599: nrows = *(buf_ri_k[k]);
4600: nextrow[k] = buf_ri_k[k]+1; /* next row number of k-th recved i-structure */
4601: nextai[k] = buf_ri_k[k] + (nrows + 1); /* poins to the next i-structure of k-th recved i-structure */
4602: }
4604: /* set values of ba */
4605: m = merge->rowmap->n;
4606: for (i=0; i<m; i++) {
4607: arow = owners[rank] + i;
4608: bj_i = bj+bi[i]; /* col indices of the i-th row of mpimat */
4609: bnzi = bi[i+1] - bi[i];
4610: PetscMemzero(ba_i,bnzi*sizeof(PetscScalar));
4612: /* add local non-zero vals of this proc's seqmat into ba */
4613: anzi = ai[arow+1] - ai[arow];
4614: aj = a->j + ai[arow];
4615: aa = a->a + ai[arow];
4616: nextaj = 0;
4617: for (j=0; nextaj<anzi; j++) {
4618: if (*(bj_i + j) == aj[nextaj]) { /* bcol == acol */
4619: ba_i[j] += aa[nextaj++];
4620: }
4621: }
4623: /* add received vals into ba */
4624: for (k=0; k<merge->nrecv; k++) { /* k-th received message */
4625: /* i-th row */
4626: if (i == *nextrow[k]) {
4627: anzi = *(nextai[k]+1) - *nextai[k];
4628: aj = buf_rj[k] + *(nextai[k]);
4629: aa = abuf_r[k] + *(nextai[k]);
4630: nextaj = 0;
4631: for (j=0; nextaj<anzi; j++) {
4632: if (*(bj_i + j) == aj[nextaj]) { /* bcol == acol */
4633: ba_i[j] += aa[nextaj++];
4634: }
4635: }
4636: nextrow[k]++; nextai[k]++;
4637: }
4638: }
4639: MatSetValues(mpimat,1,&arow,bnzi,bj_i,ba_i,INSERT_VALUES);
4640: }
4641: MatAssemblyBegin(mpimat,MAT_FINAL_ASSEMBLY);
4642: MatAssemblyEnd(mpimat,MAT_FINAL_ASSEMBLY);
4644: PetscFree(abuf_r[0]);
4645: PetscFree(abuf_r);
4646: PetscFree(ba_i);
4647: PetscFree3(buf_ri_k,nextrow,nextai);
4648: PetscLogEventEnd(MAT_Seqstompinum,seqmat,0,0,0);
4649: return(0);
4650: }
4652: PetscErrorCode MatCreateMPIAIJSumSeqAIJSymbolic(MPI_Comm comm,Mat seqmat,PetscInt m,PetscInt n,Mat *mpimat)
4653: {
4654: PetscErrorCode ierr;
4655: Mat B_mpi;
4656: Mat_SeqAIJ *a=(Mat_SeqAIJ*)seqmat->data;
4657: PetscMPIInt size,rank,tagi,tagj,*len_s,*len_si,*len_ri;
4658: PetscInt **buf_rj,**buf_ri,**buf_ri_k;
4659: PetscInt M=seqmat->rmap->n,N=seqmat->cmap->n,i,*owners,*ai=a->i,*aj=a->j;
4660: PetscInt len,proc,*dnz,*onz,bs,cbs;
4661: PetscInt k,anzi,*bi,*bj,*lnk,nlnk,arow,bnzi,nspacedouble=0;
4662: PetscInt nrows,*buf_s,*buf_si,*buf_si_i,**nextrow,**nextai;
4663: MPI_Request *si_waits,*sj_waits,*ri_waits,*rj_waits;
4664: MPI_Status *status;
4665: PetscFreeSpaceList free_space=NULL,current_space=NULL;
4666: PetscBT lnkbt;
4667: Mat_Merge_SeqsToMPI *merge;
4668: PetscContainer container;
4671: PetscLogEventBegin(MAT_Seqstompisym,seqmat,0,0,0);
4673: /* make sure it is a PETSc comm */
4674: PetscCommDuplicate(comm,&comm,NULL);
4675: MPI_Comm_size(comm,&size);
4676: MPI_Comm_rank(comm,&rank);
4678: PetscNew(&merge);
4679: PetscMalloc1(size,&status);
4681: /* determine row ownership */
4682: /*---------------------------------------------------------*/
4683: PetscLayoutCreate(comm,&merge->rowmap);
4684: PetscLayoutSetLocalSize(merge->rowmap,m);
4685: PetscLayoutSetSize(merge->rowmap,M);
4686: PetscLayoutSetBlockSize(merge->rowmap,1);
4687: PetscLayoutSetUp(merge->rowmap);
4688: PetscMalloc1(size,&len_si);
4689: PetscMalloc1(size,&merge->len_s);
4691: m = merge->rowmap->n;
4692: owners = merge->rowmap->range;
4694: /* determine the number of messages to send, their lengths */
4695: /*---------------------------------------------------------*/
4696: len_s = merge->len_s;
4698: len = 0; /* length of buf_si[] */
4699: merge->nsend = 0;
4700: for (proc=0; proc<size; proc++) {
4701: len_si[proc] = 0;
4702: if (proc == rank) {
4703: len_s[proc] = 0;
4704: } else {
4705: len_si[proc] = owners[proc+1] - owners[proc] + 1;
4706: len_s[proc] = ai[owners[proc+1]] - ai[owners[proc]]; /* num of rows to be sent to [proc] */
4707: }
4708: if (len_s[proc]) {
4709: merge->nsend++;
4710: nrows = 0;
4711: for (i=owners[proc]; i<owners[proc+1]; i++) {
4712: if (ai[i+1] > ai[i]) nrows++;
4713: }
4714: len_si[proc] = 2*(nrows+1);
4715: len += len_si[proc];
4716: }
4717: }
4719: /* determine the number and length of messages to receive for ij-structure */
4720: /*-------------------------------------------------------------------------*/
4721: PetscGatherNumberOfMessages(comm,NULL,len_s,&merge->nrecv);
4722: PetscGatherMessageLengths2(comm,merge->nsend,merge->nrecv,len_s,len_si,&merge->id_r,&merge->len_r,&len_ri);
4724: /* post the Irecv of j-structure */
4725: /*-------------------------------*/
4726: PetscCommGetNewTag(comm,&tagj);
4727: PetscPostIrecvInt(comm,tagj,merge->nrecv,merge->id_r,merge->len_r,&buf_rj,&rj_waits);
4729: /* post the Isend of j-structure */
4730: /*--------------------------------*/
4731: PetscMalloc2(merge->nsend,&si_waits,merge->nsend,&sj_waits);
4733: for (proc=0, k=0; proc<size; proc++) {
4734: if (!len_s[proc]) continue;
4735: i = owners[proc];
4736: MPI_Isend(aj+ai[i],len_s[proc],MPIU_INT,proc,tagj,comm,sj_waits+k);
4737: k++;
4738: }
4740: /* receives and sends of j-structure are complete */
4741: /*------------------------------------------------*/
4742: if (merge->nrecv) {MPI_Waitall(merge->nrecv,rj_waits,status);}
4743: if (merge->nsend) {MPI_Waitall(merge->nsend,sj_waits,status);}
4745: /* send and recv i-structure */
4746: /*---------------------------*/
4747: PetscCommGetNewTag(comm,&tagi);
4748: PetscPostIrecvInt(comm,tagi,merge->nrecv,merge->id_r,len_ri,&buf_ri,&ri_waits);
4750: PetscMalloc1(len+1,&buf_s);
4751: buf_si = buf_s; /* points to the beginning of k-th msg to be sent */
4752: for (proc=0,k=0; proc<size; proc++) {
4753: if (!len_s[proc]) continue;
4754: /* form outgoing message for i-structure:
4755: buf_si[0]: nrows to be sent
4756: [1:nrows]: row index (global)
4757: [nrows+1:2*nrows+1]: i-structure index
4758: */
4759: /*-------------------------------------------*/
4760: nrows = len_si[proc]/2 - 1;
4761: buf_si_i = buf_si + nrows+1;
4762: buf_si[0] = nrows;
4763: buf_si_i[0] = 0;
4764: nrows = 0;
4765: for (i=owners[proc]; i<owners[proc+1]; i++) {
4766: anzi = ai[i+1] - ai[i];
4767: if (anzi) {
4768: buf_si_i[nrows+1] = buf_si_i[nrows] + anzi; /* i-structure */
4769: buf_si[nrows+1] = i-owners[proc]; /* local row index */
4770: nrows++;
4771: }
4772: }
4773: MPI_Isend(buf_si,len_si[proc],MPIU_INT,proc,tagi,comm,si_waits+k);
4774: k++;
4775: buf_si += len_si[proc];
4776: }
4778: if (merge->nrecv) {MPI_Waitall(merge->nrecv,ri_waits,status);}
4779: if (merge->nsend) {MPI_Waitall(merge->nsend,si_waits,status);}
4781: PetscInfo2(seqmat,"nsend: %D, nrecv: %D\n",merge->nsend,merge->nrecv);
4782: for (i=0; i<merge->nrecv; i++) {
4783: PetscInfo3(seqmat,"recv len_ri=%D, len_rj=%D from [%D]\n",len_ri[i],merge->len_r[i],merge->id_r[i]);
4784: }
4786: PetscFree(len_si);
4787: PetscFree(len_ri);
4788: PetscFree(rj_waits);
4789: PetscFree2(si_waits,sj_waits);
4790: PetscFree(ri_waits);
4791: PetscFree(buf_s);
4792: PetscFree(status);
4794: /* compute a local seq matrix in each processor */
4795: /*----------------------------------------------*/
4796: /* allocate bi array and free space for accumulating nonzero column info */
4797: PetscMalloc1(m+1,&bi);
4798: bi[0] = 0;
4800: /* create and initialize a linked list */
4801: nlnk = N+1;
4802: PetscLLCreate(N,N,nlnk,lnk,lnkbt);
4804: /* initial FreeSpace size is 2*(num of local nnz(seqmat)) */
4805: len = ai[owners[rank+1]] - ai[owners[rank]];
4806: PetscFreeSpaceGet(PetscIntMultTruncate(2,len)+1,&free_space);
4808: current_space = free_space;
4810: /* determine symbolic info for each local row */
4811: PetscMalloc3(merge->nrecv,&buf_ri_k,merge->nrecv,&nextrow,merge->nrecv,&nextai);
4813: for (k=0; k<merge->nrecv; k++) {
4814: buf_ri_k[k] = buf_ri[k]; /* beginning of k-th recved i-structure */
4815: nrows = *buf_ri_k[k];
4816: nextrow[k] = buf_ri_k[k] + 1; /* next row number of k-th recved i-structure */
4817: nextai[k] = buf_ri_k[k] + (nrows + 1); /* poins to the next i-structure of k-th recved i-structure */
4818: }
4820: MatPreallocateInitialize(comm,m,n,dnz,onz);
4821: len = 0;
4822: for (i=0; i<m; i++) {
4823: bnzi = 0;
4824: /* add local non-zero cols of this proc's seqmat into lnk */
4825: arow = owners[rank] + i;
4826: anzi = ai[arow+1] - ai[arow];
4827: aj = a->j + ai[arow];
4828: PetscLLAddSorted(anzi,aj,N,nlnk,lnk,lnkbt);
4829: bnzi += nlnk;
4830: /* add received col data into lnk */
4831: for (k=0; k<merge->nrecv; k++) { /* k-th received message */
4832: if (i == *nextrow[k]) { /* i-th row */
4833: anzi = *(nextai[k]+1) - *nextai[k];
4834: aj = buf_rj[k] + *nextai[k];
4835: PetscLLAddSorted(anzi,aj,N,nlnk,lnk,lnkbt);
4836: bnzi += nlnk;
4837: nextrow[k]++; nextai[k]++;
4838: }
4839: }
4840: if (len < bnzi) len = bnzi; /* =max(bnzi) */
4842: /* if free space is not available, make more free space */
4843: if (current_space->local_remaining<bnzi) {
4844: PetscFreeSpaceGet(PetscIntSumTruncate(bnzi,current_space->total_array_size),¤t_space);
4845: nspacedouble++;
4846: }
4847: /* copy data into free space, then initialize lnk */
4848: PetscLLClean(N,N,bnzi,lnk,current_space->array,lnkbt);
4849: MatPreallocateSet(i+owners[rank],bnzi,current_space->array,dnz,onz);
4851: current_space->array += bnzi;
4852: current_space->local_used += bnzi;
4853: current_space->local_remaining -= bnzi;
4855: bi[i+1] = bi[i] + bnzi;
4856: }
4858: PetscFree3(buf_ri_k,nextrow,nextai);
4860: PetscMalloc1(bi[m]+1,&bj);
4861: PetscFreeSpaceContiguous(&free_space,bj);
4862: PetscLLDestroy(lnk,lnkbt);
4864: /* create symbolic parallel matrix B_mpi */
4865: /*---------------------------------------*/
4866: MatGetBlockSizes(seqmat,&bs,&cbs);
4867: MatCreate(comm,&B_mpi);
4868: if (n==PETSC_DECIDE) {
4869: MatSetSizes(B_mpi,m,n,PETSC_DETERMINE,N);
4870: } else {
4871: MatSetSizes(B_mpi,m,n,PETSC_DETERMINE,PETSC_DETERMINE);
4872: }
4873: MatSetBlockSizes(B_mpi,bs,cbs);
4874: MatSetType(B_mpi,MATMPIAIJ);
4875: MatMPIAIJSetPreallocation(B_mpi,0,dnz,0,onz);
4876: MatPreallocateFinalize(dnz,onz);
4877: MatSetOption(B_mpi,MAT_NEW_NONZERO_ALLOCATION_ERR,PETSC_FALSE);
4879: /* B_mpi is not ready for use - assembly will be done by MatCreateMPIAIJSumSeqAIJNumeric() */
4880: B_mpi->assembled = PETSC_FALSE;
4881: B_mpi->ops->destroy = MatDestroy_MPIAIJ_SeqsToMPI;
4882: merge->bi = bi;
4883: merge->bj = bj;
4884: merge->buf_ri = buf_ri;
4885: merge->buf_rj = buf_rj;
4886: merge->coi = NULL;
4887: merge->coj = NULL;
4888: merge->owners_co = NULL;
4890: PetscCommDestroy(&comm);
4892: /* attach the supporting struct to B_mpi for reuse */
4893: PetscContainerCreate(PETSC_COMM_SELF,&container);
4894: PetscContainerSetPointer(container,merge);
4895: PetscObjectCompose((PetscObject)B_mpi,"MatMergeSeqsToMPI",(PetscObject)container);
4896: PetscContainerDestroy(&container);
4897: *mpimat = B_mpi;
4899: PetscLogEventEnd(MAT_Seqstompisym,seqmat,0,0,0);
4900: return(0);
4901: }
4903: /*@C
4904: MatCreateMPIAIJSumSeqAIJ - Creates a MATMPIAIJ matrix by adding sequential
4905: matrices from each processor
4907: Collective on MPI_Comm
4909: Input Parameters:
4910: + comm - the communicators the parallel matrix will live on
4911: . seqmat - the input sequential matrices
4912: . m - number of local rows (or PETSC_DECIDE)
4913: . n - number of local columns (or PETSC_DECIDE)
4914: - scall - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
4916: Output Parameter:
4917: . mpimat - the parallel matrix generated
4919: Level: advanced
4921: Notes:
4922: The dimensions of the sequential matrix in each processor MUST be the same.
4923: The input seqmat is included into the container "Mat_Merge_SeqsToMPI", and will be
4924: destroyed when mpimat is destroyed. Call PetscObjectQuery() to access seqmat.
4925: @*/
4926: PetscErrorCode MatCreateMPIAIJSumSeqAIJ(MPI_Comm comm,Mat seqmat,PetscInt m,PetscInt n,MatReuse scall,Mat *mpimat)
4927: {
4929: PetscMPIInt size;
4932: MPI_Comm_size(comm,&size);
4933: if (size == 1) {
4934: PetscLogEventBegin(MAT_Seqstompi,seqmat,0,0,0);
4935: if (scall == MAT_INITIAL_MATRIX) {
4936: MatDuplicate(seqmat,MAT_COPY_VALUES,mpimat);
4937: } else {
4938: MatCopy(seqmat,*mpimat,SAME_NONZERO_PATTERN);
4939: }
4940: PetscLogEventEnd(MAT_Seqstompi,seqmat,0,0,0);
4941: return(0);
4942: }
4943: PetscLogEventBegin(MAT_Seqstompi,seqmat,0,0,0);
4944: if (scall == MAT_INITIAL_MATRIX) {
4945: MatCreateMPIAIJSumSeqAIJSymbolic(comm,seqmat,m,n,mpimat);
4946: }
4947: MatCreateMPIAIJSumSeqAIJNumeric(seqmat,*mpimat);
4948: PetscLogEventEnd(MAT_Seqstompi,seqmat,0,0,0);
4949: return(0);
4950: }
4952: /*@
4953: MatMPIAIJGetLocalMat - Creates a SeqAIJ from a MATMPIAIJ matrix by taking all its local rows and putting them into a sequential matrix with
4954: mlocal rows and n columns. Where mlocal is the row count obtained with MatGetLocalSize() and n is the global column count obtained
4955: with MatGetSize()
4957: Not Collective
4959: Input Parameters:
4960: + A - the matrix
4961: . scall - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
4963: Output Parameter:
4964: . A_loc - the local sequential matrix generated
4966: Level: developer
4968: .seealso: MatGetOwnershipRange(), MatMPIAIJGetLocalMatCondensed()
4970: @*/
4971: PetscErrorCode MatMPIAIJGetLocalMat(Mat A,MatReuse scall,Mat *A_loc)
4972: {
4974: Mat_MPIAIJ *mpimat=(Mat_MPIAIJ*)A->data;
4975: Mat_SeqAIJ *mat,*a,*b;
4976: PetscInt *ai,*aj,*bi,*bj,*cmap=mpimat->garray;
4977: MatScalar *aa,*ba,*cam;
4978: PetscScalar *ca;
4979: PetscInt am=A->rmap->n,i,j,k,cstart=A->cmap->rstart;
4980: PetscInt *ci,*cj,col,ncols_d,ncols_o,jo;
4981: PetscBool match;
4982: MPI_Comm comm;
4983: PetscMPIInt size;
4986: PetscStrbeginswith(((PetscObject)A)->type_name,MATMPIAIJ,&match);
4987: if (!match) SETERRQ(PetscObjectComm((PetscObject)A), PETSC_ERR_SUP,"Requires MATMPIAIJ matrix as input");
4988: PetscObjectGetComm((PetscObject)A,&comm);
4989: MPI_Comm_size(comm,&size);
4990: if (size == 1 && scall == MAT_REUSE_MATRIX) return(0);
4992: PetscLogEventBegin(MAT_Getlocalmat,A,0,0,0);
4993: a = (Mat_SeqAIJ*)(mpimat->A)->data;
4994: b = (Mat_SeqAIJ*)(mpimat->B)->data;
4995: ai = a->i; aj = a->j; bi = b->i; bj = b->j;
4996: aa = a->a; ba = b->a;
4997: if (scall == MAT_INITIAL_MATRIX) {
4998: if (size == 1) {
4999: MatCreateSeqAIJWithArrays(PETSC_COMM_SELF,am,A->cmap->N,ai,aj,aa,A_loc);
5000: return(0);
5001: }
5003: PetscMalloc1(1+am,&ci);
5004: ci[0] = 0;
5005: for (i=0; i<am; i++) {
5006: ci[i+1] = ci[i] + (ai[i+1] - ai[i]) + (bi[i+1] - bi[i]);
5007: }
5008: PetscMalloc1(1+ci[am],&cj);
5009: PetscMalloc1(1+ci[am],&ca);
5010: k = 0;
5011: for (i=0; i<am; i++) {
5012: ncols_o = bi[i+1] - bi[i];
5013: ncols_d = ai[i+1] - ai[i];
5014: /* off-diagonal portion of A */
5015: for (jo=0; jo<ncols_o; jo++) {
5016: col = cmap[*bj];
5017: if (col >= cstart) break;
5018: cj[k] = col; bj++;
5019: ca[k++] = *ba++;
5020: }
5021: /* diagonal portion of A */
5022: for (j=0; j<ncols_d; j++) {
5023: cj[k] = cstart + *aj++;
5024: ca[k++] = *aa++;
5025: }
5026: /* off-diagonal portion of A */
5027: for (j=jo; j<ncols_o; j++) {
5028: cj[k] = cmap[*bj++];
5029: ca[k++] = *ba++;
5030: }
5031: }
5032: /* put together the new matrix */
5033: MatCreateSeqAIJWithArrays(PETSC_COMM_SELF,am,A->cmap->N,ci,cj,ca,A_loc);
5034: /* MatCreateSeqAIJWithArrays flags matrix so PETSc doesn't free the user's arrays. */
5035: /* Since these are PETSc arrays, change flags to free them as necessary. */
5036: mat = (Mat_SeqAIJ*)(*A_loc)->data;
5037: mat->free_a = PETSC_TRUE;
5038: mat->free_ij = PETSC_TRUE;
5039: mat->nonew = 0;
5040: } else if (scall == MAT_REUSE_MATRIX) {
5041: mat=(Mat_SeqAIJ*)(*A_loc)->data;
5042: ci = mat->i; cj = mat->j; cam = mat->a;
5043: for (i=0; i<am; i++) {
5044: /* off-diagonal portion of A */
5045: ncols_o = bi[i+1] - bi[i];
5046: for (jo=0; jo<ncols_o; jo++) {
5047: col = cmap[*bj];
5048: if (col >= cstart) break;
5049: *cam++ = *ba++; bj++;
5050: }
5051: /* diagonal portion of A */
5052: ncols_d = ai[i+1] - ai[i];
5053: for (j=0; j<ncols_d; j++) *cam++ = *aa++;
5054: /* off-diagonal portion of A */
5055: for (j=jo; j<ncols_o; j++) {
5056: *cam++ = *ba++; bj++;
5057: }
5058: }
5059: } else SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"Invalid MatReuse %d",(int)scall);
5060: PetscLogEventEnd(MAT_Getlocalmat,A,0,0,0);
5061: return(0);
5062: }
5064: /*@C
5065: MatMPIAIJGetLocalMatCondensed - Creates a SeqAIJ matrix from an MATMPIAIJ matrix by taking all its local rows and NON-ZERO columns
5067: Not Collective
5069: Input Parameters:
5070: + A - the matrix
5071: . scall - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
5072: - row, col - index sets of rows and columns to extract (or NULL)
5074: Output Parameter:
5075: . A_loc - the local sequential matrix generated
5077: Level: developer
5079: .seealso: MatGetOwnershipRange(), MatMPIAIJGetLocalMat()
5081: @*/
5082: PetscErrorCode MatMPIAIJGetLocalMatCondensed(Mat A,MatReuse scall,IS *row,IS *col,Mat *A_loc)
5083: {
5084: Mat_MPIAIJ *a=(Mat_MPIAIJ*)A->data;
5086: PetscInt i,start,end,ncols,nzA,nzB,*cmap,imark,*idx;
5087: IS isrowa,iscola;
5088: Mat *aloc;
5089: PetscBool match;
5092: PetscObjectTypeCompare((PetscObject)A,MATMPIAIJ,&match);
5093: if (!match) SETERRQ(PetscObjectComm((PetscObject)A), PETSC_ERR_SUP,"Requires MATMPIAIJ matrix as input");
5094: PetscLogEventBegin(MAT_Getlocalmatcondensed,A,0,0,0);
5095: if (!row) {
5096: start = A->rmap->rstart; end = A->rmap->rend;
5097: ISCreateStride(PETSC_COMM_SELF,end-start,start,1,&isrowa);
5098: } else {
5099: isrowa = *row;
5100: }
5101: if (!col) {
5102: start = A->cmap->rstart;
5103: cmap = a->garray;
5104: nzA = a->A->cmap->n;
5105: nzB = a->B->cmap->n;
5106: PetscMalloc1(nzA+nzB, &idx);
5107: ncols = 0;
5108: for (i=0; i<nzB; i++) {
5109: if (cmap[i] < start) idx[ncols++] = cmap[i];
5110: else break;
5111: }
5112: imark = i;
5113: for (i=0; i<nzA; i++) idx[ncols++] = start + i;
5114: for (i=imark; i<nzB; i++) idx[ncols++] = cmap[i];
5115: ISCreateGeneral(PETSC_COMM_SELF,ncols,idx,PETSC_OWN_POINTER,&iscola);
5116: } else {
5117: iscola = *col;
5118: }
5119: if (scall != MAT_INITIAL_MATRIX) {
5120: PetscMalloc1(1,&aloc);
5121: aloc[0] = *A_loc;
5122: }
5123: MatCreateSubMatrices(A,1,&isrowa,&iscola,scall,&aloc);
5124: if (!col) { /* attach global id of condensed columns */
5125: PetscObjectCompose((PetscObject)aloc[0],"_petsc_GetLocalMatCondensed_iscol",(PetscObject)iscola);
5126: }
5127: *A_loc = aloc[0];
5128: PetscFree(aloc);
5129: if (!row) {
5130: ISDestroy(&isrowa);
5131: }
5132: if (!col) {
5133: ISDestroy(&iscola);
5134: }
5135: PetscLogEventEnd(MAT_Getlocalmatcondensed,A,0,0,0);
5136: return(0);
5137: }
5139: /*@C
5140: MatGetBrowsOfAcols - Creates a SeqAIJ matrix by taking rows of B that equal to nonzero columns of local A
5142: Collective on Mat
5144: Input Parameters:
5145: + A,B - the matrices in mpiaij format
5146: . scall - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
5147: - rowb, colb - index sets of rows and columns of B to extract (or NULL)
5149: Output Parameter:
5150: + rowb, colb - index sets of rows and columns of B to extract
5151: - B_seq - the sequential matrix generated
5153: Level: developer
5155: @*/
5156: PetscErrorCode MatGetBrowsOfAcols(Mat A,Mat B,MatReuse scall,IS *rowb,IS *colb,Mat *B_seq)
5157: {
5158: Mat_MPIAIJ *a=(Mat_MPIAIJ*)A->data;
5160: PetscInt *idx,i,start,ncols,nzA,nzB,*cmap,imark;
5161: IS isrowb,iscolb;
5162: Mat *bseq=NULL;
5165: if (A->cmap->rstart != B->rmap->rstart || A->cmap->rend != B->rmap->rend) {
5166: SETERRQ4(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Matrix local dimensions are incompatible, (%D, %D) != (%D,%D)",A->cmap->rstart,A->cmap->rend,B->rmap->rstart,B->rmap->rend);
5167: }
5168: PetscLogEventBegin(MAT_GetBrowsOfAcols,A,B,0,0);
5170: if (scall == MAT_INITIAL_MATRIX) {
5171: start = A->cmap->rstart;
5172: cmap = a->garray;
5173: nzA = a->A->cmap->n;
5174: nzB = a->B->cmap->n;
5175: PetscMalloc1(nzA+nzB, &idx);
5176: ncols = 0;
5177: for (i=0; i<nzB; i++) { /* row < local row index */
5178: if (cmap[i] < start) idx[ncols++] = cmap[i];
5179: else break;
5180: }
5181: imark = i;
5182: for (i=0; i<nzA; i++) idx[ncols++] = start + i; /* local rows */
5183: for (i=imark; i<nzB; i++) idx[ncols++] = cmap[i]; /* row > local row index */
5184: ISCreateGeneral(PETSC_COMM_SELF,ncols,idx,PETSC_OWN_POINTER,&isrowb);
5185: ISCreateStride(PETSC_COMM_SELF,B->cmap->N,0,1,&iscolb);
5186: } else {
5187: if (!rowb || !colb) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"IS rowb and colb must be provided for MAT_REUSE_MATRIX");
5188: isrowb = *rowb; iscolb = *colb;
5189: PetscMalloc1(1,&bseq);
5190: bseq[0] = *B_seq;
5191: }
5192: MatCreateSubMatrices(B,1,&isrowb,&iscolb,scall,&bseq);
5193: *B_seq = bseq[0];
5194: PetscFree(bseq);
5195: if (!rowb) {
5196: ISDestroy(&isrowb);
5197: } else {
5198: *rowb = isrowb;
5199: }
5200: if (!colb) {
5201: ISDestroy(&iscolb);
5202: } else {
5203: *colb = iscolb;
5204: }
5205: PetscLogEventEnd(MAT_GetBrowsOfAcols,A,B,0,0);
5206: return(0);
5207: }
5209: /*
5210: MatGetBrowsOfAoCols_MPIAIJ - Creates a SeqAIJ matrix by taking rows of B that equal to nonzero columns
5211: of the OFF-DIAGONAL portion of local A
5213: Collective on Mat
5215: Input Parameters:
5216: + A,B - the matrices in mpiaij format
5217: - scall - either MAT_INITIAL_MATRIX or MAT_REUSE_MATRIX
5219: Output Parameter:
5220: + startsj_s - starting point in B's sending j-arrays, saved for MAT_REUSE (or NULL)
5221: . startsj_r - starting point in B's receiving j-arrays, saved for MAT_REUSE (or NULL)
5222: . bufa_ptr - array for sending matrix values, saved for MAT_REUSE (or NULL)
5223: - B_oth - the sequential matrix generated with size aBn=a->B->cmap->n by B->cmap->N
5225: Developer Notes: This directly accesses information inside the VecScatter associated with the matrix-vector product
5226: for this matrix. This is not desirable..
5228: Level: developer
5230: */
5231: PetscErrorCode MatGetBrowsOfAoCols_MPIAIJ(Mat A,Mat B,MatReuse scall,PetscInt **startsj_s,PetscInt **startsj_r,MatScalar **bufa_ptr,Mat *B_oth)
5232: {
5233: PetscErrorCode ierr;
5234: Mat_MPIAIJ *a=(Mat_MPIAIJ*)A->data;
5235: Mat_SeqAIJ *b_oth;
5236: VecScatter ctx;
5237: MPI_Comm comm;
5238: const PetscMPIInt *rprocs,*sprocs;
5239: const PetscInt *srow,*rstarts,*sstarts;
5240: PetscInt *rowlen,*bufj,*bufJ,ncols,aBn=a->B->cmap->n,row,*b_othi,*b_othj,*rvalues=NULL,*svalues=NULL,*cols,sbs,rbs;
5241: PetscInt i,j,k=0,l,ll,nrecvs,nsends,nrows,*rstartsj = 0,*sstartsj,len;
5242: PetscScalar *b_otha,*bufa,*bufA,*vals;
5243: MPI_Request *rwaits = NULL,*swaits = NULL;
5244: MPI_Status rstatus;
5245: PetscMPIInt jj,size,tag,rank,nsends_mpi,nrecvs_mpi;
5248: PetscObjectGetComm((PetscObject)A,&comm);
5249: MPI_Comm_size(comm,&size);
5251: if (A->cmap->rstart != B->rmap->rstart || A->cmap->rend != B->rmap->rend) {
5252: SETERRQ4(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Matrix local dimensions are incompatible, (%d, %d) != (%d,%d)",A->cmap->rstart,A->cmap->rend,B->rmap->rstart,B->rmap->rend);
5253: }
5254: PetscLogEventBegin(MAT_GetBrowsOfAocols,A,B,0,0);
5255: MPI_Comm_rank(comm,&rank);
5257: if (size == 1) {
5258: startsj_s = NULL;
5259: bufa_ptr = NULL;
5260: *B_oth = NULL;
5261: return(0);
5262: }
5264: ctx = a->Mvctx;
5265: tag = ((PetscObject)ctx)->tag;
5267: if (ctx->inuse) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE," Scatter ctx already in use");
5268: VecScatterGetRemote_Private(ctx,PETSC_TRUE/*send*/,&nsends,&sstarts,&srow,&sprocs,&sbs);
5269: /* rprocs[] must be ordered so that indices received from them are ordered in rvalues[], which is key to algorithms used in this subroutine */
5270: VecScatterGetRemoteOrdered_Private(ctx,PETSC_FALSE/*recv*/,&nrecvs,&rstarts,NULL/*indices not needed*/,&rprocs,&rbs);
5271: PetscMPIIntCast(nsends,&nsends_mpi);
5272: PetscMPIIntCast(nrecvs,&nrecvs_mpi);
5273: PetscMalloc2(nrecvs,&rwaits,nsends,&swaits);
5275: if (!startsj_s || !bufa_ptr) scall = MAT_INITIAL_MATRIX;
5276: if (scall == MAT_INITIAL_MATRIX) {
5277: /* i-array */
5278: /*---------*/
5279: /* post receives */
5280: if (nrecvs) {PetscMalloc1(rbs*(rstarts[nrecvs] - rstarts[0]),&rvalues);} /* rstarts can be NULL when nrecvs=0 */
5281: for (i=0; i<nrecvs; i++) {
5282: rowlen = rvalues + rstarts[i]*rbs;
5283: nrows = (rstarts[i+1]-rstarts[i])*rbs; /* num of indices to be received */
5284: MPI_Irecv(rowlen,nrows,MPIU_INT,rprocs[i],tag,comm,rwaits+i);
5285: }
5287: /* pack the outgoing message */
5288: PetscMalloc2(nsends+1,&sstartsj,nrecvs+1,&rstartsj);
5290: sstartsj[0] = 0;
5291: rstartsj[0] = 0;
5292: len = 0; /* total length of j or a array to be sent */
5293: if (nsends) {
5294: k = sstarts[0]; /* ATTENTION: sstarts[0] and rstarts[0] are not necessarily zero */
5295: PetscMalloc1(sbs*(sstarts[nsends]-sstarts[0]),&svalues);
5296: }
5297: for (i=0; i<nsends; i++) {
5298: rowlen = svalues + (sstarts[i]-sstarts[0])*sbs;
5299: nrows = sstarts[i+1]-sstarts[i]; /* num of block rows */
5300: for (j=0; j<nrows; j++) {
5301: row = srow[k] + B->rmap->range[rank]; /* global row idx */
5302: for (l=0; l<sbs; l++) {
5303: MatGetRow_MPIAIJ(B,row+l,&ncols,NULL,NULL); /* rowlength */
5305: rowlen[j*sbs+l] = ncols;
5307: len += ncols;
5308: MatRestoreRow_MPIAIJ(B,row+l,&ncols,NULL,NULL);
5309: }
5310: k++;
5311: }
5312: MPI_Isend(rowlen,nrows*sbs,MPIU_INT,sprocs[i],tag,comm,swaits+i);
5314: sstartsj[i+1] = len; /* starting point of (i+1)-th outgoing msg in bufj and bufa */
5315: }
5316: /* recvs and sends of i-array are completed */
5317: i = nrecvs;
5318: while (i--) {
5319: MPI_Waitany(nrecvs_mpi,rwaits,&jj,&rstatus);
5320: }
5321: if (nsends) {MPI_Waitall(nsends_mpi,swaits,MPI_STATUSES_IGNORE);}
5322: PetscFree(svalues);
5324: /* allocate buffers for sending j and a arrays */
5325: PetscMalloc1(len+1,&bufj);
5326: PetscMalloc1(len+1,&bufa);
5328: /* create i-array of B_oth */
5329: PetscMalloc1(aBn+2,&b_othi);
5331: b_othi[0] = 0;
5332: len = 0; /* total length of j or a array to be received */
5333: k = 0;
5334: for (i=0; i<nrecvs; i++) {
5335: rowlen = rvalues + (rstarts[i]-rstarts[0])*rbs;
5336: nrows = (rstarts[i+1]-rstarts[i])*rbs; /* num of rows to be received */
5337: for (j=0; j<nrows; j++) {
5338: b_othi[k+1] = b_othi[k] + rowlen[j];
5339: PetscIntSumError(rowlen[j],len,&len);
5340: k++;
5341: }
5342: rstartsj[i+1] = len; /* starting point of (i+1)-th incoming msg in bufj and bufa */
5343: }
5344: PetscFree(rvalues);
5346: /* allocate space for j and a arrrays of B_oth */
5347: PetscMalloc1(b_othi[aBn]+1,&b_othj);
5348: PetscMalloc1(b_othi[aBn]+1,&b_otha);
5350: /* j-array */
5351: /*---------*/
5352: /* post receives of j-array */
5353: for (i=0; i<nrecvs; i++) {
5354: nrows = rstartsj[i+1]-rstartsj[i]; /* length of the msg received */
5355: MPI_Irecv(b_othj+rstartsj[i],nrows,MPIU_INT,rprocs[i],tag,comm,rwaits+i);
5356: }
5358: /* pack the outgoing message j-array */
5359: if (nsends) k = sstarts[0];
5360: for (i=0; i<nsends; i++) {
5361: nrows = sstarts[i+1]-sstarts[i]; /* num of block rows */
5362: bufJ = bufj+sstartsj[i];
5363: for (j=0; j<nrows; j++) {
5364: row = srow[k++] + B->rmap->range[rank]; /* global row idx */
5365: for (ll=0; ll<sbs; ll++) {
5366: MatGetRow_MPIAIJ(B,row+ll,&ncols,&cols,NULL);
5367: for (l=0; l<ncols; l++) {
5368: *bufJ++ = cols[l];
5369: }
5370: MatRestoreRow_MPIAIJ(B,row+ll,&ncols,&cols,NULL);
5371: }
5372: }
5373: MPI_Isend(bufj+sstartsj[i],sstartsj[i+1]-sstartsj[i],MPIU_INT,sprocs[i],tag,comm,swaits+i);
5374: }
5376: /* recvs and sends of j-array are completed */
5377: i = nrecvs;
5378: while (i--) {
5379: MPI_Waitany(nrecvs_mpi,rwaits,&jj,&rstatus);
5380: }
5381: if (nsends) {MPI_Waitall(nsends_mpi,swaits,MPI_STATUSES_IGNORE);}
5382: } else if (scall == MAT_REUSE_MATRIX) {
5383: sstartsj = *startsj_s;
5384: rstartsj = *startsj_r;
5385: bufa = *bufa_ptr;
5386: b_oth = (Mat_SeqAIJ*)(*B_oth)->data;
5387: b_otha = b_oth->a;
5388: } else SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE, "Matrix P does not posses an object container");
5390: /* a-array */
5391: /*---------*/
5392: /* post receives of a-array */
5393: for (i=0; i<nrecvs; i++) {
5394: nrows = rstartsj[i+1]-rstartsj[i]; /* length of the msg received */
5395: MPI_Irecv(b_otha+rstartsj[i],nrows,MPIU_SCALAR,rprocs[i],tag,comm,rwaits+i);
5396: }
5398: /* pack the outgoing message a-array */
5399: if (nsends) k = sstarts[0];
5400: for (i=0; i<nsends; i++) {
5401: nrows = sstarts[i+1]-sstarts[i]; /* num of block rows */
5402: bufA = bufa+sstartsj[i];
5403: for (j=0; j<nrows; j++) {
5404: row = srow[k++] + B->rmap->range[rank]; /* global row idx */
5405: for (ll=0; ll<sbs; ll++) {
5406: MatGetRow_MPIAIJ(B,row+ll,&ncols,NULL,&vals);
5407: for (l=0; l<ncols; l++) {
5408: *bufA++ = vals[l];
5409: }
5410: MatRestoreRow_MPIAIJ(B,row+ll,&ncols,NULL,&vals);
5411: }
5412: }
5413: MPI_Isend(bufa+sstartsj[i],sstartsj[i+1]-sstartsj[i],MPIU_SCALAR,sprocs[i],tag,comm,swaits+i);
5414: }
5415: /* recvs and sends of a-array are completed */
5416: i = nrecvs;
5417: while (i--) {
5418: MPI_Waitany(nrecvs_mpi,rwaits,&jj,&rstatus);
5419: }
5420: if (nsends) {MPI_Waitall(nsends_mpi,swaits,MPI_STATUSES_IGNORE);}
5421: PetscFree2(rwaits,swaits);
5423: if (scall == MAT_INITIAL_MATRIX) {
5424: /* put together the new matrix */
5425: MatCreateSeqAIJWithArrays(PETSC_COMM_SELF,aBn,B->cmap->N,b_othi,b_othj,b_otha,B_oth);
5427: /* MatCreateSeqAIJWithArrays flags matrix so PETSc doesn't free the user's arrays. */
5428: /* Since these are PETSc arrays, change flags to free them as necessary. */
5429: b_oth = (Mat_SeqAIJ*)(*B_oth)->data;
5430: b_oth->free_a = PETSC_TRUE;
5431: b_oth->free_ij = PETSC_TRUE;
5432: b_oth->nonew = 0;
5434: PetscFree(bufj);
5435: if (!startsj_s || !bufa_ptr) {
5436: PetscFree2(sstartsj,rstartsj);
5437: PetscFree(bufa_ptr);
5438: } else {
5439: *startsj_s = sstartsj;
5440: *startsj_r = rstartsj;
5441: *bufa_ptr = bufa;
5442: }
5443: }
5445: VecScatterRestoreRemote_Private(ctx,PETSC_TRUE,&nsends,&sstarts,&srow,&sprocs,&sbs);
5446: VecScatterRestoreRemoteOrdered_Private(ctx,PETSC_FALSE,&nrecvs,&rstarts,NULL,&rprocs,&rbs);
5447: PetscLogEventEnd(MAT_GetBrowsOfAocols,A,B,0,0);
5448: return(0);
5449: }
5451: /*@C
5452: MatGetCommunicationStructs - Provides access to the communication structures used in matrix-vector multiplication.
5454: Not Collective
5456: Input Parameters:
5457: . A - The matrix in mpiaij format
5459: Output Parameter:
5460: + lvec - The local vector holding off-process values from the argument to a matrix-vector product
5461: . colmap - A map from global column index to local index into lvec
5462: - multScatter - A scatter from the argument of a matrix-vector product to lvec
5464: Level: developer
5466: @*/
5467: #if defined(PETSC_USE_CTABLE)
5468: PetscErrorCode MatGetCommunicationStructs(Mat A, Vec *lvec, PetscTable *colmap, VecScatter *multScatter)
5469: #else
5470: PetscErrorCode MatGetCommunicationStructs(Mat A, Vec *lvec, PetscInt *colmap[], VecScatter *multScatter)
5471: #endif
5472: {
5473: Mat_MPIAIJ *a;
5480: a = (Mat_MPIAIJ*) A->data;
5481: if (lvec) *lvec = a->lvec;
5482: if (colmap) *colmap = a->colmap;
5483: if (multScatter) *multScatter = a->Mvctx;
5484: return(0);
5485: }
5487: PETSC_INTERN PetscErrorCode MatConvert_MPIAIJ_MPIAIJCRL(Mat,MatType,MatReuse,Mat*);
5488: PETSC_INTERN PetscErrorCode MatConvert_MPIAIJ_MPIAIJPERM(Mat,MatType,MatReuse,Mat*);
5489: PETSC_INTERN PetscErrorCode MatConvert_MPIAIJ_MPIAIJSELL(Mat,MatType,MatReuse,Mat*);
5490: #if defined(PETSC_HAVE_MKL_SPARSE)
5491: PETSC_INTERN PetscErrorCode MatConvert_MPIAIJ_MPIAIJMKL(Mat,MatType,MatReuse,Mat*);
5492: #endif
5493: PETSC_INTERN PetscErrorCode MatConvert_MPIAIJ_MPISBAIJ(Mat,MatType,MatReuse,Mat*);
5494: #if defined(PETSC_HAVE_ELEMENTAL)
5495: PETSC_INTERN PetscErrorCode MatConvert_MPIAIJ_Elemental(Mat,MatType,MatReuse,Mat*);
5496: #endif
5497: #if defined(PETSC_HAVE_HYPRE)
5498: PETSC_INTERN PetscErrorCode MatConvert_AIJ_HYPRE(Mat,MatType,MatReuse,Mat*);
5499: PETSC_INTERN PetscErrorCode MatMatMatMult_Transpose_AIJ_AIJ(Mat,Mat,Mat,MatReuse,PetscReal,Mat*);
5500: #endif
5501: PETSC_INTERN PetscErrorCode MatConvert_XAIJ_IS(Mat,MatType,MatReuse,Mat*);
5502: PETSC_INTERN PetscErrorCode MatConvert_MPIAIJ_MPISELL(Mat,MatType,MatReuse,Mat*);
5503: PETSC_INTERN PetscErrorCode MatPtAP_IS_XAIJ(Mat,Mat,MatReuse,PetscReal,Mat*);
5505: /*
5506: Computes (B'*A')' since computing B*A directly is untenable
5508: n p p
5509: ( ) ( ) ( )
5510: m ( A ) * n ( B ) = m ( C )
5511: ( ) ( ) ( )
5513: */
5514: PetscErrorCode MatMatMultNumeric_MPIDense_MPIAIJ(Mat A,Mat B,Mat C)
5515: {
5517: Mat At,Bt,Ct;
5520: MatTranspose(A,MAT_INITIAL_MATRIX,&At);
5521: MatTranspose(B,MAT_INITIAL_MATRIX,&Bt);
5522: MatMatMult(Bt,At,MAT_INITIAL_MATRIX,1.0,&Ct);
5523: MatDestroy(&At);
5524: MatDestroy(&Bt);
5525: MatTranspose(Ct,MAT_REUSE_MATRIX,&C);
5526: MatDestroy(&Ct);
5527: return(0);
5528: }
5530: PetscErrorCode MatMatMultSymbolic_MPIDense_MPIAIJ(Mat A,Mat B,PetscReal fill,Mat *C)
5531: {
5533: PetscInt m=A->rmap->n,n=B->cmap->n;
5534: Mat Cmat;
5537: if (A->cmap->n != B->rmap->n) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"A->cmap->n %d != B->rmap->n %d\n",A->cmap->n,B->rmap->n);
5538: MatCreate(PetscObjectComm((PetscObject)A),&Cmat);
5539: MatSetSizes(Cmat,m,n,PETSC_DETERMINE,PETSC_DETERMINE);
5540: MatSetBlockSizesFromMats(Cmat,A,B);
5541: MatSetType(Cmat,MATMPIDENSE);
5542: MatMPIDenseSetPreallocation(Cmat,NULL);
5543: MatAssemblyBegin(Cmat,MAT_FINAL_ASSEMBLY);
5544: MatAssemblyEnd(Cmat,MAT_FINAL_ASSEMBLY);
5546: Cmat->ops->matmultnumeric = MatMatMultNumeric_MPIDense_MPIAIJ;
5548: *C = Cmat;
5549: return(0);
5550: }
5552: /* ----------------------------------------------------------------*/
5553: PETSC_INTERN PetscErrorCode MatMatMult_MPIDense_MPIAIJ(Mat A,Mat B,MatReuse scall,PetscReal fill,Mat *C)
5554: {
5558: if (scall == MAT_INITIAL_MATRIX) {
5559: PetscLogEventBegin(MAT_MatMultSymbolic,A,B,0,0);
5560: MatMatMultSymbolic_MPIDense_MPIAIJ(A,B,fill,C);
5561: PetscLogEventEnd(MAT_MatMultSymbolic,A,B,0,0);
5562: }
5563: PetscLogEventBegin(MAT_MatMultNumeric,A,B,0,0);
5564: MatMatMultNumeric_MPIDense_MPIAIJ(A,B,*C);
5565: PetscLogEventEnd(MAT_MatMultNumeric,A,B,0,0);
5566: return(0);
5567: }
5569: /*MC
5570: MATMPIAIJ - MATMPIAIJ = "mpiaij" - A matrix type to be used for parallel sparse matrices.
5572: Options Database Keys:
5573: . -mat_type mpiaij - sets the matrix type to "mpiaij" during a call to MatSetFromOptions()
5575: Level: beginner
5577: .seealso: MatCreateAIJ()
5578: M*/
5580: PETSC_EXTERN PetscErrorCode MatCreate_MPIAIJ(Mat B)
5581: {
5582: Mat_MPIAIJ *b;
5584: PetscMPIInt size;
5587: MPI_Comm_size(PetscObjectComm((PetscObject)B),&size);
5589: PetscNewLog(B,&b);
5590: B->data = (void*)b;
5591: PetscMemcpy(B->ops,&MatOps_Values,sizeof(struct _MatOps));
5592: B->assembled = PETSC_FALSE;
5593: B->insertmode = NOT_SET_VALUES;
5594: b->size = size;
5596: MPI_Comm_rank(PetscObjectComm((PetscObject)B),&b->rank);
5598: /* build cache for off array entries formed */
5599: MatStashCreate_Private(PetscObjectComm((PetscObject)B),1,&B->stash);
5601: b->donotstash = PETSC_FALSE;
5602: b->colmap = 0;
5603: b->garray = 0;
5604: b->roworiented = PETSC_TRUE;
5606: /* stuff used for matrix vector multiply */
5607: b->lvec = NULL;
5608: b->Mvctx = NULL;
5610: /* stuff for MatGetRow() */
5611: b->rowindices = 0;
5612: b->rowvalues = 0;
5613: b->getrowactive = PETSC_FALSE;
5615: /* flexible pointer used in CUSP/CUSPARSE classes */
5616: b->spptr = NULL;
5618: PetscObjectComposeFunction((PetscObject)B,"MatMPIAIJSetUseScalableIncreaseOverlap_C",MatMPIAIJSetUseScalableIncreaseOverlap_MPIAIJ);
5619: PetscObjectComposeFunction((PetscObject)B,"MatStoreValues_C",MatStoreValues_MPIAIJ);
5620: PetscObjectComposeFunction((PetscObject)B,"MatRetrieveValues_C",MatRetrieveValues_MPIAIJ);
5621: PetscObjectComposeFunction((PetscObject)B,"MatIsTranspose_C",MatIsTranspose_MPIAIJ);
5622: PetscObjectComposeFunction((PetscObject)B,"MatMPIAIJSetPreallocation_C",MatMPIAIJSetPreallocation_MPIAIJ);
5623: PetscObjectComposeFunction((PetscObject)B,"MatResetPreallocation_C",MatResetPreallocation_MPIAIJ);
5624: PetscObjectComposeFunction((PetscObject)B,"MatMPIAIJSetPreallocationCSR_C",MatMPIAIJSetPreallocationCSR_MPIAIJ);
5625: PetscObjectComposeFunction((PetscObject)B,"MatDiagonalScaleLocal_C",MatDiagonalScaleLocal_MPIAIJ);
5626: PetscObjectComposeFunction((PetscObject)B,"MatConvert_mpiaij_mpiaijperm_C",MatConvert_MPIAIJ_MPIAIJPERM);
5627: PetscObjectComposeFunction((PetscObject)B,"MatConvert_mpiaij_mpiaijsell_C",MatConvert_MPIAIJ_MPIAIJSELL);
5628: #if defined(PETSC_HAVE_MKL_SPARSE)
5629: PetscObjectComposeFunction((PetscObject)B,"MatConvert_mpiaij_mpiaijmkl_C",MatConvert_MPIAIJ_MPIAIJMKL);
5630: #endif
5631: PetscObjectComposeFunction((PetscObject)B,"MatConvert_mpiaij_mpiaijcrl_C",MatConvert_MPIAIJ_MPIAIJCRL);
5632: PetscObjectComposeFunction((PetscObject)B,"MatConvert_mpiaij_mpisbaij_C",MatConvert_MPIAIJ_MPISBAIJ);
5633: #if defined(PETSC_HAVE_ELEMENTAL)
5634: PetscObjectComposeFunction((PetscObject)B,"MatConvert_mpiaij_elemental_C",MatConvert_MPIAIJ_Elemental);
5635: #endif
5636: #if defined(PETSC_HAVE_HYPRE)
5637: PetscObjectComposeFunction((PetscObject)B,"MatConvert_mpiaij_hypre_C",MatConvert_AIJ_HYPRE);
5638: #endif
5639: PetscObjectComposeFunction((PetscObject)B,"MatConvert_mpiaij_is_C",MatConvert_XAIJ_IS);
5640: PetscObjectComposeFunction((PetscObject)B,"MatConvert_mpiaij_mpisell_C",MatConvert_MPIAIJ_MPISELL);
5641: PetscObjectComposeFunction((PetscObject)B,"MatMatMult_mpidense_mpiaij_C",MatMatMult_MPIDense_MPIAIJ);
5642: PetscObjectComposeFunction((PetscObject)B,"MatMatMultSymbolic_mpidense_mpiaij_C",MatMatMultSymbolic_MPIDense_MPIAIJ);
5643: PetscObjectComposeFunction((PetscObject)B,"MatMatMultNumeric_mpidense_mpiaij_C",MatMatMultNumeric_MPIDense_MPIAIJ);
5644: #if defined(PETSC_HAVE_HYPRE)
5645: PetscObjectComposeFunction((PetscObject)B,"MatMatMatMult_transpose_mpiaij_mpiaij_C",MatMatMatMult_Transpose_AIJ_AIJ);
5646: #endif
5647: PetscObjectComposeFunction((PetscObject)B,"MatPtAP_is_mpiaij_C",MatPtAP_IS_XAIJ);
5648: PetscObjectChangeTypeName((PetscObject)B,MATMPIAIJ);
5649: return(0);
5650: }
5652: /*@C
5653: MatCreateMPIAIJWithSplitArrays - creates a MPI AIJ matrix using arrays that contain the "diagonal"
5654: and "off-diagonal" part of the matrix in CSR format.
5656: Collective on MPI_Comm
5658: Input Parameters:
5659: + comm - MPI communicator
5660: . m - number of local rows (Cannot be PETSC_DECIDE)
5661: . n - This value should be the same as the local size used in creating the
5662: x vector for the matrix-vector product y = Ax. (or PETSC_DECIDE to have
5663: calculated if N is given) For square matrices n is almost always m.
5664: . M - number of global rows (or PETSC_DETERMINE to have calculated if m is given)
5665: . N - number of global columns (or PETSC_DETERMINE to have calculated if n is given)
5666: . i - row indices for "diagonal" portion of matrix; that is i[0] = 0, i[row] = i[row-1] + number of elements in that row of the matrix
5667: . j - column indices
5668: . a - matrix values
5669: . oi - row indices for "off-diagonal" portion of matrix; that is oi[0] = 0, oi[row] = oi[row-1] + number of elements in that row of the matrix
5670: . oj - column indices
5671: - oa - matrix values
5673: Output Parameter:
5674: . mat - the matrix
5676: Level: advanced
5678: Notes:
5679: The i, j, and a arrays ARE NOT copied by this routine into the internal format used by PETSc. The user
5680: must free the arrays once the matrix has been destroyed and not before.
5682: The i and j indices are 0 based
5684: See MatCreateAIJ() for the definition of "diagonal" and "off-diagonal" portion of the matrix
5686: This sets local rows and cannot be used to set off-processor values.
5688: Use of this routine is discouraged because it is inflexible and cumbersome to use. It is extremely rare that a
5689: legacy application natively assembles into exactly this split format. The code to do so is nontrivial and does
5690: not easily support in-place reassembly. It is recommended to use MatSetValues() (or a variant thereof) because
5691: the resulting assembly is easier to implement, will work with any matrix format, and the user does not have to
5692: keep track of the underlying array. Use MatSetOption(A,MAT_NO_OFF_PROC_ENTRIES,PETSC_TRUE) to disable all
5693: communication if it is known that only local entries will be set.
5695: .keywords: matrix, aij, compressed row, sparse, parallel
5697: .seealso: MatCreate(), MatCreateSeqAIJ(), MatSetValues(), MatMPIAIJSetPreallocation(), MatMPIAIJSetPreallocationCSR(),
5698: MATMPIAIJ, MatCreateAIJ(), MatCreateMPIAIJWithArrays()
5699: @*/
5700: PetscErrorCode MatCreateMPIAIJWithSplitArrays(MPI_Comm comm,PetscInt m,PetscInt n,PetscInt M,PetscInt N,PetscInt i[],PetscInt j[],PetscScalar a[],PetscInt oi[], PetscInt oj[],PetscScalar oa[],Mat *mat)
5701: {
5703: Mat_MPIAIJ *maij;
5706: if (m < 0) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"local number of rows (m) cannot be PETSC_DECIDE, or negative");
5707: if (i[0]) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"i (row indices) must start with 0");
5708: if (oi[0]) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"oi (row indices) must start with 0");
5709: MatCreate(comm,mat);
5710: MatSetSizes(*mat,m,n,M,N);
5711: MatSetType(*mat,MATMPIAIJ);
5712: maij = (Mat_MPIAIJ*) (*mat)->data;
5714: (*mat)->preallocated = PETSC_TRUE;
5716: PetscLayoutSetUp((*mat)->rmap);
5717: PetscLayoutSetUp((*mat)->cmap);
5719: MatCreateSeqAIJWithArrays(PETSC_COMM_SELF,m,n,i,j,a,&maij->A);
5720: MatCreateSeqAIJWithArrays(PETSC_COMM_SELF,m,(*mat)->cmap->N,oi,oj,oa,&maij->B);
5722: MatAssemblyBegin(maij->A,MAT_FINAL_ASSEMBLY);
5723: MatAssemblyEnd(maij->A,MAT_FINAL_ASSEMBLY);
5724: MatAssemblyBegin(maij->B,MAT_FINAL_ASSEMBLY);
5725: MatAssemblyEnd(maij->B,MAT_FINAL_ASSEMBLY);
5727: MatSetOption(*mat,MAT_NO_OFF_PROC_ENTRIES,PETSC_TRUE);
5728: MatAssemblyBegin(*mat,MAT_FINAL_ASSEMBLY);
5729: MatAssemblyEnd(*mat,MAT_FINAL_ASSEMBLY);
5730: MatSetOption(*mat,MAT_NO_OFF_PROC_ENTRIES,PETSC_FALSE);
5731: MatSetOption(*mat,MAT_NEW_NONZERO_LOCATION_ERR,PETSC_TRUE);
5732: return(0);
5733: }
5735: /*
5736: Special version for direct calls from Fortran
5737: */
5738: #include <petsc/private/fortranimpl.h>
5740: /* Change these macros so can be used in void function */
5741: #undef CHKERRQ
5742: #define CHKERRQ(ierr) CHKERRABORT(PETSC_COMM_WORLD,ierr)
5743: #undef SETERRQ2
5744: #define SETERRQ2(comm,ierr,b,c,d) CHKERRABORT(comm,ierr)
5745: #undef SETERRQ3
5746: #define SETERRQ3(comm,ierr,b,c,d,e) CHKERRABORT(comm,ierr)
5747: #undef SETERRQ
5748: #define SETERRQ(c,ierr,b) CHKERRABORT(c,ierr)
5750: #if defined(PETSC_HAVE_FORTRAN_CAPS)
5751: #define matsetvaluesmpiaij_ MATSETVALUESMPIAIJ
5752: #elif !defined(PETSC_HAVE_FORTRAN_UNDERSCORE)
5753: #define matsetvaluesmpiaij_ matsetvaluesmpiaij
5754: #else
5755: #endif
5756: PETSC_EXTERN void PETSC_STDCALL matsetvaluesmpiaij_(Mat *mmat,PetscInt *mm,const PetscInt im[],PetscInt *mn,const PetscInt in[],const PetscScalar v[],InsertMode *maddv,PetscErrorCode *_ierr)
5757: {
5758: Mat mat = *mmat;
5759: PetscInt m = *mm, n = *mn;
5760: InsertMode addv = *maddv;
5761: Mat_MPIAIJ *aij = (Mat_MPIAIJ*)mat->data;
5762: PetscScalar value;
5765: MatCheckPreallocated(mat,1);
5766: if (mat->insertmode == NOT_SET_VALUES) mat->insertmode = addv;
5768: #if defined(PETSC_USE_DEBUG)
5769: else if (mat->insertmode != addv) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Cannot mix add values and insert values");
5770: #endif
5771: {
5772: PetscInt i,j,rstart = mat->rmap->rstart,rend = mat->rmap->rend;
5773: PetscInt cstart = mat->cmap->rstart,cend = mat->cmap->rend,row,col;
5774: PetscBool roworiented = aij->roworiented;
5776: /* Some Variables required in the macro */
5777: Mat A = aij->A;
5778: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
5779: PetscInt *aimax = a->imax,*ai = a->i,*ailen = a->ilen,*aj = a->j;
5780: MatScalar *aa = a->a;
5781: PetscBool ignorezeroentries = (((a->ignorezeroentries)&&(addv==ADD_VALUES)) ? PETSC_TRUE : PETSC_FALSE);
5782: Mat B = aij->B;
5783: Mat_SeqAIJ *b = (Mat_SeqAIJ*)B->data;
5784: PetscInt *bimax = b->imax,*bi = b->i,*bilen = b->ilen,*bj = b->j,bm = aij->B->rmap->n,am = aij->A->rmap->n;
5785: MatScalar *ba = b->a;
5787: PetscInt *rp1,*rp2,ii,nrow1,nrow2,_i,rmax1,rmax2,N,low1,high1,low2,high2,t,lastcol1,lastcol2;
5788: PetscInt nonew = a->nonew;
5789: MatScalar *ap1,*ap2;
5792: for (i=0; i<m; i++) {
5793: if (im[i] < 0) continue;
5794: #if defined(PETSC_USE_DEBUG)
5795: if (im[i] >= mat->rmap->N) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Row too large: row %D max %D",im[i],mat->rmap->N-1);
5796: #endif
5797: if (im[i] >= rstart && im[i] < rend) {
5798: row = im[i] - rstart;
5799: lastcol1 = -1;
5800: rp1 = aj + ai[row];
5801: ap1 = aa + ai[row];
5802: rmax1 = aimax[row];
5803: nrow1 = ailen[row];
5804: low1 = 0;
5805: high1 = nrow1;
5806: lastcol2 = -1;
5807: rp2 = bj + bi[row];
5808: ap2 = ba + bi[row];
5809: rmax2 = bimax[row];
5810: nrow2 = bilen[row];
5811: low2 = 0;
5812: high2 = nrow2;
5814: for (j=0; j<n; j++) {
5815: if (roworiented) value = v[i*n+j];
5816: else value = v[i+j*m];
5817: if (in[j] >= cstart && in[j] < cend) {
5818: col = in[j] - cstart;
5819: if (ignorezeroentries && value == 0.0 && (addv == ADD_VALUES) && row != col) continue;
5820: MatSetValues_SeqAIJ_A_Private(row,col,value,addv,im[i],in[j]);
5821: } else if (in[j] < 0) continue;
5822: #if defined(PETSC_USE_DEBUG)
5823: /* extra brace on SETERRQ2() is required for --with-errorchecking=0 - due to the next 'else' clause */
5824: else if (in[j] >= mat->cmap->N) {SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Column too large: col %D max %D",in[j],mat->cmap->N-1);}
5825: #endif
5826: else {
5827: if (mat->was_assembled) {
5828: if (!aij->colmap) {
5829: MatCreateColmap_MPIAIJ_Private(mat);
5830: }
5831: #if defined(PETSC_USE_CTABLE)
5832: PetscTableFind(aij->colmap,in[j]+1,&col);
5833: col--;
5834: #else
5835: col = aij->colmap[in[j]] - 1;
5836: #endif
5837: if (ignorezeroentries && value == 0.0 && (addv == ADD_VALUES) && row != col) continue;
5838: if (col < 0 && !((Mat_SeqAIJ*)(aij->A->data))->nonew) {
5839: MatDisAssemble_MPIAIJ(mat);
5840: col = in[j];
5841: /* Reinitialize the variables required by MatSetValues_SeqAIJ_B_Private() */
5842: B = aij->B;
5843: b = (Mat_SeqAIJ*)B->data;
5844: bimax = b->imax; bi = b->i; bilen = b->ilen; bj = b->j;
5845: rp2 = bj + bi[row];
5846: ap2 = ba + bi[row];
5847: rmax2 = bimax[row];
5848: nrow2 = bilen[row];
5849: low2 = 0;
5850: high2 = nrow2;
5851: bm = aij->B->rmap->n;
5852: ba = b->a;
5853: }
5854: } else col = in[j];
5855: MatSetValues_SeqAIJ_B_Private(row,col,value,addv,im[i],in[j]);
5856: }
5857: }
5858: } else if (!aij->donotstash) {
5859: if (roworiented) {
5860: MatStashValuesRow_Private(&mat->stash,im[i],n,in,v+i*n,(PetscBool)(ignorezeroentries && (addv == ADD_VALUES)));
5861: } else {
5862: MatStashValuesCol_Private(&mat->stash,im[i],n,in,v+i,m,(PetscBool)(ignorezeroentries && (addv == ADD_VALUES)));
5863: }
5864: }
5865: }
5866: }
5867: PetscFunctionReturnVoid();
5868: }