Actual source code: mpibaij.c
1: /*$Id: mpibaij.c,v 1.234 2001/09/25 22:56:49 balay Exp $*/
3: #include src/mat/impls/baij/mpi/mpibaij.h
5: EXTERN int MatSetUpMultiply_MPIBAIJ(Mat);
6: EXTERN int DisAssemble_MPIBAIJ(Mat);
7: EXTERN int MatIncreaseOverlap_MPIBAIJ(Mat,int,IS[],int);
8: EXTERN int MatGetSubMatrices_MPIBAIJ(Mat,int,const IS[],const IS[],MatReuse,Mat *[]);
9: EXTERN int MatGetValues_SeqBAIJ(Mat,int,const int[],int,const int [],PetscScalar []);
10: EXTERN int MatSetValues_SeqBAIJ(Mat,int,const int[],int,const int [],const PetscScalar [],InsertMode);
11: EXTERN int MatSetValuesBlocked_SeqBAIJ(Mat,int,const int[],int,const int[],const PetscScalar[],InsertMode);
12: EXTERN int MatGetRow_SeqBAIJ(Mat,int,int*,int*[],PetscScalar*[]);
13: EXTERN int MatRestoreRow_SeqBAIJ(Mat,int,int*,int*[],PetscScalar*[]);
14: EXTERN int MatPrintHelp_SeqBAIJ(Mat);
15: EXTERN int MatZeroRows_SeqBAIJ(Mat,IS,const PetscScalar*);
17: /* UGLY, ugly, ugly
18: When MatScalar == PetscScalar the function MatSetValuesBlocked_MPIBAIJ_MatScalar() does
19: not exist. Otherwise ..._MatScalar() takes matrix elements in single precision and
20: inserts them into the single precision data structure. The function MatSetValuesBlocked_MPIBAIJ()
21: converts the entries into single precision and then calls ..._MatScalar() to put them
22: into the single precision data structures.
23: */
24: #if defined(PETSC_USE_MAT_SINGLE)
25: EXTERN int MatSetValuesBlocked_SeqBAIJ_MatScalar(Mat,int,const int*,int,const int*,const MatScalar*,InsertMode);
26: EXTERN int MatSetValues_MPIBAIJ_MatScalar(Mat,int,const int*,int,const int*,const MatScalar*,InsertMode);
27: EXTERN int MatSetValuesBlocked_MPIBAIJ_MatScalar(Mat,int,const int*,int,const int*,const MatScalar*,InsertMode);
28: EXTERN int MatSetValues_MPIBAIJ_HT_MatScalar(Mat,int,const int*,int,const int*,const MatScalar*,InsertMode);
29: EXTERN int MatSetValuesBlocked_MPIBAIJ_HT_MatScalar(Mat,int,const int*,int,const int*,const MatScalar*,InsertMode);
30: #else
31: #define MatSetValuesBlocked_SeqBAIJ_MatScalar MatSetValuesBlocked_SeqBAIJ
32: #define MatSetValues_MPIBAIJ_MatScalar MatSetValues_MPIBAIJ
33: #define MatSetValuesBlocked_MPIBAIJ_MatScalar MatSetValuesBlocked_MPIBAIJ
34: #define MatSetValues_MPIBAIJ_HT_MatScalar MatSetValues_MPIBAIJ_HT
35: #define MatSetValuesBlocked_MPIBAIJ_HT_MatScalar MatSetValuesBlocked_MPIBAIJ_HT
36: #endif
40: int MatGetRowMax_MPIBAIJ(Mat A,Vec v)
41: {
42: Mat_MPIBAIJ *a = (Mat_MPIBAIJ*)A->data;
43: int ierr,i;
44: PetscScalar *va,*vb;
45: Vec vtmp;
48:
49: MatGetRowMax(a->A,v);
50: VecGetArray(v,&va);
52: VecCreateSeq(PETSC_COMM_SELF,A->m,&vtmp);
53: MatGetRowMax(a->B,vtmp);
54: VecGetArray(vtmp,&vb);
56: for (i=0; i<A->m; i++){
57: if (PetscAbsScalar(va[i]) < PetscAbsScalar(vb[i])) va[i] = vb[i];
58: }
60: VecRestoreArray(v,&va);
61: VecRestoreArray(vtmp,&vb);
62: VecDestroy(vtmp);
63:
64: return(0);
65: }
67: EXTERN_C_BEGIN
70: int MatStoreValues_MPIBAIJ(Mat mat)
71: {
72: Mat_MPIBAIJ *aij = (Mat_MPIBAIJ *)mat->data;
73: int ierr;
76: MatStoreValues(aij->A);
77: MatStoreValues(aij->B);
78: return(0);
79: }
80: EXTERN_C_END
82: EXTERN_C_BEGIN
85: int MatRetrieveValues_MPIBAIJ(Mat mat)
86: {
87: Mat_MPIBAIJ *aij = (Mat_MPIBAIJ *)mat->data;
88: int ierr;
91: MatRetrieveValues(aij->A);
92: MatRetrieveValues(aij->B);
93: return(0);
94: }
95: EXTERN_C_END
97: /*
98: Local utility routine that creates a mapping from the global column
99: number to the local number in the off-diagonal part of the local
100: storage of the matrix. This is done in a non scable way since the
101: length of colmap equals the global matrix length.
102: */
105: int CreateColmap_MPIBAIJ_Private(Mat mat)
106: {
107: Mat_MPIBAIJ *baij = (Mat_MPIBAIJ*)mat->data;
108: Mat_SeqBAIJ *B = (Mat_SeqBAIJ*)baij->B->data;
109: int nbs = B->nbs,i,bs=B->bs,ierr;
112: #if defined (PETSC_USE_CTABLE)
113: PetscTableCreate(baij->nbs,&baij->colmap);
114: for (i=0; i<nbs; i++){
115: PetscTableAdd(baij->colmap,baij->garray[i]+1,i*bs+1);
116: }
117: #else
118: PetscMalloc((baij->Nbs+1)*sizeof(int),&baij->colmap);
119: PetscLogObjectMemory(mat,baij->Nbs*sizeof(int));
120: PetscMemzero(baij->colmap,baij->Nbs*sizeof(int));
121: for (i=0; i<nbs; i++) baij->colmap[baij->garray[i]] = i*bs+1;
122: #endif
123: return(0);
124: }
126: #define CHUNKSIZE 10
128: #define MatSetValues_SeqBAIJ_A_Private(row,col,value,addv) \
129: { \
130: \
131: brow = row/bs; \
132: rp = aj + ai[brow]; ap = aa + bs2*ai[brow]; \
133: rmax = aimax[brow]; nrow = ailen[brow]; \
134: bcol = col/bs; \
135: ridx = row % bs; cidx = col % bs; \
136: low = 0; high = nrow; \
137: while (high-low > 3) { \
138: t = (low+high)/2; \
139: if (rp[t] > bcol) high = t; \
140: else low = t; \
141: } \
142: for (_i=low; _i<high; _i++) { \
143: if (rp[_i] > bcol) break; \
144: if (rp[_i] == bcol) { \
145: bap = ap + bs2*_i + bs*cidx + ridx; \
146: if (addv == ADD_VALUES) *bap += value; \
147: else *bap = value; \
148: goto a_noinsert; \
149: } \
150: } \
151: if (a->nonew == 1) goto a_noinsert; \
152: else if (a->nonew == -1) SETERRQ2(PETSC_ERR_ARG_OUTOFRANGE,"Inserting a new nonzero (%d, %d) into matrix", row, col); \
153: if (nrow >= rmax) { \
154: /* there is no extra room in row, therefore enlarge */ \
155: int new_nz = ai[a->mbs] + CHUNKSIZE,len,*new_i,*new_j; \
156: MatScalar *new_a; \
157: \
158: if (a->nonew == -2) SETERRQ2(PETSC_ERR_ARG_OUTOFRANGE,"Inserting a new nonzero (%d, %d) in the matrix", row, col); \
159: \
160: /* malloc new storage space */ \
161: len = new_nz*(sizeof(int)+bs2*sizeof(MatScalar))+(a->mbs+1)*sizeof(int); \
162: PetscMalloc(len,&new_a); \
163: new_j = (int*)(new_a + bs2*new_nz); \
164: new_i = new_j + new_nz; \
165: \
166: /* copy over old data into new slots */ \
167: for (ii=0; ii<brow+1; ii++) {new_i[ii] = ai[ii];} \
168: for (ii=brow+1; ii<a->mbs+1; ii++) {new_i[ii] = ai[ii]+CHUNKSIZE;} \
169: PetscMemcpy(new_j,aj,(ai[brow]+nrow)*sizeof(int)); \
170: len = (new_nz - CHUNKSIZE - ai[brow] - nrow); \
171: PetscMemcpy(new_j+ai[brow]+nrow+CHUNKSIZE,aj+ai[brow]+nrow,len*sizeof(int)); \
172: PetscMemcpy(new_a,aa,(ai[brow]+nrow)*bs2*sizeof(MatScalar)); \
173: PetscMemzero(new_a+bs2*(ai[brow]+nrow),bs2*CHUNKSIZE*sizeof(PetscScalar)); \
174: PetscMemcpy(new_a+bs2*(ai[brow]+nrow+CHUNKSIZE), \
175: aa+bs2*(ai[brow]+nrow),bs2*len*sizeof(MatScalar)); \
176: /* free up old matrix storage */ \
177: PetscFree(a->a); \
178: if (!a->singlemalloc) { \
179: PetscFree(a->i); \
180: PetscFree(a->j);\
181: } \
182: aa = a->a = new_a; ai = a->i = new_i; aj = a->j = new_j; \
183: a->singlemalloc = PETSC_TRUE; \
184: \
185: rp = aj + ai[brow]; ap = aa + bs2*ai[brow]; \
186: rmax = aimax[brow] = aimax[brow] + CHUNKSIZE; \
187: PetscLogObjectMemory(A,CHUNKSIZE*(sizeof(int) + bs2*sizeof(MatScalar))); \
188: a->maxnz += bs2*CHUNKSIZE; \
189: a->reallocs++; \
190: a->nz++; \
191: } \
192: N = nrow++ - 1; \
193: /* shift up all the later entries in this row */ \
194: for (ii=N; ii>=_i; ii--) { \
195: rp[ii+1] = rp[ii]; \
196: PetscMemcpy(ap+bs2*(ii+1),ap+bs2*(ii),bs2*sizeof(MatScalar)); \
197: } \
198: if (N>=_i) { PetscMemzero(ap+bs2*_i,bs2*sizeof(MatScalar)); } \
199: rp[_i] = bcol; \
200: ap[bs2*_i + bs*cidx + ridx] = value; \
201: a_noinsert:; \
202: ailen[brow] = nrow; \
203: }
205: #define MatSetValues_SeqBAIJ_B_Private(row,col,value,addv) \
206: { \
207: brow = row/bs; \
208: rp = bj + bi[brow]; ap = ba + bs2*bi[brow]; \
209: rmax = bimax[brow]; nrow = bilen[brow]; \
210: bcol = col/bs; \
211: ridx = row % bs; cidx = col % bs; \
212: low = 0; high = nrow; \
213: while (high-low > 3) { \
214: t = (low+high)/2; \
215: if (rp[t] > bcol) high = t; \
216: else low = t; \
217: } \
218: for (_i=low; _i<high; _i++) { \
219: if (rp[_i] > bcol) break; \
220: if (rp[_i] == bcol) { \
221: bap = ap + bs2*_i + bs*cidx + ridx; \
222: if (addv == ADD_VALUES) *bap += value; \
223: else *bap = value; \
224: goto b_noinsert; \
225: } \
226: } \
227: if (b->nonew == 1) goto b_noinsert; \
228: else if (b->nonew == -1) SETERRQ2(PETSC_ERR_ARG_OUTOFRANGE,"Inserting a new nonzero (%d, %d) into matrix", row, col); \
229: if (nrow >= rmax) { \
230: /* there is no extra room in row, therefore enlarge */ \
231: int new_nz = bi[b->mbs] + CHUNKSIZE,len,*new_i,*new_j; \
232: MatScalar *new_a; \
233: \
234: if (b->nonew == -2) SETERRQ2(PETSC_ERR_ARG_OUTOFRANGE,"Inserting a new nonzero (%d, %d) in the matrix", row, col); \
235: \
236: /* malloc new storage space */ \
237: len = new_nz*(sizeof(int)+bs2*sizeof(MatScalar))+(b->mbs+1)*sizeof(int); \
238: PetscMalloc(len,&new_a); \
239: new_j = (int*)(new_a + bs2*new_nz); \
240: new_i = new_j + new_nz; \
241: \
242: /* copy over old data into new slots */ \
243: for (ii=0; ii<brow+1; ii++) {new_i[ii] = bi[ii];} \
244: for (ii=brow+1; ii<b->mbs+1; ii++) {new_i[ii] = bi[ii]+CHUNKSIZE;} \
245: PetscMemcpy(new_j,bj,(bi[brow]+nrow)*sizeof(int)); \
246: len = (new_nz - CHUNKSIZE - bi[brow] - nrow); \
247: PetscMemcpy(new_j+bi[brow]+nrow+CHUNKSIZE,bj+bi[brow]+nrow,len*sizeof(int)); \
248: PetscMemcpy(new_a,ba,(bi[brow]+nrow)*bs2*sizeof(MatScalar)); \
249: PetscMemzero(new_a+bs2*(bi[brow]+nrow),bs2*CHUNKSIZE*sizeof(MatScalar)); \
250: PetscMemcpy(new_a+bs2*(bi[brow]+nrow+CHUNKSIZE), \
251: ba+bs2*(bi[brow]+nrow),bs2*len*sizeof(MatScalar)); \
252: /* free up old matrix storage */ \
253: PetscFree(b->a); \
254: if (!b->singlemalloc) { \
255: PetscFree(b->i); \
256: PetscFree(b->j); \
257: } \
258: ba = b->a = new_a; bi = b->i = new_i; bj = b->j = new_j; \
259: b->singlemalloc = PETSC_TRUE; \
260: \
261: rp = bj + bi[brow]; ap = ba + bs2*bi[brow]; \
262: rmax = bimax[brow] = bimax[brow] + CHUNKSIZE; \
263: PetscLogObjectMemory(B,CHUNKSIZE*(sizeof(int) + bs2*sizeof(MatScalar))); \
264: b->maxnz += bs2*CHUNKSIZE; \
265: b->reallocs++; \
266: b->nz++; \
267: } \
268: N = nrow++ - 1; \
269: /* shift up all the later entries in this row */ \
270: for (ii=N; ii>=_i; ii--) { \
271: rp[ii+1] = rp[ii]; \
272: PetscMemcpy(ap+bs2*(ii+1),ap+bs2*(ii),bs2*sizeof(MatScalar)); \
273: } \
274: if (N>=_i) { PetscMemzero(ap+bs2*_i,bs2*sizeof(MatScalar));} \
275: rp[_i] = bcol; \
276: ap[bs2*_i + bs*cidx + ridx] = value; \
277: b_noinsert:; \
278: bilen[brow] = nrow; \
279: }
281: #if defined(PETSC_USE_MAT_SINGLE)
284: int MatSetValues_MPIBAIJ(Mat mat,int m,const int im[],int n,const int in[],const PetscScalar v[],InsertMode addv)
285: {
286: Mat_MPIBAIJ *b = (Mat_MPIBAIJ*)mat->data;
287: int ierr,i,N = m*n;
288: MatScalar *vsingle;
291: if (N > b->setvalueslen) {
292: if (b->setvaluescopy) {PetscFree(b->setvaluescopy);}
293: PetscMalloc(N*sizeof(MatScalar),&b->setvaluescopy);
294: b->setvalueslen = N;
295: }
296: vsingle = b->setvaluescopy;
298: for (i=0; i<N; i++) {
299: vsingle[i] = v[i];
300: }
301: MatSetValues_MPIBAIJ_MatScalar(mat,m,im,n,in,vsingle,addv);
302: return(0);
303: }
307: int MatSetValuesBlocked_MPIBAIJ(Mat mat,int m,const int im[],int n,const int in[],const PetscScalar v[],InsertMode addv)
308: {
309: Mat_MPIBAIJ *b = (Mat_MPIBAIJ*)mat->data;
310: int ierr,i,N = m*n*b->bs2;
311: MatScalar *vsingle;
314: if (N > b->setvalueslen) {
315: if (b->setvaluescopy) {PetscFree(b->setvaluescopy);}
316: PetscMalloc(N*sizeof(MatScalar),&b->setvaluescopy);
317: b->setvalueslen = N;
318: }
319: vsingle = b->setvaluescopy;
320: for (i=0; i<N; i++) {
321: vsingle[i] = v[i];
322: }
323: MatSetValuesBlocked_MPIBAIJ_MatScalar(mat,m,im,n,in,vsingle,addv);
324: return(0);
325: }
329: int MatSetValues_MPIBAIJ_HT(Mat mat,int m,const int im[],int n,const int in[],const PetscScalar v[],InsertMode addv)
330: {
331: Mat_MPIBAIJ *b = (Mat_MPIBAIJ*)mat->data;
332: int ierr,i,N = m*n;
333: MatScalar *vsingle;
336: if (N > b->setvalueslen) {
337: if (b->setvaluescopy) {PetscFree(b->setvaluescopy);}
338: PetscMalloc(N*sizeof(MatScalar),&b->setvaluescopy);
339: b->setvalueslen = N;
340: }
341: vsingle = b->setvaluescopy;
342: for (i=0; i<N; i++) {
343: vsingle[i] = v[i];
344: }
345: MatSetValues_MPIBAIJ_HT_MatScalar(mat,m,im,n,in,vsingle,addv);
346: return(0);
347: }
351: int MatSetValuesBlocked_MPIBAIJ_HT(Mat mat,int m,const int im[],int n,const int in[],const PetscScalar v[],InsertMode addv)
352: {
353: Mat_MPIBAIJ *b = (Mat_MPIBAIJ*)mat->data;
354: int ierr,i,N = m*n*b->bs2;
355: MatScalar *vsingle;
358: if (N > b->setvalueslen) {
359: if (b->setvaluescopy) {PetscFree(b->setvaluescopy);}
360: PetscMalloc(N*sizeof(MatScalar),&b->setvaluescopy);
361: b->setvalueslen = N;
362: }
363: vsingle = b->setvaluescopy;
364: for (i=0; i<N; i++) {
365: vsingle[i] = v[i];
366: }
367: MatSetValuesBlocked_MPIBAIJ_HT_MatScalar(mat,m,im,n,in,vsingle,addv);
368: return(0);
369: }
370: #endif
374: int MatSetValues_MPIBAIJ_MatScalar(Mat mat,int m,const int im[],int n,const int in[],const MatScalar v[],InsertMode addv)
375: {
376: Mat_MPIBAIJ *baij = (Mat_MPIBAIJ*)mat->data;
377: MatScalar value;
378: PetscTruth roworiented = baij->roworiented;
379: int ierr,i,j,row,col;
380: int rstart_orig=baij->rstart_bs;
381: int rend_orig=baij->rend_bs,cstart_orig=baij->cstart_bs;
382: int cend_orig=baij->cend_bs,bs=baij->bs;
384: /* Some Variables required in the macro */
385: Mat A = baij->A;
386: Mat_SeqBAIJ *a = (Mat_SeqBAIJ*)(A)->data;
387: int *aimax=a->imax,*ai=a->i,*ailen=a->ilen,*aj=a->j;
388: MatScalar *aa=a->a;
390: Mat B = baij->B;
391: Mat_SeqBAIJ *b = (Mat_SeqBAIJ*)(B)->data;
392: int *bimax=b->imax,*bi=b->i,*bilen=b->ilen,*bj=b->j;
393: MatScalar *ba=b->a;
395: int *rp,ii,nrow,_i,rmax,N,brow,bcol;
396: int low,high,t,ridx,cidx,bs2=a->bs2;
397: MatScalar *ap,*bap;
400: for (i=0; i<m; i++) {
401: if (im[i] < 0) continue;
402: #if defined(PETSC_USE_BOPT_g)
403: if (im[i] >= mat->M) SETERRQ2(PETSC_ERR_ARG_OUTOFRANGE,"Row too large: row %d max %d",im[i],mat->M-1);
404: #endif
405: if (im[i] >= rstart_orig && im[i] < rend_orig) {
406: row = im[i] - rstart_orig;
407: for (j=0; j<n; j++) {
408: if (in[j] >= cstart_orig && in[j] < cend_orig){
409: col = in[j] - cstart_orig;
410: if (roworiented) value = v[i*n+j]; else value = v[i+j*m];
411: MatSetValues_SeqBAIJ_A_Private(row,col,value,addv);
412: /* MatSetValues_SeqBAIJ(baij->A,1,&row,1,&col,&value,addv); */
413: } else if (in[j] < 0) continue;
414: #if defined(PETSC_USE_BOPT_g)
415: else if (in[j] >= mat->N) {SETERRQ2(PETSC_ERR_ARG_OUTOFRANGE,"Column too large: col %d max %d",in[i],mat->N-1);}
416: #endif
417: else {
418: if (mat->was_assembled) {
419: if (!baij->colmap) {
420: CreateColmap_MPIBAIJ_Private(mat);
421: }
422: #if defined (PETSC_USE_CTABLE)
423: PetscTableFind(baij->colmap,in[j]/bs + 1,&col);
424: col = col - 1;
425: #else
426: col = baij->colmap[in[j]/bs] - 1;
427: #endif
428: if (col < 0 && !((Mat_SeqBAIJ*)(baij->A->data))->nonew) {
429: DisAssemble_MPIBAIJ(mat);
430: col = in[j];
431: /* Reinitialize the variables required by MatSetValues_SeqBAIJ_B_Private() */
432: B = baij->B;
433: b = (Mat_SeqBAIJ*)(B)->data;
434: bimax=b->imax;bi=b->i;bilen=b->ilen;bj=b->j;
435: ba=b->a;
436: } else col += in[j]%bs;
437: } else col = in[j];
438: if (roworiented) value = v[i*n+j]; else value = v[i+j*m];
439: MatSetValues_SeqBAIJ_B_Private(row,col,value,addv);
440: /* MatSetValues_SeqBAIJ(baij->B,1,&row,1,&col,&value,addv); */
441: }
442: }
443: } else {
444: if (!baij->donotstash) {
445: if (roworiented) {
446: MatStashValuesRow_Private(&mat->stash,im[i],n,in,v+i*n);
447: } else {
448: MatStashValuesCol_Private(&mat->stash,im[i],n,in,v+i,m);
449: }
450: }
451: }
452: }
453: return(0);
454: }
458: int MatSetValuesBlocked_MPIBAIJ_MatScalar(Mat mat,int m,const int im[],int n,const int in[],const MatScalar v[],InsertMode addv)
459: {
460: Mat_MPIBAIJ *baij = (Mat_MPIBAIJ*)mat->data;
461: const MatScalar *value;
462: MatScalar *barray=baij->barray;
463: PetscTruth roworiented = baij->roworiented;
464: int ierr,i,j,ii,jj,row,col,rstart=baij->rstart;
465: int rend=baij->rend,cstart=baij->cstart,stepval;
466: int cend=baij->cend,bs=baij->bs,bs2=baij->bs2;
467:
469: if(!barray) {
470: PetscMalloc(bs2*sizeof(MatScalar),&barray);
471: baij->barray = barray;
472: }
474: if (roworiented) {
475: stepval = (n-1)*bs;
476: } else {
477: stepval = (m-1)*bs;
478: }
479: for (i=0; i<m; i++) {
480: if (im[i] < 0) continue;
481: #if defined(PETSC_USE_BOPT_g)
482: if (im[i] >= baij->Mbs) SETERRQ2(PETSC_ERR_ARG_OUTOFRANGE,"Row too large, row %d max %d",im[i],baij->Mbs-1);
483: #endif
484: if (im[i] >= rstart && im[i] < rend) {
485: row = im[i] - rstart;
486: for (j=0; j<n; j++) {
487: /* If NumCol = 1 then a copy is not required */
488: if ((roworiented) && (n == 1)) {
489: barray = (MatScalar*)v + i*bs2;
490: } else if((!roworiented) && (m == 1)) {
491: barray = (MatScalar*)v + j*bs2;
492: } else { /* Here a copy is required */
493: if (roworiented) {
494: value = v + i*(stepval+bs)*bs + j*bs;
495: } else {
496: value = v + j*(stepval+bs)*bs + i*bs;
497: }
498: for (ii=0; ii<bs; ii++,value+=stepval) {
499: for (jj=0; jj<bs; jj++) {
500: *barray++ = *value++;
501: }
502: }
503: barray -=bs2;
504: }
505:
506: if (in[j] >= cstart && in[j] < cend){
507: col = in[j] - cstart;
508: MatSetValuesBlocked_SeqBAIJ_MatScalar(baij->A,1,&row,1,&col,barray,addv);
509: }
510: else if (in[j] < 0) continue;
511: #if defined(PETSC_USE_BOPT_g)
512: else if (in[j] >= baij->Nbs) {SETERRQ2(PETSC_ERR_ARG_OUTOFRANGE,"Column too large, col %d max %d",in[j],baij->Nbs-1);}
513: #endif
514: else {
515: if (mat->was_assembled) {
516: if (!baij->colmap) {
517: CreateColmap_MPIBAIJ_Private(mat);
518: }
520: #if defined(PETSC_USE_BOPT_g)
521: #if defined (PETSC_USE_CTABLE)
522: { int data;
523: PetscTableFind(baij->colmap,in[j]+1,&data);
524: if ((data - 1) % bs) SETERRQ(PETSC_ERR_PLIB,"Incorrect colmap");
525: }
526: #else
527: if ((baij->colmap[in[j]] - 1) % bs) SETERRQ(PETSC_ERR_PLIB,"Incorrect colmap");
528: #endif
529: #endif
530: #if defined (PETSC_USE_CTABLE)
531: PetscTableFind(baij->colmap,in[j]+1,&col);
532: col = (col - 1)/bs;
533: #else
534: col = (baij->colmap[in[j]] - 1)/bs;
535: #endif
536: if (col < 0 && !((Mat_SeqBAIJ*)(baij->A->data))->nonew) {
537: DisAssemble_MPIBAIJ(mat);
538: col = in[j];
539: }
540: }
541: else col = in[j];
542: MatSetValuesBlocked_SeqBAIJ_MatScalar(baij->B,1,&row,1,&col,barray,addv);
543: }
544: }
545: } else {
546: if (!baij->donotstash) {
547: if (roworiented) {
548: MatStashValuesRowBlocked_Private(&mat->bstash,im[i],n,in,v,m,n,i);
549: } else {
550: MatStashValuesColBlocked_Private(&mat->bstash,im[i],n,in,v,m,n,i);
551: }
552: }
553: }
554: }
555: return(0);
556: }
558: #define HASH_KEY 0.6180339887
559: #define HASH(size,key,tmp) (tmp = (key)*HASH_KEY,(int)((size)*(tmp-(int)tmp)))
560: /* #define HASH(size,key) ((int)((size)*fmod(((key)*HASH_KEY),1))) */
561: /* #define HASH(size,key,tmp) ((int)((size)*fmod(((key)*HASH_KEY),1))) */
564: int MatSetValues_MPIBAIJ_HT_MatScalar(Mat mat,int m,const int im[],int n,const int in[],const MatScalar v[],InsertMode addv)
565: {
566: Mat_MPIBAIJ *baij = (Mat_MPIBAIJ*)mat->data;
567: PetscTruth roworiented = baij->roworiented;
568: int ierr,i,j,row,col;
569: int rstart_orig=baij->rstart_bs;
570: int rend_orig=baij->rend_bs,Nbs=baij->Nbs;
571: int h1,key,size=baij->ht_size,bs=baij->bs,*HT=baij->ht,idx;
572: PetscReal tmp;
573: MatScalar **HD = baij->hd,value;
574: #if defined(PETSC_USE_BOPT_g)
575: int total_ct=baij->ht_total_ct,insert_ct=baij->ht_insert_ct;
576: #endif
580: for (i=0; i<m; i++) {
581: #if defined(PETSC_USE_BOPT_g)
582: if (im[i] < 0) SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,"Negative row");
583: if (im[i] >= mat->M) SETERRQ2(PETSC_ERR_ARG_OUTOFRANGE,"Row too large: row %d max %d",im[i],mat->M-1);
584: #endif
585: row = im[i];
586: if (row >= rstart_orig && row < rend_orig) {
587: for (j=0; j<n; j++) {
588: col = in[j];
589: if (roworiented) value = v[i*n+j]; else value = v[i+j*m];
590: /* Look up into the Hash Table */
591: key = (row/bs)*Nbs+(col/bs)+1;
592: h1 = HASH(size,key,tmp);
594:
595: idx = h1;
596: #if defined(PETSC_USE_BOPT_g)
597: insert_ct++;
598: total_ct++;
599: if (HT[idx] != key) {
600: for (idx=h1; (idx<size) && (HT[idx]!=key); idx++,total_ct++);
601: if (idx == size) {
602: for (idx=0; (idx<h1) && (HT[idx]!=key); idx++,total_ct++);
603: if (idx == h1) {
604: SETERRQ2(PETSC_ERR_ARG_OUTOFRANGE,"(%d,%d) has no entry in the hash table", row, col);
605: }
606: }
607: }
608: #else
609: if (HT[idx] != key) {
610: for (idx=h1; (idx<size) && (HT[idx]!=key); idx++);
611: if (idx == size) {
612: for (idx=0; (idx<h1) && (HT[idx]!=key); idx++);
613: if (idx == h1) {
614: SETERRQ2(PETSC_ERR_ARG_OUTOFRANGE,"(%d,%d) has no entry in the hash table", row, col);
615: }
616: }
617: }
618: #endif
619: /* A HASH table entry is found, so insert the values at the correct address */
620: if (addv == ADD_VALUES) *(HD[idx]+ (col % bs)*bs + (row % bs)) += value;
621: else *(HD[idx]+ (col % bs)*bs + (row % bs)) = value;
622: }
623: } else {
624: if (!baij->donotstash) {
625: if (roworiented) {
626: MatStashValuesRow_Private(&mat->stash,im[i],n,in,v+i*n);
627: } else {
628: MatStashValuesCol_Private(&mat->stash,im[i],n,in,v+i,m);
629: }
630: }
631: }
632: }
633: #if defined(PETSC_USE_BOPT_g)
634: baij->ht_total_ct = total_ct;
635: baij->ht_insert_ct = insert_ct;
636: #endif
637: return(0);
638: }
642: int MatSetValuesBlocked_MPIBAIJ_HT_MatScalar(Mat mat,int m,const int im[],int n,const int in[],const MatScalar v[],InsertMode addv)
643: {
644: Mat_MPIBAIJ *baij = (Mat_MPIBAIJ*)mat->data;
645: PetscTruth roworiented = baij->roworiented;
646: int ierr,i,j,ii,jj,row,col;
647: int rstart=baij->rstart ;
648: int rend=baij->rend,stepval,bs=baij->bs,bs2=baij->bs2;
649: int h1,key,size=baij->ht_size,idx,*HT=baij->ht,Nbs=baij->Nbs;
650: PetscReal tmp;
651: MatScalar **HD = baij->hd,*baij_a;
652: const MatScalar *v_t,*value;
653: #if defined(PETSC_USE_BOPT_g)
654: int total_ct=baij->ht_total_ct,insert_ct=baij->ht_insert_ct;
655: #endif
656:
659: if (roworiented) {
660: stepval = (n-1)*bs;
661: } else {
662: stepval = (m-1)*bs;
663: }
664: for (i=0; i<m; i++) {
665: #if defined(PETSC_USE_BOPT_g)
666: if (im[i] < 0) SETERRQ1(PETSC_ERR_ARG_OUTOFRANGE,"Negative row: %d",im[i]);
667: if (im[i] >= baij->Mbs) SETERRQ2(PETSC_ERR_ARG_OUTOFRANGE,"Row too large: row %d max %d",im[i],baij->Mbs-1);
668: #endif
669: row = im[i];
670: v_t = v + i*bs2;
671: if (row >= rstart && row < rend) {
672: for (j=0; j<n; j++) {
673: col = in[j];
675: /* Look up into the Hash Table */
676: key = row*Nbs+col+1;
677: h1 = HASH(size,key,tmp);
678:
679: idx = h1;
680: #if defined(PETSC_USE_BOPT_g)
681: total_ct++;
682: insert_ct++;
683: if (HT[idx] != key) {
684: for (idx=h1; (idx<size) && (HT[idx]!=key); idx++,total_ct++);
685: if (idx == size) {
686: for (idx=0; (idx<h1) && (HT[idx]!=key); idx++,total_ct++);
687: if (idx == h1) {
688: SETERRQ2(PETSC_ERR_ARG_OUTOFRANGE,"(%d,%d) has no entry in the hash table", row, col);
689: }
690: }
691: }
692: #else
693: if (HT[idx] != key) {
694: for (idx=h1; (idx<size) && (HT[idx]!=key); idx++);
695: if (idx == size) {
696: for (idx=0; (idx<h1) && (HT[idx]!=key); idx++);
697: if (idx == h1) {
698: SETERRQ2(PETSC_ERR_ARG_OUTOFRANGE,"(%d,%d) has no entry in the hash table", row, col);
699: }
700: }
701: }
702: #endif
703: baij_a = HD[idx];
704: if (roworiented) {
705: /*value = v + i*(stepval+bs)*bs + j*bs;*/
706: /* value = v + (i*(stepval+bs)+j)*bs; */
707: value = v_t;
708: v_t += bs;
709: if (addv == ADD_VALUES) {
710: for (ii=0; ii<bs; ii++,value+=stepval) {
711: for (jj=ii; jj<bs2; jj+=bs) {
712: baij_a[jj] += *value++;
713: }
714: }
715: } else {
716: for (ii=0; ii<bs; ii++,value+=stepval) {
717: for (jj=ii; jj<bs2; jj+=bs) {
718: baij_a[jj] = *value++;
719: }
720: }
721: }
722: } else {
723: value = v + j*(stepval+bs)*bs + i*bs;
724: if (addv == ADD_VALUES) {
725: for (ii=0; ii<bs; ii++,value+=stepval,baij_a+=bs) {
726: for (jj=0; jj<bs; jj++) {
727: baij_a[jj] += *value++;
728: }
729: }
730: } else {
731: for (ii=0; ii<bs; ii++,value+=stepval,baij_a+=bs) {
732: for (jj=0; jj<bs; jj++) {
733: baij_a[jj] = *value++;
734: }
735: }
736: }
737: }
738: }
739: } else {
740: if (!baij->donotstash) {
741: if (roworiented) {
742: MatStashValuesRowBlocked_Private(&mat->bstash,im[i],n,in,v,m,n,i);
743: } else {
744: MatStashValuesColBlocked_Private(&mat->bstash,im[i],n,in,v,m,n,i);
745: }
746: }
747: }
748: }
749: #if defined(PETSC_USE_BOPT_g)
750: baij->ht_total_ct = total_ct;
751: baij->ht_insert_ct = insert_ct;
752: #endif
753: return(0);
754: }
758: int MatGetValues_MPIBAIJ(Mat mat,int m,const int idxm[],int n,const int idxn[],PetscScalar v[])
759: {
760: Mat_MPIBAIJ *baij = (Mat_MPIBAIJ*)mat->data;
761: int bs=baij->bs,ierr,i,j,bsrstart = baij->rstart*bs,bsrend = baij->rend*bs;
762: int bscstart = baij->cstart*bs,bscend = baij->cend*bs,row,col,data;
765: for (i=0; i<m; i++) {
766: if (idxm[i] < 0) SETERRQ1(PETSC_ERR_ARG_OUTOFRANGE,"Negative row: %d",idxm[i]);
767: if (idxm[i] >= mat->M) SETERRQ2(PETSC_ERR_ARG_OUTOFRANGE,"Row too large: row %d max %d",idxm[i],mat->M-1);
768: if (idxm[i] >= bsrstart && idxm[i] < bsrend) {
769: row = idxm[i] - bsrstart;
770: for (j=0; j<n; j++) {
771: if (idxn[j] < 0) SETERRQ1(PETSC_ERR_ARG_OUTOFRANGE,"Negative column: %d",idxn[j]);
772: if (idxn[j] >= mat->N) SETERRQ2(PETSC_ERR_ARG_OUTOFRANGE,"Column too large: col %d max %d",idxn[j],mat->N-1);
773: if (idxn[j] >= bscstart && idxn[j] < bscend){
774: col = idxn[j] - bscstart;
775: MatGetValues_SeqBAIJ(baij->A,1,&row,1,&col,v+i*n+j);
776: } else {
777: if (!baij->colmap) {
778: CreateColmap_MPIBAIJ_Private(mat);
779: }
780: #if defined (PETSC_USE_CTABLE)
781: PetscTableFind(baij->colmap,idxn[j]/bs+1,&data);
782: data --;
783: #else
784: data = baij->colmap[idxn[j]/bs]-1;
785: #endif
786: if((data < 0) || (baij->garray[data/bs] != idxn[j]/bs)) *(v+i*n+j) = 0.0;
787: else {
788: col = data + idxn[j]%bs;
789: MatGetValues_SeqBAIJ(baij->B,1,&row,1,&col,v+i*n+j);
790: }
791: }
792: }
793: } else {
794: SETERRQ(PETSC_ERR_SUP,"Only local values currently supported");
795: }
796: }
797: return(0);
798: }
802: int MatNorm_MPIBAIJ(Mat mat,NormType type,PetscReal *nrm)
803: {
804: Mat_MPIBAIJ *baij = (Mat_MPIBAIJ*)mat->data;
805: Mat_SeqBAIJ *amat = (Mat_SeqBAIJ*)baij->A->data,*bmat = (Mat_SeqBAIJ*)baij->B->data;
806: int ierr,i,bs2=baij->bs2;
807: PetscReal sum = 0.0;
808: MatScalar *v;
811: if (baij->size == 1) {
812: MatNorm(baij->A,type,nrm);
813: } else {
814: if (type == NORM_FROBENIUS) {
815: v = amat->a;
816: for (i=0; i<amat->nz*bs2; i++) {
817: #if defined(PETSC_USE_COMPLEX)
818: sum += PetscRealPart(PetscConj(*v)*(*v)); v++;
819: #else
820: sum += (*v)*(*v); v++;
821: #endif
822: }
823: v = bmat->a;
824: for (i=0; i<bmat->nz*bs2; i++) {
825: #if defined(PETSC_USE_COMPLEX)
826: sum += PetscRealPart(PetscConj(*v)*(*v)); v++;
827: #else
828: sum += (*v)*(*v); v++;
829: #endif
830: }
831: MPI_Allreduce(&sum,nrm,1,MPIU_REAL,MPI_SUM,mat->comm);
832: *nrm = sqrt(*nrm);
833: } else {
834: SETERRQ(PETSC_ERR_SUP,"No support for this norm yet");
835: }
836: }
837: return(0);
838: }
841: /*
842: Creates the hash table, and sets the table
843: This table is created only once.
844: If new entried need to be added to the matrix
845: then the hash table has to be destroyed and
846: recreated.
847: */
850: int MatCreateHashTable_MPIBAIJ_Private(Mat mat,PetscReal factor)
851: {
852: Mat_MPIBAIJ *baij = (Mat_MPIBAIJ*)mat->data;
853: Mat A = baij->A,B=baij->B;
854: Mat_SeqBAIJ *a=(Mat_SeqBAIJ *)A->data,*b=(Mat_SeqBAIJ *)B->data;
855: int i,j,k,nz=a->nz+b->nz,h1,*ai=a->i,*aj=a->j,*bi=b->i,*bj=b->j;
856: int size,bs2=baij->bs2,rstart=baij->rstart,ierr;
857: int cstart=baij->cstart,*garray=baij->garray,row,col,Nbs=baij->Nbs;
858: int *HT,key;
859: MatScalar **HD;
860: PetscReal tmp;
861: #if defined(PETSC_USE_BOPT_g)
862: int ct=0,max=0;
863: #endif
866: baij->ht_size=(int)(factor*nz);
867: size = baij->ht_size;
869: if (baij->ht) {
870: return(0);
871: }
872:
873: /* Allocate Memory for Hash Table */
874: PetscMalloc((size)*(sizeof(int)+sizeof(MatScalar*))+1,&baij->hd);
875: baij->ht = (int*)(baij->hd + size);
876: HD = baij->hd;
877: HT = baij->ht;
880: PetscMemzero(HD,size*(sizeof(int)+sizeof(PetscScalar*)));
881:
883: /* Loop Over A */
884: for (i=0; i<a->mbs; i++) {
885: for (j=ai[i]; j<ai[i+1]; j++) {
886: row = i+rstart;
887: col = aj[j]+cstart;
888:
889: key = row*Nbs + col + 1;
890: h1 = HASH(size,key,tmp);
891: for (k=0; k<size; k++){
892: if (HT[(h1+k)%size] == 0.0) {
893: HT[(h1+k)%size] = key;
894: HD[(h1+k)%size] = a->a + j*bs2;
895: break;
896: #if defined(PETSC_USE_BOPT_g)
897: } else {
898: ct++;
899: #endif
900: }
901: }
902: #if defined(PETSC_USE_BOPT_g)
903: if (k> max) max = k;
904: #endif
905: }
906: }
907: /* Loop Over B */
908: for (i=0; i<b->mbs; i++) {
909: for (j=bi[i]; j<bi[i+1]; j++) {
910: row = i+rstart;
911: col = garray[bj[j]];
912: key = row*Nbs + col + 1;
913: h1 = HASH(size,key,tmp);
914: for (k=0; k<size; k++){
915: if (HT[(h1+k)%size] == 0.0) {
916: HT[(h1+k)%size] = key;
917: HD[(h1+k)%size] = b->a + j*bs2;
918: break;
919: #if defined(PETSC_USE_BOPT_g)
920: } else {
921: ct++;
922: #endif
923: }
924: }
925: #if defined(PETSC_USE_BOPT_g)
926: if (k> max) max = k;
927: #endif
928: }
929: }
930:
931: /* Print Summary */
932: #if defined(PETSC_USE_BOPT_g)
933: for (i=0,j=0; i<size; i++) {
934: if (HT[i]) {j++;}
935: }
936: PetscLogInfo(0,"MatCreateHashTable_MPIBAIJ_Private: Average Search = %5.2f,max search = %d\n",(j== 0)? 0.0:((PetscReal)(ct+j))/j,max);
937: #endif
938: return(0);
939: }
943: int MatAssemblyBegin_MPIBAIJ(Mat mat,MatAssemblyType mode)
944: {
945: Mat_MPIBAIJ *baij = (Mat_MPIBAIJ*)mat->data;
946: int ierr,nstash,reallocs;
947: InsertMode addv;
950: if (baij->donotstash) {
951: return(0);
952: }
954: /* make sure all processors are either in INSERTMODE or ADDMODE */
955: MPI_Allreduce(&mat->insertmode,&addv,1,MPI_INT,MPI_BOR,mat->comm);
956: if (addv == (ADD_VALUES|INSERT_VALUES)) {
957: SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Some processors inserted others added");
958: }
959: mat->insertmode = addv; /* in case this processor had no cache */
961: MatStashScatterBegin_Private(&mat->stash,baij->rowners_bs);
962: MatStashScatterBegin_Private(&mat->bstash,baij->rowners);
963: MatStashGetInfo_Private(&mat->stash,&nstash,&reallocs);
964: PetscLogInfo(0,"MatAssemblyBegin_MPIBAIJ:Stash has %d entries,uses %d mallocs.\n",nstash,reallocs);
965: MatStashGetInfo_Private(&mat->bstash,&nstash,&reallocs);
966: PetscLogInfo(0,"MatAssemblyBegin_MPIBAIJ:Block-Stash has %d entries, uses %d mallocs.\n",nstash,reallocs);
967: return(0);
968: }
972: int MatAssemblyEnd_MPIBAIJ(Mat mat,MatAssemblyType mode)
973: {
974: Mat_MPIBAIJ *baij=(Mat_MPIBAIJ*)mat->data;
975: Mat_SeqBAIJ *a=(Mat_SeqBAIJ*)baij->A->data,*b=(Mat_SeqBAIJ*)baij->B->data;
976: int i,j,rstart,ncols,n,ierr,flg,bs2=baij->bs2;
977: int *row,*col,other_disassembled;
978: PetscTruth r1,r2,r3;
979: MatScalar *val;
980: InsertMode addv = mat->insertmode;
983: if (!baij->donotstash) {
984: while (1) {
985: MatStashScatterGetMesg_Private(&mat->stash,&n,&row,&col,&val,&flg);
986: if (!flg) break;
988: for (i=0; i<n;) {
989: /* Now identify the consecutive vals belonging to the same row */
990: for (j=i,rstart=row[j]; j<n; j++) { if (row[j] != rstart) break; }
991: if (j < n) ncols = j-i;
992: else ncols = n-i;
993: /* Now assemble all these values with a single function call */
994: MatSetValues_MPIBAIJ_MatScalar(mat,1,row+i,ncols,col+i,val+i,addv);
995: i = j;
996: }
997: }
998: MatStashScatterEnd_Private(&mat->stash);
999: /* Now process the block-stash. Since the values are stashed column-oriented,
1000: set the roworiented flag to column oriented, and after MatSetValues()
1001: restore the original flags */
1002: r1 = baij->roworiented;
1003: r2 = a->roworiented;
1004: r3 = b->roworiented;
1005: baij->roworiented = PETSC_FALSE;
1006: a->roworiented = PETSC_FALSE;
1007: b->roworiented = PETSC_FALSE;
1008: while (1) {
1009: MatStashScatterGetMesg_Private(&mat->bstash,&n,&row,&col,&val,&flg);
1010: if (!flg) break;
1011:
1012: for (i=0; i<n;) {
1013: /* Now identify the consecutive vals belonging to the same row */
1014: for (j=i,rstart=row[j]; j<n; j++) { if (row[j] != rstart) break; }
1015: if (j < n) ncols = j-i;
1016: else ncols = n-i;
1017: MatSetValuesBlocked_MPIBAIJ_MatScalar(mat,1,row+i,ncols,col+i,val+i*bs2,addv);
1018: i = j;
1019: }
1020: }
1021: MatStashScatterEnd_Private(&mat->bstash);
1022: baij->roworiented = r1;
1023: a->roworiented = r2;
1024: b->roworiented = r3;
1025: }
1027: MatAssemblyBegin(baij->A,mode);
1028: MatAssemblyEnd(baij->A,mode);
1030: /* determine if any processor has disassembled, if so we must
1031: also disassemble ourselfs, in order that we may reassemble. */
1032: /*
1033: if nonzero structure of submatrix B cannot change then we know that
1034: no processor disassembled thus we can skip this stuff
1035: */
1036: if (!((Mat_SeqBAIJ*)baij->B->data)->nonew) {
1037: MPI_Allreduce(&mat->was_assembled,&other_disassembled,1,MPI_INT,MPI_PROD,mat->comm);
1038: if (mat->was_assembled && !other_disassembled) {
1039: DisAssemble_MPIBAIJ(mat);
1040: }
1041: }
1043: if (!mat->was_assembled && mode == MAT_FINAL_ASSEMBLY) {
1044: MatSetUpMultiply_MPIBAIJ(mat);
1045: }
1046: MatAssemblyBegin(baij->B,mode);
1047: MatAssemblyEnd(baij->B,mode);
1048:
1049: #if defined(PETSC_USE_BOPT_g)
1050: if (baij->ht && mode== MAT_FINAL_ASSEMBLY) {
1051: PetscLogInfo(0,"MatAssemblyEnd_MPIBAIJ:Average Hash Table Search in MatSetValues = %5.2f\n",((PetscReal)baij->ht_total_ct)/baij->ht_insert_ct);
1052: baij->ht_total_ct = 0;
1053: baij->ht_insert_ct = 0;
1054: }
1055: #endif
1056: if (baij->ht_flag && !baij->ht && mode == MAT_FINAL_ASSEMBLY) {
1057: MatCreateHashTable_MPIBAIJ_Private(mat,baij->ht_fact);
1058: mat->ops->setvalues = MatSetValues_MPIBAIJ_HT;
1059: mat->ops->setvaluesblocked = MatSetValuesBlocked_MPIBAIJ_HT;
1060: }
1062: if (baij->rowvalues) {
1063: PetscFree(baij->rowvalues);
1064: baij->rowvalues = 0;
1065: }
1066: return(0);
1067: }
1071: static int MatView_MPIBAIJ_ASCIIorDraworSocket(Mat mat,PetscViewer viewer)
1072: {
1073: Mat_MPIBAIJ *baij = (Mat_MPIBAIJ*)mat->data;
1074: int ierr,bs = baij->bs,size = baij->size,rank = baij->rank;
1075: PetscTruth isascii,isdraw;
1076: PetscViewer sviewer;
1077: PetscViewerFormat format;
1080: /* printf(" MatView_MPIBAIJ_ASCIIorDraworSocket is called ...\n"); */
1081: PetscTypeCompare((PetscObject)viewer,PETSC_VIEWER_ASCII,&isascii);
1082: PetscTypeCompare((PetscObject)viewer,PETSC_VIEWER_DRAW,&isdraw);
1083: if (isascii) {
1084: PetscViewerGetFormat(viewer,&format);
1085: if (format == PETSC_VIEWER_ASCII_INFO_DETAIL) {
1086: MatInfo info;
1087: MPI_Comm_rank(mat->comm,&rank);
1088: MatGetInfo(mat,MAT_LOCAL,&info);
1089: PetscViewerASCIISynchronizedPrintf(viewer,"[%d] Local rows %d nz %d nz alloced %d bs %d mem %d\n",
1090: rank,mat->m,(int)info.nz_used*bs,(int)info.nz_allocated*bs,
1091: baij->bs,(int)info.memory);
1092: MatGetInfo(baij->A,MAT_LOCAL,&info);
1093: PetscViewerASCIISynchronizedPrintf(viewer,"[%d] on-diagonal part: nz %d \n",rank,(int)info.nz_used*bs);
1094: MatGetInfo(baij->B,MAT_LOCAL,&info);
1095: PetscViewerASCIISynchronizedPrintf(viewer,"[%d] off-diagonal part: nz %d \n",rank,(int)info.nz_used*bs);
1096: PetscViewerFlush(viewer);
1097: VecScatterView(baij->Mvctx,viewer);
1098: return(0);
1099: } else if (format == PETSC_VIEWER_ASCII_INFO) {
1100: PetscViewerASCIIPrintf(viewer," block size is %d\n",bs);
1101: return(0);
1102: } else if (format == PETSC_VIEWER_ASCII_FACTOR_INFO) {
1103: return(0);
1104: }
1105: }
1107: if (isdraw) {
1108: PetscDraw draw;
1109: PetscTruth isnull;
1110: PetscViewerDrawGetDraw(viewer,0,&draw);
1111: PetscDrawIsNull(draw,&isnull); if (isnull) return(0);
1112: }
1114: if (size == 1) {
1115: PetscObjectSetName((PetscObject)baij->A,mat->name);
1116: MatView(baij->A,viewer);
1117: } else {
1118: /* assemble the entire matrix onto first processor. */
1119: Mat A;
1120: Mat_SeqBAIJ *Aloc;
1121: int M = mat->M,N = mat->N,*ai,*aj,col,i,j,k,*rvals,mbs = baij->mbs;
1122: MatScalar *a;
1124: /* Here we are creating a temporary matrix, so will assume MPIBAIJ is acceptable */
1125: /* Perhaps this should be the type of mat? */
1126: if (!rank) {
1127: MatCreate(mat->comm,M,N,M,N,&A);
1128: } else {
1129: MatCreate(mat->comm,0,0,M,N,&A);
1130: }
1131: MatSetType(A,MATMPIBAIJ);
1132: MatMPIBAIJSetPreallocation(A,baij->bs,0,PETSC_NULL,0,PETSC_NULL);
1133: PetscLogObjectParent(mat,A);
1135: /* copy over the A part */
1136: Aloc = (Mat_SeqBAIJ*)baij->A->data;
1137: ai = Aloc->i; aj = Aloc->j; a = Aloc->a;
1138: PetscMalloc(bs*sizeof(int),&rvals);
1140: for (i=0; i<mbs; i++) {
1141: rvals[0] = bs*(baij->rstart + i);
1142: for (j=1; j<bs; j++) { rvals[j] = rvals[j-1] + 1; }
1143: for (j=ai[i]; j<ai[i+1]; j++) {
1144: col = (baij->cstart+aj[j])*bs;
1145: for (k=0; k<bs; k++) {
1146: MatSetValues_MPIBAIJ_MatScalar(A,bs,rvals,1,&col,a,INSERT_VALUES);
1147: col++; a += bs;
1148: }
1149: }
1150: }
1151: /* copy over the B part */
1152: Aloc = (Mat_SeqBAIJ*)baij->B->data;
1153: ai = Aloc->i; aj = Aloc->j; a = Aloc->a;
1154: for (i=0; i<mbs; i++) {
1155: rvals[0] = bs*(baij->rstart + i);
1156: for (j=1; j<bs; j++) { rvals[j] = rvals[j-1] + 1; }
1157: for (j=ai[i]; j<ai[i+1]; j++) {
1158: col = baij->garray[aj[j]]*bs;
1159: for (k=0; k<bs; k++) {
1160: MatSetValues_MPIBAIJ_MatScalar(A,bs,rvals,1,&col,a,INSERT_VALUES);
1161: col++; a += bs;
1162: }
1163: }
1164: }
1165: PetscFree(rvals);
1166: MatAssemblyBegin(A,MAT_FINAL_ASSEMBLY);
1167: MatAssemblyEnd(A,MAT_FINAL_ASSEMBLY);
1168: /*
1169: Everyone has to call to draw the matrix since the graphics waits are
1170: synchronized across all processors that share the PetscDraw object
1171: */
1172: PetscViewerGetSingleton(viewer,&sviewer);
1173: if (!rank) {
1174: PetscObjectSetName((PetscObject)((Mat_MPIBAIJ*)(A->data))->A,mat->name);
1175: MatView(((Mat_MPIBAIJ*)(A->data))->A,sviewer);
1176: }
1177: PetscViewerRestoreSingleton(viewer,&sviewer);
1178: MatDestroy(A);
1179: }
1180: return(0);
1181: }
1185: int MatView_MPIBAIJ(Mat mat,PetscViewer viewer)
1186: {
1187: int ierr;
1188: PetscTruth isascii,isdraw,issocket,isbinary;
1191: PetscTypeCompare((PetscObject)viewer,PETSC_VIEWER_ASCII,&isascii);
1192: PetscTypeCompare((PetscObject)viewer,PETSC_VIEWER_DRAW,&isdraw);
1193: PetscTypeCompare((PetscObject)viewer,PETSC_VIEWER_SOCKET,&issocket);
1194: PetscTypeCompare((PetscObject)viewer,PETSC_VIEWER_BINARY,&isbinary);
1195: if (isascii || isdraw || issocket || isbinary) {
1196: MatView_MPIBAIJ_ASCIIorDraworSocket(mat,viewer);
1197: } else {
1198: SETERRQ1(1,"Viewer type %s not supported by MPIBAIJ matrices",((PetscObject)viewer)->type_name);
1199: }
1200: return(0);
1201: }
1205: int MatDestroy_MPIBAIJ(Mat mat)
1206: {
1207: Mat_MPIBAIJ *baij = (Mat_MPIBAIJ*)mat->data;
1208: int ierr;
1211: #if defined(PETSC_USE_LOG)
1212: PetscLogObjectState((PetscObject)mat,"Rows=%d,Cols=%d",mat->M,mat->N);
1213: #endif
1214: MatStashDestroy_Private(&mat->stash);
1215: MatStashDestroy_Private(&mat->bstash);
1216: PetscFree(baij->rowners);
1217: MatDestroy(baij->A);
1218: MatDestroy(baij->B);
1219: #if defined (PETSC_USE_CTABLE)
1220: if (baij->colmap) {PetscTableDelete(baij->colmap);}
1221: #else
1222: if (baij->colmap) {PetscFree(baij->colmap);}
1223: #endif
1224: if (baij->garray) {PetscFree(baij->garray);}
1225: if (baij->lvec) {VecDestroy(baij->lvec);}
1226: if (baij->Mvctx) {VecScatterDestroy(baij->Mvctx);}
1227: if (baij->rowvalues) {PetscFree(baij->rowvalues);}
1228: if (baij->barray) {PetscFree(baij->barray);}
1229: if (baij->hd) {PetscFree(baij->hd);}
1230: #if defined(PETSC_USE_MAT_SINGLE)
1231: if (baij->setvaluescopy) {PetscFree(baij->setvaluescopy);}
1232: #endif
1233: PetscFree(baij);
1234: return(0);
1235: }
1239: int MatMult_MPIBAIJ(Mat A,Vec xx,Vec yy)
1240: {
1241: Mat_MPIBAIJ *a = (Mat_MPIBAIJ*)A->data;
1242: int ierr,nt;
1245: VecGetLocalSize(xx,&nt);
1246: if (nt != A->n) {
1247: SETERRQ(PETSC_ERR_ARG_SIZ,"Incompatible partition of A and xx");
1248: }
1249: VecGetLocalSize(yy,&nt);
1250: if (nt != A->m) {
1251: SETERRQ(PETSC_ERR_ARG_SIZ,"Incompatible parition of A and yy");
1252: }
1253: VecScatterBegin(xx,a->lvec,INSERT_VALUES,SCATTER_FORWARD,a->Mvctx);
1254: (*a->A->ops->mult)(a->A,xx,yy);
1255: VecScatterEnd(xx,a->lvec,INSERT_VALUES,SCATTER_FORWARD,a->Mvctx);
1256: (*a->B->ops->multadd)(a->B,a->lvec,yy,yy);
1257: VecScatterPostRecvs(xx,a->lvec,INSERT_VALUES,SCATTER_FORWARD,a->Mvctx);
1258: return(0);
1259: }
1263: int MatMultAdd_MPIBAIJ(Mat A,Vec xx,Vec yy,Vec zz)
1264: {
1265: Mat_MPIBAIJ *a = (Mat_MPIBAIJ*)A->data;
1266: int ierr;
1269: VecScatterBegin(xx,a->lvec,INSERT_VALUES,SCATTER_FORWARD,a->Mvctx);
1270: (*a->A->ops->multadd)(a->A,xx,yy,zz);
1271: VecScatterEnd(xx,a->lvec,INSERT_VALUES,SCATTER_FORWARD,a->Mvctx);
1272: (*a->B->ops->multadd)(a->B,a->lvec,zz,zz);
1273: return(0);
1274: }
1278: int MatMultTranspose_MPIBAIJ(Mat A,Vec xx,Vec yy)
1279: {
1280: Mat_MPIBAIJ *a = (Mat_MPIBAIJ*)A->data;
1281: int ierr;
1284: /* do nondiagonal part */
1285: (*a->B->ops->multtranspose)(a->B,xx,a->lvec);
1286: /* send it on its way */
1287: VecScatterBegin(a->lvec,yy,ADD_VALUES,SCATTER_REVERSE,a->Mvctx);
1288: /* do local part */
1289: (*a->A->ops->multtranspose)(a->A,xx,yy);
1290: /* receive remote parts: note this assumes the values are not actually */
1291: /* inserted in yy until the next line, which is true for my implementation*/
1292: /* but is not perhaps always true. */
1293: VecScatterEnd(a->lvec,yy,ADD_VALUES,SCATTER_REVERSE,a->Mvctx);
1294: return(0);
1295: }
1299: int MatMultTransposeAdd_MPIBAIJ(Mat A,Vec xx,Vec yy,Vec zz)
1300: {
1301: Mat_MPIBAIJ *a = (Mat_MPIBAIJ*)A->data;
1302: int ierr;
1305: /* do nondiagonal part */
1306: (*a->B->ops->multtranspose)(a->B,xx,a->lvec);
1307: /* send it on its way */
1308: VecScatterBegin(a->lvec,zz,ADD_VALUES,SCATTER_REVERSE,a->Mvctx);
1309: /* do local part */
1310: (*a->A->ops->multtransposeadd)(a->A,xx,yy,zz);
1311: /* receive remote parts: note this assumes the values are not actually */
1312: /* inserted in yy until the next line, which is true for my implementation*/
1313: /* but is not perhaps always true. */
1314: VecScatterEnd(a->lvec,zz,ADD_VALUES,SCATTER_REVERSE,a->Mvctx);
1315: return(0);
1316: }
1318: /*
1319: This only works correctly for square matrices where the subblock A->A is the
1320: diagonal block
1321: */
1324: int MatGetDiagonal_MPIBAIJ(Mat A,Vec v)
1325: {
1326: Mat_MPIBAIJ *a = (Mat_MPIBAIJ*)A->data;
1327: int ierr;
1330: if (A->M != A->N) SETERRQ(PETSC_ERR_SUP,"Supports only square matrix where A->A is diag block");
1331: MatGetDiagonal(a->A,v);
1332: return(0);
1333: }
1337: int MatScale_MPIBAIJ(const PetscScalar *aa,Mat A)
1338: {
1339: Mat_MPIBAIJ *a = (Mat_MPIBAIJ*)A->data;
1340: int ierr;
1343: MatScale(aa,a->A);
1344: MatScale(aa,a->B);
1345: return(0);
1346: }
1350: int MatGetRow_MPIBAIJ(Mat matin,int row,int *nz,int **idx,PetscScalar **v)
1351: {
1352: Mat_MPIBAIJ *mat = (Mat_MPIBAIJ*)matin->data;
1353: PetscScalar *vworkA,*vworkB,**pvA,**pvB,*v_p;
1354: int bs = mat->bs,bs2 = mat->bs2,i,ierr,*cworkA,*cworkB,**pcA,**pcB;
1355: int nztot,nzA,nzB,lrow,brstart = mat->rstart*bs,brend = mat->rend*bs;
1356: int *cmap,*idx_p,cstart = mat->cstart;
1359: if (mat->getrowactive == PETSC_TRUE) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Already active");
1360: mat->getrowactive = PETSC_TRUE;
1362: if (!mat->rowvalues && (idx || v)) {
1363: /*
1364: allocate enough space to hold information from the longest row.
1365: */
1366: Mat_SeqBAIJ *Aa = (Mat_SeqBAIJ*)mat->A->data,*Ba = (Mat_SeqBAIJ*)mat->B->data;
1367: int max = 1,mbs = mat->mbs,tmp;
1368: for (i=0; i<mbs; i++) {
1369: tmp = Aa->i[i+1] - Aa->i[i] + Ba->i[i+1] - Ba->i[i];
1370: if (max < tmp) { max = tmp; }
1371: }
1372: PetscMalloc(max*bs2*(sizeof(int)+sizeof(PetscScalar)),&mat->rowvalues);
1373: mat->rowindices = (int*)(mat->rowvalues + max*bs2);
1374: }
1375:
1376: if (row < brstart || row >= brend) SETERRQ(PETSC_ERR_SUP,"Only local rows")
1377: lrow = row - brstart;
1379: pvA = &vworkA; pcA = &cworkA; pvB = &vworkB; pcB = &cworkB;
1380: if (!v) {pvA = 0; pvB = 0;}
1381: if (!idx) {pcA = 0; if (!v) pcB = 0;}
1382: (*mat->A->ops->getrow)(mat->A,lrow,&nzA,pcA,pvA);
1383: (*mat->B->ops->getrow)(mat->B,lrow,&nzB,pcB,pvB);
1384: nztot = nzA + nzB;
1386: cmap = mat->garray;
1387: if (v || idx) {
1388: if (nztot) {
1389: /* Sort by increasing column numbers, assuming A and B already sorted */
1390: int imark = -1;
1391: if (v) {
1392: *v = v_p = mat->rowvalues;
1393: for (i=0; i<nzB; i++) {
1394: if (cmap[cworkB[i]/bs] < cstart) v_p[i] = vworkB[i];
1395: else break;
1396: }
1397: imark = i;
1398: for (i=0; i<nzA; i++) v_p[imark+i] = vworkA[i];
1399: for (i=imark; i<nzB; i++) v_p[nzA+i] = vworkB[i];
1400: }
1401: if (idx) {
1402: *idx = idx_p = mat->rowindices;
1403: if (imark > -1) {
1404: for (i=0; i<imark; i++) {
1405: idx_p[i] = cmap[cworkB[i]/bs]*bs + cworkB[i]%bs;
1406: }
1407: } else {
1408: for (i=0; i<nzB; i++) {
1409: if (cmap[cworkB[i]/bs] < cstart)
1410: idx_p[i] = cmap[cworkB[i]/bs]*bs + cworkB[i]%bs ;
1411: else break;
1412: }
1413: imark = i;
1414: }
1415: for (i=0; i<nzA; i++) idx_p[imark+i] = cstart*bs + cworkA[i];
1416: for (i=imark; i<nzB; i++) idx_p[nzA+i] = cmap[cworkB[i]/bs]*bs + cworkB[i]%bs ;
1417: }
1418: } else {
1419: if (idx) *idx = 0;
1420: if (v) *v = 0;
1421: }
1422: }
1423: *nz = nztot;
1424: (*mat->A->ops->restorerow)(mat->A,lrow,&nzA,pcA,pvA);
1425: (*mat->B->ops->restorerow)(mat->B,lrow,&nzB,pcB,pvB);
1426: return(0);
1427: }
1431: int MatRestoreRow_MPIBAIJ(Mat mat,int row,int *nz,int **idx,PetscScalar **v)
1432: {
1433: Mat_MPIBAIJ *baij = (Mat_MPIBAIJ*)mat->data;
1436: if (baij->getrowactive == PETSC_FALSE) {
1437: SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"MatGetRow not called");
1438: }
1439: baij->getrowactive = PETSC_FALSE;
1440: return(0);
1441: }
1445: int MatGetBlockSize_MPIBAIJ(Mat mat,int *bs)
1446: {
1447: Mat_MPIBAIJ *baij = (Mat_MPIBAIJ*)mat->data;
1450: *bs = baij->bs;
1451: return(0);
1452: }
1456: int MatZeroEntries_MPIBAIJ(Mat A)
1457: {
1458: Mat_MPIBAIJ *l = (Mat_MPIBAIJ*)A->data;
1459: int ierr;
1462: MatZeroEntries(l->A);
1463: MatZeroEntries(l->B);
1464: return(0);
1465: }
1469: int MatGetInfo_MPIBAIJ(Mat matin,MatInfoType flag,MatInfo *info)
1470: {
1471: Mat_MPIBAIJ *a = (Mat_MPIBAIJ*)matin->data;
1472: Mat A = a->A,B = a->B;
1473: int ierr;
1474: PetscReal isend[5],irecv[5];
1477: info->block_size = (PetscReal)a->bs;
1478: MatGetInfo(A,MAT_LOCAL,info);
1479: isend[0] = info->nz_used; isend[1] = info->nz_allocated; isend[2] = info->nz_unneeded;
1480: isend[3] = info->memory; isend[4] = info->mallocs;
1481: MatGetInfo(B,MAT_LOCAL,info);
1482: isend[0] += info->nz_used; isend[1] += info->nz_allocated; isend[2] += info->nz_unneeded;
1483: isend[3] += info->memory; isend[4] += info->mallocs;
1484: if (flag == MAT_LOCAL) {
1485: info->nz_used = isend[0];
1486: info->nz_allocated = isend[1];
1487: info->nz_unneeded = isend[2];
1488: info->memory = isend[3];
1489: info->mallocs = isend[4];
1490: } else if (flag == MAT_GLOBAL_MAX) {
1491: MPI_Allreduce(isend,irecv,5,MPIU_REAL,MPI_MAX,matin->comm);
1492: info->nz_used = irecv[0];
1493: info->nz_allocated = irecv[1];
1494: info->nz_unneeded = irecv[2];
1495: info->memory = irecv[3];
1496: info->mallocs = irecv[4];
1497: } else if (flag == MAT_GLOBAL_SUM) {
1498: MPI_Allreduce(isend,irecv,5,MPIU_REAL,MPI_SUM,matin->comm);
1499: info->nz_used = irecv[0];
1500: info->nz_allocated = irecv[1];
1501: info->nz_unneeded = irecv[2];
1502: info->memory = irecv[3];
1503: info->mallocs = irecv[4];
1504: } else {
1505: SETERRQ1(1,"Unknown MatInfoType argument %d",flag);
1506: }
1507: info->rows_global = (PetscReal)A->M;
1508: info->columns_global = (PetscReal)A->N;
1509: info->rows_local = (PetscReal)A->m;
1510: info->columns_local = (PetscReal)A->N;
1511: info->fill_ratio_given = 0; /* no parallel LU/ILU/Cholesky */
1512: info->fill_ratio_needed = 0;
1513: info->factor_mallocs = 0;
1514: return(0);
1515: }
1519: int MatSetOption_MPIBAIJ(Mat A,MatOption op)
1520: {
1521: Mat_MPIBAIJ *a = (Mat_MPIBAIJ*)A->data;
1522: int ierr;
1525: switch (op) {
1526: case MAT_NO_NEW_NONZERO_LOCATIONS:
1527: case MAT_YES_NEW_NONZERO_LOCATIONS:
1528: case MAT_COLUMNS_UNSORTED:
1529: case MAT_COLUMNS_SORTED:
1530: case MAT_NEW_NONZERO_ALLOCATION_ERR:
1531: case MAT_KEEP_ZEROED_ROWS:
1532: case MAT_NEW_NONZERO_LOCATION_ERR:
1533: MatSetOption(a->A,op);
1534: MatSetOption(a->B,op);
1535: break;
1536: case MAT_ROW_ORIENTED:
1537: a->roworiented = PETSC_TRUE;
1538: MatSetOption(a->A,op);
1539: MatSetOption(a->B,op);
1540: break;
1541: case MAT_ROWS_SORTED:
1542: case MAT_ROWS_UNSORTED:
1543: case MAT_YES_NEW_DIAGONALS:
1544: PetscLogInfo(A,"Info:MatSetOption_MPIBAIJ:Option ignored\n");
1545: break;
1546: case MAT_COLUMN_ORIENTED:
1547: a->roworiented = PETSC_FALSE;
1548: MatSetOption(a->A,op);
1549: MatSetOption(a->B,op);
1550: break;
1551: case MAT_IGNORE_OFF_PROC_ENTRIES:
1552: a->donotstash = PETSC_TRUE;
1553: break;
1554: case MAT_NO_NEW_DIAGONALS:
1555: SETERRQ(PETSC_ERR_SUP,"MAT_NO_NEW_DIAGONALS");
1556: case MAT_USE_HASH_TABLE:
1557: a->ht_flag = PETSC_TRUE;
1558: break;
1559: case MAT_SYMMETRIC:
1560: case MAT_STRUCTURALLY_SYMMETRIC:
1561: case MAT_NOT_SYMMETRIC:
1562: case MAT_NOT_STRUCTURALLY_SYMMETRIC:
1563: case MAT_HERMITIAN:
1564: case MAT_NOT_HERMITIAN:
1565: case MAT_SYMMETRY_ETERNAL:
1566: case MAT_NOT_SYMMETRY_ETERNAL:
1567: break;
1568: default:
1569: SETERRQ(PETSC_ERR_SUP,"unknown option");
1570: }
1571: return(0);
1572: }
1576: int MatTranspose_MPIBAIJ(Mat A,Mat *matout)
1577: {
1578: Mat_MPIBAIJ *baij = (Mat_MPIBAIJ*)A->data;
1579: Mat_SeqBAIJ *Aloc;
1580: Mat B;
1581: int ierr,M=A->M,N=A->N,*ai,*aj,i,*rvals,j,k,col;
1582: int bs=baij->bs,mbs=baij->mbs;
1583: MatScalar *a;
1584:
1586: if (!matout && M != N) SETERRQ(PETSC_ERR_ARG_SIZ,"Square matrix only for in-place");
1587: MatCreate(A->comm,A->n,A->m,N,M,&B);
1588: MatSetType(B,A->type_name);
1589: MatMPIBAIJSetPreallocation(B,baij->bs,0,PETSC_NULL,0,PETSC_NULL);
1590:
1591: /* copy over the A part */
1592: Aloc = (Mat_SeqBAIJ*)baij->A->data;
1593: ai = Aloc->i; aj = Aloc->j; a = Aloc->a;
1594: PetscMalloc(bs*sizeof(int),&rvals);
1595:
1596: for (i=0; i<mbs; i++) {
1597: rvals[0] = bs*(baij->rstart + i);
1598: for (j=1; j<bs; j++) { rvals[j] = rvals[j-1] + 1; }
1599: for (j=ai[i]; j<ai[i+1]; j++) {
1600: col = (baij->cstart+aj[j])*bs;
1601: for (k=0; k<bs; k++) {
1602: MatSetValues_MPIBAIJ_MatScalar(B,1,&col,bs,rvals,a,INSERT_VALUES);
1603: col++; a += bs;
1604: }
1605: }
1606: }
1607: /* copy over the B part */
1608: Aloc = (Mat_SeqBAIJ*)baij->B->data;
1609: ai = Aloc->i; aj = Aloc->j; a = Aloc->a;
1610: for (i=0; i<mbs; i++) {
1611: rvals[0] = bs*(baij->rstart + i);
1612: for (j=1; j<bs; j++) { rvals[j] = rvals[j-1] + 1; }
1613: for (j=ai[i]; j<ai[i+1]; j++) {
1614: col = baij->garray[aj[j]]*bs;
1615: for (k=0; k<bs; k++) {
1616: MatSetValues_MPIBAIJ_MatScalar(B,1,&col,bs,rvals,a,INSERT_VALUES);
1617: col++; a += bs;
1618: }
1619: }
1620: }
1621: PetscFree(rvals);
1622: MatAssemblyBegin(B,MAT_FINAL_ASSEMBLY);
1623: MatAssemblyEnd(B,MAT_FINAL_ASSEMBLY);
1624:
1625: if (matout) {
1626: *matout = B;
1627: } else {
1628: MatHeaderCopy(A,B);
1629: }
1630: return(0);
1631: }
1635: int MatDiagonalScale_MPIBAIJ(Mat mat,Vec ll,Vec rr)
1636: {
1637: Mat_MPIBAIJ *baij = (Mat_MPIBAIJ*)mat->data;
1638: Mat a = baij->A,b = baij->B;
1639: int ierr,s1,s2,s3;
1642: MatGetLocalSize(mat,&s2,&s3);
1643: if (rr) {
1644: VecGetLocalSize(rr,&s1);
1645: if (s1!=s3) SETERRQ(PETSC_ERR_ARG_SIZ,"right vector non-conforming local size");
1646: /* Overlap communication with computation. */
1647: VecScatterBegin(rr,baij->lvec,INSERT_VALUES,SCATTER_FORWARD,baij->Mvctx);
1648: }
1649: if (ll) {
1650: VecGetLocalSize(ll,&s1);
1651: if (s1!=s2) SETERRQ(PETSC_ERR_ARG_SIZ,"left vector non-conforming local size");
1652: (*b->ops->diagonalscale)(b,ll,PETSC_NULL);
1653: }
1654: /* scale the diagonal block */
1655: (*a->ops->diagonalscale)(a,ll,rr);
1657: if (rr) {
1658: /* Do a scatter end and then right scale the off-diagonal block */
1659: VecScatterEnd(rr,baij->lvec,INSERT_VALUES,SCATTER_FORWARD,baij->Mvctx);
1660: (*b->ops->diagonalscale)(b,PETSC_NULL,baij->lvec);
1661: }
1662:
1663: return(0);
1664: }
1668: int MatZeroRows_MPIBAIJ(Mat A,IS is,const PetscScalar *diag)
1669: {
1670: Mat_MPIBAIJ *l = (Mat_MPIBAIJ*)A->data;
1671: int i,ierr,N,*rows,*owners = l->rowners,size = l->size;
1672: int *nprocs,j,idx,nsends,row;
1673: int nmax,*svalues,*starts,*owner,nrecvs,rank = l->rank;
1674: int *rvalues,tag = A->tag,count,base,slen,n,*source;
1675: int *lens,imdex,*lrows,*values,bs=l->bs,rstart_bs=l->rstart_bs;
1676: MPI_Comm comm = A->comm;
1677: MPI_Request *send_waits,*recv_waits;
1678: MPI_Status recv_status,*send_status;
1679: IS istmp;
1680: PetscTruth found;
1681:
1683: ISGetLocalSize(is,&N);
1684: ISGetIndices(is,&rows);
1685:
1686: /* first count number of contributors to each processor */
1687: PetscMalloc(2*size*sizeof(int),&nprocs);
1688: PetscMemzero(nprocs,2*size*sizeof(int));
1689: PetscMalloc((N+1)*sizeof(int),&owner); /* see note*/
1690: for (i=0; i<N; i++) {
1691: idx = rows[i];
1692: found = PETSC_FALSE;
1693: for (j=0; j<size; j++) {
1694: if (idx >= owners[j]*bs && idx < owners[j+1]*bs) {
1695: nprocs[2*j]++; nprocs[2*j+1] = 1; owner[i] = j; found = PETSC_TRUE; break;
1696: }
1697: }
1698: if (!found) SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,"Index out of range");
1699: }
1700: nsends = 0; for (i=0; i<size; i++) { nsends += nprocs[2*i+1];}
1701:
1702: /* inform other processors of number of messages and max length*/
1703: PetscMaxSum(comm,nprocs,&nmax,&nrecvs);
1704:
1705: /* post receives: */
1706: PetscMalloc((nrecvs+1)*(nmax+1)*sizeof(int),&rvalues);
1707: PetscMalloc((nrecvs+1)*sizeof(MPI_Request),&recv_waits);
1708: for (i=0; i<nrecvs; i++) {
1709: MPI_Irecv(rvalues+nmax*i,nmax,MPI_INT,MPI_ANY_SOURCE,tag,comm,recv_waits+i);
1710: }
1711:
1712: /* do sends:
1713: 1) starts[i] gives the starting index in svalues for stuff going to
1714: the ith processor
1715: */
1716: PetscMalloc((N+1)*sizeof(int),&svalues);
1717: PetscMalloc((nsends+1)*sizeof(MPI_Request),&send_waits);
1718: PetscMalloc((size+1)*sizeof(int),&starts);
1719: starts[0] = 0;
1720: for (i=1; i<size; i++) { starts[i] = starts[i-1] + nprocs[2*i-2];}
1721: for (i=0; i<N; i++) {
1722: svalues[starts[owner[i]]++] = rows[i];
1723: }
1724: ISRestoreIndices(is,&rows);
1725:
1726: starts[0] = 0;
1727: for (i=1; i<size+1; i++) { starts[i] = starts[i-1] + nprocs[2*i-2];}
1728: count = 0;
1729: for (i=0; i<size; i++) {
1730: if (nprocs[2*i+1]) {
1731: MPI_Isend(svalues+starts[i],nprocs[2*i],MPI_INT,i,tag,comm,send_waits+count++);
1732: }
1733: }
1734: PetscFree(starts);
1736: base = owners[rank]*bs;
1737:
1738: /* wait on receives */
1739: PetscMalloc(2*(nrecvs+1)*sizeof(int),&lens);
1740: source = lens + nrecvs;
1741: count = nrecvs; slen = 0;
1742: while (count) {
1743: MPI_Waitany(nrecvs,recv_waits,&imdex,&recv_status);
1744: /* unpack receives into our local space */
1745: MPI_Get_count(&recv_status,MPI_INT,&n);
1746: source[imdex] = recv_status.MPI_SOURCE;
1747: lens[imdex] = n;
1748: slen += n;
1749: count--;
1750: }
1751: PetscFree(recv_waits);
1752:
1753: /* move the data into the send scatter */
1754: PetscMalloc((slen+1)*sizeof(int),&lrows);
1755: count = 0;
1756: for (i=0; i<nrecvs; i++) {
1757: values = rvalues + i*nmax;
1758: for (j=0; j<lens[i]; j++) {
1759: lrows[count++] = values[j] - base;
1760: }
1761: }
1762: PetscFree(rvalues);
1763: PetscFree(lens);
1764: PetscFree(owner);
1765: PetscFree(nprocs);
1766:
1767: /* actually zap the local rows */
1768: ISCreateGeneral(PETSC_COMM_SELF,slen,lrows,&istmp);
1769: PetscLogObjectParent(A,istmp);
1771: /*
1772: Zero the required rows. If the "diagonal block" of the matrix
1773: is square and the user wishes to set the diagonal we use seperate
1774: code so that MatSetValues() is not called for each diagonal allocating
1775: new memory, thus calling lots of mallocs and slowing things down.
1777: Contributed by: Mathew Knepley
1778: */
1779: /* must zero l->B before l->A because the (diag) case below may put values into l->B*/
1780: MatZeroRows_SeqBAIJ(l->B,istmp,0);
1781: if (diag && (l->A->M == l->A->N)) {
1782: MatZeroRows_SeqBAIJ(l->A,istmp,diag);
1783: } else if (diag) {
1784: MatZeroRows_SeqBAIJ(l->A,istmp,0);
1785: if (((Mat_SeqBAIJ*)l->A->data)->nonew) {
1786: SETERRQ(PETSC_ERR_SUP,"MatZeroRows() on rectangular matrices cannot be used with the Mat options \n\
1787: MAT_NO_NEW_NONZERO_LOCATIONS,MAT_NEW_NONZERO_LOCATION_ERR,MAT_NEW_NONZERO_ALLOCATION_ERR");
1788: }
1789: for (i=0; i<slen; i++) {
1790: row = lrows[i] + rstart_bs;
1791: MatSetValues(A,1,&row,1,&row,diag,INSERT_VALUES);
1792: }
1793: MatAssemblyBegin(A,MAT_FINAL_ASSEMBLY);
1794: MatAssemblyEnd(A,MAT_FINAL_ASSEMBLY);
1795: } else {
1796: MatZeroRows_SeqBAIJ(l->A,istmp,0);
1797: }
1799: ISDestroy(istmp);
1800: PetscFree(lrows);
1802: /* wait on sends */
1803: if (nsends) {
1804: PetscMalloc(nsends*sizeof(MPI_Status),&send_status);
1805: MPI_Waitall(nsends,send_waits,send_status);
1806: PetscFree(send_status);
1807: }
1808: PetscFree(send_waits);
1809: PetscFree(svalues);
1811: return(0);
1812: }
1816: int MatPrintHelp_MPIBAIJ(Mat A)
1817: {
1818: Mat_MPIBAIJ *a = (Mat_MPIBAIJ*)A->data;
1819: MPI_Comm comm = A->comm;
1820: static int called = 0;
1821: int ierr;
1824: if (!a->rank) {
1825: MatPrintHelp_SeqBAIJ(a->A);
1826: }
1827: if (called) {return(0);} else called = 1;
1828: (*PetscHelpPrintf)(comm," Options for MATMPIBAIJ matrix format (the defaults):\n");
1829: (*PetscHelpPrintf)(comm," -mat_use_hash_table <factor>: Use hashtable for efficient matrix assembly\n");
1830: return(0);
1831: }
1835: int MatSetUnfactored_MPIBAIJ(Mat A)
1836: {
1837: Mat_MPIBAIJ *a = (Mat_MPIBAIJ*)A->data;
1838: int ierr;
1841: MatSetUnfactored(a->A);
1842: return(0);
1843: }
1845: static int MatDuplicate_MPIBAIJ(Mat,MatDuplicateOption,Mat *);
1849: int MatEqual_MPIBAIJ(Mat A,Mat B,PetscTruth *flag)
1850: {
1851: Mat_MPIBAIJ *matB = (Mat_MPIBAIJ*)B->data,*matA = (Mat_MPIBAIJ*)A->data;
1852: Mat a,b,c,d;
1853: PetscTruth flg;
1854: int ierr;
1857: a = matA->A; b = matA->B;
1858: c = matB->A; d = matB->B;
1860: MatEqual(a,c,&flg);
1861: if (flg == PETSC_TRUE) {
1862: MatEqual(b,d,&flg);
1863: }
1864: MPI_Allreduce(&flg,flag,1,MPI_INT,MPI_LAND,A->comm);
1865: return(0);
1866: }
1871: int MatSetUpPreallocation_MPIBAIJ(Mat A)
1872: {
1873: int ierr;
1876: MatMPIBAIJSetPreallocation(A,1,PETSC_DEFAULT,0,PETSC_DEFAULT,0);
1877: return(0);
1878: }
1880: /* -------------------------------------------------------------------*/
1881: static struct _MatOps MatOps_Values = {
1882: MatSetValues_MPIBAIJ,
1883: MatGetRow_MPIBAIJ,
1884: MatRestoreRow_MPIBAIJ,
1885: MatMult_MPIBAIJ,
1886: /* 4*/ MatMultAdd_MPIBAIJ,
1887: MatMultTranspose_MPIBAIJ,
1888: MatMultTransposeAdd_MPIBAIJ,
1889: 0,
1890: 0,
1891: 0,
1892: /*10*/ 0,
1893: 0,
1894: 0,
1895: 0,
1896: MatTranspose_MPIBAIJ,
1897: /*15*/ MatGetInfo_MPIBAIJ,
1898: MatEqual_MPIBAIJ,
1899: MatGetDiagonal_MPIBAIJ,
1900: MatDiagonalScale_MPIBAIJ,
1901: MatNorm_MPIBAIJ,
1902: /*20*/ MatAssemblyBegin_MPIBAIJ,
1903: MatAssemblyEnd_MPIBAIJ,
1904: 0,
1905: MatSetOption_MPIBAIJ,
1906: MatZeroEntries_MPIBAIJ,
1907: /*25*/ MatZeroRows_MPIBAIJ,
1908: 0,
1909: 0,
1910: 0,
1911: 0,
1912: /*30*/ MatSetUpPreallocation_MPIBAIJ,
1913: 0,
1914: 0,
1915: 0,
1916: 0,
1917: /*35*/ MatDuplicate_MPIBAIJ,
1918: 0,
1919: 0,
1920: 0,
1921: 0,
1922: /*40*/ 0,
1923: MatGetSubMatrices_MPIBAIJ,
1924: MatIncreaseOverlap_MPIBAIJ,
1925: MatGetValues_MPIBAIJ,
1926: 0,
1927: /*45*/ MatPrintHelp_MPIBAIJ,
1928: MatScale_MPIBAIJ,
1929: 0,
1930: 0,
1931: 0,
1932: /*50*/ MatGetBlockSize_MPIBAIJ,
1933: 0,
1934: 0,
1935: 0,
1936: 0,
1937: /*55*/ 0,
1938: 0,
1939: MatSetUnfactored_MPIBAIJ,
1940: 0,
1941: MatSetValuesBlocked_MPIBAIJ,
1942: /*60*/ 0,
1943: MatDestroy_MPIBAIJ,
1944: MatView_MPIBAIJ,
1945: MatGetPetscMaps_Petsc,
1946: 0,
1947: /*65*/ 0,
1948: 0,
1949: 0,
1950: 0,
1951: 0,
1952: /*70*/ MatGetRowMax_MPIBAIJ,
1953: 0,
1954: 0,
1955: 0,
1956: 0,
1957: /*75*/ 0,
1958: 0,
1959: 0,
1960: 0,
1961: 0,
1962: /*80*/ 0,
1963: 0,
1964: 0,
1965: 0,
1966: /*85*/ MatLoad_MPIBAIJ
1967: };
1970: EXTERN_C_BEGIN
1973: int MatGetDiagonalBlock_MPIBAIJ(Mat A,PetscTruth *iscopy,MatReuse reuse,Mat *a)
1974: {
1976: *a = ((Mat_MPIBAIJ *)A->data)->A;
1977: *iscopy = PETSC_FALSE;
1978: return(0);
1979: }
1980: EXTERN_C_END
1982: EXTERN_C_BEGIN
1983: extern int MatConvert_MPIBAIJ_MPISBAIJ(Mat,const MatType,Mat*);
1984: EXTERN_C_END
1986: EXTERN_C_BEGIN
1989: int MatMPIBAIJSetPreallocation_MPIBAIJ(Mat B,int bs,int d_nz,int *d_nnz,int o_nz,int *o_nnz)
1990: {
1991: Mat_MPIBAIJ *b;
1992: int ierr,i;
1995: B->preallocated = PETSC_TRUE;
1996: PetscOptionsGetInt(PETSC_NULL,"-mat_block_size",&bs,PETSC_NULL);
1998: if (bs < 1) SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,"Invalid block size specified, must be positive");
1999: if (d_nz == PETSC_DEFAULT || d_nz == PETSC_DECIDE) d_nz = 5;
2000: if (o_nz == PETSC_DEFAULT || o_nz == PETSC_DECIDE) o_nz = 2;
2001: if (d_nz < 0) SETERRQ1(PETSC_ERR_ARG_OUTOFRANGE,"d_nz cannot be less than 0: value %d",d_nz);
2002: if (o_nz < 0) SETERRQ1(PETSC_ERR_ARG_OUTOFRANGE,"o_nz cannot be less than 0: value %d",o_nz);
2003: if (d_nnz) {
2004: for (i=0; i<B->m/bs; i++) {
2005: if (d_nnz[i] < 0) SETERRQ2(PETSC_ERR_ARG_OUTOFRANGE,"d_nnz cannot be less than -1: local row %d value %d",i,d_nnz[i]);
2006: }
2007: }
2008: if (o_nnz) {
2009: for (i=0; i<B->m/bs; i++) {
2010: if (o_nnz[i] < 0) SETERRQ2(PETSC_ERR_ARG_OUTOFRANGE,"o_nnz cannot be less than -1: local row %d value %d",i,o_nnz[i]);
2011: }
2012: }
2013:
2014: PetscSplitOwnershipBlock(B->comm,bs,&B->m,&B->M);
2015: PetscSplitOwnershipBlock(B->comm,bs,&B->n,&B->N);
2016: PetscMapCreateMPI(B->comm,B->m,B->M,&B->rmap);
2017: PetscMapCreateMPI(B->comm,B->n,B->N,&B->cmap);
2019: b = (Mat_MPIBAIJ*)B->data;
2020: b->bs = bs;
2021: b->bs2 = bs*bs;
2022: b->mbs = B->m/bs;
2023: b->nbs = B->n/bs;
2024: b->Mbs = B->M/bs;
2025: b->Nbs = B->N/bs;
2027: MPI_Allgather(&b->mbs,1,MPI_INT,b->rowners+1,1,MPI_INT,B->comm);
2028: b->rowners[0] = 0;
2029: for (i=2; i<=b->size; i++) {
2030: b->rowners[i] += b->rowners[i-1];
2031: }
2032: b->rstart = b->rowners[b->rank];
2033: b->rend = b->rowners[b->rank+1];
2035: MPI_Allgather(&b->nbs,1,MPI_INT,b->cowners+1,1,MPI_INT,B->comm);
2036: b->cowners[0] = 0;
2037: for (i=2; i<=b->size; i++) {
2038: b->cowners[i] += b->cowners[i-1];
2039: }
2040: b->cstart = b->cowners[b->rank];
2041: b->cend = b->cowners[b->rank+1];
2043: for (i=0; i<=b->size; i++) {
2044: b->rowners_bs[i] = b->rowners[i]*bs;
2045: }
2046: b->rstart_bs = b->rstart*bs;
2047: b->rend_bs = b->rend*bs;
2048: b->cstart_bs = b->cstart*bs;
2049: b->cend_bs = b->cend*bs;
2051: MatCreate(PETSC_COMM_SELF,B->m,B->n,B->m,B->n,&b->A);
2052: MatSetType(b->A,MATSEQBAIJ);
2053: MatSeqBAIJSetPreallocation(b->A,bs,d_nz,d_nnz);
2054: PetscLogObjectParent(B,b->A);
2055: MatCreate(PETSC_COMM_SELF,B->m,B->N,B->m,B->N,&b->B);
2056: MatSetType(b->B,MATSEQBAIJ);
2057: MatSeqBAIJSetPreallocation(b->B,bs,o_nz,o_nnz);
2058: PetscLogObjectParent(B,b->B);
2060: MatStashCreate_Private(B->comm,bs,&B->bstash);
2062: return(0);
2063: }
2064: EXTERN_C_END
2066: EXTERN_C_BEGIN
2067: EXTERN int MatDiagonalScaleLocal_MPIBAIJ(Mat,Vec);
2068: EXTERN int MatSetHashTableFactor_MPIBAIJ(Mat,PetscReal);
2069: EXTERN_C_END
2071: /*MC
2072: MATMPIBAIJ - MATMPIBAIJ = "mpibaij" - A matrix type to be used for distributed block sparse matrices.
2074: Options Database Keys:
2075: . -mat_type mpibaij - sets the matrix type to "mpibaij" during a call to MatSetFromOptions()
2077: Level: beginner
2079: .seealso: MatCreateMPIBAIJ
2080: M*/
2082: EXTERN_C_BEGIN
2085: int MatCreate_MPIBAIJ(Mat B)
2086: {
2087: Mat_MPIBAIJ *b;
2088: int ierr;
2089: PetscTruth flg;
2092: PetscNew(Mat_MPIBAIJ,&b);
2093: B->data = (void*)b;
2095: PetscMemzero(b,sizeof(Mat_MPIBAIJ));
2096: PetscMemcpy(B->ops,&MatOps_Values,sizeof(struct _MatOps));
2097: B->mapping = 0;
2098: B->factor = 0;
2099: B->assembled = PETSC_FALSE;
2101: B->insertmode = NOT_SET_VALUES;
2102: MPI_Comm_rank(B->comm,&b->rank);
2103: MPI_Comm_size(B->comm,&b->size);
2105: /* build local table of row and column ownerships */
2106: PetscMalloc(3*(b->size+2)*sizeof(int),&b->rowners);
2107: PetscLogObjectMemory(B,3*(b->size+2)*sizeof(int)+sizeof(struct _p_Mat)+sizeof(Mat_MPIBAIJ));
2108: b->cowners = b->rowners + b->size + 2;
2109: b->rowners_bs = b->cowners + b->size + 2;
2111: /* build cache for off array entries formed */
2112: MatStashCreate_Private(B->comm,1,&B->stash);
2113: b->donotstash = PETSC_FALSE;
2114: b->colmap = PETSC_NULL;
2115: b->garray = PETSC_NULL;
2116: b->roworiented = PETSC_TRUE;
2118: #if defined(PETSC_USE_MAT_SINGLE)
2119: /* stuff for MatSetValues_XXX in single precision */
2120: b->setvalueslen = 0;
2121: b->setvaluescopy = PETSC_NULL;
2122: #endif
2124: /* stuff used in block assembly */
2125: b->barray = 0;
2127: /* stuff used for matrix vector multiply */
2128: b->lvec = 0;
2129: b->Mvctx = 0;
2131: /* stuff for MatGetRow() */
2132: b->rowindices = 0;
2133: b->rowvalues = 0;
2134: b->getrowactive = PETSC_FALSE;
2136: /* hash table stuff */
2137: b->ht = 0;
2138: b->hd = 0;
2139: b->ht_size = 0;
2140: b->ht_flag = PETSC_FALSE;
2141: b->ht_fact = 0;
2142: b->ht_total_ct = 0;
2143: b->ht_insert_ct = 0;
2145: PetscOptionsHasName(PETSC_NULL,"-mat_use_hash_table",&flg);
2146: if (flg) {
2147: PetscReal fact = 1.39;
2148: MatSetOption(B,MAT_USE_HASH_TABLE);
2149: PetscOptionsGetReal(PETSC_NULL,"-mat_use_hash_table",&fact,PETSC_NULL);
2150: if (fact <= 1.0) fact = 1.39;
2151: MatMPIBAIJSetHashTableFactor(B,fact);
2152: PetscLogInfo(0,"MatCreateMPIBAIJ:Hash table Factor used %5.2f\n",fact);
2153: }
2154: PetscObjectComposeFunctionDynamic((PetscObject)B,"MatStoreValues_C",
2155: "MatStoreValues_MPIBAIJ",
2156: MatStoreValues_MPIBAIJ);
2157: PetscObjectComposeFunctionDynamic((PetscObject)B,"MatRetrieveValues_C",
2158: "MatRetrieveValues_MPIBAIJ",
2159: MatRetrieveValues_MPIBAIJ);
2160: PetscObjectComposeFunctionDynamic((PetscObject)B,"MatGetDiagonalBlock_C",
2161: "MatGetDiagonalBlock_MPIBAIJ",
2162: MatGetDiagonalBlock_MPIBAIJ);
2163: PetscObjectComposeFunctionDynamic((PetscObject)B,"MatMPIBAIJSetPreallocation_C",
2164: "MatMPIBAIJSetPreallocation_MPIBAIJ",
2165: MatMPIBAIJSetPreallocation_MPIBAIJ);
2166: PetscObjectComposeFunctionDynamic((PetscObject)B,"MatDiagonalScaleLocal_C",
2167: "MatDiagonalScaleLocal_MPIBAIJ",
2168: MatDiagonalScaleLocal_MPIBAIJ);
2169: PetscObjectComposeFunctionDynamic((PetscObject)B,"MatSetHashTableFactor_C",
2170: "MatSetHashTableFactor_MPIBAIJ",
2171: MatSetHashTableFactor_MPIBAIJ);
2172: return(0);
2173: }
2174: EXTERN_C_END
2176: /*MC
2177: MATBAIJ - MATBAIJ = "baij" - A matrix type to be used for block sparse matrices.
2179: This matrix type is identical to MATSEQBAIJ when constructed with a single process communicator,
2180: and MATMPIBAIJ otherwise.
2182: Options Database Keys:
2183: . -mat_type baij - sets the matrix type to "baij" during a call to MatSetFromOptions()
2185: Level: beginner
2187: .seealso: MatCreateMPIBAIJ,MATSEQBAIJ,MATMPIBAIJ
2188: M*/
2190: EXTERN_C_BEGIN
2193: int MatCreate_BAIJ(Mat A) {
2194: int ierr,size;
2197: PetscObjectChangeTypeName((PetscObject)A,MATBAIJ);
2198: MPI_Comm_size(A->comm,&size);
2199: if (size == 1) {
2200: MatSetType(A,MATSEQBAIJ);
2201: } else {
2202: MatSetType(A,MATMPIBAIJ);
2203: }
2204: return(0);
2205: }
2206: EXTERN_C_END
2210: /*@C
2211: MatMPIBAIJSetPreallocation - Creates a sparse parallel matrix in block AIJ format
2212: (block compressed row). For good matrix assembly performance
2213: the user should preallocate the matrix storage by setting the parameters
2214: d_nz (or d_nnz) and o_nz (or o_nnz). By setting these parameters accurately,
2215: performance can be increased by more than a factor of 50.
2217: Collective on Mat
2219: Input Parameters:
2220: + A - the matrix
2221: . bs - size of blockk
2222: . d_nz - number of block nonzeros per block row in diagonal portion of local
2223: submatrix (same for all local rows)
2224: . d_nnz - array containing the number of block nonzeros in the various block rows
2225: of the in diagonal portion of the local (possibly different for each block
2226: row) or PETSC_NULL. You must leave room for the diagonal entry even if it is zero.
2227: . o_nz - number of block nonzeros per block row in the off-diagonal portion of local
2228: submatrix (same for all local rows).
2229: - o_nnz - array containing the number of nonzeros in the various block rows of the
2230: off-diagonal portion of the local submatrix (possibly different for
2231: each block row) or PETSC_NULL.
2233: Output Parameter:
2236: Options Database Keys:
2237: . -mat_no_unroll - uses code that does not unroll the loops in the
2238: block calculations (much slower)
2239: . -mat_block_size - size of the blocks to use
2241: Notes:
2242: If PETSC_DECIDE or PETSC_DETERMINE is used for a particular argument on one processor
2243: than it must be used on all processors that share the object for that argument.
2245: Storage Information:
2246: For a square global matrix we define each processor's diagonal portion
2247: to be its local rows and the corresponding columns (a square submatrix);
2248: each processor's off-diagonal portion encompasses the remainder of the
2249: local matrix (a rectangular submatrix).
2251: The user can specify preallocated storage for the diagonal part of
2252: the local submatrix with either d_nz or d_nnz (not both). Set
2253: d_nz=PETSC_DEFAULT and d_nnz=PETSC_NULL for PETSc to control dynamic
2254: memory allocation. Likewise, specify preallocated storage for the
2255: off-diagonal part of the local submatrix with o_nz or o_nnz (not both).
2257: Consider a processor that owns rows 3, 4 and 5 of a parallel matrix. In
2258: the figure below we depict these three local rows and all columns (0-11).
2260: .vb
2261: 0 1 2 3 4 5 6 7 8 9 10 11
2262: -------------------
2263: row 3 | o o o d d d o o o o o o
2264: row 4 | o o o d d d o o o o o o
2265: row 5 | o o o d d d o o o o o o
2266: -------------------
2267: .ve
2268:
2269: Thus, any entries in the d locations are stored in the d (diagonal)
2270: submatrix, and any entries in the o locations are stored in the
2271: o (off-diagonal) submatrix. Note that the d and the o submatrices are
2272: stored simply in the MATSEQBAIJ format for compressed row storage.
2274: Now d_nz should indicate the number of block nonzeros per row in the d matrix,
2275: and o_nz should indicate the number of block nonzeros per row in the o matrix.
2276: In general, for PDE problems in which most nonzeros are near the diagonal,
2277: one expects d_nz >> o_nz. For large problems you MUST preallocate memory
2278: or you will get TERRIBLE performance; see the users' manual chapter on
2279: matrices.
2281: Level: intermediate
2283: .keywords: matrix, block, aij, compressed row, sparse, parallel
2285: .seealso: MatCreate(), MatCreateSeqBAIJ(), MatSetValues(), MatCreateMPIBAIJ()
2286: @*/
2287: int MatMPIBAIJSetPreallocation(Mat B,int bs,int d_nz,const int d_nnz[],int o_nz,const int o_nnz[])
2288: {
2289: int ierr,(*f)(Mat,int,int,const int[],int,const int[]);
2292: PetscObjectQueryFunction((PetscObject)B,"MatMPIBAIJSetPreallocation_C",(void (**)(void))&f);
2293: if (f) {
2294: (*f)(B,bs,d_nz,d_nnz,o_nz,o_nnz);
2295: }
2296: return(0);
2297: }
2301: /*@C
2302: MatCreateMPIBAIJ - Creates a sparse parallel matrix in block AIJ format
2303: (block compressed row). For good matrix assembly performance
2304: the user should preallocate the matrix storage by setting the parameters
2305: d_nz (or d_nnz) and o_nz (or o_nnz). By setting these parameters accurately,
2306: performance can be increased by more than a factor of 50.
2308: Collective on MPI_Comm
2310: Input Parameters:
2311: + comm - MPI communicator
2312: . bs - size of blockk
2313: . m - number of local rows (or PETSC_DECIDE to have calculated if M is given)
2314: This value should be the same as the local size used in creating the
2315: y vector for the matrix-vector product y = Ax.
2316: . n - number of local columns (or PETSC_DECIDE to have calculated if N is given)
2317: This value should be the same as the local size used in creating the
2318: x vector for the matrix-vector product y = Ax.
2319: . M - number of global rows (or PETSC_DETERMINE to have calculated if m is given)
2320: . N - number of global columns (or PETSC_DETERMINE to have calculated if n is given)
2321: . d_nz - number of nonzero blocks per block row in diagonal portion of local
2322: submatrix (same for all local rows)
2323: . d_nnz - array containing the number of nonzero blocks in the various block rows
2324: of the in diagonal portion of the local (possibly different for each block
2325: row) or PETSC_NULL. You must leave room for the diagonal entry even if it is zero.
2326: . o_nz - number of nonzero blocks per block row in the off-diagonal portion of local
2327: submatrix (same for all local rows).
2328: - o_nnz - array containing the number of nonzero blocks in the various block rows of the
2329: off-diagonal portion of the local submatrix (possibly different for
2330: each block row) or PETSC_NULL.
2332: Output Parameter:
2333: . A - the matrix
2335: Options Database Keys:
2336: . -mat_no_unroll - uses code that does not unroll the loops in the
2337: block calculations (much slower)
2338: . -mat_block_size - size of the blocks to use
2340: Notes:
2341: A nonzero block is any block that as 1 or more nonzeros in it
2343: The user MUST specify either the local or global matrix dimensions
2344: (possibly both).
2346: If PETSC_DECIDE or PETSC_DETERMINE is used for a particular argument on one processor
2347: than it must be used on all processors that share the object for that argument.
2349: Storage Information:
2350: For a square global matrix we define each processor's diagonal portion
2351: to be its local rows and the corresponding columns (a square submatrix);
2352: each processor's off-diagonal portion encompasses the remainder of the
2353: local matrix (a rectangular submatrix).
2355: The user can specify preallocated storage for the diagonal part of
2356: the local submatrix with either d_nz or d_nnz (not both). Set
2357: d_nz=PETSC_DEFAULT and d_nnz=PETSC_NULL for PETSc to control dynamic
2358: memory allocation. Likewise, specify preallocated storage for the
2359: off-diagonal part of the local submatrix with o_nz or o_nnz (not both).
2361: Consider a processor that owns rows 3, 4 and 5 of a parallel matrix. In
2362: the figure below we depict these three local rows and all columns (0-11).
2364: .vb
2365: 0 1 2 3 4 5 6 7 8 9 10 11
2366: -------------------
2367: row 3 | o o o d d d o o o o o o
2368: row 4 | o o o d d d o o o o o o
2369: row 5 | o o o d d d o o o o o o
2370: -------------------
2371: .ve
2372:
2373: Thus, any entries in the d locations are stored in the d (diagonal)
2374: submatrix, and any entries in the o locations are stored in the
2375: o (off-diagonal) submatrix. Note that the d and the o submatrices are
2376: stored simply in the MATSEQBAIJ format for compressed row storage.
2378: Now d_nz should indicate the number of block nonzeros per row in the d matrix,
2379: and o_nz should indicate the number of block nonzeros per row in the o matrix.
2380: In general, for PDE problems in which most nonzeros are near the diagonal,
2381: one expects d_nz >> o_nz. For large problems you MUST preallocate memory
2382: or you will get TERRIBLE performance; see the users' manual chapter on
2383: matrices.
2385: Level: intermediate
2387: .keywords: matrix, block, aij, compressed row, sparse, parallel
2389: .seealso: MatCreate(), MatCreateSeqBAIJ(), MatSetValues(), MatCreateMPIBAIJ()
2390: @*/
2391: int MatCreateMPIBAIJ(MPI_Comm comm,int bs,int m,int n,int M,int N,int d_nz,const int d_nnz[],int o_nz,const int o_nnz[],Mat *A)
2392: {
2393: int ierr,size;
2396: MatCreate(comm,m,n,M,N,A);
2397: MPI_Comm_size(comm,&size);
2398: if (size > 1) {
2399: MatSetType(*A,MATMPIBAIJ);
2400: MatMPIBAIJSetPreallocation(*A,bs,d_nz,d_nnz,o_nz,o_nnz);
2401: } else {
2402: MatSetType(*A,MATSEQBAIJ);
2403: MatSeqBAIJSetPreallocation(*A,bs,d_nz,d_nnz);
2404: }
2405: return(0);
2406: }
2410: static int MatDuplicate_MPIBAIJ(Mat matin,MatDuplicateOption cpvalues,Mat *newmat)
2411: {
2412: Mat mat;
2413: Mat_MPIBAIJ *a,*oldmat = (Mat_MPIBAIJ*)matin->data;
2414: int ierr,len=0;
2417: *newmat = 0;
2418: MatCreate(matin->comm,matin->m,matin->n,matin->M,matin->N,&mat);
2419: MatSetType(mat,matin->type_name);
2421: PetscMemcpy(mat->ops,&MatOps_Values,sizeof(struct _MatOps));
2422: mat->factor = matin->factor;
2423: mat->preallocated = PETSC_TRUE;
2424: mat->assembled = PETSC_TRUE;
2425: mat->insertmode = NOT_SET_VALUES;
2427: a = (Mat_MPIBAIJ*)mat->data;
2428: a->bs = oldmat->bs;
2429: a->bs2 = oldmat->bs2;
2430: a->mbs = oldmat->mbs;
2431: a->nbs = oldmat->nbs;
2432: a->Mbs = oldmat->Mbs;
2433: a->Nbs = oldmat->Nbs;
2434:
2435: a->rstart = oldmat->rstart;
2436: a->rend = oldmat->rend;
2437: a->cstart = oldmat->cstart;
2438: a->cend = oldmat->cend;
2439: a->size = oldmat->size;
2440: a->rank = oldmat->rank;
2441: a->donotstash = oldmat->donotstash;
2442: a->roworiented = oldmat->roworiented;
2443: a->rowindices = 0;
2444: a->rowvalues = 0;
2445: a->getrowactive = PETSC_FALSE;
2446: a->barray = 0;
2447: a->rstart_bs = oldmat->rstart_bs;
2448: a->rend_bs = oldmat->rend_bs;
2449: a->cstart_bs = oldmat->cstart_bs;
2450: a->cend_bs = oldmat->cend_bs;
2452: /* hash table stuff */
2453: a->ht = 0;
2454: a->hd = 0;
2455: a->ht_size = 0;
2456: a->ht_flag = oldmat->ht_flag;
2457: a->ht_fact = oldmat->ht_fact;
2458: a->ht_total_ct = 0;
2459: a->ht_insert_ct = 0;
2461: PetscMemcpy(a->rowners,oldmat->rowners,3*(a->size+2)*sizeof(int));
2462: MatStashCreate_Private(matin->comm,1,&mat->stash);
2463: MatStashCreate_Private(matin->comm,oldmat->bs,&mat->bstash);
2464: if (oldmat->colmap) {
2465: #if defined (PETSC_USE_CTABLE)
2466: PetscTableCreateCopy(oldmat->colmap,&a->colmap);
2467: #else
2468: PetscMalloc((a->Nbs)*sizeof(int),&a->colmap);
2469: PetscLogObjectMemory(mat,(a->Nbs)*sizeof(int));
2470: PetscMemcpy(a->colmap,oldmat->colmap,(a->Nbs)*sizeof(int));
2471: #endif
2472: } else a->colmap = 0;
2474: if (oldmat->garray && (len = ((Mat_SeqBAIJ*)(oldmat->B->data))->nbs)) {
2475: PetscMalloc(len*sizeof(int),&a->garray);
2476: PetscLogObjectMemory(mat,len*sizeof(int));
2477: PetscMemcpy(a->garray,oldmat->garray,len*sizeof(int));
2478: } else a->garray = 0;
2479:
2480: VecDuplicate(oldmat->lvec,&a->lvec);
2481: PetscLogObjectParent(mat,a->lvec);
2482: VecScatterCopy(oldmat->Mvctx,&a->Mvctx);
2483: PetscLogObjectParent(mat,a->Mvctx);
2485: MatDuplicate(oldmat->A,cpvalues,&a->A);
2486: PetscLogObjectParent(mat,a->A);
2487: MatDuplicate(oldmat->B,cpvalues,&a->B);
2488: PetscLogObjectParent(mat,a->B);
2489: PetscFListDuplicate(matin->qlist,&mat->qlist);
2490: *newmat = mat;
2492: return(0);
2493: }
2495: #include petscsys.h
2499: int MatLoad_MPIBAIJ(PetscViewer viewer,const MatType type,Mat *newmat)
2500: {
2501: Mat A;
2502: int i,nz,ierr,j,rstart,rend,fd;
2503: PetscScalar *vals,*buf;
2504: MPI_Comm comm = ((PetscObject)viewer)->comm;
2505: MPI_Status status;
2506: int header[4],rank,size,*rowlengths = 0,M,N,m,*rowners,*browners,maxnz,*cols;
2507: int *locrowlens,*sndcounts = 0,*procsnz = 0,jj,*mycols,*ibuf;
2508: int tag = ((PetscObject)viewer)->tag,bs=1,Mbs,mbs,extra_rows;
2509: int *dlens,*odlens,*mask,*masked1,*masked2,rowcount,odcount;
2510: int dcount,kmax,k,nzcount,tmp;
2511:
2513: PetscOptionsGetInt(PETSC_NULL,"-matload_block_size",&bs,PETSC_NULL);
2515: MPI_Comm_size(comm,&size);
2516: MPI_Comm_rank(comm,&rank);
2517: if (!rank) {
2518: PetscViewerBinaryGetDescriptor(viewer,&fd);
2519: PetscBinaryRead(fd,(char *)header,4,PETSC_INT);
2520: if (header[0] != MAT_FILE_COOKIE) SETERRQ(PETSC_ERR_FILE_UNEXPECTED,"not matrix object");
2521: if (header[3] < 0) {
2522: SETERRQ(PETSC_ERR_FILE_UNEXPECTED,"Matrix stored in special format, cannot load as MPIBAIJ");
2523: }
2524: }
2526: MPI_Bcast(header+1,3,MPI_INT,0,comm);
2527: M = header[1]; N = header[2];
2529: if (M != N) SETERRQ(PETSC_ERR_SUP,"Can only do square matrices");
2531: /*
2532: This code adds extra rows to make sure the number of rows is
2533: divisible by the blocksize
2534: */
2535: Mbs = M/bs;
2536: extra_rows = bs - M + bs*(Mbs);
2537: if (extra_rows == bs) extra_rows = 0;
2538: else Mbs++;
2539: if (extra_rows &&!rank) {
2540: PetscLogInfo(0,"MatLoad_MPIBAIJ:Padding loaded matrix to match blocksize\n");
2541: }
2543: /* determine ownership of all rows */
2544: mbs = Mbs/size + ((Mbs % size) > rank);
2545: m = mbs*bs;
2546: PetscMalloc(2*(size+2)*sizeof(int),&rowners);
2547: browners = rowners + size + 1;
2548: MPI_Allgather(&mbs,1,MPI_INT,rowners+1,1,MPI_INT,comm);
2549: rowners[0] = 0;
2550: for (i=2; i<=size; i++) rowners[i] += rowners[i-1];
2551: for (i=0; i<=size; i++) browners[i] = rowners[i]*bs;
2552: rstart = rowners[rank];
2553: rend = rowners[rank+1];
2555: /* distribute row lengths to all processors */
2556: PetscMalloc((rend-rstart)*bs*sizeof(int),&locrowlens);
2557: if (!rank) {
2558: PetscMalloc((M+extra_rows)*sizeof(int),&rowlengths);
2559: PetscBinaryRead(fd,rowlengths,M,PETSC_INT);
2560: for (i=0; i<extra_rows; i++) rowlengths[M+i] = 1;
2561: PetscMalloc(size*sizeof(int),&sndcounts);
2562: for (i=0; i<size; i++) sndcounts[i] = browners[i+1] - browners[i];
2563: MPI_Scatterv(rowlengths,sndcounts,browners,MPI_INT,locrowlens,(rend-rstart)*bs,MPI_INT,0,comm);
2564: PetscFree(sndcounts);
2565: } else {
2566: MPI_Scatterv(0,0,0,MPI_INT,locrowlens,(rend-rstart)*bs,MPI_INT,0,comm);
2567: }
2569: if (!rank) {
2570: /* calculate the number of nonzeros on each processor */
2571: PetscMalloc(size*sizeof(int),&procsnz);
2572: PetscMemzero(procsnz,size*sizeof(int));
2573: for (i=0; i<size; i++) {
2574: for (j=rowners[i]*bs; j< rowners[i+1]*bs; j++) {
2575: procsnz[i] += rowlengths[j];
2576: }
2577: }
2578: PetscFree(rowlengths);
2579:
2580: /* determine max buffer needed and allocate it */
2581: maxnz = 0;
2582: for (i=0; i<size; i++) {
2583: maxnz = PetscMax(maxnz,procsnz[i]);
2584: }
2585: PetscMalloc(maxnz*sizeof(int),&cols);
2587: /* read in my part of the matrix column indices */
2588: nz = procsnz[0];
2589: PetscMalloc(nz*sizeof(int),&ibuf);
2590: mycols = ibuf;
2591: if (size == 1) nz -= extra_rows;
2592: PetscBinaryRead(fd,mycols,nz,PETSC_INT);
2593: if (size == 1) for (i=0; i< extra_rows; i++) { mycols[nz+i] = M+i; }
2595: /* read in every ones (except the last) and ship off */
2596: for (i=1; i<size-1; i++) {
2597: nz = procsnz[i];
2598: PetscBinaryRead(fd,cols,nz,PETSC_INT);
2599: MPI_Send(cols,nz,MPI_INT,i,tag,comm);
2600: }
2601: /* read in the stuff for the last proc */
2602: if (size != 1) {
2603: nz = procsnz[size-1] - extra_rows; /* the extra rows are not on the disk */
2604: PetscBinaryRead(fd,cols,nz,PETSC_INT);
2605: for (i=0; i<extra_rows; i++) cols[nz+i] = M+i;
2606: MPI_Send(cols,nz+extra_rows,MPI_INT,size-1,tag,comm);
2607: }
2608: PetscFree(cols);
2609: } else {
2610: /* determine buffer space needed for message */
2611: nz = 0;
2612: for (i=0; i<m; i++) {
2613: nz += locrowlens[i];
2614: }
2615: PetscMalloc(nz*sizeof(int),&ibuf);
2616: mycols = ibuf;
2617: /* receive message of column indices*/
2618: MPI_Recv(mycols,nz,MPI_INT,0,tag,comm,&status);
2619: MPI_Get_count(&status,MPI_INT,&maxnz);
2620: if (maxnz != nz) SETERRQ(PETSC_ERR_FILE_UNEXPECTED,"something is wrong with file");
2621: }
2622:
2623: /* loop over local rows, determining number of off diagonal entries */
2624: PetscMalloc(2*(rend-rstart+1)*sizeof(int),&dlens);
2625: odlens = dlens + (rend-rstart);
2626: PetscMalloc(3*Mbs*sizeof(int),&mask);
2627: PetscMemzero(mask,3*Mbs*sizeof(int));
2628: masked1 = mask + Mbs;
2629: masked2 = masked1 + Mbs;
2630: rowcount = 0; nzcount = 0;
2631: for (i=0; i<mbs; i++) {
2632: dcount = 0;
2633: odcount = 0;
2634: for (j=0; j<bs; j++) {
2635: kmax = locrowlens[rowcount];
2636: for (k=0; k<kmax; k++) {
2637: tmp = mycols[nzcount++]/bs;
2638: if (!mask[tmp]) {
2639: mask[tmp] = 1;
2640: if (tmp < rstart || tmp >= rend) masked2[odcount++] = tmp;
2641: else masked1[dcount++] = tmp;
2642: }
2643: }
2644: rowcount++;
2645: }
2646:
2647: dlens[i] = dcount;
2648: odlens[i] = odcount;
2650: /* zero out the mask elements we set */
2651: for (j=0; j<dcount; j++) mask[masked1[j]] = 0;
2652: for (j=0; j<odcount; j++) mask[masked2[j]] = 0;
2653: }
2655: /* create our matrix */
2656: MatCreate(comm,m,m,M+extra_rows,N+extra_rows,&A);
2657: MatSetType(A,type);CHKERRQ(ierr)
2658: MatMPIBAIJSetPreallocation(A,bs,0,dlens,0,odlens);
2660: /* Why doesn't this called using MatSetOption(A,MAT_COLUMNS_SORTED); */
2661: MatSetOption(A,MAT_COLUMNS_SORTED);
2662:
2663: if (!rank) {
2664: PetscMalloc(maxnz*sizeof(PetscScalar),&buf);
2665: /* read in my part of the matrix numerical values */
2666: nz = procsnz[0];
2667: vals = buf;
2668: mycols = ibuf;
2669: if (size == 1) nz -= extra_rows;
2670: PetscBinaryRead(fd,vals,nz,PETSC_SCALAR);
2671: if (size == 1) for (i=0; i< extra_rows; i++) { vals[nz+i] = 1.0; }
2673: /* insert into matrix */
2674: jj = rstart*bs;
2675: for (i=0; i<m; i++) {
2676: MatSetValues(A,1,&jj,locrowlens[i],mycols,vals,INSERT_VALUES);
2677: mycols += locrowlens[i];
2678: vals += locrowlens[i];
2679: jj++;
2680: }
2681: /* read in other processors (except the last one) and ship out */
2682: for (i=1; i<size-1; i++) {
2683: nz = procsnz[i];
2684: vals = buf;
2685: PetscBinaryRead(fd,vals,nz,PETSC_SCALAR);
2686: MPI_Send(vals,nz,MPIU_SCALAR,i,A->tag,comm);
2687: }
2688: /* the last proc */
2689: if (size != 1){
2690: nz = procsnz[i] - extra_rows;
2691: vals = buf;
2692: PetscBinaryRead(fd,vals,nz,PETSC_SCALAR);
2693: for (i=0; i<extra_rows; i++) vals[nz+i] = 1.0;
2694: MPI_Send(vals,nz+extra_rows,MPIU_SCALAR,size-1,A->tag,comm);
2695: }
2696: PetscFree(procsnz);
2697: } else {
2698: /* receive numeric values */
2699: PetscMalloc(nz*sizeof(PetscScalar),&buf);
2701: /* receive message of values*/
2702: vals = buf;
2703: mycols = ibuf;
2704: MPI_Recv(vals,nz,MPIU_SCALAR,0,A->tag,comm,&status);
2705: MPI_Get_count(&status,MPIU_SCALAR,&maxnz);
2706: if (maxnz != nz) SETERRQ(PETSC_ERR_FILE_UNEXPECTED,"something is wrong with file");
2708: /* insert into matrix */
2709: jj = rstart*bs;
2710: for (i=0; i<m; i++) {
2711: MatSetValues(A,1,&jj,locrowlens[i],mycols,vals,INSERT_VALUES);
2712: mycols += locrowlens[i];
2713: vals += locrowlens[i];
2714: jj++;
2715: }
2716: }
2717: PetscFree(locrowlens);
2718: PetscFree(buf);
2719: PetscFree(ibuf);
2720: PetscFree(rowners);
2721: PetscFree(dlens);
2722: PetscFree(mask);
2723: MatAssemblyBegin(A,MAT_FINAL_ASSEMBLY);
2724: MatAssemblyEnd(A,MAT_FINAL_ASSEMBLY);
2726: *newmat = A;
2727: return(0);
2728: }
2732: /*@
2733: MatMPIBAIJSetHashTableFactor - Sets the factor required to compute the size of the HashTable.
2735: Input Parameters:
2736: . mat - the matrix
2737: . fact - factor
2739: Collective on Mat
2741: Level: advanced
2743: Notes:
2744: This can also be set by the command line option: -mat_use_hash_table fact
2746: .keywords: matrix, hashtable, factor, HT
2748: .seealso: MatSetOption()
2749: @*/
2750: int MatMPIBAIJSetHashTableFactor(Mat mat,PetscReal fact)
2751: {
2752: int ierr,(*f)(Mat,PetscReal);
2755: PetscObjectQueryFunction((PetscObject)mat,"MatSetHashTableFactor_C",(void (**)(void))&f);
2756: if (f) {
2757: (*f)(mat,fact);
2758: }
2759: return(0);
2760: }
2764: int MatSetHashTableFactor_MPIBAIJ(Mat mat,PetscReal fact)
2765: {
2766: Mat_MPIBAIJ *baij;
2769: baij = (Mat_MPIBAIJ*)mat->data;
2770: baij->ht_fact = fact;
2771: return(0);
2772: }
2776: int MatMPIBAIJGetSeqBAIJ(Mat A,Mat *Ad,Mat *Ao,int *colmap[])
2777: {
2778: Mat_MPIBAIJ *a = (Mat_MPIBAIJ *)A->data;
2780: *Ad = a->A;
2781: *Ao = a->B;
2782: *colmap = a->garray;
2783: return(0);
2784: }