Actual source code: aij.c
petsc-master 2020-04-04
2: /*
3: Defines the basic matrix operations for the AIJ (compressed row)
4: matrix storage format.
5: */
8: #include <../src/mat/impls/aij/seq/aij.h>
9: #include <petscblaslapack.h>
10: #include <petscbt.h>
11: #include <petsc/private/kernels/blocktranspose.h>
13: PetscErrorCode MatSeqAIJSetTypeFromOptions(Mat A)
14: {
15: PetscErrorCode ierr;
16: PetscBool flg;
17: char type[256];
20: PetscObjectOptionsBegin((PetscObject)A);
21: PetscOptionsFList("-mat_seqaij_type","Matrix SeqAIJ type","MatSeqAIJSetType",MatSeqAIJList,"seqaij",type,256,&flg);
22: if (flg) {
23: MatSeqAIJSetType(A,type);
24: }
25: PetscOptionsEnd();
26: return(0);
27: }
29: PetscErrorCode MatGetColumnNorms_SeqAIJ(Mat A,NormType type,PetscReal *norms)
30: {
32: PetscInt i,m,n;
33: Mat_SeqAIJ *aij = (Mat_SeqAIJ*)A->data;
36: MatGetSize(A,&m,&n);
37: PetscArrayzero(norms,n);
38: if (type == NORM_2) {
39: for (i=0; i<aij->i[m]; i++) {
40: norms[aij->j[i]] += PetscAbsScalar(aij->a[i]*aij->a[i]);
41: }
42: } else if (type == NORM_1) {
43: for (i=0; i<aij->i[m]; i++) {
44: norms[aij->j[i]] += PetscAbsScalar(aij->a[i]);
45: }
46: } else if (type == NORM_INFINITY) {
47: for (i=0; i<aij->i[m]; i++) {
48: norms[aij->j[i]] = PetscMax(PetscAbsScalar(aij->a[i]),norms[aij->j[i]]);
49: }
50: } else SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"Unknown NormType");
52: if (type == NORM_2) {
53: for (i=0; i<n; i++) norms[i] = PetscSqrtReal(norms[i]);
54: }
55: return(0);
56: }
58: PetscErrorCode MatFindOffBlockDiagonalEntries_SeqAIJ(Mat A,IS *is)
59: {
60: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
61: PetscInt i,m=A->rmap->n,cnt = 0, bs = A->rmap->bs;
62: const PetscInt *jj = a->j,*ii = a->i;
63: PetscInt *rows;
64: PetscErrorCode ierr;
67: for (i=0; i<m; i++) {
68: if ((ii[i] != ii[i+1]) && ((jj[ii[i]] < bs*(i/bs)) || (jj[ii[i+1]-1] > bs*((i+bs)/bs)-1))) {
69: cnt++;
70: }
71: }
72: PetscMalloc1(cnt,&rows);
73: cnt = 0;
74: for (i=0; i<m; i++) {
75: if ((ii[i] != ii[i+1]) && ((jj[ii[i]] < bs*(i/bs)) || (jj[ii[i+1]-1] > bs*((i+bs)/bs)-1))) {
76: rows[cnt] = i;
77: cnt++;
78: }
79: }
80: ISCreateGeneral(PETSC_COMM_SELF,cnt,rows,PETSC_OWN_POINTER,is);
81: return(0);
82: }
84: PetscErrorCode MatFindZeroDiagonals_SeqAIJ_Private(Mat A,PetscInt *nrows,PetscInt **zrows)
85: {
86: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
87: const MatScalar *aa = a->a;
88: PetscInt i,m=A->rmap->n,cnt = 0;
89: const PetscInt *ii = a->i,*jj = a->j,*diag;
90: PetscInt *rows;
91: PetscErrorCode ierr;
94: MatMarkDiagonal_SeqAIJ(A);
95: diag = a->diag;
96: for (i=0; i<m; i++) {
97: if ((diag[i] >= ii[i+1]) || (jj[diag[i]] != i) || (aa[diag[i]] == 0.0)) {
98: cnt++;
99: }
100: }
101: PetscMalloc1(cnt,&rows);
102: cnt = 0;
103: for (i=0; i<m; i++) {
104: if ((diag[i] >= ii[i+1]) || (jj[diag[i]] != i) || (aa[diag[i]] == 0.0)) {
105: rows[cnt++] = i;
106: }
107: }
108: *nrows = cnt;
109: *zrows = rows;
110: return(0);
111: }
113: PetscErrorCode MatFindZeroDiagonals_SeqAIJ(Mat A,IS *zrows)
114: {
115: PetscInt nrows,*rows;
119: *zrows = NULL;
120: MatFindZeroDiagonals_SeqAIJ_Private(A,&nrows,&rows);
121: ISCreateGeneral(PetscObjectComm((PetscObject)A),nrows,rows,PETSC_OWN_POINTER,zrows);
122: return(0);
123: }
125: PetscErrorCode MatFindNonzeroRows_SeqAIJ(Mat A,IS *keptrows)
126: {
127: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
128: const MatScalar *aa;
129: PetscInt m=A->rmap->n,cnt = 0;
130: const PetscInt *ii;
131: PetscInt n,i,j,*rows;
132: PetscErrorCode ierr;
135: *keptrows = 0;
136: ii = a->i;
137: for (i=0; i<m; i++) {
138: n = ii[i+1] - ii[i];
139: if (!n) {
140: cnt++;
141: goto ok1;
142: }
143: aa = a->a + ii[i];
144: for (j=0; j<n; j++) {
145: if (aa[j] != 0.0) goto ok1;
146: }
147: cnt++;
148: ok1:;
149: }
150: if (!cnt) return(0);
151: PetscMalloc1(A->rmap->n-cnt,&rows);
152: cnt = 0;
153: for (i=0; i<m; i++) {
154: n = ii[i+1] - ii[i];
155: if (!n) continue;
156: aa = a->a + ii[i];
157: for (j=0; j<n; j++) {
158: if (aa[j] != 0.0) {
159: rows[cnt++] = i;
160: break;
161: }
162: }
163: }
164: ISCreateGeneral(PETSC_COMM_SELF,cnt,rows,PETSC_OWN_POINTER,keptrows);
165: return(0);
166: }
168: PetscErrorCode MatDiagonalSet_SeqAIJ(Mat Y,Vec D,InsertMode is)
169: {
170: PetscErrorCode ierr;
171: Mat_SeqAIJ *aij = (Mat_SeqAIJ*) Y->data;
172: PetscInt i,m = Y->rmap->n;
173: const PetscInt *diag;
174: MatScalar *aa = aij->a;
175: const PetscScalar *v;
176: PetscBool missing;
177: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
178: PetscBool inserted = PETSC_FALSE;
179: #endif
182: if (Y->assembled) {
183: MatMissingDiagonal_SeqAIJ(Y,&missing,NULL);
184: if (!missing) {
185: diag = aij->diag;
186: VecGetArrayRead(D,&v);
187: if (is == INSERT_VALUES) {
188: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
189: inserted = PETSC_TRUE;
190: #endif
191: for (i=0; i<m; i++) {
192: aa[diag[i]] = v[i];
193: }
194: } else {
195: for (i=0; i<m; i++) {
196: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
197: if (v[i] != 0.0) inserted = PETSC_TRUE;
198: #endif
199: aa[diag[i]] += v[i];
200: }
201: }
202: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
203: if (inserted) Y->offloadmask = PETSC_OFFLOAD_CPU;
204: #endif
205: VecRestoreArrayRead(D,&v);
206: return(0);
207: }
208: MatSeqAIJInvalidateDiagonal(Y);
209: }
210: MatDiagonalSet_Default(Y,D,is);
211: return(0);
212: }
214: PetscErrorCode MatGetRowIJ_SeqAIJ(Mat A,PetscInt oshift,PetscBool symmetric,PetscBool inodecompressed,PetscInt *m,const PetscInt *ia[],const PetscInt *ja[],PetscBool *done)
215: {
216: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
218: PetscInt i,ishift;
221: *m = A->rmap->n;
222: if (!ia) return(0);
223: ishift = 0;
224: if (symmetric && !A->structurally_symmetric) {
225: MatToSymmetricIJ_SeqAIJ(A->rmap->n,a->i,a->j,PETSC_TRUE,ishift,oshift,(PetscInt**)ia,(PetscInt**)ja);
226: } else if (oshift == 1) {
227: PetscInt *tia;
228: PetscInt nz = a->i[A->rmap->n];
229: /* malloc space and add 1 to i and j indices */
230: PetscMalloc1(A->rmap->n+1,&tia);
231: for (i=0; i<A->rmap->n+1; i++) tia[i] = a->i[i] + 1;
232: *ia = tia;
233: if (ja) {
234: PetscInt *tja;
235: PetscMalloc1(nz+1,&tja);
236: for (i=0; i<nz; i++) tja[i] = a->j[i] + 1;
237: *ja = tja;
238: }
239: } else {
240: *ia = a->i;
241: if (ja) *ja = a->j;
242: }
243: return(0);
244: }
246: PetscErrorCode MatRestoreRowIJ_SeqAIJ(Mat A,PetscInt oshift,PetscBool symmetric,PetscBool inodecompressed,PetscInt *n,const PetscInt *ia[],const PetscInt *ja[],PetscBool *done)
247: {
251: if (!ia) return(0);
252: if ((symmetric && !A->structurally_symmetric) || oshift == 1) {
253: PetscFree(*ia);
254: if (ja) {PetscFree(*ja);}
255: }
256: return(0);
257: }
259: PetscErrorCode MatGetColumnIJ_SeqAIJ(Mat A,PetscInt oshift,PetscBool symmetric,PetscBool inodecompressed,PetscInt *nn,const PetscInt *ia[],const PetscInt *ja[],PetscBool *done)
260: {
261: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
263: PetscInt i,*collengths,*cia,*cja,n = A->cmap->n,m = A->rmap->n;
264: PetscInt nz = a->i[m],row,*jj,mr,col;
267: *nn = n;
268: if (!ia) return(0);
269: if (symmetric) {
270: MatToSymmetricIJ_SeqAIJ(A->rmap->n,a->i,a->j,PETSC_TRUE,0,oshift,(PetscInt**)ia,(PetscInt**)ja);
271: } else {
272: PetscCalloc1(n,&collengths);
273: PetscMalloc1(n+1,&cia);
274: PetscMalloc1(nz,&cja);
275: jj = a->j;
276: for (i=0; i<nz; i++) {
277: collengths[jj[i]]++;
278: }
279: cia[0] = oshift;
280: for (i=0; i<n; i++) {
281: cia[i+1] = cia[i] + collengths[i];
282: }
283: PetscArrayzero(collengths,n);
284: jj = a->j;
285: for (row=0; row<m; row++) {
286: mr = a->i[row+1] - a->i[row];
287: for (i=0; i<mr; i++) {
288: col = *jj++;
290: cja[cia[col] + collengths[col]++ - oshift] = row + oshift;
291: }
292: }
293: PetscFree(collengths);
294: *ia = cia; *ja = cja;
295: }
296: return(0);
297: }
299: PetscErrorCode MatRestoreColumnIJ_SeqAIJ(Mat A,PetscInt oshift,PetscBool symmetric,PetscBool inodecompressed,PetscInt *n,const PetscInt *ia[],const PetscInt *ja[],PetscBool *done)
300: {
304: if (!ia) return(0);
306: PetscFree(*ia);
307: PetscFree(*ja);
308: return(0);
309: }
311: /*
312: MatGetColumnIJ_SeqAIJ_Color() and MatRestoreColumnIJ_SeqAIJ_Color() are customized from
313: MatGetColumnIJ_SeqAIJ() and MatRestoreColumnIJ_SeqAIJ() by adding an output
314: spidx[], index of a->a, to be used in MatTransposeColoringCreate_SeqAIJ() and MatFDColoringCreate_SeqXAIJ()
315: */
316: PetscErrorCode MatGetColumnIJ_SeqAIJ_Color(Mat A,PetscInt oshift,PetscBool symmetric,PetscBool inodecompressed,PetscInt *nn,const PetscInt *ia[],const PetscInt *ja[],PetscInt *spidx[],PetscBool *done)
317: {
318: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
320: PetscInt i,*collengths,*cia,*cja,n = A->cmap->n,m = A->rmap->n;
321: PetscInt nz = a->i[m],row,mr,col,tmp;
322: PetscInt *cspidx;
323: const PetscInt *jj;
326: *nn = n;
327: if (!ia) return(0);
329: PetscCalloc1(n,&collengths);
330: PetscMalloc1(n+1,&cia);
331: PetscMalloc1(nz,&cja);
332: PetscMalloc1(nz,&cspidx);
333: jj = a->j;
334: for (i=0; i<nz; i++) {
335: collengths[jj[i]]++;
336: }
337: cia[0] = oshift;
338: for (i=0; i<n; i++) {
339: cia[i+1] = cia[i] + collengths[i];
340: }
341: PetscArrayzero(collengths,n);
342: jj = a->j;
343: for (row=0; row<m; row++) {
344: mr = a->i[row+1] - a->i[row];
345: for (i=0; i<mr; i++) {
346: col = *jj++;
347: tmp = cia[col] + collengths[col]++ - oshift;
348: cspidx[tmp] = a->i[row] + i; /* index of a->j */
349: cja[tmp] = row + oshift;
350: }
351: }
352: PetscFree(collengths);
353: *ia = cia;
354: *ja = cja;
355: *spidx = cspidx;
356: return(0);
357: }
359: PetscErrorCode MatRestoreColumnIJ_SeqAIJ_Color(Mat A,PetscInt oshift,PetscBool symmetric,PetscBool inodecompressed,PetscInt *n,const PetscInt *ia[],const PetscInt *ja[],PetscInt *spidx[],PetscBool *done)
360: {
364: MatRestoreColumnIJ_SeqAIJ(A,oshift,symmetric,inodecompressed,n,ia,ja,done);
365: PetscFree(*spidx);
366: return(0);
367: }
369: PetscErrorCode MatSetValuesRow_SeqAIJ(Mat A,PetscInt row,const PetscScalar v[])
370: {
371: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
372: PetscInt *ai = a->i;
376: PetscArraycpy(a->a+ai[row],v,ai[row+1]-ai[row]);
377: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
378: if (A->offloadmask != PETSC_OFFLOAD_UNALLOCATED && ai[row+1]-ai[row]) A->offloadmask = PETSC_OFFLOAD_CPU;
379: #endif
380: return(0);
381: }
383: /*
384: MatSeqAIJSetValuesLocalFast - An optimized version of MatSetValuesLocal() for SeqAIJ matrices with several assumptions
386: - a single row of values is set with each call
387: - no row or column indices are negative or (in error) larger than the number of rows or columns
388: - the values are always added to the matrix, not set
389: - no new locations are introduced in the nonzero structure of the matrix
391: This does NOT assume the global column indices are sorted
393: */
395: #include <petsc/private/isimpl.h>
396: PetscErrorCode MatSeqAIJSetValuesLocalFast(Mat A,PetscInt m,const PetscInt im[],PetscInt n,const PetscInt in[],const PetscScalar v[],InsertMode is)
397: {
398: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
399: PetscInt low,high,t,row,nrow,i,col,l;
400: const PetscInt *rp,*ai = a->i,*ailen = a->ilen,*aj = a->j;
401: PetscInt lastcol = -1;
402: MatScalar *ap,value,*aa = a->a;
403: const PetscInt *ridx = A->rmap->mapping->indices,*cidx = A->cmap->mapping->indices;
405: row = ridx[im[0]];
406: rp = aj + ai[row];
407: ap = aa + ai[row];
408: nrow = ailen[row];
409: low = 0;
410: high = nrow;
411: for (l=0; l<n; l++) { /* loop over added columns */
412: col = cidx[in[l]];
413: value = v[l];
415: if (col <= lastcol) low = 0;
416: else high = nrow;
417: lastcol = col;
418: while (high-low > 5) {
419: t = (low+high)/2;
420: if (rp[t] > col) high = t;
421: else low = t;
422: }
423: for (i=low; i<high; i++) {
424: if (rp[i] == col) {
425: ap[i] += value;
426: low = i + 1;
427: break;
428: }
429: }
430: }
431: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
432: if (A->offloadmask != PETSC_OFFLOAD_UNALLOCATED && m && n) A->offloadmask = PETSC_OFFLOAD_CPU;
433: #endif
434: return 0;
435: }
437: PetscErrorCode MatSetValues_SeqAIJ(Mat A,PetscInt m,const PetscInt im[],PetscInt n,const PetscInt in[],const PetscScalar v[],InsertMode is)
438: {
439: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
440: PetscInt *rp,k,low,high,t,ii,row,nrow,i,col,l,rmax,N;
441: PetscInt *imax = a->imax,*ai = a->i,*ailen = a->ilen;
443: PetscInt *aj = a->j,nonew = a->nonew,lastcol = -1;
444: MatScalar *ap=NULL,value=0.0,*aa = a->a;
445: PetscBool ignorezeroentries = a->ignorezeroentries;
446: PetscBool roworiented = a->roworiented;
447: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
448: PetscBool inserted = PETSC_FALSE;
449: #endif
452: for (k=0; k<m; k++) { /* loop over added rows */
453: row = im[k];
454: if (row < 0) continue;
455: #if defined(PETSC_USE_DEBUG)
456: if (row >= A->rmap->n) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Row too large: row %D max %D",row,A->rmap->n-1);
457: #endif
458: rp = aj + ai[row];
459: if (!A->structure_only) ap = aa + ai[row];
460: rmax = imax[row]; nrow = ailen[row];
461: low = 0;
462: high = nrow;
463: for (l=0; l<n; l++) { /* loop over added columns */
464: if (in[l] < 0) continue;
465: #if defined(PETSC_USE_DEBUG)
466: if (in[l] >= A->cmap->n) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Column too large: col %D max %D",in[l],A->cmap->n-1);
467: #endif
468: col = in[l];
469: if (v && !A->structure_only) value = roworiented ? v[l + k*n] : v[k + l*m];
470: if (!A->structure_only && value == 0.0 && ignorezeroentries && is == ADD_VALUES && row != col) continue;
472: if (col <= lastcol) low = 0;
473: else high = nrow;
474: lastcol = col;
475: while (high-low > 5) {
476: t = (low+high)/2;
477: if (rp[t] > col) high = t;
478: else low = t;
479: }
480: for (i=low; i<high; i++) {
481: if (rp[i] > col) break;
482: if (rp[i] == col) {
483: if (!A->structure_only) {
484: if (is == ADD_VALUES) {
485: ap[i] += value;
486: (void)PetscLogFlops(1.0);
487: }
488: else ap[i] = value;
489: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
490: inserted = PETSC_TRUE;
491: #endif
492: }
493: low = i + 1;
494: goto noinsert;
495: }
496: }
497: if (value == 0.0 && ignorezeroentries && row != col) goto noinsert;
498: if (nonew == 1) goto noinsert;
499: if (nonew == -1) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Inserting a new nonzero at (%D,%D) in the matrix",row,col);
500: if (A->structure_only) {
501: MatSeqXAIJReallocateAIJ_structure_only(A,A->rmap->n,1,nrow,row,col,rmax,ai,aj,rp,imax,nonew,MatScalar);
502: } else {
503: MatSeqXAIJReallocateAIJ(A,A->rmap->n,1,nrow,row,col,rmax,aa,ai,aj,rp,ap,imax,nonew,MatScalar);
504: }
505: N = nrow++ - 1; a->nz++; high++;
506: /* shift up all the later entries in this row */
507: PetscArraymove(rp+i+1,rp+i,N-i+1);
508: rp[i] = col;
509: if (!A->structure_only){
510: PetscArraymove(ap+i+1,ap+i,N-i+1);
511: ap[i] = value;
512: }
513: low = i + 1;
514: A->nonzerostate++;
515: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
516: inserted = PETSC_TRUE;
517: #endif
518: noinsert:;
519: }
520: ailen[row] = nrow;
521: }
522: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
523: if (A->offloadmask != PETSC_OFFLOAD_UNALLOCATED && inserted) A->offloadmask = PETSC_OFFLOAD_CPU;
524: #endif
525: return(0);
526: }
528: PetscErrorCode MatSetValues_SeqAIJ_SortedFull(Mat A,PetscInt m,const PetscInt im[],PetscInt n,const PetscInt in[],const PetscScalar v[],InsertMode is)
529: {
530: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
531: PetscInt *rp,k,row;
532: PetscInt *ai = a->i,*ailen = a->ilen;
534: PetscInt *aj = a->j;
535: MatScalar *aa = a->a,*ap;
538: for (k=0; k<m; k++) { /* loop over added rows */
539: row = im[k];
540: rp = aj + ai[row];
541: ap = aa + ai[row];
542: if (!A->was_assembled) {
543: PetscMemcpy(rp,in,n*sizeof(PetscInt));
544: }
545: if (!A->structure_only) {
546: if (v) {
547: PetscMemcpy(ap,v,n*sizeof(PetscScalar));
548: v += n;
549: } else {
550: PetscMemzero(ap,n*sizeof(PetscScalar));
551: }
552: }
553: ailen[row] = n;
554: a->nz += n;
555: }
556: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
557: if (A->offloadmask != PETSC_OFFLOAD_UNALLOCATED && m && n) A->offloadmask = PETSC_OFFLOAD_CPU;
558: #endif
559: return(0);
560: }
563: PetscErrorCode MatGetValues_SeqAIJ(Mat A,PetscInt m,const PetscInt im[],PetscInt n,const PetscInt in[],PetscScalar v[])
564: {
565: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
566: PetscInt *rp,k,low,high,t,row,nrow,i,col,l,*aj = a->j;
567: PetscInt *ai = a->i,*ailen = a->ilen;
568: MatScalar *ap,*aa = a->a;
571: for (k=0; k<m; k++) { /* loop over rows */
572: row = im[k];
573: if (row < 0) {v += n; continue;} /* SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Negative row: %D",row); */
574: if (row >= A->rmap->n) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Row too large: row %D max %D",row,A->rmap->n-1);
575: rp = aj + ai[row]; ap = aa + ai[row];
576: nrow = ailen[row];
577: for (l=0; l<n; l++) { /* loop over columns */
578: if (in[l] < 0) {v++; continue;} /* SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Negative column: %D",in[l]); */
579: if (in[l] >= A->cmap->n) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Column too large: col %D max %D",in[l],A->cmap->n-1);
580: col = in[l];
581: high = nrow; low = 0; /* assume unsorted */
582: while (high-low > 5) {
583: t = (low+high)/2;
584: if (rp[t] > col) high = t;
585: else low = t;
586: }
587: for (i=low; i<high; i++) {
588: if (rp[i] > col) break;
589: if (rp[i] == col) {
590: *v++ = ap[i];
591: goto finished;
592: }
593: }
594: *v++ = 0.0;
595: finished:;
596: }
597: }
598: return(0);
599: }
601: PetscErrorCode MatView_SeqAIJ_Binary(Mat mat,PetscViewer viewer)
602: {
603: Mat_SeqAIJ *A = (Mat_SeqAIJ*)mat->data;
604: PetscInt header[4],M,N,m,nz,i;
605: PetscInt *rowlens;
609: PetscViewerSetUp(viewer);
611: M = mat->rmap->N;
612: N = mat->cmap->N;
613: m = mat->rmap->n;
614: nz = A->nz;
616: /* write matrix header */
617: header[0] = MAT_FILE_CLASSID;
618: header[1] = M; header[2] = N; header[3] = nz;
619: PetscViewerBinaryWrite(viewer,header,4,PETSC_INT);
621: /* fill in and store row lengths */
622: PetscMalloc1(m,&rowlens);
623: for (i=0; i<m; i++) rowlens[i] = A->i[i+1] - A->i[i];
624: PetscViewerBinaryWrite(viewer,rowlens,m,PETSC_INT);
625: PetscFree(rowlens);
626: /* store column indices */
627: PetscViewerBinaryWrite(viewer,A->j,nz,PETSC_INT);
628: /* store nonzero values */
629: PetscViewerBinaryWrite(viewer,A->a,nz,PETSC_SCALAR);
631: /* write block size option to the viewer's .info file */
632: MatView_Binary_BlockSizes(mat,viewer);
633: return(0);
634: }
636: static PetscErrorCode MatView_SeqAIJ_ASCII_structonly(Mat A,PetscViewer viewer)
637: {
639: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
640: PetscInt i,k,m=A->rmap->N;
643: PetscViewerASCIIUseTabs(viewer,PETSC_FALSE);
644: for (i=0; i<m; i++) {
645: PetscViewerASCIIPrintf(viewer,"row %D:",i);
646: for (k=a->i[i]; k<a->i[i+1]; k++) {
647: PetscViewerASCIIPrintf(viewer," (%D) ",a->j[k]);
648: }
649: PetscViewerASCIIPrintf(viewer,"\n");
650: }
651: PetscViewerASCIIUseTabs(viewer,PETSC_TRUE);
652: return(0);
653: }
655: extern PetscErrorCode MatSeqAIJFactorInfo_Matlab(Mat,PetscViewer);
657: PetscErrorCode MatView_SeqAIJ_ASCII(Mat A,PetscViewer viewer)
658: {
659: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
660: PetscErrorCode ierr;
661: PetscInt i,j,m = A->rmap->n;
662: const char *name;
663: PetscViewerFormat format;
666: if (A->structure_only) {
667: MatView_SeqAIJ_ASCII_structonly(A,viewer);
668: return(0);
669: }
671: PetscViewerGetFormat(viewer,&format);
672: if (format == PETSC_VIEWER_ASCII_MATLAB) {
673: PetscInt nofinalvalue = 0;
674: if (m && ((a->i[m] == a->i[m-1]) || (a->j[a->nz-1] != A->cmap->n-1))) {
675: /* Need a dummy value to ensure the dimension of the matrix. */
676: nofinalvalue = 1;
677: }
678: PetscViewerASCIIUseTabs(viewer,PETSC_FALSE);
679: PetscViewerASCIIPrintf(viewer,"%% Size = %D %D \n",m,A->cmap->n);
680: PetscViewerASCIIPrintf(viewer,"%% Nonzeros = %D \n",a->nz);
681: #if defined(PETSC_USE_COMPLEX)
682: PetscViewerASCIIPrintf(viewer,"zzz = zeros(%D,4);\n",a->nz+nofinalvalue);
683: #else
684: PetscViewerASCIIPrintf(viewer,"zzz = zeros(%D,3);\n",a->nz+nofinalvalue);
685: #endif
686: PetscViewerASCIIPrintf(viewer,"zzz = [\n");
688: for (i=0; i<m; i++) {
689: for (j=a->i[i]; j<a->i[i+1]; j++) {
690: #if defined(PETSC_USE_COMPLEX)
691: PetscViewerASCIIPrintf(viewer,"%D %D %18.16e %18.16e\n",i+1,a->j[j]+1,(double)PetscRealPart(a->a[j]),(double)PetscImaginaryPart(a->a[j]));
692: #else
693: PetscViewerASCIIPrintf(viewer,"%D %D %18.16e\n",i+1,a->j[j]+1,(double)a->a[j]);
694: #endif
695: }
696: }
697: if (nofinalvalue) {
698: #if defined(PETSC_USE_COMPLEX)
699: PetscViewerASCIIPrintf(viewer,"%D %D %18.16e %18.16e\n",m,A->cmap->n,0.,0.);
700: #else
701: PetscViewerASCIIPrintf(viewer,"%D %D %18.16e\n",m,A->cmap->n,0.0);
702: #endif
703: }
704: PetscObjectGetName((PetscObject)A,&name);
705: PetscViewerASCIIPrintf(viewer,"];\n %s = spconvert(zzz);\n",name);
706: PetscViewerASCIIUseTabs(viewer,PETSC_TRUE);
707: } else if (format == PETSC_VIEWER_ASCII_FACTOR_INFO || format == PETSC_VIEWER_ASCII_INFO || format == PETSC_VIEWER_ASCII_INFO_DETAIL) {
708: return(0);
709: } else if (format == PETSC_VIEWER_ASCII_COMMON) {
710: PetscViewerASCIIUseTabs(viewer,PETSC_FALSE);
711: for (i=0; i<m; i++) {
712: PetscViewerASCIIPrintf(viewer,"row %D:",i);
713: for (j=a->i[i]; j<a->i[i+1]; j++) {
714: #if defined(PETSC_USE_COMPLEX)
715: if (PetscImaginaryPart(a->a[j]) > 0.0 && PetscRealPart(a->a[j]) != 0.0) {
716: PetscViewerASCIIPrintf(viewer," (%D, %g + %g i)",a->j[j],(double)PetscRealPart(a->a[j]),(double)PetscImaginaryPart(a->a[j]));
717: } else if (PetscImaginaryPart(a->a[j]) < 0.0 && PetscRealPart(a->a[j]) != 0.0) {
718: PetscViewerASCIIPrintf(viewer," (%D, %g - %g i)",a->j[j],(double)PetscRealPart(a->a[j]),(double)-PetscImaginaryPart(a->a[j]));
719: } else if (PetscRealPart(a->a[j]) != 0.0) {
720: PetscViewerASCIIPrintf(viewer," (%D, %g) ",a->j[j],(double)PetscRealPart(a->a[j]));
721: }
722: #else
723: if (a->a[j] != 0.0) {PetscViewerASCIIPrintf(viewer," (%D, %g) ",a->j[j],(double)a->a[j]);}
724: #endif
725: }
726: PetscViewerASCIIPrintf(viewer,"\n");
727: }
728: PetscViewerASCIIUseTabs(viewer,PETSC_TRUE);
729: } else if (format == PETSC_VIEWER_ASCII_SYMMODU) {
730: PetscInt nzd=0,fshift=1,*sptr;
731: PetscViewerASCIIUseTabs(viewer,PETSC_FALSE);
732: PetscMalloc1(m+1,&sptr);
733: for (i=0; i<m; i++) {
734: sptr[i] = nzd+1;
735: for (j=a->i[i]; j<a->i[i+1]; j++) {
736: if (a->j[j] >= i) {
737: #if defined(PETSC_USE_COMPLEX)
738: if (PetscImaginaryPart(a->a[j]) != 0.0 || PetscRealPart(a->a[j]) != 0.0) nzd++;
739: #else
740: if (a->a[j] != 0.0) nzd++;
741: #endif
742: }
743: }
744: }
745: sptr[m] = nzd+1;
746: PetscViewerASCIIPrintf(viewer," %D %D\n\n",m,nzd);
747: for (i=0; i<m+1; i+=6) {
748: if (i+4<m) {
749: PetscViewerASCIIPrintf(viewer," %D %D %D %D %D %D\n",sptr[i],sptr[i+1],sptr[i+2],sptr[i+3],sptr[i+4],sptr[i+5]);
750: } else if (i+3<m) {
751: PetscViewerASCIIPrintf(viewer," %D %D %D %D %D\n",sptr[i],sptr[i+1],sptr[i+2],sptr[i+3],sptr[i+4]);
752: } else if (i+2<m) {
753: PetscViewerASCIIPrintf(viewer," %D %D %D %D\n",sptr[i],sptr[i+1],sptr[i+2],sptr[i+3]);
754: } else if (i+1<m) {
755: PetscViewerASCIIPrintf(viewer," %D %D %D\n",sptr[i],sptr[i+1],sptr[i+2]);
756: } else if (i<m) {
757: PetscViewerASCIIPrintf(viewer," %D %D\n",sptr[i],sptr[i+1]);
758: } else {
759: PetscViewerASCIIPrintf(viewer," %D\n",sptr[i]);
760: }
761: }
762: PetscViewerASCIIPrintf(viewer,"\n");
763: PetscFree(sptr);
764: for (i=0; i<m; i++) {
765: for (j=a->i[i]; j<a->i[i+1]; j++) {
766: if (a->j[j] >= i) {PetscViewerASCIIPrintf(viewer," %D ",a->j[j]+fshift);}
767: }
768: PetscViewerASCIIPrintf(viewer,"\n");
769: }
770: PetscViewerASCIIPrintf(viewer,"\n");
771: for (i=0; i<m; i++) {
772: for (j=a->i[i]; j<a->i[i+1]; j++) {
773: if (a->j[j] >= i) {
774: #if defined(PETSC_USE_COMPLEX)
775: if (PetscImaginaryPart(a->a[j]) != 0.0 || PetscRealPart(a->a[j]) != 0.0) {
776: PetscViewerASCIIPrintf(viewer," %18.16e %18.16e ",(double)PetscRealPart(a->a[j]),(double)PetscImaginaryPart(a->a[j]));
777: }
778: #else
779: if (a->a[j] != 0.0) {PetscViewerASCIIPrintf(viewer," %18.16e ",(double)a->a[j]);}
780: #endif
781: }
782: }
783: PetscViewerASCIIPrintf(viewer,"\n");
784: }
785: PetscViewerASCIIUseTabs(viewer,PETSC_TRUE);
786: } else if (format == PETSC_VIEWER_ASCII_DENSE) {
787: PetscInt cnt = 0,jcnt;
788: PetscScalar value;
789: #if defined(PETSC_USE_COMPLEX)
790: PetscBool realonly = PETSC_TRUE;
792: for (i=0; i<a->i[m]; i++) {
793: if (PetscImaginaryPart(a->a[i]) != 0.0) {
794: realonly = PETSC_FALSE;
795: break;
796: }
797: }
798: #endif
800: PetscViewerASCIIUseTabs(viewer,PETSC_FALSE);
801: for (i=0; i<m; i++) {
802: jcnt = 0;
803: for (j=0; j<A->cmap->n; j++) {
804: if (jcnt < a->i[i+1]-a->i[i] && j == a->j[cnt]) {
805: value = a->a[cnt++];
806: jcnt++;
807: } else {
808: value = 0.0;
809: }
810: #if defined(PETSC_USE_COMPLEX)
811: if (realonly) {
812: PetscViewerASCIIPrintf(viewer," %7.5e ",(double)PetscRealPart(value));
813: } else {
814: PetscViewerASCIIPrintf(viewer," %7.5e+%7.5e i ",(double)PetscRealPart(value),(double)PetscImaginaryPart(value));
815: }
816: #else
817: PetscViewerASCIIPrintf(viewer," %7.5e ",(double)value);
818: #endif
819: }
820: PetscViewerASCIIPrintf(viewer,"\n");
821: }
822: PetscViewerASCIIUseTabs(viewer,PETSC_TRUE);
823: } else if (format == PETSC_VIEWER_ASCII_MATRIXMARKET) {
824: PetscInt fshift=1;
825: PetscViewerASCIIUseTabs(viewer,PETSC_FALSE);
826: #if defined(PETSC_USE_COMPLEX)
827: PetscViewerASCIIPrintf(viewer,"%%%%MatrixMarket matrix coordinate complex general\n");
828: #else
829: PetscViewerASCIIPrintf(viewer,"%%%%MatrixMarket matrix coordinate real general\n");
830: #endif
831: PetscViewerASCIIPrintf(viewer,"%D %D %D\n", m, A->cmap->n, a->nz);
832: for (i=0; i<m; i++) {
833: for (j=a->i[i]; j<a->i[i+1]; j++) {
834: #if defined(PETSC_USE_COMPLEX)
835: PetscViewerASCIIPrintf(viewer,"%D %D %g %g\n", i+fshift,a->j[j]+fshift,(double)PetscRealPart(a->a[j]),(double)PetscImaginaryPart(a->a[j]));
836: #else
837: PetscViewerASCIIPrintf(viewer,"%D %D %g\n", i+fshift, a->j[j]+fshift, (double)a->a[j]);
838: #endif
839: }
840: }
841: PetscViewerASCIIUseTabs(viewer,PETSC_TRUE);
842: } else {
843: PetscViewerASCIIUseTabs(viewer,PETSC_FALSE);
844: if (A->factortype) {
845: for (i=0; i<m; i++) {
846: PetscViewerASCIIPrintf(viewer,"row %D:",i);
847: /* L part */
848: for (j=a->i[i]; j<a->i[i+1]; j++) {
849: #if defined(PETSC_USE_COMPLEX)
850: if (PetscImaginaryPart(a->a[j]) > 0.0) {
851: PetscViewerASCIIPrintf(viewer," (%D, %g + %g i)",a->j[j],(double)PetscRealPart(a->a[j]),(double)PetscImaginaryPart(a->a[j]));
852: } else if (PetscImaginaryPart(a->a[j]) < 0.0) {
853: PetscViewerASCIIPrintf(viewer," (%D, %g - %g i)",a->j[j],(double)PetscRealPart(a->a[j]),(double)(-PetscImaginaryPart(a->a[j])));
854: } else {
855: PetscViewerASCIIPrintf(viewer," (%D, %g) ",a->j[j],(double)PetscRealPart(a->a[j]));
856: }
857: #else
858: PetscViewerASCIIPrintf(viewer," (%D, %g) ",a->j[j],(double)a->a[j]);
859: #endif
860: }
861: /* diagonal */
862: j = a->diag[i];
863: #if defined(PETSC_USE_COMPLEX)
864: if (PetscImaginaryPart(a->a[j]) > 0.0) {
865: PetscViewerASCIIPrintf(viewer," (%D, %g + %g i)",a->j[j],(double)PetscRealPart(1.0/a->a[j]),(double)PetscImaginaryPart(1.0/a->a[j]));
866: } else if (PetscImaginaryPart(a->a[j]) < 0.0) {
867: PetscViewerASCIIPrintf(viewer," (%D, %g - %g i)",a->j[j],(double)PetscRealPart(1.0/a->a[j]),(double)(-PetscImaginaryPart(1.0/a->a[j])));
868: } else {
869: PetscViewerASCIIPrintf(viewer," (%D, %g) ",a->j[j],(double)PetscRealPart(1.0/a->a[j]));
870: }
871: #else
872: PetscViewerASCIIPrintf(viewer," (%D, %g) ",a->j[j],(double)(1.0/a->a[j]));
873: #endif
875: /* U part */
876: for (j=a->diag[i+1]+1; j<a->diag[i]; j++) {
877: #if defined(PETSC_USE_COMPLEX)
878: if (PetscImaginaryPart(a->a[j]) > 0.0) {
879: PetscViewerASCIIPrintf(viewer," (%D, %g + %g i)",a->j[j],(double)PetscRealPart(a->a[j]),(double)PetscImaginaryPart(a->a[j]));
880: } else if (PetscImaginaryPart(a->a[j]) < 0.0) {
881: PetscViewerASCIIPrintf(viewer," (%D, %g - %g i)",a->j[j],(double)PetscRealPart(a->a[j]),(double)(-PetscImaginaryPart(a->a[j])));
882: } else {
883: PetscViewerASCIIPrintf(viewer," (%D, %g) ",a->j[j],(double)PetscRealPart(a->a[j]));
884: }
885: #else
886: PetscViewerASCIIPrintf(viewer," (%D, %g) ",a->j[j],(double)a->a[j]);
887: #endif
888: }
889: PetscViewerASCIIPrintf(viewer,"\n");
890: }
891: } else {
892: for (i=0; i<m; i++) {
893: PetscViewerASCIIPrintf(viewer,"row %D:",i);
894: for (j=a->i[i]; j<a->i[i+1]; j++) {
895: #if defined(PETSC_USE_COMPLEX)
896: if (PetscImaginaryPart(a->a[j]) > 0.0) {
897: PetscViewerASCIIPrintf(viewer," (%D, %g + %g i)",a->j[j],(double)PetscRealPart(a->a[j]),(double)PetscImaginaryPart(a->a[j]));
898: } else if (PetscImaginaryPart(a->a[j]) < 0.0) {
899: PetscViewerASCIIPrintf(viewer," (%D, %g - %g i)",a->j[j],(double)PetscRealPart(a->a[j]),(double)-PetscImaginaryPart(a->a[j]));
900: } else {
901: PetscViewerASCIIPrintf(viewer," (%D, %g) ",a->j[j],(double)PetscRealPart(a->a[j]));
902: }
903: #else
904: PetscViewerASCIIPrintf(viewer," (%D, %g) ",a->j[j],(double)a->a[j]);
905: #endif
906: }
907: PetscViewerASCIIPrintf(viewer,"\n");
908: }
909: }
910: PetscViewerASCIIUseTabs(viewer,PETSC_TRUE);
911: }
912: PetscViewerFlush(viewer);
913: return(0);
914: }
916: #include <petscdraw.h>
917: PetscErrorCode MatView_SeqAIJ_Draw_Zoom(PetscDraw draw,void *Aa)
918: {
919: Mat A = (Mat) Aa;
920: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
921: PetscErrorCode ierr;
922: PetscInt i,j,m = A->rmap->n;
923: int color;
924: PetscReal xl,yl,xr,yr,x_l,x_r,y_l,y_r;
925: PetscViewer viewer;
926: PetscViewerFormat format;
929: PetscObjectQuery((PetscObject)A,"Zoomviewer",(PetscObject*)&viewer);
930: PetscViewerGetFormat(viewer,&format);
931: PetscDrawGetCoordinates(draw,&xl,&yl,&xr,&yr);
933: /* loop over matrix elements drawing boxes */
935: if (format != PETSC_VIEWER_DRAW_CONTOUR) {
936: PetscDrawCollectiveBegin(draw);
937: /* Blue for negative, Cyan for zero and Red for positive */
938: color = PETSC_DRAW_BLUE;
939: for (i=0; i<m; i++) {
940: y_l = m - i - 1.0; y_r = y_l + 1.0;
941: for (j=a->i[i]; j<a->i[i+1]; j++) {
942: x_l = a->j[j]; x_r = x_l + 1.0;
943: if (PetscRealPart(a->a[j]) >= 0.) continue;
944: PetscDrawRectangle(draw,x_l,y_l,x_r,y_r,color,color,color,color);
945: }
946: }
947: color = PETSC_DRAW_CYAN;
948: for (i=0; i<m; i++) {
949: y_l = m - i - 1.0; y_r = y_l + 1.0;
950: for (j=a->i[i]; j<a->i[i+1]; j++) {
951: x_l = a->j[j]; x_r = x_l + 1.0;
952: if (a->a[j] != 0.) continue;
953: PetscDrawRectangle(draw,x_l,y_l,x_r,y_r,color,color,color,color);
954: }
955: }
956: color = PETSC_DRAW_RED;
957: for (i=0; i<m; i++) {
958: y_l = m - i - 1.0; y_r = y_l + 1.0;
959: for (j=a->i[i]; j<a->i[i+1]; j++) {
960: x_l = a->j[j]; x_r = x_l + 1.0;
961: if (PetscRealPart(a->a[j]) <= 0.) continue;
962: PetscDrawRectangle(draw,x_l,y_l,x_r,y_r,color,color,color,color);
963: }
964: }
965: PetscDrawCollectiveEnd(draw);
966: } else {
967: /* use contour shading to indicate magnitude of values */
968: /* first determine max of all nonzero values */
969: PetscReal minv = 0.0, maxv = 0.0;
970: PetscInt nz = a->nz, count = 0;
971: PetscDraw popup;
973: for (i=0; i<nz; i++) {
974: if (PetscAbsScalar(a->a[i]) > maxv) maxv = PetscAbsScalar(a->a[i]);
975: }
976: if (minv >= maxv) maxv = minv + PETSC_SMALL;
977: PetscDrawGetPopup(draw,&popup);
978: PetscDrawScalePopup(popup,minv,maxv);
980: PetscDrawCollectiveBegin(draw);
981: for (i=0; i<m; i++) {
982: y_l = m - i - 1.0;
983: y_r = y_l + 1.0;
984: for (j=a->i[i]; j<a->i[i+1]; j++) {
985: x_l = a->j[j];
986: x_r = x_l + 1.0;
987: color = PetscDrawRealToColor(PetscAbsScalar(a->a[count]),minv,maxv);
988: PetscDrawRectangle(draw,x_l,y_l,x_r,y_r,color,color,color,color);
989: count++;
990: }
991: }
992: PetscDrawCollectiveEnd(draw);
993: }
994: return(0);
995: }
997: #include <petscdraw.h>
998: PetscErrorCode MatView_SeqAIJ_Draw(Mat A,PetscViewer viewer)
999: {
1001: PetscDraw draw;
1002: PetscReal xr,yr,xl,yl,h,w;
1003: PetscBool isnull;
1006: PetscViewerDrawGetDraw(viewer,0,&draw);
1007: PetscDrawIsNull(draw,&isnull);
1008: if (isnull) return(0);
1010: xr = A->cmap->n; yr = A->rmap->n; h = yr/10.0; w = xr/10.0;
1011: xr += w; yr += h; xl = -w; yl = -h;
1012: PetscDrawSetCoordinates(draw,xl,yl,xr,yr);
1013: PetscObjectCompose((PetscObject)A,"Zoomviewer",(PetscObject)viewer);
1014: PetscDrawZoom(draw,MatView_SeqAIJ_Draw_Zoom,A);
1015: PetscObjectCompose((PetscObject)A,"Zoomviewer",NULL);
1016: PetscDrawSave(draw);
1017: return(0);
1018: }
1020: PetscErrorCode MatView_SeqAIJ(Mat A,PetscViewer viewer)
1021: {
1023: PetscBool iascii,isbinary,isdraw;
1026: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERASCII,&iascii);
1027: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERBINARY,&isbinary);
1028: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERDRAW,&isdraw);
1029: if (iascii) {
1030: MatView_SeqAIJ_ASCII(A,viewer);
1031: } else if (isbinary) {
1032: MatView_SeqAIJ_Binary(A,viewer);
1033: } else if (isdraw) {
1034: MatView_SeqAIJ_Draw(A,viewer);
1035: }
1036: MatView_SeqAIJ_Inode(A,viewer);
1037: return(0);
1038: }
1040: PetscErrorCode MatAssemblyEnd_SeqAIJ(Mat A,MatAssemblyType mode)
1041: {
1042: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1044: PetscInt fshift = 0,i,*ai = a->i,*aj = a->j,*imax = a->imax;
1045: PetscInt m = A->rmap->n,*ip,N,*ailen = a->ilen,rmax = 0;
1046: MatScalar *aa = a->a,*ap;
1047: PetscReal ratio = 0.6;
1050: if (mode == MAT_FLUSH_ASSEMBLY) return(0);
1051: MatSeqAIJInvalidateDiagonal(A);
1052: if (A->was_assembled && A->ass_nonzerostate == A->nonzerostate) return(0);
1054: if (m) rmax = ailen[0]; /* determine row with most nonzeros */
1055: for (i=1; i<m; i++) {
1056: /* move each row back by the amount of empty slots (fshift) before it*/
1057: fshift += imax[i-1] - ailen[i-1];
1058: rmax = PetscMax(rmax,ailen[i]);
1059: if (fshift) {
1060: ip = aj + ai[i];
1061: ap = aa + ai[i];
1062: N = ailen[i];
1063: PetscArraymove(ip-fshift,ip,N);
1064: if (!A->structure_only) {
1065: PetscArraymove(ap-fshift,ap,N);
1066: }
1067: }
1068: ai[i] = ai[i-1] + ailen[i-1];
1069: }
1070: if (m) {
1071: fshift += imax[m-1] - ailen[m-1];
1072: ai[m] = ai[m-1] + ailen[m-1];
1073: }
1075: /* reset ilen and imax for each row */
1076: a->nonzerorowcnt = 0;
1077: if (A->structure_only) {
1078: PetscFree(a->imax);
1079: PetscFree(a->ilen);
1080: } else { /* !A->structure_only */
1081: for (i=0; i<m; i++) {
1082: ailen[i] = imax[i] = ai[i+1] - ai[i];
1083: a->nonzerorowcnt += ((ai[i+1] - ai[i]) > 0);
1084: }
1085: }
1086: a->nz = ai[m];
1087: if (fshift && a->nounused == -1) SETERRQ3(PETSC_COMM_SELF,PETSC_ERR_PLIB, "Unused space detected in matrix: %D X %D, %D unneeded", m, A->cmap->n, fshift);
1089: MatMarkDiagonal_SeqAIJ(A);
1090: PetscInfo4(A,"Matrix size: %D X %D; storage space: %D unneeded,%D used\n",m,A->cmap->n,fshift,a->nz);
1091: PetscInfo1(A,"Number of mallocs during MatSetValues() is %D\n",a->reallocs);
1092: PetscInfo1(A,"Maximum nonzeros in any row is %D\n",rmax);
1094: A->info.mallocs += a->reallocs;
1095: a->reallocs = 0;
1096: A->info.nz_unneeded = (PetscReal)fshift;
1097: a->rmax = rmax;
1099: if (!A->structure_only) {
1100: MatCheckCompressedRow(A,a->nonzerorowcnt,&a->compressedrow,a->i,m,ratio);
1101: }
1102: MatAssemblyEnd_SeqAIJ_Inode(A,mode);
1103: return(0);
1104: }
1106: PetscErrorCode MatRealPart_SeqAIJ(Mat A)
1107: {
1108: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1109: PetscInt i,nz = a->nz;
1110: MatScalar *aa = a->a;
1114: for (i=0; i<nz; i++) aa[i] = PetscRealPart(aa[i]);
1115: MatSeqAIJInvalidateDiagonal(A);
1116: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
1117: if (A->offloadmask != PETSC_OFFLOAD_UNALLOCATED) A->offloadmask = PETSC_OFFLOAD_CPU;
1118: #endif
1119: return(0);
1120: }
1122: PetscErrorCode MatImaginaryPart_SeqAIJ(Mat A)
1123: {
1124: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1125: PetscInt i,nz = a->nz;
1126: MatScalar *aa = a->a;
1130: for (i=0; i<nz; i++) aa[i] = PetscImaginaryPart(aa[i]);
1131: MatSeqAIJInvalidateDiagonal(A);
1132: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
1133: if (A->offloadmask != PETSC_OFFLOAD_UNALLOCATED) A->offloadmask = PETSC_OFFLOAD_CPU;
1134: #endif
1135: return(0);
1136: }
1138: PetscErrorCode MatZeroEntries_SeqAIJ(Mat A)
1139: {
1140: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1144: PetscArrayzero(a->a,a->i[A->rmap->n]);
1145: MatSeqAIJInvalidateDiagonal(A);
1146: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
1147: if (A->offloadmask != PETSC_OFFLOAD_UNALLOCATED) A->offloadmask = PETSC_OFFLOAD_CPU;
1148: #endif
1149: return(0);
1150: }
1152: PetscErrorCode MatDestroy_SeqAIJ(Mat A)
1153: {
1154: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1158: #if defined(PETSC_USE_LOG)
1159: PetscLogObjectState((PetscObject)A,"Rows=%D, Cols=%D, NZ=%D",A->rmap->n,A->cmap->n,a->nz);
1160: #endif
1161: MatSeqXAIJFreeAIJ(A,&a->a,&a->j,&a->i);
1162: ISDestroy(&a->row);
1163: ISDestroy(&a->col);
1164: PetscFree(a->diag);
1165: PetscFree(a->ibdiag);
1166: PetscFree(a->imax);
1167: PetscFree(a->ilen);
1168: PetscFree(a->ipre);
1169: PetscFree3(a->idiag,a->mdiag,a->ssor_work);
1170: PetscFree(a->solve_work);
1171: ISDestroy(&a->icol);
1172: PetscFree(a->saved_values);
1173: ISColoringDestroy(&a->coloring);
1174: PetscFree2(a->compressedrow.i,a->compressedrow.rindex);
1175: PetscFree(a->matmult_abdense);
1177: MatDestroy_SeqAIJ_Inode(A);
1178: PetscFree(A->data);
1180: PetscObjectChangeTypeName((PetscObject)A,0);
1181: PetscObjectComposeFunction((PetscObject)A,"MatSeqAIJSetColumnIndices_C",NULL);
1182: PetscObjectComposeFunction((PetscObject)A,"MatStoreValues_C",NULL);
1183: PetscObjectComposeFunction((PetscObject)A,"MatRetrieveValues_C",NULL);
1184: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqaij_seqsbaij_C",NULL);
1185: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqaij_seqbaij_C",NULL);
1186: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqaij_seqaijperm_C",NULL);
1188: #if defined(PETSC_HAVE_CUDA)
1189: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqaij_seqaijcusparse_C",NULL);
1190: PetscObjectComposeFunction((PetscObject)A,"MatSetFromOptions_seqaijcusparse_seqaij_C",NULL);
1191: #endif
1192: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqaij_seqaijcrl_C",NULL);
1193: #if defined(PETSC_HAVE_ELEMENTAL)
1194: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqaij_elemental_C",NULL);
1195: #endif
1196: #if defined(PETSC_HAVE_HYPRE)
1197: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqaij_hypre_C",NULL);
1198: PetscObjectComposeFunction((PetscObject)A,"MatProductSetFromOptions_transpose_seqaij_seqaij_C",NULL);
1199: #endif
1200: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqaij_seqdense_C",NULL);
1201: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqaij_seqsell_C",NULL);
1202: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqaij_is_C",NULL);
1203: PetscObjectComposeFunction((PetscObject)A,"MatIsTranspose_C",NULL);
1204: PetscObjectComposeFunction((PetscObject)A,"MatSeqAIJSetPreallocation_C",NULL);
1205: PetscObjectComposeFunction((PetscObject)A,"MatResetPreallocation_C",NULL);
1206: PetscObjectComposeFunction((PetscObject)A,"MatSeqAIJSetPreallocationCSR_C",NULL);
1207: PetscObjectComposeFunction((PetscObject)A,"MatReorderForNonzeroDiagonal_C",NULL);
1208: PetscObjectComposeFunction((PetscObject)A,"MatProductSetFromOptions_is_seqaij_C",NULL);
1209: PetscObjectComposeFunction((PetscObject)A,"MatProductSetFromOptions_seqdense_seqaij_C",NULL);
1210: PetscObjectComposeFunction((PetscObject)A,"MatProductSetFromOptions_seqaij_seqaij_C",NULL);
1211: return(0);
1212: }
1214: PetscErrorCode MatSetOption_SeqAIJ(Mat A,MatOption op,PetscBool flg)
1215: {
1216: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1220: switch (op) {
1221: case MAT_ROW_ORIENTED:
1222: a->roworiented = flg;
1223: break;
1224: case MAT_KEEP_NONZERO_PATTERN:
1225: a->keepnonzeropattern = flg;
1226: break;
1227: case MAT_NEW_NONZERO_LOCATIONS:
1228: a->nonew = (flg ? 0 : 1);
1229: break;
1230: case MAT_NEW_NONZERO_LOCATION_ERR:
1231: a->nonew = (flg ? -1 : 0);
1232: break;
1233: case MAT_NEW_NONZERO_ALLOCATION_ERR:
1234: a->nonew = (flg ? -2 : 0);
1235: break;
1236: case MAT_UNUSED_NONZERO_LOCATION_ERR:
1237: a->nounused = (flg ? -1 : 0);
1238: break;
1239: case MAT_IGNORE_ZERO_ENTRIES:
1240: a->ignorezeroentries = flg;
1241: break;
1242: case MAT_SPD:
1243: case MAT_SYMMETRIC:
1244: case MAT_STRUCTURALLY_SYMMETRIC:
1245: case MAT_HERMITIAN:
1246: case MAT_SYMMETRY_ETERNAL:
1247: case MAT_STRUCTURE_ONLY:
1248: /* These options are handled directly by MatSetOption() */
1249: break;
1250: case MAT_NEW_DIAGONALS:
1251: case MAT_IGNORE_OFF_PROC_ENTRIES:
1252: case MAT_USE_HASH_TABLE:
1253: PetscInfo1(A,"Option %s ignored\n",MatOptions[op]);
1254: break;
1255: case MAT_USE_INODES:
1256: /* Not an error because MatSetOption_SeqAIJ_Inode handles this one */
1257: break;
1258: case MAT_SUBMAT_SINGLEIS:
1259: A->submat_singleis = flg;
1260: break;
1261: case MAT_SORTED_FULL:
1262: if (flg) A->ops->setvalues = MatSetValues_SeqAIJ_SortedFull;
1263: else A->ops->setvalues = MatSetValues_SeqAIJ;
1264: break;
1265: default:
1266: SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"unknown option %d",op);
1267: }
1268: MatSetOption_SeqAIJ_Inode(A,op,flg);
1269: return(0);
1270: }
1272: PetscErrorCode MatGetDiagonal_SeqAIJ(Mat A,Vec v)
1273: {
1274: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1276: PetscInt i,j,n,*ai=a->i,*aj=a->j;
1277: PetscScalar *aa=a->a,*x;
1280: VecGetLocalSize(v,&n);
1281: if (n != A->rmap->n) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Nonconforming matrix and vector");
1283: if (A->factortype == MAT_FACTOR_ILU || A->factortype == MAT_FACTOR_LU) {
1284: PetscInt *diag=a->diag;
1285: VecGetArrayWrite(v,&x);
1286: for (i=0; i<n; i++) x[i] = 1.0/aa[diag[i]];
1287: VecRestoreArrayWrite(v,&x);
1288: return(0);
1289: }
1291: VecGetArrayWrite(v,&x);
1292: for (i=0; i<n; i++) {
1293: x[i] = 0.0;
1294: for (j=ai[i]; j<ai[i+1]; j++) {
1295: if (aj[j] == i) {
1296: x[i] = aa[j];
1297: break;
1298: }
1299: }
1300: }
1301: VecRestoreArrayWrite(v,&x);
1302: return(0);
1303: }
1305: #include <../src/mat/impls/aij/seq/ftn-kernels/fmult.h>
1306: PetscErrorCode MatMultTransposeAdd_SeqAIJ(Mat A,Vec xx,Vec zz,Vec yy)
1307: {
1308: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1309: PetscScalar *y;
1310: const PetscScalar *x;
1311: PetscErrorCode ierr;
1312: PetscInt m = A->rmap->n;
1313: #if !defined(PETSC_USE_FORTRAN_KERNEL_MULTTRANSPOSEAIJ)
1314: const MatScalar *v;
1315: PetscScalar alpha;
1316: PetscInt n,i,j;
1317: const PetscInt *idx,*ii,*ridx=NULL;
1318: Mat_CompressedRow cprow = a->compressedrow;
1319: PetscBool usecprow = cprow.use;
1320: #endif
1323: if (zz != yy) {VecCopy(zz,yy);}
1324: VecGetArrayRead(xx,&x);
1325: VecGetArray(yy,&y);
1327: #if defined(PETSC_USE_FORTRAN_KERNEL_MULTTRANSPOSEAIJ)
1328: fortranmulttransposeaddaij_(&m,x,a->i,a->j,a->a,y);
1329: #else
1330: if (usecprow) {
1331: m = cprow.nrows;
1332: ii = cprow.i;
1333: ridx = cprow.rindex;
1334: } else {
1335: ii = a->i;
1336: }
1337: for (i=0; i<m; i++) {
1338: idx = a->j + ii[i];
1339: v = a->a + ii[i];
1340: n = ii[i+1] - ii[i];
1341: if (usecprow) {
1342: alpha = x[ridx[i]];
1343: } else {
1344: alpha = x[i];
1345: }
1346: for (j=0; j<n; j++) y[idx[j]] += alpha*v[j];
1347: }
1348: #endif
1349: PetscLogFlops(2.0*a->nz);
1350: VecRestoreArrayRead(xx,&x);
1351: VecRestoreArray(yy,&y);
1352: return(0);
1353: }
1355: PetscErrorCode MatMultTranspose_SeqAIJ(Mat A,Vec xx,Vec yy)
1356: {
1360: VecSet(yy,0.0);
1361: MatMultTransposeAdd_SeqAIJ(A,xx,yy,yy);
1362: return(0);
1363: }
1365: #include <../src/mat/impls/aij/seq/ftn-kernels/fmult.h>
1367: PetscErrorCode MatMult_SeqAIJ(Mat A,Vec xx,Vec yy)
1368: {
1369: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1370: PetscScalar *y;
1371: const PetscScalar *x;
1372: const MatScalar *aa;
1373: PetscErrorCode ierr;
1374: PetscInt m=A->rmap->n;
1375: const PetscInt *aj,*ii,*ridx=NULL;
1376: PetscInt n,i;
1377: PetscScalar sum;
1378: PetscBool usecprow=a->compressedrow.use;
1380: #if defined(PETSC_HAVE_PRAGMA_DISJOINT)
1381: #pragma disjoint(*x,*y,*aa)
1382: #endif
1385: VecGetArrayRead(xx,&x);
1386: VecGetArray(yy,&y);
1387: ii = a->i;
1388: if (usecprow) { /* use compressed row format */
1389: PetscArrayzero(y,m);
1390: m = a->compressedrow.nrows;
1391: ii = a->compressedrow.i;
1392: ridx = a->compressedrow.rindex;
1393: for (i=0; i<m; i++) {
1394: n = ii[i+1] - ii[i];
1395: aj = a->j + ii[i];
1396: aa = a->a + ii[i];
1397: sum = 0.0;
1398: PetscSparseDensePlusDot(sum,x,aa,aj,n);
1399: /* for (j=0; j<n; j++) sum += (*aa++)*x[*aj++]; */
1400: y[*ridx++] = sum;
1401: }
1402: } else { /* do not use compressed row format */
1403: #if defined(PETSC_USE_FORTRAN_KERNEL_MULTAIJ)
1404: aj = a->j;
1405: aa = a->a;
1406: fortranmultaij_(&m,x,ii,aj,aa,y);
1407: #else
1408: for (i=0; i<m; i++) {
1409: n = ii[i+1] - ii[i];
1410: aj = a->j + ii[i];
1411: aa = a->a + ii[i];
1412: sum = 0.0;
1413: PetscSparseDensePlusDot(sum,x,aa,aj,n);
1414: y[i] = sum;
1415: }
1416: #endif
1417: }
1418: PetscLogFlops(2.0*a->nz - a->nonzerorowcnt);
1419: VecRestoreArrayRead(xx,&x);
1420: VecRestoreArray(yy,&y);
1421: return(0);
1422: }
1424: PetscErrorCode MatMultMax_SeqAIJ(Mat A,Vec xx,Vec yy)
1425: {
1426: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1427: PetscScalar *y;
1428: const PetscScalar *x;
1429: const MatScalar *aa;
1430: PetscErrorCode ierr;
1431: PetscInt m=A->rmap->n;
1432: const PetscInt *aj,*ii,*ridx=NULL;
1433: PetscInt n,i,nonzerorow=0;
1434: PetscScalar sum;
1435: PetscBool usecprow=a->compressedrow.use;
1437: #if defined(PETSC_HAVE_PRAGMA_DISJOINT)
1438: #pragma disjoint(*x,*y,*aa)
1439: #endif
1442: VecGetArrayRead(xx,&x);
1443: VecGetArray(yy,&y);
1444: if (usecprow) { /* use compressed row format */
1445: m = a->compressedrow.nrows;
1446: ii = a->compressedrow.i;
1447: ridx = a->compressedrow.rindex;
1448: for (i=0; i<m; i++) {
1449: n = ii[i+1] - ii[i];
1450: aj = a->j + ii[i];
1451: aa = a->a + ii[i];
1452: sum = 0.0;
1453: nonzerorow += (n>0);
1454: PetscSparseDenseMaxDot(sum,x,aa,aj,n);
1455: /* for (j=0; j<n; j++) sum += (*aa++)*x[*aj++]; */
1456: y[*ridx++] = sum;
1457: }
1458: } else { /* do not use compressed row format */
1459: ii = a->i;
1460: for (i=0; i<m; i++) {
1461: n = ii[i+1] - ii[i];
1462: aj = a->j + ii[i];
1463: aa = a->a + ii[i];
1464: sum = 0.0;
1465: nonzerorow += (n>0);
1466: PetscSparseDenseMaxDot(sum,x,aa,aj,n);
1467: y[i] = sum;
1468: }
1469: }
1470: PetscLogFlops(2.0*a->nz - nonzerorow);
1471: VecRestoreArrayRead(xx,&x);
1472: VecRestoreArray(yy,&y);
1473: return(0);
1474: }
1476: PetscErrorCode MatMultAddMax_SeqAIJ(Mat A,Vec xx,Vec yy,Vec zz)
1477: {
1478: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1479: PetscScalar *y,*z;
1480: const PetscScalar *x;
1481: const MatScalar *aa;
1482: PetscErrorCode ierr;
1483: PetscInt m = A->rmap->n,*aj,*ii;
1484: PetscInt n,i,*ridx=NULL;
1485: PetscScalar sum;
1486: PetscBool usecprow=a->compressedrow.use;
1489: VecGetArrayRead(xx,&x);
1490: VecGetArrayPair(yy,zz,&y,&z);
1491: if (usecprow) { /* use compressed row format */
1492: if (zz != yy) {
1493: PetscArraycpy(z,y,m);
1494: }
1495: m = a->compressedrow.nrows;
1496: ii = a->compressedrow.i;
1497: ridx = a->compressedrow.rindex;
1498: for (i=0; i<m; i++) {
1499: n = ii[i+1] - ii[i];
1500: aj = a->j + ii[i];
1501: aa = a->a + ii[i];
1502: sum = y[*ridx];
1503: PetscSparseDenseMaxDot(sum,x,aa,aj,n);
1504: z[*ridx++] = sum;
1505: }
1506: } else { /* do not use compressed row format */
1507: ii = a->i;
1508: for (i=0; i<m; i++) {
1509: n = ii[i+1] - ii[i];
1510: aj = a->j + ii[i];
1511: aa = a->a + ii[i];
1512: sum = y[i];
1513: PetscSparseDenseMaxDot(sum,x,aa,aj,n);
1514: z[i] = sum;
1515: }
1516: }
1517: PetscLogFlops(2.0*a->nz);
1518: VecRestoreArrayRead(xx,&x);
1519: VecRestoreArrayPair(yy,zz,&y,&z);
1520: return(0);
1521: }
1523: #include <../src/mat/impls/aij/seq/ftn-kernels/fmultadd.h>
1524: PetscErrorCode MatMultAdd_SeqAIJ(Mat A,Vec xx,Vec yy,Vec zz)
1525: {
1526: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1527: PetscScalar *y,*z;
1528: const PetscScalar *x;
1529: const MatScalar *aa;
1530: PetscErrorCode ierr;
1531: const PetscInt *aj,*ii,*ridx=NULL;
1532: PetscInt m = A->rmap->n,n,i;
1533: PetscScalar sum;
1534: PetscBool usecprow=a->compressedrow.use;
1537: VecGetArrayRead(xx,&x);
1538: VecGetArrayPair(yy,zz,&y,&z);
1539: if (usecprow) { /* use compressed row format */
1540: if (zz != yy) {
1541: PetscArraycpy(z,y,m);
1542: }
1543: m = a->compressedrow.nrows;
1544: ii = a->compressedrow.i;
1545: ridx = a->compressedrow.rindex;
1546: for (i=0; i<m; i++) {
1547: n = ii[i+1] - ii[i];
1548: aj = a->j + ii[i];
1549: aa = a->a + ii[i];
1550: sum = y[*ridx];
1551: PetscSparseDensePlusDot(sum,x,aa,aj,n);
1552: z[*ridx++] = sum;
1553: }
1554: } else { /* do not use compressed row format */
1555: ii = a->i;
1556: #if defined(PETSC_USE_FORTRAN_KERNEL_MULTADDAIJ)
1557: aj = a->j;
1558: aa = a->a;
1559: fortranmultaddaij_(&m,x,ii,aj,aa,y,z);
1560: #else
1561: for (i=0; i<m; i++) {
1562: n = ii[i+1] - ii[i];
1563: aj = a->j + ii[i];
1564: aa = a->a + ii[i];
1565: sum = y[i];
1566: PetscSparseDensePlusDot(sum,x,aa,aj,n);
1567: z[i] = sum;
1568: }
1569: #endif
1570: }
1571: PetscLogFlops(2.0*a->nz);
1572: VecRestoreArrayRead(xx,&x);
1573: VecRestoreArrayPair(yy,zz,&y,&z);
1574: return(0);
1575: }
1577: /*
1578: Adds diagonal pointers to sparse matrix structure.
1579: */
1580: PetscErrorCode MatMarkDiagonal_SeqAIJ(Mat A)
1581: {
1582: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1584: PetscInt i,j,m = A->rmap->n;
1587: if (!a->diag) {
1588: PetscMalloc1(m,&a->diag);
1589: PetscLogObjectMemory((PetscObject)A, m*sizeof(PetscInt));
1590: }
1591: for (i=0; i<A->rmap->n; i++) {
1592: a->diag[i] = a->i[i+1];
1593: for (j=a->i[i]; j<a->i[i+1]; j++) {
1594: if (a->j[j] == i) {
1595: a->diag[i] = j;
1596: break;
1597: }
1598: }
1599: }
1600: return(0);
1601: }
1603: PetscErrorCode MatShift_SeqAIJ(Mat A,PetscScalar v)
1604: {
1605: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1606: const PetscInt *diag = (const PetscInt*)a->diag;
1607: const PetscInt *ii = (const PetscInt*) a->i;
1608: PetscInt i,*mdiag = NULL;
1609: PetscErrorCode ierr;
1610: PetscInt cnt = 0; /* how many diagonals are missing */
1613: if (!A->preallocated || !a->nz) {
1614: MatSeqAIJSetPreallocation(A,1,NULL);
1615: MatShift_Basic(A,v);
1616: return(0);
1617: }
1619: if (a->diagonaldense) {
1620: cnt = 0;
1621: } else {
1622: PetscCalloc1(A->rmap->n,&mdiag);
1623: for (i=0; i<A->rmap->n; i++) {
1624: if (diag[i] >= ii[i+1]) {
1625: cnt++;
1626: mdiag[i] = 1;
1627: }
1628: }
1629: }
1630: if (!cnt) {
1631: MatShift_Basic(A,v);
1632: } else {
1633: PetscScalar *olda = a->a; /* preserve pointers to current matrix nonzeros structure and values */
1634: PetscInt *oldj = a->j, *oldi = a->i;
1635: PetscBool singlemalloc = a->singlemalloc,free_a = a->free_a,free_ij = a->free_ij;
1637: a->a = NULL;
1638: a->j = NULL;
1639: a->i = NULL;
1640: /* increase the values in imax for each row where a diagonal is being inserted then reallocate the matrix data structures */
1641: for (i=0; i<A->rmap->n; i++) {
1642: a->imax[i] += mdiag[i];
1643: a->imax[i] = PetscMin(a->imax[i],A->cmap->n);
1644: }
1645: MatSeqAIJSetPreallocation_SeqAIJ(A,0,a->imax);
1647: /* copy old values into new matrix data structure */
1648: for (i=0; i<A->rmap->n; i++) {
1649: MatSetValues(A,1,&i,a->imax[i] - mdiag[i],&oldj[oldi[i]],&olda[oldi[i]],ADD_VALUES);
1650: if (i < A->cmap->n) {
1651: MatSetValue(A,i,i,v,ADD_VALUES);
1652: }
1653: }
1654: MatAssemblyBegin(A,MAT_FINAL_ASSEMBLY);
1655: MatAssemblyEnd(A,MAT_FINAL_ASSEMBLY);
1656: if (singlemalloc) {
1657: PetscFree3(olda,oldj,oldi);
1658: } else {
1659: if (free_a) {PetscFree(olda);}
1660: if (free_ij) {PetscFree(oldj);}
1661: if (free_ij) {PetscFree(oldi);}
1662: }
1663: }
1664: PetscFree(mdiag);
1665: a->diagonaldense = PETSC_TRUE;
1666: return(0);
1667: }
1669: /*
1670: Checks for missing diagonals
1671: */
1672: PetscErrorCode MatMissingDiagonal_SeqAIJ(Mat A,PetscBool *missing,PetscInt *d)
1673: {
1674: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1675: PetscInt *diag,*ii = a->i,i;
1679: *missing = PETSC_FALSE;
1680: if (A->rmap->n > 0 && !ii) {
1681: *missing = PETSC_TRUE;
1682: if (d) *d = 0;
1683: PetscInfo(A,"Matrix has no entries therefore is missing diagonal\n");
1684: } else {
1685: PetscInt n;
1686: n = PetscMin(A->rmap->n, A->cmap->n);
1687: diag = a->diag;
1688: for (i=0; i<n; i++) {
1689: if (diag[i] >= ii[i+1]) {
1690: *missing = PETSC_TRUE;
1691: if (d) *d = i;
1692: PetscInfo1(A,"Matrix is missing diagonal number %D\n",i);
1693: break;
1694: }
1695: }
1696: }
1697: return(0);
1698: }
1700: #include <petscblaslapack.h>
1701: #include <petsc/private/kernels/blockinvert.h>
1703: /*
1704: Note that values is allocated externally by the PC and then passed into this routine
1705: */
1706: PetscErrorCode MatInvertVariableBlockDiagonal_SeqAIJ(Mat A,PetscInt nblocks,const PetscInt *bsizes,PetscScalar *diag)
1707: {
1708: PetscErrorCode ierr;
1709: PetscInt n = A->rmap->n, i, ncnt = 0, *indx,j,bsizemax = 0,*v_pivots;
1710: PetscBool allowzeropivot,zeropivotdetected=PETSC_FALSE;
1711: const PetscReal shift = 0.0;
1712: PetscInt ipvt[5];
1713: PetscScalar work[25],*v_work;
1716: allowzeropivot = PetscNot(A->erroriffailure);
1717: for (i=0; i<nblocks; i++) ncnt += bsizes[i];
1718: if (ncnt != n) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Total blocksizes %D doesn't match number matrix rows %D",ncnt,n);
1719: for (i=0; i<nblocks; i++) {
1720: bsizemax = PetscMax(bsizemax,bsizes[i]);
1721: }
1722: PetscMalloc1(bsizemax,&indx);
1723: if (bsizemax > 7) {
1724: PetscMalloc2(bsizemax,&v_work,bsizemax,&v_pivots);
1725: }
1726: ncnt = 0;
1727: for (i=0; i<nblocks; i++) {
1728: for (j=0; j<bsizes[i]; j++) indx[j] = ncnt+j;
1729: MatGetValues(A,bsizes[i],indx,bsizes[i],indx,diag);
1730: switch (bsizes[i]) {
1731: case 1:
1732: *diag = 1.0/(*diag);
1733: break;
1734: case 2:
1735: PetscKernel_A_gets_inverse_A_2(diag,shift,allowzeropivot,&zeropivotdetected);
1736: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
1737: PetscKernel_A_gets_transpose_A_2(diag);
1738: break;
1739: case 3:
1740: PetscKernel_A_gets_inverse_A_3(diag,shift,allowzeropivot,&zeropivotdetected);
1741: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
1742: PetscKernel_A_gets_transpose_A_3(diag);
1743: break;
1744: case 4:
1745: PetscKernel_A_gets_inverse_A_4(diag,shift,allowzeropivot,&zeropivotdetected);
1746: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
1747: PetscKernel_A_gets_transpose_A_4(diag);
1748: break;
1749: case 5:
1750: PetscKernel_A_gets_inverse_A_5(diag,ipvt,work,shift,allowzeropivot,&zeropivotdetected);
1751: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
1752: PetscKernel_A_gets_transpose_A_5(diag);
1753: break;
1754: case 6:
1755: PetscKernel_A_gets_inverse_A_6(diag,shift,allowzeropivot,&zeropivotdetected);
1756: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
1757: PetscKernel_A_gets_transpose_A_6(diag);
1758: break;
1759: case 7:
1760: PetscKernel_A_gets_inverse_A_7(diag,shift,allowzeropivot,&zeropivotdetected);
1761: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
1762: PetscKernel_A_gets_transpose_A_7(diag);
1763: break;
1764: default:
1765: PetscKernel_A_gets_inverse_A(bsizes[i],diag,v_pivots,v_work,allowzeropivot,&zeropivotdetected);
1766: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
1767: PetscKernel_A_gets_transpose_A_N(diag,bsizes[i]);
1768: }
1769: ncnt += bsizes[i];
1770: diag += bsizes[i]*bsizes[i];
1771: }
1772: if (bsizemax > 7) {
1773: PetscFree2(v_work,v_pivots);
1774: }
1775: PetscFree(indx);
1776: return(0);
1777: }
1779: /*
1780: Negative shift indicates do not generate an error if there is a zero diagonal, just invert it anyways
1781: */
1782: PetscErrorCode MatInvertDiagonal_SeqAIJ(Mat A,PetscScalar omega,PetscScalar fshift)
1783: {
1784: Mat_SeqAIJ *a = (Mat_SeqAIJ*) A->data;
1786: PetscInt i,*diag,m = A->rmap->n;
1787: MatScalar *v = a->a;
1788: PetscScalar *idiag,*mdiag;
1791: if (a->idiagvalid) return(0);
1792: MatMarkDiagonal_SeqAIJ(A);
1793: diag = a->diag;
1794: if (!a->idiag) {
1795: PetscMalloc3(m,&a->idiag,m,&a->mdiag,m,&a->ssor_work);
1796: PetscLogObjectMemory((PetscObject)A, 3*m*sizeof(PetscScalar));
1797: v = a->a;
1798: }
1799: mdiag = a->mdiag;
1800: idiag = a->idiag;
1802: if (omega == 1.0 && PetscRealPart(fshift) <= 0.0) {
1803: for (i=0; i<m; i++) {
1804: mdiag[i] = v[diag[i]];
1805: if (!PetscAbsScalar(mdiag[i])) { /* zero diagonal */
1806: if (PetscRealPart(fshift)) {
1807: PetscInfo1(A,"Zero diagonal on row %D\n",i);
1808: A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
1809: A->factorerror_zeropivot_value = 0.0;
1810: A->factorerror_zeropivot_row = i;
1811: } else SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_INCOMP,"Zero diagonal on row %D",i);
1812: }
1813: idiag[i] = 1.0/v[diag[i]];
1814: }
1815: PetscLogFlops(m);
1816: } else {
1817: for (i=0; i<m; i++) {
1818: mdiag[i] = v[diag[i]];
1819: idiag[i] = omega/(fshift + v[diag[i]]);
1820: }
1821: PetscLogFlops(2.0*m);
1822: }
1823: a->idiagvalid = PETSC_TRUE;
1824: return(0);
1825: }
1827: #include <../src/mat/impls/aij/seq/ftn-kernels/frelax.h>
1828: PetscErrorCode MatSOR_SeqAIJ(Mat A,Vec bb,PetscReal omega,MatSORType flag,PetscReal fshift,PetscInt its,PetscInt lits,Vec xx)
1829: {
1830: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1831: PetscScalar *x,d,sum,*t,scale;
1832: const MatScalar *v,*idiag=0,*mdiag;
1833: const PetscScalar *b, *bs,*xb, *ts;
1834: PetscErrorCode ierr;
1835: PetscInt n,m = A->rmap->n,i;
1836: const PetscInt *idx,*diag;
1839: its = its*lits;
1841: if (fshift != a->fshift || omega != a->omega) a->idiagvalid = PETSC_FALSE; /* must recompute idiag[] */
1842: if (!a->idiagvalid) {MatInvertDiagonal_SeqAIJ(A,omega,fshift);}
1843: a->fshift = fshift;
1844: a->omega = omega;
1846: diag = a->diag;
1847: t = a->ssor_work;
1848: idiag = a->idiag;
1849: mdiag = a->mdiag;
1851: VecGetArray(xx,&x);
1852: VecGetArrayRead(bb,&b);
1853: /* We count flops by assuming the upper triangular and lower triangular parts have the same number of nonzeros */
1854: if (flag == SOR_APPLY_UPPER) {
1855: /* apply (U + D/omega) to the vector */
1856: bs = b;
1857: for (i=0; i<m; i++) {
1858: d = fshift + mdiag[i];
1859: n = a->i[i+1] - diag[i] - 1;
1860: idx = a->j + diag[i] + 1;
1861: v = a->a + diag[i] + 1;
1862: sum = b[i]*d/omega;
1863: PetscSparseDensePlusDot(sum,bs,v,idx,n);
1864: x[i] = sum;
1865: }
1866: VecRestoreArray(xx,&x);
1867: VecRestoreArrayRead(bb,&b);
1868: PetscLogFlops(a->nz);
1869: return(0);
1870: }
1872: if (flag == SOR_APPLY_LOWER) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"SOR_APPLY_LOWER is not implemented");
1873: else if (flag & SOR_EISENSTAT) {
1874: /* Let A = L + U + D; where L is lower triangular,
1875: U is upper triangular, E = D/omega; This routine applies
1877: (L + E)^{-1} A (U + E)^{-1}
1879: to a vector efficiently using Eisenstat's trick.
1880: */
1881: scale = (2.0/omega) - 1.0;
1883: /* x = (E + U)^{-1} b */
1884: for (i=m-1; i>=0; i--) {
1885: n = a->i[i+1] - diag[i] - 1;
1886: idx = a->j + diag[i] + 1;
1887: v = a->a + diag[i] + 1;
1888: sum = b[i];
1889: PetscSparseDenseMinusDot(sum,x,v,idx,n);
1890: x[i] = sum*idiag[i];
1891: }
1893: /* t = b - (2*E - D)x */
1894: v = a->a;
1895: for (i=0; i<m; i++) t[i] = b[i] - scale*(v[*diag++])*x[i];
1897: /* t = (E + L)^{-1}t */
1898: ts = t;
1899: diag = a->diag;
1900: for (i=0; i<m; i++) {
1901: n = diag[i] - a->i[i];
1902: idx = a->j + a->i[i];
1903: v = a->a + a->i[i];
1904: sum = t[i];
1905: PetscSparseDenseMinusDot(sum,ts,v,idx,n);
1906: t[i] = sum*idiag[i];
1907: /* x = x + t */
1908: x[i] += t[i];
1909: }
1911: PetscLogFlops(6.0*m-1 + 2.0*a->nz);
1912: VecRestoreArray(xx,&x);
1913: VecRestoreArrayRead(bb,&b);
1914: return(0);
1915: }
1916: if (flag & SOR_ZERO_INITIAL_GUESS) {
1917: if (flag & SOR_FORWARD_SWEEP || flag & SOR_LOCAL_FORWARD_SWEEP) {
1918: for (i=0; i<m; i++) {
1919: n = diag[i] - a->i[i];
1920: idx = a->j + a->i[i];
1921: v = a->a + a->i[i];
1922: sum = b[i];
1923: PetscSparseDenseMinusDot(sum,x,v,idx,n);
1924: t[i] = sum;
1925: x[i] = sum*idiag[i];
1926: }
1927: xb = t;
1928: PetscLogFlops(a->nz);
1929: } else xb = b;
1930: if (flag & SOR_BACKWARD_SWEEP || flag & SOR_LOCAL_BACKWARD_SWEEP) {
1931: for (i=m-1; i>=0; i--) {
1932: n = a->i[i+1] - diag[i] - 1;
1933: idx = a->j + diag[i] + 1;
1934: v = a->a + diag[i] + 1;
1935: sum = xb[i];
1936: PetscSparseDenseMinusDot(sum,x,v,idx,n);
1937: if (xb == b) {
1938: x[i] = sum*idiag[i];
1939: } else {
1940: x[i] = (1-omega)*x[i] + sum*idiag[i]; /* omega in idiag */
1941: }
1942: }
1943: PetscLogFlops(a->nz); /* assumes 1/2 in upper */
1944: }
1945: its--;
1946: }
1947: while (its--) {
1948: if (flag & SOR_FORWARD_SWEEP || flag & SOR_LOCAL_FORWARD_SWEEP) {
1949: for (i=0; i<m; i++) {
1950: /* lower */
1951: n = diag[i] - a->i[i];
1952: idx = a->j + a->i[i];
1953: v = a->a + a->i[i];
1954: sum = b[i];
1955: PetscSparseDenseMinusDot(sum,x,v,idx,n);
1956: t[i] = sum; /* save application of the lower-triangular part */
1957: /* upper */
1958: n = a->i[i+1] - diag[i] - 1;
1959: idx = a->j + diag[i] + 1;
1960: v = a->a + diag[i] + 1;
1961: PetscSparseDenseMinusDot(sum,x,v,idx,n);
1962: x[i] = (1. - omega)*x[i] + sum*idiag[i]; /* omega in idiag */
1963: }
1964: xb = t;
1965: PetscLogFlops(2.0*a->nz);
1966: } else xb = b;
1967: if (flag & SOR_BACKWARD_SWEEP || flag & SOR_LOCAL_BACKWARD_SWEEP) {
1968: for (i=m-1; i>=0; i--) {
1969: sum = xb[i];
1970: if (xb == b) {
1971: /* whole matrix (no checkpointing available) */
1972: n = a->i[i+1] - a->i[i];
1973: idx = a->j + a->i[i];
1974: v = a->a + a->i[i];
1975: PetscSparseDenseMinusDot(sum,x,v,idx,n);
1976: x[i] = (1. - omega)*x[i] + (sum + mdiag[i]*x[i])*idiag[i];
1977: } else { /* lower-triangular part has been saved, so only apply upper-triangular */
1978: n = a->i[i+1] - diag[i] - 1;
1979: idx = a->j + diag[i] + 1;
1980: v = a->a + diag[i] + 1;
1981: PetscSparseDenseMinusDot(sum,x,v,idx,n);
1982: x[i] = (1. - omega)*x[i] + sum*idiag[i]; /* omega in idiag */
1983: }
1984: }
1985: if (xb == b) {
1986: PetscLogFlops(2.0*a->nz);
1987: } else {
1988: PetscLogFlops(a->nz); /* assumes 1/2 in upper */
1989: }
1990: }
1991: }
1992: VecRestoreArray(xx,&x);
1993: VecRestoreArrayRead(bb,&b);
1994: return(0);
1995: }
1998: PetscErrorCode MatGetInfo_SeqAIJ(Mat A,MatInfoType flag,MatInfo *info)
1999: {
2000: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
2003: info->block_size = 1.0;
2004: info->nz_allocated = a->maxnz;
2005: info->nz_used = a->nz;
2006: info->nz_unneeded = (a->maxnz - a->nz);
2007: info->assemblies = A->num_ass;
2008: info->mallocs = A->info.mallocs;
2009: info->memory = ((PetscObject)A)->mem;
2010: if (A->factortype) {
2011: info->fill_ratio_given = A->info.fill_ratio_given;
2012: info->fill_ratio_needed = A->info.fill_ratio_needed;
2013: info->factor_mallocs = A->info.factor_mallocs;
2014: } else {
2015: info->fill_ratio_given = 0;
2016: info->fill_ratio_needed = 0;
2017: info->factor_mallocs = 0;
2018: }
2019: return(0);
2020: }
2022: PetscErrorCode MatZeroRows_SeqAIJ(Mat A,PetscInt N,const PetscInt rows[],PetscScalar diag,Vec x,Vec b)
2023: {
2024: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
2025: PetscInt i,m = A->rmap->n - 1;
2026: PetscErrorCode ierr;
2027: const PetscScalar *xx;
2028: PetscScalar *bb;
2029: PetscInt d = 0;
2032: if (x && b) {
2033: VecGetArrayRead(x,&xx);
2034: VecGetArray(b,&bb);
2035: for (i=0; i<N; i++) {
2036: if (rows[i] < 0 || rows[i] > m) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"row %D out of range", rows[i]);
2037: if (rows[i] >= A->cmap->n) continue;
2038: bb[rows[i]] = diag*xx[rows[i]];
2039: }
2040: VecRestoreArrayRead(x,&xx);
2041: VecRestoreArray(b,&bb);
2042: }
2044: if (a->keepnonzeropattern) {
2045: for (i=0; i<N; i++) {
2046: if (rows[i] < 0 || rows[i] > m) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"row %D out of range", rows[i]);
2047: PetscArrayzero(&a->a[a->i[rows[i]]],a->ilen[rows[i]]);
2048: }
2049: if (diag != 0.0) {
2050: for (i=0; i<N; i++) {
2051: d = rows[i];
2052: if (rows[i] >= A->cmap->n) continue;
2053: if (a->diag[d] >= a->i[d+1]) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Matrix is missing diagonal entry in the zeroed row %D",d);
2054: }
2055: for (i=0; i<N; i++) {
2056: if (rows[i] >= A->cmap->n) continue;
2057: a->a[a->diag[rows[i]]] = diag;
2058: }
2059: }
2060: } else {
2061: if (diag != 0.0) {
2062: for (i=0; i<N; i++) {
2063: if (rows[i] < 0 || rows[i] > m) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"row %D out of range", rows[i]);
2064: if (a->ilen[rows[i]] > 0) {
2065: if (rows[i] >= A->cmap->n) {
2066: a->ilen[rows[i]] = 0;
2067: } else {
2068: a->ilen[rows[i]] = 1;
2069: a->a[a->i[rows[i]]] = diag;
2070: a->j[a->i[rows[i]]] = rows[i];
2071: }
2072: } else if (rows[i] < A->cmap->n) { /* in case row was completely empty */
2073: MatSetValues_SeqAIJ(A,1,&rows[i],1,&rows[i],&diag,INSERT_VALUES);
2074: }
2075: }
2076: } else {
2077: for (i=0; i<N; i++) {
2078: if (rows[i] < 0 || rows[i] > m) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"row %D out of range", rows[i]);
2079: a->ilen[rows[i]] = 0;
2080: }
2081: }
2082: A->nonzerostate++;
2083: }
2084: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
2085: if (A->offloadmask != PETSC_OFFLOAD_UNALLOCATED) A->offloadmask = PETSC_OFFLOAD_CPU;
2086: #endif
2087: (*A->ops->assemblyend)(A,MAT_FINAL_ASSEMBLY);
2088: return(0);
2089: }
2091: PetscErrorCode MatZeroRowsColumns_SeqAIJ(Mat A,PetscInt N,const PetscInt rows[],PetscScalar diag,Vec x,Vec b)
2092: {
2093: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
2094: PetscInt i,j,m = A->rmap->n - 1,d = 0;
2095: PetscErrorCode ierr;
2096: PetscBool missing,*zeroed,vecs = PETSC_FALSE;
2097: const PetscScalar *xx;
2098: PetscScalar *bb;
2101: if (x && b) {
2102: VecGetArrayRead(x,&xx);
2103: VecGetArray(b,&bb);
2104: vecs = PETSC_TRUE;
2105: }
2106: PetscCalloc1(A->rmap->n,&zeroed);
2107: for (i=0; i<N; i++) {
2108: if (rows[i] < 0 || rows[i] > m) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"row %D out of range", rows[i]);
2109: PetscArrayzero(&a->a[a->i[rows[i]]],a->ilen[rows[i]]);
2111: zeroed[rows[i]] = PETSC_TRUE;
2112: }
2113: for (i=0; i<A->rmap->n; i++) {
2114: if (!zeroed[i]) {
2115: for (j=a->i[i]; j<a->i[i+1]; j++) {
2116: if (a->j[j] < A->rmap->n && zeroed[a->j[j]]) {
2117: if (vecs) bb[i] -= a->a[j]*xx[a->j[j]];
2118: a->a[j] = 0.0;
2119: }
2120: }
2121: } else if (vecs && i < A->cmap->N) bb[i] = diag*xx[i];
2122: }
2123: if (x && b) {
2124: VecRestoreArrayRead(x,&xx);
2125: VecRestoreArray(b,&bb);
2126: }
2127: PetscFree(zeroed);
2128: if (diag != 0.0) {
2129: MatMissingDiagonal_SeqAIJ(A,&missing,&d);
2130: if (missing) {
2131: for (i=0; i<N; i++) {
2132: if (rows[i] >= A->cmap->N) continue;
2133: if (a->nonew && rows[i] >= d) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Matrix is missing diagonal entry in row %D (%D)",d,rows[i]);
2134: MatSetValues_SeqAIJ(A,1,&rows[i],1,&rows[i],&diag,INSERT_VALUES);
2135: }
2136: } else {
2137: for (i=0; i<N; i++) {
2138: a->a[a->diag[rows[i]]] = diag;
2139: }
2140: }
2141: }
2142: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
2143: if (A->offloadmask != PETSC_OFFLOAD_UNALLOCATED) A->offloadmask = PETSC_OFFLOAD_CPU;
2144: #endif
2145: (*A->ops->assemblyend)(A,MAT_FINAL_ASSEMBLY);
2146: return(0);
2147: }
2149: PetscErrorCode MatGetRow_SeqAIJ(Mat A,PetscInt row,PetscInt *nz,PetscInt **idx,PetscScalar **v)
2150: {
2151: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
2152: PetscInt *itmp;
2155: if (row < 0 || row >= A->rmap->n) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Row %D out of range",row);
2157: *nz = a->i[row+1] - a->i[row];
2158: if (v) *v = a->a + a->i[row];
2159: if (idx) {
2160: itmp = a->j + a->i[row];
2161: if (*nz) *idx = itmp;
2162: else *idx = 0;
2163: }
2164: return(0);
2165: }
2167: /* remove this function? */
2168: PetscErrorCode MatRestoreRow_SeqAIJ(Mat A,PetscInt row,PetscInt *nz,PetscInt **idx,PetscScalar **v)
2169: {
2171: return(0);
2172: }
2174: PetscErrorCode MatNorm_SeqAIJ(Mat A,NormType type,PetscReal *nrm)
2175: {
2176: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
2177: MatScalar *v = a->a;
2178: PetscReal sum = 0.0;
2180: PetscInt i,j;
2183: if (type == NORM_FROBENIUS) {
2184: #if defined(PETSC_USE_REAL___FP16)
2185: PetscBLASInt one = 1,nz = a->nz;
2186: *nrm = BLASnrm2_(&nz,v,&one);
2187: #else
2188: for (i=0; i<a->nz; i++) {
2189: sum += PetscRealPart(PetscConj(*v)*(*v)); v++;
2190: }
2191: *nrm = PetscSqrtReal(sum);
2192: #endif
2193: PetscLogFlops(2*a->nz);
2194: } else if (type == NORM_1) {
2195: PetscReal *tmp;
2196: PetscInt *jj = a->j;
2197: PetscCalloc1(A->cmap->n+1,&tmp);
2198: *nrm = 0.0;
2199: for (j=0; j<a->nz; j++) {
2200: tmp[*jj++] += PetscAbsScalar(*v); v++;
2201: }
2202: for (j=0; j<A->cmap->n; j++) {
2203: if (tmp[j] > *nrm) *nrm = tmp[j];
2204: }
2205: PetscFree(tmp);
2206: PetscLogFlops(PetscMax(a->nz-1,0));
2207: } else if (type == NORM_INFINITY) {
2208: *nrm = 0.0;
2209: for (j=0; j<A->rmap->n; j++) {
2210: v = a->a + a->i[j];
2211: sum = 0.0;
2212: for (i=0; i<a->i[j+1]-a->i[j]; i++) {
2213: sum += PetscAbsScalar(*v); v++;
2214: }
2215: if (sum > *nrm) *nrm = sum;
2216: }
2217: PetscLogFlops(PetscMax(a->nz-1,0));
2218: } else SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"No support for two norm");
2219: return(0);
2220: }
2222: /* Merged from MatGetSymbolicTranspose_SeqAIJ() - replace MatGetSymbolicTranspose_SeqAIJ()? */
2223: PetscErrorCode MatTransposeSymbolic_SeqAIJ(Mat A,Mat *B)
2224: {
2226: PetscInt i,j,anzj;
2227: Mat_SeqAIJ *a=(Mat_SeqAIJ*)A->data,*b;
2228: PetscInt an=A->cmap->N,am=A->rmap->N;
2229: PetscInt *ati,*atj,*atfill,*ai=a->i,*aj=a->j;
2232: /* Allocate space for symbolic transpose info and work array */
2233: PetscCalloc1(an+1,&ati);
2234: PetscMalloc1(ai[am],&atj);
2235: PetscMalloc1(an,&atfill);
2237: /* Walk through aj and count ## of non-zeros in each row of A^T. */
2238: /* Note: offset by 1 for fast conversion into csr format. */
2239: for (i=0;i<ai[am];i++) ati[aj[i]+1] += 1;
2240: /* Form ati for csr format of A^T. */
2241: for (i=0;i<an;i++) ati[i+1] += ati[i];
2243: /* Copy ati into atfill so we have locations of the next free space in atj */
2244: PetscArraycpy(atfill,ati,an);
2246: /* Walk through A row-wise and mark nonzero entries of A^T. */
2247: for (i=0;i<am;i++) {
2248: anzj = ai[i+1] - ai[i];
2249: for (j=0;j<anzj;j++) {
2250: atj[atfill[*aj]] = i;
2251: atfill[*aj++] += 1;
2252: }
2253: }
2255: /* Clean up temporary space and complete requests. */
2256: PetscFree(atfill);
2257: MatCreateSeqAIJWithArrays(PetscObjectComm((PetscObject)A),an,am,ati,atj,NULL,B);
2258: MatSetBlockSizes(*B,PetscAbs(A->cmap->bs),PetscAbs(A->rmap->bs));
2259: MatSetType(*B,((PetscObject)A)->type_name);
2261: b = (Mat_SeqAIJ*)((*B)->data);
2262: b->free_a = PETSC_FALSE;
2263: b->free_ij = PETSC_TRUE;
2264: b->nonew = 0;
2265: return(0);
2266: }
2268: PetscErrorCode MatIsTranspose_SeqAIJ(Mat A,Mat B,PetscReal tol,PetscBool *f)
2269: {
2270: Mat_SeqAIJ *aij = (Mat_SeqAIJ*) A->data,*bij = (Mat_SeqAIJ*) B->data;
2271: PetscInt *adx,*bdx,*aii,*bii,*aptr,*bptr;
2272: MatScalar *va,*vb;
2274: PetscInt ma,na,mb,nb, i;
2277: MatGetSize(A,&ma,&na);
2278: MatGetSize(B,&mb,&nb);
2279: if (ma!=nb || na!=mb) {
2280: *f = PETSC_FALSE;
2281: return(0);
2282: }
2283: aii = aij->i; bii = bij->i;
2284: adx = aij->j; bdx = bij->j;
2285: va = aij->a; vb = bij->a;
2286: PetscMalloc1(ma,&aptr);
2287: PetscMalloc1(mb,&bptr);
2288: for (i=0; i<ma; i++) aptr[i] = aii[i];
2289: for (i=0; i<mb; i++) bptr[i] = bii[i];
2291: *f = PETSC_TRUE;
2292: for (i=0; i<ma; i++) {
2293: while (aptr[i]<aii[i+1]) {
2294: PetscInt idc,idr;
2295: PetscScalar vc,vr;
2296: /* column/row index/value */
2297: idc = adx[aptr[i]];
2298: idr = bdx[bptr[idc]];
2299: vc = va[aptr[i]];
2300: vr = vb[bptr[idc]];
2301: if (i!=idr || PetscAbsScalar(vc-vr) > tol) {
2302: *f = PETSC_FALSE;
2303: goto done;
2304: } else {
2305: aptr[i]++;
2306: if (B || i!=idc) bptr[idc]++;
2307: }
2308: }
2309: }
2310: done:
2311: PetscFree(aptr);
2312: PetscFree(bptr);
2313: return(0);
2314: }
2316: PetscErrorCode MatIsHermitianTranspose_SeqAIJ(Mat A,Mat B,PetscReal tol,PetscBool *f)
2317: {
2318: Mat_SeqAIJ *aij = (Mat_SeqAIJ*) A->data,*bij = (Mat_SeqAIJ*) B->data;
2319: PetscInt *adx,*bdx,*aii,*bii,*aptr,*bptr;
2320: MatScalar *va,*vb;
2322: PetscInt ma,na,mb,nb, i;
2325: MatGetSize(A,&ma,&na);
2326: MatGetSize(B,&mb,&nb);
2327: if (ma!=nb || na!=mb) {
2328: *f = PETSC_FALSE;
2329: return(0);
2330: }
2331: aii = aij->i; bii = bij->i;
2332: adx = aij->j; bdx = bij->j;
2333: va = aij->a; vb = bij->a;
2334: PetscMalloc1(ma,&aptr);
2335: PetscMalloc1(mb,&bptr);
2336: for (i=0; i<ma; i++) aptr[i] = aii[i];
2337: for (i=0; i<mb; i++) bptr[i] = bii[i];
2339: *f = PETSC_TRUE;
2340: for (i=0; i<ma; i++) {
2341: while (aptr[i]<aii[i+1]) {
2342: PetscInt idc,idr;
2343: PetscScalar vc,vr;
2344: /* column/row index/value */
2345: idc = adx[aptr[i]];
2346: idr = bdx[bptr[idc]];
2347: vc = va[aptr[i]];
2348: vr = vb[bptr[idc]];
2349: if (i!=idr || PetscAbsScalar(vc-PetscConj(vr)) > tol) {
2350: *f = PETSC_FALSE;
2351: goto done;
2352: } else {
2353: aptr[i]++;
2354: if (B || i!=idc) bptr[idc]++;
2355: }
2356: }
2357: }
2358: done:
2359: PetscFree(aptr);
2360: PetscFree(bptr);
2361: return(0);
2362: }
2364: PetscErrorCode MatIsSymmetric_SeqAIJ(Mat A,PetscReal tol,PetscBool *f)
2365: {
2369: MatIsTranspose_SeqAIJ(A,A,tol,f);
2370: return(0);
2371: }
2373: PetscErrorCode MatIsHermitian_SeqAIJ(Mat A,PetscReal tol,PetscBool *f)
2374: {
2378: MatIsHermitianTranspose_SeqAIJ(A,A,tol,f);
2379: return(0);
2380: }
2382: PetscErrorCode MatDiagonalScale_SeqAIJ(Mat A,Vec ll,Vec rr)
2383: {
2384: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
2385: const PetscScalar *l,*r;
2386: PetscScalar x;
2387: MatScalar *v;
2388: PetscErrorCode ierr;
2389: PetscInt i,j,m = A->rmap->n,n = A->cmap->n,M,nz = a->nz;
2390: const PetscInt *jj;
2393: if (ll) {
2394: /* The local size is used so that VecMPI can be passed to this routine
2395: by MatDiagonalScale_MPIAIJ */
2396: VecGetLocalSize(ll,&m);
2397: if (m != A->rmap->n) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Left scaling vector wrong length");
2398: VecGetArrayRead(ll,&l);
2399: v = a->a;
2400: for (i=0; i<m; i++) {
2401: x = l[i];
2402: M = a->i[i+1] - a->i[i];
2403: for (j=0; j<M; j++) (*v++) *= x;
2404: }
2405: VecRestoreArrayRead(ll,&l);
2406: PetscLogFlops(nz);
2407: }
2408: if (rr) {
2409: VecGetLocalSize(rr,&n);
2410: if (n != A->cmap->n) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Right scaling vector wrong length");
2411: VecGetArrayRead(rr,&r);
2412: v = a->a; jj = a->j;
2413: for (i=0; i<nz; i++) (*v++) *= r[*jj++];
2414: VecRestoreArrayRead(rr,&r);
2415: PetscLogFlops(nz);
2416: }
2417: MatSeqAIJInvalidateDiagonal(A);
2418: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
2419: if (A->offloadmask != PETSC_OFFLOAD_UNALLOCATED) A->offloadmask = PETSC_OFFLOAD_CPU;
2420: #endif
2421: return(0);
2422: }
2424: PetscErrorCode MatCreateSubMatrix_SeqAIJ(Mat A,IS isrow,IS iscol,PetscInt csize,MatReuse scall,Mat *B)
2425: {
2426: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data,*c;
2428: PetscInt *smap,i,k,kstart,kend,oldcols = A->cmap->n,*lens;
2429: PetscInt row,mat_i,*mat_j,tcol,first,step,*mat_ilen,sum,lensi;
2430: const PetscInt *irow,*icol;
2431: PetscInt nrows,ncols;
2432: PetscInt *starts,*j_new,*i_new,*aj = a->j,*ai = a->i,ii,*ailen = a->ilen;
2433: MatScalar *a_new,*mat_a;
2434: Mat C;
2435: PetscBool stride;
2439: ISGetIndices(isrow,&irow);
2440: ISGetLocalSize(isrow,&nrows);
2441: ISGetLocalSize(iscol,&ncols);
2443: PetscObjectTypeCompare((PetscObject)iscol,ISSTRIDE,&stride);
2444: if (stride) {
2445: ISStrideGetInfo(iscol,&first,&step);
2446: } else {
2447: first = 0;
2448: step = 0;
2449: }
2450: if (stride && step == 1) {
2451: /* special case of contiguous rows */
2452: PetscMalloc2(nrows,&lens,nrows,&starts);
2453: /* loop over new rows determining lens and starting points */
2454: for (i=0; i<nrows; i++) {
2455: kstart = ai[irow[i]];
2456: kend = kstart + ailen[irow[i]];
2457: starts[i] = kstart;
2458: for (k=kstart; k<kend; k++) {
2459: if (aj[k] >= first) {
2460: starts[i] = k;
2461: break;
2462: }
2463: }
2464: sum = 0;
2465: while (k < kend) {
2466: if (aj[k++] >= first+ncols) break;
2467: sum++;
2468: }
2469: lens[i] = sum;
2470: }
2471: /* create submatrix */
2472: if (scall == MAT_REUSE_MATRIX) {
2473: PetscInt n_cols,n_rows;
2474: MatGetSize(*B,&n_rows,&n_cols);
2475: if (n_rows != nrows || n_cols != ncols) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Reused submatrix wrong size");
2476: MatZeroEntries(*B);
2477: C = *B;
2478: } else {
2479: PetscInt rbs,cbs;
2480: MatCreate(PetscObjectComm((PetscObject)A),&C);
2481: MatSetSizes(C,nrows,ncols,PETSC_DETERMINE,PETSC_DETERMINE);
2482: ISGetBlockSize(isrow,&rbs);
2483: ISGetBlockSize(iscol,&cbs);
2484: MatSetBlockSizes(C,rbs,cbs);
2485: MatSetType(C,((PetscObject)A)->type_name);
2486: MatSeqAIJSetPreallocation_SeqAIJ(C,0,lens);
2487: }
2488: c = (Mat_SeqAIJ*)C->data;
2490: /* loop over rows inserting into submatrix */
2491: a_new = c->a;
2492: j_new = c->j;
2493: i_new = c->i;
2495: for (i=0; i<nrows; i++) {
2496: ii = starts[i];
2497: lensi = lens[i];
2498: for (k=0; k<lensi; k++) {
2499: *j_new++ = aj[ii+k] - first;
2500: }
2501: PetscArraycpy(a_new,a->a + starts[i],lensi);
2502: a_new += lensi;
2503: i_new[i+1] = i_new[i] + lensi;
2504: c->ilen[i] = lensi;
2505: }
2506: PetscFree2(lens,starts);
2507: } else {
2508: ISGetIndices(iscol,&icol);
2509: PetscCalloc1(oldcols,&smap);
2510: PetscMalloc1(1+nrows,&lens);
2511: for (i=0; i<ncols; i++) {
2512: #if defined(PETSC_USE_DEBUG)
2513: if (icol[i] >= oldcols) SETERRQ3(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Requesting column beyond largest column icol[%D] %D <= A->cmap->n %D",i,icol[i],oldcols);
2514: #endif
2515: smap[icol[i]] = i+1;
2516: }
2518: /* determine lens of each row */
2519: for (i=0; i<nrows; i++) {
2520: kstart = ai[irow[i]];
2521: kend = kstart + a->ilen[irow[i]];
2522: lens[i] = 0;
2523: for (k=kstart; k<kend; k++) {
2524: if (smap[aj[k]]) {
2525: lens[i]++;
2526: }
2527: }
2528: }
2529: /* Create and fill new matrix */
2530: if (scall == MAT_REUSE_MATRIX) {
2531: PetscBool equal;
2533: c = (Mat_SeqAIJ*)((*B)->data);
2534: if ((*B)->rmap->n != nrows || (*B)->cmap->n != ncols) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Cannot reuse matrix. wrong size");
2535: PetscArraycmp(c->ilen,lens,(*B)->rmap->n,&equal);
2536: if (!equal) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Cannot reuse matrix. wrong no of nonzeros");
2537: PetscArrayzero(c->ilen,(*B)->rmap->n);
2538: C = *B;
2539: } else {
2540: PetscInt rbs,cbs;
2541: MatCreate(PetscObjectComm((PetscObject)A),&C);
2542: MatSetSizes(C,nrows,ncols,PETSC_DETERMINE,PETSC_DETERMINE);
2543: ISGetBlockSize(isrow,&rbs);
2544: ISGetBlockSize(iscol,&cbs);
2545: MatSetBlockSizes(C,rbs,cbs);
2546: MatSetType(C,((PetscObject)A)->type_name);
2547: MatSeqAIJSetPreallocation_SeqAIJ(C,0,lens);
2548: }
2549: c = (Mat_SeqAIJ*)(C->data);
2550: for (i=0; i<nrows; i++) {
2551: row = irow[i];
2552: kstart = ai[row];
2553: kend = kstart + a->ilen[row];
2554: mat_i = c->i[i];
2555: mat_j = c->j + mat_i;
2556: mat_a = c->a + mat_i;
2557: mat_ilen = c->ilen + i;
2558: for (k=kstart; k<kend; k++) {
2559: if ((tcol=smap[a->j[k]])) {
2560: *mat_j++ = tcol - 1;
2561: *mat_a++ = a->a[k];
2562: (*mat_ilen)++;
2564: }
2565: }
2566: }
2567: /* Free work space */
2568: ISRestoreIndices(iscol,&icol);
2569: PetscFree(smap);
2570: PetscFree(lens);
2571: /* sort */
2572: for (i = 0; i < nrows; i++) {
2573: PetscInt ilen;
2575: mat_i = c->i[i];
2576: mat_j = c->j + mat_i;
2577: mat_a = c->a + mat_i;
2578: ilen = c->ilen[i];
2579: PetscSortIntWithScalarArray(ilen,mat_j,mat_a);
2580: }
2581: }
2582: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
2583: MatBindToCPU(C,A->boundtocpu);
2584: #endif
2585: MatAssemblyBegin(C,MAT_FINAL_ASSEMBLY);
2586: MatAssemblyEnd(C,MAT_FINAL_ASSEMBLY);
2588: ISRestoreIndices(isrow,&irow);
2589: *B = C;
2590: return(0);
2591: }
2593: PetscErrorCode MatGetMultiProcBlock_SeqAIJ(Mat mat,MPI_Comm subComm,MatReuse scall,Mat *subMat)
2594: {
2596: Mat B;
2599: if (scall == MAT_INITIAL_MATRIX) {
2600: MatCreate(subComm,&B);
2601: MatSetSizes(B,mat->rmap->n,mat->cmap->n,mat->rmap->n,mat->cmap->n);
2602: MatSetBlockSizesFromMats(B,mat,mat);
2603: MatSetType(B,MATSEQAIJ);
2604: MatDuplicateNoCreate_SeqAIJ(B,mat,MAT_COPY_VALUES,PETSC_TRUE);
2605: *subMat = B;
2606: } else {
2607: MatCopy_SeqAIJ(mat,*subMat,SAME_NONZERO_PATTERN);
2608: }
2609: return(0);
2610: }
2612: PetscErrorCode MatILUFactor_SeqAIJ(Mat inA,IS row,IS col,const MatFactorInfo *info)
2613: {
2614: Mat_SeqAIJ *a = (Mat_SeqAIJ*)inA->data;
2616: Mat outA;
2617: PetscBool row_identity,col_identity;
2620: if (info->levels != 0) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Only levels=0 supported for in-place ilu");
2622: ISIdentity(row,&row_identity);
2623: ISIdentity(col,&col_identity);
2625: outA = inA;
2626: outA->factortype = MAT_FACTOR_LU;
2627: PetscFree(inA->solvertype);
2628: PetscStrallocpy(MATSOLVERPETSC,&inA->solvertype);
2630: PetscObjectReference((PetscObject)row);
2631: ISDestroy(&a->row);
2633: a->row = row;
2635: PetscObjectReference((PetscObject)col);
2636: ISDestroy(&a->col);
2638: a->col = col;
2640: /* Create the inverse permutation so that it can be used in MatLUFactorNumeric() */
2641: ISDestroy(&a->icol);
2642: ISInvertPermutation(col,PETSC_DECIDE,&a->icol);
2643: PetscLogObjectParent((PetscObject)inA,(PetscObject)a->icol);
2645: if (!a->solve_work) { /* this matrix may have been factored before */
2646: PetscMalloc1(inA->rmap->n+1,&a->solve_work);
2647: PetscLogObjectMemory((PetscObject)inA, (inA->rmap->n+1)*sizeof(PetscScalar));
2648: }
2650: MatMarkDiagonal_SeqAIJ(inA);
2651: if (row_identity && col_identity) {
2652: MatLUFactorNumeric_SeqAIJ_inplace(outA,inA,info);
2653: } else {
2654: MatLUFactorNumeric_SeqAIJ_InplaceWithPerm(outA,inA,info);
2655: }
2656: return(0);
2657: }
2659: PetscErrorCode MatScale_SeqAIJ(Mat inA,PetscScalar alpha)
2660: {
2661: Mat_SeqAIJ *a = (Mat_SeqAIJ*)inA->data;
2662: PetscScalar oalpha = alpha;
2664: PetscBLASInt one = 1,bnz;
2667: PetscBLASIntCast(a->nz,&bnz);
2668: PetscStackCallBLAS("BLASscal",BLASscal_(&bnz,&oalpha,a->a,&one));
2669: PetscLogFlops(a->nz);
2670: MatSeqAIJInvalidateDiagonal(inA);
2671: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
2672: if (inA->offloadmask != PETSC_OFFLOAD_UNALLOCATED) inA->offloadmask = PETSC_OFFLOAD_CPU;
2673: #endif
2674: return(0);
2675: }
2677: PetscErrorCode MatDestroySubMatrix_Private(Mat_SubSppt *submatj)
2678: {
2680: PetscInt i;
2683: if (!submatj->id) { /* delete data that are linked only to submats[id=0] */
2684: PetscFree4(submatj->sbuf1,submatj->ptr,submatj->tmp,submatj->ctr);
2686: for (i=0; i<submatj->nrqr; ++i) {
2687: PetscFree(submatj->sbuf2[i]);
2688: }
2689: PetscFree3(submatj->sbuf2,submatj->req_size,submatj->req_source1);
2691: if (submatj->rbuf1) {
2692: PetscFree(submatj->rbuf1[0]);
2693: PetscFree(submatj->rbuf1);
2694: }
2696: for (i=0; i<submatj->nrqs; ++i) {
2697: PetscFree(submatj->rbuf3[i]);
2698: }
2699: PetscFree3(submatj->req_source2,submatj->rbuf2,submatj->rbuf3);
2700: PetscFree(submatj->pa);
2701: }
2703: #if defined(PETSC_USE_CTABLE)
2704: PetscTableDestroy((PetscTable*)&submatj->rmap);
2705: if (submatj->cmap_loc) {PetscFree(submatj->cmap_loc);}
2706: PetscFree(submatj->rmap_loc);
2707: #else
2708: PetscFree(submatj->rmap);
2709: #endif
2711: if (!submatj->allcolumns) {
2712: #if defined(PETSC_USE_CTABLE)
2713: PetscTableDestroy((PetscTable*)&submatj->cmap);
2714: #else
2715: PetscFree(submatj->cmap);
2716: #endif
2717: }
2718: PetscFree(submatj->row2proc);
2720: PetscFree(submatj);
2721: return(0);
2722: }
2724: PetscErrorCode MatDestroySubMatrix_SeqAIJ(Mat C)
2725: {
2727: Mat_SeqAIJ *c = (Mat_SeqAIJ*)C->data;
2728: Mat_SubSppt *submatj = c->submatis1;
2731: (*submatj->destroy)(C);
2732: MatDestroySubMatrix_Private(submatj);
2733: return(0);
2734: }
2736: PetscErrorCode MatDestroySubMatrices_SeqAIJ(PetscInt n,Mat *mat[])
2737: {
2739: PetscInt i;
2740: Mat C;
2741: Mat_SeqAIJ *c;
2742: Mat_SubSppt *submatj;
2745: for (i=0; i<n; i++) {
2746: C = (*mat)[i];
2747: c = (Mat_SeqAIJ*)C->data;
2748: submatj = c->submatis1;
2749: if (submatj) {
2750: if (--((PetscObject)C)->refct <= 0) {
2751: (*submatj->destroy)(C);
2752: MatDestroySubMatrix_Private(submatj);
2753: PetscFree(C->defaultvectype);
2754: PetscLayoutDestroy(&C->rmap);
2755: PetscLayoutDestroy(&C->cmap);
2756: PetscHeaderDestroy(&C);
2757: }
2758: } else {
2759: MatDestroy(&C);
2760: }
2761: }
2763: /* Destroy Dummy submatrices created for reuse */
2764: MatDestroySubMatrices_Dummy(n,mat);
2766: PetscFree(*mat);
2767: return(0);
2768: }
2770: PetscErrorCode MatCreateSubMatrices_SeqAIJ(Mat A,PetscInt n,const IS irow[],const IS icol[],MatReuse scall,Mat *B[])
2771: {
2773: PetscInt i;
2776: if (scall == MAT_INITIAL_MATRIX) {
2777: PetscCalloc1(n+1,B);
2778: }
2780: for (i=0; i<n; i++) {
2781: MatCreateSubMatrix_SeqAIJ(A,irow[i],icol[i],PETSC_DECIDE,scall,&(*B)[i]);
2782: }
2783: return(0);
2784: }
2786: PetscErrorCode MatIncreaseOverlap_SeqAIJ(Mat A,PetscInt is_max,IS is[],PetscInt ov)
2787: {
2788: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
2790: PetscInt row,i,j,k,l,m,n,*nidx,isz,val;
2791: const PetscInt *idx;
2792: PetscInt start,end,*ai,*aj;
2793: PetscBT table;
2796: m = A->rmap->n;
2797: ai = a->i;
2798: aj = a->j;
2800: if (ov < 0) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"illegal negative overlap value used");
2802: PetscMalloc1(m+1,&nidx);
2803: PetscBTCreate(m,&table);
2805: for (i=0; i<is_max; i++) {
2806: /* Initialize the two local arrays */
2807: isz = 0;
2808: PetscBTMemzero(m,table);
2810: /* Extract the indices, assume there can be duplicate entries */
2811: ISGetIndices(is[i],&idx);
2812: ISGetLocalSize(is[i],&n);
2814: /* Enter these into the temp arrays. I.e., mark table[row], enter row into new index */
2815: for (j=0; j<n; ++j) {
2816: if (!PetscBTLookupSet(table,idx[j])) nidx[isz++] = idx[j];
2817: }
2818: ISRestoreIndices(is[i],&idx);
2819: ISDestroy(&is[i]);
2821: k = 0;
2822: for (j=0; j<ov; j++) { /* for each overlap */
2823: n = isz;
2824: for (; k<n; k++) { /* do only those rows in nidx[k], which are not done yet */
2825: row = nidx[k];
2826: start = ai[row];
2827: end = ai[row+1];
2828: for (l = start; l<end; l++) {
2829: val = aj[l];
2830: if (!PetscBTLookupSet(table,val)) nidx[isz++] = val;
2831: }
2832: }
2833: }
2834: ISCreateGeneral(PETSC_COMM_SELF,isz,nidx,PETSC_COPY_VALUES,(is+i));
2835: }
2836: PetscBTDestroy(&table);
2837: PetscFree(nidx);
2838: return(0);
2839: }
2841: /* -------------------------------------------------------------- */
2842: PetscErrorCode MatPermute_SeqAIJ(Mat A,IS rowp,IS colp,Mat *B)
2843: {
2844: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
2846: PetscInt i,nz = 0,m = A->rmap->n,n = A->cmap->n;
2847: const PetscInt *row,*col;
2848: PetscInt *cnew,j,*lens;
2849: IS icolp,irowp;
2850: PetscInt *cwork = NULL;
2851: PetscScalar *vwork = NULL;
2854: ISInvertPermutation(rowp,PETSC_DECIDE,&irowp);
2855: ISGetIndices(irowp,&row);
2856: ISInvertPermutation(colp,PETSC_DECIDE,&icolp);
2857: ISGetIndices(icolp,&col);
2859: /* determine lengths of permuted rows */
2860: PetscMalloc1(m+1,&lens);
2861: for (i=0; i<m; i++) lens[row[i]] = a->i[i+1] - a->i[i];
2862: MatCreate(PetscObjectComm((PetscObject)A),B);
2863: MatSetSizes(*B,m,n,m,n);
2864: MatSetBlockSizesFromMats(*B,A,A);
2865: MatSetType(*B,((PetscObject)A)->type_name);
2866: MatSeqAIJSetPreallocation_SeqAIJ(*B,0,lens);
2867: PetscFree(lens);
2869: PetscMalloc1(n,&cnew);
2870: for (i=0; i<m; i++) {
2871: MatGetRow_SeqAIJ(A,i,&nz,&cwork,&vwork);
2872: for (j=0; j<nz; j++) cnew[j] = col[cwork[j]];
2873: MatSetValues_SeqAIJ(*B,1,&row[i],nz,cnew,vwork,INSERT_VALUES);
2874: MatRestoreRow_SeqAIJ(A,i,&nz,&cwork,&vwork);
2875: }
2876: PetscFree(cnew);
2878: (*B)->assembled = PETSC_FALSE;
2880: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
2881: MatBindToCPU(*B,A->boundtocpu);
2882: #endif
2883: MatAssemblyBegin(*B,MAT_FINAL_ASSEMBLY);
2884: MatAssemblyEnd(*B,MAT_FINAL_ASSEMBLY);
2885: ISRestoreIndices(irowp,&row);
2886: ISRestoreIndices(icolp,&col);
2887: ISDestroy(&irowp);
2888: ISDestroy(&icolp);
2889: if (rowp == colp) {
2890: if (A->symmetric) {
2891: MatSetOption(*B,MAT_SYMMETRIC,PETSC_TRUE);
2892: }
2893: if (A->hermitian) {
2894: MatSetOption(*B,MAT_HERMITIAN,PETSC_TRUE);
2895: }
2896: }
2897: return(0);
2898: }
2900: PetscErrorCode MatCopy_SeqAIJ(Mat A,Mat B,MatStructure str)
2901: {
2905: /* If the two matrices have the same copy implementation, use fast copy. */
2906: if (str == SAME_NONZERO_PATTERN && (A->ops->copy == B->ops->copy)) {
2907: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
2908: Mat_SeqAIJ *b = (Mat_SeqAIJ*)B->data;
2910: if (a->i[A->rmap->n] != b->i[B->rmap->n]) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_INCOMP,"Number of nonzeros in two matrices are different %D != %D",a->i[A->rmap->n],b->i[B->rmap->n]);
2911: PetscArraycpy(b->a,a->a,a->i[A->rmap->n]);
2912: PetscObjectStateIncrease((PetscObject)B);
2913: } else {
2914: MatCopy_Basic(A,B,str);
2915: }
2916: return(0);
2917: }
2919: PetscErrorCode MatSetUp_SeqAIJ(Mat A)
2920: {
2924: MatSeqAIJSetPreallocation_SeqAIJ(A,PETSC_DEFAULT,0);
2925: return(0);
2926: }
2928: PETSC_INTERN PetscErrorCode MatSeqAIJGetArray_SeqAIJ(Mat A,PetscScalar *array[])
2929: {
2930: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
2933: *array = a->a;
2934: return(0);
2935: }
2937: PETSC_INTERN PetscErrorCode MatSeqAIJRestoreArray_SeqAIJ(Mat A,PetscScalar *array[])
2938: {
2940: *array = NULL;
2941: return(0);
2942: }
2944: /*
2945: Computes the number of nonzeros per row needed for preallocation when X and Y
2946: have different nonzero structure.
2947: */
2948: PetscErrorCode MatAXPYGetPreallocation_SeqX_private(PetscInt m,const PetscInt *xi,const PetscInt *xj,const PetscInt *yi,const PetscInt *yj,PetscInt *nnz)
2949: {
2950: PetscInt i,j,k,nzx,nzy;
2953: /* Set the number of nonzeros in the new matrix */
2954: for (i=0; i<m; i++) {
2955: const PetscInt *xjj = xj+xi[i],*yjj = yj+yi[i];
2956: nzx = xi[i+1] - xi[i];
2957: nzy = yi[i+1] - yi[i];
2958: nnz[i] = 0;
2959: for (j=0,k=0; j<nzx; j++) { /* Point in X */
2960: for (; k<nzy && yjj[k]<xjj[j]; k++) nnz[i]++; /* Catch up to X */
2961: if (k<nzy && yjj[k]==xjj[j]) k++; /* Skip duplicate */
2962: nnz[i]++;
2963: }
2964: for (; k<nzy; k++) nnz[i]++;
2965: }
2966: return(0);
2967: }
2969: PetscErrorCode MatAXPYGetPreallocation_SeqAIJ(Mat Y,Mat X,PetscInt *nnz)
2970: {
2971: PetscInt m = Y->rmap->N;
2972: Mat_SeqAIJ *x = (Mat_SeqAIJ*)X->data;
2973: Mat_SeqAIJ *y = (Mat_SeqAIJ*)Y->data;
2977: /* Set the number of nonzeros in the new matrix */
2978: MatAXPYGetPreallocation_SeqX_private(m,x->i,x->j,y->i,y->j,nnz);
2979: return(0);
2980: }
2982: PetscErrorCode MatAXPY_SeqAIJ(Mat Y,PetscScalar a,Mat X,MatStructure str)
2983: {
2985: Mat_SeqAIJ *x = (Mat_SeqAIJ*)X->data,*y = (Mat_SeqAIJ*)Y->data;
2986: PetscBLASInt one=1,bnz;
2989: PetscBLASIntCast(x->nz,&bnz);
2990: if (str == SAME_NONZERO_PATTERN) {
2991: PetscScalar alpha = a;
2992: PetscStackCallBLAS("BLASaxpy",BLASaxpy_(&bnz,&alpha,x->a,&one,y->a,&one));
2993: MatSeqAIJInvalidateDiagonal(Y);
2994: PetscObjectStateIncrease((PetscObject)Y);
2995: /* the MatAXPY_Basic* subroutines calls MatAssembly, so the matrix on the GPU
2996: will be updated */
2997: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
2998: if (Y->offloadmask != PETSC_OFFLOAD_UNALLOCATED) {
2999: Y->offloadmask = PETSC_OFFLOAD_CPU;
3000: }
3001: #endif
3002: } else if (str == SUBSET_NONZERO_PATTERN) { /* nonzeros of X is a subset of Y's */
3003: MatAXPY_Basic(Y,a,X,str);
3004: } else {
3005: Mat B;
3006: PetscInt *nnz;
3007: PetscMalloc1(Y->rmap->N,&nnz);
3008: MatCreate(PetscObjectComm((PetscObject)Y),&B);
3009: PetscObjectSetName((PetscObject)B,((PetscObject)Y)->name);
3010: MatSetSizes(B,Y->rmap->n,Y->cmap->n,Y->rmap->N,Y->cmap->N);
3011: MatSetBlockSizesFromMats(B,Y,Y);
3012: MatSetType(B,(MatType) ((PetscObject)Y)->type_name);
3013: MatAXPYGetPreallocation_SeqAIJ(Y,X,nnz);
3014: MatSeqAIJSetPreallocation(B,0,nnz);
3015: MatAXPY_BasicWithPreallocation(B,Y,a,X,str);
3016: MatHeaderReplace(Y,&B);
3017: PetscFree(nnz);
3018: }
3019: return(0);
3020: }
3022: PetscErrorCode MatConjugate_SeqAIJ(Mat mat)
3023: {
3024: #if defined(PETSC_USE_COMPLEX)
3025: Mat_SeqAIJ *aij = (Mat_SeqAIJ*)mat->data;
3026: PetscInt i,nz;
3027: PetscScalar *a;
3030: nz = aij->nz;
3031: a = aij->a;
3032: for (i=0; i<nz; i++) a[i] = PetscConj(a[i]);
3033: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
3034: if (mat->offloadmask != PETSC_OFFLOAD_UNALLOCATED) mat->offloadmask = PETSC_OFFLOAD_CPU;
3035: #endif
3036: #else
3038: #endif
3039: return(0);
3040: }
3042: PetscErrorCode MatGetRowMaxAbs_SeqAIJ(Mat A,Vec v,PetscInt idx[])
3043: {
3044: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
3046: PetscInt i,j,m = A->rmap->n,*ai,*aj,ncols,n;
3047: PetscReal atmp;
3048: PetscScalar *x;
3049: MatScalar *aa;
3052: if (A->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3053: aa = a->a;
3054: ai = a->i;
3055: aj = a->j;
3057: VecSet(v,0.0);
3058: VecGetArray(v,&x);
3059: VecGetLocalSize(v,&n);
3060: if (n != A->rmap->n) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Nonconforming matrix and vector");
3061: for (i=0; i<m; i++) {
3062: ncols = ai[1] - ai[0]; ai++;
3063: x[i] = 0.0;
3064: for (j=0; j<ncols; j++) {
3065: atmp = PetscAbsScalar(*aa);
3066: if (PetscAbsScalar(x[i]) < atmp) {x[i] = atmp; if (idx) idx[i] = *aj;}
3067: aa++; aj++;
3068: }
3069: }
3070: VecRestoreArray(v,&x);
3071: return(0);
3072: }
3074: PetscErrorCode MatGetRowMax_SeqAIJ(Mat A,Vec v,PetscInt idx[])
3075: {
3076: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
3078: PetscInt i,j,m = A->rmap->n,*ai,*aj,ncols,n;
3079: PetscScalar *x;
3080: MatScalar *aa;
3083: if (A->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3084: aa = a->a;
3085: ai = a->i;
3086: aj = a->j;
3088: VecSet(v,0.0);
3089: VecGetArray(v,&x);
3090: VecGetLocalSize(v,&n);
3091: if (n != A->rmap->n) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Nonconforming matrix and vector");
3092: for (i=0; i<m; i++) {
3093: ncols = ai[1] - ai[0]; ai++;
3094: if (ncols == A->cmap->n) { /* row is dense */
3095: x[i] = *aa; if (idx) idx[i] = 0;
3096: } else { /* row is sparse so already KNOW maximum is 0.0 or higher */
3097: x[i] = 0.0;
3098: if (idx) {
3099: idx[i] = 0; /* in case ncols is zero */
3100: for (j=0;j<ncols;j++) { /* find first implicit 0.0 in the row */
3101: if (aj[j] > j) {
3102: idx[i] = j;
3103: break;
3104: }
3105: }
3106: }
3107: }
3108: for (j=0; j<ncols; j++) {
3109: if (PetscRealPart(x[i]) < PetscRealPart(*aa)) {x[i] = *aa; if (idx) idx[i] = *aj;}
3110: aa++; aj++;
3111: }
3112: }
3113: VecRestoreArray(v,&x);
3114: return(0);
3115: }
3117: PetscErrorCode MatGetRowMinAbs_SeqAIJ(Mat A,Vec v,PetscInt idx[])
3118: {
3119: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
3121: PetscInt i,j,m = A->rmap->n,*ai,*aj,ncols,n;
3122: PetscReal atmp;
3123: PetscScalar *x;
3124: MatScalar *aa;
3127: if (A->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3128: aa = a->a;
3129: ai = a->i;
3130: aj = a->j;
3132: VecSet(v,0.0);
3133: VecGetArray(v,&x);
3134: VecGetLocalSize(v,&n);
3135: if (n != A->rmap->n) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Nonconforming matrix and vector, %D vs. %D rows", A->rmap->n, n);
3136: for (i=0; i<m; i++) {
3137: ncols = ai[1] - ai[0]; ai++;
3138: if (ncols) {
3139: /* Get first nonzero */
3140: for (j = 0; j < ncols; j++) {
3141: atmp = PetscAbsScalar(aa[j]);
3142: if (atmp > 1.0e-12) {
3143: x[i] = atmp;
3144: if (idx) idx[i] = aj[j];
3145: break;
3146: }
3147: }
3148: if (j == ncols) {x[i] = PetscAbsScalar(*aa); if (idx) idx[i] = *aj;}
3149: } else {
3150: x[i] = 0.0; if (idx) idx[i] = 0;
3151: }
3152: for (j = 0; j < ncols; j++) {
3153: atmp = PetscAbsScalar(*aa);
3154: if (atmp > 1.0e-12 && PetscAbsScalar(x[i]) > atmp) {x[i] = atmp; if (idx) idx[i] = *aj;}
3155: aa++; aj++;
3156: }
3157: }
3158: VecRestoreArray(v,&x);
3159: return(0);
3160: }
3162: PetscErrorCode MatGetRowMin_SeqAIJ(Mat A,Vec v,PetscInt idx[])
3163: {
3164: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
3165: PetscErrorCode ierr;
3166: PetscInt i,j,m = A->rmap->n,ncols,n;
3167: const PetscInt *ai,*aj;
3168: PetscScalar *x;
3169: const MatScalar *aa;
3172: if (A->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3173: aa = a->a;
3174: ai = a->i;
3175: aj = a->j;
3177: VecSet(v,0.0);
3178: VecGetArray(v,&x);
3179: VecGetLocalSize(v,&n);
3180: if (n != A->rmap->n) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Nonconforming matrix and vector");
3181: for (i=0; i<m; i++) {
3182: ncols = ai[1] - ai[0]; ai++;
3183: if (ncols == A->cmap->n) { /* row is dense */
3184: x[i] = *aa; if (idx) idx[i] = 0;
3185: } else { /* row is sparse so already KNOW minimum is 0.0 or lower */
3186: x[i] = 0.0;
3187: if (idx) { /* find first implicit 0.0 in the row */
3188: idx[i] = 0; /* in case ncols is zero */
3189: for (j=0; j<ncols; j++) {
3190: if (aj[j] > j) {
3191: idx[i] = j;
3192: break;
3193: }
3194: }
3195: }
3196: }
3197: for (j=0; j<ncols; j++) {
3198: if (PetscRealPart(x[i]) > PetscRealPart(*aa)) {x[i] = *aa; if (idx) idx[i] = *aj;}
3199: aa++; aj++;
3200: }
3201: }
3202: VecRestoreArray(v,&x);
3203: return(0);
3204: }
3206: PetscErrorCode MatInvertBlockDiagonal_SeqAIJ(Mat A,const PetscScalar **values)
3207: {
3208: Mat_SeqAIJ *a = (Mat_SeqAIJ*) A->data;
3209: PetscErrorCode ierr;
3210: PetscInt i,bs = PetscAbs(A->rmap->bs),mbs = A->rmap->n/bs,ipvt[5],bs2 = bs*bs,*v_pivots,ij[7],*IJ,j;
3211: MatScalar *diag,work[25],*v_work;
3212: const PetscReal shift = 0.0;
3213: PetscBool allowzeropivot,zeropivotdetected=PETSC_FALSE;
3216: allowzeropivot = PetscNot(A->erroriffailure);
3217: if (a->ibdiagvalid) {
3218: if (values) *values = a->ibdiag;
3219: return(0);
3220: }
3221: MatMarkDiagonal_SeqAIJ(A);
3222: if (!a->ibdiag) {
3223: PetscMalloc1(bs2*mbs,&a->ibdiag);
3224: PetscLogObjectMemory((PetscObject)A,bs2*mbs*sizeof(PetscScalar));
3225: }
3226: diag = a->ibdiag;
3227: if (values) *values = a->ibdiag;
3228: /* factor and invert each block */
3229: switch (bs) {
3230: case 1:
3231: for (i=0; i<mbs; i++) {
3232: MatGetValues(A,1,&i,1,&i,diag+i);
3233: if (PetscAbsScalar(diag[i] + shift) < PETSC_MACHINE_EPSILON) {
3234: if (allowzeropivot) {
3235: A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
3236: A->factorerror_zeropivot_value = PetscAbsScalar(diag[i]);
3237: A->factorerror_zeropivot_row = i;
3238: PetscInfo3(A,"Zero pivot, row %D pivot %g tolerance %g\n",i,(double)PetscAbsScalar(diag[i]),(double)PETSC_MACHINE_EPSILON);
3239: } else SETERRQ3(PETSC_COMM_SELF,PETSC_ERR_MAT_LU_ZRPVT,"Zero pivot, row %D pivot %g tolerance %g",i,(double)PetscAbsScalar(diag[i]),(double)PETSC_MACHINE_EPSILON);
3240: }
3241: diag[i] = (PetscScalar)1.0 / (diag[i] + shift);
3242: }
3243: break;
3244: case 2:
3245: for (i=0; i<mbs; i++) {
3246: ij[0] = 2*i; ij[1] = 2*i + 1;
3247: MatGetValues(A,2,ij,2,ij,diag);
3248: PetscKernel_A_gets_inverse_A_2(diag,shift,allowzeropivot,&zeropivotdetected);
3249: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
3250: PetscKernel_A_gets_transpose_A_2(diag);
3251: diag += 4;
3252: }
3253: break;
3254: case 3:
3255: for (i=0; i<mbs; i++) {
3256: ij[0] = 3*i; ij[1] = 3*i + 1; ij[2] = 3*i + 2;
3257: MatGetValues(A,3,ij,3,ij,diag);
3258: PetscKernel_A_gets_inverse_A_3(diag,shift,allowzeropivot,&zeropivotdetected);
3259: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
3260: PetscKernel_A_gets_transpose_A_3(diag);
3261: diag += 9;
3262: }
3263: break;
3264: case 4:
3265: for (i=0; i<mbs; i++) {
3266: ij[0] = 4*i; ij[1] = 4*i + 1; ij[2] = 4*i + 2; ij[3] = 4*i + 3;
3267: MatGetValues(A,4,ij,4,ij,diag);
3268: PetscKernel_A_gets_inverse_A_4(diag,shift,allowzeropivot,&zeropivotdetected);
3269: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
3270: PetscKernel_A_gets_transpose_A_4(diag);
3271: diag += 16;
3272: }
3273: break;
3274: case 5:
3275: for (i=0; i<mbs; i++) {
3276: ij[0] = 5*i; ij[1] = 5*i + 1; ij[2] = 5*i + 2; ij[3] = 5*i + 3; ij[4] = 5*i + 4;
3277: MatGetValues(A,5,ij,5,ij,diag);
3278: PetscKernel_A_gets_inverse_A_5(diag,ipvt,work,shift,allowzeropivot,&zeropivotdetected);
3279: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
3280: PetscKernel_A_gets_transpose_A_5(diag);
3281: diag += 25;
3282: }
3283: break;
3284: case 6:
3285: for (i=0; i<mbs; i++) {
3286: ij[0] = 6*i; ij[1] = 6*i + 1; ij[2] = 6*i + 2; ij[3] = 6*i + 3; ij[4] = 6*i + 4; ij[5] = 6*i + 5;
3287: MatGetValues(A,6,ij,6,ij,diag);
3288: PetscKernel_A_gets_inverse_A_6(diag,shift,allowzeropivot,&zeropivotdetected);
3289: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
3290: PetscKernel_A_gets_transpose_A_6(diag);
3291: diag += 36;
3292: }
3293: break;
3294: case 7:
3295: for (i=0; i<mbs; i++) {
3296: ij[0] = 7*i; ij[1] = 7*i + 1; ij[2] = 7*i + 2; ij[3] = 7*i + 3; ij[4] = 7*i + 4; ij[5] = 7*i + 5; ij[5] = 7*i + 6;
3297: MatGetValues(A,7,ij,7,ij,diag);
3298: PetscKernel_A_gets_inverse_A_7(diag,shift,allowzeropivot,&zeropivotdetected);
3299: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
3300: PetscKernel_A_gets_transpose_A_7(diag);
3301: diag += 49;
3302: }
3303: break;
3304: default:
3305: PetscMalloc3(bs,&v_work,bs,&v_pivots,bs,&IJ);
3306: for (i=0; i<mbs; i++) {
3307: for (j=0; j<bs; j++) {
3308: IJ[j] = bs*i + j;
3309: }
3310: MatGetValues(A,bs,IJ,bs,IJ,diag);
3311: PetscKernel_A_gets_inverse_A(bs,diag,v_pivots,v_work,allowzeropivot,&zeropivotdetected);
3312: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
3313: PetscKernel_A_gets_transpose_A_N(diag,bs);
3314: diag += bs2;
3315: }
3316: PetscFree3(v_work,v_pivots,IJ);
3317: }
3318: a->ibdiagvalid = PETSC_TRUE;
3319: return(0);
3320: }
3322: static PetscErrorCode MatSetRandom_SeqAIJ(Mat x,PetscRandom rctx)
3323: {
3325: Mat_SeqAIJ *aij = (Mat_SeqAIJ*)x->data;
3326: PetscScalar a;
3327: PetscInt m,n,i,j,col;
3330: if (!x->assembled) {
3331: MatGetSize(x,&m,&n);
3332: for (i=0; i<m; i++) {
3333: for (j=0; j<aij->imax[i]; j++) {
3334: PetscRandomGetValue(rctx,&a);
3335: col = (PetscInt)(n*PetscRealPart(a));
3336: MatSetValues(x,1,&i,1,&col,&a,ADD_VALUES);
3337: }
3338: }
3339: } else {
3340: for (i=0; i<aij->nz; i++) {PetscRandomGetValue(rctx,aij->a+i);}
3341: }
3342: MatAssemblyBegin(x,MAT_FINAL_ASSEMBLY);
3343: MatAssemblyEnd(x,MAT_FINAL_ASSEMBLY);
3344: return(0);
3345: }
3347: /* Like MatSetRandom_SeqAIJ, but do not set values on columns in range of [low, high) */
3348: PetscErrorCode MatSetRandomSkipColumnRange_SeqAIJ_Private(Mat x,PetscInt low,PetscInt high,PetscRandom rctx)
3349: {
3351: Mat_SeqAIJ *aij = (Mat_SeqAIJ*)x->data;
3352: PetscScalar a;
3353: PetscInt m,n,i,j,col,nskip;
3356: nskip = high - low;
3357: MatGetSize(x,&m,&n);
3358: n -= nskip; /* shrink number of columns where nonzeros can be set */
3359: for (i=0; i<m; i++) {
3360: for (j=0; j<aij->imax[i]; j++) {
3361: PetscRandomGetValue(rctx,&a);
3362: col = (PetscInt)(n*PetscRealPart(a));
3363: if (col >= low) col += nskip; /* shift col rightward to skip the hole */
3364: MatSetValues(x,1,&i,1,&col,&a,ADD_VALUES);
3365: }
3366: }
3367: MatAssemblyBegin(x,MAT_FINAL_ASSEMBLY);
3368: MatAssemblyEnd(x,MAT_FINAL_ASSEMBLY);
3369: return(0);
3370: }
3373: /* -------------------------------------------------------------------*/
3374: static struct _MatOps MatOps_Values = { MatSetValues_SeqAIJ,
3375: MatGetRow_SeqAIJ,
3376: MatRestoreRow_SeqAIJ,
3377: MatMult_SeqAIJ,
3378: /* 4*/ MatMultAdd_SeqAIJ,
3379: MatMultTranspose_SeqAIJ,
3380: MatMultTransposeAdd_SeqAIJ,
3381: 0,
3382: 0,
3383: 0,
3384: /* 10*/ 0,
3385: MatLUFactor_SeqAIJ,
3386: 0,
3387: MatSOR_SeqAIJ,
3388: MatTranspose_SeqAIJ,
3389: /*1 5*/ MatGetInfo_SeqAIJ,
3390: MatEqual_SeqAIJ,
3391: MatGetDiagonal_SeqAIJ,
3392: MatDiagonalScale_SeqAIJ,
3393: MatNorm_SeqAIJ,
3394: /* 20*/ 0,
3395: MatAssemblyEnd_SeqAIJ,
3396: MatSetOption_SeqAIJ,
3397: MatZeroEntries_SeqAIJ,
3398: /* 24*/ MatZeroRows_SeqAIJ,
3399: 0,
3400: 0,
3401: 0,
3402: 0,
3403: /* 29*/ MatSetUp_SeqAIJ,
3404: 0,
3405: 0,
3406: 0,
3407: 0,
3408: /* 34*/ MatDuplicate_SeqAIJ,
3409: 0,
3410: 0,
3411: MatILUFactor_SeqAIJ,
3412: 0,
3413: /* 39*/ MatAXPY_SeqAIJ,
3414: MatCreateSubMatrices_SeqAIJ,
3415: MatIncreaseOverlap_SeqAIJ,
3416: MatGetValues_SeqAIJ,
3417: MatCopy_SeqAIJ,
3418: /* 44*/ MatGetRowMax_SeqAIJ,
3419: MatScale_SeqAIJ,
3420: MatShift_SeqAIJ,
3421: MatDiagonalSet_SeqAIJ,
3422: MatZeroRowsColumns_SeqAIJ,
3423: /* 49*/ MatSetRandom_SeqAIJ,
3424: MatGetRowIJ_SeqAIJ,
3425: MatRestoreRowIJ_SeqAIJ,
3426: MatGetColumnIJ_SeqAIJ,
3427: MatRestoreColumnIJ_SeqAIJ,
3428: /* 54*/ MatFDColoringCreate_SeqXAIJ,
3429: 0,
3430: 0,
3431: MatPermute_SeqAIJ,
3432: 0,
3433: /* 59*/ 0,
3434: MatDestroy_SeqAIJ,
3435: MatView_SeqAIJ,
3436: 0,
3437: 0,
3438: /* 64*/ 0,
3439: MatMatMatMultNumeric_SeqAIJ_SeqAIJ_SeqAIJ,
3440: 0,
3441: 0,
3442: 0,
3443: /* 69*/ MatGetRowMaxAbs_SeqAIJ,
3444: MatGetRowMinAbs_SeqAIJ,
3445: 0,
3446: 0,
3447: 0,
3448: /* 74*/ 0,
3449: MatFDColoringApply_AIJ,
3450: 0,
3451: 0,
3452: 0,
3453: /* 79*/ MatFindZeroDiagonals_SeqAIJ,
3454: 0,
3455: 0,
3456: 0,
3457: MatLoad_SeqAIJ,
3458: /* 84*/ MatIsSymmetric_SeqAIJ,
3459: MatIsHermitian_SeqAIJ,
3460: 0,
3461: 0,
3462: 0,
3463: /* 89*/ 0,
3464: 0,
3465: MatMatMultNumeric_SeqAIJ_SeqAIJ,
3466: 0,
3467: 0,
3468: /* 94*/ MatPtAPNumeric_SeqAIJ_SeqAIJ_SparseAxpy,
3469: 0,
3470: 0,
3471: MatMatTransposeMultNumeric_SeqAIJ_SeqAIJ,
3472: 0,
3473: /* 99*/ MatProductSetFromOptions_SeqAIJ,
3474: 0,
3475: 0,
3476: MatConjugate_SeqAIJ,
3477: 0,
3478: /*104*/ MatSetValuesRow_SeqAIJ,
3479: MatRealPart_SeqAIJ,
3480: MatImaginaryPart_SeqAIJ,
3481: 0,
3482: 0,
3483: /*109*/ MatMatSolve_SeqAIJ,
3484: 0,
3485: MatGetRowMin_SeqAIJ,
3486: 0,
3487: MatMissingDiagonal_SeqAIJ,
3488: /*114*/ 0,
3489: 0,
3490: 0,
3491: 0,
3492: 0,
3493: /*119*/ 0,
3494: 0,
3495: 0,
3496: 0,
3497: MatGetMultiProcBlock_SeqAIJ,
3498: /*124*/ MatFindNonzeroRows_SeqAIJ,
3499: MatGetColumnNorms_SeqAIJ,
3500: MatInvertBlockDiagonal_SeqAIJ,
3501: MatInvertVariableBlockDiagonal_SeqAIJ,
3502: 0,
3503: /*129*/ 0,
3504: 0,
3505: 0,
3506: MatTransposeMatMultNumeric_SeqAIJ_SeqAIJ,
3507: MatTransposeColoringCreate_SeqAIJ,
3508: /*134*/ MatTransColoringApplySpToDen_SeqAIJ,
3509: MatTransColoringApplyDenToSp_SeqAIJ,
3510: 0,
3511: 0,
3512: MatRARtNumeric_SeqAIJ_SeqAIJ,
3513: /*139*/0,
3514: 0,
3515: 0,
3516: MatFDColoringSetUp_SeqXAIJ,
3517: MatFindOffBlockDiagonalEntries_SeqAIJ,
3518: MatCreateMPIMatConcatenateSeqMat_SeqAIJ,
3519: /*145*/MatDestroySubMatrices_SeqAIJ,
3520: 0,
3521: 0
3522: };
3524: PetscErrorCode MatSeqAIJSetColumnIndices_SeqAIJ(Mat mat,PetscInt *indices)
3525: {
3526: Mat_SeqAIJ *aij = (Mat_SeqAIJ*)mat->data;
3527: PetscInt i,nz,n;
3530: nz = aij->maxnz;
3531: n = mat->rmap->n;
3532: for (i=0; i<nz; i++) {
3533: aij->j[i] = indices[i];
3534: }
3535: aij->nz = nz;
3536: for (i=0; i<n; i++) {
3537: aij->ilen[i] = aij->imax[i];
3538: }
3539: return(0);
3540: }
3542: /*
3543: * When a sparse matrix has many zero columns, we should compact them out to save the space
3544: * This happens in MatPtAPSymbolic_MPIAIJ_MPIAIJ_scalable()
3545: * */
3546: PetscErrorCode MatSeqAIJCompactOutExtraColumns_SeqAIJ(Mat mat, ISLocalToGlobalMapping *mapping)
3547: {
3548: Mat_SeqAIJ *aij = (Mat_SeqAIJ*)mat->data;
3549: PetscTable gid1_lid1;
3550: PetscTablePosition tpos;
3551: PetscInt gid,lid,i,j,ncols,ec;
3552: PetscInt *garray;
3553: PetscErrorCode ierr;
3558: /* use a table */
3559: PetscTableCreate(mat->rmap->n,mat->cmap->N+1,&gid1_lid1);
3560: ec = 0;
3561: for (i=0; i<mat->rmap->n; i++) {
3562: ncols = aij->i[i+1] - aij->i[i];
3563: for (j=0; j<ncols; j++) {
3564: PetscInt data,gid1 = aij->j[aij->i[i] + j] + 1;
3565: PetscTableFind(gid1_lid1,gid1,&data);
3566: if (!data) {
3567: /* one based table */
3568: PetscTableAdd(gid1_lid1,gid1,++ec,INSERT_VALUES);
3569: }
3570: }
3571: }
3572: /* form array of columns we need */
3573: PetscMalloc1(ec+1,&garray);
3574: PetscTableGetHeadPosition(gid1_lid1,&tpos);
3575: while (tpos) {
3576: PetscTableGetNext(gid1_lid1,&tpos,&gid,&lid);
3577: gid--;
3578: lid--;
3579: garray[lid] = gid;
3580: }
3581: PetscSortInt(ec,garray); /* sort, and rebuild */
3582: PetscTableRemoveAll(gid1_lid1);
3583: for (i=0; i<ec; i++) {
3584: PetscTableAdd(gid1_lid1,garray[i]+1,i+1,INSERT_VALUES);
3585: }
3586: /* compact out the extra columns in B */
3587: for (i=0; i<mat->rmap->n; i++) {
3588: ncols = aij->i[i+1] - aij->i[i];
3589: for (j=0; j<ncols; j++) {
3590: PetscInt gid1 = aij->j[aij->i[i] + j] + 1;
3591: PetscTableFind(gid1_lid1,gid1,&lid);
3592: lid--;
3593: aij->j[aij->i[i] + j] = lid;
3594: }
3595: }
3596: PetscLayoutDestroy(&mat->cmap);
3597: PetscLayoutCreateFromSizes(PetscObjectComm((PetscObject)mat),ec,ec,1,&mat->cmap);
3598: PetscTableDestroy(&gid1_lid1);
3599: ISLocalToGlobalMappingCreate(PETSC_COMM_SELF,mat->cmap->bs,mat->cmap->n,garray,PETSC_OWN_POINTER,mapping);
3600: ISLocalToGlobalMappingSetType(*mapping,ISLOCALTOGLOBALMAPPINGHASH);
3601: return(0);
3602: }
3604: /*@
3605: MatSeqAIJSetColumnIndices - Set the column indices for all the rows
3606: in the matrix.
3608: Input Parameters:
3609: + mat - the SeqAIJ matrix
3610: - indices - the column indices
3612: Level: advanced
3614: Notes:
3615: This can be called if you have precomputed the nonzero structure of the
3616: matrix and want to provide it to the matrix object to improve the performance
3617: of the MatSetValues() operation.
3619: You MUST have set the correct numbers of nonzeros per row in the call to
3620: MatCreateSeqAIJ(), and the columns indices MUST be sorted.
3622: MUST be called before any calls to MatSetValues();
3624: The indices should start with zero, not one.
3626: @*/
3627: PetscErrorCode MatSeqAIJSetColumnIndices(Mat mat,PetscInt *indices)
3628: {
3634: PetscUseMethod(mat,"MatSeqAIJSetColumnIndices_C",(Mat,PetscInt*),(mat,indices));
3635: return(0);
3636: }
3638: /* ----------------------------------------------------------------------------------------*/
3640: PetscErrorCode MatStoreValues_SeqAIJ(Mat mat)
3641: {
3642: Mat_SeqAIJ *aij = (Mat_SeqAIJ*)mat->data;
3644: size_t nz = aij->i[mat->rmap->n];
3647: if (!aij->nonew) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Must call MatSetOption(A,MAT_NEW_NONZERO_LOCATIONS,PETSC_FALSE);first");
3649: /* allocate space for values if not already there */
3650: if (!aij->saved_values) {
3651: PetscMalloc1(nz+1,&aij->saved_values);
3652: PetscLogObjectMemory((PetscObject)mat,(nz+1)*sizeof(PetscScalar));
3653: }
3655: /* copy values over */
3656: PetscArraycpy(aij->saved_values,aij->a,nz);
3657: return(0);
3658: }
3660: /*@
3661: MatStoreValues - Stashes a copy of the matrix values; this allows, for
3662: example, reuse of the linear part of a Jacobian, while recomputing the
3663: nonlinear portion.
3665: Collect on Mat
3667: Input Parameters:
3668: . mat - the matrix (currently only AIJ matrices support this option)
3670: Level: advanced
3672: Common Usage, with SNESSolve():
3673: $ Create Jacobian matrix
3674: $ Set linear terms into matrix
3675: $ Apply boundary conditions to matrix, at this time matrix must have
3676: $ final nonzero structure (i.e. setting the nonlinear terms and applying
3677: $ boundary conditions again will not change the nonzero structure
3678: $ MatSetOption(mat,MAT_NEW_NONZERO_LOCATIONS,PETSC_FALSE);
3679: $ MatStoreValues(mat);
3680: $ Call SNESSetJacobian() with matrix
3681: $ In your Jacobian routine
3682: $ MatRetrieveValues(mat);
3683: $ Set nonlinear terms in matrix
3685: Common Usage without SNESSolve(), i.e. when you handle nonlinear solve yourself:
3686: $ // build linear portion of Jacobian
3687: $ MatSetOption(mat,MAT_NEW_NONZERO_LOCATIONS,PETSC_FALSE);
3688: $ MatStoreValues(mat);
3689: $ loop over nonlinear iterations
3690: $ MatRetrieveValues(mat);
3691: $ // call MatSetValues(mat,...) to set nonliner portion of Jacobian
3692: $ // call MatAssemblyBegin/End() on matrix
3693: $ Solve linear system with Jacobian
3694: $ endloop
3696: Notes:
3697: Matrix must already be assemblied before calling this routine
3698: Must set the matrix option MatSetOption(mat,MAT_NEW_NONZERO_LOCATIONS,PETSC_FALSE); before
3699: calling this routine.
3701: When this is called multiple times it overwrites the previous set of stored values
3702: and does not allocated additional space.
3704: .seealso: MatRetrieveValues()
3706: @*/
3707: PetscErrorCode MatStoreValues(Mat mat)
3708: {
3713: if (!mat->assembled) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3714: if (mat->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3715: PetscUseMethod(mat,"MatStoreValues_C",(Mat),(mat));
3716: return(0);
3717: }
3719: PetscErrorCode MatRetrieveValues_SeqAIJ(Mat mat)
3720: {
3721: Mat_SeqAIJ *aij = (Mat_SeqAIJ*)mat->data;
3723: PetscInt nz = aij->i[mat->rmap->n];
3726: if (!aij->nonew) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Must call MatSetOption(A,MAT_NEW_NONZERO_LOCATIONS,PETSC_FALSE);first");
3727: if (!aij->saved_values) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Must call MatStoreValues(A);first");
3728: /* copy values over */
3729: PetscArraycpy(aij->a,aij->saved_values,nz);
3730: return(0);
3731: }
3733: /*@
3734: MatRetrieveValues - Retrieves the copy of the matrix values; this allows, for
3735: example, reuse of the linear part of a Jacobian, while recomputing the
3736: nonlinear portion.
3738: Collect on Mat
3740: Input Parameters:
3741: . mat - the matrix (currently only AIJ matrices support this option)
3743: Level: advanced
3745: .seealso: MatStoreValues()
3747: @*/
3748: PetscErrorCode MatRetrieveValues(Mat mat)
3749: {
3754: if (!mat->assembled) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3755: if (mat->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3756: PetscUseMethod(mat,"MatRetrieveValues_C",(Mat),(mat));
3757: return(0);
3758: }
3761: /* --------------------------------------------------------------------------------*/
3762: /*@C
3763: MatCreateSeqAIJ - Creates a sparse matrix in AIJ (compressed row) format
3764: (the default parallel PETSc format). For good matrix assembly performance
3765: the user should preallocate the matrix storage by setting the parameter nz
3766: (or the array nnz). By setting these parameters accurately, performance
3767: during matrix assembly can be increased by more than a factor of 50.
3769: Collective
3771: Input Parameters:
3772: + comm - MPI communicator, set to PETSC_COMM_SELF
3773: . m - number of rows
3774: . n - number of columns
3775: . nz - number of nonzeros per row (same for all rows)
3776: - nnz - array containing the number of nonzeros in the various rows
3777: (possibly different for each row) or NULL
3779: Output Parameter:
3780: . A - the matrix
3782: It is recommended that one use the MatCreate(), MatSetType() and/or MatSetFromOptions(),
3783: MatXXXXSetPreallocation() paradigm instead of this routine directly.
3784: [MatXXXXSetPreallocation() is, for example, MatSeqAIJSetPreallocation]
3786: Notes:
3787: If nnz is given then nz is ignored
3789: The AIJ format (also called the Yale sparse matrix format or
3790: compressed row storage), is fully compatible with standard Fortran 77
3791: storage. That is, the stored row and column indices can begin at
3792: either one (as in Fortran) or zero. See the users' manual for details.
3794: Specify the preallocated storage with either nz or nnz (not both).
3795: Set nz=PETSC_DEFAULT and nnz=NULL for PETSc to control dynamic memory
3796: allocation. For large problems you MUST preallocate memory or you
3797: will get TERRIBLE performance, see the users' manual chapter on matrices.
3799: By default, this format uses inodes (identical nodes) when possible, to
3800: improve numerical efficiency of matrix-vector products and solves. We
3801: search for consecutive rows with the same nonzero structure, thereby
3802: reusing matrix information to achieve increased efficiency.
3804: Options Database Keys:
3805: + -mat_no_inode - Do not use inodes
3806: - -mat_inode_limit <limit> - Sets inode limit (max limit=5)
3808: Level: intermediate
3810: .seealso: MatCreate(), MatCreateAIJ(), MatSetValues(), MatSeqAIJSetColumnIndices(), MatCreateSeqAIJWithArrays()
3812: @*/
3813: PetscErrorCode MatCreateSeqAIJ(MPI_Comm comm,PetscInt m,PetscInt n,PetscInt nz,const PetscInt nnz[],Mat *A)
3814: {
3818: MatCreate(comm,A);
3819: MatSetSizes(*A,m,n,m,n);
3820: MatSetType(*A,MATSEQAIJ);
3821: MatSeqAIJSetPreallocation_SeqAIJ(*A,nz,nnz);
3822: return(0);
3823: }
3825: /*@C
3826: MatSeqAIJSetPreallocation - For good matrix assembly performance
3827: the user should preallocate the matrix storage by setting the parameter nz
3828: (or the array nnz). By setting these parameters accurately, performance
3829: during matrix assembly can be increased by more than a factor of 50.
3831: Collective
3833: Input Parameters:
3834: + B - The matrix
3835: . nz - number of nonzeros per row (same for all rows)
3836: - nnz - array containing the number of nonzeros in the various rows
3837: (possibly different for each row) or NULL
3839: Notes:
3840: If nnz is given then nz is ignored
3842: The AIJ format (also called the Yale sparse matrix format or
3843: compressed row storage), is fully compatible with standard Fortran 77
3844: storage. That is, the stored row and column indices can begin at
3845: either one (as in Fortran) or zero. See the users' manual for details.
3847: Specify the preallocated storage with either nz or nnz (not both).
3848: Set nz=PETSC_DEFAULT and nnz=NULL for PETSc to control dynamic memory
3849: allocation. For large problems you MUST preallocate memory or you
3850: will get TERRIBLE performance, see the users' manual chapter on matrices.
3852: You can call MatGetInfo() to get information on how effective the preallocation was;
3853: for example the fields mallocs,nz_allocated,nz_used,nz_unneeded;
3854: You can also run with the option -info and look for messages with the string
3855: malloc in them to see if additional memory allocation was needed.
3857: Developers: Use nz of MAT_SKIP_ALLOCATION to not allocate any space for the matrix
3858: entries or columns indices
3860: By default, this format uses inodes (identical nodes) when possible, to
3861: improve numerical efficiency of matrix-vector products and solves. We
3862: search for consecutive rows with the same nonzero structure, thereby
3863: reusing matrix information to achieve increased efficiency.
3865: Options Database Keys:
3866: + -mat_no_inode - Do not use inodes
3867: - -mat_inode_limit <limit> - Sets inode limit (max limit=5)
3869: Level: intermediate
3871: .seealso: MatCreate(), MatCreateAIJ(), MatSetValues(), MatSeqAIJSetColumnIndices(), MatCreateSeqAIJWithArrays(), MatGetInfo()
3873: @*/
3874: PetscErrorCode MatSeqAIJSetPreallocation(Mat B,PetscInt nz,const PetscInt nnz[])
3875: {
3881: PetscTryMethod(B,"MatSeqAIJSetPreallocation_C",(Mat,PetscInt,const PetscInt[]),(B,nz,nnz));
3882: return(0);
3883: }
3885: PetscErrorCode MatSeqAIJSetPreallocation_SeqAIJ(Mat B,PetscInt nz,const PetscInt *nnz)
3886: {
3887: Mat_SeqAIJ *b;
3888: PetscBool skipallocation = PETSC_FALSE,realalloc = PETSC_FALSE;
3890: PetscInt i;
3893: if (nz >= 0 || nnz) realalloc = PETSC_TRUE;
3894: if (nz == MAT_SKIP_ALLOCATION) {
3895: skipallocation = PETSC_TRUE;
3896: nz = 0;
3897: }
3898: PetscLayoutSetUp(B->rmap);
3899: PetscLayoutSetUp(B->cmap);
3901: if (nz == PETSC_DEFAULT || nz == PETSC_DECIDE) nz = 5;
3902: if (nz < 0) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"nz cannot be less than 0: value %D",nz);
3903: #if defined(PETSC_USE_DEBUG)
3904: if (nnz) {
3905: for (i=0; i<B->rmap->n; i++) {
3906: if (nnz[i] < 0) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"nnz cannot be less than 0: local row %D value %D",i,nnz[i]);
3907: if (nnz[i] > B->cmap->n) SETERRQ3(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"nnz cannot be greater than row length: local row %D value %d rowlength %D",i,nnz[i],B->cmap->n);
3908: }
3909: }
3910: #endif
3912: B->preallocated = PETSC_TRUE;
3914: b = (Mat_SeqAIJ*)B->data;
3916: if (!skipallocation) {
3917: if (!b->imax) {
3918: PetscMalloc1(B->rmap->n,&b->imax);
3919: PetscLogObjectMemory((PetscObject)B,B->rmap->n*sizeof(PetscInt));
3920: }
3921: if (!b->ilen) {
3922: /* b->ilen will count nonzeros in each row so far. */
3923: PetscCalloc1(B->rmap->n,&b->ilen);
3924: PetscLogObjectMemory((PetscObject)B,B->rmap->n*sizeof(PetscInt));
3925: } else {
3926: PetscMemzero(b->ilen,B->rmap->n*sizeof(PetscInt));
3927: }
3928: if (!b->ipre) {
3929: PetscMalloc1(B->rmap->n,&b->ipre);
3930: PetscLogObjectMemory((PetscObject)B,B->rmap->n*sizeof(PetscInt));
3931: }
3932: if (!nnz) {
3933: if (nz == PETSC_DEFAULT || nz == PETSC_DECIDE) nz = 10;
3934: else if (nz < 0) nz = 1;
3935: nz = PetscMin(nz,B->cmap->n);
3936: for (i=0; i<B->rmap->n; i++) b->imax[i] = nz;
3937: nz = nz*B->rmap->n;
3938: } else {
3939: PetscInt64 nz64 = 0;
3940: for (i=0; i<B->rmap->n; i++) {b->imax[i] = nnz[i]; nz64 += nnz[i];}
3941: PetscIntCast(nz64,&nz);
3942: }
3944: /* allocate the matrix space */
3945: /* FIXME: should B's old memory be unlogged? */
3946: MatSeqXAIJFreeAIJ(B,&b->a,&b->j,&b->i);
3947: if (B->structure_only) {
3948: PetscMalloc1(nz,&b->j);
3949: PetscMalloc1(B->rmap->n+1,&b->i);
3950: PetscLogObjectMemory((PetscObject)B,(B->rmap->n+1)*sizeof(PetscInt)+nz*sizeof(PetscInt));
3951: } else {
3952: PetscMalloc3(nz,&b->a,nz,&b->j,B->rmap->n+1,&b->i);
3953: PetscLogObjectMemory((PetscObject)B,(B->rmap->n+1)*sizeof(PetscInt)+nz*(sizeof(PetscScalar)+sizeof(PetscInt)));
3954: }
3955: b->i[0] = 0;
3956: for (i=1; i<B->rmap->n+1; i++) {
3957: b->i[i] = b->i[i-1] + b->imax[i-1];
3958: }
3959: if (B->structure_only) {
3960: b->singlemalloc = PETSC_FALSE;
3961: b->free_a = PETSC_FALSE;
3962: } else {
3963: b->singlemalloc = PETSC_TRUE;
3964: b->free_a = PETSC_TRUE;
3965: }
3966: b->free_ij = PETSC_TRUE;
3967: } else {
3968: b->free_a = PETSC_FALSE;
3969: b->free_ij = PETSC_FALSE;
3970: }
3972: if (b->ipre && nnz != b->ipre && b->imax) {
3973: /* reserve user-requested sparsity */
3974: PetscArraycpy(b->ipre,b->imax,B->rmap->n);
3975: }
3978: b->nz = 0;
3979: b->maxnz = nz;
3980: B->info.nz_unneeded = (double)b->maxnz;
3981: if (realalloc) {
3982: MatSetOption(B,MAT_NEW_NONZERO_ALLOCATION_ERR,PETSC_TRUE);
3983: }
3984: B->was_assembled = PETSC_FALSE;
3985: B->assembled = PETSC_FALSE;
3986: return(0);
3987: }
3990: PetscErrorCode MatResetPreallocation_SeqAIJ(Mat A)
3991: {
3992: Mat_SeqAIJ *a;
3993: PetscInt i;
3999: /* Check local size. If zero, then return */
4000: if (!A->rmap->n) return(0);
4002: a = (Mat_SeqAIJ*)A->data;
4003: /* if no saved info, we error out */
4004: if (!a->ipre) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_NULL,"No saved preallocation info \n");
4006: if (!a->i || !a->j || !a->a || !a->imax || !a->ilen) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_NULL,"Memory info is incomplete, and can not reset preallocation \n");
4008: PetscArraycpy(a->imax,a->ipre,A->rmap->n);
4009: PetscArrayzero(a->ilen,A->rmap->n);
4010: a->i[0] = 0;
4011: for (i=1; i<A->rmap->n+1; i++) {
4012: a->i[i] = a->i[i-1] + a->imax[i-1];
4013: }
4014: A->preallocated = PETSC_TRUE;
4015: a->nz = 0;
4016: a->maxnz = a->i[A->rmap->n];
4017: A->info.nz_unneeded = (double)a->maxnz;
4018: A->was_assembled = PETSC_FALSE;
4019: A->assembled = PETSC_FALSE;
4020: return(0);
4021: }
4023: /*@
4024: MatSeqAIJSetPreallocationCSR - Allocates memory for a sparse sequential matrix in AIJ format.
4026: Input Parameters:
4027: + B - the matrix
4028: . i - the indices into j for the start of each row (starts with zero)
4029: . j - the column indices for each row (starts with zero) these must be sorted for each row
4030: - v - optional values in the matrix
4032: Level: developer
4034: The i,j,v values are COPIED with this routine; to avoid the copy use MatCreateSeqAIJWithArrays()
4036: .seealso: MatCreate(), MatCreateSeqAIJ(), MatSetValues(), MatSeqAIJSetPreallocation(), MatCreateSeqAIJ(), MATSEQAIJ
4037: @*/
4038: PetscErrorCode MatSeqAIJSetPreallocationCSR(Mat B,const PetscInt i[],const PetscInt j[],const PetscScalar v[])
4039: {
4045: PetscTryMethod(B,"MatSeqAIJSetPreallocationCSR_C",(Mat,const PetscInt[],const PetscInt[],const PetscScalar[]),(B,i,j,v));
4046: return(0);
4047: }
4049: PetscErrorCode MatSeqAIJSetPreallocationCSR_SeqAIJ(Mat B,const PetscInt Ii[],const PetscInt J[],const PetscScalar v[])
4050: {
4051: PetscInt i;
4052: PetscInt m,n;
4053: PetscInt nz;
4054: PetscInt *nnz, nz_max = 0;
4058: if (Ii[0]) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE, "Ii[0] must be 0 it is %D", Ii[0]);
4060: PetscLayoutSetUp(B->rmap);
4061: PetscLayoutSetUp(B->cmap);
4063: MatGetSize(B, &m, &n);
4064: PetscMalloc1(m+1, &nnz);
4065: for (i = 0; i < m; i++) {
4066: nz = Ii[i+1]- Ii[i];
4067: nz_max = PetscMax(nz_max, nz);
4068: if (nz < 0) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE, "Local row %D has a negative number of columns %D", i, nnz);
4069: nnz[i] = nz;
4070: }
4071: MatSeqAIJSetPreallocation(B, 0, nnz);
4072: PetscFree(nnz);
4074: for (i = 0; i < m; i++) {
4075: MatSetValues_SeqAIJ(B, 1, &i, Ii[i+1] - Ii[i], J+Ii[i], v ? v + Ii[i] : NULL, INSERT_VALUES);
4076: }
4078: MatAssemblyBegin(B,MAT_FINAL_ASSEMBLY);
4079: MatAssemblyEnd(B,MAT_FINAL_ASSEMBLY);
4081: MatSetOption(B,MAT_NEW_NONZERO_LOCATION_ERR,PETSC_TRUE);
4082: return(0);
4083: }
4085: #include <../src/mat/impls/dense/seq/dense.h>
4086: #include <petsc/private/kernels/petscaxpy.h>
4088: /*
4089: Computes (B'*A')' since computing B*A directly is untenable
4091: n p p
4092: ( ) ( ) ( )
4093: m ( A ) * n ( B ) = m ( C )
4094: ( ) ( ) ( )
4096: */
4097: PetscErrorCode MatMatMultNumeric_SeqDense_SeqAIJ(Mat A,Mat B,Mat C)
4098: {
4099: PetscErrorCode ierr;
4100: Mat_SeqDense *sub_a = (Mat_SeqDense*)A->data;
4101: Mat_SeqAIJ *sub_b = (Mat_SeqAIJ*)B->data;
4102: Mat_SeqDense *sub_c = (Mat_SeqDense*)C->data;
4103: PetscInt i,n,m,q,p;
4104: const PetscInt *ii,*idx;
4105: const PetscScalar *b,*a,*a_q;
4106: PetscScalar *c,*c_q;
4109: m = A->rmap->n;
4110: n = A->cmap->n;
4111: p = B->cmap->n;
4112: a = sub_a->v;
4113: b = sub_b->a;
4114: c = sub_c->v;
4115: PetscArrayzero(c,m*p);
4117: ii = sub_b->i;
4118: idx = sub_b->j;
4119: for (i=0; i<n; i++) {
4120: q = ii[i+1] - ii[i];
4121: while (q-->0) {
4122: c_q = c + m*(*idx);
4123: a_q = a + m*i;
4124: PetscKernelAXPY(c_q,*b,a_q,m);
4125: idx++;
4126: b++;
4127: }
4128: }
4129: return(0);
4130: }
4132: PetscErrorCode MatMatMultSymbolic_SeqDense_SeqAIJ(Mat A,Mat B,PetscReal fill,Mat C)
4133: {
4135: PetscInt m=A->rmap->n,n=B->cmap->n;
4138: 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);
4139: MatSetSizes(C,m,n,m,n);
4140: MatSetBlockSizesFromMats(C,A,B);
4141: MatSetType(C,MATSEQDENSE);
4142: MatSeqDenseSetPreallocation(C,NULL);
4144: C->ops->matmultnumeric = MatMatMultNumeric_SeqDense_SeqAIJ;
4145: return(0);
4146: }
4148: /* ----------------------------------------------------------------*/
4149: /*MC
4150: MATSEQAIJ - MATSEQAIJ = "seqaij" - A matrix type to be used for sequential sparse matrices,
4151: based on compressed sparse row format.
4153: Options Database Keys:
4154: . -mat_type seqaij - sets the matrix type to "seqaij" during a call to MatSetFromOptions()
4156: Level: beginner
4158: Notes:
4159: MatSetValues() may be called for this matrix type with a NULL argument for the numerical values,
4160: in this case the values associated with the rows and columns one passes in are set to zero
4161: in the matrix
4163: MatSetOptions(,MAT_STRUCTURE_ONLY,PETSC_TRUE) may be called for this matrix type. In this no
4164: space is allocated for the nonzero entries and any entries passed with MatSetValues() are ignored
4166: Developer Notes:
4167: It would be nice if all matrix formats supported passing NULL in for the numerical values
4169: .seealso: MatCreateSeqAIJ(), MatSetFromOptions(), MatSetType(), MatCreate(), MatType
4170: M*/
4172: /*MC
4173: MATAIJ - MATAIJ = "aij" - A matrix type to be used for sparse matrices.
4175: This matrix type is identical to MATSEQAIJ when constructed with a single process communicator,
4176: and MATMPIAIJ otherwise. As a result, for single process communicators,
4177: MatSeqAIJSetPreallocation is supported, and similarly MatMPIAIJSetPreallocation() is supported
4178: for communicators controlling multiple processes. It is recommended that you call both of
4179: the above preallocation routines for simplicity.
4181: Options Database Keys:
4182: . -mat_type aij - sets the matrix type to "aij" during a call to MatSetFromOptions()
4184: Developer Notes:
4185: Subclasses include MATAIJCUSPARSE, MATAIJPERM, MATAIJSELL, MATAIJMKL, MATAIJCRL, and also automatically switches over to use inodes when
4186: enough exist.
4188: Level: beginner
4190: .seealso: MatCreateAIJ(), MatCreateSeqAIJ(), MATSEQAIJ,MATMPIAIJ
4191: M*/
4193: /*MC
4194: MATAIJCRL - MATAIJCRL = "aijcrl" - A matrix type to be used for sparse matrices.
4196: This matrix type is identical to MATSEQAIJCRL when constructed with a single process communicator,
4197: and MATMPIAIJCRL otherwise. As a result, for single process communicators,
4198: MatSeqAIJSetPreallocation() is supported, and similarly MatMPIAIJSetPreallocation() is supported
4199: for communicators controlling multiple processes. It is recommended that you call both of
4200: the above preallocation routines for simplicity.
4202: Options Database Keys:
4203: . -mat_type aijcrl - sets the matrix type to "aijcrl" during a call to MatSetFromOptions()
4205: Level: beginner
4207: .seealso: MatCreateMPIAIJCRL,MATSEQAIJCRL,MATMPIAIJCRL, MATSEQAIJCRL, MATMPIAIJCRL
4208: M*/
4210: PETSC_INTERN PetscErrorCode MatConvert_SeqAIJ_SeqAIJCRL(Mat,MatType,MatReuse,Mat*);
4211: #if defined(PETSC_HAVE_ELEMENTAL)
4212: PETSC_INTERN PetscErrorCode MatConvert_SeqAIJ_Elemental(Mat,MatType,MatReuse,Mat*);
4213: #endif
4214: #if defined(PETSC_HAVE_HYPRE)
4215: PETSC_INTERN PetscErrorCode MatConvert_AIJ_HYPRE(Mat A,MatType,MatReuse,Mat*);
4216: #endif
4217: PETSC_INTERN PetscErrorCode MatConvert_SeqAIJ_SeqDense(Mat,MatType,MatReuse,Mat*);
4219: PETSC_EXTERN PetscErrorCode MatConvert_SeqAIJ_SeqSELL(Mat,MatType,MatReuse,Mat*);
4220: PETSC_INTERN PetscErrorCode MatConvert_XAIJ_IS(Mat,MatType,MatReuse,Mat*);
4221: PETSC_INTERN PetscErrorCode MatProductSetFromOptions_IS_XAIJ(Mat);
4223: /*@C
4224: MatSeqAIJGetArray - gives read/write access to the array where the data for a MATSEQAIJ matrix is stored
4226: Not Collective
4228: Input Parameter:
4229: . mat - a MATSEQAIJ matrix
4231: Output Parameter:
4232: . array - pointer to the data
4234: Level: intermediate
4236: .seealso: MatSeqAIJRestoreArray(), MatSeqAIJGetArrayF90()
4237: @*/
4238: PetscErrorCode MatSeqAIJGetArray(Mat A,PetscScalar **array)
4239: {
4243: PetscUseMethod(A,"MatSeqAIJGetArray_C",(Mat,PetscScalar**),(A,array));
4244: return(0);
4245: }
4247: /*@C
4248: MatSeqAIJGetArrayRead - gives read-only access to the array where the data for a MATSEQAIJ matrix is stored
4250: Not Collective
4252: Input Parameter:
4253: . mat - a MATSEQAIJ matrix
4255: Output Parameter:
4256: . array - pointer to the data
4258: Level: intermediate
4260: .seealso: MatSeqAIJGetArray(), MatSeqAIJRestoreArrayRead()
4261: @*/
4262: PetscErrorCode MatSeqAIJGetArrayRead(Mat A,const PetscScalar **array)
4263: {
4264: #if defined(PETSC_HAVE_CUDA) || defined(PETSC_HAVE_VIENNACL)
4265: PetscOffloadMask oval;
4266: #endif
4270: #if defined(PETSC_HAVE_CUDA) || defined(PETSC_HAVE_VIENNACL)
4271: oval = A->offloadmask;
4272: #endif
4273: MatSeqAIJGetArray(A,(PetscScalar**)array);
4274: #if defined(PETSC_HAVE_CUDA) || defined(PETSC_HAVE_VIENNACL)
4275: if (oval == PETSC_OFFLOAD_GPU || oval == PETSC_OFFLOAD_BOTH) A->offloadmask = PETSC_OFFLOAD_BOTH;
4276: #endif
4277: return(0);
4278: }
4280: /*@C
4281: MatSeqAIJRestoreArrayRead - restore the read-only access array obtained from MatSeqAIJGetArrayRead
4283: Not Collective
4285: Input Parameter:
4286: . mat - a MATSEQAIJ matrix
4288: Output Parameter:
4289: . array - pointer to the data
4291: Level: intermediate
4293: .seealso: MatSeqAIJGetArray(), MatSeqAIJGetArrayRead()
4294: @*/
4295: PetscErrorCode MatSeqAIJRestoreArrayRead(Mat A,const PetscScalar **array)
4296: {
4297: #if defined(PETSC_HAVE_CUDA) || defined(PETSC_HAVE_VIENNACL)
4298: PetscOffloadMask oval;
4299: #endif
4303: #if defined(PETSC_HAVE_CUDA) || defined(PETSC_HAVE_VIENNACL)
4304: oval = A->offloadmask;
4305: #endif
4306: MatSeqAIJRestoreArray(A,(PetscScalar**)array);
4307: #if defined(PETSC_HAVE_CUDA) || defined(PETSC_HAVE_VIENNACL)
4308: A->offloadmask = oval;
4309: #endif
4310: return(0);
4311: }
4313: /*@C
4314: MatSeqAIJGetMaxRowNonzeros - returns the maximum number of nonzeros in any row
4316: Not Collective
4318: Input Parameter:
4319: . mat - a MATSEQAIJ matrix
4321: Output Parameter:
4322: . nz - the maximum number of nonzeros in any row
4324: Level: intermediate
4326: .seealso: MatSeqAIJRestoreArray(), MatSeqAIJGetArrayF90()
4327: @*/
4328: PetscErrorCode MatSeqAIJGetMaxRowNonzeros(Mat A,PetscInt *nz)
4329: {
4330: Mat_SeqAIJ *aij = (Mat_SeqAIJ*)A->data;
4333: *nz = aij->rmax;
4334: return(0);
4335: }
4337: /*@C
4338: MatSeqAIJRestoreArray - returns access to the array where the data for a MATSEQAIJ matrix is stored obtained by MatSeqAIJGetArray()
4340: Not Collective
4342: Input Parameters:
4343: + mat - a MATSEQAIJ matrix
4344: - array - pointer to the data
4346: Level: intermediate
4348: .seealso: MatSeqAIJGetArray(), MatSeqAIJRestoreArrayF90()
4349: @*/
4350: PetscErrorCode MatSeqAIJRestoreArray(Mat A,PetscScalar **array)
4351: {
4355: PetscUseMethod(A,"MatSeqAIJRestoreArray_C",(Mat,PetscScalar**),(A,array));
4356: return(0);
4357: }
4359: #if defined(PETSC_HAVE_CUDA)
4360: PETSC_EXTERN PetscErrorCode MatConvert_SeqAIJ_SeqAIJCUSPARSE(Mat);
4361: #endif
4363: PETSC_EXTERN PetscErrorCode MatCreate_SeqAIJ(Mat B)
4364: {
4365: Mat_SeqAIJ *b;
4367: PetscMPIInt size;
4370: MPI_Comm_size(PetscObjectComm((PetscObject)B),&size);
4371: if (size > 1) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Comm must be of size 1");
4373: PetscNewLog(B,&b);
4375: B->data = (void*)b;
4377: PetscMemcpy(B->ops,&MatOps_Values,sizeof(struct _MatOps));
4378: if (B->sortedfull) B->ops->setvalues = MatSetValues_SeqAIJ_SortedFull;
4380: b->row = 0;
4381: b->col = 0;
4382: b->icol = 0;
4383: b->reallocs = 0;
4384: b->ignorezeroentries = PETSC_FALSE;
4385: b->roworiented = PETSC_TRUE;
4386: b->nonew = 0;
4387: b->diag = 0;
4388: b->solve_work = 0;
4389: B->spptr = 0;
4390: b->saved_values = 0;
4391: b->idiag = 0;
4392: b->mdiag = 0;
4393: b->ssor_work = 0;
4394: b->omega = 1.0;
4395: b->fshift = 0.0;
4396: b->idiagvalid = PETSC_FALSE;
4397: b->ibdiagvalid = PETSC_FALSE;
4398: b->keepnonzeropattern = PETSC_FALSE;
4400: PetscObjectChangeTypeName((PetscObject)B,MATSEQAIJ);
4401: PetscObjectComposeFunction((PetscObject)B,"MatSeqAIJGetArray_C",MatSeqAIJGetArray_SeqAIJ);
4402: PetscObjectComposeFunction((PetscObject)B,"MatSeqAIJRestoreArray_C",MatSeqAIJRestoreArray_SeqAIJ);
4404: #if defined(PETSC_HAVE_MATLAB_ENGINE)
4405: PetscObjectComposeFunction((PetscObject)B,"PetscMatlabEnginePut_C",MatlabEnginePut_SeqAIJ);
4406: PetscObjectComposeFunction((PetscObject)B,"PetscMatlabEngineGet_C",MatlabEngineGet_SeqAIJ);
4407: #endif
4409: PetscObjectComposeFunction((PetscObject)B,"MatSeqAIJSetColumnIndices_C",MatSeqAIJSetColumnIndices_SeqAIJ);
4410: PetscObjectComposeFunction((PetscObject)B,"MatStoreValues_C",MatStoreValues_SeqAIJ);
4411: PetscObjectComposeFunction((PetscObject)B,"MatRetrieveValues_C",MatRetrieveValues_SeqAIJ);
4412: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqaij_seqsbaij_C",MatConvert_SeqAIJ_SeqSBAIJ);
4413: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqaij_seqbaij_C",MatConvert_SeqAIJ_SeqBAIJ);
4414: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqaij_seqaijperm_C",MatConvert_SeqAIJ_SeqAIJPERM);
4415: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqaij_seqaijsell_C",MatConvert_SeqAIJ_SeqAIJSELL);
4416: #if defined(PETSC_HAVE_MKL_SPARSE)
4417: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqaij_seqaijmkl_C",MatConvert_SeqAIJ_SeqAIJMKL);
4418: #endif
4419: #if defined(PETSC_HAVE_CUDA)
4420: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqaij_seqaijcusparse_C",MatConvert_SeqAIJ_SeqAIJCUSPARSE);
4421: PetscObjectComposeFunction((PetscObject)B,"MatProductSetFromOptions_seqaijcusparse_seqaij_C",MatProductSetFromOptions_SeqAIJ);
4422: #endif
4423: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqaij_seqaijcrl_C",MatConvert_SeqAIJ_SeqAIJCRL);
4424: #if defined(PETSC_HAVE_ELEMENTAL)
4425: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqaij_elemental_C",MatConvert_SeqAIJ_Elemental);
4426: #endif
4427: #if defined(PETSC_HAVE_HYPRE)
4428: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqaij_hypre_C",MatConvert_AIJ_HYPRE);
4429: PetscObjectComposeFunction((PetscObject)B,"MatProductSetFromOptions_transpose_seqaij_seqaij_C",MatProductSetFromOptions_Transpose_AIJ_AIJ);
4430: #endif
4431: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqaij_seqdense_C",MatConvert_SeqAIJ_SeqDense);
4432: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqaij_seqsell_C",MatConvert_SeqAIJ_SeqSELL);
4433: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqaij_is_C",MatConvert_XAIJ_IS);
4434: PetscObjectComposeFunction((PetscObject)B,"MatIsTranspose_C",MatIsTranspose_SeqAIJ);
4435: PetscObjectComposeFunction((PetscObject)B,"MatIsHermitianTranspose_C",MatIsTranspose_SeqAIJ);
4436: PetscObjectComposeFunction((PetscObject)B,"MatSeqAIJSetPreallocation_C",MatSeqAIJSetPreallocation_SeqAIJ);
4437: PetscObjectComposeFunction((PetscObject)B,"MatResetPreallocation_C",MatResetPreallocation_SeqAIJ);
4438: PetscObjectComposeFunction((PetscObject)B,"MatSeqAIJSetPreallocationCSR_C",MatSeqAIJSetPreallocationCSR_SeqAIJ);
4439: PetscObjectComposeFunction((PetscObject)B,"MatReorderForNonzeroDiagonal_C",MatReorderForNonzeroDiagonal_SeqAIJ);
4440: PetscObjectComposeFunction((PetscObject)B,"MatMatMultSymbolic_seqdense_seqaij_C",MatMatMultSymbolic_SeqDense_SeqAIJ);
4441: PetscObjectComposeFunction((PetscObject)B,"MatMatMultNumeric_seqdense_seqaij_C",MatMatMultNumeric_SeqDense_SeqAIJ);
4442: PetscObjectComposeFunction((PetscObject)B,"MatProductSetFromOptions_is_seqaij_C",MatProductSetFromOptions_IS_XAIJ);
4443: PetscObjectComposeFunction((PetscObject)B,"MatProductSetFromOptions_seqdense_seqaij_C",MatProductSetFromOptions_SeqDense_SeqAIJ);
4444: PetscObjectComposeFunction((PetscObject)B,"MatProductSetFromOptions_seqaij_seqaij_C",MatProductSetFromOptions_SeqAIJ);
4445: MatCreate_SeqAIJ_Inode(B);
4446: PetscObjectChangeTypeName((PetscObject)B,MATSEQAIJ);
4447: MatSeqAIJSetTypeFromOptions(B); /* this allows changing the matrix subtype to say MATSEQAIJPERM */
4448: return(0);
4449: }
4451: /*
4452: Given a matrix generated with MatGetFactor() duplicates all the information in A into B
4453: */
4454: PetscErrorCode MatDuplicateNoCreate_SeqAIJ(Mat C,Mat A,MatDuplicateOption cpvalues,PetscBool mallocmatspace)
4455: {
4456: Mat_SeqAIJ *c,*a = (Mat_SeqAIJ*)A->data;
4458: PetscInt m = A->rmap->n,i;
4461: c = (Mat_SeqAIJ*)C->data;
4463: C->factortype = A->factortype;
4464: c->row = 0;
4465: c->col = 0;
4466: c->icol = 0;
4467: c->reallocs = 0;
4469: C->assembled = PETSC_TRUE;
4471: PetscLayoutReference(A->rmap,&C->rmap);
4472: PetscLayoutReference(A->cmap,&C->cmap);
4474: PetscMalloc1(m,&c->imax);
4475: PetscMemcpy(c->imax,a->imax,m*sizeof(PetscInt));
4476: PetscMalloc1(m,&c->ilen);
4477: PetscMemcpy(c->ilen,a->ilen,m*sizeof(PetscInt));
4478: PetscLogObjectMemory((PetscObject)C, 2*m*sizeof(PetscInt));
4480: /* allocate the matrix space */
4481: if (mallocmatspace) {
4482: PetscMalloc3(a->i[m],&c->a,a->i[m],&c->j,m+1,&c->i);
4483: PetscLogObjectMemory((PetscObject)C, a->i[m]*(sizeof(PetscScalar)+sizeof(PetscInt))+(m+1)*sizeof(PetscInt));
4485: c->singlemalloc = PETSC_TRUE;
4487: PetscArraycpy(c->i,a->i,m+1);
4488: if (m > 0) {
4489: PetscArraycpy(c->j,a->j,a->i[m]);
4490: if (cpvalues == MAT_COPY_VALUES) {
4491: PetscArraycpy(c->a,a->a,a->i[m]);
4492: } else {
4493: PetscArrayzero(c->a,a->i[m]);
4494: }
4495: }
4496: }
4498: c->ignorezeroentries = a->ignorezeroentries;
4499: c->roworiented = a->roworiented;
4500: c->nonew = a->nonew;
4501: if (a->diag) {
4502: PetscMalloc1(m+1,&c->diag);
4503: PetscMemcpy(c->diag,a->diag,m*sizeof(PetscInt));
4504: PetscLogObjectMemory((PetscObject)C,(m+1)*sizeof(PetscInt));
4505: } else c->diag = NULL;
4507: c->solve_work = 0;
4508: c->saved_values = 0;
4509: c->idiag = 0;
4510: c->ssor_work = 0;
4511: c->keepnonzeropattern = a->keepnonzeropattern;
4512: c->free_a = PETSC_TRUE;
4513: c->free_ij = PETSC_TRUE;
4515: c->rmax = a->rmax;
4516: c->nz = a->nz;
4517: c->maxnz = a->nz; /* Since we allocate exactly the right amount */
4518: C->preallocated = PETSC_TRUE;
4520: c->compressedrow.use = a->compressedrow.use;
4521: c->compressedrow.nrows = a->compressedrow.nrows;
4522: if (a->compressedrow.use) {
4523: i = a->compressedrow.nrows;
4524: PetscMalloc2(i+1,&c->compressedrow.i,i,&c->compressedrow.rindex);
4525: PetscArraycpy(c->compressedrow.i,a->compressedrow.i,i+1);
4526: PetscArraycpy(c->compressedrow.rindex,a->compressedrow.rindex,i);
4527: } else {
4528: c->compressedrow.use = PETSC_FALSE;
4529: c->compressedrow.i = NULL;
4530: c->compressedrow.rindex = NULL;
4531: }
4532: c->nonzerorowcnt = a->nonzerorowcnt;
4533: C->nonzerostate = A->nonzerostate;
4535: MatDuplicate_SeqAIJ_Inode(A,cpvalues,&C);
4536: PetscFunctionListDuplicate(((PetscObject)A)->qlist,&((PetscObject)C)->qlist);
4537: return(0);
4538: }
4540: PetscErrorCode MatDuplicate_SeqAIJ(Mat A,MatDuplicateOption cpvalues,Mat *B)
4541: {
4545: MatCreate(PetscObjectComm((PetscObject)A),B);
4546: MatSetSizes(*B,A->rmap->n,A->cmap->n,A->rmap->n,A->cmap->n);
4547: if (!(A->rmap->n % A->rmap->bs) && !(A->cmap->n % A->cmap->bs)) {
4548: MatSetBlockSizesFromMats(*B,A,A);
4549: }
4550: MatSetType(*B,((PetscObject)A)->type_name);
4551: MatDuplicateNoCreate_SeqAIJ(*B,A,cpvalues,PETSC_TRUE);
4552: return(0);
4553: }
4555: PetscErrorCode MatLoad_SeqAIJ(Mat newMat, PetscViewer viewer)
4556: {
4557: PetscBool isbinary, ishdf5;
4563: /* force binary viewer to load .info file if it has not yet done so */
4564: PetscViewerSetUp(viewer);
4565: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERBINARY,&isbinary);
4566: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERHDF5, &ishdf5);
4567: if (isbinary) {
4568: MatLoad_SeqAIJ_Binary(newMat,viewer);
4569: } else if (ishdf5) {
4570: #if defined(PETSC_HAVE_HDF5)
4571: MatLoad_AIJ_HDF5(newMat,viewer);
4572: #else
4573: SETERRQ(PetscObjectComm((PetscObject)newMat),PETSC_ERR_SUP,"HDF5 not supported in this build.\nPlease reconfigure using --download-hdf5");
4574: #endif
4575: } else {
4576: 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);
4577: }
4578: return(0);
4579: }
4581: PetscErrorCode MatLoad_SeqAIJ_Binary(Mat mat, PetscViewer viewer)
4582: {
4583: Mat_SeqAIJ *a = (Mat_SeqAIJ*)mat->data;
4585: PetscInt header[4],*rowlens,M,N,nz,sum,rows,cols,i;
4588: PetscViewerSetUp(viewer);
4590: /* read in matrix header */
4591: PetscViewerBinaryRead(viewer,header,4,NULL,PETSC_INT);
4592: if (header[0] != MAT_FILE_CLASSID) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_FILE_UNEXPECTED,"Not a matrix object in file");
4593: M = header[1]; N = header[2]; nz = header[3];
4594: if (M < 0) SETERRQ1(PetscObjectComm((PetscObject)viewer),PETSC_ERR_FILE_UNEXPECTED,"Matrix row size (%D) in file is negative",M);
4595: if (N < 0) SETERRQ1(PetscObjectComm((PetscObject)viewer),PETSC_ERR_FILE_UNEXPECTED,"Matrix column size (%D) in file is negative",N);
4596: if (nz < 0) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_FILE_UNEXPECTED,"Matrix stored in special format on disk, cannot load as SeqAIJ");
4598: /* set block sizes from the viewer's .info file */
4599: MatLoad_Binary_BlockSizes(mat,viewer);
4600: /* set local and global sizes if not set already */
4601: if (mat->rmap->n < 0) mat->rmap->n = M;
4602: if (mat->cmap->n < 0) mat->cmap->n = N;
4603: if (mat->rmap->N < 0) mat->rmap->N = M;
4604: if (mat->cmap->N < 0) mat->cmap->N = N;
4605: PetscLayoutSetUp(mat->rmap);
4606: PetscLayoutSetUp(mat->cmap);
4608: /* check if the matrix sizes are correct */
4609: MatGetSize(mat,&rows,&cols);
4610: if (M != rows || N != cols) SETERRQ4(PETSC_COMM_SELF,PETSC_ERR_FILE_UNEXPECTED, "Matrix in file of different sizes (%D, %D) than the input matrix (%D, %D)",M,N,rows,cols);
4612: /* read in row lengths */
4613: PetscMalloc1(M,&rowlens);
4614: PetscViewerBinaryRead(viewer,rowlens,M,NULL,PETSC_INT);
4615: /* check if sum(rowlens) is same as nz */
4616: sum = 0; for (i=0; i<M; i++) sum += rowlens[i];
4617: if (sum != nz) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_FILE_UNEXPECTED,"Inconsistent matrix data in file: nonzeros = %D, sum-row-lengths = %D\n",nz,sum);
4618: /* preallocate and check sizes */
4619: MatSeqAIJSetPreallocation_SeqAIJ(mat,0,rowlens);
4620: MatGetSize(mat,&rows,&cols);
4621: if (M != rows || N != cols) SETERRQ4(PETSC_COMM_SELF,PETSC_ERR_FILE_UNEXPECTED, "Matrix in file of different length (%D, %D) than the input matrix (%D, %D)",M,N,rows,cols);
4622: /* store row lengths */
4623: PetscArraycpy(a->ilen,rowlens,M);
4624: PetscFree(rowlens);
4626: /* fill in "i" row pointers */
4627: a->i[0] = 0; for (i=0; i<M; i++) a->i[i+1] = a->i[i] + a->ilen[i];
4628: /* read in "j" column indices */
4629: PetscViewerBinaryRead(viewer,a->j,nz,NULL,PETSC_INT);
4630: /* read in "a" nonzero values */
4631: PetscViewerBinaryRead(viewer,a->a,nz,NULL,PETSC_SCALAR);
4633: MatAssemblyBegin(mat,MAT_FINAL_ASSEMBLY);
4634: MatAssemblyEnd(mat,MAT_FINAL_ASSEMBLY);
4635: return(0);
4636: }
4638: PetscErrorCode MatEqual_SeqAIJ(Mat A,Mat B,PetscBool * flg)
4639: {
4640: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data,*b = (Mat_SeqAIJ*)B->data;
4642: #if defined(PETSC_USE_COMPLEX)
4643: PetscInt k;
4644: #endif
4647: /* If the matrix dimensions are not equal,or no of nonzeros */
4648: if ((A->rmap->n != B->rmap->n) || (A->cmap->n != B->cmap->n) ||(a->nz != b->nz)) {
4649: *flg = PETSC_FALSE;
4650: return(0);
4651: }
4653: /* if the a->i are the same */
4654: PetscArraycmp(a->i,b->i,A->rmap->n+1,flg);
4655: if (!*flg) return(0);
4657: /* if a->j are the same */
4658: PetscArraycmp(a->j,b->j,a->nz,flg);
4659: if (!*flg) return(0);
4661: /* if a->a are the same */
4662: #if defined(PETSC_USE_COMPLEX)
4663: for (k=0; k<a->nz; k++) {
4664: if (PetscRealPart(a->a[k]) != PetscRealPart(b->a[k]) || PetscImaginaryPart(a->a[k]) != PetscImaginaryPart(b->a[k])) {
4665: *flg = PETSC_FALSE;
4666: return(0);
4667: }
4668: }
4669: #else
4670: PetscArraycmp(a->a,b->a,a->nz,flg);
4671: #endif
4672: return(0);
4673: }
4675: /*@
4676: MatCreateSeqAIJWithArrays - Creates an sequential AIJ matrix using matrix elements (in CSR format)
4677: provided by the user.
4679: Collective
4681: Input Parameters:
4682: + comm - must be an MPI communicator of size 1
4683: . m - number of rows
4684: . n - number of columns
4685: . i - row indices; that is i[0] = 0, i[row] = i[row-1] + number of elements in that row of the matrix
4686: . j - column indices
4687: - a - matrix values
4689: Output Parameter:
4690: . mat - the matrix
4692: Level: intermediate
4694: Notes:
4695: The i, j, and a arrays are not copied by this routine, the user must free these arrays
4696: once the matrix is destroyed and not before
4698: You cannot set new nonzero locations into this matrix, that will generate an error.
4700: The i and j indices are 0 based
4702: The format which is used for the sparse matrix input, is equivalent to a
4703: row-major ordering.. i.e for the following matrix, the input data expected is
4704: as shown
4706: $ 1 0 0
4707: $ 2 0 3
4708: $ 4 5 6
4709: $
4710: $ i = {0,1,3,6} [size = nrow+1 = 3+1]
4711: $ j = {0,0,2,0,1,2} [size = 6]; values must be sorted for each row
4712: $ v = {1,2,3,4,5,6} [size = 6]
4715: .seealso: MatCreate(), MatCreateAIJ(), MatCreateSeqAIJ(), MatCreateMPIAIJWithArrays(), MatMPIAIJSetPreallocationCSR()
4717: @*/
4718: PetscErrorCode MatCreateSeqAIJWithArrays(MPI_Comm comm,PetscInt m,PetscInt n,PetscInt i[],PetscInt j[],PetscScalar a[],Mat *mat)
4719: {
4721: PetscInt ii;
4722: Mat_SeqAIJ *aij;
4723: #if defined(PETSC_USE_DEBUG)
4724: PetscInt jj;
4725: #endif
4728: if (m > 0 && i[0]) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"i (row indices) must start with 0");
4729: MatCreate(comm,mat);
4730: MatSetSizes(*mat,m,n,m,n);
4731: /* MatSetBlockSizes(*mat,,); */
4732: MatSetType(*mat,MATSEQAIJ);
4733: MatSeqAIJSetPreallocation_SeqAIJ(*mat,MAT_SKIP_ALLOCATION,0);
4734: aij = (Mat_SeqAIJ*)(*mat)->data;
4735: PetscMalloc1(m,&aij->imax);
4736: PetscMalloc1(m,&aij->ilen);
4738: aij->i = i;
4739: aij->j = j;
4740: aij->a = a;
4741: aij->singlemalloc = PETSC_FALSE;
4742: aij->nonew = -1; /*this indicates that inserting a new value in the matrix that generates a new nonzero is an error*/
4743: aij->free_a = PETSC_FALSE;
4744: aij->free_ij = PETSC_FALSE;
4746: for (ii=0; ii<m; ii++) {
4747: aij->ilen[ii] = aij->imax[ii] = i[ii+1] - i[ii];
4748: #if defined(PETSC_USE_DEBUG)
4749: if (i[ii+1] - i[ii] < 0) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Negative row length in i (row indices) row = %D length = %D",ii,i[ii+1] - i[ii]);
4750: for (jj=i[ii]+1; jj<i[ii+1]; jj++) {
4751: if (j[jj] < j[jj-1]) SETERRQ3(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Column entry number %D (actual column %D) in row %D is not sorted",jj-i[ii],j[jj],ii);
4752: if (j[jj] == j[jj-1]) SETERRQ3(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Column entry number %D (actual column %D) in row %D is identical to previous entry",jj-i[ii],j[jj],ii);
4753: }
4754: #endif
4755: }
4756: #if defined(PETSC_USE_DEBUG)
4757: for (ii=0; ii<aij->i[m]; ii++) {
4758: if (j[ii] < 0) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Negative column index at location = %D index = %D",ii,j[ii]);
4759: if (j[ii] > n - 1) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Column index to large at location = %D index = %D",ii,j[ii]);
4760: }
4761: #endif
4763: MatAssemblyBegin(*mat,MAT_FINAL_ASSEMBLY);
4764: MatAssemblyEnd(*mat,MAT_FINAL_ASSEMBLY);
4765: return(0);
4766: }
4767: /*@C
4768: MatCreateSeqAIJFromTriple - Creates an sequential AIJ matrix using matrix elements (in COO format)
4769: provided by the user.
4771: Collective
4773: Input Parameters:
4774: + comm - must be an MPI communicator of size 1
4775: . m - number of rows
4776: . n - number of columns
4777: . i - row indices
4778: . j - column indices
4779: . a - matrix values
4780: . nz - number of nonzeros
4781: - idx - 0 or 1 based
4783: Output Parameter:
4784: . mat - the matrix
4786: Level: intermediate
4788: Notes:
4789: The i and j indices are 0 based
4791: The format which is used for the sparse matrix input, is equivalent to a
4792: row-major ordering.. i.e for the following matrix, the input data expected is
4793: as shown:
4795: 1 0 0
4796: 2 0 3
4797: 4 5 6
4799: i = {0,1,1,2,2,2}
4800: j = {0,0,2,0,1,2}
4801: v = {1,2,3,4,5,6}
4804: .seealso: MatCreate(), MatCreateAIJ(), MatCreateSeqAIJ(), MatCreateSeqAIJWithArrays(), MatMPIAIJSetPreallocationCSR()
4806: @*/
4807: PetscErrorCode MatCreateSeqAIJFromTriple(MPI_Comm comm,PetscInt m,PetscInt n,PetscInt i[],PetscInt j[],PetscScalar a[],Mat *mat,PetscInt nz,PetscBool idx)
4808: {
4810: PetscInt ii, *nnz, one = 1,row,col;
4814: PetscCalloc1(m,&nnz);
4815: for (ii = 0; ii < nz; ii++) {
4816: nnz[i[ii] - !!idx] += 1;
4817: }
4818: MatCreate(comm,mat);
4819: MatSetSizes(*mat,m,n,m,n);
4820: MatSetType(*mat,MATSEQAIJ);
4821: MatSeqAIJSetPreallocation_SeqAIJ(*mat,0,nnz);
4822: for (ii = 0; ii < nz; ii++) {
4823: if (idx) {
4824: row = i[ii] - 1;
4825: col = j[ii] - 1;
4826: } else {
4827: row = i[ii];
4828: col = j[ii];
4829: }
4830: MatSetValues(*mat,one,&row,one,&col,&a[ii],ADD_VALUES);
4831: }
4832: MatAssemblyBegin(*mat,MAT_FINAL_ASSEMBLY);
4833: MatAssemblyEnd(*mat,MAT_FINAL_ASSEMBLY);
4834: PetscFree(nnz);
4835: return(0);
4836: }
4838: PetscErrorCode MatSeqAIJInvalidateDiagonal(Mat A)
4839: {
4840: Mat_SeqAIJ *a=(Mat_SeqAIJ*)A->data;
4844: a->idiagvalid = PETSC_FALSE;
4845: a->ibdiagvalid = PETSC_FALSE;
4847: MatSeqAIJInvalidateDiagonal_Inode(A);
4848: return(0);
4849: }
4851: PetscErrorCode MatCreateMPIMatConcatenateSeqMat_SeqAIJ(MPI_Comm comm,Mat inmat,PetscInt n,MatReuse scall,Mat *outmat)
4852: {
4854: PetscMPIInt size;
4857: MPI_Comm_size(comm,&size);
4858: if (size == 1) {
4859: if (scall == MAT_INITIAL_MATRIX) {
4860: MatDuplicate(inmat,MAT_COPY_VALUES,outmat);
4861: } else {
4862: MatCopy(inmat,*outmat,SAME_NONZERO_PATTERN);
4863: }
4864: } else {
4865: MatCreateMPIMatConcatenateSeqMat_MPIAIJ(comm,inmat,n,scall,outmat);
4866: }
4867: return(0);
4868: }
4870: /*
4871: Permute A into C's *local* index space using rowemb,colemb.
4872: The embedding are supposed to be injections and the above implies that the range of rowemb is a subset
4873: of [0,m), colemb is in [0,n).
4874: If pattern == DIFFERENT_NONZERO_PATTERN, C is preallocated according to A.
4875: */
4876: PetscErrorCode MatSetSeqMat_SeqAIJ(Mat C,IS rowemb,IS colemb,MatStructure pattern,Mat B)
4877: {
4878: /* If making this function public, change the error returned in this function away from _PLIB. */
4880: Mat_SeqAIJ *Baij;
4881: PetscBool seqaij;
4882: PetscInt m,n,*nz,i,j,count;
4883: PetscScalar v;
4884: const PetscInt *rowindices,*colindices;
4887: if (!B) return(0);
4888: /* Check to make sure the target matrix (and embeddings) are compatible with C and each other. */
4889: PetscObjectBaseTypeCompare((PetscObject)B,MATSEQAIJ,&seqaij);
4890: if (!seqaij) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_PLIB,"Input matrix is of wrong type");
4891: if (rowemb) {
4892: ISGetLocalSize(rowemb,&m);
4893: if (m != B->rmap->n) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_PLIB,"Row IS of size %D is incompatible with matrix row size %D",m,B->rmap->n);
4894: } else {
4895: if (C->rmap->n != B->rmap->n) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_PLIB,"Input matrix is row-incompatible with the target matrix");
4896: }
4897: if (colemb) {
4898: ISGetLocalSize(colemb,&n);
4899: if (n != B->cmap->n) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_PLIB,"Diag col IS of size %D is incompatible with input matrix col size %D",n,B->cmap->n);
4900: } else {
4901: if (C->cmap->n != B->cmap->n) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_PLIB,"Input matrix is col-incompatible with the target matrix");
4902: }
4904: Baij = (Mat_SeqAIJ*)(B->data);
4905: if (pattern == DIFFERENT_NONZERO_PATTERN) {
4906: PetscMalloc1(B->rmap->n,&nz);
4907: for (i=0; i<B->rmap->n; i++) {
4908: nz[i] = Baij->i[i+1] - Baij->i[i];
4909: }
4910: MatSeqAIJSetPreallocation(C,0,nz);
4911: PetscFree(nz);
4912: }
4913: if (pattern == SUBSET_NONZERO_PATTERN) {
4914: MatZeroEntries(C);
4915: }
4916: count = 0;
4917: rowindices = NULL;
4918: colindices = NULL;
4919: if (rowemb) {
4920: ISGetIndices(rowemb,&rowindices);
4921: }
4922: if (colemb) {
4923: ISGetIndices(colemb,&colindices);
4924: }
4925: for (i=0; i<B->rmap->n; i++) {
4926: PetscInt row;
4927: row = i;
4928: if (rowindices) row = rowindices[i];
4929: for (j=Baij->i[i]; j<Baij->i[i+1]; j++) {
4930: PetscInt col;
4931: col = Baij->j[count];
4932: if (colindices) col = colindices[col];
4933: v = Baij->a[count];
4934: MatSetValues(C,1,&row,1,&col,&v,INSERT_VALUES);
4935: ++count;
4936: }
4937: }
4938: /* FIXME: set C's nonzerostate correctly. */
4939: /* Assembly for C is necessary. */
4940: C->preallocated = PETSC_TRUE;
4941: C->assembled = PETSC_TRUE;
4942: C->was_assembled = PETSC_FALSE;
4943: return(0);
4944: }
4946: PetscFunctionList MatSeqAIJList = NULL;
4948: /*@C
4949: MatSeqAIJSetType - Converts a MATSEQAIJ matrix to a subtype
4951: Collective on Mat
4953: Input Parameters:
4954: + mat - the matrix object
4955: - matype - matrix type
4957: Options Database Key:
4958: . -mat_seqai_type <method> - for example seqaijcrl
4961: Level: intermediate
4963: .seealso: PCSetType(), VecSetType(), MatCreate(), MatType, Mat
4964: @*/
4965: PetscErrorCode MatSeqAIJSetType(Mat mat, MatType matype)
4966: {
4967: PetscErrorCode ierr,(*r)(Mat,MatType,MatReuse,Mat*);
4968: PetscBool sametype;
4972: PetscObjectTypeCompare((PetscObject)mat,matype,&sametype);
4973: if (sametype) return(0);
4975: PetscFunctionListFind(MatSeqAIJList,matype,&r);
4976: if (!r) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_UNKNOWN_TYPE,"Unknown Mat type given: %s",matype);
4977: (*r)(mat,matype,MAT_INPLACE_MATRIX,&mat);
4978: return(0);
4979: }
4982: /*@C
4983: MatSeqAIJRegister - - Adds a new sub-matrix type for sequential AIJ matrices
4985: Not Collective
4987: Input Parameters:
4988: + name - name of a new user-defined matrix type, for example MATSEQAIJCRL
4989: - function - routine to convert to subtype
4991: Notes:
4992: MatSeqAIJRegister() may be called multiple times to add several user-defined solvers.
4995: Then, your matrix can be chosen with the procedural interface at runtime via the option
4996: $ -mat_seqaij_type my_mat
4998: Level: advanced
5000: .seealso: MatSeqAIJRegisterAll()
5003: Level: advanced
5004: @*/
5005: PetscErrorCode MatSeqAIJRegister(const char sname[],PetscErrorCode (*function)(Mat,MatType,MatReuse,Mat *))
5006: {
5010: MatInitializePackage();
5011: PetscFunctionListAdd(&MatSeqAIJList,sname,function);
5012: return(0);
5013: }
5015: PetscBool MatSeqAIJRegisterAllCalled = PETSC_FALSE;
5017: /*@C
5018: MatSeqAIJRegisterAll - Registers all of the matrix subtypes of SeqAIJ
5020: Not Collective
5022: Level: advanced
5024: Developers Note: CUSP and CUSPARSE do not yet support the MatConvert_SeqAIJ..() paradigm and thus cannot be registered here
5026: .seealso: MatRegisterAll(), MatSeqAIJRegister()
5027: @*/
5028: PetscErrorCode MatSeqAIJRegisterAll(void)
5029: {
5033: if (MatSeqAIJRegisterAllCalled) return(0);
5034: MatSeqAIJRegisterAllCalled = PETSC_TRUE;
5036: MatSeqAIJRegister(MATSEQAIJCRL, MatConvert_SeqAIJ_SeqAIJCRL);
5037: MatSeqAIJRegister(MATSEQAIJPERM, MatConvert_SeqAIJ_SeqAIJPERM);
5038: MatSeqAIJRegister(MATSEQAIJSELL, MatConvert_SeqAIJ_SeqAIJSELL);
5039: #if defined(PETSC_HAVE_MKL_SPARSE)
5040: MatSeqAIJRegister(MATSEQAIJMKL, MatConvert_SeqAIJ_SeqAIJMKL);
5041: #endif
5042: #if defined(PETSC_HAVE_VIENNACL) && defined(PETSC_HAVE_VIENNACL_NO_CUDA)
5043: MatSeqAIJRegister(MATMPIAIJVIENNACL, MatConvert_SeqAIJ_SeqAIJViennaCL);
5044: #endif
5045: return(0);
5046: }
5048: /*
5049: Special version for direct calls from Fortran
5050: */
5051: #include <petsc/private/fortranimpl.h>
5052: #if defined(PETSC_HAVE_FORTRAN_CAPS)
5053: #define matsetvaluesseqaij_ MATSETVALUESSEQAIJ
5054: #elif !defined(PETSC_HAVE_FORTRAN_UNDERSCORE)
5055: #define matsetvaluesseqaij_ matsetvaluesseqaij
5056: #endif
5058: /* Change these macros so can be used in void function */
5059: #undef CHKERRQ
5060: #define CHKERRQ(ierr) CHKERRABORT(PetscObjectComm((PetscObject)A),ierr)
5061: #undef SETERRQ2
5062: #define SETERRQ2(comm,ierr,b,c,d) CHKERRABORT(comm,ierr)
5063: #undef SETERRQ3
5064: #define SETERRQ3(comm,ierr,b,c,d,e) CHKERRABORT(comm,ierr)
5066: PETSC_EXTERN void matsetvaluesseqaij_(Mat *AA,PetscInt *mm,const PetscInt im[],PetscInt *nn,const PetscInt in[],const PetscScalar v[],InsertMode *isis, PetscErrorCode *_ierr)
5067: {
5068: Mat A = *AA;
5069: PetscInt m = *mm, n = *nn;
5070: InsertMode is = *isis;
5071: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
5072: PetscInt *rp,k,low,high,t,ii,row,nrow,i,col,l,rmax,N;
5073: PetscInt *imax,*ai,*ailen;
5075: PetscInt *aj,nonew = a->nonew,lastcol = -1;
5076: MatScalar *ap,value,*aa;
5077: PetscBool ignorezeroentries = a->ignorezeroentries;
5078: PetscBool roworiented = a->roworiented;
5081: MatCheckPreallocated(A,1);
5082: imax = a->imax;
5083: ai = a->i;
5084: ailen = a->ilen;
5085: aj = a->j;
5086: aa = a->a;
5088: for (k=0; k<m; k++) { /* loop over added rows */
5089: row = im[k];
5090: if (row < 0) continue;
5091: #if defined(PETSC_USE_DEBUG)
5092: if (row >= A->rmap->n) SETERRABORT(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_OUTOFRANGE,"Row too large");
5093: #endif
5094: rp = aj + ai[row]; ap = aa + ai[row];
5095: rmax = imax[row]; nrow = ailen[row];
5096: low = 0;
5097: high = nrow;
5098: for (l=0; l<n; l++) { /* loop over added columns */
5099: if (in[l] < 0) continue;
5100: #if defined(PETSC_USE_DEBUG)
5101: if (in[l] >= A->cmap->n) SETERRABORT(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_OUTOFRANGE,"Column too large");
5102: #endif
5103: col = in[l];
5104: if (roworiented) value = v[l + k*n];
5105: else value = v[k + l*m];
5107: if (value == 0.0 && ignorezeroentries && (is == ADD_VALUES)) continue;
5109: if (col <= lastcol) low = 0;
5110: else high = nrow;
5111: lastcol = col;
5112: while (high-low > 5) {
5113: t = (low+high)/2;
5114: if (rp[t] > col) high = t;
5115: else low = t;
5116: }
5117: for (i=low; i<high; i++) {
5118: if (rp[i] > col) break;
5119: if (rp[i] == col) {
5120: if (is == ADD_VALUES) ap[i] += value;
5121: else ap[i] = value;
5122: goto noinsert;
5123: }
5124: }
5125: if (value == 0.0 && ignorezeroentries) goto noinsert;
5126: if (nonew == 1) goto noinsert;
5127: if (nonew == -1) SETERRABORT(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_OUTOFRANGE,"Inserting a new nonzero in the matrix");
5128: MatSeqXAIJReallocateAIJ(A,A->rmap->n,1,nrow,row,col,rmax,aa,ai,aj,rp,ap,imax,nonew,MatScalar);
5129: N = nrow++ - 1; a->nz++; high++;
5130: /* shift up all the later entries in this row */
5131: for (ii=N; ii>=i; ii--) {
5132: rp[ii+1] = rp[ii];
5133: ap[ii+1] = ap[ii];
5134: }
5135: rp[i] = col;
5136: ap[i] = value;
5137: A->nonzerostate++;
5138: noinsert:;
5139: low = i + 1;
5140: }
5141: ailen[row] = nrow;
5142: }
5143: PetscFunctionReturnVoid();
5144: }