Actual source code: aij.c
petsc-3.11.0 2019-03-29
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: PetscMemzero(norms,n*sizeof(PetscReal));
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;
179: if (Y->assembled) {
180: MatMissingDiagonal_SeqAIJ(Y,&missing,NULL);
181: if (!missing) {
182: diag = aij->diag;
183: VecGetArrayRead(D,&v);
184: if (is == INSERT_VALUES) {
185: for (i=0; i<m; i++) {
186: aa[diag[i]] = v[i];
187: }
188: } else {
189: for (i=0; i<m; i++) {
190: aa[diag[i]] += v[i];
191: }
192: }
193: VecRestoreArrayRead(D,&v);
194: return(0);
195: }
196: MatSeqAIJInvalidateDiagonal(Y);
197: }
198: MatDiagonalSet_Default(Y,D,is);
199: return(0);
200: }
202: PetscErrorCode MatGetRowIJ_SeqAIJ(Mat A,PetscInt oshift,PetscBool symmetric,PetscBool inodecompressed,PetscInt *m,const PetscInt *ia[],const PetscInt *ja[],PetscBool *done)
203: {
204: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
206: PetscInt i,ishift;
209: *m = A->rmap->n;
210: if (!ia) return(0);
211: ishift = 0;
212: if (symmetric && !A->structurally_symmetric) {
213: MatToSymmetricIJ_SeqAIJ(A->rmap->n,a->i,a->j,PETSC_TRUE,ishift,oshift,(PetscInt**)ia,(PetscInt**)ja);
214: } else if (oshift == 1) {
215: PetscInt *tia;
216: PetscInt nz = a->i[A->rmap->n];
217: /* malloc space and add 1 to i and j indices */
218: PetscMalloc1(A->rmap->n+1,&tia);
219: for (i=0; i<A->rmap->n+1; i++) tia[i] = a->i[i] + 1;
220: *ia = tia;
221: if (ja) {
222: PetscInt *tja;
223: PetscMalloc1(nz+1,&tja);
224: for (i=0; i<nz; i++) tja[i] = a->j[i] + 1;
225: *ja = tja;
226: }
227: } else {
228: *ia = a->i;
229: if (ja) *ja = a->j;
230: }
231: return(0);
232: }
234: PetscErrorCode MatRestoreRowIJ_SeqAIJ(Mat A,PetscInt oshift,PetscBool symmetric,PetscBool inodecompressed,PetscInt *n,const PetscInt *ia[],const PetscInt *ja[],PetscBool *done)
235: {
239: if (!ia) return(0);
240: if ((symmetric && !A->structurally_symmetric) || oshift == 1) {
241: PetscFree(*ia);
242: if (ja) {PetscFree(*ja);}
243: }
244: return(0);
245: }
247: PetscErrorCode MatGetColumnIJ_SeqAIJ(Mat A,PetscInt oshift,PetscBool symmetric,PetscBool inodecompressed,PetscInt *nn,const PetscInt *ia[],const PetscInt *ja[],PetscBool *done)
248: {
249: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
251: PetscInt i,*collengths,*cia,*cja,n = A->cmap->n,m = A->rmap->n;
252: PetscInt nz = a->i[m],row,*jj,mr,col;
255: *nn = n;
256: if (!ia) return(0);
257: if (symmetric) {
258: MatToSymmetricIJ_SeqAIJ(A->rmap->n,a->i,a->j,PETSC_TRUE,0,oshift,(PetscInt**)ia,(PetscInt**)ja);
259: } else {
260: PetscCalloc1(n+1,&collengths);
261: PetscMalloc1(n+1,&cia);
262: PetscMalloc1(nz+1,&cja);
263: jj = a->j;
264: for (i=0; i<nz; i++) {
265: collengths[jj[i]]++;
266: }
267: cia[0] = oshift;
268: for (i=0; i<n; i++) {
269: cia[i+1] = cia[i] + collengths[i];
270: }
271: PetscMemzero(collengths,n*sizeof(PetscInt));
272: jj = a->j;
273: for (row=0; row<m; row++) {
274: mr = a->i[row+1] - a->i[row];
275: for (i=0; i<mr; i++) {
276: col = *jj++;
278: cja[cia[col] + collengths[col]++ - oshift] = row + oshift;
279: }
280: }
281: PetscFree(collengths);
282: *ia = cia; *ja = cja;
283: }
284: return(0);
285: }
287: PetscErrorCode MatRestoreColumnIJ_SeqAIJ(Mat A,PetscInt oshift,PetscBool symmetric,PetscBool inodecompressed,PetscInt *n,const PetscInt *ia[],const PetscInt *ja[],PetscBool *done)
288: {
292: if (!ia) return(0);
294: PetscFree(*ia);
295: PetscFree(*ja);
296: return(0);
297: }
299: /*
300: MatGetColumnIJ_SeqAIJ_Color() and MatRestoreColumnIJ_SeqAIJ_Color() are customized from
301: MatGetColumnIJ_SeqAIJ() and MatRestoreColumnIJ_SeqAIJ() by adding an output
302: spidx[], index of a->a, to be used in MatTransposeColoringCreate_SeqAIJ() and MatFDColoringCreate_SeqXAIJ()
303: */
304: PetscErrorCode MatGetColumnIJ_SeqAIJ_Color(Mat A,PetscInt oshift,PetscBool symmetric,PetscBool inodecompressed,PetscInt *nn,const PetscInt *ia[],const PetscInt *ja[],PetscInt *spidx[],PetscBool *done)
305: {
306: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
308: PetscInt i,*collengths,*cia,*cja,n = A->cmap->n,m = A->rmap->n;
309: PetscInt nz = a->i[m],row,*jj,mr,col;
310: PetscInt *cspidx;
313: *nn = n;
314: if (!ia) return(0);
316: PetscCalloc1(n+1,&collengths);
317: PetscMalloc1(n+1,&cia);
318: PetscMalloc1(nz+1,&cja);
319: PetscMalloc1(nz+1,&cspidx);
320: jj = a->j;
321: for (i=0; i<nz; i++) {
322: collengths[jj[i]]++;
323: }
324: cia[0] = oshift;
325: for (i=0; i<n; i++) {
326: cia[i+1] = cia[i] + collengths[i];
327: }
328: PetscMemzero(collengths,n*sizeof(PetscInt));
329: jj = a->j;
330: for (row=0; row<m; row++) {
331: mr = a->i[row+1] - a->i[row];
332: for (i=0; i<mr; i++) {
333: col = *jj++;
334: cspidx[cia[col] + collengths[col] - oshift] = a->i[row] + i; /* index of a->j */
335: cja[cia[col] + collengths[col]++ - oshift] = row + oshift;
336: }
337: }
338: PetscFree(collengths);
339: *ia = cia; *ja = cja;
340: *spidx = cspidx;
341: return(0);
342: }
344: PetscErrorCode MatRestoreColumnIJ_SeqAIJ_Color(Mat A,PetscInt oshift,PetscBool symmetric,PetscBool inodecompressed,PetscInt *n,const PetscInt *ia[],const PetscInt *ja[],PetscInt *spidx[],PetscBool *done)
345: {
349: MatRestoreColumnIJ_SeqAIJ(A,oshift,symmetric,inodecompressed,n,ia,ja,done);
350: PetscFree(*spidx);
351: return(0);
352: }
354: PetscErrorCode MatSetValuesRow_SeqAIJ(Mat A,PetscInt row,const PetscScalar v[])
355: {
356: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
357: PetscInt *ai = a->i;
361: PetscMemcpy(a->a+ai[row],v,(ai[row+1]-ai[row])*sizeof(PetscScalar));
362: return(0);
363: }
365: /*
366: MatSeqAIJSetValuesLocalFast - An optimized version of MatSetValuesLocal() for SeqAIJ matrices with several assumptions
368: - a single row of values is set with each call
369: - no row or column indices are negative or (in error) larger than the number of rows or columns
370: - the values are always added to the matrix, not set
371: - no new locations are introduced in the nonzero structure of the matrix
373: This does NOT assume the global column indices are sorted
375: */
377: #include <petsc/private/isimpl.h>
378: PetscErrorCode MatSeqAIJSetValuesLocalFast(Mat A,PetscInt m,const PetscInt im[],PetscInt n,const PetscInt in[],const PetscScalar v[],InsertMode is)
379: {
380: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
381: PetscInt low,high,t,row,nrow,i,col,l;
382: const PetscInt *rp,*ai = a->i,*ailen = a->ilen,*aj = a->j;
383: PetscInt lastcol = -1;
384: MatScalar *ap,value,*aa = a->a;
385: const PetscInt *ridx = A->rmap->mapping->indices,*cidx = A->cmap->mapping->indices;
387: row = ridx[im[0]];
388: rp = aj + ai[row];
389: ap = aa + ai[row];
390: nrow = ailen[row];
391: low = 0;
392: high = nrow;
393: for (l=0; l<n; l++) { /* loop over added columns */
394: col = cidx[in[l]];
395: value = v[l];
397: if (col <= lastcol) low = 0;
398: else high = nrow;
399: lastcol = col;
400: while (high-low > 5) {
401: t = (low+high)/2;
402: if (rp[t] > col) high = t;
403: else low = t;
404: }
405: for (i=low; i<high; i++) {
406: if (rp[i] == col) {
407: ap[i] += value;
408: low = i + 1;
409: break;
410: }
411: }
412: }
413: return 0;
414: }
416: PetscErrorCode MatSetValues_SeqAIJ(Mat A,PetscInt m,const PetscInt im[],PetscInt n,const PetscInt in[],const PetscScalar v[],InsertMode is)
417: {
418: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
419: PetscInt *rp,k,low,high,t,ii,row,nrow,i,col,l,rmax,N;
420: PetscInt *imax = a->imax,*ai = a->i,*ailen = a->ilen;
422: PetscInt *aj = a->j,nonew = a->nonew,lastcol = -1;
423: MatScalar *ap=NULL,value=0.0,*aa = a->a;
424: PetscBool ignorezeroentries = a->ignorezeroentries;
425: PetscBool roworiented = a->roworiented;
428: for (k=0; k<m; k++) { /* loop over added rows */
429: row = im[k];
430: if (row < 0) continue;
431: #if defined(PETSC_USE_DEBUG)
432: 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);
433: #endif
434: rp = aj + ai[row];
435: if (!A->structure_only) ap = aa + ai[row];
436: rmax = imax[row]; nrow = ailen[row];
437: low = 0;
438: high = nrow;
439: for (l=0; l<n; l++) { /* loop over added columns */
440: if (in[l] < 0) continue;
441: #if defined(PETSC_USE_DEBUG)
442: 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);
443: #endif
444: col = in[l];
445: if (!A->structure_only) {
446: if (roworiented) {
447: value = v[l + k*n];
448: } else {
449: value = v[k + l*m];
450: }
451: } else { /* A->structure_only */
452: value = 1; /* avoid 'continue' below? */
453: }
454: if ((value == 0.0 && ignorezeroentries) && (is == ADD_VALUES) && row != col) continue;
456: if (col <= lastcol) low = 0;
457: else high = nrow;
458: lastcol = col;
459: while (high-low > 5) {
460: t = (low+high)/2;
461: if (rp[t] > col) high = t;
462: else low = t;
463: }
464: for (i=low; i<high; i++) {
465: if (rp[i] > col) break;
466: if (rp[i] == col) {
467: if (!A->structure_only) {
468: if (is == ADD_VALUES) ap[i] += value;
469: else ap[i] = value;
470: }
471: low = i + 1;
472: goto noinsert;
473: }
474: }
475: if (value == 0.0 && ignorezeroentries && row != col) goto noinsert;
476: if (nonew == 1) goto noinsert;
477: if (nonew == -1) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Inserting a new nonzero at (%D,%D) in the matrix",row,col);
478: if (A->structure_only) {
479: MatSeqXAIJReallocateAIJ_structure_only(A,A->rmap->n,1,nrow,row,col,rmax,ai,aj,rp,imax,nonew,MatScalar);
480: } else {
481: MatSeqXAIJReallocateAIJ(A,A->rmap->n,1,nrow,row,col,rmax,aa,ai,aj,rp,ap,imax,nonew,MatScalar);
482: }
483: N = nrow++ - 1; a->nz++; high++;
484: /* shift up all the later entries in this row */
485: for (ii=N; ii>=i; ii--) {
486: rp[ii+1] = rp[ii];
487: if (!A->structure_only) ap[ii+1] = ap[ii];
488: }
489: rp[i] = col;
490: if (!A->structure_only) ap[i] = value;
491: low = i + 1;
492: A->nonzerostate++;
493: noinsert:;
494: }
495: ailen[row] = nrow;
496: }
497: return(0);
498: }
501: PetscErrorCode MatGetValues_SeqAIJ(Mat A,PetscInt m,const PetscInt im[],PetscInt n,const PetscInt in[],PetscScalar v[])
502: {
503: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
504: PetscInt *rp,k,low,high,t,row,nrow,i,col,l,*aj = a->j;
505: PetscInt *ai = a->i,*ailen = a->ilen;
506: MatScalar *ap,*aa = a->a;
509: for (k=0; k<m; k++) { /* loop over rows */
510: row = im[k];
511: if (row < 0) {v += n; continue;} /* SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Negative row: %D",row); */
512: 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);
513: rp = aj + ai[row]; ap = aa + ai[row];
514: nrow = ailen[row];
515: for (l=0; l<n; l++) { /* loop over columns */
516: if (in[l] < 0) {v++; continue;} /* SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Negative column: %D",in[l]); */
517: 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);
518: col = in[l];
519: high = nrow; low = 0; /* assume unsorted */
520: while (high-low > 5) {
521: t = (low+high)/2;
522: if (rp[t] > col) high = t;
523: else low = t;
524: }
525: for (i=low; i<high; i++) {
526: if (rp[i] > col) break;
527: if (rp[i] == col) {
528: *v++ = ap[i];
529: goto finished;
530: }
531: }
532: *v++ = 0.0;
533: finished:;
534: }
535: }
536: return(0);
537: }
540: PetscErrorCode MatView_SeqAIJ_Binary(Mat A,PetscViewer viewer)
541: {
542: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
544: PetscInt i,*col_lens;
545: int fd;
546: FILE *file;
549: PetscViewerBinaryGetDescriptor(viewer,&fd);
550: PetscMalloc1(4+A->rmap->n,&col_lens);
552: col_lens[0] = MAT_FILE_CLASSID;
553: col_lens[1] = A->rmap->n;
554: col_lens[2] = A->cmap->n;
555: col_lens[3] = a->nz;
557: /* store lengths of each row and write (including header) to file */
558: for (i=0; i<A->rmap->n; i++) {
559: col_lens[4+i] = a->i[i+1] - a->i[i];
560: }
561: PetscBinaryWrite(fd,col_lens,4+A->rmap->n,PETSC_INT,PETSC_TRUE);
562: PetscFree(col_lens);
564: /* store column indices (zero start index) */
565: PetscBinaryWrite(fd,a->j,a->nz,PETSC_INT,PETSC_FALSE);
567: /* store nonzero values */
568: PetscBinaryWrite(fd,a->a,a->nz,PETSC_SCALAR,PETSC_FALSE);
570: PetscViewerBinaryGetInfoPointer(viewer,&file);
571: if (file) {
572: fprintf(file,"-matload_block_size %d\n",(int)PetscAbs(A->rmap->bs));
573: }
574: return(0);
575: }
577: static PetscErrorCode MatView_SeqAIJ_ASCII_structonly(Mat A,PetscViewer viewer)
578: {
580: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
581: PetscInt i,k,m=A->rmap->N;
584: PetscViewerASCIIUseTabs(viewer,PETSC_FALSE);
585: for (i=0; i<m; i++) {
586: PetscViewerASCIIPrintf(viewer,"row %D:",i);
587: for (k=a->i[i]; k<a->i[i+1]; k++) {
588: PetscViewerASCIIPrintf(viewer," (%D) ",a->j[k]);
589: }
590: PetscViewerASCIIPrintf(viewer,"\n");
591: }
592: PetscViewerASCIIUseTabs(viewer,PETSC_TRUE);
593: return(0);
594: }
596: extern PetscErrorCode MatSeqAIJFactorInfo_Matlab(Mat,PetscViewer);
598: PetscErrorCode MatView_SeqAIJ_ASCII(Mat A,PetscViewer viewer)
599: {
600: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
601: PetscErrorCode ierr;
602: PetscInt i,j,m = A->rmap->n;
603: const char *name;
604: PetscViewerFormat format;
607: if (A->structure_only) {
608: MatView_SeqAIJ_ASCII_structonly(A,viewer);
609: return(0);
610: }
612: PetscViewerGetFormat(viewer,&format);
613: if (format == PETSC_VIEWER_ASCII_MATLAB) {
614: PetscInt nofinalvalue = 0;
615: if (m && ((a->i[m] == a->i[m-1]) || (a->j[a->nz-1] != A->cmap->n-1))) {
616: /* Need a dummy value to ensure the dimension of the matrix. */
617: nofinalvalue = 1;
618: }
619: PetscViewerASCIIUseTabs(viewer,PETSC_FALSE);
620: PetscViewerASCIIPrintf(viewer,"%% Size = %D %D \n",m,A->cmap->n);
621: PetscViewerASCIIPrintf(viewer,"%% Nonzeros = %D \n",a->nz);
622: #if defined(PETSC_USE_COMPLEX)
623: PetscViewerASCIIPrintf(viewer,"zzz = zeros(%D,4);\n",a->nz+nofinalvalue);
624: #else
625: PetscViewerASCIIPrintf(viewer,"zzz = zeros(%D,3);\n",a->nz+nofinalvalue);
626: #endif
627: PetscViewerASCIIPrintf(viewer,"zzz = [\n");
629: for (i=0; i<m; i++) {
630: for (j=a->i[i]; j<a->i[i+1]; j++) {
631: #if defined(PETSC_USE_COMPLEX)
632: 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]));
633: #else
634: PetscViewerASCIIPrintf(viewer,"%D %D %18.16e\n",i+1,a->j[j]+1,(double)a->a[j]);
635: #endif
636: }
637: }
638: if (nofinalvalue) {
639: #if defined(PETSC_USE_COMPLEX)
640: PetscViewerASCIIPrintf(viewer,"%D %D %18.16e %18.16e\n",m,A->cmap->n,0.,0.);
641: #else
642: PetscViewerASCIIPrintf(viewer,"%D %D %18.16e\n",m,A->cmap->n,0.0);
643: #endif
644: }
645: PetscObjectGetName((PetscObject)A,&name);
646: PetscViewerASCIIPrintf(viewer,"];\n %s = spconvert(zzz);\n",name);
647: PetscViewerASCIIUseTabs(viewer,PETSC_TRUE);
648: } else if (format == PETSC_VIEWER_ASCII_FACTOR_INFO || format == PETSC_VIEWER_ASCII_INFO) {
649: return(0);
650: } else if (format == PETSC_VIEWER_ASCII_COMMON) {
651: PetscViewerASCIIUseTabs(viewer,PETSC_FALSE);
652: for (i=0; i<m; i++) {
653: PetscViewerASCIIPrintf(viewer,"row %D:",i);
654: for (j=a->i[i]; j<a->i[i+1]; j++) {
655: #if defined(PETSC_USE_COMPLEX)
656: if (PetscImaginaryPart(a->a[j]) > 0.0 && PetscRealPart(a->a[j]) != 0.0) {
657: PetscViewerASCIIPrintf(viewer," (%D, %g + %g i)",a->j[j],(double)PetscRealPart(a->a[j]),(double)PetscImaginaryPart(a->a[j]));
658: } else if (PetscImaginaryPart(a->a[j]) < 0.0 && PetscRealPart(a->a[j]) != 0.0) {
659: PetscViewerASCIIPrintf(viewer," (%D, %g - %g i)",a->j[j],(double)PetscRealPart(a->a[j]),(double)-PetscImaginaryPart(a->a[j]));
660: } else if (PetscRealPart(a->a[j]) != 0.0) {
661: PetscViewerASCIIPrintf(viewer," (%D, %g) ",a->j[j],(double)PetscRealPart(a->a[j]));
662: }
663: #else
664: if (a->a[j] != 0.0) {PetscViewerASCIIPrintf(viewer," (%D, %g) ",a->j[j],(double)a->a[j]);}
665: #endif
666: }
667: PetscViewerASCIIPrintf(viewer,"\n");
668: }
669: PetscViewerASCIIUseTabs(viewer,PETSC_TRUE);
670: } else if (format == PETSC_VIEWER_ASCII_SYMMODU) {
671: PetscInt nzd=0,fshift=1,*sptr;
672: PetscViewerASCIIUseTabs(viewer,PETSC_FALSE);
673: PetscMalloc1(m+1,&sptr);
674: for (i=0; i<m; i++) {
675: sptr[i] = nzd+1;
676: for (j=a->i[i]; j<a->i[i+1]; j++) {
677: if (a->j[j] >= i) {
678: #if defined(PETSC_USE_COMPLEX)
679: if (PetscImaginaryPart(a->a[j]) != 0.0 || PetscRealPart(a->a[j]) != 0.0) nzd++;
680: #else
681: if (a->a[j] != 0.0) nzd++;
682: #endif
683: }
684: }
685: }
686: sptr[m] = nzd+1;
687: PetscViewerASCIIPrintf(viewer," %D %D\n\n",m,nzd);
688: for (i=0; i<m+1; i+=6) {
689: if (i+4<m) {
690: 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]);
691: } else if (i+3<m) {
692: PetscViewerASCIIPrintf(viewer," %D %D %D %D %D\n",sptr[i],sptr[i+1],sptr[i+2],sptr[i+3],sptr[i+4]);
693: } else if (i+2<m) {
694: PetscViewerASCIIPrintf(viewer," %D %D %D %D\n",sptr[i],sptr[i+1],sptr[i+2],sptr[i+3]);
695: } else if (i+1<m) {
696: PetscViewerASCIIPrintf(viewer," %D %D %D\n",sptr[i],sptr[i+1],sptr[i+2]);
697: } else if (i<m) {
698: PetscViewerASCIIPrintf(viewer," %D %D\n",sptr[i],sptr[i+1]);
699: } else {
700: PetscViewerASCIIPrintf(viewer," %D\n",sptr[i]);
701: }
702: }
703: PetscViewerASCIIPrintf(viewer,"\n");
704: PetscFree(sptr);
705: for (i=0; i<m; i++) {
706: for (j=a->i[i]; j<a->i[i+1]; j++) {
707: if (a->j[j] >= i) {PetscViewerASCIIPrintf(viewer," %D ",a->j[j]+fshift);}
708: }
709: PetscViewerASCIIPrintf(viewer,"\n");
710: }
711: PetscViewerASCIIPrintf(viewer,"\n");
712: for (i=0; i<m; i++) {
713: for (j=a->i[i]; j<a->i[i+1]; j++) {
714: if (a->j[j] >= i) {
715: #if defined(PETSC_USE_COMPLEX)
716: if (PetscImaginaryPart(a->a[j]) != 0.0 || PetscRealPart(a->a[j]) != 0.0) {
717: PetscViewerASCIIPrintf(viewer," %18.16e %18.16e ",(double)PetscRealPart(a->a[j]),(double)PetscImaginaryPart(a->a[j]));
718: }
719: #else
720: if (a->a[j] != 0.0) {PetscViewerASCIIPrintf(viewer," %18.16e ",(double)a->a[j]);}
721: #endif
722: }
723: }
724: PetscViewerASCIIPrintf(viewer,"\n");
725: }
726: PetscViewerASCIIUseTabs(viewer,PETSC_TRUE);
727: } else if (format == PETSC_VIEWER_ASCII_DENSE) {
728: PetscInt cnt = 0,jcnt;
729: PetscScalar value;
730: #if defined(PETSC_USE_COMPLEX)
731: PetscBool realonly = PETSC_TRUE;
733: for (i=0; i<a->i[m]; i++) {
734: if (PetscImaginaryPart(a->a[i]) != 0.0) {
735: realonly = PETSC_FALSE;
736: break;
737: }
738: }
739: #endif
741: PetscViewerASCIIUseTabs(viewer,PETSC_FALSE);
742: for (i=0; i<m; i++) {
743: jcnt = 0;
744: for (j=0; j<A->cmap->n; j++) {
745: if (jcnt < a->i[i+1]-a->i[i] && j == a->j[cnt]) {
746: value = a->a[cnt++];
747: jcnt++;
748: } else {
749: value = 0.0;
750: }
751: #if defined(PETSC_USE_COMPLEX)
752: if (realonly) {
753: PetscViewerASCIIPrintf(viewer," %7.5e ",(double)PetscRealPart(value));
754: } else {
755: PetscViewerASCIIPrintf(viewer," %7.5e+%7.5e i ",(double)PetscRealPart(value),(double)PetscImaginaryPart(value));
756: }
757: #else
758: PetscViewerASCIIPrintf(viewer," %7.5e ",(double)value);
759: #endif
760: }
761: PetscViewerASCIIPrintf(viewer,"\n");
762: }
763: PetscViewerASCIIUseTabs(viewer,PETSC_TRUE);
764: } else if (format == PETSC_VIEWER_ASCII_MATRIXMARKET) {
765: PetscInt fshift=1;
766: PetscViewerASCIIUseTabs(viewer,PETSC_FALSE);
767: #if defined(PETSC_USE_COMPLEX)
768: PetscViewerASCIIPrintf(viewer,"%%%%MatrixMarket matrix coordinate complex general\n");
769: #else
770: PetscViewerASCIIPrintf(viewer,"%%%%MatrixMarket matrix coordinate real general\n");
771: #endif
772: PetscViewerASCIIPrintf(viewer,"%D %D %D\n", m, A->cmap->n, a->nz);
773: for (i=0; i<m; i++) {
774: for (j=a->i[i]; j<a->i[i+1]; j++) {
775: #if defined(PETSC_USE_COMPLEX)
776: PetscViewerASCIIPrintf(viewer,"%D %D %g %g\n", i+fshift,a->j[j]+fshift,(double)PetscRealPart(a->a[j]),(double)PetscImaginaryPart(a->a[j]));
777: #else
778: PetscViewerASCIIPrintf(viewer,"%D %D %g\n", i+fshift, a->j[j]+fshift, (double)a->a[j]);
779: #endif
780: }
781: }
782: PetscViewerASCIIUseTabs(viewer,PETSC_TRUE);
783: } else {
784: PetscViewerASCIIUseTabs(viewer,PETSC_FALSE);
785: if (A->factortype) {
786: for (i=0; i<m; i++) {
787: PetscViewerASCIIPrintf(viewer,"row %D:",i);
788: /* L part */
789: for (j=a->i[i]; j<a->i[i+1]; j++) {
790: #if defined(PETSC_USE_COMPLEX)
791: if (PetscImaginaryPart(a->a[j]) > 0.0) {
792: PetscViewerASCIIPrintf(viewer," (%D, %g + %g i)",a->j[j],(double)PetscRealPart(a->a[j]),(double)PetscImaginaryPart(a->a[j]));
793: } else if (PetscImaginaryPart(a->a[j]) < 0.0) {
794: PetscViewerASCIIPrintf(viewer," (%D, %g - %g i)",a->j[j],(double)PetscRealPart(a->a[j]),(double)(-PetscImaginaryPart(a->a[j])));
795: } else {
796: PetscViewerASCIIPrintf(viewer," (%D, %g) ",a->j[j],(double)PetscRealPart(a->a[j]));
797: }
798: #else
799: PetscViewerASCIIPrintf(viewer," (%D, %g) ",a->j[j],(double)a->a[j]);
800: #endif
801: }
802: /* diagonal */
803: j = a->diag[i];
804: #if defined(PETSC_USE_COMPLEX)
805: if (PetscImaginaryPart(a->a[j]) > 0.0) {
806: PetscViewerASCIIPrintf(viewer," (%D, %g + %g i)",a->j[j],(double)PetscRealPart(1.0/a->a[j]),(double)PetscImaginaryPart(1.0/a->a[j]));
807: } else if (PetscImaginaryPart(a->a[j]) < 0.0) {
808: PetscViewerASCIIPrintf(viewer," (%D, %g - %g i)",a->j[j],(double)PetscRealPart(1.0/a->a[j]),(double)(-PetscImaginaryPart(1.0/a->a[j])));
809: } else {
810: PetscViewerASCIIPrintf(viewer," (%D, %g) ",a->j[j],(double)PetscRealPart(1.0/a->a[j]));
811: }
812: #else
813: PetscViewerASCIIPrintf(viewer," (%D, %g) ",a->j[j],(double)(1.0/a->a[j]));
814: #endif
816: /* U part */
817: for (j=a->diag[i+1]+1; j<a->diag[i]; j++) {
818: #if defined(PETSC_USE_COMPLEX)
819: if (PetscImaginaryPart(a->a[j]) > 0.0) {
820: PetscViewerASCIIPrintf(viewer," (%D, %g + %g i)",a->j[j],(double)PetscRealPart(a->a[j]),(double)PetscImaginaryPart(a->a[j]));
821: } else if (PetscImaginaryPart(a->a[j]) < 0.0) {
822: PetscViewerASCIIPrintf(viewer," (%D, %g - %g i)",a->j[j],(double)PetscRealPart(a->a[j]),(double)(-PetscImaginaryPart(a->a[j])));
823: } else {
824: PetscViewerASCIIPrintf(viewer," (%D, %g) ",a->j[j],(double)PetscRealPart(a->a[j]));
825: }
826: #else
827: PetscViewerASCIIPrintf(viewer," (%D, %g) ",a->j[j],(double)a->a[j]);
828: #endif
829: }
830: PetscViewerASCIIPrintf(viewer,"\n");
831: }
832: } else {
833: for (i=0; i<m; i++) {
834: PetscViewerASCIIPrintf(viewer,"row %D:",i);
835: for (j=a->i[i]; j<a->i[i+1]; j++) {
836: #if defined(PETSC_USE_COMPLEX)
837: if (PetscImaginaryPart(a->a[j]) > 0.0) {
838: PetscViewerASCIIPrintf(viewer," (%D, %g + %g i)",a->j[j],(double)PetscRealPart(a->a[j]),(double)PetscImaginaryPart(a->a[j]));
839: } else if (PetscImaginaryPart(a->a[j]) < 0.0) {
840: PetscViewerASCIIPrintf(viewer," (%D, %g - %g i)",a->j[j],(double)PetscRealPart(a->a[j]),(double)-PetscImaginaryPart(a->a[j]));
841: } else {
842: PetscViewerASCIIPrintf(viewer," (%D, %g) ",a->j[j],(double)PetscRealPart(a->a[j]));
843: }
844: #else
845: PetscViewerASCIIPrintf(viewer," (%D, %g) ",a->j[j],(double)a->a[j]);
846: #endif
847: }
848: PetscViewerASCIIPrintf(viewer,"\n");
849: }
850: }
851: PetscViewerASCIIUseTabs(viewer,PETSC_TRUE);
852: }
853: PetscViewerFlush(viewer);
854: return(0);
855: }
857: #include <petscdraw.h>
858: PetscErrorCode MatView_SeqAIJ_Draw_Zoom(PetscDraw draw,void *Aa)
859: {
860: Mat A = (Mat) Aa;
861: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
862: PetscErrorCode ierr;
863: PetscInt i,j,m = A->rmap->n;
864: int color;
865: PetscReal xl,yl,xr,yr,x_l,x_r,y_l,y_r;
866: PetscViewer viewer;
867: PetscViewerFormat format;
870: PetscObjectQuery((PetscObject)A,"Zoomviewer",(PetscObject*)&viewer);
871: PetscViewerGetFormat(viewer,&format);
872: PetscDrawGetCoordinates(draw,&xl,&yl,&xr,&yr);
874: /* loop over matrix elements drawing boxes */
876: if (format != PETSC_VIEWER_DRAW_CONTOUR) {
877: PetscDrawCollectiveBegin(draw);
878: /* Blue for negative, Cyan for zero and Red for positive */
879: color = PETSC_DRAW_BLUE;
880: for (i=0; i<m; i++) {
881: y_l = m - i - 1.0; y_r = y_l + 1.0;
882: for (j=a->i[i]; j<a->i[i+1]; j++) {
883: x_l = a->j[j]; x_r = x_l + 1.0;
884: if (PetscRealPart(a->a[j]) >= 0.) continue;
885: PetscDrawRectangle(draw,x_l,y_l,x_r,y_r,color,color,color,color);
886: }
887: }
888: color = PETSC_DRAW_CYAN;
889: for (i=0; i<m; i++) {
890: y_l = m - i - 1.0; y_r = y_l + 1.0;
891: for (j=a->i[i]; j<a->i[i+1]; j++) {
892: x_l = a->j[j]; x_r = x_l + 1.0;
893: if (a->a[j] != 0.) continue;
894: PetscDrawRectangle(draw,x_l,y_l,x_r,y_r,color,color,color,color);
895: }
896: }
897: color = PETSC_DRAW_RED;
898: for (i=0; i<m; i++) {
899: y_l = m - i - 1.0; y_r = y_l + 1.0;
900: for (j=a->i[i]; j<a->i[i+1]; j++) {
901: x_l = a->j[j]; x_r = x_l + 1.0;
902: if (PetscRealPart(a->a[j]) <= 0.) continue;
903: PetscDrawRectangle(draw,x_l,y_l,x_r,y_r,color,color,color,color);
904: }
905: }
906: PetscDrawCollectiveEnd(draw);
907: } else {
908: /* use contour shading to indicate magnitude of values */
909: /* first determine max of all nonzero values */
910: PetscReal minv = 0.0, maxv = 0.0;
911: PetscInt nz = a->nz, count = 0;
912: PetscDraw popup;
914: for (i=0; i<nz; i++) {
915: if (PetscAbsScalar(a->a[i]) > maxv) maxv = PetscAbsScalar(a->a[i]);
916: }
917: if (minv >= maxv) maxv = minv + PETSC_SMALL;
918: PetscDrawGetPopup(draw,&popup);
919: PetscDrawScalePopup(popup,minv,maxv);
921: PetscDrawCollectiveBegin(draw);
922: for (i=0; i<m; i++) {
923: y_l = m - i - 1.0;
924: y_r = y_l + 1.0;
925: for (j=a->i[i]; j<a->i[i+1]; j++) {
926: x_l = a->j[j];
927: x_r = x_l + 1.0;
928: color = PetscDrawRealToColor(PetscAbsScalar(a->a[count]),minv,maxv);
929: PetscDrawRectangle(draw,x_l,y_l,x_r,y_r,color,color,color,color);
930: count++;
931: }
932: }
933: PetscDrawCollectiveEnd(draw);
934: }
935: return(0);
936: }
938: #include <petscdraw.h>
939: PetscErrorCode MatView_SeqAIJ_Draw(Mat A,PetscViewer viewer)
940: {
942: PetscDraw draw;
943: PetscReal xr,yr,xl,yl,h,w;
944: PetscBool isnull;
947: PetscViewerDrawGetDraw(viewer,0,&draw);
948: PetscDrawIsNull(draw,&isnull);
949: if (isnull) return(0);
951: xr = A->cmap->n; yr = A->rmap->n; h = yr/10.0; w = xr/10.0;
952: xr += w; yr += h; xl = -w; yl = -h;
953: PetscDrawSetCoordinates(draw,xl,yl,xr,yr);
954: PetscObjectCompose((PetscObject)A,"Zoomviewer",(PetscObject)viewer);
955: PetscDrawZoom(draw,MatView_SeqAIJ_Draw_Zoom,A);
956: PetscObjectCompose((PetscObject)A,"Zoomviewer",NULL);
957: PetscDrawSave(draw);
958: return(0);
959: }
961: PetscErrorCode MatView_SeqAIJ(Mat A,PetscViewer viewer)
962: {
964: PetscBool iascii,isbinary,isdraw;
967: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERASCII,&iascii);
968: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERBINARY,&isbinary);
969: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERDRAW,&isdraw);
970: if (iascii) {
971: MatView_SeqAIJ_ASCII(A,viewer);
972: } else if (isbinary) {
973: MatView_SeqAIJ_Binary(A,viewer);
974: } else if (isdraw) {
975: MatView_SeqAIJ_Draw(A,viewer);
976: }
977: MatView_SeqAIJ_Inode(A,viewer);
978: return(0);
979: }
981: PetscErrorCode MatAssemblyEnd_SeqAIJ(Mat A,MatAssemblyType mode)
982: {
983: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
985: PetscInt fshift = 0,i,j,*ai = a->i,*aj = a->j,*imax = a->imax;
986: PetscInt m = A->rmap->n,*ip,N,*ailen = a->ilen,rmax = 0;
987: MatScalar *aa = a->a,*ap;
988: PetscReal ratio = 0.6;
991: if (mode == MAT_FLUSH_ASSEMBLY) return(0);
993: if (m) rmax = ailen[0]; /* determine row with most nonzeros */
994: for (i=1; i<m; i++) {
995: /* move each row back by the amount of empty slots (fshift) before it*/
996: fshift += imax[i-1] - ailen[i-1];
997: rmax = PetscMax(rmax,ailen[i]);
998: if (fshift) {
999: ip = aj + ai[i];
1000: ap = aa + ai[i];
1001: N = ailen[i];
1002: for (j=0; j<N; j++) {
1003: ip[j-fshift] = ip[j];
1004: if (!A->structure_only) ap[j-fshift] = ap[j];
1005: }
1006: }
1007: ai[i] = ai[i-1] + ailen[i-1];
1008: }
1009: if (m) {
1010: fshift += imax[m-1] - ailen[m-1];
1011: ai[m] = ai[m-1] + ailen[m-1];
1012: }
1014: /* reset ilen and imax for each row */
1015: a->nonzerorowcnt = 0;
1016: if (A->structure_only) {
1017: PetscFree2(a->imax,a->ilen);
1018: } else { /* !A->structure_only */
1019: for (i=0; i<m; i++) {
1020: ailen[i] = imax[i] = ai[i+1] - ai[i];
1021: a->nonzerorowcnt += ((ai[i+1] - ai[i]) > 0);
1022: }
1023: }
1024: a->nz = ai[m];
1025: 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);
1027: MatMarkDiagonal_SeqAIJ(A);
1028: PetscInfo4(A,"Matrix size: %D X %D; storage space: %D unneeded,%D used\n",m,A->cmap->n,fshift,a->nz);
1029: PetscInfo1(A,"Number of mallocs during MatSetValues() is %D\n",a->reallocs);
1030: PetscInfo1(A,"Maximum nonzeros in any row is %D\n",rmax);
1032: A->info.mallocs += a->reallocs;
1033: a->reallocs = 0;
1034: A->info.nz_unneeded = (PetscReal)fshift;
1035: a->rmax = rmax;
1037: if (!A->structure_only) {
1038: MatCheckCompressedRow(A,a->nonzerorowcnt,&a->compressedrow,a->i,m,ratio);
1039: }
1040: MatAssemblyEnd_SeqAIJ_Inode(A,mode);
1041: MatSeqAIJInvalidateDiagonal(A);
1042: return(0);
1043: }
1045: PetscErrorCode MatRealPart_SeqAIJ(Mat A)
1046: {
1047: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1048: PetscInt i,nz = a->nz;
1049: MatScalar *aa = a->a;
1053: for (i=0; i<nz; i++) aa[i] = PetscRealPart(aa[i]);
1054: MatSeqAIJInvalidateDiagonal(A);
1055: return(0);
1056: }
1058: PetscErrorCode MatImaginaryPart_SeqAIJ(Mat A)
1059: {
1060: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1061: PetscInt i,nz = a->nz;
1062: MatScalar *aa = a->a;
1066: for (i=0; i<nz; i++) aa[i] = PetscImaginaryPart(aa[i]);
1067: MatSeqAIJInvalidateDiagonal(A);
1068: return(0);
1069: }
1071: PetscErrorCode MatZeroEntries_SeqAIJ(Mat A)
1072: {
1073: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1077: PetscMemzero(a->a,(a->i[A->rmap->n])*sizeof(PetscScalar));
1078: MatSeqAIJInvalidateDiagonal(A);
1079: return(0);
1080: }
1082: PetscErrorCode MatDestroy_SeqAIJ(Mat A)
1083: {
1084: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1088: #if defined(PETSC_USE_LOG)
1089: PetscLogObjectState((PetscObject)A,"Rows=%D, Cols=%D, NZ=%D",A->rmap->n,A->cmap->n,a->nz);
1090: #endif
1091: MatSeqXAIJFreeAIJ(A,&a->a,&a->j,&a->i);
1092: ISDestroy(&a->row);
1093: ISDestroy(&a->col);
1094: PetscFree(a->diag);
1095: PetscFree(a->ibdiag);
1096: PetscFree2(a->imax,a->ilen);
1097: PetscFree(a->ipre);
1098: PetscFree3(a->idiag,a->mdiag,a->ssor_work);
1099: PetscFree(a->solve_work);
1100: ISDestroy(&a->icol);
1101: PetscFree(a->saved_values);
1102: ISColoringDestroy(&a->coloring);
1103: PetscFree2(a->compressedrow.i,a->compressedrow.rindex);
1104: PetscFree(a->matmult_abdense);
1106: MatDestroy_SeqAIJ_Inode(A);
1107: PetscFree(A->data);
1109: PetscObjectChangeTypeName((PetscObject)A,0);
1110: PetscObjectComposeFunction((PetscObject)A,"MatSeqAIJSetColumnIndices_C",NULL);
1111: PetscObjectComposeFunction((PetscObject)A,"MatStoreValues_C",NULL);
1112: PetscObjectComposeFunction((PetscObject)A,"MatRetrieveValues_C",NULL);
1113: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqaij_seqsbaij_C",NULL);
1114: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqaij_seqbaij_C",NULL);
1115: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqaij_seqaijperm_C",NULL);
1116: #if defined(PETSC_HAVE_ELEMENTAL)
1117: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqaij_elemental_C",NULL);
1118: #endif
1119: #if defined(PETSC_HAVE_HYPRE)
1120: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqaij_hypre_C",NULL);
1121: PetscObjectComposeFunction((PetscObject)A,"MatMatMatMult_transpose_seqaij_seqaij_C",NULL);
1122: #endif
1123: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqaij_seqdense_C",NULL);
1124: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqaij_seqsell_C",NULL);
1125: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqaij_is_C",NULL);
1126: PetscObjectComposeFunction((PetscObject)A,"MatIsTranspose_C",NULL);
1127: PetscObjectComposeFunction((PetscObject)A,"MatSeqAIJSetPreallocation_C",NULL);
1128: PetscObjectComposeFunction((PetscObject)A,"MatResetPreallocation_C",NULL);
1129: PetscObjectComposeFunction((PetscObject)A,"MatSeqAIJSetPreallocationCSR_C",NULL);
1130: PetscObjectComposeFunction((PetscObject)A,"MatReorderForNonzeroDiagonal_C",NULL);
1131: PetscObjectComposeFunction((PetscObject)A,"MatPtAP_is_seqaij_C",NULL);
1132: return(0);
1133: }
1135: PetscErrorCode MatSetOption_SeqAIJ(Mat A,MatOption op,PetscBool flg)
1136: {
1137: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1141: switch (op) {
1142: case MAT_ROW_ORIENTED:
1143: a->roworiented = flg;
1144: break;
1145: case MAT_KEEP_NONZERO_PATTERN:
1146: a->keepnonzeropattern = flg;
1147: break;
1148: case MAT_NEW_NONZERO_LOCATIONS:
1149: a->nonew = (flg ? 0 : 1);
1150: break;
1151: case MAT_NEW_NONZERO_LOCATION_ERR:
1152: a->nonew = (flg ? -1 : 0);
1153: break;
1154: case MAT_NEW_NONZERO_ALLOCATION_ERR:
1155: a->nonew = (flg ? -2 : 0);
1156: break;
1157: case MAT_UNUSED_NONZERO_LOCATION_ERR:
1158: a->nounused = (flg ? -1 : 0);
1159: break;
1160: case MAT_IGNORE_ZERO_ENTRIES:
1161: a->ignorezeroentries = flg;
1162: break;
1163: case MAT_SPD:
1164: case MAT_SYMMETRIC:
1165: case MAT_STRUCTURALLY_SYMMETRIC:
1166: case MAT_HERMITIAN:
1167: case MAT_SYMMETRY_ETERNAL:
1168: case MAT_STRUCTURE_ONLY:
1169: /* These options are handled directly by MatSetOption() */
1170: break;
1171: case MAT_NEW_DIAGONALS:
1172: case MAT_IGNORE_OFF_PROC_ENTRIES:
1173: case MAT_USE_HASH_TABLE:
1174: PetscInfo1(A,"Option %s ignored\n",MatOptions[op]);
1175: break;
1176: case MAT_USE_INODES:
1177: /* Not an error because MatSetOption_SeqAIJ_Inode handles this one */
1178: break;
1179: case MAT_SUBMAT_SINGLEIS:
1180: A->submat_singleis = flg;
1181: break;
1182: default:
1183: SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"unknown option %d",op);
1184: }
1185: MatSetOption_SeqAIJ_Inode(A,op,flg);
1186: return(0);
1187: }
1189: PetscErrorCode MatGetDiagonal_SeqAIJ(Mat A,Vec v)
1190: {
1191: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1193: PetscInt i,j,n,*ai=a->i,*aj=a->j,nz;
1194: PetscScalar *aa=a->a,*x,zero=0.0;
1197: VecGetLocalSize(v,&n);
1198: if (n != A->rmap->n) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Nonconforming matrix and vector");
1200: if (A->factortype == MAT_FACTOR_ILU || A->factortype == MAT_FACTOR_LU) {
1201: PetscInt *diag=a->diag;
1202: VecGetArray(v,&x);
1203: for (i=0; i<n; i++) x[i] = 1.0/aa[diag[i]];
1204: VecRestoreArray(v,&x);
1205: return(0);
1206: }
1208: VecSet(v,zero);
1209: VecGetArray(v,&x);
1210: for (i=0; i<n; i++) {
1211: nz = ai[i+1] - ai[i];
1212: if (!nz) x[i] = 0.0;
1213: for (j=ai[i]; j<ai[i+1]; j++) {
1214: if (aj[j] == i) {
1215: x[i] = aa[j];
1216: break;
1217: }
1218: }
1219: }
1220: VecRestoreArray(v,&x);
1221: return(0);
1222: }
1224: #include <../src/mat/impls/aij/seq/ftn-kernels/fmult.h>
1225: PetscErrorCode MatMultTransposeAdd_SeqAIJ(Mat A,Vec xx,Vec zz,Vec yy)
1226: {
1227: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1228: PetscScalar *y;
1229: const PetscScalar *x;
1230: PetscErrorCode ierr;
1231: PetscInt m = A->rmap->n;
1232: #if !defined(PETSC_USE_FORTRAN_KERNEL_MULTTRANSPOSEAIJ)
1233: const MatScalar *v;
1234: PetscScalar alpha;
1235: PetscInt n,i,j;
1236: const PetscInt *idx,*ii,*ridx=NULL;
1237: Mat_CompressedRow cprow = a->compressedrow;
1238: PetscBool usecprow = cprow.use;
1239: #endif
1242: if (zz != yy) {VecCopy(zz,yy);}
1243: VecGetArrayRead(xx,&x);
1244: VecGetArray(yy,&y);
1246: #if defined(PETSC_USE_FORTRAN_KERNEL_MULTTRANSPOSEAIJ)
1247: fortranmulttransposeaddaij_(&m,x,a->i,a->j,a->a,y);
1248: #else
1249: if (usecprow) {
1250: m = cprow.nrows;
1251: ii = cprow.i;
1252: ridx = cprow.rindex;
1253: } else {
1254: ii = a->i;
1255: }
1256: for (i=0; i<m; i++) {
1257: idx = a->j + ii[i];
1258: v = a->a + ii[i];
1259: n = ii[i+1] - ii[i];
1260: if (usecprow) {
1261: alpha = x[ridx[i]];
1262: } else {
1263: alpha = x[i];
1264: }
1265: for (j=0; j<n; j++) y[idx[j]] += alpha*v[j];
1266: }
1267: #endif
1268: PetscLogFlops(2.0*a->nz);
1269: VecRestoreArrayRead(xx,&x);
1270: VecRestoreArray(yy,&y);
1271: return(0);
1272: }
1274: PetscErrorCode MatMultTranspose_SeqAIJ(Mat A,Vec xx,Vec yy)
1275: {
1279: VecSet(yy,0.0);
1280: MatMultTransposeAdd_SeqAIJ(A,xx,yy,yy);
1281: return(0);
1282: }
1284: #include <../src/mat/impls/aij/seq/ftn-kernels/fmult.h>
1286: PetscErrorCode MatMult_SeqAIJ(Mat A,Vec xx,Vec yy)
1287: {
1288: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1289: PetscScalar *y;
1290: const PetscScalar *x;
1291: const MatScalar *aa;
1292: PetscErrorCode ierr;
1293: PetscInt m=A->rmap->n;
1294: const PetscInt *aj,*ii,*ridx=NULL;
1295: PetscInt n,i;
1296: PetscScalar sum;
1297: PetscBool usecprow=a->compressedrow.use;
1299: #if defined(PETSC_HAVE_PRAGMA_DISJOINT)
1300: #pragma disjoint(*x,*y,*aa)
1301: #endif
1304: VecGetArrayRead(xx,&x);
1305: VecGetArray(yy,&y);
1306: ii = a->i;
1307: if (usecprow) { /* use compressed row format */
1308: PetscMemzero(y,m*sizeof(PetscScalar));
1309: m = a->compressedrow.nrows;
1310: ii = a->compressedrow.i;
1311: ridx = a->compressedrow.rindex;
1312: for (i=0; i<m; i++) {
1313: n = ii[i+1] - ii[i];
1314: aj = a->j + ii[i];
1315: aa = a->a + ii[i];
1316: sum = 0.0;
1317: PetscSparseDensePlusDot(sum,x,aa,aj,n);
1318: /* for (j=0; j<n; j++) sum += (*aa++)*x[*aj++]; */
1319: y[*ridx++] = sum;
1320: }
1321: } else { /* do not use compressed row format */
1322: #if defined(PETSC_USE_FORTRAN_KERNEL_MULTAIJ)
1323: aj = a->j;
1324: aa = a->a;
1325: fortranmultaij_(&m,x,ii,aj,aa,y);
1326: #else
1327: for (i=0; i<m; i++) {
1328: n = ii[i+1] - ii[i];
1329: aj = a->j + ii[i];
1330: aa = a->a + ii[i];
1331: sum = 0.0;
1332: PetscSparseDensePlusDot(sum,x,aa,aj,n);
1333: y[i] = sum;
1334: }
1335: #endif
1336: }
1337: PetscLogFlops(2.0*a->nz - a->nonzerorowcnt);
1338: VecRestoreArrayRead(xx,&x);
1339: VecRestoreArray(yy,&y);
1340: return(0);
1341: }
1343: PetscErrorCode MatMultMax_SeqAIJ(Mat A,Vec xx,Vec yy)
1344: {
1345: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1346: PetscScalar *y;
1347: const PetscScalar *x;
1348: const MatScalar *aa;
1349: PetscErrorCode ierr;
1350: PetscInt m=A->rmap->n;
1351: const PetscInt *aj,*ii,*ridx=NULL;
1352: PetscInt n,i,nonzerorow=0;
1353: PetscScalar sum;
1354: PetscBool usecprow=a->compressedrow.use;
1356: #if defined(PETSC_HAVE_PRAGMA_DISJOINT)
1357: #pragma disjoint(*x,*y,*aa)
1358: #endif
1361: VecGetArrayRead(xx,&x);
1362: VecGetArray(yy,&y);
1363: if (usecprow) { /* use compressed row format */
1364: m = a->compressedrow.nrows;
1365: ii = a->compressedrow.i;
1366: ridx = a->compressedrow.rindex;
1367: for (i=0; i<m; i++) {
1368: n = ii[i+1] - ii[i];
1369: aj = a->j + ii[i];
1370: aa = a->a + ii[i];
1371: sum = 0.0;
1372: nonzerorow += (n>0);
1373: PetscSparseDenseMaxDot(sum,x,aa,aj,n);
1374: /* for (j=0; j<n; j++) sum += (*aa++)*x[*aj++]; */
1375: y[*ridx++] = sum;
1376: }
1377: } else { /* do not use compressed row format */
1378: ii = a->i;
1379: for (i=0; i<m; i++) {
1380: n = ii[i+1] - ii[i];
1381: aj = a->j + ii[i];
1382: aa = a->a + ii[i];
1383: sum = 0.0;
1384: nonzerorow += (n>0);
1385: PetscSparseDenseMaxDot(sum,x,aa,aj,n);
1386: y[i] = sum;
1387: }
1388: }
1389: PetscLogFlops(2.0*a->nz - nonzerorow);
1390: VecRestoreArrayRead(xx,&x);
1391: VecRestoreArray(yy,&y);
1392: return(0);
1393: }
1395: PetscErrorCode MatMultAddMax_SeqAIJ(Mat A,Vec xx,Vec yy,Vec zz)
1396: {
1397: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1398: PetscScalar *y,*z;
1399: const PetscScalar *x;
1400: const MatScalar *aa;
1401: PetscErrorCode ierr;
1402: PetscInt m = A->rmap->n,*aj,*ii;
1403: PetscInt n,i,*ridx=NULL;
1404: PetscScalar sum;
1405: PetscBool usecprow=a->compressedrow.use;
1408: VecGetArrayRead(xx,&x);
1409: VecGetArrayPair(yy,zz,&y,&z);
1410: if (usecprow) { /* use compressed row format */
1411: if (zz != yy) {
1412: PetscMemcpy(z,y,m*sizeof(PetscScalar));
1413: }
1414: m = a->compressedrow.nrows;
1415: ii = a->compressedrow.i;
1416: ridx = a->compressedrow.rindex;
1417: for (i=0; i<m; i++) {
1418: n = ii[i+1] - ii[i];
1419: aj = a->j + ii[i];
1420: aa = a->a + ii[i];
1421: sum = y[*ridx];
1422: PetscSparseDenseMaxDot(sum,x,aa,aj,n);
1423: z[*ridx++] = sum;
1424: }
1425: } else { /* do not use compressed row format */
1426: ii = a->i;
1427: for (i=0; i<m; i++) {
1428: n = ii[i+1] - ii[i];
1429: aj = a->j + ii[i];
1430: aa = a->a + ii[i];
1431: sum = y[i];
1432: PetscSparseDenseMaxDot(sum,x,aa,aj,n);
1433: z[i] = sum;
1434: }
1435: }
1436: PetscLogFlops(2.0*a->nz);
1437: VecRestoreArrayRead(xx,&x);
1438: VecRestoreArrayPair(yy,zz,&y,&z);
1439: return(0);
1440: }
1442: #include <../src/mat/impls/aij/seq/ftn-kernels/fmultadd.h>
1443: PetscErrorCode MatMultAdd_SeqAIJ(Mat A,Vec xx,Vec yy,Vec zz)
1444: {
1445: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1446: PetscScalar *y,*z;
1447: const PetscScalar *x;
1448: const MatScalar *aa;
1449: PetscErrorCode ierr;
1450: const PetscInt *aj,*ii,*ridx=NULL;
1451: PetscInt m = A->rmap->n,n,i;
1452: PetscScalar sum;
1453: PetscBool usecprow=a->compressedrow.use;
1456: VecGetArrayRead(xx,&x);
1457: VecGetArrayPair(yy,zz,&y,&z);
1458: if (usecprow) { /* use compressed row format */
1459: if (zz != yy) {
1460: PetscMemcpy(z,y,m*sizeof(PetscScalar));
1461: }
1462: m = a->compressedrow.nrows;
1463: ii = a->compressedrow.i;
1464: ridx = a->compressedrow.rindex;
1465: for (i=0; i<m; i++) {
1466: n = ii[i+1] - ii[i];
1467: aj = a->j + ii[i];
1468: aa = a->a + ii[i];
1469: sum = y[*ridx];
1470: PetscSparseDensePlusDot(sum,x,aa,aj,n);
1471: z[*ridx++] = sum;
1472: }
1473: } else { /* do not use compressed row format */
1474: ii = a->i;
1475: #if defined(PETSC_USE_FORTRAN_KERNEL_MULTADDAIJ)
1476: aj = a->j;
1477: aa = a->a;
1478: fortranmultaddaij_(&m,x,ii,aj,aa,y,z);
1479: #else
1480: for (i=0; i<m; i++) {
1481: n = ii[i+1] - ii[i];
1482: aj = a->j + ii[i];
1483: aa = a->a + ii[i];
1484: sum = y[i];
1485: PetscSparseDensePlusDot(sum,x,aa,aj,n);
1486: z[i] = sum;
1487: }
1488: #endif
1489: }
1490: PetscLogFlops(2.0*a->nz);
1491: VecRestoreArrayRead(xx,&x);
1492: VecRestoreArrayPair(yy,zz,&y,&z);
1493: return(0);
1494: }
1496: /*
1497: Adds diagonal pointers to sparse matrix structure.
1498: */
1499: PetscErrorCode MatMarkDiagonal_SeqAIJ(Mat A)
1500: {
1501: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1503: PetscInt i,j,m = A->rmap->n;
1506: if (!a->diag) {
1507: PetscMalloc1(m,&a->diag);
1508: PetscLogObjectMemory((PetscObject)A, m*sizeof(PetscInt));
1509: }
1510: for (i=0; i<A->rmap->n; i++) {
1511: a->diag[i] = a->i[i+1];
1512: for (j=a->i[i]; j<a->i[i+1]; j++) {
1513: if (a->j[j] == i) {
1514: a->diag[i] = j;
1515: break;
1516: }
1517: }
1518: }
1519: return(0);
1520: }
1522: PetscErrorCode MatShift_SeqAIJ(Mat A,PetscScalar v)
1523: {
1524: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1525: const PetscInt *diag = (const PetscInt*)a->diag;
1526: const PetscInt *ii = (const PetscInt*) a->i;
1527: PetscInt i,*mdiag = NULL;
1528: PetscErrorCode ierr;
1529: PetscInt cnt = 0; /* how many diagonals are missing */
1532: if (!A->preallocated || !a->nz) {
1533: MatSeqAIJSetPreallocation(A,1,NULL);
1534: MatShift_Basic(A,v);
1535: return(0);
1536: }
1538: if (a->diagonaldense) {
1539: cnt = 0;
1540: } else {
1541: PetscCalloc1(A->rmap->n,&mdiag);
1542: for (i=0; i<A->rmap->n; i++) {
1543: if (diag[i] >= ii[i+1]) {
1544: cnt++;
1545: mdiag[i] = 1;
1546: }
1547: }
1548: }
1549: if (!cnt) {
1550: MatShift_Basic(A,v);
1551: } else {
1552: PetscScalar *olda = a->a; /* preserve pointers to current matrix nonzeros structure and values */
1553: PetscInt *oldj = a->j, *oldi = a->i;
1554: PetscBool singlemalloc = a->singlemalloc,free_a = a->free_a,free_ij = a->free_ij;
1556: a->a = NULL;
1557: a->j = NULL;
1558: a->i = NULL;
1559: /* increase the values in imax for each row where a diagonal is being inserted then reallocate the matrix data structures */
1560: for (i=0; i<A->rmap->n; i++) {
1561: a->imax[i] += mdiag[i];
1562: }
1563: MatSeqAIJSetPreallocation_SeqAIJ(A,0,a->imax);
1565: /* copy old values into new matrix data structure */
1566: for (i=0; i<A->rmap->n; i++) {
1567: MatSetValues(A,1,&i,a->imax[i] - mdiag[i],&oldj[oldi[i]],&olda[oldi[i]],ADD_VALUES);
1568: MatSetValue(A,i,i,v,ADD_VALUES);
1569: }
1570: MatAssemblyBegin(A,MAT_FINAL_ASSEMBLY);
1571: MatAssemblyEnd(A,MAT_FINAL_ASSEMBLY);
1572: if (singlemalloc) {
1573: PetscFree3(olda,oldj,oldi);
1574: } else {
1575: if (free_a) {PetscFree(olda);}
1576: if (free_ij) {PetscFree(oldj);}
1577: if (free_ij) {PetscFree(oldi);}
1578: }
1579: }
1580: PetscFree(mdiag);
1581: a->diagonaldense = PETSC_TRUE;
1582: return(0);
1583: }
1585: /*
1586: Checks for missing diagonals
1587: */
1588: PetscErrorCode MatMissingDiagonal_SeqAIJ(Mat A,PetscBool *missing,PetscInt *d)
1589: {
1590: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1591: PetscInt *diag,*ii = a->i,i;
1595: *missing = PETSC_FALSE;
1596: if (A->rmap->n > 0 && !ii) {
1597: *missing = PETSC_TRUE;
1598: if (d) *d = 0;
1599: PetscInfo(A,"Matrix has no entries therefore is missing diagonal\n");
1600: } else {
1601: diag = a->diag;
1602: for (i=0; i<A->rmap->n; i++) {
1603: if (diag[i] >= ii[i+1]) {
1604: *missing = PETSC_TRUE;
1605: if (d) *d = i;
1606: PetscInfo1(A,"Matrix is missing diagonal number %D\n",i);
1607: break;
1608: }
1609: }
1610: }
1611: return(0);
1612: }
1614: #include <petscblaslapack.h>
1615: #include <petsc/private/kernels/blockinvert.h>
1617: /*
1618: Note that values is allocated externally by the PC and then passed into this routine
1619: */
1620: PetscErrorCode MatInvertVariableBlockDiagonal_SeqAIJ(Mat A,PetscInt nblocks,const PetscInt *bsizes,PetscScalar *diag)
1621: {
1622: PetscErrorCode ierr;
1623: PetscInt n = A->rmap->n, i, ncnt = 0, *indx,j,bsizemax = 0,*v_pivots;
1624: PetscBool allowzeropivot,zeropivotdetected=PETSC_FALSE;
1625: const PetscReal shift = 0.0;
1626: PetscInt ipvt[5];
1627: PetscScalar work[25],*v_work;
1630: allowzeropivot = PetscNot(A->erroriffailure);
1631: for (i=0; i<nblocks; i++) ncnt += bsizes[i];
1632: if (ncnt != n) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Total blocksizes %D doesn't match number matrix rows %D",ncnt,n);
1633: for (i=0; i<nblocks; i++) {
1634: bsizemax = PetscMax(bsizemax,bsizes[i]);
1635: }
1636: PetscMalloc1(bsizemax,&indx);
1637: if (bsizemax > 7) {
1638: PetscMalloc2(bsizemax,&v_work,bsizemax,&v_pivots);
1639: }
1640: ncnt = 0;
1641: for (i=0; i<nblocks; i++) {
1642: for (j=0; j<bsizes[i]; j++) indx[j] = ncnt+j;
1643: MatGetValues(A,bsizes[i],indx,bsizes[i],indx,diag);
1644: switch (bsizes[i]) {
1645: case 1:
1646: *diag = 1.0/(*diag);
1647: break;
1648: case 2:
1649: PetscKernel_A_gets_inverse_A_2(diag,shift,allowzeropivot,&zeropivotdetected);
1650: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
1651: PetscKernel_A_gets_transpose_A_2(diag);
1652: break;
1653: case 3:
1654: PetscKernel_A_gets_inverse_A_3(diag,shift,allowzeropivot,&zeropivotdetected);
1655: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
1656: PetscKernel_A_gets_transpose_A_3(diag);
1657: break;
1658: case 4:
1659: PetscKernel_A_gets_inverse_A_4(diag,shift,allowzeropivot,&zeropivotdetected);
1660: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
1661: PetscKernel_A_gets_transpose_A_4(diag);
1662: break;
1663: case 5:
1664: PetscKernel_A_gets_inverse_A_5(diag,ipvt,work,shift,allowzeropivot,&zeropivotdetected);
1665: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
1666: PetscKernel_A_gets_transpose_A_5(diag);
1667: break;
1668: case 6:
1669: PetscKernel_A_gets_inverse_A_6(diag,shift,allowzeropivot,&zeropivotdetected);
1670: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
1671: PetscKernel_A_gets_transpose_A_6(diag);
1672: break;
1673: case 7:
1674: PetscKernel_A_gets_inverse_A_7(diag,shift,allowzeropivot,&zeropivotdetected);
1675: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
1676: PetscKernel_A_gets_transpose_A_7(diag);
1677: break;
1678: default:
1679: PetscKernel_A_gets_inverse_A(bsizes[i],diag,v_pivots,v_work,allowzeropivot,&zeropivotdetected);
1680: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
1681: PetscKernel_A_gets_transpose_A_N(diag,bsizes[i]);
1682: }
1683: ncnt += bsizes[i];
1684: diag += bsizes[i]*bsizes[i];
1685: }
1686: if (bsizemax > 7) {
1687: PetscFree2(v_work,v_pivots);
1688: }
1689: PetscFree(indx);
1690: return(0);
1691: }
1693: /*
1694: Negative shift indicates do not generate an error if there is a zero diagonal, just invert it anyways
1695: */
1696: PetscErrorCode MatInvertDiagonal_SeqAIJ(Mat A,PetscScalar omega,PetscScalar fshift)
1697: {
1698: Mat_SeqAIJ *a = (Mat_SeqAIJ*) A->data;
1700: PetscInt i,*diag,m = A->rmap->n;
1701: MatScalar *v = a->a;
1702: PetscScalar *idiag,*mdiag;
1705: if (a->idiagvalid) return(0);
1706: MatMarkDiagonal_SeqAIJ(A);
1707: diag = a->diag;
1708: if (!a->idiag) {
1709: PetscMalloc3(m,&a->idiag,m,&a->mdiag,m,&a->ssor_work);
1710: PetscLogObjectMemory((PetscObject)A, 3*m*sizeof(PetscScalar));
1711: v = a->a;
1712: }
1713: mdiag = a->mdiag;
1714: idiag = a->idiag;
1716: if (omega == 1.0 && PetscRealPart(fshift) <= 0.0) {
1717: for (i=0; i<m; i++) {
1718: mdiag[i] = v[diag[i]];
1719: if (!PetscAbsScalar(mdiag[i])) { /* zero diagonal */
1720: if (PetscRealPart(fshift)) {
1721: PetscInfo1(A,"Zero diagonal on row %D\n",i);
1722: A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
1723: A->factorerror_zeropivot_value = 0.0;
1724: A->factorerror_zeropivot_row = i;
1725: } else SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_INCOMP,"Zero diagonal on row %D",i);
1726: }
1727: idiag[i] = 1.0/v[diag[i]];
1728: }
1729: PetscLogFlops(m);
1730: } else {
1731: for (i=0; i<m; i++) {
1732: mdiag[i] = v[diag[i]];
1733: idiag[i] = omega/(fshift + v[diag[i]]);
1734: }
1735: PetscLogFlops(2.0*m);
1736: }
1737: a->idiagvalid = PETSC_TRUE;
1738: return(0);
1739: }
1741: #include <../src/mat/impls/aij/seq/ftn-kernels/frelax.h>
1742: PetscErrorCode MatSOR_SeqAIJ(Mat A,Vec bb,PetscReal omega,MatSORType flag,PetscReal fshift,PetscInt its,PetscInt lits,Vec xx)
1743: {
1744: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1745: PetscScalar *x,d,sum,*t,scale;
1746: const MatScalar *v,*idiag=0,*mdiag;
1747: const PetscScalar *b, *bs,*xb, *ts;
1748: PetscErrorCode ierr;
1749: PetscInt n,m = A->rmap->n,i;
1750: const PetscInt *idx,*diag;
1753: its = its*lits;
1755: if (fshift != a->fshift || omega != a->omega) a->idiagvalid = PETSC_FALSE; /* must recompute idiag[] */
1756: if (!a->idiagvalid) {MatInvertDiagonal_SeqAIJ(A,omega,fshift);}
1757: a->fshift = fshift;
1758: a->omega = omega;
1760: diag = a->diag;
1761: t = a->ssor_work;
1762: idiag = a->idiag;
1763: mdiag = a->mdiag;
1765: VecGetArray(xx,&x);
1766: VecGetArrayRead(bb,&b);
1767: /* We count flops by assuming the upper triangular and lower triangular parts have the same number of nonzeros */
1768: if (flag == SOR_APPLY_UPPER) {
1769: /* apply (U + D/omega) to the vector */
1770: bs = b;
1771: for (i=0; i<m; i++) {
1772: d = fshift + mdiag[i];
1773: n = a->i[i+1] - diag[i] - 1;
1774: idx = a->j + diag[i] + 1;
1775: v = a->a + diag[i] + 1;
1776: sum = b[i]*d/omega;
1777: PetscSparseDensePlusDot(sum,bs,v,idx,n);
1778: x[i] = sum;
1779: }
1780: VecRestoreArray(xx,&x);
1781: VecRestoreArrayRead(bb,&b);
1782: PetscLogFlops(a->nz);
1783: return(0);
1784: }
1786: if (flag == SOR_APPLY_LOWER) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"SOR_APPLY_LOWER is not implemented");
1787: else if (flag & SOR_EISENSTAT) {
1788: /* Let A = L + U + D; where L is lower trianglar,
1789: U is upper triangular, E = D/omega; This routine applies
1791: (L + E)^{-1} A (U + E)^{-1}
1793: to a vector efficiently using Eisenstat's trick.
1794: */
1795: scale = (2.0/omega) - 1.0;
1797: /* x = (E + U)^{-1} b */
1798: for (i=m-1; i>=0; i--) {
1799: n = a->i[i+1] - diag[i] - 1;
1800: idx = a->j + diag[i] + 1;
1801: v = a->a + diag[i] + 1;
1802: sum = b[i];
1803: PetscSparseDenseMinusDot(sum,x,v,idx,n);
1804: x[i] = sum*idiag[i];
1805: }
1807: /* t = b - (2*E - D)x */
1808: v = a->a;
1809: for (i=0; i<m; i++) t[i] = b[i] - scale*(v[*diag++])*x[i];
1811: /* t = (E + L)^{-1}t */
1812: ts = t;
1813: diag = a->diag;
1814: for (i=0; i<m; i++) {
1815: n = diag[i] - a->i[i];
1816: idx = a->j + a->i[i];
1817: v = a->a + a->i[i];
1818: sum = t[i];
1819: PetscSparseDenseMinusDot(sum,ts,v,idx,n);
1820: t[i] = sum*idiag[i];
1821: /* x = x + t */
1822: x[i] += t[i];
1823: }
1825: PetscLogFlops(6.0*m-1 + 2.0*a->nz);
1826: VecRestoreArray(xx,&x);
1827: VecRestoreArrayRead(bb,&b);
1828: return(0);
1829: }
1830: if (flag & SOR_ZERO_INITIAL_GUESS) {
1831: if (flag & SOR_FORWARD_SWEEP || flag & SOR_LOCAL_FORWARD_SWEEP) {
1832: for (i=0; i<m; i++) {
1833: n = diag[i] - a->i[i];
1834: idx = a->j + a->i[i];
1835: v = a->a + a->i[i];
1836: sum = b[i];
1837: PetscSparseDenseMinusDot(sum,x,v,idx,n);
1838: t[i] = sum;
1839: x[i] = sum*idiag[i];
1840: }
1841: xb = t;
1842: PetscLogFlops(a->nz);
1843: } else xb = b;
1844: if (flag & SOR_BACKWARD_SWEEP || flag & SOR_LOCAL_BACKWARD_SWEEP) {
1845: for (i=m-1; i>=0; i--) {
1846: n = a->i[i+1] - diag[i] - 1;
1847: idx = a->j + diag[i] + 1;
1848: v = a->a + diag[i] + 1;
1849: sum = xb[i];
1850: PetscSparseDenseMinusDot(sum,x,v,idx,n);
1851: if (xb == b) {
1852: x[i] = sum*idiag[i];
1853: } else {
1854: x[i] = (1-omega)*x[i] + sum*idiag[i]; /* omega in idiag */
1855: }
1856: }
1857: PetscLogFlops(a->nz); /* assumes 1/2 in upper */
1858: }
1859: its--;
1860: }
1861: while (its--) {
1862: if (flag & SOR_FORWARD_SWEEP || flag & SOR_LOCAL_FORWARD_SWEEP) {
1863: for (i=0; i<m; i++) {
1864: /* lower */
1865: n = diag[i] - a->i[i];
1866: idx = a->j + a->i[i];
1867: v = a->a + a->i[i];
1868: sum = b[i];
1869: PetscSparseDenseMinusDot(sum,x,v,idx,n);
1870: t[i] = sum; /* save application of the lower-triangular part */
1871: /* upper */
1872: n = a->i[i+1] - diag[i] - 1;
1873: idx = a->j + diag[i] + 1;
1874: v = a->a + diag[i] + 1;
1875: PetscSparseDenseMinusDot(sum,x,v,idx,n);
1876: x[i] = (1. - omega)*x[i] + sum*idiag[i]; /* omega in idiag */
1877: }
1878: xb = t;
1879: PetscLogFlops(2.0*a->nz);
1880: } else xb = b;
1881: if (flag & SOR_BACKWARD_SWEEP || flag & SOR_LOCAL_BACKWARD_SWEEP) {
1882: for (i=m-1; i>=0; i--) {
1883: sum = xb[i];
1884: if (xb == b) {
1885: /* whole matrix (no checkpointing available) */
1886: n = a->i[i+1] - a->i[i];
1887: idx = a->j + a->i[i];
1888: v = a->a + a->i[i];
1889: PetscSparseDenseMinusDot(sum,x,v,idx,n);
1890: x[i] = (1. - omega)*x[i] + (sum + mdiag[i]*x[i])*idiag[i];
1891: } else { /* lower-triangular part has been saved, so only apply upper-triangular */
1892: n = a->i[i+1] - diag[i] - 1;
1893: idx = a->j + diag[i] + 1;
1894: v = a->a + diag[i] + 1;
1895: PetscSparseDenseMinusDot(sum,x,v,idx,n);
1896: x[i] = (1. - omega)*x[i] + sum*idiag[i]; /* omega in idiag */
1897: }
1898: }
1899: if (xb == b) {
1900: PetscLogFlops(2.0*a->nz);
1901: } else {
1902: PetscLogFlops(a->nz); /* assumes 1/2 in upper */
1903: }
1904: }
1905: }
1906: VecRestoreArray(xx,&x);
1907: VecRestoreArrayRead(bb,&b);
1908: return(0);
1909: }
1912: PetscErrorCode MatGetInfo_SeqAIJ(Mat A,MatInfoType flag,MatInfo *info)
1913: {
1914: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1917: info->block_size = 1.0;
1918: info->nz_allocated = (double)a->maxnz;
1919: info->nz_used = (double)a->nz;
1920: info->nz_unneeded = (double)(a->maxnz - a->nz);
1921: info->assemblies = (double)A->num_ass;
1922: info->mallocs = (double)A->info.mallocs;
1923: info->memory = ((PetscObject)A)->mem;
1924: if (A->factortype) {
1925: info->fill_ratio_given = A->info.fill_ratio_given;
1926: info->fill_ratio_needed = A->info.fill_ratio_needed;
1927: info->factor_mallocs = A->info.factor_mallocs;
1928: } else {
1929: info->fill_ratio_given = 0;
1930: info->fill_ratio_needed = 0;
1931: info->factor_mallocs = 0;
1932: }
1933: return(0);
1934: }
1936: PetscErrorCode MatZeroRows_SeqAIJ(Mat A,PetscInt N,const PetscInt rows[],PetscScalar diag,Vec x,Vec b)
1937: {
1938: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1939: PetscInt i,m = A->rmap->n - 1;
1940: PetscErrorCode ierr;
1941: const PetscScalar *xx;
1942: PetscScalar *bb;
1943: PetscInt d = 0;
1946: if (x && b) {
1947: VecGetArrayRead(x,&xx);
1948: VecGetArray(b,&bb);
1949: for (i=0; i<N; i++) {
1950: if (rows[i] < 0 || rows[i] > m) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"row %D out of range", rows[i]);
1951: bb[rows[i]] = diag*xx[rows[i]];
1952: }
1953: VecRestoreArrayRead(x,&xx);
1954: VecRestoreArray(b,&bb);
1955: }
1957: if (a->keepnonzeropattern) {
1958: for (i=0; i<N; i++) {
1959: if (rows[i] < 0 || rows[i] > m) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"row %D out of range", rows[i]);
1960: PetscMemzero(&a->a[a->i[rows[i]]],a->ilen[rows[i]]*sizeof(PetscScalar));
1961: }
1962: if (diag != 0.0) {
1963: for (i=0; i<N; i++) {
1964: d = rows[i];
1965: 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);
1966: }
1967: for (i=0; i<N; i++) {
1968: a->a[a->diag[rows[i]]] = diag;
1969: }
1970: }
1971: } else {
1972: if (diag != 0.0) {
1973: for (i=0; i<N; i++) {
1974: if (rows[i] < 0 || rows[i] > m) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"row %D out of range", rows[i]);
1975: if (a->ilen[rows[i]] > 0) {
1976: a->ilen[rows[i]] = 1;
1977: a->a[a->i[rows[i]]] = diag;
1978: a->j[a->i[rows[i]]] = rows[i];
1979: } else { /* in case row was completely empty */
1980: MatSetValues_SeqAIJ(A,1,&rows[i],1,&rows[i],&diag,INSERT_VALUES);
1981: }
1982: }
1983: } else {
1984: for (i=0; i<N; i++) {
1985: if (rows[i] < 0 || rows[i] > m) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"row %D out of range", rows[i]);
1986: a->ilen[rows[i]] = 0;
1987: }
1988: }
1989: A->nonzerostate++;
1990: }
1991: (*A->ops->assemblyend)(A,MAT_FINAL_ASSEMBLY);
1992: return(0);
1993: }
1995: PetscErrorCode MatZeroRowsColumns_SeqAIJ(Mat A,PetscInt N,const PetscInt rows[],PetscScalar diag,Vec x,Vec b)
1996: {
1997: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1998: PetscInt i,j,m = A->rmap->n - 1,d = 0;
1999: PetscErrorCode ierr;
2000: PetscBool missing,*zeroed,vecs = PETSC_FALSE;
2001: const PetscScalar *xx;
2002: PetscScalar *bb;
2005: if (x && b) {
2006: VecGetArrayRead(x,&xx);
2007: VecGetArray(b,&bb);
2008: vecs = PETSC_TRUE;
2009: }
2010: PetscCalloc1(A->rmap->n,&zeroed);
2011: for (i=0; i<N; i++) {
2012: if (rows[i] < 0 || rows[i] > m) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"row %D out of range", rows[i]);
2013: PetscMemzero(&a->a[a->i[rows[i]]],a->ilen[rows[i]]*sizeof(PetscScalar));
2015: zeroed[rows[i]] = PETSC_TRUE;
2016: }
2017: for (i=0; i<A->rmap->n; i++) {
2018: if (!zeroed[i]) {
2019: for (j=a->i[i]; j<a->i[i+1]; j++) {
2020: if (zeroed[a->j[j]]) {
2021: if (vecs) bb[i] -= a->a[j]*xx[a->j[j]];
2022: a->a[j] = 0.0;
2023: }
2024: }
2025: } else if (vecs) bb[i] = diag*xx[i];
2026: }
2027: if (x && b) {
2028: VecRestoreArrayRead(x,&xx);
2029: VecRestoreArray(b,&bb);
2030: }
2031: PetscFree(zeroed);
2032: if (diag != 0.0) {
2033: MatMissingDiagonal_SeqAIJ(A,&missing,&d);
2034: if (missing) {
2035: if (a->nonew) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Matrix is missing diagonal entry in row %D",d);
2036: else {
2037: for (i=0; i<N; i++) {
2038: MatSetValues_SeqAIJ(A,1,&rows[i],1,&rows[i],&diag,INSERT_VALUES);
2039: }
2040: }
2041: } else {
2042: for (i=0; i<N; i++) {
2043: a->a[a->diag[rows[i]]] = diag;
2044: }
2045: }
2046: }
2047: (*A->ops->assemblyend)(A,MAT_FINAL_ASSEMBLY);
2048: return(0);
2049: }
2051: PetscErrorCode MatGetRow_SeqAIJ(Mat A,PetscInt row,PetscInt *nz,PetscInt **idx,PetscScalar **v)
2052: {
2053: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
2054: PetscInt *itmp;
2057: if (row < 0 || row >= A->rmap->n) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Row %D out of range",row);
2059: *nz = a->i[row+1] - a->i[row];
2060: if (v) *v = a->a + a->i[row];
2061: if (idx) {
2062: itmp = a->j + a->i[row];
2063: if (*nz) *idx = itmp;
2064: else *idx = 0;
2065: }
2066: return(0);
2067: }
2069: /* remove this function? */
2070: PetscErrorCode MatRestoreRow_SeqAIJ(Mat A,PetscInt row,PetscInt *nz,PetscInt **idx,PetscScalar **v)
2071: {
2073: return(0);
2074: }
2076: PetscErrorCode MatNorm_SeqAIJ(Mat A,NormType type,PetscReal *nrm)
2077: {
2078: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
2079: MatScalar *v = a->a;
2080: PetscReal sum = 0.0;
2082: PetscInt i,j;
2085: if (type == NORM_FROBENIUS) {
2086: #if defined(PETSC_USE_REAL___FP16)
2087: PetscBLASInt one = 1,nz = a->nz;
2088: *nrm = BLASnrm2_(&nz,v,&one);
2089: #else
2090: for (i=0; i<a->nz; i++) {
2091: sum += PetscRealPart(PetscConj(*v)*(*v)); v++;
2092: }
2093: *nrm = PetscSqrtReal(sum);
2094: #endif
2095: PetscLogFlops(2*a->nz);
2096: } else if (type == NORM_1) {
2097: PetscReal *tmp;
2098: PetscInt *jj = a->j;
2099: PetscCalloc1(A->cmap->n+1,&tmp);
2100: *nrm = 0.0;
2101: for (j=0; j<a->nz; j++) {
2102: tmp[*jj++] += PetscAbsScalar(*v); v++;
2103: }
2104: for (j=0; j<A->cmap->n; j++) {
2105: if (tmp[j] > *nrm) *nrm = tmp[j];
2106: }
2107: PetscFree(tmp);
2108: PetscLogFlops(PetscMax(a->nz-1,0));
2109: } else if (type == NORM_INFINITY) {
2110: *nrm = 0.0;
2111: for (j=0; j<A->rmap->n; j++) {
2112: v = a->a + a->i[j];
2113: sum = 0.0;
2114: for (i=0; i<a->i[j+1]-a->i[j]; i++) {
2115: sum += PetscAbsScalar(*v); v++;
2116: }
2117: if (sum > *nrm) *nrm = sum;
2118: }
2119: PetscLogFlops(PetscMax(a->nz-1,0));
2120: } else SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"No support for two norm");
2121: return(0);
2122: }
2124: /* Merged from MatGetSymbolicTranspose_SeqAIJ() - replace MatGetSymbolicTranspose_SeqAIJ()? */
2125: PetscErrorCode MatTransposeSymbolic_SeqAIJ(Mat A,Mat *B)
2126: {
2128: PetscInt i,j,anzj;
2129: Mat_SeqAIJ *a=(Mat_SeqAIJ*)A->data,*b;
2130: PetscInt an=A->cmap->N,am=A->rmap->N;
2131: PetscInt *ati,*atj,*atfill,*ai=a->i,*aj=a->j;
2134: /* Allocate space for symbolic transpose info and work array */
2135: PetscCalloc1(an+1,&ati);
2136: PetscMalloc1(ai[am],&atj);
2137: PetscMalloc1(an,&atfill);
2139: /* Walk through aj and count ## of non-zeros in each row of A^T. */
2140: /* Note: offset by 1 for fast conversion into csr format. */
2141: for (i=0;i<ai[am];i++) ati[aj[i]+1] += 1;
2142: /* Form ati for csr format of A^T. */
2143: for (i=0;i<an;i++) ati[i+1] += ati[i];
2145: /* Copy ati into atfill so we have locations of the next free space in atj */
2146: PetscMemcpy(atfill,ati,an*sizeof(PetscInt));
2148: /* Walk through A row-wise and mark nonzero entries of A^T. */
2149: for (i=0;i<am;i++) {
2150: anzj = ai[i+1] - ai[i];
2151: for (j=0;j<anzj;j++) {
2152: atj[atfill[*aj]] = i;
2153: atfill[*aj++] += 1;
2154: }
2155: }
2157: /* Clean up temporary space and complete requests. */
2158: PetscFree(atfill);
2159: MatCreateSeqAIJWithArrays(PetscObjectComm((PetscObject)A),an,am,ati,atj,NULL,B);
2160: MatSetBlockSizes(*B,PetscAbs(A->cmap->bs),PetscAbs(A->rmap->bs));
2162: b = (Mat_SeqAIJ*)((*B)->data);
2163: b->free_a = PETSC_FALSE;
2164: b->free_ij = PETSC_TRUE;
2165: b->nonew = 0;
2166: return(0);
2167: }
2169: PetscErrorCode MatTranspose_SeqAIJ(Mat A,MatReuse reuse,Mat *B)
2170: {
2171: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
2172: Mat C;
2174: PetscInt i,*aj = a->j,*ai = a->i,m = A->rmap->n,len,*col;
2175: MatScalar *array = a->a;
2178: if (reuse == MAT_INITIAL_MATRIX || reuse == MAT_INPLACE_MATRIX) {
2179: PetscCalloc1(1+A->cmap->n,&col);
2181: for (i=0; i<ai[m]; i++) col[aj[i]] += 1;
2182: MatCreate(PetscObjectComm((PetscObject)A),&C);
2183: MatSetSizes(C,A->cmap->n,m,A->cmap->n,m);
2184: MatSetBlockSizes(C,PetscAbs(A->cmap->bs),PetscAbs(A->rmap->bs));
2185: MatSetType(C,((PetscObject)A)->type_name);
2186: MatSeqAIJSetPreallocation_SeqAIJ(C,0,col);
2187: PetscFree(col);
2188: } else {
2189: C = *B;
2190: }
2192: for (i=0; i<m; i++) {
2193: len = ai[i+1]-ai[i];
2194: MatSetValues_SeqAIJ(C,len,aj,1,&i,array,INSERT_VALUES);
2195: array += len;
2196: aj += len;
2197: }
2198: MatAssemblyBegin(C,MAT_FINAL_ASSEMBLY);
2199: MatAssemblyEnd(C,MAT_FINAL_ASSEMBLY);
2201: if (reuse == MAT_INITIAL_MATRIX || reuse == MAT_REUSE_MATRIX) {
2202: *B = C;
2203: } else {
2204: MatHeaderMerge(A,&C);
2205: }
2206: return(0);
2207: }
2209: PetscErrorCode MatIsTranspose_SeqAIJ(Mat A,Mat B,PetscReal tol,PetscBool *f)
2210: {
2211: Mat_SeqAIJ *aij = (Mat_SeqAIJ*) A->data,*bij = (Mat_SeqAIJ*) B->data;
2212: PetscInt *adx,*bdx,*aii,*bii,*aptr,*bptr;
2213: MatScalar *va,*vb;
2215: PetscInt ma,na,mb,nb, i;
2218: MatGetSize(A,&ma,&na);
2219: MatGetSize(B,&mb,&nb);
2220: if (ma!=nb || na!=mb) {
2221: *f = PETSC_FALSE;
2222: return(0);
2223: }
2224: aii = aij->i; bii = bij->i;
2225: adx = aij->j; bdx = bij->j;
2226: va = aij->a; vb = bij->a;
2227: PetscMalloc1(ma,&aptr);
2228: PetscMalloc1(mb,&bptr);
2229: for (i=0; i<ma; i++) aptr[i] = aii[i];
2230: for (i=0; i<mb; i++) bptr[i] = bii[i];
2232: *f = PETSC_TRUE;
2233: for (i=0; i<ma; i++) {
2234: while (aptr[i]<aii[i+1]) {
2235: PetscInt idc,idr;
2236: PetscScalar vc,vr;
2237: /* column/row index/value */
2238: idc = adx[aptr[i]];
2239: idr = bdx[bptr[idc]];
2240: vc = va[aptr[i]];
2241: vr = vb[bptr[idc]];
2242: if (i!=idr || PetscAbsScalar(vc-vr) > tol) {
2243: *f = PETSC_FALSE;
2244: goto done;
2245: } else {
2246: aptr[i]++;
2247: if (B || i!=idc) bptr[idc]++;
2248: }
2249: }
2250: }
2251: done:
2252: PetscFree(aptr);
2253: PetscFree(bptr);
2254: return(0);
2255: }
2257: PetscErrorCode MatIsHermitianTranspose_SeqAIJ(Mat A,Mat B,PetscReal tol,PetscBool *f)
2258: {
2259: Mat_SeqAIJ *aij = (Mat_SeqAIJ*) A->data,*bij = (Mat_SeqAIJ*) B->data;
2260: PetscInt *adx,*bdx,*aii,*bii,*aptr,*bptr;
2261: MatScalar *va,*vb;
2263: PetscInt ma,na,mb,nb, i;
2266: MatGetSize(A,&ma,&na);
2267: MatGetSize(B,&mb,&nb);
2268: if (ma!=nb || na!=mb) {
2269: *f = PETSC_FALSE;
2270: return(0);
2271: }
2272: aii = aij->i; bii = bij->i;
2273: adx = aij->j; bdx = bij->j;
2274: va = aij->a; vb = bij->a;
2275: PetscMalloc1(ma,&aptr);
2276: PetscMalloc1(mb,&bptr);
2277: for (i=0; i<ma; i++) aptr[i] = aii[i];
2278: for (i=0; i<mb; i++) bptr[i] = bii[i];
2280: *f = PETSC_TRUE;
2281: for (i=0; i<ma; i++) {
2282: while (aptr[i]<aii[i+1]) {
2283: PetscInt idc,idr;
2284: PetscScalar vc,vr;
2285: /* column/row index/value */
2286: idc = adx[aptr[i]];
2287: idr = bdx[bptr[idc]];
2288: vc = va[aptr[i]];
2289: vr = vb[bptr[idc]];
2290: if (i!=idr || PetscAbsScalar(vc-PetscConj(vr)) > tol) {
2291: *f = PETSC_FALSE;
2292: goto done;
2293: } else {
2294: aptr[i]++;
2295: if (B || i!=idc) bptr[idc]++;
2296: }
2297: }
2298: }
2299: done:
2300: PetscFree(aptr);
2301: PetscFree(bptr);
2302: return(0);
2303: }
2305: PetscErrorCode MatIsSymmetric_SeqAIJ(Mat A,PetscReal tol,PetscBool *f)
2306: {
2310: MatIsTranspose_SeqAIJ(A,A,tol,f);
2311: return(0);
2312: }
2314: PetscErrorCode MatIsHermitian_SeqAIJ(Mat A,PetscReal tol,PetscBool *f)
2315: {
2319: MatIsHermitianTranspose_SeqAIJ(A,A,tol,f);
2320: return(0);
2321: }
2323: PetscErrorCode MatDiagonalScale_SeqAIJ(Mat A,Vec ll,Vec rr)
2324: {
2325: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
2326: const PetscScalar *l,*r;
2327: PetscScalar x;
2328: MatScalar *v;
2329: PetscErrorCode ierr;
2330: PetscInt i,j,m = A->rmap->n,n = A->cmap->n,M,nz = a->nz;
2331: const PetscInt *jj;
2334: if (ll) {
2335: /* The local size is used so that VecMPI can be passed to this routine
2336: by MatDiagonalScale_MPIAIJ */
2337: VecGetLocalSize(ll,&m);
2338: if (m != A->rmap->n) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Left scaling vector wrong length");
2339: VecGetArrayRead(ll,&l);
2340: v = a->a;
2341: for (i=0; i<m; i++) {
2342: x = l[i];
2343: M = a->i[i+1] - a->i[i];
2344: for (j=0; j<M; j++) (*v++) *= x;
2345: }
2346: VecRestoreArrayRead(ll,&l);
2347: PetscLogFlops(nz);
2348: }
2349: if (rr) {
2350: VecGetLocalSize(rr,&n);
2351: if (n != A->cmap->n) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Right scaling vector wrong length");
2352: VecGetArrayRead(rr,&r);
2353: v = a->a; jj = a->j;
2354: for (i=0; i<nz; i++) (*v++) *= r[*jj++];
2355: VecRestoreArrayRead(rr,&r);
2356: PetscLogFlops(nz);
2357: }
2358: MatSeqAIJInvalidateDiagonal(A);
2359: return(0);
2360: }
2362: PetscErrorCode MatCreateSubMatrix_SeqAIJ(Mat A,IS isrow,IS iscol,PetscInt csize,MatReuse scall,Mat *B)
2363: {
2364: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data,*c;
2366: PetscInt *smap,i,k,kstart,kend,oldcols = A->cmap->n,*lens;
2367: PetscInt row,mat_i,*mat_j,tcol,first,step,*mat_ilen,sum,lensi;
2368: const PetscInt *irow,*icol;
2369: PetscInt nrows,ncols;
2370: PetscInt *starts,*j_new,*i_new,*aj = a->j,*ai = a->i,ii,*ailen = a->ilen;
2371: MatScalar *a_new,*mat_a;
2372: Mat C;
2373: PetscBool stride;
2377: ISGetIndices(isrow,&irow);
2378: ISGetLocalSize(isrow,&nrows);
2379: ISGetLocalSize(iscol,&ncols);
2381: PetscObjectTypeCompare((PetscObject)iscol,ISSTRIDE,&stride);
2382: if (stride) {
2383: ISStrideGetInfo(iscol,&first,&step);
2384: } else {
2385: first = 0;
2386: step = 0;
2387: }
2388: if (stride && step == 1) {
2389: /* special case of contiguous rows */
2390: PetscMalloc2(nrows,&lens,nrows,&starts);
2391: /* loop over new rows determining lens and starting points */
2392: for (i=0; i<nrows; i++) {
2393: kstart = ai[irow[i]];
2394: kend = kstart + ailen[irow[i]];
2395: starts[i] = kstart;
2396: for (k=kstart; k<kend; k++) {
2397: if (aj[k] >= first) {
2398: starts[i] = k;
2399: break;
2400: }
2401: }
2402: sum = 0;
2403: while (k < kend) {
2404: if (aj[k++] >= first+ncols) break;
2405: sum++;
2406: }
2407: lens[i] = sum;
2408: }
2409: /* create submatrix */
2410: if (scall == MAT_REUSE_MATRIX) {
2411: PetscInt n_cols,n_rows;
2412: MatGetSize(*B,&n_rows,&n_cols);
2413: if (n_rows != nrows || n_cols != ncols) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Reused submatrix wrong size");
2414: MatZeroEntries(*B);
2415: C = *B;
2416: } else {
2417: PetscInt rbs,cbs;
2418: MatCreate(PetscObjectComm((PetscObject)A),&C);
2419: MatSetSizes(C,nrows,ncols,PETSC_DETERMINE,PETSC_DETERMINE);
2420: ISGetBlockSize(isrow,&rbs);
2421: ISGetBlockSize(iscol,&cbs);
2422: MatSetBlockSizes(C,rbs,cbs);
2423: MatSetType(C,((PetscObject)A)->type_name);
2424: MatSeqAIJSetPreallocation_SeqAIJ(C,0,lens);
2425: }
2426: c = (Mat_SeqAIJ*)C->data;
2428: /* loop over rows inserting into submatrix */
2429: a_new = c->a;
2430: j_new = c->j;
2431: i_new = c->i;
2433: for (i=0; i<nrows; i++) {
2434: ii = starts[i];
2435: lensi = lens[i];
2436: for (k=0; k<lensi; k++) {
2437: *j_new++ = aj[ii+k] - first;
2438: }
2439: PetscMemcpy(a_new,a->a + starts[i],lensi*sizeof(PetscScalar));
2440: a_new += lensi;
2441: i_new[i+1] = i_new[i] + lensi;
2442: c->ilen[i] = lensi;
2443: }
2444: PetscFree2(lens,starts);
2445: } else {
2446: ISGetIndices(iscol,&icol);
2447: PetscCalloc1(oldcols,&smap);
2448: PetscMalloc1(1+nrows,&lens);
2449: for (i=0; i<ncols; i++) {
2450: #if defined(PETSC_USE_DEBUG)
2451: 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);
2452: #endif
2453: smap[icol[i]] = i+1;
2454: }
2456: /* determine lens of each row */
2457: for (i=0; i<nrows; i++) {
2458: kstart = ai[irow[i]];
2459: kend = kstart + a->ilen[irow[i]];
2460: lens[i] = 0;
2461: for (k=kstart; k<kend; k++) {
2462: if (smap[aj[k]]) {
2463: lens[i]++;
2464: }
2465: }
2466: }
2467: /* Create and fill new matrix */
2468: if (scall == MAT_REUSE_MATRIX) {
2469: PetscBool equal;
2471: c = (Mat_SeqAIJ*)((*B)->data);
2472: if ((*B)->rmap->n != nrows || (*B)->cmap->n != ncols) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Cannot reuse matrix. wrong size");
2473: PetscMemcmp(c->ilen,lens,(*B)->rmap->n*sizeof(PetscInt),&equal);
2474: if (!equal) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Cannot reuse matrix. wrong no of nonzeros");
2475: PetscMemzero(c->ilen,(*B)->rmap->n*sizeof(PetscInt));
2476: C = *B;
2477: } else {
2478: PetscInt rbs,cbs;
2479: MatCreate(PetscObjectComm((PetscObject)A),&C);
2480: MatSetSizes(C,nrows,ncols,PETSC_DETERMINE,PETSC_DETERMINE);
2481: ISGetBlockSize(isrow,&rbs);
2482: ISGetBlockSize(iscol,&cbs);
2483: MatSetBlockSizes(C,rbs,cbs);
2484: MatSetType(C,((PetscObject)A)->type_name);
2485: MatSeqAIJSetPreallocation_SeqAIJ(C,0,lens);
2486: }
2487: c = (Mat_SeqAIJ*)(C->data);
2488: for (i=0; i<nrows; i++) {
2489: row = irow[i];
2490: kstart = ai[row];
2491: kend = kstart + a->ilen[row];
2492: mat_i = c->i[i];
2493: mat_j = c->j + mat_i;
2494: mat_a = c->a + mat_i;
2495: mat_ilen = c->ilen + i;
2496: for (k=kstart; k<kend; k++) {
2497: if ((tcol=smap[a->j[k]])) {
2498: *mat_j++ = tcol - 1;
2499: *mat_a++ = a->a[k];
2500: (*mat_ilen)++;
2502: }
2503: }
2504: }
2505: /* Free work space */
2506: ISRestoreIndices(iscol,&icol);
2507: PetscFree(smap);
2508: PetscFree(lens);
2509: /* sort */
2510: for (i = 0; i < nrows; i++) {
2511: PetscInt ilen;
2513: mat_i = c->i[i];
2514: mat_j = c->j + mat_i;
2515: mat_a = c->a + mat_i;
2516: ilen = c->ilen[i];
2517: PetscSortIntWithScalarArray(ilen,mat_j,mat_a);
2518: }
2519: }
2520: MatAssemblyBegin(C,MAT_FINAL_ASSEMBLY);
2521: MatAssemblyEnd(C,MAT_FINAL_ASSEMBLY);
2523: ISRestoreIndices(isrow,&irow);
2524: *B = C;
2525: return(0);
2526: }
2528: PetscErrorCode MatGetMultiProcBlock_SeqAIJ(Mat mat,MPI_Comm subComm,MatReuse scall,Mat *subMat)
2529: {
2531: Mat B;
2534: if (scall == MAT_INITIAL_MATRIX) {
2535: MatCreate(subComm,&B);
2536: MatSetSizes(B,mat->rmap->n,mat->cmap->n,mat->rmap->n,mat->cmap->n);
2537: MatSetBlockSizesFromMats(B,mat,mat);
2538: MatSetType(B,MATSEQAIJ);
2539: MatDuplicateNoCreate_SeqAIJ(B,mat,MAT_COPY_VALUES,PETSC_TRUE);
2540: *subMat = B;
2541: } else {
2542: MatCopy_SeqAIJ(mat,*subMat,SAME_NONZERO_PATTERN);
2543: }
2544: return(0);
2545: }
2547: PetscErrorCode MatILUFactor_SeqAIJ(Mat inA,IS row,IS col,const MatFactorInfo *info)
2548: {
2549: Mat_SeqAIJ *a = (Mat_SeqAIJ*)inA->data;
2551: Mat outA;
2552: PetscBool row_identity,col_identity;
2555: if (info->levels != 0) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Only levels=0 supported for in-place ilu");
2557: ISIdentity(row,&row_identity);
2558: ISIdentity(col,&col_identity);
2560: outA = inA;
2561: outA->factortype = MAT_FACTOR_LU;
2562: PetscFree(inA->solvertype);
2563: PetscStrallocpy(MATSOLVERPETSC,&inA->solvertype);
2565: PetscObjectReference((PetscObject)row);
2566: ISDestroy(&a->row);
2568: a->row = row;
2570: PetscObjectReference((PetscObject)col);
2571: ISDestroy(&a->col);
2573: a->col = col;
2575: /* Create the inverse permutation so that it can be used in MatLUFactorNumeric() */
2576: ISDestroy(&a->icol);
2577: ISInvertPermutation(col,PETSC_DECIDE,&a->icol);
2578: PetscLogObjectParent((PetscObject)inA,(PetscObject)a->icol);
2580: if (!a->solve_work) { /* this matrix may have been factored before */
2581: PetscMalloc1(inA->rmap->n+1,&a->solve_work);
2582: PetscLogObjectMemory((PetscObject)inA, (inA->rmap->n+1)*sizeof(PetscScalar));
2583: }
2585: MatMarkDiagonal_SeqAIJ(inA);
2586: if (row_identity && col_identity) {
2587: MatLUFactorNumeric_SeqAIJ_inplace(outA,inA,info);
2588: } else {
2589: MatLUFactorNumeric_SeqAIJ_InplaceWithPerm(outA,inA,info);
2590: }
2591: return(0);
2592: }
2594: PetscErrorCode MatScale_SeqAIJ(Mat inA,PetscScalar alpha)
2595: {
2596: Mat_SeqAIJ *a = (Mat_SeqAIJ*)inA->data;
2597: PetscScalar oalpha = alpha;
2599: PetscBLASInt one = 1,bnz;
2602: PetscBLASIntCast(a->nz,&bnz);
2603: PetscStackCallBLAS("BLASscal",BLASscal_(&bnz,&oalpha,a->a,&one));
2604: PetscLogFlops(a->nz);
2605: MatSeqAIJInvalidateDiagonal(inA);
2606: return(0);
2607: }
2609: PetscErrorCode MatDestroySubMatrix_Private(Mat_SubSppt *submatj)
2610: {
2612: PetscInt i;
2615: if (!submatj->id) { /* delete data that are linked only to submats[id=0] */
2616: PetscFree4(submatj->sbuf1,submatj->ptr,submatj->tmp,submatj->ctr);
2618: for (i=0; i<submatj->nrqr; ++i) {
2619: PetscFree(submatj->sbuf2[i]);
2620: }
2621: PetscFree3(submatj->sbuf2,submatj->req_size,submatj->req_source1);
2623: if (submatj->rbuf1) {
2624: PetscFree(submatj->rbuf1[0]);
2625: PetscFree(submatj->rbuf1);
2626: }
2628: for (i=0; i<submatj->nrqs; ++i) {
2629: PetscFree(submatj->rbuf3[i]);
2630: }
2631: PetscFree3(submatj->req_source2,submatj->rbuf2,submatj->rbuf3);
2632: PetscFree(submatj->pa);
2633: }
2635: #if defined(PETSC_USE_CTABLE)
2636: PetscTableDestroy((PetscTable*)&submatj->rmap);
2637: if (submatj->cmap_loc) {PetscFree(submatj->cmap_loc);}
2638: PetscFree(submatj->rmap_loc);
2639: #else
2640: PetscFree(submatj->rmap);
2641: #endif
2643: if (!submatj->allcolumns) {
2644: #if defined(PETSC_USE_CTABLE)
2645: PetscTableDestroy((PetscTable*)&submatj->cmap);
2646: #else
2647: PetscFree(submatj->cmap);
2648: #endif
2649: }
2650: PetscFree(submatj->row2proc);
2652: PetscFree(submatj);
2653: return(0);
2654: }
2656: PetscErrorCode MatDestroySubMatrix_SeqAIJ(Mat C)
2657: {
2659: Mat_SeqAIJ *c = (Mat_SeqAIJ*)C->data;
2660: Mat_SubSppt *submatj = c->submatis1;
2663: (*submatj->destroy)(C);
2664: MatDestroySubMatrix_Private(submatj);
2665: return(0);
2666: }
2668: PetscErrorCode MatDestroySubMatrices_SeqAIJ(PetscInt n,Mat *mat[])
2669: {
2671: PetscInt i;
2672: Mat C;
2673: Mat_SeqAIJ *c;
2674: Mat_SubSppt *submatj;
2677: for (i=0; i<n; i++) {
2678: C = (*mat)[i];
2679: c = (Mat_SeqAIJ*)C->data;
2680: submatj = c->submatis1;
2681: if (submatj) {
2682: if (--((PetscObject)C)->refct <= 0) {
2683: (*submatj->destroy)(C);
2684: MatDestroySubMatrix_Private(submatj);
2685: PetscFree(C->defaultvectype);
2686: PetscLayoutDestroy(&C->rmap);
2687: PetscLayoutDestroy(&C->cmap);
2688: PetscHeaderDestroy(&C);
2689: }
2690: } else {
2691: MatDestroy(&C);
2692: }
2693: }
2695: /* Destroy Dummy submatrices created for reuse */
2696: MatDestroySubMatrices_Dummy(n,mat);
2698: PetscFree(*mat);
2699: return(0);
2700: }
2702: PetscErrorCode MatCreateSubMatrices_SeqAIJ(Mat A,PetscInt n,const IS irow[],const IS icol[],MatReuse scall,Mat *B[])
2703: {
2705: PetscInt i;
2708: if (scall == MAT_INITIAL_MATRIX) {
2709: PetscCalloc1(n+1,B);
2710: }
2712: for (i=0; i<n; i++) {
2713: MatCreateSubMatrix_SeqAIJ(A,irow[i],icol[i],PETSC_DECIDE,scall,&(*B)[i]);
2714: }
2715: return(0);
2716: }
2718: PetscErrorCode MatIncreaseOverlap_SeqAIJ(Mat A,PetscInt is_max,IS is[],PetscInt ov)
2719: {
2720: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
2722: PetscInt row,i,j,k,l,m,n,*nidx,isz,val;
2723: const PetscInt *idx;
2724: PetscInt start,end,*ai,*aj;
2725: PetscBT table;
2728: m = A->rmap->n;
2729: ai = a->i;
2730: aj = a->j;
2732: if (ov < 0) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"illegal negative overlap value used");
2734: PetscMalloc1(m+1,&nidx);
2735: PetscBTCreate(m,&table);
2737: for (i=0; i<is_max; i++) {
2738: /* Initialize the two local arrays */
2739: isz = 0;
2740: PetscBTMemzero(m,table);
2742: /* Extract the indices, assume there can be duplicate entries */
2743: ISGetIndices(is[i],&idx);
2744: ISGetLocalSize(is[i],&n);
2746: /* Enter these into the temp arrays. I.e., mark table[row], enter row into new index */
2747: for (j=0; j<n; ++j) {
2748: if (!PetscBTLookupSet(table,idx[j])) nidx[isz++] = idx[j];
2749: }
2750: ISRestoreIndices(is[i],&idx);
2751: ISDestroy(&is[i]);
2753: k = 0;
2754: for (j=0; j<ov; j++) { /* for each overlap */
2755: n = isz;
2756: for (; k<n; k++) { /* do only those rows in nidx[k], which are not done yet */
2757: row = nidx[k];
2758: start = ai[row];
2759: end = ai[row+1];
2760: for (l = start; l<end; l++) {
2761: val = aj[l];
2762: if (!PetscBTLookupSet(table,val)) nidx[isz++] = val;
2763: }
2764: }
2765: }
2766: ISCreateGeneral(PETSC_COMM_SELF,isz,nidx,PETSC_COPY_VALUES,(is+i));
2767: }
2768: PetscBTDestroy(&table);
2769: PetscFree(nidx);
2770: return(0);
2771: }
2773: /* -------------------------------------------------------------- */
2774: PetscErrorCode MatPermute_SeqAIJ(Mat A,IS rowp,IS colp,Mat *B)
2775: {
2776: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
2778: PetscInt i,nz = 0,m = A->rmap->n,n = A->cmap->n;
2779: const PetscInt *row,*col;
2780: PetscInt *cnew,j,*lens;
2781: IS icolp,irowp;
2782: PetscInt *cwork = NULL;
2783: PetscScalar *vwork = NULL;
2786: ISInvertPermutation(rowp,PETSC_DECIDE,&irowp);
2787: ISGetIndices(irowp,&row);
2788: ISInvertPermutation(colp,PETSC_DECIDE,&icolp);
2789: ISGetIndices(icolp,&col);
2791: /* determine lengths of permuted rows */
2792: PetscMalloc1(m+1,&lens);
2793: for (i=0; i<m; i++) lens[row[i]] = a->i[i+1] - a->i[i];
2794: MatCreate(PetscObjectComm((PetscObject)A),B);
2795: MatSetSizes(*B,m,n,m,n);
2796: MatSetBlockSizesFromMats(*B,A,A);
2797: MatSetType(*B,((PetscObject)A)->type_name);
2798: MatSeqAIJSetPreallocation_SeqAIJ(*B,0,lens);
2799: PetscFree(lens);
2801: PetscMalloc1(n,&cnew);
2802: for (i=0; i<m; i++) {
2803: MatGetRow_SeqAIJ(A,i,&nz,&cwork,&vwork);
2804: for (j=0; j<nz; j++) cnew[j] = col[cwork[j]];
2805: MatSetValues_SeqAIJ(*B,1,&row[i],nz,cnew,vwork,INSERT_VALUES);
2806: MatRestoreRow_SeqAIJ(A,i,&nz,&cwork,&vwork);
2807: }
2808: PetscFree(cnew);
2810: (*B)->assembled = PETSC_FALSE;
2812: MatAssemblyBegin(*B,MAT_FINAL_ASSEMBLY);
2813: MatAssemblyEnd(*B,MAT_FINAL_ASSEMBLY);
2814: ISRestoreIndices(irowp,&row);
2815: ISRestoreIndices(icolp,&col);
2816: ISDestroy(&irowp);
2817: ISDestroy(&icolp);
2818: return(0);
2819: }
2821: PetscErrorCode MatCopy_SeqAIJ(Mat A,Mat B,MatStructure str)
2822: {
2826: /* If the two matrices have the same copy implementation, use fast copy. */
2827: if (str == SAME_NONZERO_PATTERN && (A->ops->copy == B->ops->copy)) {
2828: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
2829: Mat_SeqAIJ *b = (Mat_SeqAIJ*)B->data;
2831: if (a->i[A->rmap->n] != b->i[B->rmap->n]) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_INCOMP,"Number of nonzeros in two matrices are different");
2832: PetscMemcpy(b->a,a->a,(a->i[A->rmap->n])*sizeof(PetscScalar));
2833: PetscObjectStateIncrease((PetscObject)B);
2834: } else {
2835: MatCopy_Basic(A,B,str);
2836: }
2837: return(0);
2838: }
2840: PetscErrorCode MatSetUp_SeqAIJ(Mat A)
2841: {
2845: MatSeqAIJSetPreallocation_SeqAIJ(A,PETSC_DEFAULT,0);
2846: return(0);
2847: }
2849: PetscErrorCode MatSeqAIJGetArray_SeqAIJ(Mat A,PetscScalar *array[])
2850: {
2851: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
2854: *array = a->a;
2855: return(0);
2856: }
2858: PetscErrorCode MatSeqAIJRestoreArray_SeqAIJ(Mat A,PetscScalar *array[])
2859: {
2861: return(0);
2862: }
2864: /*
2865: Computes the number of nonzeros per row needed for preallocation when X and Y
2866: have different nonzero structure.
2867: */
2868: PetscErrorCode MatAXPYGetPreallocation_SeqX_private(PetscInt m,const PetscInt *xi,const PetscInt *xj,const PetscInt *yi,const PetscInt *yj,PetscInt *nnz)
2869: {
2870: PetscInt i,j,k,nzx,nzy;
2873: /* Set the number of nonzeros in the new matrix */
2874: for (i=0; i<m; i++) {
2875: const PetscInt *xjj = xj+xi[i],*yjj = yj+yi[i];
2876: nzx = xi[i+1] - xi[i];
2877: nzy = yi[i+1] - yi[i];
2878: nnz[i] = 0;
2879: for (j=0,k=0; j<nzx; j++) { /* Point in X */
2880: for (; k<nzy && yjj[k]<xjj[j]; k++) nnz[i]++; /* Catch up to X */
2881: if (k<nzy && yjj[k]==xjj[j]) k++; /* Skip duplicate */
2882: nnz[i]++;
2883: }
2884: for (; k<nzy; k++) nnz[i]++;
2885: }
2886: return(0);
2887: }
2889: PetscErrorCode MatAXPYGetPreallocation_SeqAIJ(Mat Y,Mat X,PetscInt *nnz)
2890: {
2891: PetscInt m = Y->rmap->N;
2892: Mat_SeqAIJ *x = (Mat_SeqAIJ*)X->data;
2893: Mat_SeqAIJ *y = (Mat_SeqAIJ*)Y->data;
2897: /* Set the number of nonzeros in the new matrix */
2898: MatAXPYGetPreallocation_SeqX_private(m,x->i,x->j,y->i,y->j,nnz);
2899: return(0);
2900: }
2902: PetscErrorCode MatAXPY_SeqAIJ(Mat Y,PetscScalar a,Mat X,MatStructure str)
2903: {
2905: Mat_SeqAIJ *x = (Mat_SeqAIJ*)X->data,*y = (Mat_SeqAIJ*)Y->data;
2906: PetscBLASInt one=1,bnz;
2909: PetscBLASIntCast(x->nz,&bnz);
2910: if (str == SAME_NONZERO_PATTERN) {
2911: PetscScalar alpha = a;
2912: PetscStackCallBLAS("BLASaxpy",BLASaxpy_(&bnz,&alpha,x->a,&one,y->a,&one));
2913: MatSeqAIJInvalidateDiagonal(Y);
2914: PetscObjectStateIncrease((PetscObject)Y);
2915: } else if (str == SUBSET_NONZERO_PATTERN) { /* nonzeros of X is a subset of Y's */
2916: MatAXPY_Basic(Y,a,X,str);
2917: } else {
2918: Mat B;
2919: PetscInt *nnz;
2920: PetscMalloc1(Y->rmap->N,&nnz);
2921: MatCreate(PetscObjectComm((PetscObject)Y),&B);
2922: PetscObjectSetName((PetscObject)B,((PetscObject)Y)->name);
2923: MatSetSizes(B,Y->rmap->n,Y->cmap->n,Y->rmap->N,Y->cmap->N);
2924: MatSetBlockSizesFromMats(B,Y,Y);
2925: MatSetType(B,(MatType) ((PetscObject)Y)->type_name);
2926: MatAXPYGetPreallocation_SeqAIJ(Y,X,nnz);
2927: MatSeqAIJSetPreallocation(B,0,nnz);
2928: MatAXPY_BasicWithPreallocation(B,Y,a,X,str);
2929: MatHeaderReplace(Y,&B);
2930: PetscFree(nnz);
2931: }
2932: return(0);
2933: }
2935: PetscErrorCode MatConjugate_SeqAIJ(Mat mat)
2936: {
2937: #if defined(PETSC_USE_COMPLEX)
2938: Mat_SeqAIJ *aij = (Mat_SeqAIJ*)mat->data;
2939: PetscInt i,nz;
2940: PetscScalar *a;
2943: nz = aij->nz;
2944: a = aij->a;
2945: for (i=0; i<nz; i++) a[i] = PetscConj(a[i]);
2946: #else
2948: #endif
2949: return(0);
2950: }
2952: PetscErrorCode MatGetRowMaxAbs_SeqAIJ(Mat A,Vec v,PetscInt idx[])
2953: {
2954: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
2956: PetscInt i,j,m = A->rmap->n,*ai,*aj,ncols,n;
2957: PetscReal atmp;
2958: PetscScalar *x;
2959: MatScalar *aa;
2962: if (A->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2963: aa = a->a;
2964: ai = a->i;
2965: aj = a->j;
2967: VecSet(v,0.0);
2968: VecGetArray(v,&x);
2969: VecGetLocalSize(v,&n);
2970: if (n != A->rmap->n) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Nonconforming matrix and vector");
2971: for (i=0; i<m; i++) {
2972: ncols = ai[1] - ai[0]; ai++;
2973: x[i] = 0.0;
2974: for (j=0; j<ncols; j++) {
2975: atmp = PetscAbsScalar(*aa);
2976: if (PetscAbsScalar(x[i]) < atmp) {x[i] = atmp; if (idx) idx[i] = *aj;}
2977: aa++; aj++;
2978: }
2979: }
2980: VecRestoreArray(v,&x);
2981: return(0);
2982: }
2984: PetscErrorCode MatGetRowMax_SeqAIJ(Mat A,Vec v,PetscInt idx[])
2985: {
2986: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
2988: PetscInt i,j,m = A->rmap->n,*ai,*aj,ncols,n;
2989: PetscScalar *x;
2990: MatScalar *aa;
2993: if (A->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2994: aa = a->a;
2995: ai = a->i;
2996: aj = a->j;
2998: VecSet(v,0.0);
2999: VecGetArray(v,&x);
3000: VecGetLocalSize(v,&n);
3001: if (n != A->rmap->n) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Nonconforming matrix and vector");
3002: for (i=0; i<m; i++) {
3003: ncols = ai[1] - ai[0]; ai++;
3004: if (ncols == A->cmap->n) { /* row is dense */
3005: x[i] = *aa; if (idx) idx[i] = 0;
3006: } else { /* row is sparse so already KNOW maximum is 0.0 or higher */
3007: x[i] = 0.0;
3008: if (idx) {
3009: idx[i] = 0; /* in case ncols is zero */
3010: for (j=0;j<ncols;j++) { /* find first implicit 0.0 in the row */
3011: if (aj[j] > j) {
3012: idx[i] = j;
3013: break;
3014: }
3015: }
3016: }
3017: }
3018: for (j=0; j<ncols; j++) {
3019: if (PetscRealPart(x[i]) < PetscRealPart(*aa)) {x[i] = *aa; if (idx) idx[i] = *aj;}
3020: aa++; aj++;
3021: }
3022: }
3023: VecRestoreArray(v,&x);
3024: return(0);
3025: }
3027: PetscErrorCode MatGetRowMinAbs_SeqAIJ(Mat A,Vec v,PetscInt idx[])
3028: {
3029: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
3031: PetscInt i,j,m = A->rmap->n,*ai,*aj,ncols,n;
3032: PetscReal atmp;
3033: PetscScalar *x;
3034: MatScalar *aa;
3037: if (A->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3038: aa = a->a;
3039: ai = a->i;
3040: aj = a->j;
3042: VecSet(v,0.0);
3043: VecGetArray(v,&x);
3044: VecGetLocalSize(v,&n);
3045: 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);
3046: for (i=0; i<m; i++) {
3047: ncols = ai[1] - ai[0]; ai++;
3048: if (ncols) {
3049: /* Get first nonzero */
3050: for (j = 0; j < ncols; j++) {
3051: atmp = PetscAbsScalar(aa[j]);
3052: if (atmp > 1.0e-12) {
3053: x[i] = atmp;
3054: if (idx) idx[i] = aj[j];
3055: break;
3056: }
3057: }
3058: if (j == ncols) {x[i] = PetscAbsScalar(*aa); if (idx) idx[i] = *aj;}
3059: } else {
3060: x[i] = 0.0; if (idx) idx[i] = 0;
3061: }
3062: for (j = 0; j < ncols; j++) {
3063: atmp = PetscAbsScalar(*aa);
3064: if (atmp > 1.0e-12 && PetscAbsScalar(x[i]) > atmp) {x[i] = atmp; if (idx) idx[i] = *aj;}
3065: aa++; aj++;
3066: }
3067: }
3068: VecRestoreArray(v,&x);
3069: return(0);
3070: }
3072: PetscErrorCode MatGetRowMin_SeqAIJ(Mat A,Vec v,PetscInt idx[])
3073: {
3074: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
3075: PetscErrorCode ierr;
3076: PetscInt i,j,m = A->rmap->n,ncols,n;
3077: const PetscInt *ai,*aj;
3078: PetscScalar *x;
3079: const MatScalar *aa;
3082: if (A->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3083: aa = a->a;
3084: ai = a->i;
3085: aj = a->j;
3087: VecSet(v,0.0);
3088: VecGetArray(v,&x);
3089: VecGetLocalSize(v,&n);
3090: if (n != A->rmap->n) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Nonconforming matrix and vector");
3091: for (i=0; i<m; i++) {
3092: ncols = ai[1] - ai[0]; ai++;
3093: if (ncols == A->cmap->n) { /* row is dense */
3094: x[i] = *aa; if (idx) idx[i] = 0;
3095: } else { /* row is sparse so already KNOW minimum is 0.0 or lower */
3096: x[i] = 0.0;
3097: if (idx) { /* find first implicit 0.0 in the row */
3098: idx[i] = 0; /* in case ncols is zero */
3099: for (j=0; j<ncols; j++) {
3100: if (aj[j] > j) {
3101: idx[i] = j;
3102: break;
3103: }
3104: }
3105: }
3106: }
3107: for (j=0; j<ncols; j++) {
3108: if (PetscRealPart(x[i]) > PetscRealPart(*aa)) {x[i] = *aa; if (idx) idx[i] = *aj;}
3109: aa++; aj++;
3110: }
3111: }
3112: VecRestoreArray(v,&x);
3113: return(0);
3114: }
3116: PetscErrorCode MatInvertBlockDiagonal_SeqAIJ(Mat A,const PetscScalar **values)
3117: {
3118: Mat_SeqAIJ *a = (Mat_SeqAIJ*) A->data;
3119: PetscErrorCode ierr;
3120: PetscInt i,bs = PetscAbs(A->rmap->bs),mbs = A->rmap->n/bs,ipvt[5],bs2 = bs*bs,*v_pivots,ij[7],*IJ,j;
3121: MatScalar *diag,work[25],*v_work;
3122: const PetscReal shift = 0.0;
3123: PetscBool allowzeropivot,zeropivotdetected=PETSC_FALSE;
3126: allowzeropivot = PetscNot(A->erroriffailure);
3127: if (a->ibdiagvalid) {
3128: if (values) *values = a->ibdiag;
3129: return(0);
3130: }
3131: MatMarkDiagonal_SeqAIJ(A);
3132: if (!a->ibdiag) {
3133: PetscMalloc1(bs2*mbs,&a->ibdiag);
3134: PetscLogObjectMemory((PetscObject)A,bs2*mbs*sizeof(PetscScalar));
3135: }
3136: diag = a->ibdiag;
3137: if (values) *values = a->ibdiag;
3138: /* factor and invert each block */
3139: switch (bs) {
3140: case 1:
3141: for (i=0; i<mbs; i++) {
3142: MatGetValues(A,1,&i,1,&i,diag+i);
3143: if (PetscAbsScalar(diag[i] + shift) < PETSC_MACHINE_EPSILON) {
3144: if (allowzeropivot) {
3145: A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
3146: A->factorerror_zeropivot_value = PetscAbsScalar(diag[i]);
3147: A->factorerror_zeropivot_row = i;
3148: PetscInfo3(A,"Zero pivot, row %D pivot %g tolerance %g\n",i,(double)PetscAbsScalar(diag[i]),(double)PETSC_MACHINE_EPSILON);
3149: } 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);
3150: }
3151: diag[i] = (PetscScalar)1.0 / (diag[i] + shift);
3152: }
3153: break;
3154: case 2:
3155: for (i=0; i<mbs; i++) {
3156: ij[0] = 2*i; ij[1] = 2*i + 1;
3157: MatGetValues(A,2,ij,2,ij,diag);
3158: PetscKernel_A_gets_inverse_A_2(diag,shift,allowzeropivot,&zeropivotdetected);
3159: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
3160: PetscKernel_A_gets_transpose_A_2(diag);
3161: diag += 4;
3162: }
3163: break;
3164: case 3:
3165: for (i=0; i<mbs; i++) {
3166: ij[0] = 3*i; ij[1] = 3*i + 1; ij[2] = 3*i + 2;
3167: MatGetValues(A,3,ij,3,ij,diag);
3168: PetscKernel_A_gets_inverse_A_3(diag,shift,allowzeropivot,&zeropivotdetected);
3169: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
3170: PetscKernel_A_gets_transpose_A_3(diag);
3171: diag += 9;
3172: }
3173: break;
3174: case 4:
3175: for (i=0; i<mbs; i++) {
3176: ij[0] = 4*i; ij[1] = 4*i + 1; ij[2] = 4*i + 2; ij[3] = 4*i + 3;
3177: MatGetValues(A,4,ij,4,ij,diag);
3178: PetscKernel_A_gets_inverse_A_4(diag,shift,allowzeropivot,&zeropivotdetected);
3179: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
3180: PetscKernel_A_gets_transpose_A_4(diag);
3181: diag += 16;
3182: }
3183: break;
3184: case 5:
3185: for (i=0; i<mbs; i++) {
3186: ij[0] = 5*i; ij[1] = 5*i + 1; ij[2] = 5*i + 2; ij[3] = 5*i + 3; ij[4] = 5*i + 4;
3187: MatGetValues(A,5,ij,5,ij,diag);
3188: PetscKernel_A_gets_inverse_A_5(diag,ipvt,work,shift,allowzeropivot,&zeropivotdetected);
3189: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
3190: PetscKernel_A_gets_transpose_A_5(diag);
3191: diag += 25;
3192: }
3193: break;
3194: case 6:
3195: for (i=0; i<mbs; i++) {
3196: 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;
3197: MatGetValues(A,6,ij,6,ij,diag);
3198: PetscKernel_A_gets_inverse_A_6(diag,shift,allowzeropivot,&zeropivotdetected);
3199: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
3200: PetscKernel_A_gets_transpose_A_6(diag);
3201: diag += 36;
3202: }
3203: break;
3204: case 7:
3205: for (i=0; i<mbs; i++) {
3206: 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;
3207: MatGetValues(A,7,ij,7,ij,diag);
3208: PetscKernel_A_gets_inverse_A_7(diag,shift,allowzeropivot,&zeropivotdetected);
3209: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
3210: PetscKernel_A_gets_transpose_A_7(diag);
3211: diag += 49;
3212: }
3213: break;
3214: default:
3215: PetscMalloc3(bs,&v_work,bs,&v_pivots,bs,&IJ);
3216: for (i=0; i<mbs; i++) {
3217: for (j=0; j<bs; j++) {
3218: IJ[j] = bs*i + j;
3219: }
3220: MatGetValues(A,bs,IJ,bs,IJ,diag);
3221: PetscKernel_A_gets_inverse_A(bs,diag,v_pivots,v_work,allowzeropivot,&zeropivotdetected);
3222: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
3223: PetscKernel_A_gets_transpose_A_N(diag,bs);
3224: diag += bs2;
3225: }
3226: PetscFree3(v_work,v_pivots,IJ);
3227: }
3228: a->ibdiagvalid = PETSC_TRUE;
3229: return(0);
3230: }
3232: static PetscErrorCode MatSetRandom_SeqAIJ(Mat x,PetscRandom rctx)
3233: {
3235: Mat_SeqAIJ *aij = (Mat_SeqAIJ*)x->data;
3236: PetscScalar a;
3237: PetscInt m,n,i,j,col;
3240: if (!x->assembled) {
3241: MatGetSize(x,&m,&n);
3242: for (i=0; i<m; i++) {
3243: for (j=0; j<aij->imax[i]; j++) {
3244: PetscRandomGetValue(rctx,&a);
3245: col = (PetscInt)(n*PetscRealPart(a));
3246: MatSetValues(x,1,&i,1,&col,&a,ADD_VALUES);
3247: }
3248: }
3249: } else SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Not yet coded");
3250: MatAssemblyBegin(x,MAT_FINAL_ASSEMBLY);
3251: MatAssemblyEnd(x,MAT_FINAL_ASSEMBLY);
3252: return(0);
3253: }
3255: /* -------------------------------------------------------------------*/
3256: static struct _MatOps MatOps_Values = { MatSetValues_SeqAIJ,
3257: MatGetRow_SeqAIJ,
3258: MatRestoreRow_SeqAIJ,
3259: MatMult_SeqAIJ,
3260: /* 4*/ MatMultAdd_SeqAIJ,
3261: MatMultTranspose_SeqAIJ,
3262: MatMultTransposeAdd_SeqAIJ,
3263: 0,
3264: 0,
3265: 0,
3266: /* 10*/ 0,
3267: MatLUFactor_SeqAIJ,
3268: 0,
3269: MatSOR_SeqAIJ,
3270: MatTranspose_SeqAIJ_FAST,
3271: /*1 5*/ MatGetInfo_SeqAIJ,
3272: MatEqual_SeqAIJ,
3273: MatGetDiagonal_SeqAIJ,
3274: MatDiagonalScale_SeqAIJ,
3275: MatNorm_SeqAIJ,
3276: /* 20*/ 0,
3277: MatAssemblyEnd_SeqAIJ,
3278: MatSetOption_SeqAIJ,
3279: MatZeroEntries_SeqAIJ,
3280: /* 24*/ MatZeroRows_SeqAIJ,
3281: 0,
3282: 0,
3283: 0,
3284: 0,
3285: /* 29*/ MatSetUp_SeqAIJ,
3286: 0,
3287: 0,
3288: 0,
3289: 0,
3290: /* 34*/ MatDuplicate_SeqAIJ,
3291: 0,
3292: 0,
3293: MatILUFactor_SeqAIJ,
3294: 0,
3295: /* 39*/ MatAXPY_SeqAIJ,
3296: MatCreateSubMatrices_SeqAIJ,
3297: MatIncreaseOverlap_SeqAIJ,
3298: MatGetValues_SeqAIJ,
3299: MatCopy_SeqAIJ,
3300: /* 44*/ MatGetRowMax_SeqAIJ,
3301: MatScale_SeqAIJ,
3302: MatShift_SeqAIJ,
3303: MatDiagonalSet_SeqAIJ,
3304: MatZeroRowsColumns_SeqAIJ,
3305: /* 49*/ MatSetRandom_SeqAIJ,
3306: MatGetRowIJ_SeqAIJ,
3307: MatRestoreRowIJ_SeqAIJ,
3308: MatGetColumnIJ_SeqAIJ,
3309: MatRestoreColumnIJ_SeqAIJ,
3310: /* 54*/ MatFDColoringCreate_SeqXAIJ,
3311: 0,
3312: 0,
3313: MatPermute_SeqAIJ,
3314: 0,
3315: /* 59*/ 0,
3316: MatDestroy_SeqAIJ,
3317: MatView_SeqAIJ,
3318: 0,
3319: MatMatMatMult_SeqAIJ_SeqAIJ_SeqAIJ,
3320: /* 64*/ MatMatMatMultSymbolic_SeqAIJ_SeqAIJ_SeqAIJ,
3321: MatMatMatMultNumeric_SeqAIJ_SeqAIJ_SeqAIJ,
3322: 0,
3323: 0,
3324: 0,
3325: /* 69*/ MatGetRowMaxAbs_SeqAIJ,
3326: MatGetRowMinAbs_SeqAIJ,
3327: 0,
3328: 0,
3329: 0,
3330: /* 74*/ 0,
3331: MatFDColoringApply_AIJ,
3332: 0,
3333: 0,
3334: 0,
3335: /* 79*/ MatFindZeroDiagonals_SeqAIJ,
3336: 0,
3337: 0,
3338: 0,
3339: MatLoad_SeqAIJ,
3340: /* 84*/ MatIsSymmetric_SeqAIJ,
3341: MatIsHermitian_SeqAIJ,
3342: 0,
3343: 0,
3344: 0,
3345: /* 89*/ MatMatMult_SeqAIJ_SeqAIJ,
3346: MatMatMultSymbolic_SeqAIJ_SeqAIJ,
3347: MatMatMultNumeric_SeqAIJ_SeqAIJ,
3348: MatPtAP_SeqAIJ_SeqAIJ,
3349: MatPtAPSymbolic_SeqAIJ_SeqAIJ_SparseAxpy,
3350: /* 94*/ MatPtAPNumeric_SeqAIJ_SeqAIJ_SparseAxpy,
3351: MatMatTransposeMult_SeqAIJ_SeqAIJ,
3352: MatMatTransposeMultSymbolic_SeqAIJ_SeqAIJ,
3353: MatMatTransposeMultNumeric_SeqAIJ_SeqAIJ,
3354: 0,
3355: /* 99*/ 0,
3356: 0,
3357: 0,
3358: MatConjugate_SeqAIJ,
3359: 0,
3360: /*104*/ MatSetValuesRow_SeqAIJ,
3361: MatRealPart_SeqAIJ,
3362: MatImaginaryPart_SeqAIJ,
3363: 0,
3364: 0,
3365: /*109*/ MatMatSolve_SeqAIJ,
3366: 0,
3367: MatGetRowMin_SeqAIJ,
3368: 0,
3369: MatMissingDiagonal_SeqAIJ,
3370: /*114*/ 0,
3371: 0,
3372: 0,
3373: 0,
3374: 0,
3375: /*119*/ 0,
3376: 0,
3377: 0,
3378: 0,
3379: MatGetMultiProcBlock_SeqAIJ,
3380: /*124*/ MatFindNonzeroRows_SeqAIJ,
3381: MatGetColumnNorms_SeqAIJ,
3382: MatInvertBlockDiagonal_SeqAIJ,
3383: MatInvertVariableBlockDiagonal_SeqAIJ,
3384: 0,
3385: /*129*/ 0,
3386: MatTransposeMatMult_SeqAIJ_SeqAIJ,
3387: MatTransposeMatMultSymbolic_SeqAIJ_SeqAIJ,
3388: MatTransposeMatMultNumeric_SeqAIJ_SeqAIJ,
3389: MatTransposeColoringCreate_SeqAIJ,
3390: /*134*/ MatTransColoringApplySpToDen_SeqAIJ,
3391: MatTransColoringApplyDenToSp_SeqAIJ,
3392: MatRARt_SeqAIJ_SeqAIJ,
3393: MatRARtSymbolic_SeqAIJ_SeqAIJ,
3394: MatRARtNumeric_SeqAIJ_SeqAIJ,
3395: /*139*/0,
3396: 0,
3397: 0,
3398: MatFDColoringSetUp_SeqXAIJ,
3399: MatFindOffBlockDiagonalEntries_SeqAIJ,
3400: /*144*/MatCreateMPIMatConcatenateSeqMat_SeqAIJ,
3401: MatDestroySubMatrices_SeqAIJ
3402: };
3404: PetscErrorCode MatSeqAIJSetColumnIndices_SeqAIJ(Mat mat,PetscInt *indices)
3405: {
3406: Mat_SeqAIJ *aij = (Mat_SeqAIJ*)mat->data;
3407: PetscInt i,nz,n;
3410: nz = aij->maxnz;
3411: n = mat->rmap->n;
3412: for (i=0; i<nz; i++) {
3413: aij->j[i] = indices[i];
3414: }
3415: aij->nz = nz;
3416: for (i=0; i<n; i++) {
3417: aij->ilen[i] = aij->imax[i];
3418: }
3419: return(0);
3420: }
3422: /*
3423: * When a sparse matrix has many zero columns, we should compact them out to save the space
3424: * This happens in MatPtAPSymbolic_MPIAIJ_MPIAIJ_scalable()
3425: * */
3426: PetscErrorCode MatSeqAIJCompactOutExtraColumns_SeqAIJ(Mat mat, ISLocalToGlobalMapping *mapping)
3427: {
3428: Mat_SeqAIJ *aij = (Mat_SeqAIJ*)mat->data;
3429: PetscTable gid1_lid1;
3430: PetscTablePosition tpos;
3431: PetscInt gid,lid,i,j,ncols,ec;
3432: PetscInt *garray;
3433: PetscErrorCode ierr;
3438: /* use a table */
3439: PetscTableCreate(mat->rmap->n,mat->cmap->N+1,&gid1_lid1);
3440: ec = 0;
3441: for (i=0; i<mat->rmap->n; i++) {
3442: ncols = aij->i[i+1] - aij->i[i];
3443: for (j=0; j<ncols; j++) {
3444: PetscInt data,gid1 = aij->j[aij->i[i] + j] + 1;
3445: PetscTableFind(gid1_lid1,gid1,&data);
3446: if (!data) {
3447: /* one based table */
3448: PetscTableAdd(gid1_lid1,gid1,++ec,INSERT_VALUES);
3449: }
3450: }
3451: }
3452: /* form array of columns we need */
3453: PetscMalloc1(ec+1,&garray);
3454: PetscTableGetHeadPosition(gid1_lid1,&tpos);
3455: while (tpos) {
3456: PetscTableGetNext(gid1_lid1,&tpos,&gid,&lid);
3457: gid--;
3458: lid--;
3459: garray[lid] = gid;
3460: }
3461: PetscSortInt(ec,garray); /* sort, and rebuild */
3462: PetscTableRemoveAll(gid1_lid1);
3463: for (i=0; i<ec; i++) {
3464: PetscTableAdd(gid1_lid1,garray[i]+1,i+1,INSERT_VALUES);
3465: }
3466: /* compact out the extra columns in B */
3467: for (i=0; i<mat->rmap->n; i++) {
3468: ncols = aij->i[i+1] - aij->i[i];
3469: for (j=0; j<ncols; j++) {
3470: PetscInt gid1 = aij->j[aij->i[i] + j] + 1;
3471: PetscTableFind(gid1_lid1,gid1,&lid);
3472: lid--;
3473: aij->j[aij->i[i] + j] = lid;
3474: }
3475: }
3476: mat->cmap->n = mat->cmap->N = ec;
3477: mat->cmap->bs = 1;
3479: PetscTableDestroy(&gid1_lid1);
3480: PetscLayoutSetUp((mat->cmap));
3481: ISLocalToGlobalMappingCreate(PETSC_COMM_SELF,mat->cmap->bs,mat->cmap->n,garray,PETSC_OWN_POINTER,mapping);
3482: ISLocalToGlobalMappingSetType(*mapping,ISLOCALTOGLOBALMAPPINGHASH);
3483: return(0);
3484: }
3486: /*@
3487: MatSeqAIJSetColumnIndices - Set the column indices for all the rows
3488: in the matrix.
3490: Input Parameters:
3491: + mat - the SeqAIJ matrix
3492: - indices - the column indices
3494: Level: advanced
3496: Notes:
3497: This can be called if you have precomputed the nonzero structure of the
3498: matrix and want to provide it to the matrix object to improve the performance
3499: of the MatSetValues() operation.
3501: You MUST have set the correct numbers of nonzeros per row in the call to
3502: MatCreateSeqAIJ(), and the columns indices MUST be sorted.
3504: MUST be called before any calls to MatSetValues();
3506: The indices should start with zero, not one.
3508: @*/
3509: PetscErrorCode MatSeqAIJSetColumnIndices(Mat mat,PetscInt *indices)
3510: {
3516: PetscUseMethod(mat,"MatSeqAIJSetColumnIndices_C",(Mat,PetscInt*),(mat,indices));
3517: return(0);
3518: }
3520: /* ----------------------------------------------------------------------------------------*/
3522: PetscErrorCode MatStoreValues_SeqAIJ(Mat mat)
3523: {
3524: Mat_SeqAIJ *aij = (Mat_SeqAIJ*)mat->data;
3526: size_t nz = aij->i[mat->rmap->n];
3529: if (!aij->nonew) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Must call MatSetOption(A,MAT_NEW_NONZERO_LOCATIONS,PETSC_FALSE);first");
3531: /* allocate space for values if not already there */
3532: if (!aij->saved_values) {
3533: PetscMalloc1(nz+1,&aij->saved_values);
3534: PetscLogObjectMemory((PetscObject)mat,(nz+1)*sizeof(PetscScalar));
3535: }
3537: /* copy values over */
3538: PetscMemcpy(aij->saved_values,aij->a,nz*sizeof(PetscScalar));
3539: return(0);
3540: }
3542: /*@
3543: MatStoreValues - Stashes a copy of the matrix values; this allows, for
3544: example, reuse of the linear part of a Jacobian, while recomputing the
3545: nonlinear portion.
3547: Collect on Mat
3549: Input Parameters:
3550: . mat - the matrix (currently only AIJ matrices support this option)
3552: Level: advanced
3554: Common Usage, with SNESSolve():
3555: $ Create Jacobian matrix
3556: $ Set linear terms into matrix
3557: $ Apply boundary conditions to matrix, at this time matrix must have
3558: $ final nonzero structure (i.e. setting the nonlinear terms and applying
3559: $ boundary conditions again will not change the nonzero structure
3560: $ MatSetOption(mat,MAT_NEW_NONZERO_LOCATIONS,PETSC_FALSE);
3561: $ MatStoreValues(mat);
3562: $ Call SNESSetJacobian() with matrix
3563: $ In your Jacobian routine
3564: $ MatRetrieveValues(mat);
3565: $ Set nonlinear terms in matrix
3567: Common Usage without SNESSolve(), i.e. when you handle nonlinear solve yourself:
3568: $ // build linear portion of Jacobian
3569: $ MatSetOption(mat,MAT_NEW_NONZERO_LOCATIONS,PETSC_FALSE);
3570: $ MatStoreValues(mat);
3571: $ loop over nonlinear iterations
3572: $ MatRetrieveValues(mat);
3573: $ // call MatSetValues(mat,...) to set nonliner portion of Jacobian
3574: $ // call MatAssemblyBegin/End() on matrix
3575: $ Solve linear system with Jacobian
3576: $ endloop
3578: Notes:
3579: Matrix must already be assemblied before calling this routine
3580: Must set the matrix option MatSetOption(mat,MAT_NEW_NONZERO_LOCATIONS,PETSC_FALSE); before
3581: calling this routine.
3583: When this is called multiple times it overwrites the previous set of stored values
3584: and does not allocated additional space.
3586: .seealso: MatRetrieveValues()
3588: @*/
3589: PetscErrorCode MatStoreValues(Mat mat)
3590: {
3595: if (!mat->assembled) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3596: if (mat->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3597: PetscUseMethod(mat,"MatStoreValues_C",(Mat),(mat));
3598: return(0);
3599: }
3601: PetscErrorCode MatRetrieveValues_SeqAIJ(Mat mat)
3602: {
3603: Mat_SeqAIJ *aij = (Mat_SeqAIJ*)mat->data;
3605: PetscInt nz = aij->i[mat->rmap->n];
3608: if (!aij->nonew) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Must call MatSetOption(A,MAT_NEW_NONZERO_LOCATIONS,PETSC_FALSE);first");
3609: if (!aij->saved_values) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Must call MatStoreValues(A);first");
3610: /* copy values over */
3611: PetscMemcpy(aij->a,aij->saved_values,nz*sizeof(PetscScalar));
3612: return(0);
3613: }
3615: /*@
3616: MatRetrieveValues - Retrieves the copy of the matrix values; this allows, for
3617: example, reuse of the linear part of a Jacobian, while recomputing the
3618: nonlinear portion.
3620: Collect on Mat
3622: Input Parameters:
3623: . mat - the matrix (currently only AIJ matrices support this option)
3625: Level: advanced
3627: .seealso: MatStoreValues()
3629: @*/
3630: PetscErrorCode MatRetrieveValues(Mat mat)
3631: {
3636: if (!mat->assembled) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3637: if (mat->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3638: PetscUseMethod(mat,"MatRetrieveValues_C",(Mat),(mat));
3639: return(0);
3640: }
3643: /* --------------------------------------------------------------------------------*/
3644: /*@C
3645: MatCreateSeqAIJ - Creates a sparse matrix in AIJ (compressed row) format
3646: (the default parallel PETSc format). For good matrix assembly performance
3647: the user should preallocate the matrix storage by setting the parameter nz
3648: (or the array nnz). By setting these parameters accurately, performance
3649: during matrix assembly can be increased by more than a factor of 50.
3651: Collective on MPI_Comm
3653: Input Parameters:
3654: + comm - MPI communicator, set to PETSC_COMM_SELF
3655: . m - number of rows
3656: . n - number of columns
3657: . nz - number of nonzeros per row (same for all rows)
3658: - nnz - array containing the number of nonzeros in the various rows
3659: (possibly different for each row) or NULL
3661: Output Parameter:
3662: . A - the matrix
3664: It is recommended that one use the MatCreate(), MatSetType() and/or MatSetFromOptions(),
3665: MatXXXXSetPreallocation() paradgm instead of this routine directly.
3666: [MatXXXXSetPreallocation() is, for example, MatSeqAIJSetPreallocation]
3668: Notes:
3669: If nnz is given then nz is ignored
3671: The AIJ format (also called the Yale sparse matrix format or
3672: compressed row storage), is fully compatible with standard Fortran 77
3673: storage. That is, the stored row and column indices can begin at
3674: either one (as in Fortran) or zero. See the users' manual for details.
3676: Specify the preallocated storage with either nz or nnz (not both).
3677: Set nz=PETSC_DEFAULT and nnz=NULL for PETSc to control dynamic memory
3678: allocation. For large problems you MUST preallocate memory or you
3679: will get TERRIBLE performance, see the users' manual chapter on matrices.
3681: By default, this format uses inodes (identical nodes) when possible, to
3682: improve numerical efficiency of matrix-vector products and solves. We
3683: search for consecutive rows with the same nonzero structure, thereby
3684: reusing matrix information to achieve increased efficiency.
3686: Options Database Keys:
3687: + -mat_no_inode - Do not use inodes
3688: - -mat_inode_limit <limit> - Sets inode limit (max limit=5)
3690: Level: intermediate
3692: .seealso: MatCreate(), MatCreateAIJ(), MatSetValues(), MatSeqAIJSetColumnIndices(), MatCreateSeqAIJWithArrays()
3694: @*/
3695: PetscErrorCode MatCreateSeqAIJ(MPI_Comm comm,PetscInt m,PetscInt n,PetscInt nz,const PetscInt nnz[],Mat *A)
3696: {
3700: MatCreate(comm,A);
3701: MatSetSizes(*A,m,n,m,n);
3702: MatSetType(*A,MATSEQAIJ);
3703: MatSeqAIJSetPreallocation_SeqAIJ(*A,nz,nnz);
3704: return(0);
3705: }
3707: /*@C
3708: MatSeqAIJSetPreallocation - For good matrix assembly performance
3709: the user should preallocate the matrix storage by setting the parameter nz
3710: (or the array nnz). By setting these parameters accurately, performance
3711: during matrix assembly can be increased by more than a factor of 50.
3713: Collective on MPI_Comm
3715: Input Parameters:
3716: + B - The matrix
3717: . nz - number of nonzeros per row (same for all rows)
3718: - nnz - array containing the number of nonzeros in the various rows
3719: (possibly different for each row) or NULL
3721: Notes:
3722: If nnz is given then nz is ignored
3724: The AIJ format (also called the Yale sparse matrix format or
3725: compressed row storage), is fully compatible with standard Fortran 77
3726: storage. That is, the stored row and column indices can begin at
3727: either one (as in Fortran) or zero. See the users' manual for details.
3729: Specify the preallocated storage with either nz or nnz (not both).
3730: Set nz=PETSC_DEFAULT and nnz=NULL for PETSc to control dynamic memory
3731: allocation. For large problems you MUST preallocate memory or you
3732: will get TERRIBLE performance, see the users' manual chapter on matrices.
3734: You can call MatGetInfo() to get information on how effective the preallocation was;
3735: for example the fields mallocs,nz_allocated,nz_used,nz_unneeded;
3736: You can also run with the option -info and look for messages with the string
3737: malloc in them to see if additional memory allocation was needed.
3739: Developers: Use nz of MAT_SKIP_ALLOCATION to not allocate any space for the matrix
3740: entries or columns indices
3742: By default, this format uses inodes (identical nodes) when possible, to
3743: improve numerical efficiency of matrix-vector products and solves. We
3744: search for consecutive rows with the same nonzero structure, thereby
3745: reusing matrix information to achieve increased efficiency.
3747: Options Database Keys:
3748: + -mat_no_inode - Do not use inodes
3749: - -mat_inode_limit <limit> - Sets inode limit (max limit=5)
3751: Level: intermediate
3753: .seealso: MatCreate(), MatCreateAIJ(), MatSetValues(), MatSeqAIJSetColumnIndices(), MatCreateSeqAIJWithArrays(), MatGetInfo()
3755: @*/
3756: PetscErrorCode MatSeqAIJSetPreallocation(Mat B,PetscInt nz,const PetscInt nnz[])
3757: {
3763: PetscTryMethod(B,"MatSeqAIJSetPreallocation_C",(Mat,PetscInt,const PetscInt[]),(B,nz,nnz));
3764: return(0);
3765: }
3767: PetscErrorCode MatSeqAIJSetPreallocation_SeqAIJ(Mat B,PetscInt nz,const PetscInt *nnz)
3768: {
3769: Mat_SeqAIJ *b;
3770: PetscBool skipallocation = PETSC_FALSE,realalloc = PETSC_FALSE;
3772: PetscInt i;
3775: if (nz >= 0 || nnz) realalloc = PETSC_TRUE;
3776: if (nz == MAT_SKIP_ALLOCATION) {
3777: skipallocation = PETSC_TRUE;
3778: nz = 0;
3779: }
3780: PetscLayoutSetUp(B->rmap);
3781: PetscLayoutSetUp(B->cmap);
3783: if (nz == PETSC_DEFAULT || nz == PETSC_DECIDE) nz = 5;
3784: if (nz < 0) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"nz cannot be less than 0: value %D",nz);
3785: if (nnz) {
3786: for (i=0; i<B->rmap->n; i++) {
3787: 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]);
3788: 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);
3789: }
3790: }
3792: B->preallocated = PETSC_TRUE;
3794: b = (Mat_SeqAIJ*)B->data;
3796: if (!skipallocation) {
3797: if (!b->imax) {
3798: PetscMalloc2(B->rmap->n,&b->imax,B->rmap->n,&b->ilen);
3799: PetscLogObjectMemory((PetscObject)B,2*B->rmap->n*sizeof(PetscInt));
3800: }
3801: if (!b->ipre) {
3802: PetscMalloc1(B->rmap->n,&b->ipre);
3803: PetscLogObjectMemory((PetscObject)B,B->rmap->n*sizeof(PetscInt));
3804: }
3805: if (!nnz) {
3806: if (nz == PETSC_DEFAULT || nz == PETSC_DECIDE) nz = 10;
3807: else if (nz < 0) nz = 1;
3808: for (i=0; i<B->rmap->n; i++) b->imax[i] = nz;
3809: nz = nz*B->rmap->n;
3810: } else {
3811: nz = 0;
3812: for (i=0; i<B->rmap->n; i++) {b->imax[i] = nnz[i]; nz += nnz[i];}
3813: }
3814: /* b->ilen will count nonzeros in each row so far. */
3815: for (i=0; i<B->rmap->n; i++) b->ilen[i] = 0;
3817: /* allocate the matrix space */
3818: /* FIXME: should B's old memory be unlogged? */
3819: MatSeqXAIJFreeAIJ(B,&b->a,&b->j,&b->i);
3820: if (B->structure_only) {
3821: PetscMalloc1(nz,&b->j);
3822: PetscMalloc1(B->rmap->n+1,&b->i);
3823: PetscLogObjectMemory((PetscObject)B,(B->rmap->n+1)*sizeof(PetscInt)+nz*sizeof(PetscInt));
3824: } else {
3825: PetscMalloc3(nz,&b->a,nz,&b->j,B->rmap->n+1,&b->i);
3826: PetscLogObjectMemory((PetscObject)B,(B->rmap->n+1)*sizeof(PetscInt)+nz*(sizeof(PetscScalar)+sizeof(PetscInt)));
3827: }
3828: b->i[0] = 0;
3829: for (i=1; i<B->rmap->n+1; i++) {
3830: b->i[i] = b->i[i-1] + b->imax[i-1];
3831: }
3832: if (B->structure_only) {
3833: b->singlemalloc = PETSC_FALSE;
3834: b->free_a = PETSC_FALSE;
3835: } else {
3836: b->singlemalloc = PETSC_TRUE;
3837: b->free_a = PETSC_TRUE;
3838: }
3839: b->free_ij = PETSC_TRUE;
3840: } else {
3841: b->free_a = PETSC_FALSE;
3842: b->free_ij = PETSC_FALSE;
3843: }
3845: if (b->ipre && nnz != b->ipre && b->imax) {
3846: /* reserve user-requested sparsity */
3847: PetscMemcpy(b->ipre,b->imax,B->rmap->n*sizeof(PetscInt));
3848: }
3851: b->nz = 0;
3852: b->maxnz = nz;
3853: B->info.nz_unneeded = (double)b->maxnz;
3854: if (realalloc) {
3855: MatSetOption(B,MAT_NEW_NONZERO_ALLOCATION_ERR,PETSC_TRUE);
3856: }
3857: B->was_assembled = PETSC_FALSE;
3858: B->assembled = PETSC_FALSE;
3859: return(0);
3860: }
3863: PetscErrorCode MatResetPreallocation_SeqAIJ(Mat A)
3864: {
3865: Mat_SeqAIJ *a;
3866: PetscInt i;
3871: a = (Mat_SeqAIJ*)A->data;
3872: /* if no saved info, we error out */
3873: if (!a->ipre) SETERRQ(PETSC_COMM_WORLD,PETSC_ERR_ARG_NULL,"No saved preallocation info \n");
3875: if (!a->i || !a->j || !a->a || !a->imax || !a->ilen) SETERRQ(PETSC_COMM_WORLD,PETSC_ERR_ARG_NULL,"Memory info is incomplete, and can not reset preallocation \n");
3877: PetscMemcpy(a->imax,a->ipre,A->rmap->n*sizeof(PetscInt));
3878: PetscMemzero(a->ilen,A->rmap->n*sizeof(PetscInt));
3879: a->i[0] = 0;
3880: for (i=1; i<A->rmap->n+1; i++) {
3881: a->i[i] = a->i[i-1] + a->imax[i-1];
3882: }
3883: A->preallocated = PETSC_TRUE;
3884: a->nz = 0;
3885: a->maxnz = a->i[A->rmap->n];
3886: A->info.nz_unneeded = (double)a->maxnz;
3887: A->was_assembled = PETSC_FALSE;
3888: A->assembled = PETSC_FALSE;
3889: return(0);
3890: }
3892: /*@
3893: MatSeqAIJSetPreallocationCSR - Allocates memory for a sparse sequential matrix in AIJ format.
3895: Input Parameters:
3896: + B - the matrix
3897: . i - the indices into j for the start of each row (starts with zero)
3898: . j - the column indices for each row (starts with zero) these must be sorted for each row
3899: - v - optional values in the matrix
3901: Level: developer
3903: The i,j,v values are COPIED with this routine; to avoid the copy use MatCreateSeqAIJWithArrays()
3905: .keywords: matrix, aij, compressed row, sparse, sequential
3907: .seealso: MatCreate(), MatCreateSeqAIJ(), MatSetValues(), MatSeqAIJSetPreallocation(), MatCreateSeqAIJ(), MATSEQAIJ
3908: @*/
3909: PetscErrorCode MatSeqAIJSetPreallocationCSR(Mat B,const PetscInt i[],const PetscInt j[],const PetscScalar v[])
3910: {
3916: PetscTryMethod(B,"MatSeqAIJSetPreallocationCSR_C",(Mat,const PetscInt[],const PetscInt[],const PetscScalar[]),(B,i,j,v));
3917: return(0);
3918: }
3920: PetscErrorCode MatSeqAIJSetPreallocationCSR_SeqAIJ(Mat B,const PetscInt Ii[],const PetscInt J[],const PetscScalar v[])
3921: {
3922: PetscInt i;
3923: PetscInt m,n;
3924: PetscInt nz;
3925: PetscInt *nnz, nz_max = 0;
3926: PetscScalar *values;
3930: if (Ii[0]) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE, "Ii[0] must be 0 it is %D", Ii[0]);
3932: PetscLayoutSetUp(B->rmap);
3933: PetscLayoutSetUp(B->cmap);
3935: MatGetSize(B, &m, &n);
3936: PetscMalloc1(m+1, &nnz);
3937: for (i = 0; i < m; i++) {
3938: nz = Ii[i+1]- Ii[i];
3939: nz_max = PetscMax(nz_max, nz);
3940: if (nz < 0) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE, "Local row %D has a negative number of columns %D", i, nnz);
3941: nnz[i] = nz;
3942: }
3943: MatSeqAIJSetPreallocation(B, 0, nnz);
3944: PetscFree(nnz);
3946: if (v) {
3947: values = (PetscScalar*) v;
3948: } else {
3949: PetscCalloc1(nz_max, &values);
3950: }
3952: for (i = 0; i < m; i++) {
3953: nz = Ii[i+1] - Ii[i];
3954: MatSetValues_SeqAIJ(B, 1, &i, nz, J+Ii[i], values + (v ? Ii[i] : 0), INSERT_VALUES);
3955: }
3957: MatAssemblyBegin(B,MAT_FINAL_ASSEMBLY);
3958: MatAssemblyEnd(B,MAT_FINAL_ASSEMBLY);
3960: if (!v) {
3961: PetscFree(values);
3962: }
3963: MatSetOption(B,MAT_NEW_NONZERO_LOCATION_ERR,PETSC_TRUE);
3964: return(0);
3965: }
3967: #include <../src/mat/impls/dense/seq/dense.h>
3968: #include <petsc/private/kernels/petscaxpy.h>
3970: /*
3971: Computes (B'*A')' since computing B*A directly is untenable
3973: n p p
3974: ( ) ( ) ( )
3975: m ( A ) * n ( B ) = m ( C )
3976: ( ) ( ) ( )
3978: */
3979: PetscErrorCode MatMatMultNumeric_SeqDense_SeqAIJ(Mat A,Mat B,Mat C)
3980: {
3981: PetscErrorCode ierr;
3982: Mat_SeqDense *sub_a = (Mat_SeqDense*)A->data;
3983: Mat_SeqAIJ *sub_b = (Mat_SeqAIJ*)B->data;
3984: Mat_SeqDense *sub_c = (Mat_SeqDense*)C->data;
3985: PetscInt i,n,m,q,p;
3986: const PetscInt *ii,*idx;
3987: const PetscScalar *b,*a,*a_q;
3988: PetscScalar *c,*c_q;
3991: m = A->rmap->n;
3992: n = A->cmap->n;
3993: p = B->cmap->n;
3994: a = sub_a->v;
3995: b = sub_b->a;
3996: c = sub_c->v;
3997: PetscMemzero(c,m*p*sizeof(PetscScalar));
3999: ii = sub_b->i;
4000: idx = sub_b->j;
4001: for (i=0; i<n; i++) {
4002: q = ii[i+1] - ii[i];
4003: while (q-->0) {
4004: c_q = c + m*(*idx);
4005: a_q = a + m*i;
4006: PetscKernelAXPY(c_q,*b,a_q,m);
4007: idx++;
4008: b++;
4009: }
4010: }
4011: return(0);
4012: }
4014: PetscErrorCode MatMatMultSymbolic_SeqDense_SeqAIJ(Mat A,Mat B,PetscReal fill,Mat *C)
4015: {
4017: PetscInt m=A->rmap->n,n=B->cmap->n;
4018: Mat Cmat;
4021: 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);
4022: MatCreate(PetscObjectComm((PetscObject)A),&Cmat);
4023: MatSetSizes(Cmat,m,n,m,n);
4024: MatSetBlockSizesFromMats(Cmat,A,B);
4025: MatSetType(Cmat,MATSEQDENSE);
4026: MatSeqDenseSetPreallocation(Cmat,NULL);
4028: Cmat->ops->matmultnumeric = MatMatMultNumeric_SeqDense_SeqAIJ;
4030: *C = Cmat;
4031: return(0);
4032: }
4034: /* ----------------------------------------------------------------*/
4035: PETSC_INTERN PetscErrorCode MatMatMult_SeqDense_SeqAIJ(Mat A,Mat B,MatReuse scall,PetscReal fill,Mat *C)
4036: {
4040: if (scall == MAT_INITIAL_MATRIX) {
4041: PetscLogEventBegin(MAT_MatMultSymbolic,A,B,0,0);
4042: MatMatMultSymbolic_SeqDense_SeqAIJ(A,B,fill,C);
4043: PetscLogEventEnd(MAT_MatMultSymbolic,A,B,0,0);
4044: }
4045: PetscLogEventBegin(MAT_MatMultNumeric,A,B,0,0);
4046: MatMatMultNumeric_SeqDense_SeqAIJ(A,B,*C);
4047: PetscLogEventEnd(MAT_MatMultNumeric,A,B,0,0);
4048: return(0);
4049: }
4052: /*MC
4053: MATSEQAIJ - MATSEQAIJ = "seqaij" - A matrix type to be used for sequential sparse matrices,
4054: based on compressed sparse row format.
4056: Options Database Keys:
4057: . -mat_type seqaij - sets the matrix type to "seqaij" during a call to MatSetFromOptions()
4059: Level: beginner
4061: .seealso: MatCreateSeqAIJ(), MatSetFromOptions(), MatSetType(), MatCreate(), MatType
4062: M*/
4064: /*MC
4065: MATAIJ - MATAIJ = "aij" - A matrix type to be used for sparse matrices.
4067: This matrix type is identical to MATSEQAIJ when constructed with a single process communicator,
4068: and MATMPIAIJ otherwise. As a result, for single process communicators,
4069: MatSeqAIJSetPreallocation is supported, and similarly MatMPIAIJSetPreallocation is supported
4070: for communicators controlling multiple processes. It is recommended that you call both of
4071: the above preallocation routines for simplicity.
4073: Options Database Keys:
4074: . -mat_type aij - sets the matrix type to "aij" during a call to MatSetFromOptions()
4076: Developer Notes:
4077: Subclasses include MATAIJCUSPARSE, MATAIJPERM, MATAIJSELL, MATAIJMKL, MATAIJCRL, and also automatically switches over to use inodes when
4078: enough exist.
4080: Level: beginner
4082: .seealso: MatCreateAIJ(), MatCreateSeqAIJ(), MATSEQAIJ,MATMPIAIJ
4083: M*/
4085: /*MC
4086: MATAIJCRL - MATAIJCRL = "aijcrl" - A matrix type to be used for sparse matrices.
4088: This matrix type is identical to MATSEQAIJCRL when constructed with a single process communicator,
4089: and MATMPIAIJCRL otherwise. As a result, for single process communicators,
4090: MatSeqAIJSetPreallocation() is supported, and similarly MatMPIAIJSetPreallocation() is supported
4091: for communicators controlling multiple processes. It is recommended that you call both of
4092: the above preallocation routines for simplicity.
4094: Options Database Keys:
4095: . -mat_type aijcrl - sets the matrix type to "aijcrl" during a call to MatSetFromOptions()
4097: Level: beginner
4099: .seealso: MatCreateMPIAIJCRL,MATSEQAIJCRL,MATMPIAIJCRL, MATSEQAIJCRL, MATMPIAIJCRL
4100: M*/
4102: PETSC_INTERN PetscErrorCode MatConvert_SeqAIJ_SeqAIJCRL(Mat,MatType,MatReuse,Mat*);
4103: #if defined(PETSC_HAVE_ELEMENTAL)
4104: PETSC_INTERN PetscErrorCode MatConvert_SeqAIJ_Elemental(Mat,MatType,MatReuse,Mat*);
4105: #endif
4106: #if defined(PETSC_HAVE_HYPRE)
4107: PETSC_INTERN PetscErrorCode MatConvert_AIJ_HYPRE(Mat A,MatType,MatReuse,Mat*);
4108: PETSC_INTERN PetscErrorCode MatMatMatMult_Transpose_AIJ_AIJ(Mat,Mat,Mat,MatReuse,PetscReal,Mat*);
4109: #endif
4110: PETSC_INTERN PetscErrorCode MatConvert_SeqAIJ_SeqDense(Mat,MatType,MatReuse,Mat*);
4112: PETSC_EXTERN PetscErrorCode MatConvert_SeqAIJ_SeqSELL(Mat,MatType,MatReuse,Mat*);
4113: PETSC_INTERN PetscErrorCode MatConvert_XAIJ_IS(Mat,MatType,MatReuse,Mat*);
4114: PETSC_INTERN PetscErrorCode MatPtAP_IS_XAIJ(Mat,Mat,MatReuse,PetscReal,Mat*);
4116: /*@C
4117: MatSeqAIJGetArray - gives access to the array where the data for a MATSEQAIJ matrix is stored
4119: Not Collective
4121: Input Parameter:
4122: . mat - a MATSEQAIJ matrix
4124: Output Parameter:
4125: . array - pointer to the data
4127: Level: intermediate
4129: .seealso: MatSeqAIJRestoreArray(), MatSeqAIJGetArrayF90()
4130: @*/
4131: PetscErrorCode MatSeqAIJGetArray(Mat A,PetscScalar **array)
4132: {
4136: PetscUseMethod(A,"MatSeqAIJGetArray_C",(Mat,PetscScalar**),(A,array));
4137: return(0);
4138: }
4140: /*@C
4141: MatSeqAIJGetMaxRowNonzeros - returns the maximum number of nonzeros in any row
4143: Not Collective
4145: Input Parameter:
4146: . mat - a MATSEQAIJ matrix
4148: Output Parameter:
4149: . nz - the maximum number of nonzeros in any row
4151: Level: intermediate
4153: .seealso: MatSeqAIJRestoreArray(), MatSeqAIJGetArrayF90()
4154: @*/
4155: PetscErrorCode MatSeqAIJGetMaxRowNonzeros(Mat A,PetscInt *nz)
4156: {
4157: Mat_SeqAIJ *aij = (Mat_SeqAIJ*)A->data;
4160: *nz = aij->rmax;
4161: return(0);
4162: }
4164: /*@C
4165: MatSeqAIJRestoreArray - returns access to the array where the data for a MATSEQAIJ matrix is stored obtained by MatSeqAIJGetArray()
4167: Not Collective
4169: Input Parameters:
4170: . mat - a MATSEQAIJ matrix
4171: . array - pointer to the data
4173: Level: intermediate
4175: .seealso: MatSeqAIJGetArray(), MatSeqAIJRestoreArrayF90()
4176: @*/
4177: PetscErrorCode MatSeqAIJRestoreArray(Mat A,PetscScalar **array)
4178: {
4182: PetscUseMethod(A,"MatSeqAIJRestoreArray_C",(Mat,PetscScalar**),(A,array));
4183: return(0);
4184: }
4186: #if defined(PETSC_HAVE_CUDA)
4187: PETSC_EXTERN PetscErrorCode MatConvert_SeqAIJ_SeqAIJCUSPARSE(Mat);
4188: #endif
4190: PETSC_EXTERN PetscErrorCode MatCreate_SeqAIJ(Mat B)
4191: {
4192: Mat_SeqAIJ *b;
4194: PetscMPIInt size;
4197: MPI_Comm_size(PetscObjectComm((PetscObject)B),&size);
4198: if (size > 1) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Comm must be of size 1");
4200: PetscNewLog(B,&b);
4202: B->data = (void*)b;
4204: PetscMemcpy(B->ops,&MatOps_Values,sizeof(struct _MatOps));
4206: b->row = 0;
4207: b->col = 0;
4208: b->icol = 0;
4209: b->reallocs = 0;
4210: b->ignorezeroentries = PETSC_FALSE;
4211: b->roworiented = PETSC_TRUE;
4212: b->nonew = 0;
4213: b->diag = 0;
4214: b->solve_work = 0;
4215: B->spptr = 0;
4216: b->saved_values = 0;
4217: b->idiag = 0;
4218: b->mdiag = 0;
4219: b->ssor_work = 0;
4220: b->omega = 1.0;
4221: b->fshift = 0.0;
4222: b->idiagvalid = PETSC_FALSE;
4223: b->ibdiagvalid = PETSC_FALSE;
4224: b->keepnonzeropattern = PETSC_FALSE;
4226: PetscObjectChangeTypeName((PetscObject)B,MATSEQAIJ);
4227: PetscObjectComposeFunction((PetscObject)B,"MatSeqAIJGetArray_C",MatSeqAIJGetArray_SeqAIJ);
4228: PetscObjectComposeFunction((PetscObject)B,"MatSeqAIJRestoreArray_C",MatSeqAIJRestoreArray_SeqAIJ);
4230: #if defined(PETSC_HAVE_MATLAB_ENGINE)
4231: PetscObjectComposeFunction((PetscObject)B,"PetscMatlabEnginePut_C",MatlabEnginePut_SeqAIJ);
4232: PetscObjectComposeFunction((PetscObject)B,"PetscMatlabEngineGet_C",MatlabEngineGet_SeqAIJ);
4233: #endif
4235: PetscObjectComposeFunction((PetscObject)B,"MatSeqAIJSetColumnIndices_C",MatSeqAIJSetColumnIndices_SeqAIJ);
4236: PetscObjectComposeFunction((PetscObject)B,"MatStoreValues_C",MatStoreValues_SeqAIJ);
4237: PetscObjectComposeFunction((PetscObject)B,"MatRetrieveValues_C",MatRetrieveValues_SeqAIJ);
4238: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqaij_seqsbaij_C",MatConvert_SeqAIJ_SeqSBAIJ);
4239: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqaij_seqbaij_C",MatConvert_SeqAIJ_SeqBAIJ);
4240: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqaij_seqaijperm_C",MatConvert_SeqAIJ_SeqAIJPERM);
4241: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqaij_seqaijsell_C",MatConvert_SeqAIJ_SeqAIJSELL);
4242: #if defined(PETSC_HAVE_MKL_SPARSE)
4243: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqaij_seqaijmkl_C",MatConvert_SeqAIJ_SeqAIJMKL);
4244: #endif
4245: #if defined(PETSC_HAVE_CUDA)
4246: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqaij_seqaijcusparse_C",MatConvert_SeqAIJ_SeqAIJCUSPARSE);
4247: #endif
4248: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqaij_seqaijcrl_C",MatConvert_SeqAIJ_SeqAIJCRL);
4249: #if defined(PETSC_HAVE_ELEMENTAL)
4250: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqaij_elemental_C",MatConvert_SeqAIJ_Elemental);
4251: #endif
4252: #if defined(PETSC_HAVE_HYPRE)
4253: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqaij_hypre_C",MatConvert_AIJ_HYPRE);
4254: PetscObjectComposeFunction((PetscObject)B,"MatMatMatMult_transpose_seqaij_seqaij_C",MatMatMatMult_Transpose_AIJ_AIJ);
4255: #endif
4256: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqaij_seqdense_C",MatConvert_SeqAIJ_SeqDense);
4257: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqaij_seqsell_C",MatConvert_SeqAIJ_SeqSELL);
4258: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqaij_is_C",MatConvert_XAIJ_IS);
4259: PetscObjectComposeFunction((PetscObject)B,"MatIsTranspose_C",MatIsTranspose_SeqAIJ);
4260: PetscObjectComposeFunction((PetscObject)B,"MatIsHermitianTranspose_C",MatIsTranspose_SeqAIJ);
4261: PetscObjectComposeFunction((PetscObject)B,"MatSeqAIJSetPreallocation_C",MatSeqAIJSetPreallocation_SeqAIJ);
4262: PetscObjectComposeFunction((PetscObject)B,"MatResetPreallocation_C",MatResetPreallocation_SeqAIJ);
4263: PetscObjectComposeFunction((PetscObject)B,"MatSeqAIJSetPreallocationCSR_C",MatSeqAIJSetPreallocationCSR_SeqAIJ);
4264: PetscObjectComposeFunction((PetscObject)B,"MatReorderForNonzeroDiagonal_C",MatReorderForNonzeroDiagonal_SeqAIJ);
4265: PetscObjectComposeFunction((PetscObject)B,"MatMatMult_seqdense_seqaij_C",MatMatMult_SeqDense_SeqAIJ);
4266: PetscObjectComposeFunction((PetscObject)B,"MatMatMultSymbolic_seqdense_seqaij_C",MatMatMultSymbolic_SeqDense_SeqAIJ);
4267: PetscObjectComposeFunction((PetscObject)B,"MatMatMultNumeric_seqdense_seqaij_C",MatMatMultNumeric_SeqDense_SeqAIJ);
4268: PetscObjectComposeFunction((PetscObject)B,"MatPtAP_is_seqaij_C",MatPtAP_IS_XAIJ);
4269: MatCreate_SeqAIJ_Inode(B);
4270: PetscObjectChangeTypeName((PetscObject)B,MATSEQAIJ);
4271: MatSeqAIJSetTypeFromOptions(B); /* this allows changing the matrix subtype to say MATSEQAIJPERM */
4272: return(0);
4273: }
4275: /*
4276: Given a matrix generated with MatGetFactor() duplicates all the information in A into B
4277: */
4278: PetscErrorCode MatDuplicateNoCreate_SeqAIJ(Mat C,Mat A,MatDuplicateOption cpvalues,PetscBool mallocmatspace)
4279: {
4280: Mat_SeqAIJ *c,*a = (Mat_SeqAIJ*)A->data;
4282: PetscInt i,m = A->rmap->n;
4285: c = (Mat_SeqAIJ*)C->data;
4287: C->factortype = A->factortype;
4288: c->row = 0;
4289: c->col = 0;
4290: c->icol = 0;
4291: c->reallocs = 0;
4293: C->assembled = PETSC_TRUE;
4295: PetscLayoutReference(A->rmap,&C->rmap);
4296: PetscLayoutReference(A->cmap,&C->cmap);
4298: PetscMalloc2(m,&c->imax,m,&c->ilen);
4299: PetscLogObjectMemory((PetscObject)C, 2*m*sizeof(PetscInt));
4300: for (i=0; i<m; i++) {
4301: c->imax[i] = a->imax[i];
4302: c->ilen[i] = a->ilen[i];
4303: }
4305: /* allocate the matrix space */
4306: if (mallocmatspace) {
4307: PetscMalloc3(a->i[m],&c->a,a->i[m],&c->j,m+1,&c->i);
4308: PetscLogObjectMemory((PetscObject)C, a->i[m]*(sizeof(PetscScalar)+sizeof(PetscInt))+(m+1)*sizeof(PetscInt));
4310: c->singlemalloc = PETSC_TRUE;
4312: PetscMemcpy(c->i,a->i,(m+1)*sizeof(PetscInt));
4313: if (m > 0) {
4314: PetscMemcpy(c->j,a->j,(a->i[m])*sizeof(PetscInt));
4315: if (cpvalues == MAT_COPY_VALUES) {
4316: PetscMemcpy(c->a,a->a,(a->i[m])*sizeof(PetscScalar));
4317: } else {
4318: PetscMemzero(c->a,(a->i[m])*sizeof(PetscScalar));
4319: }
4320: }
4321: }
4323: c->ignorezeroentries = a->ignorezeroentries;
4324: c->roworiented = a->roworiented;
4325: c->nonew = a->nonew;
4326: if (a->diag) {
4327: PetscMalloc1(m+1,&c->diag);
4328: PetscLogObjectMemory((PetscObject)C,(m+1)*sizeof(PetscInt));
4329: for (i=0; i<m; i++) {
4330: c->diag[i] = a->diag[i];
4331: }
4332: } else c->diag = 0;
4334: c->solve_work = 0;
4335: c->saved_values = 0;
4336: c->idiag = 0;
4337: c->ssor_work = 0;
4338: c->keepnonzeropattern = a->keepnonzeropattern;
4339: c->free_a = PETSC_TRUE;
4340: c->free_ij = PETSC_TRUE;
4342: c->rmax = a->rmax;
4343: c->nz = a->nz;
4344: c->maxnz = a->nz; /* Since we allocate exactly the right amount */
4345: C->preallocated = PETSC_TRUE;
4347: c->compressedrow.use = a->compressedrow.use;
4348: c->compressedrow.nrows = a->compressedrow.nrows;
4349: if (a->compressedrow.use) {
4350: i = a->compressedrow.nrows;
4351: PetscMalloc2(i+1,&c->compressedrow.i,i,&c->compressedrow.rindex);
4352: PetscMemcpy(c->compressedrow.i,a->compressedrow.i,(i+1)*sizeof(PetscInt));
4353: PetscMemcpy(c->compressedrow.rindex,a->compressedrow.rindex,i*sizeof(PetscInt));
4354: } else {
4355: c->compressedrow.use = PETSC_FALSE;
4356: c->compressedrow.i = NULL;
4357: c->compressedrow.rindex = NULL;
4358: }
4359: c->nonzerorowcnt = a->nonzerorowcnt;
4360: C->nonzerostate = A->nonzerostate;
4362: MatDuplicate_SeqAIJ_Inode(A,cpvalues,&C);
4363: PetscFunctionListDuplicate(((PetscObject)A)->qlist,&((PetscObject)C)->qlist);
4364: return(0);
4365: }
4367: PetscErrorCode MatDuplicate_SeqAIJ(Mat A,MatDuplicateOption cpvalues,Mat *B)
4368: {
4372: MatCreate(PetscObjectComm((PetscObject)A),B);
4373: MatSetSizes(*B,A->rmap->n,A->cmap->n,A->rmap->n,A->cmap->n);
4374: if (!(A->rmap->n % A->rmap->bs) && !(A->cmap->n % A->cmap->bs)) {
4375: MatSetBlockSizesFromMats(*B,A,A);
4376: }
4377: MatSetType(*B,((PetscObject)A)->type_name);
4378: MatDuplicateNoCreate_SeqAIJ(*B,A,cpvalues,PETSC_TRUE);
4379: return(0);
4380: }
4382: PetscErrorCode MatLoad_SeqAIJ(Mat newMat, PetscViewer viewer)
4383: {
4384: PetscBool isbinary, ishdf5;
4390: /* force binary viewer to load .info file if it has not yet done so */
4391: PetscViewerSetUp(viewer);
4392: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERBINARY,&isbinary);
4393: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERHDF5, &ishdf5);
4394: if (isbinary) {
4395: MatLoad_SeqAIJ_Binary(newMat,viewer);
4396: } else if (ishdf5) {
4397: #if defined(PETSC_HAVE_HDF5)
4398: MatLoad_AIJ_HDF5(newMat,viewer);
4399: #else
4400: SETERRQ(PetscObjectComm((PetscObject)newMat),PETSC_ERR_SUP,"HDF5 not supported in this build.\nPlease reconfigure using --download-hdf5");
4401: #endif
4402: } else {
4403: 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);
4404: }
4405: return(0);
4406: }
4408: PetscErrorCode MatLoad_SeqAIJ_Binary(Mat newMat, PetscViewer viewer)
4409: {
4410: Mat_SeqAIJ *a;
4412: PetscInt i,sum,nz,header[4],*rowlengths = 0,M,N,rows,cols;
4413: int fd;
4414: PetscMPIInt size;
4415: MPI_Comm comm;
4416: PetscInt bs = newMat->rmap->bs;
4419: PetscObjectGetComm((PetscObject)viewer,&comm);
4420: MPI_Comm_size(comm,&size);
4421: if (size > 1) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"view must have one processor");
4423: PetscOptionsBegin(comm,NULL,"Options for loading SEQAIJ matrix","Mat");
4424: PetscOptionsInt("-matload_block_size","Set the blocksize used to store the matrix","MatLoad",bs,&bs,NULL);
4425: PetscOptionsEnd();
4426: if (bs < 0) bs = 1;
4427: MatSetBlockSize(newMat,bs);
4429: PetscViewerBinaryGetDescriptor(viewer,&fd);
4430: PetscBinaryRead(fd,header,4,PETSC_INT);
4431: if (header[0] != MAT_FILE_CLASSID) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_FILE_UNEXPECTED,"not matrix object in file");
4432: M = header[1]; N = header[2]; nz = header[3];
4434: if (nz < 0) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_FILE_UNEXPECTED,"Matrix stored in special format on disk,cannot load as SeqAIJ");
4436: /* read in row lengths */
4437: PetscMalloc1(M,&rowlengths);
4438: PetscBinaryRead(fd,rowlengths,M,PETSC_INT);
4440: /* check if sum of rowlengths is same as nz */
4441: for (i=0,sum=0; i< M; i++) sum +=rowlengths[i];
4442: if (sum != nz) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_FILE_READ,"Inconsistant matrix data in file. no-nonzeros = %dD, sum-row-lengths = %D\n",nz,sum);
4444: /* set global size if not set already*/
4445: if (newMat->rmap->n < 0 && newMat->rmap->N < 0 && newMat->cmap->n < 0 && newMat->cmap->N < 0) {
4446: MatSetSizes(newMat,PETSC_DECIDE,PETSC_DECIDE,M,N);
4447: } else {
4448: /* if sizes and type are already set, check if the matrix global sizes are correct */
4449: MatGetSize(newMat,&rows,&cols);
4450: if (rows < 0 && cols < 0) { /* user might provide local size instead of global size */
4451: MatGetLocalSize(newMat,&rows,&cols);
4452: }
4453: 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);
4454: }
4455: MatSeqAIJSetPreallocation_SeqAIJ(newMat,0,rowlengths);
4456: a = (Mat_SeqAIJ*)newMat->data;
4458: PetscBinaryRead(fd,a->j,nz,PETSC_INT);
4460: /* read in nonzero values */
4461: PetscBinaryRead(fd,a->a,nz,PETSC_SCALAR);
4463: /* set matrix "i" values */
4464: a->i[0] = 0;
4465: for (i=1; i<= M; i++) {
4466: a->i[i] = a->i[i-1] + rowlengths[i-1];
4467: a->ilen[i-1] = rowlengths[i-1];
4468: }
4469: PetscFree(rowlengths);
4471: MatAssemblyBegin(newMat,MAT_FINAL_ASSEMBLY);
4472: MatAssemblyEnd(newMat,MAT_FINAL_ASSEMBLY);
4473: return(0);
4474: }
4476: PetscErrorCode MatEqual_SeqAIJ(Mat A,Mat B,PetscBool * flg)
4477: {
4478: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data,*b = (Mat_SeqAIJ*)B->data;
4480: #if defined(PETSC_USE_COMPLEX)
4481: PetscInt k;
4482: #endif
4485: /* If the matrix dimensions are not equal,or no of nonzeros */
4486: if ((A->rmap->n != B->rmap->n) || (A->cmap->n != B->cmap->n) ||(a->nz != b->nz)) {
4487: *flg = PETSC_FALSE;
4488: return(0);
4489: }
4491: /* if the a->i are the same */
4492: PetscMemcmp(a->i,b->i,(A->rmap->n+1)*sizeof(PetscInt),flg);
4493: if (!*flg) return(0);
4495: /* if a->j are the same */
4496: PetscMemcmp(a->j,b->j,(a->nz)*sizeof(PetscInt),flg);
4497: if (!*flg) return(0);
4499: /* if a->a are the same */
4500: #if defined(PETSC_USE_COMPLEX)
4501: for (k=0; k<a->nz; k++) {
4502: if (PetscRealPart(a->a[k]) != PetscRealPart(b->a[k]) || PetscImaginaryPart(a->a[k]) != PetscImaginaryPart(b->a[k])) {
4503: *flg = PETSC_FALSE;
4504: return(0);
4505: }
4506: }
4507: #else
4508: PetscMemcmp(a->a,b->a,(a->nz)*sizeof(PetscScalar),flg);
4509: #endif
4510: return(0);
4511: }
4513: /*@
4514: MatCreateSeqAIJWithArrays - Creates an sequential AIJ matrix using matrix elements (in CSR format)
4515: provided by the user.
4517: Collective on MPI_Comm
4519: Input Parameters:
4520: + comm - must be an MPI communicator of size 1
4521: . m - number of rows
4522: . n - number of columns
4523: . i - row indices; that is i[0] = 0, i[row] = i[row-1] + number of elements in that row of the matrix
4524: . j - column indices
4525: - a - matrix values
4527: Output Parameter:
4528: . mat - the matrix
4530: Level: intermediate
4532: Notes:
4533: The i, j, and a arrays are not copied by this routine, the user must free these arrays
4534: once the matrix is destroyed and not before
4536: You cannot set new nonzero locations into this matrix, that will generate an error.
4538: The i and j indices are 0 based
4540: The format which is used for the sparse matrix input, is equivalent to a
4541: row-major ordering.. i.e for the following matrix, the input data expected is
4542: as shown
4544: $ 1 0 0
4545: $ 2 0 3
4546: $ 4 5 6
4547: $
4548: $ i = {0,1,3,6} [size = nrow+1 = 3+1]
4549: $ j = {0,0,2,0,1,2} [size = 6]; values must be sorted for each row
4550: $ v = {1,2,3,4,5,6} [size = 6]
4553: .seealso: MatCreate(), MatCreateAIJ(), MatCreateSeqAIJ(), MatCreateMPIAIJWithArrays(), MatMPIAIJSetPreallocationCSR()
4555: @*/
4556: PetscErrorCode MatCreateSeqAIJWithArrays(MPI_Comm comm,PetscInt m,PetscInt n,PetscInt i[],PetscInt j[],PetscScalar a[],Mat *mat)
4557: {
4559: PetscInt ii;
4560: Mat_SeqAIJ *aij;
4561: #if defined(PETSC_USE_DEBUG)
4562: PetscInt jj;
4563: #endif
4566: if (m > 0 && i[0]) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"i (row indices) must start with 0");
4567: MatCreate(comm,mat);
4568: MatSetSizes(*mat,m,n,m,n);
4569: /* MatSetBlockSizes(*mat,,); */
4570: MatSetType(*mat,MATSEQAIJ);
4571: MatSeqAIJSetPreallocation_SeqAIJ(*mat,MAT_SKIP_ALLOCATION,0);
4572: aij = (Mat_SeqAIJ*)(*mat)->data;
4573: PetscMalloc2(m,&aij->imax,m,&aij->ilen);
4575: aij->i = i;
4576: aij->j = j;
4577: aij->a = a;
4578: aij->singlemalloc = PETSC_FALSE;
4579: aij->nonew = -1; /*this indicates that inserting a new value in the matrix that generates a new nonzero is an error*/
4580: aij->free_a = PETSC_FALSE;
4581: aij->free_ij = PETSC_FALSE;
4583: for (ii=0; ii<m; ii++) {
4584: aij->ilen[ii] = aij->imax[ii] = i[ii+1] - i[ii];
4585: #if defined(PETSC_USE_DEBUG)
4586: 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]);
4587: for (jj=i[ii]+1; jj<i[ii+1]; jj++) {
4588: 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);
4589: 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);
4590: }
4591: #endif
4592: }
4593: #if defined(PETSC_USE_DEBUG)
4594: for (ii=0; ii<aij->i[m]; ii++) {
4595: if (j[ii] < 0) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Negative column index at location = %D index = %D",ii,j[ii]);
4596: 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]);
4597: }
4598: #endif
4600: MatAssemblyBegin(*mat,MAT_FINAL_ASSEMBLY);
4601: MatAssemblyEnd(*mat,MAT_FINAL_ASSEMBLY);
4602: return(0);
4603: }
4604: /*@C
4605: MatCreateSeqAIJFromTriple - Creates an sequential AIJ matrix using matrix elements (in COO format)
4606: provided by the user.
4608: Collective on MPI_Comm
4610: Input Parameters:
4611: + comm - must be an MPI communicator of size 1
4612: . m - number of rows
4613: . n - number of columns
4614: . i - row indices
4615: . j - column indices
4616: . a - matrix values
4617: . nz - number of nonzeros
4618: - idx - 0 or 1 based
4620: Output Parameter:
4621: . mat - the matrix
4623: Level: intermediate
4625: Notes:
4626: The i and j indices are 0 based
4628: The format which is used for the sparse matrix input, is equivalent to a
4629: row-major ordering.. i.e for the following matrix, the input data expected is
4630: as shown:
4632: 1 0 0
4633: 2 0 3
4634: 4 5 6
4636: i = {0,1,1,2,2,2}
4637: j = {0,0,2,0,1,2}
4638: v = {1,2,3,4,5,6}
4641: .seealso: MatCreate(), MatCreateAIJ(), MatCreateSeqAIJ(), MatCreateSeqAIJWithArrays(), MatMPIAIJSetPreallocationCSR()
4643: @*/
4644: PetscErrorCode MatCreateSeqAIJFromTriple(MPI_Comm comm,PetscInt m,PetscInt n,PetscInt i[],PetscInt j[],PetscScalar a[],Mat *mat,PetscInt nz,PetscBool idx)
4645: {
4647: PetscInt ii, *nnz, one = 1,row,col;
4651: PetscCalloc1(m,&nnz);
4652: for (ii = 0; ii < nz; ii++) {
4653: nnz[i[ii] - !!idx] += 1;
4654: }
4655: MatCreate(comm,mat);
4656: MatSetSizes(*mat,m,n,m,n);
4657: MatSetType(*mat,MATSEQAIJ);
4658: MatSeqAIJSetPreallocation_SeqAIJ(*mat,0,nnz);
4659: for (ii = 0; ii < nz; ii++) {
4660: if (idx) {
4661: row = i[ii] - 1;
4662: col = j[ii] - 1;
4663: } else {
4664: row = i[ii];
4665: col = j[ii];
4666: }
4667: MatSetValues(*mat,one,&row,one,&col,&a[ii],ADD_VALUES);
4668: }
4669: MatAssemblyBegin(*mat,MAT_FINAL_ASSEMBLY);
4670: MatAssemblyEnd(*mat,MAT_FINAL_ASSEMBLY);
4671: PetscFree(nnz);
4672: return(0);
4673: }
4675: PetscErrorCode MatSeqAIJInvalidateDiagonal(Mat A)
4676: {
4677: Mat_SeqAIJ *a=(Mat_SeqAIJ*)A->data;
4681: a->idiagvalid = PETSC_FALSE;
4682: a->ibdiagvalid = PETSC_FALSE;
4684: MatSeqAIJInvalidateDiagonal_Inode(A);
4685: return(0);
4686: }
4688: PetscErrorCode MatCreateMPIMatConcatenateSeqMat_SeqAIJ(MPI_Comm comm,Mat inmat,PetscInt n,MatReuse scall,Mat *outmat)
4689: {
4691: PetscMPIInt size;
4694: MPI_Comm_size(comm,&size);
4695: if (size == 1) {
4696: if (scall == MAT_INITIAL_MATRIX) {
4697: MatDuplicate(inmat,MAT_COPY_VALUES,outmat);
4698: } else {
4699: MatCopy(inmat,*outmat,SAME_NONZERO_PATTERN);
4700: }
4701: } else {
4702: MatCreateMPIMatConcatenateSeqMat_MPIAIJ(comm,inmat,n,scall,outmat);
4703: }
4704: return(0);
4705: }
4707: /*
4708: Permute A into C's *local* index space using rowemb,colemb.
4709: The embedding are supposed to be injections and the above implies that the range of rowemb is a subset
4710: of [0,m), colemb is in [0,n).
4711: If pattern == DIFFERENT_NONZERO_PATTERN, C is preallocated according to A.
4712: */
4713: PetscErrorCode MatSetSeqMat_SeqAIJ(Mat C,IS rowemb,IS colemb,MatStructure pattern,Mat B)
4714: {
4715: /* If making this function public, change the error returned in this function away from _PLIB. */
4717: Mat_SeqAIJ *Baij;
4718: PetscBool seqaij;
4719: PetscInt m,n,*nz,i,j,count;
4720: PetscScalar v;
4721: const PetscInt *rowindices,*colindices;
4724: if (!B) return(0);
4725: /* Check to make sure the target matrix (and embeddings) are compatible with C and each other. */
4726: PetscObjectBaseTypeCompare((PetscObject)B,MATSEQAIJ,&seqaij);
4727: if (!seqaij) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_PLIB,"Input matrix is of wrong type");
4728: if (rowemb) {
4729: ISGetLocalSize(rowemb,&m);
4730: 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);
4731: } else {
4732: if (C->rmap->n != B->rmap->n) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_PLIB,"Input matrix is row-incompatible with the target matrix");
4733: }
4734: if (colemb) {
4735: ISGetLocalSize(colemb,&n);
4736: 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);
4737: } else {
4738: if (C->cmap->n != B->cmap->n) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_PLIB,"Input matrix is col-incompatible with the target matrix");
4739: }
4741: Baij = (Mat_SeqAIJ*)(B->data);
4742: if (pattern == DIFFERENT_NONZERO_PATTERN) {
4743: PetscMalloc1(B->rmap->n,&nz);
4744: for (i=0; i<B->rmap->n; i++) {
4745: nz[i] = Baij->i[i+1] - Baij->i[i];
4746: }
4747: MatSeqAIJSetPreallocation(C,0,nz);
4748: PetscFree(nz);
4749: }
4750: if (pattern == SUBSET_NONZERO_PATTERN) {
4751: MatZeroEntries(C);
4752: }
4753: count = 0;
4754: rowindices = NULL;
4755: colindices = NULL;
4756: if (rowemb) {
4757: ISGetIndices(rowemb,&rowindices);
4758: }
4759: if (colemb) {
4760: ISGetIndices(colemb,&colindices);
4761: }
4762: for (i=0; i<B->rmap->n; i++) {
4763: PetscInt row;
4764: row = i;
4765: if (rowindices) row = rowindices[i];
4766: for (j=Baij->i[i]; j<Baij->i[i+1]; j++) {
4767: PetscInt col;
4768: col = Baij->j[count];
4769: if (colindices) col = colindices[col];
4770: v = Baij->a[count];
4771: MatSetValues(C,1,&row,1,&col,&v,INSERT_VALUES);
4772: ++count;
4773: }
4774: }
4775: /* FIXME: set C's nonzerostate correctly. */
4776: /* Assembly for C is necessary. */
4777: C->preallocated = PETSC_TRUE;
4778: C->assembled = PETSC_TRUE;
4779: C->was_assembled = PETSC_FALSE;
4780: return(0);
4781: }
4783: PetscFunctionList MatSeqAIJList = NULL;
4785: /*@C
4786: MatSeqAIJSetType - Converts a MATSEQAIJ matrix to a subtype
4788: Collective on Mat
4790: Input Parameters:
4791: + mat - the matrix object
4792: - matype - matrix type
4794: Options Database Key:
4795: . -mat_seqai_type <method> - for example seqaijcrl
4798: Level: intermediate
4800: .keywords: Mat, MatType, set, method
4802: .seealso: PCSetType(), VecSetType(), MatCreate(), MatType, Mat
4803: @*/
4804: PetscErrorCode MatSeqAIJSetType(Mat mat, MatType matype)
4805: {
4806: PetscErrorCode ierr,(*r)(Mat,MatType,MatReuse,Mat*);
4807: PetscBool sametype;
4811: PetscObjectTypeCompare((PetscObject)mat,matype,&sametype);
4812: if (sametype) return(0);
4814: PetscFunctionListFind(MatSeqAIJList,matype,&r);
4815: if (!r) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_UNKNOWN_TYPE,"Unknown Mat type given: %s",matype);
4816: (*r)(mat,matype,MAT_INPLACE_MATRIX,&mat);
4817: return(0);
4818: }
4821: /*@C
4822: MatSeqAIJRegister - - Adds a new sub-matrix type for sequential AIJ matrices
4824: Not Collective
4826: Input Parameters:
4827: + name - name of a new user-defined matrix type, for example MATSEQAIJCRL
4828: - function - routine to convert to subtype
4830: Notes:
4831: MatSeqAIJRegister() may be called multiple times to add several user-defined solvers.
4834: Then, your matrix can be chosen with the procedural interface at runtime via the option
4835: $ -mat_seqaij_type my_mat
4837: Level: advanced
4839: .keywords: Mat, register
4841: .seealso: MatSeqAIJRegisterAll()
4844: Level: advanced
4845: @*/
4846: PetscErrorCode MatSeqAIJRegister(const char sname[],PetscErrorCode (*function)(Mat,MatType,MatReuse,Mat *))
4847: {
4851: MatInitializePackage();
4852: PetscFunctionListAdd(&MatSeqAIJList,sname,function);
4853: return(0);
4854: }
4856: PetscBool MatSeqAIJRegisterAllCalled = PETSC_FALSE;
4858: /*@C
4859: MatSeqAIJRegisterAll - Registers all of the matrix subtypes of SeqAIJ
4861: Not Collective
4863: Level: advanced
4865: Developers Note: CUSP and CUSPARSE do not yet support the MatConvert_SeqAIJ..() paradigm and thus cannot be registered here
4867: .keywords: KSP, register, all
4869: .seealso: MatRegisterAll(), MatSeqAIJRegister()
4870: @*/
4871: PetscErrorCode MatSeqAIJRegisterAll(void)
4872: {
4876: if (MatSeqAIJRegisterAllCalled) return(0);
4877: MatSeqAIJRegisterAllCalled = PETSC_TRUE;
4879: MatSeqAIJRegister(MATSEQAIJCRL, MatConvert_SeqAIJ_SeqAIJCRL);
4880: MatSeqAIJRegister(MATSEQAIJPERM, MatConvert_SeqAIJ_SeqAIJPERM);
4881: MatSeqAIJRegister(MATSEQAIJSELL, MatConvert_SeqAIJ_SeqAIJSELL);
4882: #if defined(PETSC_HAVE_MKL_SPARSE)
4883: MatSeqAIJRegister(MATSEQAIJMKL, MatConvert_SeqAIJ_SeqAIJMKL);
4884: #endif
4885: #if defined(PETSC_HAVE_VIENNACL) && defined(PETSC_HAVE_VIENNACL_NO_CUDA)
4886: MatSeqAIJRegister(MATMPIAIJVIENNACL, MatConvert_SeqAIJ_SeqAIJViennaCL);
4887: #endif
4888: return(0);
4889: }
4891: /*
4892: Special version for direct calls from Fortran
4893: */
4894: #include <petsc/private/fortranimpl.h>
4895: #if defined(PETSC_HAVE_FORTRAN_CAPS)
4896: #define matsetvaluesseqaij_ MATSETVALUESSEQAIJ
4897: #elif !defined(PETSC_HAVE_FORTRAN_UNDERSCORE)
4898: #define matsetvaluesseqaij_ matsetvaluesseqaij
4899: #endif
4901: /* Change these macros so can be used in void function */
4902: #undef CHKERRQ
4903: #define CHKERRQ(ierr) CHKERRABORT(PetscObjectComm((PetscObject)A),ierr)
4904: #undef SETERRQ2
4905: #define SETERRQ2(comm,ierr,b,c,d) CHKERRABORT(comm,ierr)
4906: #undef SETERRQ3
4907: #define SETERRQ3(comm,ierr,b,c,d,e) CHKERRABORT(comm,ierr)
4909: PETSC_EXTERN void PETSC_STDCALL matsetvaluesseqaij_(Mat *AA,PetscInt *mm,const PetscInt im[],PetscInt *nn,const PetscInt in[],const PetscScalar v[],InsertMode *isis, PetscErrorCode *_ierr)
4910: {
4911: Mat A = *AA;
4912: PetscInt m = *mm, n = *nn;
4913: InsertMode is = *isis;
4914: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
4915: PetscInt *rp,k,low,high,t,ii,row,nrow,i,col,l,rmax,N;
4916: PetscInt *imax,*ai,*ailen;
4918: PetscInt *aj,nonew = a->nonew,lastcol = -1;
4919: MatScalar *ap,value,*aa;
4920: PetscBool ignorezeroentries = a->ignorezeroentries;
4921: PetscBool roworiented = a->roworiented;
4924: MatCheckPreallocated(A,1);
4925: imax = a->imax;
4926: ai = a->i;
4927: ailen = a->ilen;
4928: aj = a->j;
4929: aa = a->a;
4931: for (k=0; k<m; k++) { /* loop over added rows */
4932: row = im[k];
4933: if (row < 0) continue;
4934: #if defined(PETSC_USE_DEBUG)
4935: if (row >= A->rmap->n) SETERRABORT(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_OUTOFRANGE,"Row too large");
4936: #endif
4937: rp = aj + ai[row]; ap = aa + ai[row];
4938: rmax = imax[row]; nrow = ailen[row];
4939: low = 0;
4940: high = nrow;
4941: for (l=0; l<n; l++) { /* loop over added columns */
4942: if (in[l] < 0) continue;
4943: #if defined(PETSC_USE_DEBUG)
4944: if (in[l] >= A->cmap->n) SETERRABORT(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_OUTOFRANGE,"Column too large");
4945: #endif
4946: col = in[l];
4947: if (roworiented) value = v[l + k*n];
4948: else value = v[k + l*m];
4950: if (value == 0.0 && ignorezeroentries && (is == ADD_VALUES)) continue;
4952: if (col <= lastcol) low = 0;
4953: else high = nrow;
4954: lastcol = col;
4955: while (high-low > 5) {
4956: t = (low+high)/2;
4957: if (rp[t] > col) high = t;
4958: else low = t;
4959: }
4960: for (i=low; i<high; i++) {
4961: if (rp[i] > col) break;
4962: if (rp[i] == col) {
4963: if (is == ADD_VALUES) ap[i] += value;
4964: else ap[i] = value;
4965: goto noinsert;
4966: }
4967: }
4968: if (value == 0.0 && ignorezeroentries) goto noinsert;
4969: if (nonew == 1) goto noinsert;
4970: if (nonew == -1) SETERRABORT(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_OUTOFRANGE,"Inserting a new nonzero in the matrix");
4971: MatSeqXAIJReallocateAIJ(A,A->rmap->n,1,nrow,row,col,rmax,aa,ai,aj,rp,ap,imax,nonew,MatScalar);
4972: N = nrow++ - 1; a->nz++; high++;
4973: /* shift up all the later entries in this row */
4974: for (ii=N; ii>=i; ii--) {
4975: rp[ii+1] = rp[ii];
4976: ap[ii+1] = ap[ii];
4977: }
4978: rp[i] = col;
4979: ap[i] = value;
4980: A->nonzerostate++;
4981: noinsert:;
4982: low = i + 1;
4983: }
4984: ailen[row] = nrow;
4985: }
4986: PetscFunctionReturnVoid();
4987: }