Actual source code: aij.c
petsc-master 2020-07-04
1: /*
2: Defines the basic matrix operations for the AIJ (compressed row)
3: matrix storage format.
4: */
7: #include <../src/mat/impls/aij/seq/aij.h>
8: #include <petscblaslapack.h>
9: #include <petscbt.h>
10: #include <petsc/private/kernels/blocktranspose.h>
12: PetscErrorCode MatSeqAIJSetTypeFromOptions(Mat A)
13: {
14: PetscErrorCode ierr;
15: PetscBool flg;
16: char type[256];
19: PetscObjectOptionsBegin((PetscObject)A);
20: PetscOptionsFList("-mat_seqaij_type","Matrix SeqAIJ type","MatSeqAIJSetType",MatSeqAIJList,"seqaij",type,256,&flg);
21: if (flg) {
22: MatSeqAIJSetType(A,type);
23: }
24: PetscOptionsEnd();
25: return(0);
26: }
28: PetscErrorCode MatGetColumnNorms_SeqAIJ(Mat A,NormType type,PetscReal *norms)
29: {
31: PetscInt i,m,n;
32: Mat_SeqAIJ *aij = (Mat_SeqAIJ*)A->data;
35: MatGetSize(A,&m,&n);
36: PetscArrayzero(norms,n);
37: if (type == NORM_2) {
38: for (i=0; i<aij->i[m]; i++) {
39: norms[aij->j[i]] += PetscAbsScalar(aij->a[i]*aij->a[i]);
40: }
41: } else if (type == NORM_1) {
42: for (i=0; i<aij->i[m]; i++) {
43: norms[aij->j[i]] += PetscAbsScalar(aij->a[i]);
44: }
45: } else if (type == NORM_INFINITY) {
46: for (i=0; i<aij->i[m]; i++) {
47: norms[aij->j[i]] = PetscMax(PetscAbsScalar(aij->a[i]),norms[aij->j[i]]);
48: }
49: } else SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"Unknown NormType");
51: if (type == NORM_2) {
52: for (i=0; i<n; i++) norms[i] = PetscSqrtReal(norms[i]);
53: }
54: return(0);
55: }
57: PetscErrorCode MatFindOffBlockDiagonalEntries_SeqAIJ(Mat A,IS *is)
58: {
59: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
60: PetscInt i,m=A->rmap->n,cnt = 0, bs = A->rmap->bs;
61: const PetscInt *jj = a->j,*ii = a->i;
62: PetscInt *rows;
63: PetscErrorCode ierr;
66: for (i=0; i<m; i++) {
67: if ((ii[i] != ii[i+1]) && ((jj[ii[i]] < bs*(i/bs)) || (jj[ii[i+1]-1] > bs*((i+bs)/bs)-1))) {
68: cnt++;
69: }
70: }
71: PetscMalloc1(cnt,&rows);
72: cnt = 0;
73: for (i=0; i<m; i++) {
74: if ((ii[i] != ii[i+1]) && ((jj[ii[i]] < bs*(i/bs)) || (jj[ii[i+1]-1] > bs*((i+bs)/bs)-1))) {
75: rows[cnt] = i;
76: cnt++;
77: }
78: }
79: ISCreateGeneral(PETSC_COMM_SELF,cnt,rows,PETSC_OWN_POINTER,is);
80: return(0);
81: }
83: PetscErrorCode MatFindZeroDiagonals_SeqAIJ_Private(Mat A,PetscInt *nrows,PetscInt **zrows)
84: {
85: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
86: const MatScalar *aa = a->a;
87: PetscInt i,m=A->rmap->n,cnt = 0;
88: const PetscInt *ii = a->i,*jj = a->j,*diag;
89: PetscInt *rows;
90: PetscErrorCode ierr;
93: MatMarkDiagonal_SeqAIJ(A);
94: diag = a->diag;
95: for (i=0; i<m; i++) {
96: if ((diag[i] >= ii[i+1]) || (jj[diag[i]] != i) || (aa[diag[i]] == 0.0)) {
97: cnt++;
98: }
99: }
100: PetscMalloc1(cnt,&rows);
101: cnt = 0;
102: for (i=0; i<m; i++) {
103: if ((diag[i] >= ii[i+1]) || (jj[diag[i]] != i) || (aa[diag[i]] == 0.0)) {
104: rows[cnt++] = i;
105: }
106: }
107: *nrows = cnt;
108: *zrows = rows;
109: return(0);
110: }
112: PetscErrorCode MatFindZeroDiagonals_SeqAIJ(Mat A,IS *zrows)
113: {
114: PetscInt nrows,*rows;
118: *zrows = NULL;
119: MatFindZeroDiagonals_SeqAIJ_Private(A,&nrows,&rows);
120: ISCreateGeneral(PetscObjectComm((PetscObject)A),nrows,rows,PETSC_OWN_POINTER,zrows);
121: return(0);
122: }
124: PetscErrorCode MatFindNonzeroRows_SeqAIJ(Mat A,IS *keptrows)
125: {
126: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
127: const MatScalar *aa;
128: PetscInt m=A->rmap->n,cnt = 0;
129: const PetscInt *ii;
130: PetscInt n,i,j,*rows;
131: PetscErrorCode ierr;
134: *keptrows = 0;
135: ii = a->i;
136: for (i=0; i<m; i++) {
137: n = ii[i+1] - ii[i];
138: if (!n) {
139: cnt++;
140: goto ok1;
141: }
142: aa = a->a + ii[i];
143: for (j=0; j<n; j++) {
144: if (aa[j] != 0.0) goto ok1;
145: }
146: cnt++;
147: ok1:;
148: }
149: if (!cnt) return(0);
150: PetscMalloc1(A->rmap->n-cnt,&rows);
151: cnt = 0;
152: for (i=0; i<m; i++) {
153: n = ii[i+1] - ii[i];
154: if (!n) continue;
155: aa = a->a + ii[i];
156: for (j=0; j<n; j++) {
157: if (aa[j] != 0.0) {
158: rows[cnt++] = i;
159: break;
160: }
161: }
162: }
163: ISCreateGeneral(PETSC_COMM_SELF,cnt,rows,PETSC_OWN_POINTER,keptrows);
164: return(0);
165: }
167: PetscErrorCode MatDiagonalSet_SeqAIJ(Mat Y,Vec D,InsertMode is)
168: {
169: PetscErrorCode ierr;
170: Mat_SeqAIJ *aij = (Mat_SeqAIJ*) Y->data;
171: PetscInt i,m = Y->rmap->n;
172: const PetscInt *diag;
173: MatScalar *aa = aij->a;
174: const PetscScalar *v;
175: PetscBool missing;
176: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
177: PetscBool inserted = PETSC_FALSE;
178: #endif
181: if (Y->assembled) {
182: MatMissingDiagonal_SeqAIJ(Y,&missing,NULL);
183: if (!missing) {
184: diag = aij->diag;
185: VecGetArrayRead(D,&v);
186: if (is == INSERT_VALUES) {
187: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
188: inserted = PETSC_TRUE;
189: #endif
190: for (i=0; i<m; i++) {
191: aa[diag[i]] = v[i];
192: }
193: } else {
194: for (i=0; i<m; i++) {
195: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
196: if (v[i] != 0.0) inserted = PETSC_TRUE;
197: #endif
198: aa[diag[i]] += v[i];
199: }
200: }
201: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
202: if (inserted) Y->offloadmask = PETSC_OFFLOAD_CPU;
203: #endif
204: VecRestoreArrayRead(D,&v);
205: return(0);
206: }
207: MatSeqAIJInvalidateDiagonal(Y);
208: }
209: MatDiagonalSet_Default(Y,D,is);
210: return(0);
211: }
213: PetscErrorCode MatGetRowIJ_SeqAIJ(Mat A,PetscInt oshift,PetscBool symmetric,PetscBool inodecompressed,PetscInt *m,const PetscInt *ia[],const PetscInt *ja[],PetscBool *done)
214: {
215: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
217: PetscInt i,ishift;
220: *m = A->rmap->n;
221: if (!ia) return(0);
222: ishift = 0;
223: if (symmetric && !A->structurally_symmetric) {
224: MatToSymmetricIJ_SeqAIJ(A->rmap->n,a->i,a->j,PETSC_TRUE,ishift,oshift,(PetscInt**)ia,(PetscInt**)ja);
225: } else if (oshift == 1) {
226: PetscInt *tia;
227: PetscInt nz = a->i[A->rmap->n];
228: /* malloc space and add 1 to i and j indices */
229: PetscMalloc1(A->rmap->n+1,&tia);
230: for (i=0; i<A->rmap->n+1; i++) tia[i] = a->i[i] + 1;
231: *ia = tia;
232: if (ja) {
233: PetscInt *tja;
234: PetscMalloc1(nz+1,&tja);
235: for (i=0; i<nz; i++) tja[i] = a->j[i] + 1;
236: *ja = tja;
237: }
238: } else {
239: *ia = a->i;
240: if (ja) *ja = a->j;
241: }
242: return(0);
243: }
245: PetscErrorCode MatRestoreRowIJ_SeqAIJ(Mat A,PetscInt oshift,PetscBool symmetric,PetscBool inodecompressed,PetscInt *n,const PetscInt *ia[],const PetscInt *ja[],PetscBool *done)
246: {
250: if (!ia) return(0);
251: if ((symmetric && !A->structurally_symmetric) || oshift == 1) {
252: PetscFree(*ia);
253: if (ja) {PetscFree(*ja);}
254: }
255: return(0);
256: }
258: PetscErrorCode MatGetColumnIJ_SeqAIJ(Mat A,PetscInt oshift,PetscBool symmetric,PetscBool inodecompressed,PetscInt *nn,const PetscInt *ia[],const PetscInt *ja[],PetscBool *done)
259: {
260: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
262: PetscInt i,*collengths,*cia,*cja,n = A->cmap->n,m = A->rmap->n;
263: PetscInt nz = a->i[m],row,*jj,mr,col;
266: *nn = n;
267: if (!ia) return(0);
268: if (symmetric) {
269: MatToSymmetricIJ_SeqAIJ(A->rmap->n,a->i,a->j,PETSC_TRUE,0,oshift,(PetscInt**)ia,(PetscInt**)ja);
270: } else {
271: PetscCalloc1(n,&collengths);
272: PetscMalloc1(n+1,&cia);
273: PetscMalloc1(nz,&cja);
274: jj = a->j;
275: for (i=0; i<nz; i++) {
276: collengths[jj[i]]++;
277: }
278: cia[0] = oshift;
279: for (i=0; i<n; i++) {
280: cia[i+1] = cia[i] + collengths[i];
281: }
282: PetscArrayzero(collengths,n);
283: jj = a->j;
284: for (row=0; row<m; row++) {
285: mr = a->i[row+1] - a->i[row];
286: for (i=0; i<mr; i++) {
287: col = *jj++;
289: cja[cia[col] + collengths[col]++ - oshift] = row + oshift;
290: }
291: }
292: PetscFree(collengths);
293: *ia = cia; *ja = cja;
294: }
295: return(0);
296: }
298: PetscErrorCode MatRestoreColumnIJ_SeqAIJ(Mat A,PetscInt oshift,PetscBool symmetric,PetscBool inodecompressed,PetscInt *n,const PetscInt *ia[],const PetscInt *ja[],PetscBool *done)
299: {
303: if (!ia) return(0);
305: PetscFree(*ia);
306: PetscFree(*ja);
307: return(0);
308: }
310: /*
311: MatGetColumnIJ_SeqAIJ_Color() and MatRestoreColumnIJ_SeqAIJ_Color() are customized from
312: MatGetColumnIJ_SeqAIJ() and MatRestoreColumnIJ_SeqAIJ() by adding an output
313: spidx[], index of a->a, to be used in MatTransposeColoringCreate_SeqAIJ() and MatFDColoringCreate_SeqXAIJ()
314: */
315: PetscErrorCode MatGetColumnIJ_SeqAIJ_Color(Mat A,PetscInt oshift,PetscBool symmetric,PetscBool inodecompressed,PetscInt *nn,const PetscInt *ia[],const PetscInt *ja[],PetscInt *spidx[],PetscBool *done)
316: {
317: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
319: PetscInt i,*collengths,*cia,*cja,n = A->cmap->n,m = A->rmap->n;
320: PetscInt nz = a->i[m],row,mr,col,tmp;
321: PetscInt *cspidx;
322: const PetscInt *jj;
325: *nn = n;
326: if (!ia) return(0);
328: PetscCalloc1(n,&collengths);
329: PetscMalloc1(n+1,&cia);
330: PetscMalloc1(nz,&cja);
331: PetscMalloc1(nz,&cspidx);
332: jj = a->j;
333: for (i=0; i<nz; i++) {
334: collengths[jj[i]]++;
335: }
336: cia[0] = oshift;
337: for (i=0; i<n; i++) {
338: cia[i+1] = cia[i] + collengths[i];
339: }
340: PetscArrayzero(collengths,n);
341: jj = a->j;
342: for (row=0; row<m; row++) {
343: mr = a->i[row+1] - a->i[row];
344: for (i=0; i<mr; i++) {
345: col = *jj++;
346: tmp = cia[col] + collengths[col]++ - oshift;
347: cspidx[tmp] = a->i[row] + i; /* index of a->j */
348: cja[tmp] = row + oshift;
349: }
350: }
351: PetscFree(collengths);
352: *ia = cia;
353: *ja = cja;
354: *spidx = cspidx;
355: return(0);
356: }
358: PetscErrorCode MatRestoreColumnIJ_SeqAIJ_Color(Mat A,PetscInt oshift,PetscBool symmetric,PetscBool inodecompressed,PetscInt *n,const PetscInt *ia[],const PetscInt *ja[],PetscInt *spidx[],PetscBool *done)
359: {
363: MatRestoreColumnIJ_SeqAIJ(A,oshift,symmetric,inodecompressed,n,ia,ja,done);
364: PetscFree(*spidx);
365: return(0);
366: }
368: PetscErrorCode MatSetValuesRow_SeqAIJ(Mat A,PetscInt row,const PetscScalar v[])
369: {
370: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
371: PetscInt *ai = a->i;
375: PetscArraycpy(a->a+ai[row],v,ai[row+1]-ai[row]);
376: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
377: if (A->offloadmask != PETSC_OFFLOAD_UNALLOCATED && ai[row+1]-ai[row]) A->offloadmask = PETSC_OFFLOAD_CPU;
378: #endif
379: return(0);
380: }
382: /*
383: MatSeqAIJSetValuesLocalFast - An optimized version of MatSetValuesLocal() for SeqAIJ matrices with several assumptions
385: - a single row of values is set with each call
386: - no row or column indices are negative or (in error) larger than the number of rows or columns
387: - the values are always added to the matrix, not set
388: - no new locations are introduced in the nonzero structure of the matrix
390: This does NOT assume the global column indices are sorted
392: */
394: #include <petsc/private/isimpl.h>
395: PetscErrorCode MatSeqAIJSetValuesLocalFast(Mat A,PetscInt m,const PetscInt im[],PetscInt n,const PetscInt in[],const PetscScalar v[],InsertMode is)
396: {
397: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
398: PetscInt low,high,t,row,nrow,i,col,l;
399: const PetscInt *rp,*ai = a->i,*ailen = a->ilen,*aj = a->j;
400: PetscInt lastcol = -1;
401: MatScalar *ap,value,*aa = a->a;
402: const PetscInt *ridx = A->rmap->mapping->indices,*cidx = A->cmap->mapping->indices;
404: row = ridx[im[0]];
405: rp = aj + ai[row];
406: ap = aa + ai[row];
407: nrow = ailen[row];
408: low = 0;
409: high = nrow;
410: for (l=0; l<n; l++) { /* loop over added columns */
411: col = cidx[in[l]];
412: value = v[l];
414: if (col <= lastcol) low = 0;
415: else high = nrow;
416: lastcol = col;
417: while (high-low > 5) {
418: t = (low+high)/2;
419: if (rp[t] > col) high = t;
420: else low = t;
421: }
422: for (i=low; i<high; i++) {
423: if (rp[i] == col) {
424: ap[i] += value;
425: low = i + 1;
426: break;
427: }
428: }
429: }
430: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
431: if (A->offloadmask != PETSC_OFFLOAD_UNALLOCATED && m && n) A->offloadmask = PETSC_OFFLOAD_CPU;
432: #endif
433: return 0;
434: }
436: PetscErrorCode MatSetValues_SeqAIJ(Mat A,PetscInt m,const PetscInt im[],PetscInt n,const PetscInt in[],const PetscScalar v[],InsertMode is)
437: {
438: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
439: PetscInt *rp,k,low,high,t,ii,row,nrow,i,col,l,rmax,N;
440: PetscInt *imax = a->imax,*ai = a->i,*ailen = a->ilen;
442: PetscInt *aj = a->j,nonew = a->nonew,lastcol = -1;
443: MatScalar *ap=NULL,value=0.0,*aa = a->a;
444: PetscBool ignorezeroentries = a->ignorezeroentries;
445: PetscBool roworiented = a->roworiented;
446: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
447: PetscBool inserted = PETSC_FALSE;
448: #endif
451: for (k=0; k<m; k++) { /* loop over added rows */
452: row = im[k];
453: if (row < 0) continue;
454: if (PetscUnlikelyDebug(row >= A->rmap->n)) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Row too large: row %D max %D",row,A->rmap->n-1);
455: rp = aj + ai[row];
456: if (!A->structure_only) ap = aa + ai[row];
457: rmax = imax[row]; nrow = ailen[row];
458: low = 0;
459: high = nrow;
460: for (l=0; l<n; l++) { /* loop over added columns */
461: if (in[l] < 0) continue;
462: if (PetscUnlikelyDebug(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);
463: col = in[l];
464: if (v && !A->structure_only) value = roworiented ? v[l + k*n] : v[k + l*m];
465: if (!A->structure_only && value == 0.0 && ignorezeroentries && is == ADD_VALUES && row != col) continue;
467: if (col <= lastcol) low = 0;
468: else high = nrow;
469: lastcol = col;
470: while (high-low > 5) {
471: t = (low+high)/2;
472: if (rp[t] > col) high = t;
473: else low = t;
474: }
475: for (i=low; i<high; i++) {
476: if (rp[i] > col) break;
477: if (rp[i] == col) {
478: if (!A->structure_only) {
479: if (is == ADD_VALUES) {
480: ap[i] += value;
481: (void)PetscLogFlops(1.0);
482: }
483: else ap[i] = value;
484: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
485: inserted = PETSC_TRUE;
486: #endif
487: }
488: low = i + 1;
489: goto noinsert;
490: }
491: }
492: if (value == 0.0 && ignorezeroentries && row != col) goto noinsert;
493: if (nonew == 1) goto noinsert;
494: if (nonew == -1) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Inserting a new nonzero at (%D,%D) in the matrix",row,col);
495: if (A->structure_only) {
496: MatSeqXAIJReallocateAIJ_structure_only(A,A->rmap->n,1,nrow,row,col,rmax,ai,aj,rp,imax,nonew,MatScalar);
497: } else {
498: MatSeqXAIJReallocateAIJ(A,A->rmap->n,1,nrow,row,col,rmax,aa,ai,aj,rp,ap,imax,nonew,MatScalar);
499: }
500: N = nrow++ - 1; a->nz++; high++;
501: /* shift up all the later entries in this row */
502: PetscArraymove(rp+i+1,rp+i,N-i+1);
503: rp[i] = col;
504: if (!A->structure_only){
505: PetscArraymove(ap+i+1,ap+i,N-i+1);
506: ap[i] = value;
507: }
508: low = i + 1;
509: A->nonzerostate++;
510: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
511: inserted = PETSC_TRUE;
512: #endif
513: noinsert:;
514: }
515: ailen[row] = nrow;
516: }
517: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
518: if (A->offloadmask != PETSC_OFFLOAD_UNALLOCATED && inserted) A->offloadmask = PETSC_OFFLOAD_CPU;
519: #endif
520: return(0);
521: }
523: PetscErrorCode MatSetValues_SeqAIJ_SortedFull(Mat A,PetscInt m,const PetscInt im[],PetscInt n,const PetscInt in[],const PetscScalar v[],InsertMode is)
524: {
525: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
526: PetscInt *rp,k,row;
527: PetscInt *ai = a->i,*ailen = a->ilen;
529: PetscInt *aj = a->j;
530: MatScalar *aa = a->a,*ap;
533: for (k=0; k<m; k++) { /* loop over added rows */
534: row = im[k];
535: rp = aj + ai[row];
536: ap = aa + ai[row];
537: if (!A->was_assembled) {
538: PetscMemcpy(rp,in,n*sizeof(PetscInt));
539: }
540: if (!A->structure_only) {
541: if (v) {
542: PetscMemcpy(ap,v,n*sizeof(PetscScalar));
543: v += n;
544: } else {
545: PetscMemzero(ap,n*sizeof(PetscScalar));
546: }
547: }
548: ailen[row] = n;
549: a->nz += n;
550: }
551: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
552: if (A->offloadmask != PETSC_OFFLOAD_UNALLOCATED && m && n) A->offloadmask = PETSC_OFFLOAD_CPU;
553: #endif
554: return(0);
555: }
558: PetscErrorCode MatGetValues_SeqAIJ(Mat A,PetscInt m,const PetscInt im[],PetscInt n,const PetscInt in[],PetscScalar v[])
559: {
560: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
561: PetscInt *rp,k,low,high,t,row,nrow,i,col,l,*aj = a->j;
562: PetscInt *ai = a->i,*ailen = a->ilen;
563: MatScalar *ap,*aa = a->a;
566: for (k=0; k<m; k++) { /* loop over rows */
567: row = im[k];
568: if (row < 0) {v += n; continue;} /* SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Negative row: %D",row); */
569: 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);
570: rp = aj + ai[row]; ap = aa + ai[row];
571: nrow = ailen[row];
572: for (l=0; l<n; l++) { /* loop over columns */
573: if (in[l] < 0) {v++; continue;} /* SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Negative column: %D",in[l]); */
574: 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);
575: col = in[l];
576: high = nrow; low = 0; /* assume unsorted */
577: while (high-low > 5) {
578: t = (low+high)/2;
579: if (rp[t] > col) high = t;
580: else low = t;
581: }
582: for (i=low; i<high; i++) {
583: if (rp[i] > col) break;
584: if (rp[i] == col) {
585: *v++ = ap[i];
586: goto finished;
587: }
588: }
589: *v++ = 0.0;
590: finished:;
591: }
592: }
593: return(0);
594: }
596: PetscErrorCode MatView_SeqAIJ_Binary(Mat mat,PetscViewer viewer)
597: {
598: Mat_SeqAIJ *A = (Mat_SeqAIJ*)mat->data;
599: PetscInt header[4],M,N,m,nz,i;
600: PetscInt *rowlens;
604: PetscViewerSetUp(viewer);
606: M = mat->rmap->N;
607: N = mat->cmap->N;
608: m = mat->rmap->n;
609: nz = A->nz;
611: /* write matrix header */
612: header[0] = MAT_FILE_CLASSID;
613: header[1] = M; header[2] = N; header[3] = nz;
614: PetscViewerBinaryWrite(viewer,header,4,PETSC_INT);
616: /* fill in and store row lengths */
617: PetscMalloc1(m,&rowlens);
618: for (i=0; i<m; i++) rowlens[i] = A->i[i+1] - A->i[i];
619: PetscViewerBinaryWrite(viewer,rowlens,m,PETSC_INT);
620: PetscFree(rowlens);
621: /* store column indices */
622: PetscViewerBinaryWrite(viewer,A->j,nz,PETSC_INT);
623: /* store nonzero values */
624: PetscViewerBinaryWrite(viewer,A->a,nz,PETSC_SCALAR);
626: /* write block size option to the viewer's .info file */
627: MatView_Binary_BlockSizes(mat,viewer);
628: return(0);
629: }
631: static PetscErrorCode MatView_SeqAIJ_ASCII_structonly(Mat A,PetscViewer viewer)
632: {
634: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
635: PetscInt i,k,m=A->rmap->N;
638: PetscViewerASCIIUseTabs(viewer,PETSC_FALSE);
639: for (i=0; i<m; i++) {
640: PetscViewerASCIIPrintf(viewer,"row %D:",i);
641: for (k=a->i[i]; k<a->i[i+1]; k++) {
642: PetscViewerASCIIPrintf(viewer," (%D) ",a->j[k]);
643: }
644: PetscViewerASCIIPrintf(viewer,"\n");
645: }
646: PetscViewerASCIIUseTabs(viewer,PETSC_TRUE);
647: return(0);
648: }
650: extern PetscErrorCode MatSeqAIJFactorInfo_Matlab(Mat,PetscViewer);
652: PetscErrorCode MatView_SeqAIJ_ASCII(Mat A,PetscViewer viewer)
653: {
654: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
655: PetscErrorCode ierr;
656: PetscInt i,j,m = A->rmap->n;
657: const char *name;
658: PetscViewerFormat format;
661: if (A->structure_only) {
662: MatView_SeqAIJ_ASCII_structonly(A,viewer);
663: return(0);
664: }
666: PetscViewerGetFormat(viewer,&format);
667: if (format == PETSC_VIEWER_ASCII_MATLAB) {
668: PetscInt nofinalvalue = 0;
669: if (m && ((a->i[m] == a->i[m-1]) || (a->j[a->nz-1] != A->cmap->n-1))) {
670: /* Need a dummy value to ensure the dimension of the matrix. */
671: nofinalvalue = 1;
672: }
673: PetscViewerASCIIUseTabs(viewer,PETSC_FALSE);
674: PetscViewerASCIIPrintf(viewer,"%% Size = %D %D \n",m,A->cmap->n);
675: PetscViewerASCIIPrintf(viewer,"%% Nonzeros = %D \n",a->nz);
676: #if defined(PETSC_USE_COMPLEX)
677: PetscViewerASCIIPrintf(viewer,"zzz = zeros(%D,4);\n",a->nz+nofinalvalue);
678: #else
679: PetscViewerASCIIPrintf(viewer,"zzz = zeros(%D,3);\n",a->nz+nofinalvalue);
680: #endif
681: PetscViewerASCIIPrintf(viewer,"zzz = [\n");
683: for (i=0; i<m; i++) {
684: for (j=a->i[i]; j<a->i[i+1]; j++) {
685: #if defined(PETSC_USE_COMPLEX)
686: 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]));
687: #else
688: PetscViewerASCIIPrintf(viewer,"%D %D %18.16e\n",i+1,a->j[j]+1,(double)a->a[j]);
689: #endif
690: }
691: }
692: if (nofinalvalue) {
693: #if defined(PETSC_USE_COMPLEX)
694: PetscViewerASCIIPrintf(viewer,"%D %D %18.16e %18.16e\n",m,A->cmap->n,0.,0.);
695: #else
696: PetscViewerASCIIPrintf(viewer,"%D %D %18.16e\n",m,A->cmap->n,0.0);
697: #endif
698: }
699: PetscObjectGetName((PetscObject)A,&name);
700: PetscViewerASCIIPrintf(viewer,"];\n %s = spconvert(zzz);\n",name);
701: PetscViewerASCIIUseTabs(viewer,PETSC_TRUE);
702: } else if (format == PETSC_VIEWER_ASCII_FACTOR_INFO || format == PETSC_VIEWER_ASCII_INFO || format == PETSC_VIEWER_ASCII_INFO_DETAIL) {
703: return(0);
704: } else if (format == PETSC_VIEWER_ASCII_COMMON) {
705: PetscViewerASCIIUseTabs(viewer,PETSC_FALSE);
706: for (i=0; i<m; i++) {
707: PetscViewerASCIIPrintf(viewer,"row %D:",i);
708: for (j=a->i[i]; j<a->i[i+1]; j++) {
709: #if defined(PETSC_USE_COMPLEX)
710: if (PetscImaginaryPart(a->a[j]) > 0.0 && PetscRealPart(a->a[j]) != 0.0) {
711: PetscViewerASCIIPrintf(viewer," (%D, %g + %g i)",a->j[j],(double)PetscRealPart(a->a[j]),(double)PetscImaginaryPart(a->a[j]));
712: } else if (PetscImaginaryPart(a->a[j]) < 0.0 && PetscRealPart(a->a[j]) != 0.0) {
713: PetscViewerASCIIPrintf(viewer," (%D, %g - %g i)",a->j[j],(double)PetscRealPart(a->a[j]),(double)-PetscImaginaryPart(a->a[j]));
714: } else if (PetscRealPart(a->a[j]) != 0.0) {
715: PetscViewerASCIIPrintf(viewer," (%D, %g) ",a->j[j],(double)PetscRealPart(a->a[j]));
716: }
717: #else
718: if (a->a[j] != 0.0) {PetscViewerASCIIPrintf(viewer," (%D, %g) ",a->j[j],(double)a->a[j]);}
719: #endif
720: }
721: PetscViewerASCIIPrintf(viewer,"\n");
722: }
723: PetscViewerASCIIUseTabs(viewer,PETSC_TRUE);
724: } else if (format == PETSC_VIEWER_ASCII_SYMMODU) {
725: PetscInt nzd=0,fshift=1,*sptr;
726: PetscViewerASCIIUseTabs(viewer,PETSC_FALSE);
727: PetscMalloc1(m+1,&sptr);
728: for (i=0; i<m; i++) {
729: sptr[i] = nzd+1;
730: for (j=a->i[i]; j<a->i[i+1]; j++) {
731: if (a->j[j] >= i) {
732: #if defined(PETSC_USE_COMPLEX)
733: if (PetscImaginaryPart(a->a[j]) != 0.0 || PetscRealPart(a->a[j]) != 0.0) nzd++;
734: #else
735: if (a->a[j] != 0.0) nzd++;
736: #endif
737: }
738: }
739: }
740: sptr[m] = nzd+1;
741: PetscViewerASCIIPrintf(viewer," %D %D\n\n",m,nzd);
742: for (i=0; i<m+1; i+=6) {
743: if (i+4<m) {
744: 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]);
745: } else if (i+3<m) {
746: PetscViewerASCIIPrintf(viewer," %D %D %D %D %D\n",sptr[i],sptr[i+1],sptr[i+2],sptr[i+3],sptr[i+4]);
747: } else if (i+2<m) {
748: PetscViewerASCIIPrintf(viewer," %D %D %D %D\n",sptr[i],sptr[i+1],sptr[i+2],sptr[i+3]);
749: } else if (i+1<m) {
750: PetscViewerASCIIPrintf(viewer," %D %D %D\n",sptr[i],sptr[i+1],sptr[i+2]);
751: } else if (i<m) {
752: PetscViewerASCIIPrintf(viewer," %D %D\n",sptr[i],sptr[i+1]);
753: } else {
754: PetscViewerASCIIPrintf(viewer," %D\n",sptr[i]);
755: }
756: }
757: PetscViewerASCIIPrintf(viewer,"\n");
758: PetscFree(sptr);
759: for (i=0; i<m; i++) {
760: for (j=a->i[i]; j<a->i[i+1]; j++) {
761: if (a->j[j] >= i) {PetscViewerASCIIPrintf(viewer," %D ",a->j[j]+fshift);}
762: }
763: PetscViewerASCIIPrintf(viewer,"\n");
764: }
765: PetscViewerASCIIPrintf(viewer,"\n");
766: for (i=0; i<m; i++) {
767: for (j=a->i[i]; j<a->i[i+1]; j++) {
768: if (a->j[j] >= i) {
769: #if defined(PETSC_USE_COMPLEX)
770: if (PetscImaginaryPart(a->a[j]) != 0.0 || PetscRealPart(a->a[j]) != 0.0) {
771: PetscViewerASCIIPrintf(viewer," %18.16e %18.16e ",(double)PetscRealPart(a->a[j]),(double)PetscImaginaryPart(a->a[j]));
772: }
773: #else
774: if (a->a[j] != 0.0) {PetscViewerASCIIPrintf(viewer," %18.16e ",(double)a->a[j]);}
775: #endif
776: }
777: }
778: PetscViewerASCIIPrintf(viewer,"\n");
779: }
780: PetscViewerASCIIUseTabs(viewer,PETSC_TRUE);
781: } else if (format == PETSC_VIEWER_ASCII_DENSE) {
782: PetscInt cnt = 0,jcnt;
783: PetscScalar value;
784: #if defined(PETSC_USE_COMPLEX)
785: PetscBool realonly = PETSC_TRUE;
787: for (i=0; i<a->i[m]; i++) {
788: if (PetscImaginaryPart(a->a[i]) != 0.0) {
789: realonly = PETSC_FALSE;
790: break;
791: }
792: }
793: #endif
795: PetscViewerASCIIUseTabs(viewer,PETSC_FALSE);
796: for (i=0; i<m; i++) {
797: jcnt = 0;
798: for (j=0; j<A->cmap->n; j++) {
799: if (jcnt < a->i[i+1]-a->i[i] && j == a->j[cnt]) {
800: value = a->a[cnt++];
801: jcnt++;
802: } else {
803: value = 0.0;
804: }
805: #if defined(PETSC_USE_COMPLEX)
806: if (realonly) {
807: PetscViewerASCIIPrintf(viewer," %7.5e ",(double)PetscRealPart(value));
808: } else {
809: PetscViewerASCIIPrintf(viewer," %7.5e+%7.5e i ",(double)PetscRealPart(value),(double)PetscImaginaryPart(value));
810: }
811: #else
812: PetscViewerASCIIPrintf(viewer," %7.5e ",(double)value);
813: #endif
814: }
815: PetscViewerASCIIPrintf(viewer,"\n");
816: }
817: PetscViewerASCIIUseTabs(viewer,PETSC_TRUE);
818: } else if (format == PETSC_VIEWER_ASCII_MATRIXMARKET) {
819: PetscInt fshift=1;
820: PetscViewerASCIIUseTabs(viewer,PETSC_FALSE);
821: #if defined(PETSC_USE_COMPLEX)
822: PetscViewerASCIIPrintf(viewer,"%%%%MatrixMarket matrix coordinate complex general\n");
823: #else
824: PetscViewerASCIIPrintf(viewer,"%%%%MatrixMarket matrix coordinate real general\n");
825: #endif
826: PetscViewerASCIIPrintf(viewer,"%D %D %D\n", m, A->cmap->n, a->nz);
827: for (i=0; i<m; i++) {
828: for (j=a->i[i]; j<a->i[i+1]; j++) {
829: #if defined(PETSC_USE_COMPLEX)
830: PetscViewerASCIIPrintf(viewer,"%D %D %g %g\n", i+fshift,a->j[j]+fshift,(double)PetscRealPart(a->a[j]),(double)PetscImaginaryPart(a->a[j]));
831: #else
832: PetscViewerASCIIPrintf(viewer,"%D %D %g\n", i+fshift, a->j[j]+fshift, (double)a->a[j]);
833: #endif
834: }
835: }
836: PetscViewerASCIIUseTabs(viewer,PETSC_TRUE);
837: } else {
838: PetscViewerASCIIUseTabs(viewer,PETSC_FALSE);
839: if (A->factortype) {
840: for (i=0; i<m; i++) {
841: PetscViewerASCIIPrintf(viewer,"row %D:",i);
842: /* L part */
843: for (j=a->i[i]; j<a->i[i+1]; j++) {
844: #if defined(PETSC_USE_COMPLEX)
845: if (PetscImaginaryPart(a->a[j]) > 0.0) {
846: PetscViewerASCIIPrintf(viewer," (%D, %g + %g i)",a->j[j],(double)PetscRealPart(a->a[j]),(double)PetscImaginaryPart(a->a[j]));
847: } else if (PetscImaginaryPart(a->a[j]) < 0.0) {
848: PetscViewerASCIIPrintf(viewer," (%D, %g - %g i)",a->j[j],(double)PetscRealPart(a->a[j]),(double)(-PetscImaginaryPart(a->a[j])));
849: } else {
850: PetscViewerASCIIPrintf(viewer," (%D, %g) ",a->j[j],(double)PetscRealPart(a->a[j]));
851: }
852: #else
853: PetscViewerASCIIPrintf(viewer," (%D, %g) ",a->j[j],(double)a->a[j]);
854: #endif
855: }
856: /* diagonal */
857: j = a->diag[i];
858: #if defined(PETSC_USE_COMPLEX)
859: if (PetscImaginaryPart(a->a[j]) > 0.0) {
860: PetscViewerASCIIPrintf(viewer," (%D, %g + %g i)",a->j[j],(double)PetscRealPart(1.0/a->a[j]),(double)PetscImaginaryPart(1.0/a->a[j]));
861: } else if (PetscImaginaryPart(a->a[j]) < 0.0) {
862: PetscViewerASCIIPrintf(viewer," (%D, %g - %g i)",a->j[j],(double)PetscRealPart(1.0/a->a[j]),(double)(-PetscImaginaryPart(1.0/a->a[j])));
863: } else {
864: PetscViewerASCIIPrintf(viewer," (%D, %g) ",a->j[j],(double)PetscRealPart(1.0/a->a[j]));
865: }
866: #else
867: PetscViewerASCIIPrintf(viewer," (%D, %g) ",a->j[j],(double)(1.0/a->a[j]));
868: #endif
870: /* U part */
871: for (j=a->diag[i+1]+1; j<a->diag[i]; j++) {
872: #if defined(PETSC_USE_COMPLEX)
873: if (PetscImaginaryPart(a->a[j]) > 0.0) {
874: PetscViewerASCIIPrintf(viewer," (%D, %g + %g i)",a->j[j],(double)PetscRealPart(a->a[j]),(double)PetscImaginaryPart(a->a[j]));
875: } else if (PetscImaginaryPart(a->a[j]) < 0.0) {
876: PetscViewerASCIIPrintf(viewer," (%D, %g - %g i)",a->j[j],(double)PetscRealPart(a->a[j]),(double)(-PetscImaginaryPart(a->a[j])));
877: } else {
878: PetscViewerASCIIPrintf(viewer," (%D, %g) ",a->j[j],(double)PetscRealPart(a->a[j]));
879: }
880: #else
881: PetscViewerASCIIPrintf(viewer," (%D, %g) ",a->j[j],(double)a->a[j]);
882: #endif
883: }
884: PetscViewerASCIIPrintf(viewer,"\n");
885: }
886: } else {
887: for (i=0; i<m; i++) {
888: PetscViewerASCIIPrintf(viewer,"row %D:",i);
889: for (j=a->i[i]; j<a->i[i+1]; j++) {
890: #if defined(PETSC_USE_COMPLEX)
891: if (PetscImaginaryPart(a->a[j]) > 0.0) {
892: PetscViewerASCIIPrintf(viewer," (%D, %g + %g i)",a->j[j],(double)PetscRealPart(a->a[j]),(double)PetscImaginaryPart(a->a[j]));
893: } else if (PetscImaginaryPart(a->a[j]) < 0.0) {
894: PetscViewerASCIIPrintf(viewer," (%D, %g - %g i)",a->j[j],(double)PetscRealPart(a->a[j]),(double)-PetscImaginaryPart(a->a[j]));
895: } else {
896: PetscViewerASCIIPrintf(viewer," (%D, %g) ",a->j[j],(double)PetscRealPart(a->a[j]));
897: }
898: #else
899: PetscViewerASCIIPrintf(viewer," (%D, %g) ",a->j[j],(double)a->a[j]);
900: #endif
901: }
902: PetscViewerASCIIPrintf(viewer,"\n");
903: }
904: }
905: PetscViewerASCIIUseTabs(viewer,PETSC_TRUE);
906: }
907: PetscViewerFlush(viewer);
908: return(0);
909: }
911: #include <petscdraw.h>
912: PetscErrorCode MatView_SeqAIJ_Draw_Zoom(PetscDraw draw,void *Aa)
913: {
914: Mat A = (Mat) Aa;
915: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
916: PetscErrorCode ierr;
917: PetscInt i,j,m = A->rmap->n;
918: int color;
919: PetscReal xl,yl,xr,yr,x_l,x_r,y_l,y_r;
920: PetscViewer viewer;
921: PetscViewerFormat format;
924: PetscObjectQuery((PetscObject)A,"Zoomviewer",(PetscObject*)&viewer);
925: PetscViewerGetFormat(viewer,&format);
926: PetscDrawGetCoordinates(draw,&xl,&yl,&xr,&yr);
928: /* loop over matrix elements drawing boxes */
930: if (format != PETSC_VIEWER_DRAW_CONTOUR) {
931: PetscDrawCollectiveBegin(draw);
932: /* Blue for negative, Cyan for zero and Red for positive */
933: color = PETSC_DRAW_BLUE;
934: for (i=0; i<m; i++) {
935: y_l = m - i - 1.0; y_r = y_l + 1.0;
936: for (j=a->i[i]; j<a->i[i+1]; j++) {
937: x_l = a->j[j]; x_r = x_l + 1.0;
938: if (PetscRealPart(a->a[j]) >= 0.) continue;
939: PetscDrawRectangle(draw,x_l,y_l,x_r,y_r,color,color,color,color);
940: }
941: }
942: color = PETSC_DRAW_CYAN;
943: for (i=0; i<m; i++) {
944: y_l = m - i - 1.0; y_r = y_l + 1.0;
945: for (j=a->i[i]; j<a->i[i+1]; j++) {
946: x_l = a->j[j]; x_r = x_l + 1.0;
947: if (a->a[j] != 0.) continue;
948: PetscDrawRectangle(draw,x_l,y_l,x_r,y_r,color,color,color,color);
949: }
950: }
951: color = PETSC_DRAW_RED;
952: for (i=0; i<m; i++) {
953: y_l = m - i - 1.0; y_r = y_l + 1.0;
954: for (j=a->i[i]; j<a->i[i+1]; j++) {
955: x_l = a->j[j]; x_r = x_l + 1.0;
956: if (PetscRealPart(a->a[j]) <= 0.) continue;
957: PetscDrawRectangle(draw,x_l,y_l,x_r,y_r,color,color,color,color);
958: }
959: }
960: PetscDrawCollectiveEnd(draw);
961: } else {
962: /* use contour shading to indicate magnitude of values */
963: /* first determine max of all nonzero values */
964: PetscReal minv = 0.0, maxv = 0.0;
965: PetscInt nz = a->nz, count = 0;
966: PetscDraw popup;
968: for (i=0; i<nz; i++) {
969: if (PetscAbsScalar(a->a[i]) > maxv) maxv = PetscAbsScalar(a->a[i]);
970: }
971: if (minv >= maxv) maxv = minv + PETSC_SMALL;
972: PetscDrawGetPopup(draw,&popup);
973: PetscDrawScalePopup(popup,minv,maxv);
975: PetscDrawCollectiveBegin(draw);
976: for (i=0; i<m; i++) {
977: y_l = m - i - 1.0;
978: y_r = y_l + 1.0;
979: for (j=a->i[i]; j<a->i[i+1]; j++) {
980: x_l = a->j[j];
981: x_r = x_l + 1.0;
982: color = PetscDrawRealToColor(PetscAbsScalar(a->a[count]),minv,maxv);
983: PetscDrawRectangle(draw,x_l,y_l,x_r,y_r,color,color,color,color);
984: count++;
985: }
986: }
987: PetscDrawCollectiveEnd(draw);
988: }
989: return(0);
990: }
992: #include <petscdraw.h>
993: PetscErrorCode MatView_SeqAIJ_Draw(Mat A,PetscViewer viewer)
994: {
996: PetscDraw draw;
997: PetscReal xr,yr,xl,yl,h,w;
998: PetscBool isnull;
1001: PetscViewerDrawGetDraw(viewer,0,&draw);
1002: PetscDrawIsNull(draw,&isnull);
1003: if (isnull) return(0);
1005: xr = A->cmap->n; yr = A->rmap->n; h = yr/10.0; w = xr/10.0;
1006: xr += w; yr += h; xl = -w; yl = -h;
1007: PetscDrawSetCoordinates(draw,xl,yl,xr,yr);
1008: PetscObjectCompose((PetscObject)A,"Zoomviewer",(PetscObject)viewer);
1009: PetscDrawZoom(draw,MatView_SeqAIJ_Draw_Zoom,A);
1010: PetscObjectCompose((PetscObject)A,"Zoomviewer",NULL);
1011: PetscDrawSave(draw);
1012: return(0);
1013: }
1015: PetscErrorCode MatView_SeqAIJ(Mat A,PetscViewer viewer)
1016: {
1018: PetscBool iascii,isbinary,isdraw;
1021: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERASCII,&iascii);
1022: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERBINARY,&isbinary);
1023: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERDRAW,&isdraw);
1024: if (iascii) {
1025: MatView_SeqAIJ_ASCII(A,viewer);
1026: } else if (isbinary) {
1027: MatView_SeqAIJ_Binary(A,viewer);
1028: } else if (isdraw) {
1029: MatView_SeqAIJ_Draw(A,viewer);
1030: }
1031: MatView_SeqAIJ_Inode(A,viewer);
1032: return(0);
1033: }
1035: PetscErrorCode MatAssemblyEnd_SeqAIJ(Mat A,MatAssemblyType mode)
1036: {
1037: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1039: PetscInt fshift = 0,i,*ai = a->i,*aj = a->j,*imax = a->imax;
1040: PetscInt m = A->rmap->n,*ip,N,*ailen = a->ilen,rmax = 0;
1041: MatScalar *aa = a->a,*ap;
1042: PetscReal ratio = 0.6;
1045: if (mode == MAT_FLUSH_ASSEMBLY) return(0);
1046: MatSeqAIJInvalidateDiagonal(A);
1047: if (A->was_assembled && A->ass_nonzerostate == A->nonzerostate) return(0);
1049: if (m) rmax = ailen[0]; /* determine row with most nonzeros */
1050: for (i=1; i<m; i++) {
1051: /* move each row back by the amount of empty slots (fshift) before it*/
1052: fshift += imax[i-1] - ailen[i-1];
1053: rmax = PetscMax(rmax,ailen[i]);
1054: if (fshift) {
1055: ip = aj + ai[i];
1056: ap = aa + ai[i];
1057: N = ailen[i];
1058: PetscArraymove(ip-fshift,ip,N);
1059: if (!A->structure_only) {
1060: PetscArraymove(ap-fshift,ap,N);
1061: }
1062: }
1063: ai[i] = ai[i-1] + ailen[i-1];
1064: }
1065: if (m) {
1066: fshift += imax[m-1] - ailen[m-1];
1067: ai[m] = ai[m-1] + ailen[m-1];
1068: }
1070: /* reset ilen and imax for each row */
1071: a->nonzerorowcnt = 0;
1072: if (A->structure_only) {
1073: PetscFree(a->imax);
1074: PetscFree(a->ilen);
1075: } else { /* !A->structure_only */
1076: for (i=0; i<m; i++) {
1077: ailen[i] = imax[i] = ai[i+1] - ai[i];
1078: a->nonzerorowcnt += ((ai[i+1] - ai[i]) > 0);
1079: }
1080: }
1081: a->nz = ai[m];
1082: 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);
1084: MatMarkDiagonal_SeqAIJ(A);
1085: PetscInfo4(A,"Matrix size: %D X %D; storage space: %D unneeded,%D used\n",m,A->cmap->n,fshift,a->nz);
1086: PetscInfo1(A,"Number of mallocs during MatSetValues() is %D\n",a->reallocs);
1087: PetscInfo1(A,"Maximum nonzeros in any row is %D\n",rmax);
1089: A->info.mallocs += a->reallocs;
1090: a->reallocs = 0;
1091: A->info.nz_unneeded = (PetscReal)fshift;
1092: a->rmax = rmax;
1094: if (!A->structure_only) {
1095: MatCheckCompressedRow(A,a->nonzerorowcnt,&a->compressedrow,a->i,m,ratio);
1096: }
1097: MatAssemblyEnd_SeqAIJ_Inode(A,mode);
1098: return(0);
1099: }
1101: PetscErrorCode MatRealPart_SeqAIJ(Mat A)
1102: {
1103: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1104: PetscInt i,nz = a->nz;
1105: MatScalar *aa = a->a;
1109: for (i=0; i<nz; i++) aa[i] = PetscRealPart(aa[i]);
1110: MatSeqAIJInvalidateDiagonal(A);
1111: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
1112: if (A->offloadmask != PETSC_OFFLOAD_UNALLOCATED) A->offloadmask = PETSC_OFFLOAD_CPU;
1113: #endif
1114: return(0);
1115: }
1117: PetscErrorCode MatImaginaryPart_SeqAIJ(Mat A)
1118: {
1119: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1120: PetscInt i,nz = a->nz;
1121: MatScalar *aa = a->a;
1125: for (i=0; i<nz; i++) aa[i] = PetscImaginaryPart(aa[i]);
1126: MatSeqAIJInvalidateDiagonal(A);
1127: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
1128: if (A->offloadmask != PETSC_OFFLOAD_UNALLOCATED) A->offloadmask = PETSC_OFFLOAD_CPU;
1129: #endif
1130: return(0);
1131: }
1133: PetscErrorCode MatZeroEntries_SeqAIJ(Mat A)
1134: {
1135: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1139: PetscArrayzero(a->a,a->i[A->rmap->n]);
1140: MatSeqAIJInvalidateDiagonal(A);
1141: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
1142: if (A->offloadmask != PETSC_OFFLOAD_UNALLOCATED) A->offloadmask = PETSC_OFFLOAD_CPU;
1143: #endif
1144: return(0);
1145: }
1147: PetscErrorCode MatDestroy_SeqAIJ(Mat A)
1148: {
1149: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1153: #if defined(PETSC_USE_LOG)
1154: PetscLogObjectState((PetscObject)A,"Rows=%D, Cols=%D, NZ=%D",A->rmap->n,A->cmap->n,a->nz);
1155: #endif
1156: MatSeqXAIJFreeAIJ(A,&a->a,&a->j,&a->i);
1157: ISDestroy(&a->row);
1158: ISDestroy(&a->col);
1159: PetscFree(a->diag);
1160: PetscFree(a->ibdiag);
1161: PetscFree(a->imax);
1162: PetscFree(a->ilen);
1163: PetscFree(a->ipre);
1164: PetscFree3(a->idiag,a->mdiag,a->ssor_work);
1165: PetscFree(a->solve_work);
1166: ISDestroy(&a->icol);
1167: PetscFree(a->saved_values);
1168: PetscFree2(a->compressedrow.i,a->compressedrow.rindex);
1170: MatDestroy_SeqAIJ_Inode(A);
1171: PetscFree(A->data);
1173: /* MatMatMultNumeric_SeqAIJ_SeqAIJ_Sorted may allocate this.
1174: That function is so heavily used (sometimes in an hidden way through multnumeric function pointers)
1175: that is hard to properly add this data to the MatProduct data. We free it here to avoid
1176: users reusing the matrix object with different data to incur in obscure segmentation faults
1177: due to different matrix sizes */
1178: PetscObjectCompose((PetscObject)A,"__PETSc__ab_dense",NULL);
1180: PetscObjectChangeTypeName((PetscObject)A,0);
1181: PetscObjectComposeFunction((PetscObject)A,"MatSeqAIJSetColumnIndices_C",NULL);
1182: PetscObjectComposeFunction((PetscObject)A,"MatStoreValues_C",NULL);
1183: PetscObjectComposeFunction((PetscObject)A,"MatRetrieveValues_C",NULL);
1184: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqaij_seqsbaij_C",NULL);
1185: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqaij_seqbaij_C",NULL);
1186: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqaij_seqaijperm_C",NULL);
1188: #if defined(PETSC_HAVE_CUDA)
1189: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqaij_seqaijcusparse_C",NULL);
1190: PetscObjectComposeFunction((PetscObject)A,"MatProductSetFromOptions_seqaijcusparse_seqaij_C",NULL);
1191: #endif
1192: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqaij_seqaijcrl_C",NULL);
1193: #if defined(PETSC_HAVE_ELEMENTAL)
1194: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqaij_elemental_C",NULL);
1195: #endif
1196: #if defined(PETSC_HAVE_SCALAPACK)
1197: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqaij_scalapack_C",NULL);
1198: #endif
1199: #if defined(PETSC_HAVE_HYPRE)
1200: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqaij_hypre_C",NULL);
1201: PetscObjectComposeFunction((PetscObject)A,"MatProductSetFromOptions_transpose_seqaij_seqaij_C",NULL);
1202: #endif
1203: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqaij_seqdense_C",NULL);
1204: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqaij_seqsell_C",NULL);
1205: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqaij_is_C",NULL);
1206: PetscObjectComposeFunction((PetscObject)A,"MatIsTranspose_C",NULL);
1207: PetscObjectComposeFunction((PetscObject)A,"MatSeqAIJSetPreallocation_C",NULL);
1208: PetscObjectComposeFunction((PetscObject)A,"MatResetPreallocation_C",NULL);
1209: PetscObjectComposeFunction((PetscObject)A,"MatSeqAIJSetPreallocationCSR_C",NULL);
1210: PetscObjectComposeFunction((PetscObject)A,"MatReorderForNonzeroDiagonal_C",NULL);
1211: PetscObjectComposeFunction((PetscObject)A,"MatProductSetFromOptions_is_seqaij_C",NULL);
1212: PetscObjectComposeFunction((PetscObject)A,"MatProductSetFromOptions_seqdense_seqaij_C",NULL);
1213: PetscObjectComposeFunction((PetscObject)A,"MatProductSetFromOptions_seqaij_seqaij_C",NULL);
1214: return(0);
1215: }
1217: PetscErrorCode MatSetOption_SeqAIJ(Mat A,MatOption op,PetscBool flg)
1218: {
1219: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1223: switch (op) {
1224: case MAT_ROW_ORIENTED:
1225: a->roworiented = flg;
1226: break;
1227: case MAT_KEEP_NONZERO_PATTERN:
1228: a->keepnonzeropattern = flg;
1229: break;
1230: case MAT_NEW_NONZERO_LOCATIONS:
1231: a->nonew = (flg ? 0 : 1);
1232: break;
1233: case MAT_NEW_NONZERO_LOCATION_ERR:
1234: a->nonew = (flg ? -1 : 0);
1235: break;
1236: case MAT_NEW_NONZERO_ALLOCATION_ERR:
1237: a->nonew = (flg ? -2 : 0);
1238: break;
1239: case MAT_UNUSED_NONZERO_LOCATION_ERR:
1240: a->nounused = (flg ? -1 : 0);
1241: break;
1242: case MAT_IGNORE_ZERO_ENTRIES:
1243: a->ignorezeroentries = flg;
1244: break;
1245: case MAT_SPD:
1246: case MAT_SYMMETRIC:
1247: case MAT_STRUCTURALLY_SYMMETRIC:
1248: case MAT_HERMITIAN:
1249: case MAT_SYMMETRY_ETERNAL:
1250: case MAT_STRUCTURE_ONLY:
1251: /* These options are handled directly by MatSetOption() */
1252: break;
1253: case MAT_NEW_DIAGONALS:
1254: case MAT_IGNORE_OFF_PROC_ENTRIES:
1255: case MAT_USE_HASH_TABLE:
1256: PetscInfo1(A,"Option %s ignored\n",MatOptions[op]);
1257: break;
1258: case MAT_USE_INODES:
1259: /* Not an error because MatSetOption_SeqAIJ_Inode handles this one */
1260: break;
1261: case MAT_SUBMAT_SINGLEIS:
1262: A->submat_singleis = flg;
1263: break;
1264: case MAT_SORTED_FULL:
1265: if (flg) A->ops->setvalues = MatSetValues_SeqAIJ_SortedFull;
1266: else A->ops->setvalues = MatSetValues_SeqAIJ;
1267: break;
1268: default:
1269: SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"unknown option %d",op);
1270: }
1271: MatSetOption_SeqAIJ_Inode(A,op,flg);
1272: return(0);
1273: }
1275: PetscErrorCode MatGetDiagonal_SeqAIJ(Mat A,Vec v)
1276: {
1277: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1279: PetscInt i,j,n,*ai=a->i,*aj=a->j;
1280: PetscScalar *aa=a->a,*x;
1283: VecGetLocalSize(v,&n);
1284: if (n != A->rmap->n) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Nonconforming matrix and vector");
1286: if (A->factortype == MAT_FACTOR_ILU || A->factortype == MAT_FACTOR_LU) {
1287: PetscInt *diag=a->diag;
1288: VecGetArrayWrite(v,&x);
1289: for (i=0; i<n; i++) x[i] = 1.0/aa[diag[i]];
1290: VecRestoreArrayWrite(v,&x);
1291: return(0);
1292: }
1294: VecGetArrayWrite(v,&x);
1295: for (i=0; i<n; i++) {
1296: x[i] = 0.0;
1297: for (j=ai[i]; j<ai[i+1]; j++) {
1298: if (aj[j] == i) {
1299: x[i] = aa[j];
1300: break;
1301: }
1302: }
1303: }
1304: VecRestoreArrayWrite(v,&x);
1305: return(0);
1306: }
1308: #include <../src/mat/impls/aij/seq/ftn-kernels/fmult.h>
1309: PetscErrorCode MatMultTransposeAdd_SeqAIJ(Mat A,Vec xx,Vec zz,Vec yy)
1310: {
1311: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1312: PetscScalar *y;
1313: const PetscScalar *x;
1314: PetscErrorCode ierr;
1315: PetscInt m = A->rmap->n;
1316: #if !defined(PETSC_USE_FORTRAN_KERNEL_MULTTRANSPOSEAIJ)
1317: const MatScalar *v;
1318: PetscScalar alpha;
1319: PetscInt n,i,j;
1320: const PetscInt *idx,*ii,*ridx=NULL;
1321: Mat_CompressedRow cprow = a->compressedrow;
1322: PetscBool usecprow = cprow.use;
1323: #endif
1326: if (zz != yy) {VecCopy(zz,yy);}
1327: VecGetArrayRead(xx,&x);
1328: VecGetArray(yy,&y);
1330: #if defined(PETSC_USE_FORTRAN_KERNEL_MULTTRANSPOSEAIJ)
1331: fortranmulttransposeaddaij_(&m,x,a->i,a->j,a->a,y);
1332: #else
1333: if (usecprow) {
1334: m = cprow.nrows;
1335: ii = cprow.i;
1336: ridx = cprow.rindex;
1337: } else {
1338: ii = a->i;
1339: }
1340: for (i=0; i<m; i++) {
1341: idx = a->j + ii[i];
1342: v = a->a + ii[i];
1343: n = ii[i+1] - ii[i];
1344: if (usecprow) {
1345: alpha = x[ridx[i]];
1346: } else {
1347: alpha = x[i];
1348: }
1349: for (j=0; j<n; j++) y[idx[j]] += alpha*v[j];
1350: }
1351: #endif
1352: PetscLogFlops(2.0*a->nz);
1353: VecRestoreArrayRead(xx,&x);
1354: VecRestoreArray(yy,&y);
1355: return(0);
1356: }
1358: PetscErrorCode MatMultTranspose_SeqAIJ(Mat A,Vec xx,Vec yy)
1359: {
1363: VecSet(yy,0.0);
1364: MatMultTransposeAdd_SeqAIJ(A,xx,yy,yy);
1365: return(0);
1366: }
1368: #include <../src/mat/impls/aij/seq/ftn-kernels/fmult.h>
1370: PetscErrorCode MatMult_SeqAIJ(Mat A,Vec xx,Vec yy)
1371: {
1372: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1373: PetscScalar *y;
1374: const PetscScalar *x;
1375: const MatScalar *aa;
1376: PetscErrorCode ierr;
1377: PetscInt m=A->rmap->n;
1378: const PetscInt *aj,*ii,*ridx=NULL;
1379: PetscInt n,i;
1380: PetscScalar sum;
1381: PetscBool usecprow=a->compressedrow.use;
1383: #if defined(PETSC_HAVE_PRAGMA_DISJOINT)
1384: #pragma disjoint(*x,*y,*aa)
1385: #endif
1388: VecGetArrayRead(xx,&x);
1389: VecGetArray(yy,&y);
1390: ii = a->i;
1391: if (usecprow) { /* use compressed row format */
1392: PetscArrayzero(y,m);
1393: m = a->compressedrow.nrows;
1394: ii = a->compressedrow.i;
1395: ridx = a->compressedrow.rindex;
1396: for (i=0; i<m; i++) {
1397: n = ii[i+1] - ii[i];
1398: aj = a->j + ii[i];
1399: aa = a->a + ii[i];
1400: sum = 0.0;
1401: PetscSparseDensePlusDot(sum,x,aa,aj,n);
1402: /* for (j=0; j<n; j++) sum += (*aa++)*x[*aj++]; */
1403: y[*ridx++] = sum;
1404: }
1405: } else { /* do not use compressed row format */
1406: #if defined(PETSC_USE_FORTRAN_KERNEL_MULTAIJ)
1407: aj = a->j;
1408: aa = a->a;
1409: fortranmultaij_(&m,x,ii,aj,aa,y);
1410: #else
1411: for (i=0; i<m; i++) {
1412: n = ii[i+1] - ii[i];
1413: aj = a->j + ii[i];
1414: aa = a->a + ii[i];
1415: sum = 0.0;
1416: PetscSparseDensePlusDot(sum,x,aa,aj,n);
1417: y[i] = sum;
1418: }
1419: #endif
1420: }
1421: PetscLogFlops(2.0*a->nz - a->nonzerorowcnt);
1422: VecRestoreArrayRead(xx,&x);
1423: VecRestoreArray(yy,&y);
1424: return(0);
1425: }
1427: PetscErrorCode MatMultMax_SeqAIJ(Mat A,Vec xx,Vec yy)
1428: {
1429: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1430: PetscScalar *y;
1431: const PetscScalar *x;
1432: const MatScalar *aa;
1433: PetscErrorCode ierr;
1434: PetscInt m=A->rmap->n;
1435: const PetscInt *aj,*ii,*ridx=NULL;
1436: PetscInt n,i,nonzerorow=0;
1437: PetscScalar sum;
1438: PetscBool usecprow=a->compressedrow.use;
1440: #if defined(PETSC_HAVE_PRAGMA_DISJOINT)
1441: #pragma disjoint(*x,*y,*aa)
1442: #endif
1445: VecGetArrayRead(xx,&x);
1446: VecGetArray(yy,&y);
1447: if (usecprow) { /* use compressed row format */
1448: m = a->compressedrow.nrows;
1449: ii = a->compressedrow.i;
1450: ridx = a->compressedrow.rindex;
1451: for (i=0; i<m; i++) {
1452: n = ii[i+1] - ii[i];
1453: aj = a->j + ii[i];
1454: aa = a->a + ii[i];
1455: sum = 0.0;
1456: nonzerorow += (n>0);
1457: PetscSparseDenseMaxDot(sum,x,aa,aj,n);
1458: /* for (j=0; j<n; j++) sum += (*aa++)*x[*aj++]; */
1459: y[*ridx++] = sum;
1460: }
1461: } else { /* do not use compressed row format */
1462: ii = a->i;
1463: for (i=0; i<m; i++) {
1464: n = ii[i+1] - ii[i];
1465: aj = a->j + ii[i];
1466: aa = a->a + ii[i];
1467: sum = 0.0;
1468: nonzerorow += (n>0);
1469: PetscSparseDenseMaxDot(sum,x,aa,aj,n);
1470: y[i] = sum;
1471: }
1472: }
1473: PetscLogFlops(2.0*a->nz - nonzerorow);
1474: VecRestoreArrayRead(xx,&x);
1475: VecRestoreArray(yy,&y);
1476: return(0);
1477: }
1479: PetscErrorCode MatMultAddMax_SeqAIJ(Mat A,Vec xx,Vec yy,Vec zz)
1480: {
1481: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1482: PetscScalar *y,*z;
1483: const PetscScalar *x;
1484: const MatScalar *aa;
1485: PetscErrorCode ierr;
1486: PetscInt m = A->rmap->n,*aj,*ii;
1487: PetscInt n,i,*ridx=NULL;
1488: PetscScalar sum;
1489: PetscBool usecprow=a->compressedrow.use;
1492: VecGetArrayRead(xx,&x);
1493: VecGetArrayPair(yy,zz,&y,&z);
1494: if (usecprow) { /* use compressed row format */
1495: if (zz != yy) {
1496: PetscArraycpy(z,y,m);
1497: }
1498: m = a->compressedrow.nrows;
1499: ii = a->compressedrow.i;
1500: ridx = a->compressedrow.rindex;
1501: for (i=0; i<m; i++) {
1502: n = ii[i+1] - ii[i];
1503: aj = a->j + ii[i];
1504: aa = a->a + ii[i];
1505: sum = y[*ridx];
1506: PetscSparseDenseMaxDot(sum,x,aa,aj,n);
1507: z[*ridx++] = sum;
1508: }
1509: } else { /* do not use compressed row format */
1510: ii = a->i;
1511: for (i=0; i<m; i++) {
1512: n = ii[i+1] - ii[i];
1513: aj = a->j + ii[i];
1514: aa = a->a + ii[i];
1515: sum = y[i];
1516: PetscSparseDenseMaxDot(sum,x,aa,aj,n);
1517: z[i] = sum;
1518: }
1519: }
1520: PetscLogFlops(2.0*a->nz);
1521: VecRestoreArrayRead(xx,&x);
1522: VecRestoreArrayPair(yy,zz,&y,&z);
1523: return(0);
1524: }
1526: #include <../src/mat/impls/aij/seq/ftn-kernels/fmultadd.h>
1527: PetscErrorCode MatMultAdd_SeqAIJ(Mat A,Vec xx,Vec yy,Vec zz)
1528: {
1529: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1530: PetscScalar *y,*z;
1531: const PetscScalar *x;
1532: const MatScalar *aa;
1533: PetscErrorCode ierr;
1534: const PetscInt *aj,*ii,*ridx=NULL;
1535: PetscInt m = A->rmap->n,n,i;
1536: PetscScalar sum;
1537: PetscBool usecprow=a->compressedrow.use;
1540: VecGetArrayRead(xx,&x);
1541: VecGetArrayPair(yy,zz,&y,&z);
1542: if (usecprow) { /* use compressed row format */
1543: if (zz != yy) {
1544: PetscArraycpy(z,y,m);
1545: }
1546: m = a->compressedrow.nrows;
1547: ii = a->compressedrow.i;
1548: ridx = a->compressedrow.rindex;
1549: for (i=0; i<m; i++) {
1550: n = ii[i+1] - ii[i];
1551: aj = a->j + ii[i];
1552: aa = a->a + ii[i];
1553: sum = y[*ridx];
1554: PetscSparseDensePlusDot(sum,x,aa,aj,n);
1555: z[*ridx++] = sum;
1556: }
1557: } else { /* do not use compressed row format */
1558: ii = a->i;
1559: #if defined(PETSC_USE_FORTRAN_KERNEL_MULTADDAIJ)
1560: aj = a->j;
1561: aa = a->a;
1562: fortranmultaddaij_(&m,x,ii,aj,aa,y,z);
1563: #else
1564: for (i=0; i<m; i++) {
1565: n = ii[i+1] - ii[i];
1566: aj = a->j + ii[i];
1567: aa = a->a + ii[i];
1568: sum = y[i];
1569: PetscSparseDensePlusDot(sum,x,aa,aj,n);
1570: z[i] = sum;
1571: }
1572: #endif
1573: }
1574: PetscLogFlops(2.0*a->nz);
1575: VecRestoreArrayRead(xx,&x);
1576: VecRestoreArrayPair(yy,zz,&y,&z);
1577: return(0);
1578: }
1580: /*
1581: Adds diagonal pointers to sparse matrix structure.
1582: */
1583: PetscErrorCode MatMarkDiagonal_SeqAIJ(Mat A)
1584: {
1585: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1587: PetscInt i,j,m = A->rmap->n;
1590: if (!a->diag) {
1591: PetscMalloc1(m,&a->diag);
1592: PetscLogObjectMemory((PetscObject)A, m*sizeof(PetscInt));
1593: }
1594: for (i=0; i<A->rmap->n; i++) {
1595: a->diag[i] = a->i[i+1];
1596: for (j=a->i[i]; j<a->i[i+1]; j++) {
1597: if (a->j[j] == i) {
1598: a->diag[i] = j;
1599: break;
1600: }
1601: }
1602: }
1603: return(0);
1604: }
1606: PetscErrorCode MatShift_SeqAIJ(Mat A,PetscScalar v)
1607: {
1608: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1609: const PetscInt *diag = (const PetscInt*)a->diag;
1610: const PetscInt *ii = (const PetscInt*) a->i;
1611: PetscInt i,*mdiag = NULL;
1612: PetscErrorCode ierr;
1613: PetscInt cnt = 0; /* how many diagonals are missing */
1616: if (!A->preallocated || !a->nz) {
1617: MatSeqAIJSetPreallocation(A,1,NULL);
1618: MatShift_Basic(A,v);
1619: return(0);
1620: }
1622: if (a->diagonaldense) {
1623: cnt = 0;
1624: } else {
1625: PetscCalloc1(A->rmap->n,&mdiag);
1626: for (i=0; i<A->rmap->n; i++) {
1627: if (diag[i] >= ii[i+1]) {
1628: cnt++;
1629: mdiag[i] = 1;
1630: }
1631: }
1632: }
1633: if (!cnt) {
1634: MatShift_Basic(A,v);
1635: } else {
1636: PetscScalar *olda = a->a; /* preserve pointers to current matrix nonzeros structure and values */
1637: PetscInt *oldj = a->j, *oldi = a->i;
1638: PetscBool singlemalloc = a->singlemalloc,free_a = a->free_a,free_ij = a->free_ij;
1640: a->a = NULL;
1641: a->j = NULL;
1642: a->i = NULL;
1643: /* increase the values in imax for each row where a diagonal is being inserted then reallocate the matrix data structures */
1644: for (i=0; i<A->rmap->n; i++) {
1645: a->imax[i] += mdiag[i];
1646: a->imax[i] = PetscMin(a->imax[i],A->cmap->n);
1647: }
1648: MatSeqAIJSetPreallocation_SeqAIJ(A,0,a->imax);
1650: /* copy old values into new matrix data structure */
1651: for (i=0; i<A->rmap->n; i++) {
1652: MatSetValues(A,1,&i,a->imax[i] - mdiag[i],&oldj[oldi[i]],&olda[oldi[i]],ADD_VALUES);
1653: if (i < A->cmap->n) {
1654: MatSetValue(A,i,i,v,ADD_VALUES);
1655: }
1656: }
1657: MatAssemblyBegin(A,MAT_FINAL_ASSEMBLY);
1658: MatAssemblyEnd(A,MAT_FINAL_ASSEMBLY);
1659: if (singlemalloc) {
1660: PetscFree3(olda,oldj,oldi);
1661: } else {
1662: if (free_a) {PetscFree(olda);}
1663: if (free_ij) {PetscFree(oldj);}
1664: if (free_ij) {PetscFree(oldi);}
1665: }
1666: }
1667: PetscFree(mdiag);
1668: a->diagonaldense = PETSC_TRUE;
1669: return(0);
1670: }
1672: /*
1673: Checks for missing diagonals
1674: */
1675: PetscErrorCode MatMissingDiagonal_SeqAIJ(Mat A,PetscBool *missing,PetscInt *d)
1676: {
1677: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1678: PetscInt *diag,*ii = a->i,i;
1682: *missing = PETSC_FALSE;
1683: if (A->rmap->n > 0 && !ii) {
1684: *missing = PETSC_TRUE;
1685: if (d) *d = 0;
1686: PetscInfo(A,"Matrix has no entries therefore is missing diagonal\n");
1687: } else {
1688: PetscInt n;
1689: n = PetscMin(A->rmap->n, A->cmap->n);
1690: diag = a->diag;
1691: for (i=0; i<n; i++) {
1692: if (diag[i] >= ii[i+1]) {
1693: *missing = PETSC_TRUE;
1694: if (d) *d = i;
1695: PetscInfo1(A,"Matrix is missing diagonal number %D\n",i);
1696: break;
1697: }
1698: }
1699: }
1700: return(0);
1701: }
1703: #include <petscblaslapack.h>
1704: #include <petsc/private/kernels/blockinvert.h>
1706: /*
1707: Note that values is allocated externally by the PC and then passed into this routine
1708: */
1709: PetscErrorCode MatInvertVariableBlockDiagonal_SeqAIJ(Mat A,PetscInt nblocks,const PetscInt *bsizes,PetscScalar *diag)
1710: {
1711: PetscErrorCode ierr;
1712: PetscInt n = A->rmap->n, i, ncnt = 0, *indx,j,bsizemax = 0,*v_pivots;
1713: PetscBool allowzeropivot,zeropivotdetected=PETSC_FALSE;
1714: const PetscReal shift = 0.0;
1715: PetscInt ipvt[5];
1716: PetscScalar work[25],*v_work;
1719: allowzeropivot = PetscNot(A->erroriffailure);
1720: for (i=0; i<nblocks; i++) ncnt += bsizes[i];
1721: if (ncnt != n) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Total blocksizes %D doesn't match number matrix rows %D",ncnt,n);
1722: for (i=0; i<nblocks; i++) {
1723: bsizemax = PetscMax(bsizemax,bsizes[i]);
1724: }
1725: PetscMalloc1(bsizemax,&indx);
1726: if (bsizemax > 7) {
1727: PetscMalloc2(bsizemax,&v_work,bsizemax,&v_pivots);
1728: }
1729: ncnt = 0;
1730: for (i=0; i<nblocks; i++) {
1731: for (j=0; j<bsizes[i]; j++) indx[j] = ncnt+j;
1732: MatGetValues(A,bsizes[i],indx,bsizes[i],indx,diag);
1733: switch (bsizes[i]) {
1734: case 1:
1735: *diag = 1.0/(*diag);
1736: break;
1737: case 2:
1738: PetscKernel_A_gets_inverse_A_2(diag,shift,allowzeropivot,&zeropivotdetected);
1739: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
1740: PetscKernel_A_gets_transpose_A_2(diag);
1741: break;
1742: case 3:
1743: PetscKernel_A_gets_inverse_A_3(diag,shift,allowzeropivot,&zeropivotdetected);
1744: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
1745: PetscKernel_A_gets_transpose_A_3(diag);
1746: break;
1747: case 4:
1748: PetscKernel_A_gets_inverse_A_4(diag,shift,allowzeropivot,&zeropivotdetected);
1749: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
1750: PetscKernel_A_gets_transpose_A_4(diag);
1751: break;
1752: case 5:
1753: PetscKernel_A_gets_inverse_A_5(diag,ipvt,work,shift,allowzeropivot,&zeropivotdetected);
1754: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
1755: PetscKernel_A_gets_transpose_A_5(diag);
1756: break;
1757: case 6:
1758: PetscKernel_A_gets_inverse_A_6(diag,shift,allowzeropivot,&zeropivotdetected);
1759: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
1760: PetscKernel_A_gets_transpose_A_6(diag);
1761: break;
1762: case 7:
1763: PetscKernel_A_gets_inverse_A_7(diag,shift,allowzeropivot,&zeropivotdetected);
1764: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
1765: PetscKernel_A_gets_transpose_A_7(diag);
1766: break;
1767: default:
1768: PetscKernel_A_gets_inverse_A(bsizes[i],diag,v_pivots,v_work,allowzeropivot,&zeropivotdetected);
1769: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
1770: PetscKernel_A_gets_transpose_A_N(diag,bsizes[i]);
1771: }
1772: ncnt += bsizes[i];
1773: diag += bsizes[i]*bsizes[i];
1774: }
1775: if (bsizemax > 7) {
1776: PetscFree2(v_work,v_pivots);
1777: }
1778: PetscFree(indx);
1779: return(0);
1780: }
1782: /*
1783: Negative shift indicates do not generate an error if there is a zero diagonal, just invert it anyways
1784: */
1785: PetscErrorCode MatInvertDiagonal_SeqAIJ(Mat A,PetscScalar omega,PetscScalar fshift)
1786: {
1787: Mat_SeqAIJ *a = (Mat_SeqAIJ*) A->data;
1789: PetscInt i,*diag,m = A->rmap->n;
1790: MatScalar *v = a->a;
1791: PetscScalar *idiag,*mdiag;
1794: if (a->idiagvalid) return(0);
1795: MatMarkDiagonal_SeqAIJ(A);
1796: diag = a->diag;
1797: if (!a->idiag) {
1798: PetscMalloc3(m,&a->idiag,m,&a->mdiag,m,&a->ssor_work);
1799: PetscLogObjectMemory((PetscObject)A, 3*m*sizeof(PetscScalar));
1800: v = a->a;
1801: }
1802: mdiag = a->mdiag;
1803: idiag = a->idiag;
1805: if (omega == 1.0 && PetscRealPart(fshift) <= 0.0) {
1806: for (i=0; i<m; i++) {
1807: mdiag[i] = v[diag[i]];
1808: if (!PetscAbsScalar(mdiag[i])) { /* zero diagonal */
1809: if (PetscRealPart(fshift)) {
1810: PetscInfo1(A,"Zero diagonal on row %D\n",i);
1811: A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
1812: A->factorerror_zeropivot_value = 0.0;
1813: A->factorerror_zeropivot_row = i;
1814: } else SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_INCOMP,"Zero diagonal on row %D",i);
1815: }
1816: idiag[i] = 1.0/v[diag[i]];
1817: }
1818: PetscLogFlops(m);
1819: } else {
1820: for (i=0; i<m; i++) {
1821: mdiag[i] = v[diag[i]];
1822: idiag[i] = omega/(fshift + v[diag[i]]);
1823: }
1824: PetscLogFlops(2.0*m);
1825: }
1826: a->idiagvalid = PETSC_TRUE;
1827: return(0);
1828: }
1830: #include <../src/mat/impls/aij/seq/ftn-kernels/frelax.h>
1831: PetscErrorCode MatSOR_SeqAIJ(Mat A,Vec bb,PetscReal omega,MatSORType flag,PetscReal fshift,PetscInt its,PetscInt lits,Vec xx)
1832: {
1833: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1834: PetscScalar *x,d,sum,*t,scale;
1835: const MatScalar *v,*idiag=0,*mdiag;
1836: const PetscScalar *b, *bs,*xb, *ts;
1837: PetscErrorCode ierr;
1838: PetscInt n,m = A->rmap->n,i;
1839: const PetscInt *idx,*diag;
1842: its = its*lits;
1844: if (fshift != a->fshift || omega != a->omega) a->idiagvalid = PETSC_FALSE; /* must recompute idiag[] */
1845: if (!a->idiagvalid) {MatInvertDiagonal_SeqAIJ(A,omega,fshift);}
1846: a->fshift = fshift;
1847: a->omega = omega;
1849: diag = a->diag;
1850: t = a->ssor_work;
1851: idiag = a->idiag;
1852: mdiag = a->mdiag;
1854: VecGetArray(xx,&x);
1855: VecGetArrayRead(bb,&b);
1856: /* We count flops by assuming the upper triangular and lower triangular parts have the same number of nonzeros */
1857: if (flag == SOR_APPLY_UPPER) {
1858: /* apply (U + D/omega) to the vector */
1859: bs = b;
1860: for (i=0; i<m; i++) {
1861: d = fshift + mdiag[i];
1862: n = a->i[i+1] - diag[i] - 1;
1863: idx = a->j + diag[i] + 1;
1864: v = a->a + diag[i] + 1;
1865: sum = b[i]*d/omega;
1866: PetscSparseDensePlusDot(sum,bs,v,idx,n);
1867: x[i] = sum;
1868: }
1869: VecRestoreArray(xx,&x);
1870: VecRestoreArrayRead(bb,&b);
1871: PetscLogFlops(a->nz);
1872: return(0);
1873: }
1875: if (flag == SOR_APPLY_LOWER) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"SOR_APPLY_LOWER is not implemented");
1876: else if (flag & SOR_EISENSTAT) {
1877: /* Let A = L + U + D; where L is lower triangular,
1878: U is upper triangular, E = D/omega; This routine applies
1880: (L + E)^{-1} A (U + E)^{-1}
1882: to a vector efficiently using Eisenstat's trick.
1883: */
1884: scale = (2.0/omega) - 1.0;
1886: /* x = (E + U)^{-1} b */
1887: for (i=m-1; i>=0; i--) {
1888: n = a->i[i+1] - diag[i] - 1;
1889: idx = a->j + diag[i] + 1;
1890: v = a->a + diag[i] + 1;
1891: sum = b[i];
1892: PetscSparseDenseMinusDot(sum,x,v,idx,n);
1893: x[i] = sum*idiag[i];
1894: }
1896: /* t = b - (2*E - D)x */
1897: v = a->a;
1898: for (i=0; i<m; i++) t[i] = b[i] - scale*(v[*diag++])*x[i];
1900: /* t = (E + L)^{-1}t */
1901: ts = t;
1902: diag = a->diag;
1903: for (i=0; i<m; i++) {
1904: n = diag[i] - a->i[i];
1905: idx = a->j + a->i[i];
1906: v = a->a + a->i[i];
1907: sum = t[i];
1908: PetscSparseDenseMinusDot(sum,ts,v,idx,n);
1909: t[i] = sum*idiag[i];
1910: /* x = x + t */
1911: x[i] += t[i];
1912: }
1914: PetscLogFlops(6.0*m-1 + 2.0*a->nz);
1915: VecRestoreArray(xx,&x);
1916: VecRestoreArrayRead(bb,&b);
1917: return(0);
1918: }
1919: if (flag & SOR_ZERO_INITIAL_GUESS) {
1920: if (flag & SOR_FORWARD_SWEEP || flag & SOR_LOCAL_FORWARD_SWEEP) {
1921: for (i=0; i<m; i++) {
1922: n = diag[i] - a->i[i];
1923: idx = a->j + a->i[i];
1924: v = a->a + a->i[i];
1925: sum = b[i];
1926: PetscSparseDenseMinusDot(sum,x,v,idx,n);
1927: t[i] = sum;
1928: x[i] = sum*idiag[i];
1929: }
1930: xb = t;
1931: PetscLogFlops(a->nz);
1932: } else xb = b;
1933: if (flag & SOR_BACKWARD_SWEEP || flag & SOR_LOCAL_BACKWARD_SWEEP) {
1934: for (i=m-1; i>=0; i--) {
1935: n = a->i[i+1] - diag[i] - 1;
1936: idx = a->j + diag[i] + 1;
1937: v = a->a + diag[i] + 1;
1938: sum = xb[i];
1939: PetscSparseDenseMinusDot(sum,x,v,idx,n);
1940: if (xb == b) {
1941: x[i] = sum*idiag[i];
1942: } else {
1943: x[i] = (1-omega)*x[i] + sum*idiag[i]; /* omega in idiag */
1944: }
1945: }
1946: PetscLogFlops(a->nz); /* assumes 1/2 in upper */
1947: }
1948: its--;
1949: }
1950: while (its--) {
1951: if (flag & SOR_FORWARD_SWEEP || flag & SOR_LOCAL_FORWARD_SWEEP) {
1952: for (i=0; i<m; i++) {
1953: /* lower */
1954: n = diag[i] - a->i[i];
1955: idx = a->j + a->i[i];
1956: v = a->a + a->i[i];
1957: sum = b[i];
1958: PetscSparseDenseMinusDot(sum,x,v,idx,n);
1959: t[i] = sum; /* save application of the lower-triangular part */
1960: /* upper */
1961: n = a->i[i+1] - diag[i] - 1;
1962: idx = a->j + diag[i] + 1;
1963: v = a->a + diag[i] + 1;
1964: PetscSparseDenseMinusDot(sum,x,v,idx,n);
1965: x[i] = (1. - omega)*x[i] + sum*idiag[i]; /* omega in idiag */
1966: }
1967: xb = t;
1968: PetscLogFlops(2.0*a->nz);
1969: } else xb = b;
1970: if (flag & SOR_BACKWARD_SWEEP || flag & SOR_LOCAL_BACKWARD_SWEEP) {
1971: for (i=m-1; i>=0; i--) {
1972: sum = xb[i];
1973: if (xb == b) {
1974: /* whole matrix (no checkpointing available) */
1975: n = a->i[i+1] - a->i[i];
1976: idx = a->j + a->i[i];
1977: v = a->a + a->i[i];
1978: PetscSparseDenseMinusDot(sum,x,v,idx,n);
1979: x[i] = (1. - omega)*x[i] + (sum + mdiag[i]*x[i])*idiag[i];
1980: } else { /* lower-triangular part has been saved, so only apply upper-triangular */
1981: n = a->i[i+1] - diag[i] - 1;
1982: idx = a->j + diag[i] + 1;
1983: v = a->a + diag[i] + 1;
1984: PetscSparseDenseMinusDot(sum,x,v,idx,n);
1985: x[i] = (1. - omega)*x[i] + sum*idiag[i]; /* omega in idiag */
1986: }
1987: }
1988: if (xb == b) {
1989: PetscLogFlops(2.0*a->nz);
1990: } else {
1991: PetscLogFlops(a->nz); /* assumes 1/2 in upper */
1992: }
1993: }
1994: }
1995: VecRestoreArray(xx,&x);
1996: VecRestoreArrayRead(bb,&b);
1997: return(0);
1998: }
2001: PetscErrorCode MatGetInfo_SeqAIJ(Mat A,MatInfoType flag,MatInfo *info)
2002: {
2003: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
2006: info->block_size = 1.0;
2007: info->nz_allocated = a->maxnz;
2008: info->nz_used = a->nz;
2009: info->nz_unneeded = (a->maxnz - a->nz);
2010: info->assemblies = A->num_ass;
2011: info->mallocs = A->info.mallocs;
2012: info->memory = ((PetscObject)A)->mem;
2013: if (A->factortype) {
2014: info->fill_ratio_given = A->info.fill_ratio_given;
2015: info->fill_ratio_needed = A->info.fill_ratio_needed;
2016: info->factor_mallocs = A->info.factor_mallocs;
2017: } else {
2018: info->fill_ratio_given = 0;
2019: info->fill_ratio_needed = 0;
2020: info->factor_mallocs = 0;
2021: }
2022: return(0);
2023: }
2025: PetscErrorCode MatZeroRows_SeqAIJ(Mat A,PetscInt N,const PetscInt rows[],PetscScalar diag,Vec x,Vec b)
2026: {
2027: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
2028: PetscInt i,m = A->rmap->n - 1;
2029: PetscErrorCode ierr;
2030: const PetscScalar *xx;
2031: PetscScalar *bb;
2032: PetscInt d = 0;
2035: if (x && b) {
2036: VecGetArrayRead(x,&xx);
2037: VecGetArray(b,&bb);
2038: for (i=0; i<N; i++) {
2039: if (rows[i] < 0 || rows[i] > m) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"row %D out of range", rows[i]);
2040: if (rows[i] >= A->cmap->n) continue;
2041: bb[rows[i]] = diag*xx[rows[i]];
2042: }
2043: VecRestoreArrayRead(x,&xx);
2044: VecRestoreArray(b,&bb);
2045: }
2047: if (a->keepnonzeropattern) {
2048: for (i=0; i<N; i++) {
2049: if (rows[i] < 0 || rows[i] > m) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"row %D out of range", rows[i]);
2050: PetscArrayzero(&a->a[a->i[rows[i]]],a->ilen[rows[i]]);
2051: }
2052: if (diag != 0.0) {
2053: for (i=0; i<N; i++) {
2054: d = rows[i];
2055: if (rows[i] >= A->cmap->n) continue;
2056: 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);
2057: }
2058: for (i=0; i<N; i++) {
2059: if (rows[i] >= A->cmap->n) continue;
2060: a->a[a->diag[rows[i]]] = diag;
2061: }
2062: }
2063: } else {
2064: if (diag != 0.0) {
2065: for (i=0; i<N; i++) {
2066: if (rows[i] < 0 || rows[i] > m) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"row %D out of range", rows[i]);
2067: if (a->ilen[rows[i]] > 0) {
2068: if (rows[i] >= A->cmap->n) {
2069: a->ilen[rows[i]] = 0;
2070: } else {
2071: a->ilen[rows[i]] = 1;
2072: a->a[a->i[rows[i]]] = diag;
2073: a->j[a->i[rows[i]]] = rows[i];
2074: }
2075: } else if (rows[i] < A->cmap->n) { /* in case row was completely empty */
2076: MatSetValues_SeqAIJ(A,1,&rows[i],1,&rows[i],&diag,INSERT_VALUES);
2077: }
2078: }
2079: } else {
2080: for (i=0; i<N; i++) {
2081: if (rows[i] < 0 || rows[i] > m) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"row %D out of range", rows[i]);
2082: a->ilen[rows[i]] = 0;
2083: }
2084: }
2085: A->nonzerostate++;
2086: }
2087: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
2088: if (A->offloadmask != PETSC_OFFLOAD_UNALLOCATED) A->offloadmask = PETSC_OFFLOAD_CPU;
2089: #endif
2090: (*A->ops->assemblyend)(A,MAT_FINAL_ASSEMBLY);
2091: return(0);
2092: }
2094: PetscErrorCode MatZeroRowsColumns_SeqAIJ(Mat A,PetscInt N,const PetscInt rows[],PetscScalar diag,Vec x,Vec b)
2095: {
2096: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
2097: PetscInt i,j,m = A->rmap->n - 1,d = 0;
2098: PetscErrorCode ierr;
2099: PetscBool missing,*zeroed,vecs = PETSC_FALSE;
2100: const PetscScalar *xx;
2101: PetscScalar *bb;
2104: if (x && b) {
2105: VecGetArrayRead(x,&xx);
2106: VecGetArray(b,&bb);
2107: vecs = PETSC_TRUE;
2108: }
2109: PetscCalloc1(A->rmap->n,&zeroed);
2110: for (i=0; i<N; i++) {
2111: if (rows[i] < 0 || rows[i] > m) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"row %D out of range", rows[i]);
2112: PetscArrayzero(&a->a[a->i[rows[i]]],a->ilen[rows[i]]);
2114: zeroed[rows[i]] = PETSC_TRUE;
2115: }
2116: for (i=0; i<A->rmap->n; i++) {
2117: if (!zeroed[i]) {
2118: for (j=a->i[i]; j<a->i[i+1]; j++) {
2119: if (a->j[j] < A->rmap->n && zeroed[a->j[j]]) {
2120: if (vecs) bb[i] -= a->a[j]*xx[a->j[j]];
2121: a->a[j] = 0.0;
2122: }
2123: }
2124: } else if (vecs && i < A->cmap->N) bb[i] = diag*xx[i];
2125: }
2126: if (x && b) {
2127: VecRestoreArrayRead(x,&xx);
2128: VecRestoreArray(b,&bb);
2129: }
2130: PetscFree(zeroed);
2131: if (diag != 0.0) {
2132: MatMissingDiagonal_SeqAIJ(A,&missing,&d);
2133: if (missing) {
2134: for (i=0; i<N; i++) {
2135: if (rows[i] >= A->cmap->N) continue;
2136: if (a->nonew && rows[i] >= d) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Matrix is missing diagonal entry in row %D (%D)",d,rows[i]);
2137: MatSetValues_SeqAIJ(A,1,&rows[i],1,&rows[i],&diag,INSERT_VALUES);
2138: }
2139: } else {
2140: for (i=0; i<N; i++) {
2141: a->a[a->diag[rows[i]]] = diag;
2142: }
2143: }
2144: }
2145: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
2146: if (A->offloadmask != PETSC_OFFLOAD_UNALLOCATED) A->offloadmask = PETSC_OFFLOAD_CPU;
2147: #endif
2148: (*A->ops->assemblyend)(A,MAT_FINAL_ASSEMBLY);
2149: return(0);
2150: }
2152: PetscErrorCode MatGetRow_SeqAIJ(Mat A,PetscInt row,PetscInt *nz,PetscInt **idx,PetscScalar **v)
2153: {
2154: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
2155: PetscInt *itmp;
2158: if (row < 0 || row >= A->rmap->n) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Row %D out of range",row);
2160: *nz = a->i[row+1] - a->i[row];
2161: if (v) *v = a->a + a->i[row];
2162: if (idx) {
2163: itmp = a->j + a->i[row];
2164: if (*nz) *idx = itmp;
2165: else *idx = 0;
2166: }
2167: return(0);
2168: }
2170: /* remove this function? */
2171: PetscErrorCode MatRestoreRow_SeqAIJ(Mat A,PetscInt row,PetscInt *nz,PetscInt **idx,PetscScalar **v)
2172: {
2174: return(0);
2175: }
2177: PetscErrorCode MatNorm_SeqAIJ(Mat A,NormType type,PetscReal *nrm)
2178: {
2179: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
2180: MatScalar *v = a->a;
2181: PetscReal sum = 0.0;
2183: PetscInt i,j;
2186: if (type == NORM_FROBENIUS) {
2187: #if defined(PETSC_USE_REAL___FP16)
2188: PetscBLASInt one = 1,nz = a->nz;
2189: *nrm = BLASnrm2_(&nz,v,&one);
2190: #else
2191: for (i=0; i<a->nz; i++) {
2192: sum += PetscRealPart(PetscConj(*v)*(*v)); v++;
2193: }
2194: *nrm = PetscSqrtReal(sum);
2195: #endif
2196: PetscLogFlops(2*a->nz);
2197: } else if (type == NORM_1) {
2198: PetscReal *tmp;
2199: PetscInt *jj = a->j;
2200: PetscCalloc1(A->cmap->n+1,&tmp);
2201: *nrm = 0.0;
2202: for (j=0; j<a->nz; j++) {
2203: tmp[*jj++] += PetscAbsScalar(*v); v++;
2204: }
2205: for (j=0; j<A->cmap->n; j++) {
2206: if (tmp[j] > *nrm) *nrm = tmp[j];
2207: }
2208: PetscFree(tmp);
2209: PetscLogFlops(PetscMax(a->nz-1,0));
2210: } else if (type == NORM_INFINITY) {
2211: *nrm = 0.0;
2212: for (j=0; j<A->rmap->n; j++) {
2213: v = a->a + a->i[j];
2214: sum = 0.0;
2215: for (i=0; i<a->i[j+1]-a->i[j]; i++) {
2216: sum += PetscAbsScalar(*v); v++;
2217: }
2218: if (sum > *nrm) *nrm = sum;
2219: }
2220: PetscLogFlops(PetscMax(a->nz-1,0));
2221: } else SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"No support for two norm");
2222: return(0);
2223: }
2225: /* Merged from MatGetSymbolicTranspose_SeqAIJ() - replace MatGetSymbolicTranspose_SeqAIJ()? */
2226: PetscErrorCode MatTransposeSymbolic_SeqAIJ(Mat A,Mat *B)
2227: {
2229: PetscInt i,j,anzj;
2230: Mat_SeqAIJ *a=(Mat_SeqAIJ*)A->data,*b;
2231: PetscInt an=A->cmap->N,am=A->rmap->N;
2232: PetscInt *ati,*atj,*atfill,*ai=a->i,*aj=a->j;
2235: /* Allocate space for symbolic transpose info and work array */
2236: PetscCalloc1(an+1,&ati);
2237: PetscMalloc1(ai[am],&atj);
2238: PetscMalloc1(an,&atfill);
2240: /* Walk through aj and count ## of non-zeros in each row of A^T. */
2241: /* Note: offset by 1 for fast conversion into csr format. */
2242: for (i=0;i<ai[am];i++) ati[aj[i]+1] += 1;
2243: /* Form ati for csr format of A^T. */
2244: for (i=0;i<an;i++) ati[i+1] += ati[i];
2246: /* Copy ati into atfill so we have locations of the next free space in atj */
2247: PetscArraycpy(atfill,ati,an);
2249: /* Walk through A row-wise and mark nonzero entries of A^T. */
2250: for (i=0;i<am;i++) {
2251: anzj = ai[i+1] - ai[i];
2252: for (j=0;j<anzj;j++) {
2253: atj[atfill[*aj]] = i;
2254: atfill[*aj++] += 1;
2255: }
2256: }
2258: /* Clean up temporary space and complete requests. */
2259: PetscFree(atfill);
2260: MatCreateSeqAIJWithArrays(PetscObjectComm((PetscObject)A),an,am,ati,atj,NULL,B);
2261: MatSetBlockSizes(*B,PetscAbs(A->cmap->bs),PetscAbs(A->rmap->bs));
2262: MatSetType(*B,((PetscObject)A)->type_name);
2264: b = (Mat_SeqAIJ*)((*B)->data);
2265: b->free_a = PETSC_FALSE;
2266: b->free_ij = PETSC_TRUE;
2267: b->nonew = 0;
2268: return(0);
2269: }
2271: PetscErrorCode MatIsTranspose_SeqAIJ(Mat A,Mat B,PetscReal tol,PetscBool *f)
2272: {
2273: Mat_SeqAIJ *aij = (Mat_SeqAIJ*) A->data,*bij = (Mat_SeqAIJ*) B->data;
2274: PetscInt *adx,*bdx,*aii,*bii,*aptr,*bptr;
2275: MatScalar *va,*vb;
2277: PetscInt ma,na,mb,nb, i;
2280: MatGetSize(A,&ma,&na);
2281: MatGetSize(B,&mb,&nb);
2282: if (ma!=nb || na!=mb) {
2283: *f = PETSC_FALSE;
2284: return(0);
2285: }
2286: aii = aij->i; bii = bij->i;
2287: adx = aij->j; bdx = bij->j;
2288: va = aij->a; vb = bij->a;
2289: PetscMalloc1(ma,&aptr);
2290: PetscMalloc1(mb,&bptr);
2291: for (i=0; i<ma; i++) aptr[i] = aii[i];
2292: for (i=0; i<mb; i++) bptr[i] = bii[i];
2294: *f = PETSC_TRUE;
2295: for (i=0; i<ma; i++) {
2296: while (aptr[i]<aii[i+1]) {
2297: PetscInt idc,idr;
2298: PetscScalar vc,vr;
2299: /* column/row index/value */
2300: idc = adx[aptr[i]];
2301: idr = bdx[bptr[idc]];
2302: vc = va[aptr[i]];
2303: vr = vb[bptr[idc]];
2304: if (i!=idr || PetscAbsScalar(vc-vr) > tol) {
2305: *f = PETSC_FALSE;
2306: goto done;
2307: } else {
2308: aptr[i]++;
2309: if (B || i!=idc) bptr[idc]++;
2310: }
2311: }
2312: }
2313: done:
2314: PetscFree(aptr);
2315: PetscFree(bptr);
2316: return(0);
2317: }
2319: PetscErrorCode MatIsHermitianTranspose_SeqAIJ(Mat A,Mat B,PetscReal tol,PetscBool *f)
2320: {
2321: Mat_SeqAIJ *aij = (Mat_SeqAIJ*) A->data,*bij = (Mat_SeqAIJ*) B->data;
2322: PetscInt *adx,*bdx,*aii,*bii,*aptr,*bptr;
2323: MatScalar *va,*vb;
2325: PetscInt ma,na,mb,nb, i;
2328: MatGetSize(A,&ma,&na);
2329: MatGetSize(B,&mb,&nb);
2330: if (ma!=nb || na!=mb) {
2331: *f = PETSC_FALSE;
2332: return(0);
2333: }
2334: aii = aij->i; bii = bij->i;
2335: adx = aij->j; bdx = bij->j;
2336: va = aij->a; vb = bij->a;
2337: PetscMalloc1(ma,&aptr);
2338: PetscMalloc1(mb,&bptr);
2339: for (i=0; i<ma; i++) aptr[i] = aii[i];
2340: for (i=0; i<mb; i++) bptr[i] = bii[i];
2342: *f = PETSC_TRUE;
2343: for (i=0; i<ma; i++) {
2344: while (aptr[i]<aii[i+1]) {
2345: PetscInt idc,idr;
2346: PetscScalar vc,vr;
2347: /* column/row index/value */
2348: idc = adx[aptr[i]];
2349: idr = bdx[bptr[idc]];
2350: vc = va[aptr[i]];
2351: vr = vb[bptr[idc]];
2352: if (i!=idr || PetscAbsScalar(vc-PetscConj(vr)) > tol) {
2353: *f = PETSC_FALSE;
2354: goto done;
2355: } else {
2356: aptr[i]++;
2357: if (B || i!=idc) bptr[idc]++;
2358: }
2359: }
2360: }
2361: done:
2362: PetscFree(aptr);
2363: PetscFree(bptr);
2364: return(0);
2365: }
2367: PetscErrorCode MatIsSymmetric_SeqAIJ(Mat A,PetscReal tol,PetscBool *f)
2368: {
2372: MatIsTranspose_SeqAIJ(A,A,tol,f);
2373: return(0);
2374: }
2376: PetscErrorCode MatIsHermitian_SeqAIJ(Mat A,PetscReal tol,PetscBool *f)
2377: {
2381: MatIsHermitianTranspose_SeqAIJ(A,A,tol,f);
2382: return(0);
2383: }
2385: PetscErrorCode MatDiagonalScale_SeqAIJ(Mat A,Vec ll,Vec rr)
2386: {
2387: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
2388: const PetscScalar *l,*r;
2389: PetscScalar x;
2390: MatScalar *v;
2391: PetscErrorCode ierr;
2392: PetscInt i,j,m = A->rmap->n,n = A->cmap->n,M,nz = a->nz;
2393: const PetscInt *jj;
2396: if (ll) {
2397: /* The local size is used so that VecMPI can be passed to this routine
2398: by MatDiagonalScale_MPIAIJ */
2399: VecGetLocalSize(ll,&m);
2400: if (m != A->rmap->n) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Left scaling vector wrong length");
2401: VecGetArrayRead(ll,&l);
2402: v = a->a;
2403: for (i=0; i<m; i++) {
2404: x = l[i];
2405: M = a->i[i+1] - a->i[i];
2406: for (j=0; j<M; j++) (*v++) *= x;
2407: }
2408: VecRestoreArrayRead(ll,&l);
2409: PetscLogFlops(nz);
2410: }
2411: if (rr) {
2412: VecGetLocalSize(rr,&n);
2413: if (n != A->cmap->n) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Right scaling vector wrong length");
2414: VecGetArrayRead(rr,&r);
2415: v = a->a; jj = a->j;
2416: for (i=0; i<nz; i++) (*v++) *= r[*jj++];
2417: VecRestoreArrayRead(rr,&r);
2418: PetscLogFlops(nz);
2419: }
2420: MatSeqAIJInvalidateDiagonal(A);
2421: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
2422: if (A->offloadmask != PETSC_OFFLOAD_UNALLOCATED) A->offloadmask = PETSC_OFFLOAD_CPU;
2423: #endif
2424: return(0);
2425: }
2427: PetscErrorCode MatCreateSubMatrix_SeqAIJ(Mat A,IS isrow,IS iscol,PetscInt csize,MatReuse scall,Mat *B)
2428: {
2429: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data,*c;
2431: PetscInt *smap,i,k,kstart,kend,oldcols = A->cmap->n,*lens;
2432: PetscInt row,mat_i,*mat_j,tcol,first,step,*mat_ilen,sum,lensi;
2433: const PetscInt *irow,*icol;
2434: PetscInt nrows,ncols;
2435: PetscInt *starts,*j_new,*i_new,*aj = a->j,*ai = a->i,ii,*ailen = a->ilen;
2436: MatScalar *a_new,*mat_a;
2437: Mat C;
2438: PetscBool stride;
2442: ISGetIndices(isrow,&irow);
2443: ISGetLocalSize(isrow,&nrows);
2444: ISGetLocalSize(iscol,&ncols);
2446: PetscObjectTypeCompare((PetscObject)iscol,ISSTRIDE,&stride);
2447: if (stride) {
2448: ISStrideGetInfo(iscol,&first,&step);
2449: } else {
2450: first = 0;
2451: step = 0;
2452: }
2453: if (stride && step == 1) {
2454: /* special case of contiguous rows */
2455: PetscMalloc2(nrows,&lens,nrows,&starts);
2456: /* loop over new rows determining lens and starting points */
2457: for (i=0; i<nrows; i++) {
2458: kstart = ai[irow[i]];
2459: kend = kstart + ailen[irow[i]];
2460: starts[i] = kstart;
2461: for (k=kstart; k<kend; k++) {
2462: if (aj[k] >= first) {
2463: starts[i] = k;
2464: break;
2465: }
2466: }
2467: sum = 0;
2468: while (k < kend) {
2469: if (aj[k++] >= first+ncols) break;
2470: sum++;
2471: }
2472: lens[i] = sum;
2473: }
2474: /* create submatrix */
2475: if (scall == MAT_REUSE_MATRIX) {
2476: PetscInt n_cols,n_rows;
2477: MatGetSize(*B,&n_rows,&n_cols);
2478: if (n_rows != nrows || n_cols != ncols) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Reused submatrix wrong size");
2479: MatZeroEntries(*B);
2480: C = *B;
2481: } else {
2482: PetscInt rbs,cbs;
2483: MatCreate(PetscObjectComm((PetscObject)A),&C);
2484: MatSetSizes(C,nrows,ncols,PETSC_DETERMINE,PETSC_DETERMINE);
2485: ISGetBlockSize(isrow,&rbs);
2486: ISGetBlockSize(iscol,&cbs);
2487: MatSetBlockSizes(C,rbs,cbs);
2488: MatSetType(C,((PetscObject)A)->type_name);
2489: MatSeqAIJSetPreallocation_SeqAIJ(C,0,lens);
2490: }
2491: c = (Mat_SeqAIJ*)C->data;
2493: /* loop over rows inserting into submatrix */
2494: a_new = c->a;
2495: j_new = c->j;
2496: i_new = c->i;
2498: for (i=0; i<nrows; i++) {
2499: ii = starts[i];
2500: lensi = lens[i];
2501: for (k=0; k<lensi; k++) {
2502: *j_new++ = aj[ii+k] - first;
2503: }
2504: PetscArraycpy(a_new,a->a + starts[i],lensi);
2505: a_new += lensi;
2506: i_new[i+1] = i_new[i] + lensi;
2507: c->ilen[i] = lensi;
2508: }
2509: PetscFree2(lens,starts);
2510: } else {
2511: ISGetIndices(iscol,&icol);
2512: PetscCalloc1(oldcols,&smap);
2513: PetscMalloc1(1+nrows,&lens);
2514: for (i=0; i<ncols; i++) {
2515: if (PetscUnlikelyDebug(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);
2516: smap[icol[i]] = i+1;
2517: }
2519: /* determine lens of each row */
2520: for (i=0; i<nrows; i++) {
2521: kstart = ai[irow[i]];
2522: kend = kstart + a->ilen[irow[i]];
2523: lens[i] = 0;
2524: for (k=kstart; k<kend; k++) {
2525: if (smap[aj[k]]) {
2526: lens[i]++;
2527: }
2528: }
2529: }
2530: /* Create and fill new matrix */
2531: if (scall == MAT_REUSE_MATRIX) {
2532: PetscBool equal;
2534: c = (Mat_SeqAIJ*)((*B)->data);
2535: if ((*B)->rmap->n != nrows || (*B)->cmap->n != ncols) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Cannot reuse matrix. wrong size");
2536: PetscArraycmp(c->ilen,lens,(*B)->rmap->n,&equal);
2537: if (!equal) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Cannot reuse matrix. wrong no of nonzeros");
2538: PetscArrayzero(c->ilen,(*B)->rmap->n);
2539: C = *B;
2540: } else {
2541: PetscInt rbs,cbs;
2542: MatCreate(PetscObjectComm((PetscObject)A),&C);
2543: MatSetSizes(C,nrows,ncols,PETSC_DETERMINE,PETSC_DETERMINE);
2544: ISGetBlockSize(isrow,&rbs);
2545: ISGetBlockSize(iscol,&cbs);
2546: MatSetBlockSizes(C,rbs,cbs);
2547: MatSetType(C,((PetscObject)A)->type_name);
2548: MatSeqAIJSetPreallocation_SeqAIJ(C,0,lens);
2549: }
2550: c = (Mat_SeqAIJ*)(C->data);
2551: for (i=0; i<nrows; i++) {
2552: row = irow[i];
2553: kstart = ai[row];
2554: kend = kstart + a->ilen[row];
2555: mat_i = c->i[i];
2556: mat_j = c->j + mat_i;
2557: mat_a = c->a + mat_i;
2558: mat_ilen = c->ilen + i;
2559: for (k=kstart; k<kend; k++) {
2560: if ((tcol=smap[a->j[k]])) {
2561: *mat_j++ = tcol - 1;
2562: *mat_a++ = a->a[k];
2563: (*mat_ilen)++;
2565: }
2566: }
2567: }
2568: /* Free work space */
2569: ISRestoreIndices(iscol,&icol);
2570: PetscFree(smap);
2571: PetscFree(lens);
2572: /* sort */
2573: for (i = 0; i < nrows; i++) {
2574: PetscInt ilen;
2576: mat_i = c->i[i];
2577: mat_j = c->j + mat_i;
2578: mat_a = c->a + mat_i;
2579: ilen = c->ilen[i];
2580: PetscSortIntWithScalarArray(ilen,mat_j,mat_a);
2581: }
2582: }
2583: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
2584: MatBindToCPU(C,A->boundtocpu);
2585: #endif
2586: MatAssemblyBegin(C,MAT_FINAL_ASSEMBLY);
2587: MatAssemblyEnd(C,MAT_FINAL_ASSEMBLY);
2589: ISRestoreIndices(isrow,&irow);
2590: *B = C;
2591: return(0);
2592: }
2594: PetscErrorCode MatGetMultiProcBlock_SeqAIJ(Mat mat,MPI_Comm subComm,MatReuse scall,Mat *subMat)
2595: {
2597: Mat B;
2600: if (scall == MAT_INITIAL_MATRIX) {
2601: MatCreate(subComm,&B);
2602: MatSetSizes(B,mat->rmap->n,mat->cmap->n,mat->rmap->n,mat->cmap->n);
2603: MatSetBlockSizesFromMats(B,mat,mat);
2604: MatSetType(B,MATSEQAIJ);
2605: MatDuplicateNoCreate_SeqAIJ(B,mat,MAT_COPY_VALUES,PETSC_TRUE);
2606: *subMat = B;
2607: } else {
2608: MatCopy_SeqAIJ(mat,*subMat,SAME_NONZERO_PATTERN);
2609: }
2610: return(0);
2611: }
2613: PetscErrorCode MatILUFactor_SeqAIJ(Mat inA,IS row,IS col,const MatFactorInfo *info)
2614: {
2615: Mat_SeqAIJ *a = (Mat_SeqAIJ*)inA->data;
2617: Mat outA;
2618: PetscBool row_identity,col_identity;
2621: if (info->levels != 0) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Only levels=0 supported for in-place ilu");
2623: ISIdentity(row,&row_identity);
2624: ISIdentity(col,&col_identity);
2626: outA = inA;
2627: outA->factortype = MAT_FACTOR_LU;
2628: PetscFree(inA->solvertype);
2629: PetscStrallocpy(MATSOLVERPETSC,&inA->solvertype);
2631: PetscObjectReference((PetscObject)row);
2632: ISDestroy(&a->row);
2634: a->row = row;
2636: PetscObjectReference((PetscObject)col);
2637: ISDestroy(&a->col);
2639: a->col = col;
2641: /* Create the inverse permutation so that it can be used in MatLUFactorNumeric() */
2642: ISDestroy(&a->icol);
2643: ISInvertPermutation(col,PETSC_DECIDE,&a->icol);
2644: PetscLogObjectParent((PetscObject)inA,(PetscObject)a->icol);
2646: if (!a->solve_work) { /* this matrix may have been factored before */
2647: PetscMalloc1(inA->rmap->n+1,&a->solve_work);
2648: PetscLogObjectMemory((PetscObject)inA, (inA->rmap->n+1)*sizeof(PetscScalar));
2649: }
2651: MatMarkDiagonal_SeqAIJ(inA);
2652: if (row_identity && col_identity) {
2653: MatLUFactorNumeric_SeqAIJ_inplace(outA,inA,info);
2654: } else {
2655: MatLUFactorNumeric_SeqAIJ_InplaceWithPerm(outA,inA,info);
2656: }
2657: return(0);
2658: }
2660: PetscErrorCode MatScale_SeqAIJ(Mat inA,PetscScalar alpha)
2661: {
2662: Mat_SeqAIJ *a = (Mat_SeqAIJ*)inA->data;
2663: PetscScalar oalpha = alpha;
2665: PetscBLASInt one = 1,bnz;
2668: PetscBLASIntCast(a->nz,&bnz);
2669: PetscStackCallBLAS("BLASscal",BLASscal_(&bnz,&oalpha,a->a,&one));
2670: PetscLogFlops(a->nz);
2671: MatSeqAIJInvalidateDiagonal(inA);
2672: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
2673: if (inA->offloadmask != PETSC_OFFLOAD_UNALLOCATED) inA->offloadmask = PETSC_OFFLOAD_CPU;
2674: #endif
2675: return(0);
2676: }
2678: PetscErrorCode MatDestroySubMatrix_Private(Mat_SubSppt *submatj)
2679: {
2681: PetscInt i;
2684: if (!submatj->id) { /* delete data that are linked only to submats[id=0] */
2685: PetscFree4(submatj->sbuf1,submatj->ptr,submatj->tmp,submatj->ctr);
2687: for (i=0; i<submatj->nrqr; ++i) {
2688: PetscFree(submatj->sbuf2[i]);
2689: }
2690: PetscFree3(submatj->sbuf2,submatj->req_size,submatj->req_source1);
2692: if (submatj->rbuf1) {
2693: PetscFree(submatj->rbuf1[0]);
2694: PetscFree(submatj->rbuf1);
2695: }
2697: for (i=0; i<submatj->nrqs; ++i) {
2698: PetscFree(submatj->rbuf3[i]);
2699: }
2700: PetscFree3(submatj->req_source2,submatj->rbuf2,submatj->rbuf3);
2701: PetscFree(submatj->pa);
2702: }
2704: #if defined(PETSC_USE_CTABLE)
2705: PetscTableDestroy((PetscTable*)&submatj->rmap);
2706: if (submatj->cmap_loc) {PetscFree(submatj->cmap_loc);}
2707: PetscFree(submatj->rmap_loc);
2708: #else
2709: PetscFree(submatj->rmap);
2710: #endif
2712: if (!submatj->allcolumns) {
2713: #if defined(PETSC_USE_CTABLE)
2714: PetscTableDestroy((PetscTable*)&submatj->cmap);
2715: #else
2716: PetscFree(submatj->cmap);
2717: #endif
2718: }
2719: PetscFree(submatj->row2proc);
2721: PetscFree(submatj);
2722: return(0);
2723: }
2725: PetscErrorCode MatDestroySubMatrix_SeqAIJ(Mat C)
2726: {
2728: Mat_SeqAIJ *c = (Mat_SeqAIJ*)C->data;
2729: Mat_SubSppt *submatj = c->submatis1;
2732: (*submatj->destroy)(C);
2733: MatDestroySubMatrix_Private(submatj);
2734: return(0);
2735: }
2737: PetscErrorCode MatDestroySubMatrices_SeqAIJ(PetscInt n,Mat *mat[])
2738: {
2740: PetscInt i;
2741: Mat C;
2742: Mat_SeqAIJ *c;
2743: Mat_SubSppt *submatj;
2746: for (i=0; i<n; i++) {
2747: C = (*mat)[i];
2748: c = (Mat_SeqAIJ*)C->data;
2749: submatj = c->submatis1;
2750: if (submatj) {
2751: if (--((PetscObject)C)->refct <= 0) {
2752: (*submatj->destroy)(C);
2753: MatDestroySubMatrix_Private(submatj);
2754: PetscFree(C->defaultvectype);
2755: PetscLayoutDestroy(&C->rmap);
2756: PetscLayoutDestroy(&C->cmap);
2757: PetscHeaderDestroy(&C);
2758: }
2759: } else {
2760: MatDestroy(&C);
2761: }
2762: }
2764: /* Destroy Dummy submatrices created for reuse */
2765: MatDestroySubMatrices_Dummy(n,mat);
2767: PetscFree(*mat);
2768: return(0);
2769: }
2771: PetscErrorCode MatCreateSubMatrices_SeqAIJ(Mat A,PetscInt n,const IS irow[],const IS icol[],MatReuse scall,Mat *B[])
2772: {
2774: PetscInt i;
2777: if (scall == MAT_INITIAL_MATRIX) {
2778: PetscCalloc1(n+1,B);
2779: }
2781: for (i=0; i<n; i++) {
2782: MatCreateSubMatrix_SeqAIJ(A,irow[i],icol[i],PETSC_DECIDE,scall,&(*B)[i]);
2783: }
2784: return(0);
2785: }
2787: PetscErrorCode MatIncreaseOverlap_SeqAIJ(Mat A,PetscInt is_max,IS is[],PetscInt ov)
2788: {
2789: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
2791: PetscInt row,i,j,k,l,m,n,*nidx,isz,val;
2792: const PetscInt *idx;
2793: PetscInt start,end,*ai,*aj;
2794: PetscBT table;
2797: m = A->rmap->n;
2798: ai = a->i;
2799: aj = a->j;
2801: if (ov < 0) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"illegal negative overlap value used");
2803: PetscMalloc1(m+1,&nidx);
2804: PetscBTCreate(m,&table);
2806: for (i=0; i<is_max; i++) {
2807: /* Initialize the two local arrays */
2808: isz = 0;
2809: PetscBTMemzero(m,table);
2811: /* Extract the indices, assume there can be duplicate entries */
2812: ISGetIndices(is[i],&idx);
2813: ISGetLocalSize(is[i],&n);
2815: /* Enter these into the temp arrays. I.e., mark table[row], enter row into new index */
2816: for (j=0; j<n; ++j) {
2817: if (!PetscBTLookupSet(table,idx[j])) nidx[isz++] = idx[j];
2818: }
2819: ISRestoreIndices(is[i],&idx);
2820: ISDestroy(&is[i]);
2822: k = 0;
2823: for (j=0; j<ov; j++) { /* for each overlap */
2824: n = isz;
2825: for (; k<n; k++) { /* do only those rows in nidx[k], which are not done yet */
2826: row = nidx[k];
2827: start = ai[row];
2828: end = ai[row+1];
2829: for (l = start; l<end; l++) {
2830: val = aj[l];
2831: if (!PetscBTLookupSet(table,val)) nidx[isz++] = val;
2832: }
2833: }
2834: }
2835: ISCreateGeneral(PETSC_COMM_SELF,isz,nidx,PETSC_COPY_VALUES,(is+i));
2836: }
2837: PetscBTDestroy(&table);
2838: PetscFree(nidx);
2839: return(0);
2840: }
2842: /* -------------------------------------------------------------- */
2843: PetscErrorCode MatPermute_SeqAIJ(Mat A,IS rowp,IS colp,Mat *B)
2844: {
2845: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
2847: PetscInt i,nz = 0,m = A->rmap->n,n = A->cmap->n;
2848: const PetscInt *row,*col;
2849: PetscInt *cnew,j,*lens;
2850: IS icolp,irowp;
2851: PetscInt *cwork = NULL;
2852: PetscScalar *vwork = NULL;
2855: ISInvertPermutation(rowp,PETSC_DECIDE,&irowp);
2856: ISGetIndices(irowp,&row);
2857: ISInvertPermutation(colp,PETSC_DECIDE,&icolp);
2858: ISGetIndices(icolp,&col);
2860: /* determine lengths of permuted rows */
2861: PetscMalloc1(m+1,&lens);
2862: for (i=0; i<m; i++) lens[row[i]] = a->i[i+1] - a->i[i];
2863: MatCreate(PetscObjectComm((PetscObject)A),B);
2864: MatSetSizes(*B,m,n,m,n);
2865: MatSetBlockSizesFromMats(*B,A,A);
2866: MatSetType(*B,((PetscObject)A)->type_name);
2867: MatSeqAIJSetPreallocation_SeqAIJ(*B,0,lens);
2868: PetscFree(lens);
2870: PetscMalloc1(n,&cnew);
2871: for (i=0; i<m; i++) {
2872: MatGetRow_SeqAIJ(A,i,&nz,&cwork,&vwork);
2873: for (j=0; j<nz; j++) cnew[j] = col[cwork[j]];
2874: MatSetValues_SeqAIJ(*B,1,&row[i],nz,cnew,vwork,INSERT_VALUES);
2875: MatRestoreRow_SeqAIJ(A,i,&nz,&cwork,&vwork);
2876: }
2877: PetscFree(cnew);
2879: (*B)->assembled = PETSC_FALSE;
2881: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
2882: MatBindToCPU(*B,A->boundtocpu);
2883: #endif
2884: MatAssemblyBegin(*B,MAT_FINAL_ASSEMBLY);
2885: MatAssemblyEnd(*B,MAT_FINAL_ASSEMBLY);
2886: ISRestoreIndices(irowp,&row);
2887: ISRestoreIndices(icolp,&col);
2888: ISDestroy(&irowp);
2889: ISDestroy(&icolp);
2890: if (rowp == colp) {
2891: if (A->symmetric) {
2892: MatSetOption(*B,MAT_SYMMETRIC,PETSC_TRUE);
2893: }
2894: if (A->hermitian) {
2895: MatSetOption(*B,MAT_HERMITIAN,PETSC_TRUE);
2896: }
2897: }
2898: return(0);
2899: }
2901: PetscErrorCode MatCopy_SeqAIJ(Mat A,Mat B,MatStructure str)
2902: {
2906: /* If the two matrices have the same copy implementation, use fast copy. */
2907: if (str == SAME_NONZERO_PATTERN && (A->ops->copy == B->ops->copy)) {
2908: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
2909: Mat_SeqAIJ *b = (Mat_SeqAIJ*)B->data;
2911: if (a->i[A->rmap->n] != b->i[B->rmap->n]) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_INCOMP,"Number of nonzeros in two matrices are different %D != %D",a->i[A->rmap->n],b->i[B->rmap->n]);
2912: PetscArraycpy(b->a,a->a,a->i[A->rmap->n]);
2913: PetscObjectStateIncrease((PetscObject)B);
2914: } else {
2915: MatCopy_Basic(A,B,str);
2916: }
2917: return(0);
2918: }
2920: PetscErrorCode MatSetUp_SeqAIJ(Mat A)
2921: {
2925: MatSeqAIJSetPreallocation_SeqAIJ(A,PETSC_DEFAULT,0);
2926: return(0);
2927: }
2929: PETSC_INTERN PetscErrorCode MatSeqAIJGetArray_SeqAIJ(Mat A,PetscScalar *array[])
2930: {
2931: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
2934: *array = a->a;
2935: return(0);
2936: }
2938: PETSC_INTERN PetscErrorCode MatSeqAIJRestoreArray_SeqAIJ(Mat A,PetscScalar *array[])
2939: {
2941: *array = NULL;
2942: return(0);
2943: }
2945: /*
2946: Computes the number of nonzeros per row needed for preallocation when X and Y
2947: have different nonzero structure.
2948: */
2949: PetscErrorCode MatAXPYGetPreallocation_SeqX_private(PetscInt m,const PetscInt *xi,const PetscInt *xj,const PetscInt *yi,const PetscInt *yj,PetscInt *nnz)
2950: {
2951: PetscInt i,j,k,nzx,nzy;
2954: /* Set the number of nonzeros in the new matrix */
2955: for (i=0; i<m; i++) {
2956: const PetscInt *xjj = xj+xi[i],*yjj = yj+yi[i];
2957: nzx = xi[i+1] - xi[i];
2958: nzy = yi[i+1] - yi[i];
2959: nnz[i] = 0;
2960: for (j=0,k=0; j<nzx; j++) { /* Point in X */
2961: for (; k<nzy && yjj[k]<xjj[j]; k++) nnz[i]++; /* Catch up to X */
2962: if (k<nzy && yjj[k]==xjj[j]) k++; /* Skip duplicate */
2963: nnz[i]++;
2964: }
2965: for (; k<nzy; k++) nnz[i]++;
2966: }
2967: return(0);
2968: }
2970: PetscErrorCode MatAXPYGetPreallocation_SeqAIJ(Mat Y,Mat X,PetscInt *nnz)
2971: {
2972: PetscInt m = Y->rmap->N;
2973: Mat_SeqAIJ *x = (Mat_SeqAIJ*)X->data;
2974: Mat_SeqAIJ *y = (Mat_SeqAIJ*)Y->data;
2978: /* Set the number of nonzeros in the new matrix */
2979: MatAXPYGetPreallocation_SeqX_private(m,x->i,x->j,y->i,y->j,nnz);
2980: return(0);
2981: }
2983: PetscErrorCode MatAXPY_SeqAIJ(Mat Y,PetscScalar a,Mat X,MatStructure str)
2984: {
2986: Mat_SeqAIJ *x = (Mat_SeqAIJ*)X->data,*y = (Mat_SeqAIJ*)Y->data;
2987: PetscBLASInt one=1,bnz;
2990: PetscBLASIntCast(x->nz,&bnz);
2991: if (str == SAME_NONZERO_PATTERN) {
2992: PetscScalar alpha = a;
2993: PetscStackCallBLAS("BLASaxpy",BLASaxpy_(&bnz,&alpha,x->a,&one,y->a,&one));
2994: MatSeqAIJInvalidateDiagonal(Y);
2995: PetscObjectStateIncrease((PetscObject)Y);
2996: /* the MatAXPY_Basic* subroutines calls MatAssembly, so the matrix on the GPU
2997: will be updated */
2998: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
2999: if (Y->offloadmask != PETSC_OFFLOAD_UNALLOCATED) {
3000: Y->offloadmask = PETSC_OFFLOAD_CPU;
3001: }
3002: #endif
3003: } else if (str == SUBSET_NONZERO_PATTERN) { /* nonzeros of X is a subset of Y's */
3004: MatAXPY_Basic(Y,a,X,str);
3005: } else {
3006: Mat B;
3007: PetscInt *nnz;
3008: PetscMalloc1(Y->rmap->N,&nnz);
3009: MatCreate(PetscObjectComm((PetscObject)Y),&B);
3010: PetscObjectSetName((PetscObject)B,((PetscObject)Y)->name);
3011: MatSetSizes(B,Y->rmap->n,Y->cmap->n,Y->rmap->N,Y->cmap->N);
3012: MatSetBlockSizesFromMats(B,Y,Y);
3013: MatSetType(B,(MatType) ((PetscObject)Y)->type_name);
3014: MatAXPYGetPreallocation_SeqAIJ(Y,X,nnz);
3015: MatSeqAIJSetPreallocation(B,0,nnz);
3016: MatAXPY_BasicWithPreallocation(B,Y,a,X,str);
3017: MatHeaderReplace(Y,&B);
3018: PetscFree(nnz);
3019: }
3020: return(0);
3021: }
3023: PetscErrorCode MatConjugate_SeqAIJ(Mat mat)
3024: {
3025: #if defined(PETSC_USE_COMPLEX)
3026: Mat_SeqAIJ *aij = (Mat_SeqAIJ*)mat->data;
3027: PetscInt i,nz;
3028: PetscScalar *a;
3031: nz = aij->nz;
3032: a = aij->a;
3033: for (i=0; i<nz; i++) a[i] = PetscConj(a[i]);
3034: #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
3035: if (mat->offloadmask != PETSC_OFFLOAD_UNALLOCATED) mat->offloadmask = PETSC_OFFLOAD_CPU;
3036: #endif
3037: #else
3039: #endif
3040: return(0);
3041: }
3043: PetscErrorCode MatGetRowMaxAbs_SeqAIJ(Mat A,Vec v,PetscInt idx[])
3044: {
3045: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
3047: PetscInt i,j,m = A->rmap->n,*ai,*aj,ncols,n;
3048: PetscReal atmp;
3049: PetscScalar *x;
3050: MatScalar *aa;
3053: if (A->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3054: aa = a->a;
3055: ai = a->i;
3056: aj = a->j;
3058: VecSet(v,0.0);
3059: VecGetArray(v,&x);
3060: VecGetLocalSize(v,&n);
3061: if (n != A->rmap->n) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Nonconforming matrix and vector");
3062: for (i=0; i<m; i++) {
3063: ncols = ai[1] - ai[0]; ai++;
3064: x[i] = 0.0;
3065: for (j=0; j<ncols; j++) {
3066: atmp = PetscAbsScalar(*aa);
3067: if (PetscAbsScalar(x[i]) < atmp) {x[i] = atmp; if (idx) idx[i] = *aj;}
3068: aa++; aj++;
3069: }
3070: }
3071: VecRestoreArray(v,&x);
3072: return(0);
3073: }
3075: PetscErrorCode MatGetRowMax_SeqAIJ(Mat A,Vec v,PetscInt idx[])
3076: {
3077: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
3079: PetscInt i,j,m = A->rmap->n,*ai,*aj,ncols,n;
3080: PetscScalar *x;
3081: MatScalar *aa;
3084: if (A->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3085: aa = a->a;
3086: ai = a->i;
3087: aj = a->j;
3089: VecSet(v,0.0);
3090: VecGetArray(v,&x);
3091: VecGetLocalSize(v,&n);
3092: if (n != A->rmap->n) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Nonconforming matrix and vector");
3093: for (i=0; i<m; i++) {
3094: ncols = ai[1] - ai[0]; ai++;
3095: if (ncols == A->cmap->n) { /* row is dense */
3096: x[i] = *aa; if (idx) idx[i] = 0;
3097: } else { /* row is sparse so already KNOW maximum is 0.0 or higher */
3098: x[i] = 0.0;
3099: if (idx) {
3100: idx[i] = 0; /* in case ncols is zero */
3101: for (j=0;j<ncols;j++) { /* find first implicit 0.0 in the row */
3102: if (aj[j] > j) {
3103: idx[i] = j;
3104: break;
3105: }
3106: }
3107: }
3108: }
3109: for (j=0; j<ncols; j++) {
3110: if (PetscRealPart(x[i]) < PetscRealPart(*aa)) {x[i] = *aa; if (idx) idx[i] = *aj;}
3111: aa++; aj++;
3112: }
3113: }
3114: VecRestoreArray(v,&x);
3115: return(0);
3116: }
3118: PetscErrorCode MatGetRowMinAbs_SeqAIJ(Mat A,Vec v,PetscInt idx[])
3119: {
3120: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
3122: PetscInt i,j,m = A->rmap->n,*ai,*aj,ncols,n;
3123: PetscReal atmp;
3124: PetscScalar *x;
3125: MatScalar *aa;
3128: if (A->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3129: aa = a->a;
3130: ai = a->i;
3131: aj = a->j;
3133: VecSet(v,0.0);
3134: VecGetArray(v,&x);
3135: VecGetLocalSize(v,&n);
3136: 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);
3137: for (i=0; i<m; i++) {
3138: ncols = ai[1] - ai[0]; ai++;
3139: if (ncols) {
3140: /* Get first nonzero */
3141: for (j = 0; j < ncols; j++) {
3142: atmp = PetscAbsScalar(aa[j]);
3143: if (atmp > 1.0e-12) {
3144: x[i] = atmp;
3145: if (idx) idx[i] = aj[j];
3146: break;
3147: }
3148: }
3149: if (j == ncols) {x[i] = PetscAbsScalar(*aa); if (idx) idx[i] = *aj;}
3150: } else {
3151: x[i] = 0.0; if (idx) idx[i] = 0;
3152: }
3153: for (j = 0; j < ncols; j++) {
3154: atmp = PetscAbsScalar(*aa);
3155: if (atmp > 1.0e-12 && PetscAbsScalar(x[i]) > atmp) {x[i] = atmp; if (idx) idx[i] = *aj;}
3156: aa++; aj++;
3157: }
3158: }
3159: VecRestoreArray(v,&x);
3160: return(0);
3161: }
3163: PetscErrorCode MatGetRowMin_SeqAIJ(Mat A,Vec v,PetscInt idx[])
3164: {
3165: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
3166: PetscErrorCode ierr;
3167: PetscInt i,j,m = A->rmap->n,ncols,n;
3168: const PetscInt *ai,*aj;
3169: PetscScalar *x;
3170: const MatScalar *aa;
3173: if (A->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3174: aa = a->a;
3175: ai = a->i;
3176: aj = a->j;
3178: VecSet(v,0.0);
3179: VecGetArray(v,&x);
3180: VecGetLocalSize(v,&n);
3181: if (n != A->rmap->n) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Nonconforming matrix and vector");
3182: for (i=0; i<m; i++) {
3183: ncols = ai[1] - ai[0]; ai++;
3184: if (ncols == A->cmap->n) { /* row is dense */
3185: x[i] = *aa; if (idx) idx[i] = 0;
3186: } else { /* row is sparse so already KNOW minimum is 0.0 or lower */
3187: x[i] = 0.0;
3188: if (idx) { /* find first implicit 0.0 in the row */
3189: idx[i] = 0; /* in case ncols is zero */
3190: for (j=0; j<ncols; j++) {
3191: if (aj[j] > j) {
3192: idx[i] = j;
3193: break;
3194: }
3195: }
3196: }
3197: }
3198: for (j=0; j<ncols; j++) {
3199: if (PetscRealPart(x[i]) > PetscRealPart(*aa)) {x[i] = *aa; if (idx) idx[i] = *aj;}
3200: aa++; aj++;
3201: }
3202: }
3203: VecRestoreArray(v,&x);
3204: return(0);
3205: }
3207: PetscErrorCode MatInvertBlockDiagonal_SeqAIJ(Mat A,const PetscScalar **values)
3208: {
3209: Mat_SeqAIJ *a = (Mat_SeqAIJ*) A->data;
3210: PetscErrorCode ierr;
3211: PetscInt i,bs = PetscAbs(A->rmap->bs),mbs = A->rmap->n/bs,ipvt[5],bs2 = bs*bs,*v_pivots,ij[7],*IJ,j;
3212: MatScalar *diag,work[25],*v_work;
3213: const PetscReal shift = 0.0;
3214: PetscBool allowzeropivot,zeropivotdetected=PETSC_FALSE;
3217: allowzeropivot = PetscNot(A->erroriffailure);
3218: if (a->ibdiagvalid) {
3219: if (values) *values = a->ibdiag;
3220: return(0);
3221: }
3222: MatMarkDiagonal_SeqAIJ(A);
3223: if (!a->ibdiag) {
3224: PetscMalloc1(bs2*mbs,&a->ibdiag);
3225: PetscLogObjectMemory((PetscObject)A,bs2*mbs*sizeof(PetscScalar));
3226: }
3227: diag = a->ibdiag;
3228: if (values) *values = a->ibdiag;
3229: /* factor and invert each block */
3230: switch (bs) {
3231: case 1:
3232: for (i=0; i<mbs; i++) {
3233: MatGetValues(A,1,&i,1,&i,diag+i);
3234: if (PetscAbsScalar(diag[i] + shift) < PETSC_MACHINE_EPSILON) {
3235: if (allowzeropivot) {
3236: A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
3237: A->factorerror_zeropivot_value = PetscAbsScalar(diag[i]);
3238: A->factorerror_zeropivot_row = i;
3239: PetscInfo3(A,"Zero pivot, row %D pivot %g tolerance %g\n",i,(double)PetscAbsScalar(diag[i]),(double)PETSC_MACHINE_EPSILON);
3240: } 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);
3241: }
3242: diag[i] = (PetscScalar)1.0 / (diag[i] + shift);
3243: }
3244: break;
3245: case 2:
3246: for (i=0; i<mbs; i++) {
3247: ij[0] = 2*i; ij[1] = 2*i + 1;
3248: MatGetValues(A,2,ij,2,ij,diag);
3249: PetscKernel_A_gets_inverse_A_2(diag,shift,allowzeropivot,&zeropivotdetected);
3250: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
3251: PetscKernel_A_gets_transpose_A_2(diag);
3252: diag += 4;
3253: }
3254: break;
3255: case 3:
3256: for (i=0; i<mbs; i++) {
3257: ij[0] = 3*i; ij[1] = 3*i + 1; ij[2] = 3*i + 2;
3258: MatGetValues(A,3,ij,3,ij,diag);
3259: PetscKernel_A_gets_inverse_A_3(diag,shift,allowzeropivot,&zeropivotdetected);
3260: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
3261: PetscKernel_A_gets_transpose_A_3(diag);
3262: diag += 9;
3263: }
3264: break;
3265: case 4:
3266: for (i=0; i<mbs; i++) {
3267: ij[0] = 4*i; ij[1] = 4*i + 1; ij[2] = 4*i + 2; ij[3] = 4*i + 3;
3268: MatGetValues(A,4,ij,4,ij,diag);
3269: PetscKernel_A_gets_inverse_A_4(diag,shift,allowzeropivot,&zeropivotdetected);
3270: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
3271: PetscKernel_A_gets_transpose_A_4(diag);
3272: diag += 16;
3273: }
3274: break;
3275: case 5:
3276: for (i=0; i<mbs; i++) {
3277: ij[0] = 5*i; ij[1] = 5*i + 1; ij[2] = 5*i + 2; ij[3] = 5*i + 3; ij[4] = 5*i + 4;
3278: MatGetValues(A,5,ij,5,ij,diag);
3279: PetscKernel_A_gets_inverse_A_5(diag,ipvt,work,shift,allowzeropivot,&zeropivotdetected);
3280: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
3281: PetscKernel_A_gets_transpose_A_5(diag);
3282: diag += 25;
3283: }
3284: break;
3285: case 6:
3286: for (i=0; i<mbs; i++) {
3287: 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;
3288: MatGetValues(A,6,ij,6,ij,diag);
3289: PetscKernel_A_gets_inverse_A_6(diag,shift,allowzeropivot,&zeropivotdetected);
3290: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
3291: PetscKernel_A_gets_transpose_A_6(diag);
3292: diag += 36;
3293: }
3294: break;
3295: case 7:
3296: for (i=0; i<mbs; i++) {
3297: 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;
3298: MatGetValues(A,7,ij,7,ij,diag);
3299: PetscKernel_A_gets_inverse_A_7(diag,shift,allowzeropivot,&zeropivotdetected);
3300: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
3301: PetscKernel_A_gets_transpose_A_7(diag);
3302: diag += 49;
3303: }
3304: break;
3305: default:
3306: PetscMalloc3(bs,&v_work,bs,&v_pivots,bs,&IJ);
3307: for (i=0; i<mbs; i++) {
3308: for (j=0; j<bs; j++) {
3309: IJ[j] = bs*i + j;
3310: }
3311: MatGetValues(A,bs,IJ,bs,IJ,diag);
3312: PetscKernel_A_gets_inverse_A(bs,diag,v_pivots,v_work,allowzeropivot,&zeropivotdetected);
3313: if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
3314: PetscKernel_A_gets_transpose_A_N(diag,bs);
3315: diag += bs2;
3316: }
3317: PetscFree3(v_work,v_pivots,IJ);
3318: }
3319: a->ibdiagvalid = PETSC_TRUE;
3320: return(0);
3321: }
3323: static PetscErrorCode MatSetRandom_SeqAIJ(Mat x,PetscRandom rctx)
3324: {
3326: Mat_SeqAIJ *aij = (Mat_SeqAIJ*)x->data;
3327: PetscScalar a;
3328: PetscInt m,n,i,j,col;
3331: if (!x->assembled) {
3332: MatGetSize(x,&m,&n);
3333: for (i=0; i<m; i++) {
3334: for (j=0; j<aij->imax[i]; j++) {
3335: PetscRandomGetValue(rctx,&a);
3336: col = (PetscInt)(n*PetscRealPart(a));
3337: MatSetValues(x,1,&i,1,&col,&a,ADD_VALUES);
3338: }
3339: }
3340: } else {
3341: for (i=0; i<aij->nz; i++) {PetscRandomGetValue(rctx,aij->a+i);}
3342: }
3343: MatAssemblyBegin(x,MAT_FINAL_ASSEMBLY);
3344: MatAssemblyEnd(x,MAT_FINAL_ASSEMBLY);
3345: return(0);
3346: }
3348: /* Like MatSetRandom_SeqAIJ, but do not set values on columns in range of [low, high) */
3349: PetscErrorCode MatSetRandomSkipColumnRange_SeqAIJ_Private(Mat x,PetscInt low,PetscInt high,PetscRandom rctx)
3350: {
3352: Mat_SeqAIJ *aij = (Mat_SeqAIJ*)x->data;
3353: PetscScalar a;
3354: PetscInt m,n,i,j,col,nskip;
3357: nskip = high - low;
3358: MatGetSize(x,&m,&n);
3359: n -= nskip; /* shrink number of columns where nonzeros can be set */
3360: for (i=0; i<m; i++) {
3361: for (j=0; j<aij->imax[i]; j++) {
3362: PetscRandomGetValue(rctx,&a);
3363: col = (PetscInt)(n*PetscRealPart(a));
3364: if (col >= low) col += nskip; /* shift col rightward to skip the hole */
3365: MatSetValues(x,1,&i,1,&col,&a,ADD_VALUES);
3366: }
3367: }
3368: MatAssemblyBegin(x,MAT_FINAL_ASSEMBLY);
3369: MatAssemblyEnd(x,MAT_FINAL_ASSEMBLY);
3370: return(0);
3371: }
3374: /* -------------------------------------------------------------------*/
3375: static struct _MatOps MatOps_Values = { MatSetValues_SeqAIJ,
3376: MatGetRow_SeqAIJ,
3377: MatRestoreRow_SeqAIJ,
3378: MatMult_SeqAIJ,
3379: /* 4*/ MatMultAdd_SeqAIJ,
3380: MatMultTranspose_SeqAIJ,
3381: MatMultTransposeAdd_SeqAIJ,
3382: 0,
3383: 0,
3384: 0,
3385: /* 10*/ 0,
3386: MatLUFactor_SeqAIJ,
3387: 0,
3388: MatSOR_SeqAIJ,
3389: MatTranspose_SeqAIJ,
3390: /*1 5*/ MatGetInfo_SeqAIJ,
3391: MatEqual_SeqAIJ,
3392: MatGetDiagonal_SeqAIJ,
3393: MatDiagonalScale_SeqAIJ,
3394: MatNorm_SeqAIJ,
3395: /* 20*/ 0,
3396: MatAssemblyEnd_SeqAIJ,
3397: MatSetOption_SeqAIJ,
3398: MatZeroEntries_SeqAIJ,
3399: /* 24*/ MatZeroRows_SeqAIJ,
3400: 0,
3401: 0,
3402: 0,
3403: 0,
3404: /* 29*/ MatSetUp_SeqAIJ,
3405: 0,
3406: 0,
3407: 0,
3408: 0,
3409: /* 34*/ MatDuplicate_SeqAIJ,
3410: 0,
3411: 0,
3412: MatILUFactor_SeqAIJ,
3413: 0,
3414: /* 39*/ MatAXPY_SeqAIJ,
3415: MatCreateSubMatrices_SeqAIJ,
3416: MatIncreaseOverlap_SeqAIJ,
3417: MatGetValues_SeqAIJ,
3418: MatCopy_SeqAIJ,
3419: /* 44*/ MatGetRowMax_SeqAIJ,
3420: MatScale_SeqAIJ,
3421: MatShift_SeqAIJ,
3422: MatDiagonalSet_SeqAIJ,
3423: MatZeroRowsColumns_SeqAIJ,
3424: /* 49*/ MatSetRandom_SeqAIJ,
3425: MatGetRowIJ_SeqAIJ,
3426: MatRestoreRowIJ_SeqAIJ,
3427: MatGetColumnIJ_SeqAIJ,
3428: MatRestoreColumnIJ_SeqAIJ,
3429: /* 54*/ MatFDColoringCreate_SeqXAIJ,
3430: 0,
3431: 0,
3432: MatPermute_SeqAIJ,
3433: 0,
3434: /* 59*/ 0,
3435: MatDestroy_SeqAIJ,
3436: MatView_SeqAIJ,
3437: 0,
3438: 0,
3439: /* 64*/ 0,
3440: MatMatMatMultNumeric_SeqAIJ_SeqAIJ_SeqAIJ,
3441: 0,
3442: 0,
3443: 0,
3444: /* 69*/ MatGetRowMaxAbs_SeqAIJ,
3445: MatGetRowMinAbs_SeqAIJ,
3446: 0,
3447: 0,
3448: 0,
3449: /* 74*/ 0,
3450: MatFDColoringApply_AIJ,
3451: 0,
3452: 0,
3453: 0,
3454: /* 79*/ MatFindZeroDiagonals_SeqAIJ,
3455: 0,
3456: 0,
3457: 0,
3458: MatLoad_SeqAIJ,
3459: /* 84*/ MatIsSymmetric_SeqAIJ,
3460: MatIsHermitian_SeqAIJ,
3461: 0,
3462: 0,
3463: 0,
3464: /* 89*/ 0,
3465: 0,
3466: MatMatMultNumeric_SeqAIJ_SeqAIJ,
3467: 0,
3468: 0,
3469: /* 94*/ MatPtAPNumeric_SeqAIJ_SeqAIJ_SparseAxpy,
3470: 0,
3471: 0,
3472: MatMatTransposeMultNumeric_SeqAIJ_SeqAIJ,
3473: 0,
3474: /* 99*/ MatProductSetFromOptions_SeqAIJ,
3475: 0,
3476: 0,
3477: MatConjugate_SeqAIJ,
3478: 0,
3479: /*104*/ MatSetValuesRow_SeqAIJ,
3480: MatRealPart_SeqAIJ,
3481: MatImaginaryPart_SeqAIJ,
3482: 0,
3483: 0,
3484: /*109*/ MatMatSolve_SeqAIJ,
3485: 0,
3486: MatGetRowMin_SeqAIJ,
3487: 0,
3488: MatMissingDiagonal_SeqAIJ,
3489: /*114*/ 0,
3490: 0,
3491: 0,
3492: 0,
3493: 0,
3494: /*119*/ 0,
3495: 0,
3496: 0,
3497: 0,
3498: MatGetMultiProcBlock_SeqAIJ,
3499: /*124*/ MatFindNonzeroRows_SeqAIJ,
3500: MatGetColumnNorms_SeqAIJ,
3501: MatInvertBlockDiagonal_SeqAIJ,
3502: MatInvertVariableBlockDiagonal_SeqAIJ,
3503: 0,
3504: /*129*/ 0,
3505: 0,
3506: 0,
3507: MatTransposeMatMultNumeric_SeqAIJ_SeqAIJ,
3508: MatTransposeColoringCreate_SeqAIJ,
3509: /*134*/ MatTransColoringApplySpToDen_SeqAIJ,
3510: MatTransColoringApplyDenToSp_SeqAIJ,
3511: 0,
3512: 0,
3513: MatRARtNumeric_SeqAIJ_SeqAIJ,
3514: /*139*/0,
3515: 0,
3516: 0,
3517: MatFDColoringSetUp_SeqXAIJ,
3518: MatFindOffBlockDiagonalEntries_SeqAIJ,
3519: MatCreateMPIMatConcatenateSeqMat_SeqAIJ,
3520: /*145*/MatDestroySubMatrices_SeqAIJ,
3521: 0,
3522: 0
3523: };
3525: PetscErrorCode MatSeqAIJSetColumnIndices_SeqAIJ(Mat mat,PetscInt *indices)
3526: {
3527: Mat_SeqAIJ *aij = (Mat_SeqAIJ*)mat->data;
3528: PetscInt i,nz,n;
3531: nz = aij->maxnz;
3532: n = mat->rmap->n;
3533: for (i=0; i<nz; i++) {
3534: aij->j[i] = indices[i];
3535: }
3536: aij->nz = nz;
3537: for (i=0; i<n; i++) {
3538: aij->ilen[i] = aij->imax[i];
3539: }
3540: return(0);
3541: }
3543: /*
3544: * When a sparse matrix has many zero columns, we should compact them out to save the space
3545: * This happens in MatPtAPSymbolic_MPIAIJ_MPIAIJ_scalable()
3546: * */
3547: PetscErrorCode MatSeqAIJCompactOutExtraColumns_SeqAIJ(Mat mat, ISLocalToGlobalMapping *mapping)
3548: {
3549: Mat_SeqAIJ *aij = (Mat_SeqAIJ*)mat->data;
3550: PetscTable gid1_lid1;
3551: PetscTablePosition tpos;
3552: PetscInt gid,lid,i,j,ncols,ec;
3553: PetscInt *garray;
3554: PetscErrorCode ierr;
3559: /* use a table */
3560: PetscTableCreate(mat->rmap->n,mat->cmap->N+1,&gid1_lid1);
3561: ec = 0;
3562: for (i=0; i<mat->rmap->n; i++) {
3563: ncols = aij->i[i+1] - aij->i[i];
3564: for (j=0; j<ncols; j++) {
3565: PetscInt data,gid1 = aij->j[aij->i[i] + j] + 1;
3566: PetscTableFind(gid1_lid1,gid1,&data);
3567: if (!data) {
3568: /* one based table */
3569: PetscTableAdd(gid1_lid1,gid1,++ec,INSERT_VALUES);
3570: }
3571: }
3572: }
3573: /* form array of columns we need */
3574: PetscMalloc1(ec+1,&garray);
3575: PetscTableGetHeadPosition(gid1_lid1,&tpos);
3576: while (tpos) {
3577: PetscTableGetNext(gid1_lid1,&tpos,&gid,&lid);
3578: gid--;
3579: lid--;
3580: garray[lid] = gid;
3581: }
3582: PetscSortInt(ec,garray); /* sort, and rebuild */
3583: PetscTableRemoveAll(gid1_lid1);
3584: for (i=0; i<ec; i++) {
3585: PetscTableAdd(gid1_lid1,garray[i]+1,i+1,INSERT_VALUES);
3586: }
3587: /* compact out the extra columns in B */
3588: for (i=0; i<mat->rmap->n; i++) {
3589: ncols = aij->i[i+1] - aij->i[i];
3590: for (j=0; j<ncols; j++) {
3591: PetscInt gid1 = aij->j[aij->i[i] + j] + 1;
3592: PetscTableFind(gid1_lid1,gid1,&lid);
3593: lid--;
3594: aij->j[aij->i[i] + j] = lid;
3595: }
3596: }
3597: PetscLayoutDestroy(&mat->cmap);
3598: PetscLayoutCreateFromSizes(PetscObjectComm((PetscObject)mat),ec,ec,1,&mat->cmap);
3599: PetscTableDestroy(&gid1_lid1);
3600: ISLocalToGlobalMappingCreate(PETSC_COMM_SELF,mat->cmap->bs,mat->cmap->n,garray,PETSC_OWN_POINTER,mapping);
3601: ISLocalToGlobalMappingSetType(*mapping,ISLOCALTOGLOBALMAPPINGHASH);
3602: return(0);
3603: }
3605: /*@
3606: MatSeqAIJSetColumnIndices - Set the column indices for all the rows
3607: in the matrix.
3609: Input Parameters:
3610: + mat - the SeqAIJ matrix
3611: - indices - the column indices
3613: Level: advanced
3615: Notes:
3616: This can be called if you have precomputed the nonzero structure of the
3617: matrix and want to provide it to the matrix object to improve the performance
3618: of the MatSetValues() operation.
3620: You MUST have set the correct numbers of nonzeros per row in the call to
3621: MatCreateSeqAIJ(), and the columns indices MUST be sorted.
3623: MUST be called before any calls to MatSetValues();
3625: The indices should start with zero, not one.
3627: @*/
3628: PetscErrorCode MatSeqAIJSetColumnIndices(Mat mat,PetscInt *indices)
3629: {
3635: PetscUseMethod(mat,"MatSeqAIJSetColumnIndices_C",(Mat,PetscInt*),(mat,indices));
3636: return(0);
3637: }
3639: /* ----------------------------------------------------------------------------------------*/
3641: PetscErrorCode MatStoreValues_SeqAIJ(Mat mat)
3642: {
3643: Mat_SeqAIJ *aij = (Mat_SeqAIJ*)mat->data;
3645: size_t nz = aij->i[mat->rmap->n];
3648: if (!aij->nonew) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Must call MatSetOption(A,MAT_NEW_NONZERO_LOCATIONS,PETSC_FALSE);first");
3650: /* allocate space for values if not already there */
3651: if (!aij->saved_values) {
3652: PetscMalloc1(nz+1,&aij->saved_values);
3653: PetscLogObjectMemory((PetscObject)mat,(nz+1)*sizeof(PetscScalar));
3654: }
3656: /* copy values over */
3657: PetscArraycpy(aij->saved_values,aij->a,nz);
3658: return(0);
3659: }
3661: /*@
3662: MatStoreValues - Stashes a copy of the matrix values; this allows, for
3663: example, reuse of the linear part of a Jacobian, while recomputing the
3664: nonlinear portion.
3666: Collect on Mat
3668: Input Parameters:
3669: . mat - the matrix (currently only AIJ matrices support this option)
3671: Level: advanced
3673: Common Usage, with SNESSolve():
3674: $ Create Jacobian matrix
3675: $ Set linear terms into matrix
3676: $ Apply boundary conditions to matrix, at this time matrix must have
3677: $ final nonzero structure (i.e. setting the nonlinear terms and applying
3678: $ boundary conditions again will not change the nonzero structure
3679: $ MatSetOption(mat,MAT_NEW_NONZERO_LOCATIONS,PETSC_FALSE);
3680: $ MatStoreValues(mat);
3681: $ Call SNESSetJacobian() with matrix
3682: $ In your Jacobian routine
3683: $ MatRetrieveValues(mat);
3684: $ Set nonlinear terms in matrix
3686: Common Usage without SNESSolve(), i.e. when you handle nonlinear solve yourself:
3687: $ // build linear portion of Jacobian
3688: $ MatSetOption(mat,MAT_NEW_NONZERO_LOCATIONS,PETSC_FALSE);
3689: $ MatStoreValues(mat);
3690: $ loop over nonlinear iterations
3691: $ MatRetrieveValues(mat);
3692: $ // call MatSetValues(mat,...) to set nonliner portion of Jacobian
3693: $ // call MatAssemblyBegin/End() on matrix
3694: $ Solve linear system with Jacobian
3695: $ endloop
3697: Notes:
3698: Matrix must already be assemblied before calling this routine
3699: Must set the matrix option MatSetOption(mat,MAT_NEW_NONZERO_LOCATIONS,PETSC_FALSE); before
3700: calling this routine.
3702: When this is called multiple times it overwrites the previous set of stored values
3703: and does not allocated additional space.
3705: .seealso: MatRetrieveValues()
3707: @*/
3708: PetscErrorCode MatStoreValues(Mat mat)
3709: {
3714: if (!mat->assembled) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3715: if (mat->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3716: PetscUseMethod(mat,"MatStoreValues_C",(Mat),(mat));
3717: return(0);
3718: }
3720: PetscErrorCode MatRetrieveValues_SeqAIJ(Mat mat)
3721: {
3722: Mat_SeqAIJ *aij = (Mat_SeqAIJ*)mat->data;
3724: PetscInt nz = aij->i[mat->rmap->n];
3727: if (!aij->nonew) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Must call MatSetOption(A,MAT_NEW_NONZERO_LOCATIONS,PETSC_FALSE);first");
3728: if (!aij->saved_values) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Must call MatStoreValues(A);first");
3729: /* copy values over */
3730: PetscArraycpy(aij->a,aij->saved_values,nz);
3731: return(0);
3732: }
3734: /*@
3735: MatRetrieveValues - Retrieves the copy of the matrix values; this allows, for
3736: example, reuse of the linear part of a Jacobian, while recomputing the
3737: nonlinear portion.
3739: Collect on Mat
3741: Input Parameters:
3742: . mat - the matrix (currently only AIJ matrices support this option)
3744: Level: advanced
3746: .seealso: MatStoreValues()
3748: @*/
3749: PetscErrorCode MatRetrieveValues(Mat mat)
3750: {
3755: if (!mat->assembled) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
3756: if (mat->factortype) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
3757: PetscUseMethod(mat,"MatRetrieveValues_C",(Mat),(mat));
3758: return(0);
3759: }
3762: /* --------------------------------------------------------------------------------*/
3763: /*@C
3764: MatCreateSeqAIJ - Creates a sparse matrix in AIJ (compressed row) format
3765: (the default parallel PETSc format). For good matrix assembly performance
3766: the user should preallocate the matrix storage by setting the parameter nz
3767: (or the array nnz). By setting these parameters accurately, performance
3768: during matrix assembly can be increased by more than a factor of 50.
3770: Collective
3772: Input Parameters:
3773: + comm - MPI communicator, set to PETSC_COMM_SELF
3774: . m - number of rows
3775: . n - number of columns
3776: . nz - number of nonzeros per row (same for all rows)
3777: - nnz - array containing the number of nonzeros in the various rows
3778: (possibly different for each row) or NULL
3780: Output Parameter:
3781: . A - the matrix
3783: It is recommended that one use the MatCreate(), MatSetType() and/or MatSetFromOptions(),
3784: MatXXXXSetPreallocation() paradigm instead of this routine directly.
3785: [MatXXXXSetPreallocation() is, for example, MatSeqAIJSetPreallocation]
3787: Notes:
3788: If nnz is given then nz is ignored
3790: The AIJ format (also called the Yale sparse matrix format or
3791: compressed row storage), is fully compatible with standard Fortran 77
3792: storage. That is, the stored row and column indices can begin at
3793: either one (as in Fortran) or zero. See the users' manual for details.
3795: Specify the preallocated storage with either nz or nnz (not both).
3796: Set nz=PETSC_DEFAULT and nnz=NULL for PETSc to control dynamic memory
3797: allocation. For large problems you MUST preallocate memory or you
3798: will get TERRIBLE performance, see the users' manual chapter on matrices.
3800: By default, this format uses inodes (identical nodes) when possible, to
3801: improve numerical efficiency of matrix-vector products and solves. We
3802: search for consecutive rows with the same nonzero structure, thereby
3803: reusing matrix information to achieve increased efficiency.
3805: Options Database Keys:
3806: + -mat_no_inode - Do not use inodes
3807: - -mat_inode_limit <limit> - Sets inode limit (max limit=5)
3809: Level: intermediate
3811: .seealso: MatCreate(), MatCreateAIJ(), MatSetValues(), MatSeqAIJSetColumnIndices(), MatCreateSeqAIJWithArrays()
3813: @*/
3814: PetscErrorCode MatCreateSeqAIJ(MPI_Comm comm,PetscInt m,PetscInt n,PetscInt nz,const PetscInt nnz[],Mat *A)
3815: {
3819: MatCreate(comm,A);
3820: MatSetSizes(*A,m,n,m,n);
3821: MatSetType(*A,MATSEQAIJ);
3822: MatSeqAIJSetPreallocation_SeqAIJ(*A,nz,nnz);
3823: return(0);
3824: }
3826: /*@C
3827: MatSeqAIJSetPreallocation - For good matrix assembly performance
3828: the user should preallocate the matrix storage by setting the parameter nz
3829: (or the array nnz). By setting these parameters accurately, performance
3830: during matrix assembly can be increased by more than a factor of 50.
3832: Collective
3834: Input Parameters:
3835: + B - The matrix
3836: . nz - number of nonzeros per row (same for all rows)
3837: - nnz - array containing the number of nonzeros in the various rows
3838: (possibly different for each row) or NULL
3840: Notes:
3841: If nnz is given then nz is ignored
3843: The AIJ format (also called the Yale sparse matrix format or
3844: compressed row storage), is fully compatible with standard Fortran 77
3845: storage. That is, the stored row and column indices can begin at
3846: either one (as in Fortran) or zero. See the users' manual for details.
3848: Specify the preallocated storage with either nz or nnz (not both).
3849: Set nz=PETSC_DEFAULT and nnz=NULL for PETSc to control dynamic memory
3850: allocation. For large problems you MUST preallocate memory or you
3851: will get TERRIBLE performance, see the users' manual chapter on matrices.
3853: You can call MatGetInfo() to get information on how effective the preallocation was;
3854: for example the fields mallocs,nz_allocated,nz_used,nz_unneeded;
3855: You can also run with the option -info and look for messages with the string
3856: malloc in them to see if additional memory allocation was needed.
3858: Developers: Use nz of MAT_SKIP_ALLOCATION to not allocate any space for the matrix
3859: entries or columns indices
3861: By default, this format uses inodes (identical nodes) when possible, to
3862: improve numerical efficiency of matrix-vector products and solves. We
3863: search for consecutive rows with the same nonzero structure, thereby
3864: reusing matrix information to achieve increased efficiency.
3866: Options Database Keys:
3867: + -mat_no_inode - Do not use inodes
3868: - -mat_inode_limit <limit> - Sets inode limit (max limit=5)
3870: Level: intermediate
3872: .seealso: MatCreate(), MatCreateAIJ(), MatSetValues(), MatSeqAIJSetColumnIndices(), MatCreateSeqAIJWithArrays(), MatGetInfo()
3874: @*/
3875: PetscErrorCode MatSeqAIJSetPreallocation(Mat B,PetscInt nz,const PetscInt nnz[])
3876: {
3882: PetscTryMethod(B,"MatSeqAIJSetPreallocation_C",(Mat,PetscInt,const PetscInt[]),(B,nz,nnz));
3883: return(0);
3884: }
3886: PetscErrorCode MatSeqAIJSetPreallocation_SeqAIJ(Mat B,PetscInt nz,const PetscInt *nnz)
3887: {
3888: Mat_SeqAIJ *b;
3889: PetscBool skipallocation = PETSC_FALSE,realalloc = PETSC_FALSE;
3891: PetscInt i;
3894: if (nz >= 0 || nnz) realalloc = PETSC_TRUE;
3895: if (nz == MAT_SKIP_ALLOCATION) {
3896: skipallocation = PETSC_TRUE;
3897: nz = 0;
3898: }
3899: PetscLayoutSetUp(B->rmap);
3900: PetscLayoutSetUp(B->cmap);
3902: if (nz == PETSC_DEFAULT || nz == PETSC_DECIDE) nz = 5;
3903: if (nz < 0) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"nz cannot be less than 0: value %D",nz);
3904: if (PetscUnlikelyDebug(nnz)) {
3905: for (i=0; i<B->rmap->n; i++) {
3906: if (nnz[i] < 0) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"nnz cannot be less than 0: local row %D value %D",i,nnz[i]);
3907: if (nnz[i] > B->cmap->n) SETERRQ3(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"nnz cannot be greater than row length: local row %D value %d rowlength %D",i,nnz[i],B->cmap->n);
3908: }
3909: }
3911: B->preallocated = PETSC_TRUE;
3913: b = (Mat_SeqAIJ*)B->data;
3915: if (!skipallocation) {
3916: if (!b->imax) {
3917: PetscMalloc1(B->rmap->n,&b->imax);
3918: PetscLogObjectMemory((PetscObject)B,B->rmap->n*sizeof(PetscInt));
3919: }
3920: if (!b->ilen) {
3921: /* b->ilen will count nonzeros in each row so far. */
3922: PetscCalloc1(B->rmap->n,&b->ilen);
3923: PetscLogObjectMemory((PetscObject)B,B->rmap->n*sizeof(PetscInt));
3924: } else {
3925: PetscMemzero(b->ilen,B->rmap->n*sizeof(PetscInt));
3926: }
3927: if (!b->ipre) {
3928: PetscMalloc1(B->rmap->n,&b->ipre);
3929: PetscLogObjectMemory((PetscObject)B,B->rmap->n*sizeof(PetscInt));
3930: }
3931: if (!nnz) {
3932: if (nz == PETSC_DEFAULT || nz == PETSC_DECIDE) nz = 10;
3933: else if (nz < 0) nz = 1;
3934: nz = PetscMin(nz,B->cmap->n);
3935: for (i=0; i<B->rmap->n; i++) b->imax[i] = nz;
3936: nz = nz*B->rmap->n;
3937: } else {
3938: PetscInt64 nz64 = 0;
3939: for (i=0; i<B->rmap->n; i++) {b->imax[i] = nnz[i]; nz64 += nnz[i];}
3940: PetscIntCast(nz64,&nz);
3941: }
3943: /* allocate the matrix space */
3944: /* FIXME: should B's old memory be unlogged? */
3945: MatSeqXAIJFreeAIJ(B,&b->a,&b->j,&b->i);
3946: if (B->structure_only) {
3947: PetscMalloc1(nz,&b->j);
3948: PetscMalloc1(B->rmap->n+1,&b->i);
3949: PetscLogObjectMemory((PetscObject)B,(B->rmap->n+1)*sizeof(PetscInt)+nz*sizeof(PetscInt));
3950: } else {
3951: PetscMalloc3(nz,&b->a,nz,&b->j,B->rmap->n+1,&b->i);
3952: PetscLogObjectMemory((PetscObject)B,(B->rmap->n+1)*sizeof(PetscInt)+nz*(sizeof(PetscScalar)+sizeof(PetscInt)));
3953: }
3954: b->i[0] = 0;
3955: for (i=1; i<B->rmap->n+1; i++) {
3956: b->i[i] = b->i[i-1] + b->imax[i-1];
3957: }
3958: if (B->structure_only) {
3959: b->singlemalloc = PETSC_FALSE;
3960: b->free_a = PETSC_FALSE;
3961: } else {
3962: b->singlemalloc = PETSC_TRUE;
3963: b->free_a = PETSC_TRUE;
3964: }
3965: b->free_ij = PETSC_TRUE;
3966: } else {
3967: b->free_a = PETSC_FALSE;
3968: b->free_ij = PETSC_FALSE;
3969: }
3971: if (b->ipre && nnz != b->ipre && b->imax) {
3972: /* reserve user-requested sparsity */
3973: PetscArraycpy(b->ipre,b->imax,B->rmap->n);
3974: }
3977: b->nz = 0;
3978: b->maxnz = nz;
3979: B->info.nz_unneeded = (double)b->maxnz;
3980: if (realalloc) {
3981: MatSetOption(B,MAT_NEW_NONZERO_ALLOCATION_ERR,PETSC_TRUE);
3982: }
3983: B->was_assembled = PETSC_FALSE;
3984: B->assembled = PETSC_FALSE;
3985: return(0);
3986: }
3989: PetscErrorCode MatResetPreallocation_SeqAIJ(Mat A)
3990: {
3991: Mat_SeqAIJ *a;
3992: PetscInt i;
3998: /* Check local size. If zero, then return */
3999: if (!A->rmap->n) return(0);
4001: a = (Mat_SeqAIJ*)A->data;
4002: /* if no saved info, we error out */
4003: if (!a->ipre) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_NULL,"No saved preallocation info \n");
4005: if (!a->i || !a->j || !a->a || !a->imax || !a->ilen) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_NULL,"Memory info is incomplete, and can not reset preallocation \n");
4007: PetscArraycpy(a->imax,a->ipre,A->rmap->n);
4008: PetscArrayzero(a->ilen,A->rmap->n);
4009: a->i[0] = 0;
4010: for (i=1; i<A->rmap->n+1; i++) {
4011: a->i[i] = a->i[i-1] + a->imax[i-1];
4012: }
4013: A->preallocated = PETSC_TRUE;
4014: a->nz = 0;
4015: a->maxnz = a->i[A->rmap->n];
4016: A->info.nz_unneeded = (double)a->maxnz;
4017: A->was_assembled = PETSC_FALSE;
4018: A->assembled = PETSC_FALSE;
4019: return(0);
4020: }
4022: /*@
4023: MatSeqAIJSetPreallocationCSR - Allocates memory for a sparse sequential matrix in AIJ format.
4025: Input Parameters:
4026: + B - the matrix
4027: . i - the indices into j for the start of each row (starts with zero)
4028: . j - the column indices for each row (starts with zero) these must be sorted for each row
4029: - v - optional values in the matrix
4031: Level: developer
4033: The i,j,v values are COPIED with this routine; to avoid the copy use MatCreateSeqAIJWithArrays()
4035: .seealso: MatCreate(), MatCreateSeqAIJ(), MatSetValues(), MatSeqAIJSetPreallocation(), MatCreateSeqAIJ(), MATSEQAIJ
4036: @*/
4037: PetscErrorCode MatSeqAIJSetPreallocationCSR(Mat B,const PetscInt i[],const PetscInt j[],const PetscScalar v[])
4038: {
4044: PetscTryMethod(B,"MatSeqAIJSetPreallocationCSR_C",(Mat,const PetscInt[],const PetscInt[],const PetscScalar[]),(B,i,j,v));
4045: return(0);
4046: }
4048: PetscErrorCode MatSeqAIJSetPreallocationCSR_SeqAIJ(Mat B,const PetscInt Ii[],const PetscInt J[],const PetscScalar v[])
4049: {
4050: PetscInt i;
4051: PetscInt m,n;
4052: PetscInt nz;
4053: PetscInt *nnz, nz_max = 0;
4057: if (Ii[0]) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE, "Ii[0] must be 0 it is %D", Ii[0]);
4059: PetscLayoutSetUp(B->rmap);
4060: PetscLayoutSetUp(B->cmap);
4062: MatGetSize(B, &m, &n);
4063: PetscMalloc1(m+1, &nnz);
4064: for (i = 0; i < m; i++) {
4065: nz = Ii[i+1]- Ii[i];
4066: nz_max = PetscMax(nz_max, nz);
4067: if (nz < 0) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE, "Local row %D has a negative number of columns %D", i, nnz);
4068: nnz[i] = nz;
4069: }
4070: MatSeqAIJSetPreallocation(B, 0, nnz);
4071: PetscFree(nnz);
4073: for (i = 0; i < m; i++) {
4074: MatSetValues_SeqAIJ(B, 1, &i, Ii[i+1] - Ii[i], J+Ii[i], v ? v + Ii[i] : NULL, INSERT_VALUES);
4075: }
4077: MatAssemblyBegin(B,MAT_FINAL_ASSEMBLY);
4078: MatAssemblyEnd(B,MAT_FINAL_ASSEMBLY);
4080: MatSetOption(B,MAT_NEW_NONZERO_LOCATION_ERR,PETSC_TRUE);
4081: return(0);
4082: }
4084: #include <../src/mat/impls/dense/seq/dense.h>
4085: #include <petsc/private/kernels/petscaxpy.h>
4087: /*
4088: Computes (B'*A')' since computing B*A directly is untenable
4090: n p p
4091: ( ) ( ) ( )
4092: m ( A ) * n ( B ) = m ( C )
4093: ( ) ( ) ( )
4095: */
4096: PetscErrorCode MatMatMultNumeric_SeqDense_SeqAIJ(Mat A,Mat B,Mat C)
4097: {
4098: PetscErrorCode ierr;
4099: Mat_SeqDense *sub_a = (Mat_SeqDense*)A->data;
4100: Mat_SeqAIJ *sub_b = (Mat_SeqAIJ*)B->data;
4101: Mat_SeqDense *sub_c = (Mat_SeqDense*)C->data;
4102: PetscInt i,j,n,m,q,p;
4103: const PetscInt *ii,*idx;
4104: const PetscScalar *b,*a,*a_q;
4105: PetscScalar *c,*c_q;
4106: PetscInt clda = sub_c->lda;
4107: PetscInt alda = sub_a->lda;
4110: m = A->rmap->n;
4111: n = A->cmap->n;
4112: p = B->cmap->n;
4113: a = sub_a->v;
4114: b = sub_b->a;
4115: c = sub_c->v;
4116: if (clda == m) {
4117: PetscArrayzero(c,m*p);
4118: } else {
4119: for (j=0;j<p;j++)
4120: for (i=0;i<m;i++)
4121: c[j*clda + i] = 0.0;
4122: }
4123: ii = sub_b->i;
4124: idx = sub_b->j;
4125: for (i=0; i<n; i++) {
4126: q = ii[i+1] - ii[i];
4127: while (q-->0) {
4128: c_q = c + clda*(*idx);
4129: a_q = a + alda*i;
4130: PetscKernelAXPY(c_q,*b,a_q,m);
4131: idx++;
4132: b++;
4133: }
4134: }
4135: return(0);
4136: }
4138: PetscErrorCode MatMatMultSymbolic_SeqDense_SeqAIJ(Mat A,Mat B,PetscReal fill,Mat C)
4139: {
4141: PetscInt m=A->rmap->n,n=B->cmap->n;
4142: PetscBool cisdense;
4145: 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);
4146: MatSetSizes(C,m,n,m,n);
4147: MatSetBlockSizesFromMats(C,A,B);
4148: PetscObjectTypeCompareAny((PetscObject)C,&cisdense,MATSEQDENSE,MATSEQDENSECUDA,"");
4149: if (!cisdense) {
4150: MatSetType(C,MATDENSE);
4151: }
4152: MatSetUp(C);
4154: C->ops->matmultnumeric = MatMatMultNumeric_SeqDense_SeqAIJ;
4155: return(0);
4156: }
4158: /* ----------------------------------------------------------------*/
4159: /*MC
4160: MATSEQAIJ - MATSEQAIJ = "seqaij" - A matrix type to be used for sequential sparse matrices,
4161: based on compressed sparse row format.
4163: Options Database Keys:
4164: . -mat_type seqaij - sets the matrix type to "seqaij" during a call to MatSetFromOptions()
4166: Level: beginner
4168: Notes:
4169: MatSetValues() may be called for this matrix type with a NULL argument for the numerical values,
4170: in this case the values associated with the rows and columns one passes in are set to zero
4171: in the matrix
4173: MatSetOptions(,MAT_STRUCTURE_ONLY,PETSC_TRUE) may be called for this matrix type. In this no
4174: space is allocated for the nonzero entries and any entries passed with MatSetValues() are ignored
4176: Developer Notes:
4177: It would be nice if all matrix formats supported passing NULL in for the numerical values
4179: .seealso: MatCreateSeqAIJ(), MatSetFromOptions(), MatSetType(), MatCreate(), MatType
4180: M*/
4182: /*MC
4183: MATAIJ - MATAIJ = "aij" - A matrix type to be used for sparse matrices.
4185: This matrix type is identical to MATSEQAIJ when constructed with a single process communicator,
4186: and MATMPIAIJ otherwise. As a result, for single process communicators,
4187: MatSeqAIJSetPreallocation is supported, and similarly MatMPIAIJSetPreallocation() is supported
4188: for communicators controlling multiple processes. It is recommended that you call both of
4189: the above preallocation routines for simplicity.
4191: Options Database Keys:
4192: . -mat_type aij - sets the matrix type to "aij" during a call to MatSetFromOptions()
4194: Developer Notes:
4195: Subclasses include MATAIJCUSPARSE, MATAIJPERM, MATAIJSELL, MATAIJMKL, MATAIJCRL, and also automatically switches over to use inodes when
4196: enough exist.
4198: Level: beginner
4200: .seealso: MatCreateAIJ(), MatCreateSeqAIJ(), MATSEQAIJ,MATMPIAIJ
4201: M*/
4203: /*MC
4204: MATAIJCRL - MATAIJCRL = "aijcrl" - A matrix type to be used for sparse matrices.
4206: This matrix type is identical to MATSEQAIJCRL when constructed with a single process communicator,
4207: and MATMPIAIJCRL otherwise. As a result, for single process communicators,
4208: MatSeqAIJSetPreallocation() is supported, and similarly MatMPIAIJSetPreallocation() is supported
4209: for communicators controlling multiple processes. It is recommended that you call both of
4210: the above preallocation routines for simplicity.
4212: Options Database Keys:
4213: . -mat_type aijcrl - sets the matrix type to "aijcrl" during a call to MatSetFromOptions()
4215: Level: beginner
4217: .seealso: MatCreateMPIAIJCRL,MATSEQAIJCRL,MATMPIAIJCRL, MATSEQAIJCRL, MATMPIAIJCRL
4218: M*/
4220: PETSC_INTERN PetscErrorCode MatConvert_SeqAIJ_SeqAIJCRL(Mat,MatType,MatReuse,Mat*);
4221: #if defined(PETSC_HAVE_ELEMENTAL)
4222: PETSC_INTERN PetscErrorCode MatConvert_SeqAIJ_Elemental(Mat,MatType,MatReuse,Mat*);
4223: #endif
4224: #if defined(PETSC_HAVE_SCALAPACK)
4225: PETSC_INTERN PetscErrorCode MatConvert_AIJ_ScaLAPACK(Mat,MatType,MatReuse,Mat*);
4226: #endif
4227: #if defined(PETSC_HAVE_HYPRE)
4228: PETSC_INTERN PetscErrorCode MatConvert_AIJ_HYPRE(Mat A,MatType,MatReuse,Mat*);
4229: #endif
4230: PETSC_INTERN PetscErrorCode MatConvert_SeqAIJ_SeqDense(Mat,MatType,MatReuse,Mat*);
4232: PETSC_EXTERN PetscErrorCode MatConvert_SeqAIJ_SeqSELL(Mat,MatType,MatReuse,Mat*);
4233: PETSC_INTERN PetscErrorCode MatConvert_XAIJ_IS(Mat,MatType,MatReuse,Mat*);
4234: PETSC_INTERN PetscErrorCode MatProductSetFromOptions_IS_XAIJ(Mat);
4236: /*@C
4237: MatSeqAIJGetArray - gives read/write access to the array where the data for a MATSEQAIJ matrix is stored
4239: Not Collective
4241: Input Parameter:
4242: . mat - a MATSEQAIJ matrix
4244: Output Parameter:
4245: . array - pointer to the data
4247: Level: intermediate
4249: .seealso: MatSeqAIJRestoreArray(), MatSeqAIJGetArrayF90()
4250: @*/
4251: PetscErrorCode MatSeqAIJGetArray(Mat A,PetscScalar **array)
4252: {
4256: PetscUseMethod(A,"MatSeqAIJGetArray_C",(Mat,PetscScalar**),(A,array));
4257: return(0);
4258: }
4260: /*@C
4261: MatSeqAIJGetArrayRead - gives read-only access to the array where the data for a MATSEQAIJ matrix is stored
4263: Not Collective
4265: Input Parameter:
4266: . mat - a MATSEQAIJ matrix
4268: Output Parameter:
4269: . array - pointer to the data
4271: Level: intermediate
4273: .seealso: MatSeqAIJGetArray(), MatSeqAIJRestoreArrayRead()
4274: @*/
4275: PetscErrorCode MatSeqAIJGetArrayRead(Mat A,const PetscScalar **array)
4276: {
4277: #if defined(PETSC_HAVE_CUDA) || defined(PETSC_HAVE_VIENNACL)
4278: PetscOffloadMask oval;
4279: #endif
4283: #if defined(PETSC_HAVE_CUDA) || defined(PETSC_HAVE_VIENNACL)
4284: oval = A->offloadmask;
4285: #endif
4286: MatSeqAIJGetArray(A,(PetscScalar**)array);
4287: #if defined(PETSC_HAVE_CUDA) || defined(PETSC_HAVE_VIENNACL)
4288: if (oval == PETSC_OFFLOAD_GPU || oval == PETSC_OFFLOAD_BOTH) A->offloadmask = PETSC_OFFLOAD_BOTH;
4289: #endif
4290: return(0);
4291: }
4293: /*@C
4294: MatSeqAIJRestoreArrayRead - restore the read-only access array obtained from MatSeqAIJGetArrayRead
4296: Not Collective
4298: Input Parameter:
4299: . mat - a MATSEQAIJ matrix
4301: Output Parameter:
4302: . array - pointer to the data
4304: Level: intermediate
4306: .seealso: MatSeqAIJGetArray(), MatSeqAIJGetArrayRead()
4307: @*/
4308: PetscErrorCode MatSeqAIJRestoreArrayRead(Mat A,const PetscScalar **array)
4309: {
4310: #if defined(PETSC_HAVE_CUDA) || defined(PETSC_HAVE_VIENNACL)
4311: PetscOffloadMask oval;
4312: #endif
4316: #if defined(PETSC_HAVE_CUDA) || defined(PETSC_HAVE_VIENNACL)
4317: oval = A->offloadmask;
4318: #endif
4319: MatSeqAIJRestoreArray(A,(PetscScalar**)array);
4320: #if defined(PETSC_HAVE_CUDA) || defined(PETSC_HAVE_VIENNACL)
4321: A->offloadmask = oval;
4322: #endif
4323: return(0);
4324: }
4326: /*@C
4327: MatSeqAIJGetMaxRowNonzeros - returns the maximum number of nonzeros in any row
4329: Not Collective
4331: Input Parameter:
4332: . mat - a MATSEQAIJ matrix
4334: Output Parameter:
4335: . nz - the maximum number of nonzeros in any row
4337: Level: intermediate
4339: .seealso: MatSeqAIJRestoreArray(), MatSeqAIJGetArrayF90()
4340: @*/
4341: PetscErrorCode MatSeqAIJGetMaxRowNonzeros(Mat A,PetscInt *nz)
4342: {
4343: Mat_SeqAIJ *aij = (Mat_SeqAIJ*)A->data;
4346: *nz = aij->rmax;
4347: return(0);
4348: }
4350: /*@C
4351: MatSeqAIJRestoreArray - returns access to the array where the data for a MATSEQAIJ matrix is stored obtained by MatSeqAIJGetArray()
4353: Not Collective
4355: Input Parameters:
4356: + mat - a MATSEQAIJ matrix
4357: - array - pointer to the data
4359: Level: intermediate
4361: .seealso: MatSeqAIJGetArray(), MatSeqAIJRestoreArrayF90()
4362: @*/
4363: PetscErrorCode MatSeqAIJRestoreArray(Mat A,PetscScalar **array)
4364: {
4368: PetscUseMethod(A,"MatSeqAIJRestoreArray_C",(Mat,PetscScalar**),(A,array));
4369: return(0);
4370: }
4372: #if defined(PETSC_HAVE_CUDA)
4373: PETSC_INTERN PetscErrorCode MatConvert_SeqAIJ_SeqAIJCUSPARSE(Mat);
4374: #endif
4376: PETSC_EXTERN PetscErrorCode MatCreate_SeqAIJ(Mat B)
4377: {
4378: Mat_SeqAIJ *b;
4380: PetscMPIInt size;
4383: MPI_Comm_size(PetscObjectComm((PetscObject)B),&size);
4384: if (size > 1) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Comm must be of size 1");
4386: PetscNewLog(B,&b);
4388: B->data = (void*)b;
4390: PetscMemcpy(B->ops,&MatOps_Values,sizeof(struct _MatOps));
4391: if (B->sortedfull) B->ops->setvalues = MatSetValues_SeqAIJ_SortedFull;
4393: b->row = 0;
4394: b->col = 0;
4395: b->icol = 0;
4396: b->reallocs = 0;
4397: b->ignorezeroentries = PETSC_FALSE;
4398: b->roworiented = PETSC_TRUE;
4399: b->nonew = 0;
4400: b->diag = 0;
4401: b->solve_work = 0;
4402: B->spptr = 0;
4403: b->saved_values = 0;
4404: b->idiag = 0;
4405: b->mdiag = 0;
4406: b->ssor_work = 0;
4407: b->omega = 1.0;
4408: b->fshift = 0.0;
4409: b->idiagvalid = PETSC_FALSE;
4410: b->ibdiagvalid = PETSC_FALSE;
4411: b->keepnonzeropattern = PETSC_FALSE;
4413: PetscObjectChangeTypeName((PetscObject)B,MATSEQAIJ);
4414: PetscObjectComposeFunction((PetscObject)B,"MatSeqAIJGetArray_C",MatSeqAIJGetArray_SeqAIJ);
4415: PetscObjectComposeFunction((PetscObject)B,"MatSeqAIJRestoreArray_C",MatSeqAIJRestoreArray_SeqAIJ);
4417: #if defined(PETSC_HAVE_MATLAB_ENGINE)
4418: PetscObjectComposeFunction((PetscObject)B,"PetscMatlabEnginePut_C",MatlabEnginePut_SeqAIJ);
4419: PetscObjectComposeFunction((PetscObject)B,"PetscMatlabEngineGet_C",MatlabEngineGet_SeqAIJ);
4420: #endif
4422: PetscObjectComposeFunction((PetscObject)B,"MatSeqAIJSetColumnIndices_C",MatSeqAIJSetColumnIndices_SeqAIJ);
4423: PetscObjectComposeFunction((PetscObject)B,"MatStoreValues_C",MatStoreValues_SeqAIJ);
4424: PetscObjectComposeFunction((PetscObject)B,"MatRetrieveValues_C",MatRetrieveValues_SeqAIJ);
4425: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqaij_seqsbaij_C",MatConvert_SeqAIJ_SeqSBAIJ);
4426: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqaij_seqbaij_C",MatConvert_SeqAIJ_SeqBAIJ);
4427: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqaij_seqaijperm_C",MatConvert_SeqAIJ_SeqAIJPERM);
4428: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqaij_seqaijsell_C",MatConvert_SeqAIJ_SeqAIJSELL);
4429: #if defined(PETSC_HAVE_MKL_SPARSE)
4430: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqaij_seqaijmkl_C",MatConvert_SeqAIJ_SeqAIJMKL);
4431: #endif
4432: #if defined(PETSC_HAVE_CUDA)
4433: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqaij_seqaijcusparse_C",MatConvert_SeqAIJ_SeqAIJCUSPARSE);
4434: PetscObjectComposeFunction((PetscObject)B,"MatProductSetFromOptions_seqaijcusparse_seqaij_C",MatProductSetFromOptions_SeqAIJ);
4435: #endif
4436: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqaij_seqaijcrl_C",MatConvert_SeqAIJ_SeqAIJCRL);
4437: #if defined(PETSC_HAVE_ELEMENTAL)
4438: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqaij_elemental_C",MatConvert_SeqAIJ_Elemental);
4439: #endif
4440: #if defined(PETSC_HAVE_SCALAPACK)
4441: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqaij_scalapack_C",MatConvert_AIJ_ScaLAPACK);
4442: #endif
4443: #if defined(PETSC_HAVE_HYPRE)
4444: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqaij_hypre_C",MatConvert_AIJ_HYPRE);
4445: PetscObjectComposeFunction((PetscObject)B,"MatProductSetFromOptions_transpose_seqaij_seqaij_C",MatProductSetFromOptions_Transpose_AIJ_AIJ);
4446: #endif
4447: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqaij_seqdense_C",MatConvert_SeqAIJ_SeqDense);
4448: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqaij_seqsell_C",MatConvert_SeqAIJ_SeqSELL);
4449: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqaij_is_C",MatConvert_XAIJ_IS);
4450: PetscObjectComposeFunction((PetscObject)B,"MatIsTranspose_C",MatIsTranspose_SeqAIJ);
4451: PetscObjectComposeFunction((PetscObject)B,"MatIsHermitianTranspose_C",MatIsTranspose_SeqAIJ);
4452: PetscObjectComposeFunction((PetscObject)B,"MatSeqAIJSetPreallocation_C",MatSeqAIJSetPreallocation_SeqAIJ);
4453: PetscObjectComposeFunction((PetscObject)B,"MatResetPreallocation_C",MatResetPreallocation_SeqAIJ);
4454: PetscObjectComposeFunction((PetscObject)B,"MatSeqAIJSetPreallocationCSR_C",MatSeqAIJSetPreallocationCSR_SeqAIJ);
4455: PetscObjectComposeFunction((PetscObject)B,"MatReorderForNonzeroDiagonal_C",MatReorderForNonzeroDiagonal_SeqAIJ);
4456: PetscObjectComposeFunction((PetscObject)B,"MatProductSetFromOptions_is_seqaij_C",MatProductSetFromOptions_IS_XAIJ);
4457: PetscObjectComposeFunction((PetscObject)B,"MatProductSetFromOptions_seqdense_seqaij_C",MatProductSetFromOptions_SeqDense_SeqAIJ);
4458: PetscObjectComposeFunction((PetscObject)B,"MatProductSetFromOptions_seqaij_seqaij_C",MatProductSetFromOptions_SeqAIJ);
4459: MatCreate_SeqAIJ_Inode(B);
4460: PetscObjectChangeTypeName((PetscObject)B,MATSEQAIJ);
4461: MatSeqAIJSetTypeFromOptions(B); /* this allows changing the matrix subtype to say MATSEQAIJPERM */
4462: return(0);
4463: }
4465: /*
4466: Given a matrix generated with MatGetFactor() duplicates all the information in A into B
4467: */
4468: PetscErrorCode MatDuplicateNoCreate_SeqAIJ(Mat C,Mat A,MatDuplicateOption cpvalues,PetscBool mallocmatspace)
4469: {
4470: Mat_SeqAIJ *c,*a = (Mat_SeqAIJ*)A->data;
4472: PetscInt m = A->rmap->n,i;
4475: c = (Mat_SeqAIJ*)C->data;
4477: C->factortype = A->factortype;
4478: c->row = 0;
4479: c->col = 0;
4480: c->icol = 0;
4481: c->reallocs = 0;
4483: C->assembled = PETSC_TRUE;
4485: PetscLayoutReference(A->rmap,&C->rmap);
4486: PetscLayoutReference(A->cmap,&C->cmap);
4488: PetscMalloc1(m,&c->imax);
4489: PetscMemcpy(c->imax,a->imax,m*sizeof(PetscInt));
4490: PetscMalloc1(m,&c->ilen);
4491: PetscMemcpy(c->ilen,a->ilen,m*sizeof(PetscInt));
4492: PetscLogObjectMemory((PetscObject)C, 2*m*sizeof(PetscInt));
4494: /* allocate the matrix space */
4495: if (mallocmatspace) {
4496: PetscMalloc3(a->i[m],&c->a,a->i[m],&c->j,m+1,&c->i);
4497: PetscLogObjectMemory((PetscObject)C, a->i[m]*(sizeof(PetscScalar)+sizeof(PetscInt))+(m+1)*sizeof(PetscInt));
4499: c->singlemalloc = PETSC_TRUE;
4501: PetscArraycpy(c->i,a->i,m+1);
4502: if (m > 0) {
4503: PetscArraycpy(c->j,a->j,a->i[m]);
4504: if (cpvalues == MAT_COPY_VALUES) {
4505: PetscArraycpy(c->a,a->a,a->i[m]);
4506: } else {
4507: PetscArrayzero(c->a,a->i[m]);
4508: }
4509: }
4510: }
4512: c->ignorezeroentries = a->ignorezeroentries;
4513: c->roworiented = a->roworiented;
4514: c->nonew = a->nonew;
4515: if (a->diag) {
4516: PetscMalloc1(m+1,&c->diag);
4517: PetscMemcpy(c->diag,a->diag,m*sizeof(PetscInt));
4518: PetscLogObjectMemory((PetscObject)C,(m+1)*sizeof(PetscInt));
4519: } else c->diag = NULL;
4521: c->solve_work = 0;
4522: c->saved_values = 0;
4523: c->idiag = 0;
4524: c->ssor_work = 0;
4525: c->keepnonzeropattern = a->keepnonzeropattern;
4526: c->free_a = PETSC_TRUE;
4527: c->free_ij = PETSC_TRUE;
4529: c->rmax = a->rmax;
4530: c->nz = a->nz;
4531: c->maxnz = a->nz; /* Since we allocate exactly the right amount */
4532: C->preallocated = PETSC_TRUE;
4534: c->compressedrow.use = a->compressedrow.use;
4535: c->compressedrow.nrows = a->compressedrow.nrows;
4536: if (a->compressedrow.use) {
4537: i = a->compressedrow.nrows;
4538: PetscMalloc2(i+1,&c->compressedrow.i,i,&c->compressedrow.rindex);
4539: PetscArraycpy(c->compressedrow.i,a->compressedrow.i,i+1);
4540: PetscArraycpy(c->compressedrow.rindex,a->compressedrow.rindex,i);
4541: } else {
4542: c->compressedrow.use = PETSC_FALSE;
4543: c->compressedrow.i = NULL;
4544: c->compressedrow.rindex = NULL;
4545: }
4546: c->nonzerorowcnt = a->nonzerorowcnt;
4547: C->nonzerostate = A->nonzerostate;
4549: MatDuplicate_SeqAIJ_Inode(A,cpvalues,&C);
4550: PetscFunctionListDuplicate(((PetscObject)A)->qlist,&((PetscObject)C)->qlist);
4551: return(0);
4552: }
4554: PetscErrorCode MatDuplicate_SeqAIJ(Mat A,MatDuplicateOption cpvalues,Mat *B)
4555: {
4559: MatCreate(PetscObjectComm((PetscObject)A),B);
4560: MatSetSizes(*B,A->rmap->n,A->cmap->n,A->rmap->n,A->cmap->n);
4561: if (!(A->rmap->n % A->rmap->bs) && !(A->cmap->n % A->cmap->bs)) {
4562: MatSetBlockSizesFromMats(*B,A,A);
4563: }
4564: MatSetType(*B,((PetscObject)A)->type_name);
4565: MatDuplicateNoCreate_SeqAIJ(*B,A,cpvalues,PETSC_TRUE);
4566: return(0);
4567: }
4569: PetscErrorCode MatLoad_SeqAIJ(Mat newMat, PetscViewer viewer)
4570: {
4571: PetscBool isbinary, ishdf5;
4577: /* force binary viewer to load .info file if it has not yet done so */
4578: PetscViewerSetUp(viewer);
4579: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERBINARY,&isbinary);
4580: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERHDF5, &ishdf5);
4581: if (isbinary) {
4582: MatLoad_SeqAIJ_Binary(newMat,viewer);
4583: } else if (ishdf5) {
4584: #if defined(PETSC_HAVE_HDF5)
4585: MatLoad_AIJ_HDF5(newMat,viewer);
4586: #else
4587: SETERRQ(PetscObjectComm((PetscObject)newMat),PETSC_ERR_SUP,"HDF5 not supported in this build.\nPlease reconfigure using --download-hdf5");
4588: #endif
4589: } else {
4590: 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);
4591: }
4592: return(0);
4593: }
4595: PetscErrorCode MatLoad_SeqAIJ_Binary(Mat mat, PetscViewer viewer)
4596: {
4597: Mat_SeqAIJ *a = (Mat_SeqAIJ*)mat->data;
4599: PetscInt header[4],*rowlens,M,N,nz,sum,rows,cols,i;
4602: PetscViewerSetUp(viewer);
4604: /* read in matrix header */
4605: PetscViewerBinaryRead(viewer,header,4,NULL,PETSC_INT);
4606: if (header[0] != MAT_FILE_CLASSID) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_FILE_UNEXPECTED,"Not a matrix object in file");
4607: M = header[1]; N = header[2]; nz = header[3];
4608: if (M < 0) SETERRQ1(PetscObjectComm((PetscObject)viewer),PETSC_ERR_FILE_UNEXPECTED,"Matrix row size (%D) in file is negative",M);
4609: if (N < 0) SETERRQ1(PetscObjectComm((PetscObject)viewer),PETSC_ERR_FILE_UNEXPECTED,"Matrix column size (%D) in file is negative",N);
4610: if (nz < 0) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_FILE_UNEXPECTED,"Matrix stored in special format on disk, cannot load as SeqAIJ");
4612: /* set block sizes from the viewer's .info file */
4613: MatLoad_Binary_BlockSizes(mat,viewer);
4614: /* set local and global sizes if not set already */
4615: if (mat->rmap->n < 0) mat->rmap->n = M;
4616: if (mat->cmap->n < 0) mat->cmap->n = N;
4617: if (mat->rmap->N < 0) mat->rmap->N = M;
4618: if (mat->cmap->N < 0) mat->cmap->N = N;
4619: PetscLayoutSetUp(mat->rmap);
4620: PetscLayoutSetUp(mat->cmap);
4622: /* check if the matrix sizes are correct */
4623: MatGetSize(mat,&rows,&cols);
4624: if (M != rows || N != cols) SETERRQ4(PETSC_COMM_SELF,PETSC_ERR_FILE_UNEXPECTED, "Matrix in file of different sizes (%D, %D) than the input matrix (%D, %D)",M,N,rows,cols);
4626: /* read in row lengths */
4627: PetscMalloc1(M,&rowlens);
4628: PetscViewerBinaryRead(viewer,rowlens,M,NULL,PETSC_INT);
4629: /* check if sum(rowlens) is same as nz */
4630: sum = 0; for (i=0; i<M; i++) sum += rowlens[i];
4631: if (sum != nz) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_FILE_UNEXPECTED,"Inconsistent matrix data in file: nonzeros = %D, sum-row-lengths = %D\n",nz,sum);
4632: /* preallocate and check sizes */
4633: MatSeqAIJSetPreallocation_SeqAIJ(mat,0,rowlens);
4634: MatGetSize(mat,&rows,&cols);
4635: 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);
4636: /* store row lengths */
4637: PetscArraycpy(a->ilen,rowlens,M);
4638: PetscFree(rowlens);
4640: /* fill in "i" row pointers */
4641: a->i[0] = 0; for (i=0; i<M; i++) a->i[i+1] = a->i[i] + a->ilen[i];
4642: /* read in "j" column indices */
4643: PetscViewerBinaryRead(viewer,a->j,nz,NULL,PETSC_INT);
4644: /* read in "a" nonzero values */
4645: PetscViewerBinaryRead(viewer,a->a,nz,NULL,PETSC_SCALAR);
4647: MatAssemblyBegin(mat,MAT_FINAL_ASSEMBLY);
4648: MatAssemblyEnd(mat,MAT_FINAL_ASSEMBLY);
4649: return(0);
4650: }
4652: PetscErrorCode MatEqual_SeqAIJ(Mat A,Mat B,PetscBool * flg)
4653: {
4654: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data,*b = (Mat_SeqAIJ*)B->data;
4656: #if defined(PETSC_USE_COMPLEX)
4657: PetscInt k;
4658: #endif
4661: /* If the matrix dimensions are not equal,or no of nonzeros */
4662: if ((A->rmap->n != B->rmap->n) || (A->cmap->n != B->cmap->n) ||(a->nz != b->nz)) {
4663: *flg = PETSC_FALSE;
4664: return(0);
4665: }
4667: /* if the a->i are the same */
4668: PetscArraycmp(a->i,b->i,A->rmap->n+1,flg);
4669: if (!*flg) return(0);
4671: /* if a->j are the same */
4672: PetscArraycmp(a->j,b->j,a->nz,flg);
4673: if (!*flg) return(0);
4675: /* if a->a are the same */
4676: #if defined(PETSC_USE_COMPLEX)
4677: for (k=0; k<a->nz; k++) {
4678: if (PetscRealPart(a->a[k]) != PetscRealPart(b->a[k]) || PetscImaginaryPart(a->a[k]) != PetscImaginaryPart(b->a[k])) {
4679: *flg = PETSC_FALSE;
4680: return(0);
4681: }
4682: }
4683: #else
4684: PetscArraycmp(a->a,b->a,a->nz,flg);
4685: #endif
4686: return(0);
4687: }
4689: /*@
4690: MatCreateSeqAIJWithArrays - Creates an sequential AIJ matrix using matrix elements (in CSR format)
4691: provided by the user.
4693: Collective
4695: Input Parameters:
4696: + comm - must be an MPI communicator of size 1
4697: . m - number of rows
4698: . n - number of columns
4699: . i - row indices; that is i[0] = 0, i[row] = i[row-1] + number of elements in that row of the matrix
4700: . j - column indices
4701: - a - matrix values
4703: Output Parameter:
4704: . mat - the matrix
4706: Level: intermediate
4708: Notes:
4709: The i, j, and a arrays are not copied by this routine, the user must free these arrays
4710: once the matrix is destroyed and not before
4712: You cannot set new nonzero locations into this matrix, that will generate an error.
4714: The i and j indices are 0 based
4716: The format which is used for the sparse matrix input, is equivalent to a
4717: row-major ordering.. i.e for the following matrix, the input data expected is
4718: as shown
4720: $ 1 0 0
4721: $ 2 0 3
4722: $ 4 5 6
4723: $
4724: $ i = {0,1,3,6} [size = nrow+1 = 3+1]
4725: $ j = {0,0,2,0,1,2} [size = 6]; values must be sorted for each row
4726: $ v = {1,2,3,4,5,6} [size = 6]
4729: .seealso: MatCreate(), MatCreateAIJ(), MatCreateSeqAIJ(), MatCreateMPIAIJWithArrays(), MatMPIAIJSetPreallocationCSR()
4731: @*/
4732: PetscErrorCode MatCreateSeqAIJWithArrays(MPI_Comm comm,PetscInt m,PetscInt n,PetscInt i[],PetscInt j[],PetscScalar a[],Mat *mat)
4733: {
4735: PetscInt ii;
4736: Mat_SeqAIJ *aij;
4737: PetscInt jj;
4740: if (m > 0 && i[0]) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"i (row indices) must start with 0");
4741: MatCreate(comm,mat);
4742: MatSetSizes(*mat,m,n,m,n);
4743: /* MatSetBlockSizes(*mat,,); */
4744: MatSetType(*mat,MATSEQAIJ);
4745: MatSeqAIJSetPreallocation_SeqAIJ(*mat,MAT_SKIP_ALLOCATION,0);
4746: aij = (Mat_SeqAIJ*)(*mat)->data;
4747: PetscMalloc1(m,&aij->imax);
4748: PetscMalloc1(m,&aij->ilen);
4750: aij->i = i;
4751: aij->j = j;
4752: aij->a = a;
4753: aij->singlemalloc = PETSC_FALSE;
4754: aij->nonew = -1; /*this indicates that inserting a new value in the matrix that generates a new nonzero is an error*/
4755: aij->free_a = PETSC_FALSE;
4756: aij->free_ij = PETSC_FALSE;
4758: for (ii=0; ii<m; ii++) {
4759: aij->ilen[ii] = aij->imax[ii] = i[ii+1] - i[ii];
4760: if (PetscDefined(USE_DEBUG)) {
4761: 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]);
4762: for (jj=i[ii]+1; jj<i[ii+1]; jj++) {
4763: 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);
4764: 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);
4765: }
4766: }
4767: }
4768: if (PetscDefined(USE_DEBUG)) {
4769: for (ii=0; ii<aij->i[m]; ii++) {
4770: if (j[ii] < 0) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Negative column index at location = %D index = %D",ii,j[ii]);
4771: 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]);
4772: }
4773: }
4775: MatAssemblyBegin(*mat,MAT_FINAL_ASSEMBLY);
4776: MatAssemblyEnd(*mat,MAT_FINAL_ASSEMBLY);
4777: return(0);
4778: }
4779: /*@C
4780: MatCreateSeqAIJFromTriple - Creates an sequential AIJ matrix using matrix elements (in COO format)
4781: provided by the user.
4783: Collective
4785: Input Parameters:
4786: + comm - must be an MPI communicator of size 1
4787: . m - number of rows
4788: . n - number of columns
4789: . i - row indices
4790: . j - column indices
4791: . a - matrix values
4792: . nz - number of nonzeros
4793: - idx - 0 or 1 based
4795: Output Parameter:
4796: . mat - the matrix
4798: Level: intermediate
4800: Notes:
4801: The i and j indices are 0 based
4803: The format which is used for the sparse matrix input, is equivalent to a
4804: row-major ordering.. i.e for the following matrix, the input data expected is
4805: as shown:
4807: 1 0 0
4808: 2 0 3
4809: 4 5 6
4811: i = {0,1,1,2,2,2}
4812: j = {0,0,2,0,1,2}
4813: v = {1,2,3,4,5,6}
4816: .seealso: MatCreate(), MatCreateAIJ(), MatCreateSeqAIJ(), MatCreateSeqAIJWithArrays(), MatMPIAIJSetPreallocationCSR()
4818: @*/
4819: PetscErrorCode MatCreateSeqAIJFromTriple(MPI_Comm comm,PetscInt m,PetscInt n,PetscInt i[],PetscInt j[],PetscScalar a[],Mat *mat,PetscInt nz,PetscBool idx)
4820: {
4822: PetscInt ii, *nnz, one = 1,row,col;
4826: PetscCalloc1(m,&nnz);
4827: for (ii = 0; ii < nz; ii++) {
4828: nnz[i[ii] - !!idx] += 1;
4829: }
4830: MatCreate(comm,mat);
4831: MatSetSizes(*mat,m,n,m,n);
4832: MatSetType(*mat,MATSEQAIJ);
4833: MatSeqAIJSetPreallocation_SeqAIJ(*mat,0,nnz);
4834: for (ii = 0; ii < nz; ii++) {
4835: if (idx) {
4836: row = i[ii] - 1;
4837: col = j[ii] - 1;
4838: } else {
4839: row = i[ii];
4840: col = j[ii];
4841: }
4842: MatSetValues(*mat,one,&row,one,&col,&a[ii],ADD_VALUES);
4843: }
4844: MatAssemblyBegin(*mat,MAT_FINAL_ASSEMBLY);
4845: MatAssemblyEnd(*mat,MAT_FINAL_ASSEMBLY);
4846: PetscFree(nnz);
4847: return(0);
4848: }
4850: PetscErrorCode MatSeqAIJInvalidateDiagonal(Mat A)
4851: {
4852: Mat_SeqAIJ *a=(Mat_SeqAIJ*)A->data;
4856: a->idiagvalid = PETSC_FALSE;
4857: a->ibdiagvalid = PETSC_FALSE;
4859: MatSeqAIJInvalidateDiagonal_Inode(A);
4860: return(0);
4861: }
4863: PetscErrorCode MatCreateMPIMatConcatenateSeqMat_SeqAIJ(MPI_Comm comm,Mat inmat,PetscInt n,MatReuse scall,Mat *outmat)
4864: {
4866: PetscMPIInt size;
4869: MPI_Comm_size(comm,&size);
4870: if (size == 1) {
4871: if (scall == MAT_INITIAL_MATRIX) {
4872: MatDuplicate(inmat,MAT_COPY_VALUES,outmat);
4873: } else {
4874: MatCopy(inmat,*outmat,SAME_NONZERO_PATTERN);
4875: }
4876: } else {
4877: MatCreateMPIMatConcatenateSeqMat_MPIAIJ(comm,inmat,n,scall,outmat);
4878: }
4879: return(0);
4880: }
4882: /*
4883: Permute A into C's *local* index space using rowemb,colemb.
4884: The embedding are supposed to be injections and the above implies that the range of rowemb is a subset
4885: of [0,m), colemb is in [0,n).
4886: If pattern == DIFFERENT_NONZERO_PATTERN, C is preallocated according to A.
4887: */
4888: PetscErrorCode MatSetSeqMat_SeqAIJ(Mat C,IS rowemb,IS colemb,MatStructure pattern,Mat B)
4889: {
4890: /* If making this function public, change the error returned in this function away from _PLIB. */
4892: Mat_SeqAIJ *Baij;
4893: PetscBool seqaij;
4894: PetscInt m,n,*nz,i,j,count;
4895: PetscScalar v;
4896: const PetscInt *rowindices,*colindices;
4899: if (!B) return(0);
4900: /* Check to make sure the target matrix (and embeddings) are compatible with C and each other. */
4901: PetscObjectBaseTypeCompare((PetscObject)B,MATSEQAIJ,&seqaij);
4902: if (!seqaij) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_PLIB,"Input matrix is of wrong type");
4903: if (rowemb) {
4904: ISGetLocalSize(rowemb,&m);
4905: 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);
4906: } else {
4907: if (C->rmap->n != B->rmap->n) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_PLIB,"Input matrix is row-incompatible with the target matrix");
4908: }
4909: if (colemb) {
4910: ISGetLocalSize(colemb,&n);
4911: 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);
4912: } else {
4913: if (C->cmap->n != B->cmap->n) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_PLIB,"Input matrix is col-incompatible with the target matrix");
4914: }
4916: Baij = (Mat_SeqAIJ*)(B->data);
4917: if (pattern == DIFFERENT_NONZERO_PATTERN) {
4918: PetscMalloc1(B->rmap->n,&nz);
4919: for (i=0; i<B->rmap->n; i++) {
4920: nz[i] = Baij->i[i+1] - Baij->i[i];
4921: }
4922: MatSeqAIJSetPreallocation(C,0,nz);
4923: PetscFree(nz);
4924: }
4925: if (pattern == SUBSET_NONZERO_PATTERN) {
4926: MatZeroEntries(C);
4927: }
4928: count = 0;
4929: rowindices = NULL;
4930: colindices = NULL;
4931: if (rowemb) {
4932: ISGetIndices(rowemb,&rowindices);
4933: }
4934: if (colemb) {
4935: ISGetIndices(colemb,&colindices);
4936: }
4937: for (i=0; i<B->rmap->n; i++) {
4938: PetscInt row;
4939: row = i;
4940: if (rowindices) row = rowindices[i];
4941: for (j=Baij->i[i]; j<Baij->i[i+1]; j++) {
4942: PetscInt col;
4943: col = Baij->j[count];
4944: if (colindices) col = colindices[col];
4945: v = Baij->a[count];
4946: MatSetValues(C,1,&row,1,&col,&v,INSERT_VALUES);
4947: ++count;
4948: }
4949: }
4950: /* FIXME: set C's nonzerostate correctly. */
4951: /* Assembly for C is necessary. */
4952: C->preallocated = PETSC_TRUE;
4953: C->assembled = PETSC_TRUE;
4954: C->was_assembled = PETSC_FALSE;
4955: return(0);
4956: }
4958: PetscFunctionList MatSeqAIJList = NULL;
4960: /*@C
4961: MatSeqAIJSetType - Converts a MATSEQAIJ matrix to a subtype
4963: Collective on Mat
4965: Input Parameters:
4966: + mat - the matrix object
4967: - matype - matrix type
4969: Options Database Key:
4970: . -mat_seqai_type <method> - for example seqaijcrl
4973: Level: intermediate
4975: .seealso: PCSetType(), VecSetType(), MatCreate(), MatType, Mat
4976: @*/
4977: PetscErrorCode MatSeqAIJSetType(Mat mat, MatType matype)
4978: {
4979: PetscErrorCode ierr,(*r)(Mat,MatType,MatReuse,Mat*);
4980: PetscBool sametype;
4984: PetscObjectTypeCompare((PetscObject)mat,matype,&sametype);
4985: if (sametype) return(0);
4987: PetscFunctionListFind(MatSeqAIJList,matype,&r);
4988: if (!r) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_UNKNOWN_TYPE,"Unknown Mat type given: %s",matype);
4989: (*r)(mat,matype,MAT_INPLACE_MATRIX,&mat);
4990: return(0);
4991: }
4994: /*@C
4995: MatSeqAIJRegister - - Adds a new sub-matrix type for sequential AIJ matrices
4997: Not Collective
4999: Input Parameters:
5000: + name - name of a new user-defined matrix type, for example MATSEQAIJCRL
5001: - function - routine to convert to subtype
5003: Notes:
5004: MatSeqAIJRegister() may be called multiple times to add several user-defined solvers.
5007: Then, your matrix can be chosen with the procedural interface at runtime via the option
5008: $ -mat_seqaij_type my_mat
5010: Level: advanced
5012: .seealso: MatSeqAIJRegisterAll()
5015: Level: advanced
5016: @*/
5017: PetscErrorCode MatSeqAIJRegister(const char sname[],PetscErrorCode (*function)(Mat,MatType,MatReuse,Mat *))
5018: {
5022: MatInitializePackage();
5023: PetscFunctionListAdd(&MatSeqAIJList,sname,function);
5024: return(0);
5025: }
5027: PetscBool MatSeqAIJRegisterAllCalled = PETSC_FALSE;
5029: /*@C
5030: MatSeqAIJRegisterAll - Registers all of the matrix subtypes of SeqAIJ
5032: Not Collective
5034: Level: advanced
5036: Developers Note: CUSP and CUSPARSE do not yet support the MatConvert_SeqAIJ..() paradigm and thus cannot be registered here
5038: .seealso: MatRegisterAll(), MatSeqAIJRegister()
5039: @*/
5040: PetscErrorCode MatSeqAIJRegisterAll(void)
5041: {
5045: if (MatSeqAIJRegisterAllCalled) return(0);
5046: MatSeqAIJRegisterAllCalled = PETSC_TRUE;
5048: MatSeqAIJRegister(MATSEQAIJCRL, MatConvert_SeqAIJ_SeqAIJCRL);
5049: MatSeqAIJRegister(MATSEQAIJPERM, MatConvert_SeqAIJ_SeqAIJPERM);
5050: MatSeqAIJRegister(MATSEQAIJSELL, MatConvert_SeqAIJ_SeqAIJSELL);
5051: #if defined(PETSC_HAVE_MKL_SPARSE)
5052: MatSeqAIJRegister(MATSEQAIJMKL, MatConvert_SeqAIJ_SeqAIJMKL);
5053: #endif
5054: #if defined(PETSC_HAVE_VIENNACL) && defined(PETSC_HAVE_VIENNACL_NO_CUDA)
5055: MatSeqAIJRegister(MATMPIAIJVIENNACL, MatConvert_SeqAIJ_SeqAIJViennaCL);
5056: #endif
5057: return(0);
5058: }
5060: /*
5061: Special version for direct calls from Fortran
5062: */
5063: #include <petsc/private/fortranimpl.h>
5064: #if defined(PETSC_HAVE_FORTRAN_CAPS)
5065: #define matsetvaluesseqaij_ MATSETVALUESSEQAIJ
5066: #elif !defined(PETSC_HAVE_FORTRAN_UNDERSCORE)
5067: #define matsetvaluesseqaij_ matsetvaluesseqaij
5068: #endif
5070: /* Change these macros so can be used in void function */
5071: #undef CHKERRQ
5072: #define CHKERRQ(ierr) CHKERRABORT(PetscObjectComm((PetscObject)A),ierr)
5073: #undef SETERRQ2
5074: #define SETERRQ2(comm,ierr,b,c,d) CHKERRABORT(comm,ierr)
5075: #undef SETERRQ3
5076: #define SETERRQ3(comm,ierr,b,c,d,e) CHKERRABORT(comm,ierr)
5078: PETSC_EXTERN void matsetvaluesseqaij_(Mat *AA,PetscInt *mm,const PetscInt im[],PetscInt *nn,const PetscInt in[],const PetscScalar v[],InsertMode *isis, PetscErrorCode *_ierr)
5079: {
5080: Mat A = *AA;
5081: PetscInt m = *mm, n = *nn;
5082: InsertMode is = *isis;
5083: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
5084: PetscInt *rp,k,low,high,t,ii,row,nrow,i,col,l,rmax,N;
5085: PetscInt *imax,*ai,*ailen;
5087: PetscInt *aj,nonew = a->nonew,lastcol = -1;
5088: MatScalar *ap,value,*aa;
5089: PetscBool ignorezeroentries = a->ignorezeroentries;
5090: PetscBool roworiented = a->roworiented;
5093: MatCheckPreallocated(A,1);
5094: imax = a->imax;
5095: ai = a->i;
5096: ailen = a->ilen;
5097: aj = a->j;
5098: aa = a->a;
5100: for (k=0; k<m; k++) { /* loop over added rows */
5101: row = im[k];
5102: if (row < 0) continue;
5103: if (PetscUnlikelyDebug(row >= A->rmap->n)) SETERRABORT(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_OUTOFRANGE,"Row too large");
5104: rp = aj + ai[row]; ap = aa + ai[row];
5105: rmax = imax[row]; nrow = ailen[row];
5106: low = 0;
5107: high = nrow;
5108: for (l=0; l<n; l++) { /* loop over added columns */
5109: if (in[l] < 0) continue;
5110: if (PetscUnlikelyDebug(in[l] >= A->cmap->n)) SETERRABORT(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_OUTOFRANGE,"Column too large");
5111: col = in[l];
5112: if (roworiented) value = v[l + k*n];
5113: else value = v[k + l*m];
5115: if (value == 0.0 && ignorezeroentries && (is == ADD_VALUES)) continue;
5117: if (col <= lastcol) low = 0;
5118: else high = nrow;
5119: lastcol = col;
5120: while (high-low > 5) {
5121: t = (low+high)/2;
5122: if (rp[t] > col) high = t;
5123: else low = t;
5124: }
5125: for (i=low; i<high; i++) {
5126: if (rp[i] > col) break;
5127: if (rp[i] == col) {
5128: if (is == ADD_VALUES) ap[i] += value;
5129: else ap[i] = value;
5130: goto noinsert;
5131: }
5132: }
5133: if (value == 0.0 && ignorezeroentries) goto noinsert;
5134: if (nonew == 1) goto noinsert;
5135: if (nonew == -1) SETERRABORT(PetscObjectComm((PetscObject)A),PETSC_ERR_ARG_OUTOFRANGE,"Inserting a new nonzero in the matrix");
5136: MatSeqXAIJReallocateAIJ(A,A->rmap->n,1,nrow,row,col,rmax,aa,ai,aj,rp,ap,imax,nonew,MatScalar);
5137: N = nrow++ - 1; a->nz++; high++;
5138: /* shift up all the later entries in this row */
5139: for (ii=N; ii>=i; ii--) {
5140: rp[ii+1] = rp[ii];
5141: ap[ii+1] = ap[ii];
5142: }
5143: rp[i] = col;
5144: ap[i] = value;
5145: A->nonzerostate++;
5146: noinsert:;
5147: low = i + 1;
5148: }
5149: ailen[row] = nrow;
5150: }
5151: PetscFunctionReturnVoid();
5152: }