Actual source code: aij.c
1: #define PETSCMAT_DLL
3: /*
4: Defines the basic matrix operations for the AIJ (compressed row)
5: matrix storage format.
6: */
9: #include ../src/mat/impls/aij/seq/aij.h
10: #include ../src/inline/spops.h
11: #include ../src/inline/dot.h
12: #include petscbt.h
16: PetscErrorCode MatDiagonalSet_SeqAIJ(Mat Y,Vec D,InsertMode is)
17: {
19: Mat_SeqAIJ *aij = (Mat_SeqAIJ*) Y->data;
20: PetscInt i,*diag, m = Y->rmap->n;
21: MatScalar *aa = aij->a;
22: PetscScalar *v;
23: PetscTruth missing;
26: if (Y->assembled) {
27: MatMissingDiagonal_SeqAIJ(Y,&missing,PETSC_NULL);
28: if (!missing) {
29: diag = aij->diag;
30: VecGetArray(D,&v);
31: if (is == INSERT_VALUES) {
32: for (i=0; i<m; i++) {
33: aa[diag[i]] = v[i];
34: }
35: } else {
36: for (i=0; i<m; i++) {
37: aa[diag[i]] += v[i];
38: }
39: }
40: VecRestoreArray(D,&v);
41: return(0);
42: }
43: }
44: MatDiagonalSet_Default(Y,D,is);
45: return(0);
46: }
50: PetscErrorCode MatGetRowIJ_SeqAIJ(Mat A,PetscInt oshift,PetscTruth symmetric,PetscTruth blockcompressed,PetscInt *m,PetscInt *ia[],PetscInt *ja[],PetscTruth *done)
51: {
52: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
54: PetscInt i,ishift;
55:
57: *m = A->rmap->n;
58: if (!ia) return(0);
59: ishift = 0;
60: if (symmetric && !A->structurally_symmetric) {
61: MatToSymmetricIJ_SeqAIJ(A->rmap->n,a->i,a->j,ishift,oshift,ia,ja);
62: } else if (oshift == 1) {
63: PetscInt nz = a->i[A->rmap->n];
64: /* malloc space and add 1 to i and j indices */
65: PetscMalloc((A->rmap->n+1)*sizeof(PetscInt),ia);
66: for (i=0; i<A->rmap->n+1; i++) (*ia)[i] = a->i[i] + 1;
67: if (ja) {
68: PetscMalloc((nz+1)*sizeof(PetscInt),ja);
69: for (i=0; i<nz; i++) (*ja)[i] = a->j[i] + 1;
70: }
71: } else {
72: *ia = a->i;
73: if (ja) *ja = a->j;
74: }
75: return(0);
76: }
80: PetscErrorCode MatRestoreRowIJ_SeqAIJ(Mat A,PetscInt oshift,PetscTruth symmetric,PetscTruth blockcompressed,PetscInt *n,PetscInt *ia[],PetscInt *ja[],PetscTruth *done)
81: {
83:
85: if (!ia) return(0);
86: if ((symmetric && !A->structurally_symmetric) || oshift == 1) {
87: PetscFree(*ia);
88: if (ja) {PetscFree(*ja);}
89: }
90: return(0);
91: }
95: PetscErrorCode MatGetColumnIJ_SeqAIJ(Mat A,PetscInt oshift,PetscTruth symmetric,PetscTruth blockcompressed,PetscInt *nn,PetscInt *ia[],PetscInt *ja[],PetscTruth *done)
96: {
97: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
99: PetscInt i,*collengths,*cia,*cja,n = A->cmap->n,m = A->rmap->n;
100: PetscInt nz = a->i[m],row,*jj,mr,col;
101:
103: *nn = n;
104: if (!ia) return(0);
105: if (symmetric) {
106: MatToSymmetricIJ_SeqAIJ(A->rmap->n,a->i,a->j,0,oshift,ia,ja);
107: } else {
108: PetscMalloc((n+1)*sizeof(PetscInt),&collengths);
109: PetscMemzero(collengths,n*sizeof(PetscInt));
110: PetscMalloc((n+1)*sizeof(PetscInt),&cia);
111: PetscMalloc((nz+1)*sizeof(PetscInt),&cja);
112: jj = a->j;
113: for (i=0; i<nz; i++) {
114: collengths[jj[i]]++;
115: }
116: cia[0] = oshift;
117: for (i=0; i<n; i++) {
118: cia[i+1] = cia[i] + collengths[i];
119: }
120: PetscMemzero(collengths,n*sizeof(PetscInt));
121: jj = a->j;
122: for (row=0; row<m; row++) {
123: mr = a->i[row+1] - a->i[row];
124: for (i=0; i<mr; i++) {
125: col = *jj++;
126: cja[cia[col] + collengths[col]++ - oshift] = row + oshift;
127: }
128: }
129: PetscFree(collengths);
130: *ia = cia; *ja = cja;
131: }
132: return(0);
133: }
137: PetscErrorCode MatRestoreColumnIJ_SeqAIJ(Mat A,PetscInt oshift,PetscTruth symmetric,PetscTruth blockcompressed,PetscInt *n,PetscInt *ia[],PetscInt *ja[],PetscTruth *done)
138: {
142: if (!ia) return(0);
144: PetscFree(*ia);
145: PetscFree(*ja);
146:
147: return(0);
148: }
152: PetscErrorCode MatSetValuesRow_SeqAIJ(Mat A,PetscInt row,const PetscScalar v[])
153: {
154: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
155: PetscInt *ai = a->i;
159: PetscMemcpy(a->a+ai[row],v,(ai[row+1]-ai[row])*sizeof(PetscScalar));
160: return(0);
161: }
163: #define CHUNKSIZE 15
167: PetscErrorCode MatSetValues_SeqAIJ(Mat A,PetscInt m,const PetscInt im[],PetscInt n,const PetscInt in[],const PetscScalar v[],InsertMode is)
168: {
169: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
170: PetscInt *rp,k,low,high,t,ii,row,nrow,i,col,l,rmax,N;
171: PetscInt *imax = a->imax,*ai = a->i,*ailen = a->ilen;
173: PetscInt *aj = a->j,nonew = a->nonew,lastcol = -1;
174: MatScalar *ap,value,*aa = a->a;
175: PetscTruth ignorezeroentries = a->ignorezeroentries;
176: PetscTruth roworiented = a->roworiented;
179: for (k=0; k<m; k++) { /* loop over added rows */
180: row = im[k];
181: if (row < 0) continue;
182: #if defined(PETSC_USE_DEBUG)
183: if (row >= A->rmap->n) SETERRQ2(PETSC_ERR_ARG_OUTOFRANGE,"Row too large: row %D max %D",row,A->rmap->n-1);
184: #endif
185: rp = aj + ai[row]; ap = aa + ai[row];
186: rmax = imax[row]; nrow = ailen[row];
187: low = 0;
188: high = nrow;
189: for (l=0; l<n; l++) { /* loop over added columns */
190: if (in[l] < 0) continue;
191: #if defined(PETSC_USE_DEBUG)
192: if (in[l] >= A->cmap->n) SETERRQ2(PETSC_ERR_ARG_OUTOFRANGE,"Column too large: col %D max %D",in[l],A->cmap->n-1);
193: #endif
194: col = in[l];
195: if (v) {
196: if (roworiented) {
197: value = v[l + k*n];
198: } else {
199: value = v[k + l*m];
200: }
201: } else {
202: value = 0;
203: }
204: if (value == 0.0 && ignorezeroentries && (is == ADD_VALUES)) continue;
206: if (col <= lastcol) low = 0; else high = nrow;
207: lastcol = col;
208: while (high-low > 5) {
209: t = (low+high)/2;
210: if (rp[t] > col) high = t;
211: else low = t;
212: }
213: for (i=low; i<high; i++) {
214: if (rp[i] > col) break;
215: if (rp[i] == col) {
216: if (is == ADD_VALUES) ap[i] += value;
217: else ap[i] = value;
218: low = i + 1;
219: goto noinsert;
220: }
221: }
222: if (value == 0.0 && ignorezeroentries) goto noinsert;
223: if (nonew == 1) goto noinsert;
224: if (nonew == -1) SETERRQ2(PETSC_ERR_ARG_OUTOFRANGE,"Inserting a new nonzero at (%D,%D) in the matrix",row,col);
225: MatSeqXAIJReallocateAIJ(A,A->rmap->n,1,nrow,row,col,rmax,aa,ai,aj,rp,ap,imax,nonew,MatScalar);
226: N = nrow++ - 1; a->nz++; high++;
227: /* shift up all the later entries in this row */
228: for (ii=N; ii>=i; ii--) {
229: rp[ii+1] = rp[ii];
230: ap[ii+1] = ap[ii];
231: }
232: rp[i] = col;
233: ap[i] = value;
234: low = i + 1;
235: noinsert:;
236: }
237: ailen[row] = nrow;
238: }
239: A->same_nonzero = PETSC_FALSE;
240: return(0);
241: }
246: PetscErrorCode MatGetValues_SeqAIJ(Mat A,PetscInt m,const PetscInt im[],PetscInt n,const PetscInt in[],PetscScalar v[])
247: {
248: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
249: PetscInt *rp,k,low,high,t,row,nrow,i,col,l,*aj = a->j;
250: PetscInt *ai = a->i,*ailen = a->ilen;
251: MatScalar *ap,*aa = a->a;
254: for (k=0; k<m; k++) { /* loop over rows */
255: row = im[k];
256: if (row < 0) {v += n; continue;} /* SETERRQ1(PETSC_ERR_ARG_OUTOFRANGE,"Negative row: %D",row); */
257: if (row >= A->rmap->n) SETERRQ2(PETSC_ERR_ARG_OUTOFRANGE,"Row too large: row %D max %D",row,A->rmap->n-1);
258: rp = aj + ai[row]; ap = aa + ai[row];
259: nrow = ailen[row];
260: for (l=0; l<n; l++) { /* loop over columns */
261: if (in[l] < 0) {v++; continue;} /* SETERRQ1(PETSC_ERR_ARG_OUTOFRANGE,"Negative column: %D",in[l]); */
262: if (in[l] >= A->cmap->n) SETERRQ2(PETSC_ERR_ARG_OUTOFRANGE,"Column too large: col %D max %D",in[l],A->cmap->n-1);
263: col = in[l] ;
264: high = nrow; low = 0; /* assume unsorted */
265: while (high-low > 5) {
266: t = (low+high)/2;
267: if (rp[t] > col) high = t;
268: else low = t;
269: }
270: for (i=low; i<high; i++) {
271: if (rp[i] > col) break;
272: if (rp[i] == col) {
273: *v++ = ap[i];
274: goto finished;
275: }
276: }
277: *v++ = 0.0;
278: finished:;
279: }
280: }
281: return(0);
282: }
287: PetscErrorCode MatView_SeqAIJ_Binary(Mat A,PetscViewer viewer)
288: {
289: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
291: PetscInt i,*col_lens;
292: int fd;
295: PetscViewerBinaryGetDescriptor(viewer,&fd);
296: PetscMalloc((4+A->rmap->n)*sizeof(PetscInt),&col_lens);
297: col_lens[0] = MAT_FILE_COOKIE;
298: col_lens[1] = A->rmap->n;
299: col_lens[2] = A->cmap->n;
300: col_lens[3] = a->nz;
302: /* store lengths of each row and write (including header) to file */
303: for (i=0; i<A->rmap->n; i++) {
304: col_lens[4+i] = a->i[i+1] - a->i[i];
305: }
306: PetscBinaryWrite(fd,col_lens,4+A->rmap->n,PETSC_INT,PETSC_TRUE);
307: PetscFree(col_lens);
309: /* store column indices (zero start index) */
310: PetscBinaryWrite(fd,a->j,a->nz,PETSC_INT,PETSC_FALSE);
312: /* store nonzero values */
313: PetscBinaryWrite(fd,a->a,a->nz,PETSC_SCALAR,PETSC_FALSE);
314: return(0);
315: }
317: EXTERN PetscErrorCode MatSeqAIJFactorInfo_Matlab(Mat,PetscViewer);
321: PetscErrorCode MatView_SeqAIJ_ASCII(Mat A,PetscViewer viewer)
322: {
323: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
324: PetscErrorCode ierr;
325: PetscInt i,j,m = A->rmap->n,shift=0;
326: const char *name;
327: PetscViewerFormat format;
330: PetscObjectGetName((PetscObject)A,&name);
331: PetscViewerGetFormat(viewer,&format);
332: if (format == PETSC_VIEWER_ASCII_MATLAB) {
333: PetscInt nofinalvalue = 0;
334: if ((a->i[m] == a->i[m-1]) || (a->j[a->nz-1] != A->cmap->n-!shift)) {
335: nofinalvalue = 1;
336: }
337: PetscViewerASCIIUseTabs(viewer,PETSC_NO);
338: PetscViewerASCIIPrintf(viewer,"%% Size = %D %D \n",m,A->cmap->n);
339: PetscViewerASCIIPrintf(viewer,"%% Nonzeros = %D \n",a->nz);
340: PetscViewerASCIIPrintf(viewer,"zzz = zeros(%D,3);\n",a->nz+nofinalvalue);
341: PetscViewerASCIIPrintf(viewer,"zzz = [\n");
343: for (i=0; i<m; i++) {
344: for (j=a->i[i]+shift; j<a->i[i+1]+shift; j++) {
345: #if defined(PETSC_USE_COMPLEX)
346: PetscViewerASCIIPrintf(viewer,"%D %D %18.16e + %18.16ei \n",i+1,a->j[j]+!shift,PetscRealPart(a->a[j]),PetscImaginaryPart(a->a[j]));
347: #else
348: PetscViewerASCIIPrintf(viewer,"%D %D %18.16e\n",i+1,a->j[j]+!shift,a->a[j]);
349: #endif
350: }
351: }
352: if (nofinalvalue) {
353: PetscViewerASCIIPrintf(viewer,"%D %D %18.16e\n",m,A->cmap->n,0.0);
354: }
355: PetscViewerASCIIPrintf(viewer,"];\n %s = spconvert(zzz);\n",name);
356: PetscViewerASCIIUseTabs(viewer,PETSC_YES);
357: } else if (format == PETSC_VIEWER_ASCII_FACTOR_INFO || format == PETSC_VIEWER_ASCII_INFO) {
358: return(0);
359: } else if (format == PETSC_VIEWER_ASCII_COMMON) {
360: PetscViewerASCIIUseTabs(viewer,PETSC_NO);
361: for (i=0; i<m; i++) {
362: PetscViewerASCIIPrintf(viewer,"row %D:",i);
363: for (j=a->i[i]+shift; j<a->i[i+1]+shift; j++) {
364: #if defined(PETSC_USE_COMPLEX)
365: if (PetscImaginaryPart(a->a[j]) > 0.0 && PetscRealPart(a->a[j]) != 0.0) {
366: PetscViewerASCIIPrintf(viewer," (%D, %G + %G i)",a->j[j]+shift,PetscRealPart(a->a[j]),PetscImaginaryPart(a->a[j]));
367: } else if (PetscImaginaryPart(a->a[j]) < 0.0 && PetscRealPart(a->a[j]) != 0.0) {
368: PetscViewerASCIIPrintf(viewer," (%D, %G - %G i)",a->j[j]+shift,PetscRealPart(a->a[j]),-PetscImaginaryPart(a->a[j]));
369: } else if (PetscRealPart(a->a[j]) != 0.0) {
370: PetscViewerASCIIPrintf(viewer," (%D, %G) ",a->j[j]+shift,PetscRealPart(a->a[j]));
371: }
372: #else
373: if (a->a[j] != 0.0) {PetscViewerASCIIPrintf(viewer," (%D, %G) ",a->j[j]+shift,a->a[j]);}
374: #endif
375: }
376: PetscViewerASCIIPrintf(viewer,"\n");
377: }
378: PetscViewerASCIIUseTabs(viewer,PETSC_YES);
379: } else if (format == PETSC_VIEWER_ASCII_SYMMODU) {
380: PetscInt nzd=0,fshift=1,*sptr;
381: PetscViewerASCIIUseTabs(viewer,PETSC_NO);
382: PetscMalloc((m+1)*sizeof(PetscInt),&sptr);
383: for (i=0; i<m; i++) {
384: sptr[i] = nzd+1;
385: for (j=a->i[i]+shift; j<a->i[i+1]+shift; j++) {
386: if (a->j[j] >= i) {
387: #if defined(PETSC_USE_COMPLEX)
388: if (PetscImaginaryPart(a->a[j]) != 0.0 || PetscRealPart(a->a[j]) != 0.0) nzd++;
389: #else
390: if (a->a[j] != 0.0) nzd++;
391: #endif
392: }
393: }
394: }
395: sptr[m] = nzd+1;
396: PetscViewerASCIIPrintf(viewer," %D %D\n\n",m,nzd);
397: for (i=0; i<m+1; i+=6) {
398: if (i+4<m) {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]);}
399: else if (i+3<m) {PetscViewerASCIIPrintf(viewer," %D %D %D %D %D\n",sptr[i],sptr[i+1],sptr[i+2],sptr[i+3],sptr[i+4]);}
400: else if (i+2<m) {PetscViewerASCIIPrintf(viewer," %D %D %D %D\n",sptr[i],sptr[i+1],sptr[i+2],sptr[i+3]);}
401: else if (i+1<m) {PetscViewerASCIIPrintf(viewer," %D %D %D\n",sptr[i],sptr[i+1],sptr[i+2]);}
402: else if (i<m) {PetscViewerASCIIPrintf(viewer," %D %D\n",sptr[i],sptr[i+1]);}
403: else {PetscViewerASCIIPrintf(viewer," %D\n",sptr[i]);}
404: }
405: PetscViewerASCIIPrintf(viewer,"\n");
406: PetscFree(sptr);
407: for (i=0; i<m; i++) {
408: for (j=a->i[i]+shift; j<a->i[i+1]+shift; j++) {
409: if (a->j[j] >= i) {PetscViewerASCIIPrintf(viewer," %D ",a->j[j]+fshift);}
410: }
411: PetscViewerASCIIPrintf(viewer,"\n");
412: }
413: PetscViewerASCIIPrintf(viewer,"\n");
414: for (i=0; i<m; i++) {
415: for (j=a->i[i]+shift; j<a->i[i+1]+shift; j++) {
416: if (a->j[j] >= i) {
417: #if defined(PETSC_USE_COMPLEX)
418: if (PetscImaginaryPart(a->a[j]) != 0.0 || PetscRealPart(a->a[j]) != 0.0) {
419: PetscViewerASCIIPrintf(viewer," %18.16e %18.16e ",PetscRealPart(a->a[j]),PetscImaginaryPart(a->a[j]));
420: }
421: #else
422: if (a->a[j] != 0.0) {PetscViewerASCIIPrintf(viewer," %18.16e ",a->a[j]);}
423: #endif
424: }
425: }
426: PetscViewerASCIIPrintf(viewer,"\n");
427: }
428: PetscViewerASCIIUseTabs(viewer,PETSC_YES);
429: } else if (format == PETSC_VIEWER_ASCII_DENSE) {
430: PetscInt cnt = 0,jcnt;
431: PetscScalar value;
433: PetscViewerASCIIUseTabs(viewer,PETSC_NO);
434: for (i=0; i<m; i++) {
435: jcnt = 0;
436: for (j=0; j<A->cmap->n; j++) {
437: if (jcnt < a->i[i+1]-a->i[i] && j == a->j[cnt]) {
438: value = a->a[cnt++];
439: jcnt++;
440: } else {
441: value = 0.0;
442: }
443: #if defined(PETSC_USE_COMPLEX)
444: PetscViewerASCIIPrintf(viewer," %7.5e+%7.5e i ",PetscRealPart(value),PetscImaginaryPart(value));
445: #else
446: PetscViewerASCIIPrintf(viewer," %7.5e ",value);
447: #endif
448: }
449: PetscViewerASCIIPrintf(viewer,"\n");
450: }
451: PetscViewerASCIIUseTabs(viewer,PETSC_YES);
452: } else if (format == PETSC_VIEWER_ASCII_MATRIXMARKET) {
453: PetscViewerASCIIUseTabs(viewer,PETSC_NO);
454: #if defined(PETSC_USE_COMPLEX)
455: PetscViewerASCIIPrintf(viewer,"%%matrix complex general\n");
456: #else
457: PetscViewerASCIIPrintf(viewer,"%%matrix real general\n");
458: #endif
459: PetscViewerASCIIPrintf(viewer,"%D %D %D\n", m, A->cmap->n, a->nz);
460: for (i=0; i<m; i++) {
461: for (j=a->i[i]+shift; j<a->i[i+1]+shift; j++) {
462: #if defined(PETSC_USE_COMPLEX)
463: if (PetscImaginaryPart(a->a[j]) > 0.0) {
464: PetscViewerASCIIPrintf(viewer,"%D %D, %G %G\n", i+shift,a->j[j]+shift,PetscRealPart(a->a[j]),PetscImaginaryPart(a->a[j]));
465: } else if (PetscImaginaryPart(a->a[j]) < 0.0) {
466: PetscViewerASCIIPrintf(viewer,"%D %D, %G -%G\n", i+shift,a->j[j]+shift,PetscRealPart(a->a[j]),-PetscImaginaryPart(a->a[j]));
467: } else {
468: PetscViewerASCIIPrintf(viewer,"%D %D, %G\n", i+shift,a->j[j]+shift,PetscRealPart(a->a[j]));
469: }
470: #else
471: PetscViewerASCIIPrintf(viewer,"%D %D %G\n", i+shift, a->j[j]+shift, a->a[j]);
472: #endif
473: }
474: }
475: PetscViewerASCIIUseTabs(viewer,PETSC_YES);
476: } else {
477: PetscViewerASCIIUseTabs(viewer,PETSC_NO);
478: for (i=0; i<m; i++) {
479: PetscViewerASCIIPrintf(viewer,"row %D:",i);
480: for (j=a->i[i]+shift; j<a->i[i+1]+shift; j++) {
481: #if defined(PETSC_USE_COMPLEX)
482: if (PetscImaginaryPart(a->a[j]) > 0.0) {
483: PetscViewerASCIIPrintf(viewer," (%D, %G + %G i)",a->j[j]+shift,PetscRealPart(a->a[j]),PetscImaginaryPart(a->a[j]));
484: } else if (PetscImaginaryPart(a->a[j]) < 0.0) {
485: PetscViewerASCIIPrintf(viewer," (%D, %G - %G i)",a->j[j]+shift,PetscRealPart(a->a[j]),-PetscImaginaryPart(a->a[j]));
486: } else {
487: PetscViewerASCIIPrintf(viewer," (%D, %G) ",a->j[j]+shift,PetscRealPart(a->a[j]));
488: }
489: #else
490: PetscViewerASCIIPrintf(viewer," (%D, %G) ",a->j[j]+shift,a->a[j]);
491: #endif
492: }
493: PetscViewerASCIIPrintf(viewer,"\n");
494: }
495: PetscViewerASCIIUseTabs(viewer,PETSC_YES);
496: }
497: PetscViewerFlush(viewer);
498: return(0);
499: }
503: PetscErrorCode MatView_SeqAIJ_Draw_Zoom(PetscDraw draw,void *Aa)
504: {
505: Mat A = (Mat) Aa;
506: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
507: PetscErrorCode ierr;
508: PetscInt i,j,m = A->rmap->n,color;
509: PetscReal xl,yl,xr,yr,x_l,x_r,y_l,y_r,maxv = 0.0;
510: PetscViewer viewer;
511: PetscViewerFormat format;
514: PetscObjectQuery((PetscObject)A,"Zoomviewer",(PetscObject*)&viewer);
515: PetscViewerGetFormat(viewer,&format);
517: PetscDrawGetCoordinates(draw,&xl,&yl,&xr,&yr);
518: /* loop over matrix elements drawing boxes */
520: if (format != PETSC_VIEWER_DRAW_CONTOUR) {
521: /* Blue for negative, Cyan for zero and Red for positive */
522: color = PETSC_DRAW_BLUE;
523: for (i=0; i<m; i++) {
524: y_l = m - i - 1.0; y_r = y_l + 1.0;
525: for (j=a->i[i]; j<a->i[i+1]; j++) {
526: x_l = a->j[j] ; x_r = x_l + 1.0;
527: #if defined(PETSC_USE_COMPLEX)
528: if (PetscRealPart(a->a[j]) >= 0.) continue;
529: #else
530: if (a->a[j] >= 0.) continue;
531: #endif
532: PetscDrawRectangle(draw,x_l,y_l,x_r,y_r,color,color,color,color);
533: }
534: }
535: color = PETSC_DRAW_CYAN;
536: for (i=0; i<m; i++) {
537: y_l = m - i - 1.0; y_r = y_l + 1.0;
538: for (j=a->i[i]; j<a->i[i+1]; j++) {
539: x_l = a->j[j]; x_r = x_l + 1.0;
540: if (a->a[j] != 0.) continue;
541: PetscDrawRectangle(draw,x_l,y_l,x_r,y_r,color,color,color,color);
542: }
543: }
544: color = PETSC_DRAW_RED;
545: for (i=0; i<m; i++) {
546: y_l = m - i - 1.0; y_r = y_l + 1.0;
547: for (j=a->i[i]; j<a->i[i+1]; j++) {
548: x_l = a->j[j]; x_r = x_l + 1.0;
549: #if defined(PETSC_USE_COMPLEX)
550: if (PetscRealPart(a->a[j]) <= 0.) continue;
551: #else
552: if (a->a[j] <= 0.) continue;
553: #endif
554: PetscDrawRectangle(draw,x_l,y_l,x_r,y_r,color,color,color,color);
555: }
556: }
557: } else {
558: /* use contour shading to indicate magnitude of values */
559: /* first determine max of all nonzero values */
560: PetscInt nz = a->nz,count;
561: PetscDraw popup;
562: PetscReal scale;
564: for (i=0; i<nz; i++) {
565: if (PetscAbsScalar(a->a[i]) > maxv) maxv = PetscAbsScalar(a->a[i]);
566: }
567: scale = (245.0 - PETSC_DRAW_BASIC_COLORS)/maxv;
568: PetscDrawGetPopup(draw,&popup);
569: if (popup) {PetscDrawScalePopup(popup,0.0,maxv);}
570: count = 0;
571: for (i=0; i<m; i++) {
572: y_l = m - i - 1.0; y_r = y_l + 1.0;
573: for (j=a->i[i]; j<a->i[i+1]; j++) {
574: x_l = a->j[j]; x_r = x_l + 1.0;
575: color = PETSC_DRAW_BASIC_COLORS + (PetscInt)(scale*PetscAbsScalar(a->a[count]));
576: PetscDrawRectangle(draw,x_l,y_l,x_r,y_r,color,color,color,color);
577: count++;
578: }
579: }
580: }
581: return(0);
582: }
586: PetscErrorCode MatView_SeqAIJ_Draw(Mat A,PetscViewer viewer)
587: {
589: PetscDraw draw;
590: PetscReal xr,yr,xl,yl,h,w;
591: PetscTruth isnull;
594: PetscViewerDrawGetDraw(viewer,0,&draw);
595: PetscDrawIsNull(draw,&isnull);
596: if (isnull) return(0);
598: PetscObjectCompose((PetscObject)A,"Zoomviewer",(PetscObject)viewer);
599: xr = A->cmap->n; yr = A->rmap->n; h = yr/10.0; w = xr/10.0;
600: xr += w; yr += h; xl = -w; yl = -h;
601: PetscDrawSetCoordinates(draw,xl,yl,xr,yr);
602: PetscDrawZoom(draw,MatView_SeqAIJ_Draw_Zoom,A);
603: PetscObjectCompose((PetscObject)A,"Zoomviewer",PETSC_NULL);
604: return(0);
605: }
609: PetscErrorCode MatView_SeqAIJ(Mat A,PetscViewer viewer)
610: {
612: PetscTruth iascii,isbinary,isdraw;
615: PetscTypeCompare((PetscObject)viewer,PETSC_VIEWER_ASCII,&iascii);
616: PetscTypeCompare((PetscObject)viewer,PETSC_VIEWER_BINARY,&isbinary);
617: PetscTypeCompare((PetscObject)viewer,PETSC_VIEWER_DRAW,&isdraw);
618: if (iascii) {
619: MatView_SeqAIJ_ASCII(A,viewer);
620: } else if (isbinary) {
621: MatView_SeqAIJ_Binary(A,viewer);
622: } else if (isdraw) {
623: MatView_SeqAIJ_Draw(A,viewer);
624: } else {
625: SETERRQ1(PETSC_ERR_SUP,"Viewer type %s not supported by SeqAIJ matrices",((PetscObject)viewer)->type_name);
626: }
627: MatView_Inode(A,viewer);
628: return(0);
629: }
633: PetscErrorCode MatAssemblyEnd_SeqAIJ(Mat A,MatAssemblyType mode)
634: {
635: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
637: PetscInt fshift = 0,i,j,*ai = a->i,*aj = a->j,*imax = a->imax;
638: PetscInt m = A->rmap->n,*ip,N,*ailen = a->ilen,rmax = 0;
639: MatScalar *aa = a->a,*ap;
640: PetscReal ratio=0.6;
643: if (mode == MAT_FLUSH_ASSEMBLY) return(0);
645: if (m) rmax = ailen[0]; /* determine row with most nonzeros */
646: for (i=1; i<m; i++) {
647: /* move each row back by the amount of empty slots (fshift) before it*/
648: fshift += imax[i-1] - ailen[i-1];
649: rmax = PetscMax(rmax,ailen[i]);
650: if (fshift) {
651: ip = aj + ai[i] ;
652: ap = aa + ai[i] ;
653: N = ailen[i];
654: for (j=0; j<N; j++) {
655: ip[j-fshift] = ip[j];
656: ap[j-fshift] = ap[j];
657: }
658: }
659: ai[i] = ai[i-1] + ailen[i-1];
660: }
661: if (m) {
662: fshift += imax[m-1] - ailen[m-1];
663: ai[m] = ai[m-1] + ailen[m-1];
664: }
665: /* reset ilen and imax for each row */
666: for (i=0; i<m; i++) {
667: ailen[i] = imax[i] = ai[i+1] - ai[i];
668: }
669: a->nz = ai[m];
670: if (fshift && a->nounused == -1) {
671: SETERRQ3(PETSC_ERR_PLIB, "Unused space detected in matrix: %D X %D, %D unneeded", m, A->cmap->n, fshift);
672: }
674: MatMarkDiagonal_SeqAIJ(A);
675: PetscInfo4(A,"Matrix size: %D X %D; storage space: %D unneeded,%D used\n",m,A->cmap->n,fshift,a->nz);
676: PetscInfo1(A,"Number of mallocs during MatSetValues() is %D\n",a->reallocs);
677: PetscInfo1(A,"Maximum nonzeros in any row is %D\n",rmax);
679: a->reallocs = 0;
680: A->info.nz_unneeded = (double)fshift;
681: a->rmax = rmax;
683: /* check for zero rows. If found a large number of zero rows, use CompressedRow functions */
684: Mat_CheckCompressedRow(A,&a->compressedrow,a->i,m,ratio);
685: A->same_nonzero = PETSC_TRUE;
687: MatAssemblyEnd_Inode(A,mode);
689: a->idiagvalid = PETSC_FALSE;
690: return(0);
691: }
695: PetscErrorCode MatRealPart_SeqAIJ(Mat A)
696: {
697: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
698: PetscInt i,nz = a->nz;
699: MatScalar *aa = a->a;
702: for (i=0; i<nz; i++) aa[i] = PetscRealPart(aa[i]);
703: return(0);
704: }
708: PetscErrorCode MatImaginaryPart_SeqAIJ(Mat A)
709: {
710: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
711: PetscInt i,nz = a->nz;
712: MatScalar *aa = a->a;
715: for (i=0; i<nz; i++) aa[i] = PetscImaginaryPart(aa[i]);
716: return(0);
717: }
721: PetscErrorCode MatZeroEntries_SeqAIJ(Mat A)
722: {
723: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
727: PetscMemzero(a->a,(a->i[A->rmap->n])*sizeof(PetscScalar));
728: return(0);
729: }
733: PetscErrorCode MatDestroy_SeqAIJ(Mat A)
734: {
735: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
739: #if defined(PETSC_USE_LOG)
740: PetscLogObjectState((PetscObject)A,"Rows=%D, Cols=%D, NZ=%D",A->rmap->n,A->cmap->n,a->nz);
741: #endif
742: MatSeqXAIJFreeAIJ(A,&a->a,&a->j,&a->i);
743: if (a->row) {
744: ISDestroy(a->row);
745: }
746: if (a->col) {
747: ISDestroy(a->col);
748: }
749: PetscFree(a->diag);
750: PetscFree2(a->imax,a->ilen);
751: PetscFree3(a->idiag,a->mdiag,a->ssor_work);
752: PetscFree(a->solve_work);
753: if (a->icol) {ISDestroy(a->icol);}
754: PetscFree(a->saved_values);
755: if (a->coloring) {ISColoringDestroy(a->coloring);}
756: PetscFree(a->xtoy);
757: if (a->XtoY) {MatDestroy(a->XtoY);}
758: if (a->compressedrow.use){PetscFree(a->compressedrow.i);}
760: MatDestroy_Inode(A);
762: PetscFree(a);
764: PetscObjectChangeTypeName((PetscObject)A,0);
765: PetscObjectComposeFunctionDynamic((PetscObject)A,"MatSeqAIJSetColumnIndices_C","",PETSC_NULL);
766: PetscObjectComposeFunctionDynamic((PetscObject)A,"MatStoreValues_C","",PETSC_NULL);
767: PetscObjectComposeFunctionDynamic((PetscObject)A,"MatRetrieveValues_C","",PETSC_NULL);
768: PetscObjectComposeFunctionDynamic((PetscObject)A,"MatConvert_seqaij_seqsbaij_C","",PETSC_NULL);
769: PetscObjectComposeFunctionDynamic((PetscObject)A,"MatConvert_seqaij_seqbaij_C","",PETSC_NULL);
770: PetscObjectComposeFunctionDynamic((PetscObject)A,"MatConvert_seqaij_seqcsrperm_C","",PETSC_NULL);
771: PetscObjectComposeFunctionDynamic((PetscObject)A,"MatIsTranspose_C","",PETSC_NULL);
772: PetscObjectComposeFunctionDynamic((PetscObject)A,"MatSeqAIJSetPreallocation_C","",PETSC_NULL);
773: PetscObjectComposeFunctionDynamic((PetscObject)A,"MatSeqAIJSetPreallocationCSR_C","",PETSC_NULL);
774: PetscObjectComposeFunctionDynamic((PetscObject)A,"MatReorderForNonzeroDiagonal_C","",PETSC_NULL);
775: return(0);
776: }
780: PetscErrorCode MatCompress_SeqAIJ(Mat A)
781: {
783: return(0);
784: }
788: PetscErrorCode MatSetOption_SeqAIJ(Mat A,MatOption op,PetscTruth flg)
789: {
790: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
794: switch (op) {
795: case MAT_ROW_ORIENTED:
796: a->roworiented = flg;
797: break;
798: case MAT_KEEP_ZEROED_ROWS:
799: a->keepzeroedrows = flg;
800: break;
801: case MAT_NEW_NONZERO_LOCATIONS:
802: a->nonew = (flg ? 0 : 1);
803: break;
804: case MAT_NEW_NONZERO_LOCATION_ERR:
805: a->nonew = (flg ? -1 : 0);
806: break;
807: case MAT_NEW_NONZERO_ALLOCATION_ERR:
808: a->nonew = (flg ? -2 : 0);
809: break;
810: case MAT_UNUSED_NONZERO_LOCATION_ERR:
811: a->nounused = (flg ? -1 : 0);
812: break;
813: case MAT_IGNORE_ZERO_ENTRIES:
814: a->ignorezeroentries = flg;
815: break;
816: case MAT_USE_COMPRESSEDROW:
817: a->compressedrow.use = flg;
818: break;
819: case MAT_NEW_DIAGONALS:
820: case MAT_IGNORE_OFF_PROC_ENTRIES:
821: case MAT_USE_HASH_TABLE:
822: PetscInfo1(A,"Option %s ignored\n",MatOptions[op]);
823: break;
824: default:
825: break;
826: }
827: MatSetOption_Inode(A,op,flg);
828: return(0);
829: }
833: PetscErrorCode MatGetDiagonal_SeqAIJ(Mat A,Vec v)
834: {
835: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
837: PetscInt i,j,n;
838: PetscScalar *x,zero = 0.0;
841: VecSet(v,zero);
842: VecGetArray(v,&x);
843: VecGetLocalSize(v,&n);
844: if (n != A->rmap->n) SETERRQ(PETSC_ERR_ARG_SIZ,"Nonconforming matrix and vector");
845: for (i=0; i<A->rmap->n; i++) {
846: for (j=a->i[i]; j<a->i[i+1]; j++) {
847: if (a->j[j] == i) {
848: x[i] = a->a[j];
849: break;
850: }
851: }
852: }
853: VecRestoreArray(v,&x);
854: return(0);
855: }
859: PetscErrorCode MatMultTransposeAdd_SeqAIJ(Mat A,Vec xx,Vec zz,Vec yy)
860: {
861: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
862: PetscScalar *x,*y;
863: PetscErrorCode ierr;
864: PetscInt m = A->rmap->n;
865: #if !defined(PETSC_USE_FORTRAN_KERNEL_MULTTRANSPOSEAIJ)
866: MatScalar *v;
867: PetscScalar alpha;
868: PetscInt n,i,*idx,*ii,*ridx=PETSC_NULL;
869: Mat_CompressedRow cprow = a->compressedrow;
870: PetscTruth usecprow = cprow.use;
871: #endif
874: if (zz != yy) {VecCopy(zz,yy);}
875: VecGetArray(xx,&x);
876: VecGetArray(yy,&y);
878: #if defined(PETSC_USE_FORTRAN_KERNEL_MULTTRANSPOSEAIJ)
879: fortranmulttransposeaddaij_(&m,x,a->i,a->j,a->a,y);
880: #else
881: if (usecprow){
882: m = cprow.nrows;
883: ii = cprow.i;
884: ridx = cprow.rindex;
885: } else {
886: ii = a->i;
887: }
888: for (i=0; i<m; i++) {
889: idx = a->j + ii[i] ;
890: v = a->a + ii[i] ;
891: n = ii[i+1] - ii[i];
892: if (usecprow){
893: alpha = x[ridx[i]];
894: } else {
895: alpha = x[i];
896: }
897: while (n-->0) {y[*idx++] += alpha * *v++;}
898: }
899: #endif
900: PetscLogFlops(2*a->nz);
901: VecRestoreArray(xx,&x);
902: VecRestoreArray(yy,&y);
903: return(0);
904: }
908: PetscErrorCode MatMultTranspose_SeqAIJ(Mat A,Vec xx,Vec yy)
909: {
913: VecSet(yy,0.0);
914: MatMultTransposeAdd_SeqAIJ(A,xx,yy,yy);
915: return(0);
916: }
921: PetscErrorCode MatMult_SeqAIJ(Mat A,Vec xx,Vec yy)
922: {
923: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
924: PetscScalar *y;
925: const PetscScalar *x;
926: const MatScalar *aa;
927: PetscErrorCode ierr;
928: PetscInt m=A->rmap->n,*aj,*ii;
929: PetscInt n,i,j,nonzerorow=0,*ridx=PETSC_NULL;
930: PetscScalar sum;
931: PetscTruth usecprow=a->compressedrow.use;
932: #if !defined(PETSC_USE_FORTRAN_KERNEL_MULTAIJ)
933: PetscInt jrow;
934: #endif
936: #if defined(PETSC_HAVE_PRAGMA_DISJOINT)
937: #pragma disjoint(*x,*y,*aa)
938: #endif
941: VecGetArray(xx,(PetscScalar**)&x);
942: VecGetArray(yy,&y);
943: aj = a->j;
944: aa = a->a;
945: ii = a->i;
946: if (usecprow){ /* use compressed row format */
947: m = a->compressedrow.nrows;
948: ii = a->compressedrow.i;
949: ridx = a->compressedrow.rindex;
950: for (i=0; i<m; i++){
951: n = ii[i+1] - ii[i];
952: aj = a->j + ii[i];
953: aa = a->a + ii[i];
954: sum = 0.0;
955: nonzerorow += (n>0);
956: for (j=0; j<n; j++) sum += (*aa++)*x[*aj++];
957: y[*ridx++] = sum;
958: }
959: } else { /* do not use compressed row format */
960: #if defined(PETSC_USE_FORTRAN_KERNEL_MULTAIJ)
961: fortranmultaij_(&m,x,ii,aj,aa,y);
962: #else
963: for (i=0; i<m; i++) {
964: jrow = ii[i];
965: n = ii[i+1] - jrow;
966: sum = 0.0;
967: nonzerorow += (n>0);
968: for (j=0; j<n; j++) {
969: sum += aa[jrow]*x[aj[jrow]]; jrow++;
970: }
971: y[i] = sum;
972: }
973: #endif
974: }
975: PetscLogFlops(2*a->nz - nonzerorow);
976: VecRestoreArray(xx,(PetscScalar**)&x);
977: VecRestoreArray(yy,&y);
978: return(0);
979: }
983: PetscErrorCode MatMultAdd_SeqAIJ(Mat A,Vec xx,Vec yy,Vec zz)
984: {
985: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
986: PetscScalar *x,*y,*z;
987: const MatScalar *aa;
988: PetscErrorCode ierr;
989: PetscInt m = A->rmap->n,*aj,*ii;
990: #if !defined(PETSC_USE_FORTRAN_KERNEL_MULTADDAIJ)
991: PetscInt n,i,jrow,j,*ridx=PETSC_NULL;
992: PetscScalar sum;
993: PetscTruth usecprow=a->compressedrow.use;
994: #endif
997: VecGetArray(xx,&x);
998: VecGetArray(yy,&y);
999: if (zz != yy) {
1000: VecGetArray(zz,&z);
1001: } else {
1002: z = y;
1003: }
1005: aj = a->j;
1006: aa = a->a;
1007: ii = a->i;
1008: #if defined(PETSC_USE_FORTRAN_KERNEL_MULTADDAIJ)
1009: fortranmultaddaij_(&m,x,ii,aj,aa,y,z);
1010: #else
1011: if (usecprow){ /* use compressed row format */
1012: if (zz != yy){
1013: PetscMemcpy(z,y,m*sizeof(PetscScalar));
1014: }
1015: m = a->compressedrow.nrows;
1016: ii = a->compressedrow.i;
1017: ridx = a->compressedrow.rindex;
1018: for (i=0; i<m; i++){
1019: n = ii[i+1] - ii[i];
1020: aj = a->j + ii[i];
1021: aa = a->a + ii[i];
1022: sum = y[*ridx];
1023: for (j=0; j<n; j++) sum += (*aa++)*x[*aj++];
1024: z[*ridx++] = sum;
1025: }
1026: } else { /* do not use compressed row format */
1027: for (i=0; i<m; i++) {
1028: jrow = ii[i];
1029: n = ii[i+1] - jrow;
1030: sum = y[i];
1031: for (j=0; j<n; j++) {
1032: sum += aa[jrow]*x[aj[jrow]]; jrow++;
1033: }
1034: z[i] = sum;
1035: }
1036: }
1037: #endif
1038: PetscLogFlops(2*a->nz);
1039: VecRestoreArray(xx,&x);
1040: VecRestoreArray(yy,&y);
1041: if (zz != yy) {
1042: VecRestoreArray(zz,&z);
1043: }
1044: return(0);
1045: }
1047: /*
1048: Adds diagonal pointers to sparse matrix structure.
1049: */
1052: PetscErrorCode MatMarkDiagonal_SeqAIJ(Mat A)
1053: {
1054: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1056: PetscInt i,j,m = A->rmap->n;
1059: if (!a->diag) {
1060: PetscMalloc(m*sizeof(PetscInt),&a->diag);
1061: PetscLogObjectMemory(A, m*sizeof(PetscInt));
1062: }
1063: for (i=0; i<A->rmap->n; i++) {
1064: a->diag[i] = a->i[i+1];
1065: for (j=a->i[i]; j<a->i[i+1]; j++) {
1066: if (a->j[j] == i) {
1067: a->diag[i] = j;
1068: break;
1069: }
1070: }
1071: }
1072: return(0);
1073: }
1075: /*
1076: Checks for missing diagonals
1077: */
1080: PetscErrorCode MatMissingDiagonal_SeqAIJ(Mat A,PetscTruth *missing,PetscInt *d)
1081: {
1082: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1083: PetscInt *diag,*jj = a->j,i;
1086: *missing = PETSC_FALSE;
1087: if (A->rmap->n > 0 && !jj) {
1088: *missing = PETSC_TRUE;
1089: if (d) *d = 0;
1090: PetscInfo(A,"Matrix has no entries therefor is missing diagonal");
1091: } else {
1092: diag = a->diag;
1093: for (i=0; i<A->rmap->n; i++) {
1094: if (jj[diag[i]] != i) {
1095: *missing = PETSC_TRUE;
1096: if (d) *d = i;
1097: PetscInfo1(A,"Matrix is missing diagonal number %D",i);
1098: }
1099: }
1100: }
1101: return(0);
1102: }
1107: PetscErrorCode MatInvertDiagonal_SeqAIJ(Mat A,PetscScalar omega,PetscScalar fshift)
1108: {
1109: Mat_SeqAIJ *a = (Mat_SeqAIJ*) A->data;
1111: PetscInt i,*diag,m = A->rmap->n;
1112: MatScalar *v = a->a;
1113: PetscScalar *idiag,*mdiag;
1116: if (a->idiagvalid) return(0);
1117: MatMarkDiagonal_SeqAIJ(A);
1118: diag = a->diag;
1119: if (!a->idiag) {
1120: PetscMalloc3(m,PetscScalar,&a->idiag,m,PetscScalar,&a->mdiag,m,PetscScalar,&a->ssor_work);
1121: PetscLogObjectMemory(A, 3*m*sizeof(PetscScalar));
1122: v = a->a;
1123: }
1124: mdiag = a->mdiag;
1125: idiag = a->idiag;
1126:
1127: if (omega == 1.0 && !PetscAbsScalar(fshift)) {
1128: for (i=0; i<m; i++) {
1129: mdiag[i] = v[diag[i]];
1130: if (!PetscAbsScalar(mdiag[i])) SETERRQ1(PETSC_ERR_ARG_INCOMP,"Zero diagonal on row %D",i);
1131: idiag[i] = 1.0/v[diag[i]];
1132: }
1133: PetscLogFlops(m);
1134: } else {
1135: for (i=0; i<m; i++) {
1136: mdiag[i] = v[diag[i]];
1137: idiag[i] = omega/(fshift + v[diag[i]]);
1138: }
1139: PetscLogFlops(2*m);
1140: }
1141: a->idiagvalid = PETSC_TRUE;
1142: return(0);
1143: }
1148: PetscErrorCode MatRelax_SeqAIJ(Mat A,Vec bb,PetscReal omega,MatSORType flag,PetscReal fshift,PetscInt its,PetscInt lits,Vec xx)
1149: {
1150: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1151: PetscScalar *x,d,*xs,sum,*t,scale,*idiag=0,*mdiag;
1152: const MatScalar *v = a->a;
1153: const PetscScalar *b, *bs,*xb, *ts;
1154: PetscErrorCode ierr;
1155: PetscInt n = A->cmap->n,m = A->rmap->n,i;
1156: const PetscInt *idx,*diag;
1159: its = its*lits;
1161: if (fshift != a->fshift || omega != a->omega) a->idiagvalid = PETSC_FALSE; /* must recompute idiag[] */
1162: if (!a->idiagvalid) {MatInvertDiagonal_SeqAIJ(A,omega,fshift);}
1163: a->fshift = fshift;
1164: a->omega = omega;
1166: diag = a->diag;
1167: t = a->ssor_work;
1168: idiag = a->idiag;
1169: mdiag = a->mdiag;
1171: VecGetArray(xx,&x);
1172: if (xx != bb) {
1173: VecGetArray(bb,(PetscScalar**)&b);
1174: } else {
1175: b = x;
1176: }
1177: CHKMEMQ;
1178: /* We count flops by assuming the upper triangular and lower triangular parts have the same number of nonzeros */
1179: xs = x;
1180: if (flag == SOR_APPLY_UPPER) {
1181: /* apply (U + D/omega) to the vector */
1182: bs = b;
1183: for (i=0; i<m; i++) {
1184: d = fshift + mdiag[i];
1185: n = a->i[i+1] - diag[i] - 1;
1186: idx = a->j + diag[i] + 1;
1187: v = a->a + diag[i] + 1;
1188: sum = b[i]*d/omega;
1189: SPARSEDENSEDOT(sum,bs,v,idx,n);
1190: x[i] = sum;
1191: }
1192: VecRestoreArray(xx,&x);
1193: if (bb != xx) {VecRestoreArray(bb,(PetscScalar**)&b);}
1194: PetscLogFlops(a->nz);
1195: return(0);
1196: }
1198: if (flag == SOR_APPLY_LOWER) {
1199: SETERRQ(PETSC_ERR_SUP,"SOR_APPLY_LOWER is not implemented");
1200: } else if (flag & SOR_EISENSTAT) {
1201: /* Let A = L + U + D; where L is lower trianglar,
1202: U is upper triangular, E is diagonal; This routine applies
1204: (L + E)^{-1} A (U + E)^{-1}
1206: to a vector efficiently using Eisenstat's trick. This is for
1207: the case of SSOR preconditioner, so E is D/omega where omega
1208: is the relaxation factor.
1209: */
1210: scale = (2.0/omega) - 1.0;
1212: /* x = (E + U)^{-1} b */
1213: for (i=m-1; i>=0; i--) {
1214: n = a->i[i+1] - diag[i] - 1;
1215: idx = a->j + diag[i] + 1;
1216: v = a->a + diag[i] + 1;
1217: sum = b[i];
1218: SPARSEDENSEMDOT(sum,xs,v,idx,n);
1219: x[i] = sum*idiag[i];
1220: }
1222: /* t = b - (2*E - D)x */
1223: v = a->a;
1224: for (i=0; i<m; i++) { t[i] = b[i] - scale*(v[*diag++])*x[i]; }
1226: /* t = (E + L)^{-1}t */
1227: ts = t;
1228: diag = a->diag;
1229: for (i=0; i<m; i++) {
1230: n = diag[i] - a->i[i];
1231: idx = a->j + a->i[i];
1232: v = a->a + a->i[i];
1233: sum = t[i];
1234: SPARSEDENSEMDOT(sum,ts,v,idx,n);
1235: t[i] = sum*idiag[i];
1236: /* x = x + t */
1237: x[i] += t[i];
1238: }
1240: PetscLogFlops(6*m-1 + 2*a->nz);
1241: VecRestoreArray(xx,&x);
1242: if (bb != xx) {VecRestoreArray(bb,(PetscScalar**)&b);}
1243: return(0);
1244: }
1245: if (flag & SOR_ZERO_INITIAL_GUESS) {
1246: if (flag & SOR_FORWARD_SWEEP || flag & SOR_LOCAL_FORWARD_SWEEP){
1247: #if defined(PETSC_USE_FORTRAN_KERNEL_RELAXAIJ)
1248: fortranrelaxaijforwardzero_(&m,&omega,x,a->i,a->j,(PetscInt*)diag,idiag,a->a,(void*)b);
1249: #else
1250: for (i=0; i<m; i++) {
1251: n = diag[i] - a->i[i];
1252: idx = a->j + a->i[i];
1253: v = a->a + a->i[i];
1254: sum = b[i];
1255: SPARSEDENSEMDOT(sum,xs,v,idx,n);
1256: x[i] = sum*idiag[i];
1257: }
1258: #endif
1259: xb = x;
1260: PetscLogFlops(a->nz);
1261: } else xb = b;
1262: if ((flag & SOR_FORWARD_SWEEP || flag & SOR_LOCAL_FORWARD_SWEEP) &&
1263: (flag & SOR_BACKWARD_SWEEP || flag & SOR_LOCAL_BACKWARD_SWEEP)) {
1264: for (i=0; i<m; i++) {
1265: x[i] *= mdiag[i];
1266: }
1267: PetscLogFlops(m);
1268: }
1269: if (flag & SOR_BACKWARD_SWEEP || flag & SOR_LOCAL_BACKWARD_SWEEP){
1270: #if defined(PETSC_USE_FORTRAN_KERNEL_RELAXAIJ)
1271: fortranrelaxaijbackwardzero_(&m,&omega,x,a->i,a->j,(PetscInt*)diag,idiag,a->a,(void*)xb);
1272: #else
1273: for (i=m-1; i>=0; i--) {
1274: n = a->i[i+1] - diag[i] - 1;
1275: idx = a->j + diag[i] + 1;
1276: v = a->a + diag[i] + 1;
1277: sum = xb[i];
1278: SPARSEDENSEMDOT(sum,xs,v,idx,n);
1279: x[i] = sum*idiag[i];
1280: }
1281: #endif
1282: PetscLogFlops(a->nz);
1283: }
1284: its--;
1285: }
1286: while (its--) {
1287: if (flag & SOR_FORWARD_SWEEP || flag & SOR_LOCAL_FORWARD_SWEEP){
1288: #if defined(PETSC_USE_FORTRAN_KERNEL_RELAXAIJ)
1289: fortranrelaxaijforward_(&m,&omega,x,a->i,a->j,(PetscInt*)diag,a->a,(void*)b);
1290: #else
1291: for (i=0; i<m; i++) {
1292: n = a->i[i+1] - a->i[i];
1293: idx = a->j + a->i[i];
1294: v = a->a + a->i[i];
1295: sum = b[i];
1296: SPARSEDENSEMDOT(sum,xs,v,idx,n);
1297: x[i] = (1. - omega)*x[i] + (sum + mdiag[i]*x[i])*idiag[i];
1298: }
1299: #endif
1300: PetscLogFlops(a->nz);
1301: }
1302: if (flag & SOR_BACKWARD_SWEEP || flag & SOR_LOCAL_BACKWARD_SWEEP){
1303: #if defined(PETSC_USE_FORTRAN_KERNEL_RELAXAIJ)
1304: fortranrelaxaijbackward_(&m,&omega,x,a->i,a->j,(PetscInt*)diag,a->a,(void*)b);
1305: #else
1306: for (i=m-1; i>=0; i--) {
1307: n = a->i[i+1] - a->i[i];
1308: idx = a->j + a->i[i];
1309: v = a->a + a->i[i];
1310: sum = b[i];
1311: SPARSEDENSEMDOT(sum,xs,v,idx,n);
1312: x[i] = (1. - omega)*x[i] + (sum + mdiag[i]*x[i])*idiag[i];
1313: }
1314: #endif
1315: PetscLogFlops(a->nz);
1316: }
1317: }
1318: VecRestoreArray(xx,&x);
1319: if (bb != xx) {VecRestoreArray(bb,(PetscScalar**)&b);}
1320: CHKMEMQ; return(0);
1321: }
1326: PetscErrorCode MatGetInfo_SeqAIJ(Mat A,MatInfoType flag,MatInfo *info)
1327: {
1328: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1331: info->block_size = 1.0;
1332: info->nz_allocated = (double)a->maxnz;
1333: info->nz_used = (double)a->nz;
1334: info->nz_unneeded = (double)(a->maxnz - a->nz);
1335: info->assemblies = (double)A->num_ass;
1336: info->mallocs = (double)a->reallocs;
1337: info->memory = ((PetscObject)A)->mem;
1338: if (A->factor) {
1339: info->fill_ratio_given = A->info.fill_ratio_given;
1340: info->fill_ratio_needed = A->info.fill_ratio_needed;
1341: info->factor_mallocs = A->info.factor_mallocs;
1342: } else {
1343: info->fill_ratio_given = 0;
1344: info->fill_ratio_needed = 0;
1345: info->factor_mallocs = 0;
1346: }
1347: return(0);
1348: }
1352: PetscErrorCode MatZeroRows_SeqAIJ(Mat A,PetscInt N,const PetscInt rows[],PetscScalar diag)
1353: {
1354: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1355: PetscInt i,m = A->rmap->n - 1,d;
1357: PetscTruth missing;
1360: if (a->keepzeroedrows) {
1361: for (i=0; i<N; i++) {
1362: if (rows[i] < 0 || rows[i] > m) SETERRQ1(PETSC_ERR_ARG_OUTOFRANGE,"row %D out of range", rows[i]);
1363: PetscMemzero(&a->a[a->i[rows[i]]],a->ilen[rows[i]]*sizeof(PetscScalar));
1364: }
1365: if (diag != 0.0) {
1366: MatMissingDiagonal_SeqAIJ(A,&missing,&d);
1367: if (missing) SETERRQ1(PETSC_ERR_ARG_WRONGSTATE,"Matrix is missing diagonal entry in row %D",d);
1368: for (i=0; i<N; i++) {
1369: a->a[a->diag[rows[i]]] = diag;
1370: }
1371: }
1372: A->same_nonzero = PETSC_TRUE;
1373: } else {
1374: if (diag != 0.0) {
1375: for (i=0; i<N; i++) {
1376: if (rows[i] < 0 || rows[i] > m) SETERRQ1(PETSC_ERR_ARG_OUTOFRANGE,"row %D out of range", rows[i]);
1377: if (a->ilen[rows[i]] > 0) {
1378: a->ilen[rows[i]] = 1;
1379: a->a[a->i[rows[i]]] = diag;
1380: a->j[a->i[rows[i]]] = rows[i];
1381: } else { /* in case row was completely empty */
1382: MatSetValues_SeqAIJ(A,1,&rows[i],1,&rows[i],&diag,INSERT_VALUES);
1383: }
1384: }
1385: } else {
1386: for (i=0; i<N; i++) {
1387: if (rows[i] < 0 || rows[i] > m) SETERRQ1(PETSC_ERR_ARG_OUTOFRANGE,"row %D out of range", rows[i]);
1388: a->ilen[rows[i]] = 0;
1389: }
1390: }
1391: A->same_nonzero = PETSC_FALSE;
1392: }
1393: MatAssemblyEnd_SeqAIJ(A,MAT_FINAL_ASSEMBLY);
1394: return(0);
1395: }
1399: PetscErrorCode MatGetRow_SeqAIJ(Mat A,PetscInt row,PetscInt *nz,PetscInt **idx,PetscScalar **v)
1400: {
1401: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1402: PetscInt *itmp;
1405: if (row < 0 || row >= A->rmap->n) SETERRQ1(PETSC_ERR_ARG_OUTOFRANGE,"Row %D out of range",row);
1407: *nz = a->i[row+1] - a->i[row];
1408: if (v) *v = a->a + a->i[row];
1409: if (idx) {
1410: itmp = a->j + a->i[row];
1411: if (*nz) {
1412: *idx = itmp;
1413: }
1414: else *idx = 0;
1415: }
1416: return(0);
1417: }
1419: /* remove this function? */
1422: PetscErrorCode MatRestoreRow_SeqAIJ(Mat A,PetscInt row,PetscInt *nz,PetscInt **idx,PetscScalar **v)
1423: {
1425: return(0);
1426: }
1430: PetscErrorCode MatNorm_SeqAIJ(Mat A,NormType type,PetscReal *nrm)
1431: {
1432: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1433: MatScalar *v = a->a;
1434: PetscReal sum = 0.0;
1436: PetscInt i,j;
1439: if (type == NORM_FROBENIUS) {
1440: for (i=0; i<a->nz; i++) {
1441: #if defined(PETSC_USE_COMPLEX)
1442: sum += PetscRealPart(PetscConj(*v)*(*v)); v++;
1443: #else
1444: sum += (*v)*(*v); v++;
1445: #endif
1446: }
1447: *nrm = sqrt(sum);
1448: } else if (type == NORM_1) {
1449: PetscReal *tmp;
1450: PetscInt *jj = a->j;
1451: PetscMalloc((A->cmap->n+1)*sizeof(PetscReal),&tmp);
1452: PetscMemzero(tmp,A->cmap->n*sizeof(PetscReal));
1453: *nrm = 0.0;
1454: for (j=0; j<a->nz; j++) {
1455: tmp[*jj++] += PetscAbsScalar(*v); v++;
1456: }
1457: for (j=0; j<A->cmap->n; j++) {
1458: if (tmp[j] > *nrm) *nrm = tmp[j];
1459: }
1460: PetscFree(tmp);
1461: } else if (type == NORM_INFINITY) {
1462: *nrm = 0.0;
1463: for (j=0; j<A->rmap->n; j++) {
1464: v = a->a + a->i[j];
1465: sum = 0.0;
1466: for (i=0; i<a->i[j+1]-a->i[j]; i++) {
1467: sum += PetscAbsScalar(*v); v++;
1468: }
1469: if (sum > *nrm) *nrm = sum;
1470: }
1471: } else {
1472: SETERRQ(PETSC_ERR_SUP,"No support for two norm");
1473: }
1474: return(0);
1475: }
1479: PetscErrorCode MatTranspose_SeqAIJ(Mat A,MatReuse reuse,Mat *B)
1480: {
1481: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1482: Mat C;
1484: PetscInt i,*aj = a->j,*ai = a->i,m = A->rmap->n,len,*col;
1485: MatScalar *array = a->a;
1488: if (reuse == MAT_REUSE_MATRIX && A == *B && m != A->cmap->n) SETERRQ(PETSC_ERR_ARG_SIZ,"Square matrix only for in-place");
1490: if (reuse == MAT_INITIAL_MATRIX || *B == A) {
1491: PetscMalloc((1+A->cmap->n)*sizeof(PetscInt),&col);
1492: PetscMemzero(col,(1+A->cmap->n)*sizeof(PetscInt));
1493:
1494: for (i=0; i<ai[m]; i++) col[aj[i]] += 1;
1495: MatCreate(((PetscObject)A)->comm,&C);
1496: MatSetSizes(C,A->cmap->n,m,A->cmap->n,m);
1497: MatSetType(C,((PetscObject)A)->type_name);
1498: MatSeqAIJSetPreallocation_SeqAIJ(C,0,col);
1499: PetscFree(col);
1500: } else {
1501: C = *B;
1502: }
1504: for (i=0; i<m; i++) {
1505: len = ai[i+1]-ai[i];
1506: MatSetValues_SeqAIJ(C,len,aj,1,&i,array,INSERT_VALUES);
1507: array += len;
1508: aj += len;
1509: }
1510: MatAssemblyBegin(C,MAT_FINAL_ASSEMBLY);
1511: MatAssemblyEnd(C,MAT_FINAL_ASSEMBLY);
1513: if (reuse == MAT_INITIAL_MATRIX || *B != A) {
1514: *B = C;
1515: } else {
1516: MatHeaderCopy(A,C);
1517: }
1518: return(0);
1519: }
1524: PetscErrorCode MatIsTranspose_SeqAIJ(Mat A,Mat B,PetscReal tol,PetscTruth *f)
1525: {
1526: Mat_SeqAIJ *aij = (Mat_SeqAIJ *) A->data,*bij = (Mat_SeqAIJ*) A->data;
1527: PetscInt *adx,*bdx,*aii,*bii,*aptr,*bptr;
1528: MatScalar *va,*vb;
1530: PetscInt ma,na,mb,nb, i;
1533: bij = (Mat_SeqAIJ *) B->data;
1534:
1535: MatGetSize(A,&ma,&na);
1536: MatGetSize(B,&mb,&nb);
1537: if (ma!=nb || na!=mb){
1538: *f = PETSC_FALSE;
1539: return(0);
1540: }
1541: aii = aij->i; bii = bij->i;
1542: adx = aij->j; bdx = bij->j;
1543: va = aij->a; vb = bij->a;
1544: PetscMalloc(ma*sizeof(PetscInt),&aptr);
1545: PetscMalloc(mb*sizeof(PetscInt),&bptr);
1546: for (i=0; i<ma; i++) aptr[i] = aii[i];
1547: for (i=0; i<mb; i++) bptr[i] = bii[i];
1549: *f = PETSC_TRUE;
1550: for (i=0; i<ma; i++) {
1551: while (aptr[i]<aii[i+1]) {
1552: PetscInt idc,idr;
1553: PetscScalar vc,vr;
1554: /* column/row index/value */
1555: idc = adx[aptr[i]];
1556: idr = bdx[bptr[idc]];
1557: vc = va[aptr[i]];
1558: vr = vb[bptr[idc]];
1559: if (i!=idr || PetscAbsScalar(vc-vr) > tol) {
1560: *f = PETSC_FALSE;
1561: goto done;
1562: } else {
1563: aptr[i]++;
1564: if (B || i!=idc) bptr[idc]++;
1565: }
1566: }
1567: }
1568: done:
1569: PetscFree(aptr);
1570: if (B) {
1571: PetscFree(bptr);
1572: }
1573: return(0);
1574: }
1580: PetscErrorCode MatIsHermitianTranspose_SeqAIJ(Mat A,Mat B,PetscReal tol,PetscTruth *f)
1581: {
1582: Mat_SeqAIJ *aij = (Mat_SeqAIJ *) A->data,*bij = (Mat_SeqAIJ*) A->data;
1583: PetscInt *adx,*bdx,*aii,*bii,*aptr,*bptr;
1584: MatScalar *va,*vb;
1586: PetscInt ma,na,mb,nb, i;
1589: bij = (Mat_SeqAIJ *) B->data;
1590:
1591: MatGetSize(A,&ma,&na);
1592: MatGetSize(B,&mb,&nb);
1593: if (ma!=nb || na!=mb){
1594: *f = PETSC_FALSE;
1595: return(0);
1596: }
1597: aii = aij->i; bii = bij->i;
1598: adx = aij->j; bdx = bij->j;
1599: va = aij->a; vb = bij->a;
1600: PetscMalloc(ma*sizeof(PetscInt),&aptr);
1601: PetscMalloc(mb*sizeof(PetscInt),&bptr);
1602: for (i=0; i<ma; i++) aptr[i] = aii[i];
1603: for (i=0; i<mb; i++) bptr[i] = bii[i];
1605: *f = PETSC_TRUE;
1606: for (i=0; i<ma; i++) {
1607: while (aptr[i]<aii[i+1]) {
1608: PetscInt idc,idr;
1609: PetscScalar vc,vr;
1610: /* column/row index/value */
1611: idc = adx[aptr[i]];
1612: idr = bdx[bptr[idc]];
1613: vc = va[aptr[i]];
1614: vr = vb[bptr[idc]];
1615: if (i!=idr || PetscAbsScalar(vc-PetscConj(vr)) > tol) {
1616: *f = PETSC_FALSE;
1617: goto done;
1618: } else {
1619: aptr[i]++;
1620: if (B || i!=idc) bptr[idc]++;
1621: }
1622: }
1623: }
1624: done:
1625: PetscFree(aptr);
1626: if (B) {
1627: PetscFree(bptr);
1628: }
1629: return(0);
1630: }
1635: PetscErrorCode MatIsSymmetric_SeqAIJ(Mat A,PetscReal tol,PetscTruth *f)
1636: {
1639: MatIsTranspose_SeqAIJ(A,A,tol,f);
1640: return(0);
1641: }
1645: PetscErrorCode MatIsHermitian_SeqAIJ(Mat A,PetscReal tol,PetscTruth *f)
1646: {
1649: MatIsHermitianTranspose_SeqAIJ(A,A,tol,f);
1650: return(0);
1651: }
1655: PetscErrorCode MatDiagonalScale_SeqAIJ(Mat A,Vec ll,Vec rr)
1656: {
1657: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1658: PetscScalar *l,*r,x;
1659: MatScalar *v;
1661: PetscInt i,j,m = A->rmap->n,n = A->cmap->n,M,nz = a->nz,*jj;
1664: if (ll) {
1665: /* The local size is used so that VecMPI can be passed to this routine
1666: by MatDiagonalScale_MPIAIJ */
1667: VecGetLocalSize(ll,&m);
1668: if (m != A->rmap->n) SETERRQ(PETSC_ERR_ARG_SIZ,"Left scaling vector wrong length");
1669: VecGetArray(ll,&l);
1670: v = a->a;
1671: for (i=0; i<m; i++) {
1672: x = l[i];
1673: M = a->i[i+1] - a->i[i];
1674: for (j=0; j<M; j++) { (*v++) *= x;}
1675: }
1676: VecRestoreArray(ll,&l);
1677: PetscLogFlops(nz);
1678: }
1679: if (rr) {
1680: VecGetLocalSize(rr,&n);
1681: if (n != A->cmap->n) SETERRQ(PETSC_ERR_ARG_SIZ,"Right scaling vector wrong length");
1682: VecGetArray(rr,&r);
1683: v = a->a; jj = a->j;
1684: for (i=0; i<nz; i++) {
1685: (*v++) *= r[*jj++];
1686: }
1687: VecRestoreArray(rr,&r);
1688: PetscLogFlops(nz);
1689: }
1690: return(0);
1691: }
1695: PetscErrorCode MatGetSubMatrix_SeqAIJ(Mat A,IS isrow,IS iscol,PetscInt csize,MatReuse scall,Mat *B)
1696: {
1697: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data,*c;
1699: PetscInt *smap,i,k,kstart,kend,oldcols = A->cmap->n,*lens;
1700: PetscInt row,mat_i,*mat_j,tcol,first,step,*mat_ilen,sum,lensi;
1701: const PetscInt *irow,*icol;
1702: PetscInt nrows,ncols;
1703: PetscInt *starts,*j_new,*i_new,*aj = a->j,*ai = a->i,ii,*ailen = a->ilen;
1704: MatScalar *a_new,*mat_a;
1705: Mat C;
1706: PetscTruth stride,sorted;
1709: ISSorted(isrow,&sorted);
1710: if (!sorted) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"ISrow is not sorted");
1711: ISSorted(iscol,&sorted);
1712: if (!sorted) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"IScol is not sorted");
1714: ISGetIndices(isrow,&irow);
1715: ISGetLocalSize(isrow,&nrows);
1716: ISGetLocalSize(iscol,&ncols);
1718: ISStrideGetInfo(iscol,&first,&step);
1719: ISStride(iscol,&stride);
1720: if (stride && step == 1) {
1721: /* special case of contiguous rows */
1722: PetscMalloc((2*nrows+1)*sizeof(PetscInt),&lens);
1723: starts = lens + nrows;
1724: /* loop over new rows determining lens and starting points */
1725: for (i=0; i<nrows; i++) {
1726: kstart = ai[irow[i]];
1727: kend = kstart + ailen[irow[i]];
1728: for (k=kstart; k<kend; k++) {
1729: if (aj[k] >= first) {
1730: starts[i] = k;
1731: break;
1732: }
1733: }
1734: sum = 0;
1735: while (k < kend) {
1736: if (aj[k++] >= first+ncols) break;
1737: sum++;
1738: }
1739: lens[i] = sum;
1740: }
1741: /* create submatrix */
1742: if (scall == MAT_REUSE_MATRIX) {
1743: PetscInt n_cols,n_rows;
1744: MatGetSize(*B,&n_rows,&n_cols);
1745: if (n_rows != nrows || n_cols != ncols) SETERRQ(PETSC_ERR_ARG_SIZ,"Reused submatrix wrong size");
1746: MatZeroEntries(*B);
1747: C = *B;
1748: } else {
1749: MatCreate(((PetscObject)A)->comm,&C);
1750: MatSetSizes(C,nrows,ncols,PETSC_DETERMINE,PETSC_DETERMINE);
1751: MatSetType(C,((PetscObject)A)->type_name);
1752: MatSeqAIJSetPreallocation_SeqAIJ(C,0,lens);
1753: }
1754: c = (Mat_SeqAIJ*)C->data;
1756: /* loop over rows inserting into submatrix */
1757: a_new = c->a;
1758: j_new = c->j;
1759: i_new = c->i;
1761: for (i=0; i<nrows; i++) {
1762: ii = starts[i];
1763: lensi = lens[i];
1764: for (k=0; k<lensi; k++) {
1765: *j_new++ = aj[ii+k] - first;
1766: }
1767: PetscMemcpy(a_new,a->a + starts[i],lensi*sizeof(PetscScalar));
1768: a_new += lensi;
1769: i_new[i+1] = i_new[i] + lensi;
1770: c->ilen[i] = lensi;
1771: }
1772: PetscFree(lens);
1773: } else {
1774: ISGetIndices(iscol,&icol);
1775: PetscMalloc((1+oldcols)*sizeof(PetscInt),&smap);
1776:
1777: PetscMalloc((1+nrows)*sizeof(PetscInt),&lens);
1778: PetscMemzero(smap,oldcols*sizeof(PetscInt));
1779: for (i=0; i<ncols; i++) smap[icol[i]] = i+1;
1780: /* determine lens of each row */
1781: for (i=0; i<nrows; i++) {
1782: kstart = ai[irow[i]];
1783: kend = kstart + a->ilen[irow[i]];
1784: lens[i] = 0;
1785: for (k=kstart; k<kend; k++) {
1786: if (smap[aj[k]]) {
1787: lens[i]++;
1788: }
1789: }
1790: }
1791: /* Create and fill new matrix */
1792: if (scall == MAT_REUSE_MATRIX) {
1793: PetscTruth equal;
1795: c = (Mat_SeqAIJ *)((*B)->data);
1796: if ((*B)->rmap->n != nrows || (*B)->cmap->n != ncols) SETERRQ(PETSC_ERR_ARG_SIZ,"Cannot reuse matrix. wrong size");
1797: PetscMemcmp(c->ilen,lens,(*B)->rmap->n*sizeof(PetscInt),&equal);
1798: if (!equal) {
1799: SETERRQ(PETSC_ERR_ARG_SIZ,"Cannot reuse matrix. wrong no of nonzeros");
1800: }
1801: PetscMemzero(c->ilen,(*B)->rmap->n*sizeof(PetscInt));
1802: C = *B;
1803: } else {
1804: MatCreate(((PetscObject)A)->comm,&C);
1805: MatSetSizes(C,nrows,ncols,PETSC_DETERMINE,PETSC_DETERMINE);
1806: MatSetType(C,((PetscObject)A)->type_name);
1807: MatSeqAIJSetPreallocation_SeqAIJ(C,0,lens);
1808: }
1809: c = (Mat_SeqAIJ *)(C->data);
1810: for (i=0; i<nrows; i++) {
1811: row = irow[i];
1812: kstart = ai[row];
1813: kend = kstart + a->ilen[row];
1814: mat_i = c->i[i];
1815: mat_j = c->j + mat_i;
1816: mat_a = c->a + mat_i;
1817: mat_ilen = c->ilen + i;
1818: for (k=kstart; k<kend; k++) {
1819: if ((tcol=smap[a->j[k]])) {
1820: *mat_j++ = tcol - 1;
1821: *mat_a++ = a->a[k];
1822: (*mat_ilen)++;
1824: }
1825: }
1826: }
1827: /* Free work space */
1828: ISRestoreIndices(iscol,&icol);
1829: PetscFree(smap);
1830: PetscFree(lens);
1831: }
1832: MatAssemblyBegin(C,MAT_FINAL_ASSEMBLY);
1833: MatAssemblyEnd(C,MAT_FINAL_ASSEMBLY);
1835: ISRestoreIndices(isrow,&irow);
1836: *B = C;
1837: return(0);
1838: }
1840: /*
1841: */
1844: PetscErrorCode MatILUFactor_SeqAIJ(Mat inA,IS row,IS col,const MatFactorInfo *info)
1845: {
1846: Mat_SeqAIJ *a = (Mat_SeqAIJ*)inA->data;
1848: Mat outA;
1849: PetscTruth row_identity,col_identity;
1852: if (info->levels != 0) SETERRQ(PETSC_ERR_SUP,"Only levels=0 supported for in-place ilu");
1853: ISIdentity(row,&row_identity);
1854: ISIdentity(col,&col_identity);
1856: outA = inA;
1857: inA->factor = MAT_FACTOR_LU;
1858: PetscObjectReference((PetscObject)row);
1859: if (a->row) { ISDestroy(a->row); }
1860: a->row = row;
1861: PetscObjectReference((PetscObject)col);
1862: if (a->col) { ISDestroy(a->col); }
1863: a->col = col;
1865: /* Create the inverse permutation so that it can be used in MatLUFactorNumeric() */
1866: if (a->icol) {ISDestroy(a->icol);} /* need to remove old one */
1867: ISInvertPermutation(col,PETSC_DECIDE,&a->icol);
1868: PetscLogObjectParent(inA,a->icol);
1870: if (!a->solve_work) { /* this matrix may have been factored before */
1871: PetscMalloc((inA->rmap->n+1)*sizeof(PetscScalar),&a->solve_work);
1872: PetscLogObjectMemory(inA, (inA->rmap->n+1)*sizeof(PetscScalar));
1873: }
1875: MatMarkDiagonal_SeqAIJ(inA);
1876: if (row_identity && col_identity) {
1877: MatLUFactorNumeric_SeqAIJ(outA,inA,info);
1878: } else {
1879: MatLUFactorNumeric_SeqAIJ_InplaceWithPerm(outA,inA,info);
1880: }
1881: return(0);
1882: }
1884: #include petscblaslapack.h
1887: PetscErrorCode MatScale_SeqAIJ(Mat inA,PetscScalar alpha)
1888: {
1889: Mat_SeqAIJ *a = (Mat_SeqAIJ*)inA->data;
1890: PetscScalar oalpha = alpha;
1892: PetscBLASInt one = 1,bnz = PetscBLASIntCast(a->nz);
1895: BLASscal_(&bnz,&oalpha,a->a,&one);
1896: PetscLogFlops(a->nz);
1897: return(0);
1898: }
1902: PetscErrorCode MatGetSubMatrices_SeqAIJ(Mat A,PetscInt n,const IS irow[],const IS icol[],MatReuse scall,Mat *B[])
1903: {
1905: PetscInt i;
1908: if (scall == MAT_INITIAL_MATRIX) {
1909: PetscMalloc((n+1)*sizeof(Mat),B);
1910: }
1912: for (i=0; i<n; i++) {
1913: MatGetSubMatrix_SeqAIJ(A,irow[i],icol[i],PETSC_DECIDE,scall,&(*B)[i]);
1914: }
1915: return(0);
1916: }
1920: PetscErrorCode MatIncreaseOverlap_SeqAIJ(Mat A,PetscInt is_max,IS is[],PetscInt ov)
1921: {
1922: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1924: PetscInt row,i,j,k,l,m,n,*nidx,isz,val;
1925: const PetscInt *idx;
1926: PetscInt start,end,*ai,*aj;
1927: PetscBT table;
1930: m = A->rmap->n;
1931: ai = a->i;
1932: aj = a->j;
1934: if (ov < 0) SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,"illegal negative overlap value used");
1936: PetscMalloc((m+1)*sizeof(PetscInt),&nidx);
1937: PetscBTCreate(m,table);
1939: for (i=0; i<is_max; i++) {
1940: /* Initialize the two local arrays */
1941: isz = 0;
1942: PetscBTMemzero(m,table);
1943:
1944: /* Extract the indices, assume there can be duplicate entries */
1945: ISGetIndices(is[i],&idx);
1946: ISGetLocalSize(is[i],&n);
1947:
1948: /* Enter these into the temp arrays. I.e., mark table[row], enter row into new index */
1949: for (j=0; j<n ; ++j){
1950: if(!PetscBTLookupSet(table,idx[j])) { nidx[isz++] = idx[j];}
1951: }
1952: ISRestoreIndices(is[i],&idx);
1953: ISDestroy(is[i]);
1954:
1955: k = 0;
1956: for (j=0; j<ov; j++){ /* for each overlap */
1957: n = isz;
1958: for (; k<n ; k++){ /* do only those rows in nidx[k], which are not done yet */
1959: row = nidx[k];
1960: start = ai[row];
1961: end = ai[row+1];
1962: for (l = start; l<end ; l++){
1963: val = aj[l] ;
1964: if (!PetscBTLookupSet(table,val)) {nidx[isz++] = val;}
1965: }
1966: }
1967: }
1968: ISCreateGeneral(PETSC_COMM_SELF,isz,nidx,(is+i));
1969: }
1970: PetscBTDestroy(table);
1971: PetscFree(nidx);
1972: return(0);
1973: }
1975: /* -------------------------------------------------------------- */
1978: PetscErrorCode MatPermute_SeqAIJ(Mat A,IS rowp,IS colp,Mat *B)
1979: {
1980: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
1982: PetscInt i,nz,m = A->rmap->n,n = A->cmap->n;
1983: const PetscInt *row,*col;
1984: PetscInt *cnew,j,*lens;
1985: IS icolp,irowp;
1986: PetscInt *cwork;
1987: PetscScalar *vwork;
1990: ISInvertPermutation(rowp,PETSC_DECIDE,&irowp);
1991: ISGetIndices(irowp,&row);
1992: ISInvertPermutation(colp,PETSC_DECIDE,&icolp);
1993: ISGetIndices(icolp,&col);
1994:
1995: /* determine lengths of permuted rows */
1996: PetscMalloc((m+1)*sizeof(PetscInt),&lens);
1997: for (i=0; i<m; i++) {
1998: lens[row[i]] = a->i[i+1] - a->i[i];
1999: }
2000: MatCreate(((PetscObject)A)->comm,B);
2001: MatSetSizes(*B,m,n,m,n);
2002: MatSetType(*B,((PetscObject)A)->type_name);
2003: MatSeqAIJSetPreallocation_SeqAIJ(*B,0,lens);
2004: PetscFree(lens);
2006: PetscMalloc(n*sizeof(PetscInt),&cnew);
2007: for (i=0; i<m; i++) {
2008: MatGetRow_SeqAIJ(A,i,&nz,&cwork,&vwork);
2009: for (j=0; j<nz; j++) { cnew[j] = col[cwork[j]];}
2010: MatSetValues_SeqAIJ(*B,1,&row[i],nz,cnew,vwork,INSERT_VALUES);
2011: MatRestoreRow_SeqAIJ(A,i,&nz,&cwork,&vwork);
2012: }
2013: PetscFree(cnew);
2014: (*B)->assembled = PETSC_FALSE;
2015: MatAssemblyBegin(*B,MAT_FINAL_ASSEMBLY);
2016: MatAssemblyEnd(*B,MAT_FINAL_ASSEMBLY);
2017: ISRestoreIndices(irowp,&row);
2018: ISRestoreIndices(icolp,&col);
2019: ISDestroy(irowp);
2020: ISDestroy(icolp);
2021: return(0);
2022: }
2026: PetscErrorCode MatCopy_SeqAIJ(Mat A,Mat B,MatStructure str)
2027: {
2031: /* If the two matrices have the same copy implementation, use fast copy. */
2032: if (str == SAME_NONZERO_PATTERN && (A->ops->copy == B->ops->copy)) {
2033: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
2034: Mat_SeqAIJ *b = (Mat_SeqAIJ*)B->data;
2036: if (a->i[A->rmap->n] != b->i[B->rmap->n]) {
2037: SETERRQ(PETSC_ERR_ARG_INCOMP,"Number of nonzeros in two matrices are different");
2038: }
2039: PetscMemcpy(b->a,a->a,(a->i[A->rmap->n])*sizeof(PetscScalar));
2040: } else {
2041: MatCopy_Basic(A,B,str);
2042: }
2043: return(0);
2044: }
2048: PetscErrorCode MatSetUpPreallocation_SeqAIJ(Mat A)
2049: {
2053: MatSeqAIJSetPreallocation_SeqAIJ(A,PETSC_DEFAULT,0);
2054: return(0);
2055: }
2059: PetscErrorCode MatGetArray_SeqAIJ(Mat A,PetscScalar *array[])
2060: {
2061: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
2063: *array = a->a;
2064: return(0);
2065: }
2069: PetscErrorCode MatRestoreArray_SeqAIJ(Mat A,PetscScalar *array[])
2070: {
2072: return(0);
2073: }
2077: PetscErrorCode MatFDColoringApply_SeqAIJ(Mat J,MatFDColoring coloring,Vec x1,MatStructure *flag,void *sctx)
2078: {
2079: PetscErrorCode (*f)(void*,Vec,Vec,void*) = (PetscErrorCode (*)(void*,Vec,Vec,void *))coloring->f;
2081: PetscInt k,N,start,end,l,row,col,srow,**vscaleforrow,m1,m2;
2082: PetscScalar dx,*y,*xx,*w3_array;
2083: PetscScalar *vscale_array;
2084: PetscReal epsilon = coloring->error_rel,umin = coloring->umin;
2085: Vec w1,w2,w3;
2086: void *fctx = coloring->fctx;
2087: PetscTruth flg;
2090: if (!coloring->w1) {
2091: VecDuplicate(x1,&coloring->w1);
2092: PetscLogObjectParent(coloring,coloring->w1);
2093: VecDuplicate(x1,&coloring->w2);
2094: PetscLogObjectParent(coloring,coloring->w2);
2095: VecDuplicate(x1,&coloring->w3);
2096: PetscLogObjectParent(coloring,coloring->w3);
2097: }
2098: w1 = coloring->w1; w2 = coloring->w2; w3 = coloring->w3;
2100: MatSetUnfactored(J);
2101: PetscOptionsHasName(((PetscObject)coloring)->prefix,"-mat_fd_coloring_dont_rezero",&flg);
2102: if (flg) {
2103: PetscInfo(coloring,"Not calling MatZeroEntries()\n");
2104: } else {
2105: PetscTruth assembled;
2106: MatAssembled(J,&assembled);
2107: if (assembled) {
2108: MatZeroEntries(J);
2109: }
2110: }
2112: VecGetOwnershipRange(x1,&start,&end);
2113: VecGetSize(x1,&N);
2115: /*
2116: This is a horrible, horrible, hack. See DMMGComputeJacobian_Multigrid() it inproperly sets
2117: coloring->F for the coarser grids from the finest
2118: */
2119: if (coloring->F) {
2120: VecGetLocalSize(coloring->F,&m1);
2121: VecGetLocalSize(w1,&m2);
2122: if (m1 != m2) {
2123: coloring->F = 0;
2124: }
2125: }
2127: if (coloring->F) {
2128: w1 = coloring->F;
2129: coloring->F = 0;
2130: } else {
2131: PetscLogEventBegin(MAT_FDColoringFunction,0,0,0,0);
2132: (*f)(sctx,x1,w1,fctx);
2133: PetscLogEventEnd(MAT_FDColoringFunction,0,0,0,0);
2134: }
2136: /*
2137: Compute all the scale factors and share with other processors
2138: */
2139: VecGetArray(x1,&xx);xx = xx - start;
2140: VecGetArray(coloring->vscale,&vscale_array);vscale_array = vscale_array - start;
2141: for (k=0; k<coloring->ncolors; k++) {
2142: /*
2143: Loop over each column associated with color adding the
2144: perturbation to the vector w3.
2145: */
2146: for (l=0; l<coloring->ncolumns[k]; l++) {
2147: col = coloring->columns[k][l]; /* column of the matrix we are probing for */
2148: dx = xx[col];
2149: if (dx == 0.0) dx = 1.0;
2150: #if !defined(PETSC_USE_COMPLEX)
2151: if (dx < umin && dx >= 0.0) dx = umin;
2152: else if (dx < 0.0 && dx > -umin) dx = -umin;
2153: #else
2154: if (PetscAbsScalar(dx) < umin && PetscRealPart(dx) >= 0.0) dx = umin;
2155: else if (PetscRealPart(dx) < 0.0 && PetscAbsScalar(dx) < umin) dx = -umin;
2156: #endif
2157: dx *= epsilon;
2158: vscale_array[col] = 1.0/dx;
2159: }
2160: }
2161: vscale_array = vscale_array + start;VecRestoreArray(coloring->vscale,&vscale_array);
2162: VecGhostUpdateBegin(coloring->vscale,INSERT_VALUES,SCATTER_FORWARD);
2163: VecGhostUpdateEnd(coloring->vscale,INSERT_VALUES,SCATTER_FORWARD);
2165: /* VecView(coloring->vscale,PETSC_VIEWER_STDOUT_WORLD);
2166: VecView(x1,PETSC_VIEWER_STDOUT_WORLD);*/
2168: if (coloring->vscaleforrow) vscaleforrow = coloring->vscaleforrow;
2169: else vscaleforrow = coloring->columnsforrow;
2171: VecGetArray(coloring->vscale,&vscale_array);
2172: /*
2173: Loop over each color
2174: */
2175: for (k=0; k<coloring->ncolors; k++) {
2176: coloring->currentcolor = k;
2177: VecCopy(x1,w3);
2178: VecGetArray(w3,&w3_array);w3_array = w3_array - start;
2179: /*
2180: Loop over each column associated with color adding the
2181: perturbation to the vector w3.
2182: */
2183: for (l=0; l<coloring->ncolumns[k]; l++) {
2184: col = coloring->columns[k][l]; /* column of the matrix we are probing for */
2185: dx = xx[col];
2186: if (dx == 0.0) dx = 1.0;
2187: #if !defined(PETSC_USE_COMPLEX)
2188: if (dx < umin && dx >= 0.0) dx = umin;
2189: else if (dx < 0.0 && dx > -umin) dx = -umin;
2190: #else
2191: if (PetscAbsScalar(dx) < umin && PetscRealPart(dx) >= 0.0) dx = umin;
2192: else if (PetscRealPart(dx) < 0.0 && PetscAbsScalar(dx) < umin) dx = -umin;
2193: #endif
2194: dx *= epsilon;
2195: if (!PetscAbsScalar(dx)) SETERRQ(PETSC_ERR_PLIB,"Computed 0 differencing parameter");
2196: w3_array[col] += dx;
2197: }
2198: w3_array = w3_array + start; VecRestoreArray(w3,&w3_array);
2200: /*
2201: Evaluate function at x1 + dx (here dx is a vector of perturbations)
2202: */
2204: PetscLogEventBegin(MAT_FDColoringFunction,0,0,0,0);
2205: (*f)(sctx,w3,w2,fctx);
2206: PetscLogEventEnd(MAT_FDColoringFunction,0,0,0,0);
2207: VecAXPY(w2,-1.0,w1);
2209: /*
2210: Loop over rows of vector, putting results into Jacobian matrix
2211: */
2212: VecGetArray(w2,&y);
2213: for (l=0; l<coloring->nrows[k]; l++) {
2214: row = coloring->rows[k][l];
2215: col = coloring->columnsforrow[k][l];
2216: y[row] *= vscale_array[vscaleforrow[k][l]];
2217: srow = row + start;
2218: MatSetValues_SeqAIJ(J,1,&srow,1,&col,y+row,INSERT_VALUES);
2219: }
2220: VecRestoreArray(w2,&y);
2221: }
2222: coloring->currentcolor = k;
2223: VecRestoreArray(coloring->vscale,&vscale_array);
2224: xx = xx + start; VecRestoreArray(x1,&xx);
2225: MatAssemblyBegin(J,MAT_FINAL_ASSEMBLY);
2226: MatAssemblyEnd(J,MAT_FINAL_ASSEMBLY);
2227: return(0);
2228: }
2230: #include petscblaslapack.h
2233: PetscErrorCode MatAXPY_SeqAIJ(Mat Y,PetscScalar a,Mat X,MatStructure str)
2234: {
2236: PetscInt i;
2237: Mat_SeqAIJ *x = (Mat_SeqAIJ *)X->data,*y = (Mat_SeqAIJ *)Y->data;
2238: PetscBLASInt one=1,bnz = PetscBLASIntCast(x->nz);
2241: if (str == SAME_NONZERO_PATTERN) {
2242: PetscScalar alpha = a;
2243: BLASaxpy_(&bnz,&alpha,x->a,&one,y->a,&one);
2244: } else if (str == SUBSET_NONZERO_PATTERN) { /* nonzeros of X is a subset of Y's */
2245: if (y->xtoy && y->XtoY != X) {
2246: PetscFree(y->xtoy);
2247: MatDestroy(y->XtoY);
2248: }
2249: if (!y->xtoy) { /* get xtoy */
2250: MatAXPYGetxtoy_Private(X->rmap->n,x->i,x->j,PETSC_NULL, y->i,y->j,PETSC_NULL, &y->xtoy);
2251: y->XtoY = X;
2252: PetscObjectReference((PetscObject)X);
2253: }
2254: for (i=0; i<x->nz; i++) y->a[y->xtoy[i]] += a*(x->a[i]);
2255: PetscInfo3(Y,"ratio of nnz(X)/nnz(Y): %d/%d = %G\n",x->nz,y->nz,(PetscReal)(x->nz)/y->nz);
2256: } else {
2257: MatAXPY_Basic(Y,a,X,str);
2258: }
2259: return(0);
2260: }
2264: PetscErrorCode MatSetBlockSize_SeqAIJ(Mat A,PetscInt bs)
2265: {
2267: return(0);
2268: }
2272: PetscErrorCode MatConjugate_SeqAIJ(Mat mat)
2273: {
2274: #if defined(PETSC_USE_COMPLEX)
2275: Mat_SeqAIJ *aij = (Mat_SeqAIJ *)mat->data;
2276: PetscInt i,nz;
2277: PetscScalar *a;
2280: nz = aij->nz;
2281: a = aij->a;
2282: for (i=0; i<nz; i++) {
2283: a[i] = PetscConj(a[i]);
2284: }
2285: #else
2287: #endif
2288: return(0);
2289: }
2293: PetscErrorCode MatGetRowMaxAbs_SeqAIJ(Mat A,Vec v,PetscInt idx[])
2294: {
2295: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
2297: PetscInt i,j,m = A->rmap->n,*ai,*aj,ncols,n;
2298: PetscReal atmp;
2299: PetscScalar *x;
2300: MatScalar *aa;
2303: if (A->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2304: aa = a->a;
2305: ai = a->i;
2306: aj = a->j;
2308: VecSet(v,0.0);
2309: VecGetArray(v,&x);
2310: VecGetLocalSize(v,&n);
2311: if (n != A->rmap->n) SETERRQ(PETSC_ERR_ARG_SIZ,"Nonconforming matrix and vector");
2312: for (i=0; i<m; i++) {
2313: ncols = ai[1] - ai[0]; ai++;
2314: x[i] = 0.0;
2315: for (j=0; j<ncols; j++){
2316: atmp = PetscAbsScalar(*aa);
2317: if (PetscAbsScalar(x[i]) < atmp) {x[i] = atmp; if (idx) idx[i] = *aj;}
2318: aa++; aj++;
2319: }
2320: }
2321: VecRestoreArray(v,&x);
2322: return(0);
2323: }
2327: PetscErrorCode MatGetRowMax_SeqAIJ(Mat A,Vec v,PetscInt idx[])
2328: {
2329: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
2331: PetscInt i,j,m = A->rmap->n,*ai,*aj,ncols,n;
2332: PetscScalar *x;
2333: MatScalar *aa;
2336: if (A->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2337: aa = a->a;
2338: ai = a->i;
2339: aj = a->j;
2341: VecSet(v,0.0);
2342: VecGetArray(v,&x);
2343: VecGetLocalSize(v,&n);
2344: if (n != A->rmap->n) SETERRQ(PETSC_ERR_ARG_SIZ,"Nonconforming matrix and vector");
2345: for (i=0; i<m; i++) {
2346: ncols = ai[1] - ai[0]; ai++;
2347: if (ncols == A->cmap->n) { /* row is dense */
2348: x[i] = *aa; if (idx) idx[i] = 0;
2349: } else { /* row is sparse so already KNOW maximum is 0.0 or higher */
2350: x[i] = 0.0;
2351: if (idx) {
2352: idx[i] = 0; /* in case ncols is zero */
2353: for (j=0;j<ncols;j++) { /* find first implicit 0.0 in the row */
2354: if (aj[j] > j) {
2355: idx[i] = j;
2356: break;
2357: }
2358: }
2359: }
2360: }
2361: for (j=0; j<ncols; j++){
2362: if (PetscRealPart(x[i]) < PetscRealPart(*aa)) {x[i] = *aa; if (idx) idx[i] = *aj;}
2363: aa++; aj++;
2364: }
2365: }
2366: VecRestoreArray(v,&x);
2367: return(0);
2368: }
2372: PetscErrorCode MatGetRowMinAbs_SeqAIJ(Mat A,Vec v,PetscInt idx[])
2373: {
2374: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
2376: PetscInt i,j,m = A->rmap->n,*ai,*aj,ncols,n;
2377: PetscReal atmp;
2378: PetscScalar *x;
2379: MatScalar *aa;
2382: if (A->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2383: aa = a->a;
2384: ai = a->i;
2385: aj = a->j;
2387: VecSet(v,0.0);
2388: VecGetArray(v,&x);
2389: VecGetLocalSize(v,&n);
2390: if (n != A->rmap->n) SETERRQ(PETSC_ERR_ARG_SIZ,"Nonconforming matrix and vector");
2391: for (i=0; i<m; i++) {
2392: ncols = ai[1] - ai[0]; ai++;
2393: if (ncols) {
2394: /* Get first nonzero */
2395: for(j = 0; j < ncols; j++) {
2396: atmp = PetscAbsScalar(aa[j]);
2397: if (atmp > 1.0e-12) {x[i] = atmp; if (idx) idx[i] = aj[j]; break;}
2398: }
2399: if (j == ncols) {x[i] = *aa; if (idx) idx[i] = *aj;}
2400: } else {
2401: x[i] = 0.0; if (idx) idx[i] = 0;
2402: }
2403: for(j = 0; j < ncols; j++) {
2404: atmp = PetscAbsScalar(*aa);
2405: if (atmp > 1.0e-12 && PetscAbsScalar(x[i]) > atmp) {x[i] = atmp; if (idx) idx[i] = *aj;}
2406: aa++; aj++;
2407: }
2408: }
2409: VecRestoreArray(v,&x);
2410: return(0);
2411: }
2415: PetscErrorCode MatGetRowMin_SeqAIJ(Mat A,Vec v,PetscInt idx[])
2416: {
2417: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
2419: PetscInt i,j,m = A->rmap->n,*ai,*aj,ncols,n;
2420: PetscScalar *x;
2421: MatScalar *aa;
2424: if (A->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2425: aa = a->a;
2426: ai = a->i;
2427: aj = a->j;
2429: VecSet(v,0.0);
2430: VecGetArray(v,&x);
2431: VecGetLocalSize(v,&n);
2432: if (n != A->rmap->n) SETERRQ(PETSC_ERR_ARG_SIZ,"Nonconforming matrix and vector");
2433: for (i=0; i<m; i++) {
2434: ncols = ai[1] - ai[0]; ai++;
2435: if (ncols == A->cmap->n) { /* row is dense */
2436: x[i] = *aa; if (idx) idx[i] = 0;
2437: } else { /* row is sparse so already KNOW minimum is 0.0 or lower */
2438: x[i] = 0.0;
2439: if (idx) { /* find first implicit 0.0 in the row */
2440: idx[i] = 0; /* in case ncols is zero */
2441: for (j=0;j<ncols;j++) {
2442: if (aj[j] > j) {
2443: idx[i] = j;
2444: break;
2445: }
2446: }
2447: }
2448: }
2449: for (j=0; j<ncols; j++){
2450: if (PetscRealPart(x[i]) > PetscRealPart(*aa)) {x[i] = *aa; if (idx) idx[i] = *aj;}
2451: aa++; aj++;
2452: }
2453: }
2454: VecRestoreArray(v,&x);
2455: return(0);
2456: }
2458: /* -------------------------------------------------------------------*/
2459: static struct _MatOps MatOps_Values = {MatSetValues_SeqAIJ,
2460: MatGetRow_SeqAIJ,
2461: MatRestoreRow_SeqAIJ,
2462: MatMult_SeqAIJ,
2463: /* 4*/ MatMultAdd_SeqAIJ,
2464: MatMultTranspose_SeqAIJ,
2465: MatMultTransposeAdd_SeqAIJ,
2466: 0,
2467: 0,
2468: 0,
2469: /*10*/ 0,
2470: MatLUFactor_SeqAIJ,
2471: 0,
2472: MatRelax_SeqAIJ,
2473: MatTranspose_SeqAIJ,
2474: /*15*/ MatGetInfo_SeqAIJ,
2475: MatEqual_SeqAIJ,
2476: MatGetDiagonal_SeqAIJ,
2477: MatDiagonalScale_SeqAIJ,
2478: MatNorm_SeqAIJ,
2479: /*20*/ 0,
2480: MatAssemblyEnd_SeqAIJ,
2481: MatCompress_SeqAIJ,
2482: MatSetOption_SeqAIJ,
2483: MatZeroEntries_SeqAIJ,
2484: /*25*/ MatZeroRows_SeqAIJ,
2485: 0,
2486: 0,
2487: 0,
2488: 0,
2489: /*30*/ MatSetUpPreallocation_SeqAIJ,
2490: 0,
2491: 0,
2492: MatGetArray_SeqAIJ,
2493: MatRestoreArray_SeqAIJ,
2494: /*35*/ MatDuplicate_SeqAIJ,
2495: 0,
2496: 0,
2497: MatILUFactor_SeqAIJ,
2498: 0,
2499: /*40*/ MatAXPY_SeqAIJ,
2500: MatGetSubMatrices_SeqAIJ,
2501: MatIncreaseOverlap_SeqAIJ,
2502: MatGetValues_SeqAIJ,
2503: MatCopy_SeqAIJ,
2504: /*45*/ MatGetRowMax_SeqAIJ,
2505: MatScale_SeqAIJ,
2506: 0,
2507: MatDiagonalSet_SeqAIJ,
2508: MatILUDTFactor_SeqAIJ,
2509: /*50*/ MatSetBlockSize_SeqAIJ,
2510: MatGetRowIJ_SeqAIJ,
2511: MatRestoreRowIJ_SeqAIJ,
2512: MatGetColumnIJ_SeqAIJ,
2513: MatRestoreColumnIJ_SeqAIJ,
2514: /*55*/ MatFDColoringCreate_SeqAIJ,
2515: 0,
2516: 0,
2517: MatPermute_SeqAIJ,
2518: 0,
2519: /*60*/ 0,
2520: MatDestroy_SeqAIJ,
2521: MatView_SeqAIJ,
2522: 0,
2523: 0,
2524: /*65*/ 0,
2525: 0,
2526: 0,
2527: 0,
2528: 0,
2529: /*70*/ MatGetRowMaxAbs_SeqAIJ,
2530: MatGetRowMinAbs_SeqAIJ,
2531: 0,
2532: MatSetColoring_SeqAIJ,
2533: #if defined(PETSC_HAVE_ADIC)
2534: MatSetValuesAdic_SeqAIJ,
2535: #else
2536: 0,
2537: #endif
2538: /*75*/ MatSetValuesAdifor_SeqAIJ,
2539: 0,
2540: 0,
2541: 0,
2542: 0,
2543: /*80*/ 0,
2544: 0,
2545: 0,
2546: 0,
2547: MatLoad_SeqAIJ,
2548: /*85*/ MatIsSymmetric_SeqAIJ,
2549: MatIsHermitian_SeqAIJ,
2550: 0,
2551: 0,
2552: 0,
2553: /*90*/ MatMatMult_SeqAIJ_SeqAIJ,
2554: MatMatMultSymbolic_SeqAIJ_SeqAIJ,
2555: MatMatMultNumeric_SeqAIJ_SeqAIJ,
2556: MatPtAP_Basic,
2557: MatPtAPSymbolic_SeqAIJ,
2558: /*95*/ MatPtAPNumeric_SeqAIJ,
2559: MatMatMultTranspose_SeqAIJ_SeqAIJ,
2560: MatMatMultTransposeSymbolic_SeqAIJ_SeqAIJ,
2561: MatMatMultTransposeNumeric_SeqAIJ_SeqAIJ,
2562: MatPtAPSymbolic_SeqAIJ_SeqAIJ,
2563: /*100*/MatPtAPNumeric_SeqAIJ_SeqAIJ,
2564: 0,
2565: 0,
2566: MatConjugate_SeqAIJ,
2567: 0,
2568: /*105*/MatSetValuesRow_SeqAIJ,
2569: MatRealPart_SeqAIJ,
2570: MatImaginaryPart_SeqAIJ,
2571: 0,
2572: 0,
2573: /*110*/0,
2574: 0,
2575: MatGetRowMin_SeqAIJ,
2576: 0,
2577: MatMissingDiagonal_SeqAIJ
2578: };
2583: PetscErrorCode MatSeqAIJSetColumnIndices_SeqAIJ(Mat mat,PetscInt *indices)
2584: {
2585: Mat_SeqAIJ *aij = (Mat_SeqAIJ *)mat->data;
2586: PetscInt i,nz,n;
2590: nz = aij->maxnz;
2591: n = mat->rmap->n;
2592: for (i=0; i<nz; i++) {
2593: aij->j[i] = indices[i];
2594: }
2595: aij->nz = nz;
2596: for (i=0; i<n; i++) {
2597: aij->ilen[i] = aij->imax[i];
2598: }
2600: return(0);
2601: }
2606: /*@
2607: MatSeqAIJSetColumnIndices - Set the column indices for all the rows
2608: in the matrix.
2610: Input Parameters:
2611: + mat - the SeqAIJ matrix
2612: - indices - the column indices
2614: Level: advanced
2616: Notes:
2617: This can be called if you have precomputed the nonzero structure of the
2618: matrix and want to provide it to the matrix object to improve the performance
2619: of the MatSetValues() operation.
2621: You MUST have set the correct numbers of nonzeros per row in the call to
2622: MatCreateSeqAIJ(), and the columns indices MUST be sorted.
2624: MUST be called before any calls to MatSetValues();
2626: The indices should start with zero, not one.
2628: @*/
2629: PetscErrorCode MatSeqAIJSetColumnIndices(Mat mat,PetscInt *indices)
2630: {
2631: PetscErrorCode ierr,(*f)(Mat,PetscInt *);
2636: PetscObjectQueryFunction((PetscObject)mat,"MatSeqAIJSetColumnIndices_C",(void (**)(void))&f);
2637: if (f) {
2638: (*f)(mat,indices);
2639: } else {
2640: SETERRQ(PETSC_ERR_SUP,"Wrong type of matrix to set column indices");
2641: }
2642: return(0);
2643: }
2645: /* ----------------------------------------------------------------------------------------*/
2650: PetscErrorCode MatStoreValues_SeqAIJ(Mat mat)
2651: {
2652: Mat_SeqAIJ *aij = (Mat_SeqAIJ *)mat->data;
2654: size_t nz = aij->i[mat->rmap->n];
2657: if (aij->nonew != 1) {
2658: SETERRQ(PETSC_ERR_ORDER,"Must call MatSetOption(A,MAT_NEW_NONZERO_LOCATIONS,PETSC_FALSE);first");
2659: }
2661: /* allocate space for values if not already there */
2662: if (!aij->saved_values) {
2663: PetscMalloc((nz+1)*sizeof(PetscScalar),&aij->saved_values);
2664: PetscLogObjectMemory(mat,(nz+1)*sizeof(PetscScalar));
2665: }
2667: /* copy values over */
2668: PetscMemcpy(aij->saved_values,aij->a,nz*sizeof(PetscScalar));
2669: return(0);
2670: }
2675: /*@
2676: MatStoreValues - Stashes a copy of the matrix values; this allows, for
2677: example, reuse of the linear part of a Jacobian, while recomputing the
2678: nonlinear portion.
2680: Collect on Mat
2682: Input Parameters:
2683: . mat - the matrix (currently only AIJ matrices support this option)
2685: Level: advanced
2687: Common Usage, with SNESSolve():
2688: $ Create Jacobian matrix
2689: $ Set linear terms into matrix
2690: $ Apply boundary conditions to matrix, at this time matrix must have
2691: $ final nonzero structure (i.e. setting the nonlinear terms and applying
2692: $ boundary conditions again will not change the nonzero structure
2693: $ MatSetOption(mat,MAT_NEW_NONZERO_LOCATIONS,PETSC_FALSE);
2694: $ MatStoreValues(mat);
2695: $ Call SNESSetJacobian() with matrix
2696: $ In your Jacobian routine
2697: $ MatRetrieveValues(mat);
2698: $ Set nonlinear terms in matrix
2699:
2700: Common Usage without SNESSolve(), i.e. when you handle nonlinear solve yourself:
2701: $ // build linear portion of Jacobian
2702: $ MatSetOption(mat,MAT_NEW_NONZERO_LOCATIONS,PETSC_FALSE);
2703: $ MatStoreValues(mat);
2704: $ loop over nonlinear iterations
2705: $ MatRetrieveValues(mat);
2706: $ // call MatSetValues(mat,...) to set nonliner portion of Jacobian
2707: $ // call MatAssemblyBegin/End() on matrix
2708: $ Solve linear system with Jacobian
2709: $ endloop
2711: Notes:
2712: Matrix must already be assemblied before calling this routine
2713: Must set the matrix option MatSetOption(mat,MAT_NEW_NONZERO_LOCATIONS,PETSC_FALSE); before
2714: calling this routine.
2716: When this is called multiple times it overwrites the previous set of stored values
2717: and does not allocated additional space.
2719: .seealso: MatRetrieveValues()
2721: @*/
2722: PetscErrorCode MatStoreValues(Mat mat)
2723: {
2724: PetscErrorCode ierr,(*f)(Mat);
2728: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2729: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2731: PetscObjectQueryFunction((PetscObject)mat,"MatStoreValues_C",(void (**)(void))&f);
2732: if (f) {
2733: (*f)(mat);
2734: } else {
2735: SETERRQ(PETSC_ERR_SUP,"Wrong type of matrix to store values");
2736: }
2737: return(0);
2738: }
2743: PetscErrorCode MatRetrieveValues_SeqAIJ(Mat mat)
2744: {
2745: Mat_SeqAIJ *aij = (Mat_SeqAIJ *)mat->data;
2747: PetscInt nz = aij->i[mat->rmap->n];
2750: if (aij->nonew != 1) {
2751: SETERRQ(PETSC_ERR_ORDER,"Must call MatSetOption(A,MAT_NEW_NONZERO_LOCATIONS,PETSC_FALSE);first");
2752: }
2753: if (!aij->saved_values) {
2754: SETERRQ(PETSC_ERR_ORDER,"Must call MatStoreValues(A);first");
2755: }
2756: /* copy values over */
2757: PetscMemcpy(aij->a,aij->saved_values,nz*sizeof(PetscScalar));
2758: return(0);
2759: }
2764: /*@
2765: MatRetrieveValues - Retrieves the copy of the matrix values; this allows, for
2766: example, reuse of the linear part of a Jacobian, while recomputing the
2767: nonlinear portion.
2769: Collect on Mat
2771: Input Parameters:
2772: . mat - the matrix (currently on AIJ matrices support this option)
2774: Level: advanced
2776: .seealso: MatStoreValues()
2778: @*/
2779: PetscErrorCode MatRetrieveValues(Mat mat)
2780: {
2781: PetscErrorCode ierr,(*f)(Mat);
2785: if (!mat->assembled) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for unassembled matrix");
2786: if (mat->factor) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Not for factored matrix");
2788: PetscObjectQueryFunction((PetscObject)mat,"MatRetrieveValues_C",(void (**)(void))&f);
2789: if (f) {
2790: (*f)(mat);
2791: } else {
2792: SETERRQ(PETSC_ERR_SUP,"Wrong type of matrix to retrieve values");
2793: }
2794: return(0);
2795: }
2798: /* --------------------------------------------------------------------------------*/
2801: /*@C
2802: MatCreateSeqAIJ - Creates a sparse matrix in AIJ (compressed row) format
2803: (the default parallel PETSc format). For good matrix assembly performance
2804: the user should preallocate the matrix storage by setting the parameter nz
2805: (or the array nnz). By setting these parameters accurately, performance
2806: during matrix assembly can be increased by more than a factor of 50.
2808: Collective on MPI_Comm
2810: Input Parameters:
2811: + comm - MPI communicator, set to PETSC_COMM_SELF
2812: . m - number of rows
2813: . n - number of columns
2814: . nz - number of nonzeros per row (same for all rows)
2815: - nnz - array containing the number of nonzeros in the various rows
2816: (possibly different for each row) or PETSC_NULL
2818: Output Parameter:
2819: . A - the matrix
2821: It is recommended that one use the MatCreate(), MatSetType() and/or MatSetFromOptions(),
2822: MatXXXXSetPreallocation() paradgm instead of this routine directly. This is definitely
2823: true if you plan to use the external direct solvers such as SuperLU, MUMPS or Spooles.
2824: [MatXXXXSetPreallocation() is, for example, MatSeqAIJSetPreallocation]
2826: Notes:
2827: If nnz is given then nz is ignored
2829: The AIJ format (also called the Yale sparse matrix format or
2830: compressed row storage), is fully compatible with standard Fortran 77
2831: storage. That is, the stored row and column indices can begin at
2832: either one (as in Fortran) or zero. See the users' manual for details.
2834: Specify the preallocated storage with either nz or nnz (not both).
2835: Set nz=PETSC_DEFAULT and nnz=PETSC_NULL for PETSc to control dynamic memory
2836: allocation. For large problems you MUST preallocate memory or you
2837: will get TERRIBLE performance, see the users' manual chapter on matrices.
2839: By default, this format uses inodes (identical nodes) when possible, to
2840: improve numerical efficiency of matrix-vector products and solves. We
2841: search for consecutive rows with the same nonzero structure, thereby
2842: reusing matrix information to achieve increased efficiency.
2844: Options Database Keys:
2845: + -mat_no_inode - Do not use inodes
2846: . -mat_inode_limit <limit> - Sets inode limit (max limit=5)
2847: - -mat_aij_oneindex - Internally use indexing starting at 1
2848: rather than 0. Note that when calling MatSetValues(),
2849: the user still MUST index entries starting at 0!
2851: Level: intermediate
2853: .seealso: MatCreate(), MatCreateMPIAIJ(), MatSetValues(), MatSeqAIJSetColumnIndices(), MatCreateSeqAIJWithArrays()
2855: @*/
2856: PetscErrorCode MatCreateSeqAIJ(MPI_Comm comm,PetscInt m,PetscInt n,PetscInt nz,const PetscInt nnz[],Mat *A)
2857: {
2861: MatCreate(comm,A);
2862: MatSetSizes(*A,m,n,m,n);
2863: MatSetType(*A,MATSEQAIJ);
2864: MatSeqAIJSetPreallocation_SeqAIJ(*A,nz,(PetscInt*)nnz);
2865: return(0);
2866: }
2870: /*@C
2871: MatSeqAIJSetPreallocation - For good matrix assembly performance
2872: the user should preallocate the matrix storage by setting the parameter nz
2873: (or the array nnz). By setting these parameters accurately, performance
2874: during matrix assembly can be increased by more than a factor of 50.
2876: Collective on MPI_Comm
2878: Input Parameters:
2879: + B - The matrix-free
2880: . nz - number of nonzeros per row (same for all rows)
2881: - nnz - array containing the number of nonzeros in the various rows
2882: (possibly different for each row) or PETSC_NULL
2884: Notes:
2885: If nnz is given then nz is ignored
2887: The AIJ format (also called the Yale sparse matrix format or
2888: compressed row storage), is fully compatible with standard Fortran 77
2889: storage. That is, the stored row and column indices can begin at
2890: either one (as in Fortran) or zero. See the users' manual for details.
2892: Specify the preallocated storage with either nz or nnz (not both).
2893: Set nz=PETSC_DEFAULT and nnz=PETSC_NULL for PETSc to control dynamic memory
2894: allocation. For large problems you MUST preallocate memory or you
2895: will get TERRIBLE performance, see the users' manual chapter on matrices.
2897: You can call MatGetInfo() to get information on how effective the preallocation was;
2898: for example the fields mallocs,nz_allocated,nz_used,nz_unneeded;
2899: You can also run with the option -info and look for messages with the string
2900: malloc in them to see if additional memory allocation was needed.
2902: Developers: Use nz of MAT_SKIP_ALLOCATION to not allocate any space for the matrix
2903: entries or columns indices
2905: By default, this format uses inodes (identical nodes) when possible, to
2906: improve numerical efficiency of matrix-vector products and solves. We
2907: search for consecutive rows with the same nonzero structure, thereby
2908: reusing matrix information to achieve increased efficiency.
2910: Options Database Keys:
2911: + -mat_no_inode - Do not use inodes
2912: . -mat_inode_limit <limit> - Sets inode limit (max limit=5)
2913: - -mat_aij_oneindex - Internally use indexing starting at 1
2914: rather than 0. Note that when calling MatSetValues(),
2915: the user still MUST index entries starting at 0!
2917: Level: intermediate
2919: .seealso: MatCreate(), MatCreateMPIAIJ(), MatSetValues(), MatSeqAIJSetColumnIndices(), MatCreateSeqAIJWithArrays(), MatGetInfo()
2921: @*/
2922: PetscErrorCode MatSeqAIJSetPreallocation(Mat B,PetscInt nz,const PetscInt nnz[])
2923: {
2924: PetscErrorCode ierr,(*f)(Mat,PetscInt,const PetscInt[]);
2927: PetscObjectQueryFunction((PetscObject)B,"MatSeqAIJSetPreallocation_C",(void (**)(void))&f);
2928: if (f) {
2929: (*f)(B,nz,nnz);
2930: }
2931: return(0);
2932: }
2937: PetscErrorCode MatSeqAIJSetPreallocation_SeqAIJ(Mat B,PetscInt nz,PetscInt *nnz)
2938: {
2939: Mat_SeqAIJ *b;
2940: PetscTruth skipallocation = PETSC_FALSE;
2942: PetscInt i;
2945:
2946: if (nz == MAT_SKIP_ALLOCATION) {
2947: skipallocation = PETSC_TRUE;
2948: nz = 0;
2949: }
2951: PetscMapSetBlockSize(B->rmap,1);
2952: PetscMapSetBlockSize(B->cmap,1);
2953: PetscMapSetUp(B->rmap);
2954: PetscMapSetUp(B->cmap);
2956: if (nz == PETSC_DEFAULT || nz == PETSC_DECIDE) nz = 5;
2957: if (nz < 0) SETERRQ1(PETSC_ERR_ARG_OUTOFRANGE,"nz cannot be less than 0: value %d",nz);
2958: if (nnz) {
2959: for (i=0; i<B->rmap->n; i++) {
2960: if (nnz[i] < 0) SETERRQ2(PETSC_ERR_ARG_OUTOFRANGE,"nnz cannot be less than 0: local row %d value %d",i,nnz[i]);
2961: if (nnz[i] > B->cmap->n) SETERRQ3(PETSC_ERR_ARG_OUTOFRANGE,"nnz cannot be greater than row length: local row %d value %d rowlength %d",i,nnz[i],B->cmap->n);
2962: }
2963: }
2965: B->preallocated = PETSC_TRUE;
2966: b = (Mat_SeqAIJ*)B->data;
2968: if (!skipallocation) {
2969: if (!b->imax) {
2970: PetscMalloc2(B->rmap->n,PetscInt,&b->imax,B->rmap->n,PetscInt,&b->ilen);
2971: PetscLogObjectMemory(B,2*B->rmap->n*sizeof(PetscInt));
2972: }
2973: if (!nnz) {
2974: if (nz == PETSC_DEFAULT || nz == PETSC_DECIDE) nz = 10;
2975: else if (nz <= 0) nz = 1;
2976: for (i=0; i<B->rmap->n; i++) b->imax[i] = nz;
2977: nz = nz*B->rmap->n;
2978: } else {
2979: nz = 0;
2980: for (i=0; i<B->rmap->n; i++) {b->imax[i] = nnz[i]; nz += nnz[i];}
2981: }
2982: /* b->ilen will count nonzeros in each row so far. */
2983: for (i=0; i<B->rmap->n; i++) { b->ilen[i] = 0; }
2985: /* allocate the matrix space */
2986: MatSeqXAIJFreeAIJ(B,&b->a,&b->j,&b->i);
2987: PetscMalloc3(nz,PetscScalar,&b->a,nz,PetscInt,&b->j,B->rmap->n+1,PetscInt,&b->i);
2988: PetscLogObjectMemory(B,(B->rmap->n+1)*sizeof(PetscInt)+nz*(sizeof(PetscScalar)+sizeof(PetscInt)));
2989: b->i[0] = 0;
2990: for (i=1; i<B->rmap->n+1; i++) {
2991: b->i[i] = b->i[i-1] + b->imax[i-1];
2992: }
2993: b->singlemalloc = PETSC_TRUE;
2994: b->free_a = PETSC_TRUE;
2995: b->free_ij = PETSC_TRUE;
2996: } else {
2997: b->free_a = PETSC_FALSE;
2998: b->free_ij = PETSC_FALSE;
2999: }
3001: b->nz = 0;
3002: b->maxnz = nz;
3003: B->info.nz_unneeded = (double)b->maxnz;
3004: return(0);
3005: }
3008: #undef __FUNCT__
3010: /*@
3011: MatSeqAIJSetPreallocationCSR - Allocates memory for a sparse sequential matrix in AIJ format.
3013: Input Parameters:
3014: + B - the matrix
3015: . i - the indices into j for the start of each row (starts with zero)
3016: . j - the column indices for each row (starts with zero) these must be sorted for each row
3017: - v - optional values in the matrix
3019: Contributed by: Lisandro Dalchin
3021: Level: developer
3023: The i,j,v values are COPIED with this routine; to avoid the copy use MatCreateSeqAIJWithArrays()
3025: .keywords: matrix, aij, compressed row, sparse, sequential
3027: .seealso: MatCreate(), MatCreateSeqAIJ(), MatSetValues(), MatSeqAIJSetPreallocation(), MatCreateSeqAIJ(), SeqAIJ
3028: @*/
3029: PetscErrorCode MatSeqAIJSetPreallocationCSR(Mat B,const PetscInt i[],const PetscInt j[],const PetscScalar v[])
3030: {
3031: PetscErrorCode (*f)(Mat,const PetscInt[],const PetscInt[],const PetscScalar[]);
3036: PetscObjectQueryFunction((PetscObject)B,"MatSeqAIJSetPreallocationCSR_C",(void (**)(void))&f);
3037: if (f) {
3038: (*f)(B,i,j,v);
3039: }
3040: return(0);
3041: }
3044: #undef __FUNCT__
3046: PetscErrorCode MatSeqAIJSetPreallocationCSR_SeqAIJ(Mat B,const PetscInt Ii[],const PetscInt J[],const PetscScalar v[])
3047: {
3048: PetscInt i;
3049: PetscInt m,n;
3050: PetscInt nz;
3051: PetscInt *nnz, nz_max = 0;
3052: PetscScalar *values;
3056: MatGetSize(B, &m, &n);
3058: if (Ii[0]) {
3059: SETERRQ1(PETSC_ERR_ARG_OUTOFRANGE, "Ii[0] must be 0 it is %D", Ii[0]);
3060: }
3061: PetscMalloc((m+1) * sizeof(PetscInt), &nnz);
3062: for(i = 0; i < m; i++) {
3063: nz = Ii[i+1]- Ii[i];
3064: nz_max = PetscMax(nz_max, nz);
3065: if (nz < 0) {
3066: SETERRQ2(PETSC_ERR_ARG_OUTOFRANGE, "Local row %D has a negative number of columns %D", i, nnz);
3067: }
3068: nnz[i] = nz;
3069: }
3070: MatSeqAIJSetPreallocation(B, 0, nnz);
3071: PetscFree(nnz);
3073: if (v) {
3074: values = (PetscScalar*) v;
3075: } else {
3076: PetscMalloc((nz_max+1)*sizeof(PetscScalar), &values);
3077: PetscMemzero(values, nz_max*sizeof(PetscScalar));
3078: }
3080: for(i = 0; i < m; i++) {
3081: nz = Ii[i+1] - Ii[i];
3082: MatSetValues_SeqAIJ(B, 1, &i, nz, J+Ii[i], values + (v ? Ii[i] : 0), INSERT_VALUES);
3083: }
3085: MatAssemblyBegin(B,MAT_FINAL_ASSEMBLY);
3086: MatAssemblyEnd(B,MAT_FINAL_ASSEMBLY);
3088: if (!v) {
3089: PetscFree(values);
3090: }
3091: return(0);
3092: }
3095: #include ../src/mat/impls/dense/seq/dense.h
3096: #include ../src/inline/axpy.h
3100: /*
3101: Computes (B'*A')' since computing B*A directly is untenable
3103: n p p
3104: ( ) ( ) ( )
3105: m ( A ) * n ( B ) = m ( C )
3106: ( ) ( ) ( )
3108: */
3109: PetscErrorCode MatMatMultNumeric_SeqDense_SeqAIJ(Mat A,Mat B,Mat C)
3110: {
3111: PetscErrorCode ierr;
3112: Mat_SeqDense *sub_a = (Mat_SeqDense*)A->data;
3113: Mat_SeqAIJ *sub_b = (Mat_SeqAIJ*)B->data;
3114: Mat_SeqDense *sub_c = (Mat_SeqDense*)C->data;
3115: PetscInt i,n,m,q,p;
3116: const PetscInt *ii,*idx;
3117: const PetscScalar *b,*a,*a_q;
3118: PetscScalar *c,*c_q;
3121: m = A->rmap->n;
3122: n = A->cmap->n;
3123: p = B->cmap->n;
3124: a = sub_a->v;
3125: b = sub_b->a;
3126: c = sub_c->v;
3127: PetscMemzero(c,m*p*sizeof(PetscScalar));
3129: ii = sub_b->i;
3130: idx = sub_b->j;
3131: for (i=0; i<n; i++) {
3132: q = ii[i+1] - ii[i];
3133: while (q-->0) {
3134: c_q = c + m*(*idx);
3135: a_q = a + m*i;
3136: APXY(c_q,*b,a_q,m);
3137: idx++;
3138: b++;
3139: }
3140: }
3141: return(0);
3142: }
3146: PetscErrorCode MatMatMultSymbolic_SeqDense_SeqAIJ(Mat A,Mat B,PetscReal fill,Mat *C)
3147: {
3149: PetscInt m=A->rmap->n,n=B->cmap->n;
3150: Mat Cmat;
3153: if (A->cmap->n != B->rmap->n) SETERRQ2(PETSC_ERR_ARG_SIZ,"A->cmap->n %d != B->rmap->n %d\n",A->cmap->n,B->rmap->n);
3154: MatCreate(((PetscObject)A)->comm,&Cmat);
3155: MatSetSizes(Cmat,m,n,m,n);
3156: MatSetType(Cmat,MATSEQDENSE);
3157: MatSeqDenseSetPreallocation(Cmat,PETSC_NULL);
3158: Cmat->assembled = PETSC_TRUE;
3159: *C = Cmat;
3160: return(0);
3161: }
3163: /* ----------------------------------------------------------------*/
3166: PetscErrorCode MatMatMult_SeqDense_SeqAIJ(Mat A,Mat B,MatReuse scall,PetscReal fill,Mat *C)
3167: {
3171: if (scall == MAT_INITIAL_MATRIX){
3172: MatMatMultSymbolic_SeqDense_SeqAIJ(A,B,fill,C);
3173: }
3174: MatMatMultNumeric_SeqDense_SeqAIJ(A,B,*C);
3175: return(0);
3176: }
3179: /*MC
3180: MATSEQAIJ - MATSEQAIJ = "seqaij" - A matrix type to be used for sequential sparse matrices,
3181: based on compressed sparse row format.
3183: Options Database Keys:
3184: . -mat_type seqaij - sets the matrix type to "seqaij" during a call to MatSetFromOptions()
3186: Level: beginner
3188: .seealso: MatCreateSeqAIJ(), MatSetFromOptions(), MatSetType(), MatCreate(), MatType
3189: M*/
3192: #if defined(PETSC_HAVE_PASTIX)
3194: #endif
3195: #if defined(PETSC_HAVE_ESSL)
3197: #endif
3201: #if defined(PETSC_HAVE_MUMPS)
3203: #endif
3204: #if defined(PETSC_HAVE_SUPERLU)
3206: #endif
3207: #if defined(PETSC_HAVE_SUPERLU_DIST)
3209: #endif
3210: #if defined(PETSC_HAVE_SPOOLES)
3212: #endif
3213: #if defined(PETSC_HAVE_UMFPACK)
3215: #endif
3216: #if defined(PETSC_HAVE_LUSOL)
3218: #endif
3225: PetscErrorCode MatCreate_SeqAIJ(Mat B)
3226: {
3227: Mat_SeqAIJ *b;
3229: PetscMPIInt size;
3232: MPI_Comm_size(((PetscObject)B)->comm,&size);
3233: if (size > 1) SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,"Comm must be of size 1");
3235: PetscNewLog(B,Mat_SeqAIJ,&b);
3236: B->data = (void*)b;
3237: PetscMemcpy(B->ops,&MatOps_Values,sizeof(struct _MatOps));
3238: B->mapping = 0;
3239: b->row = 0;
3240: b->col = 0;
3241: b->icol = 0;
3242: b->reallocs = 0;
3243: b->ignorezeroentries = PETSC_FALSE;
3244: b->roworiented = PETSC_TRUE;
3245: b->nonew = 0;
3246: b->diag = 0;
3247: b->solve_work = 0;
3248: B->spptr = 0;
3249: b->saved_values = 0;
3250: b->idiag = 0;
3251: b->mdiag = 0;
3252: b->ssor_work = 0;
3253: b->omega = 1.0;
3254: b->fshift = 0.0;
3255: b->idiagvalid = PETSC_FALSE;
3256: b->keepzeroedrows = PETSC_FALSE;
3257: b->xtoy = 0;
3258: b->XtoY = 0;
3259: b->compressedrow.use = PETSC_FALSE;
3260: b->compressedrow.nrows = B->rmap->n;
3261: b->compressedrow.i = PETSC_NULL;
3262: b->compressedrow.rindex = PETSC_NULL;
3263: b->compressedrow.checked = PETSC_FALSE;
3264: B->same_nonzero = PETSC_FALSE;
3266: PetscObjectChangeTypeName((PetscObject)B,MATSEQAIJ);
3267: #if defined(PETSC_HAVE_PASTIX)
3268: PetscObjectComposeFunctionDynamic((PetscObject)B,"MatGetFactor_seqaij_pastix_C",
3269: "MatGetFactor_seqaij_pastix",
3270: MatGetFactor_seqaij_pastix);
3271: #endif
3272: #if defined(PETSC_HAVE_ESSL)
3273: PetscObjectComposeFunctionDynamic((PetscObject)B,"MatGetFactor_seqaij_essl_C",
3274: "MatGetFactor_seqaij_essl",
3275: MatGetFactor_seqaij_essl);
3276: #endif
3277: #if defined(PETSC_HAVE_SUPERLU)
3278: PetscObjectComposeFunctionDynamic((PetscObject)B,"MatGetFactor_seqaij_superlu_C",
3279: "MatGetFactor_seqaij_superlu",
3280: MatGetFactor_seqaij_superlu);
3281: #endif
3282: #if defined(PETSC_HAVE_SUPERLU_DIST)
3283: PetscObjectComposeFunctionDynamic((PetscObject)B,"MatGetFactor_seqaij_superlu_dist_C",
3284: "MatGetFactor_seqaij_superlu_dist",
3285: MatGetFactor_seqaij_superlu_dist);
3286: #endif
3287: #if defined(PETSC_HAVE_SPOOLES)
3288: PetscObjectComposeFunctionDynamic((PetscObject)B,"MatGetFactor_seqaij_spooles_C",
3289: "MatGetFactor_seqaij_spooles",
3290: MatGetFactor_seqaij_spooles);
3291: #endif
3292: #if defined(PETSC_HAVE_MUMPS)
3293: PetscObjectComposeFunctionDynamic((PetscObject)B,"MatGetFactor_seqaij_mumps_C",
3294: "MatGetFactor_seqaij_mumps",
3295: MatGetFactor_seqaij_mumps);
3296: #endif
3297: #if defined(PETSC_HAVE_UMFPACK)
3298: PetscObjectComposeFunctionDynamic((PetscObject)B,"MatGetFactor_seqaij_umfpack_C",
3299: "MatGetFactor_seqaij_umfpack",
3300: MatGetFactor_seqaij_umfpack);
3301: #endif
3302: #if defined(PETSC_HAVE_LUSOL)
3303: PetscObjectComposeFunctionDynamic((PetscObject)B,"MatGetFactor_seqaij_lusol_C",
3304: "MatGetFactor_seqaij_lusol",
3305: MatGetFactor_seqaij_lusol);
3306: #endif
3307: PetscObjectComposeFunctionDynamic((PetscObject)B,"MatGetFactor_seqaij_petsc_C",
3308: "MatGetFactor_seqaij_petsc",
3309: MatGetFactor_seqaij_petsc);
3310: PetscObjectComposeFunctionDynamic((PetscObject)B,"MatGetFactorAvailable_seqaij_petsc_C",
3311: "MatGetFactorAvailable_seqaij_petsc",
3312: MatGetFactorAvailable_seqaij_petsc);
3313: PetscObjectComposeFunctionDynamic((PetscObject)B,"MatSeqAIJSetColumnIndices_C",
3314: "MatSeqAIJSetColumnIndices_SeqAIJ",
3315: MatSeqAIJSetColumnIndices_SeqAIJ);
3316: PetscObjectComposeFunctionDynamic((PetscObject)B,"MatStoreValues_C",
3317: "MatStoreValues_SeqAIJ",
3318: MatStoreValues_SeqAIJ);
3319: PetscObjectComposeFunctionDynamic((PetscObject)B,"MatRetrieveValues_C",
3320: "MatRetrieveValues_SeqAIJ",
3321: MatRetrieveValues_SeqAIJ);
3322: PetscObjectComposeFunctionDynamic((PetscObject)B,"MatConvert_seqaij_seqsbaij_C",
3323: "MatConvert_SeqAIJ_SeqSBAIJ",
3324: MatConvert_SeqAIJ_SeqSBAIJ);
3325: PetscObjectComposeFunctionDynamic((PetscObject)B,"MatConvert_seqaij_seqbaij_C",
3326: "MatConvert_SeqAIJ_SeqBAIJ",
3327: MatConvert_SeqAIJ_SeqBAIJ);
3328: PetscObjectComposeFunctionDynamic((PetscObject)B,"MatConvert_seqaij_seqcsrperm_C",
3329: "MatConvert_SeqAIJ_SeqCSRPERM",
3330: MatConvert_SeqAIJ_SeqCSRPERM);
3331: PetscObjectComposeFunctionDynamic((PetscObject)B,"MatConvert_seqaij_seqcrl_C",
3332: "MatConvert_SeqAIJ_SeqCRL",
3333: MatConvert_SeqAIJ_SeqCRL);
3334: PetscObjectComposeFunctionDynamic((PetscObject)B,"MatIsTranspose_C",
3335: "MatIsTranspose_SeqAIJ",
3336: MatIsTranspose_SeqAIJ);
3337: PetscObjectComposeFunctionDynamic((PetscObject)B,"MatIsHermitianTranspose_C",
3338: "MatIsHermitianTranspose_SeqAIJ",
3339: MatIsTranspose_SeqAIJ);
3340: PetscObjectComposeFunctionDynamic((PetscObject)B,"MatSeqAIJSetPreallocation_C",
3341: "MatSeqAIJSetPreallocation_SeqAIJ",
3342: MatSeqAIJSetPreallocation_SeqAIJ);
3343: PetscObjectComposeFunctionDynamic((PetscObject)B,"MatSeqAIJSetPreallocationCSR_C",
3344: "MatSeqAIJSetPreallocationCSR_SeqAIJ",
3345: MatSeqAIJSetPreallocationCSR_SeqAIJ);
3346: PetscObjectComposeFunctionDynamic((PetscObject)B,"MatReorderForNonzeroDiagonal_C",
3347: "MatReorderForNonzeroDiagonal_SeqAIJ",
3348: MatReorderForNonzeroDiagonal_SeqAIJ);
3349: PetscObjectComposeFunctionDynamic((PetscObject)B,"MatMatMult_seqdense_seqaij_C",
3350: "MatMatMult_SeqDense_SeqAIJ",
3351: MatMatMult_SeqDense_SeqAIJ);
3352: PetscObjectComposeFunctionDynamic((PetscObject)B,"MatMatMultSymbolic_seqdense_seqaij_C",
3353: "MatMatMultSymbolic_SeqDense_SeqAIJ",
3354: MatMatMultSymbolic_SeqDense_SeqAIJ);
3355: PetscObjectComposeFunctionDynamic((PetscObject)B,"MatMatMultNumeric_seqdense_seqaij_C",
3356: "MatMatMultNumeric_SeqDense_SeqAIJ",
3357: MatMatMultNumeric_SeqDense_SeqAIJ);
3358: MatCreate_Inode(B);
3359: PetscObjectChangeTypeName((PetscObject)B,MATSEQAIJ);
3360: return(0);
3361: }
3366: /*
3367: Given a matrix generated with MatGetFactor() duplicates all the information in A into B
3368: */
3369: PetscErrorCode MatDuplicateNoCreate_SeqAIJ(Mat C,Mat A,MatDuplicateOption cpvalues)
3370: {
3371: Mat_SeqAIJ *c,*a = (Mat_SeqAIJ*)A->data;
3373: PetscInt i,m = A->rmap->n;
3376: PetscMemcpy(C->ops,A->ops,sizeof(struct _MatOps));
3377:
3378: c = (Mat_SeqAIJ*)C->data;
3380: C->factor = A->factor;
3382: c->row = 0;
3383: c->col = 0;
3384: c->icol = 0;
3385: c->reallocs = 0;
3387: C->assembled = PETSC_TRUE;
3388:
3389: PetscMapSetBlockSize(C->rmap,1);
3390: PetscMapSetBlockSize(C->cmap,1);
3391: PetscMapSetUp(C->rmap);
3392: PetscMapSetUp(C->cmap);
3394: PetscMalloc2(m,PetscInt,&c->imax,m,PetscInt,&c->ilen);
3395: PetscLogObjectMemory(C, 2*m*sizeof(PetscInt));
3396: for (i=0; i<m; i++) {
3397: c->imax[i] = a->imax[i];
3398: c->ilen[i] = a->ilen[i];
3399: }
3401: /* allocate the matrix space */
3402: PetscMalloc3(a->i[m],PetscScalar,&c->a,a->i[m],PetscInt,&c->j,m+1,PetscInt,&c->i);
3403: PetscLogObjectMemory(C, a->i[m]*(sizeof(PetscScalar)+sizeof(PetscInt))+(m+1)*sizeof(PetscInt));
3404: c->singlemalloc = PETSC_TRUE;
3405: PetscMemcpy(c->i,a->i,(m+1)*sizeof(PetscInt));
3406: if (m > 0) {
3407: PetscMemcpy(c->j,a->j,(a->i[m])*sizeof(PetscInt));
3408: if (cpvalues == MAT_COPY_VALUES) {
3409: PetscMemcpy(c->a,a->a,(a->i[m])*sizeof(PetscScalar));
3410: } else {
3411: PetscMemzero(c->a,(a->i[m])*sizeof(PetscScalar));
3412: }
3413: }
3415: c->ignorezeroentries = a->ignorezeroentries;
3416: c->roworiented = a->roworiented;
3417: c->nonew = a->nonew;
3418: if (a->diag) {
3419: PetscMalloc((m+1)*sizeof(PetscInt),&c->diag);
3420: PetscLogObjectMemory(C,(m+1)*sizeof(PetscInt));
3421: for (i=0; i<m; i++) {
3422: c->diag[i] = a->diag[i];
3423: }
3424: } else c->diag = 0;
3425: c->solve_work = 0;
3426: c->saved_values = 0;
3427: c->idiag = 0;
3428: c->ssor_work = 0;
3429: c->keepzeroedrows = a->keepzeroedrows;
3430: c->free_a = PETSC_TRUE;
3431: c->free_ij = PETSC_TRUE;
3432: c->xtoy = 0;
3433: c->XtoY = 0;
3435: c->nz = a->nz;
3436: c->maxnz = a->maxnz;
3437: C->preallocated = PETSC_TRUE;
3439: c->compressedrow.use = a->compressedrow.use;
3440: c->compressedrow.nrows = a->compressedrow.nrows;
3441: c->compressedrow.checked = a->compressedrow.checked;
3442: if ( a->compressedrow.checked && a->compressedrow.use){
3443: i = a->compressedrow.nrows;
3444: PetscMalloc((2*i+1)*sizeof(PetscInt),&c->compressedrow.i);
3445: c->compressedrow.rindex = c->compressedrow.i + i + 1;
3446: PetscMemcpy(c->compressedrow.i,a->compressedrow.i,(i+1)*sizeof(PetscInt));
3447: PetscMemcpy(c->compressedrow.rindex,a->compressedrow.rindex,i*sizeof(PetscInt));
3448: } else {
3449: c->compressedrow.use = PETSC_FALSE;
3450: c->compressedrow.i = PETSC_NULL;
3451: c->compressedrow.rindex = PETSC_NULL;
3452: }
3453: C->same_nonzero = A->same_nonzero;
3454: MatDuplicate_Inode(A,cpvalues,&C);
3456: PetscFListDuplicate(((PetscObject)A)->qlist,&((PetscObject)C)->qlist);
3457: return(0);
3458: }
3462: PetscErrorCode MatDuplicate_SeqAIJ(Mat A,MatDuplicateOption cpvalues,Mat *B)
3463: {
3465: PetscInt n = A->rmap->n;
3468: MatCreate(((PetscObject)A)->comm,B);
3469: MatSetSizes(*B,n,n,n,n);
3470: MatSetType(*B,MATSEQAIJ);
3471: MatDuplicateNoCreate_SeqAIJ(*B,A,cpvalues);
3472: return(0);
3473: }
3477: PetscErrorCode MatLoad_SeqAIJ(PetscViewer viewer, const MatType type,Mat *A)
3478: {
3479: Mat_SeqAIJ *a;
3480: Mat B;
3482: PetscInt i,sum,nz,header[4],*rowlengths = 0,M,N;
3483: int fd;
3484: PetscMPIInt size;
3485: MPI_Comm comm;
3486:
3488: PetscObjectGetComm((PetscObject)viewer,&comm);
3489: MPI_Comm_size(comm,&size);
3490: if (size > 1) SETERRQ(PETSC_ERR_ARG_SIZ,"view must have one processor");
3491: PetscViewerBinaryGetDescriptor(viewer,&fd);
3492: PetscBinaryRead(fd,header,4,PETSC_INT);
3493: if (header[0] != MAT_FILE_COOKIE) SETERRQ(PETSC_ERR_FILE_UNEXPECTED,"not matrix object in file");
3494: M = header[1]; N = header[2]; nz = header[3];
3496: if (nz < 0) {
3497: SETERRQ(PETSC_ERR_FILE_UNEXPECTED,"Matrix stored in special format on disk,cannot load as SeqAIJ");
3498: }
3500: /* read in row lengths */
3501: PetscMalloc(M*sizeof(PetscInt),&rowlengths);
3502: PetscBinaryRead(fd,rowlengths,M,PETSC_INT);
3504: /* check if sum of rowlengths is same as nz */
3505: for (i=0,sum=0; i< M; i++) sum +=rowlengths[i];
3506: if (sum != nz) SETERRQ2(PETSC_ERR_FILE_READ,"Inconsistant matrix data in file. no-nonzeros = %d, sum-row-lengths = %d\n",nz,sum);
3508: /* create our matrix */
3509: MatCreate(comm,&B);
3510: MatSetSizes(B,PETSC_DECIDE,PETSC_DECIDE,M,N);
3511: MatSetType(B,type);
3512: MatSeqAIJSetPreallocation_SeqAIJ(B,0,rowlengths);
3513: a = (Mat_SeqAIJ*)B->data;
3515: PetscBinaryRead(fd,a->j,nz,PETSC_INT);
3517: /* read in nonzero values */
3518: PetscBinaryRead(fd,a->a,nz,PETSC_SCALAR);
3520: /* set matrix "i" values */
3521: a->i[0] = 0;
3522: for (i=1; i<= M; i++) {
3523: a->i[i] = a->i[i-1] + rowlengths[i-1];
3524: a->ilen[i-1] = rowlengths[i-1];
3525: }
3526: PetscFree(rowlengths);
3528: MatAssemblyBegin(B,MAT_FINAL_ASSEMBLY);
3529: MatAssemblyEnd(B,MAT_FINAL_ASSEMBLY);
3530: *A = B;
3531: return(0);
3532: }
3536: PetscErrorCode MatEqual_SeqAIJ(Mat A,Mat B,PetscTruth* flg)
3537: {
3538: Mat_SeqAIJ *a = (Mat_SeqAIJ *)A->data,*b = (Mat_SeqAIJ *)B->data;
3542: /* If the matrix dimensions are not equal,or no of nonzeros */
3543: if ((A->rmap->n != B->rmap->n) || (A->cmap->n != B->cmap->n) ||(a->nz != b->nz)) {
3544: *flg = PETSC_FALSE;
3545: return(0);
3546: }
3547:
3548: /* if the a->i are the same */
3549: PetscMemcmp(a->i,b->i,(A->rmap->n+1)*sizeof(PetscInt),flg);
3550: if (!*flg) return(0);
3551:
3552: /* if a->j are the same */
3553: PetscMemcmp(a->j,b->j,(a->nz)*sizeof(PetscInt),flg);
3554: if (!*flg) return(0);
3555:
3556: /* if a->a are the same */
3557: PetscMemcmp(a->a,b->a,(a->nz)*sizeof(PetscScalar),flg);
3559: return(0);
3560:
3561: }
3565: /*@
3566: MatCreateSeqAIJWithArrays - Creates an sequential AIJ matrix using matrix elements (in CSR format)
3567: provided by the user.
3569: Collective on MPI_Comm
3571: Input Parameters:
3572: + comm - must be an MPI communicator of size 1
3573: . m - number of rows
3574: . n - number of columns
3575: . i - row indices
3576: . j - column indices
3577: - a - matrix values
3579: Output Parameter:
3580: . mat - the matrix
3582: Level: intermediate
3584: Notes:
3585: The i, j, and a arrays are not copied by this routine, the user must free these arrays
3586: once the matrix is destroyed
3588: You cannot set new nonzero locations into this matrix, that will generate an error.
3590: The i and j indices are 0 based
3592: The format which is used for the sparse matrix input, is equivalent to a
3593: row-major ordering.. i.e for the following matrix, the input data expected is
3594: as shown:
3596: 1 0 0
3597: 2 0 3
3598: 4 5 6
3600: i = {0,1,3,6} [size = nrow+1 = 3+1]
3601: j = {0,0,2,0,1,2} [size = nz = 6]; values must be sorted for each row
3602: v = {1,2,3,4,5,6} [size = nz = 6]
3604:
3605: .seealso: MatCreate(), MatCreateMPIAIJ(), MatCreateSeqAIJ(), MatCreateMPIAIJWithArrays(), MatMPIAIJSetPreallocationCSR()
3607: @*/
3608: PetscErrorCode MatCreateSeqAIJWithArrays(MPI_Comm comm,PetscInt m,PetscInt n,PetscInt* i,PetscInt*j,PetscScalar *a,Mat *mat)
3609: {
3611: PetscInt ii;
3612: Mat_SeqAIJ *aij;
3613: #if defined(PETSC_USE_DEBUG)
3614: PetscInt jj;
3615: #endif
3618: if (i[0]) {
3619: SETERRQ(PETSC_ERR_ARG_OUTOFRANGE,"i (row indices) must start with 0");
3620: }
3621: MatCreate(comm,mat);
3622: MatSetSizes(*mat,m,n,m,n);
3623: MatSetType(*mat,MATSEQAIJ);
3624: MatSeqAIJSetPreallocation_SeqAIJ(*mat,MAT_SKIP_ALLOCATION,0);
3625: aij = (Mat_SeqAIJ*)(*mat)->data;
3626: PetscMalloc2(m,PetscInt,&aij->imax,m,PetscInt,&aij->ilen);
3628: aij->i = i;
3629: aij->j = j;
3630: aij->a = a;
3631: aij->singlemalloc = PETSC_FALSE;
3632: aij->nonew = -1; /*this indicates that inserting a new value in the matrix that generates a new nonzero is an error*/
3633: aij->free_a = PETSC_FALSE;
3634: aij->free_ij = PETSC_FALSE;
3636: for (ii=0; ii<m; ii++) {
3637: aij->ilen[ii] = aij->imax[ii] = i[ii+1] - i[ii];
3638: #if defined(PETSC_USE_DEBUG)
3639: if (i[ii+1] - i[ii] < 0) SETERRQ2(PETSC_ERR_ARG_OUTOFRANGE,"Negative row length in i (row indices) row = %d length = %d",ii,i[ii+1] - i[ii]);
3640: for (jj=i[ii]+1; jj<i[ii+1]; jj++) {
3641: if (j[jj] < j[jj-1]) SETERRQ3(PETSC_ERR_ARG_OUTOFRANGE,"Column entry number %D (actual colum %D) in row %D is not sorted",jj-i[ii],j[jj],ii);
3642: if (j[jj] == j[jj]-1) SETERRQ3(PETSC_ERR_ARG_OUTOFRANGE,"Column entry number %D (actual colum %D) in row %D is identical to previous entry",jj-i[ii],j[jj],ii);
3643: }
3644: #endif
3645: }
3646: #if defined(PETSC_USE_DEBUG)
3647: for (ii=0; ii<aij->i[m]; ii++) {
3648: if (j[ii] < 0) SETERRQ2(PETSC_ERR_ARG_OUTOFRANGE,"Negative column index at location = %d index = %d",ii,j[ii]);
3649: if (j[ii] > n - 1) SETERRQ2(PETSC_ERR_ARG_OUTOFRANGE,"Column index to large at location = %d index = %d",ii,j[ii]);
3650: }
3651: #endif
3653: MatAssemblyBegin(*mat,MAT_FINAL_ASSEMBLY);
3654: MatAssemblyEnd(*mat,MAT_FINAL_ASSEMBLY);
3655: return(0);
3656: }
3660: PetscErrorCode MatSetColoring_SeqAIJ(Mat A,ISColoring coloring)
3661: {
3663: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
3666: if (coloring->ctype == IS_COLORING_GLOBAL) {
3667: ISColoringReference(coloring);
3668: a->coloring = coloring;
3669: } else if (coloring->ctype == IS_COLORING_GHOSTED) {
3670: PetscInt i,*larray;
3671: ISColoring ocoloring;
3672: ISColoringValue *colors;
3674: /* set coloring for diagonal portion */
3675: PetscMalloc((A->cmap->n+1)*sizeof(PetscInt),&larray);
3676: for (i=0; i<A->cmap->n; i++) {
3677: larray[i] = i;
3678: }
3679: ISGlobalToLocalMappingApply(A->mapping,IS_GTOLM_MASK,A->cmap->n,larray,PETSC_NULL,larray);
3680: PetscMalloc((A->cmap->n+1)*sizeof(ISColoringValue),&colors);
3681: for (i=0; i<A->cmap->n; i++) {
3682: colors[i] = coloring->colors[larray[i]];
3683: }
3684: PetscFree(larray);
3685: ISColoringCreate(PETSC_COMM_SELF,coloring->n,A->cmap->n,colors,&ocoloring);
3686: a->coloring = ocoloring;
3687: }
3688: return(0);
3689: }
3691: #if defined(PETSC_HAVE_ADIC)
3693: #include "adic/ad_utils.h"
3698: PetscErrorCode MatSetValuesAdic_SeqAIJ(Mat A,void *advalues)
3699: {
3700: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
3701: PetscInt m = A->rmap->n,*ii = a->i,*jj = a->j,nz,i,j,nlen;
3702: PetscScalar *v = a->a,*values = ((PetscScalar*)advalues)+1;
3703: ISColoringValue *color;
3706: if (!a->coloring) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Coloring not set for matrix");
3707: nlen = PetscADGetDerivTypeSize()/sizeof(PetscScalar);
3708: color = a->coloring->colors;
3709: /* loop over rows */
3710: for (i=0; i<m; i++) {
3711: nz = ii[i+1] - ii[i];
3712: /* loop over columns putting computed value into matrix */
3713: for (j=0; j<nz; j++) {
3714: *v++ = values[color[*jj++]];
3715: }
3716: values += nlen; /* jump to next row of derivatives */
3717: }
3718: return(0);
3719: }
3720: #endif
3724: PetscErrorCode MatSetValuesAdifor_SeqAIJ(Mat A,PetscInt nl,void *advalues)
3725: {
3726: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
3727: PetscInt m = A->rmap->n,*ii = a->i,*jj = a->j,nz,i,j;
3728: MatScalar *v = a->a;
3729: PetscScalar *values = (PetscScalar *)advalues;
3730: ISColoringValue *color;
3733: if (!a->coloring) SETERRQ(PETSC_ERR_ARG_WRONGSTATE,"Coloring not set for matrix");
3734: color = a->coloring->colors;
3735: /* loop over rows */
3736: for (i=0; i<m; i++) {
3737: nz = ii[i+1] - ii[i];
3738: /* loop over columns putting computed value into matrix */
3739: for (j=0; j<nz; j++) {
3740: *v++ = values[color[*jj++]];
3741: }
3742: values += nl; /* jump to next row of derivatives */
3743: }
3744: return(0);
3745: }
3747: /*
3748: Special version for direct calls from Fortran
3749: */
3750: #if defined(PETSC_HAVE_FORTRAN_CAPS)
3751: #define matsetvaluesseqaij_ MATSETVALUESSEQAIJ
3752: #elif !defined(PETSC_HAVE_FORTRAN_UNDERSCORE)
3753: #define matsetvaluesseqaij_ matsetvaluesseqaij
3754: #endif
3756: /* Change these macros so can be used in void function */
3757: #undef CHKERRQ
3758: #define CHKERRQ(ierr) CHKERRABORT(((PetscObject)A)->comm,ierr)
3759: #undef SETERRQ2
3760: #define SETERRQ2(ierr,b,c,d) CHKERRABORT(((PetscObject)A)->comm,ierr)
3765: void PETSC_STDCALL matsetvaluesseqaij_(Mat *AA,PetscInt *mm,const PetscInt im[],PetscInt *nn,const PetscInt in[],const PetscScalar v[],InsertMode *isis, PetscErrorCode *_ierr)
3766: {
3767: Mat A = *AA;
3768: PetscInt m = *mm, n = *nn;
3769: InsertMode is = *isis;
3770: Mat_SeqAIJ *a = (Mat_SeqAIJ*)A->data;
3771: PetscInt *rp,k,low,high,t,ii,row,nrow,i,col,l,rmax,N;
3772: PetscInt *imax,*ai,*ailen;
3774: PetscInt *aj,nonew = a->nonew,lastcol = -1;
3775: MatScalar *ap,value,*aa;
3776: PetscTruth ignorezeroentries = a->ignorezeroentries;
3777: PetscTruth roworiented = a->roworiented;
3780: MatPreallocated(A);
3781: imax = a->imax;
3782: ai = a->i;
3783: ailen = a->ilen;
3784: aj = a->j;
3785: aa = a->a;
3787: for (k=0; k<m; k++) { /* loop over added rows */
3788: row = im[k];
3789: if (row < 0) continue;
3790: #if defined(PETSC_USE_DEBUG)
3791: if (row >= A->rmap->n) SETERRABORT(((PetscObject)A)->comm,PETSC_ERR_ARG_OUTOFRANGE,"Row too large");
3792: #endif
3793: rp = aj + ai[row]; ap = aa + ai[row];
3794: rmax = imax[row]; nrow = ailen[row];
3795: low = 0;
3796: high = nrow;
3797: for (l=0; l<n; l++) { /* loop over added columns */
3798: if (in[l] < 0) continue;
3799: #if defined(PETSC_USE_DEBUG)
3800: if (in[l] >= A->cmap->n) SETERRABORT(((PetscObject)A)->comm,PETSC_ERR_ARG_OUTOFRANGE,"Column too large");
3801: #endif
3802: col = in[l];
3803: if (roworiented) {
3804: value = v[l + k*n];
3805: } else {
3806: value = v[k + l*m];
3807: }
3808: if (value == 0.0 && ignorezeroentries && (is == ADD_VALUES)) continue;
3810: if (col <= lastcol) low = 0; else high = nrow;
3811: lastcol = col;
3812: while (high-low > 5) {
3813: t = (low+high)/2;
3814: if (rp[t] > col) high = t;
3815: else low = t;
3816: }
3817: for (i=low; i<high; i++) {
3818: if (rp[i] > col) break;
3819: if (rp[i] == col) {
3820: if (is == ADD_VALUES) ap[i] += value;
3821: else ap[i] = value;
3822: goto noinsert;
3823: }
3824: }
3825: if (value == 0.0 && ignorezeroentries) goto noinsert;
3826: if (nonew == 1) goto noinsert;
3827: if (nonew == -1) SETERRABORT(((PetscObject)A)->comm,PETSC_ERR_ARG_OUTOFRANGE,"Inserting a new nonzero in the matrix");
3828: MatSeqXAIJReallocateAIJ(A,A->rmap->n,1,nrow,row,col,rmax,aa,ai,aj,rp,ap,imax,nonew,MatScalar);
3829: N = nrow++ - 1; a->nz++; high++;
3830: /* shift up all the later entries in this row */
3831: for (ii=N; ii>=i; ii--) {
3832: rp[ii+1] = rp[ii];
3833: ap[ii+1] = ap[ii];
3834: }
3835: rp[i] = col;
3836: ap[i] = value;
3837: noinsert:;
3838: low = i + 1;
3839: }
3840: ailen[row] = nrow;
3841: }
3842: A->same_nonzero = PETSC_FALSE;
3843: PetscFunctionReturnVoid();
3844: }