Actual source code: sbaij.c
petsc-3.8.4 2018-03-24
2: /*
3: Defines the basic matrix operations for the SBAIJ (compressed row)
4: matrix storage format.
5: */
6: #include <../src/mat/impls/baij/seq/baij.h>
7: #include <../src/mat/impls/sbaij/seq/sbaij.h>
8: #include <petscblaslapack.h>
10: #include <../src/mat/impls/sbaij/seq/relax.h>
11: #define USESHORT
12: #include <../src/mat/impls/sbaij/seq/relax.h>
14: #if defined(PETSC_HAVE_ELEMENTAL)
15: PETSC_INTERN PetscErrorCode MatConvert_SeqSBAIJ_Elemental(Mat,MatType,MatReuse,Mat*);
16: #endif
18: /*
19: Checks for missing diagonals
20: */
21: PetscErrorCode MatMissingDiagonal_SeqSBAIJ(Mat A,PetscBool *missing,PetscInt *dd)
22: {
23: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
25: PetscInt *diag,*ii = a->i,i;
28: MatMarkDiagonal_SeqSBAIJ(A);
29: *missing = PETSC_FALSE;
30: if (A->rmap->n > 0 && !ii) {
31: *missing = PETSC_TRUE;
32: if (dd) *dd = 0;
33: PetscInfo(A,"Matrix has no entries therefore is missing diagonal\n");
34: } else {
35: diag = a->diag;
36: for (i=0; i<a->mbs; i++) {
37: if (diag[i] >= ii[i+1]) {
38: *missing = PETSC_TRUE;
39: if (dd) *dd = i;
40: break;
41: }
42: }
43: }
44: return(0);
45: }
47: PetscErrorCode MatMarkDiagonal_SeqSBAIJ(Mat A)
48: {
49: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
51: PetscInt i,j;
54: if (!a->diag) {
55: PetscMalloc1(a->mbs,&a->diag);
56: PetscLogObjectMemory((PetscObject)A,a->mbs*sizeof(PetscInt));
57: a->free_diag = PETSC_TRUE;
58: }
59: for (i=0; i<a->mbs; i++) {
60: a->diag[i] = a->i[i+1];
61: for (j=a->i[i]; j<a->i[i+1]; j++) {
62: if (a->j[j] == i) {
63: a->diag[i] = j;
64: break;
65: }
66: }
67: }
68: return(0);
69: }
71: static PetscErrorCode MatGetRowIJ_SeqSBAIJ(Mat A,PetscInt oshift,PetscBool symmetric,PetscBool blockcompressed,PetscInt *nn,const PetscInt *inia[],const PetscInt *inja[],PetscBool *done)
72: {
73: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
75: PetscInt i,j,n = a->mbs,nz = a->i[n],*tia,*tja,bs = A->rmap->bs,k,l,cnt;
76: PetscInt **ia = (PetscInt**)inia,**ja = (PetscInt**)inja;
79: *nn = n;
80: if (!ia) return(0);
81: if (symmetric) {
82: MatToSymmetricIJ_SeqAIJ(n,a->i,a->j,PETSC_FALSE,0,0,&tia,&tja);
83: nz = tia[n];
84: } else {
85: tia = a->i; tja = a->j;
86: }
88: if (!blockcompressed && bs > 1) {
89: (*nn) *= bs;
90: /* malloc & create the natural set of indices */
91: PetscMalloc1((n+1)*bs,ia);
92: if (n) {
93: (*ia)[0] = oshift;
94: for (j=1; j<bs; j++) {
95: (*ia)[j] = (tia[1]-tia[0])*bs+(*ia)[j-1];
96: }
97: }
99: for (i=1; i<n; i++) {
100: (*ia)[i*bs] = (tia[i]-tia[i-1])*bs + (*ia)[i*bs-1];
101: for (j=1; j<bs; j++) {
102: (*ia)[i*bs+j] = (tia[i+1]-tia[i])*bs + (*ia)[i*bs+j-1];
103: }
104: }
105: if (n) {
106: (*ia)[n*bs] = (tia[n]-tia[n-1])*bs + (*ia)[n*bs-1];
107: }
109: if (inja) {
110: PetscMalloc1(nz*bs*bs,ja);
111: cnt = 0;
112: for (i=0; i<n; i++) {
113: for (j=0; j<bs; j++) {
114: for (k=tia[i]; k<tia[i+1]; k++) {
115: for (l=0; l<bs; l++) {
116: (*ja)[cnt++] = bs*tja[k] + l;
117: }
118: }
119: }
120: }
121: }
123: if (symmetric) { /* deallocate memory allocated in MatToSymmetricIJ_SeqAIJ() */
124: PetscFree(tia);
125: PetscFree(tja);
126: }
127: } else if (oshift == 1) {
128: if (symmetric) {
129: nz = tia[A->rmap->n/bs];
130: /* add 1 to i and j indices */
131: for (i=0; i<A->rmap->n/bs+1; i++) tia[i] = tia[i] + 1;
132: *ia = tia;
133: if (ja) {
134: for (i=0; i<nz; i++) tja[i] = tja[i] + 1;
135: *ja = tja;
136: }
137: } else {
138: nz = a->i[A->rmap->n/bs];
139: /* malloc space and add 1 to i and j indices */
140: PetscMalloc1(A->rmap->n/bs+1,ia);
141: for (i=0; i<A->rmap->n/bs+1; i++) (*ia)[i] = a->i[i] + 1;
142: if (ja) {
143: PetscMalloc1(nz,ja);
144: for (i=0; i<nz; i++) (*ja)[i] = a->j[i] + 1;
145: }
146: }
147: } else {
148: *ia = tia;
149: if (ja) *ja = tja;
150: }
151: return(0);
152: }
154: static PetscErrorCode MatRestoreRowIJ_SeqSBAIJ(Mat A,PetscInt oshift,PetscBool symmetric,PetscBool blockcompressed,PetscInt *nn,const PetscInt *ia[],const PetscInt *ja[],PetscBool *done)
155: {
159: if (!ia) return(0);
160: if ((!blockcompressed && A->rmap->bs > 1) || (symmetric || oshift == 1)) {
161: PetscFree(*ia);
162: if (ja) {PetscFree(*ja);}
163: }
164: return(0);
165: }
167: PetscErrorCode MatDestroy_SeqSBAIJ(Mat A)
168: {
169: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
173: #if defined(PETSC_USE_LOG)
174: PetscLogObjectState((PetscObject)A,"Rows=%D, NZ=%D",A->rmap->N,a->nz);
175: #endif
176: MatSeqXAIJFreeAIJ(A,&a->a,&a->j,&a->i);
177: if (a->free_diag) {PetscFree(a->diag);}
178: ISDestroy(&a->row);
179: ISDestroy(&a->col);
180: ISDestroy(&a->icol);
181: PetscFree(a->idiag);
182: PetscFree(a->inode.size);
183: if (a->free_imax_ilen) {PetscFree2(a->imax,a->ilen);}
184: PetscFree(a->solve_work);
185: PetscFree(a->sor_work);
186: PetscFree(a->solves_work);
187: PetscFree(a->mult_work);
188: PetscFree(a->saved_values);
189: if (a->free_jshort) {PetscFree(a->jshort);}
190: PetscFree(a->inew);
191: MatDestroy(&a->parent);
192: PetscFree(A->data);
194: PetscObjectChangeTypeName((PetscObject)A,0);
195: PetscObjectComposeFunction((PetscObject)A,"MatStoreValues_C",NULL);
196: PetscObjectComposeFunction((PetscObject)A,"MatRetrieveValues_C",NULL);
197: PetscObjectComposeFunction((PetscObject)A,"MatSeqSBAIJSetColumnIndices_C",NULL);
198: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqsbaij_seqaij_C",NULL);
199: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqsbaij_seqbaij_C",NULL);
200: PetscObjectComposeFunction((PetscObject)A,"MatSeqSBAIJSetPreallocation_C",NULL);
201: PetscObjectComposeFunction((PetscObject)A,"MatSeqSBAIJSetPreallocationCSR_C",NULL);
202: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqsbaij_seqsbstrm_C",NULL);
203: #if defined(PETSC_HAVE_ELEMENTAL)
204: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqsbaij_elemental_C",NULL);
205: #endif
206: return(0);
207: }
209: PetscErrorCode MatSetOption_SeqSBAIJ(Mat A,MatOption op,PetscBool flg)
210: {
211: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
215: switch (op) {
216: case MAT_ROW_ORIENTED:
217: a->roworiented = flg;
218: break;
219: case MAT_KEEP_NONZERO_PATTERN:
220: a->keepnonzeropattern = flg;
221: break;
222: case MAT_NEW_NONZERO_LOCATIONS:
223: a->nonew = (flg ? 0 : 1);
224: break;
225: case MAT_NEW_NONZERO_LOCATION_ERR:
226: a->nonew = (flg ? -1 : 0);
227: break;
228: case MAT_NEW_NONZERO_ALLOCATION_ERR:
229: a->nonew = (flg ? -2 : 0);
230: break;
231: case MAT_UNUSED_NONZERO_LOCATION_ERR:
232: a->nounused = (flg ? -1 : 0);
233: break;
234: case MAT_NEW_DIAGONALS:
235: case MAT_IGNORE_OFF_PROC_ENTRIES:
236: case MAT_USE_HASH_TABLE:
237: PetscInfo1(A,"Option %s ignored\n",MatOptions[op]);
238: break;
239: case MAT_HERMITIAN:
240: #if defined(PETSC_USE_COMPLEX) /* MAT_HERMITIAN is a synonym for MAT_SYMMETRIC with reals */
241: if (!A->assembled) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Must call MatAssemblyEnd() first");
242: if (A->cmap->n < 65536 && A->cmap->bs == 1) {
243: A->ops->mult = MatMult_SeqSBAIJ_1_Hermitian_ushort;
244: } else if (A->cmap->bs == 1) {
245: A->ops->mult = MatMult_SeqSBAIJ_1_Hermitian;
246: } else SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"No support for Hermitian with block size greater than 1");
247: #endif
248: break;
249: case MAT_SPD:
250: /* These options are handled directly by MatSetOption() */
251: break;
252: case MAT_SYMMETRIC:
253: case MAT_STRUCTURALLY_SYMMETRIC:
254: case MAT_SYMMETRY_ETERNAL:
255: case MAT_STRUCTURE_ONLY:
256: /* These options are handled directly by MatSetOption() */
257: break;
258: case MAT_IGNORE_LOWER_TRIANGULAR:
259: a->ignore_ltriangular = flg;
260: break;
261: case MAT_ERROR_LOWER_TRIANGULAR:
262: a->ignore_ltriangular = flg;
263: break;
264: case MAT_GETROW_UPPERTRIANGULAR:
265: a->getrow_utriangular = flg;
266: break;
267: case MAT_SUBMAT_SINGLEIS:
268: break;
269: default:
270: SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"unknown option %d",op);
271: }
272: return(0);
273: }
275: PetscErrorCode MatGetRow_SeqSBAIJ(Mat A,PetscInt row,PetscInt *nz,PetscInt **idx,PetscScalar **v)
276: {
277: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
281: if (A && !a->getrow_utriangular) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"MatGetRow is not supported for SBAIJ matrix format. Getting the upper triangular part of row, run with -mat_getrow_uppertriangular, call MatSetOption(mat,MAT_GETROW_UPPERTRIANGULAR,PETSC_TRUE) or MatGetRowUpperTriangular()");
283: /* Get the upper triangular part of the row */
284: MatGetRow_SeqBAIJ_private(A,row,nz,idx,v,a->i,a->j,a->a);
285: return(0);
286: }
288: PetscErrorCode MatRestoreRow_SeqSBAIJ(Mat A,PetscInt row,PetscInt *nz,PetscInt **idx,PetscScalar **v)
289: {
293: if (idx) {PetscFree(*idx);}
294: if (v) {PetscFree(*v);}
295: return(0);
296: }
298: PetscErrorCode MatGetRowUpperTriangular_SeqSBAIJ(Mat A)
299: {
300: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
303: a->getrow_utriangular = PETSC_TRUE;
304: return(0);
305: }
306: PetscErrorCode MatRestoreRowUpperTriangular_SeqSBAIJ(Mat A)
307: {
308: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
311: a->getrow_utriangular = PETSC_FALSE;
312: return(0);
313: }
315: PetscErrorCode MatTranspose_SeqSBAIJ(Mat A,MatReuse reuse,Mat *B)
316: {
320: if (reuse == MAT_INITIAL_MATRIX) {
321: MatDuplicate(A,MAT_COPY_VALUES,B);
322: } else if (reuse == MAT_REUSE_MATRIX) {
323: MatCopy(A,*B,SAME_NONZERO_PATTERN);
324: }
325: return(0);
326: }
328: PetscErrorCode MatView_SeqSBAIJ_ASCII(Mat A,PetscViewer viewer)
329: {
330: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
331: PetscErrorCode ierr;
332: PetscInt i,j,bs = A->rmap->bs,k,l,bs2=a->bs2;
333: PetscViewerFormat format;
334: PetscInt *diag;
337: PetscViewerGetFormat(viewer,&format);
338: if (format == PETSC_VIEWER_ASCII_INFO || format == PETSC_VIEWER_ASCII_INFO_DETAIL) {
339: PetscViewerASCIIPrintf(viewer," block size is %D\n",bs);
340: } else if (format == PETSC_VIEWER_ASCII_MATLAB) {
341: Mat aij;
342: const char *matname;
344: if (A->factortype && bs>1) {
345: PetscPrintf(PETSC_COMM_SELF,"Warning: matrix is factored with bs>1. MatView() with PETSC_VIEWER_ASCII_MATLAB is not supported and ignored!\n");
346: return(0);
347: }
348: MatConvert(A,MATSEQAIJ,MAT_INITIAL_MATRIX,&aij);
349: PetscObjectGetName((PetscObject)A,&matname);
350: PetscObjectSetName((PetscObject)aij,matname);
351: MatView(aij,viewer);
352: MatDestroy(&aij);
353: } else if (format == PETSC_VIEWER_ASCII_COMMON) {
354: PetscViewerASCIIUseTabs(viewer,PETSC_FALSE);
355: for (i=0; i<a->mbs; i++) {
356: for (j=0; j<bs; j++) {
357: PetscViewerASCIIPrintf(viewer,"row %D:",i*bs+j);
358: for (k=a->i[i]; k<a->i[i+1]; k++) {
359: for (l=0; l<bs; l++) {
360: #if defined(PETSC_USE_COMPLEX)
361: if (PetscImaginaryPart(a->a[bs2*k + l*bs + j]) > 0.0 && PetscRealPart(a->a[bs2*k + l*bs + j]) != 0.0) {
362: PetscViewerASCIIPrintf(viewer," (%D, %g + %g i) ",bs*a->j[k]+l,
363: (double)PetscRealPart(a->a[bs2*k + l*bs + j]),(double)PetscImaginaryPart(a->a[bs2*k + l*bs + j]));
364: } else if (PetscImaginaryPart(a->a[bs2*k + l*bs + j]) < 0.0 && PetscRealPart(a->a[bs2*k + l*bs + j]) != 0.0) {
365: PetscViewerASCIIPrintf(viewer," (%D, %g - %g i) ",bs*a->j[k]+l,
366: (double)PetscRealPart(a->a[bs2*k + l*bs + j]),-(double)PetscImaginaryPart(a->a[bs2*k + l*bs + j]));
367: } else if (PetscRealPart(a->a[bs2*k + l*bs + j]) != 0.0) {
368: PetscViewerASCIIPrintf(viewer," (%D, %g) ",bs*a->j[k]+l,(double)PetscRealPart(a->a[bs2*k + l*bs + j]));
369: }
370: #else
371: if (a->a[bs2*k + l*bs + j] != 0.0) {
372: PetscViewerASCIIPrintf(viewer," (%D, %g) ",bs*a->j[k]+l,(double)a->a[bs2*k + l*bs + j]);
373: }
374: #endif
375: }
376: }
377: PetscViewerASCIIPrintf(viewer,"\n");
378: }
379: }
380: PetscViewerASCIIUseTabs(viewer,PETSC_TRUE);
381: } else if (format == PETSC_VIEWER_ASCII_FACTOR_INFO) {
382: return(0);
383: } else {
384: PetscViewerASCIIUseTabs(viewer,PETSC_FALSE);
385: if (A->factortype) { /* for factored matrix */
386: if (bs>1) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"matrix is factored with bs>1. Not implemented yet");
388: diag=a->diag;
389: for (i=0; i<a->mbs; i++) { /* for row block i */
390: PetscViewerASCIIPrintf(viewer,"row %D:",i);
391: /* diagonal entry */
392: #if defined(PETSC_USE_COMPLEX)
393: if (PetscImaginaryPart(a->a[diag[i]]) > 0.0) {
394: PetscViewerASCIIPrintf(viewer," (%D, %g + %g i) ",a->j[diag[i]],(double)PetscRealPart(1.0/a->a[diag[i]]),(double)PetscImaginaryPart(1.0/a->a[diag[i]]));
395: } else if (PetscImaginaryPart(a->a[diag[i]]) < 0.0) {
396: PetscViewerASCIIPrintf(viewer," (%D, %g - %g i) ",a->j[diag[i]],(double)PetscRealPart(1.0/a->a[diag[i]]),-(double)PetscImaginaryPart(1.0/a->a[diag[i]]));
397: } else {
398: PetscViewerASCIIPrintf(viewer," (%D, %g) ",a->j[diag[i]],(double)PetscRealPart(1.0/a->a[diag[i]]));
399: }
400: #else
401: PetscViewerASCIIPrintf(viewer," (%D, %g) ",a->j[diag[i]],(double)(1.0/a->a[diag[i]]));
402: #endif
403: /* off-diagonal entries */
404: for (k=a->i[i]; k<a->i[i+1]-1; k++) {
405: #if defined(PETSC_USE_COMPLEX)
406: if (PetscImaginaryPart(a->a[k]) > 0.0) {
407: PetscViewerASCIIPrintf(viewer," (%D, %g + %g i) ",bs*a->j[k],(double)PetscRealPart(a->a[k]),(double)PetscImaginaryPart(a->a[k]));
408: } else if (PetscImaginaryPart(a->a[k]) < 0.0) {
409: PetscViewerASCIIPrintf(viewer," (%D, %g - %g i) ",bs*a->j[k],(double)PetscRealPart(a->a[k]),-(double)PetscImaginaryPart(a->a[k]));
410: } else {
411: PetscViewerASCIIPrintf(viewer," (%D, %g) ",bs*a->j[k],(double)PetscRealPart(a->a[k]));
412: }
413: #else
414: PetscViewerASCIIPrintf(viewer," (%D, %g) ",a->j[k],(double)a->a[k]);
415: #endif
416: }
417: PetscViewerASCIIPrintf(viewer,"\n");
418: }
420: } else { /* for non-factored matrix */
421: for (i=0; i<a->mbs; i++) { /* for row block i */
422: for (j=0; j<bs; j++) { /* for row bs*i + j */
423: PetscViewerASCIIPrintf(viewer,"row %D:",i*bs+j);
424: for (k=a->i[i]; k<a->i[i+1]; k++) { /* for column block */
425: for (l=0; l<bs; l++) { /* for column */
426: #if defined(PETSC_USE_COMPLEX)
427: if (PetscImaginaryPart(a->a[bs2*k + l*bs + j]) > 0.0) {
428: PetscViewerASCIIPrintf(viewer," (%D, %g + %g i) ",bs*a->j[k]+l,
429: (double)PetscRealPart(a->a[bs2*k + l*bs + j]),(double)PetscImaginaryPart(a->a[bs2*k + l*bs + j]));
430: } else if (PetscImaginaryPart(a->a[bs2*k + l*bs + j]) < 0.0) {
431: PetscViewerASCIIPrintf(viewer," (%D, %g - %g i) ",bs*a->j[k]+l,
432: (double)PetscRealPart(a->a[bs2*k + l*bs + j]),-(double)PetscImaginaryPart(a->a[bs2*k + l*bs + j]));
433: } else {
434: PetscViewerASCIIPrintf(viewer," (%D, %g) ",bs*a->j[k]+l,(double)PetscRealPart(a->a[bs2*k + l*bs + j]));
435: }
436: #else
437: PetscViewerASCIIPrintf(viewer," (%D, %g) ",bs*a->j[k]+l,(double)a->a[bs2*k + l*bs + j]);
438: #endif
439: }
440: }
441: PetscViewerASCIIPrintf(viewer,"\n");
442: }
443: }
444: }
445: PetscViewerASCIIUseTabs(viewer,PETSC_TRUE);
446: }
447: PetscViewerFlush(viewer);
448: return(0);
449: }
451: #include <petscdraw.h>
452: static PetscErrorCode MatView_SeqSBAIJ_Draw_Zoom(PetscDraw draw,void *Aa)
453: {
454: Mat A = (Mat) Aa;
455: Mat_SeqSBAIJ *a=(Mat_SeqSBAIJ*)A->data;
457: PetscInt row,i,j,k,l,mbs=a->mbs,color,bs=A->rmap->bs,bs2=a->bs2;
458: PetscReal xl,yl,xr,yr,x_l,x_r,y_l,y_r;
459: MatScalar *aa;
460: PetscViewer viewer;
463: PetscObjectQuery((PetscObject)A,"Zoomviewer",(PetscObject*)&viewer);
464: PetscDrawGetCoordinates(draw,&xl,&yl,&xr,&yr);
466: /* loop over matrix elements drawing boxes */
468: PetscDrawCollectiveBegin(draw);
469: PetscDrawString(draw, .3*(xl+xr), .3*(yl+yr), PETSC_DRAW_BLACK, "symmetric");
470: /* Blue for negative, Cyan for zero and Red for positive */
471: color = PETSC_DRAW_BLUE;
472: for (i=0,row=0; i<mbs; i++,row+=bs) {
473: for (j=a->i[i]; j<a->i[i+1]; j++) {
474: y_l = A->rmap->N - row - 1.0; y_r = y_l + 1.0;
475: x_l = a->j[j]*bs; x_r = x_l + 1.0;
476: aa = a->a + j*bs2;
477: for (k=0; k<bs; k++) {
478: for (l=0; l<bs; l++) {
479: if (PetscRealPart(*aa++) >= 0.) continue;
480: PetscDrawRectangle(draw,x_l+k,y_l-l,x_r+k,y_r-l,color,color,color,color);
481: }
482: }
483: }
484: }
485: color = PETSC_DRAW_CYAN;
486: for (i=0,row=0; i<mbs; i++,row+=bs) {
487: for (j=a->i[i]; j<a->i[i+1]; j++) {
488: y_l = A->rmap->N - row - 1.0; y_r = y_l + 1.0;
489: x_l = a->j[j]*bs; x_r = x_l + 1.0;
490: aa = a->a + j*bs2;
491: for (k=0; k<bs; k++) {
492: for (l=0; l<bs; l++) {
493: if (PetscRealPart(*aa++) != 0.) continue;
494: PetscDrawRectangle(draw,x_l+k,y_l-l,x_r+k,y_r-l,color,color,color,color);
495: }
496: }
497: }
498: }
499: color = PETSC_DRAW_RED;
500: for (i=0,row=0; i<mbs; i++,row+=bs) {
501: for (j=a->i[i]; j<a->i[i+1]; j++) {
502: y_l = A->rmap->N - row - 1.0; y_r = y_l + 1.0;
503: x_l = a->j[j]*bs; x_r = x_l + 1.0;
504: aa = a->a + j*bs2;
505: for (k=0; k<bs; k++) {
506: for (l=0; l<bs; l++) {
507: if (PetscRealPart(*aa++) <= 0.) continue;
508: PetscDrawRectangle(draw,x_l+k,y_l-l,x_r+k,y_r-l,color,color,color,color);
509: }
510: }
511: }
512: }
513: PetscDrawCollectiveEnd(draw);
514: return(0);
515: }
517: static PetscErrorCode MatView_SeqSBAIJ_Draw(Mat A,PetscViewer viewer)
518: {
520: PetscReal xl,yl,xr,yr,w,h;
521: PetscDraw draw;
522: PetscBool isnull;
525: PetscViewerDrawGetDraw(viewer,0,&draw);
526: PetscDrawIsNull(draw,&isnull);
527: if (isnull) return(0);
529: xr = A->rmap->N; yr = A->rmap->N; h = yr/10.0; w = xr/10.0;
530: xr += w; yr += h; xl = -w; yl = -h;
531: PetscDrawSetCoordinates(draw,xl,yl,xr,yr);
532: PetscObjectCompose((PetscObject)A,"Zoomviewer",(PetscObject)viewer);
533: PetscDrawZoom(draw,MatView_SeqSBAIJ_Draw_Zoom,A);
534: PetscObjectCompose((PetscObject)A,"Zoomviewer",NULL);
535: PetscDrawSave(draw);
536: return(0);
537: }
539: PetscErrorCode MatView_SeqSBAIJ(Mat A,PetscViewer viewer)
540: {
542: PetscBool iascii,isdraw;
543: FILE *file = 0;
546: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERASCII,&iascii);
547: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERDRAW,&isdraw);
548: if (iascii) {
549: MatView_SeqSBAIJ_ASCII(A,viewer);
550: } else if (isdraw) {
551: MatView_SeqSBAIJ_Draw(A,viewer);
552: } else {
553: Mat B;
554: const char *matname;
555: MatConvert(A,MATSEQAIJ,MAT_INITIAL_MATRIX,&B);
556: PetscObjectGetName((PetscObject)A,&matname);
557: PetscObjectSetName((PetscObject)B,matname);
558: MatView(B,viewer);
559: MatDestroy(&B);
560: PetscViewerBinaryGetInfoPointer(viewer,&file);
561: if (file) {
562: fprintf(file,"-matload_block_size %d\n",(int)A->rmap->bs);
563: }
564: }
565: return(0);
566: }
569: PetscErrorCode MatGetValues_SeqSBAIJ(Mat A,PetscInt m,const PetscInt im[],PetscInt n,const PetscInt in[],PetscScalar v[])
570: {
571: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
572: PetscInt *rp,k,low,high,t,row,nrow,i,col,l,*aj = a->j;
573: PetscInt *ai = a->i,*ailen = a->ilen;
574: PetscInt brow,bcol,ridx,cidx,bs=A->rmap->bs,bs2=a->bs2;
575: MatScalar *ap,*aa = a->a;
578: for (k=0; k<m; k++) { /* loop over rows */
579: row = im[k]; brow = row/bs;
580: if (row < 0) {v += n; continue;} /* SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Negative row: %D",row); */
581: if (row >= A->rmap->N) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Row too large: row %D max %D",row,A->rmap->N-1);
582: rp = aj + ai[brow]; ap = aa + bs2*ai[brow];
583: nrow = ailen[brow];
584: for (l=0; l<n; l++) { /* loop over columns */
585: if (in[l] < 0) {v++; continue;} /* SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Negative column: %D",in[l]); */
586: if (in[l] >= A->cmap->n) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Column too large: col %D max %D",in[l],A->cmap->n-1);
587: col = in[l];
588: bcol = col/bs;
589: cidx = col%bs;
590: ridx = row%bs;
591: high = nrow;
592: low = 0; /* assume unsorted */
593: while (high-low > 5) {
594: t = (low+high)/2;
595: if (rp[t] > bcol) high = t;
596: else low = t;
597: }
598: for (i=low; i<high; i++) {
599: if (rp[i] > bcol) break;
600: if (rp[i] == bcol) {
601: *v++ = ap[bs2*i+bs*cidx+ridx];
602: goto finished;
603: }
604: }
605: *v++ = 0.0;
606: finished:;
607: }
608: }
609: return(0);
610: }
613: PetscErrorCode MatSetValuesBlocked_SeqSBAIJ(Mat A,PetscInt m,const PetscInt im[],PetscInt n,const PetscInt in[],const PetscScalar v[],InsertMode is)
614: {
615: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
616: PetscErrorCode ierr;
617: PetscInt *rp,k,low,high,t,ii,jj,row,nrow,i,col,l,rmax,N,lastcol = -1;
618: PetscInt *imax =a->imax,*ai=a->i,*ailen=a->ilen;
619: PetscInt *aj =a->j,nonew=a->nonew,bs2=a->bs2,bs=A->rmap->bs,stepval;
620: PetscBool roworiented=a->roworiented;
621: const PetscScalar *value = v;
622: MatScalar *ap,*aa = a->a,*bap;
625: if (roworiented) stepval = (n-1)*bs;
626: else stepval = (m-1)*bs;
628: for (k=0; k<m; k++) { /* loop over added rows */
629: row = im[k];
630: if (row < 0) continue;
631: #if defined(PETSC_USE_DEBUG)
632: if (row >= a->mbs) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Block index row too large %D max %D",row,a->mbs-1);
633: #endif
634: rp = aj + ai[row];
635: ap = aa + bs2*ai[row];
636: rmax = imax[row];
637: nrow = ailen[row];
638: low = 0;
639: high = nrow;
640: for (l=0; l<n; l++) { /* loop over added columns */
641: if (in[l] < 0) continue;
642: col = in[l];
643: #if defined(PETSC_USE_DEBUG)
644: if (col >= a->nbs) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Block index column too large %D max %D",col,a->nbs-1);
645: #endif
646: if (col < row) {
647: if (a->ignore_ltriangular) continue; /* ignore lower triangular block */
648: else SETERRQ(PETSC_COMM_SELF,PETSC_ERR_USER,"Lower triangular value cannot be set for sbaij format. Ignoring these values, run with -mat_ignore_lower_triangular or call MatSetOption(mat,MAT_IGNORE_LOWER_TRIANGULAR,PETSC_TRUE)");
649: }
650: if (roworiented) value = v + k*(stepval+bs)*bs + l*bs;
651: else value = v + l*(stepval+bs)*bs + k*bs;
653: if (col <= lastcol) low = 0;
654: else high = nrow;
656: lastcol = col;
657: while (high-low > 7) {
658: t = (low+high)/2;
659: if (rp[t] > col) high = t;
660: else low = t;
661: }
662: for (i=low; i<high; i++) {
663: if (rp[i] > col) break;
664: if (rp[i] == col) {
665: bap = ap + bs2*i;
666: if (roworiented) {
667: if (is == ADD_VALUES) {
668: for (ii=0; ii<bs; ii++,value+=stepval) {
669: for (jj=ii; jj<bs2; jj+=bs) {
670: bap[jj] += *value++;
671: }
672: }
673: } else {
674: for (ii=0; ii<bs; ii++,value+=stepval) {
675: for (jj=ii; jj<bs2; jj+=bs) {
676: bap[jj] = *value++;
677: }
678: }
679: }
680: } else {
681: if (is == ADD_VALUES) {
682: for (ii=0; ii<bs; ii++,value+=stepval) {
683: for (jj=0; jj<bs; jj++) {
684: *bap++ += *value++;
685: }
686: }
687: } else {
688: for (ii=0; ii<bs; ii++,value+=stepval) {
689: for (jj=0; jj<bs; jj++) {
690: *bap++ = *value++;
691: }
692: }
693: }
694: }
695: goto noinsert2;
696: }
697: }
698: if (nonew == 1) goto noinsert2;
699: if (nonew == -1) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Inserting a new block index nonzero block (%D, %D) in the matrix", row, col);
700: MatSeqXAIJReallocateAIJ(A,a->mbs,bs2,nrow,row,col,rmax,aa,ai,aj,rp,ap,imax,nonew,MatScalar);
701: N = nrow++ - 1; high++;
702: /* shift up all the later entries in this row */
703: for (ii=N; ii>=i; ii--) {
704: rp[ii+1] = rp[ii];
705: PetscMemcpy(ap+bs2*(ii+1),ap+bs2*(ii),bs2*sizeof(MatScalar));
706: }
707: if (N >= i) {
708: PetscMemzero(ap+bs2*i,bs2*sizeof(MatScalar));
709: }
710: rp[i] = col;
711: bap = ap + bs2*i;
712: if (roworiented) {
713: for (ii=0; ii<bs; ii++,value+=stepval) {
714: for (jj=ii; jj<bs2; jj+=bs) {
715: bap[jj] = *value++;
716: }
717: }
718: } else {
719: for (ii=0; ii<bs; ii++,value+=stepval) {
720: for (jj=0; jj<bs; jj++) {
721: *bap++ = *value++;
722: }
723: }
724: }
725: noinsert2:;
726: low = i;
727: }
728: ailen[row] = nrow;
729: }
730: return(0);
731: }
733: /*
734: This is not yet used
735: */
736: PetscErrorCode MatAssemblyEnd_SeqSBAIJ_SeqAIJ_Inode(Mat A)
737: {
738: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
740: const PetscInt *ai = a->i, *aj = a->j,*cols;
741: PetscInt i = 0,j,blk_size,m = A->rmap->n,node_count = 0,nzx,nzy,*ns,row,nz,cnt,cnt2,*counts;
742: PetscBool flag;
745: PetscMalloc1(m,&ns);
746: while (i < m) {
747: nzx = ai[i+1] - ai[i]; /* Number of nonzeros */
748: /* Limits the number of elements in a node to 'a->inode.limit' */
749: for (j=i+1,blk_size=1; j<m && blk_size <a->inode.limit; ++j,++blk_size) {
750: nzy = ai[j+1] - ai[j];
751: if (nzy != (nzx - j + i)) break;
752: PetscMemcmp(aj + ai[i] + j - i,aj + ai[j],nzy*sizeof(PetscInt),&flag);
753: if (!flag) break;
754: }
755: ns[node_count++] = blk_size;
757: i = j;
758: }
759: if (!a->inode.size && m && node_count > .9*m) {
760: PetscFree(ns);
761: PetscInfo2(A,"Found %D nodes out of %D rows. Not using Inode routines\n",node_count,m);
762: } else {
763: a->inode.node_count = node_count;
765: PetscMalloc1(node_count,&a->inode.size);
766: PetscLogObjectMemory((PetscObject)A,node_count*sizeof(PetscInt));
767: PetscMemcpy(a->inode.size,ns,node_count*sizeof(PetscInt));
768: PetscFree(ns);
769: PetscInfo3(A,"Found %D nodes of %D. Limit used: %D. Using Inode routines\n",node_count,m,a->inode.limit);
771: /* count collections of adjacent columns in each inode */
772: row = 0;
773: cnt = 0;
774: for (i=0; i<node_count; i++) {
775: cols = aj + ai[row] + a->inode.size[i];
776: nz = ai[row+1] - ai[row] - a->inode.size[i];
777: for (j=1; j<nz; j++) {
778: if (cols[j] != cols[j-1]+1) cnt++;
779: }
780: cnt++;
781: row += a->inode.size[i];
782: }
783: PetscMalloc1(2*cnt,&counts);
784: cnt = 0;
785: row = 0;
786: for (i=0; i<node_count; i++) {
787: cols = aj + ai[row] + a->inode.size[i];
788: counts[2*cnt] = cols[0];
789: nz = ai[row+1] - ai[row] - a->inode.size[i];
790: cnt2 = 1;
791: for (j=1; j<nz; j++) {
792: if (cols[j] != cols[j-1]+1) {
793: counts[2*(cnt++)+1] = cnt2;
794: counts[2*cnt] = cols[j];
795: cnt2 = 1;
796: } else cnt2++;
797: }
798: counts[2*(cnt++)+1] = cnt2;
799: row += a->inode.size[i];
800: }
801: PetscIntView(2*cnt,counts,0);
802: }
803: return(0);
804: }
806: PetscErrorCode MatAssemblyEnd_SeqSBAIJ(Mat A,MatAssemblyType mode)
807: {
808: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
810: PetscInt fshift = 0,i,j,*ai = a->i,*aj = a->j,*imax = a->imax;
811: PetscInt m = A->rmap->N,*ip,N,*ailen = a->ilen;
812: PetscInt mbs = a->mbs,bs2 = a->bs2,rmax = 0;
813: MatScalar *aa = a->a,*ap;
816: if (mode == MAT_FLUSH_ASSEMBLY) return(0);
818: if (m) rmax = ailen[0];
819: for (i=1; i<mbs; i++) {
820: /* move each row back by the amount of empty slots (fshift) before it*/
821: fshift += imax[i-1] - ailen[i-1];
822: rmax = PetscMax(rmax,ailen[i]);
823: if (fshift) {
824: ip = aj + ai[i]; ap = aa + bs2*ai[i];
825: N = ailen[i];
826: for (j=0; j<N; j++) {
827: ip[j-fshift] = ip[j];
828: PetscMemcpy(ap+(j-fshift)*bs2,ap+j*bs2,bs2*sizeof(MatScalar));
829: }
830: }
831: ai[i] = ai[i-1] + ailen[i-1];
832: }
833: if (mbs) {
834: fshift += imax[mbs-1] - ailen[mbs-1];
835: ai[mbs] = ai[mbs-1] + ailen[mbs-1];
836: }
837: /* reset ilen and imax for each row */
838: for (i=0; i<mbs; i++) {
839: ailen[i] = imax[i] = ai[i+1] - ai[i];
840: }
841: a->nz = ai[mbs];
843: /* diagonals may have moved, reset it */
844: if (a->diag) {
845: PetscMemcpy(a->diag,ai,mbs*sizeof(PetscInt));
846: }
847: if (fshift && a->nounused == -1) SETERRQ4(PETSC_COMM_SELF,PETSC_ERR_PLIB, "Unused space detected in matrix: %D X %D block size %D, %D unneeded", m, A->cmap->n, A->rmap->bs, fshift*bs2);
849: PetscInfo5(A,"Matrix size: %D X %D, block size %D; storage space: %D unneeded, %D used\n",m,A->rmap->N,A->rmap->bs,fshift*bs2,a->nz*bs2);
850: PetscInfo1(A,"Number of mallocs during MatSetValues is %D\n",a->reallocs);
851: PetscInfo1(A,"Most nonzeros blocks in any row is %D\n",rmax);
853: A->info.mallocs += a->reallocs;
854: a->reallocs = 0;
855: A->info.nz_unneeded = (PetscReal)fshift*bs2;
856: a->idiagvalid = PETSC_FALSE;
857: a->rmax = rmax;
859: if (A->cmap->n < 65536 && A->cmap->bs == 1) {
860: if (a->jshort && a->free_jshort) {
861: /* when matrix data structure is changed, previous jshort must be replaced */
862: PetscFree(a->jshort);
863: }
864: PetscMalloc1(a->i[A->rmap->n],&a->jshort);
865: PetscLogObjectMemory((PetscObject)A,a->i[A->rmap->n]*sizeof(unsigned short));
866: for (i=0; i<a->i[A->rmap->n]; i++) a->jshort[i] = a->j[i];
867: A->ops->mult = MatMult_SeqSBAIJ_1_ushort;
868: A->ops->sor = MatSOR_SeqSBAIJ_ushort;
869: a->free_jshort = PETSC_TRUE;
870: }
871: return(0);
872: }
874: /*
875: This function returns an array of flags which indicate the locations of contiguous
876: blocks that should be zeroed. for eg: if bs = 3 and is = [0,1,2,3,5,6,7,8,9]
877: then the resulting sizes = [3,1,1,3,1] correspondig to sets [(0,1,2),(3),(5),(6,7,8),(9)]
878: Assume: sizes should be long enough to hold all the values.
879: */
880: PetscErrorCode MatZeroRows_SeqSBAIJ_Check_Blocks(PetscInt idx[],PetscInt n,PetscInt bs,PetscInt sizes[], PetscInt *bs_max)
881: {
882: PetscInt i,j,k,row;
883: PetscBool flg;
886: for (i=0,j=0; i<n; j++) {
887: row = idx[i];
888: if (row%bs!=0) { /* Not the begining of a block */
889: sizes[j] = 1;
890: i++;
891: } else if (i+bs > n) { /* Beginning of a block, but complete block doesn't exist (at idx end) */
892: sizes[j] = 1; /* Also makes sure atleast 'bs' values exist for next else */
893: i++;
894: } else { /* Begining of the block, so check if the complete block exists */
895: flg = PETSC_TRUE;
896: for (k=1; k<bs; k++) {
897: if (row+k != idx[i+k]) { /* break in the block */
898: flg = PETSC_FALSE;
899: break;
900: }
901: }
902: if (flg) { /* No break in the bs */
903: sizes[j] = bs;
904: i += bs;
905: } else {
906: sizes[j] = 1;
907: i++;
908: }
909: }
910: }
911: *bs_max = j;
912: return(0);
913: }
916: /* Only add/insert a(i,j) with i<=j (blocks).
917: Any a(i,j) with i>j input by user is ingored.
918: */
920: PetscErrorCode MatSetValues_SeqSBAIJ(Mat A,PetscInt m,const PetscInt im[],PetscInt n,const PetscInt in[],const PetscScalar v[],InsertMode is)
921: {
922: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
924: PetscInt *rp,k,low,high,t,ii,row,nrow,i,col,l,rmax,N,lastcol = -1;
925: PetscInt *imax=a->imax,*ai=a->i,*ailen=a->ilen,roworiented=a->roworiented;
926: PetscInt *aj =a->j,nonew=a->nonew,bs=A->rmap->bs,brow,bcol;
927: PetscInt ridx,cidx,bs2=a->bs2;
928: MatScalar *ap,value,*aa=a->a,*bap;
931: for (k=0; k<m; k++) { /* loop over added rows */
932: row = im[k]; /* row number */
933: brow = row/bs; /* block row number */
934: if (row < 0) continue;
935: #if defined(PETSC_USE_DEBUG)
936: if (row >= A->rmap->N) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Row too large: row %D max %D",row,A->rmap->N-1);
937: #endif
938: rp = aj + ai[brow]; /*ptr to beginning of column value of the row block*/
939: ap = aa + bs2*ai[brow]; /*ptr to beginning of element value of the row block*/
940: rmax = imax[brow]; /* maximum space allocated for this row */
941: nrow = ailen[brow]; /* actual length of this row */
942: low = 0;
944: for (l=0; l<n; l++) { /* loop over added columns */
945: if (in[l] < 0) continue;
946: #if defined(PETSC_USE_DEBUG)
947: if (in[l] >= A->rmap->N) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Column too large: col %D max %D",in[l],A->rmap->N-1);
948: #endif
949: col = in[l];
950: bcol = col/bs; /* block col number */
952: if (brow > bcol) {
953: if (a->ignore_ltriangular) continue; /* ignore lower triangular values */
954: else SETERRQ(PETSC_COMM_SELF,PETSC_ERR_USER,"Lower triangular value cannot be set for sbaij format. Ignoring these values, run with -mat_ignore_lower_triangular or call MatSetOption(mat,MAT_IGNORE_LOWER_TRIANGULAR,PETSC_TRUE)");
955: }
957: ridx = row % bs; cidx = col % bs; /*row and col index inside the block */
958: if ((brow==bcol && ridx<=cidx) || (brow<bcol)) {
959: /* element value a(k,l) */
960: if (roworiented) value = v[l + k*n];
961: else value = v[k + l*m];
963: /* move pointer bap to a(k,l) quickly and add/insert value */
964: if (col <= lastcol) low = 0;
965: high = nrow;
966: lastcol = col;
967: while (high-low > 7) {
968: t = (low+high)/2;
969: if (rp[t] > bcol) high = t;
970: else low = t;
971: }
972: for (i=low; i<high; i++) {
973: if (rp[i] > bcol) break;
974: if (rp[i] == bcol) {
975: bap = ap + bs2*i + bs*cidx + ridx;
976: if (is == ADD_VALUES) *bap += value;
977: else *bap = value;
978: /* for diag block, add/insert its symmetric element a(cidx,ridx) */
979: if (brow == bcol && ridx < cidx) {
980: bap = ap + bs2*i + bs*ridx + cidx;
981: if (is == ADD_VALUES) *bap += value;
982: else *bap = value;
983: }
984: goto noinsert1;
985: }
986: }
988: if (nonew == 1) goto noinsert1;
989: if (nonew == -1) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Inserting a new nonzero (%D, %D) in the matrix", row, col);
990: MatSeqXAIJReallocateAIJ(A,a->mbs,bs2,nrow,brow,bcol,rmax,aa,ai,aj,rp,ap,imax,nonew,MatScalar);
992: N = nrow++ - 1; high++;
993: /* shift up all the later entries in this row */
994: for (ii=N; ii>=i; ii--) {
995: rp[ii+1] = rp[ii];
996: PetscMemcpy(ap+bs2*(ii+1),ap+bs2*(ii),bs2*sizeof(MatScalar));
997: }
998: if (N>=i) {
999: PetscMemzero(ap+bs2*i,bs2*sizeof(MatScalar));
1000: }
1001: rp[i] = bcol;
1002: ap[bs2*i + bs*cidx + ridx] = value;
1003: A->nonzerostate++;
1004: noinsert1:;
1005: low = i;
1006: }
1007: } /* end of loop over added columns */
1008: ailen[brow] = nrow;
1009: } /* end of loop over added rows */
1010: return(0);
1011: }
1013: PetscErrorCode MatICCFactor_SeqSBAIJ(Mat inA,IS row,const MatFactorInfo *info)
1014: {
1015: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)inA->data;
1016: Mat outA;
1018: PetscBool row_identity;
1021: if (info->levels != 0) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Only levels=0 is supported for in-place icc");
1022: ISIdentity(row,&row_identity);
1023: if (!row_identity) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Matrix reordering is not supported");
1024: if (inA->rmap->bs != 1) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"Matrix block size %D is not supported",inA->rmap->bs); /* Need to replace MatCholeskyFactorSymbolic_SeqSBAIJ_MSR()! */
1026: outA = inA;
1027: inA->factortype = MAT_FACTOR_ICC;
1028: PetscFree(inA->solvertype);
1029: PetscStrallocpy(MATSOLVERPETSC,&inA->solvertype);
1031: MatMarkDiagonal_SeqSBAIJ(inA);
1032: MatSeqSBAIJSetNumericFactorization_inplace(inA,row_identity);
1034: PetscObjectReference((PetscObject)row);
1035: ISDestroy(&a->row);
1036: a->row = row;
1037: PetscObjectReference((PetscObject)row);
1038: ISDestroy(&a->col);
1039: a->col = row;
1041: /* Create the invert permutation so that it can be used in MatCholeskyFactorNumeric() */
1042: if (a->icol) {ISInvertPermutation(row,PETSC_DECIDE, &a->icol);}
1043: PetscLogObjectParent((PetscObject)inA,(PetscObject)a->icol);
1045: if (!a->solve_work) {
1046: PetscMalloc1(inA->rmap->N+inA->rmap->bs,&a->solve_work);
1047: PetscLogObjectMemory((PetscObject)inA,(inA->rmap->N+inA->rmap->bs)*sizeof(PetscScalar));
1048: }
1050: MatCholeskyFactorNumeric(outA,inA,info);
1051: return(0);
1052: }
1054: PetscErrorCode MatSeqSBAIJSetColumnIndices_SeqSBAIJ(Mat mat,PetscInt *indices)
1055: {
1056: Mat_SeqSBAIJ *baij = (Mat_SeqSBAIJ*)mat->data;
1057: PetscInt i,nz,n;
1061: nz = baij->maxnz;
1062: n = mat->cmap->n;
1063: for (i=0; i<nz; i++) baij->j[i] = indices[i];
1065: baij->nz = nz;
1066: for (i=0; i<n; i++) baij->ilen[i] = baij->imax[i];
1068: MatSetOption(mat,MAT_NEW_NONZERO_LOCATION_ERR,PETSC_TRUE);
1069: return(0);
1070: }
1072: /*@
1073: MatSeqSBAIJSetColumnIndices - Set the column indices for all the rows
1074: in the matrix.
1076: Input Parameters:
1077: + mat - the SeqSBAIJ matrix
1078: - indices - the column indices
1080: Level: advanced
1082: Notes:
1083: This can be called if you have precomputed the nonzero structure of the
1084: matrix and want to provide it to the matrix object to improve the performance
1085: of the MatSetValues() operation.
1087: You MUST have set the correct numbers of nonzeros per row in the call to
1088: MatCreateSeqSBAIJ(), and the columns indices MUST be sorted.
1090: MUST be called before any calls to MatSetValues()
1092: .seealso: MatCreateSeqSBAIJ
1093: @*/
1094: PetscErrorCode MatSeqSBAIJSetColumnIndices(Mat mat,PetscInt *indices)
1095: {
1101: PetscUseMethod(mat,"MatSeqSBAIJSetColumnIndices_C",(Mat,PetscInt*),(mat,indices));
1102: return(0);
1103: }
1105: PetscErrorCode MatCopy_SeqSBAIJ(Mat A,Mat B,MatStructure str)
1106: {
1110: /* If the two matrices have the same copy implementation, use fast copy. */
1111: if (str == SAME_NONZERO_PATTERN && (A->ops->copy == B->ops->copy)) {
1112: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
1113: Mat_SeqSBAIJ *b = (Mat_SeqSBAIJ*)B->data;
1115: if (a->i[A->rmap->N] != b->i[B->rmap->N]) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_INCOMP,"Number of nonzeros in two matrices are different");
1116: PetscMemcpy(b->a,a->a,(a->i[A->rmap->N])*sizeof(PetscScalar));
1117: PetscObjectStateIncrease((PetscObject)B);
1118: } else {
1119: MatGetRowUpperTriangular(A);
1120: MatCopy_Basic(A,B,str);
1121: MatRestoreRowUpperTriangular(A);
1122: }
1123: return(0);
1124: }
1126: PetscErrorCode MatSetUp_SeqSBAIJ(Mat A)
1127: {
1131: MatSeqSBAIJSetPreallocation(A,A->rmap->bs,PETSC_DEFAULT,0);
1132: return(0);
1133: }
1135: PetscErrorCode MatSeqSBAIJGetArray_SeqSBAIJ(Mat A,PetscScalar *array[])
1136: {
1137: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
1140: *array = a->a;
1141: return(0);
1142: }
1144: PetscErrorCode MatSeqSBAIJRestoreArray_SeqSBAIJ(Mat A,PetscScalar *array[])
1145: {
1147: return(0);
1148: }
1150: PetscErrorCode MatAXPYGetPreallocation_SeqSBAIJ(Mat Y,Mat X,PetscInt *nnz)
1151: {
1152: PetscInt bs = Y->rmap->bs,mbs = Y->rmap->N/bs;
1153: Mat_SeqSBAIJ *x = (Mat_SeqSBAIJ*)X->data;
1154: Mat_SeqSBAIJ *y = (Mat_SeqSBAIJ*)Y->data;
1158: /* Set the number of nonzeros in the new matrix */
1159: MatAXPYGetPreallocation_SeqX_private(mbs,x->i,x->j,y->i,y->j,nnz);
1160: return(0);
1161: }
1163: PetscErrorCode MatAXPY_SeqSBAIJ(Mat Y,PetscScalar a,Mat X,MatStructure str)
1164: {
1165: Mat_SeqSBAIJ *x=(Mat_SeqSBAIJ*)X->data, *y=(Mat_SeqSBAIJ*)Y->data;
1167: PetscInt bs=Y->rmap->bs,bs2=bs*bs;
1168: PetscBLASInt one = 1;
1171: if (str == SAME_NONZERO_PATTERN) {
1172: PetscScalar alpha = a;
1173: PetscBLASInt bnz;
1174: PetscBLASIntCast(x->nz*bs2,&bnz);
1175: PetscStackCallBLAS("BLASaxpy",BLASaxpy_(&bnz,&alpha,x->a,&one,y->a,&one));
1176: PetscObjectStateIncrease((PetscObject)Y);
1177: } else if (str == SUBSET_NONZERO_PATTERN) { /* nonzeros of X is a subset of Y's */
1178: MatSetOption(X,MAT_GETROW_UPPERTRIANGULAR,PETSC_TRUE);
1179: MatAXPY_Basic(Y,a,X,str);
1180: MatSetOption(X,MAT_GETROW_UPPERTRIANGULAR,PETSC_FALSE);
1181: } else {
1182: Mat B;
1183: PetscInt *nnz;
1184: if (bs != X->rmap->bs) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Matrices must have same block size");
1185: MatGetRowUpperTriangular(X);
1186: MatGetRowUpperTriangular(Y);
1187: PetscMalloc1(Y->rmap->N,&nnz);
1188: MatCreate(PetscObjectComm((PetscObject)Y),&B);
1189: PetscObjectSetName((PetscObject)B,((PetscObject)Y)->name);
1190: MatSetSizes(B,Y->rmap->n,Y->cmap->n,Y->rmap->N,Y->cmap->N);
1191: MatSetBlockSizesFromMats(B,Y,Y);
1192: MatSetType(B,(MatType) ((PetscObject)Y)->type_name);
1193: MatAXPYGetPreallocation_SeqSBAIJ(Y,X,nnz);
1194: MatSeqSBAIJSetPreallocation(B,bs,0,nnz);
1196: MatAXPY_BasicWithPreallocation(B,Y,a,X,str);
1198: MatHeaderReplace(Y,&B);
1199: PetscFree(nnz);
1200: MatRestoreRowUpperTriangular(X);
1201: MatRestoreRowUpperTriangular(Y);
1202: }
1203: return(0);
1204: }
1206: PetscErrorCode MatIsSymmetric_SeqSBAIJ(Mat A,PetscReal tol,PetscBool *flg)
1207: {
1209: *flg = PETSC_TRUE;
1210: return(0);
1211: }
1213: PetscErrorCode MatIsStructurallySymmetric_SeqSBAIJ(Mat A,PetscBool *flg)
1214: {
1216: *flg = PETSC_TRUE;
1217: return(0);
1218: }
1220: PetscErrorCode MatIsHermitian_SeqSBAIJ(Mat A,PetscReal tol,PetscBool *flg)
1221: {
1223: *flg = PETSC_FALSE;
1224: return(0);
1225: }
1227: PetscErrorCode MatRealPart_SeqSBAIJ(Mat A)
1228: {
1229: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
1230: PetscInt i,nz = a->bs2*a->i[a->mbs];
1231: MatScalar *aa = a->a;
1234: for (i=0; i<nz; i++) aa[i] = PetscRealPart(aa[i]);
1235: return(0);
1236: }
1238: PetscErrorCode MatImaginaryPart_SeqSBAIJ(Mat A)
1239: {
1240: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
1241: PetscInt i,nz = a->bs2*a->i[a->mbs];
1242: MatScalar *aa = a->a;
1245: for (i=0; i<nz; i++) aa[i] = PetscImaginaryPart(aa[i]);
1246: return(0);
1247: }
1249: PetscErrorCode MatZeroRowsColumns_SeqSBAIJ(Mat A,PetscInt is_n,const PetscInt is_idx[],PetscScalar diag,Vec x, Vec b)
1250: {
1251: Mat_SeqSBAIJ *baij=(Mat_SeqSBAIJ*)A->data;
1252: PetscErrorCode ierr;
1253: PetscInt i,j,k,count;
1254: PetscInt bs =A->rmap->bs,bs2=baij->bs2,row,col;
1255: PetscScalar zero = 0.0;
1256: MatScalar *aa;
1257: const PetscScalar *xx;
1258: PetscScalar *bb;
1259: PetscBool *zeroed,vecs = PETSC_FALSE;
1262: /* fix right hand side if needed */
1263: if (x && b) {
1264: VecGetArrayRead(x,&xx);
1265: VecGetArray(b,&bb);
1266: vecs = PETSC_TRUE;
1267: }
1269: /* zero the columns */
1270: PetscCalloc1(A->rmap->n,&zeroed);
1271: for (i=0; i<is_n; i++) {
1272: if (is_idx[i] < 0 || is_idx[i] >= A->rmap->N) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"row %D out of range",is_idx[i]);
1273: zeroed[is_idx[i]] = PETSC_TRUE;
1274: }
1275: if (vecs) {
1276: for (i=0; i<A->rmap->N; i++) {
1277: row = i/bs;
1278: for (j=baij->i[row]; j<baij->i[row+1]; j++) {
1279: for (k=0; k<bs; k++) {
1280: col = bs*baij->j[j] + k;
1281: if (col <= i) continue;
1282: aa = ((MatScalar*)(baij->a)) + j*bs2 + (i%bs) + bs*k;
1283: if (!zeroed[i] && zeroed[col]) bb[i] -= aa[0]*xx[col];
1284: if (zeroed[i] && !zeroed[col]) bb[col] -= aa[0]*xx[i];
1285: }
1286: }
1287: }
1288: for (i=0; i<is_n; i++) bb[is_idx[i]] = diag*xx[is_idx[i]];
1289: }
1291: for (i=0; i<A->rmap->N; i++) {
1292: if (!zeroed[i]) {
1293: row = i/bs;
1294: for (j=baij->i[row]; j<baij->i[row+1]; j++) {
1295: for (k=0; k<bs; k++) {
1296: col = bs*baij->j[j] + k;
1297: if (zeroed[col]) {
1298: aa = ((MatScalar*)(baij->a)) + j*bs2 + (i%bs) + bs*k;
1299: aa[0] = 0.0;
1300: }
1301: }
1302: }
1303: }
1304: }
1305: PetscFree(zeroed);
1306: if (vecs) {
1307: VecRestoreArrayRead(x,&xx);
1308: VecRestoreArray(b,&bb);
1309: }
1311: /* zero the rows */
1312: for (i=0; i<is_n; i++) {
1313: row = is_idx[i];
1314: count = (baij->i[row/bs +1] - baij->i[row/bs])*bs;
1315: aa = ((MatScalar*)(baij->a)) + baij->i[row/bs]*bs2 + (row%bs);
1316: for (k=0; k<count; k++) {
1317: aa[0] = zero;
1318: aa += bs;
1319: }
1320: if (diag != 0.0) {
1321: (*A->ops->setvalues)(A,1,&row,1,&row,&diag,INSERT_VALUES);
1322: }
1323: }
1324: MatAssemblyEnd_SeqSBAIJ(A,MAT_FINAL_ASSEMBLY);
1325: return(0);
1326: }
1328: PetscErrorCode MatShift_SeqSBAIJ(Mat Y,PetscScalar a)
1329: {
1331: Mat_SeqSBAIJ *aij = (Mat_SeqSBAIJ*)Y->data;
1334: if (!Y->preallocated || !aij->nz) {
1335: MatSeqSBAIJSetPreallocation(Y,Y->rmap->bs,1,NULL);
1336: }
1337: MatShift_Basic(Y,a);
1338: return(0);
1339: }
1341: /* -------------------------------------------------------------------*/
1342: static struct _MatOps MatOps_Values = {MatSetValues_SeqSBAIJ,
1343: MatGetRow_SeqSBAIJ,
1344: MatRestoreRow_SeqSBAIJ,
1345: MatMult_SeqSBAIJ_N,
1346: /* 4*/ MatMultAdd_SeqSBAIJ_N,
1347: MatMult_SeqSBAIJ_N, /* transpose versions are same as non-transpose versions */
1348: MatMultAdd_SeqSBAIJ_N,
1349: 0,
1350: 0,
1351: 0,
1352: /* 10*/ 0,
1353: 0,
1354: MatCholeskyFactor_SeqSBAIJ,
1355: MatSOR_SeqSBAIJ,
1356: MatTranspose_SeqSBAIJ,
1357: /* 15*/ MatGetInfo_SeqSBAIJ,
1358: MatEqual_SeqSBAIJ,
1359: MatGetDiagonal_SeqSBAIJ,
1360: MatDiagonalScale_SeqSBAIJ,
1361: MatNorm_SeqSBAIJ,
1362: /* 20*/ 0,
1363: MatAssemblyEnd_SeqSBAIJ,
1364: MatSetOption_SeqSBAIJ,
1365: MatZeroEntries_SeqSBAIJ,
1366: /* 24*/ 0,
1367: 0,
1368: 0,
1369: 0,
1370: 0,
1371: /* 29*/ MatSetUp_SeqSBAIJ,
1372: 0,
1373: 0,
1374: 0,
1375: 0,
1376: /* 34*/ MatDuplicate_SeqSBAIJ,
1377: 0,
1378: 0,
1379: 0,
1380: MatICCFactor_SeqSBAIJ,
1381: /* 39*/ MatAXPY_SeqSBAIJ,
1382: MatCreateSubMatrices_SeqSBAIJ,
1383: MatIncreaseOverlap_SeqSBAIJ,
1384: MatGetValues_SeqSBAIJ,
1385: MatCopy_SeqSBAIJ,
1386: /* 44*/ 0,
1387: MatScale_SeqSBAIJ,
1388: MatShift_SeqSBAIJ,
1389: 0,
1390: MatZeroRowsColumns_SeqSBAIJ,
1391: /* 49*/ 0,
1392: MatGetRowIJ_SeqSBAIJ,
1393: MatRestoreRowIJ_SeqSBAIJ,
1394: 0,
1395: 0,
1396: /* 54*/ 0,
1397: 0,
1398: 0,
1399: 0,
1400: MatSetValuesBlocked_SeqSBAIJ,
1401: /* 59*/ MatCreateSubMatrix_SeqSBAIJ,
1402: 0,
1403: 0,
1404: 0,
1405: 0,
1406: /* 64*/ 0,
1407: 0,
1408: 0,
1409: 0,
1410: 0,
1411: /* 69*/ MatGetRowMaxAbs_SeqSBAIJ,
1412: 0,
1413: 0,
1414: 0,
1415: 0,
1416: /* 74*/ 0,
1417: 0,
1418: 0,
1419: 0,
1420: 0,
1421: /* 79*/ 0,
1422: 0,
1423: 0,
1424: MatGetInertia_SeqSBAIJ,
1425: MatLoad_SeqSBAIJ,
1426: /* 84*/ MatIsSymmetric_SeqSBAIJ,
1427: MatIsHermitian_SeqSBAIJ,
1428: MatIsStructurallySymmetric_SeqSBAIJ,
1429: 0,
1430: 0,
1431: /* 89*/ 0,
1432: 0,
1433: 0,
1434: 0,
1435: 0,
1436: /* 94*/ 0,
1437: 0,
1438: 0,
1439: 0,
1440: 0,
1441: /* 99*/ 0,
1442: 0,
1443: 0,
1444: 0,
1445: 0,
1446: /*104*/ 0,
1447: MatRealPart_SeqSBAIJ,
1448: MatImaginaryPart_SeqSBAIJ,
1449: MatGetRowUpperTriangular_SeqSBAIJ,
1450: MatRestoreRowUpperTriangular_SeqSBAIJ,
1451: /*109*/ 0,
1452: 0,
1453: 0,
1454: 0,
1455: MatMissingDiagonal_SeqSBAIJ,
1456: /*114*/ 0,
1457: 0,
1458: 0,
1459: 0,
1460: 0,
1461: /*119*/ 0,
1462: 0,
1463: 0,
1464: 0,
1465: 0,
1466: /*124*/ 0,
1467: 0,
1468: 0,
1469: 0,
1470: 0,
1471: /*129*/ 0,
1472: 0,
1473: 0,
1474: 0,
1475: 0,
1476: /*134*/ 0,
1477: 0,
1478: 0,
1479: 0,
1480: 0,
1481: /*139*/ MatSetBlockSizes_Default,
1482: 0,
1483: 0,
1484: 0,
1485: 0,
1486: /*144*/MatCreateMPIMatConcatenateSeqMat_SeqSBAIJ
1487: };
1489: PetscErrorCode MatStoreValues_SeqSBAIJ(Mat mat)
1490: {
1491: Mat_SeqSBAIJ *aij = (Mat_SeqSBAIJ*)mat->data;
1492: PetscInt nz = aij->i[mat->rmap->N]*mat->rmap->bs*aij->bs2;
1496: if (aij->nonew != 1) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Must call MatSetOption(A,MAT_NEW_NONZERO_LOCATIONS,PETSC_FALSE);first");
1498: /* allocate space for values if not already there */
1499: if (!aij->saved_values) {
1500: PetscMalloc1(nz+1,&aij->saved_values);
1501: }
1503: /* copy values over */
1504: PetscMemcpy(aij->saved_values,aij->a,nz*sizeof(PetscScalar));
1505: return(0);
1506: }
1508: PetscErrorCode MatRetrieveValues_SeqSBAIJ(Mat mat)
1509: {
1510: Mat_SeqSBAIJ *aij = (Mat_SeqSBAIJ*)mat->data;
1512: PetscInt nz = aij->i[mat->rmap->N]*mat->rmap->bs*aij->bs2;
1515: if (aij->nonew != 1) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Must call MatSetOption(A,MAT_NEW_NONZERO_LOCATIONS,PETSC_FALSE);first");
1516: if (!aij->saved_values) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Must call MatStoreValues(A);first");
1518: /* copy values over */
1519: PetscMemcpy(aij->a,aij->saved_values,nz*sizeof(PetscScalar));
1520: return(0);
1521: }
1523: static PetscErrorCode MatSeqSBAIJSetPreallocation_SeqSBAIJ(Mat B,PetscInt bs,PetscInt nz,PetscInt *nnz)
1524: {
1525: Mat_SeqSBAIJ *b = (Mat_SeqSBAIJ*)B->data;
1527: PetscInt i,mbs,nbs,bs2;
1528: PetscBool skipallocation = PETSC_FALSE,flg = PETSC_FALSE,realalloc = PETSC_FALSE;
1531: if (nz >= 0 || nnz) realalloc = PETSC_TRUE;
1533: MatSetBlockSize(B,PetscAbs(bs));
1534: PetscLayoutSetUp(B->rmap);
1535: PetscLayoutSetUp(B->cmap);
1536: PetscLayoutGetBlockSize(B->rmap,&bs);
1538: B->preallocated = PETSC_TRUE;
1540: mbs = B->rmap->N/bs;
1541: nbs = B->cmap->n/bs;
1542: bs2 = bs*bs;
1544: if (mbs*bs != B->rmap->N || nbs*bs!=B->cmap->n) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Number rows, cols must be divisible by blocksize");
1546: if (nz == MAT_SKIP_ALLOCATION) {
1547: skipallocation = PETSC_TRUE;
1548: nz = 0;
1549: }
1551: if (nz == PETSC_DEFAULT || nz == PETSC_DECIDE) nz = 3;
1552: if (nz < 0) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"nz cannot be less than 0: value %D",nz);
1553: if (nnz) {
1554: for (i=0; i<mbs; i++) {
1555: if (nnz[i] < 0) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"nnz cannot be less than 0: local row %D value %D",i,nnz[i]);
1556: if (nnz[i] > nbs) SETERRQ3(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"nnz cannot be greater than block row length: local row %D value %D block rowlength %D",i,nnz[i],nbs);
1557: }
1558: }
1560: B->ops->mult = MatMult_SeqSBAIJ_N;
1561: B->ops->multadd = MatMultAdd_SeqSBAIJ_N;
1562: B->ops->multtranspose = MatMult_SeqSBAIJ_N;
1563: B->ops->multtransposeadd = MatMultAdd_SeqSBAIJ_N;
1565: PetscOptionsGetBool(((PetscObject)B)->options,((PetscObject)B)->prefix,"-mat_no_unroll",&flg,NULL);
1566: if (!flg) {
1567: switch (bs) {
1568: case 1:
1569: B->ops->mult = MatMult_SeqSBAIJ_1;
1570: B->ops->multadd = MatMultAdd_SeqSBAIJ_1;
1571: B->ops->multtranspose = MatMult_SeqSBAIJ_1;
1572: B->ops->multtransposeadd = MatMultAdd_SeqSBAIJ_1;
1573: break;
1574: case 2:
1575: B->ops->mult = MatMult_SeqSBAIJ_2;
1576: B->ops->multadd = MatMultAdd_SeqSBAIJ_2;
1577: B->ops->multtranspose = MatMult_SeqSBAIJ_2;
1578: B->ops->multtransposeadd = MatMultAdd_SeqSBAIJ_2;
1579: break;
1580: case 3:
1581: B->ops->mult = MatMult_SeqSBAIJ_3;
1582: B->ops->multadd = MatMultAdd_SeqSBAIJ_3;
1583: B->ops->multtranspose = MatMult_SeqSBAIJ_3;
1584: B->ops->multtransposeadd = MatMultAdd_SeqSBAIJ_3;
1585: break;
1586: case 4:
1587: B->ops->mult = MatMult_SeqSBAIJ_4;
1588: B->ops->multadd = MatMultAdd_SeqSBAIJ_4;
1589: B->ops->multtranspose = MatMult_SeqSBAIJ_4;
1590: B->ops->multtransposeadd = MatMultAdd_SeqSBAIJ_4;
1591: break;
1592: case 5:
1593: B->ops->mult = MatMult_SeqSBAIJ_5;
1594: B->ops->multadd = MatMultAdd_SeqSBAIJ_5;
1595: B->ops->multtranspose = MatMult_SeqSBAIJ_5;
1596: B->ops->multtransposeadd = MatMultAdd_SeqSBAIJ_5;
1597: break;
1598: case 6:
1599: B->ops->mult = MatMult_SeqSBAIJ_6;
1600: B->ops->multadd = MatMultAdd_SeqSBAIJ_6;
1601: B->ops->multtranspose = MatMult_SeqSBAIJ_6;
1602: B->ops->multtransposeadd = MatMultAdd_SeqSBAIJ_6;
1603: break;
1604: case 7:
1605: B->ops->mult = MatMult_SeqSBAIJ_7;
1606: B->ops->multadd = MatMultAdd_SeqSBAIJ_7;
1607: B->ops->multtranspose = MatMult_SeqSBAIJ_7;
1608: B->ops->multtransposeadd = MatMultAdd_SeqSBAIJ_7;
1609: break;
1610: }
1611: }
1613: b->mbs = mbs;
1614: b->nbs = nbs;
1615: if (!skipallocation) {
1616: if (!b->imax) {
1617: PetscMalloc2(mbs,&b->imax,mbs,&b->ilen);
1619: b->free_imax_ilen = PETSC_TRUE;
1621: PetscLogObjectMemory((PetscObject)B,2*mbs*sizeof(PetscInt));
1622: }
1623: if (!nnz) {
1624: if (nz == PETSC_DEFAULT || nz == PETSC_DECIDE) nz = 5;
1625: else if (nz <= 0) nz = 1;
1626: for (i=0; i<mbs; i++) b->imax[i] = nz;
1627: nz = nz*mbs; /* total nz */
1628: } else {
1629: nz = 0;
1630: for (i=0; i<mbs; i++) {b->imax[i] = nnz[i]; nz += nnz[i];}
1631: }
1632: /* b->ilen will count nonzeros in each block row so far. */
1633: for (i=0; i<mbs; i++) b->ilen[i] = 0;
1634: /* nz=(nz+mbs)/2; */ /* total diagonal and superdiagonal nonzero blocks */
1636: /* allocate the matrix space */
1637: MatSeqXAIJFreeAIJ(B,&b->a,&b->j,&b->i);
1638: PetscMalloc3(bs2*nz,&b->a,nz,&b->j,B->rmap->N+1,&b->i);
1639: PetscLogObjectMemory((PetscObject)B,(B->rmap->N+1)*sizeof(PetscInt)+nz*(bs2*sizeof(PetscScalar)+sizeof(PetscInt)));
1640: PetscMemzero(b->a,nz*bs2*sizeof(MatScalar));
1641: PetscMemzero(b->j,nz*sizeof(PetscInt));
1643: b->singlemalloc = PETSC_TRUE;
1645: /* pointer to beginning of each row */
1646: b->i[0] = 0;
1647: for (i=1; i<mbs+1; i++) b->i[i] = b->i[i-1] + b->imax[i-1];
1649: b->free_a = PETSC_TRUE;
1650: b->free_ij = PETSC_TRUE;
1651: } else {
1652: b->free_a = PETSC_FALSE;
1653: b->free_ij = PETSC_FALSE;
1654: }
1656: b->bs2 = bs2;
1657: b->nz = 0;
1658: b->maxnz = nz;
1659: b->inew = 0;
1660: b->jnew = 0;
1661: b->anew = 0;
1662: b->a2anew = 0;
1663: b->permute = PETSC_FALSE;
1665: B->was_assembled = PETSC_FALSE;
1666: B->assembled = PETSC_FALSE;
1667: if (realalloc) {MatSetOption(B,MAT_NEW_NONZERO_ALLOCATION_ERR,PETSC_TRUE);}
1668: return(0);
1669: }
1671: PetscErrorCode MatSeqSBAIJSetPreallocationCSR_SeqSBAIJ(Mat B,PetscInt bs,const PetscInt ii[],const PetscInt jj[], const PetscScalar V[])
1672: {
1673: PetscInt i,j,m,nz,nz_max=0,*nnz;
1674: PetscScalar *values=0;
1675: PetscBool roworiented = ((Mat_SeqSBAIJ*)B->data)->roworiented;
1678: if (bs < 1) SETERRQ1(PetscObjectComm((PetscObject)B),PETSC_ERR_ARG_OUTOFRANGE,"Invalid block size specified, must be positive but it is %D",bs);
1679: PetscLayoutSetBlockSize(B->rmap,bs);
1680: PetscLayoutSetBlockSize(B->cmap,bs);
1681: PetscLayoutSetUp(B->rmap);
1682: PetscLayoutSetUp(B->cmap);
1683: PetscLayoutGetBlockSize(B->rmap,&bs);
1684: m = B->rmap->n/bs;
1686: if (ii[0]) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"ii[0] must be 0 but it is %D",ii[0]);
1687: PetscMalloc1(m+1,&nnz);
1688: for (i=0; i<m; i++) {
1689: nz = ii[i+1] - ii[i];
1690: if (nz < 0) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Row %D has a negative number of columns %D",i,nz);
1691: nz_max = PetscMax(nz_max,nz);
1692: nnz[i] = nz;
1693: }
1694: MatSeqSBAIJSetPreallocation(B,bs,0,nnz);
1695: PetscFree(nnz);
1697: values = (PetscScalar*)V;
1698: if (!values) {
1699: PetscCalloc1(bs*bs*nz_max,&values);
1700: }
1701: for (i=0; i<m; i++) {
1702: PetscInt ncols = ii[i+1] - ii[i];
1703: const PetscInt *icols = jj + ii[i];
1704: if (!roworiented || bs == 1) {
1705: const PetscScalar *svals = values + (V ? (bs*bs*ii[i]) : 0);
1706: MatSetValuesBlocked_SeqSBAIJ(B,1,&i,ncols,icols,svals,INSERT_VALUES);
1707: } else {
1708: for (j=0; j<ncols; j++) {
1709: const PetscScalar *svals = values + (V ? (bs*bs*(ii[i]+j)) : 0);
1710: MatSetValuesBlocked_SeqSBAIJ(B,1,&i,1,&icols[j],svals,INSERT_VALUES);
1711: }
1712: }
1713: }
1714: if (!V) { PetscFree(values); }
1715: MatAssemblyBegin(B,MAT_FINAL_ASSEMBLY);
1716: MatAssemblyEnd(B,MAT_FINAL_ASSEMBLY);
1717: MatSetOption(B,MAT_NEW_NONZERO_LOCATION_ERR,PETSC_TRUE);
1718: return(0);
1719: }
1721: /*
1722: This is used to set the numeric factorization for both Cholesky and ICC symbolic factorization
1723: */
1724: PetscErrorCode MatSeqSBAIJSetNumericFactorization_inplace(Mat B,PetscBool natural)
1725: {
1727: PetscBool flg = PETSC_FALSE;
1728: PetscInt bs = B->rmap->bs;
1731: PetscOptionsGetBool(((PetscObject)B)->options,((PetscObject)B)->prefix,"-mat_no_unroll",&flg,NULL);
1732: if (flg) bs = 8;
1734: if (!natural) {
1735: switch (bs) {
1736: case 1:
1737: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_1_inplace;
1738: break;
1739: case 2:
1740: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_2;
1741: break;
1742: case 3:
1743: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_3;
1744: break;
1745: case 4:
1746: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_4;
1747: break;
1748: case 5:
1749: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_5;
1750: break;
1751: case 6:
1752: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_6;
1753: break;
1754: case 7:
1755: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_7;
1756: break;
1757: default:
1758: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_N;
1759: break;
1760: }
1761: } else {
1762: switch (bs) {
1763: case 1:
1764: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_1_NaturalOrdering_inplace;
1765: break;
1766: case 2:
1767: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_2_NaturalOrdering;
1768: break;
1769: case 3:
1770: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_3_NaturalOrdering;
1771: break;
1772: case 4:
1773: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_4_NaturalOrdering;
1774: break;
1775: case 5:
1776: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_5_NaturalOrdering;
1777: break;
1778: case 6:
1779: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_6_NaturalOrdering;
1780: break;
1781: case 7:
1782: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_7_NaturalOrdering;
1783: break;
1784: default:
1785: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_N_NaturalOrdering;
1786: break;
1787: }
1788: }
1789: return(0);
1790: }
1792: PETSC_INTERN PetscErrorCode MatConvert_SeqSBAIJ_SeqAIJ(Mat, MatType,MatReuse,Mat*);
1793: PETSC_INTERN PetscErrorCode MatConvert_SeqSBAIJ_SeqBAIJ(Mat, MatType,MatReuse,Mat*);
1795: PETSC_INTERN PetscErrorCode MatGetFactor_seqsbaij_petsc(Mat A,MatFactorType ftype,Mat *B)
1796: {
1797: PetscInt n = A->rmap->n;
1801: #if defined(PETSC_USE_COMPLEX)
1802: if (A->hermitian) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Hermitian Factor is not supported");
1803: #endif
1804: MatCreate(PetscObjectComm((PetscObject)A),B);
1805: MatSetSizes(*B,n,n,n,n);
1806: if (ftype == MAT_FACTOR_CHOLESKY || ftype == MAT_FACTOR_ICC) {
1807: MatSetType(*B,MATSEQSBAIJ);
1808: MatSeqSBAIJSetPreallocation(*B,A->rmap->bs,MAT_SKIP_ALLOCATION,NULL);
1810: (*B)->ops->choleskyfactorsymbolic = MatCholeskyFactorSymbolic_SeqSBAIJ;
1811: (*B)->ops->iccfactorsymbolic = MatICCFactorSymbolic_SeqSBAIJ;
1812: } else SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Factor type not supported");
1814: (*B)->factortype = ftype;
1815: PetscFree((*B)->solvertype);
1816: PetscStrallocpy(MATSOLVERPETSC,&(*B)->solvertype);
1817: return(0);
1818: }
1820: /*@C
1821: MatSeqSBAIJGetArray - gives access to the array where the data for a MATSEQSBAIJ matrix is stored
1823: Not Collective
1825: Input Parameter:
1826: . mat - a MATSEQSBAIJ matrix
1828: Output Parameter:
1829: . array - pointer to the data
1831: Level: intermediate
1833: .seealso: MatSeqSBAIJRestoreArray(), MatSeqAIJGetArray(), MatSeqAIJRestoreArray()
1834: @*/
1835: PetscErrorCode MatSeqSBAIJGetArray(Mat A,PetscScalar **array)
1836: {
1840: PetscUseMethod(A,"MatSeqSBAIJGetArray_C",(Mat,PetscScalar**),(A,array));
1841: return(0);
1842: }
1844: /*@C
1845: MatSeqSBAIJRestoreArray - returns access to the array where the data for a MATSEQSBAIJ matrix is stored obtained by MatSeqSBAIJGetArray()
1847: Not Collective
1849: Input Parameters:
1850: . mat - a MATSEQSBAIJ matrix
1851: . array - pointer to the data
1853: Level: intermediate
1855: .seealso: MatSeqSBAIJGetArray(), MatSeqAIJGetArray(), MatSeqAIJRestoreArray()
1856: @*/
1857: PetscErrorCode MatSeqSBAIJRestoreArray(Mat A,PetscScalar **array)
1858: {
1862: PetscUseMethod(A,"MatSeqSBAIJRestoreArray_C",(Mat,PetscScalar**),(A,array));
1863: return(0);
1864: }
1866: /*MC
1867: MATSEQSBAIJ - MATSEQSBAIJ = "seqsbaij" - A matrix type to be used for sequential symmetric block sparse matrices,
1868: based on block compressed sparse row format. Only the upper triangular portion of the matrix is stored.
1870: For complex numbers by default this matrix is symmetric, NOT Hermitian symmetric. To make it Hermitian symmetric you
1871: can call MatSetOption(Mat, MAT_HERMITIAN); after MatAssemblyEnd()
1873: Options Database Keys:
1874: . -mat_type seqsbaij - sets the matrix type to "seqsbaij" during a call to MatSetFromOptions()
1876: Notes: By default if you insert values into the lower triangular part of the matrix they are simply ignored (since they are not
1877: stored and it is assumed they symmetric to the upper triangular). If you call MatSetOption(Mat,MAT_IGNORE_LOWER_TRIANGULAR,PETSC_FALSE) or use
1878: the options database -mat_ignore_lower_triangular false it will generate an error if you try to set a value in the lower triangular portion.
1881: Level: beginner
1883: .seealso: MatCreateSeqSBAIJ
1884: M*/
1886: PETSC_INTERN PetscErrorCode MatConvert_SeqSBAIJ_SeqSBSTRM(Mat, MatType,MatReuse,Mat*);
1888: PETSC_EXTERN PetscErrorCode MatCreate_SeqSBAIJ(Mat B)
1889: {
1890: Mat_SeqSBAIJ *b;
1892: PetscMPIInt size;
1893: PetscBool no_unroll = PETSC_FALSE,no_inode = PETSC_FALSE;
1896: MPI_Comm_size(PetscObjectComm((PetscObject)B),&size);
1897: if (size > 1) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"Comm must be of size 1");
1899: PetscNewLog(B,&b);
1900: B->data = (void*)b;
1901: PetscMemcpy(B->ops,&MatOps_Values,sizeof(struct _MatOps));
1903: B->ops->destroy = MatDestroy_SeqSBAIJ;
1904: B->ops->view = MatView_SeqSBAIJ;
1905: b->row = 0;
1906: b->icol = 0;
1907: b->reallocs = 0;
1908: b->saved_values = 0;
1909: b->inode.limit = 5;
1910: b->inode.max_limit = 5;
1912: b->roworiented = PETSC_TRUE;
1913: b->nonew = 0;
1914: b->diag = 0;
1915: b->solve_work = 0;
1916: b->mult_work = 0;
1917: B->spptr = 0;
1918: B->info.nz_unneeded = (PetscReal)b->maxnz*b->bs2;
1919: b->keepnonzeropattern = PETSC_FALSE;
1921: b->inew = 0;
1922: b->jnew = 0;
1923: b->anew = 0;
1924: b->a2anew = 0;
1925: b->permute = PETSC_FALSE;
1927: b->ignore_ltriangular = PETSC_TRUE;
1929: PetscOptionsGetBool(((PetscObject)B)->options,((PetscObject)B)->prefix,"-mat_ignore_lower_triangular",&b->ignore_ltriangular,NULL);
1931: b->getrow_utriangular = PETSC_FALSE;
1933: PetscOptionsGetBool(((PetscObject)B)->options,((PetscObject)B)->prefix,"-mat_getrow_uppertriangular",&b->getrow_utriangular,NULL);
1935: PetscObjectComposeFunction((PetscObject)B,"MatSeqSBAIJGetArray_C",MatSeqSBAIJGetArray_SeqSBAIJ);
1936: PetscObjectComposeFunction((PetscObject)B,"MatSeqSBAIJRestoreArray_C",MatSeqSBAIJRestoreArray_SeqSBAIJ);
1937: PetscObjectComposeFunction((PetscObject)B,"MatStoreValues_C",MatStoreValues_SeqSBAIJ);
1938: PetscObjectComposeFunction((PetscObject)B,"MatRetrieveValues_C",MatRetrieveValues_SeqSBAIJ);
1939: PetscObjectComposeFunction((PetscObject)B,"MatSeqSBAIJSetColumnIndices_C",MatSeqSBAIJSetColumnIndices_SeqSBAIJ);
1940: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqsbaij_seqaij_C",MatConvert_SeqSBAIJ_SeqAIJ);
1941: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqsbaij_seqbaij_C",MatConvert_SeqSBAIJ_SeqBAIJ);
1942: PetscObjectComposeFunction((PetscObject)B,"MatSeqSBAIJSetPreallocation_C",MatSeqSBAIJSetPreallocation_SeqSBAIJ);
1943: PetscObjectComposeFunction((PetscObject)B,"MatSeqSBAIJSetPreallocationCSR_C",MatSeqSBAIJSetPreallocationCSR_SeqSBAIJ);
1944: #if defined(PETSC_HAVE_ELEMENTAL)
1945: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqsbaij_elemental_C",MatConvert_SeqSBAIJ_Elemental);
1946: #endif
1948: B->symmetric = PETSC_TRUE;
1949: B->structurally_symmetric = PETSC_TRUE;
1950: B->symmetric_set = PETSC_TRUE;
1951: B->structurally_symmetric_set = PETSC_TRUE;
1953: PetscObjectChangeTypeName((PetscObject)B,MATSEQSBAIJ);
1955: PetscOptionsBegin(PetscObjectComm((PetscObject)B),((PetscObject)B)->prefix,"Options for SEQSBAIJ matrix","Mat");
1956: PetscOptionsBool("-mat_no_unroll","Do not optimize for inodes (slower)",NULL,no_unroll,&no_unroll,NULL);
1957: if (no_unroll) {
1958: PetscInfo(B,"Not using Inode routines due to -mat_no_unroll\n");
1959: }
1960: PetscOptionsBool("-mat_no_inode","Do not optimize for inodes (slower)",NULL,no_inode,&no_inode,NULL);
1961: if (no_inode) {
1962: PetscInfo(B,"Not using Inode routines due to -mat_no_inode\n");
1963: }
1964: PetscOptionsInt("-mat_inode_limit","Do not use inodes larger then this value",NULL,b->inode.limit,&b->inode.limit,NULL);
1965: PetscOptionsEnd();
1966: b->inode.use = (PetscBool)(!(no_unroll || no_inode));
1967: if (b->inode.limit > b->inode.max_limit) b->inode.limit = b->inode.max_limit;
1968: return(0);
1969: }
1971: /*@C
1972: MatSeqSBAIJSetPreallocation - Creates a sparse symmetric matrix in block AIJ (block
1973: compressed row) format. For good matrix assembly performance the
1974: user should preallocate the matrix storage by setting the parameter nz
1975: (or the array nnz). By setting these parameters accurately, performance
1976: during matrix assembly can be increased by more than a factor of 50.
1978: Collective on Mat
1980: Input Parameters:
1981: + B - the symmetric matrix
1982: . bs - size of block, the blocks are ALWAYS square. One can use MatSetBlockSizes() to set a different row and column blocksize but the row
1983: blocksize always defines the size of the blocks. The column blocksize sets the blocksize of the vectors obtained with MatCreateVecs()
1984: . nz - number of block nonzeros per block row (same for all rows)
1985: - nnz - array containing the number of block nonzeros in the upper triangular plus
1986: diagonal portion of each block (possibly different for each block row) or NULL
1988: Options Database Keys:
1989: . -mat_no_unroll - uses code that does not unroll the loops in the
1990: block calculations (much slower)
1991: . -mat_block_size - size of the blocks to use (only works if a negative bs is passed in
1993: Level: intermediate
1995: Notes:
1996: Specify the preallocated storage with either nz or nnz (not both).
1997: Set nz=PETSC_DEFAULT and nnz=NULL for PETSc to control dynamic memory
1998: allocation. See Users-Manual: ch_mat for details.
2000: You can call MatGetInfo() to get information on how effective the preallocation was;
2001: for example the fields mallocs,nz_allocated,nz_used,nz_unneeded;
2002: You can also run with the option -info and look for messages with the string
2003: malloc in them to see if additional memory allocation was needed.
2005: If the nnz parameter is given then the nz parameter is ignored
2008: .seealso: MatCreate(), MatCreateSeqAIJ(), MatSetValues(), MatCreateSBAIJ()
2009: @*/
2010: PetscErrorCode MatSeqSBAIJSetPreallocation(Mat B,PetscInt bs,PetscInt nz,const PetscInt nnz[])
2011: {
2018: PetscTryMethod(B,"MatSeqSBAIJSetPreallocation_C",(Mat,PetscInt,PetscInt,const PetscInt[]),(B,bs,nz,nnz));
2019: return(0);
2020: }
2022: /*@C
2023: MatSeqSBAIJSetPreallocationCSR - Allocates memory for a sparse sequential matrix in symmetric block AIJ format.
2025: Input Parameters:
2026: + B - the matrix
2027: . bs - size of block, the blocks are ALWAYS square.
2028: . i - the indices into j for the start of each local row (starts with zero)
2029: . j - the column indices for each local row (starts with zero) these must be sorted for each row
2030: - v - optional values in the matrix
2032: Level: developer
2034: Notes:
2035: The order of the entries in values is specified by the MatOption MAT_ROW_ORIENTED. For example, C programs
2036: may want to use the default MAT_ROW_ORIENTED=PETSC_TRUE and use an array v[nnz][bs][bs] where the second index is
2037: over rows within a block and the last index is over columns within a block row. Fortran programs will likely set
2038: MAT_ROW_ORIENTED=PETSC_FALSE and use a Fortran array v(bs,bs,nnz) in which the first index is over rows within a
2039: block column and the second index is over columns within a block.
2041: .keywords: matrix, block, aij, compressed row, sparse
2043: .seealso: MatCreate(), MatCreateSeqSBAIJ(), MatSetValuesBlocked(), MatSeqSBAIJSetPreallocation(), MATSEQSBAIJ
2044: @*/
2045: PetscErrorCode MatSeqSBAIJSetPreallocationCSR(Mat B,PetscInt bs,const PetscInt i[],const PetscInt j[], const PetscScalar v[])
2046: {
2053: PetscTryMethod(B,"MatSeqSBAIJSetPreallocationCSR_C",(Mat,PetscInt,const PetscInt[],const PetscInt[],const PetscScalar[]),(B,bs,i,j,v));
2054: return(0);
2055: }
2057: /*@C
2058: MatCreateSeqSBAIJ - Creates a sparse symmetric matrix in block AIJ (block
2059: compressed row) format. For good matrix assembly performance the
2060: user should preallocate the matrix storage by setting the parameter nz
2061: (or the array nnz). By setting these parameters accurately, performance
2062: during matrix assembly can be increased by more than a factor of 50.
2064: Collective on MPI_Comm
2066: Input Parameters:
2067: + comm - MPI communicator, set to PETSC_COMM_SELF
2068: . bs - size of block, the blocks are ALWAYS square. One can use MatSetBlockSizes() to set a different row and column blocksize but the row
2069: blocksize always defines the size of the blocks. The column blocksize sets the blocksize of the vectors obtained with MatCreateVecs()
2070: . m - number of rows, or number of columns
2071: . nz - number of block nonzeros per block row (same for all rows)
2072: - nnz - array containing the number of block nonzeros in the upper triangular plus
2073: diagonal portion of each block (possibly different for each block row) or NULL
2075: Output Parameter:
2076: . A - the symmetric matrix
2078: Options Database Keys:
2079: . -mat_no_unroll - uses code that does not unroll the loops in the
2080: block calculations (much slower)
2081: . -mat_block_size - size of the blocks to use
2083: Level: intermediate
2085: It is recommended that one use the MatCreate(), MatSetType() and/or MatSetFromOptions(),
2086: MatXXXXSetPreallocation() paradgm instead of this routine directly.
2087: [MatXXXXSetPreallocation() is, for example, MatSeqAIJSetPreallocation]
2089: Notes:
2090: The number of rows and columns must be divisible by blocksize.
2091: This matrix type does not support complex Hermitian operation.
2093: Specify the preallocated storage with either nz or nnz (not both).
2094: Set nz=PETSC_DEFAULT and nnz=NULL for PETSc to control dynamic memory
2095: allocation. See Users-Manual: ch_mat for details.
2097: If the nnz parameter is given then the nz parameter is ignored
2099: .seealso: MatCreate(), MatCreateSeqAIJ(), MatSetValues(), MatCreateSBAIJ()
2100: @*/
2101: PetscErrorCode MatCreateSeqSBAIJ(MPI_Comm comm,PetscInt bs,PetscInt m,PetscInt n,PetscInt nz,const PetscInt nnz[],Mat *A)
2102: {
2106: MatCreate(comm,A);
2107: MatSetSizes(*A,m,n,m,n);
2108: MatSetType(*A,MATSEQSBAIJ);
2109: MatSeqSBAIJSetPreallocation(*A,bs,nz,(PetscInt*)nnz);
2110: return(0);
2111: }
2113: PetscErrorCode MatDuplicate_SeqSBAIJ(Mat A,MatDuplicateOption cpvalues,Mat *B)
2114: {
2115: Mat C;
2116: Mat_SeqSBAIJ *c,*a = (Mat_SeqSBAIJ*)A->data;
2118: PetscInt i,mbs = a->mbs,nz = a->nz,bs2 =a->bs2;
2121: if (a->i[mbs] != nz) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_PLIB,"Corrupt matrix");
2123: *B = 0;
2124: MatCreate(PetscObjectComm((PetscObject)A),&C);
2125: MatSetSizes(C,A->rmap->N,A->cmap->n,A->rmap->N,A->cmap->n);
2126: MatSetType(C,MATSEQSBAIJ);
2127: PetscMemcpy(C->ops,A->ops,sizeof(struct _MatOps));
2128: c = (Mat_SeqSBAIJ*)C->data;
2130: C->preallocated = PETSC_TRUE;
2131: C->factortype = A->factortype;
2132: c->row = 0;
2133: c->icol = 0;
2134: c->saved_values = 0;
2135: c->keepnonzeropattern = a->keepnonzeropattern;
2136: C->assembled = PETSC_TRUE;
2138: PetscLayoutReference(A->rmap,&C->rmap);
2139: PetscLayoutReference(A->cmap,&C->cmap);
2140: c->bs2 = a->bs2;
2141: c->mbs = a->mbs;
2142: c->nbs = a->nbs;
2144: if (cpvalues == MAT_SHARE_NONZERO_PATTERN) {
2145: c->imax = a->imax;
2146: c->ilen = a->ilen;
2147: c->free_imax_ilen = PETSC_FALSE;
2148: } else {
2149: PetscMalloc2((mbs+1),&c->imax,(mbs+1),&c->ilen);
2150: PetscLogObjectMemory((PetscObject)C,2*(mbs+1)*sizeof(PetscInt));
2151: for (i=0; i<mbs; i++) {
2152: c->imax[i] = a->imax[i];
2153: c->ilen[i] = a->ilen[i];
2154: }
2155: c->free_imax_ilen = PETSC_TRUE;
2156: }
2158: /* allocate the matrix space */
2159: if (cpvalues == MAT_SHARE_NONZERO_PATTERN) {
2160: PetscMalloc1(bs2*nz,&c->a);
2161: PetscLogObjectMemory((PetscObject)C,nz*bs2*sizeof(MatScalar));
2162: c->i = a->i;
2163: c->j = a->j;
2164: c->singlemalloc = PETSC_FALSE;
2165: c->free_a = PETSC_TRUE;
2166: c->free_ij = PETSC_FALSE;
2167: c->parent = A;
2168: PetscObjectReference((PetscObject)A);
2169: MatSetOption(A,MAT_NEW_NONZERO_LOCATION_ERR,PETSC_TRUE);
2170: MatSetOption(C,MAT_NEW_NONZERO_LOCATION_ERR,PETSC_TRUE);
2171: } else {
2172: PetscMalloc3(bs2*nz,&c->a,nz,&c->j,mbs+1,&c->i);
2173: PetscMemcpy(c->i,a->i,(mbs+1)*sizeof(PetscInt));
2174: PetscLogObjectMemory((PetscObject)C,(mbs+1)*sizeof(PetscInt) + nz*(bs2*sizeof(MatScalar) + sizeof(PetscInt)));
2175: c->singlemalloc = PETSC_TRUE;
2176: c->free_a = PETSC_TRUE;
2177: c->free_ij = PETSC_TRUE;
2178: }
2179: if (mbs > 0) {
2180: if (cpvalues != MAT_SHARE_NONZERO_PATTERN) {
2181: PetscMemcpy(c->j,a->j,nz*sizeof(PetscInt));
2182: }
2183: if (cpvalues == MAT_COPY_VALUES) {
2184: PetscMemcpy(c->a,a->a,bs2*nz*sizeof(MatScalar));
2185: } else {
2186: PetscMemzero(c->a,bs2*nz*sizeof(MatScalar));
2187: }
2188: if (a->jshort) {
2189: /* cannot share jshort, it is reallocated in MatAssemblyEnd_SeqSBAIJ() */
2190: /* if the parent matrix is reassembled, this child matrix will never notice */
2191: PetscMalloc1(nz,&c->jshort);
2192: PetscLogObjectMemory((PetscObject)C,nz*sizeof(unsigned short));
2193: PetscMemcpy(c->jshort,a->jshort,nz*sizeof(unsigned short));
2195: c->free_jshort = PETSC_TRUE;
2196: }
2197: }
2199: c->roworiented = a->roworiented;
2200: c->nonew = a->nonew;
2202: if (a->diag) {
2203: if (cpvalues == MAT_SHARE_NONZERO_PATTERN) {
2204: c->diag = a->diag;
2205: c->free_diag = PETSC_FALSE;
2206: } else {
2207: PetscMalloc1(mbs,&c->diag);
2208: PetscLogObjectMemory((PetscObject)C,mbs*sizeof(PetscInt));
2209: for (i=0; i<mbs; i++) c->diag[i] = a->diag[i];
2210: c->free_diag = PETSC_TRUE;
2211: }
2212: }
2213: c->nz = a->nz;
2214: c->maxnz = a->nz; /* Since we allocate exactly the right amount */
2215: c->solve_work = 0;
2216: c->mult_work = 0;
2218: *B = C;
2219: PetscFunctionListDuplicate(((PetscObject)A)->qlist,&((PetscObject)C)->qlist);
2220: return(0);
2221: }
2223: PetscErrorCode MatLoad_SeqSBAIJ(Mat newmat,PetscViewer viewer)
2224: {
2225: Mat_SeqSBAIJ *a;
2227: int fd;
2228: PetscMPIInt size;
2229: PetscInt i,nz,header[4],*rowlengths=0,M,N,bs = newmat->rmap->bs;
2230: PetscInt *mask,mbs,*jj,j,rowcount,nzcount,k,*s_browlengths,maskcount;
2231: PetscInt kmax,jcount,block,idx,point,nzcountb,extra_rows,rows,cols;
2232: PetscInt *masked,nmask,tmp,bs2,ishift;
2233: PetscScalar *aa;
2234: MPI_Comm comm;
2237: /* force binary viewer to load .info file if it has not yet done so */
2238: PetscViewerSetUp(viewer);
2239: PetscObjectGetComm((PetscObject)viewer,&comm);
2240: PetscOptionsGetInt(((PetscObject)newmat)->options,((PetscObject)newmat)->prefix,"-matload_block_size",&bs,NULL);
2241: if (bs < 0) bs = 1;
2242: bs2 = bs*bs;
2244: MPI_Comm_size(comm,&size);
2245: if (size > 1) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"view must have one processor");
2246: PetscViewerBinaryGetDescriptor(viewer,&fd);
2247: PetscBinaryRead(fd,header,4,PETSC_INT);
2248: if (header[0] != MAT_FILE_CLASSID) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_FILE_UNEXPECTED,"not Mat object");
2249: M = header[1]; N = header[2]; nz = header[3];
2251: if (header[3] < 0) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_FILE_UNEXPECTED,"Matrix stored in special format, cannot load as SeqSBAIJ");
2253: if (M != N) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Can only do square matrices");
2255: /*
2256: This code adds extra rows to make sure the number of rows is
2257: divisible by the blocksize
2258: */
2259: mbs = M/bs;
2260: extra_rows = bs - M + bs*(mbs);
2261: if (extra_rows == bs) extra_rows = 0;
2262: else mbs++;
2263: if (extra_rows) {
2264: PetscInfo(viewer,"Padding loaded matrix to match blocksize\n");
2265: }
2267: /* Set global sizes if not already set */
2268: if (newmat->rmap->n < 0 && newmat->rmap->N < 0 && newmat->cmap->n < 0 && newmat->cmap->N < 0) {
2269: MatSetSizes(newmat,PETSC_DECIDE,PETSC_DECIDE,M+extra_rows,N+extra_rows);
2270: } else { /* Check if the matrix global sizes are correct */
2271: MatGetSize(newmat,&rows,&cols);
2272: if (M != rows || N != cols) SETERRQ4(PETSC_COMM_SELF,PETSC_ERR_FILE_UNEXPECTED,"Matrix in file of different length (%d, %d) than the input matrix (%d, %d)",M,N,rows,cols);
2273: }
2275: /* read in row lengths */
2276: PetscMalloc1(M+extra_rows,&rowlengths);
2277: PetscBinaryRead(fd,rowlengths,M,PETSC_INT);
2278: for (i=0; i<extra_rows; i++) rowlengths[M+i] = 1;
2280: /* read in column indices */
2281: PetscMalloc1(nz+extra_rows,&jj);
2282: PetscBinaryRead(fd,jj,nz,PETSC_INT);
2283: for (i=0; i<extra_rows; i++) jj[nz+i] = M+i;
2285: /* loop over row lengths determining block row lengths */
2286: PetscCalloc1(mbs,&s_browlengths);
2287: PetscMalloc2(mbs,&mask,mbs,&masked);
2288: PetscMemzero(mask,mbs*sizeof(PetscInt));
2289: rowcount = 0;
2290: nzcount = 0;
2291: for (i=0; i<mbs; i++) {
2292: nmask = 0;
2293: for (j=0; j<bs; j++) {
2294: kmax = rowlengths[rowcount];
2295: for (k=0; k<kmax; k++) {
2296: tmp = jj[nzcount++]/bs; /* block col. index */
2297: if (!mask[tmp] && tmp >= i) {masked[nmask++] = tmp; mask[tmp] = 1;}
2298: }
2299: rowcount++;
2300: }
2301: s_browlengths[i] += nmask;
2303: /* zero out the mask elements we set */
2304: for (j=0; j<nmask; j++) mask[masked[j]] = 0;
2305: }
2307: /* Do preallocation */
2308: MatSeqSBAIJSetPreallocation(newmat,bs,0,s_browlengths);
2309: a = (Mat_SeqSBAIJ*)newmat->data;
2311: /* set matrix "i" values */
2312: a->i[0] = 0;
2313: for (i=1; i<= mbs; i++) {
2314: a->i[i] = a->i[i-1] + s_browlengths[i-1];
2315: a->ilen[i-1] = s_browlengths[i-1];
2316: }
2317: a->nz = a->i[mbs];
2319: /* read in nonzero values */
2320: PetscMalloc1(nz+extra_rows,&aa);
2321: PetscBinaryRead(fd,aa,nz,PETSC_SCALAR);
2322: for (i=0; i<extra_rows; i++) aa[nz+i] = 1.0;
2324: /* set "a" and "j" values into matrix */
2325: nzcount = 0; jcount = 0;
2326: for (i=0; i<mbs; i++) {
2327: nzcountb = nzcount;
2328: nmask = 0;
2329: for (j=0; j<bs; j++) {
2330: kmax = rowlengths[i*bs+j];
2331: for (k=0; k<kmax; k++) {
2332: tmp = jj[nzcount++]/bs; /* block col. index */
2333: if (!mask[tmp] && tmp >= i) { masked[nmask++] = tmp; mask[tmp] = 1;}
2334: }
2335: }
2336: /* sort the masked values */
2337: PetscSortInt(nmask,masked);
2339: /* set "j" values into matrix */
2340: maskcount = 1;
2341: for (j=0; j<nmask; j++) {
2342: a->j[jcount++] = masked[j];
2343: mask[masked[j]] = maskcount++;
2344: }
2346: /* set "a" values into matrix */
2347: ishift = bs2*a->i[i];
2348: for (j=0; j<bs; j++) {
2349: kmax = rowlengths[i*bs+j];
2350: for (k=0; k<kmax; k++) {
2351: tmp = jj[nzcountb]/bs; /* block col. index */
2352: if (tmp >= i) {
2353: block = mask[tmp] - 1;
2354: point = jj[nzcountb] - bs*tmp;
2355: idx = ishift + bs2*block + j + bs*point;
2356: a->a[idx] = aa[nzcountb];
2357: }
2358: nzcountb++;
2359: }
2360: }
2361: /* zero out the mask elements we set */
2362: for (j=0; j<nmask; j++) mask[masked[j]] = 0;
2363: }
2364: if (jcount != a->nz) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_FILE_UNEXPECTED,"Bad binary matrix");
2366: PetscFree(rowlengths);
2367: PetscFree(s_browlengths);
2368: PetscFree(aa);
2369: PetscFree(jj);
2370: PetscFree2(mask,masked);
2372: MatAssemblyBegin(newmat,MAT_FINAL_ASSEMBLY);
2373: MatAssemblyEnd(newmat,MAT_FINAL_ASSEMBLY);
2374: return(0);
2375: }
2377: /*@
2378: MatCreateSeqSBAIJWithArrays - Creates an sequential SBAIJ matrix using matrix elements
2379: (upper triangular entries in CSR format) provided by the user.
2381: Collective on MPI_Comm
2383: Input Parameters:
2384: + comm - must be an MPI communicator of size 1
2385: . bs - size of block
2386: . m - number of rows
2387: . n - number of columns
2388: . i - row indices
2389: . j - column indices
2390: - a - matrix values
2392: Output Parameter:
2393: . mat - the matrix
2395: Level: advanced
2397: Notes:
2398: The i, j, and a arrays are not copied by this routine, the user must free these arrays
2399: once the matrix is destroyed
2401: You cannot set new nonzero locations into this matrix, that will generate an error.
2403: The i and j indices are 0 based
2405: When block size is greater than 1 the matrix values must be stored using the SBAIJ storage format (see the SBAIJ code to determine this). For block size of 1
2406: it is the regular CSR format excluding the lower triangular elements.
2408: .seealso: MatCreate(), MatCreateSBAIJ(), MatCreateSeqSBAIJ()
2410: @*/
2411: PetscErrorCode MatCreateSeqSBAIJWithArrays(MPI_Comm comm,PetscInt bs,PetscInt m,PetscInt n,PetscInt i[],PetscInt j[],PetscScalar a[],Mat *mat)
2412: {
2414: PetscInt ii;
2415: Mat_SeqSBAIJ *sbaij;
2418: if (bs != 1) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"block size %D > 1 is not supported yet",bs);
2419: if (m > 0 && i[0]) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"i (row indices) must start with 0");
2421: MatCreate(comm,mat);
2422: MatSetSizes(*mat,m,n,m,n);
2423: MatSetType(*mat,MATSEQSBAIJ);
2424: MatSeqSBAIJSetPreallocation(*mat,bs,MAT_SKIP_ALLOCATION,0);
2425: sbaij = (Mat_SeqSBAIJ*)(*mat)->data;
2426: PetscMalloc2(m,&sbaij->imax,m,&sbaij->ilen);
2427: PetscLogObjectMemory((PetscObject)*mat,2*m*sizeof(PetscInt));
2429: sbaij->i = i;
2430: sbaij->j = j;
2431: sbaij->a = a;
2433: sbaij->singlemalloc = PETSC_FALSE;
2434: sbaij->nonew = -1; /*this indicates that inserting a new value in the matrix that generates a new nonzero is an error*/
2435: sbaij->free_a = PETSC_FALSE;
2436: sbaij->free_ij = PETSC_FALSE;
2438: for (ii=0; ii<m; ii++) {
2439: sbaij->ilen[ii] = sbaij->imax[ii] = i[ii+1] - i[ii];
2440: #if defined(PETSC_USE_DEBUG)
2441: if (i[ii+1] - i[ii] < 0) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Negative row length in i (row indices) row = %d length = %d",ii,i[ii+1] - i[ii]);
2442: #endif
2443: }
2444: #if defined(PETSC_USE_DEBUG)
2445: for (ii=0; ii<sbaij->i[m]; ii++) {
2446: if (j[ii] < 0) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Negative column index at location = %d index = %d",ii,j[ii]);
2447: if (j[ii] > n - 1) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Column index to large at location = %d index = %d",ii,j[ii]);
2448: }
2449: #endif
2451: MatAssemblyBegin(*mat,MAT_FINAL_ASSEMBLY);
2452: MatAssemblyEnd(*mat,MAT_FINAL_ASSEMBLY);
2453: return(0);
2454: }
2456: PetscErrorCode MatCreateMPIMatConcatenateSeqMat_SeqSBAIJ(MPI_Comm comm,Mat inmat,PetscInt n,MatReuse scall,Mat *outmat)
2457: {
2459: PetscMPIInt size;
2462: MPI_Comm_size(comm,&size);
2463: if (size == 1 && scall == MAT_REUSE_MATRIX) {
2464: MatCopy(inmat,*outmat,SAME_NONZERO_PATTERN);
2465: } else {
2466: MatCreateMPIMatConcatenateSeqMat_MPISBAIJ(comm,inmat,n,scall,outmat);
2467: }
2468: return(0);
2469: }