Actual source code: sbaij.c
petsc-3.6.1 2015-08-06
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> /*I "petscmat.h" I*/
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: extern PetscErrorCode MatSeqSBAIJSetNumericFactorization_inplace(Mat,PetscBool);
15: #if defined(PETSC_HAVE_ELEMENTAL)
16: PETSC_EXTERN PetscErrorCode MatConvert_SeqSBAIJ_Elemental(Mat,MatType,MatReuse,Mat*);
17: #endif
19: /*
20: Checks for missing diagonals
21: */
24: PetscErrorCode MatMissingDiagonal_SeqSBAIJ(Mat A,PetscBool *missing,PetscInt *dd)
25: {
26: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
28: PetscInt *diag,*ii = a->i,i;
31: MatMarkDiagonal_SeqSBAIJ(A);
32: *missing = PETSC_FALSE;
33: if (A->rmap->n > 0 && !ii) {
34: *missing = PETSC_TRUE;
35: if (dd) *dd = 0;
36: PetscInfo(A,"Matrix has no entries therefore is missing diagonal\n");
37: } else {
38: diag = a->diag;
39: for (i=0; i<a->mbs; i++) {
40: if (diag[i] >= ii[i+1]) {
41: *missing = PETSC_TRUE;
42: if (dd) *dd = i;
43: break;
44: }
45: }
46: }
47: return(0);
48: }
52: PetscErrorCode MatMarkDiagonal_SeqSBAIJ(Mat A)
53: {
54: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
56: PetscInt i,j;
59: if (!a->diag) {
60: PetscMalloc1(a->mbs,&a->diag);
61: PetscLogObjectMemory((PetscObject)A,a->mbs*sizeof(PetscInt));
62: a->free_diag = PETSC_TRUE;
63: }
64: for (i=0; i<a->mbs; i++) {
65: a->diag[i] = a->i[i+1];
66: for (j=a->i[i]; j<a->i[i+1]; j++) {
67: if (a->j[j] == i) {
68: a->diag[i] = j;
69: break;
70: }
71: }
72: }
73: return(0);
74: }
78: static PetscErrorCode MatGetRowIJ_SeqSBAIJ(Mat A,PetscInt oshift,PetscBool symmetric,PetscBool blockcompressed,PetscInt *nn,const PetscInt *inia[],const PetscInt *inja[],PetscBool *done)
79: {
80: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
81: PetscInt i,j,n = a->mbs,nz = a->i[n],bs = A->rmap->bs;
82: PetscInt **ia = (PetscInt**)inia,**ja = (PetscInt**)inja;
86: *nn = n;
87: if (!ia) return(0);
88: if (!blockcompressed) {
89: /* malloc & create the natural set of indices */
90: PetscMalloc2((n+1)*bs,ia,nz*bs,ja);
91: for (i=0; i<n+1; i++) {
92: for (j=0; j<bs; j++) {
93: (*ia)[i*bs+j] = a->i[i]*bs+j+oshift;
94: }
95: }
96: for (i=0; i<nz; i++) {
97: for (j=0; j<bs; j++) {
98: (*ja)[i*bs+j] = a->j[i]*bs+j+oshift;
99: }
100: }
101: } else { /* blockcompressed */
102: if (oshift == 1) {
103: /* temporarily add 1 to i and j indices */
104: for (i=0; i<nz; i++) a->j[i]++;
105: for (i=0; i<n+1; i++) a->i[i]++;
106: }
107: *ia = a->i; *ja = a->j;
108: }
109: return(0);
110: }
114: static PetscErrorCode MatRestoreRowIJ_SeqSBAIJ(Mat A,PetscInt oshift,PetscBool symmetric,PetscBool blockcompressed,PetscInt *nn,const PetscInt *ia[],const PetscInt *ja[],PetscBool *done)
115: {
116: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
117: PetscInt i,n = a->mbs,nz = a->i[n];
121: if (!ia) return(0);
123: if (!blockcompressed) {
124: PetscFree2(*ia,*ja);
125: } else if (oshift == 1) { /* blockcompressed */
126: for (i=0; i<nz; i++) a->j[i]--;
127: for (i=0; i<n+1; i++) a->i[i]--;
128: }
129: return(0);
130: }
134: PetscErrorCode MatDestroy_SeqSBAIJ(Mat A)
135: {
136: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
140: #if defined(PETSC_USE_LOG)
141: PetscLogObjectState((PetscObject)A,"Rows=%D, NZ=%D",A->rmap->N,a->nz);
142: #endif
143: MatSeqXAIJFreeAIJ(A,&a->a,&a->j,&a->i);
144: if (a->free_diag) {PetscFree(a->diag);}
145: ISDestroy(&a->row);
146: ISDestroy(&a->col);
147: ISDestroy(&a->icol);
148: PetscFree(a->idiag);
149: PetscFree(a->inode.size);
150: if (a->free_imax_ilen) {PetscFree2(a->imax,a->ilen);}
151: PetscFree(a->solve_work);
152: PetscFree(a->sor_work);
153: PetscFree(a->solves_work);
154: PetscFree(a->mult_work);
155: PetscFree(a->saved_values);
156: if (a->free_jshort) {PetscFree(a->jshort);}
157: PetscFree(a->inew);
158: MatDestroy(&a->parent);
159: PetscFree(A->data);
161: PetscObjectChangeTypeName((PetscObject)A,0);
162: PetscObjectComposeFunction((PetscObject)A,"MatStoreValues_C",NULL);
163: PetscObjectComposeFunction((PetscObject)A,"MatRetrieveValues_C",NULL);
164: PetscObjectComposeFunction((PetscObject)A,"MatSeqSBAIJSetColumnIndices_C",NULL);
165: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqsbaij_seqaij_C",NULL);
166: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqsbaij_seqbaij_C",NULL);
167: PetscObjectComposeFunction((PetscObject)A,"MatSeqSBAIJSetPreallocation_C",NULL);
168: PetscObjectComposeFunction((PetscObject)A,"MatSeqSBAIJSetPreallocationCSR_C",NULL);
169: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqsbaij_seqsbstrm_C",NULL);
170: #if defined(PETSC_HAVE_ELEMENTAL)
171: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqsbaij_elemental_C",NULL);
172: #endif
173: return(0);
174: }
178: PetscErrorCode MatSetOption_SeqSBAIJ(Mat A,MatOption op,PetscBool flg)
179: {
180: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
184: switch (op) {
185: case MAT_ROW_ORIENTED:
186: a->roworiented = flg;
187: break;
188: case MAT_KEEP_NONZERO_PATTERN:
189: a->keepnonzeropattern = flg;
190: break;
191: case MAT_NEW_NONZERO_LOCATIONS:
192: a->nonew = (flg ? 0 : 1);
193: break;
194: case MAT_NEW_NONZERO_LOCATION_ERR:
195: a->nonew = (flg ? -1 : 0);
196: break;
197: case MAT_NEW_NONZERO_ALLOCATION_ERR:
198: a->nonew = (flg ? -2 : 0);
199: break;
200: case MAT_UNUSED_NONZERO_LOCATION_ERR:
201: a->nounused = (flg ? -1 : 0);
202: break;
203: case MAT_NEW_DIAGONALS:
204: case MAT_IGNORE_OFF_PROC_ENTRIES:
205: case MAT_USE_HASH_TABLE:
206: PetscInfo1(A,"Option %s ignored\n",MatOptions[op]);
207: break;
208: case MAT_HERMITIAN:
209: if (!A->assembled) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Must call MatAssemblyEnd() first");
210: if (A->cmap->n < 65536 && A->cmap->bs == 1) {
211: A->ops->mult = MatMult_SeqSBAIJ_1_Hermitian_ushort;
212: } else if (A->cmap->bs == 1) {
213: A->ops->mult = MatMult_SeqSBAIJ_1_Hermitian;
214: } else SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"No support for Hermitian with block size greater than 1");
215: break;
216: case MAT_SPD:
217: /* These options are handled directly by MatSetOption() */
218: break;
219: case MAT_SYMMETRIC:
220: case MAT_STRUCTURALLY_SYMMETRIC:
221: case MAT_SYMMETRY_ETERNAL:
222: /* These options are handled directly by MatSetOption() */
223: break;
224: case MAT_IGNORE_LOWER_TRIANGULAR:
225: a->ignore_ltriangular = flg;
226: break;
227: case MAT_ERROR_LOWER_TRIANGULAR:
228: a->ignore_ltriangular = flg;
229: break;
230: case MAT_GETROW_UPPERTRIANGULAR:
231: a->getrow_utriangular = flg;
232: break;
233: default:
234: SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"unknown option %d",op);
235: }
236: return(0);
237: }
241: PetscErrorCode MatGetRow_SeqSBAIJ(Mat A,PetscInt row,PetscInt *nz,PetscInt **idx,PetscScalar **v)
242: {
243: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
247: 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()");
249: /* Get the upper triangular part of the row */
250: MatGetRow_SeqBAIJ_private(A,row,nz,idx,v,a->i,a->j,a->a);
251: return(0);
252: }
256: PetscErrorCode MatRestoreRow_SeqSBAIJ(Mat A,PetscInt row,PetscInt *nz,PetscInt **idx,PetscScalar **v)
257: {
261: if (idx) {PetscFree(*idx);}
262: if (v) {PetscFree(*v);}
263: return(0);
264: }
268: PetscErrorCode MatGetRowUpperTriangular_SeqSBAIJ(Mat A)
269: {
270: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
273: a->getrow_utriangular = PETSC_TRUE;
274: return(0);
275: }
278: PetscErrorCode MatRestoreRowUpperTriangular_SeqSBAIJ(Mat A)
279: {
280: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
283: a->getrow_utriangular = PETSC_FALSE;
284: return(0);
285: }
289: PetscErrorCode MatTranspose_SeqSBAIJ(Mat A,MatReuse reuse,Mat *B)
290: {
294: if (reuse == MAT_INITIAL_MATRIX || *B != A) {
295: MatDuplicate(A,MAT_COPY_VALUES,B);
296: }
297: return(0);
298: }
302: PetscErrorCode MatView_SeqSBAIJ_ASCII(Mat A,PetscViewer viewer)
303: {
304: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
305: PetscErrorCode ierr;
306: PetscInt i,j,bs = A->rmap->bs,k,l,bs2=a->bs2;
307: PetscViewerFormat format;
308: PetscInt *diag;
311: PetscViewerGetFormat(viewer,&format);
312: if (format == PETSC_VIEWER_ASCII_INFO || format == PETSC_VIEWER_ASCII_INFO_DETAIL) {
313: PetscViewerASCIIPrintf(viewer," block size is %D\n",bs);
314: } else if (format == PETSC_VIEWER_ASCII_MATLAB) {
315: Mat aij;
316: const char *matname;
318: if (A->factortype && bs>1) {
319: PetscPrintf(PETSC_COMM_SELF,"Warning: matrix is factored with bs>1. MatView() with PETSC_VIEWER_ASCII_MATLAB is not supported and ignored!\n");
320: return(0);
321: }
322: MatConvert(A,MATSEQAIJ,MAT_INITIAL_MATRIX,&aij);
323: PetscObjectGetName((PetscObject)A,&matname);
324: PetscObjectSetName((PetscObject)aij,matname);
325: MatView(aij,viewer);
326: MatDestroy(&aij);
327: } else if (format == PETSC_VIEWER_ASCII_COMMON) {
328: PetscViewerASCIIUseTabs(viewer,PETSC_FALSE);
329: for (i=0; i<a->mbs; i++) {
330: for (j=0; j<bs; j++) {
331: PetscViewerASCIIPrintf(viewer,"row %D:",i*bs+j);
332: for (k=a->i[i]; k<a->i[i+1]; k++) {
333: for (l=0; l<bs; l++) {
334: #if defined(PETSC_USE_COMPLEX)
335: if (PetscImaginaryPart(a->a[bs2*k + l*bs + j]) > 0.0 && PetscRealPart(a->a[bs2*k + l*bs + j]) != 0.0) {
336: PetscViewerASCIIPrintf(viewer," (%D, %g + %g i) ",bs*a->j[k]+l,
337: (double)PetscRealPart(a->a[bs2*k + l*bs + j]),(double)PetscImaginaryPart(a->a[bs2*k + l*bs + j]));
338: } else if (PetscImaginaryPart(a->a[bs2*k + l*bs + j]) < 0.0 && PetscRealPart(a->a[bs2*k + l*bs + j]) != 0.0) {
339: PetscViewerASCIIPrintf(viewer," (%D, %g - %g i) ",bs*a->j[k]+l,
340: (double)PetscRealPart(a->a[bs2*k + l*bs + j]),-(double)PetscImaginaryPart(a->a[bs2*k + l*bs + j]));
341: } else if (PetscRealPart(a->a[bs2*k + l*bs + j]) != 0.0) {
342: PetscViewerASCIIPrintf(viewer," (%D, %g) ",bs*a->j[k]+l,(double)PetscRealPart(a->a[bs2*k + l*bs + j]));
343: }
344: #else
345: if (a->a[bs2*k + l*bs + j] != 0.0) {
346: PetscViewerASCIIPrintf(viewer," (%D, %g) ",bs*a->j[k]+l,(double)a->a[bs2*k + l*bs + j]);
347: }
348: #endif
349: }
350: }
351: PetscViewerASCIIPrintf(viewer,"\n");
352: }
353: }
354: PetscViewerASCIIUseTabs(viewer,PETSC_TRUE);
355: } else if (format == PETSC_VIEWER_ASCII_FACTOR_INFO) {
356: return(0);
357: } else {
358: PetscViewerASCIIUseTabs(viewer,PETSC_FALSE);
359: if (A->factortype) { /* for factored matrix */
360: if (bs>1) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"matrix is factored with bs>1. Not implemented yet");
362: diag=a->diag;
363: for (i=0; i<a->mbs; i++) { /* for row block i */
364: PetscViewerASCIIPrintf(viewer,"row %D:",i);
365: /* diagonal entry */
366: #if defined(PETSC_USE_COMPLEX)
367: if (PetscImaginaryPart(a->a[diag[i]]) > 0.0) {
368: 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]]));
369: } else if (PetscImaginaryPart(a->a[diag[i]]) < 0.0) {
370: 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]]));
371: } else {
372: PetscViewerASCIIPrintf(viewer," (%D, %g) ",a->j[diag[i]],(double)PetscRealPart(1.0/a->a[diag[i]]));
373: }
374: #else
375: PetscViewerASCIIPrintf(viewer," (%D, %g) ",a->j[diag[i]],(double)(1.0/a->a[diag[i]]));
376: #endif
377: /* off-diagonal entries */
378: for (k=a->i[i]; k<a->i[i+1]-1; k++) {
379: #if defined(PETSC_USE_COMPLEX)
380: if (PetscImaginaryPart(a->a[k]) > 0.0) {
381: PetscViewerASCIIPrintf(viewer," (%D, %g + %g i) ",bs*a->j[k],(double)PetscRealPart(a->a[k]),(double)PetscImaginaryPart(a->a[k]));
382: } else if (PetscImaginaryPart(a->a[k]) < 0.0) {
383: PetscViewerASCIIPrintf(viewer," (%D, %g - %g i) ",bs*a->j[k],(double)PetscRealPart(a->a[k]),-(double)PetscImaginaryPart(a->a[k]));
384: } else {
385: PetscViewerASCIIPrintf(viewer," (%D, %g) ",bs*a->j[k],(double)PetscRealPart(a->a[k]));
386: }
387: #else
388: PetscViewerASCIIPrintf(viewer," (%D, %g) ",a->j[k],(double)a->a[k]);
389: #endif
390: }
391: PetscViewerASCIIPrintf(viewer,"\n");
392: }
394: } else { /* for non-factored matrix */
395: for (i=0; i<a->mbs; i++) { /* for row block i */
396: for (j=0; j<bs; j++) { /* for row bs*i + j */
397: PetscViewerASCIIPrintf(viewer,"row %D:",i*bs+j);
398: for (k=a->i[i]; k<a->i[i+1]; k++) { /* for column block */
399: for (l=0; l<bs; l++) { /* for column */
400: #if defined(PETSC_USE_COMPLEX)
401: if (PetscImaginaryPart(a->a[bs2*k + l*bs + j]) > 0.0) {
402: PetscViewerASCIIPrintf(viewer," (%D, %g + %g i) ",bs*a->j[k]+l,
403: (double)PetscRealPart(a->a[bs2*k + l*bs + j]),(double)PetscImaginaryPart(a->a[bs2*k + l*bs + j]));
404: } else if (PetscImaginaryPart(a->a[bs2*k + l*bs + j]) < 0.0) {
405: PetscViewerASCIIPrintf(viewer," (%D, %g - %g i) ",bs*a->j[k]+l,
406: (double)PetscRealPart(a->a[bs2*k + l*bs + j]),-(double)PetscImaginaryPart(a->a[bs2*k + l*bs + j]));
407: } else {
408: PetscViewerASCIIPrintf(viewer," (%D, %g) ",bs*a->j[k]+l,(double)PetscRealPart(a->a[bs2*k + l*bs + j]));
409: }
410: #else
411: PetscViewerASCIIPrintf(viewer," (%D, %g) ",bs*a->j[k]+l,(double)a->a[bs2*k + l*bs + j]);
412: #endif
413: }
414: }
415: PetscViewerASCIIPrintf(viewer,"\n");
416: }
417: }
418: }
419: PetscViewerASCIIUseTabs(viewer,PETSC_TRUE);
420: }
421: PetscViewerFlush(viewer);
422: return(0);
423: }
425: #include <petscdraw.h>
428: static PetscErrorCode MatView_SeqSBAIJ_Draw_Zoom(PetscDraw draw,void *Aa)
429: {
430: Mat A = (Mat) Aa;
431: Mat_SeqSBAIJ *a=(Mat_SeqSBAIJ*)A->data;
433: PetscInt row,i,j,k,l,mbs=a->mbs,color,bs=A->rmap->bs,bs2=a->bs2;
434: PetscMPIInt rank;
435: PetscReal xl,yl,xr,yr,x_l,x_r,y_l,y_r;
436: MatScalar *aa;
437: MPI_Comm comm;
438: PetscViewer viewer;
441: /*
442: This is nasty. If this is called from an originally parallel matrix
443: then all processes call this,but only the first has the matrix so the
444: rest should return immediately.
445: */
446: PetscObjectGetComm((PetscObject)draw,&comm);
447: MPI_Comm_rank(comm,&rank);
448: if (rank) return(0);
450: PetscObjectQuery((PetscObject)A,"Zoomviewer",(PetscObject*)&viewer);
452: PetscDrawGetCoordinates(draw,&xl,&yl,&xr,&yr);
453: PetscDrawString(draw, .3*(xl+xr), .3*(yl+yr), PETSC_DRAW_BLACK, "symmetric");
455: /* loop over matrix elements drawing boxes */
456: color = PETSC_DRAW_BLUE;
457: for (i=0,row=0; i<mbs; i++,row+=bs) {
458: for (j=a->i[i]; j<a->i[i+1]; j++) {
459: y_l = A->rmap->N - row - 1.0; y_r = y_l + 1.0;
460: x_l = a->j[j]*bs; x_r = x_l + 1.0;
461: aa = a->a + j*bs2;
462: for (k=0; k<bs; k++) {
463: for (l=0; l<bs; l++) {
464: if (PetscRealPart(*aa++) >= 0.) continue;
465: PetscDrawRectangle(draw,x_l+k,y_l-l,x_r+k,y_r-l,color,color,color,color);
466: }
467: }
468: }
469: }
470: color = PETSC_DRAW_CYAN;
471: for (i=0,row=0; i<mbs; i++,row+=bs) {
472: for (j=a->i[i]; j<a->i[i+1]; j++) {
473: y_l = A->rmap->N - row - 1.0; y_r = y_l + 1.0;
474: x_l = a->j[j]*bs; x_r = x_l + 1.0;
475: aa = a->a + j*bs2;
476: for (k=0; k<bs; k++) {
477: for (l=0; l<bs; l++) {
478: if (PetscRealPart(*aa++) != 0.) continue;
479: PetscDrawRectangle(draw,x_l+k,y_l-l,x_r+k,y_r-l,color,color,color,color);
480: }
481: }
482: }
483: }
485: color = PETSC_DRAW_RED;
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: return(0);
500: }
504: static PetscErrorCode MatView_SeqSBAIJ_Draw(Mat A,PetscViewer viewer)
505: {
507: PetscReal xl,yl,xr,yr,w,h;
508: PetscDraw draw;
509: PetscBool isnull;
512: PetscViewerDrawGetDraw(viewer,0,&draw);
513: PetscDrawIsNull(draw,&isnull); if (isnull) return(0);
515: PetscObjectCompose((PetscObject)A,"Zoomviewer",(PetscObject)viewer);
516: xr = A->rmap->N; yr = A->rmap->N; h = yr/10.0; w = xr/10.0;
517: xr += w; yr += h; xl = -w; yl = -h;
518: PetscDrawSetCoordinates(draw,xl,yl,xr,yr);
519: PetscDrawZoom(draw,MatView_SeqSBAIJ_Draw_Zoom,A);
520: PetscObjectCompose((PetscObject)A,"Zoomviewer",NULL);
521: return(0);
522: }
526: PetscErrorCode MatView_SeqSBAIJ(Mat A,PetscViewer viewer)
527: {
529: PetscBool iascii,isdraw;
530: FILE *file = 0;
533: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERASCII,&iascii);
534: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERDRAW,&isdraw);
535: if (iascii) {
536: MatView_SeqSBAIJ_ASCII(A,viewer);
537: } else if (isdraw) {
538: MatView_SeqSBAIJ_Draw(A,viewer);
539: } else {
540: Mat B;
541: const char *matname;
542: MatConvert(A,MATSEQAIJ,MAT_INITIAL_MATRIX,&B);
543: PetscObjectGetName((PetscObject)A,&matname);
544: PetscObjectSetName((PetscObject)B,matname);
545: MatView(B,viewer);
546: MatDestroy(&B);
547: PetscViewerBinaryGetInfoPointer(viewer,&file);
548: if (file) {
549: fprintf(file,"-matload_block_size %d\n",(int)A->rmap->bs);
550: }
551: }
552: return(0);
553: }
558: PetscErrorCode MatGetValues_SeqSBAIJ(Mat A,PetscInt m,const PetscInt im[],PetscInt n,const PetscInt in[],PetscScalar v[])
559: {
560: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
561: PetscInt *rp,k,low,high,t,row,nrow,i,col,l,*aj = a->j;
562: PetscInt *ai = a->i,*ailen = a->ilen;
563: PetscInt brow,bcol,ridx,cidx,bs=A->rmap->bs,bs2=a->bs2;
564: MatScalar *ap,*aa = a->a;
567: for (k=0; k<m; k++) { /* loop over rows */
568: row = im[k]; brow = row/bs;
569: if (row < 0) {v += n; continue;} /* SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Negative row: %D",row); */
570: 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);
571: rp = aj + ai[brow]; ap = aa + bs2*ai[brow];
572: nrow = ailen[brow];
573: for (l=0; l<n; l++) { /* loop over columns */
574: if (in[l] < 0) {v++; continue;} /* SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Negative column: %D",in[l]); */
575: 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);
576: col = in[l];
577: bcol = col/bs;
578: cidx = col%bs;
579: ridx = row%bs;
580: high = nrow;
581: low = 0; /* assume unsorted */
582: while (high-low > 5) {
583: t = (low+high)/2;
584: if (rp[t] > bcol) high = t;
585: else low = t;
586: }
587: for (i=low; i<high; i++) {
588: if (rp[i] > bcol) break;
589: if (rp[i] == bcol) {
590: *v++ = ap[bs2*i+bs*cidx+ridx];
591: goto finished;
592: }
593: }
594: *v++ = 0.0;
595: finished:;
596: }
597: }
598: return(0);
599: }
604: PetscErrorCode MatSetValuesBlocked_SeqSBAIJ(Mat A,PetscInt m,const PetscInt im[],PetscInt n,const PetscInt in[],const PetscScalar v[],InsertMode is)
605: {
606: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
607: PetscErrorCode ierr;
608: PetscInt *rp,k,low,high,t,ii,jj,row,nrow,i,col,l,rmax,N,lastcol = -1;
609: PetscInt *imax =a->imax,*ai=a->i,*ailen=a->ilen;
610: PetscInt *aj =a->j,nonew=a->nonew,bs2=a->bs2,bs=A->rmap->bs,stepval;
611: PetscBool roworiented=a->roworiented;
612: const PetscScalar *value = v;
613: MatScalar *ap,*aa = a->a,*bap;
616: if (roworiented) stepval = (n-1)*bs;
617: else stepval = (m-1)*bs;
619: for (k=0; k<m; k++) { /* loop over added rows */
620: row = im[k];
621: if (row < 0) continue;
622: #if defined(PETSC_USE_DEBUG)
623: if (row >= a->mbs) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Block index row too large %D max %D",row,a->mbs-1);
624: #endif
625: rp = aj + ai[row];
626: ap = aa + bs2*ai[row];
627: rmax = imax[row];
628: nrow = ailen[row];
629: low = 0;
630: high = nrow;
631: for (l=0; l<n; l++) { /* loop over added columns */
632: if (in[l] < 0) continue;
633: col = in[l];
634: #if defined(PETSC_USE_DEBUG)
635: if (col >= a->nbs) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Block index column too large %D max %D",col,a->nbs-1);
636: #endif
637: if (col < row) {
638: if (a->ignore_ltriangular) continue; /* ignore lower triangular block */
639: 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)");
640: }
641: if (roworiented) value = v + k*(stepval+bs)*bs + l*bs;
642: else value = v + l*(stepval+bs)*bs + k*bs;
644: if (col <= lastcol) low = 0;
645: else high = nrow;
647: lastcol = col;
648: while (high-low > 7) {
649: t = (low+high)/2;
650: if (rp[t] > col) high = t;
651: else low = t;
652: }
653: for (i=low; i<high; i++) {
654: if (rp[i] > col) break;
655: if (rp[i] == col) {
656: bap = ap + bs2*i;
657: if (roworiented) {
658: if (is == ADD_VALUES) {
659: for (ii=0; ii<bs; ii++,value+=stepval) {
660: for (jj=ii; jj<bs2; jj+=bs) {
661: bap[jj] += *value++;
662: }
663: }
664: } else {
665: for (ii=0; ii<bs; ii++,value+=stepval) {
666: for (jj=ii; jj<bs2; jj+=bs) {
667: bap[jj] = *value++;
668: }
669: }
670: }
671: } else {
672: if (is == ADD_VALUES) {
673: for (ii=0; ii<bs; ii++,value+=stepval) {
674: for (jj=0; jj<bs; jj++) {
675: *bap++ += *value++;
676: }
677: }
678: } else {
679: for (ii=0; ii<bs; ii++,value+=stepval) {
680: for (jj=0; jj<bs; jj++) {
681: *bap++ = *value++;
682: }
683: }
684: }
685: }
686: goto noinsert2;
687: }
688: }
689: if (nonew == 1) goto noinsert2;
690: 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);
691: MatSeqXAIJReallocateAIJ(A,a->mbs,bs2,nrow,row,col,rmax,aa,ai,aj,rp,ap,imax,nonew,MatScalar);
692: N = nrow++ - 1; high++;
693: /* shift up all the later entries in this row */
694: for (ii=N; ii>=i; ii--) {
695: rp[ii+1] = rp[ii];
696: PetscMemcpy(ap+bs2*(ii+1),ap+bs2*(ii),bs2*sizeof(MatScalar));
697: }
698: if (N >= i) {
699: PetscMemzero(ap+bs2*i,bs2*sizeof(MatScalar));
700: }
701: rp[i] = col;
702: bap = ap + bs2*i;
703: if (roworiented) {
704: for (ii=0; ii<bs; ii++,value+=stepval) {
705: for (jj=ii; jj<bs2; jj+=bs) {
706: bap[jj] = *value++;
707: }
708: }
709: } else {
710: for (ii=0; ii<bs; ii++,value+=stepval) {
711: for (jj=0; jj<bs; jj++) {
712: *bap++ = *value++;
713: }
714: }
715: }
716: noinsert2:;
717: low = i;
718: }
719: ailen[row] = nrow;
720: }
721: return(0);
722: }
724: /*
725: This is not yet used
726: */
729: PetscErrorCode MatAssemblyEnd_SeqSBAIJ_SeqAIJ_Inode(Mat A)
730: {
731: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
733: const PetscInt *ai = a->i, *aj = a->j,*cols;
734: PetscInt i = 0,j,blk_size,m = A->rmap->n,node_count = 0,nzx,nzy,*ns,row,nz,cnt,cnt2,*counts;
735: PetscBool flag;
738: PetscMalloc1(m,&ns);
739: while (i < m) {
740: nzx = ai[i+1] - ai[i]; /* Number of nonzeros */
741: /* Limits the number of elements in a node to 'a->inode.limit' */
742: for (j=i+1,blk_size=1; j<m && blk_size <a->inode.limit; ++j,++blk_size) {
743: nzy = ai[j+1] - ai[j];
744: if (nzy != (nzx - j + i)) break;
745: PetscMemcmp(aj + ai[i] + j - i,aj + ai[j],nzy*sizeof(PetscInt),&flag);
746: if (!flag) break;
747: }
748: ns[node_count++] = blk_size;
750: i = j;
751: }
752: if (!a->inode.size && m && node_count > .9*m) {
753: PetscFree(ns);
754: PetscInfo2(A,"Found %D nodes out of %D rows. Not using Inode routines\n",node_count,m);
755: } else {
756: a->inode.node_count = node_count;
758: PetscMalloc1(node_count,&a->inode.size);
759: PetscLogObjectMemory((PetscObject)A,node_count*sizeof(PetscInt));
760: PetscMemcpy(a->inode.size,ns,node_count*sizeof(PetscInt));
761: PetscFree(ns);
762: PetscInfo3(A,"Found %D nodes of %D. Limit used: %D. Using Inode routines\n",node_count,m,a->inode.limit);
764: /* count collections of adjacent columns in each inode */
765: row = 0;
766: cnt = 0;
767: for (i=0; i<node_count; i++) {
768: cols = aj + ai[row] + a->inode.size[i];
769: nz = ai[row+1] - ai[row] - a->inode.size[i];
770: for (j=1; j<nz; j++) {
771: if (cols[j] != cols[j-1]+1) cnt++;
772: }
773: cnt++;
774: row += a->inode.size[i];
775: }
776: PetscMalloc1(2*cnt,&counts);
777: cnt = 0;
778: row = 0;
779: for (i=0; i<node_count; i++) {
780: cols = aj + ai[row] + a->inode.size[i];
781: counts[2*cnt] = cols[0];
782: nz = ai[row+1] - ai[row] - a->inode.size[i];
783: cnt2 = 1;
784: for (j=1; j<nz; j++) {
785: if (cols[j] != cols[j-1]+1) {
786: counts[2*(cnt++)+1] = cnt2;
787: counts[2*cnt] = cols[j];
788: cnt2 = 1;
789: } else cnt2++;
790: }
791: counts[2*(cnt++)+1] = cnt2;
792: row += a->inode.size[i];
793: }
794: PetscIntView(2*cnt,counts,0);
795: }
796: return(0);
797: }
801: PetscErrorCode MatAssemblyEnd_SeqSBAIJ(Mat A,MatAssemblyType mode)
802: {
803: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
805: PetscInt fshift = 0,i,j,*ai = a->i,*aj = a->j,*imax = a->imax;
806: PetscInt m = A->rmap->N,*ip,N,*ailen = a->ilen;
807: PetscInt mbs = a->mbs,bs2 = a->bs2,rmax = 0;
808: MatScalar *aa = a->a,*ap;
811: if (mode == MAT_FLUSH_ASSEMBLY) return(0);
813: if (m) rmax = ailen[0];
814: for (i=1; i<mbs; i++) {
815: /* move each row back by the amount of empty slots (fshift) before it*/
816: fshift += imax[i-1] - ailen[i-1];
817: rmax = PetscMax(rmax,ailen[i]);
818: if (fshift) {
819: ip = aj + ai[i]; ap = aa + bs2*ai[i];
820: N = ailen[i];
821: for (j=0; j<N; j++) {
822: ip[j-fshift] = ip[j];
823: PetscMemcpy(ap+(j-fshift)*bs2,ap+j*bs2,bs2*sizeof(MatScalar));
824: }
825: }
826: ai[i] = ai[i-1] + ailen[i-1];
827: }
828: if (mbs) {
829: fshift += imax[mbs-1] - ailen[mbs-1];
830: ai[mbs] = ai[mbs-1] + ailen[mbs-1];
831: }
832: /* reset ilen and imax for each row */
833: for (i=0; i<mbs; i++) {
834: ailen[i] = imax[i] = ai[i+1] - ai[i];
835: }
836: a->nz = ai[mbs];
838: /* diagonals may have moved, reset it */
839: if (a->diag) {
840: PetscMemcpy(a->diag,ai,mbs*sizeof(PetscInt));
841: }
842: 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);
844: 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);
845: PetscInfo1(A,"Number of mallocs during MatSetValues is %D\n",a->reallocs);
846: PetscInfo1(A,"Most nonzeros blocks in any row is %D\n",rmax);
848: A->info.mallocs += a->reallocs;
849: a->reallocs = 0;
850: A->info.nz_unneeded = (PetscReal)fshift*bs2;
851: a->idiagvalid = PETSC_FALSE;
852: a->rmax = rmax;
854: if (A->cmap->n < 65536 && A->cmap->bs == 1) {
855: if (a->jshort && a->free_jshort) {
856: /* when matrix data structure is changed, previous jshort must be replaced */
857: PetscFree(a->jshort);
858: }
859: PetscMalloc1(a->i[A->rmap->n],&a->jshort);
860: PetscLogObjectMemory((PetscObject)A,a->i[A->rmap->n]*sizeof(unsigned short));
861: for (i=0; i<a->i[A->rmap->n]; i++) a->jshort[i] = a->j[i];
862: A->ops->mult = MatMult_SeqSBAIJ_1_ushort;
863: A->ops->sor = MatSOR_SeqSBAIJ_ushort;
864: a->free_jshort = PETSC_TRUE;
865: }
866: return(0);
867: }
869: /*
870: This function returns an array of flags which indicate the locations of contiguous
871: blocks that should be zeroed. for eg: if bs = 3 and is = [0,1,2,3,5,6,7,8,9]
872: then the resulting sizes = [3,1,1,3,1] correspondig to sets [(0,1,2),(3),(5),(6,7,8),(9)]
873: Assume: sizes should be long enough to hold all the values.
874: */
877: PetscErrorCode MatZeroRows_SeqSBAIJ_Check_Blocks(PetscInt idx[],PetscInt n,PetscInt bs,PetscInt sizes[], PetscInt *bs_max)
878: {
879: PetscInt i,j,k,row;
880: PetscBool flg;
883: for (i=0,j=0; i<n; j++) {
884: row = idx[i];
885: if (row%bs!=0) { /* Not the begining of a block */
886: sizes[j] = 1;
887: i++;
888: } else if (i+bs > n) { /* Beginning of a block, but complete block doesn't exist (at idx end) */
889: sizes[j] = 1; /* Also makes sure atleast 'bs' values exist for next else */
890: i++;
891: } else { /* Begining of the block, so check if the complete block exists */
892: flg = PETSC_TRUE;
893: for (k=1; k<bs; k++) {
894: if (row+k != idx[i+k]) { /* break in the block */
895: flg = PETSC_FALSE;
896: break;
897: }
898: }
899: if (flg) { /* No break in the bs */
900: sizes[j] = bs;
901: i += bs;
902: } else {
903: sizes[j] = 1;
904: i++;
905: }
906: }
907: }
908: *bs_max = j;
909: return(0);
910: }
913: /* Only add/insert a(i,j) with i<=j (blocks).
914: Any a(i,j) with i>j input by user is ingored.
915: */
919: PetscErrorCode MatSetValues_SeqSBAIJ(Mat A,PetscInt m,const PetscInt im[],PetscInt n,const PetscInt in[],const PetscScalar v[],InsertMode is)
920: {
921: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
923: PetscInt *rp,k,low,high,t,ii,row,nrow,i,col,l,rmax,N,lastcol = -1;
924: PetscInt *imax=a->imax,*ai=a->i,*ailen=a->ilen,roworiented=a->roworiented;
925: PetscInt *aj =a->j,nonew=a->nonew,bs=A->rmap->bs,brow,bcol;
926: PetscInt ridx,cidx,bs2=a->bs2;
927: MatScalar *ap,value,*aa=a->a,*bap;
930: for (k=0; k<m; k++) { /* loop over added rows */
931: row = im[k]; /* row number */
932: brow = row/bs; /* block row number */
933: if (row < 0) continue;
934: #if defined(PETSC_USE_DEBUG)
935: 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);
936: #endif
937: rp = aj + ai[brow]; /*ptr to beginning of column value of the row block*/
938: ap = aa + bs2*ai[brow]; /*ptr to beginning of element value of the row block*/
939: rmax = imax[brow]; /* maximum space allocated for this row */
940: nrow = ailen[brow]; /* actual length of this row */
941: low = 0;
943: for (l=0; l<n; l++) { /* loop over added columns */
944: if (in[l] < 0) continue;
945: #if defined(PETSC_USE_DEBUG)
946: 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);
947: #endif
948: col = in[l];
949: bcol = col/bs; /* block col number */
951: if (brow > bcol) {
952: if (a->ignore_ltriangular) continue; /* ignore lower triangular values */
953: 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)");
954: }
956: ridx = row % bs; cidx = col % bs; /*row and col index inside the block */
957: if ((brow==bcol && ridx<=cidx) || (brow<bcol)) {
958: /* element value a(k,l) */
959: if (roworiented) value = v[l + k*n];
960: else value = v[k + l*m];
962: /* move pointer bap to a(k,l) quickly and add/insert value */
963: if (col <= lastcol) low = 0;
964: high = nrow;
965: lastcol = col;
966: while (high-low > 7) {
967: t = (low+high)/2;
968: if (rp[t] > bcol) high = t;
969: else low = t;
970: }
971: for (i=low; i<high; i++) {
972: if (rp[i] > bcol) break;
973: if (rp[i] == bcol) {
974: bap = ap + bs2*i + bs*cidx + ridx;
975: if (is == ADD_VALUES) *bap += value;
976: else *bap = value;
977: /* for diag block, add/insert its symmetric element a(cidx,ridx) */
978: if (brow == bcol && ridx < cidx) {
979: bap = ap + bs2*i + bs*ridx + cidx;
980: if (is == ADD_VALUES) *bap += value;
981: else *bap = value;
982: }
983: goto noinsert1;
984: }
985: }
987: if (nonew == 1) goto noinsert1;
988: if (nonew == -1) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Inserting a new nonzero (%D, %D) in the matrix", row, col);
989: MatSeqXAIJReallocateAIJ(A,a->mbs,bs2,nrow,brow,bcol,rmax,aa,ai,aj,rp,ap,imax,nonew,MatScalar);
991: N = nrow++ - 1; high++;
992: /* shift up all the later entries in this row */
993: for (ii=N; ii>=i; ii--) {
994: rp[ii+1] = rp[ii];
995: PetscMemcpy(ap+bs2*(ii+1),ap+bs2*(ii),bs2*sizeof(MatScalar));
996: }
997: if (N>=i) {
998: PetscMemzero(ap+bs2*i,bs2*sizeof(MatScalar));
999: }
1000: rp[i] = bcol;
1001: ap[bs2*i + bs*cidx + ridx] = value;
1002: A->nonzerostate++;
1003: noinsert1:;
1004: low = i;
1005: }
1006: } /* end of loop over added columns */
1007: ailen[brow] = nrow;
1008: } /* end of loop over added rows */
1009: return(0);
1010: }
1014: PetscErrorCode MatICCFactor_SeqSBAIJ(Mat inA,IS row,const MatFactorInfo *info)
1015: {
1016: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)inA->data;
1017: Mat outA;
1019: PetscBool row_identity;
1022: if (info->levels != 0) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Only levels=0 is supported for in-place icc");
1023: ISIdentity(row,&row_identity);
1024: if (!row_identity) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Matrix reordering is not supported");
1025: 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()! */
1027: outA = inA;
1028: inA->factortype = MAT_FACTOR_ICC;
1030: MatMarkDiagonal_SeqSBAIJ(inA);
1031: MatSeqSBAIJSetNumericFactorization_inplace(inA,row_identity);
1033: PetscObjectReference((PetscObject)row);
1034: ISDestroy(&a->row);
1035: a->row = row;
1036: PetscObjectReference((PetscObject)row);
1037: ISDestroy(&a->col);
1038: a->col = row;
1040: /* Create the invert permutation so that it can be used in MatCholeskyFactorNumeric() */
1041: if (a->icol) {ISInvertPermutation(row,PETSC_DECIDE, &a->icol);}
1042: PetscLogObjectParent((PetscObject)inA,(PetscObject)a->icol);
1044: if (!a->solve_work) {
1045: PetscMalloc1(inA->rmap->N+inA->rmap->bs,&a->solve_work);
1046: PetscLogObjectMemory((PetscObject)inA,(inA->rmap->N+inA->rmap->bs)*sizeof(PetscScalar));
1047: }
1049: MatCholeskyFactorNumeric(outA,inA,info);
1050: return(0);
1051: }
1055: PetscErrorCode MatSeqSBAIJSetColumnIndices_SeqSBAIJ(Mat mat,PetscInt *indices)
1056: {
1057: Mat_SeqSBAIJ *baij = (Mat_SeqSBAIJ*)mat->data;
1058: PetscInt i,nz,n;
1062: nz = baij->maxnz;
1063: n = mat->cmap->n;
1064: for (i=0; i<nz; i++) baij->j[i] = indices[i];
1066: baij->nz = nz;
1067: for (i=0; i<n; i++) baij->ilen[i] = baij->imax[i];
1069: MatSetOption(mat,MAT_NEW_NONZERO_LOCATION_ERR,PETSC_TRUE);
1070: return(0);
1071: }
1075: /*@
1076: MatSeqSBAIJSetColumnIndices - Set the column indices for all the rows
1077: in the matrix.
1079: Input Parameters:
1080: + mat - the SeqSBAIJ matrix
1081: - indices - the column indices
1083: Level: advanced
1085: Notes:
1086: This can be called if you have precomputed the nonzero structure of the
1087: matrix and want to provide it to the matrix object to improve the performance
1088: of the MatSetValues() operation.
1090: You MUST have set the correct numbers of nonzeros per row in the call to
1091: MatCreateSeqSBAIJ(), and the columns indices MUST be sorted.
1093: MUST be called before any calls to MatSetValues()
1095: .seealso: MatCreateSeqSBAIJ
1096: @*/
1097: PetscErrorCode MatSeqSBAIJSetColumnIndices(Mat mat,PetscInt *indices)
1098: {
1104: PetscUseMethod(mat,"MatSeqSBAIJSetColumnIndices_C",(Mat,PetscInt*),(mat,indices));
1105: return(0);
1106: }
1110: PetscErrorCode MatCopy_SeqSBAIJ(Mat A,Mat B,MatStructure str)
1111: {
1115: /* If the two matrices have the same copy implementation, use fast copy. */
1116: if (str == SAME_NONZERO_PATTERN && (A->ops->copy == B->ops->copy)) {
1117: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
1118: Mat_SeqSBAIJ *b = (Mat_SeqSBAIJ*)B->data;
1120: 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");
1121: PetscMemcpy(b->a,a->a,(a->i[A->rmap->N])*sizeof(PetscScalar));
1122: } else {
1123: MatGetRowUpperTriangular(A);
1124: MatCopy_Basic(A,B,str);
1125: MatRestoreRowUpperTriangular(A);
1126: }
1127: return(0);
1128: }
1132: PetscErrorCode MatSetUp_SeqSBAIJ(Mat A)
1133: {
1137: MatSeqSBAIJSetPreallocation_SeqSBAIJ(A,A->rmap->bs,PETSC_DEFAULT,0);
1138: return(0);
1139: }
1143: PetscErrorCode MatSeqSBAIJGetArray_SeqSBAIJ(Mat A,PetscScalar *array[])
1144: {
1145: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
1148: *array = a->a;
1149: return(0);
1150: }
1154: PetscErrorCode MatSeqSBAIJRestoreArray_SeqSBAIJ(Mat A,PetscScalar *array[])
1155: {
1157: return(0);
1158: }
1162: PetscErrorCode MatAXPYGetPreallocation_SeqSBAIJ(Mat Y,Mat X,PetscInt *nnz)
1163: {
1164: PetscInt bs = Y->rmap->bs,mbs = Y->rmap->N/bs;
1165: Mat_SeqSBAIJ *x = (Mat_SeqSBAIJ*)X->data;
1166: Mat_SeqSBAIJ *y = (Mat_SeqSBAIJ*)Y->data;
1170: /* Set the number of nonzeros in the new matrix */
1171: MatAXPYGetPreallocation_SeqX_private(mbs,x->i,x->j,y->i,y->j,nnz);
1172: return(0);
1173: }
1177: PetscErrorCode MatAXPY_SeqSBAIJ(Mat Y,PetscScalar a,Mat X,MatStructure str)
1178: {
1179: Mat_SeqSBAIJ *x=(Mat_SeqSBAIJ*)X->data, *y=(Mat_SeqSBAIJ*)Y->data;
1181: PetscInt bs=Y->rmap->bs,bs2=bs*bs;
1182: PetscBLASInt one = 1;
1185: if (str == SAME_NONZERO_PATTERN) {
1186: PetscScalar alpha = a;
1187: PetscBLASInt bnz;
1188: PetscBLASIntCast(x->nz*bs2,&bnz);
1189: PetscStackCallBLAS("BLASaxpy",BLASaxpy_(&bnz,&alpha,x->a,&one,y->a,&one));
1190: PetscObjectStateIncrease((PetscObject)Y);
1191: } else if (str == SUBSET_NONZERO_PATTERN) { /* nonzeros of X is a subset of Y's */
1192: MatSetOption(X,MAT_GETROW_UPPERTRIANGULAR,PETSC_TRUE);
1193: MatAXPY_Basic(Y,a,X,str);
1194: MatSetOption(X,MAT_GETROW_UPPERTRIANGULAR,PETSC_FALSE);
1195: } else {
1196: Mat B;
1197: PetscInt *nnz;
1198: if (bs != X->rmap->bs) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Matrices must have same block size");
1199: MatGetRowUpperTriangular(X);
1200: MatGetRowUpperTriangular(Y);
1201: PetscMalloc1(Y->rmap->N,&nnz);
1202: MatCreate(PetscObjectComm((PetscObject)Y),&B);
1203: PetscObjectSetName((PetscObject)B,((PetscObject)Y)->name);
1204: MatSetSizes(B,Y->rmap->n,Y->cmap->n,Y->rmap->N,Y->cmap->N);
1205: MatSetBlockSizesFromMats(B,Y,Y);
1206: MatSetType(B,(MatType) ((PetscObject)Y)->type_name);
1207: MatAXPYGetPreallocation_SeqSBAIJ(Y,X,nnz);
1208: MatSeqSBAIJSetPreallocation(B,bs,0,nnz);
1210: MatAXPY_BasicWithPreallocation(B,Y,a,X,str);
1212: MatHeaderReplace(Y,B);
1213: PetscFree(nnz);
1214: MatRestoreRowUpperTriangular(X);
1215: MatRestoreRowUpperTriangular(Y);
1216: }
1217: return(0);
1218: }
1222: PetscErrorCode MatIsSymmetric_SeqSBAIJ(Mat A,PetscReal tol,PetscBool *flg)
1223: {
1225: *flg = PETSC_TRUE;
1226: return(0);
1227: }
1231: PetscErrorCode MatIsStructurallySymmetric_SeqSBAIJ(Mat A,PetscBool *flg)
1232: {
1234: *flg = PETSC_TRUE;
1235: return(0);
1236: }
1240: PetscErrorCode MatIsHermitian_SeqSBAIJ(Mat A,PetscReal tol,PetscBool *flg)
1241: {
1243: *flg = PETSC_FALSE;
1244: return(0);
1245: }
1249: PetscErrorCode MatRealPart_SeqSBAIJ(Mat A)
1250: {
1251: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
1252: PetscInt i,nz = a->bs2*a->i[a->mbs];
1253: MatScalar *aa = a->a;
1256: for (i=0; i<nz; i++) aa[i] = PetscRealPart(aa[i]);
1257: return(0);
1258: }
1262: PetscErrorCode MatImaginaryPart_SeqSBAIJ(Mat A)
1263: {
1264: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
1265: PetscInt i,nz = a->bs2*a->i[a->mbs];
1266: MatScalar *aa = a->a;
1269: for (i=0; i<nz; i++) aa[i] = PetscImaginaryPart(aa[i]);
1270: return(0);
1271: }
1275: PetscErrorCode MatZeroRowsColumns_SeqSBAIJ(Mat A,PetscInt is_n,const PetscInt is_idx[],PetscScalar diag,Vec x, Vec b)
1276: {
1277: Mat_SeqSBAIJ *baij=(Mat_SeqSBAIJ*)A->data;
1278: PetscErrorCode ierr;
1279: PetscInt i,j,k,count;
1280: PetscInt bs =A->rmap->bs,bs2=baij->bs2,row,col;
1281: PetscScalar zero = 0.0;
1282: MatScalar *aa;
1283: const PetscScalar *xx;
1284: PetscScalar *bb;
1285: PetscBool *zeroed,vecs = PETSC_FALSE;
1288: /* fix right hand side if needed */
1289: if (x && b) {
1290: VecGetArrayRead(x,&xx);
1291: VecGetArray(b,&bb);
1292: vecs = PETSC_TRUE;
1293: }
1295: /* zero the columns */
1296: PetscCalloc1(A->rmap->n,&zeroed);
1297: for (i=0; i<is_n; i++) {
1298: 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]);
1299: zeroed[is_idx[i]] = PETSC_TRUE;
1300: }
1301: if (vecs) {
1302: for (i=0; i<A->rmap->N; i++) {
1303: row = i/bs;
1304: for (j=baij->i[row]; j<baij->i[row+1]; j++) {
1305: for (k=0; k<bs; k++) {
1306: col = bs*baij->j[j] + k;
1307: if (col <= i) continue;
1308: aa = ((MatScalar*)(baij->a)) + j*bs2 + (i%bs) + bs*k;
1309: if (!zeroed[i] && zeroed[col]) bb[i] -= aa[0]*xx[col];
1310: if (zeroed[i] && !zeroed[col]) bb[col] -= aa[0]*xx[i];
1311: }
1312: }
1313: }
1314: for (i=0; i<is_n; i++) bb[is_idx[i]] = diag*xx[is_idx[i]];
1315: }
1317: for (i=0; i<A->rmap->N; i++) {
1318: if (!zeroed[i]) {
1319: row = i/bs;
1320: for (j=baij->i[row]; j<baij->i[row+1]; j++) {
1321: for (k=0; k<bs; k++) {
1322: col = bs*baij->j[j] + k;
1323: if (zeroed[col]) {
1324: aa = ((MatScalar*)(baij->a)) + j*bs2 + (i%bs) + bs*k;
1325: aa[0] = 0.0;
1326: }
1327: }
1328: }
1329: }
1330: }
1331: PetscFree(zeroed);
1332: if (vecs) {
1333: VecRestoreArrayRead(x,&xx);
1334: VecRestoreArray(b,&bb);
1335: }
1337: /* zero the rows */
1338: for (i=0; i<is_n; i++) {
1339: row = is_idx[i];
1340: count = (baij->i[row/bs +1] - baij->i[row/bs])*bs;
1341: aa = ((MatScalar*)(baij->a)) + baij->i[row/bs]*bs2 + (row%bs);
1342: for (k=0; k<count; k++) {
1343: aa[0] = zero;
1344: aa += bs;
1345: }
1346: if (diag != 0.0) {
1347: (*A->ops->setvalues)(A,1,&row,1,&row,&diag,INSERT_VALUES);
1348: }
1349: }
1350: MatAssemblyEnd_SeqSBAIJ(A,MAT_FINAL_ASSEMBLY);
1351: return(0);
1352: }
1356: PetscErrorCode MatShift_SeqSBAIJ(Mat Y,PetscScalar a)
1357: {
1359: Mat_SeqSBAIJ *aij = (Mat_SeqSBAIJ*)Y->data;
1362: if (!aij->nz) {
1363: MatSeqSBAIJSetPreallocation(Y,Y->rmap->bs,1,NULL);
1364: }
1365: MatShift_Basic(Y,a);
1366: return(0);
1367: }
1369: /* -------------------------------------------------------------------*/
1370: static struct _MatOps MatOps_Values = {MatSetValues_SeqSBAIJ,
1371: MatGetRow_SeqSBAIJ,
1372: MatRestoreRow_SeqSBAIJ,
1373: MatMult_SeqSBAIJ_N,
1374: /* 4*/ MatMultAdd_SeqSBAIJ_N,
1375: MatMult_SeqSBAIJ_N, /* transpose versions are same as non-transpose versions */
1376: MatMultAdd_SeqSBAIJ_N,
1377: 0,
1378: 0,
1379: 0,
1380: /* 10*/ 0,
1381: 0,
1382: MatCholeskyFactor_SeqSBAIJ,
1383: MatSOR_SeqSBAIJ,
1384: MatTranspose_SeqSBAIJ,
1385: /* 15*/ MatGetInfo_SeqSBAIJ,
1386: MatEqual_SeqSBAIJ,
1387: MatGetDiagonal_SeqSBAIJ,
1388: MatDiagonalScale_SeqSBAIJ,
1389: MatNorm_SeqSBAIJ,
1390: /* 20*/ 0,
1391: MatAssemblyEnd_SeqSBAIJ,
1392: MatSetOption_SeqSBAIJ,
1393: MatZeroEntries_SeqSBAIJ,
1394: /* 24*/ 0,
1395: 0,
1396: 0,
1397: 0,
1398: 0,
1399: /* 29*/ MatSetUp_SeqSBAIJ,
1400: 0,
1401: 0,
1402: 0,
1403: 0,
1404: /* 34*/ MatDuplicate_SeqSBAIJ,
1405: 0,
1406: 0,
1407: 0,
1408: MatICCFactor_SeqSBAIJ,
1409: /* 39*/ MatAXPY_SeqSBAIJ,
1410: MatGetSubMatrices_SeqSBAIJ,
1411: MatIncreaseOverlap_SeqSBAIJ,
1412: MatGetValues_SeqSBAIJ,
1413: MatCopy_SeqSBAIJ,
1414: /* 44*/ 0,
1415: MatScale_SeqSBAIJ,
1416: MatShift_SeqSBAIJ,
1417: 0,
1418: MatZeroRowsColumns_SeqSBAIJ,
1419: /* 49*/ 0,
1420: MatGetRowIJ_SeqSBAIJ,
1421: MatRestoreRowIJ_SeqSBAIJ,
1422: 0,
1423: 0,
1424: /* 54*/ 0,
1425: 0,
1426: 0,
1427: 0,
1428: MatSetValuesBlocked_SeqSBAIJ,
1429: /* 59*/ MatGetSubMatrix_SeqSBAIJ,
1430: 0,
1431: 0,
1432: 0,
1433: 0,
1434: /* 64*/ 0,
1435: 0,
1436: 0,
1437: 0,
1438: 0,
1439: /* 69*/ MatGetRowMaxAbs_SeqSBAIJ,
1440: 0,
1441: 0,
1442: 0,
1443: 0,
1444: /* 74*/ 0,
1445: 0,
1446: 0,
1447: 0,
1448: 0,
1449: /* 79*/ 0,
1450: 0,
1451: 0,
1452: MatGetInertia_SeqSBAIJ,
1453: MatLoad_SeqSBAIJ,
1454: /* 84*/ MatIsSymmetric_SeqSBAIJ,
1455: MatIsHermitian_SeqSBAIJ,
1456: MatIsStructurallySymmetric_SeqSBAIJ,
1457: 0,
1458: 0,
1459: /* 89*/ 0,
1460: 0,
1461: 0,
1462: 0,
1463: 0,
1464: /* 94*/ 0,
1465: 0,
1466: 0,
1467: 0,
1468: 0,
1469: /* 99*/ 0,
1470: 0,
1471: 0,
1472: 0,
1473: 0,
1474: /*104*/ 0,
1475: MatRealPart_SeqSBAIJ,
1476: MatImaginaryPart_SeqSBAIJ,
1477: MatGetRowUpperTriangular_SeqSBAIJ,
1478: MatRestoreRowUpperTriangular_SeqSBAIJ,
1479: /*109*/ 0,
1480: 0,
1481: 0,
1482: 0,
1483: MatMissingDiagonal_SeqSBAIJ,
1484: /*114*/ 0,
1485: 0,
1486: 0,
1487: 0,
1488: 0,
1489: /*119*/ 0,
1490: 0,
1491: 0,
1492: 0,
1493: 0,
1494: /*124*/ 0,
1495: 0,
1496: 0,
1497: 0,
1498: 0,
1499: /*129*/ 0,
1500: 0,
1501: 0,
1502: 0,
1503: 0,
1504: /*134*/ 0,
1505: 0,
1506: 0,
1507: 0,
1508: 0,
1509: /*139*/ 0,
1510: 0,
1511: 0,
1512: 0,
1513: 0,
1514: /*144*/MatCreateMPIMatConcatenateSeqMat_SeqSBAIJ
1515: };
1519: PetscErrorCode MatStoreValues_SeqSBAIJ(Mat mat)
1520: {
1521: Mat_SeqSBAIJ *aij = (Mat_SeqSBAIJ*)mat->data;
1522: PetscInt nz = aij->i[mat->rmap->N]*mat->rmap->bs*aij->bs2;
1526: if (aij->nonew != 1) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Must call MatSetOption(A,MAT_NEW_NONZERO_LOCATIONS,PETSC_FALSE);first");
1528: /* allocate space for values if not already there */
1529: if (!aij->saved_values) {
1530: PetscMalloc1(nz+1,&aij->saved_values);
1531: }
1533: /* copy values over */
1534: PetscMemcpy(aij->saved_values,aij->a,nz*sizeof(PetscScalar));
1535: return(0);
1536: }
1540: PetscErrorCode MatRetrieveValues_SeqSBAIJ(Mat mat)
1541: {
1542: Mat_SeqSBAIJ *aij = (Mat_SeqSBAIJ*)mat->data;
1544: PetscInt nz = aij->i[mat->rmap->N]*mat->rmap->bs*aij->bs2;
1547: if (aij->nonew != 1) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Must call MatSetOption(A,MAT_NEW_NONZERO_LOCATIONS,PETSC_FALSE);first");
1548: if (!aij->saved_values) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Must call MatStoreValues(A);first");
1550: /* copy values over */
1551: PetscMemcpy(aij->a,aij->saved_values,nz*sizeof(PetscScalar));
1552: return(0);
1553: }
1557: PetscErrorCode MatSeqSBAIJSetPreallocation_SeqSBAIJ(Mat B,PetscInt bs,PetscInt nz,PetscInt *nnz)
1558: {
1559: Mat_SeqSBAIJ *b = (Mat_SeqSBAIJ*)B->data;
1561: PetscInt i,mbs,nbs,bs2;
1562: PetscBool skipallocation = PETSC_FALSE,flg = PETSC_FALSE,realalloc = PETSC_FALSE;
1565: if (nz >= 0 || nnz) realalloc = PETSC_TRUE;
1566: B->preallocated = PETSC_TRUE;
1568: MatSetBlockSize(B,PetscAbs(bs));
1569: PetscLayoutSetUp(B->rmap);
1570: PetscLayoutSetUp(B->cmap);
1571: PetscLayoutGetBlockSize(B->rmap,&bs);
1573: mbs = B->rmap->N/bs;
1574: nbs = B->cmap->n/bs;
1575: bs2 = bs*bs;
1577: 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");
1579: if (nz == MAT_SKIP_ALLOCATION) {
1580: skipallocation = PETSC_TRUE;
1581: nz = 0;
1582: }
1584: if (nz == PETSC_DEFAULT || nz == PETSC_DECIDE) nz = 3;
1585: if (nz < 0) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"nz cannot be less than 0: value %D",nz);
1586: if (nnz) {
1587: for (i=0; i<mbs; i++) {
1588: 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]);
1589: 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);
1590: }
1591: }
1593: B->ops->mult = MatMult_SeqSBAIJ_N;
1594: B->ops->multadd = MatMultAdd_SeqSBAIJ_N;
1595: B->ops->multtranspose = MatMult_SeqSBAIJ_N;
1596: B->ops->multtransposeadd = MatMultAdd_SeqSBAIJ_N;
1598: PetscOptionsGetBool(((PetscObject)B)->prefix,"-mat_no_unroll",&flg,NULL);
1599: if (!flg) {
1600: switch (bs) {
1601: case 1:
1602: B->ops->mult = MatMult_SeqSBAIJ_1;
1603: B->ops->multadd = MatMultAdd_SeqSBAIJ_1;
1604: B->ops->multtranspose = MatMult_SeqSBAIJ_1;
1605: B->ops->multtransposeadd = MatMultAdd_SeqSBAIJ_1;
1606: break;
1607: case 2:
1608: B->ops->mult = MatMult_SeqSBAIJ_2;
1609: B->ops->multadd = MatMultAdd_SeqSBAIJ_2;
1610: B->ops->multtranspose = MatMult_SeqSBAIJ_2;
1611: B->ops->multtransposeadd = MatMultAdd_SeqSBAIJ_2;
1612: break;
1613: case 3:
1614: B->ops->mult = MatMult_SeqSBAIJ_3;
1615: B->ops->multadd = MatMultAdd_SeqSBAIJ_3;
1616: B->ops->multtranspose = MatMult_SeqSBAIJ_3;
1617: B->ops->multtransposeadd = MatMultAdd_SeqSBAIJ_3;
1618: break;
1619: case 4:
1620: B->ops->mult = MatMult_SeqSBAIJ_4;
1621: B->ops->multadd = MatMultAdd_SeqSBAIJ_4;
1622: B->ops->multtranspose = MatMult_SeqSBAIJ_4;
1623: B->ops->multtransposeadd = MatMultAdd_SeqSBAIJ_4;
1624: break;
1625: case 5:
1626: B->ops->mult = MatMult_SeqSBAIJ_5;
1627: B->ops->multadd = MatMultAdd_SeqSBAIJ_5;
1628: B->ops->multtranspose = MatMult_SeqSBAIJ_5;
1629: B->ops->multtransposeadd = MatMultAdd_SeqSBAIJ_5;
1630: break;
1631: case 6:
1632: B->ops->mult = MatMult_SeqSBAIJ_6;
1633: B->ops->multadd = MatMultAdd_SeqSBAIJ_6;
1634: B->ops->multtranspose = MatMult_SeqSBAIJ_6;
1635: B->ops->multtransposeadd = MatMultAdd_SeqSBAIJ_6;
1636: break;
1637: case 7:
1638: B->ops->mult = MatMult_SeqSBAIJ_7;
1639: B->ops->multadd = MatMultAdd_SeqSBAIJ_7;
1640: B->ops->multtranspose = MatMult_SeqSBAIJ_7;
1641: B->ops->multtransposeadd = MatMultAdd_SeqSBAIJ_7;
1642: break;
1643: }
1644: }
1646: b->mbs = mbs;
1647: b->nbs = nbs;
1648: if (!skipallocation) {
1649: if (!b->imax) {
1650: PetscMalloc2(mbs,&b->imax,mbs,&b->ilen);
1652: b->free_imax_ilen = PETSC_TRUE;
1654: PetscLogObjectMemory((PetscObject)B,2*mbs*sizeof(PetscInt));
1655: }
1656: if (!nnz) {
1657: if (nz == PETSC_DEFAULT || nz == PETSC_DECIDE) nz = 5;
1658: else if (nz <= 0) nz = 1;
1659: for (i=0; i<mbs; i++) b->imax[i] = nz;
1660: nz = nz*mbs; /* total nz */
1661: } else {
1662: nz = 0;
1663: for (i=0; i<mbs; i++) {b->imax[i] = nnz[i]; nz += nnz[i];}
1664: }
1665: /* b->ilen will count nonzeros in each block row so far. */
1666: for (i=0; i<mbs; i++) b->ilen[i] = 0;
1667: /* nz=(nz+mbs)/2; */ /* total diagonal and superdiagonal nonzero blocks */
1669: /* allocate the matrix space */
1670: MatSeqXAIJFreeAIJ(B,&b->a,&b->j,&b->i);
1671: PetscMalloc3(bs2*nz,&b->a,nz,&b->j,B->rmap->N+1,&b->i);
1672: PetscLogObjectMemory((PetscObject)B,(B->rmap->N+1)*sizeof(PetscInt)+nz*(bs2*sizeof(PetscScalar)+sizeof(PetscInt)));
1673: PetscMemzero(b->a,nz*bs2*sizeof(MatScalar));
1674: PetscMemzero(b->j,nz*sizeof(PetscInt));
1676: b->singlemalloc = PETSC_TRUE;
1678: /* pointer to beginning of each row */
1679: b->i[0] = 0;
1680: for (i=1; i<mbs+1; i++) b->i[i] = b->i[i-1] + b->imax[i-1];
1682: b->free_a = PETSC_TRUE;
1683: b->free_ij = PETSC_TRUE;
1684: } else {
1685: b->free_a = PETSC_FALSE;
1686: b->free_ij = PETSC_FALSE;
1687: }
1689: B->rmap->bs = bs;
1690: b->bs2 = bs2;
1691: b->nz = 0;
1692: b->maxnz = nz;
1694: b->inew = 0;
1695: b->jnew = 0;
1696: b->anew = 0;
1697: b->a2anew = 0;
1698: b->permute = PETSC_FALSE;
1699: if (realalloc) {MatSetOption(B,MAT_NEW_NONZERO_ALLOCATION_ERR,PETSC_TRUE);}
1700: return(0);
1701: }
1705: PetscErrorCode MatSeqSBAIJSetPreallocationCSR_SeqSBAIJ(Mat B,PetscInt bs,const PetscInt ii[],const PetscInt jj[], const PetscScalar V[])
1706: {
1707: PetscInt i,j,m,nz,nz_max=0,*nnz;
1708: PetscScalar *values=0;
1709: PetscBool roworiented = ((Mat_SeqSBAIJ*)B->data)->roworiented;
1712: if (bs < 1) SETERRQ1(PetscObjectComm((PetscObject)B),PETSC_ERR_ARG_OUTOFRANGE,"Invalid block size specified, must be positive but it is %D",bs);
1713: PetscLayoutSetBlockSize(B->rmap,bs);
1714: PetscLayoutSetBlockSize(B->cmap,bs);
1715: PetscLayoutSetUp(B->rmap);
1716: PetscLayoutSetUp(B->cmap);
1717: PetscLayoutGetBlockSize(B->rmap,&bs);
1718: m = B->rmap->n/bs;
1720: if (ii[0]) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"ii[0] must be 0 but it is %D",ii[0]);
1721: PetscMalloc1(m+1,&nnz);
1722: for (i=0; i<m; i++) {
1723: nz = ii[i+1] - ii[i];
1724: if (nz < 0) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Row %D has a negative number of columns %D",i,nz);
1725: nz_max = PetscMax(nz_max,nz);
1726: nnz[i] = nz;
1727: }
1728: MatSeqSBAIJSetPreallocation(B,bs,0,nnz);
1729: PetscFree(nnz);
1731: values = (PetscScalar*)V;
1732: if (!values) {
1733: PetscCalloc1(bs*bs*nz_max,&values);
1734: }
1735: for (i=0; i<m; i++) {
1736: PetscInt ncols = ii[i+1] - ii[i];
1737: const PetscInt *icols = jj + ii[i];
1738: if (!roworiented || bs == 1) {
1739: const PetscScalar *svals = values + (V ? (bs*bs*ii[i]) : 0);
1740: MatSetValuesBlocked_SeqSBAIJ(B,1,&i,ncols,icols,svals,INSERT_VALUES);
1741: } else {
1742: for (j=0; j<ncols; j++) {
1743: const PetscScalar *svals = values + (V ? (bs*bs*(ii[i]+j)) : 0);
1744: MatSetValuesBlocked_SeqSBAIJ(B,1,&i,1,&icols[j],svals,INSERT_VALUES);
1745: }
1746: }
1747: }
1748: if (!V) { PetscFree(values); }
1749: MatAssemblyBegin(B,MAT_FINAL_ASSEMBLY);
1750: MatAssemblyEnd(B,MAT_FINAL_ASSEMBLY);
1751: MatSetOption(B,MAT_NEW_NONZERO_LOCATION_ERR,PETSC_TRUE);
1752: return(0);
1753: }
1755: /*
1756: This is used to set the numeric factorization for both Cholesky and ICC symbolic factorization
1757: */
1760: PetscErrorCode MatSeqSBAIJSetNumericFactorization_inplace(Mat B,PetscBool natural)
1761: {
1763: PetscBool flg = PETSC_FALSE;
1764: PetscInt bs = B->rmap->bs;
1767: PetscOptionsGetBool(((PetscObject)B)->prefix,"-mat_no_unroll",&flg,NULL);
1768: if (flg) bs = 8;
1770: if (!natural) {
1771: switch (bs) {
1772: case 1:
1773: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_1_inplace;
1774: break;
1775: case 2:
1776: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_2;
1777: break;
1778: case 3:
1779: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_3;
1780: break;
1781: case 4:
1782: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_4;
1783: break;
1784: case 5:
1785: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_5;
1786: break;
1787: case 6:
1788: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_6;
1789: break;
1790: case 7:
1791: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_7;
1792: break;
1793: default:
1794: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_N;
1795: break;
1796: }
1797: } else {
1798: switch (bs) {
1799: case 1:
1800: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_1_NaturalOrdering_inplace;
1801: break;
1802: case 2:
1803: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_2_NaturalOrdering;
1804: break;
1805: case 3:
1806: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_3_NaturalOrdering;
1807: break;
1808: case 4:
1809: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_4_NaturalOrdering;
1810: break;
1811: case 5:
1812: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_5_NaturalOrdering;
1813: break;
1814: case 6:
1815: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_6_NaturalOrdering;
1816: break;
1817: case 7:
1818: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_7_NaturalOrdering;
1819: break;
1820: default:
1821: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_N_NaturalOrdering;
1822: break;
1823: }
1824: }
1825: return(0);
1826: }
1828: PETSC_EXTERN PetscErrorCode MatConvert_SeqSBAIJ_SeqAIJ(Mat, MatType,MatReuse,Mat*);
1829: PETSC_EXTERN PetscErrorCode MatConvert_SeqSBAIJ_SeqBAIJ(Mat, MatType,MatReuse,Mat*);
1833: PETSC_EXTERN PetscErrorCode MatGetFactor_seqsbaij_petsc(Mat A,MatFactorType ftype,Mat *B)
1834: {
1835: PetscInt n = A->rmap->n;
1839: #if defined(PETSC_USE_COMPLEX)
1840: if (A->hermitian) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Hermitian Factor is not supported");
1841: #endif
1842: MatCreate(PetscObjectComm((PetscObject)A),B);
1843: MatSetSizes(*B,n,n,n,n);
1844: if (ftype == MAT_FACTOR_CHOLESKY || ftype == MAT_FACTOR_ICC) {
1845: MatSetType(*B,MATSEQSBAIJ);
1846: MatSeqSBAIJSetPreallocation(*B,A->rmap->bs,MAT_SKIP_ALLOCATION,NULL);
1848: (*B)->ops->choleskyfactorsymbolic = MatCholeskyFactorSymbolic_SeqSBAIJ;
1849: (*B)->ops->iccfactorsymbolic = MatICCFactorSymbolic_SeqSBAIJ;
1850: } else SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Factor type not supported");
1851: (*B)->factortype = ftype;
1852: return(0);
1853: }
1855: /*MC
1856: MATSEQSBAIJ - MATSEQSBAIJ = "seqsbaij" - A matrix type to be used for sequential symmetric block sparse matrices,
1857: based on block compressed sparse row format. Only the upper triangular portion of the matrix is stored.
1859: For complex numbers by default this matrix is symmetric, NOT Hermitian symmetric. To make it Hermitian symmetric you
1860: can call MatSetOption(Mat, MAT_HERMITIAN); after MatAssemblyEnd()
1862: Options Database Keys:
1863: . -mat_type seqsbaij - sets the matrix type to "seqsbaij" during a call to MatSetFromOptions()
1865: Notes: By default if you insert values into the lower triangular part of the matrix they are simply ignored (since they are not
1866: stored and it is assumed they symmetric to the upper triangular). If you call MatSetOption(Mat,MAT_IGNORE_LOWER_TRIANGULAR,PETSC_FALSE) or use
1867: 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.
1870: Level: beginner
1872: .seealso: MatCreateSeqSBAIJ
1873: M*/
1875: PETSC_EXTERN PetscErrorCode MatConvert_SeqSBAIJ_SeqSBSTRM(Mat, MatType,MatReuse,Mat*);
1879: PETSC_EXTERN PetscErrorCode MatCreate_SeqSBAIJ(Mat B)
1880: {
1881: Mat_SeqSBAIJ *b;
1883: PetscMPIInt size;
1884: PetscBool no_unroll = PETSC_FALSE,no_inode = PETSC_FALSE;
1887: MPI_Comm_size(PetscObjectComm((PetscObject)B),&size);
1888: if (size > 1) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"Comm must be of size 1");
1890: PetscNewLog(B,&b);
1891: B->data = (void*)b;
1892: PetscMemcpy(B->ops,&MatOps_Values,sizeof(struct _MatOps));
1894: B->ops->destroy = MatDestroy_SeqSBAIJ;
1895: B->ops->view = MatView_SeqSBAIJ;
1896: b->row = 0;
1897: b->icol = 0;
1898: b->reallocs = 0;
1899: b->saved_values = 0;
1900: b->inode.limit = 5;
1901: b->inode.max_limit = 5;
1903: b->roworiented = PETSC_TRUE;
1904: b->nonew = 0;
1905: b->diag = 0;
1906: b->solve_work = 0;
1907: b->mult_work = 0;
1908: B->spptr = 0;
1909: B->info.nz_unneeded = (PetscReal)b->maxnz*b->bs2;
1910: b->keepnonzeropattern = PETSC_FALSE;
1912: b->inew = 0;
1913: b->jnew = 0;
1914: b->anew = 0;
1915: b->a2anew = 0;
1916: b->permute = PETSC_FALSE;
1918: b->ignore_ltriangular = PETSC_TRUE;
1920: PetscOptionsGetBool(((PetscObject)B)->prefix,"-mat_ignore_lower_triangular",&b->ignore_ltriangular,NULL);
1922: b->getrow_utriangular = PETSC_FALSE;
1924: PetscOptionsGetBool(((PetscObject)B)->prefix,"-mat_getrow_uppertriangular",&b->getrow_utriangular,NULL);
1926: PetscObjectComposeFunction((PetscObject)B,"MatStoreValues_C",MatStoreValues_SeqSBAIJ);
1927: PetscObjectComposeFunction((PetscObject)B,"MatRetrieveValues_C",MatRetrieveValues_SeqSBAIJ);
1928: PetscObjectComposeFunction((PetscObject)B,"MatSeqSBAIJSetColumnIndices_C",MatSeqSBAIJSetColumnIndices_SeqSBAIJ);
1929: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqsbaij_seqaij_C",MatConvert_SeqSBAIJ_SeqAIJ);
1930: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqsbaij_seqbaij_C",MatConvert_SeqSBAIJ_SeqBAIJ);
1931: PetscObjectComposeFunction((PetscObject)B,"MatSeqSBAIJSetPreallocation_C",MatSeqSBAIJSetPreallocation_SeqSBAIJ);
1932: PetscObjectComposeFunction((PetscObject)B,"MatSeqSBAIJSetPreallocationCSR_C",MatSeqSBAIJSetPreallocationCSR_SeqSBAIJ);
1933: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqsbaij_seqsbstrm_C",MatConvert_SeqSBAIJ_SeqSBSTRM);
1934: #if defined(PETSC_HAVE_ELEMENTAL)
1935: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqsbaij_elemental_C",MatConvert_SeqSBAIJ_Elemental);
1936: #endif
1938: B->symmetric = PETSC_TRUE;
1939: B->structurally_symmetric = PETSC_TRUE;
1940: B->symmetric_set = PETSC_TRUE;
1941: B->structurally_symmetric_set = PETSC_TRUE;
1943: PetscObjectChangeTypeName((PetscObject)B,MATSEQSBAIJ);
1945: PetscOptionsBegin(PetscObjectComm((PetscObject)B),((PetscObject)B)->prefix,"Options for SEQSBAIJ matrix","Mat");
1946: PetscOptionsBool("-mat_no_unroll","Do not optimize for inodes (slower)",NULL,no_unroll,&no_unroll,NULL);
1947: if (no_unroll) {
1948: PetscInfo(B,"Not using Inode routines due to -mat_no_unroll\n");
1949: }
1950: PetscOptionsBool("-mat_no_inode","Do not optimize for inodes (slower)",NULL,no_inode,&no_inode,NULL);
1951: if (no_inode) {
1952: PetscInfo(B,"Not using Inode routines due to -mat_no_inode\n");
1953: }
1954: PetscOptionsInt("-mat_inode_limit","Do not use inodes larger then this value",NULL,b->inode.limit,&b->inode.limit,NULL);
1955: PetscOptionsEnd();
1956: b->inode.use = (PetscBool)(!(no_unroll || no_inode));
1957: if (b->inode.limit > b->inode.max_limit) b->inode.limit = b->inode.max_limit;
1958: return(0);
1959: }
1963: /*@C
1964: MatSeqSBAIJSetPreallocation - Creates a sparse symmetric matrix in block AIJ (block
1965: compressed row) format. For good matrix assembly performance the
1966: user should preallocate the matrix storage by setting the parameter nz
1967: (or the array nnz). By setting these parameters accurately, performance
1968: during matrix assembly can be increased by more than a factor of 50.
1970: Collective on Mat
1972: Input Parameters:
1973: + B - the symmetric matrix
1974: . bs - size of block, the blocks are ALWAYS square. One can use MatSetBlockSizes() to set a different row and column blocksize but the row
1975: blocksize always defines the size of the blocks. The column blocksize sets the blocksize of the vectors obtained with MatCreateVecs()
1976: . nz - number of block nonzeros per block row (same for all rows)
1977: - nnz - array containing the number of block nonzeros in the upper triangular plus
1978: diagonal portion of each block (possibly different for each block row) or NULL
1980: Options Database Keys:
1981: . -mat_no_unroll - uses code that does not unroll the loops in the
1982: block calculations (much slower)
1983: . -mat_block_size - size of the blocks to use (only works if a negative bs is passed in
1985: Level: intermediate
1987: Notes:
1988: Specify the preallocated storage with either nz or nnz (not both).
1989: Set nz=PETSC_DEFAULT and nnz=NULL for PETSc to control dynamic memory
1990: allocation. See Users-Manual: ch_mat for details.
1992: You can call MatGetInfo() to get information on how effective the preallocation was;
1993: for example the fields mallocs,nz_allocated,nz_used,nz_unneeded;
1994: You can also run with the option -info and look for messages with the string
1995: malloc in them to see if additional memory allocation was needed.
1997: If the nnz parameter is given then the nz parameter is ignored
2000: .seealso: MatCreate(), MatCreateSeqAIJ(), MatSetValues(), MatCreateSBAIJ()
2001: @*/
2002: PetscErrorCode MatSeqSBAIJSetPreallocation(Mat B,PetscInt bs,PetscInt nz,const PetscInt nnz[])
2003: {
2010: PetscTryMethod(B,"MatSeqSBAIJSetPreallocation_C",(Mat,PetscInt,PetscInt,const PetscInt[]),(B,bs,nz,nnz));
2011: return(0);
2012: }
2014: #undef __FUNCT__
2016: /*@C
2017: MatSeqSBAIJSetPreallocationCSR - Allocates memory for a sparse sequential matrix in symmetric block AIJ format.
2019: Input Parameters:
2020: + B - the matrix
2021: . i - the indices into j for the start of each local row (starts with zero)
2022: . j - the column indices for each local row (starts with zero) these must be sorted for each row
2023: - v - optional values in the matrix
2025: Level: developer
2027: Notes:
2028: The order of the entries in values is specified by the MatOption MAT_ROW_ORIENTED. For example, C programs
2029: may want to use the default MAT_ROW_ORIENTED=PETSC_TRUE and use an array v[nnz][bs][bs] where the second index is
2030: over rows within a block and the last index is over columns within a block row. Fortran programs will likely set
2031: MAT_ROW_ORIENTED=PETSC_FALSE and use a Fortran array v(bs,bs,nnz) in which the first index is over rows within a
2032: block column and the second index is over columns within a block.
2034: .keywords: matrix, block, aij, compressed row, sparse
2036: .seealso: MatCreate(), MatCreateSeqSBAIJ(), MatSetValuesBlocked(), MatSeqSBAIJSetPreallocation(), MATSEQSBAIJ
2037: @*/
2038: PetscErrorCode MatSeqSBAIJSetPreallocationCSR(Mat B,PetscInt bs,const PetscInt i[],const PetscInt j[], const PetscScalar v[])
2039: {
2046: PetscTryMethod(B,"MatSeqSBAIJSetPreallocationCSR_C",(Mat,PetscInt,const PetscInt[],const PetscInt[],const PetscScalar[]),(B,bs,i,j,v));
2047: return(0);
2048: }
2052: /*@C
2053: MatCreateSeqSBAIJ - Creates a sparse symmetric matrix in block AIJ (block
2054: compressed row) format. For good matrix assembly performance the
2055: user should preallocate the matrix storage by setting the parameter nz
2056: (or the array nnz). By setting these parameters accurately, performance
2057: during matrix assembly can be increased by more than a factor of 50.
2059: Collective on MPI_Comm
2061: Input Parameters:
2062: + comm - MPI communicator, set to PETSC_COMM_SELF
2063: . bs - size of block, the blocks are ALWAYS square. One can use MatSetBlockSizes() to set a different row and column blocksize but the row
2064: blocksize always defines the size of the blocks. The column blocksize sets the blocksize of the vectors obtained with MatCreateVecs()
2065: . m - number of rows, or number of columns
2066: . nz - number of block nonzeros per block row (same for all rows)
2067: - nnz - array containing the number of block nonzeros in the upper triangular plus
2068: diagonal portion of each block (possibly different for each block row) or NULL
2070: Output Parameter:
2071: . A - the symmetric matrix
2073: Options Database Keys:
2074: . -mat_no_unroll - uses code that does not unroll the loops in the
2075: block calculations (much slower)
2076: . -mat_block_size - size of the blocks to use
2078: Level: intermediate
2080: It is recommended that one use the MatCreate(), MatSetType() and/or MatSetFromOptions(),
2081: MatXXXXSetPreallocation() paradgm instead of this routine directly.
2082: [MatXXXXSetPreallocation() is, for example, MatSeqAIJSetPreallocation]
2084: Notes:
2085: The number of rows and columns must be divisible by blocksize.
2086: This matrix type does not support complex Hermitian operation.
2088: Specify the preallocated storage with either nz or nnz (not both).
2089: Set nz=PETSC_DEFAULT and nnz=NULL for PETSc to control dynamic memory
2090: allocation. See Users-Manual: ch_mat for details.
2092: If the nnz parameter is given then the nz parameter is ignored
2094: .seealso: MatCreate(), MatCreateSeqAIJ(), MatSetValues(), MatCreateSBAIJ()
2095: @*/
2096: PetscErrorCode MatCreateSeqSBAIJ(MPI_Comm comm,PetscInt bs,PetscInt m,PetscInt n,PetscInt nz,const PetscInt nnz[],Mat *A)
2097: {
2101: MatCreate(comm,A);
2102: MatSetSizes(*A,m,n,m,n);
2103: MatSetType(*A,MATSEQSBAIJ);
2104: MatSeqSBAIJSetPreallocation_SeqSBAIJ(*A,bs,nz,(PetscInt*)nnz);
2105: return(0);
2106: }
2110: PetscErrorCode MatDuplicate_SeqSBAIJ(Mat A,MatDuplicateOption cpvalues,Mat *B)
2111: {
2112: Mat C;
2113: Mat_SeqSBAIJ *c,*a = (Mat_SeqSBAIJ*)A->data;
2115: PetscInt i,mbs = a->mbs,nz = a->nz,bs2 =a->bs2;
2118: if (a->i[mbs] != nz) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_PLIB,"Corrupt matrix");
2120: *B = 0;
2121: MatCreate(PetscObjectComm((PetscObject)A),&C);
2122: MatSetSizes(C,A->rmap->N,A->cmap->n,A->rmap->N,A->cmap->n);
2123: MatSetType(C,MATSEQSBAIJ);
2124: PetscMemcpy(C->ops,A->ops,sizeof(struct _MatOps));
2125: c = (Mat_SeqSBAIJ*)C->data;
2127: C->preallocated = PETSC_TRUE;
2128: C->factortype = A->factortype;
2129: c->row = 0;
2130: c->icol = 0;
2131: c->saved_values = 0;
2132: c->keepnonzeropattern = a->keepnonzeropattern;
2133: C->assembled = PETSC_TRUE;
2135: PetscLayoutReference(A->rmap,&C->rmap);
2136: PetscLayoutReference(A->cmap,&C->cmap);
2137: c->bs2 = a->bs2;
2138: c->mbs = a->mbs;
2139: c->nbs = a->nbs;
2141: if (cpvalues == MAT_SHARE_NONZERO_PATTERN) {
2142: c->imax = a->imax;
2143: c->ilen = a->ilen;
2144: c->free_imax_ilen = PETSC_FALSE;
2145: } else {
2146: PetscMalloc2((mbs+1),&c->imax,(mbs+1),&c->ilen);
2147: PetscLogObjectMemory((PetscObject)C,2*(mbs+1)*sizeof(PetscInt));
2148: for (i=0; i<mbs; i++) {
2149: c->imax[i] = a->imax[i];
2150: c->ilen[i] = a->ilen[i];
2151: }
2152: c->free_imax_ilen = PETSC_TRUE;
2153: }
2155: /* allocate the matrix space */
2156: if (cpvalues == MAT_SHARE_NONZERO_PATTERN) {
2157: PetscMalloc1(bs2*nz,&c->a);
2158: PetscLogObjectMemory((PetscObject)C,nz*bs2*sizeof(MatScalar));
2159: c->i = a->i;
2160: c->j = a->j;
2161: c->singlemalloc = PETSC_FALSE;
2162: c->free_a = PETSC_TRUE;
2163: c->free_ij = PETSC_FALSE;
2164: c->parent = A;
2165: PetscObjectReference((PetscObject)A);
2166: MatSetOption(A,MAT_NEW_NONZERO_LOCATION_ERR,PETSC_TRUE);
2167: MatSetOption(C,MAT_NEW_NONZERO_LOCATION_ERR,PETSC_TRUE);
2168: } else {
2169: PetscMalloc3(bs2*nz,&c->a,nz,&c->j,mbs+1,&c->i);
2170: PetscMemcpy(c->i,a->i,(mbs+1)*sizeof(PetscInt));
2171: PetscLogObjectMemory((PetscObject)C,(mbs+1)*sizeof(PetscInt) + nz*(bs2*sizeof(MatScalar) + sizeof(PetscInt)));
2172: c->singlemalloc = PETSC_TRUE;
2173: c->free_a = PETSC_TRUE;
2174: c->free_ij = PETSC_TRUE;
2175: }
2176: if (mbs > 0) {
2177: if (cpvalues != MAT_SHARE_NONZERO_PATTERN) {
2178: PetscMemcpy(c->j,a->j,nz*sizeof(PetscInt));
2179: }
2180: if (cpvalues == MAT_COPY_VALUES) {
2181: PetscMemcpy(c->a,a->a,bs2*nz*sizeof(MatScalar));
2182: } else {
2183: PetscMemzero(c->a,bs2*nz*sizeof(MatScalar));
2184: }
2185: if (a->jshort) {
2186: /* cannot share jshort, it is reallocated in MatAssemblyEnd_SeqSBAIJ() */
2187: /* if the parent matrix is reassembled, this child matrix will never notice */
2188: PetscMalloc1(nz,&c->jshort);
2189: PetscLogObjectMemory((PetscObject)C,nz*sizeof(unsigned short));
2190: PetscMemcpy(c->jshort,a->jshort,nz*sizeof(unsigned short));
2192: c->free_jshort = PETSC_TRUE;
2193: }
2194: }
2196: c->roworiented = a->roworiented;
2197: c->nonew = a->nonew;
2199: if (a->diag) {
2200: if (cpvalues == MAT_SHARE_NONZERO_PATTERN) {
2201: c->diag = a->diag;
2202: c->free_diag = PETSC_FALSE;
2203: } else {
2204: PetscMalloc1(mbs,&c->diag);
2205: PetscLogObjectMemory((PetscObject)C,mbs*sizeof(PetscInt));
2206: for (i=0; i<mbs; i++) c->diag[i] = a->diag[i];
2207: c->free_diag = PETSC_TRUE;
2208: }
2209: }
2210: c->nz = a->nz;
2211: c->maxnz = a->nz; /* Since we allocate exactly the right amount */
2212: c->solve_work = 0;
2213: c->mult_work = 0;
2215: *B = C;
2216: PetscFunctionListDuplicate(((PetscObject)A)->qlist,&((PetscObject)C)->qlist);
2217: return(0);
2218: }
2222: PetscErrorCode MatLoad_SeqSBAIJ(Mat newmat,PetscViewer viewer)
2223: {
2224: Mat_SeqSBAIJ *a;
2226: int fd;
2227: PetscMPIInt size;
2228: PetscInt i,nz,header[4],*rowlengths=0,M,N,bs = newmat->rmap->bs;
2229: PetscInt *mask,mbs,*jj,j,rowcount,nzcount,k,*s_browlengths,maskcount;
2230: PetscInt kmax,jcount,block,idx,point,nzcountb,extra_rows,rows,cols;
2231: PetscInt *masked,nmask,tmp,bs2,ishift;
2232: PetscScalar *aa;
2233: MPI_Comm comm;
2236: /* force binary viewer to load .info file if it has not yet done so */
2237: PetscViewerSetUp(viewer);
2238: PetscObjectGetComm((PetscObject)viewer,&comm);
2239: PetscOptionsGetInt(((PetscObject)newmat)->prefix,"-matload_block_size",&bs,NULL);
2240: if (bs < 0) bs = 1;
2241: bs2 = bs*bs;
2243: MPI_Comm_size(comm,&size);
2244: if (size > 1) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"view must have one processor");
2245: PetscViewerBinaryGetDescriptor(viewer,&fd);
2246: PetscBinaryRead(fd,header,4,PETSC_INT);
2247: if (header[0] != MAT_FILE_CLASSID) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_FILE_UNEXPECTED,"not Mat object");
2248: M = header[1]; N = header[2]; nz = header[3];
2250: if (header[3] < 0) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_FILE_UNEXPECTED,"Matrix stored in special format, cannot load as SeqSBAIJ");
2252: if (M != N) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Can only do square matrices");
2254: /*
2255: This code adds extra rows to make sure the number of rows is
2256: divisible by the blocksize
2257: */
2258: mbs = M/bs;
2259: extra_rows = bs - M + bs*(mbs);
2260: if (extra_rows == bs) extra_rows = 0;
2261: else mbs++;
2262: if (extra_rows) {
2263: PetscInfo(viewer,"Padding loaded matrix to match blocksize\n");
2264: }
2266: /* Set global sizes if not already set */
2267: if (newmat->rmap->n < 0 && newmat->rmap->N < 0 && newmat->cmap->n < 0 && newmat->cmap->N < 0) {
2268: MatSetSizes(newmat,PETSC_DECIDE,PETSC_DECIDE,M+extra_rows,N+extra_rows);
2269: } else { /* Check if the matrix global sizes are correct */
2270: MatGetSize(newmat,&rows,&cols);
2271: 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);
2272: }
2274: /* read in row lengths */
2275: PetscMalloc1(M+extra_rows,&rowlengths);
2276: PetscBinaryRead(fd,rowlengths,M,PETSC_INT);
2277: for (i=0; i<extra_rows; i++) rowlengths[M+i] = 1;
2279: /* read in column indices */
2280: PetscMalloc1(nz+extra_rows,&jj);
2281: PetscBinaryRead(fd,jj,nz,PETSC_INT);
2282: for (i=0; i<extra_rows; i++) jj[nz+i] = M+i;
2284: /* loop over row lengths determining block row lengths */
2285: PetscCalloc1(mbs,&s_browlengths);
2286: PetscMalloc2(mbs,&mask,mbs,&masked);
2287: PetscMemzero(mask,mbs*sizeof(PetscInt));
2288: rowcount = 0;
2289: nzcount = 0;
2290: for (i=0; i<mbs; i++) {
2291: nmask = 0;
2292: for (j=0; j<bs; j++) {
2293: kmax = rowlengths[rowcount];
2294: for (k=0; k<kmax; k++) {
2295: tmp = jj[nzcount++]/bs; /* block col. index */
2296: if (!mask[tmp] && tmp >= i) {masked[nmask++] = tmp; mask[tmp] = 1;}
2297: }
2298: rowcount++;
2299: }
2300: s_browlengths[i] += nmask;
2302: /* zero out the mask elements we set */
2303: for (j=0; j<nmask; j++) mask[masked[j]] = 0;
2304: }
2306: /* Do preallocation */
2307: MatSeqSBAIJSetPreallocation_SeqSBAIJ(newmat,bs,0,s_browlengths);
2308: a = (Mat_SeqSBAIJ*)newmat->data;
2310: /* set matrix "i" values */
2311: a->i[0] = 0;
2312: for (i=1; i<= mbs; i++) {
2313: a->i[i] = a->i[i-1] + s_browlengths[i-1];
2314: a->ilen[i-1] = s_browlengths[i-1];
2315: }
2316: a->nz = a->i[mbs];
2318: /* read in nonzero values */
2319: PetscMalloc1(nz+extra_rows,&aa);
2320: PetscBinaryRead(fd,aa,nz,PETSC_SCALAR);
2321: for (i=0; i<extra_rows; i++) aa[nz+i] = 1.0;
2323: /* set "a" and "j" values into matrix */
2324: nzcount = 0; jcount = 0;
2325: for (i=0; i<mbs; i++) {
2326: nzcountb = nzcount;
2327: nmask = 0;
2328: for (j=0; j<bs; j++) {
2329: kmax = rowlengths[i*bs+j];
2330: for (k=0; k<kmax; k++) {
2331: tmp = jj[nzcount++]/bs; /* block col. index */
2332: if (!mask[tmp] && tmp >= i) { masked[nmask++] = tmp; mask[tmp] = 1;}
2333: }
2334: }
2335: /* sort the masked values */
2336: PetscSortInt(nmask,masked);
2338: /* set "j" values into matrix */
2339: maskcount = 1;
2340: for (j=0; j<nmask; j++) {
2341: a->j[jcount++] = masked[j];
2342: mask[masked[j]] = maskcount++;
2343: }
2345: /* set "a" values into matrix */
2346: ishift = bs2*a->i[i];
2347: for (j=0; j<bs; j++) {
2348: kmax = rowlengths[i*bs+j];
2349: for (k=0; k<kmax; k++) {
2350: tmp = jj[nzcountb]/bs; /* block col. index */
2351: if (tmp >= i) {
2352: block = mask[tmp] - 1;
2353: point = jj[nzcountb] - bs*tmp;
2354: idx = ishift + bs2*block + j + bs*point;
2355: a->a[idx] = aa[nzcountb];
2356: }
2357: nzcountb++;
2358: }
2359: }
2360: /* zero out the mask elements we set */
2361: for (j=0; j<nmask; j++) mask[masked[j]] = 0;
2362: }
2363: if (jcount != a->nz) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_FILE_UNEXPECTED,"Bad binary matrix");
2365: PetscFree(rowlengths);
2366: PetscFree(s_browlengths);
2367: PetscFree(aa);
2368: PetscFree(jj);
2369: PetscFree2(mask,masked);
2371: MatAssemblyBegin(newmat,MAT_FINAL_ASSEMBLY);
2372: MatAssemblyEnd(newmat,MAT_FINAL_ASSEMBLY);
2373: return(0);
2374: }
2378: /*@
2379: MatCreateSeqSBAIJWithArrays - Creates an sequential SBAIJ matrix using matrix elements
2380: (upper triangular entries in CSR format) provided by the user.
2382: Collective on MPI_Comm
2384: Input Parameters:
2385: + comm - must be an MPI communicator of size 1
2386: . bs - size of block
2387: . m - number of rows
2388: . n - number of columns
2389: . i - row indices
2390: . j - column indices
2391: - a - matrix values
2393: Output Parameter:
2394: . mat - the matrix
2396: Level: advanced
2398: Notes:
2399: The i, j, and a arrays are not copied by this routine, the user must free these arrays
2400: once the matrix is destroyed
2402: You cannot set new nonzero locations into this matrix, that will generate an error.
2404: The i and j indices are 0 based
2406: 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
2407: it is the regular CSR format excluding the lower triangular elements.
2409: .seealso: MatCreate(), MatCreateSBAIJ(), MatCreateSeqSBAIJ()
2411: @*/
2412: PetscErrorCode MatCreateSeqSBAIJWithArrays(MPI_Comm comm,PetscInt bs,PetscInt m,PetscInt n,PetscInt *i,PetscInt *j,PetscScalar *a,Mat *mat)
2413: {
2415: PetscInt ii;
2416: Mat_SeqSBAIJ *sbaij;
2419: if (bs != 1) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"block size %D > 1 is not supported yet",bs);
2420: if (i[0]) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"i (row indices) must start with 0");
2422: MatCreate(comm,mat);
2423: MatSetSizes(*mat,m,n,m,n);
2424: MatSetType(*mat,MATSEQSBAIJ);
2425: MatSeqSBAIJSetPreallocation_SeqSBAIJ(*mat,bs,MAT_SKIP_ALLOCATION,0);
2426: sbaij = (Mat_SeqSBAIJ*)(*mat)->data;
2427: PetscMalloc2(m,&sbaij->imax,m,&sbaij->ilen);
2428: PetscLogObjectMemory((PetscObject)*mat,2*m*sizeof(PetscInt));
2430: sbaij->i = i;
2431: sbaij->j = j;
2432: sbaij->a = a;
2434: sbaij->singlemalloc = PETSC_FALSE;
2435: sbaij->nonew = -1; /*this indicates that inserting a new value in the matrix that generates a new nonzero is an error*/
2436: sbaij->free_a = PETSC_FALSE;
2437: sbaij->free_ij = PETSC_FALSE;
2439: for (ii=0; ii<m; ii++) {
2440: sbaij->ilen[ii] = sbaij->imax[ii] = i[ii+1] - i[ii];
2441: #if defined(PETSC_USE_DEBUG)
2442: 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]);
2443: #endif
2444: }
2445: #if defined(PETSC_USE_DEBUG)
2446: for (ii=0; ii<sbaij->i[m]; ii++) {
2447: if (j[ii] < 0) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Negative column index at location = %d index = %d",ii,j[ii]);
2448: 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]);
2449: }
2450: #endif
2452: MatAssemblyBegin(*mat,MAT_FINAL_ASSEMBLY);
2453: MatAssemblyEnd(*mat,MAT_FINAL_ASSEMBLY);
2454: return(0);
2455: }
2459: PetscErrorCode MatCreateMPIMatConcatenateSeqMat_SeqSBAIJ(MPI_Comm comm,Mat inmat,PetscInt n,MatReuse scall,Mat *outmat)
2460: {
2464: MatCreateMPIMatConcatenateSeqMat_MPISBAIJ(comm,inmat,n,scall,outmat);
2465: return(0);
2466: }