Actual source code: sbaij.c
petsc-3.5.1 2014-07-24
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> /*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);
16: /*
17: Checks for missing diagonals
18: */
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");
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: }
49: PetscErrorCode MatMarkDiagonal_SeqSBAIJ(Mat A)
50: {
51: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
53: PetscInt i,j;
56: if (!a->diag) {
57: PetscMalloc1(a->mbs,&a->diag);
58: PetscLogObjectMemory((PetscObject)A,a->mbs*sizeof(PetscInt));
59: a->free_diag = PETSC_TRUE;
60: }
61: for (i=0; i<a->mbs; i++) {
62: a->diag[i] = a->i[i+1];
63: for (j=a->i[i]; j<a->i[i+1]; j++) {
64: if (a->j[j] == i) {
65: a->diag[i] = j;
66: break;
67: }
68: }
69: }
70: return(0);
71: }
75: static PetscErrorCode MatGetRowIJ_SeqSBAIJ(Mat A,PetscInt oshift,PetscBool symmetric,PetscBool blockcompressed,PetscInt *nn,const PetscInt *inia[],const PetscInt *inja[],PetscBool *done)
76: {
77: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
78: PetscInt i,j,n = a->mbs,nz = a->i[n],bs = A->rmap->bs;
79: PetscInt **ia = (PetscInt**)inia,**ja = (PetscInt**)inja;
83: *nn = n;
84: if (!ia) return(0);
85: if (!blockcompressed) {
86: /* malloc & create the natural set of indices */
87: PetscMalloc2((n+1)*bs,ia,nz*bs,ja);
88: for (i=0; i<n+1; i++) {
89: for (j=0; j<bs; j++) {
90: *ia[i*bs+j] = a->i[i]*bs+j+oshift;
91: }
92: }
93: for (i=0; i<nz; i++) {
94: for (j=0; j<bs; j++) {
95: *ja[i*bs+j] = a->j[i]*bs+j+oshift;
96: }
97: }
98: } else { /* blockcompressed */
99: if (oshift == 1) {
100: /* temporarily add 1 to i and j indices */
101: for (i=0; i<nz; i++) a->j[i]++;
102: for (i=0; i<n+1; i++) a->i[i]++;
103: }
104: *ia = a->i; *ja = a->j;
105: }
106: return(0);
107: }
111: static PetscErrorCode MatRestoreRowIJ_SeqSBAIJ(Mat A,PetscInt oshift,PetscBool symmetric,PetscBool blockcompressed,PetscInt *nn,const PetscInt *ia[],const PetscInt *ja[],PetscBool *done)
112: {
113: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
114: PetscInt i,n = a->mbs,nz = a->i[n];
118: if (!ia) return(0);
120: if (!blockcompressed) {
121: PetscFree2(*ia,*ja);
122: } else if (oshift == 1) { /* blockcompressed */
123: for (i=0; i<nz; i++) a->j[i]--;
124: for (i=0; i<n+1; i++) a->i[i]--;
125: }
126: return(0);
127: }
131: PetscErrorCode MatDestroy_SeqSBAIJ(Mat A)
132: {
133: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
137: #if defined(PETSC_USE_LOG)
138: PetscLogObjectState((PetscObject)A,"Rows=%D, NZ=%D",A->rmap->N,a->nz);
139: #endif
140: MatSeqXAIJFreeAIJ(A,&a->a,&a->j,&a->i);
141: if (a->free_diag) {PetscFree(a->diag);}
142: ISDestroy(&a->row);
143: ISDestroy(&a->col);
144: ISDestroy(&a->icol);
145: PetscFree(a->idiag);
146: PetscFree(a->inode.size);
147: if (a->free_imax_ilen) {PetscFree2(a->imax,a->ilen);}
148: PetscFree(a->solve_work);
149: PetscFree(a->sor_work);
150: PetscFree(a->solves_work);
151: PetscFree(a->mult_work);
152: PetscFree(a->saved_values);
153: PetscFree(a->xtoy);
154: if (a->free_jshort) {PetscFree(a->jshort);}
155: PetscFree(a->inew);
156: MatDestroy(&a->parent);
157: PetscFree(A->data);
159: PetscObjectChangeTypeName((PetscObject)A,0);
160: PetscObjectComposeFunction((PetscObject)A,"MatStoreValues_C",NULL);
161: PetscObjectComposeFunction((PetscObject)A,"MatRetrieveValues_C",NULL);
162: PetscObjectComposeFunction((PetscObject)A,"MatSeqSBAIJSetColumnIndices_C",NULL);
163: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqsbaij_seqaij_C",NULL);
164: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqsbaij_seqbaij_C",NULL);
165: PetscObjectComposeFunction((PetscObject)A,"MatSeqSBAIJSetPreallocation_C",NULL);
166: PetscObjectComposeFunction((PetscObject)A,"MatSeqSBAIJSetPreallocationCSR_C",NULL);
167: PetscObjectComposeFunction((PetscObject)A,"MatConvert_seqsbaij_seqsbstrm_C",NULL);
168: return(0);
169: }
173: PetscErrorCode MatSetOption_SeqSBAIJ(Mat A,MatOption op,PetscBool flg)
174: {
175: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
179: switch (op) {
180: case MAT_ROW_ORIENTED:
181: a->roworiented = flg;
182: break;
183: case MAT_KEEP_NONZERO_PATTERN:
184: a->keepnonzeropattern = flg;
185: break;
186: case MAT_NEW_NONZERO_LOCATIONS:
187: a->nonew = (flg ? 0 : 1);
188: break;
189: case MAT_NEW_NONZERO_LOCATION_ERR:
190: a->nonew = (flg ? -1 : 0);
191: break;
192: case MAT_NEW_NONZERO_ALLOCATION_ERR:
193: a->nonew = (flg ? -2 : 0);
194: break;
195: case MAT_UNUSED_NONZERO_LOCATION_ERR:
196: a->nounused = (flg ? -1 : 0);
197: break;
198: case MAT_NEW_DIAGONALS:
199: case MAT_IGNORE_OFF_PROC_ENTRIES:
200: case MAT_USE_HASH_TABLE:
201: PetscInfo1(A,"Option %s ignored\n",MatOptions[op]);
202: break;
203: case MAT_HERMITIAN:
204: if (!A->assembled) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONGSTATE,"Must call MatAssemblyEnd() first");
205: if (A->cmap->n < 65536 && A->cmap->bs == 1) {
206: A->ops->mult = MatMult_SeqSBAIJ_1_Hermitian_ushort;
207: } else if (A->cmap->bs == 1) {
208: A->ops->mult = MatMult_SeqSBAIJ_1_Hermitian;
209: } else SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"No support for Hermitian with block size greater than 1");
210: break;
211: case MAT_SPD:
212: /* These options are handled directly by MatSetOption() */
213: break;
214: case MAT_SYMMETRIC:
215: case MAT_STRUCTURALLY_SYMMETRIC:
216: case MAT_SYMMETRY_ETERNAL:
217: if (!flg) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Matrix must be symmetric");
218: PetscInfo1(A,"Option %s not relevent\n",MatOptions[op]);
219: break;
220: case MAT_IGNORE_LOWER_TRIANGULAR:
221: a->ignore_ltriangular = flg;
222: break;
223: case MAT_ERROR_LOWER_TRIANGULAR:
224: a->ignore_ltriangular = flg;
225: break;
226: case MAT_GETROW_UPPERTRIANGULAR:
227: a->getrow_utriangular = flg;
228: break;
229: default:
230: SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"unknown option %d",op);
231: }
232: return(0);
233: }
237: PetscErrorCode MatGetRow_SeqSBAIJ(Mat A,PetscInt row,PetscInt *ncols,PetscInt **cols,PetscScalar **v)
238: {
239: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
241: PetscInt itmp,i,j,k,M,*ai,*aj,bs,bn,bp,*cols_i,bs2;
242: MatScalar *aa,*aa_i;
243: PetscScalar *v_i;
246: 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()");
247: /* Get the upper triangular part of the row */
248: bs = A->rmap->bs;
249: ai = a->i;
250: aj = a->j;
251: aa = a->a;
252: bs2 = a->bs2;
254: if (row < 0 || row >= A->rmap->N) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE, "Row %D out of range", row);
256: bn = row/bs; /* Block number */
257: bp = row % bs; /* Block position */
258: M = ai[bn+1] - ai[bn];
259: *ncols = bs*M;
261: if (v) {
262: *v = 0;
263: if (*ncols) {
264: PetscMalloc1((*ncols+row),v);
265: for (i=0; i<M; i++) { /* for each block in the block row */
266: v_i = *v + i*bs;
267: aa_i = aa + bs2*(ai[bn] + i);
268: for (j=bp,k=0; j<bs2; j+=bs,k++) v_i[k] = aa_i[j];
269: }
270: }
271: }
273: if (cols) {
274: *cols = 0;
275: if (*ncols) {
276: PetscMalloc1((*ncols+row),cols);
277: for (i=0; i<M; i++) { /* for each block in the block row */
278: cols_i = *cols + i*bs;
279: itmp = bs*aj[ai[bn] + i];
280: for (j=0; j<bs; j++) cols_i[j] = itmp++;
281: }
282: }
283: }
285: /*search column A(0:row-1,row) (=A(row,0:row-1)). Could be expensive! */
286: /* this segment is currently removed, so only entries in the upper triangle are obtained */
287: #if defined(column_search)
288: v_i = *v + M*bs;
289: cols_i = *cols + M*bs;
290: for (i=0; i<bn; i++) { /* for each block row */
291: M = ai[i+1] - ai[i];
292: for (j=0; j<M; j++) {
293: itmp = aj[ai[i] + j]; /* block column value */
294: if (itmp == bn) {
295: aa_i = aa + bs2*(ai[i] + j) + bs*bp;
296: for (k=0; k<bs; k++) {
297: *cols_i++ = i*bs+k;
298: *v_i++ = aa_i[k];
299: }
300: *ncols += bs;
301: break;
302: }
303: }
304: }
305: #endif
306: return(0);
307: }
311: PetscErrorCode MatRestoreRow_SeqSBAIJ(Mat A,PetscInt row,PetscInt *nz,PetscInt **idx,PetscScalar **v)
312: {
316: if (idx) {PetscFree(*idx);}
317: if (v) {PetscFree(*v);}
318: return(0);
319: }
323: PetscErrorCode MatGetRowUpperTriangular_SeqSBAIJ(Mat A)
324: {
325: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
328: a->getrow_utriangular = PETSC_TRUE;
329: return(0);
330: }
333: PetscErrorCode MatRestoreRowUpperTriangular_SeqSBAIJ(Mat A)
334: {
335: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
338: a->getrow_utriangular = PETSC_FALSE;
339: return(0);
340: }
344: PetscErrorCode MatTranspose_SeqSBAIJ(Mat A,MatReuse reuse,Mat *B)
345: {
349: if (reuse == MAT_INITIAL_MATRIX || *B != A) {
350: MatDuplicate(A,MAT_COPY_VALUES,B);
351: }
352: return(0);
353: }
357: PetscErrorCode MatView_SeqSBAIJ_ASCII(Mat A,PetscViewer viewer)
358: {
359: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
360: PetscErrorCode ierr;
361: PetscInt i,j,bs = A->rmap->bs,k,l,bs2=a->bs2;
362: PetscViewerFormat format;
363: PetscInt *diag;
366: PetscViewerGetFormat(viewer,&format);
367: if (format == PETSC_VIEWER_ASCII_INFO || format == PETSC_VIEWER_ASCII_INFO_DETAIL) {
368: PetscViewerASCIIPrintf(viewer," block size is %D\n",bs);
369: } else if (format == PETSC_VIEWER_ASCII_MATLAB) {
370: Mat aij;
371: if (A->factortype && bs>1) {
372: PetscPrintf(PETSC_COMM_SELF,"Warning: matrix is factored with bs>1. MatView() with PETSC_VIEWER_ASCII_MATLAB is not supported and ignored!\n");
373: return(0);
374: }
375: MatConvert(A,MATSEQAIJ,MAT_INITIAL_MATRIX,&aij);
376: MatView(aij,viewer);
377: MatDestroy(&aij);
378: } else if (format == PETSC_VIEWER_ASCII_COMMON) {
379: PetscViewerASCIIUseTabs(viewer,PETSC_FALSE);
380: for (i=0; i<a->mbs; i++) {
381: for (j=0; j<bs; j++) {
382: PetscViewerASCIIPrintf(viewer,"row %D:",i*bs+j);
383: for (k=a->i[i]; k<a->i[i+1]; k++) {
384: for (l=0; l<bs; l++) {
385: #if defined(PETSC_USE_COMPLEX)
386: if (PetscImaginaryPart(a->a[bs2*k + l*bs + j]) > 0.0 && PetscRealPart(a->a[bs2*k + l*bs + j]) != 0.0) {
387: PetscViewerASCIIPrintf(viewer," (%D, %g + %g i) ",bs*a->j[k]+l,
388: (double)PetscRealPart(a->a[bs2*k + l*bs + j]),(double)PetscImaginaryPart(a->a[bs2*k + l*bs + j]));
389: } else if (PetscImaginaryPart(a->a[bs2*k + l*bs + j]) < 0.0 && PetscRealPart(a->a[bs2*k + l*bs + j]) != 0.0) {
390: PetscViewerASCIIPrintf(viewer," (%D, %g - %g i) ",bs*a->j[k]+l,
391: (double)PetscRealPart(a->a[bs2*k + l*bs + j]),-(double)PetscImaginaryPart(a->a[bs2*k + l*bs + j]));
392: } else if (PetscRealPart(a->a[bs2*k + l*bs + j]) != 0.0) {
393: PetscViewerASCIIPrintf(viewer," (%D, %g) ",bs*a->j[k]+l,(double)PetscRealPart(a->a[bs2*k + l*bs + j]));
394: }
395: #else
396: if (a->a[bs2*k + l*bs + j] != 0.0) {
397: PetscViewerASCIIPrintf(viewer," (%D, %g) ",bs*a->j[k]+l,(double)a->a[bs2*k + l*bs + j]);
398: }
399: #endif
400: }
401: }
402: PetscViewerASCIIPrintf(viewer,"\n");
403: }
404: }
405: PetscViewerASCIIUseTabs(viewer,PETSC_TRUE);
406: } else if (format == PETSC_VIEWER_ASCII_FACTOR_INFO) {
407: return(0);
408: } else {
409: PetscViewerASCIIUseTabs(viewer,PETSC_FALSE);
410: if (A->factortype) { /* for factored matrix */
411: if (bs>1) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"matrix is factored with bs>1. Not implemented yet");
413: diag=a->diag;
414: for (i=0; i<a->mbs; i++) { /* for row block i */
415: PetscViewerASCIIPrintf(viewer,"row %D:",i);
416: /* diagonal entry */
417: #if defined(PETSC_USE_COMPLEX)
418: if (PetscImaginaryPart(a->a[diag[i]]) > 0.0) {
419: 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]]));
420: } else if (PetscImaginaryPart(a->a[diag[i]]) < 0.0) {
421: 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]]));
422: } else {
423: PetscViewerASCIIPrintf(viewer," (%D, %g) ",a->j[diag[i]],(double)PetscRealPart(1.0/a->a[diag[i]]));
424: }
425: #else
426: PetscViewerASCIIPrintf(viewer," (%D, %g) ",a->j[diag[i]],(double)(1.0/a->a[diag[i]]));
427: #endif
428: /* off-diagonal entries */
429: for (k=a->i[i]; k<a->i[i+1]-1; k++) {
430: #if defined(PETSC_USE_COMPLEX)
431: if (PetscImaginaryPart(a->a[k]) > 0.0) {
432: PetscViewerASCIIPrintf(viewer," (%D, %g + %g i) ",bs*a->j[k],(double)PetscRealPart(a->a[k]),(double)PetscImaginaryPart(a->a[k]));
433: } else if (PetscImaginaryPart(a->a[k]) < 0.0) {
434: PetscViewerASCIIPrintf(viewer," (%D, %g - %g i) ",bs*a->j[k],(double)PetscRealPart(a->a[k]),-(double)PetscImaginaryPart(a->a[k]));
435: } else {
436: PetscViewerASCIIPrintf(viewer," (%D, %g) ",bs*a->j[k],(double)PetscRealPart(a->a[k]));
437: }
438: #else
439: PetscViewerASCIIPrintf(viewer," (%D, %g) ",a->j[k],(double)a->a[k]);
440: #endif
441: }
442: PetscViewerASCIIPrintf(viewer,"\n");
443: }
445: } else { /* for non-factored matrix */
446: for (i=0; i<a->mbs; i++) { /* for row block i */
447: for (j=0; j<bs; j++) { /* for row bs*i + j */
448: PetscViewerASCIIPrintf(viewer,"row %D:",i*bs+j);
449: for (k=a->i[i]; k<a->i[i+1]; k++) { /* for column block */
450: for (l=0; l<bs; l++) { /* for column */
451: #if defined(PETSC_USE_COMPLEX)
452: if (PetscImaginaryPart(a->a[bs2*k + l*bs + j]) > 0.0) {
453: PetscViewerASCIIPrintf(viewer," (%D, %g + %g i) ",bs*a->j[k]+l,
454: (double)PetscRealPart(a->a[bs2*k + l*bs + j]),(double)PetscImaginaryPart(a->a[bs2*k + l*bs + j]));
455: } else if (PetscImaginaryPart(a->a[bs2*k + l*bs + j]) < 0.0) {
456: PetscViewerASCIIPrintf(viewer," (%D, %g - %g i) ",bs*a->j[k]+l,
457: (double)PetscRealPart(a->a[bs2*k + l*bs + j]),-(double)PetscImaginaryPart(a->a[bs2*k + l*bs + j]));
458: } else {
459: PetscViewerASCIIPrintf(viewer," (%D, %g) ",bs*a->j[k]+l,(double)PetscRealPart(a->a[bs2*k + l*bs + j]));
460: }
461: #else
462: PetscViewerASCIIPrintf(viewer," (%D, %g) ",bs*a->j[k]+l,(double)a->a[bs2*k + l*bs + j]);
463: #endif
464: }
465: }
466: PetscViewerASCIIPrintf(viewer,"\n");
467: }
468: }
469: }
470: PetscViewerASCIIUseTabs(viewer,PETSC_TRUE);
471: }
472: PetscViewerFlush(viewer);
473: return(0);
474: }
476: #include <petscdraw.h>
479: static PetscErrorCode MatView_SeqSBAIJ_Draw_Zoom(PetscDraw draw,void *Aa)
480: {
481: Mat A = (Mat) Aa;
482: Mat_SeqSBAIJ *a=(Mat_SeqSBAIJ*)A->data;
484: PetscInt row,i,j,k,l,mbs=a->mbs,color,bs=A->rmap->bs,bs2=a->bs2;
485: PetscMPIInt rank;
486: PetscReal xl,yl,xr,yr,x_l,x_r,y_l,y_r;
487: MatScalar *aa;
488: MPI_Comm comm;
489: PetscViewer viewer;
492: /*
493: This is nasty. If this is called from an originally parallel matrix
494: then all processes call this,but only the first has the matrix so the
495: rest should return immediately.
496: */
497: PetscObjectGetComm((PetscObject)draw,&comm);
498: MPI_Comm_rank(comm,&rank);
499: if (rank) return(0);
501: PetscObjectQuery((PetscObject)A,"Zoomviewer",(PetscObject*)&viewer);
503: PetscDrawGetCoordinates(draw,&xl,&yl,&xr,&yr);
504: PetscDrawString(draw, .3*(xl+xr), .3*(yl+yr), PETSC_DRAW_BLACK, "symmetric");
506: /* loop over matrix elements drawing boxes */
507: color = PETSC_DRAW_BLUE;
508: for (i=0,row=0; i<mbs; i++,row+=bs) {
509: for (j=a->i[i]; j<a->i[i+1]; j++) {
510: y_l = A->rmap->N - row - 1.0; y_r = y_l + 1.0;
511: x_l = a->j[j]*bs; x_r = x_l + 1.0;
512: aa = a->a + j*bs2;
513: for (k=0; k<bs; k++) {
514: for (l=0; l<bs; l++) {
515: if (PetscRealPart(*aa++) >= 0.) continue;
516: PetscDrawRectangle(draw,x_l+k,y_l-l,x_r+k,y_r-l,color,color,color,color);
517: }
518: }
519: }
520: }
521: color = PETSC_DRAW_CYAN;
522: for (i=0,row=0; i<mbs; i++,row+=bs) {
523: for (j=a->i[i]; j<a->i[i+1]; j++) {
524: y_l = A->rmap->N - row - 1.0; y_r = y_l + 1.0;
525: x_l = a->j[j]*bs; x_r = x_l + 1.0;
526: aa = a->a + j*bs2;
527: for (k=0; k<bs; k++) {
528: for (l=0; l<bs; l++) {
529: if (PetscRealPart(*aa++) != 0.) continue;
530: PetscDrawRectangle(draw,x_l+k,y_l-l,x_r+k,y_r-l,color,color,color,color);
531: }
532: }
533: }
534: }
536: color = PETSC_DRAW_RED;
537: for (i=0,row=0; i<mbs; i++,row+=bs) {
538: for (j=a->i[i]; j<a->i[i+1]; j++) {
539: y_l = A->rmap->N - row - 1.0; y_r = y_l + 1.0;
540: x_l = a->j[j]*bs; x_r = x_l + 1.0;
541: aa = a->a + j*bs2;
542: for (k=0; k<bs; k++) {
543: for (l=0; l<bs; l++) {
544: if (PetscRealPart(*aa++) <= 0.) continue;
545: PetscDrawRectangle(draw,x_l+k,y_l-l,x_r+k,y_r-l,color,color,color,color);
546: }
547: }
548: }
549: }
550: return(0);
551: }
555: static PetscErrorCode MatView_SeqSBAIJ_Draw(Mat A,PetscViewer viewer)
556: {
558: PetscReal xl,yl,xr,yr,w,h;
559: PetscDraw draw;
560: PetscBool isnull;
563: PetscViewerDrawGetDraw(viewer,0,&draw);
564: PetscDrawIsNull(draw,&isnull); if (isnull) return(0);
566: PetscObjectCompose((PetscObject)A,"Zoomviewer",(PetscObject)viewer);
567: xr = A->rmap->N; yr = A->rmap->N; h = yr/10.0; w = xr/10.0;
568: xr += w; yr += h; xl = -w; yl = -h;
569: PetscDrawSetCoordinates(draw,xl,yl,xr,yr);
570: PetscDrawZoom(draw,MatView_SeqSBAIJ_Draw_Zoom,A);
571: PetscObjectCompose((PetscObject)A,"Zoomviewer",NULL);
572: return(0);
573: }
577: PetscErrorCode MatView_SeqSBAIJ(Mat A,PetscViewer viewer)
578: {
580: PetscBool iascii,isdraw;
581: FILE *file = 0;
584: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERASCII,&iascii);
585: PetscObjectTypeCompare((PetscObject)viewer,PETSCVIEWERDRAW,&isdraw);
586: if (iascii) {
587: MatView_SeqSBAIJ_ASCII(A,viewer);
588: } else if (isdraw) {
589: MatView_SeqSBAIJ_Draw(A,viewer);
590: } else {
591: Mat B;
592: MatConvert(A,MATSEQAIJ,MAT_INITIAL_MATRIX,&B);
593: MatView(B,viewer);
594: MatDestroy(&B);
595: PetscViewerBinaryGetInfoPointer(viewer,&file);
596: if (file) {
597: fprintf(file,"-matload_block_size %d\n",(int)A->rmap->bs);
598: }
599: }
600: return(0);
601: }
606: PetscErrorCode MatGetValues_SeqSBAIJ(Mat A,PetscInt m,const PetscInt im[],PetscInt n,const PetscInt in[],PetscScalar v[])
607: {
608: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
609: PetscInt *rp,k,low,high,t,row,nrow,i,col,l,*aj = a->j;
610: PetscInt *ai = a->i,*ailen = a->ilen;
611: PetscInt brow,bcol,ridx,cidx,bs=A->rmap->bs,bs2=a->bs2;
612: MatScalar *ap,*aa = a->a;
615: for (k=0; k<m; k++) { /* loop over rows */
616: row = im[k]; brow = row/bs;
617: if (row < 0) {v += n; continue;} /* SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Negative row: %D",row); */
618: 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);
619: rp = aj + ai[brow]; ap = aa + bs2*ai[brow];
620: nrow = ailen[brow];
621: for (l=0; l<n; l++) { /* loop over columns */
622: if (in[l] < 0) {v++; continue;} /* SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Negative column: %D",in[l]); */
623: 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);
624: col = in[l];
625: bcol = col/bs;
626: cidx = col%bs;
627: ridx = row%bs;
628: high = nrow;
629: low = 0; /* assume unsorted */
630: while (high-low > 5) {
631: t = (low+high)/2;
632: if (rp[t] > bcol) high = t;
633: else low = t;
634: }
635: for (i=low; i<high; i++) {
636: if (rp[i] > bcol) break;
637: if (rp[i] == bcol) {
638: *v++ = ap[bs2*i+bs*cidx+ridx];
639: goto finished;
640: }
641: }
642: *v++ = 0.0;
643: finished:;
644: }
645: }
646: return(0);
647: }
652: PetscErrorCode MatSetValuesBlocked_SeqSBAIJ(Mat A,PetscInt m,const PetscInt im[],PetscInt n,const PetscInt in[],const PetscScalar v[],InsertMode is)
653: {
654: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
655: PetscErrorCode ierr;
656: PetscInt *rp,k,low,high,t,ii,jj,row,nrow,i,col,l,rmax,N,lastcol = -1;
657: PetscInt *imax =a->imax,*ai=a->i,*ailen=a->ilen;
658: PetscInt *aj =a->j,nonew=a->nonew,bs2=a->bs2,bs=A->rmap->bs,stepval;
659: PetscBool roworiented=a->roworiented;
660: const PetscScalar *value = v;
661: MatScalar *ap,*aa = a->a,*bap;
664: if (roworiented) stepval = (n-1)*bs;
665: else stepval = (m-1)*bs;
667: for (k=0; k<m; k++) { /* loop over added rows */
668: row = im[k];
669: if (row < 0) continue;
670: #if defined(PETSC_USE_DEBUG)
671: if (row >= a->mbs) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Row too large: row %D max %D",row,a->mbs-1);
672: #endif
673: rp = aj + ai[row];
674: ap = aa + bs2*ai[row];
675: rmax = imax[row];
676: nrow = ailen[row];
677: low = 0;
678: high = nrow;
679: for (l=0; l<n; l++) { /* loop over added columns */
680: if (in[l] < 0) continue;
681: col = in[l];
682: #if defined(PETSC_USE_DEBUG)
683: if (col >= a->nbs) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Column too large: col %D max %D",col,a->nbs-1);
684: #endif
685: if (col < row) {
686: if (a->ignore_ltriangular) continue; /* ignore lower triangular block */
687: 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)");
688: }
689: if (roworiented) value = v + k*(stepval+bs)*bs + l*bs;
690: else value = v + l*(stepval+bs)*bs + k*bs;
692: if (col <= lastcol) low = 0;
693: else high = nrow;
695: lastcol = col;
696: while (high-low > 7) {
697: t = (low+high)/2;
698: if (rp[t] > col) high = t;
699: else low = t;
700: }
701: for (i=low; i<high; i++) {
702: if (rp[i] > col) break;
703: if (rp[i] == col) {
704: bap = ap + bs2*i;
705: if (roworiented) {
706: if (is == ADD_VALUES) {
707: for (ii=0; ii<bs; ii++,value+=stepval) {
708: for (jj=ii; jj<bs2; jj+=bs) {
709: bap[jj] += *value++;
710: }
711: }
712: } else {
713: for (ii=0; ii<bs; ii++,value+=stepval) {
714: for (jj=ii; jj<bs2; jj+=bs) {
715: bap[jj] = *value++;
716: }
717: }
718: }
719: } else {
720: if (is == ADD_VALUES) {
721: for (ii=0; ii<bs; ii++,value+=stepval) {
722: for (jj=0; jj<bs; jj++) {
723: *bap++ += *value++;
724: }
725: }
726: } else {
727: for (ii=0; ii<bs; ii++,value+=stepval) {
728: for (jj=0; jj<bs; jj++) {
729: *bap++ = *value++;
730: }
731: }
732: }
733: }
734: goto noinsert2;
735: }
736: }
737: if (nonew == 1) goto noinsert2;
738: if (nonew == -1) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Inserting a new nonzero (%D, %D) in the matrix", row, col);
739: MatSeqXAIJReallocateAIJ(A,a->mbs,bs2,nrow,row,col,rmax,aa,ai,aj,rp,ap,imax,nonew,MatScalar);
740: N = nrow++ - 1; high++;
741: /* shift up all the later entries in this row */
742: for (ii=N; ii>=i; ii--) {
743: rp[ii+1] = rp[ii];
744: PetscMemcpy(ap+bs2*(ii+1),ap+bs2*(ii),bs2*sizeof(MatScalar));
745: }
746: if (N >= i) {
747: PetscMemzero(ap+bs2*i,bs2*sizeof(MatScalar));
748: }
749: rp[i] = col;
750: bap = ap + bs2*i;
751: if (roworiented) {
752: for (ii=0; ii<bs; ii++,value+=stepval) {
753: for (jj=ii; jj<bs2; jj+=bs) {
754: bap[jj] = *value++;
755: }
756: }
757: } else {
758: for (ii=0; ii<bs; ii++,value+=stepval) {
759: for (jj=0; jj<bs; jj++) {
760: *bap++ = *value++;
761: }
762: }
763: }
764: noinsert2:;
765: low = i;
766: }
767: ailen[row] = nrow;
768: }
769: return(0);
770: }
772: /*
773: This is not yet used
774: */
777: PetscErrorCode MatAssemblyEnd_SeqSBAIJ_SeqAIJ_Inode(Mat A)
778: {
779: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
781: const PetscInt *ai = a->i, *aj = a->j,*cols;
782: PetscInt i = 0,j,blk_size,m = A->rmap->n,node_count = 0,nzx,nzy,*ns,row,nz,cnt,cnt2,*counts;
783: PetscBool flag;
786: PetscMalloc1(m,&ns);
787: while (i < m) {
788: nzx = ai[i+1] - ai[i]; /* Number of nonzeros */
789: /* Limits the number of elements in a node to 'a->inode.limit' */
790: for (j=i+1,blk_size=1; j<m && blk_size <a->inode.limit; ++j,++blk_size) {
791: nzy = ai[j+1] - ai[j];
792: if (nzy != (nzx - j + i)) break;
793: PetscMemcmp(aj + ai[i] + j - i,aj + ai[j],nzy*sizeof(PetscInt),&flag);
794: if (!flag) break;
795: }
796: ns[node_count++] = blk_size;
798: i = j;
799: }
800: if (!a->inode.size && m && node_count > .9*m) {
801: PetscFree(ns);
802: PetscInfo2(A,"Found %D nodes out of %D rows. Not using Inode routines\n",node_count,m);
803: } else {
804: a->inode.node_count = node_count;
806: PetscMalloc1(node_count,&a->inode.size);
807: PetscLogObjectMemory((PetscObject)A,node_count*sizeof(PetscInt));
808: PetscMemcpy(a->inode.size,ns,node_count*sizeof(PetscInt));
809: PetscFree(ns);
810: PetscInfo3(A,"Found %D nodes of %D. Limit used: %D. Using Inode routines\n",node_count,m,a->inode.limit);
812: /* count collections of adjacent columns in each inode */
813: row = 0;
814: cnt = 0;
815: for (i=0; i<node_count; i++) {
816: cols = aj + ai[row] + a->inode.size[i];
817: nz = ai[row+1] - ai[row] - a->inode.size[i];
818: for (j=1; j<nz; j++) {
819: if (cols[j] != cols[j-1]+1) cnt++;
820: }
821: cnt++;
822: row += a->inode.size[i];
823: }
824: PetscMalloc1(2*cnt,&counts);
825: cnt = 0;
826: row = 0;
827: for (i=0; i<node_count; i++) {
828: cols = aj + ai[row] + a->inode.size[i];
829: counts[2*cnt] = cols[0];
830: nz = ai[row+1] - ai[row] - a->inode.size[i];
831: cnt2 = 1;
832: for (j=1; j<nz; j++) {
833: if (cols[j] != cols[j-1]+1) {
834: counts[2*(cnt++)+1] = cnt2;
835: counts[2*cnt] = cols[j];
836: cnt2 = 1;
837: } else cnt2++;
838: }
839: counts[2*(cnt++)+1] = cnt2;
840: row += a->inode.size[i];
841: }
842: PetscIntView(2*cnt,counts,0);
843: }
844: return(0);
845: }
849: PetscErrorCode MatAssemblyEnd_SeqSBAIJ(Mat A,MatAssemblyType mode)
850: {
851: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
853: PetscInt fshift = 0,i,j,*ai = a->i,*aj = a->j,*imax = a->imax;
854: PetscInt m = A->rmap->N,*ip,N,*ailen = a->ilen;
855: PetscInt mbs = a->mbs,bs2 = a->bs2,rmax = 0;
856: MatScalar *aa = a->a,*ap;
859: if (mode == MAT_FLUSH_ASSEMBLY) return(0);
861: if (m) rmax = ailen[0];
862: for (i=1; i<mbs; i++) {
863: /* move each row back by the amount of empty slots (fshift) before it*/
864: fshift += imax[i-1] - ailen[i-1];
865: rmax = PetscMax(rmax,ailen[i]);
866: if (fshift) {
867: ip = aj + ai[i]; ap = aa + bs2*ai[i];
868: N = ailen[i];
869: for (j=0; j<N; j++) {
870: ip[j-fshift] = ip[j];
871: PetscMemcpy(ap+(j-fshift)*bs2,ap+j*bs2,bs2*sizeof(MatScalar));
872: }
873: }
874: ai[i] = ai[i-1] + ailen[i-1];
875: }
876: if (mbs) {
877: fshift += imax[mbs-1] - ailen[mbs-1];
878: ai[mbs] = ai[mbs-1] + ailen[mbs-1];
879: }
880: /* reset ilen and imax for each row */
881: for (i=0; i<mbs; i++) {
882: ailen[i] = imax[i] = ai[i+1] - ai[i];
883: }
884: a->nz = ai[mbs];
886: /* diagonals may have moved, reset it */
887: if (a->diag) {
888: PetscMemcpy(a->diag,ai,mbs*sizeof(PetscInt));
889: }
890: 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);
892: 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);
893: PetscInfo1(A,"Number of mallocs during MatSetValues is %D\n",a->reallocs);
894: PetscInfo1(A,"Most nonzeros blocks in any row is %D\n",rmax);
896: A->info.mallocs += a->reallocs;
897: a->reallocs = 0;
898: A->info.nz_unneeded = (PetscReal)fshift*bs2;
899: a->idiagvalid = PETSC_FALSE;
901: if (A->cmap->n < 65536 && A->cmap->bs == 1) {
902: if (a->jshort && a->free_jshort) {
903: /* when matrix data structure is changed, previous jshort must be replaced */
904: PetscFree(a->jshort);
905: }
906: PetscMalloc1(a->i[A->rmap->n],&a->jshort);
907: PetscLogObjectMemory((PetscObject)A,a->i[A->rmap->n]*sizeof(unsigned short));
908: for (i=0; i<a->i[A->rmap->n]; i++) a->jshort[i] = a->j[i];
909: A->ops->mult = MatMult_SeqSBAIJ_1_ushort;
910: A->ops->sor = MatSOR_SeqSBAIJ_ushort;
911: a->free_jshort = PETSC_TRUE;
912: }
913: return(0);
914: }
916: /*
917: This function returns an array of flags which indicate the locations of contiguous
918: blocks that should be zeroed. for eg: if bs = 3 and is = [0,1,2,3,5,6,7,8,9]
919: then the resulting sizes = [3,1,1,3,1] correspondig to sets [(0,1,2),(3),(5),(6,7,8),(9)]
920: Assume: sizes should be long enough to hold all the values.
921: */
924: PetscErrorCode MatZeroRows_SeqSBAIJ_Check_Blocks(PetscInt idx[],PetscInt n,PetscInt bs,PetscInt sizes[], PetscInt *bs_max)
925: {
926: PetscInt i,j,k,row;
927: PetscBool flg;
930: for (i=0,j=0; i<n; j++) {
931: row = idx[i];
932: if (row%bs!=0) { /* Not the begining of a block */
933: sizes[j] = 1;
934: i++;
935: } else if (i+bs > n) { /* Beginning of a block, but complete block doesn't exist (at idx end) */
936: sizes[j] = 1; /* Also makes sure atleast 'bs' values exist for next else */
937: i++;
938: } else { /* Begining of the block, so check if the complete block exists */
939: flg = PETSC_TRUE;
940: for (k=1; k<bs; k++) {
941: if (row+k != idx[i+k]) { /* break in the block */
942: flg = PETSC_FALSE;
943: break;
944: }
945: }
946: if (flg) { /* No break in the bs */
947: sizes[j] = bs;
948: i += bs;
949: } else {
950: sizes[j] = 1;
951: i++;
952: }
953: }
954: }
955: *bs_max = j;
956: return(0);
957: }
960: /* Only add/insert a(i,j) with i<=j (blocks).
961: Any a(i,j) with i>j input by user is ingored.
962: */
966: PetscErrorCode MatSetValues_SeqSBAIJ(Mat A,PetscInt m,const PetscInt im[],PetscInt n,const PetscInt in[],const PetscScalar v[],InsertMode is)
967: {
968: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
970: PetscInt *rp,k,low,high,t,ii,row,nrow,i,col,l,rmax,N,lastcol = -1;
971: PetscInt *imax=a->imax,*ai=a->i,*ailen=a->ilen,roworiented=a->roworiented;
972: PetscInt *aj =a->j,nonew=a->nonew,bs=A->rmap->bs,brow,bcol;
973: PetscInt ridx,cidx,bs2=a->bs2;
974: MatScalar *ap,value,*aa=a->a,*bap;
977: for (k=0; k<m; k++) { /* loop over added rows */
978: row = im[k]; /* row number */
979: brow = row/bs; /* block row number */
980: if (row < 0) continue;
981: #if defined(PETSC_USE_DEBUG)
982: 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);
983: #endif
984: rp = aj + ai[brow]; /*ptr to beginning of column value of the row block*/
985: ap = aa + bs2*ai[brow]; /*ptr to beginning of element value of the row block*/
986: rmax = imax[brow]; /* maximum space allocated for this row */
987: nrow = ailen[brow]; /* actual length of this row */
988: low = 0;
990: for (l=0; l<n; l++) { /* loop over added columns */
991: if (in[l] < 0) continue;
992: #if defined(PETSC_USE_DEBUG)
993: 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);
994: #endif
995: col = in[l];
996: bcol = col/bs; /* block col number */
998: if (brow > bcol) {
999: if (a->ignore_ltriangular) continue; /* ignore lower triangular values */
1000: 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)");
1001: }
1003: ridx = row % bs; cidx = col % bs; /*row and col index inside the block */
1004: if ((brow==bcol && ridx<=cidx) || (brow<bcol)) {
1005: /* element value a(k,l) */
1006: if (roworiented) value = v[l + k*n];
1007: else value = v[k + l*m];
1009: /* move pointer bap to a(k,l) quickly and add/insert value */
1010: if (col <= lastcol) low = 0;
1011: high = nrow;
1012: lastcol = col;
1013: while (high-low > 7) {
1014: t = (low+high)/2;
1015: if (rp[t] > bcol) high = t;
1016: else low = t;
1017: }
1018: for (i=low; i<high; i++) {
1019: if (rp[i] > bcol) break;
1020: if (rp[i] == bcol) {
1021: bap = ap + bs2*i + bs*cidx + ridx;
1022: if (is == ADD_VALUES) *bap += value;
1023: else *bap = value;
1024: /* for diag block, add/insert its symmetric element a(cidx,ridx) */
1025: if (brow == bcol && ridx < cidx) {
1026: bap = ap + bs2*i + bs*ridx + cidx;
1027: if (is == ADD_VALUES) *bap += value;
1028: else *bap = value;
1029: }
1030: goto noinsert1;
1031: }
1032: }
1034: if (nonew == 1) goto noinsert1;
1035: if (nonew == -1) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Inserting a new nonzero (%D, %D) in the matrix", row, col);
1036: MatSeqXAIJReallocateAIJ(A,a->mbs,bs2,nrow,brow,bcol,rmax,aa,ai,aj,rp,ap,imax,nonew,MatScalar);
1038: N = nrow++ - 1; high++;
1039: /* shift up all the later entries in this row */
1040: for (ii=N; ii>=i; ii--) {
1041: rp[ii+1] = rp[ii];
1042: PetscMemcpy(ap+bs2*(ii+1),ap+bs2*(ii),bs2*sizeof(MatScalar));
1043: }
1044: if (N>=i) {
1045: PetscMemzero(ap+bs2*i,bs2*sizeof(MatScalar));
1046: }
1047: rp[i] = bcol;
1048: ap[bs2*i + bs*cidx + ridx] = value;
1049: A->nonzerostate++;
1050: noinsert1:;
1051: low = i;
1052: }
1053: } /* end of loop over added columns */
1054: ailen[brow] = nrow;
1055: } /* end of loop over added rows */
1056: return(0);
1057: }
1061: PetscErrorCode MatICCFactor_SeqSBAIJ(Mat inA,IS row,const MatFactorInfo *info)
1062: {
1063: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)inA->data;
1064: Mat outA;
1066: PetscBool row_identity;
1069: if (info->levels != 0) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Only levels=0 is supported for in-place icc");
1070: ISIdentity(row,&row_identity);
1071: if (!row_identity) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Matrix reordering is not supported");
1072: 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()! */
1074: outA = inA;
1075: inA->factortype = MAT_FACTOR_ICC;
1077: MatMarkDiagonal_SeqSBAIJ(inA);
1078: MatSeqSBAIJSetNumericFactorization_inplace(inA,row_identity);
1080: PetscObjectReference((PetscObject)row);
1081: ISDestroy(&a->row);
1082: a->row = row;
1083: PetscObjectReference((PetscObject)row);
1084: ISDestroy(&a->col);
1085: a->col = row;
1087: /* Create the invert permutation so that it can be used in MatCholeskyFactorNumeric() */
1088: if (a->icol) {ISInvertPermutation(row,PETSC_DECIDE, &a->icol);}
1089: PetscLogObjectParent((PetscObject)inA,(PetscObject)a->icol);
1091: if (!a->solve_work) {
1092: PetscMalloc1((inA->rmap->N+inA->rmap->bs),&a->solve_work);
1093: PetscLogObjectMemory((PetscObject)inA,(inA->rmap->N+inA->rmap->bs)*sizeof(PetscScalar));
1094: }
1096: MatCholeskyFactorNumeric(outA,inA,info);
1097: return(0);
1098: }
1102: PetscErrorCode MatSeqSBAIJSetColumnIndices_SeqSBAIJ(Mat mat,PetscInt *indices)
1103: {
1104: Mat_SeqSBAIJ *baij = (Mat_SeqSBAIJ*)mat->data;
1105: PetscInt i,nz,n;
1109: nz = baij->maxnz;
1110: n = mat->cmap->n;
1111: for (i=0; i<nz; i++) baij->j[i] = indices[i];
1113: baij->nz = nz;
1114: for (i=0; i<n; i++) baij->ilen[i] = baij->imax[i];
1116: MatSetOption(mat,MAT_NEW_NONZERO_LOCATION_ERR,PETSC_TRUE);
1117: return(0);
1118: }
1122: /*@
1123: MatSeqSBAIJSetColumnIndices - Set the column indices for all the rows
1124: in the matrix.
1126: Input Parameters:
1127: + mat - the SeqSBAIJ matrix
1128: - indices - the column indices
1130: Level: advanced
1132: Notes:
1133: This can be called if you have precomputed the nonzero structure of the
1134: matrix and want to provide it to the matrix object to improve the performance
1135: of the MatSetValues() operation.
1137: You MUST have set the correct numbers of nonzeros per row in the call to
1138: MatCreateSeqSBAIJ(), and the columns indices MUST be sorted.
1140: MUST be called before any calls to MatSetValues()
1142: .seealso: MatCreateSeqSBAIJ
1143: @*/
1144: PetscErrorCode MatSeqSBAIJSetColumnIndices(Mat mat,PetscInt *indices)
1145: {
1151: PetscUseMethod(mat,"MatSeqSBAIJSetColumnIndices_C",(Mat,PetscInt*),(mat,indices));
1152: return(0);
1153: }
1157: PetscErrorCode MatCopy_SeqSBAIJ(Mat A,Mat B,MatStructure str)
1158: {
1162: /* If the two matrices have the same copy implementation, use fast copy. */
1163: if (str == SAME_NONZERO_PATTERN && (A->ops->copy == B->ops->copy)) {
1164: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
1165: Mat_SeqSBAIJ *b = (Mat_SeqSBAIJ*)B->data;
1167: 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");
1168: PetscMemcpy(b->a,a->a,(a->i[A->rmap->N])*sizeof(PetscScalar));
1169: } else {
1170: MatGetRowUpperTriangular(A);
1171: MatCopy_Basic(A,B,str);
1172: MatRestoreRowUpperTriangular(A);
1173: }
1174: return(0);
1175: }
1179: PetscErrorCode MatSetUp_SeqSBAIJ(Mat A)
1180: {
1184: MatSeqSBAIJSetPreallocation_SeqSBAIJ(A,A->rmap->bs,PETSC_DEFAULT,0);
1185: return(0);
1186: }
1190: PetscErrorCode MatSeqSBAIJGetArray_SeqSBAIJ(Mat A,PetscScalar *array[])
1191: {
1192: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
1195: *array = a->a;
1196: return(0);
1197: }
1201: PetscErrorCode MatSeqSBAIJRestoreArray_SeqSBAIJ(Mat A,PetscScalar *array[])
1202: {
1204: return(0);
1205: }
1209: PetscErrorCode MatAXPY_SeqSBAIJ(Mat Y,PetscScalar a,Mat X,MatStructure str)
1210: {
1211: Mat_SeqSBAIJ *x=(Mat_SeqSBAIJ*)X->data, *y=(Mat_SeqSBAIJ*)Y->data;
1213: PetscInt i,bs=Y->rmap->bs,bs2=bs*bs,j;
1214: PetscBLASInt one = 1;
1217: if (str == SAME_NONZERO_PATTERN) {
1218: PetscScalar alpha = a;
1219: PetscBLASInt bnz;
1220: PetscBLASIntCast(x->nz*bs2,&bnz);
1221: PetscStackCallBLAS("BLASaxpy",BLASaxpy_(&bnz,&alpha,x->a,&one,y->a,&one));
1222: PetscObjectStateIncrease((PetscObject)Y);
1223: } else if (str == SUBSET_NONZERO_PATTERN) { /* nonzeros of X is a subset of Y's */
1224: if (y->xtoy && y->XtoY != X) {
1225: PetscFree(y->xtoy);
1226: MatDestroy(&y->XtoY);
1227: }
1228: if (!y->xtoy) { /* get xtoy */
1229: MatAXPYGetxtoy_Private(x->mbs,x->i,x->j,NULL, y->i,y->j,NULL, &y->xtoy);
1230: y->XtoY = X;
1231: }
1232: for (i=0; i<x->nz; i++) {
1233: j = 0;
1234: while (j < bs2) {
1235: y->a[bs2*y->xtoy[i]+j] += a*(x->a[bs2*i+j]);
1236: j++;
1237: }
1238: }
1239: PetscObjectStateIncrease((PetscObject)Y);
1240: PetscInfo3(Y,"ratio of nnz_s(X)/nnz_s(Y): %D/%D = %g\n",bs2*x->nz,bs2*y->nz,(double)((PetscReal)(bs2*x->nz)/(PetscReal)(bs2*y->nz)));
1241: } else {
1242: MatGetRowUpperTriangular(X);
1243: MatAXPY_Basic(Y,a,X,str);
1244: MatRestoreRowUpperTriangular(X);
1245: }
1246: return(0);
1247: }
1251: PetscErrorCode MatIsSymmetric_SeqSBAIJ(Mat A,PetscReal tol,PetscBool *flg)
1252: {
1254: *flg = PETSC_TRUE;
1255: return(0);
1256: }
1260: PetscErrorCode MatIsStructurallySymmetric_SeqSBAIJ(Mat A,PetscBool *flg)
1261: {
1263: *flg = PETSC_TRUE;
1264: return(0);
1265: }
1269: PetscErrorCode MatIsHermitian_SeqSBAIJ(Mat A,PetscReal tol,PetscBool *flg)
1270: {
1272: *flg = PETSC_FALSE;
1273: return(0);
1274: }
1278: PetscErrorCode MatRealPart_SeqSBAIJ(Mat A)
1279: {
1280: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
1281: PetscInt i,nz = a->bs2*a->i[a->mbs];
1282: MatScalar *aa = a->a;
1285: for (i=0; i<nz; i++) aa[i] = PetscRealPart(aa[i]);
1286: return(0);
1287: }
1291: PetscErrorCode MatImaginaryPart_SeqSBAIJ(Mat A)
1292: {
1293: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ*)A->data;
1294: PetscInt i,nz = a->bs2*a->i[a->mbs];
1295: MatScalar *aa = a->a;
1298: for (i=0; i<nz; i++) aa[i] = PetscImaginaryPart(aa[i]);
1299: return(0);
1300: }
1304: PetscErrorCode MatZeroRowsColumns_SeqSBAIJ(Mat A,PetscInt is_n,const PetscInt is_idx[],PetscScalar diag,Vec x, Vec b)
1305: {
1306: Mat_SeqSBAIJ *baij=(Mat_SeqSBAIJ*)A->data;
1307: PetscErrorCode ierr;
1308: PetscInt i,j,k,count;
1309: PetscInt bs =A->rmap->bs,bs2=baij->bs2,row,col;
1310: PetscScalar zero = 0.0;
1311: MatScalar *aa;
1312: const PetscScalar *xx;
1313: PetscScalar *bb;
1314: PetscBool *zeroed,vecs = PETSC_FALSE;
1317: /* fix right hand side if needed */
1318: if (x && b) {
1319: VecGetArrayRead(x,&xx);
1320: VecGetArray(b,&bb);
1321: vecs = PETSC_TRUE;
1322: }
1324: /* zero the columns */
1325: PetscCalloc1(A->rmap->n,&zeroed);
1326: for (i=0; i<is_n; i++) {
1327: 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]);
1328: zeroed[is_idx[i]] = PETSC_TRUE;
1329: }
1330: if (vecs) {
1331: for (i=0; i<A->rmap->N; i++) {
1332: row = i/bs;
1333: for (j=baij->i[row]; j<baij->i[row+1]; j++) {
1334: for (k=0; k<bs; k++) {
1335: col = bs*baij->j[j] + k;
1336: if (col <= i) continue;
1337: aa = ((MatScalar*)(baij->a)) + j*bs2 + (i%bs) + bs*k;
1338: if (!zeroed[i] && zeroed[col]) bb[i] -= aa[0]*xx[col];
1339: if (zeroed[i] && !zeroed[col]) bb[col] -= aa[0]*xx[i];
1340: }
1341: }
1342: }
1343: for (i=0; i<is_n; i++) bb[is_idx[i]] = diag*xx[is_idx[i]];
1344: }
1346: for (i=0; i<A->rmap->N; i++) {
1347: if (!zeroed[i]) {
1348: row = i/bs;
1349: for (j=baij->i[row]; j<baij->i[row+1]; j++) {
1350: for (k=0; k<bs; k++) {
1351: col = bs*baij->j[j] + k;
1352: if (zeroed[col]) {
1353: aa = ((MatScalar*)(baij->a)) + j*bs2 + (i%bs) + bs*k;
1354: aa[0] = 0.0;
1355: }
1356: }
1357: }
1358: }
1359: }
1360: PetscFree(zeroed);
1361: if (vecs) {
1362: VecRestoreArrayRead(x,&xx);
1363: VecRestoreArray(b,&bb);
1364: }
1366: /* zero the rows */
1367: for (i=0; i<is_n; i++) {
1368: row = is_idx[i];
1369: count = (baij->i[row/bs +1] - baij->i[row/bs])*bs;
1370: aa = ((MatScalar*)(baij->a)) + baij->i[row/bs]*bs2 + (row%bs);
1371: for (k=0; k<count; k++) {
1372: aa[0] = zero;
1373: aa += bs;
1374: }
1375: if (diag != 0.0) {
1376: (*A->ops->setvalues)(A,1,&row,1,&row,&diag,INSERT_VALUES);
1377: }
1378: }
1379: MatAssemblyEnd_SeqSBAIJ(A,MAT_FINAL_ASSEMBLY);
1380: return(0);
1381: }
1383: /* -------------------------------------------------------------------*/
1384: static struct _MatOps MatOps_Values = {MatSetValues_SeqSBAIJ,
1385: MatGetRow_SeqSBAIJ,
1386: MatRestoreRow_SeqSBAIJ,
1387: MatMult_SeqSBAIJ_N,
1388: /* 4*/ MatMultAdd_SeqSBAIJ_N,
1389: MatMult_SeqSBAIJ_N, /* transpose versions are same as non-transpose versions */
1390: MatMultAdd_SeqSBAIJ_N,
1391: 0,
1392: 0,
1393: 0,
1394: /* 10*/ 0,
1395: 0,
1396: MatCholeskyFactor_SeqSBAIJ,
1397: MatSOR_SeqSBAIJ,
1398: MatTranspose_SeqSBAIJ,
1399: /* 15*/ MatGetInfo_SeqSBAIJ,
1400: MatEqual_SeqSBAIJ,
1401: MatGetDiagonal_SeqSBAIJ,
1402: MatDiagonalScale_SeqSBAIJ,
1403: MatNorm_SeqSBAIJ,
1404: /* 20*/ 0,
1405: MatAssemblyEnd_SeqSBAIJ,
1406: MatSetOption_SeqSBAIJ,
1407: MatZeroEntries_SeqSBAIJ,
1408: /* 24*/ 0,
1409: 0,
1410: 0,
1411: 0,
1412: 0,
1413: /* 29*/ MatSetUp_SeqSBAIJ,
1414: 0,
1415: 0,
1416: 0,
1417: 0,
1418: /* 34*/ MatDuplicate_SeqSBAIJ,
1419: 0,
1420: 0,
1421: 0,
1422: MatICCFactor_SeqSBAIJ,
1423: /* 39*/ MatAXPY_SeqSBAIJ,
1424: MatGetSubMatrices_SeqSBAIJ,
1425: MatIncreaseOverlap_SeqSBAIJ,
1426: MatGetValues_SeqSBAIJ,
1427: MatCopy_SeqSBAIJ,
1428: /* 44*/ 0,
1429: MatScale_SeqSBAIJ,
1430: 0,
1431: 0,
1432: MatZeroRowsColumns_SeqSBAIJ,
1433: /* 49*/ 0,
1434: MatGetRowIJ_SeqSBAIJ,
1435: MatRestoreRowIJ_SeqSBAIJ,
1436: 0,
1437: 0,
1438: /* 54*/ 0,
1439: 0,
1440: 0,
1441: 0,
1442: MatSetValuesBlocked_SeqSBAIJ,
1443: /* 59*/ MatGetSubMatrix_SeqSBAIJ,
1444: 0,
1445: 0,
1446: 0,
1447: 0,
1448: /* 64*/ 0,
1449: 0,
1450: 0,
1451: 0,
1452: 0,
1453: /* 69*/ MatGetRowMaxAbs_SeqSBAIJ,
1454: 0,
1455: 0,
1456: 0,
1457: 0,
1458: /* 74*/ 0,
1459: 0,
1460: 0,
1461: 0,
1462: 0,
1463: /* 79*/ 0,
1464: 0,
1465: 0,
1466: MatGetInertia_SeqSBAIJ,
1467: MatLoad_SeqSBAIJ,
1468: /* 84*/ MatIsSymmetric_SeqSBAIJ,
1469: MatIsHermitian_SeqSBAIJ,
1470: MatIsStructurallySymmetric_SeqSBAIJ,
1471: 0,
1472: 0,
1473: /* 89*/ 0,
1474: 0,
1475: 0,
1476: 0,
1477: 0,
1478: /* 94*/ 0,
1479: 0,
1480: 0,
1481: 0,
1482: 0,
1483: /* 99*/ 0,
1484: 0,
1485: 0,
1486: 0,
1487: 0,
1488: /*104*/ 0,
1489: MatRealPart_SeqSBAIJ,
1490: MatImaginaryPart_SeqSBAIJ,
1491: MatGetRowUpperTriangular_SeqSBAIJ,
1492: MatRestoreRowUpperTriangular_SeqSBAIJ,
1493: /*109*/ 0,
1494: 0,
1495: 0,
1496: 0,
1497: MatMissingDiagonal_SeqSBAIJ,
1498: /*114*/ 0,
1499: 0,
1500: 0,
1501: 0,
1502: 0,
1503: /*119*/ 0,
1504: 0,
1505: 0,
1506: 0,
1507: 0,
1508: /*124*/ 0,
1509: 0,
1510: 0,
1511: 0,
1512: 0,
1513: /*129*/ 0,
1514: 0,
1515: 0,
1516: 0,
1517: 0,
1518: /*134*/ 0,
1519: 0,
1520: 0,
1521: 0,
1522: 0,
1523: /*139*/ 0,
1524: 0,
1525: 0
1526: };
1530: PetscErrorCode MatStoreValues_SeqSBAIJ(Mat mat)
1531: {
1532: Mat_SeqSBAIJ *aij = (Mat_SeqSBAIJ*)mat->data;
1533: PetscInt nz = aij->i[mat->rmap->N]*mat->rmap->bs*aij->bs2;
1537: if (aij->nonew != 1) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Must call MatSetOption(A,MAT_NEW_NONZERO_LOCATIONS,PETSC_FALSE);first");
1539: /* allocate space for values if not already there */
1540: if (!aij->saved_values) {
1541: PetscMalloc1((nz+1),&aij->saved_values);
1542: }
1544: /* copy values over */
1545: PetscMemcpy(aij->saved_values,aij->a,nz*sizeof(PetscScalar));
1546: return(0);
1547: }
1551: PetscErrorCode MatRetrieveValues_SeqSBAIJ(Mat mat)
1552: {
1553: Mat_SeqSBAIJ *aij = (Mat_SeqSBAIJ*)mat->data;
1555: PetscInt nz = aij->i[mat->rmap->N]*mat->rmap->bs*aij->bs2;
1558: if (aij->nonew != 1) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Must call MatSetOption(A,MAT_NEW_NONZERO_LOCATIONS,PETSC_FALSE);first");
1559: if (!aij->saved_values) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ORDER,"Must call MatStoreValues(A);first");
1561: /* copy values over */
1562: PetscMemcpy(aij->a,aij->saved_values,nz*sizeof(PetscScalar));
1563: return(0);
1564: }
1568: PetscErrorCode MatSeqSBAIJSetPreallocation_SeqSBAIJ(Mat B,PetscInt bs,PetscInt nz,PetscInt *nnz)
1569: {
1570: Mat_SeqSBAIJ *b = (Mat_SeqSBAIJ*)B->data;
1572: PetscInt i,mbs,bs2;
1573: PetscBool skipallocation = PETSC_FALSE,flg = PETSC_FALSE,realalloc = PETSC_FALSE;
1576: if (nz >= 0 || nnz) realalloc = PETSC_TRUE;
1577: B->preallocated = PETSC_TRUE;
1579: MatSetBlockSize(B,PetscAbs(bs));
1580: PetscLayoutSetUp(B->rmap);
1581: PetscLayoutSetUp(B->cmap);
1582: PetscLayoutGetBlockSize(B->rmap,&bs);
1584: mbs = B->rmap->N/bs;
1585: bs2 = bs*bs;
1587: if (mbs*bs != B->rmap->N) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_SIZ,"Number rows, cols must be divisible by blocksize");
1589: if (nz == MAT_SKIP_ALLOCATION) {
1590: skipallocation = PETSC_TRUE;
1591: nz = 0;
1592: }
1594: if (nz == PETSC_DEFAULT || nz == PETSC_DECIDE) nz = 3;
1595: if (nz < 0) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"nz cannot be less than 0: value %D",nz);
1596: if (nnz) {
1597: for (i=0; i<mbs; i++) {
1598: 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]);
1599: if (nnz[i] > mbs) SETERRQ3(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"nnz cannot be greater than block row length: local row %D value %D rowlength %D",i,nnz[i],mbs);
1600: }
1601: }
1603: B->ops->mult = MatMult_SeqSBAIJ_N;
1604: B->ops->multadd = MatMultAdd_SeqSBAIJ_N;
1605: B->ops->multtranspose = MatMult_SeqSBAIJ_N;
1606: B->ops->multtransposeadd = MatMultAdd_SeqSBAIJ_N;
1608: PetscOptionsGetBool(((PetscObject)B)->prefix,"-mat_no_unroll",&flg,NULL);
1609: if (!flg) {
1610: switch (bs) {
1611: case 1:
1612: B->ops->mult = MatMult_SeqSBAIJ_1;
1613: B->ops->multadd = MatMultAdd_SeqSBAIJ_1;
1614: B->ops->multtranspose = MatMult_SeqSBAIJ_1;
1615: B->ops->multtransposeadd = MatMultAdd_SeqSBAIJ_1;
1616: break;
1617: case 2:
1618: B->ops->mult = MatMult_SeqSBAIJ_2;
1619: B->ops->multadd = MatMultAdd_SeqSBAIJ_2;
1620: B->ops->multtranspose = MatMult_SeqSBAIJ_2;
1621: B->ops->multtransposeadd = MatMultAdd_SeqSBAIJ_2;
1622: break;
1623: case 3:
1624: B->ops->mult = MatMult_SeqSBAIJ_3;
1625: B->ops->multadd = MatMultAdd_SeqSBAIJ_3;
1626: B->ops->multtranspose = MatMult_SeqSBAIJ_3;
1627: B->ops->multtransposeadd = MatMultAdd_SeqSBAIJ_3;
1628: break;
1629: case 4:
1630: B->ops->mult = MatMult_SeqSBAIJ_4;
1631: B->ops->multadd = MatMultAdd_SeqSBAIJ_4;
1632: B->ops->multtranspose = MatMult_SeqSBAIJ_4;
1633: B->ops->multtransposeadd = MatMultAdd_SeqSBAIJ_4;
1634: break;
1635: case 5:
1636: B->ops->mult = MatMult_SeqSBAIJ_5;
1637: B->ops->multadd = MatMultAdd_SeqSBAIJ_5;
1638: B->ops->multtranspose = MatMult_SeqSBAIJ_5;
1639: B->ops->multtransposeadd = MatMultAdd_SeqSBAIJ_5;
1640: break;
1641: case 6:
1642: B->ops->mult = MatMult_SeqSBAIJ_6;
1643: B->ops->multadd = MatMultAdd_SeqSBAIJ_6;
1644: B->ops->multtranspose = MatMult_SeqSBAIJ_6;
1645: B->ops->multtransposeadd = MatMultAdd_SeqSBAIJ_6;
1646: break;
1647: case 7:
1648: B->ops->mult = MatMult_SeqSBAIJ_7;
1649: B->ops->multadd = MatMultAdd_SeqSBAIJ_7;
1650: B->ops->multtranspose = MatMult_SeqSBAIJ_7;
1651: B->ops->multtransposeadd = MatMultAdd_SeqSBAIJ_7;
1652: break;
1653: }
1654: }
1656: b->mbs = mbs;
1657: b->nbs = mbs;
1658: if (!skipallocation) {
1659: if (!b->imax) {
1660: PetscMalloc2(mbs,&b->imax,mbs,&b->ilen);
1662: b->free_imax_ilen = PETSC_TRUE;
1664: PetscLogObjectMemory((PetscObject)B,2*mbs*sizeof(PetscInt));
1665: }
1666: if (!nnz) {
1667: if (nz == PETSC_DEFAULT || nz == PETSC_DECIDE) nz = 5;
1668: else if (nz <= 0) nz = 1;
1669: for (i=0; i<mbs; i++) b->imax[i] = nz;
1670: nz = nz*mbs; /* total nz */
1671: } else {
1672: nz = 0;
1673: for (i=0; i<mbs; i++) {b->imax[i] = nnz[i]; nz += nnz[i];}
1674: }
1675: /* b->ilen will count nonzeros in each block row so far. */
1676: for (i=0; i<mbs; i++) b->ilen[i] = 0;
1677: /* nz=(nz+mbs)/2; */ /* total diagonal and superdiagonal nonzero blocks */
1679: /* allocate the matrix space */
1680: MatSeqXAIJFreeAIJ(B,&b->a,&b->j,&b->i);
1681: PetscMalloc3(bs2*nz,&b->a,nz,&b->j,B->rmap->N+1,&b->i);
1682: PetscLogObjectMemory((PetscObject)B,(B->rmap->N+1)*sizeof(PetscInt)+nz*(bs2*sizeof(PetscScalar)+sizeof(PetscInt)));
1683: PetscMemzero(b->a,nz*bs2*sizeof(MatScalar));
1684: PetscMemzero(b->j,nz*sizeof(PetscInt));
1686: b->singlemalloc = PETSC_TRUE;
1688: /* pointer to beginning of each row */
1689: b->i[0] = 0;
1690: for (i=1; i<mbs+1; i++) b->i[i] = b->i[i-1] + b->imax[i-1];
1692: b->free_a = PETSC_TRUE;
1693: b->free_ij = PETSC_TRUE;
1694: } else {
1695: b->free_a = PETSC_FALSE;
1696: b->free_ij = PETSC_FALSE;
1697: }
1699: B->rmap->bs = bs;
1700: b->bs2 = bs2;
1701: b->nz = 0;
1702: b->maxnz = nz;
1704: b->inew = 0;
1705: b->jnew = 0;
1706: b->anew = 0;
1707: b->a2anew = 0;
1708: b->permute = PETSC_FALSE;
1709: if (realalloc) {MatSetOption(B,MAT_NEW_NONZERO_ALLOCATION_ERR,PETSC_TRUE);}
1710: return(0);
1711: }
1715: PetscErrorCode MatSeqSBAIJSetPreallocationCSR_SeqSBAIJ(Mat B,PetscInt bs,const PetscInt ii[],const PetscInt jj[], const PetscScalar V[])
1716: {
1717: PetscInt i,j,m,nz,nz_max=0,*nnz;
1718: PetscScalar *values=0;
1719: PetscBool roworiented = ((Mat_SeqSBAIJ*)B->data)->roworiented;
1722: if (bs < 1) SETERRQ1(PetscObjectComm((PetscObject)B),PETSC_ERR_ARG_OUTOFRANGE,"Invalid block size specified, must be positive but it is %D",bs);
1723: PetscLayoutSetBlockSize(B->rmap,bs);
1724: PetscLayoutSetBlockSize(B->cmap,bs);
1725: PetscLayoutSetUp(B->rmap);
1726: PetscLayoutSetUp(B->cmap);
1727: PetscLayoutGetBlockSize(B->rmap,&bs);
1728: m = B->rmap->n/bs;
1730: if (ii[0]) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"ii[0] must be 0 but it is %D",ii[0]);
1731: PetscMalloc1((m+1),&nnz);
1732: for (i=0; i<m; i++) {
1733: nz = ii[i+1] - ii[i];
1734: if (nz < 0) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Row %D has a negative number of columns %D",i,nz);
1735: nz_max = PetscMax(nz_max,nz);
1736: nnz[i] = nz;
1737: }
1738: MatSeqSBAIJSetPreallocation(B,bs,0,nnz);
1739: PetscFree(nnz);
1741: values = (PetscScalar*)V;
1742: if (!values) {
1743: PetscCalloc1(bs*bs*nz_max,&values);
1744: }
1745: for (i=0; i<m; i++) {
1746: PetscInt ncols = ii[i+1] - ii[i];
1747: const PetscInt *icols = jj + ii[i];
1748: if (!roworiented || bs == 1) {
1749: const PetscScalar *svals = values + (V ? (bs*bs*ii[i]) : 0);
1750: MatSetValuesBlocked_SeqSBAIJ(B,1,&i,ncols,icols,svals,INSERT_VALUES);
1751: } else {
1752: for (j=0; j<ncols; j++) {
1753: const PetscScalar *svals = values + (V ? (bs*bs*(ii[i]+j)) : 0);
1754: MatSetValuesBlocked_SeqSBAIJ(B,1,&i,1,&icols[j],svals,INSERT_VALUES);
1755: }
1756: }
1757: }
1758: if (!V) { PetscFree(values); }
1759: MatAssemblyBegin(B,MAT_FINAL_ASSEMBLY);
1760: MatAssemblyEnd(B,MAT_FINAL_ASSEMBLY);
1761: MatSetOption(B,MAT_NEW_NONZERO_LOCATION_ERR,PETSC_TRUE);
1762: return(0);
1763: }
1765: /*
1766: This is used to set the numeric factorization for both Cholesky and ICC symbolic factorization
1767: */
1770: PetscErrorCode MatSeqSBAIJSetNumericFactorization_inplace(Mat B,PetscBool natural)
1771: {
1773: PetscBool flg = PETSC_FALSE;
1774: PetscInt bs = B->rmap->bs;
1777: PetscOptionsGetBool(((PetscObject)B)->prefix,"-mat_no_unroll",&flg,NULL);
1778: if (flg) bs = 8;
1780: if (!natural) {
1781: switch (bs) {
1782: case 1:
1783: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_1_inplace;
1784: break;
1785: case 2:
1786: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_2;
1787: break;
1788: case 3:
1789: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_3;
1790: break;
1791: case 4:
1792: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_4;
1793: break;
1794: case 5:
1795: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_5;
1796: break;
1797: case 6:
1798: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_6;
1799: break;
1800: case 7:
1801: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_7;
1802: break;
1803: default:
1804: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_N;
1805: break;
1806: }
1807: } else {
1808: switch (bs) {
1809: case 1:
1810: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_1_NaturalOrdering_inplace;
1811: break;
1812: case 2:
1813: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_2_NaturalOrdering;
1814: break;
1815: case 3:
1816: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_3_NaturalOrdering;
1817: break;
1818: case 4:
1819: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_4_NaturalOrdering;
1820: break;
1821: case 5:
1822: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_5_NaturalOrdering;
1823: break;
1824: case 6:
1825: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_6_NaturalOrdering;
1826: break;
1827: case 7:
1828: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_7_NaturalOrdering;
1829: break;
1830: default:
1831: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_N_NaturalOrdering;
1832: break;
1833: }
1834: }
1835: return(0);
1836: }
1838: PETSC_EXTERN PetscErrorCode MatConvert_SeqSBAIJ_SeqAIJ(Mat, MatType,MatReuse,Mat*);
1839: PETSC_EXTERN PetscErrorCode MatConvert_SeqSBAIJ_SeqBAIJ(Mat, MatType,MatReuse,Mat*);
1843: PETSC_EXTERN PetscErrorCode MatGetFactor_seqsbaij_petsc(Mat A,MatFactorType ftype,Mat *B)
1844: {
1845: PetscInt n = A->rmap->n;
1849: #if defined(PETSC_USE_COMPLEX)
1850: if (A->hermitian) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Hermitian Factor is not supported");
1851: #endif
1852: MatCreate(PetscObjectComm((PetscObject)A),B);
1853: MatSetSizes(*B,n,n,n,n);
1854: if (ftype == MAT_FACTOR_CHOLESKY || ftype == MAT_FACTOR_ICC) {
1855: MatSetType(*B,MATSEQSBAIJ);
1856: MatSeqSBAIJSetPreallocation(*B,A->rmap->bs,MAT_SKIP_ALLOCATION,NULL);
1858: (*B)->ops->choleskyfactorsymbolic = MatCholeskyFactorSymbolic_SeqSBAIJ;
1859: (*B)->ops->iccfactorsymbolic = MatICCFactorSymbolic_SeqSBAIJ;
1860: } else SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Factor type not supported");
1861: (*B)->factortype = ftype;
1862: return(0);
1863: }
1867: PetscErrorCode MatGetFactorAvailable_seqsbaij_petsc(Mat A,MatFactorType ftype,PetscBool *flg)
1868: {
1870: if (ftype == MAT_FACTOR_CHOLESKY || ftype == MAT_FACTOR_ICC) {
1871: *flg = PETSC_TRUE;
1872: } else {
1873: *flg = PETSC_FALSE;
1874: }
1875: return(0);
1876: }
1878: #if defined(PETSC_HAVE_MUMPS)
1879: PETSC_EXTERN PetscErrorCode MatGetFactor_sbaij_mumps(Mat,MatFactorType,Mat*);
1880: #endif
1881: #if defined(PETSC_HAVE_PASTIX)
1882: PETSC_EXTERN PetscErrorCode MatGetFactor_seqsbaij_pastix(Mat,MatFactorType,Mat*);
1883: #endif
1884: #if defined(PETSC_HAVE_SUITESPARSE)
1885: PETSC_EXTERN PetscErrorCode MatGetFactor_seqsbaij_cholmod(Mat,MatFactorType,Mat*);
1886: #endif
1887: PETSC_EXTERN PetscErrorCode MatGetFactor_seqsbaij_sbstrm(Mat,MatFactorType,Mat*);
1889: /*MC
1890: MATSEQSBAIJ - MATSEQSBAIJ = "seqsbaij" - A matrix type to be used for sequential symmetric block sparse matrices,
1891: based on block compressed sparse row format. Only the upper triangular portion of the matrix is stored.
1893: For complex numbers by default this matrix is symmetric, NOT Hermitian symmetric. To make it Hermitian symmetric you
1894: can call MatSetOption(Mat, MAT_HERMITIAN); after MatAssemblyEnd()
1896: Options Database Keys:
1897: . -mat_type seqsbaij - sets the matrix type to "seqsbaij" during a call to MatSetFromOptions()
1899: Notes: By default if you insert values into the lower triangular part of the matrix they are simply ignored (since they are not
1900: stored and it is assumed they symmetric to the upper triangular). If you call MatSetOption(Mat,MAT_IGNORE_LOWER_TRIANGULAR,PETSC_FALSE) or use
1901: 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.
1904: Level: beginner
1906: .seealso: MatCreateSeqSBAIJ
1907: M*/
1909: PETSC_EXTERN PetscErrorCode MatConvert_SeqSBAIJ_SeqSBSTRM(Mat, MatType,MatReuse,Mat*);
1913: PETSC_EXTERN PetscErrorCode MatCreate_SeqSBAIJ(Mat B)
1914: {
1915: Mat_SeqSBAIJ *b;
1917: PetscMPIInt size;
1918: PetscBool no_unroll = PETSC_FALSE,no_inode = PETSC_FALSE;
1921: MPI_Comm_size(PetscObjectComm((PetscObject)B),&size);
1922: if (size > 1) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"Comm must be of size 1");
1924: PetscNewLog(B,&b);
1925: B->data = (void*)b;
1926: PetscMemcpy(B->ops,&MatOps_Values,sizeof(struct _MatOps));
1928: B->ops->destroy = MatDestroy_SeqSBAIJ;
1929: B->ops->view = MatView_SeqSBAIJ;
1930: b->row = 0;
1931: b->icol = 0;
1932: b->reallocs = 0;
1933: b->saved_values = 0;
1934: b->inode.limit = 5;
1935: b->inode.max_limit = 5;
1937: b->roworiented = PETSC_TRUE;
1938: b->nonew = 0;
1939: b->diag = 0;
1940: b->solve_work = 0;
1941: b->mult_work = 0;
1942: B->spptr = 0;
1943: B->info.nz_unneeded = (PetscReal)b->maxnz*b->bs2;
1944: b->keepnonzeropattern = PETSC_FALSE;
1945: b->xtoy = 0;
1946: b->XtoY = 0;
1948: b->inew = 0;
1949: b->jnew = 0;
1950: b->anew = 0;
1951: b->a2anew = 0;
1952: b->permute = PETSC_FALSE;
1954: b->ignore_ltriangular = PETSC_TRUE;
1956: PetscOptionsGetBool(((PetscObject)B)->prefix,"-mat_ignore_lower_triangular",&b->ignore_ltriangular,NULL);
1958: b->getrow_utriangular = PETSC_FALSE;
1960: PetscOptionsGetBool(((PetscObject)B)->prefix,"-mat_getrow_uppertriangular",&b->getrow_utriangular,NULL);
1962: #if defined(PETSC_HAVE_PASTIX)
1963: PetscObjectComposeFunction((PetscObject)B,"MatGetFactor_pastix_C",MatGetFactor_seqsbaij_pastix);
1964: #endif
1965: #if defined(PETSC_HAVE_MUMPS)
1966: PetscObjectComposeFunction((PetscObject)B,"MatGetFactor_mumps_C",MatGetFactor_sbaij_mumps);
1967: #endif
1968: #if defined(PETSC_HAVE_SUITESPARSE)
1969: PetscObjectComposeFunction((PetscObject)B,"MatGetFactor_cholmod_C",MatGetFactor_seqsbaij_cholmod);
1970: #endif
1971: PetscObjectComposeFunction((PetscObject)B,"MatGetFactorAvailable_petsc_C",MatGetFactorAvailable_seqsbaij_petsc);
1972: PetscObjectComposeFunction((PetscObject)B,"MatGetFactor_petsc_C",MatGetFactor_seqsbaij_petsc);
1973: PetscObjectComposeFunction((PetscObject)B,"MatGetFactor_sbstrm_C",MatGetFactor_seqsbaij_sbstrm);
1974: PetscObjectComposeFunction((PetscObject)B,"MatStoreValues_C",MatStoreValues_SeqSBAIJ);
1975: PetscObjectComposeFunction((PetscObject)B,"MatRetrieveValues_C",MatRetrieveValues_SeqSBAIJ);
1976: PetscObjectComposeFunction((PetscObject)B,"MatSeqSBAIJSetColumnIndices_C",MatSeqSBAIJSetColumnIndices_SeqSBAIJ);
1977: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqsbaij_seqaij_C",MatConvert_SeqSBAIJ_SeqAIJ);
1978: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqsbaij_seqbaij_C",MatConvert_SeqSBAIJ_SeqBAIJ);
1979: PetscObjectComposeFunction((PetscObject)B,"MatSeqSBAIJSetPreallocation_C",MatSeqSBAIJSetPreallocation_SeqSBAIJ);
1980: PetscObjectComposeFunction((PetscObject)B,"MatSeqSBAIJSetPreallocationCSR_C",MatSeqSBAIJSetPreallocationCSR_SeqSBAIJ);
1981: PetscObjectComposeFunction((PetscObject)B,"MatConvert_seqsbaij_seqsbstrm_C",MatConvert_SeqSBAIJ_SeqSBSTRM);
1983: B->symmetric = PETSC_TRUE;
1984: B->structurally_symmetric = PETSC_TRUE;
1985: B->symmetric_set = PETSC_TRUE;
1986: B->structurally_symmetric_set = PETSC_TRUE;
1988: PetscObjectChangeTypeName((PetscObject)B,MATSEQSBAIJ);
1990: PetscOptionsBegin(PetscObjectComm((PetscObject)B),((PetscObject)B)->prefix,"Options for SEQSBAIJ matrix","Mat");
1991: PetscOptionsBool("-mat_no_unroll","Do not optimize for inodes (slower)",NULL,no_unroll,&no_unroll,NULL);
1992: if (no_unroll) {
1993: PetscInfo(B,"Not using Inode routines due to -mat_no_unroll\n");
1994: }
1995: PetscOptionsBool("-mat_no_inode","Do not optimize for inodes (slower)",NULL,no_inode,&no_inode,NULL);
1996: if (no_inode) {
1997: PetscInfo(B,"Not using Inode routines due to -mat_no_inode\n");
1998: }
1999: PetscOptionsInt("-mat_inode_limit","Do not use inodes larger then this value",NULL,b->inode.limit,&b->inode.limit,NULL);
2000: PetscOptionsEnd();
2001: b->inode.use = (PetscBool)(!(no_unroll || no_inode));
2002: if (b->inode.limit > b->inode.max_limit) b->inode.limit = b->inode.max_limit;
2003: return(0);
2004: }
2008: /*@C
2009: MatSeqSBAIJSetPreallocation - Creates a sparse symmetric matrix in block AIJ (block
2010: compressed row) format. For good matrix assembly performance the
2011: user should preallocate the matrix storage by setting the parameter nz
2012: (or the array nnz). By setting these parameters accurately, performance
2013: during matrix assembly can be increased by more than a factor of 50.
2015: Collective on Mat
2017: Input Parameters:
2018: + B - the symmetric matrix
2019: . bs - size of block
2020: . nz - number of block nonzeros per block row (same for all rows)
2021: - nnz - array containing the number of block nonzeros in the upper triangular plus
2022: diagonal portion of each block (possibly different for each block row) or NULL
2024: Options Database Keys:
2025: . -mat_no_unroll - uses code that does not unroll the loops in the
2026: block calculations (much slower)
2027: . -mat_block_size - size of the blocks to use (only works if a negative bs is passed in
2029: Level: intermediate
2031: Notes:
2032: Specify the preallocated storage with either nz or nnz (not both).
2033: Set nz=PETSC_DEFAULT and nnz=NULL for PETSc to control dynamic memory
2034: allocation. See Users-Manual: ch_mat for details.
2036: You can call MatGetInfo() to get information on how effective the preallocation was;
2037: for example the fields mallocs,nz_allocated,nz_used,nz_unneeded;
2038: You can also run with the option -info and look for messages with the string
2039: malloc in them to see if additional memory allocation was needed.
2041: If the nnz parameter is given then the nz parameter is ignored
2044: .seealso: MatCreate(), MatCreateSeqAIJ(), MatSetValues(), MatCreateSBAIJ()
2045: @*/
2046: PetscErrorCode MatSeqSBAIJSetPreallocation(Mat B,PetscInt bs,PetscInt nz,const PetscInt nnz[])
2047: {
2054: PetscTryMethod(B,"MatSeqSBAIJSetPreallocation_C",(Mat,PetscInt,PetscInt,const PetscInt[]),(B,bs,nz,nnz));
2055: return(0);
2056: }
2058: #undef __FUNCT__
2060: /*@C
2061: MatSeqSBAIJSetPreallocationCSR - Allocates memory for a sparse sequential matrix in symmetric block AIJ format.
2063: Input Parameters:
2064: + B - the matrix
2065: . i - the indices into j for the start of each local row (starts with zero)
2066: . j - the column indices for each local row (starts with zero) these must be sorted for each row
2067: - v - optional values in the matrix
2069: Level: developer
2071: Notes:
2072: The order of the entries in values is specified by the MatOption MAT_ROW_ORIENTED. For example, C programs
2073: may want to use the default MAT_ROW_ORIENTED=PETSC_TRUE and use an array v[nnz][bs][bs] where the second index is
2074: over rows within a block and the last index is over columns within a block row. Fortran programs will likely set
2075: MAT_ROW_ORIENTED=PETSC_FALSE and use a Fortran array v(bs,bs,nnz) in which the first index is over rows within a
2076: block column and the second index is over columns within a block.
2078: .keywords: matrix, block, aij, compressed row, sparse
2080: .seealso: MatCreate(), MatCreateSeqSBAIJ(), MatSetValuesBlocked(), MatSeqSBAIJSetPreallocation(), MATSEQSBAIJ
2081: @*/
2082: PetscErrorCode MatSeqSBAIJSetPreallocationCSR(Mat B,PetscInt bs,const PetscInt i[],const PetscInt j[], const PetscScalar v[])
2083: {
2090: PetscTryMethod(B,"MatSeqSBAIJSetPreallocationCSR_C",(Mat,PetscInt,const PetscInt[],const PetscInt[],const PetscScalar[]),(B,bs,i,j,v));
2091: return(0);
2092: }
2096: /*@C
2097: MatCreateSeqSBAIJ - Creates a sparse symmetric matrix in block AIJ (block
2098: compressed row) format. For good matrix assembly performance the
2099: user should preallocate the matrix storage by setting the parameter nz
2100: (or the array nnz). By setting these parameters accurately, performance
2101: during matrix assembly can be increased by more than a factor of 50.
2103: Collective on MPI_Comm
2105: Input Parameters:
2106: + comm - MPI communicator, set to PETSC_COMM_SELF
2107: . bs - size of block
2108: . m - number of rows, or number of columns
2109: . nz - number of block nonzeros per block row (same for all rows)
2110: - nnz - array containing the number of block nonzeros in the upper triangular plus
2111: diagonal portion of each block (possibly different for each block row) or NULL
2113: Output Parameter:
2114: . A - the symmetric matrix
2116: Options Database Keys:
2117: . -mat_no_unroll - uses code that does not unroll the loops in the
2118: block calculations (much slower)
2119: . -mat_block_size - size of the blocks to use
2121: Level: intermediate
2123: It is recommended that one use the MatCreate(), MatSetType() and/or MatSetFromOptions(),
2124: MatXXXXSetPreallocation() paradgm instead of this routine directly.
2125: [MatXXXXSetPreallocation() is, for example, MatSeqAIJSetPreallocation]
2127: Notes:
2128: The number of rows and columns must be divisible by blocksize.
2129: This matrix type does not support complex Hermitian operation.
2131: Specify the preallocated storage with either nz or nnz (not both).
2132: Set nz=PETSC_DEFAULT and nnz=NULL for PETSc to control dynamic memory
2133: allocation. See Users-Manual: ch_mat for details.
2135: If the nnz parameter is given then the nz parameter is ignored
2137: .seealso: MatCreate(), MatCreateSeqAIJ(), MatSetValues(), MatCreateSBAIJ()
2138: @*/
2139: PetscErrorCode MatCreateSeqSBAIJ(MPI_Comm comm,PetscInt bs,PetscInt m,PetscInt n,PetscInt nz,const PetscInt nnz[],Mat *A)
2140: {
2144: MatCreate(comm,A);
2145: MatSetSizes(*A,m,n,m,n);
2146: MatSetType(*A,MATSEQSBAIJ);
2147: MatSeqSBAIJSetPreallocation_SeqSBAIJ(*A,bs,nz,(PetscInt*)nnz);
2148: return(0);
2149: }
2153: PetscErrorCode MatDuplicate_SeqSBAIJ(Mat A,MatDuplicateOption cpvalues,Mat *B)
2154: {
2155: Mat C;
2156: Mat_SeqSBAIJ *c,*a = (Mat_SeqSBAIJ*)A->data;
2158: PetscInt i,mbs = a->mbs,nz = a->nz,bs2 =a->bs2;
2161: if (a->i[mbs] != nz) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_PLIB,"Corrupt matrix");
2163: *B = 0;
2164: MatCreate(PetscObjectComm((PetscObject)A),&C);
2165: MatSetSizes(C,A->rmap->N,A->cmap->n,A->rmap->N,A->cmap->n);
2166: MatSetType(C,MATSEQSBAIJ);
2167: PetscMemcpy(C->ops,A->ops,sizeof(struct _MatOps));
2168: c = (Mat_SeqSBAIJ*)C->data;
2170: C->preallocated = PETSC_TRUE;
2171: C->factortype = A->factortype;
2172: c->row = 0;
2173: c->icol = 0;
2174: c->saved_values = 0;
2175: c->keepnonzeropattern = a->keepnonzeropattern;
2176: C->assembled = PETSC_TRUE;
2178: PetscLayoutReference(A->rmap,&C->rmap);
2179: PetscLayoutReference(A->cmap,&C->cmap);
2180: c->bs2 = a->bs2;
2181: c->mbs = a->mbs;
2182: c->nbs = a->nbs;
2184: if (cpvalues == MAT_SHARE_NONZERO_PATTERN) {
2185: c->imax = a->imax;
2186: c->ilen = a->ilen;
2187: c->free_imax_ilen = PETSC_FALSE;
2188: } else {
2189: PetscMalloc2((mbs+1),&c->imax,(mbs+1),&c->ilen);
2190: PetscLogObjectMemory((PetscObject)C,2*(mbs+1)*sizeof(PetscInt));
2191: for (i=0; i<mbs; i++) {
2192: c->imax[i] = a->imax[i];
2193: c->ilen[i] = a->ilen[i];
2194: }
2195: c->free_imax_ilen = PETSC_TRUE;
2196: }
2198: /* allocate the matrix space */
2199: if (cpvalues == MAT_SHARE_NONZERO_PATTERN) {
2200: PetscMalloc1(bs2*nz,&c->a);
2201: PetscLogObjectMemory((PetscObject)C,nz*bs2*sizeof(MatScalar));
2202: c->i = a->i;
2203: c->j = a->j;
2204: c->singlemalloc = PETSC_FALSE;
2205: c->free_a = PETSC_TRUE;
2206: c->free_ij = PETSC_FALSE;
2207: c->parent = A;
2208: PetscObjectReference((PetscObject)A);
2209: MatSetOption(A,MAT_NEW_NONZERO_LOCATION_ERR,PETSC_TRUE);
2210: MatSetOption(C,MAT_NEW_NONZERO_LOCATION_ERR,PETSC_TRUE);
2211: } else {
2212: PetscMalloc3(bs2*nz,&c->a,nz,&c->j,mbs+1,&c->i);
2213: PetscMemcpy(c->i,a->i,(mbs+1)*sizeof(PetscInt));
2214: PetscLogObjectMemory((PetscObject)C,(mbs+1)*sizeof(PetscInt) + nz*(bs2*sizeof(MatScalar) + sizeof(PetscInt)));
2215: c->singlemalloc = PETSC_TRUE;
2216: c->free_a = PETSC_TRUE;
2217: c->free_ij = PETSC_TRUE;
2218: }
2219: if (mbs > 0) {
2220: if (cpvalues != MAT_SHARE_NONZERO_PATTERN) {
2221: PetscMemcpy(c->j,a->j,nz*sizeof(PetscInt));
2222: }
2223: if (cpvalues == MAT_COPY_VALUES) {
2224: PetscMemcpy(c->a,a->a,bs2*nz*sizeof(MatScalar));
2225: } else {
2226: PetscMemzero(c->a,bs2*nz*sizeof(MatScalar));
2227: }
2228: if (a->jshort) {
2229: /* cannot share jshort, it is reallocated in MatAssemblyEnd_SeqSBAIJ() */
2230: /* if the parent matrix is reassembled, this child matrix will never notice */
2231: PetscMalloc1(nz,&c->jshort);
2232: PetscLogObjectMemory((PetscObject)C,nz*sizeof(unsigned short));
2233: PetscMemcpy(c->jshort,a->jshort,nz*sizeof(unsigned short));
2235: c->free_jshort = PETSC_TRUE;
2236: }
2237: }
2239: c->roworiented = a->roworiented;
2240: c->nonew = a->nonew;
2242: if (a->diag) {
2243: if (cpvalues == MAT_SHARE_NONZERO_PATTERN) {
2244: c->diag = a->diag;
2245: c->free_diag = PETSC_FALSE;
2246: } else {
2247: PetscMalloc1(mbs,&c->diag);
2248: PetscLogObjectMemory((PetscObject)C,mbs*sizeof(PetscInt));
2249: for (i=0; i<mbs; i++) c->diag[i] = a->diag[i];
2250: c->free_diag = PETSC_TRUE;
2251: }
2252: }
2253: c->nz = a->nz;
2254: c->maxnz = a->nz; /* Since we allocate exactly the right amount */
2255: c->solve_work = 0;
2256: c->mult_work = 0;
2258: *B = C;
2259: PetscFunctionListDuplicate(((PetscObject)A)->qlist,&((PetscObject)C)->qlist);
2260: return(0);
2261: }
2265: PetscErrorCode MatLoad_SeqSBAIJ(Mat newmat,PetscViewer viewer)
2266: {
2267: Mat_SeqSBAIJ *a;
2269: int fd;
2270: PetscMPIInt size;
2271: PetscInt i,nz,header[4],*rowlengths=0,M,N,bs=1;
2272: PetscInt *mask,mbs,*jj,j,rowcount,nzcount,k,*s_browlengths,maskcount;
2273: PetscInt kmax,jcount,block,idx,point,nzcountb,extra_rows,rows,cols;
2274: PetscInt *masked,nmask,tmp,bs2,ishift;
2275: PetscScalar *aa;
2276: MPI_Comm comm;
2279: PetscObjectGetComm((PetscObject)viewer,&comm);
2280: PetscOptionsGetInt(((PetscObject)newmat)->prefix,"-matload_block_size",&bs,NULL);
2281: bs2 = bs*bs;
2283: MPI_Comm_size(comm,&size);
2284: if (size > 1) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_WRONG,"view must have one processor");
2285: PetscViewerBinaryGetDescriptor(viewer,&fd);
2286: PetscBinaryRead(fd,header,4,PETSC_INT);
2287: if (header[0] != MAT_FILE_CLASSID) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_FILE_UNEXPECTED,"not Mat object");
2288: M = header[1]; N = header[2]; nz = header[3];
2290: if (header[3] < 0) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_FILE_UNEXPECTED,"Matrix stored in special format, cannot load as SeqSBAIJ");
2292: if (M != N) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Can only do square matrices");
2294: /*
2295: This code adds extra rows to make sure the number of rows is
2296: divisible by the blocksize
2297: */
2298: mbs = M/bs;
2299: extra_rows = bs - M + bs*(mbs);
2300: if (extra_rows == bs) extra_rows = 0;
2301: else mbs++;
2302: if (extra_rows) {
2303: PetscInfo(viewer,"Padding loaded matrix to match blocksize\n");
2304: }
2306: /* Set global sizes if not already set */
2307: if (newmat->rmap->n < 0 && newmat->rmap->N < 0 && newmat->cmap->n < 0 && newmat->cmap->N < 0) {
2308: MatSetSizes(newmat,PETSC_DECIDE,PETSC_DECIDE,M+extra_rows,N+extra_rows);
2309: } else { /* Check if the matrix global sizes are correct */
2310: MatGetSize(newmat,&rows,&cols);
2311: 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);
2312: }
2314: /* read in row lengths */
2315: PetscMalloc1((M+extra_rows),&rowlengths);
2316: PetscBinaryRead(fd,rowlengths,M,PETSC_INT);
2317: for (i=0; i<extra_rows; i++) rowlengths[M+i] = 1;
2319: /* read in column indices */
2320: PetscMalloc1((nz+extra_rows),&jj);
2321: PetscBinaryRead(fd,jj,nz,PETSC_INT);
2322: for (i=0; i<extra_rows; i++) jj[nz+i] = M+i;
2324: /* loop over row lengths determining block row lengths */
2325: PetscCalloc1(mbs,&s_browlengths);
2326: PetscMalloc2(mbs,&mask,mbs,&masked);
2327: PetscMemzero(mask,mbs*sizeof(PetscInt));
2328: rowcount = 0;
2329: nzcount = 0;
2330: for (i=0; i<mbs; i++) {
2331: nmask = 0;
2332: for (j=0; j<bs; j++) {
2333: kmax = rowlengths[rowcount];
2334: for (k=0; k<kmax; k++) {
2335: tmp = jj[nzcount++]/bs; /* block col. index */
2336: if (!mask[tmp] && tmp >= i) {masked[nmask++] = tmp; mask[tmp] = 1;}
2337: }
2338: rowcount++;
2339: }
2340: s_browlengths[i] += nmask;
2342: /* zero out the mask elements we set */
2343: for (j=0; j<nmask; j++) mask[masked[j]] = 0;
2344: }
2346: /* Do preallocation */
2347: MatSeqSBAIJSetPreallocation_SeqSBAIJ(newmat,bs,0,s_browlengths);
2348: a = (Mat_SeqSBAIJ*)newmat->data;
2350: /* set matrix "i" values */
2351: a->i[0] = 0;
2352: for (i=1; i<= mbs; i++) {
2353: a->i[i] = a->i[i-1] + s_browlengths[i-1];
2354: a->ilen[i-1] = s_browlengths[i-1];
2355: }
2356: a->nz = a->i[mbs];
2358: /* read in nonzero values */
2359: PetscMalloc1((nz+extra_rows),&aa);
2360: PetscBinaryRead(fd,aa,nz,PETSC_SCALAR);
2361: for (i=0; i<extra_rows; i++) aa[nz+i] = 1.0;
2363: /* set "a" and "j" values into matrix */
2364: nzcount = 0; jcount = 0;
2365: for (i=0; i<mbs; i++) {
2366: nzcountb = nzcount;
2367: nmask = 0;
2368: for (j=0; j<bs; j++) {
2369: kmax = rowlengths[i*bs+j];
2370: for (k=0; k<kmax; k++) {
2371: tmp = jj[nzcount++]/bs; /* block col. index */
2372: if (!mask[tmp] && tmp >= i) { masked[nmask++] = tmp; mask[tmp] = 1;}
2373: }
2374: }
2375: /* sort the masked values */
2376: PetscSortInt(nmask,masked);
2378: /* set "j" values into matrix */
2379: maskcount = 1;
2380: for (j=0; j<nmask; j++) {
2381: a->j[jcount++] = masked[j];
2382: mask[masked[j]] = maskcount++;
2383: }
2385: /* set "a" values into matrix */
2386: ishift = bs2*a->i[i];
2387: for (j=0; j<bs; j++) {
2388: kmax = rowlengths[i*bs+j];
2389: for (k=0; k<kmax; k++) {
2390: tmp = jj[nzcountb]/bs; /* block col. index */
2391: if (tmp >= i) {
2392: block = mask[tmp] - 1;
2393: point = jj[nzcountb] - bs*tmp;
2394: idx = ishift + bs2*block + j + bs*point;
2395: a->a[idx] = aa[nzcountb];
2396: }
2397: nzcountb++;
2398: }
2399: }
2400: /* zero out the mask elements we set */
2401: for (j=0; j<nmask; j++) mask[masked[j]] = 0;
2402: }
2403: if (jcount != a->nz) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_FILE_UNEXPECTED,"Bad binary matrix");
2405: PetscFree(rowlengths);
2406: PetscFree(s_browlengths);
2407: PetscFree(aa);
2408: PetscFree(jj);
2409: PetscFree2(mask,masked);
2411: MatAssemblyBegin(newmat,MAT_FINAL_ASSEMBLY);
2412: MatAssemblyEnd(newmat,MAT_FINAL_ASSEMBLY);
2413: return(0);
2414: }
2418: /*@
2419: MatCreateSeqSBAIJWithArrays - Creates an sequential SBAIJ matrix using matrix elements
2420: (upper triangular entries in CSR format) provided by the user.
2422: Collective on MPI_Comm
2424: Input Parameters:
2425: + comm - must be an MPI communicator of size 1
2426: . bs - size of block
2427: . m - number of rows
2428: . n - number of columns
2429: . i - row indices
2430: . j - column indices
2431: - a - matrix values
2433: Output Parameter:
2434: . mat - the matrix
2436: Level: advanced
2438: Notes:
2439: The i, j, and a arrays are not copied by this routine, the user must free these arrays
2440: once the matrix is destroyed
2442: You cannot set new nonzero locations into this matrix, that will generate an error.
2444: The i and j indices are 0 based
2446: 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
2447: it is the regular CSR format excluding the lower triangular elements.
2449: .seealso: MatCreate(), MatCreateSBAIJ(), MatCreateSeqSBAIJ()
2451: @*/
2452: PetscErrorCode MatCreateSeqSBAIJWithArrays(MPI_Comm comm,PetscInt bs,PetscInt m,PetscInt n,PetscInt *i,PetscInt *j,PetscScalar *a,Mat *mat)
2453: {
2455: PetscInt ii;
2456: Mat_SeqSBAIJ *sbaij;
2459: if (bs != 1) SETERRQ1(PETSC_COMM_SELF,PETSC_ERR_SUP,"block size %D > 1 is not supported yet",bs);
2460: if (i[0]) SETERRQ(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"i (row indices) must start with 0");
2462: MatCreate(comm,mat);
2463: MatSetSizes(*mat,m,n,m,n);
2464: MatSetType(*mat,MATSEQSBAIJ);
2465: MatSeqSBAIJSetPreallocation_SeqSBAIJ(*mat,bs,MAT_SKIP_ALLOCATION,0);
2466: sbaij = (Mat_SeqSBAIJ*)(*mat)->data;
2467: PetscMalloc2(m,&sbaij->imax,m,&sbaij->ilen);
2468: PetscLogObjectMemory((PetscObject)*mat,2*m*sizeof(PetscInt));
2470: sbaij->i = i;
2471: sbaij->j = j;
2472: sbaij->a = a;
2474: sbaij->singlemalloc = PETSC_FALSE;
2475: sbaij->nonew = -1; /*this indicates that inserting a new value in the matrix that generates a new nonzero is an error*/
2476: sbaij->free_a = PETSC_FALSE;
2477: sbaij->free_ij = PETSC_FALSE;
2479: for (ii=0; ii<m; ii++) {
2480: sbaij->ilen[ii] = sbaij->imax[ii] = i[ii+1] - i[ii];
2481: #if defined(PETSC_USE_DEBUG)
2482: 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]);
2483: #endif
2484: }
2485: #if defined(PETSC_USE_DEBUG)
2486: for (ii=0; ii<sbaij->i[m]; ii++) {
2487: if (j[ii] < 0) SETERRQ2(PETSC_COMM_SELF,PETSC_ERR_ARG_OUTOFRANGE,"Negative column index at location = %d index = %d",ii,j[ii]);
2488: 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]);
2489: }
2490: #endif
2492: MatAssemblyBegin(*mat,MAT_FINAL_ASSEMBLY);
2493: MatAssemblyEnd(*mat,MAT_FINAL_ASSEMBLY);
2494: return(0);
2495: }