Actual source code: sbaij.c
1: /*
2: Defines the basic matrix operations for the SBAIJ (compressed row)
3: matrix storage format.
4: */
5: #include <../src/mat/impls/baij/seq/baij.h>
6: #include <../src/mat/impls/sbaij/seq/sbaij.h>
7: #include <petscblaslapack.h>
9: #include <../src/mat/impls/sbaij/seq/relax.h>
10: #define USESHORT
11: #include <../src/mat/impls/sbaij/seq/relax.h>
13: /* defines MatSetValues_Seq_Hash(), MatAssemblyEnd_Seq_Hash(), MatSetUp_Seq_Hash() */
14: #define TYPE SBAIJ
15: #define TYPE_SBAIJ
16: #define TYPE_BS
17: #include "../src/mat/impls/aij/seq/seqhashmatsetvalues.h"
18: #undef TYPE_BS
19: #define TYPE_BS _BS
20: #define TYPE_BS_ON
21: #include "../src/mat/impls/aij/seq/seqhashmatsetvalues.h"
22: #undef TYPE_BS
23: #undef TYPE_SBAIJ
24: #include "../src/mat/impls/aij/seq/seqhashmat.h"
25: #undef TYPE
26: #undef TYPE_BS_ON
28: #if defined(PETSC_HAVE_ELEMENTAL)
29: PETSC_INTERN PetscErrorCode MatConvert_SeqSBAIJ_Elemental(Mat, MatType, MatReuse, Mat *);
30: #endif
31: #if defined(PETSC_HAVE_SCALAPACK)
32: PETSC_INTERN PetscErrorCode MatConvert_SBAIJ_ScaLAPACK(Mat, MatType, MatReuse, Mat *);
33: #endif
34: PETSC_INTERN PetscErrorCode MatConvert_MPISBAIJ_Basic(Mat, MatType, MatReuse, Mat *);
36: /*
37: Checks for missing diagonals
38: */
39: static PetscErrorCode MatMissingDiagonal_SeqSBAIJ(Mat A, PetscBool *missing, PetscInt *dd)
40: {
41: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ *)A->data;
42: PetscInt *diag, *ii = a->i, i;
44: PetscFunctionBegin;
45: PetscCall(MatMarkDiagonal_SeqSBAIJ(A));
46: *missing = PETSC_FALSE;
47: if (A->rmap->n > 0 && !ii) {
48: *missing = PETSC_TRUE;
49: if (dd) *dd = 0;
50: PetscCall(PetscInfo(A, "Matrix has no entries therefore is missing diagonal\n"));
51: } else {
52: diag = a->diag;
53: for (i = 0; i < a->mbs; i++) {
54: if (diag[i] >= ii[i + 1]) {
55: *missing = PETSC_TRUE;
56: if (dd) *dd = i;
57: break;
58: }
59: }
60: }
61: PetscFunctionReturn(PETSC_SUCCESS);
62: }
64: PetscErrorCode MatMarkDiagonal_SeqSBAIJ(Mat A)
65: {
66: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ *)A->data;
67: PetscInt i, j;
69: PetscFunctionBegin;
70: if (!a->diag) {
71: PetscCall(PetscMalloc1(a->mbs, &a->diag));
72: a->free_diag = PETSC_TRUE;
73: }
74: for (i = 0; i < a->mbs; i++) {
75: a->diag[i] = a->i[i + 1];
76: for (j = a->i[i]; j < a->i[i + 1]; j++) {
77: if (a->j[j] == i) {
78: a->diag[i] = j;
79: break;
80: }
81: }
82: }
83: PetscFunctionReturn(PETSC_SUCCESS);
84: }
86: static PetscErrorCode MatGetRowIJ_SeqSBAIJ(Mat A, PetscInt oshift, PetscBool symmetric, PetscBool blockcompressed, PetscInt *nn, const PetscInt *inia[], const PetscInt *inja[], PetscBool *done)
87: {
88: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ *)A->data;
89: PetscInt i, j, n = a->mbs, nz = a->i[n], *tia, *tja, bs = A->rmap->bs, k, l, cnt;
90: PetscInt **ia = (PetscInt **)inia, **ja = (PetscInt **)inja;
92: PetscFunctionBegin;
93: *nn = n;
94: if (!ia) PetscFunctionReturn(PETSC_SUCCESS);
95: if (symmetric) {
96: PetscCall(MatToSymmetricIJ_SeqAIJ(n, a->i, a->j, PETSC_FALSE, 0, 0, &tia, &tja));
97: nz = tia[n];
98: } else {
99: tia = a->i;
100: tja = a->j;
101: }
103: if (!blockcompressed && bs > 1) {
104: (*nn) *= bs;
105: /* malloc & create the natural set of indices */
106: PetscCall(PetscMalloc1((n + 1) * bs, ia));
107: if (n) {
108: (*ia)[0] = oshift;
109: for (j = 1; j < bs; j++) (*ia)[j] = (tia[1] - tia[0]) * bs + (*ia)[j - 1];
110: }
112: for (i = 1; i < n; i++) {
113: (*ia)[i * bs] = (tia[i] - tia[i - 1]) * bs + (*ia)[i * bs - 1];
114: for (j = 1; j < bs; j++) (*ia)[i * bs + j] = (tia[i + 1] - tia[i]) * bs + (*ia)[i * bs + j - 1];
115: }
116: if (n) (*ia)[n * bs] = (tia[n] - tia[n - 1]) * bs + (*ia)[n * bs - 1];
118: if (inja) {
119: PetscCall(PetscMalloc1(nz * bs * bs, ja));
120: cnt = 0;
121: for (i = 0; i < n; i++) {
122: for (j = 0; j < bs; j++) {
123: for (k = tia[i]; k < tia[i + 1]; k++) {
124: for (l = 0; l < bs; l++) (*ja)[cnt++] = bs * tja[k] + l;
125: }
126: }
127: }
128: }
130: if (symmetric) { /* deallocate memory allocated in MatToSymmetricIJ_SeqAIJ() */
131: PetscCall(PetscFree(tia));
132: PetscCall(PetscFree(tja));
133: }
134: } else if (oshift == 1) {
135: if (symmetric) {
136: nz = tia[A->rmap->n / bs];
137: /* add 1 to i and j indices */
138: for (i = 0; i < A->rmap->n / bs + 1; i++) tia[i] = tia[i] + 1;
139: *ia = tia;
140: if (ja) {
141: for (i = 0; i < nz; i++) tja[i] = tja[i] + 1;
142: *ja = tja;
143: }
144: } else {
145: nz = a->i[A->rmap->n / bs];
146: /* malloc space and add 1 to i and j indices */
147: PetscCall(PetscMalloc1(A->rmap->n / bs + 1, ia));
148: for (i = 0; i < A->rmap->n / bs + 1; i++) (*ia)[i] = a->i[i] + 1;
149: if (ja) {
150: PetscCall(PetscMalloc1(nz, ja));
151: for (i = 0; i < nz; i++) (*ja)[i] = a->j[i] + 1;
152: }
153: }
154: } else {
155: *ia = tia;
156: if (ja) *ja = tja;
157: }
158: PetscFunctionReturn(PETSC_SUCCESS);
159: }
161: static PetscErrorCode MatRestoreRowIJ_SeqSBAIJ(Mat A, PetscInt oshift, PetscBool symmetric, PetscBool blockcompressed, PetscInt *nn, const PetscInt *ia[], const PetscInt *ja[], PetscBool *done)
162: {
163: PetscFunctionBegin;
164: if (!ia) PetscFunctionReturn(PETSC_SUCCESS);
165: if ((!blockcompressed && A->rmap->bs > 1) || (symmetric || oshift == 1)) {
166: PetscCall(PetscFree(*ia));
167: if (ja) PetscCall(PetscFree(*ja));
168: }
169: PetscFunctionReturn(PETSC_SUCCESS);
170: }
172: PetscErrorCode MatDestroy_SeqSBAIJ(Mat A)
173: {
174: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ *)A->data;
176: PetscFunctionBegin;
177: if (A->hash_active) {
178: PetscInt bs;
179: A->ops[0] = a->cops;
180: PetscCall(PetscHMapIJVDestroy(&a->ht));
181: PetscCall(MatGetBlockSize(A, &bs));
182: if (bs > 1) PetscCall(PetscHSetIJDestroy(&a->bht));
183: PetscCall(PetscFree(a->dnz));
184: PetscCall(PetscFree(a->bdnz));
185: A->hash_active = PETSC_FALSE;
186: }
187: PetscCall(PetscLogObjectState((PetscObject)A, "Rows=%" PetscInt_FMT ", NZ=%" PetscInt_FMT, A->rmap->N, a->nz));
188: PetscCall(MatSeqXAIJFreeAIJ(A, &a->a, &a->j, &a->i));
189: if (a->free_diag) PetscCall(PetscFree(a->diag));
190: PetscCall(ISDestroy(&a->row));
191: PetscCall(ISDestroy(&a->col));
192: PetscCall(ISDestroy(&a->icol));
193: PetscCall(PetscFree(a->idiag));
194: PetscCall(PetscFree(a->inode.size));
195: if (a->free_imax_ilen) PetscCall(PetscFree2(a->imax, a->ilen));
196: PetscCall(PetscFree(a->solve_work));
197: PetscCall(PetscFree(a->sor_work));
198: PetscCall(PetscFree(a->solves_work));
199: PetscCall(PetscFree(a->mult_work));
200: PetscCall(PetscFree(a->saved_values));
201: if (a->free_jshort) PetscCall(PetscFree(a->jshort));
202: PetscCall(PetscFree(a->inew));
203: PetscCall(MatDestroy(&a->parent));
204: PetscCall(PetscFree(A->data));
206: PetscCall(PetscObjectChangeTypeName((PetscObject)A, NULL));
207: PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatSeqSBAIJGetArray_C", NULL));
208: PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatSeqSBAIJRestoreArray_C", NULL));
209: PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatStoreValues_C", NULL));
210: PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatRetrieveValues_C", NULL));
211: PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatSeqSBAIJSetColumnIndices_C", NULL));
212: PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatConvert_seqsbaij_seqaij_C", NULL));
213: PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatConvert_seqsbaij_seqbaij_C", NULL));
214: PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatSeqSBAIJSetPreallocation_C", NULL));
215: PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatSeqSBAIJSetPreallocationCSR_C", NULL));
216: #if defined(PETSC_HAVE_ELEMENTAL)
217: PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatConvert_seqsbaij_elemental_C", NULL));
218: #endif
219: #if defined(PETSC_HAVE_SCALAPACK)
220: PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatConvert_seqsbaij_scalapack_C", NULL));
221: #endif
222: PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatFactorGetSolverType_C", NULL));
223: PetscFunctionReturn(PETSC_SUCCESS);
224: }
226: static PetscErrorCode MatSetOption_SeqSBAIJ(Mat A, MatOption op, PetscBool flg)
227: {
228: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ *)A->data;
229: #if defined(PETSC_USE_COMPLEX)
230: PetscInt bs;
231: #endif
233: PetscFunctionBegin;
234: #if defined(PETSC_USE_COMPLEX)
235: PetscCall(MatGetBlockSize(A, &bs));
236: #endif
237: switch (op) {
238: case MAT_ROW_ORIENTED:
239: a->roworiented = flg;
240: break;
241: case MAT_KEEP_NONZERO_PATTERN:
242: a->keepnonzeropattern = flg;
243: break;
244: case MAT_NEW_NONZERO_LOCATIONS:
245: a->nonew = (flg ? 0 : 1);
246: break;
247: case MAT_NEW_NONZERO_LOCATION_ERR:
248: a->nonew = (flg ? -1 : 0);
249: break;
250: case MAT_NEW_NONZERO_ALLOCATION_ERR:
251: a->nonew = (flg ? -2 : 0);
252: break;
253: case MAT_UNUSED_NONZERO_LOCATION_ERR:
254: a->nounused = (flg ? -1 : 0);
255: break;
256: case MAT_FORCE_DIAGONAL_ENTRIES:
257: case MAT_IGNORE_OFF_PROC_ENTRIES:
258: case MAT_USE_HASH_TABLE:
259: case MAT_SORTED_FULL:
260: PetscCall(PetscInfo(A, "Option %s ignored\n", MatOptions[op]));
261: break;
262: case MAT_HERMITIAN:
263: #if defined(PETSC_USE_COMPLEX)
264: if (flg) { /* disable transpose ops */
265: PetscCheck(bs <= 1, PETSC_COMM_SELF, PETSC_ERR_SUP, "No support for Hermitian with block size greater than 1");
266: A->ops->multtranspose = NULL;
267: A->ops->multtransposeadd = NULL;
268: A->symmetric = PETSC_BOOL3_FALSE;
269: }
270: #endif
271: break;
272: case MAT_SYMMETRIC:
273: case MAT_SPD:
274: #if defined(PETSC_USE_COMPLEX)
275: if (flg) { /* An hermitian and symmetric matrix has zero imaginary part (restore back transpose ops) */
276: A->ops->multtranspose = A->ops->mult;
277: A->ops->multtransposeadd = A->ops->multadd;
278: }
279: #endif
280: break;
281: /* These options are handled directly by MatSetOption() */
282: case MAT_STRUCTURALLY_SYMMETRIC:
283: case MAT_SYMMETRY_ETERNAL:
284: case MAT_STRUCTURAL_SYMMETRY_ETERNAL:
285: case MAT_STRUCTURE_ONLY:
286: case MAT_SPD_ETERNAL:
287: /* These options are handled directly by MatSetOption() */
288: break;
289: case MAT_IGNORE_LOWER_TRIANGULAR:
290: a->ignore_ltriangular = flg;
291: break;
292: case MAT_ERROR_LOWER_TRIANGULAR:
293: a->ignore_ltriangular = flg;
294: break;
295: case MAT_GETROW_UPPERTRIANGULAR:
296: a->getrow_utriangular = flg;
297: break;
298: case MAT_SUBMAT_SINGLEIS:
299: break;
300: default:
301: SETERRQ(PETSC_COMM_SELF, PETSC_ERR_SUP, "unknown option %d", op);
302: }
303: PetscFunctionReturn(PETSC_SUCCESS);
304: }
306: PetscErrorCode MatGetRow_SeqSBAIJ(Mat A, PetscInt row, PetscInt *nz, PetscInt **idx, PetscScalar **v)
307: {
308: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ *)A->data;
310: PetscFunctionBegin;
311: PetscCheck(!A || a->getrow_utriangular, 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()");
313: /* Get the upper triangular part of the row */
314: PetscCall(MatGetRow_SeqBAIJ_private(A, row, nz, idx, v, a->i, a->j, a->a));
315: PetscFunctionReturn(PETSC_SUCCESS);
316: }
318: PetscErrorCode MatRestoreRow_SeqSBAIJ(Mat A, PetscInt row, PetscInt *nz, PetscInt **idx, PetscScalar **v)
319: {
320: PetscFunctionBegin;
321: if (idx) PetscCall(PetscFree(*idx));
322: if (v) PetscCall(PetscFree(*v));
323: PetscFunctionReturn(PETSC_SUCCESS);
324: }
326: static PetscErrorCode MatGetRowUpperTriangular_SeqSBAIJ(Mat A)
327: {
328: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ *)A->data;
330: PetscFunctionBegin;
331: a->getrow_utriangular = PETSC_TRUE;
332: PetscFunctionReturn(PETSC_SUCCESS);
333: }
335: static PetscErrorCode MatRestoreRowUpperTriangular_SeqSBAIJ(Mat A)
336: {
337: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ *)A->data;
339: PetscFunctionBegin;
340: a->getrow_utriangular = PETSC_FALSE;
341: PetscFunctionReturn(PETSC_SUCCESS);
342: }
344: static PetscErrorCode MatTranspose_SeqSBAIJ(Mat A, MatReuse reuse, Mat *B)
345: {
346: PetscFunctionBegin;
347: if (reuse == MAT_REUSE_MATRIX) PetscCall(MatTransposeCheckNonzeroState_Private(A, *B));
348: if (reuse == MAT_INITIAL_MATRIX) {
349: PetscCall(MatDuplicate(A, MAT_COPY_VALUES, B));
350: } else if (reuse == MAT_REUSE_MATRIX) {
351: PetscCall(MatCopy(A, *B, SAME_NONZERO_PATTERN));
352: }
353: PetscFunctionReturn(PETSC_SUCCESS);
354: }
356: static PetscErrorCode MatView_SeqSBAIJ_ASCII(Mat A, PetscViewer viewer)
357: {
358: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ *)A->data;
359: PetscInt i, j, bs = A->rmap->bs, k, l, bs2 = a->bs2;
360: PetscViewerFormat format;
361: PetscInt *diag;
362: const char *matname;
364: PetscFunctionBegin;
365: PetscCall(PetscViewerGetFormat(viewer, &format));
366: if (format == PETSC_VIEWER_ASCII_INFO || format == PETSC_VIEWER_ASCII_INFO_DETAIL) {
367: PetscCall(PetscViewerASCIIPrintf(viewer, " block size is %" PetscInt_FMT "\n", bs));
368: } else if (format == PETSC_VIEWER_ASCII_MATLAB) {
369: Mat aij;
371: if (A->factortype && bs > 1) {
372: PetscCall(PetscPrintf(PETSC_COMM_SELF, "Warning: matrix is factored with bs>1. MatView() with PETSC_VIEWER_ASCII_MATLAB is not supported and ignored!\n"));
373: PetscFunctionReturn(PETSC_SUCCESS);
374: }
375: PetscCall(MatConvert(A, MATSEQAIJ, MAT_INITIAL_MATRIX, &aij));
376: if (((PetscObject)A)->name) PetscCall(PetscObjectGetName((PetscObject)A, &matname));
377: if (((PetscObject)A)->name) PetscCall(PetscObjectSetName((PetscObject)aij, matname));
378: PetscCall(MatView_SeqAIJ(aij, viewer));
379: PetscCall(MatDestroy(&aij));
380: } else if (format == PETSC_VIEWER_ASCII_COMMON) {
381: Mat B;
383: PetscCall(MatConvert(A, MATSEQAIJ, MAT_INITIAL_MATRIX, &B));
384: if (((PetscObject)A)->name) PetscCall(PetscObjectGetName((PetscObject)A, &matname));
385: if (((PetscObject)A)->name) PetscCall(PetscObjectSetName((PetscObject)B, matname));
386: PetscCall(MatView_SeqAIJ(B, viewer));
387: PetscCall(MatDestroy(&B));
388: } else if (format == PETSC_VIEWER_ASCII_FACTOR_INFO) {
389: PetscFunctionReturn(PETSC_SUCCESS);
390: } else {
391: PetscCall(PetscViewerASCIIUseTabs(viewer, PETSC_FALSE));
392: if (A->factortype) { /* for factored matrix */
393: PetscCheck(bs <= 1, PETSC_COMM_SELF, PETSC_ERR_SUP, "matrix is factored with bs>1. Not implemented yet");
395: diag = a->diag;
396: for (i = 0; i < a->mbs; i++) { /* for row block i */
397: PetscCall(PetscViewerASCIIPrintf(viewer, "row %" PetscInt_FMT ":", i));
398: /* diagonal entry */
399: #if defined(PETSC_USE_COMPLEX)
400: if (PetscImaginaryPart(a->a[diag[i]]) > 0.0) {
401: PetscCall(PetscViewerASCIIPrintf(viewer, " (%" PetscInt_FMT ", %g + %g i) ", a->j[diag[i]], (double)PetscRealPart(1.0 / a->a[diag[i]]), (double)PetscImaginaryPart(1.0 / a->a[diag[i]])));
402: } else if (PetscImaginaryPart(a->a[diag[i]]) < 0.0) {
403: PetscCall(PetscViewerASCIIPrintf(viewer, " (%" PetscInt_FMT ", %g - %g i) ", a->j[diag[i]], (double)PetscRealPart(1.0 / a->a[diag[i]]), -(double)PetscImaginaryPart(1.0 / a->a[diag[i]])));
404: } else {
405: PetscCall(PetscViewerASCIIPrintf(viewer, " (%" PetscInt_FMT ", %g) ", a->j[diag[i]], (double)PetscRealPart(1.0 / a->a[diag[i]])));
406: }
407: #else
408: PetscCall(PetscViewerASCIIPrintf(viewer, " (%" PetscInt_FMT ", %g) ", a->j[diag[i]], (double)(1 / a->a[diag[i]])));
409: #endif
410: /* off-diagonal entries */
411: for (k = a->i[i]; k < a->i[i + 1] - 1; k++) {
412: #if defined(PETSC_USE_COMPLEX)
413: if (PetscImaginaryPart(a->a[k]) > 0.0) {
414: PetscCall(PetscViewerASCIIPrintf(viewer, " (%" PetscInt_FMT ", %g + %g i) ", bs * a->j[k], (double)PetscRealPart(a->a[k]), (double)PetscImaginaryPart(a->a[k])));
415: } else if (PetscImaginaryPart(a->a[k]) < 0.0) {
416: PetscCall(PetscViewerASCIIPrintf(viewer, " (%" PetscInt_FMT ", %g - %g i) ", bs * a->j[k], (double)PetscRealPart(a->a[k]), -(double)PetscImaginaryPart(a->a[k])));
417: } else {
418: PetscCall(PetscViewerASCIIPrintf(viewer, " (%" PetscInt_FMT ", %g) ", bs * a->j[k], (double)PetscRealPart(a->a[k])));
419: }
420: #else
421: PetscCall(PetscViewerASCIIPrintf(viewer, " (%" PetscInt_FMT ", %g) ", a->j[k], (double)a->a[k]));
422: #endif
423: }
424: PetscCall(PetscViewerASCIIPrintf(viewer, "\n"));
425: }
427: } else { /* for non-factored matrix */
428: for (i = 0; i < a->mbs; i++) { /* for row block i */
429: for (j = 0; j < bs; j++) { /* for row bs*i + j */
430: PetscCall(PetscViewerASCIIPrintf(viewer, "row %" PetscInt_FMT ":", i * bs + j));
431: for (k = a->i[i]; k < a->i[i + 1]; k++) { /* for column block */
432: for (l = 0; l < bs; l++) { /* for column */
433: #if defined(PETSC_USE_COMPLEX)
434: if (PetscImaginaryPart(a->a[bs2 * k + l * bs + j]) > 0.0) {
435: PetscCall(PetscViewerASCIIPrintf(viewer, " (%" PetscInt_FMT ", %g + %g i) ", bs * a->j[k] + l, (double)PetscRealPart(a->a[bs2 * k + l * bs + j]), (double)PetscImaginaryPart(a->a[bs2 * k + l * bs + j])));
436: } else if (PetscImaginaryPart(a->a[bs2 * k + l * bs + j]) < 0.0) {
437: PetscCall(PetscViewerASCIIPrintf(viewer, " (%" PetscInt_FMT ", %g - %g i) ", bs * a->j[k] + l, (double)PetscRealPart(a->a[bs2 * k + l * bs + j]), -(double)PetscImaginaryPart(a->a[bs2 * k + l * bs + j])));
438: } else {
439: PetscCall(PetscViewerASCIIPrintf(viewer, " (%" PetscInt_FMT ", %g) ", bs * a->j[k] + l, (double)PetscRealPart(a->a[bs2 * k + l * bs + j])));
440: }
441: #else
442: PetscCall(PetscViewerASCIIPrintf(viewer, " (%" PetscInt_FMT ", %g) ", bs * a->j[k] + l, (double)a->a[bs2 * k + l * bs + j]));
443: #endif
444: }
445: }
446: PetscCall(PetscViewerASCIIPrintf(viewer, "\n"));
447: }
448: }
449: }
450: PetscCall(PetscViewerASCIIUseTabs(viewer, PETSC_TRUE));
451: }
452: PetscCall(PetscViewerFlush(viewer));
453: PetscFunctionReturn(PETSC_SUCCESS);
454: }
456: #include <petscdraw.h>
457: static PetscErrorCode MatView_SeqSBAIJ_Draw_Zoom(PetscDraw draw, void *Aa)
458: {
459: Mat A = (Mat)Aa;
460: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ *)A->data;
461: PetscInt row, i, j, k, l, mbs = a->mbs, bs = A->rmap->bs, bs2 = a->bs2;
462: PetscReal xl, yl, xr, yr, x_l, x_r, y_l, y_r;
463: MatScalar *aa;
464: PetscViewer viewer;
465: int color;
467: PetscFunctionBegin;
468: PetscCall(PetscObjectQuery((PetscObject)A, "Zoomviewer", (PetscObject *)&viewer));
469: PetscCall(PetscDrawGetCoordinates(draw, &xl, &yl, &xr, &yr));
471: /* loop over matrix elements drawing boxes */
473: PetscDrawCollectiveBegin(draw);
474: PetscCall(PetscDrawString(draw, .3 * (xl + xr), .3 * (yl + yr), PETSC_DRAW_BLACK, "symmetric"));
475: /* Blue for negative, Cyan for zero and Red for positive */
476: color = PETSC_DRAW_BLUE;
477: for (i = 0, row = 0; i < mbs; i++, row += bs) {
478: for (j = a->i[i]; j < a->i[i + 1]; j++) {
479: y_l = A->rmap->N - row - 1.0;
480: y_r = y_l + 1.0;
481: x_l = a->j[j] * bs;
482: x_r = x_l + 1.0;
483: aa = a->a + j * bs2;
484: for (k = 0; k < bs; k++) {
485: for (l = 0; l < bs; l++) {
486: if (PetscRealPart(*aa++) >= 0.) continue;
487: PetscCall(PetscDrawRectangle(draw, x_l + k, y_l - l, x_r + k, y_r - l, color, color, color, color));
488: }
489: }
490: }
491: }
492: color = PETSC_DRAW_CYAN;
493: for (i = 0, row = 0; i < mbs; i++, row += bs) {
494: for (j = a->i[i]; j < a->i[i + 1]; j++) {
495: y_l = A->rmap->N - row - 1.0;
496: y_r = y_l + 1.0;
497: x_l = a->j[j] * bs;
498: x_r = x_l + 1.0;
499: aa = a->a + j * bs2;
500: for (k = 0; k < bs; k++) {
501: for (l = 0; l < bs; l++) {
502: if (PetscRealPart(*aa++) != 0.) continue;
503: PetscCall(PetscDrawRectangle(draw, x_l + k, y_l - l, x_r + k, y_r - l, color, color, color, color));
504: }
505: }
506: }
507: }
508: color = PETSC_DRAW_RED;
509: for (i = 0, row = 0; i < mbs; i++, row += bs) {
510: for (j = a->i[i]; j < a->i[i + 1]; j++) {
511: y_l = A->rmap->N - row - 1.0;
512: y_r = y_l + 1.0;
513: x_l = a->j[j] * bs;
514: x_r = x_l + 1.0;
515: aa = a->a + j * bs2;
516: for (k = 0; k < bs; k++) {
517: for (l = 0; l < bs; l++) {
518: if (PetscRealPart(*aa++) <= 0.) continue;
519: PetscCall(PetscDrawRectangle(draw, x_l + k, y_l - l, x_r + k, y_r - l, color, color, color, color));
520: }
521: }
522: }
523: }
524: PetscDrawCollectiveEnd(draw);
525: PetscFunctionReturn(PETSC_SUCCESS);
526: }
528: static PetscErrorCode MatView_SeqSBAIJ_Draw(Mat A, PetscViewer viewer)
529: {
530: PetscReal xl, yl, xr, yr, w, h;
531: PetscDraw draw;
532: PetscBool isnull;
534: PetscFunctionBegin;
535: PetscCall(PetscViewerDrawGetDraw(viewer, 0, &draw));
536: PetscCall(PetscDrawIsNull(draw, &isnull));
537: if (isnull) PetscFunctionReturn(PETSC_SUCCESS);
539: xr = A->rmap->N;
540: yr = A->rmap->N;
541: h = yr / 10.0;
542: w = xr / 10.0;
543: xr += w;
544: yr += h;
545: xl = -w;
546: yl = -h;
547: PetscCall(PetscDrawSetCoordinates(draw, xl, yl, xr, yr));
548: PetscCall(PetscObjectCompose((PetscObject)A, "Zoomviewer", (PetscObject)viewer));
549: PetscCall(PetscDrawZoom(draw, MatView_SeqSBAIJ_Draw_Zoom, A));
550: PetscCall(PetscObjectCompose((PetscObject)A, "Zoomviewer", NULL));
551: PetscCall(PetscDrawSave(draw));
552: PetscFunctionReturn(PETSC_SUCCESS);
553: }
555: /* Used for both MPIBAIJ and MPISBAIJ matrices */
556: #define MatView_SeqSBAIJ_Binary MatView_SeqBAIJ_Binary
558: PetscErrorCode MatView_SeqSBAIJ(Mat A, PetscViewer viewer)
559: {
560: PetscBool iascii, isbinary, isdraw;
562: PetscFunctionBegin;
563: PetscCall(PetscObjectTypeCompare((PetscObject)viewer, PETSCVIEWERASCII, &iascii));
564: PetscCall(PetscObjectTypeCompare((PetscObject)viewer, PETSCVIEWERBINARY, &isbinary));
565: PetscCall(PetscObjectTypeCompare((PetscObject)viewer, PETSCVIEWERDRAW, &isdraw));
566: if (iascii) {
567: PetscCall(MatView_SeqSBAIJ_ASCII(A, viewer));
568: } else if (isbinary) {
569: PetscCall(MatView_SeqSBAIJ_Binary(A, viewer));
570: } else if (isdraw) {
571: PetscCall(MatView_SeqSBAIJ_Draw(A, viewer));
572: } else {
573: Mat B;
574: const char *matname;
575: PetscCall(MatConvert(A, MATSEQAIJ, MAT_INITIAL_MATRIX, &B));
576: if (((PetscObject)A)->name) PetscCall(PetscObjectGetName((PetscObject)A, &matname));
577: if (((PetscObject)A)->name) PetscCall(PetscObjectSetName((PetscObject)B, matname));
578: PetscCall(MatView(B, viewer));
579: PetscCall(MatDestroy(&B));
580: }
581: PetscFunctionReturn(PETSC_SUCCESS);
582: }
584: PetscErrorCode MatGetValues_SeqSBAIJ(Mat A, PetscInt m, const PetscInt im[], PetscInt n, const PetscInt in[], PetscScalar v[])
585: {
586: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ *)A->data;
587: PetscInt *rp, k, low, high, t, row, nrow, i, col, l, *aj = a->j;
588: PetscInt *ai = a->i, *ailen = a->ilen;
589: PetscInt brow, bcol, ridx, cidx, bs = A->rmap->bs, bs2 = a->bs2;
590: MatScalar *ap, *aa = a->a;
592: PetscFunctionBegin;
593: for (k = 0; k < m; k++) { /* loop over rows */
594: row = im[k];
595: brow = row / bs;
596: if (row < 0) {
597: v += n;
598: continue;
599: } /* negative row */
600: PetscCheck(row < A->rmap->N, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Row too large: row %" PetscInt_FMT " max %" PetscInt_FMT, row, A->rmap->N - 1);
601: rp = aj + ai[brow];
602: ap = aa + bs2 * ai[brow];
603: nrow = ailen[brow];
604: for (l = 0; l < n; l++) { /* loop over columns */
605: if (in[l] < 0) {
606: v++;
607: continue;
608: } /* negative column */
609: PetscCheck(in[l] < A->cmap->n, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Column too large: col %" PetscInt_FMT " max %" PetscInt_FMT, in[l], A->cmap->n - 1);
610: col = in[l];
611: bcol = col / bs;
612: cidx = col % bs;
613: ridx = row % bs;
614: high = nrow;
615: low = 0; /* assume unsorted */
616: while (high - low > 5) {
617: t = (low + high) / 2;
618: if (rp[t] > bcol) high = t;
619: else low = t;
620: }
621: for (i = low; i < high; i++) {
622: if (rp[i] > bcol) break;
623: if (rp[i] == bcol) {
624: *v++ = ap[bs2 * i + bs * cidx + ridx];
625: goto finished;
626: }
627: }
628: *v++ = 0.0;
629: finished:;
630: }
631: }
632: PetscFunctionReturn(PETSC_SUCCESS);
633: }
635: static PetscErrorCode MatPermute_SeqSBAIJ(Mat A, IS rowp, IS colp, Mat *B)
636: {
637: Mat C;
638: PetscBool flg = (PetscBool)(rowp == colp);
640: PetscFunctionBegin;
641: PetscCall(MatConvert(A, MATSEQBAIJ, MAT_INITIAL_MATRIX, &C));
642: PetscCall(MatPermute(C, rowp, colp, B));
643: PetscCall(MatDestroy(&C));
644: if (!flg) PetscCall(ISEqual(rowp, colp, &flg));
645: if (flg) PetscCall(MatConvert(*B, MATSEQSBAIJ, MAT_INPLACE_MATRIX, B));
646: PetscFunctionReturn(PETSC_SUCCESS);
647: }
649: PetscErrorCode MatSetValuesBlocked_SeqSBAIJ(Mat A, PetscInt m, const PetscInt im[], PetscInt n, const PetscInt in[], const PetscScalar v[], InsertMode is)
650: {
651: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ *)A->data;
652: PetscInt *rp, k, low, high, t, ii, jj, row, nrow, i, col, l, rmax, N, lastcol = -1;
653: PetscInt *imax = a->imax, *ai = a->i, *ailen = a->ilen;
654: PetscInt *aj = a->j, nonew = a->nonew, bs2 = a->bs2, bs = A->rmap->bs, stepval;
655: PetscBool roworiented = a->roworiented;
656: const PetscScalar *value = v;
657: MatScalar *ap, *aa = a->a, *bap;
659: PetscFunctionBegin;
660: if (roworiented) stepval = (n - 1) * bs;
661: else stepval = (m - 1) * bs;
662: for (k = 0; k < m; k++) { /* loop over added rows */
663: row = im[k];
664: if (row < 0) continue;
665: PetscCheck(row < a->mbs, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Block index row too large %" PetscInt_FMT " max %" PetscInt_FMT, row, a->mbs - 1);
666: rp = aj + ai[row];
667: ap = aa + bs2 * ai[row];
668: rmax = imax[row];
669: nrow = ailen[row];
670: low = 0;
671: high = nrow;
672: for (l = 0; l < n; l++) { /* loop over added columns */
673: if (in[l] < 0) continue;
674: col = in[l];
675: PetscCheck(col < a->nbs, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Block index column too large %" PetscInt_FMT " max %" PetscInt_FMT, col, a->nbs - 1);
676: if (col < row) {
677: if (a->ignore_ltriangular) continue; /* ignore lower triangular block */
678: 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)");
679: }
680: if (roworiented) value = v + k * (stepval + bs) * bs + l * bs;
681: else value = v + l * (stepval + bs) * bs + k * bs;
683: if (col <= lastcol) low = 0;
684: else high = nrow;
686: lastcol = col;
687: while (high - low > 7) {
688: t = (low + high) / 2;
689: if (rp[t] > col) high = t;
690: else low = t;
691: }
692: for (i = low; i < high; i++) {
693: if (rp[i] > col) break;
694: if (rp[i] == col) {
695: bap = ap + bs2 * i;
696: if (roworiented) {
697: if (is == ADD_VALUES) {
698: for (ii = 0; ii < bs; ii++, value += stepval) {
699: for (jj = ii; jj < bs2; jj += bs) bap[jj] += *value++;
700: }
701: } else {
702: for (ii = 0; ii < bs; ii++, value += stepval) {
703: for (jj = ii; jj < bs2; jj += bs) bap[jj] = *value++;
704: }
705: }
706: } else {
707: if (is == ADD_VALUES) {
708: for (ii = 0; ii < bs; ii++, value += stepval) {
709: for (jj = 0; jj < bs; jj++) *bap++ += *value++;
710: }
711: } else {
712: for (ii = 0; ii < bs; ii++, value += stepval) {
713: for (jj = 0; jj < bs; jj++) *bap++ = *value++;
714: }
715: }
716: }
717: goto noinsert2;
718: }
719: }
720: if (nonew == 1) goto noinsert2;
721: PetscCheck(nonew != -1, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Inserting a new block index nonzero block (%" PetscInt_FMT ", %" PetscInt_FMT ") in the matrix", row, col);
722: MatSeqXAIJReallocateAIJ(A, a->mbs, bs2, nrow, row, col, rmax, aa, ai, aj, rp, ap, imax, nonew, MatScalar);
723: N = nrow++ - 1;
724: high++;
725: /* shift up all the later entries in this row */
726: PetscCall(PetscArraymove(rp + i + 1, rp + i, N - i + 1));
727: PetscCall(PetscArraymove(ap + bs2 * (i + 1), ap + bs2 * i, bs2 * (N - i + 1)));
728: PetscCall(PetscArrayzero(ap + bs2 * i, bs2));
729: rp[i] = col;
730: bap = ap + bs2 * i;
731: if (roworiented) {
732: for (ii = 0; ii < bs; ii++, value += stepval) {
733: for (jj = ii; jj < bs2; jj += bs) bap[jj] = *value++;
734: }
735: } else {
736: for (ii = 0; ii < bs; ii++, value += stepval) {
737: for (jj = 0; jj < bs; jj++) *bap++ = *value++;
738: }
739: }
740: noinsert2:;
741: low = i;
742: }
743: ailen[row] = nrow;
744: }
745: PetscFunctionReturn(PETSC_SUCCESS);
746: }
748: static PetscErrorCode MatAssemblyEnd_SeqSBAIJ(Mat A, MatAssemblyType mode)
749: {
750: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ *)A->data;
751: PetscInt fshift = 0, i, *ai = a->i, *aj = a->j, *imax = a->imax;
752: PetscInt m = A->rmap->N, *ip, N, *ailen = a->ilen;
753: PetscInt mbs = a->mbs, bs2 = a->bs2, rmax = 0;
754: MatScalar *aa = a->a, *ap;
756: PetscFunctionBegin;
757: if (mode == MAT_FLUSH_ASSEMBLY || (A->was_assembled && A->ass_nonzerostate == A->nonzerostate)) PetscFunctionReturn(PETSC_SUCCESS);
759: if (m) rmax = ailen[0];
760: for (i = 1; i < mbs; i++) {
761: /* move each row back by the amount of empty slots (fshift) before it*/
762: fshift += imax[i - 1] - ailen[i - 1];
763: rmax = PetscMax(rmax, ailen[i]);
764: if (fshift) {
765: ip = aj + ai[i];
766: ap = aa + bs2 * ai[i];
767: N = ailen[i];
768: PetscCall(PetscArraymove(ip - fshift, ip, N));
769: PetscCall(PetscArraymove(ap - bs2 * fshift, ap, bs2 * N));
770: }
771: ai[i] = ai[i - 1] + ailen[i - 1];
772: }
773: if (mbs) {
774: fshift += imax[mbs - 1] - ailen[mbs - 1];
775: ai[mbs] = ai[mbs - 1] + ailen[mbs - 1];
776: }
777: /* reset ilen and imax for each row */
778: for (i = 0; i < mbs; i++) ailen[i] = imax[i] = ai[i + 1] - ai[i];
779: a->nz = ai[mbs];
781: /* diagonals may have moved, reset it */
782: if (a->diag) PetscCall(PetscArraycpy(a->diag, ai, mbs));
783: PetscCheck(!fshift || a->nounused != -1, PETSC_COMM_SELF, PETSC_ERR_PLIB, "Unused space detected in matrix: %" PetscInt_FMT " X %" PetscInt_FMT " block size %" PetscInt_FMT ", %" PetscInt_FMT " unneeded", m, A->cmap->n, A->rmap->bs, fshift * bs2);
785: PetscCall(PetscInfo(A, "Matrix size: %" PetscInt_FMT " X %" PetscInt_FMT ", block size %" PetscInt_FMT "; storage space: %" PetscInt_FMT " unneeded, %" PetscInt_FMT " used\n", m, A->rmap->N, A->rmap->bs, fshift * bs2, a->nz * bs2));
786: PetscCall(PetscInfo(A, "Number of mallocs during MatSetValues is %" PetscInt_FMT "\n", a->reallocs));
787: PetscCall(PetscInfo(A, "Most nonzeros blocks in any row is %" PetscInt_FMT "\n", rmax));
789: A->info.mallocs += a->reallocs;
790: a->reallocs = 0;
791: A->info.nz_unneeded = (PetscReal)fshift * bs2;
792: a->idiagvalid = PETSC_FALSE;
793: a->rmax = rmax;
795: if (A->cmap->n < 65536 && A->cmap->bs == 1) {
796: if (a->jshort && a->free_jshort) {
797: /* when matrix data structure is changed, previous jshort must be replaced */
798: PetscCall(PetscFree(a->jshort));
799: }
800: PetscCall(PetscMalloc1(a->i[A->rmap->n], &a->jshort));
801: for (i = 0; i < a->i[A->rmap->n]; i++) a->jshort[i] = (short)a->j[i];
802: A->ops->mult = MatMult_SeqSBAIJ_1_ushort;
803: A->ops->sor = MatSOR_SeqSBAIJ_ushort;
804: a->free_jshort = PETSC_TRUE;
805: }
806: PetscFunctionReturn(PETSC_SUCCESS);
807: }
809: /* Only add/insert a(i,j) with i<=j (blocks).
810: Any a(i,j) with i>j input by user is ignored.
811: */
813: PetscErrorCode MatSetValues_SeqSBAIJ(Mat A, PetscInt m, const PetscInt im[], PetscInt n, const PetscInt in[], const PetscScalar v[], InsertMode is)
814: {
815: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ *)A->data;
816: PetscInt *rp, k, low, high, t, ii, row, nrow, i, col, l, rmax, N, lastcol = -1;
817: PetscInt *imax = a->imax, *ai = a->i, *ailen = a->ilen, roworiented = a->roworiented;
818: PetscInt *aj = a->j, nonew = a->nonew, bs = A->rmap->bs, brow, bcol;
819: PetscInt ridx, cidx, bs2 = a->bs2;
820: MatScalar *ap, value, *aa = a->a, *bap;
822: PetscFunctionBegin;
823: for (k = 0; k < m; k++) { /* loop over added rows */
824: row = im[k]; /* row number */
825: brow = row / bs; /* block row number */
826: if (row < 0) continue;
827: PetscCheck(row < A->rmap->N, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Row too large: row %" PetscInt_FMT " max %" PetscInt_FMT, row, A->rmap->N - 1);
828: rp = aj + ai[brow]; /*ptr to beginning of column value of the row block*/
829: ap = aa + bs2 * ai[brow]; /*ptr to beginning of element value of the row block*/
830: rmax = imax[brow]; /* maximum space allocated for this row */
831: nrow = ailen[brow]; /* actual length of this row */
832: low = 0;
833: high = nrow;
834: for (l = 0; l < n; l++) { /* loop over added columns */
835: if (in[l] < 0) continue;
836: PetscCheck(in[l] < A->cmap->N, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Column too large: col %" PetscInt_FMT " max %" PetscInt_FMT, in[l], A->cmap->N - 1);
837: col = in[l];
838: bcol = col / bs; /* block col number */
840: if (brow > bcol) {
841: if (a->ignore_ltriangular) continue; /* ignore lower triangular values */
842: 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)");
843: }
845: ridx = row % bs;
846: cidx = col % bs; /*row and col index inside the block */
847: if ((brow == bcol && ridx <= cidx) || (brow < bcol)) {
848: /* element value a(k,l) */
849: if (roworiented) value = v[l + k * n];
850: else value = v[k + l * m];
852: /* move pointer bap to a(k,l) quickly and add/insert value */
853: if (col <= lastcol) low = 0;
854: else high = nrow;
856: lastcol = col;
857: while (high - low > 7) {
858: t = (low + high) / 2;
859: if (rp[t] > bcol) high = t;
860: else low = t;
861: }
862: for (i = low; i < high; i++) {
863: if (rp[i] > bcol) break;
864: if (rp[i] == bcol) {
865: bap = ap + bs2 * i + bs * cidx + ridx;
866: if (is == ADD_VALUES) *bap += value;
867: else *bap = value;
868: /* for diag block, add/insert its symmetric element a(cidx,ridx) */
869: if (brow == bcol && ridx < cidx) {
870: bap = ap + bs2 * i + bs * ridx + cidx;
871: if (is == ADD_VALUES) *bap += value;
872: else *bap = value;
873: }
874: goto noinsert1;
875: }
876: }
878: if (nonew == 1) goto noinsert1;
879: PetscCheck(nonew != -1, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Inserting a new nonzero (%" PetscInt_FMT ", %" PetscInt_FMT ") in the matrix", row, col);
880: MatSeqXAIJReallocateAIJ(A, a->mbs, bs2, nrow, brow, bcol, rmax, aa, ai, aj, rp, ap, imax, nonew, MatScalar);
882: N = nrow++ - 1;
883: high++;
884: /* shift up all the later entries in this row */
885: PetscCall(PetscArraymove(rp + i + 1, rp + i, N - i + 1));
886: PetscCall(PetscArraymove(ap + bs2 * (i + 1), ap + bs2 * i, bs2 * (N - i + 1)));
887: PetscCall(PetscArrayzero(ap + bs2 * i, bs2));
888: rp[i] = bcol;
889: ap[bs2 * i + bs * cidx + ridx] = value;
890: /* for diag block, add/insert its symmetric element a(cidx,ridx) */
891: if (brow == bcol && ridx < cidx) ap[bs2 * i + bs * ridx + cidx] = value;
892: noinsert1:;
893: low = i;
894: }
895: } /* end of loop over added columns */
896: ailen[brow] = nrow;
897: } /* end of loop over added rows */
898: PetscFunctionReturn(PETSC_SUCCESS);
899: }
901: static PetscErrorCode MatICCFactor_SeqSBAIJ(Mat inA, IS row, const MatFactorInfo *info)
902: {
903: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ *)inA->data;
904: Mat outA;
905: PetscBool row_identity;
907: PetscFunctionBegin;
908: PetscCheck(info->levels == 0, PETSC_COMM_SELF, PETSC_ERR_SUP, "Only levels=0 is supported for in-place icc");
909: PetscCall(ISIdentity(row, &row_identity));
910: PetscCheck(row_identity, PETSC_COMM_SELF, PETSC_ERR_SUP, "Matrix reordering is not supported");
911: PetscCheck(inA->rmap->bs == 1, PETSC_COMM_SELF, PETSC_ERR_SUP, "Matrix block size %" PetscInt_FMT " is not supported", inA->rmap->bs); /* Need to replace MatCholeskyFactorSymbolic_SeqSBAIJ_MSR()! */
913: outA = inA;
914: inA->factortype = MAT_FACTOR_ICC;
915: PetscCall(PetscFree(inA->solvertype));
916: PetscCall(PetscStrallocpy(MATSOLVERPETSC, &inA->solvertype));
918: PetscCall(MatMarkDiagonal_SeqSBAIJ(inA));
919: PetscCall(MatSeqSBAIJSetNumericFactorization_inplace(inA, row_identity));
921: PetscCall(PetscObjectReference((PetscObject)row));
922: PetscCall(ISDestroy(&a->row));
923: a->row = row;
924: PetscCall(PetscObjectReference((PetscObject)row));
925: PetscCall(ISDestroy(&a->col));
926: a->col = row;
928: /* Create the invert permutation so that it can be used in MatCholeskyFactorNumeric() */
929: if (a->icol) PetscCall(ISInvertPermutation(row, PETSC_DECIDE, &a->icol));
931: if (!a->solve_work) PetscCall(PetscMalloc1(inA->rmap->N + inA->rmap->bs, &a->solve_work));
933: PetscCall(MatCholeskyFactorNumeric(outA, inA, info));
934: PetscFunctionReturn(PETSC_SUCCESS);
935: }
937: static PetscErrorCode MatSeqSBAIJSetColumnIndices_SeqSBAIJ(Mat mat, PetscInt *indices)
938: {
939: Mat_SeqSBAIJ *baij = (Mat_SeqSBAIJ *)mat->data;
940: PetscInt i, nz, n;
942: PetscFunctionBegin;
943: nz = baij->maxnz;
944: n = mat->cmap->n;
945: for (i = 0; i < nz; i++) baij->j[i] = indices[i];
947: baij->nz = nz;
948: for (i = 0; i < n; i++) baij->ilen[i] = baij->imax[i];
950: PetscCall(MatSetOption(mat, MAT_NEW_NONZERO_LOCATION_ERR, PETSC_TRUE));
951: PetscFunctionReturn(PETSC_SUCCESS);
952: }
954: /*@
955: MatSeqSBAIJSetColumnIndices - Set the column indices for all the rows
956: in a `MATSEQSBAIJ` matrix.
958: Input Parameters:
959: + mat - the `MATSEQSBAIJ` matrix
960: - indices - the column indices
962: Level: advanced
964: Notes:
965: This can be called if you have precomputed the nonzero structure of the
966: matrix and want to provide it to the matrix object to improve the performance
967: of the `MatSetValues()` operation.
969: You MUST have set the correct numbers of nonzeros per row in the call to
970: `MatCreateSeqSBAIJ()`, and the columns indices MUST be sorted.
972: MUST be called before any calls to `MatSetValues()`
974: .seealso: [](ch_matrices), `Mat`, `MATSEQSBAIJ`, `MatCreateSeqSBAIJ`
975: @*/
976: PetscErrorCode MatSeqSBAIJSetColumnIndices(Mat mat, PetscInt *indices)
977: {
978: PetscFunctionBegin;
980: PetscAssertPointer(indices, 2);
981: PetscUseMethod(mat, "MatSeqSBAIJSetColumnIndices_C", (Mat, PetscInt *), (mat, indices));
982: PetscFunctionReturn(PETSC_SUCCESS);
983: }
985: static PetscErrorCode MatCopy_SeqSBAIJ(Mat A, Mat B, MatStructure str)
986: {
987: PetscBool isbaij;
989: PetscFunctionBegin;
990: PetscCall(PetscObjectTypeCompareAny((PetscObject)B, &isbaij, MATSEQSBAIJ, MATMPISBAIJ, ""));
991: PetscCheck(isbaij, PetscObjectComm((PetscObject)B), PETSC_ERR_SUP, "Not for matrix type %s", ((PetscObject)B)->type_name);
992: /* If the two matrices have the same copy implementation and nonzero pattern, use fast copy. */
993: if (str == SAME_NONZERO_PATTERN && A->ops->copy == B->ops->copy) {
994: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ *)A->data;
995: Mat_SeqSBAIJ *b = (Mat_SeqSBAIJ *)B->data;
997: PetscCheck(a->i[a->mbs] == b->i[b->mbs], PETSC_COMM_SELF, PETSC_ERR_ARG_INCOMP, "Number of nonzeros in two matrices are different");
998: PetscCheck(a->mbs == b->mbs, PETSC_COMM_SELF, PETSC_ERR_ARG_INCOMP, "Number of rows in two matrices are different");
999: PetscCheck(a->bs2 == b->bs2, PETSC_COMM_SELF, PETSC_ERR_ARG_INCOMP, "Different block size");
1000: PetscCall(PetscArraycpy(b->a, a->a, a->bs2 * a->i[a->mbs]));
1001: PetscCall(PetscObjectStateIncrease((PetscObject)B));
1002: } else {
1003: PetscCall(MatGetRowUpperTriangular(A));
1004: PetscCall(MatCopy_Basic(A, B, str));
1005: PetscCall(MatRestoreRowUpperTriangular(A));
1006: }
1007: PetscFunctionReturn(PETSC_SUCCESS);
1008: }
1010: static PetscErrorCode MatSeqSBAIJGetArray_SeqSBAIJ(Mat A, PetscScalar *array[])
1011: {
1012: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ *)A->data;
1014: PetscFunctionBegin;
1015: *array = a->a;
1016: PetscFunctionReturn(PETSC_SUCCESS);
1017: }
1019: static PetscErrorCode MatSeqSBAIJRestoreArray_SeqSBAIJ(Mat A, PetscScalar *array[])
1020: {
1021: PetscFunctionBegin;
1022: *array = NULL;
1023: PetscFunctionReturn(PETSC_SUCCESS);
1024: }
1026: PetscErrorCode MatAXPYGetPreallocation_SeqSBAIJ(Mat Y, Mat X, PetscInt *nnz)
1027: {
1028: PetscInt bs = Y->rmap->bs, mbs = Y->rmap->N / bs;
1029: Mat_SeqSBAIJ *x = (Mat_SeqSBAIJ *)X->data;
1030: Mat_SeqSBAIJ *y = (Mat_SeqSBAIJ *)Y->data;
1032: PetscFunctionBegin;
1033: /* Set the number of nonzeros in the new matrix */
1034: PetscCall(MatAXPYGetPreallocation_SeqX_private(mbs, x->i, x->j, y->i, y->j, nnz));
1035: PetscFunctionReturn(PETSC_SUCCESS);
1036: }
1038: static PetscErrorCode MatAXPY_SeqSBAIJ(Mat Y, PetscScalar a, Mat X, MatStructure str)
1039: {
1040: Mat_SeqSBAIJ *x = (Mat_SeqSBAIJ *)X->data, *y = (Mat_SeqSBAIJ *)Y->data;
1041: PetscInt bs = Y->rmap->bs, bs2 = bs * bs;
1042: PetscBLASInt one = 1;
1044: PetscFunctionBegin;
1045: if (str == UNKNOWN_NONZERO_PATTERN || (PetscDefined(USE_DEBUG) && str == SAME_NONZERO_PATTERN)) {
1046: PetscBool e = x->nz == y->nz && x->mbs == y->mbs ? PETSC_TRUE : PETSC_FALSE;
1047: if (e) {
1048: PetscCall(PetscArraycmp(x->i, y->i, x->mbs + 1, &e));
1049: if (e) {
1050: PetscCall(PetscArraycmp(x->j, y->j, x->i[x->mbs], &e));
1051: if (e) str = SAME_NONZERO_PATTERN;
1052: }
1053: }
1054: if (!e) PetscCheck(str != SAME_NONZERO_PATTERN, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONG, "MatStructure is not SAME_NONZERO_PATTERN");
1055: }
1056: if (str == SAME_NONZERO_PATTERN) {
1057: PetscScalar alpha = a;
1058: PetscBLASInt bnz;
1059: PetscCall(PetscBLASIntCast(x->nz * bs2, &bnz));
1060: PetscCallBLAS("BLASaxpy", BLASaxpy_(&bnz, &alpha, x->a, &one, y->a, &one));
1061: PetscCall(PetscObjectStateIncrease((PetscObject)Y));
1062: } else if (str == SUBSET_NONZERO_PATTERN) { /* nonzeros of X is a subset of Y's */
1063: PetscCall(MatSetOption(X, MAT_GETROW_UPPERTRIANGULAR, PETSC_TRUE));
1064: PetscCall(MatAXPY_Basic(Y, a, X, str));
1065: PetscCall(MatSetOption(X, MAT_GETROW_UPPERTRIANGULAR, PETSC_FALSE));
1066: } else {
1067: Mat B;
1068: PetscInt *nnz;
1069: PetscCheck(bs == X->rmap->bs, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Matrices must have same block size");
1070: PetscCall(MatGetRowUpperTriangular(X));
1071: PetscCall(MatGetRowUpperTriangular(Y));
1072: PetscCall(PetscMalloc1(Y->rmap->N, &nnz));
1073: PetscCall(MatCreate(PetscObjectComm((PetscObject)Y), &B));
1074: PetscCall(PetscObjectSetName((PetscObject)B, ((PetscObject)Y)->name));
1075: PetscCall(MatSetSizes(B, Y->rmap->n, Y->cmap->n, Y->rmap->N, Y->cmap->N));
1076: PetscCall(MatSetBlockSizesFromMats(B, Y, Y));
1077: PetscCall(MatSetType(B, ((PetscObject)Y)->type_name));
1078: PetscCall(MatAXPYGetPreallocation_SeqSBAIJ(Y, X, nnz));
1079: PetscCall(MatSeqSBAIJSetPreallocation(B, bs, 0, nnz));
1081: PetscCall(MatAXPY_BasicWithPreallocation(B, Y, a, X, str));
1083: PetscCall(MatHeaderMerge(Y, &B));
1084: PetscCall(PetscFree(nnz));
1085: PetscCall(MatRestoreRowUpperTriangular(X));
1086: PetscCall(MatRestoreRowUpperTriangular(Y));
1087: }
1088: PetscFunctionReturn(PETSC_SUCCESS);
1089: }
1091: static PetscErrorCode MatIsStructurallySymmetric_SeqSBAIJ(Mat A, PetscBool *flg)
1092: {
1093: PetscFunctionBegin;
1094: *flg = PETSC_TRUE;
1095: PetscFunctionReturn(PETSC_SUCCESS);
1096: }
1098: static PetscErrorCode MatConjugate_SeqSBAIJ(Mat A)
1099: {
1100: #if defined(PETSC_USE_COMPLEX)
1101: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ *)A->data;
1102: PetscInt i, nz = a->bs2 * a->i[a->mbs];
1103: MatScalar *aa = a->a;
1105: PetscFunctionBegin;
1106: for (i = 0; i < nz; i++) aa[i] = PetscConj(aa[i]);
1107: #else
1108: PetscFunctionBegin;
1109: #endif
1110: PetscFunctionReturn(PETSC_SUCCESS);
1111: }
1113: static PetscErrorCode MatRealPart_SeqSBAIJ(Mat A)
1114: {
1115: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ *)A->data;
1116: PetscInt i, nz = a->bs2 * a->i[a->mbs];
1117: MatScalar *aa = a->a;
1119: PetscFunctionBegin;
1120: for (i = 0; i < nz; i++) aa[i] = PetscRealPart(aa[i]);
1121: PetscFunctionReturn(PETSC_SUCCESS);
1122: }
1124: static PetscErrorCode MatImaginaryPart_SeqSBAIJ(Mat A)
1125: {
1126: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ *)A->data;
1127: PetscInt i, nz = a->bs2 * a->i[a->mbs];
1128: MatScalar *aa = a->a;
1130: PetscFunctionBegin;
1131: for (i = 0; i < nz; i++) aa[i] = PetscImaginaryPart(aa[i]);
1132: PetscFunctionReturn(PETSC_SUCCESS);
1133: }
1135: static PetscErrorCode MatZeroRowsColumns_SeqSBAIJ(Mat A, PetscInt is_n, const PetscInt is_idx[], PetscScalar diag, Vec x, Vec b)
1136: {
1137: Mat_SeqSBAIJ *baij = (Mat_SeqSBAIJ *)A->data;
1138: PetscInt i, j, k, count;
1139: PetscInt bs = A->rmap->bs, bs2 = baij->bs2, row, col;
1140: PetscScalar zero = 0.0;
1141: MatScalar *aa;
1142: const PetscScalar *xx;
1143: PetscScalar *bb;
1144: PetscBool *zeroed, vecs = PETSC_FALSE;
1146: PetscFunctionBegin;
1147: /* fix right-hand side if needed */
1148: if (x && b) {
1149: PetscCall(VecGetArrayRead(x, &xx));
1150: PetscCall(VecGetArray(b, &bb));
1151: vecs = PETSC_TRUE;
1152: }
1154: /* zero the columns */
1155: PetscCall(PetscCalloc1(A->rmap->n, &zeroed));
1156: for (i = 0; i < is_n; i++) {
1157: PetscCheck(is_idx[i] >= 0 && is_idx[i] < A->rmap->N, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "row %" PetscInt_FMT " out of range", is_idx[i]);
1158: zeroed[is_idx[i]] = PETSC_TRUE;
1159: }
1160: if (vecs) {
1161: for (i = 0; i < A->rmap->N; i++) {
1162: row = i / bs;
1163: for (j = baij->i[row]; j < baij->i[row + 1]; j++) {
1164: for (k = 0; k < bs; k++) {
1165: col = bs * baij->j[j] + k;
1166: if (col <= i) continue;
1167: aa = baij->a + j * bs2 + (i % bs) + bs * k;
1168: if (!zeroed[i] && zeroed[col]) bb[i] -= aa[0] * xx[col];
1169: if (zeroed[i] && !zeroed[col]) bb[col] -= aa[0] * xx[i];
1170: }
1171: }
1172: }
1173: for (i = 0; i < is_n; i++) bb[is_idx[i]] = diag * xx[is_idx[i]];
1174: }
1176: for (i = 0; i < A->rmap->N; i++) {
1177: if (!zeroed[i]) {
1178: row = i / bs;
1179: for (j = baij->i[row]; j < baij->i[row + 1]; j++) {
1180: for (k = 0; k < bs; k++) {
1181: col = bs * baij->j[j] + k;
1182: if (zeroed[col]) {
1183: aa = baij->a + j * bs2 + (i % bs) + bs * k;
1184: aa[0] = 0.0;
1185: }
1186: }
1187: }
1188: }
1189: }
1190: PetscCall(PetscFree(zeroed));
1191: if (vecs) {
1192: PetscCall(VecRestoreArrayRead(x, &xx));
1193: PetscCall(VecRestoreArray(b, &bb));
1194: }
1196: /* zero the rows */
1197: for (i = 0; i < is_n; i++) {
1198: row = is_idx[i];
1199: count = (baij->i[row / bs + 1] - baij->i[row / bs]) * bs;
1200: aa = baij->a + baij->i[row / bs] * bs2 + (row % bs);
1201: for (k = 0; k < count; k++) {
1202: aa[0] = zero;
1203: aa += bs;
1204: }
1205: if (diag != 0.0) PetscUseTypeMethod(A, setvalues, 1, &row, 1, &row, &diag, INSERT_VALUES);
1206: }
1207: PetscCall(MatAssemblyEnd_SeqSBAIJ(A, MAT_FINAL_ASSEMBLY));
1208: PetscFunctionReturn(PETSC_SUCCESS);
1209: }
1211: static PetscErrorCode MatShift_SeqSBAIJ(Mat Y, PetscScalar a)
1212: {
1213: Mat_SeqSBAIJ *aij = (Mat_SeqSBAIJ *)Y->data;
1215: PetscFunctionBegin;
1216: if (!Y->preallocated || !aij->nz) PetscCall(MatSeqSBAIJSetPreallocation(Y, Y->rmap->bs, 1, NULL));
1217: PetscCall(MatShift_Basic(Y, a));
1218: PetscFunctionReturn(PETSC_SUCCESS);
1219: }
1221: PetscErrorCode MatEliminateZeros_SeqSBAIJ(Mat A, PetscBool keep)
1222: {
1223: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ *)A->data;
1224: PetscInt fshift = 0, fshift_prev = 0, i, *ai = a->i, *aj = a->j, *imax = a->imax, j, k;
1225: PetscInt m = A->rmap->N, *ailen = a->ilen;
1226: PetscInt mbs = a->mbs, bs2 = a->bs2, rmax = 0;
1227: MatScalar *aa = a->a, *ap;
1228: PetscBool zero;
1230: PetscFunctionBegin;
1231: PetscCheck(A->assembled, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Cannot eliminate zeros for unassembled matrix");
1232: if (m) rmax = ailen[0];
1233: for (i = 1; i <= mbs; i++) {
1234: for (k = ai[i - 1]; k < ai[i]; k++) {
1235: zero = PETSC_TRUE;
1236: ap = aa + bs2 * k;
1237: for (j = 0; j < bs2 && zero; j++) {
1238: if (ap[j] != 0.0) zero = PETSC_FALSE;
1239: }
1240: if (zero && (aj[k] != i - 1 || !keep)) fshift++;
1241: else {
1242: if (zero && aj[k] == i - 1) PetscCall(PetscInfo(A, "Keep the diagonal block at row %" PetscInt_FMT "\n", i - 1));
1243: aj[k - fshift] = aj[k];
1244: PetscCall(PetscArraymove(ap - bs2 * fshift, ap, bs2));
1245: }
1246: }
1247: ai[i - 1] -= fshift_prev;
1248: fshift_prev = fshift;
1249: ailen[i - 1] = imax[i - 1] = ai[i] - fshift - ai[i - 1];
1250: a->nonzerorowcnt += ((ai[i] - fshift - ai[i - 1]) > 0);
1251: rmax = PetscMax(rmax, ailen[i - 1]);
1252: }
1253: if (fshift) {
1254: if (mbs) {
1255: ai[mbs] -= fshift;
1256: a->nz = ai[mbs];
1257: }
1258: PetscCall(PetscInfo(A, "Matrix size: %" PetscInt_FMT " X %" PetscInt_FMT "; zeros eliminated: %" PetscInt_FMT "; nonzeros left: %" PetscInt_FMT "\n", m, A->cmap->n, fshift, a->nz));
1259: A->nonzerostate++;
1260: A->info.nz_unneeded += (PetscReal)fshift;
1261: a->rmax = rmax;
1262: PetscCall(MatAssemblyBegin(A, MAT_FINAL_ASSEMBLY));
1263: PetscCall(MatAssemblyEnd(A, MAT_FINAL_ASSEMBLY));
1264: }
1265: PetscFunctionReturn(PETSC_SUCCESS);
1266: }
1268: static struct _MatOps MatOps_Values = {MatSetValues_SeqSBAIJ,
1269: MatGetRow_SeqSBAIJ,
1270: MatRestoreRow_SeqSBAIJ,
1271: MatMult_SeqSBAIJ_N,
1272: /* 4*/ MatMultAdd_SeqSBAIJ_N,
1273: MatMult_SeqSBAIJ_N, /* transpose versions are same as non-transpose versions */
1274: MatMultAdd_SeqSBAIJ_N,
1275: NULL,
1276: NULL,
1277: NULL,
1278: /* 10*/ NULL,
1279: NULL,
1280: MatCholeskyFactor_SeqSBAIJ,
1281: MatSOR_SeqSBAIJ,
1282: MatTranspose_SeqSBAIJ,
1283: /* 15*/ MatGetInfo_SeqSBAIJ,
1284: MatEqual_SeqSBAIJ,
1285: MatGetDiagonal_SeqSBAIJ,
1286: MatDiagonalScale_SeqSBAIJ,
1287: MatNorm_SeqSBAIJ,
1288: /* 20*/ NULL,
1289: MatAssemblyEnd_SeqSBAIJ,
1290: MatSetOption_SeqSBAIJ,
1291: MatZeroEntries_SeqSBAIJ,
1292: /* 24*/ NULL,
1293: NULL,
1294: NULL,
1295: NULL,
1296: NULL,
1297: /* 29*/ MatSetUp_Seq_Hash,
1298: NULL,
1299: NULL,
1300: NULL,
1301: NULL,
1302: /* 34*/ MatDuplicate_SeqSBAIJ,
1303: NULL,
1304: NULL,
1305: NULL,
1306: MatICCFactor_SeqSBAIJ,
1307: /* 39*/ MatAXPY_SeqSBAIJ,
1308: MatCreateSubMatrices_SeqSBAIJ,
1309: MatIncreaseOverlap_SeqSBAIJ,
1310: MatGetValues_SeqSBAIJ,
1311: MatCopy_SeqSBAIJ,
1312: /* 44*/ NULL,
1313: MatScale_SeqSBAIJ,
1314: MatShift_SeqSBAIJ,
1315: NULL,
1316: MatZeroRowsColumns_SeqSBAIJ,
1317: /* 49*/ NULL,
1318: MatGetRowIJ_SeqSBAIJ,
1319: MatRestoreRowIJ_SeqSBAIJ,
1320: NULL,
1321: NULL,
1322: /* 54*/ NULL,
1323: NULL,
1324: NULL,
1325: MatPermute_SeqSBAIJ,
1326: MatSetValuesBlocked_SeqSBAIJ,
1327: /* 59*/ MatCreateSubMatrix_SeqSBAIJ,
1328: NULL,
1329: NULL,
1330: NULL,
1331: NULL,
1332: /* 64*/ NULL,
1333: NULL,
1334: NULL,
1335: NULL,
1336: NULL,
1337: /* 69*/ MatGetRowMaxAbs_SeqSBAIJ,
1338: NULL,
1339: MatConvert_MPISBAIJ_Basic,
1340: NULL,
1341: NULL,
1342: /* 74*/ NULL,
1343: NULL,
1344: NULL,
1345: NULL,
1346: NULL,
1347: /* 79*/ NULL,
1348: NULL,
1349: NULL,
1350: MatGetInertia_SeqSBAIJ,
1351: MatLoad_SeqSBAIJ,
1352: /* 84*/ NULL,
1353: NULL,
1354: MatIsStructurallySymmetric_SeqSBAIJ,
1355: NULL,
1356: NULL,
1357: /* 89*/ NULL,
1358: NULL,
1359: NULL,
1360: NULL,
1361: NULL,
1362: /* 94*/ NULL,
1363: NULL,
1364: NULL,
1365: NULL,
1366: NULL,
1367: /* 99*/ NULL,
1368: NULL,
1369: NULL,
1370: MatConjugate_SeqSBAIJ,
1371: NULL,
1372: /*104*/ NULL,
1373: MatRealPart_SeqSBAIJ,
1374: MatImaginaryPart_SeqSBAIJ,
1375: MatGetRowUpperTriangular_SeqSBAIJ,
1376: MatRestoreRowUpperTriangular_SeqSBAIJ,
1377: /*109*/ NULL,
1378: NULL,
1379: NULL,
1380: NULL,
1381: MatMissingDiagonal_SeqSBAIJ,
1382: /*114*/ NULL,
1383: NULL,
1384: NULL,
1385: NULL,
1386: NULL,
1387: /*119*/ NULL,
1388: NULL,
1389: NULL,
1390: NULL,
1391: NULL,
1392: /*124*/ NULL,
1393: NULL,
1394: NULL,
1395: NULL,
1396: NULL,
1397: /*129*/ NULL,
1398: NULL,
1399: NULL,
1400: NULL,
1401: NULL,
1402: /*134*/ NULL,
1403: NULL,
1404: NULL,
1405: NULL,
1406: NULL,
1407: /*139*/ MatSetBlockSizes_Default,
1408: NULL,
1409: NULL,
1410: NULL,
1411: NULL,
1412: /*144*/ MatCreateMPIMatConcatenateSeqMat_SeqSBAIJ,
1413: NULL,
1414: NULL,
1415: NULL,
1416: NULL,
1417: NULL,
1418: /*150*/ NULL,
1419: MatEliminateZeros_SeqSBAIJ,
1420: NULL,
1421: NULL,
1422: NULL,
1423: /*155*/ NULL,
1424: MatCopyHashToXAIJ_Seq_Hash};
1426: static PetscErrorCode MatStoreValues_SeqSBAIJ(Mat mat)
1427: {
1428: Mat_SeqSBAIJ *aij = (Mat_SeqSBAIJ *)mat->data;
1429: PetscInt nz = aij->i[mat->rmap->N] * mat->rmap->bs * aij->bs2;
1431: PetscFunctionBegin;
1432: PetscCheck(aij->nonew == 1, PETSC_COMM_SELF, PETSC_ERR_ORDER, "Must call MatSetOption(A,MAT_NEW_NONZERO_LOCATIONS,PETSC_FALSE);first");
1434: /* allocate space for values if not already there */
1435: if (!aij->saved_values) PetscCall(PetscMalloc1(nz + 1, &aij->saved_values));
1437: /* copy values over */
1438: PetscCall(PetscArraycpy(aij->saved_values, aij->a, nz));
1439: PetscFunctionReturn(PETSC_SUCCESS);
1440: }
1442: static PetscErrorCode MatRetrieveValues_SeqSBAIJ(Mat mat)
1443: {
1444: Mat_SeqSBAIJ *aij = (Mat_SeqSBAIJ *)mat->data;
1445: PetscInt nz = aij->i[mat->rmap->N] * mat->rmap->bs * aij->bs2;
1447: PetscFunctionBegin;
1448: PetscCheck(aij->nonew == 1, PETSC_COMM_SELF, PETSC_ERR_ORDER, "Must call MatSetOption(A,MAT_NEW_NONZERO_LOCATIONS,PETSC_FALSE);first");
1449: PetscCheck(aij->saved_values, PETSC_COMM_SELF, PETSC_ERR_ORDER, "Must call MatStoreValues(A);first");
1451: /* copy values over */
1452: PetscCall(PetscArraycpy(aij->a, aij->saved_values, nz));
1453: PetscFunctionReturn(PETSC_SUCCESS);
1454: }
1456: static PetscErrorCode MatSeqSBAIJSetPreallocation_SeqSBAIJ(Mat B, PetscInt bs, PetscInt nz, const PetscInt nnz[])
1457: {
1458: Mat_SeqSBAIJ *b = (Mat_SeqSBAIJ *)B->data;
1459: PetscInt i, mbs, nbs, bs2;
1460: PetscBool skipallocation = PETSC_FALSE, flg = PETSC_FALSE, realalloc = PETSC_FALSE;
1462: PetscFunctionBegin;
1463: if (B->hash_active) {
1464: PetscInt bs;
1465: B->ops[0] = b->cops;
1466: PetscCall(PetscHMapIJVDestroy(&b->ht));
1467: PetscCall(MatGetBlockSize(B, &bs));
1468: if (bs > 1) PetscCall(PetscHSetIJDestroy(&b->bht));
1469: PetscCall(PetscFree(b->dnz));
1470: PetscCall(PetscFree(b->bdnz));
1471: B->hash_active = PETSC_FALSE;
1472: }
1473: if (nz >= 0 || nnz) realalloc = PETSC_TRUE;
1475: PetscCall(MatSetBlockSize(B, PetscAbs(bs)));
1476: PetscCall(PetscLayoutSetUp(B->rmap));
1477: PetscCall(PetscLayoutSetUp(B->cmap));
1478: PetscCheck(B->rmap->N <= B->cmap->N, PETSC_COMM_SELF, PETSC_ERR_SUP, "SEQSBAIJ matrix cannot have more rows %" PetscInt_FMT " than columns %" PetscInt_FMT, B->rmap->N, B->cmap->N);
1479: PetscCall(PetscLayoutGetBlockSize(B->rmap, &bs));
1481: B->preallocated = PETSC_TRUE;
1483: mbs = B->rmap->N / bs;
1484: nbs = B->cmap->n / bs;
1485: bs2 = bs * bs;
1487: PetscCheck(mbs * bs == B->rmap->N && nbs * bs == B->cmap->n, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Number rows, cols must be divisible by blocksize");
1489: if (nz == MAT_SKIP_ALLOCATION) {
1490: skipallocation = PETSC_TRUE;
1491: nz = 0;
1492: }
1494: if (nz == PETSC_DEFAULT || nz == PETSC_DECIDE) nz = 3;
1495: PetscCheck(nz >= 0, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "nz cannot be less than 0: value %" PetscInt_FMT, nz);
1496: if (nnz) {
1497: for (i = 0; i < mbs; i++) {
1498: PetscCheck(nnz[i] >= 0, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "nnz cannot be less than 0: local row %" PetscInt_FMT " value %" PetscInt_FMT, i, nnz[i]);
1499: PetscCheck(nnz[i] <= nbs, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "nnz cannot be greater than block row length: local row %" PetscInt_FMT " value %" PetscInt_FMT " block rowlength %" PetscInt_FMT, i, nnz[i], nbs);
1500: }
1501: }
1503: B->ops->mult = MatMult_SeqSBAIJ_N;
1504: B->ops->multadd = MatMultAdd_SeqSBAIJ_N;
1505: B->ops->multtranspose = MatMult_SeqSBAIJ_N;
1506: B->ops->multtransposeadd = MatMultAdd_SeqSBAIJ_N;
1508: PetscCall(PetscOptionsGetBool(((PetscObject)B)->options, ((PetscObject)B)->prefix, "-mat_no_unroll", &flg, NULL));
1509: if (!flg) {
1510: switch (bs) {
1511: case 1:
1512: B->ops->mult = MatMult_SeqSBAIJ_1;
1513: B->ops->multadd = MatMultAdd_SeqSBAIJ_1;
1514: B->ops->multtranspose = MatMult_SeqSBAIJ_1;
1515: B->ops->multtransposeadd = MatMultAdd_SeqSBAIJ_1;
1516: break;
1517: case 2:
1518: B->ops->mult = MatMult_SeqSBAIJ_2;
1519: B->ops->multadd = MatMultAdd_SeqSBAIJ_2;
1520: B->ops->multtranspose = MatMult_SeqSBAIJ_2;
1521: B->ops->multtransposeadd = MatMultAdd_SeqSBAIJ_2;
1522: break;
1523: case 3:
1524: B->ops->mult = MatMult_SeqSBAIJ_3;
1525: B->ops->multadd = MatMultAdd_SeqSBAIJ_3;
1526: B->ops->multtranspose = MatMult_SeqSBAIJ_3;
1527: B->ops->multtransposeadd = MatMultAdd_SeqSBAIJ_3;
1528: break;
1529: case 4:
1530: B->ops->mult = MatMult_SeqSBAIJ_4;
1531: B->ops->multadd = MatMultAdd_SeqSBAIJ_4;
1532: B->ops->multtranspose = MatMult_SeqSBAIJ_4;
1533: B->ops->multtransposeadd = MatMultAdd_SeqSBAIJ_4;
1534: break;
1535: case 5:
1536: B->ops->mult = MatMult_SeqSBAIJ_5;
1537: B->ops->multadd = MatMultAdd_SeqSBAIJ_5;
1538: B->ops->multtranspose = MatMult_SeqSBAIJ_5;
1539: B->ops->multtransposeadd = MatMultAdd_SeqSBAIJ_5;
1540: break;
1541: case 6:
1542: B->ops->mult = MatMult_SeqSBAIJ_6;
1543: B->ops->multadd = MatMultAdd_SeqSBAIJ_6;
1544: B->ops->multtranspose = MatMult_SeqSBAIJ_6;
1545: B->ops->multtransposeadd = MatMultAdd_SeqSBAIJ_6;
1546: break;
1547: case 7:
1548: B->ops->mult = MatMult_SeqSBAIJ_7;
1549: B->ops->multadd = MatMultAdd_SeqSBAIJ_7;
1550: B->ops->multtranspose = MatMult_SeqSBAIJ_7;
1551: B->ops->multtransposeadd = MatMultAdd_SeqSBAIJ_7;
1552: break;
1553: }
1554: }
1556: b->mbs = mbs;
1557: b->nbs = nbs;
1558: if (!skipallocation) {
1559: if (!b->imax) {
1560: PetscCall(PetscMalloc2(mbs, &b->imax, mbs, &b->ilen));
1562: b->free_imax_ilen = PETSC_TRUE;
1563: }
1564: if (!nnz) {
1565: if (nz == PETSC_DEFAULT || nz == PETSC_DECIDE) nz = 5;
1566: else if (nz <= 0) nz = 1;
1567: nz = PetscMin(nbs, nz);
1568: for (i = 0; i < mbs; i++) b->imax[i] = nz;
1569: PetscCall(PetscIntMultError(nz, mbs, &nz));
1570: } else {
1571: PetscInt64 nz64 = 0;
1572: for (i = 0; i < mbs; i++) {
1573: b->imax[i] = nnz[i];
1574: nz64 += nnz[i];
1575: }
1576: PetscCall(PetscIntCast(nz64, &nz));
1577: }
1578: /* b->ilen will count nonzeros in each block row so far. */
1579: for (i = 0; i < mbs; i++) b->ilen[i] = 0;
1580: /* nz=(nz+mbs)/2; */ /* total diagonal and superdiagonal nonzero blocks */
1582: /* allocate the matrix space */
1583: PetscCall(MatSeqXAIJFreeAIJ(B, &b->a, &b->j, &b->i));
1584: PetscCall(PetscShmgetAllocateArray(bs2 * nz, sizeof(PetscScalar), (void **)&b->a));
1585: PetscCall(PetscShmgetAllocateArray(nz, sizeof(PetscInt), (void **)&b->j));
1586: PetscCall(PetscShmgetAllocateArray(B->rmap->n + 1, sizeof(PetscInt), (void **)&b->i));
1587: b->free_a = PETSC_TRUE;
1588: b->free_ij = PETSC_TRUE;
1589: PetscCall(PetscArrayzero(b->a, nz * bs2));
1590: PetscCall(PetscArrayzero(b->j, nz));
1591: b->free_a = PETSC_TRUE;
1592: b->free_ij = PETSC_TRUE;
1594: /* pointer to beginning of each row */
1595: b->i[0] = 0;
1596: for (i = 1; i < mbs + 1; i++) b->i[i] = b->i[i - 1] + b->imax[i - 1];
1598: } else {
1599: b->free_a = PETSC_FALSE;
1600: b->free_ij = PETSC_FALSE;
1601: }
1603: b->bs2 = bs2;
1604: b->nz = 0;
1605: b->maxnz = nz;
1606: b->inew = NULL;
1607: b->jnew = NULL;
1608: b->anew = NULL;
1609: b->a2anew = NULL;
1610: b->permute = PETSC_FALSE;
1612: B->was_assembled = PETSC_FALSE;
1613: B->assembled = PETSC_FALSE;
1614: if (realalloc) PetscCall(MatSetOption(B, MAT_NEW_NONZERO_ALLOCATION_ERR, PETSC_TRUE));
1615: PetscFunctionReturn(PETSC_SUCCESS);
1616: }
1618: static PetscErrorCode MatSeqSBAIJSetPreallocationCSR_SeqSBAIJ(Mat B, PetscInt bs, const PetscInt ii[], const PetscInt jj[], const PetscScalar V[])
1619: {
1620: PetscInt i, j, m, nz, anz, nz_max = 0, *nnz;
1621: PetscScalar *values = NULL;
1622: Mat_SeqSBAIJ *b = (Mat_SeqSBAIJ *)B->data;
1623: PetscBool roworiented = b->roworiented;
1624: PetscBool ilw = b->ignore_ltriangular;
1626: PetscFunctionBegin;
1627: PetscCheck(bs >= 1, PetscObjectComm((PetscObject)B), PETSC_ERR_ARG_OUTOFRANGE, "Invalid block size specified, must be positive but it is %" PetscInt_FMT, bs);
1628: PetscCall(PetscLayoutSetBlockSize(B->rmap, bs));
1629: PetscCall(PetscLayoutSetBlockSize(B->cmap, bs));
1630: PetscCall(PetscLayoutSetUp(B->rmap));
1631: PetscCall(PetscLayoutSetUp(B->cmap));
1632: PetscCall(PetscLayoutGetBlockSize(B->rmap, &bs));
1633: m = B->rmap->n / bs;
1635: PetscCheck(!ii[0], PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "ii[0] must be 0 but it is %" PetscInt_FMT, ii[0]);
1636: PetscCall(PetscMalloc1(m + 1, &nnz));
1637: for (i = 0; i < m; i++) {
1638: nz = ii[i + 1] - ii[i];
1639: PetscCheck(nz >= 0, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Row %" PetscInt_FMT " has a negative number of columns %" PetscInt_FMT, i, nz);
1640: PetscCheckSorted(nz, jj + ii[i]);
1641: anz = 0;
1642: for (j = 0; j < nz; j++) {
1643: /* count only values on the diagonal or above */
1644: if (jj[ii[i] + j] >= i) {
1645: anz = nz - j;
1646: break;
1647: }
1648: }
1649: nz_max = PetscMax(nz_max, nz);
1650: nnz[i] = anz;
1651: }
1652: PetscCall(MatSeqSBAIJSetPreallocation(B, bs, 0, nnz));
1653: PetscCall(PetscFree(nnz));
1655: values = (PetscScalar *)V;
1656: if (!values) PetscCall(PetscCalloc1(bs * bs * nz_max, &values));
1657: b->ignore_ltriangular = PETSC_TRUE;
1658: for (i = 0; i < m; i++) {
1659: PetscInt ncols = ii[i + 1] - ii[i];
1660: const PetscInt *icols = jj + ii[i];
1662: if (!roworiented || bs == 1) {
1663: const PetscScalar *svals = values + (V ? (bs * bs * ii[i]) : 0);
1664: PetscCall(MatSetValuesBlocked_SeqSBAIJ(B, 1, &i, ncols, icols, svals, INSERT_VALUES));
1665: } else {
1666: for (j = 0; j < ncols; j++) {
1667: const PetscScalar *svals = values + (V ? (bs * bs * (ii[i] + j)) : 0);
1668: PetscCall(MatSetValuesBlocked_SeqSBAIJ(B, 1, &i, 1, &icols[j], svals, INSERT_VALUES));
1669: }
1670: }
1671: }
1672: if (!V) PetscCall(PetscFree(values));
1673: PetscCall(MatAssemblyBegin(B, MAT_FINAL_ASSEMBLY));
1674: PetscCall(MatAssemblyEnd(B, MAT_FINAL_ASSEMBLY));
1675: PetscCall(MatSetOption(B, MAT_NEW_NONZERO_LOCATION_ERR, PETSC_TRUE));
1676: b->ignore_ltriangular = ilw;
1677: PetscFunctionReturn(PETSC_SUCCESS);
1678: }
1680: /*
1681: This is used to set the numeric factorization for both Cholesky and ICC symbolic factorization
1682: */
1683: PetscErrorCode MatSeqSBAIJSetNumericFactorization_inplace(Mat B, PetscBool natural)
1684: {
1685: PetscBool flg = PETSC_FALSE;
1686: PetscInt bs = B->rmap->bs;
1688: PetscFunctionBegin;
1689: PetscCall(PetscOptionsGetBool(((PetscObject)B)->options, ((PetscObject)B)->prefix, "-mat_no_unroll", &flg, NULL));
1690: if (flg) bs = 8;
1692: if (!natural) {
1693: switch (bs) {
1694: case 1:
1695: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_1_inplace;
1696: break;
1697: case 2:
1698: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_2;
1699: break;
1700: case 3:
1701: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_3;
1702: break;
1703: case 4:
1704: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_4;
1705: break;
1706: case 5:
1707: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_5;
1708: break;
1709: case 6:
1710: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_6;
1711: break;
1712: case 7:
1713: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_7;
1714: break;
1715: default:
1716: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_N;
1717: break;
1718: }
1719: } else {
1720: switch (bs) {
1721: case 1:
1722: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_1_NaturalOrdering_inplace;
1723: break;
1724: case 2:
1725: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_2_NaturalOrdering;
1726: break;
1727: case 3:
1728: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_3_NaturalOrdering;
1729: break;
1730: case 4:
1731: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_4_NaturalOrdering;
1732: break;
1733: case 5:
1734: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_5_NaturalOrdering;
1735: break;
1736: case 6:
1737: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_6_NaturalOrdering;
1738: break;
1739: case 7:
1740: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_7_NaturalOrdering;
1741: break;
1742: default:
1743: B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqSBAIJ_N_NaturalOrdering;
1744: break;
1745: }
1746: }
1747: PetscFunctionReturn(PETSC_SUCCESS);
1748: }
1750: PETSC_INTERN PetscErrorCode MatConvert_SeqSBAIJ_SeqAIJ(Mat, MatType, MatReuse, Mat *);
1751: PETSC_INTERN PetscErrorCode MatConvert_SeqSBAIJ_SeqBAIJ(Mat, MatType, MatReuse, Mat *);
1752: static PetscErrorCode MatFactorGetSolverType_petsc(Mat A, MatSolverType *type)
1753: {
1754: PetscFunctionBegin;
1755: *type = MATSOLVERPETSC;
1756: PetscFunctionReturn(PETSC_SUCCESS);
1757: }
1759: PETSC_INTERN PetscErrorCode MatGetFactor_seqsbaij_petsc(Mat A, MatFactorType ftype, Mat *B)
1760: {
1761: PetscInt n = A->rmap->n;
1763: PetscFunctionBegin;
1764: #if defined(PETSC_USE_COMPLEX)
1765: if ((ftype == MAT_FACTOR_CHOLESKY || ftype == MAT_FACTOR_ICC) && A->hermitian == PETSC_BOOL3_TRUE && A->symmetric != PETSC_BOOL3_TRUE) {
1766: PetscCall(PetscInfo(A, "Hermitian MAT_FACTOR_CHOLESKY or MAT_FACTOR_ICC are not supported. Use MAT_FACTOR_LU instead.\n"));
1767: *B = NULL;
1768: PetscFunctionReturn(PETSC_SUCCESS);
1769: }
1770: #endif
1772: PetscCall(MatCreate(PetscObjectComm((PetscObject)A), B));
1773: PetscCall(MatSetSizes(*B, n, n, n, n));
1774: if (ftype == MAT_FACTOR_CHOLESKY || ftype == MAT_FACTOR_ICC) {
1775: PetscCall(MatSetType(*B, MATSEQSBAIJ));
1776: PetscCall(MatSeqSBAIJSetPreallocation(*B, A->rmap->bs, MAT_SKIP_ALLOCATION, NULL));
1778: (*B)->ops->choleskyfactorsymbolic = MatCholeskyFactorSymbolic_SeqSBAIJ;
1779: (*B)->ops->iccfactorsymbolic = MatICCFactorSymbolic_SeqSBAIJ;
1780: PetscCall(PetscStrallocpy(MATORDERINGNATURAL, (char **)&(*B)->preferredordering[MAT_FACTOR_CHOLESKY]));
1781: PetscCall(PetscStrallocpy(MATORDERINGNATURAL, (char **)&(*B)->preferredordering[MAT_FACTOR_ICC]));
1782: } else SETERRQ(PETSC_COMM_SELF, PETSC_ERR_SUP, "Factor type not supported");
1784: (*B)->factortype = ftype;
1785: (*B)->canuseordering = PETSC_TRUE;
1786: PetscCall(PetscFree((*B)->solvertype));
1787: PetscCall(PetscStrallocpy(MATSOLVERPETSC, &(*B)->solvertype));
1788: PetscCall(PetscObjectComposeFunction((PetscObject)*B, "MatFactorGetSolverType_C", MatFactorGetSolverType_petsc));
1789: PetscFunctionReturn(PETSC_SUCCESS);
1790: }
1792: /*@C
1793: MatSeqSBAIJGetArray - gives access to the array where the numerical data for a `MATSEQSBAIJ` matrix is stored
1795: Not Collective
1797: Input Parameter:
1798: . A - a `MATSEQSBAIJ` matrix
1800: Output Parameter:
1801: . array - pointer to the data
1803: Level: intermediate
1805: .seealso: [](ch_matrices), `Mat`, `MATSEQSBAIJ`, `MatSeqSBAIJRestoreArray()`, `MatSeqAIJGetArray()`, `MatSeqAIJRestoreArray()`
1806: @*/
1807: PetscErrorCode MatSeqSBAIJGetArray(Mat A, PetscScalar *array[])
1808: {
1809: PetscFunctionBegin;
1810: PetscUseMethod(A, "MatSeqSBAIJGetArray_C", (Mat, PetscScalar **), (A, array));
1811: PetscFunctionReturn(PETSC_SUCCESS);
1812: }
1814: /*@C
1815: MatSeqSBAIJRestoreArray - returns access to the array where the numerical data for a `MATSEQSBAIJ` matrix is stored obtained by `MatSeqSBAIJGetArray()`
1817: Not Collective
1819: Input Parameters:
1820: + A - a `MATSEQSBAIJ` matrix
1821: - array - pointer to the data
1823: Level: intermediate
1825: .seealso: [](ch_matrices), `Mat`, `MATSEQSBAIJ`, `MatSeqSBAIJGetArray()`, `MatSeqAIJGetArray()`, `MatSeqAIJRestoreArray()`
1826: @*/
1827: PetscErrorCode MatSeqSBAIJRestoreArray(Mat A, PetscScalar *array[])
1828: {
1829: PetscFunctionBegin;
1830: PetscUseMethod(A, "MatSeqSBAIJRestoreArray_C", (Mat, PetscScalar **), (A, array));
1831: PetscFunctionReturn(PETSC_SUCCESS);
1832: }
1834: /*MC
1835: MATSEQSBAIJ - MATSEQSBAIJ = "seqsbaij" - A matrix type to be used for sequential symmetric block sparse matrices,
1836: based on block compressed sparse row format. Only the upper triangular portion of the matrix is stored.
1838: For complex numbers by default this matrix is symmetric, NOT Hermitian symmetric. To make it Hermitian symmetric you
1839: can call `MatSetOption`(`Mat`, `MAT_HERMITIAN`).
1841: Options Database Key:
1842: . -mat_type seqsbaij - sets the matrix type to "seqsbaij" during a call to `MatSetFromOptions()`
1844: Level: beginner
1846: Notes:
1847: By default if you insert values into the lower triangular part of the matrix they are simply ignored (since they are not
1848: stored and it is assumed they symmetric to the upper triangular). If you call `MatSetOption`(`Mat`,`MAT_IGNORE_LOWER_TRIANGULAR`,`PETSC_FALSE`) or use
1849: 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.
1851: The number of rows in the matrix must be less than or equal to the number of columns
1853: .seealso: [](ch_matrices), `Mat`, `MATSEQSBAIJ`, `MatCreateSeqSBAIJ()`, `MatType`, `MATMPISBAIJ`
1854: M*/
1855: PETSC_EXTERN PetscErrorCode MatCreate_SeqSBAIJ(Mat B)
1856: {
1857: Mat_SeqSBAIJ *b;
1858: PetscMPIInt size;
1859: PetscBool no_unroll = PETSC_FALSE, no_inode = PETSC_FALSE;
1861: PetscFunctionBegin;
1862: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)B), &size));
1863: PetscCheck(size <= 1, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONG, "Comm must be of size 1");
1865: PetscCall(PetscNew(&b));
1866: B->data = (void *)b;
1867: B->ops[0] = MatOps_Values;
1869: B->ops->destroy = MatDestroy_SeqSBAIJ;
1870: B->ops->view = MatView_SeqSBAIJ;
1871: b->row = NULL;
1872: b->icol = NULL;
1873: b->reallocs = 0;
1874: b->saved_values = NULL;
1875: b->inode.limit = 5;
1876: b->inode.max_limit = 5;
1878: b->roworiented = PETSC_TRUE;
1879: b->nonew = 0;
1880: b->diag = NULL;
1881: b->solve_work = NULL;
1882: b->mult_work = NULL;
1883: B->spptr = NULL;
1884: B->info.nz_unneeded = (PetscReal)b->maxnz * b->bs2;
1885: b->keepnonzeropattern = PETSC_FALSE;
1887: b->inew = NULL;
1888: b->jnew = NULL;
1889: b->anew = NULL;
1890: b->a2anew = NULL;
1891: b->permute = PETSC_FALSE;
1893: b->ignore_ltriangular = PETSC_TRUE;
1895: PetscCall(PetscOptionsGetBool(((PetscObject)B)->options, ((PetscObject)B)->prefix, "-mat_ignore_lower_triangular", &b->ignore_ltriangular, NULL));
1897: b->getrow_utriangular = PETSC_FALSE;
1899: PetscCall(PetscOptionsGetBool(((PetscObject)B)->options, ((PetscObject)B)->prefix, "-mat_getrow_uppertriangular", &b->getrow_utriangular, NULL));
1901: PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatSeqSBAIJGetArray_C", MatSeqSBAIJGetArray_SeqSBAIJ));
1902: PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatSeqSBAIJRestoreArray_C", MatSeqSBAIJRestoreArray_SeqSBAIJ));
1903: PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatStoreValues_C", MatStoreValues_SeqSBAIJ));
1904: PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatRetrieveValues_C", MatRetrieveValues_SeqSBAIJ));
1905: PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatSeqSBAIJSetColumnIndices_C", MatSeqSBAIJSetColumnIndices_SeqSBAIJ));
1906: PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatConvert_seqsbaij_seqaij_C", MatConvert_SeqSBAIJ_SeqAIJ));
1907: PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatConvert_seqsbaij_seqbaij_C", MatConvert_SeqSBAIJ_SeqBAIJ));
1908: PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatSeqSBAIJSetPreallocation_C", MatSeqSBAIJSetPreallocation_SeqSBAIJ));
1909: PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatSeqSBAIJSetPreallocationCSR_C", MatSeqSBAIJSetPreallocationCSR_SeqSBAIJ));
1910: #if defined(PETSC_HAVE_ELEMENTAL)
1911: PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatConvert_seqsbaij_elemental_C", MatConvert_SeqSBAIJ_Elemental));
1912: #endif
1913: #if defined(PETSC_HAVE_SCALAPACK)
1914: PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatConvert_seqsbaij_scalapack_C", MatConvert_SBAIJ_ScaLAPACK));
1915: #endif
1917: B->symmetry_eternal = PETSC_TRUE;
1918: B->structural_symmetry_eternal = PETSC_TRUE;
1919: B->symmetric = PETSC_BOOL3_TRUE;
1920: B->structurally_symmetric = PETSC_BOOL3_TRUE;
1921: #if defined(PETSC_USE_COMPLEX)
1922: B->hermitian = PETSC_BOOL3_FALSE;
1923: #else
1924: B->hermitian = PETSC_BOOL3_TRUE;
1925: #endif
1927: PetscCall(PetscObjectChangeTypeName((PetscObject)B, MATSEQSBAIJ));
1929: PetscOptionsBegin(PetscObjectComm((PetscObject)B), ((PetscObject)B)->prefix, "Options for SEQSBAIJ matrix", "Mat");
1930: PetscCall(PetscOptionsBool("-mat_no_unroll", "Do not optimize for inodes (slower)", NULL, no_unroll, &no_unroll, NULL));
1931: if (no_unroll) PetscCall(PetscInfo(B, "Not using Inode routines due to -mat_no_unroll\n"));
1932: PetscCall(PetscOptionsBool("-mat_no_inode", "Do not optimize for inodes (slower)", NULL, no_inode, &no_inode, NULL));
1933: if (no_inode) PetscCall(PetscInfo(B, "Not using Inode routines due to -mat_no_inode\n"));
1934: PetscCall(PetscOptionsInt("-mat_inode_limit", "Do not use inodes larger then this value", NULL, b->inode.limit, &b->inode.limit, NULL));
1935: PetscOptionsEnd();
1936: b->inode.use = (PetscBool)(!(no_unroll || no_inode));
1937: if (b->inode.limit > b->inode.max_limit) b->inode.limit = b->inode.max_limit;
1938: PetscFunctionReturn(PETSC_SUCCESS);
1939: }
1941: /*@
1942: MatSeqSBAIJSetPreallocation - Creates a sparse symmetric matrix in block AIJ (block
1943: compressed row) `MATSEQSBAIJ` format. For good matrix assembly performance the
1944: user should preallocate the matrix storage by setting the parameter `nz`
1945: (or the array `nnz`).
1947: Collective
1949: Input Parameters:
1950: + B - the symmetric matrix
1951: . bs - size of block, the blocks are ALWAYS square. One can use `MatSetBlockSizes()` to set a different row and column blocksize but the row
1952: blocksize always defines the size of the blocks. The column blocksize sets the blocksize of the vectors obtained with `MatCreateVecs()`
1953: . nz - number of block nonzeros per block row (same for all rows)
1954: - nnz - array containing the number of block nonzeros in the upper triangular plus
1955: diagonal portion of each block (possibly different for each block row) or `NULL`
1957: Options Database Keys:
1958: + -mat_no_unroll - uses code that does not unroll the loops in the block calculations (much slower)
1959: - -mat_block_size - size of the blocks to use (only works if a negative bs is passed in
1961: Level: intermediate
1963: Notes:
1964: Specify the preallocated storage with either `nz` or `nnz` (not both).
1965: Set `nz` = `PETSC_DEFAULT` and `nnz` = `NULL` for PETSc to control dynamic memory
1966: allocation. See [Sparse Matrices](sec_matsparse) for details.
1968: You can call `MatGetInfo()` to get information on how effective the preallocation was;
1969: for example the fields mallocs,nz_allocated,nz_used,nz_unneeded;
1970: You can also run with the option `-info` and look for messages with the string
1971: malloc in them to see if additional memory allocation was needed.
1973: If the `nnz` parameter is given then the `nz` parameter is ignored
1975: .seealso: [](ch_matrices), `Mat`, [Sparse Matrices](sec_matsparse), `MATSEQSBAIJ`, `MatCreate()`, `MatCreateSeqAIJ()`, `MatSetValues()`, `MatCreateSBAIJ()`
1976: @*/
1977: PetscErrorCode MatSeqSBAIJSetPreallocation(Mat B, PetscInt bs, PetscInt nz, const PetscInt nnz[])
1978: {
1979: PetscFunctionBegin;
1983: PetscTryMethod(B, "MatSeqSBAIJSetPreallocation_C", (Mat, PetscInt, PetscInt, const PetscInt[]), (B, bs, nz, nnz));
1984: PetscFunctionReturn(PETSC_SUCCESS);
1985: }
1987: /*@C
1988: MatSeqSBAIJSetPreallocationCSR - Creates a sparse parallel matrix in `MATSEQSBAIJ` format using the given nonzero structure and (optional) numerical values
1990: Input Parameters:
1991: + B - the matrix
1992: . bs - size of block, the blocks are ALWAYS square.
1993: . i - the indices into `j` for the start of each local row (indices start with zero)
1994: . j - the column indices for each local row (indices start with zero) these must be sorted for each row
1995: - v - optional values in the matrix, use `NULL` if not provided
1997: Level: advanced
1999: Notes:
2000: The `i`,`j`,`v` values are COPIED with this routine; to avoid the copy use `MatCreateSeqSBAIJWithArrays()`
2002: The order of the entries in values is specified by the `MatOption` `MAT_ROW_ORIENTED`. For example, C programs
2003: may want to use the default `MAT_ROW_ORIENTED` = `PETSC_TRUE` and use an array v[nnz][bs][bs] where the second index is
2004: over rows within a block and the last index is over columns within a block row. Fortran programs will likely set
2005: `MAT_ROW_ORIENTED` = `PETSC_FALSE` and use a Fortran array v(bs,bs,nnz) in which the first index is over rows within a
2006: block column and the second index is over columns within a block.
2008: Any entries provided that lie below the diagonal are ignored
2010: Though this routine has Preallocation() in the name it also sets the exact nonzero locations of the matrix entries
2011: and usually the numerical values as well
2013: .seealso: [](ch_matrices), `Mat`, `MATSEQSBAIJ`, `MatCreate()`, `MatCreateSeqSBAIJ()`, `MatSetValuesBlocked()`, `MatSeqSBAIJSetPreallocation()`
2014: @*/
2015: PetscErrorCode MatSeqSBAIJSetPreallocationCSR(Mat B, PetscInt bs, const PetscInt i[], const PetscInt j[], const PetscScalar v[])
2016: {
2017: PetscFunctionBegin;
2021: PetscTryMethod(B, "MatSeqSBAIJSetPreallocationCSR_C", (Mat, PetscInt, const PetscInt[], const PetscInt[], const PetscScalar[]), (B, bs, i, j, v));
2022: PetscFunctionReturn(PETSC_SUCCESS);
2023: }
2025: /*@
2026: MatCreateSeqSBAIJ - Creates a sparse symmetric matrix in (block
2027: compressed row) `MATSEQSBAIJ` format. For good matrix assembly performance the
2028: user should preallocate the matrix storage by setting the parameter `nz`
2029: (or the array `nnz`).
2031: Collective
2033: Input Parameters:
2034: + comm - MPI communicator, set to `PETSC_COMM_SELF`
2035: . bs - size of block, the blocks are ALWAYS square. One can use `MatSetBlockSizes()` to set a different row and column blocksize but the row
2036: blocksize always defines the size of the blocks. The column blocksize sets the blocksize of the vectors obtained with MatCreateVecs()
2037: . m - number of rows
2038: . n - number of columns
2039: . nz - number of block nonzeros per block row (same for all rows)
2040: - nnz - array containing the number of block nonzeros in the upper triangular plus
2041: diagonal portion of each block (possibly different for each block row) or `NULL`
2043: Output Parameter:
2044: . A - the symmetric matrix
2046: Options Database Keys:
2047: + -mat_no_unroll - uses code that does not unroll the loops in the block calculations (much slower)
2048: - -mat_block_size - size of the blocks to use
2050: Level: intermediate
2052: Notes:
2053: It is recommended that one use `MatCreateFromOptions()` or the `MatCreate()`, `MatSetType()` and/or `MatSetFromOptions()`,
2054: MatXXXXSetPreallocation() paradigm instead of this routine directly.
2055: [MatXXXXSetPreallocation() is, for example, `MatSeqAIJSetPreallocation()`]
2057: The number of rows and columns must be divisible by blocksize.
2058: This matrix type does not support complex Hermitian operation.
2060: Specify the preallocated storage with either `nz` or `nnz` (not both).
2061: Set `nz` = `PETSC_DEFAULT` and `nnz` = `NULL` for PETSc to control dynamic memory
2062: allocation. See [Sparse Matrices](sec_matsparse) for details.
2064: If the `nnz` parameter is given then the `nz` parameter is ignored
2066: .seealso: [](ch_matrices), `Mat`, [Sparse Matrices](sec_matsparse), `MATSEQSBAIJ`, `MatCreate()`, `MatCreateSeqAIJ()`, `MatSetValues()`, `MatCreateSBAIJ()`
2067: @*/
2068: PetscErrorCode MatCreateSeqSBAIJ(MPI_Comm comm, PetscInt bs, PetscInt m, PetscInt n, PetscInt nz, const PetscInt nnz[], Mat *A)
2069: {
2070: PetscFunctionBegin;
2071: PetscCall(MatCreate(comm, A));
2072: PetscCall(MatSetSizes(*A, m, n, m, n));
2073: PetscCall(MatSetType(*A, MATSEQSBAIJ));
2074: PetscCall(MatSeqSBAIJSetPreallocation(*A, bs, nz, (PetscInt *)nnz));
2075: PetscFunctionReturn(PETSC_SUCCESS);
2076: }
2078: PetscErrorCode MatDuplicate_SeqSBAIJ(Mat A, MatDuplicateOption cpvalues, Mat *B)
2079: {
2080: Mat C;
2081: Mat_SeqSBAIJ *c, *a = (Mat_SeqSBAIJ *)A->data;
2082: PetscInt i, mbs = a->mbs, nz = a->nz, bs2 = a->bs2;
2084: PetscFunctionBegin;
2085: PetscCheck(A->assembled, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONGSTATE, "Cannot duplicate unassembled matrix");
2086: PetscCheck(a->i[mbs] == nz, PETSC_COMM_SELF, PETSC_ERR_PLIB, "Corrupt matrix");
2088: *B = NULL;
2089: PetscCall(MatCreate(PetscObjectComm((PetscObject)A), &C));
2090: PetscCall(MatSetSizes(C, A->rmap->N, A->cmap->n, A->rmap->N, A->cmap->n));
2091: PetscCall(MatSetBlockSizesFromMats(C, A, A));
2092: PetscCall(MatSetType(C, MATSEQSBAIJ));
2093: c = (Mat_SeqSBAIJ *)C->data;
2095: C->preallocated = PETSC_TRUE;
2096: C->factortype = A->factortype;
2097: c->row = NULL;
2098: c->icol = NULL;
2099: c->saved_values = NULL;
2100: c->keepnonzeropattern = a->keepnonzeropattern;
2101: C->assembled = PETSC_TRUE;
2103: PetscCall(PetscLayoutReference(A->rmap, &C->rmap));
2104: PetscCall(PetscLayoutReference(A->cmap, &C->cmap));
2105: c->bs2 = a->bs2;
2106: c->mbs = a->mbs;
2107: c->nbs = a->nbs;
2109: if (cpvalues == MAT_SHARE_NONZERO_PATTERN) {
2110: c->imax = a->imax;
2111: c->ilen = a->ilen;
2112: c->free_imax_ilen = PETSC_FALSE;
2113: } else {
2114: PetscCall(PetscMalloc2(mbs + 1, &c->imax, mbs + 1, &c->ilen));
2115: for (i = 0; i < mbs; i++) {
2116: c->imax[i] = a->imax[i];
2117: c->ilen[i] = a->ilen[i];
2118: }
2119: c->free_imax_ilen = PETSC_TRUE;
2120: }
2122: /* allocate the matrix space */
2123: PetscCall(PetscShmgetAllocateArray(bs2 * nz, sizeof(PetscScalar), (void **)&c->a));
2124: c->free_a = PETSC_TRUE;
2125: if (cpvalues == MAT_SHARE_NONZERO_PATTERN) {
2126: PetscCall(PetscArrayzero(c->a, bs2 * nz));
2127: c->i = a->i;
2128: c->j = a->j;
2129: c->free_ij = PETSC_FALSE;
2130: c->parent = A;
2131: PetscCall(PetscObjectReference((PetscObject)A));
2132: PetscCall(MatSetOption(A, MAT_NEW_NONZERO_LOCATION_ERR, PETSC_TRUE));
2133: PetscCall(MatSetOption(C, MAT_NEW_NONZERO_LOCATION_ERR, PETSC_TRUE));
2134: } else {
2135: PetscCall(PetscShmgetAllocateArray(nz, sizeof(PetscInt), (void **)&c->j));
2136: PetscCall(PetscShmgetAllocateArray(mbs + 1, sizeof(PetscInt), (void **)&c->i));
2137: PetscCall(PetscArraycpy(c->i, a->i, mbs + 1));
2138: c->free_ij = PETSC_TRUE;
2139: }
2140: if (mbs > 0) {
2141: if (cpvalues != MAT_SHARE_NONZERO_PATTERN) PetscCall(PetscArraycpy(c->j, a->j, nz));
2142: if (cpvalues == MAT_COPY_VALUES) {
2143: PetscCall(PetscArraycpy(c->a, a->a, bs2 * nz));
2144: } else {
2145: PetscCall(PetscArrayzero(c->a, bs2 * nz));
2146: }
2147: if (a->jshort) {
2148: /* cannot share jshort, it is reallocated in MatAssemblyEnd_SeqSBAIJ() */
2149: /* if the parent matrix is reassembled, this child matrix will never notice */
2150: PetscCall(PetscMalloc1(nz, &c->jshort));
2151: PetscCall(PetscArraycpy(c->jshort, a->jshort, nz));
2153: c->free_jshort = PETSC_TRUE;
2154: }
2155: }
2157: c->roworiented = a->roworiented;
2158: c->nonew = a->nonew;
2160: if (a->diag) {
2161: if (cpvalues == MAT_SHARE_NONZERO_PATTERN) {
2162: c->diag = a->diag;
2163: c->free_diag = PETSC_FALSE;
2164: } else {
2165: PetscCall(PetscMalloc1(mbs, &c->diag));
2166: for (i = 0; i < mbs; i++) c->diag[i] = a->diag[i];
2167: c->free_diag = PETSC_TRUE;
2168: }
2169: }
2170: c->nz = a->nz;
2171: c->maxnz = a->nz; /* Since we allocate exactly the right amount */
2172: c->solve_work = NULL;
2173: c->mult_work = NULL;
2175: *B = C;
2176: PetscCall(PetscFunctionListDuplicate(((PetscObject)A)->qlist, &((PetscObject)C)->qlist));
2177: PetscFunctionReturn(PETSC_SUCCESS);
2178: }
2180: /* Used for both SeqBAIJ and SeqSBAIJ matrices */
2181: #define MatLoad_SeqSBAIJ_Binary MatLoad_SeqBAIJ_Binary
2183: PetscErrorCode MatLoad_SeqSBAIJ(Mat mat, PetscViewer viewer)
2184: {
2185: PetscBool isbinary;
2187: PetscFunctionBegin;
2188: PetscCall(PetscObjectTypeCompare((PetscObject)viewer, PETSCVIEWERBINARY, &isbinary));
2189: PetscCheck(isbinary, PetscObjectComm((PetscObject)viewer), PETSC_ERR_SUP, "Viewer type %s not yet supported for reading %s matrices", ((PetscObject)viewer)->type_name, ((PetscObject)mat)->type_name);
2190: PetscCall(MatLoad_SeqSBAIJ_Binary(mat, viewer));
2191: PetscFunctionReturn(PETSC_SUCCESS);
2192: }
2194: /*@
2195: MatCreateSeqSBAIJWithArrays - Creates an sequential `MATSEQSBAIJ` matrix using matrix elements
2196: (upper triangular entries in CSR format) provided by the user.
2198: Collective
2200: Input Parameters:
2201: + comm - must be an MPI communicator of size 1
2202: . bs - size of block
2203: . m - number of rows
2204: . n - number of columns
2205: . i - row indices; that is i[0] = 0, i[row] = i[row-1] + number of block elements in that row block row of the matrix
2206: . j - column indices
2207: - a - matrix values
2209: Output Parameter:
2210: . mat - the matrix
2212: Level: advanced
2214: Notes:
2215: The `i`, `j`, and `a` arrays are not copied by this routine, the user must free these arrays
2216: once the matrix is destroyed
2218: You cannot set new nonzero locations into this matrix, that will generate an error.
2220: The `i` and `j` indices are 0 based
2222: When block size is greater than 1 the matrix values must be stored using the `MATSBAIJ` storage format. For block size of 1
2223: it is the regular CSR format excluding the lower triangular elements.
2225: .seealso: [](ch_matrices), `Mat`, `MATSEQSBAIJ`, `MatCreate()`, `MatCreateSBAIJ()`, `MatCreateSeqSBAIJ()`
2226: @*/
2227: PetscErrorCode MatCreateSeqSBAIJWithArrays(MPI_Comm comm, PetscInt bs, PetscInt m, PetscInt n, PetscInt i[], PetscInt j[], PetscScalar a[], Mat *mat)
2228: {
2229: PetscInt ii;
2230: Mat_SeqSBAIJ *sbaij;
2232: PetscFunctionBegin;
2233: PetscCheck(bs == 1, PETSC_COMM_SELF, PETSC_ERR_SUP, "block size %" PetscInt_FMT " > 1 is not supported yet", bs);
2234: PetscCheck(m == 0 || i[0] == 0, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "i (row indices) must start with 0");
2236: PetscCall(MatCreate(comm, mat));
2237: PetscCall(MatSetSizes(*mat, m, n, m, n));
2238: PetscCall(MatSetType(*mat, MATSEQSBAIJ));
2239: PetscCall(MatSeqSBAIJSetPreallocation(*mat, bs, MAT_SKIP_ALLOCATION, NULL));
2240: sbaij = (Mat_SeqSBAIJ *)(*mat)->data;
2241: PetscCall(PetscMalloc2(m, &sbaij->imax, m, &sbaij->ilen));
2243: sbaij->i = i;
2244: sbaij->j = j;
2245: sbaij->a = a;
2247: sbaij->nonew = -1; /*this indicates that inserting a new value in the matrix that generates a new nonzero is an error*/
2248: sbaij->free_a = PETSC_FALSE;
2249: sbaij->free_ij = PETSC_FALSE;
2250: sbaij->free_imax_ilen = PETSC_TRUE;
2252: for (ii = 0; ii < m; ii++) {
2253: sbaij->ilen[ii] = sbaij->imax[ii] = i[ii + 1] - i[ii];
2254: PetscCheck(i[ii + 1] >= i[ii], PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Negative row length in i (row indices) row = %" PetscInt_FMT " length = %" PetscInt_FMT, ii, i[ii + 1] - i[ii]);
2255: }
2256: if (PetscDefined(USE_DEBUG)) {
2257: for (ii = 0; ii < sbaij->i[m]; ii++) {
2258: PetscCheck(j[ii] >= 0, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Negative column index at location = %" PetscInt_FMT " index = %" PetscInt_FMT, ii, j[ii]);
2259: PetscCheck(j[ii] < n, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Column index too large at location = %" PetscInt_FMT " index = %" PetscInt_FMT, ii, j[ii]);
2260: }
2261: }
2263: PetscCall(MatAssemblyBegin(*mat, MAT_FINAL_ASSEMBLY));
2264: PetscCall(MatAssemblyEnd(*mat, MAT_FINAL_ASSEMBLY));
2265: PetscFunctionReturn(PETSC_SUCCESS);
2266: }
2268: PetscErrorCode MatCreateMPIMatConcatenateSeqMat_SeqSBAIJ(MPI_Comm comm, Mat inmat, PetscInt n, MatReuse scall, Mat *outmat)
2269: {
2270: PetscFunctionBegin;
2271: PetscCall(MatCreateMPIMatConcatenateSeqMat_MPISBAIJ(comm, inmat, n, scall, outmat));
2272: PetscFunctionReturn(PETSC_SUCCESS);
2273: }