Actual source code: mpisbaij.c
1: #include <../src/mat/impls/baij/mpi/mpibaij.h>
2: #include <../src/mat/impls/sbaij/mpi/mpisbaij.h>
3: #include <../src/mat/impls/sbaij/seq/sbaij.h>
4: #include <petscblaslapack.h>
6: static PetscErrorCode MatDestroy_MPISBAIJ(Mat mat)
7: {
8: Mat_MPISBAIJ *baij = (Mat_MPISBAIJ *)mat->data;
10: PetscFunctionBegin;
11: PetscCall(PetscLogObjectState((PetscObject)mat, "Rows=%" PetscInt_FMT ",Cols=%" PetscInt_FMT, mat->rmap->N, mat->cmap->N));
12: PetscCall(MatStashDestroy_Private(&mat->stash));
13: PetscCall(MatStashDestroy_Private(&mat->bstash));
14: PetscCall(MatDestroy(&baij->A));
15: PetscCall(MatDestroy(&baij->B));
16: #if defined(PETSC_USE_CTABLE)
17: PetscCall(PetscHMapIDestroy(&baij->colmap));
18: #else
19: PetscCall(PetscFree(baij->colmap));
20: #endif
21: PetscCall(PetscFree(baij->garray));
22: PetscCall(VecDestroy(&baij->lvec));
23: PetscCall(VecScatterDestroy(&baij->Mvctx));
24: PetscCall(VecDestroy(&baij->slvec0));
25: PetscCall(VecDestroy(&baij->slvec0b));
26: PetscCall(VecDestroy(&baij->slvec1));
27: PetscCall(VecDestroy(&baij->slvec1a));
28: PetscCall(VecDestroy(&baij->slvec1b));
29: PetscCall(VecScatterDestroy(&baij->sMvctx));
30: PetscCall(PetscFree2(baij->rowvalues, baij->rowindices));
31: PetscCall(PetscFree(baij->barray));
32: PetscCall(PetscFree(baij->hd));
33: PetscCall(VecDestroy(&baij->diag));
34: PetscCall(VecDestroy(&baij->bb1));
35: PetscCall(VecDestroy(&baij->xx1));
36: #if defined(PETSC_USE_REAL_MAT_SINGLE)
37: PetscCall(PetscFree(baij->setvaluescopy));
38: #endif
39: PetscCall(PetscFree(baij->in_loc));
40: PetscCall(PetscFree(baij->v_loc));
41: PetscCall(PetscFree(baij->rangebs));
42: PetscCall(PetscFree(mat->data));
44: PetscCall(PetscObjectChangeTypeName((PetscObject)mat, NULL));
45: PetscCall(PetscObjectComposeFunction((PetscObject)mat, "MatStoreValues_C", NULL));
46: PetscCall(PetscObjectComposeFunction((PetscObject)mat, "MatRetrieveValues_C", NULL));
47: PetscCall(PetscObjectComposeFunction((PetscObject)mat, "MatMPISBAIJSetPreallocation_C", NULL));
48: PetscCall(PetscObjectComposeFunction((PetscObject)mat, "MatMPISBAIJSetPreallocationCSR_C", NULL));
49: #if defined(PETSC_HAVE_ELEMENTAL)
50: PetscCall(PetscObjectComposeFunction((PetscObject)mat, "MatConvert_mpisbaij_elemental_C", NULL));
51: #endif
52: #if defined(PETSC_HAVE_SCALAPACK)
53: PetscCall(PetscObjectComposeFunction((PetscObject)mat, "MatConvert_mpisbaij_scalapack_C", NULL));
54: #endif
55: PetscCall(PetscObjectComposeFunction((PetscObject)mat, "MatConvert_mpisbaij_mpiaij_C", NULL));
56: PetscCall(PetscObjectComposeFunction((PetscObject)mat, "MatConvert_mpisbaij_mpibaij_C", NULL));
57: PetscFunctionReturn(PETSC_SUCCESS);
58: }
60: /* defines MatSetValues_MPI_Hash(), MatAssemblyBegin_MPI_Hash(), MatAssemblyEnd_MPI_Hash(), MatSetUp_MPI_Hash() */
61: #define TYPE SBAIJ
62: #define TYPE_SBAIJ
63: #include "../src/mat/impls/aij/mpi/mpihashmat.h"
64: #undef TYPE
65: #undef TYPE_SBAIJ
67: #if defined(PETSC_HAVE_ELEMENTAL)
68: PETSC_INTERN PetscErrorCode MatConvert_MPISBAIJ_Elemental(Mat, MatType, MatReuse, Mat *);
69: #endif
70: #if defined(PETSC_HAVE_SCALAPACK)
71: PETSC_INTERN PetscErrorCode MatConvert_SBAIJ_ScaLAPACK(Mat, MatType, MatReuse, Mat *);
72: #endif
74: /* This could be moved to matimpl.h */
75: static PetscErrorCode MatPreallocateWithMats_Private(Mat B, PetscInt nm, Mat X[], PetscBool symm[], PetscBool fill)
76: {
77: Mat preallocator;
78: PetscInt r, rstart, rend;
79: PetscInt bs, i, m, n, M, N;
80: PetscBool cong = PETSC_TRUE;
82: PetscFunctionBegin;
85: for (i = 0; i < nm; i++) {
87: PetscCall(PetscLayoutCompare(B->rmap, X[i]->rmap, &cong));
88: PetscCheck(cong, PetscObjectComm((PetscObject)B), PETSC_ERR_SUP, "Not for different layouts");
89: }
91: PetscCall(MatGetBlockSize(B, &bs));
92: PetscCall(MatGetSize(B, &M, &N));
93: PetscCall(MatGetLocalSize(B, &m, &n));
94: PetscCall(MatCreate(PetscObjectComm((PetscObject)B), &preallocator));
95: PetscCall(MatSetType(preallocator, MATPREALLOCATOR));
96: PetscCall(MatSetBlockSize(preallocator, bs));
97: PetscCall(MatSetSizes(preallocator, m, n, M, N));
98: PetscCall(MatSetUp(preallocator));
99: PetscCall(MatGetOwnershipRange(preallocator, &rstart, &rend));
100: for (r = rstart; r < rend; ++r) {
101: PetscInt ncols;
102: const PetscInt *row;
103: const PetscScalar *vals;
105: for (i = 0; i < nm; i++) {
106: PetscCall(MatGetRow(X[i], r, &ncols, &row, &vals));
107: PetscCall(MatSetValues(preallocator, 1, &r, ncols, row, vals, INSERT_VALUES));
108: if (symm && symm[i]) PetscCall(MatSetValues(preallocator, ncols, row, 1, &r, vals, INSERT_VALUES));
109: PetscCall(MatRestoreRow(X[i], r, &ncols, &row, &vals));
110: }
111: }
112: PetscCall(MatAssemblyBegin(preallocator, MAT_FINAL_ASSEMBLY));
113: PetscCall(MatAssemblyEnd(preallocator, MAT_FINAL_ASSEMBLY));
114: PetscCall(MatPreallocatorPreallocate(preallocator, fill, B));
115: PetscCall(MatDestroy(&preallocator));
116: PetscFunctionReturn(PETSC_SUCCESS);
117: }
119: PETSC_INTERN PetscErrorCode MatConvert_MPISBAIJ_Basic(Mat A, MatType newtype, MatReuse reuse, Mat *newmat)
120: {
121: Mat B;
122: PetscInt r;
124: PetscFunctionBegin;
125: if (reuse != MAT_REUSE_MATRIX) {
126: PetscBool symm = PETSC_TRUE, isdense;
127: PetscInt bs;
129: PetscCall(MatCreate(PetscObjectComm((PetscObject)A), &B));
130: PetscCall(MatSetSizes(B, A->rmap->n, A->cmap->n, A->rmap->N, A->cmap->N));
131: PetscCall(MatSetType(B, newtype));
132: PetscCall(MatGetBlockSize(A, &bs));
133: PetscCall(MatSetBlockSize(B, bs));
134: PetscCall(PetscLayoutSetUp(B->rmap));
135: PetscCall(PetscLayoutSetUp(B->cmap));
136: PetscCall(PetscObjectTypeCompareAny((PetscObject)B, &isdense, MATSEQDENSE, MATMPIDENSE, MATSEQDENSECUDA, ""));
137: if (!isdense) {
138: PetscCall(MatGetRowUpperTriangular(A));
139: PetscCall(MatPreallocateWithMats_Private(B, 1, &A, &symm, PETSC_TRUE));
140: PetscCall(MatRestoreRowUpperTriangular(A));
141: } else {
142: PetscCall(MatSetUp(B));
143: }
144: } else {
145: B = *newmat;
146: PetscCall(MatZeroEntries(B));
147: }
149: PetscCall(MatGetRowUpperTriangular(A));
150: for (r = A->rmap->rstart; r < A->rmap->rend; r++) {
151: PetscInt ncols;
152: const PetscInt *row;
153: const PetscScalar *vals;
155: PetscCall(MatGetRow(A, r, &ncols, &row, &vals));
156: PetscCall(MatSetValues(B, 1, &r, ncols, row, vals, INSERT_VALUES));
157: #if defined(PETSC_USE_COMPLEX)
158: if (A->hermitian == PETSC_BOOL3_TRUE) {
159: PetscInt i;
160: for (i = 0; i < ncols; i++) PetscCall(MatSetValue(B, row[i], r, PetscConj(vals[i]), INSERT_VALUES));
161: } else {
162: PetscCall(MatSetValues(B, ncols, row, 1, &r, vals, INSERT_VALUES));
163: }
164: #else
165: PetscCall(MatSetValues(B, ncols, row, 1, &r, vals, INSERT_VALUES));
166: #endif
167: PetscCall(MatRestoreRow(A, r, &ncols, &row, &vals));
168: }
169: PetscCall(MatRestoreRowUpperTriangular(A));
170: PetscCall(MatAssemblyBegin(B, MAT_FINAL_ASSEMBLY));
171: PetscCall(MatAssemblyEnd(B, MAT_FINAL_ASSEMBLY));
173: if (reuse == MAT_INPLACE_MATRIX) {
174: PetscCall(MatHeaderReplace(A, &B));
175: } else {
176: *newmat = B;
177: }
178: PetscFunctionReturn(PETSC_SUCCESS);
179: }
181: static PetscErrorCode MatStoreValues_MPISBAIJ(Mat mat)
182: {
183: Mat_MPISBAIJ *aij = (Mat_MPISBAIJ *)mat->data;
185: PetscFunctionBegin;
186: PetscCall(MatStoreValues(aij->A));
187: PetscCall(MatStoreValues(aij->B));
188: PetscFunctionReturn(PETSC_SUCCESS);
189: }
191: static PetscErrorCode MatRetrieveValues_MPISBAIJ(Mat mat)
192: {
193: Mat_MPISBAIJ *aij = (Mat_MPISBAIJ *)mat->data;
195: PetscFunctionBegin;
196: PetscCall(MatRetrieveValues(aij->A));
197: PetscCall(MatRetrieveValues(aij->B));
198: PetscFunctionReturn(PETSC_SUCCESS);
199: }
201: #define MatSetValues_SeqSBAIJ_A_Private(row, col, value, addv, orow, ocol) \
202: do { \
203: brow = row / bs; \
204: rp = aj + ai[brow]; \
205: ap = aa + bs2 * ai[brow]; \
206: rmax = aimax[brow]; \
207: nrow = ailen[brow]; \
208: bcol = col / bs; \
209: ridx = row % bs; \
210: cidx = col % bs; \
211: low = 0; \
212: high = nrow; \
213: while (high - low > 3) { \
214: t = (low + high) / 2; \
215: if (rp[t] > bcol) high = t; \
216: else low = t; \
217: } \
218: for (_i = low; _i < high; _i++) { \
219: if (rp[_i] > bcol) break; \
220: if (rp[_i] == bcol) { \
221: bap = ap + bs2 * _i + bs * cidx + ridx; \
222: if (addv == ADD_VALUES) *bap += value; \
223: else *bap = value; \
224: goto a_noinsert; \
225: } \
226: } \
227: if (a->nonew == 1) goto a_noinsert; \
228: PetscCheck(a->nonew != -1, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Inserting a new nonzero at global row/column (%" PetscInt_FMT ", %" PetscInt_FMT ") into matrix", orow, ocol); \
229: MatSeqXAIJReallocateAIJ(A, a->mbs, bs2, nrow, brow, bcol, rmax, aa, ai, aj, rp, ap, aimax, a->nonew, MatScalar); \
230: N = nrow++ - 1; \
231: /* shift up all the later entries in this row */ \
232: PetscCall(PetscArraymove(rp + _i + 1, rp + _i, N - _i + 1)); \
233: PetscCall(PetscArraymove(ap + bs2 * (_i + 1), ap + bs2 * _i, bs2 * (N - _i + 1))); \
234: PetscCall(PetscArrayzero(ap + bs2 * _i, bs2)); \
235: rp[_i] = bcol; \
236: ap[bs2 * _i + bs * cidx + ridx] = value; \
237: A->nonzerostate++; \
238: a_noinsert:; \
239: ailen[brow] = nrow; \
240: } while (0)
242: #define MatSetValues_SeqSBAIJ_B_Private(row, col, value, addv, orow, ocol) \
243: do { \
244: brow = row / bs; \
245: rp = bj + bi[brow]; \
246: ap = ba + bs2 * bi[brow]; \
247: rmax = bimax[brow]; \
248: nrow = bilen[brow]; \
249: bcol = col / bs; \
250: ridx = row % bs; \
251: cidx = col % bs; \
252: low = 0; \
253: high = nrow; \
254: while (high - low > 3) { \
255: t = (low + high) / 2; \
256: if (rp[t] > bcol) high = t; \
257: else low = t; \
258: } \
259: for (_i = low; _i < high; _i++) { \
260: if (rp[_i] > bcol) break; \
261: if (rp[_i] == bcol) { \
262: bap = ap + bs2 * _i + bs * cidx + ridx; \
263: if (addv == ADD_VALUES) *bap += value; \
264: else *bap = value; \
265: goto b_noinsert; \
266: } \
267: } \
268: if (b->nonew == 1) goto b_noinsert; \
269: PetscCheck(b->nonew != -1, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Inserting a new nonzero at global row/column (%" PetscInt_FMT ", %" PetscInt_FMT ") into matrix", orow, ocol); \
270: MatSeqXAIJReallocateAIJ(B, b->mbs, bs2, nrow, brow, bcol, rmax, ba, bi, bj, rp, ap, bimax, b->nonew, MatScalar); \
271: N = nrow++ - 1; \
272: /* shift up all the later entries in this row */ \
273: PetscCall(PetscArraymove(rp + _i + 1, rp + _i, N - _i + 1)); \
274: PetscCall(PetscArraymove(ap + bs2 * (_i + 1), ap + bs2 * _i, bs2 * (N - _i + 1))); \
275: PetscCall(PetscArrayzero(ap + bs2 * _i, bs2)); \
276: rp[_i] = bcol; \
277: ap[bs2 * _i + bs * cidx + ridx] = value; \
278: B->nonzerostate++; \
279: b_noinsert:; \
280: bilen[brow] = nrow; \
281: } while (0)
283: /* Only add/insert a(i,j) with i<=j (blocks).
284: Any a(i,j) with i>j input by user is ignored or generates an error
285: */
286: static PetscErrorCode MatSetValues_MPISBAIJ(Mat mat, PetscInt m, const PetscInt im[], PetscInt n, const PetscInt in[], const PetscScalar v[], InsertMode addv)
287: {
288: Mat_MPISBAIJ *baij = (Mat_MPISBAIJ *)mat->data;
289: MatScalar value;
290: PetscBool roworiented = baij->roworiented;
291: PetscInt i, j, row, col;
292: PetscInt rstart_orig = mat->rmap->rstart;
293: PetscInt rend_orig = mat->rmap->rend, cstart_orig = mat->cmap->rstart;
294: PetscInt cend_orig = mat->cmap->rend, bs = mat->rmap->bs;
296: /* Some Variables required in the macro */
297: Mat A = baij->A;
298: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ *)(A)->data;
299: PetscInt *aimax = a->imax, *ai = a->i, *ailen = a->ilen, *aj = a->j;
300: MatScalar *aa = a->a;
302: Mat B = baij->B;
303: Mat_SeqBAIJ *b = (Mat_SeqBAIJ *)(B)->data;
304: PetscInt *bimax = b->imax, *bi = b->i, *bilen = b->ilen, *bj = b->j;
305: MatScalar *ba = b->a;
307: PetscInt *rp, ii, nrow, _i, rmax, N, brow, bcol;
308: PetscInt low, high, t, ridx, cidx, bs2 = a->bs2;
309: MatScalar *ap, *bap;
311: /* for stash */
312: PetscInt n_loc, *in_loc = NULL;
313: MatScalar *v_loc = NULL;
315: PetscFunctionBegin;
316: if (!baij->donotstash) {
317: if (n > baij->n_loc) {
318: PetscCall(PetscFree(baij->in_loc));
319: PetscCall(PetscFree(baij->v_loc));
320: PetscCall(PetscMalloc1(n, &baij->in_loc));
321: PetscCall(PetscMalloc1(n, &baij->v_loc));
323: baij->n_loc = n;
324: }
325: in_loc = baij->in_loc;
326: v_loc = baij->v_loc;
327: }
329: for (i = 0; i < m; i++) {
330: if (im[i] < 0) continue;
331: PetscCheck(im[i] < mat->rmap->N, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Row too large: row %" PetscInt_FMT " max %" PetscInt_FMT, im[i], mat->rmap->N - 1);
332: if (im[i] >= rstart_orig && im[i] < rend_orig) { /* this processor entry */
333: row = im[i] - rstart_orig; /* local row index */
334: for (j = 0; j < n; j++) {
335: if (im[i] / bs > in[j] / bs) {
336: if (a->ignore_ltriangular) {
337: continue; /* ignore lower triangular blocks */
338: } 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)");
339: }
340: if (in[j] >= cstart_orig && in[j] < cend_orig) { /* diag entry (A) */
341: col = in[j] - cstart_orig; /* local col index */
342: brow = row / bs;
343: bcol = col / bs;
344: if (brow > bcol) continue; /* ignore lower triangular blocks of A */
345: if (roworiented) value = v[i * n + j];
346: else value = v[i + j * m];
347: MatSetValues_SeqSBAIJ_A_Private(row, col, value, addv, im[i], in[j]);
348: /* PetscCall(MatSetValues_SeqBAIJ(baij->A,1,&row,1,&col,&value,addv)); */
349: } else if (in[j] < 0) {
350: continue;
351: } else {
352: PetscCheck(in[j] < mat->cmap->N, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Column too large: col %" PetscInt_FMT " max %" PetscInt_FMT, in[j], mat->cmap->N - 1);
353: /* off-diag entry (B) */
354: if (mat->was_assembled) {
355: if (!baij->colmap) PetscCall(MatCreateColmap_MPIBAIJ_Private(mat));
356: #if defined(PETSC_USE_CTABLE)
357: PetscCall(PetscHMapIGetWithDefault(baij->colmap, in[j] / bs + 1, 0, &col));
358: col = col - 1;
359: #else
360: col = baij->colmap[in[j] / bs] - 1;
361: #endif
362: if (col < 0 && !((Mat_SeqSBAIJ *)baij->A->data)->nonew) {
363: PetscCall(MatDisAssemble_MPISBAIJ(mat));
364: col = in[j];
365: /* Reinitialize the variables required by MatSetValues_SeqBAIJ_B_Private() */
366: B = baij->B;
367: b = (Mat_SeqBAIJ *)(B)->data;
368: bimax = b->imax;
369: bi = b->i;
370: bilen = b->ilen;
371: bj = b->j;
372: ba = b->a;
373: } else col += in[j] % bs;
374: } else col = in[j];
375: if (roworiented) value = v[i * n + j];
376: else value = v[i + j * m];
377: MatSetValues_SeqSBAIJ_B_Private(row, col, value, addv, im[i], in[j]);
378: /* PetscCall(MatSetValues_SeqBAIJ(baij->B,1,&row,1,&col,&value,addv)); */
379: }
380: }
381: } else { /* off processor entry */
382: PetscCheck(!mat->nooffprocentries, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONG, "Setting off process row %" PetscInt_FMT " even though MatSetOption(,MAT_NO_OFF_PROC_ENTRIES,PETSC_TRUE) was set", im[i]);
383: if (!baij->donotstash) {
384: mat->assembled = PETSC_FALSE;
385: n_loc = 0;
386: for (j = 0; j < n; j++) {
387: if (im[i] / bs > in[j] / bs) continue; /* ignore lower triangular blocks */
388: in_loc[n_loc] = in[j];
389: if (roworiented) {
390: v_loc[n_loc] = v[i * n + j];
391: } else {
392: v_loc[n_loc] = v[j * m + i];
393: }
394: n_loc++;
395: }
396: PetscCall(MatStashValuesRow_Private(&mat->stash, im[i], n_loc, in_loc, v_loc, PETSC_FALSE));
397: }
398: }
399: }
400: PetscFunctionReturn(PETSC_SUCCESS);
401: }
403: static inline PetscErrorCode MatSetValuesBlocked_SeqSBAIJ_Inlined(Mat A, PetscInt row, PetscInt col, const PetscScalar v[], InsertMode is, PetscInt orow, PetscInt ocol)
404: {
405: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ *)A->data;
406: PetscInt *rp, low, high, t, ii, jj, nrow, i, rmax, N;
407: PetscInt *imax = a->imax, *ai = a->i, *ailen = a->ilen;
408: PetscInt *aj = a->j, nonew = a->nonew, bs2 = a->bs2, bs = A->rmap->bs;
409: PetscBool roworiented = a->roworiented;
410: const PetscScalar *value = v;
411: MatScalar *ap, *aa = a->a, *bap;
413: PetscFunctionBegin;
414: if (col < row) {
415: if (a->ignore_ltriangular) PetscFunctionReturn(PETSC_SUCCESS); /* ignore lower triangular block */
416: 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)");
417: }
418: rp = aj + ai[row];
419: ap = aa + bs2 * ai[row];
420: rmax = imax[row];
421: nrow = ailen[row];
422: value = v;
423: low = 0;
424: high = nrow;
426: while (high - low > 7) {
427: t = (low + high) / 2;
428: if (rp[t] > col) high = t;
429: else low = t;
430: }
431: for (i = low; i < high; i++) {
432: if (rp[i] > col) break;
433: if (rp[i] == col) {
434: bap = ap + bs2 * i;
435: if (roworiented) {
436: if (is == ADD_VALUES) {
437: for (ii = 0; ii < bs; ii++) {
438: for (jj = ii; jj < bs2; jj += bs) bap[jj] += *value++;
439: }
440: } else {
441: for (ii = 0; ii < bs; ii++) {
442: for (jj = ii; jj < bs2; jj += bs) bap[jj] = *value++;
443: }
444: }
445: } else {
446: if (is == ADD_VALUES) {
447: for (ii = 0; ii < bs; ii++) {
448: for (jj = 0; jj < bs; jj++) *bap++ += *value++;
449: }
450: } else {
451: for (ii = 0; ii < bs; ii++) {
452: for (jj = 0; jj < bs; jj++) *bap++ = *value++;
453: }
454: }
455: }
456: goto noinsert2;
457: }
458: }
459: if (nonew == 1) goto noinsert2;
460: PetscCheck(nonew != -1, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Inserting a new block index nonzero block (%" PetscInt_FMT ", %" PetscInt_FMT ") in the matrix", orow, ocol);
461: MatSeqXAIJReallocateAIJ(A, a->mbs, bs2, nrow, row, col, rmax, aa, ai, aj, rp, ap, imax, nonew, MatScalar);
462: N = nrow++ - 1;
463: high++;
464: /* shift up all the later entries in this row */
465: PetscCall(PetscArraymove(rp + i + 1, rp + i, N - i + 1));
466: PetscCall(PetscArraymove(ap + bs2 * (i + 1), ap + bs2 * i, bs2 * (N - i + 1)));
467: rp[i] = col;
468: bap = ap + bs2 * i;
469: if (roworiented) {
470: for (ii = 0; ii < bs; ii++) {
471: for (jj = ii; jj < bs2; jj += bs) bap[jj] = *value++;
472: }
473: } else {
474: for (ii = 0; ii < bs; ii++) {
475: for (jj = 0; jj < bs; jj++) *bap++ = *value++;
476: }
477: }
478: noinsert2:;
479: ailen[row] = nrow;
480: PetscFunctionReturn(PETSC_SUCCESS);
481: }
483: /*
484: This routine is exactly duplicated in mpibaij.c
485: */
486: static inline PetscErrorCode MatSetValuesBlocked_SeqBAIJ_Inlined(Mat A, PetscInt row, PetscInt col, const PetscScalar v[], InsertMode is, PetscInt orow, PetscInt ocol)
487: {
488: Mat_SeqBAIJ *a = (Mat_SeqBAIJ *)A->data;
489: PetscInt *rp, low, high, t, ii, jj, nrow, i, rmax, N;
490: PetscInt *imax = a->imax, *ai = a->i, *ailen = a->ilen;
491: PetscInt *aj = a->j, nonew = a->nonew, bs2 = a->bs2, bs = A->rmap->bs;
492: PetscBool roworiented = a->roworiented;
493: const PetscScalar *value = v;
494: MatScalar *ap, *aa = a->a, *bap;
496: PetscFunctionBegin;
497: rp = aj + ai[row];
498: ap = aa + bs2 * ai[row];
499: rmax = imax[row];
500: nrow = ailen[row];
501: low = 0;
502: high = nrow;
503: value = v;
504: while (high - low > 7) {
505: t = (low + high) / 2;
506: if (rp[t] > col) high = t;
507: else low = t;
508: }
509: for (i = low; i < high; i++) {
510: if (rp[i] > col) break;
511: if (rp[i] == col) {
512: bap = ap + bs2 * i;
513: if (roworiented) {
514: if (is == ADD_VALUES) {
515: for (ii = 0; ii < bs; ii++) {
516: for (jj = ii; jj < bs2; jj += bs) bap[jj] += *value++;
517: }
518: } else {
519: for (ii = 0; ii < bs; ii++) {
520: for (jj = ii; jj < bs2; jj += bs) bap[jj] = *value++;
521: }
522: }
523: } else {
524: if (is == ADD_VALUES) {
525: for (ii = 0; ii < bs; ii++, value += bs) {
526: for (jj = 0; jj < bs; jj++) bap[jj] += value[jj];
527: bap += bs;
528: }
529: } else {
530: for (ii = 0; ii < bs; ii++, value += bs) {
531: for (jj = 0; jj < bs; jj++) bap[jj] = value[jj];
532: bap += bs;
533: }
534: }
535: }
536: goto noinsert2;
537: }
538: }
539: if (nonew == 1) goto noinsert2;
540: PetscCheck(nonew != -1, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Inserting a new global block indexed nonzero block (%" PetscInt_FMT ", %" PetscInt_FMT ") in the matrix", orow, ocol);
541: MatSeqXAIJReallocateAIJ(A, a->mbs, bs2, nrow, row, col, rmax, aa, ai, aj, rp, ap, imax, nonew, MatScalar);
542: N = nrow++ - 1;
543: high++;
544: /* shift up all the later entries in this row */
545: PetscCall(PetscArraymove(rp + i + 1, rp + i, N - i + 1));
546: PetscCall(PetscArraymove(ap + bs2 * (i + 1), ap + bs2 * i, bs2 * (N - i + 1)));
547: rp[i] = col;
548: bap = ap + bs2 * i;
549: if (roworiented) {
550: for (ii = 0; ii < bs; ii++) {
551: for (jj = ii; jj < bs2; jj += bs) bap[jj] = *value++;
552: }
553: } else {
554: for (ii = 0; ii < bs; ii++) {
555: for (jj = 0; jj < bs; jj++) *bap++ = *value++;
556: }
557: }
558: noinsert2:;
559: ailen[row] = nrow;
560: PetscFunctionReturn(PETSC_SUCCESS);
561: }
563: /*
564: This routine could be optimized by removing the need for the block copy below and passing stride information
565: to the above inline routines; similarly in MatSetValuesBlocked_MPIBAIJ()
566: */
567: static PetscErrorCode MatSetValuesBlocked_MPISBAIJ(Mat mat, PetscInt m, const PetscInt im[], PetscInt n, const PetscInt in[], const MatScalar v[], InsertMode addv)
568: {
569: Mat_MPISBAIJ *baij = (Mat_MPISBAIJ *)mat->data;
570: const MatScalar *value;
571: MatScalar *barray = baij->barray;
572: PetscBool roworiented = baij->roworiented, ignore_ltriangular = ((Mat_SeqSBAIJ *)baij->A->data)->ignore_ltriangular;
573: PetscInt i, j, ii, jj, row, col, rstart = baij->rstartbs;
574: PetscInt rend = baij->rendbs, cstart = baij->cstartbs, stepval;
575: PetscInt cend = baij->cendbs, bs = mat->rmap->bs, bs2 = baij->bs2;
577: PetscFunctionBegin;
578: if (!barray) {
579: PetscCall(PetscMalloc1(bs2, &barray));
580: baij->barray = barray;
581: }
583: if (roworiented) {
584: stepval = (n - 1) * bs;
585: } else {
586: stepval = (m - 1) * bs;
587: }
588: for (i = 0; i < m; i++) {
589: if (im[i] < 0) continue;
590: PetscCheck(im[i] < baij->Mbs, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Block indexed row too large %" PetscInt_FMT " max %" PetscInt_FMT, im[i], baij->Mbs - 1);
591: if (im[i] >= rstart && im[i] < rend) {
592: row = im[i] - rstart;
593: for (j = 0; j < n; j++) {
594: if (im[i] > in[j]) {
595: if (ignore_ltriangular) continue; /* ignore lower triangular blocks */
596: 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)");
597: }
598: /* If NumCol = 1 then a copy is not required */
599: if ((roworiented) && (n == 1)) {
600: barray = (MatScalar *)v + i * bs2;
601: } else if ((!roworiented) && (m == 1)) {
602: barray = (MatScalar *)v + j * bs2;
603: } else { /* Here a copy is required */
604: if (roworiented) {
605: value = v + i * (stepval + bs) * bs + j * bs;
606: } else {
607: value = v + j * (stepval + bs) * bs + i * bs;
608: }
609: for (ii = 0; ii < bs; ii++, value += stepval) {
610: for (jj = 0; jj < bs; jj++) *barray++ = *value++;
611: }
612: barray -= bs2;
613: }
615: if (in[j] >= cstart && in[j] < cend) {
616: col = in[j] - cstart;
617: PetscCall(MatSetValuesBlocked_SeqSBAIJ_Inlined(baij->A, row, col, barray, addv, im[i], in[j]));
618: } else if (in[j] < 0) {
619: continue;
620: } else {
621: PetscCheck(in[j] < baij->Nbs, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Block indexed column too large %" PetscInt_FMT " max %" PetscInt_FMT, in[j], baij->Nbs - 1);
622: if (mat->was_assembled) {
623: if (!baij->colmap) PetscCall(MatCreateColmap_MPIBAIJ_Private(mat));
625: #if defined(PETSC_USE_DEBUG)
626: #if defined(PETSC_USE_CTABLE)
627: {
628: PetscInt data;
629: PetscCall(PetscHMapIGetWithDefault(baij->colmap, in[j] + 1, 0, &data));
630: PetscCheck((data - 1) % bs == 0, PETSC_COMM_SELF, PETSC_ERR_PLIB, "Incorrect colmap");
631: }
632: #else
633: PetscCheck((baij->colmap[in[j]] - 1) % bs == 0, PETSC_COMM_SELF, PETSC_ERR_PLIB, "Incorrect colmap");
634: #endif
635: #endif
636: #if defined(PETSC_USE_CTABLE)
637: PetscCall(PetscHMapIGetWithDefault(baij->colmap, in[j] + 1, 0, &col));
638: col = (col - 1) / bs;
639: #else
640: col = (baij->colmap[in[j]] - 1) / bs;
641: #endif
642: if (col < 0 && !((Mat_SeqBAIJ *)baij->A->data)->nonew) {
643: PetscCall(MatDisAssemble_MPISBAIJ(mat));
644: col = in[j];
645: }
646: } else col = in[j];
647: PetscCall(MatSetValuesBlocked_SeqBAIJ_Inlined(baij->B, row, col, barray, addv, im[i], in[j]));
648: }
649: }
650: } else {
651: PetscCheck(!mat->nooffprocentries, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONG, "Setting off process block indexed row %" PetscInt_FMT " even though MatSetOption(,MAT_NO_OFF_PROC_ENTRIES,PETSC_TRUE) was set", im[i]);
652: if (!baij->donotstash) {
653: if (roworiented) {
654: PetscCall(MatStashValuesRowBlocked_Private(&mat->bstash, im[i], n, in, v, m, n, i));
655: } else {
656: PetscCall(MatStashValuesColBlocked_Private(&mat->bstash, im[i], n, in, v, m, n, i));
657: }
658: }
659: }
660: }
661: PetscFunctionReturn(PETSC_SUCCESS);
662: }
664: static PetscErrorCode MatGetValues_MPISBAIJ(Mat mat, PetscInt m, const PetscInt idxm[], PetscInt n, const PetscInt idxn[], PetscScalar v[])
665: {
666: Mat_MPISBAIJ *baij = (Mat_MPISBAIJ *)mat->data;
667: PetscInt bs = mat->rmap->bs, i, j, bsrstart = mat->rmap->rstart, bsrend = mat->rmap->rend;
668: PetscInt bscstart = mat->cmap->rstart, bscend = mat->cmap->rend, row, col, data;
670: PetscFunctionBegin;
671: for (i = 0; i < m; i++) {
672: if (idxm[i] < 0) continue; /* negative row */
673: PetscCheck(idxm[i] < mat->rmap->N, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Row too large: row %" PetscInt_FMT " max %" PetscInt_FMT, idxm[i], mat->rmap->N - 1);
674: if (idxm[i] >= bsrstart && idxm[i] < bsrend) {
675: row = idxm[i] - bsrstart;
676: for (j = 0; j < n; j++) {
677: if (idxn[j] < 0) continue; /* negative column */
678: PetscCheck(idxn[j] < mat->cmap->N, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Column too large: col %" PetscInt_FMT " max %" PetscInt_FMT, idxn[j], mat->cmap->N - 1);
679: if (idxn[j] >= bscstart && idxn[j] < bscend) {
680: col = idxn[j] - bscstart;
681: PetscCall(MatGetValues_SeqSBAIJ(baij->A, 1, &row, 1, &col, v + i * n + j));
682: } else {
683: if (!baij->colmap) PetscCall(MatCreateColmap_MPIBAIJ_Private(mat));
684: #if defined(PETSC_USE_CTABLE)
685: PetscCall(PetscHMapIGetWithDefault(baij->colmap, idxn[j] / bs + 1, 0, &data));
686: data--;
687: #else
688: data = baij->colmap[idxn[j] / bs] - 1;
689: #endif
690: if ((data < 0) || (baij->garray[data / bs] != idxn[j] / bs)) *(v + i * n + j) = 0.0;
691: else {
692: col = data + idxn[j] % bs;
693: PetscCall(MatGetValues_SeqBAIJ(baij->B, 1, &row, 1, &col, v + i * n + j));
694: }
695: }
696: }
697: } else SETERRQ(PETSC_COMM_SELF, PETSC_ERR_SUP, "Only local values currently supported");
698: }
699: PetscFunctionReturn(PETSC_SUCCESS);
700: }
702: static PetscErrorCode MatNorm_MPISBAIJ(Mat mat, NormType type, PetscReal *norm)
703: {
704: Mat_MPISBAIJ *baij = (Mat_MPISBAIJ *)mat->data;
705: PetscReal sum[2], *lnorm2;
707: PetscFunctionBegin;
708: if (baij->size == 1) {
709: PetscCall(MatNorm(baij->A, type, norm));
710: } else {
711: if (type == NORM_FROBENIUS) {
712: PetscCall(PetscMalloc1(2, &lnorm2));
713: PetscCall(MatNorm(baij->A, type, lnorm2));
714: *lnorm2 = (*lnorm2) * (*lnorm2);
715: lnorm2++; /* squar power of norm(A) */
716: PetscCall(MatNorm(baij->B, type, lnorm2));
717: *lnorm2 = (*lnorm2) * (*lnorm2);
718: lnorm2--; /* squar power of norm(B) */
719: PetscCall(MPIU_Allreduce(lnorm2, sum, 2, MPIU_REAL, MPIU_SUM, PetscObjectComm((PetscObject)mat)));
720: *norm = PetscSqrtReal(sum[0] + 2 * sum[1]);
721: PetscCall(PetscFree(lnorm2));
722: } else if (type == NORM_INFINITY || type == NORM_1) { /* max row/column sum */
723: Mat_SeqSBAIJ *amat = (Mat_SeqSBAIJ *)baij->A->data;
724: Mat_SeqBAIJ *bmat = (Mat_SeqBAIJ *)baij->B->data;
725: PetscReal *rsum, *rsum2, vabs;
726: PetscInt *jj, *garray = baij->garray, rstart = baij->rstartbs, nz;
727: PetscInt brow, bcol, col, bs = baij->A->rmap->bs, row, grow, gcol, mbs = amat->mbs;
728: MatScalar *v;
730: PetscCall(PetscMalloc2(mat->cmap->N, &rsum, mat->cmap->N, &rsum2));
731: PetscCall(PetscArrayzero(rsum, mat->cmap->N));
732: /* Amat */
733: v = amat->a;
734: jj = amat->j;
735: for (brow = 0; brow < mbs; brow++) {
736: grow = bs * (rstart + brow);
737: nz = amat->i[brow + 1] - amat->i[brow];
738: for (bcol = 0; bcol < nz; bcol++) {
739: gcol = bs * (rstart + *jj);
740: jj++;
741: for (col = 0; col < bs; col++) {
742: for (row = 0; row < bs; row++) {
743: vabs = PetscAbsScalar(*v);
744: v++;
745: rsum[gcol + col] += vabs;
746: /* non-diagonal block */
747: if (bcol > 0 && vabs > 0.0) rsum[grow + row] += vabs;
748: }
749: }
750: }
751: PetscCall(PetscLogFlops(nz * bs * bs));
752: }
753: /* Bmat */
754: v = bmat->a;
755: jj = bmat->j;
756: for (brow = 0; brow < mbs; brow++) {
757: grow = bs * (rstart + brow);
758: nz = bmat->i[brow + 1] - bmat->i[brow];
759: for (bcol = 0; bcol < nz; bcol++) {
760: gcol = bs * garray[*jj];
761: jj++;
762: for (col = 0; col < bs; col++) {
763: for (row = 0; row < bs; row++) {
764: vabs = PetscAbsScalar(*v);
765: v++;
766: rsum[gcol + col] += vabs;
767: rsum[grow + row] += vabs;
768: }
769: }
770: }
771: PetscCall(PetscLogFlops(nz * bs * bs));
772: }
773: PetscCall(MPIU_Allreduce(rsum, rsum2, mat->cmap->N, MPIU_REAL, MPIU_SUM, PetscObjectComm((PetscObject)mat)));
774: *norm = 0.0;
775: for (col = 0; col < mat->cmap->N; col++) {
776: if (rsum2[col] > *norm) *norm = rsum2[col];
777: }
778: PetscCall(PetscFree2(rsum, rsum2));
779: } else SETERRQ(PETSC_COMM_SELF, PETSC_ERR_SUP, "No support for this norm yet");
780: }
781: PetscFunctionReturn(PETSC_SUCCESS);
782: }
784: static PetscErrorCode MatAssemblyBegin_MPISBAIJ(Mat mat, MatAssemblyType mode)
785: {
786: Mat_MPISBAIJ *baij = (Mat_MPISBAIJ *)mat->data;
787: PetscInt nstash, reallocs;
789: PetscFunctionBegin;
790: if (baij->donotstash || mat->nooffprocentries) PetscFunctionReturn(PETSC_SUCCESS);
792: PetscCall(MatStashScatterBegin_Private(mat, &mat->stash, mat->rmap->range));
793: PetscCall(MatStashScatterBegin_Private(mat, &mat->bstash, baij->rangebs));
794: PetscCall(MatStashGetInfo_Private(&mat->stash, &nstash, &reallocs));
795: PetscCall(PetscInfo(mat, "Stash has %" PetscInt_FMT " entries,uses %" PetscInt_FMT " mallocs.\n", nstash, reallocs));
796: PetscCall(MatStashGetInfo_Private(&mat->stash, &nstash, &reallocs));
797: PetscCall(PetscInfo(mat, "Block-Stash has %" PetscInt_FMT " entries, uses %" PetscInt_FMT " mallocs.\n", nstash, reallocs));
798: PetscFunctionReturn(PETSC_SUCCESS);
799: }
801: static PetscErrorCode MatAssemblyEnd_MPISBAIJ(Mat mat, MatAssemblyType mode)
802: {
803: Mat_MPISBAIJ *baij = (Mat_MPISBAIJ *)mat->data;
804: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ *)baij->A->data;
805: PetscInt i, j, rstart, ncols, flg, bs2 = baij->bs2;
806: PetscInt *row, *col;
807: PetscBool other_disassembled;
808: PetscMPIInt n;
809: PetscBool r1, r2, r3;
810: MatScalar *val;
812: /* do not use 'b=(Mat_SeqBAIJ*)baij->B->data' as B can be reset in disassembly */
813: PetscFunctionBegin;
814: if (!baij->donotstash && !mat->nooffprocentries) {
815: while (1) {
816: PetscCall(MatStashScatterGetMesg_Private(&mat->stash, &n, &row, &col, &val, &flg));
817: if (!flg) break;
819: for (i = 0; i < n;) {
820: /* Now identify the consecutive vals belonging to the same row */
821: for (j = i, rstart = row[j]; j < n; j++) {
822: if (row[j] != rstart) break;
823: }
824: if (j < n) ncols = j - i;
825: else ncols = n - i;
826: /* Now assemble all these values with a single function call */
827: PetscCall(MatSetValues_MPISBAIJ(mat, 1, row + i, ncols, col + i, val + i, mat->insertmode));
828: i = j;
829: }
830: }
831: PetscCall(MatStashScatterEnd_Private(&mat->stash));
832: /* Now process the block-stash. Since the values are stashed column-oriented,
833: set the row-oriented flag to column-oriented, and after MatSetValues()
834: restore the original flags */
835: r1 = baij->roworiented;
836: r2 = a->roworiented;
837: r3 = ((Mat_SeqBAIJ *)baij->B->data)->roworiented;
839: baij->roworiented = PETSC_FALSE;
840: a->roworiented = PETSC_FALSE;
842: ((Mat_SeqBAIJ *)baij->B->data)->roworiented = PETSC_FALSE; /* b->roworinted */
843: while (1) {
844: PetscCall(MatStashScatterGetMesg_Private(&mat->bstash, &n, &row, &col, &val, &flg));
845: if (!flg) break;
847: for (i = 0; i < n;) {
848: /* Now identify the consecutive vals belonging to the same row */
849: for (j = i, rstart = row[j]; j < n; j++) {
850: if (row[j] != rstart) break;
851: }
852: if (j < n) ncols = j - i;
853: else ncols = n - i;
854: PetscCall(MatSetValuesBlocked_MPISBAIJ(mat, 1, row + i, ncols, col + i, val + i * bs2, mat->insertmode));
855: i = j;
856: }
857: }
858: PetscCall(MatStashScatterEnd_Private(&mat->bstash));
860: baij->roworiented = r1;
861: a->roworiented = r2;
863: ((Mat_SeqBAIJ *)baij->B->data)->roworiented = r3; /* b->roworinted */
864: }
866: PetscCall(MatAssemblyBegin(baij->A, mode));
867: PetscCall(MatAssemblyEnd(baij->A, mode));
869: /* determine if any processor has disassembled, if so we must
870: also disassemble ourselves, in order that we may reassemble. */
871: /*
872: if nonzero structure of submatrix B cannot change then we know that
873: no processor disassembled thus we can skip this stuff
874: */
875: if (!((Mat_SeqBAIJ *)baij->B->data)->nonew) {
876: PetscCall(MPIU_Allreduce(&mat->was_assembled, &other_disassembled, 1, MPIU_BOOL, MPI_LAND, PetscObjectComm((PetscObject)mat)));
877: if (mat->was_assembled && !other_disassembled) PetscCall(MatDisAssemble_MPISBAIJ(mat));
878: }
880: if (!mat->was_assembled && mode == MAT_FINAL_ASSEMBLY) { PetscCall(MatSetUpMultiply_MPISBAIJ(mat)); /* setup Mvctx and sMvctx */ }
881: PetscCall(MatAssemblyBegin(baij->B, mode));
882: PetscCall(MatAssemblyEnd(baij->B, mode));
884: PetscCall(PetscFree2(baij->rowvalues, baij->rowindices));
886: baij->rowvalues = NULL;
888: /* if no new nonzero locations are allowed in matrix then only set the matrix state the first time through */
889: if ((!mat->was_assembled && mode == MAT_FINAL_ASSEMBLY) || !((Mat_SeqBAIJ *)baij->A->data)->nonew) {
890: PetscObjectState state = baij->A->nonzerostate + baij->B->nonzerostate;
891: PetscCall(MPIU_Allreduce(&state, &mat->nonzerostate, 1, MPIU_INT64, MPI_SUM, PetscObjectComm((PetscObject)mat)));
892: }
893: PetscFunctionReturn(PETSC_SUCCESS);
894: }
896: extern PetscErrorCode MatSetValues_MPIBAIJ(Mat, PetscInt, const PetscInt[], PetscInt, const PetscInt[], const PetscScalar[], InsertMode);
897: #include <petscdraw.h>
898: static PetscErrorCode MatView_MPISBAIJ_ASCIIorDraworSocket(Mat mat, PetscViewer viewer)
899: {
900: Mat_MPISBAIJ *baij = (Mat_MPISBAIJ *)mat->data;
901: PetscInt bs = mat->rmap->bs;
902: PetscMPIInt rank = baij->rank;
903: PetscBool iascii, isdraw;
904: PetscViewer sviewer;
905: PetscViewerFormat format;
907: PetscFunctionBegin;
908: PetscCall(PetscObjectTypeCompare((PetscObject)viewer, PETSCVIEWERASCII, &iascii));
909: PetscCall(PetscObjectTypeCompare((PetscObject)viewer, PETSCVIEWERDRAW, &isdraw));
910: if (iascii) {
911: PetscCall(PetscViewerGetFormat(viewer, &format));
912: if (format == PETSC_VIEWER_ASCII_INFO_DETAIL) {
913: MatInfo info;
914: PetscCallMPI(MPI_Comm_rank(PetscObjectComm((PetscObject)mat), &rank));
915: PetscCall(MatGetInfo(mat, MAT_LOCAL, &info));
916: PetscCall(PetscViewerASCIIPushSynchronized(viewer));
917: PetscCall(PetscViewerASCIISynchronizedPrintf(viewer, "[%d] Local rows %" PetscInt_FMT " nz %" PetscInt_FMT " nz alloced %" PetscInt_FMT " bs %" PetscInt_FMT " mem %g\n", rank, mat->rmap->n, (PetscInt)info.nz_used, (PetscInt)info.nz_allocated,
918: mat->rmap->bs, (double)info.memory));
919: PetscCall(MatGetInfo(baij->A, MAT_LOCAL, &info));
920: PetscCall(PetscViewerASCIISynchronizedPrintf(viewer, "[%d] on-diagonal part: nz %" PetscInt_FMT " \n", rank, (PetscInt)info.nz_used));
921: PetscCall(MatGetInfo(baij->B, MAT_LOCAL, &info));
922: PetscCall(PetscViewerASCIISynchronizedPrintf(viewer, "[%d] off-diagonal part: nz %" PetscInt_FMT " \n", rank, (PetscInt)info.nz_used));
923: PetscCall(PetscViewerFlush(viewer));
924: PetscCall(PetscViewerASCIIPopSynchronized(viewer));
925: PetscCall(PetscViewerASCIIPrintf(viewer, "Information on VecScatter used in matrix-vector product: \n"));
926: PetscCall(VecScatterView(baij->Mvctx, viewer));
927: PetscFunctionReturn(PETSC_SUCCESS);
928: } else if (format == PETSC_VIEWER_ASCII_INFO) {
929: PetscCall(PetscViewerASCIIPrintf(viewer, " block size is %" PetscInt_FMT "\n", bs));
930: PetscFunctionReturn(PETSC_SUCCESS);
931: } else if (format == PETSC_VIEWER_ASCII_FACTOR_INFO) {
932: PetscFunctionReturn(PETSC_SUCCESS);
933: }
934: }
936: if (isdraw) {
937: PetscDraw draw;
938: PetscBool isnull;
939: PetscCall(PetscViewerDrawGetDraw(viewer, 0, &draw));
940: PetscCall(PetscDrawIsNull(draw, &isnull));
941: if (isnull) PetscFunctionReturn(PETSC_SUCCESS);
942: }
944: {
945: /* assemble the entire matrix onto first processor. */
946: Mat A;
947: Mat_SeqSBAIJ *Aloc;
948: Mat_SeqBAIJ *Bloc;
949: PetscInt M = mat->rmap->N, N = mat->cmap->N, *ai, *aj, col, i, j, k, *rvals, mbs = baij->mbs;
950: MatScalar *a;
951: const char *matname;
953: /* Should this be the same type as mat? */
954: PetscCall(MatCreate(PetscObjectComm((PetscObject)mat), &A));
955: if (rank == 0) {
956: PetscCall(MatSetSizes(A, M, N, M, N));
957: } else {
958: PetscCall(MatSetSizes(A, 0, 0, M, N));
959: }
960: PetscCall(MatSetType(A, MATMPISBAIJ));
961: PetscCall(MatMPISBAIJSetPreallocation(A, mat->rmap->bs, 0, NULL, 0, NULL));
962: PetscCall(MatSetOption(A, MAT_NEW_NONZERO_LOCATION_ERR, PETSC_FALSE));
964: /* copy over the A part */
965: Aloc = (Mat_SeqSBAIJ *)baij->A->data;
966: ai = Aloc->i;
967: aj = Aloc->j;
968: a = Aloc->a;
969: PetscCall(PetscMalloc1(bs, &rvals));
971: for (i = 0; i < mbs; i++) {
972: rvals[0] = bs * (baij->rstartbs + i);
973: for (j = 1; j < bs; j++) rvals[j] = rvals[j - 1] + 1;
974: for (j = ai[i]; j < ai[i + 1]; j++) {
975: col = (baij->cstartbs + aj[j]) * bs;
976: for (k = 0; k < bs; k++) {
977: PetscCall(MatSetValues_MPISBAIJ(A, bs, rvals, 1, &col, a, INSERT_VALUES));
978: col++;
979: a += bs;
980: }
981: }
982: }
983: /* copy over the B part */
984: Bloc = (Mat_SeqBAIJ *)baij->B->data;
985: ai = Bloc->i;
986: aj = Bloc->j;
987: a = Bloc->a;
988: for (i = 0; i < mbs; i++) {
989: rvals[0] = bs * (baij->rstartbs + i);
990: for (j = 1; j < bs; j++) rvals[j] = rvals[j - 1] + 1;
991: for (j = ai[i]; j < ai[i + 1]; j++) {
992: col = baij->garray[aj[j]] * bs;
993: for (k = 0; k < bs; k++) {
994: PetscCall(MatSetValues_MPIBAIJ(A, bs, rvals, 1, &col, a, INSERT_VALUES));
995: col++;
996: a += bs;
997: }
998: }
999: }
1000: PetscCall(PetscFree(rvals));
1001: PetscCall(MatAssemblyBegin(A, MAT_FINAL_ASSEMBLY));
1002: PetscCall(MatAssemblyEnd(A, MAT_FINAL_ASSEMBLY));
1003: /*
1004: Everyone has to call to draw the matrix since the graphics waits are
1005: synchronized across all processors that share the PetscDraw object
1006: */
1007: PetscCall(PetscViewerGetSubViewer(viewer, PETSC_COMM_SELF, &sviewer));
1008: if (((PetscObject)mat)->name) PetscCall(PetscObjectGetName((PetscObject)mat, &matname));
1009: if (rank == 0) {
1010: if (((PetscObject)mat)->name) PetscCall(PetscObjectSetName((PetscObject)((Mat_MPISBAIJ *)A->data)->A, matname));
1011: PetscCall(MatView_SeqSBAIJ(((Mat_MPISBAIJ *)A->data)->A, sviewer));
1012: }
1013: PetscCall(PetscViewerRestoreSubViewer(viewer, PETSC_COMM_SELF, &sviewer));
1014: PetscCall(MatDestroy(&A));
1015: }
1016: PetscFunctionReturn(PETSC_SUCCESS);
1017: }
1019: /* Used for both MPIBAIJ and MPISBAIJ matrices */
1020: #define MatView_MPISBAIJ_Binary MatView_MPIBAIJ_Binary
1022: static PetscErrorCode MatView_MPISBAIJ(Mat mat, PetscViewer viewer)
1023: {
1024: PetscBool iascii, isdraw, issocket, isbinary;
1026: PetscFunctionBegin;
1027: PetscCall(PetscObjectTypeCompare((PetscObject)viewer, PETSCVIEWERASCII, &iascii));
1028: PetscCall(PetscObjectTypeCompare((PetscObject)viewer, PETSCVIEWERDRAW, &isdraw));
1029: PetscCall(PetscObjectTypeCompare((PetscObject)viewer, PETSCVIEWERSOCKET, &issocket));
1030: PetscCall(PetscObjectTypeCompare((PetscObject)viewer, PETSCVIEWERBINARY, &isbinary));
1031: if (iascii || isdraw || issocket) {
1032: PetscCall(MatView_MPISBAIJ_ASCIIorDraworSocket(mat, viewer));
1033: } else if (isbinary) PetscCall(MatView_MPISBAIJ_Binary(mat, viewer));
1034: PetscFunctionReturn(PETSC_SUCCESS);
1035: }
1037: #if defined(PETSC_USE_COMPLEX)
1038: static PetscErrorCode MatMult_MPISBAIJ_Hermitian(Mat A, Vec xx, Vec yy)
1039: {
1040: Mat_MPISBAIJ *a = (Mat_MPISBAIJ *)A->data;
1041: PetscInt mbs = a->mbs, bs = A->rmap->bs;
1042: PetscScalar *from;
1043: const PetscScalar *x;
1045: PetscFunctionBegin;
1046: /* diagonal part */
1047: PetscCall((*a->A->ops->mult)(a->A, xx, a->slvec1a));
1048: /* since a->slvec1b shares memory (dangerously) with a->slec1 changes to a->slec1 will affect it */
1049: PetscCall(PetscObjectStateIncrease((PetscObject)a->slvec1b));
1050: PetscCall(VecZeroEntries(a->slvec1b));
1052: /* subdiagonal part */
1053: PetscCheck(a->B->ops->multhermitiantranspose, PetscObjectComm((PetscObject)a->B), PETSC_ERR_SUP, "Not for type %s", ((PetscObject)a->B)->type_name);
1054: PetscCall((*a->B->ops->multhermitiantranspose)(a->B, xx, a->slvec0b));
1056: /* copy x into the vec slvec0 */
1057: PetscCall(VecGetArray(a->slvec0, &from));
1058: PetscCall(VecGetArrayRead(xx, &x));
1060: PetscCall(PetscArraycpy(from, x, bs * mbs));
1061: PetscCall(VecRestoreArray(a->slvec0, &from));
1062: PetscCall(VecRestoreArrayRead(xx, &x));
1064: PetscCall(VecScatterBegin(a->sMvctx, a->slvec0, a->slvec1, ADD_VALUES, SCATTER_FORWARD));
1065: PetscCall(VecScatterEnd(a->sMvctx, a->slvec0, a->slvec1, ADD_VALUES, SCATTER_FORWARD));
1066: /* supperdiagonal part */
1067: PetscCall((*a->B->ops->multadd)(a->B, a->slvec1b, a->slvec1a, yy));
1068: PetscFunctionReturn(PETSC_SUCCESS);
1069: }
1070: #endif
1072: static PetscErrorCode MatMult_MPISBAIJ(Mat A, Vec xx, Vec yy)
1073: {
1074: Mat_MPISBAIJ *a = (Mat_MPISBAIJ *)A->data;
1075: PetscInt mbs = a->mbs, bs = A->rmap->bs;
1076: PetscScalar *from;
1077: const PetscScalar *x;
1079: PetscFunctionBegin;
1080: /* diagonal part */
1081: PetscCall((*a->A->ops->mult)(a->A, xx, a->slvec1a));
1082: /* since a->slvec1b shares memory (dangerously) with a->slec1 changes to a->slec1 will affect it */
1083: PetscCall(PetscObjectStateIncrease((PetscObject)a->slvec1b));
1084: PetscCall(VecZeroEntries(a->slvec1b));
1086: /* subdiagonal part */
1087: PetscCall((*a->B->ops->multtranspose)(a->B, xx, a->slvec0b));
1089: /* copy x into the vec slvec0 */
1090: PetscCall(VecGetArray(a->slvec0, &from));
1091: PetscCall(VecGetArrayRead(xx, &x));
1093: PetscCall(PetscArraycpy(from, x, bs * mbs));
1094: PetscCall(VecRestoreArray(a->slvec0, &from));
1095: PetscCall(VecRestoreArrayRead(xx, &x));
1097: PetscCall(VecScatterBegin(a->sMvctx, a->slvec0, a->slvec1, ADD_VALUES, SCATTER_FORWARD));
1098: PetscCall(VecScatterEnd(a->sMvctx, a->slvec0, a->slvec1, ADD_VALUES, SCATTER_FORWARD));
1099: /* supperdiagonal part */
1100: PetscCall((*a->B->ops->multadd)(a->B, a->slvec1b, a->slvec1a, yy));
1101: PetscFunctionReturn(PETSC_SUCCESS);
1102: }
1104: #if PetscDefined(USE_COMPLEX)
1105: static PetscErrorCode MatMultAdd_MPISBAIJ_Hermitian(Mat A, Vec xx, Vec yy, Vec zz)
1106: {
1107: Mat_MPISBAIJ *a = (Mat_MPISBAIJ *)A->data;
1108: PetscInt mbs = a->mbs, bs = A->rmap->bs;
1109: PetscScalar *from;
1110: const PetscScalar *x;
1112: PetscFunctionBegin;
1113: /* diagonal part */
1114: PetscCall((*a->A->ops->multadd)(a->A, xx, yy, a->slvec1a));
1115: PetscCall(PetscObjectStateIncrease((PetscObject)a->slvec1b));
1116: PetscCall(VecZeroEntries(a->slvec1b));
1118: /* subdiagonal part */
1119: PetscCheck(a->B->ops->multhermitiantranspose, PetscObjectComm((PetscObject)a->B), PETSC_ERR_SUP, "Not for type %s", ((PetscObject)a->B)->type_name);
1120: PetscCall((*a->B->ops->multhermitiantranspose)(a->B, xx, a->slvec0b));
1122: /* copy x into the vec slvec0 */
1123: PetscCall(VecGetArray(a->slvec0, &from));
1124: PetscCall(VecGetArrayRead(xx, &x));
1125: PetscCall(PetscArraycpy(from, x, bs * mbs));
1126: PetscCall(VecRestoreArray(a->slvec0, &from));
1128: PetscCall(VecScatterBegin(a->sMvctx, a->slvec0, a->slvec1, ADD_VALUES, SCATTER_FORWARD));
1129: PetscCall(VecRestoreArrayRead(xx, &x));
1130: PetscCall(VecScatterEnd(a->sMvctx, a->slvec0, a->slvec1, ADD_VALUES, SCATTER_FORWARD));
1132: /* supperdiagonal part */
1133: PetscCall((*a->B->ops->multadd)(a->B, a->slvec1b, a->slvec1a, zz));
1134: PetscFunctionReturn(PETSC_SUCCESS);
1135: }
1136: #endif
1138: static PetscErrorCode MatMultAdd_MPISBAIJ(Mat A, Vec xx, Vec yy, Vec zz)
1139: {
1140: Mat_MPISBAIJ *a = (Mat_MPISBAIJ *)A->data;
1141: PetscInt mbs = a->mbs, bs = A->rmap->bs;
1142: PetscScalar *from;
1143: const PetscScalar *x;
1145: PetscFunctionBegin;
1146: /* diagonal part */
1147: PetscCall((*a->A->ops->multadd)(a->A, xx, yy, a->slvec1a));
1148: PetscCall(PetscObjectStateIncrease((PetscObject)a->slvec1b));
1149: PetscCall(VecZeroEntries(a->slvec1b));
1151: /* subdiagonal part */
1152: PetscCall((*a->B->ops->multtranspose)(a->B, xx, a->slvec0b));
1154: /* copy x into the vec slvec0 */
1155: PetscCall(VecGetArray(a->slvec0, &from));
1156: PetscCall(VecGetArrayRead(xx, &x));
1157: PetscCall(PetscArraycpy(from, x, bs * mbs));
1158: PetscCall(VecRestoreArray(a->slvec0, &from));
1160: PetscCall(VecScatterBegin(a->sMvctx, a->slvec0, a->slvec1, ADD_VALUES, SCATTER_FORWARD));
1161: PetscCall(VecRestoreArrayRead(xx, &x));
1162: PetscCall(VecScatterEnd(a->sMvctx, a->slvec0, a->slvec1, ADD_VALUES, SCATTER_FORWARD));
1164: /* supperdiagonal part */
1165: PetscCall((*a->B->ops->multadd)(a->B, a->slvec1b, a->slvec1a, zz));
1166: PetscFunctionReturn(PETSC_SUCCESS);
1167: }
1169: /*
1170: This only works correctly for square matrices where the subblock A->A is the
1171: diagonal block
1172: */
1173: static PetscErrorCode MatGetDiagonal_MPISBAIJ(Mat A, Vec v)
1174: {
1175: Mat_MPISBAIJ *a = (Mat_MPISBAIJ *)A->data;
1177: PetscFunctionBegin;
1178: /* PetscCheck(a->rmap->N == a->cmap->N,PETSC_COMM_SELF,PETSC_ERR_SUP,"Supports only square matrix where A->A is diag block"); */
1179: PetscCall(MatGetDiagonal(a->A, v));
1180: PetscFunctionReturn(PETSC_SUCCESS);
1181: }
1183: static PetscErrorCode MatScale_MPISBAIJ(Mat A, PetscScalar aa)
1184: {
1185: Mat_MPISBAIJ *a = (Mat_MPISBAIJ *)A->data;
1187: PetscFunctionBegin;
1188: PetscCall(MatScale(a->A, aa));
1189: PetscCall(MatScale(a->B, aa));
1190: PetscFunctionReturn(PETSC_SUCCESS);
1191: }
1193: static PetscErrorCode MatGetRow_MPISBAIJ(Mat matin, PetscInt row, PetscInt *nz, PetscInt **idx, PetscScalar **v)
1194: {
1195: Mat_MPISBAIJ *mat = (Mat_MPISBAIJ *)matin->data;
1196: PetscScalar *vworkA, *vworkB, **pvA, **pvB, *v_p;
1197: PetscInt bs = matin->rmap->bs, bs2 = mat->bs2, i, *cworkA, *cworkB, **pcA, **pcB;
1198: PetscInt nztot, nzA, nzB, lrow, brstart = matin->rmap->rstart, brend = matin->rmap->rend;
1199: PetscInt *cmap, *idx_p, cstart = mat->rstartbs;
1201: PetscFunctionBegin;
1202: PetscCheck(!mat->getrowactive, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Already active");
1203: mat->getrowactive = PETSC_TRUE;
1205: if (!mat->rowvalues && (idx || v)) {
1206: /*
1207: allocate enough space to hold information from the longest row.
1208: */
1209: Mat_SeqSBAIJ *Aa = (Mat_SeqSBAIJ *)mat->A->data;
1210: Mat_SeqBAIJ *Ba = (Mat_SeqBAIJ *)mat->B->data;
1211: PetscInt max = 1, mbs = mat->mbs, tmp;
1212: for (i = 0; i < mbs; i++) {
1213: tmp = Aa->i[i + 1] - Aa->i[i] + Ba->i[i + 1] - Ba->i[i]; /* row length */
1214: if (max < tmp) max = tmp;
1215: }
1216: PetscCall(PetscMalloc2(max * bs2, &mat->rowvalues, max * bs2, &mat->rowindices));
1217: }
1219: PetscCheck(row >= brstart && row < brend, PETSC_COMM_SELF, PETSC_ERR_SUP, "Only local rows");
1220: lrow = row - brstart; /* local row index */
1222: pvA = &vworkA;
1223: pcA = &cworkA;
1224: pvB = &vworkB;
1225: pcB = &cworkB;
1226: if (!v) {
1227: pvA = NULL;
1228: pvB = NULL;
1229: }
1230: if (!idx) {
1231: pcA = NULL;
1232: if (!v) pcB = NULL;
1233: }
1234: PetscCall((*mat->A->ops->getrow)(mat->A, lrow, &nzA, pcA, pvA));
1235: PetscCall((*mat->B->ops->getrow)(mat->B, lrow, &nzB, pcB, pvB));
1236: nztot = nzA + nzB;
1238: cmap = mat->garray;
1239: if (v || idx) {
1240: if (nztot) {
1241: /* Sort by increasing column numbers, assuming A and B already sorted */
1242: PetscInt imark = -1;
1243: if (v) {
1244: *v = v_p = mat->rowvalues;
1245: for (i = 0; i < nzB; i++) {
1246: if (cmap[cworkB[i] / bs] < cstart) v_p[i] = vworkB[i];
1247: else break;
1248: }
1249: imark = i;
1250: for (i = 0; i < nzA; i++) v_p[imark + i] = vworkA[i];
1251: for (i = imark; i < nzB; i++) v_p[nzA + i] = vworkB[i];
1252: }
1253: if (idx) {
1254: *idx = idx_p = mat->rowindices;
1255: if (imark > -1) {
1256: for (i = 0; i < imark; i++) idx_p[i] = cmap[cworkB[i] / bs] * bs + cworkB[i] % bs;
1257: } else {
1258: for (i = 0; i < nzB; i++) {
1259: if (cmap[cworkB[i] / bs] < cstart) idx_p[i] = cmap[cworkB[i] / bs] * bs + cworkB[i] % bs;
1260: else break;
1261: }
1262: imark = i;
1263: }
1264: for (i = 0; i < nzA; i++) idx_p[imark + i] = cstart * bs + cworkA[i];
1265: for (i = imark; i < nzB; i++) idx_p[nzA + i] = cmap[cworkB[i] / bs] * bs + cworkB[i] % bs;
1266: }
1267: } else {
1268: if (idx) *idx = NULL;
1269: if (v) *v = NULL;
1270: }
1271: }
1272: *nz = nztot;
1273: PetscCall((*mat->A->ops->restorerow)(mat->A, lrow, &nzA, pcA, pvA));
1274: PetscCall((*mat->B->ops->restorerow)(mat->B, lrow, &nzB, pcB, pvB));
1275: PetscFunctionReturn(PETSC_SUCCESS);
1276: }
1278: static PetscErrorCode MatRestoreRow_MPISBAIJ(Mat mat, PetscInt row, PetscInt *nz, PetscInt **idx, PetscScalar **v)
1279: {
1280: Mat_MPISBAIJ *baij = (Mat_MPISBAIJ *)mat->data;
1282: PetscFunctionBegin;
1283: PetscCheck(baij->getrowactive, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "MatGetRow() must be called first");
1284: baij->getrowactive = PETSC_FALSE;
1285: PetscFunctionReturn(PETSC_SUCCESS);
1286: }
1288: static PetscErrorCode MatGetRowUpperTriangular_MPISBAIJ(Mat A)
1289: {
1290: Mat_MPISBAIJ *a = (Mat_MPISBAIJ *)A->data;
1291: Mat_SeqSBAIJ *aA = (Mat_SeqSBAIJ *)a->A->data;
1293: PetscFunctionBegin;
1294: aA->getrow_utriangular = PETSC_TRUE;
1295: PetscFunctionReturn(PETSC_SUCCESS);
1296: }
1297: static PetscErrorCode MatRestoreRowUpperTriangular_MPISBAIJ(Mat A)
1298: {
1299: Mat_MPISBAIJ *a = (Mat_MPISBAIJ *)A->data;
1300: Mat_SeqSBAIJ *aA = (Mat_SeqSBAIJ *)a->A->data;
1302: PetscFunctionBegin;
1303: aA->getrow_utriangular = PETSC_FALSE;
1304: PetscFunctionReturn(PETSC_SUCCESS);
1305: }
1307: static PetscErrorCode MatConjugate_MPISBAIJ(Mat mat)
1308: {
1309: PetscFunctionBegin;
1310: if (PetscDefined(USE_COMPLEX)) {
1311: Mat_MPISBAIJ *a = (Mat_MPISBAIJ *)mat->data;
1313: PetscCall(MatConjugate(a->A));
1314: PetscCall(MatConjugate(a->B));
1315: }
1316: PetscFunctionReturn(PETSC_SUCCESS);
1317: }
1319: static PetscErrorCode MatRealPart_MPISBAIJ(Mat A)
1320: {
1321: Mat_MPISBAIJ *a = (Mat_MPISBAIJ *)A->data;
1323: PetscFunctionBegin;
1324: PetscCall(MatRealPart(a->A));
1325: PetscCall(MatRealPart(a->B));
1326: PetscFunctionReturn(PETSC_SUCCESS);
1327: }
1329: static PetscErrorCode MatImaginaryPart_MPISBAIJ(Mat A)
1330: {
1331: Mat_MPISBAIJ *a = (Mat_MPISBAIJ *)A->data;
1333: PetscFunctionBegin;
1334: PetscCall(MatImaginaryPart(a->A));
1335: PetscCall(MatImaginaryPart(a->B));
1336: PetscFunctionReturn(PETSC_SUCCESS);
1337: }
1339: /* Check if isrow is a subset of iscol_local, called by MatCreateSubMatrix_MPISBAIJ()
1340: Input: isrow - distributed(parallel),
1341: iscol_local - locally owned (seq)
1342: */
1343: static PetscErrorCode ISEqual_private(IS isrow, IS iscol_local, PetscBool *flg)
1344: {
1345: PetscInt sz1, sz2, *a1, *a2, i, j, k, nmatch;
1346: const PetscInt *ptr1, *ptr2;
1348: PetscFunctionBegin;
1349: *flg = PETSC_FALSE;
1350: PetscCall(ISGetLocalSize(isrow, &sz1));
1351: PetscCall(ISGetLocalSize(iscol_local, &sz2));
1352: if (sz1 > sz2) PetscFunctionReturn(PETSC_SUCCESS);
1354: PetscCall(ISGetIndices(isrow, &ptr1));
1355: PetscCall(ISGetIndices(iscol_local, &ptr2));
1357: PetscCall(PetscMalloc1(sz1, &a1));
1358: PetscCall(PetscMalloc1(sz2, &a2));
1359: PetscCall(PetscArraycpy(a1, ptr1, sz1));
1360: PetscCall(PetscArraycpy(a2, ptr2, sz2));
1361: PetscCall(PetscSortInt(sz1, a1));
1362: PetscCall(PetscSortInt(sz2, a2));
1364: nmatch = 0;
1365: k = 0;
1366: for (i = 0; i < sz1; i++) {
1367: for (j = k; j < sz2; j++) {
1368: if (a1[i] == a2[j]) {
1369: k = j;
1370: nmatch++;
1371: break;
1372: }
1373: }
1374: }
1375: PetscCall(ISRestoreIndices(isrow, &ptr1));
1376: PetscCall(ISRestoreIndices(iscol_local, &ptr2));
1377: PetscCall(PetscFree(a1));
1378: PetscCall(PetscFree(a2));
1379: if (nmatch < sz1) {
1380: *flg = PETSC_FALSE;
1381: } else {
1382: *flg = PETSC_TRUE;
1383: }
1384: PetscFunctionReturn(PETSC_SUCCESS);
1385: }
1387: static PetscErrorCode MatCreateSubMatrix_MPISBAIJ(Mat mat, IS isrow, IS iscol, MatReuse call, Mat *newmat)
1388: {
1389: Mat C[2];
1390: IS iscol_local, isrow_local;
1391: PetscInt csize, csize_local, rsize;
1392: PetscBool isequal, issorted, isidentity = PETSC_FALSE;
1394: PetscFunctionBegin;
1395: PetscCall(ISGetLocalSize(iscol, &csize));
1396: PetscCall(ISGetLocalSize(isrow, &rsize));
1397: if (call == MAT_REUSE_MATRIX) {
1398: PetscCall(PetscObjectQuery((PetscObject)*newmat, "ISAllGather", (PetscObject *)&iscol_local));
1399: PetscCheck(iscol_local, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Submatrix passed in was not used before, cannot reuse");
1400: } else {
1401: PetscCall(ISAllGather(iscol, &iscol_local));
1402: PetscCall(ISSorted(iscol_local, &issorted));
1403: PetscCheck(issorted, PETSC_COMM_SELF, PETSC_ERR_ARG_INCOMP, "For symmetric format, iscol must be sorted");
1404: }
1405: PetscCall(ISEqual_private(isrow, iscol_local, &isequal));
1406: if (!isequal) {
1407: PetscCall(ISGetLocalSize(iscol_local, &csize_local));
1408: isidentity = (PetscBool)(mat->cmap->N == csize_local);
1409: if (!isidentity) {
1410: if (call == MAT_REUSE_MATRIX) {
1411: PetscCall(PetscObjectQuery((PetscObject)*newmat, "ISAllGather_other", (PetscObject *)&isrow_local));
1412: PetscCheck(isrow_local, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Submatrix passed in was not used before, cannot reuse");
1413: } else {
1414: PetscCall(ISAllGather(isrow, &isrow_local));
1415: PetscCall(ISSorted(isrow_local, &issorted));
1416: PetscCheck(issorted, PETSC_COMM_SELF, PETSC_ERR_ARG_INCOMP, "For symmetric format, isrow must be sorted");
1417: }
1418: }
1419: }
1420: /* now call MatCreateSubMatrix_MPIBAIJ() */
1421: PetscCall(MatCreateSubMatrix_MPIBAIJ_Private(mat, isrow, iscol_local, csize, isequal || isidentity ? call : MAT_INITIAL_MATRIX, isequal || isidentity ? newmat : C, (PetscBool)(isequal || isidentity)));
1422: if (!isequal && !isidentity) {
1423: if (call == MAT_INITIAL_MATRIX) {
1424: IS intersect;
1425: PetscInt ni;
1427: PetscCall(ISIntersect(isrow_local, iscol_local, &intersect));
1428: PetscCall(ISGetLocalSize(intersect, &ni));
1429: PetscCall(ISDestroy(&intersect));
1430: PetscCheck(ni == 0, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONG, "Cannot create such a submatrix: for symmetric format, when requesting an off-diagonal submatrix, isrow and iscol should have an empty intersection (number of common indices is %" PetscInt_FMT ")", ni);
1431: }
1432: PetscCall(MatCreateSubMatrix_MPIBAIJ_Private(mat, iscol, isrow_local, rsize, MAT_INITIAL_MATRIX, C + 1, PETSC_FALSE));
1433: PetscCall(MatTranspose(C[1], MAT_INPLACE_MATRIX, C + 1));
1434: PetscCall(MatAXPY(C[0], 1.0, C[1], DIFFERENT_NONZERO_PATTERN));
1435: if (call == MAT_REUSE_MATRIX) PetscCall(MatCopy(C[0], *newmat, SAME_NONZERO_PATTERN));
1436: else if (mat->rmap->bs == 1) PetscCall(MatConvert(C[0], MATAIJ, MAT_INITIAL_MATRIX, newmat));
1437: else PetscCall(MatCopy(C[0], *newmat, SAME_NONZERO_PATTERN));
1438: PetscCall(MatDestroy(C));
1439: PetscCall(MatDestroy(C + 1));
1440: }
1441: if (call == MAT_INITIAL_MATRIX) {
1442: if (!isequal && !isidentity) {
1443: PetscCall(PetscObjectCompose((PetscObject)*newmat, "ISAllGather_other", (PetscObject)isrow_local));
1444: PetscCall(ISDestroy(&isrow_local));
1445: }
1446: PetscCall(PetscObjectCompose((PetscObject)*newmat, "ISAllGather", (PetscObject)iscol_local));
1447: PetscCall(ISDestroy(&iscol_local));
1448: }
1449: PetscFunctionReturn(PETSC_SUCCESS);
1450: }
1452: static PetscErrorCode MatZeroEntries_MPISBAIJ(Mat A)
1453: {
1454: Mat_MPISBAIJ *l = (Mat_MPISBAIJ *)A->data;
1456: PetscFunctionBegin;
1457: PetscCall(MatZeroEntries(l->A));
1458: PetscCall(MatZeroEntries(l->B));
1459: PetscFunctionReturn(PETSC_SUCCESS);
1460: }
1462: static PetscErrorCode MatGetInfo_MPISBAIJ(Mat matin, MatInfoType flag, MatInfo *info)
1463: {
1464: Mat_MPISBAIJ *a = (Mat_MPISBAIJ *)matin->data;
1465: Mat A = a->A, B = a->B;
1466: PetscLogDouble isend[5], irecv[5];
1468: PetscFunctionBegin;
1469: info->block_size = (PetscReal)matin->rmap->bs;
1471: PetscCall(MatGetInfo(A, MAT_LOCAL, info));
1473: isend[0] = info->nz_used;
1474: isend[1] = info->nz_allocated;
1475: isend[2] = info->nz_unneeded;
1476: isend[3] = info->memory;
1477: isend[4] = info->mallocs;
1479: PetscCall(MatGetInfo(B, MAT_LOCAL, info));
1481: isend[0] += info->nz_used;
1482: isend[1] += info->nz_allocated;
1483: isend[2] += info->nz_unneeded;
1484: isend[3] += info->memory;
1485: isend[4] += info->mallocs;
1486: if (flag == MAT_LOCAL) {
1487: info->nz_used = isend[0];
1488: info->nz_allocated = isend[1];
1489: info->nz_unneeded = isend[2];
1490: info->memory = isend[3];
1491: info->mallocs = isend[4];
1492: } else if (flag == MAT_GLOBAL_MAX) {
1493: PetscCall(MPIU_Allreduce(isend, irecv, 5, MPIU_PETSCLOGDOUBLE, MPI_MAX, PetscObjectComm((PetscObject)matin)));
1495: info->nz_used = irecv[0];
1496: info->nz_allocated = irecv[1];
1497: info->nz_unneeded = irecv[2];
1498: info->memory = irecv[3];
1499: info->mallocs = irecv[4];
1500: } else if (flag == MAT_GLOBAL_SUM) {
1501: PetscCall(MPIU_Allreduce(isend, irecv, 5, MPIU_PETSCLOGDOUBLE, MPI_SUM, PetscObjectComm((PetscObject)matin)));
1503: info->nz_used = irecv[0];
1504: info->nz_allocated = irecv[1];
1505: info->nz_unneeded = irecv[2];
1506: info->memory = irecv[3];
1507: info->mallocs = irecv[4];
1508: } else SETERRQ(PETSC_COMM_SELF, PETSC_ERR_ARG_WRONG, "Unknown MatInfoType argument %d", (int)flag);
1509: info->fill_ratio_given = 0; /* no parallel LU/ILU/Cholesky */
1510: info->fill_ratio_needed = 0;
1511: info->factor_mallocs = 0;
1512: PetscFunctionReturn(PETSC_SUCCESS);
1513: }
1515: static PetscErrorCode MatSetOption_MPISBAIJ(Mat A, MatOption op, PetscBool flg)
1516: {
1517: Mat_MPISBAIJ *a = (Mat_MPISBAIJ *)A->data;
1518: Mat_SeqSBAIJ *aA = (Mat_SeqSBAIJ *)a->A->data;
1520: PetscFunctionBegin;
1521: switch (op) {
1522: case MAT_NEW_NONZERO_LOCATIONS:
1523: case MAT_NEW_NONZERO_ALLOCATION_ERR:
1524: case MAT_UNUSED_NONZERO_LOCATION_ERR:
1525: case MAT_KEEP_NONZERO_PATTERN:
1526: case MAT_SUBMAT_SINGLEIS:
1527: case MAT_NEW_NONZERO_LOCATION_ERR:
1528: MatCheckPreallocated(A, 1);
1529: PetscCall(MatSetOption(a->A, op, flg));
1530: PetscCall(MatSetOption(a->B, op, flg));
1531: break;
1532: case MAT_ROW_ORIENTED:
1533: MatCheckPreallocated(A, 1);
1534: a->roworiented = flg;
1536: PetscCall(MatSetOption(a->A, op, flg));
1537: PetscCall(MatSetOption(a->B, op, flg));
1538: break;
1539: case MAT_FORCE_DIAGONAL_ENTRIES:
1540: case MAT_SORTED_FULL:
1541: PetscCall(PetscInfo(A, "Option %s ignored\n", MatOptions[op]));
1542: break;
1543: case MAT_IGNORE_OFF_PROC_ENTRIES:
1544: a->donotstash = flg;
1545: break;
1546: case MAT_USE_HASH_TABLE:
1547: a->ht_flag = flg;
1548: break;
1549: case MAT_HERMITIAN:
1550: MatCheckPreallocated(A, 1);
1551: PetscCall(MatSetOption(a->A, op, flg));
1552: #if defined(PETSC_USE_COMPLEX)
1553: if (flg) { /* need different mat-vec ops */
1554: A->ops->mult = MatMult_MPISBAIJ_Hermitian;
1555: A->ops->multadd = MatMultAdd_MPISBAIJ_Hermitian;
1556: A->ops->multtranspose = NULL;
1557: A->ops->multtransposeadd = NULL;
1558: A->symmetric = PETSC_BOOL3_FALSE;
1559: }
1560: #endif
1561: break;
1562: case MAT_SPD:
1563: case MAT_SYMMETRIC:
1564: MatCheckPreallocated(A, 1);
1565: PetscCall(MatSetOption(a->A, op, flg));
1566: #if defined(PETSC_USE_COMPLEX)
1567: if (flg) { /* restore to use default mat-vec ops */
1568: A->ops->mult = MatMult_MPISBAIJ;
1569: A->ops->multadd = MatMultAdd_MPISBAIJ;
1570: A->ops->multtranspose = MatMult_MPISBAIJ;
1571: A->ops->multtransposeadd = MatMultAdd_MPISBAIJ;
1572: }
1573: #endif
1574: break;
1575: case MAT_STRUCTURALLY_SYMMETRIC:
1576: MatCheckPreallocated(A, 1);
1577: PetscCall(MatSetOption(a->A, op, flg));
1578: break;
1579: case MAT_SYMMETRY_ETERNAL:
1580: case MAT_STRUCTURAL_SYMMETRY_ETERNAL:
1581: PetscCheck(flg, PETSC_COMM_SELF, PETSC_ERR_SUP, "Matrix must be symmetric");
1582: PetscCall(PetscInfo(A, "Option %s ignored\n", MatOptions[op]));
1583: break;
1584: case MAT_SPD_ETERNAL:
1585: break;
1586: case MAT_IGNORE_LOWER_TRIANGULAR:
1587: aA->ignore_ltriangular = flg;
1588: break;
1589: case MAT_ERROR_LOWER_TRIANGULAR:
1590: aA->ignore_ltriangular = flg;
1591: break;
1592: case MAT_GETROW_UPPERTRIANGULAR:
1593: aA->getrow_utriangular = flg;
1594: break;
1595: default:
1596: SETERRQ(PETSC_COMM_SELF, PETSC_ERR_SUP, "unknown option %d", op);
1597: }
1598: PetscFunctionReturn(PETSC_SUCCESS);
1599: }
1601: static PetscErrorCode MatTranspose_MPISBAIJ(Mat A, MatReuse reuse, Mat *B)
1602: {
1603: PetscFunctionBegin;
1604: if (reuse == MAT_REUSE_MATRIX) PetscCall(MatTransposeCheckNonzeroState_Private(A, *B));
1605: if (reuse == MAT_INITIAL_MATRIX) {
1606: PetscCall(MatDuplicate(A, MAT_COPY_VALUES, B));
1607: } else if (reuse == MAT_REUSE_MATRIX) {
1608: PetscCall(MatCopy(A, *B, SAME_NONZERO_PATTERN));
1609: }
1610: PetscFunctionReturn(PETSC_SUCCESS);
1611: }
1613: static PetscErrorCode MatDiagonalScale_MPISBAIJ(Mat mat, Vec ll, Vec rr)
1614: {
1615: Mat_MPISBAIJ *baij = (Mat_MPISBAIJ *)mat->data;
1616: Mat a = baij->A, b = baij->B;
1617: PetscInt nv, m, n;
1618: PetscBool flg;
1620: PetscFunctionBegin;
1621: if (ll != rr) {
1622: PetscCall(VecEqual(ll, rr, &flg));
1623: PetscCheck(flg, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "For symmetric format, left and right scaling vectors must be same");
1624: }
1625: if (!ll) PetscFunctionReturn(PETSC_SUCCESS);
1627: PetscCall(MatGetLocalSize(mat, &m, &n));
1628: PetscCheck(m == n, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "For symmetric format, local size %" PetscInt_FMT " %" PetscInt_FMT " must be same", m, n);
1630: PetscCall(VecGetLocalSize(rr, &nv));
1631: PetscCheck(nv == n, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Left and right vector non-conforming local size");
1633: PetscCall(VecScatterBegin(baij->Mvctx, rr, baij->lvec, INSERT_VALUES, SCATTER_FORWARD));
1635: /* left diagonalscale the off-diagonal part */
1636: PetscUseTypeMethod(b, diagonalscale, ll, NULL);
1638: /* scale the diagonal part */
1639: PetscUseTypeMethod(a, diagonalscale, ll, rr);
1641: /* right diagonalscale the off-diagonal part */
1642: PetscCall(VecScatterEnd(baij->Mvctx, rr, baij->lvec, INSERT_VALUES, SCATTER_FORWARD));
1643: PetscUseTypeMethod(b, diagonalscale, NULL, baij->lvec);
1644: PetscFunctionReturn(PETSC_SUCCESS);
1645: }
1647: static PetscErrorCode MatSetUnfactored_MPISBAIJ(Mat A)
1648: {
1649: Mat_MPISBAIJ *a = (Mat_MPISBAIJ *)A->data;
1651: PetscFunctionBegin;
1652: PetscCall(MatSetUnfactored(a->A));
1653: PetscFunctionReturn(PETSC_SUCCESS);
1654: }
1656: static PetscErrorCode MatDuplicate_MPISBAIJ(Mat, MatDuplicateOption, Mat *);
1658: static PetscErrorCode MatEqual_MPISBAIJ(Mat A, Mat B, PetscBool *flag)
1659: {
1660: Mat_MPISBAIJ *matB = (Mat_MPISBAIJ *)B->data, *matA = (Mat_MPISBAIJ *)A->data;
1661: Mat a, b, c, d;
1662: PetscBool flg;
1664: PetscFunctionBegin;
1665: a = matA->A;
1666: b = matA->B;
1667: c = matB->A;
1668: d = matB->B;
1670: PetscCall(MatEqual(a, c, &flg));
1671: if (flg) PetscCall(MatEqual(b, d, &flg));
1672: PetscCall(MPIU_Allreduce(&flg, flag, 1, MPIU_BOOL, MPI_LAND, PetscObjectComm((PetscObject)A)));
1673: PetscFunctionReturn(PETSC_SUCCESS);
1674: }
1676: static PetscErrorCode MatCopy_MPISBAIJ(Mat A, Mat B, MatStructure str)
1677: {
1678: PetscBool isbaij;
1680: PetscFunctionBegin;
1681: PetscCall(PetscObjectTypeCompareAny((PetscObject)B, &isbaij, MATSEQSBAIJ, MATMPISBAIJ, ""));
1682: PetscCheck(isbaij, PetscObjectComm((PetscObject)B), PETSC_ERR_SUP, "Not for matrix type %s", ((PetscObject)B)->type_name);
1683: /* If the two matrices don't have the same copy implementation, they aren't compatible for fast copy. */
1684: if ((str != SAME_NONZERO_PATTERN) || (A->ops->copy != B->ops->copy)) {
1685: PetscCall(MatGetRowUpperTriangular(A));
1686: PetscCall(MatCopy_Basic(A, B, str));
1687: PetscCall(MatRestoreRowUpperTriangular(A));
1688: } else {
1689: Mat_MPISBAIJ *a = (Mat_MPISBAIJ *)A->data;
1690: Mat_MPISBAIJ *b = (Mat_MPISBAIJ *)B->data;
1692: PetscCall(MatCopy(a->A, b->A, str));
1693: PetscCall(MatCopy(a->B, b->B, str));
1694: }
1695: PetscCall(PetscObjectStateIncrease((PetscObject)B));
1696: PetscFunctionReturn(PETSC_SUCCESS);
1697: }
1699: static PetscErrorCode MatAXPY_MPISBAIJ(Mat Y, PetscScalar a, Mat X, MatStructure str)
1700: {
1701: Mat_MPISBAIJ *xx = (Mat_MPISBAIJ *)X->data, *yy = (Mat_MPISBAIJ *)Y->data;
1702: PetscBLASInt bnz, one = 1;
1703: Mat_SeqSBAIJ *xa, *ya;
1704: Mat_SeqBAIJ *xb, *yb;
1706: PetscFunctionBegin;
1707: if (str == SAME_NONZERO_PATTERN) {
1708: PetscScalar alpha = a;
1709: xa = (Mat_SeqSBAIJ *)xx->A->data;
1710: ya = (Mat_SeqSBAIJ *)yy->A->data;
1711: PetscCall(PetscBLASIntCast(xa->nz, &bnz));
1712: PetscCallBLAS("BLASaxpy", BLASaxpy_(&bnz, &alpha, xa->a, &one, ya->a, &one));
1713: xb = (Mat_SeqBAIJ *)xx->B->data;
1714: yb = (Mat_SeqBAIJ *)yy->B->data;
1715: PetscCall(PetscBLASIntCast(xb->nz, &bnz));
1716: PetscCallBLAS("BLASaxpy", BLASaxpy_(&bnz, &alpha, xb->a, &one, yb->a, &one));
1717: PetscCall(PetscObjectStateIncrease((PetscObject)Y));
1718: } else if (str == SUBSET_NONZERO_PATTERN) { /* nonzeros of X is a subset of Y's */
1719: PetscCall(MatSetOption(X, MAT_GETROW_UPPERTRIANGULAR, PETSC_TRUE));
1720: PetscCall(MatAXPY_Basic(Y, a, X, str));
1721: PetscCall(MatSetOption(X, MAT_GETROW_UPPERTRIANGULAR, PETSC_FALSE));
1722: } else {
1723: Mat B;
1724: PetscInt *nnz_d, *nnz_o, bs = Y->rmap->bs;
1725: PetscCheck(bs == X->rmap->bs, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Matrices must have same block size");
1726: PetscCall(MatGetRowUpperTriangular(X));
1727: PetscCall(MatGetRowUpperTriangular(Y));
1728: PetscCall(PetscMalloc1(yy->A->rmap->N, &nnz_d));
1729: PetscCall(PetscMalloc1(yy->B->rmap->N, &nnz_o));
1730: PetscCall(MatCreate(PetscObjectComm((PetscObject)Y), &B));
1731: PetscCall(PetscObjectSetName((PetscObject)B, ((PetscObject)Y)->name));
1732: PetscCall(MatSetSizes(B, Y->rmap->n, Y->cmap->n, Y->rmap->N, Y->cmap->N));
1733: PetscCall(MatSetBlockSizesFromMats(B, Y, Y));
1734: PetscCall(MatSetType(B, MATMPISBAIJ));
1735: PetscCall(MatAXPYGetPreallocation_SeqSBAIJ(yy->A, xx->A, nnz_d));
1736: PetscCall(MatAXPYGetPreallocation_MPIBAIJ(yy->B, yy->garray, xx->B, xx->garray, nnz_o));
1737: PetscCall(MatMPISBAIJSetPreallocation(B, bs, 0, nnz_d, 0, nnz_o));
1738: PetscCall(MatAXPY_BasicWithPreallocation(B, Y, a, X, str));
1739: PetscCall(MatHeaderMerge(Y, &B));
1740: PetscCall(PetscFree(nnz_d));
1741: PetscCall(PetscFree(nnz_o));
1742: PetscCall(MatRestoreRowUpperTriangular(X));
1743: PetscCall(MatRestoreRowUpperTriangular(Y));
1744: }
1745: PetscFunctionReturn(PETSC_SUCCESS);
1746: }
1748: static PetscErrorCode MatCreateSubMatrices_MPISBAIJ(Mat A, PetscInt n, const IS irow[], const IS icol[], MatReuse scall, Mat *B[])
1749: {
1750: PetscInt i;
1751: PetscBool flg;
1753: PetscFunctionBegin;
1754: PetscCall(MatCreateSubMatrices_MPIBAIJ(A, n, irow, icol, scall, B)); /* B[] are sbaij matrices */
1755: for (i = 0; i < n; i++) {
1756: PetscCall(ISEqual(irow[i], icol[i], &flg));
1757: if (!flg) PetscCall(MatSeqSBAIJZeroOps_Private(*B[i]));
1758: }
1759: PetscFunctionReturn(PETSC_SUCCESS);
1760: }
1762: static PetscErrorCode MatShift_MPISBAIJ(Mat Y, PetscScalar a)
1763: {
1764: Mat_MPISBAIJ *maij = (Mat_MPISBAIJ *)Y->data;
1765: Mat_SeqSBAIJ *aij = (Mat_SeqSBAIJ *)maij->A->data;
1767: PetscFunctionBegin;
1768: if (!Y->preallocated) {
1769: PetscCall(MatMPISBAIJSetPreallocation(Y, Y->rmap->bs, 1, NULL, 0, NULL));
1770: } else if (!aij->nz) {
1771: PetscInt nonew = aij->nonew;
1772: PetscCall(MatSeqSBAIJSetPreallocation(maij->A, Y->rmap->bs, 1, NULL));
1773: aij->nonew = nonew;
1774: }
1775: PetscCall(MatShift_Basic(Y, a));
1776: PetscFunctionReturn(PETSC_SUCCESS);
1777: }
1779: static PetscErrorCode MatMissingDiagonal_MPISBAIJ(Mat A, PetscBool *missing, PetscInt *d)
1780: {
1781: Mat_MPISBAIJ *a = (Mat_MPISBAIJ *)A->data;
1783: PetscFunctionBegin;
1784: PetscCheck(A->rmap->n == A->cmap->n, PETSC_COMM_SELF, PETSC_ERR_SUP, "Only works for square matrices");
1785: PetscCall(MatMissingDiagonal(a->A, missing, d));
1786: if (d) {
1787: PetscInt rstart;
1788: PetscCall(MatGetOwnershipRange(A, &rstart, NULL));
1789: *d += rstart / A->rmap->bs;
1790: }
1791: PetscFunctionReturn(PETSC_SUCCESS);
1792: }
1794: static PetscErrorCode MatGetDiagonalBlock_MPISBAIJ(Mat A, Mat *a)
1795: {
1796: PetscFunctionBegin;
1797: *a = ((Mat_MPISBAIJ *)A->data)->A;
1798: PetscFunctionReturn(PETSC_SUCCESS);
1799: }
1801: static PetscErrorCode MatEliminateZeros_MPISBAIJ(Mat A, PetscBool keep)
1802: {
1803: Mat_MPISBAIJ *a = (Mat_MPISBAIJ *)A->data;
1805: PetscFunctionBegin;
1806: PetscCall(MatEliminateZeros_SeqSBAIJ(a->A, keep)); // possibly keep zero diagonal coefficients
1807: PetscCall(MatEliminateZeros_SeqBAIJ(a->B, PETSC_FALSE)); // never keep zero diagonal coefficients
1808: PetscFunctionReturn(PETSC_SUCCESS);
1809: }
1811: static PetscErrorCode MatLoad_MPISBAIJ(Mat, PetscViewer);
1812: static PetscErrorCode MatGetRowMaxAbs_MPISBAIJ(Mat, Vec, PetscInt[]);
1813: static PetscErrorCode MatSOR_MPISBAIJ(Mat, Vec, PetscReal, MatSORType, PetscReal, PetscInt, PetscInt, Vec);
1815: static struct _MatOps MatOps_Values = {MatSetValues_MPISBAIJ,
1816: MatGetRow_MPISBAIJ,
1817: MatRestoreRow_MPISBAIJ,
1818: MatMult_MPISBAIJ,
1819: /* 4*/ MatMultAdd_MPISBAIJ,
1820: MatMult_MPISBAIJ, /* transpose versions are same as non-transpose */
1821: MatMultAdd_MPISBAIJ,
1822: NULL,
1823: NULL,
1824: NULL,
1825: /* 10*/ NULL,
1826: NULL,
1827: NULL,
1828: MatSOR_MPISBAIJ,
1829: MatTranspose_MPISBAIJ,
1830: /* 15*/ MatGetInfo_MPISBAIJ,
1831: MatEqual_MPISBAIJ,
1832: MatGetDiagonal_MPISBAIJ,
1833: MatDiagonalScale_MPISBAIJ,
1834: MatNorm_MPISBAIJ,
1835: /* 20*/ MatAssemblyBegin_MPISBAIJ,
1836: MatAssemblyEnd_MPISBAIJ,
1837: MatSetOption_MPISBAIJ,
1838: MatZeroEntries_MPISBAIJ,
1839: /* 24*/ NULL,
1840: NULL,
1841: NULL,
1842: NULL,
1843: NULL,
1844: /* 29*/ MatSetUp_MPI_Hash,
1845: NULL,
1846: NULL,
1847: MatGetDiagonalBlock_MPISBAIJ,
1848: NULL,
1849: /* 34*/ MatDuplicate_MPISBAIJ,
1850: NULL,
1851: NULL,
1852: NULL,
1853: NULL,
1854: /* 39*/ MatAXPY_MPISBAIJ,
1855: MatCreateSubMatrices_MPISBAIJ,
1856: MatIncreaseOverlap_MPISBAIJ,
1857: MatGetValues_MPISBAIJ,
1858: MatCopy_MPISBAIJ,
1859: /* 44*/ NULL,
1860: MatScale_MPISBAIJ,
1861: MatShift_MPISBAIJ,
1862: NULL,
1863: NULL,
1864: /* 49*/ NULL,
1865: NULL,
1866: NULL,
1867: NULL,
1868: NULL,
1869: /* 54*/ NULL,
1870: NULL,
1871: MatSetUnfactored_MPISBAIJ,
1872: NULL,
1873: MatSetValuesBlocked_MPISBAIJ,
1874: /* 59*/ MatCreateSubMatrix_MPISBAIJ,
1875: NULL,
1876: NULL,
1877: NULL,
1878: NULL,
1879: /* 64*/ NULL,
1880: NULL,
1881: NULL,
1882: NULL,
1883: NULL,
1884: /* 69*/ MatGetRowMaxAbs_MPISBAIJ,
1885: NULL,
1886: MatConvert_MPISBAIJ_Basic,
1887: NULL,
1888: NULL,
1889: /* 74*/ NULL,
1890: NULL,
1891: NULL,
1892: NULL,
1893: NULL,
1894: /* 79*/ NULL,
1895: NULL,
1896: NULL,
1897: NULL,
1898: MatLoad_MPISBAIJ,
1899: /* 84*/ NULL,
1900: NULL,
1901: NULL,
1902: NULL,
1903: NULL,
1904: /* 89*/ NULL,
1905: NULL,
1906: NULL,
1907: NULL,
1908: NULL,
1909: /* 94*/ NULL,
1910: NULL,
1911: NULL,
1912: NULL,
1913: NULL,
1914: /* 99*/ NULL,
1915: NULL,
1916: NULL,
1917: MatConjugate_MPISBAIJ,
1918: NULL,
1919: /*104*/ NULL,
1920: MatRealPart_MPISBAIJ,
1921: MatImaginaryPart_MPISBAIJ,
1922: MatGetRowUpperTriangular_MPISBAIJ,
1923: MatRestoreRowUpperTriangular_MPISBAIJ,
1924: /*109*/ NULL,
1925: NULL,
1926: NULL,
1927: NULL,
1928: MatMissingDiagonal_MPISBAIJ,
1929: /*114*/ NULL,
1930: NULL,
1931: NULL,
1932: NULL,
1933: NULL,
1934: /*119*/ NULL,
1935: NULL,
1936: NULL,
1937: NULL,
1938: NULL,
1939: /*124*/ NULL,
1940: NULL,
1941: NULL,
1942: NULL,
1943: NULL,
1944: /*129*/ NULL,
1945: NULL,
1946: NULL,
1947: NULL,
1948: NULL,
1949: /*134*/ NULL,
1950: NULL,
1951: NULL,
1952: NULL,
1953: NULL,
1954: /*139*/ MatSetBlockSizes_Default,
1955: NULL,
1956: NULL,
1957: NULL,
1958: NULL,
1959: /*144*/ MatCreateMPIMatConcatenateSeqMat_MPISBAIJ,
1960: NULL,
1961: NULL,
1962: NULL,
1963: NULL,
1964: NULL,
1965: /*150*/ NULL,
1966: MatEliminateZeros_MPISBAIJ,
1967: NULL};
1969: static PetscErrorCode MatMPISBAIJSetPreallocation_MPISBAIJ(Mat B, PetscInt bs, PetscInt d_nz, const PetscInt *d_nnz, PetscInt o_nz, const PetscInt *o_nnz)
1970: {
1971: Mat_MPISBAIJ *b = (Mat_MPISBAIJ *)B->data;
1972: PetscInt i, mbs, Mbs;
1973: PetscMPIInt size;
1975: PetscFunctionBegin;
1976: if (B->hash_active) {
1977: B->ops[0] = b->cops;
1978: B->hash_active = PETSC_FALSE;
1979: }
1980: if (!B->preallocated) PetscCall(MatStashCreate_Private(PetscObjectComm((PetscObject)B), bs, &B->bstash));
1981: PetscCall(MatSetBlockSize(B, PetscAbs(bs)));
1982: PetscCall(PetscLayoutSetUp(B->rmap));
1983: PetscCall(PetscLayoutSetUp(B->cmap));
1984: PetscCall(PetscLayoutGetBlockSize(B->rmap, &bs));
1985: PetscCheck(B->rmap->N <= B->cmap->N, PetscObjectComm((PetscObject)B), PETSC_ERR_SUP, "MPISBAIJ matrix cannot have more rows %" PetscInt_FMT " than columns %" PetscInt_FMT, B->rmap->N, B->cmap->N);
1986: PetscCheck(B->rmap->n <= B->cmap->n, PETSC_COMM_SELF, PETSC_ERR_SUP, "MPISBAIJ matrix cannot have more local rows %" PetscInt_FMT " than columns %" PetscInt_FMT, B->rmap->n, B->cmap->n);
1988: mbs = B->rmap->n / bs;
1989: Mbs = B->rmap->N / bs;
1990: PetscCheck(mbs * bs == B->rmap->n, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "No of local rows %" PetscInt_FMT " must be divisible by blocksize %" PetscInt_FMT, B->rmap->N, bs);
1992: B->rmap->bs = bs;
1993: b->bs2 = bs * bs;
1994: b->mbs = mbs;
1995: b->Mbs = Mbs;
1996: b->nbs = B->cmap->n / bs;
1997: b->Nbs = B->cmap->N / bs;
1999: for (i = 0; i <= b->size; i++) b->rangebs[i] = B->rmap->range[i] / bs;
2000: b->rstartbs = B->rmap->rstart / bs;
2001: b->rendbs = B->rmap->rend / bs;
2003: b->cstartbs = B->cmap->rstart / bs;
2004: b->cendbs = B->cmap->rend / bs;
2006: #if defined(PETSC_USE_CTABLE)
2007: PetscCall(PetscHMapIDestroy(&b->colmap));
2008: #else
2009: PetscCall(PetscFree(b->colmap));
2010: #endif
2011: PetscCall(PetscFree(b->garray));
2012: PetscCall(VecDestroy(&b->lvec));
2013: PetscCall(VecScatterDestroy(&b->Mvctx));
2014: PetscCall(VecDestroy(&b->slvec0));
2015: PetscCall(VecDestroy(&b->slvec0b));
2016: PetscCall(VecDestroy(&b->slvec1));
2017: PetscCall(VecDestroy(&b->slvec1a));
2018: PetscCall(VecDestroy(&b->slvec1b));
2019: PetscCall(VecScatterDestroy(&b->sMvctx));
2021: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)B), &size));
2023: MatSeqXAIJGetOptions_Private(b->B);
2024: PetscCall(MatDestroy(&b->B));
2025: PetscCall(MatCreate(PETSC_COMM_SELF, &b->B));
2026: PetscCall(MatSetSizes(b->B, B->rmap->n, size > 1 ? B->cmap->N : 0, B->rmap->n, size > 1 ? B->cmap->N : 0));
2027: PetscCall(MatSetType(b->B, MATSEQBAIJ));
2028: MatSeqXAIJRestoreOptions_Private(b->B);
2030: MatSeqXAIJGetOptions_Private(b->A);
2031: PetscCall(MatDestroy(&b->A));
2032: PetscCall(MatCreate(PETSC_COMM_SELF, &b->A));
2033: PetscCall(MatSetSizes(b->A, B->rmap->n, B->cmap->n, B->rmap->n, B->cmap->n));
2034: PetscCall(MatSetType(b->A, MATSEQSBAIJ));
2035: MatSeqXAIJRestoreOptions_Private(b->A);
2037: PetscCall(MatSeqSBAIJSetPreallocation(b->A, bs, d_nz, d_nnz));
2038: PetscCall(MatSeqBAIJSetPreallocation(b->B, bs, o_nz, o_nnz));
2040: B->preallocated = PETSC_TRUE;
2041: B->was_assembled = PETSC_FALSE;
2042: B->assembled = PETSC_FALSE;
2043: PetscFunctionReturn(PETSC_SUCCESS);
2044: }
2046: static PetscErrorCode MatMPISBAIJSetPreallocationCSR_MPISBAIJ(Mat B, PetscInt bs, const PetscInt ii[], const PetscInt jj[], const PetscScalar V[])
2047: {
2048: PetscInt m, rstart, cend;
2049: PetscInt i, j, d, nz, bd, nz_max = 0, *d_nnz = NULL, *o_nnz = NULL;
2050: const PetscInt *JJ = NULL;
2051: PetscScalar *values = NULL;
2052: PetscBool roworiented = ((Mat_MPISBAIJ *)B->data)->roworiented;
2053: PetscBool nooffprocentries;
2055: PetscFunctionBegin;
2056: PetscCheck(bs >= 1, PetscObjectComm((PetscObject)B), PETSC_ERR_ARG_OUTOFRANGE, "Invalid block size specified, must be positive but it is %" PetscInt_FMT, bs);
2057: PetscCall(PetscLayoutSetBlockSize(B->rmap, bs));
2058: PetscCall(PetscLayoutSetBlockSize(B->cmap, bs));
2059: PetscCall(PetscLayoutSetUp(B->rmap));
2060: PetscCall(PetscLayoutSetUp(B->cmap));
2061: PetscCall(PetscLayoutGetBlockSize(B->rmap, &bs));
2062: m = B->rmap->n / bs;
2063: rstart = B->rmap->rstart / bs;
2064: cend = B->cmap->rend / bs;
2066: PetscCheck(!ii[0], PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "ii[0] must be 0 but it is %" PetscInt_FMT, ii[0]);
2067: PetscCall(PetscMalloc2(m, &d_nnz, m, &o_nnz));
2068: for (i = 0; i < m; i++) {
2069: nz = ii[i + 1] - ii[i];
2070: PetscCheck(nz >= 0, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Local row %" PetscInt_FMT " has a negative number of columns %" PetscInt_FMT, i, nz);
2071: /* count the ones on the diagonal and above, split into diagonal and off-diagonal portions. */
2072: JJ = jj + ii[i];
2073: bd = 0;
2074: for (j = 0; j < nz; j++) {
2075: if (*JJ >= i + rstart) break;
2076: JJ++;
2077: bd++;
2078: }
2079: d = 0;
2080: for (; j < nz; j++) {
2081: if (*JJ++ >= cend) break;
2082: d++;
2083: }
2084: d_nnz[i] = d;
2085: o_nnz[i] = nz - d - bd;
2086: nz = nz - bd;
2087: nz_max = PetscMax(nz_max, nz);
2088: }
2089: PetscCall(MatMPISBAIJSetPreallocation(B, bs, 0, d_nnz, 0, o_nnz));
2090: PetscCall(MatSetOption(B, MAT_IGNORE_LOWER_TRIANGULAR, PETSC_TRUE));
2091: PetscCall(PetscFree2(d_nnz, o_nnz));
2093: values = (PetscScalar *)V;
2094: if (!values) PetscCall(PetscCalloc1(bs * bs * nz_max, &values));
2095: for (i = 0; i < m; i++) {
2096: PetscInt row = i + rstart;
2097: PetscInt ncols = ii[i + 1] - ii[i];
2098: const PetscInt *icols = jj + ii[i];
2099: if (bs == 1 || !roworiented) { /* block ordering matches the non-nested layout of MatSetValues so we can insert entire rows */
2100: const PetscScalar *svals = values + (V ? (bs * bs * ii[i]) : 0);
2101: PetscCall(MatSetValuesBlocked_MPISBAIJ(B, 1, &row, ncols, icols, svals, INSERT_VALUES));
2102: } else { /* block ordering does not match so we can only insert one block at a time. */
2103: PetscInt j;
2104: for (j = 0; j < ncols; j++) {
2105: const PetscScalar *svals = values + (V ? (bs * bs * (ii[i] + j)) : 0);
2106: PetscCall(MatSetValuesBlocked_MPISBAIJ(B, 1, &row, 1, &icols[j], svals, INSERT_VALUES));
2107: }
2108: }
2109: }
2111: if (!V) PetscCall(PetscFree(values));
2112: nooffprocentries = B->nooffprocentries;
2113: B->nooffprocentries = PETSC_TRUE;
2114: PetscCall(MatAssemblyBegin(B, MAT_FINAL_ASSEMBLY));
2115: PetscCall(MatAssemblyEnd(B, MAT_FINAL_ASSEMBLY));
2116: B->nooffprocentries = nooffprocentries;
2118: PetscCall(MatSetOption(B, MAT_NEW_NONZERO_LOCATION_ERR, PETSC_TRUE));
2119: PetscFunctionReturn(PETSC_SUCCESS);
2120: }
2122: /*MC
2123: MATMPISBAIJ - MATMPISBAIJ = "mpisbaij" - A matrix type to be used for distributed symmetric sparse block matrices,
2124: based on block compressed sparse row format. Only the upper triangular portion of the "diagonal" portion of
2125: the matrix is stored.
2127: For complex numbers by default this matrix is symmetric, NOT Hermitian symmetric. To make it Hermitian symmetric you
2128: can call `MatSetOption`(`Mat`, `MAT_HERMITIAN`);
2130: Options Database Key:
2131: . -mat_type mpisbaij - sets the matrix type to "mpisbaij" during a call to `MatSetFromOptions()`
2133: Level: beginner
2135: Note:
2136: The number of rows in the matrix must be less than or equal to the number of columns. Similarly the number of rows in the
2137: diagonal portion of the matrix of each process has to less than or equal the number of columns.
2139: .seealso: [](ch_matrices), `Mat`, `MATSBAIJ`, `MATBAIJ`, `MatCreateBAIJ()`, `MATSEQSBAIJ`, `MatType`
2140: M*/
2142: PETSC_EXTERN PetscErrorCode MatCreate_MPISBAIJ(Mat B)
2143: {
2144: Mat_MPISBAIJ *b;
2145: PetscBool flg = PETSC_FALSE;
2147: PetscFunctionBegin;
2148: PetscCall(PetscNew(&b));
2149: B->data = (void *)b;
2150: B->ops[0] = MatOps_Values;
2152: B->ops->destroy = MatDestroy_MPISBAIJ;
2153: B->ops->view = MatView_MPISBAIJ;
2154: B->assembled = PETSC_FALSE;
2155: B->insertmode = NOT_SET_VALUES;
2157: PetscCallMPI(MPI_Comm_rank(PetscObjectComm((PetscObject)B), &b->rank));
2158: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)B), &b->size));
2160: /* build local table of row and column ownerships */
2161: PetscCall(PetscMalloc1(b->size + 2, &b->rangebs));
2163: /* build cache for off array entries formed */
2164: PetscCall(MatStashCreate_Private(PetscObjectComm((PetscObject)B), 1, &B->stash));
2166: b->donotstash = PETSC_FALSE;
2167: b->colmap = NULL;
2168: b->garray = NULL;
2169: b->roworiented = PETSC_TRUE;
2171: /* stuff used in block assembly */
2172: b->barray = NULL;
2174: /* stuff used for matrix vector multiply */
2175: b->lvec = NULL;
2176: b->Mvctx = NULL;
2177: b->slvec0 = NULL;
2178: b->slvec0b = NULL;
2179: b->slvec1 = NULL;
2180: b->slvec1a = NULL;
2181: b->slvec1b = NULL;
2182: b->sMvctx = NULL;
2184: /* stuff for MatGetRow() */
2185: b->rowindices = NULL;
2186: b->rowvalues = NULL;
2187: b->getrowactive = PETSC_FALSE;
2189: /* hash table stuff */
2190: b->ht = NULL;
2191: b->hd = NULL;
2192: b->ht_size = 0;
2193: b->ht_flag = PETSC_FALSE;
2194: b->ht_fact = 0;
2195: b->ht_total_ct = 0;
2196: b->ht_insert_ct = 0;
2198: /* stuff for MatCreateSubMatrices_MPIBAIJ_local() */
2199: b->ijonly = PETSC_FALSE;
2201: b->in_loc = NULL;
2202: b->v_loc = NULL;
2203: b->n_loc = 0;
2205: PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatStoreValues_C", MatStoreValues_MPISBAIJ));
2206: PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatRetrieveValues_C", MatRetrieveValues_MPISBAIJ));
2207: PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatMPISBAIJSetPreallocation_C", MatMPISBAIJSetPreallocation_MPISBAIJ));
2208: PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatMPISBAIJSetPreallocationCSR_C", MatMPISBAIJSetPreallocationCSR_MPISBAIJ));
2209: #if defined(PETSC_HAVE_ELEMENTAL)
2210: PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatConvert_mpisbaij_elemental_C", MatConvert_MPISBAIJ_Elemental));
2211: #endif
2212: #if defined(PETSC_HAVE_SCALAPACK)
2213: PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatConvert_mpisbaij_scalapack_C", MatConvert_SBAIJ_ScaLAPACK));
2214: #endif
2215: PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatConvert_mpisbaij_mpiaij_C", MatConvert_MPISBAIJ_Basic));
2216: PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatConvert_mpisbaij_mpibaij_C", MatConvert_MPISBAIJ_Basic));
2218: B->symmetric = PETSC_BOOL3_TRUE;
2219: B->structurally_symmetric = PETSC_BOOL3_TRUE;
2220: B->symmetry_eternal = PETSC_TRUE;
2221: B->structural_symmetry_eternal = PETSC_TRUE;
2222: #if defined(PETSC_USE_COMPLEX)
2223: B->hermitian = PETSC_BOOL3_FALSE;
2224: #else
2225: B->hermitian = PETSC_BOOL3_TRUE;
2226: #endif
2228: PetscCall(PetscObjectChangeTypeName((PetscObject)B, MATMPISBAIJ));
2229: PetscOptionsBegin(PetscObjectComm((PetscObject)B), NULL, "Options for loading MPISBAIJ matrix 1", "Mat");
2230: PetscCall(PetscOptionsBool("-mat_use_hash_table", "Use hash table to save memory in constructing matrix", "MatSetOption", flg, &flg, NULL));
2231: if (flg) {
2232: PetscReal fact = 1.39;
2233: PetscCall(MatSetOption(B, MAT_USE_HASH_TABLE, PETSC_TRUE));
2234: PetscCall(PetscOptionsReal("-mat_use_hash_table", "Use hash table factor", "MatMPIBAIJSetHashTableFactor", fact, &fact, NULL));
2235: if (fact <= 1.0) fact = 1.39;
2236: PetscCall(MatMPIBAIJSetHashTableFactor(B, fact));
2237: PetscCall(PetscInfo(B, "Hash table Factor used %5.2g\n", (double)fact));
2238: }
2239: PetscOptionsEnd();
2240: PetscFunctionReturn(PETSC_SUCCESS);
2241: }
2243: // PetscClangLinter pragma disable: -fdoc-section-header-unknown
2244: /*MC
2245: MATSBAIJ - MATSBAIJ = "sbaij" - A matrix type to be used for symmetric block sparse matrices.
2247: This matrix type is identical to `MATSEQSBAIJ` when constructed with a single process communicator,
2248: and `MATMPISBAIJ` otherwise.
2250: Options Database Key:
2251: . -mat_type sbaij - sets the matrix type to `MATSBAIJ` during a call to `MatSetFromOptions()`
2253: Level: beginner
2255: .seealso: [](ch_matrices), `Mat`, `MATSEQSBAIJ`, `MATMPISBAIJ`, `MatCreateSBAIJ()`, `MATSEQSBAIJ`, `MATMPISBAIJ`
2256: M*/
2258: /*@C
2259: MatMPISBAIJSetPreallocation - For good matrix assembly performance
2260: the user should preallocate the matrix storage by setting the parameters
2261: d_nz (or d_nnz) and o_nz (or o_nnz). By setting these parameters accurately,
2262: performance can be increased by more than a factor of 50.
2264: Collective
2266: Input Parameters:
2267: + B - the matrix
2268: . bs - size of block, the blocks are ALWAYS square. One can use MatSetBlockSizes() to set a different row and column blocksize but the row
2269: blocksize always defines the size of the blocks. The column blocksize sets the blocksize of the vectors obtained with MatCreateVecs()
2270: . d_nz - number of block nonzeros per block row in diagonal portion of local
2271: submatrix (same for all local rows)
2272: . d_nnz - array containing the number of block nonzeros in the various block rows
2273: in the upper triangular and diagonal part of the in diagonal portion of the local
2274: (possibly different for each block row) or `NULL`. If you plan to factor the matrix you must leave room
2275: for the diagonal entry and set a value even if it is zero.
2276: . o_nz - number of block nonzeros per block row in the off-diagonal portion of local
2277: submatrix (same for all local rows).
2278: - o_nnz - array containing the number of nonzeros in the various block rows of the
2279: off-diagonal portion of the local submatrix that is right of the diagonal
2280: (possibly different for each block row) or `NULL`.
2282: Options Database Keys:
2283: + -mat_no_unroll - uses code that does not unroll the loops in the
2284: block calculations (much slower)
2285: - -mat_block_size - size of the blocks to use
2287: Level: intermediate
2289: Notes:
2291: If `PETSC_DECIDE` or `PETSC_DETERMINE` is used for a particular argument on one processor
2292: than it must be used on all processors that share the object for that argument.
2294: If the *_nnz parameter is given then the *_nz parameter is ignored
2296: Storage Information:
2297: For a square global matrix we define each processor's diagonal portion
2298: to be its local rows and the corresponding columns (a square submatrix);
2299: each processor's off-diagonal portion encompasses the remainder of the
2300: local matrix (a rectangular submatrix).
2302: The user can specify preallocated storage for the diagonal part of
2303: the local submatrix with either `d_nz` or `d_nnz` (not both). Set
2304: `d_nz` = `PETSC_DEFAULT` and `d_nnz` = `NULL` for PETSc to control dynamic
2305: memory allocation. Likewise, specify preallocated storage for the
2306: off-diagonal part of the local submatrix with `o_nz` or `o_nnz` (not both).
2308: You can call `MatGetInfo()` to get information on how effective the preallocation was;
2309: for example the fields mallocs,nz_allocated,nz_used,nz_unneeded;
2310: You can also run with the option `-info` and look for messages with the string
2311: malloc in them to see if additional memory allocation was needed.
2313: Consider a processor that owns rows 3, 4 and 5 of a parallel matrix. In
2314: the figure below we depict these three local rows and all columns (0-11).
2316: .vb
2317: 0 1 2 3 4 5 6 7 8 9 10 11
2318: --------------------------
2319: row 3 |. . . d d d o o o o o o
2320: row 4 |. . . d d d o o o o o o
2321: row 5 |. . . d d d o o o o o o
2322: --------------------------
2323: .ve
2325: Thus, any entries in the d locations are stored in the d (diagonal)
2326: submatrix, and any entries in the o locations are stored in the
2327: o (off-diagonal) submatrix. Note that the d matrix is stored in
2328: `MATSEQSBAIJ` format and the o submatrix in `MATSEQBAIJ` format.
2330: Now `d_nz` should indicate the number of block nonzeros per row in the upper triangular
2331: plus the diagonal part of the d matrix,
2332: and `o_nz` should indicate the number of block nonzeros per row in the o matrix
2334: In general, for PDE problems in which most nonzeros are near the diagonal,
2335: one expects `d_nz` >> `o_nz`.
2337: .seealso: [](ch_matrices), `Mat`, `MATMPISBAIJ`, `MATSBAIJ`, `MatCreate()`, `MatCreateSeqSBAIJ()`, `MatSetValues()`, `MatCreateBAIJ()`, `PetscSplitOwnership()`
2338: @*/
2339: PetscErrorCode MatMPISBAIJSetPreallocation(Mat B, PetscInt bs, PetscInt d_nz, const PetscInt d_nnz[], PetscInt o_nz, const PetscInt o_nnz[])
2340: {
2341: PetscFunctionBegin;
2345: PetscTryMethod(B, "MatMPISBAIJSetPreallocation_C", (Mat, PetscInt, PetscInt, const PetscInt[], PetscInt, const PetscInt[]), (B, bs, d_nz, d_nnz, o_nz, o_nnz));
2346: PetscFunctionReturn(PETSC_SUCCESS);
2347: }
2349: // PetscClangLinter pragma disable: -fdoc-section-header-unknown
2350: /*@C
2351: MatCreateSBAIJ - Creates a sparse parallel matrix in symmetric block AIJ format, `MATSBAIJ`,
2352: (block compressed row). For good matrix assembly performance
2353: the user should preallocate the matrix storage by setting the parameters
2354: `d_nz` (or `d_nnz`) and `o_nz` (or `o_nnz`).
2356: Collective
2358: Input Parameters:
2359: + comm - MPI communicator
2360: . bs - size of block, the blocks are ALWAYS square. One can use `MatSetBlockSizes()` to set a different row and column blocksize but the row
2361: blocksize always defines the size of the blocks. The column blocksize sets the blocksize of the vectors obtained with `MatCreateVecs()`
2362: . m - number of local rows (or `PETSC_DECIDE` to have calculated if `M` is given)
2363: This value should be the same as the local size used in creating the
2364: y vector for the matrix-vector product y = Ax.
2365: . n - number of local columns (or `PETSC_DECIDE` to have calculated if `N` is given)
2366: This value should be the same as the local size used in creating the
2367: x vector for the matrix-vector product y = Ax.
2368: . M - number of global rows (or `PETSC_DETERMINE` to have calculated if `m` is given)
2369: . N - number of global columns (or `PETSC_DETERMINE` to have calculated if `n` is given)
2370: . d_nz - number of block nonzeros per block row in diagonal portion of local
2371: submatrix (same for all local rows)
2372: . d_nnz - array containing the number of block nonzeros in the various block rows
2373: in the upper triangular portion of the in diagonal portion of the local
2374: (possibly different for each block block row) or `NULL`.
2375: If you plan to factor the matrix you must leave room for the diagonal entry and
2376: set its value even if it is zero.
2377: . o_nz - number of block nonzeros per block row in the off-diagonal portion of local
2378: submatrix (same for all local rows).
2379: - o_nnz - array containing the number of nonzeros in the various block rows of the
2380: off-diagonal portion of the local submatrix (possibly different for
2381: each block row) or `NULL`.
2383: Output Parameter:
2384: . A - the matrix
2386: Options Database Keys:
2387: + -mat_no_unroll - uses code that does not unroll the loops in the
2388: block calculations (much slower)
2389: . -mat_block_size - size of the blocks to use
2390: - -mat_mpi - use the parallel matrix data structures even on one processor
2391: (defaults to using SeqBAIJ format on one processor)
2393: Level: intermediate
2395: Notes:
2396: It is recommended that one use `MatCreateFromOptions()` or the `MatCreate()`, `MatSetType()` and/or `MatSetFromOptions()`,
2397: MatXXXXSetPreallocation() paradigm instead of this routine directly.
2398: [MatXXXXSetPreallocation() is, for example, `MatSeqAIJSetPreallocation()`]
2400: The number of rows and columns must be divisible by blocksize.
2401: This matrix type does not support complex Hermitian operation.
2403: The user MUST specify either the local or global matrix dimensions
2404: (possibly both).
2406: If `PETSC_DECIDE` or `PETSC_DETERMINE` is used for a particular argument on one processor
2407: than it must be used on all processors that share the object for that argument.
2409: If the *_nnz parameter is given then the *_nz parameter is ignored
2411: Storage Information:
2412: For a square global matrix we define each processor's diagonal portion
2413: to be its local rows and the corresponding columns (a square submatrix);
2414: each processor's off-diagonal portion encompasses the remainder of the
2415: local matrix (a rectangular submatrix).
2417: The user can specify preallocated storage for the diagonal part of
2418: the local submatrix with either `d_nz` or `d_nnz` (not both). Set
2419: `d_nz` = `PETSC_DEFAULT` and `d_nnz` = `NULL` for PETSc to control dynamic
2420: memory allocation. Likewise, specify preallocated storage for the
2421: off-diagonal part of the local submatrix with `o_nz` or `o_nnz` (not both).
2423: Consider a processor that owns rows 3, 4 and 5 of a parallel matrix. In
2424: the figure below we depict these three local rows and all columns (0-11).
2426: .vb
2427: 0 1 2 3 4 5 6 7 8 9 10 11
2428: --------------------------
2429: row 3 |. . . d d d o o o o o o
2430: row 4 |. . . d d d o o o o o o
2431: row 5 |. . . d d d o o o o o o
2432: --------------------------
2433: .ve
2435: Thus, any entries in the d locations are stored in the d (diagonal)
2436: submatrix, and any entries in the o locations are stored in the
2437: o (off-diagonal) submatrix. Note that the d matrix is stored in
2438: `MATSEQSBAIJ` format and the o submatrix in `MATSEQBAIJ` format.
2440: Now `d_nz` should indicate the number of block nonzeros per row in the upper triangular
2441: plus the diagonal part of the d matrix,
2442: and `o_nz` should indicate the number of block nonzeros per row in the o matrix.
2443: In general, for PDE problems in which most nonzeros are near the diagonal,
2444: one expects `d_nz` >> `o_nz`.
2446: .seealso: [](ch_matrices), `Mat`, `MATSBAIJ`, `MatCreate()`, `MatCreateSeqSBAIJ()`, `MatSetValues()`, `MatCreateBAIJ()`
2447: @*/
2448: PetscErrorCode MatCreateSBAIJ(MPI_Comm comm, PetscInt bs, PetscInt m, PetscInt n, PetscInt M, PetscInt N, PetscInt d_nz, const PetscInt d_nnz[], PetscInt o_nz, const PetscInt o_nnz[], Mat *A)
2449: {
2450: PetscMPIInt size;
2452: PetscFunctionBegin;
2453: PetscCall(MatCreate(comm, A));
2454: PetscCall(MatSetSizes(*A, m, n, M, N));
2455: PetscCallMPI(MPI_Comm_size(comm, &size));
2456: if (size > 1) {
2457: PetscCall(MatSetType(*A, MATMPISBAIJ));
2458: PetscCall(MatMPISBAIJSetPreallocation(*A, bs, d_nz, d_nnz, o_nz, o_nnz));
2459: } else {
2460: PetscCall(MatSetType(*A, MATSEQSBAIJ));
2461: PetscCall(MatSeqSBAIJSetPreallocation(*A, bs, d_nz, d_nnz));
2462: }
2463: PetscFunctionReturn(PETSC_SUCCESS);
2464: }
2466: static PetscErrorCode MatDuplicate_MPISBAIJ(Mat matin, MatDuplicateOption cpvalues, Mat *newmat)
2467: {
2468: Mat mat;
2469: Mat_MPISBAIJ *a, *oldmat = (Mat_MPISBAIJ *)matin->data;
2470: PetscInt len = 0, nt, bs = matin->rmap->bs, mbs = oldmat->mbs;
2471: PetscScalar *array;
2473: PetscFunctionBegin;
2474: *newmat = NULL;
2476: PetscCall(MatCreate(PetscObjectComm((PetscObject)matin), &mat));
2477: PetscCall(MatSetSizes(mat, matin->rmap->n, matin->cmap->n, matin->rmap->N, matin->cmap->N));
2478: PetscCall(MatSetType(mat, ((PetscObject)matin)->type_name));
2479: PetscCall(PetscLayoutReference(matin->rmap, &mat->rmap));
2480: PetscCall(PetscLayoutReference(matin->cmap, &mat->cmap));
2482: if (matin->hash_active) {
2483: PetscCall(MatSetUp(mat));
2484: } else {
2485: mat->factortype = matin->factortype;
2486: mat->preallocated = PETSC_TRUE;
2487: mat->assembled = PETSC_TRUE;
2488: mat->insertmode = NOT_SET_VALUES;
2490: a = (Mat_MPISBAIJ *)mat->data;
2491: a->bs2 = oldmat->bs2;
2492: a->mbs = oldmat->mbs;
2493: a->nbs = oldmat->nbs;
2494: a->Mbs = oldmat->Mbs;
2495: a->Nbs = oldmat->Nbs;
2497: a->size = oldmat->size;
2498: a->rank = oldmat->rank;
2499: a->donotstash = oldmat->donotstash;
2500: a->roworiented = oldmat->roworiented;
2501: a->rowindices = NULL;
2502: a->rowvalues = NULL;
2503: a->getrowactive = PETSC_FALSE;
2504: a->barray = NULL;
2505: a->rstartbs = oldmat->rstartbs;
2506: a->rendbs = oldmat->rendbs;
2507: a->cstartbs = oldmat->cstartbs;
2508: a->cendbs = oldmat->cendbs;
2510: /* hash table stuff */
2511: a->ht = NULL;
2512: a->hd = NULL;
2513: a->ht_size = 0;
2514: a->ht_flag = oldmat->ht_flag;
2515: a->ht_fact = oldmat->ht_fact;
2516: a->ht_total_ct = 0;
2517: a->ht_insert_ct = 0;
2519: PetscCall(PetscArraycpy(a->rangebs, oldmat->rangebs, a->size + 2));
2520: if (oldmat->colmap) {
2521: #if defined(PETSC_USE_CTABLE)
2522: PetscCall(PetscHMapIDuplicate(oldmat->colmap, &a->colmap));
2523: #else
2524: PetscCall(PetscMalloc1(a->Nbs, &a->colmap));
2525: PetscCall(PetscArraycpy(a->colmap, oldmat->colmap, a->Nbs));
2526: #endif
2527: } else a->colmap = NULL;
2529: if (oldmat->garray && (len = ((Mat_SeqBAIJ *)oldmat->B->data)->nbs)) {
2530: PetscCall(PetscMalloc1(len, &a->garray));
2531: PetscCall(PetscArraycpy(a->garray, oldmat->garray, len));
2532: } else a->garray = NULL;
2534: PetscCall(MatStashCreate_Private(PetscObjectComm((PetscObject)matin), matin->rmap->bs, &mat->bstash));
2535: PetscCall(VecDuplicate(oldmat->lvec, &a->lvec));
2536: PetscCall(VecScatterCopy(oldmat->Mvctx, &a->Mvctx));
2538: PetscCall(VecDuplicate(oldmat->slvec0, &a->slvec0));
2539: PetscCall(VecDuplicate(oldmat->slvec1, &a->slvec1));
2541: PetscCall(VecGetLocalSize(a->slvec1, &nt));
2542: PetscCall(VecGetArray(a->slvec1, &array));
2543: PetscCall(VecCreateSeqWithArray(PETSC_COMM_SELF, 1, bs * mbs, array, &a->slvec1a));
2544: PetscCall(VecCreateSeqWithArray(PETSC_COMM_SELF, 1, nt - bs * mbs, array + bs * mbs, &a->slvec1b));
2545: PetscCall(VecRestoreArray(a->slvec1, &array));
2546: PetscCall(VecGetArray(a->slvec0, &array));
2547: PetscCall(VecCreateSeqWithArray(PETSC_COMM_SELF, 1, nt - bs * mbs, array + bs * mbs, &a->slvec0b));
2548: PetscCall(VecRestoreArray(a->slvec0, &array));
2550: /* ierr = VecScatterCopy(oldmat->sMvctx,&a->sMvctx); - not written yet, replaced by the lazy trick: */
2551: PetscCall(PetscObjectReference((PetscObject)oldmat->sMvctx));
2552: a->sMvctx = oldmat->sMvctx;
2554: PetscCall(MatDuplicate(oldmat->A, cpvalues, &a->A));
2555: PetscCall(MatDuplicate(oldmat->B, cpvalues, &a->B));
2556: }
2557: PetscCall(PetscFunctionListDuplicate(((PetscObject)matin)->qlist, &((PetscObject)mat)->qlist));
2558: *newmat = mat;
2559: PetscFunctionReturn(PETSC_SUCCESS);
2560: }
2562: /* Used for both MPIBAIJ and MPISBAIJ matrices */
2563: #define MatLoad_MPISBAIJ_Binary MatLoad_MPIBAIJ_Binary
2565: static PetscErrorCode MatLoad_MPISBAIJ(Mat mat, PetscViewer viewer)
2566: {
2567: PetscBool isbinary;
2569: PetscFunctionBegin;
2570: PetscCall(PetscObjectTypeCompare((PetscObject)viewer, PETSCVIEWERBINARY, &isbinary));
2571: 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);
2572: PetscCall(MatLoad_MPISBAIJ_Binary(mat, viewer));
2573: PetscFunctionReturn(PETSC_SUCCESS);
2574: }
2576: static PetscErrorCode MatGetRowMaxAbs_MPISBAIJ(Mat A, Vec v, PetscInt idx[])
2577: {
2578: Mat_MPISBAIJ *a = (Mat_MPISBAIJ *)A->data;
2579: Mat_SeqBAIJ *b = (Mat_SeqBAIJ *)a->B->data;
2580: PetscReal atmp;
2581: PetscReal *work, *svalues, *rvalues;
2582: PetscInt i, bs, mbs, *bi, *bj, brow, j, ncols, krow, kcol, col, row, Mbs, bcol;
2583: PetscMPIInt rank, size;
2584: PetscInt *rowners_bs, dest, count, source;
2585: PetscScalar *va;
2586: MatScalar *ba;
2587: MPI_Status stat;
2589: PetscFunctionBegin;
2590: PetscCheck(!idx, PETSC_COMM_SELF, PETSC_ERR_SUP, "Send email to petsc-maint@mcs.anl.gov");
2591: PetscCall(MatGetRowMaxAbs(a->A, v, NULL));
2592: PetscCall(VecGetArray(v, &va));
2594: PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)A), &size));
2595: PetscCallMPI(MPI_Comm_rank(PetscObjectComm((PetscObject)A), &rank));
2597: bs = A->rmap->bs;
2598: mbs = a->mbs;
2599: Mbs = a->Mbs;
2600: ba = b->a;
2601: bi = b->i;
2602: bj = b->j;
2604: /* find ownerships */
2605: rowners_bs = A->rmap->range;
2607: /* each proc creates an array to be distributed */
2608: PetscCall(PetscCalloc1(bs * Mbs, &work));
2610: /* row_max for B */
2611: if (rank != size - 1) {
2612: for (i = 0; i < mbs; i++) {
2613: ncols = bi[1] - bi[0];
2614: bi++;
2615: brow = bs * i;
2616: for (j = 0; j < ncols; j++) {
2617: bcol = bs * (*bj);
2618: for (kcol = 0; kcol < bs; kcol++) {
2619: col = bcol + kcol; /* local col index */
2620: col += rowners_bs[rank + 1]; /* global col index */
2621: for (krow = 0; krow < bs; krow++) {
2622: atmp = PetscAbsScalar(*ba);
2623: ba++;
2624: row = brow + krow; /* local row index */
2625: if (PetscRealPart(va[row]) < atmp) va[row] = atmp;
2626: if (work[col] < atmp) work[col] = atmp;
2627: }
2628: }
2629: bj++;
2630: }
2631: }
2633: /* send values to its owners */
2634: for (dest = rank + 1; dest < size; dest++) {
2635: svalues = work + rowners_bs[dest];
2636: count = rowners_bs[dest + 1] - rowners_bs[dest];
2637: PetscCallMPI(MPI_Send(svalues, count, MPIU_REAL, dest, rank, PetscObjectComm((PetscObject)A)));
2638: }
2639: }
2641: /* receive values */
2642: if (rank) {
2643: rvalues = work;
2644: count = rowners_bs[rank + 1] - rowners_bs[rank];
2645: for (source = 0; source < rank; source++) {
2646: PetscCallMPI(MPI_Recv(rvalues, count, MPIU_REAL, MPI_ANY_SOURCE, MPI_ANY_TAG, PetscObjectComm((PetscObject)A), &stat));
2647: /* process values */
2648: for (i = 0; i < count; i++) {
2649: if (PetscRealPart(va[i]) < rvalues[i]) va[i] = rvalues[i];
2650: }
2651: }
2652: }
2654: PetscCall(VecRestoreArray(v, &va));
2655: PetscCall(PetscFree(work));
2656: PetscFunctionReturn(PETSC_SUCCESS);
2657: }
2659: static PetscErrorCode MatSOR_MPISBAIJ(Mat matin, Vec bb, PetscReal omega, MatSORType flag, PetscReal fshift, PetscInt its, PetscInt lits, Vec xx)
2660: {
2661: Mat_MPISBAIJ *mat = (Mat_MPISBAIJ *)matin->data;
2662: PetscInt mbs = mat->mbs, bs = matin->rmap->bs;
2663: PetscScalar *x, *ptr, *from;
2664: Vec bb1;
2665: const PetscScalar *b;
2667: PetscFunctionBegin;
2668: PetscCheck(its > 0 && lits > 0, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONG, "Relaxation requires global its %" PetscInt_FMT " and local its %" PetscInt_FMT " both positive", its, lits);
2669: PetscCheck(bs <= 1, PETSC_COMM_SELF, PETSC_ERR_SUP, "SSOR for block size > 1 is not yet implemented");
2671: if (flag == SOR_APPLY_UPPER) {
2672: PetscCall((*mat->A->ops->sor)(mat->A, bb, omega, flag, fshift, lits, 1, xx));
2673: PetscFunctionReturn(PETSC_SUCCESS);
2674: }
2676: if ((flag & SOR_LOCAL_SYMMETRIC_SWEEP) == SOR_LOCAL_SYMMETRIC_SWEEP) {
2677: if (flag & SOR_ZERO_INITIAL_GUESS) {
2678: PetscCall((*mat->A->ops->sor)(mat->A, bb, omega, flag, fshift, lits, lits, xx));
2679: its--;
2680: }
2682: PetscCall(VecDuplicate(bb, &bb1));
2683: while (its--) {
2684: /* lower triangular part: slvec0b = - B^T*xx */
2685: PetscCall((*mat->B->ops->multtranspose)(mat->B, xx, mat->slvec0b));
2687: /* copy xx into slvec0a */
2688: PetscCall(VecGetArray(mat->slvec0, &ptr));
2689: PetscCall(VecGetArray(xx, &x));
2690: PetscCall(PetscArraycpy(ptr, x, bs * mbs));
2691: PetscCall(VecRestoreArray(mat->slvec0, &ptr));
2693: PetscCall(VecScale(mat->slvec0, -1.0));
2695: /* copy bb into slvec1a */
2696: PetscCall(VecGetArray(mat->slvec1, &ptr));
2697: PetscCall(VecGetArrayRead(bb, &b));
2698: PetscCall(PetscArraycpy(ptr, b, bs * mbs));
2699: PetscCall(VecRestoreArray(mat->slvec1, &ptr));
2701: /* set slvec1b = 0 */
2702: PetscCall(PetscObjectStateIncrease((PetscObject)mat->slvec1b));
2703: PetscCall(VecZeroEntries(mat->slvec1b));
2705: PetscCall(VecScatterBegin(mat->sMvctx, mat->slvec0, mat->slvec1, ADD_VALUES, SCATTER_FORWARD));
2706: PetscCall(VecRestoreArray(xx, &x));
2707: PetscCall(VecRestoreArrayRead(bb, &b));
2708: PetscCall(VecScatterEnd(mat->sMvctx, mat->slvec0, mat->slvec1, ADD_VALUES, SCATTER_FORWARD));
2710: /* upper triangular part: bb1 = bb1 - B*x */
2711: PetscCall((*mat->B->ops->multadd)(mat->B, mat->slvec1b, mat->slvec1a, bb1));
2713: /* local diagonal sweep */
2714: PetscCall((*mat->A->ops->sor)(mat->A, bb1, omega, SOR_SYMMETRIC_SWEEP, fshift, lits, lits, xx));
2715: }
2716: PetscCall(VecDestroy(&bb1));
2717: } else if ((flag & SOR_LOCAL_FORWARD_SWEEP) && (its == 1) && (flag & SOR_ZERO_INITIAL_GUESS)) {
2718: PetscCall((*mat->A->ops->sor)(mat->A, bb, omega, flag, fshift, lits, 1, xx));
2719: } else if ((flag & SOR_LOCAL_BACKWARD_SWEEP) && (its == 1) && (flag & SOR_ZERO_INITIAL_GUESS)) {
2720: PetscCall((*mat->A->ops->sor)(mat->A, bb, omega, flag, fshift, lits, 1, xx));
2721: } else if (flag & SOR_EISENSTAT) {
2722: Vec xx1;
2723: PetscBool hasop;
2724: const PetscScalar *diag;
2725: PetscScalar *sl, scale = (omega - 2.0) / omega;
2726: PetscInt i, n;
2728: if (!mat->xx1) {
2729: PetscCall(VecDuplicate(bb, &mat->xx1));
2730: PetscCall(VecDuplicate(bb, &mat->bb1));
2731: }
2732: xx1 = mat->xx1;
2733: bb1 = mat->bb1;
2735: PetscCall((*mat->A->ops->sor)(mat->A, bb, omega, (MatSORType)(SOR_ZERO_INITIAL_GUESS | SOR_LOCAL_BACKWARD_SWEEP), fshift, lits, 1, xx));
2737: if (!mat->diag) {
2738: /* this is wrong for same matrix with new nonzero values */
2739: PetscCall(MatCreateVecs(matin, &mat->diag, NULL));
2740: PetscCall(MatGetDiagonal(matin, mat->diag));
2741: }
2742: PetscCall(MatHasOperation(matin, MATOP_MULT_DIAGONAL_BLOCK, &hasop));
2744: if (hasop) {
2745: PetscCall(MatMultDiagonalBlock(matin, xx, bb1));
2746: PetscCall(VecAYPX(mat->slvec1a, scale, bb));
2747: } else {
2748: /*
2749: These two lines are replaced by code that may be a bit faster for a good compiler
2750: PetscCall(VecPointwiseMult(mat->slvec1a,mat->diag,xx));
2751: PetscCall(VecAYPX(mat->slvec1a,scale,bb));
2752: */
2753: PetscCall(VecGetArray(mat->slvec1a, &sl));
2754: PetscCall(VecGetArrayRead(mat->diag, &diag));
2755: PetscCall(VecGetArrayRead(bb, &b));
2756: PetscCall(VecGetArray(xx, &x));
2757: PetscCall(VecGetLocalSize(xx, &n));
2758: if (omega == 1.0) {
2759: for (i = 0; i < n; i++) sl[i] = b[i] - diag[i] * x[i];
2760: PetscCall(PetscLogFlops(2.0 * n));
2761: } else {
2762: for (i = 0; i < n; i++) sl[i] = b[i] + scale * diag[i] * x[i];
2763: PetscCall(PetscLogFlops(3.0 * n));
2764: }
2765: PetscCall(VecRestoreArray(mat->slvec1a, &sl));
2766: PetscCall(VecRestoreArrayRead(mat->diag, &diag));
2767: PetscCall(VecRestoreArrayRead(bb, &b));
2768: PetscCall(VecRestoreArray(xx, &x));
2769: }
2771: /* multiply off-diagonal portion of matrix */
2772: PetscCall(PetscObjectStateIncrease((PetscObject)mat->slvec1b));
2773: PetscCall(VecZeroEntries(mat->slvec1b));
2774: PetscCall((*mat->B->ops->multtranspose)(mat->B, xx, mat->slvec0b));
2775: PetscCall(VecGetArray(mat->slvec0, &from));
2776: PetscCall(VecGetArray(xx, &x));
2777: PetscCall(PetscArraycpy(from, x, bs * mbs));
2778: PetscCall(VecRestoreArray(mat->slvec0, &from));
2779: PetscCall(VecRestoreArray(xx, &x));
2780: PetscCall(VecScatterBegin(mat->sMvctx, mat->slvec0, mat->slvec1, ADD_VALUES, SCATTER_FORWARD));
2781: PetscCall(VecScatterEnd(mat->sMvctx, mat->slvec0, mat->slvec1, ADD_VALUES, SCATTER_FORWARD));
2782: PetscCall((*mat->B->ops->multadd)(mat->B, mat->slvec1b, mat->slvec1a, mat->slvec1a));
2784: /* local sweep */
2785: PetscCall((*mat->A->ops->sor)(mat->A, mat->slvec1a, omega, (MatSORType)(SOR_ZERO_INITIAL_GUESS | SOR_LOCAL_FORWARD_SWEEP), fshift, lits, 1, xx1));
2786: PetscCall(VecAXPY(xx, 1.0, xx1));
2787: } else SETERRQ(PETSC_COMM_SELF, PETSC_ERR_SUP, "MatSORType is not supported for SBAIJ matrix format");
2788: PetscFunctionReturn(PETSC_SUCCESS);
2789: }
2791: /*@
2792: MatCreateMPISBAIJWithArrays - creates a `MATMPISBAIJ` matrix using arrays that contain in standard CSR format for the local rows.
2794: Collective
2796: Input Parameters:
2797: + comm - MPI communicator
2798: . bs - the block size, only a block size of 1 is supported
2799: . m - number of local rows (Cannot be `PETSC_DECIDE`)
2800: . n - This value should be the same as the local size used in creating the
2801: x vector for the matrix-vector product $ y = Ax $. (or `PETSC_DECIDE` to have
2802: calculated if `N` is given) For square matrices `n` is almost always `m`.
2803: . M - number of global rows (or `PETSC_DETERMINE` to have calculated if `m` is given)
2804: . N - number of global columns (or `PETSC_DETERMINE` to have calculated if `n` is given)
2805: . 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
2806: . j - column indices
2807: - a - matrix values
2809: Output Parameter:
2810: . mat - the matrix
2812: Level: intermediate
2814: Notes:
2815: The `i`, `j`, and `a` arrays ARE copied by this routine into the internal format used by PETSc;
2816: thus you CANNOT change the matrix entries by changing the values of `a` after you have
2817: called this routine. Use `MatCreateMPIAIJWithSplitArrays()` to avoid needing to copy the arrays.
2819: The `i` and `j` indices are 0 based, and `i` indices are indices corresponding to the local `j` array.
2821: .seealso: [](ch_matrices), `Mat`, `MATMPISBAIJ`, `MatCreate()`, `MatCreateSeqAIJ()`, `MatSetValues()`, `MatMPIAIJSetPreallocation()`, `MatMPIAIJSetPreallocationCSR()`,
2822: `MATMPIAIJ`, `MatCreateAIJ()`, `MatCreateMPIAIJWithSplitArrays()`, `MatMPISBAIJSetPreallocationCSR()`
2823: @*/
2824: PetscErrorCode MatCreateMPISBAIJWithArrays(MPI_Comm comm, PetscInt bs, PetscInt m, PetscInt n, PetscInt M, PetscInt N, const PetscInt i[], const PetscInt j[], const PetscScalar a[], Mat *mat)
2825: {
2826: PetscFunctionBegin;
2827: PetscCheck(!i[0], PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "i (row indices) must start with 0");
2828: PetscCheck(m >= 0, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "local number of rows (m) cannot be PETSC_DECIDE, or negative");
2829: PetscCall(MatCreate(comm, mat));
2830: PetscCall(MatSetSizes(*mat, m, n, M, N));
2831: PetscCall(MatSetType(*mat, MATMPISBAIJ));
2832: PetscCall(MatMPISBAIJSetPreallocationCSR(*mat, bs, i, j, a));
2833: PetscFunctionReturn(PETSC_SUCCESS);
2834: }
2836: /*@C
2837: MatMPISBAIJSetPreallocationCSR - Creates a sparse parallel matrix in `MATMPISBAIJ` format using the given nonzero structure and (optional) numerical values
2839: Collective
2841: Input Parameters:
2842: + B - the matrix
2843: . bs - the block size
2844: . i - the indices into `j` for the start of each local row (indices start with zero)
2845: . j - the column indices for each local row (indices start with zero) these must be sorted for each row
2846: - v - optional values in the matrix, pass `NULL` if not provided
2848: Level: advanced
2850: Notes:
2851: The `i`, `j`, and `v` arrays ARE copied by this routine into the internal format used by PETSc;
2852: thus you CANNOT change the matrix entries by changing the values of `v` after you have
2853: called this routine.
2855: Though this routine has Preallocation() in the name it also sets the exact nonzero locations of the matrix entries
2856: and usually the numerical values as well
2858: Any entries passed in that are below the diagonal are ignored
2860: .seealso: [](ch_matrices), `Mat`, `MATMPISBAIJ`, `MatCreate()`, `MatCreateSeqAIJ()`, `MatSetValues()`, `MatMPIBAIJSetPreallocation()`, `MatCreateAIJ()`, `MATMPIAIJ`,
2861: `MatCreateMPISBAIJWithArrays()`
2862: @*/
2863: PetscErrorCode MatMPISBAIJSetPreallocationCSR(Mat B, PetscInt bs, const PetscInt i[], const PetscInt j[], const PetscScalar v[])
2864: {
2865: PetscFunctionBegin;
2866: PetscTryMethod(B, "MatMPISBAIJSetPreallocationCSR_C", (Mat, PetscInt, const PetscInt[], const PetscInt[], const PetscScalar[]), (B, bs, i, j, v));
2867: PetscFunctionReturn(PETSC_SUCCESS);
2868: }
2870: PetscErrorCode MatCreateMPIMatConcatenateSeqMat_MPISBAIJ(MPI_Comm comm, Mat inmat, PetscInt n, MatReuse scall, Mat *outmat)
2871: {
2872: PetscInt m, N, i, rstart, nnz, Ii, bs, cbs;
2873: PetscInt *indx;
2874: PetscScalar *values;
2876: PetscFunctionBegin;
2877: PetscCall(MatGetSize(inmat, &m, &N));
2878: if (scall == MAT_INITIAL_MATRIX) { /* symbolic phase */
2879: Mat_SeqSBAIJ *a = (Mat_SeqSBAIJ *)inmat->data;
2880: PetscInt *dnz, *onz, mbs, Nbs, nbs;
2881: PetscInt *bindx, rmax = a->rmax, j;
2882: PetscMPIInt rank, size;
2884: PetscCall(MatGetBlockSizes(inmat, &bs, &cbs));
2885: mbs = m / bs;
2886: Nbs = N / cbs;
2887: if (n == PETSC_DECIDE) PetscCall(PetscSplitOwnershipBlock(comm, cbs, &n, &N));
2888: nbs = n / cbs;
2890: PetscCall(PetscMalloc1(rmax, &bindx));
2891: MatPreallocateBegin(comm, mbs, nbs, dnz, onz); /* inline function, output __end and __rstart are used below */
2893: PetscCallMPI(MPI_Comm_rank(comm, &rank));
2894: PetscCallMPI(MPI_Comm_rank(comm, &size));
2895: if (rank == size - 1) {
2896: /* Check sum(nbs) = Nbs */
2897: PetscCheck(__end == Nbs, PETSC_COMM_SELF, PETSC_ERR_ARG_INCOMP, "Sum of local block columns %" PetscInt_FMT " != global block columns %" PetscInt_FMT, __end, Nbs);
2898: }
2900: rstart = __rstart; /* block rstart of *outmat; see inline function MatPreallocateBegin */
2901: PetscCall(MatSetOption(inmat, MAT_GETROW_UPPERTRIANGULAR, PETSC_TRUE));
2902: for (i = 0; i < mbs; i++) {
2903: PetscCall(MatGetRow_SeqSBAIJ(inmat, i * bs, &nnz, &indx, NULL)); /* non-blocked nnz and indx */
2904: nnz = nnz / bs;
2905: for (j = 0; j < nnz; j++) bindx[j] = indx[j * bs] / bs;
2906: PetscCall(MatPreallocateSet(i + rstart, nnz, bindx, dnz, onz));
2907: PetscCall(MatRestoreRow_SeqSBAIJ(inmat, i * bs, &nnz, &indx, NULL));
2908: }
2909: PetscCall(MatSetOption(inmat, MAT_GETROW_UPPERTRIANGULAR, PETSC_FALSE));
2910: PetscCall(PetscFree(bindx));
2912: PetscCall(MatCreate(comm, outmat));
2913: PetscCall(MatSetSizes(*outmat, m, n, PETSC_DETERMINE, PETSC_DETERMINE));
2914: PetscCall(MatSetBlockSizes(*outmat, bs, cbs));
2915: PetscCall(MatSetType(*outmat, MATSBAIJ));
2916: PetscCall(MatSeqSBAIJSetPreallocation(*outmat, bs, 0, dnz));
2917: PetscCall(MatMPISBAIJSetPreallocation(*outmat, bs, 0, dnz, 0, onz));
2918: MatPreallocateEnd(dnz, onz);
2919: }
2921: /* numeric phase */
2922: PetscCall(MatGetBlockSizes(inmat, &bs, &cbs));
2923: PetscCall(MatGetOwnershipRange(*outmat, &rstart, NULL));
2925: PetscCall(MatSetOption(inmat, MAT_GETROW_UPPERTRIANGULAR, PETSC_TRUE));
2926: for (i = 0; i < m; i++) {
2927: PetscCall(MatGetRow_SeqSBAIJ(inmat, i, &nnz, &indx, &values));
2928: Ii = i + rstart;
2929: PetscCall(MatSetValues(*outmat, 1, &Ii, nnz, indx, values, INSERT_VALUES));
2930: PetscCall(MatRestoreRow_SeqSBAIJ(inmat, i, &nnz, &indx, &values));
2931: }
2932: PetscCall(MatSetOption(inmat, MAT_GETROW_UPPERTRIANGULAR, PETSC_FALSE));
2933: PetscCall(MatAssemblyBegin(*outmat, MAT_FINAL_ASSEMBLY));
2934: PetscCall(MatAssemblyEnd(*outmat, MAT_FINAL_ASSEMBLY));
2935: PetscFunctionReturn(PETSC_SUCCESS);
2936: }