xref: /petsc/src/mat/impls/aij/seq/aij.c (revision f13dfd9ea68e0ddeee984e65c377a1819eab8a8a)
1 /*
2     Defines the basic matrix operations for the AIJ (compressed row)
3   matrix storage format.
4 */
5 
6 #include <../src/mat/impls/aij/seq/aij.h> /*I "petscmat.h" I*/
7 #include <petscblaslapack.h>
8 #include <petscbt.h>
9 #include <petsc/private/kernels/blocktranspose.h>
10 
11 /* defines MatSetValues_Seq_Hash(), MatAssemblyEnd_Seq_Hash(), MatSetUp_Seq_Hash() */
12 #define TYPE AIJ
13 #define TYPE_BS
14 #include "../src/mat/impls/aij/seq/seqhashmatsetvalues.h"
15 #include "../src/mat/impls/aij/seq/seqhashmat.h"
16 #undef TYPE
17 #undef TYPE_BS
18 
19 static PetscErrorCode MatSeqAIJSetTypeFromOptions(Mat A)
20 {
21   PetscBool flg;
22   char      type[256];
23 
24   PetscFunctionBegin;
25   PetscObjectOptionsBegin((PetscObject)A);
26   PetscCall(PetscOptionsFList("-mat_seqaij_type", "Matrix SeqAIJ type", "MatSeqAIJSetType", MatSeqAIJList, "seqaij", type, 256, &flg));
27   if (flg) PetscCall(MatSeqAIJSetType(A, type));
28   PetscOptionsEnd();
29   PetscFunctionReturn(PETSC_SUCCESS);
30 }
31 
32 static PetscErrorCode MatGetColumnReductions_SeqAIJ(Mat A, PetscInt type, PetscReal *reductions)
33 {
34   PetscInt    i, m, n;
35   Mat_SeqAIJ *aij = (Mat_SeqAIJ *)A->data;
36 
37   PetscFunctionBegin;
38   PetscCall(MatGetSize(A, &m, &n));
39   PetscCall(PetscArrayzero(reductions, n));
40   if (type == NORM_2) {
41     for (i = 0; i < aij->i[m]; i++) reductions[aij->j[i]] += PetscAbsScalar(aij->a[i] * aij->a[i]);
42   } else if (type == NORM_1) {
43     for (i = 0; i < aij->i[m]; i++) reductions[aij->j[i]] += PetscAbsScalar(aij->a[i]);
44   } else if (type == NORM_INFINITY) {
45     for (i = 0; i < aij->i[m]; i++) reductions[aij->j[i]] = PetscMax(PetscAbsScalar(aij->a[i]), reductions[aij->j[i]]);
46   } else if (type == REDUCTION_SUM_REALPART || type == REDUCTION_MEAN_REALPART) {
47     for (i = 0; i < aij->i[m]; i++) reductions[aij->j[i]] += PetscRealPart(aij->a[i]);
48   } else if (type == REDUCTION_SUM_IMAGINARYPART || type == REDUCTION_MEAN_IMAGINARYPART) {
49     for (i = 0; i < aij->i[m]; i++) reductions[aij->j[i]] += PetscImaginaryPart(aij->a[i]);
50   } else SETERRQ(PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONG, "Unknown reduction type");
51 
52   if (type == NORM_2) {
53     for (i = 0; i < n; i++) reductions[i] = PetscSqrtReal(reductions[i]);
54   } else if (type == REDUCTION_MEAN_REALPART || type == REDUCTION_MEAN_IMAGINARYPART) {
55     for (i = 0; i < n; i++) reductions[i] /= m;
56   }
57   PetscFunctionReturn(PETSC_SUCCESS);
58 }
59 
60 static PetscErrorCode MatFindOffBlockDiagonalEntries_SeqAIJ(Mat A, IS *is)
61 {
62   Mat_SeqAIJ     *a = (Mat_SeqAIJ *)A->data;
63   PetscInt        i, m = A->rmap->n, cnt = 0, bs = A->rmap->bs;
64   const PetscInt *jj = a->j, *ii = a->i;
65   PetscInt       *rows;
66 
67   PetscFunctionBegin;
68   for (i = 0; i < m; i++) {
69     if ((ii[i] != ii[i + 1]) && ((jj[ii[i]] < bs * (i / bs)) || (jj[ii[i + 1] - 1] > bs * ((i + bs) / bs) - 1))) cnt++;
70   }
71   PetscCall(PetscMalloc1(cnt, &rows));
72   cnt = 0;
73   for (i = 0; i < m; i++) {
74     if ((ii[i] != ii[i + 1]) && ((jj[ii[i]] < bs * (i / bs)) || (jj[ii[i + 1] - 1] > bs * ((i + bs) / bs) - 1))) {
75       rows[cnt] = i;
76       cnt++;
77     }
78   }
79   PetscCall(ISCreateGeneral(PETSC_COMM_SELF, cnt, rows, PETSC_OWN_POINTER, is));
80   PetscFunctionReturn(PETSC_SUCCESS);
81 }
82 
83 PetscErrorCode MatFindZeroDiagonals_SeqAIJ_Private(Mat A, PetscInt *nrows, PetscInt **zrows)
84 {
85   Mat_SeqAIJ      *a = (Mat_SeqAIJ *)A->data;
86   const MatScalar *aa;
87   PetscInt         i, m = A->rmap->n, cnt = 0;
88   const PetscInt  *ii = a->i, *jj = a->j, *diag;
89   PetscInt        *rows;
90 
91   PetscFunctionBegin;
92   PetscCall(MatSeqAIJGetArrayRead(A, &aa));
93   PetscCall(MatMarkDiagonal_SeqAIJ(A));
94   diag = a->diag;
95   for (i = 0; i < m; i++) {
96     if ((diag[i] >= ii[i + 1]) || (jj[diag[i]] != i) || (aa[diag[i]] == 0.0)) cnt++;
97   }
98   PetscCall(PetscMalloc1(cnt, &rows));
99   cnt = 0;
100   for (i = 0; i < m; i++) {
101     if ((diag[i] >= ii[i + 1]) || (jj[diag[i]] != i) || (aa[diag[i]] == 0.0)) rows[cnt++] = i;
102   }
103   *nrows = cnt;
104   *zrows = rows;
105   PetscCall(MatSeqAIJRestoreArrayRead(A, &aa));
106   PetscFunctionReturn(PETSC_SUCCESS);
107 }
108 
109 static PetscErrorCode MatFindZeroDiagonals_SeqAIJ(Mat A, IS *zrows)
110 {
111   PetscInt nrows, *rows;
112 
113   PetscFunctionBegin;
114   *zrows = NULL;
115   PetscCall(MatFindZeroDiagonals_SeqAIJ_Private(A, &nrows, &rows));
116   PetscCall(ISCreateGeneral(PetscObjectComm((PetscObject)A), nrows, rows, PETSC_OWN_POINTER, zrows));
117   PetscFunctionReturn(PETSC_SUCCESS);
118 }
119 
120 static PetscErrorCode MatFindNonzeroRows_SeqAIJ(Mat A, IS *keptrows)
121 {
122   Mat_SeqAIJ      *a = (Mat_SeqAIJ *)A->data;
123   const MatScalar *aa;
124   PetscInt         m = A->rmap->n, cnt = 0;
125   const PetscInt  *ii;
126   PetscInt         n, i, j, *rows;
127 
128   PetscFunctionBegin;
129   PetscCall(MatSeqAIJGetArrayRead(A, &aa));
130   *keptrows = NULL;
131   ii        = a->i;
132   for (i = 0; i < m; i++) {
133     n = ii[i + 1] - ii[i];
134     if (!n) {
135       cnt++;
136       goto ok1;
137     }
138     for (j = ii[i]; j < ii[i + 1]; j++) {
139       if (aa[j] != 0.0) goto ok1;
140     }
141     cnt++;
142   ok1:;
143   }
144   if (!cnt) {
145     PetscCall(MatSeqAIJRestoreArrayRead(A, &aa));
146     PetscFunctionReturn(PETSC_SUCCESS);
147   }
148   PetscCall(PetscMalloc1(A->rmap->n - cnt, &rows));
149   cnt = 0;
150   for (i = 0; i < m; i++) {
151     n = ii[i + 1] - ii[i];
152     if (!n) continue;
153     for (j = ii[i]; j < ii[i + 1]; j++) {
154       if (aa[j] != 0.0) {
155         rows[cnt++] = i;
156         break;
157       }
158     }
159   }
160   PetscCall(MatSeqAIJRestoreArrayRead(A, &aa));
161   PetscCall(ISCreateGeneral(PETSC_COMM_SELF, cnt, rows, PETSC_OWN_POINTER, keptrows));
162   PetscFunctionReturn(PETSC_SUCCESS);
163 }
164 
165 PetscErrorCode MatDiagonalSet_SeqAIJ(Mat Y, Vec D, InsertMode is)
166 {
167   Mat_SeqAIJ        *aij = (Mat_SeqAIJ *)Y->data;
168   PetscInt           i, m = Y->rmap->n;
169   const PetscInt    *diag;
170   MatScalar         *aa;
171   const PetscScalar *v;
172   PetscBool          missing;
173 
174   PetscFunctionBegin;
175   if (Y->assembled) {
176     PetscCall(MatMissingDiagonal_SeqAIJ(Y, &missing, NULL));
177     if (!missing) {
178       diag = aij->diag;
179       PetscCall(VecGetArrayRead(D, &v));
180       PetscCall(MatSeqAIJGetArray(Y, &aa));
181       if (is == INSERT_VALUES) {
182         for (i = 0; i < m; i++) aa[diag[i]] = v[i];
183       } else {
184         for (i = 0; i < m; i++) aa[diag[i]] += v[i];
185       }
186       PetscCall(MatSeqAIJRestoreArray(Y, &aa));
187       PetscCall(VecRestoreArrayRead(D, &v));
188       PetscFunctionReturn(PETSC_SUCCESS);
189     }
190     PetscCall(MatSeqAIJInvalidateDiagonal(Y));
191   }
192   PetscCall(MatDiagonalSet_Default(Y, D, is));
193   PetscFunctionReturn(PETSC_SUCCESS);
194 }
195 
196 PetscErrorCode MatGetRowIJ_SeqAIJ(Mat A, PetscInt oshift, PetscBool symmetric, PetscBool inodecompressed, PetscInt *m, const PetscInt *ia[], const PetscInt *ja[], PetscBool *done)
197 {
198   Mat_SeqAIJ *a = (Mat_SeqAIJ *)A->data;
199   PetscInt    i, ishift;
200 
201   PetscFunctionBegin;
202   if (m) *m = A->rmap->n;
203   if (!ia) PetscFunctionReturn(PETSC_SUCCESS);
204   ishift = 0;
205   if (symmetric && A->structurally_symmetric != PETSC_BOOL3_TRUE) {
206     PetscCall(MatToSymmetricIJ_SeqAIJ(A->rmap->n, a->i, a->j, PETSC_TRUE, ishift, oshift, (PetscInt **)ia, (PetscInt **)ja));
207   } else if (oshift == 1) {
208     PetscInt *tia;
209     PetscInt  nz = a->i[A->rmap->n];
210     /* malloc space and  add 1 to i and j indices */
211     PetscCall(PetscMalloc1(A->rmap->n + 1, &tia));
212     for (i = 0; i < A->rmap->n + 1; i++) tia[i] = a->i[i] + 1;
213     *ia = tia;
214     if (ja) {
215       PetscInt *tja;
216       PetscCall(PetscMalloc1(nz + 1, &tja));
217       for (i = 0; i < nz; i++) tja[i] = a->j[i] + 1;
218       *ja = tja;
219     }
220   } else {
221     *ia = a->i;
222     if (ja) *ja = a->j;
223   }
224   PetscFunctionReturn(PETSC_SUCCESS);
225 }
226 
227 PetscErrorCode MatRestoreRowIJ_SeqAIJ(Mat A, PetscInt oshift, PetscBool symmetric, PetscBool inodecompressed, PetscInt *n, const PetscInt *ia[], const PetscInt *ja[], PetscBool *done)
228 {
229   PetscFunctionBegin;
230   if (!ia) PetscFunctionReturn(PETSC_SUCCESS);
231   if ((symmetric && A->structurally_symmetric != PETSC_BOOL3_TRUE) || oshift == 1) {
232     PetscCall(PetscFree(*ia));
233     if (ja) PetscCall(PetscFree(*ja));
234   }
235   PetscFunctionReturn(PETSC_SUCCESS);
236 }
237 
238 PetscErrorCode MatGetColumnIJ_SeqAIJ(Mat A, PetscInt oshift, PetscBool symmetric, PetscBool inodecompressed, PetscInt *nn, const PetscInt *ia[], const PetscInt *ja[], PetscBool *done)
239 {
240   Mat_SeqAIJ *a = (Mat_SeqAIJ *)A->data;
241   PetscInt    i, *collengths, *cia, *cja, n = A->cmap->n, m = A->rmap->n;
242   PetscInt    nz = a->i[m], row, *jj, mr, col;
243 
244   PetscFunctionBegin;
245   *nn = n;
246   if (!ia) PetscFunctionReturn(PETSC_SUCCESS);
247   if (symmetric) {
248     PetscCall(MatToSymmetricIJ_SeqAIJ(A->rmap->n, a->i, a->j, PETSC_TRUE, 0, oshift, (PetscInt **)ia, (PetscInt **)ja));
249   } else {
250     PetscCall(PetscCalloc1(n, &collengths));
251     PetscCall(PetscMalloc1(n + 1, &cia));
252     PetscCall(PetscMalloc1(nz, &cja));
253     jj = a->j;
254     for (i = 0; i < nz; i++) collengths[jj[i]]++;
255     cia[0] = oshift;
256     for (i = 0; i < n; i++) cia[i + 1] = cia[i] + collengths[i];
257     PetscCall(PetscArrayzero(collengths, n));
258     jj = a->j;
259     for (row = 0; row < m; row++) {
260       mr = a->i[row + 1] - a->i[row];
261       for (i = 0; i < mr; i++) {
262         col = *jj++;
263 
264         cja[cia[col] + collengths[col]++ - oshift] = row + oshift;
265       }
266     }
267     PetscCall(PetscFree(collengths));
268     *ia = cia;
269     *ja = cja;
270   }
271   PetscFunctionReturn(PETSC_SUCCESS);
272 }
273 
274 PetscErrorCode MatRestoreColumnIJ_SeqAIJ(Mat A, PetscInt oshift, PetscBool symmetric, PetscBool inodecompressed, PetscInt *n, const PetscInt *ia[], const PetscInt *ja[], PetscBool *done)
275 {
276   PetscFunctionBegin;
277   if (!ia) PetscFunctionReturn(PETSC_SUCCESS);
278 
279   PetscCall(PetscFree(*ia));
280   PetscCall(PetscFree(*ja));
281   PetscFunctionReturn(PETSC_SUCCESS);
282 }
283 
284 /*
285  MatGetColumnIJ_SeqAIJ_Color() and MatRestoreColumnIJ_SeqAIJ_Color() are customized from
286  MatGetColumnIJ_SeqAIJ() and MatRestoreColumnIJ_SeqAIJ() by adding an output
287  spidx[], index of a->a, to be used in MatTransposeColoringCreate_SeqAIJ() and MatFDColoringCreate_SeqXAIJ()
288 */
289 PetscErrorCode MatGetColumnIJ_SeqAIJ_Color(Mat A, PetscInt oshift, PetscBool symmetric, PetscBool inodecompressed, PetscInt *nn, const PetscInt *ia[], const PetscInt *ja[], PetscInt *spidx[], PetscBool *done)
290 {
291   Mat_SeqAIJ     *a = (Mat_SeqAIJ *)A->data;
292   PetscInt        i, *collengths, *cia, *cja, n = A->cmap->n, m = A->rmap->n;
293   PetscInt        nz = a->i[m], row, mr, col, tmp;
294   PetscInt       *cspidx;
295   const PetscInt *jj;
296 
297   PetscFunctionBegin;
298   *nn = n;
299   if (!ia) PetscFunctionReturn(PETSC_SUCCESS);
300 
301   PetscCall(PetscCalloc1(n, &collengths));
302   PetscCall(PetscMalloc1(n + 1, &cia));
303   PetscCall(PetscMalloc1(nz, &cja));
304   PetscCall(PetscMalloc1(nz, &cspidx));
305   jj = a->j;
306   for (i = 0; i < nz; i++) collengths[jj[i]]++;
307   cia[0] = oshift;
308   for (i = 0; i < n; i++) cia[i + 1] = cia[i] + collengths[i];
309   PetscCall(PetscArrayzero(collengths, n));
310   jj = a->j;
311   for (row = 0; row < m; row++) {
312     mr = a->i[row + 1] - a->i[row];
313     for (i = 0; i < mr; i++) {
314       col         = *jj++;
315       tmp         = cia[col] + collengths[col]++ - oshift;
316       cspidx[tmp] = a->i[row] + i; /* index of a->j */
317       cja[tmp]    = row + oshift;
318     }
319   }
320   PetscCall(PetscFree(collengths));
321   *ia    = cia;
322   *ja    = cja;
323   *spidx = cspidx;
324   PetscFunctionReturn(PETSC_SUCCESS);
325 }
326 
327 PetscErrorCode MatRestoreColumnIJ_SeqAIJ_Color(Mat A, PetscInt oshift, PetscBool symmetric, PetscBool inodecompressed, PetscInt *n, const PetscInt *ia[], const PetscInt *ja[], PetscInt *spidx[], PetscBool *done)
328 {
329   PetscFunctionBegin;
330   PetscCall(MatRestoreColumnIJ_SeqAIJ(A, oshift, symmetric, inodecompressed, n, ia, ja, done));
331   PetscCall(PetscFree(*spidx));
332   PetscFunctionReturn(PETSC_SUCCESS);
333 }
334 
335 static PetscErrorCode MatSetValuesRow_SeqAIJ(Mat A, PetscInt row, const PetscScalar v[])
336 {
337   Mat_SeqAIJ  *a  = (Mat_SeqAIJ *)A->data;
338   PetscInt    *ai = a->i;
339   PetscScalar *aa;
340 
341   PetscFunctionBegin;
342   PetscCall(MatSeqAIJGetArray(A, &aa));
343   PetscCall(PetscArraycpy(aa + ai[row], v, ai[row + 1] - ai[row]));
344   PetscCall(MatSeqAIJRestoreArray(A, &aa));
345   PetscFunctionReturn(PETSC_SUCCESS);
346 }
347 
348 /*
349     MatSeqAIJSetValuesLocalFast - An optimized version of MatSetValuesLocal() for SeqAIJ matrices with several assumptions
350 
351       -   a single row of values is set with each call
352       -   no row or column indices are negative or (in error) larger than the number of rows or columns
353       -   the values are always added to the matrix, not set
354       -   no new locations are introduced in the nonzero structure of the matrix
355 
356      This does NOT assume the global column indices are sorted
357 
358 */
359 
360 #include <petsc/private/isimpl.h>
361 PetscErrorCode MatSeqAIJSetValuesLocalFast(Mat A, PetscInt m, const PetscInt im[], PetscInt n, const PetscInt in[], const PetscScalar v[], InsertMode is)
362 {
363   Mat_SeqAIJ     *a = (Mat_SeqAIJ *)A->data;
364   PetscInt        low, high, t, row, nrow, i, col, l;
365   const PetscInt *rp, *ai = a->i, *ailen = a->ilen, *aj = a->j;
366   PetscInt        lastcol = -1;
367   MatScalar      *ap, value, *aa;
368   const PetscInt *ridx = A->rmap->mapping->indices, *cidx = A->cmap->mapping->indices;
369 
370   PetscFunctionBegin;
371   PetscCall(MatSeqAIJGetArray(A, &aa));
372   row  = ridx[im[0]];
373   rp   = aj + ai[row];
374   ap   = aa + ai[row];
375   nrow = ailen[row];
376   low  = 0;
377   high = nrow;
378   for (l = 0; l < n; l++) { /* loop over added columns */
379     col   = cidx[in[l]];
380     value = v[l];
381 
382     if (col <= lastcol) low = 0;
383     else high = nrow;
384     lastcol = col;
385     while (high - low > 5) {
386       t = (low + high) / 2;
387       if (rp[t] > col) high = t;
388       else low = t;
389     }
390     for (i = low; i < high; i++) {
391       if (rp[i] == col) {
392         ap[i] += value;
393         low = i + 1;
394         break;
395       }
396     }
397   }
398   PetscCall(MatSeqAIJRestoreArray(A, &aa));
399   return PETSC_SUCCESS;
400 }
401 
402 PetscErrorCode MatSetValues_SeqAIJ(Mat A, PetscInt m, const PetscInt im[], PetscInt n, const PetscInt in[], const PetscScalar v[], InsertMode is)
403 {
404   Mat_SeqAIJ *a = (Mat_SeqAIJ *)A->data;
405   PetscInt   *rp, k, low, high, t, ii, row, nrow, i, col, l, rmax, N;
406   PetscInt   *imax = a->imax, *ai = a->i, *ailen = a->ilen;
407   PetscInt   *aj = a->j, nonew = a->nonew, lastcol = -1;
408   MatScalar  *ap = NULL, value = 0.0, *aa;
409   PetscBool   ignorezeroentries = a->ignorezeroentries;
410   PetscBool   roworiented       = a->roworiented;
411 
412   PetscFunctionBegin;
413   PetscCall(MatSeqAIJGetArray(A, &aa));
414   for (k = 0; k < m; k++) { /* loop over added rows */
415     row = im[k];
416     if (row < 0) continue;
417     PetscCheck(row < A->rmap->n, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Row too large: row %" PetscInt_FMT " max %" PetscInt_FMT, row, A->rmap->n - 1);
418     rp = PetscSafePointerPlusOffset(aj, ai[row]);
419     if (!A->structure_only) ap = PetscSafePointerPlusOffset(aa, ai[row]);
420     rmax = imax[row];
421     nrow = ailen[row];
422     low  = 0;
423     high = nrow;
424     for (l = 0; l < n; l++) { /* loop over added columns */
425       if (in[l] < 0) continue;
426       PetscCheck(in[l] < A->cmap->n, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Column too large: col %" PetscInt_FMT " max %" PetscInt_FMT, in[l], A->cmap->n - 1);
427       col = in[l];
428       if (v && !A->structure_only) value = roworiented ? v[l + k * n] : v[k + l * m];
429       if (!A->structure_only && value == 0.0 && ignorezeroentries && is == ADD_VALUES && row != col) continue;
430 
431       if (col <= lastcol) low = 0;
432       else high = nrow;
433       lastcol = col;
434       while (high - low > 5) {
435         t = (low + high) / 2;
436         if (rp[t] > col) high = t;
437         else low = t;
438       }
439       for (i = low; i < high; i++) {
440         if (rp[i] > col) break;
441         if (rp[i] == col) {
442           if (!A->structure_only) {
443             if (is == ADD_VALUES) {
444               ap[i] += value;
445               (void)PetscLogFlops(1.0);
446             } else ap[i] = value;
447           }
448           low = i + 1;
449           goto noinsert;
450         }
451       }
452       if (value == 0.0 && ignorezeroentries && row != col) goto noinsert;
453       if (nonew == 1) goto noinsert;
454       PetscCheck(nonew != -1, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Inserting a new nonzero at (%" PetscInt_FMT ",%" PetscInt_FMT ") in the matrix", row, col);
455       if (A->structure_only) {
456         MatSeqXAIJReallocateAIJ_structure_only(A, A->rmap->n, 1, nrow, row, col, rmax, ai, aj, rp, imax, nonew, MatScalar);
457       } else {
458         MatSeqXAIJReallocateAIJ(A, A->rmap->n, 1, nrow, row, col, rmax, aa, ai, aj, rp, ap, imax, nonew, MatScalar);
459       }
460       N = nrow++ - 1;
461       a->nz++;
462       high++;
463       /* shift up all the later entries in this row */
464       PetscCall(PetscArraymove(rp + i + 1, rp + i, N - i + 1));
465       rp[i] = col;
466       if (!A->structure_only) {
467         PetscCall(PetscArraymove(ap + i + 1, ap + i, N - i + 1));
468         ap[i] = value;
469       }
470       low = i + 1;
471     noinsert:;
472     }
473     ailen[row] = nrow;
474   }
475   PetscCall(MatSeqAIJRestoreArray(A, &aa));
476   PetscFunctionReturn(PETSC_SUCCESS);
477 }
478 
479 static PetscErrorCode MatSetValues_SeqAIJ_SortedFullNoPreallocation(Mat A, PetscInt m, const PetscInt im[], PetscInt n, const PetscInt in[], const PetscScalar v[], InsertMode is)
480 {
481   Mat_SeqAIJ *a = (Mat_SeqAIJ *)A->data;
482   PetscInt   *rp, k, row;
483   PetscInt   *ai = a->i;
484   PetscInt   *aj = a->j;
485   MatScalar  *aa, *ap;
486 
487   PetscFunctionBegin;
488   PetscCheck(!A->was_assembled, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Cannot call on assembled matrix.");
489   PetscCheck(m * n + a->nz <= a->maxnz, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Number of entries in matrix will be larger than maximum nonzeros allocated for %" PetscInt_FMT " in MatSeqAIJSetTotalPreallocation()", a->maxnz);
490 
491   PetscCall(MatSeqAIJGetArray(A, &aa));
492   for (k = 0; k < m; k++) { /* loop over added rows */
493     row = im[k];
494     rp  = aj + ai[row];
495     ap  = PetscSafePointerPlusOffset(aa, ai[row]);
496 
497     PetscCall(PetscMemcpy(rp, in, n * sizeof(PetscInt)));
498     if (!A->structure_only) {
499       if (v) {
500         PetscCall(PetscMemcpy(ap, v, n * sizeof(PetscScalar)));
501         v += n;
502       } else {
503         PetscCall(PetscMemzero(ap, n * sizeof(PetscScalar)));
504       }
505     }
506     a->ilen[row]  = n;
507     a->imax[row]  = n;
508     a->i[row + 1] = a->i[row] + n;
509     a->nz += n;
510   }
511   PetscCall(MatSeqAIJRestoreArray(A, &aa));
512   PetscFunctionReturn(PETSC_SUCCESS);
513 }
514 
515 /*@
516   MatSeqAIJSetTotalPreallocation - Sets an upper bound on the total number of expected nonzeros in the matrix.
517 
518   Input Parameters:
519 + A       - the `MATSEQAIJ` matrix
520 - nztotal - bound on the number of nonzeros
521 
522   Level: advanced
523 
524   Notes:
525   This can be called if you will be provided the matrix row by row (from row zero) with sorted column indices for each row.
526   Simply call `MatSetValues()` after this call to provide the matrix entries in the usual manner. This matrix may be used
527   as always with multiple matrix assemblies.
528 
529 .seealso: [](ch_matrices), `Mat`, `MatSetOption()`, `MAT_SORTED_FULL`, `MatSetValues()`, `MatSeqAIJSetPreallocation()`
530 @*/
531 PetscErrorCode MatSeqAIJSetTotalPreallocation(Mat A, PetscInt nztotal)
532 {
533   Mat_SeqAIJ *a = (Mat_SeqAIJ *)A->data;
534 
535   PetscFunctionBegin;
536   PetscCall(PetscLayoutSetUp(A->rmap));
537   PetscCall(PetscLayoutSetUp(A->cmap));
538   a->maxnz = nztotal;
539   if (!a->imax) { PetscCall(PetscMalloc1(A->rmap->n, &a->imax)); }
540   if (!a->ilen) {
541     PetscCall(PetscMalloc1(A->rmap->n, &a->ilen));
542   } else {
543     PetscCall(PetscMemzero(a->ilen, A->rmap->n * sizeof(PetscInt)));
544   }
545 
546   /* allocate the matrix space */
547   if (A->structure_only) {
548     PetscCall(PetscMalloc1(nztotal, &a->j));
549     PetscCall(PetscMalloc1(A->rmap->n + 1, &a->i));
550   } else {
551     PetscCall(PetscMalloc3(nztotal, &a->a, nztotal, &a->j, A->rmap->n + 1, &a->i));
552   }
553   a->i[0] = 0;
554   if (A->structure_only) {
555     a->singlemalloc = PETSC_FALSE;
556     a->free_a       = PETSC_FALSE;
557   } else {
558     a->singlemalloc = PETSC_TRUE;
559     a->free_a       = PETSC_TRUE;
560   }
561   a->free_ij        = PETSC_TRUE;
562   A->ops->setvalues = MatSetValues_SeqAIJ_SortedFullNoPreallocation;
563   A->preallocated   = PETSC_TRUE;
564   PetscFunctionReturn(PETSC_SUCCESS);
565 }
566 
567 static PetscErrorCode MatSetValues_SeqAIJ_SortedFull(Mat A, PetscInt m, const PetscInt im[], PetscInt n, const PetscInt in[], const PetscScalar v[], InsertMode is)
568 {
569   Mat_SeqAIJ *a = (Mat_SeqAIJ *)A->data;
570   PetscInt   *rp, k, row;
571   PetscInt   *ai = a->i, *ailen = a->ilen;
572   PetscInt   *aj = a->j;
573   MatScalar  *aa, *ap;
574 
575   PetscFunctionBegin;
576   PetscCall(MatSeqAIJGetArray(A, &aa));
577   for (k = 0; k < m; k++) { /* loop over added rows */
578     row = im[k];
579     PetscCheck(n <= a->imax[row], PETSC_COMM_SELF, PETSC_ERR_ARG_WRONG, "Preallocation for row %" PetscInt_FMT " does not match number of columns provided", n);
580     rp = aj + ai[row];
581     ap = aa + ai[row];
582     if (!A->was_assembled) PetscCall(PetscMemcpy(rp, in, n * sizeof(PetscInt)));
583     if (!A->structure_only) {
584       if (v) {
585         PetscCall(PetscMemcpy(ap, v, n * sizeof(PetscScalar)));
586         v += n;
587       } else {
588         PetscCall(PetscMemzero(ap, n * sizeof(PetscScalar)));
589       }
590     }
591     ailen[row] = n;
592     a->nz += n;
593   }
594   PetscCall(MatSeqAIJRestoreArray(A, &aa));
595   PetscFunctionReturn(PETSC_SUCCESS);
596 }
597 
598 static PetscErrorCode MatGetValues_SeqAIJ(Mat A, PetscInt m, const PetscInt im[], PetscInt n, const PetscInt in[], PetscScalar v[])
599 {
600   Mat_SeqAIJ      *a = (Mat_SeqAIJ *)A->data;
601   PetscInt        *rp, k, low, high, t, row, nrow, i, col, l, *aj = a->j;
602   PetscInt        *ai = a->i, *ailen = a->ilen;
603   const MatScalar *ap, *aa;
604 
605   PetscFunctionBegin;
606   PetscCall(MatSeqAIJGetArrayRead(A, &aa));
607   for (k = 0; k < m; k++) { /* loop over rows */
608     row = im[k];
609     if (row < 0) {
610       v += n;
611       continue;
612     } /* negative row */
613     PetscCheck(row < A->rmap->n, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Row too large: row %" PetscInt_FMT " max %" PetscInt_FMT, row, A->rmap->n - 1);
614     rp   = PetscSafePointerPlusOffset(aj, ai[row]);
615     ap   = PetscSafePointerPlusOffset(aa, ai[row]);
616     nrow = ailen[row];
617     for (l = 0; l < n; l++) { /* loop over columns */
618       if (in[l] < 0) {
619         v++;
620         continue;
621       } /* negative column */
622       PetscCheck(in[l] < A->cmap->n, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Column too large: col %" PetscInt_FMT " max %" PetscInt_FMT, in[l], A->cmap->n - 1);
623       col  = in[l];
624       high = nrow;
625       low  = 0; /* assume unsorted */
626       while (high - low > 5) {
627         t = (low + high) / 2;
628         if (rp[t] > col) high = t;
629         else low = t;
630       }
631       for (i = low; i < high; i++) {
632         if (rp[i] > col) break;
633         if (rp[i] == col) {
634           *v++ = ap[i];
635           goto finished;
636         }
637       }
638       *v++ = 0.0;
639     finished:;
640     }
641   }
642   PetscCall(MatSeqAIJRestoreArrayRead(A, &aa));
643   PetscFunctionReturn(PETSC_SUCCESS);
644 }
645 
646 static PetscErrorCode MatView_SeqAIJ_Binary(Mat mat, PetscViewer viewer)
647 {
648   Mat_SeqAIJ        *A = (Mat_SeqAIJ *)mat->data;
649   const PetscScalar *av;
650   PetscInt           header[4], M, N, m, nz, i;
651   PetscInt          *rowlens;
652 
653   PetscFunctionBegin;
654   PetscCall(PetscViewerSetUp(viewer));
655 
656   M  = mat->rmap->N;
657   N  = mat->cmap->N;
658   m  = mat->rmap->n;
659   nz = A->nz;
660 
661   /* write matrix header */
662   header[0] = MAT_FILE_CLASSID;
663   header[1] = M;
664   header[2] = N;
665   header[3] = nz;
666   PetscCall(PetscViewerBinaryWrite(viewer, header, 4, PETSC_INT));
667 
668   /* fill in and store row lengths */
669   PetscCall(PetscMalloc1(m, &rowlens));
670   for (i = 0; i < m; i++) rowlens[i] = A->i[i + 1] - A->i[i];
671   PetscCall(PetscViewerBinaryWrite(viewer, rowlens, m, PETSC_INT));
672   PetscCall(PetscFree(rowlens));
673   /* store column indices */
674   PetscCall(PetscViewerBinaryWrite(viewer, A->j, nz, PETSC_INT));
675   /* store nonzero values */
676   PetscCall(MatSeqAIJGetArrayRead(mat, &av));
677   PetscCall(PetscViewerBinaryWrite(viewer, av, nz, PETSC_SCALAR));
678   PetscCall(MatSeqAIJRestoreArrayRead(mat, &av));
679 
680   /* write block size option to the viewer's .info file */
681   PetscCall(MatView_Binary_BlockSizes(mat, viewer));
682   PetscFunctionReturn(PETSC_SUCCESS);
683 }
684 
685 static PetscErrorCode MatView_SeqAIJ_ASCII_structonly(Mat A, PetscViewer viewer)
686 {
687   Mat_SeqAIJ *a = (Mat_SeqAIJ *)A->data;
688   PetscInt    i, k, m = A->rmap->N;
689 
690   PetscFunctionBegin;
691   PetscCall(PetscViewerASCIIUseTabs(viewer, PETSC_FALSE));
692   for (i = 0; i < m; i++) {
693     PetscCall(PetscViewerASCIIPrintf(viewer, "row %" PetscInt_FMT ":", i));
694     for (k = a->i[i]; k < a->i[i + 1]; k++) PetscCall(PetscViewerASCIIPrintf(viewer, " (%" PetscInt_FMT ") ", a->j[k]));
695     PetscCall(PetscViewerASCIIPrintf(viewer, "\n"));
696   }
697   PetscCall(PetscViewerASCIIUseTabs(viewer, PETSC_TRUE));
698   PetscFunctionReturn(PETSC_SUCCESS);
699 }
700 
701 extern PetscErrorCode MatSeqAIJFactorInfo_Matlab(Mat, PetscViewer);
702 
703 static PetscErrorCode MatView_SeqAIJ_ASCII(Mat A, PetscViewer viewer)
704 {
705   Mat_SeqAIJ        *a = (Mat_SeqAIJ *)A->data;
706   const PetscScalar *av;
707   PetscInt           i, j, m = A->rmap->n;
708   const char        *name;
709   PetscViewerFormat  format;
710 
711   PetscFunctionBegin;
712   if (A->structure_only) {
713     PetscCall(MatView_SeqAIJ_ASCII_structonly(A, viewer));
714     PetscFunctionReturn(PETSC_SUCCESS);
715   }
716 
717   PetscCall(PetscViewerGetFormat(viewer, &format));
718   if (format == PETSC_VIEWER_ASCII_FACTOR_INFO || format == PETSC_VIEWER_ASCII_INFO || format == PETSC_VIEWER_ASCII_INFO_DETAIL) PetscFunctionReturn(PETSC_SUCCESS);
719 
720   /* trigger copy to CPU if needed */
721   PetscCall(MatSeqAIJGetArrayRead(A, &av));
722   PetscCall(MatSeqAIJRestoreArrayRead(A, &av));
723   if (format == PETSC_VIEWER_ASCII_MATLAB) {
724     PetscInt nofinalvalue = 0;
725     if (m && ((a->i[m] == a->i[m - 1]) || (a->j[a->nz - 1] != A->cmap->n - 1))) {
726       /* Need a dummy value to ensure the dimension of the matrix. */
727       nofinalvalue = 1;
728     }
729     PetscCall(PetscViewerASCIIUseTabs(viewer, PETSC_FALSE));
730     PetscCall(PetscViewerASCIIPrintf(viewer, "%% Size = %" PetscInt_FMT " %" PetscInt_FMT " \n", m, A->cmap->n));
731     PetscCall(PetscViewerASCIIPrintf(viewer, "%% Nonzeros = %" PetscInt_FMT " \n", a->nz));
732 #if defined(PETSC_USE_COMPLEX)
733     PetscCall(PetscViewerASCIIPrintf(viewer, "zzz = zeros(%" PetscInt_FMT ",4);\n", a->nz + nofinalvalue));
734 #else
735     PetscCall(PetscViewerASCIIPrintf(viewer, "zzz = zeros(%" PetscInt_FMT ",3);\n", a->nz + nofinalvalue));
736 #endif
737     PetscCall(PetscViewerASCIIPrintf(viewer, "zzz = [\n"));
738 
739     for (i = 0; i < m; i++) {
740       for (j = a->i[i]; j < a->i[i + 1]; j++) {
741 #if defined(PETSC_USE_COMPLEX)
742         PetscCall(PetscViewerASCIIPrintf(viewer, "%" PetscInt_FMT " %" PetscInt_FMT "  %18.16e %18.16e\n", i + 1, a->j[j] + 1, (double)PetscRealPart(a->a[j]), (double)PetscImaginaryPart(a->a[j])));
743 #else
744         PetscCall(PetscViewerASCIIPrintf(viewer, "%" PetscInt_FMT " %" PetscInt_FMT "  %18.16e\n", i + 1, a->j[j] + 1, (double)a->a[j]));
745 #endif
746       }
747     }
748     if (nofinalvalue) {
749 #if defined(PETSC_USE_COMPLEX)
750       PetscCall(PetscViewerASCIIPrintf(viewer, "%" PetscInt_FMT " %" PetscInt_FMT "  %18.16e %18.16e\n", m, A->cmap->n, 0., 0.));
751 #else
752       PetscCall(PetscViewerASCIIPrintf(viewer, "%" PetscInt_FMT " %" PetscInt_FMT "  %18.16e\n", m, A->cmap->n, 0.0));
753 #endif
754     }
755     PetscCall(PetscObjectGetName((PetscObject)A, &name));
756     PetscCall(PetscViewerASCIIPrintf(viewer, "];\n %s = spconvert(zzz);\n", name));
757     PetscCall(PetscViewerASCIIUseTabs(viewer, PETSC_TRUE));
758   } else if (format == PETSC_VIEWER_ASCII_COMMON) {
759     PetscCall(PetscViewerASCIIUseTabs(viewer, PETSC_FALSE));
760     for (i = 0; i < m; i++) {
761       PetscCall(PetscViewerASCIIPrintf(viewer, "row %" PetscInt_FMT ":", i));
762       for (j = a->i[i]; j < a->i[i + 1]; j++) {
763 #if defined(PETSC_USE_COMPLEX)
764         if (PetscImaginaryPart(a->a[j]) > 0.0 && PetscRealPart(a->a[j]) != 0.0) {
765           PetscCall(PetscViewerASCIIPrintf(viewer, " (%" PetscInt_FMT ", %g + %g i)", a->j[j], (double)PetscRealPart(a->a[j]), (double)PetscImaginaryPart(a->a[j])));
766         } else if (PetscImaginaryPart(a->a[j]) < 0.0 && PetscRealPart(a->a[j]) != 0.0) {
767           PetscCall(PetscViewerASCIIPrintf(viewer, " (%" PetscInt_FMT ", %g - %g i)", a->j[j], (double)PetscRealPart(a->a[j]), (double)-PetscImaginaryPart(a->a[j])));
768         } else if (PetscRealPart(a->a[j]) != 0.0) {
769           PetscCall(PetscViewerASCIIPrintf(viewer, " (%" PetscInt_FMT ", %g) ", a->j[j], (double)PetscRealPart(a->a[j])));
770         }
771 #else
772         if (a->a[j] != 0.0) PetscCall(PetscViewerASCIIPrintf(viewer, " (%" PetscInt_FMT ", %g) ", a->j[j], (double)a->a[j]));
773 #endif
774       }
775       PetscCall(PetscViewerASCIIPrintf(viewer, "\n"));
776     }
777     PetscCall(PetscViewerASCIIUseTabs(viewer, PETSC_TRUE));
778   } else if (format == PETSC_VIEWER_ASCII_SYMMODU) {
779     PetscInt nzd = 0, fshift = 1, *sptr;
780     PetscCall(PetscViewerASCIIUseTabs(viewer, PETSC_FALSE));
781     PetscCall(PetscMalloc1(m + 1, &sptr));
782     for (i = 0; i < m; i++) {
783       sptr[i] = nzd + 1;
784       for (j = a->i[i]; j < a->i[i + 1]; j++) {
785         if (a->j[j] >= i) {
786 #if defined(PETSC_USE_COMPLEX)
787           if (PetscImaginaryPart(a->a[j]) != 0.0 || PetscRealPart(a->a[j]) != 0.0) nzd++;
788 #else
789           if (a->a[j] != 0.0) nzd++;
790 #endif
791         }
792       }
793     }
794     sptr[m] = nzd + 1;
795     PetscCall(PetscViewerASCIIPrintf(viewer, " %" PetscInt_FMT " %" PetscInt_FMT "\n\n", m, nzd));
796     for (i = 0; i < m + 1; i += 6) {
797       if (i + 4 < m) {
798         PetscCall(PetscViewerASCIIPrintf(viewer, " %" PetscInt_FMT " %" PetscInt_FMT " %" PetscInt_FMT " %" PetscInt_FMT " %" PetscInt_FMT " %" PetscInt_FMT "\n", sptr[i], sptr[i + 1], sptr[i + 2], sptr[i + 3], sptr[i + 4], sptr[i + 5]));
799       } else if (i + 3 < m) {
800         PetscCall(PetscViewerASCIIPrintf(viewer, " %" PetscInt_FMT " %" PetscInt_FMT " %" PetscInt_FMT " %" PetscInt_FMT " %" PetscInt_FMT "\n", sptr[i], sptr[i + 1], sptr[i + 2], sptr[i + 3], sptr[i + 4]));
801       } else if (i + 2 < m) {
802         PetscCall(PetscViewerASCIIPrintf(viewer, " %" PetscInt_FMT " %" PetscInt_FMT " %" PetscInt_FMT " %" PetscInt_FMT "\n", sptr[i], sptr[i + 1], sptr[i + 2], sptr[i + 3]));
803       } else if (i + 1 < m) {
804         PetscCall(PetscViewerASCIIPrintf(viewer, " %" PetscInt_FMT " %" PetscInt_FMT " %" PetscInt_FMT "\n", sptr[i], sptr[i + 1], sptr[i + 2]));
805       } else if (i < m) {
806         PetscCall(PetscViewerASCIIPrintf(viewer, " %" PetscInt_FMT " %" PetscInt_FMT "\n", sptr[i], sptr[i + 1]));
807       } else {
808         PetscCall(PetscViewerASCIIPrintf(viewer, " %" PetscInt_FMT "\n", sptr[i]));
809       }
810     }
811     PetscCall(PetscViewerASCIIPrintf(viewer, "\n"));
812     PetscCall(PetscFree(sptr));
813     for (i = 0; i < m; i++) {
814       for (j = a->i[i]; j < a->i[i + 1]; j++) {
815         if (a->j[j] >= i) PetscCall(PetscViewerASCIIPrintf(viewer, " %" PetscInt_FMT " ", a->j[j] + fshift));
816       }
817       PetscCall(PetscViewerASCIIPrintf(viewer, "\n"));
818     }
819     PetscCall(PetscViewerASCIIPrintf(viewer, "\n"));
820     for (i = 0; i < m; i++) {
821       for (j = a->i[i]; j < a->i[i + 1]; j++) {
822         if (a->j[j] >= i) {
823 #if defined(PETSC_USE_COMPLEX)
824           if (PetscImaginaryPart(a->a[j]) != 0.0 || PetscRealPart(a->a[j]) != 0.0) PetscCall(PetscViewerASCIIPrintf(viewer, " %18.16e %18.16e ", (double)PetscRealPart(a->a[j]), (double)PetscImaginaryPart(a->a[j])));
825 #else
826           if (a->a[j] != 0.0) PetscCall(PetscViewerASCIIPrintf(viewer, " %18.16e ", (double)a->a[j]));
827 #endif
828         }
829       }
830       PetscCall(PetscViewerASCIIPrintf(viewer, "\n"));
831     }
832     PetscCall(PetscViewerASCIIUseTabs(viewer, PETSC_TRUE));
833   } else if (format == PETSC_VIEWER_ASCII_DENSE) {
834     PetscInt    cnt = 0, jcnt;
835     PetscScalar value;
836 #if defined(PETSC_USE_COMPLEX)
837     PetscBool realonly = PETSC_TRUE;
838 
839     for (i = 0; i < a->i[m]; i++) {
840       if (PetscImaginaryPart(a->a[i]) != 0.0) {
841         realonly = PETSC_FALSE;
842         break;
843       }
844     }
845 #endif
846 
847     PetscCall(PetscViewerASCIIUseTabs(viewer, PETSC_FALSE));
848     for (i = 0; i < m; i++) {
849       jcnt = 0;
850       for (j = 0; j < A->cmap->n; j++) {
851         if (jcnt < a->i[i + 1] - a->i[i] && j == a->j[cnt]) {
852           value = a->a[cnt++];
853           jcnt++;
854         } else {
855           value = 0.0;
856         }
857 #if defined(PETSC_USE_COMPLEX)
858         if (realonly) {
859           PetscCall(PetscViewerASCIIPrintf(viewer, " %7.5e ", (double)PetscRealPart(value)));
860         } else {
861           PetscCall(PetscViewerASCIIPrintf(viewer, " %7.5e+%7.5e i ", (double)PetscRealPart(value), (double)PetscImaginaryPart(value)));
862         }
863 #else
864         PetscCall(PetscViewerASCIIPrintf(viewer, " %7.5e ", (double)value));
865 #endif
866       }
867       PetscCall(PetscViewerASCIIPrintf(viewer, "\n"));
868     }
869     PetscCall(PetscViewerASCIIUseTabs(viewer, PETSC_TRUE));
870   } else if (format == PETSC_VIEWER_ASCII_MATRIXMARKET) {
871     PetscInt fshift = 1;
872     PetscCall(PetscViewerASCIIUseTabs(viewer, PETSC_FALSE));
873 #if defined(PETSC_USE_COMPLEX)
874     PetscCall(PetscViewerASCIIPrintf(viewer, "%%%%MatrixMarket matrix coordinate complex general\n"));
875 #else
876     PetscCall(PetscViewerASCIIPrintf(viewer, "%%%%MatrixMarket matrix coordinate real general\n"));
877 #endif
878     PetscCall(PetscViewerASCIIPrintf(viewer, "%" PetscInt_FMT " %" PetscInt_FMT " %" PetscInt_FMT "\n", m, A->cmap->n, a->nz));
879     for (i = 0; i < m; i++) {
880       for (j = a->i[i]; j < a->i[i + 1]; j++) {
881 #if defined(PETSC_USE_COMPLEX)
882         PetscCall(PetscViewerASCIIPrintf(viewer, "%" PetscInt_FMT " %" PetscInt_FMT " %g %g\n", i + fshift, a->j[j] + fshift, (double)PetscRealPart(a->a[j]), (double)PetscImaginaryPart(a->a[j])));
883 #else
884         PetscCall(PetscViewerASCIIPrintf(viewer, "%" PetscInt_FMT " %" PetscInt_FMT " %g\n", i + fshift, a->j[j] + fshift, (double)a->a[j]));
885 #endif
886       }
887     }
888     PetscCall(PetscViewerASCIIUseTabs(viewer, PETSC_TRUE));
889   } else {
890     PetscCall(PetscViewerASCIIUseTabs(viewer, PETSC_FALSE));
891     if (A->factortype) {
892       for (i = 0; i < m; i++) {
893         PetscCall(PetscViewerASCIIPrintf(viewer, "row %" PetscInt_FMT ":", i));
894         /* L part */
895         for (j = a->i[i]; j < a->i[i + 1]; j++) {
896 #if defined(PETSC_USE_COMPLEX)
897           if (PetscImaginaryPart(a->a[j]) > 0.0) {
898             PetscCall(PetscViewerASCIIPrintf(viewer, " (%" PetscInt_FMT ", %g + %g i)", a->j[j], (double)PetscRealPart(a->a[j]), (double)PetscImaginaryPart(a->a[j])));
899           } else if (PetscImaginaryPart(a->a[j]) < 0.0) {
900             PetscCall(PetscViewerASCIIPrintf(viewer, " (%" PetscInt_FMT ", %g - %g i)", a->j[j], (double)PetscRealPart(a->a[j]), (double)(-PetscImaginaryPart(a->a[j]))));
901           } else {
902             PetscCall(PetscViewerASCIIPrintf(viewer, " (%" PetscInt_FMT ", %g) ", a->j[j], (double)PetscRealPart(a->a[j])));
903           }
904 #else
905           PetscCall(PetscViewerASCIIPrintf(viewer, " (%" PetscInt_FMT ", %g) ", a->j[j], (double)a->a[j]));
906 #endif
907         }
908         /* diagonal */
909         j = a->diag[i];
910 #if defined(PETSC_USE_COMPLEX)
911         if (PetscImaginaryPart(a->a[j]) > 0.0) {
912           PetscCall(PetscViewerASCIIPrintf(viewer, " (%" PetscInt_FMT ", %g + %g i)", a->j[j], (double)PetscRealPart(1.0 / a->a[j]), (double)PetscImaginaryPart(1.0 / a->a[j])));
913         } else if (PetscImaginaryPart(a->a[j]) < 0.0) {
914           PetscCall(PetscViewerASCIIPrintf(viewer, " (%" PetscInt_FMT ", %g - %g i)", a->j[j], (double)PetscRealPart(1.0 / a->a[j]), (double)(-PetscImaginaryPart(1.0 / a->a[j]))));
915         } else {
916           PetscCall(PetscViewerASCIIPrintf(viewer, " (%" PetscInt_FMT ", %g) ", a->j[j], (double)PetscRealPart(1.0 / a->a[j])));
917         }
918 #else
919         PetscCall(PetscViewerASCIIPrintf(viewer, " (%" PetscInt_FMT ", %g) ", a->j[j], (double)(1.0 / a->a[j])));
920 #endif
921 
922         /* U part */
923         for (j = a->diag[i + 1] + 1; j < a->diag[i]; j++) {
924 #if defined(PETSC_USE_COMPLEX)
925           if (PetscImaginaryPart(a->a[j]) > 0.0) {
926             PetscCall(PetscViewerASCIIPrintf(viewer, " (%" PetscInt_FMT ", %g + %g i)", a->j[j], (double)PetscRealPart(a->a[j]), (double)PetscImaginaryPart(a->a[j])));
927           } else if (PetscImaginaryPart(a->a[j]) < 0.0) {
928             PetscCall(PetscViewerASCIIPrintf(viewer, " (%" PetscInt_FMT ", %g - %g i)", a->j[j], (double)PetscRealPart(a->a[j]), (double)(-PetscImaginaryPart(a->a[j]))));
929           } else {
930             PetscCall(PetscViewerASCIIPrintf(viewer, " (%" PetscInt_FMT ", %g) ", a->j[j], (double)PetscRealPart(a->a[j])));
931           }
932 #else
933           PetscCall(PetscViewerASCIIPrintf(viewer, " (%" PetscInt_FMT ", %g) ", a->j[j], (double)a->a[j]));
934 #endif
935         }
936         PetscCall(PetscViewerASCIIPrintf(viewer, "\n"));
937       }
938     } else {
939       for (i = 0; i < m; i++) {
940         PetscCall(PetscViewerASCIIPrintf(viewer, "row %" PetscInt_FMT ":", i));
941         for (j = a->i[i]; j < a->i[i + 1]; j++) {
942 #if defined(PETSC_USE_COMPLEX)
943           if (PetscImaginaryPart(a->a[j]) > 0.0) {
944             PetscCall(PetscViewerASCIIPrintf(viewer, " (%" PetscInt_FMT ", %g + %g i)", a->j[j], (double)PetscRealPart(a->a[j]), (double)PetscImaginaryPart(a->a[j])));
945           } else if (PetscImaginaryPart(a->a[j]) < 0.0) {
946             PetscCall(PetscViewerASCIIPrintf(viewer, " (%" PetscInt_FMT ", %g - %g i)", a->j[j], (double)PetscRealPart(a->a[j]), (double)-PetscImaginaryPart(a->a[j])));
947           } else {
948             PetscCall(PetscViewerASCIIPrintf(viewer, " (%" PetscInt_FMT ", %g) ", a->j[j], (double)PetscRealPart(a->a[j])));
949           }
950 #else
951           PetscCall(PetscViewerASCIIPrintf(viewer, " (%" PetscInt_FMT ", %g) ", a->j[j], (double)a->a[j]));
952 #endif
953         }
954         PetscCall(PetscViewerASCIIPrintf(viewer, "\n"));
955       }
956     }
957     PetscCall(PetscViewerASCIIUseTabs(viewer, PETSC_TRUE));
958   }
959   PetscCall(PetscViewerFlush(viewer));
960   PetscFunctionReturn(PETSC_SUCCESS);
961 }
962 
963 #include <petscdraw.h>
964 static PetscErrorCode MatView_SeqAIJ_Draw_Zoom(PetscDraw draw, void *Aa)
965 {
966   Mat                A = (Mat)Aa;
967   Mat_SeqAIJ        *a = (Mat_SeqAIJ *)A->data;
968   PetscInt           i, j, m = A->rmap->n;
969   int                color;
970   PetscReal          xl, yl, xr, yr, x_l, x_r, y_l, y_r;
971   PetscViewer        viewer;
972   PetscViewerFormat  format;
973   const PetscScalar *aa;
974 
975   PetscFunctionBegin;
976   PetscCall(PetscObjectQuery((PetscObject)A, "Zoomviewer", (PetscObject *)&viewer));
977   PetscCall(PetscViewerGetFormat(viewer, &format));
978   PetscCall(PetscDrawGetCoordinates(draw, &xl, &yl, &xr, &yr));
979 
980   /* loop over matrix elements drawing boxes */
981   PetscCall(MatSeqAIJGetArrayRead(A, &aa));
982   if (format != PETSC_VIEWER_DRAW_CONTOUR) {
983     PetscDrawCollectiveBegin(draw);
984     /* Blue for negative, Cyan for zero and  Red for positive */
985     color = PETSC_DRAW_BLUE;
986     for (i = 0; i < m; i++) {
987       y_l = m - i - 1.0;
988       y_r = y_l + 1.0;
989       for (j = a->i[i]; j < a->i[i + 1]; j++) {
990         x_l = a->j[j];
991         x_r = x_l + 1.0;
992         if (PetscRealPart(aa[j]) >= 0.) continue;
993         PetscCall(PetscDrawRectangle(draw, x_l, y_l, x_r, y_r, color, color, color, color));
994       }
995     }
996     color = PETSC_DRAW_CYAN;
997     for (i = 0; i < m; i++) {
998       y_l = m - i - 1.0;
999       y_r = y_l + 1.0;
1000       for (j = a->i[i]; j < a->i[i + 1]; j++) {
1001         x_l = a->j[j];
1002         x_r = x_l + 1.0;
1003         if (aa[j] != 0.) continue;
1004         PetscCall(PetscDrawRectangle(draw, x_l, y_l, x_r, y_r, color, color, color, color));
1005       }
1006     }
1007     color = PETSC_DRAW_RED;
1008     for (i = 0; i < m; i++) {
1009       y_l = m - i - 1.0;
1010       y_r = y_l + 1.0;
1011       for (j = a->i[i]; j < a->i[i + 1]; j++) {
1012         x_l = a->j[j];
1013         x_r = x_l + 1.0;
1014         if (PetscRealPart(aa[j]) <= 0.) continue;
1015         PetscCall(PetscDrawRectangle(draw, x_l, y_l, x_r, y_r, color, color, color, color));
1016       }
1017     }
1018     PetscDrawCollectiveEnd(draw);
1019   } else {
1020     /* use contour shading to indicate magnitude of values */
1021     /* first determine max of all nonzero values */
1022     PetscReal minv = 0.0, maxv = 0.0;
1023     PetscInt  nz = a->nz, count = 0;
1024     PetscDraw popup;
1025 
1026     for (i = 0; i < nz; i++) {
1027       if (PetscAbsScalar(aa[i]) > maxv) maxv = PetscAbsScalar(aa[i]);
1028     }
1029     if (minv >= maxv) maxv = minv + PETSC_SMALL;
1030     PetscCall(PetscDrawGetPopup(draw, &popup));
1031     PetscCall(PetscDrawScalePopup(popup, minv, maxv));
1032 
1033     PetscDrawCollectiveBegin(draw);
1034     for (i = 0; i < m; i++) {
1035       y_l = m - i - 1.0;
1036       y_r = y_l + 1.0;
1037       for (j = a->i[i]; j < a->i[i + 1]; j++) {
1038         x_l   = a->j[j];
1039         x_r   = x_l + 1.0;
1040         color = PetscDrawRealToColor(PetscAbsScalar(aa[count]), minv, maxv);
1041         PetscCall(PetscDrawRectangle(draw, x_l, y_l, x_r, y_r, color, color, color, color));
1042         count++;
1043       }
1044     }
1045     PetscDrawCollectiveEnd(draw);
1046   }
1047   PetscCall(MatSeqAIJRestoreArrayRead(A, &aa));
1048   PetscFunctionReturn(PETSC_SUCCESS);
1049 }
1050 
1051 #include <petscdraw.h>
1052 static PetscErrorCode MatView_SeqAIJ_Draw(Mat A, PetscViewer viewer)
1053 {
1054   PetscDraw draw;
1055   PetscReal xr, yr, xl, yl, h, w;
1056   PetscBool isnull;
1057 
1058   PetscFunctionBegin;
1059   PetscCall(PetscViewerDrawGetDraw(viewer, 0, &draw));
1060   PetscCall(PetscDrawIsNull(draw, &isnull));
1061   if (isnull) PetscFunctionReturn(PETSC_SUCCESS);
1062 
1063   xr = A->cmap->n;
1064   yr = A->rmap->n;
1065   h  = yr / 10.0;
1066   w  = xr / 10.0;
1067   xr += w;
1068   yr += h;
1069   xl = -w;
1070   yl = -h;
1071   PetscCall(PetscDrawSetCoordinates(draw, xl, yl, xr, yr));
1072   PetscCall(PetscObjectCompose((PetscObject)A, "Zoomviewer", (PetscObject)viewer));
1073   PetscCall(PetscDrawZoom(draw, MatView_SeqAIJ_Draw_Zoom, A));
1074   PetscCall(PetscObjectCompose((PetscObject)A, "Zoomviewer", NULL));
1075   PetscCall(PetscDrawSave(draw));
1076   PetscFunctionReturn(PETSC_SUCCESS);
1077 }
1078 
1079 PetscErrorCode MatView_SeqAIJ(Mat A, PetscViewer viewer)
1080 {
1081   PetscBool iascii, isbinary, isdraw;
1082 
1083   PetscFunctionBegin;
1084   PetscCall(PetscObjectTypeCompare((PetscObject)viewer, PETSCVIEWERASCII, &iascii));
1085   PetscCall(PetscObjectTypeCompare((PetscObject)viewer, PETSCVIEWERBINARY, &isbinary));
1086   PetscCall(PetscObjectTypeCompare((PetscObject)viewer, PETSCVIEWERDRAW, &isdraw));
1087   if (iascii) PetscCall(MatView_SeqAIJ_ASCII(A, viewer));
1088   else if (isbinary) PetscCall(MatView_SeqAIJ_Binary(A, viewer));
1089   else if (isdraw) PetscCall(MatView_SeqAIJ_Draw(A, viewer));
1090   PetscCall(MatView_SeqAIJ_Inode(A, viewer));
1091   PetscFunctionReturn(PETSC_SUCCESS);
1092 }
1093 
1094 PetscErrorCode MatAssemblyEnd_SeqAIJ(Mat A, MatAssemblyType mode)
1095 {
1096   Mat_SeqAIJ *a      = (Mat_SeqAIJ *)A->data;
1097   PetscInt    fshift = 0, i, *ai = a->i, *aj = a->j, *imax = a->imax;
1098   PetscInt    m = A->rmap->n, *ip, N, *ailen = a->ilen, rmax = 0, n;
1099   MatScalar  *aa    = a->a, *ap;
1100   PetscReal   ratio = 0.6;
1101 
1102   PetscFunctionBegin;
1103   if (mode == MAT_FLUSH_ASSEMBLY) PetscFunctionReturn(PETSC_SUCCESS);
1104   PetscCall(MatSeqAIJInvalidateDiagonal(A));
1105   if (A->was_assembled && A->ass_nonzerostate == A->nonzerostate) {
1106     /* we need to respect users asking to use or not the inodes routine in between matrix assemblies */
1107     PetscCall(MatAssemblyEnd_SeqAIJ_Inode(A, mode));
1108     PetscFunctionReturn(PETSC_SUCCESS);
1109   }
1110 
1111   if (m) rmax = ailen[0]; /* determine row with most nonzeros */
1112   for (i = 1; i < m; i++) {
1113     /* move each row back by the amount of empty slots (fshift) before it*/
1114     fshift += imax[i - 1] - ailen[i - 1];
1115     rmax = PetscMax(rmax, ailen[i]);
1116     if (fshift) {
1117       ip = aj + ai[i];
1118       ap = aa + ai[i];
1119       N  = ailen[i];
1120       PetscCall(PetscArraymove(ip - fshift, ip, N));
1121       if (!A->structure_only) PetscCall(PetscArraymove(ap - fshift, ap, N));
1122     }
1123     ai[i] = ai[i - 1] + ailen[i - 1];
1124   }
1125   if (m) {
1126     fshift += imax[m - 1] - ailen[m - 1];
1127     ai[m] = ai[m - 1] + ailen[m - 1];
1128   }
1129   /* reset ilen and imax for each row */
1130   a->nonzerorowcnt = 0;
1131   if (A->structure_only) {
1132     PetscCall(PetscFree(a->imax));
1133     PetscCall(PetscFree(a->ilen));
1134   } else { /* !A->structure_only */
1135     for (i = 0; i < m; i++) {
1136       ailen[i] = imax[i] = ai[i + 1] - ai[i];
1137       a->nonzerorowcnt += ((ai[i + 1] - ai[i]) > 0);
1138     }
1139   }
1140   a->nz = ai[m];
1141   PetscCheck(!fshift || a->nounused != -1, PETSC_COMM_SELF, PETSC_ERR_PLIB, "Unused space detected in matrix: %" PetscInt_FMT " X %" PetscInt_FMT ", %" PetscInt_FMT " unneeded", m, A->cmap->n, fshift);
1142   PetscCall(MatMarkDiagonal_SeqAIJ(A)); // since diagonal info is used a lot, it is helpful to set them up at the end of assembly
1143   a->diagonaldense = PETSC_TRUE;
1144   n                = PetscMin(A->rmap->n, A->cmap->n);
1145   for (i = 0; i < n; i++) {
1146     if (a->diag[i] >= ai[i + 1]) {
1147       a->diagonaldense = PETSC_FALSE;
1148       break;
1149     }
1150   }
1151   PetscCall(PetscInfo(A, "Matrix size: %" PetscInt_FMT " X %" PetscInt_FMT "; storage space: %" PetscInt_FMT " unneeded,%" PetscInt_FMT " used\n", m, A->cmap->n, fshift, a->nz));
1152   PetscCall(PetscInfo(A, "Number of mallocs during MatSetValues() is %" PetscInt_FMT "\n", a->reallocs));
1153   PetscCall(PetscInfo(A, "Maximum nonzeros in any row is %" PetscInt_FMT "\n", rmax));
1154 
1155   A->info.mallocs += a->reallocs;
1156   a->reallocs         = 0;
1157   A->info.nz_unneeded = (PetscReal)fshift;
1158   a->rmax             = rmax;
1159 
1160   if (!A->structure_only) PetscCall(MatCheckCompressedRow(A, a->nonzerorowcnt, &a->compressedrow, a->i, m, ratio));
1161   PetscCall(MatAssemblyEnd_SeqAIJ_Inode(A, mode));
1162   PetscFunctionReturn(PETSC_SUCCESS);
1163 }
1164 
1165 static PetscErrorCode MatRealPart_SeqAIJ(Mat A)
1166 {
1167   Mat_SeqAIJ *a = (Mat_SeqAIJ *)A->data;
1168   PetscInt    i, nz = a->nz;
1169   MatScalar  *aa;
1170 
1171   PetscFunctionBegin;
1172   PetscCall(MatSeqAIJGetArray(A, &aa));
1173   for (i = 0; i < nz; i++) aa[i] = PetscRealPart(aa[i]);
1174   PetscCall(MatSeqAIJRestoreArray(A, &aa));
1175   PetscCall(MatSeqAIJInvalidateDiagonal(A));
1176   PetscFunctionReturn(PETSC_SUCCESS);
1177 }
1178 
1179 static PetscErrorCode MatImaginaryPart_SeqAIJ(Mat A)
1180 {
1181   Mat_SeqAIJ *a = (Mat_SeqAIJ *)A->data;
1182   PetscInt    i, nz = a->nz;
1183   MatScalar  *aa;
1184 
1185   PetscFunctionBegin;
1186   PetscCall(MatSeqAIJGetArray(A, &aa));
1187   for (i = 0; i < nz; i++) aa[i] = PetscImaginaryPart(aa[i]);
1188   PetscCall(MatSeqAIJRestoreArray(A, &aa));
1189   PetscCall(MatSeqAIJInvalidateDiagonal(A));
1190   PetscFunctionReturn(PETSC_SUCCESS);
1191 }
1192 
1193 PetscErrorCode MatZeroEntries_SeqAIJ(Mat A)
1194 {
1195   Mat_SeqAIJ *a = (Mat_SeqAIJ *)A->data;
1196   MatScalar  *aa;
1197 
1198   PetscFunctionBegin;
1199   PetscCall(MatSeqAIJGetArrayWrite(A, &aa));
1200   PetscCall(PetscArrayzero(aa, a->i[A->rmap->n]));
1201   PetscCall(MatSeqAIJRestoreArrayWrite(A, &aa));
1202   PetscCall(MatSeqAIJInvalidateDiagonal(A));
1203   PetscFunctionReturn(PETSC_SUCCESS);
1204 }
1205 
1206 PetscErrorCode MatDestroy_SeqAIJ(Mat A)
1207 {
1208   Mat_SeqAIJ *a = (Mat_SeqAIJ *)A->data;
1209 
1210   PetscFunctionBegin;
1211   if (A->hash_active) {
1212     A->ops[0] = a->cops;
1213     PetscCall(PetscHMapIJVDestroy(&a->ht));
1214     PetscCall(PetscFree(a->dnz));
1215     A->hash_active = PETSC_FALSE;
1216   }
1217 
1218   PetscCall(PetscLogObjectState((PetscObject)A, "Rows=%" PetscInt_FMT ", Cols=%" PetscInt_FMT ", NZ=%" PetscInt_FMT, A->rmap->n, A->cmap->n, a->nz));
1219   PetscCall(MatSeqXAIJFreeAIJ(A, &a->a, &a->j, &a->i));
1220   PetscCall(ISDestroy(&a->row));
1221   PetscCall(ISDestroy(&a->col));
1222   PetscCall(PetscFree(a->diag));
1223   PetscCall(PetscFree(a->ibdiag));
1224   PetscCall(PetscFree(a->imax));
1225   PetscCall(PetscFree(a->ilen));
1226   PetscCall(PetscFree(a->ipre));
1227   PetscCall(PetscFree3(a->idiag, a->mdiag, a->ssor_work));
1228   PetscCall(PetscFree(a->solve_work));
1229   PetscCall(ISDestroy(&a->icol));
1230   PetscCall(PetscFree(a->saved_values));
1231   PetscCall(PetscFree2(a->compressedrow.i, a->compressedrow.rindex));
1232   PetscCall(MatDestroy_SeqAIJ_Inode(A));
1233   PetscCall(PetscFree(A->data));
1234 
1235   /* MatMatMultNumeric_SeqAIJ_SeqAIJ_Sorted may allocate this.
1236      That function is so heavily used (sometimes in an hidden way through multnumeric function pointers)
1237      that is hard to properly add this data to the MatProduct data. We free it here to avoid
1238      users reusing the matrix object with different data to incur in obscure segmentation faults
1239      due to different matrix sizes */
1240   PetscCall(PetscObjectCompose((PetscObject)A, "__PETSc__ab_dense", NULL));
1241 
1242   PetscCall(PetscObjectChangeTypeName((PetscObject)A, NULL));
1243   PetscCall(PetscObjectComposeFunction((PetscObject)A, "PetscMatlabEnginePut_C", NULL));
1244   PetscCall(PetscObjectComposeFunction((PetscObject)A, "PetscMatlabEngineGet_C", NULL));
1245   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatSeqAIJSetColumnIndices_C", NULL));
1246   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatStoreValues_C", NULL));
1247   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatRetrieveValues_C", NULL));
1248   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatConvert_seqaij_seqsbaij_C", NULL));
1249   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatConvert_seqaij_seqbaij_C", NULL));
1250   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatConvert_seqaij_seqaijperm_C", NULL));
1251   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatConvert_seqaij_seqaijsell_C", NULL));
1252 #if defined(PETSC_HAVE_MKL_SPARSE)
1253   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatConvert_seqaij_seqaijmkl_C", NULL));
1254 #endif
1255 #if defined(PETSC_HAVE_CUDA)
1256   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatConvert_seqaij_seqaijcusparse_C", NULL));
1257   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatProductSetFromOptions_seqaijcusparse_seqaij_C", NULL));
1258   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatProductSetFromOptions_seqaij_seqaijcusparse_C", NULL));
1259 #endif
1260 #if defined(PETSC_HAVE_HIP)
1261   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatConvert_seqaij_seqaijhipsparse_C", NULL));
1262   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatProductSetFromOptions_seqaijhipsparse_seqaij_C", NULL));
1263   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatProductSetFromOptions_seqaij_seqaijhipsparse_C", NULL));
1264 #endif
1265 #if defined(PETSC_HAVE_KOKKOS_KERNELS)
1266   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatConvert_seqaij_seqaijkokkos_C", NULL));
1267 #endif
1268   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatConvert_seqaij_seqaijcrl_C", NULL));
1269 #if defined(PETSC_HAVE_ELEMENTAL)
1270   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatConvert_seqaij_elemental_C", NULL));
1271 #endif
1272 #if defined(PETSC_HAVE_SCALAPACK)
1273   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatConvert_seqaij_scalapack_C", NULL));
1274 #endif
1275 #if defined(PETSC_HAVE_HYPRE)
1276   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatConvert_seqaij_hypre_C", NULL));
1277   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatProductSetFromOptions_transpose_seqaij_seqaij_C", NULL));
1278 #endif
1279   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatConvert_seqaij_seqdense_C", NULL));
1280   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatConvert_seqaij_seqsell_C", NULL));
1281   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatConvert_seqaij_is_C", NULL));
1282   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatIsTranspose_C", NULL));
1283   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatIsHermitianTranspose_C", NULL));
1284   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatSeqAIJSetPreallocation_C", NULL));
1285   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatResetPreallocation_C", NULL));
1286   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatSeqAIJSetPreallocationCSR_C", NULL));
1287   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatReorderForNonzeroDiagonal_C", NULL));
1288   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatProductSetFromOptions_is_seqaij_C", NULL));
1289   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatProductSetFromOptions_seqdense_seqaij_C", NULL));
1290   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatProductSetFromOptions_seqaij_seqaij_C", NULL));
1291   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatSeqAIJKron_C", NULL));
1292   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatSetPreallocationCOO_C", NULL));
1293   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatSetValuesCOO_C", NULL));
1294   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatFactorGetSolverType_C", NULL));
1295   /* these calls do not belong here: the subclasses Duplicate/Destroy are wrong */
1296   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatConvert_seqaijsell_seqaij_C", NULL));
1297   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatConvert_seqaijperm_seqaij_C", NULL));
1298   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatConvert_seqaij_seqaijviennacl_C", NULL));
1299   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatProductSetFromOptions_seqaijviennacl_seqdense_C", NULL));
1300   PetscCall(PetscObjectComposeFunction((PetscObject)A, "MatProductSetFromOptions_seqaijviennacl_seqaij_C", NULL));
1301   PetscFunctionReturn(PETSC_SUCCESS);
1302 }
1303 
1304 PetscErrorCode MatSetOption_SeqAIJ(Mat A, MatOption op, PetscBool flg)
1305 {
1306   Mat_SeqAIJ *a = (Mat_SeqAIJ *)A->data;
1307 
1308   PetscFunctionBegin;
1309   switch (op) {
1310   case MAT_ROW_ORIENTED:
1311     a->roworiented = flg;
1312     break;
1313   case MAT_KEEP_NONZERO_PATTERN:
1314     a->keepnonzeropattern = flg;
1315     break;
1316   case MAT_NEW_NONZERO_LOCATIONS:
1317     a->nonew = (flg ? 0 : 1);
1318     break;
1319   case MAT_NEW_NONZERO_LOCATION_ERR:
1320     a->nonew = (flg ? -1 : 0);
1321     break;
1322   case MAT_NEW_NONZERO_ALLOCATION_ERR:
1323     a->nonew = (flg ? -2 : 0);
1324     break;
1325   case MAT_UNUSED_NONZERO_LOCATION_ERR:
1326     a->nounused = (flg ? -1 : 0);
1327     break;
1328   case MAT_IGNORE_ZERO_ENTRIES:
1329     a->ignorezeroentries = flg;
1330     break;
1331   case MAT_SPD:
1332   case MAT_SYMMETRIC:
1333   case MAT_STRUCTURALLY_SYMMETRIC:
1334   case MAT_HERMITIAN:
1335   case MAT_SYMMETRY_ETERNAL:
1336   case MAT_STRUCTURE_ONLY:
1337   case MAT_STRUCTURAL_SYMMETRY_ETERNAL:
1338   case MAT_SPD_ETERNAL:
1339     /* if the diagonal matrix is square it inherits some of the properties above */
1340     break;
1341   case MAT_FORCE_DIAGONAL_ENTRIES:
1342   case MAT_IGNORE_OFF_PROC_ENTRIES:
1343   case MAT_USE_HASH_TABLE:
1344     PetscCall(PetscInfo(A, "Option %s ignored\n", MatOptions[op]));
1345     break;
1346   case MAT_USE_INODES:
1347     PetscCall(MatSetOption_SeqAIJ_Inode(A, MAT_USE_INODES, flg));
1348     break;
1349   case MAT_SUBMAT_SINGLEIS:
1350     A->submat_singleis = flg;
1351     break;
1352   case MAT_SORTED_FULL:
1353     if (flg) A->ops->setvalues = MatSetValues_SeqAIJ_SortedFull;
1354     else A->ops->setvalues = MatSetValues_SeqAIJ;
1355     break;
1356   case MAT_FORM_EXPLICIT_TRANSPOSE:
1357     A->form_explicit_transpose = flg;
1358     break;
1359   default:
1360     SETERRQ(PETSC_COMM_SELF, PETSC_ERR_SUP, "unknown option %d", op);
1361   }
1362   PetscFunctionReturn(PETSC_SUCCESS);
1363 }
1364 
1365 static PetscErrorCode MatGetDiagonal_SeqAIJ(Mat A, Vec v)
1366 {
1367   Mat_SeqAIJ        *a = (Mat_SeqAIJ *)A->data;
1368   PetscInt           i, j, n, *ai = a->i, *aj = a->j;
1369   PetscScalar       *x;
1370   const PetscScalar *aa;
1371 
1372   PetscFunctionBegin;
1373   PetscCall(VecGetLocalSize(v, &n));
1374   PetscCheck(n == A->rmap->n, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Nonconforming matrix and vector");
1375   PetscCall(MatSeqAIJGetArrayRead(A, &aa));
1376   if (A->factortype == MAT_FACTOR_ILU || A->factortype == MAT_FACTOR_LU) {
1377     PetscInt *diag = a->diag;
1378     PetscCall(VecGetArrayWrite(v, &x));
1379     for (i = 0; i < n; i++) x[i] = 1.0 / aa[diag[i]];
1380     PetscCall(VecRestoreArrayWrite(v, &x));
1381     PetscCall(MatSeqAIJRestoreArrayRead(A, &aa));
1382     PetscFunctionReturn(PETSC_SUCCESS);
1383   }
1384 
1385   PetscCall(VecGetArrayWrite(v, &x));
1386   for (i = 0; i < n; i++) {
1387     x[i] = 0.0;
1388     for (j = ai[i]; j < ai[i + 1]; j++) {
1389       if (aj[j] == i) {
1390         x[i] = aa[j];
1391         break;
1392       }
1393     }
1394   }
1395   PetscCall(VecRestoreArrayWrite(v, &x));
1396   PetscCall(MatSeqAIJRestoreArrayRead(A, &aa));
1397   PetscFunctionReturn(PETSC_SUCCESS);
1398 }
1399 
1400 #include <../src/mat/impls/aij/seq/ftn-kernels/fmult.h>
1401 PetscErrorCode MatMultTransposeAdd_SeqAIJ(Mat A, Vec xx, Vec zz, Vec yy)
1402 {
1403   Mat_SeqAIJ        *a = (Mat_SeqAIJ *)A->data;
1404   const MatScalar   *aa;
1405   PetscScalar       *y;
1406   const PetscScalar *x;
1407   PetscInt           m = A->rmap->n;
1408 #if !defined(PETSC_USE_FORTRAN_KERNEL_MULTTRANSPOSEAIJ)
1409   const MatScalar  *v;
1410   PetscScalar       alpha;
1411   PetscInt          n, i, j;
1412   const PetscInt   *idx, *ii, *ridx = NULL;
1413   Mat_CompressedRow cprow    = a->compressedrow;
1414   PetscBool         usecprow = cprow.use;
1415 #endif
1416 
1417   PetscFunctionBegin;
1418   if (zz != yy) PetscCall(VecCopy(zz, yy));
1419   PetscCall(VecGetArrayRead(xx, &x));
1420   PetscCall(VecGetArray(yy, &y));
1421   PetscCall(MatSeqAIJGetArrayRead(A, &aa));
1422 
1423 #if defined(PETSC_USE_FORTRAN_KERNEL_MULTTRANSPOSEAIJ)
1424   fortranmulttransposeaddaij_(&m, x, a->i, a->j, aa, y);
1425 #else
1426   if (usecprow) {
1427     m    = cprow.nrows;
1428     ii   = cprow.i;
1429     ridx = cprow.rindex;
1430   } else {
1431     ii = a->i;
1432   }
1433   for (i = 0; i < m; i++) {
1434     idx = a->j + ii[i];
1435     v   = aa + ii[i];
1436     n   = ii[i + 1] - ii[i];
1437     if (usecprow) {
1438       alpha = x[ridx[i]];
1439     } else {
1440       alpha = x[i];
1441     }
1442     for (j = 0; j < n; j++) y[idx[j]] += alpha * v[j];
1443   }
1444 #endif
1445   PetscCall(PetscLogFlops(2.0 * a->nz));
1446   PetscCall(VecRestoreArrayRead(xx, &x));
1447   PetscCall(VecRestoreArray(yy, &y));
1448   PetscCall(MatSeqAIJRestoreArrayRead(A, &aa));
1449   PetscFunctionReturn(PETSC_SUCCESS);
1450 }
1451 
1452 PetscErrorCode MatMultTranspose_SeqAIJ(Mat A, Vec xx, Vec yy)
1453 {
1454   PetscFunctionBegin;
1455   PetscCall(VecSet(yy, 0.0));
1456   PetscCall(MatMultTransposeAdd_SeqAIJ(A, xx, yy, yy));
1457   PetscFunctionReturn(PETSC_SUCCESS);
1458 }
1459 
1460 #include <../src/mat/impls/aij/seq/ftn-kernels/fmult.h>
1461 
1462 PetscErrorCode MatMult_SeqAIJ(Mat A, Vec xx, Vec yy)
1463 {
1464   Mat_SeqAIJ        *a = (Mat_SeqAIJ *)A->data;
1465   PetscScalar       *y;
1466   const PetscScalar *x;
1467   const MatScalar   *aa, *a_a;
1468   PetscInt           m = A->rmap->n;
1469   const PetscInt    *aj, *ii, *ridx = NULL;
1470   PetscInt           n, i;
1471   PetscScalar        sum;
1472   PetscBool          usecprow = a->compressedrow.use;
1473 
1474 #if defined(PETSC_HAVE_PRAGMA_DISJOINT)
1475   #pragma disjoint(*x, *y, *aa)
1476 #endif
1477 
1478   PetscFunctionBegin;
1479   if (a->inode.use && a->inode.checked) {
1480     PetscCall(MatMult_SeqAIJ_Inode(A, xx, yy));
1481     PetscFunctionReturn(PETSC_SUCCESS);
1482   }
1483   PetscCall(MatSeqAIJGetArrayRead(A, &a_a));
1484   PetscCall(VecGetArrayRead(xx, &x));
1485   PetscCall(VecGetArray(yy, &y));
1486   ii = a->i;
1487   if (usecprow) { /* use compressed row format */
1488     PetscCall(PetscArrayzero(y, m));
1489     m    = a->compressedrow.nrows;
1490     ii   = a->compressedrow.i;
1491     ridx = a->compressedrow.rindex;
1492     for (i = 0; i < m; i++) {
1493       n   = ii[i + 1] - ii[i];
1494       aj  = a->j + ii[i];
1495       aa  = a_a + ii[i];
1496       sum = 0.0;
1497       PetscSparseDensePlusDot(sum, x, aa, aj, n);
1498       /* for (j=0; j<n; j++) sum += (*aa++)*x[*aj++]; */
1499       y[*ridx++] = sum;
1500     }
1501   } else { /* do not use compressed row format */
1502 #if defined(PETSC_USE_FORTRAN_KERNEL_MULTAIJ)
1503     aj = a->j;
1504     aa = a_a;
1505     fortranmultaij_(&m, x, ii, aj, aa, y);
1506 #else
1507     for (i = 0; i < m; i++) {
1508       n   = ii[i + 1] - ii[i];
1509       aj  = a->j + ii[i];
1510       aa  = a_a + ii[i];
1511       sum = 0.0;
1512       PetscSparseDensePlusDot(sum, x, aa, aj, n);
1513       y[i] = sum;
1514     }
1515 #endif
1516   }
1517   PetscCall(PetscLogFlops(2.0 * a->nz - a->nonzerorowcnt));
1518   PetscCall(VecRestoreArrayRead(xx, &x));
1519   PetscCall(VecRestoreArray(yy, &y));
1520   PetscCall(MatSeqAIJRestoreArrayRead(A, &a_a));
1521   PetscFunctionReturn(PETSC_SUCCESS);
1522 }
1523 
1524 // HACK!!!!! Used by src/mat/tests/ex170.c
1525 PETSC_EXTERN PetscErrorCode MatMultMax_SeqAIJ(Mat A, Vec xx, Vec yy)
1526 {
1527   Mat_SeqAIJ        *a = (Mat_SeqAIJ *)A->data;
1528   PetscScalar       *y;
1529   const PetscScalar *x;
1530   const MatScalar   *aa, *a_a;
1531   PetscInt           m = A->rmap->n;
1532   const PetscInt    *aj, *ii, *ridx   = NULL;
1533   PetscInt           n, i, nonzerorow = 0;
1534   PetscScalar        sum;
1535   PetscBool          usecprow = a->compressedrow.use;
1536 
1537 #if defined(PETSC_HAVE_PRAGMA_DISJOINT)
1538   #pragma disjoint(*x, *y, *aa)
1539 #endif
1540 
1541   PetscFunctionBegin;
1542   PetscCall(MatSeqAIJGetArrayRead(A, &a_a));
1543   PetscCall(VecGetArrayRead(xx, &x));
1544   PetscCall(VecGetArray(yy, &y));
1545   if (usecprow) { /* use compressed row format */
1546     m    = a->compressedrow.nrows;
1547     ii   = a->compressedrow.i;
1548     ridx = a->compressedrow.rindex;
1549     for (i = 0; i < m; i++) {
1550       n   = ii[i + 1] - ii[i];
1551       aj  = a->j + ii[i];
1552       aa  = a_a + ii[i];
1553       sum = 0.0;
1554       nonzerorow += (n > 0);
1555       PetscSparseDenseMaxDot(sum, x, aa, aj, n);
1556       /* for (j=0; j<n; j++) sum += (*aa++)*x[*aj++]; */
1557       y[*ridx++] = sum;
1558     }
1559   } else { /* do not use compressed row format */
1560     ii = a->i;
1561     for (i = 0; i < m; i++) {
1562       n   = ii[i + 1] - ii[i];
1563       aj  = a->j + ii[i];
1564       aa  = a_a + ii[i];
1565       sum = 0.0;
1566       nonzerorow += (n > 0);
1567       PetscSparseDenseMaxDot(sum, x, aa, aj, n);
1568       y[i] = sum;
1569     }
1570   }
1571   PetscCall(PetscLogFlops(2.0 * a->nz - nonzerorow));
1572   PetscCall(VecRestoreArrayRead(xx, &x));
1573   PetscCall(VecRestoreArray(yy, &y));
1574   PetscCall(MatSeqAIJRestoreArrayRead(A, &a_a));
1575   PetscFunctionReturn(PETSC_SUCCESS);
1576 }
1577 
1578 // HACK!!!!! Used by src/mat/tests/ex170.c
1579 PETSC_EXTERN PetscErrorCode MatMultAddMax_SeqAIJ(Mat A, Vec xx, Vec yy, Vec zz)
1580 {
1581   Mat_SeqAIJ        *a = (Mat_SeqAIJ *)A->data;
1582   PetscScalar       *y, *z;
1583   const PetscScalar *x;
1584   const MatScalar   *aa, *a_a;
1585   PetscInt           m = A->rmap->n, *aj, *ii;
1586   PetscInt           n, i, *ridx = NULL;
1587   PetscScalar        sum;
1588   PetscBool          usecprow = a->compressedrow.use;
1589 
1590   PetscFunctionBegin;
1591   PetscCall(MatSeqAIJGetArrayRead(A, &a_a));
1592   PetscCall(VecGetArrayRead(xx, &x));
1593   PetscCall(VecGetArrayPair(yy, zz, &y, &z));
1594   if (usecprow) { /* use compressed row format */
1595     if (zz != yy) PetscCall(PetscArraycpy(z, y, m));
1596     m    = a->compressedrow.nrows;
1597     ii   = a->compressedrow.i;
1598     ridx = a->compressedrow.rindex;
1599     for (i = 0; i < m; i++) {
1600       n   = ii[i + 1] - ii[i];
1601       aj  = a->j + ii[i];
1602       aa  = a_a + ii[i];
1603       sum = y[*ridx];
1604       PetscSparseDenseMaxDot(sum, x, aa, aj, n);
1605       z[*ridx++] = sum;
1606     }
1607   } else { /* do not use compressed row format */
1608     ii = a->i;
1609     for (i = 0; i < m; i++) {
1610       n   = ii[i + 1] - ii[i];
1611       aj  = a->j + ii[i];
1612       aa  = a_a + ii[i];
1613       sum = y[i];
1614       PetscSparseDenseMaxDot(sum, x, aa, aj, n);
1615       z[i] = sum;
1616     }
1617   }
1618   PetscCall(PetscLogFlops(2.0 * a->nz));
1619   PetscCall(VecRestoreArrayRead(xx, &x));
1620   PetscCall(VecRestoreArrayPair(yy, zz, &y, &z));
1621   PetscCall(MatSeqAIJRestoreArrayRead(A, &a_a));
1622   PetscFunctionReturn(PETSC_SUCCESS);
1623 }
1624 
1625 #include <../src/mat/impls/aij/seq/ftn-kernels/fmultadd.h>
1626 PetscErrorCode MatMultAdd_SeqAIJ(Mat A, Vec xx, Vec yy, Vec zz)
1627 {
1628   Mat_SeqAIJ        *a = (Mat_SeqAIJ *)A->data;
1629   PetscScalar       *y, *z;
1630   const PetscScalar *x;
1631   const MatScalar   *aa, *a_a;
1632   const PetscInt    *aj, *ii, *ridx = NULL;
1633   PetscInt           m = A->rmap->n, n, i;
1634   PetscScalar        sum;
1635   PetscBool          usecprow = a->compressedrow.use;
1636 
1637   PetscFunctionBegin;
1638   if (a->inode.use && a->inode.checked) {
1639     PetscCall(MatMultAdd_SeqAIJ_Inode(A, xx, yy, zz));
1640     PetscFunctionReturn(PETSC_SUCCESS);
1641   }
1642   PetscCall(MatSeqAIJGetArrayRead(A, &a_a));
1643   PetscCall(VecGetArrayRead(xx, &x));
1644   PetscCall(VecGetArrayPair(yy, zz, &y, &z));
1645   if (usecprow) { /* use compressed row format */
1646     if (zz != yy) PetscCall(PetscArraycpy(z, y, m));
1647     m    = a->compressedrow.nrows;
1648     ii   = a->compressedrow.i;
1649     ridx = a->compressedrow.rindex;
1650     for (i = 0; i < m; i++) {
1651       n   = ii[i + 1] - ii[i];
1652       aj  = a->j + ii[i];
1653       aa  = a_a + ii[i];
1654       sum = y[*ridx];
1655       PetscSparseDensePlusDot(sum, x, aa, aj, n);
1656       z[*ridx++] = sum;
1657     }
1658   } else { /* do not use compressed row format */
1659     ii = a->i;
1660 #if defined(PETSC_USE_FORTRAN_KERNEL_MULTADDAIJ)
1661     aj = a->j;
1662     aa = a_a;
1663     fortranmultaddaij_(&m, x, ii, aj, aa, y, z);
1664 #else
1665     for (i = 0; i < m; i++) {
1666       n   = ii[i + 1] - ii[i];
1667       aj  = a->j + ii[i];
1668       aa  = a_a + ii[i];
1669       sum = y[i];
1670       PetscSparseDensePlusDot(sum, x, aa, aj, n);
1671       z[i] = sum;
1672     }
1673 #endif
1674   }
1675   PetscCall(PetscLogFlops(2.0 * a->nz));
1676   PetscCall(VecRestoreArrayRead(xx, &x));
1677   PetscCall(VecRestoreArrayPair(yy, zz, &y, &z));
1678   PetscCall(MatSeqAIJRestoreArrayRead(A, &a_a));
1679   PetscFunctionReturn(PETSC_SUCCESS);
1680 }
1681 
1682 /*
1683      Adds diagonal pointers to sparse matrix structure.
1684 */
1685 PetscErrorCode MatMarkDiagonal_SeqAIJ(Mat A)
1686 {
1687   Mat_SeqAIJ *a = (Mat_SeqAIJ *)A->data;
1688   PetscInt    i, j, m = A->rmap->n;
1689   PetscBool   alreadySet = PETSC_TRUE;
1690 
1691   PetscFunctionBegin;
1692   if (!a->diag) {
1693     PetscCall(PetscMalloc1(m, &a->diag));
1694     alreadySet = PETSC_FALSE;
1695   }
1696   for (i = 0; i < A->rmap->n; i++) {
1697     /* If A's diagonal is already correctly set, this fast track enables cheap and repeated MatMarkDiagonal_SeqAIJ() calls */
1698     if (alreadySet) {
1699       PetscInt pos = a->diag[i];
1700       if (pos >= a->i[i] && pos < a->i[i + 1] && a->j[pos] == i) continue;
1701     }
1702 
1703     a->diag[i] = a->i[i + 1];
1704     for (j = a->i[i]; j < a->i[i + 1]; j++) {
1705       if (a->j[j] == i) {
1706         a->diag[i] = j;
1707         break;
1708       }
1709     }
1710   }
1711   PetscFunctionReturn(PETSC_SUCCESS);
1712 }
1713 
1714 static PetscErrorCode MatShift_SeqAIJ(Mat A, PetscScalar v)
1715 {
1716   Mat_SeqAIJ     *a    = (Mat_SeqAIJ *)A->data;
1717   const PetscInt *diag = (const PetscInt *)a->diag;
1718   const PetscInt *ii   = (const PetscInt *)a->i;
1719   PetscInt        i, *mdiag = NULL;
1720   PetscInt        cnt = 0; /* how many diagonals are missing */
1721 
1722   PetscFunctionBegin;
1723   if (!A->preallocated || !a->nz) {
1724     PetscCall(MatSeqAIJSetPreallocation(A, 1, NULL));
1725     PetscCall(MatShift_Basic(A, v));
1726     PetscFunctionReturn(PETSC_SUCCESS);
1727   }
1728 
1729   if (a->diagonaldense) {
1730     cnt = 0;
1731   } else {
1732     PetscCall(PetscCalloc1(A->rmap->n, &mdiag));
1733     for (i = 0; i < A->rmap->n; i++) {
1734       if (i < A->cmap->n && diag[i] >= ii[i + 1]) { /* 'out of range' rows never have diagonals */
1735         cnt++;
1736         mdiag[i] = 1;
1737       }
1738     }
1739   }
1740   if (!cnt) {
1741     PetscCall(MatShift_Basic(A, v));
1742   } else {
1743     PetscScalar       *olda = a->a; /* preserve pointers to current matrix nonzeros structure and values */
1744     PetscInt          *oldj = a->j, *oldi = a->i;
1745     PetscBool          singlemalloc = a->singlemalloc, free_a = a->free_a, free_ij = a->free_ij;
1746     const PetscScalar *Aa;
1747 
1748     PetscCall(MatSeqAIJGetArrayRead(A, &Aa)); // sync the host
1749     PetscCall(MatSeqAIJRestoreArrayRead(A, &Aa));
1750 
1751     a->a = NULL;
1752     a->j = NULL;
1753     a->i = NULL;
1754     /* increase the values in imax for each row where a diagonal is being inserted then reallocate the matrix data structures */
1755     for (i = 0; i < PetscMin(A->rmap->n, A->cmap->n); i++) a->imax[i] += mdiag[i];
1756     PetscCall(MatSeqAIJSetPreallocation_SeqAIJ(A, 0, a->imax));
1757 
1758     /* copy old values into new matrix data structure */
1759     for (i = 0; i < A->rmap->n; i++) {
1760       PetscCall(MatSetValues(A, 1, &i, a->imax[i] - mdiag[i], &oldj[oldi[i]], &olda[oldi[i]], ADD_VALUES));
1761       if (i < A->cmap->n) PetscCall(MatSetValue(A, i, i, v, ADD_VALUES));
1762     }
1763     PetscCall(MatAssemblyBegin(A, MAT_FINAL_ASSEMBLY));
1764     PetscCall(MatAssemblyEnd(A, MAT_FINAL_ASSEMBLY));
1765     if (singlemalloc) {
1766       PetscCall(PetscFree3(olda, oldj, oldi));
1767     } else {
1768       if (free_a) PetscCall(PetscFree(olda));
1769       if (free_ij) PetscCall(PetscFree(oldj));
1770       if (free_ij) PetscCall(PetscFree(oldi));
1771     }
1772   }
1773   PetscCall(PetscFree(mdiag));
1774   a->diagonaldense = PETSC_TRUE;
1775   PetscFunctionReturn(PETSC_SUCCESS);
1776 }
1777 
1778 /*
1779      Checks for missing diagonals
1780 */
1781 PetscErrorCode MatMissingDiagonal_SeqAIJ(Mat A, PetscBool *missing, PetscInt *d)
1782 {
1783   Mat_SeqAIJ *a = (Mat_SeqAIJ *)A->data;
1784   PetscInt   *diag, *ii = a->i, i;
1785 
1786   PetscFunctionBegin;
1787   *missing = PETSC_FALSE;
1788   if (A->rmap->n > 0 && !ii) {
1789     *missing = PETSC_TRUE;
1790     if (d) *d = 0;
1791     PetscCall(PetscInfo(A, "Matrix has no entries therefore is missing diagonal\n"));
1792   } else {
1793     PetscInt n;
1794     n    = PetscMin(A->rmap->n, A->cmap->n);
1795     diag = a->diag;
1796     for (i = 0; i < n; i++) {
1797       if (diag[i] >= ii[i + 1]) {
1798         *missing = PETSC_TRUE;
1799         if (d) *d = i;
1800         PetscCall(PetscInfo(A, "Matrix is missing diagonal number %" PetscInt_FMT "\n", i));
1801         break;
1802       }
1803     }
1804   }
1805   PetscFunctionReturn(PETSC_SUCCESS);
1806 }
1807 
1808 #include <petscblaslapack.h>
1809 #include <petsc/private/kernels/blockinvert.h>
1810 
1811 /*
1812     Note that values is allocated externally by the PC and then passed into this routine
1813 */
1814 static PetscErrorCode MatInvertVariableBlockDiagonal_SeqAIJ(Mat A, PetscInt nblocks, const PetscInt *bsizes, PetscScalar *diag)
1815 {
1816   PetscInt        n = A->rmap->n, i, ncnt = 0, *indx, j, bsizemax = 0, *v_pivots;
1817   PetscBool       allowzeropivot, zeropivotdetected = PETSC_FALSE;
1818   const PetscReal shift = 0.0;
1819   PetscInt        ipvt[5];
1820   PetscCount      flops = 0;
1821   PetscScalar     work[25], *v_work;
1822 
1823   PetscFunctionBegin;
1824   allowzeropivot = PetscNot(A->erroriffailure);
1825   for (i = 0; i < nblocks; i++) ncnt += bsizes[i];
1826   PetscCheck(ncnt == n, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Total blocksizes %" PetscInt_FMT " doesn't match number matrix rows %" PetscInt_FMT, ncnt, n);
1827   for (i = 0; i < nblocks; i++) bsizemax = PetscMax(bsizemax, bsizes[i]);
1828   PetscCall(PetscMalloc1(bsizemax, &indx));
1829   if (bsizemax > 7) PetscCall(PetscMalloc2(bsizemax, &v_work, bsizemax, &v_pivots));
1830   ncnt = 0;
1831   for (i = 0; i < nblocks; i++) {
1832     for (j = 0; j < bsizes[i]; j++) indx[j] = ncnt + j;
1833     PetscCall(MatGetValues(A, bsizes[i], indx, bsizes[i], indx, diag));
1834     switch (bsizes[i]) {
1835     case 1:
1836       *diag = 1.0 / (*diag);
1837       break;
1838     case 2:
1839       PetscCall(PetscKernel_A_gets_inverse_A_2(diag, shift, allowzeropivot, &zeropivotdetected));
1840       if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
1841       PetscCall(PetscKernel_A_gets_transpose_A_2(diag));
1842       break;
1843     case 3:
1844       PetscCall(PetscKernel_A_gets_inverse_A_3(diag, shift, allowzeropivot, &zeropivotdetected));
1845       if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
1846       PetscCall(PetscKernel_A_gets_transpose_A_3(diag));
1847       break;
1848     case 4:
1849       PetscCall(PetscKernel_A_gets_inverse_A_4(diag, shift, allowzeropivot, &zeropivotdetected));
1850       if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
1851       PetscCall(PetscKernel_A_gets_transpose_A_4(diag));
1852       break;
1853     case 5:
1854       PetscCall(PetscKernel_A_gets_inverse_A_5(diag, ipvt, work, shift, allowzeropivot, &zeropivotdetected));
1855       if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
1856       PetscCall(PetscKernel_A_gets_transpose_A_5(diag));
1857       break;
1858     case 6:
1859       PetscCall(PetscKernel_A_gets_inverse_A_6(diag, shift, allowzeropivot, &zeropivotdetected));
1860       if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
1861       PetscCall(PetscKernel_A_gets_transpose_A_6(diag));
1862       break;
1863     case 7:
1864       PetscCall(PetscKernel_A_gets_inverse_A_7(diag, shift, allowzeropivot, &zeropivotdetected));
1865       if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
1866       PetscCall(PetscKernel_A_gets_transpose_A_7(diag));
1867       break;
1868     default:
1869       PetscCall(PetscKernel_A_gets_inverse_A(bsizes[i], diag, v_pivots, v_work, allowzeropivot, &zeropivotdetected));
1870       if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
1871       PetscCall(PetscKernel_A_gets_transpose_A_N(diag, bsizes[i]));
1872     }
1873     ncnt += bsizes[i];
1874     diag += bsizes[i] * bsizes[i];
1875     flops += 2 * PetscPowInt(bsizes[i], 3) / 3;
1876   }
1877   PetscCall(PetscLogFlops(flops));
1878   if (bsizemax > 7) PetscCall(PetscFree2(v_work, v_pivots));
1879   PetscCall(PetscFree(indx));
1880   PetscFunctionReturn(PETSC_SUCCESS);
1881 }
1882 
1883 /*
1884    Negative shift indicates do not generate an error if there is a zero diagonal, just invert it anyways
1885 */
1886 static PetscErrorCode MatInvertDiagonal_SeqAIJ(Mat A, PetscScalar omega, PetscScalar fshift)
1887 {
1888   Mat_SeqAIJ      *a = (Mat_SeqAIJ *)A->data;
1889   PetscInt         i, *diag, m = A->rmap->n;
1890   const MatScalar *v;
1891   PetscScalar     *idiag, *mdiag;
1892 
1893   PetscFunctionBegin;
1894   if (a->idiagvalid) PetscFunctionReturn(PETSC_SUCCESS);
1895   PetscCall(MatMarkDiagonal_SeqAIJ(A));
1896   diag = a->diag;
1897   if (!a->idiag) { PetscCall(PetscMalloc3(m, &a->idiag, m, &a->mdiag, m, &a->ssor_work)); }
1898 
1899   mdiag = a->mdiag;
1900   idiag = a->idiag;
1901   PetscCall(MatSeqAIJGetArrayRead(A, &v));
1902   if (omega == 1.0 && PetscRealPart(fshift) <= 0.0) {
1903     for (i = 0; i < m; i++) {
1904       mdiag[i] = v[diag[i]];
1905       if (!PetscAbsScalar(mdiag[i])) { /* zero diagonal */
1906         if (PetscRealPart(fshift)) {
1907           PetscCall(PetscInfo(A, "Zero diagonal on row %" PetscInt_FMT "\n", i));
1908           A->factorerrortype             = MAT_FACTOR_NUMERIC_ZEROPIVOT;
1909           A->factorerror_zeropivot_value = 0.0;
1910           A->factorerror_zeropivot_row   = i;
1911         } else SETERRQ(PETSC_COMM_SELF, PETSC_ERR_ARG_INCOMP, "Zero diagonal on row %" PetscInt_FMT, i);
1912       }
1913       idiag[i] = 1.0 / v[diag[i]];
1914     }
1915     PetscCall(PetscLogFlops(m));
1916   } else {
1917     for (i = 0; i < m; i++) {
1918       mdiag[i] = v[diag[i]];
1919       idiag[i] = omega / (fshift + v[diag[i]]);
1920     }
1921     PetscCall(PetscLogFlops(2.0 * m));
1922   }
1923   a->idiagvalid = PETSC_TRUE;
1924   PetscCall(MatSeqAIJRestoreArrayRead(A, &v));
1925   PetscFunctionReturn(PETSC_SUCCESS);
1926 }
1927 
1928 PetscErrorCode MatSOR_SeqAIJ(Mat A, Vec bb, PetscReal omega, MatSORType flag, PetscReal fshift, PetscInt its, PetscInt lits, Vec xx)
1929 {
1930   Mat_SeqAIJ        *a = (Mat_SeqAIJ *)A->data;
1931   PetscScalar       *x, d, sum, *t, scale;
1932   const MatScalar   *v, *idiag = NULL, *mdiag, *aa;
1933   const PetscScalar *b, *bs, *xb, *ts;
1934   PetscInt           n, m = A->rmap->n, i;
1935   const PetscInt    *idx, *diag;
1936 
1937   PetscFunctionBegin;
1938   if (a->inode.use && a->inode.checked && omega == 1.0 && fshift == 0.0) {
1939     PetscCall(MatSOR_SeqAIJ_Inode(A, bb, omega, flag, fshift, its, lits, xx));
1940     PetscFunctionReturn(PETSC_SUCCESS);
1941   }
1942   its = its * lits;
1943 
1944   if (fshift != a->fshift || omega != a->omega) a->idiagvalid = PETSC_FALSE; /* must recompute idiag[] */
1945   if (!a->idiagvalid) PetscCall(MatInvertDiagonal_SeqAIJ(A, omega, fshift));
1946   a->fshift = fshift;
1947   a->omega  = omega;
1948 
1949   diag  = a->diag;
1950   t     = a->ssor_work;
1951   idiag = a->idiag;
1952   mdiag = a->mdiag;
1953 
1954   PetscCall(MatSeqAIJGetArrayRead(A, &aa));
1955   PetscCall(VecGetArray(xx, &x));
1956   PetscCall(VecGetArrayRead(bb, &b));
1957   /* We count flops by assuming the upper triangular and lower triangular parts have the same number of nonzeros */
1958   if (flag == SOR_APPLY_UPPER) {
1959     /* apply (U + D/omega) to the vector */
1960     bs = b;
1961     for (i = 0; i < m; i++) {
1962       d   = fshift + mdiag[i];
1963       n   = a->i[i + 1] - diag[i] - 1;
1964       idx = a->j + diag[i] + 1;
1965       v   = aa + diag[i] + 1;
1966       sum = b[i] * d / omega;
1967       PetscSparseDensePlusDot(sum, bs, v, idx, n);
1968       x[i] = sum;
1969     }
1970     PetscCall(VecRestoreArray(xx, &x));
1971     PetscCall(VecRestoreArrayRead(bb, &b));
1972     PetscCall(MatSeqAIJRestoreArrayRead(A, &aa));
1973     PetscCall(PetscLogFlops(a->nz));
1974     PetscFunctionReturn(PETSC_SUCCESS);
1975   }
1976 
1977   PetscCheck(flag != SOR_APPLY_LOWER, PETSC_COMM_SELF, PETSC_ERR_SUP, "SOR_APPLY_LOWER is not implemented");
1978   if (flag & SOR_EISENSTAT) {
1979     /* Let  A = L + U + D; where L is lower triangular,
1980     U is upper triangular, E = D/omega; This routine applies
1981 
1982             (L + E)^{-1} A (U + E)^{-1}
1983 
1984     to a vector efficiently using Eisenstat's trick.
1985     */
1986     scale = (2.0 / omega) - 1.0;
1987 
1988     /*  x = (E + U)^{-1} b */
1989     for (i = m - 1; i >= 0; i--) {
1990       n   = a->i[i + 1] - diag[i] - 1;
1991       idx = a->j + diag[i] + 1;
1992       v   = aa + diag[i] + 1;
1993       sum = b[i];
1994       PetscSparseDenseMinusDot(sum, x, v, idx, n);
1995       x[i] = sum * idiag[i];
1996     }
1997 
1998     /*  t = b - (2*E - D)x */
1999     v = aa;
2000     for (i = 0; i < m; i++) t[i] = b[i] - scale * (v[*diag++]) * x[i];
2001 
2002     /*  t = (E + L)^{-1}t */
2003     ts   = t;
2004     diag = a->diag;
2005     for (i = 0; i < m; i++) {
2006       n   = diag[i] - a->i[i];
2007       idx = a->j + a->i[i];
2008       v   = aa + a->i[i];
2009       sum = t[i];
2010       PetscSparseDenseMinusDot(sum, ts, v, idx, n);
2011       t[i] = sum * idiag[i];
2012       /*  x = x + t */
2013       x[i] += t[i];
2014     }
2015 
2016     PetscCall(PetscLogFlops(6.0 * m - 1 + 2.0 * a->nz));
2017     PetscCall(VecRestoreArray(xx, &x));
2018     PetscCall(VecRestoreArrayRead(bb, &b));
2019     PetscFunctionReturn(PETSC_SUCCESS);
2020   }
2021   if (flag & SOR_ZERO_INITIAL_GUESS) {
2022     if (flag & SOR_FORWARD_SWEEP || flag & SOR_LOCAL_FORWARD_SWEEP) {
2023       for (i = 0; i < m; i++) {
2024         n   = diag[i] - a->i[i];
2025         idx = a->j + a->i[i];
2026         v   = aa + a->i[i];
2027         sum = b[i];
2028         PetscSparseDenseMinusDot(sum, x, v, idx, n);
2029         t[i] = sum;
2030         x[i] = sum * idiag[i];
2031       }
2032       xb = t;
2033       PetscCall(PetscLogFlops(a->nz));
2034     } else xb = b;
2035     if (flag & SOR_BACKWARD_SWEEP || flag & SOR_LOCAL_BACKWARD_SWEEP) {
2036       for (i = m - 1; i >= 0; i--) {
2037         n   = a->i[i + 1] - diag[i] - 1;
2038         idx = a->j + diag[i] + 1;
2039         v   = aa + diag[i] + 1;
2040         sum = xb[i];
2041         PetscSparseDenseMinusDot(sum, x, v, idx, n);
2042         if (xb == b) {
2043           x[i] = sum * idiag[i];
2044         } else {
2045           x[i] = (1 - omega) * x[i] + sum * idiag[i]; /* omega in idiag */
2046         }
2047       }
2048       PetscCall(PetscLogFlops(a->nz)); /* assumes 1/2 in upper */
2049     }
2050     its--;
2051   }
2052   while (its--) {
2053     if (flag & SOR_FORWARD_SWEEP || flag & SOR_LOCAL_FORWARD_SWEEP) {
2054       for (i = 0; i < m; i++) {
2055         /* lower */
2056         n   = diag[i] - a->i[i];
2057         idx = a->j + a->i[i];
2058         v   = aa + a->i[i];
2059         sum = b[i];
2060         PetscSparseDenseMinusDot(sum, x, v, idx, n);
2061         t[i] = sum; /* save application of the lower-triangular part */
2062         /* upper */
2063         n   = a->i[i + 1] - diag[i] - 1;
2064         idx = a->j + diag[i] + 1;
2065         v   = aa + diag[i] + 1;
2066         PetscSparseDenseMinusDot(sum, x, v, idx, n);
2067         x[i] = (1. - omega) * x[i] + sum * idiag[i]; /* omega in idiag */
2068       }
2069       xb = t;
2070       PetscCall(PetscLogFlops(2.0 * a->nz));
2071     } else xb = b;
2072     if (flag & SOR_BACKWARD_SWEEP || flag & SOR_LOCAL_BACKWARD_SWEEP) {
2073       for (i = m - 1; i >= 0; i--) {
2074         sum = xb[i];
2075         if (xb == b) {
2076           /* whole matrix (no checkpointing available) */
2077           n   = a->i[i + 1] - a->i[i];
2078           idx = a->j + a->i[i];
2079           v   = aa + a->i[i];
2080           PetscSparseDenseMinusDot(sum, x, v, idx, n);
2081           x[i] = (1. - omega) * x[i] + (sum + mdiag[i] * x[i]) * idiag[i];
2082         } else { /* lower-triangular part has been saved, so only apply upper-triangular */
2083           n   = a->i[i + 1] - diag[i] - 1;
2084           idx = a->j + diag[i] + 1;
2085           v   = aa + diag[i] + 1;
2086           PetscSparseDenseMinusDot(sum, x, v, idx, n);
2087           x[i] = (1. - omega) * x[i] + sum * idiag[i]; /* omega in idiag */
2088         }
2089       }
2090       if (xb == b) {
2091         PetscCall(PetscLogFlops(2.0 * a->nz));
2092       } else {
2093         PetscCall(PetscLogFlops(a->nz)); /* assumes 1/2 in upper */
2094       }
2095     }
2096   }
2097   PetscCall(MatSeqAIJRestoreArrayRead(A, &aa));
2098   PetscCall(VecRestoreArray(xx, &x));
2099   PetscCall(VecRestoreArrayRead(bb, &b));
2100   PetscFunctionReturn(PETSC_SUCCESS);
2101 }
2102 
2103 static PetscErrorCode MatGetInfo_SeqAIJ(Mat A, MatInfoType flag, MatInfo *info)
2104 {
2105   Mat_SeqAIJ *a = (Mat_SeqAIJ *)A->data;
2106 
2107   PetscFunctionBegin;
2108   info->block_size   = 1.0;
2109   info->nz_allocated = a->maxnz;
2110   info->nz_used      = a->nz;
2111   info->nz_unneeded  = (a->maxnz - a->nz);
2112   info->assemblies   = A->num_ass;
2113   info->mallocs      = A->info.mallocs;
2114   info->memory       = 0; /* REVIEW ME */
2115   if (A->factortype) {
2116     info->fill_ratio_given  = A->info.fill_ratio_given;
2117     info->fill_ratio_needed = A->info.fill_ratio_needed;
2118     info->factor_mallocs    = A->info.factor_mallocs;
2119   } else {
2120     info->fill_ratio_given  = 0;
2121     info->fill_ratio_needed = 0;
2122     info->factor_mallocs    = 0;
2123   }
2124   PetscFunctionReturn(PETSC_SUCCESS);
2125 }
2126 
2127 static PetscErrorCode MatZeroRows_SeqAIJ(Mat A, PetscInt N, const PetscInt rows[], PetscScalar diag, Vec x, Vec b)
2128 {
2129   Mat_SeqAIJ        *a = (Mat_SeqAIJ *)A->data;
2130   PetscInt           i, m = A->rmap->n - 1;
2131   const PetscScalar *xx;
2132   PetscScalar       *bb, *aa;
2133   PetscInt           d = 0;
2134 
2135   PetscFunctionBegin;
2136   if (x && b) {
2137     PetscCall(VecGetArrayRead(x, &xx));
2138     PetscCall(VecGetArray(b, &bb));
2139     for (i = 0; i < N; i++) {
2140       PetscCheck(rows[i] >= 0 && rows[i] <= m, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "row %" PetscInt_FMT " out of range", rows[i]);
2141       if (rows[i] >= A->cmap->n) continue;
2142       bb[rows[i]] = diag * xx[rows[i]];
2143     }
2144     PetscCall(VecRestoreArrayRead(x, &xx));
2145     PetscCall(VecRestoreArray(b, &bb));
2146   }
2147 
2148   PetscCall(MatSeqAIJGetArray(A, &aa));
2149   if (a->keepnonzeropattern) {
2150     for (i = 0; i < N; i++) {
2151       PetscCheck(rows[i] >= 0 && rows[i] <= m, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "row %" PetscInt_FMT " out of range", rows[i]);
2152       PetscCall(PetscArrayzero(&aa[a->i[rows[i]]], a->ilen[rows[i]]));
2153     }
2154     if (diag != 0.0) {
2155       for (i = 0; i < N; i++) {
2156         d = rows[i];
2157         if (rows[i] >= A->cmap->n) continue;
2158         PetscCheck(a->diag[d] < a->i[d + 1], PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Matrix is missing diagonal entry in the zeroed row %" PetscInt_FMT, d);
2159       }
2160       for (i = 0; i < N; i++) {
2161         if (rows[i] >= A->cmap->n) continue;
2162         aa[a->diag[rows[i]]] = diag;
2163       }
2164     }
2165   } else {
2166     if (diag != 0.0) {
2167       for (i = 0; i < N; i++) {
2168         PetscCheck(rows[i] >= 0 && rows[i] <= m, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "row %" PetscInt_FMT " out of range", rows[i]);
2169         if (a->ilen[rows[i]] > 0) {
2170           if (rows[i] >= A->cmap->n) {
2171             a->ilen[rows[i]] = 0;
2172           } else {
2173             a->ilen[rows[i]]    = 1;
2174             aa[a->i[rows[i]]]   = diag;
2175             a->j[a->i[rows[i]]] = rows[i];
2176           }
2177         } else if (rows[i] < A->cmap->n) { /* in case row was completely empty */
2178           PetscCall(MatSetValues_SeqAIJ(A, 1, &rows[i], 1, &rows[i], &diag, INSERT_VALUES));
2179         }
2180       }
2181     } else {
2182       for (i = 0; i < N; i++) {
2183         PetscCheck(rows[i] >= 0 && rows[i] <= m, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "row %" PetscInt_FMT " out of range", rows[i]);
2184         a->ilen[rows[i]] = 0;
2185       }
2186     }
2187     A->nonzerostate++;
2188   }
2189   PetscCall(MatSeqAIJRestoreArray(A, &aa));
2190   PetscUseTypeMethod(A, assemblyend, MAT_FINAL_ASSEMBLY);
2191   PetscFunctionReturn(PETSC_SUCCESS);
2192 }
2193 
2194 static PetscErrorCode MatZeroRowsColumns_SeqAIJ(Mat A, PetscInt N, const PetscInt rows[], PetscScalar diag, Vec x, Vec b)
2195 {
2196   Mat_SeqAIJ        *a = (Mat_SeqAIJ *)A->data;
2197   PetscInt           i, j, m = A->rmap->n - 1, d = 0;
2198   PetscBool          missing, *zeroed, vecs = PETSC_FALSE;
2199   const PetscScalar *xx;
2200   PetscScalar       *bb, *aa;
2201 
2202   PetscFunctionBegin;
2203   if (!N) PetscFunctionReturn(PETSC_SUCCESS);
2204   PetscCall(MatSeqAIJGetArray(A, &aa));
2205   if (x && b) {
2206     PetscCall(VecGetArrayRead(x, &xx));
2207     PetscCall(VecGetArray(b, &bb));
2208     vecs = PETSC_TRUE;
2209   }
2210   PetscCall(PetscCalloc1(A->rmap->n, &zeroed));
2211   for (i = 0; i < N; i++) {
2212     PetscCheck(rows[i] >= 0 && rows[i] <= m, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "row %" PetscInt_FMT " out of range", rows[i]);
2213     PetscCall(PetscArrayzero(PetscSafePointerPlusOffset(aa, a->i[rows[i]]), a->ilen[rows[i]]));
2214 
2215     zeroed[rows[i]] = PETSC_TRUE;
2216   }
2217   for (i = 0; i < A->rmap->n; i++) {
2218     if (!zeroed[i]) {
2219       for (j = a->i[i]; j < a->i[i + 1]; j++) {
2220         if (a->j[j] < A->rmap->n && zeroed[a->j[j]]) {
2221           if (vecs) bb[i] -= aa[j] * xx[a->j[j]];
2222           aa[j] = 0.0;
2223         }
2224       }
2225     } else if (vecs && i < A->cmap->N) bb[i] = diag * xx[i];
2226   }
2227   if (x && b) {
2228     PetscCall(VecRestoreArrayRead(x, &xx));
2229     PetscCall(VecRestoreArray(b, &bb));
2230   }
2231   PetscCall(PetscFree(zeroed));
2232   if (diag != 0.0) {
2233     PetscCall(MatMissingDiagonal_SeqAIJ(A, &missing, &d));
2234     if (missing) {
2235       for (i = 0; i < N; i++) {
2236         if (rows[i] >= A->cmap->N) continue;
2237         PetscCheck(!a->nonew || rows[i] < d, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Matrix is missing diagonal entry in row %" PetscInt_FMT " (%" PetscInt_FMT ")", d, rows[i]);
2238         PetscCall(MatSetValues_SeqAIJ(A, 1, &rows[i], 1, &rows[i], &diag, INSERT_VALUES));
2239       }
2240     } else {
2241       for (i = 0; i < N; i++) aa[a->diag[rows[i]]] = diag;
2242     }
2243   }
2244   PetscCall(MatSeqAIJRestoreArray(A, &aa));
2245   PetscUseTypeMethod(A, assemblyend, MAT_FINAL_ASSEMBLY);
2246   PetscFunctionReturn(PETSC_SUCCESS);
2247 }
2248 
2249 PetscErrorCode MatGetRow_SeqAIJ(Mat A, PetscInt row, PetscInt *nz, PetscInt **idx, PetscScalar **v)
2250 {
2251   Mat_SeqAIJ        *a = (Mat_SeqAIJ *)A->data;
2252   const PetscScalar *aa;
2253 
2254   PetscFunctionBegin;
2255   PetscCall(MatSeqAIJGetArrayRead(A, &aa));
2256   *nz = a->i[row + 1] - a->i[row];
2257   if (v) *v = PetscSafePointerPlusOffset((PetscScalar *)aa, a->i[row]);
2258   if (idx) {
2259     if (*nz && a->j) *idx = a->j + a->i[row];
2260     else *idx = NULL;
2261   }
2262   PetscCall(MatSeqAIJRestoreArrayRead(A, &aa));
2263   PetscFunctionReturn(PETSC_SUCCESS);
2264 }
2265 
2266 PetscErrorCode MatRestoreRow_SeqAIJ(Mat A, PetscInt row, PetscInt *nz, PetscInt **idx, PetscScalar **v)
2267 {
2268   PetscFunctionBegin;
2269   PetscFunctionReturn(PETSC_SUCCESS);
2270 }
2271 
2272 static PetscErrorCode MatNorm_SeqAIJ(Mat A, NormType type, PetscReal *nrm)
2273 {
2274   Mat_SeqAIJ      *a = (Mat_SeqAIJ *)A->data;
2275   const MatScalar *v;
2276   PetscReal        sum = 0.0;
2277   PetscInt         i, j;
2278 
2279   PetscFunctionBegin;
2280   PetscCall(MatSeqAIJGetArrayRead(A, &v));
2281   if (type == NORM_FROBENIUS) {
2282 #if defined(PETSC_USE_REAL___FP16)
2283     PetscBLASInt one = 1, nz = a->nz;
2284     PetscCallBLAS("BLASnrm2", *nrm = BLASnrm2_(&nz, v, &one));
2285 #else
2286     for (i = 0; i < a->nz; i++) {
2287       sum += PetscRealPart(PetscConj(*v) * (*v));
2288       v++;
2289     }
2290     *nrm = PetscSqrtReal(sum);
2291 #endif
2292     PetscCall(PetscLogFlops(2.0 * a->nz));
2293   } else if (type == NORM_1) {
2294     PetscReal *tmp;
2295     PetscInt  *jj = a->j;
2296     PetscCall(PetscCalloc1(A->cmap->n + 1, &tmp));
2297     *nrm = 0.0;
2298     for (j = 0; j < a->nz; j++) {
2299       tmp[*jj++] += PetscAbsScalar(*v);
2300       v++;
2301     }
2302     for (j = 0; j < A->cmap->n; j++) {
2303       if (tmp[j] > *nrm) *nrm = tmp[j];
2304     }
2305     PetscCall(PetscFree(tmp));
2306     PetscCall(PetscLogFlops(PetscMax(a->nz - 1, 0)));
2307   } else if (type == NORM_INFINITY) {
2308     *nrm = 0.0;
2309     for (j = 0; j < A->rmap->n; j++) {
2310       const PetscScalar *v2 = PetscSafePointerPlusOffset(v, a->i[j]);
2311       sum                   = 0.0;
2312       for (i = 0; i < a->i[j + 1] - a->i[j]; i++) {
2313         sum += PetscAbsScalar(*v2);
2314         v2++;
2315       }
2316       if (sum > *nrm) *nrm = sum;
2317     }
2318     PetscCall(PetscLogFlops(PetscMax(a->nz - 1, 0)));
2319   } else SETERRQ(PETSC_COMM_SELF, PETSC_ERR_SUP, "No support for two norm");
2320   PetscCall(MatSeqAIJRestoreArrayRead(A, &v));
2321   PetscFunctionReturn(PETSC_SUCCESS);
2322 }
2323 
2324 static PetscErrorCode MatIsTranspose_SeqAIJ(Mat A, Mat B, PetscReal tol, PetscBool *f)
2325 {
2326   Mat_SeqAIJ      *aij = (Mat_SeqAIJ *)A->data, *bij = (Mat_SeqAIJ *)B->data;
2327   PetscInt        *adx, *bdx, *aii, *bii, *aptr, *bptr;
2328   const MatScalar *va, *vb;
2329   PetscInt         ma, na, mb, nb, i;
2330 
2331   PetscFunctionBegin;
2332   PetscCall(MatGetSize(A, &ma, &na));
2333   PetscCall(MatGetSize(B, &mb, &nb));
2334   if (ma != nb || na != mb) {
2335     *f = PETSC_FALSE;
2336     PetscFunctionReturn(PETSC_SUCCESS);
2337   }
2338   PetscCall(MatSeqAIJGetArrayRead(A, &va));
2339   PetscCall(MatSeqAIJGetArrayRead(B, &vb));
2340   aii = aij->i;
2341   bii = bij->i;
2342   adx = aij->j;
2343   bdx = bij->j;
2344   PetscCall(PetscMalloc1(ma, &aptr));
2345   PetscCall(PetscMalloc1(mb, &bptr));
2346   for (i = 0; i < ma; i++) aptr[i] = aii[i];
2347   for (i = 0; i < mb; i++) bptr[i] = bii[i];
2348 
2349   *f = PETSC_TRUE;
2350   for (i = 0; i < ma; i++) {
2351     while (aptr[i] < aii[i + 1]) {
2352       PetscInt    idc, idr;
2353       PetscScalar vc, vr;
2354       /* column/row index/value */
2355       idc = adx[aptr[i]];
2356       idr = bdx[bptr[idc]];
2357       vc  = va[aptr[i]];
2358       vr  = vb[bptr[idc]];
2359       if (i != idr || PetscAbsScalar(vc - vr) > tol) {
2360         *f = PETSC_FALSE;
2361         goto done;
2362       } else {
2363         aptr[i]++;
2364         if (B || i != idc) bptr[idc]++;
2365       }
2366     }
2367   }
2368 done:
2369   PetscCall(PetscFree(aptr));
2370   PetscCall(PetscFree(bptr));
2371   PetscCall(MatSeqAIJRestoreArrayRead(A, &va));
2372   PetscCall(MatSeqAIJRestoreArrayRead(B, &vb));
2373   PetscFunctionReturn(PETSC_SUCCESS);
2374 }
2375 
2376 static PetscErrorCode MatIsHermitianTranspose_SeqAIJ(Mat A, Mat B, PetscReal tol, PetscBool *f)
2377 {
2378   Mat_SeqAIJ *aij = (Mat_SeqAIJ *)A->data, *bij = (Mat_SeqAIJ *)B->data;
2379   PetscInt   *adx, *bdx, *aii, *bii, *aptr, *bptr;
2380   MatScalar  *va, *vb;
2381   PetscInt    ma, na, mb, nb, i;
2382 
2383   PetscFunctionBegin;
2384   PetscCall(MatGetSize(A, &ma, &na));
2385   PetscCall(MatGetSize(B, &mb, &nb));
2386   if (ma != nb || na != mb) {
2387     *f = PETSC_FALSE;
2388     PetscFunctionReturn(PETSC_SUCCESS);
2389   }
2390   aii = aij->i;
2391   bii = bij->i;
2392   adx = aij->j;
2393   bdx = bij->j;
2394   va  = aij->a;
2395   vb  = bij->a;
2396   PetscCall(PetscMalloc1(ma, &aptr));
2397   PetscCall(PetscMalloc1(mb, &bptr));
2398   for (i = 0; i < ma; i++) aptr[i] = aii[i];
2399   for (i = 0; i < mb; i++) bptr[i] = bii[i];
2400 
2401   *f = PETSC_TRUE;
2402   for (i = 0; i < ma; i++) {
2403     while (aptr[i] < aii[i + 1]) {
2404       PetscInt    idc, idr;
2405       PetscScalar vc, vr;
2406       /* column/row index/value */
2407       idc = adx[aptr[i]];
2408       idr = bdx[bptr[idc]];
2409       vc  = va[aptr[i]];
2410       vr  = vb[bptr[idc]];
2411       if (i != idr || PetscAbsScalar(vc - PetscConj(vr)) > tol) {
2412         *f = PETSC_FALSE;
2413         goto done;
2414       } else {
2415         aptr[i]++;
2416         if (B || i != idc) bptr[idc]++;
2417       }
2418     }
2419   }
2420 done:
2421   PetscCall(PetscFree(aptr));
2422   PetscCall(PetscFree(bptr));
2423   PetscFunctionReturn(PETSC_SUCCESS);
2424 }
2425 
2426 PetscErrorCode MatDiagonalScale_SeqAIJ(Mat A, Vec ll, Vec rr)
2427 {
2428   Mat_SeqAIJ        *a = (Mat_SeqAIJ *)A->data;
2429   const PetscScalar *l, *r;
2430   PetscScalar        x;
2431   MatScalar         *v;
2432   PetscInt           i, j, m = A->rmap->n, n = A->cmap->n, M, nz = a->nz;
2433   const PetscInt    *jj;
2434 
2435   PetscFunctionBegin;
2436   if (ll) {
2437     /* The local size is used so that VecMPI can be passed to this routine
2438        by MatDiagonalScale_MPIAIJ */
2439     PetscCall(VecGetLocalSize(ll, &m));
2440     PetscCheck(m == A->rmap->n, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Left scaling vector wrong length");
2441     PetscCall(VecGetArrayRead(ll, &l));
2442     PetscCall(MatSeqAIJGetArray(A, &v));
2443     for (i = 0; i < m; i++) {
2444       x = l[i];
2445       M = a->i[i + 1] - a->i[i];
2446       for (j = 0; j < M; j++) (*v++) *= x;
2447     }
2448     PetscCall(VecRestoreArrayRead(ll, &l));
2449     PetscCall(PetscLogFlops(nz));
2450     PetscCall(MatSeqAIJRestoreArray(A, &v));
2451   }
2452   if (rr) {
2453     PetscCall(VecGetLocalSize(rr, &n));
2454     PetscCheck(n == A->cmap->n, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Right scaling vector wrong length");
2455     PetscCall(VecGetArrayRead(rr, &r));
2456     PetscCall(MatSeqAIJGetArray(A, &v));
2457     jj = a->j;
2458     for (i = 0; i < nz; i++) (*v++) *= r[*jj++];
2459     PetscCall(MatSeqAIJRestoreArray(A, &v));
2460     PetscCall(VecRestoreArrayRead(rr, &r));
2461     PetscCall(PetscLogFlops(nz));
2462   }
2463   PetscCall(MatSeqAIJInvalidateDiagonal(A));
2464   PetscFunctionReturn(PETSC_SUCCESS);
2465 }
2466 
2467 PetscErrorCode MatCreateSubMatrix_SeqAIJ(Mat A, IS isrow, IS iscol, PetscInt csize, MatReuse scall, Mat *B)
2468 {
2469   Mat_SeqAIJ        *a = (Mat_SeqAIJ *)A->data, *c;
2470   PetscInt          *smap, i, k, kstart, kend, oldcols = A->cmap->n, *lens;
2471   PetscInt           row, mat_i, *mat_j, tcol, first, step, *mat_ilen, sum, lensi;
2472   const PetscInt    *irow, *icol;
2473   const PetscScalar *aa;
2474   PetscInt           nrows, ncols;
2475   PetscInt          *starts, *j_new, *i_new, *aj = a->j, *ai = a->i, ii, *ailen = a->ilen;
2476   MatScalar         *a_new, *mat_a, *c_a;
2477   Mat                C;
2478   PetscBool          stride;
2479 
2480   PetscFunctionBegin;
2481   PetscCall(ISGetIndices(isrow, &irow));
2482   PetscCall(ISGetLocalSize(isrow, &nrows));
2483   PetscCall(ISGetLocalSize(iscol, &ncols));
2484 
2485   PetscCall(PetscObjectTypeCompare((PetscObject)iscol, ISSTRIDE, &stride));
2486   if (stride) {
2487     PetscCall(ISStrideGetInfo(iscol, &first, &step));
2488   } else {
2489     first = 0;
2490     step  = 0;
2491   }
2492   if (stride && step == 1) {
2493     /* special case of contiguous rows */
2494     PetscCall(PetscMalloc2(nrows, &lens, nrows, &starts));
2495     /* loop over new rows determining lens and starting points */
2496     for (i = 0; i < nrows; i++) {
2497       kstart    = ai[irow[i]];
2498       kend      = kstart + ailen[irow[i]];
2499       starts[i] = kstart;
2500       for (k = kstart; k < kend; k++) {
2501         if (aj[k] >= first) {
2502           starts[i] = k;
2503           break;
2504         }
2505       }
2506       sum = 0;
2507       while (k < kend) {
2508         if (aj[k++] >= first + ncols) break;
2509         sum++;
2510       }
2511       lens[i] = sum;
2512     }
2513     /* create submatrix */
2514     if (scall == MAT_REUSE_MATRIX) {
2515       PetscInt n_cols, n_rows;
2516       PetscCall(MatGetSize(*B, &n_rows, &n_cols));
2517       PetscCheck(n_rows == nrows && n_cols == ncols, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Reused submatrix wrong size");
2518       PetscCall(MatZeroEntries(*B));
2519       C = *B;
2520     } else {
2521       PetscInt rbs, cbs;
2522       PetscCall(MatCreate(PetscObjectComm((PetscObject)A), &C));
2523       PetscCall(MatSetSizes(C, nrows, ncols, PETSC_DETERMINE, PETSC_DETERMINE));
2524       PetscCall(ISGetBlockSize(isrow, &rbs));
2525       PetscCall(ISGetBlockSize(iscol, &cbs));
2526       PetscCall(MatSetBlockSizes(C, rbs, cbs));
2527       PetscCall(MatSetType(C, ((PetscObject)A)->type_name));
2528       PetscCall(MatSeqAIJSetPreallocation_SeqAIJ(C, 0, lens));
2529     }
2530     c = (Mat_SeqAIJ *)C->data;
2531 
2532     /* loop over rows inserting into submatrix */
2533     PetscCall(MatSeqAIJGetArrayWrite(C, &a_new)); // Not 'a_new = c->a-new', since that raw usage ignores offload state of C
2534     j_new = c->j;
2535     i_new = c->i;
2536     PetscCall(MatSeqAIJGetArrayRead(A, &aa));
2537     for (i = 0; i < nrows; i++) {
2538       ii    = starts[i];
2539       lensi = lens[i];
2540       if (lensi) {
2541         for (k = 0; k < lensi; k++) *j_new++ = aj[ii + k] - first;
2542         PetscCall(PetscArraycpy(a_new, aa + starts[i], lensi));
2543         a_new += lensi;
2544       }
2545       i_new[i + 1] = i_new[i] + lensi;
2546       c->ilen[i]   = lensi;
2547     }
2548     PetscCall(MatSeqAIJRestoreArrayWrite(C, &a_new)); // Set C's offload state properly
2549     PetscCall(MatSeqAIJRestoreArrayRead(A, &aa));
2550     PetscCall(PetscFree2(lens, starts));
2551   } else {
2552     PetscCall(ISGetIndices(iscol, &icol));
2553     PetscCall(PetscCalloc1(oldcols, &smap));
2554     PetscCall(PetscMalloc1(1 + nrows, &lens));
2555     for (i = 0; i < ncols; i++) {
2556       PetscCheck(icol[i] < oldcols, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Requesting column beyond largest column icol[%" PetscInt_FMT "] %" PetscInt_FMT " >= A->cmap->n %" PetscInt_FMT, i, icol[i], oldcols);
2557       smap[icol[i]] = i + 1;
2558     }
2559 
2560     /* determine lens of each row */
2561     for (i = 0; i < nrows; i++) {
2562       kstart  = ai[irow[i]];
2563       kend    = kstart + a->ilen[irow[i]];
2564       lens[i] = 0;
2565       for (k = kstart; k < kend; k++) {
2566         if (smap[aj[k]]) lens[i]++;
2567       }
2568     }
2569     /* Create and fill new matrix */
2570     if (scall == MAT_REUSE_MATRIX) {
2571       PetscBool equal;
2572 
2573       c = (Mat_SeqAIJ *)((*B)->data);
2574       PetscCheck((*B)->rmap->n == nrows && (*B)->cmap->n == ncols, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Cannot reuse matrix. wrong size");
2575       PetscCall(PetscArraycmp(c->ilen, lens, (*B)->rmap->n, &equal));
2576       PetscCheck(equal, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Cannot reuse matrix. wrong number of nonzeros");
2577       PetscCall(PetscArrayzero(c->ilen, (*B)->rmap->n));
2578       C = *B;
2579     } else {
2580       PetscInt rbs, cbs;
2581       PetscCall(MatCreate(PetscObjectComm((PetscObject)A), &C));
2582       PetscCall(MatSetSizes(C, nrows, ncols, PETSC_DETERMINE, PETSC_DETERMINE));
2583       PetscCall(ISGetBlockSize(isrow, &rbs));
2584       PetscCall(ISGetBlockSize(iscol, &cbs));
2585       if (rbs > 1 || cbs > 1) PetscCall(MatSetBlockSizes(C, rbs, cbs));
2586       PetscCall(MatSetType(C, ((PetscObject)A)->type_name));
2587       PetscCall(MatSeqAIJSetPreallocation_SeqAIJ(C, 0, lens));
2588     }
2589     PetscCall(MatSeqAIJGetArrayRead(A, &aa));
2590 
2591     c = (Mat_SeqAIJ *)C->data;
2592     PetscCall(MatSeqAIJGetArrayWrite(C, &c_a)); // Not 'c->a', since that raw usage ignores offload state of C
2593     for (i = 0; i < nrows; i++) {
2594       row      = irow[i];
2595       kstart   = ai[row];
2596       kend     = kstart + a->ilen[row];
2597       mat_i    = c->i[i];
2598       mat_j    = PetscSafePointerPlusOffset(c->j, mat_i);
2599       mat_a    = PetscSafePointerPlusOffset(c_a, mat_i);
2600       mat_ilen = c->ilen + i;
2601       for (k = kstart; k < kend; k++) {
2602         if ((tcol = smap[a->j[k]])) {
2603           *mat_j++ = tcol - 1;
2604           *mat_a++ = aa[k];
2605           (*mat_ilen)++;
2606         }
2607       }
2608     }
2609     PetscCall(MatSeqAIJRestoreArrayRead(A, &aa));
2610     /* Free work space */
2611     PetscCall(ISRestoreIndices(iscol, &icol));
2612     PetscCall(PetscFree(smap));
2613     PetscCall(PetscFree(lens));
2614     /* sort */
2615     for (i = 0; i < nrows; i++) {
2616       PetscInt ilen;
2617 
2618       mat_i = c->i[i];
2619       mat_j = PetscSafePointerPlusOffset(c->j, mat_i);
2620       mat_a = PetscSafePointerPlusOffset(c_a, mat_i);
2621       ilen  = c->ilen[i];
2622       PetscCall(PetscSortIntWithScalarArray(ilen, mat_j, mat_a));
2623     }
2624     PetscCall(MatSeqAIJRestoreArrayWrite(C, &c_a));
2625   }
2626 #if defined(PETSC_HAVE_DEVICE)
2627   PetscCall(MatBindToCPU(C, A->boundtocpu));
2628 #endif
2629   PetscCall(MatAssemblyBegin(C, MAT_FINAL_ASSEMBLY));
2630   PetscCall(MatAssemblyEnd(C, MAT_FINAL_ASSEMBLY));
2631 
2632   PetscCall(ISRestoreIndices(isrow, &irow));
2633   *B = C;
2634   PetscFunctionReturn(PETSC_SUCCESS);
2635 }
2636 
2637 static PetscErrorCode MatGetMultiProcBlock_SeqAIJ(Mat mat, MPI_Comm subComm, MatReuse scall, Mat *subMat)
2638 {
2639   Mat B;
2640 
2641   PetscFunctionBegin;
2642   if (scall == MAT_INITIAL_MATRIX) {
2643     PetscCall(MatCreate(subComm, &B));
2644     PetscCall(MatSetSizes(B, mat->rmap->n, mat->cmap->n, mat->rmap->n, mat->cmap->n));
2645     PetscCall(MatSetBlockSizesFromMats(B, mat, mat));
2646     PetscCall(MatSetType(B, MATSEQAIJ));
2647     PetscCall(MatDuplicateNoCreate_SeqAIJ(B, mat, MAT_COPY_VALUES, PETSC_TRUE));
2648     *subMat = B;
2649   } else {
2650     PetscCall(MatCopy_SeqAIJ(mat, *subMat, SAME_NONZERO_PATTERN));
2651   }
2652   PetscFunctionReturn(PETSC_SUCCESS);
2653 }
2654 
2655 static PetscErrorCode MatILUFactor_SeqAIJ(Mat inA, IS row, IS col, const MatFactorInfo *info)
2656 {
2657   Mat_SeqAIJ *a = (Mat_SeqAIJ *)inA->data;
2658   Mat         outA;
2659   PetscBool   row_identity, col_identity;
2660 
2661   PetscFunctionBegin;
2662   PetscCheck(info->levels == 0, PETSC_COMM_SELF, PETSC_ERR_SUP, "Only levels=0 supported for in-place ilu");
2663 
2664   PetscCall(ISIdentity(row, &row_identity));
2665   PetscCall(ISIdentity(col, &col_identity));
2666 
2667   outA             = inA;
2668   outA->factortype = MAT_FACTOR_LU;
2669   PetscCall(PetscFree(inA->solvertype));
2670   PetscCall(PetscStrallocpy(MATSOLVERPETSC, &inA->solvertype));
2671 
2672   PetscCall(PetscObjectReference((PetscObject)row));
2673   PetscCall(ISDestroy(&a->row));
2674 
2675   a->row = row;
2676 
2677   PetscCall(PetscObjectReference((PetscObject)col));
2678   PetscCall(ISDestroy(&a->col));
2679 
2680   a->col = col;
2681 
2682   /* Create the inverse permutation so that it can be used in MatLUFactorNumeric() */
2683   PetscCall(ISDestroy(&a->icol));
2684   PetscCall(ISInvertPermutation(col, PETSC_DECIDE, &a->icol));
2685 
2686   if (!a->solve_work) { /* this matrix may have been factored before */
2687     PetscCall(PetscMalloc1(inA->rmap->n + 1, &a->solve_work));
2688   }
2689 
2690   PetscCall(MatMarkDiagonal_SeqAIJ(inA));
2691   if (row_identity && col_identity) {
2692     PetscCall(MatLUFactorNumeric_SeqAIJ_inplace(outA, inA, info));
2693   } else {
2694     PetscCall(MatLUFactorNumeric_SeqAIJ_InplaceWithPerm(outA, inA, info));
2695   }
2696   PetscFunctionReturn(PETSC_SUCCESS);
2697 }
2698 
2699 PetscErrorCode MatScale_SeqAIJ(Mat inA, PetscScalar alpha)
2700 {
2701   Mat_SeqAIJ  *a = (Mat_SeqAIJ *)inA->data;
2702   PetscScalar *v;
2703   PetscBLASInt one = 1, bnz;
2704 
2705   PetscFunctionBegin;
2706   PetscCall(MatSeqAIJGetArray(inA, &v));
2707   PetscCall(PetscBLASIntCast(a->nz, &bnz));
2708   PetscCallBLAS("BLASscal", BLASscal_(&bnz, &alpha, v, &one));
2709   PetscCall(PetscLogFlops(a->nz));
2710   PetscCall(MatSeqAIJRestoreArray(inA, &v));
2711   PetscCall(MatSeqAIJInvalidateDiagonal(inA));
2712   PetscFunctionReturn(PETSC_SUCCESS);
2713 }
2714 
2715 PetscErrorCode MatDestroySubMatrix_Private(Mat_SubSppt *submatj)
2716 {
2717   PetscInt i;
2718 
2719   PetscFunctionBegin;
2720   if (!submatj->id) { /* delete data that are linked only to submats[id=0] */
2721     PetscCall(PetscFree4(submatj->sbuf1, submatj->ptr, submatj->tmp, submatj->ctr));
2722 
2723     for (i = 0; i < submatj->nrqr; ++i) PetscCall(PetscFree(submatj->sbuf2[i]));
2724     PetscCall(PetscFree3(submatj->sbuf2, submatj->req_size, submatj->req_source1));
2725 
2726     if (submatj->rbuf1) {
2727       PetscCall(PetscFree(submatj->rbuf1[0]));
2728       PetscCall(PetscFree(submatj->rbuf1));
2729     }
2730 
2731     for (i = 0; i < submatj->nrqs; ++i) PetscCall(PetscFree(submatj->rbuf3[i]));
2732     PetscCall(PetscFree3(submatj->req_source2, submatj->rbuf2, submatj->rbuf3));
2733     PetscCall(PetscFree(submatj->pa));
2734   }
2735 
2736 #if defined(PETSC_USE_CTABLE)
2737   PetscCall(PetscHMapIDestroy(&submatj->rmap));
2738   if (submatj->cmap_loc) PetscCall(PetscFree(submatj->cmap_loc));
2739   PetscCall(PetscFree(submatj->rmap_loc));
2740 #else
2741   PetscCall(PetscFree(submatj->rmap));
2742 #endif
2743 
2744   if (!submatj->allcolumns) {
2745 #if defined(PETSC_USE_CTABLE)
2746     PetscCall(PetscHMapIDestroy((PetscHMapI *)&submatj->cmap));
2747 #else
2748     PetscCall(PetscFree(submatj->cmap));
2749 #endif
2750   }
2751   PetscCall(PetscFree(submatj->row2proc));
2752 
2753   PetscCall(PetscFree(submatj));
2754   PetscFunctionReturn(PETSC_SUCCESS);
2755 }
2756 
2757 PetscErrorCode MatDestroySubMatrix_SeqAIJ(Mat C)
2758 {
2759   Mat_SeqAIJ  *c       = (Mat_SeqAIJ *)C->data;
2760   Mat_SubSppt *submatj = c->submatis1;
2761 
2762   PetscFunctionBegin;
2763   PetscCall((*submatj->destroy)(C));
2764   PetscCall(MatDestroySubMatrix_Private(submatj));
2765   PetscFunctionReturn(PETSC_SUCCESS);
2766 }
2767 
2768 /* Note this has code duplication with MatDestroySubMatrices_SeqBAIJ() */
2769 static PetscErrorCode MatDestroySubMatrices_SeqAIJ(PetscInt n, Mat *mat[])
2770 {
2771   PetscInt     i;
2772   Mat          C;
2773   Mat_SeqAIJ  *c;
2774   Mat_SubSppt *submatj;
2775 
2776   PetscFunctionBegin;
2777   for (i = 0; i < n; i++) {
2778     C       = (*mat)[i];
2779     c       = (Mat_SeqAIJ *)C->data;
2780     submatj = c->submatis1;
2781     if (submatj) {
2782       if (--((PetscObject)C)->refct <= 0) {
2783         PetscCall(PetscFree(C->factorprefix));
2784         PetscCall((*submatj->destroy)(C));
2785         PetscCall(MatDestroySubMatrix_Private(submatj));
2786         PetscCall(PetscFree(C->defaultvectype));
2787         PetscCall(PetscFree(C->defaultrandtype));
2788         PetscCall(PetscLayoutDestroy(&C->rmap));
2789         PetscCall(PetscLayoutDestroy(&C->cmap));
2790         PetscCall(PetscHeaderDestroy(&C));
2791       }
2792     } else {
2793       PetscCall(MatDestroy(&C));
2794     }
2795   }
2796 
2797   /* Destroy Dummy submatrices created for reuse */
2798   PetscCall(MatDestroySubMatrices_Dummy(n, mat));
2799 
2800   PetscCall(PetscFree(*mat));
2801   PetscFunctionReturn(PETSC_SUCCESS);
2802 }
2803 
2804 static PetscErrorCode MatCreateSubMatrices_SeqAIJ(Mat A, PetscInt n, const IS irow[], const IS icol[], MatReuse scall, Mat *B[])
2805 {
2806   PetscInt i;
2807 
2808   PetscFunctionBegin;
2809   if (scall == MAT_INITIAL_MATRIX) PetscCall(PetscCalloc1(n + 1, B));
2810 
2811   for (i = 0; i < n; i++) PetscCall(MatCreateSubMatrix_SeqAIJ(A, irow[i], icol[i], PETSC_DECIDE, scall, &(*B)[i]));
2812   PetscFunctionReturn(PETSC_SUCCESS);
2813 }
2814 
2815 static PetscErrorCode MatIncreaseOverlap_SeqAIJ(Mat A, PetscInt is_max, IS is[], PetscInt ov)
2816 {
2817   Mat_SeqAIJ     *a = (Mat_SeqAIJ *)A->data;
2818   PetscInt        row, i, j, k, l, ll, m, n, *nidx, isz, val;
2819   const PetscInt *idx;
2820   PetscInt        start, end, *ai, *aj, bs = (A->rmap->bs > 0 && A->rmap->bs == A->cmap->bs) ? A->rmap->bs : 1;
2821   PetscBT         table;
2822 
2823   PetscFunctionBegin;
2824   m  = A->rmap->n / bs;
2825   ai = a->i;
2826   aj = a->j;
2827 
2828   PetscCheck(ov >= 0, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "illegal negative overlap value used");
2829 
2830   PetscCall(PetscMalloc1(m + 1, &nidx));
2831   PetscCall(PetscBTCreate(m, &table));
2832 
2833   for (i = 0; i < is_max; i++) {
2834     /* Initialize the two local arrays */
2835     isz = 0;
2836     PetscCall(PetscBTMemzero(m, table));
2837 
2838     /* Extract the indices, assume there can be duplicate entries */
2839     PetscCall(ISGetIndices(is[i], &idx));
2840     PetscCall(ISGetLocalSize(is[i], &n));
2841 
2842     if (bs > 1) {
2843       /* Enter these into the temp arrays. I.e., mark table[row], enter row into new index */
2844       for (j = 0; j < n; ++j) {
2845         if (!PetscBTLookupSet(table, idx[j] / bs)) nidx[isz++] = idx[j] / bs;
2846       }
2847       PetscCall(ISRestoreIndices(is[i], &idx));
2848       PetscCall(ISDestroy(&is[i]));
2849 
2850       k = 0;
2851       for (j = 0; j < ov; j++) { /* for each overlap */
2852         n = isz;
2853         for (; k < n; k++) { /* do only those rows in nidx[k], which are not done yet */
2854           for (ll = 0; ll < bs; ll++) {
2855             row   = bs * nidx[k] + ll;
2856             start = ai[row];
2857             end   = ai[row + 1];
2858             for (l = start; l < end; l++) {
2859               val = aj[l] / bs;
2860               if (!PetscBTLookupSet(table, val)) nidx[isz++] = val;
2861             }
2862           }
2863         }
2864       }
2865       PetscCall(ISCreateBlock(PETSC_COMM_SELF, bs, isz, nidx, PETSC_COPY_VALUES, (is + i)));
2866     } else {
2867       /* Enter these into the temp arrays. I.e., mark table[row], enter row into new index */
2868       for (j = 0; j < n; ++j) {
2869         if (!PetscBTLookupSet(table, idx[j])) nidx[isz++] = idx[j];
2870       }
2871       PetscCall(ISRestoreIndices(is[i], &idx));
2872       PetscCall(ISDestroy(&is[i]));
2873 
2874       k = 0;
2875       for (j = 0; j < ov; j++) { /* for each overlap */
2876         n = isz;
2877         for (; k < n; k++) { /* do only those rows in nidx[k], which are not done yet */
2878           row   = nidx[k];
2879           start = ai[row];
2880           end   = ai[row + 1];
2881           for (l = start; l < end; l++) {
2882             val = aj[l];
2883             if (!PetscBTLookupSet(table, val)) nidx[isz++] = val;
2884           }
2885         }
2886       }
2887       PetscCall(ISCreateGeneral(PETSC_COMM_SELF, isz, nidx, PETSC_COPY_VALUES, (is + i)));
2888     }
2889   }
2890   PetscCall(PetscBTDestroy(&table));
2891   PetscCall(PetscFree(nidx));
2892   PetscFunctionReturn(PETSC_SUCCESS);
2893 }
2894 
2895 static PetscErrorCode MatPermute_SeqAIJ(Mat A, IS rowp, IS colp, Mat *B)
2896 {
2897   Mat_SeqAIJ     *a = (Mat_SeqAIJ *)A->data;
2898   PetscInt        i, nz = 0, m = A->rmap->n, n = A->cmap->n;
2899   const PetscInt *row, *col;
2900   PetscInt       *cnew, j, *lens;
2901   IS              icolp, irowp;
2902   PetscInt       *cwork = NULL;
2903   PetscScalar    *vwork = NULL;
2904 
2905   PetscFunctionBegin;
2906   PetscCall(ISInvertPermutation(rowp, PETSC_DECIDE, &irowp));
2907   PetscCall(ISGetIndices(irowp, &row));
2908   PetscCall(ISInvertPermutation(colp, PETSC_DECIDE, &icolp));
2909   PetscCall(ISGetIndices(icolp, &col));
2910 
2911   /* determine lengths of permuted rows */
2912   PetscCall(PetscMalloc1(m + 1, &lens));
2913   for (i = 0; i < m; i++) lens[row[i]] = a->i[i + 1] - a->i[i];
2914   PetscCall(MatCreate(PetscObjectComm((PetscObject)A), B));
2915   PetscCall(MatSetSizes(*B, m, n, m, n));
2916   PetscCall(MatSetBlockSizesFromMats(*B, A, A));
2917   PetscCall(MatSetType(*B, ((PetscObject)A)->type_name));
2918   PetscCall(MatSeqAIJSetPreallocation_SeqAIJ(*B, 0, lens));
2919   PetscCall(PetscFree(lens));
2920 
2921   PetscCall(PetscMalloc1(n, &cnew));
2922   for (i = 0; i < m; i++) {
2923     PetscCall(MatGetRow_SeqAIJ(A, i, &nz, &cwork, &vwork));
2924     for (j = 0; j < nz; j++) cnew[j] = col[cwork[j]];
2925     PetscCall(MatSetValues_SeqAIJ(*B, 1, &row[i], nz, cnew, vwork, INSERT_VALUES));
2926     PetscCall(MatRestoreRow_SeqAIJ(A, i, &nz, &cwork, &vwork));
2927   }
2928   PetscCall(PetscFree(cnew));
2929 
2930   (*B)->assembled = PETSC_FALSE;
2931 
2932 #if defined(PETSC_HAVE_DEVICE)
2933   PetscCall(MatBindToCPU(*B, A->boundtocpu));
2934 #endif
2935   PetscCall(MatAssemblyBegin(*B, MAT_FINAL_ASSEMBLY));
2936   PetscCall(MatAssemblyEnd(*B, MAT_FINAL_ASSEMBLY));
2937   PetscCall(ISRestoreIndices(irowp, &row));
2938   PetscCall(ISRestoreIndices(icolp, &col));
2939   PetscCall(ISDestroy(&irowp));
2940   PetscCall(ISDestroy(&icolp));
2941   if (rowp == colp) PetscCall(MatPropagateSymmetryOptions(A, *B));
2942   PetscFunctionReturn(PETSC_SUCCESS);
2943 }
2944 
2945 PetscErrorCode MatCopy_SeqAIJ(Mat A, Mat B, MatStructure str)
2946 {
2947   PetscFunctionBegin;
2948   /* If the two matrices have the same copy implementation, use fast copy. */
2949   if (str == SAME_NONZERO_PATTERN && (A->ops->copy == B->ops->copy)) {
2950     Mat_SeqAIJ        *a = (Mat_SeqAIJ *)A->data;
2951     Mat_SeqAIJ        *b = (Mat_SeqAIJ *)B->data;
2952     const PetscScalar *aa;
2953 
2954     PetscCall(MatSeqAIJGetArrayRead(A, &aa));
2955     PetscCheck(a->i[A->rmap->n] == b->i[B->rmap->n], PETSC_COMM_SELF, PETSC_ERR_ARG_INCOMP, "Number of nonzeros in two matrices are different %" PetscInt_FMT " != %" PetscInt_FMT, a->i[A->rmap->n], b->i[B->rmap->n]);
2956     PetscCall(PetscArraycpy(b->a, aa, a->i[A->rmap->n]));
2957     PetscCall(PetscObjectStateIncrease((PetscObject)B));
2958     PetscCall(MatSeqAIJRestoreArrayRead(A, &aa));
2959   } else {
2960     PetscCall(MatCopy_Basic(A, B, str));
2961   }
2962   PetscFunctionReturn(PETSC_SUCCESS);
2963 }
2964 
2965 PETSC_INTERN PetscErrorCode MatSeqAIJGetArray_SeqAIJ(Mat A, PetscScalar *array[])
2966 {
2967   Mat_SeqAIJ *a = (Mat_SeqAIJ *)A->data;
2968 
2969   PetscFunctionBegin;
2970   *array = a->a;
2971   PetscFunctionReturn(PETSC_SUCCESS);
2972 }
2973 
2974 PETSC_INTERN PetscErrorCode MatSeqAIJRestoreArray_SeqAIJ(Mat A, PetscScalar *array[])
2975 {
2976   PetscFunctionBegin;
2977   *array = NULL;
2978   PetscFunctionReturn(PETSC_SUCCESS);
2979 }
2980 
2981 /*
2982    Computes the number of nonzeros per row needed for preallocation when X and Y
2983    have different nonzero structure.
2984 */
2985 PetscErrorCode MatAXPYGetPreallocation_SeqX_private(PetscInt m, const PetscInt *xi, const PetscInt *xj, const PetscInt *yi, const PetscInt *yj, PetscInt *nnz)
2986 {
2987   PetscInt i, j, k, nzx, nzy;
2988 
2989   PetscFunctionBegin;
2990   /* Set the number of nonzeros in the new matrix */
2991   for (i = 0; i < m; i++) {
2992     const PetscInt *xjj = PetscSafePointerPlusOffset(xj, xi[i]), *yjj = PetscSafePointerPlusOffset(yj, yi[i]);
2993     nzx    = xi[i + 1] - xi[i];
2994     nzy    = yi[i + 1] - yi[i];
2995     nnz[i] = 0;
2996     for (j = 0, k = 0; j < nzx; j++) {                  /* Point in X */
2997       for (; k < nzy && yjj[k] < xjj[j]; k++) nnz[i]++; /* Catch up to X */
2998       if (k < nzy && yjj[k] == xjj[j]) k++;             /* Skip duplicate */
2999       nnz[i]++;
3000     }
3001     for (; k < nzy; k++) nnz[i]++;
3002   }
3003   PetscFunctionReturn(PETSC_SUCCESS);
3004 }
3005 
3006 PetscErrorCode MatAXPYGetPreallocation_SeqAIJ(Mat Y, Mat X, PetscInt *nnz)
3007 {
3008   PetscInt    m = Y->rmap->N;
3009   Mat_SeqAIJ *x = (Mat_SeqAIJ *)X->data;
3010   Mat_SeqAIJ *y = (Mat_SeqAIJ *)Y->data;
3011 
3012   PetscFunctionBegin;
3013   /* Set the number of nonzeros in the new matrix */
3014   PetscCall(MatAXPYGetPreallocation_SeqX_private(m, x->i, x->j, y->i, y->j, nnz));
3015   PetscFunctionReturn(PETSC_SUCCESS);
3016 }
3017 
3018 PetscErrorCode MatAXPY_SeqAIJ(Mat Y, PetscScalar a, Mat X, MatStructure str)
3019 {
3020   Mat_SeqAIJ *x = (Mat_SeqAIJ *)X->data, *y = (Mat_SeqAIJ *)Y->data;
3021 
3022   PetscFunctionBegin;
3023   if (str == UNKNOWN_NONZERO_PATTERN || (PetscDefined(USE_DEBUG) && str == SAME_NONZERO_PATTERN)) {
3024     PetscBool e = x->nz == y->nz ? PETSC_TRUE : PETSC_FALSE;
3025     if (e) {
3026       PetscCall(PetscArraycmp(x->i, y->i, Y->rmap->n + 1, &e));
3027       if (e) {
3028         PetscCall(PetscArraycmp(x->j, y->j, y->nz, &e));
3029         if (e) str = SAME_NONZERO_PATTERN;
3030       }
3031     }
3032     if (!e) PetscCheck(str != SAME_NONZERO_PATTERN, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONG, "MatStructure is not SAME_NONZERO_PATTERN");
3033   }
3034   if (str == SAME_NONZERO_PATTERN) {
3035     const PetscScalar *xa;
3036     PetscScalar       *ya, alpha = a;
3037     PetscBLASInt       one = 1, bnz;
3038 
3039     PetscCall(PetscBLASIntCast(x->nz, &bnz));
3040     PetscCall(MatSeqAIJGetArray(Y, &ya));
3041     PetscCall(MatSeqAIJGetArrayRead(X, &xa));
3042     PetscCallBLAS("BLASaxpy", BLASaxpy_(&bnz, &alpha, xa, &one, ya, &one));
3043     PetscCall(MatSeqAIJRestoreArrayRead(X, &xa));
3044     PetscCall(MatSeqAIJRestoreArray(Y, &ya));
3045     PetscCall(PetscLogFlops(2.0 * bnz));
3046     PetscCall(MatSeqAIJInvalidateDiagonal(Y));
3047     PetscCall(PetscObjectStateIncrease((PetscObject)Y));
3048   } else if (str == SUBSET_NONZERO_PATTERN) { /* nonzeros of X is a subset of Y's */
3049     PetscCall(MatAXPY_Basic(Y, a, X, str));
3050   } else {
3051     Mat       B;
3052     PetscInt *nnz;
3053     PetscCall(PetscMalloc1(Y->rmap->N, &nnz));
3054     PetscCall(MatCreate(PetscObjectComm((PetscObject)Y), &B));
3055     PetscCall(PetscObjectSetName((PetscObject)B, ((PetscObject)Y)->name));
3056     PetscCall(MatSetLayouts(B, Y->rmap, Y->cmap));
3057     PetscCall(MatSetType(B, ((PetscObject)Y)->type_name));
3058     PetscCall(MatAXPYGetPreallocation_SeqAIJ(Y, X, nnz));
3059     PetscCall(MatSeqAIJSetPreallocation(B, 0, nnz));
3060     PetscCall(MatAXPY_BasicWithPreallocation(B, Y, a, X, str));
3061     PetscCall(MatHeaderMerge(Y, &B));
3062     PetscCall(MatSeqAIJCheckInode(Y));
3063     PetscCall(PetscFree(nnz));
3064   }
3065   PetscFunctionReturn(PETSC_SUCCESS);
3066 }
3067 
3068 PETSC_INTERN PetscErrorCode MatConjugate_SeqAIJ(Mat mat)
3069 {
3070 #if defined(PETSC_USE_COMPLEX)
3071   Mat_SeqAIJ  *aij = (Mat_SeqAIJ *)mat->data;
3072   PetscInt     i, nz;
3073   PetscScalar *a;
3074 
3075   PetscFunctionBegin;
3076   nz = aij->nz;
3077   PetscCall(MatSeqAIJGetArray(mat, &a));
3078   for (i = 0; i < nz; i++) a[i] = PetscConj(a[i]);
3079   PetscCall(MatSeqAIJRestoreArray(mat, &a));
3080 #else
3081   PetscFunctionBegin;
3082 #endif
3083   PetscFunctionReturn(PETSC_SUCCESS);
3084 }
3085 
3086 static PetscErrorCode MatGetRowMaxAbs_SeqAIJ(Mat A, Vec v, PetscInt idx[])
3087 {
3088   Mat_SeqAIJ      *a = (Mat_SeqAIJ *)A->data;
3089   PetscInt         i, j, m = A->rmap->n, *ai, *aj, ncols, n;
3090   PetscReal        atmp;
3091   PetscScalar     *x;
3092   const MatScalar *aa, *av;
3093 
3094   PetscFunctionBegin;
3095   PetscCheck(!A->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
3096   PetscCall(MatSeqAIJGetArrayRead(A, &av));
3097   aa = av;
3098   ai = a->i;
3099   aj = a->j;
3100 
3101   PetscCall(VecSet(v, 0.0));
3102   PetscCall(VecGetArrayWrite(v, &x));
3103   PetscCall(VecGetLocalSize(v, &n));
3104   PetscCheck(n == A->rmap->n, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Nonconforming matrix and vector");
3105   for (i = 0; i < m; i++) {
3106     ncols = ai[1] - ai[0];
3107     ai++;
3108     for (j = 0; j < ncols; j++) {
3109       atmp = PetscAbsScalar(*aa);
3110       if (PetscAbsScalar(x[i]) < atmp) {
3111         x[i] = atmp;
3112         if (idx) idx[i] = *aj;
3113       }
3114       aa++;
3115       aj++;
3116     }
3117   }
3118   PetscCall(VecRestoreArrayWrite(v, &x));
3119   PetscCall(MatSeqAIJRestoreArrayRead(A, &av));
3120   PetscFunctionReturn(PETSC_SUCCESS);
3121 }
3122 
3123 static PetscErrorCode MatGetRowSumAbs_SeqAIJ(Mat A, Vec v)
3124 {
3125   Mat_SeqAIJ      *a = (Mat_SeqAIJ *)A->data;
3126   PetscInt         i, j, m = A->rmap->n, *ai, ncols, n;
3127   PetscScalar     *x;
3128   const MatScalar *aa, *av;
3129 
3130   PetscFunctionBegin;
3131   PetscCheck(!A->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
3132   PetscCall(MatSeqAIJGetArrayRead(A, &av));
3133   aa = av;
3134   ai = a->i;
3135 
3136   PetscCall(VecSet(v, 0.0));
3137   PetscCall(VecGetArrayWrite(v, &x));
3138   PetscCall(VecGetLocalSize(v, &n));
3139   PetscCheck(n == A->rmap->n, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Nonconforming matrix and vector");
3140   for (i = 0; i < m; i++) {
3141     ncols = ai[1] - ai[0];
3142     ai++;
3143     for (j = 0; j < ncols; j++) {
3144       x[i] += PetscAbsScalar(*aa);
3145       aa++;
3146     }
3147   }
3148   PetscCall(VecRestoreArrayWrite(v, &x));
3149   PetscCall(MatSeqAIJRestoreArrayRead(A, &av));
3150   PetscFunctionReturn(PETSC_SUCCESS);
3151 }
3152 
3153 static PetscErrorCode MatGetRowMax_SeqAIJ(Mat A, Vec v, PetscInt idx[])
3154 {
3155   Mat_SeqAIJ      *a = (Mat_SeqAIJ *)A->data;
3156   PetscInt         i, j, m = A->rmap->n, *ai, *aj, ncols, n;
3157   PetscScalar     *x;
3158   const MatScalar *aa, *av;
3159 
3160   PetscFunctionBegin;
3161   PetscCheck(!A->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
3162   PetscCall(MatSeqAIJGetArrayRead(A, &av));
3163   aa = av;
3164   ai = a->i;
3165   aj = a->j;
3166 
3167   PetscCall(VecSet(v, 0.0));
3168   PetscCall(VecGetArrayWrite(v, &x));
3169   PetscCall(VecGetLocalSize(v, &n));
3170   PetscCheck(n == A->rmap->n, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Nonconforming matrix and vector");
3171   for (i = 0; i < m; i++) {
3172     ncols = ai[1] - ai[0];
3173     ai++;
3174     if (ncols == A->cmap->n) { /* row is dense */
3175       x[i] = *aa;
3176       if (idx) idx[i] = 0;
3177     } else { /* row is sparse so already KNOW maximum is 0.0 or higher */
3178       x[i] = 0.0;
3179       if (idx) {
3180         for (j = 0; j < ncols; j++) { /* find first implicit 0.0 in the row */
3181           if (aj[j] > j) {
3182             idx[i] = j;
3183             break;
3184           }
3185         }
3186         /* in case first implicit 0.0 in the row occurs at ncols-th column */
3187         if (j == ncols && j < A->cmap->n) idx[i] = j;
3188       }
3189     }
3190     for (j = 0; j < ncols; j++) {
3191       if (PetscRealPart(x[i]) < PetscRealPart(*aa)) {
3192         x[i] = *aa;
3193         if (idx) idx[i] = *aj;
3194       }
3195       aa++;
3196       aj++;
3197     }
3198   }
3199   PetscCall(VecRestoreArrayWrite(v, &x));
3200   PetscCall(MatSeqAIJRestoreArrayRead(A, &av));
3201   PetscFunctionReturn(PETSC_SUCCESS);
3202 }
3203 
3204 static PetscErrorCode MatGetRowMinAbs_SeqAIJ(Mat A, Vec v, PetscInt idx[])
3205 {
3206   Mat_SeqAIJ      *a = (Mat_SeqAIJ *)A->data;
3207   PetscInt         i, j, m = A->rmap->n, *ai, *aj, ncols, n;
3208   PetscScalar     *x;
3209   const MatScalar *aa, *av;
3210 
3211   PetscFunctionBegin;
3212   PetscCall(MatSeqAIJGetArrayRead(A, &av));
3213   aa = av;
3214   ai = a->i;
3215   aj = a->j;
3216 
3217   PetscCall(VecSet(v, 0.0));
3218   PetscCall(VecGetArrayWrite(v, &x));
3219   PetscCall(VecGetLocalSize(v, &n));
3220   PetscCheck(n == m, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Nonconforming matrix and vector, %" PetscInt_FMT " vs. %" PetscInt_FMT " rows", m, n);
3221   for (i = 0; i < m; i++) {
3222     ncols = ai[1] - ai[0];
3223     ai++;
3224     if (ncols == A->cmap->n) { /* row is dense */
3225       x[i] = *aa;
3226       if (idx) idx[i] = 0;
3227     } else { /* row is sparse so already KNOW minimum is 0.0 or higher */
3228       x[i] = 0.0;
3229       if (idx) { /* find first implicit 0.0 in the row */
3230         for (j = 0; j < ncols; j++) {
3231           if (aj[j] > j) {
3232             idx[i] = j;
3233             break;
3234           }
3235         }
3236         /* in case first implicit 0.0 in the row occurs at ncols-th column */
3237         if (j == ncols && j < A->cmap->n) idx[i] = j;
3238       }
3239     }
3240     for (j = 0; j < ncols; j++) {
3241       if (PetscAbsScalar(x[i]) > PetscAbsScalar(*aa)) {
3242         x[i] = *aa;
3243         if (idx) idx[i] = *aj;
3244       }
3245       aa++;
3246       aj++;
3247     }
3248   }
3249   PetscCall(VecRestoreArrayWrite(v, &x));
3250   PetscCall(MatSeqAIJRestoreArrayRead(A, &av));
3251   PetscFunctionReturn(PETSC_SUCCESS);
3252 }
3253 
3254 static PetscErrorCode MatGetRowMin_SeqAIJ(Mat A, Vec v, PetscInt idx[])
3255 {
3256   Mat_SeqAIJ      *a = (Mat_SeqAIJ *)A->data;
3257   PetscInt         i, j, m = A->rmap->n, ncols, n;
3258   const PetscInt  *ai, *aj;
3259   PetscScalar     *x;
3260   const MatScalar *aa, *av;
3261 
3262   PetscFunctionBegin;
3263   PetscCheck(!A->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
3264   PetscCall(MatSeqAIJGetArrayRead(A, &av));
3265   aa = av;
3266   ai = a->i;
3267   aj = a->j;
3268 
3269   PetscCall(VecSet(v, 0.0));
3270   PetscCall(VecGetArrayWrite(v, &x));
3271   PetscCall(VecGetLocalSize(v, &n));
3272   PetscCheck(n == m, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "Nonconforming matrix and vector");
3273   for (i = 0; i < m; i++) {
3274     ncols = ai[1] - ai[0];
3275     ai++;
3276     if (ncols == A->cmap->n) { /* row is dense */
3277       x[i] = *aa;
3278       if (idx) idx[i] = 0;
3279     } else { /* row is sparse so already KNOW minimum is 0.0 or lower */
3280       x[i] = 0.0;
3281       if (idx) { /* find first implicit 0.0 in the row */
3282         for (j = 0; j < ncols; j++) {
3283           if (aj[j] > j) {
3284             idx[i] = j;
3285             break;
3286           }
3287         }
3288         /* in case first implicit 0.0 in the row occurs at ncols-th column */
3289         if (j == ncols && j < A->cmap->n) idx[i] = j;
3290       }
3291     }
3292     for (j = 0; j < ncols; j++) {
3293       if (PetscRealPart(x[i]) > PetscRealPart(*aa)) {
3294         x[i] = *aa;
3295         if (idx) idx[i] = *aj;
3296       }
3297       aa++;
3298       aj++;
3299     }
3300   }
3301   PetscCall(VecRestoreArrayWrite(v, &x));
3302   PetscCall(MatSeqAIJRestoreArrayRead(A, &av));
3303   PetscFunctionReturn(PETSC_SUCCESS);
3304 }
3305 
3306 static PetscErrorCode MatInvertBlockDiagonal_SeqAIJ(Mat A, const PetscScalar **values)
3307 {
3308   Mat_SeqAIJ     *a = (Mat_SeqAIJ *)A->data;
3309   PetscInt        i, bs = PetscAbs(A->rmap->bs), mbs = A->rmap->n / bs, ipvt[5], bs2 = bs * bs, *v_pivots, ij[7], *IJ, j;
3310   MatScalar      *diag, work[25], *v_work;
3311   const PetscReal shift = 0.0;
3312   PetscBool       allowzeropivot, zeropivotdetected = PETSC_FALSE;
3313 
3314   PetscFunctionBegin;
3315   allowzeropivot = PetscNot(A->erroriffailure);
3316   if (a->ibdiagvalid) {
3317     if (values) *values = a->ibdiag;
3318     PetscFunctionReturn(PETSC_SUCCESS);
3319   }
3320   PetscCall(MatMarkDiagonal_SeqAIJ(A));
3321   if (!a->ibdiag) { PetscCall(PetscMalloc1(bs2 * mbs, &a->ibdiag)); }
3322   diag = a->ibdiag;
3323   if (values) *values = a->ibdiag;
3324   /* factor and invert each block */
3325   switch (bs) {
3326   case 1:
3327     for (i = 0; i < mbs; i++) {
3328       PetscCall(MatGetValues(A, 1, &i, 1, &i, diag + i));
3329       if (PetscAbsScalar(diag[i] + shift) < PETSC_MACHINE_EPSILON) {
3330         if (allowzeropivot) {
3331           A->factorerrortype             = MAT_FACTOR_NUMERIC_ZEROPIVOT;
3332           A->factorerror_zeropivot_value = PetscAbsScalar(diag[i]);
3333           A->factorerror_zeropivot_row   = i;
3334           PetscCall(PetscInfo(A, "Zero pivot, row %" PetscInt_FMT " pivot %g tolerance %g\n", i, (double)PetscAbsScalar(diag[i]), (double)PETSC_MACHINE_EPSILON));
3335         } else SETERRQ(PETSC_COMM_SELF, PETSC_ERR_MAT_LU_ZRPVT, "Zero pivot, row %" PetscInt_FMT " pivot %g tolerance %g", i, (double)PetscAbsScalar(diag[i]), (double)PETSC_MACHINE_EPSILON);
3336       }
3337       diag[i] = (PetscScalar)1.0 / (diag[i] + shift);
3338     }
3339     break;
3340   case 2:
3341     for (i = 0; i < mbs; i++) {
3342       ij[0] = 2 * i;
3343       ij[1] = 2 * i + 1;
3344       PetscCall(MatGetValues(A, 2, ij, 2, ij, diag));
3345       PetscCall(PetscKernel_A_gets_inverse_A_2(diag, shift, allowzeropivot, &zeropivotdetected));
3346       if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
3347       PetscCall(PetscKernel_A_gets_transpose_A_2(diag));
3348       diag += 4;
3349     }
3350     break;
3351   case 3:
3352     for (i = 0; i < mbs; i++) {
3353       ij[0] = 3 * i;
3354       ij[1] = 3 * i + 1;
3355       ij[2] = 3 * i + 2;
3356       PetscCall(MatGetValues(A, 3, ij, 3, ij, diag));
3357       PetscCall(PetscKernel_A_gets_inverse_A_3(diag, shift, allowzeropivot, &zeropivotdetected));
3358       if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
3359       PetscCall(PetscKernel_A_gets_transpose_A_3(diag));
3360       diag += 9;
3361     }
3362     break;
3363   case 4:
3364     for (i = 0; i < mbs; i++) {
3365       ij[0] = 4 * i;
3366       ij[1] = 4 * i + 1;
3367       ij[2] = 4 * i + 2;
3368       ij[3] = 4 * i + 3;
3369       PetscCall(MatGetValues(A, 4, ij, 4, ij, diag));
3370       PetscCall(PetscKernel_A_gets_inverse_A_4(diag, shift, allowzeropivot, &zeropivotdetected));
3371       if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
3372       PetscCall(PetscKernel_A_gets_transpose_A_4(diag));
3373       diag += 16;
3374     }
3375     break;
3376   case 5:
3377     for (i = 0; i < mbs; i++) {
3378       ij[0] = 5 * i;
3379       ij[1] = 5 * i + 1;
3380       ij[2] = 5 * i + 2;
3381       ij[3] = 5 * i + 3;
3382       ij[4] = 5 * i + 4;
3383       PetscCall(MatGetValues(A, 5, ij, 5, ij, diag));
3384       PetscCall(PetscKernel_A_gets_inverse_A_5(diag, ipvt, work, shift, allowzeropivot, &zeropivotdetected));
3385       if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
3386       PetscCall(PetscKernel_A_gets_transpose_A_5(diag));
3387       diag += 25;
3388     }
3389     break;
3390   case 6:
3391     for (i = 0; i < mbs; i++) {
3392       ij[0] = 6 * i;
3393       ij[1] = 6 * i + 1;
3394       ij[2] = 6 * i + 2;
3395       ij[3] = 6 * i + 3;
3396       ij[4] = 6 * i + 4;
3397       ij[5] = 6 * i + 5;
3398       PetscCall(MatGetValues(A, 6, ij, 6, ij, diag));
3399       PetscCall(PetscKernel_A_gets_inverse_A_6(diag, shift, allowzeropivot, &zeropivotdetected));
3400       if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
3401       PetscCall(PetscKernel_A_gets_transpose_A_6(diag));
3402       diag += 36;
3403     }
3404     break;
3405   case 7:
3406     for (i = 0; i < mbs; i++) {
3407       ij[0] = 7 * i;
3408       ij[1] = 7 * i + 1;
3409       ij[2] = 7 * i + 2;
3410       ij[3] = 7 * i + 3;
3411       ij[4] = 7 * i + 4;
3412       ij[5] = 7 * i + 5;
3413       ij[6] = 7 * i + 6;
3414       PetscCall(MatGetValues(A, 7, ij, 7, ij, diag));
3415       PetscCall(PetscKernel_A_gets_inverse_A_7(diag, shift, allowzeropivot, &zeropivotdetected));
3416       if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
3417       PetscCall(PetscKernel_A_gets_transpose_A_7(diag));
3418       diag += 49;
3419     }
3420     break;
3421   default:
3422     PetscCall(PetscMalloc3(bs, &v_work, bs, &v_pivots, bs, &IJ));
3423     for (i = 0; i < mbs; i++) {
3424       for (j = 0; j < bs; j++) IJ[j] = bs * i + j;
3425       PetscCall(MatGetValues(A, bs, IJ, bs, IJ, diag));
3426       PetscCall(PetscKernel_A_gets_inverse_A(bs, diag, v_pivots, v_work, allowzeropivot, &zeropivotdetected));
3427       if (zeropivotdetected) A->factorerrortype = MAT_FACTOR_NUMERIC_ZEROPIVOT;
3428       PetscCall(PetscKernel_A_gets_transpose_A_N(diag, bs));
3429       diag += bs2;
3430     }
3431     PetscCall(PetscFree3(v_work, v_pivots, IJ));
3432   }
3433   a->ibdiagvalid = PETSC_TRUE;
3434   PetscFunctionReturn(PETSC_SUCCESS);
3435 }
3436 
3437 static PetscErrorCode MatSetRandom_SeqAIJ(Mat x, PetscRandom rctx)
3438 {
3439   Mat_SeqAIJ *aij = (Mat_SeqAIJ *)x->data;
3440   PetscScalar a, *aa;
3441   PetscInt    m, n, i, j, col;
3442 
3443   PetscFunctionBegin;
3444   if (!x->assembled) {
3445     PetscCall(MatGetSize(x, &m, &n));
3446     for (i = 0; i < m; i++) {
3447       for (j = 0; j < aij->imax[i]; j++) {
3448         PetscCall(PetscRandomGetValue(rctx, &a));
3449         col = (PetscInt)(n * PetscRealPart(a));
3450         PetscCall(MatSetValues(x, 1, &i, 1, &col, &a, ADD_VALUES));
3451       }
3452     }
3453   } else {
3454     PetscCall(MatSeqAIJGetArrayWrite(x, &aa));
3455     for (i = 0; i < aij->nz; i++) PetscCall(PetscRandomGetValue(rctx, aa + i));
3456     PetscCall(MatSeqAIJRestoreArrayWrite(x, &aa));
3457   }
3458   PetscCall(MatAssemblyBegin(x, MAT_FINAL_ASSEMBLY));
3459   PetscCall(MatAssemblyEnd(x, MAT_FINAL_ASSEMBLY));
3460   PetscFunctionReturn(PETSC_SUCCESS);
3461 }
3462 
3463 /* Like MatSetRandom_SeqAIJ, but do not set values on columns in range of [low, high) */
3464 PetscErrorCode MatSetRandomSkipColumnRange_SeqAIJ_Private(Mat x, PetscInt low, PetscInt high, PetscRandom rctx)
3465 {
3466   Mat_SeqAIJ *aij = (Mat_SeqAIJ *)x->data;
3467   PetscScalar a;
3468   PetscInt    m, n, i, j, col, nskip;
3469 
3470   PetscFunctionBegin;
3471   nskip = high - low;
3472   PetscCall(MatGetSize(x, &m, &n));
3473   n -= nskip; /* shrink number of columns where nonzeros can be set */
3474   for (i = 0; i < m; i++) {
3475     for (j = 0; j < aij->imax[i]; j++) {
3476       PetscCall(PetscRandomGetValue(rctx, &a));
3477       col = (PetscInt)(n * PetscRealPart(a));
3478       if (col >= low) col += nskip; /* shift col rightward to skip the hole */
3479       PetscCall(MatSetValues(x, 1, &i, 1, &col, &a, ADD_VALUES));
3480     }
3481   }
3482   PetscCall(MatAssemblyBegin(x, MAT_FINAL_ASSEMBLY));
3483   PetscCall(MatAssemblyEnd(x, MAT_FINAL_ASSEMBLY));
3484   PetscFunctionReturn(PETSC_SUCCESS);
3485 }
3486 
3487 static struct _MatOps MatOps_Values = {MatSetValues_SeqAIJ,
3488                                        MatGetRow_SeqAIJ,
3489                                        MatRestoreRow_SeqAIJ,
3490                                        MatMult_SeqAIJ,
3491                                        /*  4*/ MatMultAdd_SeqAIJ,
3492                                        MatMultTranspose_SeqAIJ,
3493                                        MatMultTransposeAdd_SeqAIJ,
3494                                        NULL,
3495                                        NULL,
3496                                        NULL,
3497                                        /* 10*/ NULL,
3498                                        MatLUFactor_SeqAIJ,
3499                                        NULL,
3500                                        MatSOR_SeqAIJ,
3501                                        MatTranspose_SeqAIJ,
3502                                        /*1 5*/ MatGetInfo_SeqAIJ,
3503                                        MatEqual_SeqAIJ,
3504                                        MatGetDiagonal_SeqAIJ,
3505                                        MatDiagonalScale_SeqAIJ,
3506                                        MatNorm_SeqAIJ,
3507                                        /* 20*/ NULL,
3508                                        MatAssemblyEnd_SeqAIJ,
3509                                        MatSetOption_SeqAIJ,
3510                                        MatZeroEntries_SeqAIJ,
3511                                        /* 24*/ MatZeroRows_SeqAIJ,
3512                                        NULL,
3513                                        NULL,
3514                                        NULL,
3515                                        NULL,
3516                                        /* 29*/ MatSetUp_Seq_Hash,
3517                                        NULL,
3518                                        NULL,
3519                                        NULL,
3520                                        NULL,
3521                                        /* 34*/ MatDuplicate_SeqAIJ,
3522                                        NULL,
3523                                        NULL,
3524                                        MatILUFactor_SeqAIJ,
3525                                        NULL,
3526                                        /* 39*/ MatAXPY_SeqAIJ,
3527                                        MatCreateSubMatrices_SeqAIJ,
3528                                        MatIncreaseOverlap_SeqAIJ,
3529                                        MatGetValues_SeqAIJ,
3530                                        MatCopy_SeqAIJ,
3531                                        /* 44*/ MatGetRowMax_SeqAIJ,
3532                                        MatScale_SeqAIJ,
3533                                        MatShift_SeqAIJ,
3534                                        MatDiagonalSet_SeqAIJ,
3535                                        MatZeroRowsColumns_SeqAIJ,
3536                                        /* 49*/ MatSetRandom_SeqAIJ,
3537                                        MatGetRowIJ_SeqAIJ,
3538                                        MatRestoreRowIJ_SeqAIJ,
3539                                        MatGetColumnIJ_SeqAIJ,
3540                                        MatRestoreColumnIJ_SeqAIJ,
3541                                        /* 54*/ MatFDColoringCreate_SeqXAIJ,
3542                                        NULL,
3543                                        NULL,
3544                                        MatPermute_SeqAIJ,
3545                                        NULL,
3546                                        /* 59*/ NULL,
3547                                        MatDestroy_SeqAIJ,
3548                                        MatView_SeqAIJ,
3549                                        NULL,
3550                                        NULL,
3551                                        /* 64*/ NULL,
3552                                        MatMatMatMultNumeric_SeqAIJ_SeqAIJ_SeqAIJ,
3553                                        NULL,
3554                                        NULL,
3555                                        NULL,
3556                                        /* 69*/ MatGetRowMaxAbs_SeqAIJ,
3557                                        MatGetRowMinAbs_SeqAIJ,
3558                                        NULL,
3559                                        NULL,
3560                                        NULL,
3561                                        /* 74*/ NULL,
3562                                        MatFDColoringApply_AIJ,
3563                                        NULL,
3564                                        NULL,
3565                                        NULL,
3566                                        /* 79*/ MatFindZeroDiagonals_SeqAIJ,
3567                                        NULL,
3568                                        NULL,
3569                                        NULL,
3570                                        MatLoad_SeqAIJ,
3571                                        /* 84*/ NULL,
3572                                        NULL,
3573                                        NULL,
3574                                        NULL,
3575                                        NULL,
3576                                        /* 89*/ NULL,
3577                                        NULL,
3578                                        MatMatMultNumeric_SeqAIJ_SeqAIJ,
3579                                        NULL,
3580                                        NULL,
3581                                        /* 94*/ MatPtAPNumeric_SeqAIJ_SeqAIJ_SparseAxpy,
3582                                        NULL,
3583                                        NULL,
3584                                        MatMatTransposeMultNumeric_SeqAIJ_SeqAIJ,
3585                                        NULL,
3586                                        /* 99*/ MatProductSetFromOptions_SeqAIJ,
3587                                        NULL,
3588                                        NULL,
3589                                        MatConjugate_SeqAIJ,
3590                                        NULL,
3591                                        /*104*/ MatSetValuesRow_SeqAIJ,
3592                                        MatRealPart_SeqAIJ,
3593                                        MatImaginaryPart_SeqAIJ,
3594                                        NULL,
3595                                        NULL,
3596                                        /*109*/ MatMatSolve_SeqAIJ,
3597                                        NULL,
3598                                        MatGetRowMin_SeqAIJ,
3599                                        NULL,
3600                                        MatMissingDiagonal_SeqAIJ,
3601                                        /*114*/ NULL,
3602                                        NULL,
3603                                        NULL,
3604                                        NULL,
3605                                        NULL,
3606                                        /*119*/ NULL,
3607                                        NULL,
3608                                        NULL,
3609                                        NULL,
3610                                        MatGetMultiProcBlock_SeqAIJ,
3611                                        /*124*/ MatFindNonzeroRows_SeqAIJ,
3612                                        MatGetColumnReductions_SeqAIJ,
3613                                        MatInvertBlockDiagonal_SeqAIJ,
3614                                        MatInvertVariableBlockDiagonal_SeqAIJ,
3615                                        NULL,
3616                                        /*129*/ NULL,
3617                                        NULL,
3618                                        NULL,
3619                                        MatTransposeMatMultNumeric_SeqAIJ_SeqAIJ,
3620                                        MatTransposeColoringCreate_SeqAIJ,
3621                                        /*134*/ MatTransColoringApplySpToDen_SeqAIJ,
3622                                        MatTransColoringApplyDenToSp_SeqAIJ,
3623                                        NULL,
3624                                        NULL,
3625                                        MatRARtNumeric_SeqAIJ_SeqAIJ,
3626                                        /*139*/ NULL,
3627                                        NULL,
3628                                        NULL,
3629                                        MatFDColoringSetUp_SeqXAIJ,
3630                                        MatFindOffBlockDiagonalEntries_SeqAIJ,
3631                                        MatCreateMPIMatConcatenateSeqMat_SeqAIJ,
3632                                        /*145*/ MatDestroySubMatrices_SeqAIJ,
3633                                        NULL,
3634                                        NULL,
3635                                        MatCreateGraph_Simple_AIJ,
3636                                        NULL,
3637                                        /*150*/ MatTransposeSymbolic_SeqAIJ,
3638                                        MatEliminateZeros_SeqAIJ,
3639                                        MatGetRowSumAbs_SeqAIJ};
3640 
3641 static PetscErrorCode MatSeqAIJSetColumnIndices_SeqAIJ(Mat mat, PetscInt *indices)
3642 {
3643   Mat_SeqAIJ *aij = (Mat_SeqAIJ *)mat->data;
3644   PetscInt    i, nz, n;
3645 
3646   PetscFunctionBegin;
3647   nz = aij->maxnz;
3648   n  = mat->rmap->n;
3649   for (i = 0; i < nz; i++) aij->j[i] = indices[i];
3650   aij->nz = nz;
3651   for (i = 0; i < n; i++) aij->ilen[i] = aij->imax[i];
3652   PetscFunctionReturn(PETSC_SUCCESS);
3653 }
3654 
3655 /*
3656  * Given a sparse matrix with global column indices, compact it by using a local column space.
3657  * The result matrix helps saving memory in other algorithms, such as MatPtAPSymbolic_MPIAIJ_MPIAIJ_scalable()
3658  */
3659 PetscErrorCode MatSeqAIJCompactOutExtraColumns_SeqAIJ(Mat mat, ISLocalToGlobalMapping *mapping)
3660 {
3661   Mat_SeqAIJ   *aij = (Mat_SeqAIJ *)mat->data;
3662   PetscHMapI    gid1_lid1;
3663   PetscHashIter tpos;
3664   PetscInt      gid, lid, i, ec, nz = aij->nz;
3665   PetscInt     *garray, *jj = aij->j;
3666 
3667   PetscFunctionBegin;
3668   PetscValidHeaderSpecific(mat, MAT_CLASSID, 1);
3669   PetscAssertPointer(mapping, 2);
3670   /* use a table */
3671   PetscCall(PetscHMapICreateWithSize(mat->rmap->n, &gid1_lid1));
3672   ec = 0;
3673   for (i = 0; i < nz; i++) {
3674     PetscInt data, gid1 = jj[i] + 1;
3675     PetscCall(PetscHMapIGetWithDefault(gid1_lid1, gid1, 0, &data));
3676     if (!data) {
3677       /* one based table */
3678       PetscCall(PetscHMapISet(gid1_lid1, gid1, ++ec));
3679     }
3680   }
3681   /* form array of columns we need */
3682   PetscCall(PetscMalloc1(ec, &garray));
3683   PetscHashIterBegin(gid1_lid1, tpos);
3684   while (!PetscHashIterAtEnd(gid1_lid1, tpos)) {
3685     PetscHashIterGetKey(gid1_lid1, tpos, gid);
3686     PetscHashIterGetVal(gid1_lid1, tpos, lid);
3687     PetscHashIterNext(gid1_lid1, tpos);
3688     gid--;
3689     lid--;
3690     garray[lid] = gid;
3691   }
3692   PetscCall(PetscSortInt(ec, garray)); /* sort, and rebuild */
3693   PetscCall(PetscHMapIClear(gid1_lid1));
3694   for (i = 0; i < ec; i++) PetscCall(PetscHMapISet(gid1_lid1, garray[i] + 1, i + 1));
3695   /* compact out the extra columns in B */
3696   for (i = 0; i < nz; i++) {
3697     PetscInt gid1 = jj[i] + 1;
3698     PetscCall(PetscHMapIGetWithDefault(gid1_lid1, gid1, 0, &lid));
3699     lid--;
3700     jj[i] = lid;
3701   }
3702   PetscCall(PetscLayoutDestroy(&mat->cmap));
3703   PetscCall(PetscHMapIDestroy(&gid1_lid1));
3704   PetscCall(PetscLayoutCreateFromSizes(PetscObjectComm((PetscObject)mat), ec, ec, 1, &mat->cmap));
3705   PetscCall(ISLocalToGlobalMappingCreate(PETSC_COMM_SELF, mat->cmap->bs, mat->cmap->n, garray, PETSC_OWN_POINTER, mapping));
3706   PetscCall(ISLocalToGlobalMappingSetType(*mapping, ISLOCALTOGLOBALMAPPINGHASH));
3707   PetscFunctionReturn(PETSC_SUCCESS);
3708 }
3709 
3710 /*@
3711   MatSeqAIJSetColumnIndices - Set the column indices for all the rows
3712   in the matrix.
3713 
3714   Input Parameters:
3715 + mat     - the `MATSEQAIJ` matrix
3716 - indices - the column indices
3717 
3718   Level: advanced
3719 
3720   Notes:
3721   This can be called if you have precomputed the nonzero structure of the
3722   matrix and want to provide it to the matrix object to improve the performance
3723   of the `MatSetValues()` operation.
3724 
3725   You MUST have set the correct numbers of nonzeros per row in the call to
3726   `MatCreateSeqAIJ()`, and the columns indices MUST be sorted.
3727 
3728   MUST be called before any calls to `MatSetValues()`
3729 
3730   The indices should start with zero, not one.
3731 
3732 .seealso: [](ch_matrices), `Mat`, `MATSEQAIJ`
3733 @*/
3734 PetscErrorCode MatSeqAIJSetColumnIndices(Mat mat, PetscInt *indices)
3735 {
3736   PetscFunctionBegin;
3737   PetscValidHeaderSpecific(mat, MAT_CLASSID, 1);
3738   PetscAssertPointer(indices, 2);
3739   PetscUseMethod(mat, "MatSeqAIJSetColumnIndices_C", (Mat, PetscInt *), (mat, indices));
3740   PetscFunctionReturn(PETSC_SUCCESS);
3741 }
3742 
3743 static PetscErrorCode MatStoreValues_SeqAIJ(Mat mat)
3744 {
3745   Mat_SeqAIJ *aij = (Mat_SeqAIJ *)mat->data;
3746   size_t      nz  = aij->i[mat->rmap->n];
3747 
3748   PetscFunctionBegin;
3749   PetscCheck(aij->nonew, PETSC_COMM_SELF, PETSC_ERR_ORDER, "Must call MatSetOption(A,MAT_NEW_NONZERO_LOCATIONS,PETSC_FALSE);first");
3750 
3751   /* allocate space for values if not already there */
3752   if (!aij->saved_values) { PetscCall(PetscMalloc1(nz + 1, &aij->saved_values)); }
3753 
3754   /* copy values over */
3755   PetscCall(PetscArraycpy(aij->saved_values, aij->a, nz));
3756   PetscFunctionReturn(PETSC_SUCCESS);
3757 }
3758 
3759 /*@
3760   MatStoreValues - Stashes a copy of the matrix values; this allows reusing of the linear part of a Jacobian, while recomputing only the
3761   nonlinear portion.
3762 
3763   Logically Collect
3764 
3765   Input Parameter:
3766 . mat - the matrix (currently only `MATAIJ` matrices support this option)
3767 
3768   Level: advanced
3769 
3770   Example Usage:
3771 .vb
3772     Using SNES
3773     Create Jacobian matrix
3774     Set linear terms into matrix
3775     Apply boundary conditions to matrix, at this time matrix must have
3776       final nonzero structure (i.e. setting the nonlinear terms and applying
3777       boundary conditions again will not change the nonzero structure
3778     MatSetOption(mat, MAT_NEW_NONZERO_LOCATIONS, PETSC_FALSE);
3779     MatStoreValues(mat);
3780     Call SNESSetJacobian() with matrix
3781     In your Jacobian routine
3782       MatRetrieveValues(mat);
3783       Set nonlinear terms in matrix
3784 
3785     Without `SNESSolve()`, i.e. when you handle nonlinear solve yourself:
3786     // build linear portion of Jacobian
3787     MatSetOption(mat, MAT_NEW_NONZERO_LOCATIONS, PETSC_FALSE);
3788     MatStoreValues(mat);
3789     loop over nonlinear iterations
3790        MatRetrieveValues(mat);
3791        // call MatSetValues(mat,...) to set nonliner portion of Jacobian
3792        // call MatAssemblyBegin/End() on matrix
3793        Solve linear system with Jacobian
3794     endloop
3795 .ve
3796 
3797   Notes:
3798   Matrix must already be assembled before calling this routine
3799   Must set the matrix option `MatSetOption`(mat,`MAT_NEW_NONZERO_LOCATIONS`,`PETSC_FALSE`); before
3800   calling this routine.
3801 
3802   When this is called multiple times it overwrites the previous set of stored values
3803   and does not allocated additional space.
3804 
3805 .seealso: [](ch_matrices), `Mat`, `MatRetrieveValues()`
3806 @*/
3807 PetscErrorCode MatStoreValues(Mat mat)
3808 {
3809   PetscFunctionBegin;
3810   PetscValidHeaderSpecific(mat, MAT_CLASSID, 1);
3811   PetscCheck(mat->assembled, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3812   PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
3813   PetscUseMethod(mat, "MatStoreValues_C", (Mat), (mat));
3814   PetscFunctionReturn(PETSC_SUCCESS);
3815 }
3816 
3817 static PetscErrorCode MatRetrieveValues_SeqAIJ(Mat mat)
3818 {
3819   Mat_SeqAIJ *aij = (Mat_SeqAIJ *)mat->data;
3820   PetscInt    nz  = aij->i[mat->rmap->n];
3821 
3822   PetscFunctionBegin;
3823   PetscCheck(aij->nonew, PETSC_COMM_SELF, PETSC_ERR_ORDER, "Must call MatSetOption(A,MAT_NEW_NONZERO_LOCATIONS,PETSC_FALSE);first");
3824   PetscCheck(aij->saved_values, PETSC_COMM_SELF, PETSC_ERR_ORDER, "Must call MatStoreValues(A);first");
3825   /* copy values over */
3826   PetscCall(PetscArraycpy(aij->a, aij->saved_values, nz));
3827   PetscFunctionReturn(PETSC_SUCCESS);
3828 }
3829 
3830 /*@
3831   MatRetrieveValues - Retrieves the copy of the matrix values that was stored with `MatStoreValues()`
3832 
3833   Logically Collect
3834 
3835   Input Parameter:
3836 . mat - the matrix (currently only `MATAIJ` matrices support this option)
3837 
3838   Level: advanced
3839 
3840 .seealso: [](ch_matrices), `Mat`, `MatStoreValues()`
3841 @*/
3842 PetscErrorCode MatRetrieveValues(Mat mat)
3843 {
3844   PetscFunctionBegin;
3845   PetscValidHeaderSpecific(mat, MAT_CLASSID, 1);
3846   PetscCheck(mat->assembled, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
3847   PetscCheck(!mat->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
3848   PetscUseMethod(mat, "MatRetrieveValues_C", (Mat), (mat));
3849   PetscFunctionReturn(PETSC_SUCCESS);
3850 }
3851 
3852 /*@C
3853   MatCreateSeqAIJ - Creates a sparse matrix in `MATSEQAIJ` (compressed row) format
3854   (the default parallel PETSc format).  For good matrix assembly performance
3855   the user should preallocate the matrix storage by setting the parameter `nz`
3856   (or the array `nnz`).
3857 
3858   Collective
3859 
3860   Input Parameters:
3861 + comm - MPI communicator, set to `PETSC_COMM_SELF`
3862 . m    - number of rows
3863 . n    - number of columns
3864 . nz   - number of nonzeros per row (same for all rows)
3865 - nnz  - array containing the number of nonzeros in the various rows
3866          (possibly different for each row) or NULL
3867 
3868   Output Parameter:
3869 . A - the matrix
3870 
3871   Options Database Keys:
3872 + -mat_no_inode            - Do not use inodes
3873 - -mat_inode_limit <limit> - Sets inode limit (max limit=5)
3874 
3875   Level: intermediate
3876 
3877   Notes:
3878   It is recommend to use `MatCreateFromOptions()` instead of this routine
3879 
3880   If `nnz` is given then `nz` is ignored
3881 
3882   The `MATSEQAIJ` format, also called
3883   compressed row storage, is fully compatible with standard Fortran
3884   storage.  That is, the stored row and column indices can begin at
3885   either one (as in Fortran) or zero.
3886 
3887   Specify the preallocated storage with either `nz` or `nnz` (not both).
3888   Set `nz` = `PETSC_DEFAULT` and `nnz` = `NULL` for PETSc to control dynamic memory
3889   allocation.
3890 
3891   By default, this format uses inodes (identical nodes) when possible, to
3892   improve numerical efficiency of matrix-vector products and solves. We
3893   search for consecutive rows with the same nonzero structure, thereby
3894   reusing matrix information to achieve increased efficiency.
3895 
3896 .seealso: [](ch_matrices), `Mat`, [Sparse Matrix Creation](sec_matsparse), `MatCreate()`, `MatCreateAIJ()`, `MatSetValues()`, `MatSeqAIJSetColumnIndices()`, `MatCreateSeqAIJWithArrays()`
3897 @*/
3898 PetscErrorCode MatCreateSeqAIJ(MPI_Comm comm, PetscInt m, PetscInt n, PetscInt nz, const PetscInt nnz[], Mat *A)
3899 {
3900   PetscFunctionBegin;
3901   PetscCall(MatCreate(comm, A));
3902   PetscCall(MatSetSizes(*A, m, n, m, n));
3903   PetscCall(MatSetType(*A, MATSEQAIJ));
3904   PetscCall(MatSeqAIJSetPreallocation_SeqAIJ(*A, nz, nnz));
3905   PetscFunctionReturn(PETSC_SUCCESS);
3906 }
3907 
3908 /*@C
3909   MatSeqAIJSetPreallocation - For good matrix assembly performance
3910   the user should preallocate the matrix storage by setting the parameter nz
3911   (or the array nnz).  By setting these parameters accurately, performance
3912   during matrix assembly can be increased by more than a factor of 50.
3913 
3914   Collective
3915 
3916   Input Parameters:
3917 + B   - The matrix
3918 . nz  - number of nonzeros per row (same for all rows)
3919 - nnz - array containing the number of nonzeros in the various rows
3920          (possibly different for each row) or NULL
3921 
3922   Options Database Keys:
3923 + -mat_no_inode            - Do not use inodes
3924 - -mat_inode_limit <limit> - Sets inode limit (max limit=5)
3925 
3926   Level: intermediate
3927 
3928   Notes:
3929   If `nnz` is given then `nz` is ignored
3930 
3931   The `MATSEQAIJ` format also called
3932   compressed row storage, is fully compatible with standard Fortran
3933   storage.  That is, the stored row and column indices can begin at
3934   either one (as in Fortran) or zero.  See the users' manual for details.
3935 
3936   Specify the preallocated storage with either `nz` or `nnz` (not both).
3937   Set nz = `PETSC_DEFAULT` and `nnz` = `NULL` for PETSc to control dynamic memory
3938   allocation.
3939 
3940   You can call `MatGetInfo()` to get information on how effective the preallocation was;
3941   for example the fields mallocs,nz_allocated,nz_used,nz_unneeded;
3942   You can also run with the option -info and look for messages with the string
3943   malloc in them to see if additional memory allocation was needed.
3944 
3945   Developer Notes:
3946   Use nz of `MAT_SKIP_ALLOCATION` to not allocate any space for the matrix
3947   entries or columns indices
3948 
3949   By default, this format uses inodes (identical nodes) when possible, to
3950   improve numerical efficiency of matrix-vector products and solves. We
3951   search for consecutive rows with the same nonzero structure, thereby
3952   reusing matrix information to achieve increased efficiency.
3953 
3954 .seealso: [](ch_matrices), `Mat`, `MatCreate()`, `MatCreateAIJ()`, `MatSetValues()`, `MatSeqAIJSetColumnIndices()`, `MatCreateSeqAIJWithArrays()`, `MatGetInfo()`,
3955           `MatSeqAIJSetTotalPreallocation()`
3956 @*/
3957 PetscErrorCode MatSeqAIJSetPreallocation(Mat B, PetscInt nz, const PetscInt nnz[])
3958 {
3959   PetscFunctionBegin;
3960   PetscValidHeaderSpecific(B, MAT_CLASSID, 1);
3961   PetscValidType(B, 1);
3962   PetscTryMethod(B, "MatSeqAIJSetPreallocation_C", (Mat, PetscInt, const PetscInt[]), (B, nz, nnz));
3963   PetscFunctionReturn(PETSC_SUCCESS);
3964 }
3965 
3966 PetscErrorCode MatSeqAIJSetPreallocation_SeqAIJ(Mat B, PetscInt nz, const PetscInt *nnz)
3967 {
3968   Mat_SeqAIJ *b              = (Mat_SeqAIJ *)B->data;
3969   PetscBool   skipallocation = PETSC_FALSE, realalloc = PETSC_FALSE;
3970   PetscInt    i;
3971 
3972   PetscFunctionBegin;
3973   if (B->hash_active) {
3974     B->ops[0] = b->cops;
3975     PetscCall(PetscHMapIJVDestroy(&b->ht));
3976     PetscCall(PetscFree(b->dnz));
3977     B->hash_active = PETSC_FALSE;
3978   }
3979   if (nz >= 0 || nnz) realalloc = PETSC_TRUE;
3980   if (nz == MAT_SKIP_ALLOCATION) {
3981     skipallocation = PETSC_TRUE;
3982     nz             = 0;
3983   }
3984   PetscCall(PetscLayoutSetUp(B->rmap));
3985   PetscCall(PetscLayoutSetUp(B->cmap));
3986 
3987   if (nz == PETSC_DEFAULT || nz == PETSC_DECIDE) nz = 5;
3988   PetscCheck(nz >= 0, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "nz cannot be less than 0: value %" PetscInt_FMT, nz);
3989   if (PetscUnlikelyDebug(nnz)) {
3990     for (i = 0; i < B->rmap->n; i++) {
3991       PetscCheck(nnz[i] >= 0, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "nnz cannot be less than 0: local row %" PetscInt_FMT " value %" PetscInt_FMT, i, nnz[i]);
3992       PetscCheck(nnz[i] <= B->cmap->n, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "nnz cannot be greater than row length: local row %" PetscInt_FMT " value %" PetscInt_FMT " rowlength %" PetscInt_FMT, i, nnz[i], B->cmap->n);
3993     }
3994   }
3995 
3996   B->preallocated = PETSC_TRUE;
3997   if (!skipallocation) {
3998     if (!b->imax) { PetscCall(PetscMalloc1(B->rmap->n, &b->imax)); }
3999     if (!b->ilen) {
4000       /* b->ilen will count nonzeros in each row so far. */
4001       PetscCall(PetscCalloc1(B->rmap->n, &b->ilen));
4002     } else {
4003       PetscCall(PetscMemzero(b->ilen, B->rmap->n * sizeof(PetscInt)));
4004     }
4005     if (!b->ipre) PetscCall(PetscMalloc1(B->rmap->n, &b->ipre));
4006     if (!nnz) {
4007       if (nz == PETSC_DEFAULT || nz == PETSC_DECIDE) nz = 10;
4008       else if (nz < 0) nz = 1;
4009       nz = PetscMin(nz, B->cmap->n);
4010       for (i = 0; i < B->rmap->n; i++) b->imax[i] = nz;
4011       PetscCall(PetscIntMultError(nz, B->rmap->n, &nz));
4012     } else {
4013       PetscInt64 nz64 = 0;
4014       for (i = 0; i < B->rmap->n; i++) {
4015         b->imax[i] = nnz[i];
4016         nz64 += nnz[i];
4017       }
4018       PetscCall(PetscIntCast(nz64, &nz));
4019     }
4020 
4021     /* allocate the matrix space */
4022     /* FIXME: should B's old memory be unlogged? */
4023     PetscCall(MatSeqXAIJFreeAIJ(B, &b->a, &b->j, &b->i));
4024     if (B->structure_only) {
4025       PetscCall(PetscMalloc1(nz, &b->j));
4026       PetscCall(PetscMalloc1(B->rmap->n + 1, &b->i));
4027     } else {
4028       PetscCall(PetscMalloc3(nz, &b->a, nz, &b->j, B->rmap->n + 1, &b->i));
4029     }
4030     b->i[0] = 0;
4031     for (i = 1; i < B->rmap->n + 1; i++) b->i[i] = b->i[i - 1] + b->imax[i - 1];
4032     if (B->structure_only) {
4033       b->singlemalloc = PETSC_FALSE;
4034       b->free_a       = PETSC_FALSE;
4035     } else {
4036       b->singlemalloc = PETSC_TRUE;
4037       b->free_a       = PETSC_TRUE;
4038     }
4039     b->free_ij = PETSC_TRUE;
4040   } else {
4041     b->free_a  = PETSC_FALSE;
4042     b->free_ij = PETSC_FALSE;
4043   }
4044 
4045   if (b->ipre && nnz != b->ipre && b->imax) {
4046     /* reserve user-requested sparsity */
4047     PetscCall(PetscArraycpy(b->ipre, b->imax, B->rmap->n));
4048   }
4049 
4050   b->nz               = 0;
4051   b->maxnz            = nz;
4052   B->info.nz_unneeded = (double)b->maxnz;
4053   if (realalloc) PetscCall(MatSetOption(B, MAT_NEW_NONZERO_ALLOCATION_ERR, PETSC_TRUE));
4054   B->was_assembled = PETSC_FALSE;
4055   B->assembled     = PETSC_FALSE;
4056   /* We simply deem preallocation has changed nonzero state. Updating the state
4057      will give clients (like AIJKokkos) a chance to know something has happened.
4058   */
4059   B->nonzerostate++;
4060   PetscFunctionReturn(PETSC_SUCCESS);
4061 }
4062 
4063 static PetscErrorCode MatResetPreallocation_SeqAIJ(Mat A)
4064 {
4065   Mat_SeqAIJ *a;
4066   PetscInt    i;
4067   PetscBool   skipreset;
4068 
4069   PetscFunctionBegin;
4070   PetscValidHeaderSpecific(A, MAT_CLASSID, 1);
4071 
4072   /* Check local size. If zero, then return */
4073   if (!A->rmap->n) PetscFunctionReturn(PETSC_SUCCESS);
4074 
4075   a = (Mat_SeqAIJ *)A->data;
4076   /* if no saved info, we error out */
4077   PetscCheck(a->ipre, PETSC_COMM_SELF, PETSC_ERR_ARG_NULL, "No saved preallocation info ");
4078 
4079   PetscCheck(a->i && a->imax && a->ilen, PETSC_COMM_SELF, PETSC_ERR_ARG_NULL, "Memory info is incomplete, and can not reset preallocation ");
4080 
4081   PetscCall(PetscArraycmp(a->ipre, a->ilen, A->rmap->n, &skipreset));
4082   if (!skipreset) {
4083     PetscCall(PetscArraycpy(a->imax, a->ipre, A->rmap->n));
4084     PetscCall(PetscArrayzero(a->ilen, A->rmap->n));
4085     a->i[0] = 0;
4086     for (i = 1; i < A->rmap->n + 1; i++) a->i[i] = a->i[i - 1] + a->imax[i - 1];
4087     A->preallocated     = PETSC_TRUE;
4088     a->nz               = 0;
4089     a->maxnz            = a->i[A->rmap->n];
4090     A->info.nz_unneeded = (double)a->maxnz;
4091     A->was_assembled    = PETSC_FALSE;
4092     A->assembled        = PETSC_FALSE;
4093   }
4094   PetscFunctionReturn(PETSC_SUCCESS);
4095 }
4096 
4097 /*@
4098   MatSeqAIJSetPreallocationCSR - Allocates memory for a sparse sequential matrix in `MATSEQAIJ` format.
4099 
4100   Input Parameters:
4101 + B - the matrix
4102 . i - the indices into `j` for the start of each row (indices start with zero)
4103 . j - the column indices for each row (indices start with zero) these must be sorted for each row
4104 - v - optional values in the matrix, use `NULL` if not provided
4105 
4106   Level: developer
4107 
4108   Notes:
4109   The `i`,`j`,`v` values are COPIED with this routine; to avoid the copy use `MatCreateSeqAIJWithArrays()`
4110 
4111   This routine may be called multiple times with different nonzero patterns (or the same nonzero pattern). The nonzero
4112   structure will be the union of all the previous nonzero structures.
4113 
4114   Developer Notes:
4115   An optimization could be added to the implementation where it checks if the `i`, and `j` are identical to the current `i` and `j` and
4116   then just copies the `v` values directly with `PetscMemcpy()`.
4117 
4118   This routine could also take a `PetscCopyMode` argument to allow sharing the values instead of always copying them.
4119 
4120 .seealso: [](ch_matrices), `Mat`, `MatCreate()`, `MatCreateSeqAIJ()`, `MatSetValues()`, `MatSeqAIJSetPreallocation()`, `MATSEQAIJ`, `MatResetPreallocation()`
4121 @*/
4122 PetscErrorCode MatSeqAIJSetPreallocationCSR(Mat B, const PetscInt i[], const PetscInt j[], const PetscScalar v[])
4123 {
4124   PetscFunctionBegin;
4125   PetscValidHeaderSpecific(B, MAT_CLASSID, 1);
4126   PetscValidType(B, 1);
4127   PetscTryMethod(B, "MatSeqAIJSetPreallocationCSR_C", (Mat, const PetscInt[], const PetscInt[], const PetscScalar[]), (B, i, j, v));
4128   PetscFunctionReturn(PETSC_SUCCESS);
4129 }
4130 
4131 static PetscErrorCode MatSeqAIJSetPreallocationCSR_SeqAIJ(Mat B, const PetscInt Ii[], const PetscInt J[], const PetscScalar v[])
4132 {
4133   PetscInt  i;
4134   PetscInt  m, n;
4135   PetscInt  nz;
4136   PetscInt *nnz;
4137 
4138   PetscFunctionBegin;
4139   PetscCheck(Ii[0] == 0, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Ii[0] must be 0 it is %" PetscInt_FMT, Ii[0]);
4140 
4141   PetscCall(PetscLayoutSetUp(B->rmap));
4142   PetscCall(PetscLayoutSetUp(B->cmap));
4143 
4144   PetscCall(MatGetSize(B, &m, &n));
4145   PetscCall(PetscMalloc1(m + 1, &nnz));
4146   for (i = 0; i < m; i++) {
4147     nz = Ii[i + 1] - Ii[i];
4148     PetscCheck(nz >= 0, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Local row %" PetscInt_FMT " has a negative number of columns %" PetscInt_FMT, i, nz);
4149     nnz[i] = nz;
4150   }
4151   PetscCall(MatSeqAIJSetPreallocation(B, 0, nnz));
4152   PetscCall(PetscFree(nnz));
4153 
4154   for (i = 0; i < m; i++) PetscCall(MatSetValues_SeqAIJ(B, 1, &i, Ii[i + 1] - Ii[i], J + Ii[i], PetscSafePointerPlusOffset(v, Ii[i]), INSERT_VALUES));
4155 
4156   PetscCall(MatAssemblyBegin(B, MAT_FINAL_ASSEMBLY));
4157   PetscCall(MatAssemblyEnd(B, MAT_FINAL_ASSEMBLY));
4158 
4159   PetscCall(MatSetOption(B, MAT_NEW_NONZERO_LOCATION_ERR, PETSC_TRUE));
4160   PetscFunctionReturn(PETSC_SUCCESS);
4161 }
4162 
4163 /*@
4164   MatSeqAIJKron - Computes `C`, the Kronecker product of `A` and `B`.
4165 
4166   Input Parameters:
4167 + A     - left-hand side matrix
4168 . B     - right-hand side matrix
4169 - reuse - either `MAT_INITIAL_MATRIX` or `MAT_REUSE_MATRIX`
4170 
4171   Output Parameter:
4172 . C - Kronecker product of `A` and `B`
4173 
4174   Level: intermediate
4175 
4176   Note:
4177   `MAT_REUSE_MATRIX` can only be used when the nonzero structure of the product matrix has not changed from that last call to `MatSeqAIJKron()`.
4178 
4179 .seealso: [](ch_matrices), `Mat`, `MatCreateSeqAIJ()`, `MATSEQAIJ`, `MATKAIJ`, `MatReuse`
4180 @*/
4181 PetscErrorCode MatSeqAIJKron(Mat A, Mat B, MatReuse reuse, Mat *C)
4182 {
4183   PetscFunctionBegin;
4184   PetscValidHeaderSpecific(A, MAT_CLASSID, 1);
4185   PetscValidType(A, 1);
4186   PetscValidHeaderSpecific(B, MAT_CLASSID, 2);
4187   PetscValidType(B, 2);
4188   PetscAssertPointer(C, 4);
4189   if (reuse == MAT_REUSE_MATRIX) {
4190     PetscValidHeaderSpecific(*C, MAT_CLASSID, 4);
4191     PetscValidType(*C, 4);
4192   }
4193   PetscTryMethod(A, "MatSeqAIJKron_C", (Mat, Mat, MatReuse, Mat *), (A, B, reuse, C));
4194   PetscFunctionReturn(PETSC_SUCCESS);
4195 }
4196 
4197 static PetscErrorCode MatSeqAIJKron_SeqAIJ(Mat A, Mat B, MatReuse reuse, Mat *C)
4198 {
4199   Mat                newmat;
4200   Mat_SeqAIJ        *a = (Mat_SeqAIJ *)A->data;
4201   Mat_SeqAIJ        *b = (Mat_SeqAIJ *)B->data;
4202   PetscScalar       *v;
4203   const PetscScalar *aa, *ba;
4204   PetscInt          *i, *j, m, n, p, q, nnz = 0, am = A->rmap->n, bm = B->rmap->n, an = A->cmap->n, bn = B->cmap->n;
4205   PetscBool          flg;
4206 
4207   PetscFunctionBegin;
4208   PetscCheck(!A->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
4209   PetscCheck(A->assembled, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
4210   PetscCheck(!B->factortype, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for factored matrix");
4211   PetscCheck(B->assembled, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Not for unassembled matrix");
4212   PetscCall(PetscObjectTypeCompare((PetscObject)B, MATSEQAIJ, &flg));
4213   PetscCheck(flg, PETSC_COMM_SELF, PETSC_ERR_SUP, "MatType %s", ((PetscObject)B)->type_name);
4214   PetscCheck(reuse == MAT_INITIAL_MATRIX || reuse == MAT_REUSE_MATRIX, PETSC_COMM_SELF, PETSC_ERR_SUP, "MatReuse %d", (int)reuse);
4215   if (reuse == MAT_INITIAL_MATRIX) {
4216     PetscCall(PetscMalloc2(am * bm + 1, &i, a->i[am] * b->i[bm], &j));
4217     PetscCall(MatCreate(PETSC_COMM_SELF, &newmat));
4218     PetscCall(MatSetSizes(newmat, am * bm, an * bn, am * bm, an * bn));
4219     PetscCall(MatSetType(newmat, MATAIJ));
4220     i[0] = 0;
4221     for (m = 0; m < am; ++m) {
4222       for (p = 0; p < bm; ++p) {
4223         i[m * bm + p + 1] = i[m * bm + p] + (a->i[m + 1] - a->i[m]) * (b->i[p + 1] - b->i[p]);
4224         for (n = a->i[m]; n < a->i[m + 1]; ++n) {
4225           for (q = b->i[p]; q < b->i[p + 1]; ++q) j[nnz++] = a->j[n] * bn + b->j[q];
4226         }
4227       }
4228     }
4229     PetscCall(MatSeqAIJSetPreallocationCSR(newmat, i, j, NULL));
4230     *C = newmat;
4231     PetscCall(PetscFree2(i, j));
4232     nnz = 0;
4233   }
4234   PetscCall(MatSeqAIJGetArray(*C, &v));
4235   PetscCall(MatSeqAIJGetArrayRead(A, &aa));
4236   PetscCall(MatSeqAIJGetArrayRead(B, &ba));
4237   for (m = 0; m < am; ++m) {
4238     for (p = 0; p < bm; ++p) {
4239       for (n = a->i[m]; n < a->i[m + 1]; ++n) {
4240         for (q = b->i[p]; q < b->i[p + 1]; ++q) v[nnz++] = aa[n] * ba[q];
4241       }
4242     }
4243   }
4244   PetscCall(MatSeqAIJRestoreArray(*C, &v));
4245   PetscCall(MatSeqAIJRestoreArrayRead(A, &aa));
4246   PetscCall(MatSeqAIJRestoreArrayRead(B, &ba));
4247   PetscFunctionReturn(PETSC_SUCCESS);
4248 }
4249 
4250 #include <../src/mat/impls/dense/seq/dense.h>
4251 #include <petsc/private/kernels/petscaxpy.h>
4252 
4253 /*
4254     Computes (B'*A')' since computing B*A directly is untenable
4255 
4256                n                       p                          p
4257         [             ]       [             ]         [                 ]
4258       m [      A      ]  *  n [       B     ]   =   m [         C       ]
4259         [             ]       [             ]         [                 ]
4260 
4261 */
4262 PetscErrorCode MatMatMultNumeric_SeqDense_SeqAIJ(Mat A, Mat B, Mat C)
4263 {
4264   Mat_SeqDense      *sub_a = (Mat_SeqDense *)A->data;
4265   Mat_SeqAIJ        *sub_b = (Mat_SeqAIJ *)B->data;
4266   Mat_SeqDense      *sub_c = (Mat_SeqDense *)C->data;
4267   PetscInt           i, j, n, m, q, p;
4268   const PetscInt    *ii, *idx;
4269   const PetscScalar *b, *a, *a_q;
4270   PetscScalar       *c, *c_q;
4271   PetscInt           clda = sub_c->lda;
4272   PetscInt           alda = sub_a->lda;
4273 
4274   PetscFunctionBegin;
4275   m = A->rmap->n;
4276   n = A->cmap->n;
4277   p = B->cmap->n;
4278   a = sub_a->v;
4279   b = sub_b->a;
4280   c = sub_c->v;
4281   if (clda == m) {
4282     PetscCall(PetscArrayzero(c, m * p));
4283   } else {
4284     for (j = 0; j < p; j++)
4285       for (i = 0; i < m; i++) c[j * clda + i] = 0.0;
4286   }
4287   ii  = sub_b->i;
4288   idx = sub_b->j;
4289   for (i = 0; i < n; i++) {
4290     q = ii[i + 1] - ii[i];
4291     while (q-- > 0) {
4292       c_q = c + clda * (*idx);
4293       a_q = a + alda * i;
4294       PetscKernelAXPY(c_q, *b, a_q, m);
4295       idx++;
4296       b++;
4297     }
4298   }
4299   PetscFunctionReturn(PETSC_SUCCESS);
4300 }
4301 
4302 PetscErrorCode MatMatMultSymbolic_SeqDense_SeqAIJ(Mat A, Mat B, PetscReal fill, Mat C)
4303 {
4304   PetscInt  m = A->rmap->n, n = B->cmap->n;
4305   PetscBool cisdense;
4306 
4307   PetscFunctionBegin;
4308   PetscCheck(A->cmap->n == B->rmap->n, PETSC_COMM_SELF, PETSC_ERR_ARG_SIZ, "A->cmap->n %" PetscInt_FMT " != B->rmap->n %" PetscInt_FMT, A->cmap->n, B->rmap->n);
4309   PetscCall(MatSetSizes(C, m, n, m, n));
4310   PetscCall(MatSetBlockSizesFromMats(C, A, B));
4311   PetscCall(PetscObjectTypeCompareAny((PetscObject)C, &cisdense, MATSEQDENSE, MATSEQDENSECUDA, MATSEQDENSEHIP, ""));
4312   if (!cisdense) PetscCall(MatSetType(C, MATDENSE));
4313   PetscCall(MatSetUp(C));
4314 
4315   C->ops->matmultnumeric = MatMatMultNumeric_SeqDense_SeqAIJ;
4316   PetscFunctionReturn(PETSC_SUCCESS);
4317 }
4318 
4319 /*MC
4320    MATSEQAIJ - MATSEQAIJ = "seqaij" - A matrix type to be used for sequential sparse matrices,
4321    based on compressed sparse row format.
4322 
4323    Options Database Key:
4324 . -mat_type seqaij - sets the matrix type to "seqaij" during a call to MatSetFromOptions()
4325 
4326    Level: beginner
4327 
4328    Notes:
4329     `MatSetValues()` may be called for this matrix type with a `NULL` argument for the numerical values,
4330     in this case the values associated with the rows and columns one passes in are set to zero
4331     in the matrix
4332 
4333     `MatSetOptions`(,`MAT_STRUCTURE_ONLY`,`PETSC_TRUE`) may be called for this matrix type. In this no
4334     space is allocated for the nonzero entries and any entries passed with `MatSetValues()` are ignored
4335 
4336   Developer Note:
4337     It would be nice if all matrix formats supported passing `NULL` in for the numerical values
4338 
4339 .seealso: [](ch_matrices), `Mat`, `MatCreateSeqAIJ()`, `MatSetFromOptions()`, `MatSetType()`, `MatCreate()`, `MatType`, `MATSELL`, `MATSEQSELL`, `MATMPISELL`
4340 M*/
4341 
4342 /*MC
4343    MATAIJ - MATAIJ = "aij" - A matrix type to be used for sparse matrices.
4344 
4345    This matrix type is identical to `MATSEQAIJ` when constructed with a single process communicator,
4346    and `MATMPIAIJ` otherwise.  As a result, for single process communicators,
4347    `MatSeqAIJSetPreallocation()` is supported, and similarly `MatMPIAIJSetPreallocation()` is supported
4348    for communicators controlling multiple processes.  It is recommended that you call both of
4349    the above preallocation routines for simplicity.
4350 
4351    Options Database Key:
4352 . -mat_type aij - sets the matrix type to "aij" during a call to `MatSetFromOptions()`
4353 
4354   Level: beginner
4355 
4356    Note:
4357    Subclasses include `MATAIJCUSPARSE`, `MATAIJPERM`, `MATAIJSELL`, `MATAIJMKL`, `MATAIJCRL`, and also automatically switches over to use inodes when
4358    enough exist.
4359 
4360 .seealso: [](ch_matrices), `Mat`, `MatCreateAIJ()`, `MatCreateSeqAIJ()`, `MATSEQAIJ`, `MATMPIAIJ`, `MATSELL`, `MATSEQSELL`, `MATMPISELL`
4361 M*/
4362 
4363 /*MC
4364    MATAIJCRL - MATAIJCRL = "aijcrl" - A matrix type to be used for sparse matrices.
4365 
4366    Options Database Key:
4367 . -mat_type aijcrl - sets the matrix type to "aijcrl" during a call to `MatSetFromOptions()`
4368 
4369   Level: beginner
4370 
4371    Note:
4372    This matrix type is identical to `MATSEQAIJCRL` when constructed with a single process communicator,
4373    and `MATMPIAIJCRL` otherwise.  As a result, for single process communicators,
4374    `MatSeqAIJSetPreallocation()` is supported, and similarly `MatMPIAIJSetPreallocation()` is supported
4375    for communicators controlling multiple processes.  It is recommended that you call both of
4376    the above preallocation routines for simplicity.
4377 
4378 .seealso: [](ch_matrices), `Mat`, `MatCreateMPIAIJCRL`, `MATSEQAIJCRL`, `MATMPIAIJCRL`, `MATSEQAIJCRL`, `MATMPIAIJCRL`
4379 M*/
4380 
4381 PETSC_INTERN PetscErrorCode MatConvert_SeqAIJ_SeqAIJCRL(Mat, MatType, MatReuse, Mat *);
4382 #if defined(PETSC_HAVE_ELEMENTAL)
4383 PETSC_INTERN PetscErrorCode MatConvert_SeqAIJ_Elemental(Mat, MatType, MatReuse, Mat *);
4384 #endif
4385 #if defined(PETSC_HAVE_SCALAPACK)
4386 PETSC_INTERN PetscErrorCode MatConvert_AIJ_ScaLAPACK(Mat, MatType, MatReuse, Mat *);
4387 #endif
4388 #if defined(PETSC_HAVE_HYPRE)
4389 PETSC_INTERN PetscErrorCode MatConvert_AIJ_HYPRE(Mat A, MatType, MatReuse, Mat *);
4390 #endif
4391 
4392 PETSC_EXTERN PetscErrorCode MatConvert_SeqAIJ_SeqSELL(Mat, MatType, MatReuse, Mat *);
4393 PETSC_INTERN PetscErrorCode MatConvert_XAIJ_IS(Mat, MatType, MatReuse, Mat *);
4394 PETSC_INTERN PetscErrorCode MatProductSetFromOptions_IS_XAIJ(Mat);
4395 
4396 /*@C
4397   MatSeqAIJGetArray - gives read/write access to the array where the data for a `MATSEQAIJ` matrix is stored
4398 
4399   Not Collective
4400 
4401   Input Parameter:
4402 . A - a `MATSEQAIJ` matrix
4403 
4404   Output Parameter:
4405 . array - pointer to the data
4406 
4407   Level: intermediate
4408 
4409   Fortran Notes:
4410   `MatSeqAIJGetArray()` Fortran binding is deprecated (since PETSc 3.19), use `MatSeqAIJGetArrayF90()`
4411 
4412 .seealso: [](ch_matrices), `Mat`, `MatSeqAIJRestoreArray()`, `MatSeqAIJGetArrayF90()`
4413 @*/
4414 PetscErrorCode MatSeqAIJGetArray(Mat A, PetscScalar **array)
4415 {
4416   Mat_SeqAIJ *aij = (Mat_SeqAIJ *)A->data;
4417 
4418   PetscFunctionBegin;
4419   if (aij->ops->getarray) {
4420     PetscCall((*aij->ops->getarray)(A, array));
4421   } else {
4422     *array = aij->a;
4423   }
4424   PetscFunctionReturn(PETSC_SUCCESS);
4425 }
4426 
4427 /*@C
4428   MatSeqAIJRestoreArray - returns access to the array where the data for a `MATSEQAIJ` matrix is stored obtained by `MatSeqAIJGetArray()`
4429 
4430   Not Collective
4431 
4432   Input Parameters:
4433 + A     - a `MATSEQAIJ` matrix
4434 - array - pointer to the data
4435 
4436   Level: intermediate
4437 
4438   Fortran Notes:
4439   `MatSeqAIJRestoreArray()` Fortran binding is deprecated (since PETSc 3.19), use `MatSeqAIJRestoreArrayF90()`
4440 
4441 .seealso: [](ch_matrices), `Mat`, `MatSeqAIJGetArray()`, `MatSeqAIJRestoreArrayF90()`
4442 @*/
4443 PetscErrorCode MatSeqAIJRestoreArray(Mat A, PetscScalar **array)
4444 {
4445   Mat_SeqAIJ *aij = (Mat_SeqAIJ *)A->data;
4446 
4447   PetscFunctionBegin;
4448   if (aij->ops->restorearray) {
4449     PetscCall((*aij->ops->restorearray)(A, array));
4450   } else {
4451     *array = NULL;
4452   }
4453   PetscCall(MatSeqAIJInvalidateDiagonal(A));
4454   PetscCall(PetscObjectStateIncrease((PetscObject)A));
4455   PetscFunctionReturn(PETSC_SUCCESS);
4456 }
4457 
4458 /*@C
4459   MatSeqAIJGetArrayRead - gives read-only access to the array where the data for a `MATSEQAIJ` matrix is stored
4460 
4461   Not Collective; No Fortran Support
4462 
4463   Input Parameter:
4464 . A - a `MATSEQAIJ` matrix
4465 
4466   Output Parameter:
4467 . array - pointer to the data
4468 
4469   Level: intermediate
4470 
4471 .seealso: [](ch_matrices), `Mat`, `MatSeqAIJGetArray()`, `MatSeqAIJRestoreArrayRead()`
4472 @*/
4473 PetscErrorCode MatSeqAIJGetArrayRead(Mat A, const PetscScalar **array)
4474 {
4475   Mat_SeqAIJ *aij = (Mat_SeqAIJ *)A->data;
4476 
4477   PetscFunctionBegin;
4478   if (aij->ops->getarrayread) {
4479     PetscCall((*aij->ops->getarrayread)(A, array));
4480   } else {
4481     *array = aij->a;
4482   }
4483   PetscFunctionReturn(PETSC_SUCCESS);
4484 }
4485 
4486 /*@C
4487   MatSeqAIJRestoreArrayRead - restore the read-only access array obtained from `MatSeqAIJGetArrayRead()`
4488 
4489   Not Collective; No Fortran Support
4490 
4491   Input Parameter:
4492 . A - a `MATSEQAIJ` matrix
4493 
4494   Output Parameter:
4495 . array - pointer to the data
4496 
4497   Level: intermediate
4498 
4499 .seealso: [](ch_matrices), `Mat`, `MatSeqAIJGetArray()`, `MatSeqAIJGetArrayRead()`
4500 @*/
4501 PetscErrorCode MatSeqAIJRestoreArrayRead(Mat A, const PetscScalar **array)
4502 {
4503   Mat_SeqAIJ *aij = (Mat_SeqAIJ *)A->data;
4504 
4505   PetscFunctionBegin;
4506   if (aij->ops->restorearrayread) {
4507     PetscCall((*aij->ops->restorearrayread)(A, array));
4508   } else {
4509     *array = NULL;
4510   }
4511   PetscFunctionReturn(PETSC_SUCCESS);
4512 }
4513 
4514 /*@C
4515   MatSeqAIJGetArrayWrite - gives write-only access to the array where the data for a `MATSEQAIJ` matrix is stored
4516 
4517   Not Collective; No Fortran Support
4518 
4519   Input Parameter:
4520 . A - a `MATSEQAIJ` matrix
4521 
4522   Output Parameter:
4523 . array - pointer to the data
4524 
4525   Level: intermediate
4526 
4527 .seealso: [](ch_matrices), `Mat`, `MatSeqAIJGetArray()`, `MatSeqAIJRestoreArrayRead()`
4528 @*/
4529 PetscErrorCode MatSeqAIJGetArrayWrite(Mat A, PetscScalar **array)
4530 {
4531   Mat_SeqAIJ *aij = (Mat_SeqAIJ *)A->data;
4532 
4533   PetscFunctionBegin;
4534   if (aij->ops->getarraywrite) {
4535     PetscCall((*aij->ops->getarraywrite)(A, array));
4536   } else {
4537     *array = aij->a;
4538   }
4539   PetscCall(MatSeqAIJInvalidateDiagonal(A));
4540   PetscCall(PetscObjectStateIncrease((PetscObject)A));
4541   PetscFunctionReturn(PETSC_SUCCESS);
4542 }
4543 
4544 /*@C
4545   MatSeqAIJRestoreArrayWrite - restore the read-only access array obtained from MatSeqAIJGetArrayRead
4546 
4547   Not Collective; No Fortran Support
4548 
4549   Input Parameter:
4550 . A - a MATSEQAIJ matrix
4551 
4552   Output Parameter:
4553 . array - pointer to the data
4554 
4555   Level: intermediate
4556 
4557 .seealso: [](ch_matrices), `Mat`, `MatSeqAIJGetArray()`, `MatSeqAIJGetArrayRead()`
4558 @*/
4559 PetscErrorCode MatSeqAIJRestoreArrayWrite(Mat A, PetscScalar **array)
4560 {
4561   Mat_SeqAIJ *aij = (Mat_SeqAIJ *)A->data;
4562 
4563   PetscFunctionBegin;
4564   if (aij->ops->restorearraywrite) {
4565     PetscCall((*aij->ops->restorearraywrite)(A, array));
4566   } else {
4567     *array = NULL;
4568   }
4569   PetscFunctionReturn(PETSC_SUCCESS);
4570 }
4571 
4572 /*@C
4573   MatSeqAIJGetCSRAndMemType - Get the CSR arrays and the memory type of the `MATSEQAIJ` matrix
4574 
4575   Not Collective; No Fortran Support
4576 
4577   Input Parameter:
4578 . mat - a matrix of type `MATSEQAIJ` or its subclasses
4579 
4580   Output Parameters:
4581 + i     - row map array of the matrix
4582 . j     - column index array of the matrix
4583 . a     - data array of the matrix
4584 - mtype - memory type of the arrays
4585 
4586   Level: developer
4587 
4588   Notes:
4589   Any of the output parameters can be `NULL`, in which case the corresponding value is not returned.
4590   If mat is a device matrix, the arrays are on the device. Otherwise, they are on the host.
4591 
4592   One can call this routine on a preallocated but not assembled matrix to just get the memory of the CSR underneath the matrix.
4593   If the matrix is assembled, the data array `a` is guaranteed to have the latest values of the matrix.
4594 
4595 .seealso: [](ch_matrices), `Mat`, `MatSeqAIJGetArray()`, `MatSeqAIJGetArrayRead()`
4596 @*/
4597 PetscErrorCode MatSeqAIJGetCSRAndMemType(Mat mat, const PetscInt **i, const PetscInt **j, PetscScalar **a, PetscMemType *mtype)
4598 {
4599   Mat_SeqAIJ *aij = (Mat_SeqAIJ *)mat->data;
4600 
4601   PetscFunctionBegin;
4602   PetscCheck(mat->preallocated, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "matrix is not preallocated");
4603   if (aij->ops->getcsrandmemtype) {
4604     PetscCall((*aij->ops->getcsrandmemtype)(mat, i, j, a, mtype));
4605   } else {
4606     if (i) *i = aij->i;
4607     if (j) *j = aij->j;
4608     if (a) *a = aij->a;
4609     if (mtype) *mtype = PETSC_MEMTYPE_HOST;
4610   }
4611   PetscFunctionReturn(PETSC_SUCCESS);
4612 }
4613 
4614 /*@C
4615   MatSeqAIJGetMaxRowNonzeros - returns the maximum number of nonzeros in any row
4616 
4617   Not Collective
4618 
4619   Input Parameter:
4620 . A - a `MATSEQAIJ` matrix
4621 
4622   Output Parameter:
4623 . nz - the maximum number of nonzeros in any row
4624 
4625   Level: intermediate
4626 
4627 .seealso: [](ch_matrices), `Mat`, `MatSeqAIJRestoreArray()`, `MatSeqAIJGetArrayF90()`
4628 @*/
4629 PetscErrorCode MatSeqAIJGetMaxRowNonzeros(Mat A, PetscInt *nz)
4630 {
4631   Mat_SeqAIJ *aij = (Mat_SeqAIJ *)A->data;
4632 
4633   PetscFunctionBegin;
4634   *nz = aij->rmax;
4635   PetscFunctionReturn(PETSC_SUCCESS);
4636 }
4637 
4638 static PetscErrorCode MatCOOStructDestroy_SeqAIJ(void *data)
4639 {
4640   MatCOOStruct_SeqAIJ *coo = (MatCOOStruct_SeqAIJ *)data;
4641 
4642   PetscFunctionBegin;
4643   PetscCall(PetscFree(coo->perm));
4644   PetscCall(PetscFree(coo->jmap));
4645   PetscCall(PetscFree(coo));
4646   PetscFunctionReturn(PETSC_SUCCESS);
4647 }
4648 
4649 PetscErrorCode MatSetPreallocationCOO_SeqAIJ(Mat mat, PetscCount coo_n, PetscInt coo_i[], PetscInt coo_j[])
4650 {
4651   MPI_Comm             comm;
4652   PetscInt            *i, *j;
4653   PetscInt             M, N, row, iprev;
4654   PetscCount           k, p, q, nneg, nnz, start, end; /* Index the coo array, so use PetscCount as their type */
4655   PetscInt            *Ai;                             /* Change to PetscCount once we use it for row pointers */
4656   PetscInt            *Aj;
4657   PetscScalar         *Aa;
4658   Mat_SeqAIJ          *seqaij = (Mat_SeqAIJ *)mat->data;
4659   MatType              rtype;
4660   PetscCount          *perm, *jmap;
4661   PetscContainer       container;
4662   MatCOOStruct_SeqAIJ *coo;
4663   PetscBool            isorted;
4664 
4665   PetscFunctionBegin;
4666   PetscCall(PetscObjectGetComm((PetscObject)mat, &comm));
4667   PetscCall(MatGetSize(mat, &M, &N));
4668   i = coo_i;
4669   j = coo_j;
4670   PetscCall(PetscMalloc1(coo_n, &perm));
4671 
4672   /* Ignore entries with negative row or col indices; at the same time, check if i[] is already sorted (e.g., MatConvert_AlJ_HYPRE results in this case) */
4673   isorted = PETSC_TRUE;
4674   iprev   = PETSC_INT_MIN;
4675   for (k = 0; k < coo_n; k++) {
4676     if (j[k] < 0) i[k] = -1;
4677     if (isorted) {
4678       if (i[k] < iprev) isorted = PETSC_FALSE;
4679       else iprev = i[k];
4680     }
4681     perm[k] = k;
4682   }
4683 
4684   /* Sort by row if not already */
4685   if (!isorted) PetscCall(PetscSortIntWithIntCountArrayPair(coo_n, i, j, perm));
4686 
4687   /* Advance k to the first row with a non-negative index */
4688   for (k = 0; k < coo_n; k++)
4689     if (i[k] >= 0) break;
4690   nneg = k;
4691   PetscCall(PetscMalloc1(coo_n - nneg + 1, &jmap)); /* +1 to make a CSR-like data structure. jmap[i] originally is the number of repeats for i-th nonzero */
4692   nnz = 0;                                          /* Total number of unique nonzeros to be counted */
4693   jmap++;                                           /* Inc jmap by 1 for convenience */
4694 
4695   PetscCall(PetscCalloc1(M + 1, &Ai));        /* CSR of A */
4696   PetscCall(PetscMalloc1(coo_n - nneg, &Aj)); /* We have at most coo_n-nneg unique nonzeros */
4697 
4698   /* Support for HYPRE */
4699   PetscBool   hypre;
4700   const char *name;
4701   PetscCall(PetscObjectGetName((PetscObject)mat, &name));
4702   PetscCall(PetscStrcmp("_internal_COO_mat_for_hypre", name, &hypre));
4703 
4704   /* In each row, sort by column, then unique column indices to get row length */
4705   Ai++;  /* Inc by 1 for convenience */
4706   q = 0; /* q-th unique nonzero, with q starting from 0 */
4707   while (k < coo_n) {
4708     PetscBool strictly_sorted; // this row is strictly sorted?
4709     PetscInt  jprev;
4710 
4711     /* get [start,end) indices for this row; also check if cols in this row are strictly sorted */
4712     row             = i[k];
4713     start           = k;
4714     jprev           = PETSC_INT_MIN;
4715     strictly_sorted = PETSC_TRUE;
4716     while (k < coo_n && i[k] == row) {
4717       if (strictly_sorted) {
4718         if (j[k] <= jprev) strictly_sorted = PETSC_FALSE;
4719         else jprev = j[k];
4720       }
4721       k++;
4722     }
4723     end = k;
4724 
4725     /* hack for HYPRE: swap min column to diag so that diagonal values will go first */
4726     if (hypre) {
4727       PetscInt  minj    = PETSC_MAX_INT;
4728       PetscBool hasdiag = PETSC_FALSE;
4729 
4730       if (strictly_sorted) { // fast path to swap the first and the diag
4731         PetscCount tmp;
4732         for (p = start; p < end; p++) {
4733           if (j[p] == row && p != start) {
4734             j[p]        = j[start];
4735             j[start]    = row;
4736             tmp         = perm[start];
4737             perm[start] = perm[p];
4738             perm[p]     = tmp;
4739             break;
4740           }
4741         }
4742       } else {
4743         for (p = start; p < end; p++) {
4744           hasdiag = (PetscBool)(hasdiag || (j[p] == row));
4745           minj    = PetscMin(minj, j[p]);
4746         }
4747 
4748         if (hasdiag) {
4749           for (p = start; p < end; p++) {
4750             if (j[p] == minj) j[p] = row;
4751             else if (j[p] == row) j[p] = minj;
4752           }
4753         }
4754       }
4755     }
4756     // sort by columns in a row
4757     if (!strictly_sorted) PetscCall(PetscSortIntWithCountArray(end - start, j + start, perm + start));
4758 
4759     if (strictly_sorted) { // fast path to set Aj[], jmap[], Ai[], nnz, q
4760       for (p = start; p < end; p++, q++) {
4761         Aj[q]   = j[p];
4762         jmap[q] = 1;
4763       }
4764       Ai[row] = end - start;
4765       nnz += Ai[row]; // q is already advanced
4766     } else {
4767       /* Find number of unique col entries in this row */
4768       Aj[q]   = j[start]; /* Log the first nonzero in this row */
4769       jmap[q] = 1;        /* Number of repeats of this nonzero entry */
4770       Ai[row] = 1;
4771       nnz++;
4772 
4773       for (p = start + 1; p < end; p++) { /* Scan remaining nonzero in this row */
4774         if (j[p] != j[p - 1]) {           /* Meet a new nonzero */
4775           q++;
4776           jmap[q] = 1;
4777           Aj[q]   = j[p];
4778           Ai[row]++;
4779           nnz++;
4780         } else {
4781           jmap[q]++;
4782         }
4783       }
4784       q++; /* Move to next row and thus next unique nonzero */
4785     }
4786   }
4787 
4788   Ai--; /* Back to the beginning of Ai[] */
4789   for (k = 0; k < M; k++) Ai[k + 1] += Ai[k];
4790   jmap--; // Back to the beginning of jmap[]
4791   jmap[0] = 0;
4792   for (k = 0; k < nnz; k++) jmap[k + 1] += jmap[k];
4793 
4794   if (nnz < coo_n - nneg) { /* Realloc with actual number of unique nonzeros */
4795     PetscCount *jmap_new;
4796     PetscInt   *Aj_new;
4797 
4798     PetscCall(PetscMalloc1(nnz + 1, &jmap_new));
4799     PetscCall(PetscArraycpy(jmap_new, jmap, nnz + 1));
4800     PetscCall(PetscFree(jmap));
4801     jmap = jmap_new;
4802 
4803     PetscCall(PetscMalloc1(nnz, &Aj_new));
4804     PetscCall(PetscArraycpy(Aj_new, Aj, nnz));
4805     PetscCall(PetscFree(Aj));
4806     Aj = Aj_new;
4807   }
4808 
4809   if (nneg) { /* Discard heading entries with negative indices in perm[], as we'll access it from index 0 in MatSetValuesCOO */
4810     PetscCount *perm_new;
4811 
4812     PetscCall(PetscMalloc1(coo_n - nneg, &perm_new));
4813     PetscCall(PetscArraycpy(perm_new, perm + nneg, coo_n - nneg));
4814     PetscCall(PetscFree(perm));
4815     perm = perm_new;
4816   }
4817 
4818   PetscCall(MatGetRootType_Private(mat, &rtype));
4819   PetscCall(PetscCalloc1(nnz, &Aa)); /* Zero the matrix */
4820   PetscCall(MatSetSeqAIJWithArrays_private(PETSC_COMM_SELF, M, N, Ai, Aj, Aa, rtype, mat));
4821 
4822   seqaij->singlemalloc = PETSC_FALSE;            /* Ai, Aj and Aa are not allocated in one big malloc */
4823   seqaij->free_a = seqaij->free_ij = PETSC_TRUE; /* Let newmat own Ai, Aj and Aa */
4824 
4825   // Put the COO struct in a container and then attach that to the matrix
4826   PetscCall(PetscMalloc1(1, &coo));
4827   coo->nz   = nnz;
4828   coo->n    = coo_n;
4829   coo->Atot = coo_n - nneg; // Annz is seqaij->nz, so no need to record that again
4830   coo->jmap = jmap;         // of length nnz+1
4831   coo->perm = perm;
4832   PetscCall(PetscContainerCreate(PETSC_COMM_SELF, &container));
4833   PetscCall(PetscContainerSetPointer(container, coo));
4834   PetscCall(PetscContainerSetUserDestroy(container, MatCOOStructDestroy_SeqAIJ));
4835   PetscCall(PetscObjectCompose((PetscObject)mat, "__PETSc_MatCOOStruct_Host", (PetscObject)container));
4836   PetscCall(PetscContainerDestroy(&container));
4837   PetscFunctionReturn(PETSC_SUCCESS);
4838 }
4839 
4840 static PetscErrorCode MatSetValuesCOO_SeqAIJ(Mat A, const PetscScalar v[], InsertMode imode)
4841 {
4842   Mat_SeqAIJ          *aseq = (Mat_SeqAIJ *)A->data;
4843   PetscCount           i, j, Annz = aseq->nz;
4844   PetscCount          *perm, *jmap;
4845   PetscScalar         *Aa;
4846   PetscContainer       container;
4847   MatCOOStruct_SeqAIJ *coo;
4848 
4849   PetscFunctionBegin;
4850   PetscCall(PetscObjectQuery((PetscObject)A, "__PETSc_MatCOOStruct_Host", (PetscObject *)&container));
4851   PetscCheck(container, PETSC_COMM_SELF, PETSC_ERR_PLIB, "Not found MatCOOStruct on this matrix");
4852   PetscCall(PetscContainerGetPointer(container, (void **)&coo));
4853   perm = coo->perm;
4854   jmap = coo->jmap;
4855   PetscCall(MatSeqAIJGetArray(A, &Aa));
4856   for (i = 0; i < Annz; i++) {
4857     PetscScalar sum = 0.0;
4858     for (j = jmap[i]; j < jmap[i + 1]; j++) sum += v[perm[j]];
4859     Aa[i] = (imode == INSERT_VALUES ? 0.0 : Aa[i]) + sum;
4860   }
4861   PetscCall(MatSeqAIJRestoreArray(A, &Aa));
4862   PetscFunctionReturn(PETSC_SUCCESS);
4863 }
4864 
4865 #if defined(PETSC_HAVE_CUDA)
4866 PETSC_INTERN PetscErrorCode MatConvert_SeqAIJ_SeqAIJCUSPARSE(Mat, MatType, MatReuse, Mat *);
4867 #endif
4868 #if defined(PETSC_HAVE_HIP)
4869 PETSC_INTERN PetscErrorCode MatConvert_SeqAIJ_SeqAIJHIPSPARSE(Mat, MatType, MatReuse, Mat *);
4870 #endif
4871 #if defined(PETSC_HAVE_KOKKOS_KERNELS)
4872 PETSC_INTERN PetscErrorCode MatConvert_SeqAIJ_SeqAIJKokkos(Mat, MatType, MatReuse, Mat *);
4873 #endif
4874 
4875 PETSC_EXTERN PetscErrorCode MatCreate_SeqAIJ(Mat B)
4876 {
4877   Mat_SeqAIJ *b;
4878   PetscMPIInt size;
4879 
4880   PetscFunctionBegin;
4881   PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)B), &size));
4882   PetscCheck(size <= 1, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Comm must be of size 1");
4883 
4884   PetscCall(PetscNew(&b));
4885 
4886   B->data   = (void *)b;
4887   B->ops[0] = MatOps_Values;
4888   if (B->sortedfull) B->ops->setvalues = MatSetValues_SeqAIJ_SortedFull;
4889 
4890   b->row                = NULL;
4891   b->col                = NULL;
4892   b->icol               = NULL;
4893   b->reallocs           = 0;
4894   b->ignorezeroentries  = PETSC_FALSE;
4895   b->roworiented        = PETSC_TRUE;
4896   b->nonew              = 0;
4897   b->diag               = NULL;
4898   b->solve_work         = NULL;
4899   B->spptr              = NULL;
4900   b->saved_values       = NULL;
4901   b->idiag              = NULL;
4902   b->mdiag              = NULL;
4903   b->ssor_work          = NULL;
4904   b->omega              = 1.0;
4905   b->fshift             = 0.0;
4906   b->idiagvalid         = PETSC_FALSE;
4907   b->ibdiagvalid        = PETSC_FALSE;
4908   b->keepnonzeropattern = PETSC_FALSE;
4909 
4910   PetscCall(PetscObjectChangeTypeName((PetscObject)B, MATSEQAIJ));
4911 #if defined(PETSC_HAVE_MATLAB)
4912   PetscCall(PetscObjectComposeFunction((PetscObject)B, "PetscMatlabEnginePut_C", MatlabEnginePut_SeqAIJ));
4913   PetscCall(PetscObjectComposeFunction((PetscObject)B, "PetscMatlabEngineGet_C", MatlabEngineGet_SeqAIJ));
4914 #endif
4915   PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatSeqAIJSetColumnIndices_C", MatSeqAIJSetColumnIndices_SeqAIJ));
4916   PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatStoreValues_C", MatStoreValues_SeqAIJ));
4917   PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatRetrieveValues_C", MatRetrieveValues_SeqAIJ));
4918   PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatConvert_seqaij_seqsbaij_C", MatConvert_SeqAIJ_SeqSBAIJ));
4919   PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatConvert_seqaij_seqbaij_C", MatConvert_SeqAIJ_SeqBAIJ));
4920   PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatConvert_seqaij_seqaijperm_C", MatConvert_SeqAIJ_SeqAIJPERM));
4921   PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatConvert_seqaij_seqaijsell_C", MatConvert_SeqAIJ_SeqAIJSELL));
4922 #if defined(PETSC_HAVE_MKL_SPARSE)
4923   PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatConvert_seqaij_seqaijmkl_C", MatConvert_SeqAIJ_SeqAIJMKL));
4924 #endif
4925 #if defined(PETSC_HAVE_CUDA)
4926   PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatConvert_seqaij_seqaijcusparse_C", MatConvert_SeqAIJ_SeqAIJCUSPARSE));
4927   PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatProductSetFromOptions_seqaijcusparse_seqaij_C", MatProductSetFromOptions_SeqAIJ));
4928   PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatProductSetFromOptions_seqaij_seqaijcusparse_C", MatProductSetFromOptions_SeqAIJ));
4929 #endif
4930 #if defined(PETSC_HAVE_HIP)
4931   PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatConvert_seqaij_seqaijhipsparse_C", MatConvert_SeqAIJ_SeqAIJHIPSPARSE));
4932   PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatProductSetFromOptions_seqaijhipsparse_seqaij_C", MatProductSetFromOptions_SeqAIJ));
4933   PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatProductSetFromOptions_seqaij_seqaijhipsparse_C", MatProductSetFromOptions_SeqAIJ));
4934 #endif
4935 #if defined(PETSC_HAVE_KOKKOS_KERNELS)
4936   PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatConvert_seqaij_seqaijkokkos_C", MatConvert_SeqAIJ_SeqAIJKokkos));
4937 #endif
4938   PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatConvert_seqaij_seqaijcrl_C", MatConvert_SeqAIJ_SeqAIJCRL));
4939 #if defined(PETSC_HAVE_ELEMENTAL)
4940   PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatConvert_seqaij_elemental_C", MatConvert_SeqAIJ_Elemental));
4941 #endif
4942 #if defined(PETSC_HAVE_SCALAPACK)
4943   PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatConvert_seqaij_scalapack_C", MatConvert_AIJ_ScaLAPACK));
4944 #endif
4945 #if defined(PETSC_HAVE_HYPRE)
4946   PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatConvert_seqaij_hypre_C", MatConvert_AIJ_HYPRE));
4947   PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatProductSetFromOptions_transpose_seqaij_seqaij_C", MatProductSetFromOptions_Transpose_AIJ_AIJ));
4948 #endif
4949   PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatConvert_seqaij_seqdense_C", MatConvert_SeqAIJ_SeqDense));
4950   PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatConvert_seqaij_seqsell_C", MatConvert_SeqAIJ_SeqSELL));
4951   PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatConvert_seqaij_is_C", MatConvert_XAIJ_IS));
4952   PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatIsTranspose_C", MatIsTranspose_SeqAIJ));
4953   PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatIsHermitianTranspose_C", MatIsHermitianTranspose_SeqAIJ));
4954   PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatSeqAIJSetPreallocation_C", MatSeqAIJSetPreallocation_SeqAIJ));
4955   PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatResetPreallocation_C", MatResetPreallocation_SeqAIJ));
4956   PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatSeqAIJSetPreallocationCSR_C", MatSeqAIJSetPreallocationCSR_SeqAIJ));
4957   PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatReorderForNonzeroDiagonal_C", MatReorderForNonzeroDiagonal_SeqAIJ));
4958   PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatProductSetFromOptions_is_seqaij_C", MatProductSetFromOptions_IS_XAIJ));
4959   PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatProductSetFromOptions_seqdense_seqaij_C", MatProductSetFromOptions_SeqDense_SeqAIJ));
4960   PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatProductSetFromOptions_seqaij_seqaij_C", MatProductSetFromOptions_SeqAIJ));
4961   PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatSeqAIJKron_C", MatSeqAIJKron_SeqAIJ));
4962   PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatSetPreallocationCOO_C", MatSetPreallocationCOO_SeqAIJ));
4963   PetscCall(PetscObjectComposeFunction((PetscObject)B, "MatSetValuesCOO_C", MatSetValuesCOO_SeqAIJ));
4964   PetscCall(MatCreate_SeqAIJ_Inode(B));
4965   PetscCall(PetscObjectChangeTypeName((PetscObject)B, MATSEQAIJ));
4966   PetscCall(MatSeqAIJSetTypeFromOptions(B)); /* this allows changing the matrix subtype to say MATSEQAIJPERM */
4967   PetscFunctionReturn(PETSC_SUCCESS);
4968 }
4969 
4970 /*
4971     Given a matrix generated with MatGetFactor() duplicates all the information in A into C
4972 */
4973 PetscErrorCode MatDuplicateNoCreate_SeqAIJ(Mat C, Mat A, MatDuplicateOption cpvalues, PetscBool mallocmatspace)
4974 {
4975   Mat_SeqAIJ *c = (Mat_SeqAIJ *)C->data, *a = (Mat_SeqAIJ *)A->data;
4976   PetscInt    m = A->rmap->n, i;
4977 
4978   PetscFunctionBegin;
4979   PetscCheck(A->assembled || cpvalues == MAT_DO_NOT_COPY_VALUES, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Cannot duplicate unassembled matrix");
4980 
4981   C->factortype    = A->factortype;
4982   c->row           = NULL;
4983   c->col           = NULL;
4984   c->icol          = NULL;
4985   c->reallocs      = 0;
4986   c->diagonaldense = a->diagonaldense;
4987 
4988   C->assembled = A->assembled;
4989 
4990   if (A->preallocated) {
4991     PetscCall(PetscLayoutReference(A->rmap, &C->rmap));
4992     PetscCall(PetscLayoutReference(A->cmap, &C->cmap));
4993 
4994     if (!A->hash_active) {
4995       PetscCall(PetscMalloc1(m, &c->imax));
4996       PetscCall(PetscMemcpy(c->imax, a->imax, m * sizeof(PetscInt)));
4997       PetscCall(PetscMalloc1(m, &c->ilen));
4998       PetscCall(PetscMemcpy(c->ilen, a->ilen, m * sizeof(PetscInt)));
4999 
5000       /* allocate the matrix space */
5001       if (mallocmatspace) {
5002         PetscCall(PetscMalloc3(a->i[m], &c->a, a->i[m], &c->j, m + 1, &c->i));
5003 
5004         c->singlemalloc = PETSC_TRUE;
5005 
5006         PetscCall(PetscArraycpy(c->i, a->i, m + 1));
5007         if (m > 0) {
5008           PetscCall(PetscArraycpy(c->j, a->j, a->i[m]));
5009           if (cpvalues == MAT_COPY_VALUES) {
5010             const PetscScalar *aa;
5011 
5012             PetscCall(MatSeqAIJGetArrayRead(A, &aa));
5013             PetscCall(PetscArraycpy(c->a, aa, a->i[m]));
5014             PetscCall(MatSeqAIJGetArrayRead(A, &aa));
5015           } else {
5016             PetscCall(PetscArrayzero(c->a, a->i[m]));
5017           }
5018         }
5019       }
5020       C->preallocated = PETSC_TRUE;
5021     } else {
5022       PetscCheck(mallocmatspace, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_WRONGSTATE, "Cannot malloc matrix memory from a non-preallocated matrix");
5023       PetscCall(MatSetUp(C));
5024     }
5025 
5026     c->ignorezeroentries = a->ignorezeroentries;
5027     c->roworiented       = a->roworiented;
5028     c->nonew             = a->nonew;
5029     if (a->diag) {
5030       PetscCall(PetscMalloc1(m + 1, &c->diag));
5031       PetscCall(PetscMemcpy(c->diag, a->diag, m * sizeof(PetscInt)));
5032     } else c->diag = NULL;
5033 
5034     c->solve_work         = NULL;
5035     c->saved_values       = NULL;
5036     c->idiag              = NULL;
5037     c->ssor_work          = NULL;
5038     c->keepnonzeropattern = a->keepnonzeropattern;
5039     c->free_a             = PETSC_TRUE;
5040     c->free_ij            = PETSC_TRUE;
5041 
5042     c->rmax  = a->rmax;
5043     c->nz    = a->nz;
5044     c->maxnz = a->nz; /* Since we allocate exactly the right amount */
5045 
5046     c->compressedrow.use   = a->compressedrow.use;
5047     c->compressedrow.nrows = a->compressedrow.nrows;
5048     if (a->compressedrow.use) {
5049       i = a->compressedrow.nrows;
5050       PetscCall(PetscMalloc2(i + 1, &c->compressedrow.i, i, &c->compressedrow.rindex));
5051       PetscCall(PetscArraycpy(c->compressedrow.i, a->compressedrow.i, i + 1));
5052       PetscCall(PetscArraycpy(c->compressedrow.rindex, a->compressedrow.rindex, i));
5053     } else {
5054       c->compressedrow.use    = PETSC_FALSE;
5055       c->compressedrow.i      = NULL;
5056       c->compressedrow.rindex = NULL;
5057     }
5058     c->nonzerorowcnt = a->nonzerorowcnt;
5059     C->nonzerostate  = A->nonzerostate;
5060 
5061     PetscCall(MatDuplicate_SeqAIJ_Inode(A, cpvalues, &C));
5062   }
5063   PetscCall(PetscFunctionListDuplicate(((PetscObject)A)->qlist, &((PetscObject)C)->qlist));
5064   PetscFunctionReturn(PETSC_SUCCESS);
5065 }
5066 
5067 PetscErrorCode MatDuplicate_SeqAIJ(Mat A, MatDuplicateOption cpvalues, Mat *B)
5068 {
5069   PetscFunctionBegin;
5070   PetscCall(MatCreate(PetscObjectComm((PetscObject)A), B));
5071   PetscCall(MatSetSizes(*B, A->rmap->n, A->cmap->n, A->rmap->n, A->cmap->n));
5072   if (!(A->rmap->n % A->rmap->bs) && !(A->cmap->n % A->cmap->bs)) PetscCall(MatSetBlockSizesFromMats(*B, A, A));
5073   PetscCall(MatSetType(*B, ((PetscObject)A)->type_name));
5074   PetscCall(MatDuplicateNoCreate_SeqAIJ(*B, A, cpvalues, PETSC_TRUE));
5075   PetscFunctionReturn(PETSC_SUCCESS);
5076 }
5077 
5078 PetscErrorCode MatLoad_SeqAIJ(Mat newMat, PetscViewer viewer)
5079 {
5080   PetscBool isbinary, ishdf5;
5081 
5082   PetscFunctionBegin;
5083   PetscValidHeaderSpecific(newMat, MAT_CLASSID, 1);
5084   PetscValidHeaderSpecific(viewer, PETSC_VIEWER_CLASSID, 2);
5085   /* force binary viewer to load .info file if it has not yet done so */
5086   PetscCall(PetscViewerSetUp(viewer));
5087   PetscCall(PetscObjectTypeCompare((PetscObject)viewer, PETSCVIEWERBINARY, &isbinary));
5088   PetscCall(PetscObjectTypeCompare((PetscObject)viewer, PETSCVIEWERHDF5, &ishdf5));
5089   if (isbinary) {
5090     PetscCall(MatLoad_SeqAIJ_Binary(newMat, viewer));
5091   } else if (ishdf5) {
5092 #if defined(PETSC_HAVE_HDF5)
5093     PetscCall(MatLoad_AIJ_HDF5(newMat, viewer));
5094 #else
5095     SETERRQ(PetscObjectComm((PetscObject)newMat), PETSC_ERR_SUP, "HDF5 not supported in this build.\nPlease reconfigure using --download-hdf5");
5096 #endif
5097   } else {
5098     SETERRQ(PetscObjectComm((PetscObject)newMat), PETSC_ERR_SUP, "Viewer type %s not yet supported for reading %s matrices", ((PetscObject)viewer)->type_name, ((PetscObject)newMat)->type_name);
5099   }
5100   PetscFunctionReturn(PETSC_SUCCESS);
5101 }
5102 
5103 PetscErrorCode MatLoad_SeqAIJ_Binary(Mat mat, PetscViewer viewer)
5104 {
5105   Mat_SeqAIJ *a = (Mat_SeqAIJ *)mat->data;
5106   PetscInt    header[4], *rowlens, M, N, nz, sum, rows, cols, i;
5107 
5108   PetscFunctionBegin;
5109   PetscCall(PetscViewerSetUp(viewer));
5110 
5111   /* read in matrix header */
5112   PetscCall(PetscViewerBinaryRead(viewer, header, 4, NULL, PETSC_INT));
5113   PetscCheck(header[0] == MAT_FILE_CLASSID, PETSC_COMM_SELF, PETSC_ERR_FILE_UNEXPECTED, "Not a matrix object in file");
5114   M  = header[1];
5115   N  = header[2];
5116   nz = header[3];
5117   PetscCheck(M >= 0, PetscObjectComm((PetscObject)viewer), PETSC_ERR_FILE_UNEXPECTED, "Matrix row size (%" PetscInt_FMT ") in file is negative", M);
5118   PetscCheck(N >= 0, PetscObjectComm((PetscObject)viewer), PETSC_ERR_FILE_UNEXPECTED, "Matrix column size (%" PetscInt_FMT ") in file is negative", N);
5119   PetscCheck(nz >= 0, PETSC_COMM_SELF, PETSC_ERR_FILE_UNEXPECTED, "Matrix stored in special format on disk, cannot load as SeqAIJ");
5120 
5121   /* set block sizes from the viewer's .info file */
5122   PetscCall(MatLoad_Binary_BlockSizes(mat, viewer));
5123   /* set local and global sizes if not set already */
5124   if (mat->rmap->n < 0) mat->rmap->n = M;
5125   if (mat->cmap->n < 0) mat->cmap->n = N;
5126   if (mat->rmap->N < 0) mat->rmap->N = M;
5127   if (mat->cmap->N < 0) mat->cmap->N = N;
5128   PetscCall(PetscLayoutSetUp(mat->rmap));
5129   PetscCall(PetscLayoutSetUp(mat->cmap));
5130 
5131   /* check if the matrix sizes are correct */
5132   PetscCall(MatGetSize(mat, &rows, &cols));
5133   PetscCheck(M == rows && N == cols, PETSC_COMM_SELF, PETSC_ERR_FILE_UNEXPECTED, "Matrix in file of different sizes (%" PetscInt_FMT ", %" PetscInt_FMT ") than the input matrix (%" PetscInt_FMT ", %" PetscInt_FMT ")", M, N, rows, cols);
5134 
5135   /* read in row lengths */
5136   PetscCall(PetscMalloc1(M, &rowlens));
5137   PetscCall(PetscViewerBinaryRead(viewer, rowlens, M, NULL, PETSC_INT));
5138   /* check if sum(rowlens) is same as nz */
5139   sum = 0;
5140   for (i = 0; i < M; i++) sum += rowlens[i];
5141   PetscCheck(sum == nz, PETSC_COMM_SELF, PETSC_ERR_FILE_UNEXPECTED, "Inconsistent matrix data in file: nonzeros = %" PetscInt_FMT ", sum-row-lengths = %" PetscInt_FMT, nz, sum);
5142   /* preallocate and check sizes */
5143   PetscCall(MatSeqAIJSetPreallocation_SeqAIJ(mat, 0, rowlens));
5144   PetscCall(MatGetSize(mat, &rows, &cols));
5145   PetscCheck(M == rows && N == cols, PETSC_COMM_SELF, PETSC_ERR_FILE_UNEXPECTED, "Matrix in file of different length (%" PetscInt_FMT ", %" PetscInt_FMT ") than the input matrix (%" PetscInt_FMT ", %" PetscInt_FMT ")", M, N, rows, cols);
5146   /* store row lengths */
5147   PetscCall(PetscArraycpy(a->ilen, rowlens, M));
5148   PetscCall(PetscFree(rowlens));
5149 
5150   /* fill in "i" row pointers */
5151   a->i[0] = 0;
5152   for (i = 0; i < M; i++) a->i[i + 1] = a->i[i] + a->ilen[i];
5153   /* read in "j" column indices */
5154   PetscCall(PetscViewerBinaryRead(viewer, a->j, nz, NULL, PETSC_INT));
5155   /* read in "a" nonzero values */
5156   PetscCall(PetscViewerBinaryRead(viewer, a->a, nz, NULL, PETSC_SCALAR));
5157 
5158   PetscCall(MatAssemblyBegin(mat, MAT_FINAL_ASSEMBLY));
5159   PetscCall(MatAssemblyEnd(mat, MAT_FINAL_ASSEMBLY));
5160   PetscFunctionReturn(PETSC_SUCCESS);
5161 }
5162 
5163 PetscErrorCode MatEqual_SeqAIJ(Mat A, Mat B, PetscBool *flg)
5164 {
5165   Mat_SeqAIJ        *a = (Mat_SeqAIJ *)A->data, *b = (Mat_SeqAIJ *)B->data;
5166   const PetscScalar *aa, *ba;
5167 #if defined(PETSC_USE_COMPLEX)
5168   PetscInt k;
5169 #endif
5170 
5171   PetscFunctionBegin;
5172   /* If the  matrix dimensions are not equal,or no of nonzeros */
5173   if ((A->rmap->n != B->rmap->n) || (A->cmap->n != B->cmap->n) || (a->nz != b->nz)) {
5174     *flg = PETSC_FALSE;
5175     PetscFunctionReturn(PETSC_SUCCESS);
5176   }
5177 
5178   /* if the a->i are the same */
5179   PetscCall(PetscArraycmp(a->i, b->i, A->rmap->n + 1, flg));
5180   if (!*flg) PetscFunctionReturn(PETSC_SUCCESS);
5181 
5182   /* if a->j are the same */
5183   PetscCall(PetscArraycmp(a->j, b->j, a->nz, flg));
5184   if (!*flg) PetscFunctionReturn(PETSC_SUCCESS);
5185 
5186   PetscCall(MatSeqAIJGetArrayRead(A, &aa));
5187   PetscCall(MatSeqAIJGetArrayRead(B, &ba));
5188   /* if a->a are the same */
5189 #if defined(PETSC_USE_COMPLEX)
5190   for (k = 0; k < a->nz; k++) {
5191     if (PetscRealPart(aa[k]) != PetscRealPart(ba[k]) || PetscImaginaryPart(aa[k]) != PetscImaginaryPart(ba[k])) {
5192       *flg = PETSC_FALSE;
5193       PetscFunctionReturn(PETSC_SUCCESS);
5194     }
5195   }
5196 #else
5197   PetscCall(PetscArraycmp(aa, ba, a->nz, flg));
5198 #endif
5199   PetscCall(MatSeqAIJRestoreArrayRead(A, &aa));
5200   PetscCall(MatSeqAIJRestoreArrayRead(B, &ba));
5201   PetscFunctionReturn(PETSC_SUCCESS);
5202 }
5203 
5204 /*@
5205   MatCreateSeqAIJWithArrays - Creates an sequential `MATSEQAIJ` matrix using matrix elements (in CSR format)
5206   provided by the user.
5207 
5208   Collective
5209 
5210   Input Parameters:
5211 + comm - must be an MPI communicator of size 1
5212 . m    - number of rows
5213 . n    - number of columns
5214 . i    - row indices; that is i[0] = 0, i[row] = i[row-1] + number of elements in that row of the matrix
5215 . j    - column indices
5216 - a    - matrix values
5217 
5218   Output Parameter:
5219 . mat - the matrix
5220 
5221   Level: intermediate
5222 
5223   Notes:
5224   The `i`, `j`, and `a` arrays are not copied by this routine, the user must free these arrays
5225   once the matrix is destroyed and not before
5226 
5227   You cannot set new nonzero locations into this matrix, that will generate an error.
5228 
5229   The `i` and `j` indices are 0 based
5230 
5231   The format which is used for the sparse matrix input, is equivalent to a
5232   row-major ordering.. i.e for the following matrix, the input data expected is
5233   as shown
5234 .vb
5235         1 0 0
5236         2 0 3
5237         4 5 6
5238 
5239         i =  {0,1,3,6}  [size = nrow+1  = 3+1]
5240         j =  {0,0,2,0,1,2}  [size = 6]; values must be sorted for each row
5241         v =  {1,2,3,4,5,6}  [size = 6]
5242 .ve
5243 
5244 .seealso: [](ch_matrices), `Mat`, `MatCreate()`, `MatCreateAIJ()`, `MatCreateSeqAIJ()`, `MatCreateMPIAIJWithArrays()`, `MatMPIAIJSetPreallocationCSR()`
5245 @*/
5246 PetscErrorCode MatCreateSeqAIJWithArrays(MPI_Comm comm, PetscInt m, PetscInt n, PetscInt i[], PetscInt j[], PetscScalar a[], Mat *mat)
5247 {
5248   PetscInt    ii;
5249   Mat_SeqAIJ *aij;
5250   PetscInt    jj;
5251 
5252   PetscFunctionBegin;
5253   PetscCheck(m <= 0 || i[0] == 0, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "i (row indices) must start with 0");
5254   PetscCall(MatCreate(comm, mat));
5255   PetscCall(MatSetSizes(*mat, m, n, m, n));
5256   /* PetscCall(MatSetBlockSizes(*mat,,)); */
5257   PetscCall(MatSetType(*mat, MATSEQAIJ));
5258   PetscCall(MatSeqAIJSetPreallocation_SeqAIJ(*mat, MAT_SKIP_ALLOCATION, NULL));
5259   aij = (Mat_SeqAIJ *)(*mat)->data;
5260   PetscCall(PetscMalloc1(m, &aij->imax));
5261   PetscCall(PetscMalloc1(m, &aij->ilen));
5262 
5263   aij->i            = i;
5264   aij->j            = j;
5265   aij->a            = a;
5266   aij->singlemalloc = PETSC_FALSE;
5267   aij->nonew        = -1; /*this indicates that inserting a new value in the matrix that generates a new nonzero is an error*/
5268   aij->free_a       = PETSC_FALSE;
5269   aij->free_ij      = PETSC_FALSE;
5270 
5271   for (ii = 0, aij->nonzerorowcnt = 0, aij->rmax = 0; ii < m; ii++) {
5272     aij->ilen[ii] = aij->imax[ii] = i[ii + 1] - i[ii];
5273     if (PetscDefined(USE_DEBUG)) {
5274       PetscCheck(i[ii + 1] - i[ii] >= 0, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Negative row length in i (row indices) row = %" PetscInt_FMT " length = %" PetscInt_FMT, ii, i[ii + 1] - i[ii]);
5275       for (jj = i[ii] + 1; jj < i[ii + 1]; jj++) {
5276         PetscCheck(j[jj] >= j[jj - 1], PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Column entry number %" PetscInt_FMT " (actual column %" PetscInt_FMT ") in row %" PetscInt_FMT " is not sorted", jj - i[ii], j[jj], ii);
5277         PetscCheck(j[jj] != j[jj - 1], PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Column entry number %" PetscInt_FMT " (actual column %" PetscInt_FMT ") in row %" PetscInt_FMT " is identical to previous entry", jj - i[ii], j[jj], ii);
5278       }
5279     }
5280   }
5281   if (PetscDefined(USE_DEBUG)) {
5282     for (ii = 0; ii < aij->i[m]; ii++) {
5283       PetscCheck(j[ii] >= 0, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Negative column index at location = %" PetscInt_FMT " index = %" PetscInt_FMT, ii, j[ii]);
5284       PetscCheck(j[ii] <= n - 1, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Column index to large at location = %" PetscInt_FMT " index = %" PetscInt_FMT " last column = %" PetscInt_FMT, ii, j[ii], n - 1);
5285     }
5286   }
5287 
5288   PetscCall(MatAssemblyBegin(*mat, MAT_FINAL_ASSEMBLY));
5289   PetscCall(MatAssemblyEnd(*mat, MAT_FINAL_ASSEMBLY));
5290   PetscFunctionReturn(PETSC_SUCCESS);
5291 }
5292 
5293 /*@
5294   MatCreateSeqAIJFromTriple - Creates an sequential `MATSEQAIJ` matrix using matrix elements (in COO format)
5295   provided by the user.
5296 
5297   Collective
5298 
5299   Input Parameters:
5300 + comm - must be an MPI communicator of size 1
5301 . m    - number of rows
5302 . n    - number of columns
5303 . i    - row indices
5304 . j    - column indices
5305 . a    - matrix values
5306 . nz   - number of nonzeros
5307 - idx  - if the `i` and `j` indices start with 1 use `PETSC_TRUE` otherwise use `PETSC_FALSE`
5308 
5309   Output Parameter:
5310 . mat - the matrix
5311 
5312   Level: intermediate
5313 
5314   Example:
5315   For the following matrix, the input data expected is as shown (using 0 based indexing)
5316 .vb
5317         1 0 0
5318         2 0 3
5319         4 5 6
5320 
5321         i =  {0,1,1,2,2,2}
5322         j =  {0,0,2,0,1,2}
5323         v =  {1,2,3,4,5,6}
5324 .ve
5325 
5326   Note:
5327   Instead of using this function, users should also consider `MatSetPreallocationCOO()` and `MatSetValuesCOO()`, which allow repeated or remote entries,
5328   and are particularly useful in iterative applications.
5329 
5330 .seealso: [](ch_matrices), `Mat`, `MatCreate()`, `MatCreateAIJ()`, `MatCreateSeqAIJ()`, `MatCreateSeqAIJWithArrays()`, `MatMPIAIJSetPreallocationCSR()`, `MatSetValuesCOO()`, `MatSetPreallocationCOO()`
5331 @*/
5332 PetscErrorCode MatCreateSeqAIJFromTriple(MPI_Comm comm, PetscInt m, PetscInt n, PetscInt i[], PetscInt j[], PetscScalar a[], Mat *mat, PetscInt nz, PetscBool idx)
5333 {
5334   PetscInt ii, *nnz, one = 1, row, col;
5335 
5336   PetscFunctionBegin;
5337   PetscCall(PetscCalloc1(m, &nnz));
5338   for (ii = 0; ii < nz; ii++) nnz[i[ii] - !!idx] += 1;
5339   PetscCall(MatCreate(comm, mat));
5340   PetscCall(MatSetSizes(*mat, m, n, m, n));
5341   PetscCall(MatSetType(*mat, MATSEQAIJ));
5342   PetscCall(MatSeqAIJSetPreallocation_SeqAIJ(*mat, 0, nnz));
5343   for (ii = 0; ii < nz; ii++) {
5344     if (idx) {
5345       row = i[ii] - 1;
5346       col = j[ii] - 1;
5347     } else {
5348       row = i[ii];
5349       col = j[ii];
5350     }
5351     PetscCall(MatSetValues(*mat, one, &row, one, &col, &a[ii], ADD_VALUES));
5352   }
5353   PetscCall(MatAssemblyBegin(*mat, MAT_FINAL_ASSEMBLY));
5354   PetscCall(MatAssemblyEnd(*mat, MAT_FINAL_ASSEMBLY));
5355   PetscCall(PetscFree(nnz));
5356   PetscFunctionReturn(PETSC_SUCCESS);
5357 }
5358 
5359 PetscErrorCode MatSeqAIJInvalidateDiagonal(Mat A)
5360 {
5361   Mat_SeqAIJ *a = (Mat_SeqAIJ *)A->data;
5362 
5363   PetscFunctionBegin;
5364   a->idiagvalid  = PETSC_FALSE;
5365   a->ibdiagvalid = PETSC_FALSE;
5366 
5367   PetscCall(MatSeqAIJInvalidateDiagonal_Inode(A));
5368   PetscFunctionReturn(PETSC_SUCCESS);
5369 }
5370 
5371 PetscErrorCode MatCreateMPIMatConcatenateSeqMat_SeqAIJ(MPI_Comm comm, Mat inmat, PetscInt n, MatReuse scall, Mat *outmat)
5372 {
5373   PetscFunctionBegin;
5374   PetscCall(MatCreateMPIMatConcatenateSeqMat_MPIAIJ(comm, inmat, n, scall, outmat));
5375   PetscFunctionReturn(PETSC_SUCCESS);
5376 }
5377 
5378 /*
5379  Permute A into C's *local* index space using rowemb,colemb.
5380  The embedding are supposed to be injections and the above implies that the range of rowemb is a subset
5381  of [0,m), colemb is in [0,n).
5382  If pattern == DIFFERENT_NONZERO_PATTERN, C is preallocated according to A.
5383  */
5384 PetscErrorCode MatSetSeqMat_SeqAIJ(Mat C, IS rowemb, IS colemb, MatStructure pattern, Mat B)
5385 {
5386   /* If making this function public, change the error returned in this function away from _PLIB. */
5387   Mat_SeqAIJ     *Baij;
5388   PetscBool       seqaij;
5389   PetscInt        m, n, *nz, i, j, count;
5390   PetscScalar     v;
5391   const PetscInt *rowindices, *colindices;
5392 
5393   PetscFunctionBegin;
5394   if (!B) PetscFunctionReturn(PETSC_SUCCESS);
5395   /* Check to make sure the target matrix (and embeddings) are compatible with C and each other. */
5396   PetscCall(PetscObjectBaseTypeCompare((PetscObject)B, MATSEQAIJ, &seqaij));
5397   PetscCheck(seqaij, PETSC_COMM_SELF, PETSC_ERR_PLIB, "Input matrix is of wrong type");
5398   if (rowemb) {
5399     PetscCall(ISGetLocalSize(rowemb, &m));
5400     PetscCheck(m == B->rmap->n, PETSC_COMM_SELF, PETSC_ERR_PLIB, "Row IS of size %" PetscInt_FMT " is incompatible with matrix row size %" PetscInt_FMT, m, B->rmap->n);
5401   } else {
5402     PetscCheck(C->rmap->n == B->rmap->n, PETSC_COMM_SELF, PETSC_ERR_PLIB, "Input matrix is row-incompatible with the target matrix");
5403   }
5404   if (colemb) {
5405     PetscCall(ISGetLocalSize(colemb, &n));
5406     PetscCheck(n == B->cmap->n, PETSC_COMM_SELF, PETSC_ERR_PLIB, "Diag col IS of size %" PetscInt_FMT " is incompatible with input matrix col size %" PetscInt_FMT, n, B->cmap->n);
5407   } else {
5408     PetscCheck(C->cmap->n == B->cmap->n, PETSC_COMM_SELF, PETSC_ERR_PLIB, "Input matrix is col-incompatible with the target matrix");
5409   }
5410 
5411   Baij = (Mat_SeqAIJ *)B->data;
5412   if (pattern == DIFFERENT_NONZERO_PATTERN) {
5413     PetscCall(PetscMalloc1(B->rmap->n, &nz));
5414     for (i = 0; i < B->rmap->n; i++) nz[i] = Baij->i[i + 1] - Baij->i[i];
5415     PetscCall(MatSeqAIJSetPreallocation(C, 0, nz));
5416     PetscCall(PetscFree(nz));
5417   }
5418   if (pattern == SUBSET_NONZERO_PATTERN) PetscCall(MatZeroEntries(C));
5419   count      = 0;
5420   rowindices = NULL;
5421   colindices = NULL;
5422   if (rowemb) PetscCall(ISGetIndices(rowemb, &rowindices));
5423   if (colemb) PetscCall(ISGetIndices(colemb, &colindices));
5424   for (i = 0; i < B->rmap->n; i++) {
5425     PetscInt row;
5426     row = i;
5427     if (rowindices) row = rowindices[i];
5428     for (j = Baij->i[i]; j < Baij->i[i + 1]; j++) {
5429       PetscInt col;
5430       col = Baij->j[count];
5431       if (colindices) col = colindices[col];
5432       v = Baij->a[count];
5433       PetscCall(MatSetValues(C, 1, &row, 1, &col, &v, INSERT_VALUES));
5434       ++count;
5435     }
5436   }
5437   /* FIXME: set C's nonzerostate correctly. */
5438   /* Assembly for C is necessary. */
5439   C->preallocated  = PETSC_TRUE;
5440   C->assembled     = PETSC_TRUE;
5441   C->was_assembled = PETSC_FALSE;
5442   PetscFunctionReturn(PETSC_SUCCESS);
5443 }
5444 
5445 PetscErrorCode MatEliminateZeros_SeqAIJ(Mat A, PetscBool keep)
5446 {
5447   Mat_SeqAIJ *a  = (Mat_SeqAIJ *)A->data;
5448   MatScalar  *aa = a->a;
5449   PetscInt    m = A->rmap->n, fshift = 0, fshift_prev = 0, i, k;
5450   PetscInt   *ailen = a->ilen, *imax = a->imax, *ai = a->i, *aj = a->j, rmax = 0;
5451 
5452   PetscFunctionBegin;
5453   PetscCheck(A->assembled, PETSC_COMM_SELF, PETSC_ERR_ARG_WRONGSTATE, "Cannot eliminate zeros for unassembled matrix");
5454   if (m) rmax = ailen[0]; /* determine row with most nonzeros */
5455   for (i = 1; i <= m; i++) {
5456     /* move each nonzero entry back by the amount of zero slots (fshift) before it*/
5457     for (k = ai[i - 1]; k < ai[i]; k++) {
5458       if (aa[k] == 0 && (aj[k] != i - 1 || !keep)) fshift++;
5459       else {
5460         if (aa[k] == 0 && aj[k] == i - 1) PetscCall(PetscInfo(A, "Keep the diagonal zero at row %" PetscInt_FMT "\n", i - 1));
5461         aa[k - fshift] = aa[k];
5462         aj[k - fshift] = aj[k];
5463       }
5464     }
5465     ai[i - 1] -= fshift_prev; // safe to update ai[i-1] now since it will not be used in the next iteration
5466     fshift_prev = fshift;
5467     /* reset ilen and imax for each row */
5468     ailen[i - 1] = imax[i - 1] = ai[i] - fshift - ai[i - 1];
5469     a->nonzerorowcnt += ((ai[i] - fshift - ai[i - 1]) > 0);
5470     rmax = PetscMax(rmax, ailen[i - 1]);
5471   }
5472   if (fshift) {
5473     if (m) {
5474       ai[m] -= fshift;
5475       a->nz = ai[m];
5476     }
5477     PetscCall(PetscInfo(A, "Matrix size: %" PetscInt_FMT " X %" PetscInt_FMT "; zeros eliminated: %" PetscInt_FMT "; nonzeros left: %" PetscInt_FMT "\n", m, A->cmap->n, fshift, a->nz));
5478     A->nonzerostate++;
5479     A->info.nz_unneeded += (PetscReal)fshift;
5480     a->rmax = rmax;
5481     if (a->inode.use && a->inode.checked) PetscCall(MatSeqAIJCheckInode(A));
5482     PetscCall(MatAssemblyBegin(A, MAT_FINAL_ASSEMBLY));
5483     PetscCall(MatAssemblyEnd(A, MAT_FINAL_ASSEMBLY));
5484   }
5485   PetscFunctionReturn(PETSC_SUCCESS);
5486 }
5487 
5488 PetscFunctionList MatSeqAIJList = NULL;
5489 
5490 /*@C
5491   MatSeqAIJSetType - Converts a `MATSEQAIJ` matrix to a subtype
5492 
5493   Collective
5494 
5495   Input Parameters:
5496 + mat    - the matrix object
5497 - matype - matrix type
5498 
5499   Options Database Key:
5500 . -mat_seqaij_type  <method> - for example seqaijcrl
5501 
5502   Level: intermediate
5503 
5504 .seealso: [](ch_matrices), `Mat`, `PCSetType()`, `VecSetType()`, `MatCreate()`, `MatType`
5505 @*/
5506 PetscErrorCode MatSeqAIJSetType(Mat mat, MatType matype)
5507 {
5508   PetscBool sametype;
5509   PetscErrorCode (*r)(Mat, MatType, MatReuse, Mat *);
5510 
5511   PetscFunctionBegin;
5512   PetscValidHeaderSpecific(mat, MAT_CLASSID, 1);
5513   PetscCall(PetscObjectTypeCompare((PetscObject)mat, matype, &sametype));
5514   if (sametype) PetscFunctionReturn(PETSC_SUCCESS);
5515 
5516   PetscCall(PetscFunctionListFind(MatSeqAIJList, matype, &r));
5517   PetscCheck(r, PetscObjectComm((PetscObject)mat), PETSC_ERR_ARG_UNKNOWN_TYPE, "Unknown Mat type given: %s", matype);
5518   PetscCall((*r)(mat, matype, MAT_INPLACE_MATRIX, &mat));
5519   PetscFunctionReturn(PETSC_SUCCESS);
5520 }
5521 
5522 /*@C
5523   MatSeqAIJRegister -  - Adds a new sub-matrix type for sequential `MATSEQAIJ` matrices
5524 
5525   Not Collective
5526 
5527   Input Parameters:
5528 + sname    - name of a new user-defined matrix type, for example `MATSEQAIJCRL`
5529 - function - routine to convert to subtype
5530 
5531   Level: advanced
5532 
5533   Notes:
5534   `MatSeqAIJRegister()` may be called multiple times to add several user-defined solvers.
5535 
5536   Then, your matrix can be chosen with the procedural interface at runtime via the option
5537 $     -mat_seqaij_type my_mat
5538 
5539 .seealso: [](ch_matrices), `Mat`, `MatSeqAIJRegisterAll()`
5540 @*/
5541 PetscErrorCode MatSeqAIJRegister(const char sname[], PetscErrorCode (*function)(Mat, MatType, MatReuse, Mat *))
5542 {
5543   PetscFunctionBegin;
5544   PetscCall(MatInitializePackage());
5545   PetscCall(PetscFunctionListAdd(&MatSeqAIJList, sname, function));
5546   PetscFunctionReturn(PETSC_SUCCESS);
5547 }
5548 
5549 PetscBool MatSeqAIJRegisterAllCalled = PETSC_FALSE;
5550 
5551 /*@C
5552   MatSeqAIJRegisterAll - Registers all of the matrix subtypes of `MATSSEQAIJ`
5553 
5554   Not Collective
5555 
5556   Level: advanced
5557 
5558   Note:
5559   This registers the versions of `MATSEQAIJ` for GPUs
5560 
5561 .seealso: [](ch_matrices), `Mat`, `MatRegisterAll()`, `MatSeqAIJRegister()`
5562 @*/
5563 PetscErrorCode MatSeqAIJRegisterAll(void)
5564 {
5565   PetscFunctionBegin;
5566   if (MatSeqAIJRegisterAllCalled) PetscFunctionReturn(PETSC_SUCCESS);
5567   MatSeqAIJRegisterAllCalled = PETSC_TRUE;
5568 
5569   PetscCall(MatSeqAIJRegister(MATSEQAIJCRL, MatConvert_SeqAIJ_SeqAIJCRL));
5570   PetscCall(MatSeqAIJRegister(MATSEQAIJPERM, MatConvert_SeqAIJ_SeqAIJPERM));
5571   PetscCall(MatSeqAIJRegister(MATSEQAIJSELL, MatConvert_SeqAIJ_SeqAIJSELL));
5572 #if defined(PETSC_HAVE_MKL_SPARSE)
5573   PetscCall(MatSeqAIJRegister(MATSEQAIJMKL, MatConvert_SeqAIJ_SeqAIJMKL));
5574 #endif
5575 #if defined(PETSC_HAVE_CUDA)
5576   PetscCall(MatSeqAIJRegister(MATSEQAIJCUSPARSE, MatConvert_SeqAIJ_SeqAIJCUSPARSE));
5577 #endif
5578 #if defined(PETSC_HAVE_HIP)
5579   PetscCall(MatSeqAIJRegister(MATSEQAIJHIPSPARSE, MatConvert_SeqAIJ_SeqAIJHIPSPARSE));
5580 #endif
5581 #if defined(PETSC_HAVE_KOKKOS_KERNELS)
5582   PetscCall(MatSeqAIJRegister(MATSEQAIJKOKKOS, MatConvert_SeqAIJ_SeqAIJKokkos));
5583 #endif
5584 #if defined(PETSC_HAVE_VIENNACL) && defined(PETSC_HAVE_VIENNACL_NO_CUDA)
5585   PetscCall(MatSeqAIJRegister(MATMPIAIJVIENNACL, MatConvert_SeqAIJ_SeqAIJViennaCL));
5586 #endif
5587   PetscFunctionReturn(PETSC_SUCCESS);
5588 }
5589 
5590 /*
5591     Special version for direct calls from Fortran
5592 */
5593 #if defined(PETSC_HAVE_FORTRAN_CAPS)
5594   #define matsetvaluesseqaij_ MATSETVALUESSEQAIJ
5595 #elif !defined(PETSC_HAVE_FORTRAN_UNDERSCORE)
5596   #define matsetvaluesseqaij_ matsetvaluesseqaij
5597 #endif
5598 
5599 /* Change these macros so can be used in void function */
5600 
5601 /* Change these macros so can be used in void function */
5602 /* Identical to PetscCallVoid, except it assigns to *_ierr */
5603 #undef PetscCall
5604 #define PetscCall(...) \
5605   do { \
5606     PetscErrorCode ierr_msv_mpiaij = __VA_ARGS__; \
5607     if (PetscUnlikely(ierr_msv_mpiaij)) { \
5608       *_ierr = PetscError(PETSC_COMM_SELF, __LINE__, PETSC_FUNCTION_NAME, __FILE__, ierr_msv_mpiaij, PETSC_ERROR_REPEAT, " "); \
5609       return; \
5610     } \
5611   } while (0)
5612 
5613 #undef SETERRQ
5614 #define SETERRQ(comm, ierr, ...) \
5615   do { \
5616     *_ierr = PetscError(comm, __LINE__, PETSC_FUNCTION_NAME, __FILE__, ierr, PETSC_ERROR_INITIAL, __VA_ARGS__); \
5617     return; \
5618   } while (0)
5619 
5620 PETSC_EXTERN void matsetvaluesseqaij_(Mat *AA, PetscInt *mm, const PetscInt im[], PetscInt *nn, const PetscInt in[], const PetscScalar v[], InsertMode *isis, PetscErrorCode *_ierr)
5621 {
5622   Mat         A = *AA;
5623   PetscInt    m = *mm, n = *nn;
5624   InsertMode  is = *isis;
5625   Mat_SeqAIJ *a  = (Mat_SeqAIJ *)A->data;
5626   PetscInt   *rp, k, low, high, t, ii, row, nrow, i, col, l, rmax, N;
5627   PetscInt   *imax, *ai, *ailen;
5628   PetscInt   *aj, nonew = a->nonew, lastcol = -1;
5629   MatScalar  *ap, value, *aa;
5630   PetscBool   ignorezeroentries = a->ignorezeroentries;
5631   PetscBool   roworiented       = a->roworiented;
5632 
5633   PetscFunctionBegin;
5634   MatCheckPreallocated(A, 1);
5635   imax  = a->imax;
5636   ai    = a->i;
5637   ailen = a->ilen;
5638   aj    = a->j;
5639   aa    = a->a;
5640 
5641   for (k = 0; k < m; k++) { /* loop over added rows */
5642     row = im[k];
5643     if (row < 0) continue;
5644     PetscCheck(row < A->rmap->n, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_OUTOFRANGE, "Row too large");
5645     rp   = aj + ai[row];
5646     ap   = aa + ai[row];
5647     rmax = imax[row];
5648     nrow = ailen[row];
5649     low  = 0;
5650     high = nrow;
5651     for (l = 0; l < n; l++) { /* loop over added columns */
5652       if (in[l] < 0) continue;
5653       PetscCheck(in[l] < A->cmap->n, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_OUTOFRANGE, "Column too large");
5654       col = in[l];
5655       if (roworiented) value = v[l + k * n];
5656       else value = v[k + l * m];
5657 
5658       if (value == 0.0 && ignorezeroentries && (is == ADD_VALUES)) continue;
5659 
5660       if (col <= lastcol) low = 0;
5661       else high = nrow;
5662       lastcol = col;
5663       while (high - low > 5) {
5664         t = (low + high) / 2;
5665         if (rp[t] > col) high = t;
5666         else low = t;
5667       }
5668       for (i = low; i < high; i++) {
5669         if (rp[i] > col) break;
5670         if (rp[i] == col) {
5671           if (is == ADD_VALUES) ap[i] += value;
5672           else ap[i] = value;
5673           goto noinsert;
5674         }
5675       }
5676       if (value == 0.0 && ignorezeroentries) goto noinsert;
5677       if (nonew == 1) goto noinsert;
5678       PetscCheck(nonew != -1, PetscObjectComm((PetscObject)A), PETSC_ERR_ARG_OUTOFRANGE, "Inserting a new nonzero in the matrix");
5679       MatSeqXAIJReallocateAIJ(A, A->rmap->n, 1, nrow, row, col, rmax, aa, ai, aj, rp, ap, imax, nonew, MatScalar);
5680       N = nrow++ - 1;
5681       a->nz++;
5682       high++;
5683       /* shift up all the later entries in this row */
5684       for (ii = N; ii >= i; ii--) {
5685         rp[ii + 1] = rp[ii];
5686         ap[ii + 1] = ap[ii];
5687       }
5688       rp[i] = col;
5689       ap[i] = value;
5690     noinsert:;
5691       low = i + 1;
5692     }
5693     ailen[row] = nrow;
5694   }
5695   PetscFunctionReturnVoid();
5696 }
5697 /* Undefining these here since they were redefined from their original definition above! No
5698  * other PETSc functions should be defined past this point, as it is impossible to recover the
5699  * original definitions */
5700 #undef PetscCall
5701 #undef SETERRQ
5702