Actual source code: aij.h
1: #pragma once
3: #include <petsc/private/matimpl.h>
4: #include <petsc/private/hashmapi.h>
5: #include <petsc/private/hashmapijv.h>
7: /*
8: Used by MatCreateSubMatrices_MPIXAIJ_Local()
9: */
10: typedef struct { /* used by MatCreateSubMatrices_MPIAIJ_SingleIS_Local() and MatCreateSubMatrices_MPIAIJ_Local */
11: PetscInt id; /* index of submats, only submats[0] is responsible for deleting some arrays below */
12: PetscMPIInt nrqs, nrqr;
13: PetscInt **rbuf1, **rbuf2, **rbuf3, **sbuf1, **sbuf2;
14: PetscInt **ptr;
15: PetscInt *tmp;
16: PetscInt *ctr;
17: PetscMPIInt *pa; /* process array */
18: PetscInt *req_size;
19: PetscMPIInt *req_source1, *req_source2;
20: PetscBool allcolumns, allrows;
21: PetscBool singleis;
22: PetscMPIInt *row2proc; /* row to process (MPI rank) map */
23: PetscInt nstages;
24: #if defined(PETSC_USE_CTABLE)
25: PetscHMapI cmap, rmap;
26: PetscInt *cmap_loc, *rmap_loc;
27: #else
28: PetscInt *cmap, *rmap;
29: #endif
30: PetscErrorCode (*destroy)(Mat);
31: } Mat_SubSppt;
33: /* Operations provided by MATSEQAIJ and its subclasses */
34: typedef struct {
35: PetscErrorCode (*getarray)(Mat, PetscScalar **);
36: PetscErrorCode (*restorearray)(Mat, PetscScalar **);
37: PetscErrorCode (*getarrayread)(Mat, const PetscScalar **);
38: PetscErrorCode (*restorearrayread)(Mat, const PetscScalar **);
39: PetscErrorCode (*getarraywrite)(Mat, PetscScalar **);
40: PetscErrorCode (*restorearraywrite)(Mat, PetscScalar **);
41: PetscErrorCode (*getcsrandmemtype)(Mat, const PetscInt **, const PetscInt **, PetscScalar **, PetscMemType *);
42: } Mat_SeqAIJOps;
44: /*
45: Struct header shared by SeqAIJ, SeqBAIJ and SeqSBAIJ matrix formats
46: */
47: #define SEQAIJHEADER(datatype) \
48: PetscBool roworiented; /* if true, row-oriented input, default */ \
49: PetscInt nonew; /* 1 don't add new nonzeros, -1 generate error on new */ \
50: PetscInt nounused; /* -1 generate error on unused space */ \
51: PetscInt maxnz; /* allocated nonzeros */ \
52: PetscInt *imax; /* maximum space allocated for each row */ \
53: PetscInt *ilen; /* actual length of each row */ \
54: PetscInt *ipre; /* space preallocated for each row by user */ \
55: PetscBool free_imax_ilen; \
56: PetscInt reallocs; /* number of mallocs done during MatSetValues() \
57: as more values are set than were prealloced */ \
58: PetscInt rmax; /* max nonzeros in any row */ \
59: PetscBool keepnonzeropattern; /* keeps matrix nonzero structure same in calls to MatZeroRows()*/ \
60: PetscBool ignorezeroentries; \
61: PetscBool free_ij; /* free the column indices j and row offsets i when the matrix is destroyed */ \
62: PetscBool free_a; /* free the numerical values when matrix is destroy */ \
63: Mat_CompressedRow compressedrow; /* use compressed row format */ \
64: PetscInt nz; /* nonzeros */ \
65: PetscInt *i; /* pointer to beginning of each row */ \
66: PetscInt *j; /* column values: j + i[k] - 1 is start of row k */ \
67: PetscInt *diag; /* pointers to diagonal elements */ \
68: PetscObjectState diagNonzeroState; /* nonzero state of the matrix when diag was obtained */ \
69: PetscBool diagDense; /* all entries along the diagonal have been set; i.e. no missing diagonal terms */ \
70: PetscInt nonzerorowcnt; /* how many rows have nonzero entries */ \
71: datatype *a; /* nonzero elements */ \
72: PetscScalar *solve_work; /* work space used in MatSolve */ \
73: IS row, col, icol; /* index sets, used for reorderings */ \
74: PetscBool pivotinblocks; /* pivot inside factorization of each diagonal block */ \
75: Mat parent; /* set if this matrix was formed with MatDuplicate(...,MAT_SHARE_NONZERO_PATTERN,....); \
76: means that this shares some data structures with the parent including diag, ilen, imax, i, j */ \
77: Mat_SubSppt *submatis1; /* used by MatCreateSubMatrices_MPIXAIJ_Local */ \
78: Mat_SeqAIJOps ops[1] /* operations for SeqAIJ and its subclasses */
80: typedef struct {
81: MatTransposeColoring matcoloring;
82: Mat Bt_den; /* dense matrix of B^T */
83: Mat ABt_den; /* dense matrix of A*B^T */
84: PetscBool usecoloring;
85: } MatProductCtx_MatMatTransMult;
87: typedef struct { /* used by MatTransposeMatMult() */
88: Mat At; /* transpose of the first matrix */
89: Mat mA; /* maij matrix of A */
90: Vec bt, ct; /* vectors to hold locally transposed arrays of B and C */
91: /* used by PtAP */
92: void *data;
93: PetscCtxDestroyFn *destroy;
94: } MatProductCtx_MatTransMatMult;
96: typedef struct {
97: PetscInt *api, *apj; /* symbolic structure of A*P */
98: PetscScalar *apa; /* temporary array for storing one row of A*P */
99: } MatProductCtx_AP;
101: typedef struct {
102: MatTransposeColoring matcoloring;
103: Mat Rt; /* sparse or dense matrix of R^T */
104: Mat RARt; /* dense matrix of R*A*R^T */
105: Mat ARt; /* A*R^T used for the case -matrart_color_art */
106: MatScalar *work; /* work array to store columns of A*R^T used in MatMatMatMultNumeric_SeqAIJ_SeqAIJ_SeqDense() */
107: /* free intermediate products needed for PtAP */
108: void *data;
109: PetscCtxDestroyFn *destroy;
110: } MatProductCtx_RARt;
112: typedef struct {
113: Mat BC; /* temp matrix for storing B*C */
114: } MatProductCtx_MatMatMatMult;
116: /*
117: MATSEQAIJ format - Compressed row storage (also called Yale sparse matrix
118: format) or compressed sparse row (CSR). The i[] and j[] arrays start at 0. For example,
119: j[i[k]+p] is the pth column in row k. Note that the diagonal
120: matrix elements are stored with the rest of the nonzeros (not separately).
121: */
123: /* Info about i-nodes (identical nodes) helper class for SeqAIJ */
124: typedef struct {
125: /* data for MatSOR_SeqAIJ_Inode() */
126: MatScalar *bdiag, *ibdiag, *ssor_work; /* diagonal blocks of matrices */
127: PetscInt bdiagsize; /* length of bdiag and ibdiag */
128: PetscObjectState ibdiagState; /* state of the matrix when ibdiag[] and bdiag[] were constructed */
130: PetscBool use;
131: PetscInt node_count; /* number of inodes */
132: PetscInt *size_csr; /* inode sizes in csr with size_csr[0] = 0 and i-th node size = size_csr[i+1] - size_csr[i], to facilitate parallel computation */
133: PetscInt limit; /* inode limit */
134: PetscInt max_limit; /* maximum supported inode limit */
135: PetscBool checked; /* if inodes have been checked for */
136: PetscObjectState mat_nonzerostate; /* non-zero state when inodes were checked for */
137: } Mat_SeqAIJ_Inode;
139: PETSC_INTERN PetscErrorCode MatView_SeqAIJ_Inode(Mat, PetscViewer);
140: PETSC_INTERN PetscErrorCode MatAssemblyEnd_SeqAIJ_Inode(Mat, MatAssemblyType);
141: PETSC_INTERN PetscErrorCode MatDestroy_SeqAIJ_Inode(Mat);
142: PETSC_INTERN PetscErrorCode MatCreate_SeqAIJ_Inode(Mat);
143: PETSC_INTERN PetscErrorCode MatSetOption_SeqAIJ_Inode(Mat, MatOption, PetscBool);
144: PETSC_INTERN PetscErrorCode MatDuplicate_SeqAIJ_Inode(Mat, MatDuplicateOption, Mat *);
145: PETSC_INTERN PetscErrorCode MatDuplicateNoCreate_SeqAIJ(Mat, Mat, MatDuplicateOption, PetscBool);
146: PETSC_INTERN PetscErrorCode MatLUFactorNumeric_SeqAIJ_Inode(Mat, Mat, const MatFactorInfo *);
147: PETSC_INTERN PetscErrorCode MatSeqAIJGetArray_SeqAIJ(Mat, PetscScalar **);
148: PETSC_INTERN PetscErrorCode MatSeqAIJRestoreArray_SeqAIJ(Mat, PetscScalar **);
150: typedef struct {
151: SEQAIJHEADER(MatScalar);
152: Mat_SeqAIJ_Inode inode;
153: MatScalar *saved_values; /* location for stashing nonzero values of matrix */
155: /* data needed for MatSOR_SeqAIJ() */
156: PetscScalar *mdiag, *idiag; /* diagonal values, inverse of diagonal entries */
157: PetscScalar *ssor_work; /* workspace for Eisenstat trick */
158: PetscObjectState idiagState; /* state of the matrix when mdiag and idiag was obtained */
159: PetscScalar fshift, omega; /* last used omega and fshift */
161: PetscScalar *ibdiag; /* inverses of block diagonals */
162: PetscBool ibdiagvalid; /* inverses of block diagonals are valid. */
164: /* MatSetValues() via hash related fields */
165: PetscHMapIJV ht;
166: PetscInt *dnz;
167: struct _MatOps cops;
168: } Mat_SeqAIJ;
170: typedef struct {
171: PetscInt nz; /* nz of the matrix after assembly */
172: PetscCount n; /* Number of entries in MatSetPreallocationCOO() */
173: PetscCount Atot; /* Total number of valid (i.e., w/ non-negative indices) entries in the COO array */
174: PetscCount *jmap; /* perm[jmap[i]..jmap[i+1]) give indices of entries in v[] associated with i-th nonzero of the matrix */
175: PetscCount *perm; /* The permutation array in sorting (i,j) by row and then by col */
176: } MatCOOStruct_SeqAIJ;
178: #define MatSeqXAIJGetOptions_Private(A) \
179: { \
180: const PetscBool oldvalues = (PetscBool)(A != PETSC_NULLPTR); \
181: PetscInt nonew = 0, nounused = 0; \
182: PetscBool roworiented = PETSC_FALSE; \
183: if (oldvalues) { \
184: nonew = ((Mat_SeqAIJ *)A->data)->nonew; \
185: nounused = ((Mat_SeqAIJ *)A->data)->nounused; \
186: roworiented = ((Mat_SeqAIJ *)A->data)->roworiented; \
187: } \
188: (void)0
190: #define MatSeqXAIJRestoreOptions_Private(A) \
191: if (oldvalues) { \
192: ((Mat_SeqAIJ *)A->data)->nonew = nonew; \
193: ((Mat_SeqAIJ *)A->data)->nounused = nounused; \
194: ((Mat_SeqAIJ *)A->data)->roworiented = roworiented; \
195: } \
196: } \
197: (void)0
199: static inline PetscErrorCode MatXAIJAllocatea(Mat A, PetscInt nz, PetscScalar **array)
200: {
201: Mat_SeqAIJ *a = (Mat_SeqAIJ *)A->data;
203: PetscFunctionBegin;
204: PetscCall(PetscShmgetAllocateArray(nz, sizeof(PetscScalar), (void **)array));
205: a->free_a = PETSC_TRUE;
206: PetscFunctionReturn(PETSC_SUCCESS);
207: }
209: static inline PetscErrorCode MatXAIJDeallocatea(Mat A, PetscScalar **array)
210: {
211: Mat_SeqAIJ *a = (Mat_SeqAIJ *)A->data;
213: PetscFunctionBegin;
214: if (a->free_a) PetscCall(PetscShmgetDeallocateArray((void **)array));
215: a->free_a = PETSC_FALSE;
216: PetscFunctionReturn(PETSC_SUCCESS);
217: }
219: /*
220: Frees the a, i, and j arrays from the XAIJ (AIJ, BAIJ, and SBAIJ) matrix types
221: */
222: static inline PetscErrorCode MatSeqXAIJFreeAIJ(Mat AA, MatScalar **a, PetscInt **j, PetscInt **i)
223: {
224: Mat_SeqAIJ *A = (Mat_SeqAIJ *)AA->data;
226: PetscFunctionBegin;
227: if (A->free_a) PetscCall(PetscShmgetDeallocateArray((void **)a));
228: if (A->free_ij) PetscCall(PetscShmgetDeallocateArray((void **)j));
229: if (A->free_ij) PetscCall(PetscShmgetDeallocateArray((void **)i));
230: PetscFunctionReturn(PETSC_SUCCESS);
231: }
232: /*
233: Allocates larger a, i, and j arrays for the XAIJ (AIJ, BAIJ, and SBAIJ) matrix types
234: This is a macro because it takes the datatype as an argument which can be either a Mat or a MatScalar
235: */
236: #define MatSeqXAIJReallocateAIJ(Amat, AM, BS2, NROW, ROW, COL, RMAX, AA, AI, AJ, RP, AP, AIMAX, NONEW, datatype) \
237: do { \
238: if (NROW >= RMAX) { \
239: Mat_SeqAIJ *Ain = (Mat_SeqAIJ *)Amat->data; \
240: PetscInt CHUNKSIZE = 15, new_nz = AI[AM] + CHUNKSIZE, len, *new_i = NULL, *new_j = NULL; \
241: datatype *new_a; \
242: \
243: PetscCheck(NONEW != -2, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "New nonzero at (%" PetscInt_FMT ",%" PetscInt_FMT ") caused a malloc. Use MatSetOption(A, MAT_NEW_NONZERO_ALLOCATION_ERR, PETSC_FALSE) to turn off this check", ROW, COL); \
244: /* malloc new storage space */ \
245: PetscCall(PetscShmgetAllocateArray(BS2 * new_nz, sizeof(PetscScalar), (void **)&new_a)); \
246: PetscCall(PetscShmgetAllocateArray(new_nz, sizeof(PetscInt), (void **)&new_j)); \
247: PetscCall(PetscShmgetAllocateArray(AM + 1, sizeof(PetscInt), (void **)&new_i)); \
248: Ain->free_a = PETSC_TRUE; \
249: Ain->free_ij = PETSC_TRUE; \
250: /* copy over old data into new slots */ \
251: for (ii = 0; ii < ROW + 1; ii++) new_i[ii] = AI[ii]; \
252: for (ii = ROW + 1; ii < AM + 1; ii++) new_i[ii] = AI[ii] + CHUNKSIZE; \
253: PetscCall(PetscArraycpy(new_j, AJ, AI[ROW] + NROW)); \
254: len = (new_nz - CHUNKSIZE - AI[ROW] - NROW); \
255: PetscCall(PetscArraycpy(new_j + AI[ROW] + NROW + CHUNKSIZE, PetscSafePointerPlusOffset(AJ, AI[ROW] + NROW), len)); \
256: PetscCall(PetscArraycpy(new_a, AA, BS2 * (AI[ROW] + NROW))); \
257: PetscCall(PetscArrayzero(new_a + BS2 * (AI[ROW] + NROW), BS2 * CHUNKSIZE)); \
258: PetscCall(PetscArraycpy(new_a + BS2 * (AI[ROW] + NROW + CHUNKSIZE), PetscSafePointerPlusOffset(AA, BS2 * (AI[ROW] + NROW)), BS2 * len)); \
259: /* free up old matrix storage */ \
260: PetscCall(MatSeqXAIJFreeAIJ(A, &Ain->a, &Ain->j, &Ain->i)); \
261: AA = new_a; \
262: Ain->a = new_a; \
263: AI = Ain->i = new_i; \
264: AJ = Ain->j = new_j; \
265: \
266: RP = AJ + AI[ROW]; \
267: AP = AA + BS2 * AI[ROW]; \
268: RMAX = AIMAX[ROW] = AIMAX[ROW] + CHUNKSIZE; \
269: Ain->maxnz += BS2 * CHUNKSIZE; \
270: Ain->reallocs++; \
271: Amat->nonzerostate++; \
272: } \
273: } while (0)
275: #define MatSeqXAIJReallocateAIJ_structure_only(Amat, AM, BS2, NROW, ROW, COL, RMAX, AI, AJ, RP, AIMAX, NONEW, datatype) \
276: do { \
277: if (NROW >= RMAX) { \
278: Mat_SeqAIJ *Ain = (Mat_SeqAIJ *)Amat->data; \
279: /* there is no extra room in row, therefore enlarge */ \
280: PetscInt CHUNKSIZE = 15, new_nz = AI[AM] + CHUNKSIZE, len, *new_i = NULL, *new_j = NULL; \
281: \
282: PetscCheck(NONEW != -2, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "New nonzero at (%" PetscInt_FMT ",%" PetscInt_FMT ") caused a malloc. Use MatSetOption(A, MAT_NEW_NONZERO_ALLOCATION_ERR, PETSC_FALSE) to turn off this check", ROW, COL); \
283: /* malloc new storage space */ \
284: PetscCall(PetscShmgetAllocateArray(new_nz, sizeof(PetscInt), (void **)&new_j)); \
285: PetscCall(PetscShmgetAllocateArray(AM + 1, sizeof(PetscInt), (void **)&new_i)); \
286: Ain->free_a = PETSC_FALSE; \
287: Ain->free_ij = PETSC_TRUE; \
288: \
289: /* copy over old data into new slots */ \
290: for (ii = 0; ii < ROW + 1; ii++) new_i[ii] = AI[ii]; \
291: for (ii = ROW + 1; ii < AM + 1; ii++) new_i[ii] = AI[ii] + CHUNKSIZE; \
292: PetscCall(PetscArraycpy(new_j, AJ, AI[ROW] + NROW)); \
293: len = (new_nz - CHUNKSIZE - AI[ROW] - NROW); \
294: PetscCall(PetscArraycpy(new_j + AI[ROW] + NROW + CHUNKSIZE, AJ + AI[ROW] + NROW, len)); \
295: \
296: /* free up old matrix storage */ \
297: PetscCall(MatSeqXAIJFreeAIJ(A, &Ain->a, &Ain->j, &Ain->i)); \
298: Ain->a = NULL; \
299: AI = Ain->i = new_i; \
300: AJ = Ain->j = new_j; \
301: \
302: RP = AJ + AI[ROW]; \
303: RMAX = AIMAX[ROW] = AIMAX[ROW] + CHUNKSIZE; \
304: Ain->maxnz += BS2 * CHUNKSIZE; \
305: Ain->reallocs++; \
306: Amat->nonzerostate++; \
307: } \
308: } while (0)
310: PETSC_INTERN PetscErrorCode MatSeqAIJSetPreallocation_SeqAIJ(Mat, PetscInt, const PetscInt *);
311: PETSC_INTERN PetscErrorCode MatSetPreallocationCOO_SeqAIJ(Mat, PetscCount, PetscInt[], PetscInt[]);
313: PETSC_INTERN PetscErrorCode MatILUFactorSymbolic_SeqAIJ(Mat, Mat, IS, IS, const MatFactorInfo *);
314: PETSC_INTERN PetscErrorCode MatILUFactorSymbolic_SeqAIJ_ilu0(Mat, Mat, IS, IS, const MatFactorInfo *);
316: PETSC_INTERN PetscErrorCode MatICCFactorSymbolic_SeqAIJ(Mat, Mat, IS, const MatFactorInfo *);
317: PETSC_INTERN PetscErrorCode MatCholeskyFactorSymbolic_SeqAIJ(Mat, Mat, IS, const MatFactorInfo *);
318: PETSC_INTERN PetscErrorCode MatCholeskyFactorNumeric_SeqAIJ_inplace(Mat, Mat, const MatFactorInfo *);
319: PETSC_INTERN PetscErrorCode MatCholeskyFactorNumeric_SeqAIJ(Mat, Mat, const MatFactorInfo *);
320: PETSC_INTERN PetscErrorCode MatDuplicate_SeqAIJ(Mat, MatDuplicateOption, Mat *);
321: PETSC_INTERN PetscErrorCode MatCopy_SeqAIJ(Mat, Mat, MatStructure);
322: PETSC_EXTERN PetscErrorCode MatGetDiagonalMarkers_SeqAIJ(Mat, const PetscInt **, PetscBool *);
323: PETSC_INTERN PetscErrorCode MatFindZeroDiagonals_SeqAIJ_Private(Mat, PetscInt *, PetscInt **);
325: PETSC_INTERN PetscErrorCode MatMult_SeqAIJ(Mat, Vec, Vec);
326: PETSC_INTERN PetscErrorCode MatMult_SeqAIJ_Inode(Mat, Vec, Vec);
327: PETSC_INTERN PetscErrorCode MatMultAdd_SeqAIJ(Mat, Vec, Vec, Vec);
328: PETSC_INTERN PetscErrorCode MatMultAdd_SeqAIJ_Inode(Mat, Vec, Vec, Vec);
329: PETSC_INTERN PetscErrorCode MatMultTranspose_SeqAIJ(Mat, Vec, Vec);
330: PETSC_INTERN PetscErrorCode MatMultTransposeAdd_SeqAIJ(Mat, Vec, Vec, Vec);
331: PETSC_INTERN PetscErrorCode MatSOR_SeqAIJ(Mat, Vec, PetscReal, MatSORType, PetscReal, PetscInt, PetscInt, Vec);
332: PETSC_INTERN PetscErrorCode MatSOR_SeqAIJ_Inode(Mat, Vec, PetscReal, MatSORType, PetscReal, PetscInt, PetscInt, Vec);
334: PETSC_INTERN PetscErrorCode MatSetOption_SeqAIJ(Mat, MatOption, PetscBool);
336: PETSC_INTERN PetscErrorCode MatGetSymbolicTranspose_SeqAIJ(Mat, PetscInt *[], PetscInt *[]);
337: PETSC_INTERN PetscErrorCode MatRestoreSymbolicTranspose_SeqAIJ(Mat, PetscInt *[], PetscInt *[]);
338: PETSC_INTERN PetscErrorCode MatGetSymbolicTransposeReduced_SeqAIJ(Mat, PetscInt, PetscInt, PetscInt *[], PetscInt *[]);
339: PETSC_INTERN PetscErrorCode MatTransposeSymbolic_SeqAIJ(Mat, Mat *);
340: PETSC_INTERN PetscErrorCode MatTranspose_SeqAIJ(Mat, MatReuse, Mat *);
342: PETSC_INTERN PetscErrorCode MatToSymmetricIJ_SeqAIJ(PetscInt, PetscInt *, PetscInt *, PetscBool, PetscInt, PetscInt, PetscInt **, PetscInt **);
343: PETSC_INTERN PetscErrorCode MatLUFactorSymbolic_SeqAIJ(Mat, Mat, IS, IS, const MatFactorInfo *);
344: PETSC_INTERN PetscErrorCode MatLUFactorNumeric_SeqAIJ_inplace(Mat, Mat, const MatFactorInfo *);
345: PETSC_INTERN PetscErrorCode MatLUFactorNumeric_SeqAIJ(Mat, Mat, const MatFactorInfo *);
346: PETSC_INTERN PetscErrorCode MatLUFactorNumeric_SeqAIJ_InplaceWithPerm(Mat, Mat, const MatFactorInfo *);
347: PETSC_INTERN PetscErrorCode MatLUFactor_SeqAIJ(Mat, IS, IS, const MatFactorInfo *);
348: PETSC_INTERN PetscErrorCode MatSolve_SeqAIJ_inplace(Mat, Vec, Vec);
349: PETSC_INTERN PetscErrorCode MatSolve_SeqAIJ(Mat, Vec, Vec);
350: PETSC_INTERN PetscErrorCode MatSolve_SeqAIJ_Inode(Mat, Vec, Vec);
351: PETSC_INTERN PetscErrorCode MatSolve_SeqAIJ_NaturalOrdering(Mat, Vec, Vec);
352: PETSC_INTERN PetscErrorCode MatSolveAdd_SeqAIJ(Mat, Vec, Vec, Vec);
353: PETSC_INTERN PetscErrorCode MatSolveTranspose_SeqAIJ_inplace(Mat, Vec, Vec);
354: PETSC_INTERN PetscErrorCode MatSolveTranspose_SeqAIJ(Mat, Vec, Vec);
355: PETSC_INTERN PetscErrorCode MatSolveTransposeAdd_SeqAIJ_inplace(Mat, Vec, Vec, Vec);
356: PETSC_INTERN PetscErrorCode MatSolveTransposeAdd_SeqAIJ(Mat, Vec, Vec, Vec);
357: PETSC_INTERN PetscErrorCode MatMatSolve_SeqAIJ(Mat, Mat, Mat);
358: PETSC_INTERN PetscErrorCode MatMatSolveTranspose_SeqAIJ(Mat, Mat, Mat);
359: PETSC_INTERN PetscErrorCode MatEqual_SeqAIJ(Mat, Mat, PetscBool *);
360: PETSC_INTERN PetscErrorCode MatFDColoringCreate_SeqXAIJ(Mat, ISColoring, MatFDColoring);
361: PETSC_INTERN PetscErrorCode MatFDColoringSetUp_SeqXAIJ(Mat, ISColoring, MatFDColoring);
362: PETSC_INTERN PetscErrorCode MatFDColoringSetUpBlocked_AIJ_Private(Mat, MatFDColoring, PetscInt);
363: PETSC_INTERN PetscErrorCode MatLoad_AIJ_HDF5(Mat, PetscViewer);
364: PETSC_INTERN PetscErrorCode MatLoad_SeqAIJ_Binary(Mat, PetscViewer);
365: PETSC_INTERN PetscErrorCode MatLoad_SeqAIJ(Mat, PetscViewer);
366: PETSC_INTERN PetscErrorCode RegisterApplyPtAPRoutines_Private(Mat);
368: #if defined(PETSC_HAVE_HYPRE)
369: PETSC_INTERN PetscErrorCode MatProductSetFromOptions_Transpose_AIJ_AIJ(Mat);
370: #endif
371: PETSC_INTERN PetscErrorCode MatProductSetFromOptions_SeqAIJ(Mat);
373: PETSC_INTERN PetscErrorCode MatProductSymbolic_SeqAIJ_SeqAIJ(Mat);
374: PETSC_INTERN PetscErrorCode MatProductSymbolic_PtAP_SeqAIJ_SeqAIJ(Mat);
375: PETSC_INTERN PetscErrorCode MatProductSymbolic_RARt_SeqAIJ_SeqAIJ(Mat);
377: PETSC_INTERN PetscErrorCode MatMatMultSymbolic_SeqAIJ_SeqAIJ(Mat, Mat, PetscReal, Mat);
378: PETSC_INTERN PetscErrorCode MatMatMultSymbolic_SeqAIJ_SeqAIJ_Sorted(Mat, Mat, PetscReal, Mat);
379: PETSC_INTERN PetscErrorCode MatMatMultSymbolic_SeqDense_SeqAIJ(Mat, Mat, PetscReal, Mat);
380: PETSC_INTERN PetscErrorCode MatMatMultSymbolic_SeqAIJ_SeqAIJ_Scalable(Mat, Mat, PetscReal, Mat);
381: PETSC_INTERN PetscErrorCode MatMatMultSymbolic_SeqAIJ_SeqAIJ_Scalable_fast(Mat, Mat, PetscReal, Mat);
382: PETSC_INTERN PetscErrorCode MatMatMultSymbolic_SeqAIJ_SeqAIJ_Heap(Mat, Mat, PetscReal, Mat);
383: PETSC_INTERN PetscErrorCode MatMatMultSymbolic_SeqAIJ_SeqAIJ_BTHeap(Mat, Mat, PetscReal, Mat);
384: PETSC_INTERN PetscErrorCode MatMatMultSymbolic_SeqAIJ_SeqAIJ_RowMerge(Mat, Mat, PetscReal, Mat);
385: PETSC_INTERN PetscErrorCode MatMatMultSymbolic_SeqAIJ_SeqAIJ_LLCondensed(Mat, Mat, PetscReal, Mat);
386: #if defined(PETSC_HAVE_HYPRE)
387: PETSC_INTERN PetscErrorCode MatMatMultSymbolic_AIJ_AIJ_wHYPRE(Mat, Mat, PetscReal, Mat);
388: #endif
390: PETSC_INTERN PetscErrorCode MatMatMultNumeric_SeqAIJ_SeqAIJ(Mat, Mat, Mat);
391: PETSC_INTERN PetscErrorCode MatMatMultNumeric_SeqAIJ_SeqAIJ_Sorted(Mat, Mat, Mat);
393: PETSC_INTERN PetscErrorCode MatMatMultNumeric_SeqDense_SeqAIJ(Mat, Mat, Mat);
394: PETSC_INTERN PetscErrorCode MatMatMultNumeric_SeqAIJ_SeqAIJ_Scalable(Mat, Mat, Mat);
396: PETSC_INTERN PetscErrorCode MatPtAPSymbolic_SeqAIJ_SeqAIJ_SparseAxpy(Mat, Mat, PetscReal, Mat);
397: PETSC_INTERN PetscErrorCode MatPtAPNumeric_SeqAIJ_SeqAIJ(Mat, Mat, Mat);
398: PETSC_INTERN PetscErrorCode MatPtAPNumeric_SeqAIJ_SeqAIJ_SparseAxpy(Mat, Mat, Mat);
400: PETSC_INTERN PetscErrorCode MatRARtSymbolic_SeqAIJ_SeqAIJ(Mat, Mat, PetscReal, Mat);
401: PETSC_INTERN PetscErrorCode MatRARtSymbolic_SeqAIJ_SeqAIJ_matmattransposemult(Mat, Mat, PetscReal, Mat);
402: PETSC_INTERN PetscErrorCode MatRARtSymbolic_SeqAIJ_SeqAIJ_colorrart(Mat, Mat, PetscReal, Mat);
403: PETSC_INTERN PetscErrorCode MatRARtNumeric_SeqAIJ_SeqAIJ(Mat, Mat, Mat);
404: PETSC_INTERN PetscErrorCode MatRARtNumeric_SeqAIJ_SeqAIJ_matmattransposemult(Mat, Mat, Mat);
405: PETSC_INTERN PetscErrorCode MatRARtNumeric_SeqAIJ_SeqAIJ_colorrart(Mat, Mat, Mat);
407: PETSC_INTERN PetscErrorCode MatTransposeMatMultSymbolic_SeqAIJ_SeqAIJ(Mat, Mat, PetscReal, Mat);
408: PETSC_INTERN PetscErrorCode MatTransposeMatMultNumeric_SeqAIJ_SeqAIJ(Mat, Mat, Mat);
409: PETSC_INTERN PetscErrorCode MatProductCtxDestroy_SeqAIJ_MatTransMatMult(void **);
411: PETSC_INTERN PetscErrorCode MatMatTransposeMultSymbolic_SeqAIJ_SeqAIJ(Mat, Mat, PetscReal, Mat);
412: PETSC_INTERN PetscErrorCode MatMatTransposeMultNumeric_SeqAIJ_SeqAIJ(Mat, Mat, Mat);
413: PETSC_INTERN PetscErrorCode MatTransposeColoringCreate_SeqAIJ(Mat, ISColoring, MatTransposeColoring);
414: PETSC_INTERN PetscErrorCode MatTransColoringApplySpToDen_SeqAIJ(MatTransposeColoring, Mat, Mat);
415: PETSC_INTERN PetscErrorCode MatTransColoringApplyDenToSp_SeqAIJ(MatTransposeColoring, Mat, Mat);
417: PETSC_INTERN PetscErrorCode MatMatMatMultSymbolic_SeqAIJ_SeqAIJ_SeqAIJ(Mat, Mat, Mat, PetscReal, Mat);
418: PETSC_INTERN PetscErrorCode MatMatMatMultNumeric_SeqAIJ_SeqAIJ_SeqAIJ(Mat, Mat, Mat, Mat);
420: PETSC_INTERN PetscErrorCode MatSetRandomSkipColumnRange_SeqAIJ_Private(Mat, PetscInt, PetscInt, PetscRandom);
421: PETSC_INTERN PetscErrorCode MatSetValues_SeqAIJ(Mat, PetscInt, const PetscInt[], PetscInt, const PetscInt[], const PetscScalar[], InsertMode);
422: PETSC_INTERN PetscErrorCode MatGetRow_SeqAIJ(Mat, PetscInt, PetscInt *, PetscInt **, PetscScalar **);
423: PETSC_INTERN PetscErrorCode MatRestoreRow_SeqAIJ(Mat, PetscInt, PetscInt *, PetscInt **, PetscScalar **);
424: PETSC_INTERN PetscErrorCode MatScale_SeqAIJ(Mat, PetscScalar);
425: PETSC_INTERN PetscErrorCode MatDiagonalScale_SeqAIJ(Mat, Vec, Vec);
426: PETSC_INTERN PetscErrorCode MatDiagonalSet_SeqAIJ(Mat, Vec, InsertMode);
427: PETSC_INTERN PetscErrorCode MatAXPY_SeqAIJ(Mat, PetscScalar, Mat, MatStructure);
428: PETSC_INTERN PetscErrorCode MatGetRowIJ_SeqAIJ(Mat, PetscInt, PetscBool, PetscBool, PetscInt *, const PetscInt *[], const PetscInt *[], PetscBool *);
429: PETSC_INTERN PetscErrorCode MatRestoreRowIJ_SeqAIJ(Mat, PetscInt, PetscBool, PetscBool, PetscInt *, const PetscInt *[], const PetscInt *[], PetscBool *);
430: PETSC_INTERN PetscErrorCode MatGetColumnIJ_SeqAIJ(Mat, PetscInt, PetscBool, PetscBool, PetscInt *, const PetscInt *[], const PetscInt *[], PetscBool *);
431: PETSC_INTERN PetscErrorCode MatRestoreColumnIJ_SeqAIJ(Mat, PetscInt, PetscBool, PetscBool, PetscInt *, const PetscInt *[], const PetscInt *[], PetscBool *);
432: PETSC_INTERN PetscErrorCode MatGetColumnIJ_SeqAIJ_Color(Mat, PetscInt, PetscBool, PetscBool, PetscInt *, const PetscInt *[], const PetscInt *[], PetscInt *[], PetscBool *);
433: PETSC_INTERN PetscErrorCode MatRestoreColumnIJ_SeqAIJ_Color(Mat, PetscInt, PetscBool, PetscBool, PetscInt *, const PetscInt *[], const PetscInt *[], PetscInt *[], PetscBool *);
434: PETSC_INTERN PetscErrorCode MatDestroy_SeqAIJ(Mat);
435: PETSC_INTERN PetscErrorCode MatView_SeqAIJ(Mat, PetscViewer);
437: PETSC_INTERN PetscErrorCode MatSeqAIJCheckInode(Mat);
438: PETSC_INTERN PetscErrorCode MatSeqAIJCheckInode_FactorLU(Mat);
440: PETSC_INTERN PetscErrorCode MatAXPYGetPreallocation_SeqAIJ(Mat, Mat, PetscInt *);
442: #if defined(PETSC_HAVE_MATLAB)
443: PETSC_EXTERN PetscErrorCode MatlabEnginePut_SeqAIJ(PetscObject, void *);
444: PETSC_EXTERN PetscErrorCode MatlabEngineGet_SeqAIJ(PetscObject, void *);
445: #endif
446: PETSC_INTERN PetscErrorCode MatConvert_SeqAIJ_SeqSBAIJ(Mat, MatType, MatReuse, Mat *);
447: PETSC_INTERN PetscErrorCode MatConvert_SeqAIJ_SeqBAIJ(Mat, MatType, MatReuse, Mat *);
448: PETSC_INTERN PetscErrorCode MatConvert_SeqAIJ_SeqDense(Mat, MatType, MatReuse, Mat *);
449: PETSC_INTERN PetscErrorCode MatConvert_SeqAIJ_SeqAIJCRL(Mat, MatType, MatReuse, Mat *);
450: PETSC_INTERN PetscErrorCode MatConvert_SeqAIJ_Elemental(Mat, MatType, MatReuse, Mat *);
451: #if defined(PETSC_HAVE_SCALAPACK)
452: PETSC_INTERN PetscErrorCode MatConvert_AIJ_ScaLAPACK(Mat, MatType, MatReuse, Mat *);
453: #endif
454: PETSC_INTERN PetscErrorCode MatConvert_AIJ_HYPRE(Mat, MatType, MatReuse, Mat *);
455: PETSC_INTERN PetscErrorCode MatConvert_SeqAIJ_SeqAIJPERM(Mat, MatType, MatReuse, Mat *);
456: PETSC_INTERN PetscErrorCode MatConvert_SeqAIJ_SeqAIJSELL(Mat, MatType, MatReuse, Mat *);
457: PETSC_INTERN PetscErrorCode MatConvert_SeqAIJ_SeqAIJMKL(Mat, MatType, MatReuse, Mat *);
458: PETSC_INTERN PetscErrorCode MatConvert_SeqAIJ_SeqAIJViennaCL(Mat, MatType, MatReuse, Mat *);
459: PETSC_INTERN PetscErrorCode MatReorderForNonzeroDiagonal_SeqAIJ(Mat, PetscReal, IS, IS);
460: PETSC_INTERN PetscErrorCode MatRARt_SeqAIJ_SeqAIJ(Mat, Mat, MatReuse, PetscReal, Mat *);
461: PETSC_EXTERN PetscErrorCode MatCreate_SeqAIJ(Mat);
462: PETSC_INTERN PetscErrorCode MatAssemblyEnd_SeqAIJ(Mat, MatAssemblyType);
463: PETSC_INTERN PetscErrorCode MatZeroEntries_SeqAIJ(Mat);
465: PETSC_INTERN PetscErrorCode MatAXPYGetPreallocation_SeqX_private(PetscInt, const PetscInt *, const PetscInt *, const PetscInt *, const PetscInt *, PetscInt *);
466: PETSC_INTERN PetscErrorCode MatCreateMPIMatConcatenateSeqMat_SeqAIJ(MPI_Comm, Mat, PetscInt, MatReuse, Mat *);
467: PETSC_INTERN PetscErrorCode MatCreateMPIMatConcatenateSeqMat_MPIAIJ(MPI_Comm, Mat, PetscInt, MatReuse, Mat *);
469: PETSC_INTERN PetscErrorCode MatSetSeqMat_SeqAIJ(Mat, IS, IS, MatStructure, Mat);
470: PETSC_INTERN PetscErrorCode MatEliminateZeros_SeqAIJ(Mat, PetscBool);
471: PETSC_INTERN PetscErrorCode MatDestroySubMatrix_Private(Mat_SubSppt *);
472: PETSC_INTERN PetscErrorCode MatDestroySubMatrix_SeqAIJ(Mat);
473: PETSC_INTERN PetscErrorCode MatDestroySubMatrix_Dummy(Mat);
474: PETSC_INTERN PetscErrorCode MatDestroySubMatrices_Dummy(PetscInt, Mat *[]);
475: PETSC_INTERN PetscErrorCode MatCreateSubMatrix_SeqAIJ(Mat, IS, IS, PetscInt, MatReuse, Mat *);
477: PETSC_INTERN PetscErrorCode MatSetSeqAIJWithArrays_private(MPI_Comm, PetscInt, PetscInt, PetscInt[], PetscInt[], PetscScalar[], MatType, Mat);
479: PETSC_INTERN PetscErrorCode MatResetPreallocation_SeqAIJ_Private(Mat A, PetscBool *memoryreset);
481: PETSC_SINGLE_LIBRARY_INTERN PetscErrorCode MatSeqAIJCompactOutExtraColumns_SeqAIJ(Mat, ISLocalToGlobalMapping *);
483: /*
484: PetscSparseDenseMinusDot - The inner kernel of triangular solves and Gauss-Siedel smoothing. \sum_i xv[i] * r[xi[i]] for CSR storage
486: Input Parameters:
487: + nnz - the number of entries
488: . r - the array of vector values
489: . xv - the matrix values for the row
490: - xi - the column indices of the nonzeros in the row
492: Output Parameter:
493: . sum - negative the sum of results
495: PETSc compile flags:
496: + PETSC_KERNEL_USE_UNROLL_4
497: - PETSC_KERNEL_USE_UNROLL_2
499: Developer Note:
500: The macro changes sum but not other parameters
502: .seealso: `PetscSparseDensePlusDot()`
503: */
504: #if defined(PETSC_KERNEL_USE_UNROLL_4)
505: #define PetscSparseDenseMinusDot(sum, r, xv, xi, nnz) \
506: do { \
507: if (nnz > 0) { \
508: PetscInt nnz2 = nnz, rem = nnz & 0x3; \
509: switch (rem) { \
510: case 3: \
511: sum -= *xv++ * r[*xi++]; \
512: case 2: \
513: sum -= *xv++ * r[*xi++]; \
514: case 1: \
515: sum -= *xv++ * r[*xi++]; \
516: nnz2 -= rem; \
517: } \
518: while (nnz2 > 0) { \
519: sum -= xv[0] * r[xi[0]] + xv[1] * r[xi[1]] + xv[2] * r[xi[2]] + xv[3] * r[xi[3]]; \
520: xv += 4; \
521: xi += 4; \
522: nnz2 -= 4; \
523: } \
524: xv -= nnz; \
525: xi -= nnz; \
526: } \
527: } while (0)
529: #elif defined(PETSC_KERNEL_USE_UNROLL_2)
530: #define PetscSparseDenseMinusDot(sum, r, xv, xi, nnz) \
531: do { \
532: PetscInt __i, __i1, __i2; \
533: for (__i = 0; __i < nnz - 1; __i += 2) { \
534: __i1 = xi[__i]; \
535: __i2 = xi[__i + 1]; \
536: sum -= (xv[__i] * r[__i1] + xv[__i + 1] * r[__i2]); \
537: } \
538: if (nnz & 0x1) sum -= xv[__i] * r[xi[__i]]; \
539: } while (0)
541: #else
542: #define PetscSparseDenseMinusDot(sum, r, xv, xi, nnz) \
543: do { \
544: PetscInt __i; \
545: for (__i = 0; __i < nnz; __i++) sum -= xv[__i] * r[xi[__i]]; \
546: } while (0)
547: #endif
549: /*
550: PetscSparseDensePlusDot - The inner kernel of matrix-vector product \sum_i xv[i] * r[xi[i]] for CSR storage
552: Input Parameters:
553: + nnz - the number of entries
554: . r - the array of vector values
555: . xv - the matrix values for the row
556: - xi - the column indices of the nonzeros in the row
558: Output Parameter:
559: . sum - the sum of results
561: PETSc compile flags:
562: + PETSC_KERNEL_USE_UNROLL_4
563: - PETSC_KERNEL_USE_UNROLL_2
565: Developer Note:
566: The macro changes sum but not other parameters
568: .seealso: `PetscSparseDenseMinusDot()`
569: */
570: #if defined(PETSC_KERNEL_USE_UNROLL_4)
571: #define PetscSparseDensePlusDot(sum, r, xv, xi, nnz) \
572: do { \
573: if (nnz > 0) { \
574: PetscInt nnz2 = nnz, rem = nnz & 0x3; \
575: switch (rem) { \
576: case 3: \
577: sum += *xv++ * r[*xi++]; \
578: case 2: \
579: sum += *xv++ * r[*xi++]; \
580: case 1: \
581: sum += *xv++ * r[*xi++]; \
582: nnz2 -= rem; \
583: } \
584: while (nnz2 > 0) { \
585: sum += xv[0] * r[xi[0]] + xv[1] * r[xi[1]] + xv[2] * r[xi[2]] + xv[3] * r[xi[3]]; \
586: xv += 4; \
587: xi += 4; \
588: nnz2 -= 4; \
589: } \
590: xv -= nnz; \
591: xi -= nnz; \
592: } \
593: } while (0)
595: #elif defined(PETSC_KERNEL_USE_UNROLL_2)
596: #define PetscSparseDensePlusDot(sum, r, xv, xi, nnz) \
597: do { \
598: PetscInt __i, __i1, __i2; \
599: for (__i = 0; __i < nnz - 1; __i += 2) { \
600: __i1 = xi[__i]; \
601: __i2 = xi[__i + 1]; \
602: sum += (xv[__i] * r[__i1] + xv[__i + 1] * r[__i2]); \
603: } \
604: if (nnz & 0x1) sum += xv[__i] * r[xi[__i]]; \
605: } while (0)
607: #elif !(defined(__GNUC__) && defined(_OPENMP)) && defined(PETSC_USE_AVX512_KERNELS) && defined(PETSC_HAVE_IMMINTRIN_H) && defined(__AVX512F__) && defined(PETSC_USE_REAL_DOUBLE) && !defined(PETSC_USE_COMPLEX) && !defined(PETSC_USE_64BIT_INDICES) && !defined(PETSC_SKIP_IMMINTRIN_H_CUDAWORKAROUND)
608: #define PetscSparseDensePlusDot(sum, r, xv, xi, nnz) PetscSparseDensePlusDot_AVX512_Private(&(sum), (r), (xv), (xi), (nnz))
610: #else
611: #define PetscSparseDensePlusDot(sum, r, xv, xi, nnz) \
612: do { \
613: PetscInt __i; \
614: for (__i = 0; __i < nnz; __i++) sum += xv[__i] * r[xi[__i]]; \
615: } while (0)
616: #endif
618: #if defined(PETSC_USE_AVX512_KERNELS) && defined(PETSC_HAVE_IMMINTRIN_H) && defined(__AVX512F__) && defined(PETSC_USE_REAL_DOUBLE) && !defined(PETSC_USE_COMPLEX) && !defined(PETSC_USE_64BIT_INDICES) && !defined(PETSC_SKIP_IMMINTRIN_H_CUDAWORKAROUND)
619: #include <immintrin.h>
620: #if !defined(_MM_SCALE_8)
621: #define _MM_SCALE_8 8
622: #endif
624: static inline void PetscSparseDensePlusDot_AVX512_Private(PetscScalar *sum, const PetscScalar *x, const MatScalar *aa, const PetscInt *aj, PetscInt n)
625: {
626: __m512d vec_x, vec_y, vec_vals;
627: __m256i vec_idx;
628: PetscInt j;
630: vec_y = _mm512_setzero_pd();
631: for (j = 0; j < (n >> 3); j++) {
632: vec_idx = _mm256_loadu_si256((__m256i const *)aj);
633: vec_vals = _mm512_loadu_pd(aa);
634: vec_x = _mm512_i32gather_pd(vec_idx, x, _MM_SCALE_8);
635: vec_y = _mm512_fmadd_pd(vec_x, vec_vals, vec_y);
636: aj += 8;
637: aa += 8;
638: }
639: #if defined(__AVX512VL__)
640: /* masked load requires avx512vl, which is not supported by KNL */
641: if (n & 0x07) {
642: __mmask8 mask;
643: mask = (__mmask8)(0xff >> (8 - (n & 0x07)));
644: vec_idx = _mm256_mask_loadu_epi32(vec_idx, mask, aj);
645: vec_vals = _mm512_mask_loadu_pd(vec_vals, mask, aa);
646: vec_x = _mm512_mask_i32gather_pd(vec_x, mask, vec_idx, x, _MM_SCALE_8);
647: vec_y = _mm512_mask3_fmadd_pd(vec_x, vec_vals, vec_y, mask);
648: }
649: *sum += _mm512_reduce_add_pd(vec_y);
650: #else
651: *sum += _mm512_reduce_add_pd(vec_y);
652: for (j = 0; j < (n & 0x07); j++) *sum += aa[j] * x[aj[j]];
653: #endif
654: }
655: #endif
657: /*
658: PetscSparseDenseMaxDot - The inner kernel of a modified matrix-vector product \max_i xv[i] * r[xi[i]] for CSR storage
660: Input Parameters:
661: + nnz - the number of entries
662: . r - the array of vector values
663: . xv - the matrix values for the row
664: - xi - the column indices of the nonzeros in the row
666: Output Parameter:
667: . max - the max of results
669: .seealso: `PetscSparseDensePlusDot()`, `PetscSparseDenseMinusDot()`
670: */
671: #define PetscSparseDenseMaxDot(max, r, xv, xi, nnz) \
672: do { \
673: for (PetscInt __i = 0; __i < (nnz); __i++) max = PetscMax(PetscRealPart(max), PetscRealPart((xv)[__i] * (r)[(xi)[__i]])); \
674: } while (0)
676: /*
677: Add column indices into table for counting the max nonzeros of merged rows
678: */
679: #define MatRowMergeMax_SeqAIJ(mat, nrows, ta) \
680: do { \
681: if (mat) { \
682: for (PetscInt _row = 0; _row < (nrows); _row++) { \
683: const PetscInt _nz = (mat)->i[_row + 1] - (mat)->i[_row]; \
684: for (PetscInt _j = 0; _j < _nz; _j++) { \
685: PetscInt *_col = _j + (mat)->j + (mat)->i[_row]; \
686: PetscCall(PetscHMapISet((ta), *_col + 1, 1)); \
687: } \
688: } \
689: } \
690: } while (0)
692: /*
693: Add column indices into table for counting the nonzeros of merged rows
694: */
695: #define MatMergeRows_SeqAIJ(mat, nrows, rows, ta) \
696: do { \
697: for (PetscInt _i = 0; _i < (nrows); _i++) { \
698: const PetscInt _row = (rows)[_i]; \
699: const PetscInt _nz = (mat)->i[_row + 1] - (mat)->i[_row]; \
700: for (PetscInt _j = 0; _j < _nz; _j++) { \
701: PetscInt *_col = _j + (mat)->j + (mat)->i[_row]; \
702: PetscCall(PetscHMapISetWithMode((ta), *_col + 1, 1, INSERT_VALUES)); \
703: } \
704: } \
705: } while (0)