Actual source code: spbas.c
petsc-3.14.6 2021-03-30
1: #include <../src/mat/impls/aij/seq/aij.h>
2: #include <../src/mat/impls/aij/seq/bas/spbas.h>
4: /*MC
5: MATSOLVERBAS - Provides ICC(k) with drop tolerance
7: Works with MATAIJ matrices
9: Options Database Keys:
10: + -pc_factor_levels <l> - number of levels of fill
11: - -pc_factor_drop_tolerance - is not currently hooked up to do anything
13: Level: intermediate
15: Contributed by: Bas van 't Hof
17: Notes:
18: Since this currently hooked up to use drop tolerance it should produce the same factors and hence convergence as the PETSc ICC, for higher
19: levels of fill it does not. This needs to be investigated. Unless you are interested in drop tolerance ICC and willing to work through the code
20: we recommend not using this functionality.
22: .seealso: PCFactorSetMatSolverType(), MatSolverType, PCFactorSetLevels(), PCFactorSetDropTolerance()
24: M*/
26: /*
27: spbas_memory_requirement:
28: Calculate the number of bytes needed to store tha matrix
29: */
30: size_t spbas_memory_requirement(spbas_matrix matrix)
31: {
32: size_t memreq = 6 * sizeof(PetscInt) + /* nrows, ncols, nnz, n_alloc_icol, n_alloc_val, col_idx_type */
33: sizeof(PetscBool) + /* block_data */
34: sizeof(PetscScalar**) + /* values */
35: sizeof(PetscScalar*) + /* alloc_val */
36: 2 * sizeof(PetscInt**) + /* icols, icols0 */
37: 2 * sizeof(PetscInt*) + /* row_nnz, alloc_icol */
38: matrix.nrows * sizeof(PetscInt) + /* row_nnz[*] */
39: matrix.nrows * sizeof(PetscInt*); /* icols[*] */
41: /* icol0[*] */
42: if (matrix.col_idx_type == SPBAS_OFFSET_ARRAY) memreq += matrix.nrows * sizeof(PetscInt);
44: /* icols[*][*] */
45: if (matrix.block_data) memreq += matrix.n_alloc_icol * sizeof(PetscInt);
46: else memreq += matrix.nnz * sizeof(PetscInt);
48: if (matrix.values) {
49: memreq += matrix.nrows * sizeof(PetscScalar*); /* values[*] */
50: /* values[*][*] */
51: if (matrix.block_data) memreq += matrix.n_alloc_val * sizeof(PetscScalar);
52: else memreq += matrix.nnz * sizeof(PetscScalar);
53: }
54: return memreq;
55: }
57: /*
58: spbas_allocate_pattern:
59: allocate the pattern arrays row_nnz, icols and optionally values
60: */
61: PetscErrorCode spbas_allocate_pattern(spbas_matrix * result, PetscBool do_values)
62: {
64: PetscInt nrows = result->nrows;
65: PetscInt col_idx_type = result->col_idx_type;
68: /* Allocate sparseness pattern */
69: PetscMalloc1(nrows,&result->row_nnz);
70: PetscMalloc1(nrows,&result->icols);
72: /* If offsets are given wrt an array, create array */
73: if (col_idx_type == SPBAS_OFFSET_ARRAY) {
74: PetscMalloc1(nrows,&result->icol0);
75: } else {
76: result->icol0 = NULL;
77: }
79: /* If values are given, allocate values array */
80: if (do_values) {
81: PetscMalloc1(nrows,&result->values);
82: } else {
83: result->values = NULL;
84: }
85: return(0);
86: }
88: /*
89: spbas_allocate_data:
90: in case of block_data:
91: Allocate the data arrays alloc_icol and optionally alloc_val,
92: set appropriate pointers from icols and values;
93: in case of !block_data:
94: Allocate the arrays icols[i] and optionally values[i]
95: */
96: PetscErrorCode spbas_allocate_data(spbas_matrix * result)
97: {
98: PetscInt i;
99: PetscInt nnz = result->nnz;
100: PetscInt nrows = result->nrows;
101: PetscInt r_nnz;
103: PetscBool do_values = (result->values) ? PETSC_TRUE : PETSC_FALSE;
104: PetscBool block_data = result->block_data;
107: if (block_data) {
108: /* Allocate the column number array and point to it */
109: result->n_alloc_icol = nnz;
111: PetscMalloc1(nnz, &result->alloc_icol);
113: result->icols[0] = result->alloc_icol;
114: for (i=1; i<nrows; i++) {
115: result->icols[i] = result->icols[i-1] + result->row_nnz[i-1];
116: }
118: /* Allocate the value array and point to it */
119: if (do_values) {
120: result->n_alloc_val = nnz;
122: PetscMalloc1(nnz, &result->alloc_val);
124: result->values[0] = result->alloc_val;
125: for (i=1; i<nrows; i++) {
126: result->values[i] = result->values[i-1] + result->row_nnz[i-1];
127: }
128: }
129: } else {
130: for (i=0; i<nrows; i++) {
131: r_nnz = result->row_nnz[i];
132: PetscMalloc1(r_nnz, &result->icols[i]);
133: }
134: if (do_values) {
135: for (i=0; i<nrows; i++) {
136: r_nnz = result->row_nnz[i];
137: PetscMalloc1(r_nnz, &result->values[i]);
138: }
139: }
140: }
141: return(0);
142: }
144: /*
145: spbas_row_order_icol
146: determine if row i1 should come
147: + before row i2 in the sorted rows (return -1),
148: + after (return 1)
149: + is identical (return 0).
150: */
151: int spbas_row_order_icol(PetscInt i1, PetscInt i2, PetscInt *irow_in, PetscInt *icol_in,PetscInt col_idx_type)
152: {
153: PetscInt j;
154: PetscInt nnz1 = irow_in[i1+1] - irow_in[i1];
155: PetscInt nnz2 = irow_in[i2+1] - irow_in[i2];
156: PetscInt * icol1 = &icol_in[irow_in[i1]];
157: PetscInt * icol2 = &icol_in[irow_in[i2]];
159: if (nnz1<nnz2) return -1;
160: if (nnz1>nnz2) return 1;
162: if (col_idx_type == SPBAS_COLUMN_NUMBERS) {
163: for (j=0; j<nnz1; j++) {
164: if (icol1[j]< icol2[j]) return -1;
165: if (icol1[j]> icol2[j]) return 1;
166: }
167: } else if (col_idx_type == SPBAS_DIAGONAL_OFFSETS) {
168: for (j=0; j<nnz1; j++) {
169: if (icol1[j]-i1< icol2[j]-i2) return -1;
170: if (icol1[j]-i1> icol2[j]-i2) return 1;
171: }
172: } else if (col_idx_type == SPBAS_OFFSET_ARRAY) {
173: for (j=1; j<nnz1; j++) {
174: if (icol1[j]-icol1[0] < icol2[j]-icol2[0]) return -1;
175: if (icol1[j]-icol1[0] > icol2[j]-icol2[0]) return 1;
176: }
177: }
178: return 0;
179: }
181: /*
182: spbas_mergesort_icols:
183: return a sorting of the rows in which identical sparseness patterns are
184: next to each other
185: */
186: PetscErrorCode spbas_mergesort_icols(PetscInt nrows, PetscInt * irow_in, PetscInt * icol_in,PetscInt col_idx_type, PetscInt *isort)
187: {
189: PetscInt istep; /* Chunk-sizes of already sorted parts of arrays */
190: PetscInt i, i1, i2; /* Loop counters for (partly) sorted arrays */
191: PetscInt istart, i1end, i2end; /* start of newly sorted array part, end of both parts */
192: PetscInt *ialloc; /* Allocated arrays */
193: PetscInt *iswap; /* auxiliary pointers for swapping */
194: PetscInt *ihlp1; /* Pointers to new version of arrays, */
195: PetscInt *ihlp2; /* Pointers to previous version of arrays, */
198: PetscMalloc1(nrows,&ialloc);
200: ihlp1 = ialloc;
201: ihlp2 = isort;
203: /* Sorted array chunks are first 1 long, and increase until they are the complete array */
204: for (istep=1; istep<nrows; istep*=2) {
205: /*
206: Combine sorted parts
207: istart:istart+istep-1 and istart+istep-1:istart+2*istep-1
208: of ihlp2 and vhlp2
210: into one sorted part
211: istart:istart+2*istep-1
212: of ihlp1 and vhlp1
213: */
214: for (istart=0; istart<nrows; istart+=2*istep) {
215: /* Set counters and bound array part endings */
216: i1=istart; i1end = i1+istep; if (i1end>nrows) i1end=nrows;
217: i2=istart+istep; i2end = i2+istep; if (i2end>nrows) i2end=nrows;
219: /* Merge the two array parts */
220: for (i=istart; i<i2end; i++) {
221: if (i1<i1end && i2<i2end && spbas_row_order_icol(ihlp2[i1], ihlp2[i2], irow_in, icol_in, col_idx_type) < 0) {
222: ihlp1[i] = ihlp2[i1];
223: i1++;
224: } else if (i2<i2end) {
225: ihlp1[i] = ihlp2[i2];
226: i2++;
227: } else {
228: ihlp1[i] = ihlp2[i1];
229: i1++;
230: }
231: }
232: }
234: /* Swap the two array sets */
235: iswap = ihlp2; ihlp2 = ihlp1; ihlp1 = iswap;
236: }
238: /* Copy one more time in case the sorted arrays are the temporary ones */
239: if (ihlp2 != isort) {
240: for (i=0; i<nrows; i++) isort[i] = ihlp2[i];
241: }
242: PetscFree(ialloc);
243: return(0);
244: }
248: /*
249: spbas_compress_pattern:
250: calculate a compressed sparseness pattern for a sparseness pattern
251: given in compressed row storage. The compressed sparseness pattern may
252: require (much) less memory.
253: */
254: PetscErrorCode spbas_compress_pattern(PetscInt *irow_in, PetscInt *icol_in, PetscInt nrows, PetscInt ncols, PetscInt col_idx_type, spbas_matrix *B,PetscReal *mem_reduction)
255: {
256: PetscInt nnz = irow_in[nrows];
257: size_t mem_orig = (nrows + nnz) * sizeof(PetscInt);
258: size_t mem_compressed;
259: PetscErrorCode ierr;
260: PetscInt *isort;
261: PetscInt *icols;
262: PetscInt row_nnz;
263: PetscInt *ipoint;
264: PetscBool *used;
265: PetscInt ptr;
266: PetscInt i,j;
267: const PetscBool no_values = PETSC_FALSE;
270: /* Allocate the structure of the new matrix */
271: B->nrows = nrows;
272: B->ncols = ncols;
273: B->nnz = nnz;
274: B->col_idx_type = col_idx_type;
275: B->block_data = PETSC_TRUE;
277: spbas_allocate_pattern(B, no_values);
279: /* When using an offset array, set it */
280: if (col_idx_type==SPBAS_OFFSET_ARRAY) {
281: for (i=0; i<nrows; i++) B->icol0[i] = icol_in[irow_in[i]];
282: }
284: /* Allocate the ordering for the rows */
285: PetscMalloc1(nrows,&isort);
286: PetscMalloc1(nrows,&ipoint);
287: PetscCalloc1(nrows,&used);
289: for (i = 0; i<nrows; i++) {
290: B->row_nnz[i] = irow_in[i+1]-irow_in[i];
291: isort[i] = i;
292: ipoint[i] = i;
293: }
295: /* Sort the rows so that identical columns will be next to each other */
296: spbas_mergesort_icols(nrows, irow_in, icol_in, col_idx_type, isort);
297: PetscInfo(NULL,"Rows have been sorted for patterns\n");
299: /* Replace identical rows with the first one in the list */
300: for (i=1; i<nrows; i++) {
301: if (spbas_row_order_icol(isort[i-1], isort[i], irow_in, icol_in, col_idx_type) == 0) {
302: ipoint[isort[i]] = ipoint[isort[i-1]];
303: }
304: }
306: /* Collect the rows which are used*/
307: for (i=0; i<nrows; i++) used[ipoint[i]] = PETSC_TRUE;
309: /* Calculate needed memory */
310: B->n_alloc_icol = 0;
311: for (i=0; i<nrows; i++) {
312: if (used[i]) B->n_alloc_icol += B->row_nnz[i];
313: }
314: PetscMalloc1(B->n_alloc_icol,&B->alloc_icol);
316: /* Fill in the diagonal offsets for the rows which store their own data */
317: ptr = 0;
318: for (i=0; i<B->nrows; i++) {
319: if (used[i]) {
320: B->icols[i] = &B->alloc_icol[ptr];
321: icols = &icol_in[irow_in[i]];
322: row_nnz = B->row_nnz[i];
323: if (col_idx_type == SPBAS_COLUMN_NUMBERS) {
324: for (j=0; j<row_nnz; j++) {
325: B->icols[i][j] = icols[j];
326: }
327: } else if (col_idx_type == SPBAS_DIAGONAL_OFFSETS) {
328: for (j=0; j<row_nnz; j++) {
329: B->icols[i][j] = icols[j]-i;
330: }
331: } else if (col_idx_type == SPBAS_OFFSET_ARRAY) {
332: for (j=0; j<row_nnz; j++) {
333: B->icols[i][j] = icols[j]-icols[0];
334: }
335: }
336: ptr += B->row_nnz[i];
337: }
338: }
340: /* Point to the right places for all data */
341: for (i=0; i<nrows; i++) {
342: B->icols[i] = B->icols[ipoint[i]];
343: }
344: PetscInfo(NULL,"Row patterns have been compressed\n");
345: PetscInfo1(NULL," (%g nonzeros per row)\n", (double) ((PetscReal) nnz / (PetscReal) nrows));
347: ierr=PetscFree(isort);
348: ierr=PetscFree(used);
349: ierr=PetscFree(ipoint);
351: mem_compressed = spbas_memory_requirement(*B);
352: *mem_reduction = 100.0 * (PetscReal)(mem_orig-mem_compressed)/ (PetscReal) mem_orig;
353: return(0);
354: }
356: /*
357: spbas_incomplete_cholesky
358: Incomplete Cholesky decomposition
359: */
360: #include <../src/mat/impls/aij/seq/bas/spbas_cholesky.h>
362: /*
363: spbas_delete : de-allocate the arrays owned by this matrix
364: */
365: PetscErrorCode spbas_delete(spbas_matrix matrix)
366: {
367: PetscInt i;
371: if (matrix.block_data) {
372: ierr=PetscFree(matrix.alloc_icol);
373: if (matrix.values) {ierr=PetscFree(matrix.alloc_val);}
374: } else {
375: for (i=0; i<matrix.nrows; i++) { ierr=PetscFree(matrix.icols[i]);}
376: PetscFree(matrix.icols);
377: if (matrix.values) {
378: for (i=0; i<matrix.nrows; i++) { ierr=PetscFree(matrix.values[i]);}
379: }
380: }
382: ierr=PetscFree(matrix.row_nnz);
383: ierr=PetscFree(matrix.icols);
384: if (matrix.col_idx_type == SPBAS_OFFSET_ARRAY) {ierr=PetscFree(matrix.icol0);}
385: ierr=PetscFree(matrix.values);
386: return(0);
387: }
389: /*
390: spbas_matrix_to_crs:
391: Convert an spbas_matrix to compessed row storage
392: */
393: PetscErrorCode spbas_matrix_to_crs(spbas_matrix matrix_A,MatScalar **val_out, PetscInt **irow_out, PetscInt **icol_out)
394: {
395: PetscInt nrows = matrix_A.nrows;
396: PetscInt nnz = matrix_A.nnz;
397: PetscInt i,j,r_nnz,i0;
398: PetscInt *irow;
399: PetscInt *icol;
400: PetscInt *icol_A;
401: MatScalar *val;
402: PetscScalar *val_A;
403: PetscInt col_idx_type = matrix_A.col_idx_type;
404: PetscBool do_values = matrix_A.values ? PETSC_TRUE : PETSC_FALSE;
408: PetscMalloc1(nrows+1, &irow);
409: PetscMalloc1(nnz, &icol);
410: *icol_out = icol;
411: *irow_out = irow;
412: if (do_values) {
413: PetscMalloc1(nnz, &val);
414: *val_out = val; *icol_out = icol; *irow_out=irow;
415: }
417: irow[0]=0;
418: for (i=0; i<nrows; i++) {
419: r_nnz = matrix_A.row_nnz[i];
420: i0 = irow[i];
421: irow[i+1] = i0 + r_nnz;
422: icol_A = matrix_A.icols[i];
424: if (do_values) {
425: val_A = matrix_A.values[i];
426: for (j=0; j<r_nnz; j++) {
427: icol[i0+j] = icol_A[j];
428: val[i0+j] = val_A[j];
429: }
430: } else {
431: for (j=0; j<r_nnz; j++) icol[i0+j] = icol_A[j];
432: }
434: if (col_idx_type == SPBAS_DIAGONAL_OFFSETS) {
435: for (j=0; j<r_nnz; j++) icol[i0+j] += i;
436: } else if (col_idx_type == SPBAS_OFFSET_ARRAY) {
437: i0 = matrix_A.icol0[i];
438: for (j=0; j<r_nnz; j++) icol[i0+j] += i0;
439: }
440: }
441: return(0);
442: }
445: /*
446: spbas_transpose
447: return the transpose of a matrix
448: */
449: PetscErrorCode spbas_transpose(spbas_matrix in_matrix, spbas_matrix * result)
450: {
451: PetscInt col_idx_type = in_matrix.col_idx_type;
452: PetscInt nnz = in_matrix.nnz;
453: PetscInt ncols = in_matrix.nrows;
454: PetscInt nrows = in_matrix.ncols;
455: PetscInt i,j,k;
456: PetscInt r_nnz;
457: PetscInt *irow;
458: PetscInt icol0 = 0;
459: PetscScalar * val;
463: /* Copy input values */
464: result->nrows = nrows;
465: result->ncols = ncols;
466: result->nnz = nnz;
467: result->col_idx_type = SPBAS_COLUMN_NUMBERS;
468: result->block_data = PETSC_TRUE;
470: /* Allocate sparseness pattern */
471: spbas_allocate_pattern(result, in_matrix.values ? PETSC_TRUE : PETSC_FALSE);
473: /* Count the number of nonzeros in each row */
474: for (i = 0; i<nrows; i++) result->row_nnz[i] = 0;
476: for (i=0; i<ncols; i++) {
477: r_nnz = in_matrix.row_nnz[i];
478: irow = in_matrix.icols[i];
479: if (col_idx_type == SPBAS_COLUMN_NUMBERS) {
480: for (j=0; j<r_nnz; j++) result->row_nnz[irow[j]]++;
481: } else if (col_idx_type == SPBAS_DIAGONAL_OFFSETS) {
482: for (j=0; j<r_nnz; j++) result->row_nnz[i+irow[j]]++;
483: } else if (col_idx_type == SPBAS_OFFSET_ARRAY) {
484: icol0=in_matrix.icol0[i];
485: for (j=0; j<r_nnz; j++) result->row_nnz[icol0+irow[j]]++;
486: }
487: }
489: /* Set the pointers to the data */
490: spbas_allocate_data(result);
492: /* Reset the number of nonzeros in each row */
493: for (i = 0; i<nrows; i++) result->row_nnz[i] = 0;
495: /* Fill the data arrays */
496: if (in_matrix.values) {
497: for (i=0; i<ncols; i++) {
498: r_nnz = in_matrix.row_nnz[i];
499: irow = in_matrix.icols[i];
500: val = in_matrix.values[i];
502: if (col_idx_type == SPBAS_COLUMN_NUMBERS) icol0 = 0;
503: else if (col_idx_type == SPBAS_DIAGONAL_OFFSETS) icol0 = i;
504: else if (col_idx_type == SPBAS_OFFSET_ARRAY) icol0 = in_matrix.icol0[i];
505: for (j=0; j<r_nnz; j++) {
506: k = icol0 + irow[j];
507: result->icols[k][result->row_nnz[k]] = i;
508: result->values[k][result->row_nnz[k]] = val[j];
509: result->row_nnz[k]++;
510: }
511: }
512: } else {
513: for (i=0; i<ncols; i++) {
514: r_nnz = in_matrix.row_nnz[i];
515: irow = in_matrix.icols[i];
517: if (col_idx_type == SPBAS_COLUMN_NUMBERS) icol0=0;
518: else if (col_idx_type == SPBAS_DIAGONAL_OFFSETS) icol0=i;
519: else if (col_idx_type == SPBAS_OFFSET_ARRAY) icol0=in_matrix.icol0[i];
521: for (j=0; j<r_nnz; j++) {
522: k = icol0 + irow[j];
523: result->icols[k][result->row_nnz[k]] = i;
524: result->row_nnz[k]++;
525: }
526: }
527: }
528: return(0);
529: }
531: /*
532: spbas_mergesort
534: mergesort for an array of integers and an array of associated
535: reals
537: on output, icol[0..nnz-1] is increasing;
538: val[0..nnz-1] has undergone the same permutation as icol
540: NB: val may be NULL: in that case, only the integers are sorted
542: */
543: PetscErrorCode spbas_mergesort(PetscInt nnz, PetscInt *icol, PetscScalar *val)
544: {
545: PetscInt istep; /* Chunk-sizes of already sorted parts of arrays */
546: PetscInt i, i1, i2; /* Loop counters for (partly) sorted arrays */
547: PetscInt istart, i1end, i2end; /* start of newly sorted array part, end of both parts */
548: PetscInt *ialloc; /* Allocated arrays */
549: PetscScalar *valloc=NULL;
550: PetscInt *iswap; /* auxiliary pointers for swapping */
551: PetscScalar *vswap;
552: PetscInt *ihlp1; /* Pointers to new version of arrays, */
553: PetscScalar *vhlp1=NULL; /* (arrays under construction) */
554: PetscInt *ihlp2; /* Pointers to previous version of arrays, */
555: PetscScalar *vhlp2=NULL;
558: PetscMalloc1(nnz,&ialloc);
559: ihlp1 = ialloc;
560: ihlp2 = icol;
562: if (val) {
563: PetscMalloc1(nnz,&valloc);
564: vhlp1 = valloc;
565: vhlp2 = val;
566: }
569: /* Sorted array chunks are first 1 long, and increase until they are the complete array */
570: for (istep=1; istep<nnz; istep*=2) {
571: /*
572: Combine sorted parts
573: istart:istart+istep-1 and istart+istep-1:istart+2*istep-1
574: of ihlp2 and vhlp2
576: into one sorted part
577: istart:istart+2*istep-1
578: of ihlp1 and vhlp1
579: */
580: for (istart=0; istart<nnz; istart+=2*istep) {
581: /* Set counters and bound array part endings */
582: i1=istart; i1end = i1+istep; if (i1end>nnz) i1end=nnz;
583: i2=istart+istep; i2end = i2+istep; if (i2end>nnz) i2end=nnz;
585: /* Merge the two array parts */
586: if (val) {
587: for (i=istart; i<i2end; i++) {
588: if (i1<i1end && i2<i2end && ihlp2[i1] < ihlp2[i2]) {
589: ihlp1[i] = ihlp2[i1];
590: vhlp1[i] = vhlp2[i1];
591: i1++;
592: } else if (i2<i2end) {
593: ihlp1[i] = ihlp2[i2];
594: vhlp1[i] = vhlp2[i2];
595: i2++;
596: } else {
597: ihlp1[i] = ihlp2[i1];
598: vhlp1[i] = vhlp2[i1];
599: i1++;
600: }
601: }
602: } else {
603: for (i=istart; i<i2end; i++) {
604: if (i1<i1end && i2<i2end && ihlp2[i1] < ihlp2[i2]) {
605: ihlp1[i] = ihlp2[i1];
606: i1++;
607: } else if (i2<i2end) {
608: ihlp1[i] = ihlp2[i2];
609: i2++;
610: } else {
611: ihlp1[i] = ihlp2[i1];
612: i1++;
613: }
614: }
615: }
616: }
618: /* Swap the two array sets */
619: iswap = ihlp2; ihlp2 = ihlp1; ihlp1 = iswap;
620: vswap = vhlp2; vhlp2 = vhlp1; vhlp1 = vswap;
621: }
623: /* Copy one more time in case the sorted arrays are the temporary ones */
624: if (ihlp2 != icol) {
625: for (i=0; i<nnz; i++) icol[i] = ihlp2[i];
626: if (val) {
627: for (i=0; i<nnz; i++) val[i] = vhlp2[i];
628: }
629: }
631: PetscFree(ialloc);
632: if (val) {PetscFree(valloc);}
633: return(0);
634: }
636: /*
637: spbas_apply_reordering_rows:
638: apply the given reordering to the rows: matrix_A = matrix_A(perm,:);
639: */
640: PetscErrorCode spbas_apply_reordering_rows(spbas_matrix *matrix_A, const PetscInt *permutation)
641: {
642: PetscInt i,j,ip;
643: PetscInt nrows=matrix_A->nrows;
644: PetscInt * row_nnz;
645: PetscInt **icols;
646: PetscBool do_values = matrix_A->values ? PETSC_TRUE : PETSC_FALSE;
647: PetscScalar **vals = NULL;
651: if (matrix_A->col_idx_type != SPBAS_DIAGONAL_OFFSETS) SETERRQ(PETSC_COMM_SELF, PETSC_ERR_SUP_SYS,"must have diagonal offsets in pattern\n");
653: if (do_values) {
654: PetscMalloc1(nrows, &vals);
655: }
656: PetscMalloc1(nrows, &row_nnz);
657: PetscMalloc1(nrows, &icols);
659: for (i=0; i<nrows; i++) {
660: ip = permutation[i];
661: if (do_values) vals[i] = matrix_A->values[ip];
662: icols[i] = matrix_A->icols[ip];
663: row_nnz[i] = matrix_A->row_nnz[ip];
664: for (j=0; j<row_nnz[i]; j++) icols[i][j] += ip-i;
665: }
667: if (do_values) { PetscFree(matrix_A->values);}
668: PetscFree(matrix_A->icols);
669: PetscFree(matrix_A->row_nnz);
671: if (do_values) matrix_A->values = vals;
672: matrix_A->icols = icols;
673: matrix_A->row_nnz = row_nnz;
674: return(0);
675: }
678: /*
679: spbas_apply_reordering_cols:
680: apply the given reordering to the columns: matrix_A(:,perm) = matrix_A;
681: */
682: PetscErrorCode spbas_apply_reordering_cols(spbas_matrix *matrix_A,const PetscInt *permutation)
683: {
684: PetscInt i,j;
685: PetscInt nrows=matrix_A->nrows;
686: PetscInt row_nnz;
687: PetscInt *icols;
688: PetscBool do_values = matrix_A->values ? PETSC_TRUE : PETSC_FALSE;
689: PetscScalar *vals = NULL;
693: if (matrix_A->col_idx_type != SPBAS_DIAGONAL_OFFSETS) SETERRQ(PETSC_COMM_SELF, PETSC_ERR_SUP_SYS, "must have diagonal offsets in pattern\n");
695: for (i=0; i<nrows; i++) {
696: icols = matrix_A->icols[i];
697: row_nnz = matrix_A->row_nnz[i];
698: if (do_values) vals = matrix_A->values[i];
700: for (j=0; j<row_nnz; j++) {
701: icols[j] = permutation[i+icols[j]]-i;
702: }
703: spbas_mergesort(row_nnz, icols, vals);
704: }
705: return(0);
706: }
708: /*
709: spbas_apply_reordering:
710: apply the given reordering: matrix_A(perm,perm) = matrix_A;
711: */
712: PetscErrorCode spbas_apply_reordering(spbas_matrix *matrix_A, const PetscInt *permutation, const PetscInt * inv_perm)
713: {
717: spbas_apply_reordering_rows(matrix_A, inv_perm);
718: spbas_apply_reordering_cols(matrix_A, permutation);
719: return(0);
720: }
722: PetscErrorCode spbas_pattern_only(PetscInt nrows, PetscInt ncols, PetscInt *ai, PetscInt *aj, spbas_matrix * result)
723: {
724: spbas_matrix retval;
725: PetscInt i, j, i0, r_nnz;
729: /* Copy input values */
730: retval.nrows = nrows;
731: retval.ncols = ncols;
732: retval.nnz = ai[nrows];
734: retval.block_data = PETSC_TRUE;
735: retval.col_idx_type = SPBAS_DIAGONAL_OFFSETS;
737: /* Allocate output matrix */
738: spbas_allocate_pattern(&retval, PETSC_FALSE);
739: for (i=0; i<nrows; i++) retval.row_nnz[i] = ai[i+1]-ai[i];
740: spbas_allocate_data(&retval);
741: /* Copy the structure */
742: for (i = 0; i<retval.nrows; i++) {
743: i0 = ai[i];
744: r_nnz = ai[i+1]-i0;
746: for (j=0; j<r_nnz; j++) {
747: retval.icols[i][j] = aj[i0+j]-i;
748: }
749: }
750: *result = retval;
751: return(0);
752: }
755: /*
756: spbas_mark_row_power:
757: Mark the columns in row 'row' which are nonzero in
758: matrix^2log(marker).
759: */
760: PetscErrorCode spbas_mark_row_power(PetscInt *iwork, /* marker-vector */
761: PetscInt row, /* row for which the columns are marked */
762: spbas_matrix * in_matrix, /* matrix for which the power is being calculated */
763: PetscInt marker, /* marker-value: 2^power */
764: PetscInt minmrk, /* lower bound for marked points */
765: PetscInt maxmrk) /* upper bound for marked points */
766: {
768: PetscInt i,j, nnz;
771: nnz = in_matrix->row_nnz[row];
773: /* For higher powers, call this function recursively */
774: if (marker>1) {
775: for (i=0; i<nnz; i++) {
776: j = row + in_matrix->icols[row][i];
777: if (minmrk<=j && j<maxmrk && iwork[j] < marker) {
778: spbas_mark_row_power(iwork, row + in_matrix->icols[row][i],in_matrix, marker/2,minmrk,maxmrk);
779: iwork[j] |= marker;
780: }
781: }
782: } else {
783: /* Mark the columns reached */
784: for (i=0; i<nnz; i++) {
785: j = row + in_matrix->icols[row][i];
786: if (minmrk<=j && j<maxmrk) iwork[j] |= 1;
787: }
788: }
789: return(0);
790: }
793: /*
794: spbas_power
795: Calculate sparseness patterns for incomplete Cholesky decompositions
796: of a given order: (almost) all nonzeros of the matrix^(order+1) which
797: are inside the band width are found and stored in the output sparseness
798: pattern.
799: */
800: PetscErrorCode spbas_power(spbas_matrix in_matrix,PetscInt power, spbas_matrix * result)
801: {
802: spbas_matrix retval;
803: PetscInt nrows = in_matrix.nrows;
804: PetscInt ncols = in_matrix.ncols;
805: PetscInt i, j, kend;
806: PetscInt nnz, inz;
807: PetscInt *iwork;
808: PetscInt marker;
809: PetscInt maxmrk=0;
813: if (in_matrix.col_idx_type != SPBAS_DIAGONAL_OFFSETS) SETERRQ(PETSC_COMM_SELF, PETSC_ERR_SUP_SYS,"must have diagonal offsets in pattern\n");
814: if (ncols != nrows) SETERRQ(PETSC_COMM_SELF, PETSC_ERR_ARG_INCOMP, "Dimension error\n");
815: if (in_matrix.values) SETERRQ(PETSC_COMM_SELF, PETSC_ERR_ARG_INCOMP, "Input array must be sparseness pattern (no values)");
816: if (power<=0) SETERRQ(PETSC_COMM_SELF, PETSC_ERR_SUP_SYS, "Power must be 1 or up");
818: /* Copy input values*/
819: retval.nrows = ncols;
820: retval.ncols = nrows;
821: retval.nnz = 0;
822: retval.col_idx_type = SPBAS_DIAGONAL_OFFSETS;
823: retval.block_data = PETSC_FALSE;
825: /* Allocate sparseness pattern */
826: spbas_allocate_pattern(&retval, in_matrix.values ? PETSC_TRUE : PETSC_FALSE);
828: /* Allocate marker array: note sure the max needed so use the max of the two */
829: PetscCalloc1(PetscMax(ncols,nrows), &iwork);
831: /* Calculate marker values */
832: marker = 1; for (i=1; i<power; i++) marker*=2;
834: for (i=0; i<nrows; i++) {
835: /* Calculate the pattern for each row */
837: nnz = in_matrix.row_nnz[i];
838: kend = i+in_matrix.icols[i][nnz-1];
839: if (maxmrk<=kend) maxmrk=kend+1;
840: spbas_mark_row_power(iwork, i, &in_matrix, marker, i, maxmrk);
842: /* Count the columns*/
843: nnz = 0;
844: for (j=i; j<maxmrk; j++) nnz+= (iwork[j]!=0);
846: /* Allocate the column indices */
847: retval.row_nnz[i] = nnz;
848: PetscMalloc1(nnz,&retval.icols[i]);
850: /* Administrate the column indices */
851: inz = 0;
852: for (j=i; j<maxmrk; j++) {
853: if (iwork[j]) {
854: retval.icols[i][inz] = j-i;
855: inz++;
856: iwork[j] = 0;
857: }
858: }
859: retval.nnz += nnz;
860: };
861: PetscFree(iwork);
862: *result = retval;
863: return(0);
864: }
868: /*
869: spbas_keep_upper:
870: remove the lower part of the matrix: keep the upper part
871: */
872: PetscErrorCode spbas_keep_upper(spbas_matrix * inout_matrix)
873: {
874: PetscInt i, j;
875: PetscInt jstart;
878: if (inout_matrix->block_data) SETERRQ(PETSC_COMM_SELF, PETSC_ERR_SUP_SYS, "Not yet for block data matrices\n");
879: for (i=0; i<inout_matrix->nrows; i++) {
880: for (jstart=0; (jstart<inout_matrix->row_nnz[i]) && (inout_matrix->icols[i][jstart]<0); jstart++) {}
881: if (jstart>0) {
882: for (j=0; j<inout_matrix->row_nnz[i]-jstart; j++) {
883: inout_matrix->icols[i][j] = inout_matrix->icols[i][j+jstart];
884: }
886: if (inout_matrix->values) {
887: for (j=0; j<inout_matrix->row_nnz[i]-jstart; j++) {
888: inout_matrix->values[i][j] = inout_matrix->values[i][j+jstart];
889: }
890: }
892: inout_matrix->row_nnz[i] -= jstart;
894: inout_matrix->icols[i] = (PetscInt*) realloc((void*) inout_matrix->icols[i], inout_matrix->row_nnz[i]*sizeof(PetscInt));
896: if (inout_matrix->values) {
897: inout_matrix->values[i] = (PetscScalar*) realloc((void*) inout_matrix->values[i], inout_matrix->row_nnz[i]*sizeof(PetscScalar));
898: }
899: inout_matrix->nnz -= jstart;
900: }
901: }
902: return(0);
903: }