Actual source code: fdaij.c
petsc-3.7.3 2016-08-01
2: #include <../src/mat/impls/aij/seq/aij.h>
3: #include <../src/mat/impls/baij/seq/baij.h>
4: #include <petsc/private/isimpl.h>
6: /*
7: This routine is shared by SeqAIJ and SeqBAIJ matrices,
8: since it operators only on the nonzero structure of the elements or blocks.
9: */
12: PetscErrorCode MatFDColoringCreate_SeqXAIJ(Mat mat,ISColoring iscoloring,MatFDColoring c)
13: {
15: PetscInt bs,nis=iscoloring->n,m=mat->rmap->n;
16: PetscBool isBAIJ;
19: /* set default brows and bcols for speedup inserting the dense matrix into sparse Jacobian */
20: MatGetBlockSize(mat,&bs);
21: PetscObjectTypeCompare((PetscObject)mat,MATSEQBAIJ,&isBAIJ);
22: if (isBAIJ) {
23: c->brows = m;
24: c->bcols = 1;
25: } else { /* seqaij matrix */
26: /* bcols is chosen s.t. dy-array takes 50% of memory space as mat */
27: Mat_SeqAIJ *spA = (Mat_SeqAIJ*)mat->data;
28: PetscReal mem;
29: PetscInt nz,brows,bcols;
31: bs = 1; /* only bs=1 is supported for SeqAIJ matrix */
33: nz = spA->nz;
34: mem = nz*(sizeof(PetscScalar) + sizeof(PetscInt)) + 3*m*sizeof(PetscInt);
35: bcols = (PetscInt)(0.5*mem /(m*sizeof(PetscScalar)));
36: brows = 1000/bcols;
37: if (bcols > nis) bcols = nis;
38: if (brows == 0 || brows > m) brows = m;
39: c->brows = brows;
40: c->bcols = bcols;
41: }
43: c->M = mat->rmap->N/bs; /* set total rows, columns and local rows */
44: c->N = mat->cmap->N/bs;
45: c->m = mat->rmap->N/bs;
46: c->rstart = 0;
47: c->ncolors = nis;
48: c->ctype = IS_COLORING_GHOSTED;
49: return(0);
50: }
52: /*
53: Reorder Jentry such that blocked brows*bols of entries from dense matrix are inserted into Jacobian for improved cache performance
54: Input Parameters:
55: + mat - the matrix containing the nonzero structure of the Jacobian
56: . color - the coloring context
57: - nz - number of local non-zeros in mat
58: */
61: PetscErrorCode MatFDColoringSetUpBlocked_AIJ_Private(Mat mat,MatFDColoring c,PetscInt nz)
62: {
64: PetscInt i,j,nrows,nbcols,brows=c->brows,bcols=c->bcols,mbs=c->m,nis=c->ncolors;
65: PetscInt *color_start,*row_start,*nrows_new,nz_new,row_end;
68: if (brows < 1 || brows > mbs) brows = mbs;
69: PetscMalloc2(bcols+1,&color_start,bcols,&row_start);
70: PetscMalloc1(nis,&nrows_new);
71: PetscMalloc1(bcols*mat->rmap->n,&c->dy);
72: PetscLogObjectMemory((PetscObject)c,bcols*mat->rmap->n*sizeof(PetscScalar));
74: nz_new = 0;
75: nbcols = 0;
76: color_start[bcols] = 0;
78: if (c->htype[0] == 'd') { /* ---- c->htype == 'ds', use MatEntry --------*/
79: MatEntry *Jentry_new,*Jentry=c->matentry;
80: PetscMalloc1(nz,&Jentry_new);
81: for (i=0; i<nis; i+=bcols) { /* loop over colors */
82: if (i + bcols > nis) {
83: color_start[nis - i] = color_start[bcols];
84: bcols = nis - i;
85: }
87: color_start[0] = color_start[bcols];
88: for (j=0; j<bcols; j++) {
89: color_start[j+1] = c->nrows[i+j] + color_start[j];
90: row_start[j] = 0;
91: }
93: row_end = brows;
94: if (row_end > mbs) row_end = mbs;
96: while (row_end <= mbs) { /* loop over block rows */
97: for (j=0; j<bcols; j++) { /* loop over block columns */
98: nrows = c->nrows[i+j];
99: nz = color_start[j];
100: while (row_start[j] < nrows) {
101: if (Jentry[nz].row >= row_end) {
102: color_start[j] = nz;
103: break;
104: } else { /* copy Jentry[nz] to Jentry_new[nz_new] */
105: Jentry_new[nz_new].row = Jentry[nz].row + j*mbs; /* index in dy-array */
106: Jentry_new[nz_new].col = Jentry[nz].col;
107: Jentry_new[nz_new].valaddr = Jentry[nz].valaddr;
108: nz_new++; nz++; row_start[j]++;
109: }
110: }
111: }
112: if (row_end == mbs) break;
113: row_end += brows;
114: if (row_end > mbs) row_end = mbs;
115: }
116: nrows_new[nbcols++] = nz_new;
117: }
118: PetscFree(Jentry);
119: c->matentry = Jentry_new;
120: } else { /* --------- c->htype == 'wp', use MatEntry2 ------------------*/
121: MatEntry2 *Jentry2_new,*Jentry2=c->matentry2;
122: PetscMalloc1(nz,&Jentry2_new);
123: for (i=0; i<nis; i+=bcols) { /* loop over colors */
124: if (i + bcols > nis) {
125: color_start[nis - i] = color_start[bcols];
126: bcols = nis - i;
127: }
129: color_start[0] = color_start[bcols];
130: for (j=0; j<bcols; j++) {
131: color_start[j+1] = c->nrows[i+j] + color_start[j];
132: row_start[j] = 0;
133: }
135: row_end = brows;
136: if (row_end > mbs) row_end = mbs;
138: while (row_end <= mbs) { /* loop over block rows */
139: for (j=0; j<bcols; j++) { /* loop over block columns */
140: nrows = c->nrows[i+j];
141: nz = color_start[j];
142: while (row_start[j] < nrows) {
143: if (Jentry2[nz].row >= row_end) {
144: color_start[j] = nz;
145: break;
146: } else { /* copy Jentry2[nz] to Jentry2_new[nz_new] */
147: Jentry2_new[nz_new].row = Jentry2[nz].row + j*mbs; /* index in dy-array */
148: Jentry2_new[nz_new].valaddr = Jentry2[nz].valaddr;
149: nz_new++; nz++; row_start[j]++;
150: }
151: }
152: }
153: if (row_end == mbs) break;
154: row_end += brows;
155: if (row_end > mbs) row_end = mbs;
156: }
157: nrows_new[nbcols++] = nz_new;
158: }
159: PetscFree(Jentry2);
160: c->matentry2 = Jentry2_new;
161: } /* ---------------------------------------------*/
163: PetscFree2(color_start,row_start);
165: for (i=nbcols-1; i>0; i--) nrows_new[i] -= nrows_new[i-1];
166: PetscFree(c->nrows);
167: c->nrows = nrows_new;
168: return(0);
169: }
173: PetscErrorCode MatFDColoringSetUp_SeqXAIJ(Mat mat,ISColoring iscoloring,MatFDColoring c)
174: {
176: PetscInt i,n,nrows,mbs=c->m,j,k,m,ncols,col,nis=iscoloring->n,*rowhit,bs,bs2,*spidx,nz;
177: const PetscInt *is,*row,*ci,*cj;
178: IS *isa;
179: PetscBool isBAIJ;
180: PetscScalar *A_val,**valaddrhit;
181: MatEntry *Jentry;
182: MatEntry2 *Jentry2;
185: ISColoringGetIS(iscoloring,PETSC_IGNORE,&isa);
187: MatGetBlockSize(mat,&bs);
188: PetscObjectTypeCompare((PetscObject)mat,MATSEQBAIJ,&isBAIJ);
189: if (isBAIJ) {
190: Mat_SeqBAIJ *spA = (Mat_SeqBAIJ*)mat->data;
191: A_val = spA->a;
192: nz = spA->nz;
193: } else {
194: Mat_SeqAIJ *spA = (Mat_SeqAIJ*)mat->data;
195: A_val = spA->a;
196: nz = spA->nz;
197: bs = 1; /* only bs=1 is supported for SeqAIJ matrix */
198: }
200: PetscMalloc1(nis,&c->ncolumns);
201: PetscMalloc1(nis,&c->columns);
202: PetscMalloc1(nis,&c->nrows);
203: PetscLogObjectMemory((PetscObject)c,3*nis*sizeof(PetscInt));
205: if (c->htype[0] == 'd') {
206: PetscMalloc1(nz,&Jentry);
207: PetscLogObjectMemory((PetscObject)c,nz*sizeof(MatEntry));
208: c->matentry = Jentry;
209: } else if (c->htype[0] == 'w') {
210: PetscMalloc1(nz,&Jentry2);
211: PetscLogObjectMemory((PetscObject)c,nz*sizeof(MatEntry2));
212: c->matentry2 = Jentry2;
213: } else SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"htype is not supported");
215: if (isBAIJ) {
216: MatGetColumnIJ_SeqBAIJ_Color(mat,0,PETSC_FALSE,PETSC_FALSE,&ncols,&ci,&cj,&spidx,NULL);
217: } else {
218: MatGetColumnIJ_SeqAIJ_Color(mat,0,PETSC_FALSE,PETSC_FALSE,&ncols,&ci,&cj,&spidx,NULL);
219: }
221: PetscMalloc2(c->m,&rowhit,c->m,&valaddrhit);
222: PetscMemzero(rowhit,c->m*sizeof(PetscInt));
224: nz = 0;
225: for (i=0; i<nis; i++) { /* loop over colors */
226: ISGetLocalSize(isa[i],&n);
227: ISGetIndices(isa[i],&is);
229: c->ncolumns[i] = n;
230: if (n) {
231: PetscMalloc1(n,&c->columns[i]);
232: PetscLogObjectMemory((PetscObject)c,n*sizeof(PetscInt));
233: PetscMemcpy(c->columns[i],is,n*sizeof(PetscInt));
234: } else {
235: c->columns[i] = 0;
236: }
238: /* fast, crude version requires O(N*N) work */
239: bs2 = bs*bs;
240: nrows = 0;
241: for (j=0; j<n; j++) { /* loop over columns */
242: col = is[j];
243: row = cj + ci[col];
244: m = ci[col+1] - ci[col];
245: nrows += m;
246: for (k=0; k<m; k++) { /* loop over columns marking them in rowhit */
247: rowhit[*row] = col + 1;
248: valaddrhit[*row++] = &A_val[bs2*spidx[ci[col] + k]];
249: }
250: }
251: c->nrows[i] = nrows; /* total num of rows for this color */
253: if (c->htype[0] == 'd') {
254: for (j=0; j<mbs; j++) { /* loop over rows */
255: if (rowhit[j]) {
256: Jentry[nz].row = j; /* local row index */
257: Jentry[nz].col = rowhit[j] - 1; /* local column index */
258: Jentry[nz].valaddr = valaddrhit[j]; /* address of mat value for this entry */
259: nz++;
260: rowhit[j] = 0.0; /* zero rowhit for reuse */
261: }
262: }
263: } else { /* c->htype == 'wp' */
264: for (j=0; j<mbs; j++) { /* loop over rows */
265: if (rowhit[j]) {
266: Jentry2[nz].row = j; /* local row index */
267: Jentry2[nz].valaddr = valaddrhit[j]; /* address of mat value for this entry */
268: nz++;
269: rowhit[j] = 0.0; /* zero rowhit for reuse */
270: }
271: }
272: }
273: ISRestoreIndices(isa[i],&is);
274: }
276: if (c->bcols > 1) { /* reorder Jentry for faster MatFDColoringApply() */
277: MatFDColoringSetUpBlocked_AIJ_Private(mat,c,nz);
278: }
280: if (isBAIJ) {
281: MatRestoreColumnIJ_SeqBAIJ_Color(mat,0,PETSC_FALSE,PETSC_FALSE,&ncols,&ci,&cj,&spidx,NULL);
282: PetscMalloc1(bs*mat->rmap->n,&c->dy);
283: PetscLogObjectMemory((PetscObject)c,bs*mat->rmap->n*sizeof(PetscScalar));
284: } else {
285: MatRestoreColumnIJ_SeqAIJ_Color(mat,0,PETSC_FALSE,PETSC_FALSE,&ncols,&ci,&cj,&spidx,NULL);
286: }
287: PetscFree2(rowhit,valaddrhit);
288: ISColoringRestoreIS(iscoloring,&isa);
290: VecCreateGhost(PetscObjectComm((PetscObject)mat),mat->rmap->n,PETSC_DETERMINE,0,NULL,&c->vscale);
291: PetscInfo3(c,"ncolors %D, brows %D and bcols %D are used.\n",c->ncolors,c->brows,c->bcols);
292: return(0);
293: }