xref: /petsc/src/mat/impls/aij/seq/fdaij.c (revision 6a98f8dc3f2c9149905a87dc2e9d0fedaf64e09a)
1 #include <../src/mat/impls/aij/seq/aij.h>
2 #include <../src/mat/impls/baij/seq/baij.h>
3 #include <../src/mat/impls/sell/seq/sell.h>
4 #include <petsc/private/isimpl.h>
5 
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 */
10 PetscErrorCode MatFDColoringCreate_SeqXAIJ(Mat mat,ISColoring iscoloring,MatFDColoring c)
11 {
12   PetscErrorCode ierr;
13   PetscInt       bs,nis=iscoloring->n,m=mat->rmap->n;
14   PetscBool      isBAIJ,isSELL;
15 
16   PetscFunctionBegin;
17   /* set default brows and bcols for speedup inserting the dense matrix into sparse Jacobian */
18   ierr = MatGetBlockSize(mat,&bs);CHKERRQ(ierr);
19   ierr = PetscObjectBaseTypeCompare((PetscObject)mat,MATSEQBAIJ,&isBAIJ);CHKERRQ(ierr);
20   ierr = PetscObjectTypeCompare((PetscObject)mat,MATSEQSELL,&isSELL);CHKERRQ(ierr);
21   if (isBAIJ) {
22     c->brows = m;
23     c->bcols = 1;
24   } else { /* seqaij matrix */
25     /* bcols is chosen s.t. dy-array takes 50% of memory space as mat */
26     PetscReal  mem;
27     PetscInt   nz,brows,bcols;
28     if (isSELL) {
29       Mat_SeqSELL *spA = (Mat_SeqSELL*)mat->data;
30       nz = spA->nz;
31     } else {
32       Mat_SeqAIJ *spA = (Mat_SeqAIJ*)mat->data;
33       nz = spA->nz;
34     }
35 
36     bs    = 1; /* only bs=1 is supported for SeqAIJ matrix */
37     mem   = nz*(sizeof(PetscScalar) + sizeof(PetscInt)) + 3*m*sizeof(PetscInt);
38     bcols = (PetscInt)(0.5*mem /(m*sizeof(PetscScalar)));
39     brows = 1000/bcols;
40     if (bcols > nis) bcols = nis;
41     if (brows == 0 || brows > m) brows = m;
42     c->brows = brows;
43     c->bcols = bcols;
44   }
45 
46   c->M       = mat->rmap->N/bs;   /* set total rows, columns and local rows */
47   c->N       = mat->cmap->N/bs;
48   c->m       = mat->rmap->N/bs;
49   c->rstart  = 0;
50   c->ncolors = nis;
51   c->ctype   = iscoloring->ctype;
52   PetscFunctionReturn(0);
53 }
54 
55 /*
56  Reorder Jentry such that blocked brows*bols of entries from dense matrix are inserted into Jacobian for improved cache performance
57    Input Parameters:
58 +  mat - the matrix containing the nonzero structure of the Jacobian
59 .  color - the coloring context
60 -  nz - number of local non-zeros in mat
61 */
62 PetscErrorCode MatFDColoringSetUpBlocked_AIJ_Private(Mat mat,MatFDColoring c,PetscInt nz)
63 {
64   PetscErrorCode ierr;
65   PetscInt       i,j,nrows,nbcols,brows=c->brows,bcols=c->bcols,mbs=c->m,nis=c->ncolors;
66   PetscInt       *color_start,*row_start,*nrows_new,nz_new,row_end;
67 
68   PetscFunctionBegin;
69   if (brows < 1 || brows > mbs) brows = mbs;
70   ierr = PetscMalloc2(bcols+1,&color_start,bcols,&row_start);CHKERRQ(ierr);
71   ierr = PetscCalloc1(nis,&nrows_new);CHKERRQ(ierr);
72   ierr = PetscMalloc1(bcols*mat->rmap->n,&c->dy);CHKERRQ(ierr);
73   ierr = PetscLogObjectMemory((PetscObject)c,bcols*mat->rmap->n*sizeof(PetscScalar));CHKERRQ(ierr);
74 
75   nz_new = 0;
76   nbcols = 0;
77   color_start[bcols] = 0;
78 
79   if (c->htype[0] == 'd') { /* ----  c->htype == 'ds', use MatEntry --------*/
80     MatEntry *Jentry_new,*Jentry=c->matentry;
81 
82     ierr = PetscMalloc1(nz,&Jentry_new);CHKERRQ(ierr);
83     for (i=0; i<nis; i+=bcols) { /* loop over colors */
84       if (i + bcols > nis) {
85         color_start[nis - i] = color_start[bcols];
86         bcols                = nis - i;
87       }
88 
89       color_start[0] = color_start[bcols];
90       for (j=0; j<bcols; j++) {
91         color_start[j+1] = c->nrows[i+j] + color_start[j];
92         row_start[j]     = 0;
93       }
94 
95       row_end = brows;
96       if (row_end > mbs) row_end = mbs;
97 
98       while (row_end <= mbs) {   /* loop over block rows */
99         for (j=0; j<bcols; j++) {       /* loop over block columns */
100           nrows = c->nrows[i+j];
101           nz    = color_start[j];
102           while (row_start[j] < nrows) {
103             if (Jentry[nz].row >= row_end) {
104               color_start[j] = nz;
105               break;
106             } else { /* copy Jentry[nz] to Jentry_new[nz_new] */
107               Jentry_new[nz_new].row     = Jentry[nz].row + j*mbs; /* index in dy-array */
108               Jentry_new[nz_new].col     = Jentry[nz].col;
109               Jentry_new[nz_new].valaddr = Jentry[nz].valaddr;
110               nz_new++; nz++; row_start[j]++;
111             }
112           }
113         }
114         if (row_end == mbs) break;
115         row_end += brows;
116         if (row_end > mbs) row_end = mbs;
117       }
118       nrows_new[nbcols++] = nz_new;
119     }
120     ierr = PetscFree(Jentry);CHKERRQ(ierr);
121     c->matentry = Jentry_new;
122   } else { /* ---------  c->htype == 'wp', use MatEntry2 ------------------*/
123     MatEntry2 *Jentry2_new,*Jentry2=c->matentry2;
124 
125     ierr = PetscMalloc1(nz,&Jentry2_new);CHKERRQ(ierr);
126     for (i=0; i<nis; i+=bcols) { /* loop over colors */
127       if (i + bcols > nis) {
128         color_start[nis - i] = color_start[bcols];
129         bcols                = nis - i;
130       }
131 
132       color_start[0] = color_start[bcols];
133       for (j=0; j<bcols; j++) {
134         color_start[j+1] = c->nrows[i+j] + color_start[j];
135         row_start[j]     = 0;
136       }
137 
138       row_end = brows;
139       if (row_end > mbs) row_end = mbs;
140 
141       while (row_end <= mbs) {   /* loop over block rows */
142         for (j=0; j<bcols; j++) {       /* loop over block columns */
143           nrows = c->nrows[i+j];
144           nz    = color_start[j];
145           while (row_start[j] < nrows) {
146             if (Jentry2[nz].row >= row_end) {
147               color_start[j] = nz;
148               break;
149             } else { /* copy Jentry2[nz] to Jentry2_new[nz_new] */
150               Jentry2_new[nz_new].row     = Jentry2[nz].row + j*mbs; /* index in dy-array */
151               Jentry2_new[nz_new].valaddr = Jentry2[nz].valaddr;
152               nz_new++; nz++; row_start[j]++;
153             }
154           }
155         }
156         if (row_end == mbs) break;
157         row_end += brows;
158         if (row_end > mbs) row_end = mbs;
159       }
160       nrows_new[nbcols++] = nz_new;
161     }
162     ierr = PetscFree(Jentry2);CHKERRQ(ierr);
163     c->matentry2 = Jentry2_new;
164   } /* ---------------------------------------------*/
165 
166   ierr = PetscFree2(color_start,row_start);CHKERRQ(ierr);
167 
168   for (i=nbcols-1; i>0; i--) nrows_new[i] -= nrows_new[i-1];
169   ierr = PetscFree(c->nrows);CHKERRQ(ierr);
170   c->nrows = nrows_new;
171   PetscFunctionReturn(0);
172 }
173 
174 PetscErrorCode MatFDColoringSetUp_SeqXAIJ(Mat mat,ISColoring iscoloring,MatFDColoring c)
175 {
176   PetscErrorCode    ierr;
177   PetscInt          i,n,nrows,mbs=c->m,j,k,m,ncols,col,nis=iscoloring->n,*rowhit,bs,bs2,*spidx,nz,tmp;
178   const PetscInt    *is,*row,*ci,*cj;
179   PetscBool         isBAIJ,isSELL;
180   const PetscScalar *A_val;
181   PetscScalar       **valaddrhit;
182   MatEntry          *Jentry;
183   MatEntry2         *Jentry2;
184 
185   PetscFunctionBegin;
186   ierr = ISColoringGetIS(iscoloring,PETSC_OWN_POINTER,PETSC_IGNORE,&c->isa);CHKERRQ(ierr);
187 
188   ierr = MatGetBlockSize(mat,&bs);CHKERRQ(ierr);
189   ierr = PetscObjectBaseTypeCompare((PetscObject)mat,MATSEQBAIJ,&isBAIJ);CHKERRQ(ierr);
190   ierr = PetscObjectTypeCompare((PetscObject)mat,MATSEQSELL,&isSELL);CHKERRQ(ierr);
191   if (isBAIJ) {
192     Mat_SeqBAIJ *spA = (Mat_SeqBAIJ*)mat->data;
193 
194     A_val = spA->a;
195     nz    = spA->nz;
196   } else if (isSELL) {
197     Mat_SeqSELL *spA = (Mat_SeqSELL*)mat->data;
198 
199     A_val = spA->val;
200     nz    = spA->nz;
201     bs    = 1; /* only bs=1 is supported for SeqSELL matrix */
202   } else {
203     Mat_SeqAIJ *spA = (Mat_SeqAIJ*)mat->data;
204 
205     A_val = spA->a;
206     nz    = spA->nz;
207     bs    = 1; /* only bs=1 is supported for SeqAIJ matrix */
208   }
209 
210   ierr = PetscMalloc2(nis,&c->ncolumns,nis,&c->columns);CHKERRQ(ierr);
211   ierr = PetscMalloc1(nis,&c->nrows);CHKERRQ(ierr); /* nrows is freeed separately from ncolumns and columns */
212   ierr = PetscLogObjectMemory((PetscObject)c,3*nis*sizeof(PetscInt));CHKERRQ(ierr);
213 
214   if (c->htype[0] == 'd') {
215     ierr        = PetscMalloc1(nz,&Jentry);CHKERRQ(ierr);
216     ierr        = PetscLogObjectMemory((PetscObject)c,nz*sizeof(MatEntry));CHKERRQ(ierr);
217     c->matentry = Jentry;
218   } else if (c->htype[0] == 'w') {
219     ierr         = PetscMalloc1(nz,&Jentry2);CHKERRQ(ierr);
220     ierr         = PetscLogObjectMemory((PetscObject)c,nz*sizeof(MatEntry2));CHKERRQ(ierr);
221     c->matentry2 = Jentry2;
222   } else SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"htype is not supported");
223 
224   if (isBAIJ) {
225     ierr = MatGetColumnIJ_SeqBAIJ_Color(mat,0,PETSC_FALSE,PETSC_FALSE,&ncols,&ci,&cj,&spidx,NULL);CHKERRQ(ierr);
226   } else if (isSELL) {
227     ierr = MatGetColumnIJ_SeqSELL_Color(mat,0,PETSC_FALSE,PETSC_FALSE,&ncols,&ci,&cj,&spidx,NULL);CHKERRQ(ierr);
228   } else {
229     ierr = MatGetColumnIJ_SeqAIJ_Color(mat,0,PETSC_FALSE,PETSC_FALSE,&ncols,&ci,&cj,&spidx,NULL);CHKERRQ(ierr);
230   }
231 
232   ierr = PetscCalloc1(c->m,&rowhit);CHKERRQ(ierr);
233   ierr = PetscMalloc1(c->m,&valaddrhit);CHKERRQ(ierr);
234 
235   nz = 0;
236   for (i=0; i<nis; i++) { /* loop over colors */
237     ierr = ISGetLocalSize(c->isa[i],&n);CHKERRQ(ierr);
238     ierr = ISGetIndices(c->isa[i],&is);CHKERRQ(ierr);
239 
240     c->ncolumns[i] = n;
241     c->columns[i]  = (PetscInt*)is;
242     /* note: we know that c->isa is going to be around as long at the c->columns values */
243     ierr = ISRestoreIndices(c->isa[i],&is);CHKERRQ(ierr);
244 
245     /* fast, crude version requires O(N*N) work */
246     bs2   = bs*bs;
247     nrows = 0;
248     for (j=0; j<n; j++) {  /* loop over columns */
249       col    = is[j];
250       tmp    = ci[col];
251       row    = cj + tmp;
252       m      = ci[col+1] - tmp;
253       nrows += m;
254       for (k=0; k<m; k++) {  /* loop over columns marking them in rowhit */
255         rowhit[*row]       = col + 1;
256         valaddrhit[*row++] = (PetscScalar*)&A_val[bs2*spidx[tmp + k]];
257       }
258     }
259     c->nrows[i] = nrows; /* total num of rows for this color */
260 
261     if (c->htype[0] == 'd') {
262       for (j=0; j<mbs; j++) { /* loop over rows */
263         if (rowhit[j]) {
264           Jentry[nz].row     = j;              /* local row index */
265           Jentry[nz].col     = rowhit[j] - 1;  /* local column index */
266           Jentry[nz].valaddr = valaddrhit[j];  /* address of mat value for this entry */
267           nz++;
268           rowhit[j] = 0.0;                     /* zero rowhit for reuse */
269         }
270       }
271     }  else { /* c->htype == 'wp' */
272       for (j=0; j<mbs; j++) { /* loop over rows */
273         if (rowhit[j]) {
274           Jentry2[nz].row     = j;              /* local row index */
275           Jentry2[nz].valaddr = valaddrhit[j];  /* address of mat value for this entry */
276           nz++;
277           rowhit[j] = 0.0;                     /* zero rowhit for reuse */
278         }
279       }
280     }
281   }
282 
283   if (c->bcols > 1) {  /* reorder Jentry for faster MatFDColoringApply() */
284     ierr = MatFDColoringSetUpBlocked_AIJ_Private(mat,c,nz);CHKERRQ(ierr);
285   }
286 
287   if (isBAIJ) {
288     ierr = MatRestoreColumnIJ_SeqBAIJ_Color(mat,0,PETSC_FALSE,PETSC_FALSE,&ncols,&ci,&cj,&spidx,NULL);CHKERRQ(ierr);
289     ierr = PetscMalloc1(bs*mat->rmap->n,&c->dy);CHKERRQ(ierr);
290     ierr = PetscLogObjectMemory((PetscObject)c,bs*mat->rmap->n*sizeof(PetscScalar));CHKERRQ(ierr);
291   } else if (isSELL) {
292     ierr = MatRestoreColumnIJ_SeqSELL_Color(mat,0,PETSC_FALSE,PETSC_FALSE,&ncols,&ci,&cj,&spidx,NULL);CHKERRQ(ierr);
293   } else {
294     ierr = MatRestoreColumnIJ_SeqAIJ_Color(mat,0,PETSC_FALSE,PETSC_FALSE,&ncols,&ci,&cj,&spidx,NULL);CHKERRQ(ierr);
295   }
296   ierr = PetscFree(rowhit);CHKERRQ(ierr);
297   ierr = PetscFree(valaddrhit);CHKERRQ(ierr);
298   ierr = ISColoringRestoreIS(iscoloring,PETSC_OWN_POINTER,&c->isa);CHKERRQ(ierr);
299 
300   ierr = VecCreateGhost(PetscObjectComm((PetscObject)mat),mat->rmap->n,PETSC_DETERMINE,0,NULL,&c->vscale);CHKERRQ(ierr);
301   ierr = PetscInfo3(c,"ncolors %D, brows %D and bcols %D are used.\n",c->ncolors,c->brows,c->bcols);CHKERRQ(ierr);
302   PetscFunctionReturn(0);
303 }
304