xref: /petsc/src/mat/impls/aij/mpi/fdmpiaij.c (revision 7d5fd1e4d9337468ad3f05b65b7facdcd2dfd2a4) !
1 #include <../src/mat/impls/sell/mpi/mpisell.h>
2 #include <../src/mat/impls/aij/mpi/mpiaij.h>
3 #include <../src/mat/impls/baij/mpi/mpibaij.h>
4 #include <petsc/private/isimpl.h>
5 
6 PetscErrorCode MatFDColoringApply_BAIJ(Mat J,MatFDColoring coloring,Vec x1,void *sctx)
7 {
8   PetscErrorCode    (*f)(void*,Vec,Vec,void*)=(PetscErrorCode (*)(void*,Vec,Vec,void*))coloring->f;
9   PetscErrorCode    ierr;
10   PetscInt          k,cstart,cend,l,row,col,nz,spidx,i,j;
11   PetscScalar       dx=0.0,*w3_array,*dy_i,*dy=coloring->dy;
12   PetscScalar       *vscale_array;
13   const PetscScalar *xx;
14   PetscReal         epsilon=coloring->error_rel,umin=coloring->umin,unorm;
15   Vec               w1=coloring->w1,w2=coloring->w2,w3,vscale=coloring->vscale;
16   void              *fctx=coloring->fctx;
17   PetscInt          ctype=coloring->ctype,nxloc,nrows_k;
18   PetscScalar       *valaddr;
19   MatEntry          *Jentry=coloring->matentry;
20   MatEntry2         *Jentry2=coloring->matentry2;
21   const PetscInt    ncolors=coloring->ncolors,*ncolumns=coloring->ncolumns,*nrows=coloring->nrows;
22   PetscInt          bs=J->rmap->bs;
23 
24   PetscFunctionBegin;
25   ierr = VecBindToCPU(x1,PETSC_TRUE);CHKERRQ(ierr);
26   /* (1) Set w1 = F(x1) */
27   if (!coloring->fset) {
28     ierr = PetscLogEventBegin(MAT_FDColoringFunction,coloring,0,0,0);CHKERRQ(ierr);
29     ierr = (*f)(sctx,x1,w1,fctx);CHKERRQ(ierr);
30     ierr = PetscLogEventEnd(MAT_FDColoringFunction,coloring,0,0,0);CHKERRQ(ierr);
31   } else {
32     coloring->fset = PETSC_FALSE;
33   }
34 
35   /* (2) Compute vscale = 1./dx - the local scale factors, including ghost points */
36   ierr = VecGetLocalSize(x1,&nxloc);CHKERRQ(ierr);
37   if (coloring->htype[0] == 'w') {
38     /* vscale = dx is a constant scalar */
39     ierr = VecNorm(x1,NORM_2,&unorm);CHKERRQ(ierr);
40     dx = 1.0/(PetscSqrtReal(1.0 + unorm)*epsilon);
41   } else {
42     ierr = VecGetArrayRead(x1,&xx);CHKERRQ(ierr);
43     ierr = VecGetArray(vscale,&vscale_array);CHKERRQ(ierr);
44     for (col=0; col<nxloc; col++) {
45       dx = xx[col];
46       if (PetscAbsScalar(dx) < umin) {
47         if (PetscRealPart(dx) >= 0.0)      dx = umin;
48         else if (PetscRealPart(dx) < 0.0) dx = -umin;
49       }
50       dx               *= epsilon;
51       vscale_array[col] = 1.0/dx;
52     }
53     ierr = VecRestoreArrayRead(x1,&xx);CHKERRQ(ierr);
54     ierr = VecRestoreArray(vscale,&vscale_array);CHKERRQ(ierr);
55   }
56   if (ctype == IS_COLORING_GLOBAL && coloring->htype[0] == 'd') {
57     ierr = VecGhostUpdateBegin(vscale,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr);
58     ierr = VecGhostUpdateEnd(vscale,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr);
59   }
60 
61   /* (3) Loop over each color */
62   if (!coloring->w3) {
63     ierr = VecDuplicate(x1,&coloring->w3);CHKERRQ(ierr);
64     /* Vec is used intensively in particular piece of scalar CPU code; won't benefit from bouncing back and forth to the GPU */
65     ierr = VecBindToCPU(coloring->w3,PETSC_TRUE);CHKERRQ(ierr);
66     ierr = PetscLogObjectParent((PetscObject)coloring,(PetscObject)coloring->w3);CHKERRQ(ierr);
67   }
68   w3 = coloring->w3;
69 
70   ierr = VecGetOwnershipRange(x1,&cstart,&cend);CHKERRQ(ierr); /* used by ghosted vscale */
71   if (vscale) {
72     ierr = VecGetArray(vscale,&vscale_array);CHKERRQ(ierr);
73   }
74   nz = 0;
75   for (k=0; k<ncolors; k++) {
76     coloring->currentcolor = k;
77 
78     /*
79       (3-1) Loop over each column associated with color
80       adding the perturbation to the vector w3 = x1 + dx.
81     */
82     ierr = VecCopy(x1,w3);CHKERRQ(ierr);
83     dy_i = dy;
84     for (i=0; i<bs; i++) {     /* Loop over a block of columns */
85       ierr = VecGetArray(w3,&w3_array);CHKERRQ(ierr);
86       if (ctype == IS_COLORING_GLOBAL) w3_array -= cstart; /* shift pointer so global index can be used */
87       if (coloring->htype[0] == 'w') {
88         for (l=0; l<ncolumns[k]; l++) {
89           col            = i + bs*coloring->columns[k][l];  /* local column (in global index!) of the matrix we are probing for */
90           w3_array[col] += 1.0/dx;
91           if (i) w3_array[col-1] -= 1.0/dx; /* resume original w3[col-1] */
92         }
93       } else { /* htype == 'ds' */
94         vscale_array -= cstart; /* shift pointer so global index can be used */
95         for (l=0; l<ncolumns[k]; l++) {
96           col = i + bs*coloring->columns[k][l]; /* local column (in global index!) of the matrix we are probing for */
97           w3_array[col] += 1.0/vscale_array[col];
98           if (i) w3_array[col-1] -=  1.0/vscale_array[col-1]; /* resume original w3[col-1] */
99         }
100         vscale_array += cstart;
101       }
102       if (ctype == IS_COLORING_GLOBAL) w3_array += cstart;
103       ierr = VecRestoreArray(w3,&w3_array);CHKERRQ(ierr);
104 
105       /*
106        (3-2) Evaluate function at w3 = x1 + dx (here dx is a vector of perturbations)
107                            w2 = F(x1 + dx) - F(x1)
108        */
109       ierr = PetscLogEventBegin(MAT_FDColoringFunction,0,0,0,0);CHKERRQ(ierr);
110       ierr = VecPlaceArray(w2,dy_i);CHKERRQ(ierr); /* place w2 to the array dy_i */
111       ierr = (*f)(sctx,w3,w2,fctx);CHKERRQ(ierr);
112       ierr = PetscLogEventEnd(MAT_FDColoringFunction,0,0,0,0);CHKERRQ(ierr);
113       ierr = VecAXPY(w2,-1.0,w1);CHKERRQ(ierr);
114       ierr = VecResetArray(w2);CHKERRQ(ierr);
115       dy_i += nxloc; /* points to dy+i*nxloc */
116     }
117 
118     /*
119      (3-3) Loop over rows of vector, putting results into Jacobian matrix
120     */
121     nrows_k = nrows[k];
122     if (coloring->htype[0] == 'w') {
123       for (l=0; l<nrows_k; l++) {
124         row     = bs*Jentry2[nz].row;   /* local row index */
125         valaddr = Jentry2[nz++].valaddr;
126         spidx   = 0;
127         dy_i    = dy;
128         for (i=0; i<bs; i++) {   /* column of the block */
129           for (j=0; j<bs; j++) { /* row of the block */
130             valaddr[spidx++] = dy_i[row+j]*dx;
131           }
132           dy_i += nxloc; /* points to dy+i*nxloc */
133         }
134       }
135     } else { /* htype == 'ds' */
136       for (l=0; l<nrows_k; l++) {
137         row     = bs*Jentry[nz].row;   /* local row index */
138         col     = bs*Jentry[nz].col;   /* local column index */
139         valaddr = Jentry[nz++].valaddr;
140         spidx   = 0;
141         dy_i    = dy;
142         for (i=0; i<bs; i++) {   /* column of the block */
143           for (j=0; j<bs; j++) { /* row of the block */
144             valaddr[spidx++] = dy_i[row+j]*vscale_array[col+i];
145           }
146           dy_i += nxloc; /* points to dy+i*nxloc */
147         }
148       }
149     }
150   }
151   ierr = MatAssemblyBegin(J,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr);
152   ierr = MatAssemblyEnd(J,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr);
153   if (vscale) {
154     ierr = VecRestoreArray(vscale,&vscale_array);CHKERRQ(ierr);
155   }
156 
157   coloring->currentcolor = -1;
158   ierr = VecBindToCPU(x1,PETSC_FALSE);CHKERRQ(ierr);
159   PetscFunctionReturn(0);
160 }
161 
162 /* this is declared PETSC_EXTERN because it is used by MatFDColoringUseDM() which is in the DM library */
163 PetscErrorCode  MatFDColoringApply_AIJ(Mat J,MatFDColoring coloring,Vec x1,void *sctx)
164 {
165   PetscErrorCode    (*f)(void*,Vec,Vec,void*) = (PetscErrorCode (*)(void*,Vec,Vec,void*))coloring->f;
166   PetscErrorCode    ierr;
167   PetscInt          k,cstart,cend,l,row,col,nz;
168   PetscScalar       dx=0.0,*y,*w3_array;
169   const PetscScalar *xx;
170   PetscScalar       *vscale_array;
171   PetscReal         epsilon=coloring->error_rel,umin=coloring->umin,unorm;
172   Vec               w1=coloring->w1,w2=coloring->w2,w3,vscale=coloring->vscale;
173   void              *fctx=coloring->fctx;
174   ISColoringType    ctype=coloring->ctype;
175   PetscInt          nxloc,nrows_k;
176   MatEntry          *Jentry=coloring->matentry;
177   MatEntry2         *Jentry2=coloring->matentry2;
178   const PetscInt    ncolors=coloring->ncolors,*ncolumns=coloring->ncolumns,*nrows=coloring->nrows;
179 
180   PetscFunctionBegin;
181   ierr = VecBindToCPU(x1,PETSC_TRUE);CHKERRQ(ierr);
182   if ((ctype == IS_COLORING_LOCAL) && (J->ops->fdcoloringapply == MatFDColoringApply_AIJ)) SETERRQ(PetscObjectComm((PetscObject)J),PETSC_ERR_SUP,"Must call MatColoringUseDM() with IS_COLORING_LOCAL");
183   /* (1) Set w1 = F(x1) */
184   if (!coloring->fset) {
185     ierr = PetscLogEventBegin(MAT_FDColoringFunction,0,0,0,0);CHKERRQ(ierr);
186     ierr = (*f)(sctx,x1,w1,fctx);CHKERRQ(ierr);
187     ierr = PetscLogEventEnd(MAT_FDColoringFunction,0,0,0,0);CHKERRQ(ierr);
188   } else {
189     coloring->fset = PETSC_FALSE;
190   }
191 
192   /* (2) Compute vscale = 1./dx - the local scale factors, including ghost points */
193   if (coloring->htype[0] == 'w') {
194     /* vscale = 1./dx is a constant scalar */
195     ierr = VecNorm(x1,NORM_2,&unorm);CHKERRQ(ierr);
196     dx = 1.0/(PetscSqrtReal(1.0 + unorm)*epsilon);
197   } else {
198     ierr = VecGetLocalSize(x1,&nxloc);CHKERRQ(ierr);
199     ierr = VecGetArrayRead(x1,&xx);CHKERRQ(ierr);
200     ierr = VecGetArray(vscale,&vscale_array);CHKERRQ(ierr);
201     for (col=0; col<nxloc; col++) {
202       dx = xx[col];
203       if (PetscAbsScalar(dx) < umin) {
204         if (PetscRealPart(dx) >= 0.0)      dx = umin;
205         else if (PetscRealPart(dx) < 0.0) dx = -umin;
206       }
207       dx               *= epsilon;
208       vscale_array[col] = 1.0/dx;
209     }
210     ierr = VecRestoreArrayRead(x1,&xx);CHKERRQ(ierr);
211     ierr = VecRestoreArray(vscale,&vscale_array);CHKERRQ(ierr);
212   }
213   if (ctype == IS_COLORING_GLOBAL && coloring->htype[0] == 'd') {
214     ierr = VecGhostUpdateBegin(vscale,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr);
215     ierr = VecGhostUpdateEnd(vscale,INSERT_VALUES,SCATTER_FORWARD);CHKERRQ(ierr);
216   }
217 
218   /* (3) Loop over each color */
219   if (!coloring->w3) {
220     ierr = VecDuplicate(x1,&coloring->w3);CHKERRQ(ierr);
221     ierr = PetscLogObjectParent((PetscObject)coloring,(PetscObject)coloring->w3);CHKERRQ(ierr);
222   }
223   w3 = coloring->w3;
224 
225   ierr = VecGetOwnershipRange(x1,&cstart,&cend);CHKERRQ(ierr); /* used by ghosted vscale */
226   if (vscale) {
227     ierr = VecGetArray(vscale,&vscale_array);CHKERRQ(ierr);
228   }
229   nz = 0;
230 
231   if (coloring->bcols > 1) { /* use blocked insertion of Jentry */
232     PetscInt    i,m=J->rmap->n,nbcols,bcols=coloring->bcols;
233     PetscScalar *dy=coloring->dy,*dy_k;
234 
235     nbcols = 0;
236     for (k=0; k<ncolors; k+=bcols) {
237 
238       /*
239        (3-1) Loop over each column associated with color
240        adding the perturbation to the vector w3 = x1 + dx.
241        */
242 
243       dy_k = dy;
244       if (k + bcols > ncolors) bcols = ncolors - k;
245       for (i=0; i<bcols; i++) {
246         coloring->currentcolor = k+i;
247 
248         ierr = VecCopy(x1,w3);CHKERRQ(ierr);
249         ierr = VecGetArray(w3,&w3_array);CHKERRQ(ierr);
250         if (ctype == IS_COLORING_GLOBAL) w3_array -= cstart; /* shift pointer so global index can be used */
251         if (coloring->htype[0] == 'w') {
252           for (l=0; l<ncolumns[k+i]; l++) {
253             col = coloring->columns[k+i][l]; /* local column (in global index!) of the matrix we are probing for */
254             w3_array[col] += 1.0/dx;
255           }
256         } else { /* htype == 'ds' */
257           vscale_array -= cstart; /* shift pointer so global index can be used */
258           for (l=0; l<ncolumns[k+i]; l++) {
259             col = coloring->columns[k+i][l]; /* local column (in global index!) of the matrix we are probing for */
260             w3_array[col] += 1.0/vscale_array[col];
261           }
262           vscale_array += cstart;
263         }
264         if (ctype == IS_COLORING_GLOBAL) w3_array += cstart;
265         ierr = VecRestoreArray(w3,&w3_array);CHKERRQ(ierr);
266 
267         /*
268          (3-2) Evaluate function at w3 = x1 + dx (here dx is a vector of perturbations)
269                            w2 = F(x1 + dx) - F(x1)
270          */
271         ierr = PetscLogEventBegin(MAT_FDColoringFunction,0,0,0,0);CHKERRQ(ierr);
272         ierr = VecPlaceArray(w2,dy_k);CHKERRQ(ierr); /* place w2 to the array dy_i */
273         ierr = (*f)(sctx,w3,w2,fctx);CHKERRQ(ierr);
274         ierr = PetscLogEventEnd(MAT_FDColoringFunction,0,0,0,0);CHKERRQ(ierr);
275         ierr = VecAXPY(w2,-1.0,w1);CHKERRQ(ierr);
276         ierr = VecResetArray(w2);CHKERRQ(ierr);
277         dy_k += m; /* points to dy+i*nxloc */
278       }
279 
280       /*
281        (3-3) Loop over block rows of vector, putting results into Jacobian matrix
282        */
283       nrows_k = nrows[nbcols++];
284 
285       if (coloring->htype[0] == 'w') {
286         for (l=0; l<nrows_k; l++) {
287           row  = Jentry2[nz].row;   /* local row index */
288           /* The 'useless' ifdef is due to a bug in NVIDIA nvc 21.11, which triggers a segfault on this line. We write it in
289              another way, and it seems work. See https://lists.mcs.anl.gov/pipermail/petsc-users/2021-December/045158.html
290            */
291          #if defined(PETSC_USE_COMPLEX)
292           PetscScalar *tmp = Jentry2[nz].valaddr;
293           *tmp = dy[row]*dx;
294          #else
295           *(Jentry2[nz].valaddr) = dy[row]*dx;
296          #endif
297           nz++;
298         }
299       } else { /* htype == 'ds' */
300         for (l=0; l<nrows_k; l++) {
301           row = Jentry[nz].row;   /* local row index */
302          #if defined(PETSC_USE_COMPLEX) /* See https://lists.mcs.anl.gov/pipermail/petsc-users/2021-December/045158.html */
303           PetscScalar *tmp = Jentry[nz].valaddr;
304           *tmp = dy[row]*vscale_array[Jentry[nz].col];
305          #else
306           *(Jentry[nz].valaddr) = dy[row]*vscale_array[Jentry[nz].col];
307          #endif
308           nz++;
309         }
310       }
311     }
312   } else { /* bcols == 1 */
313     for (k=0; k<ncolors; k++) {
314       coloring->currentcolor = k;
315 
316       /*
317        (3-1) Loop over each column associated with color
318        adding the perturbation to the vector w3 = x1 + dx.
319        */
320       ierr = VecCopy(x1,w3);CHKERRQ(ierr);
321       ierr = VecGetArray(w3,&w3_array);CHKERRQ(ierr);
322       if (ctype == IS_COLORING_GLOBAL) w3_array -= cstart; /* shift pointer so global index can be used */
323       if (coloring->htype[0] == 'w') {
324         for (l=0; l<ncolumns[k]; l++) {
325           col = coloring->columns[k][l]; /* local column (in global index!) of the matrix we are probing for */
326           w3_array[col] += 1.0/dx;
327         }
328       } else { /* htype == 'ds' */
329         vscale_array -= cstart; /* shift pointer so global index can be used */
330         for (l=0; l<ncolumns[k]; l++) {
331           col = coloring->columns[k][l]; /* local column (in global index!) of the matrix we are probing for */
332           w3_array[col] += 1.0/vscale_array[col];
333         }
334         vscale_array += cstart;
335       }
336       if (ctype == IS_COLORING_GLOBAL) w3_array += cstart;
337       ierr = VecRestoreArray(w3,&w3_array);CHKERRQ(ierr);
338 
339       /*
340        (3-2) Evaluate function at w3 = x1 + dx (here dx is a vector of perturbations)
341                            w2 = F(x1 + dx) - F(x1)
342        */
343       ierr = PetscLogEventBegin(MAT_FDColoringFunction,0,0,0,0);CHKERRQ(ierr);
344       ierr = (*f)(sctx,w3,w2,fctx);CHKERRQ(ierr);
345       ierr = PetscLogEventEnd(MAT_FDColoringFunction,0,0,0,0);CHKERRQ(ierr);
346       ierr = VecAXPY(w2,-1.0,w1);CHKERRQ(ierr);
347 
348       /*
349        (3-3) Loop over rows of vector, putting results into Jacobian matrix
350        */
351       nrows_k = nrows[k];
352       ierr = VecGetArray(w2,&y);CHKERRQ(ierr);
353       if (coloring->htype[0] == 'w') {
354         for (l=0; l<nrows_k; l++) {
355           row  = Jentry2[nz].row;   /* local row index */
356          #if defined(PETSC_USE_COMPLEX) /* See https://lists.mcs.anl.gov/pipermail/petsc-users/2021-December/045158.html */
357           PetscScalar *tmp = Jentry2[nz].valaddr;
358           *tmp = y[row]*dx;
359          #else
360           *(Jentry2[nz].valaddr) = y[row]*dx;
361          #endif
362           nz++;
363         }
364       } else { /* htype == 'ds' */
365         for (l=0; l<nrows_k; l++) {
366           row  = Jentry[nz].row;   /* local row index */
367          #if defined(PETSC_USE_COMPLEX) /* See https://lists.mcs.anl.gov/pipermail/petsc-users/2021-December/045158.html */
368           PetscScalar *tmp = Jentry[nz].valaddr;
369           *tmp = y[row]*vscale_array[Jentry[nz].col];
370          #else
371           *(Jentry[nz].valaddr) = y[row]*vscale_array[Jentry[nz].col];
372          #endif
373           nz++;
374         }
375       }
376       ierr = VecRestoreArray(w2,&y);CHKERRQ(ierr);
377     }
378   }
379 
380 #if defined(PETSC_HAVE_VIENNACL) || defined(PETSC_HAVE_CUDA)
381   if (J->offloadmask != PETSC_OFFLOAD_UNALLOCATED) J->offloadmask = PETSC_OFFLOAD_CPU;
382 #endif
383   ierr = MatAssemblyBegin(J,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr);
384   ierr = MatAssemblyEnd(J,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr);
385   if (vscale) {
386     ierr = VecRestoreArray(vscale,&vscale_array);CHKERRQ(ierr);
387   }
388   coloring->currentcolor = -1;
389   ierr = VecBindToCPU(x1,PETSC_FALSE);CHKERRQ(ierr);
390   PetscFunctionReturn(0);
391 }
392 
393 PetscErrorCode MatFDColoringSetUp_MPIXAIJ(Mat mat,ISColoring iscoloring,MatFDColoring c)
394 {
395   PetscErrorCode         ierr;
396   PetscMPIInt            size,*ncolsonproc,*disp,nn;
397   PetscInt               i,n,nrows,nrows_i,j,k,m,ncols,col,*rowhit,cstart,cend,colb;
398   const PetscInt         *is,*A_ci,*A_cj,*B_ci,*B_cj,*row=NULL,*ltog=NULL;
399   PetscInt               nis=iscoloring->n,nctot,*cols,tmp = 0;
400   ISLocalToGlobalMapping map=mat->cmap->mapping;
401   PetscInt               ctype=c->ctype,*spidxA,*spidxB,nz,bs,bs2,spidx;
402   Mat                    A,B;
403   PetscScalar            *A_val,*B_val,**valaddrhit;
404   MatEntry               *Jentry;
405   MatEntry2              *Jentry2;
406   PetscBool              isBAIJ,isSELL;
407   PetscInt               bcols=c->bcols;
408 #if defined(PETSC_USE_CTABLE)
409   PetscTable             colmap=NULL;
410 #else
411   PetscInt               *colmap=NULL;     /* local col number of off-diag col */
412 #endif
413 
414   PetscFunctionBegin;
415   if (ctype == IS_COLORING_LOCAL) {
416     if (!map) SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_ARG_INCOMP,"When using ghosted differencing matrix must have local to global mapping provided with MatSetLocalToGlobalMapping");
417     ierr = ISLocalToGlobalMappingGetIndices(map,&ltog);CHKERRQ(ierr);
418   }
419 
420   ierr = MatGetBlockSize(mat,&bs);CHKERRQ(ierr);
421   ierr = PetscObjectBaseTypeCompare((PetscObject)mat,MATMPIBAIJ,&isBAIJ);CHKERRQ(ierr);
422   ierr = PetscObjectTypeCompare((PetscObject)mat,MATMPISELL,&isSELL);CHKERRQ(ierr);
423   if (isBAIJ) {
424     Mat_MPIBAIJ *baij=(Mat_MPIBAIJ*)mat->data;
425     Mat_SeqBAIJ *spA,*spB;
426     A = baij->A;  spA = (Mat_SeqBAIJ*)A->data; A_val = spA->a;
427     B = baij->B;  spB = (Mat_SeqBAIJ*)B->data; B_val = spB->a;
428     nz = spA->nz + spB->nz; /* total nonzero entries of mat */
429     if (!baij->colmap) {
430       ierr = MatCreateColmap_MPIBAIJ_Private(mat);CHKERRQ(ierr);
431     }
432     colmap = baij->colmap;
433     ierr = MatGetColumnIJ_SeqBAIJ_Color(A,0,PETSC_FALSE,PETSC_FALSE,&ncols,&A_ci,&A_cj,&spidxA,NULL);CHKERRQ(ierr);
434     ierr = MatGetColumnIJ_SeqBAIJ_Color(B,0,PETSC_FALSE,PETSC_FALSE,&ncols,&B_ci,&B_cj,&spidxB,NULL);CHKERRQ(ierr);
435 
436     if (ctype == IS_COLORING_GLOBAL && c->htype[0] == 'd') {  /* create vscale for storing dx */
437       PetscInt    *garray;
438       ierr = PetscMalloc1(B->cmap->n,&garray);CHKERRQ(ierr);
439       for (i=0; i<baij->B->cmap->n/bs; i++) {
440         for (j=0; j<bs; j++) {
441           garray[i*bs+j] = bs*baij->garray[i]+j;
442         }
443       }
444       ierr = VecCreateGhost(PetscObjectComm((PetscObject)mat),mat->cmap->n,PETSC_DETERMINE,B->cmap->n,garray,&c->vscale);CHKERRQ(ierr);
445       ierr = VecBindToCPU(c->vscale,PETSC_TRUE);CHKERRQ(ierr);
446       ierr = PetscFree(garray);CHKERRQ(ierr);
447     }
448   } else if (isSELL) {
449     Mat_MPISELL *sell=(Mat_MPISELL*)mat->data;
450     Mat_SeqSELL *spA,*spB;
451     A = sell->A;  spA = (Mat_SeqSELL*)A->data; A_val = spA->val;
452     B = sell->B;  spB = (Mat_SeqSELL*)B->data; B_val = spB->val;
453     nz = spA->nz + spB->nz; /* total nonzero entries of mat */
454     if (!sell->colmap) {
455       /* Allow access to data structures of local part of matrix
456        - creates aij->colmap which maps global column number to local number in part B */
457       ierr = MatCreateColmap_MPISELL_Private(mat);CHKERRQ(ierr);
458     }
459     colmap = sell->colmap;
460     ierr = MatGetColumnIJ_SeqSELL_Color(A,0,PETSC_FALSE,PETSC_FALSE,&ncols,&A_ci,&A_cj,&spidxA,NULL);CHKERRQ(ierr);
461     ierr = MatGetColumnIJ_SeqSELL_Color(B,0,PETSC_FALSE,PETSC_FALSE,&ncols,&B_ci,&B_cj,&spidxB,NULL);CHKERRQ(ierr);
462 
463     bs = 1; /* only bs=1 is supported for non MPIBAIJ matrix */
464 
465     if (ctype == IS_COLORING_GLOBAL && c->htype[0] == 'd') { /* create vscale for storing dx */
466       ierr = VecCreateGhost(PetscObjectComm((PetscObject)mat),mat->cmap->n,PETSC_DETERMINE,B->cmap->n,sell->garray,&c->vscale);CHKERRQ(ierr);
467       ierr = VecBindToCPU(c->vscale,PETSC_TRUE);CHKERRQ(ierr);
468     }
469   } else {
470     Mat_MPIAIJ *aij=(Mat_MPIAIJ*)mat->data;
471     Mat_SeqAIJ *spA,*spB;
472     A = aij->A;  spA = (Mat_SeqAIJ*)A->data; A_val = spA->a;
473     B = aij->B;  spB = (Mat_SeqAIJ*)B->data; B_val = spB->a;
474     nz = spA->nz + spB->nz; /* total nonzero entries of mat */
475     if (!aij->colmap) {
476       /* Allow access to data structures of local part of matrix
477        - creates aij->colmap which maps global column number to local number in part B */
478       ierr = MatCreateColmap_MPIAIJ_Private(mat);CHKERRQ(ierr);
479     }
480     colmap = aij->colmap;
481     ierr = MatGetColumnIJ_SeqAIJ_Color(A,0,PETSC_FALSE,PETSC_FALSE,&ncols,&A_ci,&A_cj,&spidxA,NULL);CHKERRQ(ierr);
482     ierr = MatGetColumnIJ_SeqAIJ_Color(B,0,PETSC_FALSE,PETSC_FALSE,&ncols,&B_ci,&B_cj,&spidxB,NULL);CHKERRQ(ierr);
483 
484     bs = 1; /* only bs=1 is supported for non MPIBAIJ matrix */
485 
486     if (ctype == IS_COLORING_GLOBAL && c->htype[0] == 'd') { /* create vscale for storing dx */
487       ierr = VecCreateGhost(PetscObjectComm((PetscObject)mat),mat->cmap->n,PETSC_DETERMINE,B->cmap->n,aij->garray,&c->vscale);CHKERRQ(ierr);
488       ierr = VecBindToCPU(c->vscale,PETSC_TRUE);CHKERRQ(ierr);
489     }
490   }
491 
492   m      = mat->rmap->n/bs;
493   cstart = mat->cmap->rstart/bs;
494   cend   = mat->cmap->rend/bs;
495 
496   ierr = PetscMalloc2(nis,&c->ncolumns,nis,&c->columns);CHKERRQ(ierr);
497   ierr = PetscMalloc1(nis,&c->nrows);CHKERRQ(ierr);
498   ierr = PetscLogObjectMemory((PetscObject)c,3*nis*sizeof(PetscInt));CHKERRQ(ierr);
499 
500   if (c->htype[0] == 'd') {
501     ierr        = PetscMalloc1(nz,&Jentry);CHKERRQ(ierr);
502     ierr        = PetscLogObjectMemory((PetscObject)c,nz*sizeof(MatEntry));CHKERRQ(ierr);
503     c->matentry = Jentry;
504   } else if (c->htype[0] == 'w') {
505     ierr         = PetscMalloc1(nz,&Jentry2);CHKERRQ(ierr);
506     ierr         = PetscLogObjectMemory((PetscObject)c,nz*sizeof(MatEntry2));CHKERRQ(ierr);
507     c->matentry2 = Jentry2;
508   } else SETERRQ(PetscObjectComm((PetscObject)mat),PETSC_ERR_SUP,"htype is not supported");
509 
510   ierr = PetscMalloc2(m+1,&rowhit,m+1,&valaddrhit);CHKERRQ(ierr);
511   nz   = 0;
512   ierr = ISColoringGetIS(iscoloring,PETSC_OWN_POINTER, PETSC_IGNORE,&c->isa);CHKERRQ(ierr);
513 
514   if (ctype == IS_COLORING_GLOBAL) {
515     ierr = MPI_Comm_size(PetscObjectComm((PetscObject)mat),&size);CHKERRMPI(ierr);
516     ierr = PetscMalloc2(size,&ncolsonproc,size,&disp);CHKERRQ(ierr);
517   }
518 
519   for (i=0; i<nis; i++) { /* for each local color */
520     ierr = ISGetLocalSize(c->isa[i],&n);CHKERRQ(ierr);
521     ierr = ISGetIndices(c->isa[i],&is);CHKERRQ(ierr);
522 
523     c->ncolumns[i] = n; /* local number of columns of this color on this process */
524     c->columns[i]  = (PetscInt*)is;
525 
526     if (ctype == IS_COLORING_GLOBAL) {
527       /* Determine nctot, the total (parallel) number of columns of this color */
528       /* ncolsonproc[j]: local ncolumns on proc[j] of this color */
529       ierr  = PetscMPIIntCast(n,&nn);CHKERRQ(ierr);
530       ierr  = MPI_Allgather(&nn,1,MPI_INT,ncolsonproc,1,MPI_INT,PetscObjectComm((PetscObject)mat));CHKERRMPI(ierr);
531       nctot = 0; for (j=0; j<size; j++) nctot += ncolsonproc[j];
532       if (!nctot) {
533         ierr = PetscInfo(mat,"Coloring of matrix has some unneeded colors with no corresponding rows\n");CHKERRQ(ierr);
534       }
535 
536       disp[0] = 0;
537       for (j=1; j<size; j++) {
538         disp[j] = disp[j-1] + ncolsonproc[j-1];
539       }
540 
541       /* Get cols, the complete list of columns for this color on each process */
542       ierr = PetscMalloc1(nctot+1,&cols);CHKERRQ(ierr);
543       ierr = MPI_Allgatherv((void*)is,n,MPIU_INT,cols,ncolsonproc,disp,MPIU_INT,PetscObjectComm((PetscObject)mat));CHKERRMPI(ierr);
544     } else if (ctype == IS_COLORING_LOCAL) {
545       /* Determine local number of columns of this color on this process, including ghost points */
546       nctot = n;
547       cols  = (PetscInt*)is;
548     } else SETERRQ(PETSC_COMM_SELF,PETSC_ERR_SUP,"Not provided for this MatFDColoring type");
549 
550     /* Mark all rows affect by these columns */
551     ierr    = PetscArrayzero(rowhit,m);CHKERRQ(ierr);
552     bs2     = bs*bs;
553     nrows_i = 0;
554     for (j=0; j<nctot; j++) { /* loop over columns*/
555       if (ctype == IS_COLORING_LOCAL) {
556         col = ltog[cols[j]];
557       } else {
558         col = cols[j];
559       }
560       if (col >= cstart && col < cend) { /* column is in A, diagonal block of mat */
561         tmp      = A_ci[col-cstart];
562         row      = A_cj + tmp;
563         nrows    = A_ci[col-cstart+1] - tmp;
564         nrows_i += nrows;
565 
566         /* loop over columns of A marking them in rowhit */
567         for (k=0; k<nrows; k++) {
568           /* set valaddrhit for part A */
569           spidx            = bs2*spidxA[tmp + k];
570           valaddrhit[*row] = &A_val[spidx];
571           rowhit[*row++]   = col - cstart + 1; /* local column index */
572         }
573       } else { /* column is in B, off-diagonal block of mat */
574 #if defined(PETSC_USE_CTABLE)
575         ierr = PetscTableFind(colmap,col+1,&colb);CHKERRQ(ierr);
576         colb--;
577 #else
578         colb = colmap[col] - 1; /* local column index */
579 #endif
580         if (colb == -1) {
581           nrows = 0;
582         } else {
583           colb  = colb/bs;
584           tmp   = B_ci[colb];
585           row   = B_cj + tmp;
586           nrows = B_ci[colb+1] - tmp;
587         }
588         nrows_i += nrows;
589         /* loop over columns of B marking them in rowhit */
590         for (k=0; k<nrows; k++) {
591           /* set valaddrhit for part B */
592           spidx            = bs2*spidxB[tmp + k];
593           valaddrhit[*row] = &B_val[spidx];
594           rowhit[*row++]   = colb + 1 + cend - cstart; /* local column index */
595         }
596       }
597     }
598     c->nrows[i] = nrows_i;
599 
600     if (c->htype[0] == 'd') {
601       for (j=0; j<m; j++) {
602         if (rowhit[j]) {
603           Jentry[nz].row     = j;              /* local row index */
604           Jentry[nz].col     = rowhit[j] - 1;  /* local column index */
605           Jentry[nz].valaddr = valaddrhit[j];  /* address of mat value for this entry */
606           nz++;
607         }
608       }
609     } else { /* c->htype == 'wp' */
610       for (j=0; j<m; j++) {
611         if (rowhit[j]) {
612           Jentry2[nz].row     = j;              /* local row index */
613           Jentry2[nz].valaddr = valaddrhit[j];  /* address of mat value for this entry */
614           nz++;
615         }
616       }
617     }
618     if (ctype == IS_COLORING_GLOBAL) {
619       ierr = PetscFree(cols);CHKERRQ(ierr);
620     }
621   }
622   if (ctype == IS_COLORING_GLOBAL) {
623     ierr = PetscFree2(ncolsonproc,disp);CHKERRQ(ierr);
624   }
625 
626   if (bcols > 1) { /* reorder Jentry for faster MatFDColoringApply() */
627     ierr = MatFDColoringSetUpBlocked_AIJ_Private(mat,c,nz);CHKERRQ(ierr);
628   }
629 
630   if (isBAIJ) {
631     ierr = MatRestoreColumnIJ_SeqBAIJ_Color(A,0,PETSC_FALSE,PETSC_FALSE,&ncols,&A_ci,&A_cj,&spidxA,NULL);CHKERRQ(ierr);
632     ierr = MatRestoreColumnIJ_SeqBAIJ_Color(B,0,PETSC_FALSE,PETSC_FALSE,&ncols,&B_ci,&B_cj,&spidxB,NULL);CHKERRQ(ierr);
633     ierr = PetscMalloc1(bs*mat->rmap->n,&c->dy);CHKERRQ(ierr);
634   } else if (isSELL) {
635     ierr = MatRestoreColumnIJ_SeqSELL_Color(A,0,PETSC_FALSE,PETSC_FALSE,&ncols,&A_ci,&A_cj,&spidxA,NULL);CHKERRQ(ierr);
636     ierr = MatRestoreColumnIJ_SeqSELL_Color(B,0,PETSC_FALSE,PETSC_FALSE,&ncols,&B_ci,&B_cj,&spidxB,NULL);CHKERRQ(ierr);
637   } else {
638     ierr = MatRestoreColumnIJ_SeqAIJ_Color(A,0,PETSC_FALSE,PETSC_FALSE,&ncols,&A_ci,&A_cj,&spidxA,NULL);CHKERRQ(ierr);
639     ierr = MatRestoreColumnIJ_SeqAIJ_Color(B,0,PETSC_FALSE,PETSC_FALSE,&ncols,&B_ci,&B_cj,&spidxB,NULL);CHKERRQ(ierr);
640   }
641 
642   ierr = ISColoringRestoreIS(iscoloring,PETSC_OWN_POINTER,&c->isa);CHKERRQ(ierr);
643   ierr = PetscFree2(rowhit,valaddrhit);CHKERRQ(ierr);
644 
645   if (ctype == IS_COLORING_LOCAL) {
646     ierr = ISLocalToGlobalMappingRestoreIndices(map,&ltog);CHKERRQ(ierr);
647   }
648   ierr = PetscInfo3(c,"ncolors %D, brows %D and bcols %D are used.\n",c->ncolors,c->brows,c->bcols);CHKERRQ(ierr);
649   PetscFunctionReturn(0);
650 }
651 
652 PetscErrorCode MatFDColoringCreate_MPIXAIJ(Mat mat,ISColoring iscoloring,MatFDColoring c)
653 {
654   PetscErrorCode ierr;
655   PetscInt       bs,nis=iscoloring->n,m=mat->rmap->n;
656   PetscBool      isBAIJ,isSELL;
657 
658   PetscFunctionBegin;
659   /* set default brows and bcols for speedup inserting the dense matrix into sparse Jacobian;
660    bcols is chosen s.t. dy-array takes 50% of memory space as mat */
661   ierr = MatGetBlockSize(mat,&bs);CHKERRQ(ierr);
662   ierr = PetscObjectBaseTypeCompare((PetscObject)mat,MATMPIBAIJ,&isBAIJ);CHKERRQ(ierr);
663   ierr = PetscObjectTypeCompare((PetscObject)mat,MATMPISELL,&isSELL);CHKERRQ(ierr);
664   if (isBAIJ || m == 0) {
665     c->brows = m;
666     c->bcols = 1;
667   } else if (isSELL) {
668     /* bcols is chosen s.t. dy-array takes 50% of local memory space as mat */
669     Mat_MPISELL *sell=(Mat_MPISELL*)mat->data;
670     Mat_SeqSELL *spA,*spB;
671     Mat        A,B;
672     PetscInt   nz,brows,bcols;
673     PetscReal  mem;
674 
675     bs    = 1; /* only bs=1 is supported for MPISELL matrix */
676 
677     A = sell->A;  spA = (Mat_SeqSELL*)A->data;
678     B = sell->B;  spB = (Mat_SeqSELL*)B->data;
679     nz = spA->nz + spB->nz; /* total local nonzero entries of mat */
680     mem = nz*(sizeof(PetscScalar) + sizeof(PetscInt)) + 3*m*sizeof(PetscInt);
681     bcols = (PetscInt)(0.5*mem /(m*sizeof(PetscScalar)));
682     brows = 1000/bcols;
683     if (bcols > nis) bcols = nis;
684     if (brows == 0 || brows > m) brows = m;
685     c->brows = brows;
686     c->bcols = bcols;
687   } else { /* mpiaij matrix */
688     /* bcols is chosen s.t. dy-array takes 50% of local memory space as mat */
689     Mat_MPIAIJ *aij=(Mat_MPIAIJ*)mat->data;
690     Mat_SeqAIJ *spA,*spB;
691     Mat        A,B;
692     PetscInt   nz,brows,bcols;
693     PetscReal  mem;
694 
695     bs    = 1; /* only bs=1 is supported for MPIAIJ matrix */
696 
697     A = aij->A;  spA = (Mat_SeqAIJ*)A->data;
698     B = aij->B;  spB = (Mat_SeqAIJ*)B->data;
699     nz = spA->nz + spB->nz; /* total local nonzero entries of mat */
700     mem = nz*(sizeof(PetscScalar) + sizeof(PetscInt)) + 3*m*sizeof(PetscInt);
701     bcols = (PetscInt)(0.5*mem /(m*sizeof(PetscScalar)));
702     brows = 1000/bcols;
703     if (bcols > nis) bcols = nis;
704     if (brows == 0 || brows > m) brows = m;
705     c->brows = brows;
706     c->bcols = bcols;
707   }
708 
709   c->M       = mat->rmap->N/bs;         /* set the global rows and columns and local rows */
710   c->N       = mat->cmap->N/bs;
711   c->m       = mat->rmap->n/bs;
712   c->rstart  = mat->rmap->rstart/bs;
713   c->ncolors = nis;
714   PetscFunctionReturn(0);
715 }
716 
717 /*@C
718 
719     MatFDColoringSetValues - takes a matrix in compressed color format and enters the matrix into a PETSc Mat
720 
721    Collective on J
722 
723    Input Parameters:
724 +    J - the sparse matrix
725 .    coloring - created with MatFDColoringCreate() and a local coloring
726 -    y - column major storage of matrix values with one color of values per column, the number of rows of y should match
727          the number of local rows of J and the number of columns is the number of colors.
728 
729    Level: intermediate
730 
731    Notes: the matrix in compressed color format may come from an Automatic Differentiation code
732 
733    The code will be slightly faster if MatFDColoringSetBlockSize(coloring,PETSC_DEFAULT,nc); is called immediately after creating the coloring
734 
735 .seealso: MatFDColoringCreate(), ISColoring, ISColoringCreate(), ISColoringSetType(), IS_COLORING_LOCAL, MatFDColoringSetBlockSize()
736 
737 @*/
738 PetscErrorCode  MatFDColoringSetValues(Mat J,MatFDColoring coloring,const PetscScalar *y)
739 {
740   PetscErrorCode    ierr;
741   MatEntry2         *Jentry2;
742   PetscInt          row,i,nrows_k,l,ncolors,nz = 0,bcols,nbcols = 0;
743   const PetscInt    *nrows;
744   PetscBool         eq;
745 
746   PetscFunctionBegin;
747   PetscValidHeaderSpecific(J,MAT_CLASSID,1);
748   PetscValidHeaderSpecific(coloring,MAT_FDCOLORING_CLASSID,2);
749   ierr = PetscObjectCompareId((PetscObject)J,coloring->matid,&eq);CHKERRQ(ierr);
750   if (!eq) SETERRQ(PetscObjectComm((PetscObject)J),PETSC_ERR_ARG_WRONG,"Matrix used with MatFDColoringSetValues() must be that used with MatFDColoringCreate()");
751   Jentry2 = coloring->matentry2;
752   nrows   = coloring->nrows;
753   ncolors = coloring->ncolors;
754   bcols   = coloring->bcols;
755 
756   for (i=0; i<ncolors; i+=bcols) {
757     nrows_k = nrows[nbcols++];
758     for (l=0; l<nrows_k; l++) {
759       row                      = Jentry2[nz].row;   /* local row index */
760       *(Jentry2[nz++].valaddr) = y[row];
761     }
762     y += bcols*coloring->m;
763   }
764   ierr = MatAssemblyBegin(J,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr);
765   ierr = MatAssemblyEnd(J,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr);
766   PetscFunctionReturn(0);
767 }
768