xref: /petsc/src/mat/impls/baij/seq/baijfact.c (revision b06ca6b09b3ae4e8a93ed49db16f4412dd6db9d8)
1 #define PETSCMAT_DLL
2 
3 /*
4     Factorization code for BAIJ format.
5 */
6 #include "../src/mat/impls/baij/seq/baij.h"
7 #include "../src/mat/blockinvert.h"
8 
9 /* ------------------------------------------------------------*/
10 /*
11       Version for when blocks are 2 by 2
12 */
13 /* MatLUFactorNumeric_SeqBAIJ_2_newdatastruct -
14      copied from MatLUFactorNumeric_SeqBAIJ_N_newdatastruct() and manually re-implemented
15        Kernel_A_gets_A_times_B()
16        Kernel_A_gets_A_minus_B_times_C()
17        Kernel_A_gets_inverse_A()
18 */
19 #undef __FUNCT__
20 #define __FUNCT__ "MatLUFactorNumeric_SeqBAIJ_2_newdatastruct"
21 PetscErrorCode MatLUFactorNumeric_SeqBAIJ_2_newdatastruct(Mat B,Mat A,const MatFactorInfo *info)
22 {
23   Mat            C=B;
24   Mat_SeqBAIJ    *a=(Mat_SeqBAIJ*)A->data,*b=(Mat_SeqBAIJ *)C->data;
25   IS             isrow = b->row,isicol = b->icol;
26   PetscErrorCode ierr;
27   const PetscInt *r,*ic,*ics;
28   PetscInt       i,j,k,n=a->mbs,*ai=a->i,*aj=a->j,*bi=b->i,*bj=b->j;
29   PetscInt       *ajtmp,*bjtmp,nz,nzL,row,*bdiag=b->diag,*pj;
30   MatScalar      *rtmp,*pc,*mwork,*v,*pv,*aa=a->a;
31   PetscInt       bs2 = a->bs2,flg;
32   PetscReal      shift = info->shiftinblocks;
33 
34   PetscFunctionBegin;
35   ierr = ISGetIndices(isrow,&r);CHKERRQ(ierr);
36   ierr = ISGetIndices(isicol,&ic);CHKERRQ(ierr);
37 
38   /* generate work space needed by the factorization */
39   ierr = PetscMalloc((bs2*n+bs2+1)*sizeof(MatScalar),&rtmp);CHKERRQ(ierr);
40   mwork = rtmp + bs2*n;
41   ierr = PetscMemzero(rtmp,bs2*n*sizeof(MatScalar));CHKERRQ(ierr);
42   ics  = ic;
43 
44   for (i=0; i<n; i++){
45     /* zero rtmp */
46     /* L part */
47     nz    = bi[i+1] - bi[i];
48     bjtmp = bj + bi[i];
49     for  (j=0; j<nz; j++){
50       ierr = PetscMemzero(rtmp+bs2*bjtmp[j],bs2*sizeof(MatScalar));CHKERRQ(ierr);
51     }
52 
53     /* U part */
54     nz = bi[2*n-i+1] - bi[2*n-i];
55     bjtmp = bj + bi[2*n-i];
56     for  (j=0; j<nz; j++){
57       ierr = PetscMemzero(rtmp+bs2*bjtmp[j],bs2*sizeof(MatScalar));CHKERRQ(ierr);
58     }
59 
60     /* load in initial (unfactored row) */
61     nz    = ai[r[i]+1] - ai[r[i]];
62     ajtmp = aj + ai[r[i]];
63     v     = aa + bs2*ai[r[i]];
64     for (j=0; j<nz; j++) {
65       ierr = PetscMemcpy(rtmp+bs2*ic[ajtmp[j]],v+bs2*j,bs2*sizeof(MatScalar));CHKERRQ(ierr);
66     }
67 
68     /* elimination */
69     bjtmp = bj + bi[i];
70     nzL   = bi[i+1] - bi[i];
71     for(k=0;k < nzL;k++) {
72       row = bjtmp[k];
73       pc = rtmp + bs2*row;
74       for (flg=0,j=0; j<bs2; j++) { if (pc[j]!=0.0) { flg = 1; break; }}
75       if (flg) {
76         pv = b->a + bs2*bdiag[row];
77         /* Kernel_A_gets_A_times_B(bs,pc,pv,mwork); *pc = *pc * (*pv); */
78         ierr = Kernel_A_gets_A_times_B_2(pc,pv,mwork);CHKERRQ(ierr);
79 
80         pj = b->j + bi[2*n-row]; /* begining of U(row,:) */
81         pv = b->a + bs2*bi[2*n-row];
82         nz = bi[2*n-row+1] - bi[2*n-row] - 1; /* num of entries inU(row,:), excluding diag */
83         for (j=0; j<nz; j++) {
84           /* Kernel_A_gets_A_minus_B_times_C(bs,rtmp+bs2*pj[j],pc,pv+bs2*j); */
85           /* rtmp+bs2*pj[j] = rtmp+bs2*pj[j] - (*pc)*(pv+bs2*j) */
86           v    = rtmp + 4*pj[j];
87           ierr = Kernel_A_gets_A_minus_B_times_C_2(v,pc,pv);CHKERRQ(ierr);
88           pv  += 4;
89         }
90         ierr = PetscLogFlops(16*nz+12);CHKERRQ(ierr); /* flops = 2*bs^3*nz + 2*bs^3 - bs2) */
91       }
92     }
93 
94     /* finished row so stick it into b->a */
95     /* L part */
96     pv   = b->a + bs2*bi[i] ;
97     pj   = b->j + bi[i] ;
98     nz   = bi[i+1] - bi[i];
99     for (j=0; j<nz; j++) {
100       ierr = PetscMemcpy(pv+bs2*j,rtmp+bs2*pj[j],bs2*sizeof(MatScalar));CHKERRQ(ierr);
101     }
102 
103     /* Mark diagonal and invert diagonal for simplier triangular solves */
104     pv   = b->a + bs2*bdiag[i];
105     pj   = b->j + bdiag[i];
106     ierr = PetscMemcpy(pv,rtmp+bs2*pj[0],bs2*sizeof(MatScalar));CHKERRQ(ierr);
107     /* ierr = Kernel_A_gets_inverse_A(bs,pv,v_pivots,v_work);CHKERRQ(ierr); */
108     ierr = Kernel_A_gets_inverse_A_2(pv,shift);CHKERRQ(ierr);
109 
110     /* U part */
111     pv = b->a + bs2*bi[2*n-i];
112     pj = b->j + bi[2*n-i];
113     nz = bi[2*n-i+1] - bi[2*n-i] - 1;
114     for (j=0; j<nz; j++){
115       ierr = PetscMemcpy(pv+bs2*j,rtmp+bs2*pj[j],bs2*sizeof(MatScalar));CHKERRQ(ierr);
116     }
117   }
118 
119   ierr = PetscFree(rtmp);CHKERRQ(ierr);
120   ierr = ISRestoreIndices(isicol,&ic);CHKERRQ(ierr);
121   ierr = ISRestoreIndices(isrow,&r);CHKERRQ(ierr);
122 
123   C->assembled = PETSC_TRUE;
124   ierr = PetscLogFlops(1.3333*bs2*n);CHKERRQ(ierr); /* from inverting diagonal blocks */
125   PetscFunctionReturn(0);
126 }
127 
128 /*
129   MatLUFactorNumeric_SeqBAIJ_2_NaturalOrdering_newdatastruct -
130     copied from MatLUFactorNumeric_SeqBAIJ_2_newdatastruct(), then remove ordering
131 */
132 #undef __FUNCT__
133 #define __FUNCT__ "MatLUFactorNumeric_SeqBAIJ_2_NaturalOrdering_newdatastruct"
134 PetscErrorCode MatLUFactorNumeric_SeqBAIJ_2_NaturalOrdering_newdatastruct(Mat B,Mat A,const MatFactorInfo *info)
135 {
136   Mat            C=B;
137   Mat_SeqBAIJ    *a=(Mat_SeqBAIJ*)A->data,*b=(Mat_SeqBAIJ *)C->data;
138   PetscErrorCode ierr;
139   PetscInt       i,j,k,n=a->mbs,*ai=a->i,*aj=a->j,*bi=b->i,*bj=b->j;
140   PetscInt       *ajtmp,*bjtmp,nz,nzL,row,*bdiag=b->diag,*pj;
141   MatScalar      *rtmp,*pc,*mwork,*v,*pv,*aa=a->a;
142   PetscInt       bs2 = a->bs2,flg;
143   PetscReal      shift = info->shiftinblocks;
144 
145   PetscFunctionBegin;
146   /* generate work space needed by the factorization */
147   ierr = PetscMalloc((bs2*n+bs2+1)*sizeof(MatScalar),&rtmp);CHKERRQ(ierr);
148   mwork = rtmp + bs2*n;
149   ierr = PetscMemzero(rtmp,bs2*n*sizeof(MatScalar));CHKERRQ(ierr);
150 
151   for (i=0; i<n; i++){
152     /* zero rtmp */
153     /* L part */
154     nz    = bi[i+1] - bi[i];
155     bjtmp = bj + bi[i];
156     for  (j=0; j<nz; j++){
157       ierr = PetscMemzero(rtmp+bs2*bjtmp[j],bs2*sizeof(MatScalar));CHKERRQ(ierr);
158     }
159 
160     /* U part */
161     nz = bi[2*n-i+1] - bi[2*n-i];
162     bjtmp = bj + bi[2*n-i];
163     for  (j=0; j<nz; j++){
164       ierr = PetscMemzero(rtmp+bs2*bjtmp[j],bs2*sizeof(MatScalar));CHKERRQ(ierr);
165     }
166 
167     /* load in initial (unfactored row) */
168     nz    = ai[i+1] - ai[i];
169     ajtmp = aj + ai[i];
170     v     = aa + bs2*ai[i];
171     for (j=0; j<nz; j++) {
172       ierr = PetscMemcpy(rtmp+bs2*ajtmp[j],v+bs2*j,bs2*sizeof(MatScalar));CHKERRQ(ierr);
173     }
174 
175     /* elimination */
176     bjtmp = bj + bi[i];
177     nzL   = bi[i+1] - bi[i];
178     for(k=0;k < nzL;k++) {
179       row = bjtmp[k];
180       pc = rtmp + bs2*row;
181       for (flg=0,j=0; j<bs2; j++) { if (pc[j]!=0.0) { flg = 1; break; }}
182       if (flg) {
183         pv = b->a + bs2*bdiag[row];
184         /* Kernel_A_gets_A_times_B(bs,pc,pv,mwork); *pc = *pc * (*pv); */
185         ierr = Kernel_A_gets_A_times_B_2(pc,pv,mwork);CHKERRQ(ierr);
186 
187         pj = b->j + bi[2*n-row]; /* begining of U(row,:) */
188         pv = b->a + bs2*bi[2*n-row];
189         nz = bi[2*n-row+1] - bi[2*n-row] - 1; /* num of entries inU(row,:), excluding diag */
190         for (j=0; j<nz; j++) {
191           /* Kernel_A_gets_A_minus_B_times_C(bs,rtmp+bs2*pj[j],pc,pv+bs2*j); */
192           /* rtmp+bs2*pj[j] = rtmp+bs2*pj[j] - (*pc)*(pv+bs2*j) */
193           v    = rtmp + 4*pj[j];
194           ierr = Kernel_A_gets_A_minus_B_times_C_2(v,pc,pv);CHKERRQ(ierr);
195           pv  += 4;
196         }
197         ierr = PetscLogFlops(16*nz+12);CHKERRQ(ierr); /* flops = 2*bs^3*nz + 2*bs^3 - bs2) */
198       }
199     }
200 
201     /* finished row so stick it into b->a */
202     /* L part */
203     pv   = b->a + bs2*bi[i] ;
204     pj   = b->j + bi[i] ;
205     nz   = bi[i+1] - bi[i];
206     for (j=0; j<nz; j++) {
207       ierr = PetscMemcpy(pv+bs2*j,rtmp+bs2*pj[j],bs2*sizeof(MatScalar));CHKERRQ(ierr);
208     }
209 
210     /* Mark diagonal and invert diagonal for simplier triangular solves */
211     pv   = b->a + bs2*bdiag[i];
212     pj   = b->j + bdiag[i];
213     ierr = PetscMemcpy(pv,rtmp+bs2*pj[0],bs2*sizeof(MatScalar));CHKERRQ(ierr);
214     /* ierr = Kernel_A_gets_inverse_A(bs,pv,v_pivots,v_work);CHKERRQ(ierr); */
215     ierr = Kernel_A_gets_inverse_A_2(pv,shift);CHKERRQ(ierr);
216 
217     /* U part */
218     pv = b->a + bs2*bi[2*n-i];
219     pj = b->j + bi[2*n-i];
220     nz = bi[2*n-i+1] - bi[2*n-i] - 1;
221     for (j=0; j<nz; j++){
222       ierr = PetscMemcpy(pv+bs2*j,rtmp+bs2*pj[j],bs2*sizeof(MatScalar));CHKERRQ(ierr);
223     }
224   }
225 
226   ierr = PetscFree(rtmp);CHKERRQ(ierr);
227   C->assembled = PETSC_TRUE;
228   ierr = PetscLogFlops(1.3333*bs2*n);CHKERRQ(ierr); /* from inverting diagonal blocks */
229   PetscFunctionReturn(0);
230 }
231 
232 #undef __FUNCT__
233 #define __FUNCT__ "MatLUFactorNumeric_SeqBAIJ_2"
234 PetscErrorCode MatLUFactorNumeric_SeqBAIJ_2(Mat B,Mat A,const MatFactorInfo *info)
235 {
236   Mat            C = B;
237   Mat_SeqBAIJ    *a = (Mat_SeqBAIJ*)A->data,*b = (Mat_SeqBAIJ *)C->data;
238   IS             isrow = b->row,isicol = b->icol;
239   PetscErrorCode ierr;
240   const PetscInt *r,*ic;
241   PetscInt       i,j,n = a->mbs,*bi = b->i,*bj = b->j;
242   PetscInt       *ajtmpold,*ajtmp,nz,row;
243   PetscInt       *diag_offset=b->diag,idx,*ai=a->i,*aj=a->j,*pj;
244   MatScalar      *pv,*v,*rtmp,m1,m2,m3,m4,*pc,*w,*x,x1,x2,x3,x4;
245   MatScalar      p1,p2,p3,p4;
246   MatScalar      *ba = b->a,*aa = a->a;
247   PetscReal      shift = info->shiftinblocks;
248 
249   PetscFunctionBegin;
250   ierr  = ISGetIndices(isrow,&r);CHKERRQ(ierr);
251   ierr  = ISGetIndices(isicol,&ic);CHKERRQ(ierr);
252   ierr  = PetscMalloc(4*(n+1)*sizeof(MatScalar),&rtmp);CHKERRQ(ierr);
253 
254   for (i=0; i<n; i++) {
255     nz    = bi[i+1] - bi[i];
256     ajtmp = bj + bi[i];
257     for  (j=0; j<nz; j++) {
258       x = rtmp+4*ajtmp[j]; x[0] = x[1] = x[2] = x[3] = 0.0;
259     }
260     /* load in initial (unfactored row) */
261     idx      = r[i];
262     nz       = ai[idx+1] - ai[idx];
263     ajtmpold = aj + ai[idx];
264     v        = aa + 4*ai[idx];
265     for (j=0; j<nz; j++) {
266       x    = rtmp+4*ic[ajtmpold[j]];
267       x[0] = v[0]; x[1] = v[1]; x[2] = v[2]; x[3] = v[3];
268       v    += 4;
269     }
270     row = *ajtmp++;
271     while (row < i) {
272       pc = rtmp + 4*row;
273       p1 = pc[0]; p2 = pc[1]; p3 = pc[2]; p4 = pc[3];
274       if (p1 != 0.0 || p2 != 0.0 || p3 != 0.0 || p4 != 0.0) {
275         pv = ba + 4*diag_offset[row];
276         pj = bj + diag_offset[row] + 1;
277         x1 = pv[0]; x2 = pv[1]; x3 = pv[2]; x4 = pv[3];
278         pc[0] = m1 = p1*x1 + p3*x2;
279         pc[1] = m2 = p2*x1 + p4*x2;
280         pc[2] = m3 = p1*x3 + p3*x4;
281         pc[3] = m4 = p2*x3 + p4*x4;
282         nz = bi[row+1] - diag_offset[row] - 1;
283         pv += 4;
284         for (j=0; j<nz; j++) {
285           x1   = pv[0]; x2 = pv[1]; x3 = pv[2]; x4 = pv[3];
286           x    = rtmp + 4*pj[j];
287           x[0] -= m1*x1 + m3*x2;
288           x[1] -= m2*x1 + m4*x2;
289           x[2] -= m1*x3 + m3*x4;
290           x[3] -= m2*x3 + m4*x4;
291           pv   += 4;
292         }
293         ierr = PetscLogFlops(16.0*nz+12.0);CHKERRQ(ierr);
294       }
295       row = *ajtmp++;
296     }
297     /* finished row so stick it into b->a */
298     pv = ba + 4*bi[i];
299     pj = bj + bi[i];
300     nz = bi[i+1] - bi[i];
301     for (j=0; j<nz; j++) {
302       x     = rtmp+4*pj[j];
303       pv[0] = x[0]; pv[1] = x[1]; pv[2] = x[2]; pv[3] = x[3];
304       pv   += 4;
305     }
306     /* invert diagonal block */
307     w = ba + 4*diag_offset[i];
308     ierr = Kernel_A_gets_inverse_A_2(w,shift);CHKERRQ(ierr);
309   }
310 
311   ierr = PetscFree(rtmp);CHKERRQ(ierr);
312   ierr = ISRestoreIndices(isicol,&ic);CHKERRQ(ierr);
313   ierr = ISRestoreIndices(isrow,&r);CHKERRQ(ierr);
314   C->ops->solve          = MatSolve_SeqBAIJ_2;
315   C->ops->solvetranspose = MatSolveTranspose_SeqBAIJ_2;
316   C->assembled = PETSC_TRUE;
317   ierr = PetscLogFlops(1.3333*8*b->mbs);CHKERRQ(ierr); /* from inverting diagonal blocks */
318   PetscFunctionReturn(0);
319 }
320 /*
321       Version for when blocks are 2 by 2 Using natural ordering
322 */
323 #undef __FUNCT__
324 #define __FUNCT__ "MatLUFactorNumeric_SeqBAIJ_2_NaturalOrdering"
325 PetscErrorCode MatLUFactorNumeric_SeqBAIJ_2_NaturalOrdering(Mat C,Mat A,const MatFactorInfo *info)
326 {
327   Mat_SeqBAIJ    *a = (Mat_SeqBAIJ*)A->data,*b = (Mat_SeqBAIJ *)C->data;
328   PetscErrorCode ierr;
329   PetscInt       i,j,n = a->mbs,*bi = b->i,*bj = b->j;
330   PetscInt       *ajtmpold,*ajtmp,nz,row;
331   PetscInt       *diag_offset = b->diag,*ai=a->i,*aj=a->j,*pj;
332   MatScalar      *pv,*v,*rtmp,*pc,*w,*x;
333   MatScalar      p1,p2,p3,p4,m1,m2,m3,m4,x1,x2,x3,x4;
334   MatScalar      *ba = b->a,*aa = a->a;
335   PetscReal      shift = info->shiftinblocks;
336 
337   PetscFunctionBegin;
338   ierr = PetscMalloc(4*(n+1)*sizeof(MatScalar),&rtmp);CHKERRQ(ierr);
339   for (i=0; i<n; i++) {
340     nz    = bi[i+1] - bi[i];
341     ajtmp = bj + bi[i];
342     for  (j=0; j<nz; j++) {
343       x = rtmp+4*ajtmp[j];
344       x[0]  = x[1]  = x[2]  = x[3]  = 0.0;
345     }
346     /* load in initial (unfactored row) */
347     nz       = ai[i+1] - ai[i];
348     ajtmpold = aj + ai[i];
349     v        = aa + 4*ai[i];
350     for (j=0; j<nz; j++) {
351       x    = rtmp+4*ajtmpold[j];
352       x[0]  = v[0];  x[1]  = v[1];  x[2]  = v[2];  x[3]  = v[3];
353       v    += 4;
354     }
355     row = *ajtmp++;
356     while (row < i) {
357       pc  = rtmp + 4*row;
358       p1  = pc[0];  p2  = pc[1];  p3  = pc[2];  p4  = pc[3];
359       if (p1 != 0.0 || p2 != 0.0 || p3 != 0.0 || p4 != 0.0) {
360         pv = ba + 4*diag_offset[row];
361         pj = bj + diag_offset[row] + 1;
362         x1  = pv[0];  x2  = pv[1];  x3  = pv[2];  x4  = pv[3];
363         pc[0] = m1 = p1*x1 + p3*x2;
364         pc[1] = m2 = p2*x1 + p4*x2;
365         pc[2] = m3 = p1*x3 + p3*x4;
366         pc[3] = m4 = p2*x3 + p4*x4;
367         nz = bi[row+1] - diag_offset[row] - 1;
368         pv += 4;
369         for (j=0; j<nz; j++) {
370           x1   = pv[0];  x2  = pv[1];   x3 = pv[2];  x4  = pv[3];
371           x    = rtmp + 4*pj[j];
372           x[0] -= m1*x1 + m3*x2;
373           x[1] -= m2*x1 + m4*x2;
374           x[2] -= m1*x3 + m3*x4;
375           x[3] -= m2*x3 + m4*x4;
376           pv   += 4;
377         }
378         ierr = PetscLogFlops(16.0*nz+12.0);CHKERRQ(ierr);
379       }
380       row = *ajtmp++;
381     }
382     /* finished row so stick it into b->a */
383     pv = ba + 4*bi[i];
384     pj = bj + bi[i];
385     nz = bi[i+1] - bi[i];
386     for (j=0; j<nz; j++) {
387       x      = rtmp+4*pj[j];
388       pv[0]  = x[0];  pv[1]  = x[1];  pv[2]  = x[2];  pv[3]  = x[3];
389       /*
390       printf(" col %d:",pj[j]);
391       PetscInt j1;
392       for (j1=0; j1<4; j1++) printf(" %g,",*(pv+j1));
393       printf("\n");
394       */
395       pv   += 4;
396     }
397     /* invert diagonal block */
398     w = ba + 4*diag_offset[i];
399     /*
400     printf(" \n%d -th: diag: ",i);
401     for (j=0; j<4; j++){
402       printf(" %g,",w[j]);
403     }
404     printf("\n----------------------------\n");
405     */
406     ierr = Kernel_A_gets_inverse_A_2(w,shift);CHKERRQ(ierr);
407   }
408 
409   ierr = PetscFree(rtmp);CHKERRQ(ierr);
410   C->ops->solve          = MatSolve_SeqBAIJ_2_NaturalOrdering;
411   C->ops->solvetranspose = MatSolveTranspose_SeqBAIJ_2_NaturalOrdering;
412   C->assembled = PETSC_TRUE;
413   ierr = PetscLogFlops(1.3333*8*b->mbs);CHKERRQ(ierr); /* from inverting diagonal blocks */
414   PetscFunctionReturn(0);
415 }
416 
417 /* ----------------------------------------------------------- */
418 /*
419      Version for when blocks are 1 by 1.
420 */
421 #undef __FUNCT__
422 #define __FUNCT__ "MatLUFactorNumeric_SeqBAIJ_1"
423 PetscErrorCode MatLUFactorNumeric_SeqBAIJ_1(Mat C,Mat A,const MatFactorInfo *info)
424 {
425   Mat_SeqBAIJ    *a = (Mat_SeqBAIJ*)A->data,*b = (Mat_SeqBAIJ *)C->data;
426   IS             isrow = b->row,isicol = b->icol;
427   PetscErrorCode ierr;
428   const PetscInt *r,*ic;
429   PetscInt       i,j,n = a->mbs,*bi = b->i,*bj = b->j;
430   PetscInt       *ajtmpold,*ajtmp,nz,row,*ai = a->i,*aj = a->j;
431   PetscInt       *diag_offset = b->diag,diag,*pj;
432   MatScalar      *pv,*v,*rtmp,multiplier,*pc;
433   MatScalar      *ba = b->a,*aa = a->a;
434   PetscTruth     row_identity, col_identity;
435 
436   PetscFunctionBegin;
437   ierr  = ISGetIndices(isrow,&r);CHKERRQ(ierr);
438   ierr  = ISGetIndices(isicol,&ic);CHKERRQ(ierr);
439   ierr  = PetscMalloc((n+1)*sizeof(MatScalar),&rtmp);CHKERRQ(ierr);
440 
441   for (i=0; i<n; i++) {
442     nz    = bi[i+1] - bi[i];
443     ajtmp = bj + bi[i];
444     for  (j=0; j<nz; j++) rtmp[ajtmp[j]] = 0.0;
445 
446     /* load in initial (unfactored row) */
447     nz       = ai[r[i]+1] - ai[r[i]];
448     ajtmpold = aj + ai[r[i]];
449     v        = aa + ai[r[i]];
450     for (j=0; j<nz; j++) rtmp[ic[ajtmpold[j]]] =  v[j];
451 
452     row = *ajtmp++;
453     while (row < i) {
454       pc = rtmp + row;
455       if (*pc != 0.0) {
456         pv         = ba + diag_offset[row];
457         pj         = bj + diag_offset[row] + 1;
458         multiplier = *pc * *pv++;
459         *pc        = multiplier;
460         nz         = bi[row+1] - diag_offset[row] - 1;
461         for (j=0; j<nz; j++) rtmp[pj[j]] -= multiplier * pv[j];
462         ierr = PetscLogFlops(1.0+2.0*nz);CHKERRQ(ierr);
463       }
464       row = *ajtmp++;
465     }
466     /* finished row so stick it into b->a */
467     pv = ba + bi[i];
468     pj = bj + bi[i];
469     nz = bi[i+1] - bi[i];
470     for (j=0; j<nz; j++) {pv[j] = rtmp[pj[j]];}
471     diag = diag_offset[i] - bi[i];
472     /* check pivot entry for current row */
473     if (pv[diag] == 0.0) {
474       SETERRQ2(PETSC_ERR_MAT_LU_ZRPVT,"Zero pivot: row in original ordering %D in permuted ordering %D",r[i],i);
475     }
476     pv[diag] = 1.0/pv[diag];
477   }
478 
479   ierr = PetscFree(rtmp);CHKERRQ(ierr);
480   ierr = ISRestoreIndices(isicol,&ic);CHKERRQ(ierr);
481   ierr = ISRestoreIndices(isrow,&r);CHKERRQ(ierr);
482   ierr = ISIdentity(isrow,&row_identity);CHKERRQ(ierr);
483   ierr = ISIdentity(isicol,&col_identity);CHKERRQ(ierr);
484   if (row_identity && col_identity) {
485     C->ops->solve          = MatSolve_SeqBAIJ_1_NaturalOrdering;
486     C->ops->solvetranspose = MatSolveTranspose_SeqBAIJ_1_NaturalOrdering;
487   } else {
488     C->ops->solve          = MatSolve_SeqBAIJ_1;
489     C->ops->solvetranspose = MatSolveTranspose_SeqBAIJ_1;
490   }
491   C->assembled = PETSC_TRUE;
492   ierr = PetscLogFlops(C->cmap->n);CHKERRQ(ierr);
493   PetscFunctionReturn(0);
494 }
495 
496 EXTERN_C_BEGIN
497 #undef __FUNCT__
498 #define __FUNCT__ "MatGetFactor_seqbaij_petsc"
499 PetscErrorCode MatGetFactor_seqbaij_petsc(Mat A,MatFactorType ftype,Mat *B)
500 {
501   PetscInt           n = A->rmap->n;
502   PetscErrorCode     ierr;
503 
504   PetscFunctionBegin;
505   ierr = MatCreate(((PetscObject)A)->comm,B);CHKERRQ(ierr);
506   ierr = MatSetSizes(*B,n,n,n,n);CHKERRQ(ierr);
507   if (ftype == MAT_FACTOR_LU || ftype == MAT_FACTOR_ILU || ftype == MAT_FACTOR_ILUDT) {
508     ierr = MatSetType(*B,MATSEQBAIJ);CHKERRQ(ierr);
509     (*B)->ops->lufactorsymbolic  = MatLUFactorSymbolic_SeqBAIJ;
510     (*B)->ops->ilufactorsymbolic = MatILUFactorSymbolic_SeqBAIJ;
511     (*B)->ops->iludtfactor       = MatILUDTFactor_SeqBAIJ;
512   } else if (ftype == MAT_FACTOR_CHOLESKY || ftype == MAT_FACTOR_ICC) {
513     ierr = MatSetType(*B,MATSEQSBAIJ);CHKERRQ(ierr);
514     ierr = MatSeqSBAIJSetPreallocation(*B,1,MAT_SKIP_ALLOCATION,PETSC_NULL);CHKERRQ(ierr);
515     (*B)->ops->iccfactorsymbolic      = MatICCFactorSymbolic_SeqBAIJ;
516     (*B)->ops->choleskyfactorsymbolic = MatCholeskyFactorSymbolic_SeqBAIJ;
517   } else SETERRQ(PETSC_ERR_SUP,"Factor type not supported");
518   (*B)->factor = ftype;
519   PetscFunctionReturn(0);
520 }
521 EXTERN_C_END
522 
523 EXTERN_C_BEGIN
524 #undef __FUNCT__
525 #define __FUNCT__ "MatGetFactorAvailable_seqaij_petsc"
526 PetscErrorCode MatGetFactorAvailable_seqbaij_petsc(Mat A,MatFactorType ftype,PetscTruth *flg)
527 {
528   PetscFunctionBegin;
529   *flg = PETSC_TRUE;
530   PetscFunctionReturn(0);
531 }
532 EXTERN_C_END
533 
534 /* ----------------------------------------------------------- */
535 #undef __FUNCT__
536 #define __FUNCT__ "MatLUFactor_SeqBAIJ"
537 PetscErrorCode MatLUFactor_SeqBAIJ(Mat A,IS row,IS col,const MatFactorInfo *info)
538 {
539   PetscErrorCode ierr;
540   Mat            C;
541 
542   PetscFunctionBegin;
543   ierr = MatGetFactor(A,MAT_SOLVER_PETSC,MAT_FACTOR_LU,&C);CHKERRQ(ierr);
544   ierr = MatLUFactorSymbolic(C,A,row,col,info);CHKERRQ(ierr);
545   ierr = MatLUFactorNumeric(C,A,info);CHKERRQ(ierr);
546   A->ops->solve            = C->ops->solve;
547   A->ops->solvetranspose   = C->ops->solvetranspose;
548   ierr = MatHeaderCopy(A,C);CHKERRQ(ierr);
549   ierr = PetscLogObjectParent(A,((Mat_SeqBAIJ*)(A->data))->icol);CHKERRQ(ierr);
550   PetscFunctionReturn(0);
551 }
552 
553 #include "../src/mat/impls/sbaij/seq/sbaij.h"
554 #undef __FUNCT__
555 #define __FUNCT__ "MatCholeskyFactorNumeric_SeqBAIJ_N"
556 PetscErrorCode MatCholeskyFactorNumeric_SeqBAIJ_N(Mat C,Mat A,const MatFactorInfo *info)
557 {
558   PetscErrorCode ierr;
559   Mat_SeqBAIJ    *a=(Mat_SeqBAIJ*)A->data;
560   Mat_SeqSBAIJ   *b=(Mat_SeqSBAIJ*)C->data;
561   IS             ip=b->row;
562   const PetscInt *rip;
563   PetscInt       i,j,mbs=a->mbs,bs=A->rmap->bs,*bi=b->i,*bj=b->j,*bcol;
564   PetscInt       *ai=a->i,*aj=a->j;
565   PetscInt       k,jmin,jmax,*jl,*il,col,nexti,ili,nz;
566   MatScalar      *rtmp,*ba=b->a,*bval,*aa=a->a,dk,uikdi;
567   PetscReal      zeropivot,rs,shiftnz;
568   PetscReal      shiftpd;
569   ChShift_Ctx    sctx;
570   PetscInt       newshift;
571 
572   PetscFunctionBegin;
573   if (bs > 1) {
574     if (!a->sbaijMat){
575       ierr = MatConvert(A,MATSEQSBAIJ,MAT_INITIAL_MATRIX,&a->sbaijMat);CHKERRQ(ierr);
576     }
577     ierr = (a->sbaijMat)->ops->choleskyfactornumeric(C,a->sbaijMat,info);CHKERRQ(ierr);
578     ierr = MatDestroy(a->sbaijMat);CHKERRQ(ierr);
579     a->sbaijMat = PETSC_NULL;
580     PetscFunctionReturn(0);
581   }
582 
583   /* initialization */
584   shiftnz   = info->shiftnz;
585   shiftpd   = info->shiftpd;
586   zeropivot = info->zeropivot;
587 
588   ierr  = ISGetIndices(ip,&rip);CHKERRQ(ierr);
589   nz   = (2*mbs+1)*sizeof(PetscInt)+mbs*sizeof(MatScalar);
590   ierr = PetscMalloc(nz,&il);CHKERRQ(ierr);
591   jl   = il + mbs;
592   rtmp = (MatScalar*)(jl + mbs);
593 
594   sctx.shift_amount = 0;
595   sctx.nshift       = 0;
596   do {
597     sctx.chshift = PETSC_FALSE;
598     for (i=0; i<mbs; i++) {
599       rtmp[i] = 0.0; jl[i] = mbs; il[0] = 0;
600     }
601 
602     for (k = 0; k<mbs; k++){
603       bval = ba + bi[k];
604       /* initialize k-th row by the perm[k]-th row of A */
605       jmin = ai[rip[k]]; jmax = ai[rip[k]+1];
606       for (j = jmin; j < jmax; j++){
607         col = rip[aj[j]];
608         if (col >= k){ /* only take upper triangular entry */
609           rtmp[col] = aa[j];
610           *bval++  = 0.0; /* for in-place factorization */
611         }
612       }
613 
614       /* shift the diagonal of the matrix */
615       if (sctx.nshift) rtmp[k] += sctx.shift_amount;
616 
617       /* modify k-th row by adding in those rows i with U(i,k)!=0 */
618       dk = rtmp[k];
619       i = jl[k]; /* first row to be added to k_th row  */
620 
621       while (i < k){
622         nexti = jl[i]; /* next row to be added to k_th row */
623 
624         /* compute multiplier, update diag(k) and U(i,k) */
625         ili = il[i];  /* index of first nonzero element in U(i,k:bms-1) */
626         uikdi = - ba[ili]*ba[bi[i]];  /* diagonal(k) */
627         dk += uikdi*ba[ili];
628         ba[ili] = uikdi; /* -U(i,k) */
629 
630         /* add multiple of row i to k-th row */
631         jmin = ili + 1; jmax = bi[i+1];
632         if (jmin < jmax){
633           for (j=jmin; j<jmax; j++) rtmp[bj[j]] += uikdi*ba[j];
634           /* update il and jl for row i */
635           il[i] = jmin;
636           j = bj[jmin]; jl[i] = jl[j]; jl[j] = i;
637         }
638         i = nexti;
639       }
640 
641       /* shift the diagonals when zero pivot is detected */
642       /* compute rs=sum of abs(off-diagonal) */
643       rs   = 0.0;
644       jmin = bi[k]+1;
645       nz   = bi[k+1] - jmin;
646       if (nz){
647         bcol = bj + jmin;
648         while (nz--){
649           rs += PetscAbsScalar(rtmp[*bcol]);
650           bcol++;
651         }
652       }
653 
654       sctx.rs = rs;
655       sctx.pv = dk;
656       ierr = MatCholeskyCheckShift_inline(info,sctx,k,newshift);CHKERRQ(ierr);
657       if (newshift == 1) break;
658 
659       /* copy data into U(k,:) */
660       ba[bi[k]] = 1.0/dk; /* U(k,k) */
661       jmin = bi[k]+1; jmax = bi[k+1];
662       if (jmin < jmax) {
663         for (j=jmin; j<jmax; j++){
664           col = bj[j]; ba[j] = rtmp[col]; rtmp[col] = 0.0;
665         }
666         /* add the k-th row into il and jl */
667         il[k] = jmin;
668         i = bj[jmin]; jl[k] = jl[i]; jl[i] = k;
669       }
670     }
671   } while (sctx.chshift);
672   ierr = PetscFree(il);CHKERRQ(ierr);
673 
674   ierr = ISRestoreIndices(ip,&rip);CHKERRQ(ierr);
675   C->assembled    = PETSC_TRUE;
676   C->preallocated = PETSC_TRUE;
677   ierr = PetscLogFlops(C->rmap->N);CHKERRQ(ierr);
678   if (sctx.nshift){
679     if (shiftpd) {
680       ierr = PetscInfo2(A,"number of shiftpd tries %D, shift_amount %G\n",sctx.nshift,sctx.shift_amount);CHKERRQ(ierr);
681     } else if (shiftnz) {
682       ierr = PetscInfo2(A,"number of shiftnz tries %D, shift_amount %G\n",sctx.nshift,sctx.shift_amount);CHKERRQ(ierr);
683     }
684   }
685   PetscFunctionReturn(0);
686 }
687 
688 #undef __FUNCT__
689 #define __FUNCT__ "MatCholeskyFactorNumeric_SeqBAIJ_N_NaturalOrdering"
690 PetscErrorCode MatCholeskyFactorNumeric_SeqBAIJ_N_NaturalOrdering(Mat C,Mat A,const MatFactorInfo *info)
691 {
692   Mat_SeqBAIJ    *a=(Mat_SeqBAIJ*)A->data;
693   Mat_SeqSBAIJ   *b=(Mat_SeqSBAIJ*)C->data;
694   PetscErrorCode ierr;
695   PetscInt       i,j,am=a->mbs;
696   PetscInt       *ai=a->i,*aj=a->j,*bi=b->i,*bj=b->j;
697   PetscInt       k,jmin,*jl,*il,nexti,ili,*acol,*bcol,nz;
698   MatScalar      *rtmp,*ba=b->a,*aa=a->a,dk,uikdi,*aval,*bval;
699   PetscReal      zeropivot,rs,shiftnz;
700   PetscReal      shiftpd;
701   ChShift_Ctx    sctx;
702   PetscInt       newshift;
703 
704   PetscFunctionBegin;
705   /* initialization */
706   shiftnz   = info->shiftnz;
707   shiftpd   = info->shiftpd;
708   zeropivot = info->zeropivot;
709 
710   nz   = (2*am+1)*sizeof(PetscInt)+am*sizeof(MatScalar);
711   ierr = PetscMalloc(nz,&il);CHKERRQ(ierr);
712   jl   = il + am;
713   rtmp = (MatScalar*)(jl + am);
714 
715   sctx.shift_amount = 0;
716   sctx.nshift       = 0;
717   do {
718     sctx.chshift = PETSC_FALSE;
719     for (i=0; i<am; i++) {
720       rtmp[i] = 0.0; jl[i] = am; il[0] = 0;
721     }
722 
723     for (k = 0; k<am; k++){
724     /* initialize k-th row with elements nonzero in row perm(k) of A */
725       nz   = ai[k+1] - ai[k];
726       acol = aj + ai[k];
727       aval = aa + ai[k];
728       bval = ba + bi[k];
729       while (nz -- ){
730         if (*acol < k) { /* skip lower triangular entries */
731           acol++; aval++;
732         } else {
733           rtmp[*acol++] = *aval++;
734           *bval++       = 0.0; /* for in-place factorization */
735         }
736       }
737 
738       /* shift the diagonal of the matrix */
739       if (sctx.nshift) rtmp[k] += sctx.shift_amount;
740 
741       /* modify k-th row by adding in those rows i with U(i,k)!=0 */
742       dk = rtmp[k];
743       i  = jl[k]; /* first row to be added to k_th row  */
744 
745       while (i < k){
746         nexti = jl[i]; /* next row to be added to k_th row */
747         /* compute multiplier, update D(k) and U(i,k) */
748         ili   = il[i];  /* index of first nonzero element in U(i,k:bms-1) */
749         uikdi = - ba[ili]*ba[bi[i]];
750         dk   += uikdi*ba[ili];
751         ba[ili] = uikdi; /* -U(i,k) */
752 
753         /* add multiple of row i to k-th row ... */
754         jmin = ili + 1;
755         nz   = bi[i+1] - jmin;
756         if (nz > 0){
757           bcol = bj + jmin;
758           bval = ba + jmin;
759           while (nz --) rtmp[*bcol++] += uikdi*(*bval++);
760           /* update il and jl for i-th row */
761           il[i] = jmin;
762           j = bj[jmin]; jl[i] = jl[j]; jl[j] = i;
763         }
764         i = nexti;
765       }
766 
767       /* shift the diagonals when zero pivot is detected */
768       /* compute rs=sum of abs(off-diagonal) */
769       rs   = 0.0;
770       jmin = bi[k]+1;
771       nz   = bi[k+1] - jmin;
772       if (nz){
773         bcol = bj + jmin;
774         while (nz--){
775           rs += PetscAbsScalar(rtmp[*bcol]);
776           bcol++;
777         }
778       }
779 
780       sctx.rs = rs;
781       sctx.pv = dk;
782       ierr = MatCholeskyCheckShift_inline(info,sctx,k,newshift);CHKERRQ(ierr);
783       if (newshift == 1) break;    /* sctx.shift_amount is updated */
784 
785       /* copy data into U(k,:) */
786       ba[bi[k]] = 1.0/dk;
787       jmin      = bi[k]+1;
788       nz        = bi[k+1] - jmin;
789       if (nz){
790         bcol = bj + jmin;
791         bval = ba + jmin;
792         while (nz--){
793           *bval++       = rtmp[*bcol];
794           rtmp[*bcol++] = 0.0;
795         }
796         /* add k-th row into il and jl */
797         il[k] = jmin;
798         i = bj[jmin]; jl[k] = jl[i]; jl[i] = k;
799       }
800     }
801   } while (sctx.chshift);
802   ierr = PetscFree(il);CHKERRQ(ierr);
803 
804   C->ops->solve                 = MatSolve_SeqSBAIJ_1_NaturalOrdering;
805   C->ops->solvetranspose        = MatSolve_SeqSBAIJ_1_NaturalOrdering;
806   C->assembled    = PETSC_TRUE;
807   C->preallocated = PETSC_TRUE;
808   ierr = PetscLogFlops(C->rmap->N);CHKERRQ(ierr);
809     if (sctx.nshift){
810     if (shiftnz) {
811       ierr = PetscInfo2(A,"number of shiftnz tries %D, shift_amount %G\n",sctx.nshift,sctx.shift_amount);CHKERRQ(ierr);
812     } else if (shiftpd) {
813       ierr = PetscInfo2(A,"number of shiftpd tries %D, shift_amount %G\n",sctx.nshift,sctx.shift_amount);CHKERRQ(ierr);
814     }
815   }
816   PetscFunctionReturn(0);
817 }
818 
819 #include "petscbt.h"
820 #include "../src/mat/utils/freespace.h"
821 #undef __FUNCT__
822 #define __FUNCT__ "MatICCFactorSymbolic_SeqBAIJ"
823 PetscErrorCode MatICCFactorSymbolic_SeqBAIJ(Mat fact,Mat A,IS perm,const MatFactorInfo *info)
824 {
825   Mat_SeqBAIJ        *a = (Mat_SeqBAIJ*)A->data;
826   Mat_SeqSBAIJ       *b;
827   Mat                B;
828   PetscErrorCode     ierr;
829   PetscTruth         perm_identity;
830   PetscInt           reallocs=0,i,*ai=a->i,*aj=a->j,am=a->mbs,bs=A->rmap->bs,*ui;
831   const PetscInt     *rip;
832   PetscInt           jmin,jmax,nzk,k,j,*jl,prow,*il,nextprow;
833   PetscInt           nlnk,*lnk,*lnk_lvl=PETSC_NULL,ncols,ncols_upper,*cols,*cols_lvl,*uj,**uj_ptr,**uj_lvl_ptr;
834   PetscReal          fill=info->fill,levels=info->levels;
835   PetscFreeSpaceList free_space=PETSC_NULL,current_space=PETSC_NULL;
836   PetscFreeSpaceList free_space_lvl=PETSC_NULL,current_space_lvl=PETSC_NULL;
837   PetscBT            lnkbt;
838 
839   PetscFunctionBegin;
840   if (bs > 1){
841     if (!a->sbaijMat){
842       ierr = MatConvert(A,MATSEQSBAIJ,MAT_INITIAL_MATRIX,&a->sbaijMat);CHKERRQ(ierr);
843     }
844     (fact)->ops->iccfactorsymbolic = MatICCFactorSymbolic_SeqSBAIJ;  /* undue the change made in MatGetFactor_seqbaij_petsc */
845     ierr = MatICCFactorSymbolic(fact,a->sbaijMat,perm,info);CHKERRQ(ierr);
846     PetscFunctionReturn(0);
847   }
848 
849   ierr = ISIdentity(perm,&perm_identity);CHKERRQ(ierr);
850   ierr = ISGetIndices(perm,&rip);CHKERRQ(ierr);
851 
852   /* special case that simply copies fill pattern */
853   if (!levels && perm_identity) {
854     ierr = MatMarkDiagonal_SeqBAIJ(A);CHKERRQ(ierr);
855     ierr = PetscMalloc((am+1)*sizeof(PetscInt),&ui);CHKERRQ(ierr);
856     for (i=0; i<am; i++) {
857       ui[i] = ai[i+1] - a->diag[i]; /* ui: rowlengths - changes when !perm_identity */
858     }
859     B = fact;
860     ierr = MatSeqSBAIJSetPreallocation(B,1,0,ui);CHKERRQ(ierr);
861 
862 
863     b  = (Mat_SeqSBAIJ*)B->data;
864     uj = b->j;
865     for (i=0; i<am; i++) {
866       aj = a->j + a->diag[i];
867       for (j=0; j<ui[i]; j++){
868         *uj++ = *aj++;
869       }
870       b->ilen[i] = ui[i];
871     }
872     ierr = PetscFree(ui);CHKERRQ(ierr);
873     B->factor = MAT_FACTOR_NONE;
874     ierr = MatAssemblyBegin(B,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr);
875     ierr = MatAssemblyEnd(B,MAT_FINAL_ASSEMBLY);CHKERRQ(ierr);
876     B->factor = MAT_FACTOR_ICC;
877 
878     B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqBAIJ_N_NaturalOrdering;
879     PetscFunctionReturn(0);
880   }
881 
882   /* initialization */
883   ierr  = PetscMalloc((am+1)*sizeof(PetscInt),&ui);CHKERRQ(ierr);
884   ui[0] = 0;
885   ierr  = PetscMalloc((2*am+1)*sizeof(PetscInt),&cols_lvl);CHKERRQ(ierr);
886 
887   /* jl: linked list for storing indices of the pivot rows
888      il: il[i] points to the 1st nonzero entry of U(i,k:am-1) */
889   ierr = PetscMalloc((2*am+1)*sizeof(PetscInt)+2*am*sizeof(PetscInt*),&jl);CHKERRQ(ierr);
890   il         = jl + am;
891   uj_ptr     = (PetscInt**)(il + am);
892   uj_lvl_ptr = (PetscInt**)(uj_ptr + am);
893   for (i=0; i<am; i++){
894     jl[i] = am; il[i] = 0;
895   }
896 
897   /* create and initialize a linked list for storing column indices of the active row k */
898   nlnk = am + 1;
899   ierr = PetscIncompleteLLCreate(am,am,nlnk,lnk,lnk_lvl,lnkbt);CHKERRQ(ierr);
900 
901   /* initial FreeSpace size is fill*(ai[am]+1) */
902   ierr = PetscFreeSpaceGet((PetscInt)(fill*(ai[am]+1)),&free_space);CHKERRQ(ierr);
903   current_space = free_space;
904   ierr = PetscFreeSpaceGet((PetscInt)(fill*(ai[am]+1)),&free_space_lvl);CHKERRQ(ierr);
905   current_space_lvl = free_space_lvl;
906 
907   for (k=0; k<am; k++){  /* for each active row k */
908     /* initialize lnk by the column indices of row rip[k] of A */
909     nzk   = 0;
910     ncols = ai[rip[k]+1] - ai[rip[k]];
911     ncols_upper = 0;
912     cols        = cols_lvl + am;
913     for (j=0; j<ncols; j++){
914       i = rip[*(aj + ai[rip[k]] + j)];
915       if (i >= k){ /* only take upper triangular entry */
916         cols[ncols_upper] = i;
917         cols_lvl[ncols_upper] = -1;  /* initialize level for nonzero entries */
918         ncols_upper++;
919       }
920     }
921     ierr = PetscIncompleteLLAdd(ncols_upper,cols,levels,cols_lvl,am,nlnk,lnk,lnk_lvl,lnkbt);CHKERRQ(ierr);
922     nzk += nlnk;
923 
924     /* update lnk by computing fill-in for each pivot row to be merged in */
925     prow = jl[k]; /* 1st pivot row */
926 
927     while (prow < k){
928       nextprow = jl[prow];
929 
930       /* merge prow into k-th row */
931       jmin = il[prow] + 1;  /* index of the 2nd nzero entry in U(prow,k:am-1) */
932       jmax = ui[prow+1];
933       ncols = jmax-jmin;
934       i     = jmin - ui[prow];
935       cols = uj_ptr[prow] + i; /* points to the 2nd nzero entry in U(prow,k:am-1) */
936       for (j=0; j<ncols; j++) cols_lvl[j] = *(uj_lvl_ptr[prow] + i + j);
937       ierr = PetscIncompleteLLAddSorted(ncols,cols,levels,cols_lvl,am,nlnk,lnk,lnk_lvl,lnkbt);CHKERRQ(ierr);
938       nzk += nlnk;
939 
940       /* update il and jl for prow */
941       if (jmin < jmax){
942         il[prow] = jmin;
943         j = *cols; jl[prow] = jl[j]; jl[j] = prow;
944       }
945       prow = nextprow;
946     }
947 
948     /* if free space is not available, make more free space */
949     if (current_space->local_remaining<nzk) {
950       i = am - k + 1; /* num of unfactored rows */
951       i = PetscMin(i*nzk, i*(i-1)); /* i*nzk, i*(i-1): estimated and max additional space needed */
952       ierr = PetscFreeSpaceGet(i,&current_space);CHKERRQ(ierr);
953       ierr = PetscFreeSpaceGet(i,&current_space_lvl);CHKERRQ(ierr);
954       reallocs++;
955     }
956 
957     /* copy data into free_space and free_space_lvl, then initialize lnk */
958     ierr = PetscIncompleteLLClean(am,am,nzk,lnk,lnk_lvl,current_space->array,current_space_lvl->array,lnkbt);CHKERRQ(ierr);
959 
960     /* add the k-th row into il and jl */
961     if (nzk-1 > 0){
962       i = current_space->array[1]; /* col value of the first nonzero element in U(k, k+1:am-1) */
963       jl[k] = jl[i]; jl[i] = k;
964       il[k] = ui[k] + 1;
965     }
966     uj_ptr[k]     = current_space->array;
967     uj_lvl_ptr[k] = current_space_lvl->array;
968 
969     current_space->array           += nzk;
970     current_space->local_used      += nzk;
971     current_space->local_remaining -= nzk;
972 
973     current_space_lvl->array           += nzk;
974     current_space_lvl->local_used      += nzk;
975     current_space_lvl->local_remaining -= nzk;
976 
977     ui[k+1] = ui[k] + nzk;
978   }
979 
980 #if defined(PETSC_USE_INFO)
981   if (ai[am] != 0) {
982     PetscReal af = ((PetscReal)(2*ui[am]-am))/((PetscReal)ai[am]);
983     ierr = PetscInfo3(A,"Reallocs %D Fill ratio:given %G needed %G\n",reallocs,fill,af);CHKERRQ(ierr);
984     ierr = PetscInfo1(A,"Run with -pc_factor_fill %G or use \n",af);CHKERRQ(ierr);
985     ierr = PetscInfo1(A,"PCFactorSetFill(pc,%G) for best performance.\n",af);CHKERRQ(ierr);
986   } else {
987     ierr = PetscInfo(A,"Empty matrix.\n");CHKERRQ(ierr);
988   }
989 #endif
990 
991   ierr = ISRestoreIndices(perm,&rip);CHKERRQ(ierr);
992   ierr = PetscFree(jl);CHKERRQ(ierr);
993   ierr = PetscFree(cols_lvl);CHKERRQ(ierr);
994 
995   /* destroy list of free space and other temporary array(s) */
996   ierr = PetscMalloc((ui[am]+1)*sizeof(PetscInt),&uj);CHKERRQ(ierr);
997   ierr = PetscFreeSpaceContiguous(&free_space,uj);CHKERRQ(ierr);
998   ierr = PetscIncompleteLLDestroy(lnk,lnkbt);CHKERRQ(ierr);
999   ierr = PetscFreeSpaceDestroy(free_space_lvl);CHKERRQ(ierr);
1000 
1001   /* put together the new matrix in MATSEQSBAIJ format */
1002   B = fact;
1003   ierr = MatSeqSBAIJSetPreallocation(B,1,MAT_SKIP_ALLOCATION,PETSC_NULL);CHKERRQ(ierr);
1004 
1005   b = (Mat_SeqSBAIJ*)B->data;
1006   b->singlemalloc = PETSC_FALSE;
1007   b->free_a       = PETSC_TRUE;
1008   b->free_ij       = PETSC_TRUE;
1009   ierr = PetscMalloc((ui[am]+1)*sizeof(MatScalar),&b->a);CHKERRQ(ierr);
1010   b->j    = uj;
1011   b->i    = ui;
1012   b->diag = 0;
1013   b->ilen = 0;
1014   b->imax = 0;
1015   b->row  = perm;
1016   b->pivotinblocks = PETSC_FALSE; /* need to get from MatFactorInfo */
1017   ierr    = PetscObjectReference((PetscObject)perm);CHKERRQ(ierr);
1018   b->icol = perm;
1019   ierr    = PetscObjectReference((PetscObject)perm);CHKERRQ(ierr);
1020   ierr    = PetscMalloc((am+1)*sizeof(PetscScalar),&b->solve_work);CHKERRQ(ierr);
1021   ierr    = PetscLogObjectMemory(B,(ui[am]-am)*(sizeof(PetscInt)+sizeof(MatScalar)));CHKERRQ(ierr);
1022   b->maxnz = b->nz = ui[am];
1023 
1024   B->info.factor_mallocs    = reallocs;
1025   B->info.fill_ratio_given  = fill;
1026   if (ai[am] != 0) {
1027     B->info.fill_ratio_needed = ((PetscReal)ui[am])/((PetscReal)ai[am]);
1028   } else {
1029     B->info.fill_ratio_needed = 0.0;
1030   }
1031   if (perm_identity){
1032     B->ops->solve           = MatSolve_SeqSBAIJ_1_NaturalOrdering;
1033     B->ops->solvetranspose  = MatSolve_SeqSBAIJ_1_NaturalOrdering;
1034     B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqBAIJ_N_NaturalOrdering;
1035   } else {
1036     (fact)->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqBAIJ_N;
1037   }
1038   PetscFunctionReturn(0);
1039 }
1040 
1041 #undef __FUNCT__
1042 #define __FUNCT__ "MatCholeskyFactorSymbolic_SeqBAIJ"
1043 PetscErrorCode MatCholeskyFactorSymbolic_SeqBAIJ(Mat fact,Mat A,IS perm,const MatFactorInfo *info)
1044 {
1045   Mat_SeqBAIJ        *a = (Mat_SeqBAIJ*)A->data;
1046   Mat_SeqSBAIJ       *b;
1047   Mat                B;
1048   PetscErrorCode     ierr;
1049   PetscTruth         perm_identity;
1050   PetscReal          fill = info->fill;
1051   const PetscInt     *rip;
1052   PetscInt           i,mbs=a->mbs,bs=A->rmap->bs,*ai=a->i,*aj=a->j,reallocs=0,prow;
1053   PetscInt           *jl,jmin,jmax,nzk,*ui,k,j,*il,nextprow;
1054   PetscInt           nlnk,*lnk,ncols,ncols_upper,*cols,*uj,**ui_ptr,*uj_ptr;
1055   PetscFreeSpaceList free_space=PETSC_NULL,current_space=PETSC_NULL;
1056   PetscBT            lnkbt;
1057 
1058   PetscFunctionBegin;
1059   if (bs > 1) { /* convert to seqsbaij */
1060     if (!a->sbaijMat){
1061       ierr = MatConvert(A,MATSEQSBAIJ,MAT_INITIAL_MATRIX,&a->sbaijMat);CHKERRQ(ierr);
1062     }
1063     (fact)->ops->choleskyfactorsymbolic = MatCholeskyFactorSymbolic_SeqSBAIJ; /* undue the change made in MatGetFactor_seqbaij_petsc */
1064     ierr = MatCholeskyFactorSymbolic(fact,a->sbaijMat,perm,info);CHKERRQ(ierr);
1065     PetscFunctionReturn(0);
1066   }
1067 
1068   /* check whether perm is the identity mapping */
1069   ierr = ISIdentity(perm,&perm_identity);CHKERRQ(ierr);
1070   if (!perm_identity) SETERRQ(PETSC_ERR_SUP,"Matrix reordering is not supported");
1071   ierr = ISGetIndices(perm,&rip);CHKERRQ(ierr);
1072 
1073   /* initialization */
1074   ierr  = PetscMalloc((mbs+1)*sizeof(PetscInt),&ui);CHKERRQ(ierr);
1075   ui[0] = 0;
1076 
1077   /* jl: linked list for storing indices of the pivot rows
1078      il: il[i] points to the 1st nonzero entry of U(i,k:mbs-1) */
1079   ierr = PetscMalloc((3*mbs+1)*sizeof(PetscInt)+mbs*sizeof(PetscInt*),&jl);CHKERRQ(ierr);
1080   il     = jl + mbs;
1081   cols   = il + mbs;
1082   ui_ptr = (PetscInt**)(cols + mbs);
1083   for (i=0; i<mbs; i++){
1084     jl[i] = mbs; il[i] = 0;
1085   }
1086 
1087   /* create and initialize a linked list for storing column indices of the active row k */
1088   nlnk = mbs + 1;
1089   ierr = PetscLLCreate(mbs,mbs,nlnk,lnk,lnkbt);CHKERRQ(ierr);
1090 
1091   /* initial FreeSpace size is fill*(ai[mbs]+1) */
1092   ierr = PetscFreeSpaceGet((PetscInt)(fill*(ai[mbs]+1)),&free_space);CHKERRQ(ierr);
1093   current_space = free_space;
1094 
1095   for (k=0; k<mbs; k++){  /* for each active row k */
1096     /* initialize lnk by the column indices of row rip[k] of A */
1097     nzk   = 0;
1098     ncols = ai[rip[k]+1] - ai[rip[k]];
1099     ncols_upper = 0;
1100     for (j=0; j<ncols; j++){
1101       i = rip[*(aj + ai[rip[k]] + j)];
1102       if (i >= k){ /* only take upper triangular entry */
1103         cols[ncols_upper] = i;
1104         ncols_upper++;
1105       }
1106     }
1107     ierr = PetscLLAdd(ncols_upper,cols,mbs,nlnk,lnk,lnkbt);CHKERRQ(ierr);
1108     nzk += nlnk;
1109 
1110     /* update lnk by computing fill-in for each pivot row to be merged in */
1111     prow = jl[k]; /* 1st pivot row */
1112 
1113     while (prow < k){
1114       nextprow = jl[prow];
1115       /* merge prow into k-th row */
1116       jmin = il[prow] + 1;  /* index of the 2nd nzero entry in U(prow,k:mbs-1) */
1117       jmax = ui[prow+1];
1118       ncols = jmax-jmin;
1119       uj_ptr = ui_ptr[prow] + jmin - ui[prow]; /* points to the 2nd nzero entry in U(prow,k:mbs-1) */
1120       ierr = PetscLLAddSorted(ncols,uj_ptr,mbs,nlnk,lnk,lnkbt);CHKERRQ(ierr);
1121       nzk += nlnk;
1122 
1123       /* update il and jl for prow */
1124       if (jmin < jmax){
1125         il[prow] = jmin;
1126         j = *uj_ptr; jl[prow] = jl[j]; jl[j] = prow;
1127       }
1128       prow = nextprow;
1129     }
1130 
1131     /* if free space is not available, make more free space */
1132     if (current_space->local_remaining<nzk) {
1133       i = mbs - k + 1; /* num of unfactored rows */
1134       i = PetscMin(i*nzk, i*(i-1)); /* i*nzk, i*(i-1): estimated and max additional space needed */
1135       ierr = PetscFreeSpaceGet(i,&current_space);CHKERRQ(ierr);
1136       reallocs++;
1137     }
1138 
1139     /* copy data into free space, then initialize lnk */
1140     ierr = PetscLLClean(mbs,mbs,nzk,lnk,current_space->array,lnkbt);CHKERRQ(ierr);
1141 
1142     /* add the k-th row into il and jl */
1143     if (nzk-1 > 0){
1144       i = current_space->array[1]; /* col value of the first nonzero element in U(k, k+1:mbs-1) */
1145       jl[k] = jl[i]; jl[i] = k;
1146       il[k] = ui[k] + 1;
1147     }
1148     ui_ptr[k] = current_space->array;
1149     current_space->array           += nzk;
1150     current_space->local_used      += nzk;
1151     current_space->local_remaining -= nzk;
1152 
1153     ui[k+1] = ui[k] + nzk;
1154   }
1155 
1156 #if defined(PETSC_USE_INFO)
1157   if (ai[mbs] != 0) {
1158     PetscReal af = ((PetscReal)ui[mbs])/((PetscReal)ai[mbs]);
1159     ierr = PetscInfo3(A,"Reallocs %D Fill ratio:given %G needed %G\n",reallocs,fill,af);CHKERRQ(ierr);
1160     ierr = PetscInfo1(A,"Run with -pc_factor_fill %G or use \n",af);CHKERRQ(ierr);
1161     ierr = PetscInfo1(A,"PCFactorSetFill(pc,%G) for best performance.\n",af);CHKERRQ(ierr);
1162   } else {
1163     ierr = PetscInfo(A,"Empty matrix.\n");CHKERRQ(ierr);
1164   }
1165 #endif
1166 
1167   ierr = ISRestoreIndices(perm,&rip);CHKERRQ(ierr);
1168   ierr = PetscFree(jl);CHKERRQ(ierr);
1169 
1170   /* destroy list of free space and other temporary array(s) */
1171   ierr = PetscMalloc((ui[mbs]+1)*sizeof(PetscInt),&uj);CHKERRQ(ierr);
1172   ierr = PetscFreeSpaceContiguous(&free_space,uj);CHKERRQ(ierr);
1173   ierr = PetscLLDestroy(lnk,lnkbt);CHKERRQ(ierr);
1174 
1175   /* put together the new matrix in MATSEQSBAIJ format */
1176   B    = fact;
1177   ierr = MatSeqSBAIJSetPreallocation(B,bs,MAT_SKIP_ALLOCATION,PETSC_NULL);CHKERRQ(ierr);
1178 
1179   b = (Mat_SeqSBAIJ*)B->data;
1180   b->singlemalloc = PETSC_FALSE;
1181   b->free_a       = PETSC_TRUE;
1182   b->free_ij      = PETSC_TRUE;
1183   ierr = PetscMalloc((ui[mbs]+1)*sizeof(MatScalar),&b->a);CHKERRQ(ierr);
1184   b->j    = uj;
1185   b->i    = ui;
1186   b->diag = 0;
1187   b->ilen = 0;
1188   b->imax = 0;
1189   b->row  = perm;
1190   b->pivotinblocks = PETSC_FALSE; /* need to get from MatFactorInfo */
1191   ierr    = PetscObjectReference((PetscObject)perm);CHKERRQ(ierr);
1192   b->icol = perm;
1193   ierr    = PetscObjectReference((PetscObject)perm);CHKERRQ(ierr);
1194   ierr    = PetscMalloc((mbs+1)*sizeof(PetscScalar),&b->solve_work);CHKERRQ(ierr);
1195   ierr    = PetscLogObjectMemory(B,(ui[mbs]-mbs)*(sizeof(PetscInt)+sizeof(MatScalar)));CHKERRQ(ierr);
1196   b->maxnz = b->nz = ui[mbs];
1197 
1198   B->info.factor_mallocs    = reallocs;
1199   B->info.fill_ratio_given  = fill;
1200   if (ai[mbs] != 0) {
1201     B->info.fill_ratio_needed = ((PetscReal)ui[mbs])/((PetscReal)ai[mbs]);
1202   } else {
1203     B->info.fill_ratio_needed = 0.0;
1204   }
1205   if (perm_identity){
1206     B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqBAIJ_N_NaturalOrdering;
1207   } else {
1208     B->ops->choleskyfactornumeric = MatCholeskyFactorNumeric_SeqBAIJ_N;
1209   }
1210   PetscFunctionReturn(0);
1211 }
1212 
1213 /* --------------------------------------------------------- */
1214 #undef __FUNCT__
1215 #define __FUNCT__ "MatSolve_SeqBAIJ_N_NaturalOrdering_newdatastruct"
1216 PetscErrorCode MatSolve_SeqBAIJ_N_NaturalOrdering_newdatastruct(Mat A,Vec bb,Vec xx)
1217 {
1218   Mat_SeqBAIJ    *a=(Mat_SeqBAIJ *)A->data;
1219   PetscErrorCode ierr;
1220   const PetscInt *ai=a->i,*aj=a->j,*vi;
1221   PetscInt       i,k,n=a->mbs;
1222   PetscInt       nz,bs=A->rmap->bs,bs2=a->bs2,*adiag=a->diag;
1223   MatScalar      *aa=a->a,*v;
1224   PetscScalar    *x,*b,*s,*t,*ls;
1225 
1226   PetscFunctionBegin;
1227   /* printf("MatSolve_SeqBAIJ_NaturalOrdering_iludt..bs %d\n",bs); */
1228   ierr = VecGetArray(bb,&b);CHKERRQ(ierr);
1229   ierr = VecGetArray(xx,&x);CHKERRQ(ierr);
1230   t  = a->solve_work;
1231 
1232   /* forward solve the lower triangular */
1233   ierr = PetscMemcpy(t,b,bs*sizeof(PetscScalar));CHKERRQ(ierr); /* copy 1st block of b to t */
1234 
1235   for (i=1; i<n; i++) {
1236     v   = aa + bs2*ai[i];
1237     vi  = aj + ai[i];
1238     nz = ai[i+1] - ai[i];
1239     s = t + bs*i;
1240     ierr = PetscMemcpy(s,b+bs*i,bs*sizeof(PetscScalar));CHKERRQ(ierr); /* copy i_th block of b to t */
1241     for(k=0;k<nz;k++){
1242       Kernel_v_gets_v_minus_A_times_w(bs,s,v,t+bs*vi[k]);
1243       v += bs2;
1244     }
1245   }
1246 
1247   /* backward solve the upper triangular */
1248   ls = a->solve_work + A->cmap->n;
1249   for (i=n-1; i>=0; i--){
1250     v  = aa + bs2*ai[2*n-i];
1251     vi = aj + ai[2*n-i];
1252     nz = ai[2*n-i +1] - ai[2*n-i]-1;
1253     ierr = PetscMemcpy(ls,t+i*bs,bs*sizeof(PetscScalar));CHKERRQ(ierr);
1254     for(k=0;k<nz;k++){
1255       Kernel_v_gets_v_minus_A_times_w(bs,ls,v,t+bs*vi[k]);
1256       v += bs2;
1257     }
1258     Kernel_w_gets_A_times_v(bs,ls,aa+bs2*adiag[i],t+i*bs); /* *inv(diagonal[i]) */
1259     ierr = PetscMemcpy(x+i*bs,t+i*bs,bs*sizeof(PetscScalar));CHKERRQ(ierr);
1260   }
1261 
1262   ierr = VecRestoreArray(bb,&b);CHKERRQ(ierr);
1263   ierr = VecRestoreArray(xx,&x);CHKERRQ(ierr);
1264   ierr = PetscLogFlops(2.0*(a->bs2)*(a->nz) - A->rmap->bs*A->cmap->n);CHKERRQ(ierr);
1265   PetscFunctionReturn(0);
1266 }
1267 
1268 #undef __FUNCT__
1269 #define __FUNCT__ "MatSolve_SeqBAIJ_N_newdatastruct"
1270 PetscErrorCode MatSolve_SeqBAIJ_N_newdatastruct(Mat A,Vec bb,Vec xx)
1271 {
1272   Mat_SeqBAIJ    *a=(Mat_SeqBAIJ *)A->data;
1273   IS             iscol=a->col,isrow=a->row;
1274   PetscErrorCode ierr;
1275   const PetscInt *r,*c,*rout,*cout,*ai=a->i,*aj=a->j,*vi;
1276   PetscInt       i,m,n=a->mbs;
1277   PetscInt       nz,bs=A->rmap->bs,bs2=a->bs2,k;
1278   MatScalar      *aa=a->a,*v;
1279   PetscScalar    *x,*b,*s,*t,*ls;
1280 
1281   PetscFunctionBegin;
1282   ierr = VecGetArray(bb,&b);CHKERRQ(ierr);
1283   ierr = VecGetArray(xx,&x);CHKERRQ(ierr);
1284   t  = a->solve_work;
1285 
1286   ierr = ISGetIndices(isrow,&rout);CHKERRQ(ierr); r = rout;
1287   ierr = ISGetIndices(iscol,&cout);CHKERRQ(ierr); c = cout;
1288 
1289   /* forward solve the lower triangular */
1290   ierr = PetscMemcpy(t,b+bs*r[0],bs*sizeof(PetscScalar));CHKERRQ(ierr);
1291   for (i=1; i<n; i++) {
1292     v   = aa + bs2*ai[i];
1293     vi  = aj + ai[i];
1294     nz = ai[i+1] - ai[i];
1295     s = t + bs*i;
1296     ierr = PetscMemcpy(s,b+bs*r[i],bs*sizeof(PetscScalar));CHKERRQ(ierr);
1297     for(m=0;m<nz;m++){
1298       Kernel_v_gets_v_minus_A_times_w(bs,s,v,t+bs*vi[m]);
1299       v += bs2;
1300     }
1301   }
1302 
1303   /* backward solve the upper triangular */
1304   ls = a->solve_work + A->cmap->n;
1305   for (i=n-1; i>=0; i--){
1306     k  = 2*n-i;
1307     v  = aa + bs2*ai[k];
1308     vi = aj + ai[k];
1309     nz = ai[k+1] - ai[k] - 1;
1310     ierr = PetscMemcpy(ls,t+i*bs,bs*sizeof(PetscScalar));CHKERRQ(ierr);
1311     for(m=0;m<nz;m++){
1312       Kernel_v_gets_v_minus_A_times_w(bs,ls,v,t+bs*vi[m]);
1313       v += bs2;
1314     }
1315     Kernel_w_gets_A_times_v(bs,ls,v,t+i*bs); /* *inv(diagonal[i]) */
1316     ierr = PetscMemcpy(x + bs*c[i],t+i*bs,bs*sizeof(PetscScalar));CHKERRQ(ierr);
1317   }
1318   ierr = ISRestoreIndices(isrow,&rout);CHKERRQ(ierr);
1319   ierr = ISRestoreIndices(iscol,&cout);CHKERRQ(ierr);
1320   ierr = VecRestoreArray(bb,&b);CHKERRQ(ierr);
1321   ierr = VecRestoreArray(xx,&x);CHKERRQ(ierr);
1322   ierr = PetscLogFlops(2.0*(a->bs2)*(a->nz) - A->rmap->bs*A->cmap->n);CHKERRQ(ierr);
1323   PetscFunctionReturn(0);
1324 }
1325 
1326 #undef __FUNCT__
1327 #define __FUNCT__ "BlockAbs_privat"
1328 PetscErrorCode BlockAbs_private(PetscInt nbs,PetscInt bs2,PetscScalar *blockarray,PetscReal *absarray)
1329 {
1330   PetscErrorCode     ierr;
1331   PetscInt           i,j;
1332   PetscFunctionBegin;
1333   ierr = PetscMemzero(absarray,(nbs+1)*sizeof(PetscReal));CHKERRQ(ierr);
1334   for (i=0; i<nbs; i++){
1335     for (j=0; j<bs2; j++){
1336       if (absarray[i] < PetscAbsScalar(blockarray[i*nbs+j])) absarray[i] = PetscAbsScalar(blockarray[i*nbs+j]);
1337     }
1338   }
1339   PetscFunctionReturn(0);
1340 }
1341 
1342 #undef __FUNCT__
1343 #define __FUNCT__ "MatILUDTFactor_SeqBAIJ"
1344 PetscErrorCode MatILUDTFactor_SeqBAIJ(Mat A,IS isrow,IS iscol,const MatFactorInfo *info,Mat *fact)
1345 {
1346   Mat                B = *fact;
1347   Mat_SeqBAIJ        *a=(Mat_SeqBAIJ*)A->data,*b;
1348   IS                 isicol;
1349   PetscErrorCode     ierr;
1350   const PetscInt     *r,*ic;
1351   PetscInt           i,mbs=a->mbs,bs=A->rmap->bs,bs2=a->bs2,*ai=a->i,*aj=a->j,*ajtmp,*adiag;
1352   PetscInt           *bi,*bj,*bdiag;
1353 
1354   PetscInt           row,nzi,nzi_bl,nzi_bu,*im,dtcount,nzi_al,nzi_au;
1355   PetscInt           nlnk,*lnk;
1356   PetscBT            lnkbt;
1357   PetscTruth         row_identity,icol_identity,both_identity;
1358   MatScalar          *aatmp,*pv,*batmp,*ba,*rtmp,*pc,*multiplier,*vtmp;
1359   const PetscInt     *ics;
1360   PetscInt           j,nz,*pj,*bjtmp,k,ncut,*jtmp;
1361 
1362   PetscReal          dt=info->dt; /* shift=info->shiftinblocks; */
1363   PetscInt           nnz_max;
1364   PetscTruth         missing;
1365   PetscReal          *vtmp_abs;
1366   MatScalar          *v_work;
1367   PetscInt           *v_pivots;
1368 
1369   PetscFunctionBegin;
1370   /* ------- symbolic factorization, can be reused ---------*/
1371   ierr = MatMissingDiagonal(A,&missing,&i);CHKERRQ(ierr);
1372   if (missing) SETERRQ1(PETSC_ERR_ARG_WRONGSTATE,"Matrix is missing diagonal entry %D",i);
1373   adiag=a->diag;
1374 
1375   ierr = ISInvertPermutation(iscol,PETSC_DECIDE,&isicol);CHKERRQ(ierr);
1376 
1377   /* bdiag is location of diagonal in factor */
1378   ierr = PetscMalloc((mbs+1)*sizeof(PetscInt),&bdiag);CHKERRQ(ierr);
1379 
1380   /* allocate row pointers bi */
1381   ierr = PetscMalloc((2*mbs+2)*sizeof(PetscInt),&bi);CHKERRQ(ierr);
1382 
1383   /* allocate bj and ba; max num of nonzero entries is (ai[n]+2*n*dtcount+2) */
1384   dtcount = (PetscInt)info->dtcount;
1385   if (dtcount > mbs-1) dtcount = mbs-1;
1386   nnz_max  = ai[mbs]+2*mbs*dtcount +2;
1387   /* printf("MatILUDTFactor_SeqBAIJ, bs %d, ai[mbs] %d, nnz_max  %d, dtcount %d\n",bs,ai[mbs],nnz_max,dtcount); */
1388   ierr = PetscMalloc(nnz_max*sizeof(PetscInt),&bj);CHKERRQ(ierr);
1389   nnz_max = nnz_max*bs2;
1390   ierr = PetscMalloc(nnz_max*sizeof(MatScalar),&ba);CHKERRQ(ierr);
1391 
1392   /* put together the new matrix */
1393   ierr = MatSeqBAIJSetPreallocation_SeqBAIJ(B,bs,MAT_SKIP_ALLOCATION,PETSC_NULL);CHKERRQ(ierr);
1394   ierr = PetscLogObjectParent(B,isicol);CHKERRQ(ierr);
1395   b    = (Mat_SeqBAIJ*)(B)->data;
1396   b->free_a       = PETSC_TRUE;
1397   b->free_ij      = PETSC_TRUE;
1398   b->singlemalloc = PETSC_FALSE;
1399   b->a          = ba;
1400   b->j          = bj;
1401   b->i          = bi;
1402   b->diag       = bdiag;
1403   b->ilen       = 0;
1404   b->imax       = 0;
1405   b->row        = isrow;
1406   b->col        = iscol;
1407   ierr          = PetscObjectReference((PetscObject)isrow);CHKERRQ(ierr);
1408   ierr          = PetscObjectReference((PetscObject)iscol);CHKERRQ(ierr);
1409   b->icol       = isicol;
1410   ierr = PetscMalloc((bs*(mbs+1))*sizeof(PetscScalar),&b->solve_work);CHKERRQ(ierr);
1411 
1412   ierr = PetscLogObjectMemory(B,nnz_max*(sizeof(PetscInt)+sizeof(MatScalar)));CHKERRQ(ierr);
1413   b->maxnz = nnz_max/bs2;
1414 
1415   (B)->factor                = MAT_FACTOR_ILUDT;
1416   (B)->info.factor_mallocs   = 0;
1417   (B)->info.fill_ratio_given = ((PetscReal)nnz_max)/((PetscReal)(ai[mbs]*bs2));
1418   CHKMEMQ;
1419   /* ------- end of symbolic factorization ---------*/
1420   ierr = ISGetIndices(isrow,&r);CHKERRQ(ierr);
1421   ierr = ISGetIndices(isicol,&ic);CHKERRQ(ierr);
1422   ics  = ic;
1423 
1424   /* linked list for storing column indices of the active row */
1425   nlnk = mbs + 1;
1426   ierr = PetscLLCreate(mbs,mbs,nlnk,lnk,lnkbt);CHKERRQ(ierr);
1427 
1428   /* im: used by PetscLLAddSortedLU(); jtmp: working array for column indices of active row */
1429   ierr = PetscMalloc((2*mbs+1)*sizeof(PetscInt),&im);CHKERRQ(ierr);
1430   jtmp = im + mbs;
1431   /* rtmp, vtmp: working arrays for sparse and contiguous row entries of active row */
1432   ierr = PetscMalloc((2*mbs*bs2+1)*sizeof(MatScalar),&rtmp);CHKERRQ(ierr);
1433   vtmp = rtmp + bs2*mbs;
1434   ierr = PetscMalloc((mbs+1)*sizeof(PetscReal),&vtmp_abs);CHKERRQ(ierr);
1435 
1436   ierr       = PetscMalloc(bs*sizeof(PetscInt) + (bs+bs2)*sizeof(MatScalar),&v_work);CHKERRQ(ierr);
1437   multiplier = v_work + bs;
1438   v_pivots   = (PetscInt*)(multiplier + bs2);
1439 
1440   bi[0]    = 0;
1441   bdiag[0] = (nnz_max/bs2)-1; /* location of diagonal in factor B */
1442   bi[2*mbs+1] = bdiag[0]+1; /* endof bj and ba array */
1443   for (i=0; i<mbs; i++) {
1444     /* copy initial fill into linked list */
1445     nzi = 0; /* nonzeros for active row i */
1446     nzi = ai[r[i]+1] - ai[r[i]];
1447     if (!nzi) SETERRQ2(PETSC_ERR_MAT_LU_ZRPVT,"Empty row in matrix: row in original ordering %D in permuted ordering %D",r[i],i);
1448     nzi_al = adiag[r[i]] - ai[r[i]];
1449     nzi_au = ai[r[i]+1] - adiag[r[i]] -1;
1450     /* printf("row %d, nzi_al/au %d %d\n",i,nzi_al,nzi_au); */
1451 
1452     /* load in initial unfactored row */
1453     ajtmp = aj + ai[r[i]];
1454     ierr = PetscLLAddPerm(nzi,ajtmp,ic,mbs,nlnk,lnk,lnkbt);CHKERRQ(ierr);
1455     ierr = PetscMemzero(rtmp,(mbs*bs2+1)*sizeof(PetscScalar));CHKERRQ(ierr);
1456     aatmp = a->a + bs2*ai[r[i]];
1457     for (j=0; j<nzi; j++) {
1458       ierr = PetscMemcpy(rtmp+bs2*ic[ajtmp[j]],aatmp+bs2*j,bs2*sizeof(MatScalar));CHKERRQ(ierr);
1459     }
1460 
1461     /* add pivot rows into linked list */
1462     row = lnk[mbs];
1463     while (row < i) {
1464       nzi_bl = bi[row+1] - bi[row] + 1;
1465       bjtmp = bj + bdiag[row+1]+1; /* points to 1st column next to the diagonal in U */
1466       ierr  = PetscLLAddSortedLU(bjtmp,row,nlnk,lnk,lnkbt,i,nzi_bl,im);CHKERRQ(ierr);
1467       nzi  += nlnk;
1468       row   = lnk[row];
1469     }
1470 
1471     /* copy data from lnk into jtmp, then initialize lnk */
1472     ierr = PetscLLClean(mbs,mbs,nzi,lnk,jtmp,lnkbt);CHKERRQ(ierr);
1473 
1474     /* numerical factorization */
1475     bjtmp = jtmp;
1476     row   = *bjtmp++; /* 1st pivot row */
1477 
1478     while  (row < i) {
1479       pc = rtmp + bs2*row;
1480       pv = ba + bs2*bdiag[row]; /* inv(diag) of the pivot row */
1481       Kernel_A_gets_A_times_B(bs,pc,pv,multiplier); /* pc= multiplier = pc*inv(diag[row]) */
1482       ierr = BlockAbs_private(1,bs2,pc,vtmp_abs);CHKERRQ(ierr);
1483       if (vtmp_abs[0] > dt){ /* apply tolerance dropping rule */
1484         pj         = bj + bdiag[row+1] + 1; /* point to 1st entry of U(row,:) */
1485         pv         = ba + bs2*(bdiag[row+1] + 1);
1486         nz         = bdiag[row] - bdiag[row+1] - 1; /* num of entries in U(row,:), excluding diagonal */
1487         for (j=0; j<nz; j++){
1488           Kernel_A_gets_A_minus_B_times_C(bs,rtmp+bs2*pj[j],pc,pv+bs2*j);
1489         }
1490         /* ierr = PetscLogFlops(bslog*(nz+1.0)-bs);CHKERRQ(ierr); */
1491       }
1492       row = *bjtmp++;
1493     }
1494 
1495     /* copy sparse rtmp into contiguous vtmp; separate L and U part */
1496     nzi_bl = 0; j = 0;
1497     while (jtmp[j] < i){ /* L-part. Note: jtmp is sorted */
1498       ierr = PetscMemcpy(vtmp+bs2*j,rtmp+bs2*jtmp[j],bs2*sizeof(MatScalar));CHKERRQ(ierr);
1499       nzi_bl++; j++;
1500     }
1501     nzi_bu = nzi - nzi_bl -1;
1502     /* printf("nzi %d, nzi_bl %d, nzi_bu %d\n",nzi,nzi_bl,nzi_bu); */
1503 
1504     while (j < nzi){ /* U-part */
1505       ierr = PetscMemcpy(vtmp+bs2*j,rtmp+bs2*jtmp[j],bs2*sizeof(MatScalar));CHKERRQ(ierr);
1506       /*
1507       printf(" col %d: ",jtmp[j]);
1508       for (j1=0; j1<bs2; j1++) printf(" %g",*(vtmp+bs2*j+j1));
1509       printf(" \n");
1510       */
1511       j++;
1512     }
1513 
1514     ierr = BlockAbs_private(nzi,bs2,vtmp,vtmp_abs);CHKERRQ(ierr);
1515     /*
1516     printf(" row %d, nzi %d, vtmp_abs\n",i,nzi);
1517     for (j1=0; j1<nzi; j1++) printf(" (%d %g),",jtmp[j1],vtmp_abs[j1]);
1518     printf(" \n");
1519     */
1520     bjtmp = bj + bi[i];
1521     batmp = ba + bs2*bi[i];
1522     /* apply level dropping rule to L part */
1523     ncut = nzi_al + dtcount;
1524     if (ncut < nzi_bl){
1525       ierr = PetscSortSplitReal(ncut,nzi_bl,vtmp_abs,jtmp);CHKERRQ(ierr);
1526       ierr = PetscSortIntWithScalarArray(ncut,jtmp,vtmp);CHKERRQ(ierr);
1527     } else {
1528       ncut = nzi_bl;
1529     }
1530     for (j=0; j<ncut; j++){
1531       bjtmp[j] = jtmp[j];
1532       ierr = PetscMemcpy(batmp+bs2*j,rtmp+bs2*bjtmp[j],bs2*sizeof(MatScalar));CHKERRQ(ierr);
1533       /*
1534       printf(" col %d: ",bjtmp[j]);
1535       for (j1=0; j1<bs2; j1++) printf(" %g,",*(batmp+bs2*j+j1));
1536       printf("\n");
1537       */
1538     }
1539     bi[i+1] = bi[i] + ncut;
1540     nzi = ncut + 1;
1541 
1542     /* apply level dropping rule to U part */
1543     ncut = nzi_au + dtcount;
1544     if (ncut < nzi_bu){
1545       ierr = PetscSortSplitReal(ncut,nzi_bu,vtmp_abs+nzi_bl+1,jtmp+nzi_bl+1);CHKERRQ(ierr);
1546       ierr = PetscSortIntWithScalarArray(ncut,jtmp+nzi_bl+1,vtmp+nzi_bl+1);CHKERRQ(ierr);
1547     } else {
1548       ncut = nzi_bu;
1549     }
1550     nzi += ncut;
1551 
1552     /* mark bdiagonal */
1553     bdiag[i+1]    = bdiag[i] - (ncut + 1);
1554     bi[2*mbs - i] = bi[2*mbs - i +1] - (ncut + 1);
1555 
1556     bjtmp = bj + bdiag[i];
1557     batmp = ba + bs2*bdiag[i];
1558     ierr = PetscMemcpy(batmp,rtmp+bs2*i,bs2*sizeof(MatScalar));CHKERRQ(ierr);
1559     *bjtmp = i;
1560     /*
1561     printf(" diag %d: ",*bjtmp);
1562     for (j=0; j<bs2; j++){
1563       printf(" %g,",batmp[j]);
1564     }
1565     printf("\n");
1566     */
1567     bjtmp = bj + bdiag[i+1]+1;
1568     batmp = ba + (bdiag[i+1]+1)*bs2;
1569 
1570     for (k=0; k<ncut; k++){
1571       bjtmp[k] = jtmp[nzi_bl+1+k];
1572       ierr = PetscMemcpy(batmp+bs2*k,rtmp+bs2*bjtmp[k],bs2*sizeof(MatScalar));CHKERRQ(ierr);
1573       /*
1574       printf(" col %d:",bjtmp[k]);
1575       for (j1=0; j1<bs2; j1++) printf(" %g,",*(batmp+bs2*k+j1));
1576       printf("\n");
1577       */
1578     }
1579 
1580     im[i] = nzi; /* used by PetscLLAddSortedLU() */
1581 
1582     /* invert diagonal block for simplier triangular solves - add shift??? */
1583     batmp = ba + bs2*bdiag[i];
1584     ierr = Kernel_A_gets_inverse_A(bs,batmp,v_pivots,v_work);CHKERRQ(ierr);
1585   } /* for (i=0; i<mbs; i++) */
1586 
1587   /* printf("end of L %d, beginning of U %d\n",bi[mbs],bdiag[mbs]); */
1588   if (bi[mbs] >= bdiag[mbs]) SETERRQ2(PETSC_ERR_ARG_SIZ,"end of L array %d cannot >= the beginning of U array %d",bi[mbs],bdiag[mbs]);
1589 
1590   ierr = ISRestoreIndices(isrow,&r);CHKERRQ(ierr);
1591   ierr = ISRestoreIndices(isicol,&ic);CHKERRQ(ierr);
1592 
1593   ierr = PetscLLDestroy(lnk,lnkbt);CHKERRQ(ierr);
1594 
1595   ierr = PetscFree(im);CHKERRQ(ierr);
1596   ierr = PetscFree(rtmp);CHKERRQ(ierr);
1597 
1598   ierr = PetscLogFlops(bs2*B->cmap->n);CHKERRQ(ierr);
1599   b->maxnz = b->nz = bi[mbs] + bdiag[0] - bdiag[mbs];
1600 
1601   ierr = ISIdentity(isrow,&row_identity);CHKERRQ(ierr);
1602   ierr = ISIdentity(isicol,&icol_identity);CHKERRQ(ierr);
1603   both_identity = (PetscTruth) (row_identity && icol_identity);
1604   if (row_identity && icol_identity) {
1605     B->ops->solve = MatSolve_SeqBAIJ_N_NaturalOrdering_newdatastruct;
1606   } else {
1607     B->ops->solve = MatSolve_SeqBAIJ_N_newdatastruct;
1608   }
1609 
1610   B->ops->solveadd          = 0;
1611   B->ops->solvetranspose    = 0;
1612   B->ops->solvetransposeadd = 0;
1613   B->ops->matsolve          = 0;
1614   B->assembled              = PETSC_TRUE;
1615   B->preallocated           = PETSC_TRUE;
1616   PetscFunctionReturn(0);
1617 }
1618